Pure Hydrocarbons from Petroleum

that close-boiling paraffin-naphthene mixtures may not safely be assumed ideal. A fractional distillation method is pre- sented as the most reliable m...
0 downloads 0 Views 688KB Size
Pure Hydrocarbons from Petroleum Vapor-Liquid Equilibria of Close-Boiling Hydrocarbons’ JOHN GRISWOLD

degrees. When a is constant or changes but slightly so that an average value may be used, the number of theoretical plates required t o effect a given separation may be calculated by the equation of Smoker (IS), which for total reflux reduces to the simpler equation derived by Fenske (6):

The University of Texas, Austin, Texas

Close-boiling hydrocarbon mixtures deviate from ideality in both vapor and liquid phases to the extent that the relative volatility (a = P A / P B ) calculated as the ratio of vapor pressures, may not be relied upon. Deviations from the ideal gas law and from Raoult’s law usually oppose each other in their effects on the relative volatility. The system n-heptane-methylcyclohexane deviates appreciably from ideality of the solution; this indicates that close-boiling paraffin-naphthene mixtures may not safely be assumed ideal. A fractional distillation method is presented as the most reliable means of verifying vapor-liquid equilibrium data, and’ for determining the relative volatility of the close-boiling hydrocarbon systems in which a is substantially constant. Reexamination of the vapor-liquid equilibrium and fractional distillation data on the n-heptane-methylcyclohexane system both gave the same value: a = 1.083. A determination on the n-heptane-isooctane system gave the value: a! = 1.023.

where XI = mole fraction of a component at plate 1 X N = mole fraction of the component at plate N 01 = relative volatility When the still-pot composition is used as XI, the exponent of a becomes N f 1. As an illustration of the importance of knowing exact values of a for close-boiling mixtures, suppose it is desired to fractionate a 50 mole per cent mixture of n-heptane and isooctane (2,2,4-trimethylpentane) to obtain 99 per cent pure n-heptane. Using relative volatility values of 1.052 and 1.023, the calculated numbers of theoretical plates are 91 and 202.5, respectively. Since, in practice, 2 to 10 actual plates a t partial reflux are required for each theoretical plate a t total reflux, seemingly small differences become significant when a is close to unity.

Calculation of a for Ideal Systems The relative volatility of component A to component B in a mixture (16) is, by definition,

where Y and X are mole fractions or mole percentages in equilibrium vapor and in liquid phases, respectively. When the ideal gas and solution laws apply, Equation 2 simplifies to ffideal

T

H E problem of separating pure hydrocarbons from petroleum fractions requires an accurate knowledge of their behavior on distillation. The previous paper (8) discussed the complicating effects caused by the presence of aromatics. These aromatics may be removed from a petroleum fraction by any of several processes; then resolution of the residual paraffin-naphthene mixture is much easier. A valuable property of close-boiling paraffin and naphthene mixtures is the constancy of the relative volatility, a,between two of the hydrocarbons over a wide range of concentrations, as long as the temperature does not change more than a few 1

PA/PB

(3)

where P A and P B are the vapor pressures of the pure liquid components. When the system is ideal, a may be calculated directly from vapor pressures. However, accurate vapor pressure data have not been obtained for all hydrocarbons which one may wish to separate from petroleum, and an equation relating normal boiling temperature to relative volatility will be a real convenience. Edgeworth-Johnstone (6) derived the relation for hydrocarbons a t atmospheric pressure: (4)

The temperatures are the atmospheric boiling points expressed in an absolute temperature scale. The ClausiusClapeyron equation and Trouton’s rule with a constant Qf 22.0 were assumed. The constant should be modified to obtain best accuracy for close-boiling hydrocarbons, giving TA) loglo 01 = 4.85 ( T B 2’

The first paper in this series appeared on pagea 117 to 119, January,

1943, issue. z!47

(44

248

INDUSTRIAL AND ENGINEERING CHEMISTRY

Vol. 35, No. 2

A formula which gives somewhat better accuracy for hydrocarbons boiling within 5” C. of one another is: a0=1+

TB - TA 0.0633 T

10.56

+

(5)

Equation 5 was derived by fitting an empirical linear temperature equation to the values of d t / d p for pure hydrocarbons given by Brooks ( 3 ) . For wider boiling compounds, the formula is proposed: ciw =

1

+

cia

+ 2.3 log T ) (TB -

(4.4

TA)

in which the temperatures are expressed in O K. Additional approximate relations incorporated into Equation 6 are the Kistiakowsky equation for latent heat of vaporization and the Clapeyron equation. For wide-boiling compounds, Equation 5 predicts relative volatilities somewhat low, while Equation 6 gives values slightly high.

Deviations from Ideal Gas and Solution Laws The relative volatility is strictly equal to P A / P Bonly when the laws of Amagat, Boyle, Charles, Dalton, and Raoult all apply rigorously. This is almost never the true case, yet the relation a = P A / P Bis sonietimes a satisfactory approximation. DEVIATIONS mo&iIDEAL GAS LAW. Inspection of vapor density and P-1’-T data for pure six-carbon to eight-carbon hydrocarbons (9, 14) near their normal boiling points shows gas law deviations ranging from 4 to 10 per cent ( p = PV/RT = 0.96 to 0.90). These deviations result in vapor concentrations higher than those given by the gas law, and a is affected unless the deviations happen to be equal, or P A = P B . The effect of gas lam- deviations on 01 may be eliminated by the substitution of fugacities for vapor pressures (IO), and this procedure works well at high pressures. When fugacities are used a t atmospheric pressure, the result is a calculated a which is usually lower than both P A / P Band the experimental vapor-liquid equilibrium value. This anomaly is explainable on the hypothesis that “ideal” hydrocarbon solutions are rare. I n general, the gas lam deviation of hydrocarbons is negative, and that of the more volatile component of a binary mixture (component A ) is somewhat greater in magnitude than that of compoiient B. Likewise, Raoult’s law deviations (activities) of hydrocarbons are most frequently positive, and the activity of component A is often somewhat greater than that of component B. Segative gas law deviations and positive activities oppose each other in their effects on a ; and in some cases, although neither effect is negligible, the net effect is so slight that the

P L l N T FOR SEP.4RATION O F P U R E HYDROCARBONS FROX NATURAL G.4SOLINE FR.4CTIOSS

system appears to be ideal with respect to the relation CY = PA/PB. This situation may be detected by comparison of observed total pressures of mixtures with the sum of calculated partial pressures, as mas done by Beatty and Calingaert (1). It does not follow that a = P A / P Bwhenever observed total pressures approximate ideal values.

Verification of Binary Equilibrium Data

The Duhem-Rlargules equation may be used to detect incongruities in vapor-liquid equilibrium data. Beatty and Calingaert (1) used it t o develop a number of criteria for testing such data. By relations so developed and close scrutiny of experimental data, they found certain jnconsistencies and experimental errors in most of the systems investigated, and concluded that such data were unreliable. They also developed a method of calculating vapor-liquid equilibria of nearly-ideal solutions by constructing synthetic partial pressure curves from total pressure measurements by assuming the ideal gas law and “that the deviations of these pressures from the ideal lines vary with X in a more or less uniform manner”. This procedure gave values of a which are constant or independent of concentration. Of the relative volatilities so calculated by the same authors, the benzene-cyclohexane azeotrope was undetected and this article will show that the values reported for the n-heptanemethylcyclohesane and the n-heptane-isooctane (2,2,4trinkthylpentane) systems are in serious disagreement nith exnerimmtal . ~ ~ vapor-liquid equilibria and rectificaTABLE I. REL.4TIVE HETP VALUES5 tion data. Hence, although themethod may detect deviations from ideality Citation Systems with Citation Systems with No. Acaordant Values KO. Discordant Values of the solut’ion, its merit for predicting relative volatilities of close(7) A. Carbon tetrachloride-benzene (16) A. Carbon tetrachloride-benzene E.Benzene-toluene E . Heptane-methylcyclohexane boiling hydrocarbons has never been (4,7 ) A. Carbon tetrachloride-benzene (7, 16) A. Carbon tetrachloride-benzene established. B. Heptane-toluene B. Heptane-toluene Since the use of equilibrium data (7) A. Benzene-toluene (7) A . Heptane-toluene is to comput’e separations obtainB. Heptane-toluene B. Heptane-methylcyclohexane (16) A. Carbon tetrachloride-benzene (16) A. Methylcyclohexane-toluene able by fractional distillation, direct E. AMethylcyclohexane-toluene B. Heptane-methylcyclohexane data may not be dispensed with until another method is available Comparative data for two systems obtained on the same fractionating column at total reflux over the same range of boiling rates. which shows positive agreement with fractional distillation data. An obvi-

-

~

~

~

.

February, 1943

INDUSTRIAL AND ENGINEERING CHEMISTRY

ous method of verifying vapor-liquid equilibrium data is to compare it with the results of fractional distillation.

Verification of Vapor-Liquid Equilibria by Fractional Distillation Data If it is granted that the diffusional theory of fractionation is correct, the rate a t which vapor-liquid interaction proceeds in a packed or bubble-tray column (operated a t total reflux) will be nearly the same for systems of approximately equal viscosities and normal boiling ranges a t equivalent vapor rates. This condition is manifested in packed columns by equal and constant values of the height equivalent of a theoretical plate (HETP) or height of a transfer unit (HTU) for different systems run in the same column, and by equal plate efficiencies for different systems in the same bubbletray tower. Modern data suitable for comparison are reported in several articles (4, 7, 16). Equivalent HETP values were obtained in a number of the tests, as well as a number of divergent values. All systems on which comparisons could be made are listed in Table I under “Accordant” or “Discordant” columns or both, as the case may be. There are five binary systems in the “Discordant” column. Of these, the carbon tetrachloride-benzene system has been subjected t o a great deal of experimental study over a period of years, and there is scarcely room for doubt about the essential accuracy of its vapor-liquid equilibrium. It is further verified by its frequent presence in the “Accordant” column. On the basis that the vapor-liquid equilibrium of this system and the fractionation data are correct, vapor-liquid equilibrium of the system n-heptane-toluene and the relative volatility value of 1.07 for n-heptane-methylcyclohexane are open to question. The n-heptane-toluene and carbon tetrachloride-benzene pair appears in both “Accordant” and “Discordant” columns of Table I. Of the data suitable for comparison, there was a total of thirty-five experiments on n-heptane-toluene in four fractionating columns. Vapor-liquid equilibrium data used in all cases were those determined by Bromiley and Quiggle (2). Inspection of the thirty-five experiments revealed twenty giving accordant HETP values and fifteen giving discordant values. In every case the distillate composition in the accordant runs was below 80 mole per cent n-heptane whereas the distillate composition of the discordant tests was above 80 mole per cent n-heptane. It must, therefore, be concluded that the vapor-liquid equilibrium data are essentially correct below 80 mole per cent n-heptane but appreciably in error a t concentrations above that figure. Since the system is nonideal, the relative volatility may not be assumed constant and independent of X . The discordant HETP values are low; this indicates that y/s above 80 per cent heptane is too low, or that the system deviates from ideality in this zone less than the vapor-liquid equilibria indicate. The methylcyclohexane-toluene results are slightly out of line in the same direction as those of n-heptane-toluene. The distillate composition was above 90 mole per cent methylcyclohexane in each of the four experiments on this system, which indicates that the vapor-liquid equilibrium are also in error, possibly a t high heptane concentrations only. The error in both systems is explainable by the hypothesis that the charge to the equilibrium still contained quite small amounts of benzene as an impurity, which would exert a greater and greater analytical effect on the equilibrium as the concentration of heptane or methylcyclohexane increased. As the previous paper (8) showed, benzene forms azeotropes vith six-carbon hydrocarbons and is likely to behave abnormally with seven-carbon hydrocarbons other than toluene.

249

Vapor-Liquid Equilibrium of n-Heptane-Methylcyclohexane Table I indicates an appreciable error in the vapor-liquid equilibrium of n-heptane-methylcyclohexane. This system is of particular interest, since it is widely used for testing laboratory columns. The only vapor-liquid equilibrium data are those of Bromiley and Quiggle ( 2 ) . Beatty and Calingaert (1) found errors of various magnitudes in these data and criticized them as inaccurate, inconsistent, and unreliable. The latter authors recommended the relative volatility of 1.07 derived from their total pressure measurements, and this figure was then accepted with no supporting experimental verification. I n order to arrive a t the most dependable relative volatility, values will be calculated from both equilibrium and fractional distillation data. IDEAL-SYSTEM VALUE. The vapor pressure curve for n-heptane has been accurately determined by Smith (12). The vapor pressure ratio of n-heptane to methylcyclohexane a t the accepted boiling point of the latter (100.8” C.) is 1.0694. Equation 5 indicates an increase of 0.0003 a t the boiling point of n-heptane (98.42” (2.). T

I

\

DATA Or BROMILCY AND

MOL

FIGURE1.

O/-

QUIGGLC

I

LI.l.l

I t

N - HEPTANC

RELATIVEVOLATILITY

O F n-HEPTANE-METHYLCYCLOHEXANE

RE-EXAMINATION OF EQUILIBRIUM DATA. The relative volatility was calculated for each experiment of Bromiley and Quiggle’s tabulation. The determinations in which the heptane concentration was less than 5 or greater than 85 mole per cent were rejected in order to minimize the effect of analytical error. (This will be discussed later.) The average of the eighteen determinations of intermediate concentration was 1.083, with an uncertainty of 0.002. The authors made an unfortunate choice in the use of density rather than refractive index for analysis, since use of the refractive index would have greatly reduced analytical error. The experimental values of CY are plotted on Figure 1. Except for the ends of the curve (at which errors of analysis are greatly magnified) the data are remarkably consistent. The hypothetical effect of a density error of -tO.OOOl on a relative volatility of 1.08 is shown by the dashed curves, and indicates that a precision of about *0.00005 in the density determination was realized in the great majority of the analyses. The data show a slight trend of CY, increasing with n-heptane concentration. This would result if the heptane contained a trace of low-density, volatile impurity. As an illustration, suppose the true CY is 1.08 and independent of concentration. If the heptane was 99.9 mole per cent pure and contained 0.1 mole per cent n-hexane, CY observed from exact density analyses would form the solid curve shown on Figure 1. The physical constants of the materials used in these experiments were within analytical agreement of more

250

INDUSTRIAL AND ENGINEERING CHEMISTRY

Vol. 35, No. 2

recent values. These constants with the purification history of both hydrocarbons indicate that the materials were as pure as any likely to be used for purposes of fractional distillation.

equivalent theoretical plates in the column from the carbon tetrachloridebenzene runs is (33.4 - 1) = 32.4. In Table IIB the same column packed with one-turn helices, was used, and a trend of HETP with boiling rate is apparent. Experiments a t comparable rates are runs 1-4 on n-heptanemethylcyclohexane and runs 3-5 on carbon tetrachlorideT ~ B L11. E RELATIVEVOLATILITY OF TL-HEPTAKE-METHYLbenzene, since in all of these the "liquid velocity" was beCYCLOHEXANE FROM FRACTIONAL DISTILLATION DATA tween 5.2 and 5.8 liters per hour. The number of equivalent R~~ Mole % n-Heptane ~$~~~~~~ theoretical plates in the column averaged (19,O - 1) = 18.0 N o . Top, L./Hr. Distillate Still Equation 1 from carbon tetrachloride-benzene runs 3-5. From Table IIA (eight observations), as".= 1.084, while Table IIB (four A. U-Type Packing, 32.4 Theoretical Plates 25.5 1,080 observations) gave aav.= 1.080. 1 1.8 80.5

~

2

3 5

6 7 8

/

2.0 2 , 1. 2.2 2.3 4.6 4.8 5.0

~

~

~

~

84.0 90.0 81.0 82.0 76.0 75.0 80.0

24.0 1.090 24.0 1.108 22.8 1,086 1.081 26.6 20.5 1.081 1.069 25.6 28.0 1.075 Average 1.084 B. One-Turn Helices, 18.0 Theoretical Plates 51.0 20.7 1,080 47.0 19.3 1.078 5.6 48.0 18.0 1.0S3 5.8 45.8 17.5 1.080 Average 1 . 0 8 0

f

Relative Volatility of n-Heptane-Isooctane

The system n-heptane-isooctane (2,2,4-trimethylpentane) should be nearly ideal, since it consists of two paraffins of nearly the same boiling point. Smith (12) determined accurate vapor pressure curves on highly pursed samples of each hydrocarbon. From his data 01 = PA/PB= 1.0237 1 !;$ 2 a t atmospheric pressure. 3 4 Samples of pure materials (tested a t the Xational Bureau of Standards) were obtained from the California Chemical Company (n-heptane) and TABLE 111. RELATIVE VOL-~TILITY OF ~-HEPTANE-~SOOCTANE BY FRACTIOSAL DISTILLATIOX ;\IE,IHOU Rohm & Haas Company (2,2,4-triLiquid Distillate Still P o t KO.of Theoretical methylpentane). Since the relative Rate a t Time of Top of KO. of AIole c/o Mole %, ______ Plates volatility of thess hydrocarbons is of Run, Column Theoretical n-hydron-hydroa = = close t o unity and their gravities and ours cC./H~.' Plates n? carbon 72? carbon 1.07 1.083 refractive indices also lie close toA. Equivalent Theoretical Plates from n-Heptane-hIethylcyclohexane gether, the analytical precision of a g3.7 54.1 8.5 321 63(11) 1.39068 90.4 density or refractive index determina27.0 273 73 ( 2 1 ) 1.38062 03.8 1.41868 11.2 10.7 60.1 tions obtainable with instruments of B. n-Heptane-Isooctane 4 312 55.2b 1.38898 69.8 a 1.0228C the conventional types would render S 312 55.2b 1.38898 69.8 1.39014 40.0 1 0228C vapor-liquid equilibrium determinations futile. Known Mixtures of n-Heptane-Isooctane LIole % n-hydrocarbon 0 12.1 17.6 24.3 36.3 36.2 51.9 The relative volatility was denko 1.39160 1.30110 1.39092 1.39071 1.39032 1.39023 1.38967 termined by the fractional distilla3Iole % n-hydrocarbon 53.5 54.6 57.1 60.3 65.8 67.7 74.8 tion method, using a high-temperan ho 1.38962 1.38953 1.38947 1.38943 1.38913 1.38907 1.38883 ture Podbielniak apparatus equipped ,Mole % n-hydrocarbon 77.3 80.3 82.0 82.3 89.8 100 with an ll-mm. distilling tube conn 22 1.38868 1.38858 1.38845 1.38852 1.38818 1.38870 taining 36 inches of Heli-Grid packa Still pot not sampled until end of runs; final analyses used: less t h a n 0.5 cc. distillate withdrawn. ing (11). Analysis was by refracb Using same column a s in A , and a = 1.083 for n-heptane-methylcyclohexane. Relative volatility a calculated by Equation 1. tive index, using a Bausch & Lomb precision-type oil refractometer, accurate t o 0.00003 unit. The data are given in Table 111. The experiFRACTIONATION DATA. The equivalence of experimental mental value of 01 was 1.0228. An error of one theoretical HETP values obtained on a given column with different plate would affect this value by *0.0004. liquid mixtures is sufficient evidence that such data may be Lsed t o calculate a: for a system, once the number of equivaSummary lent theoretical plates in the column has been established by data on some other system of known behavior. Two The data and observations on close-boiling binary hydrosets of data by which a for n-heptane-methylcyclohexane carbon systems are summarized in Table IV. The recommay be calculated are given in Table I1 (14). The fracmended relative volatilities of n-heptane-methylcyclohexane tionating column was 32 mm. i. d. and had 2.74 meters of and n-heptane-isooctane (2,2,4-trimethylpentane) systems packing. The second system was carbon tetrachloride-benzene, which was more extensively used and gave more consistent results in these TABLE IV. RELATIVE VOLATILITIESOF CLOSE-BOILING HYDROCARBON SYSTEMS experiments than any other sysMethod of Determining a Total Ideal Ideal Ideal VaporFraoRecomtem. presforsystem soluliquid tional mended For the data of Table IIA, a USystem surea mulab ( P a / P a ) tione equilibrium distn. Value of a type packing was employed. The reBenzene-cyclohexane 1.021 ... . . . Azeotrope 1.5544 . * . n-Heptane-methylcyclosults show no consistent trend with hexane 1.07 1.0693 1.0694 1.0650 1.083 1.080 1.083 vapor velocity and so are directly n-Heptaneisooctane 1.052 1.0237 1.0237 1.0222 ... 1.0228 1.023 a Beatty and Calingaert (1, Table I). comparable. -Rejecting runs 5 an; b From Equation 5, 8 as deviating too greatly from the 0 Fugacity calculation (IO). other six, the average number of o(

(I

~

0

7

...

February, 1943

INDUSTRIAL A N D ENGINEERING CHEMISTRY

for use in fractional distillation are 1.0832and 1.023, respectively. The former system was found to deviate appreciably from an ideal solution. From the information now available, systems of sufficient ideality to permit valid use of the relation a = P A / P Bare limited to close-boiling mixtures of the same hydrocarbon class-i. e. paraffin-paraffin, naphthene-naphthene, and aromatic-aromatic compounds. Analytical error in vapor-liquid equilibrium determinations becomes pronounced a t extremes of concentration, and consistent errors which may be due to slight impurities are most pronounced a t the low-boiling end of the curve. Such errors do not of themselves justify rejection of data on intermediate concentrations. The fractional distillation method is proposed as the most reliable to date for verification of vapor-liquid equilibrium data and determination of a for close-boiling hydrocarbon systems in which it is essentially constant.

Acknowledgment The Bureau of Industrial Chemistry of the University of Texas loaned the fractionation equipment. The relative vola2 The multiplier t o convert theoretical plates from the basis of ia 0.86,and the corresponding HETP multiplier is 1.176.

~1

=

1.07

251

tility of n-heptane-isooctane was determined by E. E. Ludwig.

Literature Cited (1) Beatty and Calingaert, IND. ENG.CHEM.,26, 504, 905 (1934). (2) Bromiley and Quiggle, Ibid., 25, 1136 (1933). (3) Brooks, Howard. and Crafton, J . Research Natl. Bur. Standards, 24, 33 (1940). (4) Bruun, J. H., IND. ENG.CHEM.,ANAL.ED., 8, 224 (1936). (5) Edgeworth-Johnstone, J. Inst. Petroleum Tech., 25, 558 (1939). (6) Fenske, M. R., IND.ENC.CHEM.,24, 482 (1932). (7) Fenske, Tongberg, and Quiggle, Ibid., 26, 1169 (1934). (8) Griswold and Ludwig, Ibid., 34, 117 (1942). (9) Kelso and Felsing, Ibid., 34, 161 (1942). (10) Lewis and Luke, Ibid., 25, 725 (1933). ENQ.CHEM.,ANAL.ED.,13, 644, Fig. (11) Podbielniak, W. J., IND. 8 (1941). (12) Smith, E. R., J . Research Nat2. Bur. Standards, 24, 299 (1940). (13) Smoker, E. H., Trans. Am. Inst. Chem. Engrs., 34, 165 (1938). (14) Timmermans, Jean, in International Critical Tables, Vol. 111, p. 244, New York, McGraw-Hill Book Co., 1928. (15) Tongberg, Quiggle, and Fenske, IND.ENQ.CHEM., 26, 1213 (1934). (16) Walker, Lewis, McAdams, and Gilliland, “Principles of Chemical Engineering”, 3rd ed.. p. 521, New York, McGraw-Hill Book Co., 1937.

Activated Carbon Treatment of Raw Whiskv J

G. C. WILLIAMS AND E. A. FALLINl University of Louisville, Louisville, Ky.

The chemical changes and changes in taste characteristics that take place when a distillate from an alcohol fermentation process is treated with each of several activated carbons are reported. The effect of one carbon was investigated over a temperature range from 20’ to 80’ C. Tables of results are included, and figures illustrate the percentage change effected by the treatment.

I

N DISTILLATION industries, adsorption methods are

often used to remove impurities, p.articularly those present a t relatively low concentrations. These methods are used although very little is known of the actual chemical changes effected by the treatment. I n a liquid mixture or solution, some constituents are preferentially attracted to the surface of different adsorbents to the almost complete exclusion of others. The program described in this paper was to determine some of the more important chemical changes brought about in the raw whisky by different kinds of activated carbons, and to determine the effect of temperature on one of the carbon treatments. Taste preference was utilized to show a correlation 1

Present address, M. F. A. Milling Company, Springfield, Mo.

At room temperature it was noted that the change in acids, esters, aldehydes, and fusel oil was slight but not consistently in one direction. Permanganate time was increased in all treatments and to approximately the same extent. However, the trkatment at elevated temperatures definitely produced an increase in acids and permanganate time, while it indicated decreases in ester, aldehydes, and fusel oil. with the extent of these chemical changes. Such a relation would substantiate the effects of the various treatments. The chemical changes involved are many and complex. Different amplitudes are affected by the kinds of material used as containers during storage, as shown by laboratory investigations of Valaer and Frazier of the Bureau of Internal Revenue (6). They found upon storing raw whisky for four years in charred white oak barrels that the most rapid changes of acids, esters, solids, and color took place in the first six months. During the four years acids increased from 24.9 to 56 mg. per 100 ml. while esters increased from 7.4 to 21.3 mg. per 100 ml. Fusel oil content, however, dropped from 69 to 58.4 mg. per 100 ml. It was also found that raw whisky changes while standing