Gas Permeability of Shale
A. Sakhaee-Pour and Steven L. Bryant, The University of Texas at Austin
Summary
Permeability is one of the most fundamental properties of any res-
ervoir rock required for modeling hydrocarbon production. How-
ever, shale permeability has not yet been understood fully because
of the complexities involved in modeling flow through nanoscale
throats. In this paper, we analyze the effects of adsorbed layers of
methane (CH4) and of gas slippage at pore walls on the flow
behavior in individual conduits of simple geometry and in net-
works of such conduits. The network is based on scanning elecron
microscope (SEM) images and a drainage experiment in shale. To
represent the effect of adsorbed gas, the effective size of each
throat in the network depends on the pressure. The hydraulic con-
ductance of each throat is determined on the basis of the Knudsen-
number (Kn) criterion (Knudsen 1909). The combined effects of
adsorption and slip depend strongly on pressure and on conduit di-
ameter. The results indicate that laboratory measurements made
with N2 at ambient temperature and 5-MPa pressure, which is typi-
cal for a transient pulse-decay (TPD) method, overestimate the gas
permeability at early life of production by a factor of four. This ra-
tio increases if the measurement is run at ambient condition
because the low pressure enhances the slippage and reduces the
thickness of the adsorbed layer. Moreover, the permeability
increases nonlinearly as the in-situ pressure decreases during pro-
duction. This effect contributes to mitigating the decline in the pro-
duction rate of shale-gas wells. Laboratory data available in the
literature for CH4 permeability at pressures below 7 MPa agree
with model predictions of the effect of pressure.
Introduction
Shale has provoked a great deal of research recently because of
the considerable volume of natural gas stored in reservoirs. The
natural gas produced from these reservoirs is called shale gas and
is envisaged to provide a substantial fraction of US gas produc-
tion, with some reports of as much as half by 2020 (Polczer
2009). However, debate continues about how much gas can be
produced from these types of reservoirs (Urbina 2011).
An accurate measurement of shale permeability is challenging
because it is so small, on the order of 10–21 m2 (nanodarcies). The
challenge appears in both experimental and theoretical investiga-
tions. For instance, constant-pressure steady-state-flow measure-
ment is not used for shale because it requires a considerable time
(Mallon and Swarbrick 2008). Consequently, other approaches
such as TPD (Billiotte et al. 2008; Cui et al. 2009) and crushed
rock (Luffel et al. 1993) are commonly employed. These methods
reduce the time required for measurement by changing the bound-
ary condition and sample size. The test is altered from steady state
to transient in TPD by changing the constant-pressure boundary
conditions to pressures that vary with time. Therefore, it is con-
ducted more quickly because sustained flow at steady state is not
attempted. In the crushed-rock method, the rock is broken into
small pieces so the transient flow of gas out of the pieces is more
rapid.
Studies of gas flow through shale have invoked a different set
of parameters. For example, the gas evacuated from the core was
modeled in terms of conventional permeability and desorption
(Javadpour et al. 2007). Because many of the voids in shale are
Copyright V
C 2012 Society of Petroleum Engineers
This paper (SPE 146944) was accepted for presentation at the SPE Annual Technical
Conference and Exhibition, Denver, 30 October–2 November 2011, and revised for
publication. Original manuscript received for review 28 June 2011. Revised manuscript
received for review 17 January 2012. Paper peer approved 29 March 2012.
August 2012 SPE Reservoir Evaluation & Engineering
within organic material, methane molecules can adsorb to an
appreciable extent. For instance, the influence of the adsorbed
layer on the void space was considered and the adsorbed porosity
was introduced (Cui et al. 2009). In addition, the volume of
adsorbed gas was estimated using reservoir simulation (Cipolla
et al. 2010). The volume and producibility of adsorbed methane
have direct implications for reserves and, thus, have been widely
discussed. However, the impact of an adsorbed layer on perme-
ability has been less widely explored. Yet because the voids are
so small, the adsorbed layer can also affect the transport proper-
ties of the shale.
This paper analyzes the dependency of the in-situ shale-matrix
permeability on reservoir pressure. In this regard, we implement
the effect of an adsorbed layer in simple conduit geometries as a
function of gas pressure. Subsequently, we adopt noncontinuum
flow models to account for nonzero slip velocity in the conduits.
The presence of an adsorbed layer of gas molecules has a geomet-
ric implication (it reduces the cross section available for transport)
and an influence on microscale boundary condition (slippage of
gas being transported through the pore). Here, we examine a mac-
roscopic consequence of these phenomena—namely, the effect on
the permeability of the rock to gas—by implementing these phe-
nomena in a network model of the void space. The effect of
adsorption and slippage depends on the size of the conduit, and
though a network is not a literal model of shale pore space, it does
serve as a convenient vehicle for examining the effect of a distri-
bution of pore sizes. The results have implications for interpreting
measurements of shale permeability to gas at laboratory condi-
tions and for interpreting flow rates from production wells over
time. The flow behavior from a rock mass into a production well
is controlled by resistance of flow paths that are combinations of
matrices and fractures. Thus, for operational decisions, the effect
of pressure on matrix permeability should be incorporated with
natural- and induced-fracture dependency on the pressure (Philip
et al. 2005).
We refer to laboratory and field conditions extensively in the
following discussion, and for the sake of clarity, we define them
here. The temperature is equal to 300 and 360 K at the laboratory
and field conditions, respectively. The pressure at the laboratory
condition is presumed to be 5 MPa, which is a common pressure
for permeability measurement using TPD (Billiote et al. 2008).
The gas permeability of shale at this condition is on the order of
10 nanodarcies. We assume the thickness of the adsorbed layer is
negligible at laboratory condition. This is a plausible assumption
because the ratio of no-slip hydraulic conductance without an
adsorbed layer to no-slip hydraulic conductance with an adsorbed
layer at 5 MPa is close to unity, regardless of throat size, as we
will show later in the paper. For estimating the gas conductance
and permeability, we further suppose that gases at the laboratory
and in-situ conditions are N2 and CH4, respectively. For conven-
ience, we use the terms “liquid conductance” and “liquid perme-
ability” to refer to hydraulic conductance computed with a
continuum model and no-slip boundary condition at pore walls.
We use the terms “gas conductance” and “gas permeability” to
refer to the conductance when slippage is taken into account.
Both “gas” and “liquid” conductance can be computed when an
adsorbed layer is present. We denote the gas and liquid perme-
abilities at the laboratory condition by kg,lab and kl,lab, respec-
tively. The pressure varies at the field condition, controlling the
thickness of the adsorbed layer. The pressure also affects the Kn
and, hence, the slippage. For the field condition, the gas and liq-
uid permeabilities are represented by kg,in-situ and kl,in-situ,
respectively.
401

TABLE 1—FLUID-FLOW REGIMES DEFINED BY
RANGES OF Kn
Kn 0–10–3 10–3–10–1 10–1–101
Flow regime Continuum Slip Transition Free-molecular
Gas-Flow Regimes
Kn (Knudsen 1909) is used to differentiate flow regimes in con-
duits at micro- and nanoscale. This nondimensional parameter is
defined as
k assumption, the gas permeability relative to the no-slip permeabil-
Kn ¼ ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð1Þ
K
where k denotes the mean free path of gas molecules and K the
characteristic length of the channel. The flow regime changes
from continuum model to discrete particles as the Kn increases.
Substituting the mean free path yields the following relation (Roy
et al. 2003):
rffiffiffiffiffiffiffiffiffi
p l
Kn ¼  ; . . . . . . . . . . . . . . . . . . . . . . . . . ð2Þ
2RT qK
where l is the gas viscosity, q the density, T the prevailing tem-
perature, and R the specific gas constant. The specific gas constant
is the ratio of universal gas constant R to the molar mass m. On
the other hand, the length scale K is equal to the diameter and
aperture height in the circular tube and slit, respectively. For N2
and CH4 at the range of temperatures (300 to 360 K) and pres-
sures (5 to 28 MPa) of interest here, the range of values of k is 0.3
to 2.2 nm. For pore sizes K in the range of 3 to 30 nm, we thus en-
counter values for Kn in the range of 10–2 to 100.
Continuum Regime. The flow regime is categorized on the basis
of Kn, as summarized in Table 1 (Roy et al. 2003). This table indi-
cates that the Navier-Stokes equations with no-slip boundary con-
dition are appropriate only for small Kn (Kn < 10–3). In other
words, the classical continuum-model assumptions (momentum
transfer by means of bulk phase viscosity, fluid velocity matches
solid velocity at walls) cannot always provide accurate results.
Slip Flow Regime. The ratio of molecule/wall to molecule/mole-
cule collisions increases as Kn increases. However, the molecule/
molecule interaction is still dominant in the slip flow regime.
Therefore, Navier-Stokes equations were suggested (Karniadakis
et al. 2005) to remain valid for this domain (103< Kn < 101).
A first-order slip-boundary condition was applied to include the
effect of molecule/wall collisions (Roy et al. 2003). After imple-
menting the tangential momentum coefficient as shown in Appen-
dix A, the gas permeability can be estimated as
kg ¼ kl ð1 þ a1 Kn Þ; . . . . . . . . . . . . . . . . . . . . . . . . . . ð3Þ
where kl is the permeability of the conduit to liquid (when the no-
slip boundary condition applies), kg the gas permeability of the
conduit, and a1 the permeability enhancement. The liquid perme-
ability is assumed constant (i.e., independent of pressure) and esti-
mated on the basis of the characteristic length size of the throat
from conventional theory. The coefficient a1 depends on the ge-
ometry of the conduit. It is approximately 5 for circular cross sec-
tions (see Appendix A).
The slip flow regime can also be simulated using the dusty gas
model (DGM) (Mason and Malinauskas 1983). This model adopts
a linear combination of gas-transport mechanisms to predict the
overall flow rate. It was mainly based on empirical observation
(Graham 1833, 1846, 1849, 1863), and the theoretical explanation
was proposed later (Mason and Malinauskas 1983). The total
mass flux of a single-size conduit is calculated as
402
kl
Jtotal ¼ Jvisc þ JKn ¼  nrp  DKn rn
l
 
kl rp
¼ p þ DKn ;
>101 l RT                   ð4Þ
where rp is the pressure gradient and DKn is the Knudsen diffu-
sivity coefficient. This relation can also be written in terms of gas
permeability, as
 
kl rp kg p rp
Jtotal ¼  p þ DKn ¼ ; . . . . . . . . . . . ð5Þ
l RT l RT
where kg is the single-phase gas permeability. By substituting the
Knudsen diffusivity coefficient and implementing the ideal-gas
ity in terms of Kn is yielded
 
kg DK l
¼ 1þ n ¼ 1 þ k1 Kn ; . . . . . . . . . . . . . . . . . ð6Þ
kl kl p
where k1 is the permeability enhancement equal to 13.58 for the
circular cross section, as clarified in Appendix A. This relation
indicates that the first-order models (Eq. 3) and the DGM (Eq. 6)
suggest a similar trend for the gas permeability (i.e., a linear func-
tion of Kn).
Eq. 6 is the analog of Klinkenberg’s correction (Klinkenberg
1941) for a single conduit. In the Klinkenberg correction, the perme-
ability to gas is evaluated at several pressures for a core sample
instead of a single conduit. For steady-flow experiments, the pres-
sure is replaced by the average of the inlet and outlet pressures across
the core. The term DKknl l is determined empirically as the slope of the
measured gas permeabilities plotted against the reciprocal pressure.
While the first-order slip model and the DGM indicate a quali-
tatively similar trend for the gas permeability, the rate of change
with Kn is different. The high Kn is attained by decreasing the
throat size in the first-order slip model. In this case, molecule/mol-
ecule collisions are not necessarily infrequent; rather, they are less
frequent than molecule/wall collisions. On the other hand, the low
density of the gas is the reason for the slippage in the DGM. Note
that reductions in both the throat size and the density increase Kn
(see Eq. 2). Consequently, the permeability enhancements are
obtained by measuring the flow rates. The in-situ condition for gas
shale corresponds to the large gas density (approximately 100 kg/
m3). Kn in this situation is large because shale conduits are very
small. Thus, in this study, we use the first-order slip model in the
range of 103< Kn < 101.
Transition Flow Regime. The physics of the transition flow re-
gime is complicated, and most models are used to predict the
computational results of the Monte Carlo simulations (Karniada-
kis et al. 2005). The proposed models (Karniadakis et al. 2005)
implement different shear stress laws in the Navier-Stokes equa-
tions. The higher-order gradients of the velocity are implemented
in the proposed models, and the velocity profile is subject to
higher-order boundary conditions. To capture the flow rate, the
corresponding mass-flow rate for a circular tube was proposed as
(Karniadakis et al. 2005)
 
  4Kn pR4
M_ ¼ 1 þ 1:358tan1 ð4Kn0:4 Þ  Kn 1 þ qrp;
1 þ Kn 8l
                   ð7Þ
where M_ is the mass flux and q is the average density. The coeffi-
cients are obtained by fitting the equation to the Monte Carlo sim-
ulation results. We can express the apparent permeability by
dividing the mass-flow rate to the no-slip condition. Therefore,
the ratio of gas permeability to no-slip permeability can be
expressed as
 
kg  1 0:4
 4Kn
¼ 1 þ 1:358tan ð4Kn Þ  Kn 1 þ : . . . . ð8Þ
kl 1 þ Kn
August 2012 SPE Reservoir Evaluation & Engineering
 14
12
10
Fraction (%)
8 phase is flowing. The purpose of this calculation is simply to iso-
6
4 sidering the SEM images (Ambrose et al. 2010; Wang and Reed
2 the conductance of the circular tube depends on the fourth power
0
0 10 20 30 40 50
d (nm)
Fig. 1—Pore-size distribution of a Barnett shale sample
obtained from mercury-intrusion capillary pressure.
This indicates that the nonlinearity of permeability increases with
Kn. For the sake of convenience, we develop a polynomial form
for the permeability enhancement for 0.1 < Kn < 0.8, which lies
within the range of interest for shale-gas reservoirs. A second-
order polynomial works well:
kg
¼ 0:8453 þ 5:4576Kn þ 0:1633Kn2 ; . . . . . . . . . . . . . ð9Þ
kl
where the coefficients are calculated using the nonlinear regres-
sion model (Mendenhall et al. 1989), and the coefficient of regres-
sion is found to be 0.99. Note that Eq. 9 is valid only for
transition flow regime. Also, the gas permeability is no longer a
linear function of Kn. This reveals that the Klinkenberg correction
cannot be employed for higher-Kn flow regimes.
Free Molecular Regime. For completeness, we include the free
molecular regime, which is relevant for gas-phase transport at am-
bient conditions. The mass-flow rate of the free molecular regime
was modeled by Knudsen (1909). This model implements flux
because of a density gradient of the molecules as follows:
JKn ¼ DKn rni ; . . . . . . . . . . . . . . . . . . . . . . . . . . ð10Þ
where JKn is the mass flux of component i, DKn the Knudsen diffu-
sivity coefficient, and rni the density gradient. Unlike ordinary
diffusion, only one component is considered to predict the flow
rate in this mechanism. This model requires Knudsen diffusivity,
which is usually measured experimentally (Reinecke and Sleep
2002). It was derived analytically only for a long tube with a con-
stant circular cross-sectional area (Roy et al. 2003), as follows:
rffiffiffiffiffiffiffiffiffi
d 8RT d
DKn ¼ ¼ v; . . . . . . . . . . . . . . . . . . . . . . ð11Þ
3 pm 3 layer and slippage, if the network elements are slits rather than
where d is the tube diameter, m the molar mass, T the ambient
temperature, R the universal gas constant that is equal to 8.314 in
SI units, and v the average velocity of the molecules.
Adsorbed Layer
The effect of adsorbed gas is neglected in modeling flow through
conventional rock. This is reasonable because the occupied vol-
ume is negligible compared to the total void space in conventional
rocks. However, the adsorbed volume of CH4 is crucial in shale
because the throats are often smaller than 10 nm and much of the
void space is in the organic material, for which CH4 has a large
affinity. While molecular adsorption can occur on any surface, the
August 2012 SPE Reservoir Evaluation & Engineering
adsorbed layer of interest here is on the organic material. This
layer diminishes in thickness as the pressure decreases.
We analyze the effect of the adsorbed layer on the single-
throat conductance. The ratio of the throat conductance without
the adsorbed layer to that of the in-situ condition (gas adsorbed at
in-situ temperature and pressure) is evaluated. The conductance
without the adsorbed layer is computed by assuming a liquid
late the geometric effect of the adsorbed layer. We assume the
thickness of the adsorbed layer is 0.7 nm at 28 MPa (Ambrose
et al. 2010) and decreases linearly with reducing pressure. We fur-
ther suppose the cross sectional area of the throat is circular, con-
2009). For the calculation of liquid (i.e. no-slip) flow, we assume
of characteristic size, in agreement with Stokes flow of a Newto-
nian fluid (Sakhaee-Pour 2012). The influence of the adsorbed
layer is important for throat sizes smaller than 50 nm, which con-
stitute a significant fraction of shale pores. For a typical shale
sample, the characteristic throat size of 6 nm corresponds to the
largest fraction of conduits, as shown in Fig. 1.
We build our pore-space model on the basis of the SEM
images of shale and the drainage experiment. Some images of
Barnett shale available in the literature are shown in Fig. 2. The
high-resolution images indicate many throats exist in organic
materials. The affinity of gas molecules for organic materials
means that these throats constitute the void space that is altered
by desorption. It has been argued that the fluid flow in organic
materials is mainly single gas phase (Wang and Reed 2009).
Therefore, we propose a network model for gas permeability that
includes only voids within the organic matter.
Results
We analyze the conductance of a single-sized throat to study the
effects of the adsorbed layer and slippage. Then, we build a net-
work model to examine these effects when the connected throats
exhibit a distribution of sizes. Although the network does not ex-
plicitly represent the arrangement of voids in a shale, it does yield
a first-order estimate of difference in the gas-flow behavior
through the shale at different conditions—for example, high pres-
sure (early in the life of a well) vs. low pressure (after substantial
gas production) or field condition vs. laboratory condition. We
take into account the effects of the adsorbed layer and slippage on
the conductance of each throat in the network. Consequently, we
develop a characteristic plot, mapping the measurements taken at
typical laboratory conditions to the in-situ conductance of each
throat.
Analysis of a Single Cylindrical Conduit. Before considering
the network, we investigate the effects of the adsorbed layer and
flow slippage on the conductance of a single throat. In this regard,
we consider pore-size distribution (inferred from mercury poros-
imetry) and SEM of the Barnett shale sample shown in Figs. 1
and 2, respectively. We assume the void space is a network of cy-
lindrical throats. Our analysis (Sakhaee-Pour 2012) shows qualita-
tively similar flow behavior, in terms of response to the adsorbed
cylinders.
We begin by analyzing the effect of the adsorbed layer on the
liquid conductance of a single-sized throat. The goal is simply to
quantify the geometric effect of reducing the throat size as the
thickness of the adsorbed layer increases. Thus, slip is ignored;
the calculations assume no slip on the adsorbed layer. The role of
pressure in this analysis is merely to change the thickness of the
layer. The results are presented in terms of the ratio of the throat
conductance without the adsorbed layer to the throat conductance
with the adsorbed layer (see Fig. 3). We observe that the adsorbed
layer is crucial in estimating the conductance (with a no-slip
boundary condition), especially at larger pressures and in smaller
pore throats in the ranges of interest. The ratio of liquid conduc-
tances is larger at smaller throat sizes because the layer thickness,
403

Fig. 2—SEM of Barnett shale (Ambrose et al. 2010; Wang and
Reed 2009).
which is constant at a given pressure, occupies a greater fraction
of the cross section in smaller throats.
Next, we explore the importance of slippage with respect to
the gas conductance. To this aim, we calculate the ratio of gas to
liquid conductances. The conductances are computed assuming
that no adsorbed layer exists. The only effect of pressure is thus to
change Kn and, hence, the degree of slip. The temperature is fixed
at 300 K. As before, the liquid conductance is computed with a
no-slip boundary condition to serve as a reference value. The ratio
404
8 P=28 MPa
P=20 MPa
7
P=10 MPa
P=5 MPa
6
k(Kn=0)/kads
5
1
101
d (nm)
Fig. 3—Effect of adsorbed CH4 layer(s) on hydraulic conduct-
ance of a single cylindrical conduit. The ratio of “liquid” con-
ductance without adsorbed layer [k(Kn 5 0)] to “liquid”
conductance with adsorbed layer (kads) increases as pressure
increases and as conduit diameter decreases. Liquid conduct-
ance refers to single-phase laminar flow with no-slip boundary.
Thickness of adsorbed layer is proportional to pressure.
of gas to liquid conductances is shown in Fig. 4. The calculation
reveals that slippage plays an important role in smaller throat
sizes and at lower pressures. For instance, the gas conductance
of a 6-nm throat, which has the largest fraction of throat sizes in
Fig. 1, is larger than the liquid conductance by a factor of 1.5 at
P ¼ 20 MPa. This ratio increases nonlinearly with the decrease of
pressure, revealing that the slippage becomes more important at
late production.
Now, we consider the combination of the competing effects of
the adsorbed layer and slippage on the gas conductance. The
adsorbed layer reduces the conductance by reducing the cross-sec-
tional area. Slippage enhances the conductance by facilitating the
molecule movements at the throat surface. Both effects depend on
pressure. To analyze this competition, we calculate the ratio of
gas conductance with the adsorbed layer to liquid conductance
without the adsorbed layer. As in the previous analysis, the liquid
conductance is calculated with a no-slip boundary condition to
serve as a reference value. The gas conductance is obtained by
3 P=28 MPa
P=20 MPa
P=10 MPa
k[Kn(P,T=300 K)]/k(Kn=0)
P=5 MPa
2.5
1.5
1
101
d (nm)
Fig. 4—Effect of slip at pore walls on hydraulic conductance of
a single cylindrical conduit [ratio of gas conductance with slip-
page (Kn evaluated at 300 K and at pressure as per legend) to
liquid conductance without slippage k(Kn 5 0)]. Both conduc-
tances assume no adsorbed layer is present. The value of the
Kn changes with pressure.
August 2012 SPE Reservoir Evaluation & Engineering
 3.5 P=28 MPa
P=20 MPa
kin-situ[Kn(P,T=360 K)]/k(Kn=0)
3 P=10 MPa
P=5 MPa
2.5
2 6
1.5
1 2
0.5
0 d (nm)
101
d (nm)
Fig. 5—Combined effect of adsorbed layer and slip on hydraulic
conductance of a single cylindrical conduit [ratio of gas con-
ductance kin-situ with adsorbed layer and slippage (in-situ con-
dition of 360 K and P as per legend) to liquid conductance
k(Kn 5 0) without adsorbed layer and without slippage]. This
behavior is relevant to evolution of permeability during
production.
computing slip flow within the open cross section of the throat
remaining after implementing the adsorbed layer at the specified
pressure. The temperature is held constant at 360 K. Fig. 5 shows
qualitatively different trends, depending on pressure. At large
pressures, the ratio of gas conductance at in-situ conditions to the
corresponding no-slip conductance increases as the pore size
increases. At small pressures, the ratio decreases as the pore size
increases.
This calculation helps us understand which effect governs the
flow behavior at different stages of production. The purpose of
the comparison between gas and liquid conductances is to analyze
the gas-conductance variation vs. a reference value. It compares
the gas conductance (360 K, various P) that is representative of
reservoir condition to liquid conductance, the reference value
evaluated in the absence of an adsorbed layer and ignoring slip.
The analysis shows the adsorbed layer dominates the flow behav-
ior at high pressure (P ¼ 28 MPa in Fig. 5) regardless of the throat
size. That is, the gas conductance is smaller than the reference
conductance for all throat sizes because the adsorbed layer signifi-
cantly reduces the cross section for gas flow. The influence is
greater for smaller throats. However, the slippage has the primary
effect at small pressures (P ¼ 10 MPa or P ¼ 5 MPa). That is, the
gas conductance is greater than the reference conductance for all
throats, with the effect being greater in narrower throats. The
implication of Fig. 5 is that the adsorbed layer should be taken into
account for modeling gas transport during early production, while
the influence of slippage is dominant during later production.
Finally, for the single-tube study, we compare the gas conduc-
tances at laboratory and in-situ conditions. This is useful for
estimating gas permeability in the field from a laboratory measure-
ment. The adsorbed layer is presumed negligible at the laboratory
condition because the pressure (5 MPa) is low; the curve for the
corresponding pressure in Fig. 3 shows that the correction is less
than 25% for all throat sizes. The thickness of the adsorbed layer
varies at the in-situ condition, depending on the pressure. Note that
the slippage occurs in both conditions. We calculate the gas con-
ductance at the laboratory condition (T ¼ 300 K) from the liquid
conductance, accounting for slippage at laboratory pressure in the
relevant flow regime (Kn criterion evaluated at 300 K and 5 MPa)
by means of Eq. 2, Eq. 3, or Eq. 9. For the in-situ condition, we
compute the cross-sectional area open to flow, depending on the
thickness of the adsorbed layer. The gas conductance is computed
from the liquid conductance at the field pressure after implement-
ing the slippage. Fig. 6 shows that the gas conductance at labora-
August 2012 SPE Reservoir Evaluation & Engineering
12 P=28 MPa
P=20 MPa
klab[Kn(P=5 MPa, T=300 K)]
10 P=10 MPa
/kin-situ[Kn(P,T=360 K)]
8
4
0
101
Fig. 6—The ratio of gas conductance klab without adsorbed
layer and with slippage (laboratory condition of 300 K and 5
MPa) to gas conductance kin-situ with adsorbed layer and slip-
page (field condition of 360 K and P as per legend) increases as
pressure increases and as cylindrical throat size decreases.
These curves are relevant to estimating field permeability from
laboratory measurements.
tory condition is notably higher than at the in-situ condition at
larger pressures. This is because of the lack of an adsorbed layer
and the enhancement of slippage at laboratory condition vs. in-situ
condition. Clearly, the laboratory measurements should be cor-
rected for any flow modeling at the field condition. Considering the
importance of this issue, we will analyze the laboratory measure-
ments using network modeling and propose a correction in the fol-
lowing paragraphs.
Liquid permeability is also of interest in shale, either for the
problem of water production or for the challenge of “tight oil”
production. To estimate liquid permeability at the field condition
from gas permeability measured at the laboratory condition, we
calculate the ratio of gas to liquid conductances of a single cylin-
drical conduit. As in the preceding, the adsorbed layer at the labo-
ratory condition is negligible because of low pressure. Therefore,
the gas conductance at the laboratory condition is computed from
the liquid conductance by implementing the slippage at the labo-
ratory pressure and temperature (5 MPa, 300 K). For the liquid
conductance at the in-situ condition, we calculate the throat area
not obstructed by the thickness of the adsorbed layer. The thick-
ness of the adsorbed layer depends on pressure. We ignore the
swelling effect for the in-situ condition, which may reduce the hy-
draulic conductance. This may occur because of the presence of
water in the formation, which is not available at the same amount
at the laboratory condition. The ratio of gas to liquid conductan-
ces at laboratory and field conditions is shown in Fig. 7. The
results reveal the gas permeability obtained at the laboratory con-
dition is significantly larger than the liquid permeability at the in-
situ condition. The ratio of permeabilities estimated here is larger
than in the preceding analysis (Fig. 6) because the liquid does not
slip at laboratory condition. In reality, the ratio of gas to liquid
conductances may be even greater than in Fig. 7 because the hy-
draulic conductance of the throat decreases because of possible
swelling clays available in the formation.
Analysis of Network of Cylindrical Conduits of Distributed
Sizes. We employ a regular square lattice network (Bennett and
Myers 1962) to model the gas flow through organic materials in
shale. We adopt the pore-size distribution from the drainage
experiment shown in Fig. 1. We also assume the width of the net-
work (transverse to the direction of flow) and the throat length are
17 and 50 nm, respectively. The model width is chosen so that the
network model yields a desired porosity. Here, 17 nm is the value
that results in the desired porosity of 10%. The small width means
that the network corresponds to a thin 2D slice through the or-
ganic material oriented along the axes of the circular holes. That
is, we consider flow perpendicular to the plane of the images in
405
 25 P=28 MPa
klab[Kn(P=5 MPa, T=300 K)]/kads P=20 MPa
P=10 MPa
20
P=5 MPa
15
10
5 1.5
0
1
10
d (nm)
Fig. 7—For a single cylindrical tube, a laboratory measurement
of gas permeability greatly overestimates the permeability to
liquid in the reservoir, as shown by the ratio of gas conduct-
ance klab with slippage and without adsorbed layer (laboratory
condition of 300 K and 5 MPa) to liquid conductance kads with-
out slippage and with adsorbed layer (field condition of 360 K
and P as in legend). Pressure determines the thickness of the
adsorbed layer.
Fig. 2. The objective is not to represent the actual connectivity of
the voids, but to capture the influence of the adsorbed layers and
slip in a connected collection of throat sizes. When the throat
sizes are distributed randomly on the network, the gas permeabil-
ity of the network is 93 nanodarcies at laboratory condition [i.e.,
without the adsorbed layer; the calculation accounts only for slip
in each throat, using Eqs. 9 and 3 (high Knudsen flow models) to
relate flow in each throat to the pressure gradient along that
throat]. The network permeability is a plausible overestimate of
the measured value because the network represents only the or-
ganic portion of the shale. The organic matter accounts for 10%
of the core volume, so prorating the network contribution accord-
ingly would yield an estimate of 9 nanodarcies, which is typical
for shale (Billiotte et al. 2008).
Effect of Laboratory Conditions vs. Field Conditions on Gas
Permeability. First, we evaluate the gas permeability at in-situ
conditions by considering the influence of the adsorbed layer on
flow through each conduit in the network. To compute the gas
conductance at the in-situ condition, we calculate the liquid con-
ductance with no-slip boundary condition based on the cross-sec-
tional area, depending on the thickness of the adsorbed layer. The
gas conductance of each throat is then computed from the liquid
conductance.
The ratio of network permeabilities at laboratory conditions to
those at in-situ conditions is presented in Fig. 8. The results show
the measurements made at the laboratory condition would overes-
timate the permeability by a factor of four at P ¼ 28 MPa (i.e., at
the start of production). This factor is almost equal to the ratio of
laboratory to in-situ permeabilities for a single tube of 6 nm (see
Fig. 6), which has the largest fraction of pore-throat sizes as
deduced from the mercury-intrusion data shown in Fig 1. This
means that the throats, of which characteristic sizes constitute the
largest population, dominate the flow behavior. The overestima-
tion would be even larger if the prediction were based on labora-
tory measurements at ambient pressure and temperature. This is
because of enhancement in slippage occurring at low pressures.
The ratio reaches unity at a slightly larger pressure than 5 MPa
because the field temperature (360 K) is greater than the labora-
tory temperature (300 K) and because the gas for the laboratory
measurement is N2 and not CH4. The difference in temperatures
makes Kn slightly larger at the laboratory condition, but for a
given pressure, temperature, and throat size, the properties of N2
yield a smaller Kn than for CH4. Thus, smaller slippage occurs for
406
4.5
4
3.5
kg, lab/kg, in-situ
3
2.5
2
1
0.5
5 10 15 20 25 30
P (MPa)
Fig. 8—The ratio of gas permeability with slippage and without
adsorbed layer (laboratory condition of 300 K and 5 MPa) to gas
permeability with adsorbed layer and slippage (field condition
of 360 K and P from the x-axis) obtained from a network of cy-
lindrical throats having the size distribution shown in Fig. 1.
N2 than for CH4 because of a smaller Kn, resulting in a smaller
permeability. Thus, the gas permeability of the network model
exposed to CH4 at a slightly larger pressure than 5 MPa at field
condition is equal to the N2 permeability at the laboratory
condition.
The ratio of laboratory to in-situ permeabilities decreases as
the pressure decreases. This is because of the opposing effects of
the adsorbed layer and slippage at different pressures. To provide
a quick tool for estimating the in-situ permeability from the labo-
ratory measurement, we fit a curve to the results in Fig. 8. The
nonlinear regression model indicates the coefficient of regression
is 0.99. Therefore, the in-situ gas permeability can be estimated
from the laboratory measurement using the following equation:
kg;lab
¼ 0:001P2 þ 0:0898P þ 0:783; . . . . . . . . . . . ð12Þ
kg;in-situ
where kg,in-situ is the gas permeability at the in-situ condition, kg,lab
is the gas permeability at the laboratory condition (300 K, 5
MPa), and P is the gas pressure in MPa.
Effects of Field Conditions on Gas Permeability During
Production. Finally, we investigate the effect of declining field
pressure on the gas permeability using the network model. In this
regard, we calculate the ratio of the field permeability at low pres-
sures (kg2,in-situ) to the gas permeability value at the start of pro-
duction (kg1,in-situ). The initial field pressure is presumed to be 28
MPa. We take into account the effect of the adsorbed layer,
depending on the pressure. In addition, we account for the effect
of pressure on slippage. The ratio of the gas permeabilities is pre-
sented in Fig. 9. The network modeling suggests that the shale
permeability at the start of production is significantly smaller than
during late production. The adsorbed layer is the reason for this
effect. Furthermore, the dependency of the permeability on the
pressure is nonlinear. Hence, the gas permeability of shale is
highly dependent on pressure, unlike the conventional reservoir.
This has major implications in reservoir simulations and ultimate-
recovery-estimation models.
Validation Against Laboratory Data. Figs. 8 and 9 summa-
rize the main implications of the theory. Fig. 8 estimates gas
permeability at the in-situ condition from the laboratory measure-
ment, and Fig. 9 predicts the enhancement of gas permeability
during production. Two sets of experiments are required to vali-
date these implications. The first set requires comparison of the
measurements with N2 at the laboratory condition and CH4 at the
reservoir condition. For the second set, we must analyze the
August 2012 SPE Reservoir Evaluation & Engineering
 4.5
4
3.5
kg 2, in-situ /kg1, in-situ
3
2.5
2
1.5
1 1
5 10 15 20 25 30
P (MPa)
Fig. 9—The ratio of gas permeability at pressures below initial
reservoir pressure (kg2,in-situ) to gas permeability at initial pres-
sure (P 5 28MPa) (kg1,in-situ) increases as production continues
and pressure declines accordingly. The ratio is calculated from
a network of cylindrical throats having the size distribution
shown in Fig. 1.
dependency of rock resistance against gas flow on the pore pres-
sure when the moving fluid is methane. The test with methane
should be run at reservoir temperature and pressures to represent
the in-situ conditions. Our model predicts that the effect of the
adsorbed layer becomes more important as the organic content of
shale increases. The organic-rich region has more tendancy to
adsorb methane to its pore wall, which must be considered when
choosing the core samples.
Data across the range of conditions of Figs. 8 and 9 for rocks
with different organic content are not currently available. How-
ever, measurements of CH4 transport reported on a particular type
of shale at pressures below 7 MPa (Letham 2011) let us partially
test the effect of slippage. The experimental data were measured
with methane for a sample with a permeability of 390 nanodarcies
at 5 MPa and 300 K. The effective stress was kept constant in the
laboratory measurements (Letham 2011); that is, the confining
stress was lowered to yield the same effective stress at a lower
pore pressure. The porosity of the laboratory sample was reported
to be 13%. The porosity of our original sample was 10%, which is
slightly different from the laboratory sample. To compensate for
these differences, we increased the pore-throat sizes of our origi-
nal network model, of which permeability
qwas
ffiffiffiffiffiffi 9 nanodarcies
at 5
390
MPa and 300 K, by a factor of 6.58 9 ¼ 6:58 . We also
increased the width of the network by a factor of 43.3 (¼ 6.582) to
keep the porosity of the network model unaffected by the change
in the pore-size distribution. Using the modified model, we com-
pute gas permeabilities for a range of small pressures (between 1
and 6 MPa) and at 300 K. Since the effective stress was kept con-
stant during laboratory measurements, we do not change the
modified pore-throat-size distribution with pore pressure. (In the
field, effective stress increases with reservoir depletion, and the
effect on pore-throat-size distribution, which is not included in
the model presented here, would diminish the slippage effect in
Fig. 10.) Fig. 10 plots these gas permeabilities, normalized by the
nominal liquid permeability at laboratory condition. Similar to ev-
ery calculation performed for the laboratory condition through
this study, the gas flow here is with slippage and without the
adsorbed layer, and the liquid flow is without slippage and with-
out the adsorbed layer. However, unlike the laboratory condition
we defined in which pressure was 5 MPa, we change the pressure
to investigate its influence on gas permeability because of slip-
page. The liquid permeability for the laboratory data is 390 nano-
darcies, estimated by extrapolating the measured gas permeability
to large pressure. Fig. 10 shows that the model slightly underesti-
August 2012 SPE Reservoir Evaluation & Engineering
1.9
1.8 Laboratory data (Letham 2011)
1.7 Corrected throats model
klab(P, T=300 K)/k(Kn=0)
1.6
1.5
1.4
1.3
1.2
1.1
0 1 2 3 4 5 6 7
P (MPa)
Fig. 10—The ratio of gas permeability without an adsorbed
layer and with slippage [klab(P, T5300 K)] to the liquid perme-
ability without an adsorbed layer and without slippage
[k(Kn 5 0)] is obtained from a network of conduits and com-
pared to the laboratory measurements (Letham 2011). The
effective stress is constant, and hence, the pore-pressure
change does not affect pore-throat-size distribution. Gas per-
meability increases with lowering pressure because of slip-
page, unlike the liquid permeability, which is constant. The
flowing gas is methane.
mates the normalized gas permeability, the maximum difference
with the laboratory measurement being only 6%. This indicates
that the slippage model provides a good estimate for the effect of
pressure on gas permeability at moderate pressures.
Conclusions
Pore throats of shale are mostly narrower than 10 nm and are
inside organic material on which CH4 adsorbs. As a result, the gas
permeability of these rocks is affected significantly by the
adsorbed gas and by the slip of flowing gas on the pore walls (or
on layers of molecules adsorbed on the walls). The adsorbed layer
does not play an important role in conventional rocks because the
pore throats are much wider, nor is the gas volume desorbed from
the organic material significant in conventional reservoirs. To bet-
ter understand gas-flow behavior through shale, we evaluated
these effects in individual conduits of a cylindrical cross section
and in simple networks of such conduits. The effect of the
adsorbed layer was treated as purely geometric: The cross section
open to flow was reduced by the thickness of the adsorbed layer,
which was assumed to vary linearly with pressure. The effect of
slip was accounted for by applying the model appropriate to the
flow regime, according to the Knudsen number (Kn). The latter
depends on pressure, temperature, and conduit diameter. For the
slip flow regime, the first-order slip model was judged more suita-
ble than the dusty gas model (DGM) for the shale-gas application.
At large pressures such as typical initial shale-gas-reservoir
pressures, the effect of the adsorbed layer dominates the effect of
slip on gas-phase permeability. Slip dominates at smaller pres-
sures typical of those after longer periods of production. Conse-
quently, the reservoir matrix permeability is predicted to increase
significantly throughout the life of a well, by a factor of 4.5, as
production continues and pressure declines. The models predict
that the typical conditions for laboratory measurements of perme-
ability cause those values to overestimate field permeability by as
much as a factor of four. The model results are captured in simple
analytical expressions that allow convenient estimation of these
effects.
For complete validation of the proposed model, laboratory
measurements at elevated pressures (on the order of 28 MPa) and
reservoir temperature made with CH4 are needed. Experimental
data at such conditions, for which the adsorbed layer is expected
407
to have the dominant effect, are not available. Comparing model burgh, Pennsylvania, USA, 23–25 February. http://dx.doi.org/10.2118/
predictions with the laboratory data at lower pressures (<7 MPa) 131772-MS.
and constant effective stress permits evaluating the importance of Bennett, C.O. and Myers, J.E. 1962. Momentum, Heat, and Mass Transfer.
slippage. When normalized by the nominal absolute permeability New York: Series in Chemical Engineering, McGraw-Hill.
of the network (i.e., the value at which both slip and adsorbed Billiotte, J., Yang, D., and Su, K. 2008. Experimental study on gas permeabil-
layers are negligible), the predicted trend agrees well with the ity of mudstones. Physics and Chemistry of the Earth, Parts A/B/C 33
measurements. The maximum difference between the predicted (Supplement 1): S231–S236. http://dx.doi.org/10.1016/j.pce.2008.10.040.
normalized gas permeability and the measured value is 6%. This Cipolla, C.L., Lolon, E.P., Erdle, J.C. et al. 2010. Reservoir Modeling in
means that our adapted network model provides a reasonable ba- Shale-Gas Reservoirs. SPE Res Eval & Eng 13 (4): 638–653. SPE-
sis for understanding the effect of gas pressure on the matrix per- 125530-PA. http://dx.doi.org/10.2118/125530-PA.
meability of a shale-gas reservoir. Cui, X., Bustin, A.M.M., and Bustin, R.M. 2009. Measurements of gas
permeability and diffusivity of tight reservoir rocks: different
approaches and their applications. Geofluids 9 (3): 208–223. http://
Nomenclature
dx.doi.org/10.1111/j.1468-8123.2009.00244.x.
d ¼ tube diameter Graham, T. 1833. On the law of the diffusion of gases. J. Membr. Sci. 100
DKn ¼ Knudsen diffusivity coefficient (1): 17–21. http://dx.doi.org/10.1016/0376-7388(94)00228-Q.
JKn ¼ Knudsen flux Graham, T. 1846. On the Motion of Gases. Phil. Trans. R. Soc. Lond. A
Jtotal ¼ total flux 136 (1846): 573–631.
Jvisc ¼ viscous flux Graham, T. 1849. On the Motion of Gases. Part II. Phil. Trans. R. Soc.
k(Kn ¼ 0) ¼ no-slip conductance for a single cylin- Lond. A 139 (1849): 349–391.
drical tube with no adsorbed layer Graham, T. 1863. On the Molecular Mobility of Gases. Phil. Trans. R.
k[Kn(P,T ¼ 300 K)] ¼ gas conductance for a single cylindri- Soc. Lond. A 153 (1863): 385–405.
cal tube with slippage and without an Javadpour, F., Fisher, D., and Unsworth, M. 2007. Nanoscale Gas Flow in
adsorbed layer
Shale Gas Sediments. J Can Pet Technol 46 (10): 55–61. JCPT Paper
kads ¼ no-slip conductance for a single cylin-
No. 07-10-06. http://dx.doi.org/10.2118/07-10-06.
drical tube in the presence of an
Karniadakis, G., Bes kök, A., and Aluru, N. 2005. Microflows and Nano-
adsorbed layer
flows: Fundamentals and Simulation, Vol. 29. New York: Interdisci-
kg ¼ single-phase gas permeability
plinary Applied Mathematics, Springer. ISBN 0387221972.
kg,in-situ ¼ gas permeability of the network
Klinkenberg, L.J. 1941. The permeability of porous media to liquids and
model at the in-situ condition
gases. API Drilling & Production Practice (1941): 200–213.
kg,lab ¼ gas permeability of the network
Knudsen, M. 1909. Die Gesetze der Molekularströmung und der inneren
model at the laboratory condition
Reibungsströmung der Gase durch Röhren (The laws of molecular and
kg1,in-situ ¼ gas permeability of the network
viscous flow of gases through tubes). Ann. Phys. 333 (1): 75–130.
model at the start of production repre-
senting the in-situ condition http://dx.doi.org/10.1002/andp.19093330106.
kg2,in-situ ¼ gas permeability of the network Letham, E.A. 2011. Matrix permeability measurements of gas shales: gas
model at a different reservoir pressure slippage and adsorption as sources of systematic errors. BSc thesis,
representing the in-situ condition University of British Columbia, Vancouver, Canada.
kin-situ[Kn(P,T ¼ 360 K)] ¼ gas conductance for a single cylindri- Luffel, D.L., Hopkins, C.W., and Schettler Jr., P.D. 1993. Matrix Perme-
cal tube at the in-situ condition with ability Measurement of Gas Productive Shales. Paper SPE 26633 pre-
slippage and an adsorbed layer sented at the SPE Annual Technical Conference and Exhibition,
kl ¼ permeability of a conduit to liquid Houston, 3–6 October. http://dx.doi.org/10.2118/26633-MS.
with no-slip boundary condition Mallon, A.J. and Swarbrick, R.E. 2008. How should permeability be meas-
klab[Kn(5 MPa,300 K)] ¼ gas conductance for a single cylindri- ured in fine-grained lithologies? Evidence from the chalk. Geofluids 8
cal tube at the laboratory condition (1): 35–45. http://dx.doi.org/10.1111/j.1468-8123.2007.00203.x.
with slippage and without an adsorbed Mason, E.A. and Malinauskas, A.P. 1983. Gas Transport in Porous
layer Media: The Dusty-Gas Model. Amsterdam, The Netherlands: Chemi-
Kn ¼ Knudsen number cal Engineering Monographs, Elsevier Science.
m ¼ molar mass Mendenhall, W., Reinmuth, J.E., and Beaver, R. 1989. Statistics for Man-
M_ ¼ mass flux agement and Economics. Boston, Massachusetts: PWS-Kent Publish-
p ¼ gas pressure ing Company.
P ¼ gas pressure in MPa Philip, Z.G., Jennings Jr., J.W., Olson, J.E. et al. 2005. Modeling Coupled
R ¼ universal gas constant Fracture-Matrix Fluid Flow in Geomechanically Simulated Fracture
R ¼ specific gas constant Networks. SPE Res Eval & Eng 8 (4): 300–309. SPE-77340-PA.
T ¼ temperature http://dx.doi.org/10.2118/77340-PA.
a1 ¼ permeability enhancement Polczer, S. 2009. Shale expected to supply half of North America’s gas.
k ¼ mean free path of gas molecule Calgary Herald, 9 April 2009.
K ¼ characteristic length of the channel Reinecke, S.A. and Sleep, B.E. 2002. Knudsen diffusion, gas permeability,
l ¼ gas viscosity and water content in an unconsolidated porous medium. Water Resour.
q ¼ gas density Res. 38 (12): 1280. http://dx.doi.org/10.1029/2002wr001278.
Roy, S., Raju, R., Chuang, H.F. et al. 2003. Modeling gas flow through
microchannels and nanopores. J. Appl. Phys. 93 (8): 4870–4879.
http://dx.doi.org/10.1063/1.1559936.
Acknowledgments
Sakhaee-Pour, A. 2012. Gas flow through shale. PhD dissertation (in pro-
The authors acknowledge financial contributions from Exxon- gress), The University of Texas at Austin, Austin, Texas.
Mobil and Jackson School of Geosciences at The University of Urbina, I. 2011. Insiders Sound an Alarm Amid a Natural Gas Rush. N.Y.
Texas at Austin. Times, 25 June 2011, http://www.nytimes.com/2011/06/26/us/26gas.
html?_r¼1&pagewanted¼all.
References Wang, F.P. and Reed, R.M. 2009. Pore Networks and Fluid Flow in Gas
Ambrose, R.J., Hartman, R.C., Diaz-Campos, M. et al. 2010. New Pore- Shales. Paper SPE 124253 presented at the SPE Annual Technical
scale Considerations in Shale Gas in Place Calculations. Paper SPE Conference and Exhibition, New Orleans, 4–7 October. http://
131772 presented at the SPE Unconventional Gas Conference, Pitts- dx.doi.org/10.2118/124253-MS.

408 August 2012 SPE Reservoir Evaluation & Engineering

Appendix A ð ð1    
2  rv
First-Order Slip Model. The velocity profile in the slip flow re- Q ¼ us ðrÞ2prdr ¼ a 1  x2 þ 2 Kn xdx
gime for a cylindrical conduit is obtained by adding a correction rv
0
term to the solution of the Navier-Stokes equation after applying
the Newtonian-fluid assumption with no-slip boundary condition ð1      
Q 2  rv 1 2  rv
(Roy et al. 2003): Q¼ 1  x2 þ 2 Kn xdx ¼ þ Kn ;
a rv 4 rv
0
us ¼ uno-slip þ ucorrection ; . . . . . . . . . . . . . . . . . . . . . ðA-1Þ
                   ðA-8Þ
where us is the velocity profile in the slip flow regime, uno-slip is where Q is the flow rate, a is a characteristic flow rate, and Q is
the no-slip velocity of the Newtonian fluid, and ucorrection is the the nondimensional flow rate. The first term on the right-hand
correction term to implement the effect of slippage. The correc- side of Eq. A-8 represents the no-slip flow rate, and the second
tion term is constant, independent of location. The slip velocity term is the enhancement because of slippage. Therefore, the effect
profile can also be expressed in a nondimensional form as of slip on permeability can be readily measured by dividing
QðKn Þ by QðKn ¼ 0Þ :
Us ¼ Uno-slip þ Ucorrection ; . . . . . . . . . . . . . . . . . . . . ðA-2Þ

1 2rv  
where the nondimensional no-slip velocity is defined as follows: kg 4 þ rv Kn 2  rv
¼ 1
¼ 1 þ 4 Kn : . . . . . . . . . ðA-9Þ
r
2 kl 4
rv
Uno-slip ¼ 1  ¼ 1  x2 ; . . . . . . . . . . . . . . . . ðA-3Þ
R Here kg is the permeability of the conduit to gas phase, and kl is
where R is the radius of the tube, r is the distance from the center the permeability of the same conduit when the no-slip boundary
of the tube, and x is the normalized radial distance from the center condition applies. The latter condition usually holds when a liquid
of the tube, which is equal to unity at the wall. To include the fills the conduit, hence the choice of subscript l. After implement-
effect of molecule/wall collisions, the first-order slip boundary ing rv ¼ 0.9, we obtain the enhancement of apparent permeability,
condition is imposed as (Klinkenberg 1941) called a1 in the text (refer to Eq. 3), to be equal to 5.

2  rv @Us  DGM Expression in Terms of Kn. The apparent permeability of
Us j w ¼ Kn ; . . . . . . . . . . . . . . . . . . . . ðA-4Þ
rv @n w the conduit based on the DGM is calculated as
 
where n is the outward (unit) vector, which is normal to the tube kg DK l b
¼ 1þ n ¼ 1 þ : . . . . . . . . . . . . . . . . . ðA-10Þ
wall, and rv is the tangential momentum-accommodation coeffi- kl kl p p
cient. The tangential momentum accommodation is calculated by
measuring the gas-flow rate at the slip flow regime and is close to The second term on the right-hand side of Eq. A-10 can be simpli-
0.9 (Roy et al. 2003). fied after substituting the Knudsen diffusivity coefficient in terms
The first-order slip boundary condition is applied to derive the of a pertinent parameter, using Eq. 11. We also use the Hagen-
nondimensional velocity profile (see Eq. A-2) (Klinkenberg Poiseuille model for the permeability term, as follows:
1941): rffiffiffiffiffiffiffiffiffi rffiffiffiffiffiffiffiffiffi
    d 8RT 8RT
l
@Us  @Uno-slip @Ucorrection  @Uno-slip  b DKn l 3 pm  l 32 1 pm
¼ þ ¼ ¼ ¼ ¼
@n w @n @n 
w @n w p kl p d2 3 d p
p
¼ ð2xÞjx¼1 ¼ 2:                  ðA-5Þ 32
rffiffiffiffiffiffiffiffiffi
2RT
The minus sign is because of the direction of the normal vector. l
64 1 pm
The correction term is constant, and, thus, its gradient vanishes. ¼ :                    ðA-11Þ
Therefore, Eq. A-4 simplifies to (Klinkenberg 1941) 3 d p
   
2  rv @Us  2  rv Here, we implement the ideal-gas assumption p ¼ q mR T ¼ qRT
K ¼ 2 Kn : . . . . . . . . . . . . . . ðA-6Þ to express the pressure in terms of density, which yields the
rv @n w
n
rv results in terms of Kn:
Thus, we can express the nondimensional velocity profile based rffiffiffiffiffiffiffiffiffi
2RT rffiffiffiffiffiffiffiffiffi
on Eqs. A-2 and A-3 as follows (Klinkenberg 1941): l
b 64 1 pm 64 l 2m
  ¼ ¼
2  rv p 3 d R 3 qd pRT
Us ¼ 1  x 2 þ 2 Kn : . . . . . . . . . . . . . . . . ðA-7Þ q T
rv r mffiffiffiffiffiffiffiffiffi
128 l p
To cast this profile in terms of permeability, we obtain the flow ¼ ¼ 13:58Kn :              ðA-12Þ
3p qd 2RT
rate as

August 2012 SPE Reservoir Evaluation & Engineering 409