Formation and Combustion of Smoke in Bunsen Flames THOMAS P. CLARK Yational Advisory Committee f o r Aeronautics, Lewis Flight Propulsion Laboratory, Cleveland, Ohio
T
4
*
-
H E problem of smoke formation and combustion during the burning of hydrocarbon fuels is of interest t o several industries and to designers of high performance furnaces and combustion chambers. I n steelmaking with open hearth furnaces, for example, large increases in capacity are possible if the luminosity of the flame is increased, but the soot thus produced must be largely consumed if the efficiency of the furnace is to be maintained. Also, in the production of acetylene by the autothermic cracking process a certain amount of by-product carbon is formed which is injected into the flame and burned at a later step in the process. The carbon black industry is frequently faced with the problem of burning waste product. The question of whether smoke will burn completely in the reaction zone of a flame has an important bearing on the direction of combustion chamber research aimed a t preventing smoking. If smoke does not burn in a flame the research emphasis should be on the suppression of the formation of smoke. If smoke does burn in a flame, a second line of attack is also possible-namely, the determination of the most efficient method of burning the smoke generated. A11 such combustion systems are highly complex, with turbulent gas flow and mixing obscuring the details of the processes of carbon formation and combustion. It seems desirable, therefore, to study the nature of carbon formation and combustion in as simple a system as possible, such as a laminar flow Bunsen flame with a single flame front and a nonturbulent outer cone. Standard tests have been devised to determine the relative smoking properties of hydrocarbons. One common method is to gradually increase the length of exposed wick in a standard lamp burning a test fuel to determine how high the flame will burn without smoking (2). Tests such as this are run under standard conditions in still air. Practical application of these results should be made with reservation inasmuch as the tests give little information about the smoke forming or smoke burning capacities of flames subjected to outside influences such as high temperature or turbulence in the air surrounding the flame. This paper presents the results of a systematic study of several variables which might be expected t o influence the smoking tendencies of flames. Included are studies of the smoke burning capacities of ethylene-air Bunsen flames, and the effects of initial gas temperature, fuel flow rate, flame length, and secondary air variation on the smoking tendencies of benzene-air Bunsen flames. The experimentation was performed on a laboratory bench scale apparatus utilizing glass equipment where feasible.
il w
BUNSEN FLAME
DETAIL TUBE OF WITHSIDE NOZZLE INSERTED FLAME
ORK STOPPER
WIRE WICK SHIELD
OUTER
GLASS
HOUSING
HOLLOW FUEL CONTAINER OF WICK LAMP
Figure 1.
Ethylene-Air Smoke Burner
The amount of smoke generated per second by the smoke lamp was determined by collecting and weighing the smoke issuing from the top of the unlighted burner tube. The weight determination was made in the following manner: A filter funnel with a sintered-glass filter element was packed with glass wool and weighed. The funnel was then attached to an aspirator and suspended over the smoke stream issuing from the burner tube. After a measured time interval the funnel was removed and rewei hed. The smoke ffow rate in milligrams per second was then cafculated.
APPARATUS AND PROCEDURE
ETHYLENE-AIBSMOKE BURNER. The ethylene-air smoke* burner is shown in Figure 1. The smoke source was a standard smoke lamp mounted in a modified housing as shown. The hot exhaust products generated a flue effect, which carried the smoke up the glass burner tube. Ethylene and air were introduced through the side tubes and the mixture was burned at the top of the burner tube as a Bunsen flame. The indeterminate amount of air entering the base of the tube prevented any direct measurement of the flame fuel air ratios. However, the test flame fuel-air ratio could be estimated with approximately a &lo% error by comparing the unknown flame with ethylene-air Bunsen flames of known fuel-air ratios. The ethylene flow rate was accurate to &5%.
BENZENE-AIRBURNER. An apparatus for making gaseous mixtures of benzene and air is shown in Figure 2,a. Liquid benzene was dripped at a constant rate from a capillary tube into a n inclined evaporating and mixing tube of borosilicate glass. Metered air was introduced into the upper end of the evaporating tube, and the benzene flow rate was such that all the benzene evaporated into the air stream before the liquid reached the bottom of the mixing tube. The benzene flow rate wa8 read on the buret. The liquid flow rate was maintained a t a constant value by means of the pressure relief by-pass shown in Figure 2, a. To start the apparatus, the stopcock was closed, the buret filled with benzene, and the stopper inserted. The benzene was allowed t o flow and was collected a t the take-off shown at the bottom of the mixing tube. The flow stopped when a pressure equilibrium had been established. The etopcock was then opened and t h e benzene was delivered a t a constant rate from the capillary, as
2185
INDUSTRIAL AND ENGINEERING CHEMISTRY
2786
long as the benzene level was above the bypass. With a constant air supply and the take-off closed, a constant fuel-air ratio can be obtained with this device if the room and apparatus temperatures are held constant. Different fuel flow rates can be obtained by using capillary tubes of different lengths. All fuel and air flows were accurate to +3.0%,. RUBBER
STOPPER
FLAME-
FLAME-
STABILIZING
STAEILIZING CHIMNEY
CAP1
CON
FUE
IC)
Figure 2.
for Geuerating and Burning Benzene-Air .Mixtures
Apparatus a. b.
c.
Benzene-air mixer High temperature burner Burner with duct
The effect of initial mixture temperature was studied with the heated burner tube shoFn in Figure 2 , b. The temperature of the heated tube xas controlled by a variable-voltage transformer. Before and after each flame test measurement, the flame was blown out and the gas temperature measured by inserting a thermocouple 5 em. into the top of the burner tube. h simple tube burner was modified to study the influence of secondary air on fuel smoking tendency. The burner tube was surrounded by both a secondary air tube and a larger duct tube, as shown in Figure 2, e. The outside air could be excluded and the flow rate of secondary air could he controlled.
Vol. 45, No. 12
the carbon burns in the outer cone with oxygen, which diffuses in through the hot reaction products. The amount of smoke that could be burned in the flame was limited by the operation of the smoke lamp. At smoke deliveries above 0.12 mg. per second the smoke formed clumps, some of which passed through the ethylene flame as particles of soot. Smoke collecting measurements made with and without a flame a t a smoke flow rate of 0.20 mg. per second indicated that 95% of the sooty smoke mas consumed in the flame. BURNINGOF SMOKE FILARIESTS. The large percentage of‘ smoke consumed in the fuel-soot mixtures appeared t o be limited by the formation of clumps of smoke which formed when the smoking rate of the wick lamp exceeded a critical point. To test the ability of the flame to burn higher smoke concentrations, the smoke of the wick lamp was concentrated in a narrow filament rather than spread uniformly, thus producing local high concentrations of smoke. Inasmuch as the uniform spreading of the smoke had been accomplished by the impinging jets of fuel and air. a t rightangles t o the flow, the spreading could be avoided bydirecting the ethylene flow parallel to the smoke filament (Figure I ) and by supplying air by convection. TT7iththis arrangement, the smoke passed up the tube as a concentrated filament of finely divided carbon smoke to the rich Bunsen flame seated on the burner. Figure 4, a, shows the rich Bunsen flame generated on the nozzle when the air flow is shut off and all the air is supplied by the chimney convection due t o the smoke lamp. The flow rate of ethylene in all the flames shown in Figure 4 was 1.2 cc. per second. Figure 4, b, c, and d, shows the appearance of flames when burning smoke filaments whose flow rates are appreciably less than 0.01 my. per second. With even the smallest filament,q, the smoke ib not completely consumed in the inner cone but extends an appreciable distance into the outer cone before it is completely burned Figure 4, e, f, g, and h, shows the appearance of the flame when burning filaments of smoke within the flow range of 0.01 t o 0.07 mg. per second. As the smoke concentration is increased, the incandescent streak of carbon extends higher into the outer cone until it reaches the limits of the outer cone. Further increase in the smoke concentration causes a thin filament of smoke to escape from the tip of the incandescent streak. From this point on, continued increases in smoke concentration increase the size of this escaping smoke filament.
COMBUSTIOh- OF SMOKE IN FL4hIES
BURNING OF SMOKE.As the first minute amount of smoke is mixed uniformly with the fuel and air, the inner cone of the ethylene-air Bunsen flame (Figure 1) flashes to a brilliant yellow, characteristic of incandescent carbon a t the flame temperature. As more carbon is added t o the mixture, a yellow corona develops around the inner cone within the carbon monoxide envelope of the outer cone. Increased amounts of added carbon cause this yellow coloration t o become brighter and to extend farther into the Outer ‘One’ Figure 3. finally the whole outer cone is filled with yellow incandescent carbon. If a lean flame is fed increased amounts of carbon until the yellow luminosity just fills the outer cone, and the flame is then made richer, the yellow luminosity continues to fill the outer cone as it increases in size. This phenomenon is shown in Figure 3. Thus, in flames of the same fuel flow rate burning equivalent amounts of carbon, a lean flame burns its carbon rn a compact corona, as shown in Figure 3, b, whereas a rich flame burns its carbon as a large brush flame, as shown in Figure 3, f. The changing pattern of the incandescent carbon zone in Bunsen flames of varying fuel-air ratios suggests that in flames burning richer than stoichiometric mixtures, the major portion of
Smoke-Fuel Mixtures Burning in Ethylene-4ir Bunsen Flames a. d.
Lean Stoichiometric
f. R i c h
The smoke flow rate is approximately 0.07 mg. per second just before smoke begins to appear a t the tip of the incandescent streak (Figure 4, h). The total ethylene flow rate is 1.2 cc. per second. .4t these conditions, measurements on Figure 4, h, show that the radius of the smoke filament is 1.0 mm. and the radius of the base of the flame cone is 7 . 5 mm. Inasmuch as the ratio of the areas of circles is directly proportional t o the square of the radii, the ratio of the cross-sectional areas of the smoke filament
December 1953
INDUSTRIAL AND ENGINEERING CHEMISTRY
and the total gas stream is 1 to 56. In the 60-cm. path of the flame tube, the fuel-air mixture should mix uniformly with the smoke in this filament. Therefore, one fifty-sixth of the fuel per second, or 0.021 cc. of ethylene, flows simultaneously through the same cross-sectional area as does 0.07 mg. of smoke. At standard conditions 22,412 cc. of ethylene contain 24 grams of carbon, or approximately 1.0 mg. of carbon per cc.; 0.02 cc. of ethylene thus contains 0.02 mg. of carbon. The ratio of smoke carbon atoms to
Figure 4.
Smoke Filaments Burning in Ethylene-Air Bunsen Flames
fuel carbon atoms within the smoke filament as it approaches the flame front is therefore 0.07 to 0.02 or 3.5 to 1. It is unlikely that this proportionally large amount of smoke could be burned in a homogeneous mixture of smoke, fuel, and air, but it is indicative of therelatively large amounts of added smoke that can be burned completely by a flame.
2787
zones and thus passes through them in a finite time. Oxygen probably reacts with the incandescent surfaces of the carbon particles to form carbon monoxide, with the result that the particles are eroded away as they pass through the flame zones. This process requires time to proceed to completion; thus, the major portion of the reaction would tend to occur in the outer cone. Oxygen must therefore exist in the outer cone to react with the hot carbon. I n lean flames the oxygen is supplied both by the excess air in the mixture and by the diffusion of oxygen into the outer cone from the surrounding atmosphere. With this excess oxygen present, the carbon particles are rapidly oxidized and the luminous yellow color characteristic of hot solid carbon dies out a short distance from the inner cone. In flames richer than stoichiometric, all the oxygen reacting with the carbon smoke must diffuse into the outer cone from the surrounding atmosphere. The longer time required for sufficient oxygen to diffuse into the outer cone to react with all the carbon present means that the smoke travels farther upward in the flame before it is completely consumed. The result is a large yellow brush flame such as the one shown in Figure 3, f. The shape of the burning filaments shown in Figure 4 is probably due to the nature of the diffusion and reaction processes. The outer layers of the hot smoke filament would be oxidized first and the central core would therefore last longer and drift farther upward before it was finally consumed. Such behavior would result in the pointed filament streaks shown in Figure 4. If the filament smoke concentration was too high, the central core would drift out of the hot reaction zone and appear as a thin filament of smoke. In similar fashion, a dense clump of soot would pass through the whole reaction zone without burning completely.
i
MECHANISiM OF SMOKE BURNING
The bright inner cone of the Bunsen flame is assumed to be the reaction zone but, in the burning of added carbon smoke, the smoke tends to burn outside this zone in the outer carbon monoxide cone. This carbon burning in the outer cone occurs even in lean flames where an excess of air exists (Figure 3, a). The most minute amounts of smoke also burn into the outer cone of a rich flame as a faint streak (Figure 4,c). The reason for this behavior may be found in the mechanism of smoke burning. Electron microscope examination of smoke collected from burning ethylene, benzene, and lamp fuel revealed that all three gave smoke particles of relatively uniform size, averaging from 300 to 500 A. in diameter. Smoke from the lamp which passed through the flame front of the ethylene-air Bunsen flame varied widely in particle size from 500 A. in diameter down to particles too small to resolve in the microscope. The different burning behavior of finely divided smoke and sooty smoke was probably due t o the physical configuration of the carbon particle chains, rather than to the carbon particles themselves. The compact soot particle would be less open to oxygen attack than would the same weight of carbon dispersed in long open chains. The finely divided smoke used in the experimental work discussed previously probably consisted of long open chains of carbon particles. The coagulated smoke, or soot particles, was probably .matted clumps of such carbon particle chains. Evidence has been found that carbon first reacts with oxygen to form carbon monoxide before reacting with oxygen to give carbon dioxide ( I ) . Utilizing this information, one can postulate the probable sequence of steps in the burning of smoke. The smoke chains are heated to incandescence in the inner cone and are kept hot in the outer cone. The smoke is in motion relative to the reaction
n n n n
310% METRIC STOICHIO-
F21%TKN jG Figure 5.
Diagrams of Benzene-Air Flames
Shaded areas indicate incandescent carbon
It would appear that a flame can burn relatively large amounts of carbon smoke if the smoke is finely divided and sufficient oxygen is present to react with all the carbon while it is in the hot outer cone. With flames leaner than stoichiometric, the maximum smoke-burning capacity would seem to be limited by the maximum smoke-to-fuel ratio that will still give a stable pilot flame when enough air is added to burn all the smoke. (If the flame burns in air, part of the oxygen can be supplied by the diffusion of air into the outer cone.) With flames richer than stoichiometric, the carbon-burning capacity of the flame becomes critically dependent on the amount of air diffusing into the outer cone. This diffusion factor is important in engineering applications where a liquid or gaseous fuel is injected into air and burned. It would seem advisable, therefore, to investigate the various factors that affect the diffusion of air into hot reaction zones in which smoke is burning.
I N D U S T R I A L A N D E N G I N E E R I N G CHEMISTRY
2788
FLAME CARBON FORMATION AND SMOKING
CARBON-FORMING AND SMOKINC POrNT O F BENZENE-AIR FLAMES.The smoke-burning ethylene-air flame had several shortcomings which made it unsuitable for a general study of atmospheric influences on flame smoking and carbon combustion. Previous experimentation with the benzene-air flame indicated that it might be satisfactory for such a study. The formation of the first free carbon, as evidenced by the appearance of the yellow pip a t the tip of the inner cone, and the saturation point of the smoke-burning capacity of the flame, as evidenced by the appearance of a smoke filament a t the tip of the outer cone, were considered to represent the lower and upper limits of the carbon concentration within the flame (Figure 5 ) . It was felt that the effect of external physical factors would be reflected in the variation of these limits. These two conditions were therefore related to the initial fuel-air ratio of the premixed gases by taking experimental data a t the fuel air ratio a t which the pip of incandescent carbon first became visible in the, flame and a t the fuel-air ratio a t which the flame started to smoke. These fuel-air-ratio points will be called the carbon-forming point and the smoking point, respectively, of the flames. TEMPERATURE. The effect of the initial temperature of the benzene-air mixture was studied in the apparatus shown in Figure 2, b, used in conjunction with the benzene-air mixer shown in Figure 2, a. The carbon-forming point and the smoking point were measured a t a series of temperatures ranging from room temperature to 450" C. The results of this experimentation are shown in Figure 6. As the initial temperature of the gas mixture was raised from room temperature to 450' C., the smoking point fuelair ratio remained unchanged, but the carbon-forming point fuel-air ratio increased from 140% of stoichiometric to over 160% of stoichiometric.
-
0
cw
SMOKING POINT 0 CARBON-FORMING 0
s
POINT
280
0
0 + v)
5
240
W
0
SPONTANEOUS I G NI TlON TEMPERATURE OF BENZENE
E
200
22
*
GE E
change in the smoking point. In this instance it is probable that the enhanced diffusion rate due to the hotter gases was compensated for by the greater diffusion path length resulting from increased gas expansion. The variation of the carbon-forming point with temperature might be explained on the basis of the preferential diffusion of the lighter oxygen and nitrogen molecules
I
I
a
I-
W
I 0
I440-
0
E
-
0 0
SMOKING SMOKING
0 CARBON-FORMING
+ 360-
z W
0
-
a
E 280P IQ
a 200r
5
A!.
w
W
_ _ _
-
V
,o
I
I
1
I
I
a J LL
320
E
0
Vol. 45, No. 12
160
7
-I
W
120
0
I60
320
480
6
UNSHIELDED THERMOCOUPLE READINGS OF I N I T I A L GAS TEMPERATURE, "C. Figure 6. Effect of Temperature on Fuel-Air Ratio at Carbon-Forming and Smoking Points of Benzene-Air Flames The Bunsen cone generates a sheath of hot gaseous combustion products through which oxygen from the surrounding air must diffuse in order to react with the incandescent carbon. Those variables which improve the diffusion of oxygen through this zone should improve the smoke-burning characteristics of the flame. The rise in initial gas mixture temperature caused no
away from the heavier hydrocarbon molecules in the rounded flame tip ( 3 ) . If the higher initial gas temperatures near the flame front enhanced the pyrolysis of the hydrocarbons to two or three carbon atom chain fragments, it might be expected that the diffusion rates of all particles would be equal. I n this case the fuel-air ratio in the tip of the flame would be the same as that in the bulk of the premixed gas. An unheated gas mixture should therefore show incandescent carbon in the flame tip even when the fuel-air ratio of the bulk gas mixture is well below 160% of stoichiometric a t incipient carbon formation, regardless of the temperature or fuel-air ratio of the bulk of the gas mixture. Apparently, no appreciable amount of oxygen was able to diffuse into the flame tip, since no effect other than temperature influenced the carbon-forming fuel-air ratio. FUELFLOW RATE. The rate of the benzene flow was varied and the air flow rate was adjusted until a smoking point and a carbon-forming point fuel-air ratio were determined for each benzene flow rate. This experiment was conducted a t room temperature and repeated a t 450" C. The flame length a t the smoking point was also measured a t each benzene flow rate. The results are shown in Figure 7 . The flame length increased linearly with benzene flow rate, although the smoking point fuel-air ratio, conversely, decreased in some exponential fashion. Temperature had no effect on the smoking point fuel-air ratio, as reported previously. Points from both temperatures fell on the same curve. Variations in fuel flow rate had no effect on the carbon-forming fuel-air ratio, although the fuel air ratio was a different constant value a t each temperature. Increased fuel flow rate lengthened the flame, but decreased the fuel-air ratio a t the smoking point. If it is assumed that the unit-volume concentration of carbon to be burned was the same in these flames of different fuel flow rates, the geometry of the flame might be responsible for the increased smoking tendency of the longer flames. As a cone with a constantdiameter base increases in length, its surface-to-volume ratio de-
December 1953
INDUSTRIAL AND ENGINEERING CHEMISTRY
creases. Since less surface in proportion to the volume of reactants is available in the longer flames, less oxygen diffuses into the outer cone and the smoking point fuel-air ratio decreases.
POINT SMOKING SMOKING SMOKING CARBONFORMING
0
a Q
280
-
LIQUID BENZENE (calmin)
0
0.24 .33 .46 .24, 3 3 AND .46
200-
a
L
14
A
4b0
*bo "I:m
A
h
ISOO"
2doo
SECONDARY AIR FLOW RATE,
;loo
JOO
cc,/min.
Figure 8. Influence of Secondary Air Flow Rate on Carbon-Forming and Smoking Points of Benzene-Air Flames Burner tube diameter, 0.6 om.
SECONDARY AIR. The heated and jacketed burner tube was removed from the benzene-air mixer and a simple apparatus for burning the flame in a protective jacket, or duct, was installed in its place (Figure 2, c). With this apparatus, the amount of air surrounding the flame could be controlled. The effect of air flow variation was determined for three fuel flow rates. The results are shown in Figure 8. Here again, change in secondary air flow rate had no effect on the carbon-forming fuel-air ratio. I n the range of higher values of secondary air flow (beyond the scale in Figure 8), the variation in secondary air flow had no appreciable effect on the smoking point fuel-air ratio. It remained substantially the value for a flame burning in open air with a protective chimney. I n the lower range of secondary air flow rates shown in the figure, the smoking point occurred a t increasingly leaner fuel-air ratios as the secondary air flow rate was decreased,
2789
Flames of all flow rates became unstable and extinguished a t a smoking point fuel-air ratio of approximately 164% of stoichiometric. This value was approached from the opposite direction by igniting a lean flame, either without flowing secondary air or with nitrogen substituted for secondary air. The flames were stable up t o a fuel-air ratio of 140% of stoichiometric a t room temperature. From 140 to 164% of stoichiometric fuel-air ratio, incandescent carbon filled the upper portion of the flame as a faint haze. At this upper value the flame extinguished. No smoke was given off up to a value of 164% of stoichiometric. If the premixed benzene and air were preheated to near their ignition temperature and burned in the absence of air, or in nitrogen, the carbon-forming point and the smoking point would both occur a t approximately 164% of stoichiometric. The stoichiometric equation for the reaction of benzene and oxygen indicates that complete conversion to carbon monoxide and water occurs a t 167% of stoichiometric. Flames burning in the absence of oxygen are stable and do not smoke to a value of approximately 164% of stoichiometric. Since no smoke is formed, the products under these conditions must be almost wholly carbon monoxide and water, since the oxygen present under these conditions would allow only one half conversion to carbon dioxide. Apparently, a t room temperatures, the incandescent carbon forming a t 140% of stoichiometric reacts further up in the flame with the oxygen which has preferentially diffused out of the flame tip, for the flame does not smoke, even though incandescent yellow carbon is visible in the outer cone. LITERATURE CITED
(1) Arthur, J. R., Trans. Faraday Soc., 47,164-78 (1951). (2) Institution of Petroleum Technologists, London, "Standard
Method of Testing Petroleum and Its Products," 3rd ed., pp. 133-6,1935.
(3) Lewis, B., and Von Elbe, G., "Combustion, Flames and Explosions of Gases," pp. 277-8, New York, Academic Press, Inc.,
1951.
ACCEPTED September 11, 1953. REOEIVED for review April 14, 1953. Presented a8 part of the Symposium on Chemistry of Combustion before the Division of Gas and Fuel Chemistry a t the 122nd Meeting of ths A M ~ R I CAN CHEMICAL SOCIETY, Atlantic City, N. J.
Isopycnics and Twin Density Lines SYSTEMS WITH TWO LIQUID PHASES ALFRED W. FRANCIS Research and Development Department, Socony-Vacuum Oil Co., Paulsboro, N . J .
I
I
T OFTEN happens that for one tie line in a ternary system
+
the two liquid phases in equilibrium have identical densities. This tie line, sometimes denoted by a straight dashed line (a), is called an isopycnic (equal densities), and is readily located with precision. Compositions within the binodal area but near each side of the curve are identified as points on the isopycnic when droplets of the smaller phase fail to rise or fall with appreciable speed (provided they have already coalesced t o macroscopic size). This is a convenient method for accurate location of one tie line. Isopycnics have considerable practical importance in solvent extraction since with compositions in their vicinity settling of the layers is extremely slow. Some otherwise suitable solvents or conditions may be thus disqualified. At least 95 of the published ternary systems have isopycnics. These are listed in Table I with the page reference in Seidell's book (3). The systems are arranged nearly in order of increasing number of carbon atoms.
However, very few of the isopycnics are noted in the original papers (indicated by footnotes), and none are called by that name. Mondain-Monval and Quiquerez (7) listed three binary and 15 ternary systems with inversions of density, and gave experimental data for two of each group. In three other published systems an inversion tie line is indicated. I n a few other papers density observations on the two phases make clear the existence of an isopycnic without showing its exact location. It is hoped that future publications of systems having isopycnics will have them marked. One system has two isopycnics. I n the water-propionic acido-toluidine system of Figure 1 (a, p. 961) a liquid phase changes in position (upper or lower layer) five times in progressing around the binodal curve, at the plait point, and a t each end of each isopycnic. The portions of the curve are marked U for upper and L for lower layer. Isopycnics occur either with a free binodal curve (a bite out of
