PETROPHYSICS, VOL. 57, NO.

1 (FEBRUARY 2016); PAGE 3040; 9 FIGURES; 3 TABLES

Low-Permeability Measurements: Insights1

Sandra Proce2, Grald Hamon2, and Benjamin Nicot2

ABSTRACT
Recent studies have reported unacceptable
discrepancies (up to several decades) between intrinsic
permeability (kl) measurements from different laboratories.
These discrepancies could be explained by (a) the
diversity of experimental procedures, since no standards
of measurement have been clearly dened for poorly
permeable samples, and (b) the validity of interpretative
models which, are rarely available.
This paper rst presents the results of an experimental
study comparing: (1) Values of the intrinsic permeability kl
from step-decay (gas), pulse-decay (gas), and steady-state
(both gas and liquid) tests; (2) Values of the Klinkenberg
coefcient (b) from step-decay and steady-state tests; (3)
Values of the porosity ( )from step-decay and pycnometry
tests.
On a homogeneous material of ultralow permeability
(pyrophyllite), different techniques achieve similar results

no matter which property is being measured (kl, b or ).

Moreover, the accurate prediction of the gas ow behavior
by interpretative models relying on the Darcy-Klinkenberg
equation highlights that gas ow in tight rocks is still
viscous ow slipping at pore walls. These conclusions
were checked for hydrocarbon-bearing rocks by repeating
the study on shale.
The paper then provides the results of steady-state
and unsteady-state measurements performed during a
round-robin test. Pyrophyllite plugs were successively
characterized by four laboratories, which selected their
methods and experimental conditions but had to work on the
whole as-received plugs and at a given effective pressure.
It arises from the round-robin test that a satisfactory
agreement of kl estimations from different laboratories
requires a proper denition of the experimental procedure.

INTRODUCTION

requiring several hours or even days when the material was

extremely tight. Alternative techniques dedicated to a faster
analysis of ultralow-permeability porous media have been
developed since the early 1950s. Bruce et al. (1953) were
the rst authors to propose an unsteady-state technique
commonly referred to as pulse decay. This pioneering
work engendered numerous other studies on the method,
aiming at deriving interpretative models (Brace et al., 1968;
Jones, 1972; Hsieh et al., 1981).
To summarize, in most of the studies, apparent
permeability (kl) is estimated separately from the porosity ()
and, in rare cases, simultaneously with b. The pulse-decay
technique is still widely used in the petroleum industry,
either on core plugs or on drill cuttings. The cheaper and
faster option consisting in working on cuttings is known as
the Gas Research Institute (GRI) method, rst described by
Luffel and Guidry (1992). Even if unsteady-state techniques
relying on the application of a pulse of pressure are popular,

Over the past 10 years oil and gas shales have become a
topic of real interest due to the large amounts of hydrocarbons
they could potentially produce. Their characterization is
a challenging task since these unconventional reservoirs
have tight pore throats (a few tens of nanometers) and
low permeabilities (from microdarcies to nanodarcies).
Consequently, shales require careful identication of reliable
methods to identify their one-phase ow properties.
Steady-state techniques are the oldest and simplest ones.
They allow the determination of the intrinsic permeability
(kl) and, when applied with gas at different mean pore
pressures, give the Klinkenberg coefcient (b) in addition
(Rushing et al., 2004; Boulin et al., 2012). Until recently,
the principal drawback of such methods lay in the time
needed to achieve steady state at each new measuring point.
The sample characterization used to be a long process,

Manuscript received by the Editor January 5, 2016; revised manuscript received January 26, 2016.
1
Originally presented at the SCA International Symposium, St. Johns Newfoundland and Labrador, Canada, 1621 August, 2015, Paper SCA2015021.
2
TOTAL CSTJF Avenue Larribau 64018 Pau Cedex France; proce.sandra@laposte.net

30

PETROPHYSICS

February 2016

Low-Permeability Measurements: Insights

they are not the only methods enabling a rapid and accurate
characterization of low-permeabilty rocks. Indeed, some
oil and gas companies have decided use different in-house
devices, such as IFPENs improved steady-state device
(Boulin et al., 2012), Shells pore-pressure oscillation
device (Wang and Knabe, 2010) and Totals step-decay
device (Lasseux et al., 2012). With the rst device, a fast
derivation of kl is allowed with a liquid. With the second
one, the measurement of k and ( b too, if several tests
are carried out) is based on the excitation of the plug by a
sinusoidal pressure wave. With the last one, the simultaneous
determination of kl, b and is ensured by the generation of a
series of pulses of pressure on the plug.
All techniques presented in the previous paragraph
involve interpretative models based on the rst assumption
that Darcys law is still valid when modeling uid ow
in low-permeability media. The second assumption, that
Klinkenbergs law (Klinkenberg, 1941) is also valid, is
made when b is estimated. However, the literature often
questions these assumptions, since the Knudsen numbers
typical of shales are out of the range of validity of the DarcyKlinkenberg law. Karniadakis et al. (2005) and Javadpour
(2009), who focused on networks composed of micropores
and nanopores respectively, rejected Darcys law and all
suggested new formulations of the gas ow rate. Javadpour
(2009) even proposed an expression of k depending on
the rock specicities as well as on the uid properties at
given values of temperature and pressure. The notion of
intrinsic permeability is completely lost in his approach.
Using Karniadakis et al. (2005) theory, Civan (2009)
deduced a relationship between k and kl applicable to the
whole Knudsen number range. More recently, Fathi et al.
(2012) have derived a new relationship between k and kl by
theorizing the phenomenon of double molecular slippage at
the pore scale.
From these last studies, a natural question arises: Are
interpretative models relying on the Darcy-Klinkenberg law
well-suited for shales? This is the rst question this paper will
try to answer. The second question regards the great number
of methods used in the industry for routine measurements.
Are the discrepancies between the results found on an
identical sample by different laboratories explained by the
diversity of their interpretative models? Indeed, several
authors (Passey et al., 2010; Sondergeld et al., 2010) recently
mentioned discrepancies between permeability estimations
reaching several decades. The question remains whether
this wide dispersion of results is due to interpretation or to
experimental procedure.

February 2016

WORK AND METHODS

Three experimental studies were devised to answer the
questions above. Study 1 aimed at comparing the values of
kl, b and estimated for homogeneous plugs of pyrophyllite
with common methods of the industry to those estimated with
our step-decay method. To be more precise, the comparison
was made between:
The values of kl given by a step-decay test, a pulsedecay test, a steady-state test with gas and a steadystate test with oil.
The values of b given by a step-decay test and a
steady-state test with gas.
The values of given by a step-decay test and a
pycnometry test.
The same approach was then applied in Study 2 to two shale
plugs sourced from an actual development target area, in
order to check whether the results from the pyrophyllite
study could be corroborated by a similar study with
reservoir rocks. Study 3 was a round-robin test involving
three commercial laboratories and Total. Its objective was
to identify the cause of the discrepancies often observed
between the kl estimations found by different laboratories
for an identical sample.
Plugs
Study 1 was carried out on ve plugs of pyrophyllite
named Pyro 1, Pyro 2, Pyro 3, Pyro 4 and Pyro 5 respectively.
Pyrophyllite is a homogeneous quarry rock sourced from the
United States. It is mostly composed of clay and consequently
has a low permeability. All plugs were successively subjected
to step-decay tests, pulse-decay tests and steady-state tests
using gas and Pyro 5, the least permeable of the ve plugs,
was also analyzed with a liquid at steady state at the end
of the experiments with gas. No special treatment, such
as cleaning or drying, was performed on the pyrophyllite
before the measurements with gas, which were conducted
therefore on plugs containing some water. This choice was
made to simplify the experimental procedure, given that the
permeability of pyrophyllite does not evolve signicantly
between the as-received and dry states. Indeed, preliminary
tests done on a plug other than Pyro 1 to Pyro 5 showed
an increase of permeability of 46% between the state where
pyrophyllite was at equilibrium with the surrounding air and
the state where pyrophyllite was dried (after nine days in an
oven at 100C). Finally, before the steady-state experiment
with liquid, Pyro 5 was saturated at 400 bar during two
weeks, after having created a vacuum over a period of four
days. Note that Pyro 1, Pyro 2, Pyro 3 and Pyro 4 were the
four plugs sent to the three commercial laboratories selected
for the round-robin test of Study 3.

PETROPHYSICS

31

Proce et al.

The two shales chosen for Study 2, Shale 1 and Shale

2, respectively, are of different origins. Shale 1 comes from
a wet-gas well and Shale 2 from an oil well. Both samples
were subjected to the gas tests (step-decay, pulse-decay and
steady-state) in native state. No steady-state analysis with
liquid was undertaken knowing that the initial oil phase in
the plug would have probably moved with the ow. This
would have led to an estimation of kl that was not comparable
with the results derived with gas. The representativeness
of the results obtained on the shale plugs is debatable for
two principal reasons: (a) The oil-shale plug was analyzed
with gas in the as-received condition, and (b) plugs possibly
contained microfractures, since the CT-scans had not
been carefully observed before the selection. However,
the objective of Study 2 was not to properly characterize
hydrocarbon-bearing rocks, but to check if different gasbased techniques provide comparable results on an identical
shale plug tested in a given state.
A mercury injection capillary pressure (MICP) analysis
was carried out on a fragment of rock from the same block
that provided Pyro 1. Figure 1, which gives the pore-throat
size distribution from the MICP analysis, reveals that the
most represented pore-throat radius in the material is around
20 nm. This result is in line with the range of a pore radius of
10 to 40 nm found by applying the density functional theory
(DFT) method on a crushed sample taken from the Pyro 1
block. Pores in shales are similarly characterized by radii of
several tens of nanometers.

measurements taken at both extremes of the plug. As shown

in Fig. 2, the plug is conned in a Hassler-sleeve core holder
connected to two tanks. The pulse of pressure is prepared
in the upstream tank V0 and liberated at the plug entrance
by opening the valve v0. The recording of P is triggered
simultaneously with the pulse emission. The gas owing
out of the plug is collected in the downstream tank V1. All
pulse-decay tests were conducted with the Pulse Decay
Permeameter 200, an automated device manufactured by
Corelab, which works with nitrogen and at a high mean
pore pressure Pp to eliminate the Klinkenberg effect. Pore
pressure was initially stabilized at:
Shale 1: 31 bar (450 psi)
Plugs of pyrophyllite and Shale 2: 69 bar (1,000 psi).
The gas ow from V0 to V1, which are both around 20 cm3, is
started by depressurizing V1 until P reaches 3 bar (40 psi).
The effective pressure Peff, dened as the difference between
the conning pressure Pc and Pp, was:
Plugs of pyrophyllite: 60 bar
Shales: 100 bar.
The temperature was xed at 20C for all our tests (unsteadystate and steady-state). The interpretation of P is based on
the approach of Jones (1972). Jones devised a method of
calculating kl by using an approximate analytical solution
of the pulse-decay problem that combines the traditional
equations of uid mechanics in porous materials (including
Darcys law) and excludes the Klinkenberg effect. The
solution is an innite sum of exponentials decreasing over
time. At long times, the rst term predominates over the
other terms. Consequently, the curve of the logarithm of P
evolves linearly over time and the derivation of kl becomes
possible from the slope.

Fig. 1Pore-throat size distribution.

Fig. 2Schematic of the pulse-decay device.

Methods and Interpretative Models

The pulse-decay technique (Brace et al., 1968) consists
in generating a pulse of pressure on one face of the plug
and recording the differential pressure P calculated from

32

PETROPHYSICS

February 2016

Low-Permeability Measurements: Insights

In the step-decay technique (Lasseux et al., 2012), the

plug is excited with several pulses of pressure. In terms of
the device, the step-decay technique differs from the pulsedecay technique only by the existence of an additional
buffer tank Vb located at the plug entrance, as represented
in Fig. 3. The rst pulse is prepared by supplying V0 and
Vb with nitrogen. Once the selected pressure setpoint is
reached, the valve vb is closed to isolate Vb from V0. The
gas is then released by opening v0. This operation triggers
the simultaneous recording of the upstream and downstream
pressures, P0 and P1 respectively. All pulses of pressure
from the second pulse are produced by pressurizing Vb and
are liberated by briey opening vb. The pulse pressures and
waiting time intervals chosen for the step-decay tests were
the following ones:
Plugs of pyrophyllite: 6 bar/16 min; 10 bar/30 min;
31 bar/5 to 15 hr
Shale 1: 3 bar/1 hr; 5 bar/1.5 hr; 9 bar/2 hr
Shale 2: 6 bar/15 min; 13 bar/2 hr; 22 bar/20 hr.
For each plug, the same Peff was applied for both pulse-decay
and step-decay tests. The values of V1 were:
Pyro 1, Pyro 2, Pyro 3 and Pyro 4: 8.43 cm3
Pyro 5, Shale 1, and Shale 2: 12.46 cm3.

Fig. 3Schematic of the step-decay device.

Fig. 4Schematic of the classical steady-state device.

In the interpretation of the raw data, the step-decay

model assumes the validity of the Darcy-Klinkenberg law,
as do most pulse-decay models found in existing studies of
experiments at low pore pressure. The step-decay model
provides kl, b and by matching a numerically simulated
P1 prole with the recorded P1 data. Note that a particular
strength of the method resides in the fact that the P0 prole
acts as an input for the interpretation. One consequence
of that is the removal of V0 from the series of parameters
required for the inversion. As a result, the interpretation is
only concerned with the measurement of V1. Regarding , as
explained in reference (Lasseux et al., 2012), an increase in
the sensitivity of P1 to was necessary to ensure its accurate
estimation by history matching. This was achieved thanks to

February 2016

the major strength of the technique resulting from the fact

that P0 is an input of the history matching: the modulation
of the excitation. By emitting successive pulses of pressure,
the phenomenon of gas accumulation happening right after
the pulse emission and creating sensitivity to is generated
several times, compared to only once in a pulse-decay test.
The results from unsteady-state experiments were
compared to the results from a steady-state analysis carried
out with nitrogen. In the case of Pyro 5, characterization at
steady state was repeated with Isopar L. Figure 4 is a diagram
of a classical steady-state device. Regardless of the uid
used, one point of measurement was made by regulating
P0 and recording at steady-state P0, P1 and the volume ow
rate Qv. For all tests involving nitrogen, several points were
taken at increasing mean pore pressures to derive kl and b
independently. For the test with Isopar L, only one point
was needed to determine kl. The parameters chosen for the
characterization with nitrogen were:
Plugs of pyrophyllite: 40 bar P0 70 bar, P 1 = Patm
(atmospheric pressure) and Peff = 60 bar
Shale 1: 40 bar P 0 70 bar, P 1 = Patm and Peff = 100
bar
Shale 2: 6 bar P 0 11 bar, P 1 = Patm and Peff = 100
bar.

In the steady-state test with Isopar L on Pyro 5, P 0 was

set at 80 bar, P 1 at 2 bar and Peff at 60 bar.
The interpretation of the steady-state raw data recorded with
nitrogen relies on:
(1)
In this relationship recognized rst by Klinkenberg
(1941), k is calculated from Eq. 2, which was derived for
an isothermal steady-state gas ow by integrating the
differential form of Darcys law including the Klinkenberg
effect.

PETROPHYSICS

(2)

33

Proce et al.

where is the viscosity, L the length, S the cross-section

area, and Pm the mean pore pressure. It appears from
Klinkenbergs law that the separate estimation of kl and b
requires the plot of k versus 1/Pm. Indeed, this plot displays
a linear behavior with a slope equal to kl*b and an intercept
with the Y-axis equal to kl. For the characterization of kl with
Isopar L, the calculation is immediate with Darcys law:
(3)
Equation 3 is simply Darcys law.
Across all tests carried out with gas on a given plug,
Pp varied approximately between 1 and 70 bar. For these
extreme values of Pp, the mean free path dened by Eq. 4
is in the range [1 to 94 nm].
(4)
where is the ideal gas constant, T the absolute temperature
and M the molecular mass. For the previous range of and
for the pyrophyllite mean pore radius Rp of 25 nm (10 nm
Rp 40 nm), the Knudsen number Kn derived from Eq. 5 is
between 0.04 and 3.8.
.

(5)

Consequently, in pores having radii of a few tens of

nanometers, either a slip-ow regime or a transition-ow
regime arises depending on the level of pressure (Ziarani
and Aguilera, 2012). For the latter, both Darcys law and
Klinkenbergs law are seriously questioned since gas
molecules collide principally with pore walls and no longer
with other molecules.
Round-Robin Test
After characterization by Total, Pyro 1, Pyro2, Pyro
3 and Pyro 4 were sent successively to three commercial
laboratories for a round robin-test comprising unsteady-state
experiments and, when possible, steady-state experiments
too. The specications emphasized conducting them with
nitrogen, on the whole plugs (no crushing), by regulating
Peff at 60 bar and without carrying out any prior treatment
on the plugs (cleaning or drying) to prevent any alteration
of the pyrophyllite due to preparation. The laboratories were
free to choose their methods and experimental conditions
(pressures, ow rates, temperature...).
Laboratory 1 (Lab 1) performed unsteady-state tests with
the pulse-decay technique and steady-state tests following a
procedure similar to ours. The interpretation of their pulsedecay data relied on the methodology implemented in the

34

Pulse Decay Permeameter 200, except that the approximate

analytical solution was not that of Jones (1972) but of
Brace et al. (1968). The unsteady-state measurements were
started by increasing P0 by a few bars (between 2.5 and 3.5
bar), after the stabilization of Pp between 15.5 and 18 bar;
V0 and V1 were 58 and 47 cm3, respectively. Estimations
of kl were deduced from the tests. Each analysis at steadystate involved several points of measurement to enable the
separate identication of kl and b. The mean Pp chosen for a
given plug varied between a few bars (from 1.5 to 4 bar) and
20 bar. Both types of test were carried out at a temperature
of 24C.
Laboratory 2 (Lab 2) delivered values of kl and b from
unsteady-state experiments done with the Automated
Permeameter - 68 manufactured by Coretest Systems. Pp
was initially set at 14 bar. Once equilibrium was achieved,
the gas ow was started by opening v1. Hence, P1 was kept
at Patm and P0, recorded in V0 of 6 cm3, was the only signal
introduced in the history-matching procedure to obtain kl
and b. The numerical model assumed as usual the validity of
the Darcy-Klinkenberg law. The temperature was of 20C.
Laboratory 3 (Lab 3) provided estimations of kl from
unsteady-state tests run with a pulse-decay device. The plug
was rst pressurized at 40 bar and the pulse of pressure was
then generated by increasing P0 to 60 bar; V0 and V1 were
15 and 13 cm3, respectively. The model and assumptions
involved in the interpretation were not available. The
temperature reached 22C.
RESULTS
Results of the Comparative Studies
Table 1 lists the results of all measurements conducted
on the ve plugs of pyrophyllite. The estimations of
given by the step-decay method were compared to those
provided by pycnometers employing helium at low
pressure. The subscripts PD, SD, SSG, SSO and
Pyc in Table 1 and the rest of the paper refer to: pulse
decay, step decay, steady-state gas, steady-state oil
and pycnometry, respectively, Moreover, the deviation
indicator D quantifying the discrepancy between 1 and 2,
both estimations of (= kl, b or ), and used throughout the
present development is dened as:

PETROPHYSICS

(6)

February 2016

Low-Permeability Measurements: Insights

Table 1Results of the Comparative Measurements for the Pyrophyllite

Plugs

For Pyro 1 to Pyro 5, Dkl derived for kl,SD and kl,PD is in

the range of 5 to 17%, vs. 19 to 40% when calculated for
kl,SSG and kl,PD. Therefore, the values of kl from tests run with
gas, i.e. kl,PD, kl,SD and kl,SSG, agree in a satisfactory manner.
As shown in Table 1, this agreement can be extended to
estimations of kl from experiments carried out with gas and
oil, at least for pyrophyllite. Indeed, in the case of Pyro 5,
Dkl determined for kl,SSO and kl,SSG is about 26%. Regarding
b and , bSD compares well with bSSG as well as SD with Pyc,
Db ranging from 10 to 43 % and D from 0 to 19 %.
Figure 5a provides the normalized difference (P1,rec P1,sim)/P1,rec for Pyro 1 where P1,rec is the recorded prole P1
and P1,sim the prole P1 simulated with the step-decay model.
Identical graphs were obtained for Pyro 2 to Pyro 5. The
signal is at and centered on 0, excepted during a period
of around one hour at the beginning of the experiment. The
discrepancies between P1,rec and P1,sim observable in Fig. 5b
are due to the combination of two elements: the poor quality
of P1,rec at short times where P1,rec takes low values and the
perturbation of P1,rec right after the emission of a pulse of
pressure. Figure 6 is an example of graph where k is plotted
against 1/Pm, resulting from the processing of the steadystate raw data corresponding to Pyro 1. A similar linearity
was observed for Pyro 2 to Pyro 5.

February 2016

Fig. 5(a) Normalized difference between P1,rec and P1,sim versus time.
(b) P1,rec and P1,sim signals at short times.

Fig. 6k vs. 1/Pm.

Table 2 provides the results from the experiments

conducted on Shale 1 and Shale 2. As emphasized for
pyrophyllite plugs, the same gas-based techniques used
on shale plugs to identify a given property, kl or b, lead to
consistent estimates. Dkl is about 9% for Shale 1 and about
19% for Shale 2, when comparing kl,SSG and the mean of kl,SD
and kl,PD, while Db is about 21% for Shale 1 and about 26%
for Shale 2. The estimations of SD are not available since
the characterization of with the step-decay technique is
uncertain when dealing with shales. The bias in is mainly
due to the fact that one major assumption of the step-decay
model, which is the homogeneity of the studied rock, is rarely
respected in the presence of shales. In the case where the rock
is highly heterogeneous, it has been proven with numerical
simulations and practical tests that the method delivers an
accurate estimation of kl, an acceptable estimation of b and
an unreliable estimation of ( Proce, 2014).

PETROPHYSICS

35

Proce et al.

Table 2Results of the Comparative Measurements for the Shales

Results of the Round-Robin Test

Table 3 lists the values of kl obtained for Pyro 1, Pyro 2,
Pyro 3 and Pyro 4 by Total and by the three other laboratories
that participated in the round-robin test. Estimations of b were
provided by some of the laboratories but are not reported
here since the focus was primarily on the characterization of
kl. The values of kl delivered by the commercial laboratories
were compared to those determined by Total. To be more
precise, for a given plug, the value of kl estimated by Lab 1
at steady state was compared to kl,SSG while the values of kl
estimated by Lab 1 to Lab 3 at unsteady state were compared
to kl,SD. The agreement between the results collected by Lab
1 and Total is excellent since the discrepancies reach at most
a factor of 1.3 at steady state and a factor of 2 at unsteady
state. The results from Lab 2 are higher than Totals results
by a factor ranging from 2 to 4. While these discrepancies
remain acceptable, those observed when comparing the
results obtained by Lab 3 and Total are not tolerable anymore.
Indeed, up to a factor of 18 was noted between the two series
of results, Totals results being globally the lowest ones.
Table 3Results of the Round-Robin Test

the rock true permeability characterized? Figures 5a and

6 enable us to answer this question. Indeed, the accurate
description of gas-ow behavior by models built from the
Darcy-Klinkenberg law points out that reliable estimations
of kl are drawn from such models. Revising classical uid
mechanics is not justied when testing low-permeabilty
rocks, at least in the range of pore pressure considered in
this paper.
From the round-robin test, it arises that the values of
kl measured by different laboratories on identical plugs are
extremely dispersed. Figures 7 to 9 were plotted to check
if high discrepancies could be justied by the diversity
of interpretative models. For Pyro 1 specically, the
experimental signals recorded by Lab 2 and Lab 3 were
superimposed to the numerical signals simulated with the
step-decay model, using their parameters and estimations
of kl. As can be seen on Fig. 7, the matching between the
experimental and numerical proles P0 is clear for Lab
2. For Lab 3, Fig. 8 shows that the matching between the
recorded and simulated proles P1 is observed when the
plug pore volume is assumed to be at atmospheric pressure
prior to the pulse emission. Yet, Lab 3 maintains that it
pressurized the whole system (device and plug) at 40 bar
before each test. When accounting for this initial condition,
Fig. 9 demonstrates that the simulated P1 does not match
the recorded P1 anymore. The indications given by Lab 3
about its procedure are doubtful. Nonetheless, Figs. 7 and 8
conrm that the dispersion of the round-robin test results is
due to experimental biases rather than to the diversity of the
interpretative models.

DISCUSSION
From the tests on pyrophyllite and shale plugs, it
appears that an excellent agreement can be achieved
between estimations of kl from methods differing in terms of
the experimental procedure and the interpretative model. All
gas-based techniques used in Study 1 and Study 2 involve
models supposing that Darcys law and Klinkenbergs law
are still applicable to gas ow in low-permeabilty media.
However, as previously mentioned (refer to Methods section),
this assumption is debatable when working on nanoporous
materials such as pyrophyllite and shale. Therefore, a natural
question arises: with the selected gas-based methods, was

36

Fig. 7Lab 2 recorded and simulated proles P0.

PETROPHYSICS

February 2016

Low-Permeability Measurements: Insights

Fig. 8Lab 3 recorded and simulated proles P1 assuming that Pp,init

= Patm.

Fig. 9Lab 3 recorded and simulated proles P1 assuming that Pp,init =

40 bar.

To start with, estimations of kl higher than Totals results

by a factor of 2 or more cannot be explained by a possible
decrease of the initial water saturations of the plugs (refer
Plugs section). The discrepancies were thus engendered by
totally different causes. Working at high pore pressure was
the strategy adopted by Lab 1 and Lab 3 to eliminate the
Klinkenberg effect and then deduce intrinsic permeabilities
from analytical solutions neglecting the Klinkenberg effect.
Our tests revealed that all plugs of pyrophyllite have values

February 2016

of b close to 20 bar. With tests at a mean Pp of 20 bar, Lab 1

estimated values of kl which are approximately twice higher
than the true values of kl, according to Eq. 1. Regarding
the tests carried out by Lab 3, if Pp was effectively equal to
atmospheric pressure (and not to 40 bar) before the pulse
release, then these tests were done at a mean Pp of 10 bar
(and not of 50 bar as asserted by Lab 3). It results from Eq.
1 that Lab 3 measured values of kl overestimated by about
a factor of 3. The insufcient level of pore pressure alone
justies the discrepancies between the results from Total and
Lab 1 but only part of the discrepancies between the results
from Total and Lab 3. Moreover, note that being at pressure
equilibrium before exciting the plug is indispensable when
a history-matching procedure is used for the interpretation
knowing that the initial pressure within the plug acts as
an input datum. This remark applies to Lab 2 specically.
Indeed, Lab 2 pressurized the plug at around 14 bar before
each test but did not intend to eliminate the Klinkenberg
effect since it estimated b as well as kl by history matching.
The commercial device bought by Lab 2 automatically stops
the acquisition after three hours without any possibility to
increase the recording time even if the prole P0 continues
to be displayed on the computer screen. To ensure an
estimation of kl and b as accurate as possible, Lab 2 added to
the interpreted data a few points read on a screenshot of P0.
This explains that the matching between the recorded and
simulated proles P0 is incomplete in some areas (see Fig.
7) and hence part of the discrepancies between the results
found by Total and Lab 2. Lab 2 took another initiative when
running its history matching procedure, which is xing the
porosity at 0, since no porosity effect can be highlighted at
short times from a zoom on a prole P0 typically observed
by Lab 2. If the recording of P0 was effectively triggered at
the opening of v1, the decrease of P0 would start after a delay
time due to the gradual depressurization of the plug pore
volume before the perturbation reaches V0. This absence
of delay time is the signature of a preferential path for the
gas ow between V0 and V1. Lab 2 used pistons and sleeves
having a diameter of 25.4 mm with plugs having diameters
close to 23 mm. Gas leakages probably happened between
the sleeve and the plug, all the more that the effective pressure
chosen for the round-robin test was low. Lab 3 also worked
with pistons and sleeves of 25.4-mm diameter. To reduce the
space between the sleeve and the plug, it rst wrapped the
plug with Teon and then applied on the whole to a heatshrink sleeve. This practice certainly generated even more
gas leakages given that Teon acted as an additional porous
medium.

PETROPHYSICS

37

Proce et al.

CONCLUSIONS
Redundant estimations of kl can be obtained for a lowpermeabilty plug by using different methods of the industry.
To be more precise, steady-state and unsteady-state techniques
applied with gas provided results in excellent agreement for
pyrophyllite and shale. The convergence of the results from
all measurements with gas and from a measurement with a
liquid at steady-state is true for pyrophyllite. The verication
of this last point for shale is ongoing. The accurate prediction
of gas-ow behavior by interpretative models based on the
Darcy-Klinkenberg law proves that revising classical uid
mechanics when testing low-permeabilty rocks is useless.
Gas ow in tight formations can still be described as viscous
ow with slippage at pore walls. Furthermore, comparable
estimations of kl can be obtained by different laboratories
using their own techniques and interpretative models if the
experimental procedure is carefully dened. Indeed, the
divergence of the results found during a round-robin test is
rather explained by experimental biases than by the diversity
of the interpretative models, at least when plugs and not
crushed samples are involved. Several recommendations
should be respected to ensure the agreement of the values
of kl measured by laboratories analyzing an identical plug:
(1) Perform the test at a mean pore pressure high enough to
completely eliminate the Klinkenberg effect, if necessary;
(2) use pistons and a sleeve adapted to the plug to avoid
leakage between the sleeve and the plug; (3) develop an inhouse experimental device to fully control the procedure.
In the case of crushed samples or drill cuttings, the reasons
generating dispersion within the results from different
laboratories are much more difcult to identify (Passey et al.,
2010; Sondergeld et al., 2010), owing to the complexity of
the problem modeling and of the sample preparation (Tinni,
2012). Consequently, we advise conducting permeability
measurements on whole-core plugs as much as possible.

D
k
kl
kl, PD
kl, SD
kl, SSG
kl, SSO
Kn
L
M
P
Patm
Pc
Peff
Pm
Pp
Pp, init
P0
P1
P1, rec
P1, sim
Qv

Rp
S
T
Vb
vb
V0
v0
V1
v1
Log(Pcap)
P
SHg

, SD
, pyc

ACKNOWLEDGEMENTS
Total is gratefully acknowledged for the permission
to publish the results. Moreover, we would like to thank
Ghislain Pujol, Jean-Michel Kluska, Frdric Plantier, Sonia
Vincent-Gill and Antoine Delafargue for their contribution
to the work presented in this paper.
NOMENCLATURE
b
b, SD
b, SSG
Db
Dkl

38

= Klinkenberg coefcient, Pa
= b from the step-decay test, Pa
= b from the steady-state test with gas, Pa
= deviation indicator on b
= deviation indicator on kl

= deviation indicator on
= apparent permeability, m2
= intrinsic permeability, m2
= kl from the pulse-decay test, m2
= kl from the step-decay test, m2
= kl from the steady-state test with gas, m2
= kl from the steady-state test with oil, m2
= Knudsen number
= length, m
= molecular mass, kg.mol-1
= pressure, Pa
= atmospheric pressure, Pa
= conning pressure, Pa
= effective pressure, Pa
= mean pore pressure, Pa
= pore pressure, Pa
= initial pore pressure, Pa
= upstream pressure, Pa
= downstream pressure, Pa
= recorded P1, Pa
= simulated P1, Pa
= volume ow rate, m3.s-1
= ideal gas constant, J.mol-1.K-1
= pore radius, m
= cross-section area, m2
= absolute temperature, K
= buffer tank volume, m3
= buffer tank valve
= upstream tank volume, m3
= upstream tank valve
= downstream tank volume, m3
= downstream tank valve
= variation of the logarithm of the capillary
pressure
= differential pressure, Pa
= variation of the mercury saturation
= molecular mean free path, m
= dynamic viscosity, Pa.s
= porosity
= from the step-decay test
= from the pycnometry test
REFERENCES

Boulin, P.F., Bretonnier, P., Gland, N., and Lombard, J.M., 2012,
Contribution of the Steady-State Method to Water Permeability
Measurement in Very Low Permeability Porous Media,
Oil and Gas Science and Technology - Revue dIFP Energies
nouvelles, 67(3), 387401. DOI: 10.2516/ogst/2011169.
https://hal-ifp.archives-ouvertes.fr/hal-00727819/document.
Accessed January 28, 2016.

PETROPHYSICS

February 2016

Low-Permeability Measurements: Insights

Brace, W.F., Walsh, J.B., and Frangos, W.T., 1968, Permeability of

Granite Under High Pressure, Journal of Geophysical
Research, 73(6), 22252236. DOI: 10.1029/JB073i006p02225
Bruce, G.H., Peaceman, D.W., Rachford H.H., Jr., and Rice, J.D.,
1953, Calculations of Unsteady-State Gas Flow Through
Porous Media, Paper SPE-221-G, Journal of Petroleum
Technology, 5(3), 7992. DOI: http://dx.doi.org/10.2118/221-G
Civan, F., 2009, Effective Correlation of Apparent Gas Permeability
in Tight Porous Media, Transport in Porous Media, 82(2),
375384. DOI: 10.1007/s11242-009-9432-z
Fathi, E., Tinni, A., and Akkutlu, I.Y., 2012, Shale Gas Correction
to Klinkenberg Slip Theory, Paper SPE-154977 presented
at the SPE Americas Unconventional Resources Conference,
Pittsburgh, Pennsylvania, USA, 57 June. DOI: http://dx.doi.
org/10.2118/154977-MS
Hsieh, P.A., Tracy, J.V., Neuzil, C.E., Bredehoeft, J.D., and Silliman,
S.E., 1981, A Transient Laboratory Method for Determining
the Hydraulic Properties of Tight Rocks1. Theory,
International Journal of Rock Mechanics and Mining Sciences
& Geomechanics Abstracts, 18(3), 245252.
Javadpour, F., 2009, Nanopores and Apparent Permeability of Gas
Flow in Mudrocks (Shales and Siltstones), Paper PETSOC09-08-16-DA, Journal of Canadian Petroleum Technology,
48(8), 1621. DOI: http://dx.doi.org/10.2118/09-08-16-DA
Jones, S.C., 1972, A Rapid Accurate Unsteady-State Klinkenberg
Permeameter, Paper SPE-3535, SPE Journal, 12(5), 383397.
DOI: http://dx.doi.org/10.2118/3535-PA
Karniadakis, G., Beskok, A., and Aluru, N., 2005, Microows
and
Nanoows,
Fundamentals
and
Simulation,
Interdisciplinary Applied Mathematics 29, Springer-Verlag.
ISBN: 978-0-387-22197-7
Klinkenberg, L.J., 1941, The Permeability of Porous Media to
Liquids and Gases, API Drilling and Production Practice 1,
200213.
Lasseux, D., Jannot, Y., Proce, S., Mallet, M., and Hamon,
G., 2012, The Step Decay: A New Transient Method for the
Simultaneous Determination of Intrinsic Permeability,
Klinkenberg Coefcient and Porosity on Very Tight Rocks,
Paper SCA2012-25 presented at the SCA International
Symposium, Aberdeen, Scotland, 2730 August.
Luffel, D.L., and Guidry, F.K., 1992, New Core Analysis Methods
for Measuring Reservoir Rock Properties of Devonian Shale,
Paper SPE-20571, Journal of Petroleum Technology, 44(11),
11841190. DOI: http://dx.doi.org/10.2118/20571-PA
Passey, Q.R., Bohacs, K.M., Esch, W.L., Klimentidis, R., and
Sinah, S., 2010, From Oil-Prone Source Rock to GasProducing Shale ReservoirGeologic and Petrophysical
Characterization of Unconventional Shale Gas-Reservoirs,
Paper SPE-131350 presented at the International Oil & Gas
Conference and Exhibition, Beijing, China, 810 June. DOI:
http://dx.doi.org/10.2118/131350-MS
Proce, S., 2014, Mesure de Proprits Monophasiques de Milieux
Poreux Peu Permables par Voie Instationnaire, Bordeaux
University, France, PhD Dissertation. http://www.theses.
fr/2014BORD0142
Rushing, J.A., Newsham, K.E., Lasswell, P.M., Cox, J.C., and
Blasingame, T.A., 2004, Klinkenberg-Corrected Permeability

February 2016

Measurements in Tight Gas Sands: Steady-State Versus

Unsteady-State Techniques, Paper SPE-89867 presented at the
SPE Annual Technical Conference and Exhibition,
Houston, Texas, USA, 2629 September. DOI: http://dx.doi.
org/10.2118/89867-MS
Sondergeld, C.H., Newsham, K.E., Comisky, J.T., Rice, M.C., and
Rai, C.S., 2010, Petrophysical Considerations in Evaluating
and Producing Shale Gas Resources, Paper SPE-131768
presented at the SPE Unconventional Gas Conference,
Pittsburgh, Pennsylvania, USA, 2325 February. DOI: http://
dx.doi.org/10.2118/131768-MS
Tinni, A.O., 2012, Permeability Measurements of Nanoporous
Rocks, University of Oklahoma, Norman Oklahoma, Masters
thesis.
Wang, Y., and Knabe, R.J., 2010, Permeability Characterization on
Tight Gas Samples Using Pore Pressure Oscillation Method,
Paper SCA2010-30 presented at the SCA International
Symposium, Halifax, Nova Scotia, Canada, 47 October.
Ziarani, A.S., and Aguilera, R., 2012, Knudsens Permeability
Correction for Tight Porous Media, Transport in Porous
Media, 91(1), 239260. DOI: 10.1007/s11242-011-9842-6

ABOUT THE AUTHORS

Sandra Proce is a reservoir engineer
with Total, currently working on the
measurement of petrophysical properties
(permeability, porosity, saturation...)
of conventional and unconventional
rocks. She holds a PhD degree in uid
mechanics from University of Bordeaux
(France).
Benjamin Nicot is now a petrophysicist
with Total involved in formation
evaluation. He holds a PhD degree in
nuclear magnetic resonance from IFP
Energies nouvelles (France) and worked
for seven years for Schlumberger as
a NMR senior research scientist and
a Petrophysics domain champion.
Specialized in NMR, both laboratory and log, he is also
involved in formation evaluation techniques in general and
log interpretation.

PETROPHYSICS

Grald Hamon is currently expert for

Petrophysics with Total. He is actively
involved in working with Totals assets
to design and implement data acquisition
programs as well as in reviewing data
about formation evaluation and recovery
processes for reserves evaluation.
Gerald is also leading several research

39

Proce et al.

projects in enhanced oil recovery and formation evaluation.

He holds a PhD degree from Institut National Polytechnique
de Grenoble (France). He received the 2007 Technical
Achievement Award from the SCA for his major contribution
in core analysis.

40

PETROPHYSICS

February 2016