Carbon formation in flames

46

British Section, Symposium 'Developments in instruments' : A. E. S. PENGELLY 'Research into the aerodynamics of furnaces': E. H. HUBBARD

These papers will be discussed at the Symposium, and further papers on the following topics will be presented: Fundamentals of combustion instrument development Pulverized fuel Oxygen in combustion processes Pressure jet oil flames Aerodynamics of furnace flames Heat transfer from industrial flames

Vol. 6

Combustion Institute in the Mechanical Engineering Department of Imperial College, London, on :27 September 1961. The meeting was attended by more than 150 members and visitors including representatives from the European Sections and two visitors from the U.S.A. The meeting was divided into three sections: (1) A review of fundamental mechanisms for carbon formation.

Each paper will be presented and the discussion opened by an acknowledged expert in the field concerned. Application forms m a y be obtained from the Institute of Fuel, 18 Devonshire Street, London, W. 1, or from the Secretary of the British Flame Research Committee at 11 Park Lane, London, W.1. The Secretary has available some copies of a provisional Conference timetable.

Carbon Formation A one-day symposium on Carbon Formation was organized by the British Section of the

(2) Carbon formation in practical combustion systems. (3) Unsolved problems and future research. The British Committee were most anxious to devote the maximum possible time to open discussion; it was therefore decided to dispense with formal contributions and instead to invite informal opening comments in each section intended as a stimulus for discussion. A selection of the contributions presented at the meeting is published in this issue of Combustion and Flame, as follows :

Opening Contribution Carbon Formation in Flames A. THOMAS "Shell" Research L t d , Thornton Research Centre, P.O. B o x I, Chester. (Received January 1962)

Mr Chairman, Ladies and Gentlemen, The problem before us is that of describing the processes whereby molecules of hydrocarbon fuels are converted into particles of soot. We want to know what is the nature of the intermediates that grow to form the particles, and we also want to know what are the 'building bricks', the species that add on the growing intermediates. We would like to be able to write down a complete sequence of chemical reactions, with their rate constants, in order to provide a complete chemical description of the process. At present, however, this is impossible, and all we can do is to start on the first job, that of defining the most likely chemical route in fairly general terms.

There have been m a n y reviews of the mechanism of carbon formation in recent years. One excellent review by Drs GAYDON and WOLFHARD was published last year as a chapter in the revised version of their book, Flames. I could not possibly cover all the ground in that review. Instead I propose merely to look at the experimental facts available, and arrange them in what seems to me to be the most logical order. What then emerges will be not so much a judicial summing up, a reasoned comparison of theories, but rather a case by counsel for the defence, the defendant being the theory that will follow from placing facts in a certain order. Now any mechanism of soot formation must explain or accommodate the experimental facts.

March

1962

Carbon

It must account for the variety of products found in flames; for the effects of additives; for the differences in smoking tendencies of fuels; for various spectroscopic observations; and for certain features of flame structure. What I mean by these features of flame structure is explained in Figure I. We find that in premixed

I

Ahphatm

l

Bluegreen cone

. ~"

Here we have data for the hydrogen content which varies from one per cent to three per cent by weight. This does not sound a large amount, but when it is converted to the percentage on an atomic basis, it becomes approximately 12 per cent to 36 per cent. This is not negligible. In fact I have some details for the hydrocarbon circumanthracene which contains only 3-2 per cent hydrogen by weight. The structure of this hydrocarbon is given in Figure 2 together with some approximate molecular dimensions.

I

slat°rgrey region

/

47

Yellow streak merging with blue-green cone

Dark space

:~

-

I

Yellow streak

~

Aromatic

formation

9~ :j

Figure 2.

Figure

I. P r e m i s e d flames j u s t sufficiently rich to [orm carbon

flames just rich enough in fuel to produce a small yellow streak, this streak is separated from the blue-green region by a dark space: in fl'~nes of aromatic fuels, however, there is no dark space and the yellow and blue-green regions merge. W h y should this be so? Before going further, let us take a look at the nature of soot itself. I say 'soot' rather than 'carbon' advisedly, because soot is not carbon. In Table I I have listed some properties of soot. Table I Mean dimensions crystallites ( A ) Substance

Acetylene black G a s b'.ack Channel black Acetylene soot

Circumanthracene (unit cell of t w o molecules)

a

28'0 19"5 21 23"8

c

13 10

o/

Separation between planes

3"47 3'55 3'55 3"40

I

I %w

hydrogen

J

I}

3tol 3"2

Circumanthracene, C40 H l ~

Now soot particles are built up of crystallites, and the dimensions of these crystallites are listed in Table I. Here you see data ranging from 47.5#, to 21A for one dimension across the planes and 13A in the other. This compares with 23"8A and 10A for the unit cell of circumanthracene, which contains only two molecules. Distances between the planes are about the same for all soots and for circumanthracene. What I am leading to, what I am in fact strongly suggesting, is that soot is not carbon, but simply has a large polybenzenoid hydrocarbon structure. Now soots have another property. They all give strong electron spin resonance signals demonstrating they possess unpaired electrons. They thus have a free radical nature. One can therefore reasonably regard soot as an agglomeration of very large polybenzenoid free radicals. So much for the nature of soot particles. Now let us return to the mechanism by which soot is formed. First consider the reaction times involved. In diffusion flames the whole process of converting simple fuel molecules into large aggregates containing fifty thousand or so carbon

48

British Section, Symposium

atoms takes place in about ten milliseconds. In premixed flames times are even shorter-only about one millisecond sufficing. Such extremely fast reactions immediately point to free radicals as intermediates so that we reasonably think of soot formation in terms of free radical polymerizations. Now consider the temperature of the process. It is extremely high, about 900 ° to 1 800°C for diffusion flames, and higher for premixed flames. We should here focus our attention on the nature of the carbon skeleton in a large molecule. This skeleton must survive these high temperatures without disintegration. It must be extremely stable. It must be able to accommodate considerable energy, acquired thermally, and possibly chemically, without breaking. If one casts about for a structure that can have this sort of property one immediately seizes on conjugated species--either conjugated polyenes, or aromatics and polyaromatics. Not only must the carbon skeleton be extremely stable, it must also grow rapidly. We need great stability, but also great reactivity. One outstanding property of conjugated species is their great ability to undergo addition reactions of the Diels-Alder type. They will add on other small unsaturated species. So we are led to consider the growing intermediate as a conjugated species, and the 'building bricks' as other unsaturated species, small or large. The problem of growth of nuclei has been studied by TESNER in the U.S.S.R., who has shown that soot particles will grow rapidly when exposed to fuel vapour at temperatures well below those at which the fuel would spontaneously form soot. Thus once nuclei are formed, they will grow rapidly. A question that has received considerable discussion is 'How are these nuclei formed?' I suggest that this is the wrong question to ask. We should really ask first 'What is a nucleus?' Is it a particle of a certain size? Or with certain surface properties ? Or what ? It seems to me that the only property of a nucleus that we can define is its chemical nature. It should be able to add on fuel molecules or fragments of fuel molecules at high temperatures to form a new larger species that is just as reactive as the first. It must also

Vol. 6

grow rather than split. So we are again led back to the concept of a conjugated free radical, which fulfils this definition of a nucleus quite well. The critical steps of nucleus formation m a y well be simply the first few reaction steps needed to produce a conjugated polyolefin.

C1and C2 species (CH3,CH4,C2H2,C2Hz,etc) /Smaller atkyt radicals i Olefins " ~ 1 / Alky[ radicals l Aliphatic fuel

Aromatic rue[

ConJugated ! po[yene radicats~ 1 ""~Aromat, radicals Polybenzenoid

radicals Soot

(Very large polybenzenoidradicals) Figure 3. Proposed scheme of reactions

On the basis of these somewhat simple and naive considerations I have prepared a reaction scheme which is illustrated in Figure 3. I have distinguished between aliphatic and aromatic fuels. The first reaction of aliphatic fuels is radical formation, say by hydrogen abstraction. The carbon skeleton of aliphatic radicals is not particularly strong so we have breakdown into simpler radicals and olefins. Olefins might then suffer dehydrogenation leading to polyolefins, or might break down to form smaller species. Once a conjugated species is formed, the less the chance of breakdown and the more the chance of addition. Conjugated polyene radicals eyclize and proceed via polybenzenoid radicals to form the very large polybenzenoid radicals that we call soot. Broadly speaking aliphatic compounds break down to simpler systems before they can build up again. Aromatic compounds similarly form radicals as a first step, but these radicals can now polymerize far more readily, and do not need to break down first. They already have fairly stable conjugated structures. Now how does such a theory fit with the facts available? Consider first flame structure and the dark space I mentioned earlier. Since

March

1962

Carbon

aliphatic fuels have to break down before they can polymerize, the process is rather long. Thus oxidation which takes place rapidly gives rise to the blue-green radiation before the carbon forms. So we have the dark space. With aromatic fuels, on the other hand, carbon formation is faster and can now compete effectively with oxidation, leaving no dark space. Then consider smoking tendencies of various fuels. SCHALLAand MACDONALDwho compared smoking tendencies of hydrocarbons found that the one with the lowest smoke point was not acetylene, not benzene, nor any other aromatic fuel, but conjugated butadiene. And here in Figure 3 we see butadiene fits extremely well, directly in line in the process. Then the effect of additives. Those that promote soot formation are mainly halogens or halogenated compounds. They probably function by promoting dehydrogenation, making the process of olefin and polyene formation easier, and assisting in ring closure. In fact the preparation of circumanthracene involves a Diels--Alder type of condensation in the presence of chloranil as a dehydrogenating agent. It may be that the promoting effect of sulphur trioxide is associated with a similar function with oxygenated species, this time effecting dehydration rather than dehydrogenation. The evidence I have used so far is simple and descriptive. I have mainly considered times and temperatures of reaction, and flame structure. How does the simple theory that I have outlined compare with other theories of soot formation? Take for example the theory of acetylene as an intermediate. Well, that is far from excluded for aliphatic fuels, for it may well be one of the species formed during breakdown before the building up process starts. There is no reason why it should be the only species, however. Then take the theory of polybenzenoid intermediates. That again is in harmony with this reaction scheme except that we now talk about polybenzenoid radicals. But what of the other theories, those based on fulvenes, or oxygencontaining compounds, or other compounds isolated from flames as intermediates. We need

formation

49

to say a special word about these, indeed about the whole relevance of products isolated from flames. I put it to you that these products are not intermediates in the process, but are species that have fallen by the wayside. They are products of premature chain termination. The very fact that they have been isolated means that they were not reactive enough to go on to form soot. But although not themselves intermediates they are still guides to the road travelled by the reaction.

Table

2.

Reacttwty to D i e l s - A l d e r a d d i t i o n s r e l a t i v e to t h a t o] a n t h r a c e n e )

Polyolefins

Linear polyacenes

Butadiene

10~

Benzene

Hexatriene Octatetrene Decapentene

10 ~ 10

1

Naphthalene Anthracene Naphthacene

l

Pentacene

Infinite polyene

Hexacene Infinite polyacene

i

Bent poly,acenes

10-'~ L Phenanthrene 10-* [ Chrysene

1

(values


10 10 ~

j Pyrene -<10 -~ I Perylene <10-' , 3.4-Benz[ phenanthrene ~_-10 *

10'

J

I 10'

[

(BRowN, R. D. J, chem. Soc. 1950, 691 and 2731)

In Table 2 I have listed the reactivities of certain compounds of the type we have been talking about. They are calculated values by BROWN, published in 1950, of reactivity towards Diels-Alder additions. You will observe that some compounds have very great reactivity-butadiene particularly, and the linear polyacenes. Once, however, a polyaeene becomes bent it loses its reactivity and becomes very inert. It is these very compounds of low reactivity that are in fact found in flame products. In conclusion, I should like to recapitulate three main points. First, the one I have just made. Compounds isolated from flames are not necessarily intermediates in the process of soot formation: in fact one might say they are necessarily not intermediates. Secondly, soot is not carbon, but an aggregate of large polybenzenoid hydrocarbon radicals. Thirdly, simple considerations of time and temperature suggest very strongly that the growing intermediates in the process are highly conjugated free radicals.

50

British

Section,

General Discussion Following the opening contribution of Dr Thomas various members contributed to a general discussion, from which the following extracts have been recorded.

Dr R. F. Strickland-Constable (Imperial College): The following general proposition is believed to be applicable to all reactions by which carbon is formed in a gaseous reaction, and is therefore applicable to the formation of carbon in flames. This proposition is to the effect that the rate of formation of solid carbon b y reactions such as 2CO ---+ C O , + C

....

[1]

must be proportional to the amount of carbon already formed. For consider a system in which solid carbon is in equilibrium with gaseous carbon monoxide and carbon dioxide inside a container with inert walls. In this system the rate of the reverse reaction

CO~+C

>2CO

....

[2]

must certainly be proportional to the surface area of the solid carbon present, since the carbon here is one of the reactants. But since the gases are in equilibrium reactions 1 and 2 must be taking place with equal velocities in opposite directions: hence the rate of reaction I must

also be proportional to the surface area of carbon present. If for instance the amount of carbon present were doubled, reaction 2 would necessarily proceed twice as fast, and so, to maintain equilibrium, reaction 1 would also have to proceed twice as fast. The proposition has been derived above for the case of gases in equilibrium. But even if the gases are not in equilibrium it is reasonable to suppose that the reaction will follow the same kinetics, and that the rate will be proportional to the amount of carbon already deposited. The same argument will apply in general to any gaseous reaction giving a solid product, and will therefore apply to all flame reactions which give rise to solid carbon. That reaction 1 is necessarily catalysed by solid carbon would appear to be related to the question of the difficulty of nucleation of a new solid phase: in the absence of solid carbon

Symposium

Vol.

(~

reaction 1 will only occur with great diffÉculty, since carbon nuclei will have to be produced. But once carbon nuclei have formed the reaction will proceed easily, the carbon being deposited on these nuclei, and it is easy to believe that the rate will then be proportional to the amount of solid already deposited. [See also STRICKLANDCONSTABLE, R. F. Trans. Faraday Soc. 1960, 56, 1492.]

Dr R. Long (University of Birmingham) found particular interest in the suggested participation of conjugated polyene radicals. The suggestion that m a n y of the substances occurring in association with soot, e.g. fulvenes, oxygenated compounds and polycyclic aromatics, are the unreactive compounds, perhaps unreactive products of chain termination steps and not necessarily intermediates, was also of very considerable interest. Some years ago, dienes had been postulated as likely intermediates in soot formation*. Dr Long went on to describe some then unpublished work~" on the inhibition of the formation of polycyclic aromatic hydrocarbons during hydrocarbon combustion, with particular reference to the carcinogen 3,4-benzpyrene. 3,4-Benzpyrene has been identified in soots, carbon black, certain smoked foods, cigarette smoke and in the exhaust smoke of diesel engines. The mode of formation of 3,4-benzpyrene had been reviewed by G. M. BADGER et al., who suggested possible intermediates during the pyrolysis of organic compounds and pointed out that the observed products could be accounted for by secondary reactions involving initial or primary free radicals. In the preliminary work described by Dr Long, known amounts of nitrocompounds, e.g. nitropropane, were introduced into a stream of 'commercial propane' (Bottogas) which was burnt as a laminar diffusion flame in air. The air supply was adjusted to give soot formation in the absence of additive and the soot was collected with and without additive present. A weighed amount of soot was extracted with *THORP, N., LONG, R. and GARNER. F. H. Fuel, Lond. 1955. 34, S l tLONG. R. and RAY, S. K. Nature. Lond. 1961, 192, 353 *BADOER, G. M., KIMEER, R. W. L. and SrorswooD, T. M. Nature, Lond. 1960, 187, No. 4738. 663

March

1962

Carbon

chloroform in a Soxhlet apparatus for about three hours. The solvent was removed by distillation in an atmosphere of nitrogen and the residue was extracted repeatedly with n-pentane. The weight (rag) of pentane-soluble material per 100 grammes of carbon input in the fuel plus additive mixture was determined. The 3,4-benzpyrene content of the pentanesoluble material was also determined by the method described by E. SAWICKI et al.§. This depends on the fact that 3,4-benzpyrene in concentrated sulphuric acid exhibits a characteristic peak in its fluorescence spectrum at 545 m/~ when irradiated by u.v. light of activating wavelength 520 m/~. The fluorescence spectrum was obtained by means of an Aminco-Bowman spectrophotofluorometer and calibration was effected using a pure sample of 8,4-benzpyrene. It could be seen (Figure 4) that small additions of nitropropane lead to a marked reduction in the formation of pentane-soluble material in the soot and in the amount of 8,4-benzpyrene. Rather surprisingly the soot formation rises to a maximum with increasing amounts of additive and then declines. It was known that the carbon zone in a diffusion flame is a region where the hydro5~

0.9

'~ e ~30 E~g

k2o C

,o

>80

-g÷

40

~

7

o>

~20, E~7_ 0

A ^ 0.05 0"10 Mols of nltropropane per tool of fuel

Figure 4. Effect of addition of nitropropane

formation

51

carbon is decomposed thermally in the absence of oxygen (although OH radicals do reach this zone)l I and it seemed plausible that the production of nitric oxide within this zone (by decomposition of the nitrocompound) might be responsible for an inhibition of free-radical reactions leading to polyeyclic aromatic hydrocarbons. However, the apparent effect on soot formation was difficult to explain and further work was necessary. Nitrocompounds were known to raise the cetane number when added to diesel fuels and although the work described had not been extended to engine studies, Dr Long suggested that the use of such additives might also lead to a reduction in the emission of 3,4-benzpyrene in diesel smoke.

Professor M. W. Thring (University o[ Sheffield): In my opinion, the Engineering Faculties of Universities must provide the fundamental applied science and the bridge between that fundamental applied science and the practical problems of industry. Figure 5 indicates the general arrangement of the bridge which we are trying to build in regard to the problems of industrial flames in general, and of soot formation causing luminous radiation in such flames in particular. The original method of calculating luminous emissivity of flames in boilers and furnaces was to take the nonluminous emissivity and multiply by a factor of two or three, according to choice. The aim of the present work was in the first place to produce a much more reliable formula, and in the second place to provide ways of obtaining a higher or lower emissivity when it was required. The special contribution of Sheffield University Department of Fuel Technology was the work on the middle pier of the bridge, the controlled mixing history furnace. In this system, one arranges that all the fuel has the same history as it passes through a flame, and that one can control the arrival of air or recirculated combustion products during the course of this history over a wide range of values corresponding to the extremes to which fuel will §SAWICKI, E.. ELBERT, W., STANLEY, T. W., HAUSER, G. R. and FOx, F. T. 1960. 2, 273 I]GAYDON, A. G. and WOLFHARD, H. G. Radiation and Temperature, 2nd ed., p 153. Chapman & Hall: London, 1960

Internat.d. Air Pollution,

Flames, their Structure.

British Section, Symposium

52

/

/

I Flow studies of simplified shapes:

|

/~ single fuel

I: +/I +rt`°l's u"I+ c I

T H E

I

l J A E

Models of actual COs

1Isothermal models of CCs I (~

"(~

°rparticle I cloud J

I R O

Isothermal

12Flow of dropletI

I

D

Y

N A

M

Vol. I~

[

C

Shell boiler

: With artificial

Water tube

density difference

, boiler ~Open hearth ~ furnace

B R I

D GE Steel reheating

.

•

Oil

E

~i~

~ ~

\ Pur~¢arbon Cn Hm

It

combustion ~ experiments ~ e ! 2 Laminar premix f I ames

THE Left

bank

Fundamental science Link

A. B. C. D. E. F. G. H. I. J.

IIIbl--~m_ ,otEl][a~ .~ +

ICMH fLames

3Cloud ignition, ~ - ( ~ Velocitymeasuremerits Pure fue'~-"'~- 4Laminar gas $ingle$ige diffusion flames

COMBUSTION

"m~"

+= E .a

o= ~ == ~o

c ~ ~ .m ~

u~

o +

% O ~ ~ E ~ E~ ~ ~ u~

/i

stills

+

Gas turbineCCSRocket CCs

/

BRIDGE

River of prejudice and ignorance

--Ri(jht bank

Practical engineering

Proper design calculations. Empirical formulae for soot formation, combustion rate, emissivity for geometrically similar flames. Semi-empirical formulae for soot formation, combustion rate, emissivity as functions of flame history. Calculation for burning time of single particles in varying atmosphere. Theories of single parhcle or droplet combustion. Scale-up laws for hot burning systems. Turbulent flow and mixing theories. Semi-empirical mixing and flow formulae. Similarity theory. Empirical formulae for aerodynamic properties of flame (length, shape, impingement, recirculation)+

Figure 5.

The building plan ]or the research bridge on turbulent diffusion combustion

be subjected in actual flames. In this way, one can obtain a semi-empirical formula for the laws of soot formation and soot combustion. We have already a soot combustion formula 1 dm =2.81 x 10 -5,h2"27 ( T - 1 280) 2 m d~tto+ where T is the absolute temperature of the flame at the point concerned, and Po2 is the partial pressure of oxygen. It is not known yet over what range of values of the temperature and oxygen partial pressure this formula applies, but it seems to be of fairly wide applicability. When we have obtained a similar formula for soot formation from the properties of the fuel and its history during the heating up period of

the flame, particularly the relative rates of arrival of oxygen and heat, we shall have a reliable semi-empirical formula which can be used for the two required purposes. We have already got a satisfactory formula for the third part of the problem, namely the amount of radiation that occurs when the soot concentration in the flame is at a certain level, namely ~ = 1 - exp ( - BA-~FL) where F is the mean soot concentration, rag/1., at flame temperature; L is the optical path length, cm; B A -= is 0-01:35 at A=2-3 microns and ~e is approximately unity. From the pier of the bridge nearer to the right-hand practical bank, we have obtained a

M a r c h 1962

Carbon formation

Smax. IS a x i a l s o o t

concentration,

m g / L at n t p

fl =20 85-749R÷O 67002 + ?'63x lOt'f +2 42xlO4"[ ~ o o 0

I

2"0

R is C / H r a h o 7 is mean

of fuel

average

bo~hng point of fuel D~stance from burner t¢ = Distance from burner to %tO,Ctllom@trlc ~Olnt

> = Measured -

values

values

: Mean

10

I

OZ,

;

[

06

I

I

0.8

i

1,0

Figure 6. Relationship between soot concentration, fuel properties and position in flame (Sheffield furnace)

still more empirical formula for the soot concentration in a turbulent jet diffusion flame, provided we have geometric similarity, and fuels within a certain range. This formula is of the form

where S ..... is the m a x i m u m soot concentration on the axis of the flame; f~ is the fuel only function of the form 20-85 - 7-49R + 0-670R 2 + 7"63 x 10 -4 T +2"42x 10 -4T ~ where R is the carbon/hydrogen ratio, and T is the mean average boiling point; f2 is a function of distance along the flame expressed as a fraction of the distance to stoichiometric mixing point L~. This function is of the form shown in Figure 6; f:~ is a curious effect which for geometrically similar flames implies the shorter the flame length, the lower the soot concentration all along the flame. W h y the soot concentration should depend on the flame length, rather than on the time and the quantity of fuel burnt is not yet by any means understood. There must be two opposing effects.

ProIessor A. H. Lefebvre (College of Aeronautics): There are several straightforward methods for reducing the concentration of fuel in the fuel-rich zone of the fuel spray, all of

53

which are effective in preventing carbon deposition and reducing exhaust smoke. They include : (I) Increasing cone angle of fuel spray (2) Increasing amount of air associated with spray (3) Increasing general level of turbulence and so promoting better mixing. Unfortunately, the first two methods have an adverse effect on ignition performance, while the third method usually entails an increase in chamber pressure loss. Thus, although exhaust smoke can be completely eliminated by straightforward means, the engineer m a y prefer to control it to an acceptable level, bearing in mind the price he must pay in terms of other important aspects of combustion performance. J. J. Macfarlane (N.G.T.E.): Exhaust smoke has always been present in some degree. In earlier generations of engines using peak pressure ratios of around 4/1, it was never more than enough to colour the exhaust slightly. This even applied in systems where primary zone mixture strengths considerably richer than stoichiometric were deliberately used. Cycle pressure has increased over the years and the present generation of engines employs maximum pressure ratios as high as 15 / 1. The use of these higher pressures has brought with it a marked increase in exhaust smoke. It should be emphasized that the effect of this smoke formation on engine performance is very small. Even in cases which have achieved some public notoriety, where the problem has been further aggravated by thrust augmentation by water injection, the measured combustion loss due to carbon was less than half of one per cent. The importance of this aspect of engine behaviour lies in its nuisance value to both airport authorities and to the public in general. The temperature limitations imposed b y gas turbine blade materials mean that there is always a considerable excess of air in the overall combustion products.

G. O. Goudie (University of Edinburgh): I would welcome the universal addition of the name of the scale to all temperatures. As long as the two scales Fahrenheit and Centigrade are

British Section, Symposium

54

in common use, it is inexcusable to specify scientific data incompletely. I have noted that a major difficulty in studying carbon formation lies in controlling the local conditions in which the carbon is formed. The basis of one of the remaining supports for Professor Thring’s bridge over the ‘chasm of conjecture’ might be studies of carbon deposits in carefully monitored steady flow and stable combustion conditions. An apparatus which might provide these requirements could be developed from the simple concentric tubular burner on which I worked with J. BARR* to produce vortex diffusion flames at high air velocity and low butane velocity. Although Barr’s deposits could be salvaged at intervals, control and continuity would have been improved if the tubular principle were developed

Vol.

Two-stage

Combustion

Figzrrc 8.

Figuw

I

7.

I-5’

A.

Rig for linear vortex

diffusion

flames

into a three-dimensional form, as shown in Figure 7. This should facilitate instrumentation, sampling and optical inspection of the deposit in situ, while preserving the mixing conditions as distinct from Barr’s in a linear vortex, toroidal vortex. ____~ ______ ‘BARR. D 768.

J. Fourth Symnosium Unrernafionnf) Williams & Wilkins: Ba:timore. 1953

on

Combustion.

Gas

E. Perth& (De’partement ‘Combustion Applique’e’, Institut FranFais du P&role): The earlier work of J. R. ARTHUR and D. H. NAPIER* on the determination of favourable conditions for the formation of acetylene and soot in reversed diffusion flames of paraffinic hydrocarbons showed that the products of combustion included detectable quantities of carbon monoxide, hydrogen and various hydrocarbons, among them acetylene and soot as well as carbon dioxide and water vapour. The results obtained with methane appeared to have possibilities of application in the utilization of natural gas. In the work carried out at the I.F.P. the natural gas was burned in two stages (Figure 8). Natural gas _ Primary -_=_v reactant NatUEll Primary, tlame gas

SUPPlY I

of Natural

6

‘*_-__,___ ----XCr-

//-me Secondary

Arrangement

mlXtUre

for two-stage

combustion

It was first partially oxidized in a duct using a stabilized reversed diffusion flame, the products which were formed and the excess natural gas being burned after the admission of secondary air. In comparison with normal combustion without preliminary oxidation the two-stage combustion system could lead to the following: (f ) An increase in the rate of reaction of the secondary flame due to the presence of the rapidly reacting constituents hydrogen and acetylene (2) An improvement in the luminosity of the flame due to the presence of particles of carbon. So far our work has been concerned solely with the speed of combustion. We have studied the flame stability of mixtures of natural gas under the conditions of two-stage combustion. A diagrammatic illustration of the experimental duct is shown in Figure 9. The primary reactant is pure oxygen, air being added for the second stage. The products leaving the primary chamber are mixed with air before the final stage of combustion. The stabilization of the second*ARTHUR. J. R. on Combustion.

and NAPIER, D. H. Fiffh Svmuosium (InternatimaO m 303-315. Reinhold: New York, 1955

March 1962

('arbon formation

Air r 5ampting-,~ point

i

Burner diam. 16mm

•

Primary

~ chamber /

Natural/ '~Naturat gas gas Oxy! en

Figure 9. Experimental duct

ary flame is ensured by a bluff body (a disc normal to the flow). Tests carried out under conditions giving k=

mass flow of natural gas = 1.02 mass flow of primary oxygen showed that for the same percentage of primary oxygen as in the secondary air for single- or two-stage combustion, the maximum relative performance was as follows: Single-stage combustion (no primary

Two-stage combustion (with primary

flame)

flame)

1

1-3

1

2"1

Mass ratio in the mixing section Velocity of flow in the mixing section

These results were obtained with the ratio

55

A systematic study of the radiation from the secondary flame has not yet been undertaken. It has been observed, however, that the flame at the exit, normally transparent, becomes more luminous when the reversed primary flame is ignited. It is agreed that the practical realization of two-stage combustion of natural gas will require modifications based on the outcome of the present studies. We have, nevertheless, formed the opinion that there is a possible application for the work of Arthur and Napier in the field of turbulent combustion. Dr G. W h i t t i n g h a m (British Petroleum Limited) referred to two important aspects of

carbon formation in industrial boilers and furnaces. First, the relation between carbon (soot) content and flame emissivity; and secondly the contribution of unburnt carbon residues to atmospheric pollution. A sharp rise and fall in carbon content along the axis of a fuel oil flame in a hot furnace has been shown (Figure 10) by the work of the

c .o_

g

Primary oxygen Secondary air =0"077 (by mass). The analysis (percentage by volume) of the products leaving the primary chamber was as follows : C0.,=3.5 Unsaturated (C 2H._,+ C2H4) = 5' 2 0=0.8

H CO _22 ---~2 and - - ~---5 CO CO.,

The flow of the primary fluids was definitely turbulent (R~:,..... = 12 600 for natural gas and 20 500 for primary flow of oxygen).

I

I

I

Distance from burner Figure I0.

C H t = 17 H 2 =34-7 H20=21.4

(estimated by mass balance) giving

'

C O = 17"4

Soot concentration along axis oI fuel oil flame in hot (uncooIed) furnace

Flame Research Foundation. The effects of burner thrust, fuel/air ratio, air pre-heat temperature, and steam as an atomizing medium, had been studied quantitatively. The maximum amount of carbon formed by hydrocarbon cracking usually did not exceed two per cent by weight of fuel and it was produced in a zone about one quarter of a flame length from the burner. In the hot

56

British Section, S y m p o s i u m

(uncooled) furnace, the carbon concentration fell to zero at the end of the flame. In a relatively cold (water-cooled) furnace, some carbon could escape from the flame zone and contribute with ash compounds to 'stack solids' in flue gases. The quantity of stack solids rose rapidly as the air/fuel ratio approached stoichiometric and, at a given air/fuel ratio, different burner systems gave different amounts of stack solids (Figure 11). 0.4

] 0.31 d 0'2

/

r0

0-1

.0

i

8

[

9

25

I

I

I

l

10 11 12 13 %CO 2 in flue gases

2b

Air/fuel ratio

I

14

ig

Figure I1. Relation between stack solids and air/ fuel ratio for different burners in a water-cooled

~urnage

The rate of mixing of a fuel oil spray and combustion air had been shown to play an important part and research was in progress on the aerodynamics of mixing and gas recirculati_o_n p_roce_sses in oil-fired combustion chambers.

J. J. Mac[arlane (N.T.G.E.) : Early attempts by Whittle to develop gas turbine combustion chambers with internal vaporizers were bedevilled by carbon formation inside the vaporizer tubes. Although vaporizing systems were subsequently used very successfully by ArmstrongSiddeley, all other British engines have used some form of pressure atomization.

Vol. 6

It is natural that we should look to laboratory research to explain the phenomenon of formation of exhaust smoke. Much work has been reported on the external observation of the effects of such parameters as pressure, flame size and geometry and fuel structure on the threshold of soot emission from diffusion flames. Valuable studies have been made--notably by Dr A. J. Lindsay--of the nature of intermediate products from such flames. The very nature of these flames makes the interpretation of quantitative measurements difficult, however. Premixed flames should permit closer control of the basic parameters but the technical difficulties of working with premixed flames at high pressures have discouraged experiment in this field--a notable exception to this being the work of Street and Thomas. It does seem that further elucidation of the several hotly contending proposals for the mechanism of soot formation is hampered by the lack of quantitative data on the effects of mixture strength, pressure, hydrocarbon type, and flame geometry and composition, on soot formation rate. It was with the hope of being able to provide some of this information that my colleague Frank Holderness and I started a programme of experiments early in 1960. Our main difficulty was the development of a burner which would give stable burning over the chosen pressure range (1 to 20 atm) and which had adequate mechanical strength for long running times. We finally arrived at the design shown in Figure 12 in which a bundle of fine refractory tubes are cemented into a water-cooled annular ring, the flame being isolated from its surroundings by a quartz sleeve. This design is a compromise between a water-cooled all-metal head in which heat losses to the water seriously restrict the stable burning range, and a completely uncooled head which fails mechanically through overheating of the refractory. While maintaining a mean (cold) gas velocity of 25 cm/sec in the burner sleeve, it is possible to vary the injection velocity over a wide range by changing the number of tubes in the bundle. Injection velocities of a few hundred cm/sec give a flat t i m e . At velocities exceeding 1 000

March

1962

Carbon

Insulated

flame sleeve

....~Jlc:. ~

/Traversing suction ," thermocouple

_-J~lll II 11111

Platinum fo,, f - Outer

JIJJIIIJJJlJ ~il/I UUIII

t

quart= sleeve J~ I " 1lilt

Inner quartz /U~j~j ~ sleeve ¢~1111~'~ Water -~rr-]~

,ac,et

IIPt

- Water ,nom,oal ore,

Water ~g I I

in

Cyhndrlcal bundle of 0 006 in bore tubes OUt

~

Thermocoupte for control " of tnlet gas temperature

Small sample of inlet gas bled off to oxygen monitor ....

--\

Gauze labyrinth for final mtxLng of mtroge% fuel and oxygen C02 Shell and tube assembly. Tubes carry nitrogen, fuel and oxygen. Shell pressurized with CO:, \Mounting nut, screws into base of pressure vessel

Nitrogen-fUelsupply" ~

/

Oxygen ~c_] [ supply -- ~ J

Fig**re 12. Details of burner design

formation

57

cm/sec, the flame becomes much longer and turbulent. We have investigated the behaviour of both types of flames. Temperature measurements in the flame showed that with a simple quartz sleeve, there was considerable loss of heat by radiation from the flame. This was reduced to an acceptable value by incorporating a second outer quartz sleeve with a platinum foil radiation shield in the resulting annulus. Fig~,re 13 is a general view of the rest of the apparatus. The test vessel is pressurized with nitrogen. C~ and C Ghydrocarbons are vaporized by heating them in a stainless steel bomb. For safety reasons, a prepared hydrocarbonnitrogen mixture, and oxygen are mixed in the burner body. "[here is provision for continuous monitoring of mixture composition during the test. Solid and liquid products can be collected quantitatively for subsequent examination. The following general observations can be made. (I) Under many conditions, the products emitted from the flame are anything but pure carbon. They are sometimes substantially Interchangeable quantitative carbon filter unit

From nitrogen supply

filter

Pressure control bypass

,l , i

Nitrogen flow to burner

Safety switch: t failure

Nitrogen C3 Oxygen E3 Fuel vapour [ ] Combustion products O2flc y

,~ ',~ '~ ........., ,, , I t: ....

Fuel vaporizer

Solenoid valve

Diluent nitrogen Pressurized CO2 supply = to flame trap Test vessel

pressure

romoX--su,y r Oxygen flow (hence air flow to burner)

Sonic analyser -concentration of fuel in n~trogen Displayed test conditions

Figure 13. Diagram of test apparatus

Oxygen Temperature meter of burner -% 02 supply gas in mixture

58

British Section, Symposium

Vol. 6

X 80 Ib/in 2 gauge working pressure

100 Ib/in 2 gauge working pressure 16 14

4-ring compounds 3- ring compounds

12

[

' "'

Fiuoranthene - X /

Phenanthrene

10

[

08

yrene"

-~ 06 I

0

~04

~.nthracene

-_~ o~0 2 ~ i

(-:- : r

0

04

5-ring compounds

j~-

3,4-benzpyren~,, P

X

- " -.. -.

I ~

i

i

~

,

,

i

J

,

,

J

J

~6-&7-rtng compounds o Kotins compound 'X-; b 1,12-benzperytene "~" [_ ~ c anthanthrene

02 0 1.4

16

Figure 14.

18 20 22 24 M~xture strength,~

26

2814

16

18 20 22 24 Mixture strength,~

26

8

Polycyclic aromatic compounds in soot from benzene [lames, using high velocity burner, 33 X 0"0075 in. diameter tubes, and insulated sleeve

soluble in benzene. Partial analyses have been made which show the presence of aromatic compounds ranging from benzene to polycyclic compounds such as coronene. Figure 14 shows the effect of mixture strength on the quantity of several known polycyclic compounds in the collected material. For quantitative estimations, samples have been collected on glass fibre filters, maintained at a temperature of 100°C to avoid condensation of water. It is hoped in our future activities to include a more detailed examination of the volatile hydrocarbon products which escape detection in the filtration method. (2) For each fuel, data covering a range of mixture strengths are obtained at each of several pressures. The complete data for a fuel can be conveniently displayed in graphs of the kind shown in Figure 15. This shows the envelope

for limits of stable burning as a function of pressure, and within this the mixture strength threshold for soot formation and lines of constant soot formation rate. In this way, we have investigated the burning of the following fuels in synthesized air, with n-pentane, isopentane, n-hexane, hexene-1, c y d o h e x a n e , c y d o h e x e n e , methylcyclopentane and benzene. (3) Equilibrium calculations show that for all these fuels, except benzene, carbon should not appear at mixture strengths weaker than 9 = 3 . 0 - - f o r benzene 9 = 2 - 5 . Our investigations show that in fact, the threshold for soot formation occurs at about 9 = 1"5 for most fuels, the value for benzene being 9 = 1-3. Maximum soot emission occurs at about 9 = 2 and for paraffin hydrocarbons, falls almost to zero again at richer mixtures. For many of the fuels used,

March 1962

Carbon formation Dotted contours indicate probable tota_._~lhydrocarbon formation rates ~

30

Rich limit of stable flame

59

Sohd line contours are of measured 'soot' formation rates expresSed as % of original fuel appearing in products as filterable solidsand tars

/o

-5% I 0%

2.0%

25 /3"0%

"0-

o20 ¢0

-.~ 1.5

~ T h r e s h o t d of'soot' formation

>

g Stoichiometric mixture .

.

.

.

.

05 Weak limit of stable flame

I

I

4

~

1

1

8

12 Pressure

I

I

I

16

$

I

20 atm

Figure 15. Soot lormation in premixed isopentane-air flames mixture strength could be increased to values in the region of ~p= 5 without reaching rich-limit' extinction. Flat flames gave consistently higher soot formation rates than turbulent flames. All fuels show soot formation rate to be strongly pressure-dependent. The importance of flame temperature can be illustrated in two ways: (a) The introduction of the extra insulation around the burner sleeve produced a considerable increase in soot formation rates (hexane x 20, benzene x 3). (b) One of the fuels, n-hexane, was reexamined using argon to replace nitrogen in the prepared mixture. At equilibrium, this would increase maximum flame temperature by 260°C at 9 = 2 . The effect on soot formation was to move the threshold to a richer mixture (9 =-1.9) and to increase soot formation rate ( x 10). The following points are suggested for future investigations :

(a) A detailed examination of product composition (gas, vapour, liquid and solid) as a function of pressure and mixture strengths for one or two selected fuels; (b) More careful measurements of flame temperature; (c) More detailed measurements of the effects of substitution of argon for nitrogen; (d) Examination of the effects of adding water vapour to the fuel-air m i x t u r e - - a device which has been shown empirically to be useful in reducing smoke formation in gas turbine combustion chambers.

Future Research and Unsolved Problems Dr F. J. Weinberg (Department of Chemical Engineering and Chemical Technology, Imperial College): The field of future research and unsolved problems has no finite boundaries. This makes me feel a little better about concentrating on one small and perhaps rather unorthodox aspect of the subject--that of the

60

British Section, Symposium

electrical properties of the depositing carbon. At the same time, I hope to leave you in some doubt as to whether we are ever justified in ignoring the effects to which these properties give rise, in the kinds of investigation we are discussing today. I must start by referring to some observations of effects* of electric fields applied to flames which are potentially capable of depositing carbon (or rather soot; as a physicist I have an even better excuse than the previous speakers for not attempting to distinguish between these terms). The first and most striking observation is that carbon is deposited generally in apparently much greater amounts on the negative electrode. However, because of the ionic wind engendered t under these conditions, the flame is usually deflected. Deposition on a cold electrode adjacent to the flame might therefore be due, at least in part, to this latter effect and no relevant comparison in the presence or absence of field is therefore possible in such a system. To remedy this, lines of force must lie

-

Steam in

Steam out

l

1

Vol. 6

shown, the flame shape does not change appreciably from its unperturbed appearance. The observations, using the system, are as follows. It is found that deposition can indeed be directed by means of a field. When the field is reversed, Figure 17, the flame appears to spill

Figure 17 downward over the edge of the burner and hardly any deposit forms on the collector plate. Instead it grows luxuriantly around the burner mouth and on the matrix electrode within the burner tube. Quantitative experiments were carried out with the collector plate negative. It was found that the form rather than the mass of the deposit is altered. In spite of appearances, the weight of the carbon was never increased by application of the field. The reason becomes obvious when a deposit is studied under a

Matrix

Burner

(a) Figure

16

approximately along lines of flow. This is achieved in a system such as that shown in Figure 16. With the field applied in the direction *PAYNE, K. G. and WEINBERG, F. J. Proc. Roy. $oc. A, 1959, 250, 316 J'PAYNE. K. G. and WEINBERG, F. J. Eighth SymooMum (InternationaD on Combustion, 1960, Paper No. 15

Figure I8

(b)

microscope. Figure 18(a) and (b) show two of a sequence of photomicrographs (constant magnification) at different current densities [greater in (b)]. The effect of increasing the field is seen to be the building of looser aggregates with larger bald patches between them. A photomicrograph of the deposit in the absence

March 1962

Carbon formation

of a field reveals no structure at this magnification. Electron diffraction records of the deposits reveal no field-induced change in the size and orientation of the fundamental crystallites. Comparison of the mass collected with the electric current indicates that there must be very many thousands of atomic masses (based on C) per positive electronic charge. Concerning the theory of these phenomena, the first part seems fairly obvious; the particles of carbon are charged usually positively. Each speck which deposits on the negative electrode increases the curvature and hence field strength locally. It would therefore be expected to attract particles which would otherwise have travelled to adjacent sites. As the peak grows, it protects an ever-increasing area around it from further deposition. This accounts for the observed change in density and the pattern of deposition. The second part of the theory concerns the mechanism whereby the carbon becomes charged. It is not necessary to assume that carbon (or any other particular species) is the parent ion. Two postulates suffice to account for the observed phenomena, viz. that the mobility of the positive parent ions differs from that of the negative ones, and that carbon particles will readily become charged. There is independent evidence for the truth of both postulates. The negative ion in a flame is

Figure 19

6t

generally thought to be an electron whose mobility would be about 1 000 times greater than that of positive ions. For a steady state (constant current) system, the concentration of a species of ions must vary inversely with mobility. Under normal conditions when a field is applied, the probability of encountering a positive ion is therefore vastly greater than that of meeting a negative particle. Furthermore, this process is obviously cumulative because accretion must further decrease mobility. Other reasons for believing this hypothesis are the large mass :charge ratio, the unchanged electron diffraction pattern and the existence* of some negatively charged carbon particles. One aspect of desirable future research undoubtedly involves the practical implications of introducing changes in the magnitude, location and form of carbon deposition in this manner. Our own work at present includes studies of ion attachment to carbon particles and the removal of carbon from flames in a reducing atmosphere. It is possible that the mass collected is limited by oxidation of the hot carbon particles on their way to the electrode. It is easy to prevent this by placing the collecting electrode entirely in the fuel stream of a diffusion flame, so that the carbon particles never approach the flame closer than the pyrolysis zone. However, all this work involves deliberately applied fields. I want to draw your attention particularly to the possibility that these effects are generally important. In a flow system surrounded by earthed surfaces, in which appreciable ionization generates charged particles of very different mobilities (probably about 1:1 000 for positive:negative) and these are carried along in one gas stream, some charge separation is quite likely to occur. We should therefore think in terms of incidental as well as deliberately applied fields. We have today already heard of a number of observations which might be considered in this light, e.g. the pronounced effect which the readily ionizable alkali metals have on the carbon collected. Figure 19, from a different study, shows an electron microscope enlargement of carbon collected from a diffusion flame in the absence of a deliberately applied *Cf. first f o o t n o t e , p a g e 60

62

British Section, Symposium

Invited Papers on Modelling Principles: C. H. BARKELEW (US), J. M. BI~ER (Holland),

field; yet the dendritic structures revealed might well serve as an illustration of the 'electrolytic' process described above. I n view of all this, I would suggest that in considering the form of carbon agglomerates it might prove unwise to concentrate on chemical effects to the exclusion of the physics of the process.

W. L. FONS (US), A. J. GERRARD (UK), R. \V. HART (US), H. ('. HOTTEL (US), R. B. LAWHEAD (US), K. R. LCBLICH (Germany), F. E. MARBLE (US), J. R. OSBORN (US), R. J. PRIESt (US), D. B. SPALDING (ILK), D. G. STEWART (Australia), P. H. THOMAS (UK), M. V~r. THRING (UK),

D. VORTMEYER (UK), E. E. ZUKOSKI (US) Invited Papers on Chemical Reactions and

Ninth International Symposium on Combustion

Phase Changes in Supersonic Flow: S. \¥. BENSON (US), K. B. C. BRAY (UK), x¥. G. COURTNEY (US), L. CROCCO (US),

(Cornell University, 27 August to 1 September) A tentative outline of the technical programme is now available. The Symposium will consist of the following sessions: A. B.

Plenary Evening Lectures: T. M. SUDDEN (UK), F. T. McCLURE (US) Discussions : Detonations (D. R. WroTE, Chairman) Fundamental Flame Processes (W. H. AVERY, Chairman) Invited Papers on Detonations: G. K. ADAMS (UK), K. ANDREEV (USSR), F. P. BOWDEN (UK), F. B. CRAMER (US),

W. C. DAvis (US), J. J. ERPENBECK (US), D. F. HORNIG (US), S. J. JACOBS (US), C. H. JOHANSSON (Sweden), G. B. KISTIAKOWSKY (US), N. MANSON (France), J. N. NICHOLLS (US), R. P. RHODES (US), G. E. SEAY (US), K. I. SHCHELKIN (USSR), D. B. SPALDING (UK), R. A. STREHLOW (US), H.

GG. WAGNER (Germany), C. I-t. YANG (US) Invited Papers on Fundamental Flame Processes :

R. S. BROKAW (US), H. F. CALCOTE (US),

C.

G. DIXON-LEWIS (UK), C. P. FENIMORE (US), R. FRIEDMAN (US), R. M. FRISTROM (US), D. GARVlN (US), E. GREENE (US), J. HIRSCrlVELDER (US), F. W. JOST (Germany), F. KAIJVMAN(US), F. MASON (US), P. j. PADLEY (UK), T. M. SUGDEN (UK), A. VAN TIGGELrN (Belgium), F. J. WEINBERG (UK), H. WISE (US) Colloquia : Modelling Principles (D. B. SPALDING, Chairman) Chemical

Reactions

and

Phase

Changes

in

Supersonic Flow (P. P. WEGENER,Chairman) Reciprocating Engine Combustion Research (E. S. STARKMAN, Chairman)

Vol. 6

J. R. KLIEGEL (US), S. S. PENNER (US), "~V. R. SEARS (US), A. A. ~VESTENBERG (US) Invited Papers on Reciprocating Engine Combustion Research : S. CURRY (US), M. EDSON (US), F. W. JOST (Germany), S. KUMAGAI (J), W. T. LYN (UK), P. S. MYERS (US), C. R. ORR (US), W. L. RICHARDSON(US), F. A. F. SCHMIDT

D.

(Germany), E. S. STARKMAN(US), I-I. GG. WAGNER (Germany), A. D. WALSH (UK) 'Contributed Papers' sessions will be held in the areas of Laminar and Turbulent Flame Studies, Ignition and Inhibition, Reaction Kinetics and Mechanisms, High Temperature Spectroscopy, Ion and Electron Phenomena, Detonations and Transitions to Detonations, Combustion of Solids, Combustion Instability, Atomization of Liquids.

Special attention is d r a w n to the two Discussions. The papers, which are to made available as preprints, will be taken 'as read'. A brief s u m m a r y b y the authors of the highlights of their papers will be followed b y extended discussions from the floor. A n y o n e p l a n n i n g to join in the discussion is urged to procure the preprints, prepare pertinent comments a n d brief contributions a n d inform the respective chairmen of their intent to participate. Significant communications will be included in the printed P r o c e e d i n g s . The Second A n n o u n c e ment, to be distributed in March, will contain order blanks for the preprint books.