Enhanced Oil Recovery

1
Defining Enhanced Oil

Recovery
Enhanced oil recovery (EOR) is oil recovery by the injection of materials not
normally present in the reservoir. This definition covers all modes of oil recovery
processes (drive, push-pull, and well treatments) and most oil recovery agents.
Enhanced oil recovery technologies are also being used for in-situ extraction of
organic pollutants from permeable media. In these applications, the extraction is
referred to as cleanup or remediation, and the hydrocarbon as product. Various
sections of this text will discuss remediation technologies specifically, although we
will mainly discuss petroleum reservoirs. The text will also describe the application
of EOR technology to carbon dioxide storage where appropriate.
The definition does not restrict EOR to a particular phase (primary,
secondary, or tertiary) in the producing life of a reservoir. Primary recovery is oil
recovery by natural drive mechanisms: solution gas, water influx, and gas cap drives,
or gravity drainage. Figure 1-1 illustrates. Secondary recovery refers to techniques,
such as gas or water injection, whose purpose is mainly to raise or maintain reservoir
pressure. Tertiary recovery is any technique applied after secondary recovery. Nearly
all EOR processes have been at least field tested as secondary displacements. Many
thermal methods are commercial in both primary and secondary modes. Much
interest has been focused on tertiary EOR, but the definition given here is not so
restricted. The definition does exclude waterflooding but is intended to exclude
all pressure maintenance processes. The distinction between pressure maintenance
2
and displacement is not clear, since some displacement occurs in all pressure
maintenance processes. Moreover, agents such as methane in a high-pressure gas
drive, or carbon dioxide in a reservoir with substantial native CO2, do not satisfy the
definition, yet both are clearly EOR processes. The same can be said of CO2 storage.
Usually the EOR cases that fall outside the definition are clearly classified by the
intent of the process.
In the last decade, improved oil recovery (IOR) has been used
interchangeably with EOR or even in place of it. Although there is no formal
definition, IOR typically refers to any process or practice that improves oil recovery
(Stosur et al., 2003). IOR therefore includes EOR processes but can also include
other practices such as waterflooding, pressure maintenance, infill drilling, and
horizontal wells.
Primary
Recovery
Artificial Lift
Natural F low
Pump - Gas Lift
Conventional
Recovery
Secondary
Recovery
Pressure Maintenance
Waterflood
Water/Gas Reinjection
Tertiary
Recovery
Enhanced
Thermal Chemical
Recovery
Solvent Other
Figure 1-1. Oil recovery classifications (adapted from the Oil and Gas Journal
biennial surveys).
1-1 EOR INTRODUCTION
The EOR Target
We are interested in EOR because of the amount of oil to which it is potentially

applicable. This EOR target oil is the amount unrecoverable by conventional means
(Fig. 1-1). A large body of statistics shows that conventional ultimate oil recovery
(the percentage of the original oil in place at the time for which further conventional
3
recovery becomes uneconomic) is about 35%. This means for example that a field
that originally contained 1 billion barrels will leave behind 650,000 barrels at the end
of its conventional life. Considering all of the reservoirs in the U.S., this value is
much larger than targets from exploration or increased drilling.
The ultimate recovery is shown in Fig. 1-2. This figure also shows that there
is enormous variability in ultimate recovery within a geographic region, which is why
we cannot target reservoirs with EOR by region. Reservoirs that have an
exceptionally large conventional recovery are not good tertiary EOR candidates.
Figure 1-2 shows also that the median ultimate recovery is the same for most regions,
a fact no doubt bolstered by the large variability within each region.
100
80
Ultimate Recovery Efficiency, %
60
40
20
0
Middle East CIS LatAm Africa Far East Europe Austral Asia US
Figure 1-2. Box plots of ultimate oil recovery efficiency. 75% of the ultimate
recoveries in a region fall within the vertical boxes; the median recovery is the
horizontal line in the box; the vertical lines give the range. Ultimate recovery is
highly variable, but the median is about the same everywhere (from Laherre, 2001).
1-2 THE NEED FOR EOR
Enhanced oil recovery is one of the technologies needed to maintain reserves.
Reserves
4
Reserves are petroleum (crude and condensate) recoverable from known reservoirs
under prevailing economics and technology. They are given by the following
material balance equation:
⎛ Production ⎞
⎛ Present ⎞ ⎛ Past ⎞ ⎛ Additions ⎞ ⎜ ⎟
⎜ ⎟=⎜ ⎟+⎜ ⎟ − ⎜ from ⎟
⎝ reserves ⎠ ⎝ reserves ⎠ ⎝ to reserves ⎠ ⎜ reserves ⎟
⎝ ⎠
There are actually several categories of reservoirs (proven, etc.) which distinctions
are very important to economic evaluation (Rose, 2001; Cronquist, 2001). Clearly,
reserves can change with time because the last two terms on the right do change with
time. It is in the best interests of producers to maintain reserves constant with time,
or even to have them increase.
Adding to Reserves
The four categories of adding to reserves are
1. Discovering new fields

2. Discovering new reservoirs
3. Extending reservoirs in known fields
4. Redefining reserves because of changes in economics of extraction
technology
We discuss category 4 in the remainder of this text. Here we substantiate its

importance by briefly discussing categories 1 to 3.
Reserves in categories 1 to 3 are added through drilling, historically the most
important way to add reserves. Given the 2% annual increase in world-wide
consumption and the already large consumption rate, it has become evident that
reserves can be maintained constant only by discovering large reservoirs.
But the discovery rate of large fields is declining. More importantly, the
discovery rate no longer depends strongly on the drilling rate. Equally important,
drilling requires a substantial capital investment even after a field is discovered. By
contrast, the majority of the capital investment for EOR has already been made (if
previous wells can be used). The location of the target field is known (no need to
explore), and targets tend to be close to existing markets.
Enhanced oil recovery is actually a competitor with conventional oil

recovery because most producers have assets or access to assets in all of the Fig. 1-1
categories. The competition then is joined largely on the basis of economics in
addition to reserve replacement. At the present, many EOR technologies are
competitive with drilling-based reserve additions. The key to economic
competitiveness is how much oil can be recovered with EOR, a topic to which we
next turn.
5
1-3 INCREMENTAL OIL
Defintion
A universal technical measure of the success of an EOR project is the amount of

incremental oil recovered. Figure 1-3 defines incremental oil. Imagine a field,
reservoir, or well whose oil rate is declining as from A to B. At B, an EOR project is
initiated and, if successful, the rate should show a deviation from the projected
decline at some time after B. Incremental oil is the difference between what was
actually recovered, B to D, and what would have been recovered had the process not
been initiated, B to C. Since areas under rate-time curves are amounts, this is the
shaded region in Fig. 1-3.
Figure 1-3. Incremental oil recovery from typical

EOR response (from Prats, 1982)
6
As simple as the concept in Fig. 1-3 is, EOR is difficult to determine in

practice. There are several reasons for this.
1. Combined (comingled) production from EOR and nonEOR wells. Such

production makes it difficult to allocate the EOR-produced oil to the EOR
project. Comingling occurs when, as is usually the case, the EOR project is
phased into a field undergoing other types of recovery.
2. Oil from other sources. Usually the EOR project has experienced substantial
well cleanup or other improvements before startup. The oil produced as a
result of such treatment is not easily differentiated from the EOR oil.
3. Inaccurate estimate of hypothetical decline. The curve from B to C in Fig. 1-
3 must be accurately estimated. But since it did not occur, there is no way of
assessing this accuracy.
Ways to infer incremental oil recovery from production data range from highly
sophisticated numerical models to graphical procedures. One of the latter, based on
decine curve analysis, is covered in the next section.
Estimating Incremental Oil Recovery Through Decline Curves
Decline curve analysis can be applied to virtually any hydrocarbon production

operation. The following is an abstraction of the practice as it applies to EOR. See
Walsh and Lake (2003) for more discussion. The objective is to derive relations
between oil rate and time, and then between cumulative production and rate.
The oil rate q changes with time t in a manner that defines a decline rate D
according to
1 dq
= −D 1.3-1
q dt
The rate has units of (or [=]) amount or volume per time and D [=]1/time. Time is in
units of days, months, or even years consistent with the units of q. D itself can be a
function of rate, but we take it to be constant. Integrating Eq. 1.3-1 gives
q = qi e− Dt 1.3-2
where qi is the initial rate or q evaluated at t = 0. Equation 1.3-2 suggests a

semilogarithmic relationship between rate and time as illustrated in Fig. 1-3.
Exponential decline is the most common type of analysis employed.
7
log (q)
qi
-D
Slope = 2.303
Decline
period
begins
q EL Life
t
0
Figure 1-3. Schematic of exponential decline on a rate-time plot.
Figure 1-3 schematically illustrates a set of data (points) which begin an

exponential decline at the ninth point where, by definition t = 0. The solid line
represents the fit of the decline curve model to the data points. qi is the rate given by
the model at t=0, not necessarily the measured rate at this point. The slope of the
model is the negative of the decline rate divided by 2.303, since standard semilog
graphs are plots of base 10 rather than natural logarithms.
Because the model is a straight line, it can be extrapolated to some future

rate. If we let qEL designate the economically limiting rate (simply the economic
limit) of the project under consideration, then where the model extrapolation attains
qEL is an estimate of the project’s (of well’s, etc.) economic life. The economic limit
is a nominal measure of the rate at which the revenues become equal to operating
expenses plus overhead. qEL can vary from a fraction to a few hundred barrels per
day depending on the operating conditions. It is also a function of the prevailing
economics: as oil price increases, qEL decreases, an important factor in reserve
considerations.
The rate-time analysis is useful, but the rate-cumulative curve is more

helpful. The cumulative oil produced is given by
ξ =t
Np = ∫ qdξ
ξ =0
.
8
The definition in this equation is general and will be employed throughout the text,
but especially in Chap. 2. To derive a rate -cumulative expression, insert Eq. 1.3-1,
integrate, and identify the resulting terms with (again) Eq. 1.3-1. This gives
q = qi − DN p 1.3-3
Equation 1.3-3 says that a plot of oil rate versus cumulative production should be a
straight line on linear coordinates. Figure 1-4 illustrates.
q
qi
Slope = -D
Mobile oil
q EL
Recoverable oil
Np
0
Figure 1-4. Schematic of exponential decline on a rate-cumulative plot.
You should note that the cumulative oil points being plotted on the horizontal axis of
this figure are from the oil rate data, not the decline curve. It this were not so, there
would be no additional information in the rate-cumulative plot. Calculating Np
normally requires a numerical integration with something like the trapezoid rule.
Using model Eqs 1.3-2 and 1.3-3 to interpret a set of data as illustrated in
Figs. 1-3 and 1-4 is the essence of reservoir engineering practice, namely
1. Develop a model as we have done to arrive at Eqs. 1.3-2 and 1.3-3. Often the
model equations are far more complicated than these, but the method is the same
regardless of the model.
2. Fit the model to the data. Remember that the points in Figs. 1-3 and 1-4 are data.
The lines are the model.
3. With the model fit to the data (the model is now calibrated), extrapolate the model
to make predictions.
9
At the onset of the decline period, the data again start to follow a straight line
through which can be fit a linear model. In effect, what has occurred with this plot is
that we have replaced time on Fig. 1-3 with cumulative oil produced on Fig. 1-4, but
there is one very important distinction: both axes in Fig. 1-4 are now linear. This has
three important consequences.
1. The slope of the model is now –D since no correction for log scales is
required.
2. The origin of the model can be shifted in either direction by simple additions.
3. The rate can now be extrapolated to zero.
Point 2 simply means that we can plot the cumulative oil produced for all
periods prior to the decline curve period (or for previous decline curve periods) on
the same rate-cumulative plot. Point 3 means that we can extrapolate the model to
find the total mobile oil (when the rate is zero) rather than just the recoverable oil
(when the rate is at the economic limit).
Rate-cumulative plots are simple yet informative tools for interpreting EOR
processes because they allow estimates of incremental oil recovery (IOR) by
distinguishing between recoverable and mobile oil. We illustrate how this comes
about through some idealized cases.
Figure 1-5 shows a rate-cumulative plot for a project having an exponential

decline just prior to and immediately after the initiation of an EOR process.
q EL
IOR
Np
Incremental
Project begins mobile oil
10
Figure 1-5. Schematic of exponential decline curve behavior on a rate-cumulative
plot. The EOR project produces both incremental oil (IOR), and increases the mobile
oil. The pre- and post-EOR decline rates are the same.
We have replaced the data points with the models only for ease of
presentation. Placing both periods on the same horizontal axis is permissible because
of the scaling arguments mentioned above. In this case, the EOR process did not
accelerate the production because the decline rates in both periods are the same;
however, the process did increase the amount of mobile oil, which in turn caused
some incremental oil production. In this case, the incremental recovery and mobile
oil are the same. Such idealized behavior would be characteristic of thermal,
micellar-polymer, and solvent processes.
IOR
q EL
Np
Project begins
Figure 1-6. Schematic of exponential decline curve behavior on a rate-cumulative

plot. The EOR project produces incremental oil at the indicated economic limit but
does not increase the mobile oil.
Figure 1-6 shows another extreme where production is only accelerated, the
pre- and post-EOR decline rates being different. Now the curves extrapolate to a
common mobile oil but with still a nonzero IOR. We expect correctly that processes
that behave as this will produce less oil than ones that increase mobile oil, but they
can still be profitable, particularly, if the agent used to bring about this result is
inexpensive. Processes that ideally behave in this manner are polymer floods and
polymer gel processes, which do not affect residual oil saturation. Acceleration
processes are especially sensitive to the economic limit; large economic limits imply
large IOR.
11
Example 1-1. Estimating incremental oil recovery.
Sometimes estimating IOR can be fairly subtle as this example illustrates. Figure 1-7
shows a portion of rate-cumulative data from a field that started EOR about half-way
through the total production shown.
0.20
Monthly Rate, M std. m3/month
0.15
Pre EOR
0.10
Post EOR
0.05
qEL
0.00
0.0
2.0 1.0
3.0 4.0 5.0
Cumulative Oil Produced, M std. m3
Figure 1-7. Rate (vertical axis) - cumulative (horizontal axis) plot for a field
undergoing and EOR process.
a. Identify the pre- and post-EOR decline periods.

The pre-EOR decline ends at about 2.5 M std. m3 of oil produced, at which time the
post-EOR period begins. This point does not necessarily coincide with the start of
the EOR process. The start cannot be inferred from the rate-cumulative plot.
b. Calculate the decline rates ([=] mo-1) for both periods.

Both decline periods are fitted by the straight lines indicated. The fitting is done
through standard means; the difficulty is always identifying when the periods start
and end. For the pre-EOR decline,
⎛ M std.m3 ⎞
⎜ ( 0.11 − 0.18 ) ⎟
D = −⎜ month ⎟ = 0.027 month
−1
⎜ ( 2.55 − 0 ) M std.m
3
⎟
⎜ ⎟
⎝ ⎠
and for the post-EOR decline,
12
⎛ M std.m 3
⎞
⎜ ( 0.09 − 0.11) ⎟
D = −⎜ month ⎟ = 0.0137 month
−1
⎜ ( 4 − 2.55 ) M std.m
3
⎟
⎜ ⎟
⎝ ⎠
The EOR project has about halved the decline rate even though there is no increase in
rate.
c. Estimate the IOR ([=] M std. m3) for this project at the indicated economic limit.
The oil to be recovered by continued operations is 4.7 M std. m3. That from EOR is
(by extrapolation) 7 M std. m3 for an incremental oil recovery of 2.3 M std. m3.
1-4 CATEGORY COMPARISONS
Comparative Performances
Most of this text covers the details of EOR processes. At this point, we compare
performances of the three basic EOR processes and introduce some issues to be
discussed later in the form of screening guides. The performance is represented as
typical oil recoveries (incremental oil expressed as a percent of original oil in place)
and by various utilization factors. Both are based on actual experience. Utilization
factors express the amount of an EOR agent required to produce a barrel of
incremental oil. They are a rough measure of process profitability.
Table 1-1 shows sensitivity to high salinities is common to all chemical
flooding EOR. Total dissolved solids should be less than 100,000 g/m3, and hardness
should be less than 2,000 g/m3. Chemical agents are also susceptible to loss through
rock–fluid interactions. Maintaining adequate injectivity is a persistent issue with
chemical methods. Historical oil recoveries have ranged from small to moderately
large. Chemical utilization factors have meaning only when compared to the costs of
the individual agents; polymer, for example, is usually three to four times as
expensive (per unit mass) as surfactants.
TABLE 1-1 CHEMICAL EOR PROCESSES

Recovery Typical Typical agent
Process mechanism Issues recovery (%) utilization*
Polymer Improves volumetric Injectivity 5 0.3–0.5 lb polymer

sweep by mobility Stability per bbl oil produced
reduction High salinity
Micellar Same as polymer plus Same as polymer 15 15–25 lb surfactant
polymer reduces capillary plus chemical per bbl oil produced
forces availability,
retention, and
high salinity
13
Alkaline Same as micellar Same as micellar 5 35–45 lb chemical
polymer polymer plus oil polymer plus oil per bbl oil produced
solubilization composition
and wettability
alteration
*1 lb/bbl ≅ 2.86 kg/m3
Table 1-2 shows a similar comparison for thermal processes. Recoveries are
generally higher for these processes than for the chemical methods. Again, the issues
are similar within a given category, centering on heat losses, override, and air
pollution. Air pollution occurs because steam is usually generated by burning a
TABLE 1-2 THERMAL EOR PROCESSES

Steam Reduces oil Depth 50–65 0.5 bbl oil consumed

(drive and viscosity Heat losses per bbl oil
stimulation) Vaporization Override produced
of light ends Pollution
In situ Same as steam Same as steam plus 10–15 10 Mscf air per bbl oil
combustion plus cracking control of produced*
combustion
3 3
*1 Mscf/stb ≅ 178std. m gas/std. m oil
portion of the resident oil. If this burning occurs on the surface, the emission products
contribute to air pollution; if the burning is in situ, production wells can be a source
of pollutants.
Table 1-3 compares solvent flooding processes. Only two groups are in this
category, corresponding to whether or not the solvent develops miscibility with the
oil. Oil recoveries are generally lower than for micellar-polymer recoveries. The
solvent utilization factors as well as the relatively low cost of the solvents have
brought these processes, particularly carbon dioxide flooding, to commercial
application. The distinction between a miscible and an immiscible process is slight.
TABLE 1-3 SOLVENT EOR METHODS

Immiscible Reduces oil Stability 5–15 10 Mscf solvent per

viscosity Override bbl oil produced
Oil swelling Supply
Solution gas
Miscible Same as immiscible Same as immiscible 5–10 10 Mscf solvent per
plus development bbl oil produced
of miscible
14
displacement
3 3
*1 Mscf/stb ≅ 178 std. m solvent/ std. m oil
Screening Guides
Many of the issues in Tables 1-1 through 1-3 can be better illustrated by giving
quantitative limits. These screening guides can also serve as a first approximate for
when a process would apply to a given reservoir. Table 1-4 gives screening guides of
EOR processes in terms of oil and reservoir properties.
TABLE 1-4. SUMMARY OF SCREENING CRITERIA FOR EOR METHODS

(adapted from Taber et al., 1997).
TABLE 3: SUMMARY OF SCREENING CRITERIA FOR EOR METHODS
Oil Properties Reservoir Characteristics
Reservoir Initial Oil Net Average
Gravity Viscosity Saturation Formation Thickness Permeability
EOR Method (ºAPI) mPa-s Compostion (%PV) Type (m) (md) Depth (m)
Solvent Methods
Nitrogen and Large % of
flue gas >35 <0.4 C1 to C7 >40 NC NC NC >1800
Large % of
Hydrocarbon >23 <3 C2 to C7 >30 NC NC NC >1250
Large % of
C02 >22 <10 C5 to C12 >20 NC NC NC >750
Immiscible
gases >12 <600 NC >35 NC NC NC >640
Chemical Methods
Light,
Miscellar/ intermediate,
polymer, some organic
ASP, and acids for
alkaline alkaline Sandstone
flooding >20 <35 floods >35 preferred NC >10 <2700
Polymer Sandstone
Flooding >15 10-150 NC >50 preferred NC >10 <2700
Thermal Methods
Some
asphaltic
Combustion >10 <5,000 components >50 >3 >50 <3450
>8 to
Steam 13.5 <200,000 NC >40 >6 >200 <1350
NC=not critical
These should be regarded as rough guidelines, not as hard limits because special
circumstances (economics, gas supply for example) can extend the applications.
The limits have a physical base as we will see. For example, the restriction
of thermal processes to relatively shallow reservoirs is because of potential heat
losses through lengthy wellbores. The restriction on many of the processes to light
crudes comes about because of sweep efficiency considerations; displacing viscous
15
oil is difficult because of the propensity for a displacing agent to channel through the
fluid being recovered. Finally, you should realize that some of categorizations in
Table 1-7 are fairly coarse. Steam methods, in particular, have additional divisions
into steam soak, steam drive, and gravity drainage methods. There are likewise
several variations of combustion and chemical methods.
1-5 UNITS AND NOTATION
SI Units
The basic set of units in the text is the System International (SI) system. We cannot
be entirely rigorous about SI units because many figures and tables has been
developed in more traditional units. It is impractical to convert these; therefore, we
give a list of the more important conversions in Table 1-7 and some helpful pointers
in this section.
TABLE 1-5 AN ABRIDGED SI UNITS GUIDE (adapted from Campbell et al.., 1977)
SI base quantities and units
SPE dimensions
Base quantity or SI unit symbol symbol
dimension SI unit
Length Meter m L
Mass Kilogram kg m
Time Second S t
Thermodynamic temperature Kelvin K T
Amount of substance Mole* mol
*When the mole is used, the elementary entities must be specified; they may be atoms, molecules, ions,
electrons, other particles, or specified groups of such particles in petroleum work. The terms kilogram
mole, pound mole, and so on are often erroneously shortened to mole.
Some common SI derived units
SI unit symbol Formula

Quantity Unit
Acceleration Meter per second squared –– m/s2

Area Square meter –– m2
Density Kilogram per cubic meter –– kg/m3
Energy, work Joule J N·m
Force Newton N kg · m/s2
Pressure Pascal Pa N/m2
Velocity Meter per second –– m/s
Viscosity, dynamic Pascal-second –– Pa · s
Viscosity, kinematic Square meter per second –– m2/s
Volume Cubic meter –– m3
Selected conversion factors

16
To convert from To Multiply by
2 2
Acre (U.S. survey) Meter (m ) 4.046 872 E+03
Acres Feet2 (ft2) 4.356 000 E+04
Atmosphere (standard) Pascal (Pa) 1.013 250 E+05
Bar Pascal (Pa) 1.000 000 E+05
Barrel (for petroleum 42 gal) Meter3 (m3) 1.589 873 E–01
Barrel Feet3 (ft3) 5.615 E+00
British thermal unit (International Table) Joule (J) 1.055 056 E+03
Darcy Meter2 (m2) 9.869 232 E–13
Day (mean solar) Second (s) 8.640 000 E+04
Dyne Newton (N) 1.000 000 E–05
Gallon (U.S. liquid) Meter3 (m3) 3.785 412 E–03
Gram Kilogram (kg) 1.000 000 E–03
Hectare Meter2 (m2) 1.000 000 E+04
Mile (U.S. survey) Meter (m) 1.609 347 E+03
Pound (lbm avoirdupois) Kilogram (kg) 4.535 924 E–01
Ton (short, 2000 lbm) Kilogram (kg) 9.071 847 E+02
TABLE 1-5 CONTINUED
Selected SI unit prefixes
SI prefix
Meaning outside
SI symbol
US
Factor prefix (use roman type) Meaning (U.S.)
1012 tera T One trillion times Billion

109 giga G One billion times Milliard
106 mega M One million times
103 kilo k One thousand times
102 hecto H One hundred times
10 deka Da Ten times
10–1 deci D One tenth of
10–2 centi c One hundredth of
10–3 milli m One thousandth of
10–6 micro μ One millionth of
10–9 nano N One billionth of Milliardth
1. There are several cognates, quantities having the exact or approximate

numerical value, between SI and practical units. The most useful for EOR are
1 cp = 1 mPa-s
1 dyne/cm = 1 mN/m
1 Btu ≅ 1 kJ
1 Darcy ≅ 1 μm2
1 ppm ≅ 1 g/m3
17
2. Use of the unit prefixes (lower part of Table 1-5) requires care. When a
prefixed unit is exponentiated, the exponent applies to the prefix as well as
the unit. Thus 1 km2 = 1(km)2 = 1(103 m)2 = 1 × 106m2. We have already
used this convention where 1 μm2 = 10–12 m2 ≅ 1 Darcy.
3. Two troublesome conversions are between pressure (147 psia ≅ 1 MPa) and
temperature (1 K = 1.8 oR). Since neither the Fahrenheit nor the Celsius scale
is absolute, an additional translation is required.
°C = K – 273
and
°F = °R – 460
The superscript ° is not used on the Kelvin scale.
4. The volume conversions are complicated by the interchangeable use of mass

and standard volumes. Thus we have
0.159 m3 = 1 reservoir barrel, or bbl

and
0.159 std. m3 = 1 standard barrel, or stb
The standard cubic meter, std. m3, is not standard SI; it represents the amount
of mass contained in one cubic meter evaluated at standard temperature and
pressure.
Consistency
Maintaining unit consistency is important in all exercises, and for this reason both
units and numerical values should be carried in all calculations. This ensures that the
unit conversions are done correctly and indicates if the calculation procedure itself is
appropriate. In maintaining consistency, three steps are required.
1. Clear all unit prefixes.

2. Reduce all units to the most primitive level necessary. For many cases, this
will mean reverting to the fundamental units given in Table 1-7.
3. After calculations are complete, reincorporate the unit prefixes so that the
numerical value of the result is as close to 1 as possible. Many adopt the
convention that only the prefixes representing multiples of 1,000 are used.
Example 1-2. Converting from Darcy units.

18
Maintaining unit consistency in an equation is easy. For example, suppose we want to
use the typical oilfield units in Darcy’s law: q in units of ([=]) bbl/day; k [=] md; A
[=] ft2; p [=] psia; μ [=] cp; and x [=] ft. First we write Darcy’s law:
kA dp
q=
μ dx
This is elementary form of Darcy's law is valid for 1-D horizontal flow Darcy's law is
self-consistent in so-called Darcy’s units; hence, a "units" balance for this equation is
⎛ q − cm3 ⎞ ( k − D ) ( A − cm ) ⎛ dp − atm ⎞
2
⎜ ⎟= ⎜ ⎟
⎝ s ⎠ ( μ − cp ) ⎝ dx − cm ⎠
where k-D means that the permeability k is in D or Darcys. The other units given in
the equation are Darcy units. Note that the minus sign is unnecessary since we are
dealing only with units. Next, we write this same equation into the units that we
want , maintaining the unit consistency. That is,
⎡ q − bbl ⎫⎪ ⎧ 1 hr ⎫ ⎧ 1 hr ⎫ ⎧⎪ ( 30.48 )3 cm3 ⎫⎪
⎤ ⎧⎪ 1 day
⎢ ⎥⎨ ⎬⎨ ⎬⎨ ⎬⎨ ⎬=
⎢⎣ day ⎥⎦ ⎪⎩ 24 hrs
⎪⎭ ⎩ 3600 s ⎭ ⎩ 3600 s ⎭ ⎪⎩ ft 3 ⎪⎭
⎧ 2 ⎫
2 ⎤ ⎪ ( 30.48 ) cm ⎪
2
⎧ 1 D ⎫⎡
⎡ k − md ⎤ ⎨ ⎬ A − ft ⎨ ⎬
⎣ ⎦ 1000 md ⎣ ⎦
⎩ ⎭ ⎩⎪ ft 2 ⎭⎪ ×
[ μ − cp ]
⎡ dp − psia ⎤ ⎧⎪ 1 atm ⎫⎪ ⎧⎪ 1 ft ⎫⎪
⎢ ⎥⎨ ⎬⎨ ⎬
⎣⎢ dx − ft ⎦⎥ ⎩⎪14.70 psia ⎭⎪ ⎩⎪ 30.48 cm ⎭⎪
Although each term is written in the units we wish, each term reduces to the units of
the original equation. This is illustrated in the above equation by canceling all
similar units. By writing the above equation and checking the unit consistency, you
are assured of making no errors. The equation also introduces the practice of putting
ratios that are conversion factors in {}.
The last step is to rewrite the equation by grouping all numerical constants
and calculating the appropriate constant that must appear before the right side of the
equation. Darcy’s law becomes
⎛ q − bbl ⎞ ⎧⎪ ( 24 )( 3600 ) ( 30.482 ) ⎫⎪ ( k − md ) ( A − ft 2 ) ⎛ dp − psia ⎞

⎜ =
⎟ ⎨ ⎬ ⎜ ⎟
⎝ day ⎠ ⎪⎩ ( 5.615 ) ( 30.48 ) (1000 )(14.70 )( 30.48 ) ⎪⎭ ( μ − cp ) ⎝ dx − ft ⎠
3
or
19
⎛ q − bbl ⎞ ⎧⎪ −3
( k − md ) ( A − ft 2
) ⎫⎪ ⎛ dp − psia ⎞ .
⎜ ⎟ = ⎨1.127 × 10 ⎬⎜ ⎟
⎝ day ⎠ ⎪⎩ ( μ − cp ) ⎪⎭ ⎝ dx − ft ⎠
The constant, which is accurate to four digits, is the well-known constant for Darcy’s
law written in oil field units. The above equation also illustrates a common practice
in petroleum engineering--in our opinion bad and used sparingly in this text--of
including a conversion factor directly in an equation.
The important point the above procedure is that there is no guessing
involved. Any equation can be converted to the desired units as long as the
procedure is followed exactly.
Naming Conventions
The diversity of EOR makes it possible to assign symbols to components without

some duplication or undue complication. In the hope of minimizing the latter by
adding a little of the former, Table 1-8 gives the naming conventions of phases and
components used throughout this text. The nomenclature section defines other
symbols.
Phase always carry the subscript j, which occupies the second position in a
doubly subscripted quantity. j = 1 is always a water-rich, or the aqueous phase, thus
freeing up the symbol w for wetting (and nw for nonwetting). The subscript s
designates the solid, nonflowing phase.
A subscript i, occurring in the first position, indicates the component. Singly
subscripted quantities indicate components. In general, i = 1 is always water; i = 2 is
oil or hydrocarbon; and i = 3 refers to a displacing component, whether surfactant or
light hydrocarbon. Component indices greater than 3 are used exclusively in Chaps.
8–10, the chemical flooding part of the text.
1-6 SUMMARY
No summary can do justice to what is a large, diverse, continuously changing, and
complicated technology. The Oil and Gas Journal has provided an excellent service
in documenting the progress of EOR, and you should consult those surveys for up to
date information. The fundamentals of the processes change more slowly than the
applications, and it is to these fundamentals that the remainder of the text is devoted.
20
TABLE 1-4 NAMING CONVENTIONS FOR PHASES AND COMPONENTS
Phases
Text
j Identity
locations
1 Water-rich or aqueous Throughout
2 Oil-rich or oleic Throughout
3 Gas-rich, gaseous or light hydrocarbon Secs. 5-6 and 7-7
Microemulsion Chap. 9
s Solid Chaps 2, 3, and 8 to10
w Wetting Throughout
nw Nonwetting Throughout
Components
Text
i Identity
locations
1 Water Throughout
2 Oil or intermediate
hydrocarbon Throughout
3 Gas Sec. 5-6
Light hydrocarbon Sec. 7-6
Surfactant Chap. 9
4 Polymer Chaps. 8 and 9
5 Anions Secs. 3-4 and 9-5
6 Divalents Secs. 3-4 and 9-5
7 Divalent-surfactant
component Sec. 9-6
8 Monovalents Secs. 3-4 and 9-5
EXERCISES
1A. Determining Incremental Oil Production. The easiest way to estimate incremental oil
recovery IOR is through decline curve analysis, which is the subject of this exercise. The oil rate and
cumulative oil produced versus time data for the Sage Spring Creek Unit A field is shown below (Mack
and Warren, 1984)
Date Oil Rate std. m3/day

1/76 274.0
7/76 258.1
1/77 231.0
7/77 213.5
1/78 191.2
7/78 175.2 (Start Polymer)
1/79 159.3
7/79 175.2
1/80 167.3
7/80 159.3
1/81 159.3
7/81 157.7
21
1/82 151.3
7/82 148.2
1/83 141.8
7/83 132.2
1/84 111.5
7/84 106.7
1/85 95.6
7/85 87.6
1/86 81.2
7/86 74.9
1/87 70.1
7/87 65.3
In 7/78 the ongoing waterflood was replaced with a polymer flood. (Actually, there was a polymer gel
treatment conducted in 1984, but we neglect it here.) The economic limit is 50 std. m3/D in this field.
(a) Plot the oil rate versus cumulative oil produced on linear axes. The oil rate axis should extend to q
= 0.
(b) Extrapolate the straight line portion of the data to determine the ultimate economic oil to be
recovered from the field and the total mobile oil, both in Mstd. m3, for both the water and the
polymer flood. Determine the incremental economic oil (IOR) and the incremental mobile oil
caused by the polymer flood.
(c) Determine the decline rates appropriate for the waterflood and polymer flood declines.
(d) Use the decline rates in step c to determine the economic life of the polymer flood. Also determine
what the economic life would have been if there were no polymer flood.
1B. Maintaining Unit Conversions (Darcy’s Law). There are several unit systems used
throughout the world and you should be able to convert equations easily between systems. Convert
Darcy’s Law for 1-D horizontal flow,
kA dp
q=
μ dx
from Darcy units to the unit system where q [=] m3/day, k [=] md, A[=] m2, μ [=] cp, p [=] kgf/cm2, and
x [=] meters. This is the reverse of that in Example 1-2.
1C. Maintaining Unit Conversions (Dimensionless Time). A dimensionless time often appears
in petroleum engineering. One definition for dimensionless time used in radial flow is
kt
tD =
φμ ct rw2
where the equation is written in Darcy units ( rw [=] cm, φ is dimensionless, ct [=] of atm-1). Convert
the equation for dimensionless time from Darcy units to
(a) oil-field units.
(b) SI units.
This means write the equation with a conversion factor in it so that quantities with the indicated units
may be substituted directly.
22
1D. Maintaining Unit Conversions (Dimensionless Pressure). A dimensionless pressure often
appears in petroleum engineering. One definition for dimensionless pressure is
2π khΔp
pD =
qμ
where the equation is written in Darcy units (h in cm, Δp in atm). Convert the equation for
dimensionless pressure from Darcy units to
(a) oil-field units.
(b) SI units.
1
Basic Equations
for Fluid Flow
in Permeable Media
Successful enhanced oil recovery requires knowledge of equal parts chemistry,
physics, geology and engineering. Each of these enters our understanding through
elements of the equations that describe flow through permeable media. Each EOR
process involves at least one flowing phase that may contain several components.
Moreover, because of varying temperature, pressure, and composition, these
components may mix completely in some regions of the flow domain, causing the
disappearance of a phase in those regions. Atmospheric pollution and chemical and
nuclear waste storage lead to similar problems.
This chapter gives the equations that describe multiphase, multicomponent
fluid flow through permeable media based on conservation laws and linear
constitutive theory. Initially, we strive for the most generality possible by
considering the transport of each component in each phase. Then, special cases are
obtained from the general equations by making additional assumptions. The approach
in arriving at the special equations is as important as the equations themselves, since
it will help to understand the specific assumptions--and the limitations--that are being
made for a particular application.
The formulation initially contains two fundamentally different forms for the
general equations: overall compositional balances, and the phase conservation
equations. The overall compositional balances are useful for modeling how
components are transported through permeable media in local thermodynamic
equilibrium. The phase conservation equations are useful for modeling finite mass
transfer among phases. Figure 2-1 illustrates the relationships among several
equations developed as special cases in this chapter.
From the overall compositional balances, the list of special cases includes the
multicomponent, single-phase flow equations (Bear, 1972) and the three-phase,
2
multicomponent equations (Crichlow, 1977; Peaceman, 1977; Coats, 1980). In
addition, others (Todd and Chase, 1979; Fleming et al., 1981; Larson, 1979) have
presented multicomponent, multiphase formulations for flow in permeable media but
with assumptions such as ideal mixing or incompressible fluids. Many of these
assumptions must be made before the equations are solved, but we try to keep the
formulation as general as possible as long as possible.
Figure 2-1 Flow diagram showing the relationships among the fundamental equations
and selected special cases. There are NC components and NP phases.
2-1 MASS CONSERVATION
This section describes the conceptual nature of multiphase, multicomponent flows

through permeable media and the mathematical formulation of the conservation
equations.
3
The four most important mechanisms causing transport of chemical
components in naturally occurring permeable media are viscous forces, gravity
forces, dispersion (diffusion), and capillary forces. The driving forces for the first
three are pressure, density, and concentration gradients, respectively. Capillary or
surface forces are caused by high-curvature boundaries between the various
homogeneous phases. This curvature is the result of such phases being constrained by
the pore walls of the permeable medium. Capillary forces imply differing pressures in
each homogeneous fluid phase so that the driving force for capillary pressure is, like
viscous forces, pressure differences.
The ratios of these forces are often given as dimensionless groups and given
particular names. For example, the ratio of gravity to capillary forces is the Bond
number. When capillary forces are small compared to gravity forces, the Bond
number is large and the process (or displacement) is said to be gravity dominated.
The ratio of viscous to capillary forces is the capillary number, a quantity that will
figure prominently through this text. The ratio of gravity to viscous forces is the
gravity or buoyancy number. The magnitude of these and other dimensionless
groups help in comparing or scaling one process to another; they will appear at
various points throughout this text.
The Continuum Assumption
Transport of chemical components in multiple homogeneous phases occurs because

of the above forces, the flow being restricted to the highly irregular flow channels
within the permeable medium. The conservation equations for each component apply
at each point in the medium, including the solid phase. In principle, given
constitutive relations, reaction rates, and boundary conditions, it is possible to
formulate a mathematical system for all flow channels in the medium. But the phase
boundaries in such are extremely tortuous and their locations are unknown; hence, we
cannot solve component conservation equations in individual channels except for
only the simplest microscopic permeable media geometry.
The practical way of avoiding this difficulty is to apply a continuum
definition to the flow so that a point within a permeable medium is associated with a
representative elementary volume (REV), a volume that is large with respect to the
pore dimensions of the solid phase but small compared to the dimensions of the
permeable medium. The REV is defined as a volume below which local fluctuations
in some primary property of the permeable medium, usually the porosity, become
large (Bear, 1972). A volume-averaged form of the component conservation
equations applies for each REV within the now-continuous domain of the
macroscopic permeable medium. (Volume averaging is actually a formal process; see
Bear, 1972; Gray, 1975; and Quintard and Whitaker, 1988.) The volume-averaged
component conservation equations are identical to the conservation equations outside
a permeable medium except for altered definitions for the accumulation, flux, and
source terms. These definitions now include permeable media porosity, permeability,
tortuosity, and dispersivity, all made locally smooth because of the definition of the
4
REV. Approximating the locally discontinuous permeable medium with a locally
smooth one is called the continuum assumption.
A good way to understand the REV scale is to consider a microscopic view
of pores and grains within a medium. Figure 2-2 illustrates a cube of small volume
placed within a permeable medium. The porosity in the cube is defined as the pore
volume within the cube divided by the bulk volume of the cube. If the cube volume
is infinitesimally small, the porosity will be either 1.0 or 0.0 depending on whether it
is initially located inside a grain or a pore. We now let the cube increase from its
original size. As the cube volume increases, the porosity changes in an erratic fashion
as more and more grains and pores pass inside the cube (see Figure 2-3).
Figure 2-2 Illustration of microscopic cube placed within a permeable medium. The
cube is initially within a pore so that its porosity is 1.0. As the cube volume
increases, it takes in more grains so that its porosity decreases.
As the cube volume increases sufficiently, the porosity approaches a constant

value representative of the porous medium. This is the porosity at the REV scale,
which defines the onset of the permeable medium domain. Above the REV size, the
cube porosity remains constant within the domain of the permeable medium. As the
cube volume increases further, however, the cube porosity is affected by layering and
other heterogeneities. The formation is homogeneous if the porosity remains fixed as
the REV-sized cube is moved to any location within the permeable medium. A
heterogeneous formation is one in which the porosity (or any other petrophysical
property) varies from one spatial location to another when measured at the REV
scale.
5
Figure 2-3 Idealization of the microscopic and permeable medium domains. The
REV size separates these two domains.
Rather than beginning with the nonpermeable media flow equations, and then
volume-averaging over the REV, we invoke the continuum assumption at the outset
and derive the mass conservation on this basis. This approach skips over many of the
physical insights obtained from volume averaging, but it is far more direct.
Mass Balance for a Component in a Phase
Consider an arbitrary, fixed volume V embedded within a permeable medium through

which is flowing an arbitrary number of chemical components and phases. You must
constantly be aware of the distinction between components and phases in this
discussion. (This can be a source of confusion because under some circumstances
phases and components are the same.) A component is any identifiable chemical
entity. Components can be pure substances such as methane, a cation, or even
combinations of elements. See Lake et al., 2002 for more discussion. A phase is a
physically distinct part of a region in space that is bound by interfaces with
macroscopic physical properties, such as density and viscosity. A phase can consist
of many components. There are up to i = 1, . . . , NC components, and up to j = 1, . .
. , NP phases. The "up to" is because both components and phases can vanish in
regions of the flow domain.
The conservation laws are written over a control volume V that is greater
than or equal to the REV but smaller than or equal to the permeable medium
dimensions. Except for some restrictions on the connectivity of the surface, V can be
quite general. As Fig. 2-4 (a) shows, the surface area A
6
Figure 2-4 Geometries for conservation

law derivations.
7
of V is made up of elemental surface areas ΔA from the center of which is pointing a
unit outward normal vector n. The sum of all the surface elements ΔA is the total
surface area A of V. This sum of all the ΔA becomes the total surface area A as the
largest ΔA approaches zero.
The conservation equation for each component in each phase in volume V is
⎧ Rate of ⎫ ⎧ Net rate of i ⎫ ⎧ Net rate of ⎫

⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎨accumulation of i ⎬ = ⎨ in phase j ⎬ + ⎨ generation of i ⎬
⎪ in phase j in V ⎪ ⎪ transported into V ⎪ ⎪in phase j inside V ⎪
⎩ ⎭ ⎩ ⎭ ⎩ ⎭
i = 1, . . . , N C j = 1, . . . , N P . (2.1-1)
There are a total of N C N P equations represented by Eq. (2.1-1). This

equation is the rate form of the conservation equation; an equivalent form based on
cumulative flow follows from integrating Eq. (2.1-1) with respect to time (see Sec.
2-5). From left to right in Eq. (2.1-1) are the accumulation, flux, and source terms,
respectively. A component can be transported within a phase by convection or
hydrodynamic dispersion. The generation of a component in a given phase can be the
result of chemical or biological reactions, injection or production of a component into
or from wells, or mass transfer from one phase to the next owing to phase changes.
These physical processes are discussed in more detail in Sec. 2.2.
The first term on the right side of Eq. (2.1-1) can be written as
⎧ Net rate of i ⎫ ⎧ Rate of i ⎫ ⎧ Rate of i ⎫

⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎨ in phase j =
⎬ ⎨ in phase j −
⎬ ⎨ in phase j ⎬
⎪ transported into V ⎪ ⎪ transported into V ⎪ ⎪ transported from V ⎪
⎩ ⎭ ⎩ ⎭ ⎩ ⎭
i = 1, . . . , N C j = 1, . . . , N P . (2.1-2)
We give mathematical form to each term in the following paragraphs.

The accumulation term for component i in phase j is
⎧ Rate of ⎫ ⎧Total mass of ⎫

⎪
⎨accumulation of i
⎪ in phase j in V
⎪ d ⎪
⎪ dt ⎪
⎪ d
⎬ = ⎨ i in phase j ⎬ =
⎪ dt
{∫ W dV }
V ij (2.1-3)
⎩ ⎭ ⎩ in V ⎭
where Wij is the component concentration in units of mass of i in phase j per unit bulk
volume. The units of Eq. (2.1-3) are mass per time. The volume integral represents
the sum of infinitesimal volume elements in V weighted by the concentration.
Since V is stationary,
d
dt
{∫ W dV } = ∫ dWdt dV .
V ij V
ij
(2.1-4)
8
This entire development may be repeated with a time-varying V with the same result
(Slattery, 1972).
The net flux term follows from considering the rate of transport across a
single surface element into V as shown in Fig. 2-4(b). Let N ij be the flux vector of
component i in phase j evaluated at the center of ΔA in units of mass of i in phase j
per surface area-time. N ij consists of components normal and tangential to n.
However, only the normal component n i N ij is crossing ΔA, and the rate of transport
across ΔA is
⎧ Rate of transport ⎫
⎪ ⎪
⎨ of i in phase j ⎬ = −n i N ij ΔA (2.1-5)
⎪across ΔA into V ⎪
⎩ ⎭
The minus sign arises because n and N ij are in opposing directions for transport
across ΔA into V ( n i N ij < 0 ), and this term must be positive from Eq. (2.1-1). An
inherent assumption in Eq. (2.1-5) is that the flux across ΔA is uniform, an
assumption that is valid as ΔA approaches zero. (Always remember that the
continuum assumption assures continuity in taking limits.) The summation of
infinitesimal surface elements yields
⎧ Net rate of i ⎫
⎪ ⎪
⎨ in phase j ⎬ = − ∫A n i N ij dA . (2.1-6)
⎪ transported into V ⎪
⎩ ⎭
Since the surface integral is over the entire surface of V, both flow into and from V
are included in Eq. (2.1-6).
The net rate of generation of i in phase j inside V is
⎧ Net rate of ⎫
⎪ ⎪
⎨ generation of i ⎬ = ∫V Rij dV + ∫V rmij dV , (2.1-7)
⎪in phase j inside V ⎪
⎩ ⎭
where Rij is the rate of mass generation in units of mass of i in phase j per bulk
volume-time. This term can account for both generation (Rij >0) and destruction (Rij
<0) of i, either through one or more chemical or biological reactions or through
physical sources (wells) in V.
The term rmij in Eq. (2.1-7) expresses the rate of mass transfer of component i
from or into phase j owing to vaporization, condensation, or sorption. We must have
∑ j =1 rmij = 0 , a relation following from the inability to accumulate mass at a
N
P
volumeless phase interface. You can show that this is true by letting V be the
interface itself. With no volume the fluxes must be continuous across an interface.
Substitution of Eqs. (2.1-3), (2.1-4), (2.1-6), and (2.1-7) into Eq. (2.1-1)
gives the following scalar equation for the conservation of i in each phase:
9
dWij
∫V dt
dV + ∫ n i N ij dA = ∫ Rij dV + ∫ rmij dV ,
A V V
i = 1, . . . , N C j = 1, . . . , N P , (2.1-8)
Equation (2.1-8) is an overall balance, or “weak” form of the conservation equation

in each phase. Versions of this equation will be used in solving for solutions that
have discontinuities, such as those that involve shocks or fronts. The weak form,
which is called an overall balance in the last section of this chapter, is also useful in
numerical simulation primarily because it is not tied to a particular coordinate
system.
The surface integral in Eq. (2.1-8) converts to a volume integral through the
divergence theorem
∫V ∇i BdV = ∫A n i BdA , (2.1-9)
where B can be any scalar, vector, or tensor function of position in V (with

appropriate changes in the operator definitions). The symbol ∇ is the divergence
operator, a kind of generalized derivative, whose specific form depends on the
coordinate system. Table 2-1 gives forms of ∇ in rectangular, cylindrical, and
spherical coordinates. The function B must be single-valued in V, a requirement met
by most physical solutions. Finally, implicit in the representation of the surface
integral of Eqs. (2.1-8) and (2.1-9) is the requirement that the integrand be evaluated
on the surface A of V.
Application of the divergence theorem to Eq. (2.1-8) gives
⎛ ∂Wij ⎞
∫ V
⎜
⎝ ∂t
+ ∇i N ij − Rij − rmij ⎟dV = 0 .
⎠
(2.1-10)
Using Eq. (2.1-10) restricts the formulation somewhat. V must now be simply
connected (a point on the exterior surface of V is always exterior) and the spatial
derivatives implied by the divergence exist because of the continuum assumption
discussed above. But since V is arbitrary in location and size, the integrand must be
zero:
∂Wij
+ ∇i N ij − Rij − rmij = 0 i = 1, . . . , N C j = 1, . . . , N P . (2.1-11)
∂t
The time derivative in Eq. (2.1-11) is now a partial derivative, the introduction of
other independent variables--the spatial coordinates--making this necessary.
Equation (2.1-11) is the differential or “strong” form for the component conservation
equation in each phase. It applies to any point (actually an REV) within the
macroscopic dimensions of the permeable medium independent of the boundary
conditions. From left to right in Eq. (2.1-11), the terms are now the accumulation,
transport, and source terms, the last consisting of two types.
10
The strong form, Eq. (2.1-11), is useful in developing analytic
solutions, a mainstay of this text. Equation (2.1-11) and its analogous conservation
equations are called the strong form because they express conservation at a point (a
REV) within a medium. The word "strong" means that if Eq. (2.1-11) is satisfied at
all points within V, Eq. (2.1-8), the weak form, is also satisfied. The converse is not
true. The exact form of the equations depends on the coordinate systems being used.
The next section gives specific definitions to the component concentration Wij,
flux N ij , and source terms Rij and rmij .
11
TABLE 2-1 SUMMARY OF DIFFERENTIAL OPERATORS IN RECTANGULAR, CYLINDRICAL,

AND SPHERICAL COORDINATES
Rectangular coordinates Cylindrical coordinates Spherical coordinates
(x, y, z) (r, θ, z) (r, θ, φ)
∂B ∂By ∂Bz 1 ∂ (rBr ) 1 ∂Bθ ∂Bz 1 ∂ (r 2 Br ) 1 ∂

∇i B = x + + ∇i B = + + ∇i B = + ( Bθ sin θ )
∂x ∂y ∂z r ∂r r ∂θ ∂z r 2
∂r r sin θ ∂θ
1 ∂Bφ
+
r sin θ ∂φ
∂S ∂S ∂S
[∇S ]x = [∇S ]r = [∇S ]r =
∂x ∂r ∂r
∂S 1 ∂S 1 ∂S
[∇S ] y = [∇S ]θ = [∇S ]θ =
∂y r ∂θ r ∂θ
∂S ∂S 1 ∂S
[∇S ]z = [∇S ]z = [∇S ]φ =
∂z ∂z r sin θ ∂φ
∂2 S ∂2 S ∂2 S 1 ∂ ⎛ ∂S ⎞ 1 ∂ 2 S ∂ 2 S 1 ∂ ⎛ 2 ∂S ⎞
∇2 S = + + ∇2 S = ⎜r ⎟+ + ∇2 S = 2 ⎜r ⎟
∂x 2 ∂y 2 ∂z 2 r ∂r ⎝ ∂r ⎠ r 2 ∂θ 2 ∂z 2 r ∂r ⎝ ∂r ⎠
1 ∂ ⎛ ∂S ⎞ 1 ∂2S
+ ⎜ sin θ ⎟+ 2 2
r sin θ ∂θ ⎝
2
∂θ ⎠ r sin θ ∂φ 2
Note: B = vector function

S = scalar function
12
2-2 DEFINITIONS AND CONSTITUTIVE EQUATIONS
FOR ISOTHERMAL FLOW
Each term in Eq. (2.1-11) represents an important physical process or mechanism.

This section examines each of these processes in further detail and defines some key
formation and fluid parameters. The units on Eq. (2.1-11) are amount per unit time.
Amount means either mass or moles; we will continue to use mass generically to
mean either mass or moles. When a distinction between the two is important, we will
be explicit.
Consider first a bulk volume V at the REV scale where NP phases exist.
Figure 2.5 illustrates such a volume that contains three phases: a solid phase
consisting of rock grains or soil, an aqueous phase, and an oleic phase. The porosity
φ is defined as the fraction of the bulk permeable medium that is pore space, that is,
the pore volume divided by the bulk volume V. The phase saturation Sj is defined as
the fraction of the pore volume occupied by phase j.
Figure 2-5 Representative bulk volume occupied by fluid and solid phases. The
volume of the fluid phases are equal to φ S jV .
The volume fraction of phase j, ε j , is the volume of phase j divided by the

bulk volume V. For fluid phases such as liquids and vapors, ε j = φ S j where φ S j is
also called the fluid content. For the solid phase, ε s = 1 − φ , which is the grain
volume divided by the bulk volume V. By definition ∑ Nj=1 ε j = 1 . The parameter ε j
P
is useful in writing general equations for any phase including the solid phase, as is
done next. In summary, the fraction of the bulk volume that is occupied by phase j is
⎧φ S for fluid phases

εj = ⎨ j (2.2-1)
⎩1 − φ for the solid phase
13
Accumulation Term
The accumulation term in Eq. (2.1-11) contains the component concentration for a
given phase (Wij), which we write now in terms of the volume fraction ε j . For solid
or fluid phases the mass of phase j in any bulk volume V is given by ε j ρ jV . As
before, the inherit assumption here is that the density of each phase is uniform in V,
which is strictly valid only as V approaches the REV scale.
We now define the mass fraction of component i in V to be ω i j . The
parameter ω i j is the mass of component i in phase j divided by the total mass of all
components in that same phase. Hence, ∑ iN=1 ωij = 1 . With that definition the total
C
mass of component i in phase j in V is WijV where Wij = ε j ρ j ωij .
Flux Terms
For advective transport of component i we assume that phase j moves with its
volume-averaged velocity u j . The component of the volumetric flow rate of
component i that enters the elemental surface ΔA is −n ⋅ u j ΔA . The mass flow rate of
component i in phase j that enters V thru the elemental surface ΔA is therefore
−n ⋅ ρ j ωij u j ΔA . Again, phase j properties are assumed uniform across ΔA, which is
valid as the elemental surface area becomes small. The component flux in phase j
owing to advective transport alone is therefore equal to jCij = ρ j ωij u j . Note that the
pore velocity for phase j is given by v j = u j ε j .
Hydrodynamic dispersion includes both molecular diffusion and mechanical
dispersion. Molecular diffusion is independent of the direction or magnitude of flow,
whereas mechanical dispersion in permeable media flows is anisotropic and depends
on the magnitude of flow. Dispersion has the same form as does diffusion in
nonpermeable media flows and, in fact, collapses to molecular diffusion in the limit
of small u j (see Chap. 5). At larger u j , the components of K ij , a second-order
tensor, can be many times larger than molecular diffusion since they now contain
contributions from fluctuations of the velocity u j and mass fraction ω i j about their
average values in the REV (Gray, 1975). Two components of K ij for a homogeneous,
isotropic permeable medium (Bear, 1972) are
Dij α lj u xj2 + α tj (u yj2 + u zj2 )

( K xx )ij = + (2.2-2)
τ φS j uj
(α lj − α tj )u xj u yj
( K xy )ij = , (2.2-3)
φSj uj
where the subscript l refers to the spatial coordinate in the direction parallel, or
longitudinal, to bulk flow, and t is any direction perpendicular, or transverse, to l. D i j
is the effective binary diffusion coefficient of component i in phase j (Bird et al.,
14
2002), α l j and α t j are the longitudinal and transverse dispersivities, and τ is the
permeable medium tortuosity. (Kt)ij is positive since (α l j – α t j ) is always positive.
As is customary, hydrodynamic dispersion is assumed to have a Fickian form
that is empirically modified for the volume fraction ε j of each phase. The flux of
component i in phase j with respect to volume-averaged velocity owing to
hydrodynamic dispersion alone is taken to be jDij = −ε j K ij∇ ( ρ jωij ) where the product
ρ jωij is the mass concentration of component i in phase j. Alternatively, the flux
with respect to mass-averaged velocity is jDij = −ε j ρ j K ij∇ω ij . The negative sign
indicates that positive flux occurs in the direction of decreasing mass fraction. Note
that ∇ωij has units of inverse length whereas K ij has units of squared length per time.
The total mass flux owing to both advective and dispersive transport is the
sum of the two. Thus, the component flux for phase j owing to both advective and
hydrodynamic dispersive transport is N ij = ρ jωij uij = jCij + jDij = ρ jωij u j − ε j K ij ∇ ( ρ jωij ) ,
where uij is the statistical average apparent velocity of a component in phase j owing
to both convection and dispersion. The velocity uij therefore is the sum of the
molecular velocities divided by the total number of molecules (Bird et al. 1960). The
convective and dispersive flux for a component in the solid phase is negligible for
most permeable media. The total mass flux of a component is just the sum of
N ij over all phases.
Volume or mass-averaged velocities are sometimes used interchangeably in
the literature even though the form of the dispersion term depends on the choice of
velocity. The volume-averaged velocity u j for a multicomponent system is
u j = ∑ i =C1 ρ j ωij uijVîj where Vîj is the partial mass volume of a component in phase j,
N
∑ ρ jωijVîj = 1 . The partial mass volume depends on

NC
where from thermodynamics i =1
the mass fractions ωij , pressure, and temperature. The mass-averaged velocity is
u j = ∑ i =C1 ρ jωij uij / ∑ i =C1 ρ jω ij = ∑ i =C1 ωij uij . Because the difference in the two velocities
N N N
can be shown to be ε j K ij i∇ ln ρ j , the volume-averaged velocity is equal to the mass-

averaged velocity for the case of an incompressible fluid. Experiments typically
measure volume-averaged velocities since volumetric rates are directly measured at
the inlet and outlet ends. Phase velocities, however, are never directly measured in
situ and are approximated from empirical flux laws such as Darcy’s law. Thus, the
choice of volume or mass-averaged velocities likely does not result in significant
errors for real mixtures owing to uncertainties in other formation parameters.
Source Terms
The source term Rij in Eq. (2.1-11) accounts for the rate of appearance or generation
of component i in phase j because of homogeneous chemical or biological reactions
15
within phase j (Levenspiel, 1999). Frequently, Rij is used to represent reactions
occurring at phase boundaries even though, strictly speaking, such reactions are the
consequence of flux terms evaluated at phase boundaries in the volume-averaging
procedures (Gray, 1975). In weak forms of the conservation laws it is also
convenient to use R i j to represent physical sources (wells) that are either specified or
related to the phase pressures and saturations.
There is no general function for Rij , although the volume fractions of each
phase are handled through Rij = ε j rij where rij is the reaction rate of component i in
phase j. Both Rij and rij have units of mass per volume per time, but rij is in terms of
the phase volume not the bulk volume. Each r i j could represent the sum of several
reactions within phase j if component i participates in simultaneous reactions. In a
given phase, mass is conserved so that ∑ i =1 rij = 0 if the rij is in mass units. This is
N C
not correct if mole units are used to express the reaction rates because moles are not
conserved in a chemical reaction. An example of a first-order reaction rate for
radioactive decay or biodegradation is rij = −ki ρ j ωij where ki is the decay constant or
reaction rate coefficient in units of inverse time.
The second source term rmij in Eq. (2.1-11) allows for the appearance or
disappearance of component i by mass transfer from phase j to another phase. This
term is difficult to calculate without detailed analysis of the transport occurring
within the phases. Thus, when mass transfer occurs, one typically simplifies the
equations by using overall compositional balances as described later.
The total mass generation rate of component i in phase j per bulk volume of
permeable medium is Rij + rmij . Injection and production of component i by wells is
not explicitly included in this term and are better treated as boundary conditions or
point sources.
Fundamental Conservation Equations for a Component in a Phase
A general set of partial differential equations for the conservation of component i in

phase j is obtained upon substitution of the definitions for concentration Wij, flux N ij ,
and source terms into Eq. (2.1-11):
∂
∂t
( )
(ε j ρ jωij ) + ∇i ρ jωij u j − ε j Kij ∇ ( ρ jωij ) = ε j rij + rmij
i = 1, . . . , N C j = 1, . . . , N P . (2.2-4)
The fundamental equations in Eq. (2.2-4) are the starting point for developing all
equations for permeable media flows. The method used in this text is to simplify Eq.
(2.2-4) by making the appropriate assumptions for the problem of interest. Several
special cases or "working equations" are given in Sec. 2-4. There are a total of N C N P
equations represented by Eq. (2.2-4).
16
Example 2-1. Writing the conservation equations for each component and phase.
Consider two-component, three-phase flow in a permeable medium, where the
components are component 1 and 2, and the phases are aqueous (w), oleic (o), and
the solid phase (s). Write the specific conservation equations for each component
and each phase assuming the solid phase is stationary (no deformation) and
dispersive transport of each component in the solid phase is negligible.
Under these assumptions, Eq. (2.2-4) consists of the following set of six
equations:
∂
∂t
( )
(φ S1 ρ1ω11 ) + ∇i ρ1ω11u1 − φ S1 K11∇ ( ρ1ω11 ) = φ S1r11 + rm11 ⎫⎪
⎪
⎬ aqueous phase
∂
∂t
( )
(φ S1 ρ1ω 21 ) + ∇i ρ1ω 21u1 − φ S1 K 21∇ ( ρ1ω 21 ) = φ S1r21 + rm 21 ⎪⎪
⎭
∂
∂t
( )
(φ S2 ρ 2ω12 ) + ∇i ρ 2ω12u2 − φ S2 K12∇ ( ρ 2ω12 ) = φ S2 r12 + rm12 ⎫⎪
⎪
⎬ oleic phase
∂
∂t
( )
(φ S2 ρ 2ω 22 ) + ∇i ρ 2ω 22u2 − φ S2 K 22∇ ( ρ 2ω 22 ) = φ S2 r22 + rm 22 ⎪⎪
⎭
∂
∂t
( (1 − φ ) ρ sω1s ) = (1 − φ ) r1s + rm1s ⎪⎪⎫
⎬ solid phase
∂
∂t
( (1 − φ ) ρ sω 2 s ) = (1 − φ ) r2 s + rm 2 s ⎪⎪
⎭
The terms that remain in the solid phase equations allow for chemical reactions and
mass transfer by adsorption. The above equations were arbitrarily grouped by the
type of phase; they could have been grouped by the type of component.
Overall Compositional Balances
Important subsets of the fundamental equations are the overall compositional balance
equations (see Fig. 2-1). Overall balances are written over all phases for each
component. These set of NC equations result from grouping Eqs. (2.1-11) or (2.2-4)
by component type and then summing those grouped equations over all phases. The
result of the summation of the component equations from Eq. (2.1-11) yields
∂Wi
+ ∇i N i = Ri i = 1, . . . , N C , (2.2-5)
∂t
where ∑ Nj=1 rmij = 0 and the equation has units of mass/bulk volume.. If we insert Eqs.
P
(2.2-7) through (2.2-9) from Table 2-2 into Eq. (2.2-5), we arrive at a general form of
the overall conservation equations
17
∂⎛ ⎞ ⎛
( ⎞
)
NP NP
⎜ φ ∑ ρ j S jωij + (1 − φ ) ρ sωis ⎟ + ∇i⎜ ∑ ρ jωij u j − φ S j Kij i∇ ( ρ jωij ) ⎟

∂t ⎝ j =1 ⎠ ⎝ j =1 ⎠
NP
= φ ∑ S j rij + (1 − φ )ris , i = 1, . . . , N C . (2.2-6)
j =1
The compositional equations are useful for modeling the overall flow of
component concentrations and are often used when mass transfer between fluid
phases occurs. Of course, some detail is lost because of the disappearance of the
interphase transport terms. The detail is partially restored by invoking the local
equilibrium assumption (Lake et al., 2002) that provides algebraic relations between
concentrations in the phases. This assumption forms the justification for the
discussion in Chap. 4. The local equilibrium assumption, used in the majority of the
cases dealt with in this text and in the remainder of this chapter, is discussed more
below.
Table 2-2 summarizes the overall balance equations needed for a complete
description of isothermal, multicomponent, multiphase flow in permeable media.
Column 1 in Table 2-2 gives the differential form of the equation named in column 2.
Column 3 gives the number of scalar equations represented by the equation in
column 1. Columns 4 and 5 give the identity and number of independent variables
added to the formulation by the equation in column 1. ND is the number of spatial
dimensions ( N D ≤ 3). The solid phase is a single homogeneous phase though more
than one solid can exist.
Example 2-2. Writing the overall compositional equations for each component.
Combine the fundamental equations given in Example 2-1 to form the overall
compositional balances. Use the same assumptions as in Example 2-1.
We first sort the equations listed in Example 2-1 by components 1 and 2. We
then sum them over the phases to obtain the overall compositional balances in the
form of Eq. (2.2-5). The resulting summation consists of the following set of two
equations:
∂ ⎫
(φ S ρ ω
1 1 11
+ φ S2 ρ 2ω12 + (1 − φ ) ρ sω1s ) +
⎪
∂t
⎪
( ) (
∇i ⎡ ρ1ω11u1 − φ S1 K11∇ ( ρ1ω11 ) + ρ 2ω12 u2 − φ S 2 K12 ∇ ( ρ 2ω12 )
⎣ ) ⎤ ⎪⎬ component 1
⎦
⎪
= φ S1 r11 + φ S 2 r12 + (1 − φ ) r1s ⎪
⎪⎭
∂ ⎫
(φ S ρ ω
1 1 21
+ φ S 2 ρ 2ω 22 + (1 − φ ) ρ sω 2 s ) +
⎪
∂t
⎪
( ) (
∇i ⎡ ρ1ω11u1 − φ S1 K 21∇ ( ρ1ω 21 ) + ρ 2ω12 u2 − φ S2 K12 ∇ ( ρ 2ω12 )
⎣ ) ⎤ ⎪⎬ component 2
⎦
⎪
= φ S1 r21 + φ S 2 r22 + (1 − φ ) r2 s ⎪
⎪⎭
18
where we have used ∑ NP
r
j =1 mij = 0 . The equations above can be rewritten as
∂⎛ 2
∂t ⎝ j =1
⎞
⎠
⎡ 2
(
⎜ φ ∑ S j ρ jω1 j + (1 − φ ) ρ sω1s ⎟ + ∇i ⎢ ∑ ρ jω1 j u j − φ S j K1 j ∇ ( ρ jω1 j )
⎣ j =1
)⎤⎦⎥ ⎫⎪⎪
2
⎬ component 1
= φ ∑ S1r11 + (1 − φ ) r1s ⎪
⎪
j =1 ⎭
∂⎛ 2 ⎞
⎠
⎡ 2
( )⎤⎫
⎜ φ ∑ S j ρ jω 2 j + (1 − φ ) ρ sω 2 s ⎟ + ∇i ⎢ ∑ ρ jω1 j u j − φ S j K 2 j ∇ ( ρ jω 2 j ) ⎥ ⎪
∂t ⎝ j =1 ⎣ j =1 ⎦ ⎪ component 2
2
⎬
= φ ∑ S j r2 j + (1 − φ ) r2 s ⎪
⎪
j =1 ⎭
19
TABLE 2-2 SUMMARY OF OVERALL CONSERVATION STRONG FORM EQUATIONS FOR ISOTHERMAL FLUID FLOW IN
PERMEABLE MEDIA
Number of
independent Dependent variables†
scalar
Equation Name equations* Identity Number
(1) (2) (3) (4) (5)
∂Wi Component i
(2.2-5) + ∇i N i = Ri NC Wi , Ri , N i 2NC + NCND
∂t conservation
NP
(2.2-7) Wi = φ ∑ρ S ω
j =1
j j ij + (1 − φ ) ρ sωis Overall concentration NC – 1 ρj, Sj, ωij, ωis 2NP + NPNC + NC
(2.2-8)
( )
NP Component i total
N i = ∑ ρ jωij u j − φ S j Kij • ∇ ( ρ jωij ) NCND uj NPND
flux
j =1
NP
(2.2-9) Ri = φ ∑S r
j =1
j ij + (1 − φ )ris Component i total
source
NC – 1 rij, ris NPNC + NC
NC
Total reaction 1
(2.2-10) ∑R
i =1
i =0
definition
(2.2-11) u j = −λrj k • (∇Pj − ρ j g ) Darcy’s law NPND λrj, Pj 2NP

(2.2-12) λrj = λrj ( S , ω , u j , x ) Relative mobility NP
(2.2-13) Pj − Pn = Pcjn ( S , ω , x ) Capillary pressure NP – 1
definition
* Total independent equations = ND(NP + NC) + 2NPNC + 4NP + 4NC

†
Total dependent variables = ND(NP + NC) + 2NPNC + 4NP + 4NC
20
TABLE 2-2 CONTINUED

Number of
scalar
Equation Name equations* Identity Number
(1) (2) (3) (4) (5)
NC
(2.2-14) ∑ω ij =1 Mass fraction NP
i =1 definition
NC
Solid phase mass 1
(2.2-15) ∑ω
i =1
is =1
fraction definition
NP
(2.2-16) ∑S
j =1
j =1 Saturation definition 1
(2.2-17) rij = rj (ωij , Pj ) Homogeneous kinetic (NC – 1)NP

reaction rates
(2.2-18) ris = ris (ωis ) Solid phase reaction NC – 1
rates
NC
Total phase reaction
(2.2-19) ∑r
i =1
ij =0
definition
NP
NC
Solid phase total 1
(2.2-20) ∑r
i =1
is =0
reaction rates
(2.2-21) ωij = ωij (ωik ) k ≠ j Equilibrium relations NC(NP – 1)
(or phase balances)
(2.2-22) ωis = ωis (ωij ) Solid phase NC
equilibrium relations
(or phase balances)
(2.2-23) ρ j = ρ j (T , Pj ) Equations of state NP
* Total independent equations = ND(NP + NC) + 2NPNC + 4NP + 4NC

†
Total dependent variables = ND(NP + NC) + 2NPNC + 4NP + 4NC
21
A normally subscripted quantity (for example ω i j ) appearing without
subscripts in Table 2-2 indicates a relationship involving, at most, all members of the
subscripted set. In the listing of dependent variables, the primary media properties,
such as porosity φ, and the permeability tensor k , are given functions of position x
within the permeable medium. These quantities are, strictly speaking, functions of the
fluid pressure within the medium (Dake, 1978), but for pressures nondestructive to
the permeable medium, this effect is generally weak. We also assume the solid-phase
density ρs is given, as is the dispersion tensor K ij , even though the latter is a function
of the phase velocities and molecular diffusivities. The remaining terms in Table 2-2
are defined in the Nomenclature and below.
The first four equations in Table 2-2 are the component conservation
Eq. (2.2-5) and definitions for the accumulation, flux, and source terms in this
equation. We take the NC conservation equations to be the independent set of
equations; the conservation of overall mass or continuity equation, which follows
from summing Eq. (2.2-5) from 1 to NC, is not listed as an independent equation (see
Sec. 2-4). In solving specific problems, it may be more convenient to take the
problem statement as the continuity equation and NC – 1 mass conservation equations
with the major component (for example, water in flow of dissolved salts in an
aqueous solution) being the one omitted.
Definition of Terms in the Overall Compositional Equations
The accumulation term Wi, the overall concentration of component i, represents the
sum of the component i in the NP flowing phases plus the solid phase as shown in Eq.
(2.2-7). There are only NC – 1 independent Wi, since summing on i with the mass
fraction definitions of Eqs. (2.2-14) and (2.2-15) gives
NC NP
∑W
i =1
i = φ ∑ ρ j S j + (1 − φ )ρ s ≡ ρ (ωi , P )
i =1
(2.2-24)
where ρ is the overall density of the permeable medium (total mass flowing plus
solid phase divided by the bulk volume). We can regard the overall density as a
complicated function of some local pressure P and the set of overall mass fractions
defined as
Wi
ωi = NC
. (2.2-25)
∑W
i =1
i
Here ω i is the mass of component i in all phases divided by the total mass of the
permeable medium. The combination of Eqs. (2.2-24) and (2.2-25) yields a
constraint on the Wi
NC
ρ (W1 , . . . , WN , P) = ∑ Wi ,
C
(2.2-26)
i =1
22
which means there are NC – 1 independent Wi, not NC. The notation on the left side of
Eq. (2.2-26) indicates that ρ is a function of two variables, the set of overall
concentrations and pressure. Equations (2.2-24) through (2.2-26) can be construed as
a constraint on the mass fraction ω i j , phase pressures Pj, and saturations Sj.
Auxiliary Relations
Equation (2.2-11) is a multiphase version of Darcy’s law for flow in permeable

media (Collins, 1976). The single-phase version of Darcy’s law is actually a volume-
averaged form of the momentum equation (Slattery, 1972; Hubbert, 1956). The form
given in Eq. (2.2-11) assumes creeping flow in the permeable medium with no fluid
slip at the solid phase boundaries. Corrections to account for non-Darcy effects
appear in standard references (Collins, 1976; Bear, 1972). The potential function for
the phase superficial velocity u j is the vectorial sum ∇Pj + ρ j g , where Pj is the
pressure within the continuous phase j. g is the gravitational vector, which is
assumed constant and directed toward the Earth’s center. Hereafter in this text, we
assume the coordinate direction parallel to g is positive upward, away from the
Earth's center. The gravitational vector can be written as
g = − g ∇Dz , (2.2-27)
where g is the magnitude of the gravitational vector and Dz is a positive distance

above some horizontal reference plane, typically depth. For Cartesian coordinate
systems with a constant inclination with the reference plane, ∇Dz becomes a vector
consisting of cosines of the inclination angles between the respective axis and the
vertical.
The tensorial form of the permeability k implies an anisotropic permeable
medium having coordinate axes unaligned with the principal axis of k . With the
inclusion of k , we have now included all the primary permeable media properties, φ,
k , αlj, αtj, and τ, into the formulation. These properties, and their spatial
distribution, are geologic in nature, being characterized by the details of the
individual flow paths.
The other quantity in Eq. (2.2-11) is the relative mobility λrj of phase j,
defined as the quotient of the relative permeability krj and viscosity μj.
krj ( S , ω , x )
λrj = . (2.2-28)
μ j (ω , u j )
Equation (2.2-28) decomposes λrj into a rock–fluid property krj and a fluid property
μ j . krj is a function of the tendency of phase j to wet the permeable medium, of pore
size distribution, and of the entire set of phase saturations (see Chap. 3). μj is a
function of the phase composition and, if phase j is non-Newtonian, the magnitude of
the superficial velocity u j (see Chap. 8). The relative permeabilities and viscosities
23
krj and μ j are usually determined experimentally to give λrj. It is slightly more general
to write the λrj k product in Eq. (2.2-11) as
kj
λrj k = (2.2-29)
μj
where k j is the phase permeability tensor. This form allows for anisotropic relative
permeabilities, but little is known about the anisotropic nature of the relative
permeabilities, so it remains a scalar function here, that is, k j = krj k . Relative
permeabilities and viscosities are chemical properties.
The difference between the phase pressures of any two phases flowing in the
REV is the capillary pressure, defined as in Eq. (2.2-13). The capillary pressure
between the phases j and n is a function of most of the same variables as the relative
permeability (Fatt and Dykstra, 1951). That there are NP – 1 independent relations
follows from considering the set of all capillary pressures with j fixed, Pc1j, Pc2j, . . . ,
PcN . Ignoring the trivial case of Pcjj (=0), there are clearly NP – 1 capillary pressures.
Pj
The capillary pressure Pckn between any two other phases k and n may be expressed
as a linear combination of members from the original set.
Pckn = Pk − Pn = ( Pk − Pj ) + ( Pj − Pn ) = Pckj + Pcjn . (2.2-30)
Hence there are only NP – 1 independent capillary pressure relations, usually

determined experimentally under static conditions. We discuss capillary pressure in
more detail in Chap. 3.
The pressures Pj are the continuous phase pressures, not the pressures that
would exist in disconnected “globules” of phase j. In the latter case, the phase
pressure differences still exist but, being a reflection of the local permeable medium
pore configuration, are not uniquely determined by the functions given in Eq.
(2.2-13).
Equations (2.2-14), (2.2-15), and (2.2-16) follow from the definitions of mass
fraction and phase saturation, respectively.
Equations (2.2-17) through (2.2-20) are definitions of the reaction rate of
component i in phase j or in the solid phase. As was true for Ri, there can be no net
accumulation of mass in a phase owing to chemical reaction. Then the reaction rate
terms rij and ris sum to zero as indicated by Eqs. (2.2-18) and (2.2-19) if the equation
is written in mass units.
Local Equilibrium
Equations (2.2-21) and (2.2-22) are relations among the mass fractions of the NP
flowing phases and the solid phase present in the REV. These relations arise from
solving the conservation equation for each component in each phase
24
∂Wij
+ ∇i N ij = Rij + rmij . (2.1-11)
∂t
Since the sum of the conservation equations over all flowing phases for component i
is Eq. (2.2-5), there are NC(NP – 1) such independent phase balances. Since there are
also NC phase balances for the solid phase, the total number of independent relations
is NCNP. There are a similar number of additional unknowns, the rmij, which must be
independently specified.
The phase balance is formally correct, but requires considerable additional
work to be useful. A much more practical approach is to assume local
thermodynamic equilibrium; that is, the mass fractions of component i are related
through thermodynamic equilibrium relations (Pope and Nelson, 1978). For flow
through naturally occurring permeable media, the assumption of local equilibria
among phases is usually adequate (Raimondi and Torcaso, 1965). Exceptions are
flows at very high rates or leachant flows such as might occur in alkaline floods. The
local equilibrium approximation, or LEA, is discussed further in Lake et al., 2002.
If local equilibrium applies, the number NCNP of independent scalar
equations may be derived from the phase rule (see Chap. 4). The equilibrium
relations themselves are very strong functions of the particular EOR process, and
much of the behavior and many of the important features of a given process can be
understood from these considerations. Chapter 4 discusses phase behavior generally;
we reserve more specifics for the relevant sections on solvent, chemical, and thermal
flooding.
The final set of equations in Table 2-2 are the equations of state,
Eq. (2.2-23), which relate each phase density to its composition, temperature, and
pressure. For LEA flows, the equilibrium relations for the flowing phases, Eqs.
(2.2-21) and (2.2-22), can be derived from the equation of state as discussed in Chap.
4. This practice enforces internal consistency between the equilibrium relations and
the equations of state. In the text, however, we will often invoke simpler equilibrium
relations for pedagogical purposes.
Phase Conservation Equations
Another important set of equations are the phase conservation equations, which are
derived from the fundamental Eqs. (2.1-11) or (2.2-4) (see Fig. 2-1). This set of NP
equations result from grouping the equations by each phase and then summing those
grouped equations over each component. The result of the summations on Eq.
(2.2-4) gives
∂ NC
∂t
( ε j ρ j ) + ∇i( ρ j u j ) = ∑ rmij j = 1, . . . , N P , (2.2-31)
i =1
∑
NC
where we used Eqs. (2.2-19) and (2.2-20) and i =1
∇ ⋅ jDij = 0 (net dispersive flux in a
phase is zero). The dispersive flux term vanishes strictly only for the case when the
25
flux is written with respect to mass-averaged velocity (see exercise 2D).
Nevertheless, we assume this is also approximately correct when the dispersive flux
is written with respect to volume-averaged velocity. The phase conservation
equations are useful for modeling fluid flow of several phases. In general, this is the
case when fluids are immiscible so that the phase compositions are fixed.
Under the assumption of no mass transfer and sorption, rmij = 0 , Eq. (2.2-31)
reduces to the immiscible phase conservation equations,
∂
∂t
(ε j ρ j ) + ∇i( ρ j u j ) = 0 j = 1, . . . , N P , (2.2-32)
where ε j = φ S j from Eq. (2.2-1) and NP represents only flowing phases. The solid
phase equation is omitted because its solution is trivial when it is non-deformable,
that is, (1 − φ ) ρ s is constant temporally.
Example 2-3. Writing the phase conservation equations for each phase.
Combine the fundamental equations given in Example 2-1 to form the phase
conservation equations for the aqueous, oleic, and solid phases. Use the same
assumptions as in Example 2-1.
The equations in Example 2-1 are already sorted by phase. Thus, we simply
sum each set of grouped equations over the components to obtain the form of Eq.
(2.2-31). The resulting summation consists of the following three equations:
∂ ⎫
(φ S1 ρ1ω11 + φ S1 ρ1ω21 ) + ⎪
∂t ⎬ aqueous phase
∇i[ ρ1ω11u1 + ρ1ω21u1 ] = φ S1r11 + φ S1 r21 + rm11 + rm 21 ⎪
⎭
∂ ⎫
(φ S2 ρ 2ω12 + φ S2 ρ 2ω22 ) + ⎪
∂t ⎬ oleic phase
∇i[ ρ 2ω12 u2 + ρ 2ω22 u2 ] = φ S 2 r12 + ε 2 r22 + rm12 + rm 22 ⎪⎭
∂
∂t
( (1 − φ ) ρ sω1s + (1 − φ ) ρ sω2 s ) = (1 − φ ) r1s + (1 − φ ) r2 s + rm1s + rm 2 s ⎫⎬ solid phase
⎭
∑
NC
where we have used i =1
∇ ⋅ jDij = 0 . These equations can be further simplified by
combining terms and using Eqs. (2.2-14), (2.2-15), (2.2-19), and (2.2-20). The final
result is
∂
(φ S1 ρ1 ) + ∇i( ρ1u1 ) = rm11 + rm 21 ⎫⎬ aqueous phase
∂t ⎭
∂
(φ S2 ρ 2 ) + ∇i( ρ 2u2 ) = rm12 + rm 22 ⎫⎬ oleic phase
∂t ⎭
26
∂
∂t
( (1 − φ ) ρ s ) = rm1s + rm 2 s ⎫⎬ solid phase .
⎭
When flow is fully immiscible and there is no adsorption, the mass transfer terms in
the above equations are zero. When there is no adsorption, the solid phase equation
can be dropped, that is, its solution is again trivial. Eqs. (2.2-32) result under these
additional assumptions.
Continuity Equation
We sum Eq. (2.2-6) over the NC components to obtain the equation of continuity, or
conservation of total mass. We could have also obtained the equation of continuity
by summing the phase conservation equations, Eq. (2.2-31), over the NP phases
including the solid phase. The equation of continuity is
∂ ⎛ NP ⎞ ⎛ NP ⎞
⎜ φ ∑ ρ j S j + (1 − φ ) ρ s ⎟ + ∇i⎜ ∑ ρ j u j ⎟ = 0 . (2.2-33)
∂t ⎝ j =1 ⎠ ⎝ j =1 ⎠
Equation (2.2-33) can be written totally in terms of pressure and saturation

derivatives using Eqs. (2.2-11) and (2.2-24); this equation is a form of the “pressure”
equation. Equation (2.2-33) is strictly correct only when the velocity is the mass-
averaged velocity, not volume-averaged.
2-3 ENERGY BALANCE EQUATIONS
For steam, hot water injection, and in situ combustion—some of the most important
EOR and remediation processes—the temperature changes with both space and time.
The equations of Table 2-2 apply equally well to nonisothermal flow but with an
additional dependent variable, temperature, added to the formulation. The additional
equation required to make the problem statement determinant is the conservation of
energy, or the first law of thermodynamics. The first law is based on our every day
observation that for any change of thermodynamic properties, total energy, which
includes internal, potential, kinetic, heat, and work, is conserved.
A statement of the energy balance or first law of thermodynamics suitable for
our purposes is
⎧ Rate of ⎫ ⎧ Net rate ⎫ ⎧ Net rate of ⎫

⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎨ accumulation =
⎬ ⎨ of energy +
⎬ ⎨ generation of energy ⎬ (2.3-1)
⎪of energy in V ⎪ ⎪ transported into V ⎪ ⎪ inside ⎪
⎩ ⎭ ⎩ ⎭ ⎩ V ⎭
where V is an arbitrary volume as in Fig. 2-4. From left to right in Eq. (2.3-1) are the
accumulation, flux, and source terms, respectively. We use the parallel between the
27
component conservation Eq. (2.1-1) and Eq. (2.3-1) to shorten the following
development. By analogy to the procedure in Sec. 2-2, the accumulation term in Eq.
(2.3-1) for a stationary V can be written as
⎧ Rate of ⎫
⎪ ⎪ d ⎧Total energy ⎫
⎨ accumulation ⎬ = ⎨ ⎬
⎪of energy in V ⎪ dt ⎩ in V ⎭
⎩ ⎭
d NP ⎛ ⎡ 1 2 ⎤⎞
= ∫V dt ∑ ⎜ ε j ρ j ⎢U j + v j − gDz ⎥ ⎟dV
j =1 ⎝ ⎣ 2 ⎦⎠
(2.3-2)
where the total energy includes internal, kinetic, and potential energy. Equation
(2.3-2) can also be written as
⎧ Rate of ⎫
⎪ ⎪ d⎛ 1 ⎞
⎬ = ∫V ⎜ ρU + ρ v − ρ gDz ⎟ dV
2
⎨ accumulation (2.3-3)
⎪of energy in V ⎪ dt ⎝ 2 ⎠
⎩ ⎭
where U is an overall internal energy (total energy/total mass), and ρ is the overall
density given by Eq. (2.2-24). In Eq. (2.3-3) the term 1/ 2 ρ v
2
( )
represents total
kinetic energy per unit bulk volume and − ρ gDz total potential energy per unit bulk
volume with reference to the depth below some horizontal plane.
The remaining terms in Eq. (2.3-1) are represented by
d⎛ 1 ⎞
∫V dt ⎜⎝ ρU + 2 ρ v − ρ gDz ⎟ dV = − ∫ ∇i EdV + W
2
(2.3-4)
⎠ V
where the terms E and W represent energy flux, and source, respectively, to which
we give specific form below. The negative sign in front of the first term on the right
is required to make energy flux positive when it flows into volume V. See Chap. 4
for more treatment of the first law of thermodynamics with application to phase
behavior.
The source term requires more elaboration than do the other terms in Eq.
(2.3-4). The form of the first law of thermodynamics for open systems expressed by
Eq. (2.3-4) requires the W term to be composed of work components only, in the
absence of external heating sources. External heating sources can often be handled
through boundary conditions. Heats of reaction, vaporization, and solution are, of
course, important in several processes, but these are implicitly present in the equation
in the concentration and flux terms. Here we consider only the rate of work done
against a pressure field WPV , although other types of work could be included (see
exercise 2P). In this derivation, there is no compression or expansion work done on
volume V since it is assumed to be fixed.
28
Returning to Fig. 2-4(b), consider an element in the multiphase,
multicomponent flow field crossing ΔA. Since work is the product of force times a
distance, the rate of work is force times a velocity. The element crossing ΔA is,
therefore, doing work ΔWPV , where
NP
ΔWPV = −∑ Pj ΔAn iu j . (2.3-5)
j =1
The term Pj ΔAn is the force exerted on ΔA by the pressure in phase j. The scalar
product in Eq. (2.3-5) expresses a general definition of work rate when using vector
forces and velocities. The negative sign in Eq. (2.3-5) is to satisfy the usual
thermodynamic sign convention for work, since ΔWPV must be positive for work
done on a fluid element flowing into V (n iu j < 0). The total pressure-volume work is
the sum of Eq. (2.3-5) over all surface elements, which, in the limit of the largest ΔA
approaching zero, becomes a surface integral. Using the divergence theorem, Eq.
(2.1-9), on this integral gives the final form for WPV .
NP
WPV = − ∫
V
∑ ∇i( P u )dV
j =1
j j . (2.3-6)
The expression for work fits well into Eq. (2.3-4). After collecting all terms
under the same volume integral and making the integrand zero because V is again
arbitrary, we have
∂⎛ 1 ⎞
NP
⎜ ρU + ρ v − ρ gDz ⎟ + ∇i E + ∑ ∇i( Pj u j ) = 0
2
(2.3-7)
∂t ⎝ 2 ⎠ j =1
The energy flux term is made up of convective contributions from the

flowing phases (internal, kinetic, and potential energy), conduction, and radiation, all
other forms being neglected.
⎡ 1 2 ⎤
NP
E = ∑ ρ j u j ⎢U j + v j − gDz ⎥ + qc + qr . (2.3-8)
j =1 ⎣ 2 ⎦
For brevity, we neglect radiation in the following discussion, though this transport
mechanism can be important in estimating heat losses from wells and in certain forms
of EOR and remediation that involve electromagnetic sources. For multiphase flow,
the conductive heat flux is from Fourier’s law,
qc = −kTt ∇T , (2.3-9)
where kTt is the scalar total thermal conductivity. kTt is a complex function of the
phase saturations and phase kTj and solid kTs thermal conductivities, which we take to
29
be known (see Chap. 11). The parallel between Eq. (2.3-8) and the dispersive flux
term in Eq. (2.2-8) is obvious. We have also invoked the requirement of local thermal
equilibrium in this definition by taking the temperature T to be the same in all phases
within the REV.
Inserting definitions (2.3-8) and (2.3-9) into Eq. (2.3-7) yields
∂⎛ 1 ⎞ ⎛ NP ⎡ 1 2 ⎤⎞
⎜ ρU + ρ v − ρ gDz ⎟ + ∇i⎜ ∑ ρ j u j ⎢U j + v j − gDz ⎥ ⎟
2
∂t ⎝ 2 ⎠ ⎝ j =1 ⎣ 2 ⎦⎠
NP
− ∇i(kTt ∇T ) + ∑ ∇i( Pj u j ) = 0 . (2.3-10)
j =1
The first sum in the energy flux and that in the pressure-volume work expression may
be combined to give
∂⎛ 1 ⎞ ⎛ NP ⎡ 1 2 ⎤⎞
⎜ ∑ ρ j u j ⎢ H j + v j − gDz ⎥ ⎟
2
⎜ ρU + ρ v − ρ gDz ⎟ + ∇ i
∂t ⎝ 2 ⎠ ⎝ j =1 ⎣ 2 ⎦⎠
− ∇i(kTt ∇T ) = 0 (2.3-11)
where Hj = Uj + Pj/ρj is defined as the enthalpy of phase j per unit mass of j.

30
TABLE 2-3 SUMMARY OF ADDITIONAL STRONG FORM EQUATIONS FOR NONISOTHERMAL FLUID FLOW
IN PERMEABLE MEDIA
Number
scalar*
Equation Name equations Identity Number
(1) (2) (3) (4) (5)
∂ 1 2
(2.3-12) ( ρU + ρ v − ρ gDz ) Energy conservation 1 U, Hj, T NP + 2
∂t 2
⎛ NP
⎡ 1 2 ⎤⎞
+∇i⎜ ∑ ρ j u j ⎢ H j + v j − gDz ⎥ ⎟
⎝ j =1 ⎣ 2 ⎦⎠
−∇i(kTt ∇T ) = 0
NP
(2.3-13) ρU = φ ∑ρ S U
j =1
j j j + (1 − φ ) ρ sU s Total internal energy 1 Uj, Us NP + 1
NC
⎫
U j = ∑ ωijU ij ⎪
NP Uij NPNC
i =1 ⎪⎪
(2.3-14) ⎬ Phase internal energy
NC ⎪
U s = ∑ ωisU is ⎪
1
i =1 ⎪⎭ Uis NC
NC
(2.3-15) H j = ∑ω H
i =1
ij ij Phase enthalpy NP Hij NPNC
U ij = U ij (T , Pj , ω ) ⎫ NPNC
⎪ Partial mass internal
(2.3-16) ⎬ energy
U is = U is (T , Pj , ω ) ⎪⎭ NC
31
(2.3-17) H ij = H ij (T , Pj , ω ) Partial mass enthalpy NPNC
*Total independent equations = 2(NPNC) + 2NP + NC + 3

†
Total dependent variables = 2(NPNC) + 2NP + NC + 3
32
The conservation equation discussed in this section and in Table 2-3 are
"overall" in the sense that the balances are written over a REV that contains all the
phases. We could have written balances on each phase or each component just as
was done in the previous section for mass conservation. These balances would have
in them terms describing the rate of energy transport between phases and would not
necessarily assume equality of temperature in all phases within a REV.
Equation (2.3-11) can also be obtained if the volume V is no longer
stationary and both pressure-volume and compression-expansion work are present
(see exercise 2Q). For example, consider a closed system where volume V expands
or contracts at the same rate as the fluid flow velocity. In that case, compress-
expansion work is done on the boundary of V as it deforms and pressure-volume
work is zero. This type of work is given by ΔWCE = Fext ⋅ u , where Fext is the
component of external force acting along the velocity vector u . Work is always
related to the external pressure or force. If the external pressure were zero, there
would be no work done by the system because the surroundings offer no resistance.
The rate of work done on a surface element of V is therefore,
NP
ΔWCE = −∑ Pj ΔAn iu j , (2.3-18)
j =1
which is identical to Eq. (2.3-5) when the volume V deforms with the phase
velocities. The development then proceeds as before, but with the accumulation term
for a time-varying V deforming at a Darcy rate u j written as (Slattery, 1972)
∂ NP ⎛ ⎡ 1 2 ⎤⎞
∑ ε j ρ j ⎢⎣U j + 2 v j − gDz ⎥⎦ ⎟dV =
∂t ∫V j =1 ⎜⎝ ⎠
∂ NP ⎛ ⎡ 1 2 ⎤⎞
NP
⎛ ⎡ 1 2 ⎤⎞
∫
V ( t ) ∂t
∑ ⎜
j =1 ⎝
ε j ρ j ⎢U j + v j − gDz ⎥ ⎟dV + ∫ ∇i∑ ⎜ ρ j u j ⎢U j + v j − gDz ⎥ ⎟dA .
⎣ 2 ⎦⎠ A( t )
j =1 ⎝ ⎣ 2 ⎦⎠
Equation (2.3-18) can be rewritten in a more familiar form by combining the

surface element areas with the component of the velocity vector perpendicular to that
surface element ( Δq j ⊥ = n iΔAu j = n iΔq j ), summing over all surface elements, and
taking the limit as the largest Δq j ⊥ approaches zero. The result is,
NP
WCE = −∑ ∫ Pj dq j ⊥ . (2.3-19)
q j ⊥∈A ( t )
j =1
Further, when the phase pressures are uniform along the boundary of V, the phase
pressures can pass through the integral so that
NP
dV j
WCE = −∑ Pj . (2.3-20)
j =1 dt
33
V j in Eq. (2.3-20) is now the total volume of phase j within V. Eq. (2.3-20) is valid
no matter the shape or size of volume V.
Auxiliary Relations
Table 2-3 summarizes the equations that, together with those of Table 2-2,
completely specify non-isothermal fluid flow problems. The first three equations we
have already discussed.
The energy concentration per unit bulk volume must include internal energy
contributions from all flowing phases and the solid phase,
NP
ρU = φ ∑ ρ j S jU j + (1 − φ ) ρ sU s (2.3-13)
j =1
where Uj is the internal energy per unit mass of phase j. The total kinetic energy term
includes kinetic energy contributions from all flowing phases,
1 ⎛1 2⎞
NP
ρ v = φ∑⎜ ρ jS j vj ⎟
2
(2.3-21)
2 j =1 ⎝ 2 ⎠
where the solid phase velocity is negligible.

The phase internal energies Uj and Us and the enthalpies Hj are functions of
temperature T, phase pressure Pj, and composition ω i j . One form this dependency
can take is Eq. (2.3-14), where the doubly subscripted internal energies (and
enthalpies) are partial mass quantities. Partial mass quantities, Eq. (2.3-16), are
analogous to partial molar quantities in solution thermodynamics (Denbigh, 1968).
For example, the partial mass internal energy of component i in phase j is the change
in Uj as ω i j is changed, all other variables being held constant,
⎛ ∂U j ⎞
U ij = ⎜ ⎟⎟ (2.3-22)
⎜ ∂ω
⎝ ij ⎠ Pj ,T ,ωkj ,k ≠i
and similarly for Uis and Hij. The partial mass properties themselves may be
calculated from equations of state, Eq. (2.2-23), or empirical correlations as functions
of temperature, pressure, and composition.
Equations (2.3-14) and (2.3-15) readily revert to simple forms. For example,
if phase j is an ideal solution, the partial mass quantities become pure component
quantities, functions of temperature and pressure only. Further, if j is an ideal gas, the
partial mass quantities are functions only of temperature.
The equations presented in Tables 2-2 and 2-3 are complete, but they can
only be solved with the specification of a similarly complete set of initial and
boundary conditions.
2-4 ENTROPY BALANCE EQUATIONS

34
The equations for conservation of total mass and energy are insufficient to solve
many thermodynamic problems. For example, experience has shown that a pond on a
hot summer day will not freeze, but instead will equilibrate to the temperature of its
surroundings. The first law of thermodynamics, however, is satisfied whether the
pond freezes (loses energy to surroundings) or heats up (gains energy from
surroundings). That is, the direction of energy transfer is irrelevant to the first law of
thermodynamics and there must be another equation that describes the approach to
equilibrium.
More generally we observe that as long as there are no outside influences,
gradients in concentration, temperature, and pressure eventually vanish reaching a
time-invariant equilibrium state with its surroundings. The second law of
thermodynamics provides a mathematical statement that describes this unidirectional
nature of spontaneous processes. A new thermodynamic property, called entropy, is
introduced to describe the tendency towards equilibrium. Although we are not as
familiar with entropy, it is a property just like temperature and pressure. The main
difference, however, is that we must infer its value from other measurable
thermodynamic properties such as temperature and pressure.
Entropy increases in spontaneous processes until equilibrium is reached. For
example, Figure 2-6 shows an initial state of a hypothetical closed volume that
contains four molecules. The molecules are numbered 1 to 4 to aid in the accounting;
in reality the molecules are indistinguishable from each other. The system is initially
partitioned into two halves or subsystems, such that the molecules from one half
cannot move into the other half. Thus, each subsystem has only one possible
configuration at zero time, the initial state. The small number of possible
configurations at the initial state corresponds to a well-ordered or small-entropy state.
When the partition is removed, the molecules from each subsystem are free
to move into the other half of the system, say by random Brownian motion. When the
elapsed time is short, the initial state is the most likely configuration. Over time,
however, a total of 16 different equally-probable configurations are possible relative
to the original partition location (see Figure 2-6). The increase in possible
configurations of the system is related to an increase in entropy or disorder. A
careful examination of the possible configurations shows further that a state with one
or two molecules on each side is more likely to exist than the initial state, which has a
probability of only 2-4 or 1/16, the initial state being just one of 16 possible states. A
true equilibrium state with two molecules on either side of the original partition (no
gradients) is also more likely to exist (probability of 6/16 or 3/8) than the initial state.
35
Figure 2-6: Illustration of configurations for four molecules in a closed composite system
(after Smith et al. 2001). The molecules on each side of the partition are initially constrained
by a partition.
The illustration of Figure 2-6 demonstrates in a simple way why equilibrium

occurs in natural processes; equilibrium is the most likely state. A more realistic
example, with many more molecules is even more convincing. For example, if one
mole of a gas (6x1023 molecules) were present in the same partitioned volume, the
probability for reaching equilibrium is astronomically close to 1.0. The apparent
equilibrium state includes many unequal configurations where the number of
molecules is not exactly the same on each side, but is sufficiently close in relative
numbers that we cannot differentiate between those configurations. They all appear
to be in equilibrium.
Let’s consider an even simpler case to understand entropy and its relation to
equilibrium. Entropy was defined by Boltzmann (1872) at the microscopic level by
the mathematical function S = −k ∑ i pi ln pi where pi is the state probability of a
given configuration and k is an arbitrary constant. The negative sign is introduced so
that entropy by this function is always positive, recognizing that probabilities must be
between 0.0 and 1.0. There is nothing magical about this function, but it has the
properties we desire in that entropy is always positive and increases uni-directionally
36
as equilibrium is approached. For example, applying this function to the simplest
case of one molecule on the right hand side of the partition in Fig. 2-6, the probability
initially is 1.0 that the molecule will be at the right-hand side of the partition just after
it is removed and 0.0 that it will be at the left-hand side. Thus, entropy given by this
function is initially S = − k ( 0.0 ln 0.0 + 1.0 ln1.0 ) or zero. As time advances, the
probability that the molecule will be found at the right-hand side will decrease,
whereas it will increase at the left-hand side. At some later time for example, the
probabilities will be 0.9 and 0.1 giving S = − k ( 0.1ln 0.1 + 0.9 ln 0.9 ) or 0.33k.
Eventually at equilibrium, the molecule is equally-likely to exist on either side of the
partition and entropy will achieve its maximum value of
S = − k ( 0.5ln 0.5 + 0.5ln 0.5 ) = − k ln pi or 0.69k.
Once equilibrium has been reached, entropy remains fixed (because the
probabilities are constant) as long as there are no other outside influences, such as the
expansion of volume or addition of more molecules to the box in Fig. 2-6. For
example, if a third equal volume is added to the other two, the probability that the
molecule is contained in either is 0.33. The entropy at equilibrium is then
S = −k ( 0.33ln 0.33 + 0.33ln 0.33 + 0.33ln 0.33) = − k ln pi or 1.11k, which is greater than
the previous value of 0.69k. Thus, an increase in volume increases the entropy at
equilibrium since the equilibrium probabilities are decreased.
The microscopic entropy function defined by Boltzmann must be related to
macroscopic thermodynamic properties for it to be practically useful. Fortunately, it
is possible to relate entropy to changes in internal energy and volume (see for
example, Schmidt et al. 2006). The easiest way to understand this is to think of the
labeled molecules in Fig. 2-6 as representing different energy quanta. We can then
assign probabilities for states based on the energy levels found on the left and right
hand side of the box. For example, if the molecules are initially very near absolute
zero, each molecule would have similar individual energies equal to the group
average energy. The internal energy in the closed volume is then the sum of the
individual molecule energies. The probability of finding a molecule at that energy
state on the right-hand side of the box is 1.0 and will not change with time unless
heat is added.
When heat is added to the box, the statistical distribution of individual
molecule energies around a group average energy will spread out according to the
Maxwell-Boltzmann distribution. That energy distribution shows that there will be a
wider range of possible individual particle energies for a group of molecules that
have a greater internal energy (or larger temperature). Hence, more possible energy
levels or configurations results in lower probabilities, thereby increasing entropy in a
similar fashion as in the molecule example of Fig. 2-6. This relationship between
increasing entropy with increasing internal energy can be expressed by the partial
derivative at constant volume ( ∂S ∂U )V = 1 T . That is, the entropy change with
internal energy is inversely proportional to temperature when the volume is fixed and
will be greatest at temperatures near absolute zero. Similar logic leads to the
relationship that for constant internal energy ( ∂S ∂V )U = P T . Because all properties
37
in these relationships are state functions (independent of the path), they can be
combined to give the macroscopic definition of entropy as
dS = ( ∂S ∂U )V dU + ( ∂S ∂V )U dV = (1 T ) dU + ( P / T ) dV . Again, there is nothing
magical about this definition except that like Boltzmann’s microscopic definition it
gives a value for entropy that is always positive and increasing monotonically
towards equilibrium.
In summary, we have shown that the equilibrium state is the state of
maximum entropy for an isolated system and statistically the most probable state.
For example, it is possible, although highly improbable, that a lake of water will
freeze on a hot summer day. The second law does not rule out this possibility, but
instead gives us the most likely direction that nature will take. A good definition of
the second law therefore is that the entropy of an isolated system will always increase
from a state of high probability (well-ordered) to a state of lower probability (more
disordered) until it reaches a state of minimum probability at equilibrium. This
increase (or creation) of entropy is always related to initial gradients in the system,
and overtime the rate of increase of entropy decreases as the fluids mix or heat is
exchanged.
An entropy balance is used to mathematically represent the tendency of
processes to approach equilibrium. The steps to write the entropy balance are similar
to those for the first law of thermodynamics, except that entropy is created as fluids
mix or heat is exchanged as illustrated in Figure 2-6. Entropy for a system that is not
isolated can increase or decrease depending on the direction and magnitude of mass
flow and heat exchange with its surroundings. A statement of the entropy balance is
⎧ Rate of ⎫ ⎧ Net rate ⎫ ⎧ Rate of ⎫

⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎨ accumulation =
⎬ ⎨ of entropy +
⎬ ⎨ generation of entropy ⎬ (2.4-1)
⎪of entropy in V ⎪ ⎪ transported into V ⎪ ⎪ inside ⎪
⎩ ⎭ ⎩ ⎭ ⎩ V ⎭
where V is an arbitrary volume as in Fig. 2-4. By analogy to the procedure for the
first law of thermodynamics in Section 2-3, the accumulation term in Eq. (2.4-1) for a
stationary V can be written as
⎧ Rate of ⎫
⎪ ⎪ d ⎧Total entropy ⎫ ∂ NP
⎨ accumulation ⎬ = ⎨
in V
⎬= ∫V ∂t ∑ (ε j ρ j S j )dV (2.4-2)
⎪of entropy in V ⎪ dt ⎩ ⎭ j =1
⎩ ⎭
where the total entropy includes contributions from all phases and S j is the specific
entropy of phase j. Equation (2.4-1) can now be expressed as
d NP
∫V dt ∑
j =1
(ε j ρ j S j )dV = − ∫V ∇iSdV + ∫V σ G dV . (2.4-3)
38
The first term on the right hand side is the entropy flux term. The second term is the
rate of entropy generation per unit bulk volume.
The net rate of entropy transport into volume V is the result of heat and mass
transfer into V from its surroundings. Entropy transport owing to convection of mass
is handled similarly to energy. Further, from the definition of entropy, the entropy
transport owing to heat exchange at fixed volume is proportional to the conductive
heat flux divided by the temperature at which that transfer occurs, that is, dS = dq / T
since dU = dq from the conservation of energy (see also Section 2.6). Thus, the total
entropy transport owing to mass convection and heat exchange is
NP
qc
S = ∑ ρ ju j S j + . (2.4-4)
j =1 T
Similar to the energy balance, we have invoked the requirement of local thermal
equilibrium in Eq. (2.4-4) by taking the temperature T to be the same in all phases
within the REV. Substitution of Eq. (2.4-4) into Eq. (2.4-3) gives,
d NP
( )
N
∫V dt ∑
q
∑
P
ε ρ S
j j j dV = − ∫ ∇ i ρ j u j S j dV − ∫ ∇i c dV + ∫ σ G dV . (2.4-5)
j =1
V
j =1
V T V
The desired result is obtained after collecting all terms under the same volume
integral and making the integrand zero because V is again arbitrary,
∂ NP NP
∑
∂t j =1
( ε j ρ j S j ) = −∇i ∑ ρ j u j S j − ∇i c + σ G .
q
T
(2.4-6)
j =1
Analogous to the procedure in Sec. 2-4, Eq. (2.4-6) could also be obtained for a time-
varying V.
Second Law of Thermodynamics
The inequality σ G ≥ 0 is the second law of thermodynamics, the equality holding at

equilibrium. This simple condition provides the restriction on bi-directional transport
of energy. The creation of entropy is related to the gradients within the system.
The relationship between entropy generation and gradients within the system
is most easily seen by assuming pure single-phase flow and neglecting changes in
kinetic or potential energies. The mass, energy, and entropy balances become
respectively,
∂
(φρ ) = −∇i( ρ u )
∂t
39
∂
(φρU ) = −∇i( ρ uU ) − ∇i( Pu ) − ∇i qc
∂t
and,
∂ 1 q
(φρ S ) = −∇i( ρ uS ) − ∇i qc + c2 i∇T + σ G
∂t T T
where the conduction term in the entropy equation has been expanded and we used
H = U + P / ρ . These equations can be combined and rearranged using the total
derivative ( D / Dt = ∂ / ∂t + v i∇ ) to give,
⎛ DS 1 DU P DV ⎞ 1 qc
φρ ⎜ − − ⎟ = u i∇P + 2 i∇T + σ G (2.4-7)
⎝ Dt T Dt T Dt ⎠ T T
where V = 1/ ρ and under the assumption of constant porosity ∇iu = φρ DV / Dt .

The left-hand side of Eq. (2.4-7) gives temporal changes in thermodynamic
properties and is zero from the definition of entropy. The right-hand side of Eq.
(2.4-7) involves terms that depend on the gradient of pressure and temperature.
Thus, since entropy creation is the result of gradients, we must have
1 q
σG = − u i∇P − c2 i∇T (2.4-8)
T T
( )
Substitution of Darcy’s law for horizontal single-phase flow, u = − k / μ i∇P , and
Fourier’s law of heat conduction, qc = −h∇T , into Eq. (2.4-8)gives our desired result,
k h
σG = ∇P i ∇ P + 2 ∇ T i ∇ T (2.4-9)
Tμ T
Equation (2.4-9) shows that entropy creation is proportional to the square of the
pressure and temperature gradients. Thus, entropy creation is minimized when
gradients are small, as they are when equilibrium is approached. Last, Eq. (2.4-9)
shows that σ G ≥ 0 since the properties in Eq. (2.4-9) and the square of the gradients
are always positive. The expression for entropy creation can be extended to account
for multiple components, diffusive flux, and other processes.
2-5 SPECIAL CASES OF THE STRONG FORM
We now turn to several important special cases of the phase conservation equations,
Eqs. (2.2-31), and the overall compositional equations, Eqs. (2.2-4) and Tables 2-2
and 2-3 (see Fig. 2-1). Each special case is applied in practice to describe various
EOR processes occurring in permeable media fluid flow. These special cases can be
40
accurately approximated by much simpler forms of the above general equations with
fewer and simpler associated auxiliary equations and boundary conditions. All flows
discussed here are in local thermodynamic equilibrium.
Fractional Flow Equations
Consider isothermal flow ( T constant) with no interphase mass transfer so that

rmij = 0 (including sorption). Because flow is immiscible, only the fluid saturations
change because phase compositions are fixed. Thus, we apply the above assumptions
to the phase conservation equations, Eq. (2.2-31), which were derived from Eq.
(2.2-4) (see Fig. 2-4). For this case, the energy conservation equation and the solid
phase equation are trivial, and Eq. (2.2-31) reduces to Eq. (2.2-32).
∂
∂t
(φ S j ρ j ) + ∇i( ρ j u j ) = 0 j = 1, . . . , N P . (2.2-32)
We specialize the equation next for one-dimensional linear flow in a medium

dipping at a constant angle α and constant rock and fluid properties ( φ constant
temporally and ρ j constant temporally and spatially). Because porosity and phase
densities are constant, these properties can be removed from the respective
derivatives and the phase densities cancelled. Equation (2.2-32) is then further
simplified to
∂S j ∂u j
φ + = 0, j = 1, . . . , N P . (2.5-1)
∂t ∂x
To eliminate the need to solve for pressure, Eq. (2.5-1) is usually written in
terms of a fractional flow function, which can be defined for the case of equal phase
pressures (Pcjn = 0) as
uj λrj ⎡ kg sin α N ⎤
∑
P
fj = = N ⎢1 − λrk ( ρ j − ρ k ) ⎥ , (2.5-2)
u ⎣ u ⎦
∑ rk
P
k =1
λ
k =1
where u = ∑ u j and α is the dip angle tanα = dDz/dx. The right side of Eq. (2.5-2)
NP
j =1
results by substitution of Darcy’s law into the fractional flow definition. This
equation will be derived and developed further in Chap. 5.
It is easily shown (Chap. 5) by summing the phase equations given in Eq.
(2.5-1) that u is a function of time only and fj is a function of saturation only. The
sum being a form of the pressure equation, we can write Eq. (2.5-1) in a final form.
∂S j u ∂f j
+ = 0, j = 1, . . . , N P . (2.5-3)
∂t φ ∂x
41
To solve Eq. (2.5-3) for the phase saturations Sj(x, t), the total volumetric fluid flux u
injected at the inflow boundary and the experimentally measured fractional flow
dependences of NP – 1 phases (note ∑ Nj=1 f j = 1 ) are needed. Buckley and Leverett
P
(1942) first solved this equation for two-phase flow, and the resulting estimation of
waterflood oil recovery is called the Buckley-Leverett theory (see Chap. 5). Other
similar cases, including three-phase flow and compositional effects such as
interphase mass transfer and adsorption, have been solved in closed form (e.g. Pope,
1980). We discuss these solutions in detail in Chaps. 7 to 9.
Miscible Flow
The above fractional flow case applies to the simultaneous flow of immiscible fluids.
We now treat the analogous but opposite case of many components flowing
simultaneously in a single fluid phase in isothermal flow. Thus only one phase flows
regardless of composition, but both convection and dispersion of these components
must be included. Miscible processes of interest include (1) true (first-contact)
miscible displacement of oil by a solvent from a reservoir; (2) chromatographic
processes of various sorts such as analytical chromatography, separation
chromatography, ion exchange processes, and adsorption of chemicals as they
percolate through soils and other naturally occurring permeable media; (3) leaching
processes such as the in situ mining of uranium; and (4) chemical reaction processes
of many types in fixed-bed reactors.
Equation (2.2-6) for single-phase flow is
∂ (φρωi ) ∂
+ [(1 − φ ) ρ sωis ] + ∇i ⎡⎢ ρωi u − φ Ki i∇ ( ρωi ) ⎤⎥ = Ri i = 1, . . . , N C . (2.5-4)
∂t ∂t ⎣ ⎦
This equation arises because the saturations for all phases but one are zero and the
saturation of the remaining phase is one. The second subscript j is now superfluous
and has been dropped. The auxiliary Eqs. (2.2-11), (2.2-12), (2.2-14), (2.2-15),and
(2.2-17) through (2.2-23) are still needed, but the others are no longer pertinent. The
principal one of these, Eq. (2.2-11) or Darcy’s law, has a considerably simpler form
as well, namely,
k
u = − i(∇P + ρ g ) . (2.5-5)
μ
Because the relative permeability is now constant (typically unity), it is lumped with
k.
For miscible solvents (see Chap. 7), the sorption term, the second term in Eq.
(2.5-4), is negligible. That assumption coupled with no chemical reactions (Ri = 0),
gives
42
∂ (φρωi )
+ ∇i ⎡⎢ ρωi u − φ Ki i∇ ( ρωi )⎤⎥ = 0, i = 1, . . . , N C . (2.5-6)
∂t ⎣ ⎦
A special one-dimensional linear case of Eq. (2.5-6) is obtained when the

porosity is constant, and K i is a constant. Letting Ci = ρωi be the mass concentration
of component i, it follows that
∂Ci ∂C ∂ 2 Ci
φ + u i = φ K li , i = 1, . . . , N C , (2.5-7)
∂t ∂x ∂x 2
where Kli, the longitudinal dispersion coefficient, is now a scalar,
Di αl | u |
K li = + (2.5-8)
τ φ
as a special case of the more general definition given by Eq. (2.2-2). Moreover, u is
at most a function of time, depending on the boundary conditions specified. Di is
usually taken as a constant, yielding the linear convection–diffusion (CD) equation,
which is alternatively termed the advection-dispersion equation. Several closed-form
solutions for simple initial and boundary conditions are available for the CD equation
(see Chaps. 5 and 7).
Chromatographic Equations
Several chromatographic processes are special cases of Eq. (2.5-7). We must restore
the Cis term ( Cis = ρωis ) that describes the accumulation of component i owing to
sorption reactions, for this is the essence of a chromatographic process. These
sorption reactions may be adsorption, the exchange of one ion by another on the solid
substrate, or precipitation–dissolution reactions (see Chaps. 8 to 10 and Lake et al.,
2002). All these processes lead to selective separation of the components as they
percolate through the permeable medium. Dispersion does not alter the separation in
chromatographic columns, so we neglect the second-order term, a step that results in
a set of strongly coupled (via the sorption term) first-order partial differential
equations
∂Ci ∂C ∂C
φ + (1 − φ ) is + u i = 0, i = 1, . . . , N C . (2.5-9)
∂t ∂t ∂x
For linear sorption Eq. (2.5-9) can be rewritten in terms of a retardation

factor by first collecting like terms:
∂ ∂C
(φ Ci + (1 − φ )Cis ) + u i = 0, i = 1, . . . , N C . (2.5-10)
∂t ∂x
43
The retardation factor for each component is defined as
(1 − φ )Cis (1 − φ )
Di = = ρ s K di (2.5-11)
φ Ci φ
ωis
where K di = is the partition coefficient for component i that relates the mass
Ci
fraction of the component adsorbed onto the solid divided by the concentration of the
component in the single-phase mixture. Substitution of the definition for the
retardation coefficient into Eq. (2.5-10) gives
∂ ∂C
∂t
(φ Ci (1 + Di ) ) + u ∂xi = 0, i = 1, . . . , N C . (2.5-12)
When the retardation factor and porosity are temporally constant,
∂Ci u ∂Ci
φ + = 0, i = 1, . . . , N C . (2.5-13)
∂t 1 + Di ∂x
This equation and Eq. (2.4-7) with dispersion neglected are nearly identical. The
only difference is that the flow rate is divided by 1 + Di , resulting in an effective flow
rate of uei = u (1 + Di ) for each component. The retardation factor is aptly named
because it causes an apparent reduction in the velocity at which the component
moves through the chromatographic column. With no sorption, the retardation
factor is zero and the component moves at the average flow rate of the fluid. Because
of the slowing of the velocity, the retardation factor is sometimes called a delay
factor (Lake et al., 2002).
Semimiscible Systems
In several EOR applications, a description of flow in permeable media based on

strictly miscible or immiscible flow is insufficient. For these situations, the equations
in Table 2-2 reduce to a simpler form consistent with the known complexities of the
flow behavior. As an example of this, consider the isothermal flow of NC components
in up to NP phases in the absence of chemical reaction. Such flows are characteristic
of solvent (see Chap. 7) and micellar-polymer flooding (see Chap. 9) EOR
applications.
We first assume that the change in pressure over the displacement length has
negligible effect on phase behavior. Equation (2.2-6) may then be divided by the
respective pure component density ρio to give
44
∂⎛ NP
⎞ ⎛ NP
⎞
⎜ φ ∑ Cij S j + (1 − φ )Cis ⎟ + ∇i⎜ ∑ Cij u j − φ S j K ij i∇Cij ⎟ = 0
∂t ⎝ j =1 ⎠ ⎝ j =1 ⎠
i = 1, . . . , N C (2.5-14)
where, under the additional assumption of ideal mixing, Cij = ρ j ωij / ρio is the volume
fraction of component i in phase j. Ideal mixing states that the volume of a mixture is
equal to the sum of the pure component volumes weighted by the component mass
fractions (Chap. 4).
We can sum Eq. (2.5-14) over the NC components as for the development of
Eq. (2.2-33), that is,
∂ ⎛ NC ⎛ N P ⎞ NC
⎞ ⎛ NC ⎛ N P ⎞ ⎞
⎜⎜ φ ∑ ⎜ Cij ∑ S j ⎟ + (1 − φ )∑ Cis ⎟⎟ + ∇i⎜⎜ ∑ ⎜ Cij ∑ u j ⎟ ⎟⎟ = 0,
∂t ⎝ i =1 ⎝ j =1 ⎠ i =1 ⎠ ⎝ i =1 ⎝ j =1 ⎠ ⎠
i = 1, . . . , N C , (2.5-15)
where the dispersive flux term vanishes. The additional assumption of constant
porosity in time leads to the result that
⎛ NP ⎞
∇i⎜ ∑ u j ⎟ = ∇iu = 0 , (2.5-16)
⎝ j =1 ⎠
where we have used Eq. (2.2-16), ∑ iN=1 Cij = 1 , and ∑ iN=1 Cis is constant. For one-
C C
dimensional flow, Eq. (2.5-16) implies that the total flow rate is only a function of
time. Thus, Eq. (2.5-16) can be used along with the definition of fractional flow
(2.5-2) to write Eq. (2.5-14) in one-dimensional form.
∂ ⎛ NP ⎞ ∂ ⎛ NP ⎞ ∂ ⎛ NP ∂Cij ⎞
⎜ ∑ ij j
φ C S + (1 − φ )Cis ⎟ + u ⎜ ∑ ij j ⎟ − ⎜ ∑ φ S j K lij
C f ⎟ = 0,
∂t ⎝ j =1 ⎠ ∂x ⎝ j =1 ⎠ ∂x ⎝ j =1 ∂x ⎠
i = 1, . . . , N C . (2.5-17)
Even with the above assumptions, Eq. (2.5-17) is still fairly general and must
be solved simultaneously with Darcy’s law and with the definitions for relative
mobility, capillary pressure, mass fractions, saturations, equations of state, and
equilibria relations, [Eqs. (2.2-11), (2.2-16), (2.2-21), (2.2-22), and (2.2-23)]. This
form is particularly convenient because many cases of binary and ternary phase
equilibria are more conventionally represented as volume fractions rather than mass
fractions (see Chap. 4).
We often define the overall component volume fraction to be Ci = ∑ Nj=1 Cij S j P
and the overall component fractional flow Fi = ∑ Nj=1 Cij f j . With those definitions,
P
constant porosity, and the assumption of dispersion-free flow with no sorption, Eq.
(2.5-17) becomes
45
∂Ci ∂F
φ + u i = 0, i = 1, . . . , N C . (2.5-18)
∂t ∂x
This equation is used in Chap. 7 to develop analytical solutions and basic insights for
miscible flooding theory.
Richards Equation for Unsaturated Flow in Aquifers
The general equations developed here also apply to groundwater flow. As an

example, consider the flow of water in the unsaturated (i.e. where both water and air
exist) zone of aquifers. Aquifers are divided into two regions, an unsaturated region
near the Earth’s surface where both water and air are present and a saturated region
near and below this where only the aqueous phase is present. The surface separating
the two zones is near the water table. The water table is also known as the phreatic
surface because water freely flows to this level in wells drilled through the aquifer.
We begin the development of the unsaturated flow equation by assuming no
mass transfer rmij = 0 in the phase conservation equations, Eqs. (2.2-31):
∂
∂t
(φ S j ρ j ) + ∇i( ρ j u j ) = 0 j = 1, . . . , N P , (2.2-32)
where there are only two phases, an aqueous phase and an air phase. The two phases
are immiscible; hence, phases and components are synonymous in this example.
Because air near the Earth’s surface is at low pressure, its density and viscosity are
near zero and it provides little resistance to water flow (except through a physical
restriction in the pore volume available for water flow). This is equivalent to
assuming that the air phase is at atmospheric pressure everywhere in the aquifer.
Thus, since only the aqueous phase flows, the solution to the air phase equation is
trivial and can be dropped along with the subscripts for the aqueous phase.
∂
(φ S ρ ) + ∇i( ρ u ) = 0 . (2.5-19)
∂t
The aqueous phase density is also assumed constant, a good assumption since the
water pressure remains near atmospheric even under flowing conditions. Equation
(2.5-19) becomes
∂
(φ S ) + ∇iu = 0 . (2.5-20)
∂t
Richards equation is often formally written in terms of the moisture content

( θ = φ S ), hydraulic head, and the hydraulic conductivity. The hydraulic head, which
is often defined with reference to atmospheric pressure and surface elevation, is
46
P − Patm
h= + z − z gs . (2.5-21)
ρg
The hydraulic conductivity tensor is defined as K = k ρ g / μ . With those definitions

and the use of Eq. (2.2-11) , Eq. (2.5-20) gives Richards equation,
∂θ
∂t
(
− ∇i kr K i∇h = 0 , ) (2.5-22)
where ∇h = ∇P ρ g and kr is the relative permeability of the aqueous phase, which is

a function of the water saturation (moisture content). Note that although we dropped
the air phase equation, the air phase does impact the flow of the aqueous phase
through relative permeability.
Philips (1957) wrote Richards equation in a form commonly used by soil
scientists. He split the hydraulic head into two parts: the suction head and elevation
head where ∇h = −∇ψ + ∇z and ψ = PC ρ g = − P ρ g . The capillary pressure is equal
to the negative of the water pressure because the air pressure is assumed constant at
atmospheric (taken as zero). The suction head is always positive because water is in
tension above the phreatic surface, e.g. the aqueous phase pressure is less than
atmospheric above the water table. The unsaturated flow equation used by Phillips as
derived from Eq. (2.5-22) is then
∂θ
∂t ( ) (
+ ∇i kr K i∇ψ = ∇i kr K i∇z . ) (2.5-23)
For Cartesian coordinates aligned in the principal flow directions, the term on the
right side of Eq. (2.5-23) can be simplified to give
∂θ
∂t ( ∂
)
+ ∇i kr K i∇ψ = ( kr K zz ) ,
∂z
(2.5-24)
where K zz is the component of the hydraulic conductivity in the z-direction. Finally,

we can rewrite Eq. (2.5-24) in terms of the capillary diffusivity D = − K Ψ ′ , where ψ ′
is the derivative of the suction head with respect to the moisture content.
∂θ
∂t ( ∂
)
= ∇i kr Di∇θ + ( kr K zz ) .
∂z
(2.5-25)
The capillary diffusivity is positive and has units of length squared per time (hence
its name). Phillips used Eq. (2.5-25) to develop semi-analytical solutions for rainfall
infiltration.
47
Standard Black-Oil Equations
A common representation (Peaceman, 1977) of the flow of fluids in oil and gas
reservoirs is the “black oil” equations wherein up to three phases, aqueous (j = 1),
oleic (j = 2), and gaseous (j = 3), flow simultaneously. The aqueous and gaseous
phases each consist of a single pseudocomponent, water (i = 1) and gas (i = 3),
respectively. The oleic phase consists of two pseudocomponents, oil (i = 2) and a
dissolved gas component. The components are pseudocomponents because each is
really a group of true components whose composition remains constant.
The use of pseudocomponents is quite common in EOR descriptions, since
this simplification often results in greater understanding of the processes and can
significantly reduce the computational effort involved in numerical simulation with
little loss of accuracy. Because of these advantages, the black-oil equations are used
often to model many EOR processes, including those where the black-oil
assumptions may not strictly apply. Black-oil equations, for example, may not
accurately model miscible EOR processes or other processes that significantly
deviate from the black-oil assumption of pseudocomponents having constant
composition. In those cases, more rigorous fully compositional simulators are used,
although they require significant computation time.
Figure 2-6 illustrates the black-oil assumptions. The model relates downhole
reservoir volumes to standard temperature and pressure (STP) conditions.
Definitions for STP vary slightly, but typical values are 14.7 psia and 60oF.
Formation volume factors are defined that express the ratio between volumes at STP
conditions and reservoir conditions as indicated by the dashed lines in Fig. 2-6. The
solution gas-oil ratio is the volume of gas that evolves from a known volume of oil at
STP conditions.
48
Figure 2-7: Illustration of standard black-oil assumptions. STP is standard

temperature and pressure.
The black-oil equations are compositional equations in that they are

developed from the overall compositional equations, Eq. (2.2-6). We invoke the
following additional assumptions: no reaction ( Ri = 0 ), no dispersion ( K ij = 0 ), and
no adsorption ( ωis = 0 ). The black oil assumption itself becomes the following: the
aqueous phases contains only water ( ω11 = 1 , ω21 = 0 , ω31 = 0 ), the oleic phase
contains only oil and gas ( ω12 = 0 , ω22 > 0 , ω32 > 0 ), and the gas phase contains only
the gas component ( ω13 = 0 , ω23 = 0 , ω33 = 1 ). For component 1 (the water
component), Eq. (2.2-6) becomes
∂
(φρ1 S1 ) + ∇i( ρ1u1 ) = 0 . (2.5-26)
∂t
For component 2 (the oil component), Eq. (2.2-6) gives
∂
(φρ 2 S2ω22 ) + ∇i( ρ 2ω22u2 ) = 0 . (2.5-27)
∂t
For component 3 (the gas component), Eq. (2.2-6) gives
∂
∂t
(φ [ ρ2 S2ω32 + ρ3 S3 ]) + ∇i( ρ2ω32u2 + ρ3u3 ) = 0 . (2.5-28)
49
We now want to eliminate the mass fractions and rewrite Eqs. (2.5-26) to
(2.5-28) in favor of formation volume factors and the solution gas-oil ratio. B1 is the
water formation volume factor (volume of a given mass of water at the prevailing
temperature and pressure divided by the volume of the same mass of water at
standard temperature and pressure).
ρ1s
B1 = .
ρ1
B2 is the oil formation volume factor (volume of a given mass of oil at prevailing
temperature and pressure divided by the volume of the same mass of oil at standard
conditions).
ρ 2s
B2 = .
ω22 ρ 2
B3 is the gas formation volume factor (volume of a given mass of gas at prevailing
temperature and pressure divided by the volume of the same mass at standard
temperature and pressure).
ρ3s
B3 = .
ρ3
Rs is the solution gas–oil ratio (volume of dissolved gas divided by volume of oil
phase, with both volumes evaluated at standard temperature and pressure).
ω32 ρ 2s
Rs = .
ω22 ρ3s
The above definitions may be introduced into the mass balances of Table 2-2
by dividing each by their respective standard densities ρ sj and recognizing that each
ρ sj is time and space independent. Equations (2.5-26) to (2.5-28) become
∂ ⎛ φS j ⎞ ⎛ uj ⎞
⎜ ⎟⎟ + ∇i⎜⎜ ⎟⎟ = 0, j = 1, 2
∂t ⎜⎝ B j ⎠ ⎝ Bj ⎠
and, for gas,
∂ ⎛ ⎡ S3 S 2 Rs ⎤ ⎞ ⎛ Rs u3 ⎞
⎜⎜ φ ⎢ + ⎥ ⎟⎟ + ∇i⎜ u2 + ⎟ = 0 .
∂t ⎝ ⎣ B3 B2 ⎦ ⎠ ⎝ B2 B3 ⎠
50
Modified black-oil models, where CO2 is added as a fourth component, are
sometimes used to account for oil displacement by carbon dioxide or other miscible
gases (Todd and Longstaff, 1972).
Steam Flooding Equations
As a special case of nonisothermal flow, we derive the “steam” equations given by

Stegemeier et al. (1977). We assume at most NP = 3 phases–an aqueous phase j = 1,
a hydrocarbon phase j = 2, and a gas phase j = 3–are present. Further, at most two
unreactive, nonsorbing pseudocomponents–water and oil–are present. We restrict
the hydrocarbon phase to contain only oil, and the aqueous and gaseous phases to
contain only water, assumptions that eliminate volatile hydrocarbons from the
equations. With these assumptions, the mass conservation equations, Eqs. (2.2-6),
become, for water,
∂
[φ ( ρ1S1 + ρ3 S3 )] + ∇i( ρ1u1 + ρ3u3 ) = 0 (2.5-29)
∂t
and, for oil,
∂
(φρ 2 S2 ) + ∇i( ρ 2 u2 ) = 0 . (2.5-30)
∂t
The dispersion terms are absent from these equations since the phase compositions
are constant. The conservation of energy Eq. (2.3-11) becomes
∂
[φ ( ρ1S1U1 + ρ 2 S2U 2 + ρ3 S3U 3 ) + (1 − φ ) ρ sU s ]
∂t
+∇i( ρ1 H1u1 + ρ 2 H 2 u2 + ρ3 H 3u3 ) − ∇i(kTt ∇T ) = 0 , (2.5-31)
where kinetic and potential energy terms have been neglected. We further neglect
pressure-volume work by letting the enthalpies equal internal energies and by taking
porosity to be constant. The derivatives in Eq. (2.5-31) may then be expanded to
give
∂( ρs H s ) ∂H ∂H 2 ∂H 3 ∂ ( ρ 3 S3 )
(1 − φ ) + φρ1 S1 1 + φρ 2 S 2 + φρ3 S3 + φ ( H 3 − H1 )
∂t ∂t ∂t ∂t ∂t
+ ρ1u1 i∇H1 + ρ 2 u2 i∇H 2 + ( H 3 − H1 )∇i( ρ3u3 ) + ρ3u3 i∇H 3 − ∇i(kTt ∇T ) = 0 , (2.5-32)
where Eqs. (2.5-29) and (2.5-30) have been used to eliminate several terms. The
term (H3 – H1) equals LV, the latent heat of vaporization of water, and we assume
enthalpies are independent of pressure dHj = CpjdT, where Cpj is the specific heat of
phase j. If the Cpj are constant, Eq. (2.5-32) becomes
∂T
M Tt + ( ρ1C p1u1 + ρ 2 C p 2 u2 )i∇T − ∇i(kTt ∇T )
∂t
51
⎧ ∂ ( ρ3 S3 ) ⎫
= − LV ⎨φ + ∇i( ρ3 S3u3 ) ⎬ , (2.5-33)
⎩ ∂t ⎭
where MTt is the overall volumetric heat capacity
M Tt = φ ( ρ1C p1 S1 + ρ 2 S 2 C p 2 ) + (1 − φ ) ρ s C ps . (2.5-34)
In this definition and in Eq. (2.5-33), the terms involving the gaseous phase density
ρ3 have been neglected since gas densities are usually much smaller than liquid
densities. The term on the right side of Eq. (2.5-33) represents the production or
destruction of the steam phase times the latent heat; it serves as a source term for the
energy equation. If steam disappears (condenses), the source term is positive, which
causes the temperature to rise. This results in a decrease in oil viscosity, the primary
recovery mechanism in thermal flooding (see Chap. 11). The latent heat, phase
pressures, and temperature are related through the vapor pressure curve for water and
capillary pressure relations.
2-6 OVERALL BALANCES
A common and useful way to apply the equations in the previous sections is in the
form of macroscopic or overall balances (Bird et al., 1960). Rather than balances
written for each point within the permeable medium, overall balances are spatially
integrated forms of the differential balances that thereby apply to finite volumes
within a reservoir, such as a cell in numerical simulation, or even, as is the case here,
the entire reservoir. Since the spatial component is absent from the equations, overall
balances are much simpler and far easier to integrate than differential balances. This
simplification is achieved at the expense of losing spatial detail of the concentration
variables; therefore, to be useful, overall balances must be supplemented with
independently derived or analytical correlations.
Material Balance
To derive the overall mass balance for component i, begin with the weak form
written on volume V in the compositional form of Eq.(2.2-4). The result is
∂Wi
∫V ∂t
dV + ∫ n i N i dA = ∫ Ri dV ,
A V
i = 1, . . . , N C . (2.6-1)
We then identify V with the total bulk volume Vb exclusive of the small
volumes associated with a finite number of sources and sinks embedded within. The
boundary of Vb may also be a fluid source or sink term, as would be the case of an oil
column abutting an aquifer or a free gas cap. If we assume the fluxes across the
boundaries of V are normal to the cross-sectional area, Eq. (2.6-1) becomes
52
dWi
Vb + N Pi − N Ji = Vb Ri , i = 1, . . . , N C , (2.6-2)
dt
where the superscript bar denotes volume-averaged quantities. The terms N Pi and
N Ji are the mass production and injection rates of component i for all the source and
sink terms in Vb. These are functions of time, since they are evaluated at fixed
positions on Vb. Ri is the volume-averaged reaction rate term of component i and is
also a function of time. We integrate Eq. (2.6-2) with respect to time to arrive at a
cumulative form of material balance
t
Vb (Wi − WiI ) = N Ji − N Pi + Vb ∫ Ri dt , i = 1, . . . , N C . (2.6-3)
0
In writing Eq. (2.6-3), we have assumed the cumulative injection and production of
component i at t = 0 is zero. In what follows, we ignore the cumulative reaction rate
term.
The most common application of Eq. (2.6-3) is to calculate Npi with Wi, WiI ,
and NJi specified. In particular, Wi (t ) is difficult to evaluate without actually
integrating the differential balances. This difficulty is circumvented by defining ERi,
the recovery efficiency of component i, as
N Pi − N Ji
ERi ≡ . (2.6-4)
VbWiI
ERi is the net amount of component i produced expressed as a fraction of the amount
of component initially present. For a component injected into the reservoir, ERi is
negative, but for a component to be recovered, oil or gas (to which it is almost
exclusively applied), ERi is positive and lies between 0 and 1. From Eq. (2.6-3), Wi is
Wi = WiI (1 − ERi ) . (2.6-5)
For either Eqs. (2.6-4) or (2.6-5) to be useful, ERi must be expressed

independently as a function of time. This is commonly done by decomposing ERi into
the displacement efficiency EDi and volumetric sweep efficiency EVi of component i
ERi = EDi EVi (2.6-6)

where
Amount of i displaced
EDi = (2.6-7)
Amount of i contacted
Amount of i displaced
EVi = (2.6-8)
Amount of i in place
53
These quantities in turn must be specified independently: EDi as a function of time
and fluid viscosities, relative permeabilities, and capillary pressures (see Chap. 5) and
EVi as a function of time, viscosities, well arrangements, heterogeneity, gravity, and
capillary forces (see Chap. 6).
Energy Balance
A similar procedure applied to the energy conservation Eq. (2.3-11) yields
d
Vb ( ρU ) + H P − H J = − ∫ qc indA = −Q (2.6-9)
dt A
where kinetic and potential energy terms have been neglected and H P and H J
represent the rates of enthalpy production and injection into and from V. Q is
positive when heat is lost from the reservoir. This equation, of course, is a version of
the first law of thermodynamics that, depending on the selection of V, will be useful
in calculating heat losses to wellbores (with the potential energy term restored) and
the overburden and underburden of a reservoir (see Chap. 11).
The time integrated or cumulative form of Eq. (2.6-9) is
Vb (( ρU ) − ( ρU ) I ) = H J − H P − Q (2.6-10)
from which we may define a thermal efficiency Ehs as the ratio of thermal energy
remaining in the volume Vb to the net thermal energy injected.
Vb (( ρU ) − ( ρU ) I ) Q
Ehs = = 1− (2.6-11)
HJ − HP HJ − HP
Equation (2.6-11) is used to independently calculate Q.

Another more familiar form of the overall energy balance pertains to phase
equilibrium thermodynamics (see Chap. 4). For brevity, we again neglect potential
and kinetic energies, but allow for both pressure-volume and compression-expansion
work (see exercise 2Q). The mass that enters volume V flows at a relative flow rate
of u fj , which is the difference between the actual flow rate and the rate of
deformation of V . Further, we rewrite the accumulation term to account for a
deforming V using the weak form. The result is
NP NP
d dV j
dt ∫V
( ρU ) dV = − ∫ A
n i ∑
j =1
ρ u
j fj H j dA − ∫ A
n i qc dA − ∑
j =1
Pj
dt
(2.6-12)
From left to right, the terms are the accumulation, energy flux owing to mass inflow,
energy flux owing to conduction, and compression-expansion work. The first three
terms in Eq. (2.6-12) must now be evaluated. The accumulation term is
54
d ∂U
( ρU ) dV =
dt ∫V
,
∂t
where U is the total energy within volume V,. The energy flux term is evaluated with
the assumption that the enthalpy of the flowing mass for any phase is constant over
the entrance or exit for mass flow. The result is
NP NP NP
− ∫ n i ∑ ρ j u fj H j dA = − ∑ H j ∫ n i( ρ j u fj ) dA = ∑ H j m j
A A
j =1 j =1 j =1
where m j is the total net mass inflow of phase j into the system. The sign is positive
to allow for the convention that mass flow rate is positive into the system. Finally,
the conduction term is,
− ∫ n iqc dA = Q .
A
With these additional assumptions, Eq. (2.6-12) reduces to
dU N P NP
dV j
= ∑ H j m j + Q − ∑ Pj . (2.6-13)
dt j =1 j =1 dt
When only one phase exists, Eq. (2.6-13) reduces to a more familiar form of the first
law of thermodynamics,
dU dV
= Hm + Q − P . (2.6-14)
dt dt
The pressure at the boundary of the system in Eq. (2.6-14) is equal to the pressure in
the system when pressure gradients are negligible within the system. Along with the
entropy balance discussed next, this equation will be used in Chap. 4 to determine the
equilibrium conditions for phase behavior calculations.
Entropy Balance
To derive the overall entropy balance, we start by rewriting Eq. (2.4-5) as
N
( )
NP
∑
P
d q
dt ∫V
ε j ρ j S j dV = − ∫ A
n i ∑ ρ j u j S j dA − ∫ n i c dA + ∫ σ G dV .
A T V
(2.6-15)
j =1 j =1
Following the same procedure as for the overall energy balance,

55
NP
dS Q
= ∑ S j m j + + SG , (2.6-16)
dt j =1 T
where we have assumed that both temperature and phase entropies along the system
boundary are independent of position. The term SG is now the total entropy
generated within the system, that is, SG = ∫ σ G dV .
V
When only one phase exists, Eq. (2.6-16) reduces to,
dS Q
= Sm + + SG . (2.6-17)
dt T
This equation will be used in Chap. 4 to determine the equilibrium conditions for
phase behavior calculations. As before, the Second Law of thermodynamics is
SG ≥ 0 , the equality applying at equilibrium.
2-7 SUMMARY
We will use the equations introduced and developed in this chapter in the remainder
of the text. Introducing all of the equations here eliminates repetitive derivation in
later chapters. The compilation also emphasizes one of the main points of this text:
the behaviors of all EOR and remediation processes are described by specializations
of the same underlying conservation laws. Solving these specializations and deducing
physical observations from the solutions will occupy much of the remainder of this
text. You should keep in mind that all of the relationships discussed above and hence
forth are ways of quantifying the chemical, physical and geologic bases of EOR.
EXERCISES
2A. Overall Compositional Equations. A simplified form of the conservation
equation can be obtained by summing Eq. (2.1-1) over all phases. The result
is the mass conservation equation for component i in volume V:
⎧ Rate of ⎫ ⎧ Net rate of i ⎫ ⎧ Rate of ⎫

⎪ ⎪ ⎪ ⎪ ⎪ ⎪
⎨accumulation ⎬ = ⎨ transported ⎬ + ⎨ production ⎬ , i = 1, . . . , N C .
⎪ of i in V ⎪ ⎪ into V ⎪ ⎪of i in V ⎪
⎩ ⎭ ⎩ ⎭ ⎩ ⎭
Show starting with the above expression that the differential form for the
component conservation equation is
∂Wi
+ ∇i N i − Ri = 0, i = 1, . . . , N C
∂t
56
where Wi is the overall concentration of each component, N i is the overall
flux of each component, and Ri is the overall production of each component.
Indicate all units in the derivation.
2B. Volatile Oil Equations. The text derived the "black oil" equations wherein
the oleic and gaseous phases are immiscible and constant composition except
for the solubility of a hydrocarbon gas in the oleic phase. The next step up in
complexity is the "volatile oil" equations that allow for the vaporization of
the oil into the gaseous phase (Walsh and Lake, 2003).
Derive the volatile oil equations in terms of the oil vaporization ratio Rv
where
S tan dard volumes of oil in gaseous phase

Rv =
S tan dard volumes of gas
State all assumptions and show all work.
2C. Hydrostatics. Show that for static (u j = 0) conditions, Eq. (2.2-11) reduces to
dPc
= ( ρ1 − ρ 2 ) g
dDz
for two-phase flow where Pc is the oil–water capillary pressure curve.
∑
NC
2D. Net Dispersion in a Phase. Show that in a given phase i =1
∇ ⋅ jDij = 0 when
dispersive flux is defined with respect to mass-averaged velocity. Use the
relations given for the average-component velocity owing to convective and
dispersive transport.
2E. Single-Phase Flow. Show that for the flow of a single phase ( j = 2) in the
presence of an immiscible, immobile phase ( j = 1) the isothermal mass
balance equations in one-dimensional radial coordinates reduce to
φ ct ∂P 1 ∂ ⎛ ∂P ⎞
− =0,
λr 2 k ∂t r ∂r ⎜⎝ ∂r ⎟⎠
r
where
ct = S1c1 + S2 c2 + c f
1 ⎛ ∂ρ j ⎞
cj = ⎜ ⎟
ρ j ⎝ ∂P ⎠T
57
1 ⎛ ∂φ ⎞
cf = ⎜
φ ⎝ ∂P ⎟⎠T
The above equation, the “diffusivity” equation, has been derived assuming
that terms of the form c2(∂P/∂r)2 are negligible and the reservoir is
homogeneous with constant thickness. The diffusivity equation forms the
basis for a large variety of well test techniques (Earlougher, 1977).
2F. Confined Flow of Water in an Aquifer. Show that the final result for
problem 2E can be rewritten for fully saturated soil as
1 ∂ ⎛ ∂h ⎞ 1 ∂h
⎜r ⎟ = ,
r ∂r ⎝ ∂r ⎠ η ∂t
where
ct = c1 + c f
K T
η= =
Ss S
kρg
K=
μ
S s = φ ct ρ g
The parameter S s is the specific storage coefficient for the aquifer in units of
inverse length. The dimensionless storage coefficient of the aquifer is equal
to , S = S s b where b is the thickness of the confined aquifer. A confined
aquifer is an aquifer fully saturated with water that cannot escape at the top
and bottom. The storage coefficient is defined as the volume of water stored
(or released) per unit area of aquifer per unit increase (or decrease) in the
hydraulic head h. The aquifer transmissivity is T = Kb . Finally, the
parameter η is known as the diffusivity constant.
2G. Unconfined Flow in an Aquifer. An unconfined aquifer is an unsaturated

aquifer that contains both air and water. The water table (phreatic surface
where capillary pressure is zero) can rise and fall as water is pumped from
the aquifer. A common assumption to model the rising or falling water table
is to assume that the hydraulic head is constant vertically and is y a function
only of the horizontal coordinates (this is a vertical equilibrium assumption
as is discussed in Chap. 6). Show that under this assumption, the equation
for flow in a 2-D unconfined homogeneous aquifer is
∂ ⎛ ∂h ⎞ S y ∂h
⎜h ⎟ = ,
∂x ⎝ ∂x ⎠ K ∂t
58
where
kρg
K=
μ
S y = φ (1 − S1r )
The parameter S y is the specific yield of the aquifer, defined as the volume of
water added (or lost) due to a unit increase (or decrease) in the water table
height per unit horizontal area of aquifer. It is dimensionless. The residual
water S1r is assumed constant whether the water table is rising or falling.
The water saturation above the water table is assumed constant at S1r .
To derive the unconfined flow equation start with the equation for a confined
aquifer given in 2F but expressed in x - and z -coordinates and integrate it
vertically from the base of the aquifer to the water table. Use the Liebnitz's
rule to evaluate integral expressions that result. Last, let the water table
surface be given by F ( x, z , t ) = z − a ( x, t ) = 0 , where a( x, t ) is the elevation of
the water table as measured from the assumed horizontal base of the aquifer.
The unit total derivative of F ( x, z , t ) must be zero, which allows one to relate
the movement of the water table in the horizontal and vertical directions to
the time rate of change of the water table. State all assumptions clearly.
2H. Convection-Dispersion Equation with Retardation. Show that when

dispersion is constant, the one dimensional convection-dispersion equation
with adsorption becomes
∂Ci u ∂Ci φ K l ∂ 2Ci

φ + − = 0, i = 1, . . . , N C ,
∂t 1 + Di ∂x 1 + Di ∂x 2
where the delay factor Di = (1 − φ )Cis / φ Ci is constant temporally and K l is the

longitudinal dispersion coefficient. State all assumptions required to obtain
this equation. How does adsorption affect the transport of a component?
2I. Simplified Combustion Model. Based on two-phase (j = 2 = liquid, j = 3 =

gas), four-component (i = 1 = water, i = 2 = oil [CnH2m], i = 3 = CO2, i = 4 =
O2), one-dimensional flow show that the energy conservation equations in
Table 2-3 reduce to
∂T ∂T ∂T ⎛ ∂T ⎞
M Tt + ( ρ 2 C p 2 u2 + ρ3C p 3u3 ) − ⎜ kTt ⎟ = φ S3 ΔH RXN
∂t ∂x ∂x ⎝ ∂x ⎠
where ∆HRXN is the heat of reaction for the gaseous phase reaction
(2n + m)O 2 + 2C n H 2 m → 2nCO 2 + 2mH 2 O

59
4
ΔH RXN = −∑ H io ri 3
i =1
Further assumptions for the above equation are there is only oil present in
liquid phase, no sorption or dispersion, ideal solution behavior (specific heat
of gaseous phase is the mass fraction – weighted sum of the component
specific heats), no heat of vaporization of oil (H22 = H23), enthalpies and
internal energies are equal, kinetic and potential energies are negligible, and
solid phase density and porosity are constant.
2J. Total Velocity. Show that for immiscible flow, constant porosity, and
incompressible flow
∇iu = 0
and that for one-dimensional flow the total velocity is a function only of
time.
2K. Continuity Equation. Derive Eq. (2.2-33) from the phase conservation
equations, Eq. (2.2-31).
2L. Volume Change on Mixing. The development of the semimiscible

equations assumed no volume change on mixing to obtain volume fractions.
When volume changes upon mixing, the total velocity can change spatially
and temporally and the use of volume fractions is not needed. Relax the no-
volume-change assumption and rederive the equations in a form similar to
Eq. (2.5-18). You may neglect dispersion.
2M. Formation Volume Factor. Show that in a black oil model equation that the
oil formation volume factor can also be written as
ρ 2s + RS ρ32
B2 =
ρ2
2N. Oil Recovery by NH3 Injection. An inventor believes that he has discovered
a new enhanced oil recovery technique that you are to simulate. The process
involves the injection of anhydrous (water-free) ammonia (NH3), which is to
dissolve in and vaporize some of the connate water. This mass transfer is
highly non-ideal in that there is a substantial heat of mixing. The heat
released by this mixing will raise the temperature of the crude and cause it to
flow more readily, just as in a thermal process.
There might also be some benefit from a pH increase, but the main objective
is to increase temperature. To this end, the inventor has commissioned a
commercial laboratory to do some displacements in a laboratory core. The
60
initial condition of the core is uniform temperature To and 100% water
saturation. The injected ammonia is pure and at the same temperature To.
Beginning with the general equations, develop a set of working equations

that will describe the experiment. There will several equations but these
should be as simple as possible without omitting the important features
discussed above. You should state all assumptions you made and ensure that
the number of equations and unknowns are equal.
2O. Work Done by Gravity. The derivation of Eq. (2.3-11) included potential
energy directly. This need not be the case if gravity work on a fluid element
in V is included in the source term by the work sum W = WPV + WG . To
account for the gravity work, take a scalar product of a velocity and the
gravity vector g ,
NP
ΔWG = ∑ ρ j u j i g ΔV .
j =1
The positive sign arises in this equation since a fluid phase flowing against
gravity (u j i g < 0) is having work done on it. Note the distinction between
the elemental forms in Eqs. (2.3-5) and the equation above. Equation (2.3-5)
is appropriate for work done against forces on the surface of V, whereas the
above equation is appropriate for work done against body forces.
Show that the potential energy terms in Eq. (2.3-11) can be derived by
inclusion as a work term, where the total work done by gravity in the volume
V is
NP
WG = ∫
V
∑ ρ u i gdV .
j =1
j j
Hint: You must use Eq. (2.2-33) and the identity

NP NP NP
∑ ρ u i g = − g ∑ ∇ i( ρ u D ) + g ∑ D ∇ i( ρ u ) .
j =1
j j
j =1
j j z
j =1
z j j
2P. First Law of Thermodynamics. Derive Eq. (2.6-12) for the first law of
thermodynamics that includes both compression-expansion and pressure-
volume work. Show that the equation reduces to Eq. (2.3-11) when the
system is stationary. Also, derive the energy equation when the system
deforms at the same rate as the phase velocities. Neglect kinetic and
potential energies in your derivations.
61
2Q. Thermal Diffusivity Equation. From Eq. (2.5-33), state the necessary
assumptions required to obtain the thermal diffusivity equation,
M Tt ∂T
∇ 2T = ,
kTt ∂t
where
M Tt = φρ1C p1 S1 + (1 − φ ) ρ s C ps .
93
Phase Behavior
and Fluid Properties
The phase behavior of crude oil, water, and enhanced oil recovery fluids is a common
basis of understanding the displacement mechanisms of EOR processes. Such
behavior includes the two- and three-phase behavior of surfactant–brine–oil systems,
the two or more phases formed in crude-oil–miscible-solvent systems, and the steam–
oil–brine phases of thermal flooding. This chapter is not an exhaustive exposition of
phase behavior. We concentrate on the aspects of phase behavior most pertinent to
EOR. (For more complete treatments of phase behavior, see Francis, 1963; Sage and
Lacey, 1939; and Standing, 1977.)
4-1 PHASE BEHAVIOR OF PURE COMPONENTS
In this section, we discuss the phase behavior of pure components in terms of

pressure–temperature (P-T) and pressure–molar-volume diagrams.
Definitions
A system is a specified amount of material to be studied. In other chapters, the word

system refers to the permeable medium, including the fluid with the pore space. In
this chapter, the word refers only to the fluids. With this definition, a system can be
described by one or more properties, any of several attributes of the system that can
be measured. This definition implies a quantitative nature to physical properties––
that is, they can be assigned a numerical value.
94
Properties are of two types: extensive properties, those dependent on the
amount of mass in the system (the mass itself, volume, enthalpy, internal energy, and
so on) and intensive properties, those independent of the amount of mass
(temperature, pressure, density, specific volume, specific enthalpy, phase
composition, and so on). Many times we designate intensive quantity by the modifier
specific (quality per unit mass) or by molar (quantity per unit mole). Thermodynamic
laws and physical properties are usually expressed in terms of intensive properties.
The most important intensive properties in this chapter are
ρ = the density, mass per volume (or g/cm3 in practical SI)

V = the specific volume, volume per mass (or the reciprocal of ρ)
VM = the specific molar volume, volume per amount (or m3/kg-mole in SI)
ρM = the molar density, moles per volume (or the reciprocal of VM )
Often the standard density of a fluid is given as the specific gravity, where
⎧ ρ
⎪ρ for liquids
⎪ water
γ =⎨ (4.1-1)
⎪ ρ for gases
⎪⎩ ρ air
All densities in Eq. (4.1-1) are evaluated at standard conditions of 273 K and
approximately 0.1 MPa. The petroleum literature uses other standards (60°F and 14.7
psia).
In all discussions of phase behavior, it is important to understand the
difference between a component and a phase. A phase is a homogeneous region of
matter. Homogeneous means it is possible to move from any point in the region to
any other without detecting a discontinuous change in a property. Such a change
occurs when the point crosses an interface, and thereby the system consists of more
than one phase. The three basic types of phases are gas, liquid, and solid, but of the
last two, there can be more than one type.
A component is any identifiable chemical entity. This definition is broad
enough to distinguish among all types of chemical isomers or even among chemical
species that are different only by the substitution of a radioactively tagged element.
Examples are H2O, CH4, C4H10, Na+, Ca2+, and CO32- . Natural systems contain many
components, and we are commonly forced to combine several components into
pseudocomponents to facilitate phase behavior representation and subsequent
calculations.
The relationship among the number of components NC, number of phases NP,
number of chemical reactions NR, and “degrees of freedom” NF of the system is given
by the Gibbs phase rule:
N F = NC − N P + 2 − N R (4.1-2)
95
In many textbooks, the number of components is defined to be NC – NR, the total
number of chemical species minus the chemical reactions involving them, or the
number of independent species. The resulting NF is the same. The +2 in Eq. (4.1-2)
accounts for the intensive properties temperature and pressure. If one or both of these
properties is specified, the equation must be reduced accordingly.
The meaning of NC and NP in Eq. (4.1-2) is clear enough, but NF invariably
requires some amplification. The degrees of freedom in the phase rule is the number
of independent intensive thermodynamic variables that must he fixed to specify the
thermodynamic state of all properties of the system. Intensive thermodynamic
variables include phase compositions (ωij in Chap. 2) as opposed to overall
compositions or volume fractions (Wi and Sj in Chap. 2), which are not
thermodynamic properties. The phase rule does not itself specify the values of the NF
variables, nor does it identify the variables; it merely gives the number required.
Intuitively, we expect a relationship among the three intensive properties,
temperature, pressure, and molar volume, for a pure component. Using density, molar
density, or specific volume in place of the molar volume would mean no loss of
generality. However, in two dimensions it is difficult to completely represent this
relationship, but we can easily plot any two of these variables.
Pressure–Temperature Diagrams
Figure 4-1 shows a schematic P-T plot for a pure component. The lines on the
diagram represent temperatures and pressures where phase transitions occur. These
lines or phase boundaries separate the diagram into regions in which the system is
Figure 4-1 Pure component PT diagram

(constant composition)
96
single phase. Specifically, the phase boundary separating the solid and liquid phases
is the fusion, or melting, curve, that between the solid and gas phases is the
sublimation curve, and that between the liquid and vapor phases is the vapor pressure
curve. Based on our definition of a phase, a discontinuous change in system
properties will occur when any phase boundary is crossed.
The phase transitions we refer to in this chapter are those of fluids in
thermodynamic equilibrium. Thus it is possible for a fluid in a particular phase to
momentarily exist at a P-T coordinate corresponding to another phase. But this
condition is not permanent since the material would eventually convert to the
appropriate stable equilibrium state.
From the phase rule we know that when two phases coexist, NF equals 1.
This can happen for a pure component only on the phase boundaries since a curve has
one degree of freedom. By the same argument, three phases can coexist at only a
single point in P-T space since NF = 0 for this condition. This single point is a triple
point, shown in Fig. 4-1 as the point where the three phase boundaries intersect.
Other triple points such as three solids may also exist. For pure components, the
phase rule says no more than three phases can form at any temperature and pressure.
Each phase boundary terminates at a critical point. The most interesting of
these is the critical point at the termination of the vapor pressure curve. The
coordinates of the critical point on a P-T plot are the critical temperature Tc and
critical pressure Pc, respectively. The formal definition of Pc is the pressure above
which a liquid cannot be vaporized into a gas regardless of the temperature. The
definition of Tc is the temperature above which a gas cannot be condensed into a
liquid regardless of the pressure. At the critical point, gas and liquid properties are
identical. Obviously, the region above the critical point represents a transition from a
liquid to gas state without a discontinuous change in properties. Since this region is
neither clearly a liquid nor a gas, it is sometimes called the supercritical fluid region.
The exact definition of the fluid region is arbitrary: Most texts take it to be the region
to the right of the critical temperature (T > Tc) though it would seem that defining it
to be the region to the right and above the critical point (T > Tc and P > Pc) would be
more consistent with the behavior of mixtures.
The behavior shown in Fig. 4-1 for a pure component is qualitatively correct
though less detailed than what can be observed. There can exist, in fact, more than
one triple point where solid–solid–liquid equilibria are observed. Water is a familiar
example of a pure component that has this behavior. Remarkably, observations of
multiple gas phases for pure components have also been reported (Schneider, 1970).
Such nuances are not the concern of this text, which emphasizes gas–liquid and
liquid–liquid equilibria. In fact, in all further discussions of phase behavior, we
ignore triple points and solid-phase equilibria. Even with these things omitted, the
P-T diagrams in this chapter are only qualitatively correct since the critical point and
the vapor pressure curve vary greatly among components. Figure 7-2 shows some
quantitative comparisons.
Critical phenomena do play an important role in the properties of EOR fluids.
If a laboratory pressure cell contains a pure component on its vapor pressure curve
97
(Fig. 4-1), the pure component exists in two phases (gas and liquid) at this point, and
the cell pressure is Pv. Thus the cell will contain two regions of distinctly different
properties. One of these properties being the density, one phase will segregate to the
top of the cell, and the other to the bottom. The phases will most likely have different
light transmittance properties so that one phase, usually the upper or light phase, will
be clear, whereas the other phase, the lower or heavy phase, will be translucent or
dark.
We can simultaneously adjust the heat transferred to the cell so that the
relative volumes of each phase remain constant, and both the temperature and
pressure of the cell increase if the fluid remains on the vapor pressure curve. For
most of the travel from the original point to the critical point, no change occurs in the
condition of the material in the cell. But the properties of the individual phases are
approaching each other. In some region near the critical point, the light phase would
become darker, and the heavy phase lighter. Very near the critical point, the interface
between the phases, which was sharp at the original temperature and pressure, will
become blurred and may even appear to take on a finite thickness. At the critical
point, these trends will continue until there is no longer a distinction between
phases–that is, two phases have ceased to exist. If we continue on an extension of
the vapor pressure curve, there would be a single-fluid phase and gradual changes in
properties.
Pressure–Molar-Volume Diagram
A way of representing how the discontinuity in intensive properties between phases

vanishes at the critical point is the pressure–molar-volume diagram. Figure 4-2
compares such a diagram with the corresponding P-T diagram. Both schematic plots
show isotherms, changes in pressure from a high pressure P1 to a lower pressure P2,
at four constant temperatures, T1 through T4.
At conditions (P, T)1, the pure component is a single-phase liquid. As
pressure decreases at constant temperature, the molar volume increases but only
slightly since liquids are relatively incompressible. At P = Pv(T1), the molar volume
increases discontinuously from some small value to a much larger value as the
material changes from a single-liquid to a single-gas phase. Since the change takes
place at constant temperature and pressure, this vaporization appears as a horizontal
line in Fig. 4-2(b). Subsequent pressure lowering again causes the molar volume to
increase, now at a much faster rate since the compressibility of the gas phase is much
greater than that of the liquid phase. The endpoints of the horizontal segment of the
pressure–molar-volume plot represent two coexisting phases in equilibrium with each
other at the same temperature and pressure. The liquid and vapor phases are said to
be saturated at P = Pv(T1).
At a higher temperature T2, the behavior is qualitatively the same. The
isotherm starts at a slightly higher molar volume, the vaporization at P = Pv(T2) is at a
higher pressure, and the discontinuous change from saturated liquid to saturated
vapor VM is not as large as at T1. Clearly, these trends continue as the isotherm
98
Figure 4-2 Schematic pressure–temperature and pressure–molar-volume diagrams
temperature approaches T3 = Tc. All isotherms on the pressure–molar-volume plot are

continuously nonincreasing functions with discontinuous first derivatives at the vapor
pressure line.
At the critical temperature, the two phases become identical, and the
saturated liquid and gas molar volumes coincide. Since this temperature is only
infinitesimally higher than one at which there would still be distinguishable liquid
and gas phases, the isotherm at T = Tc, the critical isotherm, continuously decreases
with continuous first derivatives. At the critical point P = Pc, the critical isotherm
must have zero slope and zero curvature, or
⎛ ∂P ⎞ ⎛ ∂2 P ⎞
⎜ ⎟ = ⎜ 2⎟ =0 (4.1-3)
⎝ ∂VM ⎠Tc , Pc ⎝ ∂VM ⎠Tc , Pc
These critical constraints follow from the physical argument given above and can
also be derived by requiring a minimum in the Gibbs free energy at the critical point
(Denbigh, 1968).
At isotherm temperatures above the critical temperature, T = T4 in Fig. 4-2,
the isotherm is monotonically decreasing with continuous first derivatives but
without points of zero slope or curvature.
The endpoints of all the horizontal line segments below the critical points in
the pressure–molar-volume plot define a two-phase envelope as in Fig. 4-2(b).
Though rarely done, it is also possible to show lines of constant relative amounts of
liquid and gas within the two-phase envelope. These quality lines (dotted lines in Fig.
4-2b) must converge to the critical point. The two-phase envelope on a pressure–
99
molar-volume plot for a pure component, which projects onto a line in a P-T
diagram, is not the same as the two-phase envelope on a P-T diagram for mixtures.
Both Figs. 4-2(a) and 4-2(b) are merely individual planar representations of the
three-dimensional relation among temperature, pressure, and molar volume. Figure
4-3 illustrates the three-dimensional character of this relation for water.
Finally, though we illustrate the phase envelope of a pure component on a
pressure–molar-volume diagram, discontinuities in properties below the critical point
are present in all other intensive properties except temperature and pressure (see Fig.
11-3, the pressure–enthalpy diagram for water).
4-2 PHASE BEHAVIOR OF MIXTURES
Because the purpose of EOR is to recover crude oil, an unrefined product, we need
not deal with the phase behavior of pure components except as an aid to
understanding mixtures. Since the phase behavior of hydrocarbon mixtures is so
complex, in this and the next section, we simply compare the phase behavior of
mixtures to pure components and introduce pressure–composition (P-z) and ternary
diagrams.
Pressure–Temperature Diagrams
For a multicomponent mixture, NF > 2 when two phases are present. Therefore, two
(or more) phases can coexist in a planar region in P-T space, compared to the single
component case, where two phases coexist only along a line in P-T space. Mixtures
have two phases in a region, or envelope, in P-T space (Fig. 4-4).
Consider, along with this figure, a change in pressure from P1 to P5 at
constant temperature T2. The phase envelope is fixed for constant overall composition
(ωi or zi). Since the indicated change is usually brought about by changing the
volume of a pressure cell at constant composition and temperature, the process is
frequently called a constant composition expansion.
From P1 to P3, the material in the cell is a single-liquid phase. At P3, a small
amount of vapor phase begins to form. The upper boundary of the phase envelope
passing through this point is the bubble point curve, and the y coordinate at this point
is the bubble point pressure at the fixed temperature. From P3 to P5, successively,
more gas forms as the liquid phase vaporizes. This vaporization takes place over a
finite pressure range in contrast to the behavior of a pure component. Continuing the
constant composition expansion to pressures lower than P5 would result in eventually
reaching a pressure where the liquid phase would disappear, appearing only as drops
in the cell just before this point. The pressure the liquid vanishes at is the dew point
pressure at the fixed temperature, and the lower boundary of the phase envelope is
the dew point curve.
For a pure component (Fig. 4-1), the dew and bubble point curves coincide.
Within the two-phase envelope, there exist quality lines that as before,
indicate constant relative amounts of liquid and vapor. The composition of the liquid
100
Figure 4-3 Schematic pressure-specific volume–temperature surface and projections

(from Himmelblau, 1982)
101
Figure 4-4 Schematic pressure–

temperature diagram for hydrocarbon
mixtures (constant composition)
and gas phases is different at each point within the envelope, and both change
continuously as the pressure decreases.
Phase compositions are not shown on the P-T plot. But we do know that the
liquid and gas phases are saturated with respect to each other in the two-phase
envelope. Hence at any T and P within the envelope, the liquid phase is at its bubble
point, and the gas phase at its dew point. The quality lines converge to a common
point at the critical point of the mixture though this point does not, in general, occur
at extreme values of the temperature and pressure on the phase envelope boundary.
The maximum pressure on the phase envelope boundary is the cricondenbar, the
pressure above which a liquid cannot be vaporized. The maximum temperature on the
phase envelope is the cricondentherm, the temperature above which a gas cannot be
condensed. These definitions are the same as for the critical point in pure component
systems; hence the best definition of the critical point for mixtures is the temperature
and pressure at which the two phases become identical.
For mixtures, there exists, in general, a pressure range between the
cricondenbar and Pc and between the cricondentherm and Tc where retrograde
behavior can occur. A horizontal constant pressure line in Fig. 4-4 at P = P4 begins in
the liquid region at T0 and ends in the fluid region at T4. As temperature is increased,
gas begins to form at the bubble point temperature T1 and increases in amount from
102
then on. But at T2, the amount of gas begins to decrease, and the gas phase vanishes
entirely at a second bubble point T3. From T2 to T3, the behavior is contrary to
intuition––a gas phase disappearing as temperature increases––and the phenomenon
is called retrograde vaporization.
Retrograde behavior does not occur over the entire range between the two
bubble point temperatures but only over the range from T2 to T3. By performing the
above thought experiment at several pressures, one can show that retrograde behavior
occurs only over a region bounded by the bubble point curve on the right and a curve
connecting the points of zero slope on the quality lines on the left (McCain, 1973).
Though not possible in the P-T diagram in Fig. 4-4, retrograde phenomena
are also observed for changes in pressure at constant temperature. This case, which is
of more interest to a reservoir engineer, happens when the cricondentherm is larger
than Tc and the constant temperature is between these extremes. This type of
retrograde behavior is a prominent feature of many hydrocarbon reservoirs, but it
impacts little on EOR.
We do not discuss the pressure–molar-volume behavior of hydrocarbon
mixtures in detail. The main differences between the behavior of pure components
and mixtures is that the discontinuous changes in VM do not occur at constant P, and
the critical point no longer occurs at the top of the two-phase region (see Exercise
4B). These differences cause interesting variations in the shape of pressure–molar-
volume diagrams for mixtures but, again, are not directly relevant to EOR.
Since EOR processes are highly composition dependent, the behavior of the
P-T envelope as the overall composition of the mixture changes is highly important.
Consider the dilution of a crude oil M4 with a more volatile pure component A as
shown in Fig. 4-5. As the overall mole fraction of A increases, the phase envelope
migrates toward the vertical axis, increasing the size of the gas region.
Simultaneously, the phase envelope shrinks as it approaches the vapor pressure curve
of the pure component A. There are, of course, an infinite number of mixtures (Fig.
4-5 shows only three) of the crude oil with A. Each mixture has its respective critical
point in P-T space, which also migrates to the critical point of the pure component on
a critical locus. The overall composition of a mixture at a critical point is the critical
mixture at that temperature and pressure.
Pressure–Composition Diagrams
The phase behavior of the dilution in Fig. 4-5 on a plot of mole fraction of
component A versus pressure at fixed temperature shows composition information
directly. Such a plot is a pressure composition, or P-z, plot. The P-z plot for the
sequence of mixtures in Fig. 4-5 is shown in Fig. 4-6. Since the P-T diagram in Fig.
4-5 shows only three mixtures and does not show quality lines, phase envelope
boundaries are represented at relatively few points in Fig. 4-6 (see Exercise 4C).
Starting at some high pressure in Fig. 4-5 and following a line of constant
temperature as pressure is reduced produces a dew point curve for mixture M1 at
103
Figure 4-5 Schematic dilution of a crude oil by a more volatile pure component
Figure 4-6 The pressure–composition

plot for the dilution in Fig. 4-5
104
pressure P6. Since this mixture is rich in component A, this point plots nearest the
right vertical axis in Fig. 4-6 at the pressure coordinate P6. Continuing down the
constant temperature line, at P5 the critical point for mixture M2 is encountered
(mixture M2 is the critical composition at this temperature and pressure). But this
point is also a second dew point for mixture M1; hence P5 plots at the same vertical
coordinate for both mixtures in Fig. 4-6 but with different horizontal coordinates. At
P4 there is a bubble point for mixture M3 and a dew point for M2. These points again
define the corresponding phase boundaries of the P-z plot in Fig. 4-6. The process
continues to successively lower pressures in the same manner. Each pressure below
the critical is simultaneously a bubble point and a dew point pressure for mixtures of
different overall compositions. The pressures P2 and P1 are the bubble and dew point
pressures of the undiluted crude oil. The two-phase envelope in Fig. 4-6 does not
intersect the right vertical axis since the fixed temperature is above the critical
temperature of the pure component A. The diagram shows the closure of the
two-phase envelope as well as a few quality lines.
Since the entire P-z diagram is at constant temperature, we cannot represent
the phase behavior at another temperature without showing several diagrams. More
important, the composition plotted on the horizontal axis of the P-z plot is the overall
composition, not either of the phase compositions. Thus horizontal lines do not
connect equilibrium mixtures. Such tie lines do exist but are, in general, oriented on a
horizontal line in a hyperspace whose coordinates are the phase compositions.
However, for binary mixtures, the tie lines are in the plane of the P-z plot, and the
critical point is necessarily located at the top of the two-phase region. Finally, though
Fig. 4-6 is schematic, it bears qualitative similarity to the actual P-z diagrams shown
in Figs. 7-10 through 7-12.
4-3 TERNARY DIAGRAMS
On a P-z plot, we sacrifice a degree of freedom (temperature) to obtain compositional

information. But the diagrams can show only the composition of one component, and
this representation is often insufficient for the multitude of compositions that can
form in an EOR displacement. A plot that represents more composition information
is the ternary diagram.
Definitions
Imagine a mixture, at fixed temperature and pressure, consisting of three components

1, 2, and 3. The components may be pure components. But more commonly in EOR,
they are pseudocomponents, consisting of several pure components. The composition
of the mixture will be a point on a plot of the mole fraction of component 3 versus
that of component 2. In fact, this entire two-dimensional space is made up of points
that represent the component concentrations of all possible mixtures.
We need to plot the concentrations of only two of the components since the
105
concentration of the third may always be obtained by subtracting the sum of the mole
fractions of components 2 and 3 from 1. This means all possible compositions will
plot into a right triangle whose hypotenuse is a line from the 1.0 on the y axis to the
1.0 on the x axis. Though ternary diagrams are on occasion shown this way (see Fig.
7-15), they are most commonly plotted so that the right triangle is shifted to an
equilateral triangle, as in Fig. 4-7.
Figure 4-7 Ternary Diagram
All possible ternary compositions fall on the interior of the equilateral

triangle; the boundaries of the triangle represent binary mixtures (the component at
the apex opposite to the particular side is absent), and the apexes represent pure
components. Thus in Fig. 4-7, point M1 is a mixture having 20%, 50%, and 30%
components 1, 2, and 3, respectively; point M2 is a binary mixture of 70% component
1 and 30% component 3, and point M3 is 100% component 2. Representing the
compositions in this manner is possible for any concentration variable (mole fraction,
volume fraction, mass fraction) that sums to a constant.
Ternary diagrams are extremely useful tools in EOR because they can
simultaneously represent phase and overall compositions as well as relative amounts.
The correspondence of the P-T diagram to the ternary diagram in Figs. 4-8 and 4-9
compares to the P-z diagram in Figs. 4-5 and 4-6. Here we consider a ternary system
consisting of components 1, 2, and 3, and consider the dilution of mixtures having
constant ratios of components 2 and 3 by component 1. Each dilution represents a
line corresponding to a fixed 2 : 3 ratio on the ternary in Fig. 4-9.
106
Figure 4-8 Schematic evolution of P-T diagram in three component systems
We want to follow the formation and disappearance of phases on the ternary

diagram at the fixed temperature and pressure indicated by the box in Fig. 4-8. For
the dilution of component 3 by component 1, the reference temperature and pressure
is above the critical locus in the upper left-hand panel (Fig. 4-8a). Thus the C1-C3
axis of the ternary indicates no phase changes. The C1-C2 binary dilution in the upper
right-hand panel (Fig. 4-8a) does encounter phase changes, and in fact, the reference
temperature and pressure is a bubble point for a mixture of 25% C1 and a dew point
for a mixture of 85% C1. These phase transitions are shown on the C1-C2 axis on the
107
Figure 4-9 Schematic ternary diagram of dilutions in Fig. 4-8
ternary. The dilution indicated in the middle left-hand panel (Fig. 4-8b) shows phase
transitions at 82% and 21%, respectively, which are also plotted on the ternary. For
the dilution of the 1 : 3 mixture, the critical locus passes through the fixed
temperature and pressure, and this composition, 25% C1, is the critical composition
of the ternary mixture. This composition is indicated on the ternary diagram in Fig.
4-9 as a plait point after the more common designation of the critical mixture in
liquid–liquid phase equilibria. At the fixed temperature and pressure, there can exist a
second phase transition––a dew point at 67% C1––at the same temperature and
pressure. After making several dilution passes through the ternary diagram, the points
where there are phase transitions define a closed curve in Fig. 4-9. This curve, the
binodal curve, separates regions of one- and two-phase behavior. Within the region
enclosed by the binodal curve, two phases exist, and outside this region, all
components are in a single phase.
Phase Compositions
One useful but potentially confusing feature of ternary diagrams is that it is possible
to represent the composition of the phases as well as the overall composition on the
same diagram. Consider an overall composition Ci on the inside of the binodal curve
in Fig. 4-10
Ci = Ci1 S1 + Ci 2 S2 , i = 1, 2, 3 (4.3-1)
108
Figure 4-10 Two-phase ternary equilibria
where Cij is the concentration of component i in phase j, and Sj is the relative amount
of phase j. By convention, we take phase 1 to be the C1-rich phase and phase 2 to be
the C1-lean phase. Since S1 + S2 = 1, we can eliminate S1 from two of the equations in
Eq. (4.3-1) to give
C3 − C31 C − C11
S2 = = 1 (4.3-2)
C32 − C31 C12 − C11
This equation says a line through the composition of phase 1 and the overall
composition has the same slope as a line passing through the composition of phase 2
and the overall composition. Both lines, therefore, are merely segments of the same
straight line that passes through both phase compositions and the overall
composition. The intersection of these tie lines with the binodal curve gives the phase
compositions shown in Fig. 4-10. The entire region within the binodal curve can be
filled with an infinite number of these tie lines, which must vanish as the plait point is
approached since all phase compositions are equal at this point. Of course, there are
no tie lines in the single-phase region.
Further, Eq. (4.3-2) implies, by a similar triangle argument, that the length of
the line segment between Ci and Ci1 divided by the length of the segment between Ci2
and Ci1 is the relative amount S2. This, of course, is the well-known lever rule, which
can also be derived for S1. By holding S2 constant and allowing Ci to vary, we can
construct quality lines, as indicated in Fig. 4-10, which must also converge to the
plait point as do the tie lines.
109
Tie lines are graphical representations of the equilibrium relations (Eq. 2.2-
11). Assuming, for the moment, the apexes of the ternary diagram represent true
components, the phase rule predicts there will be NF = 1 degrees of freedom for
mixtures within the binodal curve since temperature and pressure are already
specified. Thus it is sufficient to specify one concentration in either phase to
completely specify the state of the mixture. A single coordinate of any point on the
binodal curve gives both phase compositions if the tie lines are known. This exercise
does not determine the relative amounts of the phases present since these are not state
variables. Nor does specifying a single coordinate of the overall concentration suffice
since these, in general, do not lie on the binodal curve. Of course, it is possible to
calculate the phase compositions and the relative amounts from equilibrium relations,
but these must be supplemented in “flash calculations” by additional mass balance
relations to give the amounts of each phase.
Three-Phase Behavior
When three phases form, there are no degrees of freedom (NF = 0). The state of the
system is entirely determined. It follows from this that three-phase regions are
represented on ternary diagrams as smaller subtriangles embedded within the larger
ternary triangle (Fig. 4-11). Since no tie lines are in three-phase regions, the apexes
or invariant points of the subtriangle give the phase compositions of any overall
composition within that subtriangle. The graphical construction indicated in Fig. 4-11
gives the relative amounts of the three phases present (see Hougen et al., 1966, and
Exercise 4D).
Figure 4-11 Three-phase diagram example (from Lake, 1984)

110
A point on a nonapex side of the subtriangle may be regarded as being
simultaneously in the three-phase region or in a two-phase region; thus the
subtriangle must always be bounded on a nonapex side by a two-phase region for
which the side of the subtriangle is a tie line of the adjoining two-phase region. By
the same argument, the apexes of the subtriangle must adjoin, at least in some
nonzero region, a single-phase region. To be sure, the adjoining two-phase regions
can be quite small (see Fig. 9-6).
Thus points A and C in Fig. 4-11 are two-phase mixtures, point B is three
phase, and points D and E are single phase, though point D is saturated with respect
to phase 1. (For more detail on the geometric and thermodynamic restrictions of
ternary equilibria, see Francis, 1963.)
4-4 QUANTITATIVE REPRESENTATION OF TWO-PHASE

EQUILIBRIA
Several mathematical relations describe the qualitative representations in the previous

section. The most common are those based on (1) equilibrium flash vaporization
ratios, (2) equations of state, and (3) a variety of empirical relations. In this section,
we concentrate only on those two-phase equilibria aspects directly related to EOR.
Three-phase equilibria calculations are discussed elsewhere in the literature (Mehra
et al., 1980; Risnes and Dalen, 1982; Peng and Robinson, 1976) and in Chap. 9,
which covers three-phase equilibria for micellar systems.
Equilibrium Flash Vaporization Ratios
If we let xi and yi be the mole fractions of component i in a liquid and in contact with
a vapor phase, the equilibrium flash vaporization ratio for component i is defined by
yi
Ki = , i = 1, . . . , N C (4.4-1)
xi
This quantity is universally known as the K-value for component i.

At low pressures, the K-values are readily related to the mixture temperature
and pressure. The partial pressure of component i in a low-pressure gas phase is yiP
from Dalton’s law of additive pressures. The partial pressure of component i in the
vapor above an ideal liquid phase is xiPvi from Raoult’s law, where Pvi is the pure
component vapor pressure of component i (see Figs. 4-1 and 7-2). At equilibrium for
this special case, the partial pressures of component i calculated by either means must
be equal; hence
y P
K i = i = vi , i = 1, . . . , N C (4.4-2)
xi P
Equation (4.4-2) says at low pressures, a plot of the equilibrium K-value for a
particular component at a fixed temperature will be a straight line of slope –1 on a
111
log–log plot. Under these conditions, the K-value itself may be estimated from pure
component vapor pressure data.
At higher pressures, where the assumptions behind Dalton’s and Raoult’s
laws are inaccurate, the K-values are functions of overall composition. The additional
composition information, usually based on the liquid-phase composition, can be
incorporated into a convergence pressure, which is then correlated to the K-values.
Convergence pressure correlations are usually presented in graphical form (GPSA
Data Book, 1983) or as equations. The introduction of a composition variable directly
into the K-value functions adds considerable complexity to the flash procedure.
The flash calculation proceeds as follows: Let zi be the overall mole fraction
of component i in the mixture (analogous to ωi, the overall mass fraction in Chap. 2).
Then
zi = nL xi + nV yi , i = 1, . . . , N C (4.4-3)
where nL and nV are the relative molar amounts of the liquid and gas phases,
respectively. Since all quantities in Eq. (4.4-3) are relative, they are subject to the
following constraints
NC NC NC
∑ xi = ∑ yi = ∑ zi = nL + nV = 1
i =1 i =1 i =1
(4.4-4)
Eliminating nL from Eq. (4.4-3) with this equation, and substituting the definition
(Eq. 4.4-1) for yi, yields the following for the liquid-phase composition:
zi
xi = , i = 1, . . . , N C (4.4-5)
1 + ( K i − 1) nV
But these concentrations must also sum to 1.
NC
zi
∑1+ ( K =1 (4.4-6a)
i =1 i − 1) nV
Equation (4.4-6a) is a single polynomial expression for nV, with Ki and zi known, that
must be solved by trial and error. The equation itself is not unique since we could
have eliminated nV and xi from Eq. (4.4-3) to give the entirely equivalent result
NC
zi
∑
i =1 ⎛ 1 ⎞
=1 (4.4-6b)
1+ ⎜ − 1 ⎟ nL
⎝ Ki ⎠
The usual flash procedure is to calculate nV or nL by trial and error, and then use Eqs.
(4.4-1) and (4.4-5) to calculate the phase concentrations.
Alternatively, Eqs. (4.4-6a) and (4.4-6b) can be used to calculate quality lines
in a P-T diagram by specifying nV or nL and then performing trial-and-error solutions
112
for pressure at various fixed temperatures. Two special cases of the above procedure
follow directly. The bubble point curve for a mixture (nL = 1) is given implicitly from
Eq. (4.4-6b) as
NC
1 = ∑ zi K i (4.4-7a)
i =1
and the dew point curve (nV = 1) from Eq. (4.4-6a) as
NC
zi
1= ∑ (4.4-7b)
i =1 K i
These equations suggest the necessity of doing a flash calculation. Because

the K-values increase as temperature increases, a mixture of overall composition zi at
fixed temperature and pressure will be two phase only if
NC NC
zi
∑i =1
z K
i i > 1 and ∑
i =1 K i
>1 (4.4-8)
If the first inequality in Eq. (4.4-8) is violated, the mixture is a single-phase liquid; if
the second is violated, the mixture is a single-phase gas.
Equations of State
Though the K-value approach is easily the most common representation of two-phase
equilibria, it suffers from a lack of generality and may result in inaccuracies
particularly near the convergence pressure. In recent years, the trend has been toward
equation of state (EOS) representations since these are potentially able to work near
the critical point and yield internally consistent densities and molar volumes. (For
more details on EOS and its underlying thermodynamic principles, see Smith and van
Ness, 1975, and Denbigh, 1968.)
Pure components. An EOS is any mathematical relationship among the

three intensive properties molar volume, temperature, and pressure. Usually, the
relation is written in a pressure-explicit form P = f (VM , T ), and the most elementary
form is the ideal gas equation
RT
P= (4.4-9)
VM
This equation applies only to gases at low pressure. Equation (4.4-9) can be corrected
to apply to real gases by introducing a correction factor z, the compressibility factor
zRT
P= (4.4-10)
VM
The compressibility factor is itself a function of temperature and pressure that is

given in many sources (see McCain, 1973, for example). Since Eq. (4.4-10) is
113
actually a definition of the compressibility factor, the equation can also be applied to
fluids and liquids though the latter is rarely done. Given the relation between z and T
and P, Eq. (4.4-10) could predict volumetric behavior for all T and P.
Consider the pressure–molar-volume behavior of a pure component as shown
in Fig. 4-2. Figure 4-12 also shows this type of plot with two isotherms T1 and T2,
both below the critical temperature. Equation (4.4-9) is the equation of a hyperbola
on this plot that matches the experimental isotherm well at low pressure or high
molar volume. The ideal gas law fails badly in the liquid region, particularly for
pressure predictions, since it predicts a zero asymptote on the molar volume axis.
This is equivalent to saying the component molecules themselves have no intrinsic
volume even at the highest pressure, which is, of course, a basic hypothesis in the
derivation of the ideal gas law from statistical mechanics.
Figure 4-12 General features of cubic equations of state
To introduce a nonzero asymptote, we try an equation of the form
RT
P= (4.4-11)
(VM − b )
where b is now the asymptotic value of VM as pressure increases. Figure 4-12 shows
this equation can be made to match the liquid molar volumes reasonably well at high
114
pressures. The value of b, the intrinsic molecular volume, is usually so small that Eq.
(4.4-11) still provides a good estimate at low pressures.
But Eq. (4.4-11) still fails for temperature and pressure combinations that are
fairly close to the pure component vapor pressure curve. To predict the molar volume
up to and including the vapor pressure curve requires a function of the form
RT
P= − f (T ,VM ) (4.4-12)
(VM − b )
where the term f (T ,VM ) is specific to the particular EOS. Equation (4.4-12) is
frequently interpreted as a sum of forces, the first term being the force that will not
allow the molecules to be compressed to zero volume (repulsive force), and the
second being the force due to the intermolecular attraction among molecules.
A practical EOS must be accurate, internally consistent near the critical
point, and relatively simple. Moreover, since we are to use it to predict vapor–liquid
equilibria, it must predict both liquid and gas properties.
For pure components, there can exist two values of molar volume at a
particular temperature and pressure; hence Eq. (4.4-12) must have at least two real
roots at this point. Moreover, since P is a monotonically decreasing function of VM
regardless of the fluid-phase identity, f must be at least second order in VM so that the
entire function (Eq. 4.4-12) must be at least cubic in molar volume. Cubic EOS,
therefore, are the simplest form that satisfy the three criteria mentioned. Though there
have been more than 100 EOS proposed in the technical literature, many of which are
quite complicated and have more thermodynamic rigor, we discuss only cubic EOS
since these seem to be the most commonly used class of equations in EOR.
In the vicinity of the vapor pressure curve (pressures between P1 and P2 at
temperature T1 in Fig. 4-12), there are three real roots to the cubic EOS. The vapor
pressure Pv corresponding to T1 is the y-coordinate value that causes the shaded areas
above and below Pv to have equal areas (Abbott and van Ness, 1972). For pressures
above Pv, only the smallest root has physical significance and corresponds to VM of a
compressed liquid; at pressures below Pv, the largest root corresponds to VM of a
superheated vapor. At the vapor pressure, both the smallest and largest roots have
physical significance corresponding to the saturated liquid and vapor molar volumes,
respectively. The intermediate root has no physical significance.
As the critical point is approached, all three roots converge to the value of
VM at the critical point VMc . For temperatures above Tc, cubic equations have only
one real root, that of the molar volume of a fluid. For the critical isotherm itself, there
is also only one real root, and the critical constraints Eq. (4.1-3) are satisfied at the
critical pressure.
Within the two-phase region on the pressure–molar-volume plot, the
quadratic curves defined by (∂P / ∂VM )T = 0 for P < Pc are the spinodal curves. They
represent the maximum degree of supersaturation with respect to the particular phase
transition. Thus theoretically at least, we could lower the pressure on a single
compressed liquid phase at T1 to P2 without changing phase. The liquid between Pv
and P2 is supersaturated with respect to the vapor phase. A phase transition must
115
occur beyond this pressure since the partial derivative (∂P / ∂VM )T is constrained to
be negative on thermodynamic and physical grounds. Similarly, a vapor phase at
pressure P1 could be supercooled down to only temperature T2 without causing a
phase change, and the vapor at P1 and T2 is supersaturated with respect to the liquid
phase. These are metastable states that will change to stable states on perturbation.
The above discussion gives the properties of any general cubic EOS. The
particular form of such equations, of course, can take a wide variety of forms. Abbott
(1973) gives the general form
RT θ (VM − η )
P= − (4.4-13)
(VM − b ) (VM − b ) (VM 2 + δ VM + ε )
where the parameters θ, η, δ, and ε are given in Table 4-1 for nine specific equations
of state. Equation (4.4-13) is, perhaps, not the most general form of the cubic
equations available (Martin, 1979), but it does include most of the commonly
accepted equations used in predicting the phase behavior of EOR fluids.
Abbott’s original work (1973) contains complete references on each of the
equations in Table 4-1. Thus far, only two of these equations have seen extensive use
in predicting EOR phase behavior: the Soave modification (1972) of the Redlich-
Kwong equation (RKS) and the Peng-Robinson (1976) equation (PR). We discuss
these two equations here.
Except for the Clausius equation, all the equations in Table 4-1 are two-
parameter equations. The value of these parameters may be chosen to force the
equation to make internally consistent predictions in the vicinity of the critical point
for pure components. Thus the values of the parameters come from enforcing the
critical constraints (Eq. 4.1-3) and from evaluating the original equation at the critical
point. Since there are three equations, the procedure also specifies a specific value of
the critical molar volume VM or critical z-factor zc in addition to a and b.
It is somewhat easier, though entirely equivalent, to use the procedure of
Martin and Hou (1955) to determine the parameters a and b. Expressing the RKS
equation in the z-factor form will eliminate VM between Eq. (4.4-10) and the RKS
equation. By applying Descartes’ rule of roots to this equation, there is either one or
TABLE 4-1 CLASSIFICATION OF SOME CUBIC EQUATIONS OF STATE

(FROM ABBOTT, 1978)
Equation θ η δ ε
van der Waals (1873) a b 0 0

Berthelot (1900) a/T b 0 0
Clausius (1880) a/T b 2c c2
Redlich-Kwong (1949) a/T1/2 b b 0
Wilson (1964)* θw(T) b b 0
Peng-Robinson (1976) θPR(T) b 2b –b2
Lee-Erbar-Edmister (1973) θLEE(T) (T) b 0
*Similarly, Barner et al. (1966) and Soave (1972)

116
three positive and no negative real roots. The z-factor equation evaluated at the
critical point must have only one real root; hence
( z − zc ) = z 3 − 3zc z 2 + 3zc2 z − zc3 = 0
3
(4.4-14)
This equation is identically equal to the form in Table 4-2; hence equating
coefficients, we immediately have zc = 1/3 and
3 zc2 = A − B − B 2 (4.4-15a)
z = AB
3
c (4.4-15b)
Eliminating A from these equations gives the cubic form
27 B 3 + 27 B 2 + 9 B = 1 (4.4-16)
Moreover, using Descartes’ rule, it follows that this equation has only one
real positive root, which may be solved for directly to give B = (21/3–1)/3 = 0.08664.
Solving for A from Eq. (4.4-15b) gives A = (9(21/3–1))–l = 0.4247. Using the
definitions for A and B gives the forms in Table 4-2 for a and b.
Clearly, the above procedure is valid for any a and b that are a function of
temperature only. To match experimental vapor pressure data to subcritical
temperatures, the a given by this procedure is multiplied by a factor αi, a function of
temperature that reduces to unity at the critical temperature. The factor αi is also
component specific through its dependence on the acentric factor ωi. Acentric factors
roughly express the deviation of the shape of a molecule from a sphere and are
available in extensive tabulations (Reid et al., 1977).
Mixtures. The true test and practical utility of any EOS is in its prediction
of mixture properties. For mixtures, many of the arguments advanced above in
conjunction with Fig. 4-12 do not apply. In particular, the critical constraints are no
longer satisfied at the critical point since this point is no longer at the top of the two-
phase envelope.
To account for mixture behavior, the pure component parameters ai and bi
come from various mixing rules, as shown in Table 4-2. The inclusion of the
component index in Table 4-2 means the parameters used in the definitions of these
quantities––Tci, Pci, and ωi––are those for the pure component i.
The most general form of the mixing rules incorporates another parameter,
the binary interaction coefficient δij into the RKS and PR equations, which accounts
for molecular interactions between two unlike molecules. By definition δij is zero
when i and j represent the same component, small when i and j represent components
that do not differ greatly (for example, if i and j were both alkanes), and large when i
and j represent components that have substantially different properties. Ideally, the δij
are both temperature and pressure independent (Zudkevitch and Joffe, 1970),
depending only on the identities of components i and j. Though the interaction
coefficients are considerably less available than acentric factors, literature tabulations
are becoming more common (Yarborough, 1978; Whitson, 1982; Prausnitz et al.,
1980).
117
TABLE 4-2 COMPARISON OF THE RKS AND PR EQUATIONS OF STATE (FROM NGHIEM AND AZIZ, 1979)
RKS PR
RT a RT a
P= − P= −
V − b VM (VM + b )
Equation
VM − b VM2 + 2bVM − b 2
z-factor z 3 − z 2 + z ( A − B − B 2 ) − AB = 0 z 3 − (1 − B ) z 2 + ( A − 3B 2 − 2 B ) z − ( AB − B 3 − B 2 ) = 0
aP bP
A= B=
( RT )
2
RT
0.42747 R 2Tci2.5α i 0.45724 R 2Tci2.5α i
Pure component i ai = , zc = 0.333 ai = , zc = 0.307
Pci Pci
let a = ai 0.08664 RTci 0.07780 RTci
bi = bi =
b = bi Pci Pci
mi = 0.480 + 1.57ω i − 0.176ωi2 mi = 0.37464 + 1.54226ωi − 0.26992ωi2
2
⎡ ⎡ ⎛ T ⎞1/ 2 ⎤ ⎤
α i = ⎢1 + mi ⎢1 − ⎜ ⎟ ⎥ ⎥
⎣⎢ ⎣⎢ ⎝ Tci ⎠ ⎦⎥ ⎦⎥
Mixture am = ∑∑ xi x j aij , bm = ∑ xi bi
ij i
let a = am
aij = (1 − δ ij )( ai a j )
1/ 2
b = bm
Pure component f A f A
ln = z − 1 − ln ( z − B ) − . ln = z − 1 − ln ( z − B ) − .
fugacity P B P 2 2B
use zL for fL ⎡z + B⎤ ⎡ z + 2.414 B ⎤
ln ⎢ ⎥ ln ⎢ ⎥
zV for fV ⎣ z ⎦ ⎣ z − 0.414 B ⎦
Fugacity of ⎛ f ⎞ b ⎛ f ⎞ b
ln ⎜ i ⎟ = i ( z − 1) − ln ( z − B ) − ln ⎜ i ⎟ = i ( z − 1) − ln ( z − B ) −
component i ⎝ Pxi ⎠ b ⎝ xi P ⎠ b
use zL and xi for f i L ⎛ 2∑ x a ⎞ ⎛ 2∑ x a ⎞
A ⎜ j j ij bi ⎟ ⎛ z + B ⎞ A ⎜ j j ij bi ⎟ ⎛ z + 2.414 B ⎞
V V ⎜ − ⎟ ln ⎜ ⎟ ⎜ − ⎟ ln ⎜ ⎟
z and yi for fi B⎝ a b⎠ ⎝ z ⎠ 2 2B ⎝ a b ⎠ ⎝ z − 0.414 B ⎠
118
Flash calculations. To calculate vapor–liquid equilibria for mixtures from
the RKS equation, an expression for the fugacity of a component i in a mixture is
needed. This is most conveniently done by introducing a fugacity coefficient of
component i defined as
f
φi = i (4.4-17)
xi P
In Eq. (4.4-17), and all subsequent equations in this section, the composition variable
may be either the liquid-phase mole fraction xi, if calculating the fugacity coefficient
of component i in the liquid phase, or the vapor-phase mole fraction yi, if calculating
the fugacity coefficient in the vapor phase. Following the arguments presented in
standard texts (Smith and van Ness, 1975), the fugacity coefficient is, for a mixture,
∞ ⎡⎛ ∂ ( nz ) ⎞ ⎤ dV
ln φi = ∫ ⎢⎜ ⎟ − 1⎥ M
− ln z (4.4-18a)
VM
⎢⎣⎝ ∂ ni ⎠T ,V , n V
⎥⎦ M
and, for a pure component,
∞ dV
ln φ = ∫ ( z − 1) M + z − 1 − ln z (4.4-18b)
VM VM
Equation (4.4-18b) is a special case of Eq. (4.4-18a) as one of the xi becomes unity.
The partial derivative in the integral of Eq. (4.4-18a) is taken at constant temperature
and total volume V, where n is the total number of moles in the mixture, and ni is the
total number of moles of species i in the phase. Clearly,
NC
V = nVM , n = ∑ ni , and xi = ni / n.
i =1
The fugacity coefficient definition (Eq. 4.4-18a) also can be written in a

variety of equivalent forms (Smith and van Ness, 1975; Coats, 1980). To evaluate the
integral in Eq. (4.4-18a), it is convenient to express z in an explicit form
V a
z= M − (4.4-19)
VM − b RT (VM + b )
After multiplying Eq. (4.4-19) by n and introducing the mixing rules for a
and b from Table 4-2, the resulting expression may be differentiated with respect to
ni. After some algebra, this gives
⎛ ∂ ( nz ) ⎞ VM VM bi 1 ⎜⎛ ∑ j x j aij abi ⎞
⎟
⎜ ⎟ = − − − (4.4-20)
⎝ ∂ni ⎠T ,V ,n VM − b (VM − bm ) RT ⎜ VM + bm (VM + bm )2 ⎟
2
⎝ ⎠
Equation (4.4-20) is explicit in VM which, when substituted into Eq. (4.4-18a) and
integrated, leads to the closed-form expression given in Table 4-2. Similar procedures
may be used on the PR equation (see Exercise 4F).
119
The actual calculation of vapor–liquid equilibria follows from two general
procedures based on the EOS approach. From Eqs. (4.4-1) and (4.4-17), the
equilibrium K-values become
φiL
Ki = , i = 1, . . . , N C (4.4-21)
φiV
since the component fugacities are equal at equilibrium. Thus based on an initial
estimate of the Ki, a flash calculation, as described above, will obtain the vapor and
liquid compositions from Eq. (4.4-5), the K-value definition, and the Ki calculated
from Eq. (4.4-21). If the beginning and initial K-value estimates agree, the calculated
compositions are the correct values; if they do not agree, new values of the Ki must
be estimated and the entire procedure repeated until the K-values do not change.
Since the flash calculation is itself a trial-and-error procedure, this procedure is
somewhat analogous to the convergence pressure approach we already described.
The second approach to calculating vapor–liquid equilibria from EOS is to
directly use the equilibrium constraints. Thus the equations
f i L = f iV , i = 1, . . . , N C (4.4-22a)
may be regarded as a set of NC independent simultaneous nonlinear equations in

either xi or yi (but not both since xi and yi are related through the K-values) that may
be linearized, solved as a system of simultaneous linear equations, and iterated until
the phase compositions do not change. Either way the calculation is fairly
convoluted, so it is not surprising that many variations of the procedure exist (Fussell
and Fussell, 1979; Mehra et al., 1980).
Equation (4.4-22a) is easily generalized to the condition for equilibrium
among any number of phases NP
fij = f ik , i = 1, . . . , N C ; j , k = 1, . . . , N P (4.4-22b)
Empirical Representations
There are three common empirical representations of phase behavior. All are used
primarily for liquid–liquid equilibria.
Hand’s rule. Hand (1939) gave a fairly simple representation of two-phase

equilibria that has proved useful for some EOR systems (Pope and Nelson, 1978;
Young and Stephenson, 1982). The procedure is based on the empirical observation
that certain ratios of equilibrium phase concentrations are straight lines on log–log or
Hand plots.
In this section, the concentration variable Cij is the volume fraction of
component i(i = 1, 2, or 3) in phase j(j = 1 or 2). Using volume fractions has become
conventional in the Hand representation since these are convenient in liquid–liquid
equilibria.
120
Figure 4-13 shows the one- and two-phase regions on the ternary diagram
and its correspondence to the Hand plot. The line segments AP and PB represent the
binodal curve portions for phase 1 and 2, respectively, and curve CP represents the
distribution curve of the indicated components between the two phases. The ratios on
the distribution curve are analogous to, but entirely different from, the definitions of
the K-values given above. The equilibria relations based on the Hand plot are
BH
C3 j ⎛ C3 j ⎞
= AH ⎜ ⎟⎟ , j = 1, 2 (4.4-23)
C2 j ⎜C j
⎝ 1 ⎠
FH
C32 ⎛C ⎞
= EH ⎜ 31 ⎟ (4.4-24)
C22 ⎝ C11 ⎠
where AH, BH, EH, FH are empirical parameters. Equation (4.4-23) represents the
binodal curve, and Eq. (4.4-24) represents the distribution curve. In this form, these
equations require the binodal curve to enter the corresponding apex of the ternary
diagram. A simple modification overcomes this restriction (see Exercise 4G).
Figure 4-13 Correspondence between ternary diagram and Hand plot
Within the two-phase region of the binodal curve, there are six unknowns,
the Cij phase concentrations and five equations, three from Eqs. (4.4-23) and (4.4-24)
and two consistency constraints
3
∑C
i =1
ij = 1, j = 1, 2 (4.4-25)
Thus there is NF = 1 degree of freedom as required by the phase rule since

temperature and pressure are fixed for ternary equilibria.
121
A flash calculation using the Hand procedure solves for the relative amounts
of the two phases. This introduces two additional variables, S1 and S2, into the
calculation, but there are now three additional equations, the mass balance Eq. (4.3-1)
with the overall concentrations Ci known and S1 + S2 = 1. As in all the phase
equilibria flash calculations, the procedure is trial and error though for certain special
cases, phase concentrations follow from direct calculation. The iterative procedure is
to first pick a phase concentration (say, C32), calculate all the other phase
concentrations from Eqs. (4.4-23) through (4.4-25), and then substitute these into the
tie line Eq. (4.3-2). If this equation is satisfied, convergence has been attained; if it is
not satisfied, a new C32 must be picked and the procedure repeated until either C32
does not change or Eq. (4.3-2) is satisfied.
Two other empirical representations of the distribution of components
between phases are of interest: the conjugate curve and the tie line extension curve.
Both require separate representations of the binodal curve, as in Eq. (4.4-23).
Conjugate curve. The conjugate curve is a curve in ternary space whose

coordinates define the ends of the tie lines. Thus for phases 1 and 2, the conjugate
curve would be of the form
C11 = f ( C22 ) (4.4-26)
Figure 4-14 shows the projections of the coordinates of this curve onto the binodal
curve. The Hand distribution curve is of the form shown in Eq. (4.4-26). The
conjugate curve must pass through the plait point.
Tie line extension curve. The tie line extension curve is anq~her curve
C = f (C20 ) in ternary space that passes through the plait point, at which point it is
0
3
tangent to the binodal curve (Fig. 4-15a). The two-phase tie lines are extensions of
tangents from this curve through the binodal curve. Thus equations of the tie lines are
given by straight lines having the equation
C3 j − f ( C20 ) = f ′ C 0 ( C2 j − C20 ) , j = 1 or 2 (4.4-27a)

2
where f ′ C 0 is the slope of the tie line extension curve evaluated at the coordinate
2
C20 . The tie lines follow from Eq. (4.4-27a), the equation of the extension curve, and
the equation for the binodal curve.
A useful special case of the tie line extension curve occurs when all tie lines
extend to a common point, as in Fig. 4-15(b). We need specify only the coordinates
of this common point to define the equation for the tie lines
C3 j − C30 = η ( C2 j − C20 ) , j = 1 or 2 (4.4-27b)
where η is the slope of the tie line. Note that if C30 > 0, the selectivity of the
components for the two phases can reverse near the base of the ternary. The
representation is extremely simple because it requires only two values: any two of the
122
Figure 4-14 Schematic representation

of a conjugate curve
coordinates Ci0 or, alternatively, any of the plait point coordinates and one of the Ci0
since the tie lines must be tangent to the binodal curve there.
This representation is far less general than either Eq. (4.4-24), (4.4-26), or
(4.4-27). But experimental accuracy is often not enough to warrant more complicated
equations. Moreover, the form (Eq. 4.4-27b) is extremely convenient for calculating
the flow behavior of two-phase mixtures; we use it extensively in Chaps. 7 and 9.
4-5 CONCLUDING REMARKS
Multiple representations of phase behavior are clear evidence that no single method
is sufficient. In most cases we find ourselves compromising between accuracy and
mathematical ease in the resulting calculation. Our goal here is the exposition of the
123
Figure 4-15 Tie line extension

representation of phase behavior
underlying principles of EOR phenomena; hence we emphasize phase behavior

representations which lend themselves to visual or graphical analysis in later
calculations––as long as the representations themselves are qualitatively correct. The
important points to grasp in this chapter, then, are graphical representations in Sec.
4-3, particularly as related to the ternary diagram; the physical meaning of tie lines
and binodal curves; and the component distribution expressed by Eqs. (4.4-23),
(4.4-24), and (4.4-27b).
EXERCISES
4A. Pure Component Phase Behavior. Sketch the following for a pure component:
(a) Lines of constant pressure on a temperature–molar-volume plot
(b) Lines of constant temperature on a density–pressure plot
(c) Lines of constant molar volume on a temperature–pressure plot
124
4B. Paths on a Pressure–Volume Plot. Indicate the paths AA′, BB′, and DD′, shown on
the pressure–specific-volume plot in Fig. 4B, on the corresponding pressure–
temperature plot.
Figure 4B Pressure–specific-volume plot for Exercise 4B
4C. Migration of P-T Envelope. Figure 4C shows the hypothetical change in the pressure–
temperature envelope of a crude oil as it is diluted with a more volatile component
(CO2). The quality lines within each envelope are in volume percent. For this data,
sketch the pressure–composition diagram at 340 K and 359 K (152°F and 180°F).
These temperatures are the critical temperatures for the 40% and 20% CO2 mixtures.
Include as many quality lines as possible.
4D. Lever Rule Application. Consider the three-component system represented in Fig. 4-11.
(a) Estimate the relative amounts of each phase present at overall compositions A, C,
D, and E.
(b) Derive the expressions (indicated on the figure) for the relative amounts of each
phase present at the three-phase overall composition.
(c) Estimate the relative amounts of each phase present at B.
4E. Parameters for RKS and RP Equations of State
(a) Derive the parameters a and b for the RKS equation using the critical constraints
and the original equation given in Table 4-1.
(b) Derive the parameters a and b for the PR equation using the procedure of Martin
and Hou (1955). Compare your results to Table 4-2.
125
Figure 4C Change in crude oil pressure–temperature diagram with dilution by CO2
4F. Fugacity Coefficient from an Equation of State

(a) Derive the expressions for the fugacity coefficient starting with Eq. (4.4-18a) for
the PR equation.
(b) Show for both the PR and RKS equations that the fugacity coefficient for a
mixture approaches that of a pure component as one of the xi approaches 1.
4G. Partially Soluble Binaries (Welch, 1982). For cases when the partially soluble binaries
on a ternary plot have some mutually soluble region, the Hand representation may be
altered as
BH
C3′ j ⎛ C3′ j ⎞
= AH ⎜ ⎟⎟ , j = 1, 2 (4G-1)
C2′ j ⎜ C1′j
⎝ ⎠
and
FH
C32′ ⎛ C′ ⎞
= EH ⎜ 31 ⎟ (4G-2)
′
C22 ⎝ C11′ ⎠
where the C1j are normalized concentrations
C1 j − C1L
C1′j = (4G-3)
C1U − C1L
126
C2 j − (1 − C1U )
C2′ j = (4G-4)
C1U − C1L
C3 j
C3′ j = (4G-5)
C1U − C1L
C1U and C1L are the upper and lower solubility limits of the 1-2 binary. Take BH = –1
and FH = 1 in the following.
(a) Derive an expression for AH in terms of the true maximum height of the binodal
curve C1 – C2. Show that the binodal curve takes value C3max when C1′ = C2′
(symmetrical in normalized concentrations).
(b) Express EH as a function of AH and the component 1 coordinate of the plait point
(C1p). The AH and EH in parts (a) and (b) will also be a function of C1U and C1L.
(c) Plot the binodal curve and the two representative tie lines for C1U = 0.9, C1L = 0.2,
C3max = 0.5, and C1p = 0.3.
4H. Using the Hand Representations. The following data were collected from a three-
component system at fixed temperature and pressure.
Phase 1 Phase 2
Component 1 Component 2 Component 1 Component 2
0.45 0.31 0.015 0.91

0.34 0.40 0.020 0.89
0.25 0.48 0.030 0.85
0.15 0.60 0.040 0.82
The concentrations are in volume fractions.

(a) On a ternary diagram, plot as many tie lines as possible, and sketch in the binodal
curve.
(b) Make a Hand plot from the data, and determine the parameters AH, BH, EH, and FH.
(c) Estimate the coordinates of the plait point from the plot in part (b).
41. Application of Conjugate Curve. Consider the ternary diagram in Fig. 41 for a three-
component system. The binodal curve is the solid line, and the conjugate curve the
dotted line.
(a) Sketch in three representative tie lines.
(b) For the overall composition, marked as A, give the equilibrium phase compositions
and the relative amounts of both phases.
(c) Plot the two-phase equilibria on a Hand plot.
(d) If the Hand equations are appropriate, determine the parameters AH, BH, EH, and
FH.
127
Figure 4I Diagram for Exercise 4I

128
Displacement Efficiency
The definitions for recovery, displacement, and sweep efficiencies in Eq. (2.5-5)
apply to an arbitrary chemical component, but they are almost exclusively applied to
oil and gas displacement. Since displacement efficiency and sweep efficiency are
multiplied by each other, they are equally important to the magnitude of the recovery
efficiency and, hence, the oil recovery. In Chap. 6, we discuss volumetric sweep
efficiency; in this chapter, we present fundamental concepts in displacement
efficiency.
For the most part, we restrict our discussion to oil displacement efficiency
based on solutions to the fractional flow Eq. (2.4-3). We apply these equations to
displacements in one-dimensional, homogeneous, isotropic permeable media. Thus,
the results apply most realistically to displacements in laboratory floods, the
traditional means of experimentally determining displacement efficiency. These
results do not, of course, estimate recovery efficiency for three-dimensional,
nonlinear flows without correcting for volumetric sweep efficiency and without
correcting the displacement efficiency to account for differences in scale.
5-1 DEFINITIONS
If we assume constant oil density, the definition of displacement efficiency for oil
becomes
Amount of oil displaced
ED = (5.1-1)
Amount of oil contacted by displacing agent
129
ED is bounded between 0 and 1. The rate at which ED approaches 1 is strongly
affected by the initial conditions, the displacing agent, and the amount of displacing
agent. Fluid, rock, and fluid–rock properties also affect ED. If the displacement is
such that the displacing agent will contact all the oil initially present in the medium,
the volumetric sweep efficiency will be unity, and ED becomes the recovery
efficiency ER.
From Eq. (2.5-4) then,
S
ED = 1 − 2 (5.1-2)
S2 I
for an incompressible, single-component oil phase flowing in an incompressible

permeable medium. Equation (5.1-2) says ED is proportional to the average oil
saturation in the medium. For cases where the oil may occur in more than one phase,
or where more than oil can exist in the hydrocarbon phase, we must use the general
definition (Eq. 2.5-5b).
5-2 IMMISCIBLE DISPLACEMENT
Virtually all of our understanding about EOR displacements begins with an

understanding of the displacement of one fluid by an immiscible second fluid. The
specific case of water displacing oil was first solved by Buckley and Leverett (1942)
and later broadened by Welge (1952). In this section, we develop the
Buckley-Leverett theory in a manner much like the original paper and several
subsequent references (Collins, 1976; Craig, 1971; Dake, 1978).
For the isothermal flow of oil and water in two immiscible, incompressible
phases in a one-dimensional permeable medium, the mass conservation equations of
Table 2-2 reduce to
∂S ∂f
φ 1 +u 1 =0 (5.2-1)
∂t ∂x
for flow in the positive x direction, as we discussed in Chap. 2. In this equation, f1 is

the fractional flow of water,
u1 λr1 ⎛ k λ2 Δρ g sin α ⎞
f1 = = ⎜1 − ⎟ (5.2-2)
u λr1 + λr 2 ⎝ u ⎠
in the absence of capillary pressure. In Eq. (5.2-2), α is the dip angle defined to be
positive when measured in the counterclockwise direction from the horizontal, and
Δρ = ρ1 – ρ2 is the density difference between the water and oil phases.
The choice of S1 as the dependent variable in Eq. (5.2-1) is largely a matter
of convention; we could easily have chosen S2 since S2 + S1 = 1, and f2 + f1 = 1. An
important point is that in the absence of capillary pressure, f1 is uniquely determined
as a function of S1 only through the relative permeability relations λr1 = kr1/μ1 and
130
λr 2 = kr 2 / μ 2 discussed in Sec. 3-3. In fact, since the shape of the f1-S1 curve proves
to be the main factor in determining the character of the displacement, we digress
briefly to discuss how flow conditions affect this curve.
Fractional Flow Curves
If we introduce the exponential form of the oil–water relative permeability curves

(Eq. 3.3-4) into Eq. (5.2-2), we obtain
1 − N g0 (1 − S ) n2 sin α
f1 = (5.2-3a)
(1 − S ) n2
1+
M 0 S n1
where
S1 − S1r
S= = Reduced water saturation (5.2-3b)
1 − S2 r − S1r
and
k0 μ
M 0 = r1 02 = Endpoint water–oil mobility ratio (5.2-3c)
μ1kr 2
kkr02 Δρ g
N g0 = = Gravity number (5.2-3d)
μ 2u
N g0 is the ratio of gravity to viscous pressure gradients based on the endpoint oil
relative permeability. In the form of Eq. (5.2-3a), f1 depends parametrically on M0,
N g0 , α, and the shape of the relative permeability curves (n1 and n2). The f1-S1 curve
is sensitive to all these factors, but usually M0 and N g0 are most important. Figure 5-1
shows f1-S1 curves for various values of M0 and N g0 sin α with the other parameters
fixed (S1r = 0.2, S2r = 0.2, n1 = n2 = 2). The S-shaped curves have an inflection point
that varies with M0 and N g0 sin α. The curvature of all curves generally becomes
more negative as M0 increases or N g0 sin α decreases. The curves where f1 is less than
0 or greater than 1 are physically correct. This circumstance indicates a flow where
gravity forces are so strong that flow in the negative x direction occurs (water flows
in the negative x direction for f1 < 0). In Sec. 3-3, we showed that shifting the
wettability of the permeable medium from water wet to oil wet caused kr01 to increase
and kr02 to decrease. Thus for constant phase viscosities, making the medium more
oil wet is qualitatively equivalent to increasing M0. But for fixed relative permeability
curves, the effect of increasing μ1 or decreasing μ2 is to decrease M0.
Buckley-Leverett Solution
Returning now to Eq. (5.2-1), to calculate ED, we seek solutions S1(x, t) subject to the
initial and boundary conditions
S1 ( x,0) = S1I , x ≥ 0 (5.2-4a)
S1 (0, t ) = S1J , t≥0 (5.2-4b)
131
Figure 5-1 Fractional flow curves for m = n = 2 and S1r = S2r = 0.2
In core floods, a specified fractional flow is usually imposed on the inflow (x = 0) so

that we may replace Eq. (5.2-4b) with
f1 (0, t ) = f1 ( S1 (0, t )) = f1J = f1 ( S1J ), t≥0 (5.2-4c)
This equation shows that f1 is a function of x and t only through its dependence on S1.
The definition used in a given instance depends on the particular application. The
conditions (Eqs. 5.2-4) also have a convenient geometrical interpretation in xt space
because at the point t = x = 0, all values of S1 between S1I and S1J exist. The Buckley-
Leverett problem is usually posed with S1I and S1J taken to be S1r and 1 – S2r,
respectively.
For greater generality, we render Eqs. (5.2-1) and (5.2-4) into the following
dimensionless forms:
⎛ ∂S1 ⎞ ⎛ df1 ⎞⎛ ∂S1 ⎞
⎜ ⎟+⎜ ⎟⎜ ⎟=0 (5.2-5a)
⎝ ∂t D ⎠ ⎝ dS1 ⎠⎝ ∂xD ⎠
S1 ( xD ,0) = S1I , xD ≥ 0 (5.2-5b)
S1 (0, t D ) = S1J , tD ≥ 0 (5.2-5c)
where the dimensionless variables xD and tD are
x
xD = = Dimensionless position (5.2-6a)
L
132
tudt
tD = ∫ = Dimensionless time (5.2-6b)
0 φL
L is the total macroscopic permeable medium dimension in the x direction. In these

equations, u may be a function of time but not of position because of the assumption
of incompressibility. Moreover, df1/dS1 is a total derivative since f1 is a function of S1
only. Introducing dimensionless variables reduces the number of parameters in the
problem from four (φ, u, S1I, and S1J) in Eqs. (5.2-1) and (5.2-4) to two (S1I and S1J).
We could further reduce the number by redefining the dependent variable S1 (see
Exercise 5A).
The dimensionless time tD can also be expressed as
t Audt t qdt
tD = ∫ =∫ (5.2-7)
0 φ AL 0 V
p
where A is the cross-sectional area of the one-dimensional medium in the direction

perpendicular to the x axis, q is the volumetric flow rate, and Vp is the pore volume.
tD is the total volume of fluid injected up to time t divided by the total pore volume of
the medium. In principle, Vp is well defined even for a highly irregular geometry so
that tD is a scaling variable in virtually any application. In fact, tD is the fundamental
variable used to scale from the laboratory to the field. It has been used with a wide
variety of definitions for the reference volume Vp (see Table 5-1). Numerical values
of tD are frequently given as “fraction of a pore volume,” or simply “pore volume”;
thus it is easy to confuse with Vp, the actual pore volume, which has units of L3 (tD, of
course, has no units).
We seek solution to Eqs. (5.2-5) in the form S1(xD, tD). S1 may be written as a
total differential
⎛ ∂S ⎞ ⎛ ∂S ⎞
dS1 = ⎜ 1 ⎟ dxD + ⎜ 1 ⎟ dt D (5.2-8)
⎝ ∂xD ⎠t D
⎝ ∂t D ⎠ x D
TABLE 5-1 TABULATION OF VARIOUS DEFINITIONS

FOR DIMENSIONLESS TIME
Reference volume Usage
Area × length × porosity Core floods

Area × thickness × porosity General
(Aφh = Vp = total pore volume)
Vp × volumetric sweep efficiency
(Vp × Ev = floodable pore volume = VPF) Micellar polymer floods
VPF × ΔS2 = movable pore volume Waterfloods
VPF × S2I = hydrocarbon pore volume (HCPV) Miscible floods
Note: t D =
Total volume of fluid injected
Reference volume
} consistent units
133
from which follows that the velocity vS1 of a point with constant saturation S1 in xDtD
space is
⎛ dxD ⎞ (∂S1 ∂t D ) xD
⎜ ⎟ =− ≡ vS1 (5.2-9)
⎝ dt D ⎠ S (∂S1 ∂xD )tD
1
vS1 is the “specific” velocity of the saturation S1 because it has been normalized by
the bulk fluid interstitial velocity u/φ. It is dimensionless. You can see this by
converting Eq. (5.2-9) back to dimensional quantities using the definitions (Eqs.
5.2-6).
Eliminating either of the derivatives in Eq. (5.2-9) by Eq. (5.2-5a) gives
df
vS1 = 1 = f1′ (5.2-10)
dS1
This equation says the specific velocity of a constant saturation S1 is equal to the
derivative of the fractional flow curve at that saturation. In dimensional form, Eq.
(5.2-10) is the Buckley-Leverett equation. Since all saturations between S1I and S1J
are initially at the origin in xD-tD space, and vS1 is defined with S1 constant, the
position of any saturation S1I ≤ S1 ≤ S1J at a given tD is
df
xD |S1 = 1 t D = f1′( S1 )t D (5.2-11)
dS1 S
1
where we include evaluation symbols to help clarify the subsequent development.

Equation (5.2-11) is the solution to the one-dimensional water-displacing-oil
problem; by selecting several S1’s between S1I and S1J, we can construct S1(xD, tD).
Figure 5-2(a) shows the procedure for one of the fractional flow curves of Fig. 5-1.
Except for relatively simple cases (see Exercise 5E), the relation (Eq. 5.2-11)
generally cannot be solved explicitly for S1(xD, tD).
Shock Formation
Figure 5-2(a) also shows a disconcerting tendency for an S-shaped f1-S1 curve to
generate solutions that have three values of S1 at the same xD and tD. In Fig. 5-2(b),
this occurs for 0.64 < xD < 0.94. Of course, such triple values are nonphysical though
they are entirely valid mathematically. The triple values are the result of the
saturation velocity vS1 increasing over some saturation region ( S1I < S1 < S1′ in Fig.
5-2) as S1 changes from its initial (downstream) value to the final (upstream) value.
We eliminate the triple value region by invoking the formation of shocks,
discontinuous changes in a physical quantity such as pressure (as in the case of sonic
booms), concentration, or in this case, saturation. Shocks are characteristic features of
hyperbolic equations, a class of which are the dissipation-free conservation
equations. Strictly speaking, shocks are not present in nature since some dissipation
(dispersion, diffusion, capillary pressure, compressibility, and thermal conductivity)
is always present, which militates against their formation. When such effects are
present, the shocks are smeared or spread out around the shock front position, but the
position of the shock is unaltered. Despite this restriction, shocks play a central role
134
Figure 5-2 Buckley-Leverett

construction of S1(xD, tD)
in fractional flow theory, where dissipative effects are neglected, and describe many
actual flows to a good approximation.
To calculate the velocity and magnitude of the shock, we recast the
differential equations of this chapter into difference equations. This we do generally
in Sec. 5-4; here we restrict ourselves to the water-displacing-oil problem already
begun. Paradoxically, we find that calculations are considerably easier when shocks
form. Figure 5-3(a) shows a water saturation shock moving from left to right. The
water saturation ahead of the shock is S1− (downstream direction), and that behind
135
Figure 5-3 Water saturation profiles

with shocks
the shock is S1+ (upstream direction). The quantity ΔS1 = S1+ – S1− is the saturation
jump across the shock. A cumulative water balance on a control volume that contains
the shock in the time interval Δt is
( present at t + Δt ) ( present at t ) ( in during Δt ) ( out during Δt )

Volume water − Volume water = Volume water − Volume water
[(v(t + Δt ) − x1 ) S1+ + ( x2 − v(t + Δt )) S1− ] Aφ

t +Δt
−[(vt − x1 ) S1+ + ( x2 − vt ) S1− ] Aφ = [ f1 ( S1+ ) − f1 ( S1− )]∫ qdt
t
After some cancellation, we obtain a specific shock velocity
f1 ( S1+ ) − f1 ( S1− ) Δf1

vΔS1 = ≡ (5.2-12)
S1+ − S1− ΔS1
136
To incorporate shock formation into the water-displacing-oil problem,
consider a saturation profile containing a triple value over some region and
containing a single value elsewhere (Fig. 5-3b). In general, some saturation S1* will
mark the end of the continuous water saturation region and the beginning of a shock.
This saturation must simultaneously satisfy Eqs. (5.2-10) and (5.2-12); Eq. (5.2-12)
gives velocities of S1 greater than S1* , and Eq. (5.2-12) gives velocities of S1 less than
S1* . Equating Eqs. (5.2-10) and (5.2-12) yields the following equation for S1* :
f (S * ) − f (S )
f1′|S * = 1 1 * 1 1I (5.2-13)
1
S1 − S1I
where we have taken S1− = S1I in Eq. (5.2-12). Equation (5.2-13) lends itself to a
graphical solution since
f1 − f1 ( S1I ) = m( S1 − S1I ) (5.2-14)
is the equation of a straight line of slope m passing through the point (f1I, S1I) on the
fractional flow plot. If m = f1′|S1 , then m is the slope of the fractional flow plot at S1* .
Comparing Eq. (5.2-14) to Eq. (5.2-13), S1* is at the tangent to the fractional flow
curve of a straight line passing through the point (f1I, S1I). Figure 5-4 schematically
illustrates this construction. The slope of this straight line is the specific shock
Figure 5-4 Schematic illustration of shock construction

137
velocity. The shock itself is a discontinuous change in saturation from S1I to S1* at
xD = vΔS1 t D as Fig. 5-3(b) illustrates. The saturation S1* is not the same as S1− (Fig.
5-2), the saturation having the largest vS1 . S1* is the saturation whose position
requires the net area between the mathematical solution and the physical solution
(shaded region in Fig. 5-3b) to be zero. This requires the shock to preserve the
material balance. With this construction, all saturation velocities are monotonically
(though not continuously) decreasing in the upstream direction. Figure 5-3(b)
illustrates the results of the entire construction. The resulting saturation profile is
sometimes called the leaky piston profile.
Wave Classification
Before further developing this theory and its applications to EOR, we define a few
more terms used in subsequent discussions. These definitions are important to the
interpretation of xD-tD plots that graphically present the solution S1(xD, tD).
We have been discussing how to calculate water saturation as a function of
position and time for water–oil displacements. A plot of saturation, or concentration,
versus time at fixed position is a saturation history. If the fixed position in such a plot
is at the outflow end of the permeable medium, it is an effluent history. Plots of
saturation versus position at fixed time are saturation profiles. Figure 5-2(b) is a
water saturation profile. Changes in saturation with time and position are saturation
waves. Thus the previous development estimates the rate of propagation of waves
through a permeable medium.
An important and unifying aspect of our understanding of EOR
displacements is the study and characterization of the number and types of waves
they form. Depending on their spreading characters, waves may be classified into
four categories.
1. A wave that becomes more diffuse on propagation is a nonsharpening,

rarified, or spreading wave. When these waves occur, the rate of spreading is
usually much larger than that caused by dissipation.
2. A wave that becomes less diffuse on propagation is a sharpening wave. In
the absence of dissipation, these waves will become shocks even if the initial
saturation profile is diffuse. When dissipation is present, these waves will
asymptotically approach a constant pattern condition (see Sec. 5-3).
3. A wave that has both spreading and sharpening character is mixed. The
Buckley-Leverett water saturation wave of Fig. 5-2(b) is mixed, being a
sharpening wave for S1I < S1 < S1* and a spreading wave for S1* < S1 < S1J.
4. A wave that neither spreads nor sharpens on propagation is indifferent. In the
absence of dissipation, indifferent waves appear as shocks.
This behavior may be summarized by defining a dimensionless mixing or

transition zone ΔxD. This is the fraction of the total system length that lies between
arbitrary saturation limits at a given time. We take the saturation limits to be 0.1 and
138
0.9 of the span between the initial and injected saturations
ΔxD (t D ) = xD |S0.1 − xD |S0.9 (5.2-15a)
where
S0.1 = 0.1( S1J − S1I ) + S1I (5.2-15b)
S0.9 = 0.9( S1J − S1I ) + S1I (5.2-15c)
The exact value of the limits is unimportant to the behavior of the mixing zone. The
wave classification, which may be restated as ΔxD, increases with time for spreading
waves, decreases for sharpening waves, and either increases or decreases for mixed
waves depending on whether the shock portion of the wave exceeds the saturations
used to define ΔxD. The mixing zone concept has general use in classifying mixing
phenomena in a wide variety of displacements.
The final definition concerning the Buckley-Leverett development is the
time–distance diagram. These diagrams are plots of xD versus tD on which appear
lines of constant saturation. Figure 5-5 shows a time–distance diagram for the water–
oil displacement in Figs. 5-3(b) and 5-4. The constant saturation curves are straight
lines with slope given by vS1 from Eq. (5.2-9). Similarly, shocks are the bold straight
lines with slope given by Eq. (5.2-12). The region having varying saturation is
shaded. Regions of constant saturation are adjacent to the waves and have no
saturation lines. Time–distance diagrams are very convenient since they subsume
both profiles and histories.
Figure 5-5 Time–distance diagram for displacement of Figs. 5-3(b) and 5-4
139
From the definition of effluent history, the shock portion of the water–oil
displacement arrives at xD = 1 when
S* − S
t D0 = 1* 1I (5.2-16a)
f1 − f1I
from Eqs. (5.2-12) and (5.2-13). The breakthrough time t D0 is an important event in
the displacement; for values tD > t D0 , we are producing some of the water being
injected. The obvious inefficiency of this should suggest that we would like to
conduct the displacement so that t D0 is as large as possible; that is, we would like to
enhance the shock-forming character of the displacement. For tD > t D0 , the water
saturation at the outflow end is given implicitly by
1
f1′|xD =1 = (5.2-16b)
tD
from Eq. (5.2-10). In laboratory floods, it is usually more direct to measure f1 |xD =1 ,
the water “cut,” than the saturation at the effluent end. The water and oil cuts
(1 − f1 |xD =1 ) are functions of only time from Eq. (5.2-16b).
Average Saturations
In the displacement efficiency, we must have some way to calculate average

saturations since, from Eq. (5.1-2), these appear in the definition of ED. These
averages are provided by the Welge integration procedure (Welge, 1952). Consider
the saturation profile in Fig. 5-3(b) at fixed tD, and let xD1 be any dimensionless
position at or behind the shock front position, xD1 ≤ vΔS1 t D . The average water
saturation behind xD1 is
1 xD 1
Sˆ1 (t D ) =
xD1 ∫0
S1dxD (5.2-17)
Equation (5.2-17) may be integrated by parts

Sˆ1 =
1
xD1 ( S11
( xD S1 ) 0D1 − ∫ xD dS1
x
S1J
) (5.2-18)
where S11 = S1 x . Since xD1 is in the spreading portion of the saturation wave, the xD
D1
integrand may be substituted by Eq. (5.2-11)
1 S11
Sˆ1 = S11 −
xD1 ∫S1 J
t D f1′dS1 (5.2-19)
which may be readily integrated (recall tD is fixed) to

t
Sˆ1 = S11 − D ( f11 − f1J ) (5.2-20)
xD
140
Equation (5.2-20) relates the average water saturation behind xD1 to the fractional
flow and saturation at that point. tD may be replaced by Eq. (5.2-11) at this point to
give
(f − f )
Sˆ1 = S11 − 11 1J (5.2-21)
f11′
Equation (5.2-21) is the final form of the Welge integration.

The most common use of this procedure is to let xD1 = 1 after water break-
through (t D ≥ t D0 ), at which point Sˆ1 = S1 , and f11 becomes the water cut. Thus the
water saturation at the outflow end may be calculated from Eq. (5.2-20) as
S1 |xD =1 = S1 − t D ( f1J − f1 |xD =1 ) (5.2-22)
If we know the water cut and average water saturation from direct measurement,
simultaneously applying Eqs. (5.2-16) and (5.2-22) provides a way of estimating
fractional flow curves ( f1 x =1 versus S1 x =1 or f11 versus S11) from experimental data.
D D
The average water saturation follows from Eq. (5.2-21) for S1 with the f1-S1
curve known. This equation may be rearranged to give
f1 |xD =1 − f1J = f1′|xD =1 ( S1 |xD =1 − S1 ) (5.2-23)
Thus S1 at any t D ≥ t D0 is given by the extension of a straight line tangent to the

fractional flow curve at ( f1 , S1 ) xD =1 to intersect with the y coordinate at f1 = f1J. The
dimensionless time required to bring this point to xD = 1 is the reciprocal slope of this
line from Eq. (5.2-16). Figure 5-4 shows the graphical procedure for this. From the
S1 thus determined, S 2 = 1 – S1 may be used in the definition (Eq. 5.1-2) to
calculate ED.
The above construction and Eqs. (5.2-22) and (5.2-23) apply only to
dimensionless times after breakthrough. Before breakthrough the average water
saturation is
S1 = S1I + t D ( f1J − f1I ), t D < t D0 (5.2-24)
by applying the overall water material balance (Eq. 2.5-2) to this special case.
Equations (5.2-22) and (5.2-24) are identical except for the value used for the effluent
water cut.
We are now ready to demonstrate the effect of endpoint mobility ratio M0,
relative permeability, and N g0 sin α on oil displacement efficiency. Figure 5-6
schematically shows the effect of these parameters for displacements with f1I = 0 and
f1J = 1. Figure 5-6 shows, from top to bottom, plots of ED versus tD, water saturation
profiles at various tD, and the fractional flow curve that would give the indicated
behavior. From left to right, the figures show oil displacement behavior for
decreasing M0, increasing N g0 sin α, and increasing water wetness through shifts in
the relative permeability curves. Figure 5-6 represents three of the four types of
waves––spreading, mixed, and sharpening. Several important conclusions follow
directly from Fig. 5-6.
141
Figure 5-6 Schematic illustration of effect of mobility ratio on displacement

efficiency
1. Any change that increases the size of the shock portion of the water
saturation wave also increases ED at any given tD. These changes also delay
water breakthrough and decrease the time over which the permeable medium
is simultaneously producing two phases.
2. Decreasing M0, increasing N g0 sin α, and increasing water wetness improve
ED. Of these three, M0 is usually the only one we can have any impact on. In
Chap. 6, we see that decreasing mobility ratio also increases vertical and
areal sweep efficiency; hence decreasing the mobility ratio improves oil
recovery in at least three ways. EOR processes that rely, partly or totally, on
lowering the mobility ratio between the displacing and displaced fluids are
said to be based on the mobility ratio concept of oil recovery. Figure 5-6
shows that when the water saturation wave becomes a complete shock, no
142
advantage is to be gained on ED by further lowering M0. Finally, there is no
unique value of M0 at which the wave changes from spreading to sharpening
since the displacement is affected also by the shape of the relative
permeability curves.
3. However low M0 might be, the ultimate displacement efficiency
(S − S2 r )
ED∞ = 2 I
S2 I
is limited by the presence of a residual oil saturation. EOR methods that
intend to recover residual oil must rely on something other than the mobility
ratio concept, such as displacing with miscible agents (see Sec. 5-5 and
Chap. 7) or lowering the water–oil interfacial tension (see Chap. 9).
Besides M0, at least two other mobility ratios are in common use. The
average mobility ratio M , defined as
(λr1 + λr 2 ) |S1 = S1
M= (5.2-25a)
(λr1 + λr 2 ) |S1 = S1 I
is the ratio of total relative mobility at the average water saturation behind the shock
front to the same quantity evaluated at the initial water saturation. M is commonly
used to correlate the areal sweep efficiency curves (see Chap. 6). The shock front
mobility ratio Msh is
(λr1 + λr 2 ) |S = S *
M sh = 1 1
(5.2-25b)
(λr1 + λr 2 ) |S1 = S1 I
Msh is the quantity that controls the formation of viscous fingers. For pistonlike
displacements, all three definitions are the same.
The most general definition of mobility ratio is actually the ratio of pressure
gradients ahead of and behind a displacing front. The above definitions, depending
on the character of the displacing front, follow from this for the case of
incompressible flow (spatially independent volumetric flow rate). For compressible
flows or flows of condensing fluids, the general definition is more appropriate (see
Chap. 11 and Exercise 5J).
5-3 DISSIPATION IN IMMISCIBLE DISPLACEMENTS
In this section, we discuss two common dissipative effects in one-dimensional flows:

capillary pressure and fluid compressibility. Both phenomena are dissipative; they
cause mixing zones to grow faster than or differently from a dissipation-free flow.
Both phenomena also bring additional effects.
Capillary Pressure
We do not present a closed-form solution to the water conservation equation. But we

can qualitatively illustrate the effect of capillary pressure on a water–oil displacement
143
and can give, through scaling arguments, quantitative guidelines on when it might be
important. For incompressible fluids and with capillary pressure Pc included, the
water material balance (Eq. 5.2-1) still applies, but the water fractional flow (Eq.
5.2-2) becomes (see Exercise 5F)
λr1 ⎛ k λr 2 Δρ g sin α ⎞ k λr1 (∂Pc / ∂x)

f1 ( S1) = 1− ⎟+ (5.3-1)
λr1 + λr 2 ⎜⎝ u ⎠ (1 + λr1 / λr 2 )u
The first term on the right side of Eq. (5.3-1) is simply the water fractional flow in
the absence of capillary pressure (Eq. 5.2-2); thus many of the conclusions about
displacements with Pc = 0, though somewhat modified, carry over to displacements
with capillary pressure. The second right term in Eq. (5.3-1) is the contribution of Pc
to the water fractional flow. Including the capillary pressure term causes the character
of Eq. (5.2-1) to change from hyperbolic to parabolic, a general result of dissipative
effects because of the spatial Pc derivative.
The capillary pressure in Eq. (5.3-1) is the phase pressure difference between
two continuous oil and water phases (see Sec. 3-2). The derivative ∂Pc/∂x =
(dPc/dS1)•(∂S1/∂x) has a positive sign for displacements in both oil-wet or water-wet
media since dPc/dS1 is negative for both cases (see Fig. 3-5), and ∂S1/∂x is also
negative. Therefore, for waterfloods, capillary pressure increases the water fractional
flow at a given water saturation. This augmentation is particularly important in
regions having large saturation gradients, that is, around shock fronts predicted by the
Buckley-Leverett theory. In an oil displacement of water, Pc causes a smaller water
fractional flow since ∂S1/∂x > 0.
The effect of Pc on a one-dimensional displacement is to spread out the water
saturation wave, particularly around shocks; Fig. 5-7, which illustrates how this
comes about, is a simulated water saturation and pressure profile for a one-
dimensional waterflood in a water-wet medium. Figure 5-7(a) shows water saturation
profiles with and without capillary pressure; Fig. 5-7(b) shows the corresponding
pressure profiles. Both panels are at the same tD. The dotted phase pressures in Fig.
5-7(b) are those that would be present if the shock remained in the water saturation
profile. Of course, representing shock waves with Pc ≠ 0 is not correct, but such a
portrayal presents the driving force for capillary mixing.
Ahead of the front (downstream), the difference between the oil and water
phase pressures is constant and equal to the capillary pressure at S1I. At the front, the
phase pressures change rapidly. But behind the front (upstream), the difference
between the oil and water phase pressures declines to the value at S1 = S1J. Compare
these comments to Figs. 5-7(a) and 3-5. There is now a local pressure gradient at the
shock that causes oil to flow upstream (countercurrent imbibition) and water to flow
downstream faster than under the influence of viscous forces only. The resulting local
mixing causes the shock to spread (Fig. 5-7a) and the pressure discontinuity to
disappear. Behind the front, in the spreading portion of the water saturation wave, the
effect of capillary pressure is small.
Capillary pressure will be small if the system length L is large. Consider the
144
Figure 5-7 Saturation and pressure

profiles under longitudinal capillary
imbibition (Yokoyama, 1981)
dimensionless water conservation equation with Eq. (5.3-1) substituted and α = 0
∂S1 ∂ ⎛ 1 ⎞ ∂ ⎛ k λr1 ∂Pc ⎞

+ ⎟ + ∂x ⎜ ⎛ ⎟=0 (5.3-2)
∂t D ∂xD ⎜ 1 + λr 2 D ⎜ λ ⎞ ∂x
⎜⎜ ⎟⎟ uL ⎜ 1 + r1 ⎟ D ⎟
λr1 ⎜ ⎟
⎝ ⎠ ⎝ ⎝ λr 2 ⎠ ⎠
The last term on the left side of this equation is nonlinear in S1 and thus difficult to
estimate. Using the Leverett j-function expression (Eq. 3.2-2), we can write Eq.
(5.3-2) as
∂S1 ∂ ⎛ 1 ⎞ 1 ∂ ⎛ ∂S1 ⎞
+ ⎟ − N ∂x ⎜ g ( S1 ) ∂x ⎟ = 0 (5.3-3)
∂t D ∂xD ⎜ 1 + λr 2 D ⎝ D ⎠
⎜⎜ ⎟⎟ RL
⎝ λr1 ⎠
145
where g is a positive dimensionless function of water saturation
n1
⎛ 1 ⎞ ⎛ S1 − S1r ⎞ dj
g ( S1 ) = − ⎜ ⎟ ⎜ 1 − S − S ⎟ dS (5.3-4)
λ
⎜⎜ 1 + r 2 ⎟⎟ ⎝ 2r 1r ⎠ 1
⎝ λr1 ⎠
and NRL, the Rapoport and Leas number, is a dimensionless constant first implied by
these authors (1953) to indicate when capillary pressure effects will be important.
1/ 2
⎛φ ⎞ μ1uL
N RL = ⎜ ⎟ (5.3-5)
⎝ ⎠
k k 0
r1φσ 12 cosθ
Figure 5-8 is a plot of fractional oil recovery at water breakthrough versus

μ1vL (recall v = u/φ) from the experimental work of Rapoport and Leas. Since the
S1I = 0 in their cores, the vertical axis in Fig. 5-8 is the breakthrough displacement
efficiency, ED0 . As μ1vL increases, ED0 increases to a maximum of 0.58. For
larger μ1vL, ED0 is constant at the value predicted by the Buckley-Leverett theory.
Rapoport and Leas did not plot their results against the more general NRL;
however, using the given k = 0.439 μm2 and φ = 0.24, and taking kr01σ 12 cosθ = 1
mN/m (typical for water-wet media), Pc will not affect a one-dimensional water–oil
Figure 5-8 Relation between oil recovery at breakthrough and scaling coefficient
in dry-filmed alumdum cores with no connate water. Different symbols represent
varying core lengths and oil viscosities. (From Rapoport and Leas, 1953)
146
displacement if NRL is greater than about 3. Because of the length appearing in the
numerator of Eq. (5.3-5), Pc will affect the displacement front to a much greater
degree in laboratory floods than in field-scale displacements because of the large
disparity in L.
Of course, on a microscopic scale, capillary forces are important in
determining the amount of trapped or residual oil in either laboratory or field
displacements. In Sec. 3-3, we saw that the S2r depended on a local viscous-
to-capillary-force ratio, the capillary number Nvc. A common form of capillary
number Nvc = vμ1 / kr01σ 12 cosθ is embedded in the definition of NRL
1/ 2
⎛φ ⎞
N RL =⎜ ⎟ LN vc (5.3-6)
⎝k⎠
The factor, L(φ/k)1/2 , is a measure of the ratio of the macroscopic permeable medium
dimension to a characteristic rock dimension. Therefore, Nvc and NRL are expressing
the same physical idea––capillary-to-viscous-force ratios––but at different scales.
Recall that if Nvc is less than about 10–5, the residual phase saturations are
roughly constant. For well-sorted media, we can then put limits on NRL so that
capillary forces, on any scale, do not affect the displacement
1/ 2
−5 ⎛φ ⎞
3 < N RL < 10 L ⎜ ⎟ (5.3-7)
⎝k⎠
(no dissipation) (constant residual
saturations)
For large L, this is an extremely wide range and accounts for the common neglect of
all capillary forces in one-dimensional displacement calculations. For laboratory
scale, it may not be possible to satisfy both requirements.
NRL may be expressed in more direct ways. From Eq. (5.3-5), we can
substitute Darcy’s law for water evaluated at S1 = 1 – S2r for v = u/φ to obtain
1/ 2
φ ΔP1
′ = ⎜⎛ ⎟⎞
N RL (5.3-8)
⎝ k ⎠ σ 12 cosθ
where ΔP1 is the pressure drop across the permeable medium measured through the
water phase. The terms containing permeability and interfacial tension may be
expressed in terms of the Leverett j-function to give yet another approximation to NRL
ΔP
′′ = 1
N RL (5.3-9)
ΔPc
where ΔPc is the change in capillary pressure between the initial and final water
saturation states. Equation (5.3-9) is a direct comparison of viscous to capillary
pressure drops and is the least rigorous, but most direct, of all the measures.
147
For small NRL, capillary pressure will cause shock waves to spread out.
Though there is a parallel between dispersion in miscible displacements (see Sec.
5-5) and Pc effects in immiscible displacements, the analogy does not carry over to
mixing zone growth. We show in Sec. 5-5 that dispersive mixing zones grow in
proportion to the square root of time. Capillary pressure generally causes mixing
zones to grow exponentially to some asymptotic limit where it proceeds, without
further growth, in simple translation. How this comes about may be qualitatively
explained by considering: a water saturation wave that would be a shock over the
entire possible Saturation range, as in the right column in Fig. 5-6, where we
neglected Pc effects. As we have seen, Pc effects cause such a wave to spread, but
there is still a strong tendency for the wave to sharpen because of the convex-upward
shape of the fractional flow curve. These two effects tend to balance each other,
causing the wave to approach an asymptotic limit. The existence of such a limit
further restricts the importance of capillary pressure as a mixing mechanism in one
dimension. Asymptotic or “stabilized” mixing zones in one-dimensional laboratory
waterfloods have been noted by several authors (Bail and Marsden, 1957).
No discussion of how capillary pressure influences a one-dimensional dis-
placement is complete without some mention of the capillary end effect. This effect
occurs when there is a discontinuity in the capillary pressure curve as, for example,
when the one-dimensional permeable medium consists of two homogeneous media of
differing permeabilities arranged in series. But it most commonly occurs at the end of
a laboratory core where the flowing phases pass from a permeable to a nonpermeable
region. The saturation behavior at the plane of discontinuity is considerably different
from that predicted by the Buckley-Leverett theory.
Consider the water saturation and pressure profiles of a waterflood in a
water-wet medium shown in Fig. 5-9. Capillary forces are such that they cannot be
neglected. Figure 5-9(a) shows the instant that water arrives at the outflow end (x =
L), and Fig. 5-9(b) shows some time later. On the right of the outflow end, there is no
permeable medium. This region has a capillary pressure curve that is zero
everywhere except at S1 = 0, where all values of capillary pressure exist. The oil and
water phase pressures must be continuous at x = L; hence the water saturation for x >
L is constrained to be zero because there is a nonzero phase pressure difference. This,
in turn, implies water cannot flow across the outflow end of the medium until the
capillary pressure just inside the system vanishes. With no production at x = L, but
with continual water transport to the outflow end, the water saturation must build up
at x = L until Pc = 0 (S1 = 1 – S2r) at this plane. Hence the capillary end effect causes a
delay in water production and a distortion of the water saturation at x = L compared
to that predicted by the Buckley-Leverett theory (Fig. 5-9b).
This delay can cause considerable error in applying the Welge integration
procedure (Eq. 5.2-22). The capillary end effect has been observed experimentally by
Kyte and Rapoport (1958) and in simulations by Douglas et al. (1958). Figure 5-10
reproduces data reflecting the capillary end effect.
148
Figure 5-9 Schematic of the capillary end effect
To eliminate the capillary end effect, laboratory floods have been run at high
velocities and with long lengths (both increase NRL) or by placing a second permeable
material at the outflow end to ensure good capillary contact.
Fluid Compressibility
A second dissipative effect is fluid compressibility. Figure 5-11 shows water

saturation profiles for two waterfloods having compressible oil and incompressible
water (Fig. 5-11a) and compressible water and incompressible oil (Fig. 5-11b). The
completely incompressible Buckley-Leverett case is shown for comparison. These
results are from computer simulations that were at constant water injection rate (Fig.
5-11a) and constant oil production rate (Fig. 5-11b). We present the results as the
product of compressibility and total pressure drop ΔP (neglecting capillary forces)
since this quantity determines the appropriateness of the small compressibility fluid
assumption in well test analysis. For cjΔP products of 0.01 or less, the effect of fluid
compressibility is negligible; the smearing of the shock fronts for the cjΔP = 1.25 ×
10–3 runs is because of numerical dispersion, which is an artificial dissipative effect.
149
Figure 5-10 Correlation of waterflood test data in strongly water-wet alumdum

cores (from Kyte and Rapoport, 1985)
The cjΔP products shown in Fig. 5-11 are, of course, unrealistically high; we have
selected these values merely to emphasize the effect of compressibility.
The effect of either oil or water compressibility is to spread out the Buckley-
Leverett shock front in addition to the spreading caused by numerical dispersion but
the effect does not become pronounced until cjΔP is 1 or more. However, we would
expect displacements in which both fluids are compressible to experience a combined
dissipative effect with greater spreading. In Fig. 5-11(a), the water saturation exceeds
1 – S2r at the inflow end. At higher pressure, oil compression below its residual
occurs. Similarly, in Fig. 5-1l(b), the water saturation exceeds S1r at the effluent end
because, at the reduced pressure, the water will expand. These effects are
characteristics of the particular conditions the runs were made under. If we had held
the production pressure constant and not allowed phase saturations to decrease below
their respective residuals, neither effect would be present. Still, we can see from Fig.
5-11 that the effect of compressibility is qualitatively similar to that of capillary
pressure; a spreading of the shock fronts occurs but with a smaller effect on the
saturation “tail.”
150
Figure 5-11 Water saturation profiles for one-dimensional water-displacing-oil

floods at t = 200 days (adapted from Samizo, 1982)
151
5-4 IDEAL MISCIBLE DISPLACEMENTS
Two components are mutually miscible if they mix in all proportions without an
interface forming between them. The definition is translated into the fluid flow
equations by allowing a phase to be composed of several components they are
mutually miscible within.
In this section, we discuss isothermal miscible displacements using fractional
flow theory and with one or more phases present. Our presentation considers ideal
miscible displacements with components that do not change the properties of the
phases they are formed in (see Chap. 7 for more complicated displacements).
Concentration Velocities
Many of the concepts in Sec. 5-2 readily generalize to miscible displacements. We

write a one-dimensional conservation equation for i = 1, . . . , Nc components as
∂ ⎛ NP ⎛1−φ ⎞ ⎞ ∂ ⎛ NP ⎞
φ ⎜ ∑ S j Cij + ⎜ ⎟ Cis ⎟ + u ⎜ ∑ f j Cij ⎟ = 0,
∂t ⎝ j =1 ⎝ φ ⎠ ⎠ ∂x ⎝ j =1 ⎠
i = 1, . . . , NC (5.4-1)
Equation (5.4-1) is a special case of Eq. (2.4-10) with dispersion neglected. fj is the
fractional flow of phase j, given by Eq. (2.4-2) with capillary pressure neglected, and
Cij and Cis are the phase concentrations of component i in phase j and on the solid,
respectively. Of course, the assumptions associated with Eq. (2.4-10)––constant
porosity, incompressible fluids, and ideal mixing––also apply. In nondimensional
form, Eq. (5.4-1) becomes
∂ ∂F
(Ci + Cis′ ) + i = 0, i = 1, . . . , N C (5.4-2)
∂t D ∂xD
where
Ci = Overall fluid phase concentration of species i
NP
= ∑ S j Cij (5.4-3a)
j =1
Cis′ = Solid phase concentration of i on a pore volume basis

⎛1−φ ⎞
= Cis ⎜ ⎟ (5.4-3b)
⎝ φ ⎠
Fi = Overall flux of species i
NP
= ∑ f j Cij (5.4-3c)
j =1
The transform accomplished by Eq. (5.4-3b) changes the solid phase concentration
from a solid volume basis (Cis is amount i on solid/volume solid) to a pore volume
152
basis ( Cis′ is amount i on solid/pore volume). Thus Ci and Cis′ are directly comparable
and may be used together in later work without the need to manipulate units. The
definition of overall flux is from Hirasaki (1981) and Helfferich (1981).
In principle, the fluxes Fi are functions of the Ci for i = 1, . . . , NC, and we
may carry over many of the definitions, particularly those of saturation velocity,
directly from Sec. 5-2. In practice, however, the relations Fi = Fi(C1, C2, . . . , C NC )
are extremely convoluted. We discuss this in more detail later, but we can give a
summary of this relation here.
With Ci known, the Cij and Sj may be calculated from phase equilibrium
relations. The exact nature of the “flash” calculation depends on the nature of the
phase -behavior (see Sec. 4-4 and Chaps. 7 and 9). With the Sj and Cij known, the
phase relative permeabilities krj = krj(Sj, Cij) and viscosities μj = μj(Cij) may be
calculated from petrophysical relations (see Sec. 3-3). From these follow the relative
mobilities λrj = krj/μj, which lead directly to the fj from Eq. (2.4-2). If the phase
densities are also required (if, for example, the permeable medium is not horizontal),
they follow from ρj = ρj(Cij) (Eq. 2.2-12). With the fj and Cij known Fi follows from
Eq. (5.4-3c). If needed, Cis′ = Cis′ (Cij) may be calculated also from the adsorption
isotherm (see Chaps. 8 and 9).
Despite this complexity, we can write Eq. (5.4-2) as
⎛ ⎛ ∂C ′ ⎞ ⎞ ∂C ⎛ ∂F ⎞ ∂C
⎜ 1 + ⎜ is ⎟ ⎟ i + ⎜ i ⎟ i
= 0, i = 1, . . . , N C (5.4-4)
⎜ ⎝ ∂Ci ⎠ x ⎟ ∂t D ⎝ ∂Ci ⎠t ∂xD
⎝ D ⎠ D
The partial derivatives (∂Cis′ / ∂Ci ) xD and (∂Fi / ∂Ci )tD in Eq. (5.4-4) follow from the
chain rule. These derivatives are not the same as (∂Cis′ / ∂C j )Cm≠ j , which are in the
definition of the total differential. The latter derivatives may be calculated directly
from Cis′ = Cis′ (Cij) and Fi = Fi(Ci), whereas the former derivatives require knowledge
of Ci = Ci(xD, tD), which are solutions. Therefore, Eq. (5.4-4) is of little use except to
allow the definition of specific concentration velocity vCi
(∂Fi / ∂Ci )tD
vCi = , i = 1, . . . , N C (5.4-5a)
1 + (∂Cis′ / ∂Ci ) xD
by analogy with Eq. (5.2-10). The definition of the specific shock velocity vΔCi is
(ΔFi / ΔCi )
vΔCi = (5.4-5b)
1 + (ΔCis′ / ΔCi )
Without additional constraints, the definitions (Eqs. 5.4-5a and 5.4-5b)

impart no new information. But for the water-oil case of Sec. 5-2, they reduce to Ci =
S1, Fi = f1, and Cis′ = 0, giving
⎛ ∂f ⎞ df
vCi = vS1 = ⎜ 1 ⎟ = 1 = f1′( S1 ) (5.4-6)
⎝ ∂S1 ⎠t
D
dS1
The last equality is possible because f1 is a function of S1 only; hence f1′ =
(∂F1 / ∂S1 )tD = (∂F1 / ∂S1 ) xD . Certainly for more complicated cases, this simplification
153
is not possible; still, many of the displacements of interest may be solved with the
coherent or simple wave theory that we discuss in Sec. 5-5. We now discuss other
particularly simple special cases of miscible displacements.
Tracers in Two-Phase Flow
The simplest case we consider is the miscible displacement in single-phase flow of

component 2 by component 1. For this case, fj and Sj are zero for all j except 1. For
this particular j, fj and Sj are unity. If component 1 does not adsorb, the concentration
velocity becomes
vC1 = 1 (5.4-7)
from either Eq. (5.4-5a) or (5.4-5b). This seemingly trivial result has two important
consequences.
1. The dimensional velocity of component 1 is equal to the bulk fluid velocity,

meaning the dimensionless breakthrough time t D0 for component 1 is also
unity. From Eq. (5.2-7), we may estimate the pore volume of the medium by
knowing the cumulative fluid injected when breakthrough occurs (see
Exercise 5K). Components that travel at the bulk fluid velocity are
“conservative” tracers for this reason.
2. The specific concentration velocity is independent of C1, meaning waves
caused by conservative tracers are indifferent, which is generally true for
ideal miscible displacements.
Most EOR displacements are only partially miscible. To illustrate a partially

miscible displacement, we now consider a displacement of oil–water mixture at water
saturation S1I by another at a water fractional flow f1J = f1(S1J). We wish to distinguish
between the initial and injected oil and water, so let's suppose the injected fluids
contain conservative tracers. The oil-miscible tracer is completely immiscible in
water, and the water-miscible tracer is similarly immiscible in oil. The process is now
the displacement of an oil–water mixture by a tagged oil–water mixture. To keep this
simple, we assume the tracers do not affect the fractional flow functions at all. The
specific velocity of the tagged water-resident water wave is
∂ (C11 f1 ) f1
v1′ = = (5.4-8a)
∂ (C11S1 ) S1
from Eq. (5.4-5a), where C11 is the water tracer concentration. Similarly, the specific
velocity of the tagged oil is
f 1 − f1
v2′ = 2 = (5.4-8b)
S2 1 − S1
v1′ and v2′ are both independent of tracer concentration; hence the miscible tagged
water and oil waves are indifferent. Of course, since neither of the tracers affects f1,
154
the saturation velocity of the water––tagged or untagged––is given by Eq. (5.2-10) or
Eq. (5.2-12). The values of f1 and S1 in Eq. (5.4-8) are determined by the character of
the oil–water wave.
Figure 5-12 illustrates some of the cases that can occur for this displacement.
On each plot, the fractional flow curve is on the left, and a saturation-concentration
profile is on the right. In case A, S1I = S1J and the specific velocities are the slopes of
straight lines passing through (0, 0) and (f1, S1)J and (1, 1) and (f1, S1)J, respectively,
from Eqs. (5.4-8a) and (5.4-8b). v2′ > v1′, and the tagged oil wave leads the tagged
water wave.
Figure 5-12 Illustration of various partially miscible displacements

155
Figure 5-12 Continued.

156
In case B, S1J > S1I, and the f1 curve is such that the oil–water wave is a
shock. Both tagged waves lag the oil–water wave. The region between the tagged
water and oil–water waves contains a “bank” of resident water that will be produced
before the injected water breakthrough. Breakthrough of a resident water bank in this
manner has been observed experimentally (Brown, 1957) though dispersion tends to
be large in such displacements (see Fig. 5-18).
Case C illustrates a spreading water–oil wave with v2′ > v1′ but with all tagged
concentration waves having a smaller velocity than the smallest saturation velocity at
S1J.
Case D is the same as case C with the fractional flow curve more convex
upward. This shape causes the oil–water wave to spread more and the tagged oil front
to fall somewhere in the spreading portion of the oil–water wave. The saturation, S1′,
whose velocity is the same as the tagged oil wave, is given by
1 − f1 ( S1′ ) ⎛ df1 ⎞
v2′ = =⎜ ⎟ (5.4-9)
1 − S1′ ⎝ dS1 ⎠ S
1′
The line whose slope is v2′ does not pass through S1J, as it did in all previous cases.
This is because a line through (1, 1) and (S1, f1)J would have a second intersection
point with the fractional flow curve. The tagged oil front would then travel with two
different water saturations––a physical impossibility.
Case E, the traditional Buckley-Leverett problem, is the inverse of case D
where the tagged water front is now traveling in the spreading zone region. The oil–
water displacement in case E is mixed, whereas in case D, it is spreading.
The important points in Fig. 5-12 are as follows:
1. As postulated, neither the tagged oil nor the tagged water causes deviation in
the water–oil displacement character. When banks of resident fluids form,
they do so within their respective phases.
2. One can easily imagine the tagged oil to be a hydrocarbon of less value than
the oil. The tracer fronts now take on added significance since these miscible
fronts are now displacing the resident oil. The resident oil, in turn, is
completely displaced. Thus the ultimate ED for these idealized displacements
is 1.0. This maximum efficiency occurs without interfacial tension lowering,
changes in wettability, or mobility reduction.
Of course, we have not as yet discovered a fluid that is simultaneously

cheaper than and miscible with crude oil and that does not drastically change the
hydrocarbon transport properties. These changes can return the ultimate displacement
efficiency to something less than 1; still, the idea of displacing with miscible fluids,
or those that will develop miscibility, is the central concept of Chap. 7.
157
5-5 DISSIPATION IN MISCIBLE DISPLACEMENTS
Because miscible waves are ideally indifferent, they are also susceptible to
dissipation. By far the most prominent of the dissipative effects in miscible
displacements are dispersion and viscous fingering. The latter is a two-dimensional
effect, so we postpone our discussion of it to Chaps. 6 and 7. In this section, we
discuss the effects of dispersion on a miscible front.
The Error Function Solution
Consider now the isothermal miscible displacement of a component by another it is

completely miscible with in a one-dimensional, homogeneous permeable medium.
The convection–diffusion (CD) equation (Eq. 2.4-7) describes the conservation of the
displacing component with mass concentration C,
∂C ∂C ∂ 2C
φ +u − φ Kl 2 = 0 (5.5-1)
∂t ∂x ∂x
Equation (5.5-1) also assumes incompressible fluid and rock, ideal mixing, and a
single phase at unit saturation. The following development is valid if other phases are
present (Delshad, 1981) and as long as all fractional flows and saturations are
constant (see Exercise 5M). Kl is the longitudinal dispersion coefficient. In
dimensionless terms, Eq. (5.5-1) becomes
∂C ∂C 1 ∂ 2C
+ − =0 (5.5-2)
∂t D ∂xD N Pe ∂xD 2
which is solved with the following boundary and initial conditions on C(xD, tD):
C ( xD ,0) = CI , xD ≥ 0 (5.5-3a)
C ( xD → ∞, t D ) = C I , tD ≥ 0 (5.5-3b)
C (0, t D ) = C J , tD ≥ 0 (5.5-3b)
where CI and CJ are the initial and injected compositions, respectively. In Eq. (5.5-2),
NPe, the Peclet number, is defined as
uL
N Pe = (5.5-4)
φ Kl
which is the ratio of convective to dispersive transport. NPe is the analogue of NRL for
immiscible displacements as seen by comparing Eqs. (5.3-3) and (5.5-2). This
displacement must take place at constant u unlike Eqs. (5.2-6b). The equation and
boundary conditions contain three independent parameters, CI, CJ, and NPe, but the
problem may be restated with only NPe as a parameter by defining a dimensionless
158
concentration CD
C − CI
CD = (5.5-5)
C J − CI
With this definition, the equation and boundary conditions become

∂CD ∂CD 1 ∂ 2CD
+ − =0 (5.5-6)
∂t D ∂xD N Pe ∂xD 2
CD ( xD ,0) = 0, xD ≥ 0 (5.5-7a)
CD ( xD → ∞, t D ) = 0, t D ≥ 0 (5.5-7b)
CD (0, t D ) = 0, t D ≥ 0 (5.5-7c)
We have replaced the original boundary condition at xD = 0 (Eq. 5.5-3c) with one at
xD → –∞ (Eq. 5.5-7c). This is an approximation to simplify the following derivation
of an analytic solution. The approximate solution thus obtained will be valid, strictly
speaking, for large tD or large NPe where the influence of the inlet boundary appears
as though it were a great distance from the displacing front. In practice, the resulting
approximate analytic solution accurately describes single-phase displacements for all
but extreme cases.
The first step in deriving CD(xD, tD) is to transform Eqs. (5.5-6) and (5.5-7) to
a moving coordinate system xD′ where xD′ = xD – tD. This may be done by regarding
CD as a function of xD and tD where a differential change in CD caused by differential
changes in xD and tD is
⎛ ∂C ⎞ ⎛ ∂C ⎞
dCD = ⎜ D ⎟ dt D + ⎜ D ⎟ dxD (5.5-8a)
⎝ ∂t D ⎠ x
D
⎝ ∂xD ⎠t
D
But regarded as a function of xD′ and tD, dCD is

⎛ ∂C ⎞ ⎛ ∂C ⎞
dCD = ⎜ D ⎟ dt D + ⎜ D ⎟ dxD′ (5.5-8b)
⎝ ∂t D ⎠ xD′ ⎝ ∂xD′ ⎠tD
Differential changes in variables are equal regardless of the coordinate

system they are viewed in. The right- hand sides of Eqs. (5.5-8a) and (5.5-8b) are
therefore equal. But xD′ is a known function of xD and tD, from which
dxD′ = dxD − dt D (5.5-9)
When dxD′ is replaced in the above equality, we have

⎡⎛ ∂C ⎞ ⎛ ∂C ⎞ ⎤
⎢⎜ D ⎟ − ⎜ D ⎟ ⎥ dxD
⎢⎣⎝ ∂xD ⎠tD ⎝ ∂xD′ ⎠tD ⎥⎦
⎡⎛ ∂C ⎞ ⎛ ∂C ⎞ ⎛ ⎛ ∂C ⎞ ⎞ ⎤
+ ⎢⎜ D ⎟ − ⎜ D ⎟ + ⎜ ⎜ D ⎟ ⎟ ⎥ dt D = 0 (5.5-10)
⎢⎝ ∂t D ⎠ xD ⎝ ∂t D ⎠ xD′ ⎜⎝ ⎝ ∂xD′ ⎠tD ⎟⎠ ⎥
⎣ ⎦
159
Since xD and tD are independent variables, dxD and dtD are not linearly related; hence
the terms in brackets in Eq. (5.5-10) are zero, giving
⎛ ∂CD ⎞ ⎛ ∂CD ⎞
⎜ ⎟ =⎜ ⎟ (5.5-11a)
⎝ ∂xD ⎠tD ⎝ ∂xD′ ⎠tD
⎛ ∂CD ⎞ ⎛ ∂CD ⎞ ⎛ ⎛ ∂C ⎞ ⎞
⎟ =⎜ ⎟ − ⎜⎜ ⎜ ⎟ ⎟
D
⎜ (5.5-11b)
⎝ D ⎠ xD ⎝ D ⎠ xD′ ⎝ ⎝ D ⎠tD ⎠⎟
∂t ∂t ∂x ′
When these are substituted into Eq. (5.5-6), we have

⎛ ∂CD ⎞ 1 ∂ 2CD
⎜ ⎟ − =0 (5.5-12)
⎝ ∂xD ⎠ x′ N Pe ∂ ( xD′ )
2
D
and the boundary conditions retain the form of Eq. (5.5-7) thanks to the replacement
of the inlet boundary condition at xD = 0 with one at xD → –∞.
Equation (5.5-12) is now the heat conduction equation whose solution may
be obtained by the method of combination of variables (Bird et al., 1960). To do this,
we define yet another dimensionless variable η = xD′ / 2 t D / N Pe , with which the
governing equations and boundary conditions may be transformed into
dCD d 2CD
2η + =0 (5.5-13a)
dη dη 2
CD (η → ∞) = 0 (5.5-13b)
CD (η → −∞) = 1 (5.5-13c)
As required for the successful transformation of a partial to an ordinary differential

equation, the conditions (Eqs. 5.5-7a and 5.5-7b) collapse into the single condition
(Eq. 5.5-13b). The transformation to an ordinary differential equation is sometimes
called Boltzmann's transformation. Equations (5.5-13) may be separated and
integrated twice to give
1⎛ 2 η −u2 ⎞
π ∫0
CD = ⎜1 − e du ⎟ (5.5-14)
2⎝ ⎠
The product times the integral on the right side of Eq. (5.5-14) is the error function, a
widely tabulated integral (see Table 5-2 and Fig. 5-13), and abbreviated with the
symbol erf(η). By substituting the definitions for η and xD′ , we have the final form
for the approximate analytic solution.
1⎡ ⎛ x −t ⎞⎤ 1 ⎛ xD − t D ⎞
CD = ⎢ 1 − erf ⎜ D D ⎟ ⎥ = 2 erfc ⎜ ⎟ (5.5-15)
2⎢
⎜ 2 tD ⎟⎥ ⎜ 2 tD ⎟
⎢⎣ ⎜ N Pe ⎟⎥ ⎜ N Pe ⎟
⎝ ⎠⎦ ⎝ ⎠
160
TABLE 5-2 TABULATED VALUES OF ERF (x) (FROM JAHNKE AND EMDE, 1945)
x 0 1 2 3 4 5 6 7 8 9 d
0.0 0.0 000 113 226 338 451 564 676 789 901 *013 113
1 0.1 125 236 348 459 569 680 790 900 *009 *118 111
2 0.2 227 335 443 550 657 763 869 974 *079 *183 106
3 0.3 286 389 491 593 694 794 89 992 *090 *187 100
4 0.4 284 380 475 569 662 755 847 937 *027 *117 93
5 0.5 205 292 379 465 549 633 716 798 879 959 86
6 0.6 039 117 194 270 346 420 494 566 638 708 76
7 778 847 914 981 *047 *112 *175 *128 *300 *361 65
8 0.7 421 480 538 595 651 707 761 814 867 918 56
9 969 *019 *068 *116 *163 *209 *254 *299 *342 *385 46
1.0 0.8 427 468 508 548 586 624 661 698 733 768 38
1 802 835 868 900 931 961 991 *020 *048 *076 50
2 0.9 103 130 155 181 205 229 252 275 297 319 24
3 340 361 381 400 419 438 456 473 490 507 19
4 0.95 23 39 54 69 83 97 *11 *24 *37 *49 14
5 0.96 61 73 84 95 *06 *16 *26 *36 *45 *55 10
6 0.97 63 72 80 88 96 *04 *11 *18 *25 *32 8
7 0.98 38 44 50 56 61 67 72 77 82 86 6
8 91 95 99 *03 *07 *11 *15 *18 *22 *25 4
9 0.99 28 31 34 37 39 72 44 47 49 51 3
2.0 0.995 32 52 72 91 *09 *26 *42 *59 *73 *88 17
1 0.997 02 15 28 41 53 64 75 95 95 *05 11
2 0.998 14 22 31 39 46 54 61 67 74 80 8
3 86 91 97 *02 *06 *11 *15 *20 *24 *28 5
4 0.999 31 35 38 41 44 47 50 52 55 57 3
5 59 61 63 65 67 69 71 72 74 75 2
6 76 78 79 80 81 82 83 84 85 86 1
7 87 87 88 89 89 90 91 91 92 92 1
8 0.9999 25 19 33 37 41 44 48 51 54 56 3
9 59 61 64 66 68 70 72 73 75 77 2
where erfc denotes the complementary error function. The exact analytic solution as
derived by Laplace transforms is (Marle, 1981)
⎛ x −t ⎛ x +t
x N
1 ⎞ e D Pe ⎞
CD = erfc ⎜ D D ⎟ + erfc ⎜ D D ⎟ (5.5-16)
2 2
⎜ 2 tD ⎟ ⎜ 2 tD ⎟
⎜ N Pe ⎟ ⎜ N Pe ⎟
⎝ ⎠ ⎝ ⎠
The second term in Eq. (5.5-16) approaches zero exponentially as xD and NPe grow.
Figure 5-14 shows concentration profiles of CD versus xD with tD and NPe
varying. As NPe increases, the concentration profile approaches the step function at xD
= tD suggested by Eq. (5.4-7). In fact, the concentration profile given by Eq. (5.5-15)
161
1 xn − v (1/ n ) −1
n! ∫0
Figure 5-13 The function En(x) = e v dv , where n = 2 is the error
function (from Jahnke and Emde, 1945)
is symmetric and centered on this point. The complete solution (Eq. 5.5-16) is not
symmetric, but as we noted, this effect is small. Dispersion, therefore, does not affect
the rate of wave propagation, but it does affect the degree of mixing in the wave.
The displacement efficiency for the displaced component is
1
ED = CD = ∫ CD ( xD , t D ) dx =
0
⎛ tD ⎞ ⎡ ⎛ − t D N Pe ⎞ ⎛ 1 − tD ⎞⎤
⎜ ⎟ ⎢ierfc ⎜ ⎟ − ierfc ⎜ ⎟⎥ (5.5-17)
⎜ N Pe ⎟ ⎢ ⎜ 2 ⎟
⎝ ⎠ ⎝ ⎠ ⎜ 2 tD ⎟⎥
⎢ ⎜ N Pe ⎟⎥
⎣ ⎝ ⎠⎦
where ierfc (x) = ∫x∞ erfc (ξ) dξ is the integral complementary error function also
tabulated (Carslaw and Jaeger, 1959). Figure 5-15 plots ED versus tD for various NPe.
ED decreases at fixed tD as dispersion increases. Since miscible displacements do not
have residual phase saturations, ED approaches 1 as tD increases. Figures 5-14 and
5-15 indicate a stronger effect of NPe on concentration profiles than on displacement
162
Figure 5-14 Dimensionless concentration

profiles
efficiency; hence concerns about the detrimental effect of dispersion on recovery are
usually limited to slugs. This topic we defer to Sec. 7-6.
The dimensionless mixing zone, the distance between the distances where CD
= 0.1 and CD = 0.9, follows from Eq. (5.5-15),
tD
ΔxD = xD |CD = 0.1 − xD |CD = 0.9 = 3.625 (5.5-18)
N Pe
To arrive at this, invert Eq. (5.5-15) for xD |CD = 0.1 to yield
tD
xD |CD = 0.1 = t D + 2 erf −1 (0.8)
N Pe
163
Figure 5-15 Displacement efficiency for one-dimensional miscible displacements
A similar procedure yields xD |CD = 0.9 , and these substituted into the definition for ΔxD
give Eq. (5.5-18). Equation (5.5-18) shows that dispersive mixing zones grow in
proportion to the square root of time. Immiscible mixing zones grow in proportion to
time. The growth suggested by Eq. (5.5-18) is generally slower than that for an
immiscible mixing zone, particularly if NPe is large. This slow growth is a partial
justification for neglecting dispersion in modeling sernimiscible displacements
compared to fractional flow effects.
ΔxD is also useful to compare laboratory to field mixing zone lengths. An
immiscible mixing zone contains no free parameters if dissipation is small.
Therefore, if we conduct a laboratory immiscible flood under conditions as nearly
identical to a field prototype as possible (displacement in native or restored state
cores, at reservoir temperature and pressure, using actual reservoir fluids), the
laboratory ΔxD will be the same as in the field.
In miscible displacements, we are generally unable to make NPe equal
between the laboratory and the field. Moreover NPe is usually smaller in the
laboratory; thus ΔxD usually will be larger in the laboratory than in the field. Of
course, the dimensional mixing zone length, ΔxDL, will always be greater in the field
because L is much greater. Why we are unable to match NPe is derived from the
following discussion of dispersion coefficients.
Dispersivity
Bear (1972) suggests “hydrodynamic” dispersion is “the macroscopic outcome of the

actual movements of the individual tracer particles through the pores and various
physical and -chemical phenomena that take place within the pores.” This movement
164
can arise from a variety of causes. In this text, dispersion is the mixing of two
miscible fluids caused by diffusion, local velocity gradients (as between a pore wall
and a pore center), locally heterogeneous streamline lengths, and mechanical mixing
in pore bodies. Gravity tonguing and viscous fingering are two-dimensional effects
that we discuss in Chap. 6. Here we summarize experimental findings on dispersion
coefficients and some qualitative reasons for these observations.
For one-dimensional flow, the longitudinal dispersion coefficient Kl is given
by
β
Kl ⎛ | v | Dp ⎞
= C1 + C2 ⎜ ⎟ (5.5-19)
D0 ⎝ D0 ⎠
where C1, C2, and β are properties of the permeable medium and the flow regime. D0
is the effective binary molecular diffusion coefficient between the miscible
displacing and displaced fluids. Dp is an average particle diameter.
For very slow flows, the second term in Eq. (5.5-19) is negligible, and Kl is
proportional to D0. This case is analogous to a slow displacement in a wide channel
where mixing is due entirely to molecular diffusion. The constant C1 has een found to
be 1/φF, where F is the electrical formation resistivity factor (Pirson, 1983) to
account for the presence of the stationary phase.
For faster displacements, the second term in Eq. (5.5-19) becomes
significant. Deans (1963) has shown that well-stirred tanks in series give mixing
zones that can be described by dispersion coefficients proportional to velocity. Here,
mixing is the result of the highly irregular flow paths in the REV, which cause fluids
to mix completely as they are produced from each cell. Diffusion, of course, is
negligible if the fluids are well mixed.
An alternate, two-dimensional interpretation, including diffusion in this flow
regime, is the theory of Taylor (1953), whereby the flow channels are visualized as
having lateral dimensions much smaller than the longitudinal dimensions. For this
idealization, diffusion equalizes concentration gradients in the lateral direction giving
rise to an “effective” diffusion coefficient. Mixing is now the result of transverse
diffusion and variations in velocity caused by the no-slip condition at the pore wall.
Taylor’s theory predicts dispersion coefficients proportional to velocity squared.
Experimentally, it is found (Perkins and Johnston, 1963) that β = 1 to 1.25 in
Eq. (5.5-19); hence it seems the local mixing interpretation is closer to the mark than
Taylor’s theory.
This local mixing flow regime is where most EOR processes will occur. In
fact, if the interstitial velocity is greater than about 3 cm/day, the local mixing term in
Eq. (5.5-19) dominates the first term, and we can write
β
D ⎛ | v | Dp ⎞
K l = 0 + C2 ⎜ ⎟ D0 ≅ α l | v | (5.5-20)
φF ⎝ D0 ⎠
This does not imply that diffusion is categorically negligible in miscible flow.
Several phenomena involve flow around stagnant regions (for example, dead-end
pores, water blocked pores, or adjacent nonflowing zones) where diffusion rates are
165
important even in regimes that would otherwise be well described by Eq. (5.5-20). αl
in Eq. (5.5-20) is the longitudinal dispersivity of the permeable medium (Eq. 2.2-14),
a measure of the local heterogeneity scale. Bear (1970) classifies αl as one of the
fundamental properties of the medium. For the local mixing flow regime, αl is a more
fundamental measure of dispersion than Kl.
Figure 5-16 shows the three flow regimes from Perkins and Johnston (1963).
Similar data is in Bear (1970) and in several references of Perkins and Johnston.
The form of Eq. (5.5-20) is particularly convenient as the Peclet number
(Eq. 5.5-4), and the dimensionless concentration balance (Eq. 5.5-2) now become
independent of velocity
L
N Pe = (5.5-21)
αl
Therefore, the dimensionless mixing zone is directly related to αl through Eq.

(5.5-18). In fact, αl/L can be crudely regarded as the dimensionless mixing zone
length.
Suppose we try to design a laboratory displacement that has the same
dimensionless mixing zone length as a field prototype. Then we must have
⎛ αl ⎞ ⎛ αl ⎞
⎜ L ⎟ =⎜ L ⎟ (5.5-22)
⎝ ⎠field ⎝ ⎠lab
Figure 5-16 Longitudinal dispersion coefficients in permeable media flow (from

Perkins and Johnson, 1963)
166
Equation (5.5-22) clearly cannot be satisfied if the laboratory and field dispersivities
are assumed equal.
To enforce the equality in Eq. (5.5-22), we must have laboratory and field
values of αl. Laboratory-measured αl’s are available through correlations or
experiments. They are generally a few centimeters or less depending on the core
material. The estimated field-measured values of αl are far less certain. A good
summary of field-measured dispersivities is shown in Fig. 5-17. This figure shows
field-measured αl’s for several formation types plotted against the length scale it was
measured over. On the log–log scale, there is clearly considerable variation in αl at
Figure 5-17 Field and laboratory measured dispersivities (from Arya et al., 1988)
167
the same length, and even for the same formation, even though there is little or no
correlation with the latter.
Despite the scatter, there is a clear trend of increasing αl with measurement
distance. We can explain this increase qualitatively by saying the scale of
heterogeneity captured by a given measurement increases as the volume sampled
increases. Quantitatively, the phenomenon is the subject of active research (Gelhar et
al., 1979; Dagan, 1984) because of the complicated interplay between heterogeneity,
local dispersion coefficients, diffusion, and other permeable media properties that
combine to make αl length dependent.
Figure 5-17 points out the interesting and significant behavior of αl as system
macroscopic length increases. But even on a local scale, the behavior of longitudinal
dispersivity is not well known when multiple phases are flowing. Figure 5-18 gives
experimental data showing how the intraphase dispersivity changes as the phase
saturation changes. The data in this figure are for constant saturation flow of micellar
fluids for which the more general definition of Kl (Eq. 2.2-14) is appropriate. Figure
5-18 shows that aqueous phase dispersivity can increase by more than a factor of 10
as the aqueous phase saturation decreases. (This dispersivity increases as the
effective heterogeneity increases, but now the “heterogeneity” must be related to the
Figure 5-18 Dispersivities for constant saturation miscible flows (from MacAllister,
1982)
168
characteristics of the flowing fluids.) It is likely that wetting conditions play a large
role in the αl increase since no such changes in αl were observed for the nonwetting
phase in Fig. 5-18.
We summarize the most important points about the effects of dispersion on
one-dimensional miscible flow as follows:
1. Dispersion controls the rate of mixing of two fluids but does not affect wave
velocity.
2. Dispersive mixing zones can grow no faster than in proportion to the square
root of time.
3. The fluid velocity of most EOR processes is such that the flow is in the local
mixing flow regime where the dispersion coefficient is proportional to the
interstitial velocity. The proportionality constant is the longitudinal
dispersivity αl.
4. αl is a measure of the heterogeneity of the permeable medium and varies
with phase saturation and the measurement scale.
4. Neglecting dispersion in field-scale displacements is not proper because
dispersivity appears to increase with travel distance.
5-6 GENERALIZATION OF FRACTIONAL FLOW THEORY
In this section, we present the mathematical formalities to broaden the fractional flow
theories of Secs. 5-2 and 5-4 to multiple component, multiphase flow. As in those
sections, we neglect dissipative effects and restrict the equations to one-dimensional
flow. Our presentation is based on a subset of the method of characteristics (MOC)
solution technique known as simple wave theory, or coherence theory. (For more
careful mathematical detail, see Courant and Friedrichs, 1948; Helfferich and Klein,
1970; and Jeffrey and Taniuti, 1964.)
The fundamental principle in the MOC is to solve partial differential
equations (PDEs) by first converting them to a set of ordinary differential equations
(ODEs) that may then be integrated simultaneously. This set of ODEs can rarely be
integrated in closed form, but there is a large class of permeable media flow problems
for which the integrations will appear in a general form. To illustrate these ideas, we
consider first a single PDE and then pairs of PDEs in the dependent variables u and v.
The theory may be generalized to more than two PDEs, but in practice, the
procedures become cumbersome.
One Dependent Variable
Consider the following partial differential equation for u(x, t)
L(u ) = Aut + Bu x + E = 0 (5.6-1)

169
where A, B, and E are known functions of u, x, and t. The operator L(u) is linear in
the derivatives of u. The notation ux and ut means partial differentiation with respect
to x and t holding the other variable constant. We want solutions to Eq. (5.6-1) in the
form u(x, t) subject to the appropriate initial and boundary conditions. In the MOC,
we seek these solutions in the form u(s), where s is a parameter along a curve C in x-t
space such that x = x(s) and t = t(s). We may, therefore, write the total derivative of u
with respect to s as
us = ts ut + xs u x (5.6-2)
Equation (5.6-2) is a mixture of total derivatives us, ts, and xs and partial derivatives ut
and ux. However, we use the same notation for both types of derivatives since the
type of derivative should be clear from the usage. Comparing Eqs. (5.6-1) and (5.6-2)
leads to
ts = A (5.6-3a)
xs = B (5.6-3b)
us = − E (5.6-3c)
Equations (5.6-3), which imply the operator L(u) is a directed derivative along C, are
a set of three ODEs that may be integrated from an initial curve, as shown in Fig.
5-19(a), to give a characteristic curve C in xt space along which u varies as given by
the integration of Eq. (5.6-3c).
The various integrations are possible only if C is nowhere tangent to the
initial curve. Figure 5-19(a) schematically shows the integration of these equations
for a curve C that begins at the point (x0, t0) on the initial curve. We could take other
points on the initial curve and thereby cover the shaded domain of dependence in Fig.
5- 19(a) defined by the characteristics through the points A and B on ends of the
initial curve. If α is a parameter along the initial curve, the solution to Eqs. (5.6-3) is
t = t(s, α), x = x (s, α), and u = u(s, α). s and α are the coordinates of a natural,
generally curved, coordinate system for Eq. (5.6-1). Since α, in effect, determines
which curve C passes through the point (x, t), at which the value of u is desired, the
characteristics for Eq. (5.6-1) are a one-parameter (α) family of curves, and α is a
label for this one-parameter family. An important observation is that at every point
(x, t) in the shaded region in Fig. 5-19(a), the slope of the characteristic curve is given
by
xs dx B
= = = σ (u , x, t ) (5.6-4)
ts dt C A
σ is the characteristic direction at a given (x, t). Equation (5.6-4) implies it will
generally be unnecessary to determine t = t(s, α) and x = x(s, α) since t = t(x, α) will
follow directly from Eqs. (5.6-4) and (5.6-3c).
Consider now a special case of Eq. (5.6-1) where E = 0, and A and B are
functions of u only. The initial data are a curve that coincides with the x axis, where
170
Figure 5-19 Domains of dependence for one-variable hyperbolic equations
u = uI and then coincides with the t axis, where u = uJ. Thus the boundary (x = 0) and
initial (t = 0) data are uniform except for a step change at the origin where all values
of u between uI and uJ exist. It follows immediately from Eqs. (5.6-3c) and (5.6-4)
that u is constant along the characteristics C, which are themselves straight lines.
Figure 5-19(b) shows the characteristics for this case. In regions adjacent to the x and
t axes, the characteristics are parallel with slopes σ(uI) and σ(uJ), respectively. These
regions are constant-state regions since the dependent variable u is constant therein.
The shaded region in Fig. 5-19(b) is a fanlike region where σ changes continuously
between the limits imposed by the constant-state regions. Each ray emanating from
171
the origin carries a particular constant σ from the infinite numbers of u’s between uI
and uJ, and each has a slope σ evaluated at that u. Therefore, the shaded region in
Fig. 5-19(a) is a wave since, in any noncharacteristic direction, u is changing.
From Fig. 5-19(b), the characteristics cannot cross, but there is nothing that
requires σ to decrease monotonically, as in the case shown. When σ does not
decrease monotonically, a mathematically valid solution exists that leads to the
formation of shock waves, u being a physical variable.
Finally, the characteristic direction σ may clearly be interpreted as a velocity
(if x and t are distance and time) and written as
dx dx
σ= = (5.6-5)
dt C dt u
With the appropriate forms for A, B, t, and x, Eq. (5.6-5)becomes the

Buckley-Leverett equation (Eq. 5.2-10) for water displacing oil in a permeable
medium as we discussed in Sec. 5-2. Note the similarity between Figs. 5-5 and
5-19(b).
Two Dependent Variables
Let us consider now a pair of PDEs in the dependent variables u(x, t) and v(x, t)
L1 (u , v) = A1ut + B1u x + C1vt + D1vx + E1 = 0 (5.6-6a)

L2 (u , v) = A2ut + B2 u x + C2 vt + D2 vx + E2 = 0 (5.6-6b)
Initially, we consider the most general case of the coefficients A–E being functions of
x, t, u, and v. The first pair of terms in the linear operators L1 and L2 may be regarded
as directed derivatives of u and v. From the total derivative of du and dv, there are
four such directions (A1ut + B1ux, C1vt + D1vx, and so on) for each PDE. But to
transform the pair to a set of ODEs, we seek a curve in (x, t) space where u = u(s), v =
v(s), x = x(s), and t = t(s). We, therefore, seek a combination L = λ1L1 + λ2L2 so that L
is a linear function of total derivatives us and vs. As before, s is a parameter along
such a curve. For solutions to the equations, the operator L must be equal to zero,
hence
L = ( A1λ1 + A2 λ2 )ut + ( B1λ1 + B2 λ2 )u x + (C1λ1 + C2 λ2 )vt
(5.6-7)
+ ( D1λ1 + D2 λ2 )vx + ( E1λ1 + E2 λ2 ) = 0
For the directed derivatives of u and v to be colinear, it is necessary that
xs B λ + B2 λ2 D1λ1 + D2 λ2
=σ = 1 1 = (5.6-8)
tx A1λ1 + A2 λ2 C1λ1 + C2 λ2
be obtained from the total derivative for each dependent variable. The two equations
172
in Eq. (5.6-8) may be written as
( A1 xs − B1ts )λ1 + ( A2 xs − B2ts )λ2 = 0 (5.6-9a)
(C1 xs − D1ts )λ1 + (C2 xs − D2ts )λ2 = 0 (5.6-9b)
For nonzero λ1 and λ2, the determinant of the coefficient matrix must be zero; hence
( A1C2 − A2 C1 )σ 2 + ( A2 D1 − D2 A1 + C1 B2 − C2 B1 )σ
(5.6-10)
+ ( B1 D2 − D1 B2 ) = 0
where we have substituted the characteristic direction σ from Eq. (5.6-8).

Immediately, it is apparent that there are, in general, two characteristic directions, not
one as in the analogous expression (Eq. 5.6-4) for the one variable problem. Whether
or not these directions are real for all (x, t) depends on the form of the coefficients.
For permeable media flow problems, σ is real in at least some, and usually all, of the
domain (x, t). This, in fact, is the definition of hyperbolic PDEs. Further, the roots in
Eq. (5.6-10) are generally distinct. Let σ+ designate the larger root and σ– the smaller
of Eq. (5.6-10). Clearly, the corresponding characteristic curves, C+ and C–, cover the
domain of dependence in (x, t) since the slope of C+ is everywhere larger than the
slope of C–. Figure 5-20 shows these curves. The shaded domain of dependence is
bounded for the two dependent variable problem by the fast σ+ characteristic through
B and the slow σ– characteristic through A.
Each point in the domain of dependence is on the intersection of an σ+ and σ–
characteristic. The coordinates of a point may then be located as a distance s along a
particular characteristic having label α; that is, x = x(s, α), and t = t(s, α).
Alternatively, the coordinates may also be located by giving the labels of both
characteristics passing through it, or x = x(α, β) and t = t(α, β), where β is now the
label of the other characteristic. The notion of labels is somewhat confusing since α
and β can take on the same numerical values on the initial curve; however, in the
interior of the domain of influence, they are distinct.
The characteristic curves cannot be obtained, in general, unless it is known
how u and v change along the characteristic directions. This may be obtained by
replacing the coefficients of ux and vx in Eq. (5.6-7) by the numerators in Eq. (5.6-8)
( A1λ1 + A2 λ2 )us + (C1λ1 + C2 λ2 )vs + ( E1λ1 + E2 λ2 )ts = 0 (5.6-11a)
where we have rearranged with us = uxxs + utts, and so on. A similar procedure on ut
and vt gives
( B1λ1 + B2 λ2 )us + ( D1λ1 + D2 λ2 )vs + ( E1λ1 + E2 λ2 ) xs = 0 (5.6-11b)
These equations rearranged are
( A1us + C1vs + E1 xs )λ1 + ( A2us + C2 vs + E2 ts )λ2 = 0 (5.6-12a)
( B1us + D1vs + E1 xs )λ1 + ( B2us + D2 vs + E2 ts )λ2 = 0 (5.6-12b)
Again, for nontrivial λ1 and λ2, the determinant of the coefficient matrix must be
zero, and again, the characteristic equation has two real, distinct roots for us or vs.
173
Figure 5-20 Domains dependence for two-variable hyperbolic equations

174
The roots to these equations form, along with the two roots to Eq. (5.6-10), four
ODEs that may be integrated simultaneously for u, v, x, and t from an initial curve.
Rather than expound further on this, let’s consider the special case where E1
and E2 = 0, and the remaining coefficients are functions of u and v only. Such
equations are said to be reducible. The characteristic directions from Eq. (5.6-10) are
a known function of u and v only. Further, along each characteristic curve, there is a
relation between u and v given by
du
( B2 A1 − B1 A2 ) = ( B1C2 − C1 B2 ) + σ ± (C1 A2 − A1C2 ) (5.6-13)
dv
from the determinant of the singular matrix formed by Eqs. (5.6-9a) and (5.6-12a).
There are three equivalent forms to this equation, but the important point is that u and
v are related to each other along the C± curves in x-t space since all the coefficients in
Eq. (5.6-13) are known functions of u and v. The u-v plot that contains the above
solution is said to be the image, or hodograph, space and the function u = u(v) for C+
is the image curve Γ+ or C+, as is u = u(v) for Γ– the image curve for C–. In this text,
we call the u-v space the composition path diagram, and the Γ+ and Γ– curves the
composition paths.
Besides the restriction to reducible PDEs, let’s now consider the special case
of u = uI and v = vI being specified on the x axis, and u = uJ and v = vJ being specified
on the t axis. As before, this means that all (u, v) values between (u, v)I and (u, v)J
exist at the origin. Also as before, there are regions of constant state adjacent to both
axes where the characteristic directions, and hence, the labels α and β are constant.
But unlike before, there are now two fan-shaped regions (Fig. 5-20b) where first the
fast or a characteristics change slope, and then the β characteristics change. The
regions cannot overlap, or there would be finite regions where σ+ < σ–. This fact
causes the creation of a new constant-state region (u, v) between the fans that is, in
general, different from either (u, v)I or (u, v)J. Within the fan-shaped regions, the α
and β characteristics cannot both be straight, or else these would be constant-state
regions. But one of the characteristic directions must be straight in each region (σ+ in
the first shaded region in Fig. 5-20(b), and σ– in the second). This is so because two
points A and D on the boundary have the same (u, v) values since they can be
regarded as being in the constant-state region. This must be true of all other rays in
the fan-shaped region, for example, that passing through C and B. Otherwise, the ray
would be curved (from Eq. 5.6-10) and would ultimately intersect either of the
constant-state regions. It follows, then, that all points on the straight-line
characteristic carry the same (u, v) value. Since (u, v)A = (u, v)D and (u, v)B = (u, v)C,
it follows that σ B− = σ C− and the slope of the a characteristics is the same on all the σ+
characteristics. This means the u = u(v) relationship defined by Eq. (5.6-13) is the
same on any slow characteristic in the region. Thus (du / dv)σ − and (du / dv)σ +
uniquely determine the variations in u and v in the respective fan-shaped regions. The
function u = u(v) is always calculated based on the curved characteristic.
175
The above concepts apply generally to reducible sets of PDEs in any number
of N dependent variables. Stated concisely, the observations are
1. Adjacent to any constant-state region, there is a region having at least one

straight line characteristic. The second region is a simple wave region.
2. Within a simple wave region, the dependent variables are related to each
other through a set of ODEs.
3. For boundary and initial conditions that are uniform except for a step change
at the origin, the entire (x, t) domain consists of alternating constant-state and
simple wave regions. The simple waves in this case are centered simple
waves.
Coherence
The information on reducible equations may be restated with more physical insight
by referring to simple waves in the terminology of coherent waves (Helfferich and
Klein, 1970). Since (u, v) is constant on a straight-line characteristic in a simple or
coherent wave region, and since σ is a function of u and v only, it follows that
dx dx
= (5.6-14a)
dt u dt v
or for u1, . . . , uN dependent variables

dx dx dx
= , ..., = (5.6-14b)
dt u1 dt u2 dt uN
Equation (5.6-14b) states that the velocity of constant values of the dependent
variables is the same––the coherence condition. As we illustrate in Sec. 5-7, the
coherence method of calculating simple waves is more direct than using MOC.
Equation (5.6-14b) implies, further, that there can be no more than N waves.
5-7 APPLICATION TO THREE-PHASE FLOW
In this section, we apply the results of the coherence theory by calculating the
displacement efficiencies for a three- phase water (i = 1), oil (i = 2), gas (i = 3) flow
problem. We assume away dissipative effects––capillary pressure and pressure-
dependent fluid properties––and restrict the fluids to be single pseudocomponent
phases. The assumption of an incompressible gas phase is, of course, realistic only if
ΔP
c3 ΔP ≅ (5.7-1)
P
is small. This condition is not met in general although for flows in high permeability
media c3ΔP can be fairly small, particularly considering that gas viscosity is also
176
small. But even if c3ΔP is large, we have seen from Sec. 5-3 that fluid
compressibility causes waves to spread and does not affect wave velocity.
Subject to the above restrictions, the species conservation Eq. (5.4-1)
becomes
∂S j ∂f j
+ = 0, j = 1 or 2 (5.7-2)
∂t D ∂xD
in dimensionless form, where for a horizontal reservoir
⎛ λrj ⎞
fj = ⎜ 3 ⎟ (5.7-3)
⎜ ∑ λrm ⎟
⎜ ⎟
⎝ m =1 ⎠
The relative mobilities in Eq. (5.7-3) are known functions of S1 and S2. Only two
independent saturations are in this example, since S1 + S2 + S3 = 1, which we
arbitrarily take to be the water and oil saturations. Equation (5.7-3) implies that the
fractional flows are known functions of S1 and S2.
From Eq. (5.4-5), the specific velocity of a constant saturation Sj is
⎛ ∂f j ⎞
vS j = ⎜ , j = 1 or 2
⎜ ∂S ⎟⎟
(5.7-4a)
⎝ j ⎠tD
if the wave is nonsharpening and
Δf j
vΔS j = , j = 1 or 2 (5.7-4b)
ΔS j
if the wave is a shock. We cannot take the derivative in Eq. (5.7-4a) without knowing
the solution to the problem Sj(xD, tD). The results of the previous section carry over to
this problem with A1 = 1, B1 = f11, D1 = f12, B2 = f2l, C2 = 1, D2 = f22, A2 = C1 = 0, E1 =
E2 = 0. For brevity, we have adopted the convention that f12 = (∂f1/∂S2)S1, and so on.
B1, B2, C1, and C2 are known functions of S1 and S2––though perhaps very
complicated––but we can calculate them without knowing the solution S1(xD, tD) and
S2(xD, tD).
Let’s now let the initial saturations in the medium be uniform at (S1, S2)I and
impose at xD = 0 the saturations (S1, S2)J. From Sec. 5-6, we know the coherence
condition applies at all points in the domain where
df1 df 2
= =σ (5.7-5)
dS1 dS 2
from Eqs. (5.6-14b) and (5.7-4a). The derivatives in Eq. (5.7-5) are total derivatives
since the coherence condition implies the existence of a relation S2 = S2(S1) in
saturation space. We expand the derivatives in Eq. (5.7-5) and write the two
equations in matrix form as
⎛ f11 f12 ⎞⎛ dS1 ⎞ ⎛ dS1 ⎞
⎜ ⎟⎜ ⎟ =σ ⎜ ⎟ (5.7-6)
⎝ f 21 f 22 ⎠⎝ dS2 ⎠ ⎝ dS2 ⎠
177
To solve for S2(S1), we first solve this equation for the eigenvalues, σ±
1
σ ± = {( f 22 + f11 ) ± [{( f 22 + f11 ) 2 + 4 f 21 f12 ]1/ 2 } (5.7-7)
2
Both roots to Eq. (5.7-7) are real, σ+ > σ–, and both are known functions of S1 and S2.
Recall that the σ± are saturation velocities. Solving for dS1 and dS2 in Eq. (5.7-6)
gives
dS1 σ ± − f11
= (5.7-8)
dS2 f12
Equations (5.7-7) and (5.7-8)are the special cases of Eqs. (5.6-10) and (5.6-13).
Equation (5.7-8) is an ordinary differential equation whose integration gives the
function S2(S1). There are two such functions corresponding to σ+ and σ–. The
velocity of any saturation along S2(S1) is given by σ+ and σ– depending on whichever
is physically realistic.
The above procedure could perhaps be made clearer by addressing a
particular problem. Consider an oil–gas–water mixture being displaced by water. To
make the problem simple, we take the relative permeabilities to be
S j − S jr
krj = , j = 1 or 2 (5.7-9)
1 − S1r − S2 r − S3r
and let S1r = S2r = S3r = 0.1. Equation (5.7-9) is not a realistic three-phase relative
permeability function (see Exercise 5N), but it is sufficient for illustration. We
further take μ1 = 1 mPa-s, μ2 = 5 mPa-s, and μ3 = 0.01 mPa-s, and consider the initial
conditions to be S2I = 0.45, and S1I = 0.1. Therefore, the medium is initially at
residual water saturation with equal volumes of oil and gas. We are to displace this
mixture with water, that is S1J = 0.8 and S2J = 0.1. This procedure corresponds to a
waterflood initiated well into the primary production phase.
Figure 5-21 shows the functions S2(S1) obtained by numerically integrating
Eq. (5.7-8) with the indicated physical relations. The plot is on a triangular diagram
to emphasize the relation S1 + S2 + S3 = 1. The integration of Eq. (5.7-8) for various
initial values of S1 and S2 produces two families of curves corresponding to σ+ and
σ–, which are the image curves σ+ and σ– (light lines in Fig. 5-21) referred to
previously. Since σ+ > σ–, the image curves nowhere coincide, and further, to every
point in the saturation diagram, there are associated two velocities σ+ and σ–. The two
families of curves we call the saturation paths after Helfferich (1981). The particular
paths that pass from the initial to the injected condition are the saturation routes (bold
lines in Fig. 5-21). Though we henceforth restrict our attention to the saturation
routes, Fig. 5-21 gives a rapid visual perspective for any displacement having
arbitrary initial and injected conditions.
In moving from the initial to injected conditions, there are two alternative
saturation routes: (1) a σ– segment going from the initial conditions to the upper apex
of the three-phase flow region and then a σ+ segment on the gas–water boundary to
the injected condition and (2) a σ+ segment from the initial conditions to (S1, S2) =
178
Figure 5-21 Three-phase flow saturation paths
(0.36, 0.54) followed by a σ– segment along the oil–water boundary to the injected
conditions. Both routes are mathematically valid solutions to the problem; in fact, an
infinite number of mathematical solutions correspond to a route that arbitrarily
switches from al to a paths in going from (S1, S2)I to (S1, S2)J. From the Buckley-
Leverett problem in Sec. (5-2), we know that saturation velocities must decrease
monotonically (though not continuously) in the upstream direction. The only physical
solution for the problem is route (2) because σ+ > σ– forces this to be the only
possible route where σ decreases monotonically from (Si, S2)I to (S1, S2)I to (S1, S2)J.
Within a route segment, the saturation velocities must decrease monotonic-
ally in the upstream direction also. This condition is not met on the σ+ route segment
(the arrows on the saturation routes indicate the direction of increasing saturation
velocity). Such behavior indicates the wave is a shock, and we can find the shock
velocity by a procedure entirely analogous to that used in Sec. (5-2). Figure 5-22(a)
179
Figure 5-22 Diagrams for three-phase flow example
plots the oil and water fluxes (f1, f2) versus (S1, S2) along the composition route. The
shock construction is exactly as suggested in Fig. 5-4, and may be performed on
either the f1-S1 curve or the f2-S2 curve. Equation (5.7-5) guarantees this equivalence.
The only real difference between the three-phase and two-phase flow problems at this
point is the existence of the constant-state region at IJ. The time-distance diagram for
the displacement is in Fig. 5-22(b), which should be compared to Figs. 5-5 and
5-20(b).
180
Despite the simplified nature of the relative permeability curves used in this
example, Fig. 5-22 illustrates that the most important feature of three-phase oil–gas–
water flow is the extremely small gas viscosity. This viscosity causes the oil
fractional flow to be small initially and to delay the appearance of an appreciable
amount of oil at the outflow end until tD = 0.28. This delay, or “fill-up,” time is an
omnipresent feature of waterfloods begun with appreciable amounts of free gas in the
medium (Caudle, 1968). A fill-up period occurs because of the very large gas
mobility, not as the result of gas compressibility or redissolution. The last two effects
would serve to reduce the fill- up time. A second consequence of the small gas
viscosity is no simultaneous three-phase flow occurs in the medium. In fact, by
assuming an oil–water mixture banks up the free gas, it is possible to repeat the
results in Figs. 5-21 and 5-22 with much less effort (see Exercise 5O). A final
consequence of the small gas viscosity is this behavior is qualitatively accurate
regardless of the relative permeability functions used.
We end this section by discussing the displacement efficiency of the three-
phase flow problem. There is now a displacement efficiency for both oil and gas for
which we need average saturations for the definition (Eq. 5.1-2). Considering the
fractional flux-saturation curve in Fig. 5-22(a), the average saturations follow from a
procedure directly analogous to the Welge procedure in Sec. 5-2.
S j = S j |xD =1 −t D ( f j |xD =1 − f jJ ), j = 1, 2, or 3 (5.7-10)
where tD = (dfj/dSj)–1 is the reciprocal slope of the fj-Sj curve evaluated at xD = 1.

Figure 5-22(a) shows the average water saturation at water breakthrough, and Fig.
5-23 shows the displacement efficiencies for this example. Once again ED is limited
by the residual phase saturations, oil production is delayed for a fill-up period, and
Figure 5-23 Displacement efficiencies for three-phase flow problem

181
the oil displacement efficiency is determined by the water–oil relative permeabilities
and viscosities.
This example demonstrates the strength of simple wave theory. In later
chapters, we return to these procedures for specific EOR applications.
Any oil recovery calculation of a field-scale displacement based solely on the

procedures discussed in this chapter will seriously overestimate the actual recovery:
Such one-dimensional calculations neglect volumetric sweep issues which are at least
as important as displacement efficiency. Nevertheless, the fractional flow
calculations are important in establishing a framework for advancing our study. The
items important in establishing this framework are the Buckley-Leverett theory and
its generalization in Sec. 5-7, the ideas of coherent waves and their representations,
and the notion of the ideal miscible displacement.
EXERCISES
5A. Parameter-Free Statement. Show that Eqs. (5.2-5) can be reduced to a parameter-free
statement by defining and introducing a reduced saturation SD, where
S1 − S1I
SD = (5A-1)
S1J − S1I
5B. Radial Form of Water Material Balance

(a) Show that the one-dimensional water conservation Eq. (5.2-1) for incompressible
flow in radial geometry is
∂S q ⎛ ∂f1 ⎞
φ 1+ =0 (5B-1)
∂t 2π H t r ⎜⎝ ∂r ⎟⎠
where q is the volumetric flow rate, Ht the medium thickness, and f1 is the same as
Eq. (5.2-2).
(b) If we let rD = (r/R)2 and tD = ∫0t qdt / πφ H t R 2 = ∫0t qdt / V p , show that Eq. (5B-1)
becomes identical to the linear Eq. (5.2-5a).
5C. Buckley-Leverett Application. Calculate effluent histories (water cut f1 |xD =1 versus t)
for water (μl = 1 mPa-s) displacing oil given the following experimental data (Chang
et. al., 1978):
S1 kr1 kr2
0.40 0.00 0.36
0.45 0.005 0.26
0.50 0.009 0.14
0.55 0.02 0.08
0.60 0.035 0.036
0.65 0.050 0.020
0.70 0.080 0.00
182
Use three values of oil viscosity: μ2 = 1, 5, and 50 mPa-s. For μ2 = 5 mPa-s, calculate
the endpoint, shock, and average saturation mobility ratios. The dip angle is zero.
5D. Gravity and Fractional Flow Theory. For the exponential relative permeability
functions of Eq. (3.3-4), plot water saturation profiles at tD = 0.3 for dip angles of α =
0°, 30°, and –30°. Additional data are S1r = S2r = 0.2, n1 = 1, n2 = 2, kr01 = 0.1, kr02 =
0.8, μ1 = 1 mPa-s, μ2 = 10 mPa-s, k = 0.5 μm2, Δρ = 0.2 g/cm3, and μ = 0.6 cm/day.
5E. Buckley-Leverett Theory with Straight Line Relative Permeabilities. Use straight line
exponential relative permeability functions with zero residual phase saturations in the
following (n1 = n2 = 1, Slr = S2r = 0 in the exponential relative permeability functions).
Also take f1I = 0 and f1J = 1.
(a) Show that the sign of (1 – M0 + M0 N g0 sin α) uniquely determines the character
(spreading, indifferent, sharpening) of the water saturation wave.
(b) For the spreading wave case––(1 – M0 + M0 N g0 sin α) < 0––Eq. (5.2-10) may be
inverted explicitly for S1(xD, tD). Derive this expression in terms of the quadratic
formula.
(c) Use the equation in part (b) to show that for α = 0 the water saturation function is
given by
⎧ xD
⎪0, > M0
tD
⎪
⎪ ⎛ t M 0 ⎞1/ 2
⎪ D −1
S1 ( xD , tD ) = ⎨ ⎜⎝ xD ⎟⎠ 1 x (5E-1)
⎪ , ≤ D ≤ M0
⎪ M 0 −1 M 0
tD
⎪ xD 1
⎪1, < 0
⎪⎩ tD M
(d) Use Eq. (5E-1) to derive an expression for the average water saturation S (t D ) and
the displacement efficiency ED(tD).
5F. Water Fractional Flow with Capillary Pressure. Derive the expression for water
fractional flow including capillary pressure (Eq. 5.3-1).
5G. Analytic Relative Permeability Ratios (Ershaghi and Omoregie, 1978). Over
intermediate water saturation ranges, the oil–water relative permeability ratio plots
approximately as a straight line on a semilog scale, using
kr 2
= Ae − BS1 (5G-1)
kr1
where A and B are positive constants. Using the Buckley-Leverett theory, show that a
plot of the product of oil and water cuts is a straight line with slope 1/B when plotted
against 1/tD. The dip angle is zero.
5H. Fractional Flow with Two Inflections. For the fractional flow curve of Fig. 5H,
construct plots of fractional flow versus dimensionless distance at breakthrough for
saturation S1 = 1 displacing S1 = 0, and S1 = 0 displacing S1 = 1.
183
Figure 5H Fractional flow curve for

Exercise 5H
5I. The Reversibility of Dispersion and Fractional Flow. Fluid 2 is to be partially

displaced by fluid 1 in a one-dimensional permeable medium. Fluid 1 is injected until
just before it is produced, and then the flow is reversed (that is, fluid 2 is injected at the
effluent end). In all that follows, take the initial (I) condition to be 100% fluid 2
flowing and the injected (J) condition to be 100% fluid 1.
(a) Sketch two time–distance diagrams for this case using fractional flow curves like
those on the extreme right and left of Fig. 5-6.
(b) If fluids 1 and 2 are completely miscible with identical viscosities and mix only by
dispersion, use Eq. (5.5-18) to sketch the time–distance diagram.
(c) Based on the results of parts (a) and (b), what can you conclude about the mixing
caused by fractional flow compared to that caused by dispersion?
(d) If fluids 1 and 2 are water and oil and a fractional flow curve like that on the
middle panel of Fig. 5-6 applies, calculate and plot the time–distance diagram.
5J. Mobility Ratio for Compressible Flow. Consider the pistonlike displacement of fluid 2
by fluid 1 in the x direction. Use the general definition of mobility ratio (pressure
gradient ahead of front divided by pressure gradient behind front) in the following:
(a) Show that the mobility ratio becomes the endpoint mobility ratio if the volumetric
flowrate uA is not a function of x (fluids are incompressible).
(b) If the mass flux ρuA is not a function of x, on the other hand, show that the
mobility ratio becomes
ko v
M v = ro1 2 (5J-1)
kr 2 v1
where v = μ/ρ is the kinematic viscosity.
(c) Calculate both M0 and Mv for the following conditions: ρI = mg/cm3, μ1 = 1 μPa-s,
ρ2 = 0.8 g/cm3, ρ2 = 2 mPa-s, kro1 = 0.1 and kro2 = 1.0.
184
5K. Using Tracer Data. Consider a one-dimensional permeable medium containing oil at a
uniform residual saturation S2r, and through which is flowing 100% water at a constant
rate. At t = 0, a second water stream is introduced at the inlet that contains two ideal
(nondispersing and nonabsorbing) tracers. Tracer 1 remains only in the water phase,
but tracer 2 partitions into the residual oil phase with a partition coefficient of 2. The
partition coefficient is the ratio of the concentration of tracer 2 in the oil phase to that
in the water phase K 212 = C22/C21. Tracer 1 breaks through after three hours, and tracer
2 after six hours. If the volumetric injection rate is 1 cm3/min, calculate the pore
volume and S2r.
5L. Laboratory Estimation of Dispersivity. Dispersivity may be estimated from laboratory,
first-contact miscible displacements with the following development:
(a) Show from Eq. (5.5-15) that a plot of (1 – tD)/ tD versus erf–1(1 – 2Ce) will
yield a straight line with slope 2N Pe−1/ 2 . Here Ce is the effluent concentration
(CD |xD =1 )
(b) Estimate the pore volume, dispersion coefficient, and dispersivity from the
following experimental data:
Volume Effluent
produced (cm3) concentration
60 0.010
65 0.015
70 0.037
80 0.066
90 0.300
100 0.502
110 0.685
120 0.820
130 0.906
140 0.988
150 0.997
The interstitial velocity is 20 cm/day, and the length is 0.5m. Note that
erf–1(1 – 2x) is the probability axis (x axis) on probability paper.
5M. Tracers in Two-Phase Flow. Consider a permeable medium flowing oil and water at
constant oil fractional flow (case A in Fig. 5-12). Show that if a tracer with partition
coefficient defined as in Exercise 5K is introduced at tD = 0, the conservation equation
for the tracer concentration C in the aqueous phase is (Delshad, 1981)
∂C ∂C K ∂ 2C
+ − =0 (5M-1)
∂t D ∂xD vT L ∂xD2
where
vT
tD = t (5M-2)
L
q f1 + K 21 f 2
vT = (5M-3)
Aφ S1 + K 21 S 2
185
S K + K 21 S 2 K l 2
K = 1 l1 (5M-4)
S1 + K 21 S 2
Kl1 and Kl2 are the longitudinal dispersion coefficients for the tracer in the oil and water
phases. Take (q/Aφ) to be constant.
5N. Three-Phase Coherence Calculation. A more realistic three-phase relative permeability
for oil, gas, and water is
n1
⎛ S1 − S1r ⎞
kr1 = k r01 ⎜ ⎟ (5N-1)
⎝ 1 − S1r − S 2 r ⎠
n3
⎛ 1 − S1 − S2 − S3r ⎞
kr 3 = k ⎜ 0
r3 ⎟ (5N-2)
⎝ 1 − S1r − S3r ⎠
⎛k ⎞⎛ k ⎞
kr 2 = kr02 {⎜ r021 + kr1 ⎟ ⎜ r023 + kr 3 ⎟ − (kr1 + kr 3 )} (5N-3)
⎝ kr 2 ⎠ ⎝ kr 2 ⎠
where
n21
⎛ 1 − S1 − S2 r1 ⎞
kr 21 =k ⎜
0
r2 ⎟ (5N-4)
⎝ 1 − S 2 r1 − S1r ⎠
n23
⎛ S + S1 − ( S2 r 3 + S1r ) ⎞
kr 23 = kr02 ⎜ 2 ⎟ (5N-5)
⎝ 1 − ( S 2 r 3 + S1r ) − S3r ⎠
These are modifications of the Stone relative permeability model (1970).

In Eqs. (5N-1) through (5N-5)
n2l = Oil relative permeability exponent in water–oil system

n23 = Oil relative permeability exponent in gas–oil system
S2rl = Residual oil saturation in water–oil system
S2r3 = Residual oil saturation in gas–oil system
Calculate and plot the following:

(a) Lines of constant kr1, kr2, kr3, in the triangular composition space, S1, S2, and S3.
(b) The composition paths and a waterflood composition route for initial saturations of
0.5, 0.3, and 0.2 for oil, gas, and water.
(c) The wave positions in a dimensionless time–distance diagram.
Use the following data:
μ1 = 1 mPa-s μ2 = 2 mPa-s μ3 = 0.01 mPa-s

S2rl = 0.3 kr02 = 0.6 n2l = 1.5
S2r3 = 0.05 kr01 = 0.3 n23 =2
S1r = 0.2 kr03 = 0.7 n1 =3
S3r = 0.05 α =0 n3 = 2.5
This problem requires a numerical solution.

5O. Simplified Three-Phase Fractional Flow. Rework part (c) of Exercise 5N by assuming
the displacement becomes a shock wave from the initial conditions to a region of
186
simultaneous two-phase oil-water flow followed by a wave of undetermined character
to the injected conditions. The velocity of the first wave is given by
f3I f − f1+ f 2 I − f 2+
vΔS1 = = 1I = (5O-1)
S3 I − S3r S1I − S1+ S2 I − S2+
where f1+ and f1+ 1are the water fractional flow and saturation behind the shock. The
velocity of the second wave is then given by the Buckley-Leverett construction. Plot an
effluent history of oil and water cuts to demonstrate the fill-up phenomenon.
5P. Method of Characteristics for Reducible Equations. Consider the following pair of
partial differential equations for u(x, t) and v(x, t)
∂u ∂ (u 2 v)
+ =0 (5P-1)
∂t ∂x
∂u ∂v 2
+ =0 (5P-2)
∂t ∂x
where both u and v are less than or equal to 1.

(a) Write these equations in the “canonical” form of Eqs. (5.6-6). Are these reducible?
Why or why not?
(b) Write the coherence requirement for Eqs. (5P-1) and (5P-2). Use this to develop an
expression for σ, the composition velocity along the characteristic directions.
(c) Use σ to develop an expression for u = u(v) along both characteristic directions.
(d) If the boundary data are specified along a line u = 1 plot the “composition” path
grid (u, v space) for u < 1 and v < 1.
(e) On the plot of part (d) show the “composition” route for (u, v)J = (0.6, 0.2)
displacing (u, v)I = (1, 1). Treat u and v as physical variables so that the composi-
tion velocity must decrease monotonically from I to J. Plot the time (t) – distance
(x) diagram for this “displacement” where t > 0 and 1 > x > 0.
(f) Based on this problem and what you know about the ideal miscible displacement,
discuss why the constructions in Fig. 5-12 can be done without the procedures in
parts (a) through (e).
5Q. Gravity Segregation and Fractional Flow. Consider the homogeneous,
one-dimensional permeable medium shown in Fig. 5Q for which all the fractional flow
assumptions apply. Both ends of the medium are sealed. For t < 0, the medium contains
a completely saturated water zone above a saturated oil zone (0 < ε < 1). At t = 0, the
more dense water is allowed to flow downward while the less dense oil flows upward.
This results in a complete reversal of the oil and water zones after a sufficiently long
time. Figure 5Q also shows the long-time condition of the medium.
(a) Show that there is no bulk flow (u = 0) at any point in the medium.
(b) Derive a water conservation equation for this special case from the general
equations in Chap. 2. Give also the boundary conditions needed to solve this
equation for S1(x, t).
(c) Make the equation of part (b) dimensionless by introducing appropriate scaling
factors.
(d) Derive a dimensionless water flux (analogous to a fractional flow) by eliminating
the water pressure gradient from the equation of part (c). The absence of bulk flow
does not eliminate pressure gradients (Martin, 1958).
187
Figure 5Q Gravity segregation with fractional flow
(e) For the following values, plot the dimensionless water flux of part (d) versus water
saturation.
kr1 = 0.1 S14 kr2 = .8(1 – S1)2

μ1 = 1 mPa-s μ2 = 5 mPa-s
(f) Based on the curve of part (e) and ε = 0.6, construct the time–distance diagram
showing the progress to complete gravity segregation of the water and oil zones.
Estimate the dimensionless time this occurs at.
5R. An Alternate Derivation of the Characteristic Equations. Consider the following
reducible equations for u(x, t) and v(x, t):
A1ut + B1u x + C1vt + D1vx = 0

(5R-1)
A2 ut + B2 u x + C2 vt + D2 vx = 0
(a) Suppose that u and v are functions of the combined variable σ = x/t. Show that Eq.
(5R-1) can be written as
⎛ B1 − σ A1 D1 − σ C1 ⎞⎛ u ′ ⎞
⎜ ⎟⎜ ⎟ = 0 (5R-2)
⎝ B2 − σ A2 D2 − σ C2 ⎠ ⎝ v′ ⎠
where u′ = du/dσ, and so on.
(b) For a nontrivial solution, the determinant of the coefficient matrix of Eq. (5R-2)
must be zero. Show that this gives the characteristic directions given by Eq.
(5.6-10).
(c) Again, for a nontrivial solution, the determinant of the augmented matrix (matrix
with the solution vector substituted for a column) must also be zero. Show that if
we replace the second column, this operation yields the following relation between
u and v:
du B1 − σ ± A1
= (5R-3)
d v B2 − σ ± A2
Solutions that can be expressed in terms of (x/t) are said to be se4f-similar.

188
Volumetric Sweep
Efficiency
Typical values of residual oil and connate water saturations indicate ultimate
displacement efficiency should normally be between 50% and 80% of the contacted
oil in a waterflood. This range is substantially higher than the 30% average recovery
efficiency observed in waterfloods; it is also higher than recovery efficiency in most
EOR projects (see Sec. 1-4). Of course, the reason displacement efficiency is higher
than the recovery efficiency is that not all the oil is contacted by the displacing agent.
This effect is present in the oil recovery Eq. (2.5-5) where the displacement
efficiency is multiplied by the volumetric sweep efficiency EV. Based on these
approximate figures, the volumetric sweep efficiency is between 40% and 60% for a
waterflood. For many EOR processes, it can be much smaller, and for others,
effecting a large EV is a primary design objective.
In this chapter, we provide both an overview of volumetric sweep efficiency and
techniques to combine areal, vertical, and displacement sweep to arrive at a recovery
efficiency. We deal almost exclusively with the immiscible water–oil displacement
since this literature on recovery efficiency is well established and many of the more
important features also carry over to EOR. In later chapters, we discuss the
volumetric sweep efficiency of specific EOR processes. To further distinguish
between volumetric and displacement sweep efficiency, we usually deal with
indifferent or self-sharpening displacements in which dispersive effects are small. For
these cases, the calculation techniques are equally valid whether the displacement is
miscible or immiscible since there is no simultaneous flow of components.
189
6-1 DEFINITIONS
Based on the overall material balance of Sec. 2-5, the cumulative mass of oil
recovered is
N p 2 = VbW2 I ER 2
from Eq. (2.5-3) with no oil injection. We wish to convert this equation to a more
standard form by the following transformations: Eliminate the recovery efficiency
ER2 through Eq. (2.5-5a), and replace W2 I with φ(ρ2S2ω22)I, which assumes oil is in
only the liquid oleic phase. This gives
N p 2 = Vbφ ( ρ 2 S2ω 2 ) I ED EV
Next, eliminate (ρ2ω22)I with the oil formation volume factor definition Eq. (2D-5),
and let Vbφ = Vp, the pore volume, and Np = N p 2 / ρ 2o , the oil production in standard
volumes. These substitutions yield
ED EV S2 I V p
Np = (6.1-1)
B2 I
In Eq. (6.1-1), ED is the displacement sweep efficiency defined in Eq. (5.1-1),

and EV is the volumetric sweep efficiency defined as
Volumes of oil contacted by displacing agent
EV = (6.1-2)
Volumes of oil originally in place
S2 IV p
The term ( ) represents the oil in place at the start of the displacement in
B2 I
standard volumes. We have also dropped the subscript i = 2 because all efficiencies
in this chapter refer to oil recovery.
The volumetric sweep efficiency can be decomposed into the product of an
areal sweep efficiency and a vertical sweep efficiency
EV = E A EI (6.1-3)
The definition of the areal sweep efficiency is

Area contacted by displacing agent
EA = (6.1-4)
Total area
Figure 6-1(a) shows a schematic of a highly idealized pistonlike displacement in a

four-layer areally homogeneous reservoir. Figure 6-1(b) is an areal view of Figure
6-1(a). Based on the definition of Eq. (6.1-4), EA is the doubly cross-hatched area (at
t2) divided by the singly cross-hatched area. The vertical sweep efficiency,
Cross-sectional area contacted by displacing agent
EI = (6.1-5)
Total cross-sectional area
is also similarly defined in Fig. 6- 1 (a) at a particular time.

190
Figure 6-1 Sweep efficiency schematic
The definitions of Eqs. (6.1-3) through (6.1-5) have several subtle

difficultties. Both areal and vertical sweep efficiency are ratios of areas; therefore,
their product EV must be a ratio of areas squared. This observation contradicts the
definition of Eq. (6.1-2) which says EV must be a ratio of lengths cubed. The
redundant dimension in either Eq. (6.1-4) or (6.1-5) is the dimension parallel to the
displacement direction. This direction is nonlinear and varies with both position and
time. Thus the decomposition Eq. (6.1-3) transforms EV into a product of two plane
flows.
A second consequence of the redundant dimension in EV is that both EA and
EI depend on each other. Note from Fig. 6-1 that EA depends on vertical position.
Similarly, though not so obviously, EI will be different from the cross section shown
for each cross section between the injector and producer. If we restrict ourselves to
cross sections defined by pathlines between the injector and producer (dotted lines in
Fig. 6-1b), EI will be the same for each cross section if it can be expressed in a
dimensionless form independent of rate. But for the general case, EI is a function of
rate and will be different for each cross section. As we see in Sec. 6-7, the practical
consequence of this observation is that neither the areal nor the vertical sweep
191
efficiency in Eq. (6.1-3) can be evaluated at the same time for which the volumetric
sweep efficiency is desired.
To use Eq. (6.1-1), even with the above complications, we must have
independent estimates of EA and EI. For certain very special cases––confined
displacements in areally homogeneous regular patterns with no or very good vertical
communication––these are available to us through correlation (see Sec. 6-2) or
calculation (see Secs. 6-3 through 6-5). When these conditions are not met, EV must
be estimated through scaled laboratory experiments or numerical simulation. In the
latter case, though certainly possible to obtain sweep efficiency estimates, the oil
recovery itself may be obtained directly, and Eq. (6.1-1) is unnecessary. Still, the
equation provides a better understanding of sweep efficiency concepts and the factors
necessary to maximize EV than does simulation alone.
6-2 AREAL SWEEP EFFICIENCY
Though areal sweep efficiency may be determined through simulation or by

analytical methods (Morel-Seytoux, 1966), the most common source of areal sweep
efficiency data is from displacements in scaled physical models. Figures 6-2 through
Figure 6-2 Areal sweep efficiency for a confined five-spot pattern (from Dyes et
al., 1954)
192
Figure 6-3 Areal sweep efficiency for a confined direct line drive pattern,
d/a = 1 (from Dyes et al., 1954)
6-4 present three of these areal sweep “correlations” from the work of Dyes et al.
(1954) for three different regular well patterns. Several more of these correlations are
in the work of Craig (1971), and an extensive bibliography of areal sweep efficiency
is given by Claridge (1972). For the patterns shown in the lower right-hand corner,
Figs. 6-2 through 6-4 plot EA on the y axis versus the reciprocal mobility ratio on the
x axis with time as a parameter. Since mobility ratio and pattern type are fixed for a
given displacement, time is actually the dependent variable. The dimensionless time
in Figs. 6-2 through 6-4 is cumulative displaceable pore volumes of displacing agent
injected. Since time is the dependent variable in these correlations, a more direct
representation would be a plot of EA versus dimensionless time at fixed mobility ratio
and pattern type (see Exercise 6A). You should remember that these correlations are
for pistonlike displacements in regular, homogeneous, confined patterns. When the
well patterns are unconfined, the reference area in Eq. (6.1-4) can be much larger,
and EA smaller. Based on an extensive survey of the available correlations for
spreading displacements, Craig (1971) determined that the appropriate mobility ratio
for the areal sweep correlations is the average saturation mobility ratio M given by
Eq. (5.2-25a).
From the correlations, EA increases with increasing time, or throughput, and
decreasing mobility ratio. At a fixed mobility ratio, EA is equal to the displaceable
193
Figure 6-4 Areal sweep efficiency for a staggered line drive pattern, d/a = 1
(from Dyes et al., 1954)
pore volumes injected until breakthrough and then given by the indicated curves in
Figs. 6-2 through 6-4 thereafter. EA also increases as the pattern type more closely
approaches linear flow, but this sensitivity is not great for the more common patterns.
The decrease in EA with increasing M is in the same direction as the change in
displacement efficiency with mobility ratio discussed in Sec. 5-2; thus large mobility
ratios are detrimental to both areal and displacement sweep.
6-3 MEASURES OF HETEROGENEITY
Considering the manner reservoirs are deposited in and the complex diagenetic
changes that occur thereafter, it should not be surprising that no reservoir is homo-
geneous. This does not imply all reservoirs are dominated by their heterogeneity
since in many cases, one mechanism is so strong that it completely overshadows all
others. For example, gravity can be so pronounced in a high-permeability reservoir
that it may be considered homogeneous to good approximation.
Nevertheless, heterogeneity is always present in reservoirs, is the most
difficult feature to define, and usually has the largest effect on vertical sweep
efficiency. Therefore, before we explore vertical sweep efficiency, we discuss the
most common measures of heterogeneity and their limitations.
194
Definitions
The three principal forms of nonidealities in reservoirs are anisotropies,

nonuniformities, and heterogeneities. These terms can be applied to any property but
usually describe permeability, porosity, and occasionally, relative permeability. An
anisotropic property varies with the direction of measurement and, hence, has
intrinsic tensorial character (see Sec. 2-2). Following Greenkorn and Kessler (1969),
the definitions of nonuniformity and heterogeneity are closely related (Fig. 6-5). A
homogeneous, uniform property is represented on a frequency distribution plot as a
single delta function (spike), and a heterogeneous, uniform property by a finite
number of these functions. A homogeneous, nonuniform property cannot be
represented by a finite number of delta functions but can be a continuous function
having only one peak. A heterogeneous, nonuniform property is represented by a
continuous distribution function having two or more peaks. Most laboratory
displacements are homogeneous and nonuniform. Most calculation techniques
assume the reservoir is uniform and heterogeneous. Actual nonuniformities are
frequently “averaged” out by capillary pressure or dispersion.
Figure 6-5 Probability distribution functions for parameter A (from Greenkorn

and Kessler, 1965)
195
Flow and Storage Capacity
Since permeability can change several factors of 10 within a short distance in a

reservoir, whereas porosity changes by only a few percent over the same scale, it is
common to view the reservoir as homogeneous with respect to porosity and
heterogeneous with respect to permeability. Although most of the traditional
measures of heterogeneity adopt this convention it is not necessary and can even lead
to occasional errors. In the following discussion, we include porosity variations in the
definitions; the more traditional definitions can be recovered by letting porosity and
thicknesses be constant.
Imagine an ensemble of NL permeable media elements each having different
permeability kl, thickness hl, and porosities φl. The elements are arranged as
resistances parallel to flow. From Darcy’s law, the interstitial velocity of the
single-phase flow of a conservative tracer is proportional to the ratio of permeability
to porosity rl = kl/φl. Thus if rl is a random variable, we can rearrange the elements in
order of decreasing rl (this is equivalent to arranging in order of decreasing fluid
velocity), and we can define a cumulative flow capacity at a given cross section as
n
kl hl
Fn = ∑ (6.3-1a)
l =1 Ht k
where Ht is the total thickness,

NL
Ht = ∑ h (6.3-1b)
l =1
The average quantities are defined as

NL
1
k=
Ht
∑ (kh)
l =1
l (6.3-1c)
and similarly for porosity. A cumulative storage capacity follows in a similar fashion
n
φl hl
Cn = ∑ (6.3-1d)
l =1 H tφ
The physical interpretation of Fn is that if the NL elements are arranged in

parallel, Fn is the fraction of total flow of velocity rn or faster. Cn is the volume
fraction of these elements. A plot of Fn versus Cn yields the curve shown in Fig.
6-6(a); if NL becomes very large, the ensemble approaches the continuous distribution
shown in Fig. 6-6(b). We designate the continuous distribution by F and C without
subscripts. From the definitions F, C, and r, the slope of either curve at any C is the
interstitial velocity at that point divided by the average interstitial velocity of the
whole ensemble
196
Figure 6-6 Schematic of discrete and continuous flow-storage capacity plots
⎧ rn
dF ⎪⎪ r (discrete)
= F'=⎨ (6.3-2)
dC ⎪ r (continuous)
⎪⎩ r
Because the elements were rearranged, the slope is monotonically decreasing and,
from the definitions, Fn = Cn = 1, when n = NL.
Measures of Heterogeneity
A common measure of reservoir heterogeneity is the Lorenz coefficient Lc, defined as

the area between the F-C curve and a 45° line (homogeneous F-C curve) and
normalized by 0.5,
⎧ 1 1⎫
Lc = 2 ⎨ ∫ FdC − ⎬ (6.3-3)
⎩ 0 2⎭
for the continuous curve. The Lorenz coefficient varies between 0 (homogeneous)
and 1 (infinitely heterogeneous). A second, perhaps more common, measure that lies
between the same limits is the Dykstra-Parsons (1950) coefficient VDP.
( F ')C = 0.5 − ( F ')C = 0.841
VDP = (6.3-4)
( F ')C = 0.5
Both Lc and VDP are independent of the particular form of the k/φ distribution, and
both rely on the rearrangement of this ratio. As originally defined, VDP is actually
taken from a straight line fit to the k-φ data plotted on a log–probability scale. This
procedure introduces a nonuniqueness (two different distributions having the same
VDP) into VDP when the data are not lognormal (Jensen and Lake, 1986). For
lognormal data, Eq. (6.3-4) is unique.
197
To relate F to C, we assume the permeability assembly is lognormally
distributed; hence the relationship between cumulative frequency Λ and r is (Aithison
and Brown, 1957)
⎡ ⎧ ⎛ r ⎞ ⎫⎤
1 ⎢⎢ ⎪⎪ ln ⎜ rˆ ⎟ ⎪⎪⎥
Λ = 1 − erf ⎨ ⎝ ⎠ ⎬⎥ (6.3-5)
2⎢ ⎪ 2vLN ⎪⎥
⎢ ⎪⎩ ⎪⎭⎥⎦
⎣
where r̂ is the geometric or log-mean of the distribution, and vLN is the variance of
the distribution. The relationship between r̂ and r is given by
r = re
ˆ ( vLN / 2) (6.3-6)
If we identify Λ with the storage capacity C and use Eqs. (6.3-2), (6.3-5), and (6.3-6),
we obtain
1⎡ ⎧⎪ ln(evLN / 2 F ') ⎫⎪⎤
C = ⎢1 − erf ⎨ ⎬⎥ (6.3-7)
2⎢ ⎩⎪ 2 v ⎭⎪ ⎥
⎣ LN ⎦
Equation (6.3-7) may be solved for F’ and then integrated subject to the boundary
condition F = C = 0,
C ⎧v ⎫
F = ∫ exp ⎨ LN + 2vLN erf −1 (1 − 2ξ ) ⎬dξ (6.3-8)
0
⎩ 2 ⎭
We integrate Eq. (6.3-8) numerically to give the F-C curve for fixed VLN. Figure 6-7,
which uses VDP instead of vLN, shows the results of such an integration where the
Figure 6-7 Flow-capacity–storage-

capacity curves (from Paul et al., 1982)
198
filled points are the results of the integration. It follows from Eqs. (6.3-4) and (6.3-5)
that
− v
VDP = 1 − e LN (6.3-9)
and, further, that the relationship among Lorenz and Dykstra-Parsons coefficients and
vLN is
⎛ v ⎞ ⎛ − ln(1 − VDP ) ⎞
Lc = erf ⎜ LN ⎟ = erf ⎜ ⎟ (6.3-10)
⎜ 2 ⎟ ⎝ 2 ⎠
⎝ ⎠
Note from Eq. (6.3-10) that Lc and VDP are bounded, whereas vLN is not.
Considering the three heterogeneity measures in Eq. (6.3-10), it must seem
odd to propose a fourth, but none of the measures discussed so far directly relates to
flow in permeable media. To ameliorate this, Koval (1963) proposed a heterogeneity
factor HK as a fourth measure of heterogeneity. HK is defined by
1− C F
HK = i (6.3-11)
C 1− F
Equation (6.3-11) follows from observing the similarity between a homogeneous

media fractional flow curve having straight-line relative permeabilities and zero
residual phase saturations, and the points generated in Fig. 6-7. In fact, the solid lines
in Fig. 6-7 are calculated from Eq. (6.3-11), with HK adjusted to fit the calculated
points. Hence there is a unique correspondence between VDP and HK, which is shown
in Fig. 6-8 as the filled points. From Eqs. (6.3-8) and (6.3-11), it follows that HK → ∞
Figure 6-8 Relation between effective

mobility ratio and heterogeneity (from
Paul et al., 1982)
199
as VDP → 1 (infinitely heterogeneous) and HK → 1 as VDP → 0. Between these limits,
the relation between VDP and HK is given by the following empirical fit to points in
Fig. 6-8:
VDP
log( H K ) = (6.3-12)
(1 − VDP )0.2
which is also shown in Fig. 6-8.

Using the F-C curve in Eq. (6.3-11), the vertical sweep efficiency of a unit
mobility ratio displacement may be calculated by the one-dimensional theory in Sec.
5-2 (see Exercise 5E).
Table 6-1 shows various statistical information from several producing
formations. VDP varies between 0.65 and 0.89 for these formations. This rather tight
range corresponds to the range in Fig. 6-7 where HK begins to be an exceedingly
strong function of VDP. Since EI decreases with increasing HK, displacements in most
reservoirs should be affected by heterogeneity. Table 6-1 also shows the areal VDP
based on the distribution of the average well permeabilities. In only three of the
entries shown are the areal variations larger than the vertical variations. This, plus the
lack of sensitivity of EI to heterogeneous permeabilities arranged in series, partly
accounts for the popularity of the stratified, or “layer-cake,” model for reservoirs. We
use the layer-cake model (uniform and heterogeneous) in the next two sections to
calculate EI. Finally, Table 6-1 shows the number of wells in each formation whose
core plug permeabilities most closely conform to the normal, lognormal, and
exponential distributions. As you can see, the wells are usually lognormal though
there are significant exceptions (see Jensen et al., 1987). If the permeabilities are
distributed normally, the procedure for calculating VDP and Lc is still correct, but the
form of the distribution function (Eq. 6.3-5) changes (see Exercise 6B).
None of the measures of heterogeneity given above are entirely satisfactory
for predicting displacement performance. Since all the measures capture both
heterogeneities and nonuniformities, there is a persistent, and largely ignored,
question about how to use them in displacement calculations. It seems reasonable that
nommiformities would alter values of the permeable media properties such as
dispersivity and capillary pressure; however, the scale of the nonuniformity is
different between the field and laboratory measurements, and there is little to suggest
how labmeasured properties can be changed to reflect these nonuniformities. For this
reason, nonuniformities are usually ignored, and displacement calculations are based
on uniformly heterogeneous permeable media models. A second reason for the
inadequacies in the heterogeneity measures is that for many reservoirs it is
inappropriate to treat permeability and porosity as independent variables.
Correlations exist between permeability and porosity (bivariate correlations), and
these variables themselves can have spatial structure (autocorrelation). When such
structure does exist, the displacement response of a rearranged ensemble of layers
will not be the same as that of the original distribution of layers. Determining when
structure exists, and separating it from the random stochastic component, are tasks
usually left to the geological interpretation in current practice.
200
TABLE 6-1 TYPICAL VALUES OF VERTICAL AND AREAL DYKSTRA-PARSONS COEFFICIENTS

(ADAPTED FROM LAMBERT, 1981)
Number
of wells
Field Formation studied* Lognormal Exponential k φ (VDP)areal VDP
1 El Dorado Admire 262 35 42 370.14 0.2538 0.484 0.697
2 Keystone Cardium 67 61 5 15.15 0.1063 0.752 0.653
3 Garrington Meanville B 38 35 2 5.73 0.1124 0.671 0.822
4 Madison Bartlesville 36 10 10 29.95 0.1790 0.238 0.823
5 Pembina Cardium 16 13 0 273.64 0.1220 0.837 0.894
Belly River 17 15 0 12.66 0.1623 0.687 0.814
6 Hamilton Dome Tensleep 33 11 10 98.24 0.1430 0.501 0.694
7 Rozet Muddy 33 25 3 43.14 0.1708 0.457 0.846
8 Salt Creek 2nd Wall Creek 30 8 6 59.08 0.1843 0.495 0.851
9 Kitty Muddy 20 19 0 11.74 0.0871 0.795 0.731
10 E. Salt Creek 2nd Wall Creek 5 3 2 35.71 0.1660 0.124 0.840
Lakota 7 6 0 38.01 0.1540 0.424 0.899
11 Dixie West Tradewater 16 5 3 129.13 0.1880 0.202 0.598
12 Burke Ranch Dakota 14 6 3 23.18 0.1191 0.625 0.663
13 Oklahoma City Prue 14 6 4 15.90 0.1368 0.473 0.683
14 Gas Draw Muddy 14 6 0 71.61 0.1572 0.615 0.899
15 Recluse Muddy 12 9 0 74.93 0.1437 0.591 0.855
16 W. Moorcroft Muddy 8 6 0 201.39 0.2150 0.973 0.833
17 S. Rozet Minnelusa 8 7 1 135.86 0.1283 0.443 0.861
18 Ute Muddy 8 2 2 62.14 0.1790 0.752 0.758
19 Riverton Dome Tensleep 7 2 0 2.68 0.0480 0.474 0.729
20 Carson-Hamm Minnelusa 7 2 3 160.36 0.1624 0.465 0.722
21 N.W. Sumatra Heath 6 0 2 124.98 0.1285 0.254 0.890
22 Pitchfork Tensleep 5 1 4 91.54 0.1410 0.229 0.728
Phosphoria 6 4 1 18.16 0.1430 0.544 0.833
Total 689 297 102
* Difference between lognormal and exponential represents number of normally distributed wells
201
6-4 DISPLACEMENTS WITH NO VERTICAL
COMMUNICATION
In this section, we illustrate the effects of mobility ratio and heterogeneity for
noncommunicating reservoirs. We treat pistonlike displacements of oil (i = 2) by
water (i = 1) in uniformly heterogeneous, horizontal layer-cake models (see Exercise
6C). Further, we do not allow permeability or transmissibility in the vertical
direction, a condition that could apply in actual practice if the reservoir contains
impermeable and continuous shale breaks in the total interval. The reservoir now
consists of an ensemble of one-dimensional elements arranged in parallel. Since there
is no vertical communication, we can rearrange the layers in decreasing k/φ, as in
Sec. 6-3. We also ignore dissipative effects to derive the noncommunication
displacement model first proposed by Dykstra and Parsons (1950).
Subject to the above assumptions, the vertical sweep efficiency of the
reservoir is
n NL
∑ (φ h)l + ∑ (φ hxD )l
EI = l =1 l =1
(6.4-1)
φ Ht
where xDl is the dimensionless front position (xfl/L) between the displacing fluid
(water) and the displaced fluid (oil). The index n denotes the layer that has just
broken through to the producer at a particular dimensionless time tD
NL
∫ qdt
t
∑ (φ hx )
D l
tD = 0
= l =1
(6.4-2)
H tWLφ φ Ht
where xDl for l > n is greater than 1, W is the width of the cross section, and L is the
length.
Two Layers
First, let’s consider the case of a reservoir having only two layers (NL = 2) as shown
in Fig. 6-9 with water saturation change ΔS = S1J – S1I. The k/φΔS for the upper or
fast layer is greater than that for the lower or slow layer. The front position in each
layer may be determined from Darcy’s law
dx fl ⎛ k ⎞ ΔP
= vl = − ⎜ ⎟ λrel , l = 1, 2 (6.4-3)
dt ⎝ φΔS ⎠l L
where vl is the interstitial x velocity in layer l, and λrel is the effective relative
mobility in layer l defined by
⎧⎡ x (1 − xDl ) ⎤
−1
⎪ ⎢ Dl0 + ⎥ for xDl < 1

λrel = ⎨ ⎣ λr1 λr02 ⎦ (6.4-4)
⎪ 0
⎩λr1 for xDl > 1
202
Figure 6-9 Schematic illustration of

heterogeneous reservoir for Dystra-
Parsons model
Taking the ratio of the interstitial velocities in the two layers will eliminate time and
the pressure drop since both layers experience the same ΔP. This equality implies
communication in the wells even though there is no communication elsewhere.
Because ΔP cancels, the calculation is valid whether the displacement is at constant
rate or constant ΔP. Thus before breakthrough (xD1 < 1), we have
dxD1 x + M 0 (1 − xD 2 )
= r12 D 2 (6.4-5a)
dxD 2 xD1 + M 0 (1 − xD1 )
where r12, the heterogeneity contrast (k1φ2ΔS2/k2φ1ΔS1), is greater than 1. After

breakthrough (xD1 > 1), the same quantity is
dxD1
= r12 ( xD 2 + M 0 (1 − xD 2 )) (6.4-5b)
dxD 2
In both equations, M0 is the endpoint mobility ratio defined in Eq. (5.2-3c). Before
breakthrough xD1 and xD2 are less than 1, we can integrate Eq. (6.4-5a) subject to the
boundary condition that xD1 = 0 when xD2 = 0 to give
1− M 0 2 ⎛1− M 0 2 ⎞
xD1 + M 0 xD1 = r12 ⎜ x D 2 + M 0 xD 2 ⎟ (6.4-6)
2 ⎝ 2 ⎠
The front position in the lower layer at breakthrough xD2 2 follows from Eq. (6.4-6) by
setting xD1 = 1
1/ 2
⎧ 0 2 1 − ( M 0 )2 ⎫
⎨( M ) + ⎬ −M
0
r
xD0 2 = ⎩ 12 ⎭ (6.4-7a)
1− M 0
After breakthrough, the front in the upper layer (outside the reservoir) is given by
integrating Eq. (6.4-5b) with the boundary condition xD2 = xD0 2 when xD1 = 1.
⎡1 − M 0 2 ⎤
xD1 = 1 + r12 ⎢ ( xD 2 − ( xD0 2 ) 2 ) + M 0 ( xD 2 − xD0 2 ) ⎥ (6.4-7b)
⎣ 2 ⎦
203
The front “position” in the upper layer at complete sweepout is given by Eq. (6.4-7b)
with xD2 = 1
⎛1+ M 0 ⎞
xD1 = 1 + (r12 − 1) ⎜ ⎟ (6.4-7c)
⎝ 2 ⎠
For fixed values of the mobility ratio, and heterogeneity contrast, EI at a

given dimensionless cumulative injection may be obtained by substituting the front
positions calculated from Eqs. (6.4-6) and (6.4-7) in the definitions (Eqs. 6.4-1 and
6.4-2). Figure 6-10 shows the results of this procedure for three values of M0 and two
values of the permeability contrast.
Figure 6-10 Two-layer Dykstra-Parsons calculation

204
For increasing permeability contrast EI decreases (Fig. 6-10a). A decreasing
M0 improves, and an increasing M0 worsens EI, just as similar changes do for the
areal and displacement sweep efficiencies. Fig. 6-10(b) plots the ratio of the
volumetric flow rate into layer 1 to the total flow rate as a function of tD. This follows
from Eqs. (6.4-6) and (6.4-7)
q1 (φ hΔSv)1 1
= = (6.4-8)
q1 + q2 (φ hΔSv)1 + (φ hΔSv) 2 (kh) 2 ⎡ M + xD1 (1 − M 0 ) ⎤
0
1+
(kh)1 ⎢⎣ M 0 + xD 2 (1 − M 0 ) ⎥⎦
Equation (6.4-8) shows the reason for the changes in EI. For M0 < 1, the fast layer is
filling up with a low mobility fluid faster than the slow layer. Thus the fast layer
resistance to flow is increasing faster than the slow layer resistance, causing the fast
layer flow rate to decrease. For M0 > 1, the situation is exactly reversed. Of course,
for M0 = 1, there are no changes in mobility, and the ratio of fast layer rate to total
rate stays constant. Mobility ratio can have an effect on EI even if there is no vertical
communication. This effect has qualitatively the same trend as the areal and
displacement sweep.
NL Layers
The above results may be readily generalized to an ensemble of NL layers. First, we

generalize the heterogeneity contrast to be between any two layers l and n
⎛ k ⎞ ⎛ φΔS ⎞
rnl = ⎜ ⎟ ⎜ ⎟ (6.4-9)
⎝ φΔS ⎠ n ⎝ k ⎠l
At a particular time, if n is the number of the layer breaking through, the front
position in all the faster layers is given by
⎛1+ M 0 ⎞
xDl = 1 + (rln − 1) ⎜ ⎟ , l = 1, . . . , n (6.4-10a)
⎝ 2 ⎠
from Eq. (6.4-7c). Similarly, the position in all the slower layers is
1/ 2
⎧ 0 2 1 − ( M 0 )2 ⎫
⎨( M ) + ⎬ −M0
⎩ rnl ⎭
xDl = , l = n + 1, . . . , N L (6.4-10b)
1− M 0
from Eq. (6.4-7a). By letting n take on values between 1 and NL, calculating all NL
front positions, and substituting these into Eqs. (6.4-1) and (6.4-2), we can construct
a plot of EI versus tD (see Exercise 6D). The EI-tD plot follows from this by
connecting these points by straight-line segments. That this procedure is not rigorous
may be seen from Fig. 6-10, where the curves between breakthrough and sweepout
are slightly curved. But if NL is large, the error introduced by this procedure will be
small. The procedure may be easily modified to calculate EI in an ensemble having
the continuous F-C distribution discussed in Sec. 6-3. Using the water–oil ratio
205
n
∑q l
WOR = l =1
NL
(6.4-11)
∑q
l = n +1
l
as the time variable in place of tD, Johnson (1956) has presented the vertical sweep
efficiency as a function of VDP and M0 in graphical form.
6-5 Vertical Equilibrium
A useful procedure for making general oil recovery calculations is to invoke the
assumption of vertical equilibrium (VE) across the cross section of the reservoir the
displacement is taking place in. When VE applies, it is possible to combine vertical
and displacement sweep efficiencies into a pseudodisplacement sweep, which then
may be estimated by the one-dimensional theory of Sec. 5-2. This combination
means recovery efficiency ER becomes
ER = E A EI ED = E A ED (6.5-1)
where ED is the pseudodisplacement sweep efficiency. Of course, the areal sweep

efficiency EA must still be estimated and used in Eq. (6.5-1). We discuss how to
combine EA and ED in Sec. 6-7. Another consequence of the VE assumption is this
represents a state of maximum transverse fluid movement, or crossflow. Thus
calculations based on VE are useful in estimating the tendency of crossflow to affect
displacements when compared to the noncrossflowing calculations of Sec. 6-4.
The VE Assumption
Formally, vertical equilibrium is a condition where the sum of all the fluid flow
driving forces in the direction perpendicular to the direction of bulk fluid flow is
zero. We see this condition is more nearly met by flow in reservoirs having large
aspect ratios (length to thickness) and good vertical communication. Moreover, Sec.
6-6 shows that several classical displacement calculations in the petroleum literature
are, in fact, subsets of the more general theory of vertical equilibrium.
To derive a general VE theory, we restrict ourselves to incompressible,
immiscible displacements of oil by water and derive the water saturation profile in
the transverse direction (z direction) at a fixed cross section (x position). For the
assumptions listed above, the conservation Eq. (2D-1) for water becomes in x-z
coordinates,
∂S ∂u ∂u
φ 1 + x1 + z1 = 0 (6.5-2)
∂t ∂x ∂z
If we introduce Darcy’s law (Eq. 2.2-5) into Eq. (6.5-2) and scale the independent
variables x and z as
206
x z
xD = , zD = (6.5-3)
L Ht
Eq. (6.5-2) becomes
⎛ L2 ⎞ ∂S ∂ ⎛ ⎛ ∂P1 ⎞⎞
φ⎜ ⎟ 1 − ⎜⎜ λr1 ⎜ + L ρ1 g sin α ⎟ ⎟⎟
⎝ k ⎠ ∂t ∂xD ⎝ ⎝ ∂xD ⎠⎠
⎛ L2 ⎞ ∂ ⎛ ⎛ ∂P1 ⎞⎞
−⎜ 2 ⎟ ⎜⎜ k z λr1 ⎜ + H t ρ1 g cos α ⎟ ⎟⎟ = 0 (6.5-4)
⎝ H t k ⎠ ∂z D ⎝ ⎝ ∂z D ⎠⎠
The terms in this equation represent water accumulation, x-direction flow, and z-
direction flow, respectively (see Fig. 6-11). We assume flow in the z direction is
finite; therefore, if the group L2 / kH t2 large, it follows that the term it multiplies must
be small. This means the z-direction water flux is a function of x only, or
⎛ ∂P ⎞
k z λr1 ⎜ 1 + ρ1 g cosα ⎟ = f ( x) (6.5-5)
⎝ ∂z ⎠
Since the water flux in the z direction is finite, if kz is large, Eq. (6.5-5) implies
∂P1
= − ρ1 g cos α (6.5-6)
∂z
Clearly, the above reasoning breaks down at water saturations near the irreducible
water saturation where λr1 is zero. But it is true that the saturation range where Eq.
(6.5-6) breaks down is precisely the range where the analogous equation for the oil
phase is most relevant. Therefore, the arguments leading to Eq. (6.5-6) should be
valid in some average sense when applied to both the water and oil phases.
Assuming the group L2 / kH t2 large is reasonable for many practical cases.
But assuming kz is large strains credibility since for most naturally occurring media kz
is less than k. For permeable media having dispersed shale barriers, kz can be much
smaller than k.
The requirements of large L2 / kH t2 and kz may be combined into a single
requirement that the effective length-to- thickness ratio
1/ 2
L ⎛ kz ⎞
RL = ⎜ ⎟ (6.5-7a)
Ht ⎝ k ⎠
be large. In Eq. (6.5-7a), the permeabilities are an arithmetic average for

1 Ht
H t ∫0
k = kdz (6.5-7b)
and a harmonic average for kz

Ht
kz = (6.5-7c)
H t dz
∫0 k z
207
A displacement actually approaches VE asymptotically as RL becomes large. Based
on numerical solution (Zapata, 1981) and analytic solutions (Lake and Zapata, 1987),
an RL greater than 10 is sufficient to ensure that the z-direction sweep efficiency is
reasonably well described by VE. You may easily verify that RL can be large for a
wide variety of reservoirs. For example, for a 16.2 hm2 (40-acre) spacing of five-spot
patterns, the injector–producer distance is 285 m (933 ft). If we take this to be L, then
for Ht = 6.1 m (20 ft) and k z = 0.1 k , we have RL = 14.8, which is large enough for
VE to be a good approximation. By taking the k to be a harmonic average over the
reservoir interval, it is clear the k z = RL = 0 if there are one or more impermeable
barriers (for example, continuous shale layers) within the interval Ht. Clearly, the VE
assumption will not apply in this case. But the pseudodisplacement sweep efficiency
of the intervals between the barriers may be estimated based on VE, and the
combined response of all such intervals may be estimated by the communicating
methods of Sec. 6-4.
RL may be regarded as a ratio of a characteristic time for fluid to cross the
reservoir in the x direction to that in the z direction. If RL is large, saturation or
pressure fluctuations in the z direction decay much faster than those in the x direction.
Therefore, we neglect the z-direction perturbations. Thus when we say that the VE
assumption applies or that the subject reservoir is in vertical equilibrium, we are
saying, for the bulk of the reservoir, z-direction fluctuations are negligible.
Arguments based on the decay time of perturbations were originally advanced by G.
I. Taylor for flow in capillary tubes (Lake and Hirasaki, 1981).
For large RL, the P1 profile in the z direction is gien by Eq. (6.5-6) for most of
the cross sections in the reservoir. This procedure applies equally well to the oil
phase, giving
∂P1 ∂P
+ ρ1 g cos α = 0 = 2 + ρ 2 g cos α (6.5-8)
∂z ∂z
When the definition for oil–water capillary pressure Pc = P2 – P1 is introduced into

this equation, we have
∂Pc
= −( ρ1 − ρ 2 ) g cos α ≡ −Δρ g cos α (6.5-9)
∂z
Equation (6.5-9) implicitly describes the water saturation profile in the z direction
since Pc is a known function of water saturation. But this saturation distribution is
just what would be observed in the transition zone between oil and water under static
conditions. Compare Eqs. (6.5-9) and (2A-1), noting the z and Pc increase in the
opposite directions. Hence the z-direction saturation profile given by Eq. (6.5-9) is
identical to that predicted by assuming no flow in the z direction.
We stated that VE is a condition that causes maximum crossflow of fluids, so
it is surprising, to say the least, that the same equation describes the saturation profile
under conditions of zero and maximum z-direction flow. The situation is analogous
to heat conduction in metal rod where the driving force for heat transfer is a
temperature gradient along the axis of the rod (Coats et al., 1971). If the thermal
208
conductivity of the rod is not zero and no heat flows along the rod, the temperature
difference between the ends of the rod is zero. But if heat flows at a fixed finite rate
along the rod, and the thermal conductivity of the rod is large, the temperature
difference is again small. The latter case is analogous to the VE flow in the z
direction of Fig. 6-11 where, since the thermal conductivity is large, the heat transfer
rate is maximum; the former case is the analogue to hydrostatic equilibrium.
Figure 6-11 Schematic cross-section for vertical equilibrium procedure
Displacement Classification
One of the consequences of VE is a classification of displacements according to

degree of segregation. Let S1+ be some water saturation slightly below 1 – S2r, and
S1− slightly above S1r. We can define a capillary transition zone thickness zCTZ as the
z-direction distance over which the water saturation changes between these two
limits. From Eq. (6.5-9) and Fig. 6-12, this is
Pc P
− c
−
S1 S1+
zCTZ ≡ z −z = (6.5-10)
S1+ S1− Δρ g cos α
We have made the integration of Eq. (6.5-9) assuming the capillary-pressure–water-

saturation relation applies throughout zCTZ. In general, the capillary transition zone
defined by Eq. (6.5-10) is not the same as that existing at the original water–oil
contact, down structure to the left in Fig. 6-11. The idea of the capillary transition
zone in a VE reservoir allows the definition of two broad classes of displacements
(Dake, 1978). If zCTZ >> Ht, the water saturation profiles in the z direction are
essentially flat, and the flow is said to be diffuse. If zCTZ << Ht, the capillary transition
zone is small with respect to the reservoir thickness, and the flow is segregated.
These definitions suggest ideas similar to the definitions of sharpening and spreading
waves in Sec. 5-2 except that the latter definitions apply to cross-sectional averaged
209
Figure 6-12 Schematic of capillary

transition zone
saturation waves. The mixing or transition zones in Sec. 5-2 were in the x direction
only and were largely caused by chromatographic effects inherent in the permeable
medium oil–water fractional flow curves. The capillary transition zone defined by
Eq. (6.5-10) is in the z direction and defined by the capillary-pressure–
water-saturation relation, the dip angle, and the density difference.
Saturation Profile
Let’s now consider the integration of Eq. (6.5-9) at the three different cross sections
A, B, and C in Fig. 6-13. In this figure, flow is from right to left for ease of
illustration. We take S1A, S1B, and S1C to be the water saturations at the bottom (z = 0)
of the reservoir at the indicated cross sections x = xA, xB, and xC. Because of the
direction of flow, and because the initial water saturation is near the irreducible value
S1A > S1B > S1C. The water saturation profile at each of these cross sections is given
implicitly from Eq. (6.5-9)
Pc ( S1 ( xk , z )) = Pc ( S1k ) + Δρ gz cosα , k = A, B, or C (6.5-11)
We do not, at this point, know the x-direction position of the z = 0 water saturations,
which we indirectly determine below. But we can schematically sketch in lines
connecting constant values of S1, as indicated in Fig. 6- 13. For positive values of the
210
Figure 6-13 Schematic of z-direction water saturation profiles at various cross

sections
density difference, the usual case, the isosaturation lines suggest an underrunning of
the oil by the injected water. This underrunning, or gravity tongue, is a persistent
feature of reservoirs in which gravity forces are strong. Tonguing occurs even in
reservoirs that have no dip cos α = 1(Dz = –z). The extent of the tonguing is greatly
influenced by the shape of the capillary pressure curve. In Sec. 6-6, we discuss a
special case of the VE theory in which capillary forces are negligible, and the gravity
tonguing occurs as segregated flow.
Pseudoproperties
To use the z-direction S1 profile, we must convert the original two-dimensional Eq.
(6.5-2) to an equivalent one-dimensional equation. Let’s integrate Eq. (6.5-2) over
the interval thickness Ht and divide the equation by Ht
1 H t ∂S1 1 H t ∂u x1 1 H t ∂u z1
∫
H t 0 ∂t
dz + ∫
H t 0 ∂x
dz +
H t ∫0 ∂z
dz = 0 (6.5-12)
Since Ht is a constant, the integration and differentiation in the first term commutes,
and Eq. (6.5-12) becomes
∂S ∂u
φ 1 + x1 = 0 (6.5-13)
∂t ∂x
Terms involving z-direction water flux do not appear in Eq. (6.5-13) since all fluxes
211
vanish at the upper and lower impermeable boundaries of the reservoir. In Eq.
(6.5-13), the averages are
1 Ht
H tφ ∫0
S1 = φ S1dz (6.5-14a)
1 Ht 1 Ht
φ = ∫
Ht 0
φ dz , u1x =
H t ∫0
u x1dz (6.5-14b)
In these definitions, and in those that follow, all averages are arithmetic averages
except the water saturation, which is weighted by the porosity. Introducing the
definitions for dimensionless independent variables
x t u dt
xD = , t D = ∫ x (6.5-15)
L 0 φ L
into Eq. (6.5-13) yields,

∂S1 ∂f1
+ =0 (6.5-16)
∂t D ∂xD
where u x = u x1 + u x 2 , and f1 = u x1 u x is a cross-sectional averaged water fractional

flow function. Eq. (6.5-16) is identical to Eq. (5.2-5a) and can be solved in the same
manner as the Buckley-Leverett and Welge integration procedures once we define f1
in terms of S1 .
Consider the cross-sectional averaged total flux multiplied by Ht with
Darcy’s law substituted for the local flux
Ht ⎛ ∂P ⎞ Ht ⎛ ∂P ⎞
H t u x = − ∫ k λr 2 ⎜ 2 + ρ 2 g sin α ⎟ dz − ∫ k λr1 ⎜ 1 + ρ1 g sin α ⎟ dz
0
⎝ ∂x ⎠ 0
⎝ ∂x ⎠
(6.5-17)
We can express the x-direction oil phase pressure gradient in terms of the water phase
pressure gradient and factor to give
Ht ∂P Ht ∂P
H t u x = − ∫ k (λr 2 + λr1 ) 1 dz − ∫ k λr 2 c dz
0 ∂x 0 ∂x
Ht
− g sin α ∫ k (λr 2 ρ 2 + λr1 ρ1 )dz (6.5-18)
0
But from Eq. (6.5-6), it follows that
∂ 2 P1 ∂ 2 P1 ∂ ⎛ ∂P1 ⎞
= = ⎜ ⎟=0
∂x∂z ∂z∂x ∂z ⎝ ∂x ⎠
hence under VE, the water phase pressure gradient in the x direction is independent
of z, as are both ∂P2 ∂x and ∂Pc ∂x . All gradients may be factored from the
integrations and solved for as
212
⎛ ∂P ⎞ Ht Ht
H t u x + ⎜ c ⎟ ∫ λr 2 kdz + g sin α ∫ k (λr 2 ρ 2 + λr1 ρ1 )dz
∂P ⎝ ∂x ⎠ 0 0
− 1= (6.5-19)
∂x Ht
∫
0
k (λr 2 + λr1 )dz
The pressure gradient of Eq. (6.5-19) substituted into the averaged water flux
⎛ ∂P ⎞ H t Ht
H t u x1 = ⎜ − 1 ⎟ ∫ k λr1dz − g sin α ∫ k λr1 ρ1dz (6.5-20)
⎝ ∂x ⎠ 0 0
gives
u x1 ( k λr1 ) ⎧⎪ (k λr 2 ) ⎛ ∂Pc ⎞ (k λr 2 )Δρ g sin α ⎫⎪

f1 = = ⎨1 + − ⎬ (6.5-21)
u x k (λr1 + λr 2 ) ⎪⎩ u x ⎜⎝ ∂x ⎟⎠ ux ⎭⎪
Comparing this equation with Eq. (5.3-1) suggests the following definitions for
pseudorelative permeabilities
1 Ht
H t k ∫0
k r1 = kkr1dz (6.5-22a)
1 Ht
H t k ∫0
kr 2 = kkr 2 dz (6.5-22b)
The capillary pressure in Eq. (6.5-21) is the capillary-pressure–water-saturation

relation for any z position in the reservoir. It does not matter which z position since
∂Pc ∂x is equal at all z positions. From this, it does not follow that the capillary-
pressure–water-saturation relation is the same in all z positions since these can vary
with permeability. The capillary pressure in Eq. (6.5-21) is often regarded as a
pseudofunction, even though it is an actual local curve, since it must be a function of
S1 .
To use the one-dimensional theory of Sec. 5-2 on these equations, we must
neglect the x-direction capillary pressure term in Eq. (6.5-21). This omission is not
equivalent to neglecting capillary pressure entirely since the capillary pressure in the
z direction determines, in part, the z-direction saturation profile. Though it seems
inconsistent to maintain capillary pressure in the z direction and neglect it in the x
direction, one can show by scaling arguments similar to those used in Sec. 5-3 that
when the conditions for VE apply, z-direction effects are far more important than x-
direction effects (Yokayama and Lake, 1981).
The procedure for calculating pseudorelative permeability curves ( kr1 and
kr 2 versus S1 ) is as follows:
1. Select a water saturation at the bottom of the reservoir S1k.

2. Determine the z-direction water saturation profile S1(xk, z) at cross section k
using Eq. (6.5-11) and the capillary-pressure–water-saturation relation.
213
3. Calculate the average water saturation at cross section k, S1 (xk), from Eq.
(6.5-14a) and from the z-direction porosity profile.
4. Calculate the pseudorelative permeabilities corresponding to S1k from Eq.
(6.5-22) and the z-direction permeability profile.
Steps 1–4 give a single point on the pseudorelative permeability curve. To construct
the entire curve, we repeat the procedure with different values of S1k. The procedure
gives all possible water saturation profiles and average water saturations for the
reservoir (see Fig. 6-13) though it does not give the x positions of these quantities,
which come from solving the one-dimensional Eq. (6.5-13). Though the averaging
procedure is fairly straightforward, most of the integrations in it must be evaluated
numerically in the absence of analytic functions for the capillary pressure and relative
permeability curves (see Exercise 6F).
Once the pseudorelative permeabilities are constructed, the pseudodisplace-
ment sweep efficiency ED follows from Eqs. (5.1-2) and (5.2-24) with the
appropriately averaged quantities appearing in place of the local quantities.
You should appreciate the generality of the VE approach, for we now have a
means for calculating and combining displacement ED and vertical EI sweep
efficiencies with little more trouble than calculating the displacement sweep alone,
VE can greatly simplify oil recovery calculations in desktop procedures and
numerical simulations (Coats et al., 1971); however, the entire procedure is restricted
to reservoirs having a large RL.
The generalized VE approach for EOR processes has yet to be worked out.
(For miscible flow, see Lake and Hirasaki, 1981.)
6-6 SPECIAL CASES OF VERTICAL EQUILIBRIUM
Though the VE procedure in Sec. 6-5 is quite general, being restricted to reservoirs
having constant properties in the x direction and a large RL, several VE flows are
special cases, Since these cases are useful in understanding many EOR processes, in
this section we review them and show how they follow from the general theory.
Homogeneous with Large Transition Zone
In this case, k and φ are both constant in the reservoir, and zCTZ >> Ht. From the
procedure given above, the saturation profile in the z direction will be essentially flat,
and the saturation at the reservoir bottom will not differ much from the average
saturation. In this case, the pseudorelative permeabilities krj become the local (or
REV) relative permeabilities krj. Large zCTZ would be the rule in most of the longer
laboratory core floods. In the shorter core experiments, VE is usually not a good
assumption, but S1 may still be uniform in a cross section since the S1 profile has not
had much time to distort.
214
Homogeneous, Uniform with No Transition Zone
Easily the most celebrated of the VE theories is the theory of gravity tonguing, or
underrunning, originally proposed by Dietz (1953). This theory was first proposed as
an alternative to the Buckley-Leverett theory, but it is actually a special case of the
VE theory because a finite time is required for the conditions underlying the theory to
apply. Since the publication of the original Dietz paper, the theory has been applied
to gravity overrunning by a miscible gas process (Hawthorne, 1960), and other work
has been published describing the approach to VE conditions (Crane et al., 1963). In
this section, we restrict ourselves to the water-displacing-oil case though the
overrunning case can be similarly developed.
The key assumption in the Dietz theory is the absence of a transition zone, or
zCTZ = 0. This condition can be accurate only for conditions where the capillary
pressure is small (well-sorted or high-permeability media). The sharp transition zone
or macroscopic interface resulting from this condition suggests the theory is
applicable to any displacement where simultaneous flow of more than one
component or phase is absent at any point in the reservoir. If Pc is identically zero,
Eq. (6.5-8) cannot be satisfied at any point in the reservoir since the oil and water
densities are not, in general, equal. The resolution of this is to let Eq. (6.5-6) apply to
zones flowing water and to let the analogous equation for oil apply to zones flowing
oil. Figure 6-14 shows the relevant cross section and these zones.
At any cross section containing the tongue, the average water saturation from
Eq. (6.5-14a) is
1
S1 = [b(1 − S 2 r ) + S1r ( H t − b)] (6.6-1)
Ht
and the pseudorelative permeability functions from Eq. (6.5-22) are

⎛ b ⎞
kr1 = kro1 ⎜ ⎟ (6.6-2a)
⎝ Ht ⎠
Figure 6-14 Schematic cross section of a water tongue

215
⎛ H −b⎞
kr 2 = kro2 ⎜ t ⎟ (6.6-2b)
⎝ Ht ⎠
The interface height b may be eliminated between Eqs. (6.6-1) and (6.6-2) to give
⎛ S1 − S1r ⎞ o ⎛ 1 − S 2 r − S1 ⎞
kr1 = kro1 ⎜ ⎟ , kr 2 = kr 2 ⎜ ⎟ (6.6-3)
⎝ 1 − S1r − S 2 r ⎠ ⎝ 1 − S1r − S 2 r ⎠
Thus the pseudorelative permeabilities are straight-line functions of the average

water saturation.
We can also derive the tilt angle β of the oil–water interface. Consider the
rectangle ABCD of height Δb and width Δx shown in Fig. 6-14. The dimensions Δx
and Δb are small (we pass to zero limit below) so that the interface between points A
and C is the diagonal of the rectangle. Along the BC side of the rectangle, the x-
direction water flux is
kk o ⎛ P − PB ⎞
u x1 = − r1 ⎜ C + ρ1 g sin α ⎟ (6.6-4a)
μ1 ⎝ Δx ⎠
and along the AD side, the x-direction oil flux is
kk o ⎛ P − PA ⎞
ux 2 = − r 2 ⎜ D + ρ 2 g sin α ⎟ (6.6-4b)
μ 2 ⎝ Δx ⎠
In the limit of Δx → 0, these two fluxes approach a common value ux since there can
be no accumulation at the interface. Further, the pressures at A and B, and at D and C,
are related because of the VE conditions (Eq. 6.5-8)
PB − PA = ρ1 g Δb cosα , PC − PD = ρ 2 g Δb cos α (6.6-5)
The four equations (Eqs. 6.6-4 and 6.6-5) may be combined to eliminate the four
pressures. This procedure gives
(u − u M o ) μ1
tan β = x1o x 2 + tan α (6.6-6)
(kkr1Δρ g )cosα
The tangent of the tilt angle is defined as
Δb
tan β = + lim (6.6-7)
Δx → 0 Δx
β is defined to be positive and can take on the entire range of values between 0° and
90°. If β is greater than 90°, the tongue is overrunning, and this procedure must be
repeated with the displacing fluid above the resident fluid.
For β > 0––that is, the interface is not parallel to the x axis––the interface
reaches a stabilized shape where β is independent of both time and z position. This
limit is not an automatic consequence of VE, but the time interval between the onset
of the VE conditions and the attainment of the stabilized interface shape appears to
be small (Crane et al., 1963). When this steady-state tilt angle βs is reached, the x-
direction fluxes ux1 and ux2 become independent of z and equal to the cross-sectional
216
average flux u x . Equation (6.6-6) then becomes
1− M 0
tan β s = 0 0 + tan α (6.6-8)
M N g cos α
where N g0 and M 0 are gravity numbers and mobility ratios defined in Eq. (5.2-3).
Equation (6.6-8) approaches the correct limits of an interface perpendicular
to the x direction for N g0 = 0 (no tonguing) and of a horizontal interface for M 0 →1.
In the case of a stable gravity tongue, the cross-sectional average water saturation
profile approaches a “constant pattern” mixing zone, whereas the directly analogous
case of a one-dimensional displacement with straight-line relative permeabilities
approaches a shock front. This is a consequence of the finite length of time required
for the VE conditions to apply in the tonguing case.
For β < 0, the interface completely underruns the oil and is said to be
unstable. The condition for stability is, from Eq. (6.6-8),
M 0 − 1 < M 0 N g0 sin α (6.6-9)
The equality form of Eq. (6.6-9) naturally leads to definitions of a critical endpoint
mobility ratio M c0 = M 0 |β s = 0
1
M c0 = (6.6-10a)
1 − N g0 sin α
and of a critical flux or rate uc = u x |β s = 0
Δρ gkkr01
uc = sin α (6.6-10b)
μ1 ( M 0 − 1)
The conditions to prevent complete underrunning of the oil by the water are ux < uc or
M0 < M c0 . Equation (6.6-10a) indicates gravity stabilization is possible even when
M0 > 1. Equation (6.6-10b), in particular, is used in estimating the flooding rates in
gravity-stabilized miscible displacements.
Layered, Uniform Horizontal Media with Pc = zCTZ = 0
For this case of the permeable medium consisting of NL layers, each of contrasting
thickness hl, permeability kl, and porosity, φl, the integrals in the definitions (Eqs.
6.5-14 and 6.5-22) become finite sums
1 NL
krj = ∑ (khkrj )l , j = 1 or 2
H t k l =1
(6.6-11a)
NL
1
S1 =
H tφ
∑ (φ hS )
l =1
1 l (6.6-11b)
The definitions (Eqs. 6.6-11a and 6.6-11b) are valid regardless of the ordering of the
layers; hence we assume, without loss of generality, they are ordered with decreasing
velocity as in Sec. 6-3.
217
Since neither gravity nor capillary pressure is present, Eq. (6.5-8) is trivially
satisfied, and there is no constraint on the saturations in the z direction. To resolve
this, Hearn (1971) assumed segregated flow within a layer, as in Fig. 6-15(a). The
definitions become
1 NL 1 NL
k r1 = ∑ r1 l r 2 H k ∑
H t k l = n +1
( khk 0
) , k =
l =1
(khkr02 )l (6.6-12a)
t
1 ⎧ NL NL
⎫
S1 = ⎨∑ (φ h(1 − S 2 r )l + ∑ (φ hS1r )l ⎬ (6.6-12b)
H tφ ⎩ l =1 l = n +1 ⎭
where n is the number of the slowest layer (smallest k/φ) flowing water at a given
cross section. Thus the average water saturation and pseudorelative permeabilities are
parametrically functions of n and can be regarded as functions of each other in this
way.
Figure 6-15 Schematic cross section of VE in stratified reservoir with no capillary and
gravity effects
Based on arguments related to the direction of flow caused by the viscous

pressure driving forces, Zapata and Lake (1981) have shown that assuming
segregated flow within a layer is correct when the displacement is favorable M0 < 1.
In fact, if VE holds, it is possible for the M0 to be so low that the effect of the
heterogeneities is entirely suppressed (see Exercise 6E). But when the displacement
is unfavorable, the viscous forces cause a mixing zone to develop between the front
in the fastest layer and that in the slowest layer (Fig. 6-15b). This mixing zone causes
the vertical sweep efficiency to be actually greater than the corresponding segregated
flow case since the mixing zone attenuates the unfavorable mobility ratio. That
diffuse flow can occur in VE displacements in the absence of capillary pressure is a
major revelation in the understanding of these processes. The implication is clear that
such crossflow might be a source of mixing in all unstable flows.
218
Stratified, Uniform with ΔP = 0 and Constant Mobility
Here, there are no gravity forces to counteract the z-direction imbibition, and the z-
direction water saturation profile is uniform within each layer. But because of the
variable properties in the z direction, the Pc – S1 function changes. Figure 6-16(a)
illustrates this change for the four-layer medium shown. From Eq. (6.5-9), the
capillary pressure (not the capillary pressure function) is a constant through any cross
section. As indicated in Fig. 6-16, if the constant is known, this specifies the water
saturation in each layer at that cross section. Because the mobility is constant, the x-
direction viscous pressure gradient is independent of both position and time. For this
case, the average water saturation and pseudorelative permeability curves are given
by Eq. (6.6-11), but each of the water saturations S1l are determined by the relation Pc
= constant and the Pc-S1 relation. Again, the average water saturation and
pseudorelative permeabilities are parametrically related through this constant. This
procedure yields an immiscible mixing zone between the most advanced and the least
advanced front, as shown in Fig. 6-16(b).
Figure 6-16 Schematic of stratified cross section with no gravity and viscous forces
6-7 COMBINING SWEEP EFFICIENCIES
In this section, we seek to provide an estimate of the recovery efficiency ER = EDEV,

from Eq. (2.5-5), as a function of dimensionless time by combining vertical, areal,
and displacement sweeps. As we mentioned in Sec. 6-1, this procedure is
complicated because all three sweep efficiencies depend on one another, and all must
219
be evaluated at times different from that at which the recovery efficiency is desired.
If the reservoir is layered and noncommunicating, we could, of course, calculate the
areal sweep efficiency of each layer and then average the φh of each layer times its
areal sweep to obtain the volumetric sweep efficiency. The procedure we describe
here includes this method as a special case, but it is valid for combining all types of
sweep efficiency curves, not just those for layer-cake models.
Our procedure is based on the idea of apparent pore volumes first presented
by Claridge (1972). We assume we have independently determined curves for EA, EI,
and ED as functions of tD. For ED, it is more convenient to work in the average water
saturations S1 , but there is no loss of generality since the two are related through Eq.
(5.1-2). Here, we are restricted to sweep efficiency functions that depend on
dimensionless time, heterogeneity, capillary pressure, and so on but that do not
depend explicitly on rate or fluid velocity.
Combining Areal and Vertical Sweep
The definition for volumetric sweep efficiency is repeated here as EV = EAEI. From
Fig. 6-1 (or Fig. 6-17), EA depends on the z position in the reservoir, and EI on a
particular cross section between the injector and producer. Rather than directly
determining these positions, we seek to determine that value of the dimensionless
time argument at which the respective values of EA and EV will give average values.
Therefore, we can rewrite Eq. (6.1-3) as
EV (t D ) = E A (t DA ) EI (t DI ) (6.7-1)
where tDA and tDI are dimensionless times based on the apparent pore volumes for
areal and vertical sweep, respectively.
Figure 6-17(a) schematically shows the positions of a pistonlike
displacement at or after breakthrough. Imagine that the shaded volumes have no
porosity or permeability, then the volumetric sweep efficiency is equal to the vertical
sweep efficiency. The ultimate volume to be swept out, for this rather oddly shaped
reservoir then, at infinite throughput is the unshaded or apparent pore volume. But
this is just the EA times the total pore volume; hence the dimensionless time for EI is
t
t DI = D (6.7-2a)
EA
By a similar argument, though it is much more difficult to show parallel cross

sections, the dimensionless time for the EA is
t
t DA = D (6.7-2b)
EI
The equation for the volumetric sweep may be written in combined form as
⎛t ⎞ ⎛t ⎞
EV = E A ⎜ D ⎟i EI ⎜ D ⎟ (6.7-3)
⎝ EI ⎠ ⎝ E A ⎠
220
Figure 6-17 Schematics for combining

sweep efficiencies
Since EA and EI appear as both multiplications and in the arguments,

constructing a volumetric sweep function requires a trial-and-error procedure.
1. Determine the cumulative injection at breakthrough t Do . This is just the

product of EA and EI at their respective breakthrough values, E A0 and EI0 .
2. Pick some tD after breakthrough t Do .
3. Select a trial EI.
4. Calculate the areal sweep efficiency from EA = EA(tD/EI).
5. Calculate the vertical sweep from EI = EI(tD/EA) using EA.
If EI agrees with that assumed in step 3, EA • EI is the volumetric sweep efficiency at

tD. If EI is different from step 3, select a new trial EI, and return to step 4.
221
Experience has shown that the procedure converges within two to three
iterations for typical EA and EI functions using direct substitution. By repeating the
procedure for several values of tD, a volumetric sweep efficiency curve may be
calculated (see Exercise 6H).
Combining Pseudodisplacement and Areal Sweep
The pseudodisplacement sweep efficiency ED may be determined from the VE

theory of Sec. 6-5. The recovery efficiency follows from this as
ER = ED (t DA ) E A (t DD ) (6.7-4)
where tDA and tDD are the dimensionless times based on the apparent pore volumes
appropriate for the particular sweep efficiency.
Combining a displacement and areal sweep in the manner described here is
again a generalization of the procedure proposed by Claridge (1972) and is repeated
in Chap. 7 for a miscible flood. The dimensionless time for ED is the same as Eq.
(6.7-2a), but for EA, we must view the displacement differently. Consider Fig.
6.17(b), which shows an areal view of a displacement divided into a swept and an
unswept region. The unswept region contains oil and water saturations at the values
present at the initiation (S1, S2)I of the displacement. We identify the saturations S1 ,
S 2 in the swept region with the cross-sectional averaged saturations determined from
ED
S1 = S1I + ED (1 − S1I ) (6.7-5)
At a particular time, the pore volume available to flow for a pistonlike displacement
whose front occupies the same position as that shown in Fig. 6-17(b) is the water
volume in the swept region. (Another way of viewing this is to suppose the oil
saturation S 2 is part of the immobile phase.) Therefore, the dimensionless time for
EA is now
t
t DD = D (6.7-6)
S1
The procedure for calculating the recovery efficiency is similar to that given above.
1. Calculate the cumulative injection at breakthrough t Do . This is equal to the

product of the breakthrough values of ED and EA.
2. Pick some tD after breakthrough t Do .
3. Select a trial ED , and calculate S1 from Eq. (6.7-6).
4. Calculate the areal sweep efficiency from EA = EA(tD/ S1 ).
5. Calculate the pseudodisplacement sweep from ED = ED ( tD/EA).
222
If the ED agrees with that in step 3, the recovery efficiency is the product of ED and
EA. If ED does not agree, return to step 3 with a new trial value.
Combining Vertical, Areal, and Displacement

Sweep Efficiencies
If all three efficiencies are independently available, the above procedure may easily
be generalized as
⎛ t ⎞ ⎛ t ⎞ ⎛t ⎞
ER (t D ) = E A ⎜ D ⎟ EI ⎜ D ⎟ ED ⎜ D ⎟ (6.7-7)
⎝ EI S1 ⎠ ⎝ E A S1 ⎠ ⎝ EV ⎠
The procedure now requires a two-level trial and error, which is equivalent to first
combining EA and EI and then combining EV and ED. The final result in Eq. (6.7-7) is
independent of the order the combinations are carried out in.
As a conclusion to this section, we remind you of the limitations inherent in
these procedures. First, we must have independently specified functions of EA, EI,
and ED, and these functions must be independent of explicit rate dependence. If a rate
dependence is present, the function EI will depend on the particular pathline it was
evaluated on. Perhaps we could evaluate on a pathline having a fluid velocity
representative of the entire pattern (this is commonly tried), but there is considerable
uncertainty about what this representative value is even in the most well-defined
displacements. Recall that, particularly in the VE approaches, the dependence of the
sweep efficiencies on rate may not be particularly evident (for example, the Dietz
theory is strongly rate dependent, but this is not evident from the general VE
approach when capillary pressure becomes small). Further recall that independent
specifications of each of the three efficiencies are available through relatively
idealized calculations (see Secs. 6-4 through 6-6) for extremes in certain physical
properties or through physical models. When any of the above conditions are
seriously violated––and their violation significantly affects the results––one must
resort to numerical simulation, from which the oil recovery could be directly
calculated.
A second more subtle, and perhaps more serious, limitation of the combined
sweep efficiency approach deals with scaling. Scaling simply means any of the sweep
efficiencies, however determined, must themselves be adjusted for the considerably
different scale between the laboratory experiment or analytical calculation and the
field application. For example, few of the independent determinations of ED or EI
account for the nonuniformities surely present in a field displacement.
A classical example of this scale effect involves applying ED to a viscously
unstable field-scale displacement. Much theoretical and experimental work has gone
into showing that the size of the instabilities formed, and indeed, whether they
propagate or not, is a function of the characteristic lengths of the laboratory
experiment or calculation. Thus unless the scaling is such that both effects are the
same in the laboratory and in the field (a sometimes impossible task), the lab-derived
ED will be too optimistic. We cover the subject of viscous instabilities in the next
section.
223
6-8 INSTABILITY PHENOMENA
No EOR process is free from some sort of instability. Hence substantial effort has
gone into minimizing or preventing instabilities (using polymer to drive surfactants
and alkaline agents, or foaming agents to drive CO2 and steam) and into predicting
the oil recovery if fingering is inevitable. We discuss predicting the results of a
fingering process in Chap. 7 in connection with solvent flooding where instability
phenomena have received the most attention. In this section, we deal with the
formation of fingers.
We use the term fingering to describe the bypassing of a resident fluid by a
displacing agent in a homogeneous, nonuniform medium. The actual bypassing
region is a finger. This definition encompasses instabilities caused by both viscous
forces (viscous fingers) and gravity forces (gravity fingers) but does not include
bypassing by permeability heterogeneities. This definition is a little more rigid than
that used in the literature, but we believe the inherent distinction is useful because
fingering can be prevented from displacements, whereas bypassing caused by
heterogeneities cannot (though it can be reduced). In this section, we deal with
isothermal flows; in Chap. 11, we discuss the stability of a nonisothermal
displacement.
A Necessary Condition for Stability
In keeping with the notion that fingering is a general phenomenon, consider the
incompressible, dissipation-free displacement of fluid 2 by fluid 1 in a dipping
reservoir, as shown in Fig. 6-18. This figure is a cross section of a displacement, but
fingering can occur in either the vertical or areal sense. There is no z-direction
communication in this problem. We also consider a perturbation of length ε of the
displacement front (caused, perhaps, by an isolated nonuniformity in the permeability
field), and strive to determine the conditions under which ε(t) will grow or decay as a
function of time. The actual fingering phenomenon is, of course, much more random
and chaotic than that shown in Fig. 6-18, as evidenced by an areal view of a fingering
displacement in a quarter five-spot model shown in Fig. 6-19. Nevertheless, the
Figure 6-18 Viscous fingering

schematic
224
Figure 6-19 Viscous fingering in a

quarter five-spot model, Mo = 17 (from
Habermann, 1960)
simple geometry of Fig. 6-18 is tractable to mathematical analysis and yields insights
into the more complex situations.
To solve for the conditions ε will grow or decay under, we proceed by the
moving boundary technique discussed by Collins (1976). In the region behind the
displacing fluid front, x < xf, the conservation of fluid 1 gives
∂u xj
=0 (6.8-1)
∂x
where j = 1 for x < xf, and j = 2 for x > xf. The accumulation terms in both equations
are zero since there is no change in concentration in the respective regions. For the
same reason, when we substitute Darcy’s law into these equations, they become
∂ ⎛ ∂Pj ⎞
⎜ + ρ j g sin α ⎟ = 0, j = 1 or 2 (6.8-2)
∂x ⎝ ∂x ⎠
The solutions to Eq. (6.8-2) will be of the form

Pj = ( s j − ρ j g sin α ) x + b j , j = 1 or 2 (6.8-3)
where aj and bj are integration constants to be determined with appropriate boundary

conditions. If P0 and PL are the pressures at the reservoir inlet and outlet,
respectively, then bj can be determined as
b1 = P0 (6.8-4a)
b2 = PL − (a2 − ρ 2 g sin α ) L (6.8-4b)
225
Using these relations, and requiring continuous x velocities across the front
u x1 |x = x f = u x 2 |x = x f = u x (6.8-5)
gives, once again using Darcy’s law,
M 0 a1 = a2 (6.8-6)
Equation (6.8-6) determines a1, for we must have continuity of pressure at xf in the
absence of capillary pressure
P1 |x f = P2 |x f (6.8-7)
Inserting Eq. (6.8-3) into Eq. (6.8-7) and using Eqs. (6.8-4) and (6.8-6), yields
−ΔP + ρ 2 g sin α ( L − x f ) + ρ1 g sin α x f
a1 = (6.8-8)
M 0 L + (1 − M 0 ) x f
where ΔP = P0 – PL is the overall pressure drop. The rate of frontal advance is from
Darcy’s law
dx f u x |x f k λr1 ΔP + g sin α [Δρ ( L − x f ) − ρ1 L]
= = (6.8-9)
dt φΔS φΔS M 0 L + (1 − M 0 ) x f
Equation (6.8-9) applies to any point on the displacement front. We could have
equally well developed an expression for a point on the perturbation front
d ( x f + ε ) k λr1 ΔP + g sin α [Δρ ( L − x f − ε ) − ρ1 L]
= (6.8-10)
dt φΔS M 0 L + (1 − M 0 )( x f + ε )
Equation (6.8-10) is identical to Eq. (6.8-9) except xf + ε has replaced xf everywhere.

The rate of change of the perturbation is
d ε d ( x f + ε ) dx f
= − =ε (6.8-11)
dt dt dt
which yields, when Eqs. (6.8-9) and (6.8-10) are substituted,

k λ ΔP (1 − M 0 ) + Lg Δρ sin α − Lg ρ1 (1 − M 0 )sin α
ε = − r1 ε (6.8-12)
φΔS [ M 0 L + (1 − M 0 ) x f ]2
In Eq. (6.8-12), we have assumed ε << xf with the corresponding simplification.

Equation (6.8-12) could be integrated, but for our purpose it is sufficient to
investigate only the sign of ε . The perturbation will grow if ε > 0, will remain
constant if ε = 0, and will decay if ε < 0. From the equality of these three choices,
we find the condition of neutral stability as
LΔρ g sin α
−(ΔP)c = − Lg ρ1 sin α (6.8-13)
1− M 0
where (ΔP)c is a critical pressure drop. The superficial velocity corresponding to this
is the critical rate uc
226
⎡ −(ΔP ) ⎤ k λ Δρ g sin α
o
uc ≡ −k λro1 ⎢ + ρ1 g sin α ⎥ = r1
(6.8-14)
⎣ L ⎦ M 0 −1
Using the critical rate, the conditions for finger growth may be restated
⎧⎪ > uc (unstable)
u x ⎨ = uc (neutral) (6.8-15)
⎪⎩< uc (stable)
where we have also used Darcy’s law to express ux in these inequalities.

Note the similarity between Eq. (6.8-14) and Eq. (6.6-10b), the
corresponding critical rate for gravity tonguing. Analogous expressions can be
worked out for almost any segregated flow conditions, so this similarity should not
be regarded as merely fortuitous. But the differences in the two flows should be kept
in mind. The critical rate in Eq. (6.8-14) is based on an unstable displacement in a
reservoir having no z-direction communication; that in Eq. (6.6-10b) is the
consequence of a VE displacement in a reservoir with very good communication.
To further investigate the stability issue, let’s write the condition for stability
(finger decay) as
( M 0 − 1)u x < k λro1Δρ g sin α (6.8-16)
The superficial velocity ux in this inequality is always positive, but the density
difference can be negative (less dense fluid displacing more dense), as can the dip
angle (displacing down dip). Of course, M0 can take on only positive values though
over quite a large range. Table 6-2 shows typical signs of M0 and Δρ for various EOR
processes. Immediately it follows from Eq. (6.8-16) that the condition for stability in
a horizontal reservoir is simply M0 < 1. This condition is used universally throughout
the EOR literature to describe a stable displacement, particularly in laboratory floods,
though the more general Eq. (6.8-16) is actually the most appropriate form (Hill,
1952).
TABLE 6-2 TYPICAL VALUES FOR MIBILITY

RATIOS AND DENSITY DIFFERENCES
BY PROCESS TYPE
M0 < 1 M0 > 1
Waterflood Waterflood
Δρ > 0 Polymer flood Polymer flood
Micellar polymer
Δρ < 0 Foam Steam
Considering the signs possible for α and Δρ, we can divide the stability
possibilities into four cases, as Table 6-3 shows. Case 1 is unconditionally stable
regardless of the values of Δρg sin α and M0 as Δρg sin α is positive, and M0 < 1.
Similarly, if Δρg sin α < 0 and M0 > 1, case 4, the displacement is unconditionally
227
TABLE 6-3 POSSIBLE CASES FOR A STABLE
DISPLACEMENT
Case
1 M0 < 1 Δρg sin α > 0 Stable

2 M0 > 1 Δρg sin α > 0 Conditionally stable (type I)
3 M0 < 1 Δρg sin α < 0 Conditionally stable (type II)
4 M0 > 1 Δρg sin α < 0 Unstablea
a
Infinite lateral boundaries
0 o
Note: Write stability criterion as ( M − 1)u x < k λ1 Δρ g sin α . For
α = 0 (no dip), the stability criterion becomes M0 < 1.
unstable. The more interesting cases are 2 and 3, which we call type I and type II
conditional stability.
For type I stability, if we divide through Eq. (6.8-16) by the positive quantity
(M0 – 1), the stability criterion is written for ux as in Fig. 6-20. The criterion is an
upper bound for ux and a plot of sweep efficiency (vertical, areal, or volumetric)
versus the dimensionless rate uD
u ( M 0 − 1)
uD = xo (6.8-17)
k λr1Δρ g sin α
shows that EV remains essentially constant until uD = 1 and then decreases thereafter.
Since increasing the displacement velocity causes the instability to form, we see that
viscous forces destabilize the displacement (uD > 1), whereas gravity forces tend to
stabilize the displacement (uD < 1). The resulting instability is a viscous instability or
finger. For type II conditional stability, a similar plot (Fig. 6-21) shows sweep
efficiency decreasing for decreasing uD, beginning a precipitous decline at uD = 1.
This is because the stability criterion is now a lower bound since (M0 – 1) is now
negative. For type II conditional stability, viscous forces stabilize the displacement,
Figure 6-20 Type I conditional

stability
228
Figure 6-21 Type II conditional

stability
and gravity forces destabilize. The resulting instabilty is a gravity instability.
For certain values of the parameters, then, both types of displacements are or
can be made stable. The conditional stability is most useful in determining a
maximum rate in a dipping displacement where M0 > 1. But usually, this rate is
below that required for economic oil production. For type II stability, a larger rate is
required, but in practice, this situation is not commonly encountered.
Critical Wavelength
Whereas ux < uc is a necessary and sufficient condition for stability, the condition ux >
uc is, unfortunately, a necessary condition only for instability. This condition is
because dissipative effects in flows in media of limited lateral extent tend to suppress
fingering. This effect means fingering may be abnormally suppressed in laboratory
displacements compared to the same displacement under field conditions. One may
legitimately wonder, then, about the purpose of doing laboratory experiments on
unstable displacements when this scale effect is not considered.
To investigate this scale effect, we reproduce an argument based on linear
stability analysis originally given by Chouke et al. (1959) and then by Gardner and
Ypma (1982).
Based on a linear stability analysis of a downward secondary miscible
displacement of oil by a less viscous and less dense solvent in a homogeneous,
uniform medium, the critical wavelength λc of an unstable miscible displacement is
M 0 + 1 ⎛ Kl ⎞
λc = 4π ⎜ ⎟ (6.8-18)
M 0 − 1 ⎝ u x − uc ⎠
229
where the dispersion coefficient Kl is taken to be isotropic. Since the displacement is
unstable, we must have M0 > 1 and ux > uc so that λc is always positive.
The analogous expression for an initially sharp immiscible displacement was
also determined by Chouke et al. (1959) and reproduced in greater detail by Peters
(1979)
1/ 2
C⎡ k λro1σ 12 ⎤
λc = ⎢ ⎥ (6.8-19)
3 ⎣ ( M − 1)(u x − uc ) ⎦
0
The constant C in Eq. (6.8-19) is called Chouke’s constant by Peters, who also
determined values C = 25 for immiscible displacements with no residual water
initially present, and C = 190 with irreducible water present. Clearly, the critical
wavelength is greater with irreducible water initially present, but the reason for this
stabilizing effect is not well understood.
The necessary and sufficient conditions for a type I instability to form based
on this analysis are now
M 0 > 1 or u x > uc and λc < ( H t ) max (6.8-20)
where (Ht)max is the maximum lateral extent of the permeable medium. One may
readily show (see Exercise 61) that λc is of the order of a few centimeters for typical
conditions. Thus, if fingering is desired in a displacement, one must take special
precautions that conditions (Eq. 6.8-20) are met. This usually means running
displacements at excessively high rates, compared to field rates, or in systems having
at least one large transverse dimension. Such a system is the Hele Shaw cell, in which
the displacement of Fig. 6-19 is occurring. But if the intent is to suppress fingering,
systems having very small transverse dimensions, such as the slim tube experiments
we discuss in Chap. 7, are preferable.
Three things are important about both the derivation of critical velocity and
wavelength. First, neither says anything about how fingers propagate once they are
formed. A finger forms, bifurcates into two branches, one of these dominates (or
shields) the other, and the dominant one then bifurcates again to repeat the process
(Homsy, 1987). If continued indefinitely, a single finger with numerous appendages
representing the bifurcations will result. Figure 6-19 suggests the bifurcation through
the various levels of fingers each superimposed on the next larger scale. The smallest
scale corresponds to the critical wavelength.
Second, both the critical wavelength and velocity derivations depended on
the perturbation being small. It is impossible to say from this what the response to a
large perturbation would be, and we can be assured that such large perturbations do
exist. Thus Eq. (6.8-20) should also be regarded as only necessary conditions.
Finally, the issues of fingering and heterogeneity cannot be rigorously
separated. After all, heterogeneity caused the perturbation in Fig. 6-16 even though
we proceeded as though the reservoir was homogeneous. The merging of the
fingering and heterogeneity issues is one of the most interesting topics in EOR
research; in Chap. 7, we discuss some primitive attempts at this merging.
230
6-9 SUMMARY
That volumetric sweep efficiency is a complex issue accounts for the scarcity of
treatment in this text compared to displacement efficiency. Three factors account for
this complexity: a strong dependency on operational issues, nonlinear and irregular
geometries, and the difficulty in capturing realistic heterogeneities. Numerical
simulators can handle all three of these issues to some extent even though some
questions remain about how to represent heterogeneity in simulation models.
There is little in the behavior of the volumetric sweep efficiency of actual
reservoir displacements that cannot be at least qualitatively understood through the
material we present here. Examples of such behavior are reservoirs with high-
permeability thief zones that behave essentially as a two-layer medium, generally
high-permeability reservoirs dominated by gravity that conform well to the Dietz
theory, low-permeability reservoirs in which crossflow tends to be unimportant, and
high-permeability reservoirs with large well spacing that tend to the VE limit rather
quickly.
Above all, the recognition of bypassing––through channeling, viscous
fingering, gravity segregation, or some combination of these––is important, for this
seems to occur in a good many waterfloods and EOR projects.
EXERCISES
6A. Using Areal Sweep Correlations. Use the areal sweep efficiency correlations for a
confined five-spot in this exercise.
(a) Plot areal sweep efficiency EA versus dimensionless time tD for a mobility ratio of
6.5.
(b) If the pattern pore volume is 106 m3, and the average injection rate is 500 m3/D,
plot cumulative oil recovery (SCM) versus time (months or years). Assume the
displacement is pistonlike, vertical sweep is 1, and the pore volume given above is
movable. The residual water and oil saturations are 0.2 and 0.3, respectively.
6B. Heterogeneity Measures of Normal Distributions. As Table 6-1 shows, permeability
often is distributed normally rather than lognormally. When this happens, the
cumulative frequency distribution function (Eq. 6.3-5) becomes
⎧
1⎪ ⎡ r−r ⎤⎫⎪
Λ= ⎨1 − erf ⎢ ⎥⎬ (6B-1)
2⎪
⎩ ⎣ 2v N ⎦⎭⎪
where r is the average permeability–porosity ratio, and vN is the variance of the

normal distribution. Using Eqs. (6B-1), (6.3-3), and (6.3-4), derive formulas for the
Lorenz and Dykstra-Parsons coefficients in terms of vN.
6C. Vertical Sweep Efficiency in a Two-Layer Reservoir
(a) Derive Eq. (6.4-4) for flow in layer l in a horizontal reservoir.
(b) Calculate and plot the vertical sweep efficiency EI and the fraction of total flow
going into the high velocity layer for a two-layer horizontal reservoir with k1 = 2k2,
φ1 = φ2, ΔS1 = ΔS2, and h1 = 3h2. Take M0 = 0.5.
231
6D. Vertical Sweep Efficiency in a Noncommunicating Reservoir. For a reservoir having no
vertical communication, calculate and plot the vertical sweep efficiency versus
dimensionless cumulative water injected for the following five-layer cross section:
hl(m) φl kl(μm2)
5 0.2 0.100
10 0.22 0.195
2 0.23 0.560
15 0.19 0.055
4 0.15 0.023
The endpoint mobility ratio is 0.5.

6E. Vertical Equilibrium for Continuous Layers. For a reservoir for which the VE Hearn
model applies with M0 < 1 and α = 0,
(a) Show that if the permeability distribution is continuous, the cross-sectional
averaged water fractional flow may be written as
−1
⎛ (1 − C ) ⎞
f1 = ⎜ 1 + 0 ⎟ (6E-1)
⎝ HK M C ⎠
where HK is the Koval heterogeneity factor (Fig. 6-8).

(b) Recalculate and plot the vertical sweep efficiency for the two-layer model of part
(b) in Exercise 6C. Use M0 = 0.5.
(c) In a two-layer horizontal reservoir, show that the effects of the heterogeneity
contrast may be completely suppressed (that is, the fronts travel at equal velocities
in both layers) if
0 k φ
M < 2 1 (6E-2)
φ 2 k1
where 1 and 2 represent the high and low velocity layers, respectively.
6F. Calculating Pseudorelative Permeabilities. For the discrete permeability–porosity data
of Exercise 6D,
(a) Calculate and plot the pseudorelative permeabilites for a waterflood in a horizontal
reservoir using the VE Hearn model.
(b) Calculate and plot the vertical sweep efficiency for this flood.
(c) Repeat part (a) for a nonzero capillary pressure function given by
1/ 2
⎛ φ ⎞ cos θ (1 − S ) 4
Pc = σ 12 ⎜ ⎟ (6F-1)
⎝k⎠
where σ12 is the oil–water interfacial tension, θ is the contact angle, and
S1 − S1r
S= (6F-2)
1 − S1r − S 2 r
(d) Calculate and plot the vertical sweep efficiency for part (c).
Additional data for this problem are Δρ = 0, S1r = S2r = 0.2, μ1 = 1 mPa-s, μ2 = 10
232
o o
mPa-s, k r1
= 0.05, k r2
= 0.9, and the relative permeability curves are given as
o 2 o
k r1 = k r1 S , k r 2 = k r 2 (1 − S ) (6F-3)
6G. Deriving Pseudorelative Permeabilities. The water–oil capillary-pressure–water-

saturation function often may be represented as
⎛ 1 ⎞
Pc = K ⎜ 2 − 1 ⎟ (6G-1)
⎝S ⎠
where K is a constant, and S is the reduced saturation (Eq. 6F-2). If the VE assumptions
apply and the reservoir is homogeneous,
(a) Derive the water saturation profile in the dip normal or z direction in terms of a
water saturation at the bottom of the reservoir (S1B or SB).
(b) Derive an expression for the average water saturation as a function of S1B or SB.
(c) If the local (laboratory-measured) relative permeabilities are approximated by Eq.
(6F-3), show that the oil and water pseudorelative permeabilities expressed in
terms of the average saturation of part (b) are
⎧ o 2
Ng S ⎫
k
o
⎪1 + 2 ⎪ o
kr1 = ln ⎨ ⎛
r1
o Ng S ⎞ ⎬,
o 2
k r 2 = k r 2 (1 − S )
N ⎪ ⎜1 − 4 ⎟ ⎪
g
⎩ ⎝ ⎠ ⎭
where
o Δρ g cos α H t
Ng =
K
o
(d) For N g = 1 and M0 = 4, calculate and plot the pseudodisplacement sweep
efficiency versus dimensionless time. The dip angle of the reservoir is zero.
6H. Combining Sweep Efficiencies. The vertical sweep efficiency curve for a pistonlike
displacement is shown in Fig. 6H. Combine this curve with the areal sweep efficiency
curve of Exercise 6A to give the volumetric sweep efficiency curve.
6L. Viscous Fingering Calculations
(a) Calculate the critical rate for a miscible displacement having the following
properties:
k = 0.12 μm2
M0 = 50
Oil-solvent density difference = – 0.8 g/cm3
Solvent mobility = 10 (mPa-s)–1
Dip angle = –10°
(b) If the superficial velocity in the above displacement is 0.8 μm/s, calculate the
critical wavelength from stability theory. Take the dispersion coefficient to be 10–5
cm2/s.
233
Figure 6H Vertical sweep efficiency function for Exercise 6H

234
Solvent Methods
One of the earliest methods for producing additional oil is through the use of solvents
to extract the oil from the permeable media. In the early 1960s, interest centered on
injecting liquified petroleum gas (LPG) in small “slugs” and then displacing the LPG
by a dry “chase” gas. This process became economically less attractive as the value
of the solvent increased. In the late 1970s, interest in solvent methods resurged
because of an increased oil price and more confidence in the ability to estimate oil
recovery. During this period, the leading solvent became carbon dioxide though
several other fluids were used also (Stalkup, 1985).
Two fluids that mix together in all proportions within a single-fluid phase are
miscible. Therefore, miscible agents would mix in all proportions with the oil to be
displaced. But most practical miscible agents exhibit only partial miscibility toward
the crude oil itself, so we use the term solvent flooding in this text. Many solvents, of
course, will become miscible with crude under the right conditions, but all solvents of
commercial interest are immiscible to an aqueous phase.
Solvent flooding refers to those EOR techniques whose main oil recovering
function is because of extraction, dissolution, vaporization, solubilization, conden-
sation, or some other phase behavior change involving the crude. These methods
have other, sometimes very important, oil recovery mechanisms (viscosity reduction,
oil swelling, solution gas drive), but the primary mechanism must be extraction.
This oil extraction can be brought about by many fluids: organic alcohols,
ketones, refined hydrocarbons, condensed petroleum gas (LPG), natural gas and
liquified natural gas (LNG), carbon dioxide, air, nitrogen, exhaust gas, flue gas, and
235
others. In this chapter, we emphasize miscible flooding with gaseous solvents CO2,
CH4, and N2, but you should remember there are many potential agents.
7-1 GENERAL DISCUSSION OF SOLVENT FLOODING
Considering the wide variety of solvents, process types, and reservoirs, our
discussion must ignore one or more interesting variations. Thus in this section, we
discuss CO2 solvent flooding, and later, we indicate more general aspects of solvent
flooding.
Figure 7-1 shows an idealized vertical cross section between an injection and
production well. By far the most common application of solvent methods is in a
displacement mode as shown, but injection and production through the same wells
have been reported (Monger and Coma, 1986). Solvent injection commences into a
reservoir in some stage of depletion, most commonly at residual oil or true tertiary
conditions. Most solvent floods are in reservoirs containing light crudes (less than 3
mPa-s oil viscosity) though there are exceptions (Goodrich, 1980). The solvent may
be introduced continuously in undiluted form, alternated with water in the
water-alternating-gas (WAG) process as in Fig. 7-1, or even injected simultaneously
with water through paired injection wells. Water is injected with the solvent in this
fashion to reduce the usually unfavorable mobility ratio between the solvent and the
oil. Carbon dioxide, in particular, can be injected dissolved in water in a distinctly
immiscible fashion that recovers oil through swelling and viscosity reduction
(Martin, 1951).
If the solvent is completely (first-contact) miscible with the oil, the process
has a very high ultimate displacement efficiency since there can be no residual phases
(see Sec. 5-4). If the solvent is only partially miscible with the crude, the total
composition in the mixing zone (miscible zone in Fig. 7-1) between the solvent and
the oil can change to generate or develop miscibility in situ. Regardless of whether
the displacement is developed or first-contact miscible, the solvent must immiscibly
displace any mobile water present with the resident fluids.
The economics of the process usually dictates that the solvent cannot be
injected indefinitely. Therefore, a finite amount or slug of solvent is usually followed
by a chase fluid whose function is to drive the solvent toward the production wells.
This chase fluid––N2, air, water, and dry natural gas seem to be the most common
choices––may not itself be a good solvent. But it is selected to be compatible with the
solvent and because it is available in large quantities. The similarity between the
chase fluid in solvent flooding and the mobility buffer drive in micellar-polymer
flooding is evident in Figs. 7-1 and 9-1.
Though the process shown in Fig. 7-1 appears relatively simple, the
displacement efficiency and volumetric sweep efficiency are quite complex. In Secs.
7-6 and 7-8, we apply the theory of Chaps. 5 and 6 to solvent flooding, but first we
must discuss selected physical properties of solvents and solvent–crude oil systems.
236
Figure 7-1 Schematic of a solvent flooding process (drawing by Joe Lindley, U.S.
Department of Energy, Bartlesville, Okla.)
237
7-2 SOLVENT PROPERTIES
Figure 7-2 shows phase behavior data (P-T diagram) for various pure components
and air. For each curve, the line connecting the triple and critical points is the vapor
pressure curve; the extension below the triple point is the sublimation curve (see Sec.
4-1). The fusion curve is not shown. The pressure–temperature plot for air is really an
envelope, but its molecular weight distribution is so narrow that it appears as a line in
Fig. 7-2. Flue gas is also a mixture of nitrogen, carbon monoxide, and carbon dioxide
with a similarly narrow molecular weight distribution; its P-T curve would fall near
the nitrogen curve in Fig. 7-2.
The critical pressures for most components fall within a relatively narrow
range of 3.4–6.8 MPa (500–1,000 psia), but critical temperatures vary over a much
wider range. The critical temperatures of most components increase with increasing
molecular weight. Carbon dioxide (molecular weight, MW = 44) is an exception to
this trend with a critical temperature of 304 K (87.8°F), which is closer to the critical
temperature of ethane (MW = 30) than to propane (MW = 44). (See Vukalovich and
Altunin (1968) for a massive compilation of CO2 properties.) Most reservoir
applications would be in the temperature range of 294–394 K (70–250°F) and at
pressures greater than 6.8 MPa (1,000 psia); hence air, N2, and dry natural gas will all
be supercritical fluids at reservoir conditions. Solvents such as LPG, in the molecular
weight range of butane or heavier, will be liquids. Carbon dioxide will usually be a
supercritical fluid since most reservoir temperatures are above the critical
temperature. The proximity to its critical temperature gives CO2 more liquidlike
properties than the lighter solvents.
Figures 7-3 and 7-4 give compressibilities factors for air and carbon dioxide,
respectively. From these the fluid density ρ3 can be calculated
PM W
ρ3 = (7.2-1)
zRT
The formation volume factor at any temperature and pressure B3, a specific molar
volume, also follows
P T
B3 = z s (7.2-2)
P Ts
In Eq. (7.2-2), Ts and Ps are the standard temperature and pressure, respectively. All
fluids become more liquidlike, at a fixed temperature and pressure, as the molecular
weight increases. The anomalous behavior of CO2 is again manifest by comparing its
density and formation volume factor to that of air. For CO2 at 339 K (150°F) and 17
MPa (2,500 psia), ρ3 = 0.69 g/cm3, and B3 = 2.69 dm3/SCM. The values for air at the
same temperature and pressure are ρ3 = 0.16 g/cm3, and B3 = 7.31 dm3/SCM. The
CO2 density is much closer to a typical light oil density than is the air density; hence
CO2 is much less prone to gravity segregation during a displacement than is air.
Usually, gravity segregation in a CO2 flood is more likely where the water saturation
is high since CO2 tends to segregate more from water than oil.
238
Figure 7-2 Vapor pressure curves for various substances (from Gibbs, 1971)
239
Figure 7-3 Compressibility chart for air (from Gibbs, 1971)

240
Figure 7-4 Compressibility chart for carbon dioxide (CO2) (from Gibbs, 1971)
241
From the formation volume factors, 370 SCM of CO2 is required to fill one
cubic meter of reservoir volume, whereas only 140 SCM air is required at the same
temperature and pressure. Thus about three times as many moles (recall that B3 is a
specific molar volume) of CO2 are required to fill the same reservoir volume as air.
Figures 7-5 and 7-6 give the viscosities of a natural gas mixture and pure
CO2. Over the pressure and temperature range shown, which includes the conditions
of interest in EOR, the viscosities of natural gas, and CH4, air, flue gas, and N2 are
about the same. But the CO2 viscosity is generally two or three times higher. Relative
to a hydrocarbon liquid or water viscosity, the values are still low, so there should be
no appreciable difference in the ease of injection of these solvents. However, the
CO2––crude-oil mobility ratio will be two or three times smaller than the other light
solvents; hence volumetric sweep efficiency will generally be better for CO2. (For the
correlations for other solvents and solvent mixtures, see McCain, 1973; Reid et al.,
1977; and Gas Processors Suppliers Association, 1973.)
Figure 7-5 Viscosity of a natural gas

sample (from Lee et al., 1966)
242
Figure 7-6 Viscosity of carbon dioxide as a function of pressure at various

temperatures (from Goodrich, 1980)
7-3 SOLVENT–CRUDE-OIL PROPERTIES
In Secs. 4-1 and 4-2, we discussed general aspects of phase behavior for pure
components and mixtures. In this section, we give specific features of solvent–crude
phase behavior necessary for the development in later sections.
Figures 7-7 and 7-8 show pressure–composition (P-z) diagrams for two
different solvent–crude systems. Recall that these diagrams are plots, at constant
temperature, of pressure versus the overall mole percent of solvent in contact with a
crude oil. These plots show the number and types of phases and the volume percent
liquid. Figure 7-7 is for the recombined Wasson crude oil at 105°F (314 K), and Fig.
7-8 is for the Weeks Island “S” Sand crude at 225°F (381 K). Other diagrams are
reported elsewhere (Turek et al., 1980; Orr and Jensen, 1982). The data in Figs. 7-7
and 7-8 represent behavior typical of low- and high-temperature systems. Recall that
no water is present during the phase behavior measurements. For mixtures, the mole
percent can represent both phase and overall concentrations.
The P-z diagrams have the same general form regardless of the temperature.
The left vertical axis gives the phase behavior of the CO2-free crude; thus the bubble
point of the recombined Wasson crude at 314 K (105°F) is 6.81 MPa (1,000 psia)
from Fig. 7-7. The right vertical axis similarly gives pure CO2 properties, which will
243
Figure 7-7 P-z diagram for recombined

Wasson crude, CO2 system (from
Gardner et al., 1981)
be a single-phase fluid for Figs. 7-7 and 7-8 since both are above the CO2 critical
temperature. At low pressures and for all CO2 concentrations, except very near the
right axis, the mixture is two-phase liquid and vapor. The liquid volume quality lines
are also shown. At high pressures and low CO2 concentration, the mixture is single
phase. At about 60% CO2, a critical point exists through which pass two single-phase
boundaries. The CO2 composition at this point is the critical composition for the
fixed temperature and indicated pressure. The phase boundary line below the critical
point is a bubble point curve, and that above is a dew point curve. Thus the upper left
corner of the P-z diagram is a supercritical fluid region. The system could form a
liquid phase as the light component increases in concentration at a constant pressure
greater than the critical. This change is a type of retrograde behavior.
244
Figure 7-8 Phase envelope for Weeks

Island “S” Sand crude and 95% CO2, 5%
plant gas at 225°F (from Perry, 1978)
Though few P-z diagrams are for solvents other than CO2 in the literature, it
appears, based on the N2–crude-oil data in Fig 7-9, that the above qualitative
character applies to other solvents as well. The critical pressure for the N2 solvent
mixture in Fig. 7-9 is much larger (off the scale) than either of the critical pressures
of the CO2 systems in Figs. 7-7 and 7-8.
The main difference between the low- and high-temperature phase behavior
is the presence, in Fig. 7-7, of a small three-phase region just below and to the right
of the critical point. These phases are two liquids––a light or upper phase and a heavy
or lower phase––and a vapor phase. Such behavior has generally not been observed
at high temperatures (Fig. 7-8) (Turek et al., 1980). Moreover, at low temperatures, a
small amount of solid precipitate can exist over some composition and pressure
ranges. The precipitate is composed mainly of asphaltenes, the n-heptane insoluble
245
Figure 7-9 Pz diagram for reservoir fluid B-nitrogen system at 164°F (347 K)
(from Hong, 1982)
fraction of crude oil (Hirshberg et al., 1982). The region of precipitate formation may
overlap the three-phase region. This behavior offers a complication to the
displacement process and may even present operational problems since the solid
precipitate can cause formation plugging.
Consider now a displacement of a crude by a pure solvent in a permeable
medium at some time before solvent breakthrough. The conditions at the injection
end of the medium plot on the right vertical axis of the P-z diagram, and those at the
production end plot on the left axis at some lower pressure. Conditions in the medium
246
between these extremes are not represented on the P-z diagram since the relative
amounts of each hydrocarbon component do not remain constant during a
displacement, as they do in the PVT measurements of Figs. 7-7 through 7-9.
Therefore, the diagrams are not particularly useful for displacement classification,
which is based on the ternary diagrams we describe next. Still, one can see
qualitatively from these diagrams that completely miscible displacements––those that
are a single phase for all solvent concentrations––would require high reservoir
pressures, in excess of 66.7 MPa (9,800 psia) for the data in Fig. 7-8.
Ternary diagrams are more useful in classifying solvent floods because they
impart more compositional information than do P-z diagrams. Figures 7-10 through
7-12 show representations of these. On these diagrams the solvent–crude mixture is
represented by three components; a light component on the top apex, an intermediate
crude fraction on the right apex, and a heavy crude fraction on the left apex. The
exact split between intermediate and heavy crude components is immaterial to the
general features of the phase equilibria or to the miscibility classification. In Figs. 7-
10 and 7-11, the split is between the C6 and C7 molecular weight fractions. Therefore,
Figure 7-10 Ternary equilibria for CO2-recombined Wasson crude mixture

(Gardner et al., 1981)
247
Figure 7-11 Ternary equilibria for CO2-recombined Wasson crude system (from
none of the corners of these ternaries are pure components, hence the designation
pseudocomponents. As before, no water is on the diagrams. In addition to those given
here, ternary diagrams are in the literature in several other sources: for alcohol
solvents (Holm and Csaszar, 1965; Taber and Meyer, 1965), for natural gas solvents
(Rowe, 1967), for CO2 (Metcalfe and Yarborough, 1978; Orr et al., 1981; Orr and
Silva, 1982), for N2 solvents (Ahmed et al., 1981), and for mixtures of CO2, SO2, and
CH4 (Sazegh, 1981).
A good example of CO2–crude-oil equilibria is shown in Fig. 7-10 for the
recombined Wasson crude (compare Figs. 7-10 and 7-11 with Fig. 7-8, the P-z
diagram for the same mixture). In these solvent–crude systems, the phase equilibria is
strongly dependent on reservoir temperature and pressure (recall that the ternary is at
constant T and P). Typically, though, the pressure is larger than the cricondenbar of
the light–intermediate component pseudobinary; hence these two components are
miscible in all proportions. The pressure is smaller than that of the light–heavy
binary, and there is a region of limited miscibility or two-phase behavior along the
248
Figure 7-12 Methan–crude oil ternary phase behavior (from Benham et al., 1961)
light–heavy axis. This region of two-phase behavior extends into the interior of the
ternary and is bounded by a binodal curve (see Sec. 4-3). Within the binodal curve,
there are tie lines whose ends represent the composition of the equilibrium phases.
These shrink to a plait point where the properties of the two phases are indistin-
guishable. The plait point is the critical mixture at this temperature and pressure.
Of great importance in what follows is the critical tie line, the fictitious tie
line tangent to the binodal curve at the plait point. The critical tie line is the limiting
case of the actual tie lines as the plait point is approached. As pressure increases, the
two-phase region shrinks––that is, light–heavy miscibility increases. No general
249
statement is possible about the effect of temperature though the two-phase region
generally increases with increasing temperature. For low pressure and low
temperature, a three-phase region can intrude into the two-phase region (Fig. 7-11).
These general characteristics apply for solvents other than CO2 (Fig. 7-12).
The composition of the reservoir crude can be placed on the ternary, as can the
composition of the solvent. In doing this, we are neglecting the pressure change that
is, of course, an essential ingredient in making the fluids flow in the reservoir. Even
with this approximation, all compositions in the solvent crude mixing zone do not lie
on a straight line connecting the initial and injected. This is because the composition
changes are affected by the phase behavior. In fact, these changes are the basis for the
classification of solvent displacements that we give in the next few paragraphs
(Hutchinson and Braun, 1961).
We represent a one-dimensional displacement of a crude by a solvent on the
schematic ternary diagram in Fig. 7-13. The crude is in the interior of the ternary,
indicating some of the light component is present initially in the crude. If a straight-
line dilution path between the solvent and the crude does not intersect the two-phase
Figure 7-13 Schematic of the first-contact miscible process

250
region, the displacement will consist of a single hydrocarbon phase that changes in
composition from crude to undiluted solvent through the solvent oil mixing zone. The
dilution path is linear (see Sec. 7-6) since the only mechanism for mixing is
dispersion, there being no water or fractional flow effects associated with the single
hydrocarbon phase. A displacement that occurs entirely within one hydrocarbon
phase is first-contact miscible. There is a range of solvent compositions that will be
first-contact miscible with the crude at this temperature and pressure.
Suppose the solvent consists entirely of the light component (Fig. 7-14). The
displacement is not first-contact miscible since the dilution path passes through the
two-phase region. Imagine a series of well-mixed cells that represent the permeable
medium in a one-dimensional displacement. The first cell initially contains crude to
which we add an amount of solvent so that the overall composition is given by M1.
The mixture will split into two phases, a gas G1 and a liquid L1, determined by the
equilibrium tie lines. The gas G1 will have a much higher mobility than L1, and this
phase moves preferentially into the second mixing cell to form mixture M2. Liquid L1
remains behind to mix with more pure solvent. In the second cell mixture, M2 splits
into gas G2 and liquid L2, G2 flows into the third cell to form mixture M3, and so
Figure 7-14 Schematic of the vaporizing gas drive process (adapted from
Stalkup, 1983)
251
forth. At some cell beyond the third (for this diagram), the gas phase will no longer
form two phases on mixing with the crude. From this point forward, all compositions
in the displacement will be on a straight dilution path between the crude and a point
tangent to the binodal curve. The displacement will be first-contact miscible with a
solvent composition given by the point of tangency. The process has developed
miscibility since the solvent has been enriched in intermediate components to be
miscible with the crude. Since the intermediate components are vaporized from the
crude, the process is a vaporizing gas drive. Miscibility will develop in this process
as long as the injected solvent and crude are on opposite sides of the critical tie line.
Suppose the crude and solvent compositions are again on opposite sides of
the critical tie line but reversed from the vaporizing gas drive (Fig. 7-15). In the first
mixing cell, the overall composition M1 splits into gas G1 and liquid L1. Gas G1
moves on to the next mixing cell as before, and liquid L1 mixes with fresh solvent to
form mixture M2. Liquid L2 mixes with fresh solvent, and so forth. Thus in the first
mixing cell, this mixing process will ultimately result in a single-phase mixture.
Figure 7-15 Schematic of the rich-gas drive process (adapted from Stalkup,
1983)
252
Since the gas phase has already passed through the first cell, the miscibility now
develops at the rear of the solvent–crude mixing zone as a consequence of the
enrichment of the liquid phase in intermediate components there. The front of the
mixing zone is a region of immiscible flow owing to the continual contacting of the
gas phases G1, G2, and so on, with the crude (this is also true at the rear of the mixing
zone in the vaporizing gas drive). The process in Fig. 7-15 is the rich gas drive
process since intermediates were added to enrich the injected solvent. Since these
intermediates condense into the liquid phase, the process is sometimes called a
condensing gas drive. Figure 7-12 shows that more than twelve contacts are
necessary to develop miscibility in an actual system.
Figure 7-16 shows a schematic of an immiscible displacement. The crude
and solvent are in single-phase regions, but both are on the two-phase side of the
critical tie line. Now the initial mixture M1 in the first mixing cell will form gas G1,
which will flow forward to form mixture M2, and so forth. This gas is being enriched
in intermediate components at the leading edge (forward contacts) of the solvent–
crude mixing zone as in a vaporizing gas drive. But the enrichment cannot proceed
beyond the gas-phase composition given by the tie line whose extension passes
Figure 7-16 Schematic of an immiscible displacement

253
through the composition crude. At the forward contacts, there will be an immiscible
displacement of the crude by a mixture on the limiting tie line. Back at the first
mixing cell, liquid L1 mixes with solvent to form mixture M–1, just as in the
condensing gas drive. The displacement is immiscible here since a single-phase
solvent is displacing a two-phase mixture. The liquid phase becomes progressively
stripped of intermediates (L–1, L–2 and so on) until it reaches another limiting tie line.
The displacement is entirely immiscible, then, at both the forward and reverse
contacts. The intermediate components are in a gas phase near the production end of
the permeable medium, and in a liquid phase at the injection end. An immiscible
flood entirely devoid of injected intermediates is a dry gas flood.
Figure 7-17 summarizes the classification of solvent displacements. A
dilution path (I2-J3) that does not pass through the two-phase region is a first-contact
miscible displacement. A dilution path entirely on the two-phase side of the critical
tie line constitutes immiscible displacement (I1-J1). When initial and injected
compositions are on opposite sides of the critical tie line, the displacement is either a
Figure 7-17 Summary of miscibility and developed miscibility

254
vaporizing gas drive (I2-J1) or a condensing gas drive (I1-J2). The last two cases are
developed or multiple-contact miscible displacement.
At the conditions shown in Fig. 7-10, CO2 displaces oil as a vaporizing gas
drive. At comparable conditions (Figs. 7-12 and 7-18), N2 and CH4 are usually an
immiscible solvent. The CH4 in Fig. 7-12 can be converted into a condensing gas
drive by adding about 35 mole % intermediates.
The solvent flooding classifications given here are corroborated by simple
wave theory (see Sec. 7-7) and experimental results. Figure 7-19 shows the effluent
history of three CO2 floods in a Berea core. The oil in this displacement was a
mixture of 25 mole % C1, 30 mole % C4, and 45 mole % C10. The three runs were at
10.2 MPa (1,500 psia) (run 4), an immiscible displacement; 12.9 MPa (1,900 psia)
(run 5), a first-contact miscible displacement; and 11.6 MPa (1,700 psia) (run 6), a
vaporizing gas drive. The temperature was 344 K (160°F) for all runs.
The effluent histories in Fig. 7-19 are plots of C1, C4, and C10 concentrations,
normalized by their initial values, versus the hydrocarbon pore volumes (HCPV) of
CO2 injected (see Table 5-1). If the dilution path between the oil and the solvent were
a straight line, the normalized concentration of all displaced components would be
identical. They are identical in the first-contact miscible run 5. The vaporizing gas
drive run 6 shows that the normalized concentrations of the heavy component C10
declines slightly before the C4 curve declines (Fig. 7-19b). (The run 6 composition at
Figure 7-18 Ternary equilibria for N2–crude-oil mixture (from Ahmed et al.,
1981)
255
Figure 7-19 Effluent histories from laboratory displacement run 4 immiscible, run 5 first-
contact miscible, run 6 multiple-contact miscible (from Metcalfe and Yarborough, 1978)
0.9 HCPV is relatively rich in C4.) In addition, the light components C1 go through a
maximum at about the same point (Fig. 7-19a). The C1 maximum is even more
pronounced in the immiscible displacement probably because the fluids can now be
saturated with respect to C1. A similar effect should occur in a condensing gas drive
process though the enrichment will now occur at the rear of the mixing zone.
The immiscible displacement and vaporizing gas drive process are similar;
however, the oil recovery (displacement efficiency) in the immiscible run (80%) was
considerably smaller than that for either the first-contact (97%) or the vaporizing gas
drive run (90%). Developed miscibility displacements can give oil recoveries
approaching first-contact miscible displacements; immiscible processes are usually
much poorer.
Immiscible displacements have merit since pressure requirements are not
large, the solvents are usually less expensive, and they can recover some oil. The
principal recovery mechanisms for immiscible solvents are (1) a limited amount of
vaporization and extraction, (2) oil viscosity reduction, (3) oil swelling, (4) solution
gas drive during pressure decline, and (5) interfacial tension lowering. All immiscible
256
Figure 7-20 Solubility (mole fraction) of carbon dioxide in oils as a function of UOP number
(from Simon and Graue, 1965)
257
displacements recover oil in this manner though the data showing these effects are
most complete on CO2-immiscible displacements (Simon and Graue, 1965).
Figures 7-20 through 7-22 show experimental data that emphasize
immiscible recovery mechanisms 1–3. Figure 7-20(a) shows the solubility of CO2 in
oil versus temperature and saturation pressure for a crude with a Universal Oil
Figure 7-21 Viscosity correlation charts for carbon-dioxide–oil mixtures (from

Simon and Graue, 1965)
258
Figure 7-22 Swelling of oil as a function of mole fraction of dissolved carbon

dioxide (from Simon and Graue, 1965)
Characterization factor (K) of 11.7. This factor is the ratio of the cube root of the
average boiling point in degrees R to the specific gravity. It can be related to API
gravity and viscosity (Watson et al., 1935). The saturation pressure is the bubble
point pressure; hence Fig. 7-20(a) is giving the maximum solubility of CO2 at the
indicated temperature and pressure. Figure 7-20(b) corrects the solubility data to
other characterization factors. Figure 7-21 gives the viscosity ratio of a CO2-swollen
crude (µm in this figure) to the CO2-free crude (µ0) as a function of pressure. For
259
moderate saturation pressures, the viscosity reduction is pronounced, particularly for
large crude viscosities.
Figure 7-22 illustrates the oil swelling mechanism by giving crude swelling
factors correlated with ratios of molecular weight to standard density (g/cm3). Similar
data on the swelling of crude by N2 are given by Vogel and Yarborough (1980).
Figures 7-20 through 7-22 are complementary. Let’s estimate the CO2
solubility, oil viscosity reduction, and swelling factor for a crude oil at 389 K (150°F)
and 8.2 MPa (1,200 psia). Recall that we are calculating the properties of a liquid
hydrocarbon phase immiscible with CO2. Therefore, the overall CO2 mole fraction
must be large enough to be in the two-phase region of the ternary diagram. The
relevant physical properties of the crude are as follows: molecular weight = 130,
UOP characterization factor K = 11.8, specific gravity = 0.70, normal boiling point =
311 K (100°F), and viscosity = 5 mPa-s. This gives a CO2 solubility of 55 mole %
from Fig. 7-20. This solubility causes the oil viscosity to decrease to 1 mPa-s from
Fig. 7-21, and the oil to swell by about 33% from Fig. 7-22. (For additional data on
the properties of crude containing immiscible solvents, see Holm, 1961; de Nevers,
1964; Holm and Josendal, 1974; and Tumasyn et al., 1969.)
Figure 7-23 Solubility of carbon dioxide in water (from Crawford et al., 1963)
7-4 SOLVENT–WATER PROPERTIES
The solubility of CO2 in water is a function of temperature, pressure, and water

salinity (McRee, 1977). Figure 7-23 shows this solubility as a solution gas–water
ratio. The data in Fig. 7-23 give the maximum CO2 solubility at the indicated
260
tempeature and pressure; hence the horizontal axis is actually saturation pressure.
The data are entirely equivalent to the data in Fig. 7-21(a) for CO2–oil mixtures. The
solution gas–water ratio may be readily converted into mole fraction.
Carbon dioxide is the only solvent with appreciable solubility in water over
EOR temperature and pressure ranges (Culberson and McKetta, 1951). The CO2
increases the viscosity of water slightly (Tumasyn et al., 1969) and decreases the den-
sity (Parkinson and de Nevers, 1969). This density change has been shown (Welch,
1982) to be less than that predicted by ideal solution theory. Neither the change in
viscosity nor the change in density is likely to affect oil recovery very much.
7-5 SOLVENT PHASE BEHAVIOR EXPERIMENTS
Solvent phase behavior does not solely determine the character of a solvent flood, but
it is of such fundamental importance that we devote a section to some of the common
experiments used to measure phase behavior. This discussion leads naturally to the
most frequently reported characteristic of solvent phase behavior––minimum
miscibility pressure.
Single Contact
In a single-contact experiment, a known amount of solvent is charged into a

transparent pressure cell containing a known amount of crude oil. After equilibrium
is established at the desired temperature and pressure, a small amount of each phase
is withdrawn. The phase compositions represent the ends of an equilibrium tie line.
Only the composition of one phase need be measured since the composition of the
other phase can be calculated from material balance. Single-contact experiments are
useful for measuring P-z diagrams since the pressure can be changed, at fixed overall
composition, by changing the cell volume. If the experiment is repeated for various
amounts of solvent, the single-contact experiment traces a dilution path on a ternary
diagram between the solvent and crude.
Multiple Contact
The multiple-contact experiment duplicates the process described in Sec. 7-3 under
miscible process classification. In it (Fig. 7-24), known amounts of solvent and crude
are charged to a transparent pressure cell as in the single-contact experiment, but
after equilibration, the upper phase is decanted and mixed in a second cell with fresh
crude. The lower phase in the cell is similarly mixed with fresh solvent. The upper
phase is repeatedly decanted in this manner to simulate, discretely, the mixing that
would take place at the forward contacts of the solvent–crude mixing zone. The
successive mixings with the lower phase are the reverse contacts. All contacts are a
fixed temperature and pressure.
From Fig. 7-24, the multiple-contact experiment for Fig. 7-10, the solvent
261
Figure 7-24 Multiple-contact experiment in 105°F (2,000 psia) (from Gardner et

al., 1981)
enrichment in the forward contacts or the crude enrichment in the reverse contacts
can cause one of the phases to disappear. This is exactly what is predicted by the
arguments used in the process classification section: A single phase cell in the
forward contacts indicates a vaporizing gas drive; in the reserve contacts, a
condensing gas drive; and two or more phases in all contacts, an immiscible process.
If the original cell is single phase for all combinations of solvent and crude oil, the
process is first-contact miscible.
262
The experiment depends somewhat on the initial charges to the first cell, so
the results are no more than indications of process classification. If phase
compositions are measured at every step, the binodal curve and tie lines on a ternary
diagram are established. Agreement between single- and multiple-contact
experiments, as in Fig. 7-10, substantiates the pseudocomponent representation of the
multicomponent equilibria.
Both single- and multiple-contact experiments place a premium on visual
observations, but with careful selection of the initial volumes, these experiments are
convenient ways to determine complete ternary equilibria data. Orr and Silva (1982)
have proposed a method to measure phase behavior through continuous contacting.
Slim Tube
Filling the gap between the above static measurements and core floods are the slim
tube experiments. These experiments are crude displacements by solvent, in the
absence of water, at fixed temperature. The permeable medium consists of beads or
unconsolidated sands packed in tubes of very thin cross section and, frequently, large
length. The displacements are run with a fixed pressure at the one end of the system,
and because the permeability of the medium is large, pressure gradients are
negligible. Table 7-1 shows characteristics of selected slim tube experiments.
The overriding feature of slim tube experiments is the large aspect ratio
(length-to-diameter ratio). This is intended to suppress viscous fingering since the
long length means there is sufficient time during the displacement for all
perturbations to be suppressed by transverse dispersion. Small wavelength
perturbations will not form at all since the tube diameter is smaller than the critical
wavelength (see See. 6-8).
The slim tube experiment, then, is designed to provide an unambiguous
measure of solvent displacement efficiency. But because of both the highly artificial
nature of the permeable medium and the experimental conditions (no water), this is
not a realistic displacement efficiency. The results are best regarded as a dynamic
measure of phase behavior properties.
Minimum Miscibility Pressure
Although effluent compositions can be monitored during a slim tube displacement,

by far the most common information derived from the experiments is the minimum
miscibility pressure (MMP). Since solvent miscibility increases with pressure,
ultimate oil recovery should also increase with pressure. This, in fact, happens, but
there is a pressure above which a further pressure increase causes only a minimal
increase in oil recovery. The pressure at which oil recovery levels out is the MMP, or
minimum dynamic miscibility pressure. MMP is variously defined as
• The pressure at which the oil recovery at tD = 1.2 PV of CO2 injected was
equal to or very near the maximum final recovery obtained in a series of tests
(Yellig and Metcalfe, 1980)
263
TABLE 7-1 CHARACTERISTICS OF SLIM TUBE DISPLACEMENT EXPERIMENTS (ADAPTED FROM ORR ET AL., 1982)
Ktφ
Length ID Packing Permeability Porosity Rate
Author(s)* (meters) (cm) Geometry (mesh) (μm2) (%) (cm/hr) uL
Rutherford (1962) 1.5 1.98 Vertical tube 50–70 mesh 24 35 37 0.2
Ottawa sand
Yarborough and 6.7 0.46 Flat coil No. 16 AGS 2.74 66 11.9
Smith (1970)
Holm and 14.6 0.59 No. 60 3.81 109.1
Josendal (1974) 25.6 Crystal sand
Holm and 15.8 0.59 Coil No. 60 20 39 101–254 12.5
Josendal (1982) Crystal sand 14.9
Huang and Tracht 6.1 1.65 1.78 43 4.7 4.8
(1974)
Yellig and 12.2 0.64 OD Flat coil 160–200 mesh 2.5 5.2–10.2 33.4
Metcalfe (1980) sand 58.1
Peterson (1978) 17.1 0.64 60–65 mesh 19

sand
Wang and 18.0 0.62 Spiral coil 80–100 mesh 13 35 381 12.3
Locke (1980)
Orr and 12.2 0.64 Spiral coil 170–200 mesh 5.8 37 42 13.5
Taber (1981) glass beads
Gardner, Orr, 6.1 0.46 Flat coil 230–270 mesh 1.4 37 32 15.0
and Patel glass beads 64
(1981)
Sigmund et al. 17.9 0.78 140 mesh 5 42
(1979)
* References in Orr et al.
264
• The pressure that causes 80% oil recovery at CO2 breakthrough and 94%
recovery at a gas-to-oil ratio of 40,000 SCF/stb (Holm and Josendal, 1974)
• The pressure that causes 90% oil recovery at tD = 1.2 HCPV of CO2 injected
(Williams et al., 1980)
Others (Perry, 1978; Yellig and Metcalfe, 1980) emphasize the qualitative nature of
the miscibility pressure determination. The importance of the exact definition is
unknown; all definitions show the same trends in correlations.
The results of slim tube experiments are giving the minimum pressure
necessary for the displacement to develop miscibility. Thus the MMP corresponds to
the pressure at which the critical tie line passes through the crude composition. This
pressure is considerably less than that required for complete or first-contact
miscibility (compare the MMP plots with the P-z diagrams). This is the origin of the
plateau on the oil-recovery–pressure plot: Any further pressure increase does not
increase oil recovery since above the MMP the displacement will tend from
developed to first-contact miscibility. These observations are also supported by
compositional measurements wherein the properties (viscosity, density, and
composition) of phases produced below the MMP become closer to one another as
the MMP is approached.
The CO2 MMP is determined by temperature, pressure, solvent purity, and
molecular weight of the heavy fraction of the reservoir crude. Generally, the MMP
increases with temperature and heavy fraction molecular weight. Holm and Josendal
(1974 and 1982) note that the development of miscibility for CO2 solvents is the
result of extracting hydrocarbon components into a CO2-rich phase. Therefore, at a
given temperature and crude composition, sufficient compression must be applied to
the solvent to promote solvency with the crude. This solvency is manifest by the CO2
density at the temperature of the test. Figure 7-25(a) shows the CO2 density required
to develop miscibility at a given temperature with the C5–C30 percent of the C5+ crude
fraction. The CO2 density can be connected to MMP through Fig. 7-4 or Fig. 7-25(b).
CO2 MMP is affected by the type of hydrocarbons (aromatic or paraffinic) in the
crude but to a lesser degree than by temperature and CO2 density (Monger, 1985).
Several works have presented determinations of MMP for impure CO2.
Figure 7-26 shows the results of the effects of N2, CH4, H2S and H2S-CH4 mixtures
on the CO2 MMP. Methane and particularly nitrogen increase the CO2 MMP,
whereas H2S decreases it. Whether an impurity increases or decreases the MMP
depends on whether the solvency of the solvent has been enhanced. Solvency is
improved (MMP decreases) if CO2 is diluted with an impurity whose critical
temperature is more than that of CO2. Solvency deteriorates (MMP increases) if CO2
is diluted with an impurity with a critical temperature less than CO2. Compare the
trends in Fig 7-26 with the critical temperatures in Fig. 7-2.
The above idea of solvency can be used to estimate the MMP of an impure
CO2 solvent. Sebastian et al. (1984) have correlated the diluted CO2 MMP by the
following:
265
Figure 7-25 Density of CO2 required for miscible displacement of various oils at
90° to 190°F (from Holm and Josendal, 1982)
266
Figure 7-26 Effect of impurities on CO2 minimum miscibility pressure (from Johnson and
Pollin, 1981; Whitehead et al., 1980; Metcalfe, 1981)
267
PMM
= 1.0 − (2.13 × 10−2 )(Tpc − Tc )
( PMM ) CO2 (7.5-1)
−4 2 −7 3
+ (2.51 × 10 )(Tpc − Tc ) − (2.35 × 10 )(Tpc − Tc )
where Tpc = ∑ i Tci yi is the pseudocritical temperature of the mixture, and yi is the
mole fraction of species i in the solvent. The denominator of the left side of Eq.
(7.5-1) can be estimated from Fig. 7-25. (For other correlations, see Johnson and
Pollin, 1981.) No MMP correlation is especially accurate; errors as much as 0.34
MPa (50 psia) are common.
Minimum Enrichment Correlations
For a dry gas process, slim tube results will give an estimate of the amount of
intermediates that must be added to develop miscibility in a condensing gas drive.
Such experiments were precursors to the MMP experiments (Benham et al., 1961).
The oil recovery plot would consist of several experiments each with a successively
richer injected solvent but each at constant pressure. When the solvent composition
coincided with the tie line extension (through the reverse contacts), oil recovery
would cease to increase as the solvent becomes richer in intermediates.
Figure 7-27 is one of 12 plots from Benham et al. (1961) that shows the
maximum methane concentration permissible in an LPG solvent that will develop
miscibility with the subject crude. These authors correlated the maximum dilution (or
Figure 7-27 Maximum methane dilution in LPG solvent for developed miscibility
at 2,500 psia and for a reservoir fluid whose C5+ component molecular weight is
240 (from Benham et al., 1961)
268
minimum enrichment) with temperature, pressure, molecular weight of the
intermediate component in the solvent, and molecular weight of the C5+ fraction in
the crude. The minimum dilution increases with decreasing C5+ molecular weight,
pressure, and temperature, and it increases with increasing intermediate molecular
weight.
Each of these trends follows from the trends in the phase behavior and the
position of the crude and solvent on the ternary diagram. And each may be
quantitatively established on true ternaries with accurate thermodynamic properties.
But the pseudocomponent representation of more than three components on a ternary
is not rigorous, and this leads to some difficulty in quantitatively predicting both the
minimum dilution and the MMP on actual systems.
7-6 DISPERSION AND SLUG PROCESSES
In the next few sections, we look in detail at how a miscible solvent behaves during
oil displacement. You should remember that first-contact and developed miscibility
solvent behave very much alike.
Dilution Paths
The concentration of species i in a first-contact miscible displacement is from Eq.

(5.5-15)
⎡ ⎛ xD − t D ⎞⎤
(CiJ − CiI ) ⎢
Ci = CiI + 1 − erf ⎜ t ⎟⎥ (7.6-1)
2 ⎢ ⎜2 D ⎟⎥
⎜ ⎟⎥
⎣⎢ ⎝ N Pe ⎠⎦
For this equation to be valid, we cannot have viscous fingering, layering, or gravity
tonguing; hence it is restricted to constant viscosity and density floods in
one-dimensional media. In Eq. (7.6-1), xD is the dimensionless length, tD the
dimensionless time in fractional pore volumes, Npe the Peclet number, and the
subscripts I and J refer to initial and injected conditions, respectively.
If we let the component subscript i refer to the light, intermediate, and heavy
pseudocomponents of Sec. 7-3, we can easily show from Eq. (7.6-1) that dilution
paths are straight lines on a pseudoternary diagram. Eliminating the term in brackets
among the three equations gives
C1 − C1I C2 − C 2 I C3 − C3 I
= = (7.6-2)
C1J − C1I C2 J − C2 I C3 J − C3 I
The Ci in Eq. (7.6-2) lie on a straight line in composition space; hence the dilution
path of Sec. 7-3 is linear.
269
Superposition
Solvents are usually too expensive to be injected continuously. Thus a typical

displacement consists of a finite amount or slug of solvent followed by a less
expensive chase fluid. The concentration of a slug follows from Eq. (7.6-1) and the
principle of superposition. This principle applies to linear partial differential
equations, which Eq. (7.6-1) is an approximate solution to. We can, in fact, derive the
concentration response of an infinite number of step changes in the influent
concentration (see Exercise 7C), but we restrict our discussion here to the case of a
single solvent slug displaced by a chase fluid.
Let I, J, and K denote the concentrations of species i in the original fluid, the
slug, and the chase fluids, respectively. Superposition states that the sum of individ-
ual solutions to a linear differential equation is also a solution to the equation. This
seems easy enough to do in practice, but we must take care in selecting the boundary
conditions of the individual solutions to give the correct composite solution. Figure
7-28 shows the influent or imposed boundary conditions of the single front problem
Figure 7-28 Schematic of influent

boundary conditions for slugs
270
(Fig. 7-28b) and that of the composite solution (Fig. 7-28c). The composite solution
gives Ci(xD, tD) for the imposed conditions in Fig. 7-28(a), simply the sum of the
solutions to the conditions in Figs. 7-28(b) and 7-28(c), respectively. The solution to
the imposed conditions in Fig. 7-28(b) is Eq. (7.6-1), and that of the imposed
condition in Fig. 7-28(c) is
⎡ ⎛ xD − (t D − t Ds ) ⎞ ⎤
(CiK − CiJ ) ⎢
Ci = 1 − erf ⎜ ⎟⎥
⎜ 2 (t D − t Ds ) ⎟ ⎥
(7.6-3)
2 ⎢
⎢⎣ ⎜ ⎟
⎝ N Pe ⎠ ⎥⎦
By superposition Ci(xD, tD) for the influent condition in Fig. 7-28(a) is the sum of
Eqs. (7.6-1) and (7.6-3)
⎛ xD − t D ⎞
CiI + CiK ⎛ CiJ − CiI ⎞ ⎜ ⎟
Ci = +⎜ ⎟ erf ⎜ 2 t D ⎟
2 ⎝ 2 ⎠ ⎜ N ⎟
⎝ Pe ⎠
⎛ xD − (t D − t Ds ) ⎞
⎛ C − CiK
+ ⎜ iJ
⎞ erf ⎜ ⎟
⎟ ⎜ 2 t D − t Ds ⎟ , t D > t Ds (7.6-4)
⎝ 2 ⎠ ⎜ ⎟
⎝ N Pe ⎠
Equation (7.6-4) is valid for any value of the injected concentrations.

Frequently, we are interested in the concentration of the solvent at the
midpoint between xD = tD and xD = tD – tDs. Evaluating Eq. (7.6-4) at xD = tD – tDs/2
yields this midpoint concentration Ci
⎡ ⎛ t Ds ⎞ ⎤ ⎛ t Ds ⎞
CiI + CiK ⎢ ⎜ ⎟
Ci = 1 − erf tD ⎥ + CiJ erf ⎜ tD
⎟ (7.6-5)
2 ⎢ ⎜ 4 ⎟ ⎥ ⎜ 4 ⎟
⎢⎣ ⎜ ⎟ ⎜ ⎟
⎝ N Pe ⎠ ⎥⎦ ⎝ N Pe ⎠
This equation is valid only for relatively small tDs where the difference between the
square roots of tD and tD – tDs in the denominator of the error function argument is not
large. If CiJ > CiI and CiJ > CiK, the midpoint concentration is usually called the peak
concentration. For CiI = CiK = 0, the peak concentration falls with increasing time
according to
⎛ t Ds ⎞
Ci = CiJ erf ⎜ ⎟ (7.6-6)
⎜ 4 tD ⎟
⎜ N ⎟
⎝ Pe ⎠
The error function may be replaced by its argument for small values of the argument.
In this event, the peak concentration falls in inverse proportion to the square root of
time. Since xD = tD – tDs/2 at the peak concentration, this is equivalent to Ci falling in
proportion to the inverse square root of the distance traveled.
271
Figure 7-29 Miscible slug concentration profiles for matched viscosity and density
displacements (from Koch and Slobod, 1956)
272
The peak concentration falling below CiJ is the consequence of -overlapping
front and rear mixing zones. Figure 7-29 shows experimental concentration profiles
from a miscible slug displacement at different throughputs. Figure 7-29(a) has the
concentration profiles normalized to the midpoint position xD = tD – tDs/2 on the
horizontal axis. The areas under all curves are the same (material balance is
preserved), but the peak concentration falls as the number of passes (travel distance)
increases. The unnormalized profiles in Fig. 7-29(b) show that the peak concentration
falls approximately as the inverse square root of tD in experimental floods.
The midpoint concentrations also trace a straight line in the pseudoternary
diagram since the error function arguments in Eq. (7.6-5) may be eliminated to give
C1K + C1I C + C2 I C + C3 I
C1 − C2 − 2 K C3 − 3 K
2 = 2 = 2 (7.6-7)
C1K + C1I C2 K + C2 I C3 K + C3 I
C1J − C2 J − C3 J −
2 2 2
This equation says that as time increases, the midpoint concentration traces a straight
line between the injected slug concentration CiJ and the average concentration of the
Figure 7-30 Dilution of solvent slug by mixing (from Stalkup, 1983)

273
fluids ahead of and behind the slug. The midpoint concentrations at successive times
a, b, and c are shown in Fig. 7- 30, as are the dilution paths given by Eq. (7.6-4). The
dilution paths become straight line segments from CiJ to Ci and then from Ci to CiK
for tDs small. These considerations are valid only so long as the entire dilution path
stays in the single-phase region of the diagram. It is not necessary for Ci to fall into
the two-phase region for the displacement to lose first-contact miscibility (see
Exercise 7E).
7-7 TWO-PHASE FLOW IN SOLVENT FLOODS
Two or more phases are all too common in solvent floods. When this happens, the
dispersion theory of Sec. 7-6 does not apply. But general conclusions about such
displacements are still possible based on the coherent or simple wave theory first
introduced in Sec. 5-6. This theory neglects dissipative effects of any kind; hence we
omit dispersion in the following discussion and restrict our treatment to centered
simple waves (see Sec. 5-4 for definitions).
We treat two cases of two-phase flow in miscible displacements: (1) solvent
floods in the absence of an aqueous phase and (2) first-contact miscible
displacements in the presence of an aqueous phase. In both cases, fluid displacement
takes place in a one-dimensional permeable medium at constant temperature and with
incompressible fluids and solid.
Solvent Floods in the Absence of an Aqueous Phase
In this section, we give a theoretical base for the classifications of Sec. 7-3. Consider
a three-component system consisting of an intermediate hydrocarbon (i = 2), a light
hydrocarbon (i = 3), and a heavy hydrocarbon that can form no more than two phases
at constant temperature and pressure. As we discussed in Sec. 4-3, the overall
concentrations Ci, the phase concentrations Cij, and the saturations Sj (j = 2 or 3) can
be conveniently represented on ternary diagrams.
The topology within the two-phase region is important for this problem.
Figure 7-31 shows a ternary diagram with a two-phase region exaggerated to point
out certain landmarks. Within the two-phase region is a family of quality lines that do
not intersect and converge at the plait point. The binodal curve itself is a quality line.
There are also lines denoting the residual saturations of the two phases. These lines
do not, in general, coincide with quality lines since residual saturations must decrease
as the plait point is approached (see Sec. 3-4). This decrease is because the interfacial
tension between the two phases must vanish at the plait point. Along each tie line,
there exists a curve relating the fractional flow of one of the phases to its saturation.
Three of these curves, along tie lines A-A′, B-B′ and C-C′, are in the upper left insert
to Fig. 7-31. The shape of the fractional flow curves is not determined by the phase
behavior alone, but the curves become straighter (more miscible-like) with smaller
residual phases along tie lines near the plait point. Because phase compositions are
274
Figure 7-31 Landmarks on a two-phase ternary
constant along tie lines, the Cij, Sj and fj can be converted to fractional flux and
overall concentration through Eqs. (5.4-3). The upper right insert of Fig. 7-31 shows
an F3-C3 along the three tie lines.
This ternary system has only two independent components, which we
arbitrarily take to be C2 and C3. The coherence condition (Eqs. 5.4-5 and 5.6-14) for
this case becomes
dF2 dF3
vC2 = = = vC3 (7.7-1)
dC2 dC3
Using the condition f2 + f3 = 1, and the definitions for overall flux and concentration
(Eqs. 5.4-3a and 5.4-3c), Eq. (7.7-1) can be rewritten as (Helfferich, 1982)
275
f (dC22 − dC23 ) + dC23 + (C22 − C23 )df 2
vC2 = 2
S2 (dC22 − dC23 ) + dC23 + (C22 − C23 )dS 2
(7.7-2)
f (dC32 − dC33 ) + dC33 + (C32 − C33 )df 2
= 2 = vC3
S2 (dC32 − dC33 ) + dC33 + (C32 − C33 )dS2
The curve in the ternary composition space that a displacement follows (the
composition route) is quite complex, but certain segments (composition paths) are
readily apparent from Eq. (7.7-2).
1. Unit velocity paths. These occur along any direction in the single-phase
region (all directions are coherent) or along the binodal curve. In both cases,
f2 = S2 = 1 or f3 = S3 = 1, depending on the side of the plait point, and the
composition velocity is
vC2 = vC3 = 1 (7.7-3)
Equation (7.7-3) is the same result as Eq. (5.4-7).
Within the two-phase region is an equivelocity path where f2 = S2.
This path is the intersection of a straight line through f2 = S2 = 0 and f2 = S2 =
1 and the family of fractional curves (Fig. 7-31). It converges to the plait
point, but it does not, in general, coincide with a quality line.
2. Tie line paths. On tie lines in the two-phase region, dCij = 0. This also
satisfies Eq. (7.7-2). On these paths, the concentration velocities are
df 2
vCi = , i = 2, 3 (7.7-4)
dS 2
Equation (7.7-4) is the same as the saturation velocity in the

Buckley-Leverett theory (Eq. 5.2-10).
Two other types of paths are not so easily derived. Both follow from
integrating the composition path curve
dC2 vC± − F22

= (7.7-5a)
dC3 F23
where the composition velocity is

1
vC± = {( F22 + F33 ) ± [( F33 − F22 ) 2 + 4 F23 F32 ]1/ 2 } (7.7-5b)
2
and
⎛ ∂F ⎞
F22 = ⎜ 2 ⎟
⎝ ∂C2 ⎠C 3
and so on.
3. Singular curves. Along these curves, the velocity of the fast and slow paths is
equal. The curves follow from setting the discriminant of Eq. (7.7-5b) equal
276
to zero. Singular curves are composition paths since they can be generated
from Eq. (7.7-5a) as long as F23 ≠ 0.
4. Non-tie line paths. Within the two-phase region, there are also composition
paths whose trajectories are not readily apparent from the above equations.
Figure 7-32 shows each of these paths.
If the tie lines extend to a common point, the concentration velocity along the
non-tie line paths is constant (Cere and Zanotti, 1985). The phase equilibria is now
represented by Eq. (4.4-27b), which we repeat for this special case
Figure 7-32 Composition path in two-phase ternary equilibria

277
C3 j − C = η (C2 j − C ),
0
3
0
2 j = 2 or 3 (7.7-6)
A particular tie line is represented by a particular value of η. If we introduce the
definitions for overall flux and concentration into the coherence condition Eq.
(7.7-1), we have
dF2 d (C22 f 2 + C23 f 3 ) d (C32 f 2 + C33 f 3 ) dF3
= = = (7.7-7)
dC2 d (C22 S 2 + C23 S3 ) d (C32 S 2 + C33 S3 ) dC3
We can substitute Eq. (7.7-6) into the third term to give, after some rearrangement
and identification,
dF2 dη ( F2 − C20 ) + η dF2
= (7.7-8)
dC2 dη (C2 − C20 ) + η dC2
Our task is to find the combination of variables that makes this equation an identity.
Immediately we see that the equivelocity and tie line paths are returned from
Eq. (7.7-8), for the conditions F2 = C20 = C2 and dη = 0 clearly satisfy the equation.
But the existence of both paths is more general than this since it follows from Eq.
(7.7-2).
The non-tie line paths are defined by the following equations:
dF2 = dη ( F2 − C20 ) + η dF2
(7.7-9)
dC2 = dη (C2 − C20 ) + η dC2
Eliminating η between these two equations gives an ordinary differential equation

relating F2 and C2 along the non-tie line path
dF2 F2 − C20
= (7.7-10)
dC2 C2 − C20
Integrating this equation yields a linear relation between F2 and C2

F2 − C20 = I c (C − C20 ) (7.7-11a)
where Ic is an integration constant independent of either F2 and C2. Immediately it

follows from Eqs. (7.7-11a) and (7.7-1) that the velocity along the non-tie line path is
constant and that the constant velocity is, in fact, the integration constant. The path
itself is given by the linear relation
Fi − Ci0 = vC (Ci − Ci0 ), i = 2, 3 (7.7-11b)
Since the above development applies for either independent species, we drop the
subscript on vC.
As Fig. 7-32 shows, the entire two-phase region is covered with a net of non-
tie line paths along each of which the velocity is constant. Some of these paths cross
the tie line paths, but others merge continuously with it at a point where the velocity
along both paths is equal. The curve defining the locus of these intersection points is
given by
278
⎛ dF2 ⎞ ⎛ F −C ⎞ 0
⎜ dC ⎟ =⎜ 2 ⎟
2
(7.7-12)
⎝ 2 ⎠ tie line ⎝ C2 − C ⎠ nontie line
0
2
from Eqs. (7.7-4) and (7.7-11b). This curve is the singular curve discussed above
wherein the discriminant of Eq. (7.7-5b) vanishes.
Figure 7-33 shows two fractional flux curves going from points J to I along
the lines A-A′ and B-B′. These were selected because they are on tie lines that extend
to the points J and I, respectively. The curves consist of three segments: a portion of
unit slope corresponding to the single-phase regions in Fig. 7-32, horizontal portions
corresponding to single-phase flow in the presence of another residual phase, and a
curved portion corresponding to two-phase flow. The curve with the more
compressed curved portion corresponds to the tie line nearest the plait point.
Figure 7-33 also shows the construction for the singular point as suggested
by Eq. (7.7-12). Since the slope of a tie line path is the coherent velocity, the tie line
Figure 7-33 Fractional flux curves for Fig. 7-32

279
paths are slow outside the singular point intersections, and fast within the
intersections. Understanding how a given tie line path can be both fast and slow is
important for what follows.
You should compare both the above constructions with those in Fig. 5-12(d).
Just as there are a variety of possibilities in Fig. 5-12, because of the variety of
fractional flow shapes, so there are several possible behaviors for the singular curves.
For example, a fractional curve without an inflection (Fig. 5-12a) will have one
singular curve that coincides with a residual phase saturation curve.
We use the curves in Fig. 7-33 to select the only physically possible
composition routes in Fig. 7-32. Before doing this, we remind you of the principles
established in Sec. 5-6 for centered simple waves.
1. The composition route must stay on the composition path segments in the
ternary diagram.
2. The composition velocity must decrease monotonically in the upstream
direction. (This rule is actually a special case of the more general statement
that all concentrations must be single valued.)
3. The correct composition route must be insensitive to infinitesimal perturba-
tions in concentration. (This rule was not needed in Chap. 5, but it is here.)
We build the composition path from I to J in three segments, each of which

must satisfy these rules.
Consider first the displacement of J1 by J. There are an infinite number of
paths between J and J1, but we consider only the two extreme routes J → J′ → J1 and
J → B′ → J1. The second path is nonphysical because it involves a fast segment (in
the single-phase region) upstream of the slow segment between B′ and J1. We could
resolve this by putting a shock directly from J to J1 but this would no longer follow
the composition paths. Route J → J′ → J1 also contains fast paths upstream of slow
paths, but the resulting resolution into shocks (Fig. 7-34a) remains on the
composition paths. In fact, this is the only route between the two extremes we discuss
here that remains so because the switch from one tie line to the other, J to J1, takes
place along a non-tie line path.
We see that the general use of non-tie line paths is to switch between tie
lines. Figure 7-34(b) shows a composition profile for this displacement. The
displacement shocks across the residual phase 2 saturation, causing complete
recovery of this phase. This recovery, which takes place in the absence of lowered
interfacial tension and developed miscibility, occurs because phase 2 dissolves into
the injected phase J. Dissolution waves are normally inefficient since their
propagation velocity is slow (Fig. 7-34).
These comments also apply to the path from I′1 to I. Of the two extreme
routes, I′1 → I1 → I and I′1 → A → I, only the latter yields a route along which the
shock resolutions will remain on the composition path segments. Figure 7-35 shows
the construction and corresponding profiles.
By comparing the routes J → J1 and I′1 → I, we see both entry and exit from
280
Figure 7-34 Composition route and profiles for displacement J → I1
a two-phase region take place along tie line extensions. Further, both entry and exit
contain slow shocks, which are the direct result of the slow segments of the fractional
flux curves. If I′1 were above the equivelocity curve in Fig. 7-35(a), the route would
follow a fast shock along the tie line nearest the plait point.
The third segment, J1 to I1, follows a single tie line path whose velocity is
given by the Buckley-Leverett construction (Fig. 7-36) from Eq. (7.7-4).
In a sense, the above constructions, particularly I′1 to I, are misleading when
applied to the entire displacement from I to J because the rules for centered simple
waves must apply globally rather than individually to segments. To see this, consider
the four possible composition routes from J to I: J → J′ → J1 → I, J → J′1 → J1 → I,
J → I′1 → I1 → I, and J → J′ → J1 → I. After you carefully consider each case, with
shock segments interspersed, the requirement of a monotonically increasing
concentration velocity forces you to see the only correct choice is J → J′ → J1 → I.
The immiscible displacement J → I will consist of two shock segments between
which is a small spreading wave; these are sketched schematically in Fig. 7-37a.
281
Figure 7-35 Composition route and profiles for displacement I′1 → I
As in all MOC problems, an infinite number of mathematical solutions exist,

but by assumption, only one physical solution exists. Finding the physical solution
involves trial and error according to the following procedure:
1. Locate a tentative composition route between the injected and initial

conditions on the fractional flux diagram. This route consists of segments
that conform to the paths we discussed above.
2. Resolve all physical inconsistencies along the tentative route with shocks. In
doing this, assume the differential and integral composition routes are the
same.
3. Discard any tentative solution in which the shock resolution leads to a route
that does not follow the composition path sequence. When this happens,
return to step 1 with another tentative route. The correct solution is usually
clear after a few trials.
282
Figure 7-36 Composition route and profiles for displacement I1 → I′
These rules enable us to sketch the composition routes for the three types of
displacements (Fig. 7-37). The composition route of the immiscible displacements
(Fig. 7-37a) both enters and exits the two-phase region on tie line extensions. The
entering segment is an extremely slow shock (a solubilization wave), which is the
consequence of the residual phase saturations. If the system includes more than three
true components––that is, at least one apex was a pseudocomponent––the
displacement would not revert to single-phase behavior as suggested by Fig. 7-37(a)
(Gardner and Ypma, 1982). Compare Fig. 7-37(a) to Fig. 7-16.
The vaporizing gas drive process (Fig. 7-37b) shows a composition route that
approaches the binodal curve on a tie line extension and then follows the binodal.
curve until it reaches a point on a straight line tangent to the initial composition.
Compare Fig. 7-37(b) to Fig. 7-14.
283
Figure 7-37 Composition routes for immiscible and developed miscibility processes
In the condensing gas drive process (Fig. 7-37c), the composition route
enters the two-phase region through the plait point (Hutchinson and Braun, 1961),
follows the equivelocity path, and then exits from the two-phase region on a tie line
extension. The route definitely passes through the two-phase region, but it does so as
a shock since the concentration velocity on the equivelocity curve is unity, and the tie
line extension segment is a shock. Compare Fig. 7-37(c) to Fig. 7-15. Auxiette and
Chaperon (1981) give an experimental investigation of these processes.
Both developed miscibility cases will appear as a first-contact displacement
in the absence of dissipation. The similarity between developed and first-contact
displacements justifies using first-contact approximations on all the developed
miscibility displacements we discuss below.
284
First-Contact Miscible Displacements in the Presence
of an Aqueous Phase
Water does not affect hydrocarbon phase behavior, and the water solubility of most
solvents is small. But the inevitable presence of an aqueous phase can affect
displacement behavior through fractional flow effects. In this section, we investigate
the effects of an aqueous phase on a first-contact miscible displacement. Although
the treatment here can be given formally, as was that discussed above, we present
instead an entirely equivalent, but more direct, approach based on fractional flow
curves.
To do this, we assume incompressible fluids and rock, no dissipative effects,
and solvent–water relative permeabilities are the same as oil–water relative perme-
abilities. Thus a water–solvent fractional flow f1s differs from a water–oil fractional
flow f1 only by the difference between the use of solvent and oil viscosities and den-
sities. Figure 7-38 shows both the f1 and f1s curves based on relative permeabilities
from Dicharry et al. (1972). Because the relative permeabilities do not change,
residual phase saturations of both the aqueous and oleic phases are invarient. The
initial condition I in the one-dimensional displacement is uniform with water cut f1I.
We take an arbitrary injection condition J to be comprised of some pre-
specified proportion of solvent and water f1sJ given on the solvent–water curve.
Injecting water and solvent together in the so-called water-alternating-gas (WAG)
process is commonly used in solvent floods. The solvent–water mixture has better
volumetric sweep efficiency and is less prone to viscous fingering than solvent alone
(Caudle and Dyes, 1958). The volumetric flow rate ratio of water to solvent in the
injected fluid is the WAG ratio WR, given by
WR
f1sJ = (7.7-13)
1 + WR
In Eq. (7.7-13) and hereafter, we assume no solubility of solvent (i = 3) or oil (i = 2)

in the aqueous (i = j = 1) phase, and we assume no solubility of water in the
hydrocarbon (j = 2) phase.
In an actual WAG process, the water and solvent are usually injected in
alternate slugs so that the cumulative volumes of solvent and water really define the
WAG ratio rather than Eq. (7.7-13). The differences in displacement behavior caused
by simultaneous injection rather than alternating injection have been investigated by
Welch (1982).
Since the displacement is first-contact miscible, the wave between the
injected solvent and the oil is indifferent. Hence the oil–solvent wave velocity is
1 − f1sJ
v3 = (7.7-14)
1 − S1J
from Eq. (5.4-5b). Equation (7.7-14) neglects solvent adsorption. v3 can also be
written in terms of the change in water saturation change across the solvent–oil front
285
Figure 7-38 Schematic fractional flow construction for first-contact miscible

displacements in the presence of an aqueous phase
f1sJ − f1B
v3 = (7.7-15)
S1J − S1B
Equation (7.7-14) is the equation of a straight line from the upper right-hand corner
of the fractional flow plot through the injected conditions (Fig. 7-38). Equating Eq.
(7.7-14) to Eq. (7.7-15) says if this line is continued, its intersection with the water–
oil fractional flow curve will give the water and oil saturation and fractional flow in
the region ahead of the solvent–oil wave. Since 1 – f1B is larger than 1 – f1I, the
displaced oil forms a region of high oil saturation or oil bank ahead of the solvent–
water front.
The leading edge of this oil bank flows with specific velocity v2B given by
f1I − f1B
v2 B = (7.7-16)
S1I − S1B
also from Eq. (5.4-5b). This is the equation of a straight line from the initial condi-
286
tions I to the oil bank, point B in Fig. 7-38. Since the injected water miscibly
displaces the resident water, the specific velocity of the displaced water wave v1′ is
the straight line from the lower left corner of the fractional flow plot to the injected
conditions (compare the lines for v1′ and v3 with case B in Fig. 5-12).
The velocity of the connate water banked up by the injected water is
f1sJ
v1' = (7.7-17)
S1J
which is also shown in Figs. 7.38 and 7.39. The water ahead of this wave is
banked-up connate water which, for a secondary solvent flood (f1I = 0), constitutes a
waterflood ahead of the solvent front. Caudle and Dyes (1958) verified
experimentally that injecting at a WAG ratio so that the banked-up connate water
does not propagate faster than the solvent resulted in optimal oil recovery.
Figure 7-39 Time–distance diagram and effluent history plot for the displacement
287
It is also possible to treat slug behavior with fractional flow theory. Suppose
after injecting at condition J for some dimensionless time tDs, we follow with chase
water at condition K on the diagram. Since the slope of the solvent–water fractional
flow curve is monotonically decreasing from K to J, this displacement front is a
shock with specific velocity
1 − f1sJ
vcw = (7.7-18)
1 − S2 r − S1J
whose straight-line construction is also shown in Fig. 7-38. If the chase fluid were to
be a second gas having the same properties as and first-contact miscible with the
solvent, the velocity of the chase-fluid–solvent front would be given by the slope of
the f1s -S1 curve since the slope of this curve is monotonically increasing from S1 =
S1r to the injected conditions.
Figure 7-39 shows the time–distance diagram and effluent history for the
displacement in Fig. 7-38. We have taken tDs = 0.8 to avoid interference between the
oil-bank–solvent and the solvent–chase water waves. In the effluent history plot, the
miscible displacement fronts are designated by a wavy line.
Several general observations; follow from these plots. First, the ultimate oil
recovery is complete––that is, the final condition in the system is zero oil saturation.
Of course, this is the natural consequence of first-contact miscible displacements
where no residual phases are allowed. Second, oil production ceases when solvent
breaks through. The moderately early solvent breakthrough is the consequence of the
pore space inaccessible to the solvent caused by the presence of irreducible water:
With no water present, the solvent slug always breaks through near tD = 1. Based on
hydrocarbon pore volumes (1 – S1r), the solvent in Fig. 7-39 breaks through at tD =
0.96 HCPV, which is much more in line with dispersion theory. Finally, the amount
of solvent produced (0.14 PV) is considerably less than the amount injected (0.4 PV).
This reduction is the consequence of trapping of the hydrocarbon-miscible solvent by
the chase water. If the solvent slug size tDs were less than about 0.6, the chase-water–
solvent front would have overtaken the solvent–oil-bank front and trapped some oil.
Such an observation suggests a procedure whereby we could select the minimum
solvent slug size (tDs = 0.53 in Fig. 7-39) that effects complete oil recovery.
7-8 SOLVENT FLOODS WITH VISCOUS FINGERING
Unfortunately, first-contact miscible displacements actually behave considerably

differently than that shown in Fig. 7-38. Figure 7-40 shows the experimental results
of a developed miscible displacement in a Berea core in which oil initially at residual
conditions is displaced by a CO2 solvent in a WR = 0 displacement. The deviation of
this displacement from a straight-line composition route in the ternary diagram was
small. In the experimental displacement, the water cut was initially 1.0 and decreased
to about 0.15 at tD = 0.15. The water cut remained essentially constant until about
288
Figure 7-40 Effluent history of a carbon dioxide flood (from Whitehead et al.,
1981)
tD = 0.33, at which point, it decreased gradually to 0. But when the water cut
originally fell to tD = 0.14, both oil and solvent broke through. This leaves a
remaining oil saturation of about 0.25 at termination. It is unclear if 100% oil
recovery would have been obtained had the experiment been continued. Several pore
volumes of solvent injection would have been required, however.
The primary cause of the simultaneous oil and solvent breakthrough and
prolonged oil recovery in experimental displacements is viscous fingering. In Sec.
6-8, we concluded that miscible displacements with typical solvents were always
unstable, barring a gravity stabilization or a boundary effect, because the solvent–oil
mobility ratio is greater than 1. Here we give descriptions of the character of
simultaneous oil and solvent flow after the onset of fingering.
Heuristic Models
Because of the chaotic nature of viscous fingering, a rigorous mathematical theory is

not possible. The behavior of a fingering displacement may be estimated by various
heuristic theories, including (1) a modification of fractional flow theory (Koval,
1963), (2) rate-controlled mass transfer between solvent and oil fingers (Dougherty,
1963), (3) defining a suitably weighted mixture viscosity (Todd and Longstaff,
1972), (4) accounting for mixing in fingers directly (Fayers, 1984), and (5) defining a
composition-dependent dispersion coefficient (Young, 1986).
In this section, we deal exclusively with the Koval theory; we leave the
others as an exercise. By excluding the others, we do not imply the Koval approach is
289
superior since all involve empirical parameters that must be determined by history
matching. However, the Koval theory is in common use, and it fits naturally into our
fractional flow theme.
The mixing zone length (the dimensionless distance between prespecified
values of a cross-sectionally averaged concentration profile) of a fingering
displacement, in the absence of boundary effects, grows in proportion to time. This
observation prompted Koval to instigate a fractional flow theory for viscous
fingering. If viscous fingers initiate and propagate, their growth in horizontal plane
flow would look something like the cross section in Fig. 7-41, where the oil and
solvent are in segregated flow. The displacement is first-contact miscible, with no
dissipation, and without water present. If dissipation can vertically smear the fingers,
the mixing zone will grow in proportion to the square root of time, as in dispersion
theory. This growth can be quite small if longitudinal dispersion is small or the
system length is large (Hall and Geffen, 1965).
Figure 7-41 Idealization of viscous

finger propagation (from Gardner and
Ypma, 1982)
With these qualifications, the volumetric flow rate of solvent across a vertical
plane within the mixing zone is
A k ⎛ ∂P ⎞
q3 = − 3 ⎜ ⎟ (7.8-1a)
μ3 ⎝ ∂x ⎠
and that of oil is
A k ⎛ ∂P ⎞
q2 = − 2 ⎜ ⎟ (7.8-1b)
μ 2 ⎝ ∂x ⎠
where A3 and A2 are cross-sectional areas of oil and solvent. There are no relative
permeabilities or capillary pressures in these equations since the displacement is
first-contact miscible. These equations assume a horizontal displacement. The
fractional flow of solvent in the oleic phase across the same vertical plane is
290
q3
f32 =
q3 + q2
by definition, which, when Eq. (7.8-1) is substituted, yields
A3 / μ 3
f32 = (7.8-2)
A3 / μ 3 + A2 / μ 2
Equation (7.8-2) assumes the x-direction pressure gradients are equal in the oil and
solvent fingers. Because the displacement is in plane flow, the oil and solvent cross-
sectional areas are proportional to average concentrations, or
−1
⎛ 1 ⎛ 1 − C32 ⎞ ⎞
f32 = ⎜⎜ 1 + ⎜ ⎟ ⎟⎟ (7.8-3)
⎝ v ⎝ C32 ⎠ ⎠
where v is the oil–solvent viscosity ratio, and C32 is the average solvent
concentration in the oleic phase across the cross section.
Equation (7.8-3) is a description of the segregated flow fingering in Fig.
7-41. Koval had to modify the definition of v to match experimental displacements.
The final form of the solvent fractional flow is
−1
⎛ 1 ⎛ 1 − C32 ⎞ ⎞
f32 = ⎜⎜1 + ⎜ ⎟ ⎟⎟ (7.8-4)
⎝ K val ⎝ C32 ⎠ ⎠
where Kval is the Koval factor.
Koval Corrections
The Koval factor modifies the viscosity ratio to account for local
heterogeneity and transverse mixing in the following fashion:
K val = H K i E (7.8-5)
The parameter E changes the viscosity ratio to account for local mixing
E = (0.78 + 0.22v1/ 4 ) 4 (7.8-6)
The consequence of Eq. (7.8-6) is that the numerical value of E is usually smaller
than that of v. That is, the effect of fingering is not as severe as it appears from the
original viscosity ratio. The 0.22 and 0.78 factors in Eq. (7.8-8) seem to imply the
solvent fingers contain, on the average, 22% oil, which causes the viscosity ratio
attenuation through the quarter-power mixing rule. In fact, Koval eschewed this
interpretation by remarking that the numerical factors were simply to improve the
agreement with experimental results. This would seem to restrict Eq. (7.8-6) to the
exact class of experiments reported by Koval. Remarkably, Claridge (1980) has
shown that the 0.22–0.78 factors accurately describe fingering displacements over
large ranges of transverse dispersion. Very likely the finger dilution is being caused
by viscous crossflow since the mechanism is consistent with linear mixing zone
growth (Waggoner and Lake, 1987).
291
The heterogeneity factor HK corrects the reduced viscosity ratio for the local
heterogeneity of the medium. Selecting the correct value for HK is the most subjective
feature of the Koval theory. In Fig. 6-8, the heterogeneity factor was calculated from
the Dykstra-Parsons coefficient. It has also been correlated with the longitudinal
Peclet number (Gardner and Ypma, 1982).
The fractional flow expression (Eq. 7.8-4) is the same as the water fractional
flow in a waterflood where the oil and water have straight-line relative
permeabilities. For such a case (see Exercise 5E), the Buckley-Leverett equation (Eq.
5.2-10) may be integrated analytically to give the following expression for effluent
fractional flow:
⎧ 1
⎪0, tD < 1
K val
⎪ 1/ 2
⎪ ⎛K ⎞
f32 |xD =1 = ⎨ K val − ⎜ val ⎟ (7.8-7)
⎪ ⎝ tD ⎠ , 1
< t D < K val
⎪ K val − 1 K val
⎪1, K val < t D
⎩
The oil fractional flow is 1 – f32 |xD =1 . This equation has been compared to
experimental data in the original Koval paper and elsewhere (Claridge, 1980;
Gardner and Ypma, 1982).
Koval with Mobile Water
The Koval theory applies to first-contact miscible displacements in the

absence of flowing water. The theory may be readily generalized to fingering
first-contact miscible displacements with water present by modifying the overall flux
and concentration definitions (see Sec. 5-4). The overall flux for oil and solvent
becomes
F2 = (1 − f 32 ) f 2 (7.8-8a)
F3 = f 32 f 2 (7.8-8b)
where f1 and f2 are the actual water and hydrocarbon fractional flow functions, and f32
is given by Eq. (7.8-4). To be consistent with Eq. (7.8-6), the hydrocarbon phase
viscosity in both f1 and f2 is given by the quarter-power mixing rule.
The overall concentrations of the oil and solvent are
C2 = (1 − C32 ) S 2 (7.8-9a)
C3 = C32 S 2 (7.8-9b)
The water concentration is simply S1 because there is no solvent solubility in the
water phase. Equations (7.8-8) and (7.8-9), substituted into the conservation
equations for oil and solvent, may then be solved by the simple wave procedure
discussed in Sec. 5-7 for the oil–gas–water problem.
Figure 7-42 shows the effluent fluxes for four displacements using this
procedure. Figure 7-42(a) is for a non-WAG secondary flood, which is simply the
results of the original theory (Eq. 7.8-7). Figure 7-42(b) is for a tertiary non-WAG
292
Figure 7-42 Effluent histories for four fingering cases

293
displacement. Figures 7-42(c) and 7-42(d) are for secondary and tertiary WAG
displacements (WR = 2). The oil and water relative permeabilities in Figs. 7-39 and 7-
42 are the same so that comparing Fig. 7-39 and Fig. 7-42(d) should reveal the effect
of fingering on a first-contact miscible displacement with water present.
For both cases, the oil is produced as a bank of constant cut. But the bank oil
cut is smaller for the fingering displacement, and the oil breakthrough and complete
sweepout times are later. In the fingering case, oil and solvent break through together
though the solvent is at low cut. By comparing Figs. 7-42(a) and 7-42(c) and Figs.
7-42(b) and 7-42(d), we see that, regardless of the initial conditions, the WAG
procedure delays solvent breakthrough and hastens complete oil recovery.
Based on the comparisons in Fig. 7-42, it appears that WAG is universally
better than injecting solvent alone, particularly when the solvent efficiency is
considered. However, the presence of an initial mobile water saturation causes a
residual oil saturation to even a first-contact displacement (see Sec. 7-9), and it is
possible that the WAG process will cause this also.
Other methods besides WAG to improve mobility control in miscible
flooding include the use of polymers (Heller et al., 1984) and foams. To date only
foams have been extensively investigated, and since foams are envisioned to drive a
variety of EOR processes, we delay their discussion until Chap. 10, where they more
naturally fit after micellar-polymer flooding.
7-9 SOLVENT FLOODING RESIDUAL OIL SATURATION
A residual oil saturation in solvent flooding can come about by two broad
phenomena: (1) a local heterogeneity (dead-end pores) in the permeable medium and
(2) an interaction of dispersion or viscous fingering with the phase behavior. The
former phenomenon occurs in first-contact miscible displacements, and the latter in
developed miscible flood.
The definition of residual oil in a miscible flood (a paradoxical quantity) is
slightly different from that in a waterflood. In a waterflood, residual oil is left behind
as capillary-trapped globs, and no amount of throughput will displace this oil without
some imposed change in the local capillary number. In a first-contact or developed
miscible flood, all the oil, even that “trapped” by whatever mechanism, will
eventually be recovered through extraction if enough solvent is injected. By residual
oil in a miscible flood, then, we mean that quantity of oil left behind a solvent flood
at some practical extreme of oil cut, oil rate, water–oil ratio, or gas–oil ratio (the data
in Fig. 7-40 are up to a gas–oil ratio of about 550 SCM/SCM). Admittedly, this lacks
the precision of the waterflood definition, but from the practical view of recovering
oil economically, this distinction is not serious. By this definition, oil severely
bypassed by a viscous finger is residual oil. Since we discussed capillary-trapped
residual oil earlier in Sec. 3.4, we discuss other causes here.
294
Local Heterogeneity
To investigate the effects of local heterogeneity on trapped oil saturation, researchers

have conducted experiments in laboratory cores on first-contact miscible
displacements (Raimondi and Torcaso, 1964; Stalkup, 1970; Shelton and Schneider,
1975; Spence and Watkins, 1980). In these experiments, viscous fingering was
suppressed by gravity stabilization or by matching the viscosity and density of the
displacing and displaced fluids.
The miscible flood residual oil was found to depend on several things, the
most important of which is the presence of a high mobile water saturation. A sample
of this experimental data (Fig. 7-43) plots trapped oil saturation S2' r , normalized by
the waterflood residual oil saturation, versus the steady-state flowing water
saturation. The normalized trapped oil saturations approach unity at high water
saturations in these data and are close to zero for water saturations less than 50%.
The steepness of the curves and the magnitudes of the residual saturations at high
water saturation are of concern in displacements where a high water saturation is
present (tertiary floods or WAG floods). The data in Fig. 7-43 were from
displacements in strongly water-wet media. In oil-wet or intermediate-wet media, the
trapping is not nearly as pronounced. Thus the trapped oil saturation has been
correlated with capillary pressure curve hysteresis (Shelton and Schneider, 1975), a
fractional relative permeability ratio (Raimondi and Torcaso, 1964), and
Figure 7-43 Oil trapped on imbibition as a function of water saturation (from

Raimondi and Torcaso, 1964)
295
Figure 7-44 Influence of oil bank and residual oil saturation on the total stagnant
hydrocarbon saturation (from Stalkup, 1970)
dimensionless oil bank saturation (Stalkup, 1970). Figure 7-44 shows the correlation
of trapped or stagnant oil saturation with dimensionless oil bank saturation. S2B is the
oil bank saturation determined from the graphical construction in Fig. 7-39 and
should contain corrections for the wettability of the medium since wettability is
contained in the fractional flow curves. The oil bank saturation should contain
corrections for injected water since the WAG ratio also affects the construction.
The most common interpretation for the effect of mobile water in miscible
flood trapped oil saturation is that on a microscopic basis the water shields, or blocks,
the solvent from contacting the oil. This explanation also qualitatively accounts for
the effect of wettability since the oil and water phases are, depending on the
wettability, differently distributed in the medium. In water-wet media, oil is
contained in the large pores mostly away from the rock surfaces. The water phase is
far more connected compared to the oil phase and thus could serve as a shield to oil
originally present in pores not in the main flow channels. For oil-wet media, the
phase distribution is reversed––the oil phase is the more continuous, and water is a
less effective shield.
The interpretation of water blocking stagnant pores is somewhat like the
dead-end pore model used to explain the behavior of water-free, first-contact
296
displacements. The capacitance or dead-end pore model was originally proposed to
explain the concentration “tail” observed in the breakthrough curves of first-contact,
stable miscible displacements. This tail is more pronounced in carbonates than in
sandstones (Fig. 7-45) because the pore structure of a typical carbonate is more
heterogeneous (Spence and Watkins, 1980). Mathematical solutions fit the
breakthrough curves well (Fig. 7-45) even though the physical interpretation of the
parameters in the dead-end pore model has been questioned (Coats and Smith, 1964).
Figure 7-45 Typical breakthrough curves (from Spence and Watkins, 1980)
297
The dead-end pore model also qualitatively explains other features of first-contact
miscible flood trapping, so we summarize the mathematical theory here.
Consider a stable, first-contact miscible displacement, in the absence of
water, flowing in a permeable medium where a fraction fa of the pore space is
available to flow and a fraction (1 – fa) is stagnant. Solvent can flow from or into the
stagnant or dead-end pores only by diffusion, represented by a mass-transfer
coefficient km. The conservation equation for solvent becomes in the absence of
dispersion
∂C ∂C
φ f a 32 + u 32 = − km (C32 − C3s )
∂t ∂x
(7.9-1)
∂C32
(1 − f a ) = km (C32 − C3 s )
∂t
where C32 and C3, are the solvent concentrations in the flowing and dead-end pores.
With dimensionless distance and time, these equations become
∂C ∂C
f a 32 + 32 = − N Da (C32 − C3 s )
∂t D ∂x
(7.9-2)
∂C32
(1 − f a ) = N Da (C32 − C3 s )
∂t D
where NDa = kmLφ/u is the Damkohler number, a dimensionless quantity that is a

ratio of the rates of diffusion from the dead-end pores to the bulk fluid flow. Equation
(7.9-2) is a two-parameter (fa and NDa) representation of flow without dispersion.
Deans (1963) gives the analytic solution to Eq. (7.9-2) subject to a step change in
influent solvent concentration
C32 ⎧0, Z ≤0
=⎨ − Z Y −ξ (7.9-3)
CeJ ⎩1 − e ∫0 e − I 0 (2 ξ Z ) dξ , Z ≥0
where Z = NDa(tD – xDfa)/(1 – fa), Y = NDaxD, and I0 is the modified Bessel function of
the first kind, zero order. Equation (7.9-3) says the solvent concentration changes
abruptly from zero to C32/C3J = e–Y at Z = 0.
The solvent effluent history (at xD = 1) is from Eq. (7.9-3)
C32 ⎧0, tD ≤ f a
=⎨ − Z N Da −ξ (7.9-4)
CeJ ⎩1 − e ∫0 e − I 0 (2 ξ Z )dξ , tD ≥ f a
Figure 7-46 plots Eq. (7.9-4) for fixed fa and various NDa. For very small NDa, the
breakthrough curve behaves normally with the pore space contracted by (1 – fa). For
this case, the miscible flood trapped oil saturation would simply be (1 – fa) times the
oil saturation in the dead-end pores since the solvent cannot enter the stagnant pores.
But for very large NDa, the effect of the stagnant pore space vanishes since mass
transfer to and from the flowing fraction is rapid. In this extreme, the trapped oil
saturation should vanish.
These observations partly explain the dependence of miscible flood trapped
298
Figure 7-46 Effluent solvent concentration

for fixed flowing fraction fa and various
NDa; no dispersion
oil saturation on velocity and system length. As suggested by the definition of the
Damkohler number, the trapped oil saturation should decrease with decreasing
velocity and increasing system length. At field-scale conditions, large length and
small velocity, the Damkohler number is usually much larger than in a laboratory
experiment. Thus laboratory experiments may be overestimating miscible flood
trapped oil saturation.
Including dispersion in Eq. (7.9-1) requires a numerical solution (Coats and
Smith, 1964). Of course, the solutions so obtained fit experimental data better than
Eq. (7.9-4) but do not alter the general conclusions.
The effect of water blocking is difficult to see from the preceding
mathematics. For conceptual clarity, it is best to separate the water-blocking and
dead-end pore effects by dividing the permeable medium pore space into flowing,
isolated, and dendritic fractions (Salter and Mohanty, 1982). The flowing pore space
is the fraction through which a phase flows into and from at least one pore throat. The
dendritic fraction is connected to the flowing fraction through a mass transfer
coefficient as above but does not exhibit flow itself. The isolated fraction of a phase
is completely surrounded by the other phase through which no diffusion can occur.
The amounts and properties of all fractions are functions of the phase saturations, the
wettability of the medium, and the saturation history. Generally, the isolated and
dendritic fractions vanish as the nonwetting phase saturation increases. But these two
nonflowing fractions can occupy most of the total pore space at low nonwetting
phase saturations.
Phase Behavior Interference
When the miscibility of a displacement is developed, the analysis is considerably

complicated because, besides the water-blocking effect, a solvent flood can now trap
oil by interactions with the phase behavior. Fig. 7-47 gives results from a combined
experimental and theoretical study of Gardner et al. (1981) that shows the results of
CO2 displacements at two different pressure and dispersion levels. At both pressures,
299
Figure 7-47 Results of CO2 displacements at two different pressure and dispersion levels (from
300
the displacements are vaporizing gas drives. Still, the lower pressure gives a
measurably lower oil recovery than the higher pressure. The effect is relatively
insensitive to rate, and there was no mobile water, indicating the lower recovery is
caused by something more than the dead-end pore effect.
Figure 7-47(b) shows the composition route for the 13.6 MPa (2,000 psia)
displacement in Fig. 7-47(a). Dispersion causes the composition route for this
developed miscibility displacement to enter the two-phase region (compare this to the
no-dispersion extreme in Fig. 7-37b). This intrusion will lower oil recovery because
the trapped phase saturations within the two-phase region are large, the interfacial
tension between the two hydrocarbon phases being large. Though the effect of
dispersion on the experimental data (“low” dispersion level) is relatively minor, the
simulated effect at the high dispersion level is pronounced.
The displacements in Fig. 7-47 were gravity stabilized so that it would be
proper to ignore viscous fingering. That this phenomenon also contributes to the
trapped oil saturation in an unstable displacement is demonstrated by the work of
Gardner and Ypma (1982). Figure 7-48 shows literature data on trapped miscible oil
saturation plotted versus residence (Lφ/u) time for several secondary CO2 floods. The
decrease in trapped oil saturation with residence time is very much like the decrease
associated with increasing NDa in the first-contact miscible floods discussed earlier.
But the displacements in Fig. 7-48 were generally not stable, and there was no mobile
water present.
Figure 7-48 Literature data on trapped miscible flood oil saturation versus residence
time (from Gardner and Ypma, 1982)
301
Gardner and Ypma interpret the large residual oil saturations at small
residence times to be the consequence of a synergistic effect between the phase
behavior and viscous fingering. They argue that in the longitudinal direction at the tip
of the viscous finger, miscibility between the solvent and crude oil develops much
like that shown in Fig. 7-37b. In the transverse direction, mixing takes place because
of transverse dispersion and, perhaps, viscous crossflow. As we have seen, mixing
due to dispersion causes straight-line dilution paths on pseudoternary diagrams (see
Fig. 7-13). Such mixing does not cause developed miscibility unless very long
residence times or very high transverse dispersion is allowed. Thus oil is first swept
out by the longitudinal movement of a finger, the tip of which contains the
light-enriched CO2 solvent, and then reflows back into the finger from the transverse
direction into a region of pure CO2. Since CO2 and crude are not first-contact
miscible, multiple phases form in the finger, and trapping occurs. In fact, in
simulations, it was observed that the trapped oil was actually present in highest
amounts in the regions where the solvent fingers had passed because of this
resaturation and phase behavior effect. Though this seems paradoxical––that the
largest remaining oil saturation is where the solvent has swept––the contention is
supported by correlating the data in Fig. 7-48 against a transverse dispersion group,
reproducing this correlation with simulation, and finally, matching the effluent
history of laboratory floods with the simulation results. Interestingly, the composition
routes of zones both inside and outside the fingers passed well into the interior of the
two-phase region of the ternary. When transverse dispersion is large, the transverse
Figure 7-49 Oil recovery versus injected water fraction for teriary CO2
displacements in water-wet and oil-wet media (Tiffin and Yellig, 1982)
302
mixing takes place before the solvent fingers have emptied of the displacing mixture,
and trapped oil saturation goes down.
Undoubtedly, the interaction with phase behavior, dispersion, and viscous
fingering all play a part in understanding these complex phenomena. Still, it seems
persuasive that the wettability of the medium plays a central role, particularly since
there seems to be a wettability effect in even the most complicated developed
miscible, unstable, displacements (Fig. 7-49).
7-10 ESTIMATING FIELD RECOVERY
In this section, we combine the effects of areal sweep efficiency and displacement
efficiency.
Assume we have a plot of average solvent and oil concentration versus
dimensionless time in a one-dimensional displacement. This can be from an overall
material balance of a laboratory experiment or from the fractional flow calculation in
Secs. 7-7 and 7-8. Figure 7-52 shows the average concentrations from the
experimental data in Fig. 7-40. The solid lines are the fractional flow solution.
Figure 7-50 Schematic illustration of

contacted and invaded area in quarter
5-spot pattern
In this section, we illustrate the correction of this data for areal sweep
efficiency only. The procedure for correcting for vertical sweep is similar except we
must now use a volumetric sweep efficiency function rather than an areal sweep
efficiency function. The correction based on areal sweep would also be correct if the
average concentration curves are corrected for vertical sweep, that is, were they
averaged over a cross section using pseudofunctions.
Since we are explicitly including viscous fingering in the average
concentration function, it is important not to include it in the areal sweep correlation
also. Claridge anticipated this event by defining an “invaded area” sweep efficiency
303
as shown in Fig. 7-51. He determined that the areal sweep correlation of Caudle and
Witte (1959) most nearly approximated the invaded area sweep and derived
equations to describe it for flow in a confined five-spot. He also gave a procedure for
combining areal and displacement sweep for secondary, non-WAG displacements.
Our procedure is a generalization of Claridge’s to first-contact floods of arbitrary
WAG ratios and arbitrary initial conditions.
Figure 7-51 Schematic of the behavior

of average concentrations
The invaded area is defined by a curve connecting the extreme tips of the
viscous fingers (Fig. 7-50) and given by the product of EA, the invaded areal sweep
efficiency, and the pattern area. EA as a function of dimensionless time and mobility
ratio is given in Claridge’s paper; it is not repeated here though it could be given
graphically for a particular case. In Fig. 7-50, the contacted area is that actually
occupied by the solvent fingers.
Central to the procedure is the idea of average concentrations behind the
front. We define these to be the average concentrations in the invaded zone
Volume of component i in invaded zone

Ci = (7.10-1)
Volume of invaded zone
304
The average concentrations in the contacted area are (C1, C2, C3)J, the injected
concentrations. The Ci are equal to the average concentration functions after break-
through; before this, they are constant and equal to their breakthrough value Ci0 (see
Fig. 7-52).
Figure 7-52 Average concentration from the experimental displacement in Fig. 7-40
Since EA and Ci are both known functions of time, the cumulative production of
component i is
t
N pi = ∫ FiJ qdt − V p E A (Ci − Cij ), i = 1, 2, 3 (7.10-2)
0
from an overall material balance (Eq. 2.5-2). In Eq. (7.10-2), q is the injection–
production rate, Vp is the total pore volume, and t is time. All volumes in this
equation are in reservoir volumes. For oil, in particular, we can write
⎛ C ⎞
N p 2 D = E A ⎜1 − 2 ⎟ (7.10-3)
⎝ C2 I ⎠
where Np2D is the cumulative oil produced expressed as a fraction of oil in place at the
start of solvent injection (Np2/VpC2I). To express oil recovery as a fraction of original
(at discovery) oil in place, Eq. (7.10-3) should be multiplied by the ratio of C2I to (1 –
S1r), the original oil saturation.
We begin here to use tD1 as the time variable for the Ci, and tD2 for EA since,
in general, neither function depends explicitly on the actual dimensionless time tD in
Eq. (7.10-3). We relate tD1, tD2, and tD to one another below.
Breakthrough occurs at tD1 = (1 + WR) C30 where C30 is the average solvent
concentration behind the front at or before breakthrough. In Fig. 7-52, plotted curves
show the average concentration in the one-dimensional system versus tD1 (dotted
lines). In the following development, we do not use system average concentrations;
305
we show them in Fig. 7-52 for completeness. Average system concentrations and
average concentrations behind the front coincide after breakthrough.
Imagine a continuous one-dimensional permeable medium with C2 = C2 (tD1)
and C3 = C3 (tD1) known. An appropriate definition for tD1 is
Volume solvent + Water injected

t D1 = (7.10-4)
Volume invaded
If we identify the flow-excluded regions with the uninvaded regions in Fig. 7-51, the
dimensionless time tD1 becomes
t
t D1 = D (7.10-5)
EA
On the other hand, consider a homogeneous five-spot pattern with EA =
EA(tD2) known, into which solvent and water are being simultaneously injected. If the
oil and water in the invaded region are regarded as part of the rock matrix, the
appropriate dimensionless time tD2 becomes
Volume solvent injected
tD 2 = (7.10-6)
Volume solvent in invaded region
which may be decomposed into
Volume solvent injected
tD 2 =
× (7.10-7)
Volume of pattern
Volume of pattern
×
Volume solvent in invaded region
After breakthrough, tD2 may be written in our terminology as

tD
tD 2 = (7.10-8)
(1 + WR )C3
Equations (7.10-5) and (7.10-8) are the relations among the various dimensionless
times. Claridge calls tD1 and tD2 the apparent pore volumes injected for the
appropriate variable. tD may be eliminated between Eqs. (7.10-5) and (7.10-8) to give
t
t D1 = D 2 (1 + WR )C3 (7.10-9)
EA
The definitions in Eqs. (7.10-5) and (7.10-8) may be verified by observing

that when breakthrough happens in an areal sense, it also happens in a
one-dimensional sense. Thus at breakthrough, we have tD2 = EA and C3 = C30 from
which it follows that tD1 = (1 + WR) C30 from Eq. (7.10-9). Figure 7-52 shows this is
indeed the correct dimensionless breakthrough time for the one-dimensional system.
The procedure to calculate the correct tD1, tD2, and tD is iterative.
306
1. Estimate the mobility ratio M to be used in the areal sweep correlation. We
take this to be the mobility ratio based on the average concentrations behind
the front at breakthrough
(λrt )C 0
M= i
(7.10-10a)
(λrt )CiI
This requires a knowledge of the relative permeability curves. If these are not
available, M may be estimated from the one-dimensional data as
(q / ΔP )final
M≅ (7.10-10b)
(q / ΔP ) I
where q = total volumetric rate, and ΔP = overall pressure drop. The
numerical value of M does not change during the calculation.
2. For this value of M , find the breakthrough areal sweep efficiency t D0 2 = E A0 .
The dimensionless breakthrough time for the combined areal and
displacement sweep is t D0 = E A0 (1 + WR )C30 from Eq. (7.10-8). The iterative
calculations begin at t D0 .
3. Fix tD > t D0 2
4. Pick tD1 > C30 (1 + WR).
5. Calculate C3 (tD1) from the one-dimensional results.
6. Calculate tD2 from Eq. (7.10-8).
7. Estimate EA(tD2) from the areal sweep correlation.
8. Calculate tD1 from Eq. (7.10-5).
9. Test for convergence. If the tD1 estimated in steps 4 and 8 differ by less than
some small preset tolerance, the procedure has ~converged; if not, reestimate
tD1, and return to step 4.
10. Calculate cumulative oil produced from Eq. (7.10-3), and calculate the
combined fraction flow of each component from
⎛ dE ⎞ dE
Fi T = ⎜ 1 − A ⎟ Fi (t D1 ) + A FiJ (7.10-11)
⎝ dt D 2 ⎠ dt D 2
11. Increment tD, and return to step 3 for a later time. The entire procedure
continues until tD is larger than some preset maximum. The procedure
converges in two to four iterations per step by simple direct substitution. The
combined fractional flow in Eq. (7.10-11) represents contributions from the
invaded zone (first term) and the uninvaded zone (second term), with all
expressions being evaluated at consistent values of tD1, tD2, tD. The Fi terms in
Eq. (7.10-11) are from the one-dimensional curves, and the derivatives are
numerically evaluated. Once Fi T (t D ) is known, we calculate component rates
in standard volumes as
qF T
qi = i , i = 1, 2, 3 (7.10-12a)
Bi
307
corresponding to a real time t by inverting
t qdt
tD = ∫ (7.10-12b)
0 V
P
Figure 7-53 shows the results of the corrections for areal sweep applied to the
data in Fig. 7-52. The y-axis plots cumulative oil produced as a fraction of initial oil
in place at the start of solvent injection, and the x-axis plots each of the three
dimensionless times. The combined areal and displacement sweep case breaks
through earlier than the other two and, except for early time, is everywhere smaller.
The combined fractional oil recovery at a particular dimensionless time is not simply
the product of displacement and areal sweep at that time. The correct dimensionless
time for consistently evaluating the latter two is given by Eqs. (7.10-5) and (7.10-8).
For this particular case, Fig. 7-53 indicates the combined oil recovery is roughly
equally dependent on areal and displacement sweep efficiencies.
Figure 7-53 Calculated cumulative oil produced
Solvent methods currently occupy a large fraction of implemented EOR methods. For
certain classes of reservoirs––low permeability, fairly deep, and with light oil––they
are clearly the method of choice. Future technology, particularly related to gravity
stabilization and mobility control methods, could expand this range somewhat, but
the target oil is nevertheless immense.
The topics of special importance in this chapter are the solvent flooding
classifications, the usefulness of the minimum miscibility pressure correlations, and
viscous fingering. The importance of viscous fingering remains largely unappreciated
in large-scale displacement because of the obscuring effects of heterogeneity;
308
however, it is undoubtedly true that this phenomenon, perhaps in conjunction with
others, accounts for the large discrepancy between lab-scale and field-scale oil
recoveries. The material on dispersion and slugs and on solvent-water-oil fractional
flow can form the basis for many design procedures. Of course, both topics easily
lend themselves to the graphical presentation which is an essential part of this text.
EXERCISES
7A. Immiscible Solvent. A particular crude oil has a specific gravity of 0.76, normal boiling
point of 324 K (124°F), molecular weight of 210 kg/kg-mole, and viscosity of 15
mPa-s. At 8.16 MPa (1,200 psia) and 322 K (120°F), estimate
(a) The CO2 solubility in the oil
(b) The viscosity of the saturated CO2–crude-oil mixture
(c) The swelling factor of the mixture
(d) The CO2 water solubility, and express this as a mole fraction
Use the Simon and Graue correlations (Figs. 7-20 through 7-22) and the water
solubility correlations (Fig. 7-23).
7B. Calculating Minimum Miscibility Pressure. An analysis of a particular separator oil is
given below, including analyses at two different solution gas levels. Using the 1982
Holm and Josendal correlation (Fig. 7-25), estimate the minimum miscibility pressure
(MMP) for the separator oil and the oil with 53.4 and 106.9 SCM dissolved gas/SCM
dissolved oil. The reservoir temperature is 344 K (160°F).
Weight percent
Component Separator Oil + Oil +
oil 53.4 SCM gas/SCM oil 106.9 SCM gas/SCM oil
C1 21.3 53.0
C2 7.4 18.4
C3 6.1 15.1
C4 2.4 6.0
C5–C30 86 54.0 6.5
C31+ 41 8.8 1.1
What can you conclude about the effect of solution gas on the MMP? How would you
explain this with a ternary diagram?
7C. Superposition and Multiple Slugs. Using the principle of superposition applied to M
influent step changes to a one-dimensional medium, show the composite solution to the
convective–diffusion equation is
Ci 0 − CiM 1 M
⎛ j
⎞
Ci =
2
−
2
∑ (Cij − Cij −1 )φ ⎜ t D −
⎝
∑t Dk ⎟
⎠
(7C-1)
j =1 k =1
where Cij = injected concentration of component i during time interval j (Ci0 is the
same as CiI), and tDj = duration of interval j, and where
309
⎧ xD − t D
φ (t D ) = erf ⎪
⎫
⎪
⎨ 2 tD ⎬
⎪ ⎪
⎩ N Pe ⎭
j
Eq. (7C-1) is valid only for tD > ∑ t Dk .
k =1
7D. Dilution Paths on Ternary Diagrams. Plot for the following:
(a) Concentration profiles at tD = 0.5 for the displacement of an oil of composition
(C2, C3) = (0.1, 0) by a small slug (tDs = 0.1) of composition C2J = 1.0, which is
then followed by a chase gas of composition C3K = 1.0. Take the Peclet number to
be 100.
(b) The dilution path of the concentration profile in part (a) on a ternary diagram as in
Fig. 7-30.
7E. Rich Gas Dilution. Based on the ternary diagram in Fig. 7E with initial oil composition
(C2, C3)I = (0.1, 0),
(a) Determine the minimum intermediate component concentration (C2J) that may be
used in a continuous mixture of dry gas and intermediate displacing fluid that will
ensure developed miscibility.
(b) Using the C2J of part (a) as a lower bound, estimate the solvent slug size necessary
to ensure first-contact miscibility at tD = 1 for a series of C2J values. Plot the total
amount of intermediate injected (C2JtDs) versus the slug size to determine an
optimum. Take the Peclet number to be 1,000.
Figure 7E Ternary diagram for rich gas design problem
7F. Fractional Flow Solution of Immiscible Displacement. The fractional flow curves
along the three tie lines in Fig. 7E are shown in Fig. 7F. The straighter curves (with the
smaller residual phase saturations) are nearer to the plait point. Phase 3 is that richest in
component 3.
310
Figure 7F Fractional flow curve for

Exercise 7F
(a) On the ternary diagram, sketch residual saturation lines, the singular curve(s), the
equivelocity path, and as many non-tie line paths as possible.
(b) Plot all the possible fractional flux curves you can. The initial oil composition is
(C2, C3)I = (0.28, 0), and the injected solvent composition is (C2, C3)J = (0.09,
0.91). These compositions are on extensions of the lines farthest and nearest the
plait point, respectively.
(c) Pick the physically possible solutions from the curves of part (b), and plot
saturation and concentration profiles at tD = 0.8.
7G. WAG Calculations. Figure 7G gives representative relative permeability curves for the
Slaughter Estate Unit (SEU). The water, oil, and solvent viscosities are 0.5, 0.38, and
0.037 mPa-s, respectively.
(a) Plot the water–oil and water–solvent fractional flow curves. Assume the relative
permeability curves for these pairs are the same and take α = 0.
(b) Determine the optimal WAG ratio for a first-contact miscible secondary
displacement in the absence of viscous fingering and dispersion.
(c) If the optimal WAG ratio is used, calculate the minimum solvent–water slug size
(tDs) for complete displacement. The chase fluid is water.
(d) If the solvent–water slug size is 50% greater than that calculated in part (c), plot
the time–distance diagram and effluent history for this displacement.
'
(e) Estimate the miscible flood trapped oil saturation S 2r from Fig. 7-44.
7H. Solvent Velocity with Water-Oil Solubility
(a) Show that by including the solvent water solubility and the solubility of the sol-
vent in a trapped oil saturation, the solvent specific velocity (Eq. 7.7-14) becomes
s
1 − f1 J (1 − C31 )
v3 = '
(7H-1)
1 − S1 J (1 − C31 ) − S 2 r (1 − C32 )
where C31 = solvent solubility in water = R31B3/B1, and C32 = solvent solubility in
oil = R32B3/B2. Rij is the solubility of component i in phase j in standard volumes of
i per standard volumes of j. See Fig. 7-20 and 7-23.
311
Figure 7G Slaughter Estate Unit relative

permeability curves (from Ader and Stein,
1982)
'
(b) Using the S 2r from part (e) of Exercise 7G, and taking R31 = 17.8 SCM/SCM, R32
= 214 SCM/SCM, B3 = 10–3 m3/SCM, B1 = 1 m3/SCM, and B2 = 1.2 m3/SCM,
repeat parts (b–d) of Exercise 7G.
(c) Repeat parts (c) and (d) of Exercise 7G if the chase fluid is a gas having the
identical properties of the solvent instead of water.
71. Carbonated Waterflooding Fractional Flow. One of the earlier EOR techniques is
displacement by CO2-saturated water. This technique is amenable to fractional flow
analysis (de Nevers, 1964).
(a) Show that the specific velocity of a pistonlike carbonated water front is given by
3
K 21
1− 3
K 21 − 1
vΔC3 = 3
(7I-1)
K 21
1 − S2 r − 3
K 21 − 1
Equation (7I-1) assumes flow behind the front is at a CO2-saturated residual oil
phase.
(b) By matching the specific velocity of the oil bank rear to Eq. (7I-1) show the oil
bank saturation and fractional flow must satisfy
312
f1 − C32
1−
1 − C32
vΔC3 = (7I-2)
S1 − C32
1−
1 − C32
3
In these equations, K 21 is the volumetric partition coefficient of CO2 (i = 3)
between the water (j = 1) and oil (j = 2) phases, and C32 is the volume fraction of
CO2 in the oil. f1(S1) is the water fractional flow curve.
3
(c) Estimate C32 and K 21 from Fig. 7-20 at 15 MPa and 340 K. You may assume ideal
mixing in both phases.
(d) Calculate and plot the effluent oil cut of a carbonated waterflood in a
one-dimensional permeable medium with initial (uniform) oil cut of 0.1.
(e) On the same graph, plot the effluent oil fractional flow of a noncarbonated
waterflood. Finally, plot the incremental oil recovery (IOR) versus tD.
For this problem, use the following parameters in the exponential relative
0 0
permeability curves: n1 = n2 = 2, k r1 = 0.1, φ = 0.2, k r 2 = 0.8, μ1 = 0.8 mPa-s, μ2 =
5 mPa-s, S1r = S2r = 0.2, and α = 0. The oil molecular weight is 200 kg/kg-mole, its
density is 0.78 g/cm3, and the UOP factor is 11.2.
7J. Viscous Fingering and Displacement Efficiency. Using the Koval theory (Eq. 7.8-7),
plot the effluent history of a first-contact miscible displacement where the oil–solvent
viscosity ratio is 50, and the heterogeneity factor is 5.
7K. Viscous Fingering by Mixing Parameter. In the Todd-Longstaff (1972) representation
of viscous fingering, the Koval factor Kval in Eq. (7.8-4) is replaced by KTL where
M 1−ω
K TL = 2 e = v (7K-1)
M 3e
where M2e and M3e = effective solvent and oil viscosities in the mixing zone, v =
viscosity ratio, and ω = mixing parameter (0 < ω < 1).
(a) Repeat Exercise 7J with ω = 1/3.
(b) Determine the correspondence between Kval and KTL by setting KTL = Kval in Eq.
(7.8-5) and plotting ω versus v for various Hk.
7L. Dispersion as a Normal Distribution. One view of dispersion is that it is the result of
the mixing of a large number of fluid particles along independent paths. If so, the
distribution of particles should follow a normal distribution. In this exercise, we show
that the equations in Sec. 7-6 reduce to such a form.
(a) Show that Eq. (7.6-4) applied to a unit slug CiI = CiK = 0 and
t Ds CiJ = 1 (7L-1)
reduces to
1 ⎧ ⎡ xD − (t D − t Ds ) ⎤ − erf ⎡ xD − t D ⎤⎫
Ci = ⎪erf ⎢
2t Ds ⎨ ⎥ ⎢ ⎥⎪ (7L-2)
⎥⎬
t t
⎢ 2 D ⎥ ⎢2 D
⎪ ⎢ ⎥⎪
⎩ ⎣ N Pe ⎦⎥ ⎣⎢ N Pe ⎦⎭
for tD >> tDs.
(b) Using the definition for the error function (Eq. 5.5-14), show that Eq. (7L-2)
becomes
313
1/ 2
⎛ N Pe ⎞ −[( x 2
− t D ) /( 4 t D / N pe )]
Ci = ⎜ ⎟ e
D
(7L-3)
⎝ 4π t D ⎠
as tDs → 0. Equation (7L-3) says the distribution of a large number of particles at

xD = 0 initially approaches a normal distribution with mean position xD = tD and a
standard deviation of 2 t D / N Pe .
7M. Calculating Solvent Oil Recovery. Figure 7M shows the volumetric sweep efficiency of
a tertiary solvent displacement.
(a) Using the procedure in Sec. 7-10, estimate and plot cumulative oil recovery
(fraction of oil in place at start of solvent injection) and oil cut versus
dimensionless time. Use the average concentrations of Fig. 7-50.
(b) If the oil formation volume factor is 1.2 m3/SCM, the reservoir pore volume is 160
hm3, and the average injection rate is 80 m3/day, calculate and plot the cumulative
oil produced and oil rate versus time.
Figure 7M Volumetric sweep efficiency for miscible displacement

314
Polymer Methods
Polymer flooding consists of adding polymer to the water of a waterflood to decrease

its mobility. The resulting increase in viscosity, as well as a decrease in aqueous
phase permeability that occurs with some polymers, causes a lower mobility ratio.
This lowering increases the efficiency of the waterflood through greater volumetric
sweep efficiency and a lower swept zone oil saturation. Irreducible oil saturation does
not decrease although the remaining oil saturation does, approaching S2r for both
waterflooding and polymer flooding. The greater recovery efficiency constitutes the
economic incentive for polymer flooding when applicable. Generally, a polymer
flood will be economic only when the waterflood mobility ratio is high, the reservoir
heterogeneity is high, or a combination of these two occurs.
Polymers have been used in oil production in three modes.
1. As near-well treatments to improve the performance of water injectors or

watered-out producers by blocking off high-conductivity zones.
2. As agents that may be cross-linked in situ to plug high-conductivity zones at
depth in the reservoir (Needham et al., 1974).
These processes require that polymer be injected with an inorganic metal cation
that will cross-link subsequently injected polymer molecules with ones already bound
to solid surfaces.
3. As agents to lower water mobility or water–oil mobility ratio.
The first mode is not truly polymer flooding since the actual oil-displacing agent is
not the polymer. Certainly most polymer EOR projects have been in the third mode,
the one we emphasize here. We discussed how lowering the mobility ratio affects
displacement and volumetric sweep efficiency in Chaps. 5 and 6.
315
Figure 8-1 Schematic illustration of polymer flooding sequence (drawing by Joe Lindley,
U.S. Department of Energy, Bartlesville, Okla.)
316
Figure 8-1 shows a schematic of a typical polymer flood injection sequence:
a preflush usually consisting of a low-salinity brine; an oil bank; the polymer solution
itself; a freshwater buffer to protect the polymer solution from backside dilution; and
finally, chase or drive water. Many times the buffer contains polymer in decreasing
amounts (a grading or taper) to lessen the unfavorable mobility ratio between the
chase water and the polymer solution. Because of the driving nature of the process,
polymer floods are always done through separate sets of injection and production
wells.
Mobility is lowered in a polymer flood by injecting water that contains a high
molecular weight, water- soluble polymer. Since the water is usually a dilution of an
oil-field brine, interactions with salinity are important, particularly for certain classes
of polymers.
Figure 8-2 Salinities from representative oil-field brines (from Gash et al., 1981)
317
Salinity is the total dissolved solids (TDS) content of the aqueous phase.
Figure 8-2 shows typical values. Virtually all chemical flooding properties depend on
the concentrations of specific ions rather than salinity only. The aqueous phase’s total
divalent cation content (hardness) is usually more critical to chemical flood
properties than the same TDS concentration. Figure 8-2 also shows typical brine
hardnesses.
Because of the high molecular weight (1 to 3 million), only a small amount
(about 500 g/m3) of polymer will bring about a substantial increase in water
viscosity. Further, several types of polymers lower mobility by reducing water
relative permeability in addition to increasing the water viscosity. How polymer
lowers mobility, and the interactions with salinity, can be qualitatively illustrated
with some discussion of polymer chemistry.
8-1 THE POLYMERS
Several polymers have been considered for polymer flooding: Xanthan gum,
hydrolyzed polyacrylamide (HPAM), copolymers (a polymer consisting of two or
more different types of monomers) of acrylic acid and acrylamide, copolymers of
acrylamide and 2-acrylamide 2-methyl propane sulfonate (AM/AMPS),
hydroxyethylcellulose (HEC), carboxymethylhydroxyethylcellulose (CMHEC),
polyacrylamide (PAM), polyacrylic acid, glucan, dextran polyethylene oxide (PEO),
and polyvinyl alcohol. Although only the first three have actually been used in the
field, there are many potentially suitable chemicals, and some may prove to be more
effective than those now used.
Nevertheless, virtually all the commercially attractive polymers fall into two
generic classes: polyacrylamides and polysaccharides (biopolymers). In the
remainder of this discussion, we deal with these exclusively. Figure 8-3 shows
representative molecular structures.
Polyacrylamides
These are polymers whose monomeric unit is the acrylarnide molecule. As used in
polymer flooding, polyacrylamides have undergone partial hydrolysis, which causes
anionic (negatively charged) carboxyl groups (––COO–) to be scattered along the
backbone chain. The polymers are called partially hydrolyzed polyacrylamides
(HPAM) for this reason. Typical degrees of hydrolysis are 30%–35% of the
acrylamide monomers; hence the HPAM molecule is negatively charged, which
accounts for many of its physical properties.
This degree of hydrolysis has been selected to optimize certain properties
such as water solubility, viscosity, and retention. If hydrolysis is too small, the
polymer will not be water soluble. If it is too large, its properties will be too sensitive
to salinity and hardness (Shupe, 1981).
The viscosity increasing feature of HPAM lies in its large molecular weight.
318
Figure 8-3 Molecular structures (from Willhite and Dominguez, 1977)
This feature is accentuated by the anionic repulsion between polymer molecules and
between segments on the same molecule. The repulsion causes the molecule in
solution to elongate and snag on others similarly elongated, an effect that accentuates
the mobility reduction at higher concentrations.
If the brine salinity or hardness is high, this repulsion is greatly decreased
through ionic shielding since the freely rotating carbon–carbon bonds (Fig. 8-3a)
allow the molecule to coil up. The shielding causes a corresponding decrease in the
319
effectiveness of the polymer since snagging is greatly reduced. Virtually all HPAM
properties show a large sensitivity to salinity and hardness, an obstacle to using
HPAM in many reservoirs. On the other hand, HPAM is inexpensive and relatively
resistant to bacterial attack, and it exhibits permanent permeability reduction.
Polysaccharides
These polymers are formed from the polymerization of saccharide molecules (Fig.
8-3b), a bacterial fermentation process. This process leaves substantial debris in the
polymer product that must be removed before the polymer is injected (Wellington,
1980). The polymer is also susceptible to bacterial attack after it has been introduced
into the reservoir. These disadvantages are offset by the insensitivity of
polysaccharide properties to brine salinity and hardness.
Figure 8-3(b) shows the origin of this insensitivity. The polysaccharide
molecule is relatively nonionic and, therefore, free of the ionic shielding effects of
HPAM. Polysaccharides are more branched than HPAM, and the oxygen-ringed
carbon bond does not rotate fully; hence the molecule increases brine viscosity by
snagging and adding a more rigid structure to the solution. Polysaccharides do not
exhibit permeability reduction. Molecular weights of polysaccharides. are generally
around 2 million.
Today, HPAM is less expensive per unit amount than polysaccharides, but
when compared on a unit amount of mobility reduction, particularly at high salinities,
the costs are close enough so that the preferred polymer for a given application is site
specific. Historically, HPAM has been used in about 95% of the reported field
polymer floods (Manning et al., 1983). Both classes of polymers tend to chemically
degrade at elevated temperatures.
Polymer Forms
The above polymers take on three distinctly different physical forms: powders, broths
and emulsions. Powders, the oldest of the three, can be readily transported and stored
with small cost. They are difficult to mix because the first water contacting the
polymer tends to form very viscous layers of hydration around the particles, which
greatly slow subsequent dissolution- Broths arc aqueous suspensions of about 10 wt.
% polymer in water which are much easier to mix than powders. They have the
disadvantage of being rather costly because of the need to transport and store large
volumes of water. Broths are quite viscous so they can require special mixing
facilities. In fact, it is this difficulty which limits the concentration of polymer in the
broth. Emulsion polymers, the newest polymer form, contain up to 35 wt. % polymer
solution, suspended through the use of a surfactant, in an oil-carrier phase. Once this
water-in-oil emulsion is inverted (see Fig. 9-5), the polymer concentrate can be
mixed with make-up water to the desired concentration for injection. The emulsion
flows with roughly the same viscosity as the oil carrier, which can be recycled.
320
B-2 POLYMER PROPERTIES
In this section, we present qualitative trends, quantitative relations, and representative

data on the following properties: viscosity relations, non-Newtonian effects, polymer
transport, inaccessible pore volume, permeability reduction, chemical and biological
degradation, and mechanical degradation.
Viscosity Relations
Figure 8-4 shows a plot of Xanflood viscosity versus polymer concentration. This
type of curve has traditionally been modeled by the Flory-Huggins equation (Flory,
1953)
μ1′ = μ1[1 + a1C41 + a2C412 + a3C413 + ⋅ ⋅ ⋅ + ] (8.2-1)
where C41 is the polymer concentration in the aqueous phase, μ1 is the brine (solvent)
viscosity, and a1, a2, and so on are constants. In the remainder of this chapter we drop
the second subscript 1 on the polymer concentration since polymer is always in an
aqueous phase. The usual polymer concentration unit is g/m3 of solution, which is
approximately the same as ppm. The linear term in Eq. (8.2-1) accounts for the dilute
range where the polymer molecules act independently (without entanglements). For
most purposes, Eq. (8.2-1) can usually be truncated at the cubic term.
For a 1,000 g/m3 Xanflood solution at 0.1 s–1 in 1 wt % NaCl brine at 24°C,
Figure 8-4 Xanflood viscosity versus

concentration in 1% NaCl brine (from
Tsaur, 1978)
321
the viscosity is 70 mPa-s (70 cp) from Fig. 8-4. Compared to the brine at the same
conditions, this is a substantial increase in viscosity brought about by a relatively
dilute concentration (recall that 1,000 g/m3 = 0.1 wt %). Xanflood at these conditions
is an excellent thickener.
A more fundamental way of measuring the thickening power of a polymer is
through its intrinsic viscosity, defined as
⎡ μ ′ − μ1 ⎤
[ μ ] = limit ⎢ 1 ⎥ (8.2-2)
C4 → 0
⎣ μ1C4 ⎦
From its definition, [μ] is a measure of the polymer’s intrinsic thickening power. It is
insensitive to the polymer concentration. The intrinsic viscosity for the Xanflood
polymer under the conditions given above is 70 dl/g, the units being equivalent to
reciprocal weight percent. Intrinsic viscosity is the same as the a1 term in Eq. (8.2-1).
For any given polymer–solvent pair, the intrinsic viscosity increases as the
molecular weight of the polymer increases according to the following equation
(Flory, 1953):
[ μ ] = K ′M wa (8.2-3)
The exponent varies between about 0.5 and 1.5 and is higher for good solvents such
as freshwater. K′ is a polymer-specific constant.
The above relationships are useful for characterizing the polymer solutions.
For example, the size of the polymer molecules in solution can be estimated from
Flory’s (1953) equation for the mean end-to-end distance
d p = 8( M w [ μ ])1/ 3 (8.2-4)
This equation, being empirical, presumes certain units; [μ] must be in dl/g, and dp is
returned in Angstroms (10–10 m). This measure of polymer size is useful in
understanding how these very large molecules propagate through the small pore
openings of rocks. The molecular weight of Xanthan gum is about 2 million. From
Eq. (8.2-4), dp is about 0.4 μm. This is the same size as many of the pore throats in a
low-to-moderate permeability sandstone. As a result, we would expect to, and in fact
do, observe many polymer–rock interactions.
Non-Newtonian Effects
Figure 8-5 shows polymer solution viscosity μ1′ versus shear rate γ measured in a
laboratory viscometer at fixed salinity. At low shear rates, μ1′ is independent of γ
( μ1′ = 1.01), and the solution is a Newtonian fluid. At higher γ , μ1′ decreases,
approaching a limiting ( μ1′ = μ1∞ ) value not much greater than the water viscosity μ1
at some critical high shear rate. This critical shear rate is off-scale to the right in Fig.
8-5. A fluid whose viscosity decreases with increasing γ is shear thinning. The shear
thinning behavior of the polymer solution is caused by the uncoiling and unsnagging
322
Figure 8-5 Polymer solution viscosity versus shear rate and polymer concentration
(from Tsaur, 1978)
of the polymer chains when they are elongated in shear flow. Below the critical shear
rate, the behavior is part reversible.
Figure 8-6 shows a viscosity–shear-rate plot at fixed polymer concentration
with variable NaCl concentration for an AMPS polymer. The sensitivity of the
viscosity to salinity is profound. As a rule of thumb, the polymer solution viscosity
decreases a factor of 10 for every factor of 10 increase in NaCl concentration. The
viscosity of HPAM polymers and HPAM derivatives are even more sensitive to
hardness, but viscosities of polysaccharide solutions are relatively insensitive to both.
The behavior in Figs. 8-5 and 8-6 is favorable because, for the bulk of a
reservoir’s volume, γ is usually low (about 1–5 s–1), making it possible to attain a
design mobility ratio with a minimal amount of polymer. But near the injection wells,
γ can be quite high, which causes the polymer injectivity to be greater than that
expected based on μ10 . The relative magnitude of this enhanced injectivity effect can
be estimated (Sec. 8-3) once quantitative definitions of shear rate in permeable
media, and shear-rate–viscosity relations are given.
The relationship between polymer-solution viscosity and shear rate may be
described by a power-law model
n pl −1
μ1′ = K pl (γ ) (8.2-5)
323
Figure 8-6 Polymer solution viscosity versus shear rate at various brine salinities (from
Martin et al., 1981)
where Kpl and npl are the power-law coefficient and exponent, respectively. For shear
thinning fluids, 0 < npl < 1; for Newtonian fluids, npl = 1, and Kpl becomes the
viscosity. γ is always positive. Equation (8.2-5) applies only over a limited range of
shear rates: Below some low shear rate, the viscosity is constant at μ10 , and above the
critical shear rate, the viscosity is also constant μ1∞ .
The truncated nature of the power law is awkward in some calculations;
hence another useful relationship is the Meter model (Meter and Bird, 1964)
μ10 − μ1∞
μ1′ = μ1∞ + nM −1
(8.2-6)
⎛ γ ⎞
1+ ⎜ ⎟
⎝ γ 1/ 2 ⎠
where nM is an empirical coefficient, and γ 1/ 2 is the shear rate at which μ1′ is the
average of μ10 and μ1∞ . As with all polymer properties, all empirical parameters are
functions of salinity, hardness, and temperature.
When applied to permeable media flow, the above general trends and
equations continue to apply. μ1′ is usually called the apparent viscosity μapp and the
effective shear rate γ eq is based on capillary tube concepts, as we derived in Sec. 3-1
324
for Newtonian fluids. For power-law fluids, the procedure is identical (see Exercise
8B) except the beginning equation is Eq. (8.2-5). We give only the results here.
The apparent viscosity of a flowing polymer solution is
n −1
μ app = H pl u pl (8.2-7)
where (Hirasaki and Pope, 1974)
n pl −1
⎛ 1 + 3n pl ⎞ (1− n pl / 2)
H pl = K pl ⎜ ⎟⎟ (8k1φ1 ) (8.2-8)
⎜ n pl
⎝ ⎠
n −1
The right side of Eq. (8.2-7) is K pl γ eqpl which yields the equivalent shear rate for
flow of a power-law fluid
⎛ 1 + 3n pl ⎞ u
γ eq = ⎜ ⎟⎟ (8.2-9)
⎜ n
⎝ pl ⎠ 8k1φ1
In both Eqs. (8.2-8) and (8.2-9), k1, the aqueous phase permeability, is the product of
the phase’s relative permeability and the absolute permeability. φ1, the aqueous phase
porosity, is φS1.
The only difference between the equivalent shear rate and that for the
Newtonian fluid (Eq. 3.1-11) is the first term on the right-hand side. This factor is a
slowly varying function of npl; hence the Newtonian shear rate affords an excellent
approximation of the shear rate in non-Newtonian flow.
Even though γ eq has units of reciprocal time, shear rate is essentially a
steady-state representation since it can be realized in steady laminar flow in a tube.
Thus the constitutive Eqs. (8.2-5) and (8.2-6) are representing purely viscous effects
since an instantaneous change in γ eq causes a similar change in μ1′ . In reality,
fluctuations in γ eq , or elastic effects, do affect polymer properties; these we discuss
separately below.
Polymer Transport
All polymers experience retention in permeable media because of adsorption onto

solid surfaces or trapping within small pores. Polymer retention varies with polymer
type, molecular weight, rock composition, brine salinity, brine hardness, flow rate,
and temperature. Field-measured values of retention range from 7 to 150 μg
polymer/cm3 of bulk volume, with a desirable retention level being less than about 20
μg/cm3. Retention causes the loss of polymer from solution, which can also cause the
mobility control effect to be lost––a particularly pronounced effect at low polymer
concentrations. Polymer retention also causes a delay in the rate of the polymer and
generated oil bank propagation (see Sec. 8-4).
For more quantitative work, we represent polymer adsorption by a Langmuir-
type isotherm
aC
C4 s = 4 4 (8.2-10)
1 + b4C4
325
where C4 and C4s are the species concentrations in the aqueous and on the rock
phases. The units of adsorption can take on a variety of forms, but mass of polymer
per mass of rock is most common. In our notation, this is ω4s/(1 – ω4s), strictly
speaking, but ω4s is very much smaller than 1. The units conversion between
C4s(g/m3) and ω4s(μg/g-rock) are embedded in the constants a4 and b4. The b4 in Eq.
(8.2-10) controls the curvature of the isotherm, and the ratio a4/b4 determines the
plateau value for adsorption (Fig. 8-7).
Figure 8-7 Typical Langmuir isotherm shapes

326
In the original Langmuir theory, the plateau adsorption corresponded to
monolayer coverage of the surface by physical adsorption (see Exercise 8C).
Considering the anionic character of water-soluble polymers, the adsorption is more
likely to be chemical adsorption described by an exchange isotherm like Eq. (3.5-4).
In fact, polymer adsorption does increase with increasing salinity and hardness but
measured surface coverages are much smaller than monolayer coverage. Moreover, it
is unknown if adsorption is reversible. Hence Eq. (8.2-10) and Fig. 8.7 are simply
empirical representations of physical observations. (This is the origin of the term
Langmuir-type.) Typical polymer adsorption isotherms are quite steep; that is, they
attain their plateau value at very low C4. The values given above for polymer
adsorption are referring to the plateau adsorption.
Equation (8.2-10) is a general isotherm function. The specific form depends
on the units of the retention; unfortunately, no standard form exists for this. Common
ways to report retention are
Mass polymer ω4s
=
Mass solid (1 − ω 4 s )
Mass polymer ω 4 s
=
Surface area av
Mass polymer
= ω 4 s ρ s (1 − φ )
Bulk volume
Mass polymer ω 4 s ρ s (1 − φ )
= = C4 s
Pore volume φ
Volume polymer solution ω 4 s ρ s (1 − φ )
= = D4
Pore volume (φ C4 )
The last of these is often called the frontal advance loss.
Inaccessible Pore Volume
Offsetting the delay caused by retention is an acceleration of the polymer solution

through the permeable medium caused by inaccessible pore volume (IPV). The most
common explanation for IPV is that the smaller portions of the pore space will not
allow polymer molecules to enter because of their size. Thus a portion of the total
pore space is uninvaded or inaccessible to polymer, and accelerated polymer flow
results. A second explanation of IPV is based on a wall exclusion effect whereby the
polymer molecules aggregate in the center of a narrow channel (Duda et al., 1981).
The polymer fluid layer near the pore wall has a lower viscosity than the fluid in the
center, which causes an apparent fluid slip.
IPV depends on polymer molecular weight, medium permeability, porosity,
and pore size distribution and becomes more pronounced as molecular weight
increases and the ratio of permeability to porosity (characteristic pore size) decreases.
In extreme cases, IPV can be 30% of the total pore space.
327
Permeability Reduction
For many polymers, viscosity–shear-rate data derived from a viscometer ( μ1′ versus
γ ) and those derived from a flow experiment (μapp versus γ eq ) will yield essentially
the same curve. But for HPAM, the viscometer curve will be offset from the
permeable medium curve by a significant and constant amount. The polymer
evidently causes a degree of permeability reduction that reduces mobility in addition
to the viscosity increase.
Actually, permeability reduction is only one of three measures in permeable
media flow (Jennings et al., 197 1). The resistance factor RF is the ratio of the
injectivity of brine to that of a single-phase polymer solution flowing under the same
conditions
λ ⎛ k ⎞ ⎛ μ′ ⎞
RF = 1 = λ1 μ app = ⎜ 1 ⎟i⎜ 1 ⎟ (8.2-11)
λ1′ ⎝ μ1 ⎠ ⎝ k1′ ⎠
For constant flow rate experiments, RF is the inverse ratio of pressure drops; for
constant pressure drop experiments, RF is the ratio of flow rates. RF is an indication of
the total mobility lowering contribution of a polymer. To describe the permeability
reduction effect alone, a permeability reduction factor Rk is defined as
k μ
Rk = 1 = 1 RF (8.2-12)
k1′ μ1′
A final definition is the residual resistance factor RRF, which is the mobility of a
brine solution before and after (λ1a) polymer injection
λ1
RRF = (8.2-13)
λ1a
RRF indicates the permanence of the permeability reduction effect caused by the
polymer solution. It is the primary measure of the performance of a channel-blocking
application of polymer solutions. For many cases, Rk and RRF are nearly equal, but RF
is usually much larger than Rk because it contains both the viscosity-enhancing and
the permeability-reducing effects.
The most common measure of permeability reduction is Rk, which is
sensitive to polymer type, molecular weight, degree of hydrolysis, shear rate, and
permeable media pore structure. Polymers that have undergone even a small amount
of mechanical degradation seem to lose most of their permeability reduction effect.
For this reason, qualitative tests based on screen factor devices are common to
estimate polymer quality.
The screen factor device is simply two glass bulbs mounted into a glass
pipette as shown in Fig. 8-8. Into the tube on the bottom of the device are inserted
several fairly coarse wire screens through which the polymer solution is to drain. To
use the device, a solution is sucked through the screens until the solution level is
above the upper timing mark. When the solution is allowed to flow freely, the time
required to pass from the upper to the lower timing mark td is recorded. The screen
328
Figure 8-8 Screen factor device

(adapted from Foshee et al., 1976)
factor for the polymer solution is then defined as
td
SF = (8.2-14)
tds
where tds is the similar time for the polymer-free brine.

Because of the normalization, screen factors are independent of temperature, device
dimensions, and screen coarseness, and they are fairly independent of screen spacing.
The screen factor is not independent of polymer concentration, but its primary intent
is to measure the time-dependent portion of the polymer’s solution configuration; that
is, it measures the rate at which a polymer molecule returns to its steady-state flow
configuration after it has been perturbed. This relaxation time is evidently very fast
for the polysaccharides because they do not have a measurable screen factor even at
high concentrations. HPAMs have much slower relaxation times because their screen
factors can be large even at the same viscosity as a polysaccharide solution. The
above explanations are consistent with the chemical properties of the two polymer
groups given in Sec. 8-1 and can be used to deduce the sensitivity of screen factors to
brine salinity and hardness.
Screen factors are particularly sensitive to changes in the polymer molecule
itself. One definition of polymer quality is the ratio of the degraded to the undegraded
screen factors. This use is important for screen factor devices, particularly in
locations that prohibit more sophisticated equipment.
Another use for screen factors is as a correlator for RF and RRF (Fig. 8-9). The
explanation for such a correlation is consistent with that given above on polymer
329
Figure 8-9 Correlation of resistance factors with screen factors (from Jennings et
al., 1971)
relaxation. On a pore scale, steady flow in permeable media is actually a succession

of contracting and diverging channels. The frequency with which the solution
experiences these contractions, compared with the polymer relaxation time,
determines the degree of permeability reduction. Such an effect also qualitatively
explains the increase in viscometer viscosity at very high shear rates (Hirasaki and
Pope, 1974).
The relaxation time argument cannot completely account for permeability
reduction because such effects have been observed in glass capillaries. For this case,
permeability reduction seems to be caused by polymer adsorption, which decreases
the effective pore size (see Exercise 8E).
A reasonable question is whether permeability reduction is a desirable effect.
Rk is difficult to control, being sensitive to even small deteriorations in the polymer
quality. Moreover, an extremely large Rk will cause injectivity impairment. But it is
possible to achieve a predesignated degree of mobility control with less polymer if Rk
> 1. If M T0 is a design or target endpoint mobility ratio,
0
⎛ k1′ ⎞⎛ μ 2 ⎞ M Rk =1 M 0
M T = ⎜ 0 ⎟⎜ 0 ⎟ =
0
= (8.2-15)
⎝ μ1 ⎠⎝ k2 ⎠ Rk RRF
In this equation, M 0 is the mobility ratio of a polymer having no permeability

Rk =1
reduction, and M0 is the endpoint water–oil mobility ratio. Clearly, if Rk > 1, the
polymer viscosity μ10 can be smaller than if Rk = 1, which indicates a given
concentration of HPAM will have a lower mobility ratio than polysaccharide under
conditions where both polymers have the same flowing viscosity. Note that the
limiting viscosity μ10 is used to estimate M0 from Eq. (8.2-15).
330
Chemical and Biological Degradation
The average polymer molecular weight can be decreased, to the detriment of the
overall process, by chemical, biological, or mechanical degradation. We use the term
chemical degradation to denote any of several possible processes such as thermal
oxidation, free radical substitution, hydrolysis, and biological degradation.
For a given polymer solution, there will be some temperature above which
the polymer will actually thermally crack. Although not well established for most
EOR polymers, this temperature is fairly high, on the order of 400 K. Since the
original temperature of oil reservoirs is almost always below this limit, of more
practical concern for polymer flooding is the temperature other degradation reactions
occur at.
The average residence time in a reservoir is typically very long, on the order
of a few years, so even slow reactions are potentially serious. Reaction rates also
depend strongly on other variables such as pH or hardness. At neutral pH,
degradation often will not be significant, whereas at very low or very high pH, and
especially at high temperatures, it may be. In the case of HPAM, the hydrolysis will
destroy the carefully selected extent of hydrolysis present in the initial product. The
sensitivity to hardness will increase, and viscosity will plummet. For Xanthan gum,
hydrolysis is even more serious since the polymer backbone is severed, resulting in a
large decrease in viscosity.
TABLE 8-1 SELECTED BACTERICIDES AND OXYGEN

SCAVENGERS (ADAPTED FROM ENHANCED OIL RECOVERY,
NATIONAL PETROLEUM COUNCIL, 1984)
Bactericide Oxygen scavengers

Commonly used
Acrolein Hydrazine
Formaldehyde Sodium bisulfite
Sodium dichlorophenol Sodium hydrosulfite
Sodium pintachlorophenol Sulfur dioxide
Proposed or infrequent use
Acetate salts of coco amines
Acetate salts of coco diamines
Acetate salts of tallow diamines
Alkyl amino
Alkyl dimethyl ammonium chloride
Alkyl phosphates
Calcium sulfate
Coco dimethyl ammonium chloride
Gluteraldehyde
Paraformaldehyde
Sodium hydroxide
Sodium salts of phenols
Substituted phenols
331
Oxidation or free radical chemical reactions are usually considered the most
.serious source of degradation. Therefore, oxygen scavengers and antioxidants are
often added to prevent or retard these reactions. These chemicals are strong reducing
agents and have the additional advantage of reducing iron cations from the +3 to the
+2 state. They, in turn, help prevent gelation, agglomeration, and other undesirable
effects that can cause wellbore plugging and reduced injectivity. Wellington (1980)
has found that alcohols such as isopropanol and sulfur compounds such as thiourea
make good antioxidants and free radical inhibitors.
Laboratory results indicate Xanthan can be stabilized up to about 367 K, and
HPAM to about 394 K. In the case of Xanthan, the results depend strongly on the
precise conditions such as salinity and pH, with high salinity and pH between 7 and 9
being preferred. Obviously, one should test the particular polymer solution under the
particular reservoir conditions of interest to establish the expected behavior.
Biological degradation can occur with both HPAM and polysaccharides, but
is more likely with the latter. Variables affecting biological degradation include the
type of bacteria in the brine, pressure, temperature, salinity, and the other chemicals
present. As in waterflooding, the preventive use of biocide is highly recommended.
Often too little biocide is used or it is started too late, and the ensuing problems
become almost impossible to correct. Table 8-1 lists typical polymer flooding
additions.
Mechanical Degradation
Mechanical degradation is potentially present under all applications. It occurs when

polymer solutions are exposed to high velocity flows, which can be present in surface
equipment (valves, orifices, pumps, or tubing), downhole conditions (perforations or
screens), or the sand face itself. Perforated completions, particularly, are a cause for
concern as large quantities of polymer solution are being forced through several
small holes. For this reason, most polymer injections are done through open-hole or
gravel-pack completions. Partial preshearing of the polymer solution can lessen the
tendency of polymers to mechanically degrade. Because flow velocity falls off
quickly with distance from an injector, little mechanical degradation occurs within
the reservoir itself.
All polymers mechanically degrade under high enough flow rates. But
14PAMs are most susceptible under normal operating conditions, particularly if the
salinity or hardness of the brine is high. Evidently, the ionic coupling of these anionic
molecules is relatively fragile. Moreover, elongational stress is as destructive to
polymer solutions as is shear stress though the two generally accompany each other.
Maerker (1976) and Seright (1983) have correlated permanent viscosity loss of a
polymer solution to an elongational stretch rate-length product. On a viscosity-
shear-rate plot (purely shear flow), mechanical degradation usually begins at shear
rates equal to or somewhat less than the minimum viscosity shear rate.
332
8-3 CALCULATING POLYMER FLOOD INJECTIVITY
The economic success of all EOR processes is strongly tied to project life or injection
rate, but polymer flooding is particularly susceptible. In many cases, the cost of the
polymer itself is secondary compared to the present value of the incremental oil.
Because of its importance, many field floods are preceded by single-well injectivity
tests. Here we give a simple technique for analyzing injectivity tests based on the
physical properties given in the previous section.
The injectivity of a well is defined as
i
I≡ (8.3-1)
ΔP
where i is the volumetric injection rate into the well, and ΔP is the pressure drop
between the bottom-hole flowing pressure and some reference pressure. Another
useful measure is the relative injectivity
I
Ir = (8.3-2)
I1
where I1 is the water injectivity. Ir is an indicator of the injectivity decline to be

anticipated when injecting polymer. Both I and Ir are functions of time, but the
longtime limit of Ir for a Newtonian polymer solution is simply the viscosity ratio if
skin effects are small. However, the ultimate Ir for an actual polymer solution can be
higher than this because of shear-thinning.
We make several simplifying assumptions, many of which can be relaxed
(Bondor et al., 1972). The well, of radius Rw, whose injectivity we seek, is in a
horizontal, homogeneous, circular drainage area of radius Re. The pressures at Re and
Rw are Pe and Pwf, respectively. Pe is constant (steady-state flow), but Pwf can vary
with time. The fluid flowing in the reservoir is a single aqueous phase, at residual oil
saturation, which is incompressible with pressure-independent rheological properties.
Dispersion and polymer adsorption are negligible although the polymer can exhibit
permeability reduction. The flow is one-dimensional and radial. Finally, the entire
shear rate range in the reservoir lies in the power-law regime; hence Eq. (8.2-7)
describes the apparent viscosity.
Subject to these assumptions, the continuity equation (Eq. 2.4-11) reduces to
d
(rur ) = 0 (8.3-3)
dr
where ur is the radial volumetric flux. This equation implies the volumetric rate is
independent of r and equal to i since
i = 2π rH t ur (8.3-4a)
Equation (8.3-4a) is a consequence of the incompressible flow assumption; however,

i is not independent of time. Let us substitute Darcy’s law for ur in Eq. (8.3-4a)
333
2π rH t k1′ dP 2π rH t k1′ dP
i=− =− (8.3-4b)
μ app dr n −1
H pl ur pl dr
from Eq. (8.2-7). This equation has been defined so that i is positive. The
permeability reduction factor is introduced through Eq. (8.2-12). Eliminating ur with
Eq. (8.3-4a) yields an ordinary differential equation, which may be integrated
between the arbitrary limits of P1 at r1 and P2 at r2.
n pl
⎛ i ⎞ H pl Rk
P2 − P1 = ⎜ ⎟
⎝ 2π H t ⎠ k1 (1 − n pl )
(r
1
1− n pl
− r2
1− n pl
) (8.3-5a)
The Newtonian flow limit, npl = 1 = Rk and Hpl = μ1, of this equation is the familiar
steady-state radial flow equation,
i μ1 ⎛r ⎞
P2 − P1 = ln ⎜ 1 ⎟ (8.3-5b)
2π k1 H t ⎝ r2 ⎠
We now apply these equations to the polymer flood injectivity.

At some time t during the injection, the polymer front (assumed sharp) is at
radial position Rp where
t
∫ idt = π ( R − Rw2 ) H tφ (1 − S 2 r )
2
p (8.3-6)
0
The left side of this equation is the cumulative volume of polymer solution injected.
Therefore, Eq. (8.3-5a) applies in the region Rw < r < Rp, and Eq. (8.3-5b) applies in
the annular region Rp < r < Re. With the appropriate identification of variables, we
have for the second region
i μ1 ⎛R ⎞
P R − Pe = ln ⎜ e ⎟ (8.3-7a)
p
2π k1 H t ⎜⎝ R p ⎟⎠
and for the first
n pl
⎛ i ⎞ H pl 1− n pl 1− n pl
Pwf − P R =⎜ ⎟ ( Rp − Rw ) (8.3-7b)
⎝ 2π H t ⎠ k1 (1 − n pl ) Rk
p
where P R , is the pressure at the polymer–water front. Adding these two equations
p
gives the total pressure drop from Rw to Re.
n pl
⎛ i ⎞ H pl Rk 1− n pl 1− n pl
Pwf − Pe = ⎜ ⎟ ( Rp − Rw )
⎝ 2π H t ⎠ k1 (1 − n pl )
i μ1 ⎛ ⎛R ⎞ ⎞
+ ⎜ ln ⎜ e ⎟ + S w ⎟ (8.3-8)
2π k1 H t ⎜ ⎜ Rp ⎟ ⎟
⎝ ⎝ ⎠ ⎠
where Sw the intrinsic skin factor of the well, has been introduced to account for well
damage.
334
Equation (8.3-8) substituted into the injectivity definition (Eq. 8.3-1) gives
n pl
−1 ⎛ i ⎞ H pl Rk 1− n pl 1− n pl
I =⎜ ⎟ ( Rp − Rw )
⎝ 2π H t ⎠ i (1 − n pl )k1
i μ1 ⎛ ⎛R ⎞ ⎞
+ ⎜ ln ⎜ e ⎟ + S w ⎟ (8.3-9)
2π k1 H t ⎜ ⎜ Rp ⎟ ⎟
⎝ ⎝ ⎠ ⎠
The water injectivity I1 is given by Eqs. (8.3-1) and (8.3-5b), with r1 = Rw and r2 = Re.
This and I, calculated from Eq. (8.3-9), yield an expression for Ir through Eq. (8.3-2).
Both I and Ir relate to the cumulative polymer solution injection (or to time) through
Eq. (8.3-6).
8-4 FRACTIONAL FLOW IN POLYMER FLOODS
The fractional flow treatment of polymer floods resembles the water–solvent

treatment in Sec. 7-7. The only major complications are the addition of terms for
polymer retention and inaccessible pore volume (IPV). In this section, we apply the
usual fractional flow assumptions: one-dimensional flow, incompressible fluid and
rock, and nondissipative mixing.
Single-Phase Flow
First, consider the case of a water-soluble species that is being adsorbed from
solution via a Langmuir-type isotherm. The isotherm is given by Eq. (8.2-10).
Let the flow be such that species concentration C4I is being displaced by
concentration C4J in single-phase flow where C4J > C4I. From Eq. (5.4-5a), the
specific velocity of concentration C4 is
−1 −1
⎛ (1 − φ ) ρ s dω 4 s ⎞ ⎛ dC4 s ⎞
vC4 = ⎜1 + ⎟ = ⎜1 + ⎟
⎝ φ dC4 ⎠ ⎝ dC4 ⎠
From Eq. (8.2-10), the specific velocity becomes
−1
⎛ C4 ⎞
vC4 = ⎜ 1 + 2 ⎟
(8.4-1)
⎝ (1 + b4 C4 ) ⎠
But since C4J > C4I, we have vC4 > vC4 , and the displacement is a shock; if C4J <
J I
C4I (see Exercise 8J), it would be 4 spreading wave. But for C4J displacing C4I, the
front between C4J and C4I moves with specific velocity
−1 −1
⎛ (1 − φ ) ρ s Δω 4 s ⎞ ⎛ ΔC4 s ⎞
vΔC4 = ⎜1 + ⎟ = ⎜1 + ⎟ (8.4-2)
⎝ φ ΔC 4 ⎠ ⎝ ΔC4 ⎠
335
from Eq. (5.4-5b). In this equation, Δ( ) = ( )J – ( )I. If, as is usually the case for
polymer floods, C4I = 0, Eq. (8.4-2) reduces to
1 1
vΔC4 = ≡ (8.4-3)
(1 − φ ) ρ s ⎛ ω 4 s ⎞ 1 + D4
1+ ⎜ ⎟
φ ⎝ C4 ⎠J
where D4 is the frontal advance loss for the polymer. It is also called the retardation
factor because adsorption causes the front velocity to be lower than that of the ideal
miscible displacement (see Sec. 5-4). D4 is one of the most useful concepts in both
polymer and micellar-polymer flooding because it expresses retention in pore volume
units which are consistent with slug size.
Two-Phase Flow
The fractional flow treatment will consist of two phases (aqueous j = 1 and oleic j =
2) and three components (brine i = 1, oil i = 2, and polymer i = 4). Let the permeable
medium have a uniform original water saturation of S1I. We inject an oil-free polymer
solution (S1J = 1 - S2r). The initial overall polymer concentration is 0, and the polymer
concentration in the aqueous phase is C4J. Polymer and water do not dissolve in the
oil (C12 = C42 = 0); the oil has no solubility in the aqueous phase (C21 = 0).
Effect of lPV The aqueous phase porosity is φS1. Only a portion of this pore
volume fraction, (φS1 – φIPV), is accessible to the polymer; hence the overall polymer
concentration per unit bulk volume is
W4 = (φ S1 − φIPV ) ρ1ω 41 + (1 − φ ) ρ sω 4 s (8.4-4a)
Similarly, the overall water concentration is
W1 = (φ S1 − φIPV ) ρ1 (1 − ω 41 ) + φIPV ρ1 (8.4-4b)
since only water is present in the excluded pore volume φIPV. But the IPV can be
easily neglected in Eq. (8.4-4b) because the polymer concentration is very small
(ω11 ≅ 1) . The overall oil concentration and Eqs. (8.4-4a) and (8.4-4b) sum to the
porosity as required by the assumption of incompressible flow.
Oil Displacement The polymer itself alters neither the water nor the oil
relative permeabilities because, as we have seen in Sec. 3-4, the apparent viscosity
cannot be increased enough to change residual phase saturations. Moreover, when
permeability reduction is significant, it applies over the entire saturation range but
only to the wetting phase (Schneider and Owens, 1982). We may, therefore, construct
a polymer-solution-oil (polymer–oil) water fractional flow curve simply by using the
apparent viscosity in place of the water viscosity and dividing kr1 by Rk. Figure 8-10
shows both the water–oil (f1 – S1) and polymer–oil ( f1 p − S1 ) fractional flow curves.
336
Figure 8-10 Graphical construction of polymer flooding fractional flow
Because the polymer adsorption is Langmuir-like, and because the polymer

displaces the connate water miscibly, the polymer front is pistonlike and has specific
velocity
f p (S * )
vΔC4 = * 1 1 (8.4-5a)
S1 + D4 − φe
where D4 is the polymer retardation factor defined in Eq. (8.4-3), and
φIPV
φe = (8.4-5b)
φ
S1* and f1 p ( S1* ) are the water saturations and fractional flows at the polymer shock
front. S1* may also be regarded as a point in the spreading portion of the mixed
polymer–oil wave given by the Buckley-Leverett equation, whence from Eq. (8.4-5)
we can define S1*
f p (S * ) ⎛ df p ⎞
vΔC4 = * 1 1 = ⎜ 1 ⎟ = vC1 (8.4-6)
S1 + D4 − φe ⎝ dS1 ⎠ S *
1
since S1* is also in the shock portion of the polymer–oil wave. The Buckley-Leverett
treatment in Sec. 5-2 used a similar argument.
337
Equation (8.4-6) will also determine the oil bank saturation since S2 will
change discontinuously with velocity given by
f p ( S * ) − f1 ( S1B )
vΔC2 = 1 1* = vC1 (8.4-7)
S1 − S1B
Equations (8.4-6) and (8.4-7) are particular statements of the coherence condition
(Eq. 5.6-14).
As in the solvent–water treatment in Sec. 7-7, the velocity of the front of the
oil (or water) bank is given by
f − f1I
vΔC2 = 1B = vΔC1 (8.4-8)
S1B − S1I
Figure 8-11 Figures for the fractional flow curves in Fig. 8-10
338
for a pistonlike oil bank front. The construction proceeds in the same manner as in
Sec. 7-7. Figure 8-11 shows the time–distance diagram and a composition profile at
tD = 0.35 for the construction in Fig. 8-10.
Though relatively direct, the construction in Figs. 8-10 and 8-11 has several
important insights into polymer floods.
1. The oil bank breakthrough time (reciprocal of the oil bank specific velocity
vΔC2 ) increases as S1I increases, suggesting polymer floods will be more
economic if they are begun at low initial water saturation. Of course, the
lower S1I, the higher the mobile oil saturation, also a favorable indicator for
polymer floods.
2. Adsorption (large D4) causes a delay of all fronts. D4 can be large if the poro-
sity is low, the retention is high, or the injected polymer concentration C4J is
low. Usually, C4J is so low that D4 can be high even if retention is moderate.
3. Inaccessible pore volume causes an acceleration of all fronts, exactly
opposite to retention. In fact, retention and IPV can exactly cancel so that the
polymer front and the denuded water front v1′ (Fig. 8-10) travel at the same
velocity.
4. Both D4 and IPV influence the oil bank saturation, which in turn, influences
the oil bank mobility and the desired injected polymer concentration.
8-5 ELEMENTS OF POLYMER FLOOD DESIGN
Polymer flood design is a complex subject. But most of the complexity arises from
reservoir- specific aspects of a particular design. In this section, we deal in
generalities that apply to all types of polymer flooding.
A polymer flood design procedure will follow these six steps.
1. Screen the candidate reservoirs. The distinction between technical and

economic feasibility is important. Technical feasibility means a given
reservoir can be polymer flooded regardless of the funds available. Economic
feasibility means the project has a good chance of being profitable. Technical
feasibility is measured by a series of binary screening parameters (see
National Petroleum Council, 1984). But for polymer flooding, there are only
two: the reservoir temperature should be less than about 350 K to avoid
degradation, and the reservoir permeability should be greater than about 0.02
μm2 to avoid plugging. Economic feasibility can be estimated by simple hand
calculations (as in the fractional flow method) or through using predictive
models (Jones et al., 1984), which requires deciding how the polymer is to be
used.
2. Decide on the correct mode. The choices are (a) mobility control (decrease
M), (b) profile control (improve the permeability profile at the injectors or
producers), or (c) some combination of both. We have not discussed profile
control, but the concepts and goals are similar to polymer flooding. We want
to inject an agent that will alter the permeability so that more fluid will go
339
into the tight rock than into the high-permeability rock. We can do this by
using gels, polymers, and solids and by using selective perforation. When
selective perforation is ineffective or incompletely effective, we use chemical
agents or solids.
3. Select the polymer type. The requirements for EOR polymers are severe. An
outline of the principal ones is as follows:
(a) Good thickening. This means high mobility reduction per unit cost.
(b) High water solubility. The polymers must have good water solubility
under a wide range of conditions of temperature, electrolyte composition,
and in the presence of stabilizers.
(c) Low retention. All polymers adsorb on reservoir rocks to various
degrees. Retention may also be caused by plugging, trapping, phase
separation, and other mechanisms. Low here means less than 20 μg/g.
(d) Shear stability. During flow through permeable media, stress is applied
to the polymer molecules. As we discussed, if this is excessive, they may
mechanically break apart or permanently degrade, resulting in less
viscosity. HPAM is especially subject to shear degradation.
(e) Chemical stability. Polymers, like other molecules, can chemically react,
especially at high temperature and in the presence of oxygen.
Antioxidants are used to prevent this.
(f) Biological stability. Both HPAM and polysaccharides can be degraded
by bacteria, but the latter are more susceptible. Biocides are required to
prevent this.
(g) Good transport in permeable media. This catchall includes essentially the
ability to propagate the polymer through the rock intact and without
excessive pressure drop or plugging. Good transport also means good
injectivity and no problems with microgels, precipitates, and other
debris.
Obviously, no one polymer can universally meet these requirements
for all reservoir rocks. Thus we must tailor the polymer to the rock to
some extent. Some general guidelines are possible for minimum
standards, but the ultimate criterion must be economics.
4. Estimate the amount of polymer required. The amount, the total mass in
kilograms to be injected, is the product of the slug size, the pore volume, and
the average polymer concentration. Ideally, the amount would be the result of
an optimization study that weights the present value of the incremental oil
against the present value of the injected polymer. Each iteration of the
optimization procedure requires estimating the polymer concentration in
initial portion (spike) of the slug and estimating the volume of the polymer
slug (spike plus rate of taper).
(a) Estimating the spike concentration. Suppose we have decided on a target
mobility ratio that might come from simulation studies (see Chap. 6) or
simply injectivity limitations. If the target mobility ratio is MT
(λrt ) polymer (λr′1 + λr 2 ) S *
MT = = 1
(8.5-1)
(λrt )oil bank (λr1 + λr 2 ) S1B
340
Estimating the spike concentration simply means picking the value of
injected polymer concentration that gives the correct MT in this
equation. The translation between apparent viscosity follows from
permeability reduction factor correlations and shear rate data as in Fig.
8-4. The latter must be evaluated at a sheer rate corresponding to the
median velocity in the flood––usually the low shear rate plateau.
Estimating the denominator of Eq. (8.5-1), the oil bank relative mobility,
is a little more difficult.
One procedure is to estimate the oil bank saturation through the
graphical procedure of Sec. 8-4, and then estimate the oil bank mobility
from the relative permeability curves evaluated at this saturation
⎛k k ⎞
(λrt )0 B = ⎜ r1 + r 2 ⎟ (8.5-2)
⎝ μ1 μ 2 ⎠ S1 B
This procedure is iterative inasmuch as S1B depends on the polymer–oil
fractional flow curve. This, in turn, depends on the polymer apparent
viscosity whose value we are estimating in Eq. (8.5-1). Fortunately, the
dependence between S1B and apparent viscosity is weak, and a
trial-and-error procedure should converge rapidly.
A second procedure is to base the total mobility of the oil bank on
the minimum in the total relative mobility curve (Gogarty et al., 1970).
The minima in such curves do not, in general, correspond to the oil bank
saturation from fractional flow theory. However, taking MT based on the
minimum will yield a conservative design since the mobility ratio with
the actual oil bank saturation will always be less than or equal to MT. The
method has the advantage of simplicity since it is noniterative.
Both methods require care in measuring relative permeability curves
since hysteresis can render the drainage and imbibition kr’s different
(Chang et al., 1978). Such hysteresis effects are particularly difficult to
reproduce when the initial water saturation begins at an intermediate
value. The second method is also commonly used in micellar- polymer
design (see Fig. 9-34).
(b) Estimate the polymer slug volume. One way to do this is to simply let the
slug volume be somewhat larger than the retention. Although this is the
basic premise in designing a micellar slug, retention is not the dominant
factor in polymer slug sizing. The major control affecting slug size is
viscous fingering between the chase water and the polymer spike.
In predicting the extent of fingering, all the problems in estimating
the rate of finger propagation that we discussed in Sec. 7-8 apply. Once
again, we apply the Koval model, but here the effective mobility ratio
must be modified to account for the polynomial mixing expressed in Eq.
(8.2-1).
K
E = (1 + a1C4 + a2 C42 + a3C43 + ⋅ ⋅ ⋅ + ) RRF = val (8.5-3)
HK
341
where C4 = 0.22 C4J. The use of this equation, particularly the constant
mixing factor 0.22, is relatively untested in polymer flooding. We have
assumed complete analogy between the first-contact miscible flooding
case and the unstable chase water displacement in this regard.
We use the time–distance diagram to sketch slug sizing alternatives.
An obvious sizing technique is to begin chase water injection just as the
polymer breaks through (Fig. 8-12a). This is excessively conservative
since much full-strength polymer is produced. A second possibility is to
adjust the polymer slug size so that the polymer and chase water break
through simultaneously (Fig. 8-12b), leading to the following equation
for slug size
Figure 8-12 Time–distance diagrams

for polymer grading
1 − S2 r
t Ds = 1 − S 2 r − D4 − (8.5-4)
K val
342
where we have taken S1* = 1 – S2r, and φIPV = 0. An equally viable

alternative is to grade the polymer back to chase water in steps. Figure
8-12(c) shows two such steps with the size of the spike and the
intermediate step adjusted so that the chase water again breaks through
with the polymer. The isoconcentration lines become curved after there
is wave interference.
The case in Fig. 8-12(c) uses less polymer than that in Fig. 8-12(b).
In fact, Claridge (1978) has shown that a continuously graded polymer
drive uses the least amount of polymer. Such grading is impractical
except as a limiting case to compare it to the no-grading case (Fig.
8-12b). But a succession of finite grading steps is extremely difficult to
deal with theoretically because of the numerous degrees of freedom
present. That is, the engineer must decide on the N number of steps, the
N slug volumes, and the N – 1 intermediate concentrations. In practice,
single-step (no-grading) polymer floods, and logarithmic grading
(Mungan, 1968) are the most common procedures.
5. Design polymer injection facilities. Getting a good quality solution is, of
course, important, but the cost of the injection facilities is usually small
compared to well and chemical costs.
The three essential ingredients are mixing facilities, filtration, and
injection equipment. The type of mixing apparatus depends on the polymer.
For solid polymers, a skid-mounted solid mixer is required. Concentrates or
emulsion polymers require somewhat less sophistication although the latter
may require some emulsion breaking. Filtration largely depends on the
success of the mixing, but ordinarily it is no more stringent than what is
required by waterflooding. But if exotic and difficult filtration is required, the
complexity and cost can become significant. Injection equipment is the same
as that for waterflooding. All surface and downhole equipment should be
modified to avoid all forms of degradation.
6. Consider the reservoir. Little is required here beyond the usual waterflood
considerations such as the optimal well pattern and spacing, completion
strategy, pattern allocation (balance), reservoir characterization, and
allowable injection rates.
Optimal values of these quantities imply precise values that will result in
the maximum rate of return on investment. Since several quantities are
involved, it is usually not possible to perform optimizations on everything.
Hence most of the parameters must be fixed by other considerations (such as
striving for a target mobility ratio). But for the most sensitive quantities,
optimization is required.
Figure 8-13 shows a schematic optimization for the amount of polymer
injected. The vertical axes plot both an economic measure, such as the
cumulative incremental discounted cash flow (DCF), and incremental oil
343
Figure 8-13 Schematic incremental oil recovered and economic trends for a
mobility control flood.
recovery (IOR) versus the amount of polymer injected. The IOR curve is
monotonically increasing from zero. The DCF curve begins at zero,
decreases for small polymer amounts, and then rises to a maximum at
substantially larger amounts. After this point, the DCF decreases
monotonically. The DCF decreases initially because the entire expense of the
polymer is assessed in the initial stages of a project when little incremental
oil has yet been produced. This front-end loading effect is present in all EOR
processes, particularly chemical floods. Such a curve is highly instructive
because it counters a tendency to short-cut the amount of polymer injected if
the initial economics are unfavorable. Unfortunately, many actual polymer
flood applications have used less than the optimum amount of polymer.
8-6 FIELD RESULTS
The incremental oil recovery (IOR) from a polymer flood is the difference between
the cumulative oil actually produced and that which would have been produced by a
continuing waterflood (see Exercise 8L). Thus for a technical analysis of the project,
it is important to establish a polymer flood oil rate decline and an accurate waterflood
decline rate. Figure 8-14 shows the IOR for the North Burbank polymer flood.
Table 8-2 summarizes other field results on more than 250 polymer floods
based on the comprehensive survey of Manning et al. (1983). The table emphasizes
oil recovery data and screening parameters used for polymer flooding.
Approximately one third of the reported projects are commercial or field-scale
floods. The oil recovery statistics in Table 8-2 show average polymer flood
recoveries of 3.56% remaining (after waterflood) oil in place and about 1 m3 of IOR
for each kilogram of polymer injected with wide variations in both numbers. The
large variability reflects the emerging nature of polymer flooding in the previous
decades. Considering the average polymer requirement and the average costs of
crude and polymer, it appears that polymer flooding should be a highly attractive
344
Figure 8-14 Tertiary polymer flood response from North Burbank Unit, Osage County,
Okla. (from Clampitt and Reid, 1975)
EOR process. However, such costs should always be compared on a discounted basis,
reflecting the time value of money. Such a comparison will decrease the apparent
attractiveness of polymer flooding because of the decreased injectivity of the polymer
solutions.
In terms of the number of field projects, polymer flooding is the most common
enhanced oil recovery technique in existence. The reasons for this are that, short of
waterflooding, polymer flooding is the simplest technique to apply in the field and
requires a relatively small capital investment. Most of the field projects have been
small, however, as has the amount of oil recovered, a fact that should be expected
from the treatment given in this chapter. Nevertheless, there can exist significant
potential for an acceptable rate of return even when recovery is low.
The most important property covered in this chapter is the non-Newtonian
behavior of polymer solutions, because such behavior impacts on the polymer
345
TABLE 8-2 POLYMER FLOOD STATISTICS (ADAPTED FROM MANNING, 1082)
Value
standard Number Standard
(units) projects* Mean Minimum Maximum deviation
Oil recovery
(% remaining OIP) 50 3.56 0 25.3 5.63
Polymer utilization
(m3/kg polymer) 80 0.94 0 12.81 1.71
Oil recovery
(m3/hm3 bulk volume) 88 3.1 0 24.3 4.72
Permeability variation
(fraction) 118 0.70 0.06 0.96 0.19
Mobile oil saturation
(fraction) 62 0.27 0.03 0.51 0.12
Oil viscosity
(mPa-s) 153 36 0.072 1,494 110.2
Resident brine salinity
(kg/m3 TDS) 10 40.4 5.0 133.0 33.4
Water-to-oil mobility
Ratio (dimensionless) 87 5.86 0.1 51.8 11.05
Average polymer
Concentration (g/m3) 93 339 51 3,700 343
Temperature (K) 172 319 281 386 302
Average permeability
(μm2) 187 0.349 0.0015 7.400 0.720
Average porosity
(fraction) 193 0.20 0.07 0.38 0.20
*Partial data available on most projects; includes both commercial and pilot projects
requirements through the design mobility ratio, and on the ability to accurately
forecast the rate of polymer injection. Polymer injection rate determines project life
which, in turn, determines the economic rate of return. Injectivity estimates along
with estimates of mobile oil saturation and the likelihood that polymer will remain
stable in a given application are the most important determinants in polymer flooding
success.
EXERCISES
8A. Calculating Shear Rates. Calculate the equivalent shear rate under the following
conditions:
(a) In an open-hole completion (entire well cylinder open to flow) where q = 16
m3/day, Rw = 7.6 cm, and net pay Ht = 15.25 m.
(b) In the field where the interstitial velocity is 1.77 μm/s.
(c) Using the data for Xanflood at 297 K and 1% NaCl (Fig. 8-5), estimate the
effective permeable medium viscosity at the above conditions for a 600 g/m3
polymer solution.
346
(d) Suppose the well in part (a) is perforated with 1 cm (ID) holes over its entire net
pay at a density of 4 holes/m. Assuming a uniform fluid distribution, estimate the
shear rate in the perforations.
(e) Comparing the results of parts (a) and (d), what do you conclude about the
preferred completion technique in polymer flooding? Use k1 = 0.1 μm2, φ = 0.2,
and S1 = 1.0 in all parts.
8B. Derivation of Power Law in Permeable Media. Equation (8.2-9) may be derived in the
same manner as Eq. (3.1-11). The procedure is as follows:
(a) Show that a force balance on an annular element of a single-phase fluid flowing
through a tube (as in Fig. 3. 1) in Ian- iinar steady-state flow is
1 d (rτ rz ) ΔP
= (8B-1)
r dr L
where τrz is the shear stress on the cylindrical face at r, and ΔP/L is the pressure
gradient. This equation, when integrated, yields
ΔP
τ rz = r (8B-2)
2L
The shear stress must be finite at r = 0.

(b) The power-law expression relating shear stress to shear rate is
n −1
τ rz = K pl γ pl (8B-3)
where
dv
γ =− (8B-4)
dr
is the shear rate. Show that combining Eqs. (8B-2) through (8B-4) leads to a
differential equation whose solution is
1/ n pl
⎛ ΔP
v(r ) = ⎜
⎜ 2 LK
⎝ pl
⎞
⎟⎟
⎠
⎛ n pl
⎜⎜
⎝ 1 + n pl
(
⎞ 1+ n pl
⎟⎟
⎠ R
n pl
1+ n pl
n
− r pl ) (8B-5)
This equation has used the no-slip condition v(R) = 0.

(c) Using Eq. (8B-5), show that the shear rate at the wall of the tube depends on the
average velocity as
1 + 3n pl ⎛ v ⎞
γ wall = ⎜ ⎟ (8B-6)
n pl ⎝ R ⎠
(d) When the equivalent radius from Eq. (3.1-4) is substituted, this gives
⎛ 1 + 3n pl ⎞ u
γ eq = ⎜ ⎟⎟ (8B-7)
⎜ n φ 1/ 2
⎝ pl ⎠ 1 1)
(8k
With appropriate variable identifications, this equation yields Eqs. (8.2-7) and
(8.2-8) when substituted into
τ rz
μ app = (8B-8)
γ eq
347
8C. Langmuir Calculations. The Langmuir isotherm and various other insights may be
derived fairly simply. Suppose a permeable medium in contact with a solution
containing an adsorbing species consists of a fixed number of surface sites. A fraction
φ of these sites is covered when the solution concentration of an adsorbing species is C.
(a) Let the rate of adsorption be kf C (1 – φ) and the rate of desorption be krφ. kf and kr
are the forward and reverse rate constants. At equilibrium, the forward and reverse
reaction rates are equal. Show that the fractional surface coverage is
kf
C
kr
θ= (8C-1)
kf
1+ C
kr
(b) Show that φ may be related to ωs, adsorption in mass per unit of rock mass by
d p2 N A ρ r
θ= ωs (8C-2)
av M w
where ρr is the adsorbed species density, av is the specific surface area of the
medium, Mw is the molecular weight of the adsorbed species, and NA is
Avogadro’s number. Assume the adsorbed species exists on the surface as a
monolayer of cubes of diameter dp.
(c) If the observed polymer adsorption is 18 μg/g-rock, calculate φ. Take the medium
to have the Berea properties tabulated in Table 3-5. You must derive the effective
polymer sphere diameter from the intrinsic viscosity (Eq. 8.2-4) and the data in
Fig. 8-4. The polymer molecular weight is 2 million.
(d) What can you conclude about the nature of the adsorption of polymers from this?
8D. Complications to Langmuir Isotherm
(a) Suppose there is Langmuir adsorption of a single adsorbing species with a finite
mass transfer rate rmt between the bulk solution and the solid–fluid interface given
by
rmt = h(C − C ) (8D-1)
In this expression, C and C are the bulk and interface concentrations, and h is the
specific mass transfer coefficient. Show that an isotherm relating φ to C has the
same form as Eq. (8C-1) but with k ′f replacing kf where
1 1 1
= + (8D-2)
k ′f k f h
You must assume the rate of adsorption is equal to rmt,

(b) Show that if h → ∞, the isotherm approaches the expression derived in Exercise
8C.
(c) Consider now a case where h → ∞ and there are i = 1, . . . , NC adsorbing species,
each competing for a fixed number of sites. Derive the Langmuir isotherm relating
the adsorbed concentration of species i, φi, to its bulk concentration Ci.
(d) Use this expression to justify the fractional coverages calculated in part (d) of
Exercise 8C.
348
8E. Simplified Permeability Reduction. One of the explanations for permeability reduction
is that the effective pore size is decreased (or the effective grain diameter increased)
because of the adsorption of a layer of polymer on the rock surface. In the following,
take the medium to be comprised as spheres of diameter Dp:
(a) Derive an expression for the permeability reduction factor Rk based on the polymer
adsorbing as a uniform layer of thickness δ on the rock surface. You must use the
hydraulic radius concept developed in Sec. 3-1.
(b) Make two plots, at φ = 0.1 and 0.2, of polymer adsorption (in mg polymer/g-rock)
versus Rk. Take the density of the adsorbed polymer to be 1.5 g/cm3 and the
density of the rock to be 2.5 g/cm3.
8F. Representation of Linear Viscoelasticity. A powerful conceptual model of a liquid that
has some elastic effects is the Maxwell model, which is the series combination of a
spring and a dashpot
where F is the force sustained by the model, and εl and ε2 are the strains (dimensionless
deformations). Let the spring be a linear elastic element so that
F = k ε1 (8F-1)
likewise, the dashpot is a Newtonian viscous element
F = με 2 (8F-2)
where k and μ are the spring constant and viscosity of the element. Because of the
series arrangement, the force supported by both elements is the same; however, the
total strain ε is
ε = ε1 + ε 2 (8F-3)
(a) Show that the relationship between the time behavior of the force and the strain is
με = θ F + F (8F-4)
In this equation, θ = μ/k is the relaxation time of the model, and ε is the time
derivative of ε.
(b) To integrate this, we treat ε as a known function of time. Show that the general
solution is
t dε
F (t ) = e − t / θ F (0) + ke− t / θ ∫ eξ / 0 dξ (8F-5)
0 dξ
The next three steps complete the analogy between the Maxwell model and
viscoelastic flow.
(c) If the rate of strain is constant and the initial force of the model is zero, show that
F (t ) = με (1 − e − t / θ ) (8F-6)
(d) The apparent viscosity of the model is defined as F / ε . Show from Eq. (8F-4) that
this becomes
349
μ
μ app = (8F-7)
F
1+θ
F
(e) Use this equation and Eq. (8F-6) to show that
μ
μ app = (8F-8)
1 + N Deb
The quantity in the denominator of Eq. (8F-8) is the Deborah number

θ
N Deb = (8F-9)
t
This number, the ratio of relaxation time to undisturbed flow time around a rock
grain, is a measure of viscoelastic effects in permeable media flow when the
characteristic flow time t has been replaced by φDp/μ.
8G. Analysis of Screen Factor Device. The screen factor device in Fig. 8-8 may be analyzed
as a permeable medium experiencing gravity drainage. The volume V of fluid in the
bulb at any height h(h1 > h > h2) is
π
V= (h − h2 )(3R − h + h2 ) (8G-1)
3
from the bulb geometry. If we treat the screen pack as a permeable resistive element,
the flux through the screens is
⎛ k ρ gh ⎞
u = −⎜
⎜ L μ ⎟⎟
(8G-2)
⎝ app ⎠
(a) Since u = –1/πr2(dV/dt), show from these equations that the height h is the solution
to
dh r 2 ρ gkh
[(h − h2 )(2 R − (h − h2 ))] = (8G-3)
dt μ app L
L in these equations is the height of the screen pack.

(b) Neglecting the drainage times in the tubes above and below the lower bulb, derive
an expression for the drainage time for a Newtonian fluid. The drainage time td is
defined as
td = t |h = h2 −t |h = h1 (8G-4)
(c) Repeat part (b) with a viscoelastic fluid whose apparent viscosity is
HVE
μ app = (8G-5)
1 + bu
In view of Eqs. (8G-4) and (8.2-14), show that the screen factor SF is given by
HVE kρg
SF = + bI (8G-6)
μ1 μ1 L
350
where I is a geometric factor. From Exercise 8F, the screen factor is directly
proportional to the fluid’s relaxation time.
8H. Injectivity Calculation
Use the following data for the Coalinga HX sand (Tinker et al., 1976):
φ = 0.28 k1 = 0.036 μm2

n −1
Kpl = 7.5 mPa-s(s) pl μ1 = 0.64 mPa-s
npl = 0.8 Ht = 2.44 m
Rk =3 Rw = 10 cm
Re = 284 m i = 30 m3/D
S2r = 0.2
(a) Calculate the relative injectivity Ir versus cumulative polymer injected. Plot Ir
versus tD (up to tD = 0.5) on linear graph paper.
(b) Show that when Rp = Re, the Newtonian polymer case (npl = 1) reduces to
μ1
Ir = (8H-1)
K pl Rk
(c) Plot the Newtonian polymer case for the HX sand on the same plot as in part (a).
8I. Improvements to Injectivity Calculations. If the shear rate range in a cylindrical
reservoir is outside the power-law range, the following truncated form of Eq. (8.2-7)
must be used:
⎧ μ10 , u < u0
⎪⎪ n −1
μ app = ⎨ H pl u pl , u0 < u < u ∞ (8I-1)
⎪
⎪⎩ μ1∞ , u > u∞
where u0 and u∞ are superficial velocities which define the limits of the power law
range.
(a) Repeat the derivation in Sec. 8-3 for I and Ir using Eq. (8I-1), assuming both the
maximum and minimum velocities fall outside the power-law range.
(b) For numerical simulation, it may be more convenient to define injectivity in terms
of the average reservoir pressure P rather than Pe (Bondor et al., 1972). Rederive
the expression for I defined in this manner.
(c) For large numerical simulations, the entire non-Newtonian range of polymer
behavior is confined within one grid block of the well. This being the case, the
non-Newtonian effect can be effectively expressed as a time-varying skin factor in
terms of an average polymer “saturation.” Derive an expression for this skin
factor.
8J. Transport of Adsorbing Slugs. The leading edge of a polymer slug adsorbing as a
Langmuir isotherm is self-sharpening.
(a) Show that the rear of the slug (CK < CJ) is a spreading wave.
(b) If the Langmuir parameters in Eq. (8.2-10) are a = 2 and b = 20, plot the time–
distance diagram and effluent history of tDs = 0.4 slug displacement. Take CI = CK
= 0 and CJ = 1.
(c) The propagation of slugs satisfies an overall material balance
351
∞
tDs =
∫
0
(C + Cs )dxD
(8J-1)
CJ
Use the analogy to the Welge integration in Sec. 5-2 to show that Eqs. (8.4-1) and
(8.4-2) satisfy this identically. In all these calculations, take the flow to be single
phase with the usual fractional flow assumptions.
8K. Asymptotic Mixing Zone Length (Lake and Helfferich, 1978). Stabilized mixing zones
occur in miscible displacements if the transported species adsorbs according to a
Langmuir isotherm. The spreading caused by dispersion is balanced by the sharpening
caused by adsorption. In the following, take the dimensionless material balance of an
adsorbing species to be
∂ (C + Cs ) ∂C 1 ∂ 2C
+ − =0 (8K-1)
∂tD ∂xD N Pe ∂xD2
where C and Cs are the solution and adsorbed concentrations for an adsorbing species.
C is normalized so that the injected concentration is unity, CJ = 1, and CI = 0.
(a) Show that Eq. (8K-1) may be transformed to a moving coordinate system ( xD′ , tD ),
where xD′ = xD − vΔC t D , and vΔC is the shock velocity of C. This gives
∂ (C + Cs ) ∂Cs ∂C 1 ∂ 2C
− vΔC + (1 − vΔC ) − =0 (8K-2)
∂tD ∂xD′ ∂xD′ N Pe ∂ ( xD′ ) 2
(b) The displacement will asymptotically approach stabilized flow where the time
derivatives in Eq. (8K-2) are zero. Show that in this limit the resulting ordinary
differential equation may be integrated to give
1 0.1 dC
ΔxD =
N Pe ∫0.9 (1 − vΔC )C − vΔC Cs
(8K-3)
Equation (8K-3) uses the boundary conditions C(+∞) = dC(+∞)/dxD = 0 and the
definition of dimensionless mixing zone given in Eq. (5.2-15a).
(c) When CJ = 1, it is convenient to write the Langmuir isotherm (Eq. 8.2-10) so that
the plateau adsorption appears in the equation in place of the parameter a
(1 + b)CsJ C
Cs = (8K-4)
1 + bC
where Csj is the maximum adsorbed concentration. Substitute Eq. (8K-4) into Eq.
(8K-3), and perform the indicated integration to show that
1 ⎧1 + CsJ ⎫ ⎛ 2 ⎞
ΔxD = ⎨ ⎬ ⎜ 1 + ⎟ ln(9) (8K-5)
N Pe ⎩ CsJ ⎭ ⎝ b ⎠
where this equation has used a form of vΔC consistent with Eq. (8K-4).
(d) Take Eq. (8K-5) in the limits of b → ∞, b → 0, and NPe → ∞, and justify each
answer on physical grounds.
8L. Fractional Flow and Incremental Oil
(a) Calculate the polymer frontal advance lag D4 when the maximum polymer
adsorption is 38 g/m3 (bulk volume), the injected polymer concentration is 1200
g/m3, and the porosity is 0.2.
352
(b) Using the D4 of part (a) and the water–oil relative permeabilities in Fig. 8L,
calculate the effluent history of polymer and oil for a polymer flood with μ10 = 30
mPa-s. Take the oil and water viscosities to be 20 and 1 mPa-s, respectively, the
dip angle to be 0, the permeability reduction factor to be 1, and the initial water
saturation to be 0.4.
(c) The technically correct way to evaluate a polymer flood is by the incremental oil
recovery (IOR)
⎛ Polymer flood ⎞ ⎛ Waterflood ⎞

IOR = ⎜ ⎟−⎜ ⎟ (8L-1)
⎝ oil produced ⎠ ⎝ oil produced ⎠
Figure 8L Relative permeabilities for

Exercise 8L (from El Dorado, 1977).
353
Calculate and plot IOR (in SCM) versus time (years). Take the pore volume to be
1.6 × 106 m3, the injection rate constant at 480 SCM/day, and all formation volume
factors to be 1.0 m3/SCM.
8M. Fractional Flow and Slugs. Fractional flow theory can be used to gain insight into the
behavior of polymer slugs, under idealized conditions, and into the polymer utilization
factor.
(a) Assume the polymer is to be injected as a slug. If the chase water displaces the
polymer as an ideal miscible displacement at residual oil saturation, show that the
polymer chase water front travels with specific velocity
1
vCW = (8M-1)
1 − S2 r
if the polymer adsorption is irreversible and excluded pore volume negligible.

(b) Show that the polymer slug size just needed to satisfy adsorption is equal to D4.
(c) The data to use in the remainder of this exercise are
a = 1 cm3/g-rock C4J = 800 g/m3

b = 100 cm3/mg ρs = 2.65 g/cm3
φ = 0.2
Plot the time–distance and effluent histories (oil and polymer) if the slug size used
is one half that demanded by adsorption. Use the fractional flow curves and initial
condition of Exercise 8L.
8N. Polymer Flood Design. You want to design a polymer flood in a reservoir containing an
oil and brine whose viscosities are 25 mPa-s and 0.38 mPa-s, respectively, at reservoir
temperature of 73°C. The relative permeability curves of Fig. 8L apply, and conditions
indicate the Xanflood data in Figs. 8-4 and 8-5 are satisfactory for this reservoir.
(a) Plot the total relative mobility curves. If the desired mobility ratio is 0.7, estimate
the polymer concentration required to bring this about. Use the data in Fig. 8-5,
and recall that μ1′ / μ1 is essentially independent of temperature.
(b) Estimate the power-law parameters Kpl, npl, and Hpl for the polymer solution in part
(a).
(c) The flood is to be done at a constant volumetric injection rate of 20 m3/D. Estimate
and plot as a function of volume injected the bottom-hole injection pressure in
MPa. Justify the shape of this curve on physical grounds.
(d) For an open-hole completion, estimate the shear rate the polymer solution will be
exposed to. Does this portend mechanical degradation of the polymer?
Take the reservoir to be circular with Re = 950 m and Pe = 18 MPa.
Additional properties are k = 0.05 μm2, Sw = 0, Rw = 5 cm, Ht = 42 m, φ = 0.2, and
S2r = 0.3.
354
Micellar-Polymer
Flooding
From the earliest days, it was recognized that capillary forces caused large quantities
of oil to be left behind in well-swept zones of waterflooded oil reservoirs. Capillary
forces are the consequence of the interfacial tension (IFT) between the oil and water
phases that resists externally applied viscous forces and causes the injected and
banked-up connate waters to locally bypass oil. Similarly, early efforts of enhanced
oil recovery strove to displace this oil by decreasing the oil-water IFT. Though many
techniques have been proposed and field tested, the predominant EOR technique for
achieving low IFT is micellar-polymer (MP) flooding.
Lowering interfacial tension recovers additional oil by reducing the capillary
forces that leave oil behind any immiscible displacement. This trapping is best
expressed as a competition between viscous forces, which mobilize the oil, and
capillary forces, which trap the oil. The local capillary number Nvc, the dimensionless
ratio of viscous to capillary forces, determines the residual oil and water saturations
through a capillary desaturation curve (CDC). Section 3-4 gives general features
about the CDC and Nvc. In this chapter, we specialize those results to MP flooding.
Recall that ultralow IFTs are required––of the order of 1 μN/m––and that these
values can be attained only through highly surface-active chemicals.
9-1 THE MP PROCESS
MP flooding is any process that injects a surface-active agent (a surfactant) to bring

about improved oil recovery. This definition eliminates alkaline flooding (see Chap.
10) where the surfactant is generated in situ and other EOR processes where lowering
the capillary forces is not the primary means of oil recovery.
355
MP flooding has appeared in the technical literature under many names:
detergent, surfactant, low-tension, soluble oil, microemulsion, and chemical flooding.
We use the term micellar-polymer flooding because it is the least ambiguous
(chemical flooding, for example, could describe all nonthermal EOR processes) and
most comprehensive (no other name implies the polymer component). Moreover,
several names imply a specific sequence and type of injected fluids as well as the
specific nature of the oil-recovering MP slug itself. Though there are differences
among processes, in this chapter we emphasize the similarities since they are more
numerous and important.
Figure 9-1 shows an idealized version of an MP flooding sequence. The
process is usually applied to tertiary floods and is always implemented in the drive
mode (not cyclic or huff ’n puff). The complete process consists of the following:
Preflush. A volume of brine whose purpose is to change (usually lower) the

salinity of the resident brine so that mixing with the surfactant will not cause
loss of interfacial activity. Preflushes have ranged in size from 0% to 100%
of the floodable pore volume (Vpf) of a reservoir. In some processes, a
sacrificial agent is added to lessen the subsequent surfactant retention (Holm,
1982).
MP slug. This volume, ranging from 5% to 20% Vpf in field applications,
contains the main oil-recovering agent, the primary surfactant. Several other
chemicals (Fig. 9-1) are usually needed to attain the design objectives. We
discuss the purpose of these chemicals in more detail later.
Mobility buffer. This fluid is a dilute solution of a water-soluble polymer
whose purpose is to drive the MP slug and banked-up fluids to the production
wells. All the polymer flooding technology discussed in Chap. 8 carries over
to designing and implementing the mobility buffer. Thus in this chapter, we
deal relatively little with the mobility buffer though there is good evidence
(see Fig. 9-33) that this volume is very important to the oil recovering ability
Figure 9-1 Idealized cross section of a typical micellar-polymer flood (from

Lake, 1984)
356
of the entire sequence. The target oil for an MP flood––the residual oil––is
different from that of a polymer flood––the movable oil.
Mobility buffer taper. This is a volume of brine that contains polymer,
grading from that of the mobility buffer at the front end (the spike) to zero at
the back. The gradual decrease in concentration mitigates the effect of the
adverse mobility ratio between the mobility buffer and the chase water.
Chase water. The purpose of the chase water is simply to reduce the expense
of continually injecting polymer. If the taper and mobility buffer have been
designed properly, the MP slug will be produced before it is penetrated by
this fluid.
9-2 THE SURFACTANTS
Since much is required of the MP surfactant, we discuss surfactant solutions here.

This discussion can be no more than a précis of the voluminous literature on
surfactant properties. (For more on oil-recovering surfactants, see Akstinat, 1981.)
A typical surfactant monomer is composed of a nonpolar (lypophile) portion,
or moiety, and a polar (hydrophile) moiety; the entire monomer is sometimes called
an amphiphile because of this dual nature. Figures 9-2(a) and 9-2(b) show the
Figure 9-2 Representative surfactant molecular structures (from Lake, 1984)

357
molecular structure of two common surfactants and illustrate a shorthand notation for
surfactant monomers: The monomer is represented by a “tadpole” symbol, with the
nonpolar moiety being the tail and the polar being the head.
Surfactants are classified into four groups depending on their polar moieties,
(Table 9-1).
Anionics. As required by electroneutrality, the anionic (negatively charged)

surfactant molecule (distinct from monomer) is uncharged with an inorganic
metal cation (usually sodium) associated with the monomer. In an aqueous
solution, the molecule ionizes to free cations and the anionic monomer.
Anionic surfactants are the most common in MP flooding because they are
good surfactants, relatively resistant to retention, stable, and can be made
relatively cheaply.
Cationics. If the polar moiety is positively charged, the surfactants are
cationic. In this case, the surfactant molecule contains an inorganic anion to
balance the charge. Cationic surfactants are used little in MP flooding
because they are highly adsorbed by the anionic surfaces of interstitial clays.
Nonionics. A class of surfactants that have seen extensive MP use, mainly as
cosurfactants but increasingly as primary surfactants, is the nonionics. These
surfactants do not form ionic bonds but, when dissolved in aqueous solutions,
exhibit surfactant properties by electronegativity contrasts between their
constituents. Nonionics are much more tolerant of high salinities than
anionics and historically have been poorer surfactants.
Amphoterics. This class of surfactants contains aspects of two or more of the
other classes. For example, an amphoteric may contain both an anionic group
and a nonpolar group. These surfactants have not been used in oil recovery.
TABLE 9-1 CLASSIFICATION OF SURFACTANTS AND EXAMPLES (ADAPTED FROM

AKSTINAT, 1981)
Sulfonates Quaternary ammonium Alkyl-, Alkyl- aryl-, acyl-, Aminocarboxylic

Sulfates organics, pyridinum, acylamindo-, acyl- acids
Carboxylates imidazolinium, piperi- aminepolyglycol, and
Phosphates dinium, and sulfonon- polyol ethers
ium compounds Alkanolamides
Within any one class, there is a huge variety of possible surfactants. Figure
9-2 shows some of this variety by illustrating differences in nonpolar molecular
weight (C12 for the sodium dodecyl sulfate (SDS) versus C16 for Texas No. 1), polar
moiety identity (sulfate versus sulfonate), and tail branching (straight chain for SDS
versus two tails for Texas No. 1) all within the same class of anionic surfactants.
358
Besides these, there are variations in both the position of the polar moiety attachment
and the number of polar moieties (monosulfonates versus disulfonates, for example).
Even small variations can drastically change surfactant properties. For example,
sulfates tend to be less thermally stable than sulfonates. (For more details on the
effect of structure on surfactant properties, see Graciaa et al., 1981; Barakat et al.,
1983.)
The most common primary surfactant used in MP flooding is petroleum
sulfonates. These are anionic surfactants produced by sulfonating a pure organic
chemical (sometimes called synthetic sulfonates), an intermediate molecular weight
refinery stream, or when appropriate, even a crude oil itself. If R––C=C––H
represents the molecular formula of the feedstock, the sulfonation reaction proceeds
as
R––C=C––H + SO3 → R––C=C–– SO3− + H+ (9.2-1)
The reaction can also proceed to saturate the carbon–carbon double bond
R––C=C–– SO3− + H2 → R––CH––CH–– SO3− (9.2-2)
Here we adopt a shorthand notation that shows only the atoms participating in the
chemical reaction. The surfactant produced in Eq. (9.2-1) is an α-olefin sulfonate,
and that produced in Eq. (9.2-2) is an alkyl sulfonate. If the feedstock is aromatic, the
sulfonation produces an alkyl benzene sulfonate
R– + SO3 → R – – SO3− + H+ (9.2-3)
The sulfate in these reactions comes from bubbling SO3 gas through the
feedstock or through contact with a solvent the SO3 is dissolved into. The sulfonation
reactions (Eqs. 9.2-1 through 9.2-3) yield a highly acidic aqueous solution through
the parallel reactions
H2O + SO3 → H2SO4 (9.2-4)
+ − + 2−
H2SO4 → H + HSO 4 → 2H + SO 4 (9.2-5)
The solution is subsequently restored to a neutral pH by adding a strong base, such as
NaOH or NH3, dissolved in water. This neutralization step also provides the
counterion for the sulfonate; for the α-olefin sulfonate this is
Na+ + R––C=C–– SO3− → R––C=C–– SO3––Na (9.2-6)
If the feedstock is unrefined, a mixture of surfactant types will result. The

mixture can contain a distribution of isomeric forms, molecular weights, and degrees
of sulfonation (mono- versus disulfonation). The mixture is extremely difficult to
characterize except through several gross properties. Table 9-2 shows some typical
properties of commercial sulfonates. Typical molecular weights range from 350 to
450 kg/kg-mole, with the lower values indicating greater water solubility. In some
calculations, it is better to use the surfactant equivalent weight (molecular weight
divided by charge) instead of the molecular weight. Thus equivalent weight (mass per
equivalent) and molecular weight are the same for monosulfonates. Some products
contain impurities: unreacted oil from the sulfonation step and water from the
neutralization. Part of the surfactant, as purchased, is inactive. Inasmuch as it is the
359
TABLE 9-2 SELECTED PROPERTIES OF A FEW COMMERCIAL ANIONIC SURFACTANTS
Molecular Activity Oil Water Salt

Type
Company Surfactant name weight (wt %) (wt %) (wt %) (wt %)
ALCOLAC SIPONATE DS-10 350 98.0 Sodium dodecyl benzene sulfonate

ALCOLAC SIPONATE A168 350 70.0
CONOCO AES 14125 58.3 1.7 Alfonic ether sulfates
CONOCO AES 1412A 60.0 3.0 Alfonic ether sulfates
EXXON RL 3070 334 60.0 14.0 25.2 Alkyl aryl sodium sulfonates
KAO LS 8203 330 65.0 53.0 Linear alkyl sulfonate
KAO LS 8202 480 44.1 54.6 0.06 Linear alkyl sulfonate
LION LEONOX E 94.0 2.0 2.0
LION LEONOX D 350 94.0 2.0 3.0 Alpha olefin sulfonate
LION LION AJS-2 375 35.0
LION LEONOX K 570 30.0 Alfonic ether sulfates
SHELL ENORDET AOS 310–40 317 38.0 61.0 <1 Alcohol ethoxy sulfonate
SHELL ENORDET LXS 370–60 375 60.1 38.0 1.9 Linear alkyl xylene sulfonates
SHELL ENORDET 3ES–441–60 441 59.3 29.5 1.2 Linear alkyl xylene sulfonates
STEPAN PS HMW 50.7 24.4 22.1
STEPAN PS MMW 53.2 18.4 26.6
STEPAN PS 360 360 65.8 18.9 12.4
STEPAN PS 420 420 56.1 13.0 28.8
STEPAN PS 465 464 58.7 14.9 24.2
WITCO TRS 40 330–350 40–43 18.0 40.0
WITCO TRS 10–410 315–430 61–63 33.0 4–5
WITCO TRS 16 440–470 61–63 32.5 4–5
WITCO TRS 18 490–500 61–63 32.5 4–5
360
surfactant itself we are interested in, all slug concentrations should report the
surfactant concentration only (100% active basis).
In the following discussion, we ignore distinctions between surfactant types
by simply treating the surfactant as the tadpole structure of Fig. 9-2.
If an anionic surfactant is dissolved in an aqueous solution, the surfactant
disassociates into a cation and a monomer. If the surfactant concentration is then
increased, the lypophilic moieties of the surfactant begin to associate among
themselves to form aggregates or micelles containing several monomers each. A plot
of surfactant monomer concentration versus total surfactant concentration (Fig. 9-3)
is a curve that begins at the origin, increases monotonically with unit slope, and then
levels off at the critical micelle concentration (CMC). Above the CMC, all further
increases in surfactant concentration cause increases only in the micelle
concentration. Since CMCs are typically quite small (about 10–5 to 10–4
kg-moles/m3), at nearly all concentrations practical for MP flooding, the surfactant is
predominantly in the micelle form. This is the origin of the name micellar-polymer
flooding. The representations of the micelles in Fig. 9-3 and elsewhere are schematic.
The actual structures of the micelles are not static and can take on various forms.
Figure 9-3 Schematic definition of the

critical micelle concentration (from
Lake, 1984)
When this solution contacts an oleic phase (the term oleic phase indicates
this phase can contain more than oil), the surfactant tends to accumulate at the
intervening interface. The lipophilic moiety “dissolves” in the oleic phase, and the
hydrophilic in the aqueous phase. The surfactant prefers the interface to the micelle;
however, only small surfactant concentrations are needed to saturate the interface.
The dual nature of the surfactant is important since the accumulation at the interface
causes the IFT between the two phases to lower. The IFT between the two phases is a
function of the excess surfactant concentration at the interface. The excess is the
361
difference between the interface and bulk concentration. The interface blurs in much
the same manner as do vapor–liquid interfaces near a critical point.
The surfactant itself and the attending conditions should be adjusted to
maximize this effect, but this affects the solubility of the surfactant in the bulk oleic
and aqueous phases. Since this solubility also impinges on the mutual solubility of
brine and oil, which also affects IFTs, this discussion leads naturally to the topic of
surfactant–oil–brine phase behavior. Curiously, the surfactant concentration itself
plays a rather minor role in what follows compared to the temperature, brine salinity,
and hardness. This is true of many micellar properties.
9-3 SURFACTANT–BRINE–OIL PHASE BEHAVIOR
Surfactant–brine–oil phase behavior is conventionally illustrated on a ternary

diagram (see Sec. 4-3). By convention, the top apex of the ternary diagram represents
the surfactant pseudocomponent (i = 3), the lower left represents brine (i = 1), and the
lower right represents oil (i = 2). Table 9-3 summarizes these and other notational
conventions.
TABLE 9-3 NOTATION AND COMMON UNITS FOR MP FLOODING

i Specicies Concentration unit
1 Water Volume fraction

2 Oil Volume fraction
3 Surfactant Volume fraction
4 Polymer Weight percent or g/m3
5 Anion meq/cm3-pore volume
6 Divalents meq/cm3-pore volume
7 Cosurfactant Volume fraction
8 Monovalents meq/cm3-pore volume
j Phase
1 Aqueous
2 Oleic
3 Microemulsion
MP phase behavior is strongly affected by the salinity of the brine. Consider

the sequence of phase diagrams, Figs. 9-4 through 9-6, as the brine salinity is
increased. The phase behavior we now describe was originally given by Winsor
(1954) and adapted to MP flooding later (Healy et al., 1976; Nelson and Pope, 1978).
At low brine salinity, a typical MP surfactant will exhibit good
aqueous-phase solubility and poor oil-phase solubility. Thus an overall composition
near the brine–oil boundary of the ternary will split into two phases: an excess oil
phase that is essentially pure oil and a (water-external) microemulsion phase that
contains brine, surfactant, and some solubilized oil. The solubilized oil occurs when
362
globules of oil occupy the central core of the swollen micelles. The tie lines within
the two-phase region have a negative slope. This type of phase environment is
variously called a Winsor type I system, a lower-phase microemulsion (because it is
more dense than the excess oil phase), or a type II(–) system. We adopt the type II(–)
terminology here. II means no more than two phases can (not necessarily will) form,
and (–) means the tie lines have negative slope (Fig. 9-4). The plait point in such a
system PR is usually located quite close to the oil apex. Any overall composition
above the binodal curve is single phase.
Figure 9-4 Schematic representation

of the type II(–) system (from Lake,
1984)
For high brine salinities (Fig. 9-5), electrostatic forces drastically decrease
the surfactant’s solubility in the aqueous phase. An overall composition within the
two-phase region will now split into an excess brine phase and an (oil-external)
microemulsion phase that contains most of the surfactant and some solubilized brine.
The brine is solubilized through the formation of inverted swollen micelles, with
brine at their cores. The phase environment is a Winsor type II system, an
upper-phase microemulsion, or a type II(+) system. The plait point PL is now close to
the brine apex.
363
Figure 9-5 Schematic representation of

high-salinity type II(+) system (from
Lake, 1984)
The two extremes presented thus far are roughly mirror images: The
microemulsion phase is water-continuous in the type II(–) systems and oil-continuous
in type II(+) systems. The induced solubility of oil in a brine-rich phase, a type II(–)
system, suggests an extraction mechanism in oil recovery. Though extraction does
play some role, it is dwarfed by the IFT effect discussed below, particularly when
phase behavior at intermediate salinities is considered.
At salinities between those of Figs. 9-4 and 9-5, there must be a continuous
change between type II(–) and II(+) systems. The obvious change of a
counterclockwise tie line rotation and corresponding plait point migration is
incorrect; there is no salinity where the solubility of the surfactant in the brine- and
oil-rich phases are exactly equal. But there is a range of salinities where a third
surfactant-rich phase is formed (Fig. 9-6). An overall composition within the
three-phase region separates into excess oil and brine phases, as in the type II(–) and
II(+) environments, and into a microemulsion phase whose composition is
represented by an invariant point. This environment is called a Winsor type III, a
middle-phase microemulsion, or a type III system. To the upper right and left of the
364
Figure 9-6 Schematic representation of

optimal-salinity type III system (from
Lake, 1984)
three-phase region are type II(–) and II(+) lobes wherein two phases will form as
before. Below the three-phase region, there is a third two-phase region (as required
by thermodynamics) whose extent is usually so small that it is neglected (Anderson et
al., 1976). In the three-phase region, there are now two IFTs between the
microemulsion and oil σ32 and the microemulsion and water σ31.
Figure 9-7, a prism diagram, shows the entire progression of phase
environments from type II(–) to II(+). The type III region forms through the splitting
of a critical tie line that lies close to the brine–oil boundary as the salinity increases to
CSel (Bennett et al., 1981). A second critical tie line also splits at CSeu as salinity, is
decreased from a type II(+) environment. Over the type III salinity range, the
invariant point M migrates from near the oil apex to near the brine apex before
disappearing at the respective critical tie lines. Equally important, as the migration
takes place, the surfactant concentration in the microemulsion phase goes through a
minimum near where brine–oil ratio at the invariant point becomes 1.
The migration of the invariant point implies essentially unlimited solubility
365
Figure 9-7 Pseudoternary or “tent” diagram representation of micellar-polymer

phase behavior (from Lake, 1984)
of brine and oil in a single phase. This has generated intense research into the nature
of the type III microemulsion (Scriven, 1976). The middle-phase microemulsion
cannot be simultaneously oil- and water-external. Somewhere between CSel and CSeu,
the micelles contained therein undergo an inversion, and many phase properties (for
example, electrical conductivity) abruptly change from being characteristic of the
water to being characteristic of the oil. Moreover, several other properties (see Fig.
9-13) take on extreme values. Though logically appealing, the phase inversion
salinity does not necessarily indicate optimal salinity.
Several variables other than brine electrolyte content can bring about the Fig.
9-7 phase environment shifts. In general, changing any condition that enhances the
surfactant’s oil solubility will cause the shift from type II(–) to II(+). We discuss
some of the more important below.
Surfactant Structure
In general, increasing the importance of the nonpolar end of the surfactant will
increase oil solubility. Such changes include increasing the nonpolar molecular
weight, decreasing the tail branching (Graciaa et al., 1981), decreasing the number of
polar moieties (from disulfonates to monosulfonates), and decreasing the strength of
the polar moiety.
Two common measures of the competition between the hydrophile and
lipophile indicate oil solubility. The surfactant’s charge density is the number of
dissociated ions per molecule divided by the molecular size. Surfactant brine
solubility goes up as charge density increases. A second measure is the hydrophile–
lipophile balance (HLB) number. For certain types of surfactant (for example,
nonionics), the HLB number is simply related to molecular structure. But for others,
the HLB number cannot be uniquely defined apart from the oil and brine it is
366
competing for (Shinoda and Kunieda, 1979). Though both measures have enjoyed a
degree of success, they are difficult to apply to petroleum sulfonate because of the
many chemical species contained therein.
Cosurfactants
One of the first uses for cosurfactants was to adjust the surfactant pseudocomponent
so that the II(–) → II(+) transition occurs at different salinities. A water soluble
cosurfactant (for example, tertiary amyl alcohol, a second petroleum sulfate, or
n-butanol) also causes the surfactant to be more water soluble. Higher molecular
weight alcohols cause increased oil solubility (Salter, 1977). Bourrel et al. (1978)
have derived mixing rules for the properties of surfactant–cosurfactant mixtures.
Oil Properties
If the oil can be made more polar, it will act as a better solvent for the surfactant,
hastening the II(–) → II(+) transition. There are several measures for this tendency.
High specific gravity crudes tend to be rich in organic acids; thus surfactant oil
solubility is lower in high gravity oils (Puerto and Reed, 1982). Similarly, low
specific volume crudes behave in the same fashion (Nelson, 1982). Cash et al. (1976)
devised a measure of oil effects on surfactant–brine–oil phase behavior by comparing
the transitions observed with a crude to a refined hydrocarbon. The surfactant in all
cases is Texas No. 1 (Fig. 9-2) in a NaCl brine. If the transition from the II(–) →
II(+) for a crude occurs at the same salinity as the linear alkane, the alkane carbon
number (ACN) of the refined oil and the equivalent alkane carbon number (EACN)
of the crude are equal. Therefore, EACN is relatively easy to measure and gives an
indication of the model oil to be used in formulation. The same idea can be used to
categorize surfactants (Graciaa et al., 1981).
Decreasing Temperature
There is little generality in the tendency for the surfactant to dissolve in oil as
temperature increases. For most anionics, higher temperatures mean more brine
solubilities (Nelson and Pope, 1978). This trend is reversed for most nonionics.
Decreasing Pressure
MP phase behavior, being an all-liquid system, is relatively insensitive to pressure.

But Nelson (1982) has noted a substantial pressure effect in gassy crudes.
Interestingly, the trend here parallels that of the oil properties given above: As the
specific volume of the oil increases (through decreased pressure), the surfactant
becomes more water soluble.
Decreasing the surfactant’s oil solubility will cause the reverse of these
367
changes. Thus Fig. 9-7 could be redrawn with any of the above variables on the base
of the prism with the variable CSe increasing in the direction of increased oil
solubility. These observations have occupied a very great share of the MP literature.
Their utility will become apparent under our discussion of IFTs in the next section.
9-4 NONIDEAL EFFECTS
In much the same manner as the ideal gas law approximates the behavior of real
gases, Figs. 9-4 through 9-6 are approximations to actual MP phase behavior. Though
nonidealities are significant in many instances, in this section, we mention only the
most important.
1. At high surfactant concentrations or low temperatures (Scriven, 1976; Healy

and Reed, 1974) or even in the presence of pure surfactants (Salter, 1983),
phases other than those in Fig. 9-7 have been observed. These phases tend to
be high-viscosity liquid crystals or other condensed phases. The large
viscosities are detrimental to oil recovery since they can cause local viscous
instabilities during a displacement or decreased injectivity. Frequently,
low-to-medium molecular weight alcohols (cosolvents) are added to MP
formulations to “melt” these undesirable viscosities. Because most alcohols
are weak surfactants, the term cosurfactant has enjoyed popular usage for
these additions, as it has for the addition of other surfactants. When the brine
contains polymer, a condensed phase occurs at low surfactant concentration
because of exclusion of the polymer from the microemulsion phases.
Cosurfactants can be used to eliminate this polymer–surfactant
incompatibility (Trushenski, 1977).
2. When cosurfactants are present, it is often inappropriate to lump all the
chemicals into the surfactant apex of the prism in Fig. 9-7. If the
cosurfactants do not partition with the primary surfactant during a
displacement, much of the benefit from adding the chemical is lost; hence
surfactant–cosurfactant separation effects are an important concern. Efforts
to account for the preferential partitioning of the cosurfactant include a
quaternary phase behavior representation (Salter, 1978) and a pseudophase
theory (Hirasaki, 1982).
3. The type III salinity limits (CSel and CSeu) are functions of surfactant
concentration. This dependency may be visualized by tilting the vertical
triangular planes in Fig. 9-7 about their bases. This is sometimes called the
dilution effect.
One way to graphically represent the dilution effect is through the
salinity requirement diagram (Fig. 9-8). This diagram is a plot of overall
surfactant concentration C3 (horizontal axis) versus the salinity (vertical
axis). All other variables are held constant. Figure 9-8 represents salinity as
percent dilution of a particular high salinity brine. The upper curve shows the
boundary between the types II(+) and III environments or a curve of CSeu
368
Figure 9-8 Salinity-requirement

diagram (from Nelson, 1982)
versus C3. Similarly, the lower curve shows CSel versus C3; hence the region
between the two curves give the extent of the type III region as a function of
C3. Other plots (Glover et al., 1979; Bourrel et al., 1978) plot the extent of
observed three-phase behavior in a similar fashion. Figure 9-8 also shows a
three-phase region within the type III region.
The MP system in Fig. 9-8 shows a type III region that decreases with
salinity. For other surfactants and brines, the trend can be entirely opposite
(Bourrel et al., 1978). For ideal MP phase behavior, neither CSeu nor CSel
should depend on C3––that is, the salinity requirement diagram should
consist of two horizontal lines. Frequently, the behavior of soft brines will
approximate this since the dilution effect is particularly pronounced when the
brine contains significant quantities of divalent ions.
4. The phase behavior shifts are specific to the exact ionic composition of the
brine, not simply to the total salinity. Hence just as in polymer flooding, it is
insufficient to characterize the brine as merely “fresh” or in terms of its total
dissolved solids content. For anionic surfactants, other anions in solution
have little effect on the MP phase behavior, but cations readily cause phase
environment changes. Divalent cations (calcium and magnesium are the most
common) are usually 5–20 times as potent as monovalent cations (usually
sodium). Divalents are usually present in oil-field brines in smaller quantities
369
than monovalents (Fig. 8-1), but their effect is so pronounced that it is
necessary, as a minimum, to separately account for salinity––total dissolved
solids––and hardness––total divalent cation concentration. Nonconstant
monovalent–divalent ratios will also cause electrolyte interactions with clay
minerals through cation exchange. The disproportionate effects of the salinity
and hardness are accounted for by defining a weighted sum of the
monovalent and divalent concentrations as an “effective” salinity CSe.
The salinity effects discussed here are much less significant with
nonionic surfactants where there are no ionic associations. Even for anionics,
they can be greatly attenuated by adding nonionic cosurfactants. The
cosurfactant monomers add into the micelle between the larger primary
surfactant monomers, thus lessening the charge density of the micelle surface
and making the “mixed” micelle more like a nonionic.
9-5 PHASE BEHAVIOR AND INTERFACIAL TENSION
You may be wondering what this discussion of MP phase behavior has to do with the
goal of recovering oil through lowered IFT. Early MP flooding literature contains
much information about the techniques of measuring IFTs and what causes them to
be low (Cayias et al., 1975). IFTs depend on the types and concentration of
surfactant, cosurfactant, electrolyte, oil, polymer, and temperature. However, in
surely one of the most significant advances in all MP technology, all IFTs have been
shown to directly correlate with the MP phase behavior. Healy and Reed (1974)
originally proposed the correlation, which has been theoretically substantiated by
Huh (1979) and experimentally verified by several others (for example, Glinsmann,
1979; Graciaa et al., 1981).
A practical benefit of this correlation is immediately realized: Relatively
difficult measurements of IFTs can be largely supplanted by relatively easy phase
behavior measurements. Indeed, in the recent literature, the behavior of IFTs has
been inferred by a narrower subset of phase behavior studies based on the
solubilization parameter (Bourrel et al., 1978). As important as this benefit is, a more
important benefit is that the correlation logically provides a basis for MP design. We
discuss design in Sec. 9-13.
To investigate further the relation between IFTs and phase behavior, let C23,
C13, and C33 be the volume fractions of oil, brine, and surfactant in the microemulsion
phase. According to Figs. 9-4 through 9-6, the microemulsion phase is present at all
salinities; hence all three quantities are well defined and continuous. For systems
containing alcohol, C33 is the surfactant coordinate less the cosurfactant content.
Solubilization parameters between the microemulsion-oleic phases S32, for type II(–)
and III phase behavior, and between the microemulsion-aqueous phases S31 for type
II(+) and III are defined as
370
C
S32 = 23 (9.5-1a)
C33
C
S31 = 13 (9.5-1b)
C33
The interfacial tensions between the corresponding phases, σ32 and σ31, are empirical
functions only of S32 and S31. Figure 9-9 shows a typical correlation.
Figure 9-10 shows the corresponding behavior of the solubilization
parameters and IFTs in a different manner. Consider a locus at constant oil, brine,
and surfactant overall concentrations in Fig. 9-7, but with a variable salinity. If
nonideal effects are unimportant and the locus is at low surfactant concentration and
intermediate brine–oil ratios, σ32 will be defined from low salinity up to CSeu, and σ31
from CSel to high salinities. Both IFTs are the lowest in the three-phase type III region
between CSel and CSeu where both solubilization parameters are also large. Further,
there is a precise salinity where both IFTs are equal at values low enough (about 1
μN/m) for good oil recovery. This salinity is the optimal salinity CSopt for this
Figure 9-9 Correlation of solubilization parameters with interfacial tensions (from

Glinsmann, 1979)
371
Figure 9-10 Interfacial tensions and solubilization parameters (from Reed and
Healy, 1977)
particular surfactant–brine–oil combination, and the common IFT is the optimal IFT.
Optimal salinities have been defined on the basis of equal IFTs, as in Fig. 9-10, equal
solubilization ratios (Healy et al., 1976), equal contact angles (Reed and Healy,
1979), and the midpoint between CSeu and CSel. Fortunately, all definitions of optimal
salinity give roughly the same value.
The optimal salinity based on solubilization parameters also corresponds to
the salinity where oil recovery in a core flood is a maximum. Figure 9-11 illustrates
this oil-recovery optimal salinity. The middle panel, Fig. 9-1lb, shows a plot similar
to the upper panel in Fig. 9-10 for a different surfactant system; the lower panel
shows the oil recovery for a series of constant-salinity core floods. The optimal
salinity based on solubilization parameters, IFTs, and oil recovery agree well. Since
optimal phase behavior salinity is the same as maximum oil recovery salinity, clearly
one of the goals of an MP design is to generate this optimal salinity in the presence of
372
Figure 9-11 Correlation of phase volume and IFT behavior with retention and oil
recovery (from Glinsmann, 1979)
the surfactant. The optimal salinity does not correspond to minimum surfactant
retention (Fig. 9-11c), but this is because of competing effects as we mention below.
Because of the dilution effect mentioned above, maximum oil recovery is
really where the combination of electrolyte, surfactant, and cosurfactant
concentrations bring about maximum solubilization parameters. Hence one should
speak of optimal conditions rather than optimal salinity. The optimal salinity
terminology is deeply embedded within the MP literature, but it is precise only for
the ideal Fig. 9-7 phase behavior. Do not confuse the optimal salinity CSopt, an
intrinsic property of the oil–brine–surfactant combination, with the prevailing salinity
CSe, an independent variable in the MP design.
Optimal salinities can vary greatly depending on the nature of the surfactant
and brine pseudocomponents. But it is dismaying that for many commercially
attractive surfactants in most MP candidate reservoirs, the optimal salinity is smaller
than the resident brine salinity. Optimal salinities can be raised by adding to the slug
373
any chemical that increases the primary surfactant’s brine solubility. Adding
cosurfactants to the MP slug normally increases the optimal IFT.
The notion of optimal conditions is directly connected to the phase behavior
of MP systems. Even properties of MP systems apparently unrelated to phase
behavior (retention, for example) are functions of salinity, cosurfactant concentration,
and temperature. This observation leads to the interesting speculation that all MP
properties (retention, phase behavior, IFT, mobilities) correlate to optimal salinity
and, perhaps, to solubilization parameters.
9-6 OTHER PHASE PROPERTIES
Our understanding of MP phase behavior follows from the ternary representation of

the Winsor phase behavior progression. Other representations are common,
particularly to show phase properties.
A very useful phase-behavior representation is the volume fraction diagram
(VFD) (Fig. 9-12). Imagine a point of fixed overall composition (parallel to the
salinity axis) in the ternary planes in Fig 9-6. The volumes of each phase are
observed and plotted as the brine salinity changes. Starting with a low salinity, the
VFD shows a succession of decreasing oleic-phase volume and increasing aqueous-
phase volume with some three-phase overlap in the middle. If the overall surfactant
concentration is low and the brine–oil ratio (WOR) is about 1, the appearance of the
lower brine phase corresponds approximately to the onset of the type III region (CSel),
and the disappearance of the upper oleic phase corresponds approximately to the
termination of the type III region (CSeu). The salinity at which the brine and oleic
phases have equal volumes is a good approximation of the optimal salinity if the
surfactant and cosurfactant concentration is low enough. Compare the VFD in Fig.
9-11a to Fig. 9-12b.
Varying salinity while holding other variables constant is sometimes called a
salinity scan. Varying the salinity is the most common presentation of the VFD;
however, a derivative of the VFD, in which the cosurfactant concentration is varied
in place of the salinity, is sometimes useful. To minimize the number of
measurements, each scan can be relatively coarse (about ten measurements) and then
supplemented with fill-in measurements to refine the estimate of the important
events.
Of course, any phase property can be plotted in place of the phase volumes.
Figure 9-13 shows the microemulsion-phase viscosity as a function of salinity. Over
this range, the microemulsion phase, as defined above, is continuous and shows a
viscosity maximum at a salinity near the optimal. The maximum indicates molecular
ordering in the phase that seems to be the strongest at the phase inversion salinity.
Such maxima can be either beneficial, if it can be used to provide mobility control in
the slug, or detrimental, if it leads to excessively viscous fluids. It was to counteract
the latter tendency that cosurfactants were first added to MP slugs. Over the same
salinity range in Fig. 9-13, the excess phase viscosities do not change appreciably.
374
Figure 9-12 Phase volume diagrams (salinity scans) at three water–oil ratios
(from Englesen, 1981)
375
Figure 9-13 Microemulsion phase viscosity as a function of salinity (from Jones,

1981)
9-7 QUANTITATIVE REPRESENTATION OF MICELLAR

PROPERTIES
Prediction of MP response rests on being able to quantitatively represent the

foregoing behavior in equations. Given the complexities, many of which we have
only alluded to, it is not possible to derive comprehensive representations. Here we
seek to capture the major features by following these assumptions.
1. All fluids are incompressible and mix ideally.

2. Temperature does not change, and the phase behavior is insensitive to
pressure. These restrictions mean phase behavior is driven only by changes
in the effective salinity CSe.
3. The ternary equilibria in Fig. 9-11 apply. For the moment, we neglect the
nonideal effects.
4. The height of the binodal curve passes through a minimum near the optimal
salinity CSopt. The minimum forces the optimal salinity based on IFT and
376
phase behavior to be equal since the solubilization parameters will be a
maximum at CSopt.
5. The splitting of the critical tie line as the type III system forms or disappears
is so close to the ternary base that we can take the incipient invariant points
to coincide with the left and right apexes at these events.
We strive for a representation that captures the basics of the MP behavior

without becoming burdened with an excessive number of parameters. Such simplicity
means equations are used to describe as much of the behavior as possible, and we
choose these equations to have a small number of adjustable parameters. Though
many of the equations are empirical, we strive to make limiting cases theoretically
rigorous.
Salinity Events
We form the equations so that the adjustable parameters have physical significance
on the previously described diagrams. The effective salinity corresponding to the
type II(–) to III transition CSel is approximately the salinity on a VFD where the third
microemulsion phase appears. CSeu is where the microemulsion phase disappears. For
best approximation for both quantities, use a VFD with a water–oil ratio of 1. At this
water–oil ratio, the optimal salinity CSopt is the CSe where the excess phases have
equal volumes.
In what follows, the effective salinities are normalized by CSopt. The resulting
dimensionless effective salinities
C
CSeD = Se (9.7-1)
CSopt
are those that control the phase behavior. Clearly, CSeD can take on any positive value
and is equal to 1 at optimal conditions.
Other events coming from the VFD relate to the surfactant maximum
coordinate on the binodal curve at low, optimal, and high salinities. At CSeD = 1, the
surfactant coordinate of the invariant point is
C3
C3 M = = C3max1 (9.7-2)
S3
where S3 is the volume fraction (saturation) of the microemulsion phase. The oil and
brine coordinates at optimal conditions are
1 − C3 M
C1M = C2 M = (9.7-3)
2
These equations assume the excess phases are free of surfactant.

At low salinity, the height of the binodal curve is
377
C3
C3max 0 = (9.7-4a)
S3 CSel
and the similar quantity at high salinity is

C3
C3max 2 = (9.7-4b)
S3 CSeu
Equations (9.7-4a) and (9.7-4b) generally provide underestimations of the binodal

curve heights. Figure 9-14 shows the quantities in Eqs. (9.7-2) and (9.7-4).
Figure 9-14 Definition of quantities for phase-behavior representation
Events not observable from the VFDs are the plait point locations. We
assume the oil coordinates of the plait point vary linearly between the limits of CSeuD
and CSelD as shown in Fig. 9-15. C2PL and C2PR, the left and right plait point oil
coordinates, apply to the type II(–) and II(+) systems, respectively. also shows the
assumed linear variation of the invariant point oil coordinate C2M. The superscript *
refers to low and high salinity limiting cases. Typical values of C2PL and C2PR are 0.05
and 0.95, respectively.
These seven parameters (CSelD, CSeuD, C3max0, C3max1, C3max2, C2PL, and C2PR)
are sufficient to define the phase behavior with a few additional assumptions.
Binodal Curve
We use the same formalism to represent the binodal curve in all phase environments.
For type III, this means (Fig. 9-14) the two two-phase lobes are defined by a
378
Figure 9-15 Migration of plait and

invariant points with effective salinity
continuous curve. For simplicity, let the Hand representation from Eq. (4.4-23) define
the binodal curve with BH = –1.
C3 j ⎛ C1 j ⎞
= AH ⎜ , j = 1, 2, or 3
⎜ C3 j ⎟⎟
(9.7-5)
C2 j ⎝ ⎠
Equation (9.7-5) forces the binodal curve to be symmetric. Solving Eq. (9.7-5) for C3j
in terms of C2j gives (recalling that C1j + C2j + C3j = 1)
1
C3 j = ([(C2 j AH ) 2 + 4C2 j AH (1 − C2 j )]1/ 2 − C2 j AH ) (9.7-6)
2
Since the curve is symmetric, the maximum C3j occurs when C2j = C1j or,
alternatively, when C2j = (1 – C3j)/2, as in Eq. (9.7-3). This substituted into Eq.
(9.7-6) gives
2
⎛ 2C3max m ⎞
AHm =⎜ ⎟ (9.7-7)
⎝ 1 − C3max m ⎠
where m = 0, 1, or 2 corresponding to the salinity extremes. The AHm’s are linearly

interpolated as
AH = AH 0 + ( AH 1 − AH 0 )CSeD , CSeD ≤ 1
(9.7-8)
AH = AH 1 + ( AH 2 − AH 1 )(CSeD − 1), CSeD ≥ 1
Tie Lines in Two-Phase Systems
Since the treatment for the II(+) system is identical, let us deal with the tie lines in a
II(–) system only, but we use C2PL instead of C2PR. Again, using the Hand
representation, but with FH = 1, the phase distribution (Eq. 4.4-24) now becomes
379
C32 ⎛C ⎞
= EH ⎜ 33 ⎟ (9.7-9)
C22 ⎝ C13 ⎠
This equation applies at the plait point from which we have EH = C1PL/C2PL =
(1 – C2PL – C3PL)/C2PL. Since the plait point is also on the binodal curve, Eq. (9.7-6)
applies to give
1
1 − C2 PR − [( AH C2 PR ) 2 + AH C2 PR (1 − C2 PR )1/ 2 − AH C2 PR ]
EH = 2 (9.7-10)
C2 PR
C2PR being defined as a function of salinity, this equation and Eq. (9.7-8) give the
salinity dependence of EH.
Type III
The three-phase portion of this environment poses no difficulties since the excess
phases are pure by assumption, and the composition of the microemulsion phase is
given by the coordinates of the invariant point. For a given CSeD, C2M is fixed, and
C3M follows from Eq. (9.7-6).
The two-phase lobes are somewhat more trouble. Once again, we consider
only the II(–) lobe since the II(+) lobe is analogous. Let’s suppose the Hand
representations (Eqs. 9.7-5 and 9.7-9) apply to transformed concentrations (denoted
by superscript prime) where
C2′ j = C2 j sec ξ (9.7-11a)
C3′ j = C3 j − C2 j tan ξ (9.7-11b)
C1′ j = 1 − C2′ j − C3′ j , j = 2 or 3 (9.7-11c)
The angle ξ in these equations is from Fig. 9-14

C
tan ξ = 3 M (9.7-12a)
C1M
or, alternatively,
(C12M + C32M )1/ 2
sec ξ = (9.7-12b)
C1M
These relations allow the parameter EH to be expressed in terms of the untransformed

coordinates of the plait point as
C′ 1 − (sec ξ − tan ξ )C2 PR − C3 PR
EH = 1PR = (9.7-13)
C2′ PR C2 PR sec ξ
When the relation between C3 and C2 (Eq. 9.7-6) is used, this gives EH as a function
380
of salinity in the type II(–) lobe (CSel < CSe < CSeu). You can verify that these
manipulations for the type III lobes merge continuously with the two-phase
environments.
9-8 ADVANCED MP PHASE BEHAVIOR
For ideal surfactant–brine–oil systems, phase boundaries and optimal salinity would
be independent of brine salinity. This observation means plots of phase boundaries
and optimal salinity versus overall surfactant concentration––the salinity requirement
diagram (SRD)––would consist of horizontal lines. Petroleum sulfonate systems
generally do not manifest this type of behavior (Fig. 9-8 or Fig. 9-16); how ever, this
nonideality can be explained by the pseudophase theory (Hirasaki, 1982; Camilleri et
al., 1987). The theory also illustrates the correct measure of optima conditions,
accounts for preferential partitioning of a cosurfactant among the various phases, and
fits nicely into the formalism of the previous section.
Figure 9-17(b) shows a three-phase type III system with a water-external
Figure 9-16 Salinity requirement diagram for brine, decane, isobutanol, TRS 10-
410. Surfactant/alcohol = 1, Na/Ca = 10 (equivalents), Water–oil ratio = 1 (from
Prouvost, 1984)
381
middle phase, and Fig. 9-17(a) shows the pseudophase representation. Since the only
volume occupying components in the system are water, oil, surfactant, and
cosurfactant, the system is naturally represented on a quaternary diagram (Fig.
9-17a). All charged species, except those we mention below, exist in unassociated
form. The system consists of three pseudocomponents.
1. An oleic pseudocomponent consisting of the excess oil phase and the oil at
the center of the swollen micelles.
2. An aqueous pseudocomponent consisting of the excess brine phase and the
brine in the microemulsion. This phase contains all charged species not
associated with the micelles. Both the oleic and aqueous pseudocomponents
can contain cosurfactant but neither contains surfactant.
3. An interfacial pseudocomponent consisting of the surfactant, cosurfactant,
and counterions associated with the micelles. Micelles containing two or
more surfactant types are mixed micelles.
The theory has three separate facets: definition of effective salinity,

cosurfactant partitioning, and cation association with the mixed micelles.
Definition of Effective Salinity
The phase rule (Eq. 4.1-2) states there are two (NC = 7, NP = 3, NR = 2) degrees of
freedom for an optimal surfactant system at fixed temperature and pressure. Thus
there must be two variables specified to fix optimal conditions. The phase rule gives
no indication of what the two degrees of freedom should be except that they should
be intensive thermodynamic variables, an observation that rules out overall
concentrations.
Glover et al. (1979) present experimental data that suggests the divalent
cations bound to the micelles are the most direct indicator of optimality. They
suggest optimal salinity decreases linearly with f 63 , the fraction of the total divalent
cations bound to the micelles.
Optimal salinity ∼ − β 6 f 63 (9.8-1)
where β6 is a positive constant. Moreover, ample experimental evidence (Baviere et

al., 1981) suggests optimal salinity varies linearly with cosurfactant concentration.
Optimal salinity ∼ − β 7 f 73 (9.8-2)
where f 73 is the fraction of the cosurfactant associated with the micelles. The
constant β7 can be positive if the cosurfactant is more water-soluble than the
surfactant, and it can be negative otherwise. Equations (9.8-1) and (9.8-2) suggest the
following combination for the optimal salinity expressed as the anion concentration
in the aqueous phase:
C51 = C51
*
(1 − β 6 f 63 )(1 + β 7 f 73 ) (9.8-3)
382
Figure 9-17 Schematic representation of pseudophase theory for surfactant–

brine–oil–cosurfactant systems
383
*
In this equation, C51 is the optimal anion concentration in the absence of divalents
and cosurfactants. Equation (9.8-3) suggests a definition of effective salinity
C51
CSe = (9.8-4)
(1 − β 6 f 6 )(1 + β 7 f 73 )
3
CSe is the effective salinity used as a normalizing factor in Sec. 9-7. The remaining
tasks are to define f 63 and f 73
Cosurfactant Partitioning
In the following, we use Ci j to designate pseudophase compositions. Let’s estimate

the pseudocomponent compositions at point P(C1, C2, C3, C7) in Fig. 9-17(a). If the
pseudocomponents are the apexes of the indicated triangle,
C7 = C1C71 + C2 C72 + C3C73 (9.8-5)
That the pseudocomponent concentrations occupy the role of phase saturations in the
equation accounts for why the theory is called the pseudophase theory. Now let's use
partition coefficients to eliminate two of the pseudocompositions
C2 C3
7
K 21 = 71 , K 317 = 71 (9.8-6)
C7 C7
Ideally, the partition coefficients should be equal to the cosurfactant partition

coefficients in the absence of surfactant (Prouvost, 1984). These substituted into Eq.
(9.8-6) give the cosurfactant concentration in the aqueous pseudophase
⎛ 3 ⎞
C71 = C7 ⎜ ∑ C j K 7j1 ⎟ (9.8-7a)
⎝ j =1 ⎠
Because this pseudophase contains only cosurfactant and water

C11 = 1 − C71 (9.8-7b)
Equations (9.8-7b) and (9.8-6) can be used to calculate all pseudophase compositions
from overall compositions and the partition coefficients.
To use the equations of the previous section, we must express the overall and
phase compositions in terms of the pseudocomponents. We define CPi as
Volume of i + Volume of 7 associated with i
CPi = (9.8-8)
Total volume
which gives
CPi = Ci (1 + C7i ), i = 1, 2, 3 (9.8-9)
The CPi are overall concentrations and are to be used directly in the strict ternary
representation. The equations collapse to the Sec. (9-6) equations when cosurfactant
384
and divalent concentrations are zero. Further, Eq. (9.8-9), summed over three phases,
equals unity from Eq. (9.8-7a).
The fractional cosurfactant associated with the micelle follows directly from
this also. By definition
Volume of 7 in pseudophase 3
f 73 = (9.8-10)
Volume of pseudophase 3
which is simply
1
f 73 = (9.8-11)
1
1+ 3
C7
also by definition.
Divalent Cation Association
Competition for anionic sites on the micelle surface is through electrical forces.
Hence a cation exchange law of the following form applies:
⎛ (C93 ) 2 3 3 (C9 ) ⎞
2
⎜ 3 = β C3 ⎟ (9.8-12)
⎝ C6 C6 ⎠1
This equation is a form of Eq. (3.5-4) in which the constant (β3) is multiplied by a
factor that will convert the volume fraction C33 into units of meq/L3 of pore volume.
Equation (9.8-12) assumes the cosurfactant is nonionic and all the surfactant is
available to exchange.
Two types of electroneutrality now apply: on the micelle surface
3
C61 + C813 = C31 (9.8-13a)
and in the bulk aqueous phase

C61 + C81 = C51 (9.8-13b)
These form three equations in four unknowns, from which it becomes possible to
solve for the bound divalents in terms of the unassociated species concentrations
1
3
C61 = C313 + [r863 − ((r863 ) 2 + 4r863 C313 )1/ 2 ] (9.8-14a)
2
where
(C813 ) 2
r863 = β 3C33 3
(9.8-14b)
C61
Compare these equations to Eq. (3.5-7). Once the left side of Eq. (9.8-14a) is known,
the fraction of the total divalents bound to the micelle follows from
385
3
C
f 63 = 6
(9.8-15)
C3
and CSe can be estimated from Eq. (9.7-3).

The above theory will fit experimental data very well. Figure 9-16 shows the
agreement between estimated and calculated phase boundaries for a system of
petroleum sulfonate, decane, isobutanol, Na, and Ca. To construct this match,
Prouvost (1984) assumed the above theory applies to phase boundaries as well as to
optimal salinities. The theory and experiment agree well even though the SRD is far
from ideal.
9-9 HIGH CAPILLARY NUMBER RELATIVE PERMEABILITIES
A transport property that deserves treatment in a separate section is the high capillary
number relative permeability. In this section, we discuss two- and three-phase
experimental results based on the work of Delshad et al. (1987) (see Sec. 3-3 for
discussion of low capillary number relative permeabilities).
Few theoretical relations exist for relative permeabilities in general, much
less for those at high capillary number. We do know the extreme values of relative
permeability functions occur at residual phase saturations. The latter are functions o
capillary number Nvc through the capillary desaturation curve (CDC) (see Sec. 3-4).
Further, for very high values of Nvc, we expect the relative permeabilities to approach
straight-line functions between zero and unit endpoints with no residual phase
saturations. For low Nvc, the relative permeabilities should return to the two- or
three-phase high IFT functions. The variation between these extremes is not well
established.
High Nvc relative permeabilities are difficult to measure. In one type of
experiment, the large Nvc may be attained by increasing the flow rate. This technique
causes experiments to proceed rapidly since, as we saw in Sec. 3-4, Nvc must increase
by several factors of 10 before a significant effect occurs. Such high rates art clearly
unrepresentative of typical reservoir fluid velocities. If the high Nvc is established by
lowering the IFT, the experiments tend to be dominated by transient composition
changes. In principle, these transients could be analyzed by the method given in Sec.
9.10, but this requires knowing the relative permeabilities, whose measurement is the
point of the experiment.
The most reliable measurement is of steady-state relative permeabilities
using preequilibrated fluids. For micellar fluids in two-phase flow, this consists of
displacing a composition on one end of a tie line with another on the same tie line at
constant salinity. When the effluent and injected fractional flows are equal, and
transients caused by nonideal phase behavior are gone, the relative permeability to
the flowing phases may be calculated from the measured effluent cuts and pressure
drop. A similar provision exists in the three-phase ideal systems where all
386
compositions are in equilibrium at constant salinity. Of course, such transients may
take some time to die out; thus steady-state experiments can be time consuming. The
uniform saturations established by such a procedure follow from material balance or,
preferably, tracer data interpreted by a suitable numerical model (Delshad et al.,
1987).
Despite these difficulties, high Nvc relative permeabilities for two-phase flow
have been rather intensively measured, but three-phase data are rare. Figure 9-18
shows steady-state relative permeabilities to brine, oil, and microemulsion phases for
both two- and three-phase flow. The permeable medium was strongly water wet in
both cores A and B at high Nvc conditions. Nvc = 0.01 at the optimal salinity used in
the experiments. The micellar system under test closely followed ideal phase
behavior. From these high Nvc data, several observations can be made.
1. The residual phase saturations are nonzero. Of course, these values are points
on the CDC. Except for the oleic phase, whose endpoint was already high in
the water-wet medium, the endpoint relative permeabilities are substantially
different from their low Nvc values.
2. The high Nvc relative permeabilities are not straight lines. The curves in these
figures are the matches of the exponential forms Eq. (3.3-4) to the data. But
the exponents n1 and n2 in these equations are not substantially different from
their low Nvc values.
3. The two- and three-phase data follow essentially the same curves.
4. The relative permeability for all three phases are functions of their own
saturations. This observation is at odds with the high Nvc behavior of
three-phase gas, oil, and water flows (Stone, 1970).
5. Probably the most surprising conclusion is that the excess brine phase was
not the most strongly wetting phase as it was under low Nvc conditions. This
observation is supported by a variety of observations not present in Fig. 9-16.
However, the microemulsion and excess brine residual phase saturations
have about the same value at Nvc = 0.01.
6. The shape of the microemulsion curve is concave downward. This
observation is highly atypical of relative permeabilities and can be explained
only as wall or interfacial slippage.
For approximate calculation, let the exponential relative permeabilities of Eq.

(3.3-4) approximate two-phase high Nvc behavior. Suppose the CDC of a type II(–)
system is represented by Fig. 3-19, with (Nvc)c and (Nvc)t corresponding to the wetting
state of phase j. We can define linear interpolants for the endpoints and the
curvatures. For example, the endpoints vary according to
⎛ S′ ⎞
(kr02 )′ = (kr02 ) + ⎜1 − 3r ⎟ (1 − kr02 ) (9.9-1a)
⎝ S3r ⎠
S′
(kr03 )′ = ( kr03 ) + 2 r (1 − kr03 ) (9.9-1b)
S2 r
387
Figure 9-18 Two- and three-phase relative permeabilities (from Delshad et al., 1987)
where S2r ′ and S3r′ are the high Nvc residual phase saturations. This approximate
linear relation has been substantiated by Stegemeier (1976). The nonunit curvatures
of the relative permeabilities seem to persist beyond the point of zero residual phase
saturations; hence it seems reasonable that the logarithm of Nvc itself be used as an
interpolating function
388
⎡ N ⎤
n′j = n j + (1 − n j ) log ⎢ vc ⎥ , j = 2 or 3 (9.9-2)
⎣ ( N vc )c ⎦
Relations for type II(+) systems, where j = 1 or 3, follow analogous arguments.

We can now estimate two-phase relative permeabilities from phase behavior,
a solubilization parameter correlation, the CDC curves, and low Nvc relative
permeability curves. Suppose we know the overall composition of a type II(–) system
that splits into two equilibrium phases. The phase compositions follow the ternary
diagram. These can be converted to solubilization parameters using Eq. (9.5-1) and
then into IFTs from the appropriate correlations. We use this to calculate Nvc, and the
CDCs to estimate residual phase saturations. The high Nvc curves follow from Eqs.
(9.9-1) and (9.9-2). If additional data are available about viscosities, dip angles, and
densities, we can easily calculate phase fractional flows.
For three-phase flow, even such rough estimates are not warranted.
Theoretical models by Hirasaki et al. (1983) and Delshad et al. (1987), though
plausible in limiting senses, account neither for the intermediate wetting of the excess
brine phase nor for the observation that the phase relative permeabilities are functions
only of their own saturations. Clearly, we are hindered by a lack of understanding
about the pore-level nature of high capillary number flows.
9-10 FRACTIONAL FLOW THEORY IN MICELLAR-POLYMER

FLOODS
Fractional flow theory can be just as insightful for MP floods as for the solvent and
polymer floods we covered in Secs. 7-7 and 8-4. In fact, there are so many
similarities to those processes that we draw heavily on the material in those sections.
To make the analysis, we invoke the usual fractional flow assumptions:
incompressible fluid and rock, one- dimensional flow, and no dissipative effects. In
addition, we neglect the presence of the polymer drive (the polymer treatment can be
added as an exercise), assume three-component MP floods with a step change in
concentration at the origin of a time–distance diagram, and treat only those floods
with constant phase behavior environment. Further, we neglect surfactant retention
until later in this section where we invoke more restrictive assumptions about the
phase behavior. To shorten the development, we cover only the high-salinity type
II(+) floods. Fractional flow treatment for three-phase MP floods has not been
extensively investigated (Giordano and Salter, 1984), but it could be so treated with
the numerical technique of Sec. 5-7.
Ternary Landmarks
Figure 9-19(a) shows the basic phase and saturation behavior. This is very much like
the behavior in Fig. 7-31 except that the miscibility gap extends entirely across the
bottom edge of the ternary, and of course, water is explicitly included on the diagram.
389
Figure 9-19 Ternary diagram and

composition paths for micellar–polymer
system
All phase diagrams in this section have exaggerated two-phase regions. One
significant difference with solvent flooding is lines of residual oleic and aqueous
phase saturations merge with the binodal curve at some distance from the plait point.
This happens because the oleic-aqueous capillary number increases (IFT decreases)
rapidly as the plait point is approached, which causes S2r and S1r to approach zero
(see Sec. 3-4). For continuous surfactant injection, as we are treating here, this issue
is entirely secondary. But for finite slugs in highly dissipative displacements––that is,
the realistic cases––the rate of approach to zero S2r is very important. The aqueous-
oleic fractional flow curve follows from the large Nvc relative permeabilities we
discussed in Sec. 9-9.
MP Flooding without Retention
The relative permeability behavior does not affect the qualitative features of the
composition path diagram (Fig. 9-19b). The development in Sec. 7-7 applies directly:
We see the presence of “hair-pin” fast paths along tie lines, slow paths on either side,
390
and a succession on nontie line paths. Because of the graphical possibilities, we
assume the component distribution between phases is given by a family of straight
lines intersecting at C3 = 0 and C2 = C20 (the tie line envelope is a point on the C3
axis)
C3 j = η (C2 j − C20 ), j = 1 or 2 (9.10-1)
from which we have C10 = 1 − C20 . The parameter η is the slope of the phase
distribution line.
To review briefly, the component velocities along a tie line are
dF
vCi = i , i = 1, 2, 3 (9.10-2)
dCi
The nontie line paths carry the constant specific velocities given by
F − Ci0
vΔCi = i , i = 1, 2, 3 (9.10-3)
Ci − Ci0
At the tangent intersection of the tie line and nontie line paths, we must have
Fi − Ci0 dFi
= (9.10-4)
Ci − Ci0 dCi
which defines the two singular curves and allows the location of the appropriate
constructions on a fractional-flux–overall-concentration plot. Other paths include the
binodal curve itself and the equivelocity curve where f1 = S1.
In Fig. 9-19(b), we illustrate behavior for fractional flux curves whose
S-shape persists even to low IFT. Our task is to string together the paths so that the
composition route leads to monotonically decreasing composition velocities.
We focus on the three different injection conditions. Condition J1 is an
aqueous (oil-free) surfactant solution below the critical tie line extension, J2 is an
aqueous surfactant solution above the extension, and J3 is an oleic (brine-free) slug
below the extension. Conditions J1 and J2 represent low- and high-concentration
aqueous surfactant solutions, and condition J3 is an oil-soluble solution. In each case,
the initial condition will be at I, a uniform tertiary condition.
Figure 9-20 shows the composition route and the S2 and C3 profiles at fixed
tD for the low-concentration surfactant displacement. Starting at the injection
condition, the composition route enters the two-phase region along a tie line
extension, switches to the nontie line path at the second singular point on the tie line,
switches again to the fast path along the ternary base, and then to the initial condition
I. For typical fractional flux curves, this causes a shock to an oil bank saturation S2B,
and a mixed wave from S2B to S2 = 0. Following the tie line causes the curious effect
that the flowing surfactant concentration can be greater than the injected
concentration. The displacement can also be relatively inefficient if the spreading
portion of the oil bank rear is large.
For the high-concentration surfactant displacement (Fig. 9-21), the
composition route passes through the plait point, follows the equivelocity path to the
391
Figure 9-20 Composition route and

profiles for low-concentration surfactant
flood
oil bank saturation, and then on to the initial condition. This displacement is directly
analogous to a condensing gas drive miscible displacement since the surfactant wave
is indifferent and moves with unit specific velocity (compare this displacement with
the lower panel in Fig. 7-35b). As such, it is highly efficient; however, the greater
efficiency is bought with a higher surfactant concentration. The oil bank saturation is
also somewhat lower than in Fig. 9-20.
The oleic surfactant behavior is shown in Fig. 9-22. Here the composition
route also enters along a tie line extension, branches to a non-tie line path at the first
singular point, and then on to the oil bank and the initial condition. In many respects,
this displacement is the mirror image of that in Fig. 9-20. However, the ultimate
microemulsion phase saturation is unity, meaning the oil bank saturation S2B is
between the initial and 1.0. The surfactant concentration decreases monotonically in
this displacement, which as in Fig. 9-20, can also be inefficient.
There is great variety of behavior in the displacement character even under
392

profiles for high-concentration surfactant
flood
the restrictive assumptions invoked here. Some of this variety is present in the
constructions used to infer Figs. 9-20 through 9-23. The cases for type II(–) are
analogous. The nature of the composition route does not change with the shape of the
fractional flux curve even though the latter greatly affects the efficiency of the
displacement.
MP Flooding with Retention
Adding retention complicates the analysis because the composition route no longer
follow tie lines. But by making a few additional assumptions, we can develop a
fractional flow solution that uses fractional flow curves instead of fractional fluxes.
Let us now analyze the type II(–) system where the right plait point is in the
oil corner of the ternary, and the amount of solubilized oil in the microemulsion
phase is negligible. The aqueous and microemulsion phase are now equivalent (S1 =
S3). If the injected slug composition is below a tangent from the binodal curve at the
393

profiles for high-concentration oleic
surfactant flood
plait point, it must necessarily be on a tie line even if it contains no oil. Let the
residual oil saturation on this tie line be S 2′r , the ultimate value of a low IFT (high
Nvc) aqueous-phase fractional flow curve as shown in Fig. 9-24. This figure also
shows the water-oil fractional flow curve f1 along the tie line on the base of the
ternary. Since this aqueous slug miscibly displaces the irreducible water, the velocity
of the corresponding indifferent wave is
f1s
vC3 = (9.10-5)
S1 + D3
from Eq. (5.4-8a). Note that f1s is the microemulsion (aqueous) phase high Nvc
fractional flow. In this equation, D3 is the surfactant’s frontal advance loss given by
1 − φ C3 s
D3 = (9.10-6)
φ C3 J ρ s
394
Figure 9-23 Fractional flux and composition routes for aqueous and oleic surfactant
displacements
The most general case occurs when the rear of the oil bank travels as a mixed wave.
At the leading edge of the spreading portion of this wave, the specific velocity in Eq.
(9.10-5) must be equal to the specific oil velocity at some saturation S1* given
implicitly by
⎛ df1s f1s ⎞
⎜ = ⎟ (9.10-7)
⎝ dS1 S1 + D3 ⎠ S * 1
The specific velocity of the shock portion of the oil bank rear is
f 2 B − f 2s ( S2* )
vΔC2 = (9.10-8)
S2 B − S2*
This must be equal to vC3 evaluated at S 2* = 1 – S1* . If the oil bank front is a shock, it
travels with velocity given by
f − f2I
vΔC2 = 2 B (9.10-9)
S2 B − S2 I
395
Figure 9-24 Graphical construction for simplified II(–) surfactant displacements
These equations are entirely suggestive of the polymer flooding construction

in Sec. 8-4. This parallel is also apparent from comparing the construction given in
Fig. 9-24 with the one in Fig. 8-10. The construction of the time–distance and profile
diagrams corresponding to Fig. 9-24 is left as an exercise.
An issue not dealt with in Sec. 8-8 is the minimum slug size needed to satisfy
retention. Let’s suppose the surfactant displacement is pistonlike, that is, S2I =
S 2′ r = S2* . The minimum surfactant slug size is attained when the similarly pistonlike
surfactant–polymer front overtakes the surfactant front at the injection end of the
medium. This gives a minimum surfactant slug size of tDs = D3, meaning the frontal
advance loss is an expression of the retention capacity of the medium expressed in
units consistent with the slug size. Therefore, knowing D3 is the beginning point in
estimating surfactant requirement in MP flooding. The above result does not depend
on the existence of a pistonlike surfactant front.
9-11 ROCK-FLUID INTERACTIONS
Brine salinity and hardness would have far less importance to MP flooding if the host
permeable medium were unreactive. Unfortunately, in all but the most artificial
cases, reservoir minerals provide an almost limitless source of monovalent and
396
divalent cations as well as ample sites for surfactant retention. Two sources of cations
are mineral dissolution and cation exchange. Dissolution usually occurs at such a low
level that it can be neglected in MP floods (but not in high-pH floods). Cation
exchange is rarely negligible. For this reason, we discuss it at some length in this
section. In the second half of the section, we deal with surfactant retention.
Cation Exchange
We treat the simplest case of monovalent–divalent exchange in single-phase flow in

the absence of surfactant or oil. The displacement satisfies the fractional flow
assumptions. (For more complicated treatment, see Pope et al., 1978.)
Each point in the permeable medium must satisfy solution electroneutrality
C5 = C6 + C8 (9.11-1a)
and electroneutrality on the clays
Z v = C6 s + C8 s (9.11-1b)
The units on all concentrations are in equivalents per unit pore volume, and the
superfluous phase subscript has been dropped. These equations imply the
monovalent, divalent, and anion concentrations are not independent; hence for
convenience, we choose to proceed with the divalent and anion concentrations as the
dependent variables. At local equilibrium, each point in the medium must also satisfy
C82s C2
= KN 8 (9.11-2)
C6 s C6
which is simply the cation exchange isotherm of Eq. (3.5-4). Using Eq. (9. 11-1), we
can express the adsorbed divalent concentration as
⎛ K r ⎡1 ⎛ Z 1 ⎞ ⎤ ⎞⎟
1/ 2
⎜
C6 s = Z v 1 + N
⎢ −⎜ v
+ ⎟ ⎥ (9.11-3a)
⎜ Zv ⎢ 2 ⎝ K N r 4 ⎠ ⎥ ⎟
⎝ ⎣ ⎦⎠
where
C 2 (C − C6 ) 2
r= 8 = 5 (9.11-3b)
C6 C6
Equations (9.11-3) are the basic equilibrium representations.

Let’s now consider the displacement of solution I by solution J under the
above conditions. The coherence conditions (Eq. 5.6-14)
−1 −1
⎛ dC5 s ⎞ ⎛ dC6 s ⎞
⎜1 + ⎟ = vC5 = vC6 = ⎜1 + ⎟ (9.11-4)
⎝ dC5 ⎠ ⎝ dC6 ⎠
are satisfied at all points in the medium. Equation (9.11-4) implies

dC5 s dC6 s
= =λ (9.11-5)
dC5 dC6
397
where λ is the eigenvalue for this problem. In matrix form, Eq. (9.11-5) becomes
⎛ C55 C56 ⎞ ⎛ dC5 ⎞ ⎛ dC5 ⎞
⎜ ⎟⎜ ⎟ = λ⎜ ⎟ (9.11-6)
⎝ C65 C66 ⎠⎝ dC6 ⎠ ⎝ dC6 ⎠
where C65 = (∂C6 s / ∂C5 )C6 , and so on. The matrix on the left side has a row of zeros
because the anion does not adsorb. Solving Eq. (9.11-6) for the eigenvalues gives
λ − = 0, λ + = C56 (9.11-7)
From this, it is obvious that λ+ > λ– and that the wave corresponding to λ– is faster
than that corresponding to λ+ from Eq. (9.11-4).
The eigenvector corresponding to each eigenvalue gives the concentration
change across each wave. For the fast wave, inserting λ– gives
C65 dC5 + C66 dC6 = 0 (9.11-8a)
and from the slow wave, inserting λ gives
+
dC5 = 0 (9.11-8b)
Immediately we see that the anion concentration is constant across the slow
wave since dC5 = 0. The fast wave interpretation is only a little less obvious.
Equation (9.11-8a) is the change in C5 and C6 that would occur at constant C6s. We
can see this by setting the total differential dC6s
dC6 s = C66 dC6 + C65 dC5 (9.11-9)
equal to zero. The result (Eq. 9.11-9) is independent of the form of the exchange
isotherm.
The coherent solution, therefore, predicts two waves: an indifferent salinity
wave that moves at unit velocity across which the clay concentration remains
constant and an exchange wave where the clay changes to be in equilibrium with the
injected solution at constant anion concentration. The character of the latter wave
depends on the direction in which the concentration velocity is increasing through
Eqs. (9.11-3) and (9.11-4).
For example, consider Fig. 9-25(a), which shows the composition space for
the two cation system plotted on a ternary representation. Lines of constant anion
concentration are parallel to the right edge, and lines of constant clay composition are
curves converging to the lower left corner. These curves are described by Eq.
(9.11-3). Both sets of lines form the composition path diagram. The bold line
segments I-J′ (salinity wave) and J′-J (exchange wave) form the composition route.
This particular sequence is selected because it is the only one for which the
concentration velocities monotonically decrease from I to J. The exchange wave is
spreading if the concentration velocity decreases from J′ to J; otherwise, it is a shock.
Figure 9-25(b) shows a time-distance diagram for the former case.
To illustrate the accuracy of the above predictions, Fig. 9-26 shows the
effluent histories of two laboratory core floods through which are flowing solutions
containing only calcium (divalents), sodium (monovalents), and chloride (anion). In
398
Figure 9-25 Diagrams for two exchanging cation case (from Pope et al., 1978)
both cases, the injected calcium was the same; hence in the absence of exchange, the
effluent calcium concentration should not change. But because the anion
concentration changes, cation exchange occurs, and the effluent calcium does change.
In Fig. 9-26(a), the increased anion concentration causes calcium to be expelled from
the clays. In Fig. 9-26(b), the reverse occurs. In both cases, the prediction based on
the coherent treatment agrees well with the observed results. Calculated results
including dispersion match even better (Lake and Helfferich, 1978).
These results hold immense practical significance for MP flooding in general
and the use of low-salinity preflushes in particular. One of the intentions of a preflush
is to remove divalent cations so that the slug can work more effectively. However,
the above theory suggests the following hindrances:
399
Figure 9-26 Comparison between theory and experiment for two exchanging cation
displacement (from Pope et al., 1978)
400
1. It is entirely possible to inject a low-salinity preflush that actually loads the
clays with divalent cations because changes in the ratio r determine clay
loading. If r decreases, the clays will take up divalents, which regardless of
the salinity, are available for subsequent release into solution.
2. Even if r decreases so that the clays unload divalents, this normally takes a
large preflush because the exchange wave velocity is very slow at typical
cation exchange capacities and brine concentrations.
3. If the injected solution is entirely devoid of divalents, the clays will still only
partially unload because the dissolution of a small amount of divalent-
containing minerals acts as a persistent source of hardness.
One philosophy for preflushes is to avoid upsetting the clays at all costs.
Doing this is simple in principle: One just injects the preflush, slug, and polymer
drive at the same r ratio as exists in the formation brine. But in practice, this
procedure is complicated by dispersion-induced mixing (Lake and Helfferich, 1978)
and by exchange of divalents with micelles (see Sec. 9-8).
Surfactant Retention
Surfactant retention is probably the most significant barrier to the commercial

application of MP flooding. The problem here is one of selectivity. The surfactants
should have good selectivity for oil–water interfaces, but they should also have poor
selectivity for fluid–solid interfaces.
Surfactants are retained through four mechanisms.
1. On metal oxide surfaces (Fig. 9-27), the surfactant monomer will physically
adsorb through hydrogen bonding and ionically bond with cationic surface
sites. At higher surfactant concentrations, this association includes tail-to-tail
interactions with the solution monomers, resulting in proportionally greater
adsorption. At and above the CMC, the supply of monomers becomes
constant, as does the retention. The Langmuir-type isotherm of adsorption
versus overall surfactant concentration resembles the CMC plot in Fig. 9-4,
which can be expressed as
a3C3
C3 s = (9.11-10)
1 + b3C3
where a3/b3 represents the plateau adsorption value. C3 here is the surfactant
concentration in the liquid phase wetting the substrate. The parameter b3 is
large, being related to the CMC, which is very small compared to practical
surfactant concentrations (see Fig. 9-28). The surfactant isotherm therefore
attains its plateau at such a low C3 that it may be usefully represented as a
step function. This form of retention should be reversible with surfactant
concentration. The parameters a3 and b3 are functions of salinity since they
depend on the number of surface sites available for adsorption.
401
Figure 9-27 Surfactant adsorption on metal oxide surfaces (adapted from Harwell,
1983)
Figure 9-28 Effect of cosurfactant on surfactant retention. Surfactant is 4-phenyl dodecyl

benzene sulfonate. (adapted from Fernandez, 1978)
2. In hard brines, the prevalence of divalent cations causes the formation of

surfactant–divalent complexes
2R-SO3− + M 2 + → M (R-SO3 ) 2 ↓ (9.11-11)
which have a low solubility in brine. Precipitation of this complex will lead
to retention. When oil is present, it can compete for the surfactant. Of course,
402
the precipitate must also compete with the micelles for the surfactant
(Somasundrun et al., 1979).
3. At hardness levels somewhat lower than those required for precipitation, the
preferred multivalent surfactant will be a monovalent cation that can
chemically exchange with cations originally bound to the reservoir clays
(Hill and Lake, 1978).
R-SO3− + M 2 + → MR-SO3+
(9.11-12)
Na-Clay + MR-SO3+ → MR-SO3 -Clay + Na +
This effect is not unlike the divalent-micelle effect we discussed in Sec. 9-8.
The surfactant bound to the clays will exhibit tail–tail interaction as in Fig.
9-27.
As a consequence of the ionic bonding and tail–tail interactions, adding a
cosurfactant will reduce both types of retention (Fig. 9-28). Cosurfactants
perform this service in two ways: (1) by filling surface sites that might
otherwise be occupied by surfactant and (2) by mitigating the tail-to-tail
associations. The retention expressed by Eq. (9.11-12) can also be lessened
by filling the clay sites with a more preferred metal cation. This form of
retention is reversible with both M2+ and surfactant concentration.
4. In the presence of oil in a II(+) phase environment, the surfactant will reside
in the oil-external microemulsion phase. Because this region is above the
optimal salinity, the IFT is relatively large, and this phase and its dissolved
surfactant can be trapped. Figure 9-29 illustrates this phenomenon. The filled
squares represent the surfactant injected, and the open squares the surfactant
retained in a series of constant-salinity core floods. Retention increases
smoothly with salinity (both a3 and b3 are functions of salinity) until 3%
NaCl, at which point it increases so substantially that all the injected
surfactant is retained. 3% NaCl is just above CSeu for this system; hence the
deviation can be nicely explained by phase trapping. A similar phase
trapping effect does not occur in the II(–) environment because the aqueous
mobility buffer miscibly displaces the trapped aqueous-external
microemulsion phase. Using less than optimal salinities can, therefore,
eliminate phase trapping. This form of retention is strongly affected by the
MP phase behavior.
Most studies of surfactant retention have not made these mechanistic

distinctions. Therefore, which mechanism predominates in a given application is not
obvious. All mechanisms retain more surfactant at high salinity and hardness, which
in turn, can be attenuated by adding cosurfactants. Precipitation and phase trapping
can be eliminated by lowering the mobility buffer salinity at which conditions the
chemical adsorption mechanism on the reservoir clays is predominant. In this event,
there should be some correlation of surfactant retention with reservoir clay content.
Fig. 9-30 attempts to make this correlation by plotting laboratory and field surfactant
retention data against clay fraction. The correlation is by no means perfect since it
403
Figure 9-29 Surfactant retention caused by phase trapping; 3% NaCl is a type

II(+) microemulsion system (from Glover, et al., 1979)
ignores variations in MP formulation and clay distribution as well as salinity effects.

However, the figure does capture a general trend useful for first-order estimates of
retention. In addition, note that the difference between lab- and field-measured
retention is not significant. This observation implies that surfactant retention can be
effectively measured in the laboratory.
A useful way to estimate the volume of surfactant required for an MP slug is
through the dimensionless frontal advance lag D3 defined in Eq. (9.10-6). C3s is the
surfactant retention from Fig. 9-30 (the plateau value a3/b3 on the appropriate
isotherm), φ is the porosity, C3J is the surfactant concentration in the MP slug, and ρs
is the surfactant density. D3 is a fraction that expresses the volume of surfactant
retained at its injected concentration as a fraction of the floodable pore Vpf. For
optimal surfactant usage, the volume of surfactant injected should be large enough to
contact all Vpf but small enough to prevent excessive production of the surfactant.
404
Figure 9-30 Overall surfactant retention correlated with clay content (adapted
from Goldburg et al., 1983)
Besides wasting an expensive chemical, the produced surfactant can cause severe
produced emulsions. Thus the MP slug size should be no less but not much greater
than D3. The total amount of surfactant injected is independent of injected surfactant
since, from Eq. (9.10-6), D3C3J is independent of C3J.
9-12 TYPICAL PRODUCTION RESPONSES
In this section, we review responses of typical laboratory core and field flood
showing the important features and expectations of MP flooding.
Laboratory Flood
Fig. 9-31 plots an effluent response of a typical MP flood in a Berea core showing oil
cut, produced surfactant (Mahogany AA), cosurfactant, (isopropyl alcohol), polymer,
and chloride anion concentrations. All concentrations have been normalized by their
respective injected values. The chloride indicates the salinity in this flood. At the top
of the figure is the phase environment of the produced fluids. The slug size is tDs =
0.1, and the horizontal axis is tD, the volume of fluid injected since the start of the
slug expressed as a fraction of the core’s pore volume. There was no preflush. (For
further details of this and similar core floods, see Gupta, 1980.)
Figure 9-31 shows a typical, though by no means optimal, oil recovery
experiment. Before surfactant injection, the core was waterflooded so that it produces
no oil initially. Oil breaks through at about tD = 0.2, with relatively sustained cuts of
405
Figure 9-31 Typical core-flood produce response (from Gupta, 1980)
about 40% until about tD, = 0.6, at which point the surfactant appears. The behavior
in this portion of the flood is consistent with the fractional flow theory in Sec. 9-10.
About 60% of the produced oil is free of the injected chemicals. That 40% of the oil
is produced with the surfactant indicates a viscous instability apparently caused by
nonideal phase behavior. A well-designed flood will produce 80% to 90% of the oil
ahead of the surfactant. Even here, though, the oil is invariably produced early and at
fairly low cuts in laboratory experiments.
Surfactant breaks through at tD = 0.6, reaches its maximum produced
concentration of 30% of the injected concentration at tD = 0.8, and ceases at tD = 1.5.
The total amount of surfactant produced is about one half that injected, which
indicates substantial, though not excessive, retention.
The surfactant is preceding both the chloride and polymer by about 0.3 Vp.
This separation indicates preferential partitioning of the cosurfactant between the
aqueous and microemulsion phases (see Sec. 9-8 on phase behavior nonidealities).
Though this did not drastically affect oil recovery, which is in excess of 90% of the
residual oil, the separation is not a favorable indication for this design. A good MP
design should show simultaneous production of all MP slug constituents as well as
good oil recovery.
406
Field Response
As a field example, consider Fig. 9-32, which shows the produced fluid analyses of
well 12-1 in the Bell Creek (Carter and Powder River counties, Montana) MP flood.
This flood used a high oil content MP slug preceded by a preflush that contained
sodium silicate to lessen surfactant retention and reduce divalent cation
concentration. Well 12-1 was a producer in the center of an unconfined single
five-spot pattern. (For further details on the flood, see Holm, 1982; Aho and Bush,
1982.)
Figure 9-32 Production response from Bell Creek Pilot (from Holm, 1982)
407
Before MP slug injection in February, 1979, well 12-1 was experiencing low
and declining oil cuts. Beginning in late 1980, MP oil response reversed the decline
and reached peak cuts of about 13% about six months later. The pre-MP decline must
be clearly established to accurately evaluate the MP oil recovery, an unnecessary step
in evaluating the core flood. Moreover, compared to the core flood, there is no
evident clean oil production; surfactant production actually preceded the oil response.
Simultaneous oil and surfactant production is a persistent feature of field MP floods
probably because of heterogeneities and dispersive mixing. The surfactant is
preceding the oil in Fig. 9-32 because of preferentially water-soluble disulfonate
components in the MP slug. The peak oil cut is invariably lower in field floods (13%
in Fig. 9-32 versus nearly 60% in Fig. 9-31).
Other significant features in Fig. 9-32 are the evident presence of the preflush
preceding the MP slug, inferred from the maxima in the pH and silicate
concentrations, and the very efficient removal of the calcium cations ahead of the
surfactant. But when oil production commenced, calcium rose roughly to its
premicellar level.
Figure 9-33 shows ultimate oil recovery efficiency ER (ultimate oil produced
divided by oil in place at start of MP process) from a survey of more than 40 MP
field tests correlated as a function of mobility buffer slug size tDmB. Similar analyses
on other process variables showed no or weak correlation (Lake and Pope, 1979).
The strong correlation in Fig. 9-33 indicates the importance of mobility control in
MP design. Though we have largely ignored mobility control in this chapter, it is
Figure 9-33 Recovery efficiencies from 21 MP field tests (adapted from Lake and
Pope, 1979)
408
clearly an important variable. Note from Fig. 9-33 that the high oil content slugs have
generally been driven by polymer drives larger than the high water content slugs.
Ultimate oil recovery efficiency averages about 30% of residual oil
saturation in field tests (Fig. 9-33). Since oil recovery efficiency can be quite high in
core floods, it seems that the peak oil cut and ultimate oil recovery efficiency in a
technically successful MP field flood will average about one third of their respective
values in core floods.
9-13 DESIGNING AN MP FLOOD
A successful MP flood must achieve three things for efficient oil recovery (Gilliland
and Conley, 1975).
1. The MP surfactant slug must propagate in an interfacially active mode (i.e.,

at optimal conditions).
2. Enough surfactant must be injected so that some of it is unretained by the
permeable media surfaces.
3. The active surfactant must sweep a large portion of the reservoir without
excessive dissipation because of dispersion or channeling.
The first of these objectives is met through the formulation step of the MP design
procedure; the second two objectives are met through scale up. Though there is
considerable overlap, the formulation step consists mainly of test tube experiments
and core floods; the scale-up steps consist mainly of core floods and numerical
simulations.
Generating Optimal Conditions
There are three techniques for generating optimal conditions in MP floods.
1. Raise the MP slug optimal salinity to that of the resident brine salinity in the
candidate reservoir. Philosophically, this procedure is the most satisfying of
the three possibilities, and it is usually the most difficult. Though the subject
of intensive research, surfactants having high optimal salinities that are not,
at the same time, unstable at reservoir conditions, excessively retained by the
solid surfaces, or expensive are yet to be discovered. Field successes with
synthetic surfactants have demonstrated the technical feasibility of this
approach (Bragg et al., 1982). A second way to make the optimal salinity of
the MP slug closer to the resident brine salinity is to add cosurfactant. This
approach is the most common implementation to date; however, as we
409
mentioned, there are penalties in surfactant–cosurfactant separation, loss of
interfacial activity and expense.
2. Lower the resident salinity of a candidate reservoir to match the MP slug’s
optimal salinity. This common approach is the main purpose of the preflush
step illustrated in Fig. 9-1. A successful preflush is appealing because, with
the resident salinity lowered, the MP slug would displace oil wherever it goes
in the reservoir, and retention would also be low. Preflushes generally require
large volumes to significantly lower the resident salinity owing to mixing
effects and cation exchange (see Sec. 9-11). With some planning, the
function of preflush could be accomplished during the waterflood preceding
the MP flood.
3. Use the most recent salinity gradient design technique for generating active
MP slugs (Paul and Froning, 1973; Nelson and Pope, 1978; Hirasaki et al.,
1983). This technique tries to dynamically lower the resident salinity to
optimal during the course of the displacement by sandwiching the MP slug
between the overoptimal resident brine and an underoptimal mobility buffer
salinity. Table 9-4 illustrates the results of experimental core floods for
different sequences of salinities. The experiment numbers on this table match
the uncircled numbers in Fig. 9-8. Three core floods––numbers 3, 6, and 7––
stand out both with respect to their low ultimate saturation and surfactant
retention. The common feature of all these experiments is that the salinity of
the polymer drive is underoptimal. In fact, no other variable, including,
paradoxically, surfactant slug concentration, has such a similarly strong
effect (Pope et al., 1982). The salinity gradient design has several other
advantages: it is resilient to design and process uncertainties, provides a
favorable environment for the polymer in the mobility buffer, minimizes
retention, and is indifferent to the surfactant dilution effect.
TABLE 9-4 PHASE-ENVIRONMENT TYPE AND MP FLOOD PERFORMANCE FOR THE

SALINITY-REQUIREMENT DIAGRAM IN FIG. 9-8 (FROM NELSON, 1982)
Phase type promoted by the Residual oil

Chemical saturation after Injected surfactant
flood Waterflood Chemical Polymer chemical flood retained by the core
number brine slug drive (% PV) (%)
1 II(–) II(–) II(–) 29.1* 52

2 II(+)/III II(+)/III II(+)/III 25.2* 100*
3 II(+)/III II(+)/III II(–) 2.0† 61*
4 II(–) II(–) II(+)/III 17.6* 100*
5 II(–) II(+)/III II(+)/III 25.0 100
6 II(+)/III II(–) II(–) 5.6† 59†
7 II(–) II(+)/III II(–) 7.9* 73*
8 II(+)/III II(–) II(+)/III 13.7† 100*
*
Average of duplicates
†
Average of triplicates
410
Injecting Enough Surfactant
The first aspect of overcoming retention is to design the flood so that retention is as
low as possible- This includes minimizing the chemical and physical adsorption
effects discussed above and eliminating phase trapping by propagating the slug in a
low-salinity environment. Cosurfactants and sacrificial agents in a preflush may also
be appropriate. Once a low surfactant retention value is in hand, enough surfactant
must be injected so that some of it transports to the production wells. As in polymer
flooding, there are two aspects to this issue: the slug’s surfactant concentration and
the slug size
Strong theoretical or practical reasons for selecting the slug surfactant
concentration do not exist. The concentration must be large enough so that a type III
region can form when the salinity is optimal but small enough so that the slug can be
easily handled and transported. The latter requirement usually means the slug is
single-phase and not excessively viscous and the surfactant does not precipitate.
Perhaps a more stringent lower bound on surfactant concentration is in its
relative rate of propagation. The frontal advance loss of D3 contains surfactant
concentration in the denominator. This means the rate of slug propagation, as well as
the maximum oil cut calculated from fractional flow theory (Fig. 9-24), decreases as
concentration decreases. Because of the worth of the oil, the resulting delay in oil
production is a liability to the process even if the ultimate oil recovery were
unaffected. This argument suggests the concentration should be as large as possible,
and the slug size should be correspondingly small. But extremely small slugs would
seem to be sensitive to dispersive mixing in the reservoir.
Once the slug concentration is set, the slug size follows from the value of D3,
as in Sec. 9-11. To satisfy retention, the slug size, based on floodable pore volume,
must be somewhat larger than retention. Of course, how much larger is a strong
function of the prevailing economics and reservoir characteristics. (For a graphical
procedure, see Jones, 1972.)
Maintaining Good Volumetric Sweep
Figure 9-33 attests that the importance of this issue, particularly with respect to the
mobility buffer, cannot be overstated.
The mobility control agent in the slug can be polymer or oil as in Fig. 9-13.
Whatever the agent, it is of paramount importance that the slug–oil bank front be
made viscously stable since small slugs cannot tolerate even a small amount of
fingering. Thus we seek a slug less mobile than the oil bank it is to displace. To
provide a margin of safety in estimating the oil bank mobility, use the minimum in
the total relative mobility curves (see Sec. 3.3) to base the mobility control on. Such
curves (Fig. 9-34) show that the minimum can be substantially less than the total
relative mobility of either endpoint. Since these curves are subject to hysteresis, it is
important that the relative permeability curves be measured in the direction of
increasing oil saturation for tertiary floods.
411
Figure 9-34 Total relative mobilities for samples of the same reservoir (from
Gogarty et al., 1970)
Sizing the mobility buffer proceeds like the polymer drive we discussed in
Sec. 8-5. Here the spike portion of the buffer must have mobility equal to or less than
the slug mobility. Since the latter depends on the degree of oil desaturation, the
buffer mobility cannot be designed independently of the slug.
9-14 MAKING A SIMPLIFIED RECOVERY PREDICTION
In the next few paragraphs, we describe a simple procedure to estimate oil recovery
and oil rate–time curves for an interfacially active MP process. Since interfacial
activity may be lost in innumerable ways, the procedure will be most accurate for
processes that clearly satisfy the first design goal in Sec. 9-13. The procedure has two
steps: estimating the recovery efficiency of an MP flood and then proportioning this
recovery according to injectivity and fractional flow to give an oil rate–time curve.
(For further details of the procedure, see Paul et al., 1982.)
Recovery Efficiency
The recovery efficiency ER of a tertiary (S2I = S2R) MP flood is the product of a

volumetric sweep efficiency EV, a displacement efficiency ED, and a mobility buffer
efficiency EMB
412
ER = ED EV EMB (9.14-1)
Each quantity must be independently calculated.
Displacement efficiency. The displacement efficiency of an MP flood is the

ultimate (time-independent) volume of oil displaced divided by the volume of oil
contacted
S′
ED = 1 − 2 r (9.14-2)
S2 r
where S2r ′ and S2r are the residual oil saturation to an MP and a waterflood,
respectively. S2r must be known, but S 2r′ can be obtained from a large slug (free from
′ indicate
the effects of surfactant retention) laboratory core flood. Low values of S 2r
successful attainment of good interfacial activity in the MP slug. If core flood results
′ may be estimated from a CDC using a “field” capillary number
are not available, S2r
(Lake and Pope, 1978) based on the median velocity in a confined five-spot pattern.
0.565qσ
N vc = (dimensionless) (9.14-3)
H t Ap
Here, q is the volumetric injection rate and Ap is the pattern area. For approximate
calculation, assume σ = 1 μN/m in Eq. (9.14-3). The CDC chosen to estimate S 2r ′
should be consistent, as much as possible, with conditions of the candidate reservoir.
Volumetric sweep efficiency. Volumetric sweep efficiency EV is the

volume of oil contacted divided by the volume of target oil. EV is a function of MP
slug size tDs, retention D3, and heterogeneity based on the Dykstra-Parsons coefficient
VDP.
Consider the layered medium in Fig. 9-35 into which is injected an MP slug
of size tDs. If the flow is apportioned by kh, and there is no crossflow, the slug size in
layer l is
⎛ k ⎞ ⎛φ ⎞
t Dsl = t Ds ⎜ ⎟ ⎜ ⎟ = t Ds Fl ′ (9.14-4)
⎝ φ ⎠l ⎝ k ⎠
where Eq. (9.14-4) has introduced the derivative of the flow-capacity–sorage-

capacity curve (F-C curve) first discussed in Sec. 6-3. Besides invoking a continuous
permeability–porosity distribution, using F-C curves implies the layers are arranged
as decreasing (k/φ). If t Dsm > D3 in a particular layer m, it will be completely swept;
otherwise, the layer’s volumetric sweep will be in proportion to t Dsm / D3
⎧ 1, t Dsm > D3
⎪
EVm = ⎨ t Dsm (9.14-5)
⎪ , t Dsm ≤ D3
⎩ D3
413
Figure 9-35 Schematic representation of MP slug sweep in a layered medium
Figure 9-35 illustrates this division. EVm summed over n = 1, . . . , m, . . . , NL layer

after being weighted by (φh)m gives
t
EV = Cm + Ds (1 − Fm ) (9.14-6)
D3
To calculate EV with tDs, D3, and F-C curve known, begin by finding the layer number
m where t Dsm = D3 . This determines the coordinates (F, C)m in Eq. (9.14-6) to
determine EV. Equations (6.3-11) and (6.3-12) establish a relation between the F-C
curves, the heterogeneity factor HK, and VDP for a lognormal continuous permeability
distribution; thus EV in Eq. (9.14-6) can be related directly to VDP. Figure 9-36 shows
this relationship. VDP may be estimated from geologic study, matching the prior
waterflood, or core data (see Table 6-1). The D3 is from Eq. (9.10-6).
Mobility buffer efficiency. The mobility buffer efficiency EMB is a function

of EV and VDP
⎡ ⎛ −0.4t DMB ⎞ ⎤
EMB = (1 − EMBe ) ⎢1 − exp ⎜ 1.2 ⎟ ⎥ + EMBe (9.14-7a)
⎢⎣ ⎝ EV ⎠ ⎥⎦
and
EMBe = 0.71 − 0.6VDP (9.14-7b)
where EMBe is the mobility buffer efficiency extrapolated to tDMB = 0, and tDMB is the
mobility buffer volume, fraction Vpf. Equation (9.14-7) was obtained by numerical
simulation.
414
Figure 9-36 Effect of slug size–retention ratio on vertical sweep efficiency (from
Paul et al., 1982)
The recovery efficiency ER now follows from Eq. (9.14-1), which may be
checked for reasonableness against Fig. 9-33.
Calculation of an Oil-Rate–Time Plot
The production function (oil rate q2 versus time) is based on ER and the following
procedure. We assume the dimensionless production function is triangular with oil
production beginning when the oil bank arrives. From here, q2 increases linearly to a
peak (maximum) oil cut when the surfactant breaks through and then decreases
linearly to the sweep-out time. The triangular shape is imposed by the reservoir
heterogeneity.
The first step is to calculate the dimensionless oil bank and surfactant
breakthrough times for a homogeneous flood
⎛ S2 − S2 I ⎞
t DB = ⎜ B ⎟t (9.14-8a)
⎜ f 2 − f 2 I ⎟ Ds
⎝ B ⎠
t Ds = 1 + D3 − S2′r (9.14-8b)
where t DB is the dimensionless oil bank arrival time, and tDs is the surfactant arrival
time. S 2B and f 2B may be estimated from the simplified fractional flow theory (see
Sec. 9-10) or directly from laboratory experiments.
The second step is to correct these values for the heterogeneity of the
candidate reservoir using the heterogeneity factor HK defined in Eq. (6.3-11).
The corrected breakthrough times are now
tD
tˆDB = B (9.14-9a)
HK
415
t
tˆDs = Ds (9.14-9b)
HK
and the peak oil cut f2pk is

⎛ ⎛ t DB ⎞ ⎞
1/ 2
⎜ HK − HK ⎜ ⎟ ⎟
⎜ ⎝ t Ds ⎠ ⎟⎠
fˆ2 pk = ⎝ f 2B (9.14-10)
( H K − 1)
The symbol ˆ represents a quantity in a layered medium.

The final step is to convert the dimensionless production function to oil rate
q2 versus time t. This follows from
q2 = qfˆ2 (9.14-11a)
Figure 9-37 Comparison between predicted and observed oil-rate–time responses

for the Sloss micellar–polymer pilot (from Paul et al., 1982)
416
V pf tˆD
t= (9.14-11b)
q
Here fˆ2 and tˆD are any points on the triangular oil recovery curve that begins at
(tˆDB ,0), peaks at (tˆDs , fˆ2 pk ), and ends at (tˆDsw ,0). tˆDsw , the dimensionless time at
complete oil sweepout, is selected to make the area under the fˆ2 -tˆD curve equal to
ER,
2E S
tˆDsw = tˆDB + R 2 (9.14-12)
fˆ
2 pk
Figure 9-37 compares the results of this procedure with the Sloss MP pilot.
In terms of the number of design decisions required, micellar-polymer flooding is the

most complicated enhanced oil recovery process. This complexity, along with
reservoir heterogeneity and the need for a rather large capital investment, make
micellar-polymer flooding a high-risk process. Consequently, recent years have seen
a decline in interest in the process. The potential for the process is immense,
however, even slightly exceeding that of thermal methods, at least in the United
States. Moreover, both polymer and MP flooding seem uniquely suited for light-oil
reservoirs in isolated areas of the world.
Reservoirs amenable to micellar-polymer flooding contain light- to medium-
weight oils with moderate to high permeability. Since injectivity is essential in this
process as in polymer flooding, we seek reservoirs with depth sufficient to tolerate
high injection pressures but not so deep as to promote thermal degradation. Finally,
the process is sensitive to high brine salinities, although this can be dealt with
somewhat by suitable surfactant/polymer selection and design.
The important topics in this chapter deal with the association of interfacial
activity with brine salinity and hardness through phase behavior, the importance of
surfactant retention, and the need for good mobility control. In a sense, the design
criteria given in Sec. 9-13 apply to all EOR processes, but it is only in micellar-
polymer flooding that all criteria seem to apply with equal severity. Finally, the
screening estimation of recovery in Sec. 9-14 is a useful yet simple tool for assessing
the suitability of a reservoir and for estimating the risk associated with the process.
EXERCISES
9A. The Units of MP Flooding. A particular petroleum sulfonate surfactant has an average
molecular weight of 400 kg/kg-mole, a density of 1.1 g/cm3, and a monosulfonate-to-
disulfonate mole ratio of 4. Express the overall surfactant concentration of a 5 volume
percent aqueous solution in g/cm3, kg-moles/cm3, meq/cm3, mole fraction, and mass
fraction.
417
9B. Surfactant Equilibria and Aggregation. Relatively simple models can reveal much
about surfactant equilibria. The surfactant is a monosulfonate in this problem.
(a) The aggregation of surfactant monomers into micelles in a NaCl brine may be
represented by the following reaction:
N A (NA + +RSO3− ) (RSO3 Na) N A (9B-1)
where NA is the aggregation number. Using the definition for total surfactant
(monomer + micelles), derive an expression between total and monomer sulfonate
concentrations. If the equilibrium constant for Eq. (9B-1) is 1015 and NA = 10,
estimate the critical micelle concentration. The total sodium concentration is
10,000 g/m3.
(b) Consider a more complicated situation where 0.3175 kg-moles/m3 monosulfonate
surfactant solution is added to a NaCl brine. In a NaCl brine solution, five species
can form: surfactant monomer (RSO3− ), surfactant micelles [(RSO3 Na) N A ], free
sodium-surfactant (RSO3Na), precipitated sodium-surfactant (RSO3 Na ↓), and
free sodium (Na+). Calculate the concentration of each species when the overall
sodium concentration is 100 g/m3. Use the data in part (a) for the monomer–
micelle reaction, and take the equilibrium constant for the sodium-sulfonate
formation to 3 × 106 and the solubility product for the precipitate to be 10–8.
(c) Repeat the calculation of part (b) if the overall sodium concentration is 100,000
g/m3. What can you conclude about the effect of high salinities on surfactant
precipitation?
9C. Phase Ratios for Hand’s Rule. In Sec. 4.4, we saw that flash calculations for vapor–
liquid equilibria required using flash vaporization ratios or K-values. The analogous
quantities for Hand’s rule are phase ratios defined as
Cij
Rikj = (9C-1)
Ckj
for components i and k. Nearly all the flash calculation can be formulated in terms of
the phase ratios. Assume a type II(–) phase behavior (j = 2 or 3) in the following:
(a) Show that the Hand equations for the binodal curve (Eq. 4.4-23) and the
component distribution (Eq. 4.4-24) can be written as
R32j = AH ( R31j ) B , j = 2 or 3 (9C-2)

R = EH ( R )
2
32
3 F
31 (9C-3)
(b) We can interchange the roles of phase concentrations and phase ratios. Show that
the consistency relation ∑ 3i =1 Cij = 1 reduces to
−1
⎛ 3 ⎞
Cij = ⎜ ∑ Rkij ⎟ , j = 2 or 3 (9C-4)
⎝ k =1 ⎠
where Riij = 1.
There are 18 phase ratios in two-phase systems. But only 4 of these are independent
since
Rj
Rikj = ( Rkij ) −1 and Rikj = imj (9C-5)
Rkm
418
As is consistent with the phase rule, specifying any one of these will determine the
others and all phase concentrations through Eq. (9C-4). Solve for the phase
concentrations when R312 = 5. Take AH = 0.5, BH = –1.5, EH = 0.137, and FH = 0.65.
Note that the phase ratios for the microemulsion phase are the same as solubilization
parameters.
9D. Using the Hand Equations
(a) For the Hand parameters AH = 2, BH = –0.5, EH = 600, and FH = 2.3, plot the
binodal curve and at least two tie lines on triangular coordinates. Flash calculations
in two-phase regions require an additional constraint over those in Exercise 9C.
The constraint here is the equation of a tie line
C − Ci 3
S1 = i , i = 1, 2, or 3 (9D-1)
Ci1 − Ci 3
for type II(+) systems (where j = 1 replaced j = 2 in Eqs. [9C-2] and [9C-3]). Any
two of these may be used for a flash calculation, for example,
C1 − C13 C2 − C23
− = f (9D-1)
C11 − C23 C21 − C23
The flash consists of picking the correct phase ratio (see Exercise 9C) so that f = 0
with the Ci known.
(b) Calculate the compositions and amounts of each phase present if the overall
composition Ci is (0.45, 0.45, 0.1).
9E. Two-Phase Flash Calculation (Plait Point in Corner). In a type II(+) system with the
plait point in the brine corner, we have C11 = 1, and C21 = C31 = 0. The phase
distribution Eq. (9.7-9) now becomes superfluous, as does the binodal Eq. (9.7-5) for
the aqueous phase. The entire Hand representation collapses to
BH
C33 ⎛C ⎞
= AH ⎜ 33 ⎟ (9E-1)
C23 ⎝ C13 ⎠
(a) Show that the tie line equation for this special case reduces to
⎛C ⎞
C23 = C33 ⎜ 2 ⎟ (9E-2)
⎝ C3 ⎠
⎡ (1 − C1 ) ⎤
C13 = 1 − C33 ⎢ ⎥ (9E-3)
⎣ C3 ⎦
which express the microemulsion phase concentrations as ratios of each other.

(b) Show that Eqs. (9E-1) through (9E-3) may be used to solve explicitly for the
surfactant concentration in the microemulsion phase as
−1/ BH
1 1 − C1 ⎛ C3 ⎞
= +⎜ ⎟ (9E-4)
C33 C3 ⎝ AH C2 ⎠
(c) For an overall composition of Ci = (0.45, 0.45, 0.1), solve Eq. (9E-4) for the phase
composition and the saturation of the aqueous phase. Take AH = 2 and BH = –0.5.
419
(d) Compare the results of part (c) with the results of part (b) of Exercise 9D. What do
you conclude about approximating this phase behavior with the plait point in one
of the corners of the ternary?
9F. Equilibrium Calculations with Simplified Phase Behavior. Use the simplified Hand
representations with BH = –1 and FH = 1 in the following. Further, take the left and
right oil coordinates of the plait point to be 0.05 and 0.95, respectively; the low-,
optimal-, and high-salinity binodal curve heights to be 0.2, 0.1, and 0.2, respectively;
and the lower- and upper-effective salinity limits to be 0.06% and 1.4% NaCl. The
optimal salinity is at the midpoint between these two. Make all the calculations at a
salinity of 0.08% NaCl where the phase environment is type III.
(a) Calculate the Hand parameter AH and the coordinates of the two plait points and of
the invariant point.
(b) Plot the binodal curve and the three-phase region on a ternary diagram.
(c) Calculate the phase concentrations and saturations at an overall concentration of Ci
= (0.65, 0.3, 0.05).
(d) Repeat part (c) at an overall concentration of Ci = (0.44, 0.44, 0.12).
Plot both points on the diagram of part (b).
9G. Phase Behavior and IFT. Fig. 9G shows the bottom half of six surfactant–brine–oil
mixtures. These diagrams are on rectangular coordinates having a greatly expanded
vertical scale. CSe is the salinity in wt. % NaCl. In the following, the surfactant
concentration is 0.05 volume fraction:
(a) Calculate and plot volume fraction diagrams at brine–oil ratios of 0.2, 1.0, and 5.
(b) At a brine–oil ratio of 1, calculate and tabulate the solubilization parameters.
(c) Use the correlation in Fig. 9-9 to convert the solubilization parameters to
interfacial tensions. Plot these solubilization parameters against salinity, and
estimate the optimal salinity.
(d) Plot the IFTs in part (c) against salinity on semilog paper. Estimate the optimal
salinity based on IFT and the optimal IFT.
(e) Compare the optimal salinities in parts (c) and (d) to the midpoint salinity. The
latter is the salinity halfway between CSeu and CSe1.
9H. Fractional Flow Construction for Type II(–) Systems. Fig. 9H shows water fractional
flux curves for a type II(–) MP system for which all tie lines extend to the common
point Cio = (0. 1, 1. 1, 0).
(a) Calculate and plot an overall water concentration (C1) profile at oil bank
breakthrough and an effluent water flux (F1) for the following cases:
Injected composition (J) Initial composition (I)

Case C2 C3 C1 C2
1 0 0.10 0.66 0.34

2 0.97 0.03 0.66 0.34
3 0 0.10 0.20 0.80
For all cases, the displacement satisfies the fractional flow assumptions, the
surfactant is not retained by the permeable medium, and the surfactant injection is
420
Figure 9G Ternary diagrams at various salinities (from Engleson, 1981)

421
Figure 9H Water fractional flux for

Exercise 9H
continuous. All injected compositions lie on extensions of the tie lines whose
fractional flux curves are shown in the figure.
(b) On the water concentration profiles of part (a), sketch (no calculation necessary)
the microemulsion phase saturation S3 profile.
(c) On the water effluent histories of part (a), sketch the overall surfactant C3 effluent
history.
9I. Two-Phase II(–) Fractional Flow. Use the data in Figs. 9G and 9-9 in the following.
Take the oil-free injected slug concentration to be 0.05 volume fraction surfactant, and
the salinity to be constant at 0.56% NaCl. The surfactant is in an ideal mixture. The
low Nvc relative permeability curves are given by
S2r = 0.3, kr02 = 0.8, n2 = 1.5
S3r = 0.2, kr03 = 0.1, n2 = 3
Phase 3 is water when Nvc. The displacement occurs at a superficial velocity of 10

μm/s. The microemulsion, oil, and water viscosities are 2, 5, and 1 mPa-s. The medium
is horizontal. Use Fig. 3-19 as the capillary desaturation curve.
(a) Estimate and plot the relative permeability curves corresponding to the tie lines the
initial and injection conditions are on. Use the high Nvc relative permeabilities of
Eqs. (9.9-1) and (9.9-2).
(b) Estimate and plot the microemulsion ftactional flow curves along the two tie lines
in part (a).
(c) Plot the time–distance diagram and a composition profile at oil bank breakthrough
for this displacement if the injection is continuous surfactant. Use the simplified
fractional flow analysis of Eqs. (9.10-5) through (9.10-9). Take D3 = 0.1.
9J. Slugs and Simplified Fractional Flow. Use the simplified fractional flow of Eqs.
(9.10-5) through (9.10-9) in the following. The displacement is a constant II(–) phase
environment consisting of an oil-free surfactant slug followed by a polymer drive. The
water-, oil-, and microemulsion-phase viscosities are 1, 5, and 10 mPa-s, respectively,
and the relative permeability data at low and high Nvc are
422
Oleic phase Microemulsion phase
0
S2r k r2 n2 S3r kr03 n3
Low Nvc 0.3 0.8 1.5 0.2 0.1 5.0

High Nvc 0.05 0.9 1.2 0.1 0.6 2.5
(a) Estimate the polymer solution viscosity in the mobility buffer if the mobility ratio
between the slug and drive is to be 0.8. The polymer has no permeability reduction
effect.
(b) Calculate and plot the three aqueous-phase fractional flow curves (water–oil,
microemulsion–oil, polymer-solution–oil) based on the data in part (a) and the
polymer solution viscosity.
(c) Estimate the minimum slug size required to entirely sweep the one-dimensional
medium with slug. Take D3 = 0.2 and D4 = 0.1. There is no polymer in the slug.
(d) Calculate and plot the time–distance diagram if the slug size is one half that
estimated in part (c).
(e) Calculate and plot saturation profiles at tD = 0.3 and 0.8 for the conditions of part
(d).
9K. Fractional Flow with Oil-Soluble Slug. For this exercise, take the displacement to be
constant type II(+) phase environment (j = 1 or 3) with the plait point in the brine
corner. The surfactant is now dissolved in a predominantly oleic phase.
(a) Show that the surfactant-specific velocity is analogous to Eq. (9.10-5)
1 − f1s
vΔC3 = (9K-1)
1 − S1 + D3
and the oil bank saturation is given by the solution to

C f s − f1sB − C3 J 1 − f1s
vΔC2 = 2 J 1 = (9K-2)
C2 J S1 − S1B − C3 J 1 − S1 + D3
(b) Illustrate the graphical solution of Eq. (9K-2) on an aqueous-phase fractional flow
plot. What is the effect of injected oil concentration on the oil bank saturation?
Justify this observation on physical grounds.
(c) Figure 9K shows high- and low-Nvc fractional flow curves for a particular
displacement. Based on these curves, calculate and plot an oleic-phase saturation
profile at tD = 0.5. Take D3 = 0.1 and the surfactant injection to be continuous.
9L. Preflush Size Estimation. The composition of an initial reservoir brine and a possible
preflush solution are as follows:
Reservoir Preflush
Species brine (I), meq/cm3 (J), meq/cm3
Na+ 0.02 0.01

Ca2+ 0.06 0.005
Cl– 0.08 0.015
423
Figure 9K Aqueous-phase fractional

flow curves for Exercise 9K
The cation exchange capacity of the reservoir is Zv = 0.05 meq/cm3 of pore volume.
The cation exchange satisfies Eq. (9.11-3) with KN = 0.1. Assume single-phase flow of
an ideal solution that contains only the species explicitly stated above.
(a) Sketch this displacement in composition space as in Fig. 9-25a.
(b) Estimate the pore volumes of fluid J required to reduce the effluent calcium
concentration to the injected value. What percentages of the clays are in the
calcium form at this point?
(c) Calculate and plot the time–distance diagram for this displacement.
(d) State whether you think this would be an effective preflush for an MP flood.
9M. Importance of Mobility Control in MP Floods. In the absence of other data, high Nvc
relative permeabilities for a type II(–) system may be approximated by straight lines
through the points ( S3′r , 0) and (1 − S 2 r , kr03 )′ for the aqueous phase and through
( S3r , kr02 )′ and (1 − S 2′r , 0) for the oleic phase.
(a) Plot two high-Nvc fractional flow curves for aqueous-phase (j = 3) viscosities of 5
and 50 mPa-s. Take μ2 = 5, μ3 = 0.8 mPa-s, S3′r = 0.15, S 2′r = 0.05, (kr03 )′ = 0.8,
and (kr02 )′ = 0.6. The medium has no dip.
(b) Using the El Dorado relative permeabilities of Fig. 8L, illustrate the effects of
good mobility control on an MP flood by calculating oil saturation profiles for the
two cases in part (a) at tD = 0.3. The frontal advance lag D3 = 0.16. The injected
aqueous surfactant is continuous.
9N. Performance Prediction. Use the following information to perform a screening
estimation of oil recovery on an MP project. The water–oil relative permeability data in
Fig. 9K is appropriate.
(a) Estimate the swept zone oil displacement efficiency ED if the injection rate per
pattern is 65 m3/day. The pattern area is 8.1 hm2, and the formation thickness is 2
m. Take the IFT to be 1 μN/m, and use the CDC in Fig. 3-19 for the nonwetting
phase.
(b) Calculate the volumetric sweep efficiency EV. Take the Dykstra-Parsons
coefficient to be 0.5, the slug size to be 0.16, and D3 = 0.12.
(c) Estimate the recovery efficiency based on the above if the mobility buffer size is
0.8 PV.
(d) Calculate and plot the oil production rate versus time.

Enhanced Oil Recovery

Uploaded by

Copyright:

Available Formats

Enhanced Oil Recovery

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Enhanced Oil Recovery

Uploaded by

Copyright:

Available Formats

1

Defining Enhanced Oil

1-1 EOR INTRODUCTION

The EOR Target

We are interested in EOR because of the amount of oil to which it is potentially

1-2 THE NEED FOR EOR

Enhanced oil recovery is one of the technologies needed to maintain reserves.

The four categories of adding to reserves are

1. Discovering new fields

We discuss category 4 in the remainder of this text. Here we substantiate its

Enhanced oil recovery is actually a competitor with conventional oil

1-3 INCREMENTAL OIL

A universal technical measure of the success of an EOR project is the amount of

Figure 1-3. Incremental oil recovery from typical

As simple as the concept in Fig. 1-3 is, EOR is difficult to determine in

1. Combined (comingled) production from EOR and nonEOR wells. Such

Estimating Incremental Oil Recovery Through Decline Curves

Decline curve analysis can be applied to virtually any hydrocarbon production

where qi is the initial rate or q evaluated at t = 0. Equation 1.3-2 suggests a

Figure 1-3 schematically illustrates a set of data (points) which begin an

Because the model is a straight line, it can be extrapolated to some future

The rate-time analysis is useful, but the rate-cumulative curve is more

Figure 1-4. Schematic of exponential decline on a rate-cumulative plot.

Figure 1-5 shows a rate-cumulative plot for a project having an exponential

Figure 1-6. Schematic of exponential decline curve behavior on a rate-cumulative

a. Identify the pre- and post-EOR decline periods.

b. Calculate the decline rates ([=] mo-1) for both periods.

1-4 CATEGORY COMPARISONS

TABLE 1-1 CHEMICAL EOR PROCESSES

Polymer Improves volumetric Injectivity 5 0.3–0.5 lb polymer

TABLE 1-2 THERMAL EOR PROCESSES

Steam Reduces oil Depth 50–65 0.5 bbl oil consumed

TABLE 1-3 SOLVENT EOR METHODS

Immiscible Reduces oil Stability 5–15 10 Mscf solvent per

TABLE 1-4. SUMMARY OF SCREENING CRITERIA FOR EOR METHODS

1-5 UNITS AND NOTATION

Some common SI derived units

SI unit symbol Formula

Acceleration Meter per second squared –– m/s2

Selected conversion factors

TABLE 1-5 CONTINUED

Selected SI unit prefixes

1012 tera T One trillion times Billion

1. There are several cognates, quantities having the exact or approximate

The superscript ° is not used on the Kelvin scale.

4. The volume conversions are complicated by the interchangeable use of mass

0.159 m3 = 1 reservoir barrel, or bbl

0.159 std. m3 = 1 standard barrel, or stb

1. Clear all unit prefixes.

Example 1-2. Converting from Darcy units.

⎛ q − bbl ⎞ ⎧⎪ ( 24 )( 3600 ) ( 30.482 ) ⎫⎪ ( k − md ) ( A − ft 2 ) ⎛ dp − psia ⎞

The diversity of EOR makes it possible to assign symbols to components without

Date Oil Rate std. m3/day

(a) oil-field units.

(a) oil-field units.

2-1 MASS CONSERVATION

This section describes the conceptual nature of multiphase, multicomponent flows

The Continuum Assumption

Transport of chemical components in multiple homogeneous phases occurs because

As the cube volume increases sufficiently, the porosity approaches a constant