Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
THE ROLE OF CHEMISTRY IN FIRE PROBLEMS: GAS-PHASE COMBUSTION KINETICS RAYMOND F R I E D M A N Factory Mutual Research Corporation Introduction The hterature on combustion kinetics is extensive. A book entitled "Atomic and Free Radical Reactions,''^ pubhshed m 1954, contams over 2000 references A later, more selective book entitled "Chemical Kmetics of Gas Reactions,'"" con- tams 1478 references Another book, entitled "Chemistry of Combustion Reac- tions,'" contams 922 references A fourth book, of bmited scope, entitled "Chem- istry m Premixed Flames,"* contams 283 references. A rough check shows that not very many of the same references appear m two or more of these books, sug- gestmg that all four hsts of references are only samples of what is available. I do not propose to provide a guide to this vast hterature in this article, except to mention that the 16 volumes of the journal Combustion and Flame^ and the pub- hshed proceedmgs of the 13 mtemational symposia of the Combustion Institute' certainly must be perused by anyone mterested m the detailed progress m this area. What I propose to do is to review a very few sahent pomts and to provide com- mentary on how these may relate to fire problems. These are of course personal comments, based on my own limited experiences, and they must be recognized as such. The Most Baste Approach If a scientist with background m nuclear reactor physics or electromagnetic theory were assigned the problem of providmg a scientific description of fire, his approach would surely be to formulate the relevant basic differential equations, based on Newtonian mechamcs and classical thermodynamics, which he could do, and then to solve these equations, either analytically or numerically, which he could not do. He would encounter two major difficulties. First, real life fires on a scale larger than a few inches generally mvolve turbulent flow, with transport rates of species and energy up to an order of magmtude greater than for lammar flow under corresponding conditions We know no way to describe turbulence from the basic equations, except with ad hoc assumptions of limited utihty How- ever he might sidestep this difficulty by arguing that the turbulent behavior might be descnbable at least approximately (and highly precise description of fire is not really needed) by assuming a proportionahty of some sort between the lammar and the correspondmg turbulent behavior, perhaps with an empirically determmed proportionahty factor. This could give him the courage to tackle the laminar problem, which not only can be formulated and solved theoretically, at least with modem computers, but also can be studied expenmentally on a small scale m the laboratory. This leads him to the second major difficulty. For those fire problems 187
188 FIBE BESEABCH in which chemical kinetics plays a role (ignition, flammabihty limits, extmguish- ment, flame spread through premixed gas, transition to detonation, etc.), it is necessary to Imow the sequence of chemical reactions occumng which transform reactants to products (the mechanism) and the temperature-dependent rate of each step. In general, these are not known except for the very simplest fuels, such as hydrogen. If we could calculate the rate of each possible reaction, we could then decide what actual reactions will occur, smce we know from chemical thermodynamics the final state (the most stable state) which the system seeks and we know that it will go there by the easiest available route. The most basic approach would be to calculate the rates of all possible elementary reactions for all combmations of carbon-hydrogen-oxygen, say, once and for all, by the methods of quantum me- chanics. The pioneermg work of Schrodmger, de Broghe, Heisenberg, and Pauli (1925-1927) laid the foimdations of quantum mechamcs, and very soon thereafter (1927-1931), Heitler and London, Slater, Pauhng, et al, formulated ways of de- scnbmg molecular stmcture in terms of the nuclei and electrons and the quantum mechamcal relationships between them. The next step was to calculate the rates of chemical reactions m this way, m 1931 Eyrmg and Polanyi made the first pioneer- mg treatment of the reaction H -|- para-H2âKjrtho-Hj+H by quantum mechanical methods. However the treatment was semi-empincal m that it was necessary to assume an arbitrary constant ratio between coulombic bindmg energy and exchange (resonance) bmdmg energy for the diatomic mole- cules concemed, for all significant mteratonuc distances. Nonetheless, this was very excitmg, m that it suggested that some day all chemical kmetics could be revealed by appropriate solutions of the Schrodmger equation. A tremendous effort, for that tune, was mitiated to tackle this project and by 1942, the book "Theory of Rate Processes," by Glasstone, Laidler, and Eyrmg,' was published, containmg over 500 references. Even though the mitial decade of effort did not yield a final quantitative solution to the problem, the work led to fantastically improved msights mto the factors actually controllmg the rates of reactions. The mcreasmg availabihty of faster and faster computers throughout the 1950's and 1960's, as well as the general growth in scientific activity m this penod, led to contmumg onslaughts on this problem. The Journal of Chemical Physics over this penod contams hterally hundreds of pubhcations attemptmg to quantify chemical reactivity m terms of quantum mechamcs. Success has not yet occurred. When will it occur? I have confidence that, sooner or later, rf support contmues to be available for basic science, this problem will be unraveled, with benefits not only to fire research but to numerous other apphed fields. I have no idea if the time required is a decade or a century. However to significantly mcrease the prob- abihty of obtainmg useful information from quantum mechamcs on, say, fires of wood pyrolysis gases m the next decade, it would appear to be necessary to mcrease significantly the number of scientists now workmg on this problem I would judge that there are at least several hundred competent scientists m the world today workmg in this area, with free exchange of results, so an effort perhaps beyond the capabihty of the fire research commumty is needed to mfluence probabihty of a brealcthrough by this basic approach.
ABSTRACTS AND REVIEWS 189 Experiments to Produce a Compilation of Reaction Rates As an alternate to a general theory yieldmg reaction rates, one might system- atically measure thousands of reaction rates for unimolecular, bunolecular, and termolecular reactions of possible mterest m fire problems, over an appropriate range of temperatures, and tabulate these Mechamsms could then be worked out. The difficulty is that virtually all reactions of mterest mvolve free radicals and unstable species, and do not occur m isolation; i e at least one of the products of the reaction to be studied may react with one of the reactants, for example, to give other products, etc. Thus the measurement of gaseous reaction rates is very difficult and subject to uncertamties. Nevertheless, a variety of mgenious techmques have been devised to measure individual reaction rates. These mclude: (a) electrical discharge plus flow system' (b) shock tube^'i" (c) explosion limit measurements" (d) flash photolysis'^ (e) measurements m flat premixed flames^-'' (f) mteracting molecular beams." In all cases, the goal is to make the measurement m a very short tune (micro- seconds), before secondary reactions mterfere, which requires either high flow velocities or sudden mitiatmg events (shock or flash). The ultunate experimentâ two monoenergetic mtersecting molecular beams with a smgle collision per re- actant moleculeâhas not proven to have much utility because of the great ex- perimental difficulties of producmg the velocities and mtensities required. The other techniques have yielded quantitative mformation for qmte a number of reactions. The present situation is that we have information good to perhaps a factor of 2 for several dozen reactions occurnng m fires, other mformation good only to a somewhat lower accuracy, a factor of 4 to 10, for several hundred reac- tions, and there are hundreds, or perhaps thousands, of other gas reactions of probable importance m real fires of hydrocarbons, cellulosic fuels, plastics, and fabrics, with and without flame-retardants, on which we do not even know the reaction rate withm a factor of 10. Several compendia of measurements of rate constants exist. Perhaps the most carefully evaluated rate data is that issued m five bulletins by the High Tempera- ture Reaction Rate Data Centre, Leeds Umversity, England There are also compilations by Trotman-Dickmson et al'° on bunolecular reactions, and by Benson and O'Neal" on ununolecular reactions, and a new "Handbook of Kmetic Constants of Gaseous Reactions" by Kondratiev i« As an illustration of the present state of kinetic data, a recent Combustion Institute symposium contams a tabulation by Browne, White, and Smookler" of kinetic parameters of 56 bunolecular and termolecular reactions mvolvmg 14 species monatomic H , 0, N diatomic Hj , O2, Nj, OH, CO, NO tnatomic H2O, HO2, HCO, Oa, CO2
190 FIKE HESBARCH Also, in the same book, NewhalF provides a tabulation of 33 reactions involving the same species, with the exception of HCO, HO2, and O3, and the addition of N2O and NO2. In comparing the two hsts, one sees that values differ moderately on almost every reaction To compare ratios, one must note that most of the reactions are temperature-dependent, so that when different temperature de- pendencies are chosen for a given reaction by the two groups of authors, the plots of rate vs. temperature would m general cross at some temperature, yieldmg perfect agreement, while at other temperatures the values would diverge. (Com- bustion m air involves reactions over the range from about 600°K to 2400°K ) To illustrate the state of knowledge, one of the most important reaction is C 0 + 0 H - * C 0 2 - | - H . In this case, the two tabulations give rates with different temperature dependences (activation energies of 700 and 7700 calories), the values agreeing at around 1200°K, differmg by a factor of 1.8 at 1500°K and a factor of 3 3 at 2000°K. As another illustration, for the recombmation reaction H - l - O H + M ^ H j O + M , one tabulation treats the rate as mdependent of temperature while the other treats it as varymg mversely with the square root of absolute temperature. The values differ by a factor of 3 2 at 1000°K and a factor of 4 5 at 2000°K. (I selected these illustrations at random, rather than m a search for best or worst cases) When one considers reactions of hydrocarbons, hydrocarbon fragments, and partially oxidized hydrocarbons, the available data are far more mcomplete and mconsistent. To represent one extreme of opmion, I quote from a survey paper on reaction rates by G. von Elbe": "With respect to complex systems, such as hydrocarbons and oxygen, we may never attam a state of quantitative kmetic knowledge." Another survey on "fundamental and broadly appbcable combustion research" by BatteUe Memorial Institute speciahsts^* refers to hydrocarbon oxida- tion as "enormously complex," and concludes that the approach of rigorous chemi- cal kmetics to practical problems mvolvmg hydrocarbon combustion such as air pollution "does not appear to be currently feasible " The alternative of empirical and phenomenological descnptions, such as "global" reaction rate models, is suggested. Contmued efforts to improve our compilation of rate constants could be aimed in any or aU of the followmg du-ections. (1) Tabulation, evaluation, and assignment of limits of uncertainty to data already pubhshed. (2) Use of presently known techniques to check and refine values already meas- ured. (3) Use of presently known techmques to measure previously unstudied reactions. (4) Development of new or improved techmques that might yield better meas- urements. (5) Analysis of what rate information, to what precision, is most urgently needed for fire problems. To estimate the progress to be expected m the next decade, it may be helpful to consider the tune scale leadmg to the present state of capabihty of dealmg
ABSTBACTS AND REVIEWS 191 experimentally with fast gas reactions The onginal pioneering work demonstrating the cntical importance of radicals and cham reactions was done m the 1920's and early 1930's by Semenov, F . 0. Rice, Hmshelwood, Bonhoeffer, A Farkas, L . Farkas, Harteck, Lewis and von Elbe, M . Polanyi, and many others. Positive direct identification and semiquantitative measurement of radicals m flames was not accompbshed to any degree until the 1940's and 1950's, because of the need to refine detectmg techniques such as mass spectroscopy and absorption spec- troscopy. The shock tube, a powerful new tool for fast reaction kmetics, was first used by Carrmgton and Davidson m 1953 to study N2O4 decomposition. The begmmng of modem flame-denved kmetics comcides more or less with develop- ment of flat-flame burners around 1950 by Powhng and Egerton, and by Spaldmg. Fnstrom and Westenberg made important refinements m techniques of flame structure measurement by mass spectrometry m the 1950's Kmetic studies of atoms produced by electnc discharges were vigorously pursued in the 1950's and 1960's by Kaufman and others In 1967, Westenberg and de Haas successfully used electron spm resonance detection to study kmetics m a flow system. Thus, there has been an impressive continuous mcrease m capability to measure fast reactions such as occur m fire, over the past 50 years We are not yet at the pomt where we can accurately measure any desired reaction, but we appear to be con- vergmg on this capability Another decade of progress might brmg us to that pomt. In order to consider what good this would do m fire problems, let us consider specific problems Obviously, the number of different fire problems which nught be of practical importance and might mvolve chemical kinetics is enormous. However we will restrict our remarks to four rather broad areas, namely igmtion, propagation, suppression, and toxic or smoky products. Ignition and Flammabthty The minimum energy which will just be able to ignite a gas mixture is in prmciple calculable if the chemical kmetics are known For common combustibles (other than hydrogen) the kinetics are not sufficiently known and such calculations have not been successful, as far as I know However, a well developed experimental technique exists for measurmg the cntical value of capacitance spark energy for ignition,^' and we know that the energy needed is less than a miUijoule for most combustible gas mixtures From a practical viewpomt, we know that a malfimc- tionmg household electric circmt, or a common match, is capable of supplymg many orders of magnitude more energy than this, so more understandmg of the critical energy requirement for igmtion would appear to be of httle value m re- ducmg fire loss Another aspect of igmtion is the ability of a substance undergoing slow oxidation to self-heat and spontaneously igmte. Famihar examples are oily rags, stored edible grams, sawdust, powdered coal, etc Carbon disulfide vapor m air can igmte by contact with metaUic surfaces no hotter than 125°C (steam pipes at moderately elevated pressures). This type of phenomenon is well understood m prmciple, as long as the chemical kmetics are known and the geometry is simple. A thorough recent review of the principles of "thermal igmtion" (cham reactions, diffusion, and hydrodynamic factors ignored) is available,^' with 70 references.'^The theory cannot of course be apphed to any of the real cases of spontaneous igmtion just mentioned, because the kmetics are not known I t may even be mentioned that m many of
192 FIBB BESEABCH these cases of slow oxidation of organic solids at low temperature, biological oxida- tions are known to be involved, and sterilization will prevent ignition. In spite of the inabihty to treat real cases quantitatively, the availabihty of the theory certainly broadens our conceptual understanding of the phenomenon. The really important practical aspect of ignition is that some sohds or hquids are much harder to igmte with a pilot flame than others, and some are essentially nomgmtable Understandmg of the reasons for these differences would be of great value m achievmg wider use of fire-safe materials. For hqmds, the concept of flash pomt temperature and its relation to vapor pressure and lower flammabihty hmit IS enormously useful, even though lower flammabihty hmit has to be expenmentally measured and is not theoretically predictable (not even for hydrogen, because of flame cell formation with very lean mixtures). Of course, even hqmds below their flash pomts can bum if sufl&ciently preheated or if dispersed as a foam or mist. For sohds, there is no concept like flash pomt, and igmtion temperatures of sohds, which can be measured by vanous empincal tests, give results which are highly test-dependent; i e., relative rankmgs of two matenals may reverse from one test to another I t must be presumed that gas phase kmetics as well as condensed-phase and interface kmetics play roles m sohd igmtion, but practically nothmg is Imown about this. Either accidental rmpunties or dehberately introduced fire-retardmg chemicals can have a large effect on ease of igmtion of sohds, which clearly involves chemical kmetics, but to what degree m the gas phase and m the condensed phase IS usually not known. In some few cases, indirect evidence strongly suggests that the flame retardant gasifies and inhibits the gas reactions.^* A "flammabihty" test of a sohd may measure its abihty to igmte or its abihty to contmue burnmg after the igmtion source is withdrawn If it measures the latter, we may have a semantic problem as to whether we are studymg ignition, but we are certamly studymg something of great practical importance. Fenimore and Martm'^ devised a test for a sohd m which it is igmted in atmospheres with varymg O2/N2 ratio and the cntical oxygen mdex (O2/O2+N2) is detenmned for which combustion can contmue. This test, permittmg quantitative companson of a wide range of matenals, is now widely used, and clearly is strongly mfluenced by chemical kmetic parameters, since the effect of the mtrogen must be to lower the flame temperature and hence the gaseous reaction rate. No quantitative mterpre- tation m chemical kmetic terms has been made of such test results as yet, to my knowledge. I beheve that the most useful thmg that chemical kmeticists could do at this time m regard to igmtion and flammabihty is to devise exploratory experiments to reveal the role of gas kmetics m igmtion and burnmg of cellulosic and plastic sohds, including man-made fibers. For example, we could devise an apparatus for heatmg solids so as to generate pyrolysis gases at a controlled rate, detemune the compo- sition, and then feed these gases mto a laboratory flame, mtroducmg additives either as vapors or with the heated sohd. The resultmg flame structure and stability could be studied. Once we have an idea what reactions might be important, separate detailed study of the rates of these reactions could be justified. The other thmg that is needed is further development of theory capable of accountmg for the heat and mass transfer process occumng along with gas-phase and condensed-phase chemistry. Here we might note that, for a related problem, the burmng of a solid propellant, elaborate theones for both igmtion and steady burnmg have been developed over the past two decades, but controversy in regard to mechanisms
ABSTRACTS AND REVIEWS 193 still exists because sufficiently accurate experimental measurements of conditions at the igmtmg surface have not been achieved, and separate decomposition studies of the individual mgredients (e g , ammomum perchlorate) have revealed tre- mendous complexity of behavior Smce fires are characterized by lower tempera- tures, less steep gradients, and slower rates than sohd propellants, we may hope that experimental probing of igmtmg or bummg wood or plastics will be more readily handled The problem is not easy. Any help that can be gotten from kmetic understandmg m developmg better test procedures for evaluatmg igmtabihty or flammabihty of soUds would be enormously valuable. PTopagaiion Combustion may be divided mto premixed flames and diffusion flames. While fijes are usually diffusion flames, the great majonty of combustion kmetics research has dealt with premixed flames because these are simpler to deal with both experi- mentally and theoretically (a one-dimensional model is possible). Inasmuch as peak temperatures are similar for the two types of flames, it is hkely that many of the same reactions occur, so that one can hope that knowledge obtained from premixed flames may have some relevance to fires. The theory of lammar propagation of flame through a premixed gas was ap- parently first formulated m terms of reaction kinetics m 1915 by Nusselt (cf. Ref. 26), who derived a simple formula showmg the flame speed to be proportional to the square root of the product of the thermal conductivity of the gas mixture and the mean rate of heat release (or overall exothermic chemical reaction) per unit volume of flame Thus, even in 1915, if we knew the kmetics of a gaseous flame we could have made an a prion prediction of flame speed This theory has been highly refined and elaborated to mclude diffusional effects, multiple reactions, temperature-dependent rate constants, etc, prmcipally m the period 1948-1958, most thoroughly by Hirschfelder and co-workers, and may be found m standard texts."'^' The application of this theory to actual flames has been very limited because of the lack of suflScient chemical kmetic knowledge for any real flame with oxygen as oxidant except the hydrogen flame. Another limitation has been the difficulty of apphcation of theory, which requu-es a high-speed electronic computer, to complex reaction mechanisms mvolvmg branchmg cham reactions Some of the best work on this problem has been done recently by Dixon-Lewis,'' who has solved the time-dependent equations for a nch H2-O2-N2 flame to give the steady-state concentration profiles and burnmg velocity. Much has been learned about premixed flame propagation by refined experunents to measure flame structure, especially of expanded low pressure flames, by fine thermocouples, samphng with microprobes, and emission and absorption spec- troscopy."'We presently have a reasonably good concept of the kmds of chemical processes occurrmg m a premixed hydrocarbon flame, and theu- probable effects on the propagation rate, even though we lack a quantitative theory. Let us consider now how, if at all, this limited understandmg of premixed flame chemistry relates to fire propagation problems. First, consider flame spread through a cloud of flammable vapor, mist, or dust which has partially mixed with air pnor to combustion. If this occurs mdoors, we are primarily mterested in the ventmg area necessary to reheve the pressure bmld-up before the buildmg fails. This depends of course on the flame propagation rate,
194 FERE RESEAUCH which depends on the fundamental laminar bummg velocity of the mixture existing at any pomt (which is relevant to the foregomg research), but also on the turbulence present, and particularly on the additional turbulence generated by the event itself as it occurs, mfluenced strongly by any turbulence-generating objects m the path of the flame. Fortunately, there is evidence showmg a relationship between speed of turbulent flame propagation for a given aerodynamic environment and fundamental lammar bummg velocity of the combustible mixture which is present. Thus, the fundamental bummg velocity is relevant, although far from the whole answer. Of course, lammar bummg velocities can be measured directly relatively simply, if 10% uncertamty is acceptable, and such measurements have been made for hundreds of fuels over a range of fuel-air ratios, pressures and imtial tempera- tures. Kmetic knowledge of why flames propagate at the velocities they do, while interestmg, is apparently not needed for explosion ventmg research. If the flammable vapor release occurs outdoors, and is substantial m size, as from a petrochemical plant or a hquefied natural gas reservoir, we are concemed with questions such as the possibility of a blast wave bemg generated, and the destmctive radius of such a wave, which probably depends on turbulence and on lammar bummg velocity somewhat as discussed above, and we are also concerned with questions such as how far downwmd of the release pomt we will stiU be above the lower flammable limit. The latter question is purely aerodynaimc. If we con- sider active control of such releases, perhaps by mixing devices such as water sprays or air jets which reduce the mixture below the flammable limit earher than would occur naturally, agam aerodynarmcs rather than chemistry is what is important. Of course the percentage of fuel correspondmg to the lower flammable limit IS chemically controlled, but flammabihty limits can readily be measured directly, so the lack of our abihty to explain them in fundamental terms is ap- parently not holdmg up progress in deahng with vapor, mist, or dust explosions. Of course, basic chemical kmetic understandmg of Imuts might give us some valuable new insight permittmg better prediction or control of explosions, but it IS not obvious that this is so. Tummg now to flame propagation over contmuous hqmd or sohd fuels, we seek aspects that might be kmetically controlled. For flame spread over a hqmd surface above its flash pomt, the rate of flame spread is known to be essentially govemed by the rate of lammar burmng velocity m the fuel vapoi^-air mixtiu-e existmg just over the surface, as modified by any turbulence which may be present, and the previous statements about burning velocity apply. If the hqmd is below the flash pomt, the flame will spread qmte slowly, at a rate controlled by heat transfer which brmgs adjacent flmd to the flash pomt. The heat transfer depends dra- matically on cu-culatmg currents withm the hqmd, near the surface, mduced by temperature gradients from the flame." This process is not completely understood, but does not appear to be dependent on chemical kmetics, except m the sense that the flash pomt itself, which is independently measurable, is chemical-kmetic controlled. Various theories exist for flame spread rate over a solid surface which vaporizes by irreversible thermal decomposition rather than sunple vaponzation. Expen- mental data have been reviewed by Friedman. The most successful theory to date IS that of de Ris ," '" m that his theory successfully predicts the magnitude of the flame spread rate over sohds without adjustable parameters, and also cor- rectly predicts experimental variations with sample thickness and with pressure.
ABSTRACTS AND REVIEWS 195 The theory assumes that each successive element of the surface must be brought up to a critical pjTolysis temperature characteristic of the material before it can Igmte This is obviously a chemical kinetic parameter, but it can m pnnciple be measured in separate expenments and the results are not highly sensitive to the exact value used for the parameter. No gas phase kmetic parameters appear m the theory; i.e., the gas phase reactions are assumed to be mfinitely fast, once fuel vapors and oxygen have diffusively mixed. Reaction is assumed to occur all the way to the surface; m other words, no flame stand-off distance (which would be kmetically controlled) is assumed In reahty of course there is an observable dead space between flame and surface, and previous theoreticians had assumed that this was of govemmg importance m flame propagation, since the heat transfer from flame to surface might be thought to vary m an mverse manner with the width of this dead space, possibly going to mfinity if the dead space approached zero de Ris treats the heat transfer as occurring in a two-dunensional flow field, and obtams a fimte heat flux to the surface, smce only an mfinitesimal portion of the flame touches the surface, m the model The crucial heat transfer is from the flame through the gas m front of the flame to the not yet igmted portion of the surface. In view of the success of this theory, there is good reason to beheve that, at least to a first approximation, chermcal kinetics affects flame spread over hon- zontal surfaces only insofar as pyrolysis is concerned, and to treat this de Ris's theory only requires knowledge of the surface temperature which must be reached for sufficient vapor production to give ignition, and the endothermic energy re- qmrement of the sohd. Chemical kmetic studies relevant to pyrolysis of sohds would thus have a bearmg on such flame propagation. I t should be mentioned that it has been well estabhshed by McAlevy and co- workers"' that flame spread over thick horizontal plastic slabs increases with about the 2/3 power of pressure. Smce gas-phase thermal conductivity as well as sohd properties are essentially independent of pressure, it seemed logical to as- sume that a pressure-dependent chemical kmetic process m the gas phase exerted a controUmg influence However, de Ris's theory provides the needed 2/3-power pressure dependence aerodynamically, by showing that the gravitationally-mduced gas phase flow above the fuel bed and toward the flame, expressed as mass flow per unit area, mcreases with the 2/3-power of gas density, and the propagation speed for a thermally thick fuel is proportional to this mduced oxygen-supplymg flow velocity. (For a thermally thm fuel, the same mduced velocity effect is present, but the term representmg it cancels out of the final equation for propagation rate, which therefore becomes mdependent of pressure.) Let us consider one final case, flame propagation with several fuel elements separated by a gap. In this case, a fire burnmg at one location generates radiative and/or convective heat fluxes which ignite a nearby fuel element. Examples could be a forest fire, a fire m a room contammg discrete flammable furmshmgs, a fire m a corridor with carpetmg on the floor and a combustible ceihng or side walls, a row of houses, each 10 ft apart, a warehouse with 10-ft-wide aisles, etc. While the net effect is a propagation of the fire, it appears that what is really happening is Ignition at a distance from the fire, by heat transfer. Thus the chemical kinetic factors which enter mto the ignition process, as discussed m an earher section, are highly relevant to this type of propagation. Of course, we already know empirically that a heat flux of a few tenths of a calone per square cm per sec is sufficient to igmte many common combustibles, and the precise value, when measured for
196 TIKE RESEARCH any material, is found to depend considerably on sample size and onentation. A large area of ignorance in regard to the kmd of propagation under discussion is the ability to predict the magnitude of heat flux produced by a fire of given size at a given location near the fire Of course, the emissivity of a flame depends m part on the soot content, which may be lonetically controlled. (Propagation by firebrands is also a possibihty.) To summarize these examples of fire propagation, factors other than chemical kmetics require the major research effort if better scientific imderstandmg is to be achieved. Chemical kinetics however is mvolved in separately measurable param- eters important m fire propagation, such as critical ignition heat flux, pyrolysis temperature and energetics of sohds, flash pomts of hqmds, and flammabiLty limits and burning velocities of vapors, mists, or dusts. (Suppression Let us consider agents which either are applied to an existmg fire or are in- corporated as additives (fire retardants) in otherwise flammable materials. In the former case, the common agents are water (spray, fog, foam, film, gel, etc.) dry powders (NaHCOs, K H C O j , NH4H2PO4, etc.) carbon dioxide halogenated liqmds (CCI4, CFjBr , CHjBrCl , CF2BrCF2Br). These may act in basically three ways to suppress the fire. The same agent often acts simultaneously m several ways. A. The agent modifies the gas composition so as to render it nonflammable. B The agent reduces the rate of pyrolysis of the condensed-phase fuel, either by cooling or by chemical effect. C . The agent acts as a physical bamer to heat or mass transfer. Gas-phase chemical kmetics is directly mvolved in mode of action A and mdirectly m mode of action B, smce the degree to which the pyrolysis rate must be reduced by coolmg m order to extmgmsh the gas flame is a function of gas-phase kmetics. In the mcorporation of fire-retardants into materials, enormous empmcal m- formation has been collected, as reviewed by Lyons'' with about 1400 references, chiefly based on patents. Compounds of phosphorus, antimony, boron, chlorme, and bromme are promment. The modes of action are generally unknown. In many cases, the pnmary effect appears to be to modify the pyrolysis productsâfor example, givmg more char and less flammable volatiles. In other cases, the additive or reaction products thereof vaporize and perhaps mhibit the gas flame, and m yet other cases the additive forms a surface coatmg or bamer which may suppress glowmg after the gas flame is extmgmshed. In this discussion, we are concerned only with the gas phase chemical processes. The present state of ignorance prevents any quantitative statement m regard to how much the overall problem of improvmg fire suppression can be helped by better knowledge of the gas phase part Clearly, we need coordmated studies of both condensed phase and gas phase processes as mfluenced by both physical and chemical effects of additives
ABSTRACTS AND REVIEWS 197 Let US consider some simple examples of gas-phase combustion inhibition. First, we know that an inert additive such as mtrogen or a nearby cold surface (quenching) will reduce bummg velocity and ultimately render a mixture non- flammable. We beheve that the pnmary effect is a reduction of flame temperature and a correspondmg reduction in chemical reaction rates. Roughly speaking, when enough nitrogen is added to hydrocarbon-air mixtures to reduce maximum tem- perature below about 1500°K, the flammabihty limit is reached. The propagation reactions must have a finite activation energy, smce it is known that very Uttle reaction occurs in the cooler (or preheat) portion of a lammar flame. If H - l - 0 2 - » 0 H - | - 0 IS the critical cham-branching reaction, with its activation energy of 17 kcal, then probably the mtrogen reduces the temperature so that the branching rate is no longer adequate, m view of prevailmg heat losses (probably by transverse conduc- tion), or possibly radical recombmation reactions, and the flame limit is explamed, at least to this qualitative extent. Now, when we add some substance other than mtrogen, we can predict the proportion needed to reach the flammabihty Imut by producmg the same degree of coohng, strictly on the basis of well known thermo- dynamic properties. When a substance proves to be much more effective than what would be calculated m this general maimer, we say that it is a chemical inhibitor, and it must be in some way mterfermg with the chemical kmetics of the flame, and probably competmg with the cham-branchmg process by reacting with chain earners (H, OH, 0 ) . The best known example of such an inhibitor is bromme, mcorporated into almost any volatile molecule that can decompose readily m the flame Bromme is roughly an order of magmtude more effective than nitrogen, on a molecular or volume basis, and hence it is beheved that some specific chemical reaction such as H-|-HBr->H2- |-Br or 0H - | -HBr ->H20+Br must be responsible for the effect. A number of researches have been done on this subject, as reviewed m 1967 by Fnstrom,'' m 1969 by McHale," and agam, most thoroughly, m 1971 by Fnstrom and Sawyer.'' No conclusive demonstration of the reaction mechamsm has emerged. An unanswered question is whether the mhibitmg species, perhaps HBr, can only act once dunng its passage through the flame, or whether it can be regenerated to act a number of tunes. If the latter, the kmetics of the regeneration process becomes very important While most such studies have been done m premuced flames, because they are relatively simple, it has been found that these bromme compounds also mhibit diffusion flames, which are more hke fires For example, Milne et al " showed that with a counterflow methane-air diffusion flame, addition of either 0 8% CFsBr to the au- or 20% CFsBr to the methane would produce an equal and large effect on the velocity at which blow-off occurred. This large difference in effectiveness of addition on the fuel and on the air side of the flame is related to the fact that methane and air combme m stoichiometric proportions of 9 5 to 1, so that the flame will be located well on the air side of the stagnation pomt where the two
198 FIRE RESEARCH countercurrent flows meet. Thus, for this geometry, additive mtroduced into the air has a better chance to pass through the flame than additive in the fuel. Real fires would not bum m qmte this type of geometrical arrangement, and further study IS needed to understand this aspect of inhibitor effectiveness However, testmg mhibitors on real fires is not a very quantitative way of companng ef- fectiveness, as well as bemg completely useless m determinmg mechanisms. It must be said that the first practical bromme-contammg agents for suppression of gaseous flames, CH2BrCl and then CFjBr , were mtroduced two or more decades ago, a time when combustion scientists had not studied chemical flame inhibition at all, and m fact there was a widespread behef that it was not possible to mhibit or accelerate a flame with low concentrations of additives, because of the tre- mendous concentration of highly reactive radicals beheved to be already present m the flame front. The data collected m the past two decades show conclusively that this view was wrong Both practical fire-extmgmshmg tests and laboratory flame studies amply demonstrate that chemical perturbation of the flame is possible. The most spectacular cases of chemical inhibition reported are by Lask and Wagner,*" who found that roughly 2 parts m 10,000 (mole basis) of any of the followmg chromium oxychlonde Cr02Cl2, lead tetraethyl Pb(C2H6)4, and iron pentacarbonyl Fe(C0)6, could reduce the burning velocity of stoichiometric hexane-air to about 30% of its mitial value. The mechamsms are qmte unknown, and may well mvolve formation of fine metal or oxide particles m the flame. These effects deserve further mvestigation, even though the anti-knock mechanism of tetraethyl lead m mtemal combustion engmes has never been completely un- raveled, m spite of enormous research effort. Sodium bicarbonate has been widely used for many years as a dry powder extmgmshant. Comparisons of effectiveness with other salts strongly suggest a chemical kmetic effect above and beyond the obvious coolmg and dilutmg effect of the endothermic decomposition. I n the past decade, empmcal comparisons of sodium bicarbonate with potassium bicarbonate have shown that the latter is more effective per unit mass, and even more effective per mole, than the former, so potassium bicarbonate is now m widespread commercial use. Assuming a chemical mechanism of mhibition, it has never been estabhshed conclusively whether ho- mogeneous or heterogeneous reactions are of cntical importance Inasmuch as: (I) a variety of alkah metal compounds are effective; (II) there is a correlation between effectiveness and ease of decomposition or vaporization; and ( I I I ) ho- mogeneous reactions can proceed more rapidly m a flame than surface processes requirmg diffusion to and from the surface, it can be argued that any alkah metal salt may vaporize m a flame to form a common species with mhibitmg power. Friedman and Levy^^ have proposed this, and have suggested that the mhibitmg species is the alkah metal hydroxide vapor, undergomg reactions such as K O H + H - ^ H j O - l - K or K 0 H + 0 H - > H 2 0 - | - K 0 . However, no proof of such a mechamsm has appeared, and there are very few if any current research programs to mvestigate such mechanisms, to the writer's knowledge. Let us consider the inhibition of the gas flame over a burning sohd, by an additive
ABSTRACTS AND REVIEWS 199 to the sohd. As previously noted, there is httle available but speculation m this area. In one case, Fenimore and Jones^* have done some careful work to demon- strate an example of this type of mhibition. Antimony oxide is a known mhibitor of certam plastics They showed that it had no effect on polyethylene's hmitmg oxygen mdex, but a pronounced effect on partially chlormated polyethylene, the effect mcreasmg with mcreasmg chlormation This suggests formation of a volatile antimony hahde which inhibits the gas flame. (It is already known that antimony pentachlonde vapor is a good inhibitor of hydrogen flames.) They measured the antimony remaming m a partially burned piece of plastic and found a preferential loss of antimony, provmg vaporization. The antimony did not vaporize preferen- tially when no chlonne was present They replaced the oxygen m the atmosphere by nitrous oxide, thereby altermg the gaseous cham reactions in the flame (H-f O2â>0H+0, to give branchmg, no longer occurs) and found that there was no mhibition m this atmosphere This seems to be very conclusive cu-cumstantial evidence for gaseous mhibition by a volatile antimony hahde when Sb203 is added to a chlormated hydrocarbon. To summarize this section, chemical kmetics is critically involved, and poorly understood, even for a suppressant actmg by a physical mechanism, such as heat abstraction. Further, it is well estabhshed that certam suppressants m practical use, havmg been proven effective empirically, owe their effectiveness to chemical kinetic perturbations of the gas flame, the mechanisms bemg imknown. Further research could lead not only to new suppressants but also to new msights on how to use more effectively the suppressants we have We need to leam not only the chemical gas-phase inhibition mechamsms but also the role of the gaseous reactions m the overall bummg and extmguishment of real, sohd fuels. Toxic Products and Smoke Fires burning m buildmgs under poorly ventilated conditions produce carbon monoxide at concentrations more than an order of magmtude greater than the lethal level Furthermore, the products of incomplete combustion of organic materials generally contam smoke and eye-imtants which mterfere with v i s i o n and hence escape. In addition to these major problems, various secondary problems exist, which imder some conditions could become very serious (A) Smoke may cause lung damage; (B) If chlormated combustibles are involved (as from polyvmyl chloride), then very probably hydrogen chloride and possibly phosgene may be produced, (C) If nitrogen-contammg organic combustibles are mvolved (polyurethane, wool), then hydrogen cyanide may be found; (D) Partially oxidized hydrocarbons (acrolem, formaldehyde) may form; (E) Fire-retardant additives (e g., phosphoms compounds, antimony com- pounds, bromme compounds) may generate toxic products or stunulate smoke generation; (F) Oxygen depletion and carbon dioxide have well known physiological effects; (G) The toxic nature of a mixture of dangerous gases (and smoke) cannot
200 FERE RESEARCH easily be estabhshed because of synergistic mteractions between com- ponents. To put the vanous toxic hazards m perspective, relative to carbon monoxide, we hst the threshold hmits for an eight-hour workday (Federal Safety and Health Standards, 1960): Substance ppm {by vol.) CH2CHCHO 0.5 COCI2 1 H C l 5 H B r 5 H C H O 5 H O N 10 CO 100 CO2 5000 This ordering must be balanced agamst the known abihty of an underventilated fire to produce up to 10% CO (moisture-free basis), while concentrations of the more lethal gases, on the basis of measurements to date, are far lower than this. Smce smoke, carbon monoxide, and all other substances hsted above, except H C l and HBr, are products of incomplete combustion, clearly chemical kmetics plays a key role m determmmg relative quantities of these substances Both ho- mogeneous and heterogeneous reactions would be expected to be mvolved. Results are known to depend not only on the chemical matenals present, but also on the degree of ventilation of the enclosure (relative to the quantity of fuel therem), and the temperature level therem (both gas temperature and radiant flux unpmgmg on surfaces) In view of the tremendous difficulty of usmg a scientffic approach to predict toxic concentrations or smoke levels for vanous combustibles bummg imder various conditions, research mto the underlymg kmetics must, m my opmion, be justified primarily on the basis of gmdance which may be obtamed m developmg empirical matenal test methods. There is practically no hterature on this problem m the scientific journals. Empirical methods for measurmg smoke generation capabihties of matenals, by hght obscuration, has been reviewed by the A S T M Committee E - 5 on Fire Tests of Matenals and Construction.^ Smoke can consist either of tar vapors which condense (white smoke from a just-extmguished match) or soot (from a burning pool of benzene) or acid mist (if H C l is present) or vanous combmations thereof. There is of course a great difference m smoke generation by smouldermg com- bustion as contrasted with flaming combustion, for many matenals. For smoke formmg by condensation, the coolmg effect of ambient air mixmg mto the com- bustion products produces more smoke, but at the same tune this mixmg dilutes the smoke already present. Hence, at what pomt m the vicmity of the fire should smoke be measured? Rasbash and Stark^' have burned cellulosic fuels m compartments with vanable ventilation, to study effects on gaseous products Carbon monoxide concentrations up to 14% and methane concentrations up to 1 5% were found. The results were very roughly correlated on a plot of log(% CO) vs logiAH^'^/W), where A and H are area and height of wmdow openmg, m meters, and W is weight of fuel m kilo-
ABSTRACTS AND REVIEWS 201 grams: Aff"VTF % CO 10-* 13 10-' 7 10-2 2 10-» 0.2 The size of the chamber, the size of the fuel elements, and the degree of insulation of the chamber influenced the results. More recently, Tewarson** made similar expenments and found trends m his data which disagreed sharply with the trend shown above. For example, m many of Tewarson's tests the percent C O mcreased with mcreasmg ventilation m certam regions, for a given fuel content m a given chamber. He also showed complex dependence of burnmg rate and maximum concentration of other species on test parameters (wmdow openmg, chamber volume, fuel content). He found different charactenstic behavior patterns for fuel loadings above and below 12 kg/m'. For wood-bummg expenments, he re- ported CH4 up to 4%, and C2H4 up to 2%, while CO ranged from 0 6% to 10 9%, with 8% CO bemg frequently found. No such experunents have been done as yet with fuels consistmg of common plastics, and no satisfymg explanations are avad- able for the trends m the cellulose data. I t seems very clear that much more research along these hnes is reqmred. We need to be able to distmguish between these alternate mechanisms: 1. The undesirable product of mcomplete combustion was produced by pyrolysis of a sohd, and never passed through a flame front. 2. The undesirable product, produced by pyrolysis, survived m part its passage through a flame front. 3. The undesu^ble product was formed m a flame front. 4. The undesirable product was formed m a nonflame gas-phase process, associated perhaps with mixmg of ambient gases of different composition and temperature with the ongmal combustion or pyrolysis products. 5. The undesirable product was formed by interaction of products of a gas reaction with an element of sohd fuel (e g , CO2-I-Câ»2 CO). I t seems unrealistic to me that choices between alternatives such as these will be possible solely on the basis of better knowledge of the chemical kmetics of individual reaction steps. The complexity of the problem is such that experiments with at least semi-reahstic systems are probably required, perhaps mvolvmg highly locahzed samplmg, and then the results of such expenments could lead to proposed generalized reaction mechanisms which might subsequently be quantified by use of specific reaction rates, detenmned separately. Summary The rates and mechamsms of chemical reactions important m fire are generally imknown. A half-century of chemical kmetic research has shown much progress, but the goal of quantitatively descnbmg practical processes m basic terms will take some tune to achieve. While fires are basically diffusion flames, with different chemistry and perhaps different response to inhibitors than premixed flames, nevertheless certam pre-
202 FIRE RESEARCH mixed combustion parameters, such as flammabihty limit, mmimum spark ig- mtion energy, bummg velocity, and quenchmg distance, have a degree of relevance to fires. The aforementioned parameters are each dependent m part on chemical kmetic mechamsms. Inasmuch as these parameters are easdy measurable mde- pendently, it is not certam to what degree their further elucidation m chemical kinetic terms will help in solving fire problems. I t appears desirable to have a better understandmg of flammabihty hmits. There is an urgent need for exploratory experiments to define the role of gas- phase kmetics m ignition, flammabihty, fire retardancy, and toxic gas and smoke formation of common sohd combustibles bummg as diffusion flames. The ignorance m these areas is enormous and very few chemical kineticists have been willmg to tackle these problems. References 1 SrEAcrB, E W R ⢠Atoimo and Free Radical Reactions, Reinhold, 1954 2 K o N D B A T i E v , V N : Chemical Kinetics of Gas Reactions, Addison-Wesley, 1964, 812 pp 3. MiNKOFP, G. J . AND C F. H. TIPPER- Chemistry of Combustion Reactions, Butterworths, 1962, 393 pp. 4 FENMOBB, F P Chemistry m Premixed Flames, Macmillan, 1964, 119 pp 5. Combushon and Flame, American Elsevier Publishmg Company, New York, Vol 1 (1957) through Vol 16 (1971). 6 The Combustion Institute, Umon Trust Building, Pittsburgh, Pa A senes of thirteen sym- posium volumes, from 1928 to 1971, totahng about 10,000 pages 7. GLASSTONE, S , K J . LAIDLER, AND H ETRINQ Theory of Rate Processes McGraw-Hill, 1941, 611 pp. 8. KAUFMAN, F . "Oxygen Atom Chemistry," m Progress tn Readum Kinetics, G Porter, E d , Pergamon, 1961. 9 BRADLEY, J N Shock Waves m Chemistry and Physics, Methuen (London), 1962 10 GATDON, A G AND I. R H u R L B The Shock Tube m High-Temperature Chenucal Physics, Reinhold, 1963, 307 pp 11 LEWIS, B AND G . VON ELBE Combustion, Flames and Explosions m Gases, Acadenuc Press, 1961, 731 pp. 12 NoRRiSH, R G. W The Study of Combustion by Photochemical Methods, Tenlh Symposium (IrUemalional) on Combustion, pp 1-18, The Combustion Institute, 1965 13. FRISTROM, R M AND A A WBSTENBEBG Flame Structure, McGraw-Hill, 1965, 424 pp 14 ACKERMAN, M , E F . GREENE, A L MoimsxrND, AND J . Ross A Study of the Reaction of K and CHjBr m Crossed Molecular Beams, Ninth Symposium (International) on Combustion, pp. 669-677, Academic Press, 1963 15 BA0LCH, D L , D D DRTSDALE, AND A C LLOYD High Temperature Reaction Rate Data Senes, Leeds Umversity, Leeds, England, No 1 (May 1968) through No 5 (June 1970). 16 TROTMAN-DICKINSON, A F AND G S MILNE Tables of Bimolecular Gas Phase Reac- tions, National Bureau of Standards NSRDS-NBS-9, vi , 1967, 129 pp, plus supplement by RATAJCZAK, E AND A F TROTMAN-DICKINSON pubhshed by UWIST, Cardiff, Wales, 1970 17. BENSON, S W AND H E O'NEAL Kmetic Data on Gas Phase Ummolecular Reactions, National Bureau of Standards NSRDS-NBS-21, XVI, 1970, 628 pp 18 KONDRATIEV, V. N ⢠Handbook of Kinetic Constants of Gaseous Reactions, Izdatel'stvo Nauka, Moskva, 1970, translated by L Holtschlag and R M Fnstrom, to appear as NBS monograph, 1971. 19. BROWNE, W G , D R WHITE, AND G R. SMOOKLER: Twelfth Symposium {International) on Combustion, pp. 557-567, The Combustion Institute, Pittsburgh, 1969. 20. NBWHALL, H . K ⢠7 M , pp. 603-613.
ABSTRACTS AND REVIEWS 203 21 VON ELBE, G Eighth Symposium {Inlemational) on Combustion, pp 41-43, Williams and WUkina, 1961 22. WELLER, A E , A LEVY, A A PUTNAM, AND J F WALLING The Federal R&D Plan for Air Pollution Control by Combustion Process Modification, Final Report for E P A, Contract CPA 22-69-147, 1971, Battelle Memonal Institute, Columbus, Ohio 23 MERZHANOV, A G AND A E AVEKSON "Present State of the Thermal Igmtion Theory," Combust Flame 16, 89-124 (1971) 24 FENIMORE, C P AND G. W JONES "Modes of Inhibiting Polymer Flammabihty," Combust Flame 10, 295-301 (1966). 25 FENIMORE, C P. AND F J MARTIN . 'Tlammabihty of Polymers," Combust Flame 10, 135- 139 (1966). 26 J o s T , W. Explosion and Combustion Processes in Gases, McGraw-Hill, 1946, pp. 105-113 27. W i L L U M S , F A . Combustion Theory, Addison-Wesley, 1966, 477 pp 28. H n t s c H F E L D B R , J O , C F. CTOTISS, AND R B BIRD Molecular Theory of Gases and Liquids, WUey, 1954 29. DIXON-LEWIS, G Proc Roy. Soc (London) Am, 495 (1967) and A317, 235 (1970), also Combust Flame 16, 243 (1971) 30 GAYDON, A G Spectroscopy of Flames, Wiley, 1957, 279 pp 31. B u R a o T N B , J H AND A F ROBERTS Proc Roy Soc (London) AS08, 39-79 (1968) 32 FRIEDMAN, R ⢠"A Survey of Knowledge about Ideahzed Fire Spread over Surfaces," FIRE RESEAKCH ABSTRACTS AND REVIEWS 10, 1-8 (1968). 33. DB RIS, J . : Twdflh Symposium (International) on Combustion, pp 241-252, The Combustion Institute, 1969 34 DE RIS, J (comments on a paper by LASTRINA ET AL ) Thirteenth Symposium (iTUematwnal) on Combustion, p 946, The Combustion Institute, 1971 35 LYONS, J . W.: Chemistry and Uses of Fire Retardants, Wiley-Interscience, 1970, 462 pp 36 FRISTROM, R M . : "Combustion Suppression," FIRE RESEARCH ABSTRACTS AND REVIEWS 9, 125-152 (1967). 37. MCHALE, E T "Survey of Vapor Phase Chemical Agents for Combustion Suppression," FIRE RESEARCH ABSTRACTS AND REVIEWS 11, 90-104 (1969). 38 FRISTROM, R. M . AND R. F SAWYER Flame Inhibition Chemistry, Presented at AGARD (NATO) 37th Propulsion Panel Meeting, May 1971 (to be pubhshed) 39 MILNE, T . A, C L GREEN, AND D . K BENSON Combust Flame 15, 255-263 (1970) 40. LASK, G AND H G WAGNER Eighth Symposium (International) on Combustion, pp 432-438, WiUiams and Wilkins, 1962 41. FRIEDMAN, R AND J . B LEVY Combust Flame 7, 195-201 (1963). 42. Task Group, Subcommittee IV, ASTM Committee E-5 on Fire Tests of Materials and Con- struction, ASTM Materials Research and Standards, pp 16-23, Apnl 1971. 43 RASBASH, D J AND W V STARK Generation of Carbon Monoxide by Fires m Compartments, Joint Fire Research Organization, Boreham Wood, Fire Research Note No 614 (1966) 44 TEW ARSON, A Some Observations on Experimental Fires m Enclosures Part I Cellulosic Materials, Factory Mutual Research Corporation, 1971 (submitted to Combustion and Flame)
