Applications of holdup data to explain the effect of diffusivity on the vaporization of liquids in packings and t o estimate effective interfacial

areas for mass transfer have been outlined.

ACKNOWLEDGMENT

G = superficial gas rate, 1b.I (hr.)

(sq.ft.) h, = operating holdup, cu.ft./cu.ft. li, = static holdup, cu.ft.1cu.ft. ht = total holdup, cu.ft./cu.ft. k, = gas-phase mass transfer coefficient, lb. moles/ (hr.) (sq. ft.) (atm.) k , a = gas-phase mass transfer coefficient, lb. moles/ (hr.) (cu. ft.) (atrn.) L = superficial liquid rate, 1b.l (hr.) (sq.ft.) m = constant in Equation (5) M M = mean molecular weight of gas, 1b.llb. mole n = constant in Equation (5) P,, = mean partial pressure of inert gas in t h e gas phase, atm.
Greek Letters
p

= liquid density, g./ml. = gas density, 1b.lcu.ft. o = surface tension, dyneslcm. o = void fraction, cu. ft./cu.ft.
p
Subscripts

The authors wish to acknowledge support of this work ( P a r t s I, 11, and 111) under Grant G-200 of t h e National Science Foundation and Contract No. AT(30-1)-1463 of t h e Atomic Energy Commission.
NOTATION

w = f o r water systems abs = for absorption work without a chemical reaction vap = for vaporization work
LITERATURE CITED 1. Jesser, B. W., and J. C. Elgin, Trans. Am. I n s t . Chem. Engrs., 39, 277 (1943). 2. Leva, M., Tower Packings and Packed Tower Design, 2nd ed.,

a = effective interfacial area, sq.

c

= constant in Equation (5)

f t .I cu.ft.

D = diffusivity of solute in gas,

sq. ft./hr. D, = diameter of sphere possessing the same surface area a s a piece of packing, f t .

The United States Stoneware Company, Akron, Ohio (1953). 3. Mehta, J. J., and R. H. Parekh, S.M. thesis, Mass. Inst. Technol.
(1939). 4. Surosky, A. E., and B. F. Dodge, Znd. Eng. Chem., 42, 1112 (1950).
Parts I and II presented at the A . I . Ch. E. Sfiringfield mcetiny, Part III at the N e w Y-orn
?nPeting.

= liquid viscosity, centipoises = gas viscosity, 1b.l (hr.) (ft.)

CORRELATION OF DIFFUSION COEFFICIENTS IN DILUTE SOLUTIONS

C. R. WILKE and PIN CHANG
University of California, Berkeley, California

The diffusion coefficient is normally defined and assumed in this study to be the proportionality constant in the rate equation written for undirectional mass transfer as follows :

ties of dilute solutions so as to permit estimation of diffusion coefficients f o r engineering purposes.

PREVIOUS COlRRELATION

general i t was assumed t h a t this function extrapolated into the Stokes-Einstein equation a t very large solute molal volumes.
DEVELOPMENT OF NEW CORRELATION Sources of Data. A t the time of the previous work so few data were available for diffusion of single solutes in a variety of solvents t h a t the effect of solvent properties, could not be brought into a general correlation. I n a special effort t o obtain suitable data of this kind a companion experimental study ( 2 ) was conducted involving the diffusion of iodine and toluene i n a wide variety of hydrocarbon

Equation i 1) is strictly applicable in ideal dilute solutions in which convective transport due t o volume changes on mixing is negligible, and in which other possible modes of mass transfer are not operative. This paper represents an attempi- to generalize the relation of P to conveniently available proper-

I n the earlier paper by Wilke (10) a method of correlating diffusion coefficients was proposed on the basis of qualitative conclusions of the Eyring theory(3) and t h e Stokes-Einstein relation. It was shown t h a t the group TID-q, designated as the diffusion factor F, was essentially independent of temperature f o r available systems. Furthermore F could be represented as a smooth function of molal volume f o r diffusion of various solutes in a given solvent. I n

Page 264

A.1.Ch.E. Journal

June, 1955

u ;d v
I U

6
o

x v
m

polvents ranging from hexane through tetradecane. Data were also obtained f o r diffusion of organic acids in several solvents. These new data were supplemented by certain other data from the literature, including all the data reported in the previous paper ( l o ) , to provide a basis f o r the present development. All data which supplement those presented in Tables 2 through 5 of referenceil0) are presented in Table 1.

FIG.1. DIFFUSION IN WATEP..

D 1 0

20 4 0 GO 80100 200 400 600 1003 2000 SOLUTE MOL A L VOLUME, CU. Cm /qm m s l

6 4
03

__

____

1I
---

0 x

3--0 BENZENE

BROMOBE NZ E NE

v CARBON TETRACHLORIDE]
Q

IODINE OTULUENE

I 20

I 40

I
400

60 80 100 200 V, c.c / g m m o l

6oo

Effect of Solute Molal Volume. Figure 1 shows the diffusion as a function of molal volume f o r various solutes in water based on data from Table 2 of reference 10. Molal volumes used throughout this work a r e values a t the normal boi!i?g point estimated for complex molecules by the atomic contributions of LeBas (I, 6) as summarized in Table 2. As indicated in Figure 1, F is a smooth function of molal volume having a log-log slope of about 0.7 at low molal volumes and apparently merging smoothly with t h e

c I . 2

1
,SLOPE

= 0.6

vs
5:
0 ACETIC

gv

BENZOIC ClNNAMlC'0 F O R M I C A
0
v..

20

40

v,

6 0 80 100 c.c./gm. mol

ORGAPI'IC

2 00

400

20

40

60 80 1 0 0 V, c.c./qm. mol

200

408

FIG. 3. DIFFUSION O F

ACIDSI N TOLUENE.

FIG. 4. DIFFUSION OF ORGANIC ACIDSI N

CARBONTETRACHLORIDE.

Vol. 1, No. 2

A.1.Ch.E. Journal

Page 263

TABLE NAL DIFFUSION DATAFOR

VARIOUS

SYSTEMS

(Supplementary to Tables 2 to 5 of reference 10) Solute Acetic acid Solvent Acetone Benzene Carbon tetrachloride Toluene i-Amy1 alcohol Aniliie Benzene Benzoic acid
MOLECULAR W E I G H T OF SOLVENTS

DXlW DX106 Tempera- sq. cm./sec. sq. cm./sec. Ref. ture, "C. (obs.) (calc.)
40.0 25.0 15.0 25.0 5.9 6.5 14.8 25.0 40.0 25.0 15.0 6.8 20.0 18.5 7.5 15.0 15.0 13.2 25.0 40.0 15.0 25.0 40.0 14.8 25.0 40.5 16.0 25.0 40.0 7.6 7.3 7.3 7.3 7.3 7.3 7.3 7.3 7.3 7.3 20.0 20.0 20.0 20.0 7.3 . .7.3 7.3 7.3 7.3 7.5 7.5 7.5 7.4 7.3 25.0 25.0 25.0 25.0 25.0 25.0 25.0 25.0 25.0 25.0 25.0 25.0 25.0 25.0 25.0 20.0 20.0 4.044 3.309 2.916 2.081 1.587 1.151 1.267 1.490 1.780 2.265 1.905 1.661 0.78 2.70 1.02 3.70 3.70 2.368 2.622 3.054 1.170 1.379 1.762 0.776 0.908 1.168 1.289 1.493 1.851 1.32 1.41 0.90 1.34 1.44 3.50 2.60 1.31 0.47 1.52 2.74 1.77 0.97 0.76 1.04 0.85 0.34 0.149 2.15 0.226 0.36 1.24 1.52
~

Ethyl alcohol Ethyl alcohol Bromobenzene Chloroform n-Hexane Acetone Benzene Carbon tetrachloride

3.49 2.85 2.51 1.74 1.22 1.13 1.32 1.58 2.04 2.00 1.72 1.48 0.85 2.41 2.54 3.85 1.66 1.89 2.19 0.95 1.13 1.44 0.87 1.04 1.36 1.16 1.34 1.65 1.29 1.29 0.91 1.18 1.55 3.80 3.02

a
a

1.20

FIG. 5. EFFECTOF SOLVENT MOLECULAR WEIGHT.

Stokes-Einstein equation which requires a slope of 113 a t high molal volumes. On the assumption t h a t molecules a r e spherical with a radius equal to (3V/4nN)1I3 t h e Stokes-Einstein equation may be written as follows:

Toluene Benzo-trichloride Bromobenzene Toluene Benzene Cyclohexane m-Cymene Ethyl benzene Ethl ether n-Hexane Mesitylene Transdecalin m-Xylene Bromform Acetone Benzene Ethyl alcohol Bromonaphthalene Ethyl alcohol a-BromoBenzene naphthalene Cyclohexane Decalin Dibenzyl ether n-Hexane a-Methyl naphthalene Tetralin Toluene rn-Bromotoluene Toluene Carbon Benzene tetrabromide Carbn Benzene tetrachloride Carbon tetrachloride Cyclohexane Decalin Dioxane n-Heptane n-Hextane Isooctane Kerosene Tetralin Toluene Cinnamic acid Acetone Benzene Carbon tetrachloride Toluene Ethyl benzoate Acetone Benzo trichloride Ethyl acetate Nitrobenzene

d
d d

1.54

-- = 1.004 X 10' DS

d3 (2)

Equation (2) is shown as a dotted line on Figure 1. The general behavior of the curve of Figure 1 relative to the Stokes-Einstein equation constitutes reasonable evidence in favor of the proposed method of correlation. Over t h e middle range of molal volumes the curve of Figure 1may be satisfactorily represented by a line of slope 0.6. Bearing in mind the theoretical limitations of t h e assumption i t is convenient t o assume t h a t the diffusion factor is proportional to P6 over the middle range. The proportionality of Dy to V0.6 was used by Thakar and Othmer (8) in their representation of the correlation f o r diffusion of substances in water. To explore the mold volume effect in nonaqueous systems several solvents were studied as shown in Figures 2 through 4. The group DqlT may be represented satisfactorily as proportional to VO.8. It is therefore a f a i r generalization t h a t D y l T is proportional t o Vo.6 in the medium molal volume

0.48 1.69 3.22 1.88 0.96 0.72 1.05 0.72 0.39 0.17 2.44 0.22 0.38 1.31 1.48 1.27 1.87 1.74 1.32 0.733 1.01 3.25 3.11 2.86 1.03 .735 2.22 2.51 0.99 0.91 1.14 2.29 0.58 2.01 0.51

d d d d d d
d

i
b
d d d d d d

d
d d
g j g g g g g
g g g g
U

1.12 2.00 1.41 1.49 0.776 1.02 3.17 3.70 2.57 961 0.735 2.19 2.41 1.21 0.755

1.18

a
U

20.0 20.0

2.47 0.52 1.85 0.60

h h h h

Page 266

A.1.Ch.E. Journal

June, 1955

Ethylene bromide Ethylene chloride Formic acid Acetone Benzene Carbon tetrachloride Toluene n-Heptyl bromide n-Hexyl bromide Iodine Carbon tetrachloride Cyclohexane Dioxane Ethyl alcohol Heptane Hexane n-Hexane Methyl cyclohexane n-Octane n-Tetradecane Iodobenzene Methyl iodide Nitrobenzene n-Octyl bromide Benzene Toluene Methylene chloride Acetone Ethyl benzoate Ethyl acetate Octane Ethyl alcohol Benzene Benzene n-Decane n-Dodecane n-Heptane n-Hexane n-Tetradecane Heptane Hexane

7.3 25.0 18.5 6.5 6.2 13.9 25.0 8.5 25.0 15.0 6.2 14.1 25.0 7.4 7.6 25.0 15.0 25.0 40.0 25.0 25.0 25.0 25.0 40.0 30.0 15.0 25.0

1 . 1 1
3.768 3.274 3.132 1.991 2.306 2.577 1.612 1.888 1.673 2.285 2.463 2.646 1.52 2.31 1.50 1.54 1.07 1.772 1.316 3.42 4.05 4.24 2.71 2.30 2.43 0.96 1.35 2.23 2.06 2.94 0.73 2.25 1.46 1.12 1.24 1.34 2.09 1.35 4.33 3.72 2.95 4.21 1.02

1.31 3.59 3.12 2.87 1.48 1.76 2.14 1.48 2.01 1.67 1.90 2.32 2.54 1.48 2.54 2.30 1.19 1.31 1.72 1.32 4.60 4.96 4.37 2.74 2.28 3.04 1.14 1.31 2.40 2.54 2.83 0.66 2.48 1.42 1.00 1.03 1.12 1.77 1.19 4.08 3.35 2.56 4.13 0.86

d
a a

a a

d d

i
a

i
a

9 i
a

appeared to correlate the data most successfully. Figure 5 shows the group D-qlT as a function of molecular weight for diffusion of given solutes in a number of solvents. Although there is considerable scatter of the points a line of slope 112 on the log-log plots correlates each system moderately well. As in the case of the molal-volume effect there are obviously other factors involved so that use of solvent molecular weight is satisfactory only a s a first approximation. The data of Trevoy and Drickamer(9) are for 0.50 mole fraction of phenol in various hydrocarbons so that the solvent molecular weight is used primarily to show the trend.
General Correlation for Unassociated Liquids. From the results of the pre-

Iodine

7.3
7.4 7.5 20.0 20.0 20.0 7 . 5 20.0 7.6 7.6 25.0 25.0 40.0 25.0 6.9 25.0 25.0

d
d d

ceding section i t was concluded that an equation of the frllowing form would express the effects of solute and solvent:

h h h
d

= const. ___
r]

TM'/

(3)

VoP6

Pyridine
1 , 2, 4, 5-Tetrachlorobenzene

b
d

1, 2, 4-Trichlorotoluene
Toluene

d
a a
a a

a
a

Figure ti shows a log-log plot of DIT vs. the group T V O . ~ / Mfor ~/a ~ wide variety of unassociated solvents embracing the data of Table 2 of this paper and Table 5 of reference 10. The method of plotting was selected to spread the data and best illustrate the scope of the correlation. The line through the data has slope 1 as required in the assumptions of the correlation and may be expressed by the equation

@hang, Pin, and C . R. Wilke, "Some Measurements of Diffusion in Liquids," J . Phys. Chem. (in press). bznternational Critical Tables 5, 63-75 (1929). CMuchin, G. E., and G. P. Faermann, 2. physik. Chem., 121, 180 (1926). dHerzog, R. O., et al., 2. physik. Chem., ( A )167, 329 and 343 (1933). C'e Monde, H., J . phys. radium, 7, 371-8 (1936). fOholm, L. W., Medd. Nobelinst., 2, 23 (1913). gHammofid, B. R., and R. H. Stokes, personal communication to J. H. Hildebrand (Sept. 21, 1954). ILDurnmer, E., 2. akorg. u. a l l g e m Chem., 109, 49 (1919). %tokes, R. H., P. J. Dunlop, and J . R. Hall., Trans. Faraday Soc., 49, 886
(1953).

Data f o r 155 points among 123 different solute-solvent systems are expressed by the correlation with an average deviation of 12% between calculated and observed results.
Correlation of Associated Liquids. ASsociated liquids such as water and other hydrogen-bonding solvents might be expected t o show deviation from the correlation of Figure 6. Figure 7 shows the plot of DIT vs. -qVO.6/M1/2 for diffusion in water. The best line through the data falls clearly above the dotted line representing Figure ti. This deviation is in the direction corresponding to association of the solvent. By assigning a molecular weightt to the solvent equal to 2.6 times the nominal molecular weight of water one can bring the data of

jWatts, H., B. J. Alder, and J. H. Hildebrand, J . Phys. Chem., 23, 659 (1955).

range for both aqueous and nonaqueous solvents. However, it must be recognized that special structural features of molecules and other molecular interactions may be important in certain cases and that therefore the proposed relationship is a t best an oversimplification utilized to obtain a practical result.

Effect of Solvent Properties. Study of the effect of solvent properties in addition to viscosity centered on the behavior of the group D-qlT for diffusion of single solutes in a variety of solvents. A wide variety of variables such as molal volume, heat of vaporization, molecular weight, etc., were examined. Of these the solvent molecular weight

Vol. 1, No. 2

A.1.Ch.E. Journal

Page 267

TABLE2-ATOhllC

VOLUMES FOR COMPLEX MOLECULES, VOLUMES FOR SIMPLE SUBSTANCES

MOLECULAR

TABLE

3-COMPARISON

OF

ASSOCIA-

TION PARAMETERS WITH ASSOClAlION

Atomic Volumes Bromine Carbon Chlorine Hydrogen Iodine Nitrogen, doub!e bonded. Nitrogen, in primary amines 27.0 14.8 24.6 3.7 37.0 15.6 10.5

NUMBERS OF JACOBSEN -4ssociation Association parameter, x number * 2.6 1.9 1.5 1.o 1.o 1 . 0 60 3.5 2.7 1.0 1.02 1.0

12.0 Nitrogen, in secondary amines Oxygen (except as noted below) 7.4 Oxygen, in methyl esters 9.1 Oxygen, in methyl ethers 9.9 Oxygen, in higher ethers and esters 11.0 Oxygen, in acids 12.0 Sulfur 25.6
0.6 8.5 11.5 15 15 30 47.5

Solvent
Water Methyl alcohol Ethyl alcohol Benzene Ether Heptane *At 20C.

For three-membered ring, as in ethylene oxide, deduct For four-membered ring, as in cyclobutane, deduct For five-membered ring, as in furan, thiophene, deduct For pyridine, deduct For benzene ring, deduct For naphthalene ring, deduct For anthracme ring, deduct Molecular Volumes 14.3 N2 0 25.6 NH, 31.2 H,O 29.9 H& 3c1.7 cos 34.0 Clz 44.8 Br 2 23.6 12 20.0

H, 0,

N,
Air

co coz
so2
NO D,O *

36.4 25.8 18.9 32.9

*Estimated value.

hol, illustrated in Figure 9, x is found to be 1.5. It is of interest to compare t h e values of x with the association numbers recommended by Jacobsen ( 4 ) from intermolecular freelength relationships as given in Table 3. Although Jacobsens association numbers a r e larger than the present association parameters the agreement in order of t h e solvents suggests t h a t the general concept of t h e association effect may be valid. The results further suggest t h a t the methods of Jacobsen might be used to select an association parameter which normally would lie between t h e values of 2.6 for water and 1.0 f o r unassociated solvents. By use of the given association parameters the data or diffusion in water a r e correlated by Equakion (5) with a n average deviation of about 6%. Data f o r methyl alcohol It should a r e predicted within 11%. be noted t h a t the experimental data for methyl alcohol systems a r e known to be of rather low precision in the original source.
DISCUSSION General Comment. The correlation represented by Equation (5) is satisfactory for estimation of diffusion coefficients in dilute solutions with sufficient precision or most engineering purposes, i.e., about 10% average error. It must be emphasized that the diffusion process is extremely complex and that any rigorous treatment must consider solute-solvent interaction in a more detailed manner than the present relation could possibly imply.? Although the present functional relationship of diffusion coefficient to solute molal volume rests upon some qualitative theo*For 285 points
among

7v0.6

M 12
FIG.6. DIFFUSION IN
UNASSOCIATED LIQUIDS.

Figure 7 squarely onto t h e curve of Figure 6. Thus t h e correlation for diffusion in water and also in nonassociated solvents may be expressed by the general equation

The association parameter x is introduced to define the effective molecular weight of t h e solvent with respect t o t h e diffusion process. F o r nonassociated solvents z = 1 and for water x = 2.6. Diffusion in methyl alcohol is shown similarly in Figure 8, indicating a n association parameter of 1.9, and f o r diffusion in ethyl alco-

251

solute-solvent

systems of this study.

?Diffusion of iodine in aromatic hydrocarbons, for example, bas been excluded from t h e present correlation because of known comples forrnat ion.

Page 268

A.1.Ch.E. Journal

June, 1955

retical foundation, the relationship t o solvent molecular weight is strictly empirical. Some theoretical basis f o r the latter rela-

tionship or some improved correlation would be highly desirable, Only a tenfold range of viscosity is covered by t h e solvents present-

ed here. Study of the correlation and deviations from constancy of t h e group DTIT over more extensive temperature and viscosity ranges is especially needed. Although deviations in constany of DTIT have been observed and might well be expected f o r strongly interacting solute-solvent systems such as iodine in aromatic solvents and acetic acid in ethylene glycol, use of the group seems justified for prediction of t h e effect of temperature on D in absence of experimental data.
Comparison with Other Correlations.

0.4 0.6

2
M "2

8 I0

20

3 V 0.6

FIG.7. DIFFUSION IN

WATER.

Olson and Walton ( 5 ) have proposed a general form of correlation of diffusion coefficients based on surface-tension lowering of the solvent by hhe solute. In view of the special data required no attempt will be made to compare their method quantitatively with the present correlation. Scheibel(7) has fitted the correlation of Wilke to a general equation involving t h e molal volumes of solute and solvent based on the curves f o r water, methyl alcohol, and benzene. I n view of t h e special distinction developed above between water and methyl alcohol as associated solvents and benzene as an unassociated solvent t h e basic assumptions used by Scheibel are believed to be in error. Thakar and Othmer (8) have proposed t h e following general equation developed through the reference substance method :

a l
0 X

a l l -

0.4 0.6

6 8 10

20

9 vo.6 M '4
FIG.8. DIFFUSION IN METHYLALCOHOL.

For diffusion in water only i t was assumed t h a t the group D-ql.1 fitted the temperature behavior better than did the Stokes-Einstein group DqlT. However since t h e ratio [-ql.ll / [-q/ TI changes only 10% for water between 0" and 80C. it is difficult to justify a choice between the two ways of representing the temperature dependence on t h e basis of the relatively limited data presently available. Contrary to conclusions of t h e present study, for diffusion of a single solute in various solvents at 20C. Equation ( 6 ) does not permit variation of Dq,O with solvent molecular weight. Application of Equation (6) to thirty-six representative systems involving diffusion of various solutes among twenty-one unassociated solvents gave rather unsatisfactory results, with an average deviation of over 30 % between calculated and experimental diffusion coefficients.

Vol. 1, No. 2

A.1.Ch.E. Journal

Page 269

CONCLUSION

It is believed that Equation ( 5 )

represents an improvement over previous correlations of diffusion coefficients in dilute solutions. Due

caution of course should be observed in extending the method f a r beyond the range of variables and types of systems included i n the present development. To facilitate

use of t h e method a revised diffusion-faotor chart is given in Figure 10 t o include the association parameter and molecular weight of the solvent.
ACKNOWLEDGMENT

Assistance of Research Corporation through a grant-in-aid is gratefully acknowledged.

N W A T I ON

Q)

0 X

4-

0.4 0.6

4
1VO.6

6 8 1 0

20

M /2

FIG.9. DIFFUSION IN ETHYL ALCOHOL.

C = concentration, g. moles/cc. D = diffusion coefficient, sq.cm./ see. F = diffusion factor, TJDq, (OK.) (see.) (sqxm.) (centipoise) L, = latent heat of vaporization of solvent L, = latent heat of vaporization of water M = molecular weight of solvent N = Avogadros number, molecules per mole N A = diffusion rate of component A , g. moles/ (see.) (sqxm.) V = molal volume of solute a t normal boiling point, cc.lg. mole t =temperature, C . T = temperature, OK. x = association parameter, multiple of nominal molecular weight of solvent to give effective value Z = distance in direction of diffusion q = viscosity of solution, centipoise qw = viscosity of water, centipoise q S o = viscosity of solvent at 20C., centipoise

LITERATURE CITED
1. Arnold, J. H., Ind. Eng. Chem., 22, 1091 (1930). 2. Chang, Pin, and C. R. Wilke, J . Phys. Chem. (in press). 3. Eyring, H., J . Chem. Phys., 4, 283-91 (1936). 4. Jacobsen, Bertil, Association Numbers in Liquid Systems from Intermolecular Free Length Relationships, Karolinska Institute, Stockholm (in press). 5. Olson, R. L., and J. S. Walton, Ind. Eng. Chem., 43, 701 (1953). 6. Perry, J. H., Chemical Engineers Handbook, McGrawHill Book Company, Inc., New York (1950). 7. Scheibel, E. G., Ind. Eng. Chem., 46, 2007 (1954). 8. Thakar, N. S., and D. F. Othmer, Ind. Eng. Chem., 45, 589 (1953). 9. Trevoy, D. J., and H. G. Drickamer, J. Chem. Phys., 17, 1117 (1949). 10. Wilke, C. R., Chem. Eng. Progr., 45, 219 (1949).
Presented at A . I. Ch. E . brew York meetinp.

4 u d 0) v) I u
Y
0

s:

SOLUTE M O L A L VOLUME, CU. cm./gm. mol. FIG.10. GENERALIZED DIFFUSION-FACTOR CHART.

Page 270

A.1.Ch.E. Journal

June, 1955