ANALYTICAL EDITION

Industrial
VOLUME6 AND E N G I N E E R I N G MARCH
15,
NUMBER2
Chemistry
P U B L I S H E BDY T H E A M E R I C A N
CEEMICAL
SOCIETY
1934

HARRISON E. HOWE,EDITOR

Capillary Penetration of Fibrous Materials

R. L. PEEK,JR., AND D. A. MCLEAN,Bell Telephone Laboratories, Inc., New York, N. Y.

N ’UMEROUS i n d u s t r i a l This paper is a study of the penetration of the quantitative determination

processes consist essen- liquids into porous materials, with special refer- of i n t e r f a c i a l forces. By a
tially in the penetration slight extension of available
of fibrous materials by liquids.
ence to the use qf capillary rise in strips of fibrous theory, however, it has been
As examples of such processes materials as a test method for the evaluation of the found possible so to analyze the
may be cited the treatment of penetration tension of the liquid-solid system (the data from a test of this type
wood with p r e s e r v a t i v e sub- penetration tension being the product of the sur- as to obtain q u a n t i t a t i v e l y
stances, the dyeing of textiles, and face tension and the cosine of the contact angle). comparative results. Experi-
the impregnation of paper, wall mentally, the method is attrac-
board, and felt with materials
It is shown theoretically that the rate of rise, tive, as it is simple and rapid
which increase their resistance to dhldt, varies linearly with the reciprocal qf the and m a y be r e a d i l y applied
heat or moisture, or which im- height of rise, l / h , and that the slope of the under c o n d i t i o n s affording a
prove their properties as electri- straight line obtained by a plot of these quantities relatively high degree of repro-
cal insulators. Any understand- is proportional to ylr, where y is the penetration ducibility.
ing of these processes must rest
upon a knowledge of the dy-
tension and q the viscosity of the liquid. The
THEORY OF WETTING
namics of capillary penetration proportionality constant is shown to be dependent
into such materials. Moreover, not merely on the average pore size, but on the In order to present the theory
it is known that the interfacial extent of the range of pore sizes represented. a p p l y i n g to t h e method, it
forces which cause such penetra- Using a single solid medium, therefore, relative will be convenient to review
tion are related to the adhesion briefly the concepts involved in
between the penetrating material
values of the penetration tension for various the general theory of capillary
and the medium impregnated. liquids may readily be determined. Experi- a c t i o n i n t h e f o r m i n which
Hence the determination of the mental data supporting the theory are presented. they have been p r e s e n t e d by
forces effective in impregnation recent writers (6, 7). When a
should afford inform&ion bearing on the nature and extent liquid spreads over a plane surface, each unit of area covered
of the adhesion obtaining in the impregnated material. involves the loss of the energy (SSG)required to form unit
The experimental investigation of the capillary penetration area of solid-gas interface, and the addition of the energy
of fibrous materials presents a number of difficulties. Capil- (SSL)required to form unit area of solid-liquid interface,
lary forces are commonly investigated by means of the rise in and of the energy (SLG)required to form unit area of liquid-
capillary tubes or the spreading on plane surfaces. These gas interface. The increase in free energy involved in spread-
methods cannot be applied to fibrous materials. For cellulose ing over unit area is therefore
materials, use might be made of tubes or plates of regenerated
AF = SSL - SSGi-SLG (1)
cellulose, but these might well exhibit different interfacial
forces with a given liquid than would the parent substance. SLGis numerically equal to the surface tension of the liq-
In these laboratories, attempts have been made to use the uid. The quantity Ssa - SSL, a constant of the solid-
method of Bartell and Osterhof ( I ) , in which there is measured liquid pair, is the Freundlich “adhesion tension,” ASL. If
the pressure required to prevent the penetration of a liquid the ratio of the adhesion tension to the surface tension for any
into a porous material packed in a cell. These attempts have solid-liquid system is written as K , so that
not been successful, presumably because of the difficulty of
packing fibrous materials in the cell in such a way as to ASL= KSLG (2)
assure uniform pore size. the condition that spreading will occur-i. e., that AF in
It is, however, possible to study the capillary penetration Equation 1 be less than zero-can be written K > 1. If this
of fibrous materials directly, in some cases, by observing the condition does not hold, K being less than unity, the liquid
rate a t which the liquid will rise in a strip of the material will come to rest with its surface making an angle 0 with the
when dipped in the liquid. While this is a fairly common solid surface, such that cos 0 = K .
qualitative test, it does not appear to have been employed for The case of capillary rise differs from that of spreading in
86 ANALYTICAL EDITION Vol. 6, No. 2

that no liquid-gas interface is formed. Hence for each unit straight capillary, has been confirmed by experiment, as will
area of capillary wall covered, the energy loss is equal to the be shown below.
adhesion tension alone, so that the condition for such rise is I n a similar manner equations can be derived for the rate of
merely that K > 0, or that 8 < 5. The effective force per
motion in various possible experiments in which one im-
miscible liquid displaces another in a capillary tube. I n such
unit length causing capillary rise will be here termed the cases the force acting a t the interface of the two liquid ill
penetration tension, and denoted by the symbol y . The be given by 2 T Ry’, where y’ is the difference ABL - ’,
penetration tension is equal given by Equation 3.
to the adhesion t e n s i o n
when the contact angle 8 is PENETRATION
POROUS
o*
greater than zero. If K > The problem of flow into or through a porous or fibrous
1,the penetration tension is medium is obviously more complex than that of flow through
merely the surface tension a single uniform capillary. The channels in a porous medium
07
(SLB)of the liquid, as this is will be similar in shape and arrangement to the spaces be-
the greatest force which can tween spheres piled together, while in a fibrous material they
‘A?2
06
act on the periphery of the will resemble the spaces within ti bundle of cylinders. The
advancing column. When liquid flowing a t one instant in a large pore space will at the
Qh 05
.I%
K > 1, the l i q u i d will next be divided into streams flowing through smaller spaces,
I
3
spread out along the walls which will unite with each other or with other streams a t the
3
04
of the capillary faster than next large pore space. The simplest model of these channels
the column of liquid will that will bear any relation to the facts is one involving a set of
01 advance. capillaries in parallel alternating in series with larger capil-
It may be further noted laries, each capillary whether small or large varying in radius
at that while ASL cannot be along its length, and therefore similar in its viscous resistance
directly me a su r ed when to a set of shorter capillaries in series.
K > 1, it may be obtained An exact mathematical treatment for the viscous flow
indirectly by measuring the through such an aggregation of channels could be developed
0 capillary rise of the liquid only by assuming a specific arrangement of pore spaces.
“ALULS OF I(
in a tube containing it and The essential distinction between the flow through such
FIGURE1. AIR, B/R2, A l d B another liquid, immiscible channels and through a group of uniform parallel capillaries,
As OF UN1- with the first, for which ASL however, can be brought out by considering a hypothetical
FORM DISTRIBUTION OF PORE
SIZES is known (as will be the case case in which each channel consists simply of a number of
if for the second l i q u i d single capillaries in series. This case will differ from the
K < 1). For at the interface of the two liquids the equi- model described in the preceding paragraph only in that each
librium condition gives group of small capillaries in series that feed or are fed by
SLL’COS -
8’= SsL SsLt = ASL- ASL’ (3) larger capillaries will be replaced by single capillaries having
the same resistance to viscous flow.
where XLL’ is the energy per unit area of the interface be- For this hypothetical case, then, it will be assumed that
tween the two liquids and 8’ the angle this interface makes the pores in the medium vary in size in accordance with some
with the solid wall. distribution function, such that the fraction dN of the total
number of pores having a radius lying in the interval from
CAPILLARY
PENETRATION +
r to r dr is given by
The penetration of liquids into capillary tubes has been dN = @ ( r ) dr (5)
considered by Washburn (9) and Bosanquet ( 2 ) . The lat-
ter’s treatment is more complete than the former’s, in that I n any single channel, then, the various capillaries in series
inertia effects are considered. These were shown by Wash- of which it is constituted will occur with a frequency given by
burn to be trivial after the penetration has proceeded a Equation 5. Now the pressure drop per unit length in a
measurable distance, and for small capillaries may be neg- uniform capillary is given by Poiseuille’s law as
lected. Taking the simplest case, in which the viscosity of P 81 dV
the displaced gas is neglected, the treatment consists simply E=;& Zi
in substituting the force producing the motion for the total
force producing flow in Poiseuille’s law. As noted above, the Evidently the pressure drop through a series of capillaries
capillary force per unit length of periphery, or penetration in series is additive, so that the total pressure drop through a
tension, will be here written as y. The total force producing channel of length h is given by
motion is given for a circular capillary by 2 r R y , less the
hydrostatic head, so that for a vertical capillary
dh - R2P 1
,-=,zLrn7
P 87 dV @(r)dr

dt - 8 T h = 8 e-
h
(2R7 - R2ghd) (4) If the particular channel in question has, a t the height h,
a radius R, the pressure P effective in producing vertical
where R is the radius of the capillary, h the height of the capillary rise is given by
column, dhjdt the rate of penetration, g the acceleration due
to gravity, and and d are the viscosity and density of the T RZP = 2 T Rr - T r2ghd

liquid, respectively. Washburn (9) gives an integral of so that the rate of efflux from this channel a t the height h
Equation 4, but this is not convenient for comparison with will be
actual data. By evaluating the slopes of tangents drawn to
a plot of h versus t, or by means of interpolation formulas,
values of dh/dt may be determined, and if these are plotted
against l/h, a straight line should be obtained, if Equation 4
applies, The validity of this equation, as applied to a single
Marc6 15, 1934 INDUSTRIAL AND ENGINEERING CHEMISTRY 87

The total efflux will be given by adding the efflux from the integrations indicated in Equation 6, the constants A and B
various channels, and as the radii that these have a t any of Equation 7 are found to have the following values
height are distributed in accordance with Equation 5, the
9x3 log, x
total efflux is given by
A = (a! - 1) ( 2 2 + + 1)2 R
x
9x3
B =
(22 + +
x 1 ) 2 RZ
I where x = R1/R2and Rz is written simply as R. I n Figure 1
are plotted as functions of x, the values of AIR and B/R2
**r given by these equations. If the values of A and B thus
No#;& any height the rate of rise, dhldt, will not be uni- obtained are substituted in Equation 7 and the resulting
form, btrt will vary from channel to channel. If it is as- expression com-
sumed, however, that the average distance along a channel
from maximum to minimum radii is small, the rate of rise will
be apparently uniform, the variations from channel to channel
pared wit,h Equa-
tion 4. it will be
seen that the two 0
Do6rr
1 I I I' IY1
appearing only as minor irregularities in the surface marking expressions cliff er
the average height attained throughout the medium. This is, %
I:
only in that Equa- B
indeed, what is actually observed. This approximately tion 7 then in- 0

uniform rate of rise, dhldt, can be evaluated by equating the volves the func-
efflux corresponding to a uniform rate in every channel with 3 -
tions of z plotted
that given by the preceding equation. It follows that in Figure 1. If x -0

equals unity-i. e.,

from which,
Equation 6 gives the rate of rise in a porous medium having
the structure of the hypothetical case considered, in which
each channel consists simply of a number of single capillaries
in series, or of a single capillary varying in radius along its
length. This hypothetical case differs essentially from that
obtaining in any actual porous or fibrous medium only in
that in the latter each channel will include groups of small
capillaries in parallel.
For practical purposes, Equation 6 may be written in the
form
dh A y Bdg
& = 4Th - (7)
where A and B are dependent only on the distribution of pore
sizes in the medium, and are therefore constants for a given
medium. It follows that the rate of rise, dh/dt, should give a
straight line when plotted against l / h . Using the same
medium, the intercepts of such straight lines, obtained with
different liquids, should be proportional to d / v , affording a
check on the equation. As the slopes of these straight lines
should be proportional to y/q, relative values of y can be
thus determined.
Before presenting a comparison of experimental data with
Equation 7 , there will be given a qualitative demonstration of
the significance of this equation in so far as the effect of the
distribution of pore sizes is concerned. This is readily done
by assuming as a simple ideal case that all pore sizes between
upper and lower limits are equally frequent, or that the
distribution function of Equation 5 has the form
@(r) = Cfor R1 < r < Rz
@ ( r )=Oforr<R1andr>R2 (8)
As by definition the bitegral of Equation 5, s_", dN, must
be unity, the constant C of Equation 8 must equal l / ( R z -
RI).Applying to the form of (r) given by Equation 8 the
t h e p o r e size i s 0012 O M 0020 0024 0028 0032 0036 0040 004d
VALUES OF i
constant-the two
expressions are FIGURE 2. 811 dL/dT vs. 1 / L FOR WATER
IN GLASSCAPILLARY
identical, as they
should be. If x is
small, however, corresponding to a wide variation in pore
sizes, the two expressions differ in the relative magnitude of
their terms.
The other quantity, A / a , plotted in Figure 1, is ob-
tained from the expressions given above, and is a function of x
only, which can be evaluated from the ratio of the slope of the
plot of dh/dt versus l / h to the square root of the intercept,
which can be seen from Equation 7 to be equal t o
$d&
EXPERIMENTAL: SINGLEUNIFORMCAPILLARY
Experimental confirmation of the expression given above
for uniform capillaries is most readily obtained by considering
a somewhat more general case; that of a uniform capillary
inclined a t an angle a to the horizontal, and having a length
L1dipping below the free surface of the liquid. Paraphrasing
the argument used to derive Equation 4, the expression for
this more general case can be shown to be
- _- 2Ry + R811L
dL
dt
ZLldg sin CY R 2dg sin CY
811
(9)
where L is the length of tube penetrated.
dL 1
It follows that a plot of 817 against should be a straight
+
line of intercept -R2 dg sin a and of slope 2Ry R2Lldgsin a.
Figure 2, curve 1, shows such a plot of the data for dis-
tilled water penetrating a capillary inclined a t an angle of
9'22' to the horizontal. Previous to the test, the capillary
had been wet by forcing the liquid back and forth several
times. Curve 2 of the same figure shows the data for water
in the same capillary when flowing back toward the equi-
librium position from a greater height. Calculation on the
basis of Equation 9 gives the results shown in Table I.
These values of R are in excellent agreement with the value
R = 0.0250 obtained by calibration with a mercury thread.
The value 72.4 dynes per em. for y agrees well with the value
of the surface tension of water a t 20' C., 72.75 dynes per cm.
(4). The value obtained with the advancing meniscus, 70.1
dynes per cm., is significantly lower than the other, but
88 ANALYTICAL EDITIOK Vol. 6, iVo. 2

agrees closely with a number of similar determinations made suspended over the liquid in a closed system. The apparatus
by the authors, and with the value (70.2 dynes per cm.) employed in the tests reported below is shown in Figure 4.
obtained by Washburn (9) with an advancing meniscus The strip is fastened at both
(though at a slightly higher temperature). ends to hooks on a glass rod
which is suspended by a fine
TABLE I. WATER IN GLASSCAPILLARY AT 20' c. wire from a glass windlass by
CGRVE MENISCUS INTERCEPT SLOPE R Y m e a n s of which i t can be
Dynes/cm. Dynes Cm. Dyneu/cm. raised or lowered. An ordinary
1 Advancing -0,1011 3.786 0.0252 70.1 stopcock above the windlass
-
2 Receding -0.1006 3,888 0.0251 72.4
L I = 2.5 om. can be opened to evacuate the
a 9022'. system if desired. The bulb REEL

sealed to the side of the main

'The experiment just cited is of interest here in supporting tube permits the atmosphere
the theory discussed above. It has independent merit, in the latter to be dried or
however, in affording a simple method of determining surface brought t o a n y desired hu-
tension under dynamic rather than under static conditions. midity. With this arrange-
The computation of results is simplified by using a horizontal ment, the vapor in the tube can
capillary, so that a = 0, in which case Equation 9 can be be brought into equilibrium
integrated to give with the liquid, and the sample
lowered into contact with the
liquid without disturbing this
equilibrium.
Although in this case it is necessary to make a separate In Figure 5 are shown the
determination of the capillary radius, y can be computed results obtained on the penetra-
much more simply, as a direct plot of L2versus t gives a tion of liquids into a paper
straight line of slope R y / 2 q . In Figure 3 are plotted data for strip, as computed from data
the penetration of water into a dry and into a wet horizontal obtained as described above.
glass capillary of radius 0.03556 cm. a t 26' C. The value The corresponding values of
(70.3 dynes per cm.) obtained with the wet capillary agrees dh/dt and l / h give a good fit to
closely with the value (70,l dynes per cm.) cited above for a straight-line relationship, in
penetration into an agreement with the theory de-
6900
inclined capillary at veloped above. This agree-
5000 20" C. a n d w i t h ment was found with all liquids
5 4000 Washburn's value of examined-namely, benzene,
70.2 dynes per cm. ethyl benzene, carbon tetra- FIGURE 4. APPARATUSFOR
;3000 at 30" C. The ex- chloride, ethyl alcohol, methyl
tremely low value alcohol. and ethylene dichlo- sTUDYING pEmTRATIoN
0
2 zoo0 LIOUIDSINTO PAPER STRIPS
1000 (38.5 dynes per cm.) ride. In another article (5)the
found for penetra- method is shown to apply to studies of the penetration of
0
tion into a dry capil- molten waxes into paper strips.
VALUES O Ft INSECONDS
lary has been In Table I1 are listed the values of slope and intercept
FIGURE3. L2 VS. !i FOR WATER IN checked by a num-
HORIZONTAL GLASSCAPILLARY obtained in four runs made with each of the liquids tested.
ber of careful repeti- The approximate temperature of each run is noted in the
1. Capillary wet
2. Capillary dry tions of the experi- table. For each value of slope and intercept there have been
ment, and similar computed the corresponding values of Ay (or 417 times the
low values have been observed by others using different meth- slope) and of B (or 8r/dg times the intercept) as given by
ods, notably Haller (3). No discussion of this phenomenon Equation 7 . For each liquid the mean values of A y and of B
will be attempted here, but it is cited to illustrate the im- thus obtained are listed in Table 111, together with the
portance in penetration studies of determining interfacial standard (or root mean square) deviations of the individual
forces under dynamic conditions comparable with those ob- values from their mean, as a measure of the reproducibility of
taining in the type of penetration under consideration. the results. The standard deviations of B are given in the
same units as the values of B (sq. cm.), while the standard
EXPERIMENTAL : P A P ~ STRIPS
R deviations of Ay are expressed as percentages of the corre-
In studying the rate of penetration into paper, strips of sponding values of A y .
filter paper about 1.0 cm. wide and about 20 cm. long were I n Table I11 are included the grand mean value of B and the
employed. These were marked lightly with pencil at each mean values of the standard deviations of Ay and of B.
half-centimeter along the length. I n each test a strip was Taking each of these latter quantities as an estimate of the
suspended vertically with its lower end dipping into the liquid. expected value of u in samples of size four, the standard
The liquid could be seen to rise in the strips, the upper borders deviation (a') of a single determination will be given by the
of the wetted portion usually being even and horizontal and relation u = 0.798 u' (8, p. 185). The value of u' for By is
always clearly demarcated from the unwetted portion. The thus found to be 3.8 per cent (corresponding to a probable
time at which the rising liquid reached each half-centimeter error of about 2.5 per cent for a single determination, or of
marking was read with a stop watch and recorded. Having about 1.3 per cent for the mean of four determinations).
thus determined corresponding values of height of rise, h, and Variability of this magnitude may be ascribed wholly to the
time, t, values of the rate of rise, dh/dt, were obtained by experimental error of the method.
graphical differentiation, and plotted against corresponding As, for such fine capillaries, the hydrostatic head has but a
values of l / h . small effect on the rate during the earlier stage of penetration,
To reduce evaporation as much as possible, the strip was the intercept which measures this effect is so small that a
March 15, 1934 INDUSTRIAL AND ENGINEERING CHEMISTRY 89

TABLE11. VALUESOF SLOPEAND INTERCEPT cross sections of the paper used in these testsl which were
LrQUID TEMP. SLOPE INTERCEPT prepared by Miss A. K. Marshall, of these laboratories.'
C. Sq. cm./sec. Cm./sec. In Figure 6, A and B were taken from nearby points in a
Benzene 29.8 0.081 0.0030
29.8 0.082 0.0024 single section, and illustrate the considerable variation in
30.6
31.4
0.089
0.085
0.0032
0.0025
density apparent in the paper. It is evident from these
Carbon tetrachloride 24.5 0.048 0.0022 pictures that any attempt to represent the channels between
24.2 0.047 0.0026 fibers by a system of circular capillaries must be necessarily
25.3 0.052 0.0028
24.8 0.048 0.0024 highly approximate. Comparison of the spaces between
Ethyl alcohol 36.5 0.036 0.0005 fibers with the scale shown in the figures, which has a total
25.9 0.035 0.0005
24.7 0.032 0.0003 length (when reduced by the magnification factor) of 0.05
25.0 0.033 0.0006
Ethyl benzene 30.9 0.080 0.0018 cm., shows that the mean diameter of the largest of these
32.2 0.086 0.0021 spaces is of the order of.O.005 cm. This is in fairly good agree-
31.7 0.072 0.0004
27.9 0.076 0.0004 ment with the maximum mean diameter of 0.008 cm. calcu-
Ethylene dichloride 27.9 0.070 0.0024 lated as described above.
28.1 0.074 0.0022
27.8 0.076 0.0023
27.8 0.068 0.0019 TABLEIV. RELATIVE
VALUESFOR y
Methyl alcohol 25.5 0.070 0.0020
26.0 0.066 0.0017 AVERAGE SLQ
26.3 0.072 0.0020 LIQUID TEMP. (REL~TIVE) S L G (RELATIVB)
26.8 0.067 0.0009 a c. Dynes/cm.
Benzene 30 1.000 27.6 1.000
Carbon, tetra- 25 0.942 26.1 0.946
chloride
TABLE111. MEAN V.4LUES OF A? AND B Ethyl alcohol 26 0.779 21.8 0.790
LIQVID A'Y 0 B 0
Ethyl benzene 31 0.983 28.0 1.014
Ethylene dichloride 28 1.133 31.2 1.130
Dmes % Sq. cm. sq. cm. Methvl alcohol 26 0.794 22.1 0,800
Benzene 0.00188 3.3 13.05 X 10-8 1.58 X
Carbon 0.00177 3.3 11.70 X 10-8 1.03 X lo-*
tetrachloride Having thus subjected Equation 7 to such comparison with
Ethyl 0.00146 3.3 5 . 6 0 X 10-8 1.35 X
alcohol experimental results as is possible, it may be tentatively
Ethyl 0.00184 6.2 6.55 X 4.35 X
benzene assumed to apply, and relative values of y may be computed
Ethylene 0.00212 3.0 10.80 X 10-8 0.90 X for the liquids tested from the values of A y listed in Table
dichloride
Methyl 0.00149 3.8 9 . 3 0 X 10-8 2 . 5 8 X 10-8 111. In Table IV are given such relative values for y, taking
alcohol
Mean values 3.82 9 . 5 0 X 10-8 1.97 X 10-8 that for benzene as 1.00. In the same table are listed for
comparison values
(at t h e a v e r a g e 0032
quite small experimental error will serve to explain the large
percentage errors in B listed in Table 111. The quantity B temperature of test) 0020

of SLG,or the values

is of interest principally in affording a check on the theory of the surface ten-
Y 0024

given above, according to which it should depend only on the sion of these liquids, 3 0020

pore distribution in the paper strips, and be independent of

the liquid tested. Hence the values of B obtained with the
as given in the In- L
ternational Critical :
different liquids, as listed in Table 111, should show no Tables. I n a third 2 O0"
greater variation than the random one associated with varia- column are given 0000 f

bility of the order observed in individual determinations with values of SLG,rela- 0004
a single liquid. This condition requires that each individual tive t o benzene 0
determination differ from the grand mean value (9.5 X (1.00). T h e o b - 001 012 OM 000 032 036 0 4 0
020 024 028

sq. cm.) by less than 3 u', where u' is the standard deviation served a n d com-
VALUES Or IN CU-'

of individual determinations, which has the value 2.5 X
sq. cm., when computed as described above. Also this
condition requires that the mean value for each liquid shall
differ from the grand mean value by less than 3 u t / G
(8, p: 309). Except for the mean value of ethyl alcohol, which
lies just outside the limits, all these conditions are satisfied.
It seems reasonable to conclude that these data support the
theory in showing B to be independent of the liquid tested.
A complete experimental check of Equation 7 is impossible,
lacking a method for independently determining the distribu-
tion of pore sizes. A rough qualitative check, however, can
be made by computing from the experimental data the range
of pore sizes that would correspond to a uniform pore distribu-
tion. For this computation B can be taken as having the
grand mean value of Table I11 (9.5 x low8sq. cm.) and A as
given by the mean value of A y for benzene, assuming (for
reasons discussed below) that for this liquid y is equal to its
surface tension (27.6 dynes per cm.), so that A = 68.1 X lom6
cm. From these values i t follows that A/ & = 0.221,
to which corresponds the value 0.10 for x in Figure 1. This
permits A I R to be evaluated, giving R = 0.004 cm. and XR =
0.0004 cm. Hence, on the assumption of a uniform distribu-
tion, the range of pore sizes would be from 0.0004 to 0.004 cm.
That the actual pore sizes are of this order of magnitude can
be seen from the photomicrographs shown in Figure 6 of thin
puted values of rela- FIGURE5. dh/dt vs. l/h FOR SEVERAL
tive surface tension LIQUIDSI N FILTERPAPER
agree well w i t h i n
the estimated precision of A y . As it would be very unlikely
that these liquids would all have the same contact angle
unless the latter were zero, it was concluded that for all
these liquids the contact angle against paper is either zero
or very small.
DISPLACEMENT
TESTS
By putting two immiscible liquids in a single vessel with a
strip of paper projecting through the surface of separation,
one liquid can be made to displace the other, and the rate of
penetration of the displacing liquid can be observed and
measured. The theory for this case can be worked out on
lines analogous to the treatment given above of simple pene-
tration, the effective force at the interface being the difference
in adhesion tensions ( 7 ' ) given by Equation 3. Experi-
mentally it was found easy to observe and time such displace-
1 A strip of paper was embedded in a block of paraffin and sectioned with
a microtome in a plane normal to the length of the strip. The section was
fastened to a slide, the paraffin dissolved out, the fibers stained with eosin,
and photographs taken at a magnification of 122. It should be noted that
the direction in which the microtome moved in making the cut was at right
angles to the width of the strip, and it is therefore evident from the pictures
that the fibers were cut clean and not dragged out.
 90 ANALYTICAL EDITION Vol. 6, No. 2

ment in the cases of water and benzene, and water and carbon
tetrachloride. Unfortunately, however, the results were not
consistent or reproducible, presumably because of the unusual
sensitivity of water surfaces to the presence of foreign ma-
terials. As there are few pairs of immiscible liquids of prac-
tical interest that do not include water as one of the liquids,
the study of this displacement test was not pursued further.
It is possible, however, that in porous media other than paper
have been directed entirely to checking the validity of the
theory presented. Such a check cannot be complete, and
must be supported by the mutual consistency of results ob-
tained in further work. Assuming the theory to be suffi-
ciently accurate to justify its use in interpreting data, it
permits the determination of relative values of penetration
tension in paper and further studies may show it to be ap-
plicable to textiles and thin wood sections. The experi-
mental procedure reauires merelv the
measurement of the rate of rise of the

held i n a ;esse1 closed t o i r e v e n t

evaporation. By plotting the observed
liquid height, h, against time, t, the rate
of rise, dhldt, may be evaluated Eraphi-

portional to y. q (the ratio of penetra-

. ,’ tion tension to viscosity), the slope be-
ing a function of.thc distribution of pore
? sizes in the medium.
Yalues of the viicosity of the liquid
0 0.01 0.02 0.03 0.04 0.05
niultiplied by the slope of the plot of
FIGURE6. PIIOTOWCROGRIPIIS OF PAPEH USED is TESTS dh, dt versus h give, therefore, relative
values of penetration teiision for liquids
results would be olnained of sufficient constancy to make the tested with a given mediuin. Comparison of the values of
displacement test a useful method of determiniug adhesion penetration tension for a single liquid and several different
tension. The theory for such cases may be readily developed mydia can be made only when it i.5 ki1on.n that the media are
along the lines indicated. identical in pore structure or in the distribution of pore sizes,
or when tests can be macle with a reference liquid which is
SIX\l.\RY known to have a zero contact angle - with all the media to be
The theory presented above for capillary penetration into tested.
fibrous materials is an extension of that developed by Wash- While the method of analysis given applies primarily to
burn for uniform capillaries. It has been shown that this fibrous materials in which the rate of rise may be observed
simpleracase may be conveniently studied in terms of the directly, it may be used to interpret data on the rate of pene-
relation between rate of penetration and the reciprocal of tration into porous systems for which only the rate of weight
the distance penetrated, and that in this way both the radius increase by penetration is known, provided the mean cross-
of the capillary and the penetration tension may be computed sectional area of pore space is known, so that the linear rate of
from the test data. The method is of value in that the penetration can be calculated.
measurement is made under dynamic conditions. LITERATURE CITED
It has been shown that the treatment for penetration into a (1) Bartell, F. E., and Osterhof, H. J., IND.ENQ.CHEM.,19, 1277
porous medium is similar to that for a uniform capillary, save (1927).
that consideration must be given to the fact that the variation (2) Bosanquet, C. H., Phil. Mag., [6], 45, 525 (1921).
in pore size causes the resistance to viscous flow of each (3) Haller, W., Kolloid-Z., 54, 9 (1931).
channel to be similar to that of a number of capillaries in (4) Hodgman-Lange, “Handbook of Chemistry and Physics,” 16th
ed., Chemical Rubber Publishing Co., 1931.
series. The final expression therefore involves quantities (5) McLean, D. A , , and Kohman, G. T., paper to be published in
whose values depend on the distribution of pore sizes. The EZectricaZ Engineering.
general character of these quantities has been indicated by (6) McMillan, E. L., IND. ENG.CHEM.,21, 1237 (1929).
evaluating them for the case of a uniform distribution of pore (7) Osterhof, H. J., and Bartell, F. E., J . Phys. Chem., 34, 1399
(1930).
sizes. (8) Shewhart, W. A., “Economic Control of Quality of Manufactured
Data have been presented on the penetration of six organic Product,” Van Nostrand, 1931.
fluids into strips of filter paper. It has been shown that these (9) Washburn, E. W., Phys. Rev., 17, 273 (1921).
data agree with the theory in character. These experiments RECEIVED September 2, 1933.

FIRE-HAZARD TESTSWITH CIGARETS.Tests have been con-
ducted in the fire-resistance section of the Bureau of Standards
t o determine the fire hazard of discarded lighted cigarets. The
efficacy of certain modifications, such as slow-burning paper and
the application of tips over one end, was also investigated. The
burning cigarets were placed on representative specimens of
grass and forest floor materials.
Under the conditions of the test, with the most favorable
drafts and with relative humidities in the range 25 to 50 per
cent, fires were caused on the average by 9 out of 10 lighted half-
length untipped fast-burning cigarets discarded on grass, forest
litter, or duff. The percentage of cases resulting in fires in-
creased somewhat with decrease in relative humidity. The fire
hazard of the slow-burning cigaret was much lower than for the
fast-burning type. In the former the glow will not progress ap-
preciably after the cigaret is discarded, while the latter will
continue t o glow until fully consumed.
The fire hazard of discarded lighted cigarets can be decreased
by applying tips of cigaret paper. In tests with half-length fast-
burning cigarets having tips 1 inch long of paper similar to that
used on this type of cigaret, 4 fires occurred on the average for
every 10 trials. With tips of the same length made of the paper
used on slow-burning cigarets, the occurrence of fire in the ex-
posed materials was reduced t o 1 out of 4 trials.