Dec.

, 1958 PERMEATION
OF CELLOPHANE BY DIFFUSION
MEMBRANES 1483

ON THE PERMEATION OF CELLOPHANE MEMBRANES BY DIFFUSION

BY L. B. TICKNOR
Research and Development Division, American Viscose Corporation, Marcus Hook, Pa.
Received April 81, 1968

The assumption that permeation of membranes by liquids is a diffusive process leads t o equations which have the proper
form; Le. the quantity of permeant transferred is directly proportional to the pressure difference and inversely proportional
t o the bulk viscosity of the liquid. It is therefore necessary to consider diffusion as a possible mechanism in liquid permea-
tion of membranes. By comparison of permeation and self-diffusion it is found that the permeability of cellophane to water
and methanol is greater than would be expected from a diffusive flow mechanism; and, therefore, the permeation in these
cases is largely a result of viscous flow. With other alcohols the permeation is about the same as would be calculated for a
diffusive flow mechanism. An approximate calculation shows that the pore radii obtained on the assumption of Poiseuille
flow are small enough to suggest that part of the transport is by a diffusion mechanism. The variation of the permeability
constant with molecular size also suggests that the permeation is not simple viscous flow. These results and other considera-
tions mentioned suggest that, in general, permeation of cellophane membranes by liquids involves both a viscous and a dif-
fusive flow mechanism.

Introduction sides might be diffusing or moving individually

In permeability measurements a gas or a liquid is while those in the center were moving as groups.
forced through a membrane by a pressure differ- The experimental determination of whether per-
ence. I n liquid permeation the volume of material meation is a viscous flow or diffusion process is of
transferred in unit time through unit cross sectional necessity dependent on a comparison of the per-
area is found to depend linearly on the pressure meation rate with the rate one would expect if dif-
difference in accordance with the equations fusion were the only process taking place. If the
molecules of permeant moved through the mem-
Q = -A(AP) = - (1) brane in groups, the permeation rate would be ex-
1IL pected to be much larger than if they moved indi-
A is termed the “permeability”, K is the “permea- vidually. In order to make this comparison we have
bility constant,” 1 is the viscosity of the liquid and calculated a “coefficient of diffusion” from rates of
L is the membrane thickness. This equation is permeation of cellophane by water and alcohols.
derived readily on the assumption of Poiseuille flow These diffusion coefficients are compared with the
through capillaries. It also may be obtained from self-diff usion coefficient for each permeant. To
the assumption of a diffusive flow through the mem- make this comparison valid the frames of reference
brane. for the calculation of the diffusion coefficients must
Whether the permeation of cellophane mem- be equivalent in the two cases and the frictional
branes by liquids is a viscous or diffusive type of forces for the movement of the permeant molecules
transport is difficult to ascertain. These two types in the membrane must be comparable to those in
of flow differ in whether the molecules move in the bulk liquid. In the calculations given below
groups or move individually in a random walk these two requirements are essentially satisfied;
manner. Since it is conceivable that with a mem- the frames of reference are equivalent because the
brane of very small capillaries the molecules of the volume in that part of the system wherein diffusion
permeant would be separated from each other by takes place remains constant,* and the frictional
the material of which the membrane structure is forces are similar because of the hydroxyl groups
composed, the permeant molecules would only on the cellulose molecule and the solvation of the
move as individuals and their movement could be cellulose by the permeant.
in a random walk manner. The gaseous permea- Equations for Diffusive Flow
tion of polymer membranes involves a real concen-
tration gradient and has often been interpreted If we make the assumption that all the permea-
as a diffusion pr0cess.l In the permeation of tion is a diffusive flow process resulting from the
membranes by liquids there is essentially no con- difference in chemical potential of the permeant at
centration gradient; but there is a gradient in the the two sides of the membrane, we may use the
chemical potential because of the gradient in hy- procedure of Mauro2and derive the equation
drostatic pressure. Whereas the gradient in hy- DeV( A P )
& = -~
drostatic pressure would lead to mass flow, the gradi- RTL
ent in chemical potential would lead to a diffusive
flow. Which type of flow predominates would where D is the coefficient of diffusion, E is the frac-
depend on the sizes of the capillaries, the size of the tional void volume of the membrane, V is the mo-
permeating molecules and the bonding of the per- lar volume of the bulk liquid, AP is the difference
meant to the film. We might expect that in some in hydrostatic pressure on the two sides of the
cases, as has been indicated by Mauro,2 the total membrane and L is the thickness of the swollen
permeation would be a combination of viscous flow membrane. The concentration of the permeant in
and diffusion. A distribution in capillary sizes the membrane has been taken as E/V. This is
would lead to a combined mechanism; and even equivalent to correcting the area of the membrane
within some capillaries the molecules along the to that fraction of the area that is permeable to
the liquid. Equation 2 is identical in form to the
(1) R. M. Barrer, “Diffusion in and Through Solids,” Cambridge
University Press, Cambridge, 1951.
first part of equation 1.
(2) A. Mauro, Science, 126, 252 (1967). (3) G. 8. Hartley and J. Crank, Trans. Farodav SOC.,45, 801 (1949).
1484 L. B. TICKNOR Vol. 62

I n order to relate the permeation by diffusion to TABLEI

the second part of equation 1 we need a relation CALCULATED “DIFFUSION
COEFFICIENTS”
FOR THE PERMEA-
between the diffusion coefficient D and the viscos- TION OF A WATERSWOLLEN CELLOPHANE
MEMBRANE BY
ity 7. Johnson and Babb4 have shown that the WATER
Stokes-Einstein theory, the kinetic theory of Ey- Temp., OC. A X 1011 D x 105
ring, the hydrodynamic-thermodynamic approach 20 1.95 37
of Hartley and Crank, and the theory of irreversible 25 2.17 42
thermodynamics all lead to a relationship between 28 2.28 45
diffusion and viscosity coefficients of the form 33.5 2.65 53
D = BRT
-- 38 3.08 63
(3) 40 3.30 67
N?l
where B depends on various molecular dimen- TABLEI1
sions. Substitution of equation 3 into 2 gives CALCULATED “DIFFUSIONCOEFFICIENTS”
FOR THE PERMEA-
TION OF CELLOPHANE MEMBRANES BY ALCOHOLS AT 30”
-Membrane swollen in-
Self-diff. 00% Ethanol Water 3% NaOH
This equation is identical with the second part of Alcohol coef. X 106 D X 106 D X 106 D X IO5
equation 1. Thus we see that the assumption of a Methanol 2.5 2.97 7.37 10.3
diffusion mechanism leads to permeation equations Ethanol 1.16 0.87 1.95 2.3
identical to those obtained through the assumption I-Propanol 0.73 0.33 0.70 1.06
of viscous flow. Therefore, we must consider dif- I-Butanol 0.58 0.20 0.39 0.47
fusion as a possible mechanism for the permeation
of membranes by liquids. fusion coefficients to be no larger than the self-
diffusion coefficients of the bulk liquid if diffusion
Application of Equations through the membrane were the controlling process.
Madras, McIntosh and Mason5 have measured I n the case of water permeation the values in Table
the permeability of cellophane to pure liquids with I may be compared with the results of Wang, et al.,
a van Campen6 type osmometer. They used a who obtained self-diffusion coefficients for water
commercial type cellophane after swelling it in ranging from about 2.8 to 4.4 X a t 25 to 45”.
various media and conditioning it in the permeating Since the constants in Table I are about ten times
liquid. Their measurements were carried out a t as large, it appears that a faster transfer mecha-
30” except for water for which the temperature was nism, buch as viscous flow, is the predominant trans-
varied. They measured the fractional void vol- port process.
ume e by two methods. We will use the values In the case of the alcohols, the calculated dif-
derived from the thickness of the swollen mem- fusion coefficients in Table I1 are of about the same
brane because that method appears to be more reli- order of magnitude as the self-diffusion coefficients.
able.7 They found that the fractional void volume For methanol the coefficients for all three mem-
was nearly independent of the permeant, but it in- branes are somewhat larger than the self-diffusion
creased with stronger swelling agents. They ob- coefficient. This suggests a certain amount of mass
tained values of 0.38, 0.59 and 0.62, respectively, flow in these cases. For t h e other alcohols it is
for swelling in 60% ethanol, water and 3% NaOH. found that for the least swollen membranes the
From equations 4 and 1 we have self-diffusion coefficient of the liquid is larger than
D=- RTLA the calculated diffusion coefficient. In these cases
eV (5) it appears that diffusive flow may predominate.
and we may calculate D from the values of A re- It was suggested above that the size of the capil-
ported by Madras, et al. (their P values). I n laries would determine whether the flow was pre-
equation 5 D has units of cma2/sec.,A is in cm.2- dominantly viscous or diffusive. The following
sec./g., L is in cm., T is in OK., R is 8.314 X lo7 simple calculationlo shows under what conditions
ergs/mole-deg., and V is in cc./mole. one may expect a transition from one type of flow to
Application of equation 5 t o Madras’ data on the the other. Consider a system of capillaries of uni-
water permeability of cellophane (taking L equal to form
84 p ) yields the results listed in Table I.
radius r and length L passing perpendicularly
through a membrane. If the fractional void vol- ..
When we use the data of Madras, et al., for the ume is e, the number of capillaries per unit surface
permeation of alcohols, we obtain the calculated area is e/nr2. Then, according t o Poiseuille, the
“diffusion coefficients” given in Table 11. I n addi- rate of transport &’ of a liquid in cm.8/cm.2-sec.is
tion, the values of the self-diffusion coefficients for
the alcohols as estimated from a paper by Parting-
ton, et aZ.,8are given. In order to compare this result with diffusive
Discussion transport the Stokes-Einstein relation
In general, we would expect these calculated dif- _ - 6rr*q
(4) P. A. Johnson and A. L. Babb, Chem. Reus., 66, 387 (1956).
(5) S. Madras, R. L. McIntosh and S. E. Mason, Can. J . Research. where r* is the radius of the molecules, is substi-
27B,764 (1949). tuted into equation 2 giving
( G ) P. van Campen, Rec. trau. chim., 60, 915 (1931).
(7) J. M . Kuzmak, Ph.D. Thesis, McGill University, 1953. (9) J. H. Wang, C. V. Robinson and I. S. Edelman, J . A m . Chem.
(8) J. R. Partington, R. F. Hudson and K. W. Bagnall. Nalure, Soc., 76, 466 (1953).
169, 583 (1952). (IO) Suggested by Professor Carl Wagner.
Dec., 1958 OF HYDROGEN-BONDED
STABILITY POLYPEPTIDE
STRUCTURES
IN SOLUTION 1485

& = - - ET’( A P ) If viscous flow were the only transport mech-

6~r*qNL (7) anism, the permeation of a membrane by various
The two transport rates may now be compared by liquids (assuming they did not change the pore ra-
taking the ratio of Q’ to Q dius) should yield the same permeability constant.
Madras, et al., however, found permeability con-
stants for water swollen membranes varying from
3.6 X 10-l6 cm.-2 for permeation by methanol to
For close-packed hard spheres the fraction of the 1.8 x 1O-la for butanol and 0.8 X for hep-
total volume that is occupied by the spheres is tane. On the other hand if the mechanism were
about 0.74. For a real liquid we can only estimate partly viscous flow and partly diffusive flow, it
this ratio and we will take the convenient value of would be expected that an increase in the molecular
4/9. Then, substitution in equation 8 for the size of the permeant would cause a transfer from a
volume in terms of the molecular radius shows that convective to a diffusive mechanism. Thus, the
the two rates of transport would be equal if observed results can be explained by a combined
r =2r* (9) mechanism.
Therefore, within the limitations of the above as- The increase in the calculated diffusion coeffi-
sumptions, viscous flow across a membrane would cients with increased degree of swelling suggests an
predominate if *the capillary radius were much increase in the size of the capillaries and a transfer
larger than the molecular radius. If, on the other to more convective and less diffusive flow.
hand, the capillary radius were only slightly larger Other investigators have found evidence for a dif-
than the molecular radius, diffusive flow would be fusion mechanism in membrane permeation. Bar-
important. rer and others’ have treated the gaseous permeation
Madras, et al., calculated the effective pore ra- of organic membranes as a diffusive process. It
dius to be about 1.5 X lo-’ cm. for water permea- has been shown by McBain and Stuewer12 that an
tion of their cellophane membranes. Their method ordinary cellophane membrane when swollen in
of calculation (assumption of Poiseuille flow) gives water and used as an ultrafilter will partially hold
values of 6 X 10-8 cm. for permeation by ethanol back small molecules and ions such as sucrose, dex-
of a water swollen membrane and 3.6 X cm. trose and potassium chloride. C. E. Reid and E. J.
for the permeation by butanol of a membrane pre- Breton13have reported on the use of cellulose acetate
viously swollen in 60% ethan0140% water. Thus, membranes for the ultrafiltration of a 0.5% aqueous
equation 8 suggests that for water viscous flow NaCl solution; salt concentration was reduced by
would predominate and for the alcohols both types about 95%. These results suggest that, in general,
of flow would occur, which agrees with the results of permeation of membranes by liquids may be inter-
Tables 1 and 11. It would be difficult to separate preted in part by a diffusion mechanism.
the permeation into these types of flow, although it
has been attempted for gaseous permeation of (12) J. W. McBain and R.F. Stuewer, ibid., 40, 1157 (1936).
(13) Charles E. Reid and Ernest J. Breton, “Water and Ion Flow
membranes. l1 through Cellulose Acetate Membranes,” a paper presented a t the 130th
(11) H. L. Frisch, THIEJOURNAL,
60, 1177 (1950). National Colloid Symposium, Madison, Wisc., June 18-22, 1956.

T H E FACTORS AFFECTING THE STABILITY OF HYDROGEN-BONDED

POLYPEPTIDE STRUCTURES I N SOLUTION
BY JOHN
A. SCHELLMAN*
Division of Physical Chemistry, Chenzistry Department, University of Minnesota, 1cf inneapolis, Minn.
Received A p r i l 21, I968

A previous discussion of the thermal transition of helices is extended t o include the effect of fluctuations, mixed organic
solvent systems and pressure. It is found that though fluctuation effects are very large for helices of low stability, the
average thermodynamic properties are altered in a relatively minor way. The fluctuation treatment reveals an asym-
metry in transitions which has been verified experimentally. An essential difference between the transitions of helices
in pure and mixed solvents is demonstrated. The form of the results gives an explanation for the inverted transitions
which have been observed in proteins and polypeptides. The calculated properties of helices should have semiquantitative
application t o proteins.

It is now known that the various hydrogen- explained as the disruption of these weak struc-
bonded polypeptide structures which have been tures by heating, denaturants, pH changes, e t ~ . ~ J
proposed1-a have a sufficiently limited stability in This interpretation of denaturation is not always
aqueous solution that a large number of the clas- valid since cooperative transformations are known
sical denaturation reactions can be qualitatively in which the configuration of the polypeptide back-
* Department of Chemistry, University of Oregon, Eugene, Oregon.
(1) M . L. Huggins. C h e n . Revs., 3 2 , 195 (1943). (4) W. Kauzman, “Rleohanism of Enzyme Action,” W. McElroy
(2) L. Pauling, R. B. Corey and H. R. Eranson, PTOC.Nall. Acad. and B. Glass, Johns Hopkins Press, Baltimore, Md., 1954.
S c i . ( U . S.), 37,205 (1951). (5) J. Schellman, Compl. rend. trav. Lab. Cadsberg. Ser. chin., 29,
(4) B. Low and H. J. Grenville-Wells. ibid., 99, 785 (1953). No. 15 (1955).