June, 1959
INFRARED STUDIES OF PROPELLANT FLAMES
94 1
INFRARED STUDIES OF PROPELLANT FLAMES BY HENRYA . BENTA N D BRYCECRAWFORD, ,JR. School of Chemistry, University of Minnesota, Minneapolis 1 4 , Minnesota Received February 1 7 , 1969
Application of a fast-scanning infrared spectrometer to the study of the pressure-supported flame of a solid double-base propellant is described. In the region from 700 t o 4000 cm.-l the flame spectrum in emission contains twelve characteristic bands. With but one exception, these bands correspond to the prominent bands in the absorption spectrum of the quenched products and have been assigned to COZ, CO, NZO, H20, NO, NO2, CBHz and HCN. A Rice-Herzfeld-type free-radical mechanism appears capable of accounting for the principal features of the thermal decomposition of both nitroglycerine and nitrocellulose. The direct detection of NO2 is described, and the peculiar role of nitric oxide in nitrate ester flames is pointed out. Some suggestions are forwarded concerning a perplexing, broad and intense absorption a t 2600-2700 cm. -1.
fall far short of the requirements of a convenIntroduction The combustion of double-base propellants has tional spectrometer. R,ather than attempting to been under investigation a t Minnesota since 1912. slow down the flame it was decided to speed up the Early experimental studies were concerned, in part, spectrometer. To this end a fast-scanning infrawith the determination of the rate of those chemical red-recording spectrometer mas built in 1951. Earprocesses by which a solid propellant, once ignited, lier reports have described the basic instrument6 is converted spontaneously into (!argely) gaseous and its modifications for studies of gas-phase reacproducts, without the addition of any oxidizing tions7-10 and for preliminary studies on propellant agent. This over-all kinetic constant, the burning flames.’ltl2 The present report is concerned with rate, is one of the most important properties of a recent work with this fast-scanning spectrometer on propellant, and a simple, convenient method was the infrared spectra of propellant flames in the devised for direct measurement of linear burning region 700-4000 cm.-l. mtes.l The mechanism of burning of double-base Experimental propellants and the dependence of the burning rate Since propellants seldom burn with a self-sustaining flame on pressure and compositioii have been considered a t pressures below 50 p.s.i., it was necessary to construct a by Crawford, Huggett and McBrady.2 A physical special high-pressure cell (Fig. 1). This cell was designed to withstand stat.ic pressures up to 500 p.s.i.; the usual operatheory based on a three stage flame-foam-fizz ting pressure, in runs with nitrogen as the pressurizing and model was developed by Parr and Cra~vford.~sweep gas was 100 p.s.i. (gauge pressure). The main frame These theories have been reviewed by G e ~ k l e r As . ~ of the cell was constructed from t.hree-quarter inch steel stock, a logical extension of this work, recent experimental grooved to t8akea lead “0”-ring gasket. The a/!’‘ thick and rear plates were bolted to the frame with thirty-two work in this Laboratory has centered attention on front tempered steel Allen-head bolts. Three or four days were infrared studies of the propellant flame. One of the usually required to draw down the cover plates to a tight fit obvious advantages of this technique is that one after the cell had been dismantled for cleaning or modificsmay look at the flame without sensibly disturbing t,ions. The cell has two sets of windows: one set of crystal winit, and in principle one should be able not only to dows, usually KRr, for the infrared beam of the spectromidentify reactants, intermediates a i d products, but eter or emission spectra of the flame itself; and one set of also to measure their concentration and (effective) Plesiglass windows for visual or photographic observation. An important feature of this cell was provision for contemperature throughout the flame. tinuous removal of the combustion products to prevent their For flames that can be maintained accurately sta- recirculsting back into the flame. A t elevated pressures tionary for long periods of time, the infrared spec- this requires a relatively large flow of nitrogen or other sweep trum could be obtained with a conventional spec- gas during a run. A large inlet a t the end below t,he strand was connected to two commercial nitrogen cylinders; at the trometer in about 20 minutes.5 However, in the other end of the cell, an exit orifice of adjustable diameter led investigation of the hurning of solid propellants, it directly t.0 an exhaust system with provision for sampling the would be difficult to maintain a flame stationary for comhustion products. Baffle screens at B (Fig. 1) served to 20 minutes because (i) the flame is not steady smooth out the flow. :In this way a linear flow of gas matched to the natural velocity of the flame products (apenough, (ii) burning rates vary considerably from proximately 1 meter/second) could be maintained for the powder to powder and even along the length of a duration of a run (approximfitely 20-30 seconds). There given powder strand, and (iii) it is not practicable to is also provision for mechanically advancing the strand to extrude or handle the lengths of powder strands slow the motion of the flame with respect to the spectromthat would be necessary. One might hope to slow eter. In early use,’* this cell lay bolted horizontally on the main down the rate of motion of the flame with respect to the spectrometer by, say, a factor of 10 by move(6) P. J. Wheatley, E. R . Vincent, D. I,. Rotenberg and G. R. ment of the propellant strand, but this mould still Cowan, J . O p . SOC.A m . , 41, G G 5 (1951). (1) B. L. Crawford, Jr., C. Huggett, F. Daniels and R. E. Wilfong, Anal. Chem., 19, 630 (1947). JOUR(2) B. L. Crawford, Jr., C. Huggett and J. J. McBrady, THIS N A L , 64, 854 (1950). (3) R. G. Parr and B. L. Crawford, Jr., ibid., 54, 929 (1950). (4) R. D. Geclder, “Selected Combustion Problems,” AGARD, Butterworths, London, 1954, p. 289. (5) (a) D. P. Needham and J. Powling, PTOC.R o y . Soe. ( L o n d o n ) , 5 3 2 4 337 (1955); (b) D. A. Dows. E. Whittle and G . C. Pirnentel, J . Cham. Phys., 23, 499 (1955); (c) R. E. Donovan and W. A. Agnew, $bid., 23, 1692 (1955).
(7) G. R. Cowan, E. Vincent and B. Crawford, J r . , ibid., 43, 710 (1953). (8) R. E. Nightingale, G. R. Cowan and B. L. Crawford, Jr., J . Chem. Phyq., 21, 1398 (1953). (9) G. R. Cowan, D. L. Rotenberg, A. Downie, B. L. Crawford, J r . , and R. A. Ogg, Jr., ibid., 21, 1397 (1953). (IO) I. C. Hisatsune, A. P. iMcBale, R . E. Nightingale, D. L. Rotenberg and Bryce Crawford, Jr., ibid., 23, 2467 (1055). (11) H. F. White, G. R. Cowan, D. Rotenberg and B. Crawford. Jr., ibid.,21, 1399 (1953). (12) A. D. Dickson, B. L. Crawford and D. L. Rotenberg, Ind. Eng Chem., 48, 759 (195G).
942
HENRYA. BENTAND BRYCECRAWFORD, JR.
I
II
I+---~~~---.~C-~~L.-.+I
g~~----,Jc-9~.-+
Fig. 1.-High pressure cell: A, strand holder and feed; B, baffle screens; C, crystal windows; D, glass windows.
100
50
CINE-KODAK
150
2 DO
FRAME COUNTER
DIAL.
O 100pal 4/32'
40-
/
36-
/
/
r:
100 psi 8/32'
41
0
I
4
8
I2
16
20
24
I
I
28
32
J
PER CENT OXYGEN I N NITROGEN SWEEP GAS.
Fig. 3.-Propellant
burning rate us. per cent. oxygen in nitrogen sweep gas.
frame of the spectrometer; naturally, in this position the flame had a tendency to lift. For this reason, a second optical system WBS constructed and mounted at right angles to the first. This system permitted vertical mounting of the propellant strand, with the image of the flame twisted through 90" and focussed across the slit of the monochromator. The spectra displayed in the figures have been obtained with this arrangement. In a typical run, a piece of propellant measuring 4" X 1" X 3/s" was cut from standard stock on a band saw, and drilled at one end to accommodate an igniter tip and a t the other to fit onto the strand holder (Fig. 1 ) . This propellant, mounted on the strand holder, was advanced until the igniter tip pressed against a wire that could bt, heat8edelectrically. The
Vol. 63
strand in final position extended approximately one inch ahove the space normally traversed by the beam, thereby blocking the beam. This provided time for the crater-like ignition mea t o develop into a flat fl%mefront hefore the spectrometer saw the flame. Finally, the front, plexiglass observation windows were bolted into place and the prewsurizing sweep gas startted through the cell. When the cell pressure was steady, the igniter was energized and shortly thereafter the scope camera was started. For emission runs, a blank would usually be run before and immediately after each burn. When desired, the entire optical path outside the flame could he swept free of atmospheric carbon dioxide and water with dry nitrogen; however, these atmospheric bands were sometimes retained as convenient calibration pointci. The amplification ot the signal from the Golag a.nd the smmning speed of the Littrow were checked each time with a standard 60-cycle signal. The appearance of a pattern on the scope was influenced by several factors: the cell pressure, the slit opening, the scope amplification, the camera speed, the scanning speed, the prism used, the drum setting and the scanning amplitude. The higher the cell pressure, the brighter the flame in emission; thus the spectral resolution and signal-to-noise ratio could be increased by raising the pressure; however, this compressed the characteristic zones in the flame and diminished the geometrical resolution. The greater the scanning speed, the less the difficulties on any individual scan with random fluctuations in the flame, and the better the geometrical but the poorer the spectral resolution. As a compromise, most emission spectra were run a t 100 p.s.i. (gauge pressure) a t a scanning speed of 15 scans/second. In view of the instrumental factors peculiar tjo the fast scanner we calibrated each scanned spectral region directly against known standards, such as polystyrene, atmospheric water and carbon dioxide, or specific calibrating gases (chiefly NO, Not, C ~ Hand Z HCN). The spectra of the individual oxides of nitrogen have heen published previously.13 The general procedure followed in surveying the spectrum of the propellant flame was to scan the spectral region covered by each prism in two series of overlapping scans: first an initial survey with the Littrow mirror oscillating with a large amplitude (wide scan), t.hen a more detailed series of overlapping scans of smaller scanning amplitude (narrow scan), with special emphasis in the second series on the promising regions of the spectrum. For example, the calcium fluoride region (1400-5000 cm.-l) was scanned twice, in a series of 6 overlapping wide scans and a series of 26 overlapping narrow scans, with special emphasis on the region around 2600-2700 cm.-l just above the GO? band at 2350 (Figs. 1 1 , 12). The region from about 1850 to 5000 cm.? also was scanned in a series of 11 medium scans using a LiF nrism. Likewise, the sodium chloride region (670-1400 ern.?) was sc,anned twice, first in a series of 6 overlapping wide scans and later with a series of 14 overlapping narrow scnnw. Several special high-resolution scans were made in regions of special interest, as when looking specifically a t NO, NOp, HCN or the separation of the P and R branches of Cot. The procedure in the exhaust collection runs was to hurn the propellant, under pressure with the sweep gas exhausting through liquid-nitrogen surrounded traps. The contents of the traps were then fractionated, suc.cessively, with a microcolumn loaned u s by Dr. John Overend, into storage bulbs, and from thence into an 8-cm. infrared gas cell equipped with silver chloride windows. Rewesentative spectra of the gases collected in this WayareshowninFigs. 15and 16. These were obtained with a Perkin-Elmer >lode1 21 douhle-beam spectrometer. The method b.v which the linear burning mt,e of the propellant was determined photographically is summarized in Fig. 2 . It was observed that the burning rate slowly increases with increasing oxygen content of the ambient atmosphere, and with decreasing exit-orifice diameter, unt.il some critical velocity, between 1.78 and 3.10 mm./sec., is reached, when, suddenly, the propellant begiris to hurn down the side a t B much increased rate (Fig. 3). These side burns have been fairly reproducible; can be sustained to low pressures; are quite sensitive to oxygen content, pressure and sweep-gas flow rate; and are quite brilliant. In separate experiments, it has been found that to a first approximation the linear flow (13) R . E. Nightingale, A. R . Downie, D . L. Rotenberg, Bryce Crawfold, Jr., and R. A . Ogg, Jr., THISJ O U R N A ~ ,118, 1047 (1954).
June, 1959
INFRARED STUDIESOF PROPELLANT FLAMES
rate of gas through the high pressure cell is nearly independerit of the cell pressure, depending principally on the exitorifice diameter.
‘
Results and Discussion With a CaF2prism, the infrared absorption spectrum from 1400 to 4000 cm.-l of propellant flames supported by an atmosphere of nitrogen a t 100 p.s.i. (gauge pressure) was obtained, in 12 overlapping scans, a t a scanning speed of 15 scans/secmd. While some moderately strong absorption bands were observedi1 particularly in the region of the COP 2350 cm.-l band, the spectra on the whole were disappointingly like the black-body radiation of the globar. On increasing the bomb pressure to 350 p.s.i., emission of infrared radiation from the flame itself began to predominate over the black-body emission of the globar, as evidenced by the fact that the oscilloscope trace \VAS deflected off scale when the luminous part of the flame reached the level of the beam. However, it was evident from the few scans obtained on these flames that, on the whole, more structure was to be observed from the flame in emission than in absorption; hence n.e turned our attention to the emission spectrum. All flame spectra reported here are of the flame in emission. Burning a t 100 p.s.i., the peak energy collected a t the detector from the flame of a strand 1” x 3/g’’ in cross-section was roughly one-quarter that from the globar, operating a t 4.6 amperes. These figures represent our first extensive report on the spectra of solid-propellant flames as viewed by the Minnesota fast-scanning infrared spectrometer. The d i d curves in the figures are traces of the photographic oscilloscope record. Oscilloscope deflection is plotted vertically, wave numbers scanned horizontally, increasing from left to right; the approximate scan center is indicated in the upper left,hand corner, together with the prism, cell pressure and spectrometer slits in microns. The solid curve labeled I represents the spectra obtained early in a run, from a zone close to the surface of the burning propellant; I1 and 111 indicate zones progressively further out in the flame. For comparison, the dashed curve labeled GB is the globar; for calibration purposes, the dotted curve labeled PS is the absorption spectrum of polystyrene, as recorded immediately after a run. These traces are quite characteristic and, for all their peculiar structure, quite reproducible. The first question raised by these traces is whether they portray peaks of emission or valleys of absorption-or both. In a chemically and physically similar system, the flame decomposition of ethyl nitrate, Needham and Powling6”find, in the main, emission peaks (for HzO, CH, and COz). Close examination of the reyion about the COz doublet a t 2350 cm.-l, has enabled us to resolve the COz doublet in absorption-precisely where one should ordinarily find it, except for a slightly increased semration of the P and R branches. Indeed, this C02 hand in “emission” appears in nearly every way like the C02 band in absorption reported previously. l 2 There are other clues; the flames are sooty and luminous; and the “spectrophotometric gradient’’ of the early traces (I) resembles rather closely that
943
RUN 131 LiF 5200cm- (~4000-10000) IOOpri c-\ # GB
\.
100- ISO/L I I
I ~4000 p: 5200 p 10000 Fig. 4.-Flame spectrum: black-body radiation peak.
of the black-body globar blank. Increasing the cell pressure incrcases both the visible luminosity of the flame and the intensity of the black-body radiation and, also, the separation of the P and R branches of C02. Furthermore, in both absorption and emission the optical density of the C 0 2 band was roughly independent of the cross-sectional dimensions of the strand. Thece facts suggest that what we are observing is a small sui1 of luminous soot particles radiating a black-body continuum, surrounded by a mantle, of increasing thickness as one progresses outward in the flame, of cooler infrared-absorbing gases. The outer envelope of flame gases is presumably quenched by the nitrogen sweep gas, which enters the bomb some 20 or 30 degrees below room temperature a t a Reynolds number in excess of 3200. The importance of diffusion into a flame from the surrounding gases recently has beeii emphasized by Gaydon and W01fhard.l~ Under even mild conditions, mixing can occur; in our cell conditions are quite turbulent and extended mixing must certainly occur. We have estimated the temperature of the soot particles responsible for the infrared continuum in this way. Where the envelope of quenched gases is optically thin, as a t the base of the flame, the spectrometer sees directly into the hot interior of the flame. The soot particles there are probably nearly i n temperature equilibrium with their surroundi n g ~although , ~ ~ the radiation from them may be not quite Planckian. Absolute intensity measurements on the spectrophotometric gradient of the flame black-body emission are, of course, meaningless. l 6 From the black-body radiation peak, however (Fig. 6, trace I), one estimates for the case of n spectrometer sampling equal wave length intervals16 that T = (max)/3.45 = 1200-1300° (for the inner cone of a flame burning in an atmosphere of T\’2 a t 100 p.s.i. and streaming by a t approximately 1 meter/sec.). This temperature is reasonable for our (14) A . G . Gaydon and H. C . Wolfhard, “Flames,” Chapman and Hall, Ltd., London, 1953. ( 1 5 ) G . H. Dieke, “Temperature,” Vol. 2, Reinhold Publ. Corp., New York, N. Y., 1955,Chapter 3 ( 1 6 ) M a x Planck. “The Theory of Heat Radiation,” € . Blakiston’s Son and Co.,Philadelphia, 1914.
914
HENRY A. BENTAND BRYCE CRAWFORD, JR.
For this average, we fiiid values ranging from 200800°, depending on the pressure and turbulence of the nitrogen sweep gas. These temperature estimates serve to confirm the picture of z1 hot core of solid particles, radiating a continuum, surrounded by an envelope of cooler gas. With this picture in mind, the traces have been examined from the point of view of absorption. The main “bands,” whose structures have been verified repeatedly under varying conditions, are listed in the first column of Table I by their designation in the figures, together with their approximate frequency and intensity. Our confidence in this interpretation of the over-all emission spectra of the flame as being that of essentially discrete absorption bands superimposed on a continuous blackbody background will depend in part on the extent to which logical assignments can be offered for these bands.
RUN 90
100 psi
7OP
RUN 85 CaF2 3400 cm-1 7OP 6/32“
TABLE I Obsd. bands (cm.-I)
RUN 132 Li F 4000cm-I (3200- 5200)
VOl. 63
A
Figures
Z(S) 5-6000 A(ms) 3700
5 5,6
B(m) 3300
5,6 7 8,9 81 9 9 10 10 11 11,12 11 13,14
C(m) 3085 U(VS) 2600-2700 E(s) 2350 F(w) 2100 G(m) 2000 H(m) 1860 I(vs) 1620 J(ms) 1500 K(w) 1600 region L(s) 700-1000
Assignments
Overtones; combinations 0-H (Hz0, OH, HONO) coz (211 213; 2v2 211) HCN; CZHz; N-H C-H ?
+
+
coz
CN (?) NO NO2 C-NOZ H20 HCN, C2Hz; HONO; CzHn
In trying to assign these bands, several problems are encountered immediately. First, there are factors involving the flame itself: whether it contains normal molecules, excited molecules, free radicals or some complex mixture of all these. Secondly, there are purely instrumental factors : the appearance of an absorption band as recorded by the fast scanner depends on many factors, and any similarity between the appearance of a band on the fast scanner with that obtained with an unmodified 12C spectrometer on slow scan is purely coincidental! The simplest and most direct approach to these two problems is to assume that the absorbing entities in the turbulent flame envelope are normal Fig.’5.-Flnme spectrum: 0-H stretching region; CO, com- molecules in thermodynamic equilibrium with their bination bands. Fig. 6.--Flame spectrum: hydrogen stretching region (LiF environment and t o compare where possible the spectrum of the flame directly with that of susprism). pected absorbing entity in an ordinary gas cell, flames, though possihly a bit on the high side.58 either just before or immediately following a run. We shall return to this point later, in our discussion Such comparisons were made for water vapor (Fig. of the mechanism of the flame reactions and the pos- l l ) , carbon dioxide (Fig. 9), nitric oxide, nitrogen sible role of surface catalysis. dioxide and hydrogen cyanide. The temperature of the outer mantle of the flame If the principal absorbing entities are stable has been estiinnted from the separation of the peaks molecules, it should be possible to collect them from qf the P and R branches of the 2350 COZ band.” the exhaust gases and to observe their spectra as The spectrometer naturally reports an average of quenched products a t room temperature. Figure I5 what it sees along its line of sight into the flame. shows the spectrum on a model 21 double-beam spectrometer of the volatile products from two (17) G . Herzberg, “Molecular SI)cctra. I,” D . Van Nostrand Co., inodes of decomposition: thermal decomposition in New York, N. Y., 1950, Chapter 3.
INFRARED STUDIESOF PROPELLANT FLAMES
June, 1959
945
t~acuo(the dashed line), and flame decomposition a t
114 p.s.i. (the solid line). The similarities are striking: in both instances one finds COz, CO and NzO, NO, NOz, HONO, probably HN03,C2Hz,HCN and C2H,. As we shall see later, most of these products are kiiietically quite plausible. The same species, less those involving HzO, are displayed in a fractionated sample in Fig. 16. Across the bottom of this figure we have indicated the bands seen in absorption in the flame spectra. (The vertical pips indicate the bands reported by Wolfram, st aL.,l* for the gaseous products from the slow coiitrolled thermal decomposition of iiitrocelluloee.) Again, the correspondence is striking. With one exception, all of the major flame bands correspoiid to promiiient bands in the quenched products. Perhaps this is not surprising: in the one case the quenching is with gaseous nitrogen a t approximately O o , in the other with liquid nitrogen. The two bands of COz a t approximately 3600 and 3700 are clearly seen in both instances (Figs. 15, 16; and 5, 6). Removal of the wster clearly reveals band B in the queiiched exhaust gases, and the relatively weak band C. Next is the relatively strong COz band a t 2350, with the CO-NzO structure on its lower wing. From here on down to 1400 the principal absorptions are due to NO and Ye2. Ns04,which is detected in the exhaust gases, would, of course, not be present at any appreciab e concmtration a t 200' above, and in the flame the region extending from approximately 1350 down to 1050 showed no strong absorption bands. The flame spectrum from approximately 1000 to 700 is one broad fairly intense band with a certain amount of fine structure. This region is nicely covered by GH4, HONO, CzH2 and HCN in that manner. The only intense flame band unaccounted for in this way is band D a t 2600-2700 cm. -I. From the optical densities of the non-overlapping bands in Figs. 15 and 16, one finds that carbon dioxide and nitric oxide are the principal constituents of the volatile exhaust gases. Exclusive of any nitrogen, hydrogen, oxygen or water that may be formed, these two gases together account for over 80% of the volatile, infrared active products. Table I1 gives the relative partial pressures of the identifiable products based on 100 for PNO.
RUN 145 LiF 2680 cm-1 100 psi 150p
TABLE I1 Sribstanee
CO? NO HCN C2Hz
Relative partial pressure
170 100 35
Substance
Relative partial pressure
NOz
8
CnH+
6 0.2
NZO
12
The carbon-to-nitrogen ratio for the gaseous mixture represetited by Table I1 comes out to 1.68, cornpiwed to a calculated value of approximately 1.70 for the propellant itself. Since neither carbon nor nitrogen are included i n the experimental count, this agreement is either fortuitous, or indicates that the relative aniount of soot and reduced nitrogen formed is small. (18) AI. L. Wolfroin, J. H. Frazer, L. P. Krihn, E. E. Dickey, S.hI. O h , R . S. Bower, G . Q . l l a h e r , ,J. I). J l u ~ d o c k ,A . Chaney and E. Carpenter, J. A m . Chem. Soc., 78, 4095 (I95G).
Fig. 7.-Flame spectrum: C-H stretching region. Fig. 8.-Flame spectrum: band D arid COI (band E).
From Table I1 it niay be seen that nitric oxide constitutes approximately one-third of the total pressure (exclusive of any K2, 0 2 , H? and HzO). Much the same conclusion is reached from an inspection of the traces. It is interesting to compare these results with those of Needhani and Powling on the flame decomposition of ethyl nitrate a t atmospheric presThey find the products listed in Table 11, together with metane (which we would not have detected a t the low level they report), hydrogen and nitrogen, nitromethane and the partially oxidized products methyl nitrite, acetaldehyde, formaldehyde and methyl alcohol. Most strikingly, whereas we fiiid much COZ, and little CO, t,hey
HENRYA. BENTAND BRYCE CRAWFORD, JR.
946
I
!
Vol. 63
The formation of HCN is interesting, because it is usually a s ~ u m e d that ~ ~ -this ~ ~ molecule is formed by the reaction of a methyl radical with nitric oxide CHa.
+ NO +CHaNO +CHFNOH
HCN
CoF2 2220 cm-1 IOOpsi
imp
CN
RUN 93 CoF2
,‘‘GB
;PS /
1910cm-1
,.’
‘e
I
:
.. 0.:
:
:
’.;
I
, /
)735
+ Hz = HCN + H
Fig. S.-Flame
spectrum: 2200 cm.-l region. spectrum: 1900 crn.-l region.
find mainly CO, with little COz, no NO2 and no soot. Also, we find relatively more HCN--indeed, several hundred millimeters of it in the flame under combustion conditioiis. The HCN formed in the flame is probably a synthetic product, since there are relatively few C-N bonds in the double-base propellant, and none a t all in either of the major components, nitroglycerine and nitrocellulose. Other synthetic products are NzO, C02and H20. CO, KO, NOz and C Z H ~on, the other hand, presumably are decomposition products.
(2)
which has an activation energy of approximately 7 kcaLZ8 This is not to suggest that CN, for which we have some slight evidence (band F, Fig. 19), abstracts hydrogen atoms in the flame from Hz,for which we have no direct evidence, but rather, since D(H-H) and D(C-H) are comparable, that CN if present (as ultraviolet emission studies would suggest14) could abstract hydrogen atoms from hydrocarbons with an activation energy of 7 kcal. or less and form free radicals still more easily. To facilitate comparisons of this sort, we have in Table 111 listed some representative activation energies for a number of elementary kinetic processee that might be thought to occur in a nitrate-ester flame. The elementary processes in Table I11 may be divided into two classes, with activation energies, respectively, greater or less than that for rupture of the weakest bond in a nitrate ester. This activation and energy is approximately 40 kcal./mole,19~29 correPponds to the bond dissociation energy of the RO-NOs bond. The first step in the thermal decomposition of a nitrate ester is, then, generally postulated as being the homolytic scission of this bond to give two free radicals, an alkoxy1 radical and nitrogen dioxide RO-N02 = RO.
Fig. 10.-Flame
(1)
Under cracking conditions a t 1000°, HCN has, in fact, been manufactured from NO and methane.24 It is known that CHzNOH decomposes a t moderate temperatures2j and that HCN is produced in the reaction of methyl radicals with N0.23s26HCN has been reported before in flame^,^^,*^ and by Wolfrom, et U Z . , ’ ~ in the slow controlled thermal decomposition of nitrocellulose. Our system, as previously noted, is flooded with nitric oxide; however, neither of the major components contaiii methyl groups, which may come from the additives, a possibility also considered by others.I8 Another possibility is suggested by the reaction
RUN 96
8Opsi 210p
+
+ HzO
+ NOz
E* = 40 kcal./mole
Fairly conclusive proof that NOz is in fact formed in some such step as this recently has been obtained in a study of the thermal decomposition of (19) E. W. R. Steacie, “Atomio and Free Radical Reactions,” Reinhold Publ. Corp., New York, N. Y., 1954. (20) A. F. Trotman-Dickenson, ”Gas Kinetics.” Academic Press, Inc., New York, N. Y., 1955. (21) H. A. Taylor and H. Bender, J . Chem. Phyc., 9, 7G1 (1941). (22) C. S. Coe and T. F. Doumanl. J . A m . Chem. Sac., T O , 1516 (1948). (23) W. A. Bryce and K. U. Ingold. J . Chem. P k y s . , 23, 1968 (1955). (24) h l . Patry and G. Engel. Camp. rend., 231, 1302 (1950). (25) G. K. Adams, W. G. Parker and H. G. Wolfhard, Dtsc. Faradau Soc., 14, 97 (1953). (26) P. Harteck, Ber.. 66, 423 (1933). (27) E. A. Arden and J. Powling “Sixth Symposinm (International) on Combustion,” Reinhold Publ. Corp., New York, N. Y . ,1957, p. 177. (28) Ref. 19, 1). 614. (29) P. Gray,-Trans. Faraday SOC.,51, 1367 (1955).
TABLE I11 ACTIVATION ENERGIES OF SOMEELEMENTARY REACTIONS
+
E*
Reaction
+ + +
+ + +
+
+
+
+
+
+ +
+
+
-
-t
RUN
aa
aF2 1555 cm-1
A -1
/ \
19, p. 646 32 20, p. 284 20, p. 296 19, p. 616 19, p. 616 19, p. 644 39a 39, 51 19, p. 609; 20, p. 306 32
+
+
+
+
+
+
+
+
Ref.
(kcal./mole)
CH, NO 0 C=C NO2 + C(NOz)-C 0 CzH6 CzHs 1 C=C H 4-5 C=C CH3 6-7 H HzCO 3.6 HzCO 5.6 CH, CN+Hz+HCN+H 7 RO. dismutating by c-C rupture 3-13 RO. dismutating by C-H rupture 12-25 CzHz NOz + glyoxal (over-all) 15 CHO+CO H 13-14 NOz -.+ HONO HzCO CHO 15-19 CHO.CHO N0z HONO CHO.CO 20 R ON0 RO NO 20 (H) CHI NOz -.+ HONO CH3 21 R. dismutating by c-c rupture 24 R. dismutating by C-H rupture 40 HCHO oxidation (350”) 21 NO2 CO COz f NO 28 2N02 + 2 N 0 Oz 25 D(R0-NO) 38 D(RO-NOz) 40 NO CO + Coz ‘/&z 50 D(R-ONO) 57 D(R-NOz) 58 63 2NO -+ NzO 0 D(NN-0) 55 HCHO + CO Hz (510-607 ”) 45 NO Hz 49 C?Hs + CZH4 70 D (C-C) 80-85 D(C-H), prim., sec, ter. 100,94,00 114 CzNz + 2CN
+
947
INFRARED STUDIES OF PROPELLANT FLAMES
June, 1059
+
20, p. 212 398
1450
32
RUN 129
20
NaCl 1334cm-I 100psi l50-270p
20 43 20 20 39a 19, p. 241 55 39a 39a 48 48 44 47 20 20 20 19, p. 645
.../
several nitrate esters in potassium halide matrices. Free NOz is known to react, quickly a t room temperature with alkali halides according to the equation30 2NO2
+ KX = KNOB+ NOX
(3)
The nitrate ion that is produced should be readily detectable in the infrared. Using this diagnostic test for NOz we have, in fact, found strong proof of the formation of NO2 in the controlled thermal decomposition of ethyl nitrate, nitrocellulose and the double-base propellant itself.31 The nitrogen dioxide may react in a host of ways. Gray and Y ~ f f have e ~ ~recently published a system(30) D. M. Yost and H. Russell, “Systematic Inorganic Chemistry,” Prentice-Hall, Inc.. New Yo& N. Y.,1946. (31) H . A. Bent and B. L. Crawford, Jr., J . Am. Chem. Soc., 79, 1783 (1957). (32) P. Gray and A. D. YofTe, Quart. Reus., I X , 362 (1955): also, Chem. Reus., 66, 1070 (1955).
Fig. 11.-Flame spectrum: C-NO2 region (CaF2 prism). Fig. 12.-Flaine spectrum: C-NO1 region (NaCl prism).
atic review of many of its reactions. Essentially, in one way or another, NOz loses oxygen to become NO, a radical that is unusually stable under flame condition^.^^^^^ This is indeed one of the great problems of nitrate ester combustion: the reduction of nitrogen beyond NO.33 However, we defer further consideration of these radicals until we have considered the probable fate of the alkoxy1 radical, and the products derived from it, some of which are known to react quite readily with NOz. Enough is known of bond dissociation energies and activntioii energies of simple gas phase reactions (cf. Table 111))to be a fairly detailed guide as to the probable mode of decomposition of a complex ester like nitroglycerine or nitrocellulose. We shall start, somewhat arbitrarily, with nitroglycerine. (33) P. Gray and M. W. T. Pratt, “Sixth Symposium (International) on Combustion,” Reinliold Publ. Corp., New York, N. Y.. 1957, p. 183.
HENRYA. BENTA N D BRYCE CRAWFORD, YR.
918
400- 6aOP
sitioii of the p a r a f f i ~ i s . ~ ~ According ~'~ to the RiceHerzfeld mechanism, and studies of alkoxyl radic a l ~ , one ~ ~may ~ expect ~ ~ - two ~ ~reactions to dominate the further decomposition of our nitrate ester radical (b): (i) hydrogen abstraction and (ii) radical dismutation. The activation energies for these two alternatives are not known with certainty (Table 111),but presumably both alternatives are important a t any temperature where the initial RO-KOZ bond scission is occurring at a significant rate. We may, however, ignore hydrogen abstraction for the following reasons. This absbraction process forms alcohols, which are known to react rapidly with KO2 to give nitrites and nitric a ~ i d ~ ~ ~ ~
A
-" w
GB
NaCl 780 cm-I 100psi
......
50op
:PS ,
spectruin: 900 em.-' region. Fig. 14.-Flame spectrum: 780 an.-1 region.
Fig. 13.-Flame
- KO2
I ONOa 0x02 ON02
0x09 0
1 I
I
I
ON02
('h)
H
H
ON02
(?
AI -bI -H+
H&--
According to the discussion above (reaction 3), the first step in the decomposition of this ester is probably scission of an 0-K bond, There are two possibilities H
iH2C----C---CHa t I
ONOz
HZC=O
In what follows, both of these alkoxyl radicals will yield essentially the same piqoducts; we shall coiisider only oiie of them, (b). An alkoxy1 radical is isoelectronic with an ordiirary alkyl radieal, the terminal CHs group of the LaLtter having beeii i.eplaoed by a terminal oxygen ritom with an unpaired electron. As such, it may be expected $0 undergo very much the same seq u p o e of reaotion as postutated by Rice and HemfeEd for the alkyl rndicnls in the thermal rlecoinpo-
I
I
(6)
ON02
The radical thus formed will be much less stable> than the parent molecule, for it can dismutate by. breaking a relatively weak 0-N bond H
(ih)
H
+ HZC--C. ON02
H
HAC--C--CH2 ON% I bNO,(!I
(5))
This reaction is, in fact, merely the alkyl analog ot' t,he reaction of water with NO2 to yield nitrous 'Gd and nitric acid, which occurs rapidly a t room temas a normal RO-NO bond is p e r a t ~ r e . Inasmuch ~~ kinetically equivalent to an RO-NO, bond (Table 111; see also ref. 29), the mixed nitrite-nitrate ester formed in this way is essentially equivalent,, kinetically, to the original nitrate ester. Given R source of readily abstractable hydrogen, such as an aldehyde (see later), this sequence merely changes nitrate groups in our original kinetic entity to nitrite groups. Eventually the intermediate radical will presumably find an opportunity to dismutate,, particularly since this latter alternative is presum-. ably less discriminating in the type of collision required (because of the higher steric factor). An alkoxy radical may decompose in either of two ways: (i) by splitting homolytically an alpha C-H bond, or (ii) by splitting an alpha C-C bond. Since C-C bonds are the weaker, the fatter alternative is the most likely, just as with alkyl radicals (Table 111, ref. 39a, b), where, instead of forming a C=C double bond, there is formed a carbon-oxygen double bond
,P"\
---+
+ 2N02 = RO.NO + HO.NO2
ROH
RUN 121
Hs H C--C-----CH. I I
Vol. 63
H2C-C.
I
I
--+ H2C--C