1386
R. M. FRISTROM, C. GRUNFELDER AND S. FAVIN
Vola 64
R.IETHANE-OXYGEN FLAME STRUCTURE. I. CHARACTERISTIC PROFILES IX A LOW-PRESSURE, LAMINAR, LEAN, PREMIXED METHAKEOXYGEN FLAME BY R. M. FRISTROM, C. GRUNFELDER AND S. FAVIN Applied Physics Laboratory, The Johns Hopkins University, Silver Spring, Maryland Received Febiuarv 18, 1960
The results of a study of the structure of a 1 / 1 ~ atmosphere laminar, lean, premixed methane-oxygen flame are presented in the form of profiles giving the local intensive properties as a function of distance through the flame front. Thirteen different profiles have been measured including local aerodynamics, temperature and composition. All species except atoms and radicals are included. These data allow comparisons among methods of measuring temperatures in flames; satisfactory agreement was found. The validity of describing t h i flame with a one-dimensional model has been tested by measuring the profiles both on the axis and off the axis. Since no significant trends were found in these results, it was concluded that a one-dimensional model provided a good quantitative description. The precision and sources of error in the measurements are discussed and the probable accuracy of the measurements is assessed. It was concluded that the results showed sufficient precision and spatial resolution to allow a quantitative interpretation of the results.
Introduction The theor,y of flames has been sufficiently developed so that the important chemical and physical processes are well understood,’ and experimental techniques have been developed for studying the detailed microstructure of laminar flames. These measurements allow the determination of local aerodynamics12local ternperat~re,~ and local composition* at each point in the flame. Thus a quantit,ative interpretation of experimental flame structure should be possible. As is frequently the case with new techniques, questions have been raised as to the validity of these experimental methods. These questions are diflicult to answer directly since no accepted techniques are available for comparison, and the parameters required for a direct comparison with flame theory are either unavailable or of too low precision to be useful. This is true even in the case of the I12-Br2 ~ y s t e m which ,~ is the best understood flame. In the case of the methaneoxygen flame, which is even more complex, theory offers no solution. An earlier attempt was made to interpret experimental flame structure data quantitatively.6 This study of the propane-air system was promising, but the results were not of sufficient precision to establish the techniques and interpretation tieyond reasonable doubt. Therefore, the present system (Table I) was chosen as a proving ground for these tech.niques. This methane-oxygen flame was chosen because it is a simple chemical system. It has sufficient thickness so that spatial resolution is not, a. problem, and the velocities and temperatures fa 11 in an experimentally convenient range. This paper deals with the reproducibility of experimental results and probable sources of error in (1) J . 0. Hirsebrelder, C. F. Curtiss and R. B. Bird, “Molecular Theory of Gases and Liquids,“ John Wiley and Sons, Inc., Neu York. h’, P.,195.1,p. 761. (2) R. 31. Fristrom, W. H. Avery, R. Presoott and A. Mattuck, J , Chem. E h t e . , 2 9 , 106 (1954). (3) R. Friedman, “Fourth Symposium on Combustion,” Williams & LViIkins Co., Baltimore. Md., 1953, p. 259. (4) R. iU.Fristrom, R. Presoott and C. Grunfelder, Combustion and Flame, 1, 102 (1957).
(5) E. S . Camr%belland R. %I. Fristrom, Chem. Reus., 68, 173 (1958). (6) R. XI. Fristrim and A . A. Westenberg, Cornbustion and Flame, 8 , 217 (19571.
measurements. Validity of the measurements will be discussed in paper 11,where it will be shown using a one-dimensional model, that matter and energy are conserved at every point in the flame, as indeed must be the case if the data are valid and the interpretation correct. These tests are not trivial since they involve separate measurements of local flame aerodynamics, local temperature, local composition, their spatial derivatives, and a knowledge of the thermal conductivity of the mixture and the diffusion coefficients of each of the species over the entire temperature range covered by the flame. The chance of accidental agreement in such a complex interpretation without adjustable parameters seems remote. Model for Flame Structure The flame chosen for study had flat, axially symmetric geometry and approached in outward appearance the ideal one-dimensional flame usually discussed in theory. All physically realizable flat flames are three-dimensional so that a onedimensional model represents an abstraction, which greatly simplifies the problem. In such a flame, the local properties are assumed to vary only in the direction of flame propagation, the gradients along the other two directions being assumed negligible, and the aerodynamics is described by the equation pv = constant. In real flames, lateral expansion of the gas occurs and the more general formula for one-dimensional flow in a duct of changing cross section must be used, namely, pva = constant. We will refer to this second type of flow as one-dimensional, reserving the term “ideal” one-dimensional to designate systems where pv = constant. This modification of the general theory to include lateral expansion has been made by Westenberg,6 and his formulation of Hirschfelder’s flame equations will be used in the following paper I1 for the interpretation of our data. ,4s previously pointed out, the criterion for a one-dimensional description is that the spatial derivatives of the intensive properties (temperature, composition, velocity) are small along the transverse coordinates X and Y compared with the derivatives along the coordinate of flame propagation 2. This was clearly the case for the present flame since no systematic
METHLNF-OXYGEN FLAME STRTJCTTJRE
Oct., 1960
1387
TABLE I INITIAL AND FINAL STATES OF METHANE-OXYGEN FLAME T, OK. Initial Final Calcd. Exptl. Pneum. probe Particle track Thermocouple oor.
350
Oa CHc A NI COa HnO CO 0.9143 0.0785 0.0034 0,0008 0.0022 0.0006 0
0.7498 0 0.7603 0
2000
CHI
coz H20 co
OCHh
Hz
0
0
.....
2015 f 20 1930 80 1992 rt 10
*
TABLE I1 EXPERIMENTAL PRECISION OF COMPOSITION DATA 0 2
H
OH
0 0
0.0034 0.0008 0.0822 0.1569 0.0040 0.0002 0.00201 0.00528 0.00007 0.0034 0.0008 0.0779 0.1559 0.0010 0.0005 ..... .....
variation of the intensive properties was found with distance off axis (see Fig. I I I C and Table I1 and 111). This finding was in agreement with other flame structure studies on Bunsen flame micro~ t r u c t u r e . ~ , ~For - ~ quantitative study it is necessary that the flame diameter be large compared with the thickness of the reaction zone. This was the case except with respect to the slow secondary reaction of carbon monoxide where the problem is not serious since this region of the flame is not strongly coupled with the primary reaction and is not greatly influenced by gradients.' This reaction appears to be homogeneous and initiated by the primary reaction of methane.
Specic s
Hz 0
-Mean Run 1
deviationa, %Run 2 Run 3
1.92 0.945 2.14 1.63 3.28 2.07 4.25
1.01 2.26 0.642 1.46 1.48 1.41 2.79
0 374 100 075 .208 .454 .263 ,254
Entries are mean deviations of normalized experimental data Ni (defined below) from the smoothed curves of Figs. 3A and 3B. Xi(Z:i Xi(initia1) for Of, CHI, COZ,HzO IVi(') = m ( f i n n -1 Xi(initia1) )
-
Ni(Z) = -xi(z) for co, OCH~,H* Xi (max) This presentation of the data was necessary because of variations in purity of the 0 2 from run to run.
TABLE 111 EXPERIMENTAL F'RECISION O F THERMOCOUPLE TEMPERATURE DATA Mean Run Description deviation, yo
1 0.0012 cm. aoated (axial) 0.67 2 .0012 cm. uncoated (axial) Max. temp. only 3 .00:!5 cm. coated (0.5 cm. off axis) 1.57 4 ,0025 cm. coated (0.75 cm. off axis) 1.09 .0025 cm. uncoated (0.75 cm. off axis) 1.58 5
Necessary and Sufficient Profiles for the Description of a One-dimensional Flame.-A complete description of a one-dimensional flame consists of a family of profiles giving the intensive properties of the flame as a function of an independent coordi(7) R. M. Fristrom, J . Chem. Phys., 24, 888 (1956). (8) R. M. Fristrom, W. H. Avery and C. Grunfelder, "Seventh Symposium on Combustion," Butterworth Scientifio Publications, London, 1959,p. 304. (9) R. M. Fristrom, Ch. 6 of "Experimental Methods of Studying Flames AGARDOGRAPH," edited by J. 5. Rurugue ( t o be published).
nate. This description can take several different forms since the independent coordinate can be distance, time, temperature or compo~ition.~Each of these coordinates requires a different choice of dependent variables. The systems are related and can be derived from one another. We will discuss the system used experimentally. Here, distance is the independent coordinate. The minimum number of variables or independent profiles required for the description is equal to the number of distinct chemical species plus one. In such a minimal description, the composition profiles used must be measurable in some absolute units (e.g., mass per unit volume), and the remaining profile must describe the deviation from true one-dimensional behavior, ie., lateral expansion. I n practice, compositions usually were measured in relative terms so a density or temperature profile had to be provided also. The remaining profile was expressed as the cross-sectional area expansion ratio of a stream tube in the flame front. It was possible and desirable to measure more than the minimum number of profiles. We present twelve independently measured profiles; the minimum number is nine plus one for each radical species considered. Extra profiles provide a cross check on the experimental techniques and give a valuable confirmation of the validity of the temperature and aerodynamic measurements. The only information lacking from a complete description of this flame is the concentrations of the free radical and atomic species which cannot be determined using our present techniques. In principle, radical profiles might be derived from the comparison between the density measurements of the pneumatic probe and the measured composition of stable species, but present data are not precise enough to yield useful information. This lack of radical information is a common one in kinetic studies, but it is particularly unfortunate in flames since there is doubtlo as to the validity of the steadystate approximation which is often used to avoid the problem. The measurement of free radical species has been attacked e ~ p e r i m e n t a l l y , ~ ~ - ~ ~ but some improvement in resolution will be necessary to apply the techniques to the rapid reaction zone of this flame. Meanwhile, the present in(10) E. s. Campbell, "Sixth Symposium on Combustion," Reinhold Publ. Corp., New York, N. Y., 1959,p. 213. (11) A. S. Leah and N. Carpenter, "Fourth Symposium on Combustion," Williams & Wilkins Co., Baltimore, Md.. 1953,p. 274. (12) E. M. Bulewicz and T. M. Sugden, "Spectrochemica Acta IVth International Spectroscopic Colloquium," Amsterdam, Pergamon Press, New York, N. Y., 1956,p. 20. (13) C. P. Fenimore and G. W. Jones, THISJOKRNAL 62, 693 1958). (14) W. E. Kaskan. Combustion 8: Flame, 2,229 (1958). (15) Th. Grewer and H. G. Wagner, Z. physik Chem., 20, 371
(1959).
formation provides a good initial basis for discussing the kinetics of flame processes. Apparatus and Experimental Techniques The techniques for making flame structure measurements are discussed in various places in the literature. Therefore the present disciission will outline them only briefly and indicate where more detailed information may be found.2-4,7-9,1S 16 Materials.-TIP materials used iii these studies were bottled gases, hoth for the flame m d to d i b r a t e thespectromrter. Thwe g:tstv weir of the highest purity available m d analyzed by us prior to use. Analvses of gases used in several runs are given in Table I. These analyses were ot)tnined ~ i m a r i l ythrough mass spectrometry, although several doubtful points wrre confirmed with gas chromaiography. Changes in argon content of the oxygen from run to run are noticeahlc (they are within the manufacturers’ specifications of a99..5Yc),but such minor variations should have no important effect on the flame system. Burner System.-The flame chosen for this study was a flat, 1/10 atmosphere screen flame. It appeared as a flat luminous disc suspendcd above the burner. To maintain :tnd stabilize such a flame a t low pressure requires precise instrumentation. The apparatus consisted of a gas metering system, a bx-ner, a low-pressure housing, and a pump.16 The inlet flows were regulated to 0.1% by critical orifice f l o ~ m e t e r .A ~ ~modified Egerton burner of the screen-type was used. It wx; 3 2 cni in diameter, and used a 100-mesh “lektromesh scretxn .” This gave a uniform velocity profile ( k l T u ) . The burner housing was a 2 ” Pyrex pipe cross with a watcr-cooled chimney. The exhaust flow was led through a vritiral orifice so that pressure was maintained (tonstant i I the rhnmber (&O.l mm.). The pressure was monitored hv a Zimmerli-type absolute mercury manometer. Distance Measurements.-The measurement of position in the flnnie was fundnmental to these studies, and was accsomplished by two different techniques. In the aerodynamic measurements, photographic methods were used and in thc rase of temperature and composition measuremrnts, a preciaiori cathetometer. The camera is L: Nrw Vue Model VC3. It was used with a reversrd Srhncider S r n o n 1 :2 f2 Irns a t a magnification ratio of about 3.5 to 1. Cut film 4 X 5” of the contrast process Ortho-type was used arid developed in D-19. Positions were read from these pictures by making a twelve-fold enlargement on contrast pTper and measuring the enlargement directly, usin< a calibrated eyepiece mounted on a drafting maphine. Positions could be read with the drafting machine to the nearest hundredth of an inch, corresponding to about lip i n the Yeme. Short distances, such as those between ndjacw~timages i n the spot photographs, could be rend to thv nearest 0 OOJ’, corresponding to about 2 p in the flame. Rrproducibility of siieh measurements was 1% or 2p, wbichever was tl-e largrr figure. The edthrtom(>bm used in thew studies wm an eighty pomc’r miciowope with a long focal length and a crosshair ;r!;iew This tvlrscope wa9 mounted to slide along a diameter gro ind strcl rod driven by a precision screw, whirh c.oiild be rrad directly in microns. With this system the positio.1 of a sharp-edged ohiect easily could be read reproducibly to .5p. Flame Stability.-The first woblem in making Position nieasurenirntx iri flames is to determinc whethrrthe flsme stays fixed i n a space during the period of measurement. This mesnn that :L method for detecting small movements of the f l m e fronl- is needed. To the eye this flame was rompletrly stable, hut previous experience4~e indicated that movement might still be a problem. The luminous zone of this flame coilld only be located to within loop, so a mrthod of xreater delicacy was required. This was provided by placing a small (0.0025 rm. dia.) chromcl-alumel thcrmocouple 0.01 cni. above the stabilizing
-
(16) R. M Fristrcm and S. D. Raeaer, “Applied Physics Laboratory,” T h e Johns Hol,kins Univercity, CM 919, August, 1957. (17) J. Anderson and R. Friedman. Rev. S c i . Instr., 2 0 , 61 (1949). (18) R. &I. Fristrom. It. Prescott, R. Neutnann and TV. 13. Svery, “Fourth Symposium 3n Combustion.” Williame PU. n’ilkins Co., Baltimore, Md.. 1958, p. 267. (19) A. A. Westenberg. R. E. Walker and T. P. Fehlner (to be publisherl).
screen. The temperature gradient here was steep ( 5 X 10s “K./cm.) so that a small movement of the flame front was reflected as a measurable temperature change in this thermocouple. With this system i t was possible to detect movemrnts as small as a micron. This includes both slow drifts and oscillations since flame vibrations result in quite noticeable fluctuations in the sensing galvanometer of the potentiometer. The frequency response of the galvanometer was limited to about ten cycles, but it was also possible to check for higher frequency oscillations using a low impedanre amplifier and oscilloscope. The undisturbed methane flame showed no perceptible short period movements (6Z < 2 p ) and after a half hour warmup period, the drift was less than 1 p per hour. Probes and thermocouples disturbed this reading in B reproducible manner. This was not expected since previous experience had been that small thermocoupks and properly designed probes had no visible effect on the flame.* Detection of the effect was due to the increased sensitivity of the thermocouple monitor ( + 2 p ) over visual checks (f.100 p). Taken a t face value, the maximum movement of the flame amounted to 50-100 p (the limit of visibility), and between two adjacent positions, the shift amounted to no more than 2-3 p . These displacements were so small that they have been neglected, altborigh a correction could have been made using the observed temperature shift and the measured temperature profile in the region of the screen. An interesting point, which is discussed elsevhere,g is that movements of the flame were detected which were a small fraction of the diameter of the thermocouple (e.g., 1 p movemrnt could he detected by 3 25-p thermocouple). This was the case because the thermocouple reproduciblv averaged the temperature over the region it occupied. I t was this average that was read on the potentiometer and it changed reproducibly with minute shifts in position. Reference Surfaces.-Having established that the flame was stable enough for precision measurrinents, the next problem was that of establishing a fixed reference surface or point as an origin for the measurements. There were several possible rrference surfaces: The luminous zone of the flame, the surface of the burner screen, or the position of the monitoring thermocouple. The sharpest portion of the luminous zone was the inner edge, which could only be located to *I00 p . Therefore, this surface was useful only as a cross check. The burner screen offered a fixed surface, but it was rather difficult to locate accurately. Therefore, these positions were measured with reference to the upper edge of the bead of the monitor thermocouple, which could br located to rt5 w . Relative Positions and Errors in Position Measurements. -Relative positions were determined either photographicallv or with a cathetometer. The photographic method is subject to a number of other limitations. These included resolution and distortion of the camera and the enlarger lenses, optical distortion due to inhomogeneities in the windows and density gradients in the flame, uneven shrinkage ot t h r film and pagrr used, m d possible deri:ztion of scattered image from the renter of gravity of a particlr. These errors have been discussed in some detai1,2J8 and it was felt that photographic mrasiirements could he made with a precision of about 2 % , with a least count crror of the order of 2-6 p . Cathetomrter measurements used in thermocouple and probe trtiversrs were siibiect t o errors due to reading, the ralibration of the precision screw, and optical distortion in the windows and flame gases. These soiirces of error amounted to only a few microns. The mounting mas sufficiently free from vibration, and the adjustment of the tension was made so that hysteresis and mechanical reproducibility was f.2 p . The cathetometer agreed with a standard laboratory ruled meter to within 5 p in 10 cm. of travel. Absolute Position.-To compare profiles obtained by different experimental techniqucs, it was necessary to rectify the coordinate systems to a common one. All of these coordinate svstems had a common origin (the monitor thermocouple), but in general there NRS a displacement between the position of the “probe” and thr region of the flame actually measured. In the casr of particle measurements, this ctisplscemrnt a a s due t o aerodynamic lag of the particles. In the case of the thermocouple, the displarement mas due to its wake which distorted the flame slightly, so that the thermocoiiple measured a region slightly doanstrram of its
~JIETHANE-OXTGEN FLAME STRUCTURE
Oct., 1960
TIME (sec
360
0
x
1389
lo3).
0.5
1.0
1.5
2.0
I
I
I
I
2.3
2.3 2.5 3.0 3.5 4.0
I
320
280 INNER EDGE OF LUMINOUS
$ 240 u)
\
E 0
>-
200
t
I
160
u
_J
W
>
@
0
I20
6 B
80
CALCULATED FROM TEMPERATURE P R O F I L t RUN RUN RUN RUN
I
,
I
2
I
I
2
,
4ot 0
I
0
0.10
0.2
0.3
visual position. In the case of pneumatic probe and composition measurements, the probes withdrew a sample from a region slightly upstream of the measured position of the probe. Tjiese eilects are understood qualitatively, and rough qiiantitative values can be assigned to the shifts of coordinate systems.* There was sufficient ambiguity in these corrections, however, so that an independent check was desirable. This was provided by comparing the temperature profilrs which can be derived from these three types of measuremcnts. They were aligned by superimposing the “knee” point of teach of these curves. This allows a superposition of these measurements to about 100 p . Closer alignmentri requires other criteria, which are discussed in paper 11. Spatial Resolution.-The spatial resolution which the several techniques allowed is an important factor since the iuterpretation of flame structure data depends upon spatial derivatlves. This resolution was determined by the size of the instrument used; for thermocouple measurements, it was of the order of bead size; for microprobes, it was of the order of the orifics diameter; for particles, itwasof the order of the spare interval between successive images of the particle. Since the data may show reproducibility which is a small fraction of the instrument diameter, it is important to distinguish between positional reproducibility and spatial resolution. Positional reproducibility is an experimental quantity which indicates the !east positional change which results in t reproducible variation. In these measurements, this least distance was of the order of 10 p for each individual run. ReE201utioriis an abstraction which refers to the “true” function being measured. Assiiming that the experimental sensing element averages linearly over its sampling rrgion, re~olutionw ~ identified s with the minimum distancr over which a reproducible second derivative could be detected. In these measurements, the resolution was believed to be 200 1 for the particle-track aerodynamic measurements, 100 fi fo- the therniocouple temperature traverRes and 100 p for the microprobe sampling cornposition results. Aerodynamic Measurements.-Because of lateral CYpanston, it was necessary to make direct measurements of the aerodynamirs of the flame front. I t was assumed that the flame obpyed the ronservation equation pvu = constant. Therefore, since temperature and average molecular weigit vert’ known, it was necessary to measure anly :t single at rodynamic parameter. Actually, two parameters, velocity and stream tube area, were measured by introducing tracer particles of micro-
0.35 0.3 0.5
I
0.7
I
0.9
I I
I
1
2
I3
1.5
n -
scopic MgO dust into the incoming gas stream. Thcsc small particles ( < 5 fi diameter) followed the accelerations in the gas stream with negligible lag.* These were illuminated a t right angles to the direction of viewing and the images were recorded photographically. Ilhimination with zirconium flashbulbs gave streak pictures of particle paths which were measured to derive the area ratio. For velocity measurement, a double or multiple set of flashes of short duration were obtained from an electronic flash lamp and it repetitive pulser. Measurement of the resulting spots on the picture, combined with a knowledge of the time interval between flashes, allowed velocity to be calculated. It was possible to compare the directly measured velocity with that derived from the temperature measurements (Fig. 1). The agreement between the two sets of measurements was satisfactory. A detailed discussion of the techniques can be found in the litrrature.2J8 Sources of Error.-The direct errors in particle-track studies are those of position measurements on photographic film which hLtve been discussed. Errors inherent in the method include inertial lag of the particles, drag due to the thermomechanical efiects and gravity, and the asphericitv of the particles Questions involving the distnrbance of the flame by the injrrted particles also must be answered. The average error of t h t w measurements was 475, or 20 cni /sec. Temperature Measurements in Flame Fronts.-The temperature profile of this flame was measured with two instruments-a pneumatic prohe and thermocouple. Temperatures could also have been derived from aerodynamic measurcments,2 but the results were of lower precision and resolution (see velocity curve, Fig. l), 80 only the mauimum temperatiire calrulated from these measurements is presented (Table I). Initial Temperature of the Flame.-Thri effective initial temperature of the gas in the burner is an important parameter since it and the initial composition define the enthalpy flux through the flame front. Although the gas originates at room temperature, it does not follow that this is the effective initial temperature of the flame, since thc gas may either be preheated or precooled. Our burncr was not e-iternally cooled and was thermally isolated. The primary mechanism for energy transfer to the incoming gas is by convective heat transfer from the screen and burner which, in turn, receive energy by radiation from the chimnry. Since the tcJmperature of the burner and screen were lowrr (375°K.) than the tcmperature of the chimney whii~li radiated onto it ( 1200°K.), the net effect considering the
R. M. FRISTROM, C. GRUNFELDER AND S. FAVIN
1390
2?o
I
~
L_-
0
01
03
02
04
06
08
IO
2 icml,
Fig. 2. Temperature profile-temperature tance (cm).
(OK.) vs. dis-
geometry of the situation was that of preheating and an increase in specific enthalpy of the flame gas. The calculation of the effective initial temperature is not straightforward even knowing the temperatures of the gas, screen and chimney, but a lower limit was set by the incoming gas temperature (300OK.) and the upper limit was set by the screen temperature (375°K.) Using the available information the best estimat,e of the effective initial temperature was 350 &lO°K. This value gives good agreement between the calculated adiabatic flame temperature and the values measured by several techniques. Pneumatic Probe.-The pneumatic probe was a standard double orifice deviceeO which was modified for this work to minimize flame disturbance and maximize spatial resolution. They were uncooled quartz microprobes similar t o those used in composition sampling but of somewhat larger diameter (75 p ) . The pressure measurements were made in the range around 1 cm. with a Zimmerli gage. Even with a minimum volume system, it was necessary also to use a large orifice in order to make measurements on a convenient time scale (5 min. per measurement). The errors in t,he temperature measurements were dominated by those in pressure. Errors in determining second orifice temperature and in the calibration procedures were small. Spatial resolution with this probe was an important source of error due t o the large size orifice necessary. The positional uncertainty was 0.0075 cm., while the uncertainty in temperature measurement was 1%. There was good agreement between the measured maximum flame temperature (2015’K.) and t’he calculated adiabatic flame temperature (2000°K.). Thermocouple.-Flame temperature profiles were most precisely measured by thermocouple traversing techniques (see Fig. 2).a In this work, the thermocouples were made from small Pt and Pt-lO% Rh wires,21mounted on a micromcter traversing device, and the position of the bead determined with a precision cathetometer. Three major problems were associated with these measurements. The first was catalytic reaction on the wire surface; the second was the problem of radiation lossesfrom the thermocouple; and the third was the disturbance of the flame by the wake of the thermocouple. In the case of the methane-oxygen flame, the problem of catalysis o n the platinum surface was eliminated by “flame plating” the couple with silica.Z2 This coating had the advantage of‘ not only stopping catalysis but also inhibiting the evaporation of the platinum and prolonging thermocouple life in the flame. It had one principal disadvantage; it increased the size and emissivity of the thermocouple so that the effective emissivity had t o be determined by calibration. (20) P. L. Blacksnear, Jr., American Society of Mechanical Engineers paper No. 52-SA-38, Spril (1952). (21) Two wire diameters were used, 0.002 and 0.001 om. The beads were twice the wire diameter and the silica coating was 0.0003 om. thick. (22) W. E. Kaskan, “Sixth Symposium on Combustion,” Reinhold Publ. Carp., New York, N. Y.. 1957, p. 134.
Vol. 64
At temperatures above 1000”K., the radiation loss from these thermocouples was measurable. The laws governing this heat loss are well known.ea For a long wire in B slow gas stream, i t is proportional to the fourth power of temperature. The correction of thermocouple readings is straightforward3 except for the evaluation of the emissivity-area constant. Emissivity is known with moderate precision for bright platinum but values given for the (SiOz) coatinge2 were not used because they depend somewhat on the conditions of deposition. The emissivity-area constant was determined by calibrating each thermocouple under conditions of use a t the maximum temperature of the flame which was assumed to be the adiabatic value calculated thermodynamically. This is generally considered t o be a good approximation. Its validity for our flame was substantiated by the agreement between the calculated value and temperatures measured by the pneumatic probe and particle track measurements (which do not involve the radiation correction) and the thermocouple measurements of run 2. This run employed a small, uncoated thermocouple, whose radiation correction was small (about 50OK.) and could be estimated with reasonable precision. This procedure for estimating radiation corrections is selfconsistent, and although corrections as high as 300OK. were necessary, the average deviation of the four runs from the average temperature curve given in Fig. 2 was 13°K. The distortion of the flame due to the wake of the thermocouple was small, of the order of a few wire diameters. If the thermocouple wires are small compared with the flame front thickness, as was the case in these measurements, then to a first approximation this would result in a shift of the measured profile by a constant distance. The thermocouple would measure a temperature corresponding to a position slightly downstream of the measured position. Unfortunately, although this was a small absolute shift, it was not inappreciable compared with the measurements in the flame. The procedure for determining this shift was discussed in the section on position measurements. The assumption that the effect was primarily a simple shift rather than a distortion, was made reasonable by the agreement between the methods of temperature measurement and the general agreements between the temperature and composition curves. The precision of these measurements was 13OK. (see Table 111), and i t was felt that the absolute accuracy was 25°K. These errors probably were dominated by positional reproducibility (0.001 cm.). However, errors due to change in rhodium content in the wire from the drawing process or preferential evaporation in the wire might account for as much as 10°K. error. Composition Measurements in Flames.-The final piece of information necessary to describe the flame front is the profile of local concentrations. Composition measurements in this flame were obtained by sampling the flame gases with a quartz microprobe and analyzing them with a mass spectrometer (CEC-21-610 and CEC-21-620). These so the present techniques have been given elsewhere4~8~Q discussion will cover, principally, improvements in the techniques. Sampling was accomplished using a tapered quartz probe with an orifice tip. Careful control of the orifice shape was necessary to assure rapid decompression of the sample and quenching of the reaction. The techniques for accomplishing this have been given.9 In the present work, the only difficulties encountered were in initial runs where the probe was found to be badly shaped and of too small diameter (4 p ) . These preliminary runs were discarded. iill of the reported work was done with larger, carefully tapered probes whose diameters ranged from 15 to 40 p . Continuous flow sampling was found t o be superior t o the batch sampling techniques reported previously since adsorption problems, which are particularly serious for H20, were completely eliminated. By using a Teflon line inlet and a sapphire jewel inlet orifice to the spectrometer ionization chamber, the time taken for the flow system to reach equilibrium (ItO.lyo) was reduced to two minutes. This design eliminated metal surfaces and lubricants in the sampling system, and it was also possible to detect unstable species such as hydrogen peroxide and ozone. The sample was withdrawn a t a pressure around 100 M . The technique (23) M. Jakob. “Heat Transfer,” Val. I. John Wiley and Sons,Inc.. New York. N. Y., 1948, p. 23; Vol. 11, p. 147.
Oct., 1960
METHANE-OXYGEN FLAMESTRUCTURE
0.92
0.20
0.90
0.18
1391
MAJOR COMPONENTS
I INNER EDGE
OF LUMINOUS
0.88 ,i 0.16
'% r
*
0.86 ~ 0 . 1 4 0
0.
-0"z 0.84 o,* 0.12 0
E!
I-
2
r
0.82
OL
2 0.10 0
E!
LL
I
5 0.80 'sO.OS a 0
LL
E
LL
0.78 ,0.06 _I
0
0.76 E 0.04
0.74
0.72 - .-
0.02 0
0.1
0
0.2
0.3
0.35
0.4
0.6
0.8
1.0
Z (cm).
z
0
-
k'u
o a
3
MINOR COMPONENTS
0.004
0.003
E 0 0.002 UJ
t
I, 0.001 '3
0'" E
0.2
0.I
0
0.3
0.35
(cm). ERR3R CURVE FOR OXYGEN
2 0
12/22/58
0
12/24/50
I
0
o 6/16/59
H
I
I
I I-L I
I
I
I I
0.3 0.35 0.4 0.6 0.8 1.0 Z (cm), species (CHa, ,02, GO,,CO,, Hp0)-mole fraction us. distance (cm); B--Composi-
0 .I
Fig. 3A.-Composition profiles-major tion profiles--minor species (A, Hz, OCH2)-mole of three runti.
0.2
fraction us. distance (cm).
was to adjucit the probe, measuring the position of the tip with the cathetometer. -4t this fixed position, a complete analysis of the stable constituents of the flame gases was made. To be certain that equilibrium had been attained and that the apparatus was functioning properly, the runs were duplicated (each peak was reported during a run, and two runs Mere ma,de). If deviations greater than 0.5% of full-scale occurred, the run was repeated until successive runs agreed within this limit. Repetition was rarely necessary, and the average set of runs for a point could be obtained in 15-20 minutes. The three profiles taken on this system (each consisting of 25 complete analyses and their associated calibrations) were each completed during a single d a y . This was an important point in maintaining precision since calibrations of the mass spectrometer were found to
C-Error
curve-oxygen
per cent. deviation
drift several per cent. over a period of days. Ovpr :t singlr day, however, an analytical precision of better than 1% could be maintained by calibrating before and after the set of runs. Calibration of the instrument was accomplished by introducing gas samples a t known pressures into the burner and recording the spectrometer output. Calibrations for water vapor and formaldehyde were less direct; here, a mixture of oxygen partially saturated with the vapor was passed through the burner. The total pressure was measured with the Zimmerli gage and the partial pressure of the oxygen was determined using a previous calibration and the partial pressure of the vapor deduced as the difference between the total pressure and the oxygen partial pressure. In the rase of water, these calibrations could be compared with the results obtained in the high ternlieratitre equi-
It. hl. FRISTROM, C. G ~ ~ U N F E L AND D E RS. FAVIN
1392
librium region of the flame where thc water concentration could be calculated. The two methods of determining the spectrometer sensitivity to water agreed to within 2%. The accuracy of the formaldehyde calibration was only 10-2070, but fortunately this is a quantitatively unimportant constituent. These measurements allowed the determination of local concentrations of all of the constituents of this flame except the reactive atoms and free radicals. It is presumed that these latter species do not represent a quantitatively important fraction of the species being measured. Analysis of our data from the standpoint of matter consc~rvationsubstantiates this (see atom balance, paper 11). Sources of :Error.-Reading errors and crrors due to transient, disturbances were not important since each run was repested four times. The inherent stability and linearity of the instrument was a fraction of a per cent. for relativr sensitivii y over periods of 10-!.0 minutes. Thc reproducibility of analysw was l%,with a least count better than 10mole fraction. Errors in calibration were less than 2%, except in the case of formaldehyde where i t was 20%. Positional error3 ( 1 2 0 M ) are probably dominant in determining tlie reproducibility. Errors could occur because thc sample measured was not representative of the concentratlon a t the sam lingpoint. Such deviations could stem from four sources: biasing due to selective action of the probe, (2) non-linearities in averaging of the sample over the region of withdrawal, :3) adsorption of the sample in the transfer system and (4)inefficient quenching of the sample with resulting reaction in the probe. These questions have been discussed in sonic detail in the literature.6~8.9 It is believed that these sources of error do not exceed 2%, provided the following conditions are met: The reaction rates have half lives exceeding ten microseconds, the radical roncentrstions do not exceed a fraction of a per cent., and a continuous flow Teflon-lined analytical system is used.
8)
Data With these techniques, it was possible to measure the important local properties in the flame froiit. This information was collected as a set of profiles which give these intensive properties as a function of distance (Z coordinate in the flame). These curves represent a complete description of this flame excluding only the composition of the free radical and atomic species. A number of duplicate runs were made since it was desirable to establish the applicability of a one-dimensional model to this system and to establish the reproducibility of the techniques. The bulk of the data makes it impractical tc' present them all directly. Therefore, the most reliable run of each type of data is presented directly and the other runs are given in the form of an error curve for oxygen composition and error tables for the other variables. The error curve and tables indicate the precision of the data and demonstrate the one-dimensional character of the flamc sincc the off-axis runs show no significant trend.
Vol. 64
Two types of aerodynamic information were obtained-the area ratio and velocity. Area ratio is the ratio of streamtube area at any point Z to the streamtube area a t the coordinate origin (the screen). The values are recorded directly in Fig. 1 which gives the points measured on a number of streamtubes on two pictures. These data cover streamlines out as far as 1 cm. from the axis. The velocity measurements were taken from two pictures and cover the central region of the flame ( f1em.). Temperature information was derived using both thermocouples and a pneumatic probe. Four separate complete runs were made on this flame using several different thermocouples. Thermocouple diameter, coating and positions of the measurement were varied in order to assess the effects of aerodynamic wake, radiation losses and deviations from one dimensionality. The curve of Fig. 2 is an average taken a t every 50°K. interval from the experimental data interpolating linearly between adjacent points. Iluii number 2 was made with a small (0.0012 cm. dia.) uncoated couple so that the radiation correction could be estimated and the maximum temperature of the flame determined. The maximum value (1952OIi.i when corrected for radiation was identical with the calculated adiabatic flame temperature (2005 i15'K. experimental us. 2000°K. calculated). The pneumatic probe data consist of a single run made with a probe similar in geometry but somewhat larger in size (75 p ) than those used in the composition studies. This study confirmed the presumption that the flame closely approached the calculated adiabatic temperature. It followed the same pattern as the thermocouple temperature but had somewhat lower precision and resolution. Because of interaction with the screen, data were taken only beyond 0.1 cm. The composition data consist of three runs. The first run was made with a 30-p probe on the axis. The 21-610 mass spectrometer was used for aiialysis of the gases. The second run was made with a 15-p probe, with samples taken 1 cm. off axis. The 21-610 mass spectrometer was used for analysis. The third run TYas made with a 20-p probe on axis using a 21-620 mass spectrometer. This last run was considered to provide the most reliable and complete data and is presented in Fig. 3 A, B. Data from the other runs are presented in the form of error curyes of Fig. 3C and error Table 11.