OXIDATION OF LUBRICATING OILS IRON AND COPPER AS CATALYSTS I N THE OXIDATION OF HYDROCARBON LUBRICATING OILS BY J. H. T. BROOK AND J. B. MATTHEWS Received 26th January, 1951 The kinetics of the catalyzed oxidation of a lubricating oil in the liquid phase have been studied, using the rate of absorption of oxygen as a measure of the reaction velocity. The most important catalysts which come in contact with oils used for the lubrication of internal combustion engine crankcases being iron, copper and " crankcase catalyst ", i t is shown that the catalytic activity of all these is probably due to the dissolved iron and copper compounds formed by corrosion. A chain reaction mechanism for the catalyzed oxidation is proposed, and it is found that dissolved copper is a good chain-initiation catalyst, while dissolved iron is a good chain-branching catalyst. With increasing concentra- tions of dissolved copper the overall rate of oxidation reaches a maximum, and i t is suggested that this is due to micelIe formation of the dissolved copper soap.J.H. T. BROOK AND J. B. MATTHEWS 299 Whilst many catalyzed oxidation tests have been devised for pre- dicting the relative performance of hydrocarbon oils used for the lubrication of the crankcases of internal combustion engines, only Larsen and co-workers have made any systematic examination of the properties of the catalysts which are most likely to be encountered in the engine. Thus, Larsen, Thorpe and Armfield have shown that lubricating oils are mixtures of hydrocarbons stabilized by weak natural inhibitors, and Larsen and Armfield * have investigated the behaviour of several hydro- carbon oils under a wide range of conditions using lead, iron and corJper as catalysts, without, however, analyzing their results kinetically. Larsen, Armfield and Whitneya reached the conclusion that the oil- insoluble material present in used engine oils, which in the dry state was described by them as “ crankcase catalyst ”, was the most active catalyst present under service operating conditions.In addition George and Robertson 4, have investigated the catalyzed oxidation of some hydro- carbons, with special emphasis on tetralin, and presented reaction schemes in general agreement with those proposed by BollandB for the oxidation of olefins. The investigation described now was undertaken in order to examine in more detail certain parts of the field which are important with respect t o lubricating oils, and in particular to determine whether the simple kinetics found applicable to the reactions of pure hydrocarbons could be applied to complex mixtures such as lubricating oils.In addition to crankcase catalyst, metallic iron and copper are generally accepted as being the most important catalysts for the oxidation of crank- case lubricating oils in the field. It is now shown that all three catalysts owe their activity to the production of dissolved iron and copper com- pounds. Semenov’s concept of the degenerate chain-branching reaction has been applied to describe the kinetics of the liquid-phase oxidation of a typical lubricating oil in the presence of dissolved copper and iron soaps, and of these catalysts the former is found to be more effective for the chain-initiation step than for the chain-branching steps, whereas the reverse holds for the latter.Experimental Apparatus.-A circulatory oxygen absorption apparatus (Fig. I) of the type described by Dornte 7 was employed for all the measurements except when an inert solvent was used. The apparatus differed from Dornte’s in that glass valves were used in the circulating pump, a capillary flow meter was in- serted to measure oxygen flow rates, larger reaction vessels were used holding IOO g. of sample, the condensers did not return volatiles to the flask, a solid CO, cooled trap was provided, and the absorber train consisted of tubes filled with calcium chloride, activated charcoal, Hopcalite and soda asbestos.Auto- matic recording of oxygen uptake was provided. A static oxidation apparatus was employed for the measurements involving an inert solvent. This apparatus 8 consisted of a IOO ml. conical flask, con- nected to a gas burette and manometer, and gave the same induction periods for oxidation inhibited oils as the circulatory apparatus. Oil samples were degassed and saturated with oxygen. Five g. samples of oil were weighed into the flask and the apparatus was then flushed with oxygen prior to the com- mencement of each experiment. Whilst it could not be shown that a maximum oxidation rate existed in the circulatory oxygen absorption apparatus, the rates obtained with reaction vessels of varying dimensions and gas diffuser pore-size were constant within 1 Larsen, Thorpe and Armfield, Ind. Eng.Chem., 1942, 34, 183. Larsen and Armfield, I n d . Eng. Chem., 1943, 35, 581. Larsen, Armfield and Whitney, SOC. Auto Eng. J., 1943, 51(T), 310, 4 George and Robertson, Trans. Faraday SOC., 1946, 42, 217. 5 George and Robertson, J . Inst. Petrol., 1946, 32, 382. 6 Bolland, Proc. Roy. SOC. A , 1946, 186, 218. 7 Dornte, Ind. Eng. Chem., 1936, 28, 26. Reaven, Irving and Thompson, J . Inst. Petrol, 1951, 37, 25300 OXIDATION O F LUBRICATING OILS the limits of experimental error. However, the rate of oxidation depended slightly on the ratio of the oxygen flow rate to the sample weight. This was taken to indicate that the oxygen flow rate was not rate determining, but that variations in the degree of removal of reaction products affected the rate slightly.Similarly, attention had to be given to the absorption train, and the combination used was found to give the highest reaction rates. The repeatability of the measurements was variable, usually being about f 5 yo and about f 3 yo for the uninhibited and inhibited reactions respectively. n COLD TRAP FIG. I .-Circulatory oxygen absorption apparatus. Materials .-OIL.-A typical solvent refined paraffinic lubricating oil Carbon group analysis DILUENT.-B.D.H. diphenyl, twice recrystallized from 60-80 petroleum INHIBITOR.-The calcium phenate derivative of the iso-octyl phenol-formal- CATALYSTS.-Cupric stearate, recrystallized, (cu 9.9 yo : theory 10 yo). Ferric stearate, recrystallized, (Fe 7-7 yo : FeSt, requires Fe 6-18 yo, and Copper wire, soft commercial, 18 S.W.G.Reduced iron powder (specific surface of the order of 7000 sq. cm./g.). Crankcase catalyst, as described by Larsen, Armfield and W h i t n e ~ , ~ essenti- ally the oil-insoluble material which accumulates in internal combustion engine crankcase oils. r ] l o ~ ~ ~ 121.2 cs., pk5 0.879, sulphur content 0-42 % w. (Leendertse Q) : aromaticity 4.9 yo, naphthenicity 27-1 yo, paraffinicity 68 yo. spirit. dehyde resinlo hence this material is probably a basic stearate). Results Solid Catalysts.-The rates of oxidation of the oil catalyzed by copper metal (Fig. 2) proved t o be insensitive to the surface area of the catalyst over the surface area range 0.25 crn.,/g.oil to I cm.2/g. oil, and copper was found to corrode during the oxidation to an appreciable extent, giving, for example, 50 p.p.m. of dissolved copper after the absorption of 20 ml. of oxygen per g. oil, with I cm., of copper exposed per g. oil. Assuming that the amount of corrosion is proportional to the area of the catalyst surface and the amount of oxygen absorbed, it was found possible to correlate the results obtained with copper metal with results obtained using dissolved copper in the form of the stearate. Thus, from the above corrosion data one would expect to find, with I cm.2 of copper surface per g . oil, that the dissolved copper contents after absorption of I ml. O,/g. were 2-5 p.p.m., of z ml. O,/g., 5 p.p.m., of 4 ml. O,/g., 10 p.p.m., etc. It may be assumed that over the oxygen absorption range, 1-5 ml. O,/g.to 3 ml. O,/g., the average dissolved copper content was 5 p.p.m., and from the measured oxidation rate for this concentration of dissolved copper, the time taken in covering this oxygen absorption range can be determined. These various time intervals corresponding to various oxygen absorption intervals may be summed, and the results obtained in this way are presented as plot points in Fig. 2, showing good agreement between observed and computed oxygen absorptions. In the inhibited reaction (M/5o inhibitor concentration), copper metal again corroded slightly to give dissolved copper. Difficulties in the exact analysis of about 5 p.p.m. of copper in oil prevented the determination of the corrosion rate, and therefore an exact comparison of data on induction periods obtained 9 Leendertse (in press).l o U.S. Pat. 2,280,419, granted to the TJnion Oil Co. of California.J. H. T. BROOK AND J. B. MATTHEWS 301 with solid and dissolved copper could not be made. However, the induction period with I cm.2 of copper per g. oil was 870 min., while the average dissolved copper concentration appeared to be of the order of 4-5 p.p.m. and the induction period found with 5 p.p.m. of dissolved copper added as copper stearate was 830 min. From this it follows that, when metallic copper is used, dissolved copper is probably the active catalyst in the inhibited reaction, as well as in the uninhibited reaction. With iron powder as catalyst, corrosion was again observed, together with a very high degree of autocatalysis, the reaction obeying Semenov’s equation l1 dV/dt = A exp ($t), where A is the initial rate of oxidation, r j is the net chain-branching coefficient and V is the volume of oxygen absorbed. The initial rate A was very erratic, and it was not possible to apply the same reasoning as in the case of the copper catalyst in order to determine whether the catalytic activity was entirely due t o dissolved iron.In the inhibited reaction, with MI50 concentration of inhibitor and I g. of iron, giving about 70 cm.2 of iron surface per g. oil, the induction period was over 1200 min., showing that iron surfaces must be very weak catalytically. FIG. 2.-Influence of metallic copper upon the rate of ab- sorption of oxygen a t 1 6 0 O C ; ( I ) 0.10 cm.2 of Cu per g.oil; ( 2 ) 0.25 cm.2 of Cu per g. oil; (3) 0.5 cm.2 of Cu per g. oil ; (4) 1-0 cm.2 of Cu per g. oil. 0 and represent calcul- ated absorptions for 0.5 cm.2 and 1.0 cm.2 Cu per g. oil respectively. D so 100 I50 200 TIME FOR ABSORPTION MINUTES Larsen’s “ crankcase catalyst ” gave results with the uninhibited oil closely resembling those obtained with ferric stearate, taking I yo of the “ crankcase catalyst ” as being equivalent to 20 p.p.m. of dissolved iron. The high degree of dispersion of this catalyst in oxidized oils prevented accurate measurement of their dissolved iron contents, but the values obtained were of the right order of magnitude to account for the catalytic activity of this material. Since no soluble iron could be extracted from the catalyst by oils under non-oxidizing conditions, i t is probable that a corrosion mechanism is involved in the mode of action of this catalyst also. Soluble Catalysts .-Uninhibited Reactions.-Copper stearate exerts a maximum catalytic effect in this oil at about 20 p.p.m.of copper. It is possible that the reaction velocities are proportional to some simple power of the catalyst concentration a t low catalyst concentrations (of the order of magnitude of 2 p.p.m. or less), but the results a t higher catalyst concentrations (data a t 150’C are given in Fig. 3) are not easily interpreted. The apparent “ saturation ” of the inhibited oil by soluble copper with regard to catalytic activity affords an explanation for the copper metal catalyzed reaction being insensitive to the catalyst surface area, when the ratio of surface area to quantity of oil is greater than 0-25 cm.2 per g.oil. The oxidation rate curves do not fit Semenov’s equation for a reaction accelerating to a constant velocity, and are partially described by the parameters presented in Table I. l1 Sernenov, Chemical Kinetics and Chain Reactions (Clarendon Press, Oxford, 1935).302 OXIDATION OF LUBRICATING OILS The initial oxidation rates are not easily measured in the circulatory ap- paratus, since there are temperature changes in various parts of the gas flow lines on starting the oxygen flow. The reactions in the presence of dissolved iron obey Semenov's equation Catalyst Initial Rate Concentration ml.-~.~roo g. min. p.p.m. Cu 5 2'4 1 0 4'0 20 6-7 40 8.0 particularly well a t the higher catalyst concentrations as illustrated in Fig.4 where log V is plotted against time, but a t low catalyst concentrations the re- action rates fall away from the theoretical values as the reactions proceed. I500 - 0 d -- ItKQ E: 2 a 500 f 2 0" E 0 0 50 I00 IS0 200 TIME FOR ABSORPTION MINUTES FIG. 3.-Influence of dissolved cupric stearate on the rate of absorption of oxygen a t 150O C. Values for A , the initial rate of oxidation, fit an equation of the form A = k[Fe] + A , , and are thus not directly proportional to the catalyst concentration. On the other hand the values of 6, a more reliably determined constant, are approxim- ately proportional to the iron concentration (Fig. 5). and their temperature- dependence, given in Fig.6, leads t o an activation energy of 36 kcal./mole for the chain-blanching reaction. TABLE I Steady State Rate ml. 02/Ioo g. min. 5'6 8.25 10-5 9'1 The addition of soluble copper to the soluble iron catalyzed reaction gives a highly autocatalytic reaction with higher rates than are obtained with iron alone. Some of these rates are thought to be too high to be measured in the apparatus under the standard conditions without giving rise to a diffusion controlled re- action. Some values for A and 4 in Semenov's expression, where applicable, are given in Table 11. Inhibited Reactions .-In investigating the inhibited reaction, the in- hibitor was added a t a single concentration (M/50), and the measured induction period was taken to be the time required for an amount of free radicals to be produced equivalent to the amount of inhibitor originally present.The reciprocal of the induction period ~ / t , can be considered to be proportional t o the chain initiation rate. Experiments at low partial pressures of oxygen were conducted by filling the circulating system with appropriate oxygen +J. H. T. BROOK AND J. B. MATTHEWS 303 FIG. +-Influence of dis- solved ferric stearate on the rate of absorption of oxygen a t 150'C. 10 0 so 100 I so TIHE FOR ABSORPTION MINUTES FIG. 6.-Activation energy plot for chain-branching reaction in the oxidation catalyzed by 50 p.p.m. Fe as ferric stearate. FIG. 5.-Variation of 4 with concentration of dissolved ferric stearate a t 150' C. 0 and A represent values obtained with different samples of oil.-00235 00240 00245 00250 IIT304 OXIDATION O F LUBRICATING OILS nitrogen mixtures, and inserting a length of capillary tubing between the burette (which must necessarily be full of oxygen) and the circulating system to prevent mixing of the gases. As all the available inert diluents are in- conveniently volatile a t I 50' C for measurements in the circulatory apparatus, the static oxidation apparatus, in which volatility is less important, was used to obtain the dependence of the induction period upon hydrocarbon concentra- tion, diphenyl being chosen as the diluent. As will be seen from Fig. 7, 8 and g , the data for the copper catalyzed reaction can be represented by the expression, w = k [catalyst] [OJ* [RHIa, in which w is the chain-initiation rate and k is a constant, provided that the copper concentration is not greater than 30 p.p.m.The activation energy plots Simple kinetics were found for the reactions. ~~ 0 10 20 ppm Cu 30 49 so 20 40 ppm Fe 60 80 I I#) FIG. 7.-Influence of cupric and ferric stearates on the rate of chain initiation a t 150' C in the presence of M/50 concentration of inhibitor. are not good (Fig. IO), but lead to a value of about 30 kcal./mole for the activa- tion energy of the chain-initiation reaction. The chain-initiation rates are additive. For example, the value of w for 20 p.p.m. of copper in this system is 0.0038 min.-1, and for 50 p.p.m. of iron is 0.0018 min.-l. Experimentally, the value obtained from 20 p.p.m. Cu + 50 p.p.m. Fe is 0.0057 min.-l, in good agreement with the value for the sum of the separate rates, which is 0.0056 min.-l.TABLE I1 Fe p.p.m. I00 50 5 0 20 20 I 0 20 20 I 0 Cu p.p.m. A ml./roo g. min. 2.I* 1'3 1-5 * 1'1 * 0.8 4'2 2.4 1.0 1 2'1 4 min.-1 0.162 * 0.064 0.087 * 0.030 0.032 * 0.018 0.061 0.052 0.0 I 8 * Values obtained on a second sample of oil.J. H. T. BROOK AND J. B. MATTHEWS 305 FIG. 8.-Influence of partial goo01 pressure !of oxygen on the I rate of chain initiation in the 2 copper- and iron-catalyzed oxidations a t 15oO C in the presence of M/50 concen- tration of inhibitor ; ( r ) 20 p.p.m. of Cu as cupric stearate ; ( 2 ) 50 p.p.m. of Fe as ferric stearate. 0 05 I 0 F2 ( p o l PRESSURE, ATMOSPHERES) FIG. 10.-Activation energy plot for the chain-initiation reaction in the presence of 20 p.p.m.Cu as cupric stearate and 50 p.p.m. Fe as ferric stearate. FIG. 9.-Variation of chain- initiation reaction velocity a t 150' C with concentra- tion of oil in diphenyl. Catalyst: 20 p.p.m. Cu as cupric stearate. I 0 00235 ,00240 a I IT 45306 OXIDATION OF LUBRICATING OILS Discussion A simple reaction scheme may be postulated to represent the oxidation as a chain reaction with degenerate branching, including the following steps, in which the symbols have the usual significance and X represents a molecule of natural inhibitor.1 catalyst R H - R- W l - (1) R-+O2 -+ R0,- - (2) R- + X -+ inert products k4 - - (4) R 0 , - + R H --f R O O H + R - } kp : - (3) catalyst ROOH -* R- (or RO-) ROOH + inert products k6 * - (6) It is of little importance in the development of the final equations whether the radical in (4) is R- or R0,--, whether the radical produced in ( 5 ) is R- or RO--, etc.Following Semenov, and regarding the concentra- tions of radicals as being small and in equilibrium with each other, we can Put Or, since d [R-]/dt = W, - k4 [R-] + k5 [ROOH] * 0. dV/dt = kp [R-1, then dV/dt = (wl + K s [ROOH]) kp/k4. . . * (1) Now, - (ks + k6) [ROOHI. * a ( 2 ) d [ROOH]/dt = W ~ V + { K ~ ( v - I ) - k6) [ROOH] . * ( 3 ) Also v, the chain length, may be written for K,/k4. d[ROOH] dV dt dt Substitution in (2) of the value for dV/dt from eqn. (I) gives Integration of (3), with the condition that [ROOH] = o at t = o gives --- {exp [k5(v - I) - k6]t - I}. Wl V [ROOH] = K 5 ( v - I) - k ; Substituting for [ROOH] in eqn.(I) shows that This equation can be simplified under the two following conditions. Firstly, if and v the chain length is large then - I) k 6 , dV/dt = A exp ($t), . - ( 5 ) in which and $ = k,v. A = w1v = initial (unbranched) rate of oxidation, Secondly, when k d v - I) < K 6 , then dV/dt = A + B{I - exp (- $t)}, . ( 6 ) where and 4 = k6 - K ~ ( v - I). Eqn. (6) represents a reaction accelerating to a constant velocity. The oxidation of the uninhibited oil catalyzed by the higher concentra- tion of iron, and by iron and copper together, follows eqn. ( 5 ) . The experimental values for A (Table 11) may not in fact be values of wp, since they do not show the same dependence upon the catalyst concentra- tion as do the values for w, obtained from measurements on the inhibitedJ.H. T. BROOK AND J. B. MATTHEWS 307 system (Fig. 7), v being, according to the reaction scheme, independent of catalyst concentration and reaction rate. This d8ficulty in identifying A with w, is in all probability due to the nature of the oil. Thus, traces of unstable compounds or possibly of naturally occurring strong in- hibitors in the oil, might displace the reaction velocity curves in Fig. 4 along the time axis sufficiently to prevent this identification, since the values of A are derived from the intercepts of these curves on the volume axis. The chain-branching coe'fficient 4 which is given by the slopes of the curves in Fig. 4 is not affected by such considerations, and, since from Fig. 5 the value of 4 is proportional to [Fe], it may be identified with k,v.It will be noted from the few results in Table I1 that the addition of copper to iron in general increases A much more than 4. On a quali- tative basis, therefore, the data lead to the conclusion that copper is a good chain-initiation catalyst but a less effective catalyst for chain branching, Additional evidence in support of this conclusion is provided by a comparison of the shapes of the reaction velocity curves for the iron catalyzed and copper catalyzed reactions, since the former show a high degree of autocatalysis whereas the latter do not. The only example in the present work of a reaction closely following eqn. (6) is the oxidation catalyzed by 5 p.p.m. of iron, the data for which are given in Fig. 4. With copper catalysts, the reactions accelerate to a constant velocity without, however, obeying eqn.(6), as a result possibly of the value of the term {K5(u - I) - K,} being such that the conditions imposed in deriving eqn. (5) and (6) are not valid. The maxima in the rates of reactions catalyzed by copper stearate have been commented on before,2 and were thought by George and Robertson 4 to be due to the catalyst participating in a chain-termination reaction, mainly on the grounds that the addition of copper stearate reduces the rate of oxidation of long chain paraffins and tetralin catalyzed by cobalt stearate and ferric stearate respectively. However, as shown in Table 11, the addition of copper increases the already rapid rates of the iron catalyzed oxidations. Furthermore, even in the presence of a strong inhibitor, a maximum in the rate of oxidation is reached at ap- proximately the same concentration of copper as found in the uninhibited reaction. For these reasons it appears to be necessary to seek an alter- native explanation to that advanced by George and Robertson for the " saturation " effect observed with copper. The molecular weights of copper and ferric stearates have recently been measured ebullioscopically in cyclohexane in these laboratories, and were found to be 5000 for copper stearate in the concentration range zoo p.p.m. Cu to 800 p.p.m. Cu, and 1600 for ferric stearate in the concentration range IOO p.p.m. Fe t o 400 p.p.m. Fe. At these concentrations, therefore, copper stearate is highly associ- ated and ferric stearate is present largely as a dimer. Hence the possi- bility suggests itself that the maxima in the oxidation rates obtained with copper are due to micelle fortnation of the copper soap. This may be expected to occur at different copper concentrations in different systems. The authors wish to thank Mr. J. L. Dawson and Mr. D. K. Gilmour for experimental assistance and the Shell Refining and Marketing Co. Ltd., for permission to publish. Thornton Research Centre, P.O. Box No. I, Chester.