CHARGES FOR EXTRACTION
FRACTIONS FROM PRELIMINARY SEPARATIONS
WEIGHT QMS.
FRACTIONS FROM EXTRACTION
KVIOO',. STOKE
LUBRICANT
I
I/ .
f ,
1-1
20.25 KG. IN 16 FRACTIONS
EXTRACTION WITH SOz
t
/
1
IN STORAGE
492
I 0.42-0.44 1 14752-1.4764
553
I a44-0.461
550 I a46-0.471 t 4 7 5 a - 1 ~ 7 7
51
5 0 0 10.47-0.491 14760-l.4776k 452
DEWAXING AT -18.C. WITH GYCL,
FRACTIONAL DISTILLATION IN VACUO
ADSORPTION WITH SILICA GEL
*
L4758-IA769p '
H
438
0.49-0.51
14761-1.478
I 060-065 114779-1.4785
--
SERIES C
- - -0
EACH CHARGE SEPARATED INTO ABOUT 25 TO 35 FRACTIONS BY EXTRACTION WITH ACETONE PLUS . WATER IN 14,-METER COLUMNS
+
SERIES D
.-0
SERIES E
--U
KV,,e,. ~ 0 2 1 4 -113 STOKE D , ~ ~a ~4 ,7 .- 0 . ~ 7 8 VML.
TO EXTRACT
'
IN STORAGE
FIGURE1. TREATMENT AND DISPOSITION OF LUBRICANT FRACTION
LUBRICATING
0IL FRACTIONS Acetone Extraction
of Constant-Boiling Fractions BEVERIDGE J. MAIR AND S. T. SCHICKTANZ National Bureau of Standards, Washington, D. C.
oT
HIS paper presents a continuation of the work on the chemical constitution of lubricating oil, undertaken a t the National Bureau of Standards as Dart of the American Petroleum Institute Research Project 6. I n particular, it describes the of molecules. effected bv seDaration. with resaect to tvDe "* sohent extraction of oil which had previously been extensively distilled and which was substantially constant-boiling, A correlation of the physical properties of certain fractions from this extraction process and their comparison with those of synthetic hydrocarbons of high molecular weight follow in a subsequent paper (page 1452). Preliminary Treatment of Fraction The preliminary treatment of the lubricant fraction of the Midcontinent petroleum used in this work was described previously (4) and is illustrated graphically in Figure 1: The lubricant fraction was separated by successive treatments into three parts: (a) an extract portion by extraction with sulfur dioxide, ( b ) a wax portion by crystallization from ethylene chloride a t -18' C., and ( c ) a water-white oil portion by filtration through silica gel. This paper deals with the water-white portion only; the investigation was begun by distilling this portion under high vacuum with the object of separating it into fractions differing in molecular weight. The light and heavy ends from this distillation were placed in storage as indicated in Figure 1, and distillation of the remainder was continued until substantially constant-boiling fractions were obtained. Charges were prepared for extraction by mixing according to their viscosities, the distillation fractions of about 45 grams each. The weight of these charges and the approximate range in kinematic viscosity at 100' F. (37.8' C.) and refractive index a t 25' C. of the fractions which it was necessary to mix to prepare each charge 1446
DECEMBER, 1936
INDUSTRIAL AND ENGINEERING CHEMISTRY
are shown in Figure 1 under "charges for extraction." Each charge was separated into twenty-five to thirty-five fractions by extraction with acetone in 14-meter columns. Kinematic viscosities at 100" and 210" F. (37.8' and 98.9" C.) and refractive indices at 25" C. were determined on all fractions from the extractors. However, these properties are reported only for the series designated in Figure 1 as A , E , C, D, E, and F , which were regarded as typical. In addition, for the same series carbonhydrogen ratios, molecular weights, densities, dispersions, optical activities, boiling points, and aniline points were determined on certain key fractions.
Distillation The systematic distillation under high vacuum of the water-white oil was continued, using the equipment and procedure described previously (4). When the range in viscosities of the fractions resulting from the distillat,ion of each charge, particularly f,or the more volatile portion, finally became nearly as small as the range in viscosity of the fractions which were blended t o make up a charge, it was decided that it was unprofitable to continue the distillation further. In Figure 2 are plotted the viscosities of fractions resulting from the final distillation with respect t o the percentage by weight of the charge. Curves for all the charges are not included, but those given are typical. On the right-hand side are indicated the weights of the individual charges, together with the range in viscosities of the fractions which it was necessary t o mix to prepare a charge. Curves I to V are the result of seven distillations in short fractionating columns ; curves VI1 to S represent the results of four distillations in a column molecular still, followed by from four t o seven distillations in simple molecular stills. Curve VI is the result of eight stages of molecular distillation followed by one distillation in a short fractionating column. There W B S an interchange between the two types of distillation, the more volatile fractions from the molecular stills (below 0.60 stoke) being charged for the next distillation into the fractionating column, while the less volatile fractions from the fractionating columns (above 0.60 stoke) were charged into the molecular stills. A greater range in the viscosity of the fractions from the distillation of the less volatile charges is noticeable. This 135
135
25
26
0
25 PERCENT
50
75
100
OF T H E CHARGE BY WEIGHT
FIGURE2. VISCOSITIES OF FRACTIONS FROM FINAL DIETILLATION
Each point represents b y i t s ordinate the viscosity of fraction by its sbsrissa the percentage which had been distilled including the iraotion.
1447
This paper describes the separation] with respect t o the type of molecule, effected by solvent extraction with acetone of substantially constant-boiling fractions of a water-white lubricating oil. Each charge of about 500 grams was separated, by extraction in 14-meter columns, into twenty-five to thirty-five fractions. Kinematic viscosities a t 100" and 210" F. (37.8" and 98.9" C.) and refractive indices were determined on all fractions. In addition to the properties mentioned, carbon-hydrogen ratios, molecular weights, densities] dispersions, optical activities, boiling points, and aniline points were determined on about thirty key fractions. The extractor columns and their mode of operation are described.
may be due in part to the limited quantity of oil in this region, since it was necessary t o blend fractions with a fairly wide range in viscosity to prepare even a small charge. At the end of this systematic distillation, the originally water-white oil was somewhat colored; the distillates were pale yellow, while the still-pot residues were reddish brown.
Extraction with Solvents A powerful tool for investigating the composition of lubricating oil fractions is now available in extraction with solvents. Solvents have been used for many years t o effect a separation of the various types of hydrocarbons in petroleum. However, the twofold action of solvents in separating, both with respect to molecular weight and type, has only recently been clearly recognized It is evidently important t o extract fractions composed of hydrocarbons with a narrow range of molecular weights if the maximum separation with respect t o type is t o be obtained. Consequently, the extraction experiments of the earlier investigators were not particularly successful, since they used fractions of wide boiling range and obtained fractionation with respect to both type and molecular weight. Recently Saal and.Van Dyck (Y), by drawing attention to the analogies between distillation and solvent extraction and emphasizing the importance of reflux in extraction, pointed the way t o the more efficient use of solvents. Fenske and co-workers (1) have made notable contributions in this field and constructed columns which utilize the idea of reflux and give exceptionally good separations. Columns similar to Fenske's, designed for the small charges available, were erected in this laboratory and have proved effective in separating the narrow distillation fractions with respect to type. Description of Extractions Figure 3 shows one of the extractors used with acetone, or acetone with 1.5 per cent water, as a solvent: It was constructed entirely of Pyrex glass. It consisted of the 500-ml. Kjeldahl flask, E , from which the solvent was distilled to condenser J and from which it flowed through tube I , then through the oilin Erlenmeyer flask D (500 to 700 ml.), returning again saturated with oil to flask B. When the solvent in fla-k E became
those solvents wlricti give B good scpnration with respect to type ---lrave been reported. &lostof them, however, Imil :it elevated t.cmperiltnrcs and probably would have been iiiiiicult to remove completely from the oil and would have ilecn likely to attack the oil at the temperatores necessary to distill and circulate them in t h e type OS colunin just described. bletiiyl cyeiride, one of the lower boiling solvents, \\-:LS tried, but the globules of oil instead of falling Ercely stuck to the glass tubing, eventually clogging the roliimn.
thorough mixinx of ihe oil and intimate
s o l v e n t . Tuhea E and €1 fxt k v,
0 35
0 v,
K. V. OF ORIG. CHARGE 2 9 -32
5
0
>
~ 2 5 .
a z W z
> 25s
I
z
15
15 3 0
20 40 60 80 100 PERCENT BY WEIGHT OF CHARGE
0
20 40 60 80 PERCENT BY WEIGHT OF CHARGE
100
FIGURE 10. VISCOSITIES OF FRACTIONS BELONGING TO SERIES A
require very narrow distillation fractions. This point will be referred to later.
Viscosity Indices The relation between viscosity, the temperature coefficient of viscosity, and constitution has been investigated by many chemists, and is now much better understood than a few years ago. Huge1 (4) and Mikeska (IS), attacking the problem from the synthetic angle, showed that increasing the degree of cyclization causes an increase in viscosity and a decrease in viscosity index. Davis and McAllister (1) found that a linear relation exists between the viscosity index of an aromatic-free oil and the percentage of carbon atoms in naphthene rings, the viscosity index decreasing with increase in the percentage of carbon atoms in naphthene rings. They pointed out that the first extracts which they obtained from a Pennsylvania oil deviated from this linearity and explained this on the assumption that the first extracts contained aromatic or unsaturated hydrocarbons. Zt was of interest to test the relation found by Davis and McAllister (1) with the key fractions obtained in this investigation. The upper half of Figure 6 is a plot of the percentage of carbon atoms in naphthene rings (calculated from the empirical formulas on the assumption of 6 carbon atoms to the ring) with respect to the viscosity index. For oils with viscosity indices above 90 (although there is a marked scattering of points), there is some semblance of linearity between viscosity index and percentage of carbon atoms in naphthene rings. The majority of the fractions with viscosity indices below 90 (the fist and second key fractions) are definitely out of line, their viscosity indices being much higher than would be expected if only naphtheiies were present. This abnormality is undoubtedly due to aromatics or unsaturated material, the presence of which in these fractions is evidenced by the specific refraction, specific dispersion, and aniline points. The lower portion of Figure 6 is a plot of the viscosity index with respect to the percentage of carbon atoms in naphthene rings for series A together with those for a number of synthetic naphthenic hydrocarbons. The viscosity indices of the synthetic hydrocarbons show the effect of constitutive differences besides that of cyclization, but are in as good agreement with those of the fractions from petroleum as could be expected.
I n Figure 7 are plotted the kinematic viscosities at 100" F. of the key fractionsfrom petroleum with reepect to the number of rings they contain. The fractions from each extraction charge are connected with arrows which indicate the order of their extraction. The approximate number of carbon atoms in each series is indicated by the figures shown in the circles. Also represented by hexagons are a number of synthetic naphthenes, the figures within the hexagons indicating the number of carbon atoms contained in each naphthene. As has been observed by other investigators, both increase in molecular weight and increase in the degree of cyclization of the petroleum fractions cause an increase in the viscosity. As with the other properties and presumably for the same reason (i. e., the presence of aromatic and unsaturated hydrocarbons), the first key fractions show abnormal viscosities which were lower than expected. The synthetic naphthenes show the effect of constitutive differences other than cyclization but are in as good agreement with the fractions from petroleum as could be expected.
Boiling Points In Figure 8 are plotted the boiling points at 1 mm. mercury pressure of the key fractions, with respect to the number of carbon atoms. A straight line represents the data fairly well with the exception of two points. As far as can be told from these data, the boiling point is a function of the number of carbon atoms and does not d e pend (at least to within about *3" C.) on the number of naphthene rings in the molecule. Properties of Fractions from Series A In this discussion of physical properties and constitution, so far only the key fractions, on which a large number of physical properties were determined, have been considered. The properties of all the fractions from one charge, A , which was systematically extracted twice, will uow be considered. Charge A was divided into three equal parts (Figure 1, page 1446) and each part extracted separately. The resulting fractions were then blended according to their physical properties to make up three new charges, each of which was again extracted. In Figures 9, 10, and 11 are plotted the refractjive indices, viscosities, and viscosity indices of the
DECEMBER, 1936
I
INDUSTRIAL AND ENGINEERING CHERZISTRY
--
I
FIRST EXTRACTION
1459
SECOND EXTRACTION 10
20
40
40 -p9
io
lo
x
W Q
z O0
> k cn
0 EO
&I
140.
L7.- --.
40
I
160
PERCENT BY WEIGHT OF CHARGE FIQURE 11. VISCOSITY INDICES
2 >
I
60
PERCENT BY WEIGHT OF CHARGE O F FR.4CTIONS BELONGINQ TO SERIES
A
viscosity characteristics are probably attributable to one or all fractions with respect to the percentage by weight of the of these substances. charge for both the first and second extractions. The range Mikeska (13) showed from his study of the properties of in physical properties and total weight of the fractions which synthetic compounds that unsaturation does not appreciably it was necessary to blend in preparation of the second set of change the viscosity characteristics. Therefore, if the uncharges is indicated a t the right of the figures. saturated compounds in the first few fractions have the same The wide spread in the properties of the fractions from the composition (except for unsaturation) as the naphthenes first extraction is evident. The spread in properties of the contained in these fractions, then no abnormalities in viscosity fractions from each charge in the second extraction is less, would he expected. However, it is probable that the unand in the case of the less soluble fractions is scarcely more saturated material in these fractions is more nearly analogous than the spread in properties of the fractions which were in composition to the original charge and consequently has a blended to make up the charge. This indicatJesthat the limit higher viscosity index and lower viscosity than the naphof separation, with acetone as a solvent, and with the small thenes in the fractions with which it occurs. This might accharges available, has about been reached. There is a suggescount for the comparatively high viscosity indices and low tion from the constancy of physical properties indicated by viscosities of the first few fractions. The possibility that these the flat portion of curves I and I1 that compounds of one anomalies are caused bv aromatic hydrocarbons with high type are concentrating in these regions, particularly since the viscosity indices raises some interesting empirical formulas i n these regions speculations which are considered in the correspond approximately to monocyclic following section. and dicyclic naphthenes. One of the most interesting features of curve I11 is the fact that the viscosity decreases Temperature Coefficient of for the first few fractions, then rises Viscosity of Aromatic Extracts markedly; the viscosity index also shows Mikeska (IS) showed that, in changa similar peculiarity, increasing for the ingfrom a naphthene ring to an aromatic first few fractions, then d e c r e a s i n g ring without altering the rest of the markedly. These abnormal variations molecule, there is some decrease in visin viscosity and viscosity index are cosity but, no appreciable change in the evidently not due to any change in the viscosity index, and that aromatic hydroefficiency of the extracting column, since carbons with one or two rings attached the refractive indices for these fractions to long paraffin side chains have, like decrease in a perfectly normal manner, the corresponding naphthenes, remarkbut must instead be attributed to the ably high viscosity indices. We might fact that in this region we are dealing then naturally expect some material of with a mixture of s u b s t a n c e s with high viscosity index in the supposedly markedly different viscosity charaoteraromatic extracts from petroleum. Yet istics. As has already been pointed out, No. OF A ~ O M A T I CRINGSno extract material of high viscosity the first fractions extracted contain small FIGURE12. RELATION BETWEEN VISindex appears to have been found. a m o u n t s of oxygen compounds, unCOSITY INDEXAND NUMBER OF RINGS saturated material, and probably aroWhether this material does not exist FOR CYCLOPARAFFINS AND FOR AROin petroleum or whether its presence matic hydrocarbons, and the abnormal MATIC HYDROCARBONS
1460
INDUSTRIAL AND ENGINEERING CHEMISTRY
has been overlooked, is an interesting question. In Figure 12 (curve I) are plotted the viscosity indices with respect to the number of rings in the molecule of the key fractions resulting from the double extraction to which reference has already been made. The five key fractions which contain no aromatics form a smooth curve which has been extrapolated to indicate (this requires experimental verification) what would be the viscosity index of a naphthene with 6 rings and 28 carbon atoms. On the basis of Mikeska’s conclusions, curve I1 has been constructed identical in shape with curve I; it indicates what would be the viscosity indices of aromatics analogous in constitution to the naphthenes of curve I. Curve I1 is displaced from curve I in such a manner that monocyclic aromatics fall above hexacyclic naphthenes. If aromatics with one ring have solubilities not far removed from those naphthenes with six rings (indicated as probable in the discussion of “aniline points”), it is easy to see why aromatic extracts of high viscosity index have not been obtained. In the modern methods fractions with wide boiling ranges are extracted, and the mononuclear aromatics of high viscosity index would always be mixed with polynuclear naphthenes of low viscosity index. This question, in addition to its theoretical interest, is of considerable practical importance, since the petroleum industry today may be discarding, in the solvent extraction processes, aromatic material with viscosity characteristics as good as the best naphthenic material and with oxidation characteristics which may be as desirable.
Iso- or Branched-Chain Paraffins i n Lubricating Oil Whether iso- or branched-chain paraffins are important constituents of mineral oils is an unsettled question. Kyropoulos (5) came to the conclusion, as a result of comparing the refractive indices and densities of fractions of a Pennsylvania oil with those of synthetic hydrocarbons, that it was composed largely of isoparaffins. Vlugter, Waterman, and Van Westen, considering the same data on the same oils, showed that the specific refractions are too low to be accounted for on the assumption of the presence of isoparaffins and concluded that these oils are composed of naphthenes. An examination of the literature shows that of the many fractions of wax-free lubricating oils from a wide variety of ~ourcesand prepared by widely different methods, none have been found with a positive value of z in the formula CnH2n+ =. An examination of the key fractions used in this investigation shows only one fraction with a positive value for z (+0.35) in the formula C,,H2n+z. This fraction contained a small amount of wax which was removed and identified from its melting point and refractive index as being composed of normal paraffin hydrocarbons. Negative values for 2 do not, however, prove the absence of isoparaffins, since mixtures of a dicyclic naphthene with an isoparaffin may give negative values for z. Nevertheless, the fact that no fractions were
VOL. 28, NO. 12
observed richer in hydrogen than C,H2, (except for that one known to contain normal paraffins), obtained from an extraction process which was dekitely separating the oil into fractions containing, respectively, more or fewer naphthene rings and should, consequently, have separated isoparaffins, is considered strong evidence t,hat no appreciable percentage of isoparaffins exists in these fractions. Even the supposed existence of isoparaffins in wax from petroleum neede reexamination, since the determination of the empirical formulas of some waxes in this laboratory (IO) showed that certain waxes which melt some 35’ C. below the corresponding normal paraffins and were formerly supposed to consist of isoparaffins, show considerable deficiencies in hydrogen from the formula CnHzn+z.
Acknowledgment The authors desire to express their gratitude for the advice and encouragement of F. D. Rossini. They wish also to thank L. A. Mikeska of the Standard Oil Development Company of ‘New Jersey for transmitting to them prior to publication his interesting manuscript on synthetic hydrocarbons of high molecular weight. Financial assistance for this work was received from the research fund of the American Petroleum Institute. This work is part of Project 6’on the “Separation, Identificat’ion, and Determination of the Constituents of Petroleum.”
Literature Cited (1) Davis, G. H. B., and McA411ister,E. N., ISD.E x . CHEM., 22, 1326 (1930). (2) Delcourt, Y., Bull. soc. chim. Belg., 40,285-94 (1931). (3) Griffith, R. H., and Hollings, H., J . Inst. Petroleum Tech., 20, 255 (1934). (4) Hugel, G., Chimie et industrie, 26, 1282 (1931). (5) Kyropoulos, S., Z. phusik. Chem., 144A,22-48 (1929). (6) Landa, S., and &ch, J., Collection Czechoalov. Chem. Communications, 6 , 423 (1934). (7) Lane, E. C.. and Garton, E. L., Bur. Mines, Rept. Investigations 3279 (1935). (8) Lerer, M. M., Ann. combustibles Ziquides, 8, 687 (1933). (9) Mabery, C. F., ISD.ENO.CHEM.,15, 1233 (1923); 18, 814 (1926). (10) Mair, B. J., and Schicktane, S. T., Ibid., 28, 1056 (1936). (11) Ibid., 28, 1446 (1936); Natl. Bur. Standards, J . Research, 17, Dee., 1936. (12) Mair, B. J., Schicktana, S. T., and Rose, E’. F., Jr., Ibid., 15, 557-73 (1935). (13) Mikeska, L. A,, IND.ENG.CHEX, 28, 970-83 (1936). (14) Rossini, F. D., Proc. Am. PetroleztmInst., 16M (III), 63 (1935). (15) Suida, H., and Planckh, R., Ber., 66B,1453 (1933). (16) Vlugter, J. C., Waterman, H. I., and Van Westen, H. A., J. Inst. Petroleum Tech., 18,735 (1932); 21, 661, 707 (1935). RECEIVEDOctober 9, 1936. Presented before the Division of Petroleum Chemistry a t the 92nd Meeting of the American Chemical Society, Pittaburgh, Pa., September 7 to 11, 1936. Publication approved b y the Director, National Bureau of Standards, The authors are research associates a t the bureau, representing the American Petroleum Institute.
F ’rom B a n k a Tin
