Machine Translated by Google
March 1950 INDUSTRIAL AND ENGINEERING CHEMISTRY
represent the over-all process has been written. The hydro carbon
effluent when charging a straight chain Cd hydrocarbon always
contained butenes and butadiene, suggesting that these materials may
be intermediates in the reaction. The somewhat higher yields from
butenes than from butane would tend to confirm this. However, these
yield data are complicated, especially in fixed bed operation, by the
heat effects resulting from the highly exothermic nature of the reactions.
ACKNOWLEDGMENT
The authors wish to express their appreciation to J. T. Clarke, C. H.
Culnane, and B. R. Stanerson for their contributions to the experimental
work and to members of the Analytical and Testing Department of the
Beacon Laboratories for carrying out much of the analytical work
i
involved.
Diffusion Coefficients in
Multicomponent Gas Mixtures
D. F. FAIRBANKS AND C. R. WILKE
University of Callfornia, Berkeley, Calif.
An experimental study has been made of the diffusion of
vapors into multicomponent gases by vaporization of iiquids
in a long tube under conditions such that the theory of
diffusion in the semi-infinite column is appli cable. The results
have verified the relation:
D‘A = 1 - yA
YB +yc fYD +***
DAB DAC DAD
where D‘A is the effective diffusion coefficient of gas A with
respect to the total gas mixture; DAB, DAG, DAD, etc., are the
respective binary diffusion coefficients: and yA, ye, ye, etc,,
are the mole fractions of the components
in the mixture.
N A binary system the rate of unidirectional diffusion of a gas,
’
I A, through a second stagnant gas, B, may be expressed by the
equation
where NA = rate of diffusion of A, gram moles per second-sq.
cm.
DAB = diffusion coefficient, sq. cm. per second P =
total pressure, atmospheres R = gas
constant, cc.-atmospheres per gram mole-” K.
T = temperature, OK.
PA = partial pressure of component A, atmospheres X =
distance in direction of diffusion, cm. pi = pressure
of nondiffusing gas B, atmospheres
For diffusion of A into a multicomponent mixture of stagnant gases it
is convenient to express the rat,e of diffusion by an equa tion analogous
to Equation l :
Present addresa, Department of Chemical Engineering, Massachusetts
InaCitute of Technology, Cambridge, Mass.
471
LITERATURE CITED
(1) Auwers, K. V., and KohIhaas, W., J. Pract. Chem., 108, 321–31
(1924).
(2) Fawcett, F. S., J. Am. Chem. Soc., 68, 1420-2 (1946).
(3) Fawcett, F. S., and Rasmussen, H. E., Ibid., 67, 1705-9 (1945).
(4) Lowry, T., and Nasini, A,, Proc. Roy. SOC. (London), A123, 686-91
(1929).
(5) Meyer, V., and Kreis, H., Bey., 17, 1558-63 (1884).
(6) Morton, A. A., “The Chemistry of Heterocyclic Compounds,”
p. 40, New York, McGraw-Hill Rook Co., Inc., 1946.
(7) Rasmussen, H.E. , Hansford, R.C. , and Sachanen, A.N. , IND.
ENO. CHEM., 38, 376-82 (1946).
(8) Shepard, A. F., Henne, A., and Midgley, T., J. Am. Chem. SOC. 56,
1365-6 (1934).
(9) Steinkopf, W., "The Chemistry of Thiophenes," p. 115, Dresden u.,
Leipzig, Theodor Steinkopff, 1941 (J. W. Edwards, Ann Arbor,
Mich., 1944).
RECEIVED August 31,1949.
where 0; is some proper effective diffusion coefficient for com ponent
A, which will be a function of the gas composition. In this case pi is the
sum of the partial pressure of all gases other than A.
On the basis of the theories of Maxwell (6) and Stefan (8, 9), Wilke
(10) has derived an expression given in Equation 3 for the effective
diffusion coefficient and discussed its general use in dif fusion
calculations.
1 - OF
D‘A = (3)
ya+yc +E+.
DAB DAC DAD
where yB, ye, etc., are the mole fractions of components A, B, C, etc.,
and DAB, DAC, etc., are the respective binary diffusion coefficients of
component A with respect to each component of the mixture.
The present paper reports the experimental verification of Equation
3 under conditions approximating those assumed in its derivation-
namely, the diffusion of one gas into a mixture of stagnant gases.
THEORY
On the basis of experimental convenience an apparatus was
construeted to operate on the principle of diffusion in the semi infinite
column. In this method a suitable liquid is allowed to evaporate upward
into the gas mixture from the bottom of a long glass tube under
conditions such that a negligible quantity of vapor reaches the upper
end of the tube during the time of the experiment. This method has
been suggested by Arnold (d), who has integrated the differential
equations applicable to this case for diffusion in binary systems. These
binary equations may be extended to multicomponent systems if the
diffusing gas, A, is maintained at concentrations sufficiently low that
the average diffusion coefficient given by Equation 3 remains essentially
con stant during the diffusion process. This is accomplished in prac-
Machine Translated by Google

472 INDUSTRIAL AND ENGINEERING CHEMISTRY Flight. 42, No. 3

The general procedure in the application of this theory to the

TABLE I.FUNCTION F, (2) determination of the diffusion coefficient is to measure the volume
Vapor Vapor of gas, V, displaced from the top of the tube as a function of time,
Mole Mole 6. According to Equation 5 a plot of V zs. straight should be a
Fraction, Fraction,
II* R AND* R line of slope (d‘V/ d61’2j vc-hich may be determined from the best line
0.50 0.6215 through the plot of the experimental data. This formulation for the
0 1 0.55 0.6810
0.05 0.9635 0.60 0.5398 slope is convenient from the experimental standpoint, because only
0.10 0.9268 0.65 0.4976 the time must be referred to a zero value and the volume measurements
0.15 0.8900 0.70 0.4540
0.20 0.8527 0.75 0.4088 may be started at a later and more convenient period of the experiment.
0.25 0.8152 0.80 0.3616
0.30 0.7774 0.85 0.3112 Equation 5 may be solved for the diffusion coefficient:
0.35 0.7381 0.90 0.2546
0.40 0.45 0.7004 0.6613 0.95 0.1893
1 0

tice by using liquids of relatively low vapor pressure. Arnold’s

development will be reviewed briefly, in so far as it is applicable to the
present work.
Consider a vapor, A, diffusing at constant temperature and pressure
from a liquid surface into a mixture of gases in a long cylindrical tube.
The gas space will be assumed initially free of vapor, and the tube will
be assumed of infinite length. As vapori zation proceeds gas is
withdrawn from the upper end of the tube at a measured rate to
maintain the condition of constant pres sure. For this unsteady state
process the following equation is applicable :
(4)
where z = distance in direction of diffusion, cm. (measured from liquid-
gas interface)
YA = mole fraction of A at distance z y: =
mole fraction of A at interface (5 = 0), determined by liquid
vapor pressure (~YA/~z),
= mole fraction gradient at z = 0 D; . = effective
diffusion coefficient for gas A, sq. cm. per
second
e = time, seconds
The boundary conditions for integration are as follows:
y = g* at z = 0 for all values of 6
APPARATUS AND PROCEDURE
Figure 1 is a schematic diagram of the apparatus. A gas mix ture
of the desired composition was prepared prior to each ex periment in
an external mixer (not shown) connected directly to the diffusion tube.
DIFFUSIOW TUBE. The diffusion tube was a vertical borosilicate
glass tube with an average internal diameter of 1.043 em. and a length
of about 200 cm. Liquid entered the tube at the bottom, the inflow
being induced by an initial difference in the liquid level between the
two sides of the U-shaped injector section. The high side of the U was
connected to the top of the diffusion column; as the liquid entered the
bottom of the column, an equal volume of gas left it at the top, and
there was, therefore, no sudden change of pressure in the system at
the onset of the experimental run.
After the liquid was in, stopcocks were shut, isolating the column from
the U.
BAROSTAT. The barostat mas a mercurial manometer in a U
shaped tube, and its purpose was to indicate the difference in pressure
between the gas in the diffusion system and a near- vacuum
maintained in the barostat’s upper chamber.
As the liquid in the diffusion tube evaporated, the slow increase of
pressure inside the system caused the mercury surface in the up per
vacuum chamber of the barostat to rise. A sharp-pointed tungsten wire
was sealed into this upper chamber, and electrical connections were
made so that the wire and the mercury surface

y = 0 at 6 = 0 for all values of z except zero

Integration of Equation 4 with these boundary conditions gives the

relation:

where V = volume of vapor formed in time 6, CG. 6 =

time of diffusion, seconds a =
cross section of tube perpendicular to direction of diffusion, sq.
cm.
F, = factor correcting for deviation from Fick’s law

The correction factor, F,, is defined as follows:

Y:2
F, = ~

z/ F 9
where 4 is defined by the expressions

Arnold has solved Equations 6 to 8 numerically for F, for various

conditions of vaporization and absorption. Values of F, are given in
Table I for the present case of vaporization with zero initial vapor
concentration in the tube. Because F, is a function of y* only, it is
constant for any one experiment. Figure 1. Schematic Diagram of Apparatus.
Machine Translated by Google

March 1950 I N D U S T R I A L A N D ENGINEERING CHEMISTRY 473

would form the contact points of a switch which, in turn, con- trolled and minimize temperature differences between jacketed and un-
a vacuum-tube relay. The relay was constructed after a plan given jacketed portions of the equipment.
by Serfass (7). The relay operated an indicator light, and, thus, the PREPARATION OF APPARATUS FOR OPERATION. Before
making and breaking of contact at the mercury surface were made each run the apparatus was carefully brought to temperature, evacu-
detectable. No arcing was observable in the upper chamber, and the ated, and checked for leaks. The gas mixture was then admitted to
arrangement was found to be very sensi tive in indicating the the apparatus from the mixing section. The procedure of al ternate
attainment of the desired pressure. evacuation and filling was repeated several times to sweep the
apparatus free of residual gases from the preceding experi ment.
After final introduction of the gas, stopcocks were closed to isolate
V.0, I I I I I I I 4 I I the main diffusion section from the U-shaped liquid in jection tube,
and the injection tube was filled with the liquid to be evaporated.
0.7 After a few minutes' preheating by means of a portable water bath
brought up around the injection tube, the liquid was ready for injection
0.6
into the diffusion tube.

MATERIALS. The source and purity of all gases and liquids

in 0.
used in the experiments are summarized in Table 11.
8U'
10.4
3
>o 0.3
TABLE 11. SPECIFICATION OF MATERIALS
0.2 Boiling
Substance Source Point, OC. Remarks
0.1 Hydrogen Electrolytic Univ. of Calif. .. 99.9% hydrogen
'Chemis try Dept.

0 Argon Linde Air Products Go. .. 99.0% A, 1.0% $"z

0 10 PO 30 40 50
Air Dried over PnOs
TIME, SEC. Atmospheric ..
before me
Toluene Eastman Kodak Co. 110.5
'I2 Figure 2. Volume-Time Data for Typical Run Accepted b.p.
110.62" C. (1)
Ethyl pro- Eastman Kodak Co. 98.7- Accepted b.p.
pionate 99.3 99.00 c. (5)
CAPILLARY VOLUMETER. The volumeter was a vertical cali
brated capillary tube which had a nominal internal diameter of 1 mm.
It was connected at the top to the top of the diffusion tube and at the
bottom through a stopcock to a mercury reservoir in a develing bulb. RESULTS

At the start of an experimental run, the capillary was filled with
mercury. As described above, the vaporization of the liquid caused
the pressure in the apparatus to increase until, finally, the mercury
in the barostat came in contact with the tungsten wire.
The time of the contacting was recorded along with the accom-
panying level of mercury in the volumeter. Then, by releasing some
mercury from the volumeter, gas could be withdrawn from the
diffusion tube, lowering the pressure there and breaking the barostat
contact. The vaporization of the liquid continued, and the pressure
of the gas again increased until contact was again made. Once more
the time and the volumeter scale reading were noted. The process
was repeated until all the mercury had been emptied from the
volumeter.
With the aid of the volumeter calibration the volume corre
sponding to each scale reading could be calculated. Thus a series
of data giving the increase of vapor volume with time was ob tained.
Zero volume was arbitrary, only the change in volume being
obtained. Zero time was the moment that the liquid en tered the
diffusion tube. The volume-time data formed the basis for the
calculation of the diffusion Coefficient.
TEMPERATURE CONTROL. It was necessary to keep the
For each run, volumes of gas evolved from the top of the tube
were plotted against the square root of time on rectangular co
ordinates. Figure 2 shows a typical plot, in this case for the dif fusion
of ethyl propionate into a mixture of 20% air-80% hydro gen. From
the slope of the best straight line through the data the diffusion
coefficient was calculated by Equation 9.
The vapor pressure of ethyl propionate was calculated by inter
polation of data from the International Critical Tables (S), and the
vapor pressure of toluene was calculated from data published by the
National Bureau of Standards (1).
For purposes of comparison the diffusion coefficients were con
verted from their values at room temperature and prevailing
barometric pressure to the standard diffusion coefficient used in the
International Critical Tables (4) defined by the relation:
0.35 I
I I I I I I I I

diffusion tube at constant temperature to avoid thermally induced

con vection currents and volume fluctuations.
0.30

To obtain close control the diffusion tube was encased in a wa-

ter jacket through which about 2 liters per minute of water were 0.25

pumped from a 50-liter constant temperature bath. The tem perature

of the water bath was maintained within 10.01 C. by a sensitive
0.90
mercury-expansion switch which controlled an electrical heating unit
immersed in the bath. The bath was stirred continu ously and was d
well insulated. Corrections to the observed gas volumes were made 0.1 5
for slight fluctuations in the water bath tem perature.
In order to observe the temperature of the evaporating liquid and 0.10
detect any evaporative cooling effects, a fine-wire Chromel Alumel
thermocouple was sealed through the wall near the bottom of the
diffusion tube, and was placed so that the surface of the liquid was
about 1 mm. above the thermocouple junction. 0*05 t 1
The apparatus was housed in a windowless, tightly shut, sub 01 I I I I I I I I
I.

terranean room to obtain freedom from drafts and room tempera 0 20 40 60 80 100
ture fluctuations. Corrections were made for the effects of small
MOLE % OF AIR IN GAS MIXTURE
variations in room temperature on the unjacketed tubing in the
varostat and volumeter. Experiments were conducted at ap Figure 3. Diffusion of Ethyl Propionate into Hydrogen
proximately room temperature t,o facilitate temperature control Air Mixtures
Machine Translated by Google

474 INDUSTRIAL AND ENGINEERING CHEMISTRY Flight. 42, No. 3

0.5 f I 1
I

I I I DISCUSSION

The present experimental techniqoe gives values of D which are

self-consistent and reproducible. Furthermore, the straipht line
1 method.
relation obtained in all the runs in the plot of V us. is in
agreement with the theory and supports the validity of the

It is of interest to compare the iesults obtained for the binary

'\ - systems with the corresponding results of other investigators.
\
0.2 - - For diffusion of ethyl propionate into air Winkleman (4) obtained data
giving Do = 0.0653 sy ciii. per second compared to an average value
- \ -
'\ of 0.069 sq. cin. per second for the present \\orl<.
'\ Mack (5) has measured the diffusion of toluene into air, leading to a
0.1 - '\ - \
value of I), = 0.071 sq em. pel second as opposed to the value of
- - 0.081 sq. cin. pel second for the present work.
0. I I I I

I I 1 A possible explanation for the loir el diffusir rties repoi teci by

I
0 AFTER 40 60 80 these investigators may be aclvniiced 011 the basis of the dif ferent
100
MOLE 70 OF ARGON IN GAS MIXTURE experimental techniques employed. Both Winkleiiiari and Mack
einployed the steady-state evaporation of tile correspond ing liquid
Figure of Toluene into Hydrogen-Argoll
Mixtures in a tube. Because of the greater rate of diffusion ein ployed in
these steady-state methods, as opposed to the present unsteady-
state method, sigiiiticaiit evaporative cooling of the liquid may have
where D is the diffusion coefficient determined at !Po K. and P mm., occurred. The effect of evaporative cooling is to lower the reported
Do is the standard diffusion coefficient, To is the standard value of the diffusion coefficient through use of an erroneously high
'
temperature, 273.1 K., and Po is the standard pressure, 760 inni. vapor pressure in the calculations. Mack's method further involved
adsorption of tlie vqms in charcoal at the
erid of the diffusion tube opposite the
liquid, and as sunied that the adsorbed
TABLE 1x1. EXPERIMENTAL DATA AND CALCCL.4TED 1tEsULTS
vapors exerted no appreciable vapor
Pressure, dV pressure over the charcoal. Any error in
Gas Composition, Temp., Mm. D, Sq. DO, Sq.
Liquid Run Volume % C.
Hg Cc./Sec.l/* Cm./Sec. Cm./Sec. this assumption due to ewessive vapor
Ethyl pro- 34 27.9 763.0 0.0162 751.7 0.085 0.070 adsorption would lead to an erroneously
pionate 35 27.9 0.0162 755.2 0.0161 0 086 0.071
36 27.8 760.7 0.0174 759.8 0.082 0.067 low value for the diffusion coefficient.
44 28.0 0.0179 758.1 0.0209 0.098 0.081
47 29.0 761.2 0.0276 760.1 0.103 0.086
42 27.9 0.0308 753.7 0.0358 0.140 0.115 When the present method is used,
45 27.9 755.64 0.064 0.060 0,247 0,204
46 27.9 0.060 0.307 0.252 errors due to an initial disturbance of the
~~ 27.9 0.410 0.334
39 27.9 0.414 0,339
gas in the diffusion tube accompanying
40 41 27.8 0.374 0.306 injection of the liquid or any other dis
Toluene 25 100% on 27.7 70.2 0.013 0 0.081 27.6.20.20.0 % 2 0.019.0 % 0.01.0% 0.0.0 % 0 0.019.7.0 turbance such as thermal convection cur
26 % 0 0.019.0 % 28.0 756.8 0.0194 0.204 0.167 16.5% A, 83.5% Hz 28.0 757.1 0.0229 0.286
27 0.234 100~o Hs 28.0 758.8 0,0324 0.571 0.469 33.3% on , 34.2 % HI rents, room vibrations, etc., would lead to
49
50
high values for the diffusion roefficient.
51 In view of these possible sources of error.
52
63 the agreement between the results of tlie
54
28.1 759.4 0.0159 0.136 0.112 present,study and those of the investiga
t'ors noted above is believed to be satis
factorv. illthough further study is dc sirable
I

to reconcile the differences in i esults

The essential results of the experiments are tabulated in Table between these experimental techniques, this was not given further
111. Values of Do for diffusion of ethyl propionate into hydrogen air attention in the present work, because the primary ob jective was
mixtures are plotted as a function of gas composition in Figure 3. to study the mixture properties with respect to dif fusion. The data
The data for diffusion of toluene into hydrogen-argon mixtures are are believed to approximate the absolute values of the diffusion
shown in Figure 4. The solid curves in Figures 3 and 4 are calculated coefficient sufficiently closely, and to be suffi ciently reproducible
from Equation 3, based on the terminal points of the curves and internally self-consistent to constitute a valid experimental
corresponding to the respective binary diffusivities. verification of Equation 3 for the effecti\ e diffusion coefficient in
The agreement between Equation 3 and the experimental points is multiromponent mixtures.
believed to be satisfactory, and constitutes experimental verifi cation
NOMENCLATURE
of Equation 3 and the theory leading to its derivation.
The dotted lines in Figures 3 and 4 are the results obtained as a = cross-sectional area perpendicular to direction of diffusion, sq.
cm.
suming the binary diffusivities to be additive on a mole fraction basis, I) = diffusion coefficient, sq. cm. per second D'
and illustrate the serious error that may be involved in the employment = effective diffusion coefficient with respect to a multicompo nent
of this relationship. mixture, sq. cm. per second I), =
Equation 3 is further verified for a four-component system (run 54 standard diffusion coefficient at 273.1' K., 760 mm., sq. cm. per
second F, =
in Table 111) for diffusion of toluene into an air-argon hydrogen correction factor for deviation from Fick's law 1v = rate
mixture composed of approximately one third of each gas. The of diffusion, moles per second-sq. cm. p = partial
experimental value of Do is 0.112 sq. cm. per second compared to pressure, atmospheres P = total
a value of 0.104 sq. cm. per second calculated by Equation 3. Again, pressure, atmospheres R = gas
constant, c,c. atmosphere per gram mole-" K.
the simple additive relation on a mole frac tion basis is seriously in T = temperature, K.
error, giving a value for Do of 0.203 sq. cm. per second. V = volume. cc. 5
= distancl? in direction of diffusion, cm.
Machine Translated by Google
March 1950 INDUSTRIAL AND ENGINEERING CHEMISTRY
y = mole fraction y*
= equilibrium vapor mole fraction at gas-liquid interface e = time,
seconds + =
parameter defined by Equation 7, a function of y* SUBSCRIPTS
A: B, C, etc. =
components A, B, C, etc., of gas mixture i = stagnant gas or
gases o = gas-liquid interface
or standard temperature and pressure
LITERATURE CITED
(1) American Petroleum Institute Research Project 44, Table 5K,
Washington, D. C., National Bureau of Standards, 1945.
Low Temperature Stiffening
of Elastomers
S. D. GEHMAN, P. J. JONES, C. S. WILKINSON, JR., AND D. E. WOODFORD The
Goodyear Tire and Rubber Company, Akron 16, Ohio
The generally observed Stiffening of elastomers at low
temperatures may be supplemented by the occurrence of
crystallization if the elastomers have sufficient regularity of
molecular structure and other conditions are favorable.
To study these effects, observations of the progressive stiff
enings of elastomer compounds at low temperatures were
made by measurements of the relative torsional modulus of
test strips mounted in racks which were stored at low
temperature. The periods of observation extended over 30 to
60 days with temperatures in the range from -59” to -18’ C. A
correlation was found between density changes due to
crystallization as determined dilatometrically and torsional
stiffness changes. Factors studied in addition to
I N EXTENDING the low temperaturerange for useful applica- tions
of elastomers, the phenomena of spontaneous stiffening due to
crystallization or other causes must be taken into account.
Crystallization, accompanied by progressive stzening, sometimes
occurs over a long period of time. In general, this may be ex pected in
the case of elastomers which have sufficient regularity of molecular
structure to show the more familiar phenomenon of crystallization upon
stretching. Because such elastomers include some of the most useful
types of rubber, it is important to under stand under what conditions
spontaneous crystallization at low temperatures is to be anticipated.
Work reported from the National Bureau of Standards (10, 11), dealing
principally with the crystallization and melting of un vulcanized Hevea
rubber, furnishes a useful background for the crystallization phenomena
with vulcanizates. Observations of the progressive low temperature
stiffening of a class of technical vulcanizates have been published by
Gregory, Pockel, and Stiff (7). Forman (6) investigated the comparative
cold hardening of neoprene, GR-S, and Hevea vulcanizates over a
period of time at low temperatures. Conant and Liska (3) observed
progressive stiffening with plasticized butadiene-acrylonitrile vulcanizates,
which was attributed to low temperature incom patibility or crystallization
of the plasticizer. Forman reported similar effects for plasticized
compounds of Neoprene Type FR.
Beatty and Davies (8) studied effects of time, temperature, and stress
on the behavior of rubber at low temperatures.
The present work comprises a description of what appears to be an
improved technique for securing extensive data on the progressive
stiffening of elastomers, a survey of the effects of
475
(2) Arnold, J. H., Trans. Am. Inst. Chem. Engrs., 40, 361 (1944).
(3) International Critical Tables, Vol. 111, New York, McGraw- Hill
Book Co., 1929.
(4) Ibid., Vol. V.
(5) Mack, J. Am. Chem. Soc., 47, 2473 (1925).
(6) Maxwell, J. C., “Scientific Papers,” Vol. 2, p. 57, Cambridge,
England, Cambridge University Press, 1890.
(7) Serfass, E. J., IND. EN. CHEM., ANAL. ED., 13, 262-3 (1941).
(8) Stefan, Sitzber. Acad . Vienna, 63 (2), 63 (1871).
(9) Ibid., 65 (2), 323 (1872).
(10) Wilke, C. R,, Chem. Eng. Prog., 46, 95-104 (1950).
RECEIVED July 5, 1949.
temperature included compounding variables such as cure
and plasticizer content. Some data were obtained to show the
acceleration of crystallization due to compressive stress.
Progressive stiffening due to crystallization was observed for
vulcanizates of Hevea, Neoprene Type GN, and Butyl rubber
as well as for an 85-15 butadiene-styrene redox-type copolymer.
The results were interpreted in accordance with concepts of
crystal nucleation and growth with necessary modifications
due to the molecular structure of elastomers. Although the
phenomenon of spontaneous crystallization of elastomers
was found to be complicated by many factors, a useful degree
of generality appears in some of the results.
temperature and other significant variables on the crystallization of
typical vulcanizates of technically important elastomers, and a discussion
of the results in order to reach an understanding of their significance in
terms of general principles applicable to crystallization and the molecular
structure of elastomers.
EXPERIMENTAL METHOD
A method of measuring the torsional stiffness of rubber samples
previously described (1, 6) has been adapted for securing data on the
stiffening due to prolonged exposures to low temperature.
Figure 1 is a photograph of the assembled apparatus. The test strips
were mounted in supplementary racks as shown in Figure 2.
Ten of these racks were available for the investigation, so that fifty
samples could be run concurrently.
The torsional stiffness of each test piece was measured at room
temperature and at the storage temperature. The racks were then stored
in constant temperature cold boxes or cold rooms at the desired
temperatures. It was found advantageous to invert glass beakers over
the racks to minimize the chances of any tem perature variation of the
samples due to opening a cold box during the course of the tests. The
racks were removed at suitable inter vals and the test pieces were
measured without changing their temperature, by transferring the rack
to be measured into the ap- propriately precooled insulated cylinder of
the apparatus and carrying the rack to the measuring stand where the
storage tem perature was maintained. After measurement, the rack
was re- moved from the insulated cylinder in the cold box, if it was
desired to continue the test.
In the presentation of the results, the ratio of the torsional modulus
measured at various time intervals to the original