Spontaneous Ignition Properties - American Chemical Society

HE investigation of the spontaneous ignition properties of. T fuels and pure compounds has occupied the attention of many researchers in a number of d...
2 downloads 0 Views 665KB Size
Spontaneous Ignition Properties

of Fuels and Hydrocarbons J. ENOCH JOHNSON, JOHN W. CRELLIN, AND HOMER W. CARHART Naval Research Laboratory, Washington 25, D . C.

T

HE investigation of the spontaneous ignition properties of

fuels and pure compounds has occupied the attention of many researchers in a number of different fields. This is shown b y the variety of experimental methods and techniques which have been employed in the study of spontaneous ignition such as the CFR cetane number rating engine; well-instrumented single-cylinder motored engines (10, 16,19) in which single combustion cycles can be isolated; special devices such as the rapid compression machines (9, 18) and constant volume bombs (4, 16) which restrict themselves to the development of a fraction of a combustion cycle; and various combustion containers under both static and flow conditions ( I , S,6, 7,8, 11, 17, 18) in which the variables affecting ignition may be isolated more individually and in which the chemistry of the ignition and combustion processes may be studied more closely. All these methods provide valuable informrttion which can be used in the interpretations of ignition behavior, and it is noteworthy that the same particular phenomena associated with ignition and combustion are often apparent in one form or another in the results obtained with the various methods. As a n example, mention may be made of the occurrence of positive or "hot" ignitions under a given set of conditions, whereas under similar but less stringent conditions only "cool" flames occur. Another example is the appearance of "two-stage" combustion in which positive ignition, characterized by the release of considerable energy, is preceded by a very definite reaction that liberates a much smaller amount of energy and which may well be akin to cool flames. There are a t least two generalized mechanisms for the oxidative reactions which lead t o ignition of hydrocarbons. I n the low temperature mechanism, which predominates u p to temperatures of 350" to 400" C., the reactions proceed through free radical chain-branching processes involving the formation and decomposition of peroxides and hydroperoxides. The high temperature mechanism, which predominates above 350" to 400" C., depends at least partially on pyrolysis prior to and during oxidation. I n the present work interest is centered primarily on the low temperature mechanism since it is believed to be the more significant in the spontaneous ignition of fuels in a n engine. The present work was undertaken in order to further the knowledge of the relationship between fuel composition and other factors to ignition behavior. The type of experimentation used was selected because of its simplicity and speed and the fact that only small amounts of material are necessary. Yet many aspects of the results which are obtained can he correlated very well with studies made by other investigators using entirely different approaches and techniques. APPARATUS AND PROCEDURES

The apparatus (7, 8) used to obtain the data reported here consisted of a stainless steel block containing a n ignition chamber of about 21 ml. capacity with an opening at the top and two gas inlets near the bottom. The chamber had a small depression in the bottom designed to hold a shallow stainless steel crucible which could be replaced easily and which served as a receptacle for the fuel. The block was heated in a n electric furnace. I n operation, a n oxygen-nitrogen mixture of the desired concentration was. fed continuously into the chamber at 25 ml./min-

Ute, the block was heated to a high temperature and one drop of fuel was added to the crucible. The reaction which followed was then observed and recorded. Since it was easier to change temperature than oxygen concentration, the procedure was repeated a t progressively lower temperatures until the ignition pattern for a given oxygen level was established. The oxygen-nitrogen concentration was then changed and the process repeated until a delineation of the ignition zones could be obtained as shown in Figure 1. After each individual ignition test the chamber was >20

230

240

250

15

260

8

5

270

280

TEMPERATURE

2.5

290

300

310

320

330

("C.)

Figure 1. Portion of Ignition Curve for n-Octane Showing Individual Data Points Numbers at top indicate ignition lags in seconds

purged with air from above, and the crucible was replaced with a clean one. The recorded temperatures were those of the block although for many studies a thermocouple was suspended in the chamber itself to record gas temperature changes. At the lower temperatures, reproducibility was about 1 2 O C., but a t the higher temperatures the ignition boundary could not be defined as clearly. The n-hexadecane and 1-hexadecene were obtained from Humphrey-Wilkinson, Inc. The cyclohexane was Eastman Kodak white label grade. The octane was prepared at this laboratory by hydrogenolysis of n-octyl bromide with lithium aluminum hydride (6). All the hydrocarbons were purified b y percolation through silica gel, retaining a middle cut. The di-tert-butyl peroxide was obtained from the Shell Chemical Corp. and used without further treatment. The tert-butyl hydroperoxide was prepared by purification of Union Bay State Co. commercial tert-butyl hydroperoxide using the method of Milas and Surgenor (14). The FischerTropsch fuel and the 29-cetane fuel were obtained from the U. S. Naval Engineering Experiment Station, Annapolis, Md. The Fischer-Tropsch fuel is a synthetic fuel composed mainly of normal paraffins from about CS to C I ~ . The 29-cetane fuel is from a catalytically cracked stock and is rich in aromatics. Blend No. 2 is a mixture of the other two designed to give a 50cetane number fuel. 1749

INDUSTRIAL AND ENGINEERING CHEMISTRY

1750

Vol. 45, No. 8

IGNITION CURVES

IGNITION CURVES FOR PURE HYDROCARBONS

A complete ignition curve (of which Figure 1 was a part) is shown in Figure 2 and will be used as a prototype for discussion. The area of positive ignition (visible and/or audible combustion) lies above and to the right of the ignition curve. I n the area labeled “cool flames’’ a definite but much weaker type of reaction or ignition occurs which can be seen only in the dark or observed by means of thermocouple response. T o the left is the area of nonignition in which oxidative reactions still proceed but too slowly to culminate in ignition.

In order to shed more light on the ignition behavior of fuels, ignition curves for four pure hydrocarbons have been obtained. These are n-hexadecane (cetane), 1-hexadecene (cetene), aoctane, and cyclohexane. Hexadecane was chosen since it is one of the primary standards for the determination of cetane number of Diesel fuels; 1-hexadecene was chosen to show the effect of the double bond, the carbon chain being the same; octane to show the effect of chain length; and cyclohexane because it contains secondary carbons only and is cyclic. Figure 4 shows the curves for n-hesadecane and 1-hexadecene. These compounds have volatilities which fall in the Diesel fuel range and the similarity of the curves to those of whole fuels is apparent. However, there is one important difference. I-Hexadecene has a higher self-ignition point than n-hexadecane, yet it will give positive ignitions a t lower oxvgen concentrations, the curve being displaced to the right and downward. For the fuels studied thus far the displacement has been to the right and u p ward when related to decreasing cetane number, yet n-hexadecane has a cetane number of 100 and I-hexadecene 84, so that the generalization made for whole fuels no longer applies. The effect of introducing a double bond Seems to be that a greater amount of energy, as indicated by the temperature in this case, is required to initiate ignition but that the heat sensitive interme diates formed are more stable.

I 220

260

I

N

300

380 TEMPERATURE

340

420

460

500

(T.)

Figure 2. Prototype Ignition Curve (n-Octane)

The self-ignition point (SIP),is the lowest temperature at which a drop of fuel will ignite when pure oxygen is being supplied. It is a readily reproducible value characteristic of the fuel in a given ignition chamber. For many substances this temperature remains nearly constant for oxygen concentrations down to 50% and below. The ignition curve reverses itself at lower oxygen levels to give an inflection point a t ~14. As yet no good explanation can be given for this particular feature of the ignition curve, but since it appeals in a variety of fuels and pure hydrocarbons it is verv likely a reflection of a fundamental part of the ignition process. There does appear to be a relationship between the height of B and ease of ignition. A s may be seen from Figure 3, this inflection point is progressively lower as the cetane number increases; for n-hexadecane, ahich has a cetane number of 100, a rise from Jf t o B was not found a t all (Figure 4). -4s the temperature is raised past B , reactivitj again increases to a maximum value a t N . This region represents the greatest reactivity of the fuel to oxidation by the low temperature mechanism. As the temperature is increased further, this apparent reactivity decreases probzbly due to the increased thermal instability of certain heat-sensitive intermediates. A4ccordingly, more oxygen is required to furnish the necessary supply of these intermediates so that positive ignition may occur, ht higher temperatures, OP, the high temperature mechanism becomes effective so that again less oxygen is required for positive ignition. It is evident that the amount of oxygen required for the formation of cool flames is very small in that cool flames are observed even when pure nitrogen is fed into the chamber. Under these conditions the only oxygen available is that which enters the chamber from the air above by back diffusion or convection. The ignition curves given in Figure 3 are typical of those obtained with Diesel fuels. The data s h o r n are for three fuels of widely separated cetane number. As would be expected i t may be seen that as the cetane number of the fuel is decreased, the curves are displaced to higher temperature and oxygen levels. The ignition curves for three fuels of 50 cetane number (8), including one “doped” with a n ignition improver, are almost superimposable. The addition of a dope to a fuel displaces its ignition curve down and to the left in keeping with the increased cetane number.

I TEMPERATURE

(OC.1

Figure 3. Comparison of Ignition Curves for Fuels of Different Cetane Number

The ignition curves for n-octane and cyclohexane are given in Figure 5 . Data obtained by Maccormac and Townend (IS) for n-octane and by Townend (18) for cyclohexane using air a t various pressures in a closed container are included here for comparative purposes. The data in this case have been plotted on a basis of the partial pressure of oxygen, which in Townend’s work was regulated by changing the total pressure of air without changing the mole fraction and in the present mrork by changing mole fraction but not pressure. Considering the great difference in experimentation, the similarity in each set of curves is striking indeed. This similarity extends even to the first infleetion point (designated by iM in Figure 2). Particular attention is invited to the proximity of the temperatures a t the inflection points. Of added interest is the observation that the time lags for each hy-

August 1953

INDUSTRIAL AND ENGINEERING CHEMISTRY

drocarbon were reasonably close a t any given temperature. The curve of Burgoyne, Tang, and Newitt for cyclohexane (2') is similar in shape but displaced to the left. However, since the time lags are much greater, being of a different order of magnitude, a displacement in this direction is in keeping with the data shown here. 100

80 W a

?

8

60

b-

40 (L

a 20

0 200

240

Figure 4.

280

320 360 400 TEMPERATURE ("GI

440

480

Ignition Curves for n-Hexadecane and 1-Hexadecene

A comparison of the curves for n-octane and n-hexadecane in the low temperature region shows a displacement upwards and to the right with decreasing carbon chain length. This is in keeping with the generalization made for fuels since n-octane has a cetane number of 64. It is interesting to note that n-octane has the same self-ignition point as 1-hexadecene, but the rest of the curve for n-octane is displaced upward. It is obvious that cyclization makes ignition much more difficult, the curve for cyclohexane being displaced very far to the right. Since cyclohexane contains secondary carbons only, cyclization and the lack of primary carbon atoms must have a very pronounced effect on the oxidative processes leading to ignition. A complete elucidation of this is certainly to be desired. Other methods of measurement have shown that cyclization (a, 4, the incorporation of a double bond (4,18),and the branching and shortening of the carbon chain ( 3 , 4, 18) all decrease the ease of ignition, effects which are borne out in the present work.

1751

mediates are formed with more difficulty for the former type, nevertheless they must be more stable thermally. If peroxides are involved in these intermediates this is in keeping with their known thermal stability properties-namely, that primary peroxides are very unstable, whereas secondary and tertiary peroxides are increasingly more stable. A low cetane number fuel, having more branched and cyclic components, would be expected to give a higher proportion of tertiary peroxides than a fuel containing more normal paraffins. Therefore the intermediates farmed would be more stable. I n order t o test the concept that the stability of peroxides is related to the main features of the ignition curves for fuels and hydrocarbons, the ignition curves for a peroxide and hydroperoxide were determined as shown in Figure 6. The curve for di-tert-butyl peroxide will be considered first. Studies on the thermal decomposition of this peroxide have shown (14) that it decomposes rapidly a t temperatures around 200" C. and yields acetone and ethane exclusively as the pyrolytic products. The temperature coefficient of this reaction is fairly large. The peroxide is stable enough to permit positive ignition between about 185" to 240' C., and this reaction is quite independent of the oxygen concentration. This certainly indicates that a t higher temperatures the rate of thermal decomposition of the peroxide becomes so great that the slower chain branching oxidative processes no longer can yield the concentration of intermediates necessary for positive ignition. At temperatures above 500" C., positive ignition occurs again. This is in the high temperature mechanism zone and may be attributed to the ignition of the ethane and acetone formed, which ignite only in this range under these conditions. Although not as dramatic, similar results are evident with the tert-butyl hydroperoxide, which shows posiM8T

I

NRL

n- OCTANE

DISCUSSION

In the low temperature region it has been postulated (3, 11,WO) t h a t peroxides and hydroperoxides are among the f i s t oxidation products formed. These then decompose or react to initiate branched-chain reactions. Neither cool flame nor positive ignition can occur if the rate of quenching of these chains is greater than their rate of initiation and propagation. In view of the relatively long time lags before ignition, when contrasted to the short duration of the ignition itself, it appears that there is a gradual increase in concentration of intermediate oxidation products up t o a critical value. When this is reached, then either cool flame and/ or positive ignition ensues immediately. Under a given set of conditions the net increase in rate of formation of these intermediates must be greater for an easily ignitable substance like nhexadecane than for one like cyclohexane, or else the critical concentration is much lower. Also, in the vapor phase ( 3 )the rate of oxidation for normal paraffin hydrocarbons is greater than for branched or aromatic ones a t temperatures just below ignition. This would explain why higher cetane number fuels ignite more readily (see Figure 3) since, in general, they are more paraffinic. The ignition curves show that positive ignition can occur under conditions of higher temperature and lower oxygen partial pressure for a more difficultly ignitable (or lower cetane number) substance than for a more easily ignitable (or higher cetane number) substance. This would imply that although the necessary inter-

01

200

Figure 5.

240

280

320 360 400 TEMPERATURE ("C)

440

480

Ignition Curves for n-Octane and Cyclohexane

NRL-present works M & T-replotted from Maccorm a o and Townend (13); T-replotted from Townend (18)

tive ignition a t considerably higher temperatures and all the way across a t 100% oxygen. However, on pyrolysis ( I d ) , it yields methanol, formaldehyde, acetone, methane, tert-butyl alcohol, and tar. I n view of the variety of the decomposition products, it could hardly be expected to give as sharp a curve as that for the diperoxide. It is suggested, then, that the main inflection point in the ignition curves ( N in Figure 2) is related to peroxide stability. The

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

1752

shift with temperature for the various materials would be due to the different stabilities of the different peroxides formed. Of special significance in the present work are the studies on ignition lag or time delay, which is the time elapsing between addition of sample and ignition. The rapid exponential increase in lags with decreasing temperature, illustrated in Figure 7, is typical of the results obtained for all fuels and hydrocarbons studied at all oxygen concentrations. The ignition lags shown in the bottom part of Figure 7 were obtained with air for 1-hexadecene. I t was chosen here because a t this particular oxygen level there are several transitions from positive ignitions to cool flames and back, as may be seen in the ignition curve in the upper part of the figure.

160

220

Figure 6.

280 340 400 460 TEMPERATURE (“C.)

520

580

Ignition Curves for Peroxides

Particular attention is invited to the smooth continuity of the ignition lag curve regardless of the type of ignition. Furthermore, by reference to Figure 1, it is evident that although very dependent on temperature, the ignition lags are scarcely affected by changes in oxygen concentration. Also, it has been observed by other investigators (4, 10, l d , 17, 20) and in the present work that under special conditions a cool flame (or some related phenomenon) immediately precedes positive ignition, I n most instances this time interval is very short and, although this twostage process may occur generally, the cool flame usually was not observed because of the shortness of the time interval and the masking effect of the much more violent positive ignition. The following explanations are offered for the various observations given above. Similar concepts have been proposed by Walsh (20) and others. I n the low temperature mechanism zone the preignition oxidative reactions for a given substance are independent of the type of ignition that follows, whether positive or cool flame. These preignition reactions are constantly yielding intermediates, which for want of a better name, will be labeled X . These compounds, X,are continually being lost by chemical reaction, degradation, and physical removal. The formation of X requires relatively little oxygen, but the rate is very dependent on temperature. If the temperature is high enough, then the rate of formation of X will be greater than the loss of X, and eventually a critical concentration is reached and ignition occurs. This explains why a minimum temperature is necessary for ignition of a particular substance in a given apparatus; the almost perpendicular boundary between nonignition and both types of ignition and the very rapid continuous decrease in ignition lags with increasing temperature regardless of the type of ignition.

Vol. 45, No. 8

A second concept is that there are certain heat sensitive reactants, Y,part of which are dependent on the cool flame reaction, that are necessary intermediates for the reactions that result in positive ignition. I n addition, these must attain a critical concentration before positive ignition can occur. This implies that the first’reactions in ignition are the cool flame ones which, if they yield Y above the critical concentration, immediately will become positive ignition; otherwise, if the rate of destruction of Y is greater than its generation, the ignition remains a cool flame. These ignition reactions are generally very fast compared t o the preignition reactions, but the border line case exists where under certain conditions the cool flame can be observed immediately preceding positive ignition. Two-stage combustion may well be another manifestation of this same phenomenon. The sharpness of the boundaries betrveen positive ignitions and cool flames in the ignition curves shows that attainment of the critical concentration of Y a t a particular temperature is very dependent on the availability of oxygen. This is quite in contrast to the formation of intermediates, X . However, the formation of J’ is also somewhat dependent on temperature since less oxygen is required a t the point of great,est reactivity (IV in Figure 2). The rise in the positive ignition curve a t temperatures aboveN inFigure Zimplies that the reactants, Y,may well be peroxidic in nature-especially in view of the results shown in Figure 6-their tenipera.t,ure instability coefficient being greater tha.n the Coefficient of formation so that the composite gives the inflection a t :Y. A further piece of evidence for the generalizations given resulted from observation of temperature changes of the gases in the chamber following the addition of a drop of fuel. This was accomplished by means of a thermocouple suspended a short distance above the crucible in the bottom of the chamber and connected to a fast-responding electronic indicator. Under conditions of longer ignition lags the gas temperat’uredropped several degrees because of the cooling effect of evaporation. The temperature would then rise slowly to slightly above block temperature owing to the small amount of heat liberated by the preignition reactions (the formation of compounds X mentioned pre-

z

I

E

I

l

1

380

400

~

I-”

i? 022s

240

260

280

300 320 340 TEMPERATURE ( T i

360

420

Figure 7. Continuity of Ignition Lags for Cool Flames and Positive Ignitions for 1-Ilexadecene

viously). ilt the moment of ignition. either positive or cool flame, there was a very sudden temperature rise or “kick” due to the Iarge amount of heat liberated (the time at which X reached the critical concentration). The use of a thermocouple in the gas phase to measure t h e temperature kick has been very convenient experimentally in that not only can cool flames be identified without darkening the room completely, but their ignition lags can also be obtained, especially in the low temperature region.

August 1953

INDUSTRIAL AND ENGINEERING CHEMISTRY LITERATURE CITED

( 1 ) Barusch, M. R . , and Payhe, J. Q., IND.ENO.CHEM.,43, 2329 (1951); Barusch, M. R., Crandall, H. W., Neu, J. T., Payne, J. Q., and Thomas, J. R., Ibid., 43, 2761 (1951). (2) Burgoyne, J. H., Tang, T. L., and Newitt, D. M., Proc. Roy. Soc. (London),1 7 4 , 3 7 9 (1940). (3) Cullis, C. F., Hinshelwood, C. N., Mulcahy, M. F. R., and Partington, R. G., Discussions Faraday Soc., No. 2 , 1 1 1 (1947). (4) Hurn, R. W., and Smith, H. M., IND.ENG.CHEM.,43, 2788 (1951). (5) Jentzsch, H., Brennstoff-Chem.,31, 255 (1950). ( 6 ) Johnson, J. E., Blizzard, R, H,, and Carhart, H. w., J . Am. Chem. SOC.,7 0 , 3 6 6 4 (1948). (7) Johnson, J. E., Crellin, J. W., and Carhart, H. W., IND.ENG. CHEM.,4 4 , 1 6 1 2 (1952). ( 8 ) johnson, j. E., cre1lin, J. and carhart, H. w., ~~~~l R ~ search Laboratory, Rept. No. 3859 (Dee. 21,1951). (9) Jost, W., “Third Symposium on Combustion, Flame, and Explosion Phenomena,” pp. 424-32, Baltimore, The Williams and Wilkins Co., 1949. (10) Levedahl, W. J., and Howard, F. L., IND.ENG.CHEM.,4 3 , 2 8 0 5 (1951). (11) Lewis, B., and von Elbe, G., “Combustion, Flames, and Explosions of Gases,” New York, Academic Press, Inc., 1951.

w,,

1753

(12) Livengood, J. C., and Leary, W. A,, IND.ENG.CHEM.,43, 2797 (1951). (13) Maccormac, M., and Townend, D. T. A,, J . Chem. Soc., 1938, p. 238. (14) Milas, N. A,, and Surgenor, D. M., J . Am. Chem. Soc., 68, 205 (1946). (15) Pastell, 0. L., S. A . E. Quart. Trans., 4 , 571 (1950). (16) Petrov, A. D., “Dependence of the Antiknock Properties and

Pour Points of Diesel Fuel Hydrocarbons upon Their Structure,” Translation 649, UOP survey of foreign petroleum literature (1946). (17) Spence, K., and Townend, D. T. A., ‘‘ThirdSymposium on Combustion, Flame, and Explosion Phenomena,” pp. 404-14, Baltimore, The Williams and Wilkins Co., 1949. (18) Townend,D. T.A., Chem. Revs.,21, 259 (1937). (19) WalCutk C., and Rifkin, E. B.,I N D . ENG. CHEM.7 439 2844 (1951). (20) Walsh, A. D., Trans. Faraday SOC., 4 3 , 3 0 5 (1947). RECEIVED for review March 27, 1953. ACCEPTED May 20, 1953. Presented before the Division of Petroleum Chemistry at the 122nd Meeting, AMERICAN CHEMICAL SOCIETY,Atlantic City, N. J. The opinions and assertions contained in this article are the private ones of the authors and are not to be construed as reflecting the views of the Navy Department or the Naval establishment at large.

Spontaneous Ignition of Lubricating Oils CHARLES E. FRANK’, ANGUS U. BLACBHAM2, AND DONALD E. SWARTS Applied Science Research Laboratory, University of Cincinnati, Cincinnati 21, Ohio

T

HIS paper summarizes a continuation of the study on spontaneous ignition phenomena previously reported (7).

Such information is of particular importance because ignition of lubricants by hot metal surfaces is considered to contribute materially to the initiation of fires in aircraft. Earlier work was concerned with the spontaneous ignition temperatures of pure compounds and the factors influencing these values. The present paper describes the investigation of various types of lubricants and potential lubricants, and gives further observations regarding the effects of structure, additives, and metal surfaces on spontaneous ignition temperatures (8).

slight shift in the fuel-air ratio may bring about a relatively large shift in the observed ignition temperature. Accordingly, various charge sizes were tested at several air-flow rates to obtain as nearly as possible optimum conditions for ignition. I n this manner, good correlation between structure and spontaneous ignition temperature was obtained with pure compounds. This

APPARATUS AND PROCEDURE

The same crucible apparatus ( 7 ) was employed in the present investigation. This comprised a stainless steel block heated in an electric furnace, and containing a cavity into which ignition chambers prepared from various metals and alloys could be inserted; a stainless steel chamber was employed unless otherwise indicated. The procedure with the simple, volatile compounds involved heating the block to well above the ignition temperature then adding the material dropwise to the ignition chamber as the block slowly cooled. Throughout this earlier work, the size of charge was varied over a wide range to obtain the minimum spontaneous ignition temperature. The reason for this is apparent from a consideration of the general ignition curve shown in Figure 1. This curve represents the change in spontaneous ignition temperature with change in fuel-air ratio at a single pressure. With spontaneous ignition determinations which employ a single charge size for different compounds, it is difficult to obtain comparable results, as a 1 Present address, Research Division, National Distillers Products Corp., Cincinnati, Ohio. 2 Present address, Brigham Young Univeraity, Provo, Utah

1

w E

3

t

a

a W

n

I w

I-

FUEL/AIR ’

RATIO-

Figure 1. Typical Ignition Curve (1)