Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

Nanoscale Gas Flow in Shale Gas Sediments

F. JAVADPOUR, D. FISHER, M. UNSWORTH


Alberta Research Council

But shales can be difficult to evaluate using conventional labora-


Abstract tory techniques. Much of this has to do with resident clays that can
Production of gas out of low permeability shale packages is have bound water either as part of their matrix or loosely bound in
very recent in the Western Canadian Sedimentary Basin (WCSB). the interlayers in amounts of 75 to 80%. Another challenge can be
The process of gas release and production from shale gas sed- the accurate measurement of in situ permeabilities, which are on
iments is not well understood. Because of adsorptive capacity the nanoscale. Core samples have often been subjected to coring-
of certain shale constituents, including organic carbon content, induced or stress release fractures, resulting in greatly overstated
coalbed methane models are sometimes being applied to model permeability measurements.
and simulate tight shale gas production behaviour. Alternatively, While some shales in Western Canada are, at this early stage,
conventional Darcy flow models are sometimes applied to tight showing the proper geochemical and reservoir properties to sup-
shale gas. However, neither of these approaches takes into ac- port gas production, new techniques need to be developed to more
count the differences in transport mechanisms in shale due to ad- accurately understand shale properties and their productive poten-
ditional nanopore networks. Hence, the application of existing tial. Modelling shale gas data to better understand storage mech-
models for shale results in erroneous evaluation and predictions. anisms and diffusive flow would benefit the nascent shale gas
Our analysis shows that a combination of a nanopore network industry in Canada.
connected to a micrometre pore network controls the gas flow in The use of desorption curves through canister desorption testing
shale. Mathematical modelling of gas flow in nanopores is dif- is a standard laboratory procedure, which is an important step in
ficult since the standard assumption of no-slip boundary condi- the analysis of coalbed methane production. The same technique is
tions in the Navier-Stokes equation breaks down at the nanometre currently used for shale gas. The need to establish an estimate of
scale, while the computational times of applicable molecular-dy- gas volume in representative samples is the primary goal of reserve
namics (MD) codes become exorbitant. We found that the gas estimation since the value of a reservoir is related to the amount of
flow in nanopores of the shale can be modelled with a diffu- recoverable gas-in-place.
sive transport regime with a constant diffusion coefficient and Many of the models used for coal assume gas production is
negligible viscous effects. The obtained diffusion coefficient is being controlled by a constant diffusivity coefficient from per-
consistent with the Knudsen diffusivity which supports the slip fectly spherical objects connected by a relatively large fracture
boundary condition at the nanopore surfaces. This model can be network(3). Application of this general model to shale may result in
used for shale gas evaluation and production optimization. the introduction of large errors in the production model, although
in the current shale gas testing environment, the same desorption
tests developed for coal may be used as at least partial indicators of
reserves potential for shales. However, gas production from shale
Introduction is the outcome of several different types of gas release processes,
Shale gas is a type of reservoir classified under the Unconven- including, but not limited to, desorption. In fact, the data sets are
tional Gas heading. These ‘difficult to produce’ reservoirs will play incorrectly referred to as ‘desorption graphs’ and should in fact
an increasingly important role in Canadian gas production because be referred to as ‘gas evolution graphs’ in order to encompass all
they are showing the potential to offset declining conventional gas forms of gas transport mechanisms.
production. We have developed a new descriptive model of flow dynamics
Quite simply, shale gas is natural gas produced from shale se- in structures which have complex pore networks. The idea that the
nanopore networks of shale behave differently than Darcy flow has
quences. Gas shales are predominantly lithified clays with organic
shaped our preliminary observations.
material and detrital minerals present in varying amounts. Organic
matter is an integral constituent of a productive shale gas reservoir.
In addition, these fine-grained rocks are microporous, causing low
permeabilities. Physics of Gas Evolution and Production
While shale gas production has had a long history in the United Understanding the physical aspects of gas production from
States, dating back about 80 years, it is still at the very early stages shale gas sediments is important for reservoir evaluation and pro-
of commercial production in Canada. Very little public data exists duction optimization. Figure 1 shows the spider diagram of the
on shale gas production, yet industry interest is on the rise. A va- shale gas research study. The research is divided into experimental
riety of estimates indicate that between 550 and 860 trillion cubic and theoretical studies. From the experiments, pore size distribu-
feet of gas-in-place could exist in potential shale gas formations in tion and characteristic coefficients such as permeability and dis-
Western Canada(1, 2). persion coefficient can be determined. From theoretical modelling,
Peer Reviewed Paper (“Review and Publication Process” can be found on our Web site)

October 2007, Volume 46, No. 10 55


Wellbore

Production
optimization
(a) Macroscale
Reserve
Rock (Reservoir)
estimation Simulation

Kerogen & Clays


fluid
Deliverability
Fluid flow Mesoscale
Characterization
(b) (micro-fractures
Desorption
network)
Desorption in coefficient Adsorption
nanopores coefficient

Dispersion
Theoretical coefficient
modeling Experiment Microscale
(c)
* (Nanopores
Pore size
network)
* distribution
Flow in nanopores

FIGURE 1: Spider digram of the shale gas research study. The Nanoscale
materials presented in this paper are marked by *.
(d) (gas desorption
from nanopore
walls)
(a) Conventional (b) Shale gas

Molecular scale
(e) (mass transfer from
kerogen/clay bulk to
pore surface

FIGURE 3: Gas evolution and production in shale gas sediments


at different length scales. (a) Macroscale: gas flow from the
FIGURE 2: Comparison between pores in conventional (a) and reservoir to the wellbore, (b) Mesoscale: gas flow in large pores,
unconventional shale gas (b) reservoirs. (c) Microscale: gas flow in nanopores, (d) Nanoscale: gas desoption
from the nanopore walls, and (e) Molecular scale: diffusion of gas
the flow and desorption in nanopores can be studied. Combining molecules in the kerogen/clay.
experiments with the theoretical model results in shale gas sed-
iment characterization for reserve estimation and deliverability (c). During reservoir depletion the thermodynamic equilibrium be-
analysis. Furthermore, the obtained knowledge will be used to de- tween kerogen/clays and the gas phase in the pore spaces changes.
velop/modify reservoir simulators for production optimization. In Hence, gas desorbs from the surface of the kerogen/clays (d). This
this paper, we present the experimentally obtained permeabilities non-equilibrium process further drives the gas molecules to dif-
(pulse decay method) of 152 samples from nine reservoirs with fuse from the bulk of the kerogen to the surface of the kerogen ex-
pore-size distribution of several shale samples at 415 MPa (60,000 posed to the pore network (e). It is well documented that quantities
psi) mercury injection pressure, as well as desorption data for a of gas can be stored either as a dissolved phase in liquid hydrocar-
representative shale sample. For the theoretical part, we present bons, or as an adsorbed phase on other materials within the shales
gas flow in nanopores and describe gas evolution curves using of the kerogen, i.e., certain forms of illite. However for the sake
probability density functions. of simplicity, liberation of gas by these mechanisms is neglected
The phenomena of gas storage and flow in shale gas sediments in this study.
are a combination of different controlling processes. Gas is stored
as compressed gas in pores, as adsorbed gas to the pore walls and as Flow in Micropores
soluble gas in solid organic materials, i.e., kerogen and clays. Gas
flows through a network of pores with different diameters ranging This aspect of the flow in shale gas is similar to conventional
from nanometres (nm = 10-9m) to micrometres (μm = 10-6m)(4). gas reservoirs. The Darcy equation is a suitable equation for these
types of pores, i.e., laminar flow.
Figure 2 compares the pore distribution in conventional and un-
conventional (shale) reservoirs. As shown in this schematic figure,
the number of nanopores is higher in unconventional shale gas sed- Flow in Nanopores
iments. Diameter of pores in shale gas sediments range from a few In general, fluid flow through nanopores is modelled using ei-
nanometres to a few micrometres. ther the continuum or the molecular approach(5, 6). The continuum
In shale gas systems, nanopores play two important roles. First, approach solves for macroscopic fluid properties as a function of
for the same pore volume, the exposed surface area in nanopores the spatial coordinate and is used widely for fluid flow applica-
is larger than in micropores. This is because surface area is propor- tions. As the length scale of a physical system decreases, the va-
tional to 4/d, where d is the pore diameter. This large exposed area lidity of the standard continuum approach with no-slip boundary
permits large volumes of gas desorption from the surface of the conditions diminishes. Figure 4 illustrates the no-slip and slip
kerogen in nanopores. Consequently, higher mass transfer of gas boundary conditions on the inner surface of a pore. Figure 4a illus-
molecules occurs inside the bulk kerogen. Secondly, gas flow in trates the Poisseuille parabolic velocity profile in a pore with zero
nanopores is different from the Darcy flow. This issue will be dis- velocity (no-slip) on the walls. Figure 4b illustrates the slip con-
cussed in detail in the following section. dition, where gas molecules can freely slip on the surface of the
Figure 3(a-e) illustrates the sequence of gas production at dif- pores and collide with the wall or other gas molecules. A widely-
ferent length scales. Production from a new drilled wellbore (a) is recognized dimensionless parameter that determines the degree of
happening through the large pores (b), and then the smaller pores appropriateness of the continuum model is the Knudsen number(5,
56 Journal of Canadian Petroleum Technology
Table 2: Flow regime condition based on Knudsen
(a) number.
U=0
Navier-Stokes Equation
No-slip (Kn < 0.001) Slip (0.001 < Kn < 0.1)

Mm
U = Umax
Continuum flow Slip flow
Darcy flow Knudsen Diffusion
U=0

10
(b)

Transition
flow
1

0.1

Slip flow
10 nm
0.01
50 nm

Kn
0.001
300 nm

No-slip
nm

flow
1 Mm
0.0001
10 Mm
0.00001
50 Mm
FIGURE 4: Comparison of gas flow (a) in micropores where the
0.000001
flow is no-slip and (b) in nanopores where the flow is slip. 100 1,000 10,000 100,000
P (kpa)
Table 1: Mole fraction, molar mass, and collision
diameter δ for each component in typical shale gas FIGURE 6: Knudsen number for the gas mixture as a function of
sediment gas sample. pressure for different pore sizes at 350 K.

Gas Mole Collision Diameter Molar Mass Equation (2) is not very accurate at very high pressures. That is
Components (%) (δ, nm) (kg/kmol) why it is predicting a mean-free-path smaller than the collision di-
CH4 87.4 0.4 16.0 ameter at very high pressures.
C2H6 0.12 0.52 30.0 The Knudsen number is a measure of the degree of rarefaction
CO2 12.48 0.45 44.0 of gases encountered in gas flow through small pores. Table 2 de-
average 0.41 19.5 scribes different regimes of fluid flow depending on the Knudsen
number(7). At a low Knudsen number (Kn < 0.001) the no-slip
boundary condition in the continuum flow regime is valid. At a
80 high Knudsen number regime (0.001 < Kn < 0.1), the rarefaction ef-
fects become more pronounced and eventually the continuum ap-
proach breaks down. Figure 6 presents the Knudsen number as a
60 function of pressure for different pore sizes ranging from 10 nm to
350 K
50 μm. The Knudsen number decreases by increasing the pressure,
and also it is smaller at larger pores.
(nm)

40 An accurate approach to describe a high Kn flow regime is the


400 K molecular model(5), which recognizes the fluid as a swarm of dis-
crete particles. In this model, the position, inertia and state of all
20 individual particles are calculated either deterministically(7) or
probabilistically(9) at all times. The molecular approach considers
300 K individual particle dynamics based on a Boltzmann distribution at
0 the temperature of interest. The interaction between two particles
100 1,000 10,000 100,000 is given by the two-body potential energy and transient evaluation
P (kpa) of particle positions which are determined by numerical integra-
FIGURE 5: Average mean-free-path λ for the gas mixture as
tion of Newton’s equation of motion. In spite of the theoretical ap-
function of pressure and temperature. propriateness of molecular-dynamics (MD) models for the purpose
of capturing the small-scale interactions, it is unrealistic to simu-
late a reasonably practical nanoscale network in shale gas sediment
7).It is defined as the ratio of the gas mean-free-path λ and the pore
using present-day supercomputers. Most MD calculations are re-
diameter d.
stricted to femtoseconds (10-15 s) time steps limiting the results to a
short time-scale such as picoseconds (10-12 s) to nanoseconds (10-9
λ
Kn = s) phenomena(5, 10).
d ................................................................................................... (1)

Where λ is defined as(8):


Alternative Approach
An alternative approach for gas flow in nanopores is to describe
k BT the flow consistent with a diffusive transport regime with a con-
λ=
2 πδ2 P ........................................................................................... (2) stant diffusion coefficient and negligible viscous effects. Mathe-
matically, this is expressed as(10):
in which kB is the Boltzmann constant (1.3805 × 10-23J/K), T is
qρM ∆P
temperature (K), P is pressure (Pa) and δ is the collision diameter = Dk
of the gas molecule. Table 1 presents the gas properties of a typ- φA RTL
...................................................................................... (3)
ical shale gas sediment. In Figure 5, the average mean-free-path of
the gas presented in Table 1 is plotted versus pressure at three dif- where q is the flow rate, φ is porosity, A is the exposed area, ∆P
ferent temperatures. The average mean-free-path of the gas mix- is the pressure drop across the porous bed, L is the thickness, R is
ture decreases by increasing pressure or decreasing temperature. the universal gas constant, ρ is density, M is molar mass and T is
October 2007, Volume 46, No. 10 57
(a) (a)
50
106
N = 152 samples from 9 reservoirs
1 2
40 105
Frequency

3
30 104 5 4

Pc (kPa)
103
20 1: Shale I (~50 nD)
2: Shale II (~300 nD)
102 3: Shale III (~700 nD)
10 4: Crossfield Cardium Sand (230 µD)
Mode = 54 nD 101 5: Indiana Limestone (4,000 µD)

0
0 50 100 150 200 250 300 350 400 100
100 90 80 70 60 50 40 30 20 10 0
Permeability (nd) SHg (%)
(b)
(b)
100
100
Cumulative Frequency (%)

90
80 90 2
80
70
70
60 3
uSHg 60
50 1
u(lnPc) 50
40 5
40
30
30
20 20 4
10 10
0 0
0 50 100 150 200 250 300 350 400
10-3 2 10-2 2 10-1 2 100 2 101 2 102
Permeability (nd) Pore Throat Diameter ( m)

FIGURE 7: Frequency versus permeability of 152 shale gas samples FIGURE 8: (a) capillary pressure versus mercury saturation for
from nine reservoirs. (a) permeability distribution, (b) cumulative 3 shale samples, one sand sample and one limestone sample. (b)
frequency distribution. the pseudo-flow determined from the capillary pressure plots to
demonstrate dominant pore sizes to flow.
temperature. The expression Dk is the Knudsen diffusion coeffi-
cient given as: permeabilities are in nanoDarcies. Figure 7(b) presents the cu-
mulative frequency of the permeability data. From this figure, we
d 8 RT observed that 90% of the measured permeabilities are less than
Dk = 150 nD.
3 πM ........................................................................................ (4)

where d is the diameter of the nanopore. The derivation of Equa-


Extremely High Pressure Mercury (Hg)
tion (4) is presented in Appendix A. Injection
To obtain pore-size distribution of the core samples, traditional
Desorption mercury injections to obtain capillary pressure measurements were
made. The apparatus operates at extremely high pressures (up to
Gas molecules desorb from the surface of the kerogen to the
415 MPa or 60,000 psi). The results are plotted in Figure 8 for
pore space and then carry on due to pressure depletion. Desorption
three shale samples. The data of sandstone and limestone samples
in shale gas will be addressed in detail in a subsequent work.
are also plotted for comparison. As shown in Figure 8(a), capil-
lary pressure of the three shale samples (marked by 1, 2 and 3)
Diffusion in Kerogen increases rapidly by injecting mercury into the samples. For sand-
Once the gas molecules from the surface of the kerogen desorb stone (4) and limestone (5), the capillary pressure is lower. The de-
to the pore space, the gas concentration difference between the rivative of the mercury saturation SHg with respect to the natural
bulk of the kerogen and the surface of the kerogen controls gas dif- log of capillary pressure ln(Pc) is plotted versus pore throat diam-
fusion in solid kerogen. eters obtained from the Laplace equation in Figure 8(b). This de-
scribes the flow of mercury in the pores. As shown in Figure 8(b),
for the shale samples (1, 2 and 3), pore diameters in the range of 4
nm to 200 nm are the dominant flow passages, while for the sand-
Experiments stone and limestone, much larger pores contribute to flow.
The results of three sets of experiments are presented in this
section: pulse decay experiments used to measure permeability, Gas Evolution and Desorption Tests
extremely high pressure mercury injection to find pore-size dis-
tribution and gas evolution experiments from selected shale The general method of measuring the evolution of gas is based
samples. on the work of the Gas Research Institute(3) in their study of
coalbed methane. It can be summarized as follows: as soon as the
Pulse Decay Permeability core starts its trip up the wellbore, the start time is recorded; once
the core has been removed from the core barrel, it is examined and
Pulse decay is a technique used to measure permeability of cut into sections. Those sections where the on site geologist de-
tight rocks and synthetic materials(11, 12). Figure 7 presents per- cides good prospects lie, are then weighed and placed in plastic
meability data for 152 shale samples from nine reservoirs. Figure bags, and then placed in a standard core desorption canister. These
7(a) is the frequency versus permeability plot, where the measured canisters have a pressure gage and a valve on the top lid, and are
58 Journal of Canadian Petroleum Technology
closed with an o-ring seal. Once the canister is sealed, it is placed deviation (σ) is determined for each curve and converted to diffu-
in a water bath at reservoir temperature. At this point, gas volume sion coefficient (D):
measurements can begin. A schematic diagram of the set up is il-
lustrated in Figure 9.
(σVi )
2

These measurements are done in small time increments as long D=


2t ............................................................................................. (6)
as the gas production volumes are large, but these increments are
lengthened as the produced volume decreases. The first set of high
rate measurements are done in the field but once the rate slows where Vi is the interstitial velocity and t is the time at the mean.
down, the canisters are placed in an ice bath and returned to the In Figure 10(b), data points are labeled as ‘data.’ Using Equation
laboratory where the samples are transferred to the water baths (5), four different gas transport processes can be extracted from the
again. The gas volume readings are continued for several months overall gas evolution curve. Process A, as shown in Figure 10(b),
until the gas volume change becomes too small to measure. At corresponds to the gas flow in micropores which happens at the
each measurement of the gas volume in the manometre, the atmo- early times of gas flow from the sample. This can be described by
spheric pressure and ambient temperature are recorded so that the Fickian diffusion or sometimes by a Darcy equation depending on
final gas volumes can be converted to standard temperature and the original pressure in the system. Process B, in Figure 10(b), is
pressure (STP). the gas flow in nanopores where a Knudsen diffusion is valid. Pro-
In Figure 10, we show plots for coalbed methane and the shale cess C describes gas desorption from the surface of the kerogen/
gas sample. In Figure 10(a), the gas evolution of coalbed methane clays to the pore networks. Finally, Process D describes the gas
is presented. The solid line is the model prediction based on the molecule diffusion from the kerogen bulk or clays to the exposed
spherical diffusion model. Figure 10(b) presents our shale sample surfaces. This process is very slow. Note that all these processes
measurements for a typical cumulative gas volume versus time. overlap each other.
The gas evolution is clearly different from the coalbed methane The extracted data from Figure 10(b), suggests that diffusion
evolution curve. Hence, the spherical diffusion model might not be coefficient in Process B (gas flow in the nanopore) is 4 × 10-7 m2/s,
accurate for the shale gas samples. which is consistent with the value of 3.8 × 10-7 m2/s (35 nm pore
at 375 K) obtained from the analysis of the theoretical model as
Different gas production rates are evident in Figure 10(b). These
presented in Equation (4). Also, the determined diffusion coeffi-
different gas production rates are the basis of our analysis. Each
cient for the gas molecule transport in the kerogen bulk is 2 × 10-10
section of the curve represents a specified type of gas flow with
m2/s which is consistent with the measurement of the gas diffusion
overlaps of each other. The next section explains how these sec-
in bitumen(14). Assuming that kerogen and bitumen are alike, this
tions are determined.
finding supports our preliminary approach of analysis.

Mathematical Modelling Conclusions


Assuming that gas flow behaviour in shale gas sediment is a The gas flow behaviour in shale gas sediments is studied with an
combination of all the above mentioned processes, i.e., gas diffu- emphasis on nanopore flow.
sion in kerogen/clays, desorption and flow in nanopores and mi-
cropores, then a group of Gaussian distribution curves can be used
to cover all these events(13) (see Appendix B for details): (a)

 2 m −x 1,200
n
( )  + C
∑ Fx = ∑ 0.5 ⋅ 2 ⋅ π ⋅hi ⋅ σ i ⋅ erf 

t
2σ i  1,000
Spherical model
i=1  
Gas Volume (cc)

............................... (5)
800

where, σ is Gaussian standard deviation, m is Gaussian mean, x is


600
an independent variable and h is a scaling factor.
The results of the analysis are presented in Figure 10(b). Each 400
process takes a Gaussian distribution function. Gaussian standard
200
Coalbed methane

Valve 0
100 2 101 2 102 2 103 2 104 2 105 2 106 2
Time (hrs)

Manometer (b)
7,000
Canister
6,000 data
5,000
Gas Volume (cc)

4,000 D
Sample
3,000

2,000
C
1,000 B
0

-1,000 A Shale gas


-2,000
0 10 20 30 40 50 60

Time (hrs)
Water Bath
FIGURE 10: (a) Gas evolution for a coalbed methane sample with
FIGURE 9: Schematic diagram of the setup for gas evolution the spherical model fit, (b) Gas evolution and production from a
measurement. shale gas sample with the Gaussian standard deviation plots.

October 2007, Volume 46, No. 10 59


• A considerably large number of pulse decay permeability 9. HEER, C.V., Statistical Mechanics, Kinetic Theory, and Stochastic
measurements suggest an average of 54 nD permeability for Processes; Academic Press, New York, NY, 1972.
152 samples from nine selected reservoirs. 10. ROY, S., RAJU, R., CHUANG, H.F., CRUDEN, B.A. and MEYY-
• Extremely high pressure mercury injection tests revealed the APPAN, M., Modelling Gas Flow Through Microchannels and
dominance of the nanopores in shale samples. Nanopores; Journal of Applied Physics, Vol. 93, No. 8, pp. 4870-
4879, 2003.
• A method based on Gaussian distribution curves is proposed
to determine different flow behaviour in the shale gas flow. 11. FINSTERLE, S. and PERSOFF, P., Determining Permeability of
Tight Rock Samples Using Inverse Modelling; Water Resources Re-
• Our analysis showed that nanopores in shale gas sediments search, Vol. 33, No. 8, pp. 1803-1811, 1997.
contribute to the gas evolution processes. Gas flow in nano-
12. LIANG, Y., PRICE, J.D., WARK, D.A. and WATSON, E.B., Non-
pores deviates from the traditional flow models, i.e., Fick’s
linear Pressure Diffusion in a Porous Medium: Approximate Solutions
law or Darcy’s equation. An appropriate theoretical model of With Applications to Permeability Measurements Using Transient
gas flow in nanopores is presented. Pulse Decay Method; Journal of Geophysical Research, Vol. 106, No.
B1, pp. 529-536, 2001.
13. von SEGGERN, D.H., CRC Handbook of Mathematical Curves
Acknowledgements and Surfaces; CRC Press, Boca Raton, FL, 1990.
14. FISHER, D.B., SINGHAL, A.K., DAS, S.K., GOLDMAN, J. and
We would like to thank Ms. B. Jeary, Conventional Oil and JACKSON, C., Use of Magnetic Resonance Imaging and Advanced
Natural Gas (CONG) business unit manager, for her support and Image Analysis as a Tool to Extract Mass Transfer Information from
encouragement. a 2D Physical Model of The Vapex Process; paper SPE 59330 pre-
sented at the SPE/DOE Improved Oil Recovery Symposium, Tulsa,
Nomenclature OK, 3-5 April 2000.

A = surface area, m2
c = mean molecular velocity, m/s
d = pore diameter, m APPENDIX A: Derivation of Knudsen
D = diffusion coefficient, m2/s Diffusion Coefficient in a Nanopore
Dk = Knudsen diffusion coefficient, m2/s
Jk = free molecule flux, molecules/m3 Consider a gas with a molecular density of ρN (molecules/m3) at
one side of a pore and a vacuum at the other side. The free-mole-
kB = Boltzmann’s constant, J/K
cule flux Jk (molecules/m2) through the pore is given by(5, 7):
Kn = Knudsen number, dimensionless
L = thickness, m
m = Gaussian mean, hr J k = αc ρ N
.......................................................................................... (A.1)
M = molar mass, mol/g
P = pressure, Pa where, α is a dimensionless probability factor and c is the mean
q = flow rate, m3/s molecular speed (m/s).
R = universal gas constant, J/molK If there is gas on both sides of the pore, the net flux is propor-
t = time, hr tional to the difference in gas number densities at the two sides:
T = temperature, K
Vi = interstitial velocity, m/s
(
J k = αc ρ N 1 − ρ N 2 ) ............................................................................ (A.2)
Greek Letters
ρ = density, kg/m3 where one represents the side of the tube with denser gas. In order
ρN = number density, molecule/m3 to use this equation, expressions are required for the mean molec-
α = dimensionless probability factor ular speed and the dimensionless probability factor. The mean mo-
σ = Gaussian standard Deviation, hr lecular speed is readily calculated using kinetic theory(8) as:
δ = collision diameter of gas molecules, nm
λ = mean-free-path, nm c=
8 RT
φ = porosity, fraction πM .......................................................................................... (A.3)

REFERENCES where R is the universal gas constant (J/kmol K), T the tempera-
1. Petroleum Technology Alliance Canada (PTAC), ture (K) and M is the molar mass of the gas (kg/kmol). Calculation
Filling the Gap: Unconventional Gas Technology Roadmap; PTAC of the probability factor α is considerably complicated, requiring
Knowledge Centre, Calgary, AB, 2006. knowledge of the pore geometry. For a long straight circular tube of
2. Canadian Society for Unconventional Gas (CSUG), diameter d and length L (L>>d) α is given by (d/3L). The method
Shale Gas; http://csug.ca/faqs.html#Sa. of derivation of this expression is presented elsewhere(7).
3. MAVOR, M. and NELSON, C.R., Coalbed Reservoir Gas-In-Place
The widely used parallel pore model of porous media uses the
Analysis; Gas Research Institute Report No. GRI-97/0263, Gas Re-
search Institute, Chicago, IL, 1977. latter of the above simple geometry. Development of this model
4. KATSUBE, T.J., Shale Permeability and Pore-Structure Evolution is as follows. Substituting α = (d/3L) and c as given by Equation
Characteristics; Current Research 2000-E15, Geological Survey of (A.3) in Equation (A.2) gives:
Canada, Ottawa, ON, 2000.
5. BIRD, G.A., Molecular Gas Dynamics and the Direct Simulation of
Gas Flows; Oxford University Press, Oxford, UK, 1994. Jk =
d 8 RT
ρ −ρ ( )
3L πM N 1 N 2 ................................................................ (A.4)
6. KOPLIK, J. and BANAVAR, J.R., Continuum Deductions from Mo-
lecular Hydrodynamics; Annual Review of Fluid Mechanics, Vol. 27,
pp. 257-292, 1995. which can be re-written in differential form (applied along the axis
7. HADJICONSTANTINOU, N.G., The Limits of Navier-Stokes of the cylindrical pore, z) and applying the convention that flux
Theory and Kinetic Extensions for Describing Small-Scale Gaseous
Hydrodynamics; Physics of Fluids, Vol. 18, No. 11, pp. 111301-
moves from high to low concentration as:
111320, November 2006.
8. HEIDEMANN, R.A., JEJE, A.A. and MOHTADI, F., An Introduc- – d 8 RT  dρ 
tion to the Properties of Fluids and Solids; University of Calgary Jk =  N 
3 πM  dz  ....................................................................... (A.5)
Press, Calgary, AB, 1984.
60 Journal of Canadian Petroleum Technology
Considering the flux Jk to be in mole, not molecules, and con- This function is most valuable in the summation form where we
sidering the rate of change of gas concentration C (mol/m3) rather can convolute multiple distribution functions.
than molecular concentration, gives:
 2 m −x
n
( )  + C
−d 8 RT  dC  ∑ Fx = ∑ 0.5 ⋅ 2 ⋅ π ⋅hi ⋅ σ i ⋅ erf 

i
2σ i 
Jk =   i=1  
3 πM  dz  ............................(B.4)
.......................................................................... (A.6)
where the subscript i represents the parameters described above
Hence, by analogy to continuum gas diffusion, Knudsen diffu- needed to define each dispersion curve and C is the constant of
sion coefficient Dk for flow in a long straight pore is obtained as: integration.

d 8 RT
Dk =
3 πM .................................................................................... (A.7)
Provenance—Original unsolicited manuscript, Nano-Scale Gas Flow in
Shale Gas Sediments (2007-10-14). Abstract submitted for review June
Therefore, the Knudsen diffusion coefficient is propor- 5, 2007; editorial comments sent to the author(s) August 2, 2007; revised
tional to the pore diameter and the mean molecular velocity or manuscript received August 17, 2007; paper approved for pre-press August
temperature. 17, 2007; final approval September 7, 2007.

Knudsen Diffusion Coefficient in Porous


Media
Authors’ Biographies
Due to the complexity of the geometry of most porous media,
Knudsen diffusion coefficients are typically calculated from ex- Farzam Javadpour received his B.Sc. with
perimental data. If experimental data is not available, a model distinction in petroleum engineering from
must be selected. The approach is to consider porous media as con- the Petroleum University of Technology
sisting of a certain percentage of open pores (porosity) of the mean (Iran), and M.Sc. and Ph.D. degrees in
pore diameter and having a degree of interconnection resulting in chemical and petroleum engineering from
paths longer than a direct path (tortuosity). Hence, Knudsen flow the University of Calgary. He served on the
through porous media can be modelled using a modified form of Technical Committee of the Canadian Inter-
Equation (A.7). national Petroleum Conference (CIPC) in
2004 and 2005. He serves as a member of
φ the Editorial Review Board for the Journal
Dk , pm = D of Canadian Petroleum Technology (JCPT) and is Issue Coordi-
τ k ...................................................................................... (A.8)
nator of the August issue. He is currently working as Research Sci-
entist/Engineer at the Alberta Research Council on projects related
where, φ/τ is the porosity-tortuosity factor. to enhanced oil and gas recovery, shale gas, coalbed methane and
CO2 sequestration. He is also an instructor at the University of Cal-
gary. He worked as a Reservoir Engineer for the National Iranian
APPENDIX B: Determination of Diffusion Oil Company for six years. He has presented and published over 30
technical papers on topics related to reservoir engineering. Farzam
Coefficient from Gas Evolution Data is an invited speaker on the subjects related to transport in porous
The modelling of the gas evolution curves uses a class of mathe- media in Europe and the United States.
matical models based on the characteristics of the probability den-
sity function described by the Gaussian equation(13). Douglas Fisher started his career at the Pe-
troleum Recovery Institute (PRI) in 1976
 2
and joined the Alberta Research Council
(
− x − m ) 
() 1 (ARC) in 2000. It was at PRI that he, along
f x = ⋅ exp   with Selim Sayegh, developed one of the
2π ⋅ σ  2σ
2
 early high-pressure micromodel systems.
.........................................................(B.1)
Over this period, he has acquired expertise
where, σ is Gaussian standard deviation, m is Gaussian mean and x in image analysis and data mining methods
is the independent variable. In a general form, Equation (B.1) can as applied to general research problems. At
be written as: ARC, he focuses on the development of
new methods of image analysis to extract
 2 information from experimental systems such as the new Micro
− x − m( ) 
()
Slim Tube and other related high-pressure micromodels.
f x = h ⋅ exp  
2σ 2
 
Michelle Unsworth received her B.Sc. in
..................................................................(B.2)
geology, as well as a Masters in business
The term h allows this function to represent the distribution administration from the University of Al-
without it being normalized to have an area of one. This is impor- berta. She is currently working at the Al-
tant in physical modelling since, in general, we do not have data berta Research Council as Team Lead of the
sets where the integral is unity. Unconventional Gas Research Group. Her
If we assume that the flow properties of the sample are repre- activities are commercially focused on pro-
sented by pore structures and diffusion mechanisms that can be de- viding oil and gas exploration and devel-
scribed as a group of Gaussian distributions, then we can convert opment companies with accurate, reliable
this form of the equation into its cumulative form. data on natural gas contents of both coalbed
methane, as well as shale gas samples. The
Alberta Research Council has been concentrating its efforts on de-
 2 m −x
( )  + C
∫ f ( x ) dx = 0.5 ⋅
veloping better methods to measure gas evolution, evaluate gas
Fx = 2 ⋅ π ⋅hi ⋅ σ i ⋅ erf  i
 2σ i  composition, determine maturity profiles and delineate between
  different types of gas transport out of highly microporous shales.
.................(B.3)
October 2007, Volume 46, No. 10 61

You might also like