INDUSTRIAL A N D ENGINEERING CHEMISTRY
1126
Vol. 16, No. 11
Fractional Distillation for t h e Separation of t h e Constituents of Petroleum' By W.A. Peters, Jr. E. I. DU PONTDE NEMOURS & Co.,WILMINGTON, DEL.
Then Y = a x , and a will The theory of fractional distillation with the mathematical formube nearly constant for all mixture of liquids by Ias used for calculating fractionating columns is summarized. values of X below, say, 0.01. fractional distillation, Experiments in which the formulas are checked in a small bubbler cap It is convenient to use this the degree of separation atplate column using known mixtures, and tests on mixtures of petroconstant a for many mixtained depends on two facleum distillates run in the same column are dcscribcd. A method tures for which Equation 1 tors-the amount of heat is suggested for calibrating fractionating columns to be used in pedoes not hold-for example, used per pound of product, troleum work in order to determine the proper balance between heat ethyl alcohol and water, and and the type of apparatus consumption, column eficiency, and the sharpness of cut to be obaniline and water. Values employed. These two factained. of a for various pairs of tors are not independent liquids are given in Table I. and the theoretical calculations and experiments to be described in this paper were It will be noted that the value of a is about 2 to 3 for the ordicarried out to show to what extent they are related. Unfor- nary separations that are carried out in column rectifiers. tunately, the apparatus was not exactly adapted to this work Mixtures for which a is 1.3 to 1.6 may be economically sepaand the results obtained in fractionating gasoline do not show rated in columns with twenty to thirty plates. For values of the relationship as clearly as they should. The separation a below 1.3 the separation becomes increasingly difficult and by fractional distillation of a mixture of many constituents, such as petroleum, cannot be analyzed mathematically, alI I I though some empirical Composition of D i ~ f i l l a f eIn Columr s N+l rules can be deduced of Different Efficiencies andDifr'erenf regarding sharpness of Values of H cuts to be obtained unA i n c h a r q e = 0.01:X, der various conditions. Therefore. it will be
I
N SEPARATING any
'I
I
mixture of two definite liquids whose properties are known. N-l T h e possibility of separating a mixture FIG.1 by fractional distillation depends on the difference in composition between the liquid phase and the vapor phase in equilibrium. Where there is no difference no separation can be made; where the difference is great the separation is easy. It has been shown that for normal mixtures, such as mixtures of isomers, homologs, etc., the relation between the compositions of the liquid and vapor phases can be expressed as a function of the vapor pressures of the two constituents.2
1
X
A in the liquid Y =concentration of A in the vapor K = the ratio of the vapor tension of B (the less volatile component) t o that of A
Let
= concentration of more volatile component
X
=
K
+ (1 - K)x
(1)
(All concentrations are expressed in terms of equivalent latent heats.)
For low concentrations of A , X will be small compared with K , and Equation 1 becomes y = -n K
(2)
For convenience we may call 1/K = a, where a is the enrichment ratio of one liquid, A , in dilute solution in another, B. 1 Presented before the Division of Petroleum Chemistry a t t h e 67th Meeting of t h e American Chemical Society, Washington, D. C., April 21 to 26, 1924. Peters. THISJ O U R N A L , 16. 402 (1823).
*
FIG 2
November , 1924
I N D U S T R I A L A N D ENGINEERING CHEMISTRY
more efficient columns must be employed. I n the equations used for calculating the separation to be made in a plate column, a comes in as a ratio in a geometric progression, so that as a approaches 1 the separation becomes extremely difficult. The separation of ethyl bromide and ether, for which a = 1.13, is many times more difficult than that of the nitrochlorobenzenes for which a = 1.23.
Different Efficienocs and Differeni Values of H
A
In
chorqe. X i 0 4 0 ond 0.95
.98
.97
sumption is measured by the factor H,which represents the amount of vapor passed up the scrubber or fractionating column per unit of distillate withdrawn. The difference between H and the unit of distillate withdrawn is refluxed or run back down the scrubber. The equations by which the separation to be made by simple distillation or fractional condensation (Methods 1 and 2) is calculated are wellknown and need not be given here.
't-
Composition of D i s t i l l o f e in Columns o f
1127
TABLE I-APPROXIMATEVALUESOF a FOR SOMETYPICAL MIXTURES, Boili?g point
LIQUIDB O c. Ethyl bromide 38.4 o-Nitrochlorobenzene 166.0 Methanol 64.6 Acetic acid 118.1 $-Nitrotoluene 157.0 Water 100.0 Methanol 64.6 Water 100.0 Water 100.0 Water 100.0 Water 100.0 Toluene 110.4 139.2 m-Xylene 139.2 m-Xylene n-Hexane 69.0 n-Heptane 98.4 %Octane 125.8
LIQUIDA Ether 9-Nitrochlorobenzene Methylethyl ketone Water o-Nitrotoluene Butyric acid Water Aniline Methanol Ethyl alcohol Acetone Benzene Toluene Benzene n-Pentane %Hexane %Heptane
Boiling point
"C. a 34.6 1.13 1.23 160.0 79.6 1.3 100.0 1.4 145.0 1.55 1.8 162.3 100.0 2.1 184.4 5.6 64.6 7.4 78.3 11.0 56.2 20.0 80.2 2.5 110.4 2.4 80.2 5.5 36.3 3.0 69.0 2.8 98.4 2.3
Av. pres: sure Mm.
760
so
760 760 80,
760 760 760 760, 760 760' 760 760 760 760 764
THETHEORETICAL PLATE T .95
Equations for calculating the separation to be made in a column or countercurrent scrubber have been developed and published many times. Most of these equations differ only in the nomenclature used and every method of calculation except that used by Baker, of the University of Michigan, starts from the theoretical plate as a unit and basis of calculation. Fig. 1 gives a diagram of a theoretical plate. We may define the theoretical plate as a plate on which all the liquid
x
B
l.00
P
:c:.90
9
.80
.70 Reflux
.6 0
Requlotor
Theoreticpl Plofes
FIG.3
METHODS Separation by fractional distillation is never complete. A partial separation by fractional distillation may be made by three methods: (1) boiling a comparatively large amount of liquid and collecting the vapors in different parts of fractions; ( 2 ) vaporizing the liquid and condensing parts of it in differDetail of Plete ent condensers, keeping these parts separate; (3) passing the Caps have 50 Slots 063* O l 6 c m Moqneaia Immersion of l o p o f Slots O b 3 G q vapor through a countercurrent scrubber in which it is Loqq1nq Cross section A r e a d S O S I ~ S ' washed or scrubbed by liquid refluxed down the scrubbers. Ssq cm I n general, the heat expenditure per unit of products sepaWeiqht of Section 7 54 K* lo5 rated is less for the first method, greater for the second, and Hold up on ploies 130 LC may be very much greater for the third. However, the separations obtained may be much sharper by the third method. For example, consider the separation of two liquids, A and B, for which a = 2. Assume concentration of A in B to be 1 per cent. By Method 1 the first distillate, which will be the richest, will contain only 2 per cent A , and the heat FIG. 4-PLATE COLUMN WITH SINGLE5.4 CM. BUBBLERCAP expenditure will be a minimum or 1 pound of vapor per pound of product. By Method 2, 75 per cent of the vapor must be condensed before the remaining vapor has a concentration of ,is of one composition and all the vapor passed throughisin 2 per cent A . By Method 3 the separation will depend en- phase equilibrium with this liquid, or, in order to avoid con7 tirely on the efficiency of the countercurrent scrubber used and fusion when dealing with packed towers, it mqy be defined a$ on the amount of heat supplied per pound of product. Fig. 2 the distance between two points in a column ohosen so that shows the relation between these factors. The heat con- a sample of vapor from the upper point y ~ s l dbe in phase
INDUSTRIAL A N D ENGINEERING CHEMISTRY
1128
equilibrium with a sample of liquor from the lower. This height may be called the H. E. T. P., or height of equivalent theoretical plate. The fundamental equation for a concentrating column is
x,+
1
HY, - Y, = H-1
TABLE11-EQUIVALENTHEIGHTOF THEORETICAL P ~ A T FOR E VARIOUS SCRUBBERS AND COLUMNS WITH VARIOUS L 1ourns4 H. E. T. P. LIQUIDMIXTURE COLUMNOR SCRUBBER Cm. Tower filled with 8 to 10 mm. glass rings Tower filled with 8 to 10 mm. glass rings Tower filled with 8 to 10 mm. glass rings Tower filled with 8 to 10 mm. glass rings Tower filled with 8 t o 10 mm. glass rings Tower filled with 26-mm. rings Tower filled with 76-mm. rings Tower filled with paving
Acetic acid and water Ethyl alcohol and water
(3)
X , Y,and H are defined above. Y , is the concentration of A in the distillate. As before, all amounts and concentra-
Benzene and toluene Nitric acid and water Ammonia and water Benzene and toluene
,161
I
Separation ofhcefoneond Methanol in Bronze ?olurnn
I
Vol. 16, No. 11
Benzene and toluene Benzene and toluene
43
14 t o 16 43 30 12.5 1200
300 finn
stones
--1
- _ _ _ _ _ I
Plate column, any type of good design Plate column any type of good design’ Plate column, any type of good design
Acetic acid and water Acetone and methanol Ethyl alcohol and water 4
Estimated.
2 plates
l ’ / z plates 11,’s
plates
I n Table I1 are given various values for the H. E. T. P. The subject of the measurement of and the explanation for variations in the H. E. T. P. cannot be considered here. It is probable that any scrubber or column separating petroleum mixtures would have the same H. E. T. P. as that given for acetic acid and water. Fig. 2 brings out two points: first, if H is small it does not help the separation to use an efficient column, and second, if I
90
I
I
I
I
lU
Hi62
I
I
I
I
1
Wvt 1.7*
Theoretical Plates FIG.5
tions are ekpressed in terms of latent heats. From this equation and Equation l an equation giving the height of column in units of equivalent theoretical plates required to make a given separation is derived.a I n general, it will be easier to work with Equation 3 and figure a column plate by plate. I n order to apply Equation 3 to a specific problem it is necessary to know the value of K or to have a curve giving the relation between X and Y and to know the H. E. T. P. Of course, another equally important factor is the capacity of the column under consideration, The capacity may be measured by the latent heat equivalent of the vapors that can be passed up the column per square foot cross-section area. This factor was not considered in the present experiments, but it should be remembered that the cost of a column, especially of a packed column, will depend on both the height the column is ineficient a very large heat expenditure does and the cross-section area, so that in comparing different - not help much. Fig. 3 shows t h e relation between these columns it is convenient to use a quantity which is the prod- factors for other ranges of concentration of the same two uct of the two and may be called the volume efficiency factor. liquids. It will be noted that for the higher concentrations of 8
THISJOURNAL, 16, 402 (1923), equation 12.
4
THISJOURNAL, 14,466 (1922).
INDUSTRIAL AND ENGINEERING CHEMISTRY
November, 1924
A this dependence of the separation on the proper balancing of the heat consumption and the column efficiency is not so marked as in the lower concentrations. For example, with a :still charge with 40 per cent concentration of A (a = 2) a
1129
with ethyl alcohol of high concentration were not trustworthy, owing to errors in the data on the relation between X and Y available. The tests with low concentrations of acetone in methanol and lower values of H were difficult to interpret,
Gasoline Frachonafedm Enqler Distillation Ranqe 6O"to 70' Cuts Ip H = 1 3 W t 27' B H = 22 E H - 51 W t 3 5 #
FIG.7
for reasons that are evident from an inspection of Fig. 5. I n order to simplify the curves all tests were reduced to correspond to a still charge with 1 per cent concentration of acetone by weight. The circles represent the concentration of acetone in different tests. The abscissas of these circles were determined by the point on the line for H = 03, where the average of ten EXPERIMENTALtests with the 4-plate column fell. This gave 2.7 theoretical I n order to deter- plates for the 4-plate column. It is obvious that the effimine the effect of ciency of the 12-plate column cannot be determined by tests variation of these two with any value of H less than 15 or 20. Even H = 10 gives a factors on the separa- curve which is nearly horizontal to the right of Plate 7. That tion of a complex is, from the data available, the vertical line on which the tests mixture, gasoline was on the 12-plate column are plotted might be anywhere to the fractionated in the right even of Plate 5 . This does not help to check the effismall rectifier shown ciency of the column, but is a good proof of the validity of the in Fig. 4. The recti- equations on the basis of which the statement was made that fier was first checked for certain separations with a given heat expenditure it does on mixtures of ace- not pay to increase the efficiency of the column beyond a tone and methanol, certain point. A good grade of gasoline which appeared to be straight-cut acetic acid and water, and ethyl alcohol and was then fractionated. Several difficulties were encountered water. , All the tests in making these fractionations in an entirely quantitative that looked reason- manner. First, the still would hold only about 15 gallons to able gave the same a charge. During the distillation of this charge, the jacket H. E. T. P. a s t h a t temperature could be controlled so that the radiation losses shown in Table 11, were negligible, but the column itself could not be heated. t h a t is, a b o u t 2 Therefore, its temperature had to be raised by condensing a c t u a l p l a t e s f o r gasoline vapors, and the vapors so condensed gave a variable acetic acid and water, amount of reflux from the top to the bottom of the column. The value of H in Figs. 6 and 7 is measured a t the center of the 11/2 for methanol and acetone, and ll/a for column. A second and much more serious error was introe t h y l a l c o h o l and duced by the hold-up of liquid on the plates. Four plates w a t e r . T h e t e s t s held about 450 grams of liquid and 12 plates about 1350 grams.
distillate of pure A can be obtained if H = 3.5 or more, provided the column is efficient enough. With a still charge with 95 per cent concentration of A , H need be only 2.05 to give a distillate of pure A w i t h a n efficient enough column.
FIG.8
I N D UXTRIAL A N D ENGINEERING CHE&fIfJTRY
1130
All the fractions taken ran less than 1800 grams; therefore the separations, especially in the 12-plate column with the high reflux ratio, are very much worse than would be obtained under truly continuous conditions of operation. For this reason, the curves obtained cannot be considered as more than qualitative and a comparison of Curves V and VI in Fig; 7 is misleading. The amount of pentanes in Fractions 111, V, and VI was determined by careful fractionation in a special laboratory column. The results for Fractions I11 and V in comparison with the range obtained in the regular Engler distillation are plotted in Fig. 8. In all the distillations there was a loss due to not condensing some of the vapors. Cooling with liquid air would have been necessary to make the recoveries strictly quantitative, and this was not available. The losses in the Engler distillations were distributed. Those in the special column were neglected. For that reason, the curves in Fig. 8 do not line up exactly. COMPARISON OF OBSERVED AND CALCULATED RESULTS The amounts of pentanes which should be in fractions of hexane and higher oils cut between 60' and 70" C. under different conditions were calculated approximately. These amounts with those actually found are given in Table 111. (The H. E. T. P. is taken as two actual plates.) TABLE 111-AMOUNT OF No. actual
PENTANE I N A
plates in column
H
4 12 12
2
5 5
60'
70'
TO
-PER CENT Caldulated
c. CUT O F GASOLINE
PENTANB-
2.5 8.0 2
Found
19 15 9.5
Vol. 16, No. 11
There are several explanations for the large discrepancies between the amounts calculated and found. The calculation was made for a single hexane boiling at 6.5' C. The presence of others boiling around 60' C. would make a difference. Moreover, as mentioned before, the hold-up on the plates of' the column would introduce a serious error. Unfortunately, there was not time to clear up these points. A comparison of Curves I1 and V is shown in Fig. 9. Evidently there is a considerable gain in sharpness of cut in increasing the efficiency of the column 200 per cent. However, the difference cannot be directly correlated with the theory as was done in the case of the mixtures of acetone and methanol. The data only indicate the results that may be obtained by experiments made under the proper conditions. CONCLUSIONS
It is safe to say that many of the scrubbers or fractionating columns now in use in petroleum refineries are not so efficient as they should be. The most economical installation for any given separation can probably be most easily determined by experiments similar to those described. These experiments would, of course, include checking up and calibrating the small columns used and also the plant columns by the theory outlined. Comparatively simple tests would show approximately the heat expenditure needed and the efficiency of the column required to make certain rough cuts. The accurate determination of the economical balance between heat consumption, column efficiency, and sharpness of cut for each product is a more complex problem, but one for which a definite solution can be worked out.
Notes on the Preparation of Standard Cellulose-II'~z By A. B. Corey and H.LeB. Gray EASTMAN KODAKCo., ROCHESTER, N. Y.
I N A previous paper3 it was noted that the standard cellulose prepared by the tentative method described, gave too high an ash. It has been found that if the resulting cellulose, before drying, is treated with 1 per cent acetic acid solution for 2 hours a t room temperature and then washed with four changes of distilled water, the ash content is reduced to a very low quantity without any apparent injurious effect on the cellulose. Table I gives results obtained on two different samples. In each case the cellulose prepared as previously recommended was divided into two portions and one of them treated with acetic acid as described above.
The writers have observed, during the alkali boiling of different cottons, that with some a longer time is necessary to discharge the yellow color of the solution than with others. Table I1 shows that the percentage of a-cellulose may b e slightly increased by boiling with the 1 per cent sodium hydroxide solution for 10 hours instead of 6. It seems advisable to allow the displacing alkali solution to enter a little more rapidly' so that 12 liters are used during the time of boiling; and also to continue the heating and displacement with 1 per cent alkali for 1 hour after the yellow color has entirely disappeared. TABLE11
TABLE I
Wan namaker's Cleveland
Without acid treatment
American Peeler (Combed sliver)
Without acid treatment
With acetic acid treatment
With acetic acid treatment
Moisture Ash a-Cellulose Per Per Per cent cent cent 0.12 99.66
0.13 0.06 0.02 0.11 0.11 0.03 0.04
99.71 99.85 99.71 99.61 99.74 99.54 99.76
Received September 24, 1924. Communication No. 216 from the Research Laboratory of the Eastman Kodak Company. 8 THIS J O U R N A L , 16, 853 (1924). 1 2
6 hours' alkali treatment and acetic acid treatment 10 hours' alkali treatment and acetic acid treatment
Moisture Ash u-Cellulose Per Per Per cent cent cent 0.06 99.69
0.00
{ ;: E
0.07
0.04 0.04
99.48 99.48 g9.S4 99.85
The apparatus used by the writers is the she recommended: by the A. C. S. committee4and 76 grams of cellulose are used instead of 100 grams. The corrected capper numbers of the cottons, while relatively low, are not recorded in the accompanying tables on account of the erratic results obtained, which are apparently due to errars in the method oE determination as given by the committee. 4
THISJOURNAL, lS, 748 (1923).
