Capillary Conduction of Liquids Through Porous Mediums: NOVEMBER, 1931 Physics
Capillary Conduction of Liquids Through Porous Mediums: NOVEMBER, 1931 Physics
Capillary Conduction of Liquids Through Porous Mediums: NOVEMBER, 1931 Physics
CORNELL UNIVERSITY
INTRODUCTION
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
CAPILLARY CONDUCTION OF LIQUIDS 319
pressure. A study of this phase of the problem indicates that the capillary
conduction of liquids through porous mediums is similar to the flow of liquids
through pipes or fine tubes to the extent that in both cases the flow may be
expressed in terms of gravity and the pressure gradient in the liquid. There
is, however, a significant difference between the two classes of flow phe-
nomena and we are quite justified in continuing to use the term capillarity
when dealing with liquid systems where the pressure distribution is deter-
mined by the surface tension and curvature of a gas-liquid interface.
Among the expressions which have been derived for capillary flow, those
given by Gardner4 ,5 are the most noteworthy but as yet a general theory in
terms of known or measurable functions has not been published. The prob-
lem is complicated by a large number of variables and the complex con-
figuration of the region occupied by the capillary liquid has been a stumbling
block for theoretical analysis. Also, the lack of suitable apparatus has pre-
vented accurate experimental measurements. In the present work some of
the experimental difficulties have been overcome and the problem has been
approached by methods which are closely analogous, both in theory and ex-
periment, to those used in the study of thermal and electrical conduction
in metals.
FORCES AFFECTING CAPILLARY ACTION
The forces acting in the boundary surfaces of liquids are directly re-
sponsible for all capillary phenomena and have their origin in the cohesive
and adhesive attractions which are exerted between molecules. The most
common cases of capillary action occur when one surface of a liquid is in
contact with air. Because of the free energy possessed by the molecules of
the liquid lying in such a surface there is a tendency for the surface to as-
sume that configuration which makes the area a minimum. Mathematically
a uniform surface density of free energy is equivalent to a uniform tension in
the surface and for some purposes it is more convenient to express the physi-
cal properties of an air-liquid surface in terms of surface tension.
For instance, the pressure difference caused by a curved liquid surface is
usually expressed by the relation,
(1)
pw and PA being the pressures on the two sides of the surface, Tl the surface
tension, and (1/Rl + 1/R 2 ) the total curvature of the surface.
Because of the inequality between cohesive and adhesive forces at a
liquid-solid interface, these surfaces also have a free energy which is impor-
tant in determining the capillary behavior of liquids. If, for a given liquid
and solid the surface densities of free energy for the air-liquid, air-solid and
liquid-solid interfaces are respectively T 1 , T 2 , and T 12 , it may be shown 6
that,
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
320 L. A. RICHARDS
(2)
W being the work necessary to separate unit area of liquid from solid and ()
the angle of contact between liquid and solid. If W < 2T1 , () may vary any-
where between 0° and 180°. If, however, W~2Tl then () is zero and the
liquid will spread over the surface of the solid and is said to wet the solid.
Because of their effect on the angle of contact, adhesive forces are di-
rectly involved in an initial wetting process such as the spreading of a liquid
in a dry porous medium, but after the medium is wetted adhesive forces are
no longer effective in producing a motion of the liquid and influence capillary
action only to the extent that they hold a thin film firmly in contact with the
solid surface. The liquid lying outside the adsorbed films is free to move
under the action of unbalanced forces. Capillary flow may be expressed in
terms of the well-known laws of hydrodynamics and is distinguished from
other cases of liquid flow only because of the relation of surface tension and
curvature to the pressure and to the effective cross-sectional area of the
liquid-transmitting region.
The equation of motion for a viscous fluid is given in various treatises on
hydrodynamics 7 ,8 as,
pdv/dt = pF - 'i/P + fJ.('i/'i/·v/3 + 'i/''i/v), (3)
where p is the mass of unit volume of the liquid, dv/dt is the acceleration and
the right hand member is the expression for the forces acting per unit volume.
Dividing Eq. (3) by p we obtain the relation between the acceleration and
forces for unit mass,
dv/dt = F - 'i/p/p + (fJ./p) ('i/'i/. v/3 + 'i/·'i/v). (4)
F represents the external or body forces and for most capillary problems
gravity is the only external force which need be considered. The term - V p/p
is the expression for the force due to the pressure gradient and the third
term, being a function of the viscosity p" and the velocity v, is the expression
for the viscous retarding forces.
If the gravity force F is represented as the negative gradient of the po-
tential eI> = gz, where z is the height from a reference level, then F= - Vel> = - g.
The variation of the gravity force with z is neglected, g being assumed con-
stant. Also, if p is constant or depends on p alone, the force represented by
the term - V p/p may be expressed as the gradient of the potential y; = jdp/p,
the integral being taken from an arbitrarily chosen reference pressure to the
pressure at the point in question. Eq. (4) may then be written,
dv/dt = - 'i/(ep + 1/1) + (fJ./p)('i/'i/·v/3 + 'i/''i/v), (5)
or,
dv/dt =- 'i/<I> + (fJ./p)('i/'i/·v/3 + 'i/''i/v), (6)
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
CAPILLARY CONDUCTION OF LIQUIDS 321
where <I> = (¢ +1/;) and - \7<1> is the total water-moving field or the total field
tending to produce a motion of the water.
Expressions for the flow of viscous liquids through various channels have
been derived from the general equations of motion. The equation of Poi-
seuille derived for flow through small tubes is in excellent agreement with
experiment. Applications of the theory have been made by Slichter 9 for the
flow of water through saturated porous mediums but the results are less
definite in this case, because, eVen for the simplified idealized system made
~p of uniform spheres with regular packing, the complicated configuration
of the water transmitting region makes necessary the introduction of various
simplifying assumptions.
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
322 L. A. RICHARDS
In order for the capillary water to attain the correct pressure for equilibrium
at a given height, the moisture content· of the medium changes until the
curvature has the right value.
If the density of water is assumed to be constant and equal to unity, the
potential t/;=jdp/p becomes numerically equal to the pressure. t/; is a poten-
tial due to pressure forces and when the pressure is determined by the surface
tension and curvature of the air-liquid interface t/; will be called the capillary
potential. For a given value of the pressure or capillary potential the amount
of moisture present in a medium will depend on the number and kind of pore
spaces. The curves in Fig. 1 are reproduced from an earlier paper by the
Ai 51 V
V
V
-"
.":1
t
I Ii.
6xl0 5
.
Fig. 1. Curves showing the relation between capillary potential and moisture content
for, A-Bennet sandy soil, B-Greenville loam soil and C-Preston clay.
writer 3 and show the relation between moisture content and capillary po-
tential for three mediums. To obtain these data a hollow capillary cell having
a flat porous surface was filled with water at a controlled pressure. A layer
of the medium to be studied was then spread on the porous surface and the
moisture in the medium was allowed to come to pressure equilibrium with
the water in the cell. Samples of the medium were removed at each pressure
to determine the moisture content. The moisture content is given as the
number of cubic centimeters of water per gram of dry medium and will be
designated by O. The capillary potential is expressed in ergs per gram and
is numerically equal to the pressure expressed in dynes per square centimeter.
When the conditions for equilibrium under gravity, as expressed by Eq.
(7), are fulfilled, the velocity and acceleration of the capillary liquid are
everywhere zero and Eq. (5) becomes,
v(cp + 1/;) = 0 (8)
which means that the force arising from the pressure gradient just balances
gravity. If this condition does not obtain there will be a resultant water-
moving force and in general there will be capillary flow.
Because of the complex configuration of the capillary liquid it would be
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
CAPILLARY CONDUCTION OF LIQUIDS 323
difficult to derive expressions for capillary flow from the general hydro-
dynamical equations but it is possible to deduce generalizations from ex-
perimental data which enable us to set up mathematical relations between
the flow and factors causing the flow. Fourier's law and Ohm's law are just
such generalizations. An experimental law, quite analogous to these two, and
attributed to Darcey, may be used in connection with the present problem.
Darcey, working with mediums under saturated conditions, found that
the flow of water through a column of soil is directly proportional to the
pressure difference and inversely proportional to the length of the column.
For low pressure gradients it has been found by numerous investigators10,1l
that this law is in exact agreement with experiment and it is entirely analog-
ous to the well-known law of Poiseuille for the flow of liquids through capil-
lary tubes. However, both of these laws fail to hold for high pressure gradi-
ents. The limits within which they are true and the modifications which a
second approximation requires can be determined only by exhaustive ex-
periments on a wide range of materials. In view of the experimental data now
available it is assumed that Darcey's law holds for the low velocities and
pressure gradients dealt with in this paper.
Mathematically Darcey's law may be expressed by the equation
q = - K'lP, (9)
where q is the volume of water crossing unit area perpendicular to the flow,
in unit time and K is a proportionality factor which for a medium whose
pore spaces are filled with water will depend on the number and kind of pore
spaces and the viscosity. This factor has been variously known as the trans-
mission constant,9 the coefficient of permeability,t° and the conductivity.l
In C.g,s. units it is the number of cubic centimeters of water crossing one
square centimeter perpendicular to the flow in one second when there is a
water-moving force of one dyne per gram. From analogy with the thermal
and electrical cases it will here be called conductivity.
If there is a steady flow of liquid through a porous medium which is
only partially saturated, then the larger pore spaces contain air and the
effective cross-sectional area of the water conducting region is reduced. If
these air spaces could in some way be filled with solid, the condition of the
flow would be unchanged and the proportionality between the flow and the
water-moving force would still hold because Darcey's law is independent of
the size of particles or the state of packing. Hence the essential difference be-
tween flow through a porous medium which is saturated and flow through a
medium which is unsaturated is that under this latter condition the pressure
is determined by capillary forces and the conductivity depends on the mois-
ture content of the medium. The motion of liquids through unsaturated
porous mediums will be referred to as capillary flow and for this case K in
Eq. (9) will be called capillary conductivity.
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
324 L. A. RICHARDS
where \l.q is the divergence of the flow, Ps is the weight of the dry medium
in unit volume and ae/at is the rate at which the moisture content is changing
with the time. Now if e is a single valued continuous function of 1/;, ae/at
= (de/d1/;) (aNat) =AaNat and Eq. (10) becomes,
\l.q = - P8Aaif;jat. (11)
A is the rate of change of the moisture content with respect to the capillary
potential and will be called the capillary capacity of the medium. The func-
tional relation of A to 1/; or e may be determined from curves such as those
shown in Fig. 1.
Putting Eq. (9) in cartesian coordinates we have,
where i, j, and k are unit vectors along the x, y, and z axes respectively. If
the z-axis is chosen positive upward along the vertical then a¢/ax =a¢/ay = 0,
a¢/az = g, and
q = - K[ia1/;/ax + jaif;jay + k(g + a1/;/az)]. (13)
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
CAPILLARY CONDUCTION OF LIQUIDS 325
of the soil has been used as a variable in expressing capillary flow 5 ,12, but,
for uniform packing this variable is proportional to what is here called the
moisture content.
In order to make use of the above equations it is necessary to have in-
formation concerning the functions ()(t/t), and K(t/t). There is already avail-
able in the literature1 ,3,12,13,14,15 some data showing the relation between ()
and t/t for soils and it is possible to determine, for a given porous medium,
the way in which the capillary conductivity depends upon the capillary
potential.
EXPERIMENTS WITH CAPILLARY CONDUCTION
If the pressure gradient and capillary flow are known for a given column
of porous medium, the value of the conductivity may be calculated by sub-
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
326 L. A. RICHARDS
which have a flat porous wall of uniform thickness in contact with the column
of soil. To insure good capillary contact these cells were held against the
soil column with strong rubber bands. The cells are filled with water and
connected as shown to the vacuum bottle C whose pressure is automatically
controlled. It is the value of this pressure which determines the moisture
content of the column. The pressure in the upper cell is greater than in the
lower cell due to the weight of the hydrostatic column from D to E. With
the arrangement shown there is a pressure and a gravity force tending to
make the water move from the upper cell through the column to the lower
cell. The water flowing in and out of the column is measured by the burettes
F and G. By using pinch clamps at H and I the upper burette can be filled
and the lower one emptied without disturbing the cell water pressure. The
pressure difference at the two ends of the column is measured by means of
four small porous clay cups J extending across the ends of the column. The
cups are filled with water and connected to two mercurial manometers, K,
only one of which is shown in the figure. These cups, which serve as capillary
potentiometers, are 5 mm. in diameter and have a wall thickness of about
0.5 mm. Since the pressure drop along lines of flow passing through the cups
is small, it is assumed these potentiometers indicate the pressure difference
between the two ends of the soil column. If half of this difference is subtracted
from the pressure difference indicated by manometer L, the average pressure
in the water films of the soil column is obtained. Air in the column was at at-
mospheric pressure and evaporation from the cells and soil was prevented by
enclosing them in a paraffined insulite box, the air in which was kept satu-
rated by cloth wicks.
The porous clay cells, having the desired shapes and properties, were
made up in the laboratory by the ordinary plaster of Paris casting process.16
The hollow rectangular cells were cast in two pieces and cemented together
with slip when partly dried. The clay used, when fired at 1000°C was readily
permeable to water and a wall of the material 0.5 mm thick, when wet, will
stand a pressure difference of one atmosphere without leaking air.
The liquid passing through a column was used over and over so that
dissolved substances would not continue to change its surface tension and
viscosity.
Since the conductivity was to be measured at a constant film thickness
or constant curvature it was necessary to maintain a constant difference
between the capillary water pressure and atmospheric pressure. Special
apparatus17 was divised for this purpose which would automatically maintain
the pressure in flask C to within 0.5 mm of mercury of the desired value.
Room temperature was controlled by means of a mercury ether thermo-
regulator18 and temperature fluctuations as read by a sensitive thermo-
meter were less than 0.1°C. It was found necessary for some of the runs to
store the ingoing liquid in a flask, M, at reduced pressure so as to prevent
16 C. F. Binns, The Potter's Craft, New York (1922).
17 L. A. Richards, Rev. Sci. Inst. 2, 50 (1931).
18 L. A. Richards, l.O.S.A. and R.S.l. 18, 131 (1929).
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
CAPILLARY CONDUCTION OF LIQUIDS 327
dissolved air from coming out of solution under the low pressure in the
burettes and cells. The connecting tubes were arranged so that any air col-
lecting in the cells would expell itself. Four of the units shown in Fig. 2 were
constructed.
When the rate of flow into a column becomes steady and equal to the
flow out, a value of the conductivity may be calculated. If Q/t is the inflow
in cc per sec., a, the area of the soil column, L its length and !1p the pressure
difference, then, multiplying Eq. (9) by the column area to obtain the total
flow gives,
Qjt = aq = - aK\iip. (16)
the negative signs of the right hand member indicating that the force acts in
the negative z direction. Substituting (17) in (16) and solving for K gives
the curves with the corresponding letters in Fig. 1 were obtained. The A
mediums in Figs. 1 and 3 are not the same but seem to have a similar me-
chanical composition.
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
328 L. A. RICHARDS
In order to follow closely the changes in the rate of flow, readings of the
burettes were taken several times a day and except in a few cases when
difficulties with the apparatus developed the time intervals between the dates
given in Table I indicate the time required for the flow to become steady.
As was stated above, K, which is the flow across unit area in unit time
for unit potential gradient, depends on the mechanical composition and state
of packing of the medium and on the viscosity and pressure in the water
films. For a given determination of the conductivity all of these factors were
B'
-~-------
B
~
1-----1-----1-----+-'-'''-7J~_+_1100 if-
>
-'-'
()
J
-0
c
o
t)
1-------I------~~~----tr·----__410 ~
ro
0-
ro
()
4)\10 3 2 I
Capillary potential, \jJ (er$s/sram)
Fig. 3. Capillary conductivity curves for, A-a sandy soil,
B-Greenville loam soil and C-Preston clay.
held constant except the state of packing, which, under the compressive stress
of the rubber bands, continued to become more dense with time. The effect
of this change is indicated in the data.
When the apparatus was first set up the data shown in Fig. 3 as the dotted
curves and designated by the primed letters was taken. The apparatus then
stood for five months before obtaining the data shown by the full curves. In
each case there is a shifting of the curve toward lower values of the con-
ductivity.
ApPLICATIONS OF THEORY AND DATA
The curves for the moisture content and conductivity shown in Figs. 1
and 3 have had all the factors except the capillary potential held constant
and, before any general application of the flow equations can be made it
would be necessary to know something about the relations among 1/;, e and
K for different temperatures and different states of packing. However, as an
example to illustrate the usefulness of such information, the data given here
will be employed in the solution of a typical flow problem.
Let it be required to find for Greenville soil the moisture distribution and
steady state upward flow from a water table one meter below the surface.
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
CAPILLARY CONDUCTION OF LIQUIDS 329
Substituting the values of K and dfjdz given by Eqs. (20) and (22) gives
q = (aif; (25)
which reduces to
q = -l/acl = 11(1.66 X 10 19 X 5.35 X 10- 15)= 1. 13XlO-5 cc sec.- 1 cm- 2 (26)
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
330 L. A. RICHARDS
'"
"" ~~ '\ \.
1\
\
\
80-4--~--+-+-~-+--~
60-4---+~-+--~--4-~
Z
40-+--+--\-+---+---'<--+-----1
~ A
6.x.10· 5 4 :; '2 I
~o 0 0.12 0.18 Q24 0.30 0.36 0.42
Capillary potential, IjI (ergs/gram) MoisGvre content, e ecc/gra.m)
Fig. 4. Curves showing the distribution of capillary potential and moisture content with
height above a water table 100 cm below the surface of Greenville soil. A and B are for upward
capillary flow while A I and B' are for static equilibrium.
The curves A' and B ' , Fig. 4, give the distribution of pressure and mois-
ture with height when there is no surface loss. That is, when the capillary
liquid is at rest under gravity and the flow is zero.
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
CAPILLARY CONDUCTION OF LIQUIDS 331
capiIIaries and because the cellular pore spaces between soil particles have
various sized openings, it is difficult to predict whether all porous mediums
made up of particles will show a hysteresis effect.
The capiIIary-potential vs. moisture content data shown in Fig. 1 were
obtained by having a one half inch layer of soil in contact through a porous
plate with water which is at controJled pressure. This method is open to the
objections that it is impossible to tell when the soil water is at pressure
equilibrium with the water in the cell and also each point on the curve is
taken for a different mass of soil. To overcome these undesirable features a
fired clay cell like those used in the transmission experiments was suspended
on a balance and provided with a lid for covering the upper surface which was
porous. Connections to the ceJl were made by means of a glass tube three
meters long and heavy rubber tubing in such a way that when the cell con-
tains water any air accumulating inside will expel itself. The weight of the
cell system was made practically independent of changes in the water pres-
sure and a sensitivity of 5 mg in the weighings was attained. When ready for
an experiment, a known weight of dry soil is spread on the plate and the
amount of moisture in the soil at any later time may be determined from the
balance reading.
TABLE II.
The only data yet available with this apparatus were taken for Green-
ville soil and are presented in Table II and Fig. S. The cell was filled with soil
solution so that transfers of moisture between soil and plate would not cause
changes in surface tension and the temperature was controlled for the ex-
periment. Balance readings were taken at frequent intervals after a change in
pressure and except for 9 and 13, the points shown represent equilibrium
values of the moisture content for the corresponding pressures. It is apparent
for a 4 mm layer of soil in a rather loose state of packing there is a consider-
able hysteresis effect over the pressure range studied.
These results, however, are not conclusive because under the conditions
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
332 L. A. RICHARDS
T<V
L
~ <f)
.....tr.J
SL 21-------I~ "-H'-------l--------1
....,
s::
C)
..w
8.-
~31------·1---~'---------4---------~
L
<D
a...
«j
U
5,
I
,,
I
I
,
5x I 0 5 L - - - - L . - - L . - L - - - - - - l - - - - - - . . l
Fig. 5. Curves showing the variation in the moisture content of Greenville
soil for cyclic variations in the capillary potential.
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe
CAPILLARY CONDUCTION OF LIQUIDS 333
soil or clay if the pore structure were fixed. As yet no data are available from
this experiment.
Another source of information on the hysteresis effect lies in the con-
ductivity experiments. If curves like those shown in Fig. 3 retrace them-
selves as the potential is increased to zero, it would be good evidence that no
hysteresis effect exists. vVith but one exception the curves shown were ob-
tained by starting with the porous medium wet and proceeding to lower
pressures and moisture contents. For curve C, as indicated by the dates in
Table I, the data for the second point on the curve were taken after the low
conductivity determination was made. The consistant relation between the
dotted and full curves for the three mediums would seem to indicate that if a
hysteresis effect exists for the Preston clay, it does not have a very large effect
on the relation between the capillary potential and the conductivity.
The author is indebted to Professor C. C. Murdock for helpful criticism
given during the preparation of this paper.
ed 06 Jul 2011 to 155.69.4.4. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_pe