paratively large bulk of liquid in which such a solute is dirsolved. Processes employing two solvent phases, on the other hand, usually have as their objective the separation of a mixture of substances into its individual components. This latter process is closely analogous to that of fractional distillation and may be referred to as “fractional distribution,” as distinct from the first type of process which may be either a solvent refining or solvent extraction process. Although fractional distribution can be used in a greater variety of ways and i s altogether a much more flexible process than fractional distillation, it has as yet been but seldom employed even in the redearch laboratory. Jantzen (10‘) ha5 used it for the separation and isolation of isomeric quinoline bases from coal tar, and more recently Cornish and his coworkers (5)have employed it for the purification of vitamin,, while the Duo-Sol process (24) for treating lubricating or1 stocks is the only commercial application of fractional diitribution. In this process a lubricating oil stock 1s separated into two portions, one consisting of naphthenic or undesirable constituents and the other consisting mainly oi paraffinic material, a special blend of coal-tar acids and bases being employed as the solvent for the naphthenic type and propane as the solvent for the paraffinic type constituents. On the other hand, both solvent extraction and solvent refining processes have been widely employed in the laborstory for many years. Solvent extraction has also been succewfully applied commercially to the recovery of acetic acid in the cellulose acetate industry and in the dehydration of acetic acid. In recent years it has also been widely used for the dephenolation of ammonia liquor and coke-oven effluents in Germany, England, and America, Owing to the increased severity of the requirements placed on motor lubricants by the modern type of automotive engine, a new era in lubricating oil refining is being developed with great rapidity Engine speeds, bearing speeds, and pressures are all increasing. dutoinotire engine speeds of 4000 r. p. m. are not uncommon in England and the Continent, and American engines are developing rapidly in the same direction. Engine tolerances in clearances and uniformity are nom far narromr than in the past. All these tendencies in engine development are creating a steadily increasing demand for more efficient lubricants. Perhaps the most universally adopted niethod for meeting such demands is the solvent refining treatment for lubricating oils and lubricating oil stocks This process has, in consequence, been subjected to a remarkably rapid commercial development during the last two yeais, a considerable number of solvent refining plants having been erected and operated in America, England, and France during this period. In America alone it has been estimated that $5,000,000 have already been -pent on units novs in operation and under construction and that these plants when completed will have a total capacity of refined oil of 15,000 barrels per day.
Liquid-Liquid Extraction Computations for Solvent Refining of Oils Extraction methods suitable for liquidliquid systems are described and computation methods applicable to these cases are briefly discussed. A graphical method, employing triangular coordinates, for representing equilibrium in complex oil-solvent systems is developed as a basis of computation €or the solvent refining of oils. Calculated and experimental results from both multiple and countercurrent extraction processes refining different lubricating oil stocks with different lubricating oil stocks with nitrobenzene are compared, and the methods presented are shown to be sufficiently accurate for most purposes.
T, G. HUNTER AXD A. W. KASH ‘TheUniversity of Birmingham, England
URPNG the last five years considerable progress has been made in the development and industrial application of extraction processes, particularly in the special case of liquid-liquid systems. Such extraction processes are now used successfully for the recovery of phenols from effluent liquors, for the winning of halogens from brines, for the dehydration of acetic acid (2. 21)! for the regeneration of alkali phenolates ( 1 9 ) , and for the refining of petroleum products. In the petroleum industry solvents are now employed for deJT-a.uing oils, for asphalt removal (18), for refining burning oils to be used as illuminants, and finally, what is perhaps the iiewest and most important application, for the solvent refining of lubricating oils (1). A wide variety of solvents, chiefly organic, are used in these various extraction processes. Usually a sing!e solvent phase is employed but in certain case. two solvent phases, either completely or partially immiscible, are used. Processes employing a single solvent phase are usually refining or extractive in character, their main object being to remove a solute impurity or separate a valuable solute from a com-
Multiple Contact Met hods Methods for carrying out extraction or distribution processes fall into two distinct classes-namely, multiple or successire contact methods and countercurrent contact methods, The multiple contact method of working a qolvent extraction or refining process is illustrated diagrammatically in Figure 1. Mixing and separation of the treated and solvent liquids are followed by contact of the treated liquid with fresh solvent. This operation E repeated as often ab required. The method may be carried out in batches, or semi-continuour working may be employed. In the latter case the two liquids entering each stage of a multistage process are mixed continuously in a suitable mixing device and finally separated either by centrifuges or continuous settling. The mcthod 836
INDUSTRIAL AND ENGINEERING CHEMISTRY
JULY, 1935
finds its main application as a laboratory or small-scale method and is seldom used commercially, since comparatively large volumes of solvent are required to ensure satisfactory extraction. Where the two liquids, treated liquid and solvent, are completely immiscible and where the solute is distributed between them in accordance with the distribution law, computations for the multiple contact extraction method have been derived and exhaustively discussed in the literature. Calculations for successive extractions with batches of fresh solvent have been put forward by Her2 (11), Smith : ;; (25), Holleman ( l a ) , and Fischer (8),while a graphical method has been described by Evans (6). Successive extraction with batches of solvent containing some dissolved solute has been discussed by -Underwood (65). I n the multiple contact method the most efficient extraction is obtained when the solvent is subdivided into batches of equal volume; this condition has been dealt with fully by several investigators (16, 23, 26, 68). With a finite volume of solvent available for extraction, the limit of the process with the solvent subdivided into many portions has been investigated theoretically by Evans (4) and Griffin (IO). Continuous eo-current extraction in a single stage, such as would be obtained by the eo-current passage of treated liquid and solvent through an orifice column or packed tower, has been treated mathematically by Fischer (8). When the solute exists in one liquid as single molecules and in the other liquid chiefly as double molecules, the distribution equilibrium is given by a formula of the type K = Cl/dc,and computations for this special case have been derived by Friedrichs (9) and by Hunter and Nash (15). A more complicated case of solute association is often found, for example, in the distribution of phenol between benzene and water where the solute exists in one phase partly as single and partly as triple molecules. Multiple extraction calculations for this type of equilibrium have been discussed by Hunter and Nash (15). All of these computations mentioned apply to special cases where some comparatively simple distribution equilibrium is maintained over a limited solute concentration range, when the two solvent liquids are completely immiscible or SOLVEHT LlGUlD
837
where their partial miscibility is not materially altered by the presence of the distributed solute. Equilibrium relations for the general case where the partial miscibility of the two solvent liquids varies with the concentration of the distributed solute are best represented graphically on triangular coordinates. A graphical method of computation applicable to this general case has been described by Hunter and Kash SOLVENT
ilqvlo
P‘W
,P, w
/*’
W
y”w
P
FIGURE 1. MULTIPLE CONTACT EXTRK T I O N
(IC) and reviewed together with other computation methods by Evans (6). For the separation of two substances by fractional distribution employing the multiple contact method, a conventional treatment scheme is shown in Figure 2. The total amount of the mixture to be treated is dissolved in one of the solvents and contacted with the second solvent in an amount such that the constituents of the mixture are fairly evenly distributed in both layers. The layers, after separation, are each contacted with a volume of the other solvent in a n amount equal to the volume of the layer. Of the four fractions thus obtained, the middle ones are mixed and the outer two are again treated with pure solvent. This procedure is continued until the outer fractions are sufficiently pure when they are set aside and only the middle or intermediate fractions further treated. If a mixture consisting of 100 grams each of two substances, A and B , is treated a t all stages with equal volumes of two completely immiscible solvents, P and W , in which both A and B are distributed according to the distribution law, such that KA
concn. of A in P solvent phase - 0.25 = concn. of A in W solvent phase -
and concn. of B in P solvent phase - 4.00 Ke = concn. of B in W solvent phase -
50LVENT LlQUIO
P
FIGURE2. MULTIPLE FRACTIONAL DISTRIBUTION FOR SEPARATION OF TwoCOMPONENT hlIXTURE
then the degree of separation of A and B in the first three fractions of each solvent, P I , Pf,and Pa, and W , , Wa, and will be that shown in Table I. Table I shows that, provided the solvent liquid pair is chosen so that a large difference between the distribution ratios of the mixture components exists, then really effective separation can be obtained in a f e n stages. Computations for processes of this type have not been specially described in the literature but are fundamentally similar to the usual multiple extraction calculations. A useful method of fractional distribution has been employed by Jantzen (16) for the separation of mixtures of isomeric organic bases. The basis of this method is the employment as the solvent pair of an organic solvent and an aqueous solution of easily hydrolyzed compounds of the mixture components to be separated. In the separation of
ISDUSTRIAL ASUD FXGIKEERING CHEMlSTHX
838
two isomeric organic bases, for example, ether or benzene could be used as the organic solvent and an aqueous solution of the hydrochlorides of the bases as the other solvent. The method is illustrated diagrammatically in Figure 3. Ten equimolecular amounts of the mixture of bases to be separated are dissolved in ten equal volumes of a n organic solvent, and contacted with five portions of dilute hydrochloric acid. The amount of acid used in each of the acid portions is just sufficient, to combine exactly with the amount of base dissolved in each of the base solutions in organic solvent. The first portion of acid solution is contacted vith the first TABLE
1. BY
SEPARATIOX OF TWO-COUPONENT ;\IIXTURE
MULTIPLEFRACTIONAL DISTRIBUTION
'02Et'
Frsclion Pa
P2
P8
PI+P~+P,
Grams of Solute in Separated Fraction Solute 51.5
25.4 12.2 89.1
99.0
98.5 9s,4 98.8
Fraction FVl
a'u
K3
Wa+W+Wr
0 by Wb. Grams pf o'f A in Solute in Separated Fraction Solute 51.5 99.0 28.4 98.5 12,2 98.4 89.1 98.8
base solution. The organic solvent, after separation, should be free from base, but may contain a small quantity due to the hydrolysis of the hydrochloride. If necessary this can be removed by contacting with the second acid solution, All the base solutions are contacted with the aqueous solutions according to the scheme shown in Figure 3, until five basefree portions of organic solyent and five portions containing base are obtained, together with five aqueous solutions of base hydrochlorides. The first of these aqueous fractions is treated with an alkali solution and contacted with fresh organic solvent, and this procedure is continued until all the bases are obtained in ten organic solvent fractions. With this method it is also possible to return any desired proportion of the separated bases to the system by washing the final fractions of bases dissolved in organic solvent with dilute acid and returning the resulting acid extract to the system. This corresponds to the use of reflux in fractional distillation. In Figure 4 some seaults obtained by Jantzen for the separation OS a n equimolar mixture of a-naphthylamine and quinoline and also of a mixture of isoquinoline and quinoline in equal weights by treetment with dilute hydrochloric acid and benzene are shown. Good separation of the a-~aphthylamine-cluinoliiie mixture has been obtained by this method, but the separation is less complete iyith the
isoyuiiioliiie- quinoliiie mixture because in this inst'ance the dissociation aiid distribution coefficients are less favorable. The e q u i l i b r i a selationahips for the type of solvent pair system described in the foregoing can be calculated by u s i n g t h e appropriate dissociation and distribution coefficients. Such calculations in connection with liquid-liquid e x t r a c t i o 11 p r o c esse s have been discussed by J a n t a e n (16) and by Meyer (bo), These calculations are not always accurate, however, and equilibrium relationships are best determined experimentally and represented graphically by plotting the solute concentration in the organic solvent layer against total eoncentration o€ solute, as free solute and complex, in the aqueous layer. Computations for processes of this type have not been described in the literature, It is generally recognized that the multiple contact method is much less efficient than the countercurrent. method of contacting. The latter method, however, i s not always practicable, particularly on the laboratory scale, although it can readily be applied by using a taxer in place of a series of mixers; such towers, however, are generally equivalent only to a relatively small number of ideal mixers and are not very satisfactory. The multiple contact method therefore finds iti greatest application in &he laboratory and, where drastic restriction in the volume of solvents employed i s seldom essential, can be used very effectively. It is often desirable, liowever, to reproduce countcrcurrent effects in the laboratory without 81%extensive outlay of t h e and equipment. Such an effect can be very simply reproduced with ordinary laboratory apparatus by an extension of
VIELQ
OF W W E B FMCTIONS OQC?INAL
A5
WEICHT
PERCENT
OF'
MIXTURE
FIGURE 4 SEP.WLITION OF Two MIXTURES BY Hs-, DRQCHLORTC kCII> 4 x 1 BEXZEUE TREATVE~T
INDUSTAI.4L AND ENGINEERING CHEMISTRY
JULY, 1935
the multiple contact method. Again, of course, solvent economy niust not be severely limited. The process, called "pseudo countercurrent extraction," was first developed by Watanabe and Morikawa ($7). The method is illustrated diagrammatically in Figure 5. In this example a countercurrent process is simulated in which a solute in P volumes of treated liquid is removed by contacting with TV volumes of solvent in n stages, liquids P and It' being completely inimiscible. The first horizontal row in Figure 5 shows the multiple extraction of P volumes of treated liquid in n stages with FV volumes of fresh solvent a t each stage. In every row, R,, which lies below the first row, P volumes of treated liquid are extracted at each stage, n, with W volumes of / - exit solvent from stage (n 1) of row R ~ - , . If such a series is infinitely repeated, it approaches more and more closely to the true countercurrent effect. This method of extraction has not been dealt with hitherto in the literature, and no computations have as yet been derived for the system. Lengthy derivations and discussions of this method would serve no useful purpose here, but a few brief indications of the properties of the system may be mentioned. I n the multiple contact method for the simplest case, where the two liquids are immiscible and the distribution law holds, the formula KP (1) yn = [ K m F I " nhere yn = amount of solute unextracted after the nth extraction yo = original amount of solute in treated liquid K = distribution coefficient of solute in the two liquids P, W = volumes of solute-free treated liquid and solvent, respectively n = number of extractions or extraction stages
+
+
MULTIPLE EXTRACTION
lNTER
-
,w COUNTERCURRENT EhTUACTION
/ , '
may be used to compute the amount of extraction. The percentage deviation from the true countercurrent value of y, given a t each stage of a four-stage pseudo countercurrent process with varying number of rows, R, where the group [KP/(KP W ) ] is equal to 0.33, is shown in Figure 6. This figure shows that the greatest deviation from the true countercurrent value of yn occurs in the last stage of this process, and that the deviation for all stages decreased with increasing R, approaching zero as R approaches infinity. The deviation becomes negligibly small a t a finite value of R. In Figure 7 the variation of the number of rows, R, with the W ) ]to give certain definite percentage group [ K P / ( K P deviations from true countercurrent values is shown for the last stage of a four-stage process. On the same figure exactly the same effect has been plotted using W / P instead of [ K P / ( K P W ) ] ,K having been taken equal to 0.5, and n equal to 4 as previously. These curves show that it is possible to approach within 2 per cent of the correct countercurrent values for the last stage of a four-stage process a t very low values of R, especially a t high solvent ratios. Deviations from the true countercurrent values in the other stages would be still smaller Figure 8 shows the variation of R with n, the number of stages, to give definite percentage deviations from true countercurrent ralues. As in Figure 7 the deviation used is that of the last or nth stage of an n stage process, while the value of the group [KP/(KP W ) has been maintained constant throughout a t a value of 0.25. As n, the number of stages, is increased, larger numbers of rows will be required to maintain the process results a t a fixed percentage deviation from the true countercurrent values. For the majority of chemical engineering purposes an ac-
+
839
FIGURE5 , PSEUDOCOUNTERCURRENT EXTRACTION curacy of * 5 per cent is usually sufficient, while for laboratory and research purposes 1 2 per cent is likely to be the maximum desirable deviation. It is obvious therefore that for most cases the pseudo countercurrent method can be used without excessive extraction repetitions. The pseudo countercurrent method has been employed by Watanabe and Morikawa (27) for the removal of phenols from creosote oil, containing 26.1 per cent by weight of phenols, by extraction with both methyl and ethyl alcohols. Distribution coefficients of phenols between the alcohol and the creosote oil were measured and, assuming that the distribution law held, the results for a true countercurrent process were calculated by means of a formula given by Hunter and Nash (IS). The experimental results from the pseudo countercurrent process are compared with the calculated results for the true countercurrent process i n Table 11. TABLE11. PSEUDO COUNTERCURRENT EXTRACTION OF CREOSOTE OIL wt. qo of of Phenols in
Alcohol in Water, Used as Solvent 59.7 methyl 59.7 methyl 69.5 methyl 75.2 methyl 58.8ethyl
No. of Solvent/Oil Stages, n Ratio, W / P 2.0 3 1.0
5
3 3 3
1.0 0.5
1.0
No. of Rows, R 4 7 4 4 4
Wt* efined Oil Exptl. 5.7 9.3 3.6 4.7 3.1
Cslcd. 5.5 10.3 3.1 3.6 4.2
I n Table I1 a large proportion of the discrepancy between the experimental and theoretical figures must of coume be
+
+
R
NUMB-
CF R O W 5
OF
PSEUDO
COUNTER-
CURRENT
PROCESS
FIGURE 6. DEVIATION FROM TRUECOUNTERCURRENT VALUEOF Y,
840
INDUSTRIAL AND ENGINEERING CHEMISTRY
P b C U D O COUUTLR CURRENT1 E I T R A C T I O H
I
due to the assumption of a constant distribution coefficient which assumption is by no means correct. Taking this into account, however, it will be seen that the pseudo countercurrent method does reproduce true countercurrent data to a reasonably accurate degree.
VOL. 21, NO. a
been used by Jantaen and Tiedcke (1") and has been under investigation for several years by the Bataafsche Petroleum Nij. ($2). The method is illustrated in Figure 9, where F i s the feed material to be refined, S the solvent employed, (S E ) the extract solution, and E the solvent-free extract (part of which is returned as reflux to the system), and R is the refined material produced. A graphical method of computation employing triangular cobrdinates has been applied for the general case of partially miscible treated liquid and solvent to such a system by Saal and Van Dyck (22). The feed material, F , in Figure 9 may represent a liquid to be refined by the removal of an undesirable constituent or may also represent a mixture of substances to be separated-for example,a binarymixture of constituentsA and B. In the latter case one of the constituents (either A or B ) must be a liquid immiscible or partially miscible with the solvent S which is employed. Instead of returning, as reflux, extract from which all the solvent has been removed, and which consists of a large amount of the extracted constituent, A , together with a small amount of B, the reflux may, if desired, be a ternary mixture of constituents A and B and solvent S. Its composition, however, must then correspond to the composition of some point which lies on a binodial curve of a triangular diagram representing the equilibrium between A , B , and S during cxtraction conditions. If the equilibrium conditions of the three-component by+ tern A-B-X a t constant extraction temperature are represented on B triangular diagram by a binodial curve of the type shown in Figure 40(a) where component A i s completely miscible with jolvent S but coniponent B is only partially miscible, then pure A completely free from B ran never
+
Countercurrent Extraction The utility of the countercurrent method of operation i s universally recognized, and it has been widely employed in many processes, notably in distillation, absorption, adsorption, washing, and extraction operations. In this method the liquid to be treated is contacted with solvent which has been employed in a previous extraction stage, except that in the final stage of the process the treated liquid is contacted with fresh solvent. This effect is most readily produced by causing the t v o liquids to flow continuously and countercurrently through a suitable vertical tower. The method may also be applied by contacting and separating in stages, the mixers and separators so employed being provided Kith suitable cross connections to insure countercurrent flow of the two liquids. As in the multiple contact method, computations for countercurrent extraction for the simplest case of two immiscible liquids in which a solute is distributed in accordance with the distribution law have been well established. The various mathematical and graphical treatments possible have been exhaustively described by Hunter and Nash ( I S ) who have also described a graphical method of computation (14) using triangular coordinates suitable for the general case where the miwibility of the two solvent liquids is altered by the distributed solute. These methods have been summarized and reviewed in a later paper by Evans ( 5 ) . In applying the countercurrent method, it is possible to increase the efficiency of the extraction by returning extracted material t o the extraction apparatus. The resulting process is then analogous! to a distillation process employing reflux. This possibility has not been realized in extraction until recently. This modified countercurrent method has
be realized, although pure B can be produced. The extract solution richest in A which can he reached is that represented by point XS on the binodial This corresponds to an extract, after the solvent has been removed, of a coinposition represented by X . Should the equilibrium conditions be represented by a binodial of the type shomn in Figure lo@), where both components are miscible with the solvent only to a limited extent, then complete separation of A and 5 can be obtained. Khere complete separation is desired, it is essential to select working temperature conditions in conjunction with the proper solvent to give these equilibrium relations. A similar effect to the use of a n extract reflux may be brought, about by the use of two immiscible or partially miscible solvents for the separation of mixtures by counter-
JULY, 1935
INDUSTRIAL AND ENGINEERING CHEMISTRY
current fractional distribution. In this process a binary mixture of A and B is fed into a column a t some intermediate enter at the opposite point, while the two solvents, Xiand Sz, ends of the column. Solvent XIleaving the top of the column rich in A has its content of B reduced by extraction with fresh solvent Si, E while this solvent leaving the bottom of the column rich in B is denuded of A by the action of solvent SI entering here. Solvents Xz and SI act in a manner analogous to the distillate reflux and stripping steam in a fract i o n a t i n g column. This method is employed in the Duo-Sol process for the solvent refining of lubricating oil, and except in this connection its use has not been described in the literature. In the Duo-Sol process, lubricating oil stock is separated into two groups of constituents, naphthenic and paraffinic. For the naphthenic solvent a special blend of coal-tar acids and bases with a boiling range of 360’ t o F I G U R E 9. 405’ F. is used, and for the paraffinic COUNTERCURRENT solvent liquid propane is recommended. EXTRACTION These solvents are only partially misEXTRACTREFLUX cible and in themselves form a twolayer solvent system. An ingenious countercurrent method of fractional distribution has been devised by Cornish, Archibald, Murphy, and Evans (3) and employed for the purification of vitamins. In this method two solvents flow through a column in opposite directions a t constant rates. The mixture to be fractionated, consisting of several components of which A is the component desired in the pure state, is suddenly injected into the middle of the column. If the inverse ratio of flow rates of the two solvents is exactly equal to the distribution ratio of A in these solvents; then A will simply be shuttled to and fro in the various sections of the column, eventually finding its way out of the column according to the laws of probability, while the other components of the mixture will be removed from the column by the two solvent streams. Computations for this method have been derived and discussed by its originators.
E
841
following lines leads to an approximate but satisfactory solution. In the solvent refining of an oil, the solvent tends to split the oil into two fractions, soluble and less soluble, .possessing different physical characteristics. These two fractions do not represent a sharp separation of the oil into purely soluble and less soluble constituents, but are mixtures of both. In certain special cases the nature of the soluble and less soluble constituents may be known-for example, in the refining of kerosene where the more soluble constituents are aromatic hydrocarbons and the less soluble are nonaromatic. In this case where the amounts of aromatic and nonaromatic hydrocarbons are readily determined by analysis, the oil may be considered as a simple two-component system, the two components being each complex mixtures of aromatic and nonaromatic hydrocarbons, respectively. The resulting oilsolvent system may then be treated as a simple ternary system and equilibrium relations represented on triangular coordinates in terms of solvent, aromatics, and nonaromatics; the triangular coordinate graphical method (14), suitable for the general case in ternary systems where the partial miscibility of the two solvent components is altered by the distributed component, may be used for both multiple and countercurrent extraction computations. Such cases where the oil can be analyzed in terms of two groups of constituents are, however, relatively rare and unimportant, compared with cases where such analysis is impossible. The chemical nature of the hydrocarbon groups present in any lubricating oil, for instance, is not known. Such oils are usually considered to be made up of two major groups of constituents generally referred to as “naphthenic” and
II,
;ii:
Equilibria in Complex Hydrocarbon-Solvent Systems Computations for all the extraction methods discussed above have been confined to the comparatively simple ternary system possessing well-established equilibrium relationships. Where solvent extraction is employed for the removal of undesirable constituents from petroleum produots, such as the extraction of aromatics from kerosene by liquid sulfur dioxide and the extraction of the more unsaturated constituents from lubricating oils by nitrobenzene, phenol, or dichloroethyl ether, the equilibria involved cannot be represented exactly by any simple means. Exact representation can be obtained only by the use of complex methods involving a comprehensive knowledge of the constituents of such petroleum products. Even if our present state of knowledge were sufficient to enable exact equilibria relationships to be compiled, it is doubtful if the resultant phase representations would be suitable for purposes of computation, since the major essentials of a successful method of computation must combine simplicity, ease, and speed of calculation with reasonable accuracy, all of which demand the simplest possible representation of equilibrium relationships. However this problem is simplified by our ignorance of the chemical nature of petroleum products, and equilibria characterization on the
TABLE111. LUBRICATING OILSTOCKA BENZENE AT
--Reffinate Ratio, aolvent Expt. to oil 1 6 2 3 3 1
YO
10”
-Extract -
Layer?
ori.gina1 V. G. C. oil In of oil In layer layer 10.5 0.824 85.0 0.836 63.0 0.851
c.
70
solyent In layer 10.7 14.6 17.0
7 0
AND
NITROLaye-r: YO
original V. G. C. solyent oil in of oil in in layer layer layer 87.0 89.5 0.872 81.9 65.0 0.852 70.2 37.0 0.900
where the term “naphthenic” is to be interpreted as denoting a mixture of hydrocarbon groups relatively poor in hydrogen, and the term “paraffinic” as a mixture of hydrocarbon groups relatively rich in hydrogen. Lubricating oils are then designated “naphthenic in character” or “paraffinic in character” by virtue of the numerical value of certain physical properties s u c h as the viscosity index N . I.) or t h e viscosity gravity constant (V. G. C.). The number and ac(4 (6) tual c h e m i c a l FIGURE10. Two TWESOF EQUILIBRIUM nature of the hYCONDITIONS drocarbon groups present i n a n y particular product are still entirely speculative. It is known, however, that the groups comparatively poor in hydrogen form the bulk of the undesirable constituents which have to be removed to produce a good lubricating oil. It is also known that such constituents possess a high value for the V. G. C. and a low value for the V. I., while conversely the desirable or so-called paraffinic constituents have, respectively, low and high values for these physical properties. The degree of refining of a lubricating oil stock may, therefore, be followed
INDUSTRIAL AND ENGINEERING CHEMISTRY
842
VOL. 27, NO. 7
The completed i s o t h e r m a l b i n o d i a l curve and tie lines shown on Figure I1(a) then represent the equilibrium relations for this l u b r i c a t i n g oil-nitrobenzene system a t PO" C, Figure Il(a) can also be used to determine the amounts of any two phases in equilibrium, given only the quantities of oil and solvent used for the equilibrium experiment. In experiment 3 equal volumes of nitrobenzene and oil were employed, so that the resultant mixture contained 50 per cent of nitrobenzene and its composition would be represented by the intersection of khe ,o' line joining the V. 6.C. of the stock, point c, to the 100 per cent nitrobenzene apex and the line representing 50 per cent of nitrobenzene-namely, point m. This mixture, however, can exist only as two phases whose composition h and 2" is given by the tie line passing through m. The amounts of the r a f i a t e phase, h, and the extract phase, i, are then proportional .Is the lengths of the NlTQOBENZENC At.CO*OL two lines mi and mh. Therefore, knowing the volumes of solvent and oil used (=I (d ) -+ 5OLUTlON fOMP051TION5 FROM MULTIPLE. EXTRACTION PROCES5E5 originally, the volumes of the two phases 0 SOLUTION COMPOSITION5 F R O M COUNTER CURRENT EXTRACTION PROCESSES formed can be calculated from this relationship. Further, since the volume per FIGURE11. EQUILIBRIA IN COMPLEX HYDROCARBON-SOLVENT SYSTEMS cent of solvent in each phase can be read from Figure ll(a),the actual volume of solvent-free oil in each and controlled t o a certain extent by the use of these or phase can be calculated also. Such a diagram therefore gives similar physical properties. a complete representation of all essential equilibrium relation-. Therefore with our present state of knowledge, in order to ships. I n constructing these diagrams, the scale employed for represent the equilibrium relations in an oil-solvent system, it the physical property of the solvent-free oil can be chosen t o is necessary only to record by some suitable means the followgive a suitably sized binodial curve, The actual terminal ing information: values finally selected become, by the nature of the diagram, (1) Amounts of the two phases in equilibrium. equivalent to the physical properties of the soluble and in(2) Amount of solvent in each phase a t equilibrium. soluble components of the oil. This is an unavoidable as(3) A physical property of the oil present in each phase at sumption made in the actual construction of Figure l l ( a ) , but equilibrium. since such terminal values will vary with the desired degree of refining for the same oil, and with solvent and temperature Items 2 and 3 of this information can be recorded by a single conditions, it is preferable not t o consider these terminal point on a triangular graph of which one vertex represents values as indicative of soluble and insoluble group properties. pure solvent and one side, opposite this vertex, is scaled into The equilibrium diagram can be constructed from the units representing the required physical property of the experimental results obtained by the single-stage batch exsolvent-free oil. The equilibrium resulting between a lubritraction of an oil with different volumes of solvent, as in cating oil and nitrobenzene a t 10" C. is shown in Table 111. Table III. Care must be exercised to insure that the t!wo These equilibrium relationships are plotted on the triphases are in equilibrium. The diagram can also be conangular graph, Figure l l ( ~ ) The . V. G . C. of the solventstructed from multiple extraction results provided that the free oil in the raffinate layer from experiment 1 is given by compositions of conjugate layers are used and true equilibrium point b on this diagram. If raffinate oil of this V. G. C. were is attained a t each eytraction ptage; that is, each stage is mixed with a n amount of nitrobenzene such that the total equivalent to an ideal stage. In st heterogeneous system mixture contained 10.7 per cent of nitrobenzene, a e should undergoing agitation, two phases may coexist which are get the actual raffinate layer obtained. Therefore by joining approaching equilibrium with each other. I n this case the b to the apex of the triangle representing 100 per cent of nitropoints representing the composition of each phase a t any benzene and locating the point on this line where the nitroinstant during approach to equilibrium may be assumed t o benzene content is 10.7 per cent, we obtain point d representalie on the binodial curve, since this curve represents the tive of the composition of the raffinate layer. This may be compositions of all possible heterogeneous solutions, but the shown in another way. By starting with the raffinate layer tie line joining two such points would not, of course, represent of composition d and removing all the solvent present, we equilibrium. The composition of any phase from either a would obtain point b on the diagram, which is solvent-free multiple or countercurrent extraction process would then oil of V. G. C. 0.824. The composition of the various raffinate presumably lie on the binodial curve even if equilibrium is and extract layers is therefore easily ascertained and is indinot attained in each extraction stage. cated on Figure l l ( u ) by points d, f, h, and e , g, i for the exI n Figure 11 several oil-solvent equilibrium diagrams are ample quoted in Table 111. Returning again to experiillustrated. The equilibrium between nitrobenzene and three ment 1, the compositions of the raffinate and extract layers in different lubricating oil stocks a t 10' C. are shown in ( u ) , equilibrium are given by points d and e. These two solu( b ) , and (e), using the Q. G.C. for the physical property of tions are conjugate and hence may be joined by the tie line de.
'"A
~
JULY, 1935
INDUSTRIAL AND ENGINEERING CHEMISTRY
843
expressed in terms of three components, they are given in terms of solvent per cent and a physical property of the solvent-free oil. Computations for a multiple or countercurrent extraction process normally involve a knowledge of the value of some physical property of the oil to be treated, the ratio of solvent to treated oil OF STOCKNo. 2 TABLEIV. MULTIPLEEXTRACTION WITH NITROBENZENE AT 10' C. employed, and the value of that physical R a 5 n a t e Layer -Extract Layerproperty of the refined oil desired. Given Solvent-free Vol. % of Solvent-free such data, it is usually rewired to calculate oil, voi. % of v. G. C. of solvent in oil, yoi. % of V. G. C. of original stock solvent-free oil layer original stock solvent-free oil at fixed working conditions (1) the yield of Actual Calcd. Actual Calcd. Actual Calcd. Actual Calcd. Actual Calcd. refined oil, (2) the amount of solventassociated
the solvent-free oil. The binodial curves and tie lines for these three systems were constructed from batch extraction experiments. The data for stock A in Figure ll(a) were kindly communicated to the writers by S. W. Ferris and the
-
tion Stage 1 2 3
49.0 27.3 15.6
49.7 31.0 20.7
0.840 0.811 0.802
0.838 0.811 0.805
19.5 12.5 16.6
19.5 15.0 16.0
50.7 50.6 72.7 69.3 84.4 79.6
data for stocks 2 and 3 were taken from the paper of Ferris, Birkhimer, and Henderson (6). The individual points indicated by crosses and circles were obtained, respectively, by multiple and countercurrent extraction of these oils with nitrobenzene a t 10O C., employing actual stages where equilibrium between the two phases of each stage was not always obtained. The fact that these points lie for the most part on the binodial curves provides a useful check on the already established shape of the binodials. It affords proof also of the assumption that a point representing the composition of a phase from any extraction stage of an agitated heterogeneous system, whatever the stage efficiency, will lie on the binodial. Such equilibrium diagrams can therefore be employed to compute graphically the results to be expected from any solvent extraction process of oil refining. I n Figure l l ( d ) the equilibrium diagram for the system kerosene95 per cent ethyl alcohol a t 17" C. is shown, using the aniline point (miscibility temperature of equal volumes of oil and aniline) as the physical property of the solvent-free oil. This diagram is similar to that of a ternary system where both components of the binary mixture being treated are only partially miscible with the solvent. The diagrams for the lubricating oil-nitrobenzene systems on the other hand, are all similar to that of a ternary system where only one of the components of the binary mixture being treated is partially miscible with solvent, the other being completely miscible. Computations for Complex HydrocarbonSolvent Systems The triangular equilibrium diagram may be employed as the basis for computation in complex oil-solvent systems. The actual method of computation is the triangular coordinate graphical method for the general case in ternary systems which has been described by Hunter and Nash (14) and also by Evans (6). The only difference in this method is that in the present case, instead of phase compositions being
0.911 0.898 0.887
0.911 0,904 0.893
with the refined oil, (3) the amount of extract, (4) the amount of solvent in the extract and yield of solvent-free extract oil. (5) the value of a physical property of the solvent-free extract oil, an'd (6) the number of ideal extraction stages required to give the desired refined oil. This graphical method enables these results to be computed with reasonable accuracy. Further, if such results are already available from a working unit utilizing a known number of actual extraction stages, then the number of ideal stages 100 VOL5.
+
OF STOCK
100 VOL5. OF SOLVENT
-
I
RAFFINATE NQI
EXTRACT N I (
+
I20 VOL5. OF SOLVENT
I
/
EXTRACT N O 2
EXTRACT N O 3
FIGURE 12. SCHEMEOF EXTRACTION FOR STOCK No. 2 WITH NITROBENZENE obtained by such a computation may be combined with this number and a n over-all stage efficiency evaluated, where the over-all stage efficiency is defined by E1 = Ni -
x
100
(2)
N O
in which N , = number of actual extraction stages Ni = number of ideal extraction stages Since this method of computation has already been thoroughly treated in the literature in connection with ternary
COUNTERCURRENT EXTRACTION OF LUBRICATING OIL STOCESWITH NITROBENZENE AT 10' C. TABLEV. THREE-STAGE
Oil Stock Treated Stock 3, V. G. C. 0.853
Stock A, V. G. C. 0.867 Stock B, V. G. C. 0.864 Stock C, V. G. C. 0.828
Vol. of solvent per 100 vel, Oil Stock 53.3 100.0 136.4 185.0 375.0 65.0 98.0 188.0 75.0 150.0 225.0 150.0
R a 5 n a t e Layer Solvent-free oil 88 vel. % ' 0: Vol. % of solvent-free original stock solvent in layer oil Actual Calcd. Aotual Calcd. 0.817 54.8 57.7 14.2 14.0 0.811 49.0 60.5 12.5 12.5 0.807 44.0 47.5 10.8 12.5 0.804 37.7 42.0 12.5 13.0 0.799 27.9 23.0 9.0 14.5 0.845 56.6 59.3 11.1 16.0 0,835 52.5 51.4 13.2 14.0 0.831 37.9 40.4 11.2 13.0 0.828 62.7 63.0 14.6 16.0 0.814 50.5 56.0 13.8 0.811 45.7 50.6 li:7 13.3 0.803 62.5 64.0 13.6 14.3
V. G. C. of
V. G. C. of solvent-free oil Actual Calcd. 0.899 0.901 0.895 0.897 0.889 0.895 0.884 0.890 0.874 0.879 0.897 0.898 0.902 0.901 0.887 0.891 0.925 0.925 0.917 0.928 0,909 0.918 0.869 0.863
Extract Layer Solvent-free oil @,VOl. % of original stock Actual Calcd. 45.2 44.0 51.0 49.5 56.0 52.5 62.3 57.0 72.1 72.0 43.4 41.4 47.5 48.2 62.1 60.5 37.3 37.0 49.5 50.0 54.3 48.6 37.5 39.5
7
Vol. % of solvent in layer Actual Calcd. 49.8 51.0 64.8 65.0 70.1 71.1 74.3 75.5 83.7 84.0 57.0 56.0 65.5 65.0 74.7 75.0 63.3 63.0 76.6 8O:O 82.0 78.9 77.5
INDUSTRIAL AND ESGINEERING CHEMISTRI
844
systems, it will not be necessary to describe it again here, Instead, the results obtained by applying the method to several oil-solvent systems will be discussed and compared with results actually obtained.
0
50
100
150
200
250
300
$50
400
The multiple extraction of a naphthenic-distillate lubricating oil stock No. 2 with nitrobenzene a t 10" C., the equilibrium diagram of which is given in Figure l l ( b ) , has been described by Ferris and Houghton ( 7 ) . The actual extraction process employed was as follows: 100 volumes of stock \yere treated with 100 volumes of solvent giving the first raffinate and extract. The first raffinate was then extracted with 120 17.01umes of nitrobenzene giving a second raffinate, and a n extract which was added to tlie first extract giving the second extract. The second raffinate mas finally treated with 173 volumes of solvent to give the third raffinate. The third extract was obtained by adding the second extract to that obtained in the last treatment. The scheme is shown diagrammatically in Figure 12. The experimental results from this extraction process are given in Table IV together with results calculated by the triangular coordinate graphical method using the equilibrium diagram of Figure l l ( b ) . Very satisfactory agreement is obtained except in the third stage. The rather wide divergence here is undoubtedly due to the fact that insufficient data were obtained accurately t o define the lowest part of the binodial curve, where the graphical con~tructionsfor the third stage are located. Since the experimental extraction was carried out in the laboratory, it has been assumed in making this comparison that equilibrium
TABLE
VI, OVER-ALL STAGE EFFICIENCY FOR AGITATEDSTAGE
A41R-
(Three air-agitated stages were required) V. G. C. of ?\To. of Percentage Vol. Solvent Solvent-Free Ideal Extn. Over-all per 100 Raffinate Stages Stage Lubricating Oil Stock Vol. Oil Oil Produced Required Efficiency Stock KO,2, 0.853 1'. G. C, 53,3 0.817
Stock A , 0.867 V. 0.c. Stock B, 0.864 V. G. C.
100.0 136.4 155.0 375.0 65.0 98.0 188.0 75.0 180.0 225.0 150.0
0.811 0.807 0,804 0.799 0.845 0,835 0.831 0.S28 0.814 0.811 0.803
2.42 2 05 1 .SO 1.53
80:6 68.3 60.0 51.0
2.43 1.53 2 TO 2,45 1.90 2.30
si:o
51.0 90 0 81 6 03.3 76.6
Stock C, 0.828 V. G. C. a Number uncertain owing t o insufficient tie line or binodial curve data.
had been attained in each experimental stage, whereas it is possible that such stages did not reach complete equilibrium, which mould in part be the cause of a certain amount of discrepancy between experimental and calculated data. In Table V the experimental and calculated results from a three-stage countercurrent extraction process refining different lubricating oil stocks with nitrobenzene a t 10" C. are
COL. 27, NO, a
compared. The oil stocks treated were a XIidcontinent semiparaffinic stock KO.3, V. G. C. 0.853, the equilibrium diagram of which is given in Figure l l ( c ) : stock A, V, G. 6. 0.867, the equilibrium diagram of which is given in Figure 1 I. (a); stock B,T'. G, e. 0.864; and stock C, TeG. C. 0.821. The equilibrium diagrams of the last) two oils, stocks B and C, are not given a$ they are similar to that of stock A, Theqe extractions !%ere carried out in the Atlantic Rcfining Company's laboratory and kindly communicated to the authors by S.KcFerris. The extraction conditions employed \yere as follons: The extracting vessels werp all approximately the same, being pear-shaped separating 450 funnels 7 inches (1'9 8 em.) in diameter a t llie top,
to systematize stirring. Temperatures were maintained constant in all three stages a t 10 0.5"C. In arriving a t tlie calculated yalues in Table T', the yields of solvent-free raffinate and entract oil, the volume per cent of
solvent in both layers, and the V. @. C. of the extract mere computed for the tseatmenl of the given stock wit11 the given volume of solvent to obtain a solvent-free raffinate oil having the 17- C C. shown in column 3. The agreement betmeeii actual and calculated results is satisfactory, and tlie average deviation is of the order of zt: 10 per cent. The maximum deviation was found to occur for the experiments T+here tlie constructional computation lines mere located in that part of the equilibrium diagram which was outside the range of the initial equilibrium epperirnents. A. material balance check on the experimental data used for constructing the equilibrium diagrams also showed that many of these data were only correct to within + 10 per cent Unless particular precautions are taken, the error in determining thote sections of the binodials n here one phase i s large and thr other phase comparatively sinal1 may easil) he cvcn larger. 713th very accurate equilibrium diagrams the deviation between the actual and computed data could probably be reduced to less than 1.5 per cent. The accuracy of the computations i s chiefly dependent on the accuracy of the experimental extractions used for constructing the equilibriuni diagram. The number of ideal stages required to reproduce thwe countercurrent extraction results as computed graphically is shown together with the over-all stage eficiency in Table VI* The over-all stage efficiency is apparently inversely proportional to the solvent-oil ratio employed, and 100 per cent efficiency is probably approached a t solvent-oil ratios
INDUSTRIAL AND ENGINEERING CHEMISTRY
JULY, 1935
which correspond to complete miscibility between the oil stock and the solvent. Unfortunately, insufficient data are available definitely to fix the effect of solvent-oil ratio upon the over-all efficiency, and these results are further complicated by the variable nature of the air agitation employed. The method of computation is exceedingly useful, particularly for forecasting the most favorable extraction conditions to produce a maximum yield of raffinate of a given V. G. C. The data resulting from such graphical calculations for the countercurrent extraction of stock No. 2, with nitrobenzene a t 10’ C. are shown in Figures 13 and 14. The volume per cent yield of raffinates of varying V. G. C. are plotted against the solvent-oil ratio. Maximum yields of the most desirable raffinate-that is, a raffinate possessing a low V. G. C.-are obtained a t low solvent-oil ratios. At these low solvent-oil ratios, however, a rather large number of ideal extraction stages is required to give a low V. G. C. raffinate. The number of actual stages necessary can be computed from the over-all stage efficiency. The variation of the over-all efficiency for an air-agitated stage as given in Table VI has been plotted in Figure 13. The best extraction conditions to give maximum yield of low V. G. C. raffinate reconcilable with a reasonable number of actual stages and the lowest solvent-oil ratio can easily be determined from these diagrams. Aclrnowledgment The authors wish to record their grateful thanks to the Atlantic Refining Company for permission to use the experimental extraction data with nitrobenzene as a solvent, and to S. W.Ferris of that company for his assistance and interest in this investigation.
845
Literature Cited (1) Am. Petroleum Inst. Bull., 14, 39-105 (1933). (2) Clotworthy, Ind. Chemist, 7,111 (1931). ENQ.CHEnc., 26, (3) Cornish, Archibald, Murphy, and Evans, IND. 397 (1934). (4) Evans, Ibid., 26, 439 (1934). (5) Ibid., 26,860 (1934). ( 6 ) Ferris, Birkhimer, and Henderson, IND.EXQ. CHEM.,23, 753 (1931). (7) Ferris and Houghton, Refiner, 11, 560 (1932). (8) Fischer, 2. tech. Physik, 10, 153 (1929). (9) Friedrichs, C h m . Fabrik, 23, 199 (1932). (10) Griffin‘,‘IwD.ENQ.CHEM.,Anal. Ed., 6 , 4 0 (1934). (11) Herz, Der Verteilungsatz,” Stuttgart. 1909. (12), Holleman, Chem. Weekblad,29, 762 (1932). (13) Hunter and Nash, J . SOC.Chem.Ind., 51, 285T (1932). (14) Ibid., 53, 95T (1934). (15) Hunter and Nash, World Petroleum Congr., London, 1933, Proc., 2, 340. (16) Jantzen, “Das fraktionierte Destillieren und das fraktionierte Verteilen,” Dechema Monographie, Band 5, Nr. 48, p. 81, Berlin, Verlag Chemie, 1932. (17) Jantzen and Tiedcke, Ibid., p. 115. (18) Kalichevsky and Stagner, ”Chemical Refining of Petroleum,” A. C. S. Monograph Series No. 63, New York, Chemical Catalog Co., 1933. (19) Kestner, IND. ENQ.CHEW,24, 1121 (1932). (20) Meyer, J . Inst. Petroleum Tech., 17, 621 (1931). (21) Othmer, Trans. Am. Inst. Chem. Engrs., 30, 299 (1934). (22) Saal and Van Dyck, World Petroleum Congr., London, 1933, Proc., 2, 352. (23) Smith, J . SOC. Chem.Ind., 47, 159T (1928). (24) Tuttle and Miller, Am. PetroleumInst.Bull., 14,85 (1933). (25) Underwood, Ind. Chemist,10, 128 (1934). (26) Underwood, J. SOC.Chem.Ind.,47, 805 (1928). (27) Watanabe and Morikawa, J . SOC.Chem. Ind. Japan, 36, 585B (1933). (28) White, J . SOC.Chem.Ind., 47,596 (1928). R B C E I ~ BFebruary D 23, 1935.
Graphical Correlation of Solvent Extraction Data SIMPLE method of plotting data obtained in the solvent extraction of petroleum hydrocarbon mixtures (Figure l), which has been used in this laboratory for some t h e , is presented in the hope that it may be of service t o others. This method of plotting has the following advantages: (1) It is not necessary t o d e h e the components of the extracted mixture. (2) Rectangular coordinate aper is used. (3) Experimental data may !e plotted directly. These advantages have made the method more convenient for this work than the three-component phase diagrams (6) which have been applied to data on the separation of petroleum hydrocarbon mixtures by means of selective solvents or precipitants (1, 2, 8-11).
STEWART S. KURTZ, JR. Atlantic Refining Company, Philadelphia, Pa,
The principle of this method of plotting is to graduate the x-axis of a graph in terms of a physical property which is additive on a volume per cent basis (density, refractive index, or viscosity gravity constant, 7) and graduate the y-axis from 0 to 100 volume per cent of solvent. A point called the solvent apex, C, is located on the graph so that its ordinate is 100 and its abscissa is approximately equal to the physical property (for example, viscosity gravity constant) of the stock. I n plotting data for an oil-solvent mix, a line is drawn through the solvent apex and the point representing the physical property of the solvent-free oil and points representing mixtures of the solvent with this oil must fall on ~~
=
TABLE I. BATCHNITROBENZENE EXTRACTION OF STOCK No. 2 AT 0” -------Undissolved Layer---Per -Properties of oilCent Per Saybolt VisPer Solrent cent VIBcosity- cent solof A. P. I. cosity, gravity vent in in hfixture stock gravity IOO~I?. constanb layer 33.3 71.0 23.3 487 0.853 20.8 50.0 58.8 26.0 399 0.837 17.0 75.0 38.0 28.9 327 0.819 14.6 ~
-Properties -Dissolved Per cent of A. P:I. stock gravity 29.0 14.3 41.4 13.2 59.0 15.5
of oilLayer-
c. Per
Saybolt Viscent viscosity- solcosity gravity vent in 100’ F. oonstant layer 1002 0.925 52.1 1433 0.935 68.0 1041 0.918 83.4
this line. _._ . . . . .
Procedure for Constructing Graph The data used for constructing Figure 1 are for the nitrobenzene batch extraction of a naphthenic distillate a t 0’ C . and are from the psper of ~ j + ~ ~ ~ i ~ and , Henderson TheBe data are given in Table 1.
(e),