March 1953
589
INDUSTRIAL AND ENGINEERING CHEMISTRY
(4) Brown, H. C., Pearsall, H., and Eddy, I . P., Ibid., 72, 5347 (1950). (5) Egloff, G., Hulla, G., and Komarewsky, V. I., “Isomerization
of Pure Hydrocarbons,” New York, Reinhold Publishing Corp., 1942. (6) Evering, B. L., Fragen, N., and Weems, G. S., Chem. Eng. News, 22, 1898 (1944). (7) Evering, B. L., and d’ouville, E. L., J . Am. Chem. SOC., 71, 440 (19491. (8) Grummitt, O., Sensel, E. E., Smith, W. R., Burk, R. E., and Lankelma, H. P., Ibid., 67, 910 (1945). (9) Heldman, J. D., Ibid., 66, 1786 (1944). (10) Ipatieff, V. N., and Grosse, A. V., IND. ENG.CHEM.,28, 461 (1936). (11) Ipatieff, V. N., and Schmerling, L., Ibid., 40, 2354 (1948). (12) Komarewsky, V. I., and Ulick, S. C., J. Am. Chem. SOC.,69, 492 (1947). (13) Leighton, P. A., and Heldman, J. D., Ibid., 65, 2276 (1943). (14) Lien, A. P., d’Ouville, E. L., Evering, B. L., and Grubb, H. M., IND. ENG.CHEM.,44, 351 (1952). (15) McAllister, S. H., Ross, W. E., Randlett, €I. E., and Carlson, G. J., T r a n s . Am. I n s t . Chem. Engrs., 4 2 , 3 3 (1946). (16) Mavitv. J. M.. Pines. H.. Wackher. R. C.. and Brooks. J. A,. IN=:ENG.CHEM.,40, 2374 (1948). (17) Oblad, A, G., and Gorin, M. H., Ibid., 38, 822 (1946). (18) Perry, S. F., Trans. Am. I n s t . Chem. Engrs., 42, 639 (1946).
Pines, H., in “Advances in Catalysis,” Vol. I, p. 251, edited by Frankenburg, Komarewsky, and Rideal, New York, Academic Press, 1948. Pines, H., Kvetinskas, B., Kassel, L. S., and Ipatieff, V. N., J . Am. Chem. SOC.,67, 631 (1945). Pines, H., and Wackher, R. C., Ibid., 68, 595 (1946). Ibid., p. 699. Powell, T. M., and Reid, E. B., Ibid., 67, 1020 (1945). Schmerling, Louis, paper presented before Division of Petroleum Chemistry, 119th Meeting, AM. CHEM. SOC., Cleveland, Ohio. Schuit, G. C. A., Hogg, H., and Verheus, J., Rec. trau. chim., 59, 793 (1940).
Stevenson, D. P., and Beeck, O., J . Am. Chem. Soc., 70, 2890 (1948).
Swain, C. G . , Ibid., 70, 1119 (1948). Swearingen, J, E., Geckler, R. D., and Nysewander, C. W., T r a n s . Am. I n s t . Chem. Engrs., 42, 573 (1946). Wackher, R. C., and Pines, H., J . Am. Chem. SOC.,6 8 , 1642 (1946).
Wagner, C. D., Beeck, O., Otvos, Y . W., and Stevenson, D. P., J . Phys. Chem., 1 7 , 4 1 9 (1949). RECEIVED for review June 14, 1952. ACCEPTED October 20, 1952. Presented before the Division of Petroleum Chemistry at the 111th Meeting of the AMERICAN CHEMICAL SOCIETY, Atlantic City, N. J.
Correlating Diffusion Coefficients in Liquids DONALD F. OTHMER AND MAHESH S. THAKAR Polytechnic Institute of Brooklyn, Brooklyn 2, N . Y .
D
TFFUSION coefficients in liquids are increasinglv important
in many theoretical and engineering calculations involving mass transfer, such as absorption, extraction, distillation, and chemical reactions. Comparatively few such data have been published, however, and methods for their correlation and for prediction of data on other substances are needed. A simple method of correlation is presented which allows easy and accurate extrapolation and interpolation of available data, and the prediction of such data when not available. The method follows the principle used t o correlate many other properties of matter in previous papers of this series-namely, plotting logarithmically the property of one material against the same property or the vapor pressure of a reference material. I n this manner, vapor pressures (a), gas solubilities and partial pressures (1‘7), adsorption pressures ( 1 5 ) , vapor compositions, equilibrium constants, activity coefficients, relative volatilities, electromotive forces (9, 1 1 ) , viscosities (IO),reaction rate constants ( I d ) , surface tensions ( l a ) , densities ( 1 3 ) , and azeotropic compositions ( 1 6 ) have already been correlated. The theoretical background for such a plot may vary, b u t the method usually follows the simple steps for making a vapor pressure plot (8): 1. On a sheet of logarithmic paper indicate vapor pressures of a standard substance-e.g., water-on the X axis and of the substance in question on the Y axis; then calibrate the X axis with values of temperatures corresponding t o the vapor pressures of the standard substance. 2 . Erect temperature ordinates. 3. Plot points and connect with a straight line, the slope of which is the ratio of the molal latent heat of the substance to that of the reference substance. This ratio is much more nearly constant a t all temperatures-and the line more nearly straightthan the latent heat itself from a plot of log P us. 1 /T. DEVELOPMENT OF LOGARITHMIC PLOT FOR DIFFUSION COEFFICIENTS
The variation of diffusion coefficients with temperature has been assumed by Wilke ( 1 9 ) t o be related to the Stokes-Einstein
law-i.e., directly proportional to the absolute temperature and inversely proportional t o the viscosity of the liquid. Eyring (6) and others (3,4,18) have suggested that diffusion is a rate process which varies as a n exponential function of temperature. Thus:
D
= KeEd/RT
(1)
Taking logarithms and differentiating,
The Clausius-Clapeyron equation for the vapor pressure of liquids is: dT d l o g P = -L
RT2
(3)
At the same temperature, if Equation 2 is divided b y Equation 3, there results:
(4)
If E J L is assumed constant, this can be integrated t o give: Log D = Ed log P
+C
(5)
Equation 5 indicates that, if diffusion coefficients are plotted on logarithmic paper against vapor pressures of a reference liquid at the same temperatures, a straight line results with slope of EaIL. Diffusion coefficients of different substances are plotted this way in Figure 1. The X or vapor pressure axis is calibrated t o give a temperature scale by known values from a standard table. The energy of activation of diffusion can be calculated from the slope of the line and the latent heat of vaporization of the reference substance a t t h a t temperature.
INDUSTRIAL AND ENGINEERING CHEMISTRY
590 n
0 x
LogD =
TEMP ' C
VAPOR PRESSURE QF WATER
MM. Hg.
Figure 1. Logarithmic Plot of Diffusion Coefficients (x lo5) of Various Materials us. Temperature Scale Derived from Vapor Pressure of Water 1. Hydrogen in water Self-diffusion of water Phenol in benzene 4. Calcium chloride i n water a. Mannitol in water 6 . 131,2,2-Tetrabromoethanein tetrachloroethane 7. Saccharose in water 2. 3.
It can be observed in Figure 1, that, in cases where water is not the solvent, straight lines are obtained; where water is the solvent, the lines show a definite break a t about 30" C. An exactlv similar break was found t o occur near the same trmperature when viscosity of water was plotted in a similar way against the vapor pressure of water (10). This similarity in behavior supports the views of Glasstone, Laidler, and Eyring (6) that the mechanisms of viscosity and diffusion are similar, and in dilute solutions the energy of activation for viscosity, E,, is equal to the energy of activation for diffusion, E d This break, which is mainly due t o changes in internal structure of .ivater with consequent changes in energy of activation, is a disadvantage to correlating data. The temperature a t which the changes in internal structure of water occur map vary somewhat with the dissolution therein of the solute; hence the breaks in the line may be a t somewhat different temperatures in the general range of 30" C. One of the great advantages of using straight linrs as a correlating method is that slight changes in the basic physical function, due to chemical or polymeric changes of the material itself, are readily evident; with curved lines the change in the function could not be detected, and the best curve through all points would be drawn, or the best empirical equation with an additional number of terms would be drawn, to ''smooth'' the data, rather than obtaining the two lines representing the true conditions. Because water is the most interesting liquid in diffusion studies, i t is necessary to eliminate this break. This may be done in viscosities of aqueous solutions bv plotting them against viscosity of water on logarithmic paper. Similarly, diffusion coefficients of aqueous systems may be so correlated; and because of basic similarity of the water relation to each (both above and below 30' C ), the diffusion coefficients may be plotted against viscosities t o eliminate this break entirely. The corresponding equation of a straight line for the variation of viscosity with temperature, used previously for the viscosity correlation (8),is:
Vol. 45, No. 3
- dEE- , l o g p
+ C'
(7)
As the energy of activation for diffusion, Ed, is approximal(,Iy equal to the energy of activation for viscosity, E,, in dilute solutions, i t follows that the slope of the line obtained by plotting diffusion coefficients in water against the viscosity of water on logarithmic paper should be approximately equal t o - 1.0. Also, because Ed and E, are related energies, i t follows that any changes in one would be accompanied by corresponding changes in the other; and their ratio, &/'Ev, will be more nearly constant than either one alone over wide temperature ranges, with the result that the break occurring around 30" C. will be eliminated. These conclusions are borne out in Figurc 2, where diffusion coefficients in water are plotted against, viscosity of water on logarithmic paper. A temperaturc scale was constructed, as brfore, on the X axis, by indicating temperature values corresponding to viscosity values of water. The data are represented by nearly straight lines over the entire temperature range; and the slopes of the lines for a numher of systems vary between -1.07 and -1.15. A value of the slope of - 1.1was regarded as a fair average for the systems of Figure 2 and many others similarly plotted, for systems where wat:lr is the solvent and where another liquid is the solvent. Thus, if a diffusion coeffcicnt is available a t only one temperature, a straight line can be draim with a slopc of -1.1 on a reference plot, such as Figure 2, through this point; and the valucs may be picked off with a ri3asonable assurance a t any other temperature. CORRELATION O F DIFFUSION DATA IN WATER
The intercept, C' of Equation 7 represents the logarithm of diffusion coefficients of various substanccs in Figure 2. Logarithmic Plot ?f water a t a temperaDiffusion Coefficient ( X lo6) in ture when the visWater of Various &laterials us. cosity of water is 1 Temperature Scale Derived from Viscosity of Water centipoise-i.e, at 20.06' C. The rate of diffusion depends upon h o w large a hole a molecule has to create in the surrounding liquid in order to start diffusing; and i t can be expected that the intercepts (represented by C'), which can be called log D;,will be a function of the size of the diffusing molecule or its molal volume. Figure 3 is a plot of the diffusion coefficients, D:, of various substances in water a t 20" C. (7) against the molal volume of the diffusing substance, V,. The relation can be represented by the following equation, which represents all data with an average deviation of 5.05%: 14.0 uq x 105 = v7no.a
If at the same temperature this is divided by Equation 2, and the result is integrated, there follows:
The molal volume of substances was determined as the summation of atomic volumes, set forth by LeBas, and as used previously by Arnold (1) and Gilliland (5) in correlating diffusion
March 1953
INDUSTRIAL AND ENGINEERING CHEMISTRY
591
Figure 3. Logarithmic Plot of Diffusion Coefficients in Water at 20" C. against Molal Volumes of Diffusing Substances
2.
For 31 gaaes, liquids, and solids with molecular weights up t o 694 and molal volumes up t o 504. Solid dots indicate gases; open dots indicate liquids and solids. Data from (I,7)
Figure 4.
Logarithmic Plot of Diffusion Coefficients
us. Function of Viscosity of Water and Molal Volume
m
of Diffusion Substance
E!
Solid dote indicate gases; open dots indicate liquids and solids. Data are for 44 systems, a t different temperatures. Data from (a. 7)
X
4
Figure 5. Logarithmic Plot of Diffusion Coefficients for Numerous Nonaqueous Systems US. Function of Viscosity of Water, Particular Solvent Used, and Molar Volume of Diffusing Substance
W
v)
This is a general plot for all solutes and solvents. The line is t h a t of Fi ure 4 and all data of Figure 4 might also be plotted heFe t o give ad%tional weight to the general oorrelation. All data in (7) for nonionizing solutes in different solvents, for which the physical properties required were available, are plotted in this figure, including coefficients in more than 20 different solvents
water are taken as the averages suggested by Arnold ( 8 ) . It can be seen that experimental data correlate closely according t o Equation 9. This is closer than would be expected, considering the wide range of substances, of experiments, and of experimental techniques over the many years of work represented. It is suggested t h a t this correlation may be used with reasonable assurance t o predict diffusion coefficients in water when experimental data are lacking. CORRELATION OF DIFFUSION COEFFICIENTS IN VARIOUS SOLVENTS
To obtain a general correlation of diffusion coefficients in various solvents, the following procedure was adopted: at
For diffusion in any solvent: d log D, = R~
-at
For viscosity of water: d log put = RT2 E,
,
Ed,
(10) (11)
Dividing and integrating, as before:
. ... coefficients in gases. given in Table I.
The values of atomic volumes used are
TABLE I. ATOMICVOLUMES Oxygen Except as indicated I n methyl esters In methyl ethers I n higher ethers and esters In acids Molecular oxygen
Bromine Carbon Chlorine Hydrogen Iodine
27.0 14.8 24.6 3.7 37.0
Nitrogen Double bonded Primary amines Secondary amines Water Heavy water
15.6 Sulfur 10.5 12.0 I n benzene ring deduct 18.9 22.0 (estimated) I n naphthslenehng, deduct
7.4 9.1 9.9
11 .o 12.0 25.6 25.6 15 30
Substituting Equation 8 in Equation 7, the following relationship is obtained to express diffusion coefficients of various substances in dilute water solutions: (9)
Thus, a plot of diffusion coefficients in water of different substances at various temperatures against the function h1*1VrnO.e should give a straight line with a slope of -1.0 on logarithmic paper. All data available in International Critical Tables (7) of substances, whose molal volume could be determined, are plotted in this way in Figure 4. D a t a for gases diffusing through
which on rearrangement, becomes:
It has been shown (6) t h a t energy of activation is proportional to the latent heat of vaporization of the solvent; and the proportionality constant varies little for most liquids Therefore, the activation energy ratio, Eds/Edw,can be replaced by the latent Taking Eh/E,, at an average heat of vaporization ratio, L,/L,. value of 1.1., i t follows that
L L,
Log D, = -1.1 .-! log
pw
+k
(13)
I n Equation 13 k is the logarithm of diffusion coefficients at 20.06" C., where the viscosity of water is 1 centipoise. It is reasonable t o assume t h a t the diffusion coefficient of a given substance in various solvents at 20" C. will depend on the resistance which the surrounding solvent offers t o the diffusing molecule. Hence, it can be expected t h a t the diffusion coefficients of a given substance in different solvents will be inversely proportional t o the viscosities of the given solvents a t a given temperature (providing the molecular species is identical i n each solvent). From the data available for diffusion of iodine and phenol in various solvents a t 20" C., i t was found t h a t the expected relationship holdswell. I t i s obvious that the proportionality constant will be D:, as the viscosity of water is 1 centipoise a t 20"C. Therefore, therelationship can be expressed as D:/D: =
592
INDUSTRIAL AND ENGINEERING CHEMISTRY
l/p:. Assuming t h a t this relationship holds in general, and using Equations 13 and 8, there follows:
Vol. 45, No. 3
.\
TEMPERATURE OF SOLUTION'C
It is obvious that for diffusion in water, Equation 14 reduces t o Equation 9. A plot of diffusion coefficients in various solvents was made following Equation 14. All data available in the International Critical Tables ( 7 ) for diffusion in liquids are plotted in Figure 5. The line drawn in Figure 5 represents Equation 14 and also the data for diffusion in water as shown in Figure 4; and the values for water are not repeated. Whereas data in many solvents correlate closely according to Equation 14, those in ethanol and methanol show considerable scattering. This may be due to the inapplicability of the assumption made above for correlation, to inaccuracies in data which are many times large, or to some interaction of polarities between molecules, which is not taken into account in the above considerations. A very likely cause of scattering may be the presence of water as impurity in ethanol or methanol, which changes considerably the viscosity (and presumably the diffusion coefficients). In absence of experimental data, correlation in Figure 5 can be used to estimate diffusion coefficients in dilutc solutions.
\ \
REFERENCEL\INE No. 1 \
\
\
0 0 0 0 0 Z w m + m
g
P 0
-
N 0
w
0 1111 11 I I
A~-VISCOSITY
\
\
Li
in
b b b j i n IIII I
SOLVENTA T 20 C '
OF
IN
-
0
N 111l111
I l l 1
I
I I I
I
CPS.
\
/ I
/
/. /' /
\ \
/-
I
\\
//
%' MOLALV O L U M E
CONSTRUCTION OF NOMOGRAM
OF
/ ,
'
DIFFUSINGSUEISTANCE--CM.~/G.MOL
/
For ready prediction of diffusion /' \ coefficients, a nomogram has been -CM~E.EC x lo5 // DIFFUSIONCOEFFICIENT 0 constructed based on Equation 14 N and is presented in Figure 6. For I1 I l l I I I I I I convenience, the viscosity line has Figure 6. Nomogram for Predicting Diffusion Coefficients in Dilute Solutions been calibrated t o give temperature of diffusion directly. Thus, if the Use only three scales on right when water is solvent Use all scales when liquid other than water is solvent temperature of diffusion, ratio of Illustrations. the molal latent heat of vaporization O F DIFFUSIOX COEFFICIENT OF CAFFEINE I K VATER .&'r 10' C. For diffusion in water, 1. PREDICTION join 10' C. on temperature scale for water (3rd from right) with V m = 2 3 2 on molal volume scale of the solvent to water, viscosity of (summed from individual atomic volumes in Tahle I) to obtain a diffusion coefficient of 0.40 X 10-6 the so'vent a t 20" C., and molal (experimental value reported = 0.41 X 10-5) 2. PREDICTION OF DIFFUSIONCOEFFICIEST OF IODINE I N CARBON DISULFIDE AT 16' C. REQVIRED volume of the diffusing substance PROPERTIES.Ls/Lw = 0.59 (obtained from slope of log plot, of vapor pressureor from tabular data) are known, i t is possible to predict &' = 0.365 (from tabular data f o r 20' C.) diffusion coefficients by making use V m = 7 4 . 0 (summed from Table I) of the nomogram. For diffusion in Join temperature point for 16' C. on temperature of solution scale (extreme left) with value of L d L w water, L8/Lw becomes 1, and p; 0.59, t o find intersection of reference line 1; join this point with \ d u e of ~s = 0.366 on viscosity scale to obtain intersection on reference line 2 ; join this point with value Vin = 74.0 on molal volume scale to equals unity. For convenience, anobtain value of diffusion coefficient of 2.80 (corresponds to experimental value of 2.07) other temperature scale has been constructed on the second reference line, for predicting diffusion coefficient in the more For example, if it is deaired to predict the diffusion coefficient of iodine in carbon disulfide a t 16" C., the necessary properties of water. Thus, if the temperature of diffusion the system are first obtained. These are: and the molal volume of diffusing substance are known, the L,/Lw = 0.59 diffueion coefficient in water can be predicted by the use of only the three scales on the right. V , = 74.0 ml. per gram-mole pL,o = 0.365 centipoise
-
-
March 1953
INDUSTRIAL AND ENGINEERING CHEMISTRY
A line is drawn t o join the point representing 16' C. with the value of L,/L, to intersect the first reference line. L,/L, may be found directly from the slope of a vapor pressure plot as previously drawn (8),and fi: may be taken from a nomogram previously presented. From here a line joins p,O = 0.365 to obtain a point of intersection on the second reference line. From this point of intersection a line is drawn through V, = 74.0 on the molal volume line t o obtain diffusion coefficient of 2.80 X 10-6 on the diffusion coefficient line. This compares with an experimental value of 2.97 X 10-6 sq. cm. per second. Another example may be cited for diffusion of caffeine in water. T o predict the diffusion coefficient of caffeine in water a t 10" C., the molal volume of caffeine is determined from its structural formula; this is equal to 232.6. The point of 10"C. on the temperature scale constructed for water on the second reference line ' is connected with 7 , = 232.6 and the line produced t o obtain the value of the diffusion coefficient on the diffusion coefficient sq. cm. per second is read off, soale. A value of 0.40 X which is close t o the experimental value of 0.41 X 10-6 sq. cm. per second. However, the nomograph is based on Equation 14, and the limitations of that equation should be borne in mind when the nomograph is used.
-4
D, D: Ed
= diffusion coefficient in water, sq. cm. per second
= diffusion coefficient in water a t 20" C sq. cm. per second = energy of activation for diffusion, calo;:es per gram-mole
Ed. = energy of activation in any solvent, calories per gram-mole Ea, = energy of activation in water, calories per gram-mole Ey = energy of activation for viscosity, calories per gram-mole E,, = energy of activation for viscosity in water, calories per gram-mole K , C, C', k = constants L = latent heat of vaporization, calories per gram-mole L, = latent heat of vaporization of any solvent, calories per gram-mole L, = latent heat of vaporization of water, calories per grammole P = vapor pressure of a liquid R = gas constant, calories per gram-mole X O K. T = absolute temperature, K. p = viscosity, centipoises p? = viscosity of any solvent at 20" C., centipoises fi, = viscosity of water, centipoises V , = molal volume of the diffusing substances, ml. per grammole LITERATURE CITED
CONCLUSIONS
The method presented in this paper can be conveniently used
to obtain straight-line correlation of diffusion coefficients in dilute solutions. For diffusion in water, it is suggested that a plot of diffusion coefficients against viscosity rather than against vapor pressure of water be used to obtain better straight-line correlation. The slopes of these straight lines are related to the energy of activation for diffusion. Equation 9 or Figure 4 can be used t o obtain diffusion coefficients in water. To estimate diffuaion coefficients in solvents other than water Equation 14 or Figure 5 may be used. The use of the nomograph in Figure 6 facilitates the prediction of such data. The correlation is based on experimental data from widely varying sources, and some indicated assumptions. I t is to be used, therefore, with some reserve, especially with ionizing solutes; but it will often be most helpful for the many materials for which no experimental data are available. ACKNOWLEDGMENT
The help of Paul Maurer with the manuscript and figures is gratefully acknowledged. NOMENCLATURE
D D. D:
593
= diffusion coefficient, sq. cm. per second = diffusion coefficient in any solvent, sq. cm. per second = diffusion coefficient in any solvent a t 20' C., sq. cm. per
second
(1) Arnold, J. H., IND.ENQ.CHEM.,22, 1091 (1930). (2) Arnold, J. H., J . Am. Chem. Soc., 52, 3937 (1930). (3) Bradley, R. S., J.Chem. Soc., 1934,1910. (4) Braune, H., 2. physilc. Chem., 110, 147 (1924). (5) Gilliland, E. R., IND. ENG.CHEM.,26, 681 (1934). (6) Glasstone, S., Laidler, K., and Eyring, H., "Theory of Rate Processes," p. 522, New York, McGraw-Hill Book Co., 1941. (7) International Critical Tables, New York, McGraw-Hill Book Co., 1926. (8) Othmer, D. F., IND. ENG.CHEM.,32,841 (1940). (9) Ibid., 36, 669 (1944). (10) Othmer, D. F., and Conwell, J. W., Ibid., 37, 1112 (1945). (11) Othmer, D. F., and Gilmont, R., Ibid., 36, 858 (1944). (12) Othmer, D. F., Josefowitz, S., and Schmutzler, ,4.E., Ibid., 40, 286 (1948). (13) Ibid., pp. 883,886. (14)Othmer, D. F., and Luley, A. H., Ibid., 38, 408 (1946). (15) Othmer, D. F., and Sawyer, F. G., Ibid., 35,1269 (1943). (16) Othmer, D. F., and Ten Eyck, E. H., Ibid., 41,2897 (1949). (17) Othmer, D. F., and White, R. E., Ibid., 34, 952 (1942). (18) Rabinowitch, E., Trans. Faraday Soc., 33, 1225 (1937). (19) Wilke, C. R., Chem. Eng. Progr., 45, 218 (1949). RECEIVED for review December 17, 1951. ACCEPTEDOctober 16, 1952. Previous articles in this series have appeared in INDUSTRIAL AND ENQINEERINQ CHEMISTRY during 1940, 1942, 1943, 1944, 1945, 1946, 1948, 1949, 1950, and 1951; Chem. & Met. Eng., 1940; Chimie et Industrie (Paris), 1948; Euclides ( M a d r i d ) , 1948; Sugar, 1948; and Petroleum Refiner, 1961 and 1952.
Pinitol from Sugar Pine Stump Wood ARTHUR B. ANDERSON Forest Products Laboratory, University of California, Berkeley, Calif.
A '
R E C E N T investigation indicated that the amount of pinitol is not uniform in the trunk of the sugar pine tree (Pinus lambertiana Doug].) ( 2 ) . The greatest quantity of this cyclitol, which is a monomethyl ether of d-inositol, CBHB. (OH)SOCHa, is found in the butt heartwood adjacent to the stump This portion of the tree yielded from 4.8 to 8.6% pinitol, whereas the average yield for the total heartwood in the tree was about 4.0%. This suggested that sugar pine stumps now left in the forest might contain sufficient quantities of pinitol to warrant their removal and subsequent processing for the recovery of this product. In addition, there is evidence that stump removal would make the areas more adaptable for reforestation by either natural propagation or tree planting methods. Certain species of stump wood are being harvested and processed for their extractive components. Old-growth southern pine stump wood has been used for over 40 years for the recovery of rosin, turpentine, and pine oil, which augments the naval stores supplied by the living tree (12). More recently, ponderosa
pine stump wood has been found to be a good source of rosin and volatile terpenes (1) and a commercial extraction plant has been installed t o process ponderosa pine stumps (15). Sugar pine stump wood, on the other hand, is not rich in resinous components, for i t contains only 3.0 t o 6.5% benzene-soluble material and hence is not suitable for rosin production. It was nearly a hundred years ago that Berthelot first described the chemical nature of a pine tree "sugar" obtained from the exudate of sugar pine (4). Wiley later identified this product as pinite, now called pinitol (17). This cyclitol is present in redwood heartwood (Sequoia aempervirens D. Don Endl.) (14)andit hasbeen found in the heartwood of five other pines belonging to the Haploxylon subgroup or soft, white five-needled pines (11). Pinitol also occurs in red spruce (Picea rubra) (8). The yields of pinitol from these sources are reported t o vary from traces to 0.5%. Recently, it has been reported t h a t redwood stumps contain 1.9 to 2.1% cyclitols-Le., pinitol plus sequoyitol (10). None of these sources appears to contain as much pinitol as
