Energy & Fuels 1989,3,461-464
461
Insoluble Sediment Formation in Middle-Distillate Diesel Fuel: Evidence Concerning the Role of Fuel Acidity M. A. Wechter Department of Chemistry, Southeastern Massachusetts University, North Dartmouth, Massachusetts 02747
D. R. Hardy* Naval Research Laboratory, Department of the Navy, Code 6180, Washington, D.C. 20375-5000 Received February 15, 1989. Revised Manuscript Received April 27, 1989 This work focuses on attempts to link fuel acidic components to insoluble sediment formation during oxidative aging. Titratable acidic species naturally present or formed during accelerated aging for the fuels of this study could not be correlated with the amounts of fuel insolubles formed. The acids present in the fuels of this study appear not to be limiting reagents in the reactions leading to insolubles. Hence, we were not able to predict fuel oxidative instability by measuring acid strength or concentration. In addition, no evidence was found for any universal specific or general class of compounds that could correlate to a given fuel's stability.
Introduction The search for a plausible, verifiable mechanism for the formation of solids in middle distillate diesel fuel has now spanned over 40 years.'t2 Contributions from this laboratory over the past 6 years have focused on oxidative instability from catalytically cracked light cycle oils (LCO). Recently, the abilityof a simple methanol extraction of these LCO's to drastically increase their storage stability has been systematically ~ t u d i e d . This ~ work noted that the methanol extract should prove to be a convenient and simple chemical probe of the precursors of the insoluble sediments. Many different mechanisms or combinations of mechanisms have been proposed for fuel degradation product formation over the years2s4 None of these have actually been directly verified. This is due in part to the complex nature of the fuel matrix and in part to the failure of commonly used analytical techniques to isolate, as yet, the responsible fuel sediment precursors. In spite of the incredible progress made in analytical instrumentation and techniques over the last 40 years, no definitive compositional links have been proposed that are valid over a range of unstable diesel fuels. Moreover, until the recently proposed low-pressure reactor technique (LPR)5became available, obtaining accurate data by accelerated aging of samples was a lengthy process. The long aging periods greatly reduced the opportunities available to systematically study the effects of long-term storage on fuel characteristics and stability. Several workers6,' have proposed various individual "marker" compounds as an indication of the tendency of a particular fuel to produce insolubles. These markers
remain to be validated in a variety of fuels. This method suffers an additional drawback in that mechanistic proposals utilizing the markers generally fail to account for sufficient heteroatom content and the effectiveness of strong organic base inhibitors. Other workers"" have implicated the acidic components of the fuel as being mechanistically important either as reactants or through catalytic involvement. In order to further study this possibility, we used a methanol extraction of unstable fuels that spanned a wide range of total acid content (acid number) and acidity as measured by the initial emf reading on a glass electrode. The present investigation focused on attempts to validate the involvement of acidic species in insolubles formation using these fuels. This was particularly interesting since a detailed high-resolution capillary GC/MS analysis of the methanol extracts of these fuels revealed no compositional factors that could be correlated with insolubles production. In addition, a search for thermally labile reactant polar species in the extracts by reverse-phase HPLC did not meet with success.
(1) Smith, H. M.; et al. Bureau of Mines Report to Western Petroleum Refiners Association, Bartlesville, OK, 1954. (2) Sauer, R. W.; Weed, A. F.; Headington, C. G . Prepr.-Am. Chem. SOC.,Diu. Pet. Chem. 1958, 3, 95. (3) Wechter, M. A.; Hardy, D. R. Fuel Sci. Technol. Int. 1989, 7(4), 423-441. (4) Schrepfer, M. W.; Arnold, R. J.; Stansky,C. A. Oil Gas J. 1984,79. (5) Hardy, D. R.; Hazlett, R. N.; Beal, E. J.; Burnett, J. C. Energy Fuels 1989, 3, 2C-24. (6) Bhan, 0. K.; Tang, S. Y.; Brinkman, D. W.; Carley, B. Fuel 1988, 67, 227-237. (7) Pedley, J. F.; Hiley, R. W.; Hancock, R. A. Fuel 1988, 67, 1124-1 130.
(8) Pedley, J. F.; Hiley, R. W.; Hancock, R. A. Fuel 1987, 66, 1646-1651. (9) Hazlett, R. N. Fuel Sci. Technol. Int. 1988, 6(2), 185-208. (10)Beal, E. J.; Cooney, J. V.; Hazlett, R. N.; Morris, R. E.; Mushrush, G. W.; Beaver, B. B.; Hardy, D. R. DOE/BC/105245-4; Naval Research Laboratory: Washington, DC, 1982. (11) Westbrook, S. R.; Stavinoha, L. L.; Present, B. L.; Hsu, J. P.; Herrera, J. G . In Proceedings of the Third International Conference on Stability and Handling of Liquid Fuels; Institute of Petroleum: London, 1988; Vol. 2, pp 538-553. (12) Acid Number of Petroleum Products by Potentiometric Titration: ASTM D 664-87; American Society for Testing and Materials: Philadelphia, PA, 1988.
0887-0624/89/2503-0461$01.50/0
Experimental Section Fuels used in this study were obtained from several different refineries. They were selected for their relative tendencies to form insolubles and for variety in acid number as determined by ASTM D664 titration.'* Acid number, or total acid number (TAN), is expressed as milligrams of KOH per gram of sample titrated. Table I summarizes the fuels used and includes pertinent characteristics. The fuel samples included straight run stocks (SR), catalytically cracked light cycle oils (LCO), and blended fuels (B) of these two refinery streams. Blended fuels were, in general, 70130 (v/v) combinationsof straight runs and light cycle oils. One 60140 (v/v) blended fuel was used.
0 1989 American Chemical Society
462 Energy & Fuels, Vol. 3, No. 4, 1989
ident n0.O B-2 SR-3 LCO-3 B-3 SR-4 LCO-4 B-4 SR-5 B-5 SR-8 SR-9 B*-10
D86 90%, "C 317 290 311 309 302 329
Wechter and Hardy
Table I. Fuel and Blend Stock Properties viscos, D86 end acid density, wt % of CPat point, "C dmL sulfur 40 "C no. 345 0.25 0.852 0.3 2.8 0.02 0.02 339 0.03 0.841 0.2 2.0 0.25 0.02 348 0.20 0.885 0.5 2.8 0.73 324 0.28 0.866 0.2 330 0.02 0.849 0.4 2.3 352 0.08 0.848 0.6 3.0 0.02
flash point, "C 79.5
insolubles (43 "C for 18 weeks): me/100 mL 2.9 0.4 10.6 8.9 0.1 2.5 6.0
77 72
11.3 7.1 3.9 4.3
Key: SR = straight-run fuel; LCO = catalytically cracked light cycle oil; B = 70/30 (v/v) blend of SR-LCO; B* = 60/40 (v/v) blend of SR/LCO; 2-10 = refinery code. Blank spaces were not determined. *> 4 mg/100 mL is considered unstable.16 Samples were prefiltered through Gelman Type A / E filters immediately prior to any accelerated aging. Sample volumes for p H measurements were generally 12 mL of fuel in 50 mL of D664 solvent or 12 mL of methanol extract in 50 mL of D664 solvent. Long-term storage conditions were simulated by accelerated aging (stressing) of samples in a low-pressure reactor (LPR).6 Reactor temperature was maintained a t 90 "C in a n oven, and an overpressure of oxygen was maintained in the reactor a t 100 psig (690 kPa gauge). Test periods ranged between 24 and 168 h. Under these conditions, 24 h has a n effect on the samples that compares with 2-3 years of storage under ambient conditions. Sample volumes used for the aging experiments were 100 mL. They were contained in 125-mL borosilicate bottles covered with perforated aluminum foil. After being aged, samples were fiitered after their removal from the reactor, and the insoluble material was recovered and weighed by following a procedure reported by Frankenfeld, Taylor, and Brinkman13 and modified by Hazlett, Cooney, and Beal14 and Beal.15 The filtered fuels were recovered and retained for emf measurements. Filter pads were also retained for further study. All samples were stored away from light. I t has been reported elsewhere3 that extraction of fuel with methanol before stressing removes precursors that are responsible for insolubles formation. Some of the experimental procedures in this study employed methanol extraction. The emf measurements were made with a Mettler DG 112 combination electrode filled with a solution of LiCl and connected to a Mettler DL 20 Compact Titrator with automatic stirring. The titrator was used in the millivolt function mode and the system was calibrated with aqueous buffer solutions. Filtered fuel samples 12 mL in volume were diluted with 50 mL of D664 solvent for emf measurements. Methanol extra& of some of the fuel samples were also tested for initial emf values. For these measurements, a 20 mL sample of fuel would be extracted once with 12 mL of methanol. The methanol phases were decanted and diluted with 50 mL of D664 solvent and the emf values determined by using the Mettler system. The emf of methanol in D664 solvent was measured for purposes of acidity intercomparison between samples. Relative acidities of fuels and methanol extracts used in this study are reported throughout as emf values although TAN or D664 titration results were presented in some tables. The emf measurement combines acid strength and acid con~entration.~ Moreover, our work indicates that there is generally good directional agreement between emf and TAN or acid concentration measurements. All emf measurements made during this investigation used the Mettler titrator/electrode system. Reported TAN (total acid (13) Frankenfield, J. W.; Taylor, W. F.; Brinkman, D. W. Ind. Eng. Chem. Prod. Res. Deu. 1983,22, 608. (14) Hazlett, R. N.; Cooney, J. V.; Beal, E. J. DOE/BC/10526-16; Naval Research Laboratory: Washington, DC, 1987.
(15) Beal, E. J. Private communication, 1988. (16) Hardy, D. R.; Hazlett, R. N.; Gianinni, R.; Strucko, R. SAE Technical Paper Series No.860895; SAE: New York, 1986.
Table 11. Effect of MeOH Extraction on D664 Acid Number for High-Acid-Number Straight-Run Fuel SR-5" TAN, mg of TAN, mg of KOH/g of KOH/g of sample fuel sample fuel 1 0.73 3 1.14 2 1.75 4 0.33 OKey: (1) filtered fuel; (2) methanol phase from 1st extraction of 100-mL sample extracted with 15 mL of methanol; (3) methanol phase from 2nd extraction; (4) fuel SR-5 sample after two methanol extractions. Table 111. Relationship between Insolubles Formation and TAN for Blended Stocksa mg of insol/ TAN, mg of KOH/ emf*mV fuel 100 mL ident of fuel e of fuel mVI1) . , mV(2) ., B-2 1.4 0.25 39 103 B-4 6.8 0.2 45 135 B-3 9.4 0.03 8 87 B-5 11.4 0.28 71 103
-
a Key: mV(1) = emf of methanol extract before aging; mV(2) = emf of methanol extract after aging. Samples stressed for 24 h at 90 "C, 690 kPa of oxygen.
number) values were obtained by D664 titration12 using this system.
Results and Discussion P r e l i m i n a r y experimental work that centered on the reduction in insolubles formation in fuels which had been e x t r a c t e d with methanol3 yielded evidence that the fuel acidity, as determined b y TAN, decreased after methanol extraction. Table I1 presents the TAN results of an ext r a c t i o n series where a 100-mL sample of a high-acidnumber straight-run fuel was extracted with two successive 15-mL volumes of methanol. TAN determinations of each extraction phase were made b y D664 titration. As can be seen in Table 11, extraction w i t h methanol decreased the acid number of the original fuel. O t h e r investigator^^^^ have suggested a link between acid content of middistillate diesel fuels and their t e n d e n c i e s t o w a r d insolubles formation. The methanol extract has removed a significant fraction of acidic species, and t h i s extraction results i n stabilizing the r e m a i n i n g fuel. Extraction with methanol has been f o u n d to stabilize all fuels that would be marginally to very u n s t a b l e w h e n exposed to an oxygen overpressure of 690 k P a in the low pressure reactor ( L P R ) at 90 "C for 24 h (equivalent to 2-3 years of a m b i e n t storage). A fuel that would f o r m 5-6 m g of insoluble m a t e r i a l under these conditions would be cons i d e r e d u n s u i t a b l e for military use, although o r d i n a r y
Sediment Formation in Diesel Fuel Table IV. Effect of Acid Number on Sediment Formation in Fuel Blends and in Unblended Cracked and Straight-Run Stocksa TAN, mg of fuel mg of KOH/ sediment/100 ident g of fuel mL of fuel SR-3 0.02 0 LCO-3 0.02 21.7 B-3 0.03 11.3 SR-4 0.25 0 LCO-4 0.02 5.5 B-4 0.20 6.1 OSamples were stressed for 28 h at 90 "C, 690 kPa of oxygen.
commercial requirements would be somewhat less stringent. Table I11 summarizes the gravimetric results obtained for 70/30 (SR/LCO) blended stocks from four different refineries that were subjected to accelerated aging for 24 h in the low-pressure reactor. When these results are compared with the acid numbers of the same fuels there is no apparent correlation, Table IV makes the same comparison for straight-run/light cycle oil (70/30) blend samples from two different refiners. The outcomes are similar to those of Table 111. Table I11 also includes data on the emf of the methanol extracts (measured in D664 solvent) of the four fuel blends before (mV-1) and after (mV-2) aging. There appears to be a marginal correlation between the TAN results and emf values for the prestress samples. However, the emf values obtained for the aged samples show no correlation with TAN or with any of the other values. The (mV-2) values do show that this fraction has drastically increased in acidity, and this mirrors the usual finding that the acid number of a particular fuel increases with aging? However, the magnitude of this change has no relationship whatever to the tendency of the fuel to form sediment. Fuel B-3 was selected for additional acidity study, and samples were aged in the low-pressure reactor for periods ranging between 16 and 48 h. After filtration and extraction with methanol, the emf of the methanol extracts was measured in D664 solvent. It was found that the acidity tended to increase in a nonlinear fashion with increasing stress period. However, the increase in insolubles weight for these samples was linear with time. In order to further investigate the relationship between acid formation and insolubles formation during aging, two blended stocks were selected for systematic study. Fuel B-3 was a high-sedimentingfuel with a low (0.03) TAN and fuel B-2 was a very stable fuel with a considerably higher TAN (0.25). Prefiltered 100-mL samples of each were stressed at 90 O C and 690 kPa oxygen for periods of 1-7 days. The LPR timetable was as follows: Samples that were to be stressed for 4-7 days were placed in the LPR on day 1. The LPR was first vented on day 4 for removal of the 4-day stress sample and insertion of the samples for stress periods of 1-3 days. The LPR was then repressurized. Thereafter, the LPR was vented and repressurized once each day to allow for removal of appropriate sample pairs. Sample bottles were capped and stored in the dark until all seven were collected. The samples were then filtered to remove insolubles. The filtered insolubles and adherent gum fractions were combined, rinsed, dried, and weighed. The post-stress emf of each fuel sample was determined in two ways: 12 mL of the filtered fuels was diluted with 50 mL of D664 solvent and mV(f) measured; 20 mL of fuel was extracted with 12 mL of methanol and the emf of the methanol fraction was determined in D664 solvent mV(m). Table V presents the experimental results. It can be seen from the data presented that fuel B-3 is
Energy & Fuels, Vol. 3, No. 4, 1989 463 Table V. Relationship between Acidity and Insolubles Formation for Two Fuels under Accelerated Aging Conditions of 90 O C and 690 kPa of Oxygenn
0
mg of insolubles/ 100 mL of fuel Fuel B-2 0
1
2.2
2 3 4
days of stress
emf, mV mV(D mV(m) -11
6 7
3.6 4.5 5.8 7.4 8.7 10.2
55 87 141 236 265 290 306 324
0 1 2 3 4 5 6 7
Fuel B-3 0 7.0 16.7 21.0 35.3 44.5 48.2 53.5
10 50 45 87 220 268 290 306
-11
5
92 148 237 243 280 301 319
23 45 87 227 280 301 321
I, mV(0 and mV(m) are emf readings of the fuel sample and the methanol extracted sample, respectively.
considerably more unstable than is fuel B-2, yet from the TAN data before stressing (Table I) and emf results after stressing, B-2 is more acidic. The emf values of the two fuels are quite different until well into the experimental period, when they nearly converge. Further, the emf of the whole fuels in D664 solvent compare very favorably with those of the methanol extracts in the same solvent. This is further evidence that the sediment precursors and any important acidic components have been effectively partitioned into the methanol extracts. These methanol extracts were extensively analyzed by GC and GC/MS and reverse-phase HPLC. The methanol extracts were much simpler than the whole fuels and therefore much more amenable to complex compositional analysis. Even so, no indication cf directionally important compositional species was seen. Four representative LCO's (one stable, two marginal, and one unstable) were semiquantitatively analyzed for the following compounds or classes of compounds: fluorenes, phenalenes, carbazoles, indoles, and naphthalenes. The hydrocarbon species (fluorenes, phenalenes, and naphthalenes) were found in all four fuels in about the same ratio and absolute concentrations (in the methanol extracts) regardless of fuel stability. Nitrogencontaining species were missing in the most unstable fuel extract and present in the most stable extract. Sulfur compounds were not detected. The polar compounds from ten different fuels of varying stability including the four discussed above were isolated by methanol extraction and analyzed by high-resolution reverse-phase HPLC. Eight of the ten fuels showed good directional agreement of these polars with unstable product development. Of the two remaining fuels, one very unstable fuel contained very little polar material and one very stable fuel contained a great amount of polar material. If the emf data from either fuel B-2 or B-3 are plotted against stress time, the resultant resembles a titration curve with the breaks occurring between days 3 and 4 for fuel B-3 and between day 2 and 3 for fuel B-2. The results of both of these, together with the insolubles data for each fuel, are summarized in Figure 1. As can be seen the insolubles data are strikingly different for the two fuels whereas the acidity data are quite similar. Another point of interest is the very high acidities reached by day 6. This is equivalent to 12-18 years of ambient storage!
Wechter and Hardy
464 Energy & Fuels, Vol. 3, No. 4, 1989
i
3501 > 3001
1
8-3, mV
250 0
:200
3
/--A
/
1
150
8-2, mg x 4
-0
1
2
3
4
5
6
7
STRESS PERIOD, DAYS
Figure 1. LPR study of two fuels. Results were obtained at 90 OC,690 kPa of oxygen. Standard error for the insolubles weight by LPR was about &15%; standard error for the emf measurements was about *5%.
These data indicate that any acids formed well into the fuel oxidation are not participating as important reactants in insolubles production, otherwise the fuel insolubles production rates would also be higher than first order. The data also suggest that the hydrogen ion strength in the fuel necessary to invoke acid participation in insolubles production has been exceeded even before oxidation has begun. This means that a detailed analysis of fuel acids (even those present before those formed by oxidation) for concentration and strength should not be able to predict the relative tendency of a given fuel to form insolubles. This is indeed the case when one examines a wide range of fuels of varying instabilities and acidities. This would also explain the results of adding acids of known strengths and concentrations to base fuelsSgSince there is already an excess of acid in any unstable fuel, adding more should stimulate the insoluble production in a linear fashion, as has been found in studies of that type. Previous work3 revealed not only that a methanol extraction improved an extracted fuel's storage stability but also that the extract could be dissolved in a neutral solvent (dodecane/butylbenzene), which, when stressed, formed insoluble sediment. When the weights of total insolubles from the aged extract and from the aged extracted fuel were added together, a mass balance was achieved between those results and the weight of sediment obtained from the unextracted aged fuel. In one of the four fuels an excess mass was formed in the methanol extract in neutral solvent. Three additional fuels were subjected to this test regime. Two of the fuels were unstable although they contained less than 10% v/v of cracked cycle oil stocks and one fuel contained 40% v/v of cracked LCO. An additional methanol extract series was dissolved in a very stable straight-run fuel (SR-3) with an acid number of 0.02. The results of the experiment are given in Table VI and generally support the earlier work regarding mass balances of produced total insolubles. Fuel B-10 forms somewhat more sediment in the methanol extract, resulting in a slightly higher total mass. This means that the sediment precursors have indeed been effectively isolated in this methanol extract and possibly some natural inhibitors have not been partitioned into the methanol. Insolubles weights produced by the methanol extracts in the neutral solvent and in the SR-3 fuel are seen to be comparable. The initial acidities of the two diluents are quite different, yet there is no effect on insolubles formation. Conclusions In the results detailed above, it is seen that acidic fuel
Table VI. Yield of Insolubles (90 OC LPR and 690 kPa of Oxygen for 28 h) yield of insolubles, mg/100 mL of fuel MeOH- MeOH extract MeOH unextracted extracted in dodecanel extract sample control fuel butylbenzene in SR-3 5.7 3.0 3.0 2.4 SR-8 SR-9 2.1 1.3 0.9 1.0 B-10 5.9 2.0 8.7 6.5
components for the fuels in this work could not be correlated with the production of insolubles. This includes acidic fuel components initially present or those formed during oxidative aging. This is particularly evident when comparing the nonlinear rate of acid formation with the quite linear rate of insoluble sediment formation in the two fuels aged from 0 to 7 days. It is also evident from a lack of correlation between insolubles formed and emf and TAN data for a wide range of fuels. Clearly, however, acidity has been shown to be involved a t some step of the insolubles-forming process with most fuel^.^^^ This work, however, indicates that the acids present in the fuels of this study (even before any oxidation has begun) appear not to be the limiting reagent in reactions leading to insolubles. Hence any attempts to predict fuel oxidative instability by measuring in situ acid strength or concentration are not, in general, possible. A detailed analytical search for compositional links to diesel fuel sediment production in a wide range of fuels was attempted by using the simple methanol extract. Regardless of their oxidative storage stabilities, all the fuels contained homologues of fluorenes, phenalenes, carbazoles, indoles, and conjugated olefinic and aromatic species when analyzed by high-resolution GC/MS. No simple directional compositional correlation involving these types of fuel components on a quantitative basis was found. Analysis of polar fractions by reverse-phase HPLC indicated some reasonable compositional correlation, but some samples gave false positive results and some were found to give false negative results. Thus we found no evidence for any universal specific or general classes of compounds that could correlate to a given fuel's stability. This implies three possibilities: (a) synergistic/catalytic effects far outweigh specific fuel compositional effects; (b) there is no fuel compositional correlation to fuel instability; (c) analytical techniques are unable to isolate and detect the actual fuel species responsible. This last possibility is not very likely, given the tremendous chromatographic advances over the last 40 years. The first possibility has been noted by several workers even when working with simple model systems to induce sediment production.10 These results indicate that there may be a physicochemical explanation for insoluble sediment formation in middle-distillate fuels. Hypothetical fuel precursors (which are present in the methanol extract) would be the basis for particle growth following chemical oxidation (perhaps strongly influenced by fuel acidic components). These precursors should be higher in molecular weight than the so-called soluble gums. Hence, the qualitative and quantitative determinations of these precursors would require the application of analytical techniques hitherto unused in fuel compositional analysis. Acknowledgment. We wish to acknowledge Dr. Robert N. Hazlett's advice and helpful participation in technical discussions. This work was funded by the U.S. Navy Energy R&D Office and the David Taylor Research Center.
