~6fJ
A Review of Diffusion and Dispersion in Porous Media
T. K. PERKINS
O. C. JOHNSTON
MEMBERS AIME
ABSTRACT
Because of the influence of dispersion on miscible-
displacement processes, diffusion and dispersion
phenomena in parous rocks are of current interest
in the oil industry. This paper reviews and sum-
marizes a great deal of pertinent information from
the literature.
Porous media (both unconsolidated packs and
consolidated rocks) can be visualized as a network
of flow chambers, having random size and flow
conductivity, connected together by openings of
smaller size. In such a porous medium, the appar-
ent diffusion coefficient D is less than the molecular
diffusion coefficient Do, as measured in the absence
of a porous medium. For packs of unconsolidated
granular material the ratio DIDo is about 0.6 to 0.7.
For all porous rocks, both cemented and unconsoli-
dated, the ratio of diffusion coefficients can also
be represented as DIDo = _1_, where F is the for-
Fcp
mation electrical resistivity factor and cp is the
porosity.
If fluids are flowing through the porous medium,
dispersion may be greater than that due to diffusion
alone. At moderate flow rates the porous medium
will create a slightly asymmetrical mix zone (trail-
ing edge stretched out), with the longitudinal dis-
persion coefficient approximately proportional to
the first power of average fluid velocity (if compo-
sition is nearly equalized in pore spaces by diffu-
sion). If the velocity in interstices is large enough,
there will be insufficient time for diffusion to
equalize concentration within pore spaces. In this
region, longitudinal dispersion increases more
rapidly than fluid velocity.
At low velocities in interstices, transverse
dispersion is characterized by a region in which
transverse diffusion dominates. If the fluid velocity
gets high enough, there will be a transition into a
region where there is stream splitting with mass
transfer but with insufficient residence time to
completely damp-out concentration variations within
pore spaces.
There are several variables that must be con-
Original manuscript received in Society of Petroleum Engi-
neers office Sept. 10, 1962. Revised manuscript received Jan.
15, 1963.
THE ATLANTIC REFINING CO.
DALLAS, TEX.
trolled to get consistent longitudinal and transverse
dispersion results, viz., (1) edge effect in packed
tubes, (2) particle size distribution, (3) particle
shape, (4) packing or permeability heterogeneities,
(5) viscosity ratios, (6) gravity forces, (7) amount
of turbulence, and (8) effect of an immobile phase.
INTRODUCTION
Diffusion and dispersion in porous rocks are of
current interest to the oil industry. This interest
arises because of the influence of dispersion on
miscible-displacement processes.
In a recovery process utilizing a zone of miscible
fluid, there is the possibility of losing miscibility
by dissipating the miscible fluid or by channeling
or "fingering" through the miscible zone. Diffusion
and dispersion are two of the mechanisms that may
lead to mixing and dissipation of the siug. On 'the
other hand, dispersion may tend to damp-out viscous
fingers which may be channeling through the
miscible slug.58 Hence, dispersion may be detri-
mental or beneficial (if it prevents fingering through
the miscible zone). Therefore, it is doubly important
that we understand these processes.
In this paper we review, summarize and interpret
a great deal of information from the literature. In
particular, we will briefly discuss molecular diffu-
sion in miscible fluids. Then we will discuss what
differences to expect for diffusion in a porous rock.
If there is movement of the fluid through the rock,
then there may be an additional mixing or "disper-
sion". Furthermore, the dispersion longitudinally
(in the direction of gross fluid movement) and
transverse to the direction of fluid movement will
not be equal. We will discuss both types of dis-
persion as well as several variables which can
affect dispersion (viscosity differences, density
differences, turbulence, heterogeneity of media,
etc.). This group of variables has sometimes led to
difficulty when comparing literature data.
DIFFUSION OF MISCIBLE FLUIDS
If two miscible fluids are in contact, with an
initially sharp interface, they will slowly diffuse
into one another. As time passes, the sharp inter-
face between the two fluids will become a diffuse
58References given at end of paper.
SOCIETY OF PETROLEUM ENGINEERS JOURNAL
mixed zone grading from one pure fluid to the other. probability co-ordinate paper as shown on Fig. 1.
This diffusion arises because of random motion of Taylor 52 has shown that the diffusion coefficient
the molecules. If there is no change in volume upon can be calculated from a plot of experimental data,
mixing the two fluids, then the net transport of one such as Fig. 1, by Eq. 3.
of the constituents across an~ arbitrary plane can
2
be represented by the Fick 1 diffusion equation,
shown as Eq. 1. 0. - -
o - f
I [X 3.625X 90 - 10 ]
.
. . . . . . (3)

dG : _ Do A' ac . ......... (1)

df ax
where G quantity of material diffusing across a
plane,
time, second,
Do = molecular diffusion coefficient, sq cm/
sec,
A' = cross sectional area for diffusion, sq
cm,
e concentration, volume fraction, and
X distance, cm.
Typically, diffusion coefficients as defined in
Eq. 1 are a function of concentration. To be com-
pletely accurate, the relationship between diffusion
coefficient and concentration must be included in
Eq. 1 before integration. The mathematics describ-
ing behavior with a variable diffusion coefficient
are quite complicated. (The interested reader can
see Crank 16 for further discussion.) However, it is
often possible to represent diffusive behavior
approximately by selecting an effective average
diffusion coefficient. The remainder of this paper
deals with systems that can be approximated in
this manner. Hence, we will use the term "diffusion
coefficient" for a fluid system and mean a diffusion
coefficient constant and independent of concentr'a-
tion. If the diffusion coefficient varies slightly with
concentration, then Do for a 50 per cent:50 per cent
mixture can be use'd.
By using Eq. 1 and the proper boundary conditions,
an equation can be derived giving fluid composition
as a function of distance and time for two stationary
miscible fluids (with an initially sharp interface)
mixing by diffusion. Such an equation has been
derived by several investigators and IS shown as
Eq.2.
where X90 = distance from the initial interface
where the composition is 90 per cent
of the fluid under consideration, and
X 10
distance from the initial interface
where the composition is 10 per cent
of the fluid under consideration.
Now let us see what differences to expect if the
diffu sion is taking place in a porous medium. As a
first approach, let us represent the porous medium
as a bundle of straight capillary tubes.
For steady-state diffusion, Eq. 1 may be rewritten
as
~ =- 0 (area open for diffusion) ~c . (4)
where D apparent diffusion coefficient, sq cm /
sec, and
L = total length of tube.
In the oil industry, the apparent diffusion coeffi-
cient in a porous medium, D, is usually based on
the average cross-sectional area open for diffusion
and the over-all length (as shown in Eq. 4). For
the bundle of capillary tubes, the diffusion rate is
proportional to the true molecular diffusion coeffi-
cient Do times the area open for diffusion. Substi-
tuting these values into Eq. 4 shows that, for the
straight capillary model,
...!2. = I . . . (5)
Do
The straight capillary model is not a very good
representation of a porous rock. Carman,14 for
instance, has shown that in a porous rock, fluids
must move on the average at about 45° to the net

C·: - I
2 [
I ± erf ( X)] ..
2..fl5;t
. . . (2)
~
0,8

0.6

where e concentration of fluid under considera-

tion, volume fraction,
X distance measured from original position
of interface, cm, and
: 0 :
I

~
0
.L [~'-C?_-_X...l..Q.l·
I 3.625

:IO~O[O;~~~.'8J
.
I -5 cm Z
erf = error function. , 00;0: I XIO SiC
I
Note that the sign to be used within the bracket
depends on the initial conditions. If the concentra-
tion initially steps down (i.e., at t = 0, e = 1 for -0."
X < 0 and e = 0 for X > 0), then the minus sign is 0.01 0.1 0.' I 2 5 10 20 so 70 to" H U.' ".9 91.99
appropriate. If the concentration initially steps up PER CENT OF INITIAL TRACER CONCENTRATION

(at t = 0, e = 0 for X < 0 and e = 1 for X> 0), then

FIG. I-CONCENTRATION PROFILE (ON ARITHMETIC-
the plus sign is appropriate. PROBABILITY PAPER) RESULTING FROM DIFFUSION
Eq. 2 will plot as a straight line on arithmetic- OF A TRACER INTO WATER (Do = 1 x 10- 5 SQ CM/SEC).

MARCH, 1963 71
 1.0,------------------...,
L-~~I
_no~ .oo".'~" ""AT'" ")
RATIO PREDICTED BY A SIMPLE 45"

~.
--- ~ @~o
@~~ ~@O/~ O~~~
® ~ b ..

---
---
o ~ 0 <y
~ @ ~;;;. ~O
~
0'-'@;;;..
~ 0 to~ 0 (f#'" fi: /"
~ '}~ 0
0""0 0 0
0 ---
-----
---1--"7 .."-0----;"-0---

O.6r-
§ ° "y ° I I
.x. rP 0 ••• : .
x"':

---
FIG. 2 -
0d'''h 0/o~~ & 0
O~ ~ ~ 0 0 0
~OOO~O~O~O
e. . . . . 0

SKETCH SHOWING THAT, IN A POROUS

MEDIUM, FLUID IS GENERALLY FLOWING AT ABOUT
- 0.4r-
o

• PENMAN (37)
x HOOGSCHAGEN (29)

45° WITH RESPECT TO THE AVERAGE DIRECTION " ZUNKER (60)

0.2 f-
OF FLOW. o VAN SAVEL (55)

direction of flow. See Fig. 2. Hence, when the fluid

has traveled a net distance L, it has traveled an 1 I I I
actual average distance of about V1 L. Substitution 0.2 0.4 0.6 O.B 1.0
of this value into Eq. 4 shows for this model of a FRACTION POROSITY
porous rock, FIG. 3 - RELATIVE DIFFUSION COEFFICIENT IN
PACKS OF GRANULAR MATERIAL.
o = 1 = 0.707 . . . . . . . . . (6) dispersion in the longitudinal direction or direction
00 ft of gross fluid movement. The second is dispersion
transverse to the direction of gross fluid movement.
Fig. 3 shows the data of several investigators The dispersions will not be equal in these two
and indicates that this simple approach is a fair directions; hence, we must consider each separately.
approximation to the diffusion coefficient in un-
consolidated media. LONGITUDINAL DISPERSION
A more sophisticated and comprehensive approach In order to understand the increase in longitudinal
has been suggested by Brigham, et al,9 Grane and mixing resulting from fluid flow, we will first con-
Gardner,26 van der Poel58 and others. These inves- sider dispersion in a single straight capillary tube.
tigators recognized that there is an analogy between Several investigators have attacked this problem
diffusion and electrical conductivity in porous in some detail. Then we will consider a network of
media. The law governing electrical conductivity short capillaries (a little better representation of a
is completely analogous to Eq. 4. It is easily shown porous rock). Finally, we will discuss longitudinal
that diffusion in a porous medium is related to the dispersion actually observed in packs of granular
formation electrical resistivity factor by Eq. 7. * material.
This equation can be used for cemented rocks as Dispersion in a Capillary Tube
well as unconsolidated packs. Again as a first approach, let us consider longi-
tudinal dispersion in a capillary tube. Suppose the
o I • . • • . . . • . . . . . . (7) tube is filled with one fluid and a second fluid is
Do = Fe/> injected at one end of the tube. If the two fluids
are of the same viscosity, if diffusion effects are
where F formation electrical resistivity factor, of no significance and if flow is laminar, then the
and concentration of displacing fluid in the effluent
¢ fractional porosity. stream is easily determined by integrating the flow
Verification of Eq. 7 has been reported by several equation for laminar conditions. There will be no
investigators. 23 ,30, 50 injected fluid appearing in the effluent stream until
one-half of the tube volume has been injected. For
DISPERSION IN A POROUS MEDIUM continued injection, the effluent concentration is
given by Eq. 8. (Note that this equation gives the
In the previous section we have discussed diffu- instantaneous average concentration of the effluent,
sion in the absence of gross fluid movement. If not the average concentration across any cross-
fluids are flowing through the porous medium, then section of the tube.)
there may be some additional mixing taking place.
This increased mixing caused by uneven fluid flow V. 2
or caused by concentration gradients resulting from x= 1- ( 2~) (When j
V ~ - l.p) '" (8)
fluid flow will be designated dispersion. There are 2 .
two types of dispersion to be considered. One is
where x = volume fraction of injected fluid in the
*Note that if the apparent diffusion coefficient had been effluent,
defined in tenns of the totat cross-sectional area in Eq. 4 rather
than the area open for diffusion, then Eq. 7 would have 'been Vp total volume of the tube, and
DIDo = 1/F. V volume injected.
SOCIETY OF PETROLEUM ENGINEERS JOURNAl.
72
 A derivation of this equation has been shown by
Fowler and Brown,24 Rifai,44 and others.
In the actual case, molecular diffusion will cause
mixing along the interface. The net result will be
a mixed zone growing at a more rapid rate than
would obtain from diffusion alone, but less than
the rate predicted by Eq. 8. Taylor 52 and Arisl
have studied the case where the time necessary for
appreciable concentration changes to appear, owing
to convection transport, was long compared with the
"time of decay" during which radial variations of
concentration were reduced to a fraction of their
initial value through the action of molecular diffu-
sion. Theoretical equations derived by these inves-
tigators showed that, if one fluid were displaced
by another fluid under these conditions where
diffusion could nearly damp-out radial concentration
variations, then a symmetrical longitudinal mixed
zone would be established. The mixed zone would
travel with the mean speed of the injected fluid and
would be dispersed as if there were a constant
dispersion coefficient given by Eq. 9.
u2a 2
Kn = 00 + --0- . . . . . . . (9)
"'"' 48 0 Let us now proceed to a discussion of longitudinal
longitudinal dispersion coefficient, sq
cm/sec,
U average velocity, cm/sec, and
a radius of the capillary, cm.
The effluent concentration, for the capillary tube,
is given by the diffusion equation, Eq. 10.
C=t[l±
are negligible, and therefore Eq. 8 is applicable, if
2
-Ua > 250.
DoL
An intermediate region where neither Eq. 8 nor
Eq. 9 is valid has also been studied by Bosworth8
and Van Deemter, et al,56 but their studies did not
result in a quantitative representation of behavior
over the full intermediate range. Recently, Bailey
and Gogarty4 have studied dispersion over the full
range by solving the diffusion and flow equations
numerically.
Dispersion in a Network of Capillary Tubes
Again we should note that a bundle of straight
capillary tubes is not a very good representation of
a porous rock. de Jong 17 and Saffman 45 - 47 have
studied a random network of capillaries. Their
mathematical model is a more accurate description
of a porous rock than is a bundle of straight capil-
laries, but the results are still not quantitatively
correct for rock. The analysis is valuable, however,
from a conceptual point of view. Saffman's results
can be sketched as shown on Fig. 4.
Dispersion in Packs of Granular Material
dispersion in packs of granular material. Most
investigators studying this problem have filled a
packed column with one fluid, displaced it with
another fluid, and measured fluid composition at
the exit end of the tube as a function of displace-
ment. Brigham, et al, 9 have shown a convenient
method for determining the dispersion coefficient
from data of this type. In a slight modification of
their method, the function
V/V - 1
P
yV/Vp
is plotted vs

erf(,j~Oi~L) c;;~)] ... (10)

the per cent of displacing fluid (on arithmetic-
probability paper) as shown on Fig. 5. The disper-
sion coefficient can then be calculated with the aid
of Eq. 11.
Under what conditions will Eqs. 8 or 9 be valid?
Taylor has shown that the ratio of time necessary
for damping of radial concentration variations to >"90->"10\2
. • . . . . (11)
the time to get a significant change in concentration KJ.. = UL (
2
3.625 )
. .. Ua
IS proportional to the dImenSIOnless group - -
0.04,.---------------------------,
DoL
Furthermore, he has shown that diffusion effects 0.03-

100~----------------------------~--.
K.· LONGITUDINAL DISPERSION IN THE CAPILLARY
~90
NETWORK, CM.2, SEC. 0.01~- - -
Do' DIFFUSION COEFFICIENT, CM!!/SEC.
U • AVERAGE FLUID VELOCITY THROUGH 0-
THE NETWORK, CM./SEC.
10
"'1°
:.:0 .JI. • LENGTH OF EACH CAPILLARY, CM.
-0.0,:; _

-0.021-

- 0.041- I I I I I II : I I I I I I I I I
0.1 0.01 .05 0,20.5 1 2 5 10 20 30 40 50 60 70 80 90 95 98 99
PER CENT DISPLACING FLUID

FIG. 5-TYPICAL EFFLUENT-COMPOSITION CURVE

FIG.4--LONGITUDINAL DISPERSION COEFFICIENTS FOR MISCIBLE DISPLACEMENT OF A SAND-PACKED
FOR A RANDOM NETWORK OF CAPILLARIES (FROM COLUMN (ON ARITHMETIC-PROBABILITY PAPER)-
SAFFMAN, REF. 45). FROM BRIGHAM, ET AL (REF. 9).

MARCH, 1963 73
 VIVp - 1 uniform size connected in series. The chambers
where '\'90 = _~ when effluent contains 90 are filled with one fluid and a second fluid is
vVIVp injected at one end at a constant rate. The compo-
per cent displacing fluid. sition within each chamber is maintained uniform
Nearly all literature data show that the longi- at all times by complete mixing (in the case of
tudinal dispersion coefficients for unconsolidated granular packs and for some flow conditions, the
sand or bead packs can be represented as shown by concentration in each pore space is maintained
Eq. 12. essentially uniform by diffusion). Eq. 14 gives the
concentration of displaced fluid in each cell of a
U~ U~
00
E
= I. 7 5 ---0;;-;
(
2 < 00 < 50) very long series of cells.

. . . . (12)
en = I - e-qt/v [nI=n (qt/v(
l
+ I] • . . (14)
n =I
n.
where E longitudinal convective dispersion co-
=
efficient, sq cm/sec, and where en
concentration of the displaced fluid in
dp = particle diameter, cm. the nth cell,
Furthermore, in the region where both diffusion q injection rate,
and convective dispersion are important, the total t = time, and
dispersion coefficient is the sum of these two v = volume of each celL
coefficients. Hence, the total longitudinal dispersion Fig. 7 shows the concentration of displaced fluid
coefficient in an unconsolidated sand pack can be calculated from Eq. 14 when a pore volume of 100
represented as shown by Eq. 13, and is sketched cells has been injected (i.e., qtlv = 100).
on Fig. 6. This figure shows that a mixed zone is established
K{= 0 + E which moves with the mean speed of the injected
fluid; that is, the 50 per cent point is essentially
o E at the hundredth cell (if the displacement had been
00
=
00
+ Do piston-like, there would have been an abrupt change
in composition after the hundredth cell). Furthermore,
K.e = I +
00 Fq,
I. 75 Ud
00
p
; ( Ud
p
00
< 50) the growth of the mix zone, relative to the 50 per
cent composition point, can be represented by a
constant dispersion coefficient (i.e., plots as an
. • . . . • . • • • . • . . (13) essentially straight line on probability paper). Fig.
Can the dispersive behavior of granular packs be 7 is equivalent to Fig. 1. The apparent dispersion
represented fairly accurately by an "equivalent" coefficient can be calculated with Eq. 3. By noting
bundle of capillary tubes? A comparison of Eqs. 9 that the time to inject 100 cell volumes is equal to
and 12 shows that in packed columns the longitudinal the length of 100 cells divided by the average fluid
dispersion coefficient is proportional to the first velocity, it follows that the apparent dispersion
power of average velocity (if longitudinal dispersion coefficient varies as the first power of the mean
is large compared to longitudinal molecular diffu- velocity (the same behavior as exhibited by porous
sion), whereas for capillary tubes the dispersion packs, see Eq. 12).
coefficient is proportional to the second power of A pack of granular material, of course, can be
average velocity. How can we explain this signifi- thought of as a series of chambers or pore spaces
cant difference in behavior? connected by smaller openings. This is indicated
A clue is furnished in work reported by Aris and graphically on Fig. 8, which shows the shapes of
Amundson.3 These investiga·tors have studied the void spaces for various types of symmetrical pack-
dispersion to be expected in mixing chambers of ing of spheres as reported by Graton and Fraser. 27
Study of dispersion data from packed columns
COEFFICIENT. cm~/ ..<=.
KJ. -LONGITUDINAL
Do. MOLECULA.R
U • AVERAGE
DISPERSION
DIFFUS-tON
INTERSTITIAL
COEFFICIENT. em,lt/ nc .
VELOCITY. em.I •• c. I
-
y' reveals that the concentration profile in the mixed
zone is not typically a perfect S-shaped probability
dp • AVERAGE DIAMETER OF. THE PARTICLES, em. I

SOLID LINE. EQUATION (13)

,.~_
- ii-
I
120

-;--1
0 / I 110
e
1 • • •
//: I ~~:::~~ ~~!
I CONTROLS
..
II:
ID 100

F--~ r- :.

- -
z

DIFFUSION
CONTROLS
-j
-1
-l
l-
I- . '0
j
o
_ 80
I c
0.1 L--'--.J-'-'-LlllJ':-L-J.·_llLllIL-l LLLill1L---'--.L.l.J-LW":--'-L.LJ.-,-,,!:
0.001 0.1 10 100
TO
Ud p
0. 0.01 0.1 0.5 I 2 5 10 20 50 70 90 95 9. 99.5 99.9 99.99
Cn i CONCENTRATION OF DISPLACED FLUID IN CELL n
FIG. 6-LONGITUDINAL DISPERSION COEFFICIENTS
FOR UNCONSOLIDATED, RANDOM PACKS OF UNI- FIG. 7 - DISTRIBUTION OF COMPOSITION ALONG A
FORM SIZE SAND OR BEADS (DATA FROM REFS. 11, SERIES OF MIXING CELLS; CALCULATED FROM EQ.
42, 53 AND 59). 14. (qt/v) = 100.
74 SOCIETY OF PETROLEUM ENGINEERS JOURNAL
 ~
10'
INSUFFICIENT
TI ME FOR DIF- /
V!Y CASf I CASE 2 CASE J
10'
0
I
F
FUSION TO
EQUAL IZ E
COMPOSITION
IN PORE
/
F
U SPACES /
I MIXING I-
S
10' I
0
~
N
c
I T~~~~Y I
"I
><00
10· 0
N
T
R
I

~
0
10 L K/= LONGITUDINAL DISPERSION
S
COEFFICIENT
Do. DIFFUSION COEFFICI.ENT
U = INTERSTITIAL VELOCITY
CASE 4 CASE 5 d p • DIAMETER OF PARTICLE

FIG. 8 - SHAPES OF PORE SPACES IN SYMMETRI-

1
CAL PACKS OF SPHERES (FROM GRATON AND 10- -3 -2. -I 4 5

FRASER, REF. 27). 10 ~ 10 10 10 10 ~ ~

FIG. 9 - RANGE OF VALIDITY OF MECHANISMS OF

curve as predicted by theory (i.e., Eq. 10). Rather,
the mixed zone is often slightly asymmetrical with
the trailing edge being stretched out (as shown by
the data points on Fig. 5). This behavior has also
been reported by other investigators such as Von
Rosenburg.59 Scheidegger and Larson 48 studied
this phenomenon to see if it resulted from asymmetry
of boundary conditions, but they found this could
not be the explanation.
For many purposes it is permissible to neglect
the slight asymmetry of the mixed zone. If the
dispersion process needs to be described with
great precision, then the asymmetry of the trailing
edge must be included quantitatively. Several in-
vestigators have proposed a simplified "stagnant
pocket" theory to describe the asymmetry. So far
these theories do not seem adequate to fully explain
the observed phenomena. For further discussions of
the stagnant-pocket theories, the reader is referred
to Aris,2 Turner,54 K oump32 and Goodknight, et
al. 25
In summary, we can say that a granular pack may
be thought of as a network of flow chambers (having
random size and flow conductivity) connected
together by openings of smaller size. Such geometry
will result in a slightly asymmetrical mixed zone,
with the longitudinal dispersion coefficient approx-
imately proportional to the first power of interstitial
velocity (if composition is equalized in pore spaces
by diffusion).
If we flow fluid through a pack at a very low
rate, then there may well be time for diffusion to
equalize concentration within each pore space.
However, if we increase the velocity in the inter-
stices high enough, we will eventually reach a
velocity at which there is insufficient time for
diffusion to equalize concentration within each pore
space. For approximately radially symmetrical pore
spaces, the ratio of time needed to damp-out con-
centration variations to the time available is pro-
Ud
portional to --.P.. . Hence, the dimensionless gr,oup
Do
UclplDo should be a measure of how effectively
diffusion can equalize concentration within pore
spaces. Dispersion studies have shown that Kt IDo
is about proportional to Udp IDo if 4 < UdplDo < 50;
and, hence, the mixing-cell theory is apparently
valid. At values of Udp IDo greater than about 50,
MARCH, 1963
DIFFUSION AND DISPERSION IN UNCONSOLIDATED,
RANDOM PACKS OF UNIFORM SIZE BEADS (LAMINAR
FLOW).
the dispersion is greater than would be explained
by the mixing-cell theory and presumably shows
that diffusion is not entirely equalizing the concen-
tration within each pore space. Most data indicate
K[IDo is about proportional to (UclpIDo)1.2 in this
region. Fig. 9 shows our present concept of the
mechanisms of diffusion and dispersion in uncon-
solidated, random packs of uniform-size beads (for
laminar flow).
So far, the discussion has centered around
"random" packs of beads. We have also assumed
that the pore size distribution and the degree of
inhomogeneity are the same within each random
pack. Raimondi, et aZ,42 have shown that the
convective dispersion coefficient can be written
as shown in Eq. 15.
. (15)
where c = a constant, and
a = a measure of the inhomogeneity of the
pack.
Furthermore, Aris and Amundson,3 Carberry12 and
Prausnitz 40 have shown that the mixing-cell theory
leads to a value of c = 0.5. This would then give a
value of about a = 3.5 for the typical random pack.
Literature data suggest that packing of large beads
is usually better than for small beads. Hence, we
should expect the value of a to be a bit smaller for
"random" packs of large beads as shown by Fig.
18 (but a should never be less than unity). Also, a
may be larger for poorly packed beads. The final
equation describing longitudinal dispersion is shown
as Eq. 16.
= _,_ + 0.5 Uudp ; Uudp < 50
Fcp Do Do
. . . . . . . . . . . . . . . . . (16)
A convenient graphical presentation of this
equation as well as the relationship if Uadp IDo >
50 is shown on Fig: 10.
TRANSVERSE DISPERSION
If a sand-packed column is arranged as shown on
Fig. 11, a mixed zone will grow in a direction
transverse to the direction of gross fluid movement.
The mechanism leading to this type of dispersion
75
has been described by several investigators as a packs IS given by Eq. 18 and p10tted on Fig. 12.
"stream-splitting" mechanism (with diffusion caus-
ing mass transfer between streams). If a profile of K Udp
fluid composition is made along a line perpendicular
t - + 0.055 . (18)
Do - F¢ Do
to the direction of gross fluid movement, the typical The mixing - cell theory has been developed to
S-shaped concentration profile will be observed. cover transverse dispersion in a network of mixing
The transverse dispersion coefficient can be deter- chambers.5.18.43 Furthermore, the theoretical equa-
mined by plotting (per cent composition) vs (distance tion has proved to be approximately correct when
from 50 per cent composition) on arithmecic- turbulence equalizes concentration variations within
probability paper, as is shown by the typical each pore space. However, the transverse mixing-
experimental data on Fig. 11. The dispersion cell theory should not be applicable in the laminar
coefficient can then be calculated with Eq. 17. region. Fig. 12 shows that transverse dispersion is
2 large compared to transverse diffusion if (U 4> IDo )
Kf _ J!... ( X 90 - X IO ) • • • • • • (17) is greater than about 100. However, Fig. 9 shows
- L 3.625 that if (U dp ID0) > 100, then there is insufficient
time for equalization of concentration within pore
where K t transverse dispersion coefficient, sq spaces by diffusion. Hence, as the interstitial
cm/sec, velocity increases, we pass from a region where
u average longitudinal interstitial veloc- transverse diffusion dominates into a region where
ity, cm/sec, there is stream splitting with mass transfer but
L dista~ce from inlet, cm, and with insufficient residence time to completely
X transverse distance from 50 per cent damp - out concentration variations within pore
composition point, cm. spaces.
The total transverse dispersion coefficient is Blackwell's work7 suggests that pore size dis-
the sum of the diffusion coefficient in the porous tribution and inhomogeneities will have roughly the
medium (which is, of course, the same as in the same effect on transverse dispersion as on longitu-
longitudinal direction unless the porous medium is dinal dispersion. Hence, we write Eq. 19 and draw
anisotropic) and the transverse convective disper- the final correlation for transverse dispersion as
sion coefficient. The final expression for laminar shown on Fig. 13. Tpe inhomogeneity factor a for
flow conditions in typical unconsolidated, random
100 K
t
• TRA;SVERSE DISPERSION COEFFICIENT. -C;.~SEC.
KJ -LONGITUDINAL DISPERSION C~EFFICIENT. cm~/"c, Do" MOLECULAR DIFFUSION COEFFICIENT. CM~SEC.
Do -DIFFUSION COEFFICII:NT. em.I •• c. U t AVERAGE INTERSTITIAL VELOCITY. eM./5Ec.
U -AVERAGE INTERSTITIAL. VELOCITY, em.'''c. dp " AVERAGE DIAMETER OF THE PARTICLES, eM.
dp.AVERAGE PARTICLE DIAMETER, em.
a -INHOMOGENEITY FACTOR 10
F· FORMATION ELECTRICAL RESISTIVITY FACTOR SOLID LINE· EQUATION (18) _

•• POROS ITY
I
I
I
I
I CONVECTIVE

o IFFUSION ..j
-.J r- DISPERSION
CONTROLS
CONTROLS ...I l-
I-
0.1 L-L-LWlill_LLLWlIL---.l-LJ...LillJL-L-LWilll_LllillllJ
0.1 10 10 4
(.!!£fl
FIG. 12 -TRANSVERSE DISPERSION COEFFICIENTS
10 10' 10' 10'
FOR UNCONSOLIDATED. RANDOM PACKS OF
UNIFORM-SIZE SAND OR BEADS (DATA FROM REFS.
26 AND 7).
FIG. 10 LONGlTUDINAL DISPERSION COEFFI- KI • TRANSVERSE DISPERSION COEFFICIENT. em.l/ne.
CIENTS FOR POROUS MEDIA. 00 • DtFFUS ION CDEFF Ie lENT, c:m~ Inc. EXTRAPOLATION
OF DATA
U_AVERAGE INTERSTITIAL VEL.OCITY, ern./see.
2.0
d p ' AVERAGE PARTICLE DIAMETER, em.

/
CT' INHOMOGENEITY FACTOR
1.5

..j
Z
\;:- i=-= ,:--=.,-
flo ID _ -~-:...i-=---=-":
Xeo

- -[:..~O:O:IO~ LINE
I K,'
J
JL [ x.o - X10
L 3.625
t F • FORMATION ELECTRICAL RESISTIVITY
." • POROSITY
FACTOR

.
:::;

0
on
" -- ~~

c,;.OfILE ALONG /
/ I
I K,. TRANSVERSE DISPERS-
,.
0
THIS LINE ION COEFFICIENT,
I CIII.; ..<:.
0: fu. AYG. LONGitUDINAL
"- -0.5
...<> / I INTERSTITIAL VELOCITY,
I;III'/ •• C •

Z
""
I-
III
0
-1.0

-1.5
_ _ X..!!!1- _ _ _ j I
I
L'DISTANCE
,m.
IX • TRANSVERSE
FROM INLET.

DISTANCE
FROM 50.,., COMPOSIT~

-2.0 • I
'ON POINT. CIII.
o.~L.,-LWlllll-1-LLillJJL'O-LWlllll,O"-'LlJlJ.JJlL,o,,..LWlllll,o~.LlJlJ.llI,L,,..LillllJlJ •
o ,o
0.01 50 70 80 90 95 98 99 99.9 99.99
PER CENT OF TRACER FLUID

FIG. 11 - TRANSVERSE CONCENTRATION PROFILE FIG. 13 - TRANSVERSE DISPERSION COEFFICIENTS

ON ARITHMETIC-PROBABILITY PAPER. FOR POROUS MEDIA.

76 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

random packs of small particles will again average
about 3.5.
Kt I Uudp
- F¢ + 0.0157
( U~ dp < I 0 4) . • .
ADDITIONAL VARIABLES
In the previous discussion we have considered a
simplified physical system, that is, very large
columns packed with unconsolidated, uniform-size,
nearly spherical particles, and fluids of equal
viscosity and density in laminar flow. Now let us
consider the following additional variables: (1)
ratio of particle diamet~r to column diameter, (2)
particle size distribution, (3) particle shape, (4)
packing or permeability heterogeneities (as in natural
rocks), (5) fluids with different viscosities, (6) fluids
with different densities, (7) fluids in turbulent flow,
and (8) effect of an immobile phase.
RATIO OF PARTICLE DIAMETER
TO COLUMN DIAMETER
If spherical particles are packed into a cylinder,
there will be packing irregularities near the con-
tainer walls. Schwartz and Smith 49 present data
showing zones of high porosity extending two to
three particle diameters from the container walls.
Graton and Fraser 27 indicate the following, how~
ever. "Interference (to regular packing of spheres)
caused by the container wall is not limited to the
immediately adjacent region but is propagated
throughout the entire area." Packing irregularities
of this type will no doubt have an effect on both
longitudinal and transverse dispersion.
Several investigators have studied the wall effect
on transverse dispersion for turbulent flow. Fig. 14
Kt I Uudp {3t
= + 0.0157 . (20)
Do F¢ Do
where (3t = factor to account for wall effect (see
Fig. 14).
. . . . . • (19) PARTICLE SIZE DISTRIBUTION
The particle size distribution in naturally occur-
ring, unconsolidated sands often follows a log-
normal distribution. That is, the particle size vs
cumulative per cent plots as a nearly straight line
on logarithmic-probability paper. What effect will a
wide particle size distribution such as this have
on flow behavior? In a solid pack of various size
particles, the small particles can reside in the
interstices between large particles. The size of
the average pore space can therefore be character-
ized by a dimension smaller than the median
particle size. Fig. 15 shows the particle size
distribution for three mixtures of glass beads
reported by Raimondi, et a1. 42 Permeabilities of
the three mixtures were about the same. Hence, the
effective particle size (so far as calculating
permeability is concerned) for these log - normal
distributions is the particle size corresponding to
the 10 per cent cumulative fraction. If we assume
that the effective particle size characterizing
dispersion can also be taken as the particle size
corresponding to the 10 per cent cumulative fraction,
then the data can be correlated as shown on Fig. 16.
The distribution of particle sizes can be quanti-
tatively rlescribed by the slope of the log-normal
line; see Eq. 21.
dP JS4
S = log J
dp S4 - log dpJ 16 = log
dp JI6
. . . (21)
IO~-----------------------'

shows a correlation obtained from the data of

Latinen,33 Fahien and Smith: 2 and an approximate
relationship proposed by Singer and Wilhelm. 51
These experimental data suggest that the increase
in transverse dispersion in the laminar region will
be, roughly, the same order of magnitude as in the EFFECTIVE
turbulent r~gion. Hence, we can write Eq. 20.

~
II

~I"C
..
I-
*4.0 ~ dp = PARTICLE D lAM ETER
101
~ 0.1 ~_~~_ _ _~M~E~D~I~U~M~I_ _,
~

z z a::
dT = TUBE
0
iii
0:
0
<I)
a: 3.0
DIAMETER
'"
11.

"'
Q. '"
Q.
<I) <I) SINGER 8 WILHELM (51)
0 0

0.0 I 2L...-....I....-.....L...---'2-0-'-..J.....J..-L........1---'--'--:'----'

CUMULATIVE AMOUNT SMALLER THAN

STATED SIZE. "I.
II FIG. IS-PARTICLE SIZE DISTRIBUTION
cQ FOR THREE MIXTURES OF GLASS BEADS
ALL HAVING ABOUT THE SAME PERME-
FIG. 14 - EFFECT OF dp/riT ON TRANSVERSE DIS- ABILITY (FROM RAIMONDI, ET AL, REF.
PERSION (TURBULENT FLOW). 42).

MARCH, 1963 77
where S
slope of the log-normal particle size
distribution, and ....>-
dp J84 particle diameter of the 84 per cent t; en
cumulative fraction. ~ ~ 2.0
~ ~
Raimondi, et aI,42 and Orlob and Radhakrishna 36 en :;;
indicate that wide particle size distributions will
lead to increased dispersion. Fig. 16 shows the i ~
..
....:I: a:

order of magnitude of increased dispersion that ::: ~ 1.5

should be expected as a function of the slope of

the particle size distribution.
d
....
a: 2
c c
..
Q

... en
a: a:
PARTICLE SHAPE ~ ~ 1.0~______~~__L-__~~~__~~__-L~~~~
0.6
The effect of particle shape on dispersion has b b
9' , SPHERICITY
been studied by Bernard and Wilhelm,6 Carberry, 11 FIG. 17 -- APPROXIMATE EFFECT OF PARTICLE
Ebach and White,20 and others. These investigators SHAPE ON DISPERSION.
have studied spheres, cubes, rings, saddles, crushed
granular material, etc. It is generally found that PACKING OR PERMEABILITY HETEROGENEITIES
packs of nonspherical particles lead to greater In this section we will discuss the effects of
dispersion than do packs of spherical particles of two types of packing or permeability heterogeneities:
about the same size. (1) those arising in "random packs" of spheres;
The concept of "sphericiry" of a particle lO is and (2) those typically found in cemented, outcrop
often used to correlate pressure-drop data and other or reservoir rocks.
flow behavior in packed beds. The sphericity of a The effect of packing heterogeneities in "random
particle is defined by Eq. 22. packs" has been described by several investigators.
'" = particle sphericity Blackwell 7 says the following.
The increase in the dispersion coefficient with
surface area of a sphere. haVing) decreasing sand sizes for sands smaller than 20-30
'" = ( same volume as particle mesh was noted by Van Deemter, et ai,57 for longi-
surface area of particle tudinal dispersion. Van Deemter, et al. attribute this
increase to bridging by the particles and other
. . . . . . . . . . • . . . . . . . . (22) microscopic packing irregularities which occur more
Fig. 17 is a correlation developed from literature frequently as the sand size decreases or the particle
shape becomes more irregular. Ebach 19 has found
data which shows the order of magnitude of increased that for 30-mesh-or-Iarger particle sizes the longi-
dispersion to expect in packs of nonspherical tq,dinal dispersion coefficients are independent of
particles. We believe this correlation is roughly size. Based on the above observation of Ebach, it
correct for both longitudinal and transverse disper- would be expected that the lateral dispersion
coefficient for particle sizes larger than those in
sion in the laminar region. 20-30 mesh sands would be approximately the same
--. as that measured for the 20-30 mesh sand in the
o 16r-------------~
present investigation.
Ul
Z Similarly, Brigham, et al,9 reported the following.
o~
-::>
14
APPROXIMATE RELATION-' If two porous media were geometrically similar but
~ID SHIP FOR SAND PACKS.
with different particle sizes (such as two identically
ID (SUGGESTED BY THE
-w
a:N DATA OF ORlOB AND I packed bead packs different only in a diameter of
~-
UlUl 12
RADHAKRISHNA (3") I beads), they would have identical values of (a) . ...
0
However, in practice, (a) • .• has experimentally been
w found to increase for finer particles. Klinkenberg and
w...J Sjenitzer 31 report values of (a) increasing from ...
N~
-~ 10 (eightfold increase) as particle diameter drops from
Ula:
<t 1 to 0.005 mm. This undoubtedly reflects either the
wa. difficulty of obtaining uniform packing with fine
...J
Ow particles or the greater relative variation in particle
-> 8 /
~-
a:~
size usually found with fine particles.
<to
/ Fig. 18 shows the inhomogeneity factor a for
a.w
z ...
... 6 random packs of various size beads. This correla-
Ww / tion was developed from literature data. These
~w
(!);:;:
<t
/ values of a may be used in Eqs. 15, 16, 19 and 20,
wUl 4 FOR GLASS BEAD PACKS
or on Figs. 10, 13, 16 and l7.
:t: / (FROM DATA OF RAIMONDI
~w
:t: /
et. 0I,142)} Another type of inhomogeneity that is of great
a:~
2 /- interest is that in cemented outcrop or reservoir
0
u.a: k rocks. Part of the inhomogeneity is of a small
0
bU. geometric scale (from pore to pore). However, we
b
---- 0
0
S, SLOPE
0.2 0.4
OF
0.6
PARTICLE
0.8
SIZE
1.0
also know that there are larger-scale inhomogeneities
in natural sandstones. That is, the average perme-
01 STRI BUTtON ability of the rock varies over distances of a few
inches or feet.
FIG. 16 -- EFFECT OF PARTICLE SIZE
DISTRIBUTION ON DISPERSION. Dispersion in outcrop rocks has been studied by

78 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 10 coefficient. As the viscosity ratio gets very favor-
\VTYPICAL VALUES FOR able, the beneficial effect would approach a limiting
ORDINARY LABORATORY PACKS
0:
et- value.
c
•
....
..
t- If the fluid composition within pore spaces is
o

)-

...zt:
6t- . .\
\
...co \ ,
c
~
c
:I:
41- • " ~ ........
! • I .........
b 2t- • dispersion.
0 dispersion coefficient should therefore depend on
0.01 0.1
PARTICLE DIAMETER. MM.
FIG. 18 - INHOMOGENEITY FACTOR FOR RANDOM
PACKS OF SPHERES.
several investigators. They all find that dispersion
is larger than one might have suspected from parti-
cle size alone (thus reflecting the increased
heterogeneity). It is convenient to report the value
of (dp u) that will bring dispersion data for cemented
rocks into agreement with our equations and charts.
Table 1 shows values of (dp u) for natural sand-
stones.
It appears that an average value of about udp =
0.36 em can be used to approximate the behavior of
all sandstones studied. This should give the correct
amount of dispersion within blocks having dimen-
sions of a few feet.
FLUIDS WITH DIFFERENT VISCOSITIES
The additional variables previously considered
have dealt with properties of the porous medium.
Let us now consider some of the additional fluid
variables.
First, let us suppose the displaced and displac-
ing fluids have different viscosities. If the displac-
ing fluid is less viscous than the displaced fluid
(an unfavorable viscosity ratio), viscous fingers
will be formed. This topic is outside the scope of
this paper and will not be considered further here.
However, if the displacing fluid is more viscous
(a favorable viscosity ratio), the usual dispersion
mechanisms will continue to operate. A favorable
viscosity ratio will tend to suppress the effects of
packing or permeability heterogeneities. If we con-
sider a simple analogy of two parallel flow paths
not equalized by diffusion (i.e., if U dplDo is
greater than about 50, see Fig. 9), then a favorable
viscosity ratio may lead to an additional suppression
of dispersion within pore spaces. The favorable
viscosity ratio will tend to straighten interfaces
within pore spaces and thereby reduce convective
In actual porous media, suppression of the
10
the degree of inhomogeneity in the pack, perhaps
the length of the mixed zone (i.e., the mobility
ratio across the inhomogeneity or pore space), and
the relative amount of diffusion within the pore
spaces. Fig. 19 gives the order of magnitude of
suppression for one case reported.
FL UIDS WITH DIFFERENT DENSITIES
If fluids of unequal density are used during
miscible displacements, then gravity forces may
influence dispersion. For vertical displacements,
if the denser fluid is placed above the less-dense
fluid, then gravity will usually cause redistribution
or perhaps gravity "fingers". This topic is outside
the scope of this paper and will not be considered
further here. However, if the denser fluid is on the
bottom (favorable gravity forces), then a stable
displacement will usually occur.
Favorable gravity forces will suppress dispersion
in two ways. First, any unevenness of the front
caused by packing or permeability heterogeneities
will tend to be reduced by a factor which might be
as large as u (since u is a measure of the increased
dispersion due to heterogeneities).
However, if there is not complete equalization of
composition within each pore space by diffusion,
then gravity can cause further reduction of dispersion
within pore spaces. That is, large gravity forces
will tend to straighten interfaces within pores to a
more piston-like displacement.
An effect of gravity forces on transverse dis~r­
sion has been reported by Grane and Gardner, 26
and Pozzi and Blackwell. 39
1.0 I- (2.7 ~ 10 TO IOO)~
FT. LONG; Do ABOUT

with different permeabilities, it is easy to show

/
that a favorable viscosity ratio will reduce the /
0.81--
distance between moving interface s. F or a network /
of flow channels, this effect would no doubt lead /
to a shorter mix·ed zone and a smaller dispersion 0.61-- /
/
TABLE 1 - VALUES OF (u~) FOR OUTCROP SANDSTONES /
Udp
/
/
Source Dispersion Rock (em) /
Grane and
Gardner 26 Transverse Berea 0.25 FROM DATE OF BRIGHAM .t. al.(SU
I BEAD I I I I I 1
Brigham, et 0/ 9 Longitudinal Berea 0.39 RANDOM
OL-_ _~_~_~-L-L~~
rACK

Longitudinal Torpedo 0.17 0.1 I

Raimondi, et 0/ 42 Longitudinal Berea 0.46 MOBI L ITY OF DISPLACING FLUID)
M· MOBILITY RATIO' ( MOBILITY OF DISPLACED FLUID
Handy2S Longitudinal Boise 0.55
Average 0.36 FIG. 19-5UPPRESSION OF DISPERSION RESULTING
FROM FAVORABLE MOBILITY RATIOS.

MARCH, 1963 79
FLUIDS IN TURBULENT FLOW
Turbulent flow conditions are not likely to be
encountered in a petroleum reservoir. However, a
brief discussion of dispersion during turbulent flow
will be valuable for two reasons. First, turbulent
flow conditions may be encountered in laboratory
experiments, and it is therefore desirable to under-
stand turbulent effects (and avoid them, usually).
Second, it will be reassuring if dispersion phenomena
are consistent with the previously discussed con-
cepts as the laminar region is approached from the
turbulent side.
First, let us define the turbulent region for flow
in porous media. We can do this by considering
pressure-drop data. Fig. 20 shows a pressure-drop
correlation for spheres similar to one presented by
Ergun. 21 In the laminar region, the Kozeny-Carman
equation is valid (as is Darcy's law). In fully
developed turbulence, the Burke and Plummer equa-
tion is valid. Fig. 20 shows that, for spheres,
laminar flow will obtain at Reynolds numbers less
than about 10, whereas fully developed turbulence
will obtain at Reynolds numbers in excess of about
1,000.
In porous media, there are no sharp transitions
from laminar to turbulent flow (as there are in open
pipes, for instance). Rather, there is a gradual
tranSItiOn from laminar flow to fully developed
turbulence. Carman 13 has suggested that turbulent
eddies develop at selected spots throughout the
medium. As the Reynolds number increases, the
amount of turbulence increases, finally reaching a
state of "fully developed turbulence". Actually,
the amount of turbulence at a given Reynolds number
coefficients will again be different as in the case
of laminar flow.
In the turbulent region, most investigators have
reported dispersion coefficients in terms of the
"Peclet number". This dimensionless group is
defined by Eq. 23.
dpU
Pe = . . . . . (23)
K
where K == dispersion coefficient_
From theoretical studies (mixing - cell theory)
Prausnitz,40 Carbetry, 12 and Aris and Amundson 3
have shown that longitudinal Peclet numbers should
approach a value of about 2.0 for fully developed
turbulence (i.e., complete mixing in each pore
space). Similarly, Ranz 43 and Baron 5 have shown
that transverse Peclet numbers should approach
about 11 for full turbulence. Figs. 21 and 22 show
literature data of several investigators tending to
verify these limiting theoretical values for full
turbulence.
The limiting values apparently vary somewhat
depending on particle shape and pac king. In turbulent
flow, however, increased dispersion due to packing
irregularities probably will not be as severe as in
laminar flow. We can show this by a simple analogy.
Let us consider two parallel flow paths filled with
particles of differeat diameter such that the ratio
of permeabilicies is 10: 1. If the pressure drop across
oZ
J- DATA FOR
,
W
d GASEOUS SYSTEM THEORETICAL VALUE FDA FULLY

is influenced by particle shape and packing. Hence, j ":ILD~~~~-2U~B~E~E

~
L :_:_:_.._
Fig. 20 should be considered as only a typical plot. =>1""
c.:.:
." ~
. . . : . ; ~DATA FOR LlaUID SYSTEMS
« o o~~o
The theory of dispersion during turbulent flow is z ?-.
0
~ BY
fairly well advanced. Again, a pack of granular
-<:>{

o'!!
0
=>
2~x~o 0 ~
fo oR 0
o

0
0-

.,.,.,'1,
0 (; c::
I-+-

I
TURBULENCE
J-
material has been likened to a series of mixing
3'" o. I L....L-L...LLlilJ.J...,;--'-..LJ...LilJ..U..,.---1.....LLlll..l.l.L-:-LL...LJ...UW
z I
chambers. In this case, fluid streams entering a
~ ~
chamber are thoroughly mixed by turbulence before
leaving the chamber (rather than being mixed by Re • dp Up. REYNOLDS NUMBER
11-
diffusion as was sometimes the case for laminar LEGEND: ._MCHENRy(35) O-CARBERRy!1I1 X- EBACH
(19)

flow). Longitudinal and transverse dispersion

FIG. 21 ~ PECLET NUMBERS FOR LONGITUDINAL
DISPERSION IN TURBULENT FLOW.

LAMINAR -1 r-TURBULENT
" FLOW -1 ~ FLOW x ~DATA
FOR LIQUID
''l., I X SYSTEM I
".. ~TRANSITION -----.l J-
ILl
...J
X I
""'~ I ZONE 1 0 ~ I THEORETICAL VALUE FOR FULLY
=>1_ ILl • ~ _.. I DEVELOPED TURBULENCE,
" I
'J.
6P, PRESSIURE OIFF.
L ~ LENGTH
...
a.", 0-
.~~~J~e8"'~' '1<--- ___ t...
- ILl o 0 (bO
' I"
~,
dp' PARTICLE DIAMETER
.p.
en
a:: o r-
~
FRACTION POROSITY
GASE~~SA ~~~~EM
ILl
I~"'" p' FLUID DENSITY >
en
r- TURBULENCE
,,"" U • AVERAGE INTER- z ~
I ;". STITIAL VELOCITY ...:
a::
KOZENY-CAR~AN EQUATION '~:::~ JL* FLUID VISCOSITY I- I
""')l.:J,"-", I 10
BURKE a PL~MMER EQUATION ____ - ....!. ~.~~"',""._~---l
I~~~~~ __~~~~~~~~~~~wu~ Re • --r
dp
=
Up
REYNOLDS NUMBER
I 10 10'
I6
LEGEND: X - L.o,TINEN!331 . - BERNARD a WILHELM )
REYNOLDS NUMBER, (dP,.U p)
0- FAHIEN a SMITH(22)

FIG. 20 PRESSURE-DROP DATA SHOWING FLOW FIG. 22-PECLET NUMBERS FOR TRANSVERSE DIS-
REGIMES IN POROUS MEDIA. PERSION IN TURBULENT FLOW.

80 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 of Ebach,19 Carberry 11 and Liles.34
;::.
",,- 1.0
0
I- The Peelet number can now be estimated over
<>
...c O.B the whole range of Reynolds numbers. Fig. 24
shows a typical plot for a liquid system (water).
'"~
I-
:r
0.6 Fig. 25 shows a typical plot for a gaseous system
!:! (air). It is interesting to see the differences between
UJ
~ 0.4
gaseous and water systems. Most of these differ-
UJ
<>
z 0.2 ences have been experimentally verified. The
UJ
..J
:::> estimate of transverse Peelet numbers for the
CD
cr: 0 gaseous system in the transition region (from
...:::> I 10
laminar to turbulent flow) are in good agreement
REYNOLDS NUMBER, ¥ with the data of Bernard and Wilhelm,6 Fahien and
FIG. 23 - TURBULENCE WEIGHTING FACTOR. Smith,22 and Plautz and J ohnstone.38 For the liquid
system the transverse Peelet numbers are less than
both flow channels is the same, then we can esti- the values reported by Latinen. 33 However, a
mate the ratio of flow velocities. In laminar flow, smaller value of a (i.e., less heterogeneity or pack-
Darcy's law will apply and the ratio offlow velocities ing irregularity) could partially explain Latinen's
will be about equal to the ratio of permeabilities, data. The predicted longitudinal Peelet number for
or about 10: 1. However, for fully developed turbu- the liquid system agrees with most literature data.
lence, the Burke and Plummer equation will apply The predicted longitudinal Peelet numbers for the
and, if permeability is about proportional to (dp )2, gaseous system are less than the values reported by
then the ratio of flow velocities will be about McHenry.35 Again, this might be explained by a
proportional to (ratio of permeabilities) 1/4, or for less heterogeneous pack than is typical. For in-
this example (10) 1/4 • stance, Fig. 26 shows that there is good agreement
In the transition region, the Peelet numbers between McHenry's measured results and calculated
should be between the values for laminar flow and results if the inhomogeneity factor a is assumed to
values for fully developed turbulence (since there be unity.
are probably both types of flow in different pore
spaces). Eq. 24 defines an empirical turbulence EFFECT OF AN IMMOBILE PHASE
weighting factor,
The effect of an immobile gas phase on dispersion
has been studied by Orlob and Radhakrishna. 36
PeJRe =(I-v) (Pe)laminar+ These investigators found two important effects of
v (Pe)turbu lent . . . . . . . (24) immobile gas.
First, they found an unexpected reduction in
where Pe JRe average Peelet number at Reynolds
= flowable liquid. They coneluded that entrapment of
number Re. a small amount of gas in the pore spaces apparently
This factor should approach zero as the laminar traps some of the liquid in such a way that it is
region is approached and should approach unity as isolated from the main stream of flow and, henoe,
fully developed turbulence is approached. The not displaced during the passage of a tracer liquid.
factor v sketched on Fig. 23 has been estimated The volume of liquid so isolated varied considerably
from the dispersion data (in the transition region) with each individual set-up, ranging up to about 5
100 ASSUMPTIONS
d • • 0.2 CM. (10 MESH) STREAM SPLITTING WITH MASS
P • I GM/CC·. TRANSFER BY DIFFUSION

10
jJ. • I CPo
D• • 1.10'· CM.·/SEC. ..... --- ---
0:: , . . 0.7 ~TRANSYERSE PECLET
TURBULENCE
W / NUMBER
CT • 3.5 / TRANSITION FROM LAMINAR CONTROLS
en TO TURBULENT FLOW
.~ MIXING
DIFFUSION
:::> CONTROLS CELL
z THEORY

~
W DIFFUSION DOES
-.J NOT EQUALIZE
(.) 0.1 CONCENTRATION
W
a..
0.0 IIL..J_..Ji....LLllliLO-_L4Ll..lJLll,lLO-_L3.L.l..LW,lLO-_L2.L.LLlillLlO-_LI..l...L.l.llllL---l......l...L.llillLlO---l......l...L.l.LLUL--::-'-...L..LJ.J.J..I..lI..-~..J...J..J..u.w..~..J...J..J..u=10 5
o5 I
p
d Up = REYNOLDS NUMBER
p.
FIG. 24 - PECLET NUMBERS FOR AN AQUEOUS SYSTEM.

'MARCH, 1963 81
 100 ASSUMPTIONS:
d p -0.2 em. (10 MESH)
P • I. 2 x 10-' gm.lee.
~
a:
lJJ
CD
10
fL' 1.82
00 - 0.15 em.ftsee.

f4.I - 0.7
X 10-
4
POI SE

"
--- --- ---
"TRANSITION
FROM LAMINAR
TO TURBULENT FLOW
TURBULENCE
CONTROLS

~ (T • 3.5
::J DIFFUSION
Z CONTROLS

t-
lJJ
...J
u 0.1
lJJ
a..
0.01 •
10-" 10

• REYNOLDS NUMBER

FIG. 25 - PECLET NUMBERS FOR A GASEOUS SYSTEM (AIR).

per cent and averaging about 2 1/2 per cent of the
total pore volume.
Second, they found that dispersion in the liquid
phase was influenced by gas entrapment. The
amount of influence was dependent on the location
of the gas bubbles. If the gas bubbles were
located in the larger, more-easily-drained pores,
then the pore spaces effective in liquid transmission
were more uniform. This led to less dispersion
(apparently decreased u). On the other hand, a
small amount of trapped gas (say, less than 5 per
cent of the total pore volume) had little effect on
dispersion.
These authors conclude that the extent to which
dispersion is influenced by gas entrapment is
dependent on the location of gas bubbles and on
the structure of the porous medium. In systems
where gas enters only certain of the pore spaces
(thus making the medium more uniform), it is to be
expected that dispersion will decrease as gas is
accumulated. In general, they think that the more
nearly uniform the medium, the less will be the
influence of trapped gas.
ASSUMPTIONS:
dp'O.2 em.
1'- • 1.82 I 10. 4 POISE
Do: 0.15
0: ~ -0.7
"'m
~
(T',I
::>
z ; • '.
....
"'.,
..J
."'
The effect of an immobile wetting phase has been
studied by Raimondi 41 and is being studied by Fatt
and co-workers. 61
CONCLUSIONS
Porous media (both unconsolidated packs and
consolidated rocks) can be thought of as networks
of flow chambers, having random size and flow
conductivity, connected together by openings of
smaller size. The diffusion coefficient in such a
porous medium can be calculated by Eq. 7.
If fluids are flowing through the porous medium,
dispersion may be greater than that due to diffusion
alone. At moderate flow rates, the porous medium
will create a slightly asymmetrical mixed zone
(trailing edge stretched out) with the convective
dispersion coefficient approximately proportional
to the first power of average fluid velocity (if
composition is equalized in pore spaces by diffu-
sion). The longitudinal dispersion coefficient can
be calculated by Eq. 16 or from Fig. 10. The trans-
verse dispersion coefficient can be calculated by
Eq. 19 or from Fig. 13. For sandstones, the linear
dimension characterizing dispersion, (dpu), averages
about 0.36 cm.
There are also other variables that must be
properly controlled to get consistent results, viz.,
(1) edge effect in packed tubes, (2) particle size
distribution, (3) particle shape, (4) packing or
• • :-:. CCALCULATED
permeability heterogeneities, (5) viscosity ratios,
~,"ATA OF MCHENRy!3')
(6) gravity forces, (7) amount of turbulence, and
(8) effect of an immobile phase .
(LONGITUDINAL, GASEOUS SYSTEM)

NOMENCLATURE
0.1 L.l...J...ilillJJL-Ll...JJ.JWll-:-L..L.J.J.llllL::---l-LllLWL:;--'--L..J...l..llllJ
10 103 lOS
A = total cross sectional area, sq cm
REYNOLDS NUMBER. dpUp ,.
FIG. 26 - COMPARISON OF CALCULATED AND EX-
A ' = cross-sectional area open for flow or diffusion,
PERIMENTAL DATA (LONGITUDINAL DISPERSION:
GASEOUS SYSTEM). sq cm

82 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 a radius of a capillary, cm
C concentration
D apparent diffusion coefficient in a porous
medium based on the average area open for
diffusion and the over-all length
molecular diffusion coefficient, sq cm/sec
particle diameter, cm
tube diameter, cm
longitudinal convective dispersion coefficient,
sq cm/sec
F formation electrical resistivity factor
G quantity of material diffusing across a plane
(see Eq. 1)
g acceleration of gravity
Kl total longitudinal dispersion coefficient, sq
cm/sec
total transverse dispersion coefficient, sq cm
/sec
k permeability
L length of a tube or a porous block, cm
Pe Peclet number (see Eq. 23)
S slope of a log-normal particle size distribution
(see Eq. 21)
time, seconds
average interstitial velocity, cm/ sec
volume injected, cc
total volume of tube or total pore volume
distance, cm
x volume fraction of injected fluid in the effluent
(see Eq. 8)
factor to account for wall effect (see Fig. 14)
f3 t
(V/Vp ) - 1
A
V/Vp
f1 fluid viscosity
v turbulence weighting factor (see Fig. 23)
p fluid density
a a measure of the inhomogeneity of the porous
pack
porosity
sphericity of a particle (see Fig. 17)
REFERENCES
1. Aris, R.: "On the dispersion of a solute in a fluid
flowing through a tube", P,oc., Royal Soc. of London
(1956) Series A, Vol. 235, 67.
2. Aris, R.: "The Longitudinal Diffusion Coefficient in
Flow Through a Tube with Stagnant Pockets", Chem.
Eng. Sci. (1959) Vol. 11, 194.
3. Aris, R. and Amundson, N. R.: "Some Remarks on
Longitudinal Mixing or Diffusion in Fixed Beds",
AlChE JOUf'. (1957) Vol. 3, 280.
4. Bailey, H. R. and Gogarty, W. B.: "Diffusion Coeffi-
cients from Capillary Flow", Paper SPE-302, pre-
sented at SPE Production Research Symposium,
MARCH, 1963
Tulsa, Okla. (April 12-13, 1962).
5. Baron, Thomas: Chem. Eng. P,og. (1952) Vol. 48, 118.
6. Bernard, R. A. and Wilhelm, R. H.: "Turbulent Diffu-
sion in Fixed Beds of Packed Solids", Chem. Eng.
Pmg. (1950) Vol. 46,233.
7. Blackwell, R. J.: "Laboratory Studies of Microscopic
Dispersion Phenomena", Soc. Pet. Eng. Jou,. (March,
1962) 1.
8. Bosworth, R. C. L.: Phil. Mag. (1948) Vol. 39, 847.
9. Brigham, W. E., Reed, Philip W. and Dew, John N.:
"Experiments on Mixing During Miscible Displace-
ment in Porous Media", Soc. Pet. Eng. Jour. (March,
1961) 1.
10. Brown, G. G., et al: Unit Operations, John Wiley &
Sons, Inc., N. Y. (1950).
11. Carberry, J. J.: "Axial Dispersion of Mass in Flow
Through Fixed Beds", D. Eng. Dissertation, Yale U.
(1957).
12. Carberry, J. J.: "Axial Dispersion and Void-cell Mix-
ing Efficiency in Fluid Flow in Fixed Bt:ds", AIChE
Jour. (1958) Vol. 4, No.1, 13M.
13. Carman, P. C.: "Fluid Flow Through Granular Beds",
Trans., Insti. of Chern. Eng., London (1937) Vol. 15,
150.
14. Carman, P. C.: "Permeability of Saturated Sands,
Soils, and Clays", Jour. Agri. Sci. (1939) Vol. 29, 262.
15. Collins, R. E.: Flow of Fluids through Porous Media,
Reinhold Publishing Co., N. Y. (1961) 201.
16 .• Crank, J.: The Mathematics of Diffusion, Oxford at
the Clarendon Press, N. Y. (1957).
17. de J ong, G. de J 0 sselin: "Longitudinal and Trans-
verse Diffusion in Granular Deposits", Trans., AGU
(1958) Vol. 39, 67.
18. Deans, H. A. and Lapidus, Leon: "A Computational
Model for Predicting and Correlating the Behavior of
Fixed Bed Reactors: 1. Derivation of Model for Non-
reactive Systems", AIChE Jour. (Dec., 1960) 656.
19. Ebach, E. A.: "The Mixing of Liquids Flowing Through
Beds of Porous Solids", PhD Dissertation, U. of
Michigan (1957).
20. Ebach, E. A. and White, R. R.: "Mixing of Fluids
Through Beds of Packed Solids", AIChE Jour. (1958)
Vol. 6, 161.
21. Ergun, S. K.: "Fluid Flow Through Packed Columns",
Chem. Eng. Prog. (1952) Vol. 48, 89.
22. Fahien, R. W. and Smith, J. M.: "Mass Transfer in
Packed Beds", AIChE Jour. (1955) Vol. 1, 28.
23. Fatt, I.: "Pore Structure of Sintered Glass From
Diffusion and Resistance Measurements", Jour. Phys.
Chem. (1959) Vol. 63, 751.
24. Fowler, F. C. and Brown, G. G.: "Contamination by
Successive Flow in Pipe Lines", Trans., AIChE
(1943) Vol. 39, 491.
25. Goodknight, R. C., Klikoff, W. A. and Fatt, I.: "Non-
83
 Steady-State Fluid Flow and Diffusion in Porous Media
Containing Dead-End Pore Volume", Jour. Pbys.
Cbem. (1960) Vol. 64, 1162.
26. Grane, F. E. and Gardner, G. H. F.: "Measurements
of Transverse Dispersion in Granular Media", Jour.
Cbem. Eng. Data (1961) Vol. 6, 283.
27. Graton, L. C. and Fraser, H. J.: "Systematic Packing
of Spheres-With Particular Relation to Porosity and
Permeability", Jour. Geo'. (Nov.-Dec., 1935) 785.
28. Handy, L. L.: ~'An Evaluation of Diffusion Effects
in Miscible Displacement", Trans., AIME (1959) Vol.
216, 382.
29. Hoogschagen, Jan: "Diffusion in Porous Catalysts
and Adsorbents", Ind. Eng. Cbem. (1955) Vol. 47, 906.
30. Klinkenberg, L. J.: "Analog Between Diffusion and
Electrical Conductivity in Porous Rocks", Bull.,
GSA (1951) Vol. 62, 559.
31. Klirikenberg, A. and Sjenitzer, F.: "Holding-time
Distributi~ns of the Gaussian Type", Cbem. Eng. Sci.
(1956) Vol. 5, 258.
32. Koump, V.: "Study of the Mechanism of Axial Disper-
sion in Packed Beds at Low Flow Rates", D. Eng.
Dissertation, Yale U. (1959).
33. Latinen, G. A.: "Mechanism of Fluid Phase Mixing
of Fixed ~d Fluidized Beds of Uniformly Sized
Spherical Particles", PhD Dissertation, Princeton U.
(1951).
34. Liles, A. W. and Geankoplis, C. J.: "Axial Diffusion
of Liquids in Packed Beds and End Effects", AICbE
Jour. (1960) Vol. 6, 591.
35. McHenry, Keith W., Jr.: "Axial Mixing of Binary Gas
Mixtures Flowing in a Random Bed of Spheres", PhD
Dissertation, Princeton U. (1958).
36. Orlob, G. T. and Radhakrishna, G. N.: "The Effects
of Entrapped Gases on the Hydraulic Characteristics
of Porous Media", Trans., AGU (1958) Vol. 39, 648.
37. Penman, H. L.: "Gas and Vapor Movement in Soil",
Jour. Agri. Sci. (1940) Vol. 30, 438.
38. Plautz, D. A. and Johnstone, H. F.: "Heat and Mass
Transfer in Packed Beds", AICbE Jour. (1955) Vol.
1, 193.
39. Pozzi, A. L. and Blackwell, R. J.: "Design of Labor-
atory Models for Study of Miscible Displacement",
Soc. Pet. Eng. Jour. (March, 1963).
40. Prausnitz, J. M.: "Longitudinal Dispersion in a
Packed Bed", AICbE Jour. (1958) Vol. 4, No.1, 14M.
41. Raimondi, P., Torcaso, M. A. and Henderson, J. H.:
"The Effect of Interstitial Water on the Mixing of
Hydro<;:arbons During a Miscible Displacement Pro-
cess", Mi.. I,.d. Exp. Statio,.Circ. 61, ThePennsylvania
State U.
42. Raimondi, P., Gardner, G .. H. F. and Petrick, C. B.:
'''Effect of Pore Structure and Molecular Diffusion on
the Mixing of Miscible Liquids Flowing in Porous
84
Media", Preprint 43 presented at AIChE-SPE Joint
Symposium, San Francisco, Cal. (Dec. 6-9, 1959).
43. Ranz, W. E.: "Friction and Transfer Coefficients
for Single Particles and Packed Beds", Cbem. Eng.
Prog. (1952) Vol. 48, 247.
44. Rifai, M. N. E.: "An Investigation of Dispersion
Phenomena in Laminar Flow Through Porous Media",
PhD Dissertation, U. of California, Berkeley (1956).
45. Saf.fman, P. G.: "Dispersion in Flow Through a Net-
work of Capillaries", Cbem. Eng. Sci. (1959) Vol. 11,
125.
46. Saffman, P. G.: "Theory of Dispersion in Porous
Media", Jour. Fluid Mecb.(1959) Vol. 6, 321.
47. Saffman, P. G.: "Dispersion Due to Molecular Diffu-
sion Through a Network of Capillaries", Jour. Fluid
Mecb. (1960) Vol. 7, 194.
48. Scheidegger, A. E. and Larson, V. C.: "Asymmetry
of the Concentration Front During Miscible Displace~
ment in Porous Media", Can. Jour. Phys. (1958) Vol.
36, 1476.
49. Schwartz, C. E. and Smith, J. M.: "Flow Distribution
in Packed Beds", Ind. Eng. Prog. (1953) Vol. 45,
1209.
50. Scott, D. S. and Dullien, F. A. L.: "Diffusion of
Ideal Gases in Capillaries and Porous Solids", AICbE
Jour. (1962) Vol. 8 j 113.
51. Singer, E. and Wilhelm, R. H.: "Heat Transfer in
Packed Beds; Analytical Solution and Design Method;
Fluid Flow, Solids Flow and Chemical Reaction",
Cbern. Eng. Prog. (1950) Vol. 46, 343.
52. Taylor, G. I.: "Dispersion of Soluble Matter in Solvent
Flowing Slowly Through a Tube", Proc., Roy. Soc.
(1953) Vol. 219, 186. .
53. Terry, W. M., Blackwell, R. J. and Rayne, J. R.:
"Factors Influencing the Efficiency of Miscible Dis-
placement", Trans., AIME (1959) Vol. 216, 1.
54. Turner, G. A.: "The Flow-Structure in Packed Beds",
Cbem. E,.g. Sci. (1957) Vol. 7, 156.
55. van Bavel, C. H. M.: "Gaseous Diffusion and Porosity
in Porous Media", Soil Sci. (1952) Vol. 74, 91.
56. Van Deemter, J. J., Broeder, J. J. and Lauwerier, H.
A.: "Fluid Displacement in Capillaries", Appl.
Scientific Res. (1956) Vol. 5A, 374.
57. Van Deemter, J. J., BraIder and Lawrence: "Fluid
Displacement in Capillaries", Che",. E,.g. Sci. (1956)
Vol. 5, 271.
58. van der Poel, C.: "Effect of Lateral Diffusivity on
Miscible Displacement in Horizontal Reservoirs",
Soc. Pet. Eng. Jo~. (Dec., 1962) 317.
59. Von Rosenberg, D. U.: "Mechanics of Steady State
Single-phase Fluid Displacement from Porous Media",
AICbE Jour. (1956) Vol. 2, 55.
60. Zunker, F.: Blanck's Handbuck der Bodenlebre, VI,
Berlin (1930) 202.
61. ACS-PRF, 450-A Project.
***
SOCIETY OF PETROLEUM ENGINEERS JOURNAL.