You are on page 1of 32

SPE 134589

Diffusion in Naturally Fractured Reservoirs - A Review


Mudit Chordia, University of Alberta and Japan J. Trivedi, University of Alberta

Copyright 2010, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Asia Pacific Oil & Gas Conference and Exhibition held in Brisbane, Queensland, Australia, 18–20 October 2010.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been reviewed
by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or
members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is
restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract

The impetus for this work stemmed from the growing demand in the industry for Carbon Dioxide sequestration during
incremental oil recovery and gas or solvent injection for heavy oil recovery from naturally fractured reservoirs. A
thorough research and assimilation of available field data led us to believe that a better understanding of the diffusion
coefficients and their impact on the overall recovery will go a long way in providing viable techniques in the Improved
Oil Recovery sector. This knowledge will further enhance the understanding of the mass transfer process and
mechanisms at a molecular scale in the reservoirs.

Remarkable amount of experimental and computational work has been carried out to study the effect of diffusion and
dispersion mechanisms in naturally fractured reservoirs. However, what's missing is an extensive and critical review of
the efforts in this direction. The synchronization of similar efforts and recognition of all the independent studies in a
common direction would not only highlight the deficiencies and limitations of present scenario in the field, it will also
give meaningful insight into further developments.

We have investigated methods employed by various authors to arrive at concurrent values for diffusion and dispersion
coefficients. The main focus during our work was to, i) evaluate the role of diffusion and the factors that critically affect
diffusion and oil recovery, ii) investigate current methods for calculation of the coefficients and iii) various
methods/models to incorporate diffusion. Our work includes significant technological contributions and findings of
experimental and simulation studies in the field to this day.

It is our understanding that our research will bridge the gap between the fundamental understanding and application of
related processes and serve as a useful reference for engineers and researchers for present applications as well as the
future.

Introduction

In many situations, the movement of a fluid in the interstitial porous network of a material is controlled by two main
parameters such as permeability and porosity. A large amount of literature available describes techniques allowing
measurement of these quantities for applications such as the optimization of oil production or storage of gas in rock
formations. However, a number of problems concern the migration of species over very long periods without any
2 SPE 134589

significant pressure gradient, or in a low permeability material and under the influence of a gradient in chemical
composition. Under these conditions, the relevant phenomenon is molecular diffusion and the parameters to be measured
are the diffusion coefficient(s) and the accessible porosity. During injection of fluids for enhanced oil recovery, oil
recovery and transport of injectants are controlled by fracture and matrix properties in naturally fractured reservoirs. For
these reservoirs, the transfer between matrix and fracture due to diffusion constitutes a major part of recovery. The
efficiency of mixing or dissolving of injected gas or solvent with oil phase is characterized by mutual diffusion coefficient
or effective diffusion coefficient for matrix. Diffusion is also critical for CO2 sequestration into depleted oil and gas
reservoirs. With optimization of cross flow and mass transfer between matrix and fracture, significant amount of oil can
be recovered in secondary and tertiary recovery processes and greenhouse gas storage could be maximized in these
complex and challenging reservoirs. Keeping this in mind our primary focus has been to understand the factors affecting
diffusion and the methods of determining the diffusion coefficients.

Governing Mechanisms in matrix-fracture system

Holm et al. (1974) cited solution gas drive, immiscible gas drive, hydrocarbon CO2 miscible drive, hydrocarbon
vaporisation, direct miscible CO2 drive and multiple contact dynamic drive as various mechanisms by means of which
CO2 displaces oil from porous media. Upon contacting the reservoir oil CO2 promotes oil recovery by, promoting
swelling, reducing viscosity, increasing oil density, improved solubility in water, exerting acidic effect on rock,
vaporising and extracting various portions of crude oil and transporting chromatically through porous rock. In this
section we discuss all these possible governing mechanisms during gas/liquid injection into fractured reservoirs and their
effect on oil recovery. A large part of the research done in the industry towards evaluating techniques for recovering oil
more effectively from reservoirs is done by means of experimental methods. Hence, a large number of researchers
simulate reservoir conditions in their labs and study the mechanisms in place that effect ultimate oil recovery. The work
done by Perkins et al. (1963) is widely reviewed by many researchers. It provides a review of dispersion on miscible
displacement processes and dispersion and diffusion phenomena in porous rocks.

Gravity Drainage and Diffusion

Typically, when the matrix is surrounded by gas (flowing) in the fracture, oil drains from the matrix as a result of density
difference between gas in fracture and oil in matrix – called a gravity drainage mechanism or GOGD. The process
depends on size and permeability of matrix blocks, type of gas and oil, temperature and pressure condition, size fracture,
and rate of gas flowing in the fracture. Depending on which parameter influences the process, diffusion plays a role. For
instance, in small and low permeable matrix blocks gravity drainage is inefficient and diffusion dominates. Gas injection
in fractured reservoirs could increase oil recovery from the matrix trapped oil in both equilibrium and non equilibrium
conditions. In such case it's the gravity drainage that would help recover the oil (Karimaie et al., 2007). Hence, gravity
drainage is a critical process taking place in the fractured reservoirs as it played a major role in the oil recovery from low
permeability matrix blocks during gas injection process. Gravity drainage displacement was also studied by Dastyari et
al. (2005) for fractured and unfractured micro models. The drainage process was found to be fast initially and slow at
longer times. Oil recovery is found to be greater in a fractured system as compared to an unfractured one. The recovery
increases as the gravitational forces increase. However, it is unlikely that diffusion plays a critical role in the adverse
effects of large scale bypassing due to gravity segregation (Grogan and Pinczewski, 1987).

A matrix block surrounded by gas will undergo a gravity drainage process when the gravitational forces exceed the
capillary forces. However the efficiency of this process depends on the threshold height and the matrix block size. For
CO2 injection the ultimate oil recovery is approximately the same for all stack heights (Alavian et al., 2005). The
capillary entry height in CO2 injection is lower than the capillary entry height of CO2 equilibrium gas injection due to low
IFT in a developed condensing/vaporizing near miscible process. Increase in pressure would still improve oil recovery,
even in the case of a non equilibrium gas (Øyno, 1995; Thomas, 1991; Darvish, 2006). An increase in pressure causes an
increase in solubility of CO2 in oil. CO2 concentration at a given fracture location in the reservoir depends on many
SPE 134589 3

factors such as gravity drainage segregation, gas diffusion, gas injection rate, matrix block permeability, gas cap size and
location of the injector perforations. Fracture gas compositions surrounding matrix blocks may vary substantially, with
the associated impact on oil recovery (Alavian et al., 2005). Increase in pressure as a result of increased injection rate can
affect the viscous flow and also increase the density of CO2, thereby leading to less gravity domination in the process
(Mahmoud et al., 2008). In contrast to this, high solubility of CO2 reduces the IFT and improves the microscopic sweep
efficiency.

Morel et al. (1990) stated that initial gas saturation has little effect on oil recovery and the process is not governed purely
by diffusion. The changes in the interfacial tension create a capillary equilibrium at lower saturation near fractures and
hence cause liquid accumulation near the fractures. Le Romancer et al. (1994) performed similar experiments to
Karimaie et al. (2007) and found that water saturation on oil recovery depended strongly on the nature of the diffusing
gas. The importance of diffusion calculation was shown by Hu et al. (1991) who simulated Morel et al's. (1990)
experiment and showed the correction of the capillary pressure curve for changes in interfacial tension that were due to
compositional variation. Dunn et al. (1989) developed a mathematical model for the gravity drainage process using low
temperature soluble gas as an extension of the theory for the thermally assisted gravity drainage process. The measured
drainage rates were higher than those predicted by the model using molecular diffusivity data but fit the model better
using an increased effective diffusivity.

Karimaie et al. (2007) stated that gas could not directly contact with the oil in the matrix and the diffusion process had to
occur by means of an intermediate step. This was observed when CO2/N2 first diffused into water phase and then the gas
dissolved in water diffused into oil, which in turn reduced the oil recovery. In further work, Karimaie and Torsæter
(2008) investigated the gas injection in fractured reservoirs in both secondary and tertiary injection with focus on gravity
drainage using equilibrium gas followed by re-pressurization. It was observed that oil recovery showed a significant
increase as a result of re-pressurization in gravity drainage. It was also seen that low IFT gravity drainage was capable of
recovering significant amount of oil in fractured reservoirs after water injection.

There have been various studies showing the effect of gas gravity drainage during gas injection into fractured porous
media. All yielding to the conclusion that gas drainage could significantly enhance oil recovery as a secondary or tertiary
drive mechanism (Li et al. (2000), Schechter et al. (1998), Schechter et al. (1996), Li et al. (1999)). The presence of
fracture improves the efficiency in the case of CO2 injection. Further, the efficiencies of the CO2 gravity drainage
decrease as the rock permeability decreases and the initial water saturation increases (Li et al., 2000).

Extraction Mechanism and Influence of Diffusion

Li et al. (1999) noticed an increased concentration of C11-C20 components during CO2-gravity drainage EOR process
indicating extraction mechanism by CO2. However, there have not been any significant findings in the direction of how
diffusion influences the extraction in naturally fractured reservoirs. It is believed that the constant rate production of oil
without a change in composition is due to displacement and drainage. Any form of selective extraction of components
(variation of produced oil composition) from the original oil in place is due to diffusion of gas into oil. Holm et al. (1974)
found that CO2 as a flooding agent can extract heavier components (C6-C30) and also can be miscible with crude oil that
has little C2-C6 components (lighter HC) unlike the lean gases. Though their finding was in unfractured porous media, a
similar extraction phenomenon in fractured porous media has also been reported by Darvish et al. (2006a) and Trivedi and
Babadagli (2008a) recently. Both studies mentioned here were on different types of rocks. During CO2 injection into
fractures, lighter components having lower diffusion coefficients are produced at early stage and heavier components with
higher diffusion coefficient are produced at much later stage. Molecular diffusion of inert gas favours displacement of
hydrocarbons more than dry gas does because of the concentration differences.

Darvish et al. (2006a) investigated the efficiency of the tertiary CO2 injection into fractured core with focus on the mass
transfer of the CO2 and hydrocarbons between the fracture and the chalk matrix. Mass fractions of light and heavy
components were produced during the oil recovery process from the chalk core. Initially, the produced gas was enriched
4 SPE 134589

with methane and at the later stage it contained more intermediate components and then towards the end heavy
components were recovered. Darvish et al. (2006a) did not however explain clearly the cause of this phenomenon.
Diffusion, oil swelling and gravity drainage were cited as governing mechanisms. In subsequent work Darvish et al.
(2006b) stated that density difference dominates the drainage of oil from a matrix in case of tall and permeable oil
saturated matrix blocks surrounded by gas in the fracture. However, this mechanism is not as dominant in case of low
permeability and small size matrix blocks with high capillary pressure and it’s the diffusion that plays the governing role.
Also of significance here is the usage of dead oil instead of live oil in the experiments. This is due to the dangers caused
by the volatility of the components in the live oil and the large permeability contrast between matrix and fracture normal
which would cause difficulties in normal core flooding and effective saturation of the pore system. Pressure stabilizations
over short durations also do not guarantee effective diffusion in the matrix. Hence, care has to be taken while performing
experimental work that matrix is saturated with the oil effectively.

Effect of miscibility & Injected gas phase condition

It is a proven phenomena that the amount of extracted hydro-carbon increases with pressure and decreases with
temperature (Al-Marzouqi et al., 2007). This may lead to the understanding of how miscibility condition may affect the
extraction which is a consequence (result / measure) of diffusion. As reported by Wang et al. (1986) liquid CO2 has best
extractability compared with CO2 in gases or supercritical state. These were the unfractured porous media experimental
observations; similar conclusions could be drawn for fractured reservoirs also as the gas phase condition is same in both
gases; though the miscibility development could be different. Verlaan et al. (2006) stated that if the injected gas is
miscible with oil, then the density and viscosity reduction will occur. Miscibility adds the advantages of single phase
flow and interfacial tension reduction, which enhanced the Gas-oil gravity (GOGD) rates and ultimate recovery. In their
work Verlaan et al. (2006) concluded that ultimate recovery could be enhanced by injection of the gas at fully miscible
conditions. The mechanism in play are, mixing and IFT reduction. Mixing causes oil viscosity reduction, reduction of oil
density in the matrix and increase in the gas density in the fractures. Whereas, IFT reduction causes, increase in oil
relative permeability, no capillary hold-up and no capillary re-imbibition. It was observed in the mixing zone between the
gas and the oil that the drainage rates are higher than in the oil zone and the excess is flowing towards fracture. Hence,
the viscosity reduction effect on the drainage rate is larger than the density reduction. The physical mechanisms for
mixing in the matrix are molecular diffusion and convective dispersion. The mixing zone increases with time which
explains why gravity drainage rate increases with time. Miscible gas injection exhibits no re-imbibition and then profits
from the layers that are draining simultaneously. Reimbibition is the differentiating factor between miscible and
immiscible GOGD. The absence of reimbibition is due to the instantaneous mixing of oil flowing from the matrix to the
fracture with the gas in the fracture.

Asghari et al. (2007) proposed the idea of injecting CO2 at high pressures (close to maximum miscibility pressure) and
soaking for a certain period to enhance oil recovery. They used cores of varying permeability and porosity and then
injected CO2 at incremental pressures of about 250psi to study the effects of matrix permeability and pressures. The
governing factors cited by the authors towards the recovery of oil were oil swelling, hydrocarbon extraction, viscosity
reduction and relative permeability effects. Based on the final recovery of oil, the authors concluded that pressures much
higher than the miscibility pressure are not as beneficial in recovery factor and need economic analysis. They further
concluded that the matrix permeability was inconsequential in the recovery factor of CO2 huff and puff process at
miscible conditions. This contradicts with the findings of Trivedi and Babadagli (2008c) where they derived the scaling
groups for miscible injection into naturally fractured reservoirs and matrix permeability was a contributing factor. The
authors fall short in addressing the recovery mechanisms by not addressing the presence of diffusion occurring in the
reservoirs. The experiments were carried out at room temperature, which is a considerable deviation, considering the
natural reservoir temperatures are expected to be much higher. Hence the MMP pressure might be incorrect and the
behaviour of the injected gas at higher pressures also is left unaccounted for. For these reasons the value of the injection
rate is critical. It might not be possible to avoid forced displacement if the core is small in length and hence the
conclusions might be slightly incorrect. One can say that miscibility i.e. increase of pressure, is somewhat related to
SPE 134589 5

diffusion and hence the recovery. The experiments of Asghari and Torabi (2007) on fractured cores with focus to study
the effect of miscible condition using Huff-n-Puff CO2 flooding does not tell the direct effect of diffusion on miscibility
and recovery because their results are not supported with produced oil compositional analysis. Minimal increase of
recovery beyond miscibility suggests diffusion may not play a major role after certain pressure and that is around miscible
pressure. Again, this needs to be supported with composition analysis of produced oil at different pressure stages. In a
binary system (single component oil) increasing pressure above MMP results into faster recovery while same is not true
for multi-component systems (Vahapcan et al., 2010; Asghari & Torabi, 2007; Trivedi & Babadagli, 2008b).

Under reservoir condition, MMP varies with the type of injection gas and its composition injected. The gas type and
composition surrounding the matrix may vary during the process as a result of dilution with gas-cap equilibrium gas (if
present) or recycling of produced gas stream.

A drop in gas-oil IFT causes a greater capillary gradient. This coupled with small relative phase density differences
induces a Darcy flow of oil against gravity. As the oil flows upwards it comes into contact with fresher and leaner gas
and the intermediate and heavy components are vaporized. These components are transported to the fracture system by
density difference and the molecular diffusion due to compositional difference between the matrix gas and in the
extracted components. Increasing the system pressure at this point makes the displacement process more miscible.

Common practice of measuring diffusion and dispersion coefficients for porous media

The most prevalent technique to measure the diffusion and dispersion coefficients in the industry today is, by means of a
PVT cell. The work of researchers Hill and Lacey (1934), Reamer et al. (1956), Sigmund (1976) commonly referred to as
Direct methods involves compositional analysis which are both time consuming and expensive. Low pressures due to
equipment limitations and various approximations made for the models caused significant errors in determining the
diffusion coefficient values for multicomponent mixtures at high pressures. Most of the correlations are developed based
on binary diffusion coefficients of relatively light compounds at high pressures and hence extrapolation of this available
data induces inaccuracies in the model. In the paper presented by Riazi (1996) a constant volume constant temperature
PVT cell is used. Here a non equilibrium gas is brought in contact with a liquid at constant temperature in a constant
volume vessel. The system reaches its equilibrium state with the passage of time. The drawback of this technique is that
the diffusion process continues only till the liquid phase is saturated with the gas phase and hence the results of only the
first few days of the experiment are analysed. Moreover, the diffusion constant's dependence on pressure cannot be
established as pressure is constantly decaying. Grogan et al. (1988) stated that diffusion coefficients could be measured at
high pressures and compositional measurements would not be necessary. They used the Stokes-Einstein equation to
estimate the diffusivity of CO2 into water. Riazi (1996) considered a constant value of diffusion coefficient but accounted
for a thermodynamic correction factor from the compositions at the end of the preceding time step for the accuracy of
calculation. Before the experiment is carried out, the PVT behaviour of the initially pure components and final mixture in
equilibrium are determined using the Peng-Robinson Equation of State (EOS). The final equilibrium gas composition
was matched with the measured values to confirm the EOS used for phase equilibrium calculations. Hence, if the values
for diffusion coefficient are accurate in the model then the predicted rate of interface movement will be consistent with
the experimental results. This way the need for compositional measurement is avoided. It is thus clear that improvement
needs to be made to both the experiments and the data analysis in this application. By working directly with composition
profiles of diffusing species within a phase and ensuring that the physical properties of the fluids are accounted for,
consistent and reliable mutual diffusion coefficients can be obtained. Tsay and Mc Hugh (1992) used lasers to measure
concentration profiles. They also showed that the coefficient values obtained are in close agreement with those obtained
using other methods. Another challenge to the experimental setup is their opacity to visible light, large variations of
density with composition and high concentration of surface active materials. Conventional techniques such as refractive
index matching are not applicable. In this regard, X-ray transmission tomography is used to establish accurate
6 SPE 134589

composition profiles within a phase and the data analysis method that accounts explicitly for the variation of mutual
diffusion coefficient and density with composition is employed. It is important to note that all these methods and future
evolutions determines the diffusion coefficient of fluid pairs (mutual diffusion coefficient or binary diffusion coefficients)
in PVT type cells to further use in porous media application and not evaluate the coefficients directly in the porous media
(effective diffusion coefficient). Moreover for naturally fractured reservoirs application, the use of these values to
represent the diffusion in the matrix blocks (effective matrix diffusion coefficient) where the process could be governed
by convection in the fracture, the choice of using correct diffusion coefficient value is a challenge. In this section, we
discuss various direct and indirect techniques, pertaining assumptions, applications and their drawbacks.

Evolution of techniques to determine diffusion coefficient

The Direct methods as used by Hill and Lacey (1934), Reamer et al. (1956) Sigmund (1976) etc. use compositional
analysis to determine the concentration of the diffusing species along the length of the bitumen or crude sample with time.
A major drawback in determining the diffusion coefficients using the direct methods is time consumption and the
associated costs with the experimental setup. The limitations of the apparatus and measurement techniques are often an
impediment in simulating the reservoir conditions. Consequently, most of the experimental work done in the field using
direct methods is performed at room temperature and pressure. This is a significant assumption, considering the reservoir
conditions are far more severe. Most of the diffusion models in practice use the Fick's law of diffusion as their base. The
Fick's law works on the assumption that viscosity and density of diffusing fluids is equal and constant. However, this is
not the case when gas diffuses into heavy oil. Hence, these techniques seem to inherently fall short in their interpretation
and use of the Fick's model and consequently model errors are induced. In contrast to the direct techniques the methods
employed by most researchers today, do not involve measurement of the compositions and are categorised as Indirect
methods. In these methods a change in system parameter is correlated to the desired coefficient value or any other
measurable property. These parameters could be the rate of change of solution volume or movement of the gas-liquid
interface, rate of pressure drop in a confined cell or the rate of gas injection from the top to a cell in which the pressure
and volume are kept constant. A porous medium is composed of a solid matrix containing pores filled with some fluid. It
is assumed that the fluid contains one phase only (liquid or gas). The fluid may, however contain different components.
There is typically a main fluid in which dilute species migrate under the effect of concentration gradient.

CAT Scanning and Low field NMR

CAT scanning technique makes use of X-Rays to differentiate between contrasts in a measurable property in the scanned
sample. Wen et al. (2005) used this technique to distinguish between densities at different points in their sample. A CT
number, obtained by comparing the linear attenuation coefficient of a pixel with that of water, is assigned during the scan.
This number changes with the process and correlates to the property of the mixture. CT scanning is capable of providing
up to 3 dimensional images to provide information pertaining to physical properties (density, porosity, mineralogy,
heterogeneities) and fluids within the core (saturation and saturation files). Relative fractions of a fluid stream containing
oil and water can be determined by means of a single NMR measurement. The NMR technique traces the mobility of the
Hydrogen bearing molecules of both solvent and oil and it changes as heavy oil comes into contact with the crude sample.
These changes are easily detected through changes in the NMR relaxation characteristics of the solvent and oil and in turn
correlated to the mass flux and concentration changes. Similar technique is also used by Afsahi and Kantzas (2007) to
determine the diffusion using the changes in NMR measured parameters. In their results the diffusivity in higher
viscosity oil samples was much lower compared to lower viscosity samples. These results for the diffusion coefficient
were shown to be in agreement with that of Wen et al. (2005).

Wen et al. (2005) and Afsahi and Kantzas (2007) used the Fick's second law to mathematically model the flux and
consequently measure the concentration changes and determine an apparent diffusion coefficient. The experiments were
carried out at room pressure and a temperature of 30oC. This approximation is farfetched as reservoir conditions are
much higher in range. Also, since the experiments were conducted under isothermal conditions the temperature
SPE 134589 7

dependence of diffusion coefficients could not be observed and accounted for. Another factor by means of which the
technique used by Wen et al. (2005) falls short is the definition of infinite boundary conditions. Even though its
applicability is correct while simulating a reservoir but essentially in a closed system like one used the wall effects and
the finite limitations may apply dominantly. This will induce deviations in the model. The other aspect where the
methodology used by Wen et al. (2005) falls short is the inability in capturing a 3 dimensional model. Since the diffusion
occurs in all directions along the length of the core sample a better estimate would be if it were averaged based on
calculations in all directions. Further, the experiment was run for almost four weeks, but however the data for only the
first two days was accounted for, as the interface between solvent and the oil could only be maintained for a short
duration. This is a significant approximation as the distance traversed by the solvent front is unknown and thus the
effects of the boundary walls on the interface may not be judged properly. Also what is lacking is the uniformity in the
case of different solvents with varying molecular masses corresponding to difference in capacities of diffusion. Last but
not the least is the assumption of independence of diffusion coefficient on concentration. Since, there is no mechanical
stirring or mixing of fluids, the mixing between the phases occurs only as a result of the concentration gradients.
However for accurate evaluation, the diffusion coefficient needs to be considered as a part of the governing Fick's
equation.

D dependent on Concentration is defined as,

Where, C is the solvent concentration as a function of distance x and time t, with constant diffusion coefficient D.

The solution of the equation is expressed as,

1
2 2√

Based on the following initial and boundary conditions. The initial condition for t=0 is,

C = Co for x < 0

C = 0 for x > 0

The boundary condition for t > 0 is

C = 0 at x = ∞

During the NMR experiments, the concentration of solvent determined through NMR spectra change is the overall
concentration, C in the mixture area, which is function of time and diffusion coefficient. The correlation between
concentration C, time t, and D is described in the following,

, ,

, , ,
8 SPE 134589

At a given t,

, ,
,

D independent of Concentration is defined as,

As the diffusion coefficient is function of the diffusing substance, C, D varies from point to point along the diffusion path.
The resultant equation becomes,

The solution to the above equation is given by the Boltzmann transformation,

1
2

Where, C1 can be any concentration between 0 and Co and the integral is equal to zero at x = ∞ to satisfy the
boundary and conservation of mass conditions.

Magnetic Resonance Imaging

The conventional 2D tools used in the industry only provide information pertaining to vapour chamber and its
advancement. However, with the use of MRI useful information about the concentration gradients are also determined.
Gas diffusion into the bitumen reduces the viscosity enough for the gravitational forces to overcome the capillary forces.
The effect of connate water is also observed, however the relative effect of Asphaltene preparation is not quantified by
Fisher et al. (2000). The nonlinear square method is used to derive exact parametric information about the process in
question. However, this is only an approximate method and successive iterations do not guarantee the accuracy of the
results. The physical constraint in the model is the need of non metallic materials. Also, the sand sample used has to be
made free of all iron impurities found on the surface of the grains. Presence of such impurities will tend to affect the MRI
signal and thus cause distortion. Use of an MRI apparatus also limits the working space and choice of materials that can
be used. The small size of the MRO magnet restricts the researchers to a very small sand pack sample. Another
limitation caused by the small size of the sample is that the leading edge of the solvent will hit the edge of the image
before it hits the end of the sand pack. This may influence the readings from the sample and hence make the model
inaccurate. Concern is also with low relaxation time values since the average signal strength is very low. Hence to arrive
at a useable signal to noise ratio multiple averaging methods have to be used. This could mean a fair amount of waiting
time between experiments. Fisher et al. (2000) developed a mathematical model to extract parametric information from
the MRI collected data. A few assumptions were made by the researchers. Amongst them was setting of the diffusion
value to zero at a coordinate when oil drained away from that specific point. The second critical assumption is that the
horizontal or vertical selection of data is relatively unaffected by diffusion in the other direction. These assumptions are
not representative of the physical condition or the mechanisms in play in the reservoir and the inability of the MRI data
sheet and the mathematical model henceforth clearly indicates the inadequacy of the setup. Fisher et al. (2000) also
assumed that the point of the maximum concentration change is representative of the gas-oil interface. The formulation
used to describe Fick’s law however, has a slight problem when describing the flow at the edges, as the gas-oil interface
keeps on moving. Care has to be taken as the volume of oil in the chamber is constantly changing. Hence there is a need
to develop an equation that allows for moving interface and to take care of the signal strength to represent concentration.
SPE 134589 9

Modified pressure decay method

Etminan et al. (2009) proposed a modified pressure decay method, where a secondary supply cell is used to continuously
inject gas in order to maintain the pressure across the gas-liquid interface constant. The pressure decay is measured in the
supply cell, which makes the measurements easier and a simple analysis can be performed to determine the equilibrium
concentration and the diffusion coefficient. This makes the accurate determination of the amount of gas being added to
the cell or equivalently dissolved into the bitumen body, possible. Etminan et al. (2009) addressed the complexity in
modelling the physics of the interface where the pressure is declining and the nature of the mathematical solutions. Also
highlighted is the problem in the assumption of a constant equilibrium concentration which introduces significant errors
in the estimation of the diffusion coefficient. An infinite and a finite acting analytical solution are derived to fit the
boundary condition and further the diffusion coefficient and solubility is obtained using inverse techniques. The problem
with using the inverse techniques however is their infinitely non unique nature. Also, an incorrect estimation made in the
determination of the finite parameters gets amplified and reflects as errant data in the solutions. The pressure decay
technique allowed the measurement of diffusivity of the solvent gases close to their dew point pressures. The
mathematical formulations representing the effects of liquid swelling were left out of the calculations. Hence, the
swelling effect is not accounted for. Further as a blind cell is used the interface movement is inconvenient to track.
However this problem was negated to an extent by calculating the liquid volume change using solubility amount and ideal
mixing rule. The experiment was carried out at conditions not as severe as the ones experienced in normal reservoir
conditions and hence do not justify the coefficient values. The pressure supply cell and the diffusion cell are connected
by means of electronic valves. The connection is intermittent and hence there is a pressure build up every time gas goes
into the diffusion cell. Etminan et al. (2009) accounted for this by assuming, all the gas leaving the pressure cell dissolves
into liquid. The assumption of one way diffusion and non volatility of gas was made by comparison and analysis of the
experimental setup with other researchers. This assumption may further induce errors.

Dynamic Pendant Drop Shape Analysis

The DPDSA is a unique process used by Yang and Gu (2003) that correlates the interfacial tension reduction in a drop to
the oil swelling effect in heavy oil as solvent diffuses into it. This physical phenomenon is represented in theoretical form
by the Laplace equation of capillarity and the molecular diffusion process described by the mass diffusion equation. In
the experimental setup a see through windowed high pressure cell is filled with the test solvent at a desired pressure and
temperature and then a heavy oil sample is introduced slowly through a syringe delivery system to form a pendant drop
inside the pressure cell. This causes the shape and volume of the drop to change until equilibrium is reached. This
change in shape is captured by means of sequential images and digitised by applying computer aided digital image
processing techniques. To attain convergence between the experimentally observed and numerically predicted profiles an
objective function is constructed. Once the minimum objective function is achieved, the solvent diffusivity in heavy oil
and the mass transfer Biot number are determined. This technique is quick and allows mass transfer coefficient to be
determined at the solvent-heavy oil interface. The Laplace-Young equation used as the governing equation works on the
assumption that gravity forces are the only forces being experienced by the pendant drop, which is not the case in a
reservoir oil swell the researchers are attempting to simulate. Also, the equation works well for axis symmetric cases
only, which again is difficult to achieve. The influence of the shape and volume of the pendant drop on the diffusion
process is neglected. The introduction of the surfactant into the system is also not a simple process. Care has to be taken
that the surfactant spreads evenly on the drop surface. Also, the solvent has to be kept from entering the bulk. If this is
not achieved the accurate measurement of the number of molecules per surface area may not be possible. With the
volume of the drop constantly changing the even spread of the surfactant will be difficult to achieve. The profiles are
analysed under the assumption that surface tension is the same as the surface tension of a static droplet with the same
shape which again is a big approximation for a dynamic system such as this one. The boundary conditions used in the
mass transfer model are the equilibrium and non equilibrium boundary conditions. The equilibrium BC dictates that the
heavy oil interface remains saturated with the solvent at all times. This means there is no interfacial resistance to the
10 SPE 134589

mass transfer of the solvent across the interface. It is uncertain whether the equilibrium BC exists at the interface between
the heavy oil and the solvent phases given that the adjacent bulk oil phase undergoes a transient diffusion process.

Graphical Methods

Sheikha et al. (2005) used a pressure decay data to develop a mathematical model based on the Fick's second law and
mass balance for the diffusing gas and then derived graphical solutions to estimate the diffusivity coefficient. They relied
mainly on the setup used by Upreti et al. (2000) for their measurements. The experimental setup consisted of a pressure
cell placed in an isothermal water bath. A constant amount of gas filled up the top of the pressure cell while the lower
section was filled with a sample of the bitumen. The air is purged after the bitumen reaches the bath temperature.
Pressure is increased to the initial test pressure and then recorded as gas diffuses into bitumen. Factors that govern the
diffusion process are the gas solubility and diffusion coefficient. The gas solubility acts as the driving force for diffusion
and the diffusion coefficient controls the rate of mass transfer within the bitumen phase. Sheikha et al. (2005) used
certain assumptions in their work such as chemical inertness of the mixing components, isothermal process, constant
diffusivity value and compressibility factor, non volatility of bitumen, negligible swelling and thermodynamic
equilibrium at the gas-bitumen interface. The two models developed here are finite acting and infinite acting in nature.
The infinite acting model is used to evaluate the diffusion constant while the finite acting model is used to determine the
time beyond which the infinite model leads to large uncertainties in the estimated diffusion coefficient values. The
infinite acting model works under the assumption that the penetration depth of the system encompasses the total depth of
the system at all times. Before the penetration depth reaches the bottom of the cell it is assumed to be at infinity. This
section was developed for the period before the closed boundary at the bottom of the cell starts to influence the gas
diffusion in bitumen. The final relationship developed was a forward acting model. In the equation used by Sheikha et
al. (2005) it is suggested that pressure in the gas zone is a function of the square root of diffusion coefficient divided by
Henry's constant. It was observed that the early time data lacked sufficient information to determine the diffusion
coefficient and Henry's constant. The solution developed required the values of all variables including the diffusivity
coefficient. For this reason an inverse solution had to be developed. The problem with using the inverse solution
technique however is that they are ill-posed and several solutions can be achieved for them. In order to reduce the non
uniqueness, the aim is to integrate as much data as possible. A general approach to the solution of the inverse problem is
through a minimisation procedure. The parameters are determined by minimising an objective function that measures the
misfit between observed and calculated dynamic data, using static data as prior information. This does not guarantee an
accuracy in the final solution as the initial estimate could be made on any assumption and its coherence with the
governing equation may or may not be guaranteed. In the work by Sheikha et al. (2005) a couple of graphical techniques
were developed as they had the advantage in being able to segregate the data that are affected by processes that are not
included in the forward model. For example if the early time data was affected by pressure affects or late time data by the
closed boundary, it could be clearly excluded from calculations. Upon analysis it was concluded by the researchers that
the one graphical method could not be considered superior to the other. Since the pressure decay data collected did not
permit the determination of both the diffusivity and Henry's constant the runs would have to conduct for longer time
periods such that the effect of the closed boundary is seen clearly in the analytical solution. Henry's constant could also
be determined from the late time period when the effect of mass diffusion decreases as the gas concentration in the entire
bitumen sample approached the solubility limit.

Mass Transfer Models

Tharanivasan et al. (2004) used the standard pressure decay cell method developed by Riazi (1995) to define three unique
boundary conditions (BC). These are the equilibrium, quasi-equilibrium and the non equilibrium boundary conditions.

These boundary conditions are derived from the work done by various researchers previously. In their method the
analytical solution was subjected to each boundary condition in order to determine the concentration distribution of the
solvent in the heavy oil. The condition proposed by Renner (1988) suggested that the heavy oil at the interface is
SPE 134589 11

saturated with the solvent under the so called equilibrium pressure at all times or the equilibrium boundary condition.
The quasi-equilibrium BC states that the solvent concentration at the interface remains at the saturation concentration that
varies with the pressure of the solvent phase during the measurement (Upreti and Mehrotra, 2000, 2002). Finally the BC
applied by Civian and Rasmussen (2002, 2003) applied the non-equilibrium BC, which considers the interfacial resistance
to mass transfer across the interface.

In further work, Tharanivasan et al. (2006) accurately measured the pressure decay data for the heavy oil-carbon dioxide,
heavy oil-methane, and heavy oil-propane systems. The above three BCs were applied to these three heavy oil solvent
systems. For each BC, the diffusion coefficient was determined by finding the minimum objective function, which
represents the minimum average pressure difference between the theoretically calculated and experimentally measured
pressures at different times. The most suitable BC for each heavy oil-solvent system is found by comparing the minimum
objective functions for the three BCs. The history matching technique is used to determine the solvent diffusivity. It was
assumed that the diffusion coefficient remained constant and independent of the concentration throughout the experiment.
However, we know that as the solvent diffuses into the heavy oil, pressure of the system decays due to the change in the
concentration of the mixing species. Another assumption in their work was that of the chemical inertness of the mixing
species, which again is not exactly the case in a reservoir. The solvent phase is predicted by utilizing the solvent mass
concentration and the equation of state for real gas. Therefore an average solvent diffusivity is determined by matching
the numerically predicted pressures with the experimentally measured data. A simultaneous search function is also
introduced to minimize the objective function, which represents the discrepancy between the numerically predicted and
experimentally measured pressures. The average diffusion coefficient is thus calculated for all BCs and the minimized
objective functions are compared with each other to determine an appropriate BC for a system. Other significant
assumption is that of negligible oil swelling. This is assumed because the position of the interface does not change
drastically over the length of the pressure cell.

The usual procedure in history matching is to adopt a Bayesian approach with an objective function that is assumed to
have a single simple minimum at the correct model. Reservoir history matching is a difficult inverse problem arising in
the industry. The main goal behind the history matching is to find a model such that the difference between the
performance of the model and history of the model is minimized. Gradient based optimization techniques are in practice
in the industry for computer aided history matching. This offers great time saving. In most of the history matching there
is usually an assumption that there exists a simple unique solution at the correct model. Hence the inherent non-
uniqueness of the solution of the inverse problem is neglected.

PVT Cell monitored in Pressure

Creux et al. (2005) established from their experiments that diffusivity in extra heavy oils is lower than in conventional oil.
They also proposed the correlation for temperature dependence of diffusivity in the determination of diffusion
coefficients. Creux et al. (2005) attempted to validate the experimental setup by means of two complementary
experimental setups, thermo-physical and analytical. The first method is a typical setup primarily based on the one used
by Riazi (1996) and then followed suit by Zhang et al. (2000) who further simplified it. Mathematically this model
originates in the work of Crank (1995) from the gas recombination kinetics data. This method is accurate for liquids with
low Gas-Oil-Ratios as diffusion lengths have been neglected. The second method used methane marker diffusing across a
plane sheet of known thickness. The methane 13C isotope is used as the marker. The model used by Creux et al. (2005)
uses a constant value of diffusion coefficient based on Ficks law. The solution to the Ficks model is given by Crank
(1995) in various conditions in semi infinite and limited diffusion length. For the second method, the solution given by
Crank (1995) to the Ficks equation is retained but the transportation of the gas is characterized by an analytical method
based on marker diffusivity. The experiment is conducted at conditions that are far from realistic values experienced in a
reservoir. The oil is also assumed to be mono-phasic which again is bound to cause deviations in the model and make it
inaccurate. Another difficulty while performing the second experiment is the immobilization of the oil.
12 SPE 134589

Non Intrusive Method to determine Gas Diffusivity

Upreti et al. (2000) used an indirect method to measure the change in pressure with time as gas diffusion into heavy oil
occurred. A comprehensive distributed parameter model of mass transfer was applied and a functional optimization
technique was used to compute the concentration dependent gas diffusivity. The calculation techniques do not use the
empirical correlations; instead the density difference of gases in liquid mixture is used. The diffusion of gas into the
bitumen causes the pressure decay. Since bitumen is non diffusing, the recorded PVT data and the density of the gas-
bitumen mixture yield experimental mass of gas absorbed in the bitumen phase. The main objective in the experiment
designed by Upreti et al. (2000) is to find the diffusivity at a given gas concentration, depth and time, such that the
bitumen mass of gas absorbed determined from the mass transfer model becomes equal to its experimental value. Indirect
methods based on property change are limited by several simplifications in the estimation of diffusivity values. The
techniques based on self diffusion values are limited by empirical mixing rules used to determine concentration
dependence of diffusivity.

Slopes and Intercepts Analytical Technique

Aconcha et al. (2008) and Aconcha and Kantzas (2009) used the CAT scanning technique to obtain the concentration
profiles of the n-alkanes and hydrocarbon gases respectively and then used the Slopes and Intercepts analytical technique
to calculate the concentration dependent diffusion coefficients. The X-ray CAT scanner collects and reconstructs the data
of X-ray transmissions through a two dimensional slice of an object without interference from overlaying and underlying
areas of the object. The CT image hence produced represents the point by point linear attenuation coefficients in the slice
which depend on the physical density of the material, the effective atomic number of the material under study and x-ray
beam energy. Low atomic numbers absorb less X-ray radiation. The CT image is un-obscured by other regions of the
test piece and is highly sensitive to small density differences between structures. Once, the concentration profiles were
calculated using the densities obtained by the X-ray measurements three zones are defined in the diffusion process.
Initially it is the early time period that is identified by a lower slope in the curve of mass diffused against root time,
followed by the intermediate region which is bounded by the time where the slope changes in the curve, mass diffused
against root time and the time at which the diffusive front reaches the physical boundary of the diffusion cell and the last
zone is the one where the assumption of the infinite cell does not apply and the physical boundary of the cell is reached.
For this reason the slopes and intercepts technique fails when applied to the third zone due to its assumption of an infinite
system. Hence theoretically this system can be used only in the first and second zones.

Pressure Decay Method

Molecular diffusion of CO2 from fracture to the oil in the matrix affects the rate of oil production from the matrix to the
fracture. Although the microscopic sweep efficiency of CO2 flooding is very high the volumetric sweep is a cause for
concern (Bernard et al. 1980). CO2 dissolves easily into oil reducing its viscosity, swelling it and then extracting the light
components in oil at sufficiently high pressures. However the local efficiency with which CO2 gas displaces oil depends
strongly on the phase behaviour of the mixtures of gas and oil (Jessen et al., 2005). Recovery from the reservoirs is a
result of a complex combination of different mechanisms such as viscous and capillary flow, extraction by molecular
diffusion and gravity drainage. If the fractures have sufficient vertical relief, with significant density difference, CO2
injection can recover a significant amount of oil by a gravity drainage displacement mechanism (Schecter, 2000). It's
been found by researchers that gravity drainage had no significant effect at the experimental conditions used. However,
in high pressure reservoirs the density of CO2 is the same order of magnitude as that of the crude oil (Heller et al., 1985,
Darvish et al., 2006a). This may reduce the role of gravitational forces and consequently the final oil recovery. Transport
processes are usually very slow and hence it's important to evaluate their effect on the overall oil recovery in the lifetime
of a project carefully. Many correlations are available in the literature but most are inadequate when it comes to high
pressures. A number of researchers have performed compositional analysis which is expensive and time consuming
especially at high pressures. A serious limitation of this work is the early time data affected by temperature fluctuations,
SPE 134589 13

natural convection or incubation period arising from initially high mass transfer rates and surface tension driven
instabilities. Convective transport has to be effectively eliminated in order to evaluate the diffusion coefficients more
effectively. The pressure drop in a closed sapphire cell that contained column of liquid below the gas cap was monitored
in order to measure the CO2 diffusivity. The amount of gas transfer is controlled by the diffusion coefficient of the gas.
The major assumption in the work by Fjelde et al. (2008) is the assumption of equilibrium between the gas and liquid
interfaces at all times. However we know from the film theory that this assumption is applicable as long as the
temperature and pressure are kept constant. Under the current circumstances where the pressure is decaying as the
diffusion of the gas into the oil is occurring this assumption does not hold good. The value of the equilibrium pressure
was estimated by extrapolation of the experimental pressure time curve. The final pressure reached in the experiment was
close to the equilibrium pressure but for it to actually reach the equilibrium value considerable amount of time would be
required. This raises the question that the effect of any error in estimating equilibrium pressure on the value of the
calculated diffusion coefficient. The graphical technique used by the authors is not sensitive enough to discriminate
between small deviations in the value of the equilibrium pressure. However, upon further investigation it is seen that the
diffusion coefficient is very sensitive to the equilibrium pressure. Hence the graphical technique cannot be considered a
reliable technique to determine the diffusion coefficients unless the value of the equilibrium pressure is determined
experimentally with minimum uncertainty.

Limitations and Challenges

The measurement of liquid phase mutual diffusion coefficients in bitumen and light hydrocarbon or gas mixtures presents
numerous experimental and data analysis challenges due to the viscosity and opacity of the mixtures and the variability of
density (Zhang et al., 2007). The most common assumptions made while performing the experiments are: no volumetric
expansion and volatility of oil, constant gas Z-factor and diffusion coefficient independent of gas concentration. Apart
from that, most of the laboratory processes use pure gases and single component oil instead of multicomponent gases and
crude oils. High temperatures and pressures are also avoided due to the physical limitations of the apparatus and hence
there is usually a great deviation from the actual cases, in particular for such oil fields that have strong heterogeneity, low
permeability, high viscosity and paraffin content, and high miscibility pressure of oil and CO2.

Typically concentration gradient drives the diffusion process. Under specific physical conditions in the PVT cell, the
density of the oil phase will decrease when the gas molecules diffuses into the oil phase. Diffusion coefficients for gas
and reservoir fluid or heavy oil mixtures are frequently obtained by interpreting pressure drop data over time for a pure
gas above a liquid at fixed volume (Riazi, 1996; Zhang et al., 2000; Tharanivasan et al., 2004). This method can work in
principle since the mass fraction of the gas dissolved at saturation is frequently small. However, the basic deficiency of
this approach in general is that the known impact of composition on the value of mutual diffusion coefficients is not
captured and it is not clear if the method is appropriate for gases with high solubilities (Zhang et al., 2007).

Correlation Methods

Sigmund (1976) demonstrated the effects of combination of convection and diffusion to calculate concentration profiles
for fluid mixtures flowing in porous media. Continuity equation was used as the basis for expressing composition change
of each component in the diffusing mixture with respect to time and space. For an n+1 component mixture the continuity
equation was given as

Where,
14 SPE 134589

C1, is the concentration of component i

, is the Darcy velocity

j1, is the flux of component i, with respect to the reference velocity

, is the gradient operator

For a system where all components have constant partial molar volume, at constant temperature and pressure the flux
equations which express diffusion rates with respect to a coordinate moving at the reference velocity was expressed as

The reference velocity can be taken as mass, molar or volume average. For an n+1 component system the above
equation defines an n X n matrix of diffuion coefficients [D] of the type developed by Onsager (1945). The n2 elements
of the matrix [D] are termed the multicomponent diffusion coefficients. The off diagonal diffusion coefficients
have been termed the cross diffusion coefficients and the diagonal values the main diffusion coefficients. The
magnitudes of the cross coefficients are a measure of the coupling or interaction that takes place between the n+1
diffusing species. An effective diffusion coefficient , can be considered to account for the effects of coupling such
that

Methods of estimating the multicomponent diffusion coefficient and effective diffusion coefficients were further
discussed by Sigmund (1976) in their work. da Silva et al. (1989) proposed an algorithm based on the Sigmund
Correlation (1976) that was an extension to the Sigmund (1976) correlation. It was suggestively better as it required only
the component critical properties and other parameters used in any equation of state.

Transport properties are commonly correlated using hydrodynamic, kinetic or statistical mechanical theories of fluids.
The Hydrodynamic theory is usually represented by the Stokes-Einstein equation and is limited to describing diffusion in
systems constituted by a large spherical particle immersed in a continuous solvent (Reis et al., 2005). The kinetic theory,
represented through Enskog's solution of the Boltzmann equation is a sufficiently well developed theory for gases at low
densities, but there is a lack in the development of a realistic description of molecular interactions for liquids and dense
fluids especially for polymeric solutions. Mehrotra et al. (1987) summed up methods for determining mutual and infinite
dilution diffusion coefficients. They found the Umesi-Danner correlation most accurate for the prediction of gas-liquid
infinite dilution diffusion coefficient and Teja's method the best of mutual diffusion coefficient. The paucity of literature
related to diffusion of gases into bitumen necessitated a review of the existing correlations for mass diffusivity. The
correlations were used by Mehrotra et al. (1987) on the premise that in order for a method to be discarded, it had to
produce a diffusivity value higher than an experimental value. Amongst the theories such as the hydrodynamic theory, the
kinetic theory, the statistical mechanical and the irreversible thermodynamic approach, the hydrodynamic theory was
selected as the basis for application. All the equations suggested were applicable only at atmospheric conditions. This to
an extent will downplay the accuracy of the values obtained. Rocha et al. (1997) and Dariva et al. (1999) estimated
diffusivities in dense fluids and fluid mixtures with the effective sphere diameter at sub and supercritical conditions.
Chao-Hong (1997) predicted the infinite-dilution molecular diffusion coefficients in the supercritical region as the
proposed theory was independent of the fluid state. Diffusion coefficients in liquids are readily correlated by the
molecular dynamics approach employing the rough hard-sphere theory of diffusion (Dymond, 1985; Easteal and Woolf,
1984). Yu and Gao (1999) developed an equation for self diffusion coefficient for polyatomic fluid as a sum of three
different friction terms: the temperature dependent hard sphere diameter (Ben-Amotz and Herschback, (1990), the chain
connectivity contribution, and the soft contribution. Mutual diffusion coefficients are used in a number of polymer
SPE 134589 15

processing operations. Therefore, it is important to know the dependence of temperature, concentration and molecular
weight on diffusion coefficient for polymer-solvent systems. Mutual diffusion coefficient in amorphous polymers and
organic solvents is a strong function of temperature and concentration. The most commonly used method for describing
diffusion in polymer-solvent systems is the Vrentas and Duda model (Vrentas and Duda, 1977a,b), and is based on the
free volume theory. It's seen from the literature that the predictive version of the model is able to qualitatively describe
the experimental data, but significant quantitative deviations are verified mainly at low-polymer concentrations (Reis et
al., 2001). In a prior work, self diffusion coefficient data was obtained by equilibrium molecular dynamics (MD)
simulation technique (Reis et al., 2001) for freely joined Lennard-Jones chain (LJC) fluids. Based on both the Chapman-
Enskog formalism and on the MD simulation data a semi-empirical equation to correlate diffusion coefficients for LJC
was proposed (Reis et al., 2005).

Mutual Diffusion Coefficient

The Wilke Equation (1949):

The Caldwell and Babb Equation (1956):

The Leffler and Cullinan Equation (1970):

The Teja generalized corresponding states procedure (1985):

Where,

⁄ ⁄

⁄ ⁄

⁄ ⁄

⁄ ⁄

⁄ ⁄ ⁄
16 SPE 134589

And the cross parameters, i ≠ j are obtained by,

⁄ ⁄

The use of above methods required infinite dilution values for the CO2 bitumen system; hence it was necessary to first
predict the infinite dilution diffusion coefficients.

Infinite Dilution Diffusion Coefficient

The Stokes- Einstein Equation (Daniles and Alberty, 1962):

The Wilke-Chang equation (Wilke and Chang, 1955):

7.4 10 .

The Othmer-Thakar equation (Othmer and Thakar, 1953):

1.4 10
.
1.1

The Eyring-Jhon Equation (Eyring and Jhon, 1969):

6
SPE 134589 17

The Akgerman-Gainer Equation (Akgerman and Gainer, 1972a, 1972b):

Where,

ln ln
2
1 1

.
5875.3

The Sridhar-Potter Equation (Sridhar and Potter, 1977):

0.088

Where Vo = 0.31VcB

The Umesi-Danner Equation (Umesi and Danner, 1981):

2.75 10

The first four predictive methods applicable to dilute binary liquid-liquid systems, while the fifth and the seventh are used
for the prediction of gas-liquid systems and the sixth is applicable for both cases. Mehrotra et al. (1987) observed in their
work that the diffusivity values for liquid-liquid systems were lower and scattered over a wide range of values as
compared to gas-liquid predictive methods. The gas-liquid methods appear to be more reliable than the liquid-liquid
methods and suggest that the CO2/bitumen system probably behaves like a gas-liquid system.

Incorporation of Diffusion Coefficient in Simulation Techniques

Several models could be used for the simulation of flow and transport in the fractured porous media. A lot of work has
been done in the industry pertaining to the single and double porosity models. We however discuss them only in brief
and focus on recent approaches and advancement.

Single porosity model is used as an explicit computational representation for fractures. Geological extents are allowed to
vary between the matrix and the fractures. High contrast and different length scales make the approach unpractical
because of the ill-conditionality of the matrix appearing in the numerical computations (Ghorayeb and Firoozabadi,
18 SPE 134589

2000a). The small control volumes in the fracture grids also add a severe restriction on the time step size because of the
Courant-Freidricks-Levy condition if an explicit temporal scheme is used. Miscibility in conventional single porosity
reservoirs has been researched in the past few years. Experimental and numerical procedures have been proposed that
give a definitive measure of the miscible development for one dimensional flow system. It is seen that one dimensional
miscible flow process is governed by phase behaviour and not at all by rock or other fluid properties. However in case of
fractures it is not even certain that miscibility can be obtained.

Another model of significance is the dual porosity / dual permeability model. This model works under the assumption
that the medium is composed of many well connected fractures. The fracture network is described by an equivalent
porous medium that generally has the configuration of a sugar cube. The accuracy of this method depends on the
description of the transfer functions between the matrix and the fractures. These functions may or may not be properly
defined with gravity, compressibility, and compositional effects. Another limitation in this approach is the assumption
that the medium may lose accuracy when treating few discrete or connected fractures. The dual porosity method was first
conceived and proposed by Barenblat et al. (1960) for a single phase system. This was further used by Warren and Root
in their work to develop a pressure transient analysis method for naturally fractured reservoirs. Finally, Kazemi et al.
(2009) extended the work of Warren and Root (1963) to multiphase flow using a two dimensional, two phase, black oil
formulation. The two resultant equations were linked by means of a matrix-fracture transfer function. The Kazemi et al.
(1976) dual porosity approach now is the most prevalent in the industry to develop field scale simulation models for
naturally fractured reservoir performance.

The dual porosity model is the most commonly used conceptual model for simulating naturally fractured reservoirs. In
this model there are two kinds of porosity that are present in the rock volume, fracture and the matrix. Matrix blocks are
surrounded by fractures and the system is visualized as a stack of volumes, representing matrix blocks separated by
fractures. Hence there is no communication between matrix blocks in this model and the fracture network is continuous.
However, the matrix blocks do communicate with the fractures that surround them. The efficiency of the simulator
depends on the accuracy of the transfer functions that characterize fluid flow between the matrix blocks and the fractures.
Naimi-Tajdar et al. (2006) foresaw the need of developing an accurate parallel simulator for large scale naturally
fractured reservoirs that would be capable of modeling fluid flow in both rock matrix and fractures. The computational
problem could be answered by a simulator that runs parallel, 3D, fully implicit, equation of state compositional model
that solved very large, sparse linear system, arising from discretization of the governing partial differential equations. A
generalized dual porosity, the Multiple Interacting Continua (MINC) has been implemented in this simulator. One of the
important features of this simulator as claimed by the authors, which is not available in any commercial simulator is its
ability to model both gas and water processes and the ability of two dimensional matrix subgridding. The authors of this
work also claim to have verified the results of this simulator against analytical solutions and other available commercial
simulators such as Eclipse and GEM. The MINC is an approximate method for modeling fluid and heat flow in fractured-
porous media. The method is an extension of the double-porosity concept, originally developed by Barenblatt et al.
(1960) and Warren and Root (1963). It is based on the fact that fractures have large permeability and small porosity
(when averaged over a reservoir subdomain), while the intact rock has the opposite characteristics. Therefore, any
disturbance in reservoir conditions will travel rapidly through the network of interconnected fractures, while invading the
matrix blocks only slowly. The primary assumption in the MINC method is that the thermodynamic conditions in the
matrix depend on the distance from the nearest fracture. Hence the flow domain can be partitioned into compositional
domain such that all the interfaces between volume elements in the matrix are parallel to the nearest fracture.

The matrix blocks may contain a large oil reserve. Hence it's crucial to understand the mechanisms that take place in the
matrix blocks and to simulate these processes accurately. Discretizing the matrix blocks into subgrids or subdomains is a
very good solution to accurately take into account transient and spatially non linear flow behaviour in the matrix blocks.
The resultant finite difference equations are solved along with the fracture equations to calculate matrix fracture transfer
flow. The discretization of the matrix blocks depends on the proposed models, but the objective is to accurately model
pressure and saturation gradients in the matrix blocks. Ghorayeb et al. (2000b) examined the fracture extents such as
SPE 134589 19

aperture, intensity and connectivity on the fluid compositional variation in a fractured porous media. They concluded that
for high fracture aperture the convective motion is pronounced within the fracture whereas the composition is only
affected beyond a certain fracture aperture. The effect of discrete and connected fractures is also studied. They addressed
four primary reasons for compositional variation in single phase in reservoirs as, thermal diffusion, pressure diffusion,
molecular diffusion and natural convection. However what is lacking in the work so far is the response of the system to
various orientations of fracture and multiphase flow. Coefficients of molecular diffusion, pressure diffusion and thermal
diffusion are assumed constant; this assumption could lead to errors in evaluation of the coefficients and the researchers
have to find a way to measure these values for early time and equilibrium conditions. It was assumed that the conductive
flow of heat is higher than convective flow and that the thermal diffusivity is very large compared with molecular
diffusion coefficient. No basis was given for such an approximation and considering that the flow through fractures is
fairly high, there is a possibility that convective diffusion is predominant. The steady state solution is sought until the
steady state is reached. The discretization of the governing equations is a huge task and it creates a large sparse system of
linear equations. The matrixes are highly ill-conditioned because of the spatial discontinuity in permeability of matrix
and fractures and small fracture thicknesses in comparison to the matrix size. Thus special robust pre-conditioners have
to be used for the iterative systems to effective. Hence direct methods are suggested such as Gauss Elimination. This
could however have an impact on the computation time for the solvers. The results of the solver can further be solved by
developing algorithms that are capable of solving equations for matrix with varying permeability. As an alternative,
Trivedi & Babadagli (2009) used the finite element matrix fracture model to evaluate the mass transfer between fracture
and matrix and effective diffusion/dispersion in the matrix of naturally fractured subsurface reservoirs.

Zhang et al. (2004) investigated the effects of porosity of porous media and the fracture aperture on transport in fractured
porous media. They found that unlike transport in channels or in porous media, solute transport in fractured porous media
is non-Gaussian with long tails and with time dependent mean plume velocities and non-Fickian dispersion coefficients.
Due to statistically complex distribution of geological heterogeneity and the multiple length and time scales involved,
three approaches are usually used in describing fluid flow and solute transport in naturally fractured porous formations.
They are discrete fracture models, continuum models including double porosity models using effective properties to
represent the effects of fractures and hybrid models that combine the previous two. In the discrete fracture approach the
effect of porous matrix on flow and transport is usually neglected. However, as recently concluded by Kang et al. (2002)
this is not a good assumption when the matrix permeability is finite. The matrix flow has a significant effect on the
fractured system and the assumption that the matrix is impermeable is not valid. Hence the use of cubic law to calculate
the fracture permeability may cause a significant error. Thus it is essential to take into account the effect of porous matrix
and represent the interactions between the fractures and the matrix blocks in the continuum and hybrid approaches.

Miscibility in conventional one dimensional system is dependent on fluid properties alone and can be estimated by means
of common experimental and numerical procedures. This is not the case for heterogeneous and multidimensional systems
such as fractured reservoirs. Uleberg et al. (2002) proposed a method based on a multicell algorithm to determine the
conditions of developed miscibility in fractured reservoirs. They used a systematic step by step approach from a single
one dimensional matrix block to a more complex matrix fracture systems and showed that the minimum miscibility
pressure/enrichment (MMP/MME) level in a fractured reservoir is significantly higher than for a conventional one
dimensional single porosity system. This is attributed to multi dimensional flow and molecular system. Despite all this
significant enhancements can be expected by non equilibrium gas injection. It is known that a reduction in the gas-oil
interfacial tension and internal Darcy flow induced by interfacial tension gradients play a key role.

Deficiency of commercial simulators in modelling diffusion

Darvish et al. (2006b) developed an Equation of State (EOS) model to be used in compositional simulation after
comprehensive PVT studies were performed on the reservoir fluids. Cross phase diffusion is modelled and its effects on
the matrix oil are investigated. It is seen from the resulting plots that the produced oil in the early stages was very light
and included very less amounts of C23+ while the last oil sample the weight percentage increased almost four times. The
20 SPE 134589

authors cite diffusion mechanism over the others such as gravity drainage for this phenomenon. However, this conclusion
can only be verified once many cores of varying permeability are sampled by researchers under consistent conditions.
The simulation model used for this system is expected to model self as well as cross diffusion. The compositional
simulator used here is the Eclipse version 2004A, which is incapable of capturing the cross phase diffusion. For this
purpose a 2D single porosity radial dummy model is introduced. Gas and oil diffusion coefficients are set to zero for the
inert component. The extended Sigmund correlation is then used to determine the diffusion coefficients for porous media.
Further, history matching technique is used to determine the value of the cementation factor and the best match is seen at
1.7. Large pore volumes for fractures were used to simplify the model and eliminate the need for injection and
production wells. However, this is at the cost of accurate prediction of the oil production. It helps only to quantify
diffusion mechanism. It was also noticed that a certain amount of composition concentration increased in the dummy
region. It is thus evident from experimental work of Darvish et al. (2006b) that the current compositional simulators need
to be upgraded to better represent cross diffusion in reservoirs. This proves the importance of diffusion as one of the
important displacement mechanisms which causes the transport of components based on their diffusivities from a tight
matrix to a fracture network. Diffusion calculations using the current simulation models such as Eclipse 300 (2004), are
based on inconsistent upstream formulation of gas to gas and liquid to liquid component transport in both fracture and
matrix domains. It is seen that the complex component exchange involved in the dominant diffusion mechanisms results
in inaccurate formulation which results in underestimation of recovery. Das (2005) experienced problems in modeling
the bulk diffusion coefficient due to the microscopic nature of the process and suggested the use of smaller grid size along
the interface to tackle the problem. Dauba et al. (2002) performed the compositional part of their experiments on semi-
compositional and fully compositional models (STARS and GEM) (Nghiem et al., 2001). In the PVT description for the
semi composition model it was assumed that there is no dissolved gas in the oil and the model is tuned to match the Peng
Robinson Equation of State (1976) on density and viscosity using only one component to describe the initial oil and two
components for the solvent/gas phase. The Peng Robinson Equation of State (1976) was used to simultaneously match
the properties of the initial oil and that of the mixture. Lohrenz-Bray-Clark (1964) correlation is used to compare the
experimental results but the poor match confirms the inadequacy of this correlation for modelling heavy oils. Moortgat et
al. (2009) stated Fickian diffusion as the main recovery mechanism in reservoir conditions. CO2 transfer occurred from
the matrix to the fracture as a consequence of diffusion. They observed that the recovery rates to be high initially but as
the lighter components are vaporized the viscosity of the remaining oil increases and the recovery rate drops. Further,
decrease is observed due to decreasing contact surface.

Discrete Fracture Models

To overcome the shortcomings of these models, the discrete fracture model is considered as an alternative. Fractures are
treated explicitly in the medium. They are represented by (n-1) dimensional elements in an n dimensional domain. This
simplification makes discrete fracture model much more efficient than the single porosity model. The pressure in the
fracture element is assumed to be equal to the pressure in the surrounding matrix. This assumption is valid as long as the
matrix elements are not too large. Integrating the flow equations in a control volume that includes the fracture element
and the adjacent matrix elements alleviates the computation of the matrix / fracture fluxes with the cross flow equilibrium
assumption. This simplification results in computational efficiency and the reduction in CPU time can be two or three
orders of magnitude compared to the single porosity models. This also overcomes the limitations of the dual porosity
model.

Cell based finite volume schemes used for the discrete fracture model have fracture pressure unknown at the discretized
element centers. Using two point approximations would be sufficient to evaluate the flux across the intersection of two
fracture branches. However, computing flux across the intersection of three or more branches is a challenge. Most of the
prevalent techniques that describe the flux across the multi intersecting fractures have some deficiency in them. In their
approach Hoteit and Firoozabadi (2004) solved this deficiency by the Mixed Finite Element (MFE) formulation that
provided pressure at the interface of the fracture element in addition to the cell pressure average. The governing flow
equations in the 2D fractured and the un-fractured media are solved by using a numerical approach that combines the
SPE 134589 21

MFE and Discontinuous Galerkin (DG) along with the discrete fracture model. The capillary pressure and the diffusion
processes are neglected.

Hoteit and Firoozabadi (2006a) used a discrete fracture model to approximate two phase flow with mass transfer in
fractured media. This model is numerically superior to the single porosity model and overcomes limitations of the dual
porosity model including the use of the shape factor.

Multi-component Diffusion Model

Hoteit and Firoozabadi (2006b) used the multicomponent diffusion coefficient model of Ghorayeb and Firoozabadi
(2000a) to determine the full diffusion matrix as a function of pressure, temperature and composition.

The diffusion coefficients are written as,

Γα

Where the elements of matrices Bα and Гα are given by

, ,
, ,
1 1
,
, , , ,

And

,
, , ,
,
, , ,

The derivatives of the fugacities in the above equation are calculated from the Peng-Robinson (1976) equation of state.
Di,j,α is known as the Stefan-Maxwell coefficients and is defined as,

,
, ,
, , , , , , ,
,
,

i, j = 1, ... , nc-1
0
Where i,j are the infinite dilution coefficients.

Ghorayeb and Firoozabadi (2000b) used the correlation of Hayduk and Minhas (1982) to estimate these coefficients,

.
. . .
, 13.3 10 ,

Where Vi and μj are the molar volume and the viscosity at the normal boiling point of the components i and j,
respectively.
22 SPE 134589

Guo et al. (2009) used equations with established boundary conditions to describe the mutual diffusion coefficient
between different components. Effective diffusion coefficients of individual components had an impact on the time
required for the entire system to attain equilibrium. Since there are no general equations available that could be used to
accurately determine the diffusion coefficient of i-component in the oil phase and gas phase, some empirical equations
were developed. The diffusion factor of the i-component in the oil phase was calculated using the empirical equation
developed by Wilke and Chang (1955) and in the gas phase the equations developed by Chapman-Enskog (Kihara, 1975)
are used. The initial value of K of each component is calculated by the Wilson function, and was further corrected using
the fugacity coefficient of each time step, while the fugacity coefficient was calculated using the Peng-Robinson equation
of state. As compared to conventional computational models for single components, the model used by Guo et al. (2009)
take into consideration the mutual influence between the different components which describes the molecular diffusion in
process of a real gas-oil system.

Oil phase:

,0

0,
0

Gas Phase:

,0

0,

,
0

C1oi and C1gi are the initial molar concentrations of i-component of the oil phase and gas phase, respectively, in kmol/m3.
Cobi and Cgbi are the molar concentrations of i-component of the oil phase and gas phase at the oil-gas interface
respectively, in kmol/m3.

Lattice Boltzmann models

In their work Zhang et al. (2004) quantify pore scale simulations of solute transport in porous media with single fracture
of varying apertures. They used Lattice Boltzmann (LB) (Kang et al., 2002) method as tool for numerical modeling.
Unlike the conventional methods that are based on discretization of macroscopic continuum equations the LB method is
SPE 134589 23

based on microscopic models and mesoscopic kinetic equations. This feature gives the LB method the advantage of
studying non equilibrium dynamics, especially in fluid flow applications involving interfacial dynamics and complex
boundaries.

Summary

In our review we have touched upon many factors that affect diffusion mechanisms and the techniques that are used in the
industry to determine the diffusion coefficients in fractured and porous media. The dominant mechanisms that effect
recovery and greenhouse gas sequestration are extraction, gravity drainage and miscibility. The summary of our review
could be highlighted as follows:

• Gravity drainage can significantly enhance oil recovery as a secondary or tertiary drive mechanism in fractured
reservoirs.

• Miscibility condition may affect the extraction which is a consequence of diffusion. Minimal increase of
recovery beyond miscibility suggests diffusion may not play a major role after certain pressure and that is around
miscible pressure. Displacement as well as extraction of oil increases with higher pressure. Displacement seems
to be the dominant mechanism of oil recovery initially, after which oil extraction becomes important.

• Typically, the industry relies on the use of Ficks second law for representing the mass transfer models. The
researchers tend to assume diffusion coefficient as a constant which we know is not the case. The models with
temperature, pressure and concentration dependent diffusion coefficient should be considered for further
research.

• We have discussed the existing methods of determining diffusion coefficient between crude oil and solvent or
gas with limitations, assumptions and approximations. Various existing correlations are also presented here.

• The experiments results should be backed by adequate models and simulation results. In particular the use of
discrete fracture models is recommended. While modelling, it should be noted that subgridding of matrix blocks
based on distance from fractures gives rise to a pattern of nested volume elements. This increases computer time
and storage especially when all flow equations are solved implicitly. Even though serial computing and
simulation technology may be adequate for typical reservoirs, naturally fractured reservoirs need more grid
blocks to sufficiently define the transient flow in matrix media.

• Numerical simulators with of multicomponent cross diffusion modelling may overcome the deficiencies of
existing models to represent matrix-fracture diffusive transfer.

Acknowledgement

This study is supported through funds from University of Alberta and Natural Sciences and Engineering Research
Council (Discovery Grant).

Nomenclature

D, is the diffusivity (cm2/sec);

ED, is the activation energy for diffusion (J/mol);


24 SPE 134589

Eµ, is the activation energy for viscosity (J/mol);

L, is the latent heat of vaporization (J/mol);

M, is the molar mass (g/mol);

No, is the Avogadro's number (6.023 X 1023 mol-1);

P, is the pressure (Pa);

R, is the universal gas constant;

r, is the radius of the molecule (cm);

, is the radius of gyration ( );

T, is the temperature (K);

V, is the molar volume (cm3/mol);

Vo, 0.31VcB;

k, is the Boltzmann's constant (1.38 X 10-23 J/K);

x, is the mole fraction

Subscripts and Superscripts

A, solute;

B, solvent;

C, critical;

i, solute;

j, solvent;

m, mixture;

w, water; 1,2 are the reference fluids;

j, jumping energy

o, infinite dilution;

Symbols

ξ = M1/2Tc-1/2Pc-1/3

A, is the number of B molecules around a central A molecule;

Ø, is the association parameter;


SPE 134589 25

µ, is the viscosity in g/cm.sec and mPa.sec for the Wilke-Chang (1955), Othmer-Thakar (1953) and Umesi-Danner (1981)
equations;

µ, is the viscosity;

ψ, is the binary interaction parameter

References

Abukhalefeh, H., Lohi, A. and Upreti, S.R.: “A Novel Technique to Determine Concentration Dependent Solvent
Dispersion in Vapex”, Energies, 2, pp 851-872, 2009.

Afsahi, B. and Kantzas, A.: "Advances in Diffusivity Measurement of Solvents in Oil Sands", Journal of Canadian
Petroleum Technology, 46(11), pp 56-61, 2007.

Akgerman, A. and Gainer, J.L.: "Diffusion of Gases in Liquids", Ind. Eng. Chem. Fundamentals, 11(3), pp 373-379,
1972a.

Akgerman, A. and Gainer, J.L.: "Predicting Gas-Liquid Diffusivities", J. Chem. Eng. Data, 17(3), pp 372-377, 1972b.

Alavian, S.A. and Whitson, C.H.: "CO2 IOR Potential in Naturally Fractured Haft Kel Field, Iran", IPTC 10641,
International Petroleum Technology Conference, Doha, Qatar, November 21-23, 2005.

Ali H. Al-Marzouqi, Abdulrazag Y. Zekri, Baboucarr Jobe and Ali Dowaidar: "Supercritical fluid extraction for the
determination of Optimum Oil Recovery Conditions", Journal of Petroleum Science and Engineering, 55, pp 37-47, 2007.

Asghari, K. and Torabi, F.: "Laboratory Experimental Results of Huff 'n' Puff CO2 Flooding in a Fractured Core System",
SPE 110577, SPE Annual Technical Conference and Exhibition, Anaheim, California, U.S.A, November 11-14, 2007.

Barenblatt, G.E., Zheltov, I.P. and Kochina, I.N.: "Basic concepts in the Theory of Seepage of Homogeneous Liquids in
Fissured Rocks Journal of Applied Mathematics", 24 (5), pp 1286-1303, USSR, 1960.

Bear, J.: "Dynamics of Fluids in Porous Media", Elsevier, New York, 1972.

Ben-Amotz, D. and Herschbach, D.R.: "Estimation of effective diameters for molecular fluids", Journal of Physical
Chemistry, 94, pp 1038-1047, 1990.

Bernard, G.G., Holm, L.W. and Harvey, C.P.: "Use of Surfactant to Reduce CO2 Mobility in Oil Displacement", SPE
8370-PA, SPE Journal, 20(4), pp 281-292, 1980.

Blackwell, R.J.: "Laboratory Studies of Microscopic Dispersion Phenomena", Soc. Pet. Eng. J., 2, April, 1962.

Brigham, W.E., Reed, P.W. and Dew, J.N.: "Experiments on mixing During Miscible Displacement in Porous Media",
Society of Petroleum Engineering Journal, 1, March, 1961.

Butler, R.M. and Morkys, I.J.: "A new process (VAPEX) for recovering heavy oils using hot water and hydrocarbon
vapour", JCPT 91-01-09, CIM/SPE Annual Technical Conference, 30(1), pp 97-106, Jan-Feb, 1991.

Caldwell, C.S. and Babb, A.L.: "Diffusion in Ideal Binary Liquid Mixures", J. Phys. Chem., 60, pp 51-56, 1956.
26 SPE 134589

Chao-Hong, H.: "Prediction of Binary Diffusion Coefficients of Solutes in Supercritical Solvents", AIChE Journal,
43(11), pp 2944, Nov 1997.

Civan, F. and Rasmussen, M.L.: "Analysis and interpretation of gas diffusion in quiescent reservoir, drilling, and
completion fluids: equilibrium vs. non-equilibrium models", SPE 84072, SPE Annual Technical Conference and
Exhibition, Denver, Colorado, USA, October 5-8, 2003.

Civan, F. and Rasmussen, M.L.: "Improved measurement of gas diffusivity for miscible gas flooding under non-
equilibrium vs. equilibrium conditions", SPE 75135, SPE/DOE 13th Improved Oil Recovery Symposium, Tulsa, OK,
USA, April 13-17, 2002.

Computer Modelling Group LTD manuals, STARS and GEM version, Calgary, Canada.

Crank, J.: "The Mathematics of Diffusion, IInd Edition", Clarendon Press, Oxford Science Publications, 47, 1995.

Creux, P., Meyer, V., Gracia, A., Cordelier, P.R., Franco, F. and Montel, F.: "Diffusivity in Heavy Oils", SPE 97798, SPE
International Thermal Operations and Heavy Oil Symposium, Calgary, Canada, November 1-3, 2005.

da Silva, F.V. and Belery, P.: "Molecular Diffuion in Naturally Fractured Reservoirs: A Decisive Recovery Mechanism",
SPE 19672, 64th Annual Technical Conference and Exhibition, San Antonio, Texas, U.S.A., October 8-11, 1989.

Daniels, F. and Alberty, R.A.: "Physical Chemistry", Wiley, 581, 1962.

Dariva, C., Coelho, L.A.F. and Oliveira, J.V.: "A Kinetic approach for predicting diffusivities in dense fluid mixtures",
Fluid Phase Equilibria, 158-160, pp 1045-1054, 1999.

Darvish, G.R., Lindberg, E., Holt, T. and Utne, S.A.: "Laboratory Experiments of Tertiary CO2 Injection Into a Fractured
Core, SPE99649, SPE/DOE Symposium on Improved Oil Recovery, Tulsa, Oklahoma, U.S.A., April 22-26, 2006a.

Darvish, G.R., Lindberg, E., Holt, T. and Utne, S.A.: "Reservoir Conditions Laboratory Experiments of CO2 Injection
into Fractured Cores", SPE99650, SPE Europe/ EAGC Annual Conference and Exhibition, Vienna, Austria, June 12-15,
2006b.

Das, S.: “Diffusion and Dispersion in the Simulation of Vapex Process”, SPE 97924, SPE International Thermal
Operations and Heavy Oil Symposium, Calgary, Alberta, Canada, November 1-3, 2005.

Dastyari, A., Bashukooh, B., Shariatpanahi, S.F. and Haghighi, M.: "Visualization of Gravity Drainage in a Fractured
System During Gas Injection Using Glass Micromodel", SPE 93673, 14th SPE Middle East Oil & Gas Show and
Conference, Baharain, March 12-15, 2005.

Dauba, C., Quettier, L., Christensen, J., Le Goff, C. And Cordelier, P.: "An Integrated Experimental and Numerical
Approach to Assess the Performance of Solvent Injection Into Heavy Oil", SPE 77459, SPE Annual Technical
Conference and Exhibition, San Antonio, Texas, U.S.A, September 29-October2, 2002.

Dunn, S. G., Nenniger, E.H. and Rajan, V. S. V: "A Study of Bitumen Recovery by Gravity Drainage using Low
temperature Soluble-gas Injection", Can. J. Chem. Eng., 67, pp 978-991, 1989.

Dymond, J.H.: "Hard-Sphere Theories of Transport Properties", Chem. Soc. Res., 3, 317, 1985.

Easteal, A. J., and L. A. Woolf, “Diffusion in Mixtures of Hard Spheres at Liquid Densities: A Comparison of Molecular
Dynamics and Experimental Data in Equimolar Systems,” Chem. Phys., 88, pp 101, 1984.
SPE 134589 27

Eclipse 300 Technical Manual (2004).

Etminan, S.R., Maini, B.B., Hassanzadeh, H. and Chen, Z.: "Determination of Concentration Dependent Diffusivity
Coefficeint in Solvent Gas Heavy Oil System", SPE 124832, SPE Annual Technical Conference and Exhibition, New
Orleans, Louisiana, U.S.A., October 4-7, 2009.

Eyring, H. and Jhon, M.S.: "Significant Liquid Structures", Wiley, 1969.

Fisher, D.B., Singhal, A.K.,Das, S.K. and Jackson, C.: "Use of Magnetic Resonance Imaging and Advanced Image
Analysis as a Tool to Extract Mass Transfer Information from a 2D Physical Model of The Vapex Process", SPE 59330,
SPE/DOE Improved Oil RecoverySymposium, Tulsa, Oklahoma, April 3-5, 2000.

Fjelde, I., Zuta, J. and Duyilemi, O.V.: "Oil Recovery from Matrix during CO2-Foam Flooding of Fractured Carbonate
Oil Reservoirs", SPE 113880, SPE Europe/EAGE Annual Conference and Exhibition, Rome, Italy, June 9-12, 2008.

Ghedan, S.: "Global Laboratory Experience of CO2-EOR Flooding", SPE125581, SPE/EAGE Reservoir Characterization
and Simulation Conference, Abu Dabhi, UAE, October 19-21, 2009.

Ghoyayeb, K. and Firoozabadi, A.: "Molecular, pressure, and thermal diffusion in nonideal multicomponent mixtures",
AIChE J., 46(5), pp 883-891, 2000b.

Ghoyayeb, K. and Firoozabadi, A.: "Numerical Study of Natural Convection and Diffusion in Fractured Porous Media",
SPE Journal, 5(1), pp 12-20, 2000a.

Grogan, A.T. and Pinczewski, W.V.: "The Role of Molecular Diffusion Process in Tertiary CO2 Flooding", Journal of
Petroleum Technology, 39(5), pp 591-602, 1987.

Grogan, A.T., Pinczewski, V.W., Ruskauff, G.J. and Orr, Jr, F.: "Diffusion of CO2 at Reservoir Conditions: Models and
Measurements", SPE Reservoir Engineering, 3(1), pp 93-102, February 1988.

Guerrero-Aconcha, U., Salama, D. and Kantzas, A.: "Diffusion Coefficient of n-alkanes in Heavy Oil", SPE 115346, SPE
Annual Technical Conference and Exhibition, Denver, Colorado, U.S.A. September 21-24, 2008.

Guerrero-Aconcha, U. and Kantzas, A.: "Diffusion of Hydrocarbon Gases in Heavy Oil and Bitumen", SPE 122783, SPE
Latin American and Caribbean Petroleum Engineering Conference, Cartagena, Columbia, May 31-June 3, 2009.

Guo, P., Wang, Z., Shen, P. and Du, J.: "Molecular Diffusion Coefficients of the Multicomponent Gas-Crude Oil Systems
under High Temperature and Pressure", Ind. Eng. Chem. Res., 45, pp 9023-9027, 2009.

Hayduk, W. and Minhas, B.S.: "Correlations for prediction of Molecular Diffusivities in Liquids", Can. J. Chem. Eng.,
60, pp 295-299, 1982.

Heller, J.P., Lien, C.L. and Kuntamukkula, M.S.: "Foamlike Dispersions for Mobility Control in CO2 Floods`, SPE
11233-PA, SPE Journal, 25(4), pp 603-613, 1985.

Hill, E.S. and Lacey, W.N.: “Rate of solution of propane in quiescent liquid hydrocarbons”, Ind. Eng. Chem., 26(2), pp
1324-1327, 1934.

Holm, L.W. and Josendal, V.A.: “Mechanism of Oil Displacement by Carbon Dioxide”, Journal of Petroleum
Technology, pp 1427-1438, December, 1974.
28 SPE 134589

Hoteit, H. and Firoozabadi, A.: "Compositional Modeling of Discrete-Fractured Media without Transfer Functions by the
Discontinuous Galerkin and Mixed Methods", SPE Journal, 11(3), pp 341-352, 2006a.

Hoteit, H. and Firoozabadi, A.: "Compositional Modeling of Fractured Reservoirs without Transfer Functions by the
Discontinuous Galerkin and Mixed Methods", SPE90277, SPE Annual Technical Conference and Exhibition, Houston,
Texas, U.S.A., September 26-29, 2004.

Hoteit, H. and Firoozabadi, A.: "Numerical Modeling of Diffusion in Fractured Media for Gas-Injection and Recycling
Schemes", SPE 103292, SPE Annual Technical Conference and Exhibition, San Antonio, Texas, U.S.A., September 24-
27, 2006b.

Jessen, K., Kovseck, A.R. and Orr Jr. F.M.: "Increasing CO2 Storage in Oil Recovery", Energy Conversion and
Management, 46, pp 293-311, 2005.

Kang, Q., Zhang, D. and Chen, S.: “Unified lattice Boltzmann method for flow in multiscale porous media”, Phys. Rev.
E, 66, 056307, 2002.

Karimaie, H., Darvish, G.R., Lindberg, E. and Torsæter, O.: "Experimental Investigation of Secondary and Tertiary Gas
Injection in Fractured Carbonate Rock", SPE107187, SPE Europe/EAGE Annual Conference and Exhibition, London,
United Kingdom, June11-14, 2007.

Karimaie, H. and Torsæter O.: "Low IFT Gas-Oil Gravity Drainage in Fractured Carbonate Porous Media", SPE113601,
SPE Europec/EAGE Annual Conference and Exhibition, Rome, Italy, June 9-12, 2008.

Kazemi, A. and Jamialahmadi, M.: "The Effect of Oil and Gas Molecular Diffusion in Production of Fractured Reservoir
During Gravity Drainage Mechanism by CO2 Injection", SPE120894, SPE Europec/EAGE Annual Conference and
Exhibition, Amsterdam, Netherlands, June 8-11, 2009.

Kazemi, H., Merrill, L.S., Porterfield, K.L. and Zemen P.R.: “Numerical Simulation of Water-Oil Flow in Naturally
Fractured Reservoirs”, SPE 5719, SPE AIME Fourth Symposium on Numerical Simulation of Reservoir Performance,
Los Angeles, CA, U.S.A, February 19-20, 1976.

Kihara, T.: "The Chapman-Enskog and Kihara Approximations for Isotopic Thermal Diffusion in Gases", J. Stat. Phys.,
13, 137, 1975.

Le Romancer, J-F.X., Defives, D.F. and Fernandes, G.: "Mechanism of Oil Recovery by Gas Diffusion in Fractured
Reservoir in Presence of Water", SPE 27746, SPE/DOE Ninth Symposium on Improved Oil Recovery, Tulsa, Oklahoma,
U.S.A., April 17-20, 1994.

Leffler, J. and Cullinan, H.T.: "Variation of Liquid Diffusion Coefficients with Composition: Binary Systems", Ing. Eng.
Chem. Fundam., 9(1), pp 84-88, 1970.

Li, H., Putra, E., Schechter, D.S. and Grigg, R.B.: “Experimental Investigation of CO2 Gravity Drainage in a Fractured
System”, SPE 64510, SPE Asia Pacific Oil and Gas Conference and Exhibition, Brisbane, Australia, October 16-18,
2000.

Li, H.: "Experimental Investigation of CO2 Gravity Drainage in Spraberry and Berea Whole Core", the Fifth Naturally
Fractured Reservoir Symposium, Socorro, NM, U.S.A, October 28, 1999.

Lohrenz, J., Bray, B.C. and Clark, C.R.: "Calculating Viscosities of Reservoir Fluids from their Compositions", Journal of
Petroleum Technology, pp 1171-1176, October, 1964.
SPE 134589 29

Mahmoud, T.N. and Rao, D.N.: "Range of Operability of Gas-Assisted Gravity Drainage Process", SPE 113474-MS,
SPE/DOE Symposium on Improved Oil Recovery, April 20-23, Tulsa, Oklahoma, USA, 2008.

Mehrotra, A., Garg, A. and Svrcek, W.: "Prediction of Mass Diffusivity of CO2 into Bitumen", The Canadian Journal of
Chemical Engineering, 65, pp 826-832, 1987.

Moortgat, J., Firoozabadi, A. and Farshi, M.M.: "A New Approach to Compositional Modeling of CO2 Injection in
Fractured Media Compared to Experimental Data", SPE124918, SPE Annual Technical Conference and Exhibition, New
Orleans, Louisiana, U.S.A., October 4-7, 2009.

Morel, D.D., Bourbiaux, B., Latil, M. and Thiebolt, B.: "Diffusion Effects in Gas Flooded Light Oil Fractured
Reservoirs", SPE 20516, Presented at the 65th Fall Meeting, New Orleans, September, 1990.

Naimi-Tajdar, R., Han, C., Sepehrnoori, K., Arbogast, T.J. and Miller, M.A.: “A Fully Implicit, Compositional, Parallel
Simulator for IOR Processes in Fractured Reservoirs”, SPE 100079, SPE /DOE Symposium on Improved Oil Recovery,
Tulsa, Oklahoma, U.S.A., April 22-26, 2006.

Nghiem, L.X., Kohse, B.F. and Sammon, P.H.: "Compositional Simulations of the VAPEX process", JCPT, 54, August,
2001.

Onsager, L.: Ann. N.Y. Acad Sci:, 46, 241, 1945-46.

Othmer, D.F. and Thakar, M.S.: "Correlating Diffusion Coefficients in Liquids", Ind. Eng. Chem., 45(3), pp 589-593,
1953.

Øyno, L., Uleberg, K. and Whitson, C.H.: "Dry gas injection in fractured chalk reservoirs — an experimental approach",
Paper SCA 1995–27, Presented at the International Symposium of the Society of Core Analysts held in San Francisco,
California, 1995. 

Peng, D.Y. and Robinson, D. B.: “A New Two-Constant Equation of State”, Ind. Eng. Chem. Res., 15, pp 59-64, 1976.

Perkins, T.K. and Johnston, O.C.: "A Review of Diffusion and Dispersion in Porous Media", SPE 480, Society of
Petroleum Engineers Journal, January 15, 1963.

Reamer, H.H., Duffy, C.H. and Sage, B.H.: “Diffusion coefficients in hydrocarbon systems: Methane-pentane in liquid
phase”, Ind. Eng. Chem., 48, pp 275-288, 1956.

Reis, R.A., Nobrega, R., Oliveira, J.V. and Tavares, F.W.: "Self and mutual diffusion coefficient equation for pure fluids,
liquid mixtures and polymeric solutions", Chemical Engineering Science, 60(16), pp 4581-4592, 2005.

Reis, R.A., Oliveira, J.V., and Nobrega, R.: "Diffusion Coefficients in Polymer-Solvent Systems for Highly Concentrated
Polymer Solutions", Brazilian J. Chem. Eng., 18(3), pp 221-232, 2001.

Renner, T.A.: "Measurement and correlation of diffusion coefficients for CO2 and rich gas applications", SPE Reserv.
Eng., 3, pp 517– 523, 1988.

Riazi, M.R.: "A new method for experimental measurements of diffusion coefficients in reservoir fluids", Journal of
Petroleum Science and Engineering, 14, pp 235-250, 1996.

Rocha, S.R.P., Oliveira, J.V. and Rajagopal, K.: "An evaluation of density correlations for estimating diffusivities in
liquids and liquid mixtures", Chemical Engineering Science, 52(7), pp 1097-1109, 1997.
30 SPE 134589

Salama, D. And Kantzas, A.: "Experimental Observations of Miscible Displacement of Heavy Oils With Hydrocarbon
Solvents", SPE/PS-CIM-CHOA 97854, SPE International Thermal Operations and Heavy Oil Symposium, Calgary,
Canada, November 1-3, 2005.

Schechter, D.S. and Guo, B.: "Mathematical Modelling of GravityDrainage after Gas Injection into Fractured
Reservoirs", SPE 35170, SPE Improved Oil Recovery Symposium, Tulsa, Oklahoma, April 22-24, 1996.

Schechter, D.S., et al: "Advanced Reservoir Characterization and Evaluation of CO2 Gravity Drainage in the Naturally
Fractured Spraberry Trend Area", Third Annual Techical Progress Report, Contract No. DE-FC22-95BC14942, U.S
DOE, December, 1998.

Schechter, D.S.: "Advanced Reservoir Characterization and Evaluation of CO2 gravity drainage in Spraberry Trend
Area", 5th Annual Technical Progress Report, US DOE Contract No. DE-FC22-95BC14942, 2000

Sheikha, H., Pooladi-Davish, M. and Mehrotra, A.K.: "Development of Graphical Methods for Estimating the Diffusivity
Coefficeint of Gases in Bitumen from Pressure-Decay Data", Energy & Fuels, 19, pp 2041-2049, 2005.

Sigmund, P.M.: “Prediction of molecular diffusion at reservoir conditions, Part I. Measurement and prediction of Binary
dense gas diffusion coefficients”, Canadian Journal of Petroleum Technology, pp 48-57, 1976.

Sridhar, T. and Potter, O.E.: "Predicting Diffusion Coefficients", AIChE J, 23(4), pp 590-592, 1977.

Teja, A.S.: "Correlation and Prediction of Diffusion Coefficients by use of a Generalized Corresponding States Principle",
Ind. Eng. Chem. Fundam, 24(1), pp 39-44, 1985.

Tharanivasan, A.K., Yang, C. and Gu, Y.: "Comparison of three different interface mass transfer models used in the
experimental measurement of solvent diffusivity in heavy oil", Journal of Petroleum Science and Engineering, 44(3-4), pp
269-282, 15 November 2004.

Tharanivasan, A.K., Yang, C. and Gu, Y.: "Measurements of Molecular Diffusion Coefficeints of Carbon Dioxide,
Methane, and Propane in Heavy Oil under Reservoir Conditions", Energy & Fuels, 20, pp 2509-2517, 2006.

Thomas, L.K., Dixon, T.N. and Pierson, R.G.: "Ekofisk Nitrogen Injection", SPE Formation Evaluation, June, pp. 151–
160, 1991.

Torabi, F. and Asghari, K.: "Performance of CO2 Huff-and-Puff Process in Fractured Media (Experimental Results)",
Canadian International Petroleum Conference (58th Annual Technical Meeting), Calgary, Alberta, Canada, June 12-14,
2007.

Trivedi, J. and Babadagli, T.: "Experiment and Numerical Modeling of the Mass Transfer between Rock Matrix and
Fracture", Chemical Eng. J., 146(2), pp 194-204, 2009.

Trivedi, J. and Babadagli, T.: "Efficiency Analysis of Greenhouse Gas Sequestration during Miscible CO2 Injection into
Fractured Oil Reservoirs", Env. Sci. and Tech, 42(15), pp 5473-5479, Aug 1, 2008a.

Trivedi, J. and Babadagli, T.: "Efficiency of Diffusion Controlled Miscible Displacement in Fractured Porous Media",
Transport in Porous Media, 71(3): 379-394, 2008b.

Trivedi, J. and Babadagli, T.: "Scaling Miscible Displacement in Fractured Porous Media Using Dimensionless Groups",
J. Pet. Sci. Eng, 61, pp 58-66, 2008c.
SPE 134589 31

Trivedi, J. J. and Babadagli, T.: "Efficiency of Miscible Displacement in Fractured Porous Media", SPE 100411, SPE
Western Regional/AAPG Pacific Section/GSA, Cordilleran Section Joint Meeting, Anchorage, Alaska, May 8-10, 2006.

Tsay, C-S. and McHugh, A.J.A.: "Technique for rapid measurements of diffusion coefficients", Ind. Eng. Chem. Res., 31,
pp 449-452, 1992.

Uleberg, K. and Høier L.: "Miscible Gas Injection in Fractured Reservoirs", SPE75136, SPE/DOE Improved Oil
Recovery Symposium, Tulsa, Oklahoma, U.S.A., April 13-17, 2002.

Umesi, N.O. and Danner, R.P.: "Predicting Diffusion Coefficients in Nonpolar Solvents", Ind. Eng. Chem. Process Des.
Dev., 20(4), pp 662-665, 1981.

Upreti, S.R. and Mehrotra, A.K.: "Diffusivity of CO2, CH4, C2H6, and N2 in Athabasca bitumen", Can. J. Chem. Eng., 80,
pp 116– 125, 2002.

Upreti, S.R. and Mehrotra, A.K.: "Experimental Measurement of Gas Diffusivity in Bitumen: Results for Carbon
Dioxide", Ind. Eng. Chem. Res., 39, pp 1080-1087, 2000.

Upreti, S.R., Lohi, A., Kapadia, R.A. and El-Haj, R.: “Vapor Extraction of Heavy Oil and Bitumen: A Review”, Energy
& Fuels, 21, pp 1562-1574, 2007.

Vahapcan, E., Babadagli, T. and Xu, Z.: “Pore Scale Investigation of the Matrix-Fracture Interaction During CO2
Injection in Naturally Fractured Oil Reservoirs”, Energy & Fuels, 24, pp 1421-1430, 2010.

Verlaan, M. and Boerrigter, P.: "Miscible Gas/Oil Gravity Drainage", SPE 103990, International Oil Conference and
Exhibition, Cancun, Mexico, August 31-September 2, 2006.

Vrentas, J.S. and Duda, J.L., "Diffusion in Polymer-Solvent Systems II. A Predictive Theory for the Dependence of
Diffusion Coefficients on Temperature, Concentration and Molecular Weight", J. of Polym. Sci., Polymer Physics
Edition, 15, pp 417-439, 1977b.

Vrentas, J.S. and Duda, J.L.: "Diffusion in Polymer-Solvent Systems I. Re-examination of the Free- Volume Theory", J.
of Polym. Sci.", Polymer Physics Edition, 15, pp 403-416, 1977a.

Wang, G.C.: "A Study of Crude Oil Composition during CO2 Extraction Process", SPE 15085, 56th California Regional
Meeting, Oakland, CA, U.S.A, April 2-4, 1986.

Warren, J.E. and Root, P.J.: “The Behavior of Naturally Fractured Reservoirs”, Society of Petroleum Engineers Journal,
228, pp 245-255, September, 1963.

Wen, Y., Kantzas, A. and Wang, G.J.: "Estimation of Diffusion Coefficients in Bitumen Solvent Mixtures Using X-Ray
CAT Scanning and Low Field NMR Spectra", Journal of Canadian Petroleum Technology, 44(4), pp 29-35, 2005.

Wilke, C.R. and Chang, P.: "Correlation of Diffusion Coefficients in Dilute Solutions", AIChE J, 1(2), pp 264-270, 1955.

Wilke, C.R. and Chang, P.: "Estimation of Liquid Diffusion Coefficients", Chem. Eng. Prog., 45(3), pp 218-224, 1949.

Yang, C. and Gu, Y: "A New Method for measuring solvent diffusivity in Heavy Oil by Dynamic Pendant Drop Shape
Analysis, SPE 84202, SPE Annual Technical Conference and Exhibition, Denver, October 5-8, 2003.

Yu, Y.-X. and Gao, G.-H.: "Self-diffusion coefficient equation for polyatomic fluid", Fluid Phase Equilibria, 116, pp 111-
124, 1999.
32 SPE 134589

Zhang, D. and Kang, Q.: “Pore Scale simulation of solute transport in fractured porous media”, Geophysical Research
Letters, 31, L12504, 2004.

Zhang, X. and Shaw, J.M.: "Liquid-phase Mutual Diffusion Coefficients for Heavy Oil + Light Hydrocarbon Mixtures",
Petroleum Science and Technology, 25, pp 773-790, 2007.

Zhang, Y., Hyndman, C.L. and Maini, B.: "Measurement of Gas Diffusivity in Heavy Oils", Journal of Petroleum Science
and Engineering, 25(1-2), pp 37-47, 2000.

You might also like