Combustion Science and Engineering Series
Series Editor Forman A. Williams Princeton University
Advising Editors Ronald Hanson, Stanford University David T. Pratt, University of Washington Daniel J. Seery, United Technologies
To Th. von Karman, the founder of aerothermochemistry
Combustion Theory
The Fundamental Theory of Chemically Reacting Flow Systems Second Edition
Forman A Williams
Princeton University
The Benjamin/Cummings Publishing Company, Inc. Menlo Park, California Reading, Massachusetts Don Mills, Ontario Wokingham, U.K. Amsterdam Sydney Singapore Tokyo Mexico City Bogota Santiago San Juan
Sponsoring Editor: Richard W. Mixter Production Coordinator : Kristina Montague Copyeditor : Linda Thompson Cover Design : Leigh McLellan
Copyright @ 1985 by The Benjamin/Cummings Publishing Company, Inc. All rights reserved. N o part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the publisher. Printed in the United States of America. Published simultaneously in Canada.
Library of Congress Cataloging in Publication Data
Williams, F. A. (Forman Arthur), 1934Combustion theory. (Combustion, science, and engineering series) Bibliography: p. Includes index. 1. Combustion. I . Title. 11. Series. QD516.W5 1985 541.3’61 853902 ISBN 0805398015 ABCDEFGHIJMA8 98765 The Benjamin/Cummings Publishing Company, Inc 2727 Sand Hill Road Menlo Park, California 94025
Preface to the Second Edition
Truly remarkable progress has been achieved in combustion theory during the 20 years since the completion of the first edition of this book. Because of this progress, when the first edition went out of print about 5 years ago, it was clear that a reprinting could no longer serve to bring readers to the frontiers of the subject. Revision of the entire text was needed, and I undertook the revision without realizing that it would require as much time as writing the first edition. Although it is regrettable that the book has been unavailable during the preparation of the second edition, it seemed best to try to complete as good a revision as possible instead of rushing into print. Thus consideration has been given to changes throughout ; none of the chapters survived without modification, and outdated chapters have disappeared entirely and have been replaced by chapters devoted to new material of importance in the subject. Nevertheless, the general format of the book and the philosophy of the presentation remain the same. The necessary background material still appears in the same five appendixes. These appendixes needed less revision than the rest of the volume; a few of the sections survived intact. Among the specific changes in the appendixes are updating of the discussion of thermochemical equilibrium calculations to take account of the revolutionary advances in computational power of electronic computers, presentation of clarifications that have been achieved in the concepts of steadystate and partialequilibrium approximations in chemical kinetics, amplification of the discussions of branchedchain and thermal explosions on the basis of insights developed from ideas of asymptotic methods, presentation of newly available information on the chemical kinetics of hydrocarbon oxidation and of the production of oxides of nitrogen in flames, and augmentation of the treatment of transport processes to include additional aspects of radiant energy transfer. The first chapter, in which the basic mathematical formulations are presented, remains just as formidable to the fainthearted as it was in the first edition. There are still 12 chapters, but the later ones, especially, exhibit modifications in coverage as a consequence of the progress made in the field. The titles of the first six chapters remain essentially the same as in the first edition. Chapter 7 of the first edition, which was devoted mainly to a discussion of experiments on turbulent flames, has been deleted and replaced by a considerably more extensive chapter (Chapter 10) that develops turbulentflame theory from first principles; noteworthy advances in the theory during the past 20 years have made this possible. Presentation of material on turbulent combustion, formerly in Chapter 7, is postponed until Chapter 10 because the new knowledge relies in part on various results that have been obtained in studies of laminarflame instabilities (now
V
vi
Preface to the Second Edition
covered in Chapter 9j and in investigations of ignition and extinction (included in Chapter 8j. Monopropellant droplet burning, which comprised the subject of Chapter 10 of the first edition, has assumed a somewhat less prominent role in combustion in the intervening years and therefore is covered only briefly now, specifically at the end of Chapter 3, the chapter that includes detailed discussions of the burning of fuel droplets. Chapter 9 of the first edition addressed both the theory of solidpropellant deflagration and theories of combustion instability; such substantial advances have been made in these subjects that two chapters are now needed to approach the material properly, Chapter 7 on combustion of solid propellants and Chapter 9 on combustion instabilities. The general topic of Chapter 8, ignition, extinction, and flammability limits, is the same as in the first edition, as is the subject of Chapter 11, spray combustion. Also, the general thrust of Chapter 12 remains approximately the same, notwithstanding the title change, from “Chemical Reactions in Boundary Layers” to “Flame Attachment and Flame Spread,” introduced to place emphasis on specific areas of application. To delineate more thoroughly the modifications that have been introduced, it seems desirable to discuss briefly the content chapter by chapter. The basic formulations in Chapter 1 and the derivation in Chapter 2 of jump conditions across combustion waves are largely unchanged from the first edition. Therefore, the following discussion focuses on the remaining chapters. Chapter 3, on diffusion flames and droplet burning, has been expanded considerably to include, for example, greater consideration of influences of buoyancy and turbulence on jet flames, discussion of kinetics of surface oxidation of carbon, consideration of combustion mechanisms of metal particles, and reference to treatment of a number of realities of droplet burning that were not addressed in the first edition, such as influences of radiant energy transfer, the occurrence of unsteady conditions in the gaseous and condensed phases, the role of momentum conservation, influences of variable properties and of Lewis numbers differing from unity, the selection of suitable reference states for the evaluation of effects of forced and natural convection, and identification of conditions for the occurrence of disruptive burning of droplets. In addition, sections on diffusionflame structure and on the burning of monopropellant droplets have been added to Chapter 3 , since these subjects, in which significant new progress in understanding has been made, fall within the realm of the title of the chapter; this more advanced material, which introduces the mixture fraction as an important variable and which makes use of asymptotic analysis for describing aspects such as flamesheet broadening, ignition, and extinction, pedagogically should be covered only after study of Chapter 5 , where the necessary background ideas are presented.
Preface to the Second Edition
vii
The material in Chapter 4, on finiterate chemistry in flows with negligible molecular transport, has been amplified to a lesser extent beyond the presentation in the first edition. The discussions of ignition delay and of the wellstirred reactor have been expanded somewhat, and a treatment of twophase nozzle flow has been added, while elsewhere in the chapter modifications in the writing have been restricted to updating of ideas. The chapter on laminarflame theory, where molecular transport and finiterate chemistry first enter simultaneously, has been revised substantially in the light of advances in our conceptual foundations of the subject. Asymptotic methods in mathematical analysis, notably the socalled activationenergy asymptotics, have provided us with a greatly improved understanding of the elements of laminarflame propagation. Most of the ad hoc approaches that were detailed in the first edition are deleted and replaced by a systematic development of activationenergy asymptotics. The results obtained thereby are employed to explain many previously perplexing properties of laminarflame structure and fame speeds. In addition, new consideration is given to flames with multiplestep chemistry and to distributions of radicals in flames, on the basis of both asymptotic concepts and advances in computational abilities of electronic computers. Because of rapid progress in detonation theory in the 1960s, Chapter 6 of the first edition on detonations was obsolete in certain respects even before it appeared in print. In contrast, many of the revolutionary advances in this subject seem now to have been brought to fruition and appraised in reviews. Therefore, the extensive revision of this chapter needed for preparation of the second edition, appears likely to remain adequate for a longer period of time. The major changes and additions begin with the discussion of effects of threedimensional structures in detonation propagation. Consideration is given to the relevance of spin, to the mechanism of instability, to tripleshock interactions and to transverse spacings. In addition, new analyses of detonability limits and of failure of detonations are included, and criteria for direct initiation of detonations are presented along with a discussion of processes of transition from deflagration to detonation. Chapter 7, on solidpropellant combustion, focuses mostly on the deflagration of homogeneous solids. Consideration is given to deflagrations controlled by interface, condensedphase, gasphase, and dispersedphase reactions, now with utilization of concepts from asymptotics. Specific examples mentioned include the combustion of nitrocellulose and of ammonium perchlorate; attention is paid to both pressure and temperature sensitivities of burning rates of propellants. Appraisals are made of the current status of theories of combustion of heterogeneous propellants, and the combustion of black powder is addressed as an especially challenging theoretical problem involving the oldest propellant known to mankind. In
Vlll
...
Preface to the Second Edition
addition, a discussion of erosive burning is given, including the equations for steady, frictionless, adiabatic flow in a constantarea channel with mass addition. The presentation in Chapter 8 of the subject of ignition, extinction and flammability limits is a revision in nearly all respects. The new approach to this subject relies strongly on activationenergy asymptotics and thereby, I believe, achieves both enhanced understanding and improved accuracy in the results. Topics that were not present in the first edition include a discussion of physical aspects of spark ignition and an introduction to the use of activationenergy asymptotics in more complicated ignition problems, through the vehicle of the problem of ignition of a semiinfinite reactive solid by a constant energy flux applied to its surface. Chapter 9 represents an effort to provide a unified and tutorial presentation of the broad field of the theory of combustion instabilities. The length of the chapter attests to the vastness of the field and to the progress that has been made therein in recent years. The final section of this chapter, on the theory of instabilities of premixed flames, is basic to analyses of premixed turbulent flame propagation and also has a bearing on aspects of flammability limits. Chapter 10 on turbulentflame theory also is long, as it must be because so many different viewpoints and approaches to the subject now are available. Included in this chapter are discussions of analyses of effects of strain on laminar flame sheets, topics of interest in themselves as well as in connection with turbulent combustion. Evolution equations for laminar flames in nonuniform flows also are given. The results outlined for turbulent burning velocities emphasize those aspects that have the strongest basic theoretical justifications. Since turbulentflame theory is a subject of continuing development, improvements of results presented herein might be anticipated to be available in the nottoodistant future. The presentation of the subject of spray combustion in Chapter 11 is not greatly different from that in the first edition. An updated outlook on the subject has been provided, and the formulation has been generalized to admit time dependences in the conservation equations. The analysis of spray deflagration has been abbreviated, and qualitative aspects of the results therefrom have been anticipated on the basis of simplified physical reasoning. In addition, brief discussions of the topics of spray penetration and of cloud combustion have been added. The final chapter again is devoted largely to the theory of chemical reactions in boundarylayer flows. A general formulation of timedependent, axisymmetric, boundarylayer equations has been added to provide the reader with a convenient point of departure for analyzing a variety of reacting boundarylayer flows. The problem of combustion of a fuel plate in an oxidizing stream then provides an illustration of the use of the general formulation, instead of standing independently. The analysis of the problem
Preface to the Second Edition
ix
of premixed flame development in a mixing layer now is abbreviated and approached as part of the broader subject of mechanisms of flame stabilization by solid bodies and by fluid streams. The final section, addressing problems of flame spread through solid and liquid fuels from a simplified viewpoint, is new. The present volume inevitably is appreciably longer than the first edition. The increase in length is due mainly to the increase in the breadth of the coverage. Details of the mathematical development have not been amplified; on the contrary, in a number of respects fewer intermediate steps have been included, so that on the average the reader will have to invest more time per page to extract the same depth of understanding. I think that this is an unavoidable consequence of the growth of combustion theory. Even the first edition contained too much material to be covered properly in a single course. At Princeton University portions of the contents have been used in a number of graduate courses. By the time students reach the last of these, combustion theory, they have been exposed to the subjects of the appendices, of Chapters 1,2, and 4, and of most of Chapter 12, and they have studied the more qualitative aspects of Chapters 3 and 5; the students are then able to assimilate most of the rest of the book in one semester. Alternative orders of coverage would fit into differently designed curricula. It seems fair to say that at most universities courses do not cover all of the material in the text; in this respect, the presentation may serve as a reference for researchers specializing in the field. In the period since the appearance of the first edition, a number of related books on combustion have been published. Many of these are referenced throughout the new text. They are quite helpful in providing alternative viewpoints toward much of the material. The progress in the field has resulted in significantly more extensive citations to references in the second edition, even though the bibliographical information certainly is not complete. Where I have forgotten to cite an important reference, I hope that the author will recognize the magnitude of my task and accordingly will forgive my oversight. In an effort to achieve some economy in printing, three oftcited sources are abbreviated as follows: C & F for Combustion and Flame, C S T for Combustion Science and Technology, and n’th Symp. for the currently biennial international symposia on combustion, the full citations of which are:
Proceedings o f the First Symposium on Combustion and the Second Symposium on Combustion. The Combustion Institute, Pittsburg (19653, reprinted from Industrial and Engineering Chemistry (19283 and from Chemical Reviews (19373, Third Symposium on Combustion and Flame and Explosion Phenomena, Williams and Wilkins Co., Baltimore (1949),
X
Preface to the Second Edition
Fourth Symposium (International) on Combustion, Williams and Wilkins Co., Baltimore (1953), Fifth Symposium (International) on Combustion, Rienhold Publishing Corp., New York (1955), Sixth . . . , Reinhold Publishing Corp., New York (1957), Seventh.. . , Butterworths Scientific Publications, London (1959), Eighth . . . , Williams and Wilkins Co., Baltimore (1962), Ninth . . . , Academic Press, New York (1963), T e n t h . . . , The Combustion Institute, Pittsburgh (1965), Eleoenth . . . (1967), Twelfth . . . (1969), Thirteenth . . . (1971), Fourteenth . . . (1973), Fijteenth . . . (1975), Sixteenth . . . (1977), Seventeenth . . . (1979), Eighteenth . . . (1981), Nineteenth . . . (1982).
I am indebted to many people for aid in preparation of the second edition. The colleagues and scientists who contributed significant discussions of technical material are too numerous to name, for if I were to attempt to acknowledge them all, I would surely slight someone by inadvertently omitting his or her name from the list. Nevertheless, I cannot refrain from stating here that Amable Liiian has done more than any other individual in improving my understanding of combustion theory. Even with respect to specific helpful recommendations for changes in the text, I do not dare to try to acknowledge individuals, although I would like to say that observations made by the team of Japanese combustion scientists, who translated the book from the typewritten manuscript as I completed the writing, proved to be notably useful in prompting clarifications in the Englishlanguage presentation. Nearly all of the original manuscript was typed by Peggy Berger, for whose patient help over a 5year period I am greatly indebted. Finally, I wish to express my gratitude for the presence of mind encouraged by Elizabeth. Forman A. Williams Princeton, New Jersey June 1984
Preface to the First Edition
This book contains a systematic development of the theory of chemical reactions in fluid flow. It is intended to serve either as an advanced graduate text or as a reference for workers in the field of combustion. An attempt has been made to cover the entire field, not just those parts of the field in which the author has done research. However, the book does emphasize the theoretical aspects of the subject and discusses experiments only where absolutely necessary. This type of approach appears to be justified in view of the existence of a number of excellent treatises on the experimental aspects of combustion, such as B. Lewis and G. von Elbe’s classical work Combustion, Flames and Explosions of Gases (Academic Press, New York, 1st ed., 1951,2nd ed. 1961), the book by A. G. Gaydon and H. G. Wolfhard on Flames, Their Structure, Radiation and Temperature, (Chapman and Hall, Ltd., London, 1960), and W. Jost, Explosion and Combustion Processes in Gases (McGrawHill, New York, 1946). A familiarity with the material in these books will provide the reader with a proper perspective for beginning a study of combustion theory. Relevant experimental work may also be found in the mammoth biennial volumes containing the proceedings of the International Symposia on Combustion, sponsored by the Combustion Institute, and in the journal Combustion and Flame. The arrangement and the method of attack of the book require some explanation. In order to proceed with the development of the principal subject material as directly as possible, from the outset I assume that the reader has the scientific background necessary for understanding combustion theory. Thus, a knowledge of mathematics (principally, a thorough understanding of ordinary and partial differential equations), thermodynamics, statistical mechanics, chemical kinetics, and molecular transport theory is presupposed. To aid the reader who does not have a complete command of all of these subjects, and also as convenient places from which to draw specific background items used in the text, extended appendices are provided containing reviews on thermodynamics and statistical mechanics, on chemical kinetics, on the fundamental equations of fluid dynamics, and on transport properties. For courses in which students lack specific background topics, the appropriate appendices would be the best place to begin studying the book. However, some knowledge of elementary gas dynamics and thermodynamics and a thorough knowledge of applied mathematics are needed in any case, because no mathematical appendices are given, and topics such as heat engine cycle analysis, the meaning of Reynolds and Mach numbers, and the concept of turbulence are omitted. Although pedagogical considerations did go into the arrangement of the material, the subject matter itself dictates more of a whathasbeendone book than a howtodo book. I hope that by solving some problems and by pointing out related problems that have and have not been solved, 1
XI
xii
Preface to the First Edition
will be able to indicate fruitful lines of research and to suggest useful methods of solution to the reader. Although no homework problems are provided explicitly, statements such as “it can be shown” and “the reader may prove” appear often in the text and can be treated as problems for the student. My style of presentation is generally concisely deductive rather than discursively inductive. Perhaps, by stripping the presentation to the barest essentials and by removing adornments commonly incorporated for pedagogical reasons, I may clarify some possible points of confusion for a few readers. The writing of this treatise was undertaken through the encouragement of Dr. S. S. Penner. Practically all of the book was completed while I was an Assistant Professor at Harvard University holding a National Science Foundation Grant for Fundamental Studies of Sprays. During this time I spent six months at Imperial College, University of London, on a National Science Foundation Postdoctoral Fellowship and devoted most of my time there to the writing. Thanks are due to many of my colleagues with whom I have talked about the book. Deserving of special mention are Dr. William Nachbar and Professor Howard W. Emmons for informative conversations. Professor D. B. Spalding contributed many helpful suggestions and kindly read the first draft of the manuscript, recommending many important changes. But, most of all, I wish to thank my teacher, Dr. S. S . Penner, who wrote the original drafts of a few of the chapters and provided constant encouragement and help thereafter. My family also deserves thanks for their patience. F. A . Williams La Jolla, California June 1964
Contents
Preface to the Second Edition Preface to the First Edition
V
xi
1
1 SUMMARY OF RELEVANT ASPECTS OF FLUID DYNAMICS AND CHEMICAL KINETICS
1.I . The conservation equations for multicomponent, reacting, idealgas mixtures 1.2. Onedimensional flow 1.2.1. Unsteady flow 1.2.2. Steady flow 1.3. Coupling functions 1.4. Conservation conditions at an interface 1.5. Discussion of the approach adopted in the following development of combustion theory References
2 4 4 7
9
13 17 18 19 20 20 22 23 24 24 26 26 28 30 30 33 33 33 34 35 37
XI11
2 RANKINEHUGONIOT RELATIONS
2.1. General RankineHugoniot equations 2.1.1. Derivation of the equations 2.1.2. The coldboundary difficulty 2.1.3. Use of the RankineHugoniot equations 2.2. Analysis of a simplified system 2.2. I . Simplification of the RankineHugoniot equations 2.2.2. Dimensionless form 2.2.3. Properties of the Hugoniot curve 2.2.4. Analysis of the detonation branch 2.2.5. Analysis of the deflagration branch 2.2.6. Properties of ChapmanJouguet waves 2.3. Extension of the results to arbitrary systems 2.3.1. Range of validity of the results of Section 2.2 2.3.2. Frozen versus equilibrium sound speeds 2.3.3. Proof that u, = u,,, co at the ChapmanJouguet points 2.3.4. Summary of the properties of Hugoniot curves References
...
Discussion of the burningrate formula 3. The flame shape and the flame height 3. The validity of the other approximations 3.1. Analysis 3. quasionedimensional flow 4.2. The mixturefraction variable 3. Assumptions 3.3.5.3.3.4. quasionedimensional conservation equations 4.1.xiv Contents 3 DIFFUSION FLAMES AND DROPLET BURNING 3.2.5.4.1.1.4.2.6. Ignition and extinction 3. The nature of carbon combustion 3.5.1. Analysis predicting the burning rate 3.4.2. Ignition delay and the wellstirred reactor 4.4. Assumptions 3.1.2.1.2. Reactions in steady. Freezing of reactions 4.3.1. The burning of a fuel particle in an oxidizing atmosphere 3.2.7.6.2. The flame at the mouth of a tube in a duct 3.1.2. Approach to structure questions 3. The oxidation of carbon at the walls of a duct 3.2.3.3.3. Activationenergy asymptotics 3. Definition of the problem 3.2. Application to the reaction A P B with species A and B present in only trace amounts 4. Definition of the problem 3.6.4. Structure of the flame sheet 3. Solution of the species conservation equation for the coupling function p 3.3. Twophase nozzle flow .1. Nearequilibrium and nearfrozen flows 4. The flame sheet approximation 3.1. Further realities of droplet burning 3.2. Comments on the formulation and the analysis 3.2.1.1. Steadystate.4.3.3.2.1.4.2.4.2. Specific impulse of rockets 4. Predictions of other characteristics of burning droplets 3.3. Background and definition of the problem 3.5.3. Monopropellant droplet burning References 4 REACTIONS IN FLOWS WITH NEGLIGIBLE MOLECULAR TRANSPORT 38 39 39 40 41 42 44 45 46 48 48 48 50 52 52 54 56 59 61 62 69 69 73 76 80 84 86 92 92 96 96 99 100 103 105 106 4.
1. Solution of the diffusion equation when Le = 1 5. and von Karman 5.3. Preliminary relationships 4. FrankKamenetskii.1. Summary of the simplified mathematical problem 5.3.1.3.3.6.3. characteristic surfaces 4.6.3.2. The coldboundary difficulty 5.2.2. unsteady flows 4.3.2. Introductory remarks 5. Onedimensional.3. Preliminary assumptions and equations 5.2. reacting idealgas mixture for the reaction A P B 4.4.2.2. twodimensional (axially symmetrical and plane) flows 4.3.3. unsteady sound propagation in a binary. Generalizations to other flames .5. Dispersion relations 4.2.4. Reactions in unsteady.1. threedimensional flow 4. The method of characteristics for steady. Mathematical formulation 5. Approximations that further simplify the energy equation 5. Conservation equations .3.3.2.3.1.3. Linearization 4.1. An initialvalue problem 4. Bounds on the burningrate eigenvalue 5.4. Asymptotic analysis for strongly temperaturedependent rates 5.2.4. Related problems References 5 THEORY OF LAMINAR FLAMES 108 108 113 118 119 119 120 120 123 124 126 127 130 131 131 135 136 136 137 138 139 141 141 142 143 143 145 149 151 153 154 160 5.3.2. Reduction to a single partial diferential equation 4.4.3.3.3.3.4.4.7. Simplifications in the energy and diffusion equations for unimolecular reactions in binary mixtures 5.2. Iterative procedures and variational methods 5.4.1. Governing equations 5. The unimolecular decomposition flame with Lewis number of unity 5.5.Contents xv 4.6.3. Experiments 5.4. The method of characteristics for onedimensional.4.2.7.3. The approximations of Zel’dovich. Description of laminar flames 5.5.3. Phenomenological analysis of a deflagration wave 5.1.3. Dimensionless forms for the momentum equation and the species conservation equation 5.
4.1 .4.3.2. Discussion of the background and the nature of the problem 5.3. Further comments on weak detonations 6. Chemical reactions behind a shock wave produced in a shock tube 182 183 183 183 184 187 188 190 191 192 194 197 197 199 199 20 1 203 204 . Planar detonation structure 6.3.4.1. Effects of tube walls 6. A criterion for the applicability of the extended steadystate approximation 5. Methods of analysis for testing steadystate approximations 5.1. Remarks on deflagrations 6. Formulation 5. General properties of the integral curves 6.1.4.1.3. The structure of ZND detonations 6.5.1.4.6.2. Discussion of detonation structure 6.2.6.4. Conservation equations for reaction intermediaries 5.1.4.2.2.2.1. Properties of the governing equations 6.5. Basic considerations for planar waves 6. Propagation velocities for detonations traveling in tubes 6. Observations on theories of flame structure References 165 165 165 167 167 170 172 173 174 175 177 179 6 DETONATION PHENOMENA 6.4. Location o f singularities 6.2.1.2. Ambiguities associated with frozen and equilibrium sound speeds 6.1. Governing equations 6. Objectives of’ analyses 5.1.2.4. The extended steadystate approximation 5. Burningvelocity calculations 5.2.2.xvi Contents 5.4. Solutions in the neighborhoods of’ singular points 6.2. Effects of threedimensional structures 6.6.3. Literature 5.1.1.1.1.4.1.2. Approximate solution for the structure of a detonation 6.2.1.4.4.4.5.4. I .2.1. Flames with multiplestep chemistry 5.
Standing detonations 6.4. Combustion of heterogeneous propellants 7. Methods of analysis 8.2.1. Detonations in solids.3. Erosive burning References 8 IGNITION. Theories of transverse structures 6.5. and sprays References 204 204 208 212 212 217 219 22 1 229 230 232 235 238 243 249 25 1 258 26 1 265 268 27 1 27 1 277 277 279 279 284 29 1 294 295 295 298 7 COMBUSTION OF SOLID PROPELLANTS 7. Conservation of acoustic energy .4. Minimum ignition energies and quenching distances 8.8.6.3.2. Transverse structures for detonations 6.1.5.1. Approach to interfacial equilibrium 7.3.1.5.2.3. The existence of two flame speeds 8.2.3. The transition from deflagration to detonation 6. AND FLAMMABILITY LIMITS 8.4.3. Description of steady deflagration of a homogeneous solid 7.1.2.2. Pressure limits of flammability 8. Deflagration controlled by condensedphase reaction rates 7. liquids. EXTINCTION.2.1.7.3. Premixed flames with heat losses 8. Deflagration controlled by gasphase reaction rates 7.2. Spinning detonations and stability considerations 6.3. Detonability limits and quenching thickness 6. Dispersion phenomena and other influences 7.4. Concentration limits of flammability 8. Activationenergy asymptotics in ignition theory References 9 COMBUSTION INSTABILITIES 9. Acoustic instabilities in solidpropellant rocket motors 9.2. Applications of transitionstate theory 7.Contents xvii 6. Oscillation modes 9.2. Estimates of heat loss 8.1.1.3.3.
1.1.2. Production of trace species 10.3. The acoustic admittance 9. Properties of probabilitydensity functions 10. Relative importance 9.2.4.5.7. Nozzle damping 9. Probabilistic descriptions 10.1. Formulation through asymptotic methods 9.3.4.3. Scales of turbulence 10.2.1.4.5.2.1 3 . Turbulent diffusion flames 10. Heterogeneity efects 9. Inherent oscillations of burning solids 9.1. Homogeneous damping 9. Amplification mechanisms 9. Probabilitydensity functionals 10.5.4.4. Wall damping 9.3.5. Hydrodynamic and diffusive instabilities in premixed flames 9. Bodyforce instabilities 9. Use of coupling functions 10. Cellular flames 9.5.1.4.2.1.2.5.2. Damping mechanisms 9. Oscillatory burning in liquidpropellant rocket motors 9.1. 2 .1.1.1. Solid vibrations 9.6.2. Relative importance 9.1. DifSusivethermal instabilities References 10 THEORY OF TURBULENT FLAMES 10.2. Hydrodynamic instabilities 9.1.2.5.6.2. Nonlinear effects 9.xviii Contents 9.3.4.1.1.4. AmpliJication criteria 9.5. Combustion response 9.1.4.1. Par tide damping 9. Effects of strain on flame sheets 30 1 304 304 305 308 309 309 311 312 315 315 315 318 319 323 324 328 336 339 341 34 1 349 350 352 357 365 373 375 375 377 38 1 385 388 392 392 394 402 405 408 .4. Objectives of analyses 10.1.4.5.2.1.4. System instabilities in combustion equipment 9. Fourier decompositions 10.5.5.1.3.1.1.5. Timelag theories 9.4.5. Relaxation damping 9. Averages 10.1.2.2.3.1.5. Average rates of heat release 10.2.1.1.5.1.
Droplet vaporization rate 11.5. The spray equation 11.2.4. Assumptions 11.3. The spray equation 11.6.3. Droplet size distributions 11.4. The model 11. Species conservation 11.3. Particle size and shape 11. Simplified conservation equations 11. Comments on formulations 11.1.5. Extended model of combustion in a liquidpropellant rocket motor 11.1. Overall continuity 11. Flames in turbulence of high intensity or small scale References 41 1 41 1 415 423 429 437 440 446 448 448 449 449 450 450 45 1 452 453 455 458 458 459 460 460 462 462 463 463 463 464 464 465 466 466 467 468 468 11 SPRAY COMBUSTION 11.1.4.4.5. Motivation 11.2.1. Simplified model of combustion in a liquidpropellant rocket motor 11.5.2. Effects of strain on laminar flames 10.4.2. The combustion efficiency and other spray properties 11.3.2.5.5.5.1.1.4.5. Momentum Conservation 11.1.1. Solution of the spray equation 11.3.4. Spray statistics 1 1.3. Species conservation 11.3. Simplified spray equation 11.Contents XIX 10. Momentum conservation 1.4.2.3.2. Objectives of analysis 10.2. The model 1 1. Theory of wrinkled laminar flames 10. Energy conservation 11.3.5.3. Overall continuity 11.4.4.3. Turbulent premixed flames 10.4. Turbulent flame speeds 10.2.3.1.3.3. Droplet drag .3.3.1.3.2.1.2.Energy conservation 11. The conservation equations for dilute sprays 11.3. The distribution function 11.3.2.
2.6.1.6. Momentum and energy conservation 1I .2. The burning rate 12.2.1. Pertinent results of statistical mechanics A. 1.2.2.2.2.2. The laws of thermodynamics A.2.1.5. 1.4. Continuity 11.6.7. Species conservation 11. Relevance to other problems 12.5.2.3. Solution 12. General thermodynamical results A.3.6. 1. Overall continuity and the spray equation 11. Background A. Evaluation of partition functions 52 1 522 522 523 524 524 526 527 . Boundary conditions 12.2. The chamber length for complete combustion 1 1.6.6. Definition of the problem 12.5.2.2.1.1. Derivation of simplified governing equations 12.3.1. The boundarylayer approximation for laminar flows with chemical reactions 12.4. Combustion of a fuel plate in an oxidizing stream 12.7.2.5.6. The mathematical problem and boundary conditions 11. Generalizations 12.1.6.3. The force on the plate 12. Mechanisms of flame stabilization 12. Solution to the problem 11.xx Contents 11.2. Summary of results A.5. Description 11. Solution to the problem 469 469 47 1 472 472 474 475 476 478 479 480 48 1 485 486 486 489 49 5 495 496 498 500 501 502 11.4.2. Thermodynamic functions A.6. Deflagrations in sprays 11.5. Spray penetration and cloud combustion References 12 FLAME ATTACHMENT AND FLAME SPREAD 12.1.6. Processes of flame spread References 503 509 516 Appendix A SUMMARY OF APPLICABLE RESULTS OF THERMODYNAMICS AND STATISTICAL MECHANICS A.
5. Condensed phases A.5.5. l .4. and termination steps B. Differential heat of reaction A.3.5.2.5.2.5. Reactions in condensed phases A.3. Chemical equilibrium A.2.3.2.5.4. Reaction mechanisms B. Opposing reactions B. 1. Phase equilibria A. Boiling points of binary mixtures A. Higherorder reactions B.2.5.2.4. Multiple reactions.2.4. Firstorder reactions and unimolecular reactions B. Definition of heat of reaction A. Kinetics o f hydrocarbon combustion B.1.5. Initiution. Statement of the law B. Vapor pressures of binary mixtures A.3. Temperature dependence of vapor pressures of binary mixtures A.3.1.2. Heats of reaction A.l.3.2.6. Chain reactions and related processes B.4.3.4.Contents xxi A.3. The equations of Kirchhoff and van’t Hoff A.5.2.2.5.3.3.2.2. The phase rule A.2. l .5. Branchedchain explosions B. Idealgas reactions A.3. The adiabatic flame temperature A.4.5.4.2. Reaction order and molecularity B.1. Catalysis 529 529 53 1 532 533 533 535 535 538 538 539 540 54 1 543 544 544 545 547 548 550 552 554 554 554 555 557 558 558 559 56 1 56 1 563 563 565 570 576 58 1 584 . propagation. Heat of formation and other properties A.5.2. equilibrium constant B.3. Heterogeneous reactions A. Calculation of equilibrium compositions A.4.3. Colligative properties of solutions References Appendix B REVIEW OF CHEMICAL KINETICS B.5.4. Thermal explosions B. General equilibrium condition A.5. The steadystate and partialequilibrium upproximat ions B.2.4. Nonidealgas reactions A.7. 1.1.1.6. General methods B.3.2. The law of mass action B .
4. The Arrhenius law B.3.4.2. Gas transport ratecontrolling B. Description of the heterogeneous processes to be discussed B.3.5.3. Transitionstate theory B. Comparison between the conservation laws derived for independent coexistent continua and the kinetic theory results for multicomponent gas mixtures (2.3.2. Comparison between transitionstate theory and collisional theory B.xxii Contents B.4.1. The activation energy B.3. Energy equation C. Other applications of transitionstate theory B. Definitions of fluid dynamical variables D.3.2.4. Momentum equation C. Definitions and basic mathematical relations C.3.6. The equation of change . Comparison of conservation equations (2. Collision reactionrate theory B.6.1.3.4.3. Continuity equations C. The velocity distribution function and the Boltzmann equation 618 618 620 624 D.4. Definitions of kinetic theory C. 1. Modern developments in reactionrate theory B. Surface reaction ratecontrolling References 585 585 585 587 589 59 1 593 593 595 595 595 595 598 60 1 Appendix C CONTINUUM DERIVATION OF THE CONSERVATION EQUATIONS C. Proof of equation (6) References 604 605 607 608 610 612 613 614 615 617 Appendix D MOLECULAR DERIVATION OF THE CONSERVATION EQUATIONS D. The rates of heterogeneous processes B.2.3.1.3.4. Determination of the specific reactionrate constant B.2. Adsorption or desorption ratecontrolling B.7.4.3. 1.5.5.5.
Species conservation D. Overall continuity D. Physical derivation of the multicomponent diffusion equation E. The pressure tensor E.5. Macroscopic conservation equations D.5.5.1.2.5.2.2.2. 5.4. Simplified diffusion equations E.4.5. Momentum conservation D. Unified elementary treatment of transport processes E.4. Heat flux E. Thermal diffusion coefficients E. Binary diffusion coefficients E. Dimensionless ratios of transport coefficients References Subject Index Author Index .3. Summational invariants D.Contents xxiii D.2.3.1.4.4.2.5. Energy conservation D. Summary References Appendix E TRANSPORT PROPERTIES 624 625 625 625 626 626 627 627 628 629 63 1 63 1 634 635 636 637 638 640 640 64 1 642 642 643 646 647 651 667 E. Diffusion E.4.2.1. Multicomponent diffusion coefficients E.5.6.1.2.3. Viscosity E. Thermal conductivity E. The heatflux vector E.5.2. Collision integrals E. Coefficient of viscosity E.5. 1.2.5.
Most of the topics considered in this book involve the solutions of these equations for special flows.1. Readers unfamiliar with any of the background subjects are advised to read the appropriate appendixes before beginning Chapter 1. where specializations accurate for a number of combustion problems are developed.CHAPTER 1 Summary of Relevant Aspects of Fluid Dynamics and Chemical Kinetics Since combustion problems requiring theoretical analysis are primarily concerned with the flow of reacting and diffusing gases. The specialized equations given in this chapter describe the flow for most of the combustion processes that have been analyzed satisfactorily. The differential forms of the conservation equations derived in the appendixes for reacting mixtures of ideal gases are summarized in Section 1. The forms that the equations assume for (steadystate and unsteady) onedimensional flows in orthogonal. the student of combustion theory must understandin addition to chemical thermodynamics the conservation equations of fluid dynamics.3.2. From the macroscopic viewpoint (Appendix C). curvilinear coordinate systems are derived in Section 1. the governing equations (excluding the equation of state and the caloric equation of state) are not restricted to ideal gases. Simplified forms of the conservation equations applicable to steadystate problems in three dimensions are discussed in Section 1. Condensed but reasonably complete treatments of these background topics are given in the appendixes. including transport properties and chemical kinetics. 1 .
is p &/dt + pv. are derived in Appendixes C and D and may be summarized as follows.p[(Vv) + (VVy]. which may be combined to give P = [p + ($A . p is obtained from equations (E38)(E40). q is given by equation (E46). is dp/dt +v * (pv) = 0. is In equation (2). vi.P : (VV) + p 1 y i f i ..5. equation (D37).p)/p + c Xfi.Summary of Relevant Aspects of Fluid Dynamics and Chemical Kinetics Boundary and interface conditions must be known if solutions to the conservation equations are to be obtained. For this reason. (5) In equation (3). it is difficult to give general rules for stating them. equation (D33). they may require consideration of surface equilibria (discussed in Appendix A) or of surface rate processes (discussed in Appendix B). equation (D40).)(V * v)]U . for example. to be used repeatedly. THE CONSERVATION EQUATIONS FOR MULTICOMPONENT. q . momentum. Since these conditions depend strongly on the model of the particular system under study. IDEALGAS MIXTURES The conservation equations. (1) Momentum conservation. Overall continuity.1. N i= 1 (2) Energy conservation. REACTING. The organization of the remaining chapters of the book is discussed in Section 1.4. simple mass. However. 1. and energy balances at an interface often are of importance. which may be written as . (D35). v u = .v . i= 1 N (3) Conservation of species. interface conditions are derived through introduction of integral forms of the conservation equations in Section 1. is svpt + v * vv = (V .
and (B59). x. qR. . The transport coefficients appearing in equations (5)(7) are given in Appendix E. The reaction rates wi in equation (4) are determined by the phenomonological expressions of chemical kinetics. N .~T ) Be ~(&/ROT) j=l fi () X. it is found fundamentally through the integrodifferential equation of radiation transport (see Appendix E). (8) ROT which are combinations of equations (B6)(B8)..(fi  N  fj) (7) i j = l i = 1. in which case the other variables may be related to these through the idealgas equation of state.1. The N + 5 dependent variables in equations (1)(4) may be taken to be p. w i= k= 1 M 1 (vI:~ V. U is the unit tensor.. vector notation is employed.1 The Conservation Equations for Multicomponent. i = l .. T. . Reacting.1.p v. . two dots (:) imply that the tensors are to be contracted twice.3k .y. and the identity In equations (1)(7).. is also viewed here as specified. The external forces fi are specified (not derived). IdealGas Mixtures 3 In equations (3) and (4). and the superscript T denotes the transpose of the tensor. N . V is the gradient operator. the caloric equation of state. . N the thermodynamic identity.. Vi is determined by equation (E18) which is vx. and v. (B58). = j= 1 xixj (&Vj  Vi) + (X   Xi) + j$ = 1 [rz)r+ . The symbols appearing here are summarized in Table 1. The radiant flux. .)?(I%  (T)  + 1I. .
2. 1.1.2. ONEDIMENSIONAL FLOW 1. curvilinear coordinatesystems may easily be derived from equations (1)(4). the . standard reference temperature* Internal energy per unit mass for the gas mixturet Diffusion velocity of species it Massaverage velocity of the gas mixturet Molecular weight of species i* Rate of production of species i by chemical reactions (mass per unit volume per unit time)+ Mole fraction of species it Mass fraction of species it Exponent determining the temperature dependence of the frequency factor for the kth reaction* Bulk viscosity coefficient* Thermal conductivity* Coefficient of (shear) viscosity* Stoichiometric coefficient for species i appearing as a reactant in reaction k* Stoichiometric coefficient for species i appearing as a product in reaction k* Density' * Parameters that must be given in order to solve the conservation equations + Quantities determined by the equations given in this section.1 Symbols Meaning A constant in the frequency factor for the kth reaction* Specific heat at constant pressure for species i* Binary diffusion coefficient for species i and j * Thermal diffusion coefficient for species i* Activation energy for the kth reaction* External force per unit mass on species i* Specific enthalpy of species it Standard heat of formation per unit mass for species i at temperature T o* Total number of chemical reactions occurring* Total number of chemical species present* Hydrostatic pressuret Stress tensor+ Heatflux vectort Radiant heatflux vector* Universal gas constant* Temperaturet A fixed. Since the complete form is complicated and will not be required in subsequent problems.4 Summary of Relevant Aspects of Fluid Dynamics and Chemical Kinetics TABLE 1. Unsteady flow The conservation equations in orthogonal.
the coordinate in the flow direction will be taken to be the actual physical distance in that direction (gl = 1). for example. . These conditions appear to define the most general type of onedimensional flow. and K2 = 0 and K3 = 0. (13) which essentially defines (gl. andx. (Note that the numbers in brackets refer to references at the end of each chapter. Here the subscripts 2 and 3 denote the corresponding components of the vectors. g3). are flows in which the radiant heatflux vector. ds2 = g2 dx2. v 3 = 0. = dx and use x in place of x1 in the equations.2. The flow will be assumed to be in the x1 direction.* Of special interest. From equation ( 5 ) it is found that the offdiagonal elements of p vanish (p12 = p Z 3 = p31 = 0) and that the diagonal elements of p are given by * See. the (massaverage) velocity vector and the bodyforce vectors all are parallel to one of the three coordinates. however. ds3 =93 dx3.) t Of course. x2. = 0 and d/dx3 = 0 for all flow properties? and that v 2 = 0. s3) and the coordinates (xl. For brevity we shall assume that g1 dx. yields q2 (14) = and also shows that q3 = 0. g 2 and g3 may depend upon x2 and x 3 . . 1 = 1.L2 = 0. Onedimensional flow is then defined by the hypotheses that dldx. Equation (7) now reduces to 1 aT j=1 .1. in other words. and the relation between the physical distances (sl. N . where a form valid in arbitrary (nonorthogonal) coordinate systems also is given. = 0. s2. [l]. Equation (6) then 0. . .x3) will be taken as dsl = g1 dxl. the components of the vectors in the 1 (x) direction will be identified by italic type without a subscript. g 2 . and all properties are uniform along surfaces in the other two orthogonal directions. . OneDimensional Flow 5 reader is referred to the literature for these equations.
= 0. a/ax. one can show that equation (2) reduces to Equation ( 3 ) becomes and equation (4) may be written as ar + . The simplified form of equation (1) is By utilizing the expression for the divergence of a diagonal tensor in orthogonal. curvilinear coordinate systems.u. curvilinear coordinate systems..u.v 2t = ar.. = 0. In Cartesian coordinate systems g 2 = g 3 = 1. ax wi P I ~ 1 a P9293 ax ( ~ y i V & ~ g 3 ) .6 Summary of Relevant Aspects of Fluid Dynamics and Chemical Kinetics and where use has been made of expressions for the divergence and the dyadic gradient in orthogonal. and g1 = 1. In combustion problems it is usually more convenient to work with the enthalpy h = u + p / p instead of the internal energy u. A modified form of the energy equation may be derived from equation (21) by making the * There IS an error in equation (14) of [I].) . the 2 and 3 components of all vectors are zero. the pressure tensor is diagonal..i = 1. N . . in the present case./h.. (pJJ/bJ) should be replaced by (p. (22) Equations (19)(22) may also be obtained by specializing the conservation equations given in [l]* to the case in which a/dx.
+q  ( p . + * This sum often is called the stagnation enthalpy flux and includes thermal enthalpy.2.2. Equation (19) may be integrated immediately. the leading term on the lefthand side in equations (19)(22) disappears. in addition. a relationship quite similar to equation (21a) may be derived from equations (1)(3) for general flows.+ d/dx because only one independent variable remains. considerable simplification in equations (19)(22) may be effected. After recombining terms in the appropriate manner. Steady flow In certain steadyflow systems. yielding the simplified continuity equation (23) In many steadystate combustion problems. body forces are negligible and a consequent simplification in the energy equation is possible. OneDimensional Flow I substitution u = h . the result is found to be N Actually.pll)v]. may be explained as “displacement work.2.p/p in equation (21).” An integrated form of the momentum equation may be derived only under additional restrictive conditions.pll)u Equation (24) states that the sum* of the enthalpy flux and the ordered kineticenergy flux may change only because of heat flux (g2g3 q) or viscous dissipation [ g 2 g 3 ( p . 1 = constant. multiplying equation (21a) by pg2g3 and utilizing equation (23) shows that equation (21a) may be integrated once. 1. When a/& = 0 and fi = 0. if viscous forces are negligible. . and utilizing equation (19) to eliminate a( l/p)/at. then Id In p/dx I < Id In u/dx I (that is. Equation (24) also illustrates that the enthalpy is of more importance than the internal energy in steadyflow problems. giving pug2g3 = constant. In steadystate flow with fi = 0. u 2 / ( p / p ) < 1. then equation (20) reduces to pu du/dx dp/dx = 0. the additional term p/p. chemical enthalpy. If. which appears when equation (24) is written in terms of u. multiplying equation (20) by u. and d/ax . and kinetic energy.1. substituting the result into equation (21). Here a/& = 0.
(26) + since equation (23) implies that (27) when g2 = g3 = 1. i = 1 . . i 1.  q  p +K and  ): : = u . p1 = p . (25) The condition v 2 / ( p / p ) < 1 is valid for lowspeed flow. specifically. ei = pE.N .(. Utilizing g 2 = g 3 = 1.. we find that pv = constant pv h ( + Z) + . pressure forces. . the integral of which is pv2 p l l = constant.. plane.= constant d [pyi(v dx + V. and g 2 = g 3 = 1. In these cases it is not necessary to use equation (20) (except to compute the actual small pressure change. . = 0. In steady onedimensional flows it is often useful to introduce the massflux fraction of chemical species i. .)] = wi. .. after the rest of the problem has been solved with p = constant).. (29) Equations (26)(29) are the conservation equations for steady.f. i = 1.8 Summary of Relevant Aspects of Fluid Dynamics and Chemical Kinetics fractional changes in the pressure are negligibly small) and the solution of the momentum equation is p = constant.. onedimensional flow with no body forces. . (30) Although ei need not lie between 0 and 1 (it may be negative or greater than ti = 1. . Multiplication of unity).(u + F)/pv. N . equations (22) (with appropriate set of conservation equations.($p + K ) dv/dx according to equation ( 16). In (steadystate) plane. N . When equation (25) is an acceptable ap= 0) through (24) constitute an proximation. equation (20) reduces to pv du/dx + dpll/dx= 0. In equation (26). s = entropy]a measure of the ratio of the ordered kinetic energy to the random. the square of the Mach number [v2/(ap/ap). onedimensional flow ( g 2 = g 3 = l). the definitions of yi and imply that equation (22) by p g 2 g 3 and utilization of equation (23) shows that when ajat = 0. the additional restrictions imposed in the preceding paragraph are not needed in order to integrate the momentum equation. the species conservation equation is dei/dx = wi / ( pv g2g3) . (31) . . Equation (26) therefore expresses a balance between dynamic (inertial) forces. thermal kinetic energy of the moleculesmust be small. and viscous forces. When av/at = O. d / d t = 0 and the above results in equations (22) and (24).
pll)v dx dT = constant. Since the acceleration due to gravity is the same for all species.. The above equations serve to eliminate completely the diffusion velocities from the governing equations.1. Body forces disappear from equations (3) and (7) and remain only in the momentum equation. then g2g3 p u x.+ Reasons for neglecting these phenomena are discussed in Appendix E. some combustion problems remain tractable when radiation is included because the effects of scattering and absorption often are negligible except at interfaces.i).1.Xi) V p / p ] . radiation pressure and radiationenergy density.3 Coupling Functions 9 In terms of ti.X i ) . The Soret and Dufour effects (terms involving DT. and consequently fi is not negligible. N . [ (i:l C hiti +  z)  A. is the same for all i in these problems. 4. Bulk viscosity ( K ) . replacing them by the flux fractions. * In some lowspeed combustion problems natural convection is of importance. Body forces* (fi terms). . f. 2. There are a number of combustion problems in which the neglect of q. Other radiative influences. is quite complicated. Pressuregradient diffusion [ ( y .. 1. + (x . 5= dx j=1 (T)(x xixj to j). whenh equation (14) becomes =0 and thermal diffusion is neglected (for brevity). (33) In this form of equation (24).dP P dx i = 1. 3. the effect of diffusion on the enthalpy flux is included implicitly in the first term.. from which qR is obtained. 5. Radiant heat flux (qR). COUPLING FUNCTIONS A number of effects contained in equations (1)(7) are unimportant in a large class of combustion problems and will be neglected in nearly all of the applications to be considered subsequently and also in the equations to be derived in this section. is invalid. The fact that equations (31) and (33) appear to be less complicated than equations (22) (with = 0) and (24) often justifies this transformation. These include: 1.. and the only term from the heatflux vector that remains explicitly is that of heat conduction.3. always appear to be negligible in combustion. (32) Using these same assumptions and the (idealgas) condition h = hi one can show by substituting equation (15) into equation (24) that if radiant transfer is neglected.( p . Although the general formof the equation of radiation transport.
2. in a reasonable approximation. viscous effects are often negligible and. the resulting equations remain valid. for many of the problems considered in later chapters. N in the present approximation. . There are many approaches to the derivation of equations for coupling functions. approximately. species conservation. equation (4). ) ] = w. where D = D i j .(v + V . the use of coupling functions often enhances understanding of the physical processes that occur.. equation (l). In addition to providing significant simplifications in analysis. X i = 1 by definition. (pv) = 0.[pY(v + V . This enables the chemicalsource term to be removed from many equations by considering suitable linear combinations of dependent variables. Nevertheless. i = 1. Utilizing equations (10) and (34) in this expression. ) . multiplying this equation by = 1.reduces to v . (34) For lowspeed. i = 1. Since Y. The assumptions needed limit the range of applicability of the formulation to a considerably greater extent than do the approximations given in the preceding paragraph. These linear combinations often are descriptively called coupling functions. One involves equations given in early papers of Shvab [ 2 ] and Zel’dovich [ 3 ] .h. For s t e a d y j o w overall continuity..= Xi j= 1 cXjVj N  XiVi. In view of equation (34). in this approximation. ) N . the momentum equation implies that. . The substitution of equation (6) into equation (3) yields. may be rewritten as v . we obtain Y. steadyflow problems.i lV T i =1 1 = 0 (35) for the energy equation. and summing over i yields a relation for X j V j that may be ./X. . equation (7) reduces to D VX. and Y. (36) When the binary diffusion coeflcients of all pairs of species are equal.Vi = 0.10 Summary of Relevant Aspects of Fluid Dynamics and Chemical Kinetics Introduction of additional restrictive assumptions is useful in analysis of many problems. . The objective of the following manipulations is to express the equations for conservation of energy and of chemical species in a common form.. and the resulting formulation often is called the ShvabZel’dovich formulation. p is a constant. . as pointed out in Section 1.
1''' Y. not derived from fundamentals through explicit hypotheses. dT] according to equation (38). Equations (40) and (41) are the energy. [ p v ( f T l c p d T ) pi:lyiVi + N N where cp = 1 yiCP. showing that D VlnXi.i i= 1 N (38) is the specific heat of the mixture. obtained from the total enthalpy.. Coupling functions may now be identified from equations (40) and (41).and speciesconservation equations of Shvab and Zel'dovich. .3. . that is. the integration is performed * Here. V i = DVlnyi. i = 1. By substituting equations (11) and (36) into equation (39. . we obtain Fick's law..VlnXj j=1 [ N Substituting equation (12) for X i into all terms on the lefthand side of this expression. Coupling Functions 11 substituted back into the equation. N.. . ( P V ~ pDVY. 1 = Vi. (37) a simple and wellknown diffusion equation that often is merely postulated. by subtraction of the chemical enthalpy. i = 1. h. .xE. the similarity in the forms of equation (40) (for the thermal enthalpy iFo c p d T ) and equation (41) (for the mass fractions yi) is striking.) = w i . (41) In view of equation (39). . d T . we find V . N . . . N.hp. Equations (36) and (37) imply that V. precisely j& c p dT = at constant composition.1. & cp. . (39) we may substitute equation (37) into the above form of the energy equation to show that since V JFoc p dT = c p VT [(Vyi) & cp..* Assuming that the Lewis number (Le) is unity. Le = A/pDcp = 1. The derivation given for these equations required neither that any transport coefficient or the specific heat of the mixture is constant nor that the specific heats of all species are equal. This integral represents the thermal enthalpy. i = + If= 1.
(@]) = w (48) can be solved for a l .. diffusive.mi + 1v. (44) (45) and *T for equation (40). assume further that chemical changes occur by a single reaction step.. while = YI / W 1 ( V~ v:) ai.mi.a.12 Summary of Relevant Aspects of Fluid Dynamics and Chemical Kinetics To emphasize more explicitly the similarity of equations (40) and (41). . Such balances are very common in combustion and often are employed as points of departure in theories that do not begin with derivations of conservation equations.v:) = o.pD V a l . . i = 1. and reactive effects. The nonlinear rate term may be eliminated from all except one of the relations corresponding to equation (44) by subtraction. In the operator L. In this case. (47) for equation (41). then an additional term. equations (48) and (49) may be derived with this generalized definition of L. i= 1 N N i= 1 where mi is the symbol for chemical species i. N . d(pa)/at. . the first term represents convection and the second diffusion.this term represents accumulation of thermal energy or chemical species. where the linear operator L is defined by L(a) E v * [pvGl . For species conservation. E PT and P = ai .a1 = ji( i # 1). and equations (40) and (41) may both be expressed as L(a) = 0. N .. If the steadyflow approximation is relaxed. 1v.appears in L . in the absence of the assumptions of lowspeed flow and of a Lewis with P = . (43) is the same for all species (see Appendixes A and B).. . then the other flow variables are determined by the linear equations for the coupling functions P L(P) = 0. the quantity w i / w ( v i . Equation (44) therefore describes a balance of convective. for example. i = 1. if I. (49) aT .
4. be pointed out that equation (49) is deceptively simple in appearance. in addition. the retention of assumption 5. equation (39) and the approximation of lowspeed flow. unless additional approximations are made.2). burning rates often may be determined by solving equation (49). without attacking equation (48) [4]. CONSERVATION CONDITIONS AT AN INTERFACE Relations expressing the conservation of mass. In such cases. however. and equation (49) actually becomes nonlinear. Only the restrictions 1.1. to the effect that aplat is negligible in an expression like equation (21a). Integration of equation (1) over V and utilization of the divergence theorem and of equation (50) forf = p yields f ( J v p d V ) + Jdp(v  g) nd d = 0. a result like equation (49) often can be derived for the total enthalpy h through reasonable approximations (see Section 3. Nevertheless. coupling functions obeying equation (49) still may be derived through use of atom conservation. The time rate of change of the integral of an arbitrary functionf(x. pv [which is determined by equation (34). t ) over a given volume V is where d is the surface bounding V. Although the onereaction approximation is sufficiently accurate for k? many purposes in combustion. To obtain equation (44) for energy conservation in timedependent problems necessitates. 1. For systems in which reactants initially are unmixed. Conservation Conditions at an Interface 13 number of unity. and energy at an interface may be derived by writing equations (1)(4) in integral form and then passing to the limit in which the volume of integration approaches a surface. as well as replacement of the steadyflow approximation by an additional assumption. and 3. 2. more than one rate equation of the type given in equation (48) remain to be solved. and the assumption of a onestep reaction are needed . and the term involving d x l d t accounts for the changes associated with motion of the control volume. an equation of state with p = constant and in some cases an approximate form of the momentum equation] and p D may depend on pi or j T .4. momentum. thus the formulation is appreciably less general for energy conservation than for species conservation. the equality of diffusion coefficients.4. . n is the outwardpointing normal to this surface.therefore.so that L depends implicitly on p [see equation (45)]. It should. the formulation is of reasonably wide applicability for species conservation. However. equation (A36). it is often important in detailed investigations to replace equation (42) by a kinetic scheme of greater complexity.
Control volume for the derivation of interface conditions. the integral over d is composed of two parts. the thickness of which is allowed to approach zero (see Figure 1. In the limit.1). which may be integrated over Y to show that by the procedures described above. We now let the control volume ^Y be a thin slab.14 Summary of Relevant Aspects of Fluid Dynamics and Chemical Kinetics By multiplying equation (1) by v and adding the result to equation (2). integrals over each face of the slab.we find that The substitution of equation (I) into equation (4) yields equation (D39). namely. Let us arbitrarily choose one face as the “positive” side of the slab and identify quantities on this side by the subscript + and those on the other t nFIGURE 1. Integrating this result over Y. whichupon integration over V”reduces to (54) Equations (51)(54) are the integral forms of the conservation equations. then equation (1) may be used to obtain an equation which is essentially identical with equation (D36). . one obtains equation (D34). If equation (2) is dotted into v and substituted into equation (3).1.
the volume integral essentially approaches a surface integral). JdI [p+(v+  g)  p(v.. and 6 is the Dirac 6function. n+] + (P+ .1.q .yi. momentum. then wi + wi 6 ( y .4. surface reactions.N .) ...P + . n + d d 1 and [i’.P. at the surface.. limv+o is to be interpreted as the limit in which the thickness of the slab becomes infinitesimal (that is. yi+(v+ JdI + vi+  g) d  p . By substitution. (58) from equations (51)(54).v . y is the coordinate normal to the surface. i = 1. or a chemical species at the interface.P . wi dd. where wi is the surface reaction rate (mass of species i produced per unit area per unit time). energy. It is clear that n. account for the possibility of surface forces.. Conservation Conditions at an Interface 15 side by the subscript .v  ( + vi.. .+I  PV[(V  3. if the interface is taken to be a solid surface at which chemical reactions are proceeding at a finite rate. .dx)]n+ dt  d d = lim JYwi d V 770 ( JvpX d V ) . and in the limit the integral over the surface of V may be yielding replaced by an integral over the interface area d.)I n+ d d = V lim +O [it J/  d 4 (55) Jd1{P+V+[(Y+ %). . y = y . For example.+ v + . or a nonzero time rate of accumulation of mass. respectively.y l ) . The terms involving limY+oof volume integrals remaining in equations (55)(58). n+ at corresponding points on the two faces. which is we then find that in equation (58) limv+o Jy wi d V = SdI .. In equations (55)(58)..n + I dd + q+ .
for example.v . In many problems.3 are valid.n + >+ P . then equations (55)(58) simplify considerably. Although equation (60) will not be used explicitly in the subsequent applications.we find by passing to the limit of small d. If pi is the mass of species i per unit surface area. S(y . which is the negative of the rate of accumulation of mass of species i on the surface. all terms in equations (55)(58) involving are zero. (61). It is apparent that corresponding interpretations may also be given for equations (55)(57).. In equations (61) and (62). and (62) often will be needed as boundary conditions for equations (1)(4).y. then p x + pf S(y . that viscosity is negligible. with additional approximations (see Section 1. by Fick's law. In equation (61).yI). lim I V p . . The difference between the mass flux of species i leaving the surface and that entering the surface equals the mass rate of production of species i at the surface minus the mass rate of accumulation of species i on the surface.IdTpi d dt dd.that the differential forms of equations (55)(58) reduce to p + v + n+ = p . and that approximations 14 of Section 1.3). qR has been retained because radiative heat losses from surfaces often are substantial.v(v.n . Using equations (5) and (6) and the identity h = u + ( p / p ) . n + . there is usually no excess mass at the surface so that limY+of p d V = 0 in equation (55). and the remaining term on the righthand side of equation (58) is p yi d V =  d . and respectively. p + v+(v+ n + ) + P+ n+ (59) (60)  = P . equations (59).) d V dt Y0 = . Detailed discussions of the importance of . the V i values are determined in terms of concentration gradients at the interface either by the form of equation (7) in which the last three terms are omitted or. Equation (58) therefore expresses the following physically obvious conservation condition.16 Summary of Relevant Aspects of Fluid Dynamics and Chemical Kinetics the mass rate of production of species i on the surface. If it is also assumed that the interface is not moving (dx/dt = 0).
5. Chapter 12 concerns basic aspects * See. addresses a problem of engineering importance and also provides a vehicle for systematic amplification of basic concepts. None of the differential conservation equations discussed in this chapter are needed in Chapter 2. Chapter 3 is concerned with problems in which transport effects are of importance but the chemical reactionrate terms appearing in equation (4) and given by equation (8) need not be considered in obtaining first approximations for most of the information of interest. The analysis in Chapter 8 of classical aspects of flapmability emphasizes the influences of heat losses and of finiterate chemistry. DISCUSSION OF THE APPROACH ADOPTED IN THE FOLLOWING DEVELOPMENT OF COMBUSTION THEORY In the following chapters we shall begin with problems that are. the formulation developed in Section 1. where relations between the properties on the upstream and downstream sides of combustion waves are developed from overall (algebraic) conservation equations. [6] and [7] for applications and generalizations of equations (61) and (62).1. The arrangement of the remaining chapters is less strongly dependent on pedagogic considerations. turbulent combustion. for example. presented in Chapter 7. and [S] for an amplification relevant to equation (62). to problems of steadystate. developed therein. The study of combustion of solid propellants. The current status of our theoretical understanding of turbulent flames is reviewed in Chapter 10. Chapter 4 deals with problems in which finite chemical reactionrates must be considered. and spray combustion. fundamentally.* 1. but the transport phenomena [the most highly differentiated terms appearing in equations (2)(4) after use is made of equations (5)(7)] all are negligible. onedimensional burning in liquidpropellant combustors and of deflagrations in sprays.5. the least complex and proceed to successively morecomplicated problems. Processes in which both transport phenomena and finite chemical reactionrates must be considered simultaneously even in the most elementary descriptions are encountered in Chapter 5 on laminar flame theory and also appear in the discussions of detonation phenomena given in Chapter 6. namely. Results of Chapter 3 are employed in Chapter 11 to apply the spraycombustion theory. .3 is useful in these problems. Approach Adopted in the Development of Combustion Theory 17 various terms in equations (59)(62) for particular systems may be found in the literature. The bases of theoretical descriptions of unsteady burniqg in a wide variety of combustion problems are given in Chapter 9. The next three chapters address from a basic viewpoint three classes of combustion problems that are of high practical interest. [ 5 ] for an analysis of equation (61). combustion instability.
2. 141 (1959). B. D. DA04495Ord446. Sci 25. Pasadena. Shvab. B. Enery. A . J. 19. P. P. S. 18CL182. A. an attempt has been made to cite references that can be pursued in advancing to any frontier of the subject. W. Spalding. 1974. 847864. “CondensedPhase Mass and Energy Balances”. chapter 3 of Heat Transjer in Fires: Therrnophysics. Penner and F... 1296 (1950). 8. Williams. Scala and G. 1152 (1965). A. M. 5. Rept. Zhur. Sutton. Memo. (1957). 7. A. English translation. California Institute of Technology. F. izd. 1 199 (1 949). New York: Wiley. Blackshear. Economic Impact. Libby and F. A R S Journal29. 4. . Williams. A. Tech. 19. Williams. Williams. (1953).18 Summary of Relevant Aspects of Fluid Dynamics and Chemical Kinetics of problems of flame attachment in combustion devices and of flamespread processes such as those occurring in fires.. Scripta Book Co. Contract No. Fiz. No. Aero. Gos. L. Y . Tekhn. Social Aspects. No. 4th Symp. MoscowLeningrad (1948). NACA Tech. F. ed. A. AZAA Journal 3. 3. Zel’dovich. 6. S. S. 343 (1958). REFERENCES 1. V. Since a thorough presentation of every aspect of combustion theory cannot be included in a volume of this size.
2 and 6. In this chapter. plane.1). where detailed analyses of each type of wave are presented. Such a classification provides a framework within which plane deflagration and detonation waves may be investigated. In such problems. Since the major changes in the values of the flow variables in these systems usually take place over a very short distance (in nearly all cases less than a few centimeters. then the Hugoniot curve for a simplified system is studied in detail in order to delineate explicitly the various burning regimes.1. the equations derived in this chapter provide all of the information that is required concerning these waves. in which the properties become uniform as x + & co. nonheatconducting. The experimental conditions under which these waves appear are described in Chapters 5 and 6. onedimensional flows involving exothermic chemical reactions. see Sections 5. steadystate. nondiffusive. Some other texts discussing the RankineHugoniot equations are listed in [1][5]. ” 19 . except their speed of propagation. The RankineHugoniot relations are the equations relating the properties on the upstream and downstream sides of these combustion waves. in many problems deflagration and detonation waves can be treated as discontinuities (at which heat addition occurs) in the flow equations for an “ideal (inviscid.CHAPTER 2 RankineHugoniot Relations In this chapter we shall classify the various types of infinite. nonreacting) fluid. general RankineHugoniot equations are derived and discussed first.
(4) (5) The equation of state.. and equation (128) (the energy equation) implies h. The sequence of states at x = co (with fixed parameters at x = .. equation (126) reduces to Pod + Po = PWVW + Pm (2) (momentum conservation).00) will be identified by the subscript 0. 2 (1) where m is the mass flow rate per unit area.1. (3) where use has been made of equation (1). Equations (126)( 129) govern the system. viscous effects FIGURE 2.. and.1. diffusion. provide the additional requirements W ~ .. Since dvldx approaches zero as + f 00.+ 0 as x . The flow is illustrated schematically in Figure 2.1. and it will be assumed that the flow is in the +x direction and that properties are uniform in planes normal to the x axis. .~=Oi .a) for which equations (1) and (2) are satisfied is often referred to as the Rayleigh line. = l . Derivation of the equations A coordinate system that is stationary with respect to the wave will be adopted. q . N . Schematic diagram of a deflagration or detonation wave . N . equation (129). reaction. The speciesconservation equations. Zone involving heat conduction.. ~ = O i.+ f 00. = l . while downstream conditions (at x = co) are denoted by the subscript co.1.20 RankineHugoniot Relations 2. Upstream conditions (at x = . + ~ $ 2 = hw + &/2. . The continuity equation [equation (127)] implies that + povo x = pmvm = m.1. GENERAL RANKINEHUGONIOT EQUATIONS 2. and W~. Since q is proportional to temperature gradients and concentration gradients..
2. for idealgas mixtures equations ( 6 ) and (7).. K. ..k is the difference of two terms of the type appearing inside the summation in equation (18). p m . The speciesconservation equations require further discussion because equations (4) and (5) may not all be independent. h= 1 where wi. d h o .and finalflow properties. Using equation (1) to express u in terms of m and p in equation (3) yields when equation (8) is used to eliminate m2.l/po) plane. and wi._ w i = C w i .k is the mass rate of production of species i in the kth reaction and there are M independent reactions. From equation (18) we see that the mass rate of production of species i by chemical reactions may be expressed as M . Equation (9). . Since equation (1) implies that equation (2) may be written as which defines a straight line (the Rayleigh line) with a negative slope in the ( p . N . '0. General RankineHugoniot Equations 21 and the caloric equation of state. as velocities have been eliminated. and all of the information that can be obtained from equation (129) is not contained in equations (4) and (5).o) = d h m . The convention to be adopted here is slightly different from that for equation (18).. which is called the Hugoniotequation. it follows that . In future work it will be convenient to employ modified forms of equations (2) and (3). forward and backward reactions will not be considered separately as is implied by equation (1S). From the stoichiometry of the kth reaction. k . so that the present M is half of the previous M .p o ) .mL (7) constitute further relationships between the initial. respectively. reduce to equation (19) and the relation obtained by substituting equation (111) into equation (110) ( h = u + p / p ) ..1. T m . Y. PO.(l/pm . i = l . is a relationship among thermodynamic properties alone.
~ # 0. The coldboundary difficulty The restrictions given by equation (13) [or equation ( 5 ) ] are not fulfilled exactly in deflagration or detonation waves. . are linearly independent. N l v ? ) w i / w = 0. when M < N it is possible to derive additional (atomconservation) conditions from equation (129). since M I N in general. Although it may be found that not all of the L additional relations between the properties at x = +a and those at x = a. ~ ... Substituting equation (129) for w. (which is independent of i) is the net rate of the kth reaction (that is.co in Figure 2.00 to x = + oc yields 1vp)x. there are M equilibrium conditions and N ..2). However. into equation (14) and integrating from x = .1. Hence equations (12) and (15) provide a total of N independent relations among the flow properties at x = 0 0 . (13) which may constitute fewer relations than equations (4) and (5). . the stable condition is that in which the mixture is composed of products of combustion.. the difference between the forward rate and the backward rate. . . where L is the total number of atoms in the system.22 RankineHugoniot Relations where w.. The initial mixture is in a metastable state and therefore actually reacts at a small but finite rate.2.o/w= 1v ? x . J Y . The combustible mixture approaching from x = . L. it will always be true that N . u.=0... given by equation (15). M . .. L.M .3. just as in the problem of computing equilibrium compositions in closed systems (Appendix A). k = l . .7).. Equations (10) and (11) imply that equations (4) and (5) are satisfied when 0. . Hence w . .1. (12) and k = l .=O. If vy) denotes the number of atoms of kind j in a molecule of kind i. . . I= N N j = 1.1 is not in stable chemical equilibrium.M of these relations are independent (compare Section A.M atomconservation conditions. but w . j i= 1 = I . 2. see Section B... [equation (l)] and + 0 as x + co (because diffusion velocities are proportional to gradients of the flow properties). . then the fact that atoms are neither created nor destroyed by chemical reactions is expressed by the equation w. (15) 1 I= 1 since po uo = p.
However.. the values of T . As illustrated in Figure 2. The nature of this curve. equation (9). and (15). and (15) constitute the required forms since velocities appear only in equations (1) and (8) of this set. (6)(9). because the same trouble necessarily arises in studies of detonation structure (Section 6. the conditions that fix v o (to be defined in later sections) may be introduced to complete the solution. p . . thus emphasizing how elementary it is.. l/pm) plane along which the solution must lie regardless of the value of 21. and the velocity at which the combustion wave propagates down the tube will be measured. is investigated in the following section. from as many equations as possible. equation (13) will be satisfied approximately for any initial mixture composition for which the mixture can be contained long enough to support a detonation or deflagration wave. called the Hugoniot curve.h.1. (6)(9). then these N + 5 equations would determine the downstream flow variables u . Use of the RankineHugoniot equations In summary. If the problem were formulated in terms of wave propagation in a given combustible mixture..1. study of combustion waves is facilitated by eliminating u. Furthermore. may then ~ (x. If all upstream conditions (including uo) were specified. Therefore. 2. and Y. (12). After the equations that do not contain u. General RankineHugoniot Equations 23 is very small compared with its value in the reaction zone so that.=. T. The value of v. the propagation velocity u. To. and (15) comprise the independent relations between upstream and downstream conditions. it appears to be desirable to introduce this “coldboundary difficulty” at an early stage. equation (13) actually provides no reasonable restriction on the initial conditions and will be omitted from the governing equations.3. .2) and of deflagration propagation velocities (Section 5. equations (l). in order to determine a curve in the (pm. ..J * This calculation is similar to the equilibriumcomposition computations discussed in Appendix A. in a typical experiment a given mixture in a tube will be ignited. is not controlled directly but instead is determined by other experimental parameters.. given. (12). p . from equation (7). after m has been obtained from u.. Therefore. (12). h .2).3.* the enthalpy h. .2. equation (13) would not even arise. equation (13) is satisfied approximately. With p o . . the intersection of the Hugoniot curve with the Rayleigh line defined by equation (8) determines the final therinodynamic state. are solved for relations among flow properties downstream. Equations (l). in some sense. and the result may finally be substituted into the Hugoniot equation.2. p . and the chemical equilibrium composition can be calculated for various values of p4 and pa. from equations (6).2. for the particular experiment.. may then be evaluated for these same values of pa and pa. and y. Although the initial composition and thermodynamic properties are adjustable experimentally.1.
~ and (3) between the initial temperature To and any final temperature T. (1) the final equilibrium composition along the Hugoniot curve is constant (that is. . [ S ] ) . Variations of this procedure have been programmed in recent years for solution by electronic computers (for example. 2. i = .2. It will be assumed that for the given initial thermodynamic properties ( p o . The simplest system for which .=~ 1 ~ . which provide accurate Hugoniot curves based on the best values available for thermochemical properties. C ~ .24 RankineHugoniot Relations Pm t I l/PO  l/Pm FIGURE 2. We shall consider an ideal gas mixture in order to assure that equations (19)(111) are valid. p o . (2) the final average molecular weight equals the initial average molecular weight that is. The determination of downstream flow properties from the Hugoniot curve. o c p . y. are all fixed).2. along the Hugoniot curve. y. in some T range including To and T. o. ANALYSIS OF A SIMPLIFIED SYSTEM 2. the specific heat at constant pressure at the final composition equals the specific heat at constant pressure of the initial mixturethat is. be calculated from equation (1). i= 1 C X .2.1. . o cp i= 1 N N (=constant) (17) for cp. Simplification of the RankineHugoniot equations The essential characteristics of the Hugoniot curve are most easily illustrated by studying a particular simple system. and so on).
Solving equation (18) for T. is the specificheat ratio for the final mixture. The heat of reaction of the mixture may be defined as 4 i= 1 ? [(Y. Substituting equation (20) into equation (19). Equation (23) may also be obtained by considering a . which.) = y / ( y . we obtain h.  h. = c.2. Analysis of a Simp!ified System 25 equations (16) and (17) are valid is one in which all species have equal = W for all i) and constant and equal heat capacities molecular weights (c. after subtraction.~ = c.  To). in view of equation (17).cc and at x = + cc yields =u + p/p. q. The temperatures To and T. where y = CJC. = constant for all i). it follows that c..o +  Y. (21) where use has been made of equation (17) in obtaining the final equality.2.idT)].)(ho + IT:cp. = 4 dT = 4 + cJT. may be eliminated from equation (21) by using equation (18). and p m . Ro/W for an idealgas mixture. and po in addition to the variables p . solving for h and evaluating the result at x = ./(Ro/W) = c. In view of equation (16). . and To and substituting these results into equation (21) therefore yields Equations (9) and (22) show that the Hugoniot equation may be written in the form which containsonly the known constants y./(Ro/W)will enter into the result. It can be seen that the ratio c. p o . is the total chemical heat released per unit mass of combustible mixture. equation (19) reduces to (w Substituting equation (111) into equation (1lo). Since the specific heat at constant volume is given by c.l).c../(c.
A dimensionless heat of reaction is = 4Po/Po7 and a dimensionless massflow rate is P = m’lpo P o .l)/(y 1). It will be observed that p . 2. A dimensionless final pressure is the pressure ratio P =PXiPO./uo according to equation (1).4 for y = 1. Equations (8) and (23) therefore determine the complete solution for the simplified system. A dimensionless final specific volume is the density ratio 0 = Po/& > which is also equal to the velocity ratio u. .Therefore. respectively. approach the lines u = (y .3 and 2. Dimensionless form The nature of the results will become more transparent if dimensionless variables are introduced.2. u.3. and p must all be positive for physically acceptable solutions.2.( y .2.26 RankineHugoniot Relations onecomponent ideal gas to which an amount of heat q (per unit mass) is added externally.* All curves asymptotically 1) and p = .l)/(y + + * The curves for a = 0 are of significance in connection with ordinary shock waves.2. From equation (23) an explicit algebraic expression for p a as a function of p x along the Hugoniot curve may be obtained.4 and y = 1. Properties of the Hugoniot curve The Hugoniot curves given by equation (24) are plotted in Figures 2. Multiplying equation (23) by po/po yields the solution of which is Dividing equation (8) by p o po gives 2.
the slope of this straight line is negative. see equation (24)] called thedetonation branch and a lower branch [0 I p I 1.2.4 the entire range of pressure ratio (0 I p I c o )may occur. Acceptable end states must lie on one of these two branches. see equation (24)] called the deflagration branch. Each Hugoniot curve is therefore divided into two nonoverlapping branches.1)a I p I 00. but the specificvolume (or velocity) ratio is restricted to the interval (Y .p [equation (25)] establishes the final state of the system. 1 + (y . ( y . 1)with slope .l)a/y I v I 2a + ( y + l)/(y . Hugoniot diagram from (24) with y = 1.l)/(Y + 1) I v I 2a + [(r + l ) / ( Y  111.3 and 2.1).4 are physically meaningless. and end states lying in the two shaded quadrants in Figures 2. Since p > 0. The intersection between the appropriate Hugoniot curve and the straight line through the point (1. Analysis of a Simplified System 27 P=  r+l V 71 FIGURE 2.3.l)/(y + 1) I v I 1. Combustion waves are termed detonation waves or deflagration waves according to the branch of the Hugoniot curve upon which the final condition falls. . with the upper bound corresponding to the limit p + 0 [see equation (24)].2. an upper branch [l + ( y .
1).4 that a unique straight line through the point (1. in going through a deflagration the gas speeds up and expands.2.28 RankineHugoniot Relations P V= P= . 1) is tangent to the detonation branch of the Hugoniot curve. This can be proven formally by considering the equation .YT 1 I 0 1 2 3 4 U 5 6 7 = 8 9 FIGURE 2.4. in passing through a detonation wave the gas is slowed down. 2. Although this formal distinction is invaluable because of its unambiguity. Analysis of the detonation branch It is easily seen from Figure 2. Hugoniot diagram from (24) with y 1. and its pressure and density increase.Y.4. an understanding of the difference between detonation and deflagration waves is best obtained by contrasting the characteristics of each. Other striking differences between the two types of waves will appear when we discuss wave structures in subsequent chapters (see Sections 5.3 and 6. Thus (see Figures 2.2. and its pressure decreases.3 and 2. however.4).3 or Figure 2.
where typical solutions lying on the detonation branch ABC are indicated. Under most experimental conditions. one lying on the line A B (a strong detonation) and the other lying on the line BC (a weak detonation).lying on the detonation branch.2.2. detonations are ChapmanJouguet waves. When p exceeds its minimum value p+ (that is. a minimum wave speed) for detonations. a ChapmanJouguet wave. since the reasoning involves concepts of the structure of the wave. at higher wave speeds). Schematic diagram depicting the various sections of the Hugoniot curve. . which states that the slope of the Hugoniot curve equals the slope of the straight line from the Hugoniot curve to the point (1. can be shown to possess only one solution. ) on the detonation branch is said to be the upper ChapmanJouguet point. while it intersects the detonation branch at two points if the magnitude of its slope exceeds that of the tangent line. equation (24). l). u . and the wave with this end condition (point B on the Hugoniot curve illustrated in Figure 2. A straight line through the point (1. This is illustrated in Figure 2. Equation (25) therefore implies that there is a minimum value of p (that is. The experimental conditions determine whether a strong detonation.5. 1) will not intersect the detonation branch anywhere if the magnitude of its slope is less than that of the tangent line. this topic will be discussed more fully in Section 6. Equation (26). This point ( p + .5).2. u +).5. Analysis of a Simplified System 29 where p(u) and dp/dv are evaluated from the Hugoniot equation. there are two possible end states for the detonation. or a weak detonation will be observed at a given value of p. corresponding to the ChapmanJouguet end conditions (at point B in Figure 2. ( p + . P 11 FIGURE 2.5) is a ChapmanJouguet detonation.
Analysis of the deflagration branch Remarks precisely analogous to those given above for the detonation branch of the Hugoniot curve also apply to the deflagration branch. or equation (26).4. Hence the ChapmanJouguet deflagration has the maximum wave speed ( p = p .30 RankineHugoniot Relations Figure 2.3 that strong deflagrations do not occur. A wave with endpoint E is a ChapmanJouguet deflagration. Figure 2. 1) is tangent to the deflagration branch (DEF in Figure 2. 2.5).5 shows that as the nomenclature implies. both of which propagate at infinite velocity ( p = a ) .. Limiting cases are the strong deflagration with p = 0.3. v).5 clearly shows that this maximum deflagration velocity is less than the minimum detonation velocity ( p .6. Differentiating equation (24) yields . it will be indicated in Section 6.2.) of all deflagrations. Figure 2. and it intersects the deflagration branch at two points (once on DE and once on E F ) if its slope is less than that of the tangent line. From considerations of combustionwave structure. Figure 2. particularly those involving detonations. The strong detonation with p = co and the isochoric weak detonation (v = l).5) at a point ( p . and the isobaric weak deflagration ( p = l). Properties of ChapmanJouguet waves ChapmanJouguet waves assume a special significance in many systems. hence the physically meaningful section of the deflagration branch of the Hugoniot curve is DE. which propagates at zero velocity ( p = 0). Most deflagrations are. the pressure ratio and the velocity change across a strong detonation exceed those across a weak detonation. represent unattainable limiting cases. in fact. 1) fails to intersect the deflagration branch if the magnitude of its slope exceeds that of the tangent line. 2. which has a finite (nonzero) propagation speed. Waves with end states lying along line DE are weak deflagrations and those with end states on line EF are strong deflagrations. nearly isobaric. this result can also be derived from equation (26). A straight line through the point (1.< p + ) .5. It is therefore of interest to investigate the properties of the ChapmanJouguet points in greater detail. shows that a unique straight line through the point (1.2. The pressure and velocity changes across strong deflagrations exceed those across weak deflagrations.1. which is called the lower ChapmanJouguet point (point E in Figure 2.
/p./Pm (29) [see equation (l)].2. = o. p . Recalling that the speed of sound in an ideal gas is a m = Jm.which implies that urn = JYP.. From the definitions of the dimensionless variables p . Jyp. if equation (27) is solved for u./a. giving C81. the solution to which is Upper signs correspond to the detonation branch. it can be shown that (28) at the ChapmanJouguet points. and 0. the entropy assumes a local maximum value at the ChapmanJouguet point Explicit expressions for the dimensionless downstream properties (in terms of CI and y) are easily derived for ChapmanJouguet waves. A more thorough analysis shows that the downstream gas velocity relative to the wave exceeds the sound velocity ( M . > 1) for weak detonations and strong deflagrations and is less than the sound velocity ( M . the entropy assumes a local minimum value only at the upper ChapmanJouguet point and a local maximum value only at the lower ChapmanJouguet point [7] and that along each tangent line given by equation (25) (the Rayleigh line). < 1) for strong detonations and weak deflagrations. and lower signs correspond to the deflagration branch. = (s = entropy per unit mass). then a quadratic equation for p is obtained.2. For example. and this result is substituted into equation (24). we see that equation (29) implies that the downstream Mach number M . It is also possible to prove that along the Hugoniot curve. the dimensional form of equation (28) is seen to be m 2 = y p z p . By using this relation in equation (25). is unity at both ChapmanJouguet points. .rl P = YPIV (27) at either ChapmanJouguet point.Analysis of a Simplified System 31 which may be substituted into equation (26) to obtain P = m r + 1)u . Substitution of equation (31) into equation (30) yields .
6. The dependence of the properties of ChapmanJouguet waves upon the dimensionless heatrelease parameter 3. According to equation (34).25 D a d ~ $ 7 $ 2 R I FIGURE 2. . the initial Mach number always exceeds unity for ChapmanJouguet detonations and lies between zero and unity for ChapmanJouguet deflagrations.00 I 3. Equation (34) shows that both M o + and M. = uo/uo = tion of p.+ 0 as the heatrelease parameter a goes to zero. we find that The temperature ratio may be computed from T.. while M o + + c as c( r 00. we conclude that all detonations propagate at supersonic velocities and all deflagrations propagate at subsonic velocities.. Equation (33) may be used to compute the initial Mach number for ChapmanJouguet waves. = p u [see equation (IS)] .75 3. The result is = from equation (1) and the definisince M. as determined by equations (31)(34) with y = 1. Since the ChapmanJouguet waves correspond to the minimum propagation speed for detonations and the maximum propagation speed for deflagrations.50 0.3.32 RankineHugoniot Relations By using equations (31) and (32) in equation (25). uo/dyz f i 1./T.approach unity o and M. the downstream temperature is found to exceed the upstream temperature at both ChapmanJouguet points.
6 shows that. It is understood that we are always referring to downstream conditions here. p + = 20.= 15..8 (giving uo % 1000m/s) at the upper ChapmanJouguet point. approximately. and M.2. for which Figure 2. only some equations and certain limits on inequalities will require modification. Extensions of the Results to Arbitrary Systems 33 Numerical values of the parameters for ChapmanJouguet waves are plotted in Figure 2. When interpreted correctly.1 independent intensive parameters must be specified because p .46.3. The expressions which must be changed are those that involve c1 explicitly. Frozen versus equilibrium sound speeds It was shown in Section 2. and M o + = 5. However. u. u+ = 0. (Tm/To)+ = 12. N parameters besides s must be specified as constants in computing a p / d p to evaluate a’. A representative value of a is a = 30.2 Moregeneral analyses than those given in the preceding section show that the qualitative properties of the Hugoniot curve derived in Section 2. (Tm/To)= 7. a possible source of ambiguity for multicomponent systems is the fact that more than one sound speed a can be defined.2 that the downstream Mach number is unity for ChapmanJouguet waves.2 are also of general validity. this result applies to any combustible gas mixture.58.17 at the lower ChapmanJouguet point.3.2. and then we shall summarize most of the properties of arbitrary Hugoniot curves by means of a table. In accordance with our earlier discussion. Range of validity of the results of Section 2.1. these numbers are typical of experimentally observed detonations but unrealistic for deflagrations.3.5.4. then actually only N . since the single parameter c1 is not sufficient to characterize the behavior of arbitrary multicomponent mixtures.= 0.* Since there are N + 2 independent thermodynamic variables in an Ncomponent gas mixture.+ A frozen (constantcomposition) sound speed may be defined as * The subscript cc is suppressed in this section for the sake of clarity in notation. 2. Many quantitative results obtained in Section 2.= 0. and a are intensive properties that are independent of the size of the system.1.2 remain valid for all the systems that fall within the framework of the equations presented in Section 2. p.3. . EXTENSION OF THE RESULTS TO ARBITRARY SYSTEMS 2.6 as functions of the dimensionless heatrelease parameter a for y = 1. If s is the entropy per unit mass.3. We shall briefly discuss the most important complication that arises in attempting to extend our results. 2. while p .
e ( p . we find by differentiation that * This is a consequence of the fact that thermodynamically unstable equilibria are not attainable in natural processes. [lo]. at the ChapmanJouguet points Ixl For a small change along the general Hugoniot curve [equation (9)].3. = Y. for example. where y . it might be expected that the ChapmanJouguet tangency condition would lead to M = 1 with M defined in terms of a. in general. For example. e ( p ..+Further information on this topic may be found in [9] and in some of the references quoted therein (see also Section 6.. and the partial derivatives are taken in a closed system (of entropy S and volume V ) .. .s ) represents the equilibrium mass fraction of species i at the given p and s in a mixture at the specified overall atomic composition [equation (15)]. s) means that be set equal to its equilibrium function of p and s [which is determined by equation (12)] in computing the partial derivative.3.NJ 1/2. [which equals (a2U/2t2). [6].. which are comparable with the errors in the most accurate experimental measurements of this parameter (about one part in lo3). the sign of u. .. Using M = 1 to define the ChapmanJouguet point. the downstream gas velocity relative to the wave equals the equilibrium sound speed at the ChapmanJouguet points. other symbols are defined in Appendix A. Methods for computing ChapmanJouguet detonation velocities are given in [l]. with the M based on a/..uz is the same as the sign of the last factor. c . may lead to overestimates of the computed ChapmanJouguet detonationpropagation velocities. The subscript notation is to = y .instead of a. Since af 2 a. [2].4). Proof that = ae.. . Since the Hugoniot curve is an equilibrium curve. then straightforward thermodynamic calculations show that where t is the reaction progress variable defined in equation (A19). if only one chemical reaction can occur in the system under consideration. Y. 2. Since the factors preceding the brace on the righthand side of this relation are intrinsically positive. (V = internal energy)]. ( p . S K I = 1.* the problem is to determine which (if either) of these appears in the definition of M for which M = 1 at the ChapmanJouguet points. The rigorous analysis given in the following paragraph supports this guess.34 RankineHugoniot Relations (Y.2. and [ll]. The reasoning following equation (A20) implies that a small disturbance will cause the system to move away from the equilibrium point if this last factor is negative. = mass fraction of chemical species i). and an equilibrium sound speed may be defined as ae = [ ( a P / m s .
I.N ) .=. Since the entropy change along the Hugoniot curve [at Y. . . 2... which states that v. l / p o ) provided that. Extensions of the Results to Arbitrary Systems 35 The tangent to the Hugoniot curve is a straight line through the point ( p o ..3. at the ChapmanJouguet points as desired. 1. The quantities * Terms corresponding to pLd N . determined by the initial state of the mixture. Equations (8) and (36) show that where the last equality follows from equation (1). This analysis is illustrative of the type of reasoning required when the assumptions of Section 2. = d p Pa. . S m ) ( I = l ..Y. in equation (A5) do not appear in equation (38) because the condition of chemical equilibrium [equation (A1 6) with the equality sign] implies that these terms are zero in the present case.s w ) for i = 1. Equations (40) and (41) finally yield (42) v : = (aPa/aPm). w ( p w .. = 0 (39) at the ChapmanJouguet points along the Hugoniot curve. we obtain the expression 1 (37) d h . CD..2 are inapplicable. and min are fixed constants.. Hence. in equation (39... quantities identified by the subscripts +..4. xy=. max..3. in equation (36).1 summarizes many of the properties of Hugoniot curves.. Using equation (36) to eliminate d ( l / p a ) from equation (39. . = a?.=Y m ( p m . Summary of the properties of Hugoniot curves Table 2. .2. = Y. In the table. N ] is given by the thermodynamic relationship* equation (37) implies that ds..
.1. > 1 Not observed. Upper ChapmanJouguet point 1M. Point B = 1 Weak detonations Line BC u + < u < l Mo+ m < Mo < M. > 1 Weak leflagrations 0 Line DE p . p = 1 in most ex periments. forbidden by considerations of wave structure.) umin (Omin (Pm/PO) = (u. < 1 Often observed..) M.o) = (udae.TABLE 2.< p < l < Mo < Mo hi. 1 Strong detonations m Line AB < u < u+ > 0) Mo+ < Mo < < Seldom observed. Properties of Hugoniot Curves Downstream Mach number Remarks M m Section in Fig.> 1) Lower ChapmanJouguet point Point E M. 25 Pressure ratio Velocity and density ratios P= = (PO/P. Seldom observed. requires very special gas mixtures. Usually observed for waves propagating in tubes./uo) Propagation Mach number M o = (uo/a/. requires special experimental arrangement. = 1 Not observed. u = u(u. Strong leflagrations Line EF 11.
Hirschfelder. Scorah. Table 2. A. 68108. Strehlow. Dayton (March 1956). is based on the chemically frozen sound speed in the initial mixture. 606 (1962). Kompaneets. TheoryofDetonation. 8. Mullins and S. F. Zel’dovichand A.2. 6. J. 5. + REFERENCES 1. Scranton. Gross and A.are given by equations (31)(34) only under the restrictive conditions imposed in Section 2. Curtiss. “Calculation of Detonation Velocities in Gases.4172. S. Physik. 149155. A R S Journal 29. Wright Air Development Center. Fundamentals of Combustion. NASA SP273 (1971). 425 (1935). 9. 173 (1959). 2. Gordon. Chem. 8. S. R. 3. Penner. R. B. 797814. Edse. Y.” Tech. B. The limits for all of the inequalities except those in parentheses are intended to represent greatest lower bounds or least upper bounds. Chemistry Problems in Jet Propulsion. S. Zeleznik and S. New York: Wiley. C. R. 11. R. Penn. Penner. R. Becker. S. P. 4. Detonations. L. McBride. J. . Rept. J.1 should become selfexplanatory. S. Molecular Theory of Gases and Liquids. 321 (1922). New York: Pergamon Press. Elektrochem. 1954. No. 1957. Bird. Explosions. 1959. B.New York: Academic Press. J . A R S Journal32. Flammability and Ignition.258266. is based on the equilibrium sound speed in the final mixture. Oppenheim. 1960. : International Textbook. and R. 0. Phjx 3. With this information. 10. New York: Pergamon Press. and M .457 (1936). 54416. The quantity M . Z .References 37 with subscripts and . 1968. Gordon and B. F. 42. K. 7. Z .
nearly isobaric flame in which most of the reaction occurs in a narrow zone that can be approximated as a surface. we shall merely illustrate the analytical procedure by means of three examples (Sections 3. so many diffusionflame problems exist that we do not have enough space to consider them all. a diffusion flame may be defined as a nonpremixed. highspeed flow. quasisteady.CHAPTER 3 Diffusion Flames and Droplet Burning In the broadest sense. as we shall see. and highly turbulent motion. a diffusion flame may be defined as any flame in which the fuel and oxidizer initially are separated. The problem of droplet burning (Section 38 . Coupling functions usually are more useful for initially unmixed systems than for initially premixed systems because. and a fuel droplet burning in oxygen in a rocket engine. Therefore. Included here are processes involving unsteady flow. the chemical reaction rate is often of negligible importance in answering certain questions raised for diffusion flames. In a restricted sense. Even within our restricted definition.1 3. The broad definition encompasses a wide variety of processesfor example. Therefore.3) provides a convenient framework within which these problems can be studied. a forest fire. an aluminum sheet burning in a supersonic air stream. The couplingfunction formulation (Section 1. This is the class of flame problems that will be discussed in the present chapter. a pan of oil burning in air. it is unreasonable to attempt to discuss all of these processes from a unified viewpoint. a lighted candle. the term is synonymous with nonpremixed combustion. With this usage.3).
1. Definition of the problem The first successful detailed analysis of a diffusionflame problem was given by Burke and Schumann in 1928 [4]. .4 and 3. The BurkeSchumann problem is illustrated in Figure 3.5.1.3) is of importance in other contexts (for example.4. Monopropellant droplet burning.O Yo = Yo. 3.1. In a pedagogic approach.1. the relevance to diffusion flames dictates coverage here. Although study of these structures necessitates simultaneous consideration of both finiterate chemistry and diffusion processes and therefore pedagogically belongs after Chapter 4. For this reason.1 . THE FLAME AT THE MOUTH OF A TUBE IN A DUCT 3. Most of these topics are also treated in [1][3].5 after completing Chapter 5. the reactionsheet structure is analyzed in Section 3.O Underventilated f' Overventilated f u r FIGURE 3. treated in Section 3.3. = Yo. the reader may return to Sections 3.0 Y F = YF. also involves consideration of finiterate chemistry but is related to the burning of a fuel droplet in an oxidizing atmosphere (Section 3. The Flame at the Mouth of a Tube In a Duct 39 3. In recent years appreciable clarification of structures of the thin sheets at which reactions occur has been obtained through asymptotic methods. Diagram of the BurkeSchumann problem. liquidfueled engine combustion) and therefore will be considered in somewhat greater detail.3).1. Flame r I V / I Y .fuel (or oxidizer) issues from a cylindrical Z .
40 Diffusion Flames and Droplet Burning tube into a concentric cylindrical duct through which oxidizer (or fuel) is flowing. . pD = constant. Let us employ a cylindrical coordinate system ( z = distance along the axis. that is..3. (2) where e. Section 1. . Conditions at z = 0 will be identified by the subscript 0. A flame is established at the mouth of the tube and either extends to the wall of the duct (the underventilated case) or converges to the axis (the overventilated case). We shall denote the species in the inner and outer tubes by the subscripts F and 0. Assumptions In addition to all of the assumptions introduced in Section 1. The notation will be the same as in Chapter 1. v = ve.. respectively. The product pD is constant throughout the duct.laz + a r . respectively. ThAflame height. let the radii of the inner and outer tubes be a and b. that is. p . (4) 4. Burke and Schumann obtained fairly accurate predictions of the flame height and the shape of the flame for laminar flow. that is. (5) . The massaverage velocity is parallel to the axis of the tube everywhere.2) identifies the various product species [compare equation (142)]. and shall consider the onestep overall chemical reaction where the subscript i ( i = 1. . particularly. Axial diffusion is negligible in comparison with radial diffusion. the following approximations will be made. ar.2. 1. (3) 3. is a unit vector in the z direction.3.1). The total mass flux in the z direction is constant everywhere in the duct. that is. To investigate this problem. 3. N . pv = constant. defined as the axial distance from the mouth of t h m d k e point at which the flame reaches either the wall of the duct or the axis. 8 = azimuthal angle) and set the origin on the tube axis at the mouth of the inner tube (see Figure 3.2. 2.1. is required for the design of burners of this type. Y = radial distance from the axis.
= [~(a~~/a~ = )0. = 0 and Yo = Yo.o and Yo = 0 in the central pipe. By using equations (2)(5) in the definition given in equation (145). according to and a0 = .e. The boundary conditions for equation (9) are that p is bounded everywhere and that p = YF. because of the symmetry of the problem.=.)..o in the outer duct.Yo/W0v. The entire reaction occurs at a surface called the flame sheet.3.=b = [D(d&/dr)].1. l~=~ where e.1. Solution of the species conservation equation for the coupling function p The only differential equation that we shall be obliged to consider is equation (149) with where.3.1. . (10) (1 1) (12) p = . Equation (12) is a consequence of the physical requirement that neither fuel nor oxidizer can diffuse into the outer wall (at r = b): (YFvF'er). The way in which approximation 5 is used will be explained in Section 3.. or reaction sheet. is a unit vector in the r direction. 0. since YF = Y. we find that equation (149) becomes where the Laplacian has been written in cylindrical coordinates and terms involving a/ae have been set equal to zero. Equations (10) and (11) follow from equations (6)(8).ro.4.=b and = 0 ( y o v O . The Flame at the Mouth of a Tube in a Duct 41 5. while Y .O/wFvF at at at z = = 0. a < r < b. and the validity of the approximations will be discussed after the solution is obtained. z > 0. 0 5 r < a. 3.o/w and z r aglar = o = b.
Equation (22) provides the solution to equation (149) for in the present problem. OI{<C. then this surface must divide the flow field into two regions.cpo = 0). (22) where J . To solve equation (18). > qn. respectivelyand the cpn represent successive roots of the equation J .~ ~ q .42 Diffusion Flames and Droplet Burning It is convenient to introduce the dimensionless coordinates i" = r/b and y E zDfvb2 and to define the reduced parameters and and the reduced variable y PwFvF/yF. no oxidizer is present (Yo+ = 0). and dy/d( = 0 i" = 1. The flame shape and the flame height In order to obtain the shape of the flame and the flame height from equation (22). ( ( P . and therefore P+ = YF/WFVF (23) . we must introduce approximation 5. (19) (20) (21) at y = O .1. In one region (which will be identified by the subscript +).O.4. < ) e . equations (9)( 12) become y = l y=v at y=O. for example. By the method of separation of variables. at c<(<l. 3. In terms of these quantities. it can be shown (through the use of recurrence formulae and other known properties of Bessel functions of the first kind) that y = (1 + v)c'  v + 2(1 + v)c n = l (Pn c m  CJO((Pn)l2 J1(ccpn) J . subject to the boundary conditions given in equations (19)(21) and to the additional requirement that y be bounded. If the entire reaction is to occur at a flame sheet. is a straightforward mathematical problem. ( ( P )= 0 (with ordering convention (P. and J 1 are Bessel functions of the first kindof order 0 and 1. q > 0.
3.= 0).2 for c = and for two different values of v. The Flame at the Mouth of a Tube in a Duct 43 from equations (6)(8). Flame shapes for the BurkeSchumann problem (c = $). The flame height is obtained by solving equation (22) for y after setting 5 = 0 for overventilated flames or 5 = 1 for underventilated flames 0 . The shape of the surface obtained in this manner is shown in Figure 3.Y ol w o vo . This condition is compatible with the preceding inequalities only if P=O (25) at the flame sheet. in their respective regions. and therefore P = . Hence.I 0 in the oxidizer region.5 (Overventilated) FIGURE 3. . 2 v = 0.2. In view of the fact that the mass fractions YFand Yo are nonnegative. in the other region (identified by the subscript ). equation (23) shows that P + 2 0 in the fuel region. no fuel is present (YF. Since equation (22) implies that P is continuous in the interior of the duct [see equation (17)]. P+ = P at the flame surface. (24) Equations (23) and (24) actually imply that equation (22) gives the concentration fields for both the fuel and the oxidizer.1. setting y = 0 in equation (22) provides a relation between 4 and y that defines the locus of the flame. and equation (24) shows that P.
we obtain the rough approximation for the dimensionless flame height of an underventilated flame.44 Diffusion F l a w s and Droplet Burning (and. In view of equation (l). * These operations are necessary because if all of the reactants are to be consumed at the interface.* (158) with i = F by WFvF.vF = wo/Wo vo .yo+ Vo+) + Pw o vo ~ in the steady state.3. 3.41. have the same meaning as in Section 1. w o vo . The flamesheet approximation Since the flamesheet approximation is common to all of the problems considered in Sections 3.1. Neglecting all the terms except n = 1.5. . it is of interest to study the character of this approximation in greater detail. The flame shapes and flame heights obtained from equation (26) by Burke and Schumann are in fairly good agreement with experiments. Hence.13. as well as all of the other symbols. the differential form of this equation is Equation (27) states formally that fuel and oxidizer must flow into the flame sheet from opposite sides in stoichiometric proportions.4. dividing equation (158) with i = 0 by Wove. The first zero of J.(cp) is cpl = 3. we find that ~ p + yo+ (v. Since flame heights are generally large enough to cause the factor eqftq to decrease rapidly as n increases at these values of q.y r )in the notation of Section 1. considering the drastic nature of some of the approximations. wi d Y # 0 [that is.). y = 0 in either case). of course.83. When there is no fuel on one side of the surface (YFE 0) and no oxidizer on the other side (Yo+ = 0). it usually suffices to retain only the first few terms of the sum in equation (22) for this calculation.. The significance of the approximation can be addressed from the viewpoint of the general interface conservation condition given in equation (158). the agreement is surprisingly good.o f. Here the unit vector normal to the flame sheet (n. wi + wi S ( y . then lim. dividing equation and subtracting the second result from the first. which we shall now discuss. equation (143) implies wF/ W.
The Flame at the Mouth of a Tube in a Duct 45 Since equations (23)(25) imply that Y.vanish because FV. and the concentration gradients remain nonzero at the flame sheet. A velocity profile corresponding to fully developed pipe flow would be more reasonable.2). but it does not imply that the terms involving VF+ and Vo.D V F . Furthermore.6. since the heat release is not uniform in planes perpendicular to the axis of the tube.1. for a diffusion flame. some consideration has been given to influences of inlet velocity profiles [581. and this would invalidate assumption 2 in the approach stream. and the streamlines will diverge from the hot regions (compare Figure 5.(the convective terms) in equation (27) vanish. 0 [see equation (137)]. the density decrease will also cause the flow to expand. Assumptions 1 and 2 appear to constitute the most unrealistic approximations in the present analysis.1.1.4). The approach flow in the inner and outer pipes will be parallel to the axis (assumption 1) and will have constant density. thus causing the velocity to increase according to equation (3). the approximation usually is well justified.= 0. violating assumption 1. equation (27) reduces to the condition that the diffusion rates (which are now equal to the total rates) of transport of fuel and oxidizer to the flame sheet must be in stoichiometric proportions. The validity of the other approximations Let us finally consider the validity of assumptions 14 of Section 3. = 0 and Yo. they are best justified by the analytical simplifications that they produce. i = F + . Thus. and v . However. In the couplingfunction formulation. Assumption 3 has previously been shown to constitute a reasonably accurate approximation. the heat release in the flame increases the temperature and therefore decreases the density of the isobaric stream. but the initial velocity profile will not be uniform (unless special experimental precautions are taken) because the velocity must approach zero at the walls. This implies that the terms involving v. uncertainties in the temperature and composition at which p D is best evaluated arise from the difficulties associated with assumptions 1 and 2 and are particularly severe if the values of p D for fuel and oxidizer differ appreciably..2. 3. However. Therefore. Criteria for the validity of the flamesheet approximation may be developed by analyzing the structure of the sheet (see Section 3. there is no fuel or oxidizer present at the flame sheet. this condition is an automatic consequence of the continuity of the gradient of p at the flame. Assumptions 1 and 2 are mutually inconsistent in this problem. the flame produces nononedimensional flow. Assumption 4 is acceptable when the flame heigdt i$ large enough (in comparison with the tube diameter) .3. although uncertainties arise for strongly sooting flames. For calculation of flame shapes in the BurkeSchumann problem. = .
this can be shown to be true if the dimensionless parameter ub/D is sufficiently large. laminar. then surprisingly results in for conservation of fuel mass in axisymmetric flow. This estimate is not immediately applicable for deriving the correlations [9][l l] because they pertain to largescale regimes. Other refinements can also be introduced. in which D here must be viewed as an effective diffusivity that depends on v. If. Assumption 1 deserves special mention because buoyancy forces on the hotter. Fr = u’/(ag) (where g denotes the acceleration of gravity). namely. then dimensionally D a2/t and h fit. where A p is a representative density difference and p a mean density). C. giving D ii2/t.7. Roper. so that the characteristic transverse length for diffusion becomes an average value si.Imposition of a condition a%. and A. typically strongly turbulent.          . Fr 2 1). and D . instead. Smith. I7 gt A p / p g t (roughly). a. is a Pecletnumber (Pe) dependence that cannot be correlated with Fr and that N  holds in a “momentumcontrolled’’ regime that occurs for sufficiently large values of Frfor example. G. C & F 29. 227 (1977)l that one further aspect of the physics must be considered. for example. where I7 is an average velocity of the element. further complicating the algebra . roughly. however. different from a. Comments on the formulation and the analysis All the assumptions of Section 1. Assumption 4 can be eliminated at the expense of additional analytical complexity. then the flame heights are controlled by buoyancy. measures the relative importance of inertial and buoyant forces. A Froude number. Roper.46 Diffusion Flames and Droplet Burning for the transport of fuel and oxidant to the flame to occur principally through radial diffusion. then we obtain t Jh/g and h ga4/DZfrom the starting expressions for h and D. and Fr. The analysis leading to equation (26) corresponds to setting 6 equal to the exit velocity v. * If t is the time for a fuel element to travel from the burner port to the flame tip. Thus buoyancy usually plays a significant role in coflow laminar diffusion flames in the laboratory. h/a ua/D = Pe.3 underlie the formulation. buoyancycontrolled flames. Correlations of measured flame heights in the form h/a Fr”. In other configurations the cancellation does not occur. for a planar. it has been suggested by Roper [see F. For these situations. C & F 29. There is evidence that the latter estimate is better under typical laboratorv conditions and that buoyancy ceases to dominate only at burner dimensions below about 1 mm if v 10 cmjs (that is. where h is the flame height and 3 IIZ 5 *. the product pD may be assigned different values in the fuel and oxidizer regions. the reader may use these methods to derive h vaZ/Dfor the planar for D the ~)” buoyancy~ analog of the analysis underlying equation (22) but h ( v ~ ) ~ ~ ~ / ( g dominated flow. 3. The estimate might be thought to be usable directly for smallscale. a. Cunningham. Associated with the change in v is a change in flame width. For example.* The result of equation (26). C. If Fr is sufficiently small (for example. are available for b + co under buoyancyinfluenced conditions with negligible viscous forces [9][11].1. an estimate for domination by buoyancy is applied.219 (1977) and F. cancellation ofg from the estimate for h and therefore fortuitously produces a result in agreement with equation (26) concerning the dependence of h on v. less dense gas in the region of the flame often distort the shape of the flame sheet. at sufficiently large u. twodimensional slot burner of width a in the open atmosphere. Fr < A p / p . G.
However. where ji is a representative average value of the coefficient of viscosity) becomes large enough to cause the flow to be turbulent.4).for example. 4’C’ 0 obtained with co = ccp. a situation that often is encountered in practice). . there is a critical value of a above which the flow in the momentumcontrolled regime always is turbulent. if v 4 1. Many experiments and some analyses [13] apply to this limit. Re = pva/. a range that vanishes if a 2 (Re:ji2/gpAp)’13. Since increasing a increases Re and decreases Fr. then the ratio of inertial to viscous forces (characterized by a Reynolds number. The burning of a fuel jet issuing into an infinite oxidizing atmosphere may be viewed as the limit as b + 00 of the problem formulated here.cp. Turbulence occurs in most practical burners and introduces qualitative differences from the present predictions. The result. This flameheight behavior may be obtained from the present analysis by replacing D by a turbulent diffusivity that is assumed to be proportional to the product ua. The laminar analysis is justified best for relatively small laboratory flames.3. is the same representation that would be found by applying a Hankel transform to the directly formulated problem with b = a. The Flame at the Mouth of a Tube in a Duct 47 If u becomes too large. in terms of the known fuel or oxidizer fields. It may be observed that only the fuel and oxidizer concentration fields have been considered in finding the flame shape.1.2). they may be calculated a posteriori.ii.and productconcentration fields when the stated assumptions are adopted. (? + v)/(l + v) = J nu ~J.p/pa. By putting y = 0 and 5 = 0 in this formula. which becomes h/a = (ua/D)(l + v)/4v if hD/ua2 is large (that is.(o)Jo(w~/c)eW2 do. the flame fluctuates rapidly and on the average is thicker.) + nc + do. then the analysis that has been given here is justified roughtly for . The nature of the boundary conditions makes it unnecessary to study the temperature. If temperature or product concentrations are desired. Large flames often encounter turbulence already in the buoyancycontrolled regime and pass into the momentumcontrolled regime under fully turbulent conditions. If Re. and c(cp. we obtain an integral expression for the flame height. Temperatures at the flame sheet calculated in this way usually are too high (see Section 3. / 5u5 Re. with h/a approximately independent of o or a [12]. to derive the solution from equation (22) by investigating the limit in which c + 0 with </c and q/c2 held fixed.+ . Although the it is an interesting exercise problem may be formulated directly with b = a. there are many properties of turbulent diffusion flames that cannot be predicted well by a simple approach of this kind (see Section 10. by solving equation (149) for PT and Pi with a1 = aF. is the critical value of Re above which the flow becomes turbulent (a geometrydependent value perhaps on the order of 2000).
Specific examples include the oxidation of some solid fuels (for example. we shall briefly discuss a very simple example. for example. the reaction takes place only at interfaces and no flame appears in the gas. Present research is directed toward enlarging the class of soluble flow configurations. as illustrated in Figure 3. The nature of carbon combustion depends strongly on the temperature of the fuel. under certain circumstances. the walls of which are coated with carbon.6). fall within this category. Let us consider a rectangular duct (with sides of lengths a and b in the x and y directions. The entire field of heterogeneous catalysis (see Section B. analyzing more complex surface reactions. limit the range of utility of equation (28). At lower temperatures. as well as all heterogeneous processes whose basic mechanisms are treated in Section B. including effects of simple homogeneous reactions and exploring peculiarities of convectivediffusive interactions. processes of this type represent a degenerate kind of diffusion “flame” that is comparatively easy to analyze because the location of the “flame sheet” is known in advance.3. + 2co. The nature of carbon combustion Carbon combustion is an interesting illustration of gassolid reactions [24] and also is relevant to the practical subject of the combustion of coal [ZS].2. At temperatures above 4000K.2. following devolatilization.2.2.1 Definition of the problem There are many chemically reacting flow processes in which. THE OXIDATION OF CARBON AT THE WALLS OF A DUCT 3. sublimation of solid carbon becomes nonnegligible at atmospheric pressure. since later stages of coal combustion. 3. In one sense.2.48 Diffusion Flames and Droplet Burning 3. mainly involve carbon burning. For illustrative purposes. and through which oxygen is passed with velocity u (in the z direction). (28) We wish to determine the distribution of the mass fraction of oxygen ( Y ) through the tube when the mass fraction at the inlet (z = 0) is Y = Yo. oxygen . oxygen). Detailed studies of the burning of solid fuels often reveal complexities peculiar to the fuel. carbon) and the surface recombination of various dissociated species (for example. The oxygen will be assumed to react instantaneously at the walls according to the process 2 c + 0 . For carbon there are a number of complexities that affect the simplifications applicable in models and. and gasphase combustion processes may be important. Similar problems are treated in [14][23].4. respectively). the last four of which constitute fairly comprehensive reviews of the subject.
there are different structural types of carbon that have been studied. + 02.) at a finite rate. [33] seems popular. S. The StricklandConstable scheme [29].3. There are other alternatives (for example [30] and [31]). The Oxidation of Carbon at the Walls of a Duct 49 c Velocity=v IInss fraction of o2 = Yo FIGURE 3. Aside from possible catalytic influences of impurities in the carbon. S. The rate constants and activation energies for the elementary steps differ for the different schemes. Diagram of a rectangular duct with carbon oxidation occurring at its walls. followed by a minimum at a temperature a few hundred degrees above that of the maximum. a different finiterate reaction occurring subsequent to each type of adsorption. + O2 and the other to 2S. at temperatures below about 800K and CO at higher temperatures [26]. and react there. and S. may diffuse to the carbon surface. one corresponding to S. it involves two types of adsorption sites.3.2. and the other CO. but two types of equilibrium adsorption with of oxygen. The primary product of this surface reaction is CO. be adsorbed. A number of different theories have been qualitatively successful [27][31]. but data obtained by different authors differ. on which equilibrium adsorption and finiterate reaction of oxygen occur (at different rates). The surface chemistry responsible for the interchange between C 0 2 and CO as products is uncertain. One viable alternative (developed by Ubhayakar in unpublished work) involves only one type of adsorption site.. [32]. These include amorphous carbon (typically randomly oriented graphite crystals) . A number of complications lead to differences in experimental results. Each of the predictive schemes agrees with at least one set of experimental data. with the sites being interconvertible ( S l f S.. the overall rates obtained are not of the simple Arrhenius form but instead usually exhibit a maximum in the rate of carbon removal at a surface temperature between 1200K and 2500K. one of these reactions produces CO. but it is unclear that any of them are fundamentally correct. Thus equation (28) can be a good approximation only if the surface temperature of the carbon is between about 800K and 4000K (we shall see that there are additional restrictions).
[43]. An alternative to equation (28) is a model in which CO )O. Uncertainties remain concerning influences of gasphase chemistry [44]. composition. geometrical configuration. chemical kinetics in the gas can influence the carbonburning rate. + C + 2CO at the carbon surface [39]. molecules on carbon surfaces [42]. + 3. [36]. [34]. From this discussion it may be seen that. may occur [39]. An average of these rates would be relevant for most combustion processes. if the residence time is long enough [43]. The rate of oxidation of carbon by COz is so slow that this alternative seems reasonable only for fairly high surface temperatures and fairly low oxygen concentrations in the gas [41]. + CO. appreciably lower gas temperatures. Amorphous carbon often tends to be porous. The occurrence of this diffusioncontrolled regime is well established for carbon combustion [3 11. may react with CO in the gas to form CO. The following analysis will be restricted to this limit. nonreacting gas mixture and are not considering . rates of diffusion of oxygen into pores can influence burning rates of porous materials.2. 0 atoms are appreciably more reactive than 0. [40]. [44]. lowspeed flow. Pore diffusion is seldom ever significant for pyrolytic graphite. At sufficiently high temperaturesfor example. The following analysis applies in a particular regime. [45]. for laminar flows with relatively short residence times in small ducts with dimensions on the order of centimeters. the hypotheses of steady.3. which ceases to apply if the dimensions of the carbon materials become too small [39]. [37][39]. Equation (28) requires that residence times in the gas are short enough to prevent chemical depletion of 0. and assumptions 14 of Section 3. As is seen in Section B. if the reaction rate at the surface and the gas pressure are high enough. and oxygenrich atmospheres at pressures on the order of atmospheric. There are models that show how pore diffusion tends to reduce measured values of overall activation energies [26]. there are a variety of different regimes of carbon combustion. oxygen dissociation is unimportant. Analysis For simplicity. At lower temperatures. anisotropic carbon with distinctly different characteristics of base and edge planes).3 are not necessary here because we are studying a binary. typically with surface temperatures on the order of 2000K.1. The other assumptions in Section 1. from occurring before oxygen reaches the solid surface. Even when the reaction rate at the surface is effectively infinite. and gas velocity.2. above about 3500Ka substantial amount of dissociation of 0. depending on the conditions of temperature. at a flame sheet in the gas and CO. but 0. but oxidation rates differ appreciably on the base and edge surfaces [35]. we shall again employ assumptions 15 of Section 1.3.50 Diffusion Flames and Droplet Burning and pyrolytic graphite (a highdensity. [41]. then the burning rate is controlled by the rate of diffusion in the gas. [41]. pressure.4.
subject to the given boundary conditions.3. and that Y = 0 at x = 0. ) R ' ( . and steps 1 and 5 of Section B.4. It is also found that the most severe approximations of Section 3. x = a.1.)I.2). In general. equation (30) becomes Y = (:)2Yo sin(:)sin(y)exp[  ($ + (31) which shows that the initially uniform concentration profile assumes the shape of a half of a cycle of a sine wave and decays exponentially with the decay length . * Boundarylayer theory (see Chapter 12) can profitably be utilized to provide more accurate results in many of these problems. .* however. y = 0.2.2 (that is. The Oxidation of Carbon at the Walls of a Duct 51 the temperature field. the solution to equation (29).( + . Under the stated assumptions. The reader is referred to cited reviews for interesting fluidmechanical ramifications. assumptions 1 and 2) can be avoided in these relatively elementary problems [16]. In the Cartesian coordinate system shown in Figure 3. For recombination processes. and y = b (that is. as the specific rate constant for the surface reaction approaches infinity. . Systems with finite surface reaction rates have been studied in [19] and [20]. equation (144) is valid for CI = Y and with w = 0. the limit of an infinite surface rate constant is usually referred to as the case of a perfectly catalytic wall. that Y = Yo at z = 0.3 (and with assumptions 14 of Section 3. At large distances from the entrance to the duct ( z + co). at the walls of the duct). (30) Equation (30) gives the desired oxygen concentration field.)( . we shall not introduce such refinements in the present illustrative treatment. is found to be (2m m=O + 1 1)(2n x sin[ (2n + 1)ny ]exp{  [ + 1) 2m + 1 ( y ) 2 + ("b+l)2]~2(:)z}. ) Z ] .1.4 (the diffusion transport steps) become rate controlling. for example. the concentrations of the reactant on the surface and in the gas phase adjacent to the surface both approach zero. in the steady state. [23].+ By the method of separation of variables. this governing equation reduces to The boundary cdnditions for equation (29) are that Y is bounded. the boundary conditions for equation (29) become nonlinear in Y when the surface rates are finite. This last condition follows directly from an analysis in the spirit of Section B. [20].
may accumulate in a shell at a condensation sheet located outside a thin primary gasphase reaction sheet.3. The last of these processes may be enhanced by thermophoretic motion (see Section E. reactions tend to occur on the surface of a carbon particle during combustion. these complexities often are associated with the burning of solid fuels.2. followed by internal chemical reactions [48]. [46]. Here we shall consider one of the simplest situations.5) of fine particles away from the hottest reaction zone under the influence of the temperature gradient [SS]. To present analyses of these phenomena is beyond the scope of the discussion here. there are a number of possible modes of quasisteady burning. Many theoretical analyses of the various types of combustion processes have been published [55][62]. [53] and calculation of combustion oscillations associated with diffusion of oxidizing gas into the interior of the particle. then their condensation can influence the combustion mechanism.03. If reaction products are less volatile. [61] of different .3. The relative volatility of the fuel and its oxide has an important influence on the combustion mechanism of the particle [47]. may accumulate in a shell at a reaction sheet located at some distance from the surface of the particle. In the presence of product condensation. Relatively volatile fuels with nonvolatile combustion products. at the outset it is of interest to indicate some of the complexities that may arise in other situations. [41].1. this oxide is sufficiently nonvolatile for its liquid phase to play a role in the combustion of boron particles under many circumstances [54]. although boron is less volatile than B. THE BURNING OF A FUEL PARTICLE IN AN OXIDIZING ATMOSPHERE 3. the quasisteady. Because of the diversity of hightemperature oxidation mecanisms of solids [24]. producing gaseous products that flow and diffuse to infinity. Condensed products may accumulate on the surface of the particle.52 Diffusion Flames and Droplet Burning 3. Among the recent theoretical studies on the burning of carbon particles that go beyond the computation of steady rates of burning [39]. as indicated in the previous section. [52]. [48][53] are analyses of extinctions that occur in cool ambient atmospheres when the particle becomes sufficiently small [49]. investigations of conditions for ignition of a particle by a hot gas [Sl]. For example. Law’s models [60]. or may flow and diffuse to infinity in the form of fine particles. such as magnesium and aluminum. [Sl]. However. Carbon is an example of a nonvolatile solid fuel with volatilein fact gaseousreaction products. spherically symmetrical burning of a liquid fuel that vaporizes and reacts in a gasphase flame. Background and definition of the problem A variety of phenomena are exhibited by the burning of a spherical fuel particle in an infinite oxidizing atmosphere. therefore. practically always exhibit burning mechanisms influenced by product condensation.
(c) Falling droplet. internally supplied with fuelfor example. toward ascertaining the physical reason for the occurrence of one burning mode rather than another or toward inclusion of thermophoretic effects. spherically symmetrical burning of a liquid sphere. which have not yet been studied in the context of particle burning. unless the fuel particle is very small. their diameters do not attain quasisteady values during the testing time [67]. Liquid fuels with gaseous products do not share these complexities in their combustion mechanisms. The maximum test times achievable in these experiments are on the order of 2 s.3. Nevertheless. fuel has been t (a) (b) motion (. Diagrams of experiments on the burning of a fuel droplet in an oxidizing atmosphere. (b) Suspended droplet. The most ideal experiment in liquiddroplet combustion would be the steadystate. The absence of gravity is significant because buoyant forces distort the shape of the flame significantly. Although this experiment has not yet been implemented. insufficient for the complete burning of droplets with diameters on the order of 1 mm or larger.3. although spherical flames have been established. through a porous. spherical support in a gravityfree environment.) FIGURE 3. The Burning of a Fuel Particle in an Oxidizing Atmosphere 53 classes of combustion mechanisms provide a simplified but relatively accurate and useful framework within which many processes of combustion of metal particles can be placed and from which burning rates and flame locations may be obtained.4). three different experimental techniques have been devised for studying the burning of a fuel droplet in an oxidizing atmosphere (see Figure 3. droplet burning in freely falling chambers on the earlh has been studied [64][69]. Irrespective of whether a freely falling chamber is used. (a) Porous sphere. . for example. A conceptual design that most closely approximates this ideal involves the quasisteady burning of a free droplet in a laboratory in space [63]. In part (a). A method for performing this most ideal experiment has not yet been devised. there is room for much additional work. directed.4.
[93][103]. and the rate of decrease of the diameter of the droplet has been measured after ignition [65].df = K ( t . (b) yields accurate resultdor the diameter as a function of time. = riz d f o . and K . [106]. equations (32) and (33)] are found to be predicted with good accuracy by analyses that postulate spherical symmetry. Departures from the main assumptions will be discussed subsequent to the analysis. and comprehensive reviews are available [l09][115]. In experiment (a) it is observed that the total mass per second that must be supplied to the sphere in order to maintain steady burning is K'rf. In part (b). [Sl]. and [107]. Each of the techniques possesses certain intrinsic advantages over the others. and all the assumptions of Section 1. as was observed by Sreznevsky [lo41 and explained by Langmuir [l05].3. [72]. Later it will also be assumed for simplicity that the flow variables (in particular.3 will be adopted. 3. An assumption of Section 1. In experiments (b) and (c) it is found that after an initial unsteady period. (33) where df is the droplet diameter. which is just sufficient to maintain a liquid film on the surface of the sphere during burning [70][80]. for example. The theoretical model of the burning droplet is illustrated in Figure 3. Some of the objectives of theoretical analyses of droplet combustion have been to derive equations (32) and (33) and to predict observed flame 'shapes and temperature and concentration profiles. the square of the droplet diameter decreases linearly with time . these analyses are discussed in textbooks [l]. (32) where rf is the radius of the sphere and K' is independent of r l . df.3 that merits special mention for this problem at the outset is the hypothesis that steady conditions prevail. salient features other than the flame shape [for example. In part (c).5. which is called the evaporation constant. The chemical reaction will be assumed to be described by equation (l). [lOS]. temperature) are constant inside the droplet. [81][92]. and (c) is applicable to very small droplets. namely.54 Diffusion Flames and Droplet Burning supplied internally to a stationary porous sphere at a steady rate.2 Assumptions Although the shape of the reaction zone surrounding a burning droplet is not spherical in most experiments. is independent of t. t is time. freely falling fuel droplets have been ignited and observed during burning [82]. Successful analyses were first presented in [71].o is the value of df at t = t o . fuel droplets have been suspended on stationary quartz fibers.to). (a) is steady. There . The dsquare law in equation (33) also expresses the relationship between diameter and time for vaporizing droplets in the absence of combustion. Spherical symmetry is the primary simplifying hypothesis that will be adopted in the following discussion. small.
= 2r.5.3. is required when the liquid is above the critical point [I 19][124]. [SS]. The ignition period usually is predicted to occupy less than 20 % of the total burning time. in which steadystate conservation equations determine the mass per second leaving the droplet (h). there are two reasons for questioning the assumption. A theoretical study of these two effects is presented in [I 161. [117]. one which is basically unsteady. The theoretical predictions of [1191[124] are in qualitative agreement with experimental observations in [77]. . in agreement with a number of experimental observations [93]. at best quasisteady conditions can be attained. the initial unsteady ignition period may conceivably occupy the entire lifetime of the droplet. in experiments (b) and (c). Model of the burning of a fuel droplet in an oxidizing atmosphere is little doubt of the validity of this assumption for experiment (a). the combustion process resembles diffusion of a gas from a point or distributed source. First. which becomes appreciable only as the liquid approaches its critical point. since the size of the droplet is continually decreasing with time. Conditions of criticality may be attained in diesel and rocket combustion./dt I . the fractional error is somewhat less than the ratio of the gas density to pd. During the remainder of the burning time.a (pd * In highly supercritical regimes.) of the droplet. [llS]. The Burning of a Fuel Particle in an Oxidizing Atmosphere 55 cs Droplet FIGURE 3. the quasisteadystate approximation [employing equation (34)] leads to a slight overestimate of I dr. overall massconservation condition = liquid density = constant) determines the rate of change of diameter ( d . An entirely different treatment. Second.3. near critical conditions the strong dependence of the diffusion coefficient of the fuel on temperature can aid in establishing an approach to quasisteady conditions in the fuelrich zone for situations under which quasisteadiness otherwise would not be expected.after which the simple. and [92] made near the critical point. however.
and (35) can be used to express the evaporation constant K in terms of the proportionality factor K’. (37) where r is the radial coordinate and u is the radial velocity. Analysis predicting the burning rate In a onedimensional. 3. d: will vary linearly with t). (33). In terms of the dimensionless independent variable equation (39) becomes the solution to which is . namely. can be written as ri? = 4nr2pu = constant.3. ri? rJ. K =  (35) 2m/Xpdrl = 2K’/~pd: (36) Equation (36) determines K from the results of the following steadystate analysis. which explicitly shows that if equation (32) is satisfied (that is. the overall continuity equation. Equations (32). The burning rate rcZ is found to appear as an eigenvalue of equation (149) and will be obtained by applying equation (149) to PT with c( f C I ~ .3.56 Diffusion Flames and Droplet Burning Equation (34) can be written in the form d ( d f ) / d t = 2m/Xpdr[. equation (123). a steadystate theory that leads to equation (32) automatically yields equation ( 3 3 ) through the quasisteady hypothesis. spherically symmetrical system. we find that equations (1 45) and (149) give which reduces to dr dr (39) when use is made of equation (37). then equation (33) will also be satisfied (that is. Therefore. Employing the expression for the divergence in spherical coordinates.
then equations (137) and (162) give [POYO . (147). oxidizer. and product species. . and employ the massconservation conditions. The coupling between the temperature and oxidizer concentration profiles is determined by the function PT c(T . that is. N2 (43) where the standard heat of reaction at temperature T o is 4 ' h i WFVF + h$ WoVo  hywvi i= 1 (44) [see equations (146).3. and (l)].refer to conditions outside and inside the droplet.3. Equations (161) and (162) provide conditions on T and Yo that must be satisfied at the surface of the droplet. [4nr2A(dT/dr)] + = riZL. dT/qo + Yo/Wo~o. which must be discussed separately for each coupling function. which merely states that the net mass flow rate of oxidizer at the surface of the droplet is zero.= 0).~10= 6 C . Using equations (139) and (40) in equation * Radiative heat transfer is not negligible for liquids with high boiling points or for opaque droplets with highly luminous flames.6).= 0.pWdYo/dr)I+ = 0 7 (45) since there is no oxidizer inside the droplet (that is. 1 > (46) where use has been made of equations (37) and (40) and the subscript 1 identifies conditions in the gas at the droplet surface.. If the subscripts + and .yo. for the fuel. equation (162). The analysis requires modification for these systems (see Section 3. Equation (45). (47) where L is the heat of vaporization per unit mass of the fuel at temperature ?. yields the boundary condition (GO/dO. the amount of heat conducted into the droplet must be just sufficient to vaporize the fuel leaving the droplet. Yo. Equation (42) is the solution for the profile of any coupling function p. In order to obtain equation (47) formally from equation (161). The Burning of a Fuel Particle in an Oxidizing Atmosphere > .= 0 and Vi. one must neglect the radiation* and kineticenergy terms. use ( V T ) .3. = . respectively. 51 where A and B are constants. It is clear on physical grounds that when the temperature of the droplet is constant and uniform. The values of A and B depend on the boundary conditions. The corresponding constants A T and BT in equation (42) with p = PT are determined by the boundary conditions for the temperature and for the mass fraction of oxidizer. The explicit derivation of the interface condition on T from equation (161) is somewhat more lengthy.
f) + BTer'. thus. equation (43) implies that N i= 1 1yi. appear in equation (48).m < = 0 [see equation (40)]. where (52) Evaluating the same equation at we find PT. utilizing equation (50) in equation (54) shows that * If this convention is not adopted. m lT?cp. equation (49) shows that where a dimensionless burning rate is [see equation (40)]. co. m  1) (55) Hence.1 (49) In view of equation (42). L Yo. for convenience.* we obtain the boundary condition Equations (46) and (48) determine the gracfient of PT [equation (43)] at the surface of the droplet in the gas in terms of L and Yo. When equation (42) is applied to PT and evaluated at r = r l . (53) = AT + BT. wo voyo. it is found that BT = ( P T .. and taking the arbitrary standard reference temperature to be T o = 7.I . m  PT. (54) In view of the meaning of cp dT (see Section 1. An expression for the burning rate can now be obtained from equation (42).  PT.. 4O i d~ PT.58 Diffusion Flames and Droplet Burning (47). By eliminating A T between equations (52) and (53). we obtain PT. then additional terms involving (d&'dr).I =AT Y = + B.e<'. .f = + (Yo.3) and our convention T o = 7.
require more detailed discussion.l are found to be less than 1OP2. Yo. Discussion of the burningrate formula In many respects. since nearly all of the dropletburning studies it is assumed that Yo. at temperature 7. in dimensional notation.3.4). values of Yo. in the quasisteady state depends on the nature of the surface gasification process. 3. The Burning of a Fuel Particle in an Oxidizing Atmosphere 59 The fact that equation (56) is an expression for the burning rate can be seen from equation (5 1). The specific heat at constant pressure for each pure species i(cPJ the heat of vaporization per unit mass for the fuel at temperature T.Jn the flamesheet approximation. The value of 7.4. The remaining parameters.(L)and the quantity 4O/WOvo. in general. equation (57) is of more general validity than the burning rate expressions usually quoted because arbitrary temperature dependences of (pD) and c p are permitted in equation (57). However. The value of Yo.l. .4. Yo. and the flamesheet approximation has not been introduced.. and Yo = 0 inside this surface (hence. which is the heat of reaction f oxidizer consumed. [126]. In dropletburning experiments. The surface process may conceivably be a rate x.3. all the oxidizer is consumed instantaneously at the spherical surface where the reaction rate (per unit volume) is infinite. through numerical integration of the conservation equations or through asymptotic analyses of the type indicated in Section 3. to prescribed values. In most = 0. in par= 0 except very ticular. and the factor preceding the logarithm. YF.3. forgaseous reactants per unit mass o are all fundamental thermodynamic properties of the system and therefore .! = 0). 7. is no fuel in the ambient atmosphere). = 0 (there are presumed to be known. and an experimenter who controls the for the atmosphere in which the droplet burns can adjust the quantities . oxidizer and products and the temperature T. The validity of this assumption has been investigated by numerical integrations [lo71 and by asymptotic expansions for small values of the ratio of a reaction time to a residence time [125]. equation (57) will be useful for computing rit only if all quantities appearing on the righthand side of this equation are known.l oxidizer is expected to be consumed before diffusing to the surface of the droplet. It is generally quite accurate to put Yo. thus.l near extinction conditions (Section 3. depends on the reaction rate and can be determined.
aiding maintenance of spherical symmetry. for example. is not immediately apparent. and the factor preceding the logarithm becomes 4npDrI.go). is determined by the thermodynamic (phase equilibrium) condition that the partial pressure of the fuel at the surface of the droplet must be equal to the equilibrium vapor pressure of the fuel. and 7. and for many systems. except for fuels that undergo surface or condensedphase chemical reactions. Equation (58) shows that the principal dependence of m on the properties of the fluid is expressed by ri2 A/cp.’it can be shown from equation (57) that for small changes in r l . such as depolymerization of polymeric fuels. Therefore. When p D is independent of r in the vicinity of r = rl. riz rJ(1 + arl). Since the factor preceding the logarithm in equation (57) is influenced by the dependence of p D on r. that riz depends logarithmically on the heat of reaction and the heat of vaporization and that riz is insensitive to the pressure and is not strongly influenced by the temperature.i= c p = constant is reasonably accurate. greater than 10 p in diameter) that effects of surface tension on phase equilibria are negligible. Thus the surfacetensioneffects are purely mechanical.3 for a more complete discussion of surface conditions. The quantity p D can be replaced by A/cp. Therefore. would be found to depend explicitly on yiz. However. since Le = 1 [see equation (139)]. solving for /IF = uF . See Section 7..  * The droplets under consideration throughout Section 3. the heat flux to the droplet surface is large enough to lead to equilibrium droplet surface temperatures that are only slightly below the boiling temperature of the liquid (see.4 would be required in order to determine 7.. where a = constant and usually lurll 4 1. in which case an analysis analogous to that given in Section B. 7. the approximation cp. showing that riz r l .* The application of this equilibrium condition entails obtaining the coupling between the fuel and oxidizer concentration profiles (i. 7. The results of the preceding discussion can be summarized by means of the approximate formula   This expression agrees with the empirical relationship given in equations (32) and (33) and can be used in equation (36) to determine the evaporation constant. = % (% E boiling point of the fuel) is a good approximation. However. These approximations will be considered further in subsequent discussion.e. predicted by equation (57). Often p D = constant is an acceptable approximation.3 are so large (for example. the rates at which molecules of fuel enter and leave the liquid are sufficiently i commonly rapid to maintain surface equilibrium for the low values of t encountered in droplet burning. unless the heat of reaction is extremely low or the fuel is extremely nonvolatile. . the variation of m with T I . [116]).60 Diffusion Flames and Droplet Burning process. Experimental observations of the burning of single droplets small enough to exhibit this type of influence of surface tension would be difficult to obtain and have not been reported.
3. From equation (61). = 0 and ctT = PT for r < rc.p F and aT = PT = PT + / j F for r > r c . The Burning of a Fuel Particle in an Oxidizing Atmosphere 61 These results are in agreement with experiment. 3.1.m w F v F / w O v O ~ YF. which can be combined with equation (56) to show that + .5. = 0. In many respects curves in Figure 3. Predictions of other characteristics of burning droplets Since equation (42) is valid for any coupling function B. = 0 and Yo. in order to obtain explicit temperature and concentration profiles and to locate the reaction zone. then ctF = P F . by considering /3F c(F  =  YF/WFVF + Y~/Wovo. the position of the flame is determined by B F = O. These observations serve to determine the fuel and oxidizer massfraction profiles and the temperature profile through equation (42).PD(dYF/dr)l + = (Pv>+ it can be shown that the mass fraction of fuel in the gas at the surface of the droplet is L[l + yO. are to be avoided. cx0 = . 7 (59) and by using equations (45) and (56) and the surface conservation condition for fuel.~/wOvO)  provided that YF. For example. a. other results can be derived without explicitly invoking the flamesheet approximation. (61) just as in Section 3.6 are in qualitative agreement with the results of numerical integrations for reasonable chemical reaction rates [1071and with experimental results [SO].3. while aF = 0.idT) +(~OyO. Representative profiles are shown in Figure 3.i = 1 > (60) + i = 1 ( x . [ P v y F . m JT:cp. it can be shown that the value of 5 at the flame sheet is determined by e<c = I Yo. If quantities at the flame surface are identified by the subscript c. where Y.3.6. the flamesheet approximation must generally be employed as a first approximation if numerical integrations. In the flamesheet approximation. However. denotes the total mass fraction of products. Theoretical and experimental values of the e4aporation constant K are roughly of the order of lo* cmz/s. WFvF/wo~o.
(r. First. Because of the effects of natural convection in dropletburning experiments. qualitatively representing the ratio of an impetus for interphase transfer to a resistance opposing the transfer. and. the sum of the heat released per unit mass of oxidizer consumed and the difference in the thermal enthalpy per unit mass between the ambient gas and the gas at the surface of the droplet comprises the impetus for transfer. [127]) lead to the logarithmic term in equation (57) being identified as ln(1 + B).2 can modify the burning rate of a droplet. Hence.6. it is of interest to write the results that have been obtained in a simplified notation.62 Diffusion Flames and Droplet Burning 4 FIGURE 3. empirically. in equation (58) the logarithmic term is ln(1 + B).3. when Yo. the value of the flame radius r. Further realities of droplet burning A number of phenomena excluded by the assumptions given in Section 3. in fact.3. This may be accomplished by introducing the transfer number B. Generalizations in the definition of B (for example. For a vaporizing droplet.r l ) is more nearly independent of rl than is rc/rl for supported droplets with diameters on the order of a millimeter or more. Temperature and massfraction profiles for a burning fuel droplet in the flamesheet approximation. while the heat of vaporization L per unit mass of fuel vaporized is the resistance. .2. such as for the problem discussed in Section 3. 3. in droplet burning the factor ln(1 + B) always appears in the expression for the burning rate. predicted by equation ( 6 2 ) is larger than that typically observed by roughly a factor of 2. These phenomena ofen are too complex to be included in accurate theoretical burningrate analyses and necessitate semiempirical methods in deriving burningrate formulas. From the viewpoint of energetics. The most important of these additional phenomena will now be considered.l = 0 and pD = constant. The introduction of B can be useful not only in droplet burning but also for all problems involving interphase transfer. gasphase .6.
the mole fraction X F . and &. then equation (A23) may be employed with p representing the partial pressure of the fuel vapor. times the specified hydrostatic pressure.YF. .I = YF. = ln(1 + B)/ln(l + v).3. where Q is the additional energy per unit mass of fuel needed for gasification. as is readily verified by a suitably modified energy balance [compare equation (47). the product of the stoichiometric mass ratio of fuel to oxidizer with the oxidizer mass fraction at infinity.YF. or wetbulb temperature. for example]. then in view of equation (E5 l). For suspended or free droplets that are not initially at the boiling where t)/h. + (63) and it is found from equation (60) that the mass fraction of fuel at the droplet surface is ( B . the quantity L in the denominator must then be replaced by L + Q. The Burning of a Fuel Particle in an Oxidizing Atmosphere 63 diffusion of the fuel rather than heat conduction is the main driving force for d .value of 7. If itis of interest to account for the difference between 7.v)/(l B). Substitution of equation (58) into equation (36) gives YF. In the application to the burning of a porous sphere. and YF.)/( YF. where YF. The. obtained in this manner is a wetbulb temperature somewhat below &. . this causes equation (58) to become nonlinear in viz and to cease to possess a solution if rl is too large.G).m . this effectively decreases Q and may cause it to become negative in extreme circumstances. In the definition of B. *). X F .then the additional requirement in thermal enthalpy is Q = c. Then equation (62) for the flamestandoff ratio becomes dJd. vaporization. Under these conditions gasification requires an energy L + Q instead of L. X F .I is given by equation (64). If the specific heat of the liquid fuel has the constant value cI and the feed temperature is G. A practical concern with poroussphere experiments is that a significant amount of energy may be transferred to the liquid fuel from the flame that bathes the fine feed line [90] .If the fuel surface is at a sufficiently high temperature for the radiant energy lost from it to be of significance. an additional term u~T~/(viz/47tr~) appears in Q . point. usually the fuel is not fed into the sphere at its boiling point.I may be obtained from equation (ll2)namely.3. To write simple formulas it is useful to define v = (vFWF/vo Wo)Yo.(7. and a better definition of B is B = ( YF. F/WFwhere W is the average molecular weight (I7W3'. there is a contribution to Q given by ( ~ ) 7 t r ~ p d c l ( d ~ / d [ I : = . .dis the mass fraction of fuel deep in the interior of the droplet (usually unity).I = (64) (65) K = (8/Pd)(&)W + B ) for the evaporation constant. for use in equations (63)(65).
If radiant transfer from the flame or from hot * Further considerations may be found in a more recent publication.219 (1983).64 Diffusion Flames and Droplet Burning is the average temperature of the liquid. in addition to the timedependent outer zone. Matalon and C. Except under somewhat special conditions [133]for example. equation (63) usually predicts relatively small values of d J d . for burning in dilute atmospheres. For burning in oxidizerrich atmospheres. then this outer zone is swept away from the droplet relatively quickly.3. during which d. see Section 3. the flame may lie in the outer zone. during which there is a quasisteady inner zone and a timeClependent outer zone. This period is usually relatively short. Typically. to increase with time is in agreement with experiments [67] on the spherically symmetrical burning of free droplets in the absence of gravity. there must be an outer region in which quasisteady conditions are not attained.* The correction to K just cited remains consistent with the dsquare law for the total burning time.2. A correction factor to equation (65).4). such that the flame sheet lies in the quasisteady zone. there is an initial period of unsteady behavior throughout the gas. However. The predicted tendency for dJd. However. where pm is the density of the ambient atmosphere. It is evident physically that since a burning droplet exists only for a finite amount of time. causing d. and effects of gasphase unsteadiness on d. increases with time [135]. This contribution decreases with increasing time at a rate that increases as the thermal conductivity of the liquid increases. subsequent to liquid heatup [135]. approaches infinity (unless the flame is first extinguished. and the flame sheet is initially established closer to the droplet surface than a quasisteady analysis such as equation (63) would predict. There have been a number of studies of the heatup process [I 161.)ln(l + B)]'I2. to increase with time and equation (63) to be inapplicable. K. for fuels with high boiling pointsthe heatup typically occurs sufficiently rapidly to be unimportant during most of the burning. are relatively small. the gasphase unsteadiness may be significant [134][136]. Law. as the droplet disappears. [128][130]. and quasisteady conditions may prevail throughout the region of importance during most of the burning history. when ignition occurs there is relatively little fuel vapor in the gas./d. accounting for aspects of gasphase unsteadiness [135] is 1 + [(2/n)(pm/p. M. If the droplet is moving with respect to the gas. C & F 50./d. under suitable circumstances it achieves a quasisteady value given by equation (63) during most of the burning history. The possible occurrence of unsteady conditions in the gas phase for free droplets has also received further consideration [1341[136]. but there always is a short period. as was indicated in Section 3. Theory predicts that d. Thus. and the main influence of gasphase unsteadiness on the burning time and on the flamesheet location typically arises from the later period. . in spherically symmetrical burning. including consideration of influences of convective motions of the liquid within the droplet [131][133].
a Mach number) just outside the flame sheet. questions remain concerning these effects. A few relevant principles have been identified and deserve to be discussed here. however.5. [139][ 1421. a dlaw.. and in view of equation (34). (132) and (133) become good starting equations for an analysis. The Burning of a Fuel Particle in an Oxidizing Atmosphere 65 surroundings. which would contradict the original hypothesis of constant p . Since Figure 3. say q R . especially if the flamesheet approximation is adopted. Effects of variable properties and of Le # 1 are significant in that they can introduce uncertainties in excess of a factor of 2 in K and in dJd. The analysis that has been given here relies on the assumption that the Lewis number is unity [equation (139)]. since then the flux fractions t i remain constant on each side of the + . becomes a dominant source of energy input to the droplet. the paradox concerns a small effect. Equations (131). The paradox has been resolved independently by Sepri [137] and by Cooper and Clarke [138]. This is evident from the fact that the radiant energy input per unit area at the droplet surface. then the dsquare law becomes a dlaw for the usual situation in which the Planckmean absorption length of the gas (Section E.. it follows from the idealgas law with constant pressure p that the density gradient is discontinuous. which may.6 shows that the temperature gradient is discontinuous at the flame sheet.3. An interesting paradox is posed by considering momentum conservation in connection with the flamesheet approximation. If Le # 1. The phenomenon provides one mechanism for combustion to affect the fluid dynamics. ./dt = (qR/pd)/(L+ Q).2qR = yiz(L Q ) in a radiationdominated situation. this gives dr. therefore.is independent of d.. The dsquare law may be viewed as arising from the fact that the conductive input flux. gasphase [A(dT/dr)]+ in equation (47). is inversely proportional to Y gradients steepen as the droplet size decreases. and from equation (120) it is evident that there is nothing in momentum conservation that can balance this divergence. or r. A simplified energy balance for the droplet is 47cr.2) is large compared with the droplet diameter. p experiences a discontinuity in its gradient at the flame sheet and then decreases by a fractional amount of order M2 through the adjustment region. From equation (116) we see that this introduces a 6function in the divergence of the shearstress tensor at the flame sheet. in turbulent flows). in fact. The more general theories also are more complex and therefore offer greater resistance to attempts at extracting generally valid conclusions. such as hot chamber walls. who showed that there is an adjustment region with thickness of order rcM2 (where M is. Although there have been a number of theoretical studies that consider Le # 1 and that take into account variable transport properties and specific heats [106]. according to equation (37) the velocity gradient is discontinuous. except possibly a discontinuity in p at the flame sheet. Since M is small.3. then a formulation on the basis of coupling functions is less convenient. influence fluid flows appreciably in configurations that are more complex (for example.
6 times that of air [90]. When Le # 1 it is found [140] that the main transport properties affecting the burning rate are (1) the thermal conductivity in the region between the flame and the droplet surface and (2) the diffusion coefficient of oxidizer in the region outside the flame. reasonable agreement with experiment is achieved by use of the c. For burning in air. may be employed for evaluating the burning rate through use of equation (65). then in this formula the theoretical undissociated flame temperature is best used in calculating B . heat conduction is a dominant process and fuel diffusion is selfadjusting. It has been reasoned [90] that since processes occurring between the flame and the droplet surface are of greatest im. would suggest that an average specific heat of all species present in the gas is more relevant. [114]. the c the fuel vapors (and possibly their gasphase pyrolysis products). [142] with good results. and heat conduction merely provides a loss mechanism that decreases (or increases) the flame temperature.the energy of . In this respect the couplingfunction theory can tend to be misleading. if a formula for B is employed that explicitly contains a flame temperature instead of a standard heat of reaction qo. for example.4 times that of fuel . that of heat conduction between the flame and the droplet. in equation (65) should be that for portance in controlling burning. the A in equation (65) may be taken as an average of that of the fuel vapor and air. a procedure that clearly burning in airto employ c is entirely unjustified if Le # 1. for . [142]. O n the other hand. As a first approximation. In calculating the flame temperature for this evaluation. equation (58). for fuel if suitable procedures are adopted for evaluation of transport properties [90]. outside oxidizer diffusion is dominant.i combustion. The thermal conductivity outside the flame is of secondary importance and the diffusion coefficient of the fuel vapor is practically irrelevant. have in fact been employed in a number of the theories [106]. Since A and c perature must be selected at which these properties are to be evaluated for use in equation (65). Although the resulting value of c is larger by a factor of 2 than that obtained from other prescriptions. In contrast to the relevant heat capacity. since the unrealistically high temperatures often obtained when dissociation is neglected are clearly irrelevant to realistic property evaluations. equilibrium dissociation should be taken into account. 0. evident from equation (133) and also physically clear. a templus 0. These influences can be reasoned physically. This conclusion would be correct only for Le = 1 and would tempt onefor example..66 Diffusion Flames and Droplet Burning flame. vary with temperature. is that a species i must be transto play any role in the quasisteady ported (ti# 0) for its specific heat cp. The arithmetic mean of the flame temperature and the boiling temperature may be used for this purpose [90]. These equations. the only . the largest effect. A result. depending on whether the Lewis number outside is greater (or less) than unity. typically species transported in that region. or simplifications thereof. all species present contribute to the thermal conductivity. Inside the flame. for nitrogen in A/c.
the ambient gas composition. is the gas velocity far upstream. (71453. in . Pr = 1 may be employed in equation (67). and ideas for corrections extending to large convective velocities have been considered [72]. where KO is given by equation (65). Forced and free convection introduce departures from spherical symmetry that modify burning rates and flame shapes.276Re'i2Pr1i3) (67) for droplet burning in forced convection. Thorough theoretical analyses for nonspherical situations are not yet available because of the complexity of the problem. [144]. Pr and the properties p and p in Re may be evaluated at average conditions [148]. The prescriptions that have now been given enable equations (63) and (65) to be employed with reasonable accuracy for spherically symmetrical burning. Since heat transfer is usually more important than mass transfer in droplet burning. it seems reasonable [148] to replace Sc by the Prandtl number Pr (see Appendix E) in the Frossling formula to obtain K = KO(l + 0. In equation (67). results adapted from the early analysis of Frossling [1461. Gr = (SP2&/P2)(AT/T). With the prescription that has just been given for evaluation of the average temperature T and the average properties p and p. the range of practical interest. U P (66) where p and p are the density and viscosity of the gas and u. With these prescriptions. Effects of forced convection at small convection velocities have been studied [143]. equation (68) may be used with AT taken as the difference between the flame temperature (with dissociation) and the ambient temperature to calculate K .3. which may be taken to be the ambient pressure. Burningrate corrections for free convection depend on the Grashof number. [96]. equation (67) provides reasonable agreement with data on flameenveloped drops for Reynolds numbers from 1 to 2000.276 Re1/2Sc1/3. and the arithmetic mean of the ambient temperature and the flame temperature (with dissociation taken into account) [90]. [147] appear to remain as good as any currently available for forcedconvection corrections.3. Nevertheless. Here Re = PU. Frossling studied vaporizing droplets without combustion and where Sc is the Schmidt number found that K / K o = 1 + 0. (68) where g denotes the acceleration of gravity and AT denotes a temperature difference. Re is the Reynolds number. relative to the droplet. [113]. (Appendix E). and KO is the value of K for evaporation in the same atmosphere without forced convection. Within the accuracy of the correlation thereby obtained. The Burning of a Fuel Particle in an Oxidizing Atmosphere 61 dissociation is largely recovered in the cooler parts of the flame and therefore is available for driving vaporization.
In the combustion of droplets composed of binary or multicomponent miscible liquids. through which species are transported at relative rates proportional to their relative bulk concentrations in the interior. a topic considered in the following section. At sufficiently high forcedconvective velocities. the fuel constituents need not enter the gas in the same proportion in which they are present in the liquid. A pure fuel undergoing this type of fuel pyrolysis in the liquid phase gradually may become a multicomponent liquid during burning. Below or ahead of the droplet. and. good theoretical bases for specific functional forms are lacking. For hydrocarbon droplets burning in air. For fuels with high boiling points. Formulas that differ somewhat from equation (69) may be found in the literature. [149]: K = KO(l + 0. The phenomena of flame extinction are associated with finiterate gasphase chemistry. based. The forced convection may induce circulation within the droplet [131][133]. with a diffusion layer at the surface. opens and emits soot. under suitable conditions. (69) where KO is obtained from equation (65). the elongated downstream flame yellows. most practical fuels. it is increased significantly. on heattransfer analogies . but accurate theoretical results are unavailable. the distance to the flame is reduced. from the empirical formula [90]. wake burning may ensue. If rapid mixing is maintained throughout the droplet. Development of the diffusion layer is favored when liquidphase diffusion times are long compared with the droplet lifetime. for example. finiterate chemistry may occur in the liquid phase. Often the burning of initially multicomponent liquids is of interest . the flame ahead of the droplet is extinguished. a quasisteady concentration field may be established within the liquid. After vaporization of the lighter species from a thin outer layer.68 Diffusion Flames and Droplet Burning the presence of free convection with Gr < 1. are not pure compounds. Estimates are available of the distance to the flame (for example. [65] and [66]). the surface temperature of the droplet consistent with interphase equilibrium may exceed the boiling point of the . Under these conditions. Liquidphase diffusion coefficients typically are sufficiently small that thorough mixing of the liquid is not anticipated theoretically. which in turn is extinguished at a higher velocity. Fluidmechanical details may become complex. for example. A few correlations for corrections to K caused by finiterate combustion are available [90].s2). Free or forced convection strongly influences the flame shape. with the lighter (more volatile) fuels vaporizing first and the heavier ends later. then the burning may involve a batch distillation process.533Gr0. and above or behind. Just prior to extinction there is a measurable decrease in the burning rate [90].
[ISO]. * The nucleation temperature. In the latter case droplet vaporization in a hot environment without combustion is relevant.6 are idealizations. [SO]. .1. of course. This interior temperature may exceed a nucleation temperature for the interior mixture*. flame sheets may surround either each individual droplet or entire clouds. Disruptive burning has been observed experimentally [91]. Further ideas on cloud combustion appear in Section 11. Droplet interactions can be complex and can produce increases or decreases in burning or vaporization rates. both experimental [98]. Approaches to structure questions Concentration and temperature profiles like those shown in Figure 3. Depending on the fuel. [156][161]. [loll[103] for both miscible fuels and emulsified fuels. This process. the possibility of disruptive vaporization of multicomponent droplets). and its description on the basis of a nucleation temperature is a simplification. Configurations studied include porous spheres [SO] and cylinders [1633. In recent years a number of profile measurements have been made in laminar diffusion flames (for instance. since the concentration of the lowboiling constituent has been depleted to a large extent at the surface. Structure of the Flame Sheet 69 interior fuel mixture. and on the overall stoichiometry. STRUCTURE OF THE FLAME SHEET 3.7.4. Homogeneous nucleation temperatures are substantially above the boiling point. [1591for example. shares many attributes of droplet combustion (for example. 3. Since practical combustors burn clouds of droplets. [162][168]).4. for example. The topic has been considered to be of practical importance because the performance of combustors can be enhanced in various ways through disruptive burning [91].3. depending on the circumstances [1511. is the temperature at which bubbles spontaneously appear in the liquid. Since thermal diffusivities of liquids are typically much higher than their diffusion coefficients. heterogeneous nucleationaided. Theoretical models for predicting the Occurrence of the phenomenon have been developed [114]. [loo]. conduction causes the interior temperature of the droplet to increase toward the boiling point of the heavier constituent. Bubble nucleation is a rate process. which exceeds the boiling point of the species. on the convective environment. interactions between quasisteadily burning droplets have been subjected to continuing study. thereby inducing a disruptive burning process in which internal vaporization of light fuel causes the liquid to expand rapidly and then disintegrate into many tiny droplets. increases by bringing flames closer to droplets (at large separations) in convective situations [I 511 and decreases by producing larger effective diameters [155]. [151][155] and theoretical [154].4. by impurities like dustmay occur at somewhat lower temperatures that nevertheless still exceed the boiling point.
. or solid [166] fuels. These profiles are invariant with distance along the flame sheet except near the edge of the liquid. . and CH. these are produced by finiterate pyrolysis of the fuel vapor as it is heated.) and oxidizer 0. The similarity of the profiles of temperature T.H.H. such as C.70 Diffusion Flames and Droplet Burning through which fuel passes into initially stagnant or uniformly flowing oxidizer. . The particles are seen to pass through the luminous flame.6 may be seen in Figure 3. The streamlines are made visible by seeding the gas with fine particles and illuminating it from the side.8. fuel (heptane. A photograph of one such flame is shown in Figure 3.7. with the profiles of Figure 3. lighter fuel species.. the flame in the mixing layer of two parallel streams [164].7 a pool of liquid fuel is supplied with liquid from below. H. For simplicity. but this is less evident because the downward convection quickly FIGURE 3. Photograph of a typical diffusion flame for heptane burning in air with streamlines revealed by a particletrack method [165]. and the BurkeSchumann flame [1671. A rounding of the temperature profile in the reaction sheet is evident. In Figure 3.. some fuel species also penetrate to the oxidizer side. Also. Results obtained are qualitatively similar. attention here will be focused on the axisymmetric counterflow diffusion flame with a planar flame sheet.8.7. C. A recent review [1681 summarizes observations made for various counterflow configurations. Concentration and temperature profiles measured in this flow are shown in Figure 3. liquid [165]. there is a stagnation point very near the liquid surface. A small amount of oxygen penetrates the flame sheet and survives on the fuel side. are found on the fuel side of the reaction region. although behaviors are more complex in configurations that exhibit variations of profiles with distance along the flame sheet. flows out of a duct from above at a constant exit velocity U . counterflow diffusion flames with gaseous [162]. and an oxidizing gas with oxygen mass fraction Yo.
There are two phenomena that can lead to this reactionzone broadening: equilibrium dissociation and finiterate chemistry.185 U=0. leading to formulas for flame . If all Lewis numbers are unity.O tend to peak in concentration at the flame sheet and to decrease in concentration more rapidly on the oxidizer side than on the fuel side.4.305 m/s Distance from Luminous Flame Zone (mm) FIGURE 3.7) may be employed to calculate the temperature and composition at each mixture fraction.1 and 3. Rounding of the temperature profile may occur by the former effect as a consequence of energy extracted chemically by dissociation. reduces their concentrations in a convectivediffusive balance.3.8. Structure of the Flame Sheet 71 Y. Concentration and temperature profiles above a flat surface of liquid heptane burning in a downwardly directed stream of a mixture of oxygen and nitrogen [165].2 = 0.. The flame is a sheet of finite thickness. then in principle it is straightforward to account for equilibrium dissociation. again because of convective effects.3.3 in effect remain entirely applicable. [I691 and [170]). and an equilibrium chemistry program (Section A. Analyses by perturbation methods for equilibrium broadening of diffusion flames have been developed (for example. The formulations of Sections 3. a mixture fraction may be introduced as a coupling function. Products such as C 0 2 and H. The questions to be addressed here concern the rounding of the profiles of temperature and of product concentrations at the reaction sheet and the leakage of fuel and oxidizer through the flame. not precisely a surface.
eds.” DynamicsofFIamesand Reactive Systems. 19. one of the similarity groups introduced by Damkohler [174]. K. A. However. lowvelocity. Oppenheim and R. Barenblatt. Spalding. In the former the ratio of a diffusion time to a reaction time. Zhur. Zel’dovich. [171]). vol. treats the question of diffusionflame structure for a onestep.8since the broadening shown is caused mainly by finiterate chemistry. B. 1296 (1950)l and of D. Tekhn. M. Fiz. hightemperature. Fuel 33. they would not be relevant for many of the applicationsfor example. Damkohlernumber asymptotics and activationenergy asymptotics. 3760). and G. B. Numerical integrations that have been performed mainly employ onestep (first in [107]) or twostep [173] approximations to the kinetics. B. With simplified chemical kinetics. ’ Among the earliest works employing simplified kinetics and obtaining analytical approximations are those of Y. N. 1984. Williams. A . Moscow: Nauka. No. Appropriate onestep approximations are realistic+ for limited purposes over restricted ranges of conditions. Bowen. 95 of Progress in Astronautics and Aeronautics. Chapter 7 of The Mathematical Theory of Combustion and Explosion by Y.G. Librovich.* and the results could be of interest for largescale. laminar diffusion flames. and one of the major currently outstanding diffusionflame problems is to develop better simplified kinetic models for hydrocarbon diffusion flames that are capable of predicting results such as observed correlations [172] for concentration profiles of nonequilibrium species. R. there are important aspects of flame structure (for example. is treated as a large parameter. 1. being restricted mainly to dissociation of products into reactants rather than into radicals. secondorder. Arrhenius reaction in a clear fashion. V. sootconcentration profiles) that cannot be described by onestep. kinetic schemes. perturbation methods are attractive for improving understanding and also for seeking quantitative comparisons with experimental results. Makhviladze. J. Leakage of reactants through the flame is a finiterate effect that cannot be found from analyses of equilibrium broadening. but the types of dissociation thus far included in the simplified formulas are not entirely realistic.255 (1954). Soloukhin. . 1980. overall. Known rates of elementary reaction steps may be employed in the full set of conservation equations. Zel’dovich. Peters and F. Manson. for the profiles appearing in Figure 3. Complexities of diffusionflame problems cause this approach to be difficult to pursue and motivate searches for simplifications of the chemical kinetics [172]. Memo. with solutions sought by numerical integration (for example.72 Diffusion Flames and Droplet Burning thicknesses in terms of equilibrium constants. B. NACA Tech. I. Two types of perturbation approaches have been developed. “Effects of Chemical Equilibrium on the Structure and Extinction of Laminar Diffusion Flames. It would be relatively straightforward to extend the perturbation methods to include realistic types of dissociation.1199 (1949) [English translation. There are a number of possible approaches to the calculation of influences of finiterate chemistry on diffusion flames. New York: American Institute of Aeronautics and Astronautics. However. and in the latter the ratio of the energy of activation to the thermal * Recently some progress has been made along these lines (N.
They also afford possibilities of investigating other reaction regimes [176]. Structure of the Flame Sheet 73 energy. 3. Alternatively. Moreover.3. and different uniform properties over the rest of the boundary. activationenergy asymptotics may lead to results valid for all Damkohler numbers. (71) . assume that there is no oxidizer in the fuel stream and no fuel in the oxidizer stream. and 3 of Section 1. O ) / ( B F . and from equation (14) it is readily found that p  a z at + PV * V Z = V * (pDVZ). most of the analysis of its structure can be performed without reference to a particular configuration. is the molecular weight of molecule i. 0. Activationenergy asymptotics appear. regimes of small Damkohler numbersand they are extendable beyond onestep chemistry [177][I811 more easily than are activationenergy asymptotics. possibly at infinity). is taken to be large.4. the mixture fraction may be defined as z = (B . and vv). the application to diffusionflame structure was made by Liiian [184]. The mixturefraction variable Since the reaction zone is thin. emphasized as important by FrankKamenetskii [37]. where W. is the atomic weight of atom j . O  B O . [182]for example. if the binary diffusion coefficients are equal then equation (137) is obtained. to be the more useful of the two types of perturbation approaches. and in terms of the oxidizerfuel coupling function /3. in general. the form given in equation (1) may be adopted. called the fuel stream (subscript F . consider a twostream problem having uniform properties over one portion of the boundary.4. but they can be treated by activationenergy asymptotics. However analyses of phenomena such as sharp ignition and extinction events cannot be performed on the basis of Damkohlernumber asymptotics.)vv)x. For a onestep reaction. Damkohlernumber asymptotics can provide estimates of reactionzone broadening in nearequilibrium situations [125]. A book is available devoted largely to activationenergy asymptotics [1831. for general chemistry the local mass fraction of an atom of typej may be defined as Z j = (K./W. also possibly at infinity). 0. [175][181]. [126]. the mixture fraction may then be defined as in equation (70) with p replaced by Z j for a selected atom j that does not have the same value of Z j in each stream. appearing in equation (6).O ) > (70) so that Z = 1 in the fuel stream and Z = 0 in the oxidizer stream. which appears in equation (A36).2.2. To introduce a general approach of this type. is the number of atoms of kind j in molecule i. A conservation equation for Z may be derived from the general conservation equations. and therefore results of Damkohlernumber asymptotics may be extracted from those of activationenergy asymptotics. called the oxidizer stream (subscript 0.B O . Under assumptions 1.3.
. the further assumptions of a Lewis number of unity (Yc. For the coupling functions p. which were needed in Section 1. the mass fractions may be expressed explicitly in terms of Z and YF. that equation (72) may be written as a+ h+v.i = cp = constant and if the chemistry is described by equation (1). it is readily shown that = hO. This development has the important attribute of not invoking the assumptions of steady flow and of onestep chemistry.Vh=+V aP.= .mF = x/yvi Y F I W F v F .14 Diffusion Flames and Droplet Burning The chemicalsource term does not appear in equation (71) because (like p or Z j ) Z is a conserved scalar. For example.q. all conserved scalars are expressible in terms of 2. pat at Evidently. thereby giving (74) as a solution for h in terms of Z . Energy conservation warrants special consideration.mi . provided that this form is consistent with initial and boundary conditions.hO. Simplified. The sum of the thermal and chemical enthalpies is h = hiY.3. replaced by Z j . then formulas for all Zj in terms of Z are obtained by solving equation (70) for p. N * (v + Vi) v . .  (72) For flows at low Mach numbers with negligible radiant heat flux. explicit expressions for concentrations and temperature in terms of Z and one other variable can be written if cp.namely. a scalar quantity that is neither created nor destroyed by chemical reactions. = v * * (pDVPJ. Under additional assumptions. if initial conditions and boundary conditions are appropriate. . Substitution of this into equation (110) and the result into equation (13) may be shown by use of equations (11) and (12) to provide as a general form of energy conservation cr=l +v * [(pU  P)* v] +p i= 1 c Yf. From these solutions. + p  ab’i at + pv vg. it may be shown from equation (16) under the assumptions introduced above. (75) whence in view of equation (71). = pD) and negligible dp/& reduce equation (73) to the same form as equation (71). solutions in a form like equation (74) exist. 0 Olz + ( h F . 0 .
oWo~oTo. = ( 1 + (81) = yF. the quantity qo/wFvF is the heat released per unit mass of fuel consumed.oqo/cp T . without chemical reactions. In the BurkeSchumann limit.Z / Z J The temperature may be obtained as a function of Z in the BurkeSchumann limit by use of these results in equation (79). O(z  2c)/(1  zc>. The frozen concentration profiles of fuel and oxidizer would be Tf = TO. After algebraic reduction it is found that where qo is defined by equation (44). it is found that where 7 ’ is given by equation (SO).oYo. (80) give the “frozen” temperature profile that would be obtained for pure mixing. In equation (79). and all concentrations and temperatures may be written explicitly in terms of Z. 0  TO. The temperature is expressed in terms of 2 and YFthrough use of equations (74). + YF.3. a flame surface exists where YF = Yo = 0.oWFvFTiF.o .OWOVO YO. and according to equation (77).f = yO.f = yF.4.0 WFvFIFor Z > Z. (78) obtained from equation (111) with cp. = yo. From equation (77) it is seen that at YF = Yo = 0 the corresponding (stoichiometric) value of the mixture fraction is YF. yF 2.OWFVF (85) .o + YF. 0 I z . ‘F. no oxidizer is present. (76). Note that by putting Z = Zcin equation (84) the appropriate adiabatic flame temperature is obtained : + YF.O(l . 0 + (TF.i= c p = constant. and equation (77) gives (83) yo = YO.Oz? yO.no fuel is present.(T To). and the first two terms. (82) for Z < Zc.O(l  z>. Structure of the Flame Sheet 75 The similar expression is obtained directly from equation (70). and (77) in the formula h = h:YF f hgY0 + N2 hpx i= 1 + c.0 wovo/Yo..
For example./r. and equations (76). = v. in equation (25) the flamesheet condition. we therefore need an additional conservation equationfor example.. becomes 2 = Z. 3. that is obtained from this solution can be derived from the fact that 1 + v = (1 . it could have been introduced in Section 1. y e . that for fuel. the empirical reaction orders nF and no have been introduced. it is found that equation (42). .Z. becomes simply 2 = 1 . and (79) may be employed.EIRoT (87) In this onestep approximation. interpreted for 2.)' with Z . under the present assumptions. the replacement of /? by 2 achieves a convenient normalization and also can help to clarify aspects of physical interpretations. and (79) relate the other variables to this quantity and YF.6. (83). and (84) cannot be taken as generally valid relationships. the twostream problem most simply relevant to droplet burning has pure liquid = l).3. obtaining a qualitatively correct variation of the overall rate with the mass fractions of reactants.3. to replace fi in equation (149). For simplicity here we put nF = no = 1. even though YF.4.(1 + B)rl'r. From equations (14). The mixture fraction 2 could have been introduced and employed directly in earlier examples. For the dropletburning problem. The overall activation energy is E . given by equation (81). (19) and (112) we find that.76 Diffusion Flames and Droplet Burning In studying the structure of the flame sheet. it could have been used in place of b in equations (9) and (38). The equivalence of equation (63) with the expression for r. (77). and the overall frequency factor for the rate of fuel consumption is A .1.BT".l < 1 according to fuel in one of the streams (YF. To study the structure of the reaction sheet. a condition of mixturefraction stoichiometry.o equation (64). this additional equation can be written as where W F V F BT" P n + ~n o y p y . Since it is a coupling function.3. with equation (70) employed to recover b from the solution for 2. Activationenergy asymptotics Equation (71) is a conservation equation for the mixture fraction.3. but equations (76). Although such selections of variables basically are matters of personal taste. equations (82). as discussed in Section B. (77). fl = 0. In fact. since the overall stoichiometry seldom provides the correct dependence of the rate on concentrations. (18). when all the assumptions that underlie equation (58) are introduced. where B is defined at the beginning of Section 3.
This may be done in principle if pD is independent of T and if changes in the solution T have a negligible influence on the fluid dynamics. in a manner analogous to that of Crocco [l85]. and (86) it is readily shown that p aT at + pv. Equations (71) and (88)(91). prior to the study of equation (88).(ln 1 V Z I ) V . We consider that equation (71) is solved first.VT= V. namely. in general it is necessary to consider the fluid dynamics as well. and where Tf is given by equation (80). investigate the variation of T with Z and study the influence of finiterate chemistry on this variation. (79). define the flamestructure problem to be investigated. as the variable to be considered in addition to Z . Consider an orthogonal coordinate system in which 2 is one of the coordinates and the other two.(pDV. Structure of the Flame Sheet I1 Since T appears inside the exponential.T = WT T + p D I V Z 1d2 2 y + V. are distances along surfaces of constant 2.(pDVT) + WT. respectively.T) az  pDV. Since v appears in these equations. it is convenient to employ T instead of Y. in this new reference frame. From equations (71). through equations (11) and (12). T (92) .3. however. where The quantities YF and Yo in equation (89) may be expressed in terms of Z and T by use of equations (77) and (79). Otherwise it is only a conceptual aid. Let u and v be velocity components in the x and y directions. This type of investigation is facilitated by introducing into (88) the variable 2 as an independent variable. in seeking solutions. along with boundary and initial conditions.V. x and y.4. We can. When use is made of equation (71) it is found that a formal transformation of (88) into the new coordinate system yields dT pat + PVl. and we cannot investigate directly the variation of T with space and time.
The details of .Z.(1 . Therefore values of the reaction rate some distance away from the stoichiometric surface ( Z = Z....o)/(K . treating these variables as being of order unity. and those solutions will. and t.. By introducing cp and 9 into equation (93)..w ~ / ( ~ D I v z I ~ ) . where E = 2R0T. approximately. defined in Section 5.q2)e(4'+)'$) (96) is obtained.. to the total chemical heat released at the flame.. It can be shown that t is the formal small parameter of expansion for activationenergy asymptotics applied to the diffusion flame and that 1 .ZC)(TF. be exact solutions of equation (92) as well.y is twice the ratio of the heat conducted from the flame toward negative values of Z . thenin view of equations (89).6. Equation (89) then suggests that w T will be largest near the maximum temperature T. where a reduced Damkohler number is with A. E/RoT.. is the twodimensional gradient operator in x and y . (90).O .T)E/RoTf..yq. because of the Arrhenius factor.)/[E(T. and therefore equation (92) becomes.2zc(1 . This implies that to analyze the effect of w T .78 Diffusion Flames and Droplet Burning where vL is a twodimensional vector with x and y components being u and v.p/(DIVZIZ) understood to be evaluated at Z = 2.) will be very small in comparison with the values near Z = Z. is a large parameter. The quantity l/t is closely related to the Zel'dovich number. . as T decreases appreciably below T. A stretching of this kind causes the terms involving the highest Z derivative to be dominant. (95) in which Ts. As an independent variable. and expanding in €with use made of equations (89) (which has nF = no = l).Tf.To. Thus equation (93) viewed as an ordinary differential equation exactly for many problems. in fact. describes T(2) To carry out an analysis of the reaction zone by activationenergy asymptotics.T'. a2T/az2= .Z. In activationenergy asymptotics.1 . cp = (T .)/t and as a dependent variable. y . (93) If AFp/(DIVZI2) is independent of x.. Z may be stretched about Z . and (91)equation (93) possesses solutions in which T varies only with Z .3.2Z. introduce q = (2 .it is helpful to stretch the 2 variable in equation (92) about Z = 2. and that w T will decrease rapidly. and where V. (90).)] (94) and "? = 2 2 .).Z . is Tf evaluated from equation (80) at 2 = Z. and T about T. and (91)in the first approximation the differential equation d2cp/dy2 = d ( q 2 .
1. The difference T.2. over its frozen value. enparameterfree problem d’[/d(’ = 1 countered in Damkohlernumber asymptotics and integrated numerically [175]. As is usually true (Chapter 5).3. [176]. They also show. [126]. the heat release need not be very large if E is large.t2. In typical combustion applications. the fractional decrease of the maximum temperature below T. results in the ’ . and therefore no leakage of fuel or oxidizer in the first approximation. [126]. it may be shown that the matching requires dcp/dq t t. over which departures from the These . but there must be some heat release. matching to the BurkeSchumann solution given by equation (84) may be imposed for q + f co. The thickness of the reaction zone is seen to correspond to 2 .(cp either fuel or oxidizer or both leak through the reaction zone in a fractional amount of order t that increases as 6 decreases [184]. In the large4 limit. Structure of the Flame Sheet 19 deriving equation (96) are not shown here but parallel the development of activationenergy asymptotics given in Chapter 5. Solutions to equation (96) that satisfy these boundary conditions exist for a range of values of 6 [184]. d[/d( + f 1 as ( + fm. proportional to E. 1 as q + fco. The < . is the local temperature increase. being of order E .If this temperature increase is too smallthat is. that of lim.01 and 0. The large4 solution to the above equation has [ T ( + 0 as + k rx.. in the first approximation this matching requires the slopes dT/dZ to agree. scalings vary BurkeSchumann solution occur also is of order ~ / d ’ / ~ with the reaction order [125]. associated with complete combustion occurring at the flame sheet ( Z = 2. is of order ~ / 6 ” and ~ ./E. with [ = cp6”3 and = q d l i 3 treated as being of order unity. The advent of activationenergy asymptotics has helped greatly in clarifying criteria for the validity of the flamesheet approximation. From equation (84) and the definitions of cp and q. then smaller values of E still yield a small value o f t . a reactant having an overall reaction order greater than unity experiences leakage in the first approximation in Damkohlernumber asymptotics [125]. through evaluation T y ~ ) from the numerical integration of equation (96). a rescaling of variables in equation (96).2. and results of expansions to the lowest order in activationenergy asymptotics are reasonably accurate. It is significant that E .. Liiian’s paper [1841 may be consulted for methodology.Tf.). This development shows that Damkohlernumber asymptotics can be obtained as a limiting case of activationenergy asymptotics. that increases as 6 decreases. by a fractional amount.4. if the heat released in complete combustion is too smallthen t is not small enough for w T to vary rapidly enough with temperature near the flame sheet to justify use of activationenergy asymptotics. if the heat release is large. is the best formal parameter of expansion instead of simply RT. . To obtain boundary conditions for equation (96). t lies between 0. the range of 2 .. These solutions show that the maximum temperature is reduced below T. As 6 becomes large.. To have t small. defined in equation (94).
1. by turning off the chemistry outside this sheet. This situation sometimes tends to be encountered in practice for the burning of hydrocarbon fuels in air. For (y(> 1. For the more common problems in which the peak temperature occurs at the flame sheet. typically 0. so that y is near . 3. the extinction formula can be inaccurate if (y(is too close to unity. the chemistry is frozen to all algebraic orders in E on both sides of the reaction zone. These regimes will be discussed in the Section 3. This suggests that equation (84). by a correction term proportional to E. the analysis shows that equilibrium prevails to all algebraic orders in t between the hot boundary and the flame sheet. is of order E. and the thickness of the reaction zone decreases with increasing 6. The extinction phenomenon will be addressed next. and appreciable leakage of fuel into the oxidizer stream is calculated to occur at extinction. one of the two boundary temperatures exceeds the flame temperature.055(1 . The nonexistence of solutions to equation (96) when 6 is too small means that if reaction rates are not sufficiently high. Large values of the magnitude of the reaction rate in the reaction zone help to improve the validity of the flamesheet approximation. and solutions linear in Z are obtained. Ignition and extinction Analysis [184] shows that the minimum value of 6 below which solutions to equation (96) cease to exist is & = [(I . Use of equation (98) in equation (97) provides an explicit expression for conditions of extinction of the diffusion flame.4.4. Damkohlernumber asymptotics apply.Iyl)"]e (98) within 1 % for IyI < 1. and also to equation (84). and the chemistry is frozen between the flame sheet and the cold boundary. the BurkeSchumann solution.l y l ) . The extent of leakage at extinction is proportional to t.80 Diffusion Flames and Droplet Burning strong dependence of wT on T for large E helps to confine the combustion reactions to a thin sheet. The reactant that leaks is described outside the reaction zone by an equation like equation (93) with wT = 0. the value of 2. For problems in which the peak temperature occurs at a boundary instead of at the flame sheet.2.4. defines only one of many possible solutions that may apply outside the flame sheet.26(1 .05.4.1 ~ 1 + ) ~ 0. Therefore. For large values of 6. and in fact there are reaction regimes in which these equations do not apply [184]. . and an abrupt extinction event does not exist for the diffusion flame. the proportionality factor at extinction between the leakage correction to the reactant mass fraction and t becomes large. As I y I approaches unity. where 2 . where T is lower. The result is a modification to equation (82) or equation (83).(1 . in equation (95) can be small. leakage of one of the reactants through the reaction zone increases. As extinction is approached (through decreasing S).Iyl)' + 0. then extinction of the diffusion flame occurs.
. or nearequilibrium regime considered in the previous section. and even in lowest order the flame temperature becomes a premixedJame temperature less than T.. In this regime. is less than a boundary temperature (Iyl > l)..4.2. in the reaction zone. in turn. the factor cp2 . and diffusionflame solutions become accurate. nearly frozen solutions lose accuracy anL premixedflame solutions gain accuracy. and (84) then occur even in lowest order on both sides of the flame sheet. This has been termed the nearly frozen.and T ( Z ) is given by the frozen solutions. Provided that T. (83). then the solution must lie in one of the other regimes. as in a premixed flame. Structure of the Flame Sheet 81 When leakage of one reactant becomes sufficiently large. Yo(Z). in equation (96) the factor cp2 . There are no partialburning solutions when T. nearly frozen solution exist for all values of 6. hence. An asymptotic analysis may be developed for which. then as 6 decreases. There is a fourth regime in which the first approximation to YF(Z). and T ’ [see equation (80) and the sentence following it]. all of which have a thin reaction zone of elevated temperature. A modification to the BurkeSchumann solution of order unity is needed on the side of the flame sheet to which reactant has leaked. or ignition regime [184]. Qualitatively. The method of analysis bears some relationship to the methods discussed in Section B. but they tend to become inaccurate if 6 is too large. solutions are found to exist only for values of a Damkohler number. to distinguish it from the diffusionflame. Yo. Departures from equations (82). then these. Since nearly frozen solutions do not exist for 6 > 61..exceeds both boundary temperatures. that are less than a maximum value 61.4. and modification of scaling for the asymptotic analysis becomes warranted. if 6 > 6. activationenergy asymptotics show that the reaction tends to occur in a region near the hotter of the two boundaries [184]. As the premixedflame temperature decreases. who terms it thepremixedflame regime of the diffusion flame. If I y I is near enough to unity and t is not too small. related to 6. the situation in the vicinity of the reaction zone tends to begin to resemble that for a premixed flame. conditions may be calculated for which the second reactant also begins to leak through the reaction zone in appreciable quantities. although there are significant differences because of the absence of spatial homogeneity. lose accuracy at larger 6. . In this case.q)(cp + q ) of fuel and oxidizer concentrations in the reaction zone. When T. YF.q2 remains effectively constant [184]. One of these two factors tends to become constant. as 6 increases from zero with t fixed..3. depletion of only one reactant occurs.The asymptotic analysis of this regime is due to Liiian [184].q2 is the scaled product (cp . extinction of the diffusion flame occurs subsequent to its passage into the premixedflame region..5. This regime has been termed the partialburning regime [184]. is less than a boundary temperature. the value 61 may be interpreted as a critical Damkohler number for ignition.
at which point the solution must jump discontinuously to a vigorously burning state on the upper branch. [187] and [l88]) indicate that instability begins at or very near the extrema.82 Diffusion Flames and Droplet Burning . This figure applies for small 6 and only if T. and the lower branch and part of the middle branch by analysis of the ignition regime. The upper branch and part of the middle branch are predicted by analysis of the diffusionflame regime.. prior to the vertical tangent. . but it is likely that the entire partialburning regime is unstable and therefore not significant for diffusion flames. which is a sketch of the maximum temperature T. it is a very rapid. If 6 is gradually decreased for a vigorously burning diffusion flame on the upper branch.Lower branch I 0 % * 6 81 FIGURE 3. It is believed that the middle branch is unstable. leading to an ignition or extinction jump. the few studies that have been performed (for example. much of the middle branch and part of the upper branch by analysis of the premixedflame regime. then at least when 6 Eis passed. as a function of the Damkohler parameter 6. Schematic illustration of the dependence of the maximum temperature on the reduced DamkBhler number for diffusion flames. exceeds the boundary temperatures. Of course. usually in the diffusionflame regime. It is conceivable that the upper or lower branches experience instability.9.  Middle branch I 1 I I . TF.o and To. Although entirely thorough stability analyses have not been completed.9. If 6 is gradually increased for a mixing flow with negligible chemical reaction. then the system may be expected to move along the lower branch until it reaches d l . A suitable partialburning analysis can give the middle branch and both bends. . timedependent process that would . as obtained both by numerical integration of the full equations [I1861 and by activationenergy asymptotics [1841. the solution must jump discontinuously to the lower branch. flames on this branch have not been obtained experimentally. where generally there is negligible reaction. The most common situation is summarized qualitatively in Figure 3. an ignition or extinction jump itself is not instantaneous.
3. that is. notably in connection with turbulent flames (Chapter 10). personal communication) that in the premixedflame regime. The possibility of using the Crocco transformation to achieve this generality was suggested by Peters [190]. it can be helpful to eliminate one of the three formally by means of a transformation.9. The analysis exposing activationenergy asymptotics originally was developed within the context of the problem of the steady counterflow diffusion flame [184]. Since the stretching needed to describe the reaction zone by activationenergy asymptotics increases the magnitude of the diffusion terms with respect to the (less highly differentiated) convection terms. Structure of the Flame Sheet 83 not be described by Figure 3. instability occurs near (not precisely at) the extremum. since no other terms occur in the equation for species conservation. It also has been applied to the quasisteady. If this regime is unstable.9 or equation (93) but instead would require retention of p aT/& in equation (92). in this event the premixedflame regime is relevant to Figure 3. Analysis of onedimensional. timedependent mixing and reaction of initially separated fuel and oxidizer by activationenergy asymptotics [193] encounters many of the aspects that have been discussed here. The stability of solutions in the premixedflame regime for systems having T. it will prove useful later. greater than the boundary temperatures has been questioned [189]. The presentation here has not been cast within the framework of any particular configuration for the diffusion flame. as 6 decreases solutions on the upper branch in the diffusionflame regime become less accurate than those in the premixedflame regime prior to extinction. Liiian. 2 equals the normal distance (measured from the oxidizer side of the layer) divided by the thickness of the layer. it is evident that outside the reaction zone there must be a balance between diffusion and convection. Since the possible simultaneous presence of three phenomenaconvection. and reactioncomplicate analyses. The simplest configuration to which the formulation may be applied is the steady. [191]. For this problem with p D constant. spherically symmetrical burning of a droplet [1923. in the first approximation these reaction zones maintain a balance between diffusion and reaction and may be more descriptively termed reactivediflusive zones.9 in that. then only solutions in the diffusionflame and ignition regimes are significant with respect to Figure 3. if IyI is near unity and t is not too small. Thus these flames consist of convectivediffusivezones separated by thin reaction zones. diffusion. I VZ I is the reciprocal of the thickness Gf the layer.4. stagnant diffusion layer. It seems more likely (A. by finding a transformation to a new spatial variable such that the convective and diffusive terms coalesce into a . Thus the BurkeSchumann flame consists of two convectivediffusive zones separated by a reactivediffusive zone. and Z therefore may be interpreted directly as a normalized physical coordinate. For steadystate diffusion flames with thin reaction sheets. This is achieved in the analysis of the counterflow flame [184] by writing the equations in a convectionfree form.
84
Diffusion Flames and Droplet Burning
single term that resembles a purely diffusive term in the new variable. The transformation to convectionfree form is especially useful for analyzing the ignition regime, for which the reaction zone is not thin in the original spatial variable but instead extends an infinite distance into the hotter of the two streams, causing convection, diffusion, and reaction all to be important in the reaction zone. For any diffusion flame having a symmetry that eliminates the transverse gradient from equation (92), it is evident from equation (92) that the transformation to the mixture fraction as an independent variable achieves' a convectionfree form. Thus a diffusionflame prescription for transformation to convectionfree form is to introduce a mixture fraction as an independent variable. Overall rate parameters for diffusion flames, needed in applying the analysis given here, often are not available. The analysis itself may be employed to design an experimental technique for obtaining these parameters from measurements of diffusionflame extinction [194]. The counterflow configuration is convenient ; the analysis addresses experiments with liquid fuels, shown in Figure 3.7. The quantity D I VZI2 is proportional to the exit velocity U of gases leaving the oxidizer duct [195] if the experiment is carefully designed to establish a radial channel flow with a uniform rate of injection from above. To achieve reasonable accuracy in calculating D I VZ 12, it is necessary to take into account variable properties ( p and D) and the influence of the presence of the condensed phase (through parameters such as the transfer number B ) [194]. O n the basis of the theory, an Arrhenius graph can be defined from equations (97) and (98), in which a semilog plot of a quantity proportional to the measured value of U at flame extinction, as a function of l/T, provides E from the slope and AF from a value. Rate parameters obtained in this manner may be employed for calculating diffusionflame structure and extinction in other configurations. Overall rate parameters for a number of fuels have been found in this way; a critical review is available [196]. 3.5. MONOPROPELLANT DROPLET BURNING
A monopropellant is a fuel that is capable of reacting exothermically without an oxidizer. Numerous such fuels are known; many are liquids. A droplet of liquid monopropellant can burn in an inert atmosphere without an oxidizer. There are rockets that employ liquid monopropellants as fuels for propulsion ; in some applications these are better suited than bipropellant rockets that employ a liquid fuel and a liquid oxidizer. The relevance of monopropellant droplet combustion to the operation of monopropellant rockets and fundamental curiosity have prompted studies of the subject. Burning monopropellant droplets have the spherical configuration shared by burning fuel droplets in an oxidizing atmosphere and the pre
3.5. Monopropellant Droplet Burning
85
mixedness of ordinary laminar flames (Chapter 5). Consequently, it may be expected that monopropellant droplets exhibit some characteristics of each of these systems. As is often true in intermediate cases, the analysis of monopropellant droplet combustion in certain respects is more complicated than that of either the planar premixed flame or the fuel droplet. A number of theoretical studies of the quasisteady, spherically symmetrical burning of monopropellant droplets have been published. These include work involving numerical integration of the conservation equations or of approximations thereto [197], [1981, analytical approaches involving ad hoc approximations [85], [199][201], methods employing Damkohlernumber asymptotics [202][206] and procedures based on activationenergy asymptotics [207][210]. Reviews may be found in the literature [lll], [183]. Consideration has been given to burning in inert atmospheres under both adiabatic and nonadiabatic conditions, as well as to burning in reactive atmospheres containing an oxidizer that can react with the fuel or its decomposition products [SS], [205], [206]. The general character of results of these analyses may be understood on the basis of physical reasoning. Consider first the burning in an inert atmosphere, the temperature of which is at or above the adiabatic flame temperature for decomposition of the monopropellant. If the reaction rate is sufficiently low, then the combustion will have a negligible influence on the rate of mass loss of the droplet, and the dominant process will be inert vaporization in a hot atmosphere. We have seen that this process has a dsquare law, and we know how to calculate the burning rate (for example, from equation (58) with Yo,m = 0; see Section 3.3.6). The reaction occurs far from the droplet surface, in a region of large extent, whose location depends on a nondimensional activation energy and/or on a Damkohler number, D,, = rfW/pD, where W is a representative value for the mass rate of consumption of fuel. A decrease in D,, or an increase in the activation energy moves the reaction region farther from the droplet. The burningrate constant, K , is given by equation (65) and does not depend on DII. If the reaction rate is larger, then the combustion influences the rate of mass loss. For sufficiently large values of D,,, the reaction occurs in a narrow zone whose thickness is small enough for curvature effects to be negligible within it. If a nondimensional activation energy is large, then this zone is separated from the droplet surface by a convectivediffusive zone. In this case the structure of the reaction zone resembles that of the reaction zone of a premixed laminar flame (Chapter 5), with heat addition from superadiabatic products if the ambient temperature exceeds the adiabatic flame temperature. Qualitatively, the speed of propagation of the planar laminar flame begins to control the rate of mass loss of the droplet, and K becomes proportional to (see Chapter 5). Since K drf/dt and r l , the dsquare law is replaced by a dlaw in which the first power of the droplet diameter decreases linearly with time.
fi

fi
86
DiffusionFlames and Droplet Burning
The transition between these two limiting regimes occurs roughly Asymptotic where K , as given by equation (65) is of the same order as analyses give sharper criteria for the transition [207]. Over any limited range of conditions, drk/dt is approximately constant, where 1 < k I 2. In view of the dependence of D,, on rI, for large droplets the &law tends to hold; for small droplets the dsquare law holds. Additional phenomena occur if the temperature of the inert atmosphere is less than the adiabatic flame temperature. As might be expected from the discussion in the previous section, ignition and extinction events may be encountered [207]. Formulas for extinction conditions, obtained on the basis of activationenergy asymptotics, are available [207]. Within the context of activationenergy asymptotics, a partialburning regime of the type discussed in the preceding section is found to arise [207]. A diffusionflame regime, of course, does not exist for burning in an inert atmosphere; in this respect the problem of the monopropellant droplet is simpler than that of a fuel droplet in an oxidizing atmosphere. For burning in a reactive atmosphere, conditions may be encountered for which the analysis of the diffusionflame regime is relevant. Under suitable conditions there are two thin reaction sheets, an inner one at which premixed decomposition of the monopropellant occurs and an outer one having a diffusionflame character. Categorizations are available for potential limiting behaviors at large Damkohler numbers in reactive environments [206]. There are many flamestructure possibilities, not all of which have been analyzed thoroughly.
JE.
REFERENCES
1. S. S. Penner, Chemistry in Jet Propulsion, New York: Pergamon Press, 1957, Chapter XXII. 2. A. M. Kanury, Introduction to Combustion Phenomena, New York Gordon and Breach, 1975, 142269. 3. I. Glassman, Combustion, New York: Academic Press, 1977, 158193. 4. S. P. Burke and T. E. W. Schumann, Znd. Eng. Chem. 20,998 (1928). 5. J. A. Fay, J. Aero. Sci. 21, 681 (1954). 6. A. Goldburg and S. I. Cheng, C & F 9, 259 (1965). 7. J. B. Haggard, Jr. and T. H. Cochran, CST 5,291 (1972). 8. T. M . Liu, C S T 7 , 283 (1973). 9. P. H. Thomas, 9th Symp. (1963), 844859. 10. F. J . Kosdon, F. A. Williams and C. Buman, Z2th Symp. (1969), 253264. 11. F. A. Williams, Prqg. Energy Combust. Sci. 8, 317 (1982). 12. H. C. Hottel and W. R. Hawthorne, 3rd Symp. (1949), 254266. 13. M. Klajn and A. K. Oppenheim, 19th Symp. (1982), 223235. 14. 0. A. Tsukhanova, 6th Symp. (1957), 573577.
References
87
L. N. Khitrin and L. S. Solovyeva, 7th Symp. (1959), 532538. H. A. Lauwerier, Appl. Sci. Research At?, 366 (1959). D. E. Rosner, A.1.Ch.E. Journal9, 321 (1963). D. E. Rosner, Chem. Eng. Sci. 19, l(1964). P. A. Libby and F. A. Williams, AIAA Journal 3, 1152 (1965). D. E. Rosner, AZAA Journal 2, 593 (1964). D. E. Rosner, Int. J. Heat Mass Transfer 9, 1233 (1966). D. E. Rossner, 11th Symp. (1967), 181196. R. Ark, “Hierarchies of Models in Reactive Systems,” in Dynamics and Modeling ofReactive Systems, W. E. Stewart, W. H. Ray, and C. C. Conley, eds., New York: Academic Press, 1980, 135. 24. D. E. Rosner, Annual Review ofMaterials Science 2, 573 (1972). 25. I. Glassman, Combustion, New York: Academic Press, 1977, 258269. 26. M. F. R. Mulcahy and I. W. Smith, Rev. Pure and Appl. C h m . 19,81 (1969). 27. G. Blyholder, J. S. Binford, Jr., and H. Eyring, J . Phys. Chem. 62,263 (1958). 28. R. H. Essenhigh, 8th Symp. (1962), 813814. 29. J. Nagle and R. F. StricklandConstable, Proc. Fifth Conf. on Carbon, vol. 1, New York: Macmillan, 1962, 154164. 30. J. N. Ong, Jr., Carbon 2, 28 1 (1964). 31. H. G. Maahs, NASA Tech. Note No. D6310 (1971). 32. J. R. Walls and R. F. StricklandConstable, Carbon 1, 333 (1964). 33. R. F. StricklandConstable, Second Conf. on Industrial Carbon and Gruphite, Soc. Chem. Ind. London, 235 (1966). 34. P. L. Walker, Jr., F. Rusinko, Jr., and L. G. Austin, Advances in Catalysis, vol. 11, New York: Academic Press, 1959, 133221. 35. D. R. Olander, W. Seikhaus, R. Jones, and J. A. Schwarz, J. Chem. Phys. 57,408 (1972). 36. T. R. Acharya and D. R. Olander, Carbon 11, 7 (1973). 37. D. A. FrankKamenetskii, Diffusion and Heat Transfer in Chemical Kinetics, New York: Plenum Press, 1969, 6472. 38. K. Matsui, A. Koyama, and K. Uehara, C & F25, 57 (1975). 39. S. K. Ubhayakar and F. A. Williams, J . Electrochem. Soc. 123, 747 (1976). 40. C. Park and J. P. Appleton, C & F 20, 369 (1973). 41. P. A. Libby and T. R. Blake, C & F 3 6 , 139 (1979). 42. D. E. Rosner and H. D. Allendorf, Carbon 3, 153 (1965); 7, 515 (1969). 43. H. Tsuji and K. Matsui, C & F 26, 283 (1976). 44. G. Adomeit, G. Mohiuddin, and N. Peters, 16th Symp. (1977), 731743. 45. D. B. Spalding, Fuel 30, 6, 121 (1951). 46. D. E. Rosner, Oxidation of Metals 4, 1 (1972). 47. I. Glassman, “Combustion of Metals. Physical Considerations,” in Solid Propellant Rocket Research, vol. 1 of Progress in Astronautics and Rocketry, M. Summerfield, ed., New York: Academic Press, 1960, 253258. 48. N. Peters, 15th Symp. (1975), 363375. 49. S. K. Ubhayakar, C & F 26,23 (1976). 50. N. R. Amundson and E. Mon, “Diffusion and Reaction in Carbon Burning,” in Dynamics and Modeling of Reactive Systems, W. E. Stewart, W. H. Ray, and C. C. Conley, eds., New York: Academic Press, 1980,353374. 51. P. A. Libby, C & F 3 8 , 2 8 5 (1980).
15. 16. 17. 18. 19. 20. 21. 22. 23.
88
Diffusion Flames and Droplet Burning
M. Matalon, CST24, 115 (1980). M. Matalon, CST25,43 (1981). G. Mohan and F. A. Williams, AIAA Journal 10, 776 (1972). R. P. Wilson, Jr., “Studies on Combustion of Aluminum Particles,” Paper No. 693, Spring Meeting, Western States Section, The Combustion Institute, China Lake, CA (1969). 56. T. A. Brzustowski and I. Glassman, VaporPhase Diffusion Flames in the Combustion of Magnesium and Aluminum : 1 Analytical Developments,” in Heterogeneous Combustion, vol. 15 of Progress in Astronautics and Aeronautics, H. G. Wolfhard, I. Glassman, and L. Green Jr., eds., New York: Academic Press, 1964, 75116. 57. D. K. Kuehl, AIAA Journal 3, 2239 (1965). 58. L. A. Klyachko, Fiz. Gor. Vzr. 5, 404 (1969). 59. P. F. Pokhil, A. F. Belaev, Yu. V. Frolov, V. S. Logachev, and A. E. Korotkov, Combustion of Metal Powder Formulations in Reactive Environments, Moscow : Nauka, 1972. 60. C. K. Law, C S T 7 , 197 (1973). 61. C. K. Law, CST 12, 113 (1976). 62. C. K. Law and F. A. Williams, C & F22, 383 (1974). 63. F. A. Williams, “Droplet Burning,” Chapter I1 of Combustion Experiments in a ZeroGravity Laboratory, T. H. Cochran, ed., vol. 73 of Progress in Astronautics and Aeronautics, New York : American Institute of Aeronautics and Astronautics, 1981, 3160. 64. S. Kumagai, Jet Propulsion 26, 786 (1956). 65. S. Kumagai and H. Isoda, 6th Symp. (1957), 726731. 66. H. Isoda and S. Kumagai, 7th Symp. (1959), 523531. 67. S. Kumagai, T. Sakai, and S. Okajima, 13th Symp. (1971), 779785. 68. S. Okajima and S. Kumagai, 15th Symp. (1975), 401407. 69. B. Knight and F. A. Williams, C & F 38, 111 (1980). 70. G. N. Khudyakov, Izvest. Akad. Nauk. S.S.S.R., Otdel. Tech. Nauk. 4,508 (1949). 71. D. B. Spalding, Fuel29, 25 (1950); 32, 169 (1953). 72. D. B. Spalding, 4th Symp. (1953), 847864. 73. G. A. Agoston, H. Wise, and W. A. Rosser, 6th Symp. (1957), 708717. 74. G. A. Agoston, B. J. Wood, and H. Wise, Jet Propulsion 28, 181 (1958). 75. D. G. Udelson, C & F 6 , 93 (1962). 76. J. W. Aldred, J. C. Patel, and A. Williams, C & F 17, 139 (1971). 77. G. S. Canada and G. M. Faeth, 14th Symp. (1973), 13451354. 78. S. R. Gollahalli and T. A. Brzustowski, 14th Symp. (1973), 13331344. 79. D. D. Ludwig, F. V. Bracco, and D. T. Harrje, C & F25, 107 (1975). 80. S. I. AbdulKhalik, T. Tamaru, and M. M. ElWakil, 15th Symp. (1975), 389399. 81. G. A. E. Godsave, 4th Symp. (1953), 818830. 82. A. R. Hall and J. Diederichsen, 4th Symp. (1953), 837846. 83. M. Goldsmith, Jet Propulsion, 26, 172 (1956). 84. C. S. Tarifa, “On the Influence of Chemical Kinetics on the Combustion of Fuel Droplets,” Congres Internacional des Machines a Combustion, Paper B9, Copenhagen: A. Busk, 1962. 85. C. S. Tarifa, P. P. del Notario, and F. Garcia Moreno, 8th Symp. (1962), 10351056. 86. J. W. Aldred and A. Williams, C & F 10, 396 (1966).
‘I
52. 53. 54. 55.
References
89
87. M. T. Monaghan, R. G. Siddal, and M. W. Thring, C & F 12,45 (1968). 88. G. M. Faeth, D. P. Dominicis, J. F. Tulpinsky, and D. R. Olson, 12th Symp. (1969), 918. 89. G. M. Faeth and R. S. Lazar, AZAA Journal 9, 2165 (1971). 90. C. K. Law and F. A. Williams, C & I; 19,393 (1972). 91. F. L. Dryer, 16th Symp. (1979), 279295. 92. T. Kadota and H. Hiroyasu, 18 Symp. (1981), 275282. 93. H. C. Hottel, G. C. Williams, and H. C. Simpson, 5th Symp. (1955), 101129. 94. J. A. Bolt and M. A. Saad, 6th Symp. (1957), 717725. 95. B. J. Wood, W. A. Rosser, and H. Wise, AIAA Journal 1, 1076 (1963). 96. P. Eisenklam, S. A. Arunachalam, and J. A. Weston, 11th Symp. (1967), 715728. 97. B. J. Wood and W. A. Rosser, AIAA Journal 7,2288 (1969). 98. A. S. M. Nuruzzaman, A. B. Hedley, and J. M. Beer, 13th Symp. (1971), 787799. 99. J. J. Sangiovanni and A. S. Kestin, CST 16, 59 (1977). 100. J. J. Sangiovanni and L. G. Dodge, 17th Symp. (1979), 455465. 101. J. C. Lasheras, A. C. FernandezPello, and F. L. Dryer, CST21, 1 (1979). 102. J. C. Lasheras, A. C. FernandezPello, and F. L. Dryer, CST22, 195 (1980). 103. J. C. Lasheras, A. C. FernandezPello, and F. L. Dryer, 18th Symp. (1981), 293305. 104. B. Sreznevsky, Zh. R. F. Kho. 14,420,483 (1882). 105. I. Langmuir, Phys. Rev. 12, 368 (1918). 106. M . Goldsmith and S. S. Penner, Jet Propulsion 24, 245 (1954). 107. J. Lorell, H. Wise, and R. E. Carr, J. Chem. Phys. 25, 325, (1956). 108. B. P. Mullins and S. S. Penner, Explosions, Detonations, Flammability and [pition, New York: Pergamon Press, 1959, 101107. 109. H. Wise and G. A. Agoston, “Burning of a Liquid Droplet,” in Advunccs in Chemistry Series, no. 20, Am. Chem. Soc. (1958), 116135. 110. A. Williams, Oxidation and Combustion Reviews 1, 1 (1968). 111. D. E. Rosner, “Liquid Droplet Vaporization and Combustion,” Section 2.4 of Liquid Propellant Rocket Combustion Instability, D. T. Harrje and F. H. Reardon, eds., NASA SP194, Washington (1972), 74100. 112. A. Williams, C & F 21, 1 (1973). 113. G. M. Faeth, Prog. Energy Combust. Sci. 3, 191 (1977). 114. C. K. Law, Prog. Energy Combust. Sci. 8, 171 (1982). 115. C. K. Law, “Mechanisms of Droplet Combustion,” in Proceedings ofthe Second International Colloquium on Drops and Bubbles, D. H. Le Croisette, ed., JPL Publication 827, Jet Propulsion Laboratory, Pasadena (1982), 3953. 1 16. F. A. Williams, J. Chem. Phys. 33, 133 (1960). 117. K. Kobayasi, 5th Symp. (1955), 141148. 118. N. Nishiwaki, 5th Symp. (1955), 148151. 119. D. B. Spalding, ARS Journal 29, 828 (1959). 120. D. E. Rosner, AZAA Journal5, 163 (1967). 121. A. Chervinsky, AIAA Journal 7, 1815 (1969). 122. C. S. Tarifa, A. Crespo, and E. Fraga, Astronautica Acta 17, 685 (1972). 123. C. E. Polymeropoulos and R. L. Peskin, CST 5, 165 (1972). 124. D. E. Rosner and W. S . Chang, CST7, 145 (1973). 125. D. R. Kassoy and F. A. Williams, Phys. Fluids 11, 1343 (1968). 126. D. R. Kassoy, M. K. Liu, and F. A. Williams, Phys. Fluids 12, 265 (1969).
90
Diffusion Flames and Droplet Burning
127. S. H. Sohrab, A. Liiian, and F. A. Williams, CST27, 143 (1982). 128. H. Wise and C. M. Ablow, J. Chem. Phys. 27, 389 (1957). 129. C. K. Law, C & F26, 17 (1976). 130. C . K. Law and W. A. Sirignano, C & F 2 8 , 175 (1977). 131. S. Prakash and W. A. Sirignano, Int. J. Heat Mass Transfer 21, 885 (1978). 132. S. Prakash and W. A. Sirignano, h r . J. Hear Mass Transfer 23, 253 (1980). 133. P. LaraUrbaneja and W. A. Sirignano, 18th Symp. (1981), 13651374. 134. C. H. Waldman, 15th Symp. (1975), 429442. 135. A. Crespo and A. Lifiin, C S T 1 1 , 9 (1975). 136. C. K. Law, S. H. Chung, and H. Srinivasan, C & F 3 8 , 173 (1980). 137. P. Sepri, C & F26, 179 (1976). 138. C. A. Cooper and J. F. Clarke, J. qfEnq. Math. 11, 193 (1977). 139. H. Wise, J. Lorell, and B. J. Wood, 5th Symp. (1955), 132141. 140. D. R. Kassoy and F. A. Williams, AIAA Journal 6, 1961 (1968). 141. B. N. Raghunandan and H. S. Mukunda, “Analysis of Bipropellant Droplet Combustion with Variable Thermodynamic and Transport Properties,” Rept. N o . A E C&P4, Department of Aeronautical Engineering, Indian Institute of Science, Bangalore (1974); C & F 30, 71 (1977). 142. C. K. Law and H. K . Law, CST 12,207 (1976); C & F29,269 (1977). 143. F. E. Fendell, M. L. Sprankle, and D. A. Dodson, J. Fluid Mech. 26, 267 (1966). 144. F. E. Fendell, D . E. Coats, and E. B. Smith, AIAA Journal6, 1953 (1968). 145. R . Natarajan and T. A. Brzustowski, CST 2, 259 (1970). 146. N. Frossling, Gerlands Beitr. Grophys. 52, 170 (1938). 147. W. E. Ranz and W. R. Marshall, Chem. Eng. Prog. 48, 141, 173 (1952). 148. G. A. Agoston, H. Wise, and W. A. Rosser, 6th Symp. (1957), 708717. 149. S. Okajima and S. Kumagai, 19th Symp. (1982), 10211027. 150. C. K. Law, CST 17, 29 (1977); AIChE Journal 24,626 (1978). 151. J. F. Rex, A. E. Fuhs, and S. S. Penner, Jet Propulsion 26, 179 (1956). 152. N. V. Feedoseeva, Adcances in Aerosol Physics 2, 110 (1972). 153. T. A. Brzustowski, E. M. Twardus, S. Wojcicki, and A. Sobiesiak, AIAA Journal 17, 1234 (1979). 154. K. Miyasaka and C. K. Law, 18th Symp. (1981), 283292. 155. J. J. Sangiovanni and M. Labowsky, C & F 47, 15 (1982). 156. H. H. Chiu and T. M. Liu, CST 17, 127 (1977). 157. R. Samson, D. Bedeaux, M. J. Saxton, and J. M. Deutch, C & F 31,215 (1978). 158. M. Labowsky, CST 18, 145 (1978). 159. M. Labowsky, CST22,217 (1980). 160. A. Umemura, S. Ogawa, and N. Oshima, C & F 4 1 , 4 5 (1981). 161. A. Umemura, 18th Symp. (1981), 13551363. 162. T. P. Pandya and F. J. Weinberg, Proc. Roy. SOC. London 279A, 544 (1964). 163. H. Tsuji and I. Yamaoka, 12th Symp. (1969), 9971005; 13th Syvp. (1971), 72373 1. 164. A. Melvin, J. B. Moss, and . I . F. Clarke, CST4, 17 (1971). 165. J. H. Kent and F. A. Williams, 15th Symp. (1975), 315325. 166. K. Seshadri and F. A. Williams, J. Polymer Sci. :Polymer Chem. Edition 16, 1755 (1978). 167. R. E. Mitchell, A. F. Sarofim, and L. A. Clomburg, C & F37,227 (1980). 168. H. Tsuji, Prog. Energy Combust. Sci.8, 93 (1982).
References
91
P. M. Chung and V. D. Blankenship, Phys. Fluids 9, 1569 (1966). F. E. Fendell, Astronautica Acta 13, 183 (1967). T. M. L i u a n d P . A. Libby, C S T 2 , 131 (1970). R. W. Bilger, C & F30, 227 (1977). V. K. Jain and H. S. Mukunda, CST 1, 105 (1969). G. Damkohler, Zeit. Elektrochem. 42, 846 (1936). S. K. Friedlander and K. H. Keller, Chem. Eng. Sci. 18, 365 (1963). F. E. Fendell, J . Fluid Mech. 21, 281 (1965). J. F. Clarke. Proc. Roy. Soc. London 307A, 283 (1968); 312A, 65 (1969). J. F. Clarke and J. B. Moss, CST 2, 115 (1970). A. Melvin and J. B. Moss, 15th Symp. (1975), 625636. R. W. Bilger, CST22, 251 (1980). R. A. Allison and J. F. Clarke, CST23, 113 (1980). J. D. Buckmaster, C & F24, 79 (1975). J . D. Buckmaster and G. S. S. Ludford, Theory qf’Laminar Flames, Cambridge University Press, 1982. 184. A. Liiiin, Acta Astronautica 1, 1007 (1974). 185. L. Crocco, Atti di Guidoniu 17, 118 (1939). 186. F. E. Fendell, Chem. Enq. Sci.22, 1829 (1967). 187. L. L. Kirkby and R. A. Schmitz, C & F 10, 205 (1966). 188. R. A. Schmitz, C & F 11, 49 (1967). 189. N. Peters, C & F33, 315 (1978). 190. N. Peters, CST30, 1 (1982). 191. N. Peters and F. A. Williams, AZAA Journal, 21, 423 (1983). 192. C. K. Law, C & F24,89 (1975). 193. A. Liiian and A. Crespo, CST 14,95 (1976). 194. L. Krishnamurthy, F. A. Williams, and K. Seshadri, C & F 26, 363 (1976). 195. K. Seshadri and F. A. Williams, Znt. J . Heat Mass Transfer 21, 251 (1978). 196. F. A. Williams, Fire Safety Journal3, 163 (1981). 197. J. Lorell and H. Wise, J. Chem. Phys. 23, 1928 (1955). 198. V. K. Jain and N. Ramani, CST 1, 1 (1970). 199. D. B. Spalding and V. K. Jain, “Theory of the Burning of Monopropellant Droplets,” Aero. Res. Council Tech. Rept. No. 20176, Current Paper No. 447 (1959). 200. F. A. Williams, C & F 3 , 529 (1959). 201. V. K. Jain, C & F 7 , 17 (1963). 202. F. E. Fendell, Astronautica Acta 11, 418 (1965). 203. F. E. Fendell, AZAA JournaZ6, 1946 (1968). 204. F. E. Fendell, Znt. J . Heat Mass Transfer 12, 223 (1969). 205. F. E. Fendell, CST 1, 131 (1970). . 206. J. D. Buckmaster, A. K. Kapila, and G. S. S. Ludford, C ST 17, 227 (1978). 207. A. Liiian, Acta Astronautica 2, 1009 (1975). 208. A. K. Kapila, G. S. S. Ludford, and J. D. Buckmaster, C & F25, 361 (1975). 209. G. S. S. Ludford, D. W. Yannitell, and J. D. Buckmaster, CST 14, 125 (1976). 210. G. S. S. Ludford, D. W. Yannitell, and J. D. Buckmaster, CST 14, 133 (1976).
169. 170. 171. 172. 173. 174. 175. 176. 177. 178. 179. 180. 181. 182. 183.
CHAPTER 4
Reactions in Flows with Negligible Molecular Transport
There are a number of combustion problems in which transport phenomena (which were of dominant importance in the preceding chapter) are negligible. Since the transport effects give rise to the most highly differentiated terms in the conservation equations, these effects will generally be unimportant provided that the gradients of the flow variables are sufficiently small. How small these gradients must be depends, of course, on the sizes of the other terms appearing in the conservation equations. Large values of the convective and timedependent terms help to make the transport processes negligible. For example, in flows at high subsonic or supersonic speeds, transport effects are usually unimportant except in thin regions, such as shock waves and boundary layers, where the flow variables change very rapidly with position. The present chapter deals with systems involving negligible transport but finite reaction rates and focuses on methods of analysis, giving a few illustrative examples; textbooks [l] and [2] delve more deeply into details of some of the material to be covered here.
4.1. IGNITION DELAY AND THE WELLSTIRRED REACTOR
One simple example of an experiment in which transport effects are considered to be negligible is illustrated in Figure 4.1 and is described in detail in [3]. An oxidizing stream at an elevated temperature and traveling with
92
4.1. Ignition Delay and the WellStirred Reactor
93
._
V
Oxidizer
4
1 tv
Fuel FIGURE 4.1.
7
Schematic diagram of spontaneous ignition experiment
velocity V is brought into contact with a small amount of fuel which rapidly achieves this same temperature and velocity. An ignition delay z may be defined as the ratio of the distance L downstream at which detectable luminous emissions begin to the convective velocity V. If it can be assumed that mixing occurs very rapidly in comparison with the rate of the gaseous chemical reaction, then transport terms can be neglected, and equation (14) for the fuel implies that, approximately,
where p is the density of the reacting mixture at the temperature and pressure of the approaching stream, wF is the rate of production of fuel (mass per unit volume per second), evaluated at the temperature and pressure of the oxidizing stream and at a mean composition of the ignition zone, and AYF is the change in the fuel mass fraction that occurs prior to ignition. An estimate of the overall reaction rate during ignition can therefore be obtained by measuring V and L. Values of the ignition delay, obtained in experiments of this type that were performed at different temperatures, have been used in conjunction with equation (1) to deduce overall activation energies for the ignition of numerous fuels in air [3]. Various aspects of these ignitiondelay experiments deserve further discussion. As explained in Section B.2.5, the chemical kinetics occurring during the ignition delay are complex and generally involve significant chain branching. Implicit in the writing of equation (1) is the idea that however complex the chemistry is, it can be related approximately to the disappearance of fuel, possibly to a small extent, prior to ignition. This interpretation is not essential to the use of z;for example, it would be equally possible to interpret l/z as the ratio of the rate of production of a critical chain carrier to the concentration of that carrier needed for ignition. The key aspect is that it must be possible to identify, in principle, a single variable that describes the progress of the ignition reactions. When this variable is normalized so that it begins at zero and achieves a value of unity at ignition, then the time rate of increase of the normalized progress variable Y is
dY/dt
=
l/z.
(2)
94
Reactions in Flows with Negligible Molecular Transport
Use of z = L/V in equation (2) involves an approximation of relatively small changes in properties throughout the ignition region in Figure 4.1. Equation (2) affords the possibility of employing ignitiondelay data to calculate ignition processes in nonuniform flows [4]. Elements of gas containing fuel are identified and followed in a Lagrangian sense. For any such element, integration of equation (2) implies that ignition will occur at a time tI defined by
Jr(dt/~) = 1.
(3)
Application of equation (3) to different elements provides a value of t , for each element; the minimum of these values is the shortest time for ignition to occur in the nonuniform system. Approaches of this type have been employed for estimating the chamber length required for combustion of a liquid injected into a supersonic stream [4] and for identifying conditions for ignition of streams of droplets of liquid fuels in oxidizing gases [S]. These approaches inherently lack accuracy because of their neglect of interdiffusional effects but often are convenient approximations, particularly for configurations that are too complex to be amenable to analyses that are more precise. Another example of a system in which attention is focused primarily upon the rate of the chemical reaction is the conceptual wellstirred reactor [6]. It is supposed that by introducing extremely highintensity ' ; the thermodynamic properties can turbulence into a reactor of volume 9 be made to be practically uniform throughout the reactor. The rate of consumption (mass per second) of fuel in the reactor will then be given by  w F^y; where the thermodynamic variables appearing in the expression for wF [equation (1S)] are the average values of these variables in the reactor. For steady flow, the mass per second tizF of fuel entering the reactor would therefore have to be
riz~ = WFv/AYF,
(4)
where AY, is the difference between the inlet and outlet mass fractions of fuel. This is the fundamental equation of the wellstirred reactor; it is sometimes coupled with an overall energy balance in order to determine the heatrelease rate. The difficult problem in the analysis of any given system of this type is to ascertain the appropriate average properties within the reactor for computing wF. A considerable amount of attention has been devoted to this problem, and a number of results have been obtained [ 6 ] , c71, CSI.*
* The earliest detailed theoretical analysis of the behavior of the wellstirred reactor appears to be that of Y. B. Zel'dovich and Y. A. Zysin, Zhur. Tekhn. Fiz. 11,493 (1941); see [7].
4.1. Ignition Delay and the WellStirred Reactor
95
An experimental apparatus that under certain conditions approximates closely a wellstirred reactor is a spherical chamber into which premixed reactants are injected centrally through a perforated tube and out of which the reaction products flow through a number of ports located at various positions in the outer shell. The mixing is driven by the fuelinjection jets that are formed at the tube perforations, and the apparatus has been called the jetstirred reactor. The average residence time of an element of gas in the reactor is
TR =
pv/mF>
(5)
where p is the average density inside. Measurement of the average temperature and pressure inside and of the flow rate and inlet and outlet compositions for various values of riz, enables z R and W , to be obtained as functions of temperature, for example. In this manner, overall rates of reaction are derived. The overall rate of consumption of fuel obtained in a wellstirred reactor differs from that of an ignitiondelay experiment because the combustion rather than the ignition of the fuel is investigated. Overall rates of other processes, such as pollutant production, also have been obtained in experiments with jetstirred reactors by measuring levels of the appropriate products in the outlet streams. Results of some experiments on jetstirred reactors are consistent with the concept of the wellstirred reactor [6], [9][17]. Some of the more recent work cited has endeavored to account for influences of imperfect mixing. Theoretical aspects of the wellstirred reactor have been presented in a recent book.* Questions of the degree of perfection of mixing have plagued the use of the concept of the wellstirred reactor. In general, high turbulence intensities, small turbulence scales, slow rates of reaction, high reactor temperatures, small amounts of heat release, and relatively weak dependences of rates on temperature favor achievement of experimental results to which the concept can be applied, since under these conditions mixing rates are enhanced in comparison with reaction rates, and influences of nonuniformities within the reactor are minimized. Further information on favorable conditions may be found in Chapter 10. In the applications that have been discussed here, high rates of transport have, somewhat paradoxically, favored attainment of conditions under which analyses neglecting transport effects can be applied. The rapid transport helps to achieve conditions of uniformity, under which transport no longer is significant, and effects of finiterate chemistry can be studied. This same kind of situation prevails in various other experiments, such as those employing a suitably designed turbulentflow reactor [18], [19], [20]. In
* Y. B. ZePdovich, G. I. Barenblatt, V. B. Librovich and G. M. Makhviladze, The Mathematical Theory of Combustion and Explosion, Moscow: Nauka, 1980.
96
Reactions in Flows with Negligible Molecular Transport
contrast, for the problems considered in the rest of this chapter, low rates of transport (in comparison with rates of convection and of reaction) favor attainment of conditions under which transport phenomena are negligible.
4.2. REACTIONS IN STEADY, QUASIONEDIMENSIONAL FLOW
4.2.1. Steadystate, quasionedimensional conservation equations
The equations governing the steadystate, quasionedimensional flow of a reacting gas with negligible transport properties can easily be obtained from equations (119)(122). When transport by diffusion is negligible ( D i j + 0 and DT, t 0 for i, j = 1, . . . , N ) , the diffusion velocities, of course, vanish [K + 0 for i = 1 ,. . . , N , see equation (114)]. If, in addition, transport by heat conduction is negligible (1 + 0) and qR = 0, then the heat flux q vanishes [see equation (115)]. Finally, in inviscid flow ( p + 0 and K + 0), equations (116)(118) show that all diagonal elements of the pressure tensor reduce to the hydrostatic pressure, p I 1 = p 2 2 = p33 = p . The steadystate forms of equations (120), (121a), and (122) then become
u dujdx
dx (h
and
=  (ljp)
dpjdx,
(6)
+
);
dY
= 0,
(7)
l ., ..,N , dx P respectively. In obtaining equations (6) and (7) body forces have been neglected (fi = 0); this assumption, which will be retained throughout this chapter, is an excellent approximation for practically all problems of the present type. On the other hand, the use of qR = 0 excludes many problems in radiation gas dynamics [2], [21]. The form assumed by the continuity equation, equation (119), can be derived formally by integration over the total crosssectional area of the flow. The limits of the coordinates x 2 and x 3 [which appear in equation (113)] in such an integration must be independent of x because the boundaries of the cross section are streamlines and must therefore be parallel to the local x coordinate (that is, parallel to the local velocity vector). Thus, since the flow variables are independent of the coordinates x 2 and x 3 ,multiplication of equation (119) by g2g3 followed by integration over the crosssectional area shows that
u'=l ~ . =
w. ,
* numerical solutions for property profiles have been obtained (for example. . equations (6)(9) constitute N + 3 firstorder ordinary differential equations for the N + 3 unknowns (i = 1. equations (110) and (111) combined with the definition h = u + p / p ] relate the thermodynamic properties h and p to p.2. If equation (A5) is applied to a closed system consisting of a unit mass of the reacting fluid. . T. . and (i = 1.which is the total crosssectional area of the flow. . diverging portions of nozzles. and Wi is the molecular weight of species i. the double integral appearing here is simply ds.4. if all flow variables are specified at some point in the flow (for example. E A ( x ) . finiterate processes influence the choking behavior in the vicinity of the throat in a manner that is explained in [l] and [2]. The formalism involved in * These problems and approaches to their resolution are discussed in Section B.2 2 Schematic diagram of nozzle flow. N ) . i= 1 N where s is the entropy per unit mass. the regions where effects of finiterate chemistry are greatest . and v. Therefore. QuasiOneDimensional Flow 97 I I I I I I I 1 FIGURE 4. N ) . then we obtain dh = T ds + (l/p) d p + C (pi/WJ dY. p. Reactions in Steady. Hence the continuity equation is 11 d(pvA)/dx = 0.. For a number of purposes it is convenient to replace equation (6) by an alternative expression. at point 1 in Figure 4.. equation (19)] and the caloric equation of state [for example. T.. Although computational difficulties arise in forward integration procedures if some of the reactions approach equilibrium too closely. In view of equation (113). d s . thenin principlea stepwise forward integration of equations (6)(9) [employing the subsidiary relations given in equations (18)(112)] will yield all flow variables at any other point (x) in the channel. pi is the chemical potential of species i. .5. These solutions are restricted primarily to supersonic. .2. [22][26]).2) and if the area A ( x ) is given as a function of x..2. in steady flow. (9) Since the equation of state [for example.
N = N i is the total number of moles in this closed element. = y1 . where M is the mass of the closed element. The righthand side of equation (10) vanishes in two important limiting cases: (1) frozen flow. equation (8) can be replaced by an appropriate set of atom conservation and chemical equilibrium equations (see Section A. and (2) equilibrium flow.. in which pi d N i = 0 in a closed system by definition [see equation (Al6)]./N)/(M/N)] = d ( X i / F ) = d(&'K) = dx/vc/. For frozen flow. 1 while for equilibrium flow.98 Reactions in Flows with Negligible Molecular Transport deriving the form given for the last term in this expression is as follows: dNJM = d ( N i / M ) = d[(N.1).)dX = 0 N by the argument indicated above. W = M/N = X i q= is the average molecular weight. impling that zr= i= 1 1(pi/wT. for example.7 and the last part of Section 2. Equation (10) can also be expressed in an integrated form. h (h. For either frozen or equilibrium flow. 0 .)' dv equations (6) and (7) and the preceding thermodynamic relationship yield T ds/dx which can be written as + C ( p i / w )dYJdx = 0. for either frozen or equili . + v2/2 = ho pvA = (1 1) = total stagnation enthalpy) and vc1 (12) (riz 5 mass per second flowing through the channel). Since zr= X/w. equation (8) simply reduces to y.Therefore. i=1 N N ~ds/dx = (pT)' i= 1 1 (pi/W.1.3. in which the chemical reaction rates wi all vanish. and use is made of Equation (112) in obtaining the nexttolast equality.)wi (10) in view of equation (8). s = so (13) (so = initial entropy). Equations (7) and (9) can always be written in integrated forms.
h. In place of the parameter T2. and therefore u2 is divided by the standard gravitational acceleration in obtaining Isp. for example. is conventionally expressed in units of seconds. are desirable. and (19) that T2/T. for example. the impulse delivered to the rocket per unit weight of propellant ejected. 4./riz = v 2 . (c. when some of the reactions are frozen and the rest are in equilibrium. For isentropic flow [equation (13)] of a onecomponent ideal gas with c p = constant. Specific impulse of rockets One practical reason for studying chemical reactions in nozzle flow is to obtain accurate predictions of the thrust of a rocket motor. * . (7). h.2. apart from a dimensional constant.). the value of u2 is relevant. . of course. (and therefore for ISp) can be derived if a onecomponent gas with a constant specific heat flows through the nozzle.. Since (from cp = c. Hence. = specific heat at constant volume).T in this case. An important rocket performance parameter is the specific impulse I. Since h = constant + c. the thrust (force in the x direction) acting on the motor is rizv. = c pT. [29]. that is. Useful methods for computing the flow variables in these limiting cases are given in [27] and [28]. . followed by a nozzle (see Figure 2) in which the hot products of combustion are expanded to a high velocity. .. The quantity I .2. through equation (1 1) which may be rewritten as From equation (14) a simple and explicit formula for v. equation (14) yields * This same simplification also occurs. QuasiOneDimensional Flow 99 brium flow.4. where Ro is the universal gas constant and W is the molecular weight.. which is seen to be given by I. See. Reactions in Steady. it is well known and readily derived from equations (6).* and all property profiles can be determined without solving any differential equations. = mv. this value can be related to h . A force balance on the rocket motort shows that provided the gas is expanded to the local atmospheric pressure at the nozzle exit (station 2).2. because the atmospheric pressure (to which p 2 has been equated above) is usually specified in discussing propellant performance.(1 . it is generally more convenient to use p . equations (6)(9) reduce to entirely algebraic relations. = (p2/po)(y1)/y. . + Ro/W). A rocket motor consists of a chamber in which propellants burn.TJT. where the ratio of specific heats is y = cp/c.' Propellant combinations with large values of I.
a low molecular weight a high pressure ratio p o / p 2 (Po % the pressure in the combustion chamber of the rocket) and a low ratio of specific heats y all favor high motor performance. [30].3. these criteria must depend on the reactionrate constants. wall friction in the nozzle flow.p . It may be worth emphasizing first that variable specific heats and chemical reactions in the nozzle are not the only causes of departures from equation (15) in real rocket motors. then differential equations must be solved. regenerative cooling of the walls by liquid propellants. 4.100 Reactions in Flows with Negligible Molecular Transport (aside from conversion factors).). T. there has been interest in deriving simple criteria for determining whether frozen or equilibrium flow is a good approximation for any given reactiont and in developing approximate methods to account for small departures from either of these limiting cases. Nearequilibrium and nearfrozen flows Let f j represent any flow variable (for example. . the reader is referred to textbooks [32361. The occurrence and consequences of shock waves in the nozzle under certain conditions has been neglected. or effort may be placed on developing computer programs for solving the appropriate set ofalgebraic equations. Clearly. We have also neglected the effects of incomplete combustion in the chamber (a nonequilibrium. This topic will be considered in the rest of Section 4.2. to find Zsp. For background material and for discussions of these additional phenomena. heat transfer to the walls of the nozzle. and [30] are available for determining Zsp. when p 2 differs from the local atmospheric pressure). formulas for I . Since the computational procedures then become more complex. the comparison of measured composition profiles with profiles calculated for frozen and equilibrium flow provides information concerning the reactionrate constants through the use of these criteria.2. v. which are unknown for many reactions. [28]. We have not mentioned the modification to I. I. which shows that a high stagnation temperature To (To = the adiabatic flame temperature of the propellants).* If reactions proceed at finite. We have not considered ‘‘offdesign performance” (that is.. In such cases. attention may be focused on obtaining effective mean molecular weights and specificheat ratios that will again make equation (15) approximately valid [27]. caused by the divergence of the flow streamlines at the nozzle exit. burning mixture entering the nozzle). Thus a tool for measuring chemical reaction rates emerges [31]. The latter approach has given the results that can be used with greatest accuracy [28]. procedures such as those discussed in [27]. It is important to discover how equation (15) is modified when the many simplifying assumptions introduced in its derivation are removed.and consider situations in which the initial conditions at a selected position in the diverging * In deriving these procedures. and so on. nonzero rates in the nozzle. For frozen or equilibrium nozzle flow of a multicomponent idealgas mixture with variable heat capacities. in general.
in order to define nearequilibrium flow. Reactions in Steady. and the deviations in all other flow variables are related algebraically to A X . The magnitude of the difference A h = f. The general equation that has been given symbolically here can be expanded for lAh/fj. s is constant in the first approximation.i is a function of the local equilibrium and nonequilibrium flow variables (but not of the rates). finiterate chemistry may influence rn by modifying throat conditions. where F. caused by the area change. consistent with the fact that each G j = 0 if all Afj = 0.e ( x ) and are presumed known. the only remaining differential equations are those in equation (S). . and correspond to chemical equilibrium. General thermodynamic stability conditions require that Ge. possesses special properties. while the area function A ( x ) is specified. and the corresponding G j is found to be of higher order than G j for other variables when an expansion is made for small values of Ah/&. enow ) be introduced in equation (16).e A dx may then be derived. (ll). and construct linear combinations of the A yi that serve to diagonalize the coefficient matrix. Let M be the number of independent linear differential equations thereby obtained.2. the first term on the righthand side of this equation results from the rate of change of f j .el G 1. and the last term accounts for the tendency of f j to approach h. Therefore. e. F j = 0 for the entropy.. expressions of the form d (%) = (1dA F j ) .* The solutions to the equilibriumflow equations are denoted by f j . and Ge. and (12) (along with state and rate equations) for describing the flow because the expansion of the entropy s.2 are given. QuasiOneDimensional Flow 101 portion of the nozzle shown in Figure 4. > 0 in this form. Since s is constant in equilibrium. Qualitatively. The form of the expanded equation depends on the flow variable selected. The differential * With upstream stagnation conditions and A ( x ) given. thereby deriving for the new variables where Fe. It is most convenient to employ equations (8). often only one independent reaction is neither frozen nor in equilibrium. i j ( A ~ /may ~ . This effect is not addressed here. .is assumed to be small compared with the magnitude offj.4. and Gi = ~ i G e . and G j is a function of these properties and of the reaction rates as well. For brevity of notation let us now suppress the subscript j . The oneterm expansions F j = Fe. are evaluated from the local equilibriumflow variables. under the influence of the finite reaction rate.Gj dx fj. are held fixed. as given by equation (lo). From the conservation equations. (lo).
/A)dA/dx I G .102 Reactions in Flows with Negligible Molecular Transport equation for A f / h given in equation (17) is linear and possesses the solution f= exp( f e J o x G. Thus the deviation from frozen conditions does not depend principally .6 . f . in general. One obtains and where f is again an arbitrary flow variable. = t and ( F J A ) dA/dx L G. a sufficient condition for these two inequalities to hold is that that is. and therefore the criterion given in equation (19). Since G. Thus at A f /f. equation (17). from the prescription indicated above is complicated. G is proportional to the reaction rate. depend on the flow variable that f is taken to represent. dx’) d x (18) if A f / f e = 0 at x = 0. since integrals over x appear. . IAf/fel I t if and only if the magnitude of the righthand side is I t. However.f’. ] exceeds its maximum permissible value t.:)Jdx2 0. we require d(Af /J. we require d ( Af / . and G. the subscript f identifies quantities in the frozen flow.. A development for nearfrozen flows.f/.c at A f l f . dependent only on local properties. but unfortunately this criterion for nearequilibrium flow depends on the history. for nearequilibrium flow. parallel to that just discussed for nearequilibrium flows. ) / d x I 0. to find the functions F. d x ) Jo A F. = c. It may be emphasized that the functional forms of F . A reasonable condition is the requirement that d 1 A. Equation (19) constitutes an upper bound on the fractional rate of change of area. even for simple reactions. These conditions can be written as (F. An approximate condition for the occurrence of nearequilibrium flow.c. can be obtained from the differential equation for A f / f . and A f 3 f . This equation explicitly gives the fractional deviation of a property f from its local equilibrium value f. whichif satisfied everywhereis sufficient (but not. t at A f Jf .I/dxbe negative whenever 1 A f J f . and G. and at A f If. > 0 and t > 0. = .f. d dx A ex.] 2 t everywhere.( 1 : G. Equation (18) shows that given a small positive number t. is much simpler. necessary) to ensure that IAfif. = .
(25). . but instead is roughly proportional to the distance (or time) available for a departure from frozen conditions to occur.2.4. and (26) yield . enthalpy. Reactions in Steady. equations (18). For the reaction under consideration. Only the concentrations of the trace species involved in the reaction are affected appreciably by the rate of the reaction. because the reaction then has a negligible effect on the pressure. . it can be inferred that where the rate constants refer to the forward reaction A ./dx. density. identifying species A and B by the subscripts 1 and 2. the value of Y. (B1 l). Equations (22).(T) are known functions of x and are the same in equilibrium and nonequilibrium flows (since only trace amounts of A and B are present). As an explicit illustration. (23). 4.Jx) = ( c x)/(1 + K.= K. equations (8) and (21 1) yield We need not consider dY. = W. and the last equality follows from the fact that W. For equilibrium flow.+ 3.4 Application to the reaction A in only trace amounts $B with species A and B present The calculations discussed in the preceding section are particularly simple for reactions involving species that are present only in trace amounts. is the equilibrium constant for partial pressures. From rate formulas [for example. QuasiOneDimensional Flow 103 on how rapidly the frozen flow variables change with x. The other flow variables may then be considered to be known functions of x in investigating the effect of the reaction rates on the trace concentrations. respectively. velocity.(n = YJYl (24) for this reaction [see equation (A27)]. . Here K .). 1i=3 N Here (i = 3./x. and (Bl3)]. since Y2=lY1i= 3 cx N and are known for the species not entering the reaction. . for this reaction. is given by equations (23) and (24) : Y1.2. The equilibrium condition is x. let us consider the reaction A e B. N ) and K. and other variables.
. we are interested in AY..f = constant. = Y.Yl. ..f.. brium flows. Y. to retrieve the previous notation. in equation (30) define F. = . . equation (27) becomes . .l 4 1. = Y. For frozen flow. where Y l . the criterion for near equilibrium [equation (19)] is for the present reaction. = Yl.. we are interested in AY. f is the initial value of Y. where AY.d In Y./dx/d In Aldx.Y. where now AY../Y. In particular. The equivalence of equations (29) and (16) is apparent. where is a known function of x and is the same in equilibrium and in nonequili. For nearequilibrium conditions. Equation (27) shows that When lAYl/Yl.104 Reactions in Flows with Negligible Molecular Transport for nonequilibrium flow./Y. the terms on the righthand side of equation (29) will be of the same order if whence equation (29) becomes approximately for nearequilibrium flow. . Thus the transcription of the results of the preceding section for nearequilibrium flow becomes obvious.... For nearfrozen flow.. In terms of this variable.
./dx I < according to equation (37). QuasiOneDimensional Flow 105 is a known function of x. These criteria play roles in design criteria to maximize nozzle performance. 1 v dY.depending on whether oFor wB becomes small in frozen flow. wF z wE.3 and illustrated in Section 4.2.)wb/p. i = 1. that is. It has been found that for threebody atomic recombination reactions (primarily due to the cubic pressure dependence of the recombination rate). Freezing of reactions The reasoning outlined in Section 4. respectively [for the forms of these terms./dx = WF  WE. The intermediate freezingpoint condition is therefore determined by either 1 v dYJdx 1 % wF or 1 v d Y.v. . combining equations (8) and (211) yields vdx/dx = W(V’. On the other hand. (37) where wF = W. see equation (B1 l)]. that is. wE = W.4 was preceded by many approximate criteria that are less precise for the attainment of nearequilibrium flow [37][40]. When rate data are available. either wF 4 wE or wE G w F .v:)o’/~.2. When the flow is near equilibrium.2. and wk and wk are the forward and backward rate terms in w‘. In order to determine the flow variables for flows that exhibit this “suddenfreezing’’ character. . Reactions in Steady. either I v d YJdx 1 % wF or lvdY./dx( % oE. When only one reaction occurs. Letting species 1 represent a product of the reaction. it is straightforward to estimate freezing points of this type [24]. N . When freezing occurs . .’ .(vl.5. . 4./dx / 2 w E .and ( v dYJdxl is of the same order of magnitude as the larger of the two rate terms according to equation (37).)wk/p.4. after which the remaining portions of the composition profiles are practically frozen [24].(vl. we can write equation (36) as v dY. (36) where w‘ is the difference between the forward and backward rates of the reaction./dx I < wF and I v dY. nearequilibrium and nearfrozen criteria are unimportant. when the flow is near frozen. it is sufficient to obtain a condition specifying the location of the freezing point . . Thus equation (20) and the corresponding results of the preceding section are obtained. A criterion for nearfrozen flow can be written as for this reaction.2. the composition profiles in a supersonic nozzle follow equilibrium curves very closely until a point is reached at which a relatively rapid transition to frozen flow occurs.v.
This is evident from the fact that equation (13) applies in both equilibrium and frozen flow and is a further manifestation of the exceptional attributes of entropy in nozzle flow. directed toward predicting the increase in entropy that occurs during freezing. A. for example. 831 (1968). Lee.. There is a continuum of isentropic nozzle flows. A I A A Journal 6.. . S. J . In the vicinity of the freezing point a perturbation may be sought in the time interval needed for freezing to occur. Cheng and R. approximately. Therefore. This approach has not yet been pursued.6. in the first approximation sudden freezing occurs at the point where equilibrium calculations predict that the first term in equation (38) equals the second or the third. All the quantities appearing in equation (38) can be evaluated from the equilibrium flow variables and the chemical kinetic formulas for mF and 0. ranging from fully equilibrium to fully frozen flows. although detailed analyses that treat the ratio of the thermal energy to the heat of dissociation as a small parameter provide important ideas that would be helpful in beginning the study. e % W B . K . 4. 17. Good agreement of suddenfreezing predictions with experiment has been obtained for NO. A curious consequence of the suddenfreezing approximation is that the flow remains isentropic.* IUe(dY1dx)el % m F . the location of the freezing point parameterizes the continuum. recombination [41] but not for ion recombination [25]. the quantities may all be evaluated in nearequilibrium flow. e . it appears to be of interest to discuss twophase flows briefly. (38) where the subscript e identifies equilibrium flow. Gasparticle flows have much in common with reacting flows. c p and c. 823.106 Reactions in Flows with Negligible Molecular Transport suddenly. Twophase nozzle flow Many rocket propellants form solid or liquid products of combustion either in the combustion chamber or in the nozzle. in precise analogy with the effects of finiterate chemistry. This condensation tends to improve rocket performance by releasing heat but introduces the complications of twophase flow in the nozzle. Moreover. aerodynamic drag on the particles and heat transfer from the particles increase entropy. This observation suggests an approach to obtaining an improved description of influences of finiterate chemistry on nozzle flow. containing solid or liquid particles with constant * See P. 20. Blythe.2. at the freezing point. 126 (1963). with regions of equilibrium and frozen flow separated by a transition region. the twophase effects are numerically larger and often of greater practical significance in contributing to losses in performance. 243 (1964) and H. and the freezing condition then becomes. As a first step consider a onecomponent ideal gas with constant specific heats. The result would be a more accurate description of the property profiles. Nuid Mech.
For illustrative purposes let us assume that thermal equilibration occurs but allow for a particle velocity. and the equation of state in terms of p becomes the same as that for an ideal gas with molecular weight W/Z./c. is the specific gravity of the condensed phase. as may be seen by algebraic manipulation. Thus in equation (15) W is replaced by W/Z and y by [Zy (1 . (6).. In terms of the number density of particles. cs. v. where ps is the gas density. so that ps x Zp.Z increases.p(vq . = l/P> (39) where p. but the altered equation of state modifies equation (15). (42) . Since equations (9).2. = constant. respectively. + Z/P./dx = 6nr.4. n. that are sufficiently small to maintain the same velocity and temperature as the gas and that do not exchange mass with the gas through condensation or evaporation [28]. Equation (39) states that the total volume per unit mass is the sum of that of the condensed and gaseous phases.)].. Generally (1 . QuasiOneDimensional Flow 107 specific heat. the mass flow rate of particles is n s ( $ n r ~ p s ) u s= A viz.Z)c. but the consequent increase in I. and energy balances. different from that of the gas.Z)c. If the particles are spherical. T + (1 . [36].. Neglecting the contribution of Brownian motion of the particles to the pressure. as the effective specific heat at constant pressure. The penalty is greater if the particles cannot follow the changes in gas properties. in accordance with the expression for h. (40) where p is the gas viscosity.)]/[Z+ (1 .Z)(c. (41) The mass flow rate of the gas is (pW/RoT)v.dv. where Z is the mass of gas per unit mass of the system./c. even in this the factor most ideal situation. This relationship may be used in equation (14) for v 2 . the gas is excluded from the volume occupied by the solid or liquid. + (1 . while retaining all the other approximations indicated above. a performance penalty is associated with the twophase flow.Z)(e. momentum. from equation (19) we may write p = psTRo/W. in this approximation the derivation of equation (15) remains valid. and the Stokes law of drag applies. The enthalpy per unit mass of this gasparticle system is h = constant + Zc. and (7) express appropriate mass. v.. Therefore. = constant. for the twophase system. This effective y decreases as l . all with radius r. subject to the indicated replacement of molecular weight and to the use of Zc.A = viz. generally is less than the decrease associated with that arises from the replacement of W.Z)p/ps4 1. T. which is related to the total mass per unit volume P by (1  ZYP. Reactions in Steady. then a force balance on a particle gives + fi $nr3psu.. [33].us).
. and generalizations of the characteristic methods discussed below become useful. then us = v.2/2) + kg(c. Effects of temperature lags are similar to those of velocity lags. and T. We shall return to twophase flows in connection with spray combustion (Chapter 11) to give a more systematic development. axisymmetric flows must be considered in analyzing these processes [47][50]. and I may be developed [28]. such as riq'(Adp/dx).dv. [52] (see Section 9. pDv/Dt pDu/Dt = = (47) . If the lag is neither large nor small. Vp./dx + kgdu. ug. is sufficiently largethen in the first approximation us is approximately constant through. + h. 4. Perturbations about this smalllag limit have been studied because of potential practical importance [42]. In view of the simplifications that arise when transport effects are absent (see Section 4.2. is low [28]. Perturbations about this largelag limit out the nozzle. Among additional topics of interest in twophase flows are questions of sound propagation in twophase media [28]. Equations (40)(44) serve to determine the five unknown variables us. If the former is much smaller. then numerical integration is generally needed [45].108 Reactions in Flows with Negligible Molecular Transport and the total rate of mass flow is m = viz.4) and many other problems [53][55]. From equation (40) a characteristic time lag or response time for velocity equilibration can be identified as 2r.(c./dx = Adp/dx (43) (44) and k. Particles may fail to follow streamlines and produce undesirable effects by impinging on walls. equations (11)(14) become DplDt + pV * v = 0. threedimensional flow of a homogeneous reacting fluid with negligible transport properties. Condensation and agglomeration processes also may be important [46]. REACTIONS IN UNSTEADY.. Conservation equations.3. This time lag can be compared with a representative residence time in the nozzle.u. and the results described after equation (39) apply.T + u. characteristic surfaces Let us now consider the unsteady.1.1).for example. THREEDIMENSIONAL FLOW 4. [43] and have been considered in connection with optimized nozzle design [44].2pS/9.3. if Y. [28].T + $12) = constant. [Sl]. p(V * (45) (46) v). The generalizations of equations (6) and (7) for overall momentum and energy conservation are viz.1. n. p . in the first approximation. If the former is much larger.
With this interpretation.(j # i ) are meant to be proportional to derivatives with respect to the mass of species i in the system at constant values of the masses of a11 other species in the system. it is convenient to write equations (45) and (47) in alternative forms.Yi> (51) the frozen volumetric thermal expansion coefficient (which equals 1/T for an ideal gas mixture) is given by Pf = PCa(l/P)/aTI.. ThreeDimensional Flow 109 and pDY. all N of the K’s are independently variable in the differentiation. (48) respectively. = (aP/aPlS. In order to illustrate most easily the mathematical properties of (45)(48).2. and the restriction CF=. where the substantial derivative is defined as opt = a/at + v . . Reactions in Unsteady.p. By substituting equation (49) into the mathematical identity* DT Dt = ($)s. Yi? (52) * Here and in what follows./Dt = wi. .yiE jl(%)s. at constant r. The righthand member of equation (49) is the rate of production of entropy by the irreversible chemical reaction.v. When the same kind of reasoning used in deriving the expression for dh. i = 1 . Equation (49) is a useful replacement for equation (47). = 1 is not to be imposed in evaluating such derivatives. derivatives with respect to rl.. .4.yj(j+ + one can show through straightforward thermodynamic arguments that where the frozen sound speed af is determined by a. employed in obtaining equation (10) in Section 4. we find N Substituting this relationship and equation (45) into equation (47) gives where use has been made of equation (48) in the last equality.N . .3. is applied to equation (A3).yi% + (%)p.
). u ~ u. the p . . Equations (48) and (49) are already in characteristic form.) = 0. (a.. specific enthalpy of species i [that is. ' . Equations (46) and (54) involve derivatives only of p and v. x aY)ko (ax.  = 0.* they show that composition and entropy changes propagate only along streamlines. (49).. .)=O.x.thus the term o i w i is a measure of the rate of liberation of heat by the chemical reaction. N _ where i= 1 (54) (55) For an idealgas mixture with a constant average molecular weight..110 Reactions in Flows with Negligible Molecular Transport and the frozen specific heat at constant pressure is (53) Substituting equations (45) and (48) into equation (50) yields the expression c p . and (54) may be taken to be the governing differential conservation equations. (56) (57) w%..f = (ah/aT). . (48). ~ . . 2. has a characteristic surface.... ax. defined by the equation x1.' the characteristics of these equations require further study. the quantity hi given in equation (1 . and the dimensionless parameter oi becomes oi = T ) . .. for example. Equation (54) is a useful replacement for equation (45) in studies of the characteristic surfaces of the governing equations. i = 1 . see. Y. (58) * For a general discussion of characteristics and for the definition of characteristic surfaces with an arbitrary number of dependent and independent variables. It is known [56] that an arbitrary system of quasilinear partial differential equations which may be written in the form xr=. u ~ . + terms in lowerorder derivatives +f. x.. [56].. Equations (46)... y j ( j + i ) appearing in equation (55) is simply the (total) term (ah/aYi)p. ax. . N .(xo. . The use of the method of characteristics in solving equations is considered in [57]. This is the reason that equation (54) is more useful than equation (45) in the present context.ll)].. x ..
is unity in a term of the summation in equation (56). at 3 _  and + a ( . . u. ax. . . x. . .+ v1 __ + v2at ax. u 2 . In this simple case. ul. In equation (56) there are n + 1 independent variables denoted by x . ax.ax. x l . ax. where k takes on the values 0. (59) which is a firstorder partial differential equation for the characteristic surface. .k . . In this case. and equation (58) becomes i = 1 .. . ... .1. . + vlax. We may then use a new notation which is obtained by letting k. ul. u. +v 2 ax. . . equation (56) may be written as (ko.n) and N dependent variables denoted by u j . . k. u1.+ aP ax. in that term must be zero. then every other k . av3 ax.N . . + v2. t i 2 .xn . ThreeDimensional Flow 111 where vertical lines represent the determinant of the matrix within them. .u N ) = 0. 2 . if one k. . + f = 0. ax. P a x . . . the highest derivatives are all of the first 1). . (64) . may assume only the values 0 or 1. Reactions in Unsteady. . n and is equal to the 1 value of the only nonzero k l . x.  av. When there are no derivatives of order higher than the first (nj I 1 for all j ) . . av. av.3. . while fi(xo. .x l . av. . Equations (46) and (54) may be written in full in Cartesian order ( n j I coordinates as follows: av.. . . . a ~ . In equations (46) and (54). . d (xo. 1 aP v. . u N ) are the coefficients of the highest order ( n j ) derivatives of u j in the ith partial differential equation. .4. (1 = 0. . Pax. . . The quantities A $ C J . kn) + J . ~ I . uN) represents the terms in the ith equation that involve no derivatives at all. + . a ~ .... x i . : ( x ~ .. .. the integers k.. 1. . . + v 3 av 1 ap 2+ = 0. av + 1 a p = 0. + 213 2) +++av. aP ax.
. Dt P ax. . .. 0 L’l 0 0 01  P 0 1’2  0 0 0 1 Pa: Pa: Pa: Here the subscripts 1.x l . By comparing equations (61)(64) with equation (59). we obtain the following relationship defining the characteristic surface. By substituting the values of AfJ(xo. uN) given in Table 4. . 3.1 into equation (60). It is seen from equations (64) and (54) that f = oiwi/p.): z=. and u j = vj for j = 1. and the subscripts on v denote the components of v in the corresponding directions. . u.. the A f j may easily be determined. and 3 on x represent the various coordinates. 2. . If xo = t . respectively.. 2. x. . .1). Y(t. u 2 .xl.112 Reactions in Flows with Negligible Molecular Transport TABLE 4. while u4 = p then a table of the Afj may be constructed (Table 4.1 j = 1 k=O k = l 01 j = 2 k = 2 02 k = 3 03 k=O k = l k = 2 k = 3 i= 1 i=2 i = 3 i=4 1 0 0 0 0 1 0 01 0 02 0 1’3 0 0 1 j = 3 0 0 0 0 0 0 0 0 0 0 j = 4 0 1 0 0 Ik=O k = l k = 2 k = 3 I k=O k = l k = 2 k = 3 i= 1 i=2 i=3 i=4 0 0 1 0 0 0 02 0 0 03 0 0 0 1   1 0 1 ~ P 0 0 1 P 03 Pa.x. i = 1 through 4 is allowed to correspond to equations (61)(64). x 2 .
[(”)z  ax.4. and since enthalpies are usually easier to compute than are entropies. and (54) as the governing differential conservation equations. z 0. . 4. Reactions in Unsteady. in which equation (46) may be employed to show that D(h + v2/2)/Dt . equation ( 6 8 ) then reduces to (P)* a. (48).( V Y ) * ( V Y ) = 0. The other solution to equation (66) may be written as (DY/Dt)’/a? . (67) which shows that streamlines are characteristic lines of the equations. ThreeDimensional Flow 113 Evaluation of the determinant yields One solution of equation (66) is DY/Dt = 0. + ax. equation (69) reduces to the relationship + v 2 / 2 ) = 0. By substituting equation (45) and the identity h = u + p / p into equation (47). That this choice is not convenient for unsteady flow is apparent from the fact that equation (69) involves ap/ & and therefore would have to be * This result becomes more transparent if the coordinates are transformed in such a way that the fluid is brought locally to rest.3. it is generally convenient to employ h + v2/2 in place of s as a new dependent variable. (70) which states that the total stagnation enthalpy ( h + v2/2) is constant along streamlines. (68) which defines characteristic surfaces that propagate normal to themselves with the frozen sound velocity af relative to the local massaverage velocity of the fluid.3. + (q(?)’I ax. equation (49) may be replaced by an expression that is both simpler and generally more useful.(l/p)Dp/Dt = 0. v * V(h (69) For steady flow. adopting equation (70) and the steadystate forms of equations (46). we obtain the expression Dh/Dt . Since equation (70) is simpler than the steadystate form of equation (49).2.( l / p ) a p / a t = 0. twodimensional (axially symmetrical and plane) flows For steady flows. The method of characteristics for steady.* This last result has been emphasized by Wood and Kirkwood [SS] and by Chu [59].
..). vl+v2=~. v1 be the component of velocity in the x direction. dv] = . The first relation in equation (78) may be used to eliminate the velocity derivative from equation (80). w. Then equation (70) and the steadystate forms of equations (46). and (54) become.114 Reactions in Flows with Negligible Molecular Transport included along with equations (46) and (54) in calculating the characteristic surfaces for the pressure and velocity equations. ar ar. q. For steady flow with axial symmetry. 5. In terms of the magnitude of the velocity. r be the radial coordinate normal to the axis of symmetry. v = (of + v . In view of equations (75) and (76).d x sin 8 v2 = v sin 8 + dr cos 8. we find d[ = d x cos 8 + dr sin 8. (48). (75) (76) and U. = u cos 8. and the angle between the flow direction and the x axis. in place of x and r as independent variables. 8 = tan’(u. ) l i 2 . equations (71)(74) can be shown to reduce to respectively. let x be the axis of symmetry. (see Figure 4./u.3). the result is . and u2 be the component of velocity in the r direction. respectively. P i = l . and ax ar.N . .. (73) It is convenient to use the distance in the streamline direction. and the distance normal to streamlines.
Their characteristic lines may easily be obtained by the methods of the preceding section. ThreeDimensional Flow 115 r'yl \ FIGURE 4..3.[ l / ( p u ) ] Ap (where the symbol A indicates a small increment). (79). The finitedifference forms of equations (77). is the Mach number referred to the frozen speed of sound. . and (81) may be taken as the governing set of differential conservation equations for steady. . Along a streamline (Aq = 0).1 0 = 0. Equation (81) and the last relation in equation (78) constitute two independent equations for the derivatives of 6 and p and have characteristic lines [57] given by 1 0 0  1 PV2 0 1  M2 . (82) (83) (84) i = 1. equations (77) and (79) and the first relation in equation (78) reduce to total differential expressions and become. (x. V X Schematic illustration of the definitions of the (x.3. A(h A T = [ w i / ( p v ) ]A t . and (5.4. and (81) are useful for obtaining numerical solutions by the method of characteristics for supersonic flows. . for Aq = 0 and Av = . . (79). + u2/2) = 0 for Aq = 0. (78). (78). systems. for Aq = 0. respectively. Reactions in Unsteady.r ) . N . Equations (77). q) coordinate where M = u/a.y). axially symmetrical flow.
and (86) is very similar to the method of characteristics for nonreacting rotational flows [60]. (84). properties throughout a region of the shape indicated in Figure 4 3 a ) may be obtained. The extension of the streamline through the point C in the upstream direction (at angle 8. By computing the length Atc of the line DC. the properties p D .By a linear interpolation along the line AB. the basic step in the method of computation is as follows. (the pressure and the flow angle at point C).4. and y) appear in these equations. u. .* Here Al* . (83). When M > 1 [so that the values of Aq given by equation (85) are real]. Since the distances AC and BC are known. uD. and h. 8. one finds by substituting equation (85) into the finitedifference forms of equation (81) and of the last expression in equation (78). from equation (82). (84). In this manner. All properties must be known at two neighboring points.) is then found to cross the line AB at a point D. which cannot be connected by a single streamline or Mach line. equations (82). and (81). A and B in Figure 4. hD. equation (86) determines the values of pc and 8. The explicit numerical calculation procedure for solving equations (82). which defines the Mach lines based on the frozen sound speed. It is worthwhile to consider what kinds of boundaryvalue problems 4 are correctly set for equations (77). and one may proceed to the next step. All properties at the point C are now known. there are no novel basic differences in the refined methods. A total of N dependent variables ( p .and so on may be found approximately at point D. one may then determine vc from equation (84).which is the geometric distance along the Mach lines. (83). These Mach lines will intersect at a point C . subsidiary + * Expressions such as equation (86) are intended to represent two equations. h. by Au = fA C I J F T . Along these lines. that the relationship must hold. One may then construct Mach lines of opposite families at points A and B. While many refinements of the procedure outlined above yield better accuracy [60]. the principal difference is that here entropy varies along streamlines as well as normal to them. (79). is given by Al* = 4 M Aq along the lines Av] = A < / J m .116 Reactions in Flows with Negligible Molecular Transport that is. the new point at which all properties are to be computed. and (86) can be used to obtain a stepwise solution to the conservation equations. (78). one obtained by reading the upper signs and the other obtained by reading the lower signs. & from equation (83). Briefly.
see Figure 4. (c) Variables specified along intersecting Mach lines. neither a Mach line nor a streamline.3. (b) Variables specified along streamlines and Mach lines. 0 given .4. relations (such as the equation of state) enable us to express all other quantiand wi) in terms of these dependent variables. (83).4. oi. We have ties (namely. all given (b) _ _ _ _ _ Mach lines Streamlines Boundary of region of solution all given (c) FIGURE 4. (84). (a) Variables specified along a noncharacteristic line. and are determined along streamlines. in many practical + / / /’ // / 1 p . It follows that a correctly set 4 dependent Cauchy initial value problem [57] is one in which the N variables are given along a noncharacteristic curve [that is. and (86)] that relationships between the differentials of p and 0 exist along Mach lines and that the differentials of u. Unfortunately. h. M . Reactions in Unsteady.5.5(a)]. . shown [equations (82). p. ThreeDimensional Flow 117 ___. Elementary step of the characteristic calculation. Mach lines Streamlines  FIGURE 4.
5(b) and (c). and (91) are used in obtaining numerical solutions by applying the method of characteristics in the x . we obtain the following pair of differential equations : + (U +_ ax = u. as is illustrated in Figure 4. and (91) constitute a useful set of governing conservation equations for characteristic analyses of one dimensional. Explicitly. then we need only neglect the curvature terms in order to obtain equations that are valid in plane flows. 4. and equations (88) and (89) become Ayi = (wi/p)At/2. our equations and results can be made to apply to steady. In this case. plane. i= 1 N (91) Equations (88). (89). . Ax = u At. of course. u sin B/r in equation (80). i = 1. . N aspt + u aslax = (PT)' 1 (pily)wi. . twodimensional flows. unsteady flows Let us next consider onedimensional. i= 1 (87) (88) (89) and respectively. and we do not have a simple Cauchy problem. Along the streamlines. . Data of this last type can.3. . (92) .118 Reactions in Flows with Negligible Molecular Transport problems data are given along characteristics. we delete the terms pu. unsteady flows. N . C oiwi. The method of characteristics for onedimensional. The finitedifference forms of equations (88). Thus if the radial coordinate r everywhere is replaced by y . The discussion is not modified in any way by these changes. unsteady flows with velocity u in the x direction. and equation (87) by af and then add and subtract the resulting equations. and sin e / r in equations (81) and (86). (49). t plane. axially symmetrical flow.3). Throughout this section we have been discussing steady. also determine solutions in finite regions. axlat + u ayilax = wi/p. (89). Procedures slightly different from that described in the preceding paragraph are required in solving such problems numerically.3. the Cartesian coordinate perpendicular to x (see Figure 4. . for Ax = v At. and (54) reduce to p aqat + p u aqax + aplax = 0. i = 1 . If we multiply equation (90) by a. equations (46). N . With only a few minor modifications in the formulas./r in equation (74). . (48). .
) At..4.1.e [see equation (A26)]. that is. Onedimensional. and changes in pressure and velocity propagate only along Mach lines.(Y/Kp)1. 4. = W (equal molecular weights). For simplicity.) (97) r. . Corresponding analyses for a simple dissociating diatomic gas [62] and for a gas experiencing an arbitrary onestep reaction [59] may be found in the literature.w l = WBITa'l(p/Ro)eE'IRoT [ ( I . we shall consider here only a binary ideal gas mixture experiencing the reaction A B.3. i= 1 N for Ax = v At.Y pressures is given by = X.e/Pl. reacting idealgas mixture for the reaction A B 4. Analyses are greatly facilitated by utilizing the special properties [equations (92)(94)] of the differential equations along the characteristic lines [61]. and the equilibrium constant for partial = x2. X i = we then have N = 2. Reactions in Unsteady.e/xl. changes in entropy and composition propagate only along streamlines.)At.4. and [from equation (18)] w2 = . P = P(R0/W)T (95) [from equation (19)]. ThreeDimensional Flow 119 and As = (l/pT) i= 1 c (pi/K)wiAt/2 1 oiwi At/2. KA1  r. = + KJ. . Letting the subscripts 1 and 2 identify species A and B.3. Preliminary relationships Much information can be gleaned from the linearized forms of equations (87)(90) that are applicable for small departures from a uniform quiescent state of chemical equilibrium.e = K .4. (94) Thus the streamlines and Mach lines again take on special physical significance. 1 . Similarly. Wl = W (i = 1. The reader may consult [61] for details of procedures in which equations (92)(94) are employed in obtaining numerical solutions and for discussions of correctly set initialvalue problems. respectively.3. along the Mach lines. (96) + where Y = Y.Y) . which are defined by Ax equation (9 1) assumes the totaldifferential form N Ap f p a f Av = u? for Ax = (v k a. unsteady sound propagation in a binary. 2). = X 2 . All the essential physical phenomena can be exhibited in this special case. = P2. = (93) (v k a.
Linearization In order to linearize equations (87)(90). Equations (95)(98) comprise subsidiary relationships that are required in the analysis. we have u = u'. wheref.) ap'iat = (qlc. for any dependent variable f appearing in these equations. ar'lat asilat = (99) ( w y Y' 0.3.h Z . and (90) become. this result has already been emphasized in Section 4. Since the quiescent value uo of the velocity u is u. In obtaining equations (100) and (102). + wppl + wssypo.4. Reduction to a single partial differential equation t = The solution to equation (101). the righthand side of equation (90) becomes N  i= 1 1 ciwi = (q/cp._= C(aw2/as>. respectively. . In deriving equation (101).3. use is made of the fact that the righthand side of equation (89) is of order higher than the first since w2. 4.is s' = 0. = 0 for chemical equilibrium [see equation (A21)]. is aconstant (independent of x and t ) that corresponds to quiescent chemical equilibrium conditions and f ' is small.and w. we then find that equations (87). 'Ow2 5 (98) where use has been made of equation (96) and the heat of reaction (per unit mass) has been defined as q = h . = 0.. Po avlat + ap//ax = 0. / ( ~ ~ . Equation (98) has also been employed in equation (102). 4.p1).120 Reactions in Flows with Negligible Molecular Transport in which the subscript e identifies chemical equilibrium conditions. and we shall therefore omit the prime on the small quantity u.s10? w p = c(aw2/~P)Y. ~for . = 0). (102) where WY = C(aw2/ay>. Neglecting terms of order higher than the first in f ' and in each of its derivatives. (100) (101) = and po av/ax + (ila.3. ( w .slo. ~ ' wpp' + + W. (88) for i = 2.. Since ci = h .ylO. use has been made of the fact that the unperturbed state is one of chemical equilibrium (w2. = 0 at (103) . (89). = 0 and ( p z .SO.4.2. r ~ ) . we shall set f = fo + f ' .3...2. 7 ideal ' ) gases with constant molecular weights [see equation (55)Q. subject to the condition s' 0 for all x.
that 40 dY + C. we obtain dT = (l/PO) dP. In order to relate 6 to the equilibrium sound speed a..fdT.3. present gas. (105) where use has been made of equation (103) (ds = 0) and of the fact that the coefficient of dY arising from the last term in equation (A5) is zero at chemical equilibrium. we can then easily identify the coefficients of Y' and p' and can thereby obtain wy and w p . for ~ = the ~ . and. we find from equation (96) that for small departures from equilibrium.[1/(1 + K . + Y2h2. when equation (A5) is applied to a unit mass of the gas near equilibrium. dT. let us compute l/a.In this way. for our present purpose. 1/d = 1/4 + ( 8Y / @ ) . depends only on temperature.w20 = [WB. In view of equation (51). and 6 will now be shown to be simply related to the difference between the frozen and equilibrium sound speeds. ThreeDimensional Flow 1 121 Therefore. = ( 8 ~ / 8 p ) ~ . 0)1(4~/R0T2)0T'. Since the quantity in brackets in equation (96) is zero in equilibrium. By solving equation (105) for T' and substituting the result into equation (104).4.p ' (108) . we find [in view of equation (95)].we find by differentiation that dh 4 dY + ~.h. and the frozen sound speed a f . respectively.. (104) where use has been made of the fact that K . we find that the last term in the braces in this expression reduces to .. and p'. T'. Employing equation (97) and the van't Hoff equation [equation (A49)].Y = Ye( 8 P P u s . w2 . therefore. = Since h = Y.T"l'(p/Ro)eE1'RoT]o x {[I + (l/Kpo)IY' + (Yo/K%o)(dKp/dT)o T'I. and d p appearing in equation (105) may be identified with Y'.fO and that (107) The quantity z is clearly a representative reaction time. Reactions in Unsteady. the only derivatives of w 2 that we need to evaluate for use in equations (100) and (102) are wy and w. The quantities dY.
f)O.122 Reactions in Flows with Negligible Molecular Transport Here (aPlay).@W/RO T 2 ) . thus. and (A49) (the van? Hoff equation). In view of equation (103) after differentiating equation (102) with respect to t and subtracting the result (multiplied by po) from the undifferentiated form of equation (102) (multiplied by wy).)2/K..o + 6.f~o~'>/(~42>1~1. .4) (R0T/WP4){(c.(aP/an. (110) where use has been made of equations (99. Equation (105) may also be used to evaluate the remaining partial derivative in equation (109) at chemical equilibrium (the state of present interest). The only dependent variable appearing in equation (1 13) is the velocity v. (112) where 6 is given by equation (107). we obtain 1 1 aZv a2v (113) where use has been made of the definitions of z and 6 [equations (106) and (107)] and of equation (112).. (ay/aP>s.fdT/dY. = = = (l/P)/(cp. (97).lo = (4/c.we may make use of equation (100) to show that By differentiating this expression with respect to x and then using the first and second t derivatives of equation (99) to eliminate a2p'/axdt and a3p'/ax at2.o at2 ax2 = O .. Equation (1 13) implies that for very slow reactions (l/z + 0).f/4)(d[Kp/(l + Kp)l/dT)'  (R0T/WP4){1 + C(1 + ~. = (aT/aY). = (aT/ay)s. (109) where use has been made of equation (95). It is instructive [62] to reduce equations (99)(102) to a single partial differential equation.. u satisfies the simple wave equation __ 1 aZv  a2v ~ u...Y=.. (1 11) The substitution of equations (109)(111) into equation (108) yields the required relationship l/UZo = l/u. [ ( a m u s . ..IC(~. Since equation (105) is valid at constant s a n d a t Y = x.
o/aed4 + ( ~ z > ~ 1 /+ [ 1( ~ z ) ~ I l ) " " and CP (115) +tan.4.4. . v satisfies the wave equation 1 a2v a2v a:. 4. The velocity of the piston can be represented by the real part of v(O. (116) Ex f Reacting gas T' Piston. since they illustrate sound dispersion resulting from finite reaction rates. Dispersion relations [62] Solutions to equation (113) with a sinusoidal time dependence are of interest. / a ~ =~ 0. Generation of sound waves in a reacting gas.6.)x cos cp]}. v = V(O)e'mt or v = ( 0 for t < O V for t > O FIGURE 4. On the other hand. Harmonic solutions to equation (1 13) will then be given by (the real part of. (the equilibrium sound speed). (1 14) = CC(a. Reactions in Unsteady. where V(0) is the constant (small) amplitude of the piston velocity perturbation and o is the (angular) vibrational frequency. at2 ax2  0 with the propagation speed a.3.3.'{C(a. t ) = V(x)eiW'. t ) = V(0)eiw'.4.6).o/aeo>2. Suppose one generates sound waves in a gas occupying the region x > 0 by oscillating a piston harmonically about the point x = 0 (see Figure 4. (1 + i o z ) d 2 V / d x 2+ ( w / ~ ~ ~ ) ~ [ ( a + ~icoz]T/ . for very fast reactions (z + 0).. for which Vremains bounded as x + 00.)  From the solution to this equation.. where. by substitution into equation (113).) v(x. ThreeDimensional Flow 123 with a propagation speed equal to the frozen sound velocity. t ) = V(O)exp[ (wcc/afo)x sin cp]exp{io[t where (a/a.1I~z/"(a~o/aeo>' + ( ~ 0 z ) ~ I ) . we obtain v(x.
for intermediate values of w. Suppose again that the gas occupies the region x > 0 and that a piston is located at x = 0.(a. a . The product (07 appears in equations (1 15) and (1 16) as a dimensionless parameter. which. The piston is given an impulse at t = 0. aeo < a < afo .[I  ( u . 4. An initialvalue problem [59] A more thorough insight into the physical significance and implications of equation (1 13) is best obtained by solving an initialvalue problem instead of by studying the dispersion relations further. the gas is entirely quiescent and is in chemical equilibrium. The solution may easily be obtained by the Laplace transform method. upon rearrangement. Since no disturbance is imposed at x = a.4.a 2 U / d X 2 ] + (l/7)[(s/ae0)2u  a2U/dX2] = 0. Since the boundary conditions on u become u = I/ e”‘dt = V/s at x = 0 and u + 0 as x x. If the Laplace transform of the velocity isdefined as u(x. the solution to equation (117) is Jr  + u = (V/s)exp[(sx/aro){l [l . afo and 1 + 2af0 ~/[(afo/aeo)2 . causing the velocity at x = 0 to assume the constant value V for t > 0 (see Figure 4. reduces to (1 . The consequent dependence of the phase velocity on the frequency [see equations (1 15) and (1 16)] represents sound dispersion. (no attenuation). as w + 30. thus indicating that nearequilibrium sound propagation will occur for frequencies such that 07 < 1 and that nearfrozen conditions will prevail if wz $ 1.I] . ~ / u ~ ~ ) ~ ] ST’)} / ( I aZu/ax27’ + = 0. Various specific results follow directly from equations (1 14)(116). (118) . s) = s.” e%(x. we require u + 0 as x + m for all (finite) values oft.5. a + aeo and 1 + rx.o/afo)27 is a slightly modified representative reaction time. we obtain s[(s/afo)2u . t ) dt. For t 5 0. t ) to equation (113).3. As w + 0. (1 17) where = (a.o/afo)2]/(1 + SZ’)”’’~]. These initial and boundary conditions determine a unique solution u(x.124 Reactions in Flows with Negligible Molecular Transport This result shows that the phase velocity of the disturbance is a = ufo/(a cos cp) and that the disturbance attenuates as x increases with a characteristic decay length given by I = afo/(wa sin cp). where s is now the Laplace transform variable. then by taking the transform of equation ( 1 13). and the attenuation illustrates sound absorption.6).
the sign of the real part of the argument of the exponential in equation (119) is negative when ls’l + co with Re{s’) > 0. we then find that the velocity is given by x exprs’t‘ . for x > afot). the sharp frozen wave is a close approximation to reality. the reader may show that where erf denotes the error function * erf(z) = ( 2 / ~ i ) e .SIX’{ I . ThreeDimensional Flow 125 By taking the inverse transform. and the simple pole at s‘ = 0 and branch points at s‘ = . and the contribution from the semicircular arc along which Is’/ + m vanishes. the main part of the signal travels with the equilibrium sound speed x/t = aeo. Thus some disturbance always propagates with the velocity afo. By means of an interesting application of the method of steepest descents.3.( U . The intermediate transition . At small values of t / z . Reactions in Unsteady. When x’ < t’.4. In order to find the shape of the wave front at large values of x and t . analogous reasoning shows that the contour of integration must be closed to the left. For x‘ > t’ (that is. the dimensionless time is t’ = t/z‘. Since t’ is the dimensionless time variable appearing in equation (119).1 and at s’ = .7. The magnitude of the pulse at x = a f o t decreases as t increases. no disturbance propagates faster than the frozen sound speed x/t = afo. and the width of the signal broadens (as x and t increase) in proportion to 1/&. Therefore. the asymptotic shape given by equation (120) is attained earlier (at smaller values o f t and x ) for smaller values of the reaction time z. one may perform an asymptotic expansion of the integral in equation (1 19) for t‘ and x‘ approaching infinity with the ratio x‘/t’ fixed.Here y is a positive constant such that there are no singularities of the integrand to the right of the line s’ = y in the complex s’ plane. (1 19) where s’ = sz‘. The contour of integration may then be closed to the right. It is now clear that the time development of the sound wave behaves as illustrated in Figure 4. while at large values of t/z.[I  (U.~ dy.~/U~~)~]/ s ’ ( ) ) ~ + ~ 1’23 ds‘ls‘. 0 [ Equation (120) shows that at large values of x‘ and t’. ~ / U ~ ~ will ) ~ provide a nonzero contribution to v. and the dimensionless distance is x’ = x/afoz‘. thus showing that u = 0 for x’ > t‘. the spreading equilibrium wave closely represents the true wave pattern.
Lick [65] has obtained solutions to the nonlinear. many related problems can be solved. twodimensional system.4.8) can be obtained by the methods of Section 4.4. Related problems By analyses that are similar to those presented above. The wave field produced in the steady. unsteady propagation of a sound pulse into a reacting gas. Reviews of these and other studies may be found in [l]. twodimensional flow of a reacting gas past a wavy wall has been treated in [63] and [64]. 4.3. twodimensional conservation equations governing the flow of a reacting gas mixture about a blunt body.5.6. and [66]~711. The essential characteristics of the results of such an analysis can be inferred by interpreting the time t of Section 4. .. and a spreading wave develops at some distance from the wall and propagates at an angle determined by the equilibrium sound speed aeo. steady.ot FIGURE 4. This behavior is consistent with the picture developed in Section 4. region occurs when t z z.126 Reactions in Flows with Negligible Molecular Transport t" OO a. \LI' \LI' t>>T t>>>T c . Schematic diagram of the onedimensional.4. [a].3.3. twodimensional sound wave produced when a supersonic reacting gas stream is turned through an infinitesimally small angle (see Figure 4. ~ OO amt afot u .5 as the distance measured along the sound wave in the steady.7.4. the structure of the steady. For example. A decaying wave propagates at an angle determined by the frozen sound speed a f o .4.3.
Hottel. B. H.257 (1958). H. C . (1973). 9951002. 13. F. (1967). (1962). J. B. J. 5. J. Krieger. 18. 12. Williams. L. (1957). F. 970982. Odgers. 779790. Heims. 23. I. R. T. REFERENCES 1. Rocket SOC. S. W. Yumlu. S. W.. Glassman. C. Ryan. Herbert. Vulis.21. A. 9871003. 10th Symp. 10. and D. 16th Symp. 16. The Dynamics o f Real Gases. Longwell. F. Radiation and Reentry. Chem. 20. Hottel. Blazowski. and J. Jr. 3. 11th Symp. and I. Pratt. 8. F. D. Dryer. S. B. M. J. P. Note No. Williams. 15. 17th Symp. 2. and P. Vincenti and C. London: Butterworths. (1979). (1977). P. (1965). J.References Equilibrium sound wave . Eng. 0. 50. R. Introduction to Physical Gas Dynamics. Weiss. Fuel 32. Ind. L. 14th Symp. R. Chem. Williams. 47. 11. Kestin. P. Clarke and M. Thermal Regimes of Combustion. C. Spalding. P. G. T. Glassman.‘I 127 FIGURE 4. 17. (1973). (1967). Jain and D. Spalding. 37 (1965). B. 547 (1970). Am. A. 1968. Blichner. Crowe. 8th Symp. G. 4144 (1958). New York: Academic Press. Miles. C S T 2 1 . 365374. A. NACA Tech. 14th Symp. 819830. Schematic diagram of a steady. M. Santoro. Ind. J. D. Clarke. 771778. Lang. Jenkins. A. Chapter 2. Bowman. Longwell and M. Astronautica Acta 15. G. Weiss. C & F 9. 19. Kruger. (1962).451 (1953). C. 577592. D. New York: McGrawHill. . Aronowitz. H. 9. A. 1965. Mullins. Glassman. New York: Wiley. S . V. 327. 11th Symp. J. Olfe. R. S. J. E. 8th Symp. A. 633644. Dryer and I. 6th Symp. and G. 6. twodimensional sound wave in a supersonic reacting gas stream. 1634 (1955). and C. 14. F. 22. Eberstein and I. 4. J. L. 8th Symp. 1961. and M.8. K. Penner and D. Eng. B. 7. Sangiovanni and A. 87 (1980). 21. McChesney. 1964. L. V. S. 179 (1951). Baker. (1962). 982994. V. 398411.
322 (1960). 17.” in Proceedings of the Fifth AGARD Combustion and Propulsion Colloquium. Bray. Wegener. Penner and F. 1958. Phys.. “Gas Dynamics of Chemically Reacting Gas Mixtures Near Equilibrium.441 (1950). Penner and D. M. and M . Gilbert. 799810. N. 1958 Heat Transfer and Fluid Mechanics Institute. Williams. 841865. 1957. 26. Crocco. M. A. C. Lenard. M. Lancaster. 25 (1955).. S. vol. S. 37. 43. J. Bailey. Harrison. 17. I1 of High Speed Aerodynamics and Jet Propulsion. 1 (1959).. Williams. 439520. 28. A. H. (1963). 45. 20. Emmons. and J. M. F. 71. D. Franklin Inst. Fluids 2. Davis. 6 of Progress in Astronautics and Rocketry. New York: Wiley. D.2793 (1963). 44. Nilson. W. J. 9th Symp. eds. 29. M. K. 245. 25.. K. 1951. Fundamental Aspects of Solid Propellant Rockets.” in Dynamics of Manned Lifting Planetary Entry. W. A. N. 47. Rudin. 841 (1949). P. A R S Journal 31. A.” in Detonation and TwoPhase Flow. 41.” in Jet Propulsion Engines.J. M. J. Barrere. L. 42. Princeton : Princeton University Press. 35. 32. P. : AddisonWesley. 341 (1952). M. 9th Symp. Pease.421 (1948). Marble. S . Huang. S.56 (1949). 9th Symp. Chem. Franklin Inst. 34. “Perturbation Analysis of OneDimensional Heterogeneous Flow in Rocket Nozzles. A. Zupnik. Rocket Propulsion. Fraeijs de Veubeke.E. Mass. 31. Altman and S. Altman. 46. N. N. 117144. Am. Courtney. Serra. Fluid Mech. 40. L. S. Slough. A. Westenberg and S. J. Hall. B. ARS Journal 30. England : Technivision Services. SOC. F. New York: Elsevier. R. S. Phys. Favin. New York: Pergamon Press. S.” in Fundamentals of Gas Dynamics. (1963). Reading. M. Altman. Penner. : PrenticeHall. “ OneDimensional Treatment of Steady Gas Dynamics. 38. B. Stanford : Stanford University Press. Gordon and B. 111 of High Speed Aerodynamics and Jet Propulsion. E. 64349. Summerfield. Phys. S. 793 (1961). 6. 1958. E. 116. 1963.. C & F 3 . “The Liquid Propellant Rocket Engine. Barrere. Aroeste. and N. New York: Pergamon Press. J. Princeton: Princeton University Press. W. S. D. 36. Rogers. Chemistry Problems in Jet Propulsion. New York: Academic Press. C. A. 384 (1958). Penner. Mechanics and Thermodynamics of Propulsion. and D. Rannie. M. and H. AIAA Journal 1. NASA SP273 (1971). 19. F. XI1 of High Speed Aerodynamics and Jet Propulsion. P. R. R. ed. Scala. Wegener. C. G. N. P. S. 1963. McBride. 91103. F. G. Thermodynamics of Fluid Flow. 1959. Chem. 1969. and T. eds. S. Fluids 1. S. Jet Propulsion 25. Carter. N. Chem.128 Reactions in Flows with Negligible Molecular Transport 24. vol. Penner. 264 (1959). 27. W. 1956. Vandenkerckhove. 1962. Taylor. Peterson. S. E. Rudin and H. Altman and J. 30. vol. C.” in Combustion Processes. Bray. 175215. Englewood Cliffs.788 (1949). 770784. Marble. P. vol. 2663. “Expansion Processes. A. Phys. J. 273 (1959). (1963). ed. . 877 (1951). Princeton: Princeton University Press. 33. “Dynamics ofa Gas Containing Small Solid Particles. eds. 39. 785798. 1960. Jaumotte. Lewis. 249. AGARDograph No. 1965. J. 0. Hill and C.
4. 28. G. I. F. T. Carrier. Clarke. “Wave Propagation in a Reacting Mixture. R. New York: Academic Press. 1949. Hoffman and S. Annual Review of Fuid Mechanics 11.I. Williams. Nickerson. Wallis. Annual Review of Fluid Mechanics 8. 1106 (1964). Fluid Mech. New York: Interscience. Annual Review of Fluid Mechanics 4. Chapter 11. G. 377 (1976). New York: Macmillan. AZAA Journal3. Fluid Mech. 811826. 1948.8090. F. Tech. van Nostrand. R. 68. Lorenc. Phys. A. 54. Tech.” AFOSR Tech. 103 (1965). 128 (1960). 173194.577 (1960). Part B. 0. Rept. D. J. Note No. Numerical Solutions. The Mechanics of Aerosols. Soo. Appl. 376 (1958). 1969. Friedricks. “Inviscid Flow Around a Blunt Body of a Reacting Mixture of Gases. Lick. W. F. AIAA Journal2. K. 53. G. 7. A. J. J. Linear and Nonlinear Waves. 70. Fluid Mech. 27. I70 (1961). General Analysis. 6. Elements of Aerodynamics of Supersonic Flows. J. J . Partial Differential Equations. Wave Diagrams for Nonsteady Flow in Ducts. W. 61. J. 51. 581124 (R. 62. 63. Moore and W. OneDimensional TwoPhase Flow. F. S . 49. 64. Rept. 1964. Courant and K. Vincenti. New York: Wiley. 52. J. Rudinger.Z. G. S . J. Chem. No. 1974. Note No. 1958.395 (1957).P. Y . eds. 65. L. Part A. Dobbins and S. G. Gibson. Becker. 7. J. No. B.397 (1970). AE5810) and AFOSR Tech. A E 5814). New York: McGrawHill. 69. Wright Air Development Center (1957). 596 (1957). New York: Macmillan. W. Fuchs. 1962. A. E. W. S . Temkin. Aero. “Wave Propagation and the Method of Characteristics in a Reacting Gas Mixture with Application to Hypersonic Flow. 1955. 56. Wood and J.P. 50. 117 (1960). Whitham. 4. New York: D . Fluid Dynamics of Multiphase Systems. Petrovsky. 59. Kliegel and G. 71. T. G. Li. New York : Interscience. R. Chu. Rensselaer Polytechnic Institute.” Tech. 66. B . 9th Symp. 155 (1972). 58522 (R.27. Troy (1958). Sci.” in Detonation and TwoPhase Flow. 481 (1959). Section 3. Annual Review of Fluid Mechanics 2. Chapter I. R. J. Waltham. : Blaisdell. G.References 129 48. R. . 60. “Flow of GasParticle Mixtures in Axially Symmetric Nozzles. Marble. Phys. J. D. Crighton. Kliegel. Stanford : Stanford University Press. Kirkwood. Penner and F. 58. 1954. B. Note No. G. Fluid Mech. Ferri. vol. 57213. 67. A.” 1958 Heat Transfer and Fluid Mechanics Institute. V. Supersonic Flow and Shock Waves. 6 of Progress in Astronautics and Rocketry. 1967. 11 (1979).. Rusanov. E. V. 55. 169 (1966). 57. Mass. E. N. A R S Journal31. (1963). A.
the earliest combustion problem to require the simultaneous consideration of transport phenomena and of chemical kinetics was the deflagration problem. knowledge of the concepts developed and results obtained in laminarflame theory is essential for many other studies in combustion. The problem of determining the propagation velocity of a deflagration wave was first studied by Mallard and le Chatelier [l]. In more recent years. improved basic equations became available for use in theoretical analyses. Timedependent and multidimensional effects are considered in Chapter 9. or laminar flames. The essential result that the burning velocity is proportional to the square root of the reaction rate and to the square root of the ratio of the thermal conductivity to the specific heat at constant pressure was first demonstrated by Mikhel'son [2]. [4]. whose work has been discussed in more recent literature [3]. Attention here is restricted to the steadily propagating. the idea that the overall activation energy is large in comparison with the thermal enthalpy has led to the development 130 . constitute the central problem of combustion theory in at least two respects. planar laminar flame. Deflagrations. Subsequently. who considered heat loss to be of predominant importance and rates of chemical reactions to be secondary.CHAPTER 5 Theory of Laminar Flames This chapter concerns the structures and propagation velocities of the deflagration waves defined in Chapter 2. First. Independent investigations by Mallard's student Taffanel [S] and by Daniel1 [6] based on simplified models of the combustion wave reached the same conclusion. Second.
Description of Laminar Flames 131 of asymptotic analyses that lend further clarification to the wave structure and that provide formulas with improved accuracy for the burning velocity. . 5. The necessary modifications caused by the presence of chain reactions are discussed in the final section. a combustion wave spreads through the gas. wave speed. the terms are used interchangeably here.3 emphasizes the utility of asymptotic concepts. a criterion for the validity of the steadystate approximation for reaction intermediaries is described. Provided that the tube is not too short and the electrical energy released by the spark is not too much larger than the minimum energy required for ignition. Much of the recent work is covered in [lo]. DESCRIPTION OF LAMINAR FLAMES 5. The implications of this idea were first derived heuristically by Zel'dovich and FrankKamenetskii [7]. and propagation velocity. as synonyms for such terms as dejlagration velocity. many outstanding questions remain concerning deflagration structure. This choice may be justified by the fact that morecomplicated processes may be treated by extensions of the approaches developed. They all refer to velocities measured with respect to the gas ahead of the wave.5. We shall begin with a brief discussion of experiments and a simple indication of the primary physical phenomena involved. Experiments When a quiescent combustible gas mixture contained in an open tube is ignited by a spark at one end of the tube. The development of laminarflame theory to be given in Section 5.1. Systematization of the concept through formal introduction of the method of matched asymptotic expansions was achieved first by Bush and Fendell [9]. This constant speed is an empirical laminar burning velocity or flame speed. the combustion wave often is observed to be nearly planar and to travel at approximately a constant speed between (very roughly) two and ten tube diameters from the spark. Next. P = product). It will be seen that although useful approaches to the theoretical analysis are available. Many ad hoc approximations that rest in varying measure on the concept of large activation energy are reviewed in [S].1. A number of recent studies of deflagration structure have been based on asymptotic methods of this type. appropriate basic differential equations are derived for describing the structures of combustion waves.1. P (R = reactant. in particular. The essential features of the mathematical problem of predicting the propagation speed of a onedimensional laminar deflagration wave will then be presented by examining in detail the unimolecular process R .1. which is characteristic of the combustible mixture. Therefore.* * Recent efforts to distinguish between the terms burning velocity and pame speed on the basis of Eulerian and Lagrangian coordinate systems appear to introduce confusion.
a Bunsentype burner with an elongated rectangular outlet. The component of velocity normal to the flame surface. An alternative apparatus that provides burningvelocity data of greater accuracy is the Bunsentype burner schematically illustrated in Figure 5. and a steady. typically photographic. the experiment just described does not yield accurate burning velocities.1. and various Laminar flame Fuel Oxidizer experiment.132 Theory of Laminar Flames Because of various inherent difficulties.1. is the velocity of propagation (burning velocity) of the flame. conical flame sits at the mouth of the cylindrical burner if the exit velocity exceeds the flame speed. the curvature can be eliminated by use of a slot burner. The mixture composition is determined by adjusting the flow rates of the fuel and oxidizer (often air). accompanied by measurements. The component of velocity tangent to the flame surface remains constant across the flame. of the geometry of the flame surface (for example. . This propagation velocity can be determined from measurements of the flow velocity at the exit of the burner tube. as is illustrated by the vector diagrams in Figure 5. but the component of velocity normal to the flame increases. Numerous experiments of this type have been performed. The radius of curvature of the flame cone is generally so large compared with the width of the flame that the flame structure is the same as the structure of a plane flame except at the tip of the cone and at the burner port. just upstream from the flame. the apex angle of the flame cone).
Laminar deflagration waves may be produced in a number of other experiments. the steady flow is contained within a duct. .1.5. Photograph of a natural gasair flame on a rectangular burner.2. by a wire or rod inserted perpendicular to the flow direction. the propagation velocity of the wedgeshaped flame determines the minimum chamber length required for the * See [l t At a velocity below the burning velocity the flame would move upstream andjushbuck would be said to occur. Lewis and of Academic Press). In ramjet combustors. procedures for calculating burning velocities from the measurements have been developed [11]. and one or more stabilizing rods are employed. Description of Laminar Flames 133 FIGURE 5. flowing steadily at a velocity greater than the burning velocityt. a steady deflagration may be stabilized in a combustible mixture. I] for a discussion of the techniques and the accuracy of each. courtesy of B.* A photograph of a typical laminar flame produced in this way is shown in Figure 5. with very small magnesium oxide particles introduced into the flow and illuminated stroboscopically in order to make the gas streamlines visible (from [l 11.2. For example.
2 is produced by buoyancy. the turbulent burning velocity (Chapter 10) must be employed. flatflame. is ignited in the open or in a very large chamber. In using this technique. with data reduction performed on the basis of an assumed spherical shape of a flame of negligible thickness. possibly enclosed within a soap bubble. * A simple solution exists only for spherical symmetry. and horizontally in chambers usually behave differently from each other because of buoyant effects. A configuration useful for studying laminar flames in the laboratory is the flatflame burner. downward. This arrangement affords a good research tool because it offers the experimenter a relatively large number of variable parameters and facilitates analysis and interpretation of measurements by virtue of the plane. inert stream. Variants of the flatflame burner that also are useful in the laboratory are stagnationflow devices. . in which a flow with a stagnation point is established by directing a uniform combustible stream either normally onto a flat plate or counter to the flow of a uniform. Motions of propagating laminar flames and shapes of steady flames often are influenced by buoyancy and by various instabilities (Chapter 9). elliptic problem even with inviscid fluid mechanics if there is an appreciable change in density across the sheet. and stagnationflow burners are their tendencies to minimize these effects.* Nevertheless. to calculate histories of flame shapes for slow propagation of a flame sheet at a specified flame speed from an initially uniform and quiescent state in a closed chamber with fixed walls is a nontrivial. the flame speed is vo = (p.. then the rate of increase of the mass of burnt gas inside a spherical flame of radius r is 47cr2p0v. it is best to verify by observation that the flame is nearly spherical. onedimensional geometry. and since the mass of burnt gas is 47cr3pm. in which a planar flame is stabilized above one or between two porous plates in a flow that is essentially onedimensional./po) dr/dt.134 Theory of Laminar Flames combustion to be completed inside the duct. Even in the absence of both gravity and instabilities. Advantages of some Bunsen. the flame produced near the stagnation point is flat locally and can be arranged to have negligible heat loss but is “stretched” in that streamlines diverge in the vicinity of the flame. The curvature of the streamlines in the upper part of Figure 5. or “bomb” (typically spherical). The complication of pressure variation with time can be eliminated by the soapbubble method. if po is the density of the unburnt mixture and v o is the flame speed. approximate values of flame speeds often are obtained by monitoring the pressure histories in a closed chamber. Flames propagating upward. in which the combustible mixture. but since the flow in practical ramjets is turbulent. a constant fraction of the observed growth rate of the spherical flame. In this experiment. walls influence flame motions in chambers. but heat loss to the porous plate generally is significant and must be taken into account. where pm is the density of the burnt gas.
and the subscripts 0 and co identify conditions upstream and downstream of the wave.2. into which a combustible gas mixture flows at velocity uo (the laminar burning velocity). leading to a crude estimate of the burning velocity. where p o is the density of the initial combustible gas mixture. The cold incoming gas is heated to a temperature at which it reacts at an appreciable rate by means of heat conduction from the reaction zone. If the wave is adiabatic so that no energy is lost downstream or from the sides of the wave.1. then energy conservation implies that q = cp(Tm. then qw 6 is the energy released per unit area of the wave per second.TJ6. x is the distance normal to the wave. that is. respectively. Our primary objective in this chapter is to predict this burning velocity. Phenomenological analysis of a deflagration wave Consider a stationary plane deflagration wave of thickness 6. .1.1.To)/6 (energy per unit area per second). If q is the heat of reaction (energy released per unit mass of the reactant mixture) and w is the reaction rate (mass of the reactant mixture converted per unit volume per second). The mass of combustible material per unit area per second flowing into the wave is p o u o . and composition of the initial combustible mixture.To)/&These results lead to the equation c. The discussion. which varies roughly as eE1’RoTm. T is the temperature. 5. and Ro is the universal gas constant [see equation (18)]. . The factor on the righthand side of equation (2) with the strongest temperature dependence is usually w. yields Equation (2) shows that the burning velocity is proportional to the square root of a reaction rate and to the square root of the ratio of the thermal conductivity to the specific heat. At constant temperature.2 a simple physical picture of the deflagration wave is presented. the combustion wave propagates at a definite velocity that empirically depends on the pressure. in conjunction with equation (l).To)(where c p is an average specific heat of the mixture) and that the entire heat released must be conducted upstream. The deflagration wave consumes these reactants at a rate w6 (mass per unit area per second).(T. temperature. Description of Laminar Flames 135 In all these laminarflame experiments. Hence mass conservation implies that pouo = w6. where A is the mean thermal conductivity of the gas. which is similar to that of Landau and Lifshitz [3]. . The rate at which heat is conducted upstream is roughly A dT/dx = A(Tm . .5. qw6 z A ( T . which implies that the thickness of the wave is The burning velocity uo can be related to this wave thickness as follows.which. where El is an activation energy. illustrates the essential mechanism involved. In Section 5.To)w6 z A(T.
values of w are so uncertain (see Appendix B) that equation (2) is more useful for estimating w from experimental values of uo . but 6 is small compared with typical dimensions of experimental equipment (for example. which was shown in the preceding section to be valid for .2. experimental studies of the interiors of laminar flames are often performed at subatmospheric pressures. MATHEMATICAL FORMULATION 5.1. (4) With representative values for A. and po . Eliminating w between equations (1) and (2) leads to the relation 6 = A/c. Thus laminar flames are nearly isobaric. and the continum equations of fluid dynamics are valid within the deflagration wave.. Empirical laminar burning velocities lie between 1 and 1000 cm/s. Since these velocities are small compared with the speed of sound. The remainder of this chapter is devoted mainly to a mathematical amplification of the concepts introduced in this section. Equations (1) and (2) enable us to estimate the flame thickness 6 from measured values of uo .    5. appearing above is the specific heat at constant pressure. we were justified in not attaching a subscript to p in equation (3). c. and w p". Thus the approximation p = constant. and the quantity c . the diameter of the burner mouth.E ~ / 2 R o T m (3) In principle.and with vo 50 cmls. Hence the pressure and temperature dependences of uo are given approximately by uo    p(n/z)l e . Therefore 6 is large compared with a molecular mean free path (about lo' cm). in practice. However. In particular. equation (125) is valid for laminar flames. Since equations (3) and (4) imply that 6 p("/') at constant temperature. and hence the radius of curvature of the flame cone. and laminar deflagration waves may be approximated as discontinuities in many experiments.2. Many questionable aspects of the reasoning employed in this section will have been remedied before the end of this chapter. equation (2) should provide a numerical estimate for uo . for experiments with Bunsentype burners). it will be seen that the burning velocity is an eigenvalue of the conservation equations. are independent of the pressure p . A useful way to deduce an overall order and an overall activation energy for a reaction is to measure the T and p dependences of uo and to correlate these empirically with equation (3) by adjusting n and El.136 Theory of Laminar Flames A and c.pouo. equation (4) gives 6 lo' cm. Introductory remarks In this section we shall develop equations capable of describing the structures of both deflagrations and detonations. where n is the order of the reaction [see equation (141. po p according to the ideal gas law [equation (19)].
Thus pu2 + p . equation (24)]. The approach is unnecessarily complicated for deflagrations only in that the momentum conservation equation must be considered and one additional approximation [assumption 8 below] must be made.($p + K ) dvldx = constant.12) are necessary subsidiary relationships among the variables appearing in equations (5)(9). The following assumptions will then be made: 1. External forces are negligible. The Soret and Dufour effects are negligible. pv (5) = m = constant. . Our principal reason for adopting this approach is to avoid repetition in Chapter 6.5. These disadvantages appear to be insignificant in view of the greater generality of some of the results [for example. if x is the coordinate normal to the combustion wave. 3. will not be made here. see [l2] or [13]. = (6) 1. The flow is steady. plane. 6. 4.2.. (132). (131). 5. . then equations (126). Radiative heat transfer is negligible. The approximate validity of these assumptions and of others to be made subsequently can be justified by orderofmagnitude considerations for many systems.2. it will obviously be expedient to adopt a coordinate system which is at rest with respect to the wave. Diffusion caused by pressure gradients is negligible. Equations (18).2. . i = and (9) where the notation was defined in Chapter 1. For a detailed derivation of the basic relations for deflagrations alone. and (1. 2. . Preliminary assumptions and equations In order to analyze the structure of a combustion wave. Mathematical Formulation 137 ordinary deflagrations but which is invalid for detonations (see Chapter 2). In view of assumptions 16. and onedimensional in the adopted frame of reference. (111). and (133) are valid with g 2 = g3 = 1 and with the last term in equation (132) omitted. (7) dtJdx wi/m. N . (127). The gas mixture is ideal. 5. (19).
c c (where properties will be identified by the subscript 0) and at x = + c c (where properties will be identified by the subscript a ) .1. Assumption 7 is not essential because the important qualitative results to be obtained do not depend on this assumption. All chemical species may be considered to have constant and equal average specific heats at constant pressure cp. for p % constant) because the terms involving the viscosity are negligibly small. it enables the analysis to be simplified considerably and. in some respects. In view of assumption 7. however. clarifies the physical significance of the results. Appendix E or [14].3).co. properties again become uniform (dldx = 0) and chemical equilibrium is reached. At x = + m. for example.1. (10) Assumption 8 is entirely unnecessary for ordinary deflagrations (that is.3. However.T.T + u2/2) as a variable.). Section 2. which is a good approximation for many gases (see.2) and the values of all variables except uo are known (compare Section 2.(cpT0 + ~&'2)' (12) . equation (1 1) implies that the Prandtl number is 2 (see Appendix E).2. 5.& + u&Q) + &/2) . As was pointed out by Spalding [lS]. + u2/2) .1) it was shown that these conditions imply that relationships exist between the properties at x = +a and those at x = . 16). equation (111) reduces to hi = hP + c. When bulk viscosity is negligible ( K = 0). The effective coefficient of viscosity ($p + K ) equals the ratio of the thermal conductivity to the average specific heat (Ale.(c. A normalized dimensionless measure of the stagnation enthalpy is Z E (c. p. Nevertheless.T (c. this assumption leads to a considerable simplification in the analysis for detonations and for flames with nonnegligible kinetic energy. some aspects of the analysis of combustion waves in which kinetic energy is of importance can be simplified by using the stagnation enthalpy (c. all properties are uniform (d/dx = 0) (see. In Chapter 2 (Section 2. Assumption 8 states that which is a relation between transport properties.TO).138 Theory of Laminar Flames The boundary conditions for these differential equations are to be applied at x = . Approximations that further simplify the energy equation Equation (9) can be simplified by making two additional assumptions : 7. At x = rx. 8. we reemphasize that assumption 8 will not be present in subsequent sections of this chapter.(T . However.
This case has been treated in most analyses of combustion waves . T + u2/2)  ( c . By using equations (lo)( 12) (along with the boundary conditions d/dx = 0 at x = c o and the identity ti= l). The gas is a binary mixture in which the unimolecular reaction R + P occurs. ) . then equation (7) provides N . Mathematical Formulation I39 which varies (monotonically for purely exothermic reactions) from 0 at x = . In order to avoid all these complications while retaining the essential physical aspects of the problem. If only one chemical reaction occurs [ M = 1 in equation ( 1 4 1 . Thus N 4 1 i= 1 h?(ti. Assumption 9 automatically reduces the number of independent relations in each of the equations (7) and (8) to one. when the analysis is extended to include multicomponent systems in which any onestep reaction occurs ( M = 1) and for which the binary diffusion coefficients .4. But. we shall consider the following system: 9. the reader can prove that equation (9) reduces to + The denominator inside the brackets in equation (13) is the total heat release per unit mass of mixture. follows directly from equations (9) and (10) and the boundary conditions. 2 . + &/2) .1 simple relations among the N flux fractions ti.m t o l a t x = +a. even in this case. = (cPT. general conclusions regarding the wave structure are difficult to obtain. simplifying assumptions regarding the diffusion coefficients (usually that all binary diffusion coefficients are approximately equal) must be made in order to decrease the number of independent relations in equation (8). (14) where the last equality. Simplifications in the energy and diffusion equations for unimolecular reactions in binary mixtures Since equations (7) and (8) constitute 2N differential equations.ti.5 . In view of equation (14). Furthermore. 0 .(c. To + U&2 ') ] / q . equation (7) reduces to a single differential equation. q.2. (124 5.and a single flux fraction is sufficient to characterize the progress of the reaction. which is simply the statement of overall energy conservation for the wave. the definition of T [equation (12)] reduces to T = [(c. T O + ~&'2).
or formally by summing equation (7) over i].6. = W2 = W [since mass is conserved in chemical reactions. (20) which is Fick's law.co to 1 at x = (17) + co. The addition of assumption 9 to assumptions 17 also renders valid all of the results of Section 2. .If the dimensionless distance m  '2. then equation (13) reduces to dz/d< = z . 2. . ~ = v . M = 1 and N = 2./A) dx (19) is also introduced. equation (20) implies that X .t 2 .€ 2 . the quantity in the brackets in equation (13) is (€2 .0)/(€2. v . assumption 9 will have to be omitted in the final section of this chapter because the distribution of reaction intermediaries obviously cannot be studied on the basis of a onestep process. o = ti. W. Since the boundary conditions require that dK/dx = 0 at x = fa.O) (21) .t 2 .Y2. X i = Y. Hence.140 Theory of Laminar Flames of the pairs of principal species d o not differ appreciably.O)/(Y2. Assumption 9 enables us to simplify equation (13) further. ~ = v . For the present twocomponent mixture.€7 where the normalized flux fraction (or. .w 2 = . i = 1. (16) where use has been made of equation (15).2. .Y2)eEI/RoT.ti).m (the diffusion velocities vanish far upstream and downstream).)l(t2. = 0. 0  € 2 . m) = 1 . However.0>7 5= Jo(mc. (15) and equation (18) becomes w1 = . . According to assumption 9. and t1 = 1 . [from equation (112)].E ~ / R 0 T ROT = pBITa'(l . in terms of the normalized mass fraction x. Letting the subscripts 1 and 2 identify the properties of species R and P. reaction progress variable) has been defined as (€2  '2.m y = (Y2  Y2. we find that v . equation (8) reduces to d & d 'x = (m/pD12)(X . ~= 1.m  Y2. respectively. Yl = 1 . Equation (19) then reduces to p = pRoT/W. there is virtually no change in the results. which goes from 0 at x variable = .. Since t1 = 1 .O and = ti.WpB1 Ta' x l e .
= 1 ( t i . .co. since the reaction goes to completion at x = co [dtJdx + 0. Equation (12a) can be used to express the temperature T in terms of rp and z.the only acceptable value of A is A = 0. 5. Since z . then equation (5) becomes dv A convenient dimensionless velocity variable is (26) which approaches unity as x + .2. Actually. implying that Y=z. has the general solution (T  Y) = Ae’. Solution of the diffusion equation when Le = + 1. (1 l). see equation (7)]. (19. where use has been made of equations (17) and (18) and the Lewis number is Le = 3L/pDI2c.Y must remain bounded as x + co. and therefore also t2.(l/Le)(dY/d() = z .5. Mathematical Formulation 141 (which goes from 0 at x = c c to 1 at x = + co).Y .6. Equation (24) provides an explicit relationship between the composition and the stagnation enthalpy throughout the wave.5. (24) which also satisfies the correct boundary conditions at x = f c o [see equations (12) and (21)].. equation (20) becomes (22) (23) (l/Le) dY/d( = Y .m= 0). The difference of equations (19) and (22) is dz/d< . 5. which.6. the result is rp = v/vo. It is shown in Appendix E that Le = 1 is a reasonable approximation. From equations (19) and (22) it can be seen that Y is simply related to z provided that the next assumption is allowed: 10. and (18). The Lewis number is unity.2. where A is an arbitrary constant. equation (16) implies that Y2. when Le = 1.2. = 0). Dimensionless forms for the momentum equation and the speciesconservationequation If the constant in equation ( 5 ) is evaluated at x = co and use is made of equations (6).m = 1 (Yl.
in the physically acceptable range negative over the entire range 0 I T I of cp. The relation between a' and the dimensionless heat release c1 used in Section 2. which appear on the righthand side of equation (34).(z. cp). because the additional variables p and T.(t. although it becomes very small (because of the exponential factor involving T ) near z = 0 where T is small. are easily expressed in terms of z.cp. 5. which determine the variations of the variables z. = 1 and equations (16)(18). cp) is non1. use may be made of Y. (21). The function F. In equation (7) with i = 2.2.. the velocity.2 is By substituting equations (26). and (33) are three firstorder differential equations.(z.cp. (31). (32) The function F.7. cp) depends only on z. as defined in equation (34). and constants through equations (6) and (27). F. and the progress of . (27).cp.2. and it equals zero only at z = I.I)]  1).142 Theory of Laminar Flames where the ratio of specific heats is the dimensionless heat release is and the initial Mach number is Mo =V O l J r n . is a function of z and cp. Summary of the simplified mathematical problem Equations (19). and (30) into equation (25). we obtain the dimensionless form of the momentum conservation equation.(z. We are entitled to indicate that F. and constants. cp) = (cp  1) + 4 {' [1 + d z YMo cp  ( q ) M b ( c p 2 . and (24) to show that Equation (33) is a dimensionless form of the speciesconservation equation. and E (dimensionless measures of the stagnation temperature.
given by equation (32)] imply that (38) and (39) In view of equation (29).1. it is easily eliminated. Boundary conditions for equations (35) and (36) are 6 = 0 and cp = 1 at T = 0. (26). Using equations (6). The Unimolecular Decomposition Flame with Lewis Number of Unity 143 the reaction.3. one obtains p T = po To = constant. T is a measure of the static temperature. ) been obtained from equations (35) and After solutions for V ( T )and t ( ~have (36). If q and dcp/d5 remain small compared with 1/Mi (as is. and (38) in equation (39). and 6 = 1 and cp = (pa. then equations (27) and (31) [with F . Since both of these endpoints are singular points of equations (35) and (36). q) = 0 and T = 1 in equation (32) or by consulting Section 2. the boundary conditions alone do not enable us to draw any conclusions concerning the existence or uniqueness of solutions. Since the distance variable t appears only as d/dc in these equations.5. 4 1. respectively) through the combustion wave. retaining all of the assumptions of the preceding section. the profiles in the physical coordinate can be found by integrating equation (19). [which is determined either by setting F. since ordered kinetic energy is negligible. thereby reducing the total number of differential equations from three to two. where T~ is the value of 7 at x = 0.(T  m/q. equation (38) states that T = c. THE UNIMOLECULAR DECOMPOSITION FLAME WITH LEWIS NUMBER OF UNITY 5. Governing equatiow Let us now restrict our attention to deflagrations with M . 5. namely. (40) that is. found to be true from the deflagration solution). (41) . in fact.21 at T = 1. Through division of equations (31) and (33) by equation (19).3. it is seen that the two governing equations can be taken to be and over the range 0 I T I 1.3.(T.
cp) = F. (45) So that A may be treated as a constant without complicating the definition we shall (solely for simplicity) employ a final assumption: of o. is an unknown parameter that is to be determined by the structure of the wave. Since equation (39) represents an approximate solution to equation (39.(T. and the activation energy. they can be derived more directly by estimating the magnitudes of the terms in equations (5) and (9). Thus A is considered to be an eigenvalue of the nonlinear equation (42) with the boundary conditions stated above.T)eXp[P'(l  T)/(l/a' k T)]. By utilizing equations (38) and (41) in equation (34). Three constants a'. 11.1. in view of equation (15). These formal results were anticipated in Section 5. When this result is substituted into equation (36). (42) where the dimensionless reactionrate function o = W ( T ) is defined as W (1 . in which F. all of which are presumed to be known. Ta1/m2cp)eB'. 3. The first two of these constants are determined by the initial thermodynamic properties of the system. which. and A. we can obtain the T dependence of F . This latter derivation rests only on the assumptions that the Mach number is small compared with unity throughout the wave and that the coefficient of viscosity is not abnormally large.(T. according to the discussion in Section 5. enter into equation (42). we find that d€/dT = I\W/(T . and the dimensionless factor A has been defined as (44) A = (ApB. we may hope that a solution will exist only for a particular value of the constant A. = Two realistic cases in which the nonessential assumption 11 is valid are T and a1 = f. A is called the burningrate eigenvalue. b'.  8 . thus the results do not depend on assumption 8. explicitly.t). the total heat release. the third constant A depends on the mass burning velocity m.144 Theory of Laminar Flames which implies that p = constant. 3. The twopoint boundary conditions for equation (42) are t = 0 at T = 0 and E = 1 at T = 1./RoT.1. kinetic. the only differential equation that remains to be solved is equation (36). Since equation (42) is a firstorder equation with two boundary conditions. In addition to depending on known thermodynamic. AT"'' a. The temperature dependence of A is given by constant.. and transport properties. (43) the dimensionless activation energy is p' = E. cp(~)) depends only on T . = 0.
Hence. and it is a simple matter to determine the behavior of the solution in the neighborhood of this point. when ~ ( z ) given by equation (43). whence t 4 t near z = 0. The Unimolecular Decomposition Flame with Lewis Number of Unity 145 From equations (6). t = 0.l/Ao(O). and wz(l. introduced in Section 2.  5.)(1/JA)> where A. this implies that the mass fraction of products becomes negative [see equations (21) and (24)]. ( l / f i ) 0. . first arose. (16). then rn would have to be imaginary according to equation (45).1. [ 171. which implies that z becomes negative as t increases [since (dzldt). for A > 0. This can be seen more directly by difis ferentiating the reciprocal of equation (42) with respect to E .. for example. the result is (d2z/dc2). It is physically unacceptable for z to decrease as c increases near z = 0. [IS] The coldboundary condition is z = 0 at E = 0. We thereby see that d 2 z / d t 2 ./C. derived from equations (42) and (19). the reaction proceeds rapidly and heat is conducted downstream.2.2. is the value of A at T = T. z becomes negative (the temperature drops) as t increases near the cold boundary. Equation (42) is then easily integrated to show that = t2/2Aw(0) near z = 0. It is clear that the trouble lies not in the mathematics but instead in the mathematical model of the physical situation . Thus A itself should not depend strongly on the properties of the system. The comparison of this expression with equation (2) shows that the expected primary dependence of vo on the properties of the system has been included in the definition of A. in the case Y2. If o(z) does not go to zero at z = 0 [as is implied by equation (43)]. A further indication of this fact may be obtained by differentiating equation (19) with respect to 5 and using the result dt/d( = ( d t / d z ) ( d z / d t ) = A o .dz/d< =  Aq (46) * If A were negative.3. the mathematical problem defined by equations (42) and (43) with the stated boundary conditions is illposed. 6 = 0. it can depend only on the two dimensionless groups CI'and p. indeed. and hence z may be neglected in comparison to t on the righthand side. equation (42) implies that (dzldt).5. For typical activation energies.)W2(L T.) denotes the reaction rate [appearing in equation (16)] evaluated at X I = 1 and T = T. then the point (z = 0. The problem identified here is the one in which the coldboundary dificulty.0= 0. (44). = .2.3. and (45) it can be seen that 00 = (liPo)J(~. = 0. T. With t = 0 at z = 0 and o(0) # 0. t = 0) is not a singular point of equation (42). The coldboundary difficulty [ 161. = 01 for realistic* values of A.
With either of these methods of remedying the coldboundary difficulty.3). It is found that the calculated estimate for A assumes a pseudostationary value as zi. It cannot do so. At the cold boundary.)(dz/dOi.00 at a finite velocity (that is. or the heat loss to the flame holder. In the formulation of the problem. where the dashed line indicates the pseudostationary value of A. below which the chemical reaction rate vanishes [see equations (19) and (33)]. so that dz/d< and d2Tldt2 should both approach zero. there is no reaction upstream). for an infinite time) without reacting appreciably. we then see that A = 0. The equations attempt to demonstrate this result by suggesting that A = 0. for reasonable functions w(z). but this is impossible according to equation (46) if w does not approach zero as z + 0. nonzero value for A involves the assumption that a flame holder serves as a weak heat sink [lS].'. This may be seen. whence equation (42) may be integrated to show that A = rd 0e/[:(w/z) dz. This conclusion depends directly on the strong temperature dependence of the reactionrate function.146 Theory of Laminar Flames the three terms of which nondimensionally represent diffusion. it is therefore reasonable to neglect t in comparison with z in the denominator of equation (42). of course. 5 + . it will have to have reacted before it arrived. Since the integral in the denominator diverges at z = 0 if cc) > 0 and w(0) # 0. In one approach. It is reasonable to identify the pseudostationary value as the .(T. equation (19) implies that (dz/d5)i = z i . respectively. Avoidance of the appearance of the coldboundary difficulty can be achieved only by revision of the physical model on which the mathematical formulation is based. where e' is the value o f t at z = T' (near zero). by starting an integration from the hot boundary and finding that. convection. the value of A depends. and reaction.7'. the combustible mixture has been asked to flow from x = .co.3. whence the heatsink concept is seen to be mathematically equivalent to the ignitiontemperature concept at the cold boundary. on the chosen value of zi. is allowed to vary between reasonable limits [16]. There are many ways in which this can be done. so that both m and uo are infinite [see equations (45) and (6)] if w(0) # 0. The pseudostationary behavior is illustrated in Figure 5. . The introduction of this artifice is equivalent to employing the physical concept of an ignition temperature Ti. von Karman and Millan [17] replace T = 0 by z = zi (where 0 < zi < 1) as the position at which to apply the coldboundary condition t = 0. even though o # 0. for example. Since t = 0 at the flame holder (that is. An alternative procedure for determining a finite. In the vicinity of the cold boundary. removing an amount of heat per unit area per second equal to (1d T / d x ) i = mc.we expect z to approach zero smoothly. t approaches zero much more rapidly than z at the cold boundary (see the inset in Figure 5.
For this reason alternative modifications to o(z) have been sought. by arbitrarily defining A as the value corresponding to the inflection point of the curve in Figure 5. one which can lead to somewhat laborious computations if solutions are to be generated for a range of values of zi in search of pseudostationarity. and the corresponding is not to be associated in any way with a physical ignition temperature to be found experimentally. Thus a burning rate that depends only weakly on zi is obtained. .The nondimensional ignition temperature zi is small. One alternative that has enjoyed some popularity is that of Friedman and Burke [19] in which l/a’ is set equal to zero in equation (43). ziis a purely computational device. The Unimolecular Decomposition Flame with Lewis Number of Unity 147 FIGURE 5 . Schematic representation of the solution profiles and of the pseudostationary behavior of the burningrate eigenvalue as a function of the nondimensional ignition temperature. However. this dependence may be eliminated entirelyfor example. Instead. physically significant value.5. changes the slope d o / d z at z = 1. 3 . since values of zi very near zero or unity seem unrealistic. it is a modification that extends over the entire range of z and. for example. in an exponential fashion as z + 0 and results in a graph of ~ ( z that ) is not very different from that for reasonable values of l/a’ # 0. These early approaches to the elimination of the coldboundary difficulty are mathematically equivalent to reformulation of the physical model through modification of o(z) by putting o = 0 for < zi. This avoids the coldboundary difficulty by causing w(z) to approach zero rapidly.3.3.
. the maximum value of A appears physically to be the most realistic [26] and is the value that has been obtained [23]. boundary point to be z = z i . In this case. Multiplying equation (42) by (1 . It is because of this nodal character that one integral curve from the saddle passes through the cold boundary for each value of A less than or equal to the maximum. [25]. If w/z approached infinity as z + 0. then this formula gives two directions of integral curves at the cold boundary for A less than a maximum value. the requirement that the integral curve must pass through both the nonsingular point (5 = zi.. the solution to which is 1 f$ I 412 lim (w/z) g)o = 1n 2 If o / z approaches a finite limit as z + 0. we obtain a quadratic equation for (dE/dz). which is not a singularity of equation (42). changes are permissible in o(z) only near z = 0. [22]. In contrast. the behavior of w/z as z + 0 is relevant. [24]. then solutions satisfying the given boundary conditions exist for all values of A less than or equal to a certain maximum value [22]. if lim. with an the solution t = 0 for z < zicauses the relevant coldignition temperature zi. have been studied [23]. This attribute of the ignitiontemperature approach to the correction of the coldboundary difficulty helps to make it more attractive than other approaches that involve modification of o.(w/z) is finite (either positive or zero). [27]) it is found that the singular point (or critical point) at z = t = 1 is a saddle point.148 Theory of Laminar Flames thereby affecting the value of A obtained and providing an incorrect dependence of A on a'. where m and n are positive integers.Z)~Z". If such an analysis is pursued [26] by standard methods (for example. As might be expected from the above discussion. and it does so with < z. An understanding of the special characteristics of the ignitiontemperature method may be obtained from a phaseplane analysis of equation (42). E + 0. and moreover the cold boundary is found to exhibit a nodal type of behavior when A does not exceed the maximum. For nonnegative functions w(z). t = 0. E = 0.t/z) and passing to the limit 7 0.0 I t I1) is the cold boundary. If w(z) is positive except at z = 0 and z = 1. this particular modification is not recommended. For rate functions in this class. then the problem remains illposed. . [Zl].t = 0) and the hotboundary saddle serves to select a unique value of A. only one of the two integral curves from the saddle enters the physically acceptable region z < 0. Much of the work has been concerned with continuous functions w(z) that are positive for 0 < z < 1 and that vanish at z = 0 and at z = 1 [22]. If w vanishes in a finite neighborhood of the point t = 0. t < 1.. for these rate functions. then the only other singularity in the region of interest (0 s z I 1. Modifications to w of the form o = constant (1 . then A is unique [21]. z = 0 . Therefore. A certain amount of rigorous study has been given to the mathematical form of o(z) near z = 0 that renders the problem wellposed [20].
that at the cold boundary (originally g = 0 at z = 0) becomes g = zi at z = zi(O < zi < 1) in the ignitiontemperature formulation. The specific form of w(z) in the vicinity of z = 0 plays no role in generating any of the terms in the expansion. If the problem is rendered wellposed by introduction of zi.then successive orders of the asymptotic expansion yield. In this respect. Bounds on the burningrate eigenvalue When the problem is formulated in a wellposed manner. upper and lower bounds may be established rigorously for A. The analysis employs a form of equation (46) obtained by defining g = dz/d< and by treating z as the independent variable and g as the dependent variable.r = ACO. Unlike the steadyflow probIem in the infinite domain considered here. see Section 5.5. the coldboundary difficulty simply does not appear within the context of the asymptotic expansion.3. This has been accomplished by Johnson and Nachbar [21].6) occupies a unique position with respect to the coldboundary difficulty. The asymptotic method thereby provides a physically correct solution to the illposed problem. or ratefunction modifications become superfluous.3. 5. the idea of a steadily propagating premixed laminar flame is an asymptotic concept. When the asymptotic method is adopted. flame holders.3. (47) The boundary condition at the hot boundary for equation (47) is g = 0 at z = 1. The Unimolecular Decomposition Flame with Lewis Number of Unity 149 Asymptotic analysis for large nondimensional activation energies (p’ appreciably greater than unity. it is permissible but unnecessary to state the problem in a wellposed fashion. Integrating equation (47) from z to z = 1 gives g2/2 = A [ 1 o dz  [g dz. (48) which reduces to (49) . in an asymptotic sense. In a very real sense. The asymptotic approach in effect determines A through an expansion dominated by the behavior of the solution in the hotter portions of the flame. successively improved approximations to the pseudostationary value of A. If the asymptotic approach becomes inappropriate.3. From equation (46) we see that 9  g dg1d. Further clarification of the situation has been achieved recently [28] on the basis of analysis of an initialvalue problem in which the initial rate of reaction is not negligible. then the initial metastable mixture of reactants is insufficiently stable for a flame to propagate through it at an identifiable quasisteady speed. and artifices of ignition temperatures. the initialvalue problem in the infinite domain (or the steadyflow problem in the semiinfinite domain) is wellposed. [29] for the formulation based on the ignition temperature.
* On the other hand. substituting this same upper * The observant reader may have noticed that by setting T = 1 (and therefore g = 0) in equation (50). should we employ the more complicated expression given in equation (51)? The reason is that for most upperbound trial functions g(r). equation (51) yields a lower upper bound for A (a value closer to the correct value of A) than does the simpler equation. Dividing equation (47) by g and integrating from z = zi to z yields 9 = 5  A By substituting equation (50) into the integral on the righthand side of equation (49). whence the resulting value of A given by equation (51) will be an upper bound. (1 .z ) . which also yields an upper bound for A when an upper bound for g ( r ) is used to evaluate the integral. (o/g) d T = 1. Equation (51) will be used shortly to determine an upper bound for A. Since u. for which the value of A given by the simpler equation is twice the value given by equation (51). . and g are all nonnegative.150 Theory of Laminar Flames when evaluated at z = zi. one obtains A l:.” Multiplying equation (52) by exp[ .  d {g2 exp[ 2 dz [( ) : d z ] } = 2Aw exp[ 2 I(): dr]. (50) which becomes when the double integral is reduced to a single integral through integration by parts. Equation (47) may also be written in the form d(g2)/dz = 2(g2)/9 . Equation (54) will be used to determine a lower bound for A.2 (I/g) dz] in fact shows that It. then. Why. it is clear that by substituting an upper bound for g(z) into equation (51) a lower bound for the integral is obtained. it is found that s: (w/g ) dz. Integrating equation (53) from z = zito z = 1 and utilizing the boundary conditions on g yields after dividing by 7’. (53) as can be seen explicitly by expanding the derivative on the lefthand side of equation (53) by parts. (52) which resembles a linear equation for g2 and therefore suggests obtaining a “solution” by means of an “integrating factor.~ A w . This is readily seen with the upper bound g(r) = T .
Iterative procedures and variational methods An iterative technique for obtaining successively narrower upper and lower bounds for the eigenvalue A and successively lower upper bounds for the function g(z) has been developed by Johnson and Nachbar [21]. implying that the exponential and therefore the outer integral assume upper bounds. Equations (51) and (54) can therefore be used to provide limits between which the eigenvalue must lie.4. (56) after the inner integration over T is performed. Since A (wig) d z 2 0. because the inner integral will be a lower bound. 5. Equation (51) then implies that A I and equation (54) shows that [2 [:(WT) dT]'. The Unimolecular Decomposition Flame with Lewis Number of Unity 151 bound for g(r) into equation (54) will yield a lower bound for A. rigorous bounds for the burning rate eigenvalue A. (55) Hence T may be used for the upper bound of g in equations (51) and (54). Integrating equation (50) from T = zi to T yields where the last equality is obtained through integrating the last term by parts. Since they have proven that this procedure converges and that g(z) monotonically approaches the solution of equation (47).A [ w dz ri + [y ri dz. (59) .3. Equations (56) and (57) constitute explicit.5. and the present method constitutes a truly rigorous procedure for obtaining solutions of any desired accuracy.3. and equation (56) or the average of equations (56) and (57) provides a closer estimate for A than that obtained by many approximate methods. These bounds typically are reasonably narrow. it follows from equation (50) that 95 T. the oscillating divergence sometimes observed with other iterative schemes cannot appear here. Integrating equation (47) from z = zi to T gives g2/2 = z?/2 .
3 it is seen that for each grnl(z) a value of A may also be computed from equation (51). Equations (54) and (60) provide the basis for the iterative procedure. as n + m. An upper bound for g(z) is substituted into equation (54) in order to obtain a lower bound for A. Thus successively narrower bounds for the burning rate and successively smaller upper bounds for g = dz/dt are obtained. the righthand side of equation (60) will assume an upper bound. equation (60) could have been replaced by the simpler expression given in equation (50). Throughout this reasoning. convergence is usually slower when equation (50) is used instead of equation (60) (see the preceding footnote). whence equation (60) implies that the ( n + 1)st approximation is smaller than the nth approximation. gml(z) approaches the correct solution from above and Alnlapproaches the correct eigenvalue from below.1)st approximation to prove that the (n + 1)st approximation is a lower upper bound than the nth approximation.3. since the fact that the nth approximation for g is lower than the (n . . and that the resulting sequence of values will approach the correct eigenvalue from above. Since the first estimates for the bounds of A given in equations (56) and (57) are so close together. then. however. since the integral in equation (60) will clearly assume a lower bound. The fact that the resulting new upper bound for g(z) is lower than the previous upper bound requires proof by mathematical induction. If this value of A and the upper bound for g(z) are substituted into the righthand side of equation (60).152 Theory of Laminar Flames which becomes when equation (58) is substituted into the last term. then the second approximation obtained from equation (60) will be a lower upper bound. and hence a new upper bound for g2(z) will be given by equation (60). it appears likely that few iterations would be * It is obvious that if g = 5 is used as the first approximation.1)st approximation immediately shows [via equations (54) and (60)] that the righthand side of equation (60) used in obtaining the ( n + 1)st approximation is smaller than the righthand side of equation (60) used in obtaining the nth approximation. From the discussion in Section 5. This last proof is obtained quite simply from equations (54) and (60). The induction proof is then completed by using the assumption that the nth approximation for g is a lower upper bound than is the (n .* The nth approximation for g is then given by [see equations (60) and (5411 where In accordance with the preceding discussion.
6 will be sufficiently accurate. for example. the integral is convergent. either equations (56) and (57) or one of the other approximate formulas to be given in Sections 5. The approximations to be discussed next bypass the coldboundary difficulty in a natural way. may be employed. based on the introduction of q = ec as a new variable. ds/dt = c. The Unimolecular Decomposition Flame with Lewis Number of Unity 153 required to obtain an exceedingly high degree of accuracy.5. which is easily evaluated in terms of exponential integrals [33] when w is given by equation (43). For the purpose of burningvelocity calculations. so that equation (46) can be written in a standard form. and von Karman Many approximate formulas for A have been developed on the basis of the physical idea that conditions in the vicinity of the hot boundary control the burning rate (for example. and it will be unnecessary to employ the method described here. d 2 t / d q 2 = . The approximations of Zel’dovich. 1 . None of these approaches circumvent the coldboundary difficulty unless o(z) is modified suitably.t).3. Multiplying equation (42) by (1 . However.5. with the nth approximation for g(z) substituted into the integrals to obtain the ( n + 1)st approximations for A and g(z). One way to see how these formulas may be obtained is to follow the approach of von Karman [32]. For example. A variational approach also has been developed [25].z) and integrating from 0 to 1 yields AI where = 9 . von Karman suggests that an approximation to A can be obtained from equation (63) by approximating the integral therein on the basis of the behavior of z ( t ) in the vicinity of the hot boundary.Aw/q2. [S].3. with the constant c obtained in terms of A by evaluating the slope. from .z = c(l . FrankKarnenetskii. one approach [30] employs equations (48) and (49) directly. A linear relationship.t) . 5. by the introduction of zi # 0. Other iterative methods have been proposed that are less well justified.5 and 5. From equations (61) and (62) it is seen that the integrations that must be performed in carrying out the present technique are not excessively complicated. for most applications. [31]). It may be noted that if the present methods are employed. In this approximate evaluation.jol(lz) dt. then calculations must be performed for various values of zi to find conditions of pseudostationarit y.  (63) I = IO1w dz.3.(1 .3. and the coldboundary difficulty thereby disappears.
6. it is more instructive to begin with equation (46). A simpler approximation is to put c = 0. For later convenience in analysis. we obtain.154 Theory of Laminar Flames the limiting form taken by equation (42) at the hot boundary.3. typically is less than 10% when the first formula is used and less than 300. The factor (1 + ~ / p ’ ) in resulting expression for I may be expanded in powers z/p’ to obtain an asymptotic expression for I useful for large values of 8’. obtaining A = (21)’ directly from equation (63). we find that The activation energy and the heat of reaction are usually sufficiently large that a negligible error in this integral is produced by extending the upper ~ the integrand of the limit to infinity (l/fi’a’ = 0).z)/(l/a’ + z) in equation (64). which is proportional to m or u. Keeping only the first term in the expansion.. 1% ( +B’ )’ 1 lja’ ~ J : z e K 2 dz + 0($) JZ(1 = (y)2 + O($). This may be shown to give A = (1 . the combinations a = a’/( 1 + a’) (not to be confused with the a of Chapter 2) . These formulas of von Karman [32] become increasingly accurate as p‘ increases. [34] may be obtained from the simplest result of von Karman by performing an asymptotic expansion of the integral’I for large values of fi‘ and retaining only the first term in the expansion. then the error in u. Asymptotic analysis for strongly temperaturedependent rates Although the asymptotic analysis was developed first on the basis of a formulation in the variables appearing in equation (42) [9]. by transforming from the variable z to z = p’(1 .@)/(21). This formula is nearly as accurate as the von Karman expression from which it was derived and is exactly the same as the first approximation obtained by the asymptotic analysis to be presented next. The formula of Zel’dovich and FrankKamenetskii [7]. with 0 meaning terms of order. The significance of this result and of the appearance of the nonnegligible factor 1/p’ on the righthand side will become clear after the activationenergy asymptotics have been developed.. In view of equation (43).. for the second. if p’ > 2. The introduction of this result into the formula l/A= f l yields 5 I/& = + l/a’) p’ which is the expression of Zel’dovich and FrankKamentskii for the nondimensional quantity l/&.. 5.
5. Because of the many important contributions of Zel’dovich to our physical understanding of the theory. is the appropriate large parameter of expansion.(p). (0 < a < l). . may be introduced. . and large activation energy and large heat release both favor large p.rather than p’. Therefore. (67) . thus /? generally is large if p’ is large. produces the sequence of outer equations d 2 z n / d t 2= dzn/d<. passing to the limit /3 co.)/T. . Thus fl < p‘. qualitatively a measure of the strength of the dependence of the reaction rate on the extent of reaction completed. there is a zone in which the reaction rate is negligible and in which the convective and diffusiveterms in equation (46) must be in balance. Here f + + + lim CHn+ l(P)/ffn(P)I B.T’. From equation (66) it is seen that as p becomes large the reaction term in equation (46) becomes very small (“exponentially small.and so on. For the sake of being definite. In describing this convectivediffusive zone.m = 0.z of order unity. again passing to the limit /3 + 00. with HOW) = 1.2. and p is the product of a with the nondimensional activation energy E1/RoT. Hence for 1 . however.5.8 and seldom are less than 0. f Substitution of this expansion into equation (46). In terms of a and the rate expression in equation (43) becomes LC) 0. subtracting the result from equation (46). . 1.3. n = 0. (T. It will be found that j. (66) Note that a is the temperature rise divided by the final temperature. The Unimolecular Decomposition Flame with Lewis Number of Unity 155 and /? = alp’/( 1 + a‘)are employed in place of a ‘ and p‘. The attribute that is essential to the asymptotic approach is that the nonnegative singlepeaked rate function possesses a maximum that narrows and increases in magnitude and that occurs at a location progressively closer to the fully burnt condition as p co. an outer expansion of the form z = zo(5) H1(p)z1(5) H 2 ( p ) z 2 ( 5 ) .. dividing by H. the heat release cannot be too small if the asymptotic results are to be accurate. numbers that are sufficiently large for asymptotic results to be reasonably accurate. = (1 . The Zel’dovich number is a nondimensional measure of the temperature sensitivity of the overall reaction rate.. that other rate expressions can be treated in a similar way. It may be emphasized. Representative values of p in fact lie between 5 and 15. Values of a typically are 0. it seems appropriate to call p the Zel’dovich number. following the formalism of matched asymptotic expansions [35].” since p appears inside the exponential) unless z is near unity. . in addition to p’ being large. the analysis will be developed on the basis of the rate expression given specifically in equation (66) by considering the parametric limit in which /? + 00 with a of order unity and held fixed and with A required to vary with fl in such a way that a nontrivial result is obtained. .z)exp{p(l  z)/[1  a(l  z)]}.
. limB+. Any stretching of this type causes the convective term in equation (46) to become small compared with the diffusive term.156 Theory of Laminar Flames so long as limp. The solutions to equation (67) satisfying the boundary condition that z = 0 at ( = co are   T. This observation indicates the need for a downstream zone in which 1 . Study also shows that if dq/d( is of order unity in the limit.(p) p". so is introduced as well.[h. 2. In equation (69) an inner expansion of the form YO(V> + hl(P)Yl(V) + hz(B)Yz(rl) + . the necessary reactivediffusive balance with y and q of order unity will be maintained only if s2A is finite and nonzero in the limit. for H.. Equation (68) must fail if ( becomes sufficiently large.(dq/d() = co . then any selection having limp. (69) where use has been made of equation (66).(p) CUB. n = 0.z is or order p'. With the stretching q = (5 . Therefore. Thus the downstream zone cannot be thicker than the upstream zone..+ l(fi)/hn(P)] = 0. If this dependent variable is denoted by q.+l/sn) 1. the reactionrate term is no longer exponentially small. . in fact. for H. are constants to be determined by matching conditions. with h&?) = 1. From equation (69) it is seen that in the limit p + co (s + 0).T). Following a . Therefore. = A. where limB. then matching again cannot be achieved. An appropriate stretched dependent variable for this zone apparently is y = p(l . from equation (66) it is seen that when 1 . (68) where the A . This last restriction is satisfied by all algebraic (powertype) ordering functionsfor example.(o)/s(p). 1. m[Aeup/H. where lim~~m(s. equation (46) written in inner variables is d2YldV2. .. The inner zone therefore is reasonably termed a reactivediffusive zone.(dq/d() = 0 produces in the lowest order a firstorder equation that expresses a convectivereactive balance and that lacks the flexibility needed to satisfy the downstream boundary condition and simultaneously match with equation (68). where s + 0 as p + so.I = = 0.z becomes of order b'. . the expansion Y = A = (llS2>CAo+ Sl(PIA1 + s2(P)A2 + . . the reactionrate term can contribute to equation (67) only at exponentially small ordersfor example...s dYl4 = S2AY expCyl(1  aY/P>I. To ascertain formally a suitable independent variable for use with y necessitates investigating various possibilities.e5.the downstream zone is thinner than the upstream zone and may reasonably be viewed as an inner zone. and since at least two terms are needed to satisfy the matching and boundary conditions successfully. may be introduced. there must be a reactivediffusive balance in the inner zone to the lowest order.(fi)] = 0 for any fixed a > 0 of order unity. .
stands for terms that go to zero exponentially in v] as v] + . A N Ao/s2 = p2/2.(1 + y. Thus we first find that Aoe50= 1 and next find that s must be of order p.'. [36]).s.t. (74) Equation (74) is used repeatedly at successively higher orders to evaluate constants and ordering functions.[l . + . (72) where e. Unlike equation (70). a twoterm expansion for A is available. + (70) + The downstream boundary conditions.(1 yo)eyo]'i2} d y . with t ( p ) + 0 and t ( P ) / s ( P ) + m in the limit. = 3c( . The general matching condition is then written as (73) where 7 implies that when v]. (n = 1 . (71) It is found.(. . y. + .. = P". in agreement with equation (65).. it may be shown that H . By proceeding further with the expansion and matching./dv])' = A.s. . = 1.)/t(P). . /:yfeCy' dy' = Ao[l . held fixed. the first of which is d2yO/dy2= A. and a composite expansion for T ( < ) may be constructed by standard methods [35]. .3440.. the differential equations for y. the most infallible approach currently available is to investigate a parametric limit in an intermediate variable [35]. z . we then generate a sequence of inner equations. 2 . Although there are many ways to effect matching.3. where y = f : (1 . From equations (68) and (72). A further integration then gives An expansion of this formula for large negative values of v] produces yo = &v] + constant + e. that to achieve matching the negative square root must be taken in equation (71) in obtaining dy. . s. where v]. ..p + h. e c o + . + . equation (73) may be written as 7. .)eCYo]./s)y. 0 as v] co. the two asymptotic expansions must agree.5. . 7 1 + p'&(t. .t. A0e5"(1 + tqt + . These last results provide the the further matching requirement first approximation to the burningrate eigenvalue.thereby obtaining = 1.. + . = s(P)v]/t(P) = (5 .may be applied to the inner equations. With these results. as may be anticipated from the fact that z decreases (y increases) as v] decreases.~ O .). Y . A . The Unimolecular Decomposition Flame with Lewis Number of Unity 157 procedure analogous to that described for the outer expansion.) + H . Thus we consider P + co with v ] . ~ .hence we may put s = p. Completion of a solution by matched asymptotic expansions entails employing matching conditions.) are linear.y./dv]. and A.y. + H . l(y. h. 7 1 . = p". and methods for solving them are available (for example. if desired...00. = P". The first integral of equation (70) then is +(dy. is introduced on both sides.
= i. at least in principle. Section 3. The reaction occurs in the thin downstream zone because the temperature is too low for it to occur appreciably upstream. + . followed by a thinner reaction zone in which convection is negligible to the lowest order in P'.4.(a) then need not be introduced at all. For example. . for example. abbreviated presentations are made. Analyses at the depth indicated here underlie applications of activationenergy asymptotics. where t ( ~ )is sought.' y . as well as elsewhere in this volume (for example. Because of the translational invariance of the original problem.4. matching of the slopes of the inner and outer solutions [37]may then be employed directly in equation (71) for q + . z = 1 for 5 > 0. the expansions A = P2Ao + PA. often physical understanding allows shortcuts to be taken. . There are two zones through which the temperature varies: a thicker upstream zone in which the reaction rate is negligible to all algebraic orders in p'. the entire analysis can be completed in a few equations [38]. A simpler matching criterionnamely.which specifies the first approximation to the location of the inner zone. although much thought underlies the equations. the expansion in terms of H. However. The flame structure that has been calculated here is illustrated schematically in Figure 5. Suitably imposing this stipulation at all algebraic orders allows the outer solution to be written in advance as z = e5 for 5 < 0. simplification can be achieved by employing this fact to demand at the outset that the reaction occur at r = 0. . the convectivediffusive zone need not be mentioned Slopes match I 1 FIGURE 5. where the reaction is occurring. giving the first approximation for the burning rate. .rn ( y o + co) to obtain A. the reactant concentration has been depleted by diffusion so that it is of order B' times the initial reactant concentration.4. and y = yo + P .158 Theory of Laminar Flames The extended discussion of activationenergy asymptotics that has been given here was included to illustrate details of the technique that sometimes are needed in analyzing difficult problems. cannot be calculated. the value of to. matching conditions such as equation (73) seldom are written. . In the formulation based on equation (42). In most publications. + A2 + . could have been guessed in advance and later verified to work. Schematic illustration of deflagration structure obtained by activationenergy asymptotics.3). With these simplifications.
This result has come formally from the matching condition and represents the only burning velocity for which an internally consistent asymptotic structure for the deflagration can be obtained. and El. for simplicity let us ignore the correction terms in equation (75)of order 8. However. In dimensional form [see equation (45)]. notably the strongest dependences on L/cP. see [9]).4 shows the physics that have been overlooked in deriving equation (2). for many purposes. At representative values of B. In making this comparison. the latter two controlled by the Arrhenius exponential. Figure 5. The twoterm result in equation (75) lies within a few percent of the pseudostationary burning velocity calculated numerically over the entire range of values of p that could be of practical interest (fi 2 2.3.p.3).. where it can produce corrections to uo approaching 50% in extreme cases. . The value of vo obtained from the twoterm expansion typically is 10% below that given by the first term. no change occurs in ~ ( z )until z is within B. as a delta function in 5 that describes appropriate jump conditions across a reaction sheet [39]. the burning velocity given by the twoterm expansion A = +/I2 + P(3a . In the reaction zone. and /J = E.of unity (see Figure 5. T.y). is* where o! = (T. It is at least as accurate as the best results obtained by ad hoc methods prior to its derivation. This deltafunction idea underlies the early development of equation (65) [7]. the twoterm asymptotic expansion for uo is significantly better than the corresponding oneterm expansion of Zel'dovich and FrankKamenetskii obtained from equation (65) and given by the first term in equation (75). . Retention of the second term in equation (75) is especially important at the lower values of p.(T. the reactant concentration has * Here 0 means terms of order. if accurate results are wanted.To)/RoT:. The principal functional dependences of uo agree in equations (2) and (79. [34]. . It is of interest to compare equation (75) with equations (2) and (3). if w in equation (2) is evaluated at the initial concentration and at the final temperature..or higher. The results could be made to agree by evaluating w in equation (2) at a suitable lower temperature. Thus Zel'dovich numbers for real flames lie in a range that is both appropriate for application of asymptotic methods and sufficientlyfar from the limit to require a formalism capable of generating the second term in the expansion.B.To)/T. The reaction zone is sufficiently narrow in the coordinate to allow the reactionrate function to be approximated. but no physical basis exists for selecting this temperature. then the numerical value of uo obtained from equation (2) is roughly p times that of equation (75).5. The Unimolecular Decomposition Flame with Lewis Number of Unity 159 since the solution therein is t = 0. the difference is an order of magnitude.
as given by equation (4). the reaction proceeds not in a zone of thickness 6. appearing through p in equation (79.times the initial concentration (and at T. in equation (2) w should be evaluated at p. It would not be difficult to modify the physical reasoning given in Section 5.5 in the physical (x) coordinate. the total pressure drop seldom exceeds 1 mm of mercury. There are evident secondary dependences of uo on the overall activation energy and on the heat release. because of the strong temperature dependence of w.2 on the basis of considerations derived from a structure like that shown in Figure 5.5. profile is easily computed from equation (5) by using the velocity profiles obtained from the equations in this section. The actual pressure clear from equation (9. that p must decrease. however. The reader may be interested in doing this to test his or her understanding.7. ' 5.4 to obtain correct fi factors in equation (2). the reaction rate). . but rather in a thinner zone of thickness 6/p.3. therefore. Furthermore. of course. it is for example.' . . Mach numbers are sufficiently small for this fundamentally perturbative approach for obtaining p(x) to be well justified. All properties vary monotonically through the wave (except. that cannot be obtained from equation (2) without introducing these modifications. Representative variation of flow properties in a deflagration wave. since v increases through the wave. since the reaction does indeed occur at the downstream side of the deflagration).1. Increasing the activation energy makes the reactionrate curve more sharply peaked toward the hot boundary but has relatively little effect on the shapes t FIGURE 5. The small pressure drop shown is not predicted by the equations in this section (since p = constant here).160 Theory of Laminar Flames been depleted through diffusion by the factor p . Generalizations to other flames Some additional aspects of the deflagration structure are illustrated qualitatively in Figure 5. These two factors of p' produce the p' in equation (75).
the expansion of the convectivediffusive solution about 4 = 0 is Y. the structure illustrated in Figure 5. However.2. . assumption 10 is violated. u.5 will be modified can be inferred from reasoning like that in Section 5. where the conditions z = Y = 1 have been imposed at the reaction zone (5 = 0). Since w is proportional to the fuel concentration. If A/cp is increased with pDlz held constant. Since thermal effects alone were considered in deriving equation (l). may be expected physically to spread the reaction region of elevated temperature over a larger range of reactant concentration. respectively. equation (1) yields 6 &. and the analysis in this section is invalid... increasing A/cp should spread the T curve (and. w is proportional to the reactant concentration (in the reaction zone) to the nth . the fuel concentration is proportional to Le. which indicates that in the reaction zone.3. the p.. A symmetric behavior could be anticipated if the reaction extended throughout the entire thickness of the deflagration. An increase in A/cp at fixed pD. onestep chemistry... and solutions to equations (19) and (22) are z = eT and Y = eLeC. thereby effectively increasing the average reactant concentration in the reaction zone and hence increasing w. and p curves) over a somewhat larger x range.)/=. spreads all curves over a larger range of x. the spatial extent of the composition curves is narrowed. The asymmetry in the effects of variations of A/c. the way in which the curves in Figure 5. for large p arises through the influence on the reactant concentration in the reaction zone.Le 5. = Yl. while causing less change in the composition curves. and pD. A quantitative estimate of the magnitude of the effects of relative diffusion of reactant and heat can be obtained on the basis of the structure of the convectivediffusive zone of Figure 5. x Y. if the Zel'dovich number p were not largethen no influence of A/cp on w would be anticipated. The type of reasoning that has been given here can be extended to take into account reaction orders different from unijy and multiplereactant.4. independent of A/cp. Thus an increase in A/cp at constant pD. keeping Le = 1.4indicates that for reasonable values of p. increases the burning velocity proportionally but leaves the extent of the T profile in physical space unchanged. there may be a change in the average reactant concentration in the reaction zone. Increasing pD with A/c. held constant decreases m (physically by diffusively reducing the reactant concentration in the reaction zone. then Le # 1. a) .. However.. The Unimolecular Decomposition Flame with Lewis Number of Unity 161 of the other curves. consequently. and equation (2) implies that m (A/c.o(l .. Since the fuel mass fraction is Y. In this zone 6 = 0. thereby decreasing w) and consequently increases the spatial extent of the T curve (in proportion to and of the concentration curves [in proportion to (pD12)3/2J.Y). proportional to cp/A. .5.1. provided that there is no appreciable modification in w. For a onereactant proces of reaction order n.. Increasing A/cp and pD. If there were a singlezone structure to the flamethat is. because of the increase in m at constant pD.
CST. with the rate expression given by equation (18). Williams. Jm./(. At higher orders. has been evaluated by numerical integration for various values of n [36] and is shown in Figure 5.l/cp)Len. reaction. in which a balance between convection.~ [ 4+n i)(n + 2)/2 . A.equation (l). which depends on n. See B. assumption 9 and many of its consequences no longer are valid for n # 1 [for example.times the initial concentration). approximations corresponding to assumption 11 are unnecessary because the reaction rate is relevant only near the burntgas state. the extent of the concentration profiles is seen to be proporMore precisely. equations (16) and (34) are revised]. l ( Y l . without performing the asymptotic analysis. although M = 1 and N = 2. the extent of the temperature profile from equation (2) the burning velocity is is proportional to and from the Y solution in the convectiveproportional to diffusive zone. onestep chemistry. in the reaction zone is proportional to Le p.3440 for n = 1. it is sufficient to evaluate the quantities in the definition of Afor example. and y = 2. the reaction goes to completion (y = 0) at a finite value of y~ in the reaction zone.970 for n = 3.In 2) for n = 0. representative values are y = 2(1 . [13]. y = 0. to appear (1985). This modification is needed because. the reactant concentration tional to ( p D cp//z). Rogg and F. expansions of these quantities about their hotboundary values appear. For n < 1. ./m. so that w is proportional to Le”fi”.~ (76) ) where r denotes the gamma function [T(n + 1) = j. reviews also may be found elsewhere [lo].l. There have been a number of analyses employing activationenergy asymptotics for tworeactant. + 1)1>{1+ 2 ~ . Occurrence of these finer details of the reactionzone structure is comprehensible qualitatively on the basis of the relative rapidity with which the reaction rate decreases with decreasing reactant concentration at very low reactant concentrationsthis rapidity is low for small n and high for large n. and therefore from. for n > 1. in equation (45)at the hot boundary. equation (76) provides a flamespeed formula in which every factor except the constant 2r(n + 1) could have been obtained by physical reasoning based on Figure 5. Most of these have been summarized by Mitani [40]. * In equation (76) A is defined as in equation (45) but with the additional factors v ~ .1274 for n = 2. and diffusion develops at a very low concentration level (of order p(“+ ‘ ) i ( n .4. yneKydy = n!] and A = {p+l/[2r(n Here y.162 a Theory of Laminar Flames power. where 5 is of order unity.71 + ~ ( f i . To the lowest order. onestep process of order n is [36]* . In the asymptotic analysis to lowest order.y = 1.6. o p / W l ) ’ on ~ l the righthand side. as is evident from the functional form y”. there is a thick downstream region. and powers of p appear that depend on the reaction order. For Le = 1 a twoterm expansion of the burningrate eigenvalue for a onereactant.
If i = 1. . Equation (78) holds for4 > lthat is. . the formulations have been based not on equation (8) but instead on diffusion equations of the type dY. n 2 .m 1 P Le. in which a = p(4 . Equation (77) is derived formally by assuming that an inert species i = N is present in great excess so that the other species effectively diffuse through this inert."' Le. Apparently results based on equation (77) can be surprisingly good [41] if the inert is taken to be an average product. i = 1.1.3. The Unimolecular Decomposition Flame with Lewis Number of Unity 163 0 I 2 3 4 5 J PI FIGURE 5. and n2 identify the reaction orders with respect to these reactants. WlY2. are the corresponding Lewis numbers (Lei = A/pD. (77) where D i is a diffusion coefficient for species i in the mixture. . v.6. Jv.. 2 are the two reactants and n .l(v. a) = ~omy"'(y + aY2eFYdy.cp). if species 2 is .. while Le../dx = ( m / p D i ) ( q .o being the equivalence ratio.. = 1.5.0/v2W2Yl. + 1 (78) where G(nl. and Le. whose concentration profile is calculated a posteriori from I. then to the lowest order in p' the asymptotic analysis yields [40] ' 0 = [ 22. oo 72Y 7: 0'n2 2 Wnl+nZl n l + n z + l PoCp.)"'& "2B1. with 4 = v. . on the reaction order n..ti). equation (76). Dependence of the term y in the formula for the burningrate eigenvalue. .1)/Le2. N ."' x J e E1~2RaTm. With notable exceptions [41].
for example.. Assumption 7 could be circumvented in an asymptotic analysis by use of the idea + + . Consider. usually smoothly because of equilibrium dissociation of products. and equation (78) reduces to + + + (79) The principal features of these flamespeed results are readily comprehensible physically from the previous discussion. n2 1). the overall rate parameterssuch as El. if it is found that 4 < 1.. and lean mixtures (see. It must. 6. the powers of fi and of the Lewis numbers in equation (78).1 to be large compared with p'. be recognized that many of the quantities appearing in these equations vary with 4.a) = an2r(nl l). Equations (78) and (79) have the merit of exhibiting explicitly the influences of transport coefficients on flame speeds and of providing explanations for some initially counterintuitive results (for example. since El/RoTm is large.. rich.1 is small compared with p'). n 2 .2. we have G(n. Assumptions 81 1 have been largely eliminated. If conditions are sufficiently far from stoichiometry for 4 . ormore preciselyif 4 . The remaining powers for p (nl and n 2 ) and the Lewis numbers come from the rate expression w and reflect the fact that in the reaction zone the concentration of reactant i is of order Lei/p times its initial concentration because of the concentration reduction through the convectivediffusive zone. For offstoichiometric conditions. It is also quite significant that in onestep approximations. and therefore the ratereduction factor (Le2/PY2does not appear in equation (79). and n2generally differ in stoichiometric. a) = T(n. species 2 is not depleted appreciably. then G(n. and Tm peaks near 4 = 1. in particular. [42]). Formulas like equations (78) and (79) are useful for calculating the flame speed as a function of the equivalence ratio. The extra 1 in n1 n2 1 for p comes from the thinness of the reaction zone in comparison with the flame thickness. that the high diffusion coefficient of H2 tends to reduce the burning velocity of a fuellean H202 flame). such as influences of variable Lewis numbers. but these formulas cannot be employed in the absence of appropriate chemicalkinetic information for overall rate parameters. For onestep reactions the analyses that have been given here for flame structures and flame speeds have successfully removed many of the most restrictive assumptions that were introduced beginning in Section 5.164 Theory of Laminar Flames present in excess of the stoichiometric requirement or if the mixture is stoichiometric. n 2 . Usually the strongest variation occurs in the Arrhenius factor. then equation (78) may be employed merely by interchanging the definitions of species 1 and 2 throughout. however. n.2. For stoichiometric conditions (4 = 1. with some minor questions remaining.
and the average molecular weight W = l / x l q / K . The 2N dependent variables ciand X i appear in the 2N expressions given in equations (80) and (81) (with pressure constant. adiabatic deflagrations. planar.which is given by equation (40). and finally dividing by equation (13).1.4.1. then using equation (19) to eliminate W/p.2 and the isobaric approximation p = constant are retained. Equation (80) is the ratio of equation (7) to equation (13).5.4.[namely. If an ignition temperature is defined in such a way that wi = 0 for . . If assumptions 17 of Section 5. the critical remaining approximation is that of a onestep reaction. (80) and (81) where f 1 " j= 1 hy(cj .5).4. 1. The physical effects that have been considered here refer specifically to steady. N. and departures from assumptions 3 and 4 could be investigated readily. a ) / .2 and 9. compare equation (112)]. Formulation Let us now consider flames in which more than one chemical reaction occurs. then the flame equations may be taken as dtJd7 = (A/2/m2cP)(wJf). FLAMES WITH MULTIPLESTEP CHEMISTRY 5. it is unclear that much further work directed toward removal of other underlying assumptions would be warranted prior to study of the importance of the onestep approximation. I: = X i W $ T . For steady. Flames with MultipleStep Chemistry 165 that the reaction occurs mainly near the hot boundary. planar. additional physical phenomena occur if these assumptions are relaxed (see Sections 8. ]  (1  (82) is a linear function of z and tj. and equation (81) can be derived from equation (8) by first expressing in terms of X i .1. . K. . adiabatic flames (assumptions 1 and 2). i = 1. Investigations of multiplestep kinetics are addressed in the following section.and so on are known functions of z and Xi). the quantities w i . The independent variable employed here is the normalized dimensionless temperature z.. Discussion of the background and the nature of the problem 5. 5..t j .4.
2. with complicated chemistry the integration of equations (80) and (8 1) faces computational difficulties.4. namely. [45]. Thus. However. These modifications should not introduce appreciable additional complexity if solutions are to be sought by numerical integration. .2.hj. adiabatic.merely at This assumption can be removed.1.166 Theory of Laminar Flames T below the ignition temperature and wi is given by equation (18) for T above the ignition temperature.i= c p = constant. with complicated chemistry. is a convenient independent variable only if it varies monotonically with distance through the flame. For example. reliable and general approaches to the accurate calculation of flame structures and flame speeds for arbitrary reaction kinetics do not yet appear to be available. for idealgas mixtures with Dufour effects neglected. while = ti at T = 1 and chemical equilibrium exists (wi = 0) at T = 13 is expected to determine a (unique) solution to equations (80) and (81) only for a particular value of the mass burning rate m. It is also straightforward to include Soret diffusion by adding appropriate terms to equation (8 1). m)/cp(Tm . associated not only with “stiffness” (Sections B. The normalized temperature z.5. but these are difficult to begin because. tendencies for stepwise integrations in the direction of increasing T to diverge from the hotboundary point motivate integrations starting at the hot boundary. with T = IT . notwithstanding the significant advances made in recent years (see Section 5. given by equation (40). The main limitation to the formulation given is assumption 7 to the effect that cp.To). then a reasonable set of boundary conditions [for example. where each h j is a known function of T and cp is arbitrary since it cancels from the equations. Although it is conceivable that nonmonotonic behavior of T could occurfor example. thereby requiring a search in a multidimensional space to satisfy suitable conditions at the cold boundary [43]. This “multipleeigenvalue” aspect of the problem has led to questionable suggestions to the effect that flammability limits may be associated with the nonexistence of solutions to equations (80) and (8 l). f= ( h j t j . the use of T seems appropriate from the viewpoint of monotonicity. formal expansions about the hot boundary demonstrate the necessity of selecting values of more than one parameter for starting a numerical integration.To). [46] that seem better suited t o some flames than to others. = ei at z = 0 and all ti are specified at T = 0. and numerical integrations of the ordinary differential equations have been completed successfully only by the introduction of special devices 11441. .2 and 4. subject only to the steady. the formulation then becomes quite general. and isobaric approximations.To)/(Tm the expense of complicating the definition off in equation (82).co ej. for flames in which endothermic reactions become dominant at the highest temperaturesno practical examples of adiabatic flames with temperature peaks or valleys are known.3). planar.1) but also with complications from molecular transport and from eigenvalue aspects.
studies with model kinetics and studies with real kinetics. the objectives are to develop physical insight. 5.1. Burning velocities often are influenced primarily by conditions near the hot boundary. isobaric deflagrations with multiplestep reaction kinetics.5. Studies in the former category employ relatively simple rate expressions designed to retain some essentials of the full chemistry in various flames but not based on accepted rate parameters for any specific flame.1. the chemistry of flames generally involves chain reactions. planar. In the order of increasing difficulty. Flames with MultipleStep Chemistry 167 5.2. not to obtain predictions for real flames that can be compared with experimental results. and to predict the distributions of species present in lower concentrations (principally chain carriers). As indicated in Appendix B. as has been seen in the previous section. adiabatic. they depend on overall levels of chaincarrier concentrations but seldom are affected appreciably by distributions of radicals or atoms in the cooler portions of the flame. Three specific objectives of studies with real kinetics may be identified. Literature Investigations with model kinetics date from the early work of Zel'dovich and Semenov [47]. and concentration profiles of major species often can be obtained with. Studies have shown that burning velocities.4. temperature profiles. are associated with chaincarrier distributions. to predict the distributions of temperature and of major species (principally reactants and products). These predictions are of interest not only because of their relationships to fundamental transport and chemicalkinetic properties of chain carriers but also because observable properties of flames. such as emission of chemiluminescent radiation.reasonable accuracy even if predictions of chaincarrier distributions are poor. A number of different model mechanisms have been studied on the basis of analytical approximations designed to .3. their objectives are to improve knowledge of chemical processes in specific flames and to achieve agreement with experimental results.4. Studies in the latter category endeavor to employ correct kinetic data for particular flames to the extent that these data are known. while those for studies with real kinetics usually are numerical. they are to predict the burning velocity. The methods employed in studies with model kinetics often are analytical in character. the burning velocity per se seldom is a sharp discriminant of proper kinetic details. Concentration profiles of major species usually are controlled mainly by the burning velocity and transport coefficients.4. In particular. Objectives of Analyses There are two broad classes of theoretical analyses of steady. namely. Thus the most challenging problem in laminarflame theory is to predict the distributions of chain carriers in flames. thereby possibly obtaining better values for poorly known rate parameters of certain elementary steps.
if 6 is sufficiently small. P. then there usually is a single flame with two reaction zones separated by a transport zone. . for the first and second steps ( k = 1. a1 + a2 must be positive. A . however. y. C. Additional mechanisms that have received consideration are the straightchain model [lo] A + B + 2C + B. Most of the work has treated both p’s as being large and both a’s as being positive. which models a branchedchain reaction with carrier C [lo]. A brief consideration of the mechanism A + B + C may serve to illustrate the range of possibilities that occur. these quantities need not be defined precisely here). there is a thick convectivereactive zone for B + C. in which A functions as a reactant. If p2 = 0 with positive. then two separate flames may occur. B as an inert. Mechanisms that have been studied include the sequential reaction A B C . A C 2C. Related tworeactant mechanisms also have been investigated [60]. and generally both steps occur in the same thin reactivediffusive 0 and P2 is large. More recently.2). [48][51]). aC. to search for the full range of possible behaviors. in this case. For there to be a positive net heat release. traveling faster than the second. . If 6 is larger. but these two reaction zones merge at sufficiently large values of 6. If a1 I processes with highly reactive reaction intermediaries). and 6 are stoichiometric coefficients. the techniques of asymptotic analysis have been brought to bear on simple mechanisms. then no matter how small 6 is. as 6 becomes sufficiently large the step B + C begins to occur in the reactivediffusivezone of the flame and to influence the flame speed (it may be anticipated that even if bz > 0. in addition to this structure there appears zone. and let 6 be a measure of the ratio of the rate of the second step to that of the first (since the discussion is purely qualitative. the process A + B cannot proceed without energy input from B C. [%I. [57]. [62]. embedded within the convectivediffusive + + + + + + + + + + * Numerical methods rather than asymptotic analyses were employed in the first of these references. [34]. the first. A + C D + C. and B1 is large in all problems of interest. there may be a range of values of P2 and 6 for which B + C occurs in a downstream region having a 0 and Pz = 0 (a case that may model convectivereactive balance). and to calculate flame speeds and concentration and temperature profiles. and also related processes with varying extents of chain branching. downstream from the A B flame. B C. A + B. C + yB + 6D has been considered for sequential process A + PB conditions under which species B is present in excess [59]. then at sufficiently small values of 6. here a. [6 1). [52][56]* and the competing reaction A + B. When a1 I to be the possibility (for sufficiently small values of 6) of a thin. reactivediffusivezone for A + B. The tworeactant. the mechanism A + B + 2C (at a negligible rate). and A C + D C (or 2C + D). respectively.168 Theory of Laminar Flames provide estimates of flame speeds (for example. Let ak and Pk represent nondimensional heat releases and activation energies. C as a chain carrier and D as a product. endothermic.
. when they can be obtained . Thus it is seen that even for the simple example of A + B + C. [88]). and different percentages of reactant converted to B and C [%I . with different flame temperatures. Experience has shown that by marching forward in time. Flames with MultipleStep Chemistry 169 zone that precedes the exothermic reactivediffusive zone of B + C. In principle. there is even a range of parameters in which two stable steadystate solutions for the flame structure exist. then the solutions will reveal the instability. [43][46]. numerical integration of the ordinary differential equations also possesses advantages.5. the reverse is true for the laminarflame problem with complex chemical kinetics because of the eigenvalue character of the problem. the integration of ordinary differential equations that describe the steady flame and the integration of onedimensional. The timedependent approach tends to establish the steadily propagating flame in the same way that nature does. the last two references are general reviews that mention flamestructure studies by numerical methods. Standard routines for the integration of systems of parabolic equations may be employed. namely. Although solving timedependent partial differential equations numerically usually is thought to be more complicated than solving ordinary differential equations. With the availability of large computers. it should be emphasized that the extensive citation of literature given here is not complete and omits. the timedependent approach offers many advantages. in particular. the ignition procedure may be presumed to select one of the multiplicity of solutions. A + C. onedimensional * Concerning the work employing numerical methods. The approaches to numerical integration fall into two broad categories. the solutions usually converge rapidly to a description of the steadystate flame. However. partial differential equations that reduce to the equations for the steady flame when the transient terms vanish. different flame speeds. the earliest studies [63][66] and a small fraction of the more recent work [43][46] involve the former. timedependent. For the competing reaction. The majority of the recent investigations [67][87] employ the latter technique. A + B. Foremost among these are the inherently greater accuracies of the solutions. if the flame is unstable to onedimensional disturbances. with known boundary conditions upstream and downstream from the flame and with guessed initial profiles of temperature and concentrations. there are examples in which this has occurred (for example. Investigations with real kinetics have employed expansions about the hot boundary for the purpose of calculating burning velocities [33] and various numerical methods of integration to obtain burning velocities and profiles of temperature and of species concentrations [30]. many relevant papers coauthored by DixonLewis.4. [63][87]*. selected to approximate those of the steadily propagating flame. many different flame structures may be encountered.
the burning rate. the advantage would be thatirrespective of the complexity of the kineticsonly one parameter. this can be a significant benefit if different reactions are of importance in different narrow zones of the flame. In view of these difficulties. [63][66] employed z as the independent variable. there are onedimensional methods involving collocation that are better classified as twodimensional methods. .1.* In principle. there may be merit in trying to integrate first from the cold boundary with an ignitiontemperature approximation. A successful technique of this type would enable a pseudostationary value of the burning velocity to be identified and the corresponding temperature and concentration profiles to be calculated with good accuracy. these integrations generally exhibit saddlelike divergences as z is decreased as a consequence of the varying kinetics [44]. the reader is referred to the references. [44]. the burning * In this respect. the integration was started first from the hot boundary. However. Since these are too numerous to be discussed here. and criteria similar to those previously developed for identifying divergent solutions might be applicable for locating the correct integral curve. using a modern routine for “stiff’ equations.170 Theory of Laminar Flames programs achieve finer intervals in space than twodimensional programs. with complex chemistry. In addition. later studies adopted instead a variable related to 5. Successful integrations of the ordinary differential equations have been obtained for relatively few flames.4. would be available for adjustment. 5. thus the numerical solutions tend to move away from the correct integral curve whether the integration is performed in the direction of increasing or decreasing z. From this discussion it should be evident that further research on the numerical integration of the laminarflame equations seems warranted. In all these investigations. a procedure that achieves simplification with onestep chemistry and often yields good approximations to the flame speed at early stages in the computation. Criteria therefore must be developed for modifying parameters in a manner designed to overcome these divergences [44]. Thus both classes of the numerical approaches have attractive attributes.4. using a formulation like that in equations (80) and (81). Burningvelocity calculations It has been seen from the viewpoint of asymptotic analysis that when a onestep approximation is applicable for the chemical kinetics. Only the earliest work [43]. [63][85]) and in related work. Many important conclusions concerning chemical kinetics in specific flames have been obtained in the references cited (for example. which could aid in the understanding of flame structures. the ordinary differential equations can provide multiple steadystate solutions when they exist (a relatively difficult task for timedependent methods) and unstable solutions as well.
the flux fractions t i remain equal to their ~ .i = 1. with all eigenvalues being negative.T . and an equation of the form dt/dz = Ao(z) may be derived from equation (SO). then by writing D i j = D i j .Throughout the convectivediffusivezone. ) / P D ~ ~ one . w0/W [in which average molecular weight]. may derive a linear system with constant coefficients from equation (81): d ( X i . KO= (Cj”= cj. . and the derivatives d X i / d z at z = 1 are readily computed from equation (83)./W. If D i j may be assumed to have the same temperature dependence for each pair of species. N. where A is inversely proportional to m2 and o(z) is known.j + i 5 x ~ Leij.Xi. . Flames with MultipleStep Chemistry 171 velocity is determined by matching the solution in a downstream reactivediffusive zone at the hot boundary to the inert solution in the upstream convectivediffusive zone.4. and A0 are constants and the functions g and h are nondimensional.2. A progress variable t for the effective onestep process may be defined. and position is neglected. N . where D i .. .2. h( T ) . . o g ( T )and 1=1 . Then . at 0 = 0.0 >  j = l . ~i = j and The problem in equation (83) is wellposed and solvable by matrix methods. ~ . . These derivatives provide linear relationships between X i and z in the reaction zone that enable the rates wi there to be expressed explicitly in terms of z.)i#j is the initial [Xi.O)/d€J= Xi = j= 1 c CiJ(Xj  N >  xj.5. equation (81) (or the generalization thereof that eliminates assumption 7) reduces to a system of linear ordinary differential equations for X i ( z ) that may be solved subject to the known final equilibrium values X i ..5. when the generally weak dependence of 1 on cominitial values. = ti. i = 1. . a. which changes from 0 to 1 across the reaction zone. as described in Section B. at z = 1.J where X i . . ..o). E ~ . .(T. This conclusion remains true for multiplestep kinetics if the sequence of reactions can be reduced to an effective onestep process through introduction of a suitable combination of steadystate and partialequilibrium approximations. and by defining Leij. o Leij.o = 1. .
this procedure may be extended to provide the concentration and temperature profiles from the solution to equation (83) and the integral of equation (13). The function o(z) may be sufficiently complicated to necessitate a numerical evaluation of the integral for obtaining A. Criteria for the applicability of the extended steadystate approximation have been developed by Giddings and Hirschfelder [64] and improved by Millan and Da Riva [89]. This situation often is encountered in real flames and is largely responsible for their individual structural features.3. and the upstream structure and propagation velocity may be calculated first by neglecting these processes . This procedure fails if an insufficient number of justifiable steadystate and partialequilibrium approximations is available. but within which more than one independent reaction must be considered because of inapplicability of steadystate types of approximations. 5. paralleling that of Section 5. In addition to giving the burning rate. For example.5. This may occur under appropriate conditions for bromine molecules in hydrogenoxygen flames to which HBr has been added. The simplifications that have just been indicated to follow from the application of this approximation underscore the importance of having methods for ascertaining the validity of the approximation.4. Suitable knowledge of the kinetics may still enable asymptotic methods to be employed for calculating burning velocities and some structural aspects. The extended steadystate approximation When the steadystate approximation of Section B. Particular species may be produced and consumed in relatively narrow upstream reaction zones in such a manner that they exert a minimal effect on the burning velocity and can be ignored at first and then calculated later in an analysis specifically addressing their reaction zone. then the recombination region may be analyzed afterward. This situation appears to prevail for the ozone decomposition flame under many conditions. In the rest of this chapter we shall be concerned primarily with methods for testing the validity of the steadystate approximation for chain carriers. There may be a single reaction zone with a reactivediffusive character at the hot boundary.2 is applied to nonisothermal systems involving molecular transport.172 Theory of Laminar Flames provides the first approximation to the burningrate eigenvalue after a suitable expansion for o(z). The discussion . even if the flame speed is not controlled by conditions in the hottest part of the flame. has been introduced. it is sometimes referred to as the extended steadystate approximation. These and other unique characteristics of individual flames may motivate future theoretical studies.2. This necessitates development of special analyses for the structure of the reactivediffusive zone. threebody recombination processes may be sufficiently slow to occur in a downstream convectivereactive zone that does not influence the flame speed.2.6.
and the other species will be denoted by the subscriptsjo = 1. Since the production terms w. the mole fractions of all these reaction intermediaries may be eliminated from the flame equations. [90] are (1) too short a residence time in the flame for the chaincarrier concentration to reach its steadystate value and (2) the diffusion of chain carriers away from regions of maximum chaincarrier concentration. It can be shown [9l] that when X . A statement of the extended steadystate approximation is that throughout the flame.4. (87) . Since reaction intermediaries are present in low concentrations. relations such as equation (85) permit explicit representation of X . . . are functions of z and of the mole fractions X i . wr = 0. in terms of z and the remaining Xi. Factors that may contribute to inapplicability of the steadystate approximation in flames [89].) has been defined by N.1 l/Dr j= 1 C (XjIDrj). then equation (81) reduces to dXrldz = (IR'T/cppWDr)[Xr . consideration will be restricted to systems in which only one intermediary exists. including the X . . Equation (86) may be taken as qualitatively true in flames even though the diffusion velocities of major species differ. Flames with MultipleStep Chemistry 173 here largely follows [89].5. 5.1). equation (81) may be simplified for species r. Conservation equations for reaction intermediaries For the sake of simplicity. equation (80) is dcrldz = (Wm2cp)(wr/f).4. the net mass rate of production of any reaction intermediary r is negligibly small. (86) where the diffusion coefficient for species Y in the multicomponent mixture (D. N . . For species r. this will be denoted by the subscript r.(W/W)trl/f. Since stoichiometry conditions relate the remaining ti. if the diffusion velocities are the same for all species other than r. The discussion and results are applicable qualitatively for a given chain carrier in flames containing any number of chain carriers. 6 1. in terms of z. (85) The integral of equation (80) for i = r is then t r = 0 because the initial (as well as final) chaincarrier concentration is zero. and the problem is reduced to that of a onestep reaction. and the flux fractions of all of these species are zero.only one independent expression remains in equation (80). The flame equations may then be solved explicitly to give all the mole fractions.3. . Thus if equation (85) is valid for all chain carriers.
Substituting equation (89) into equation (88) yields qx./dz from equation (87).1 species other than r because X . convection. The result. < 1) and T.4.4. and reaction for the chain carriers.). (90) R = Lb/m2cp (9 1) is a dimensionless function of T and the X j ' s excluding X. Since the steadystate approximation (w. Then w. < 1 implies that the term corresponding t o j = Y will be small. = 0) implies that x.). The three terms in equation (88) represent diffusion..* 5.cp.. This term.X. = b(a . is a study of which will yield a criterion for the applicability of the extended steadystate approximation.174 Theory of Laminar Flames The sum in the quantity f [equation (82)] appearing in the denominators of equations (86) and (87) now extends only over the N . is of the form W. which goes to zero as D. Hence f depends only on T and c . since X . does not appear. A criterion for the applicability of the extended steadystate approximation For illustrative purposes let us assume that chain breaking occurs through a process that is of the first order with respect to species r. = L R 0 T / c p p W D ./W) and then differentiating the resulting equation and substituting for dE. (89) where a and b may depend on z and all X j except X. in which E . ~ It is reasonable to define the Lewis number for the radical r as Le.)( W.  which may depend on the composition (excluding X . A useful form for equations (86) and (87) may then be obtained by multiplying equation (86) by (f/Le. for j # r.za (92) * It will be noted that the diffusion coefficient of the radical affects only the first term of the operator 9in equation (88). .+ 0.) = R(a in which  X. = L/pD. produces the primary influence of diffusion upon the radical concentration profile. These same three physical effects must appear irrespective of the formulation and would also arise in testing criteria for partial equilibria.
. is a characteristic residence time in the flame. is a characteristic flame thickness [see equation (4)]. Flames with MultipleStep Chemistry 175 from equation (89). also lead essentially to equation (93). such as w.X#?)and w. of course. according to the discussion in Section 5. and [from equation (89)].X. Exactly how large R must be is.n/K”. (94) where 6 = A/mc. 9 tch.5.. Methods of analysis for testing steadystate approximations Since the steadystate approximation applies if R is sufficiently large. Substituri:ig equations (96) and (97) into equation (90) and collecting terms . equation (90) shows that this approximation will be valid if I g ( X r >I/Q 4 Xr . (95) and the steadystate approximation will be valid if t. not determined by these approximate considerations. The value of 0 may be chosen in such a way that R 5 rn’. The time ratio in equation (95) was first introduced by Damkohler [92]. It follows that 6/v = t.. [93] in an entirely different context and is referred to as Damkiihler’s first similarity group..4.‘). Hence R = tres/tch. n=O m (96) where 0 is a constant representative of the magnitude of R. = b(a .4.. hence a criterion for the applicability of the steadystate approximation is that R be large. we expect that accurate predictions of chaincarrier distributions will require a larger R than will accurate predictions of burning velocities.. = b(aX. and other forms for w. The result given in equation (95) is easily shown to be independent of the special assumption introduced with regard to the order of the reaction causing removal of the chain carriers. p / b = t. one approach to solving equation (88) for small deviations from the steady state is to assume that X r = C Xr..4.1. The physical significance of the criterion in equation (93) may be seen by noting that equation (91) implies that 0 = (6/V)/(P/b). (97) where b‘ I 1 and b‘ expresses the variation of R through the flame. 5. is a characteristic chemical time for the destruction of chain carriers. (93) Equation (93) will be satisfied if R is sufficiently large.5.
. 2. Special situations exist for which this procedure simplifies considerably. (99) for small deviations from the steady state. Therefore. ) are known functions of z. n l)/b’. On the other hand. therefore. Sharper definitions of values of consistent with the steadystate approximation (in the zone where it is applied) might thereby be developed. X. = a(z.l(z)/a. near the cold boundary b‘ is small and.o = a for the steadystate distribution and X r . napproaches zero at the hot boundary for n 2 1. )to approach zero [see equation (88)]. Under these circumstances. After this structure is determined. an approach based on matched asymptotic expansions. profile) can therefore be determined completely by setting X . . the series tends to converge slowly upstream [see equation (99)]. a. f goes to zero. . while radical distributions away from the hot reaction zone have relatively little influence on the main characteristics of the flame. may be helpful. = a(z. causing As the hot boundary is approached ( 9 ( X . X. then the burning velocity and the composition profiles of all other species in the flame are virtually unaffected by the presence of this intermediary. b‘. For very small deviations from the steady state.176 Theory of Laminar Flames involving the same powers of gives Xr. Since equation (90) then implies that X r .(z) is then computed from equation (99) and the flame equations are again solved with X . and in the solution procedure described above 9 and b’ need not be recomputed (through the flame equations) in each step because they remain unchanged.). The approach that has been described here is an example of a perturbation method for large Damkohler numbers and may be termed Damkzhlernumber asymptotics. z + l). X j ) + Xr. If the intermediary under consideration is not a chain carrier but is merely produced and consumed through unimportant side reactions. and b’(z) and 9 ( a ) are evaluated. . n = 1. it is often reasonable to omit the recomputation of 9 and b‘ in successive steps even for chain carriers. the series given in equation (96) should converge rapidly as the adiabatic flame temperature is approached. This process is continued until sufficient convergence is obtained. treating different zones differently. . It is possible that the Damkohlernumber expansion will often be good in the hot reaction zone but poor elsewhere. It has been developed on the basis of an expansion that does not distinguish among special zones within the flame. The structure of the flame (excluding the X . = 0 in the flame equations. . n= 9 ( X r . equation (90) reduces to a linear nonhomogeneous differential equation with known variable coefficients. Thus the flame equations are first solved with X . and the coefficients of the linear differential operator 9 ( X .. the operator 9 is computed.
Numerical integrations have been applied successfully to the hydrogenoxygen flame. often with conflicting conclusions.4. An alternative approach to the estimation of minimum acceptable values of is to introduce simplified functionsf(z). the validity of the steadystate approximation for bromine atoms in this flame is less clear and is likely to be significantly dependent on initial conditions. Comparison of the with those for = c c then provides an solutions for finite values of indication of how small can be if the steady state is to be reasonable. and W. there appears to be a wide range of conditions under which the steadystate approximation for hydrogen atoms is excellent. Better rateconstant data are needed to aid in the application of the methods discussed herein to specific flames. Therefore. The primary reason for the differences is uncertainty in values of reactionrate parameters. even though estimates indicate that the steady state is inapplicable. and composition of the reactant mixture.4. + 0 + M. temperature. + + . minimum values of consistent with an acceptable steadystate approximation may be estimated by comparing X. a a a a 5.5.. Key specific reactionrate constants sometimes are uncertain by an order of magnitude at representative flame temperatures. X j ) . 0. Flames with MultipleStep Chemistry 177 On the basis of the approach defined here. This has been done by different investigators. OH. Because of the complexity of the chemical kinetics. Few flames have rate parameters known as accurately as those for the hydrogenoxygen flame.. For example. Le. it appears that the steadystate approximation for oxygen atoms (with the mechanism 0. analytical approaches have been less successful. and H leads to reasonable agreement between calculated and measured burning velocities. M + 0.(z)/Q with a(z. there are conditions under which introduction of the steadystate approximation for the intermediaries 0./W that enable equation (88) to be solved analytically [89]. Observations on theories of flame structure It would be of interest to employ results like those in the preceding section to ascertain the applicability of the steadystate approximation in specific flames. This flame affords a good testing ground for theoretical methods of prediction. these conditions must be specified in stating whether the steadystate assumption is applicable. In addition.(z). 2 100 has been obtained as an approximate criterion [89]. b’(z). at least for the last two of these chain carriers [42]. for example) but inaccurate at high ozone concentrations (above 30%). ) is reasonably accurate at sufficiently low initial concentrations of ozone (between 20% and 30% by volume.6. 0 + 2 0 . the former must be sufficiently small. For the hydrogenbromine flame. For the ozone decomposition flame in an ozoneoxygen mixture initially at room temperature and atmospheric pressure. a(z). In this manner. the validity of simplifying approximations can vary strongly with the initial pressure.
178 Theory of Laminar Flames There is motivation to pursue analytical methods based on asymptotic analysis for the hydrogenoxygen flame. recent numerical integrations for methaneair flames. There are important chemical processes involving species present in low concentrations that exert little or no influence on the flame speed and on other aspects of the flame structure. there may be many different hydrogenoxygen flames. fundamentally justified simplifications applicable for these more complex fuels is a continuing. Often this a posteriori calculation can be simplified by use of activationenergy asymptotics. they may be ignored at first. asymptotic analysis may be restricted less to model kinetics and applied more to real kinetics. It is possible that. In general. A single primary reactivediffusive zone with a rate that can be approximated as a onestep process may control the flame speed for typical hydrocarbonair flames.5) because of the strong temperature dependence of the rate. for methanolair flames. So many elementary steps are involved in soot production in flames that accurate calculations of soot concentrations are difficult whether numerical or analytical methods are employed. asymptotic methods may reveal interesting intricacies of the flame structure.5. in spite of the availability of numerical methods.2. the overall rates may be modified in such a way that fewer variations in flame structure occur as initial conditions are varied. Measured burning velocities are readily correlated by expressions like equation (78). and for other flames with full kinetics (to the extent known at present) show numerous species participating in many reactions [86]. This is especially true for the production of oxides of nitrogen by the Zel’dovich mechanism (Section B. To the extent that these trace species do not influence the flame propagation. although the hydrogenoxygen chain is involved in the kinetics of hydrocarbon combustion. The complexities of the hydrogenoxygen flame underscore uncertainties associated with more complex systems. Identification of useful. such as hydrocarbonair flames. Notable among these are the production of oxides of nitrogen in fuelair combustion and the production of solid carbonaceous materials (soot) in fuelrich hydrocarbon combustion. it may be anticipated that in the future. simplified overall kinetic expressions for soot processes appear to be needed. For the hydrogenoxygen system. Activationenergy asymptotics currently comprise the most promising analytical approaches for studying the hydrogenoxygen flame and other flames. that may be hidden by purely numerical results. worthwhile topic of research. it appears that flamestructure simplifications appropriate to activationenergy asymptotics may vary appreciably with initial conditions. . However. and their production rates may be calculated after the rest of the flame structure has been obtained. Aside from the fact that successful analytical methods provide formulas for the burning velocity and for temperature and concentration profiles. such as combinations of reaction steps to form different overall effective reactions in different parts of the flame.
G. Berkeley. 296 (1957). 1959. 10. 21. D. C. Moscow (1889). Ann. 25. 381400. 13. Zel’dovich. G. Ludford. (1953). G. 389 (1957). 28.” Ph. Univ. E. New York: Wiley. 22. D. FrankKamenetskii. Mikhel’son. Bush and F. 30. Chem. 6. 1951. F. Rosen. Hirschfelder. 7. New York: Pergamon Press. Petrovskii. vol. 0. S. Spalding. V. The Molecular Theory of Gases and Liquids. P. Y. A. E. 158. 0. 27. Anal. 20. von Karman and G. Birkhoff and G. Roy. de Techniche Uitgeverji H. Jet Propulsion 28. B. Bird. Rational Mech. Moscow: Nauka. Predvoditelev. Flames and Explosions of Gases. Landau and E. E. 31 1 (1960). Zhur. NSG1059. Rend. C & F 1. Johnson and W. Piskunov. J. Thesis. Spalding. 1930. Roy. N. 23. 1957. Campbell. Zel’dovich. 22. 24. 1980. E. 100 (1938). B. W. vol. B.” Bulletin of the State University of Moscow. Chapter XIV. Layzer. Sci. 46 (1963). 19. Cambridge: Cambridge University Press. International series. 9. Astronautica Acta 7. 4th Symp. Proc. B. 12. and R. see Collected Works.” report prepared under N. Lewis and G. London: Pergamon Press. Barenblatt. H. Compt. S. Johnson. Chapter 11. D. pp. 474476. L. and N. 171 (1961). D. 29. Zel’dovich and D. Acad. Haarlem (Holland) (1953). Williams. B. and G. le Chatelier. Anniversary Volume on Applied Mechanics Dedicated to C. J. J. S.710 (1953). J. 13. 379 (1883). 8th Symp. S. . London 126A. W. Fendell. J. Znt. (1963). New York: Academic Press. M. S. Grant No. Daniell. 8. C S T 1. 1954.. New York: Wiley. “Study of the diffusion equation with a source term and its application to a biological problem. 4. Section A. Makhviladze. London 249A. E.A. 839 (1958). B. April 1955). Chem. 5 . J. S. and D. Kolmogorov. 2. (1962). Chem. Friedman and E. Paris 157. 58 (1963). Theory of Laminar Flames. 5569. Burke. M. A. Heat Mass Transfer 5. C. A. 1. W. 125 (1937). I. Curtiss.435 (1962). University of California. 219 (1980). Moscow: Novyi Agronom Press. 1st ed.. SOC. Trans. Rational Mech. Phil. Librovich. “On the Normal Ignition Rate of Fulminating Gas Mixtures. Klein. F. Phys. Fluid Mechanics. Nachbar. Mallard and H. 339341 . W. C. Chapter XXIV. 42 (1913). M. Anal. Arch. A. Taffanel. Y. Rosen. A. 14. Curtiss. Chemistry Problems in Jet Propulsion. 417423. Millan. S.421 (1970). B. S. Buckmaster and G. Penner and F. 12. 21. 1955. Th. 17. J. Lifshitz. 18. Phys. Biezeno. W.References 179 REFERENCES 1. G.D. Phys. Khim. Rota. B. 190210. 26.213220. E. (1959). B. “Contribution to the Theory of the Structure of Detonation Waves in Gases. Arch. Y. 32. C & F39. July 1960. G. Emmons. Fiz. SOC. B. 16. I. 7th Symp. Nachbar. G. V.S.A. Mines 4. Combustion. D. Johnson and W. Harwrd University Report (1950). 1.. L. Stam. 1982. Penner. 11. 393 (1930). N. Ordinary Differential Equations. 1954. J. 1959. 287. see also 9th Symp. The Mathematical Theory of Combustion and Explosion. (Sept. 714. 678689. von Elbe. 12. V. 15. 3.222 (1954). R. Feb. Hirschfelder. University of Wisconsin Naval Research Laboratory Repts.
S. C & F 9. 125. Schalit. 1964. N. I. Y . 51. 37. S. “ A Review of Some Theoretical Considerations of Turbulent Flame Structure. Greenberg. AGARD. Fundamentals und Aeronautical Applications.Hirschfelder. 128. G. 389 (1975). 0. Co. 60. 11 16. S. M. 501 (1958). SO. (1963). L. 111. Berman and Y. K. 9th Symp. 57. C & F 5 . Pittsburgh: The Combustion Institute. 56. Van Dyke.. Lovachef. Goldsworthy. 46. 6th Symp. S. CST27. 52.” Colloque Internafional BerthelorVieilleMallardle Chatelier. 41. Korman. Matkowsky. 41 1417.” in AGARD Conference Proceedings No. London : Butterworths Scientific Publications. Fiz. 47. G. B. K. B. S. Fiz. 55. NATO. Perturbation Methods in Applied Mathematics. Clavin. A. Campbell. Prikl. 541. Mass. 63. Clavin. SSSR 245. Perturbation Methods in Fluid Mechanics. 1287. Mitani and F. NACA Tech. 38. Mekh. Joulin and P. C & F29. Barrere.403 (1953). 43. S. J. L. S. 43 (1972). no. Roy. 45. Teor. B. 22. Peters. Th. de Sendagorta. Campbell. Akad. 39. 42. CST 28. Margolis and B. 305 (1961). S. J. J. 876 (1960). Williams. A.. Lovachef. CST21. London M A . 123 (1961). 40. 1271. Selected Combustion Problems. A. S . 39. 164. C & F 5 . 357 (1960). Berman and Y.S. Nauk S. H. K. SOC. J. A.180 Theory of Laminar Flames 31. ed. A. F. Ludford. D. CST2. 7280. B. 43 (1965). 48. Paris (1975). 62. 36. Lovachef. S. 169 (1980). and D. 129. E. New York: Academic Press. B.418426. Campbell. 44. (1981). B. B.R.. 35. C & F 4 . C & F25. 35 (1983). 175 (1980). I. Appl. 59. : Blaisdell Publ. Waltham. Th. English translation. S. J. Ludford. Ryazantsev. 8th Symp. F. (1962). von Karman and S. Margolis and B. E. A. Sen and G . M. 37. Proc. Prikl. Fiz. 139 (1979). von Karman. 75 (1973). 306 (1975). Khudyaev. Zhur. Adler and D. Zhur. 54. 167 (1977). S. 124. J. C & F 39. Semenov. Curtiss. Ryazantsev. G. 32. . I1 11 to I1 125. 53. 1981. J. LXXLXXIV. Heinen. I. Analytical and Numerical Methods for Investigation o f Flow Fields with Chemical Reactions.123. S. V. 61. 149 (1970). DixonLewis. Matkowsky. Seshadri and N. Prikl. A. Zel’dovich. Zel’dovich and N. Memo. F. K. A. Margolis. Akad. 155 (1979). 27 (1948). B. 1427 (1940). T. Khaikin and S . Wilde. E. 1968. 18th Symp. Kapila and G . “Asymptotic Methods in Deflagrations. Dokl. 42. 417424. A. Dokl. 49. C & F 18. S. Eksp. Cole. 34. (1957). S. 58. F. M. Mekh. C. 107 (1982). AGARD. Zhur. 1954. and F. M. Joulin and P. S . Spalding. Nauk. 131. V. Williams. 261 (1975). C & F35. 57. J . Mekh. 1049 (1973). 995 (1959). 10. Chem. Mat. V. SIAM J . F. Penner. Tekhn.S. Mitani. 193 (1982). 120. Ryazantsev. Phys. No. Math. J. Especially Related to Combustion. L. 982 (1982). Khim. A. Williams. First Specialists Meeting (International) of The Combustion Insritute. T. 33. and L. Mat. Y . CST33. Berman and Y. 1282 (1951).
G. 6th Symp. J . 84. Heimerl. K. K. 332338. C. J . (1957). B. F. 90. G. 69. (1981). L. Phys.231 (1973). S. P. 81. C & F33. 385396. Roy. and R. C & F 17. C & F 3 . 834 (1978). M. 18th Symp. Taylor. B. 66. 82. C & F 39. G. B. DixonLewis and R. Zel’dovich and G. Coffee and J. D. 42. (1971). 550 (1949). Stamp. 87. S. Elektrochem. A. 301 (1980). M. (1957). (1981). Bledjian. I. 217 (1978). Proc. Sci. 213221. Millan and I. 5 (1973). Chem. Rhodes. Stephenson and R. Williams. 495 (1967). 140162. P. Altman. 738 (1957). P. Barenblatt.203 (198 1). C & F20. (1957). Energy Combust. (1962). Roy. Penner. 80. Da Riva. D. P. (1977). SOC. S. 0. Comp. Schalit. 846 (1936). 191 (1980). 89. 6th Symp. London: Butterworths Scientific Publications. 71. 76. 86. 8th Symp. Phys. DixonLewis. 27. 1956. 410 (1978). Cook. Cherian. J. London 324A. J. 55 (1971). DixonLewis. B. 67. 68. E. 13th Symp. C & F 43. M. D. L. J. Warnatz. M. G. C & F20. C. 83. 6th Symp. J. CST 26. L. Chem. Hirschfelder. 70572 1. J. 323 (1976). 222236. 18th Symp. Chem. 75. and G. C & F 4 . K. Bunsenges Phys. L. Ber. Z 93. 70. 73. 18th Symp. 199212. 9 (1960). G. Dryer. Coffee. Margolis. L. 7. Spalding and P. and L. G. P. R. Oran and J. J.Hirschfelder. Gilbert and D. Hecker. 61. 88. Spalding. CST23. 643. Warnatz. Phil. 369384. 79. S. Warnatz. Simpson. C. 1 (1981). Taylor. 82. 7th Symp. C & F 26. Campbell. . 65. 72. 193. E. Spalding. SOC. 92. 39841 1. T. Proc. 61 (1959). Smoot. Thompson. 0. Damkohler. 1 (1957). M. S. Adams and G. Curtiss and J. (1981). C. 77. G. Giddings and J. Westbrook and F. J . J. P. (1959). G. Stephenson. 11111 119. Phys. Y. D . Stephenson. A. and G. Roy. e i t . Combustion Researches and Reviews 1955. Simpson. Boris. M. Prog.273 (1981). 0. 17. Trans. London 249A. Tsatsaronis. and G. 315 (1971). Campbell. 91. B. 749767.References 181 64. London 298A4. 85. D. Heimerl and T. K. L. J. B. 16th Symp. E. Westbrook. Soc. V. W. DixonLewis. Hirschfelder. C. 74. S. 78. Day. J.
which is invariably found to propagate at a speed correspondingly closely to the upper ChapmanJouguet point (Chapter 2). We shall 182 . This highspeed wave is a detonation.CHAPTER 6 Detonation Phenomena Let us reconsider the experiment described at the beginning of Section 5. the deflagration wave begins to accelerate markedly. Standing detonations (stationary in the laboratory) may be produced by adjusting air flow conditions in a convergingdiverging nozzle so that a Machreflected normal shock sits downstream from the nozzle exit. The rate of development of the detonation can be enhanced by closing the ends of the tube.1. When the air is preheated to a sufficiently high temperature and fuel (usually hydrogen) is added. and a highspeed (about 3 x lo5 cm/s) nearly planar combustion wave often emerges and travels at a constant velocity through the remainder of the combustible mixture to the end of the tube. plane) strong detonation.1. if the pressure ratio across the shock wave is very large. Detonations can also be initiated by passing a sufficiently strong shock wave through a combustible mixture in a shock tube. then after traveling a distance on the order of ten tube diameters. and the ensuing combustion has been observed to transform the shock into a (steady. then the shock wave ignites the mixture. An unsteady transitional region develops. these are strong detonations that propagate faster than the ChapmanJouguet velocity and that decay slowly toward the ChapmanJouguet condition. If the open tube containing the combustible gas mixture is sufficiently long.
Factors influencing detonation propagation velocities are considered in Section 6. Since the material on detonations is too extensive to be developed fully here. 6. onedimensional. Therefore. cp) and F. Important unsteady and nonplanar aspects of detonation phenomena are treated in Section 6.2.3.2 and are given by equations (535) and (536). beginning at the singular point cp = 1. Properties of the governing equations The solution to equations (535) and (536) is a line in the threedimensional (cp.2. t) space. and Doring [7]. that is. Background material and references to the large body of literature on detonations that existed prior to 1960 may be found in books [S][13] and in reviews [14].2. z is the local velocity to the upstream (x = . Governing equations The structure of a steady.7. gaseous detonations. in which FJz.1. . we shall begin by adopting assumptions 110 of Section 5. It will be recalled that cp is the ratio of the velocity [equation (526)].4.a) normalized dimensionless stagnation temperature [equation (512)]. and t is the reaction progress variable [equation (517)]. 6. [l5]. Detonations in media that are not purely gaseous will be discussed in Section 6. z. von Neumann [6]. initially we shall consider the onestep firstorder reaction R + P in a binary mixture with a Lewis number of unity and an effective Prandtl number o f t . plane. Hypotheses leading to mathematical predictions of the speed of propagation of these waves were given by Chapman [3] and Jouguet [4].1. The status of the mathematical problem is summarized in Section 5.1. the reader is directed to the literature cited above for further details. respectively.1. cp) are defined in equations (532) and (534). z = 0.6. Detonations were first investigated scientifically by Berthelot and Vieille [l] and by Mallard and le Chatelier [2]. planar detonation wave must be understood before investigations of unsteady. In the following section we shall discuss the structure of steady. PLANAR DETONATION STRUCTURE 6.2. Planar Detonation Structure 183 consider the structures and propagation velocities of detonations produced by the methods described above. the wealth of information that has been obtained more recently is covered in reviews [16][22] and in newer books [23][28]. nonplanar phenomena can properly be begun and is most easily illustrated by analyzing the simplest representative system. Significant strides toward improved understanding of detonation structure began in the late 1950s with the gradual recognition of the prevalence of unsteady. The two differential equations describing the wave structure in such a system were derived in Section 5. and reasonable postulates for the wave structure were presented by Zel’dovich [S].1.(z. nonplanar wave patterns.
1. and to sketch integral curves between the singularities using the slopes given by equations (535) and (536).2. 6. t = 1.184 Detonation Phenomena T FIGURE 6. heat conduction precisely balances viscous dissipation. (1) which defines a slanting plane containing the cp axis in Figure 6. The numerator of equation (535) vanishes when and the numerator of equation (536) vanishes when . as illustrated in Figure 6. Schematic illustration of the (cp. In this plane.2. t) region in which the solution for the t = 0 and ending at the singular point cp = cpm. equation (519)].1.1. The denominator for equation (535) and equation (536) vanishes when t = 7. for example. the singularities will be located at the points or lines where both the numerator and the denominator vanish. it is necessary to locate the singularities of equations (535) and (536) in the region of interest. and diffusion vanishes [see Section 5. structure of a combustion wave must lie.1. In order to deduce the general shape of the solution curve. Location of Singularities Since the numerator and the denominator on the righthand side of equation (535) or equation (536) are bounded in the region of interest. to determine the behavior of the solutions in the neighborhoods of these singularities. Physically acceptable solutions must lie entirely within a region shaped like the interior of a semiinfinite beam of unit square cross section. T . z = 1.1.
From equation (531) it can be seen by recalling the meaning of the variables that when equation (6) [that is.2. Solving for z. and mainly on M. (4). z) plane. E ) space] given by equation (6) with the plane given by equation (1) is the singular line of equation (535). their solutions will define cylindrical surfaces. = 0. by equation (6). heat conduction. z. Since the only tdependent condition above is t = 7.3. applicable to all of these curves and easily deduced from equations (6) and (7). are illustrated in Figure 6.2 for the curves labeled M . Planar Detonation Structure 185 Since equations (2) and (3) are independent of E . we may also want to treat the plane as a solution to equation (3) in order to remedy the coldboundary difficulty. t) space on the (cp. we find that this parabola is given by or [solving equation (6) for cp] by where the notation implies that cp+(z) 2 cp(z). The product a‘z was chosen as the vertical coordinate in Figure 6. = 5 and M . the viscous force term in equation (55) is zero.6. The construction of such a projection is aided by Figure 6. because in equation (6) the relationship between this product and cp depends only on y and M . and ( 5 ) imply that the lines t = z = 1 and E = z = 0 may be taken to be the singular lines of importance for equation (536). S ) plane.. z. The intersection of the parabolic surface [in (cp. along this line. the generatrices of which are parallel to the E axis.it will be convenient to view the location of the singularities in a projection of the (cp. z) plane defined by equation (6) is the Rayleigh line [along which equation (22) relates the properties at every point]. viscous effects. and diffusion vanish. Pertinent facts.2. as given by the Rayleighline condition [that is. . in which a’z is sketched as a function of cp for various values of M . Equations (l).1. since y generally lies within a very narrow range of values. equation ( 2 ) ] is valid. Hence the line in the (cp. Equation (532) shows that equation (2) defines a parabola in the (cp. which is the only condition with any degree of complexity]. (4) In view of the discussion in Section 5. From equation (534) it is clear that the solution to equation (3) in the region of interest is the plane 5 = 1.1.2.
4 = (q(1). c. d correspond to strong and weak detonations and weak and strong deflagrations. for M . while for a deflagration (that is. T . cp = ~ ~ (then 0 )yields the equidimensional equation .5). it can be seen that equation (10) is the hotboundary point for a weak detonation or a strong deflagration and equation (11) is the hotboundary point about T = 0. End states at points a. The most important singular points are those common to both equations (535) and (536).1. 1).2. 1. b. Figure 2. T. O)].Deflagrations FIGURE 6. (10) and (Cp. respectively. (cp. (1 1) Equation (7) shows that for a detonation (that is. 4 = (cp+U). 2. = . Schematic diagram illustrating the properties of various Rayleigh lines. < l). 11.2 (for example. equation (8) identifies the coldboundary point [(cp. for M . > l).186 Detonation Phenomena Detonations / \ M. equation (9) represents the coldboundary point. 6 ) = ( L O . By using equation (7) and the results of Section 2.namely.
only one solution reaches the singularity from the permissible region and this solution has T > t near the singularity. Planar Detonation Structure I87 for a strong detonation or a weak deflagration.6. t = 1) from ( T I 1. and therefore cp+(l) = q(l). The nature of the singularities (T = 1. t = 1) [equations (10) and (1 l)] can be deduced by expanding equation (536) about these points. t = 0) [equations (8) and (9)]. only one integral curve reaches (z = 1.t) plane. for any reasonable reaction order.1. t I 1). The slope of the curve that reaches ( T = 1.2. Solutions in the neighborhoods of singular points Let us first consider the relationship between t and z near the points defined by equations (8)(11). Hence equation (536) implies that t = 0 as T increases in the neighborhood of either of these points (that is. For ChapmanJouguet waves. T > t here). cp = ~ ~ (then where a = 0 for cpLt(0) = 1 (the cold boundary) and a = 1 for cp. equation (234) [in which CI = a'y/(y . The general solution to equation (13) is . causing equations (10) and (11) to coalesce into a single point. t = 1) from the permissible region (T I 1 . q) = 0 over a short range of T in order to remedy the coldboundary difficulty (see Section 53. E I I) is which implies that T > t along this curve.l)] implies that the quantity under the radical sign in equation (7) vanishes at T = 1. however. In the neighborhood of ( T = 0 . In this manner.2). all others bend away from the singularities.(z. 6.1. The qualitative properties of the singularity do depend on our choice of a firstorder reaction. these singularities are found to be saddle points in the (7. we must assume that F. Let us next investigate the relationship between T and cp near the points defined by equations (8)(11).(O) # 1 (the other singularity at T = 0).2. Equation (12) shows that the nature of the singularity depends quantitatively but not qualitatively on the value of cp. For equations (8) and (9) we may assume (in view of the preceding discussion) that 6 = 0. Expanding equation (535) 0 )yields the equidimensional equation about T = 0.
c = 0). Thus the cold boundary (cp = 1. and that [cp = cp(O) = 1. z = 01 is a nodal point for detonations. and such singularities are nodal points in the (cp. z = 0) is a saddle point in the (cp.l)”(Mi . z = 11 (the end state for strong detonations or weak deflagrations) always exhibits a nodal type of behavior in the (cp. this is the primary reason that the burning velocity is an eigenvalue for deflagrations. but c = 0 when ( . and the point [cp = cp(l).1) > 0.1) < 0 (that is. while wavestructure solutions exist with any propagation velocity (greater than or equal to the ChapmanJouguet wave speed) for detonations.3. z) plane for detonations. These results enable us to use equation (535) and the previously deduced properties of the singularities in the (cp. it can be deduced from equation (13) or (14) that the singularity is a saddle point in the (z. z = 11 (the end state for weak detonations or strong deflagrations) always exhibits a saddle type of behavior in the (cp. z) plane in the neighborhoods of these points is somewhat more complicated than analysis for the other singularities . Nevertheless. z) plane.1) > 0.3 clearly illustrates the nodal type of behavior of solutions about cp+(O) and cp(l) and the saddle type of behavior of solutions about cp (O) and cp+( 1). Thus an infinite number of integral curves. as stated previously. > 1 for detonations and Mg < 1 for deflagrations. The encircled signs are the signs of the slopes of the integral curves in regions bounded . z = 11 [equations (10) and (1 l)] have been considered in detail by Hirschfelder and Curtiss [29]. Since M . equation (536) and the result that z > t in the neighborhoods of the hot and cold boundaries suggest that z > t throughout the range 0 < z < 1 and that t varies monotonically with z in this region. z) plane in order to construct the projection of the solution curves on the (cp. the point [cp = cp+(l). The cold boundary (cp = 1) is located at cp (O) for deflagrations and at cp+(O) for detonations. z) plane. General properties of the integral curves The (t. z = 01 is a saddle point and [cp = cp+(O). Analysis of the integral curves in the (cp. M.depending on the type of wave. cp) plane. for all realistic reactions.188 Detonation Phenomena where c is an arbitrary constant when ( . z) plane. we conclude that [cp = cp(O). z) plane for deflagrations and a nodal point in the (cp. z = 01 is a nodal point for deflagrations.2.1)”(Mi . the precise properties of the solutions differ for various ranges of values of the parameters (that is. The singularities [cp = cp+(l). z = 11 and [cp = cp(l).3. Figure 6. in fact.l)”(Mi . The hot boundary is located at cp(l) or cp+(l).) and depend on the order of the reaction. z) plane. when (.t) relationship along the solution curves is relatively simple. pass through the singularity when (. all integral curves (with t = 0) in the neighborhood of the singular point. This projection is shown in Figure 6.1) 5 0 [since cp must approach cp*(O) as z goes to zero]. 6. z = 01 is a saddle point and [cp = cp+(O) = 1.1. On the other hand.l)”(Mi .
z. Similarly.3 corresponds to a different value of some wave parameter (for example. c) space.6. z = 1. strong detonations exist for any propagation speed (above the upper ChapmanJouguet wave speed). It should be emphasized that for given parameters ( M o . solutions to equations (535) and (536) for the structure of strong deflagrations do not exist. that is. and the Rayleigh line. an infinite number of integral curves connect these two points in Figure 6. It is interesting that consideration of the basic (elliptic versus hyperbolic) character of the external . that is. Since cp+(O) and cp(l) are both nodes. no integral curves pass between cp(O) and cp+(l)in Figure 6. that is. that is. ) . by z = 0. weak detonations can exist only for particular reactionrate functions at any given wave speed. Finally. Thus. only one integral curve (the curve labeled b) passes between cp(O) and cp( 1). each solution cutve shown in Figure 6. Planar Detonation Structure 189 i Rayleigh line FIGURE 6. Hence. and so on) the ( E . z) behavior of the solution ensures that at most one integral curve from the coldboundary singularity reaches the hotboundary point in (cp. only one integral curve (the curve labeled f ) passes between cp+(O) and cp+(l)in Figure 6. regardless of the reaction rate. z) plane] and therefore do not satisfy the hotboundary conditions.3.3. these signs depend only on the result that z > E for 0 < z < 1 and the assumption (adopted for clarity) that z < E for z > 1 and z < 0. the reactionrate frequency factor B .. B.3. Schematic diagram of the projection of solution curves of (535) and (536) on the (cp. On the other hand. weak deflagrations propagate with a definite wave speed that depends on the reaction rate.E l . 7 ) plane.1.(l).3. acceptable solutions passing between these two points will exist for a range of values of the wave parameters. z = 11do so with E # 1 [and hence with infinite slope in the (cp. all other integral curves from the cold boundary that pass through [cp = cp.
z = 0. the result has been found to be true. At t = z. Equation (14) shows that the initial .z = 1. equation (536) shows that z continues to decrease as t (or 5 ) increases. Searches for strong deflagrations in combustion could focus on systems with both extraordinarily high burning velocities and endothermic attributes near the hot boundary. Since equations (533) and (534) imply that dt/d< > 0. and therefore z becomes less than E. However. Solutions for structures of strong deflagrations may be obtained if an endothermic process follows the exothermic reaction [3 13.3. the solution curve leaves the cold boundary (cp = 1. with M o < 1. is very small compared with unity and the Rayleigh line is very nearly a straight line with slope d(cr’z)/dcp = 1 in the range 0 I z I 1 (see Figure 6. it follows that z begins to decrease as 4 (or E ) increases.1. equation (519) shows that dz/d< = 0. Since the above argument relies on the assumptions used to derive equations (535) and (536).3.190 Detonation Phenomena flow. Since the result depends on the structure of the wave. with the wave treated as a discontinuity. the results pertain to different questions and are not in conflict. a Lewis number of unity. it is impossible to invent a generally valid proof. a Prandtl number of 2. the integral curve must cross the Rayleigh line at some point between z = 0 and z = 1 in order to reach the hotboundary point [cp = cp+(l). We may therefore conclude that solutions to equations (535) and (536) cannot satisfy the boundary conditions for strong deflagrations. M . leads to roughly the opposite variation in the number of free parameters [30] . 6.2). all known deflagrations in the laboratory are weak deflagrations. z) plane of the integral curve on the Rayleigh line is infinite (unless E = z) and the slope in the region to the left of the Rayleigh line is positive (unless t 2 z) the integral curve cannot reach the Rayleigh line unless E becomes equal to z at some point in the region 0 < z < 1. t = 11 of a strong deflagration. as is illustrated at cp(O) in Figure 6. Hence z will not approach the required hotboundary condition z = 1 at E = 1. and so on. while the 5: derivative of equation (519) shows that d 2 z / d t 2 = dc/d<. z) plane which exceeds that of the Rayleigh line. These solutions have been proposed for astrophysical problems of stellar and galactic winds involving ionizationrecombination fronts [32]. Since the slope in the (cp. for all combustion systems that have been studied. we have shown only that strong deflagrations are impossible in binary systems with a firstorder reaction. Hence. From equations (7) and (14) it is easily shown that. For ordinary deflagration waves. Once t exceeds z. Currently it is unclear whether fundamental reactionrate limitations will preclude the occurrence of strong deflagrations in combustion. Remarks on deflagrations A more detailed demonstration that strong deflagrations do not exist can be given. t = 0) with z > E and with a slope in the (cp.
equations (531) and (17) determine the approximate structure of the detonation. € =z (16) throughout the detonation wave. Since the change in t through the detonation wave is A€ = 1. Since the (cp. In terms of dimensional variables. approximately. Equations (16) and (524) show that. the dimensionless reactionrate function F. Since 6 = T = 0 along curve a.4. In view of equations (534) and (15). the temperature must become relatively large in the detonation wave before T begins to change appreciably from its value (T = 0) at the cold boundary. An estimate of the order of magnitude of the lefthand side of equation (519) then shows that dz/d< 4 1 (since the change in T through the wave is AT = l).. Approximate solution for the structure of a detonation Since detonations propagate at supersonic velocities. this condition implies that T/T.3.3. + 0. approximately. diffusion is negligible. T ) projection of the solution curve lies between the Rayleigh line and the straight line through the cold boundary with this initial slope. Hence. and therefore equation (519) implies that. defined in equation (534). is much smaller for detonations.(z. t = Y in the wave.3. The solution curve c is closely approximated by curve a followed by curve b in Figure 6. 6. (17) which serves to eliminate t from the problem.1. equations (15) and (533) imply that a representative dimensionless distance scale A< is large compared with unity for detonations (it is of the order of unity for deflagrations). and the detonation structure will be represented by a curve similar to the one labeled c in Figure 6. which is equivalent to the assumption (introduced in Section 5. that is. and the structure of the upstream part of the wave is obtained by integrating .1) of an isobaric wave. Planar Detonation Structure 191 slope of the integral curve at the cold boundary approaches this same value as M . Consequently. is sufficiently small.(T. no reaction occurs along this curve. the upstream portion of the detonation structure will lie very near the line T = 0 connecting the points cp+(O) and cp(O) in Figure 6. the solution itself must have d(cc'z)/dcp = 1 everywhere when M . typically F. while deflagrations propagate at low subsonic velocities. = u/u. Equations (16) and (533) lead to d W = FAT.1.(z. the mass flow rate m is much larger for detonations than for deflagrations. cp). cp) 5 10 in a deflagration wave and F.3.cp). cp) 5 41 (15) in a detonation wave. This result has a number of important implications concerning detonation structure.6.
Discussion of detonation structure We have inferred that for planar detonations. the integral in equation (19) must generally be evaluated numerically.F (C52). The integration can be performed analytically. the solution to which is cp = cp(z).[z.4). the result is where t1is an arbitrary constant. cp(z)]/(1 r+ 1 . For the highly unrealistic case in which the activation energy is zero (El = 0) and the pressure and temperature dependences of A and B1are such that the factor in brackets in equation (534) is a constant.. . curve c in Figure 6. and the structure of the downstream part of the wave is obtained by integrating equation (17) with cp = cp(z). Since curve c lies close to curve a followed by curve b. The relationship between pressure and volume along this curve is illustrated in Figure 6. namely.5. Since . equation (19) shows that the asymptotic 5 dependence of z (as 5 .4 (the labels a through fidentify corresponding curves in Figures 6. However. 6. along curve b (the downstream part of the wave) the term dq/d< may be neglected in equation (531).3 and 6. where cp(z) is given by equation (7). Hence. (20) where t2is a constant and the dimensionless specific reactionrate constant at 5 = cc is F E lim {t'.z)} = constant. Since equation (527) shows that the temperature T depends on both cp and z and equation (7) shows that cp(z) is a fairly complicated function. cp) = 0.3 represents the most reasonable structure. curve b lies very close to the Rayleigh line. Since the characteristic distance over which properties change appreciably is A5 % 1.(z. we concluded that these two curves afford a good approximation to the structure of most planar detonations. equation (20) is the solution to equation (17) throughout the downstream portion of the wave.1.192 Detonation Phenomena equation (531) with z = 0 between cp+(O) and cp(O). The relationship between cp and z along curve b is therefore given approximately by F. cc) is * = 1 .
This structure excludes weak detonations.4). for the most commonChapmanJouguet wavesmore recent studies. which require special consideration (see Section 6. von Neumann [6].  l/p no heat release occurs along curve a.1. [34][37]. and Doring [7] and may be referred to as the ZND wave structure.2.3.2. The structure of a highspeed “deflagration” of this type necessarily differs greatly from the deflagration structure discussed in Chapter 5.6 and 6. It is likely to apply to sufficiently strong detonations (overdrivenwaves. planar detonation structure (for example. Hence a good approximation to the usual detonation structure is a shock wave followed by a deflagration. the shock wave heats the reactants to a temperature at which they can react at a rate that is high enough for the ensuing deflagration to propagate as fast as the shock.2). see review in [38]).4. Planar Detonation Structure 193 Shock Hugoniot (Hugoniot curve for zero heat release) Hugoniot curve von Neurnann spike Final state for strong detonation a Rayleigh line P I Initial state’ ( j I I I \\ 1/PO FIGURE 6.6. this curve represents the structure of a normal shock wave in the reactant gas. However. The ab (shockdeflagration) structure of detonations was first proposed by Zel’dovich [ S ] .see Sections 6. . it was widely believed that the ZND structure was representative of most real detonations. Schematic illustration of detonation structure in a pressurevolume plane.3). [33]) and of good experimental measurements performed mainly in the 1950s (for example. This is quite reasonable on physical grounds. On the basis of numerical calculations of steady. Curve b represents the structure of a deflagration with upstream conditions given by the point q ( 0 ) (which is called the von Neumann spike in Figure 6.
for example. This increase should not substantially modify our other conclusions regarding the wave structure. established that this structure usually is unstable and that. [22][25]. Kinetictheory analyses of strong shock waves in nonreactive gases (for example. The result that Ax is of the order of a few mean free paths in the shock wave casts doubt upon the validity of our continuum equations for the upstream portion of a detonation. planar waves. [16][18]. the chemical reaction rate would have to approach the maximum conceivable reaction rate (that is.5 and 6.1. is the initial velocity of sound.194 Detonation Phenomena to be discussed in Sections 6. the kinetic theory expression for the viscosity p [for example. A< y for detonations with a reasonable amount of heat release. The structure of ZND detonations The approximate solution given in equation (18) for the structure of the upstream part of the detonation wave was first derived by Becker [41] . . further deductions on the basis of the ZND structure deserve to be pursued here. [38]).1). Therefore. Knowledge of the ZND structure nevertheless is essential to the understanding of the threedimensional structure (to be discussed in Section 6. equation (E27) combined with equation (E34)] indicates that Ax is of the order of a few molecular mean free paths. as given by equation (18). planar wave only on the average (for reviews see. the shock waves that occur in detonations are strong enough to be treated as discontinuities for nearly all purposes. is of the same order of magnitude as molecular velocities. Since a.2. molecules suffer only two or three collisions in passing through shocks of the strengths ordinarily encountered in detonations. [39]) yield wave thicknesses that are nearly twice as large as those obtained from continuum theory and that are in better agreement with the few available experimental results [40]. is A< yMg/ ( M : . in fact.a0 M .3. since chemical reactions can proceed only through molecular collisions. but such analyses would be expected to show a similar increase in the width of the upstream (shock) portion of the detonation wave. There is a fundamental reason that large departures from the ZND structure (such as curve e in Figures 6. The dimensionless distance over which cp changes appreciably in the shock wave. In general. this implies that the width of the shock wave in physical dimensions is Ax 4yp/3m = 4yp/3p.    6. It appears probable that such highly reactive mixtures usually would explode in a more or less homogeneous manner instead of supporting a propagating detonation wave. where a. Hence. Thus. the intermolecular collision rate) before an appreciable amount of heat release could occur in the shock wave.4) are not to be expected for steady. that is.6.3). In view of equations (511) and (518) and the fact that K z 0. No kinetictheory analyses of detonation structure have been reported.3 and 6. these detonations generally possess a complex threedimensional substructure involving transverse waves that produce a steady.
Since no reaction occurs in the shock wave. The absence of significant molecular transport removes the eigenvalue character of the deflagration problem. Representative variation of flow properties in a detonation wave.co to + 00. it does not depend on assumptions 9 and 10 of Section 52). The spatial dependence of the flow variables for a shock wave in a gas is illustrated by the upstream (lefthand) portion of the curves in Figure 6. The spatial dependence of the flow FIGURE 6. which can be shown to violate the second law of thermodynamics in ideal gases with constant specific heats. and the chemical reaction rate is not high enough to produce these large gradients. simple analytical solutions for the shockwave structure cannot be obtained from the continuum equations unless the effective Prandtl number is 1 (assumption 8) [42]. . Planar Detonation Structure 195 for shock waves in nonreactive gases. very large gradients would be necessary to generate transport fluxes comparable in magnitude to the convective fluxes. Equation (18) shows that as 5 increases from . and diffusion) are negligible.1. (that is. equation (18) is valid for arbitrary reaction mechanisms and diffusion coefficients. The solution given in equation (19) for the downstream part of the detonation wave represents a highspeed deflagration in which the kinetic energy is appreciable but the transport processes (viscosity. thereby making it essentially an initialvalue problem .the chemical reaction rate influences the length of the reaction zone but not directly the speed of the wave.6. The difference i s primarily a consequence of the much larger mass flow rate (or convection velocity) in detonations.5.5. cp decreases monotonically from cp+(O) to cp(O). There is no solution for the structure of a wave in which cp increases from cp(O) to cp+(O) as t increases. [43]. heat conduction. this would represent a finiteamplitude expansion wave. On the other hand. This deflagration wave is therefore radically different from those discussed in Chapter 5.
5. X. Figure 6. . momentum.5. a small ambiguity arises in connecting the solutions given in equations (18) and (19). . and pressure to Xi. temperature. In view of the range of validity of equations (18) and (19) discussed above. while (for Arrhenius rate functions) the deflagration extends from a finite value of 5 to 5 = + 00. equation (19) can also be made to apply to a system involving any onestep reaction . the shock wave extends from 5 = . the continuity. of course. = J. and state equations reduce to algebraic expressions relating velocity. increases) along the Rayleigh line when the Mach number is between 1/& and 1. the temperature decreases (and the Mach number. isothermal. the equation does not depend on assumptions 8 and 10 of Section 5. . since transport by suitably modifying the definition of F c ( ~cp). N (21) X i= mole fraction of species i.cp(O)]/(p_(O) = 6.(X. ZND detonations involving chain reactions can be analyzed more easily than the deflagrations with chain reactions discussed in Section 5.00 to 5 = + co. Since transport effects are neglected in equation (19). the shock wave must be cut off at an arbitrary point and tacked onto the beginning of the deflagration in order to obtain a singlevalued solution for the detonation structure.2.5 are modified imperceptibly when 6 is varied from lo’ * The restriction to onestep reactions may be removed by replacing equation (19) by equation (21). Because of the high velocity. and in using equations (18) and (19) without modification.2. the problem of computing the structure of the downstream portion of the detonation wave for any reaction mechanism merely involves a forward integration of a set of equations of the form dXi/dx (x = distance. effects are negligible. N = number of chemical species). .4.. where 6 is a number that is small compared with unity. reacting system (see Appendix B).. provided that the rate constants and heats of reaction are known for all elementary reaction steps..* However.196 Detonation Phenomena variables in the deflagration is illustrated in the downstream (righthand) part of Figure 6. This effect is not present for most strong detonations . . it occurs for ChapmanJouguet waves because. the pressure is not constant in the deflagration region of Figure 6. Hence the two waves overlap.5 shows a small decrease in temperature as the hot boundary is approached. these two equations describe the structure of a ZND detonation involving any onestep reaction and satisfying assumptions 18 of Section 5. energy. Clearly. i = 1. Equation (21) is comparable in complexity to the equations governing the time dependence of the concentrations of chemical species in a homogeneous. Mathematically. .). In fact. The cutoff point may be defined as the point at which [cp . as heat is added. . . For typical detonations the width of the shock wave is so small in comparison with the width of the reaction zone that curves similar to those shown in Figure 6. density.
This conclusion appears to apply not only to steady.1.2. An asymptotic analysis with the reaction rate assumed to be zero below an ignition temperature and constant above it has been reported [46]. Cheng. the shock wave may be treated as a discontinuity. ZND detonation is facilitated by the introduction of simplifications in the chemical kinetics. by means of a boundarylayer expansion near the shock and an expansion in powers of l/A( near the Rayleigh line. . the “support” provided by the chemical reactions is of a nonplanar compressible gasdynamic character with negligible molecular transport. Both the viscous interaction between the shock and the reaction region and the molecular transport within the reaction region are small perturbations that do not appear to exert qualitatively significant influences on the wave structure. For purposes of further analyses of detonation structure. Asymptotic Theory of Shock Supported Arrhenius Reactions and Ignition Delay in Gas. Basic considerations for plane waves We have seen that considerations of detonation structure do not restrict the detonation propagation velocity (at least for strong and ChapmanJouguet waves). 6. threedimensional structures. “An Analytic. Analysis of the reaction region of a steady.2. K. Activationenergy asymptotics with the chemistry approximated as a onestep process have been introduced on the basis of physical reasoning [13] and through formalisms of matched asymptotic expansions [47]. Here we shall first consider detonations produced by * See also H. Therefore. The work in [47] involves a double limit that also describes slight merging of the shock with the reaction zone. planar waves but also to unsteady.” unpublished (1970) for an early version of the study in [48]. Morecomplex approximations to the chemical kinetics generally have required numerical integrations [33]. A related expansion was developed independently [45]. thereby providing an analytical approximation to curve c in Figures 6.2. [49]. planar. it affords one helpful simplification in the complicated analyses of transverse wave structures. It also alters the interpretation of a detonation as a deflagrationsupported shock.3 and 6. PROPAGATION VELOCITIES FOR DETONATIONS TRAVELING IN TUBES 6. he has obtained a natural connection between the shock and deflagration solutions. wave speeds must depend on the experimental configuration and can be discussed only within the framework of a class of experiments.6. Propagation Velocities for Detonations Traveling in Tubes I97 to Wood [44] recognized that ambiguities of this kind are characteristic of problems of the boundarylayer type.4. [48]* to identify an induction zone followed by a shorter heatrelease zone and to obtain asymptotic expressions for the length of the induction zone.
[34]). Hence (compare Figure 2.oo) 1 t ' t . detonations in such systems are found to travel at the ChapmanJouguet wave speed after an unsteady period of development. and zero for ChapmanJouguet waves. Since the Mach number behind a strong detonation is subsonic (relative to the detonation wave). this rarefaction wave will overtake the detonation and therefore weaken it. For a strong detonation propagating into a quiescent gas mixture.5) weak detonations Closed end of tube (moving with velocity .198 Detonation Phenomena igniting a combustible mixture with a spark at the closed end of a tube. all velocities in the + x direction. Detonation wave (stationary) c . strong detonations are easily ruled out by dynamic considerations. = l). the rarefaction wave can no longer affect the detonation. Brinkley and Kirkwood [SO] have formalized the dynamic argument by analyzing a model in which the detonation is of infinitesimal thickness and the pressure behind the wave initially decreases monotonically (with finite slope) toward the closed end of the tube.6.6). Since the gas velocity is necessarily zero at the closed end of the tube. Many experimental studies have been concerned with systems of this kind. Such a velocity increase requires that a rarefaction wave follows the detonation. the gas velocity behind the detonation is directed toward the wave (in the laboratory coordinate system) and is subsonic with respect to the wave (see Chapter 2). Empirically. They find that the time rate of increase of the pressure immediately behind the detonation wave is negative for strong detonations. there is a net increase in the gas velocity in the direction of propagation of the wave (in the laboratory frame) between the end of the tube and the detonation. This same qualitative argument cannot be used to rule out weak detonations. When the ChapmanJouguet condition is reached ( M . Schematic illustration of the velocity profile behind a detonation wave propagating in a tube.c FIGURE 6. decreasing the detonation velocity until the Mach number behind the wave becomes sonic. The presence of the rarefaction wave is well substantiated experimentally (for example. . this velocity change becomes an increase from a subsonic speed immediately behind the detonation to a supersonic speed at the closed end of the tube (see Figure 6. Steady. positive for weak detonations. When transformed to a coordinate system moving with the detonation velocity.
6.3. for example). A ZND structure is not possible for weak detonations. 6. it soon became apparent that these waves do not propagate precisely at the .3. Since a normal shock wave would raise the temperature above the saturation point (thus ruling out the ZND structure. the rate of condensation increases as the temperature decreases. and chemical reaction rates generally are not high enough to produce a direct transition to the end state.531.3 leads to the idea that a condensation wave can exist in a supersonic nozzle flow only at a definite location that depends on the wave structure.6. Proposals that weak detonations also represent various processes occurring in geological transformations have been presented [52]. Theoretical analyses of condensation waves for steady supersonic flows in nozzles are based primarily on the early work of Oswatitsch C.2.1. Good agreement with the experimental measurements on the location and width of the condensation wave is obtained.2. However. Further comments on weak detonations A very convincing argument against the existence of weak detonations utilizes the results on detonation structure presented in Section 6. Effects of tube walls As experiments on detonation waves in tubes were refined. it appears reasonable to assume that condensation shocks are weak detonations. this success does not produce simple formulas for predicting the onset of condensation. and the flow is observed to be supersonic downstream from the condensation wave. This hypothesis may be supported by the fact that unlike chemical reaction rates. Simplified approaches to the analysis of condensation waves can provide useful results and improved understanding of the wave structure. The amount of liquid formed is so small that the equations for purely gaseous waves are expected to apply approximately.2. [Sl].2. Propagation Velocities for Detonations Traveling in Tubes 199 (as well as strong detonations) approach the ChapmanJouguet condition as time increases for the BrinkleyKirkwood model. Numerical integrations of sets of ordinary differential equations are performed to describe homogeneous nucleation and droplet growth in supersonic flows with negligible molecular transport in the streamwise direction. Weak detonations are believed to represent the “condensation shocks” observed in supersonic wind tunnels [12]. Supercooled water vapor in a supersonic stream has been observed to condense rapidly through a narrow wave. for example. One such approach has been reported recently [56]. Consideration of Figure 6.2. The approach is discussed in [54] and [55]. This idea of treating the condensationwave problem as an eigenvalue problem reminiscent of the deflagration problem has not yet been exploited with the objective of obtaining simplified descriptions of condensation waves.
7.7. influences the detonation velocity to a negligible extent. the wave speed now depends on the Shock front ’1 Reaction zone+ ‘&(Mell TBoundary layer Tube wall FIGURE 6. [57]). Furthermore. For example. the velocity in the boundary layer exceeds the velocity in the reaction zone. standard experimental practice is to plot the detonation velocity as a function of the reciprocal of the tube diameter and to extrapolate to zero to obtain the “true” experimental wave speed (for example. Hence. who accounted for the influence of the boundary layer behind the shock front in a ChapmanJouguet wave with the ZND structure. On the other hand. the fact that the flow diverges throughout the reaction zone does affect the detonation velocity. this effect has been observed but. Schematic diagram of a detonation near a tube wall [SS]. (46) through (49)] for the reaction zone (excluding the boundary layer). The first satisfactory explanation of these effects was presented by Fay [58]. then the ideal gas equation of state implies that the density in the boundary layer exceeds the density in the reaction zone. . where a coordinate system that is stationary with respect to the detonation wave has again been adopted. since the equations of Chapter 2 must now be replaced by quasionedimensional flow equations with a small rate of increase of area [that is. detonation velocities decrease approximately linearly with the reciprocal of the tube diameter at fixed initial conditions.200 Detonation Phenomena ChapmanJouguet speed. Conditions relating properties across oblique shock waves readily show that this divergence implies that the shock front must be slightly convex toward the unburned gas. the mass flux pv in the boundary layer exceeds the mass flux in the reaction zone. Also. if it is assumed (quite reasonably) that the pressure is independent of the coordinate y (normal to the direction of propagation of the wave) and that the wall of the tube is maintained approximately at room temperature. in itself. The calculation of detonation velocities from the quasionedimensional flow equations is considerably more difficult than the calculations discussed in Chapter 2. detonation velocities have been observed to decrease approximately linearly with the reciprocal of the initial pressure in a tube of fixed diameter. For example. Fay’s model is illustrated in Figure 6. and therefore as the boundary layer grows. Geometric considerations clearly imply that in this frame. the streamlines in the reaction zone must diverge (at a rate determined by the rate of increase of the displacement thickness of the boundary layer).
On the basis of approximate calculations of the detonation velocity for representative systems containing stoichiometric hydrogenoxygen mixtures [58]. Another problem lies in determining where to apply the downstream ChapmanJouguet condition M . which is empirically [34] approximately inversely proportional to the pressure) is in rough agreement with the observed pressure dependence. from M < 1 to M > 1. Ambiguities associated with frozen and equilibrium sound speeds If a detonation is viewed as a plane ChapmanJouguet wave traveling steadily in a tube in a direction away from its closed end. the condition cannot be applied at x = co because the boundary layer would eventually occupy the entire cross section of the tube at some finite value of x. This result exhibits the observed inverse dependence on diameter. Fay solved this last problem by noting that the effects of area increase and of heat addition counteract each other in quasionedimensional flow (in the subsonic regime. Fay obtained numerical values of Avo/vo which varied from 1 % to 4 % and which were in good agreement with experimental results (see Table 6.co.3. the propagation velocity is influenced by the wave structure. Propagation Velocities for Detonations Traveling in Tubes 20 1 profiles of the static and dynamic pressures in the reaction zone. A phenomenon similar to that occurring at the throat of a convergingdivergingnozzle can therefore occur. addition of heat increases M and an increase in area decreases M).3). = 1. that is. the flow can experience a smooth transition through M = 1.1 6*/d. at a point where the rate of increase of area is in the proper proportion to the rate of increase of stagnation enthalpy. the fact that the characteristics propagate at the frozen sound speed implies [59] that the upstream end of the rarefaction wave that follows the detonation (Section .3. From the standpoint of the method of characteristics for reacting gas mixtures (Section 4.2. then we may question whether the result that v. By estimating the reaction rate from schlieren studies of detonations and by utilizing an empirical formula for the rate of growth of a supersonic turbulent boundary layer.2. = ue. the fractional decrease in the detonation speed (below its classical value) is found to be Avo/vo = 2. obtained in Section 2. 6. and (since 6* should be approximately proportional to the width of the reaction zone. The ChapmanJouguet condition must therefore be applied at the finite value of x at which the rate of growth of the boundary layer is appropriately related to the reaction rate. and d is the diameter of the tube. Events occurring downstream from this ( M = 1) plane cannot affect the detonation because in the downstream region the velocity with respect to the wave is supersonic both inside and outside of the boundary layer.1).4. is consistent with the endwall condition.6. where 6* is the boundarylayer displacement thickness at the plane where M = 1.
4 1. Comparison of calculated and observed fractional velocity deficit at 1atm pressure in 1cm radius tubes [SS] DO. 2H2 2.1) can smoothly join the downstream end of the detonation wave only at a point where the flow velocity relative to the detonation wave is the frozen sound speed af . Possibilities that have been considered for resolving these difficulties include the investigation of “pathological” [59] systems in which equation (22) (which merely states that the enthalpy change is zero when a small amount of reaction occurs at constant pressure and density) is satisfied with some w j # 0 and the analysis of systems with endothermic reactions near the hot boundary (to aid in a DC transition) [60]. the end states for detonations propagating in closed tubes correspond to the frozen ChapmanJouguet condition indicated by point C in Figure 6. where the Mach number based on the frozen sound speed is unity. From Figure 6. The gas mixture in a ZND wave would meet the equilibrium Hugoniot at point D in Figure 6.3. 16% HZ + 84% 2. uo AVO  x 102. [61].0 2.e. 2H2 + + 13A 1. a point situated slightly onto the weakdetonation branch of the equilibrium Hugoniot.1. which is the statement of chemical equilibrium for the system.6 1.8 before reaching state C . Since equation (22) will be satisfied if w j = 0 for all j .3 1. Composition AVO . p.I. 53% CzHz + 47% air 1552 2820 6.202 Detonation Phenomena TABLE 6.6 1. = af. 2H2 + + 13He 4. 2H2 0 2 0 2 0 2 0 2 5. Although it is clear that there are hotboundary complexities associated with the behaviors of planar detonations having . and (491)] N j= 1 1 wj(dh/a ?:>. from the requirement that the pressure and velocity gradients must remain finite where rarefaction joins the detonation).2.4 + + + 5He 3.8 it may be seen that this conclusion poses difficulties in connection with the wave structure. This result suggests that contrary to the conclusion reached in Section 2..5 4. From the characteristic analysis (specifically.3. it appears that the detonation will propagate steadily if its downstream (chemical equilibrium) condition is applied at v. (22) where the notation is that introduced in Chapter 1.7 1. and a transition from D to C is difficult to visualize. m . i. Y.2 2. (455).0 observed 1. m/sec 2830 3660 3823 1522 0 2 00 calculated 1.x 102.3 6.8 [59].(i + j ) = 0.2 3. it is found that the condition which must be applied at this point is [compare equations (454).
Propagation Velocities for Detonations Traveling in Tubes 203 P ilibrium Hugoniot) Equilibrium ChapmanJouguet point Frozen ChapmanJouguet point negative of the square he detonation velocity) FIGURE 6. # ue.3. To substantiate this speculation remains a challenging theoretical problem. The prevalence of the threedimensional phenomenon. largely through the onset of a reduced average extent of combustion.5. 6. the threedimensional effects will result in average propagation velocities (after removal of the wall effects discussed in Section 6.2. necessitates its consideration in the development of accurate theories for propagation velocities.3. It seems likely that when the characteristic dimensions of the nonplanar disturbances are small compared with the tube diameter.1) in weak mixtures to “turbulent” detonations [62] in strong mixtures. . Effects of threedimensional structures Observations of threedimensional structures for ChapmanJouquet detonations propagating in closed tubes range from spinning detonations (Section 6.8.2.3) slightly in excess of the theoretical onedimensional ChapmanJouguet speed calculated on the basis of chemical equilibrium downstream. When the characteristic transverse dimensions are largerfor example.2. arising from instability of the planar wave and discussed in Section 6. af. these effects are small (typically less than 1 % in the detonation velocity) and masked in the laboratory by multidimensional structures of ChapmanJouguet waves.6. Appropriate analyses that couple the threedimensional behavior to the wave field behind the detonation are difficult to develop with accuracies sufficient to distinguish between equilibrium and frozen sound speeds. for nearlimit or singlespin detonationsaverage sustained detonation velocities below the theoretical values by amounts up to about 20 % are observed. Schematic illustration of the difference between frozen and equilibrium ChapmanJouguet detonations.
[12] and [62][64]).2. The transverse structures that are observed at or near ChapmanJouguet conditions are consequences of this type of instability.3. Often the reactant material is introduced in small amounts so that the reaction occurs nearly isothermally.3. they propagate at the velocity of the shock wave initially established in the shock tube. it is merely necessary to burst a diaphragm between the reservoir and the combustible to establish a strong detonation .204 Detonation Phenomena 6. The evident stability of these waves clearly demonstrates that sufficiently strong planar detonations are stable . If the amount of heat released in the reaction is high enough to generate pressure waves locally that can interact significantly with the leading shock.1. thenin principlethe shock followed by the reaction region together comprise a strong detonation wave. TRANSVERSE STRUCTURES FOR DETONATIONS 6. 6. This is the principle of operation of a shock tube. Shock tubes have often been used to investigate strong (and ChapmanJouguet) detonations in a controlled fashion (see. It is possible instead to initiate the detonation by exploding a strong charge (of a solid explosive. If the overall reaction is exothermic at all. Gasdynamic (inviscid) interactions with boundaries tend to weaken the strong detonation rapidly to a ChapmanJouguet wave. This weakening is enhanced if the end of the tube at which the charge is exploded is open to a lowpressure reservoir. If the shock wave (or “blast” wave) from the explosive is stronger than the shock that precedes the reaction zone in the ZND structure of the ChapmanJouguet wave. then a strong detonation is established . but provision of a sufficiently highpressure reservoir at the initiation end retards the decay. Peculiar detonations that appeared to propagate along helical paths in tubes with round cross sections . Spinning detonations and stability considerations It is curious that the basis for the discovery of transverse structures in detonations existed more than 50 years ago. Shock waves in suitably designed tubes are very planar and decay extremely slowly through viscous interactions. Chemical reactions behind a shock wave produced in a shock tube Up until now in Section 6. for example.2 we have been considering detonations in closed tubes filled with combustibles and ignited weakly at one end. as discussed in the following section. then the potential for development of instability arises. the pressurized reservoir serves to maintain the downstream gas pressure above that of the ChapmanJouguet state until the reflected wave from the end of the reservoir overtakes the detonation. this is termed an overdriven wave. for example) in the tube.6. Shock tubes have become standard tools for studying chemical kinetics at high temperatures [65]. With a sufficiently large reservoir at sufficiently high pressure.
The belief that no ChapmanJouguet detonations are planar rests on the conclusion that all such planar waves are unstable to certain nonplanar disturbances. For a planar ChapmanJouguet wave in an infinitely long tube and in a frame of reference in which the burnt gas is at rest.3. Had it been realized that spin is merely a particular manifestation of the currently wellknown transverse structures. Experiments show that morecomplex modes exist. the circumferential velocity of a traveling tangential acoustic wave at the wall is a known constant times urn. This conclusion is difficult to substantiate in general because of the complexity of the needed . the leading planar shock moves at the velocity uo . The acoustic aspect of the theory suggests that transverse modes other than the spinning (travelingwave) mode may occur. Depending on the experimental conditions. with the tacit understanding that the transverse waves often do not involve rotation.the sound speed of the burnt gas. Transverse Structures for Detonations 205 were observed [66].purely spinning modes are forbidden. The writings on the walls mark the tratechniques [23][25]. The experimental prevalence of multiheaded spin has prompted categorical statements to the effect that all selfsustaining detonations exhibit significant threedimensional structure [25]. spinning detonations were thought to be unusual phenomena associated only with gas mixtures near the limits of detonability (Section 6. the general properties of the writings have been described theoretically [73][76].4).6. jectories of the strongest waves. clarifications of these structures could have been obtained much earlier (through use of the available experimental techniques to study mixtures progressively removed from limits of detonability). and they may also be studied by many types of photographic [68][72]. [67].urn.Therefore.urn). For 30 years after their discovery. but their identification is more difficult. the resulting smokedfoil patterns on walls may be very regular or highly irregular. These “spinning detonations” leave a helical inscription on a smokecovered foil that has been placed on the inside wall of the tube.3. detonations in mixtures far removed from the limit of detonability may be described roughly as multipleheaded spinning detonations. The number of transverse nodes in the tube is observed to increase with an increase in the tube diameter or with an increasing departure from the limit of detonability. the slope of the helical path on the tube wall.The resulting predictions are in reasonable agreement with experiment. then an acoustic theory may be applied to predict the spin frequency. if the markings on the wall occur where the tangential wave crest intersects the leading shock (thereby providing local maxima of pressure and temperature). in tubes with noncircular cross sections. In particular. In this frame of reference. In this respect. the slope of the helical path will be proportional to arn/(vo. If it is accepted that a spinning mode of propagation exists. and other characteristics of spinning detonations [75][76].
( n . In particular. If uo is perturbed by an amount 6 v o . The amplification rate varies with the transverse wavelength and may achieve a maximum at a particular wavelength but apparently does not vanish at either large or small wavelengths. The stability investigations address the timedependent. The analytical work introduces linearizations about this steadystate solution and employs various different approximations for the rates of chemical reactions. 1. = vt. If v is the gas velocity with respect to the shock at the Neumann condition. The results indicate that few. [28].. a model for the steady detonation structure is needed to proceed with a stability analysis. the Neumannstate conditions are expressible by v/vo = p o / p = (Y .. usually in a onedimensional approximation (for example. P = p o M . However. inviscid conservation equations having the ZND steadystate solution.MY + 1)’. may be expressed approximately in terms of the explosion time t. however. then the mechanism for the instability is absent. From normalshock relations for an ideal gas with constant specific heats in the strongshock limit. . assume that properties remain constant at their Neumannspike values for an induction distance l. ~ Y / ( Y+ I). attention may be focused first on a planar detonation subjected to a onedimensional. [87].v T P +r=ufiy .206 Detonation Phenomena stability analysis. real combustible gas mixtures admit stable planar ChapmanJouguet waves. planar ChapmanJouguet detonations are stable for suitable rate functions. then 1. a sufficient number of stability studies have now been performed [77][92] to provide a good qualitative understanding of the instability question. then a mechanism exists for amplification of disturbances. For example. To obtain a rough physical understanding of the mechanism of the instability. where 61. As the simplest structure model.. after which all of the heat of combustion is released instantaneously. the ZND structure usually is unstable even to planar disturbances.MY + I). thereby giving rise to pulsating detonations . The failure to find ChapmanJouguet detonations without transverse structures reflects the inability to encounter real chemical systems with reactionrate properties suitable for stability. these may not represent the most unstable mode and therefore may not be observed experimentally.~ Y ( Y. and [92]). It should be understood that since the stability predictions involve reactionrate properties. and T = TOM. More recent numerical studies have dealt with the nonlinear timedependent equations. given by equation (B57) as 1. 6v E 6T 6T 6P a.. if the rate of heat release decreases monotonically with an increasing extent of reaction behind the shock.1). is perturbed by an amount 61. then 1.6v 6t. if any. timedependent perturbation.. The general idea that emerges is that if the overall rate of heat release increases sufficiently rapidly with increasing temperature or extent of reaction. Since the instability depends on the wave structure.
it may be shown by use of equations (234). the strength of the compression that is generated may be insufficient to strengthen the leading shock appreciably.2 ~( 2(n  I)]. The discussion given here has been qualitative and does not constitute a correct stability analysis. if E/(RoT)+ ci + ( n . and the steady detonation would be anticipated to be stable to planar disturbances./6uo < 0 if E/(RoT)+ ci + ( n . From comparison with more complex results. if E/RoT is so small that equation (23) gives 61. Even for the simplified model adopted. consideration should be given to jump conditions for interactions at the shock and Jm + . is is the Mach number at the Neumann state. and (530) that. where M % The condition for occurrence of the onedimensional instability mechanism is thus seen to be that E/RoT be sufficiently large. 3 . Thus instability also requires that E/RoTm or Ecp/Roqbe sufficiently large for ChapmanJouguet waves. by use of equations (231) and (232). For a sufficiently strong ChapmanJouguet wave. This locally causes an increase in p and T that is convected downstream at velocity u. otherwise. somewhat like a piston motion. In general. Assume that an external perturbation of the leading shock introduces 6vo > 0. Use of the strongshock relations then gives 61. the quantity Bo in equation (B57) having been made proportional to p"Ta' [see equation (B51)]. generation of an expansion wave instead of a compression would be expected by 6vo > 0. It appears physically that for overdriven waves there will be a further requirement that the heat release be sufficiently large. (529). (23) from which it may be seen that 61. thereby strengthening it and increasing uo further. which involves convection in the downstream direction followed by propagation of an acoustic wave in the upstream direction.l)/(y 1). it is found that the specific criterion on E/RoT given above is quite conservative in that detonations are stable for a range of values of E/RoT large enough to satisfy the inequality. in terms of the specific heat c p and the heat release q per unit mass of the mixture. < 0) of the heatrelease front when the perturbation arrives at 1.1) > f./le = ( 6 ~ 0 / ~ 0 ) [1 2E/(R0T).l)/y. in terms of the final temperature it is T = Tm2(y. the temperature at the Neumann state is T = (q/cp)4y(y . The consequent decrease in the ignition time then causes an upstream displacement (61. A characteristic period for this instability mechanism./6uo > 0. which propagates upstream at velocity a .This local increase of heat release generates a compression wave. Transverse Structures for Detonations 207 in which n is the overall order of the reaction (the pressure exponent) and ci measures the temperature exponent of the preexponential factor.6 .u with respect to the leading shock (where a is the sound speed at the Neumann state) and overtakes the shock.1) > i.
208
Detonation Phenomena
reaction fronts, to allowance for outgoing waves beyond these boundaries, and to inclusion of acoustic waves propagating in both directions in between ; equation (43) may be taken as an integral condition for defining l,, with the integral performed in a Lagrangian frame. A proper analysis of the simplified model would be too lengthy to be included here. Moreover, it is better to take into account distributed heat release, as is done in current analyses (for example, [92]). Even the linear stability analyses then become quite complicated. Simplifications in the stability investigations and greater generality in presentation of results might be achieved by formal application of asymptotic analysis for large values of E/RoT with the distributed reaction. Although some consideration has been given to large values of this parameter, the full power of activationenergy asymptotics has not yet been brought to bear on the problem. Accurate analyses for real systems include such factors as multistep chemistry and variable heat capacities and become very complicated. Thus far our stability discussion has been restricted to onedimensional instabilities. Mathematical formulations of multidimensional instability problems are readily written [78][85], [SS], and although the resulting mathematical problems are not appreciably more complicated than those for onedimensional perturbations, it remains difficult to draw general conclusions from analyses. A stability approach in terms of ray trajectories for transverse acoustic waves helps to clarify physically the character of the multidimensional instability [25], [93]. The types of processes discussed abovegeneration of acoustic waves by heat release in hot, highpressure regions and their subsequent propagation (now nonplanar) and interaction with the leading shockare involved in the instability mechanism. Simplified physical models may be developed [23] based on ideas of nonplanar wave propagation. In one such model [77] a disturbance with a transverse wavelength of order 1, is considered within the context of the simplified structure discussed above, and a criterion for instability is proposed on the basis of isentropic propagation of the wirve generated at the reaction front. Although simplified models may aid in physical understanding, their results [including that obtained from equation (23)] are inaccurate and only suggestive. There is need for development of a better understanding of the instability mechanism.
6.3.2. Theories of transverse structures
Instability analyses do not provide good indications of fully developed transverse structures of detonations because these structures correspond to highly nonlinear phenomena. A great deal of nonlinear evolution would occur between onset of instability and attainment of a mature multidimensional detonation structure. Intersections of oblique shocks are known to constitute a central element in transverse structures of detonations [69], [72]. Obliqueshock relations are therefore relevant to the nonplanar structure.
6.3. Transverse Structures for Detonations
209
In a frame of reference fixed with respect to an oblique shock, let $ be the angle between the approach velocity and the shock plane and 6 be the deflection angle of the flow upon passage through the shock, as illustrated in Figure 6.9 for the wave labeled incident shock. For an ideal gas with constant specific heats, it is then readily shown [94] that in the adopted frame,
M: =
2 ( y  1)Mf 2yMf sin2 I )  (y  1)
+
+
2
+ (y  1)MI sin2$ ’
2M: cos2$
(24)
pz2yM: P1
sin’ $  (y  1)
Y + l
and
where the subscripts 1 and 2 identify conditions upstream and downstream from the shock, respectively. If two intersecting shocks are considered with different angles $ in the same uniform approach flow, then it is found to be impossible for both 6 and p 2 , given by equations (25) and (26), to be the same for both shocks; that is, steady doubleshock interactions do not exist. However, it may be shown [95] that steady tripleshock intersections do exist, as illustrated in Figure 6.9. These tripleshock interactions lie at the core of understanding of multidimensional detonation structures.
Approach
velocity
FIGURE 6.9. Schematic diagram of reactive tripleshock interaction.
\
Incident shock
210
Detonation Phenomena
The wave labeled incident shock in Figure 6.9 leads the other waves and differsfrom them in that in a frame fixed with respect to the wave pattern, the component of the approach velocity tangential to the shock is directed toward the point of intersection. It is in this respect that the leading shock is “incident,” or incoming, and the others are outgoing. The incident shock must be sufficiently oblique to have M , > 1. The transmitted, or Machstem, shock that it intersects is more nearly normal, hence stronger, and it generally has a subsonic downstream Mach number ( M , < 1). To satisfy continuity of the pressure and flow direction across the illustrated slipstream, the indicated reflected shock is needed. This reflected shock is the weakest of the three. The triple point is the point at which the three shocks meet. The pattern in Figure 6.9 is conveniently viewed as a limiting case of a problem in which a shock in one medium is incident on another medium [95], the limit being that in which the two media become identical. The two media are separated along a streamline through the triple point and the slipstream, and the Machstem shock is the shock transmitted into the second medium, the analogy with ray optics then being evident. The steady wave pattern described here is well known in nonreactive ) for the incident shock, the gasdynamics. For given values of M I and I configuration is determined more or less uniquely (see [95], for example, for a discussion of uniqueness). In selfsustaining detonations the incident shock may not be sufficiently strong to initiate chemical heat release in the time available, but the strong Machstem shock usually is. In Figure 6.9 a heatrelease front therefore is shown behind the Machstem shock. The streamwise distance from the Machstem shock to the heatrelease front is the induction distance. For the conditions represented by Figure 6.9, the induction distance behind the incident or reflected shock has been taken to be too long for a reaction front to be shown in the figure. The heat release, of course, will modify the flow field and change the conditions from those calculated for nonreactive wave interactions. A relatively simple manner in which tripleshock intersections may combine to produce transverse structures for detonations is illustrated in Figure 6.10. This figure represents wall traces left by a detonation with a twodimensional structure having conditions that are uniform in the direction normal to the figure. The markings on the smoked wall are produced by triple points like the one illustrated in Figure 6.9.* The triple points are observed to propagate along curved paths and to collide periodically. The shockwave pattern is indicated in Figure 6.10 at one point along one of the markings. At this point the incident and reflected shocks extend above the
* This has been demonstrated experimentally beyond doubt by A. K. Oppenheim and coworkers, who employed highspeed photographic techniques to photograph simultaneously in real time the wave pattern and the wall markings as they develop during detonation propagation.
6.3. Transverse Structures for Detonations
21 1
Direction of propagation of detonation
FIGURE 6.10. Sketch of a regular pattern of smokedfoil inscriptions on the wall of a detonation tube with a rectangular cross section.
trajectory of the triple point in the figure, and the Machstem shock and slipstream extend below. The figure indicates that the incident shock is slightly curved and extends to another symmetrically located triple point on an adjacent wall trace. All the shocks in the detonation in fact are curved and are intensified mainly where the triple points collide; they then tend to decay therafter. For the pattern in Figure 6.10, the incident and Machstem shocks both are strong enough to generate heat release at some distance behind them, but the rate of this heat release is considerably less than that occurring at the triplepoint collisions. The markings in Figure 6.10 may be viewed as corresponding to a standing transverse mode in a tube. A readily measurable characteristic of the detonation illustrated in Figure 6.10 is the transverse spacing s shown, often called the cell size, since detonations with transverse structures have been called cellular detonations. Calculation of s is a primary objective of theories of transverse structures. The spacing is regular in Figure 6.10; considerably more complex patterns with both regular and irregular spacings have been observed and explained at least partially on the basis of various types of triplepoint trajectories. Transverse spacings in selfsustaining ChapmanJouguet detonations are observed to decrease with an increasing pressure or with an increasing amount of heat release of the combustible mixture. For spinning detonations the transverse spacings are of the same order of magnitude as the tube diameter. Theories for calculating s are based on simplifications that adopt ray acoustics for analyzing the propagation of trapped transverse waves [25], [93], [96][98] or on attempts to integrate numerically the full twodimensional, timedependent equations of gasdynamics with shock discontinuities and with finite rates of heat release [27], [49], [99], [loo]. Even the former approach usually employs numerical integrations for calculating ray trajectories [97], and in spite of these complexities, the best theoretically predicted values of s differ from the observed values by amounts on the order of a factor of two. In principle, numerical integration of the full equations describes the process correctly, but although these calculations have
212
Detonation Phenomena
produced interesting wave dynamics similar to those observed, the computations are timeconsuming, even with the largest electronic computers [loo]. Moreover, idealizations and simplifications providing analytical results usually lie at the heart of theories that develop impoved understanding. Thus far, analytical theories for transverse structures of detonations have been developed only on the basis of the acoustic approximation for transverse rays [98]. The spacing s varies with the initial pressure, temperature, and chemical composition in the same way that the induction length 1, varies. Thus the ratio s/l, may be taken as constant in a first approximation. The value of s is appreciably larger than 1,; perhaps s/l, % 10 or 30 [1011.Thus s/vo may be on the order of the period of the longitudinal disturbance, identified in the preceding subsection. The fundamental reasons for this are not entirely clear. 6.3.3. Standing detonations Experimentally, the standing detonations, defined at the beginning of this chapter, do not exhibit the transverse structures discussed above [102][106]. For these strong waves the experimental environment is steady and threedimensional (usually axially symmetric). The velocity is subsonic behind the normal shock existing in the flow before the detonation is formed, and the rarefaction discussed in Section 6.2.1 is not present. Therefore, the fact that these are strong detonations is entirely consistent with the boundary conditions. The length of the reaction zone downstream from the normal shock typically is comparable with the characteristic longitudinal length of the wave pattern downstream from the nozzle exit. Therefore, it is currently unclear whether transversewave patterns would develop if the reaction zone were very short in comparison with the external wavepattern dimension. Based on the stability considerations of Section 6.3.1, it might be anticipated that a transverse structure would occur if the wave conditions were near enough to the ChapmanJouguet conditions but that the ZND structure would be observed for sufficiently strongly overdriven waves. These standing detonations, or variants thereof, have been suggested for use in ramjet combustion chambers for supersonic propulsion at high Mach numbers [1071[ 1lo]. 6.3.4. Detonability limits and quenching thickness A number of practically important phenomena associated with detonations pose fundamentally more difficult theoretical problems than those addressed above. These include the initiation of detonations, transmission of a detonation from confined to unconfined spaces, quenching of a detonation, and limits of detonability. Discussion of these phenomena may conveniently be introduced by considering quenching of detonations first.
6.3. Transverse Structures for Detonations
213
It is found experimentally that selfsupported propagation of a detonation through a combustible mixture contained in a tube with rigid walls cannot occur if the tube diameter is less than a critical value, a quenching diameter for the detonation. Similarly, selfsupported propagation does not occur through a combustible mixture contained between two rigid, parallel, infinite plates if the distance between the plates is less than the critical quenching distance for the detonation. Replacement of one of the two plates by an infinite, quiescent, inert gas again results in failure of the detonation if the thickness of the combustible layer is less than the critical quenching thickness for the detonation, the value of which appreciably exceeds the detonation quenching distance of the same mixture [llll[113], For an infinite, planar slab of combustible bounded on both faces by an inert gas, the critical slab thickness for selfsustained detonation is approximately twice the quenching thickness. Bases have been laid for the development of a degree of theoretical understanding of these phenomena of quenching of detonations [ l l l ] , [114][121]. Although the development of a description of detonation quenching necessitates consideration of detonation structure, an approximate physical understanding of the phenomenon may be obtained without addressing the complexities of cellular structures of detonations. The quenching typically occurs near conditions of spin, for example, with at most a small number of cells present, so that the wave that is quenched may be approximated as being planar and noncellular ; experimentally, cellular structures often are not observed during quenching [111][113]. Therefore, we may consider a model like that illustrated in Figure 6.7, with a reaction zone of length 1 following a normal shock. As discussed in Section 6.2.3, for a detonation traveling in a tube the boundarylayer growth has the effect of an area expansion on the quasionedimensional reaction zone and therefore tends to reduce the Mach number, while the heat release tends to increase the Mach number. It seems logical that for the detonation to exist the latter effect must exceed the former at the Neumann state; otherwise behind the leading shock the subsonic Mach number will begin to decrease further, with an associated temperature and reactionrate decrease, and the system will not be likely to be able to recover to reach the ChapmanJouguet state. Thus a quenching criterion is that the areaexpansion effect exceeds the heatrelease effect at the Neumann state. If A is the effective crosssectional area of the quasionedimensional flow and T ois the stagnation temperature, then for an ideal gas with constant specific heats [1223 in variablearea, isentropic flow,
214
Detonation Phenomena
while in constantarea flow with heat addition,
dM M
2yM2
(29)
According to the specified criterion, the detonation exists only if at the Neumann state, d T o / T odivided by the coefficient of dM/M on the lefthand side of equation (29) exceeds dA/A divided by the coefficient of dM/M on the lefthand side of equation (28). For sufficiently strong leading shocks, the value of M at the Neumann state is such that this criterion to reduce approximately to
dTo/To > 2 dA/A.
(30)
An equality may be used in equation (30) to define approximately a general condition for detonation quenching. If for orderofmagnitude purposes, dTo/dx is approximated as Toll,then this equality becomes dAJdx % A/(21), in which 1 may be estimated from experimental results (for example, [123]~1251). In applying the general criterion of equation (30) to detonations confined by walls, available approximations [58], [117] may be employed for relating dA/dx to the growth rate of the displacement thickness 6* of the boundary layer; roughly,
dA/dx
[pW(vo  ~ ) / ( p v ) ] ( 4 A / d6*/dx H)
= (pw/p){8A/CH(y 1>1> dd*ldx,
where pw is the density of the gas at the wall and H is the hydraulic diameter (four times the ratio of the area to the perimeter of the duct, which is the diameter for a circular tube and twice the height for a channel). In using this formula, d6*/dx may be approximated as 6*/1, where 6* is the boundarylayer thickness at the Chapmanjouguet plane; the condition for quenching then becomes very roughly
6*
( y

1)H/8.
(31)
Orderofmagnitude estimates of quenching diameters and of quenching distances may be obtained from equation (31) if boundarylayer theory is the Reynolds employed to estimate 6*namely, 6* FZ l/&where number is Re = pl(vo  v)/p, in which p is the coefficient of viscosity at the Neumann state (typically Re 2 lo5). To apply equation (30) to the estimation of a quenching thickness for a combustible layer adjacent to an inert gas, we must consider the shock pattern in the inert. An approximate model for this shock pattern is illustrated in Figure 6.11. Although not perfectly consistent with the triplepoint discussion related to Figure 6.9, in a first approximation it may be assumed that the detonation generates an oblique shock with deflection
6.3. Transverse Structures for Detonations
215
FIGURE 6.11. Schematic diagram of model for detonation adjacent to a compressible medium [121].
angle 6 in the inert gas and that this deflection angle also occurs at the boundary of the combustible and thereby imposes a divergence in the crosssectional area of the reacting flow. The geometry in Figure 6.1 1 shows that in terms of the thickness h of the detonable layer, dA/dx = (A/h)tan 6. Equation (25) may be used for 6, with the shock angle t,h obtained in terms of the pressure ratio and the upstream Mach number ( M , ) from equation (26). To achieve an approximate lateral pressure balance, the pressure ratio p2/p1 may be taken to be the pressure ratio p + across the undisturbed detonation given by equation (231) for a ChapmanJouguet wave. With MI = v,,/d, where d is the speed of sound in the inert gas, the cited formulas suffice to determine t,h and 6. If ii % a, then a1 ebraic manipulations show 1 1 4 and tan that for sufficiently strong waves, sin t j % l / 2; that is, t,h rz 7 6 rz l/y. Thus the approximate condition for quenching, obtained from the orderofmagnitude version of equation (30), is
J
1
%
(y/2)h.
(32)
Since 6*/1 is small, comparison of equations (31) and (32) shows that the hydraulic diameter for quenching under confinement is smaller than the quenching thickness for the combustible bounded by the compressible medium. The reason for the difference clearly is that the relief provided by the compressible boundary allows d In A/dx behind the leading shock to be larger than it can be for a detonation under confinement. There are procedures for making more accurate estimates of quenching conditions than those outlined here. For example, differences between properties of the inert gas and of the combustible may be taken into account [l 11). It is interesting that very weak walls can often effectively provide
216
Detonation Phenomena
complete confinement [112] because at the short time scales relevant to passage of the reaction zone, the inertia rather than the strength of the wall is relevant. A wall with mass m per unit area subjected to a pressure p moves a distance (p/rn)t2/2 in time t, and confinement will be provided if this is small compared with 1 during the passage time l/uo. Thus (p/m)(l/2ui) 4 1 leads to confinement, a condition that is roughly m po 1, since the pressure behind the shock is p = p o ui for a strong detonation, where p o is the initial density of the combustible. Since solid materials typically have densities times the length on the order of 103p0, the wall thickness need only be of the reaction zone to provide confinement in representative experiments. From these results it may be inferred that properties of bounding gaseous media may significantly influence quenching thicknesses. Quenching of detonations is relevant to the subject of detonability limits. It is known experimentally that combustible gas mixtures diluted sufficiently with inerts, or fueloxidant mixtures that are too rich or too lean, cannot be made to support a selfsustained detonation; these mixtures are said to lie outside the detonability limits. In practice, detonability limits are narrower than the flammability limits to be discussed in Chapter 8, in that there are mixtures capable of supporting deflagrations but not detonations, although theoretically this need not always be true because the phenomena differ. Explanations of detonability limits were sought first solely on the basis of chemicalkinetic mechanisms and rates [126], [127] but later it was realized that quenchingthickness phenomena provide better bases for pursuing explanations [Ill], [116], [117], [120], [128]. According to these later views, detonability limits may be equated to quenching limits for detonations and therefore depend on the size and configuration of the system. Even with these newer explanations of detonability limits, there tend to be mixtures for which these limits are insensitive to dimensions of the system at sufficiently large dimensions. If the induction time t , or the reactionzone length 1 increases greatly with a small amount of dilution for chemicalkinetic reasons, then in equation (31) or (32) the variation of 6* or 1 with dilution will be so great that no matter how large the value of the dimension H or h is within practical bounds, the quenching condition will be achieved approximately at a fixed limit composition, calculable from the equation by use of a representative value for H or h. This type of behavior enables simplified criteria for the nonexistence of selfsustained detonations to be specifiedfor example, that the temperature at the Neumann state of the ChapmanJouguet wave lie below a critical value. If chemicalkinetic mechanisms are such that there exists a critical dilution above which 1 increases strongly with increasing dilution and below which 1 is much less sensitive to dilution, then this critical dilution may represent a detonability limit determined by the chemistry and effectively independent of the configuration. In this sense, chemical kinetics sometimes may control detonability limits, with quenching details playing a very minor role. The relative
+
6.3. Transverse Structures for Detonations
217
importance of the chemical kinetics and of the system configuration to detonability limits for specific combustible mixtures currently is a topic of uncertainty. 6.3.5. The transition from deflagration to detonation Ways to initiate detonations in detonable mixtures may be subdivided into direct initiation and indirect initiation. In direct initiation a sufficient amount of energy is deposited rapidly in the mixture, which results in the immediate emergence of a detonation after an initial period of adjustment, without the intervening establishment of a deflagration. In indirect initiation the ignition source is weaker and produces a deflagration, which subsequently undergoes a transition to detonation. Criteria for direct initiation involve the energy of the igniter, while the process of indirect initiation is independent of the ignition mechanism. Since direct initiation fundamentally is a simpler process, it deserves some comment prior to a discussion of the transition from deflagration to detonation. On the basis of the ZND detonation structure, a simple criterion can be stated for direct initiation. It seems evident that the energy deposited in the combustible medium must be sufficient to maintain a shock at least as strong as the leading shock of a ChapmanJouguet wave for a time period at least as long as the chemical induction time (or perhaps the overall chemical reaction time) at the Neumann state if the detonation structure is to be established directly by the initiator. This idea was developed by Zel'dovich in the 1940s [129][131] and in essence underlies morerecent theories for direct initiation [22], [132][140]. Use may be made of the similarity solutions for strong blast waves [141][143] to show that the energy (for spherical waves, k = 3), the energy per unit length (for cylindrical waves, k = 2), or the energy per unit area (for planar waves, k = 1) that must be deposited instantaneously in order to initiate a detonation directly is approximately
uk
= Ckrk(Y)POUO
k+2 k te,
(33)
where C , = 1, C2 = 27~,C3 = 471, and the function r k ( Y ) , dependent on both k and the ratio of specific heats, typically is roughly of order unity [22]. Numerical integrations of the timedependent conservation equations [144][146] tend to provide a little additional support for equation (33). Recent work [139], [140] has focused on clarifying influences of the time history of energy deposition on the initiation conditions; in addition to a minimum critical energy, clearly there must exist a minimum critical power as well because an initiator with a power level that is too low cannot establish a shock wave as strong as the leading shock of the ChapmanJouguet wave. Indirect initiation involves a transition from deflagration to detonation. This process has been studied extensively for waves propagating in tubes, and many reviews are available [141[28]. The detonation develops
218
Detonation Phenomena
through a complicated sequence of unsteady, nononedimensional events. The earliest studies of detonation development consisted of experimental measurements of detonation induction distances (the distance required for the steady detonation to form in the tube) [147][154]. Observations that the induction distance depends on the gas composition [148], [149] indicate that the chemical reaction rates are of importance, and observations that the induction distance depends on the tube diameter [l50] suggest that wallinteraction effects are of importance. Wall roughness, turbulenceproducing grids, and vents in the tube also affect the induction distance. Refined experimental techniques have given detailed pictures of the mechanism of transition from deflagration to detonation [16][24], [155][158]. The process is known to involve acceleration of the deflagration by the expanding hot gases behind the wave, generation of Mach waves by the flame, coalescence of the Mach waves to form shock waves, development of turbulence ahead of and within the deflagration caused by the increased flow velocities, and a complicated wave interaction in the multiwave, turbulent system to produce the final ChapmanJouguet detonation. Oppenheim [158][161] and others [162], [163] have developed simplified onedimensional, laminar, multiwave models of the transition process that are capable of correlating many of the experimental observations. The simplest picture [159], [160], [163], involving only two wave discontinuities (a shock wave and a flame), predicts that both waves must propagate at velocities greater than the ChapmanJouguet speed before they can coalesce to form the ChapmanJouguet detonation. From a much more detailed graphical analysis [161], it was found to be possible to obtain a wave pattern that was in excellent agreement with the experimental results of [l55]. More recent studies employ moresophisticated computational techniques [164]. The fundamental problems posed by the transition process remain formidable. It is difficult to explain the high turbulent flame speeds needed for coalescence of a deflagration with a shock; localized initiations of explosions by focused pressure waves and a more nearly direct development of a detonation from these energy sources are likely to play an important role that causes deflagrationshock coalescence to be unnecessary. Studies of wave interactions in reacting media [165][168] therefore are relevant to the transition process, while onedimensional timedependent analyses of fast deflagrations propagating at velocities approaching sonic would not appear to be very relevant. The importance of the buildup of pressure waves in promoting transition to detonation indicates that confining a combustible mixture by walls aids in the development of a detonation. Detonations are more difficult to initiate in unconfined combustibles [169]. If a transition from deflagration to detonation is to occur in the open, then unusually large flame accelerations are needed, and these are more difficult to achieve without confinement. Numerical integrations of the conservation equations demonstrate that
6.4. Detonations in Solids, Liquids, and Sprays
219
transition to detonation is rare in the open [170]. There are similarity solutions for flamesupported pressure fields in the absence of confinement [171][177], and these solutions tend to be stable in that they resist coalescence of the flame with the pressure waves to produce a detonation [173], [1773. Therefore, provision of even partial confinement can significantly enhance the tendency for transition to occur. To obtain improved quantification of the dependence of transition on the degree of confinement is a topic of current research [178]. Since the initiation of a detonation is achieved more readily under confinement, there is interest in establishing the conditions that are necessary for a ChapmanJouguet detonation traveling in a tube to be transmitted into an unconfined combustible at an open end of the tube [179][188]. It is found experimentally that the critical tube diameter needed is much greater than the quenching diameter for the detonation; the value is approximately 13 times the cell size sthat is, roughly on the order of 2001, (see Section 6.3.2). A degree of theoretical understanding of this observation has been developed [183]. To ascertain influences on detonation behavior of the extent of confinement in various configurations remains a topic of active research.
6.4. DETONATIONS IN SOLIDS, LIQUIDS, AND SPRAYS
Our attention thus far has been focused on detonations in gaseous combustibles. There are many liquid and solid materials that are capable of experiencing exothermic decomposition in a detonative mode. These include many solid propellants as well as explosives such as trinitrotoluene (TNT). Detonative processes in liqud and solid combustibles are fundamentally like detonative processes in gases in many ways; the foregoing material is relevant. However, there are special complications that arise from the phase changes and especially from the extremely high pressures produced. The pressures achieved are so high (on the order of los atm) that an appropriate equation of state for the system is uncertain, and therefore ChapmanJouguet detonation velocities cannot be calculated by the methods of Chapter 2. In fact, measurement of detonation velocities becomes an important way to obtain information about the equation of state through use of ChapmanJouguet and Hugoniot relations. In addition, fundamentals of reactionrate theory become hazy at these high pressures. Thus it is unclear whether the instability mechanism of gaseous detonations (Section 6.3.1) is generally operative for solid explosives ; experimentally, the reaction zones of high explosives are usually found to be very thin, and local departures from planar waves often seem to be dominated by granular structures of the material. Over larger distances the wave front tends to be curved and appears to propagate in the charge as if it were unconfined because the pressures
220
Detonation Phenomena
achieved not only exceed the yield strengths of wall materials but also typically are large enough in acting on the wall inertia to produce a significant displacement over a reactionzone length. There are critical charge diameters below which detonation cannot occur, much like quenching diameters for detonations in gases, but usually corresponding to unconfined conditions (except for some liquid and dilute explosives). These often are called failure diameters because the detonation fails to propagate in charges of smaller diameters. In solid or liquid explosives, reactive molecules are continually interacting, and limitations on detonation structures associated with molecular mean free paths no longer apply. It becomes entirely possible for significant release of chemical energy to occur within the structure of the leading shock. This fact motivates new approaches to studies of detonation structure on the basis of molecular dynamics [189], [190]. Although the fundamental complexities that are encountered make the problem difficult, further pursuit of these lines of investigation seems desirable. Questions of practical importance concern failure diameters and conditions necessary for initiation of detonation. Theoretical descriptions of failure conditions in terms of chemical induction times have been developed [1911[196] for solid and liquid explosives through reasoning that bears a relationship to that of Section 6.3.4 for gases. Susceptibility to initiation is usually measured by impact tests in standard configurations that effectively expose the material to one or more shock waves and result in sensitivity classifications for the material. With this type of initiation, the process of transition from shock to detonation is relevant. Initiation also may occur as a consequence of a deflagration propagating through the material, and the process of transition from deflagration to detonation then is of interest. Although there have been many investigations of these various transition processes for propellants as well as for high explosives (see, for example, [1971[208]), numerous unknown aspects remain; representative questions concern the homogeneous and heterogeneous rate processes that may occur. Background information concerning detonations in solid and liquid explosives is available in a number of books [ll], [209][215]. Additional information may be found not only in the proceedings of the combustion symposia and of the international colloquia on gasdynamics of explosions and reactive systems but also in the proceedings of the international symposia on detonation, published by the U S . Office of Naval Research. These last symposia have been held periodically, approximately once every 5 years. Sprays of liquid fuels in gaseous oxidizers can also support detonations. Reasonably extensive theoretical [216][223] and experimental [219], [224][230] studies of detonations in sprays have been reported in the literature, and reviews are available [231][234]. Since the liquid fuels must vaporize and mix with the oxidizer before combustion can occur, it is
Rept. Rept. 0. M. FTS1227IA (GDAM AYT46). New York: Interscience (l948). Zel’dovich. The mechanisms of propagation of these detonations are not well understood. Paris (1917). E. No. Acad. Wright Patterson Air Force Base. It may be said in general that for any system capable of experiencing combustion or exothermic chemical reactions.” Tech. the possibility of detonation deserves consideration at least for safety reasons. W. Mag. 8. some degree of theoretical understanding of these detonations has been developed [243][247]. No. adjacent to oxidizing gases [237][242] . Detonations also may occur in solidgas mixtures (for example. Physik 43. Mallard and H. 1261 (1950). Dayton (1949). and reaction zones therefore are much longer than the reaction zones of gaseous detonations. Friedrichs. E. 1st ed. 3. Rept. New York. 90 (1899). No. L. NACA Tech. Supersonic Flow and Shock Waves.D. 204234. 542 (1940). the heterogeneous reactions can occur rapidly enough to support nearly ChapmanJouguet waves in systems of dimensions encountered in practice.References 221 generally more difficult to initiate detonations in sprays than in gas mixtures. Nevertheless. Otherwise. B. von Elbe. J. Among other configurations in which detonations have been observed and studied are thin films of liquid fuels on surfaces . If the fuel has a sufficiently high vapor pressure. No. Doring. 2nd ed. Compt. Dayton (1949). 9.S. (1951). breakup and vaporization of fuel droplets is a dominant aspect of the propagation mechanism. Sci. English translation. 5. Lewis and G. 549 (1942). “Contribution to the Theory of Detonation. Eksp. FTS1226IA (GDAM AYT45). R. Berthelot and P. Paris 93. 7. Flames and Explosions of Gases. 579627. 4. B. WrightPatterson Air Force Base. Academic Press. . Y. Combustion. 2. 6.. Compt. if not for scientific reasons. Courant and K . Memo.. of aluminum powder [235] or coal dust [236]) and have been involved in dust explosions in grain elevators. Fiz. 18 (1881).” Prog. Jouguet. W. REFERENCES 1. 10. Phil. (1961). 0 . Ann. “Theory of Detonation Waves. Doin. No. 47. 347 (1905). J . and the detonation velocity correspondingly is less than the theoretical maximum ChapmanJouguet value. Zhur. 238 (April 1942) O. Acad. Rend. L.R. Doring and G. for example. the existence of detonations in nonvolatile sprays under suitable conditions is now a wellestablished fact. Chapman. D. Mkchanique des Explosifs. When the solid fuel particles are sufficiently fine. Sci. Rept. 421 (1943). 511554. Burkhardt. Vieille. von Neumann.. le Chatelier. 145 (1881). then the detonation may propagate through the gas approximately independently of the liquid droplets. see also “Theory of Combustion and Detonation of Gases.” Tech. Paris93. MathPmatique 6. Rend. Teor. Incomplete combustion and losses to boundaries then tend to occur.
J. Ubbelohde and J. D. 111.. A. 25.” Tech. Kompaneets. 947 (1959). Princeton: Princeton University Press. R. B. F.. and H. “The Basic Theory of Gasdynamic Discontinuities. H. A. G. 19. Brochet. 1130 (1958). 219 (1962). Liquids and Solids. Axford. Curtiss. Baltimore: Mono Book Corp. W. 0.. and H. J . Rend. H. Taylor. 18. Strehlow. WISAF9. Chem. P. W. Rosciszewski. R. 12th Symp. 81 (1968). A. 26. 8th Symp. Pa.. Levitt and D. 33. Lee. K. Shock Waves and Detonations in Gases. A. 36. 111 of High Speed Aerodynamics and Jet Propulsion. J. 36. C. Fundamentals of Combustion. Fluid Mech. A. 3 0 4 0 .” in Fundamentals of Gas Dynamics. T. Oppenheim.222 Detonation Phenomena 10. 37. 41748 I . S. University of Wisconsin. A. 819828. Gasdynamics of Combustion. Oppenheim and R. 1. C & F 12. Bird. (1969). K. and R. N. Soloukhin. K. No. Astronautica Actu 17. “Theory of Detonations. Annual Review ofFluid Mechanics 16. Baltimore: Mono Book Corp. 1979. Introduction to Gasdynamics of Explosions. Kydd. 1966. Hirschfelder and C. H. G. N. New York: Academic Press. I. G. C. Phys. D. K. K. Flammability and Ignition. Hornig. Penner. . 264. 23. Chem. 1145 (1968). New York: SpringerVerlag. 17. Emmons. 28. 16. Lee. 32. Lee. 28. 11. Detonations. 31 (1973). 173 (1959).565 (1969). 1959. Capp.” in Combustion Processes. 797813. Troshin. 271 (1955). Compt. Edwards. 15. B. J. Astrophysical Journal 151. K. R. Soc. 1940 (1954). Wagner. C. B. Madison (1958). 1960. R. K. 424441. AZAA Journal.2243 (1963). and J. Gross and A. 22. 13. and C. W. 31. 0. I1 of High Speed Aerodynamics and Jet Propulsion. Hirschfelder. K. Princeton: Princeton University Press. eds. Curtis. Berkeley: University of California Press. London 253A. 20. Roy. 1958. 1968. 301 (1961). 12. Kistiakowsky and P. H. Rept. 455 (1972). Axford. I. Oppenheim and J. : International Textbook Co. R. 824 (1956). P. 31 1 (1984). Y. J. J . and M. Chem. A. C. Breeze. Oppenheim. New York: Wiley. Soloukhin. Barnett. 38. 24. (1962). Mullins and S. B. Scranton. 6. 34. Fickett and W. F. Manson. Berkeley: University of California Press. 4172.287352. I. ARS Journal 29. Hayes. I. Astronautica Acta 14. J. L. C. Annual Review of Fluid Mechanics 5 . H. Edwards. R. Lee. C. Hirschfelder. Phys. S. 1979. 21. Soloukhin. 1972. 29. J. Zel’dovich and A. H. Annual Review of Physical Chemistry 28. 14. 25. Phjx 22. Newman and W.. 1956. 1965. F. ed. D. Strehlow. Shchelkin and Y. vol. R. Curtiss. Pease. Numerical Modeling of Detonations. Phil. 35. 9th Symp. A. Detonation. B. Mader. Explosions. J . Leyer. R. Trans. (1963). vnl. Williams. Guerraud. I. B. Fay. W. 23. New York: Pergamon Press. H. 1954. H. Theory of Detonation. The Molecular Theory of Gases and Liquids. F. 75 (1977). 30. C. Davis. Oppenheim. 0. I. C. 577612. 27. J . Oppenheim. and A. J. J. J. S. 5 (1967). “Detonation Processes in Gases. Lewis.
706 (1960). Chem. Stever. UMM97. I11 of High Speed Aerodynamics and Jet Propulsion. 850 (1965). G . G. Fluids 4. Phys. 65. V. P. W. J. Kruger. Zhur. 467. editor. W. 43. 1. 40. 414 (1983). White. Trans. 271 (1971). Lifshitz. R. 82. Duff. Phys. Phys. 56. Libby. 60. H. J. Mekh. 1 (1942). J. London 288A. B. Aero. G. G. Campbell and D.” in Fundamentals of Gas Dynamics. 20. 55. J . C S T 2 . P. Sci. 50.” Ph. M. J. 1493 (1952). 69. G . 549 (1960). “ A Shock Tube Investigation of Detonative Combustion. Shock Waves in Chemistry. “Condensation Phenomena in High Speed Flows. Peek and R. “Theory of ShockSupported Arrhenius Reactions and Ignition Delay. H. R. D. W. Fluids 8. M. 67. Fay. Mech. ed. and J. 197 (1929). T. 61. W. 1965. Brinkley. Phys. 52. see also the discussion by B. 42. Kirkwood and W. Woodhead. Fluids 3. W. 20. 27 (1962). Phys. (1953). Thesis. p. J. Erpenbeck. J. Morrison. J. H. J. Emmons. W. 51. Thrap. 41. 22. W. Duff and H. 3010. edited by H. R. 1618 (1954). A. P. 72. S. 586590. Aero. Z . Chem. J . T. 46 (1961). E. 526573. MottSmith. 70. A. Math. 64. D. 1427 (1961). Jr. 71. 3.” Chapter 4 of Nonequilibrium Flows. Young. and M. R. 62. Phys. E. G . V. A. R. 48. Phil. 230A. T. Dove and H.. 63. A. 22. 44. C. W. Fluids 2. Wright.References 223 39. Physik 8. Lewis. (1926). Phys. no. E. Rev.” Rept. and H. A. (1953). G . Phys. Fiz.740 (1957). T. Dobbins. 465 (1961).286 (1951). W. C. Bowen. B. Picone. New York: Marcel Dekker. J. 8th Symp. Kirkwood. 66. 283 (1959). Phys. J. Salsburg. P. Wood and Z. Soc. Tekhn. A. Duff. 412424.674 (1949). M. G . P. 501507. R.. Beckstead. Wegener. Wegener. 363 (1932). 481 (1961). 54. Knight. Mitrofanov. 57. Chem. P. vol. K. Bone and R. Fluids 3. 49. 1572 (1927). vol. A. Fluids 10.. (1981). Wood. Sci. Fluids 4. S. Topchain. Vincenti and C. 1958. Prikl. Fluids 4. Phys. Voitsekhovskii. part 1. (1949). 105. Schott. V. Roy. Adamson. 290 (1967). Knight. J. Chem. (1962). 321 (1922). 1823 (1952). 46. R. Oran. 16411649.. 26. Department of Mechanical Engineering. Fluids 4. T. 464467. S. W. Z. 1915 (1954). R. Boris. New York: Wiley. J. 45. Los Angeles (1 973). 4th Symp. Princeton: Princeton University Press. E. W. Fendell. E. R. 3rd Symp. B. and M. Journal of Fluids Engineering. “ Gasdynamics of Expansion Flows with Condensation and Homogeneous Nucleation of Water Vapor. Jr.. 4th Symp. 68. Ubbelohde. Fay. Wagner. Flanigan. P. Phys. Morduchow and P. Burks. New York: Marcel Dekker. Phys. Becker. Phys. J. M. Jr. SOC. 589599. A. A. Introduction to Physical Gas Dynamics. Phys. Ann Arbor (1952). . R. E.D. Bush and F. 58. Fraser. 18. 18th Symp. 53. Libby. 1969. 22. Reed. University of Southern California. 47. Chem. J. R. Phys. 885 (1951). 16.. E. University of Michigan Engineering Research Institute. H. H. 59. Wood. Agnew. 1981. L. Oswatitsch. Chem.
R. 1780 (1961). Popov. Barthel. J. E. Morrison. J. 517 (1960). Northwestern University. Y. 353 (1976). Shchelkin. New York: Academic Press. Dabora. R. 147 (1946). C & F26. C. Westbrook. Gor. and R. Prikl. K. S. Vzr. A. (1963). 766772. J. A. Fickett. 84. L. 27. E. A. eds. Rend. G. 77. Jet Propulsion 27. 109. J. H. Guirao. 7th Symp. Acad. 1896 (1966). 3 1 (1946). 9th Symp. London: Pergamon Press. L. R. Waterloo. (1962). 74. Paris 222. 451 (1958). P. 6. 316. Fiz. Pukhnachev. 20. M. Y. 98. D. 63 (1959). V. R. AIAA Journal 16. Tekhn. 6 of Progress in Astronautics and Rocketry. M. Fluids 8. and G. 81. K . 604 (1962). Fluid Mech. and R. 600 (1959). Fluids 9. and G. 798 (1963). Adams. Jet Propulsion 28. Akad. Barthel and R. A. M. G. Mekh. A. 1982. 86. Phys. Henderson. Gross. Brehm. K . Dunlap. 79. Fay. Abouseif and T. J. 817830. K. Fluids 7. 18th Symp. B. I.. L. Erpenbeck. Paris 222. Compt. 90.224 Detonation Phenomena N. Chem. 95. 1959. 108. Eksp. Dabora. 115. Melani. M. J. Wilkinson. F. J. A. Fickett and W. Fluids 5. IOth Symp. Erpenbeck. Zel’dovich. J. Nicholls and E. 684 (1964). Wood. Nicholls. and J. J. 105. Aero. Chu. Shchelkin. J. Dokl. A. P. Fluids 17. A. ARS Journal29. Jacobson. C & F45. 8th Symp. 89. A. W. Fiz. 514 (1972). 101.2778 (1963). Teor. Zel’dovich. T. eds. Phys. Fujiwara. vol. 942 (1952).. 87. “Stability of Detonation Waves at Low Pressures. 26. 106. 43 (1972). (1965). Evanston (1956). Fluids 15. Roy. 607 (1966). J. Nicholls. Fiz. 36. Canada: University of Waterloo Press. K. Williams. Nauk SSSR 52. (1959). 93. Morrison. H. B. Strehlow. S. Phys. Sci. Nauk SSSR 149. Chinitz. D. SSSR 136. Morrison.. Fluids 9. Fay. M. 112. Manson. 4. Fujiwara. Dokl. Sci. Astronautica Acta 15. Nicholls. R. W. 107. 63 (1976). R. Lee. J. 102. Schott. A. 488498. B. Gross and W. 111. 99. P. J. Lifshitz. Medvedev. Lee and C. Gor.534 (1957). K . 442453. 73. 80. M. J. A. P. 0. Zaidel and Y. Nicholls. Rhodes Jr. Proceedings of the Gas Djmamics Symposium on Aerothermochemistry. Dokl. J. 82. Frey. H. Barthel. H. R. Zhur. R. Compt. and R. A.” in FuelAir Explosions. Toong. Akad. 9th Symp.67 (1982). Sommers and R. B. Nauk. Landau and E.. Sommers. 110. 0. Fiz. 100.. 903 (1966). Taki and T. 103. L. A. Erpenbeck. 371 (1970). Phys. 75. 85. G. H. 1547 (1974). 59 (1963). 20 (1965). AZAA Journal 16.. J. Rubins and R. Sci. L. Gealer. W. Vzr. 94. Fluid Mechanics. Chiu and J. no. Howe. A. (1981). J. 1293 (1966). I. Fluids 5. Nicholls. “Chemical Kinetics of Gaseous Detonations. S. 92. Penner and F. 39 (1968). Phys. J . J. W. 46 (1946). Zhur. H. Zaidel. J. Phys. 78. Erpenbeck. Il’kaeva and N. E. 241 (1962). Phys. 1142 (1961). 83. 16711681. 189242. Phys. T. J. W. B. 1035 (1978). (1963). Taki and T. CST 14. G. 1962. 73 (1978). 76. B. AZAA Journal 1. 113. 104. Erpenbeck. AZAA Journal 10. 0. 96. 91. Phys. 1192 (1965). Acad. W. Fluids 9. S. J. P. . A. 97. L. 951 11. Phys.” in Detonation and TwoPhase Flow. 88. 644655. Rend. Dabora. 1. V. K . 114. Akad. S. ARS Journal 31. Chernyi and S .
Soc. J. Nauk 87. G. Sichel. S. and T. Sci.795 (1933). : PrenticeHall. P. 13th Symp. Nakagawa. E.221 (1976). 137. R. Phys. Y. Paris 183. C. 130. 152 (1927). Tsuge. (1963). E. B. Dixon. Zhur. S.Fujiwara.References 225 116. Nicholls. G. Eksp. Usp. 148. T. 144. 122. Lee. Proc. 5. TsugC. A. Berthelot and P. 138. Compt. J . 131.. CST 1. N. Trans. Shchelkin. 46 (1961). CST 3. Sandia National Laboratories. F. 745751. 317 (1971). 101 (1882). P. Teor. 284 (1926). 149. Lee and K. N. J . Astronautica Acta 15. 128. Nauk SSSR 52. Knystautas. Acta Astronautica 4. C & F45. Fry and J. 152. Thermodynamics of Fluid Flow. 159 (1950). 118. Zel’dovich. 119. F. and C. Acta Astronautica 1. 516 (1927). (1968). W.. AlmaAka : Nauka. Y. J. Acta Astronautica 5. B. Furukawa. Khim. J.M. CST21. 15th Symp. and N. Tehkn. Boni. Phil. Ramamurthi. Wilson and A. Soc. Tsuge. Hall. Mech. R. Zhur. 150. Sedov. 116A. F. A. 23. L. M. Math. P. 7th Symp. M. A. Paris 176. B. 126. 132. 1362 (1949). A. Phys. Lu. 145. London 186A. Wilson. Knystautas and J. 26. 137.. E. Fujiwara and S.” in DifSerential Equations and Their Applications.157 (1974). Dokl. S. 133. Belles. H. J. Acta Astronautica4. 123. 127. 454460. and J. L. Sci. Matsukawa. H.. 273 (1965). A. 140. Roy. ARSJournal31. Sokolik and K. A. S. 195 (1971). Bach. Shchelkin. W. Acad. Bishimov. 10971110. “The Relationship between Energy and Power for Direct Initiation of HydrogenAir Detonations. Fiz.. Rend. Albuquerque. P. Kistiakowsky and P. 121. C & F 26. I. M. 331 (1976). “Numerical Solution of a Problem of a Strong Point Explosion in a Detonating Gas. Laffitte. (1975). 931952. Acad. G. Vieille. G. 146. A. 134. V. Fiz. Fiz. 1392 (1923). Laffitte. 1744 (1956). 15th Symp. Angew. Abouseif and T. Chapman. 129. Zel’dovich. Roy. (1975). Taylor. E.409 (1977). and J. 141. 237 (1972). G. A. L. 273 (1946). S.. (1982). H. 1951. 39 (1982). Nicholls. Acad. S.. Williams. Zel’dovich. Kogarko. 139. Y. H. Semenov. C & F 27. L. Toong. 345 (1974). 135. Egerton and S. 377 (1970). 3 (1947). Tsuge and T. Kydd. 25. 136. M. (1971). Paris 130. H. Semerjian. 5367. Zeit. Fiz. 1775 (1900). 409 (1977). R.” in Proceedings oj the Second International Conference on the Impact of Hydrogen on Water Reactor Safety. Lee. Soc. Zhur. Japan 33. Akad. B. 17 (1946). Zhur. Compt. M. Wagner. 21.824 (1956). 1153 (1978). K. 316 (1903). M. C. (1959). A. Fiz. 143. Compt. 4. Lee. Dumanois and P. Rend. Roy. H. Dabora. Gates. Annual Review of Fluid Mechanics 3. Sci. 54. R. Knystautas. I. le Chatelier. C . B. Boni and C. London 200A. A. E. 147. 120. A. Fiz. 117. 124. H. I. Englewood Cliffs. Scroggie. 201A. A. N. Korobeinikov. N. Acad. A. Z . . Soc. E. H. Compt. Sci. 9th Symp. W. Khim. Paris 94. 183 (1980).J. Cook. Chem. 403 (1976). I. K. Proc. Rend. Dove. A. C & F27. 65 (1969). R. 153. 4352. Y. K. Guirao. and J. and H. Dabora. H. 142. G. Patch. 1459 (1934). 125. B. Rend. Wilson. Boni. 151. London 114A. W.
Oppenheim. Bowen. J . “Effects of Cellular Structure on the Behavior of Gaseous Detonation Waves under Transient Conditions. “Diffraction of a Planar Detonation in Various FuelOxygen Mixtures at an Area Change. Kuhl. Strehlow. Cook. Edwards. Knystautas. and H. G. M. Pittsburgh : The Combustion Institute. 19. P. Oppenheim. . N. 199 (1976). 75 of Progress in Astronautics and Aeronautics. I. Schneyer. Manson. J. 379405. 156. vol. Mitrofanov and R.429 (1982). A. 159. Given. E. 7th Symp. (1979).Thomas. A I A A Journal 18. A. 399 (1958). Kurylo. and R. I. G . C. I. J. Lee. E. Soloukhin. W. 17th Symp. M. H. 543 (1978). M. E. J . Chiu. Oppenheim. K. 168. Acta Astronautica 5 . Moen.. 12351245. Blythe. Appl. Martin. 172. Schmidt. 182. H. G. Oran and J. 174. 155. Oppenheim. P. A. (1959).VieilleMallardle Chatelier. W. 173. Knystautas. 1981. 178. Oppenheim and R. and J. 26. “Transition to Detonation in an Unconfined Turbulent Medium. 14th Symp. 341 (1976).” Colloque Intcmational Berthelot. J. Nuuk S S S R 159. H. G. 856865. J. J. 1 (198 1). 11891200. vol. 95. 63 (1952). L. F. Akad. J . Kamel. Clarke. 166. A. S. Nettleton. 179. A. 177. H. 81. A. I. 89. Mech. H. G. Fluid Mech. Soloukhin and K. 169. 1978. Richeson. H. J . D.” in Gasdynamics of Detonations and Explosions. P. 837850. H. (1979).226 Detonation Phenomena 154. Chapman. G. K . R. Soloukhin. 167. Steinicke. “Transformation from Deflagration to Detonation. 163. F. 193 (1977). 7th Symp. K . (1959). Edwards. 658666. Phys. 7. CST 12. H. Soloukhin. K. Fluid Mech. L. 171. Chem. 127. J. New York: American Institute of Aeronautics and Astronautics. H. (1979). B. 82. (1953). and W. 4th Symp. I. Y.” in Fundamentals of Gas Dynamics. L. Matsui and J . 1981. Prog. Kennedy. New York: American Institute of Aeronautics and Astronautics. 94. F. D. Neubert. (1973). Boni. K. 1958. C & F 13. A. 14th Symp. 160. Oppenheim.1003 (1964). A. and G . R. 184. Abouseif and T. J. 158. I. eds. 176. “Similarity Analysis of FlameDriven Blast Waves with Real Equations of State. 778 (1961). 28 of Progress in Astronautics and Aeronautics. and R. 162. 43. R. L. Wagner. Phys. J. Toong. Kuhl. ed. 295 (1969). Urtiew and C. Nettleton. K. A. P. First Specialists Mwting (International) of the Combustion Institute. 370384. W. 164. 175. (1973). F. 909916. 17th Symp. A. Stern. Martin and D. V. 12011215. and M. Laderman and A. 0. H. A. Fluids 1. A. Appl. Dokl. E. Kistiakowsky. J. 180. 11 11 (1955). F. N. W. vol. (1953). K. Thomas. R. H.257 (1977). A. Phys. Fluid Mech. 1981. 471480. Tarver. Ragland. 645656.” in Turbulent Cornbustion. 0. Emmons. and A. 183. eds. M. 4th Symp. J. 491496. 79 (1979). 111 of High Speed Aerodynamics and Jet Propulsion. C & F 27. Manson. A. 165. R. J. Guirao. M. and U. J . Energy and Combust. Evans. and R.. Williams. Oppenheim. J. I. Fluids 4. 12691280. Lee. A. Bach. A. 75 of Progress in Astronautics and Aeronautics. Bowen. 103. A. 181. F. 170. 195 (1979). A. R. and G. vol. Boris. 1 (1981). A. Fluid Mech. 17th Symp. Clarke. Williams. 302 (1980). A. A. 161. M. ed. K. Sci. Lee. W. G . 0. 343 (1978). Taylor and R. Lee. V. A. and A. Ind. 2794 (1955). M.” in Gasdynamics of Detonations and Explosions. F. White. 341357. New York: American Institute of Aeronautics and Astronautics. J. Oppenheim. Fluid Mech. Princeton: Princeton University Press. K . Eng. 157.. Dwyer. S. Tankin..
Moen. New York: Academic Press. Lee. I. Acta Astronautica 5. 36. New York: American Institute of Aeronautics and Astronautics. M. Jr. eds. C. 498 (1961). ‘‘ SingleShock Curve Buildup and a Hydrodynamic Criterion for Initiation of Detonation. Phys. A. 189. D. J. 0. J. Fluids 4. R. 51 1 (1961). F. J. Lee. Fuhre. F. 208. vol. Thibault. 190. B. Campbell. (1965). 188. C. Wood and J. A. M. Cowperthwaite. eds. J. 1981. 207. J. Boyer and R.69 (1949). Trofimov. Bowen. London 189A. A. 517529. Manson. Soc.765 (1959). R. 198. 206. D. (1979). A. A. Travis. 1962. Initiation and Growth ofExplosions in Liquids and Solids. 75 of Progress in Astronautics and Aeronautics. 196. Jacobs. W. S. H. Chem. 6 3 (1982). E. A. and R. S. P. Grandey. A. vol. 499516. 1981. Manson. H. K. S. A. H. J . H. I. 203. N. Pack. 10th Symp. Knystautas. R. Moen. Dremin and V. H. C. Knystautas. I. 1920 (1954). M. 16151 622. New York: American Institute of Aeronautics. Kirkwood. Erying. Liddiard. F. R. vol. R. Manson. Karo. Soloukhin. Davis. Murray. W. Ramsay. R. F. 75 of Progress in Astronautics and Aeronautics.. M. Dremin and V. S. R. J. New York: Academic Press. 210. 18th Symp. N. Bowen. 193 (1962). Yoffee. 22. 635644. Hubbard and M. Donato. 197. I. 211. P. Taylor. 205.. R. Oxford : Oxford University Press. 30. Bowden. Duffey. Presles and C. 253268. 192. and B. Hjertager. Kezerle. (1979). H. Travis.” in Gasdynamics of Detonations and Explosions. Krier and J. Rinnan.” in Detonation and TwoPhase Flow. Dick. M. Fluids4. 16231629. 415 (1947). H. and J. W. (1963). J. Oppenheim. 599606. R. Phys. eds. K. Jones. 269281. Roy. “Theoretical Treatment of the Detonation Behavior of Composite Solid Propellants. N. H. Chan. H. Johnson. J. Cook. Brochet. Proc. J . M. 1958. Bowden and A. 19th Symp. Lee. J. Oppenheim. A. Powell. 0. 9th Symp. D. H. B. Oppenheim. A. Knystautas. 209. A. New York: American Institute of Aeronautics and Astronautics. B. W. 186. G. Soloukhin. and J. 820836. (1983). S. Walker. Phys. Warner. A. “The Effect of the ShockWave Front on the Origin of Reaction. 194. 1981. W. Chem. Bernecker and D. Liu. 1952. N. vol. 18th Symp. Cambridge : Cambridge University Press. 7th Symp. 200. A. J. R. A. J. Phys. J. (1963). 202. H. Y. Campbell. Y. (1981). 1041 (1978). Guirao.. 75 of Progress in Astronautics and Aeronautics.” in Gasdynamics of’ Detonations and Explosions. D. 536544.. 2334. eds. . Chem. F. and R. A. and J. C. Detonation in Condensed Explosives. (1981). Parlin. Gey. 191. N. Price. Yoffee. Revs. and C. “Induction Delay and Detonation Failure Diameter of Nitromethane Mixtures. G. 19th Symp. 5562. M. 9th Symp.. 1952. B. Klimenko. N. Hardy. R. T. (1983). Lee. 9th Symp. M. and K. 45. 199. P. Bowen. A. K. Guirao. 17th Symp. and R. Williams. and R. Knystautas.References 227 185. (1963). H. Phys. Evans. 187. 195.. 7598. Drimmer. 839843. Fast Reactions in Solids. I. A. and F. P. Davis. R. and J. P. W. H. C & F 4 8 . Soloukhin.” in Gasdynamics o f Detonations and Explosions. R. 193. W. 6 of Progress in Astronautics and Rocketry. Penner and F. Bowden and A. and W. Bjerketvedt. K. Appl. 17th Symp. W. 201. E. J. (1959). 282295. 204.
E. 209 (1972). Lee and C. Troshin. H. 214. F. A. S. Gor. K. Mitrofanov. Rao. 246. 226. Termodynamiki i Spalaia 7. Fluids 4 . Y. Phys. Astronautica Acta 12. J.” in FuelAir Explosions. Canada: University of Waterloo Press. and J. W. 39. New York: Reinhold. 224. Akad. Sichel. A. A. A. (1971). 220. A. P. Detonation Waves in Condensed Media. Acta Astronautica 1. A. Ragland and C. A. V. R. Acta Astronautica 6 . AZAA Journal 6. W. Rend. S. (1981). Fry. and D. 245. F. E. Nauk SSSR 164. 238. A. 109 (1967). A. A. Sichel. Waterloo. S. M. F. M. Ragland and J. and J. A. Bowen. 244. Dabora. Dokl. 218. K. Nicholls. and J . 241. Cramer. 222. 232. Lyubimov. B. 225. 229. K. 159 (1971). Guirao. W. E. 221. M. Gel’fand. Penner and F. Fujitsuna. Savrov. J. Ragland. Paris 234. and Y. Berger and J. (1973). 16611670. A. 218 (1979). F. eds. W. Explosions in Air. W. M. Borisov and B. V. AZAA Journal 7. K. eds. 242. Arch. 18th Symp. K. 1982. 245264. 265302. 13th Symp. A. K. Baker. Dabora. (1962). Nicholls. 785 (1979). 385 (1974). M. V. S. E. 859 (1969). Nauk SSSR 160. 512 (1952). 1926. 1962. “The Detonation of Sprays: Recent Results. H. J.” in FuelAir Explosions. Pierce and J. A. Palmer and P. Sichel. Loflin. 53 (1972). 234. Dabora. 6 of Progress in Astronautics and Rocketry. D. Burcat. and J . 247. T. Y. 237. Strauss. 235. 230. E. A. V. S. 14th Symp. G. Nicholls. K. Z. Guirao. F. T. F. 233. Nicholls. 240. C & F 18. Pinaev. F. A. and A. R. E. Schvedov.” in Detonation and TwoPhase Flow. 11411149. E. eds. 273 (1976). Akad. Steffes. “Detonations in Dilute Sprays. A. Gabrijel and J. A. 1982. 8th Symp. Akad. Weinberger. Fiz. 853 (1965). Acta Astronautica 1. 1973. (1963). J. Webber. Compt.228 Detonation Phenomena 212. W. J. 109 (1977). K. Williams. and J. K. Komov and Y . Nicholls. Nicholls. I051 (1979). K.. Tsuge. 228. Nicholls. W. K. 1753 (1968). V. Glass. M . Lee and C. (1973). Tonkin. 1970. New York: Academic Press. 12651275. V. Acta Astronautica 6. 991 14. Dabora. 269 (1979). S. (1971). W. 231. V. 15991606. 217. C. R. (1981). Nauk SSSR 175. Williams. 11291 140. Troshin. Nicholls. J. A. Williams. V. M. Rao. Phys. 1434 (1961). 9th Symp. 1958. Loison. S. Fluids 11. Garcia. C S T 4 . Paris: Dunod. Acta Astronautica 6. 223. Ragland. Komov. Sichel. 236. C & F 17. Trofimov... A. 239. and K. 1962. Nicholls. Gubin and M. Acta Astronautica 5. A. and J. 243. Kogarko. E. BarOr. Dokl. Borisov. Sci. Drernin. Waterloo. “Fundamental Mechanisms of Liquid Spray Detonations. and S . Cook. W. Sichel. S. R. 1. CST 17. S . K . 215. E. 13th Symp. S. 361 (1974). 31 (1965). vol. ACS Monograph No. M. Austin: University of Texas Press. S. S. 213. Canada: University of Waterloo Press. Zhdan. Dokl. 227. 482487. 12th Symp. Acad.. Moscow: Nauka. Ragland. N. 11311140. Viard. Ragland. K. Sichel. S. (1969). 2377 (1968). Dabora. A. K. 14th Symp. C. K.9 (1966). Fujitsuna and S . Dabora and L. 216. 18th Symp. and T. M. Nicholls. Gordeev. Physique des Explosifs Solides. Eidelman and A. The Science o f High Explosives. H. 219. K. 12771284. N. Vzr. R. . 125 (1965).
Therefore. but singlebase and triplebase propellants also have been used. In heterogeneous propellants the solid is heterogeneous. There are many different composite propellants. the solid may consist of a single constituent such as nitrocellulose. solidpropellant formulations must be sufficiently metastable to resist transition to detonation. usually consisting of a continuous fuel matrix or binder containing oxidizer particles. a typical example is a polybutadiene acrylicacid copolymer as the fuel with crystals of ammonium perchlorate as oxidizer.CHAPTER 7 Combustion of Solid Propellants Solid propellants are solid materials that are capable of experiencing exothermic reactions without the addition of any other reactants. in principle. Solid propellants conveniently are subdivided into two categories. homogeneous and heterogeneous propellants. the fuel and oxidizer being intimately mixed at the molecular level. It is intended that the propellants will deflagrate. In homogeneous propellants the solid is homogeneous. There are heterogeneous propellants in which one or both of the constituents could serve as 229 . sensitive explosives generally are not employed as solid propellants. thus forming a composite material that is termed a composite propellant. because a detonation would be damaging. so named because they are composed mostly of two exothermic compounds (usually nitrocellulose and nitroglycerin). They are employed mainly as propellants for rocket vehicles and as propelling charges for projectiles in guns. but in practice it is a mixture of constituents. The most common homogeneous propellants are doublebase propellants.
Although processes of this type are of fundamental interest and also of applied interest in materials processing. in which oxidizer crystals are added to the homogeneous propellant. different approaches to the analyses of their combustion processes will be indicated. various other heterogeneous additives often are used. the simplest example of a solid propellant would be a solid that gasifies in an exothermic rate process at a planar solidgas interface. 7. DESCRIPTION OF STEADY DEFLAGRATION OF A HOMOGENEOUS SOLID The simplest examples of deflagrating solids are those in which all reactants and products are in the solid state (for example. When solid propellants deflagrate. there are compositemodified doublebase propellants.230 Combustion of Solid Propellants a homogeneous propellant (for example. No less than three books [6][S] have been concerned solely with solidpropellant topics. Therefore. most of the attention will be devoted to homogeneous propellants . The present chapter is concerned with theories of deflagration of solid propellants.7. and even the compositepropellant constituent ammonium perchlorate can deflagrate alone at sufficiently high ambient pressures). mixtures of molybdenum with boron) [9]. where various elements of their combustion mechanisms will be reviewed. and the surface is observed to regress normal to itself at a definite rate. the burning rate is of importance to the design of systems such as rocket motors that employ solid propellants. Thus both doublebase and composite propellants may be metalized. propellant composition. A simple model mechanism for propellant deflagration involves an endothermic . In addition. the last [S] being a 700page text on the subject. and so on. Heterogeneous propellants will be considered briefly in Section 7. There are a number of useful background references on solid propellants and their combustion. Gasification steps more commonly tend to be endothermic. These materials may deflagrate in a process in which a single exothermic reaction front propagates through the solid. gaseous products of combustion are needed in propulsion applications . metal particles (such as aluminum) are added and these burn in the gaseous reaction products of the propellant constituents and enhance performance. Most books on rocket propulsion devote at least one chapter to solid propellants [1][5]. they burn from the surface inward. This subject is used here more as a means for introducing various fundamental concepts than as an accurate derivation of burning rates of real solid propellants. An objective of theories of solidpropellant deflagration is to ascertain the dependence of this linear regression rate or burning rate on the ambient pressure. Readers may consult this literature for more detailed information. In addition to being of fundamental interest.1. temperature.
.3. and the combustion of doublebase propellants then usually involves reactions occurring in the condensed phase. for example. For example. in the burning of doublebase propellants with sufficiently high contents of nitroglycerin. The foam zone may be observed. such as RDX (1. 18th Symp. Instead.5triazine. Although analyses of the reacting dispersed phase have been developed [13][ 151. However. Description of Steady Deflagration of a Homogeneous Solid 23 1 sublimation of the solid at an interface. and in the gas phase. moreover. A twophase region has been called a dispersed phase. (1981). 7777851. significant exothermic reactions were inferred to occur in the liquid phase [ll]. and a foam reaction zone may exist. Nitrocellulose burns with a subsurface reaction zone in which the solid softens appreciably. The problem becomes especially difficult if it is necessary to account for twophase flow that occurs in foam or fizz zones.' these theories will not be addressed in detail here. (1982).5trinitrohexahydro1. analysis of the deflagration of homogeneous solid propellants is a challenging field of research. a simplified model will be considered in which there is a sharp planar boundary between condensed and gaseous phases. gas producing reaction. chemical formula (N0. A model of this type can be useful for limited purposes. by a luminous gasphase flame. At low nitroglycerin fractions. bubbles may develop within it in the course of the exothermic. CH. for most doublebase formulations.N . 19th Symp. Because of these complex combustion mechanisms. one member of a class of materials. if the neglected aspects influence the phenomenon of interest to a sufficiently small extent. the burning rate appears * There have been some recent clarifications of combustion mechanisms of propellants containing these materials [see N. 187194. and then. followed by a fizz zone. ethyl nitrateburn by a mechanism that differs from this only in that vaporization replaces the sublimation step [lo]. Kubota.).3. a gasphase region in which condensed particles that leave the surface of the homogeneous solid experience exothermic gasification. A number of liquid monopropellantsfor instance. after a dark zone.* Unless the liquid layer on the surface of the solid is very thin. called nitramines. the foam zone typically is less prominent than a fizz reaction zone. by further study it was found that this solid melts and then vaporizes during burning instead of subliming directly. in the dispersed phase. followed by an exothermic gasphase reaction. including materials currently subjected to continuing investigation for use in propellants. such as for calculating burning rates.7. Burning rates may be controlled primarily by chemicalreaction processes occurring in one phase. At one time ammonium perchlorate was thought to deflagrate by the simple model mechanism just defined. .1. containing both fuel and oxidizing species in the same molecule) [12]. A number of energetic crystals behave in this same general way.
A simplified model that may provide a reasonable representation of the steadystate. the velocity at which the interface between the gaseous and condensed phases moves relative to the stationary solid).* Thus simplified models may be relevant to the burning of real propellants. * See also F.1. the condensed phase gasifies either in a simple (endothermic) sublimation process or in a more complex (endothermic or exothermic) chemical reaction. As in most steady. Thus.h a s e Condensedphase reaction zone * Gaseousphase reaction Gaseous products Nonreacting solid 0 X FIGURE 7. “Deflagration Mechanisms of DoubleBase Propellants at Moderate and High Pressures.GO. plane. Williams.1 Model of a burning solid. and the gas extends to x = + co. depending on which processes are dominant.232 Combustion of Solid Propellants . one would like to be able to predict the linear regression rate of the solid (that is. the slowest processes may be expected to govern the overall burning rate.” unpublished (1976). A. Since the processes described above occur in series. At a higher temperature. x is the coordinate normal to the surface. the interface is located at x = 0. We shall first discuss interface processes and then consider spatially distributed processes in different phases. to depend mainly on reactions occurring in the condensed phase [16][18]. in Figure 7. A number of different burningvelocity formulas may therefore be obtained.1. it is convenient here to assume that the solid and the gas are infinite in extent and to adopt a coordinate system in which the flow is steady. onedimensional burning of a homogeneous solid is illustrated schematically in Figure 7. planar flow problems. Given the fundamental physical and chemical parameters describing each of these processes.We are interested in computing the mass burning rate m (g/cm2 .Condensed phase ? J G a s e p . . the solid extends to x = . Heat conducted into the solid from the reaction region raises its temperature to a point at which a phase transition or some other (endothermic or exothermic) chemical (equilibrium or rate) process begins to occur. s) at which the solid must move in the +x direction so that the interface will remain at x = 0. The gases evolved react exothermically in the gas phase to produce the final (gaseous) reaction products.
(2) where W. y + is related to the actual number of moles of species 1 per unit area. Transitionstate theory (Appendix B) provides information concerning the rate of this gasification process. yl. However.. q 1 is the partition function per unit surface area for species 1 (with the groundstate energy factor removed).3./T.y l(kOT./h)(q# /ql)eE’RoTl./T). for illustrative purposes. respectively. Ro is the universal gas constant. We shall focus our attention on the process occurring at the interface x = 0. each of which may gasify by processes of varying complexity.4. Hence. APPLICATIONS OF TRANSITIONSTATE THEORY Let us consider the case in which the burning rate is determined by the rate at which molecules from the condensed phase enter the gas. m = ~IYlw. 7. can be expressed as a ratio of partition functions per unit surface area through reasoning similar to that leading to equation (A24) for K. the mass burning rate becomes m = W. Solid propellants generally contain a number of different chemical species. we shall study the simplest case of a onecomponent condensed phase that sublimes in a onestep. then the number of moles per unit area per second leaving the surface is Y + / T . unimolecular process : %(s) %(g). and 4 + is the same kind of partition function for the activated complex without the degree of freedom corresponding to translational motion in the direction of the reaction coordinate.2.Y. (1) where ‘9Xl is the chemical symbol for the subliming species. through the equilibrium condition y + / y l = K:. we then find that + m = W. If y+ denotes the number of moles of activated complexes (for the sublimation process) per unit surface area. (3) The equilibrium constant K . E is the activation energy for the sublimation process. Applications of TransitionState Theory 233 7. Hence. is the interface temperature (at x = 0).2. where K: is the equilibrium constant for surface concentrations for the process in which the complex is formed.7. is the molecular weight of species 1 (the species of which the solid is composed). for example. Proceeding as in Section B. h is Planck’s constant. Related analyses of more realistic systems may be found in [19] and [20]. (4) where ko is Boltzmann’s constant. Since the same number of partition functions appear in the numerator and in the denominator of equation (4) . and s and g denote solid and gas phases. From the viewpoint of the transitionstate theory. and T denotes the average time for an activated complex to sublime.
234
Combustion of Solid Propellants
(because of the assumption of a unimolecular process), the “per unit surface area” parts of q+ and q1 cancel and therefore can be neglected. The surface concentration y1 appearing in equation (4) is easily related to the density and molecular weight of the solid. Geometrically, $ I 2 is seen to be the number of moles per unit volume in the solid. Hence,
where p, is the specific gravity of the solid. In view of equation (5), equation (4) becomes
Equation (6) is the required massburningrate formula, which (under our present assumptions) determines m in terms of the interface temperature T and the properties of both the solid and the activated complex. In order to use equation ( 6 ) ,we must measure, estimate, or compute the value of T, for the burning solid. The surface temperature usually depends on the amount and rate of heat release in the gas (and perhaps in the solid) as well as on the heat transfer rates in the gas and in the solid. Therefore, equation (6) should be supplemented by an analysis of the thermal processes occurring in the system. An example of such an analysis will be considered in Section 7.5. In the remaining part of this section we shall discuss a more basic difficulty encountered in attempting to use equation (6). The structures of the solid and of the activated complex affect m through the parameters E and q2/q1 in equation (6). Generally, so many different possibilities exist for the structure of the complex that these parameters cannot be calculated from purely theoretical considerations. Semiempirical procedures must therefore be used to determine the burning rate from equation ( 6 ) even when is known. In order to amplify these assertions, let us consider the ratio q + / q l . The partition function for species 1 on the surface of the solid differs considerably from that for species 1 in the gas. The translational degrees of freedom of the gas molecule generally become vibrational degrees of freedom in the solid, and some (or all) of the rotational modes in the gas may also become vibrational modes (librations) for the solid. Furthermore, the activated complex on the surface of the solid may differ in more than one way from nonactivated surface molecules ; for example, rotational motion may be less inhibited for the complex (leading to fewer librations and a greater number of fully excited rotational modes), or vibrational modes corresponding to translational motion of the center of mass may assume different frequencies in the complex. A very simple case is that in which q+ differs from q1 principally in having one less vibrational factor, as indicated in Section B.3.6 for the unimolecular gaseous reaction; that is,
7.3. Approach to Interfacial Equilibrium
235
where q, denotes this vibrational factor. However, equation (7) represents only one of many possibilities. In view of these uncertainties, the actual structure of the complex is usually best determined by using all available data concerning the structure of the sdid (to find q l ) in conjunction with measured gasification rates, in order to obtain q + through equation (6). These rate measurements may be performed for the burning solid, in which case equation (6) merely provides a formula for correlating data, or for a gasification process in which the gaseous combustion is supressed (for example, sublimation at pressures below the limit of flammability). With the latter strategy, equation (6) is first used in order to obtain q + from experiments that do not involve combustion, and afterward equation (6) also determines m for the burning solid. Actually, a measurement of m gives q+ C E I R T T , not q t alone [see equation (6)]; since E is generally also unknown, the measurements must be performed at different interface temperatures in order to determine E (from the primary dependence of m) and q + separately. The program outlined above has not been attempted for conventional solid propellants, primarily because propellant compositions are so complex that a simple formula of the type given in equation (6) is not expected to be valid. However, a number of propellant ingredients have been studied extensively from our present viewpoint. Experiments on the rate of sublimation of a solid in a vacuum [21][26] [for which equation (6) is least questionable, since gasphase processes cannot affect the measured sublimation rates] and experiments in which the solid is pressed against a heated plate at atmospheric pressure [27][31] have been performed. Hypotheses concerning the mechanism of the sublimation process and the structure of the complex have been invented in order to correlate the measured m values The heatedplate, or linearpyrolysis, experiments [26][28], [30][33]. demand special consideration of gasphase processes in their analysis [S], [30],[33][38] as a penalty that must be paid for their abilities to approach the higher surface temperatures achieved in solidpropellant deflagration. Alternative methods for reaching high temperatures, such as laser heating, also need gasphase analyses for their full interpretation [39], [40]. The presence of the gaseous phase can cause a departure from equation ( 6 ) as a consequence of the possibility of attainment of interphase equilibrium. 7.3 APPROACH TO INTERFACIAL EQUILIBRIUM
At the surface of a burning solid, the reverse process
may be fast enough to alter appreciably the formula for the burning rate given in equation (6). Equation (8) represents the process in which the gaseous species evolved from the solid returns to the surface. This process may be described from the viewpoint of molecular gas kinetics, which is presented
236
Combustion of Solid Propellants
in Section B.4.3. By reasoning analogous to that leading to equation (B80), we find that the mass per unit area per second returning to the solid is WIPla,where 8, [which is given by equation (B78) when the vapor is an ideal gas] is the number of moles of species 1 (in the gas) per unit area per second striking the surface, and a is a (dimensionless) surface accommodation coefficient or evaporation coefficient for molecules of species 1. Thus, equation (6) should be replaced by the expression
= ~ 1 1 213 3
1
P, ( k 7;/h)(q#/ql)eE'RoT1  W l P l E .
o
(9)
The factor P, introduces into equation (9) an explicit dependence of m on the concentration c l , i of species 1 in the gas adjacent to the interface [see equation (B78)]. Except for this difference, equation (9) contains the same kinds of parameters as does equation (6), since the coefficient a can be analyzed from the viewpoint of transitionstate theory. Although a may depend in general on and the pressure and composition of the gas at the interface, a reasonable hypothesis, which enables us to express a in terms of kinetic parameters already introduced and thermodynamic properties of species 1, is that a is independent of the pressure and composition of the gas [ a = a(TJ3. Under this condition, at constant the last term in equation (9) is proportional to the concentration c l,i and the first term on the righthand side of equation (9) is independent of cl, i.Therefore, by increasing the concentration (or partial pressure) of species 1 in the gas, the surface equilibrium condition for species 1m = 0can be reached. If pl,,(TJ denotes the equilibrium partial pressure of species 1 at temperature T , then when m = 0, equation (9) reduces to
W,[pl, e( TJ/Ro T J J a = W :'3p:i3(k0 TJh)(q# / q l)e E'RoTl,
where use has been made of equation (B78) and of the idealgas equation for species 1 (cl = pl/RoT).This relation shows that a(7J is given by
a
=
[ p , , e( 331 ' J ~ ( p , / W 1 ) 2 ~ 3 ( k 0 ~ / h /)q l() eq  E,' R O T 1
(10)
under our present assumptions. In view of equation (B78) and the idealgas law, equation (10) implies that equation (9) can be written in the alternative form
= w1/3 2 / 3
1
P, ( k T/h)(q,/q,)e
0
E/RoT,
C1  ~ 1 , i / ~ l , e ( T i > l ,
(11)
where pl, is the partial pressure of species 1 in the gas at x = 0. The new parameter p1,,(TJ, appearing in equations (10) and ( l l ) , is the equilibrium vapor pressure, which is given by equation (A23) and is where L is the heat of sublimation. approximately proportional to eKLIRoTZ, If pl,e(ZJ cannot be calculated from known thermodynamic properties or measured by conventional techniques, its value may sometimes be deduced by measuring m with the solid pressed against a heated plate [34]. After pl,,(TJ has been obtained, to compute the burning rate of the solid from
7.3. Approach to Interfacial Equilibrium
237
equation (1 1) entails considering gasphase reactions in order to determine p l , i . Equation (11) is, therefore, somewhat more difficult to use than is equation (6). If pl,i/pl,e(TJ < 1, then equation (11) merely provides a small correction to equation (6). However, p l , i / p l ,,(Ti) may approach unity for some burning solids. In such cases, the gasification process, equation (l), no longer governs the overall burning rate ; some other process becomes rate controlling. In extreme cases it is sometimes reasonable and convenient to replace equation (1 1) by the simpler requirement of surface equilibrium, namely,
P1,i
= Pi,e(Th
(12)
in computing the mass burning rate from an analysis of another process, such as finiterate chemistry in the condensed or gaseous phase. If surface equilibrium prevails, then it is relatively straightforward to generalize the interface condition to chemical processes that are more complex than equations (1) and (8). This of interest, since propellant materials often experience processes of this type; for example, NH, C10, undergoes dissociative sublimation into NH, and HC10, [33]. For a general process v;YJlj in the gas, in which the condensed material is transformed to the surface equilibrium condition (for an ideal gas mixture and a solid whose thermodynamic properties are independent of pressure) is
j= 1
n
N
pyli = KJT)~
(13)
where K,(T) is an equilibrium constant for partial pressures [compare equation (A26)]. When equation (13) is applied with p j = p X , , it is important to realize that the mole fractions X j in the gas are not deducible directly from the composition of the solid. There are relationships for interface flux conservation of species, obtainable from equation (158), but these involve diffusion velocities in addition to the concentrations. A nearby source or sink for a species in the gasfor example, a reaction sheetcan influence the relative concentration of that species at the interface through differential diffusion. It is desirable to be able to ascertain in advance whether equation (6) or equation (12) is the better approximation. A few general observations can be made concerning this question. Polymeric materials generally gasify in irreversible rate processes, effectively exhibiting an infinite value of P ~ , ~ ( they T ) ; may be considered to exist in metastable states because molecules entering the interface from the gas do not combine to add to the polymer chain in the solid. On the other hand, for most simple subliming or is I much larger than the values of m typically vaporizing materials, W , ~ , C encountered in solid deflagration, and it may then be seen from equation (9) that surface equilibrium is a good approximation. The smaller CI is, the
238
Combustion of Solid Propellants
less likely it will be that equation (12) can be used. An increasing extent of molecular rearrangement upon gasification is associated with a decreasing likelihood of achieving interfacial equilibrium. For a simple dissociative sublimation such as that of NH,C10,, surface conditions may or may not correspond to equilibrium (they usually do for NH,CIO, [33], but not necessarily for the dissociative vaporization of B, 0, [41]). For processes that are more complex than these, finite surface rates usually are encountered. In doublebase propellants, the situation is more complicated in that the nitrocellulose component cannot achieve equilibrium, while equilibrium vaporization of nitroglycerin may occur. This complexity enables interfacial equilibrium to be unimportant for some processes (for example, steady deflagration [40]) but possibly of significance for others (for example, radiant ignition [42]). Use of the more general interface condition given by equation (ll),instead of equation (6) or equation (12), has not been made in studies of deflagration of solids but could be warranted if a is in an appropriate intermediate range.
7.4. DEFLAGRATION CONTROLLED BY CONDENSEDPHASE REACTION RATES
If there is exothermicity in the condensedphase reaction zone illustrated in Figure 7.1, then the burning rate may be controlled by reactions that occur in this zone. Deflagration analyses paralleling those discussed in Chapter 5 may be developed for condensedphase combustion. For illustrative purposes, let us assume that the condensedphase reaction is the onestep process
WdS)
+
%(s),
(14)
which may be considered to precede equation (1). An Arrhenius reaction of zero order will be treated here for simplicity. The most important physical difference between condensedphase and gasphase reactions is that diffusion of species is much slower in the condensed phase; here we shall neglect molecular diffusion entirely to expose this difference most strongly. A formulation like that of Chapter 5, with the assumption of low Mach number 0,approximately a constant) introduced at the ouset, results in
dYJdx
=
wl/m
(15)
and
ildT/dx = m(h0 Yo
+ h,Y,) + constant,
(16)
in place of equations (57)and (59). With the heat released per unit mass of
7.4. Deflagration Controlled by CondensedPhase Reaction Rates
239
reactant consumed in the condensed phase defined as q, (16) may be written as
=
h:

hy, equation
where the subscript 0 identifies conditions at x = a (at which point dT/dx = 0) and where c p is the specific heat at constant pressure for species 0 and 1, assumed equal. In equation (15) for a zeroorder reaction, we write
w1 = p ~ l ~ a i e  E ~ / R o T
(18)
[compare equation (516)], in which all quantities now of course refer to the condensed phase. Boundary conditions for equations (15) and (17) are that Y,= Yl,,andT=T0atx=  m a n d t h a t Y , = l a n d T = q a t x = O . The problem that has been defined here possesses the coldboundary difficulty discussed in Section 5.3.2 and can be approached by the same variety of methods presented in Section 5.3. The asymptotic approach of Section 5.3.6 has been seen to be most attractive and will be adopted here. Thus we shall treat the Zel'dovich number
as a large parameter. Although a relationship between Y, and T like equation (524) cannot be derived because the Lewis number differs from unity, the variable Yl can be eliminated by differentiating equation (17) and using equation (15). From the solution to the resulting equation for T, the value of m is then found from the integral of equation (15), namely,
= w 1 dx/(l  yl,o), in which the asymptotic expansion of this divergent integral is intended. There is an upstream convectivedifuusive zone in which only convection and heat conduction occur. In the variable 5 defined in equation (518), the solution in this zone is Y, = Y,, and
J,
0
where a matching condition has been anticipated in putting q equal to the upper limit of the integral on the lefthand side. This solution allows for a variable heat capacity, and subsurface phase changes (for example, melting) are readily included by introducing delta functions in c p . In the reactivediffusive zone the stretched variables q =  b5 and
240
Combustion of Solid Propellants
may be introduced to produce in lowest order the problem d2yldq2 = AeCY7 (23) with y = 0 at q = 0 and dy/dq + 1 as q 4 c c (through matching), where
which is assumed to be of order unity. The first integral of equation (23), subject to the matching condition, gives (25) provided that 2A I1. In the innerzone variables that have been introduced, the first approximation to equation (20) is
d y l d n = (1  2Ae')''2,
&=A
where
[
m
eYdq,
(26)
Use of equation (25) in equation (26) produces, after algebra,
A = B (1  +a).
In dimensional terms, this result is
ROpLBl T;i +221eE1/R0T~
(28)
m = ( El(1

Y,,)[ " c p d T
To
%(I 1  Y1,J 2
The subscripts i here on p and A indicate that they are to be evaluated in the condensed phase at T = ?I. The derivation of equation (29) has not employed conservation conditions at the interface with the gas; hence the value of .T; remains undetermined. Therefore, additional physical information must be introduced to calculate rn. The simplest condition to apply at the interface is adiabaticity, dT/dx = 0, which serves to isolate the solid from the gas and shows from equation (17) that l;; c,dT = qs(l  Y l , o ) therefore , giving 7;. In this case equation (29) reduces to a formula for rn derived originally by Zel'dovich [43]. It is interesting to compare the resulting expression with equation (575) for a gaseous flame with a Lewis number of unity. In the first approximation the two formulas differ only by an additional factor of j'/' on the righthand side of equation (575). This difference is readily understandable on the basis of the physical reasoning given in Section 5.3.7. Diffusion of the reactant is responsible for a factor p'/' in equation (575) through the consequent reduction in the reactant concentration (hence in the reaction rate)
1
I.
(29)
7.4. Deflagration Controlled by CondensedPhase Reaction Rates
24 1
in the reaction zone. This diffusion is not present in the condensedphase deflagration, and the corresponding factor is therefore absent. The reactant concentration in the condensedphase reaction zone is of the same order of magnitude as it is in the cold propellant. This fact could have been used to scale equation (15), thereby producing, with equation (23), a pair of simultaneous differential equations for describing the reactionzone structure 1401. The solutions obtained with the adiabatic boundary condition are unstable to onedimensional disturbances (see Section 9.2). The derivation of equation (29) is invalid if q,(l  Y1,J > c p dT, since equation (17) then implies that dT/dx < 0 at x = 0, so that contrary to assumption, the peak temperature of the condensed phase no longer occurs at its surface. In a sense, the adiabatic condition represents a limiting condition at which the condensedphase deflagration breaks away from the surface and ceases to be influenced by interface or gasphase processes. From the viewpoint of the condensedphase analysis, it is necessary that q,(1  Yl, o) I j$o c, dT. When the inequality prevails in this relationship, heat is fed into the condensed phase at its surface. With To, q,, and other condensedphase properties fixed, this energy input increases m because the consequent increase in IT; exerts a larger influence on the Arrhenius factor in the numerator of equation (29) than on the integral in the denominator. Strictly speaking, whenever the inequality holds, the burning rate is not controlled entirely by condensedphase processes because interface or gasphase processes then influence m by modifying IT;. The simplest approach to the use of equation (29) in these nonadiabatic cases is to assign Ti a constant value, a “gasification temperature” that is presumed to be known, but this approach fails to address the question of the physics that determine the value of q. If we consider Ti fixed and q, decreasing, then we find that the derivation of equation (29) remains valid as q, vanishes or becomes negative. The requirements of the analysis are that 7; > To and that there be a distributed condensedphase Arrhenius reaction with large. The reaction may be exothermic, energetically neutral, or endothermic ; in the latter case, A and B are negative. Under these conditions, the condensedphase reaction naturally does not control the deflagration rate, but equation (29) provides a relationship between m and 7; that is necessary for compatibility with the condensedphase processes. Many polymeric binders of composite propellants experience endothermic gasification of this type. In fact, the basis of the analysis given here was first developed for endothermic linear pyrolysis of polymeric materials [44], [45]. A number of modifications and generalizations of the analysis presented here can be explored [14], [40], [45]. A reaction order of zero has been selected here for simplicity of development. By working with two simultaneous differential equations in the reaction zone, the formulation is easily extended to an arbitrary reaction order n. For 0 < n < 1, a modified version
242
Combustion of Solid Propellants
of equation (29) is readily derived [45]. With the most natural assumption of unimolecular decomposition, we have n = 1, but for n 2 1, a difficulty is encountered in that equations (15) and (16) allow Y, to approach unity only as x + + co ;just as in the gaseous flame (see Section 5.3.7), an infinite time is needed for complete conversion of reactant to product at these larger values of n. To place the interface at a finite location with n 2 1, a value of Y, less than unity must be selected at the interface [45]. The selection may be made arbitrarily on the grounds of insensitivity of predictions to the value chosen within a reasonable range, or an additional physical phenomenon such as equilibrium vaporization, equilibrium solubility, or onset of dispersion may be introduced formally to fix the surface value of Y, [14]. Indeed, these physical phenomena may be relevant even for 0 5 n < 1, and equation (29) may be modified accordingly by taking Y, < 1 at the surface in carrying out the analysis. These refinements have been introduced mainly in conjunction with an assumption of surface adiabaticity [141. In addition, adiabatic analyses are available for arbitrary n 2 0 for situations in which the condensed phase extends to x = + co [14]. Effects of nonadiabaticity associated with radiant flux incident on the surface of the solid have been analyzed by methods similar to that given here [40]. So long as the finiterate chemistry occurs in a single reactivediffusive zone extending to the surface of the condensed phase, the analysis need not be restricted to a onestep reaction. Known mechanisms of homogeneous polymer degradation may be taken into accountfor example, [45]. The results continue to be expressible in formulas that resemble equation (29) but that usually are somewhat more complicated. There are numerous possibilities for using equation (29) in conjunction with interface or gasphase conditions to determine both m and 7;. For example, the condensed product of equation (14) conceivably might experience the rate process of equation (1) as soon as Y, = 1, so that equations (6) and (29) become two independent expressions for m and T . Alternatively and somewhat less unlikely, surface equilibrium may occur so that equation (12) determines in equation (29); in this case, a gasphase analysis is generally needed to find p , , i. It appears that in most real homogeneous propellants, the products of the exothermic condensedphase reactions are mainly gaseous, so that considerations of dispersion or possibly of gasphase reactions are most relevant for determining T in equation (29), and equations (6), (ll), or (12) are not directly useful. The general similarity between equations (6) and (29) may be noted. Both exhibit an Arrhenius dependence of m on T , and measurements of the variation of m with T , therefore do not distinguish between the two equations, even though the processes represented fundamentally are quite different, one being an interface reaction and the other in reality a distributed condensedphase reaction. The activation energy E for the surface reaction is seen to correspond to E,/2, half of the activationenergy for the
7.5. Deflagration Controlled by GasPhase Reaction Rates
243
bulk reaction, because of the squareroot dependence of the deflagration rate on the homogeneous reaction rate. This squareroot relationship clearly applies in endothermic linear pyrolysis as well. Thus in steady deflagration or linear pyrolysis, for many purposes the distributed condensedphase reaction may be approximated as a surface reaction having half of the true activation energy, thereby achieving simplifications in the analysis. Fundamentally, equation (6) is seldom every correct for polymers, but since equation (29) often is, equation (6) may be employed, with suitably adjusted values of E and of the preexponential factor. If the solid were inert and the interface gasification process were exothermic, releasing the energy qi per unit mass of solid gasified, and if the system were adiabatic with negligible energy input from the gas, then lF:,cpdT = q i would determine IT;., after which equation (6) would give m. This result, which constitutes the simplest complete model for solidpropellant deflagration, can be made to correspond to the adiabatic version of equation (29) by putting E , / 2 = E and q i = q,(l  Y1,J and equating the factors preceding the exponential. Gasphase reaction rates may modify these results if heat conduction from the gas to the interface occurs.
7.5. DEFLAGRATION CONTROLLED BY GASPHASE REACTION RATES
If we overlook the processes occurring in the region x I0 in Figure 7.1, then the illustrated burning mechanism of the solid closely resembles that of a premixed gaseous laminar flame. To proceed with an analysis, let us assume for illustrative purposes that the gaseous reaction is the onestep unimolecular process %(i4 1IJz2(g), (30)
+
where the subscript 2 identifies the reaction product, and also adopt (in the gaseous region) the other basic simplifying assumptions employed in Section 5.3 (namely, lowspeed flow and assumptions 111 of Chapter 5). Then the results of Section 5.3 can be applied to a deflagrating solid in which premixed gasphase reactions influence the burning rate. The purely gaseous deflagration extends from x =  co to x = + 00, with cold boundary conditions applied at x =  co ; but the gaseous region of the burning solid extends only from x = 0 to x = + co,with cold boundary conditions applied at x = 0. This is the primary difference between a gaseous laminar flame and our present model of a burning solid. Any governing equations that can be derived for the gaseous flame without making reference to the cold boundary conditions will apply equally well to the burning solid in the region x > 0. If the normalized mass flux fraction of products is defined as t 1  (tZ.m  t2)/A~2, (31)
244
Combustion of Solid Propellants
where t2 is the flux fraction of species 2, the subscript co identifies conditions at x = co,and At2 is the total change in t2 in the gas, the normalized mass fraction of products is defined as
+
Y
1  (Y2,, 
Y2)/At2,
and the dimensionless temperature is defined as
7 1 c,(T  T,)/q, (33) where c p is the (constant) specific heat at constant pressure for the gas and q is the total heat released in the gas per unit mass of the gas mixture, then, through a development similar to that given in Chapter 5, the reader may show that equations (524) and (542) can be derived without using the coldboundary conditions. Thus the relations
+
Y=T
(34)
and
dt/dT
= h / ( Z  t),
(35)
where the dimensionless reactionrate function w(z) and the (constant) dimensionless massburning rate A are given by equations (543) and (549, respectively, remain valid in the gaseous region for the burning solid. Actually, in order to remove all reference to conditions at x = a in the purely gaseous problem, the quantity a' appearing in equation (543) must be defined as
a' = q/(c,T,  d,
(36)
which is equivalent to equation (529) for the gaseous flame. The boundary conditions for equation (35) at x = co for the burning solid are the same as those for the gaseous flame; namely, t = 1 and z = 1 [see equations (31) and (33)]. However, as we have already stated, the coldboundary conditions differ in the two problems. For the burning solid, the definition of At2 implies that
+
At2 = t 2 , ,

t2,i,
where the subscript i identifies conditions in the gas at the interface (x = 0). Hence, equation (31) shows that t = 0 at x = 0. Equation (33) implies that T = ziat x = 0, where (37) Therefore, the coldboundary condition (at x = 0) for equation (35) in the burning solid is t = 0 and T = zi. It will be noted that these boundary conditions ( E = 0 at T = zi and t = 1 at T = 1) are formally identical with the boundary conditions introduced in Secti,on 5.3.2 in order to remedy the coldboundary difficulty for the
Ti
1 k
C,(T
 T,)/q.
7.5. Deflagration Controlled by GasPhase Reaction Rates
245
gaseous flame. In Chapter 5 the quantity zi was a dimensionless ignition temperature for which a pseudostationary value was chosen ; for the present problem, on the other hand, ziis a welldefined dimensionless measure of the interface temperature. Thus the coldboundary difficulty does not arise in the present problem; the presence of the condensed phase has removed it by cutting off the reaction at the interface. A consequence of the equivalence of the two problems is that any of the procedures discussed in Section 5.3 for determining the laminar flamespeed eigenvalue A of equation (545) can be applied without modification for finding A as a function of zi,a’, and [defined in equation (544)] in the present problem. In using these procedures, one should recognize that w(z) is given by equation (543) only for z > z i ;w = 0 for z < zi. Results of the asymptotic analysis of Section 5.3.6 have been used in a number of investigations of solidpropellant deflagration [43], [46][49]. From the discussion associated with equation (575) (where now a = q/c,T, and b ,’ = E,q/Roc,T2), it is seen that A is then found to be independent of when the gasphase properties and the zithat is, m is independent of overall energetics (T,) are fixed. In this respect, the gasphase reaction controls m completely; other processes serve only to establish values for other quantities, such as q ,when the asymptotic analysis that has been given is applicable. With adiabatic combustion, departure from a complete control of m by the gasphase reaction can occur only if the derivation of equation (575) becomes invalid. There are two ways in which this can happen; essentially, the value of m calculated on the basis of gasphase control may become either too low or too high to be consistent with all aspects of the problem. If the gasphase reaction is the only rate processfor example, if the condensed phase is inert and maintains interfacial equilibriumthen m may become arbitrarily small without encountering an inconsistency. However, if a finiterate process occurs at the interface or in the condensed phase, then a difficulty arises if the value of m calculated with gasphase control is decreased below a critical value. To see this, consider equation (6) or equation (29). As the value of rn obtained from the gasphase analysis decreases (for example, as a consequence of a decreased reaction rate in the gas), the interface temperature q, calculated from equation (6) or equation (29), also decreases. According to equation (37), this decreases zi.Eventually, corresponds to at a sufficiently low value of m, the calculated value of zi = 0. As this condition is approached, the gasphase solution approaches one in which d T/dx = 0 at x = 0, and the reaction zone moves to an infinite distance from the interface. The interface thus becomes adiabatic, and the gasphase processes are separated from the interface and condensedphase processes. At this limiting condition, the gasphase deflagration becomes irrelevant to the solidpropellant deflagration. Either equation (6) or (29) can now
246
Combustion of Solid Propellants
be used to calculate the burning rate in conjunction with a condition of surface adiabaticity. If L denotes the net chemical enthalpy per unit mass needed for gasification of the condensed phase, then an energy balance for the condensed phase shows that this adiabaticity condition can be written as c,dT
JT,
=

L,
which serves to determine IT; without reference to the gaseous phase. In principle, equation (38) can be applied irrespective of the sign of L and irrespective of whether the value of obtained from it is greater than or less than To. For example, for the model that led to equation (29), if species 1 undergoes endothermic sublimation or vaporization after Yl = 1 with a heat of vaporization per unit mass of L,, then L = L,  q,(l  Y1,J. For solid propellants, it is usual that L < 0 and > To when equation (38) applies.* An abrupt change from gasphase control of m to interface or condensedphase control occurs when surface adiabaticity is reached. In this sense, even if the conductive energy feedback from the gas is a small but positive fraction of the total amount of heat release that occurs prior to complete gasification, the gasphase reaction controls the burning rate. If the gasphase reaction rate is reduced further, below a value corresponding to surface adiabaticity, then there no longer exists a gasphase solution with an m that is consistent with condensedphase and interface conditions. A “breakaway” phenomenon occurs in which the burning rate is controlled by condensedphase or interface processes, and the gasphase deflagration propagates more slowly than the solidpropellant deflagration. Under these conditions, after the gasphase deflagration has moved away, the gasphase chemistry occurs in a convectivereaction process with negligible molecular transport, anchored by the condensedphase deflagration and analyzable by methods discussed in Section B.2.5, for example. Analyses have been given of this gasphase reaction zone, which no longer influences the burning rate [14], [SO], [Sl]; expressions for the distance from the interface to the region of major gasphase reaction have been obtained by application of explosion theory. The second way in which the derivation of equation (575) can fail is for the thickness of the reactivediffusive zone in the gas to become comparable in size with the thickness of the convectivediffusive zone. This occurs if (T,  7;)/T, becomes of order p’, which would be favored by low overall
* This entire discussion has been subject to the restriction that L is not sufficiently large and positive to require, 7; in equation (38) to be less than absolute zero. If the condensed phase is sufficiently endothermic to violate this restriction, then equation (38) cannot be satisfied, surface adiabaticity cannot be approached, 5, cannot approach zero, and solidpropellant deflagration with gasification cannot occur in the absence of some energy input from the gas phase. In this case the gasphase reaction must always control the burning rate (at least when the burning involves complete gasification); if the surface equilibrium defined by equation (12) applies, then unusual shapes of curves of the burning rate as a function ofpressure are obtained [8], [49].
Since the presence of the condensed phase has removed the coldboundary difficulty. 0 .)/T. we find from an overall energy balance that T. Effects of energy losses may differ for different mechanisms of burning and for different types and locations of losses. heat losses may be responsible for limits of deflagration.5. and convection all are important. The reaction zone in the gas then requires a modified analysis. then there is a single gasphase zone in which reaction. C491.. attention will be restricted to systems that are controlled by gasphase reaction rates and for which equation ( 6 ) applies. then the gasphase reaction zone continues to involve a diffusivereactive balance in the first approximation. whereupon the convectivediffusive zone in the gas disappears. For brevity. Thus far our discussion pertains to conditions of overall adiabaticity. Deflagration Controlled by GasPhase Reaction Rates 247 activation energies for the gasphase reaction. diffusion. . If q R denotes the radiantenergy flux from the surface. the gasphase reaction continues to control the burning rate. . increasing m (by increasing the gasphase reaction rate) eventually causes (T. If fi is large. Enhancement or control of burning rates may be considered by external application of radiant energy fluxes [40]. discussions of other systems may be found in the literature [S]. Analysis shows that when equation (575) fails for these reasons. C541.3 and 5. The methods of Sections 5. In laboratory experiments on linear burning of strands of solid propellants.T. to become of order b'. and a corresponding modification to equation (575) occurs. need not be large for a mathematically welldefined deflagration problem to exist. let us consider radiantenergy loss from the interface between the condensed and gaseous phases.qR/(mCp). (39) * Strands are long cylinders with burning inhibited on all faces except one of those normal to the axis. This same effect occurs with surface equilibrium [equation (12)] if increasing the pressure p is the means for increasing the gasphase reaction rate (but not if p is held constant and the reaction rate is increased instead by increasing the preexponential constant in the rate expression). To illustrate a few of the phenomena that may occur.3.3. which is readily developed. Nonadiabatic deflagration of solid propellants is of practical interest in many contexts. [53]. and the interface with the condensed phase forms a boundary of the reactivediffusive zone at a finite location in the stretched variable. but m becomes bounded in the limit of an infinite gasphase reaction rate. a process that often is significant in laboratory experiments with strand burners.* In rocket motors these losses tend to be reduced by energy input from hot surroundings. but with equation ( 6 ) or equation (29) applicable. then under the assumption that this radiation is not absorbed elsewhere in the system.4 may be used to analyze the gasphase deflagration in these cases. If b is not large. such applications of these methods have been made [52]. = T.7.
Nevertheless.the situation is somewhat more complicated if the gasphase reaction zone extends to the surface of the condensed phase. In equation (39). the derivation of which is somewhat complicated algebraically* even on the basis of equation (575). In the pressure range where two solutions for m exist.2. the equation cannot be satisfied) and the overall order (the pressure exponent) of the distributed gasphase reaction is two. in both cases it is found that if qR > 0. in equation (38) is negative (that is. the increase in qR is obtained by an increase in C. Therefore. a solution no longer exists for m when the pressure p is less than a critical limiting value. (59)]. where Tm. . It may thus be seen in general that whenever equation (575) applies. The absence of a solution at low pressures is an example of a prediction of a deflagration limit caused by heat loss (see Chapter 8). when q R # 0. T. in equation (575) for m.a is the value of T. it is usually assumed that the lower burning rate represents an unstable solution and that the upper is correct physically (compare Sections * The equation to be investigated may be shown to be roughly of the form where A. B. and D are positive constants. C. is illustrated schematically in Figure 7. which corresponds to a system in which the value of . This result.248 Combustion of Solid Propellants FIGURE 7.T.2. and there are two solutions if p exceeds this value. . qR depends on the surface temperature (typically q R is proportional to T f ) . Equation (39) can be derived formally from the general interface energy conservation condition given in equation (161) and the energyconservation equation for the gas and the solid [for example. Schematic illustration of the pressure dependence of the mass burning rate when the gasification process is an unopposed rate process. under adiabatic conditions (qR = 0) for the same initial temperature To. heat losses affect the burning rate only through their influence on T. now depends on both Ti and m through equation (39).
1. Dispersion might begin as a consequence of attainment of an equilibriumlike condition for transition to dispersion or in a rate process for subsurface generation of gases. geometric configuration of the dispersion (gaseous bubbles. and good agreement between theoretical predictions and experimental observations has been obtained [56].6.6.7. DISPERSION PHENOMENA AND OTHER INFLUENCES Let us now consider combustion mechanisms that. and other attributes.3). for which it has been observed that the ratio of the energy feedback (energy transferred from the dispersed phase to the condensed phase) to the energy released in the condensed phase varies from about 0. [SS]. If processes occurring in other zones are important as well. in the condensed or gaseous component or at their boundaries).4 and 8.1. then the theoretical problem is more complicated because accurate conditions for the onset or termination of dispersion become necessary in developing analyses with reasonable accuracy. solid rods. given the model of Section 7. Some further discussion of the function m(p) will be given in the following section.3 at 5 atm to about 0.2. The new difficulties encountered are fundamental because the physics of these processes are not well understood. The consequent occurrence of twophase flow adds substantial complication to the deflagration mechanism. Results are available for a few models that have dispersedphase control [13][15]. Specifically. the location of the heat release in the dispersed phase (for example. strictly speaking. at the interface in Figure 7. An example of a propellant in which conditions for onset of dispersion are relevant to the burning mechanism is nitrocellulose. differing according to the equilibrium versus finiterate character of the gasification step in the dispersed phase. and so on). When dispersedphase processes contribute in any essential way to the burning rate. let us introduce the dispersed phase mentioned in Section 7. Polyvinylnitrate is a material that appears to burn in a process controlled by dispersedphase combustion. since then the problem may be analyzed by asymptotic methods on the basis of purely dispersedphase control. For example. There are numerous possible models of this type. 7.1 at 100 atm [ls]. the problem is simplest if the other phases may be treated as nonreactive and the reaction rate in the dispersed phase is strongly temperaturesensitive.4 for condensedphase processes. are not represented by the model illustrated in Figure 7. liquid spheres. Dispersion Phenomena and Other Influences 249 3. species 1 of equation (14) might be presumed to initiate dispersion upon attainment of an equilibrium relationship of the form .1.4.
~is a consequence of the definitions. The identity (dn/dTo).n. Thus a wide variety of complications may arise.. T . ~ . it is quite possible that the fundamental rate constants for the condensedphase reactions become pressure dependent. is called an explosion temperature and n is a pressure exponent of the burning rate.i. A third conditionfor example. In equation (41). T. the value of m depends only on the two parameters To and p . respectively. Much research remains to be done in analyzing the deflagration of most of the homogeneous propellants for which dispersedphase processes are important. then rates of nucleation might be relevant. Various empirical relationships for m have been developed in thc past. Equation (40) and results of a condensedphase analysis provide two relationships among three unknowns. or it might be proportional to the total hydrostatic pressure if the phase transition is influenced by mechanical properties of a solid having a limiting “strength. Theoretical analyses often may provide agreement with experiment m = . for specifying the energy feedbackis needed.. . and n = In m/a In P ) . When a given propellant formulation burns in the absence of externally applied perturbations.250 Combustion of Solid Propellants where L. which must be less than unity if the propellant is to burn in a stable manner in a rocket chamber. Although the situation would be simplest if an adiabatic condition (zero temperature gradient) were achieved at the onset of dispersion. T. Use of an equation like equation (40) as a downstream boundary condition for analysis of the condensedphase reaction does not eliminate the necessity of analyzing the dispersedphase processes. B. becoming proportional to m2. and the problem of developing a correct analysis is challenging. c. applicable for a range of values of To and p . This function would be proportional to the partial pressure of species 1 in the gas if an equilibrium condition for absorption of this species applies.” If the transition to dispersion is a rate process. is a solubility constant. and n are three positive constants that serve to characterize m and its pressure and temperature sensitivities. and Yl. representing an energy of nucleation and with B. and f ( p ) is an increasing function of p. but with L.To) as well as relationships involving other functional dependences on p [8]. small amounts of energy feedback can be significant because of the strong dependences of the exponential factors on q. These include (a (41) cp”/(T. Thus a nonnegligible coupling between the reaction zones often remains. should agree with experiment concerning the values of m. A good theory of propellant deflagration. and n. is a heat of solution. = p(d~. all are of use in helping to ascertain deflagration mechanisms. Measurements of m. At the high pressures encountered in the use of gun propellants ( p on the order of 1000 atm). and a formula like equation (40) again may be derived. and rc. rc. m./dp). Temperature and pressure sensitivity coefficients are defined as rcT = (d In m/dTo).
7. For these propellants gasphase or liquidphase heat release subsequent to mixing or heterogeneous heat release at points of contact between the two constituents must play an essential role in the deflagration mechanism. These effects may be produced in conventional doublebase propellants by suitable addition of burningrate catalysts (typically certain metalorganic salts) to the propellant formulation. There are plateau propellants that exhibit an intermediate range of pressure over which m is practically independent of p and mesa propellants for which m(p) achieves a maximum at a particular value ofp then a minimum at a higher value. Thus dispersion phenomena are of importance to the deflagration of homogeneous propellants in a number of ways. for * This is especially true if at least one of the propellant constituents can experience exothermic deflagration in the absence of the other. .To)in the denominator may often be raised to a positive power k different from unity without introducing disagreement with measurements.. gasphase combustion appears to be of greater importance for heterogeneous propellants than for most homogeneous propellants. There are theories that predict positive values of k different from unity [40]. [ S S ] . Most of the currently used heterogeneous propellants are of this type. In recent years an appreciable amount of research has been devoted to the development of comprehensive models that take into account all effects that are known to be potentially relevant. the occurrence of diffusion flames between fuel and oxidizer constituentsbecome relevant to the combustion process. many other phenomenafor example. values for which can be selected within reason to obtain good agreement with measurements. the factor ( T . To include all the potentially significant phenomena in a theoretical analysis becomes a formidable task that can be pursued only through the introduction of drastic idealizations for each of the phenomena considered. In general.7.7. COMBUSTION OF HETEROGENEOUS PROPELLANTS In principle. For homogeneous propellants. Combustion of Heterogeneous Propellants 25 1 concerning c and n without predicting the temperature sensitivity correctly [40]. since measurements can be made only over a limited range of To less than T. These theories perform important functions beyond the correlation of existing data. 7.* Moreover. Model propellants can be considered in which neither constituent can burn without the other. all the theoretical complexities that we have discussed for homogeneous propellants also may arise in attempts to analyze the deflagration of heterogeneous propellants. The results are expressions for burning rates that contain several adjustable parameters. It has been shown experimentally that these catalysts usually operate by modifying the interaction between the condensedphase and dispersedphase reaction zones [57]. . such analyses must be deemed incorrect in a fundamental sense.
[63]). The principal trend of research on comprehensive models is toward additional complexity rather than toward independent evaluation of the appropriateness of the values of parameters employed. . without addressing questions of dependences on other parameters. By focusing attention on limited aspects of the combustion mechanisms.252 Combustion of Solid Propellants example. and we shall discuss in a little more detail a few of the simplified models that address only certain specific phenomena. fuel pyrolysis. In fact. and improved predictions of dependences of burning on certain parameters can be developed. The intent is to develop codes that reproduce burning rates of propellants of practical interest rather than to clarify fundamentals of the chemical kinetics and of the physical processes that occur during deflagration. the more recent publications seldom ever even bother to list the values of the parameters that were employed. However. The analytical formulations generally are expressed in terms of sets of simultaneous nonlinear algebraic equations that are solved by exercising routines for electronic computers. and so on. The train of development may be followed in a sequence of publications [60][63]. The models contain various energy balances and different prescriptions for averaging over sizes and locations of oxidizer particles in the fuel matrix. the relationships between the values so obtained and independently determined values for the fundamental processes that they are intended to describe by and large remain uncertain. different oxidizer particle size distributions. and they thereby indicate propellant modifications that may be tried for tailoring combustion characteristics in manners desired in applications. Here we shall merely cite some literature and reviews of comprehensive models. The origins of comprehensive models are usually traced to the work of Hermance [59]. primarily with ammonium perchlorate as oxidizer. better understanding of these aspects can be obtained (including unforeseen restrictions on complete models). Although values of parameters in models have been selected successfully to produce good agreement with measured burning rates. they enable estimates to be made of burning rates of heterogeneous propellant formulations that differ somewhat from those that have been subjected to measurement. they tend not to lead to detailed understanding of specific aspects of the combustion processes. This is especially true for rate parameters such as overall activation energies. The criteria for successful models mainly have been taken to be agreement between theoretical and experimental dependences of burning rates on pressure for a variety of propellants involving different binders. A topic of recent interest has been to account for effects of bimodal size distributions (for example. Later. we shall consider composite propellants. who included a gasphase decomposition flame of ammonium perchlorate. then we shall make a few comments on the burning of black powder. and a heterogeneous reaction between fuel and oxidizer at their interface. First. a gasphase diffusion flame was added as well.
One approach is to characterize the oxidizer particles as monopropellant spheres of a specified. in addition./(A i Bm.). This observation led to the twotemperature concept. it is necessary that the average surface temperatures differ. eters A and E in expressions like m = to achieve the same average values of m / p for each constituent. in which the fuel and oxidizer components were assigned different surface temperatures that served to provide them with equal regression rates. a relationship between the two temperatures is obtained. The model is beneficial in demonstrating how the packing configuration can influence even the pressure dependence of the average burning rate of the composite propellant. [28]. the average burnthrough time of the binder. Various methods (for example. [64]) have been proposed for averaging over these variables to obtain average burning rates from burning rates of individual particles. constant radius with known. and one additional relationship is then needed to determine completely the two temperatures and the regression rate of the propellant.. The simplified model that gives equation (42) helps to explain potential influences of oxidizer particle size and particle loading on propellant burning rates. to characterize the fuel binder as a barrier having a burnthrough time that is a known. This relationship has been taken to be an energy balance relating the energy feedback from a gasphase flame to the energy needed for gasification of the propellant. Thus.7 Combustion of Heterogeneous Propellants 253 A formidable problem encountered in modeling composite propellant deflagration concerns the statistics of oxidizer particle sizes and locations. as needed for steady burning. presumed known. [66].7. By calculating an average time for propagation of burning along optimum paths. in terms of the mass burning rate of the oxidizer. is proportional to the average burnthrough time of the binder. . The earliest simplified models of composite propellant combustion were models in which pyrolysis laws like equation (6) were applied separately constituents [8]. the average burning rate of the propellant can be expressed in terms of the regression rate of the oxidizer spheres. Since the paramto the fuel (F) and oxidizer (0) differ for fuel and oxidizer. increasing function of the shortest distance between the surfaces of burning and adjacent nonburning spheres and to characterize the configuration of the oxidizer spheres as being nearly a tightly packed hexagonal lattice with a probability density function assigned to a parameter that defines a small departure from symmetry [65]. (which depends upon p ) . and geometric properties that include the particle size and statistical parameters of the particle configuration [65]. If a pressure dependence of m is to be obtained in a simplified analysis. constant rates of normal surface regression. m. By equating the two regression rates. a representative result for the mass burning rate of the propellant is [65] (42) where A and B depend on geometric parameters of the oxidizer distribution within the propellant and on the densities of the constituents and B. then = m m.
Burning occurs on a face that is perpendicular to the planes of the slabs. However. Summerfield [67] first discussed a model in which the rate of diffusion of gaseous fuel and oxidizer to a conical flame sheet above a fuel pocket controlled the overall burning rate of the propellant (see also [28]). [66]. the latter was employed in the original developments [28]. [28]. This model has been augmented in various approximate ways to include premixed gasphase flames as well [4]. Many experimental studies have been made of the combustion of sandwich propellants (see [70] and references quoted therein). The concept was first discussed in the literature in [28]. only one slab of oxidizer and one of fuel or one of fuel between two of oxidizer. The former view might be better €or sufficiently finely mixed propellant constituents and the latter for coarser formulations. An analysis of an infinite array of uniform fuel slabs alternating with uniform oxidizer slabs has been completed in a flamesheet approximation for diffusion flames that either extend from the lines of contact between solid fuel and oxidizer to infinity (under conditions of overall stoichiometry. [S]. in defining precisely what heats whatand there are consequent uncertainties and inaccuracies in the twotemperature models. or it may be considered that the mixing is relatively slow and unimportant and that the oxidizer gases support a decomposition flame that is responsible for the energy feedback. a sandwich model of propellant combustion has been conceived and subjected to a number of studies. the gasphase equations . A sandwich propellant is a heterogeneous propellant composed of a series of parallel slabs of fuel and oxidizer.1 for the BurkeSchumann problem. thereby greatly improving agreement with meassured pressure dependences of burning rates of propellants but also preventing detailed analyses of gasphase flow fields from being accomplished with reasonable effort. Choices must be made in imposition of an energy balancefor example. alternately placed side by side and bonded together. To enable investigations of influences of diffusion flames to be pursued most simply. [72]). It may be imagined that mixing of fuel and oxidizer gases is rapid and that the gasphase flame is a premixed flame in the resulting gas mixture. [67][69]. but for purposes of clarifying processes occurring locally. the general conclusion that different constituents are likely to have different average surface temperatures is an important aspect that has found application in most of the newer modeling efforts. The approximations in the gasphase portion of the analysis are essentially the same as those discussed in Section 3. A heterogeneous propellant is approximated by an infinite number of slabs.254 Combustion of Solid Propellants the gasphase flame must be premixed. [71]) or close over the fuel or oxidizer components (under nonstoichiometric conditions. investigators have often studied a small number of slabsfor example. Another class of simplified models focuses on diffusion flames between gases generated from fuel and oxidizer constituents. The advantage of the sandwich geometry is the simplification in analysis produced by the absence of variations in one of the directions.
It is a mixture of saltpeter.4. Through nonlinear effects the heterogeneities themselves can modify the average burning rate [76]. and sulfur in weight percentages of about 75. the gaseous fuel being treated as inert [74]. In this model the propellant is burnt on a face parallel to the planes of the slabs. heat of gasification. (43) + + + + . timedependent conservation equations describe the burning . and the constituent most difficult to pyrolyze thereby determined the burning rate. 3C S + K2S N.7 Combustion of Heterogeneous Propellants 255 are solved in a halfspace with surface conditions describing interface species and energy conservation.7. These oscillatory burning phenomena are discussed further in Chapter 9. theoretical analyses have been performed for a semiinfinite. [77]. endothermically pyrolyzing fuel slab adjacent to a semiinfinite. Black powder is a heterogeneous propellant of unusually complex structure in that three different solid phases are involved. heat capacity. there is a similarity to the analysis of oscillatory burning of homogeneous propellants (see Section 9. and rate parameters may be considered [77]. and if mechanisms exist for communication of the unsteadiness to distant portions of the burning surface. The model has no bearing on the occurrence of diffusion flames but instead is a vehicle for studying influences of heterogeneities within the solid on the combustion behavior.1. These heterogeneities result in local unsteadiness of the burning. while the pyrolysis parameters of the other constituent determined the difference in heights of the fuel and oxidizer surfaces in an approximation in which these surfaces were planar. However. 15. Pyrolysis laws like equation (6) were applied to each propellant constituent. such as in details of the shape of the burning surface. in qualitative agreement with observed decreases in burning rates of heterogeneous propellants with increasing oxidizer particle sizes. endothermically pyrolyzing oxidizer slab [73] and for this same configuration with the oxidizer supporting a gasphase decomposition flame and with the diffusion flame neglected. there are many aspects in which the model did not attempt to achieve agreement with experiments performed with sandwich propellants. Heterogeneities in the solid’s density. then intrinsic oscillations of the propellant combustion may develop [77]. charcoal.5). Onedimensional. heat of combustion. 3c02. These analyses have provided some new understanding but have not taken into account potential influences of finiterate chemistry in the gasphase diffusion flamesinfluences that have been suggested [75] to affect burning rates of heterogeneous propellants and that may be treated with help of the asymptotic methods discussed in Section 3. The analysis predicts a decrease in burning rate with increasing slab thickness. Another type of sandwich propellant is the “sideways sandwich” suggested independently in [76] and [77]. and 10. thermal conductivity. In efforts to predict surface shapes better. Although these percentages correspond to stoichiometry for the overall reaction 2KN0.
with p representing the density of the bulk propellant. results are shown in Figure 7.. and the pellets often are glazed with graphite before use.256 Combustion of Solid Propellants the observed products of combustion are not entirely those of equation (43) but instead include a variety of species such as NO. The normal regression rate of black powder as a function of pressure at 25°C ~831. The particles are mixed and pressed into cakes that are ground into pellets having mean diameters typically between 300 p and 3000 p. 0. I Blackwood & Bowden o Belyaev & Maznev 0 1 0 Glazkova & Tereshkin Belyaev et al. spreading flames rapidly and in a reasonably reproducible manner through a somewhat loosely packed configuration of pellets.SO. The combustion of black powder has been subjected to scientific scrutiny since antiquity. elevated [SS]. NO. for transmitting combustion from the initiator to the main propelling charge in gun propellants) because it provides a margin of safety by being relatively difficult to ignite while maintaining reliability since once ignited. and K. Noble [78] and Vieille [79]. Each of the three reactants in equation (43) is present in a separate phase in particulate form. it burns vigorously even at reduced pressures. In addition. rates of spread of flames among arrays of pellets loo I 0. for example. Burning ratesthat is. where the regression rate is Y E m / p . Detailed experimental studies have been performed by use of more modern techniques [80][83].. 0 0 0 p atm FIGURE 7. This ancient art produces a propellant that even today is better suited than any other for certain specialized applications (for example. with particle sizes between 1 and 100 p. normal regression rates of surfaces of pellets or of tightly pressed powdershave been measured at reduced [S4]. contributed to knowledge of its burning characteristics.1 ' I I I I I I 1 0 I00 I000 1 0 . and very high [SS] pressures. .3.3.
Melting of S (and possible of KNO. The model involves a controlling twophase (dispersedphase) reaction zone in which carbon particles burn in a gaseous oxidant. Two candidate mechanisms for theoretical description of the flame spread are the motion of hot gases through the gaspermeable propellant and a random walk of hot K. it appears that this model does not reasonably provide agreement with observed burning rates [91]. where they rapidly induce local ignitions [83]. from a burning liquid layer on the surface of a pellet and impingement of these particles on the surfaces of adjacent pellets. mainly oxygen. which may be simplified and applied to flame spread in black powder to provide as an estimate of the spread velocity [S3] I/ = b / P g ) ( d 4 (44) where 1 is a representative length of the propellant charge. The randomwalk model is illustrated schematically in Figure 7.7. and 6 is the void fraction (the ratio of the volume not occupied by the pellets to the total chamber volume). No theories have been pursued yet for calculating the regression rate of black powder on the basis of control by condensedphase reactions. Combustion of Heterogeneous Propellants 257 have been measured [SO] and found to be much larger than normal regression rates. SO. .. replaced by u [83].4. particle velocities on the order of 5000 cm/s have been observed.7. Although the theory of the process has not been developed.) leading to mixing and exothermic reaction in a melt layer is known to be significant in ignition of black powder [80]. Flame spread among pellets of black powder appears to involve emission of fine droplets of hot salts containing K. these produce spread rates on the order of 60 cm/s. particles through the propellant matrix. the four drawings illustrate the motions of salt particles at four successive time steps for a situation in which there is multiplication by a factor of three in each step. and rates of heat release in a melt layer also may control the rate of normal regression [83]. with multiplication of the number of particles at each stop through a local surface ignition leading to renewed particle emissions [83]. Further theoretical and experimental studies of the flamespread processes could help to clarify the extent to which formulas like equation (44) may be applicable. m is the normal mass burning rate of the propellant. Despite contentions to the contrary [90]. There are theories for burning of gaspermeable propellants [92][95]. produced from the decomposition of KNO. where u is the velocity of the hot salt particle and 6 is the average distance between surfaces of adjacent blackpowder pellets. A theoretical model has been developed for the rate of normal regression of a burning surface of black powder [87][89]. the resulting spread velocity appears to resemble that of equation (44) in some respects. SO. with m / p . At atmospheric pressure. C J is the surfacetovolume ratio (the total pellet surface area per unit chamber volume). pg is the density of the product gas.
the gas velocity parallel to the burning surface can attain high subsonic values. However. More accurate experimental evidence indicates that contrary to equation (45).5. a “threshold” value of M exists. (c) (4 Schematic model of random walk with multiplication [83]. the average value of the component of velocity parallel to the surface of the solid has been assumed to be zero. In particular. and k is an erosion constant.5). Along the downstream surface of the propellant in these motors. below which M has . The first proposed empirical burningrate formulas were of the form m = mo + k M . The twodimensional nature of the process initially forced semiempiricism upon theoretical descriptions of erosive burning. which was taken to be independent of M [ZS].4. mo is the mass burning rate in the absence of erosion [as given by one of the above models]. (45) where m is the local mass burning rate. The consequent increased convective heat transfer to the downstream portion of the charge increases the burning rate in the downstream region.8.258 Combustion of Solid Propellants 0 0 FIGURE 7. practical designs for solid propellant rocket motors usually employ laterally burning charges such as that illustrated in Figure 7. 7.m is not exactly a linear function of M . M is the local average mass flux of gas parallel to the surface (averaged over a cross section of the chambersee Figure 7. EROSIVE BURNING In all the models for the deflagration of solids discussed above. This phenomenon is termed erosive burning.
Schematic diagram of a laterally burning charge in a solid propellant rocket little or no effect on m [28]. and turbulent boundary layers are anticipated. therefore. [98]. the exponential factor in this equation is intended to account for reduction in this convective heat transfer through the “blowing” effect of mass transfer from the propellant surface.and a number of different flow regimes may therefore be anticipated. The prediction that the last term in equation (46) does not depend on m0 is in agreement with experiment [97]. Equations (45) and (46) are only two of many formulas that have been used to describe erosive burning [8]. Reynolds numbers based on M and on R or z in Figure 7. motor. to the burning rate underlies equation (46) [96] .8. and z. Erosive Burning 259 FIGURE 7. to a regime in which the Reynolds numbers are well above critical values for transition to turbulence. and the values of the constants k . The idea that turbulent convective heat transfer from the hot combustion products outside the boundary layer provides an additive contribution to the heat flux reaching the propellant surface and.M . A burningrate formula that tends to reproduce this threshold phenomenon is [96] m = mo + k l ( M 4 / 5 / ~ ’ / 5 ) e .7. which in turn increases the rate of heat transfer to the propellant and hence the burning rate [99]. Among these concepts is the idea that high external velocities produce a turbulent boundary layer (see Chapter 12) on the propellant surface and thereby effectively increase the thermal diffusivity of the gas. which becomes significant in upstream regions where m is of the same order as M .5 (0 Iz I 1) typically vary from 0 to lo8. Most of the formulas that have been suggested are based on physical concepts of influences of crossflow on propellant burning.5. ranging from negligible crossflow near z = 0. (46) where z is the axial distance from the upstream point at which M = 0 and the (undetermined) constants k . are independent of mO.k 2 m / M . Most of the theoretical . can be chosen in such a way that the function m(M) agrees with results of reasonably accurate experiments [28]. to a regime having Reynolds numbers appropriate for laminar boundary layers (see Chapter 12). and k . and k . to conditions of small crossflow in which boundarylayer ideas are inapplicable.
[1001[lo41 by refining or modifying the physical descriptions of boundarylayer flows with finiterate chemistry. In boundarylayer studies. In one study [1101. In addition. respectively.5) in obtaining conditions outside the boundary layer. frictionless. which are below the threshold for erosion and within which “negative erosion” often is observed (a small decrease in m with increasing M over a range of values of M ) [8].260 Combustion of Solid Propellants investigations emphasize the latter regimes rather than the conditions of small crossflow. in which M* = yp0/[2(y2 . a procedure that may be useful in accounting for rotationality in the core [1113. represents the value of M at chocking conditions (a Mach number of unity). in a constantarea channel with gas injection normal to the flow direction. Even in laminar boundarylayer flows. no work of this kind appears to have been undertaken. Although the most recent studies of erosive burning [102][110] have helped greatly to improve our understanding of the process. Fundamental theories of negative erosion might be pursued through perturbation analyses of the deflagration theories discussed in previous sections. with the ratio of the crossflow velocity to the normal velocity treated as a small parameter. these investigations usually provide graphical results rather than burningrate formulas. numerical integrations of the boundarylayer equations with finiterate chemistry included have been performed for laminar and turbulent flows with erosiveburning applications in mind [lOS][ 1091.parabolic partial differential equations of turbulent reacting flow were integrated numerically across the entire chamber without breaking the flow into a core region and a boundary layer. and $ = [l . Equation (47) is used along with the mass balance M = J : (P/A)mdz (where the perimeter P and crosssectional area A are given by 271R and nR2. the external flow may be approximated as a steady. the theoretical problems remain difficult. erosive burning of solid propellants presents one of the more challenging problems because simplifications achievable for nonpremixed combustion (Chapter .l)c. adiabatic flow of an ideal gas with constant specific heats. The main difficulties in analyses then arise in attempting to describe the boundarylayer flows. This flow may be shown to be described by algebraic equations [8] such as where y is the ratio of specific heats.(M/M*)2]1’2. the superscript 0 identifies core conditions at z = 0 (where M = 0). There have been many newer studies in which attempts have been made to improve burningrate formulas for erosive burning [8]. for the circular cylindrical grain of Figure 7.
1967. Jaumotte. A. 15. DtfSusion and Heat Transfer in Chemical Kinetics. F. Pokhil. Novozhilov. A. 310387. K. 12231233. A. 16yh Symp. and therefore the problem is fundamentally insoluble today. 11. Basic Characteristics of Combustion. 143 (1969). F. Moscow: Nauka. 1949. 18. S. B. and N. S. 1960. Belyaev. Vzr. 99143. 9. I. for example. Huggett. Bakhman and A. Williams. S. Chapters 1012. FrankKamenetskii. S . L. Nevertheless. B.. J. A. 317. N. Williams. Merzhanov. and M. Ryazantsev. Summerfield. Combustion of Heterogeneous Condensed Systems. Merzhanov and I. aCro. 1345 (1971). Lewis. 1 16. 3. and T.. AGARDograph No. New York: Elsevier Publishing Co. C & F 13. S . F. A. 1962.References 26 1 12) do not apply to the premixed solidpropellant system. “Combustion of Solid Propellants. Rocket Propulsion. in developing welljustified simplified approximations for laminarflow situations and in ascertaining suitable descriptions of turbulent flows. BenReuven. 4. AZAA Journal 9. C. Huang. Y. New York: Gordon and Breach. B. Chapter 5. Fiz. 8. and J. Akad. Sutton. G. Maltsev. M. 9. Fundamental Aspects o f Solid Propellant Rockels. Vichnevetsky. Kashporov. V. Khaikin and A. G. N. P. New York: Wiley. Barrere. H. Internal Ballistics of SolidFuel Rockets. P. I. 1969. CST 10. Vandenkerckhove. Gor. In view of these difficulties. M. Barrere. A. Chapters 46. 5. 2. 195 (1975). 3d ed. 10. Nauk SSSR 173. . 1969. and Y. P. England : Technivision Services. eds. which are of greater practical interest. Nonstationary Combustion of Solid Rocket Fuels. M. Chapter 4. Novikov. Fraeijs de Veubeje. 17. 190368. Vzr. Moutet. Pease. much more research remains to be done on the subject. 1382 (1967). M. vol. S . 13. 1956. Borovinskaya. N. I. REFERENCES 1. B. the extent of progress that has been made recently in understanding erosive burning is remarkable. N. S. G. G. 16. Rech. R. M. New York: Plenum Press. Princeton: Princeton University Press. Rocket Propulsion Elements. Slough. C. New York: McGrawHill. 14. 9. 4. 514574. Caveny.” in Combustion Processes. 6. Moscow: Chimia. R. and H. R. Martem’yanova. V. 269317. I. B. Guirao and F. Khudyaev. (1976). Wimpress. Filonenko. Dokl.469 (1968). M. Barrere and H. A. Khaikin. 169 (1973). Chemical Rocket Propulsion and Combustion Research. (1973). C. I1 of High Speed Aerodynamics and Jet Propulsion. M. D. 1963. Taylor. 7. Fiz. Maltsev. 31 (1956). Moscow: Nauka. Merzhanov. Gor. Penner. 12. In the turbulent boundarylayer flows. and L. appreciable uncertainty remains concerning the proper partial differential equations that should be used to describe premixed combustion (see Chapter lo). 1950. 1977.
S. A. Dubovitsky. R. M. AZAA Journal 1. 32. E. 404 (1964). I. N. 34. AZAA Journal 2. A. Kubota. 2311 (1962). Dokl. H. SO.863 (1950). H. Pearson. 46. M. Johnson and W. R. and S. Sutherland. 26. 57. 38. H. Phillips. Van de Mark. 1439 (1960). D. Penner. J. Fiz. and F. 29. 73. B. S. 1737 (1967). R. 29. Caveny. G. 11631171. Ohlemiller. W. (1955). Khaikin. S. Williams. Int J. Williams. Buckmaster. K. I. J. 48. Vzr. and G. 27. Williams. J. Zhur. A. D. F. M. 71 (1969). Filonenko. G. 55. I. M. W. Jacobs. Dokl. H. 678689. P. 5th Symp. Soc. Williams. Merzhanov and F. Wilfong. 37. K. Wise. D. 135. 1328 (1973). Fiz. Maksimov. K . Znstr. 345357. Chem. Schultz. Soc. Theory o f Laminar Flames. (1979). V. Phys. J. Cantrell. Sibbett. W. A. 406 (1964). D. W. Bircumshaw and T. Ibiricu and F. Faraday Soc. Chem. Inami. (1962). Nauk SSSR 152. Z . Filonenko. (1978). R. Chem.262 Combustion of Solid Propellants R. B. H. 17th Symp. 22. phys. E. L. Spingler. I. 21. 67. 28. E. 47. 1989 (1970). S. S. 39. Fiz Gor. A. 0. A. D. Nachbar. 44. M. 51. New York: Pergamon Press. P. Rosen. G. 37. 1629 (1966). 72. Schultz and A. 591 (1968). 776 (1972). M. A. D. R. L. 4302 (1969). Chem. 35. Phys. 1544 (1963). W. Chem. B52. Jacobs and A. Williams. K. L. Merzhanov. AIAA Journal 2. J. Y. Chem. Niioka and F. RussellJones. Ludford. Maksimov and A. Kapila. Spalding.202 (1968). R. L. 203 (1968). Dekker. 14351441. Summerfield. 153 (1959). Penner. K.. AZAA Journal 10. 367420. K. 45. Phys. Ludford. Chaiken. 32. Zel’dovich. Nachbar and F. Chem. 17th Symp. 1709 (1974). 12. JR. M. 279 (1977). Schultz and A. 447456. P. 63. . T. 2. RussellJones. G. 1. Khudyaev. M. Trans. Combustion and Propulsion. Galwey and P. 25. A. H. Mohan and F. 5031 (1960). D. 4. J. H. Vzr. R. S. Dekker. 8th Symp. 41. G. 419 (1969). A. 1077 (1963). Akad. 90 (1942). S. C. C & F 13. Kubota. Wheeler. AIAA Journal 11. Eksp. and M. A. Hermance. AZAA Journal 8. (196?). Fiz. J .143 (1963). D. E. Guirao and F. 30. Phys. E. Williams. 4. 24. 40. S. W. A. 52. 54. 54. Rev. Acta Astronautica 4. Andersen.. and A. Buckmaster and G. 23. 33. 1958. 185 (1975). Acta Astronautica 3. 53. T. A. J. and A. and H. A. 837 (1959). Lengelle. Powling. S. Williams. Gor. 11th Symp. Chem. C & F 13. 0. 593 (1976). Davies. 20. Phys. Jacobs and J. 43. AZAA Journal 12. G. Akad. S. 89 (1960). W. 42. Rosser. Jacobs and G. 9th Symp. Vzr. (1963). F. 36. C. Third AGARD Colloquium. 49. Heat Mass Transfer 8.. E. J. 59 (1960). Kapila and G. F. 19. Barsh et al. and S. 58. N. Phys. C & F4. 1982. 56. W. Cantrell. C & F 24. Chem. J .26026?. 2133 (1955). Cambridge: Cambridge University Press. Sci. J . W. 47 (1966). Chem. 31. 575 (1965). P. 59. AZAA Journal 4. 498 (1942). J. Nauk SSSR 129. Merzhanov and A. Jacobs. 392 (1958). Phys. Teor. S. K. B. Daniels. J. A. 23. J. J . Green. Powling. Gor. 4741 (1957). Ludford.
Williams and G. 175185. Barrere and L. Parfenov. Novozhilov. Williams. R. Khim. Fiz. Novikov. Moe. London 213A. Fiz. AZAA Journal 8. Belyaev and S. Paris: GauthierVallars et Fils. N. M. Soc. Nachbar and G. A. Jr. Dokl. S.55. Cohen. Zakharov and L. Bowden. M. and M. AZAA Journal 20. 277 (1980). 66. Vieille. Mekh. R. Handley. Tabak. R. 128 (1968). Price. A. S. and R. Pokhil. “The Effects of Nonstoichiometric Composition and Particle Size on the Burning Rates of Composite Solid Propellants. 1960. N. K. L. A. W. 63. 143 (1970). A.” in Solid Propellant Rocket Research. Strand. V. Price. Astronautica Acta 14. F. Noble and R. Tereshkin. Korotkov. 61. Leipunskii. E. N. 65. AZAA Journal 14. Bakhman. Zhur. 75. Explosirstoff 2. Roy. H. Bills. A. Glick. 62. (1971). A. 14 (1972). 2200 (1970). New York: Academic Press. Vzr. A. 207226. Derr. “The Burning Mechanism of Ammonium PerchlorateBased Composite Solid Propellants. London 165A. 37. F. 68. C & F 3. Dokl. J. A. M. Stang. K. Webb. J. Cohen and L. B. 69. Beckstead. 141182. 3.. 824 (1959). S. 380 (1981). 384 (1974). Chapter 11. W. Summerfield ed. J. N. M. Klyachko. W. N. Sutherland. Blackwood and F. Tekhn. Fiz. 85. ed.222 1 (1959). Sukanov. Strahle. P. Bakhman and V.References 263 60. B. S. W. H. 1739 (1982). A. and A. W. no. E. C. Prikl. J. C. F. 80.” in Solid Propellant Rocket Research. C & F 12. M. Trans. P. Zhur. Dokl. A. 1631 (1976). 201 (1968). AF 49(638)412. S. 72. AZAA Journal 19. 86. Maznev. G. Summerfield. Phil. Lockheed Missiles and Space Company. 82. Glazkova and I. 67. Roy. 83. 77. 1400 (1960). 15 (1964). Sigman. Nauk SSSR 131. 256391. W. Rech.. vol. 88. C. 64. and C . Fenn. AZAA Journal 16. 1 of Progress in Astronautics and Rocketry. Mishuck. I. F. D. Faraday SOC. 87. Beckstead. F. 0. 73. 8. Ghosh. Summerfield. Palo Alto (1962). Trans. W. D. A. J. 637 (1976). 90. 35. Khim. 177 (1960). 98). 10471056. P. Nobel Hejte 2. 76. (1981). 1960. Akad. Schultz. Braunschweig (1962). A. 70. Hall. Note 364621. 78. Zhur. 843 (1978). a h o . D. Sulimov. Librovich. 1893. and L. and A.887 (1960). vol. 84. Nachbar. Steinz. Ryantsev. 13th Symp. A. F. L. 97 (1969). Gor. G. Nauk SSSR 129. Y. C. “Burning Mechanism of Ammonium Perchlorate Propellants. R. 71. P. Cline. 41 (1968). Panyam. New York: Academic Press. V. Abel. Fiz. 17 (No.” AIAA Paper No. Memorial des Poudres et Salpetres 6 . L. Andersen. Belyaev. 18th Symp.” Tech. presented at Fifth AGARD Combustion and Propulsion Colloquium. Nadaud. W. Khim. 68658 (1968).. 8 1. Nauk SSSR 131. Fiz. M. S. and K. F. Weingarten. 1 of Progress in Astronautics and Rocketry. Akad. 14. N. 89.. Akad. M. 15 (1972). I. C & F 15. R.49 (1875). 150 (1963). “A Theoretical Study of the Burning of a solid Propellant Sandwich. R. 640 (1975). B. 1622 (1961). J. Contract No. M. 285 (1952). AZAA Journal 13. Soc. Zhur. 34. K. M. B. Hintze. Campbell and G. Strahle. . 74. Summerfield. 79. W. AZAA Journal 18. P. W. 301 (1959). AZAA Journal 12. Lengellk. Proc.
D. Torda. J . 17 (1979). Blatz. AIAA Journal 18. V. P. Journal o f Spacecraji and Rockets 15. Fiz. Borisov. S. 663667. 105 (1978). M. 97. AIAA Journal 21. 110. 599 (1958). 1225 (1979). Lake. A. Khasainov. 9th Symp. L. (1963). 107. 100. Schuyler and T. A. Kuo. 1128 (1975). AIAA Journal 17. and B. R. A. AIAA Journal 13. 94. Vandenkerckhove. 635 (1947). 106. Robillard. A R S Journal. H. H. P. Marklund and A.49 (1974). London : Butterworths Scientific Publications. 102. 109. (1957). 108. 30. K. 289339. Beddini. 1954. Jet Propulsion 28. Summerfield. Glick. Zwarts. T. 122 (1982). 384393. Tsuji. B. Energy Combust. Lengelle. Beddini. 2171 (1966). Corner. 156 (1983). Gokhale and H. K. M. 173 (1960).264 Combustion of Solid Propellants 91. P. 111. G. Ermolayev. CST 18. Khim. 98. L. J. 898 (1978). Razdan and K. Fundamentals and Aeronautical Applications. 95. 745752. F. 92. J . S. R. 101. K. Zhur. Prog. 36. 8th Symp. 8. 315 (1975). 103. A. 93. R. Mukunda. AIAA Journal 4. King. 6th Symp. M. K. 99. A. Razdan and K. 96. AGARD. Lenoir and G. AIAA Journal 17. Sci. Gough and F. Geckler.43. 1803 (1962). AIAA Journal 12. 105. Novozhilov.1345 (1980). Kuo and M. . S. B. K . Kuo. AIAA Journal 13. 104. Trans. 139 (1978). J . 1 (1982). S. S. Faraday SOC. Selected Combustion Problems. J . AIAA Journal 16. Krier. AIAA Journal 20. A. M. R. (1962). K.
and flammability limits appear to be highly diverse phenomena. Extinction. Therefore. to hot inert gases. the presentation to be given here will be less precise. to a small pilot flame. Therefore. they share many common elements. or to an electrical spark discharge. It has long been of practical interest from numerous viewpoints. General principles are difficult to extract from detailed analyses of specific problems. These methods have been listed here roughly 265 . and guns. to an electrically heated wire. extinction. motors. Methods for achieving ignition include exposure of the combustible to a sufficiently hot surface.CHAPTER 8 Ignition. in an effort to expose general ideas. Description of these phenomena by theories as rigorous as those presented in Chapters 26 is a formidable task that can be completed only by treating each problem individually. it is instructive to treat them from a relatively common viewpoint. These phenomena are related in one way or another to processes of ignition or extinctionprocesses that generally are multidimensional and timedependent. to an explosive charge. Ignition is the process whereby a material capable of reacting exothermically is brought to a state of rapid combustion. ranging from the desire to prevent unwanted explosions in transportation and storage of combustibles to the need to initiate controlled combustion in furnaces. Although ignition. and Flammability Limits A number of classical combustion phenomena are fundamentally more complex than the processes that have been considered thus far. to a radiant energy source. Analyses will be restricted to the simplest illustrative models.
There are many situations in which analyses of these phenomena reasonably may employ onestep. and Flammability Limits in the order of increasing temperatures of the ignition sources.2. we exclude the consideration of special effects associated with finer details of chemical kinetics. Extinction. by explosives in large productionwell fires). [2]. for example. more emphasis has been placed on ascertaining critical conditions for extinction than on extinction times. Flammability limits are limits of composition or pressure beyond which a fueloxidizer mixture cannot be made to burn. Meanings of these tabulations and their relationships to ignition and extinction phenomena will be considered here in Section 8. Seldom ever can energy conservation be neglected in addressing . Since processes of flame extinction usually occur relatively rapidly if at all. removing an essential reactant from the system (an approach especially suitable for diffusion flames). strong sources ignite quickly. In so doing. Adoption of the energybased viewpoint may be motivated by the fact that energetic aspects practically always are relevant to the phenomena to be studied. For this reason.5). extensive tabulations of limits of flammability have been prepared [l]. General strategies for extinguishment may be listed as cooling. Rates of chemical reactions always have a bearing on ignition.3. Extinction (quenching) of deflagrations or of diffusion flames may be achieved in various ways. attention here is restricted to situations in which ignition results in a deflagration or a diffusion flame. and flammability limits. flame removal. or physically removing the flame from the reactant mixture by inducing high gas velocities (blowing the flame away. In extreme cases of strong sources. reactant removal. We shall focus our attention here on results explained by energyconservation requirements and heat losses. as accomplished. but weaker sources require longer exposures. Arrhenius approximations to the rates. quenching.266 Ignition. and flammability limits. which serve to correlate a number of different observed characteristics of ignition. including passing the flame through tubes of small diameter (a design principle of flame arrestors). They are of practical interest especially in connection with safety considerations because mixtures outside the limits of flammability can be handled without concern about ignition. and chemical inhibition. practical methods may be placed in one or more of these four categories. There tends to be an inverse relationship between the temperature or strength of the ignition source and the time of exposure required for ignition. Extinction phenomena are of interest in problems of safety and of fire suppression as well as in various aspects of control of combustion processes in industry and propulsion. such as radical diffusion or surface reactions. adding sufficiently large quantities of a material (such as water or a chemical fire extinguishant) that slows combustion. detonation can be initiated directly (see Section 6. This fact enables common theories to be developed on the basis of energy considerations. extinction.
to quenching distances (see Section 8. Hertzberg.Ignition. H . the chemical fire suppressant CF. It will be seen in Section 8. corresponding ideas were introduced in studying the simplest problem related to ignition theory.4. but rather involve questions of which of the two is of greater importance. and in Sections 3. the more complicated the analyses tend to become. If brominecontaining species (for example.4. References [3][5] are earlier books that present the material from viewpoints different from that of Sections 8. . Vapors and Dusts: Theory and Experiment. V. Waterloo.3 and that discuss the relevant experiments in greater detail. Current uncertainties lie not at the level of asking whether energetic aspects or changes in chemicalkinetic mechanisms govern the phenomena. Lovachev. With the onestep. activationenergy asymptotics (Section 5. S . Here the method will be used first (Section 8. under appropriate conditions ignition of a fuelair mixture by laser radiation is appreciably enhanced if the frequency of the light is suitable for dissociation of 0. Canada: University of Waterloo Press.3. activationenergy asymptotics were applied to the problem of diffusionflame extinction.5.4. V. A.2. V. N. + Br removes the very active H atom. and to flammability limits.4. A .2) for analyzing premixedflame propagation with homogeneously distributed heat losses and next (Section 8. the remaining influences may be attributed to particular aspects of chemical kinetics. . the problem of homogeneous thermal explosions. Guirao eds. V’yun. N. The considerations in Sections 8. The analysis to be given in Section 8. Arrhenius approximation to the chemistry.3 that the greater the degree of heterogeneity of the system. replacing it with the less active Br atom. 348.6) afford a useful vehicle for analyzing ignition and extinction phenomena.1 with some rough physical ideas and observations that serve to define terms and to relate various ignition and extinction phenomena.1 and 8. S. 259 (1973) and by M . Krivulin. C & F 2 0 . There are many situations in which chemicalkinetic details have been established as significant. Babkin. after their removal.Br) are added to fuelair flames. Bunev. and Flammability Limits 267 topics in combustion. J. Baratov. Extinction.l). In wellplanned studies of influences of specific aspects of chemicalkinetic mechanisms. We begin in Section 8. For example.” in FuelAir Explosions. In Section B. and A .2 will be seen to be relevant to premixedflame extinction. 1982. Although lists may be prepared of examples like these that illustrate the significance of specific chemicalkinetic mechanisms. M.3) in discussing a variety of ignition problems. mechanisms of reaction of oxygen atoms are important in this ignition process. we shall not discuss them further. Lee and C.* * Morerecent reviews of flammability limits are given by L. then extinction is promoted at least partially because the step H + HBr + H. A. V.3 and 3. the energetic effects are taken into account .2 and 8.2 are restricted to homogeneous mixtures of premixed combustibles. thereby reducing the overall rate of heat release. “The Flammability Limits of Gases..
equation (1) reduces to the statement that the minimum ignition energy per unit area is where . A formula expressing the rule is where His the minimum ignition energy. po is the initial density of the mixture. A is the crosssectional area of the slab. onedimensional. Ignition will occur only if enough energy is added to the gas to heat a slab about as thick as a steadily propagating adiabatic laminar Jame to the adiabaticpame temperature. it is found that the amount of energy (per unit surface area of the slab) added to the gas must exceed a definite minimum value for ignition to occur. Although experiments have been performed in a number of different configurations . Analysis (for example. ignition by means of a hot slab of gas represents one of the simplest conceivable ignition problems. 1is the average thermal conductivity of the gas and u. 6 is the (adiabatic) laminar flame thickness.. Arrhenius kinetics) that for slabs thinner than a critical thickness. cp is the average specific heat at constant pressure. [ 6 ] ) by numerical integration for slabs of various thickness. This ignition criterion is immanently reasonable. possibly. Extinction. and Tois the initial temperature of the mixture.268 Ignition. is the (adiabatic) laminar burning velocity. other effects). initially raised to the adiabatic flame temperature. discrepancies exceeding a factor of 2 or 3 will not be surprising. Flames will not propagate through very narrow channels because of heat losses to the walls (and. The rule also clearly represents an oversimplified approximation . a propagating laminar flame develops. that the rate of liberation of heat by chemical reactions inside the slab must approximately balance the rate of heat loss from the slab by thermal conduction. the temperature decays to the ambient temperature by heat conduction without producing ignition. since the plane. A minimum ignition energy may be obtained explicitly from equation (2) by identifying a minimum crosssectional area A for the slab. Theoretically. with onestep. It can also be shown to be roughly equivalent to a number of other reasonable physical statementsfor example. The minimum value of A has often been related to the results of flamequenching experiments. and Flammability Limits 8. while for thicker slabs. These results and those of various related experimental and theoretical studies (such as [4][8]) can be summarized by the following approximate rule. T. shows (for instance. timedependent conservation equations describe the process. is the adiabatic flame temperature. When use is made of equation (54) for 6. MINIMUM IGNITION ENERGIES AND QUENCHING DISTANCES When a combustible mixture is ignited by heating a plane slab of the gas.1.
The results that have been given here for the minimum ignition energy are subject to various chemicalkinetic restrictions related to the nature of the ignition source. When this condition is used in 2 ) . the results of these heat loss studies agree with other heat loss theories [16]. [5]. For systems in which heat losses control the quenching process.To)/c. 2 2 3 In view of equation (54). Equations (54) and (4) yield d = aA/. provided that the slab area A is greater than the area of a square with sides of length equal to the quenching distance (that is. A 2 d’). the minimum ignition energy is found to be equation ( H = d2A(T. alternative forms of equation (6) are H =a ’ (6) (7) 6A2(Tm . The chemistry involved in uo. [9][15].pou~ and H = ~ ~ 6 ~ p . Equation (3) will determine the minimum ignition energy in terms of fundamental properties of the system only if an expression for the quenching distance d can be found.To)/c. . a popular procedure is to confine the combustible mixture between parallel flat plates and to define the quenching distance d as the minimum plate separation for which flame propagation can be achieved. [17] and with experiment [4]. ~ (8) Equation (6) correctly predicts the approximate temperature and pressure dependences of the minimum ignition energy. The same chemical mechanism therefore must be involved in H if correlations like those of equations (3) or (6)(8) are to be obiained. [17] in predicting that [see equation (30)] d = ah. If the ignition source selectively enhances particular steps in . and the thickness 6 of the (adiabatic) flame is given by equation (54). It is reasonable to assume that multidimensional effects will be small.(T To).8.pouo. d is determined by the factors discussed in Section 8.To)/u0. and d evidently is that of a propagating deflagration. 6. [16]. Minimum Ignition Energies and Quenching Distances 269 [4].2. (3) The approximate correlation between quenching distance and minimum ignition energy expressed by equation (3) is (roughly) substantiated experimentally [4]. When the heat loss is assumed to occur through thermal conduction.1. c ./c. and therefore that the rule expressed by equation (1) will be valid.po 00 3 (5) whence equation (3) becomes H = a2A3(Tm. (4) where a(approximate1y 40) is a multiplicative factor that is not strongly dependent on the properties of the system.
From the viewpoint of chemical kinetics. [18][29]. or if it produces temperatures much higher or much lower than the adiabatic flame temperature. by turbulence [20]. For example. Extinction. it is known that there is a critical electrode spacing giving a minimum value of H [4]. [22]. when the spacing is less than the quenching distance. [26]. [25]. if it involves surface reactions. [28] also have been devised. Among the ignition procedures identified in the previous section (see also [l8] for a classification). This last effect can be included approximately in equation (7) by replacing a 6 there by the actual electrode spacing [20]. electricalwire. [24]. [25]. [24]. this does not seem to exert a significant influence on the chemical kinetics of flame development. pilot flames are ideally suited to the approach (because they provide temperatures and compositions in the range assumed). The slab geometry that has been considered here is not representative of most sparkignition arrangements. Emmons. the hotsurface. A considerable amount of attention has been devoted to studies of processes of spark ignition [4]. and Flammability Limits 3 the kinetic mechanism. although the initial temperatures produce ions and radicals and are much higher than the adiabatic flame temperature.270 Ignition. W. Simple formulas to account for increases in the ignition energy by convective gas flow past the electrodes [ZO]. personal communication. toroidal regions of hot gas influenced by buoyancy have been observed*). [28]. Most of the experiments on which the correlations are based are sparkignition experiments. [lS]. radiationsource. Approximate analyses of spherical and cylindrical ignition sources [191[21] produce expressions that differ from equation (6) at most by a roughly constant multiplicative factor that can be absorbed into a’. the spark may heat a cylindrical region of length equal to the electrode spacing. [23]. and by the presence of liquid fuel droplets [27]. the required energy for ignition may exceed the minimum ignition energy appreciably because of heatconduction losses to the electrodes. Sparks add measurable energy rapidly enough for their effects to be characterized in terms of an H . [23][27]. . Another factor influencing the minimum ignition energy for spark ignition is the spark duration [20]. There is an optimum * H. and hotgas techniques often produce temperatures below the adiabatic flame temperature by amounts sufficient to raise doubts concerning the use of equations (6)(8). [22]. then the basis of the correlations becomes questionable. thereby increasing the energy requirement as the spacing is increased. but experimentally it is difficult to introduce a pilot flame rapidly enough into a combustible mixture to avoid nonhomogeneities and associated heat losses that have not been taken into account here. [28]. and many facts have been uncovered concerning the mechanism. spherical or cylindrical ignition kernels are more realistic. although the actual shape of the region of elevated temperature may be quite complex (for example. but when the spark gap exceeds the quenching distance.
physical nature and leaves open questions. Later work involved numerical integrations [I411. 8. [35]. why is the value of a so large? These questions may be addressed by detailed analyses that begin with the conservation equations in differential form.2. and ignition analyses will be considered in Section 8. For example. and especially for heated surfaces. Premixed Flames with Heat Losses 27 1 duration for minimum energy. generated by the spark. . electrically heated wires. the energy that must be supplied by the ignition source increases because of energy lost from the ignition region by conduction and radiation prior to ignition [30]. To avoid the complexities of multidimensional analyses.1. hot gases. Analyses of this type are developed in the following sections. As the heating time is increased. This fact was recognized in early work [4]. steadyflow model. In rare instances for spark or glow discharges of very long duration. An expression for a will be derived in Section 8. 8.2. but a significant amount of energy is also lost in shock waves. it is of interest to develop the theory of flame propagation with heat losses.2.8. typically on the order of 1 mJ for stoichiometric gaseous hydrocarbonair mixtures at atmospheric pressure.[46]. PREMIXED FLAMES WITH HEAT LOSSES To address topics of flame quenching and of flammability limits more thoroughly. more recently. [9]. this last loss increases as the spark duration decreases because of a consequent increase in shock strength.3. activationenergy asymptotics [44] .2. Methods of analysis The earliest studies of heatloss effects in premixed flames were based on analytical approximations to the solution of the equation for energy conservation [35][39]. [36]. Thus the minimum value of the minimum sparkignition energy. Two such approximations that have been sufficiently popular to be presented in books are those of Spalding (see [40]) and of von Karmhn (see [S]). The material that has been presented in this section is of an approximate. the consequences of heat losses will be explored here within the framework of a onedimensional. heat losses to the surroundings of the flame practically always are important. critical conditions for ignition are characterized better by an ignition temperature [IS] (the source temperature needed to produce ignition) than by an ignition energy because losses through heat conduction and through igniterinduced fluid flow are substantial. typically on the order of 50 ps.[43] and. is attained at particular values of the electrode spacing and of the spark duration. [32][34] and was used in predicting limit conditions [17]. or pilot flames. we indicated that although changes in chemical kinetics may influence limit phenomena. In the introduction to this chapter. that propagate away quickly without contributing to ignition [31].
cpx/A and obtain the equation d 2 z / d t 2 . Under other conditions./.272 Ignition. Under certain conditions involving nonvanishing heat losses. it will be found that no solution for the burning velocity exists.3.6.A. activationenergy asymptotics have been applied to the adiabatic version of equation (9) for a particular rate function w .To)/(qoYF0). heat production. which is assumed to be determined by the equation (9) where the flame is at rest in the adopted coordinate system. onedimensional. Since equation (9) is the only differential equation that will be used in the initial discussion.3.3. x is the coordinate normal to the flame. and heat loss in an element of thickness dx. m is the constant mass flow rate per unit area in the + x direction. The first objective of the following analysis will be to compute the burning velocity for a simplified model of a plane. qo is the heat released in the chemical reaction per unit mass of fuel converted.L d z / d t = . In Section 5.w A d 2 T / d x 2 .. w is the chemical reaction rate (mass of fuel converted per unit volume per second). We shall focus our attention on the steadystate temperature distribution.mep dTfdx = wqo + L. c p is a constant average specific heat at constant pressure. one of which presumably corresponds to an unstable flame. and the initial mass fraction Y. the approach may be described as a thermal theory of flame propagation.6 for this rate function and in Section 5. A is a constant mean thermal conductivity. Extinction.. the nondimensional reaction rate Awl( Y F o m ~ chas p ) been written as the product of a burningrate eigenvalue . Thus. the burning rate is known on the basis of these results and to employ the known adiabatic mass burning rate ma for the purpose of nondimensionalization. in analogy with equation (518). we introduce the nondimensional streamwise coordinate 5 = m. it expresses a balance among heat conduction. and L is the heat loss per unit volume per second by radiation or conduction to the walls. Here we first address a model problem from the viewpoint of asymptotics. of fuel as z = c. two burning velocities will be obtained. and Flammability Limits Current computational capabilities enable twodimensional formulation to be solved numerically [47]. + K. Equation (9) follows from equation (143) and the onedimensional form of equation (140) (with A = constant and c p = constant) provided that the phenomenological heatloss term L is inserted. if multiplied by dx. nonadiabatic flame. (10) (1 1) where the burningvelocity ratio is I* = m/m. burningrate formulas are given in Section 5. thermalenthalpy convection. the nondimensional temperature z has been defined in terms of the initial temperature To of the combustible mixture. Here it is convenient to presume that for L = 0.7 for others. thus indicating the presence of flammability limits.(T .
co and the other to place z = T . As exemplified by equation (566). Downstream from the reaction zone only T constant would be consistent with a convectivediffusivebalance. the reactionrate term in equation (10) will be exponentially small. where one constant has been selected to satisfy T = 0 at = . Under the assumption that all the fuel is consumed in the reaction zone. times a suitable power of b held fixed [see. an arbitrary selection of the origin of the coordinate system. (12) The expansion parameter is the Zel'dovich number (13) = To + qoYFo/c. for example] with a nondimensional function o(T). The limit p . therefore. co is considered with A. Ro is the where universal gas constant. for example.is a critical factor in the structure of the nonadiabatic flame. at ~ = 0. the expression above equation (575)]. a contraction of the coordinate is suggested by equation (lo). In this case. and E represents the overall activation energy.T f ) = IpmK d< . Premixed Flames with Heat Losses 273 A. Use of equations (10) and (1 1) in this expression gives cf P = Jwlf .2. where f identifies conditions just downstream from the reaction zone. of the adiabatic problem [defined by equation (545) for the adiabatic problem. The heatloss term is analogous to that appearing in equation (B52) and is assumed to be nonnegative and to vanish at the initial temperature (T = 0). we may readily derive the expression my. with IC of order b'. It is found that if ~ ( z )is of order unity or larger over a range of of order unity. The heatloss distribution. such that ( = </p is treated as being of order unity. (14) as a nondimensional representation of the influences of heat losses on the decrease in flame temperature. the function W ( T ) depends exponentially on P.8. and the nondimensional heatloss rate is K(T) = kL/(qoYFO mzC. and in the first approximation the solution < < .. upstream from the4reaction zone in equation (10) there is a convectivediffusive balance with negligible loss in the first approximation. and the approximate solution < z = zfe": (15) is obtained. and for large fl we expect that there will be a narrow reaction zone centered about the position of maximum T . at first let us assume that ~ ( z ) order P' everywhere. = f?fm w dx from the fuelconservation equation. Outside this reaction zone. ~ ( z ) .(dT/dO.). is the adiabatic flame temperature. then flamestructure solutions do not exist is of with the present formulation.TO)/(ROTzf)? Tf P(1 .
ReactiveI diffusive diffusive I zone zone I Convectiveloss zone FIGURE 8. has been applied. dy/dy + 0 as y + 0 to leading order.To. The calculation of . The threezone structure of the nonadiabatic flame is shown schematically in Figure 8.a(l . The effect of diffusion is a perturbation of first order in the convectiveloss zone. is obtained. To achieve agreement with the nonadiabatic version of the theory whose adiabatic reactionrate function is given by equation (566).214 Ignition..z)]}. we may put o = (zf . If. we obtain from equation (10) (dy/dq)2 = (2Aa/P) SI. (19) .z ) / z J . a convectiveloss balance exists in equation (10) in the large downstream region. Arrhenius kinetics. and Flammability Limits Convective.an exponential decay of temperature over a length scale having 5 or order B.1. L is proportional to T . then equation (16) explicitly gives z = T f e . which is expected to exhibit a reactivediffusive balance in the first approximation. Schematic representation of the temperature distribution through a flame with heat losses.z)exp{p(l .C U ( p B ) . which therefore may be called a convectiveloss zone. It thus becomes clear that in the first approximation.z)/[1 . 0 dy. for example. A matching of this solution to equation (1 5 ) can be written as The form of the reactionrate function is germane to equation (18). Extinction.u entails investigating the structure of the reaction zone. (17) in the first approximation where a matching condition implied by equation (16). with y~ = b< being the relevant nondimensional space coordinate of order unity. the nonadiabatic form for a thermal theory possesses a greater degree of arbitrariness than does that for a complete theory with onestep. Just as equation (571) was derived but with y = P(zJ .1. so that from equation (12) K can be written as (c/P)z with c being a constant or order unity.
tworeactant systems with reaction orders different from unity). Equation (14) may be used in equation (20) to provide an expression for p in terms of a heatloss parameter. The burningvelocity ratio as a function of the heatloss parameter for nonadiabatic flames.UP*. then with L approximated as being linear in T. we find from equations (12) and (14) by use of equations (15) and (16) that Tf = Zf  2A. is shown in Figure 8. The result. giving an upper and a lower burning velocity. for 1 < l/e there are two values of p for each 1. which enables 1 to be calculated explicitly as a function of p. equation (20) may be written as p 2 = .4  I 0 0 . to the lowest order equation (18) then gives (with A.T o ) / K f and .4 1 FIGURE 8. 1 0. at 1 = l/eand that there are no solutions for 1 > l/e. If the dimensional flame temperature.3 0. From equation (16) it may be seen that (dzldt). (20) Through investigation of complete theories with various other onestep Arrhenius rate functions (for example. plotted in the form of as a function of I. (23) .L(Tf)/(m"cf).8. = P2/2) p 2 = eD(l7f). since equation (14) implies that 1 .2 0. = ic(zf)/p. it may be shown that equation (20) always follows from equation (18) to the present order of approximation. Tf = To + zfqoYFo/cp.2. whence by use of equation (15) in the integral in equation (14) it may be shown that with the definition which represents a loss parameter of order unity. where CI = (Kf.2. These are the results that were identified at the beginning of this section. (22) The solution to equation (22) for 1 is 1 = p 2 ln(l/p2).2. Premixed Flames with Heat Losses 215 0. It is seen that dl/dp = 0 at p = l/& that is.zf is of order P'. is introduced.
For example. instead of T.) is the loss function L evaluated at T = T. To the order to which equation (20) has been derived. by use of equation (518) it is readily possible to account for a variable thermal conductivity 1. in equation (23). large values of p may be approximated qualitatively by a large value of m or by zi near unity (for example. and the extinction condition is found from equations (12) and (21) to be where equation (13) has been employed in the last equality. In these theories. is not needed in applications having representative accuracies of 10 %. Pursuit of this more general analysis for problems retaining the full speciesconservation equations for onestep. Arrhenius processes reveals that solutions may be obtained when IC is of order unity over a range of 4 of order unity.4. If fi is sufficiently large. then the resulting extinction conditions reduce to those derived herein. may be applied more directly with more readily available rate parameters (for example. such as equations (24) and (25). The development that has been presented here is not the most general asymptotic analysis. zi = 0. this permits orderunity departures of T . with its greater complexity.q in equation (396)]. The doublevalued solution arises through the influence of m on T.1. It appears that generally fi is large enough that the more general analysis. .276 Ignition.the result is that A is to be evaluated at 7 ’ ’in the formulas that have been given. and Flammability Limits where L(T. and extinction occurs before the rate of heat loss becomes large enough to impose an appreciable enthalpy gradient within the thin reaction zone.z)”zm with m and n constants. Spalding took w = (1 . Extinction. while von Karma. in defining p . where ziis a nondimensional ignition temperature.. A generalization of greater fundamental significance would be to formulate the problem employing m instead of m.q2 by cp . there being a nonnegligible enthalpy gradient within the reaction zone. in forming the nondimensional coordinate 5 and T. E ) . the reactionzone equations then differ from those typified by equation (570) and resemble those of the premixedflame regime discussed in Section 3.4 [replace cpz .. it may equally well be written as which shows directly that the reduction in the burning rate is due entirely to the influence of the reduced flame temperature on the Arrhenius factor in the reactionrate expression. The rate functions considered in early analyses that did not employ asymptotics may be contrasted with equation (19). & from unity and exponentially small ratios m/m. put w constant over a finite range of 5 for which zi < z < 1 and zero otherwise. The asymptotic analysis possesses the advantage that the resulting formulas.9).. The results thereby obtained are quite similar to those derived here. it also reveals the structural aspects summarized in Figure 8.
at least for sufficiently small values of I. [48]. If the reactant concentration is . Use of the more general asymptotic approach identified in the preceding subsection can provide results that extend more accurately to smaller values of p. with 1 = l/e providing a correct extinction criterion (within about 10 % accuracy since corrections of order 1 / j may be anticipated).. it seems quite likely that the entire upper branch is stable and that the practical extinction condition therefore can be identified with I = lje. [49].2..2. then the upper branch of Figure 8. 8.2. having trnfc.3.8. The foundations of the analysis producing Figure 8. Whether both solutions may be observed experimentally depends on the stability of the steady state. Flame stability is the subject of the following chapter. Experimentally. in which the operator controls both the massflow rate and the rate of heat loss from a plane flame located above a cooled porous plate.2. However.2. Static stability arguments have suggested that the upper branch is stable and the lower unstable [36]. The stability is likely to be influenced by the experimental configuration and by parameters not appearing in equation (22). and it is the upper branch that generally is ascribed physical meaning and along which the flame possesses its usual structure. Premixed Flames with Heat Losses 217 8. it should be emphasized here that the complete stability question is complex. It seems sufficient here to reemphasize that if special types of flame instabilities are not observed in an experiment. this refinement does not appear to warrant development..p) of order unity (where t denotes time and p o the initial density).. At least for a onereactant mixture with a Lewis number of unity. from equation (22).2 fail on the lower branch when p becomes sufficiently small. there are diffusivethermal instabilities that are influenced by heat loss and that have led to various statements concerning the dependence of the stability in Figure 8.2 on the Lewis number [46]. We shall postpone any further discussion of these stability questions until the following chapter. a small portion of the lower branch in the vicinity of the maximum value of 1 apparently has been stabilized [36]. In view of the established difficulty in obtaining stable flames on the lower branch. clearly shows the existence of steadystate solutions for two different flame speeds if 0 < 1 < l/e./(Lp. the upper branch appears stable. Concentration limits of flammability The dependence of the calculated flame speed on the initial reactant concentrations arises primarily through the influence of the concentrations on the adiabatic flame temperature T. also suggest stability only for the upper branch [46]. The existence of two flame speeds Figure 8. For Lewis numbers different from unity.2 may be expected to apply. Analyses in which the flame is allowed to evolve only on a long time scale. although by means of experiments with a flatflame burner.
The evaluation of L(7”) will be considered in Section 8. then the quantity in equation (24) that varies most strongly is rnz. This enables flammability limits to be tabulated without extensive reference to the experimental configuration [2]. exhibiting extinction condition for defining flammability limits. Extinction.2. [2] may be attributed to a larger value of L ( T f )for downward propagation. That limits usually are somewhat wider for flames propagating upward in tubes than for flames propagating downward [l].3. It has already been indicated in Section 5.3. where questions will be addressed relating to the physics of losses and also to the accuracy with which equation (24) follows from equation (21). .5 2 4 FIGURE 8.3 where eE’RoTaf with its critical value for extinction. especially in the wings near the limits. An accurate application of equation (24) to the calculation of flammability limits entails accounting for variations other than that of eCEiRoTaf.5.278 Ignition. and Flammability Limits varied with the pressure and the initial temperature held fixed. where 4 denotes the equivalence ratio for a fueloxidizer mixture (see Section 5. Relevant parameters are exhibited explicitly in equation (24). In a first approximation. 0 1. Therefore. and a critical value of e P E i R o T m may f therefore be identified for extinctions from equation (24). achieves Rich and lean limits are seen to occur in Figure 8.5 1 .3. mainly because of its proportionality to eE’RoTaf. differences in values of parameters such as L( 7 ’ ’ )between different experiments produce relatively small changes in the limiting values of 4.7 xf exp (EIR0Taf) f I 0 I 0. the remaining factors may be considered constant in any given series of experiments. The situation is illustrated schematically in Figure 8. Illustration of the dependence of the adiabatic flame temperature and of the Arrhenius factor on the equivalence ratio.7).3. The comparatively weak variation of I $ causes a strong variation of eEiRoTaf.
2. given in . In cases of this kind (some of which may be envisaged from chemicalkinetic considerations but none of which have been established indisputably to occur for real combustible mixtures). then a limiting pressure is reached. is the Planck mean absorption length. since then a very small heat loss may produce the same limit as a large loss.2.5. equation (9) may be derived with L = 4uT4/1.8. however. From the definitions of B. the value of the rate of heat loss is irrelevant to the flammability limit over practical ranges of conditions. in calculating flammability limits. Estimates of heat loss Flammability limits can be truly independent of the experimental apparatus only if there is a critical point (for example.. below which flame propagation cannot be achieved. a critical value of 4) at which the exothermic chemistry is shut off completely. Pressure limits of flammability It is often observed experimentally that if the relative concentrations of the reactants and the initial temperature of the mixture are kept fixed but the pressure p is decreased. the righthand side of equation (24) decreases more rapidly than the lefthand side as p decreases. The behavior. and of the absorptivity K . by one or two orders of magnitude. 8.4. where m FZ 0 for conductive or convective losses and 0 i m I 1 for radiative losses.2.   8. The flammable range of 4 usually narrows as the pressure is decreased. m. flammability limits may effectively be controlled by chemical kinetics if there is a value or a narrow range of 4 at which the overall rate of heat release decreases abruptly. Although it is highly unlikely that this occurs. the presence of heat loss is essential. p". values that are appropriate in the vicinity of each limit separately should be employed. (26) where o is the StefanBoltzmann constant and 1. In this case. These observations are consistent with equation (24). Since n > 1 for practically all flames. in a sense. and below a critical pressure flame propagation does not occur for any value of 4 . and the limiting equality may therefore be expected to be surpassed at a sufficiently low pressure. without heat loss. by use of equation (E50) in suitable versions of equations (115) and (128). It will be seen in the following section that L ( T f ) pm. An even more likely candidate for producing flammability limits that are independent of the apparatus over a very wide range of conditions is a solely radiative heat loss from an optically thin gas. in which the main pressuredependent quantities are mi and L(Tf). corresponds to a heatloss theory with a very large value of E. Premixed Flames with Heat Losses 219 that parameters such as i l and E vary with 4.In terms of the overall order n (the pressure exponent) of the overall reaction rate.
the generality of equation (21). the pressure dependence is weaker. is proportional to the number density of emitting In equation (26). is likely to radiate much more strongly than either the intermediate oxygen atom or the product oxygen molecule. An accurate application of equation (21) with 1 = l/e to calculate extinction with radiant loss can be made only if K ( Z ) is known accurately. extinction conditions for premixed flames are unavailable. of order lip). and although the corresponding diffusionflame problem has been investigated [SO]. ozone.2. in this case. and Flammability Limits Section E. If the radiation is dominated by chemiluminescence from an intermediate that exists only over a short range AT about the value z = zi. the asymptotic analysis requires modification in that a radiant loss term belongs in equation (17). in equation (26) to vary in such a way that 1/1.2). we note that convergence of the integral in equation (21) requires ~(z) to approach zero as z approaches zero. varies strongly with position x in the flame because of the changes in the concentrations and degrees of excitation of the emitters. For radiant loss often this has been achieved by putting L proportional to T 4 .5. is proportional to pressure at constant temperature and relative composition (see Section E.5. is particularly helpful for use with radiant loss. In flames radiant losses often are negligible until a value of z is reached at which radiatively active intermediate or product species are present. until the gas begins to become optically thick. not dependent on the experimental environment. an exception to this rule may be the ozone decomposition flame. is a local property of the gas mixture. ~ ( qis ) obtained through equation (12) from an average value L in the region of emission.T:. Although the needed details of flamestructure information usually are unavailable for real flames. dimensions of the experimental apparatus do not influence any of the terms in equation (9). Joulin has completed the necessary analysis in recent work not yet published. and a better approach is to allow I . it may be seen from equation (E50) that I . in turn. 1 molecules. goes to zero in the fresh mixture. which admits a wide class of variations of L. / 1 .280 Ignition. radiation at the ambient temperature is entirely negligible. with the last term viewed as representing absorption of radiation emitted by the surroundings. to be used in calculating extinction. Therefore. then ic(zi) may be large (for instance. Extinction. when reabsorption is not negligible.* * G. a number of relevant general observations can be made. and predicted limits from radiant loss are apparatusindependent. These results were employed in the preceding subsection. It might be expected that chemiluminescent emissions will often be strong but restricted to the main reaction zone. then equation (21) becomes approximately 1 = B I C ( ~ ) A Z ~ where / ~ . in which the reactant. First. In addition. of order unity rather than of order 1/P) while still enabling equation (21) with 1 = l/e. For this reason. . which might not occur until sizes on the order of meters are reached. We may note that if Azi is small (for example. Typically. 1/1. which.
n.2. surface [SO]. The carbonaceous particlesor any other solid particleseasily can be the most powerful radiators of energy from the flame.l)". then we find from equations (12) and (26) that K is proportional to z(z + a . for spherical particles of radius r. and that in this case equation (24) may therefore be used with the factor of 2 replaced by 1.2.1)5]/5. Since the downstream zone is convective in character. Therefore. Currently it is uncertain whether any of the tabulated flammability limits are due mainly to radiant loss (since convective and diffusive phenomena will be seen below to represent more attractive alternatives). an approximate formula for 1. However.3 in discussing Figure 8. in contrast. These emissivity t.)4/(nr. appreciable changes in the loss rate usually produce relatively small changes in the limits.( a p 1. only ~ ( 7 occurs losses anywhere in the upstream convectivediffusive zone will influence 1. in typical experiments these losses are larger than radiant losses. in fact. wherein it was found that the upstream and downstream contributions were equal (their sum giving the factor of 2 in the equation). with number density ns.1 . Premixed Flames with Heat Losses 28 1 Estimates often indicate that although chemiluminescence may provide flames with their distinctive colors. which suggests that radiation emitted downstream from the reaction zone is about five times as important as that emitted upstream in contributing to extinction. and we have indicated in Section 8. = (T/T. only the loss rate at its beginning can contribute to the reactionzone behavior and to extinction ~ ) in the first term in equation (21)].1)5 . but if any of them are.O and C 0 2 in hydrocarbon flames). then the rich limits of sooting hydrocarbon flames almost certainly can be attributed to radiant loss from soot. [that is.2 or possibly unity. Near rich limits of hydrocarbon flames. and surface temperature T. the associated energy losses may be negligible in comparison with radiation from major stable products (such as from H. This may be contrasted to the situation in which L is linear in z. It appears that for radiant losses from reaction products ~ ( 7 is ~ )often appreciably more important than the integral in equation (21). they often have been excluded as mechanisms that may be responsible for fundamental limits of flammability. . as represented by equation (24). depend on the experimental configuration [l].2t.) parameters depend on the chemical kinetics of soot productiona complicated subject.3 that near the limits. is 1.for example. convective losses are good candidates for explaining tabulated limits.8. If the concentration profiles of the products in the flame are similar to those of the normalized temperature and if the rate of radiant emission from the products is proportional to T4. Since precise values of flammability limits usually do. The function ~ ( z )is difficult to compute for soot radiation for use in equation (21) because it depends on the histories of number densities and of size distributions of the particles produced ..1)4 + [(zf + a' . whence 1 in equation (21) becomes proportional to zf(zf + tll . Conductive or convective losses of energy from the flame necessarily depend on the geometry of the apparatus. soot is sometimes produced in the flame.
extinction formulas that in principle can be used with an accuracy as great as that of equation (21) for radiant loss are unavailable for convective or conductive loss.282 Ignition.5. Rough heatloss estimates are readily obtained from overall balances. [52].) estimated from an approximate analysis. for hydrogenrich mixtures that do not experience the instability.)/(D/2). The conductive heat loss per unit volume from a plane flame in a circular tube of diameter D can be estimated by the following simple reasoning.5.2. the lean limits do not depend on the size of the apparatus. of lean methaneair flames that can propagate downward in a tube 2 cm in diameter but are extinguished in a tube 5 cm in diametermay be explained on a similar basis [Sl].T'.2) that leads to fingers of flames separated by a nonreactive. Consider an Eulerian element of gas of length dx in the tube whose walls are maintained at the temperature To. The energy per second conducted to the walls from this element is the product of the thermal conductivity A. The accuracy of the extinction prediction then depends mainly on the accuracy of the heatloss estimate. Loss analyses must consider nononedimensional heat transfer and should also take flame shapes into account if high accuracy is to be achieved. and Flammability Limits Special sets of circumstances may conspire to produce limits independent of apparatus dimensions even if convective or conductive losses are important. The most convenient approach in accounting for convective or conductive losses appears to be to employ equation (24) with L(T. conductive or convective losses never can be consistent exactly with the planeflame assumption that has been employed in our development. Since the dimensions over which this conduction occurs are independent of the tube diameter. [53] that take into account both the instability and the conductive loss. The observations thus are consistent with theoretical predictions [52]. Extinction. Therefore. vertical tubes experience a cellular instability (see Section 9. Unusual observationsfor example.3) can influence extinction phenomena. The rate of heat loss per unit volume is then obtained through division by the volume of the element dxnD2/4: . the limits are observed to depend on the tube diameter [5 11. This is difficult to accomplish by methods other than numerical integration of partial differential equations. Unlike the radiant loss from an optically thin flame. in hydrogenoxygennitrogen mixtures. as is normally expected for conductive losses. and they show that diffusivethermal effects (Section 9. The sizes of the fingers are determined by the instability and are independent of the tube diameter. hydrogenlean flames propagating downward in open. cool gas from which hydrogen has been removed by diffusion [Sl]. On the other hand. a mean temperature gradient ( T . Conductive losses from the fingers to the cool gas are responsible for the lean flammability limit. and the wall area nD dx. For example.
This minimum tube diameter is called the quenching diameter.8. and results quoted in [17] and [36]. Premixed Flames with Heat Losses 283 This linear dependence of L upon T is consistent with the derivation of equation (24). The results may be described by replacing the factor 8 in equation (27) by a factor b that depends on a suitably defined Reynolds number. indicate that b z 15 for circular tubes. Rates of heat transfer to the walls in these various configurations may be estimated by use of heattransfer coefficients. for flames propagating upward. Equation (28) thus shows that L(Tf)is independent of pressure (as stated in Section 8. By substituting equation (28) into equation (24) and solving the resulting expression for D . gas velocities in the direction of flame propagation may be influenced strongly by buoyancy and tend to be positive (upward) both before and after the flame passes. Since none of the other physical parameters in equation (24) depend on D. In contrast. Putting fi z 10 . (28) Typically.4) and that L(Tf) l / D z . we obtain equation (4) from equation (30). for example. These influences may depend strongly on the configuration of the apparatus in which the flame propagates. For flames propagating downward in open tubes. Neglecting the small difference between the quenching diameter and the quenching distance.). we find   D = J2epbcP/(Am. for example. with a = J2eph. Influences of gas motion have been neglected in equation (27). It is thereby found that L(T’) = (bA/D2)(T’  To). the gas velocity relative to the walls is directed from the cold to the hot side of the flame.2. An approximate formula for the quenching distanced can be obtained from equations (24) and (28).2. in a forced flow that maintains the flame stationary in the laboratory frame.1. as well as on the tube geometry and the type of flow. (29) Equation (54) implies that equation (29) can be written as D = J2epb6. at the burning velocity upstream and at the burntgas velocity downstream. (30) where 6 is the thickness of the adiabatic laminar flame. variations in b are relatively weak. The dependence L ( T f ) 1/D2 implies that the lefthand side of equation (24) increases as the tube diameter D decreases. the velocity of the gas relative to the walls is approximately zero ahead of the flame and approximately the difference between burntgas and unburntgas velocities behind . which are obtained from heattransfer correlations that depend on suitably defined Reynolds numbers (or perhaps Grashof numbers) [54]. which is empirically approximately the same as (perhaps 25 or 50 greater than) the quenching distance d defined in Section 8. an extinction condition is reached when the tube diameter D becomes sufficiently small. Quenching phenomena can thus be attributed to conductive heat losses.
there are a number of results concerning ignition times that provide the history of the ignition process. Reviews of problems of this type are available (see the literature cited in Section B. typically it is found that multiple steadystate solutions exist and that the one with the smallest value of the maximum temperature represents a stable. Thus we see that a is large as a consequenLe of two effects: First. If interest is focused on ascertaining critical ignition conditions. the large nondimensional activation energy makes the flame highly sensitive to small rates of heat loss and capable of being extinguished by extraction of a small fraction of the heat released . There are autoignition problems for which progress is being made in describing ignition histories by asymptotic methods [57][59]. Section 3 combustible within a container whose walls are maintained at a fixed temperature. Mech. Problems of ignition by means of an externally applied stimulus have also been introduced from the viewpoint of activationenergy asymptotics [46] . Above a critical value of a parameter measuring the magnitude of the reaction rate. W. and give reference to reviews for entrance to the literature.2. second. and ignition occurs. Extinction.4 and an introduction from the viewpoint of asymptotics is available [46]. to appear. There is recent literature in which activationenergy asymptotics have been applied to the steadystate equations for various autoignition problems [55]. Dold. Quart.* an introduction to these is available [46]. . achieved in the preceding section through activationenergy asymptotics. In addition to specifications of critical conditions for ignition to occur. Math. Here we merely identify a few illustrations. the convective enhancement of heat loss combined with the geometrically favorable factor for conductive loss cause the loss rate to be about an order of magnitude greater than might be guessed in the absence of a loss model. which is not very different from the value a z 40 quoted in Section 8. has been matched and in some ways surpassed for ignition phenomena. [56]. Appl. Timedependent conservation equations must be considered if the history of an ignition process is to be studied. ACTIVATIONENERGY ASYMPTOTICS IN IGNITION THEORY The understanding of extinction phenomena.5. An example is the autoignition of a equations (for example. 4 ) . J.5.284 Ignition. and Flammability Limits and b z 15 in this formula gives a z 30.1. this slowreaction solution no longer exists. The situation has been discussed in Section B.3.2. * For an improved analysis see J. introduce one example. slow reaction. 4 . then analyses often can be performed on the basis of steadystate conservation . (1985). 8.4).
is pc. The timedependent equation for conservation of energy for the material. Let a reactive material occupy the region x > 0. and to avoid complications assume that the material remains at rest and has a constant density p. but we shall not do that here. to achieve the greatest simplicity we adopt a thermal theory. although in various applications that have been cited the full set of conservation equations has been considered. aT/at  1 a2T/aX2 = wqo. the ignition of a solid by processes involving exothermic heterogeneous reactions at its surface [69]. where 1is the constant thermal conductivity of the material. Equation (31) may be derived formally from equations (111). therefore. analogous to equation (9). t). [63]. Equation (31) expresses a balance of energy accumulation. To achieve the simplest thermal theory. Consideration of the history (to calculate quantities like ignition times) necessitates retention of time derivatives in the conservation equations. Just as in the previous section.8. We shall briefly address the first of these problems as a vehicle for introducing the techniques. . be exposed to a constant heat flux q = 1 aT/ax at x = 0 for all time t > 0. (121a) and (122) by introducing a list of explicit assumptions. and a review of applications of asymptotic methods to these problems has been published [61]. we shall neglect effects of reactant depletion on the reaction rate and write = p~eE/RoT (32) in equation (31). the ignition of a solid by a radiant energy flux for situations in which the only exothermic reaction occurs in the gas phasefor example. (115). this same problem with heat loss at the surface to an inert material [65]. [71]. the endothermic gasification of a solid by a constant energy flux [67]. after mixing of fuel vapors with an oxidizing gas [72][74]and the ignition of a reactive material exposed at time zero to Newtonian heating [75] to an elevated surface temperature [76]. where A is a constant having the dimensions of a reciprocal time.3. initially at temperature To. which may be contrasted to the balance appearing in equation (9). [70]. [66]. is soluble in principle for T ( x . Specific problems that have been addressed by asymptotics include the ignition of a reactive material that occupies a halfspace and whose planar surface is subjected to a constant energy flux beginning at time zero [62]. (31) where the remaining symbols and assumptions are the same as those of equation (9). although coordinate transformations readily enable this assumption to be removed. [68]. The problem that has been stated here is wellposed in a mathematical sense and. ActivationEnergy Asymptotics in Ignition Theory 285 an extensive review of research in this area that preceded activationenergy asymptotics is available [60]. Let the material. ignition of the same material by a radiant energy flux with indepth absorption of radiation [ H I . heat conduction and chemical heat release.
the material experiences only inert heat conduction because the heatrelease term is exponentially small in the large parameter that measures E. the properties of which are well known [77]. for example. Let t. during most of the time that t < t. This behavior is exhibited in the present problem if a nondimensional measure of the activation energy E is large. Then t = RoT:/(ETo. the time at which the rapid temperature increase occurs.) for the inert solution given by equation (33).286 Ignition. Ignition processes often are characterized by a gradual increase of temperature that is followed by a rapid increase over a very short time period. Illustration of the solution to the problem of ignition by a constant heat ux.4. (33) where the nondimensional time and space variables appropriate to the inert problem are e = 4zt/(P~c. (35) respectively. In the present problem. a more precise definition oft. arises in the course of the development.) and (34) z = qx/(ATo).. denote an ignition time. 0 X 8. has a known solution that can be derived by Laplace transforms. and where erfc denotes the complementary error function. Let denote T(0. with w = 0. and that can be written as TIT. as is true in the applications. 0 F G IRE F 0 X 0.T. Extinction. . = I + 2&ez’’(4e) z erfc[z/(2&1 =~ ( z01./&) (36) . t. and Flammability Limits Reactivediffusive zone Transientdiffusive I \\x t Increasing T O . The inert problem.4 for early t. The solution given by equation (33) is illustrated schematically by the curves in Figure 8.
and thereafterbecause of the Arrhenius factorthe chemistry will become dominant and cause a rapid temperature increase. To obtain an ignition time t. if less elegant.3. obtainable by similar reasoning simply from a nondimensionalization of equation (31). the result is a transcendental equation for t. a quantity of order unity. use is made of equation (34) and of the solution T. as calculated from the inert solution.6. becomes large enough for the rate of heat release to equal the inert rate of increase of thermal enthalpy. It is sufficient.. and T/T0are of order unity.3. = 1 [in place of the final equality in equation (37)]. ActivationEnergy Asymptotics in Ignition Theory 287 is an appropriate small parameter of expansion. this cannot be seen until the analysis is completed. .5. The smallness of t is attributable primarily to the necessary smallness of RoT. Analogous factors appear in optimum selections of t for other ignition problems. evaluated at x = 0. for example.the definition t = RoTi/[E(T.2. (37) whereD.6. that at x = 0. = To(1 + 2Jec/. the ratio of a reaction rate to a heatconduction rate. as discussed in Section 5. the temperature . the additional restrictions then appear in the course of the analysis.&) is irrelevant for large values of 8. from equation (37). = 1. simply to select 6 = RoTo/Eand to assume that 8.. The factor T. is a Damkohler number. analogous to the t of Section B. because equation (33) shows that this factor then approaches 2/n. is the value of 0 at t = t./E.3.. for ignition by a constant surface temperature T.. A similar factor has appeared in /3.To)] is best. > To. which may be solved explicitly to give In the rough approximation D. In view of equations (32) and (33). A more primitive approximation. as obtained from equation (34). is D..4 or to 1/p for the p of equation (13) or of Section 5. The motivation for writing equation (37) is that the chemistry will remain insignificant everywhere until the maximum temperature. this gives D. G (q0pLToA/q2)epE'RoTc = I/ . Here 0. A first approximation to the ignition time may be obtained by equating the heatrelease term in equation (31) to either of the other two terms. The only reason for not putting n/2 as a factor in equation (36) as well is to achieve some simplification in the final formulas. identified with ignition.8../(T. the inclusion of the additional factor TJ( To&) in the small parameter serves to assure that ignition occurs prior to homogeneous thermal explosion if t is to be small.) (38) given by equation (33).
O .. namely. According to equation (40). = 1 is easy to use.)/(€~6. we may investigate the process of transition to ignition by stretching the time variable about t = t.. Use of this expansion to achieve a linearization within the exponential enables the combination of equations (31) and (32) to be written in the variables s and X as + a$/as with a$/aX selection =  (a2*/ax2y6= ( I / & ) & + ~ . . is to be treated as being of order unity. the ignition temperature depends on the externally applied heat flux but is independent of the heating time .6. where F is defined by equation (33). To consider a small perturbation about the inert solution with a nontrivial expansion of the Arrhenius factor. z about z = 0 and 6 = O. the expansion of equation (33) r1 6 . Solutions may be sought as power series in and a &. since equation (33) shows that T differs appreciably from To when z/. it is less accurate than equation (37). (41) where c is a constant translation to be selected so that an event best identifiable with ignition occurs at a value of s of order unity. = (43) 0. matching to the inert solution requires that t. the definition * = C(T/To). more precisely. in an intermediate zone.~ + . 6)1/(4%) (42) is introduced.. Here the (44) 0 as the heatflux boundary condition at X c =1n(DII&&) has been introduced as a first approximation in equation (41) to achieve an appropriate scaling. To construct an expansion for small E in the stretched variables.co (and also as Z + 30).) + c. . Extinction. let s = (6 .. in that the reaction is negligible when the temperature is below this value and dominant when above. Substitution into the differential equation then eventually reveals that space stretching also is needed if space derivatives are to appear at the lowest order in 6 . which in turn is appreciably less accurate than the result of a more thorough asymptotic analysis.s) coordinates.F(z. with X of order unity in an inner zone and Z of order unity in an outer zone [or.. To proceed with an improved theory. F(z. 6.. is needed. AlthoughD. To accomplish this.h& is of order unity]. . .) (0 . 6) = F(0. ) / 7 + . . equation (37) gives an ignition temperature that decreases slowly as the ignition time increases. In equation (43) and also in the equation in the (2.b + 0 as s + .288 Ignition. Since it is convenient to work with the difference between the true temperature and that given by the inert solution. It is found that the variables X = z/(E&) and Z = z/& both occur. and Flammability Limits plays the role of an ignition temperature.
and equation (44) therefore yields DII= 0. ActivationEnergy Asymptotics in Ignition Theory 289 twoterm expansion of the inner solution may be developed and matched with the leading order of the intermediate solution to produce the problem a$las = a2$/az2. Equation (45) describes a transient heatconduction problem having a surface heat flux that increases exponentially with the surface temperature. It seems natural to identify this runaway as a unique ignition event and to equate the time at which it appears to the ignition time. but these departures from the inert solution are small. the equation cannot be meaningful everywhere. Since we originally defined 8 = 8.431. s) = co occurs at s = 0. followed by a shorter stage of transition to an ignited state. The view of the ignition process that arises from this analysis is illustrated in Figure 8. occurs in a narrow zone at the surface that may be described as a reactivediffusive zone because the time derivative is seen to be of higher order in t there.65/&. ] (45) Investigation of the properties of equation (45) provides the improved formula for the ignition time. and a broader transientdiffusive zone where the additional heat flux caused by the heat release is felt. where the chemical heat release causes an accelerating increase in temperature. = $ ( a . A numerical integration of equation (45) has shown that +(O. tion to There are two additional zones during the transition stage. This incremental heat flux is caused by the chemical reaction that. with t given by equation (36). of order t.at Z = 0.4. s) = 0. at ignition. s) to approach infinity at a finite value of s. equation (41) shows that c = 0. The surface temperature as a function of time exhibits an inflection and then a runaway during the transition stage. There is an initial stage of inert heat conduction. a thin reactivediffusive zone at the surface. Of course. The infinity of $ may be described as a thermal runaway that develops in the stretched temperature variable. according to equation (43). The depth of the heated layer and the elevation of surface temperature increase in proporduring the inert stage under the influence of the surface heat flux.431.3.8. Carrying the asymptotic expansion to higher order would produce 1 Lo(&) as a correction factor for equation (46). The exponential nonlinearity in the boundary condition in equation (45) is known to cause $(O. It is a consequence of the linearization within the exponential produced by the expansion in e and has + 4 . Equation (46) differs from equation (37) by the factor 0. the divergence at a finite t is not contained in the original problem defined by equations (31) and (32). a$pz = e*+' +(z. at larger values of s.65/& (46) as an improved equation defining the ignition time.
and the subsequent history of the system is observed. as a consequence of the onset of reactant depletion. Thus a great variety of mathematical problems is encountered in asymptotic analyses of ignition. It may involve partial differential equations as in equation (49. a flame develops and propagates into the material. the time at which radiation in a selected wavelength band associated with reaction intermediates or products is detected. For the problem as formulated. initially is approximately at the adiabatic flame temperature. This go. the ignition stimulus is applied for a fixed period of time and then removed instantaneously. Often the maximum temperature decreases after the ignition source is turned off and then begins to increase rapidly a short time later. The shortest exposure time that results in satisfaction of an ignition criterion within a reasonable time after shutofffor example. and this stage must be analyzed to obtain expressions giving ignition times accurately. Other definitions more closely tied to experiment include the time at which a selected. Extinction. the true surface temperature increases at a rapidly accelerating rate over time scales small compared with those having s of order unity. there always exists a degree of arbitrariness in defining an ignition time. sufficiently high temperature is achieved at a specified position in the reacting system. temperatures much higher than any of practical interest. and a go. These subsequent stages can also be described by activationenergy asymptotics [63]. ordinary differential equations in space or time.290 Ignition. within a period of one exposure timeis then taken to represent the ignition time. In fact. the time at which luminous radiant output from the system is first observed. nogo definition.2. The nature of the mathematical problem encountered in the transition stage varies with the type of ignition problem considered. . At very long times the process returns to one of inert heat conduction under the influence of the surface flux. on the order of an adiabatic flame temperature. Fundamentally. The nonlinear driving term of chemical heat release. near t. may occur in a differential equation or in a boundary condition [such as equation (45)] and may act over different time and space scales.4. With this last definition. nogo definition has been employed for our example problem in an analysis by methods in which the original partial differential equation. The definition based on thermal runaway that has been introduced here is a mathematical definition embedded within activationenergy asymptotics. the acceleration lessens at much lower temperatures. in effect. or integral equations that cannot be reduced to differential equations [76]. the acceleration begins to diminish only when surface temperatures on the order of E/Ro are reached. In reality. the time at which a specified rate of increase of temperature occurs. soon satisfying an ignition criterion according to one of the other definitions given above. Ignition problems typically exhibit a distinguished stage of transition to ignition for small c. now in a material which. and Flammability Limits been encountered in a simpler context associated with thermal explosions in Section B.5. which is characteristic of these problems. Subsequently.
Mech. A. P. B. Berlad and A. 14. Government Printing Office.” Bulletin 627. An independence of this type often arises in ignition problems because c is relatively small in ignition. Washington. 1951. G. (1955). New York: Pergamon Press. D. Jet Propulsion 27. more than one chemical process may be involved. A. the theoretical problems become challenging. SOC. Cullen. “Flammability Characteristics of Combustible Gases and Vapors. Explosions. 728735. Explosion and Combustion Processes in Gases. New York: McGrawHill. Coward and G. 352 (1958). H. 126138.. 1960. Bureau of Mines. (1949). Aircraft Engineering 25. E. 2736. 1959.. Spalding. Berlad. 1023 (1954). M. Potter. E. (1977). Penner and F. . New York: Academic Press. Jost. 1. F. there are many ignition problems for which the value of the ignition time depends strongly on the definition adopted.. 75. Potter. and the resulting ignition time was found to correlate well with equation (46). Simon. (1965). The low temperatures often associated with ignition events can in comparison with lo’ for typical flame and extinction cause c % processes . “Limits of Flammability of Gases and Vapors. Chem. E. vol. Roy. Bureau of Mines. 145181. the ignition time is practically independent of the definition employed. Spakowski. No. L. New York: Pergamon Press. E. Potter. P h p . (1957). A. U S . Jensen. B. E. 3rd Symp. 3. S. 15. Jr. 8998. Chem. 9. Detonations. Jr. Rosen. (1952). Berlad. U S . 16. A. 30. D. D R. 2. D. Mullins and S. 31. Jones. 544 (1957). 13. U S . A. A. H. J. Eng. S. A. Phys. Zabetakis. and A.C. Although methods of the type illustrated here may be employed in analyzing these more complex processes.S. London 245A. Flames and Explosions of Gases. 6th Symp. D. Lefebvre.. SOC. This occurs both experimentally and theoretically and is attributable to the process being more complex than that analyzed here. Proc. M. 4th Symp. 16th Symp. Trans. Ballal and A. 273 (1959). Williams.C. NACA Rept. W. Belles. W. 10. 12. Combustion. 11. was integrated numerically [78]. 43 (1953). 8. L. and A. L. Washington. J. Putnam and R. S. 1264 (1956). 4.References 29 1 obtained from equations (31) and (32). Am. Progress in Combustion Science and Technology. However. thus activationenergy asymptotics tend to be more accurate for ignition. B. U. Flammability and Ignition. for example. REFERENCES 1. Penner. 1st ed. 5. 7. Jr. E. Government Printing Office. R. Lewis and G. This agreement implies that for the problem considered. G. 298 (1959). F. 16891698. (1953). 264 (1953). 1946. 5th Symp. or the system may be close to a condition of transition from one ignition mechanism to another. von Elbe. A.” Bulletin 503. 58. 6.
46. L. and T. J. R. Matkowsky. No. Lefebvre. NACA Res. 61 (1981). T. 39. C & F 3. Spalding and V.412 (1980). London: Butterworths Scientific Publications. Buckmaster and G. R. A. P. Proc. (1953). Math. 373 (1979). SOC. E. Roy. 333 (1982). B. 38. 33. 34. 553 (1959). Ballal and A. Swett. 20. (1973). F. 55. “Ignition and Flammability of Hydrocarbon Fuels. A. 708714. C & F 40. 27. H. 155 (1979). M. E. Memo. C & F 24. 43. Williams. 39. C & F 6 . Ballal and A. Math. 735743. 3rd Symp. S I A M J . J. C & F 39. Ballal and A. New York: Academic Press. Ballal and A. 23. 17371746. (1981). Kono. A. C & F 4. and Flammability Limits 17. 28. H. J. 14731481. C & F 4 . 48. Olsen.9. C & F 35. S. AGARD. 13331342. C & F 1. 438 (1957). Friedman. Joulin and P. CST 19. 24. J. 18th Symp. C h e n a n d T . 277318. A. Teor. Belles and C. 33 (1978). H. Spalding. 173 (1981). 6th Symp. T. Maly. 325 (1960). E54F29a (1954). J. 35. Mitani. S. and F. 163 (1977). Proc. 13th Symp. SOC. Poland. T . C. D. Avery and H. 18. No. Appl. 26. R. H. Bregeon. H. (1975). Eng. 45. Yang. Stine. A. J. K. 523532. A. Combustion. Ind. Ballal and A.292 Ignition. NACA Rept. 41. M . Botha and D. Y. H. W. J. Kassoy and J. Zhur. F. Chem. Jr. (1981). 231 (1981). (1981). 15th Symp. 36. 11. and F. K. L. Th. 32. 30. 3 1 3 ( 1 9 6 0 ) . Lefebvre. Appl. Williams. C & F 33. 18th Symp. Swett. R. H. 143 (1982). Williams. SOC. 44. 29. R. A. 52. C & F 7. 49. R. J. D. B. CST 23. 18th Symp. S. 57. 169 (1980). D. Fenn. Extinction. 215 (1962). 173178. Sci. B. C.London 261A. Proc. Sakai. 36. 56. (1953). . Zel’dovich. I. 99 (1975). D. London 225A. (1971). C & F26. and J. Matkowsky. 47. 42. Adler and D. 145147. D. Ludford. 53 (1961). 13 (1978). 93 (1980). Glassman. L. 41. Spalding. Gerstein and W. Fiz. 31. B. C. Yang. Sohrab. 39 (1963). de Soete. (1949). 163 (1961).. 53. N . Ener. Vega. B. B. R. K. S. 38. 1977. 1300 (1959).qy Combust. 8. Buckmaster. Lefebvre. Math. Eksp. 151 (1976). 54. 110120. S. 1956.” Chapter I11 of Basic Considerations in the Combustion of Hydrocarbon Fuels with Air. B. D. 71 (1954). von Karman and Millan. Adler. A. Lefebvre. C. H. (1957). Cambridge: Cambridge University Press. 4th Symp. Mayer. Liiian. G. B. G. Smith. 25. S. 1982. Putnam and L. Roy. Gordon. 43. Archiuum Combustionis 1. London 357A. A. E. Selected Combustion Problems II. 51. 83 (1957). 40. Pr0. Y . 21. D. D. Proc. 2865 (1951). H. R. Williams. CST 27. A. B. Clavin. D. Roy. Berlad and C. Iinuma. 4th Symp. Adelman. 14th Symp. Yumlu. 37. SIAM J . 50. 391 (1980). Kumagai. 17471754. 223 (1976). Appl. A. Lefebvre. Kapila. Spalding.London UOA. G. 39. Hermance. T o o n g . 19. H. R. C & F 5. Kapila and B. G. 159 (1941). SZAMJ. 22. Yang.382 (1980). S. Acta Astronautica 3. Mitani and F. C. Theory of Laminar Flames. Roy SOC. Aly and C. H. 11091 118.
M. H. C & F 16. Math. Crespo. 66. Stegun. 1965. 259 (1983). 17th Symp. 76. 445 (1975). “Asymptotic Methods in Ignition Theory. New York: Dover Publications. Averson. International Journal o f Engineering Science. 75. 39. 19. 223 (1972). K. T. A. Williams. C & F 27. 62. W. A. 61. (1979). J. 11 (1970). A. A. Liiian and F. 11631171. M. A. Handbook of Mathematical Functions. 40. A. CST 2. (1981). Kassoy. Liiian and F. CST 6. T. Williams. Jr. 261 (1981). Acta Astronautica 2. Williams. Williams. A. 59. F. Williams. 85 (1972). A. T. 18071813.References 293 58. Liiian and F. Williams. H. 29 (1981). W. 47 (1977). CST 16. 60. 64. 91 (1971). 69. 67. Kindelan and F. R. 224227. Bush and F. A. A. 77. M. A. A. Kindelan and F. 1 (1975). 78. A. 18th Svmp. SZAM J. Williams. M. Acta Astranautica 2. Bradley. C & F 18. Niioka. 1199 (1978). Williams. B. 89 (1971). C & F 29. 73. B. Kindelan and A. G. Appl. A. 955 (1975). M. Niioka. Poland and D. 71. Acta Astronautica 5. Williams. CST 10. CST 3. A. Kapila. A. Williams. T. 72. E.. Kapila. 70. Kindelan and F. Mexico (198 l). 43 (1977). Abramowitz and I. Williams. SZAMJ. 63. 65. Liiian. K. Math. Bush and F. Oaxaca. A. 68. CST 18. Appl. 41. 36. 207 (1978). 74. Niioka and F. 495 (1981).” Memoria del VII Congreso de la Academia Nacional de Ingenieria. 587 (1979). A. C & F 50. . Liiian and A. Niioka and F. A. 21 (1980). Merzhanov and A. 321 (1976).
a thorough classification is difficult to present because so many different phenomena may be involved. for example. shock instabilities. and system instabilities. more detailed presentation will be reserved for a few of the types that have been investigated more 294 . for example. most of the known instabilities that have been observed and studied. which involve an interaction of processes occurring within a combustion chamber with processes operative in at least one other part of the system. at least in passing. or hydrodynamic instabilities. a classification is based on the components of a system (such as a motor or an industrial boiler) that participate in the instability in an essential fashion. Thus intrinsic instabilities may involve chemicalkinetic instabilities. which are specifically associated with the occurrence of combustion within a chamber. Three major categories are identified : intrinsic instabilities. Chamber instabilities may be caused by acoustic instabilities. Here we shall try to mention. and have been assigned different specific names in different contexts. Within each of the three major categories are several subcategories selected according to the nature of the physical processes that participate in the instability.CHAPTER 9 Combustion Instabilities Instabilities arise in combustion processes in many different ways. or fluiddynamic instabilities within chambers. which may develop irrespective of whether the combustion occurs within a combustion chamber. and system instabilities may be associated with feedsystem interactions or exhaustsystem interactions. chamber instabilities. diffusivethermal instabilities. In one approach [l].
3.2. unexpected irregular histories of the chamber pressure as a function of time occasionally were observed in place of the smooth history for which the motor was designed [2]. The final section addresses intrinsic instabilities of premixed flames. these subjects are touched briefly in Section B. We begin by considering a particular chamber instability for which a substantial literature exists.4.4.1. for example. but reviews are available [11. diffusion flames appear to exhibit fewer intrinsic instabilities than premixed flames. as indicated at the end of Section 9. Certain chamber instabilities that are not related to acoustic instabilities (such as Coanda effectsoscillatory attachment of flows to different walls) will not be discussed here.1. 9. ~71). acoustic instability of solidpropellant rockets. ACOUSTIC INSTABILITIES IN SOLIDPROPELLANT ROCKET MOTORS 9.4.1 and will not be revisited. which builds on the material presented in Section 7.2.5. [ S ] .9. The irregularities often were accompanied by regular pressure oscillations having frequencies comparable with the natural vibrational frequencies of sound waves in the rocket chamber [3] and can be caused by amplification of acoustic waves [4]. System instabilities in other combustion devices are discussed in Section 9. To be specific we may consider cylindrical chambers of the type illustrated . Section 9. will be addressed. Oscillation modes At an early stage in the development of solidpropellant rocket motors. Next.5. another subject that has received extensive analysis. Among the areas not covered here is that of intrinsic instabilities associated with chemicalkinetic mechanisms. Acoustic waves in chambers possess various modes of oscillation.3. The arrangement of sections does not directly follow the categorization just given.3 concerns chamber instabilities and system instabilities of liquidpropellant rocket engines. a fundamental topic of central relevance to both laminar and turbulent flame propagation. Certain aspects of intrinsic instabilities of diffusion flames were mentioned briefly in Section 3. Instabilities that result from soundwave amplification are called acoustic instabilities. an intrinsic instability of solid fuels.4. Extensive reviews of acoustic instabilities in solidpropellant rockets have been published (for example. Intrinsic instabilities of detonations were considered in Section 6.1. Acoustic Instabilities in SolidPropellant Rocket Motors 295 thoroughly.1. [6]. In extreme cases the “secondary peaks” of pressure became large enough to cause structural failure of the motor. as exhibited in coolflame phenomena. although under appropriate experimental conditions their effects can be observed.
quiescent state and neglect body forces and all transport fluxes. momentum. (446) and (447). a(V v)/& may be eliminated with the aid of the linearization and use may be made of equation (4) to obtain a2p'/at2 .which implies that with p = p(l + p') and p = p(l + p') (where jj and p are constant and where p' and p' are small compared with unity). To develop acoustic equations for the product gas in the chamber.enough for terms quadratic in them to be negligible. a velocity potential cp may be introduced such that * i w ~ b ~ C L . The equations for conservation of mass. but by making suitable use of equation (449) we need not retain equation (447). $ (. 1 %M v = VV. Then for acoustic perturbations. then the velocity field may be obtained from the time integral of the equation aqat =  (p/p)vp'. (7) .296 Combustion Instabilities in Figure 7. we first consider small departures from a uniform.a2V2p' = 0. From equation (449) we see that if the flow is either chemically frozen or in chemical equilibrium. and energy are then equations (449. Equations (445) and (446) can be written as aplat + v .3.which may be shown to reduce to the ratio of specific heats for an ideal gas with constant heat capacities. respectively. it remains constant. We assume that this condition applies and write the sound speed as a = Jdp/dp. (6) which follows from equation (2).(pv) = o (1) and aviat + (vp)/p = o (2) if v * Vv is neglected in equation (446) on the assumptions that the mean velocity is negligibly small and that velocity fluctuations are small. P ' = YP'. and if the entropy is uniform at some time. If p' is known. with burning assumed to occur only at the inner surface of the propellant charge and to go to completion in a distance small compared with all characteristic chamber dimensions. then the entropy does not vary with time along a particle path.5. (5)  which is the wave equation of acoustics. Alternatively. (3) where the derivative is evaluated at constant entropy and at either frozen or equilibrium conditions (compare Section 4. P I =  (ima d a t . (4) where y = a2p/jj.4) in the uniform state. By taking the time derivative of equation (1) and the divergence of equation (2). d p = a2dp.
. Acoustic Instabilities in SolidBfopellant Rocket Motors consistent with equation (6). . . k = m = 0. 1. The general oscillatory solution to equation (9. The simplest boundary conditions are for rigid walls. and they have velocity nodes at nz/l = 0. . . The method of separation of variables. .O k m n r . Most of the normal modes describe standing waves.(a)/da = 0. Thus each normal mode of oscillation [each term in equation (S)] defines a monochromatic wave. cp also satisfies the wave equation. as wk. on the boundary conditions which must be applied at the ends of the chamber and at the surface of the propellant. 3. at the upstream ( z = 0) and downstream ( z = 1) ends of the chamber. Bkmn and Ykmn are constant phase angles. this is called degeneracy and admits an infinite variety of monochromatic wave forms (for example. subject to these conditions. (S) where k.. see equation (6)] at the propellant surface (Y = R ) and that (Vp’)norma... according to equation (S). those with rn = n = 0 ( k # 0) correspond to radial modes. There are special shapes of chambers for which more than one mode may have the same frequency. that the normal component of velocity must be zero [and therefore = 0. The derivation of equation (5) may be shown to require that I v/a I be small. From equation (9) we see that the lowest (nonzero) frequency corresponds to the fundamental axial mode ( n = 1. [ S ] ) . Frequencies corresponding to tangential modes with rn = 1.. . m and n are integers. J .1.. a m k IS the kth root of the equation dJ. and those with k = n = 0 (m # 0) represent tangential modes. bkmn and ckmn are amplitude constants. is the Bessel function of the first kind of order m. . .2./‘ arnkl/~12(nn/02. Oscillations with only one frequency are monochromatic waves. for example. shows that solutions representing oscillatory wave propagation in a cylindrical chamber are conveniently expressed as a sum of terms. waves having nodal points for the velocity (points where the velocity is always zero) and for the amplitude of the pressure oscillations. 2. are determined by [bkmn (9) Terms with m = k = 0 ( n # 0) describe longitudinal modes. of course.. namely. w = an/l) when 1/R is sufficiently large and to a tangential mode (k = n = 0. is [ S ] p‘ = k=O m=O n = O 1 1 Jm(amkr/R)cos(nnz/j) cos(mcp + O k r n n t . grnk = 0. + .9. Properties of the solution to equation ( 5 ) are well understood (see. = a.P k m n ) + Ckrnn cos(rncp. for example. . Thus. longitudinal modes have pressure nodes at nzll = 3.Ykrnn)]. w = aarno/R)when 1/R is sufficiently small. and the (circular) frequencies w k . 5 have been observed [ 3 ] . each of which is given by the product of(a Bessel function of the radial coordinate g(a trigonometric function of the axial coordinate zka trigonometric function of the azimuthal angle andka trigonometric function of time) The precise form of the oscillatory solution depends. tangential modes in cylindrical chambers). .
we may treat the thermodynamic quantities p and h as unique functions of p in an expansion through second order about the uniform state.Ckmn. This enables questions of the occurrence of acoustic instability to be addressed by considering mechanisms of damping and amplification of acoustic waves.p / p . which may be written as u = h . and Ykmn. Nevertheless. With the definition p = j(1 p’) and with j j and h representing the values in the uniform state.t . the boundary conditions influence the solutions [for example.3. Since entropy changes are of higher order. Transverse modes (those having n = 0) represent standing waves if bkmO= & ckmO. the amplitude remains constant along cp = f cok. quiescent state can modify the modes of oscillation.2. if there were an open isobaric boundary.1.. Conservation of acoustic energy To see how damping and amplification affect amplitudes of acoustic vibrations. Between the nodal and antinodal positions. antinodes (points at which the amplitudes achieve maxima) occur for pressure at nz/l = 0.. evidence favoring traveling tangential waves in rocket motors has been obtained [3]. . Departures from the assumption of smallamplitude oscillations about a uniform. we then obtainfrom a Taylor expansion through second order + .1. the pressure and velocity oscillations are out of phase by an amount n/2. From this example. let us first define the acoustic energy. In some cases.. Multiplying this quantity by the density and using equation (1lo). . as do velocity nodes and pressure antinodes. Also. The component of velocity in the direction of propagation of a traveling wave is in phase with the pressure amplitude. the results that have been given form good first approximations for use in approaches to the analysis of acoustic instabilities. and equation (8) would be changed]. as may be verified by use of equation (6). If m # 0 and either bkmO = 0 or ckmO = 0. where u = (vl. where h is the enthalpy per unit mass. then the transverse mode describes a traveling wave that propagates in the f q direction with angular frequency okmO (that is. . in that if one is proportional to cos(o. then p’ = 0 would be appropriate there. Identification of the energy in an acoustic field followed by development of a balance equation for the acoustic energy is a useful approach to the estimation of magnitudes of amplitudes. the other is proportional to sin(o. . For these modes.298 Combustion Instabilities may be found by use of equation (6). ..t /loon). we obtain an expression for the energy per unit volume. Nothing in what has been presented so far provides a basis for calculating the constants bkmn. ..2.ot/m). 9. and for velocity at nz/l = +. The sum of the internal energy and kinetic energy per unit mass at any position and time in a gas is u + v2/2. we see that pressure nodes and velocity antinodes coincide. /?kmn.peon). Frequencies and spatial dependences of amplitudes are less strongly influenced by flow nonuniformities that are the magnitudes of the amplitudes of the various modes.
to second order with p = p(l p’).1. A similar derivation of equation (10) may be found in [ 9 ] . and it is not counted as an energy density of the sound field. Acoustic Instabilities in SolidPropellant Rocket Motors 299 + pp’d(ph)/dp + ( j ~ [’ d)2 (~ p h ) / d p 2 ] / 2. ) + pp’h + (pp’)2/(2pa2) + pv2/2. Differentiating this result with use of equation ( 3 ) gives d2(ph)/dp2 = l / ( p a 2 ) hd2p/dp2. Here ph . d(ph)/dp = 1 h/a2 in the expression for the energy of the acoustic field. By employing these results in the preceding expression.p . (4). we see from equation ( 3 ) that a shorter expression for the above energy is + If this expression were integrated over a volume containing a fixed mass. we have pp’ = pp‘ d p / d p (Ljp’)’(d2p/dp2)/2. Since. instantaneous energy per unit volume in the acoustic field is therefore defined as e = ~ p ’ ~ / ( 2 y ) pv2/2. + (10) with y defined after equation (4). and (101. Equation (10) shows that since p‘ and v are firstorder quantities. Since the linearized form of equation (1) is ap‘/at = . The local. The partial derivative of equation (10) with respect to time is aept = ( ~ j / y p’ap’/at ) + pv . and . The substitution of equation (6) here then gives ae/at = pp‘ ap‘/dt .p represents the energy of the uniform state.9. The derivation of equation (11) has neglected the mean flow of gas in the chamber. to second order. Since d h = d p / p under the present conditions according to the definition of h and to equation (449) and the equation preceding it. (6). (1 1) which is the simplest form of the desired conservation equation. we see that  &/at + pv . we find that the portion of the energy per unit volume associated with the acoustic field may be written as + + jp‘h/a2 + (Ljp’)’[hdzp/dp2]/2 + ( p ~ ’ ) ~ / ( 2 p+ a~ pv2/2.(P’v)= 0.jp‘ + p v 2 / 2 for the energy per unit volume. A conservation equation for emay be developed from equations (l). the first term thus is associated with energy changes that are due to mass changes. in which ap’/at = yap‘/at may be employed by use of equation (4).av/at. instantaneous flux of acoustic energy. and the rest of the expression is therefore to be associated with the energy of the acoustic field. the energy e is a quantity of second order.pv Vp’. Equation (1 1) shows that pp’v may be interpreted as the local.V * v. although expressions for p’ and v valid to first order may be employed in the equation to evaluate e. then the integral of the first term here would vanish because the volume integral of pp’ is zero. spatial variations of mean thermodynamic properties. where use has been made of equation (3).
Let angular brackets identify the time average. then a further decomposition of p’ into a sum over degenerate modes can be made. [17]. all cross terms average to zero. boundary conditions may be applied at a nozzle entrance plane having a low Mach number [4].t  fi) is zero for o1# o2. with use made of special analyses [18][21] of acoustic waves in the nozzle flow for ascertaining the appropriate boundary conditions.t . the average acoustic energy is ( e ) . a situation that usually is . then the integration may be performed over one period of oscillation.o. that is. A timeaverage value of e may be defined by integrating over a suitable time period. phenomenological manner by adding to equation (11) the divergence of the rate of convection of acoustic energy by the local mean velocity V and the local instantaneous rate of dissipation of acoustic energy per unit volume @. then the integration is extended over a time long compared with the longest oscillation period. With p’ represented as a superposition of monochromatic waves. If it is a superposition of monochromatic waves.(eT) + @ = 0. The modified conservation equation then becomes ae/dt + pV . in which V is to be obtained from steadyflow analyses with acoustics neglected and expressions for @ may be derived by considering specific mechanisms of dissipation. Although the Mach number is not small in a choked nozzle. and double sums may be needed for ( e ) and (p’v). To take these effects into account by starting with the full set of conservation equations is a complicated problem that has been addressed partially in the literature from various viewpoints. The Q of an acoustic cavity is essentially the ratio of the time over which (e) varies to the longest oscillation period. Especially challenging is the task of including flow nonhomogeneities that are expected to lead to refraction of sound waves and to be of significance in real rocket chambers.300 Combustion Instabilities homogeneous dissipative effects such as the viscous terms in the equation for momentum conservation. If the acoustic field is monochromatic.(p’v) + V . both ( e ) and (p’v) also may be written as a sum over monochromatic waves because the average of cos(o. To avoid a lengthy development of equations. the simple additivity of ( e ) is lost. A number of studies have been directed toward investigating the interaction between the sound and the flow [lo][16]. we may include effects of mean flow and of homogeneous dissipation in an ad hoc. but wave interference may prevent ( e ) from being expressed as a single sum over degenerate modes. usually on the basis of expansions for small values of the Mach number of the mean flow. If there is degeneracy for a particular o. and the averaging procedure is useful only for high Q. A conservation equation for ( e ) may be obtained from the time average of equation (12) if the time over which ( e ) varies appreciably is long compared with the averaging time.
(13) When balance equations for acoustic energy are used. Let the volume average of (e) be 2 = (e) d V / l v . Equation (14) shows that the homogeneous dissipation. the convection term vanishes if the acoustic energy convected into the chamber is convected out at the same rate. A boundary at which a rigidwall condition (5 * n = 0) or an isobaric condition (p' = 0) is exactly applicable clearly has (p'v) * n = 0 and therefore no boundary work. (16) where go is the value of 2 at t = 0. they almost invariably refer to time averages of the type indicated here. pp'vndd  S . The acoustic admittance A clearer interpretation of the boundary work may be obtained by neglecting convection and homogeneous dissipation and by focusing attention on a monochromatic wave field of frequency w. QdY. and define an amplification rate a by sv a= [. 9. Depending on the sign of a. p(p'v) n d d   s.1. Acoustic instability therefore may be considered to occur when ct > 0.S . e V * n d d S . Of the three contributions to a that appear in equation (15). The problem of evaluating acoustic instability then reduces to the problem of calculating a. if a remains constant. d edV =  S . Acoustic Instabilities in SolidPropellant Rocket Motors 301 encountered in practice. The volume average also may be applied to equation (13) to obtain an equation like equation (14) for (e).1. Space averages of e also may be defined by integrating over a fixed volume V of the chamber. (15) Then. (@) dV]l(Ziv). (14) where n is the unit outwardpointing normal vector at the boundary and the area integrals are carried over the entire boundary. Use of the divergence theorem in equation (12) shows that dt f.3. For . and the boundary work all influence the average e in the chamber.((e>V) + (0)= 0. the net convection of acoustic energy across the boundary. the acoustic energy in the chamber either increases or decreases exponentially with time.9.e2"'. The first term on the righthand side of equation (14) is the rate at which the boundaries do work on the acoustic field. (e) 5 n d d 9 J'. the balance like equation (14) for 2 gives 2 = c. the average of equation (12) is d(e)/dr + pV*((p'v)) + V. we shall be concerned mainly with the first and the third. With the restriction to high Q.
y).302 Combustion Instabilities monochromatic waves. corresponding to energy removal from the vibrations at the boundary. respectively. the normal specijic acoustic impedance. . an increase in pressure is associated with a positive outward velocity in the former case and with a negative outward velocity (a positive inward velocity) in the latter. The average rate of energy input per unit area to the acoustic field is found from equations (18) and (21) to be expressible as P (p’v) * n =  [(pP’)z/2]Re{CiY}.p) cos(ot  y) dt(w/2n) = . corresponding to energy addition into the vibrations at the boundary. where the constant P‘ and V are the positive amplitudes of p’ and v n. if rc < 6 < n/2 or 4 2 < 6 I rc. the boundary work depends on the phase difference between p’ and v * n. where Re{ } denotes the real part. . and p and y. It can be shown. Equation (17) may be written as = prei(wtP). This is negative. (18) where 6 = p . are their constant phase angles. The average rate of work done per unit area on the acoustic field at the boundary point then is   (17) p(p’v) . if .y is the phase difference between pressure and outwardnormal velocity oscillations.p).4 2 < 6 < 4 2 and positive. p’ = P v*n = V cos(ot . With this notation. Complex representations of acoustic waves are convenient for many purposes. that a traveling wave incident on the boundary is reflected with a diminished amplitude if Re(9Y) > 0 and with an increased amplitude if Re{Y} < 0. The admittance and its reciprocal. boundary conditions near velocity nodes are expressed most easily in terms of the complex acoustic admittance 9/ of the boundary. play many roles in analyses of monochromatic wave fields [22]. = Vei(wtY) (19) with the understanding that the real parts of these expressions represent the physical variables.n = .PP’V Jozz’w cos(ot . where v n and p’ are given by equation (19). On the average. We see from equations (19) and (20) that  the real part of which has the same sign as cos 6. Suppose that at the boundary point of interest ‘ cos(ot . which is defined as the ratio of the complex outward velocity perturbation to the complex pressure perturbation.(PP’V cos 6)/2. for example.
(27) where equation (26) and the definition of y following equation (4) have been employed.9. This enables us to obtain a simplified expression for P for monochromatic waves. the amplitude measure appearing in equation (28) is the same as that in equation (22). then substitution into equations (5) and (6) yields the Helmholtz equation for P. then iwv =    (j/P)VP (26)  J v (pP2/y)d V + Jv (PV V) d V = 0. .obtained by use of equation (25) implies that if either P = 0 or V n = 0 at all boundaries. a=  (yu/2) L IPI2 Re{y} dd/ lPI2dK v (29) where the definition in equation (23) has been introduced. where the complex amplitudes P and V are functions of position.P2w2/u2. (28) where ] P I 2represents the square of the magnitude of the complex pressure amplitude. it may be shown that the 2 defined above equation (15) is expressible as 2 = PJvIPI2dV/(2yV) = p L IV*VldV/(2V). Since I P I = P' at the boundary according to equations (19) and (24). Acoustic Instabilities in SolidPropellant Rocket Motors 303 An appropriate nondimensional measure of the admittance is Y = +v/4 (23) in which p/u is the characteristic impedance of the internal gaseous medium. The identity V (PVP) = VP VP . Equations (22) and (28) may be used in equation (15) to show that for conditions under which V = 0 and = 0. and y = 0 may be introduced as a first approximation at these boundaries in analyzing the wave field. From equation (20) we see that y = 0 at a rigidwall boundary and y = 00 at an isobaric boundary. thereby facilitating amplificationrate estimates. V2P + (w'/u')P = 0. both large and small values of the magnitude I y 1 may be expected to be found at various positions within a chamber. thus. v = Re{Ve'"'} (24) are introduced. lyl < 1. (25) and the relationship for V. By use of equation (24) and the result given in equation (27) in the definition of e that appears in equation (lo). If the complex representations p' = Re{Pe'"'}.1. At solid boundaries like the propellant surface in rocket applications.
1.304 Combustion Instabilities The contribution of the boundary work or of radiation of acoustic energy through the boundary to the rate of amplification of the acoustic field within the chamber can be estimated more readily from equation (29) than from equation (15). Since the derivation of equations (15) and (16) implicitly required I c1I 6 w. We shall not discuss these modifications. the fact that energy densities averaged over a cycle are additive may be considered to justify application of the present results separately to each frequency for acoustic fields composed of a superposition of monochromatic waves.1. Therefore. there are? modifications to the natural frequencies of vibration associated with the imaginary part. then the boundary would be well matched with the acoustic cavity. 9. Studies of boundary effects on acoustic instability in rockets therefore involve expansions for ( y ( < 1. we see from equation (30) that IReol6 1 is needed for the approach adopted to be valid. and .4. The approximate constancy of CI as oscillations grow or decrease in magnitude arises from the cancellation that results from the presence of I PI2 in both the numerator and the denominator of equation (29).4 Damping mechanisms 9. then nonlinearities begin to become important. Although attention here has been focused on monochromatic waves. if lyl were of order unity. and all the acoustic energy could be transmitted through the boundary in a time on the order of an oscillation period.1. Therefore. the main task in analyzing influences of boundary work on amplification or damping of acoustic energy may be viewed as that of finding the real part of the acoustic admittance at the boundary. Relative importance In the linear range. a dependence consistent with equation (16). results for the CI of equation (15) may be obtained separately for the various mechanisms of damping or amplification. for example] are on the order of w z u d / K since V / dis a characteristic chamber dimension. In addition to the growth or decay associated with Re{y). equation (29) shows that CI M o Re{y}. If amplitudes of oscillation become sufficiently large. Natural vibrational frequencies of sound in the chamber [see equation (9). Im{y}. In equation (28) it may be considered that 2 varies with time only because the pressure amplitude I PI is proportional to eat at every point in the chamber. and departures from the exponential dependence on t develop. which are small when [ y1 is small. (30) where Re{y} is obtained from a suitable average of the nondimensional admittance over the boundary area.
Acoustic Instabilities in SolidPropellant Rocket Motors 305 the values of a for each may be added to assess the stability. in a manner presumed known from a quasionedimensional. Losses through the choked end nozzle often are most important for acoustic modes that have longitudinal components. steadyflow nozzle analysis [20]. for example.2. Here we shall briefly consider theories for each of these effects. Damping mechanisms provide negative contributions to a. especially for modes having radial components. but homogeneous gasphase dissipation through viscosity and heat conduction generally is entirely negligible except at high frequencies. There are many mechanisms of acoustic damping in chambers. A somewhat more thorough presentation is given in [7]. then special damping devices (related. Solid propellants often have fine condensed particles in their combustion products. and there is a useful summary in [12]. p = p(l + p')] for the streamwise velocity u as wellthat is.1. Viscous and (to a lesser extent) heattransfer processes at the walls can provide nonnegligible contributions to damping for chambers with relatively large wall areas.9. and viscoelastic damping within an internally burning propellant grain then may become significant. where it may be responsible for the general absence of very high modes. timedependent forms of equations (445) and (446) by introducing linearizations of the previously indicated type (for example. The propagation of longitudinal acoustic waves in choked nozzles has been analyzed on the basis of the onedimensional. and these particles can provide appreciable damping.4. Nozzle damping. 9.1.3 for definitions of these damping devices . Chemical relaxation processes in the gas may produce small contributions to damping. The chamber walls and the propellant itself can participate in the oscillations. to the acoustic liners [23][26] or baffles [27] that have been used for sound attenuation in liquidpropellant rocket motors and in other equipment) may be employed. u = V(1 + u')and by allowing the mean quantities p.* We shall not address these topics further. The perturbation equations ap'lat + ij(ap'/az + au'/az) = o auyat + aul/az = pi (31) 2u7 d q d z ~ / ( p i j )a ] and  ppz (32) * See Section 9. Many additional damping mechanisms may be envisioned under various circumstances [7] . beginning first with processes of boundary damping. p . and ij to vary with the streamwise distance z through the nozzle. if problems of acoustic instability are encountered.
Abramowitz and I.2[1 + ik/(y + l ) ] x d P / d x .z m t . 1 . Methods of numerical integration are available not only for these quasionedimensional equations but also for twodimensional equations that allow for modes having components that are not longitudinal [28] .* where c = 1 + * The notation here is the same as that of M. . With the subscript t identifying throat conditions. A. If 6 is a linear function of z. that would apply if the nozzlechamber interface were to occur at that point. where M = fi/u is the local Mach number. then equations (33) and (34) may be reduced to the hypergeometric equation [20]. Subtracting equation (33) from equation (34) yields an expression for V that can be substituted into equation (33) to provide a secondorder differential equation for P. with Y = (fi/U)(V/P) (35) evaluated at each point to obtain the local nondimensional admittance. In terms of the hypergeometric function F . the arbitrary constant in the regular part of the expansion of the solution may be assigned any convenient value. where (36) k = N z .1)M2]. it may be shown that when 1 7 is linear in z . New York: Dover. Handbook of Mathernaticul Functions.l)(dF/dz)P = 0. Stegun. the constant multiplying of the singular terms may be set equal to zero. 1965 .306 Combustion Instabilities are obtained.[ik(l + ik/2)/(y + 1)]P = 0 may be derived. and the differential equation x(l . Equations (33) and (34) may be expanded about the throat. in which equations (4) and (24) (with v replaced by u') may be employed to derive the pair of firstorder ordinary differential equations yodV/dz and yV dVidz + fidP/dz + i o P = 0 (33) (34) + (u2/V)dP/dz + y(2du/dz + iw)V .x ) d 2 P / d x 2.x). the solution to this differential equation that remains regular at x = 1 is F(u. and a numerical integration may then be performed in the direction of decreasing z . and the condition that waves not be transmitted upstream from the supersonic section of the nozzle translates into the requirement that the solution in the subsonic part remains regular at the throat. threedimensional nozzle calculations have also been performed [29]. then x = (U/Ut)2 = (y + 1)M2/[2 + (y .El). in which the subscript 1 identifies conditions at the nozzle entrance. The nozzle throat is a singularity of this system because 3 = u there. c. defined in equation (23).(y . b .
9.Re{y. from equation (35). and nozzle losses become large. that Re{y. invalidating the approach of Section 9.3 and changing the wave form within the chamber substantially. which generally is small. There exists a degree of experimental verification of the predictions [30]. The real part of the nondimensional nozzle admittance yy. expansions for large and small values of k may be obtained more generally from equations (33) and (34). On the other hand. as a function of the nondimensional frequency k for various values of the nozzleentrance Mach number M .3. Also. a b = 2[1 ik/(y l)] and ab = ik(1 can be used along with the expression I/ = [ ( y  + + + + ik/2)/(y + 1). for the small values of M . The result is shown in Figure 9.} approaches l/y as k approaches infinity.1.1. for long nozzles or for values of M I approaching unity.} is small. without the restriction to a linear variation of V with z. It is found from the solution for V linear in z that Re{y. with y = 1. as assumed in Section 9.X)dP/dx]/[(2 + ik)y] to calculate y. Acoustic Instabilities in SolidPropellant Rocket Motors 307 0 2 k 4 FIGURE 9. It is seen from this result that for small M I . that are of greatest practical interest may be obtained from the expansion of F for small value of x [ZO].} approaches M.} > 0.2.1 for y = 1. and that Re{y. The last of these results is of greatest practical interest because [as seen from equation (36) for small MI] k is the frequency made nondimensional by the ratio of the length of the subsonic portion of the nozzle to the sound speed at the throat and is therefore roughly on the order of the ratio of the subsonic nozzle length to the chamber length.2 for longitudinal modes in a nozzle having a velocity linear with distance [20]. Transverse modes are predicted to be much less strongly influenced by the end nozzle than are modes with longitudinal components .1./M. [31]. Re{y.1.( y + 1)(1 .(y .} begins to become of order unity. This solution 1 + ik)P . Expansions of y.1)/(2y) as k approaches zero.
that the thickness of the oscillatory boundary layer is on the order of . the gas temperature in the chamber. Wall damping Damping by wall friction may be addressed on the basis of the theory of oscillatory boundary layers [35]. where T . the oscillatory boundary layer at the wall is much thinner than the meanflow boundary layer [37]./T. (37) where a noninertial coordinate system has been employed in which the wall oscillates and the gas just outside the oscillatory boundary layer is at rest. The solution to equation (37) is readily found to be (38) v = (T. Under conditions in real motors.) of velocity oscillation parallel to the surface. where equation (24) has been employed for the acoustic field in the chamber. from equation (38). the rate of energy loss by a pistonlike mechanism is roughly pal P I 2 d 2 / l 2 . so that the pressuregradient term will be cancelled by a noninertial effective bodyforce term in momentum conservation. where 1 is the wavelength of the mode and the amplitude 1 PI is evaluated at the vent location. [32][34]. Here the subscript w indicates that the density p and viscosity p are to be evaluated in the gas at the wall. with the coordinate system locally aligned in the direction of velocity oscillation. [36].is the wall temperature and T.) boundary layer. and y is the coordinate normal to the wall./T.(1 i)Jwyl).308 Combustion Instabilities [21] and might even be amplified slightly by the nozzle interaction. for a small subsonic vent of area d radiating energy from a chamber mode having its acoustic velocity perpendicular to the vent axis. There is a body of literature on acoustic radiation from subsonic orifices [101[12].4. Neglecting locally the effects of density changes and the spatial variations of acoustic amplitudes. and the tangential acoustic velocities just outside times those outside the meanflow the oscillatory boundary layer are (TWIT.1. Dissipation in the oscillatory boundary layer may be analyzed by considering a flat element of the surface exposed to the complex amplitude V (T'. the timeaverage rate of energy dissipation per unit wall area can then be shown to be ' + This outline of a theory for wall dissipation shows first.)Re(Vexp[icot . 9. whence u = (Tw/TJ Re{Ve'"'} at y = 0 and u = 0 at y = co are boundary conditions. we write the combination of equations (12) and (15) for the velocity u parallel to the wall as P w W d t = a(Pw W a Y ) / d Y .3.
then an analysis of the temperature oscillations. Here p / ( p a ) is on the order of a molecular mean free path [see. for example. in the lower modes the ratio dimension. of homogeneous damping to wall damping is on the order of the square root of the ratio of a mean free path to a chamber dimension. The dissipative energy flux is seen from equation (39) to increase with the frequency in In the same way that acoustic velocity oscillations proproportion to duce an oscillatory velocity boundary layer. 9. shows that the ratio of the rate of dissipation of acoustic energy by the oscillatory temperature boundary layer to that by the oscillatory velocity boundary layer is approximately C(a/y)’(IPl’/I VI’).1.4. which clearly is a small number under all conditions of interest. wall dissipation by velocity oscillations usually is greater than that by temperature oscillations. To see what those special conditions are we may first consider undamped oscillations in a onedimensional. acoustic temperature oscillations give rise to an oscillatory temperature boundary layer at the wall. Solid vibrations Homogeneous dissipation within a viscoelastic solid propellant can produce significant damping if the propellant participates in the oscillations.1.1.l ) ( T J f w ) d K in w which . 1 and c p are the thermal conductivity and the specific heat at constant pressure for the gas. The amplitude of acoustic temperature oscillations is proportional to IPI(y . while for lower modes. equations (E27) and (E34)]. d . If the wall temperature remains constant. Since the constant C usually is less than unity. Homogeneous damping The homogeneous rate of dissipation of acoustic energy per unit volume by viscosity in the gas is of the order of pl VI’/(a/w)’. but under special conditions it does. as is V / d . for viscosity.5. Usually the solid does not participate appreciably. since a / o is the wavelength that determines the velocity gradient associated with the sound field. so long as the solid temperature does not exactly follow the temperature oscillations of the gas (which is usual because of the greater thermal capacity of the solid).4. Acoustic Instabilities in SolidPropellant Rocket Motors 309 which practically always is less than 1 mm.9. twomedium system having . a/w is of the order of a characteristic chamber Therefore.l)/y rather than to I Vl. From this estimate and equation (39) we see that the ratio of the homogeneous rate of viscous dissipation to the rate of viscous dissipation at the walls is roughly J[p/(pu)](w/u) ( w / a ) ( V / d ) . &.4. The proportionality of this indicates that high modes can be relatively strongly damped ratio to 03/’ by homogeneous viscous dissipation. where C = (y . analogous to that indicated above for the velocity oscillations. 9. Similar estimates can be made (with similar conclusions) for homogeneous dissipation by heat conduction between temperature peaks and valleys of the sound waves.
Natural vibrational frequencies as functions of the fractionf occupied by gas for a onedimensional. where the relatively small value of R was selected to enable the relationships among the curves to be seen more easily. Also shown in Figure 9. we may readily derive the equation for the natural vibrational frequencies. twomedium system [38]. where 1 is the length of the cavity and f is the fraction of its volume occupied by gas [38]. and we enforce zerodisplacement conditions (dp’ldz = 0) at z = 0 and at z = 1 as well as equaldisplacement and equalpressure conditions for solid and gas at their interface./a be the ratio of the sound speed of the solid to that of the gas. f FIGURE 9. . where A = cos (“:f)/ cos ~ [‘““h. Typically.2 0.2. t ’ ] is the ratio of the pressure amplitude of the oscillation in the solid to that in the gas. R is on the order of lo3 and S is on the order of unity.4 0.a. with the a for the solid representing its dilational sound speed.2.)/(pa) be the ratio of impedance of the solid to that of the gas and S = a. z =f l .2 are frequencies for gas modes that correspond to vibrations of the gas only (with zerodisplacement conditions applied at the propellant surface) and frequencies of solid modes for solid vibrations with freeboundary conditions applied to the solid at both of its surfaces. for simplicity we may apply equation (5) separately to each medium. Wl  2aa First solid mode 0 0 1 1 1 0. solutions to equation (40) for R = 10 and S = 3are shown by the solid lines in Figure 9.310 Combustion Instabilities gas in the region 0 < z <f l and propellant in the region f l < z < 1.8 I. Although the wave equation for the solid involves two sound speeds [7].6 0. Letting R = (p.
5). is much larger in the solid than in the gas (that per unit mass. Relaxation damping Homogeneous dissipation by chemical or molecular relaxation processes in the gas may be addressed on the basis of a formulation like that given in Section 4. e/p. Resonance between gas and solid oscillations occurs at mode crossings. For values off such that oscillation frequencies nearly coincide for a gas mode and a solid modenamely. is comparable).f ) z (m + $)fSmore careful investigation of equations (40) and (41) is needed. the resonance may be responsible for an intermittent disappearance of instability. but the frequency eigenvalues never cross.4. we should focus our attention on gas modes. then equation (40) requires A to be large. sharp transitions between gas modes and solid modes occur when R is large. e/p. then o l f z nna. Acoustic Instabilities in SolidPropellant Rocket Motors 31 1 For large values of R . 9. Since the gas fraction f in a rocket chamber increases with time during burning. for which we then see [from equation (41)] that the pressure amplitudes are of the same order of magnitude in the gas and solid.1. This condition defines solid modes having a freeboundary condition at the gas interface. At resonance. This condition defines gas modes. while [for example. with roughly equal velocity amplitudes in the two media but a much larger pressure amplitude in the solid.4.1. n(1 . where n is an integer. we see through equation (10) that for solid modes the acoustic energy per unit volume.3. As f is varied continuously. except at the modecrossing conditions of resonance. while for gas modes the acoustic energy per unit mass. Wave equations arise in which a relaxation time z .f ) z (m + $) naS with m an integer. Therefore. From these results. partly because of viscoelastic damping but mainly because their freeboundary (constantpressure) interface condition removes the major source of amplification (see Section 9.6. Solid modes are not observed in acoustic instability of solidpropellant rockets. while between crossings the large impedance difference of the two media effectively isolates one from the other. is much larger in the gas than in the solid (that per unit volume is comparable). this may be true even if viscoelastic damping is negligible because the transfer of acoustic energy into the solid reduces that in the gas. which [from equation (41)] is seen to imply that ol(1 . appreciable acoustic energy from the gas is transmitted into the solid and may be dissipated there viscoelastically. With R large. equation (40) shows that if A is of order unity. A number of analyses of influences of solid vibrations [38][41] contribute to the understanding of their unique attributes. and we see that the oscillations of the solid are likely to be negligible.2 in the vicinity of the modecrossing points (intersections of dotted lines) occurs. from equation (6)] the velocity amplitudes are much larger in the gas. and it is found that A is of an intermediate order of magnitude and that the frequency behavior seen in Figure 9.9. if it is not true that o l f nnu. e.1.
as needed to satisfy boundary conditions for normal into equation (4113).1 is small but positive. [43].lO’w. modes. compare Section 4.4. particle damping can be of greater significance. The theory of sound attenuation by spherical inert particles in the gas is well developed [44][49]. and s are constants.Here the frequency w is real by definition and b .4.7. as was done in Section 4.o/uf  l)(wz)/[l + (oz)”. (42) More thorough derivations of formulas equivalent to equation (42) may be found in the literature [42]. The solution to the resulting equation for B is B = . generalized to threedimensional flow by representing the velocities .4.a / o . by substituting 2) = Re{ Veikx+st} where V.4). we obtain zs3 + bsz + zw2s + o2= 0. The contribution to the previously defined amplification constant CI = Re{B}. respectively. where b = u:o/u:o and o2= a f o k 2 . k . Jlul(pw) is the previously identified thickness of an oscillatory boundary layer on a flat surface.4. for frozen or equilibrium behavior. Because of this we may set s = io + B and assume that 18) < 101. at o z = 1.3.(w/2)(u. For example. Instead of looking for solutions precisely harmonic in time. 9. a onedimensional version was written in equation (4113). Particle damping As indicated in Section 9. vanishes as wz approaches zero or infinity and attains a maximum value of ( U ~ ~ / U ~ 1)/4 . which exhibits both wave damping and a small decrease in frequency. and 1 is the wavelength of the acoustic field. evaluating sz and s3 by expansion to first order in I B/w 1. Gasparticle flow with Stokes drag was considered in Section 4.1.(~0/2)(b. is a = .6. to the order of approximation to which this result applies.1.312 Combustion Instabilities and both equilibrium and frozen sound speeds appear .1. we may seek solutions harmonic in space.and most chemical times are either too short or too long. values of afo/aeofor the few that are not typically produce CI z . .4. k being a wave number. which is not a large damping effect.3. Thus the dissipation rate is greatest when the reciprocal of the relaxation time equals the frequency of the acoustic oscillations and is negligible for long or short relaxation times (that is. but that of greatest practical interest has rs @ Jam 1.~)(oz ib)/[b2 + + (oz)’].2. The result here shows that the dimensionless decay rate. Here rs is the particle radius. In the notation of that section. A number of different regimes may occur [7]. with the Stokes drag law applicable. Most vibrational relaxations are too rapid to contribute significantly to CI.
9. per unit volume. . . corresponding to o z = 1.2Ps/(9Ph (44) identified in Section 4. by the formula &nrs3Ps) = P&l .p (v. the consequent contribution to the amplification rate is r JY where the response time . it may be seen from equation (440) that the force per unit volume exerted on the gas by the particles is n. then by use of the equation of motion for a particle. = V(1 . v.e'"'}. (47) By substitution of equations (28).6. $nr:p. The dissipation rate 0 of equation (12) is the dot product of the negative of this force with gas velocity v.vs). . and (47) into equation (43). we can show that which reduces to equation (42) with aeo = &fO (49) if Z and z are uniform throughout the chamber.2. has been introduced.c = 2r.p(v. Thus equilibrium and frozen sound speeds can be defined for particlegas mixtures in a manner entirely analogous to that for chemical relaxation. Z .6nr.dv. introduced in Section 4. = Re{V. Acoustic Instabilities in SolidPropellant Rocket Motors 313 as vectors. as becomes clear from an alternative analysis [47].6. Under frozen conditions.ldt = 6nr. If equation (24) is employed for the gas velocity v. The mass of particles per unit volume appears in equation (43) and can be expressed in terms of the gas density p. Thus. (46) [which is readily inferred from equation (440) for example] it is found that v. while under equilibrium conditions (for small particles giving small z) the particles move with the gas. that is.VJ. is introduced for the particle velocity. from equation (19.. (45) Evaluation of the average in equation (43) necessitates investigation of the oscillatory velocity fields. and the mass of gas per unit mass of the system.2. the particles remain stationary. The optimum particle radius for maximum damping. and a similar representation. the negative of the rate of doing work on the gas by the particles. is found from equation (44) to be which ds).1. (49.ioz)/[l + (oz)'].a/z. which occur for large particles [large z in equation (44)].
then a = . with an additional term representing the heattransfer contribution r:G(r. In addition to the velocitylag effect analyzed here.1.1 07 1 1 0 FIGURE 9. which may achieve a value on the order of . The value of a at this maxtypically lies in the range of imum is w(l . see [7]. in the figure.." r.(W/2)(PS/P. as may be seen from Figure 9. Comparison of theory and experiment concerning particle attenuation.) dr. there can be temperaturelag effects on particle damping associated with heat transfer to the particles. being the number of particles per unit volume in the radius range dr.(a/m)3p.G(r.].71 0 0 02 0. and equation (50) was used for the theoretical curve in the figure. 0.4. cl = .2)/(42).) 1 0 m m r : ( ~ M 1 + (mz>'l>G(r. showing dimensionless damping coefficient as a function of dimensionless particle radius [SO]. about r. (50) in which z is given by equation (44).>drs. many different approaches to the analysis of various types of damping mechanisms were indicated. The experiments shown here involved a distribution of particle sizes.2 as large as 0.3.lo' o./[2nps as z in equation (44) with r. Predictions of particle damping have been found to agree well with measurement [ S O ] . The linearity implies that if there is a particle size distribution. .3 are not uncommon for certain metalized propellants.) dr.) dr./f. with G(r.3. The variety of formulations exposed may help to show alternative paths to damping calculations.. since values of 1 . and f is defined added.314 Combustion Instabilities cm to cm. In Section 9. replaced by f : r:G(rs) dr. For a discussion of these effects and for a more detailed presentation of the velocitylag analysis.
T). heat addition occurs in phase with the pressure increases during the oscillation. if we erroneously neglect any velocity changes just to get a quick result. Temkin for this reference.1. Our attention here will be focused on the first mechanism. the amplification at a burning propellant surface is considerably more complex than the phenomena described by equation (5 l). Relative importance Among the mechanisms that may lead to amplification of acoustic oscillations are combustion at the burning surface.5. The third usually is negligible.1. where Q denotes the rate of heat release per unit volume. It explains qualitatively many observations of singing and vibrating flames (and vibrations in tubes containing heated elements as well). on the average. The second would be more relevant to liquidpropellant rockets.9. Rayleigh recognized. finiterate chemistry in exothermic reactions that occur homogeneously throughout the chamber. Of course. Amplification criteria The mechanism of sound amplification by heat release was understood by Rayleigh [Sl]. as implied by equation (51). In solidpropellant rockets. multiplying the second equation by p‘/y and averaging over a cycle.1. by use of the first equation.2. Acoustic Instabilities in SolidPropellant Rocket Motors 315 9. [25]). although it may become significant under special circumstances. * I am indebted to S. quiescent state.1.1.5. much research has been done on this fluidmechanical topic in the field of aeroacoustics. constantproperty. . and conversion of meanflow energy into acoustic energy through aerodynamic noisegeneration processes of various types.5. Qualitatively. since molecular transport processes are involved in an essential way for example (as was seen for the steadystate combustion in Chapter 7). for example. 9. we obtain which in fact turns out to be correct for heat addition to an inviscid. B. and reviews are available (for example. Higgins in 1777. this is the Rayleigh criterion for sound generation by periodic addition and abstraction of heat [52]. described beautifully by Tyndall [53]* and dating back at least to an observation made by Dr. then equation (10) is e = j$i2/(2y) and equation (454) is Dp/Dt = a2Q/(c. However. the pistonlike mechanism implied by equation (51) when the heat release is localized and the result is expressed in terms of an acoustic admittance [as in equation (29)] is tied closely to the amplification by propellant combustion. that amplification results if. Amplification mechanisms 9. the first of these sources generally is the most important. ideal gas experiencing smallamplitude acoustic oscillations about a uniform.
/p. Putting m=E ( 1 m') and assuming that lm'l G 1. v n is the negative of the oscillating part of the gas velocity in the + x direction at x = + m. and since this ratio is small. analysis of the combustion zone usually is performed more conveniently in terms of the mass flux m. on the order of heatconduction lengths for deflagrations. the rate of amplification for solid modes is much smaller than that for gas modes. which ranged from .V2/2times the real part of the impedance.2. typically a few millimeters or less. However. The combustionzone analysis may be performed in a noninertial frame that moves with the acoustic velocity. we employed a coordinate x.m and at x = +so for use in calculating acoustic amplification. the value of which seldom is as large as Therefore. which is roughly (fiw/u)(p.p. seen by the solid at its interface. for a given energy of the acoustic field. In this frame. normal to the propellant surface.00 deep in the solid to + m far into the gas. an apparent body force per unit volume of order povappears in the equation for momentum conservation. v n is on the order of the ratio of the gas density to the solid density times that at x = m. Therefore. it is appropriate to approximate the combustion zone as an interface and to employ admittances to describe its influence on the sound field. The pressure oscillations of the acoustic field are imposed on the combustion zone. but at x = . The difference of the impedance at x = . The difference of these two values arises in the acoustic analysis [39].m. but the latter is negligible because of the density ratio. .1. In terms of the thickness 6 of the combustion zone./p.)/(pu.5 and identified in Figure 9. the ratio of the rate of doing work on the field per unit area at the pressurenode condition of the solid mode to that at the velocitynode condition of the gas mode is on the order of Gpou(u. effectively. appears.r: and that at x = + x. and from the solutions the admittance defined in equation (20) (with v now representing the difference of the velocity from its mean value) is evaluated at x = . r is the ratio of the complex amplitude of p' to the complex amplitude * For the solid modes discussed in Section 9. in comparison with chamber dimensions. we may express y of equation (23) as  + +  + where p is the ratio of the complex amplitude of m' to the complex amplitude of p'.4. is small (since the interface is nearly an isobaric pressure node) and is therefore more appropriate than the nondimensional admittance y (its reciprocal) for use in equation (22).* Since n is the unit normal pointing outward from the chamber in equation (20). from the viewpoint of acoustic instability. it is sufficient to evaluate CiY at x = m for use in amplification expressions such as equation (29). Although it is the velocity that appears most directly in the admittance that is responsible for the amplification. The combustionzone structure must be analyzed subject to these pressure oscillations. where the dimensions and velocities are small enough that the timevarying pressure is independent of x to a good approximation [54]. the distances over which changes occurred were small. Thus the righthand side of equation (22) is best expressed as .). solid modes are not excited. the nondimensional acoustic impedance.).316 Combustion Instabilities In analyzing solidpropellant combustion in Chapter 7.
In the isothermal limit. We shall employ r = l / y and view dissipation in the nonisentropic zone to be counted as an additional damping contribution./o.1. of order v. often it is most convenient to evaluate p just after quasisteady combustion. and r = l / y holds there. . p’ = p’ and r = 1. Models based on the hypothesis of a quasisteady gasphase flame predict a nonisentropic relationship between p and p in the gas at the downstream edge of the gasphase reaction zone. It is strictly proper to employ equation (52) in an acoustic boundary condition for amplification calculations only outside the nonisentropic zone. Acoustic Instabilities in SolidPropellant Rocket Motors 317 of p‘. ~ * Since the ratio of the decay length to the chamber dimension theoretically increases in proportion to the chamber dimension. the periodically generated nonisentropic component is transported with fixed form at the mean gas velocity. a decay time may be estimated )~ by a thermal diffusivity. However. . since equation (449) is not the wave equation. z lo2cm/s and o z 103/s)are found to lie between 1 cm and 10 cm. representative values (with v. / odivided v. the velocity amplitude has been eliminated by use of the expansion m/p = (E/p) (1 + m‘ . for example. this dissipation is much more rapid than the corresponding dissipation for acoustic waves because the wavelength. [56]. Equation (4) implies that r = l / y in equation (52). although this is somewhat of a misnomer. the effect is predicted to be greatest in large chambers. . a complication has been identified on the basis of studies of the combustionzone structure [55]. r may be evaluated in equation (52) from the quasisteady combustionzone analysis. and all quantities are to be evaluated in the gas at x = +a. Although this value assuredly is correct formally. When this is done. small changes in p that may occur between this location and the downstream edge of the reaction zone usually are not addressed in analyses. to obtain a characteristic decay length.* Thus the combustion response may create a region large in comparison with the thickness of the steadystate combustion zone. the density and temperature oscillations are found to be related to the pressure oscillations by conditions that are not isentropic. Since the combustion has been completed. within which the wave field is modified by nonisentropic effects. which in extreme cases may reach a significant fraction of the chamber dimension.9. it is expected that l / y I Re{r} I 1. The consequent spatial and temporal oscillations of entropy have been termed entropy waves. as implied by equation (449).p’ + . By balancing the rate of heat conduction per unit area with the rate of change of thermal enthalpy per unit area in a halfwave. . is shorter (typically lo’ cm to 1 cm). in the hope that use of the value so obtained may account approximately for dissipation in the nonisentropic zone.). and therefore influences of uncertainties in the value of r are not large. say v. and when this is multiplied by as ( ~ . These “waves” decay through nonquasisteady heat conduction between their temperature peaks and valleys.
which requires that n > l/y and that wz be in an appropriate range for amplification to occur. then in m(t) = EL1 + m’(t)] and p ( t ) = p[1 p’(t)]. Among the improvements to equation (54) is the introduction [17]. [57][60]. for example. equation (52) indicates that y is of the same order of magnitude as the Mach number of mean flow of the burnt gas. whence amplification is predicted at low frequencies. in equation (741)].zJ. The principal value of timelag concepts lies in the wide range of problems to which they can be applied with relative ease. we see that a necessary condition for amplification by the propellant combustion is The ratio p often is termed the response function. o < (l/zl) cos. we have m’(t) = np’(t . and if a small pressure pulse is applied to a steadily burning propellant. [59] of a sensitive part of zI that varies with pressure and gives results more versatile than equation (54).iwrl . 9. for example. Timelag theories Many early analyses of timedependent solidpropellant combustion employed the concept of a combustion time lag [4]. appropriate to the new pressure. The interval zl is the time lag. If the onedimensional. at which time m instantaneously assumes its steadystate value. Their principal deficiency lies in the difficulty of relating zl to the fundamental processes occurring. Therefore. or the burningrate response.‘[l/(ny)]. assumed here to be a known constant. assume that a pressure change applied at time t has no effect on m until time t + z l . [17]. This observation enables the order of magnitude of the growth or decay rate to be estimated directly from equation (30). . and combustionzone analyses typically provide magnitudes of order unity for it. A critique and more thorough review of the timelag ideas has been published [7]. More accurate computations necessitate calculating Re{pL).3. at all x. so that + (54) Thus the criterion in equation (53) becomes n cos(wzl) > l/y. Here we shall outline only two and comment on other approaches. timedependent conservation equations are linearized about any one of the steadystate solutions of Chapter 7 and p is calculated from a perturbation analysis.5. The many different possibilities for steadystate structures of the combustion zone indicated in Chapter 7 imply that many different analyses can be relevant to the calculation of p. As the simplest idealization of these observations. then it is found p=ne .318 Combustion Instabilities By use of equation (52) in equation (29). A given element of reactant material requires a certain amount of time to burn. If n is the pressure sensitivity [appearing.1. typically wzl is small. then the regression rate will take a certain amount of time to reach its new steady level.
and yet the results of many of the more recently published theories bear a close resemblance to its results. This idea dates back to the work of Zel'dovich [6 13.E. denote the preexponential factor and the overall activation energy for the surface process.MROT. given by equation (76) but also interpretable in terms of a distributed solidphase reaction in a thin zone (as indicated at the end of Section 7. Acoustic Instabilities in SolidPropellant Rocket Motors 319 to be difficult to express the results in terms of distributions of time lags. ~ . It is of interest to present an analysis of the last of these models because it is relatively simple. lie in the range s to lop4s. With the density.co and T = at x = 0. the equation for energy conservation in the solid becomes 7s = (56) The coordinate system is selected so that the solidgas interface. (57) where the constants B. 9.5 [64]. The interface Arrhenius law.aTlat + mc. and a more complete model has been explored. at x = . Under the assumption of .4). is maintained at x = 0. but it varies with t in this coordinate system.4..9. where conditions are identified by the subscript i. . aTlax 1 .a2T/ax2.1.1.)l. Boundary conditions for equation (56) are T = To. specific heat. but more detailed analyses of the flow processes are better when they can be performed. Thus it is attractive to assume that all processes remain quasisteady except the solidphase heat conduction. m is independent of x everywhere. [62]. Combustion response Although many different processes may be involved in solidpropellant combustion. in which use is made of the deflagration analysis of Section 7. By mass conservation. (55) where the subscripts identifies properties of the solid. which is comparable with typical values of l/o. and E . This latter characteristic time is PsMcpsm2). estimates suggest that characteristic times for all of them may be short compared with the characteristic time for heat conduction in the solid. and thermal conductivity of the solid assumed constant. Representative values of z. who considered timedependent burning within the context of an approximation of constant temperature at the solidgas interface.the initial temperature of the propellant. Timelag approaches are useful for a first look at instability problems (including liquidpropellant rocket instability). The combustion mechanism addressed involves inert heat conduction in the solid. Oscillatory burning has been analyzed under this assumption [63]. is written here in the form = p . m = 4 expr . surface gasification by an Arrhenius process and a gasphase deflagration having a high nondimensional activation energy..5.
5 and equation (5.75)]. adiabatic. gasphase combustion./A.T). and thermodynamics require L..JL. the interface energy conservation condition readily (for example.  731. (dAT/d()i = s ( A T ) ~ pP + (S  l)pP(T .that is. .To)+ E. (58) where L. represented as the sum of the particular solution and a solution to the homogeneous equation. Equations (56)(59) provide a wellposed problem for addressing influences of harmonic pressure oscillations.From the general solution. is the flame temperature (which varies with time). is defined in equation (55).. c. is the energy per unit mass required for gasification.E1/(2RoT’.where s = (1 + J1+4ioz.) (64) . and T. L..(T . (60) By putting T = T + Re{ATei”‘} and linearizing equation (56) about T (under the assumption that I AT I 4 T).T).(T.TO)/(io~. + c.)(T .(dT/d~)i= m[L. . and the solution to the homogeneous equation is a constant times es5. + (c. (59) where C and n are positive constants ( n being half of the pressure exponent for the homogeneous gasphase reaction) and where El is the overall activation energy for the gasphase combustion [see Section 7.). the pressure appears here only in equation (59).q + c. and (59) give = (AT)iE.)’pLP(T .(T.)/2.)]. The perturbations of equations (57).AT + dAT/d( + pPdT/dt = d2AT/dt2.we may readily derive the perturbation equation ioz.). P in equation (24). Analysis of the quasisteady gasphase flame by activationenergy asymptotics [65] shows that also m = Cp”exp[ E1/(2RoT. A particular solution to equation (61) is (ioz.To)e5. (63) may be derived. (62) the positive sign of the square root having been selected to satisfy AT = 0 at 5 = .c.. the relationship between the heatflux perturbation and the temperature perturbation at the interface. (61) where z./(RoT?) = nP + (AT). The steadystate solution to equation (56) is easily seen to be T = To + (T . to vary with if cps # c. and t = xi%. . (58). [65]) may be shown to be expressible as L. q is the (constant) energy per unit mass released in the gasphase reaction.co. is the (constant) specific heat at constant pressure for the gas.320 Combustion Instabilities quasisteady. = L.To)exp(xEc. An overall energy balance may be applied to steadystate conditions (identified by an overbar) to give q = c.
.T3. with n. . for calculating the response function from models that do not rely on time lags.2.( B + 1) (for A > 3 )and ( B 1) J ( B + 1)’ . the height of which depends on the values of A and B and exceeds n for a range of values of A and B. and ( A T ) ./(ROT.. where the envelope curves represent the maximum achieved for any value of A and the envelope asymptote is related to the intrinsic instability. Use of equations (62)(65) to remove iwz.9. ( ~ i A T / d t(AT)i ) ~ .T. These results  + + .( B + 1). variation have been employed with equation (58) in obtaining the last expression. values between ( B + 1) . large. increases for 07. as will be discussed in Section 9. From equations (62) and (66) we find that Re{p} = n{l for wz. reviews are available [7].4. (66) where A = (T . for wz. [75]. from the system then yields / l = nBs/[s’  (1 + A . An idea of the predictions concerning amplification may be obtained from expansions for small and large values of wz. Re{p} exhibits one broad maximum. The present model also exhibits an intrinsic instability if A or A / B is too large.)’[(B + 1)(2A 1) . A few representative theoretical response curves are shown in Figure 9. [66][74]).>E. which occurs for A 2 6 in this case. large values of A and of A / B lead to larger peak responses through the greater sensitivity provided by the high activation energies. A .J ( B + 1)’ . have appeared in the literature (for example.(~p/~ps)(AT)m + (AT)i. small and demonstrate the quasisteady response for wz. if A is small or large but an increasing response for intermediate values of A ..B)s + A]./(E1CP. in the vicinity of wz. Acoustic Instabilities in SolidPropellant Rocket Motors 321 and (dAT/dOi = ~ P ( ’ i i To) .Z) (67) and B = 2E. large and a behavior for small wz. [76]. z z 1. [12]. (65) where the overall adiabaticity and the L. The enhancement is caused by the transient buildup of energy in the solidphase heatconduction zone and is typical of predictions of morecomplicated theories.CPT2. and B appearing as the relevant steadystate combustion parameters.A’]/B2 + Re{p} = nB/& + . Typically. Some of the models admit a certain amount of nonquasisteady + (wz. (68) Equations (62) and (66) provide p as a function of wz. In general. amplification that decreases in proportion to 1 / 6 as wz. Many analyses of the general type outlined here. = 0. showing a decreasing response with increasing wz.1.
particularly with respect to condensedphase and dispersedphase processes. This belief is based on the observations that gasphase combustion is usually at least of small importance for transmitting energy in propellant burning. * It can be shown that by reinterpreting the parameters n. and B in terms of sensitivities of rn top.* In fact. even on the basis of quasisteady gasphase approximations. which can be important at high frequencies. the few results that have been reported for homogeneous propellants tend to agree qualitatively with predictions [77][80]. . These models all pertain to homogeneous propellants.322 Combustion Instabilities Envelope asymptote FIGURE 9. Thus even if the model that has been analyzed here is inaccurate in detail. and T . with A = 28. which allow them to exhibit transient energy accumulation. . Nevertheless. for which it was seen in Chapter 7 that the combustion mechanisms may be complex. while condensedphase (or possibly dispersedphase) processes have longer response times. most of the mechanisms described in Chapter 7 have not been investigated in the context of oscillatory response. A . it seems likely that the main elements of the response function for homogeneous propellants are reasonably represented by the analyses that have been performed. showing a nondimensional measure of the amplification rate as a function of the nondimensional frequency for various values of the nondimensional activation energy for gasification. A . Representative burningrate response curves obtained from equation (66). results for any model that maintains quasisteady gasphase processes can be expressed in terms of the results that have been derived here. it seems likely to retain the qualitatively essential physics for describing response to pressure oscillations. behavior in the gas phase. Although experimental measurements of the response function are difficult to perform (except by inference from measured growth rates of pressure oscillations in chambers).4. A useful philosophy is to attempt to develop general descriptions that are independent of specific details of combustion models [76]. The equations allowing fully nonquasisteady combustion in the gas phase are complicated [54] and have not been explored completely. T i .
To employ a result like equation (66) directly is questionable. wher CI and p are constants [85]. A peculiar aspect of the response of heterogeneous propellants is associated with the periodic uncovering of different propellant constituents during normal regression [85].1.7. nonacoustic oscillations may be superposed (at least locally) on the acoustic oscillations. Description of the interaction necessitates augmentation of responsefunction concepts and entails considering a synchronization time needed for the establishment of a coherent phase of the heterogeneity over the propellant surface by action of the acoustic oscillations [85]. has been explored for calculating the acoustic response [85]. frequency ranges. Response curves like Figure 9. An equation of the form d&/dt =a &  +p (69) therefore is obtained.5. an additional term arises in equation (14) that is proportional to the product of the pressure amplitude and the heterogeneity amplitude. [83]. a simplified modelincluding transient variations in stoichiometryhas been developed [84]. to amplitude thresholds that must be exceeded for growth to occur ( a > 0.4 may be found to possess more than one . then an average is perceived by the acoustic field. if the uncovering time is long. p > 0).7. although attempts have been made to evaluate parameters like A and B of equations (67) and (68) from complicated combustion models for use in responsefunction calculations [8 11. thereby leading to the possibility of significant interaction between heterogeneity and acoustics. The solutions to equation (69) differ from equation (16) and show that the heterogeneity interactions can lead to linear rather than exponential growths of acoustic amplitudes with time at low amplitudes (a > 0. and the sideways sandwich model. Relatively few theories have been addressed specifically to the acoustic response of heterogeneous propellants [82]. Heterogeneity effects For heterogeneous propellants. Uncovering times often can be comparable with acoustic periods. then lowfrequency.1. The complexity of the combustion process was discussed in Section 7. the latter measuring the magnitude of a heterogeneity of the propellant and being independent of the acoustic amplitude. If the uncovering time is short compared with an acoustic period. B < 0) and to limiting amplitudes that are approached at long times (a < 0. in which instabilities are strongest in specific. B > 0). Acoustic Instabilities in SolidPropellant Rocket Motors 323 9. have been observed experimentally for composite propellants [86] and may be attributable to the heterogeneity of the solid. described in Section 7. Applications of timelag concepts to account for various aspects of heterogeneity have been made [60]. There are reviews of the early studies [7] and of more recent work [82]. When the propellant surface responds coherently. [82].9. narrow. the current situation is much less satisfactory. Preferredfrequency phenomena.5.
a difficulty in applying results of the analyses to the calculation of growth rates of acoustic fields arises from the fact that the velocity oscillations associated with oscillations of m are normal to the velocity oscillations of the driving field. and the new mode is not likely to be a normal mode of the chamber. and instabilities that involve the propagation of steepfronted waves (identified in the introduction as shock instabilities). arose early. The oscillations of rn can excite acoustic oscillations if there are pressure oscillations at the point in question to enable work to be performed on the acoustic field. Erosive burning was addressed in Section 7. [123. If the acoustic velocity in the gas adjacent to the burning surface is small compared with the local mean erosion velocity. little work has been done on the corresponding solidpropellant problem. a corresponding response to pressure oscillations also is to be anticipated. Nonlinear effects Nonlinear phenomena.8. that of combustion of a solid fuel with a gaseous oxidizer in a laminar boundary layer. usually associated with high amplitudes of the acoustic field. The process of transfer of normal oscillations to tangential oscillations must be studied to ascertain the efficiency with which the response can produce amplification. [87][93]. indicated in Section 7. An analysis has been published [94] of a related problem. In addition to the meanflow erosion.8 where it was indicated that the burning rate m depends on the gas velocity parallel to the propellant surface. Even for these relatively simple problems. Thus a mode transverse to the driving mode is generated. even at low acoustic amplitudes. However. Of course. 9.1. Here we shall discuss only three topics involving nonlinearity: the response of the combustion zone to transverse velocity oscillations (conventionally termed velocity coupling). on par with the pressurecoupled response function p.6. Although there are special conditions under which this makes sense.324 Combustion Instabilities maximum. It has long been recognized that acoustic erosion can affect the instability [7]. can introduce many interesting effects into acoustic instability [76]. Some aspects of the results qualitatively resemble those anticipated from nonlinear effects. then it may be expected to produce smallamplitude oscillations of m that locally may be related to the amplitude of the acoustic velocity oscillation through an admittancelike expression similar to equation (52). the influences of velocity coupling usually are nonlinear. changes in the mean burning rate of the propellant in the presence of an acoustic field. . The combustionzone analyses needed for calculating the response would be complicated because of the complexity of erosive burning. and the idea of introducing a velocitycoupled response function. there is acoustic erosion from most of the modes in a chamber experiencing acoustic instability. which is more difficult because of the essential role played by finiterate chemistry. and they require special consideration.
then equation (745) or equation (746). By contrast. in boundary damping or in the pressurecoupled response. then with m = E. . which is of importance in producing “secondary peaks” of the mean pressure. 6w. This same effect would be found in the presence of mean erosion if the acoustic velocity were normal to the mean erosive velocity. for example. as the intensity of the acoustic field increases. where E. For example. are found.p)]} (71) . Another aspect of nonlinearity is a change in the mean burning rate. it may be shown by evaluating jgn (I 1 + 2 u p sin 81 . Thus the response generates harmonics but does not affect oscillations at the driving frequency. this causes m‘ to be a function of lcos cot1 = {[l + co~(2wt>]/2}”~. if equation (741) is taken to apply m’) and under conditions of oscillatory pressure. the Fourier expansion of which exhibits contributions from frequencies 204 401. With u proportional to cos cot. if the erosive velocity is u = uo(l + u’) and u’ is sinusoidal in time (so that its maximum is 2.1) d0/2n that m‘ = 0 for 2 0 < 1 and ‘ E‘ = ( k M 0 / i i i 0 ) ( 2 / n ) { ~ ~1 . where v is the acoustic velocity tangent to the surface. shows that the fluctuating response m‘ is a function of I u 1. as well as harmonics. . This type of process has been described for heaterexcited acoustics oscillations in tubes [95]./7). erosive effects (in the absence of an erosion threshold or of negative erosion) produce an increase in the mean burning rate with increasing velocity amplitude. is the value of E at p’ = 0 (so that E ‘ # 0 at finite amplitude). we find through expansion to second order (by use of the average p’ = 0) that + which exhibits a decrease in the mean burning rate with increasing pressure amplitude for 0 < y1 < 1. For example. For example. then through adoption of equation (745) with M = M. The amplitudes of the harmonics in m‘ increase as I u I increases. even though the acoustic field itself remains entirely in a range of linearity. with a nonzero mean erosive velocity and a component of the acoustic velocity parallel to it. contributions at frequency co. It provides a mechanism for reaching limiting amplitudes that is an alternative to nonlinearities in the bulk of the chamber.(1 + u’).1. It is easy to write formulas for this change if the response is assumed to be entirely quasisteady. but none from frequency w.(l p = p(1 + p’). Acoustic Instabilities in SolidPropellant Rocket Motors 325 All other types of velocity response are nonlinear in one way or another. the erosive response extracts energy from the fundamental and feeds it into harmonics. It is theoretically possible that. . More generally. if there is no mean erosion and if frequencies are low enough for the propellant to respond to the acoustic erosion in an entirely quasisteady fashion.cOs1[1/(2.9. eventually establishing a steady condition in which the amplitudes of the fundamental and of all harmonics are independent of time.
As acoustic amplitudes increase. when the maximum acoustic erosive velocity equals the mean erosive velocity. a threshold for onset of an acousticerosive contribution to E occurs at the beginning of rectification. where the inverse cosine lies between 0 and 4 2 . the same results are not expected to apply at frequencies. The departures remain small if nonlinearities develop in damping processes at sufficiently low acoustic amplitudes and serve to limit the amplitudes to low levels. the expansion is carried to second order in a small parameter t that measures the shock strength. In the theoretical analysis of shock instability. but the shock remains isentropic to this order of approximation. [loo]. [102]. shock waves that are not too strong are presumed to propagate axially back and forth in a cylindrical chamber. In this case. the wave forms begin to depart from those of the normal modes. high enough that departure from quasisteady response occurs. Particular examples involve axial modes in cylindrical motors with internally burning tubular grains [99]. In addition. discontinuities are permitted across the normal shock. [102]. Theoretical methods for analyzing nonlinear growth and limiting amplitudes in this lowamplitude regime have been investigated [96][98].326 Combustion Instabilities for 2 0 > 1. At a fixed value of the dimensional amplitude. bouncing off a planar combustion zone at one end and a short choked nozzle at the other [lo ll. A few analyses of nonquasisteady changes in the mean burning rate have been completed for various models of pressurecoupled response [76] and suggest that as the frequency increases. the largest E' in equation (71) occurs at uo = 0 (as expected. in qualitative agreement with the predictions of equations (70) and (71). Since nonlinear effects are addressed. the increases in E that are of greatest practical concern for producing increases in jj in choked chambers in all probability are usually associated with effects of acoustic erosion. shocklike disturbances. there are motors that are linearly stable but nonlinearly unstable to finiteamplitude. since this corresponds to a zero threshold) and is given by For a few propellants. There are systems in which the amplitudes become large and the acoustic waves begin to exhibit a steepfronted character like shock waves. Boundary conditions at the propellant surface ( z = 0) and at the . However. A theoretical approach has been developed for analyzing these shock instabilities [lo ll. Therefore. The onedimensional. The approach is based on methods that have been applied to related problems in nonreacting flows [103][ los]. experimental evidence exists showing that a decrease in E'is produced by pressure oscillations and an increase by erosive velocity oscillations [90]. timedependent conservation equations for an inviscid ideal gas with constant heat capacities are expanded about a uniform state having constant pressure j and constant velocity V in the axial ( z ) direction. E' decreases from the value given in equation (70). 2 u 0 0 . that is.
Yo and I . eventually approaching shocks.Y y ) . For example. ' .. The results relate the pressure and velocity fields and the shock strength to the admittances. (73) here t > 0. as is usual in perturbation procedures of this type. It is known that nonlinearity causes waves to steepen with time. and a timelag approach to the description of the combustion zone has been adopted [102]. given by equation (74). but both Ycl)and Y c zoccur ) at larger [7]. may be thought of as being set by a balance between the net work input at the boundaries and a kind of dissipation in the shock. the admittances could be functions of the pressure history over a cycle at the boundary.Calculation of these admittances entails analysis of the combustion zone and of the nozzle flow subject to shocklike perturbations. For example characteristics may be introduced as independent variables and space and time may be treated as dependent variables. Solutions cyclic in time are sought. constancy of the Mach number at the nozzle entrance has been employed [loll. it is found that the pressure increase across the shock is Ap = tpM[4y2/(y + I)]. It appears that.3) and may be performed in different ways [loll. (74) Representative pressure and velocity fields are illustrated in Figure 9. in general. Acoustic Instabilities in SolidPropellant Rocket Motors 321 nozzle entrance ( z = I ) may be expressed in terms of real admittances. The solutions obtained describe the longtime behavior of slightly nonlinear waves.. Thus in the first approximation the shock strength depends on the boundary admittances [see equations (73) and (74)]. For axial modes in tubular grains. the driving mechanism must involve a continuous nononedimensional energy input from the response of the propellant combustion to the passage of the shock over its surface and probably exhibits a significant erosive component. where the sawtooth shapes of the waves may be seen. + . In rocketchamber experiments.1 = YLyp  + Yb?](P  p)..9.5.. it is necessary to carry the analysis to second order if the complete solution at first order is required. the steep fronts may develop from sinusoidal waves or may be established directly by an imposed pressure pulse. [I1021. The limiting amplitude. thereby introducing additional complications into the formulation. for small Mach numbers M of mean flow in the chamber.Y y ) . so that more mechanical energy is fed into the wave pattern at the combustion zone than is extracted at the nozzle. [102].by expansion to second order : T (0  6)0.3. Few of the needed admittance analyses have been completed. . respectively.1. and. . The analysis makes use of characteristics (Section 4. (72) The small parameter of expansion for the shock strength is € = (j/V)(. [1011. only Y c l ) appears for small values of M .
The statement of energy conservation in the solid may be taken to be equation (56) with a heatrelease term.5.4. let us first consider the planar. gasless combustion of a solid. Theoretically calculated nonsinusoidal wave shapes of pressure and velocity fields for shock instability [loll.1 and discussed in the middle of Section 7. adiabatic. INHERENT OSCILLATIONS OF BURNING SOLIDS As an especially simple example of an intrinsic instability. If t of equation (73) is not small. 9. added to the righthand side. Accurate use of numerical methods for acoustic instability is difficult to achieve because so many physical processes with poorly known parameters usually are involved. a temperatureexplicit . methods of numerical integration are needed in describing largeamplitude behaviors accurately. There are intrinsic instabilities in burning solids that involve fewer processes and that are therefore better suited to investigation by numerical approaches. then cyclic solutions cannot be obtained by the perturbation approach that has been outlined here.2. mentioned at the beginning of Section 7. also can be applied to these intrinsic instabilities. as will be indicated in the following section. In general. I 0 2 1 FIGURE 9. somewhat analogous to the method just described. say w q .328 Combustion Instabilities P P  V V D t t f  Wave d i r e c p Wave direction t fixed I p fixed ! ~ f . Although w q properly depends on the reactant concentration. Nonlinear perturbation methods. 0 Z 1  I 0 c Wave direction f  1 Wave direction Z t I fixed .
imposition of T consistent with the differential equation.=oI2. The burning rate m is expressed in terms of w1 by equation (720).=o+I2 = (2qs/&) J.* As can be inferred from the timedependent generalization of equation (7. then this can be shown to restrict the formulation to conditions such that changes with time are not too rapid. Here x = 0 has been selected as the position at which the reactant has been consumed completely. whichthrough matchingcorrespond to conditions just on the negative side of x = 0 for the outer solution in the unstretched x variable. * On a thermalexplosion time scale.=~ is + of higher order in the nondimensional activation energy (the Zel’dovich number). (76) in which the asymptotic expansion of the integral may be evaluated relatively conveniently for temperatureexplicit rates.a2T/dx2= .identifies conditions “at minus infinity”in the stretched variable.9. t ) and m(t). Use of these equations along with T = To = constant at x = m.qswl to the first approximation in the reaction zone. It is consistent within the formal context of the asymptotics to restrict attention to problems in which (aT/a~).4. Then. equations (75) and (76) give.15). If there is a narrow reaction zone in the vicinity of x = 0. and wy = 0 for x > 0. 0 wl(aT/ax) dx. = constant at x = c c is in . the integral of this equation across the reaction zone is seen to be expressible as where the subscript x = 0.1. the formulation in employing equation (720) excludes a transient behavior within the reaction zone. Multiplying the differential equation by dT/ax and integrating over the same range yields [(aT/~x). the augmented version of equation (56) becomes . so long as attention is restricted to times small compared with the time for homogeneous thermal explosion of the reactant at temperature To. dT/dx = 0 at x = +a and a suitable initial profile for T(x) provides a wellposed problem for calculating T(x. It is attractive to seek solutions by use of activationenergy asymptotics. then by stretching the coordinate about x = 0 and excluding variations on a very short time scale. By use of equation (720). with use of equation (718). Inherent Oscillations of Burning Solids 329 formulation may be achieved by adopting an expression like equation (718). with wq = qswl for x < 0 in the notation of Section 7. in the first approximation.2.[(aT/ax). if timedependent versions of equations (715) and (716) describe the original problem.
+ = 0 provide first approximations to the results obtained by a more formal derivation of jump conditions across the thin reaction zone. +~ [ ( A . Intrinsic instability to planar disturbances occurs if solutions having a > 0 exist. it may then readily be seen that timedependent solutions in the absence of external pressure perturbations may occur if s2 .( A + B)l(a + io)zs.=."' ~. (79) (78) may be written as ( A .we may put n = 0 in equation (59) and B = c.q c. = (AT)i from equation (64). which has been assumed to be large in the course of the activationenergy asymptotics. in equation (68). (78) where now Equation 2(a + iw)z3. B.) = constant.Y1. Specifically. B ~ T ~ ' f 2 ~ . Equation (77) and (dT/dx). From the derivation leading to equation (66).    .330 Combustion Instabilities the last equality of which may also be inferred from equation (729). Here ?I: denotes the value of T at x = 0.(1 + +A  B)s + A = 0./c. we must then put cp = c. the formulation of Section 9. with the replacements E . .(T. A relatively simple technique is to look for solutions of the form T = T + Re{ATe(afi")t}.B)2 . from which it is seen that for o # 0.B ) . T remains constant for x > 0. to achieve surface adiabaticity within the context of a linear expansion about equation (60).B)2 (A + B) . obtaining ( A T ) . Thus for describing small perturbations about the steady state in gasless combustion.. = $ [ ( A ..5. / ( A B)' + 1 2(A + B)]. o = 0 and a ~=. = E1/(2R0). The resulting form of equation (58) corresponds to a surface that is adiabatic on the side x = O + and at which a constant energy per unit mass is released. to enforce a constant surface heat release per unit mass. otherwise. A stability analysis of the steady state of equations (56)(59) may be developed in a straightforward manner [64]. = { 2 R 0 ~ . = A + B + s = [I + (az . Studies of the stability of the steady state therefore may be pursued as a special case of the stability analysis of the problem defined by equations (56)(59).] (}' l' and L .7.o2+ 2iao)z.l/2. we have az. where a is the growth rate and o the frequency. in equation (65).( A . . Equation (56) applies on each side of the reaction zone in this threezone problem.B)' . The problem for x < 0 becomes identical to that described by equations (56)(58).) = qs(l .( A + B)] and co = This solution applies when ( A . the square of which yields AB + J1 + 4(a + iw)z. To the lowest order in the Zel'dovich number. and the sole remaining parameter is A [defined in equation (67)].B)' + 1 I 2(A + B).1.4 may be employed with n = 0 and B = 1. / [ E ~Y1. Jv..B ) J1 + 4 ( a + io)z.2 = [ ( A .
A stability diagram in the parameter space is shown in Figure 9. In applying these results to the model of Section 9.6 remains exactly correct. but. and the resulting stability boundary will not be expected to differ greatly from that shown in Figure 9.6. that is.6 even for relatively small values of A / B . we let B = 1 and find that intrinsic instability OO L . a generalization of equation (59) may be developed for finite values of A / B (still within the context of a quasisteady gas phase). I . . To apply the results to the gasless combustion problem introduced at the beginning of this section. Thus Figure 9. I I I I I I I 2 4 6 8 1 0 AIB FIGURE 9. it may be noted from equations (67) and (68) that A and A / B represent nondimensional activation energies for the interface and gasphase processes.1. Since equation (59) relies on an asymptotic expansion for large values of A / B .6 therefore shows that intrinsic instability occurs (even in the absence of pressure perturbations) when the nondimensional activation energies are sufficiently large. for example. it may be shown that Figure 9.9 2 Inherent Oscillations of Burning Solids 331 which is positive (implying instability) throughout its region of validity. we may conclude that in a strict interpretation of the range of validity of the analysis. If A < 1 or if A / B < 1.4.6 is expected to provide at least a qualitatively reasonable approximation to the stability boundary for all values of A and A/B. respectively. Figure 9. with suitable redefinitions of A and B.6.5.B)’ < A + B. equal to zero . instability is to be anticipated whenever A > 1. Diagram of regions of intrinsic instability of the steadystate combustion given by equation (60). A stability boundary in the AB plane may be defined by setting the first expression for M. In fact. if A > 1. the coldboundary difficulty is not present in this model. then according to the criterion obtained. the combustion is intrinsically unstable unless ( A . then there is a critical value of A / B above which instability occurs. the combustion is intrinsically stable.
)/[p. it is halted. the thickness of the heated layer of reactant is on the order of x 1 = a./(2R07..Yl.332 Combustion Instabilities is predicted to occur if A > 3.(1 . therefore. ~ ) ’ ] ) ~ x’ constant. From the second equality in equation (77). As formulated. The mechanism of the instability that has been described here suggests that intrinsic instability of gasless combustion evolves to an inherently pulsating mode of propagation rather than to a sustained explosion or . apparently we should conclude that the steadystate solution to the adiabatic version of the problem of Section 7. = {2RoBlT~1+2A. as a consequence of the fact that v1/v2is small for large Zel’dovich numbers. ~  and where 7.To)dx just ahead of the reaction sheet.. Approximate conditions for the existence of this hot flame are reasonably well satisfied because its thickness. = x:/cis. = To q. = To + 2qdl  yl. x2 = a...(~Y . After the heating is completed./v2. The mechanism of the instability encountered here can be understood physically. and thereafter practically no propagation of the reaction front can occur until a new layer having thickness of order x1 is heated.o)/cps./v. Since 3 may not be a sufficiently large number. then its velocity will be roughly v2 = C exp[ . is much less than the thickness x1 of the heated layer in which it propagates.. an adiabatic flame may again be established.El/(2RUT. pscps(T . then for the steadystate solution. this occurs over a distance on the order of x1 and. If a.)].the adiabatic flame temperature. lasts for a time of about e2 = x1/v2. where T2 + = TI + 4 S U  YI. flame propagation in the solid is associated with an excess enthalpy per unit area given by a. The possibility of the occurrence of instabilities such as this in burning solids was known to Zel’dovich [61]. this gasless combustion problem possesses the coldboundary difficulty. and the hot flame again may be initiated at its reaction front. . where v1 is the steadystate flame speed.4 is intrinsically unstable.When the hot flame encounters cold reactant. If a flame now begins to propagate adiabatically in the heated reactant.)].. [l lo].O)/C. and the results are therefore restricted to large values of A . [l09]. EE As/(pscps) denotes the thermal diffusivity of the solid. we see that s’? v1 where C = C exp[ E. The hot flame may propagate until it consumes all the heated reactant. and stability analyses for various models are available [76]. The transient heatconduction time required for this heating is roughly t .Elq. This excess provides a local reservoir of heated reactant in which a flame may propagate at an increased velocity.o)/~ps. Under the present conditions of negligible diffusion. a value considerably greater than the adiabatic flame temperature of the unheated reactant.
If the reaction is of the Arrhenius form for temperatures above the switchon temperature. onedimensional combustion is stable for values of the parameter below the critical value. then A is a suitable parameter for measuring the temperature sensitivity of the rate. and pulsating propagation has been observed experimentally for solids [ 112][114]. while the deflagration is inherently oscillatory at larger values. and a wellposed deflagration problem can then be defined without introducing asymptotic concepts. it is to be expected that there will be a critical value of a parameter that measures the temperature sensitivity of the heatrelease rate. The average velocity of propagation is XI/@. there might be a critical temperature exceeding To. which is less than the velocity u1 of the steady flame. a steady solution and an oscillatory solution) and at which the character of the solution to the problem may change. These results are in qualitative agreement with predictions of accurate theories. a value at which a splitting into more than one possible solution occurs (for example. It can be imagined that for sufficiently large Zel'dovich numbers. These results verify that during the rapid portions of pulsating propagation. a tendency for this to occur has been found in numerical integrations [lo91 which show. perhaps corresponding to a phase transition. The critical value would represent a bifurcation point.2. Inherent Oscillations of Burning Solids 333 extinction. flame temperatures in excess of the steadystate adiabatic flame temperature occur. The coldboundary difficulty then disappears. This conclusion has been substantiated theoretically by more detailed analyses [1091. The ratio of the duration of a pulse to the time between pulses is roughly which is small when the Zel'dovich number is large. Estimates of characteristics of pulsating propagation may be obtained from the model that has been described for the mechanism. Real condensedphase processes might not be strictly Arrhenius. We have seen that there are interpretive difficulties associated with the use of activationenergy asymptotics for calculating limits of intrinsic instability in gasless combustion. morerapid flames might develop and follow the first hot flame. for example. at the larger values of the activation energy. such that the steadystate. For a problem of this kind. additional. Although stability analyses based on activationenergy asymptotics may not provide accurate values of A at the bifurcation point. Numerical integrations [lo91 seem to indicate that for . with each series separated by quiet periods. below which the reaction effectively ceases.9. + t2) = 4 1 + (t2/tl)I. an irregular series of pulses. they may nevertheless yield stability limits that are roughly correct. [l 111.
Heterogeneous solid propellants possess additional mechanisms with potentials for producing oscillatory burning. approximately 4./t. particularly in the limit in which the pulse duration is short compared with the time between pulses. It was indicated in Section 9. relatively small rates of heat feedback to the surface through processes that are not too strongly dependent on temperature may produce stability of the steady deflagration. the resulting bifurcation value of A . certain metalized composite propellants have been observed to burn in the laboratory with idehtifiable ranges of frequencies of oscillation [I IS].5 that heterogeneities introduce at least local periodicities and that. then a nonlinear perturbation analysis can be developed for describing analytically the nature of the pulsating mode [Ill]. it has often been called chgfing. the difference between A and its bifurcation value is treated as a small parameter. In effect. A review is +fi . Although the theoretical explanations that have been proposed for chuffing [I 151[ 1 171 appeal to concepts of homogeneous thermal explosions and usually introduce more than one chemical step. The results exhibit the same general behavior that was found from the numerical integrations [1091 for conditions near bifurcation. Even if the solid propellant burns by a mechanism predominantly involving condensedphase reactions of large nondimensional activation energy.+ = 0. The departure of the oscillation frequency from its value at bifurcation is expressed in the same type of series.and oscillatory solutions for temperature profiles are calculated as perturbations about the steady solution in the form of a power series in . This phenomenon has been observed at low pressures in strandburner and rocketmotor tests of certain propellants. For solid propellants. and under these conditions pulsating deflagration of a homogeneous solid propellant is anticipated. so that inert heat conduction occurs for x > 0 as well. say t. give A > 2 as the criterion for instability [ill].334 Combustion Instabilities gasless combustion.5. If attention is restricted to the vicinity of the bifurcation point. the mechanism may involve chemical interactions between the metal and the oxidizer. the gasphase flame may be absent. a mechanism like that just described may be relevant. is closer to the value inferred from the numerical integrations than is that obtained from the more formally correct asymptotic analysis. obtained above from the asymptotic analysis. For example.1..24. the pulsating instability may be eliminated through stabilizing influences provided by energy input from the gasphase flame. A 2 6 typically may be a better criterion for intrinsic planar instability than the criterion A > 3. sustained coherent oscillations of the combustion will occur. The methods of analysis possess a qualitative similarity to those of the shockinstability analysis discussed at the end of the previous section. in the presence of a mechanism for synchronizing the phases of the oscillations over the surface of the propellant.=. Stability analyses in which equations (75) and (76) are employed with a &function heatrelease rate but without (i?T/dx). in this case. At sufficiently low pressures.
Higuera and A. The results show that the minimum value of A along this limit curve occurs at k.6 is changed to I/$. +fi * F. with an adiabaticity condition applied at the wall. A. R. 95 of Progress in Astronautics und Aeronuutics. = q m n u o ~ s where /a. K. An algebraic expression of the type of equation (78) is obtained. A model for spinning propagation has been proposed [120]. have shown that a similar phenomenon occurs for the model that led to equation (78). the permissible number of spin cells increases as the diameter increases. the stability boundary in Figure 9. This that result suggests that there may be materials for which 4 < A < 2 are incapable of experiencing pulsating combustion under any conditions but that can exhibit spinning combustion if the diameter of the cylinder is sufficiently large. Spinning propagation of deflagrations in cylindrical samples of certain homogeneous solid combustibles has been observed experimentally [114]. timedependent phenomena. The stability limit ( a = 0) may then be obtained in the form of a curve of k.84. vol. although the physical mechanisms involved of course are very different. “Stability of SolidPropellant Combustion to NonPlanar Perturbations” in Dvnarnlcs of Flumes and Reactioe Systems.5. as a function of A . the smallest cylinder radius for the occurrence of this spinning propagation is roughly a z 3. which approaches the boundary shown in Figure 9. Axial propagation in a cylinder of radius a is considered. At large values of the Lewis number the character of the diffusivethermal instability is that which has been described here. New York: American Institute of Aeronautics and Astronautics. which will be introduced in Section 9. 1984.. In general. J . uo is the flame speed of the unperturbed wave and qmn is the mth root of the equation d J .7 uoz. The inherent oscillations that have been addressed here in fact correspond to a limiting case of the diffusivethermal intrinsic instabilities of flames. k. the equation analogous to equation (78) becomes a dispersion relation. an additional nondimensional parameter appearsnamely. When trans= + verse disturbances are admitted. N. Soloukhin.6 for large values of A but gives A = 1/2( < 1) at large values of A / B .1). = 1/2 and has A = 4. if A z 4. and a linear stability analysis has been worked out [121].* Since the smallest nonzero value of qmn is q 1 z 1. fi .. Oppenheim and R. Lifian. Manson. r = a.2. represents a nondimensional wave number of the disturbance in the transverse direction. J..3. giving the complex frequency as a function of the wave number. It will be found in Section 9.5 that additional types of instability behavior may occur if diffusion of chemical species is not negligible. The ideas of oscillatory burning that have been considered thus far in the present section have been restricted to onedimensional. admitting nonplanar perturbations in the model of [l 111described previously. I. Bowen. eds. Inherent Oscillations of Burning Solids 335 available that includes discussion of influences of preferredfrequency combustion in heterogeneous propellants [1191. ( q ) / d q = 0 for the Bessel function of order n. Since k. There is a geometrical analogy here to spinning detonations (Section 6..9. which is below the value 2 + J5 that was obtained for the planar instability.. 248258.
Equation (5 1) is useful for calculating contributions of homogeneous processes to amplification. Here we shall merely give a brief discussion first of screaming.3. and responses of these diffusion flames to acoustic oscillations may contribute to amplification.ough viscoelastic damping and solidparticle damping usually are absent for liquid propellants. at least to some extent. A thorough presentation of the subject of combustion instability in liquidpropellant rockets has been published in a volume of more than 600 pages [122].336 Combustion Instabilities 9. most of the materials in Section 9. and the latter involve perforated sheets on the chamber walls. then of chugging. between fuel and oxidizer streams. Analysis of acoustic amplification by combustion processes in liquidpropellant rockets is more diffcult than that for solidpropellant motors .1.1. for example. Acoustic instabilities in liquidpropellant rocket motors share many features in common with those of solidpropellant motors. The former type has been called screaming and the latter chugging. although various other descriptive terms also have been applied. The former are solid surfaces extending into the flow to interfere with establishment of lowfrequency acoustic modes and to enhance attenuation. alth. by locally enhancing rates of heat transfer. Conditions might occur under which mixing is rapid and combustion slower so that homogeneous gasphase chemical rate processes provide some amplification. Diffusion flames also may be established elsewhere than around droplets. Because of their fewer natural damping mechanisms. liquidpropellant motors more often need instability “fixes” like the baffles and liners mentioned at the end of Section 9. designed according to the principles of the Helmholtz resonator (described in the following section) to increase acoustic damping by dissipation in the oscillatory flow through the orifices.1 is relevantespecially the discussions of oscillation modes and of damping mechanisms. and finally of instabilities at intermediate frequencies.4. OSCILLATORY BURNING IN LIQUIDPROPELLANT ROCKET MOTORS Liquidpropellant rocket motors are no less susceptible to oscillatory burning than are solidpropellant motors. The two major types of combustion instabilities for liquidpropellant motors are acoustic instabilities (with frequencies greater than roughly lo3 cycles per second) and instabilities of lower frequencies that typically involve interactions between chamber processes and flow in the propellant feed lines. Therefore. it can rapidly burn holes in propellant injectors or chamber walls. The main difference between the two types of motors lies in the character of the amplification mechanisms. Combustion is distributed more or less homogeneously throughout the chamber of a liquidpropellant motor and generally involves droplet combustion. Rarely has a development program for a liquidpropellant rocket been completed without encountering a problem of combustion instability. Screaming often is the more severe of the two in that.
Consequently. [122].9.2. Equilibrium and frozen sound speeds with respect to droplet behavior may be identified (compare Sections 9.1. However. These frequencies are generally sufficiently low compared with the natural acoustic vibrational frequencies of the chamber that the pressure can be considered to be spatially uniform throughout the chamber. for example. general inferences about aspects of scaling [122]. Acoustic theories may be derived from the conservation equations of Chapter I1 for the purpose of analyzing acoustic amplification in spray combustion. the oscillations depend in a lessdetailed manner on the nature of the processes occurring in the chamber. in the injection system. because of the complexity of the problem. [127]. Timelag hypotheses have become relatively sophisticated for acoustic instabilities in liquidpropellant rockets [20]. Numerical integrations of nonlinear differential equations. as a model for combustion in the wake region of a vaporizing droplet. Crocco [20] * Amplification expressions in general are not as simple as equation (51) for twophase flows in rocket motors [128]. From the viewpoint of equation (14). Numerical estimates suggest that the last of these three problems. Chugging is a system instability having frequencies less than roughly lo2 cycles per second and involving oscillations in the propellant feed lines and injection systems as well as in the combustion chamber. and the laminar jet diffusion flame [124]. and ordinary linear differential equations with time lags are well suited for describing the instability. the quantity @ must include amplification effects that arise from both energy sources and mass sources. the application to instability calculations has not been carried beyond the point of drawing rough.3 have been applied more often for liquidpropellant motors. Oscillatory Burning in LiquidPropellant Rocket Motors 331 because the direct use of admittances through approximations of planar burning surfaces seldom can be justified. possesses the greatest potential relevance to acoustic instability in liquidpropellant rockets [1293. interpreted.4.3. [123].6 and 4. Therefore. and in various processes occurring in the combustion chamber. These equations may involve a number of time lags. which correspond to delays in the feed system. [125]. also have been applied [122]. [124][127].5. with chemical kinetics and diffusion processes treated in overall quasisteady approximations. and a masssource term may arise in equation (51) in the development of a conservation equation for the acoustic energy in the gas. timelag approximations of the type discussed in Section 9. .1. the laminar boundary layer on a flat plate of vaporizing fuel 1941.6).* The three specific diffusionflame problems analyzed were the stagnationpoint boundary layer on a vaporizing fuel [126]. A few analyses have been pursued concerning timedependent vaporization and diffusionflame combustion in which linearizations about known quasisteady solutions were developed with the objective of ultimately being able to calculate amplitude and phase relationships for use in an amplification expression like equation (51) [94]. Further research on problems like these could improve our fundamental understanding of amplification mechanisms for combustion instabilities in liquidpropellant motors. [ 1221.
so that a change in the pressure at the injector produces a change in the injected fueloxidizer mixture ratio. Theoretical analyses of lowfrequency oscillations involve determining the response of one section of the rocket system to vibrations in another section. result in variations in the temperature of the burnt gas which. For example. A. [129]. [122]. [128]. in turn. The term entropy waue. A similar mechanism can readily be envisioned for an endburning solidpropellant motor. without variations in the mixture ratio. Thus vorticity may play a role in intermediatefrequency oscillations under appropriate conditions. and the term is therefore insufficiently descriptive. The mixtureratio variations are convected through the chamber and. Tsien [ 1301 has given clear illustrations of these analyses. and since half of the process is acoustic and the other half convective (which is slower).* A more detailed discussion of a phenomenon of this type was given in Section 6. E. J. if the combustion is quasisteady and the consequent nonisentropic variations in burntgas temperature with pressure. then a sustained oscillation can develop. the injection rates of fuel and oxidizer may differ in their sensitivities to chamber pressure. fails to distinguish mechanisms involving mixtureratio variations from those that do not. higher than chugging but lower than screaming [122]. in propellant motors. 83 (1984). producing pressure perturbations that propagate back upstream acoustically to influence the pressure at the injector face. A closely related mechanism also can be involved in rumble instabilities in ramjet combustors. delineating regions of selfexcited instabilities in the entire rocket system. .2. do not decay before reaching the nozzle. from which it may be seen that onedimensional instabilities of the planar detonation structure involve qualitatively the same type of process. and T. A number of reviews are available [20].1. may modify the nozzle flow.1. whose generation may be enhanced by pressure waves in suitable geometrical configurations.5. upon combustion. Y. It is easy to imagine that a sustained oscillation in a longitudinal mode may be established in this manner. discussed in Section 9.3. Toong. [1311[ 1341. Convectiveacoustic instabilities of this general kind also could involve the convection of vortices with axes transverse to the flow direction or of swirls having longitudinal axes. which often has been applied to instabilities of this type in motors. the frequency will be less than an acoustic frequency of the chamber. There are mechanisms for oscillations at intermediate frequencies. and developing feedback servomechanism techniques for stabilizing the system. Abouseif.338 Combustion Instabilities has made a considerable amount of theoretical and experimental progress in deducing properties of the time lags of the conversion processes in the combustion chamber. * G. C S T 3 6 . Keklak.
and the ratio cu. with frequency WH =a / J W . in which spatial variations of the pressure in the combustion chamber are negligible. SYSTEM INSTABILITIES IN COMBUSTION EQUIPMENT Chugging in liquidpropellant rockets is only one of many examples of system instabilities in combustion devices [135]. The simplest example of an oscillation in a bulk mode is the classical Helmholtz resonator. dthat is open to the atmosphere. [136].4. Since frictional damping always is present but has been neglected in the derivation of equation (80).4. differentiating. As may be seen from the contribution of Putnam to [137]./(a/l) is the square root of the ratio of the volume of the tube to that of the chamber. a source of excitation. and then substituting the result into the first expression. Here we shall not address problems involving acoustic wave propagation. Applying equation ( 3 ) to the gas in the chamber. we shall restrict our attention to bulk modes. since otherwise an acoustic field with a nonuniform p would have to be considered.9.p). and consider slowly varying processes for which the pressure p at the inner end of the tube is equal to the spatially uniform pressure within the chamber. then a representative acoustic frequency is all. Consider a gasfilled chamber of volume into which is inserted a tube of length 1 and of crosssectional area . a force balance on the gas in the tube is d(p . a related restriction limits the length 1 of the exit tube. By substituting this relation into the one preceding it. we have d p / d t = (dp/dt)/a2. such as a suitably . If 1 is comparable with a characteristic length of the chamber. With friction neglected. Assume that the atmospheric pressure remains constant at the outer end of the tube at a value p. System Instabilities in Combustion Equipment 339 9. Conservation of mass for the gas of density p in the chamber is V d p / d t = p d v . we find that the pressure history is described by ( l V / d ) d 2 p / d t 2 = a2(p . (80) Equation (80) possesses solutions that are oscillatory in time. and it may or may not be necessary to consider acoustic wave propagation in one or more of these components in theoretical analyses [138][142]. Interactions of processes occurring in intakes and exhausts with those occurring in the combustion chamber typically are involved. which is small for small tubes in large chambersthe configuration envisaged in the Helmholtz model. (8 1) The derivation of equation (81) is invalid unless the resulting Helmholtz frequency wHis small compared with the lowest natural frequency of acoustic waves in the chamber. a considerable amount of research has been performed on mechanisms of these instabilities.j) = ( p d ) du/dt. where j is the average gas density in the tube and u is the average outward velocity of gas in the tube.
defined as the ratio of the mass flow rate through the nozzle to the product of the throat area with the chamber pressure. The response of the burning rate to pressure variations is needed for use in equation (83). we find that co. the stability limitis defined by putting a = 0 in equation (85).p). then either by use of Laplace transforms or (more simply) by seeking soluit is readily found from equation (83) that tions proportional to e(a+iW)'.E) . instability [7]. is a suitable nondimensional frequency. = nBs/[s2 . or L*. it is reasonable to neglect spatial variations of the pressure within the chamber. by subtracting the steadystate mass balance from equation (82). and C Dis a discharge coefficient for the nozzle.340 Combustion Instabilities phased rate of heat release by combustion in the chamber [see equation (51)]. (82) where m is mass burning rate per unit area for the propellant. Instability occurs when a > 0 in the solution to equation (85). must be available if sustained oscillations are to occur in the Helmholtz mode.(1 + A . An example of bulk modes in solidpropellant rockets is afforded by the lowfrequency. (85) where A and B are given by equations (67) and (68) and s by equation (79) [compare equation (66)]. = (83) where a characteristic frequency for the instability is WL a2CD/L* (84) Since equation (84) shows that oL decreases as L* increases. Equation (81). and by employing equation (3). also often provides a reasonable approximation for frequencies of bulk modes in combustion devices even when the mean velocity of flow out of the chamber exceeds the oscillation velocity. A mass balance for the gas inside a rocket chamber with a choked nozzle is Km = pCD + d(pL*)/dt. If the mechanism of Section 9.5. and its ratio to a characteristic exhaust velocity provides an estimate of the residence time of a fluid element in the gas phase inside the chamber. In view of equation (79). it is seen that 07.B)s + A].(p . and therefore the neutral stability condition provides a relationship among the dimensionless . dp/dt = (jj//m)(m. growth of oscillations would need an excitation source even if there were no damping. which is directly useful in design of the acoustic liners defined in the previous section. 1 + (a + iW)/W. Since in rocket designs the fractional rate of increase of L* is small compared with the rate of development of the instability. this ratio often is denoted by L*.1. which are produced by propagation of acoustic waves. A characteristic length of importance in rocket design is the ratio of the gas volume in the chamber to the throat area of the nozzle. The condition for neutral stabilitythat is. K is the ratio of the burning area to the throat area.4 is considered. if L* is sufficiently large.
. [143][147]. The mechanism involved in this chuffing process need not necessarily be the same as that of the corresponding intrinsic instability since an essential role may be played by the fact that the ratio of the chamber residence time to the solid heatconduction time is small. Therefore.p in equation (83)which corresponds to a chamber with a low capacity. a propellant that does not exhibit the nonsinusoidal oscillations while burning in the open may experience them in a suitable chamber. we may therefore write (aZC.lL*)p. HYDRODYNAMIC AND DIFFUSIVE INSTABILITIES IN PREMIXED FLAMES 9. for given values of n.2. then it is observed that the nearly sinusoidal lowfrequency instability progressively evolves into a nonlinear phenomenon having the character of chuffing. A . There have been many studies of L* instability that support various inferences of the analysis outlined above [I 171.5.m2) = f(n7 A. 9. The .for the occurrence of instability. and the instability appears when this product is less than its critical vaIue.1. Here we shall only be able to touch on some of the highlights. Thus it may be expected that for a given propellant (that is. and B). It is found that L* instability occurs when oLz. Since iii is proportional to according to equation (59). where the function f is to be obtained is above from equation (85). In view of equation (55). this critical valuethat is.. when 0 .l. there is a critical value of wLz. Hydrodynamic and Diffusive Instabilities in Premixed Flames 341 parameters n. B) (86) as an equation for the stability limit. Theoretical investigations also have addressed questions of influences of spatial gasphase temperature variations [1481 and of solidphase heterogeneity [1491 on the instability. dp/dt becomes sufficiently small compared with p . A . If L*jj2"is progressively decreased below the critical value for instability. discussed in Section 9. heatconduction time in the burning solid propellant. we see from equation (86) that for a given propellant L*p2" assumes a critical value at the stability limit. In other words. a relatively short gasphase residence time.5. the term chuflng as used in practice may apply to different fundamental phenomena. The two characteristic times involved in the mechanism are the gasphase residence time in the chamber and the condensedphase. This prediction is in accord with experimental observations on L* instability [1431.5. Formulation through asymptotic methods We now restrict our attention to premixed gaseous combustibles and address the question of the intrinsic stability of their deflagrative processes./(c. The topic is a large one that has undergone extensive advancement in recent years. B..9. and wLz.
The constant a. analyses of pulsations of these " burnerstabilized" flames have been performed by both numerical and asymptotic methods [1631[ 1691. B. For more recent. is er/(bl).2 suggest that even for gaseous reactants. I. of order unity. Barenblatt and Y . Although an oscillatory mode might turn out to be more susceptible to extinction by heat loss. a direct correspondence between stability and flammability limits is not to be expected. M. Zel'dovich and G. Here we shall be concerned with the stability of planar deflagrations to nononedimensional perturbations. [l50][153]. continuing work on this problem see. Upstream heat loss to the flame holder of a flatflame burner can broaden the range of conditions over which the oscillations occur . C. 161 (1984).1. A. transport and chemicalkinetic properties of the system [155][157]. where x(t . oscillatory propagation might occur if the thermal diffusivity is sufficiently greater than the diffusivity of the limiting reactant (the stoichiometrically deficient one whose concentration vanishes at equilibrium). In particular. [159]. C & F 3 .Some of the results indicated unconditional stability in adiabatic propagation for the classes of flames investigated [154]. a sawtoothtype history of the reactionzone location is a typical result 111641. F. + . and numerical integrations with real kinetics have encountered the same phenomenon under certain conditions [ 1621. The delay is the time for a temperature perturbation to travel from the reaction sheet to the flame holder and back. B. Prikl. Most of the early analyses of the stability of planar deflagrations treated only onedimensional perturbations [1541[ 1591. Mat. where 5 represents the nondimensional distance of equation (518) between the flame holder and the steadystate reactionsheet position and b is the Zel'dovich number appearing in equation (566). Mekh. The equation is of a type that has been encountered in other contexts and may be written in nondimensional variables as adx/dt = e?(''' . 4 t h Symp. 21. fundamentally. Barenblatt. for example.342 Combustion Instabilities reader must consult reviews for more thorough presentations [137]. a nonlinear ordinary differential equation with a time delay+ may be derived by asymptotic methods for describing fully nonlinear pulsations [1681. We shall see that a richer variety * Additional studies with findings of unconditional stability are those of J. 61 (1959). [161]. Asymptotic analyses have demonstrated this to be true [160]. who addressed nonplanar disturbances as well. and of G. the limits of which depend on the thermodynamic. limits of flammability under perfectly adiabatic conditions are not anticipated to exist . while others identified regions of instability. Richardson. [lSS]. 856 (1957) and Y .1.2 suggests that an instability is likely to evolve to an oscillatory propagation rather than to extinction. McIntosh and J. The results of Section 9. Zel'dovich. I. CST 38. and the analysis of Section 9. The steadystate solution x = 0 is unstable if a < 2 h .1) signifies the function x ( t ) evaluated at b . Attempts to correlate flammability limits with stability limits have been unsuccessful. (1953). 182189.* At the time that these investigations were performed. [168]. an objective was to equate stability limits with flammability limits of the combustible mixture. Clarke. the oscillations tend to be sinusoidal if a is near 2in and sawtooth at small values of a.
cp.c13 and the burnt gas to x = + m./Am and Pr. and z = t v 2 / a m . with the unburnt gas extending to x = . equation (1l). Prandtl numbers based on the first and a modified second coefficient of viscosity. To allow for defined as Pr. the sheet may be treated as a surfacefor example. for defining nondimensional space and time variables as 5 = xu. However. and V F has no longitudinal component since by definition F is not a function of x./CI. = p/p. let the x coordinate be normal to the planar flame. tworeactant process of large nondimensional activation energy. At large values of the Zel’dovich number.. = ( i p . Attention will be restricted to a onestep./ci. is a twodimensional gradient having dimensionless components (d/dq. with uo given by equation (578). where f ( y . measured in the burnt gas. z ) may be written locally as x = F(y./A. vI is a transverse velocity vector composed of the velocity components in the y and z directions made nondimensional through division by v. An appropriate nondimensional  puu. C ./ct./(p./a. . (87) where R = pip. ) may be introduced. y~ = yv. and the thermal diffusivity evaluated at the adiabatic flame temperature. the nondimensional functions p ... a/a[). variable viscosities. = p. [ = ZU. z. Since research on this topic has not yet progressed to a point at which clear general conclusions can be drawn. The coordinate transformation also enables a nondimensional version of the equation for momentum conservation to be written from equations (12) and (15). through a transformation of variables. When analyses are pursued in outerscale variables. G(x. Employ the steady. Under the assumption that the reaction sheet is smooth and nowhere parallel to the x axis.v V F ) . t ) = Owhich in terms of the Cartesian coordinates (x. the chemical reaction is confined to a thin sheet in the flow. v. Hydrodynamic and Diffusive Instabilities in Premixed Flames 343 of behaviors is encountered. = p(vl aF/at . .. may be rc. (RVJ = 0. t ) if the x coordinate is not parallel to the sheet in its local orientation.. can be written as aR/az a ( ~ u ) / a t v. I = L.5. V . one review article is available that summarizes current knowledge on the g&e& fH3. respectively. reaction intermediaries may diffuse laterally from one part of the flame to another and thereby affect the response of the flame to the disturbance. t ) = F ( y . E pouo/p.  + + .)cP. For example.cp. J. it may readily be imagined that in the presence of a nonplanar disturbance. and p 2 = (fp + K ) / ( $ ~ K . z.9. t)v.f. adiabatic.. For the undisturbed flow. z. For all purposes except the analysis of the sheet structure. y . For this purpose. it is convenient to work in a coordinate system that moves with the sheet. Departures from onestep kinetics can produce significant influences on flame stability even if the normal flame speed can be described well by a onestep approximation. we shall not discuss influences of reaction mechanisms on stability. and the longitudinal mass flux relative to the reaction sheet is + +  (88) in which v1 is the x component of the velocity vector v. laminar flame speed.f.
we shall neglect the Dufour effect (and later the Soret effect). we shall adopt these assumptions. GI.ul = 1. and GI are quantities that are proportional to derivatives of pl. the moregeneral forms exhibited here are of interest for discussing influences of variable transport properties and for pursuing further investigations. The longitudinal and transverse components of momentum conservation then become. and it is often assumed that K = 0. in which h = hi y i .= 0 and GL = 0. so that Pr. Eventually. Most of the analyses that have been performed treat p and K as constant. both of which are accurate approximations for the deflagration problems considered. .. In addition. adiabatic deflagration. we obtain x : = p ah aP + pv*Vh = + V*(AVT) at at N  V. where pmis the downstream pressure of the steady. so that . and where the Laplacian operator in the transformed coordinate system is and GI.pm)/(pm u i ) . respectively. and a nondimensional body force per unit mass may be expressed as F = ~fiam/v the ~ longitudinal .344 Combustion Instabilities measure of the pressure is P = (p . (x) and transverse components of which will be written as Flland Fl. = SPr. The equation for conservation of energy will be written under the assumptions that the Mach number is small and that the work done by body forces is negligible. so that by use of equation (16) in equation (372). p2 = 1.
in the transformed coordinates. which appear through the H i with hi given by equation (1ll). .1. including mixtures containing hydrogen (for which the effect might be anticipated to be greatest). m). where Di is the binary diffusion coefficient for the species pair i and N [compare equation 3 = T/Tm.5. In addition. quantitative modification to stability boundaries [153]. in which it will be convenient to let the reference temperature T o be T. we can take advantage of the appreciable simplification afforded by neglect of crosstransport effects. Therefore. Analysis has shown that there are conditions under which the Soret effect may even change the character of the instability [170]. . equation (92) then becomes 9 N. c p . (E25)].mDi. to circumvent complexities associated with the multicomponent diffusion equation. XV.. but for usual flames. . = 4 ' 2 . its only influence appears to be a relatively small. N .m).cp. i = 1.The term ap/at has disappeared because it is of the order of the square of the Mach number if changes in P are assumed to be of order unity. in demonstrating the results of interest here.. Hydrodynamic and Diffusive Instabilities in Premixed Flames 345 It is not obvious that the crosstransport effects are unimportant in deflagration stability because it will be found that changes of diffusion coefficients or Lewis numbers by amounts on the order of 10 % can be significant. . The somewhat lengthy form of equation (93) is a consequence of including variable transport coefficients as well as variable heat capacities. . and Soret effects typically may reach 10% of those of the ordinary diffusion processes.VY. we shall neglect bodyforce and pressuregradient diffusion (likely good approximations) and shall employ the diffusion equation that would apply if the species i = N were present in great excess. m)'TmI. H i = hi/(cp. namely. and L = nm(V * q R ) / [ ( P m 0.1 where the Lewis numbers are Lei = Am/(p. With the nondimensionalizations i 2.9. = D. hi = PDi/(pm Di. Tm).
~) prodyct is equal to . instead of i = 1. v i = Fv. adiabatic flame would give the correct formula. then equation (578) is not quite right. v 2 = . . but a revised asymptotic analysis of the steady. (x/w) xr= (x. 1 ) = (p. (95) x. . tworeactant chemistry to be considered. 1).. we let i = 1. t'i/cIm)w/(Wlv. v. i = 1) . Equations (87).W. the species i = 2 may become the limiting reactant at various times and positions along the flame sheet. an interesting phenomenon may occur in the perturbation about the steady.modified as suggested by equation (B16). but we shall not discuss it here and shall assume that attention is restricted to conditions v. l/(Wlv. )N  1.~ w2 )/ (W. 1 ) for reaction products and vi = 0 for nonreactive species. P . J(W1v. The reactionzone equations " I> at . implies that the ratio W ~ / ( W ~for V~ any . where a dimensionless reactionrate function is + With the preceding assumptions and in the transformed coordinates. 1). For the reaction v..). and with R related to these quantities through the equation of state. Appropriate integrations of the differential equations across the reaction sheet serve to eliminate the rate term w from the system. A l # 1 or di # 1.346 Combustion Instabilities The equations for conservation of chemical species contain the reactionrate terms that will be nonnegligible only in a narrow region about 5: = 0. [ V1u.f)> where v1 = . lYJlz + produnder which it does not occur.v. thereby providing a set of reactionfree equations with reactionsheet jump conditions for analyzing the dynamics of the flame. equation (14) then becomes 6i =A X Lei +V~W + (Vld. RO = JF). vl.  (V. If specific heats vary. For nearstoichiometric conditions. . (89). and diffusion flames emanating from the stoichiometric points on the flame sheet may then be present in the downstream region.... 0. consistent with equation (578). For the onestep. and (95) are suitable conservation equations for describing the evolutions of u. lYJll ucts. . equation (1 4.1. An analysis of this effect has been developed [171]...W ~ / ( W ~ V= . (90). 2 identify the reactants and consider i = 1 to be the limiting or deficient reactant. planar solution. (93)...
9. m> is continuous.HN./a( is continuous [as may be derived directly from equation (95) by dividing by vi and adding to the equation for Y. Since the transverse coordinates and the time variable are not stretched. + Pr. oLel) be continuous. ~ ( h 7vihp)/(cp. P = EiQ/(RoL).9. from suitable combinations of equations (89) and (90) that P ./a< are continuous.in which Yl.Tm). However. Al. S)[au/d< (V.l These results require further that u. and the second states that the streamwise gradients of the components of velocity tangent to the sheet are continuous./a<] &.o is the initial mass fraction of the deficient reactant and the h: are the standard enthalpies of formation that appear in equation (11 1) and that are to be evaluated at the reference temperature T. f but this result is not sufficient because the integral is not readily expressed in terms of other variables. Hydrodynamic and Diffusive Instabilities in Premixed Flames 341 to be integrated may be developed formally by stretching the coordinate about ( = 0 by a factor fl (that is.)(dx/d()/Lei is continuous.. vL. The conditions obtained from equations (89). +  + + xY=. a more thorough asymptotic analysis of 1 (ay1/8i">/~el + . which (since vi = 0 by mass conservation in the chemical reaction) may be expressed in terms of the nondimensional heat release that (dY.5. to eliminate w prior to integration].. f ) av. p 2 .IfN. It does imply that (since v 1 = l)(ax/d()[Lel/ (Leivi)] aY. R . m .. The first of the conditions quoted from equations (89) and (90) expresses a pressure discontinuity that balances the discontinuity in the viscous stress tensor.(Pr. tion (93) that a9/a. and be continuous in the first approximation and rely on the assumptions that p l . and hi are continuous.)au/d( and (V. in the formulation that has been given here. (90). all the jump conditions except those of equation (95) are obtained relatively easily.' x (1 + IVIf12)1 Leivi wd(. simply by integrating the equations across the reaction sheet. When equation (95) is integrated across the sheet by use of equation (91) it is found that the discontinuity in a x / d ( is XI"=. (96) where the nondimensional heat released per unit mass of mixture has been defined as Q = Y l .. treating fl< as being of order unity) and then expanding the resulting equations in the small parameter fl. and (93) in effect involve only the longitudinal components of the stress tensor and of the heatflux vector. . and from equa( H i . In this way it is found from equation (87) that Ru is continuous.The relevant Zel'dovich number is < './a()Q/ appears in equation (96) as the requirement that a9/d( ( Yl. and this result may be used in the jump condition obtained from energy conservation to show that + do/at  ril i= 1 vi(Hi.
6.The differential equation describing the reactionzone structure in the first approximation. Because of the strong dependence of the reaction rate on temperature. The corresponding onereactant problem is readily . the expressions + + (97) 6 = 6.::n21 Lei' L ~ .. we have (a6/d[).1 is of order P' downstream. a). at the leading order in p. /at. Evaluation of these constants is aided by the further result that Yl = 0 (and hence aYl/a( = 0) downstream from the reaction zone.1 and has been developed for both onereactant [172]. also. in the formulation the order of magnitude of 11 .1/61] ~ 2G(n1.YiQ/(Yi. The continuity conditions derived previously enable other gradients to be obtained from this result. a)Y. Z ( .f may be treated as being of order p.oLel) and Y.348 Combustion Instabilities the reactionzone structure is needed for deriving a jump condition for a y. instantaneous values just downstream from the reaction sheet. Y2Le1/(v2Le2)are constant. Equation (99) relates the concentration gradient of the limiting reactant just upstream from the reaction sheet to the temperature and the concentration of the reactant in excess at the reaction sheet. Therefore.2.f . n2.oLel) to the lowest order. at least in the first approximations. so that for the reactionzone analysis. For most problems (for example. since 6 .V ~ ) "(98) ~' (I Le 1 + l V L f l 2 1 = in which equation (578) has been employed as a definition of G(nl. For example.= . gradients of the total enthalpy and of the mixture ratio of the reactants are negligible in the first approximation within the reaction sheet.the matching condition '. obtained from equations (94) and (95) (for i = 1) through appropriate scalings and stretching.2. The necessary application of activationenergy asymptotics parallels that given in Sections 9.oLei). [175] systems. so that in the scaled variables appropriate to the reaction zone.Q/(Yl. 5. Y 2 = Y2. for stability analyses). and 8. Introduction of equation (97) into equation (98) renders the righthand side a function of Yl. where the function G is defined below equation (578). n2.(dYl/at)c=o.' Y'2 ex~C(B/Q1(1 . to achieve the greatest flexibility in the analysis.. both 6 YIQ/(Yl.6. with 6.f constants representing the local. and Y2. [173] and tworeactant [174].3. it contains all the essential information associated with the reactionzone structure. is 1 a2y1 p + n z + 1 Y. .o.1 cannot be greater than P. thereby enabling the equation to be integrated to provide. Y2.YivzLez/Lei may be employed.
which implies that a mixture with a sufficient excess of species 2 behaves like a onereactant mixture. hydrodynamic effects and diffusivethermal effects.9.. 5 = 0. Results of this type have been developed within the approximation that the extent of thermal expansion across the flame is small [l5l]. and (95) with w = 0 comprise a reactionfree description of the flame dynamics. This would not be likely to introduce any appreciable simplification into the local analysis of the flame structure and dynamics.5.2.3). the wavelength must be large compared with the thickness of the reaction sheet and the frequency low compared with the transit time through the sheet. Disturbances with wavelengths on the order of the flame thickness (5 of order unity) are permitted here.1 becomes small. Cellular flames There are three basic distinct types of phenomena that may be responsible for intrinsic instabilities of premixed flames with onestep chemistry: bodyforce effects.. equations (87). (89).5. 9. t). If wavelengths are large compared with the entire flame thickness and frequencies are small compared with the transit time through the flame. (93). A significant contribution to the complexity of the equations that have been given here is the fact that the coordinate system adopted is noninertial and nonorthogonal. then ~Y. [176] and also for more general situations [177][180]. With equation (99) and the jump conditions that have now been obtained. These complications are offset by the simplification achieved by placing the reaction sheet at a fixed location. (90). For example.f becomes of order unity. These geometric aspects of global flame dynamics have not been investigated.3. z.. These results aid in simplifying analyses of flame dynamics and contribute to the understanding of factors influencing flame motions (see Section 10. Stability analyses are conveniently pursued by perturbing these equations about their steady. It may be noted that if a becomes large so that Y2./(v2Le2YI. These restrictions are satisfied well in a vast majority of problems. planar solutions. then a further integration across the flame can be performed to derive a single partial differential equation for f ( y .a) . but it could help in describing largescale flame motions and in allowing for situations in which the reaction sheet becomes parallel to the x axis at various points. Hydrodynamic and Diffusive Instabilities in Premixed Flames 349 recoverable by putting n2 = 0 [so that the functions G in equation (99) cancel] and then letting v2 + 0. Restrictions on the formulation pertain to wavelengths and frequencies of the disturbances. It may be possible to obtain an orthogonal system of coordinates by generating the coordinates on the basis of the instantaneous flame shape. and it is found from the definition of G that the ratio of G's in equation (99) approaches unity. Cellular flamesflames that spontaneously take on a nonplanar shapeoften have structures affected most strongly by diffusivethermal .
Therefore. However. the ordering.03 (upper left). and 458462 of [45]. propagating downward in an open tube 5 cm in diameter. in a sense.01 (lower left). 9. then hydrodynamic effects. for example.34 (lower right). then through dimensional reasoning.1. 372. with respect to the burnt mixture.Br in molar percentage of about 0. with a mole fraction of nitrogen of 0. flames propagating upward may be expected to experience an instability that mathematically is a consequence of the term F in equations (89) and (90)./g and a characteristic * See pages 371. it is desirable to discuss all three types of instabilities here. the cellular flames shown in Figure 9.57 (upper right). Bodyforce instabilities It is well known* that a lessdense fluid beneath a moredense fluid is buoyantly unstable [181]. and 2. Bregeon). phenomena.5.7. 2. Cellular premixed hydrogenoxygen flames. .8 and a molar ratio of hydrogen to oxygen of 1. is from largescale to smallscale phenomena. visibility provided by addition of CF. We shall first address bodyforce effects. 1. If the entire flame is viewed as a discontinuity moving at velocity v o with respect to the fresh gas and v. Therefore.350 Combustion Instabilities FIGURE 9.7 are largely controlled by a particular diffusivethermal process. and finally diffusivethermal effects. bodyforce and hydrodynamic phenomena also play interesting roles.2. a characteristic transverse length of the instability in the presence of a gravitational acceleration g may be guessed to be of order uou.2 (photographs taken by B.
the wavelength . Experimentally. sealed at the top and open to the atmosphere at the bottom. Nevertheless.5. theoretically [I1831 for the velocity at which an air column rises into a vertical tube filled with water. and achieves an approximately constant rise velocity of the nose (if the mixture is sufficiently weak to avoid the buildup of pressure that initiates transition toward detonation). the effective nondimensional body force per unit volume becomes R(F. Since the nose of the flame is curved and finds itself propagating into a nonuniform flow. Thus. an elongated flame having a roughly parabolic shape with its nose pointing upward rapidly develops. the streamlines of which diverge in approaching the flame. the flame motion in these upwardpropagation experiments is affected primarily by buoyancy.the measured rise velocities have been shown [182] to agree well with the value ub z 0 . propagates upward. typical wavelengths are estimated to be on the order of a few centimeters and growth times in tens of milliseconds.9. 3 m obtained .df/dz) is equivalent to a body force directed from the fresh mixture to the burnt gas. the flame shape is quite similar to the shape of the air column. A bodyforce instability is therefore associated with flame acceleration. it has been observed that for initially quiescent mixtures ignited at the closed base of a vertical tube open at the top in the laboratory. this last estimate suggests that the severity of the instability increases as the flame speed decreases.d2f/dz2). This configuration is similar to that in which the buoyantly convective Rayleigh instability involving formation of Benard cells often has been studied for nonreactive fluids. In a noninertial coordinate system that moves with the flame (or in which the flame moves at a constant velocity). By the preceding reasoning.2). an effective body force appears as a consequence of flame acceleration. Hydrodynamic and Diffusive Instabilities in Premixed Flames 351 growth time of order . As might be expected physically. Acceleration of a flame sheet produces effects of the same type as those of body forces. so long as uo < v b . . For weak combustible mixtures of different fuels in tubes of various diameters D. the flame structure at the nose will differ from that of a normal flame (see Section 10. and similar phenomena would be expected for premixed flames under appropriate conditions. but in the flame experiments stabilization effects for nonplanar disturbances are provided by interactions with the burner and by the divergence of the streamlines. In earth gravity (g z lo3cm/s2).3. instabilities have been seen in this configuration for diffusion flames on sufficiently large burners (greater than about 10cm diameter) [1871. The same mechanism is operative for steady flames of flatflame burners with the fuel flow directed downward. h o / g . Consistent with the preceding estimates. Buoyant instabilities have been observed for flames propagating in larger vented enclosures [1841[186]. . the singlelobe shape that is usually observed suggests that the wavelength of the buoyant instability is comparable with or larger than typical tube diameters. This may be seen from equation (89) by bringing Rd2f / d z 2 to the righthand side of the equation.. Thus a flame acceleration in the direction of its motion (a positive rate of increase of .
A pattern of buoyant convection then begins to be established. quantities decrease as the acceleration increases. Therefore. Because of other mechanisms of instability. analyses of flame str c ure generally neglect buoyancy. buoyancy may stabilize a deflagration without appreciably modifying the flame structure. There are special situations in which buoyancy can be responsible for instabilities in configurations that might appear to be buoyantly stable. ’ . Indeed. weak combustible mixture contained in a chamber open at the top experiences conductive cooling of the burnt gas at the walls of the chamber. dux// H 9. the bodyforce effects are stabilizing. For example. the work of G.2. a stability analysis that neglects body forces and that treats the entire flame simply as a discontinuity in density that * Body forces relevant to flame structure may be electrical forces acting on ionized species for flames in strong electric fields. the ease with which stable laminar deflagrations are observed in the laboratory may be attributable largely to the stabilizing influence of buoyancy.* Nevertheless i buoyancy is responsible for observed flame stabilitg then it must have a significant influence on conditions for the occurrence of cellular flames. flames propagating downward. Hydrodynamic instabilities A startling discovery in flame theory was made independently by Darrieus+ and Landau [190]. Phenomena of this general type have been observed experimentally [lSS]. it appears to be necessary to include buoyancy in analyses of stability boundaries. to be discussed later.v. in comparison with that of the combustion products in the center of the chamber. Paris (1945). Therefore.2./la2F/at21 and d 2 F / d t 2 ( These . This cooling increases the density of the burnt gas near the walls. a relatively slow flame propagating downward in an initially quiescent. a net rate of buoyant rise of downwardburning combustible gas in the center of a large chamber or in the open is a common phenomenon. with the flame near the wall moving well ahead of the flame in the central part of the chamber. as well as on the basis of the conservation equations of the previous section [178]. Thus body forces may produce a variety of observable effects on premixedflame propagation. we do not discuss influences of electric fields. La Technique Moderne (1938) and Congrts de Mkchanique Appliquee. This has been done with the effect of flame curvature on the burning velocity included in a phenomenological way [137]. and buoyancy forces may then promote downward motion of the gases near the wall. Normal buoyancy effects are relatively farfield phenomena in that within the thickness of the thin flame (including the preheat zone as well as the reaction zone). Because ofconditions in France during the second world war. For decelerating flames. with good justification. the buoyancy terms in the conservation equations are comparatively small. or burnerstabilized flames with the flow upward. [l89]. but he presented a paper entitled “Propagation d’un front de flamme” at two conferences in France.5.352 Combustion Instabilities and growth time of the instability may be estimated as v. Darrieus never was published.
< po (that is. with time derivatives and transverse gradients neglected within the flame. The first step is to obtain jump conditions across the flame by integrating the equations with respect to 5 through the entire flame.1/R. and integration of equation (90) yields Ru(v. The third step is to seek solutions to these linear equations with constant coefficients for perturbations that are sinusoidal in the transverse direction (for example. nevertheless is both correct and readily comprehensible (see below) and certainly requires consideration in connection with cellularflame phenomena [1371. = R(av. if Re{a} > 0. . A dispersion relation may be sought by letting the time dependence of the perturbations be proportional to ear. the linearization gives aqag + v.u .jag) .l)/RO . Since the flame speed is considered to be unmodified by the instability. = 0. that u approach U . largely contradictory to common experience in the laboratory. are q and [).P I ) .v. u + = 1 and u. proportional to Re{eik'X1}.. P = 0 for 5 > 0 and 1 . P (with R = 1 for 5 > 0 and R = R. with respect to the more dense gas demonstrates that since thermal expansion across the flame causes p. respectively). (100) Since R . which is given by u = U (a constant defined to be unity for 5 > 0 and l/Ro for 4 < 0).).jaz + u av. which satisfy the jump conditions at 5 = 0 and vanish as 5 + f co [thus requiring. and v.v. = constant for 5 < 0). The second step is to linearize the differential equations for the constantdensity flows on each side of the flame about the steady.aP/ag. (89) and (90). the planar deflagration is unconditionally unstable. in view of equation (88). u. equation (87) shows that Ru is constant throughout the flame. vI+ + u + V .. (where the subscripts + and . equation (100) gives o > 0 and hence implies instability. > 1. where CJ is the nondimensional growth rate. where k = Ikl.af/az as 5 + fco].P . R(au/az + a2f/az2 + u au/ag) = .identify conditions just downstream and upstream from the flame. the integral of equation (89) gives Ru(u+ . f . = 0. f ( P + .+ v. A quadratic equation for o is of C k g obtained. whichwhen combined with the preceding relationshipproduces the condition of continuity of the velocity components tangent to the flame. f = v.1]/(1 + Ro).= l/& (which is the ratio of densities of burnt and unburnt gas). for 5 < 0. > u. This result.V . and . Hydrodynamic and Diffusive Instabilities in Premixed Flames 353 propagates normal to itself at a constant speed u. planar solution.) = V.9.+ u . where the components of x. The hydrodynamic instability can be derived directly from equations (87). the parts of the perturbations that vanish at infinity are then found to be proportional to ekt for 5 < 0 and to a linear combination and e"< for 5 > 0.) = P .5.Clear expositions of the character of this hydrodynamic instability of flame fronts may be found in [9] and [153J. the acceptable solution to which is CJ = k[Jl + ( R i .
) at small values of (urn .vo)/v. may be written as 6. = uo. as indicated. it is clear on dimensional grounds that the growth rate must be proportional to the product of the wave number and flame speed.)./vo + 00. This causes a stream tube of initial area A (which must be the same far upstream and far downstream. Since the flame speed is the only dimensional parameter that appears in this stability problem.4  &TO/(v.8. and approaches unity as v. we see that continuity of the tangential component of velocity across a tilted flame element required the streamlines to be deflected toward the local normal to the sheet in passage through the flame. equation (100) states that the growth rate of the disturbance is proportional to the product of the transverse wave number with the geometricmean flame speed the factor of proportionality.8. approaches (v. the area increase causes a velocity decrease. . they play negligible roles in the problems addressed because the Mach numbers involved are very small.uo)/ (221. Schematic illustration of the mechanism of hydrodynamic instability. (J&+ urn00 . is positive when urn > vo. which vanishes if v. which depends only on the density ratio. That the constant of proportionality is positive may be traced to the fact that the flameshape perturbations are convected with the fluid.. since the perturbations vanish there) to be larger at the flame than far away if the sheet is convex toward the unburnt gas. In incompressible flow. Referring to Figure 9.* Since a flamesheet perturbation with an upstream flame displace * Compressibility and acoustics are absent from the equations of the previous section. + v.354 Combustion Instabilities FIGURE 9. In dimensional variables. . as illustrated.
D/6 2 100 [153]) to enforce a condition 1 < 1.. < 1 < 1. it encounters a decreased fluid velocity and therefore (propagating at its normal speed) will tend to move even farther upstream. then equation (100) becomes in which the new term is the ratio of a dimensional wavelength characteristic of the transverse disturbance../u. Hydrodynamic and Diffusive Instabilities in Premixed Flames 355 ment (F < 0) geometrically must have a curvature convex toward the unburnt gas. I = I. thereby causing the growth rate to achieve a maximum value at a particular wavelength appreciably greater than the flame thickness and to decrease thereafter with decreasing wavelength. therefore. a tendency toward further displacement downstream. nonoscillatory character (a is real). It can be shown [137]. 1 > I. [178] that if the hydrodynamic stability problem is addressed as before [190] but with the bodyforce terms in equations (89) and (90) retained in the differential equations. Discussion of these diffusivethermal phenomena is postponed to the following subsection. As the wavelength decreases and approaches the thickness 6 of the deflagration.9. then the hydrodynamic instability will not be observed. but it may be anticipated that they often introduce stabilizing influences. the resolution must lie in geometric or nonlinear hydrodynamic aspects of flame propagation and in effects of buoyancy. 2nuO u. it is no longer permissible to treat the flame as a discontinuity. Gravity thus introduces stability for disturbances of large wavelengths. planar flames in a vertically upward flow.. For these flames. 1. a downstream flame displacement results in streamtube contraction and. If a laboratory flame is contained within an apparatus whose walls exclude disturbances of wavelength greater than l. then the hydrodynamic instability is . buoyancy must be a significant contributor to stability. From equation (100) we see that the growth rate is inversely proportional to the wavelength 1 of the disturbance. which shows that since the perturbation of u vanishes at the flame.5./g E I h . stable flames can be observed in apparatuses with characteristic dimension D too large (for example. for any reasonable value of I.. The instability is of a purely diverging.. to the bodyforce wavelength. 271(a. dF/& equals the perturbation of u. Even for wavelengths on the order of 10 to 100 times the deflagration thickness. small wrinkles grow more rapidly than large wrinkles. Unfortunately..)/k = 1. Conversely. The range of wavelengths for instability now becomes finite. there are difficulties with this popular resolution of the contradiction between theory and experiment concerning hydrodynamic instability. if / h < I. For stationary. possibly becoming negative at a critical wavelength. diffusive and thermal effects within the flame may cause the flame speed to depend appreciably on the flame shape and on the local flow field [137].. A formula that contributes to this explanation is equation (88).
in a sense. leaving a stagnant layer adjacent to the wall downstream from the flame sheet [1923. downwardpropagating flames in tubes or channels with quiescent or irrotational flow upstream therefore have rotational flow downstream. Since 1.modified by allowing the area A far downstream to be less than that far upstream. From numerical experiments with equations describing the selfevolution of flame surfaces in the limit of small values of the density change across the flame. Buoyancy again may help greatly to stabilize the downwardpropagating flame against hydrodynamic instability. is proportional to u o u m . for propagation in channels this same effect is present because of the curvature of the flame. an additional factor of significance that may be dominant. the effect often can be substantial because typical values of 1.8. Thus selfturbulization of premixed flames was attributed to hydrodynamic instability (analogous. In the absence of disturbances like acoustic perturbations (which can flatten flames under appropriate circumstances [191]). for example. and moreover. However. In the early investigations of hydrodynamic instability [9]. flames propagating downward in tubes and channels tend to become convex toward the unburnt gas. The flow streamlines then assume the pattern illustrated in the upper part of Figure 9. The velocityjump conditions across the flame sheet readily show that all curved flames that are not everywhere normal to the incoming flow produce vorticity . for a given value of g the stabilization effect increases as the flame speed decreases. [190]. In most configurations flameinduced flow nonuniformities complicate matters. Thus steadily propagating. [198][200] that the hydrodynamic instability evolves . to the development of turbulence in shear flows). it has been inferred [1521.356 Combustion Instabilities eliminated entirely. the time available for growth of the instability is limited by the intersection of the ray with the wall. for horizontal flame sheets in earth gravity are on the order of centimeters. For an expanding spherical flame in an initially quiescent medium the increase of the area of the flame sheet with time is known to delay the onset of hydrodynamic instability [196]. except for the ray propagating along the centerline of the channel. since flow separation occurs at the wall as the flame discontinuity passes. curved flames in channels or tubes in the absence of buoyancy may represent a stable configuration achieved by the longtime. This viewpoint must be revised if the instability evolves to stable nonplanar structures. for flame propagation in channels in the absence of gravity is a rayoptics aspect of the development of hydrodynamic instability [193][195]. as suggested above. it was presumed that the instability evolves to a chaotic state characteristic of turbulence. but the rotationality per se does not appear to be a significant aspect of stability. Even for nonhorizontal flames such as bunsenburner flames. the stabilizing influence of buoyancy may not be negligible. [197] . It is likely that this last condition applies to experiments in which large planar flames are stabilized in a uniform flow [153]. nonlinear evolution of hydrodynamic instability.
mixture is written as a constant times 1 + 2/9. is a constant intended to describe influences of flamestructure modifications on the flame speed. a Fourier decomposition will introduce a correction term proportional to k 2 in a formula like equation (100) for 0. The flame speed with respect to the unburnt where the Markstein length. diffusive phenomena may play a role beyond their obvious relevance to the cusp structure at the ridges of the sheet. the socalled polyhedral flames [203]. Thus evidence is growing against the idea that in the absence of other physical phenomena. The changes in flame structure are associated with variations of the effective diffusivity of the limiting reactant [201]. corrugated shapes that have sharp ridges pointed toward the burnt gas (while the diffusivethermal instability evolves to a somewhat similar but unsteady shape that exhibits a chaotic behavior). the first reported observations of these effects appear to be those for tent flames with periodic ridges on bunsentype burners. Stability analyses like those discussed earlier can then be pursued for any value of 2.2.5. [202]. they develop pebbled surface structures (with average sizes on the order of 10 cm) that appear to be chaotic in some respects and.V 2 F for small deflections of the flame sheet. Significant diffusive influences in cellular flames were established in early work [137]. the hydrodynamic instability of deflagrations leads to turbulence. the formulation may be viewed as one that retains only the first term in an expansion for small flame curvatures. However. Although it would seem that the hydrodynamic instability should be the dominant phenomenon in these large flames. .9. 9.7 cannot be understood on the basis of the instabilities discussed thus far because the flame speed and the density change vary little from one photograph to another. Hydrodynamic and Diffusive Instabilities in Premixed Flames 351 to flames with steady. 2. propagate at about twice the normal burning velocity [152].5. Diffusivethermal instabilities The principal features of the flame shown in Figure 9. Of course. In particular since 1/94? = . here denoted by 9and taken to be positive for convexity toward the burnt gas and negative for convexity toward the fresh mixture). with the objective of calculating 2 from the fundamental chemical and transport properties of the system. on the average. without further inquiry into the flamestructure phenomena that determine 2. It is observed experimentally that as large (up to 10m radius) spherical flame sheets expand.and the new term will be most important for large values of k (small wavelengths). The effects were first addressed theoretically in a semiphenomenological manner [204] by allowing the flame speed to depend on the radius of curvature of the flame sheet (the reciprocal of the mean curvature. the question certainly is not completely resolved.3. a complete analysis entails studying the influences of the perturbations on the internal structure of the flamefor example. but the flame structure clearly varies greatly.
o(l . Extensive research on the theory of these flamestructure modifications has been pursued [170][180]. X Burnt FIGURE 9. = 0 at [ = 0. the burning velocity increases with increasing < * They were expressed in his early book.9 Schematic illustration of the mechanism of diffusivethermal instability. the burning velocity depends mainly on the local flame temperature.o at = c o and Y.eRut)] shows that Ru is proportional to exp[B(tlr . Nauk SSSR. The essential physical ideas apparently were first known to Zel’dovich. = Yl.1)/(2Q)]. 1944. . if this equation is integrated with Ru constant and with the boundary conditions that Y. Theory of Combustion and Gas Detonation.5.9. where the arrows indicate directions of net fluxes of heat and species. = Yl.1 but may be motivated from it by employing Ru dY.358 Combustion Instabilities Strongly diffusing diffusing reactant Normallv diffusing reactant Heat / \ U L Heat Fresh  U.* The instability mechanisms can be explained by reference to Figure 9. Moscow: Akad./a[’ as a rough approximation to equation (95) in the preheat (convectivediffusive) zone for i = 1. A key element in the explanation is the fact that for large values of the Zel’dovich number. [205][225]. This intuitively clear result is not entirely obvious from the formulation of Section 9. and we now have a reasonably good general understanding of the influence on flame stability of the diffusive and thermal processes that occur within the flame. then application of equation (99) to the solution [which is Y. that is./d< = d’Y.
then Tfdecreases at the bulge. If the thermal diffusivity equals the diffusivity of the limiting reactant (Le. and the reactant flux and conductive flux will both.Il for all species i. In this last situation. where the hot gas tends to surround the cold combustible. When Lei = 1 and hi = . In the lower part of the figure. wherever Y. However. In Figure 9. In the absence of hydrodynamic effects and any other I . it can be shown rigorously from equation (93) with L = 0 that If=. > 0 such that the flame is stable to disturbances with transverse wavelengths 1 < 1.= 1 perturbations. fundamentally simpler physical mechanism that is purely thermal in character and that tends to provide a stabilizing influence for Le.5. > l. if the thermal diffusivity is greater (Le. the heating rate is greater than that for the plane flame. ascauses the bulge to tend to become larger. In neglecting nonplanar influences within the preheat zone.* However. This reasoning suggests that the diffusive and thermal effects within the flame tend to stabilize the hydrodynamic instability (that is. and the local propagation speed therefore tends to be increased. therefore. Hydrodynamic and Diffusive Instabilities in Premixed Flames 359 flame temperature in the expected Arrhenius fashion. there is an additional. as indicated at the beginning of Section 9. < 1. From this. sociated with the increase in Tf. the temperature remains constant and equal to the adiabatic flame temperature. > 1. .) only if Le. produce a critical wavelength 1. and with all perturbation lengths large compared with the flame thickness * It must. then these increases are balanced in such a way that Tf is unchanged.9 the temperature is therefore constant everywhere on the reaction sheet and to the right thereof for this adiabatic system with equal diffusivities. = 0. heat is lost to the side from the bulge in the upper part of the figure. so that the propagation speed is reduced there (see Section 8. Conversely. which thus is seen to occur for Le. = 1. then Tf increases (as a consequence of the preferential diffusion of the reactant with respect to heat).2). however.1. the equation for evolution of the reaction sheet with . be increased.5. in comparison with those of the normal flame. be less than the critical value for pulsating instability. strongly diffusing reactant). the local increase in flame speed.9. = l). this reasoning is restricted to disturbances with transverse wavelengths that are large compared with the flame thickness. remains constant. if it is less (Le. the local gradients will be steepened sheet tends to reduce Tf. Increasing the rate of diffusion of the limiting reactant to the sheet increases the rate of heat release there and hence tends to increase the flame temperature T’. It is evident that a bulge in the reaction sheet extending toward the unburnt gas serves as a local sink for reactants and a local source for heat. while increasing the rate of heat conduction from the At the bulge. thereby giving rise to the diffusivethermal instability. weakly diffusing reactants). Hi remains constant and from equation (95) that Yi + vi Y. < 1. These effects cause the wrinkled reaction sheet to tend to become planar and therefore constitute a stabilizing influence.
At stoichiometric conditions for a mixture that has its maximum normal burning velocity at stoichiometry.360 Combustion Instabilities can be shown to be 1 + af/az = V .5. timedependent heatconduction equation for f + z in this case. often termed preferential diffusion of reactants. ( V . Upstream bulges are relatively enriched in the strongly diffusing reactant and downstream bulges are consequently enriched in the weakly diffusing reactant. If the limiting component is relatively strongly diffusing.. Our reasoning suggests that there is a critical value of the Lewis number of the limiting component. and it may be reasoned [201] the flame will evolve to relatively stable. # Le. like those seen at the top of Figure 9. they may be expected to be large compared with the smoothing effect discussed here if /? is large and if Le. which may therefore exert a largely . . For Le. and planar flames will not be observable in mixtures that satisfy this criterion. Since the Lewisnumber effects described in the previous paragraph involve changes in Tf that make themselves felt through the Arrhenius factor. The consequences of the instability in unstable situations are understood only partially. > Le. approached as z + co. For Le. nonplanar diffusivethermal phenomena will be enhanced as the instability develops. then the mixtureratio variations aggravate instability. the diffusive and thermal phenomena discussed here evidently do not always produce stabilizing influences for nonplanar disturbances. a twodimensional. therefore.2. < Le. The stable solution to this equation. Therefore. f ) . nonplanar cellular shapes.. there is an additional purely diffusive effect. mainly in their (thermal) effect on the flame speed through T in nearstoichiometric mixtures their influences are comparable in magnitude to those of the diffusivethermal interactions. that can be understood even more readily than either of the two effects indicated above. the mixtureratio variation of a weakly diffusing limiting reactant is a stabilizing effect. upstream bulges can tend to be suppressed and downstream bulges can tend to be amplified by the mixtureratio variations. These variations make themselves felt on the dynamics of the reaction sheet f . the steepening gradients associated with development of the hydrodynamic instability accentuate magnitudes of nonplanar diffusivethermal effects. there is nothing to stabilize the hydrodynamic instability at small wavelengths. In tworeactant systems with Le. Even for Le.. is the planeflame solution a f / a z = 1. < Le.7 (perhaps moving somewhat chaotically). hydrodynamic instability may prevent the plane flame from occurring unless walls or buoyancy provide sufficiently strong stabilizing influences. and planar flames in a uniform flow then cannot exist. When they do not....2. Contrary to the suggestion made in Section 9. The more strongly diffusing reactant can reach the sink at the upstreampointing bulge more readily than its weakly diffusing partner. > Le. is not too close to unity. the wrinkles induce local mixtureratio variations. say Le. such that diffusivethermal instability occurs for Le.
(102) in which the factor multiplying 1/P approaches 2[1 + 3(R. the corresponding value of Le. lowdensity layer of fluid placed in front of it).9. = 1 1 ~ P (2R..1)/4]. = p z = A1 = di = l). is just slightly below unity. which lead to cellularflame structures affected mainly by diffusivethermal phenomena. Thus it seems safe to conclude that Le. a value that may be compared with the limit 0. . the influence probably is somewhat greater than would be implied by equation (102) because the decrease in diffusivity with decreasing temperature helps to make the reaction sheet less accessible to reactants.ln R. the associated divergence of the streamlines as they approach the sheet provides a strain configuration known [221] to be strongly stabilizing to ripples for diffusivethermal processes. In fact. of the gas expansion associated with the heat release.7). Hydrodynamic and Diffusive Instabilities in Premixed Flames 361 controlling influence on the final cellular shape (a shape somewhat like that seen at the bottom right of Figure 9.37 at Ro = 2 and diverges as 4R. equals (48/n2) In 2 = 3. for most real combustible mixtures and. Theoretical analysis of a onereactant. has shown [1771 that if the transport properties are constant ( p . + 1. that the observed instabilities. That the latter influences are also stabilizing may be understood by realizing that the decrease in density with increasing temperature increases transverse distances through transverse convection and therefore reduces transverse gradients of temperature and of concentrations locally in the vicinity of the reaction sheet.9 (which is obtained at P = 20 for R . with the estimate 10 5 P 5 20. Although the curvature at the nose regions of large cells may be thought to be too small for diffusivethermal phenomena to be relevant there. + cx. Therefore. with a partially insulating. even if the original source of the instability is hydrodynamic. Le. the shapes of the cellular flames that develop may be dominated largely by diffusivethermal effects. it is important to know the value of the critical Lewis number. Since the density ratio R. Although our reasoning suggested that Le. as has been shown in a further investigation of variableproperty effects [223]. . > Le. for small reactionsheet curvatures and large Zel'dovich numbers. the reaction sheet can be considered to be moved to a less accessible position. we ignored the finite thickness of the reaction sheet (an effect of relative order 1/P) and did not specifically address influences. the diffusivethermal variations in T./h R. thereby are reduced (in effect. are in fact first initiated by the hydrodynamic instability. therefore. for R . within the flame. then Le.) x'ln(l + x) dx. as R. and the smoothing effect associated with constancy of Tf is relatively enhanced.5. typically has a value of about 5.5. adiabatic system with constant specific heats. For estimating when diffusivethermal instabilities may occur. = 1) to demonstrate the large influence of gas expansion. certainly does not exceed 0.
through equation (93). induce relative diffusion of the reactant and heat to an extent sufficient to produce fractional changes of order unity in the flame temperature if Le. In particular. The influence of viscosity within the flame on the hydrodynamic instability has been a subject of uncertainty since the early studies of instabilities [ 1371. while variableproperty effects decrease it somewhat. [219]. differs from unity by an amount of order unity. within the context of an expansion for long wavelengths of disturbances. There are nonplanar disturbances that. through Pr. showing that gasexpansion effects increase Y appreis greater ciably. [223] also provide expressions for the length Y (introduced at the beginning of the present subsection). If the coefficient of viscosity is constant. constantproperty analyses. For the investigation of diffusivethermal effects. the approach greatly simplifies the analysis and thereby enables qualitative diffusivethermal features shared by real flames to be studied without being obscured by the complexity of variations in density and in other properties. [171]. the viscosityinduced modifications in the convectivediffusive zone being cancelled by those in the reactivediffusive zone. the viscosity modifies the gas expansion only slightly. Physically. The asymptotic analyses [178]. With few exceptions [1771[ 1801. [223][225]. Therefore. shear viscosity introduces a very small stabilizing effect if the coefficient of viscosity increases as the fluid passes through the flame. does not appear in the dispersion relation and that. theoretical analyses of wrinkled flames based on activationenergy asymptotics usually treat departures of Lei from unity as being of order 1/p [170]. Changes in the flame temperature of an order no larger than 1/p times the adiabatic flame temperature are required if the response of the reaction sheet is not to become exponentially large. with this approximation it becomes feasible to admit disturbances with wavelengths less than the thickness of the preheat zone (but still large compared with the thickness of the reactivediffusive zone). In this approach it is usual to set v = 0. [175][180]. this ordering is dictated by the fact that exp[P(Bf . Through activationenergy asymptotics. recent analyses of diffusivethermal phenomena in wrinkled flames have employed approximations [208] of nearly constant density and constant transport coefficients. The known diffusivethermal instabilities all appear for [Lei . Although results obtained with these approximations are quantitatively inaccurate. [173].11 of order 1/p. it has now been shown [223] that Pr. [217][225].. [179].362 Combustion Instabilities By viewing the coefficient of 1/p in equation (102) as a quantity of order unity (even though it may be somewhat large numerically).1)/Q] appears in equation (99). thereby excluding the gasexpansion effects discussed above. and unity formally is of order 1/p. and in real gas mixtures this ordering usually is reasonable numerically. It . 9’ than predicted by constantdensity. and equations (93) and (95) are simplified somewhat. [ 1983[200]. we see that the difference between Le. equations (87)(90) are no longer needed. then there is no effect [178].
The implications of equation (103) are illustrated in Figure 9. n. u)/an. defined in equation (821).. heatloss plane.k ) = 1 + J1 + 4(a + k 2 ) . in which G and u are defined below equation (578). equation (103) may be (Le. Regions of diffusivethermal instability in a Lewisnumber. . Hydrodynamic and Diffusive Instabilities in Premixed Flames 363 is possible to include effects of heat losses by retaining L in equation (93).) applied. is defined by equation (81 1) so that in terms of the loss parameter I .9. as well as for adiabatic tworeactant flames [175].Le. L proportional to 8 . For tworeactant flames [175]. Here the heat loss has been assumed to be linear in the temperature increment [Newtonian. in which the real part of the radical is to be taken to be positive.10.5. both adiabatic [2 131 and nonadiabatic [I731 ( L # 0). replaced by the effective Lewis number Le. ln(l/p) = 1/(2p2).. where various stability boundaries are shown by solid lines in a plane of + k k k FIGURE 9. the ratio of the burning velocity of the planar flame with heat loss to that under adiabatic conditions. The nonadiabatic result is [ 1731 where r(a. with Le. we have [from equation (822)]. and p.. Dispersion relations have been developed through activationenergy asymptotics for onereactant flames. with dispersion relations illustrated by insets.O0 in equation (93) or IC proportional to z in equation (8lo)].10. . aG(n.
10 to narrow from 2 < P(Lel . > 0 for stability even in the range of diffusivethermal instability. the dashed curves represent Re{@}for conditions under which CT is complex.1) < 21/2 under adiabatic conditions ( p = 1) to 0 < P(Le. The critical value Le. the former value. of course.1) of the departure from unity of the effective Lewis number of the limiting reactant as a function of the nondimensional measure l/pz = ln(l/p2) of the rate of heat loss.1 and discussed in Section 9.f proportional to exp[z Re{o}]Re{ei[k'xl+r'm~u~l }. One is the planar pulsating instability. . the solution to equation (103) for (r as a function of k . below which the cellular instability occurs is seen to increase from 1 . 9. corresponding to the lower branch of the burningvelocity curve in Figure 8.364 Combustion Instabilities the measure P(Le.10.10. they are likely to remain qualitatively correct under more general conditions. this region includes Le. . which was based on the restriction that the wavelength be large compared with the flame thickness. . agrees with that given by equation (102) for R. = 1. this stabilization at short wavelengths. at values of k above that at which CT passes through zero). The stable and unstable behaviors are illustrated by the insets in Figure 9. this occurs at the largest values of the Lewis number. ln(l/p2) > 1 in Figure 9. As the heat loss increases. For traveling waves. the region of diffusivethermal stability is seen in Figure 9. giving . [226][229] that in many circumstances oscillations are capable of being transmitted directly along the flame surface. the dispersion relation exhibits a maximum growth rate at a preferred wavelength and damping at sufficiently small wavelengths (that is. It has been found experimentally [153]. k is positive and CJ is complex. could not be obtained in our earlier reasoning..10 for sufficiently large values of the Lewis number. Although these results depend quantitatively on the approximations. it is not yet clear whether observed traveling waves correspond to the traveling diffusivethermal instabilities illustrated in Figure 9. the range of k over which Re{@}> 0 extends to k = 0).5.2//3 under adiabatic conditions to unity at the flammability limit. producing a critical wavelength 1.10. consistent with our earlier reasoning. The other two are instabilities that involve waves traveling along the reaction sheet. which are sketches of the dispersion relation.10. Three additional types of diffusivethermal instability appear in Figure. indicated at the beginning of Section 9. = 1. of Le. In the range of cellular instability. The steadystate solutions at values of p below l/& [that is.21 are found to be unstable for all values of Le.2. There is a narrow range of parameters in which > 0) only in the traveling modes and another instability is experienced (Re{@} range in which instability is experienced in both traveling and pulsating modes (that is. Because of the constantdensity approximation that underlies equation (103). In the dispersion relations sketched in Figure 9. such as Newtonian loss.1) < 18/5 at the flammability limit ( p = 1/&). The character of these travelingwave instabilities has been investigated theoretically [220].
2. REFERENCES 1. greater than those of real gas mixtures. with no observable reaction between the stripes.References 365 It is seen in Figure 9. M . A. 4th Symp. The cells seen in Figure 9. The cell sizes can be correlated on the basis of the wavelength for the maximum rate of amplification of the cellular instability. Smith and D.10 [202]. leaving a mixture between the cells that is too lean to burn. Near the flammability limit. A similar mechanism could apply to earlier observations of instability in coflowing hydrogenair diffusion flames [233]. and the cellular and travelingwave instabilities may be expected to be observable more readily. There must be inaccuracies in the correlation because of the constantdensity approximation and because of the evidently nonlinear character of the structures in Figure 9. The increase in size with increasing hydrogenoxygen ratio is associated with the increase in this wavelength. 893906. Research on nearlimit behavior is continuing. Diffusivethermal instabilities rarely occur for diffusion flames.* The diffusivethermal instabilities that have been considered here pertain entirely to premixed flames. over most of the range of heat loss. Williams. The mechanism of this instability must be preferential diffusion of hydrogen. CST36. * A step in this direction has been taken by T. 1950. 235 (1984). seen at the bottom of Figure 9. Internal Ballistics of SolidFuel Rockets. F. the travelingwave instabilities appear to occur at values of Le. to the combustion zones. 12th Symp. (1969). [230]. patterns have been found [232] (at sufficiently large dilutions of the stream of hydrogen coming from the cylinder) in which the combustion occurs in periodically positioned stripes wrapped around the cylinder. [231]. . New York: McGrawHill. Barrere and F. 169181.10 that the stable range of parameters is narrowed near the flammability limit. produced by the increase in the effective Lewis number identified earlier. (1953). Mitani. In particular. Neither data on the stripe size and spacing nor theoretical analyses of the instability have yet been published. R. but near the limit these values decrease into an accessible range. in the forward stagnation region of a porous cylinder. N. In a diffusion flame between air and hydrogennitrogen mixtures. through nitrogen and combustion products. Sprenger.7. but they can be observed under certain circumstances. Wimpress. R. Nonlinear analyses of flame shapes of cellular flames are needed.7 arise largely through preferential diffusion of hydrogen to the bright zones. P. 3. with the consequence that there is insufficient hydrogen to support combustion in between. there are special effects that can widen somewhat the range of propagation of cellular flames [222].
Eversman. vol. Mohring. vol. H. Culick. F. Jet Propulsion 24. F. 117. SOC. AGARDograph No. Buffum. Acoust. McClure. 1420 (1982).. F.Am. SOC. 16. Van Moorhem. Acoust. Pure Appl. Grey. K . Green. Landau and E. 6. AZAA Journal 1. 38. 34. Paris: The Advisory Group for Aerospace Research and Development. (1955). Barrere. Koch and W. Am. Tsien. AGARDograph No. W. 137. L. R. Stewartson. M. 11. 36. Feshback. Crocco and S. Fluid Mechanics. Geckler. W. 475 (1 974). R. 8. Hart and F. T. 1960. W. Dehority. 25.366 Combustion Instabilities H. 21. R. 79 (1949). VI. 109 (1975). Cantrell and R. R. Culick. F. 33. E. E. Cantrell. Lambert.697 (1964). R. T. 23. F. L. W. K . 655 (1956). R. Annual Review of Fluid Mechanics 9. 1 16. . Williams. 13. 1 of Progress in 4. Crocco. 35. 22. B. AIAA Journal 1. Price. Jr. Journal of Sound and Vibration 74. 29. Hart. T. Hart and R. 2940. 15. E. F. CST 10. Rocket SOC. 18 (1963). Espander. L. Hart. A. Culick.Am. McClure. 1068 (1953). I . Baer. Mitchell and M. M. J. 9. R. Baer and C. (1965). New York: McGrawHill. 12. L. P. Hart. 0. 1953. R. Fundamental Aspects of Solid Propellant Rockets. Slough. F. A I A A Journal 12. I. W. 28. R. G. M. 1959. 37. “The Theory of Unsteady Laminar Boundary Layers. W. 30. 5. J. Theory of Combustion Instability in Liquid Propellant Rocket Motors. W. Morse and H. AZAA Journal 20. Crocco. 586 (1963). and N. 103 (1981). Methods of Theoretical Physics. Lifshitz. R. London : Butterworths Scientific Publication. 46 (1953). Acoust. 27. 24. K. 1960. 32. 135 (1977). W. S. R. 31. Cheng. Astronautica Acta 12. 36. Stewartson. L. J. Chapter 6. Bird. 18. Huang. 8. L. 5th Symp.” in Solid Propellant Rocket Research. Ffowcs Williams. Vibration and Sound. C. A. Am. Acoust. C. Jet Propulsion 26. Cantrell. AZAA Journal 13. 1964. 2. W. J. 35. M. New York: McGrawHill. Hart. L. C. 10. London: Pergamon Press. Grad. vol. and J. 1097 (1963). NATO. 1948. Culick. 17.. 447 (1977). 1967. D. P. The Theory of Laminar Boundary Layers in Compressible Fluids. CST 7. AIAA Journal 21. Behavior of Supercritical Nozzles under ThreeDimensional Oscillatory Conditions. Comm. Cantrell. Monti. 165 (1973). F. Aerotecnica 33. C. E. 25. 19. E. G. and J. Slates. 771 (1961). Am. AZAA Journal 15. H. 14. England : Technivision Services. F. and R. F. H. New York: Academic Press. J. J. 26. E. Morse. 10471065. 27. J. Crocco and W. Mitchell and W. D.500 (1963). Lindsay. T. E. H. R. 1107 (1975). and R. 1956. E. and E. Mitchell. Jr. AGARDograph No. Astley and W. AIAA Journal5. Oxford: Clarendon Press. Soc. 7. 113 (1966). C. Cheng. C. 272 (1967). R. McClure. 1969. 102 (1954). 20. Math. I. Baer. H. R. M. Mechanical Radiation. “Solid Propellant Rocket Motors as Acoustic Oscillators. C. 200 (1983). McClure. 35. Sirignano. F. S.22. ARS Journal 31. M. SOC. 139 (1952). R. 10th Symp. New York: McGrawHill.” in Advances in Applied Mechanics. R.
S. 73. R. J. (1963). 68. Hart and R. T. 1. J. Jr. Acoust. 1867. V. J. S. 64. 52. Phys. Hydrodynamics. R. L. H. Williams.. 49. J. 9th Symp. 303310. 33. H. 418 (1965). 63. Nachbar and L. “A Theory of Oscillatory Burning of Solid Propellants Assuming a Constant Surface Temperature. W. J. R. M. 54. Am. Baum. 1973. 1106 (1964). 319 (1878). J.References 367 Astronautics and Rocketry. Epstein and R. 48. F. C. R. The Theory of Sound. Lamb. Mekh. 4. (1962). 295358. London 210A. T. D. T. Trans. 645663. vol. AZAA Journal 2. Hart and F. AZAA Journal 1. M. J . 1945. 47 (1972). Phil.. 886894. Moscow : Nauka. W. Denison and E. Temkin. 70. Zhur. C. S. 577 (1960). 2241 (1968). Temkin and R. Culick. 51. 30. 884 (1960). M. 61 (1956). Culick. G. J. F. 75. T. C S T 5. E. E. Sound. Phys. J. Temkin. Dobbins. W. 45. 40. B. J. 12. Prikl. E. Novikov and Y. 378 (1962). C. and M. R. 375392.no. J. Am. W. Deters. 32. 71. Price. Green. Acoust. Summerfield. M. A. McClure. H. P. Y. F. F. 518 (1959). Hart and F. 1610 (1966). A. Novikov and Y. F. New York: Dover. Prikl. Dobbins and S. 57 (1966). F. 386 (1958). Y. A. Soc. Zel’dovich. 76. Fiz. Friedly and E. Bernard. S. Carrier. 6. Appl. 1501 (1959). Tekhn. Fluid Mech. 1 1 12 (1961). 69. Soc. Roy. ed. F. 1016 (1966). W. 278 (1968). no. Zhur. T’ien. and J. T. 59. 40. J. 61. Dobbins and S. A I A A Journal 4.” in Solid Propellant Rocket Research. Prikl. A. Sirignano.. Appl. 55. 56. 53. Imber. I. 1932. A R S Journal 31. 1960. J.. H. AZAA Journal5. J. 398 (1963). Smith. Fiz. Summerfield. 41. 31. 74. A. 50. Williams. S. Am. McClure. 1100 (1964). 67. Blackman. Green. Tekhn. M. 2182 (1967). 39. 46. Fluid Mech. Novozhilov. 498 (1942). Fluid Mech. Tekhn. 221 (1967). Petersen. Ryazantsev. Zel’dovich. Sewell. 1945. Zhur. 60. A R S Journal 32. T’ien. J. Krier. 239 (1910). A. 44. W. AIAA Journal 3. V. Barrkre and J. Lord Rayleigh. Bird. 57. New York : Academic Press. 58. New York: Academic Press. 26. 3153 (1962). Acoust. 376 (1958). Cantrell. Nonstationary Combustion o f Solid Rocket Fuels. 8th Symp. 66. 25. Nature 18. Jr. J. Jet Propulsion 28. H. B. 65. AZAA Journal 11.. ed. G. 0. Cheng. S. McClure. 72. W. no. 77 (1964). R. S. A. (1962). 9. Sci. F. New York: Dover. R.. AZAA Journal 2. Cantrell. S. 3. 1604 (1966). 1960. C. 6th ed. 2. Cantrell. F. Teor. S. AZAA Journul6. S. 1413 (1960). Zhur. 77. . Soc. Eksp. Ryazantsev. AZAA Journal 4. 553 (1953). Astronautica Acta 13. Appleton and Co. 62. Clarke. 43. 42. Hart. New York: D. Soc. W. Fiz. Phys. 7. J. Hart. R. J. Bird. R. 1328 (1973). F. 1 of Progreys in Astronautics and Rocketry. 226235.. 8th Symp. 126 (1964). J. Fiz. Mekh. AZAA Journal 6. R. 47. Lord Rayleigh. Summerfield. Tyndall. M. Aero. Horton and E. 8196.. McClure and R. Chem. Carhart. S. Mekh.
1978. F. 111. E. AZAA Journal 2. 1638 (1963). E. 799 (1979). 1 (1964). M. Culick. AZAA Journal 17. SZAM J. T. W. T.368 Combustion Instabilities 78. 63 o f Progress in Astronautics and Aeronautics. J. J . E. ARS Journal 32. J.. Comm. and S. and R. 383 (1955). Fiz. 109. CST 1. N . T. Sirignano. B. E . 18. B. Crocco. Summerfield. AZAA Journal 19. 12th Symp. 95. C. B. New York: American Institute of Aeronautics and Astronautics. 99. 9th Symp. 311315. Phys. 913 (1981). 91. Williams and G. and D. Merzhanov. 1 of Progress in Astronautics and Rocketry. vol. Bird. 1319 (1964). Culick. B. 1960. 97 (1969). 80. Wood. 12. B. Zel’dovich. 94. T. Appl. 1960. 9th Symp. Leipunskii. Witham. 1 (1971). C. S. A. I. 374 (1962). I. G. Brown. Williams. F. 107. Price. Summerfield. Fluid Mech. ARS Journal 32. and S. 101. M. L. W. 7. Math. 93. 643 (1962). Khaikin. (1963). R. 1285 (1964). H. Kumar and F. L. (1969). 335344. J. W. Brownlee. Cantrell. Cheng. B. B. AZAA Journd 2. T. JournalofMnth. 110. 79. 83. Librovich. ed. K. 1625 (1963). T. CST 20. C S T 3 . 1 of Progress in Astronautics and Rocketry. 191200. Zinn. Bird. 96. Cohen. W. McClure. 0. and B. W. Horton. Hart. F. 393422. N. Lighthill. 42345 1. Collins. W. Price. CST 15. L. D. AZAA Journal 2. 179 (1977). 1274 (1964). G. L. Sivashinsky. 33. 1975. J. . 88.. 87. W. W. R. Phil. L. A. I. and Phys. F . 465 (1978). E. 90. AZAA Journal 19. B. Powell. 10671082. Hart and F. F. 2112 (1965). New York: Academic Press. W . C. Math. Pfaff. Math. F. W. 86. C. 338 (1952). 98. T. 92. Dickinson. (1963). 104. 81. 5. ed. K. Theory ofNonsteady Powder Combustion. M. F. T . Sirignano and L.” in Solid Propellant Rocket Research. 82. R. W. 105. Pure Appl. Cheng. S. Vzr. C. Hart. AZAA Journal 3. Phys. J. Chu. F. 907 (1981). and V. 85. H. 101113. Rice. Price. Shkadinskii. W. L. A. Chu. 275 (1964). 102. New York : Academic Press. 1270 (1964). Eisel. and E. vol. 19 (1971). Chester. and F . G. (1965). Moscow: Nauka. 97. E. Lin. vol. 108. 179 (1970). AZAA Journal 2. Strittmater. E. Watermeier. E. Price. and A. in Experimentd Diagnostics in Combustion of Solids. Culick. “The Influence of Erosive Burning on Acoustic Instability in Solid Propellant Rocket Motors. Crump and E. 35. R. Appl. AZAA Journal 2. 179 (1979). A. C. R. Fluids 6. S. eds. 89. 117 (1954). I. W. Gor. 10th Symp. Zinn. S. CST 2. 84. Caveny. G. Carrier. Zinn. J. W. Fluids 6. 100. 106. 35 (1969). Quart. Astronautica Acta 14. E. “Experimental Methods for Combustion Admittance Measurements ”.” in Solid Propellant Rocket Research. C. 1179 (1949). M. L. Mitchell. R. McClure. T. Culick. A. A. M. F. E . Price. McClure. Lengelle.. Padmanabhan. Y. Boggs and B. G. A. Mag. Matkowsky and G. “Combustion Instability in Solid Rockets Using Propellants with Reactive Additives. 103. 40. Crocco and W.
Strahle. 119. A. 127. A. Appl. New York: Macmillan. Beer. 141. “Review of the Combustion Instability Characteristics of Solid Propellants. 1968. Harrje and F. 193201. Government Printing Office. (1969). C. S. W. Reardon. 138. honSteady Flame Propagation. Barrere and J. C. 1968. Nauk SSSR 239. 122.” in Combustion Technology: Some Modern Developments. M. 659663. 11011128. AGARD Conference Proceedings no.. H. NASA SP. A. Clemmow and J. 127162. U S . Ross and P. Huffington. Angelus. 183192. 1194 (1965).194. Maksimov. 139. D. 1974. E. R. 133. Dokl. Engineering Cybernetics. 1954. Datner. England : Technivision Services. Brown. 131. 129. (1969). J. P. 128. A. 137. W. A. 1128. M. S. Yount and T. 12th Symp. Fiz. 24 (1965). B. 140. de Saint Martin. 12th Symp. Trans.C. 115. C. 52. 126. G. (1963). P. W. W. 123. and V. Palmer and J. Merzhanov.50. S. T. “Les Instabilites de Combustion dans les Fusees a Propergol Liquide. Putnam and D. F. G. (1965). C. “Combustion Noise: Problems and Potentials.References 369 112. (1959). S. R. Price. England: Technivision Services. AGARD. 957 (1965). Akad. Combs. and I. 130. Tsien. M. Corbeau. 113. AZAA Journal 2. 139194. P. P. Strahle. 352376. Dokl. 747754. (1969). CST 10. P. 125. ed. Priem. G. and K. K. C. G. 11th Symp. S. “Secondary Nondestructive Instability in Medium Size Liquid Fuel Rocket Engines. S. Huffington. Penner. C. 1971. G. 117. Selected Combustion Problems. Crocco. Y. SIAM J . 134. Maidenhead.. (1955). Pariel and L. ed. Kydd. 10th Symp. E. A. Strahle. 121. 124. 40. 7th Symp. Inami and H. 5th Symp. Crocco. Math. Merzhanov. D.. New York: Academic Press. SalC. A. H. Penner. 135. (1967). 1. Shkiro. 255 (1978). A. 136. H. AIAA Journal 2. Strahle. London: Butterworths Scientific Publications. Faraday SOC. 94110. . Braunschweig (1962). Akad. P. 1. Nauk SSSR 208. I. 116. J. 1307 (1964). M . Borovinskaya. M. W. ed. 1314 (1964). Combustion Driven Oscillations in Industry. Datner. 1. 383419. 10th Symp. R. 9th Symp. 892 (1973). 1954. Mauss. Fontaine. 8599. G. Liquid Propellant Rocket Combustion Instability. A. Sivashinsky. Thring. Fundamentals and Aeronautical Applications. D. AIAA Journal 3. H. J. 120. A. Putnam. and L. A. 118. S.. h. I . Ivleva. 385 (1956). J. Merzhanov. New York: Elsevier. Shkadinskii.” presented at Fifth AGARD Combustion and Propulsion Colloquium. 12th Symp. Levine. Filonenko.” in Advances in Tactical Propulsion. 132. 942 (1954). D. 982992. P. M. Trans. Maidenhead. Penner and P. D. L. J. T. Borovinskaya. Washington. 114. (1965). 12411249. 1964. H. Perthuis and B. 13151325. L. Shanfield. (1965). 432 (1981). eds. Markstein. AGARD Conference Proceedings no. eds. Merzhanov and I. Gor: Vzr. New York: McGrawHill. G . E. 195 (1975).. Faraday SOC. R. S. 10th Symp. W.” in Advances in Tactical Rocket Propulsion. (1972). AIAA Journal 3.
Librovich and G. (1957). Agranat. J. 161. B. S. B. Ludford. A S M E Trans. F. Phys. 169. Joulin. Lawn. C. 22. eds. 18th Symp. Mathes. SIAM J . 147. J. Nauk SSSR 241. Appl. Sivashinsky. Bertsun. 237244. London 253A. 733. Tanno. GarciaYbarra and P. W. P. I. Joulin. 40. D. (1981). Appl. E.” in Combustion in Reactive Systems. Sci. R. 1980. to appear 154. 150. C & F35. 67 (1980). 193 (1982). 148. F. 551 (1981). J. G. Cambridge: Cambridge University Press. The Mathematical Theory of Combustion and Explosion. J. R. Dokl.370 Combustion Instabilities 142. Law and F. B. Clavin. G. Math. Strand. Margolis and B. 229 (1954). 156.696 (1964). BunkyoKu. K. Clarke. 335 (1973). Appl. D.530 (1974). B. A. Buckmaster and G. Rosen. 1989 (1967). 290 (1980). M. Olagunju. J. Moscow : Nauka. Manson. B. Bowen. W. I. I. J. N. 12th Symp. “Low Frequency Instability in Solid Rocket Motors. (19821. 46348 1. 170. Tokyo (1959). N. B. W. Proc. T’ien. Energy Combust. J. W. Beckstead and E. and A. SOC. Ser. C.” in Proceedings qf the First Symposium (International) on Rockets and Astronautics. 155. J . 271 (1982).Olagunju. Annual Review of Fluid Mechanics 15. I. Rosen. C & F 1. Joulin and P. 638 (1975). eds. J. 168. G. Layzer. 153. 166. Math. 157. J. I. H. 171. C. Price. 164. Math. Akad. Soloukhin. 1981. 144. CST 27. Sivashinsky. G. 145. M. Bowen. A. J. and F. N. A. 1335 (1983). Matkowsky and D.. London. Sci. E. 149. 39. Beckstead. McIntosh and J. 158. Golovichev. Clavin. Margolis. 172. Chem. 743 (1954). Soc. 152. “The Time Response of a Premixed Flame to Changes in Mixture Strength. SIAM J . Yokendo. 173. “ CrossTransport Effects in Nonadiabatic Premixed Flames. B. F. 65 (1962). 19th Symp. . C. S I A M J. V. Wehner and J. CST 22. 76 of Progress in Astronautics and Aeronautics. Matkowsky. M. W. 367 (1979). 120 (1970). 39. S. G. 7482. 146. Research A l l . Clavin. AIAA Journal 4. 162. A. J. Appl. Math.” in Combustion in Reactive Systems. McIntosh. Appl. 165. 22. M. Adler. Rosen. 62. L. 143. R. AIAA Journal 8. Barenblatt. 76 of Progress in Astronautics and Aeronautics. Proc. 1622 (1966). Summerfield. 380 (1959). J. 179 (1983). Journalof Heat Transfer 96. B. Roy. Akiba and M. (1985). Baer. 159. Y. Ryan. Clarke and A. 167. C. M. G. W. Phys. AIAA Journal 5. Makhviladze. Grishin. Williams. Sivashinsky. 1982. 339 (1957). vol. 203211. A. CST 27. Buckmaster. J. B. SZAM J . Sirignano and M. Sehgal and L.. Chapter 6. G. Prog. Theory of Laminar Flames. 139 (1979). 6th Symp. AIAA Journal 2. 1981. and V. CST 6. K. J. I. 163. I. New York: American Institute of Aeronautics and Astronautics. Beckstead. R. vol. 0. K . Oppenheim and R. A. Chapter 11. N. Zel’dovich. Oppenheim. J . I. and R. G. C . 151. Roy. M. 0. D. 372A. Chem. V. 679693. V. S. Chem. 160. C & F 46. S. Matkowsky and D. Culick. Phys. Menkes. 43. J. 236241. 443462. Price. 83 (1981). (1969). Manson. J. W. B. 305 (1978). 143 (1980). Sivashinsky. 222. 174. New York: American Institute of Aeronautics and Astronautics. S. Soloukhin. P.
181. 175. 219 (1982). J . A. C & F 40. Williams. . G. Taylor. Mat. eds. Khaikin. Sivashinsky. 433 (1969). Chem. Zel’dovich. L. Michelson and G. London 201A. P. Sivashinsky. Fluid Mech. 33 (1978). M. B. Aero. 251 (1982). 72 (1959). J . Zalosh. Istratov and V. 206. C & F 39. H. A. 385 (1950). Clavin. A. 451 (1966). 225 (1981). 787819. 177. 197. 235 (1981). Beason. Pappas. C & F 40. Istratov and V. Solberg. 1843 (1968). Uberoi. 188. J. Mekh. 203.. 771785. J. Proc. Michelson and G. A. SOC. Fiz. 215. Gor. Landau. D. Joulin and T. Williams.61. and D. G. P. 128 (1975). 48. Sivashinsky. 210. Barenblatt. Fire Safety Journal2. G. Baratov. D. J. B. 1982. Istratov. Williams. 201. 18th Symp. de Michanique 1. Astronuutica Acta 14. G. Lee and C. J. Alvares. 178. J . Guirao. W. C & F 28. Sci. 194. Y. Markstein. and V. 124. Proc. B. 125 (1976). Phys. Phys. G. Fluid Mech. A. B. G. 209. 365 (1954). I. Y. N. Zel’dovich. 205. Gor. C S T 15. Teor. Nauk SSSR 24467 (1979). 569 (1979). C & F 48. 18. 200. Vzr. D. D.L1 (1983). 11. Smithells and H. Vzr. 179. 187. N. I. Fiz. J. 134 (1965). I. Fluid Mech. Mekh. J . I. A. I (1980). J . eds. Sivashinsky. S. 1207 (1977). 11. 890 (1975). 199 (1951). M. G. Rozlovskii and Y.240 (1944).” in FuelAir Explosions. 192. 480 (1953). B. G. Dokl. 225 (1977). Proc. Levy. Clavin and F. A. Peke and P. G. Lovachev. 185. deRis. 293 (1962). C & F 33. A. Chem. Orloff and J. Parlange. Zel’dovich. 211 (1982). Lovachev. H. 190. Sivashinsky. 211. Librovich. 183. 182. 176. Fluids 2. 16071614. Prikl. A. J. 889 (1916). Kidin. Chem. J . Acta Astronautica 6. 196. Prikl. Phys. Canada: University of Waterloo Press. C & F 27. Clavin and G.” in FuelAir Explosions. M. Zel’dovich. Fiz. A. SOC. J. Ingle. M. Bregeon. A. 4. G. G. B. 239 (1983). 124. I. 212. M. Journal de PhysiqueLettres 44. Skramstad. no. (1971). I. no. Mitani. P. T. I. “Gas Explosion Research at Factory Mutual. S. 198. 184. Librovich. Lee and C. Zhur. Krivulin. 10. 204. Buckmaster. 169 (1980). I. Librovich. “Gas Explosion Research Related to Safety of Ships and Offshore Platforms. G.. 80 (1961). L. 214. S. Roy. Parlange. G. M. D.London 283A. G. Zhur. B. 192 (1950). Waterloo Canada: University of Waterloo Press. V. T. 1. SOC. A. 137 (1977). P. I. A. 193. 1177 (1977). Sivashinsky. Waterloo. B. Fiz. L. Akad. Matalon and B. C S T 24. Aldushin and S. I. London 200A. 30. Fiz. R. Matkowsky. Mekh. P. 186. Parks. Roy. Taylor. Acta Astronautica 4. A. J . and E. J. and A. 191. M. B. Kudryavtsev. Prikl. Istratov. Roy. 204. J. Dokl. Tekhn. 21 (1962). Y . Y . I. B. 217 (1882).. Zhur. Nuuk SSSR 57. Actu Astronautica 3. 213. 199. 116. 180. J . Guirao. Acta Astronautica 4. 189. Einbinder. I. 21. L. 237 (1979). (1981). 14. Y. H. 979992. 202. Davies and G. Akad. 208. N. and F. D.References 37 1 G. Mitani and F. A. Y. R. and A. A. Aldushin and B. Chu and J. 102 (1966). 207. 1982. Fluid Mech. Eckhaus. N. Kasparov. Solberg. S. 195. SOC. M. Zel’dovich. 13th Symp. Tekhn. M. Joulin. H. Gordon. E. Eksp.
J. L. 255 (1981). SZAM J. Joulin and G. T. I. Tsuji. 0. Matkowsky and D. GarciaYbarra. 38. K. Mitani. Nicoli. 18th Symp. P. Matalon and B. Math. A. Petersen and H. de Mkchanique Thkorique et Appliquke 2. 223. 217. 42. M. and G. I. CST 28. Sivashinsky. 228. Matkowsky. 225. M. 220. Joulin. Putnick. and G. 489 (1980). I. Clavin and C. 233. Ishizuka and H. Jet Propulsion 26. 669 (1979). C. 695703. J. J. 218. CST 23. . 456 (1961).1 (1985). 490 (1956). 245 (1983). C & F 6. 232. Math. W.486. Clavin and P. T. Appl. SZAMJ. 44. C & F 60. 1138 (1982). T. 327 (1984). M. 233 (1962). Appl. 221.312 Combustion Instabilities 216. Markstein. Appl. S . I. Melvin. CST 29. G. J. G. 219. 224. Phys. (1949). B. 40. 93 (1980). 230. Matkowsky and G. Williams. G. I. Math. CST 14. SZAM J. B. 227. Dongworth and A. Emmons. H. Maxworthy. G. G. L. Mitani and F. L. P. 162167. 75 (1983). (1981). 61 (1981). Math. J. I. Olagunju. Appl. Sivashinsky. 155 (1982). 226. R. Sivashinsky. Sivashinsky and B. 177 (1976). Sivashinsky. Matkowsky. SZAM J. 222. R. 37. 229. 207 (1982). CST31. SZAM J. Fluids 4. J. Math. J. Law. Matkowsky. 3rd Symp. E. H. Appl. Sivashinsky. Frankel and G. B. Markstein. 231. Architlurn Cornbustionis 1.
in a sense. Moreover.CHAPTER 10 Theory of Turbulent Flames Although Mallard and le Chatelier [11 recognized that turbulence affected burning velocities. In recent years remarkable progress has been made in the theory of turbulent flames. [12]. Accounts of a number of these early turbulentflame studies have appeared in textbooks [4][6]. it is now entirely appropriate to discuss theories of turbulent combustion here. This statement is no longer true. the theoretical understanding of premixed turbulent flames contributed by subsequent works prior to 1965 was not much greater than that originally developed by Damkohler. Therefore. In the first edition of this book the comment was made that. [27] is a 250page book devoted to the subject. the subject is so broad that we shall be able to cover only a portion of it and must refer to recent reviews [13][40] for more extensive presentations. and a number of new approaches to the subject are available for both premixed and nonpremixed systems. scientific investigations of turbulent flames began in 1940 with Damkohler’s classical theoretical and experimental study [2] of premixed flames. Damkohler’s work was soon extended by Shchelkin [3] and others. the topic of premixed turbulent flames did not belong in a book on combustion theory because the theoretical basis of the subject was practically undeveloped. [6]. and reviews of much of the early work on this subject as well are available [4]. In fact. 313 . a few of these approaches are sufficiently firmly based to enable analyses to be pursued with high confidence in limiting cases. However. Theoretical analyses of turbulent diffusion flames began in the same decade [7][ll]. In particular.
representing the joint probability that a molecule is in one volume element of phase space. Certain studies in kinetic theory in which this has been accomplished [41]. NavierStokes turbulence.1. v./dt. a more fundamental level of description may be needed to address certain phenomena. such as the homogeneous thermal explosion of turbulent hydrogenoxygen mixtures. appears in an integral expression for df. There are a number of excellent background books on NavierStokes turbulence in nonreactive fluids [46][52]. equation (D2) is not closed because a twoparticle distribution function. are deterministic in that they are believed to possess a unique solution. The Boltzmann equation must be supplemented by an equation for the twoparticle distribution in the absence of molecular chaos. either theoretically or experimentally. occurs when the inertial forces are sufficiently large in comparison with the viscous forcesthat is.3 74 Theory of Turbulent Flames The most fundamental question in turbulent combustion concerns the applicability of a continuum description. characterized by fluctuations of velocities and state variables in space and time. it cannot be claimed that theories of turbulent combustion should necessarily be developed from the NavierStokes equations. combined with a proper set of initial and boundary conditions. Turbulence develops from fluidmechanical instabilities of flows [53] when Re is large enough to prevent viscous damping from suppressing fluctuations. Without molecular chaos. Therefore. However. . the most convincing derivation of this result for gases is based on kinetic theory (Appendixes D and E). Molecular chaos is a key hypothesis underlying the derivation in that it enables the collision contribution dfi/dt in equation (D2) to be expressed in terms of integrals involving only the oneparticle velocity distribution function fi(x. However. at high Re. in which these kinetictheory effects have been found to be significant. at sufficiently large values of a representative Reynolds number Re (see Section 3. It then becomes reasonable to introduce statistical methods and to seek only probabilistic aspects of the solutions. and in reality the conditions cannot be specified accurately enough to obtain the deterministic solution. N . in reducing explosion times appreciably [43][45]. the NavierStokes equations augmented by chemicalkinetic laws) govern turbulent combustion. that is. [42] through introduction of a weakened chaos hypothesis have revealed longrange correlations that persist on macroscopic time scales and thereby modify the full NavierStokes equations. . The NavierStokes equations. to treat turbulence as a random process. t ) . . while at the same time another molecule is in another volume element. for example. There are particular problems in turbulent combustion. a NavierStokes description appears to suffice for most phenomena. the solution is strongly sensitive throughout most of the flow field to the initial and boundary conditions. It is conventional to assume that the full NavierStokes equations (that is. i = 1. . . especially the chemical reactionrate expression. The following presentation is based entirely on this continuum view.7).
Conceptually.1.* * Although it is essential conceptually to make a distinction between a stochastic function and a particular function v ( x . such as a component of velocity. Studies of strange attractors are just beginning and have not yet found specific applications in turbulent combustion. small changes in boundary or initial conditions lead to large and irregular changes in the solution away from the boundaries. turbulence possesses a “strange attractor” that draws its statistics away from input statistics and toward statistics that are characteristic of the flow system [54]. Although the statistical properties of the flow might be expected to depend on those assigned to the initial and boundary conditions. which we might call u.< . we may consider the initial and boundary conditions to be given only in a probabilistic sense. therefore. temperature. is a function of the continuous variables of space x and time t and is. and thus it is convenient to treat them separately. After presenting some background information on theories . Symbols like v always will refer to arguments of probabilities unless there is an indication to the contrary.1. P[u(x.defined by stating that the probability of finding the function in a small range 6u(x. In turbulence. [ S S ] ) .t).1.2) and of turbulent premixed flames (Section 10.1. PROBABILISTIC DESCRIPTIONS 10. a stochastic function. In this sense. for example. t ) is P[u(x. the solution in most of the flow would be pseudorandom in that it would pass tests for randomness.of turbulent combustion in the following section. It seems clear that even if a totally deterministic problem were specified. t)]du(x. t ) that appears as the argument of we choose not to do so notationally for the sake of simplicity. any stochastic variable. t ) about a particular function u(x. The solution to the deterministic problem is said to exhibit chaos in that.3). Although there are situations in which turbulence can tend to wash out the distinction between premixed and nonpremixed systems. Probabilistic Descriptions 315 Basics of the probabilistic description of turbulence will be presented in the following section. we shall discuss theoretical descriptions of turbulent diffusion flames (Section 10. the best approaches to analyses of these two limits remain distinct today. . t ) ] .10. Probabilitydensity functionals With turbulent combustion viewed as a random (or stochastic) process. this input information is modified so greatly by the many fluid instabilities that the statistical aspects of the turbulence are found to be virtually independent of those of initial and boundary fluctuations. mathematical bases are available for addressing the subject. 10. or the concentration of a chemical species. A complete statistical description of a stochastic function would be provided by a probabilitydensity functional. A number of textbooks provide introductions to stochastic processes (for example.
One approach to obtaining an equation of evolution for P(v) is to introduce the ensemble average of a finegrained density. t ) = 1.* although some interesting results have been obtained in this way for relatively simple problems (for example. This difficulty is the infamous “closure” problem of turbulence. they are of little help because useful methods of solution are navailable. from which the statistical properties can be calculated in accordance with a frequency interpretation of the stochastic process. To make progress in analyzing turbulent reacting flows. The probabilitydensity function for v may be denoted by P(o). and from the fundamental definition of measure or probability. &(x. attention is restricted entirely to particular. then S[v(x. for example. For incompressible NavierStokes turbulence. Although this fact and the ability to write the functional differential equation symbolically are comforting. t)] satisfies a functional differential equation that is linear and that in principle contains all the statistical information concerning the velocity field. If. and its statistics are described by a probabilitydensity function. . which has the same properties. we must seek a description at a lessdetailed level than that of the probabilitydensity functional. . . t)]dv(x. t ) is a volume element in the function space. For reacting flows. An avenue that has received exploration is the development of equations for evolution of probabilitydensity functions. where { . which exhibits itself in many different forms. { effectively involves a continuously infinite number of realvariable integrations extending over the entire function space. . this would be done either once for a sufficiently long period of time or many times. density. another is formally to perform suitable integrations in * Depending on the problem. temperature. Seeking lessdetailed levels of description invariably leads to formulations for which. there are more unknowns than equations. . then the variable whose value may be represented by v becomes a random variable instead of a random function. since any simplification must delete some potentially relevant information contained in the probabilitydensity functional. the current capabilities of electronic computers in general are much too limited to enable this approach to be pursued. in one sense or another. it can be shown [56] from the NavierStokes equations that if v represents the velocity vector. for P[v(x. as described by O’Brien in [27]. there is a functional differential equation. depending on the approach adopted. and JYrn P(v) d o = 1. fixed values of x and t . J Y[o(x.376 Theory of Turbulent Flames Here. for example. . One line of research is to attempt to solve the timedependent NavierStokes equations for the reacting flow numerically. so that an ensemble of solutions is obtained. t ) ] if v is interpreted as a vector in N + 4 dimensions whose components represent values of velocity. and all chemical compositions in an Ncomponent mixture [27]. By definition P(v) 2 0. where P(v) dv is the probability that the random variable lies in the range dv about the value v. [57] and [SS]) by working with simplified equations such as the Euler equations.
Even if the difficulties in the chemical source terms are overcome. the sign of the diffusivity changes somewhere.378 Theory of Turbulent Flames forms of these are called Reynolds stresses and typically are assumed to be proportional to gradients of average velocities. these methods are illsuited for many turbulent combustion problems. As a consequence of this. depending on the units. In use of moment methods in reacting flows. Since the computational difficulties in use of moment methods tend to be less severe than those for many other techniques (for example.2). [32]. the chemistry can introduce severe difficulties in moment methods for reacting flows [17].+ Because of these difficulties with moment methods for reacting flows. When both gradient and countergradient diffusion occur in different regions of the flow. they currently are being applied to turbulent combustion in relatively complex geometrical configurations [22]. with constants of proportionality called turbulent viscosities or turbulent diffusivities. . [40]. diffusion. real phenomenon for certain turbulent reacting flows [59][63]. [28].* Countergradient diffusion now is a wellestablished. and then are transported back. and these may correspondingly be modeled as proportional to the gradient of the average temperature or of the average species concentration. the transportflux and gradient vectors have opposite directions. Reynolds transport. Since it cannot be handled well by moment methods. A number of reviews are available [22]. the chemistry interacts with the transport modeling through processes in which fluid elements are transported by turbulence to a location. techniques involving evolution equations for probabilitydensity functions). now for energy or for chemical species. in a rough physical sense. the convective terms in equations for energy and species conservation produce averages of the product of a velocity component with temperature or a species concentration. Similarly. Many of the aspects of moment methods play important roles in other approaches. problems that stem from the chemistry remain in modeling the Reynolds transport [28]. Related problems can arise in the modeling needed for methods based on evolution of probabilitydensity functions. or gradient. [32]. [25]. react there. as they are needed. for example in the vicinity of equation (38). notably in those for turbulent diffusion flames (Section 10. with the coefficients again being turbulent diffusivities. There are classes of turbulent combustion problems for which moment methods are reasonably well justified [40]. they have the same direction. This sketchy description will become defined better later. * In normal. we shall not present them here. modeling is needed for the chemical source terms as well (in contrast to methods based on equations for evolution of probabilitydensity functions) because of their nonlinearity. A consequence of this interaction is the occurrence of countergradient diffusion. [27]. We shall develop those aspects later. in countergradient diffusion. [31]. the signs of the turbulent diffusivities must change within the flow if a momentmethod formulation is adopted.
Ensemble averages are relatively well adapted to application in reciprocatingengine combustion. definitions of averages per se require further discussion here because they arise in one way or another in all approaches. The statistics of streamwise and transverse velocity fluctuations must be the same in isotropic turbulence. called homogeneous. Turbulent combustion in a wellstirred reactor may be approximated as if it possessed homogeneity. There are other flows. ergodic theorems establish the equivalence of ensemble and time averages under suitable conditions of statistical stationarity of the stochastic process. we may consider a process to be stationary in a strict sense if all of its statistical properties are independent of the selection of the origin of time. and there are a number of different ways to define averages. Homogeneity must not be confused with isotropy. averages may be viewed as time averages. the collection of reproductions is the ensemble. there are corresponding simplifications in its statistical description. Although there are different definitions of stationarity. isotropy occurs only approximately even without combustion. * There are difficulties here associated with cycletocycle variation [64]. in which ensemble averages may be viewed instead as space averages. for example. The experimental evaluation of some averages that depend on coordinate directions can be simplified somewhat by use of isotropy if it applies. For example. a turbulent flame positioned normal to the average flow.* There are conditions under which ensemble averages may be replaced by space or time averages that may be obtained more simply. and ensemble averages are equivalent to averages over either of the coordinates with respect to which the process is homogeneous. the turbulent flow through a continuous combustor such as a gas turbine or a ramjet appears to be stationary on physical grounds.10. The stochastic process is imagined to be reproduced many times. where each cycle may be viewed as a separate member of the ensemble. and the ensemble average is the sum of the values at the x and t of interest in each member of the ensemble. which is considered to be different from turbulence. In these flows.1. divided by the total number of members of the ensemble. and a flame can be a source for generation of further departures from isotropy. and simplifications in statistical descriptions arise because of the stationarity. the most fundamental average is an ensemble average. Probabilistic Descriptions 319 Nevertheless. During steady operation. downstream from a uniform grid in a suitably designed combustion tunnel. There are turbulent flows in which homogeneity applies only in particular coordinate directions. which refers to invariance of statistical properties under rotation of the coordinate system. . In statistical methods. Homogeneous turbulence [47] is turbulence whose statistical properties are independent of the selection of the origin of x. for example. is nearly homogeneous in the two transverse directions parallel to the grid. in that the statistical properties are invariant under translations in these directions.
The ensemble average of equation (11) (with overbars signifying averages) then is appt +v( p+ ~ pIvI) = 0. Probabilitydensity functions appropriate for use in connection . equation (1) becomes appt + v ( p ~= ) 0. an appreciably greater compactness is obtained in the inertial terms . The economies associated with the introduction of Favre averaging therefore appear to constitute the determining factor in favor of its adoption.380 Theory of Turbulent Flames Irrespective of whether space. which have p' = 0 and 7 = 0 by definition. but in many uses of averaged equations. With p and pv as variables.pTT is found to be pv"v" = p(v"v"). but in the average of the equation for momentum conservation. The simplifications that arise in convective and inertial terms are offset somewhat by complications produced in molecular transport terms. The desirability of working with Favre averages is a topic of current controversy. In this notation. The controversy cannot be resolved on the basis of ease of measurement because some instruments provide v" most directly. the divergence term is then simply V * (pV). modeling is introduced for effects of molecular transport irrespective of the averaging selected. It is unclear whether the successful constantdensity modeling of Reynolds stresses is extendable more readily to pv" or to pv"v". A formulation in which this modification was effected has been developed by Favre [65]. The average of a product of fluctuations appears in equation (1). instantaneous fluctuations are p' = p . Therefore. for which averages are to be calculated) in analyses of turbulent combustion.i. as they must be for completeness. or ensemble averages are employed. (1) where the local. In thorough studies there is unlikely to be a fundamental reason for making a choice because the identity pv" = pV + (@)demonstrates that one average is recoverable from the other once p and p'v' are obtained. The simplification of equation (1) represented by equation (2) is relatively unimpressive. notable economies generally are achieved through introduction of these massweighted averages. while others provide C. questions arise as to what variables are to be selected for use (for example.. A simplification can be achieved in variabledensity flow by selecting the product pv as a fundamental variable to be employed instead of v. b u t 3 = 0 by definition. and we may therefore choose to work with the averages p and V. Then the average v" is not zero.V.p and v' = v . time. the ri dyadic @ . Favre defines the massweighted average of a variable v as i. a product of fluctuations does not arise in the average of the continuity equation. which involves four terms if the more conventional average is employed. Considering equation (11) we may guess that p and v are reasonable variables. as v" = v . = (p)/p and denotes the departure of v from i.
and conventionally is denoted by UI2 = 0 2 == J m (U  I$~P(u) dv = v2  ij2. Properties of probabilitydensity functions The basic definition of a probabilitydensity function is given in Section 10. with positive s implying a more gradual tailing off toward positive values of o than toward negative values. with larger values of K corresponding to a shape more like a sharp. P(v) cannot be reconstructed completely even from a knowledge of the values of all of its moments because this is possible in principle only if P(v) obeys . measures the width of the distribution of P(v). 10. . and K is a measure of how kinky P(v) is.1. here 0 represents the rootmeansquart: intensity of the fluctuations about the mean (that is. rectangular mesa than like a gently curved hill. respectively. as indicated at the end of the following section.1.0.(. Probabilistic Descriptions 381 with massweighted averages have been defined by Bilger (see [27]). The variance is the second moment about the mean.s. in general. The third and fourth central moments may be written as sc3 = J m (U m  ij)3P(~ dv ) (6) and KO4 = J. and K would provide a significant amount of information about P(v).3.1. Knowledge of the values of 6. However. is the second moment about zero.ij)4P(v) dv. average) value. From P(v). (5) m where the last equality follows directly from the definitions by expanding (v Either the variance or its positive root. also termed the second central moment. the average of the random variable is readily calculated as This is also the first moment of P(u) about zero because the nth moment about zero is defined as Jm The meansquare value.for example.1. u ' . 0.the standard deviation. W (7) where s and K are called the skewness and kurtosis (or flatness factor). s is a measure of asymmetry of P(v).10.
If g(v) possesses a powerseries expansion that converges wherever P(v) # 0. . a probabilitydensity function for a new random variable with its value represented by g . and the equation becomes Q(g)ldgl = P(v)Idvl. . .382 Theory of Turbulent Flames various continuity and convergence requirements that many probabilitydensity functions of practical interest do not satisfy. equation (9) extends this result to nonmonotonic functions by adding the contributions from the members of the ensemble having the original random variable in the range dg/ I dg/dv I about each vi for which g(vi) = g. . n) are the values of v for which g(v) takes on the value of g. where dg = Idg/dvI d v . Equation (9) may be understood best by first considering functions g(v) that are monotonic. then neither g nor Q(g) can be obtained even in principle from P(v) because it lacks information concerning the relative frequencies with which each of the different possible values of g occurs in the ensemble . so that there is only one value of v for any given g (that is. and therefore in the absence of additional information. g = go for v1 < v < v. is which may be evaluated directly by integration. Considerations like those indicated here enable Q(g) to be generated for any singlevalued function g(v). stating that all members of the ensemble for which the original random variable lies in the range d v about v also have the new random variable lying in the range dg about g . the mixture fraction usually is a doublevalued function of temperature (see Section 3. Given g(v) and P(v). say g(v).2). The average of any singlevalued function of v.4. delta functions often arise in probabilitydensity functions for temperature and species concentrations. then g may be evaluated in terms of the moments of P(v).then equation (9) simply produces infinity at g = g o . but through consideration of the fraction of members of the ensemble for which g = g o .in equilibrium diffusion flames. it becomes clear that in the vicinity of g = g o . . If g remains constant over a finite range of vfor example. a probabilitydensity function for it cannot be generated from one for temperature (which fails to distinguish rich from lean mixtures). n = l). there is a deltafunction contribution to Q given by in turbulent combustion. say Q(g). If g(v) is multiplevalued. may be constructed according to the formula where vi ( i = 1.
) is a twopoint. . v. Often it is desirable to consider not all members of the ensemble.. .. as needed in evaluating Reynolds stresses and Reynolds transports. v. restricting attention to these members is establishing a .. To increase the amount of statistical information. for example. Here P(v. twotime would be applied to P(v. In practice it is seldom convenient to consider n > 3.. .) d v . about v. then a relevant joint probabilitydensity function (for two random variables) may be denoted by P(u. while if t . if different. ... If v1 and v.. v. = t . It is always possible to recover P(vl) from P(v. .10.). . and v. joint probabilitydensity functions (now onepoint. twotime joint probabilitydensity function of one stochastic function..1.) 2 0 and .) dv. # t..). v. may be values of two different components of velocity or of one velocity component and temperature. the second) takes on a specified value or lies in a specified range. then the adjectives onepoint.). The definition requirements P(v. about v. the terminology twopoint. joint probabilitydensity functions may be defined for any number of random variables. . t 2 ) . In a similar manner. here v1 and v. . and allowing n to approach infinity suitably in principle could result in a complete statistical description of a stochastic process.. then the further qualifier auto often is used. but only those in which one random variable (for instance. . . m: [ = 1 (10) always apply. Even for x1 = x2 and t . v. Twopoint probabilitydensity functions play important roles in describing stochastic processes whose underlying conservation equations involve derivatives with respect to x. If x1 = x. v. t. Thus P(v. 0. . . . v. . [ymP(vl. If v1 is a value of a stochastic function at a particular position and time (x... but x1 # x. d v . is a value of the same stochastic function at a different position and time (x. v. dv. = t..) because from the underlying definitions.) may be introduced.. joint probabilitydensity functions may be introduced. . is defined as the probability that the first random variable lies in the range dv. cross. v. and the second in the range dv.. where P(v.viewed as the fraction of members of the ensemble in which both of the random variables are in their respective ranges. . onetime would be used.. relate to the same stochastic function.. for example. onetime) are of interest in turbulent combustion . In equation (11) P(vl) would be called a marginal probabilitydensity function. v. Probabilistic Descriptions 383 Knowledge of P(v) for one random variable at fixed x and t offers very limited information concerning a stochastic process because no aspects of relationships obeyed by the random variable at adjacent values of x and t are contained therein. but t . having been generated on a margin (the v l axis) from the projection of a joint function. dv.
the value of the temperature at a point in a turbulent flame may be expected to be related statistically to the value of the fuel concentration there.that is. ) P ( U 2 ) . In turbulent diffusion flames. These flows are said to exhibit intermittency of the turbulence . then the notation P(u. Conditioned averages and conditioned moments can be defined in a straightforward manner since conditioned probabilitydensity functions obey all the properties of unconditioned probabilitydensity functions. and the conditioning may be that rotational fluid be present.. I u2)P(u2)= P(u2 I ul)P(u. From equation (12) it then follows that P(Ul. In this case P(u.U2) = P ( U l ) P ( U 2 ) (14) for statistically independent random variables. and the value of the . P(u. (13) where Po(u) is the probabilitydensity function in the irrotational fluid and Pl(u) is that for the rotational fluid. Thus P(ul I u 2 ) du. 1 u 2 ) = P(u. If the conditioning is that the second random variable have the value u 2 . We may decompose a probabilitydensity function as P(u) = (1 . and conditioning on I = 1 may be desirable for a number of purposes. Conditioning often is needed in turbulent combustion. t ) = 1 in rotational flow. is Bayes's theorem. The meanings of the various probabilities show that P(U.384 Theory of Turbulent Flames condition on the probability.lu2) is independent of u2 for all u. t ) = 0 in irrotational flow and Z(x. the meaning of which should be clear in spite of the somewhat imprecise notation. is the probability that the first random variable lies in the range du. often there is interest in flows (for example. about u1 while the second takes on the value u 2 .in which all four P's are different functions. It is conventional to place a vertical bar inside the parentheses in the notation for P and to write the conditioning restriction after the bar. 1 u 2 ) is used for the conditioned probabilitydensity function. For example. 02) = P(U1 U . the conditioned probabilitydensity function is the same as the marginal probabilitydensity function. A random variable with a value denoted by u1 is said to be statistically independent from a random variable with a value denoted by u2 if P(u. For example. I (12) A result of the definitions. an intermittency function may be defined as a stochastic function having a value Z(x.I)P. jets and wakes) that are partly turbulent and partly nonturbulent (in that the flow is partly irrotational).(u) + IP.)that is. if the conditioning does not affect the probabilitydensity function.). Statistical independence is a strong property that is difficult to justify.(u). conditioning on the requirement that fuel be present is another example of conditioning that sometimes can be useful. and the resulting probabilitydensity functions are called conditioned probabilityclensity functions.
. (17) the first of which is similar to equation (3) and the second of which is like equation (5). the second cross central moment.) dv. Fourier transforms provide the desired decompositions.fiZ)P(V. (15) A nondimensional measure of the covariance is the correlation coefficient. since corresponding averages are obtained from S(u) by the usual rules for averaging.1.4.)(v. 0) d p / P . it is helpful for various purposes to view the processes in frequency or wavenumber spaces instead of in time or physical space.fi. A massweighted (or Favre) probabilitydensity function for the second random variable may then be defined as p"(4 = By definition. then the random variables are said to be uncorrelated. In working with . Thus. The joint probabilitydensity function for the density of the gas and another random variable may be denoted by P(p.. for the ij and u" defined above equation (2).10. = 0. 10. = 0 if the two random variables are statistically independent. a m .1. defined in equation (5). . .. Use of equations (3) and (14) in equation (15) readily shows that R . v). Fourier decompositions In analyses of turbulent combustion. 2 represent meansquare fluctuations (variances). P(u) possesses the necessary nonnegative and normalization properties of probabilitydensity functions. . v. implied by statistical independence but not implying it. is vanishing of the covariance. Probabilistic Descriptions 385 velocity at a point will be related to the value of the velocity at a nearby point. JOm PP(P. where the for i = 1. If R . It is especially useful in connection with the Favreaveraged formulation of the conservation equations. A much weaker property. R. j. Transforms with respect to different variables arise. do. sometimes called the cross correlation. ( U. = j.
t l ) = J. we may transform these functions with respect to velocity components to obtain “characteristic functions” from which the probabilitydensity functions can be recovered [55]. f. In equation (20). J. .x m m m m R b . example.. thereby effecting simplifications in the analysis. t . and the transform variable k is the wavenumber vector. (22) Cross correlations such as between temperature T and a mass fraction Y. which assumes that R is defined for all real values of x and t (for example. the formalism of FourierStieltjes transforms may be introduced to allow for anticipated nonanalytic behavior [47].. The spacetime velocity correlation tensor is defined as Its components could be evaluated from equation (1 5 ) if suitable probabilitydensity functions were known. t l > = I.) and R(xl. Here we shall consider only transforms of certain averages. t . s_. and the latter if the process is statistically stationary. The former is especially useful for coordinates in which the stochastic process is homogeneous. namely. Particularly for velocity and state variables. The quantities transformed may be the velocity components or state variables themselves or various averages of them. or else this refinement can be bypassed by accepting illbehaved transforms (having delta functions in them).)Y.t. x2.w . the transform variable o is the frequency.I. Correlations are readily recovered from their transforms by the inversion formula. t. With x = x2 . . T’(x. R(x. we may define a fourdimensional Fourier transform as w. t . e i o r f ( z ) which would be applied four times for the spacetime transforms. t . . (20) where dx is a volume element in x space. For statistically stationary turbulence. For a scalar field such as the temperature T. corresponding quantities are the autocorrelation. t.) . correlations. . x. s_..x m m m m R(x..I. w .t.) dx dz.. x 2 . x.eireik. we havef(z) = j?. with v‘ = T‘ (the temperature fluctuation) and the transform W.) dx dz. eiwrcp(o) do/271. with cp(co) = j?... for dz.)may be treated similarly. t. Equations (20) and (22) in their full forms are seldom useful. XI.(x2. .namely. t.I. and z = t ..x . The main advantage of introducing transforms in these situations is that differentiation becomes multiplication by the transform variable.ireik.386 Theory of Turbulent Flames probabilitydensity functions for velocities. x. Transforms with respect to spatial coordinates or time are encountered more often.. by putting R = 0 for values outside the bounds of the flow of interest).
Half of the trace of Q(k. t. and their transforms with respect to z alone depend only on o and the two positions.t. t l ) = (2. the space correlation obtained from equation (19) by setting t .)47cr2 dr.. Ideas about the dynamics of turbulence have been developed by reference to fully homogeneous turbulence. = t. x1. only through the difference z = t . . A spectral decomposition for intensities of fluctuations of scalar fields may be obtained from equation (21) in a similar manner. is (k/27c)’ times the traceof Q ( k . t1)/(271)3] dk.J. . = t depends only on r = I x2 . (24) in this case. play central roles. Similarly. Spectral decompositions in frequency .)since. and t . t l ) = J0%. ) v’(xl 2  1 + x. namely. t l ) = ~omeikrR(xl. then the integrand in equation (25) or in equation (26) may be viewed as the contribution (per unit range of wave number) to the turbulent kinetic energy or to the fluctuation intensity from eddies of size l/k.x.I. is related to the average turbulent kinetic energy per unit masssay q(x. t.10.)/(27~)~ in equation (23). for which the threedimensional transforms with t . .. as an integral over wavenumber space. putting x = 0 in equation (23) gives q(x. t l ) dk.. and its spatial transform therefore depends only on k E Jkl. to be denoted here by E(k.)3 ~ I.). then transforms with respect to those coordinates may be useful. ) at time t . Probabilistic Descriptions 387 depend on t . For isotropic turbulence. (23) which is independent of x .) and becomes 47ck2E(k. 4(x. from equation (19). t.  d2(x1. ) and @(k.. for example. eik’XE(k. tl). denoted here by Q(k. equation (23) shows that the average kinetic energy per unit mass can be expressed as an integral over a scalar wave number. ) dk.) and @(w.t . t.x. for homogeneous turbulence. ) and called spectra of space correlations. t . are transthese functions for x forms (or spectra) of time correlations at a particular position and can be especially useful. denoted by @(a. x.1. that is. Q ( k .xl)... t .t . t . .. its inverse transform is  1 v‘(xl. = x .). (25) where the energy spectrurn. (26) If l/k is interpreted as a turbulent eddy dimension. t ... x 2 . t . ) = SU: [ 4 z k 2 @ ( k . . x. t. x. w m a .e(k.x 1I at fixed values of x 1 and t. if there is homogeneity in some coordinate directions..t.
Scales of turbulence The dependence of e(k. . have been verified reasonably well. This is an ambitious task that has been completed successfully for a number of turbulent flows.. Taylor hypotheses are much more questionable for many turbulent flames because of rapid changes of fluctuations with position. If this is acceptable.t l ) on k affords a convenient focus for discussing many properties of turbulent flows as well as various scales of turbulence.tl). An approximation of this type.. and e(k) can be obtained from a measurement of the time history of the velocity fluctuation at a single point. Here we shall write e(k). may be poorer for many turbulent reacting flows.t 2 ) for x2 = x 1 + V ( t 2 . 10. ) v ’ ( x .5. The classical experiment concerns the flow in a constantarea wind tunnel downstream from turbulenceproducing grids. produced by the influence of combustion on the turbulence through the effects of heat release. Integral scales (also called correlation lengths) may be defined explicitly in terms of correlation coefficients. Measurements of e(k) show that it vanishes at small and large values of k and reaches a maximum at an intermediate value. t . v’(x2. Equation (25) suggests that a continuum of scales of turbulence contributes to q.Ix2 . then v‘(xl.388 Theory of Turbulent Flames for turbulent kinetic energy and for intensities of scalar fluctuations may be introduced in an analogous way for statistically stationary turbulence. the simplifications afforded by speaking in terms of a scalar k offset the greater precision associated with retaining k as a vector. moreover. v’(xl. Although the approximation of isotropy is not exactly correct even for the simplest nonreacting turbulent flows in the laboratory and. For nonisotropic turbulence. which has been called a Taylor hypothesis. The measurement task can be simplified appreciably in this experiment if it can be assumed that the turbulence is convected with negligible change past the fixed measurement tl)= station at the constant average velocity V. xl.x 1 in the direction of 1.initially our discussion will mainly concern the dynamics at smaller scales. In view of equations (19) and (24). t l ) =v‘(x. tl)v’(xl. The maximum usually occurs where l / k is on the order of the integral scale of the turbulence. thereby derived. suppressing the dependence on x1 and t. and we shall occasionally appeal implicitly to results for homogeneous turbulence because the ideas have been developed most thoroughly in that case.xll/lVl). this involves measuring velocity fluctuations simultaneously at two different locations for the full range of distances between these locations. The general physical concepts of turbulent motion. so that with x2 . .1.t . Isotropy tends to be achieved more closely at small scales than at larger scales . The contribution from each scale can be obtained experimentally by measuring e(k). is reasonably good in many nonreacting flows for finding e(k) at all but the largest scales. we may consider our remarks to apply to an average over directions. Consider any one component .
. If a Taylor hypothesis applies. fi R . the largest concentration of turbulent kinetic energy occurs at scales in the vicinity of 1. respectively. for equation (19).. depends on the direction of the unit vector. The eddies with sizes in this range have been called the energycontaining eddies.. the integral scale becomes in which the integration is performed with the point x. . and on the component of equation (19) selected.t . For other purposes integral time scales also are introduced. e(k) falls rapidly toward zero with decreasing k . For values of k above 111. ) . the size of the apparatus imposes an upper limit on the maximum eddy size achievable (the minimum k having e # 0). in equation (15) to represent the selected components of v(x.. t . .10. there may be significant additional structure in the function e(k) if the turbulence Reynolds number. R .1. Thus the integral scale and the rootmeansquare velocity fluctuation are employed in defining the turbulence Reynolds number.and allow v . maintains a fixed direction. varying from x.1. ) may be defined as an average over the diagonal elements only and over all directions. = t . of equation (16).. = x.. and let u . ) and v(x. but with u and a there replaced by and 1.. a unique integral scale 1 at (x. on x 1 and t . For homogeneous.t . . . x . Then. Under more general circumstances. isotropic turbulence at fixed t . = P&l/ji. exhibit a correlation. and t. is like that of Re of Section 3... Since e(k) peaks near k = 111. toward x. and v .. l. then a relationship between integral space and time scales can be developed. one for diagonal elements of R and the other for offdiagonal elements. of equation (16) and with T = t . fixed and with x. Probabilistic Descriptions 389 of equation (19) for t . let r = Ix. then the integral time scale or correlation time would be rm with the c.xll. in equation (15) refer to the variables at times t . for example. Integral scales may be defined for scalar fluctuations in a similar manner by considering equation (21) instead of equation (19). is sufficiently large. respectively. In general. to a boundary of the fluid (indicated by a3 in the integral) in such a way that the unit vector from x.. For values of k below those corresponding to the energycontaining eddies. # x. there are only two different values of I. t . The definition of R . and v . respectively [where the q is that of equation (25)]. in terms of the c. From its definition we see that the integral scale 1 physically measures the distance over which the fluctuations occur in a relatively fixed phasethat is. (29) . if we were to consider equation (21) with x.7. . the question of selection of a component of a tensor then does not arise. .
(30) which shows explicitly that I. 1. This deduction agrees with experimental observations for R. there is an intermediate range of k between 111 (where turbulence is generated from meanflow instabilities) and 1/1.ii/p. 1. is considered to occur through interactions between components of different wave numbers. where turbulence is dissipated by viscosity). Related constant factors that also occur in other definitions with isotropy iv/z. The inertial subrange ends when k becomes large enough for viscosity to be important. % 1. A spectral transfer of energy. the viscous * Often the intensity in R.is defined in terms of just one component of velocity so that appears instead. e(k) decreases with increasing k in proportion to k . thereby causing spectral mixing. 1.390 Theory of Turbulent Flames Since R. v = . When R . we see that + + 1. 6 1. $ 1.5 / 3 . . Turbulence dies rapidly if R . From equation (29) and the definitions. The characteristic length at which this occurs. = l/Rf'". Both scale (1) and intensity* (4%) of turbulence must be known to characterize it properly. .5 such as equation (31) are not introduced herein. over which only the inertial terms in the equation for momentum conservation are significant. in the inertial subrange. the Kolmogoroa length. produced by the nonlinearity of the inertial terms. 1. is based on properties of the turbulence. The dependence of e(k) on k in this inertial subrange or Kolmogorov subrange was deduced by Kolmogorov through dimensional analysis. > 1. (defined below. Therefore. The average rate of spectral transfer must be the same as the average rate of dissipation of turbulent kinetic energy and dimensionally may be estimated as c = ( 2 ~ ) ~ " / 1 The . is obtained through dimensional analysis as the length that may be constructed from c and a representative kinematic viscosity. only quantity with the dimensions of e(k) obtainable from c and k is a dimensionless constant times c 2 / 3 k . which is a cascade of energy from small to large wave numbers.' I 3 . The interaction becomes evident if the convolution theorem is applied to the Fourier transform of the inertial terms in the equation for momentum conservation in homogeneous turbulence [47] . For k + l / l k . a condition often encountered in practice. < 1 when R. the peak of e(k) is followed with increasing k by a tailing off that is roughly symmetrical with the rise. namely. For R. = (v3/c)"". For R. only k and the average rate of spectral transfer of energy can affect e(k). During the cascade. measurement and physical reasoning agree in ascribing a distinctive shape to e(k). the contribution at any given wave number then involves an integral over all other wave numbers. The viscous terms are the most highly differentiated terms in the equation for momentum conservation and therefore are largest relative to other terms at the smallest scales (largest k's). it is associated more closely with regimes and dynamics of turbulent motion than is Re.
Nothing special occurs in the energy spectrum at k = l/l. < 1.1. or 1. = I/&. % 1. Here D is a representative coefficient of diffusion for the scalar field. $ 1. Experiments confirm that the cascade is terminated abruptly by viscous dissipation approximately at k = Illk. A third. By equating this quantity to E . constantdensity turbulence that e(k) becomes proportional to e2vk2t [47]. then this equation is found to contain terms that may be interpreted as convection by the mean flow. 1 and I. at any given time f. this represents an exponential decrease of e(k) with increasing k 2 . Time scales related to the various lengths may be defined through division by various velocities. production of turbulent kinetic energy.. arises in estimates of the magnitude of the average rate of viscous dissipation. and the equation becomes linear. represents a cutoff length below which no turbulent eddies exist. instead of 1 may be introduced if desired (and often are). The cascade concepts may be applied to intensities of scalar fields as well as to the turbulent kinetic energy. Passive scalars (those that do not influence the velocity field) in the absence of chemical reactions can experience spectral transfers as a consequence of the convective terms in their conservation equations. and an inertialconvective subrange can exist in which the integrand of equation (26) exhibits a powerlaw dependence on k analogous to that of e(k) [66]. I. a much more rapid dropoff than the powerlaw behavior in the inertial subrange.  (32) for the variable in equation (26). in a first approximation for R. Nonlinearities of the chemical production . j = 2D(VV') (VV'). We have now seen that for R. Comparison of equations (30) and (31) shows that for R . that is.. whence it is straightforward to demonstrate for homogeneous. it may be assumed that I . % 1. the Taylor scale. Probabilistic Descriptions 391 terms are much larger than the inertial terms in the equation for momentum conservation. The dissipation term is v times the meansquare value of the magnitude of the vorticity vector (V x v) and can be written as a nondimensional constant times v(2q)/L?. There have been studies of influences of chemical reactions on the spectra of passive scalars [161. there are at least two special length scales. and Reynolds numbers based on 1. is the Taylor length. we find through use of equation (29) that 1. < 1. Therefore. where I.10. multiplications and averaging in the equation for momentum conservation an (unclosed) equation for dq/dt is derived. I. is of a magnitude intermediate between that of I and I. If through suitable substitutions.. plays a role analogous to that of E in the analyses. and dissipation of turbulent kinetic energy. turbulent diffusion. Dimensional analysis again provides the power of k in the inertialconvective subrange. The average rate of scalar dissipation.
Phys. For example. [69]. it may be inferred that if a scalar is not conserved. and additional turbulence scales with distinguished attributes may be defined. nevertheless appeal to concepts of coherent structures [68]. H. Various time scales also are associated with the * If Prandtl or Schmidt numbers differ appreciably from unity. We shall not have space here to present these approaches and must refer instead to reviews [18].4. TURBULENT DIFFUSION FLAMES 10. the increase in width of a turbulent shear layer with increasing distance downstream may be associated largely with vortexpairing processes that create larger structures from smaller structures. although not deductively based on the NavierStokes equations. then a number of additional interesting aspects of cascade phenomena arise. The volume in which [29] appears is concerned with this subject. a process that must occur. then there are hazards associated with attempts to apply the usual cascade concepts to it.2). the formulation in the following section focuses on conserved scalars.2. for example. Much of the early research on turbulent flows [48] acknowledges scaleincrease phenomena. 10. Partly for this reason. They may be responsible for an increase in 1 with increasing time as the turbulence evolves. Objectives of Analyses We saw in the last section that disparate length scales and time scales exist for turbulent flows. The focus is placed on frequencyhalving processes instead of the frequencydoubling effects operative in the cascade. or mixture fractions.392 Theory of Turbulent Flames terms can modify cascade aspects of the spectra. we shall consider a promising approach to the theoretical analysis of turbulent diffusion flames. theories that we also shall forego discussing for the sake of brevity. Fluids 11. they do). a countercascade phenomenon that meshes well with the ideas of [54]. These effects are complementary (not contradictory) to the cascade dynamics because they occur mostly at scales on the order of 1 or larger. * Most of the modern research on turbulence dynamics is addressing aspects that differ from those of the cascade concepts. We do not discuss these topics but instead refer the reader to C. . 2305 (1968).1. [40]. Instead.2. of course. There are approaches to analyses of turbulent combustion that. [27]. This research emphasizes coherent structures that occur in turbulent flows [67]. if shear layers are to grow in width (as. A conserved scalar is one that possesses no chemical source term (see Section 3. These methods share some aspects in common with “age” theories of stirred reactors [19]. as in liquids or ionized gases. Gibson. Relationships are being identified to the onset of turbulence and to the development of spottiness of turbulence at large Reynolds numbers through vortexstretching mechanisms that tend to concentrate fine structures in localized regions of space.
We shall see that good methods are available for analyzing problems of this type. are of two distinct types.2. Some that will not nevertheless have been considered to some extent in the literature. that is.1) may develop. the limit in which the chemistry is rapid compared with the fluid mechanics and the limit in which it is slow. It seems clear that. . Methods for analyzing turbulent diffusion flames with fast chemistry have been developed recently [ls]. For example. These methods. and we shall focus our attention mainly on twostream problems. Some of the specific questions asked are: What is the flame length or flame height? What is the average local volumetric rate of heat release? What are the critical conditions for a turbulentjet diffusion flame to be lifted off the fuel duct? What are the radiant energy fluxes emitted by the flame? How complete is the combustion? What is the rate of production of oxides of nitrogen? What is the rate of production and liberation of smoke and of unburnt hydrocarbons? What are the properties of noise emission from the flame? A number of these questions will be addressed here. a number of different regimes of turbulent combustion may occur. Nevertheless. Turbulent Diffusion Flames 393 chemistry. Instead. The primary combustion reactions in ordinary turbulent diffusion flames encountered in the laboratory and in industry appear to lie closer to the fastchemistry limit.2).4. There are many reasons for interest in turbulent diffusion flames.