JID:EPSL AID:14555 /SCO [m5G; v1.221; Prn:26/07/2017; 10:56] P.

1 (1-10)
Earth and Planetary Science Letters ()

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

www.elsevier.com/locate/epsl

Transient radon signals driven by uid pressure pulse, micro-crack

closure, and failure during granite deformation experiments
Frdric Girault a,b, , Alexandre Schubnel a , ric Pili c
a
Laboratoire de Gologie de lENS PSL Research University UMR 8538 CNRS, F-75005 Paris, France
b
Institut de Physique du Globe de Paris, Sorbonne Paris Cit, University Paris Diderot, UMR 7154 CNRS, F-75005 Paris, France
c
CEA, DAM, DIF, F-91297 Arpajon, France

a r t i c l e i n f o a b s t r a c t

Article history: In seismically active fault zones, various crustal uids including gases are released at the surface. Radon-
Received 15 April 2017 222, a radioactive gas naturally produced in rocks, is used in volcanic and tectonic contexts to illuminate
Received in revised form 2 July 2017 crustal deformation or earthquake mechanisms. At some locations, intriguing radon signals have been
Accepted 4 July 2017
recorded before, during, or after tectonic events, but such observations remain controversial, mainly
Available online xxxx
Editor: J. Brodholt
because physical characterization of potential radon anomalies from the upper crust is lacking.
Here we conducted several month-long deformation experiments under controlled dry upper crustal
Keywords: conditions with a triaxial cell to continuously monitor radon emission from crustal rocks affected by three
tri-axial stress main effects: a uid pressure pulse, micro-crack closure, and differential stress increase to macroscopic
radon emission failure. We found that these effects are systematically associated with a variety of radon signals that
transient signals can be explained using a rst-order advective model of radon transport. First, connection to a source of
permeability changes deep uid pressure (a uid pressure pulse) is associated with a large transient radon emission increase
pre-seismic signals
(factor of 37) compared with the background level. We reason that peak amplitude is governed by the
macroscopic failure
accumulation time and the radon source term, and that peak duration is controlled by radioactive decay,
permeability, and advective losses of radon. Second, increasing isostatic compression is rst accompanied
by an increase in radon emission followed by a decrease beyond a critical pressure representing the
depth below which crack closure hampers radon emission (150250 MPa, ca. 5.59.5 km depth in our
experiments). Third, the increase of differential stress, and associated shear and volumetric deformation,
systematically triggers signicant radon peaks (ca. 25350% above background level) before macroscopic
failure, by connecting isolated cracks, which dramatically enhances permeability.
The detection of transient radon signals before rupture indicates that connection of initially isolated
cracks in crustal rocks may occur before rupture and potentially lead to radon transients measurable
at the surface in tectonically active regions. This study offers thus an experimental and physical basis for
understanding predicted or reported radon anomalies.
2017 Elsevier B.V. All rights reserved.

1. Introduction
Various uids produced in the Earths crust migrate through
fault zones and are released at the surface in seismically active
regions worldwide (e.g., Irwin and Barnes, 1980; Sato et al., 1986;
Sugisaki et al., 1983; Kennedy et al., 1997; Perrier et al., 2009;
Walia et al., 2010). One of those uids, radon-222, is a radioac-
tive noble gas of half-life of 3.82 days and is naturally produced
in rocks. An alpha-decay product of radium-226 in the uranium-
238 decay chain, radon-222 becomes detectable when it reaches
*
Corresponding author at: Institut de Physique du Globe de Paris, Sorbonne Paris
Cit, University Paris Diderot, UMR 7154 CNRS, F-75005 Paris, France.
E-mail address: girault@ipgp.fr (F. Girault).
connected pores or cracks and then exits the porous network. The
ability of a rock to emit radon depends thus on the bulk radium
concentration in the grains and on the emanation factor (Tanner,
1964; Sakoda et al., 2011), which accounts for radium distribu-
tion and geometry of the porous network. Radon transport from
its source to the surface is governed by diffusion, advection and/or
convection (Nazaroff, 1992). Due to its relatively short half-life, ad-
vection is best able to carry radon-222 from crustal layers to the
surface fast enough to be measured before its decay. Therefore,
radon-222 is generally used to track uid migration in tectonic
and volcanic geosystems characterized by high permeability and
shallow processes (e.g., Trique et al., 1999; Cigolini et al., 2007;
Girault et al., 2014). As a chemical probe to look for physical
phenomena in the upper crust, radon-222 is sometimes consid-

http://dx.doi.org/10.1016/j.epsl.2017.07.013
0012-821X/ 2017 Elsevier B.V. All rights reserved.
JID:EPSL AID:14555 /SCO [m5G; v1.221; Prn:26/07/2017; 10:56] P.2 (1-10)
2 F. Girault et al. / Earth and Planetary Science Letters ()

ered as a potential earthquake precursor (Virk and Singh, 1994; Table 1

Igarashi et al., 1995). However, such intriguing observations remain Characteristics of the PL and AL samples including major elements concentrations,
238
U, 226 Ra, and 232 Th activity concentrations, radon-222 source term and emana-
controversial (Geller, 2011), mainly because physical characteriza-
tion, porosity and permeability.
tion of potential radon signals from the upper crust is lacking.
Many transient signals of radon-222 concentration in soil- Sample characteristics Portugal Allaire
leucogranite (PL) leucogranite (AL)
gas or in water have been recorded in the eld. Nevertheless,
only few of them are accepted by the scientic community as Concentrations of major elements (wt.%)a
SiO2 74.19 72.91
likely related to short time scale deformation processes such as Al2 O3 13.52 14.76
lake-level variations, volcanic activity and earthquakes (Cox et Fe2 O3 (T) 1.8 1.5
al., 1980; Wakita et al., 1980, 1991; Igarashi et al., 1993, 1995; MnO 0.05 0.02
Virk and Singh, 1994; Trique et al., 1999; Richon et al., 2003; MgO 0.29 0.33
CaO 0.89 0.58
Steinitz et al., 2003; Wang and Manga, 2010). All of these sig-
Na2 O 3.2 3.18
nals show a relatively narrow peak of duration between 1 day K2 O 4.5 4.5
and 1.5 week and amplitudes between 1.5 to 28 times the back- TiO2 0.196 0.197
ground levels. This peak is frequently characterized by a rapid P2 05 0.18 0.26
increase and a more progressive decrease. However, at these gen- LOI 0.6 1.41
Total 99.41 99.64
erally poorly controlled natural sites, the source of the radon sig-
nals remains dicult to constrain. As the number of natural sites Activity concentrations (Bq kg1 )b
instrumented with radon sensors installed near the surface is in- U-238 368 13 160 17
Ra-226 432 125 210 76
creasing worldwide, a better understanding of the radon concen-
Th-232 67.0 5.7 43 10
tration time-series is required. Relevant rock deformation experi-
ments have been conducted in the laboratory since the late 1970s Radon-222 source term and emanation
Effective radium-226 4.9 0.2 11.2 0.8
(Holub and Brady, 1981 and references herein; King and Luo, 1990;
concentration EC Ra (Bq kg1 ) (heat-treated)
Tuccimei et al., 2010; Mollo et al., 2011; Nicolas et al., 2014; Emanation factor E (%) 1.1 0.3 5.3 2.0
Koike et al., 2015). These studies proposed formation and growth (heat-treated)
of micro-cracks, formation of new grain surfaces or connection Physical parameters
of isolated micro-pores as the main effects leading to anoma- Connected porosity <0.02 <0.02 (heat-treated)
Axial permeability k (m2 ) at (1.5 0.1) 1018 (6.1 0.3) 1018
lous radon signals. However, little physical characterization has conning pressure of 10 MPa (heat-treated)
been attempted. For instance, the potential role of permeability
a
Determined by Fusion-Inductively Coupled Plasma (FUS-ICP).
changes (Manning and Ingebritsen, 1999; Manga et al., 2012) has b
Determined by gamma spectrometry, Ra-226 activity being determined by its
never been investigated. In addition, all but one (Nicolas et al., own ray.
2014) of the experimental studies was performed under uniax-
ial compression (i.e., different from natural conditions better rep-
resented using triaxial systems). Some studies were carried out In this study, we used two granitic rock samples typical of the
on poorly characterized samples (Holub and Brady, 1981; Belikov upper crust, a leucogranite from Portugal (PL) and a leucogranite
et al., 2014) or with only intermittent (Tuccimei et al., 2010; from Allaire, France (AL). The main characteristics of the gran-
Mollo et al., 2011) to daily (Scarlato et al., 2013; Nicolas et al., ite samples are summarized in Table 1. Their 238 U, 232 Th, and
2014) radon monitoring. Higher time resolution is needed to cap-
226
Ra activity concentrations are 368 13, 67.0 5.7, and 432
ture and physically characterize the short time scales of such tran- 125 Bq kg1 for PL, and 160 17, 43 10, and 210 76 Bq kg1
sient signals. for AL, respectively. To articially increase its initial permeability,
In this paper, we present several month-long deformation ex- AL sample was slowly heated to 575 C in order to induce thermal
periments performed on granitic rock samples with a triaxial oil- cracks (Darot et al., 1992). The radon source term EC Ra and ema-
medium cell. We studied the effects of a uid pressure pulse, nation factor E of the samples are 4.9 0.2 Bq kg1 and 1.1 0.3%
micro-crack closure, and failure on radon emission from crustal for PL, and 11.2 0.8 Bq kg1 and 5.3 2.0% for AL, respectively.
rocks. Physical characterization of the radon concentration time- The PL sample has larger 226 Ra activity concentration, smaller EC Ra
series is proposed as a basis to clarify predicted or reported tran- and smaller permeability than the heat-treated AL sample. All ex-
sient radon signals, in particular those of a pre-seismic nature. periments were conducted on cylindrical cores of diameter 4 cm
and length 8 cm, stored at 40 C for at least a week before the
2. Radon-222 source term, sample description, and experimental start of an experiment and jacketed in a neoprene sleeve during
system experiments.
Month-long experiments were conducted under controlled dry
The radon-222 source term of a sample, called the effective upper crustal conditions using a triaxial oil-medium cell designed
radium-226 concentration (EC Ra ), is dened by the product of the for long-term experiments (Schubnel et al., 2005; Fortin et al.,
bulk radium-226 concentration C Ra and the emanation factor E. It 2005). The system (Fig. 1) allows placing the sample under nat-
accounts for the ability of radon-222 to be produced within the ural conditions, controlling conning pressure (i.e., setting depth)
sample and to reach the connected pore space, which is a pre- and axial stress using two independent volumetric servo-pumps,
requisite to escape from the porous network (Sakoda et al., 2011). and controlling pore pressure using a servo-controlled delivery
To measure EC Ra from intact and fractured samples, we applied system of pressurized argon. The pore uid system can be set
the classical accumulation method (Ferry et al., 2002), placing the either to maintain a constant pore uid pressure or to inject a
sample in a cylindrical container attached to an ionization cham- pore uid volume at a constant rate. In our experiments with rel-
ber (AlphaGUARD, Saphymo, Germany) and correcting for possi- atively low permeability granite samples, the pore pressure was
ble leakage in the system (Nicolas et al., 2014). Uncertainties in- set constant, and thus the injected volume of argon varied. Room
clude experimental uncertainty related to the accumulation curve temperature was maintained quasi-constant (1 C) during the ex-
(<1%) and repeatability (5%), and common overall uncertainty of periment.
the instrument (4%). The emanation factor E is then inferred from Axial and radial deformation was continuously recorded using
EC Ra /C Ra , C Ra being determined using gamma spectrometry. two pairs of strain gauges glued to the sample surface. The total
JID:EPSL AID:14555 /SCO [m5G; v1.221; Prn:26/07/2017; 10:56] P.3 (1-10)
F. Girault et al. / Earth and Planetary Science Letters () 3

Fig. 1. (a) Sketch and (b) photograph of the experimental system. The triaxial oil-medium cell designed for long-term experiments at the Ecole Normale Suprieure is here
coupled with an ionization chamber to measure the radon concentration driven out of the sample by argon ow during deformation experiments.

Fig. 2. Overview of transient radon signals during a month-long deformation experiment on a granite sample under crustal conditions. The experiment was performed on
an 8-cm-length 2-cm-radius cylindrical, 575 C heat-treated AL sample at 1 MPa pore pressure. From bottom to top, time-series of measured radon concentration, calculated
axial gas permeability, and effective conning pressure and differential stress are shown as functions of time. In the course of the experiment, radon concentration, measured
at the exit of the sample, is controlled by: (1) a uid pressure pulse, (2) increase of the effective conning pressure, and (3) increase of the differential stress to macroscopic
failure (i.e., rupture) of the sample.

deformation was corrected from the apparatus stiffness using a lin-
ear variable displacement transducer (LVDT). We deduced axial and
volumetric strains to account for sample deformation, adopting the
convention that compressive stresses and compactive strains are
positive. P-wave velocity was measured along the axis of compres-
sion using a high-frequency ultrasonic monitoring system. Acoustic
emissions were also recorded during the experiment. Axial gas per-
meability of the core sample is deduced from the argon ow rate
and the pore pressure measured by the servo-controlled delivery
system of pressurized argon. As pore pressure was set at 2 MPa
maximum, permeability calculation is corrected for the Klinken-
berg effect (Ziarani and Aguilera, 2012).
The system can be described as a controlled-leak system (Holub
and Brady, 1981), since pore pressure was set at 1 to 2 MPa at
the top of the sample and at atmospheric pressure (0.1 MPa) at
the bottom of the sample (Fig. 1). Argon permeates through the
sample, driving the naturally produced radon-222 from the con-
nected pores and cracks of the sample before being released at
the bottom of the sample. The gas ow then enters the ioniza-
tion chamber (AlphaGUARD) where radon-222 concentration of
JID:EPSL AID:14555 /SCO
4 F. Girault et al. / Earth and Planetary Science Letters ()
the ow is measured every 10 min, and nally escapes from the
system.
3. Results of the deformation experiments
In order to study potential effects of a uid-pressure pulse,
elastic crack closure during burial and brittle fracturing on radon
emission, we adopted an experimental protocol (Fig. 2) which con-
sists of: (1) increased pore uid pressure (i.e., opening of argon
ow), followed by (2) isostatic compression increase and nally
(3) differential stress increase to sample rupture. First, when ar-
gon ow begins, a signicant peak in radon concentration is mea-
sured at the exit of the sample, with a rapid increase and more
gradual decrease. Second, the increase of conning pressure from
10 to 250 MPa, representing ca. 0.35 to 9.5 km depth, decreases
Fig. 3. Transient radon signals caused by uid pressure pulses for two granite sam-
ples. From bottom to top: time-series of measured radon concentration and radon
concentration calculated from model predictions as solid gray lines (see subsec-
tion 4.1 for model description) and measured ow rate of argon through the sample.
AL and PL samples have different effective radium-226 concentration EC Ra (radon-
222 source term) and axial gas permeability k (pore pressure 1 MPa).
Fig. 4. Transient radon signals caused by isostatic pressure increase for a PL sample. This experiment was performed under 1 MPa pore pressure. From bottom to top:
time-series of measured radon concentration, calculated axial gas permeability, and effective conning pressure and axial P-wave velocity. Radon concentration time-series
is shown in gray and the 1-hour smoothing average in black. Gray area and arrows highlight the time of signicant radon concentration decreases at the highest conning
pressure values.
[m5G; v1.221; Prn:26/07/2017; 10:56] P.4 (1-10)
permeability over two orders of magnitude and intensies radon
concentration by one order of magnitude. However, at 250 MPa
conning pressure, radon concentration decreases (Fig. 2). Third,
after decreasing conning pressure to 75 MPa (Fig. 2), the in-
creased differential stress is associated with several variations of
radon concentration, with detectable increase near sample rupture
and large decrease after rupture (Fig. 2).
When the argon ow is connected to the system (pore pres-
sure increase, time = 0 in Fig. 3), a signicant pulse of radon
concentration is produced for all granite samples. Compared with
the steady state radon concentrations measured after the pulse
(approximately 200 Bq m3 ), this transient peak is about 3 to 7
times higher. The rapid increase of radon concentration is simi-
lar in both experiments, but for the AL (more permeable) sample,
peak amplitude and subsequent decrease are larger, while peak
width is smaller than for the PL sample. The amplitude and de-
cay of the peak correlate positively and negatively with sample
permeability, respectively (Fig. 3). The observed decrease of the
transient radon signals exceeds the rate of radioactive decay of
radon-222.
Increasing isostatic pressure (Fig. 4) produces elastic volumet-
ric compaction and elastic crack closure, associated with a large
decrease in permeability and increase in P-wave velocity. This cor-
relates with an initial increase in radon concentration, followed by
a decrease beyond a critical pressure. The decrease in radon con-
centration is rst observed in Fig. 4 at a conning pressure of
150 MPa and is enhanced at a conning pressure of 250 MPa.
In Fig. 2, the initial decrease is observed at a larger conning
pressure (250 MPa). In each instance, the critical pressure is as-
sociated with similar permeability values of k < 1019 m2 . Look-
ing more closely at the radon concentration time-series, we see
that each step of increasing conning pressure initially produces
a slight decrease of radon concentration. The smaller the perme-
ability, the larger this transient decrease in radon concentration
(Figs. 2 and 4).
A signicant increase in radon concentration is measured im-
mediately following macroscopic failure of the sample (Fig. 5 and
Supplementary Fig. S1). This peak is observed in all experiments.
However, its amplitude and duration differ. The sample rupture
is characterized by a nearly 10-fold permeability increase, a large
JID:EPSL AID:14555 /SCO [m5G; v1.221; Prn:26/07/2017; 10:56] P.5 (1-10)
F. Girault et al. / Earth and Planetary Science Letters () 5

Fig. 5. Transient radon signals caused by differential stress increase for two granite samples. Experiments were performed under 1 MPa pore pressure with (a) a 575 C
heat-treated AL sample of low permeability (2 1019 m2 ) and (b) a PL sample of higher permeability (1.3 1018 m2 ). Effective conning pressure was set to 75 MPa in
(a) and 20 MPa in (b). From bottom to top: time-series of measured radon concentration, calculated axial gas permeability and volumetric strain, and differential stress and
cumulative acoustic emissions. Radon concentration time-series is shown in gray and the 1-hour smoothing average in black. Numbers refer to events leading to permeability
changes. The gray vertical line identies the time of sample rupture. Model predictions (see subsection 4.1 for model description) are shown as orange dashed lines (see
text). (For interpretation of the references to color in this gure, the reader is referred to the web version of this article.)

increase in the number of acoustic emissions, and signicant de-
creases of P-wave velocity and volumetric strain. Rupture of a low
permeability AL sample causes a 25 5% to 40 10% radon
concentration increase over about half a day (Fig. 5a and Sup-
plementary Fig. S1a). The preceding increments of increasing dif-
ferential stress cause only small radon concentration increases,
generally dicult to discriminate from the background level but
apparently associated with small permeability changes and acous-
tic emissions. Rupture of a granite sample with larger permeability
causes a 350 100% radon concentration increase over about 4 h
(Fig. 5b). In this case, the preceding increments of increasing dif-
ferential stress, which are associated with small enhancements of
permeability, acoustic emissions and axial and volumetric strains,
are easily detectable in the radon concentration time-series. Fi-
nally, rupture of an impermeable granite sample shows a delayed
but extremely large radon concentration increase (Supplementary
Fig. S1b).
4. Physical characterization of the radon anomalies
We have shown that (1) a uid pressure pulse, (2) isostatic
pressure increase, and (3) differential stress increase to macro-
scopic failure have a signicant impact on radon emission, trig-
gering a variety of radon signals. We now characterize these time-
series using a rst-order physical model of radon transport.
4.1. Physical model of radon transport in the experiments
In our experiments, radon-222 migrates along the connected
cracks of the granite sample, mainly driven by argon ow. No wa-
ter is present in the system. The Pclet number, which accounts
for the relative importance of advection versus kinematic disper-
sion, is always above unity. Thus, the radon transport is mainly
governed by advection and, at rst order, kinematic dispersion can
be neglected. Likewise, molecular diffusion process is disregarded.
JID:EPSL AID:14555 /SCO
6 F. Girault et al. / Earth and Planetary Science Letters ()
We resolve numerically (by sampling time and space) the gen-
eral radon transport equation in one dimension that includes ad-
vection, a production term, and radioactive decay (Nazaroff, 1992;
Girault and Perrier, 2014):
 
C (x, t ) u (t ) C (x, t )
= + C C (x, t ) ,
t x
where C is the radon concentration in the pore space (Bq m3 )
that evolves with time t (s) and along the longitudinal axis x of the
core (m), u is the Darcy velocity (m s1 ) that can vary with time,
is the connected porosity, is the radioactive decay of radon-222
(2.1 106 s1 ), and C = EC Ra (1 )/ is the radon concen-
tration in the pore space at equilibrium (Bq m3 ), with the bulk
density of the rock sample (kg m3 ). Darcy velocity is expressed by
u (t ) = k(t ) P /( L ), where k is the permeability (m2 ) that can vary
with time,  P is the pressure gradient (Pa), is the kinematic
viscosity of argon (Pa s), and L is the sample length. Under steady-
state conditions, the exit radon concentration at the bottom of the
sample is C ( L ) = EC Ra L 2 /(k P ). We then obtain expressions
for the radon concentration and ux at the bottom surface of the
sample as a function of time. This ux is conserved along the tub-
ing of the system until the gas ow enters the ionization chamber.
Therefore, the ionization chamber of known volume continuously
measures the radon concentration of the gas ow. We thus intro-
duce a mixing equation to account for dilution, radioactive decay,
advective losses, and the leakage of the tubing and the ionization
chamber:
 
C (t ) S S core
= (1 + aV ) + u (t ) C (t ), (2)
t V V
where S = EC Ra L is the radon ux from the bottom surface
of the sample (Bq m2 s1 ), V is the ionization chamber volume
(560 cm3 ), aV is the normalized leakage (or ventilation) rate de-
ned as aV = exp / 1, with exp being the decay constant of
the experiment, and S core is the surface area of the core sample
(m2 ). At the bottom surface of the sample (when x = L), we resolve
Eq. (2) after resolving Eq. (1) as a function of time to calculate
model predictions of radon concentration at the exit of the sample.
Based on this rst-order advective radon transport model, we can
test the effects of temporal changes of permeability k (calculated
during the experiment), and radon source term EC Ra (measured
before and after the experiment) on radon concentration. In the
following, we compare the model predictions directly with the
radon concentration time-series measured by the ionization cham-
ber in our experiments.
4.2. Characterization of radon signals during uid pressure pulse
The transient radon peaks obtained during the uid pressure
pulse experiments can be accounted for by this advective model
presented above in subsection 4.1 (Fig. 3). Although they have sim-
ilar connected porosity, the AL sample has a higher radon source
term and larger permeability than the PL sample. The amplitude
of the peak depends mainly on the length of time that the sam-
ple spends in the connement chamber of the loading cell (i.e.,
accumulation time before the pulse) and on the radon source term
of the sample. The width of the peak is predominantly controlled
by the radon-222 decay rate, the permeability of the sample and
radon losses. Since the accumulation time, radon source term and
permeability are constrained experimental quantities, the radon
losses (Eq. (2)) can be determined under these experimental condi-
tions. Furthermore, an instantaneous doubling of the pore pressure
(i.e., doubling argon ow rate) halves radon concentration (Supple-
mentary Fig. S2), as expected (Eq. (1)). In this case, the transient
decrease of radon concentration is faster than the radon-222 decay
and represents leakage of the experimental system.
[m5G; v1.221; Prn:26/07/2017; 10:56] P.6 (1-10)
The uid pressure pulse systematically triggers a large posi-
tive radon concentration peak. This experimental observation can
be directly related to some observations made in tectonically ac-
tive regions, particularly in compressional regimes. Models for uid
ow in fault zones (Rice, 1992; Byerlee, 1993), and in particu-
(1) lar the fault-valve model (Sibson, 2014), postulate that sudden
connection to a deep lithostatic pore uid pressure reservoir can
initiate a pulse of uid that migrates along connected faults and
fractures towards the surface. The carrier uid for radon can be
water (dissolved radon) or gases such as carbon dioxide, methane,
hydrogen, helium, or a mixture of these. Water ow can also push
accumulated radon in the connected pores (Trique et al., 1999;
Pili et al., 2004). Associated pore pressure migration may trigger
aftershocks, as proposed in various tectonically active contexts: a
decade ago in Italy (Miller et al., 2004) and more recently in China
(Liu et al., 2014), Japan (Terakawa et al., 2013) and West Bohemia
(Fischer et al., 2017).
4.3. Characterization of radon signals during micro-crack closure
Isostatic pressure increase initially decreases permeability and
increases radon concentration (Figs. 2 and 4). The permeability de-
crease is associated with increased volumetric strain and P-wave
velocity owing to the progressive closure of micro-cracks initially
present in the sample (Mitchell and Faulkner, 2008). Nevertheless,
due to the pore pressure (1 MPa minimum) and initial thermal
cracking by heat treatment (Darot et al., 1992), the granite samples
retain some permeability even at high effective pressure (Brace et
al., 1968). At this stage, using the advective radon transport model
described in subsection 4.1, the increase in radon concentration is
simply due to decreasing argon ow rate, and gradually increasing
transit time of the argon ow (the carrier of radon in our ex-
periments) through the sample, which mainly depends on sample
permeability. As permeability decreases, less argon ows through
the sample, so that carried radon is less diluted in the ow and
has more time to accumulate, consistently with previous consider-
ations (Girault and Perrier, 2014).
However, when permeability reaches 1019 m2 , additional de-
creases in permeability also correspond to decreased radon con-
centration. This is probably due to the conjunction of two phenom-
ena. (1) Under these conditions, the surfaces able to emit radon
into the porous network of the sample (i.e., radon exhaling sur-
faces) are drastically reduced, hence promoting radon trapping in
the adjacent grains (e.g., Sakoda et al., 2011) and decreasing radon
emission from the sample. (2) At these low permeability values,
the transit time of argon through the sample (about 1 day), is suf-
ciently large to affect the radon concentration time-series because
it is of the same order of magnitude as the radon-222 half-life (3.8
days). This observation suggests that, in the absence of uid path-
ways such as faults or fractures, at effective pressure higher than
150 MPa for the PL sample and higher than 250 MPa for the AL
sample (>5.5 km and >9.5 km depth, respectively), it may be dif-
cult for radon to be transported by advection. At larger depth, the
percolation threshold is not reached. The presence of permeable
faults or fractures may be a prerequisite for rapid radon transport
from deep sources.
4.4. Characterization of radon signals during differential stress increase
to rupture
The rst stages of applied stress slightly decrease perme-
ability (Fig. 5a and Supplementary Fig. S1a), due to the clo-
sure of pre-existing cracks oriented generally perpendicular to
the axial stress. The subsequent stages increase permeability
due to formation of dilatant cracks parallel to the axial stress
JID:EPSL AID:14555 /SCO
F. Girault et al. / Earth and Planetary Science Letters ()
(Mitchell and Faulkner, 2008). Increasing acoustic emission (Fig. 5b)
is also likely associated with the opening of new dilatant cracks
(Scholz, 1968). For a low permeability sample (Supplementary
Fig. S1b), permeability increase at the rupture point opens a path-
Fig. 6. Details of transient radon signals near the rupture of two granite samples
from the 1-hour smoothing average radon concentration time-series of (a) Fig. 5a
and (b) Fig. 5b. Time is corrected based on the transit time of argon through the
sample, knowing argon compressibility, sample permeability and experimental con-
ditions before rupture (in blue) and after rupture (in red). The gray vertical line
identies the rupture time. (For interpretation of the references to color in this g-
ure, the reader is referred to the web version of this article.)
Fig. 7. Optical photographs of a selection of granite samples used in this study, showing geometrical effects of mechanical fracturing. The connected porous networks (mainly
cracks) of the thin-sections have been colored blue experimentally during processing, and then red using image processing. Optical photographs of the thin-section show (a)
intact and (b) fractured PL samples, and (c) intact and (d) fractured AL samples. (For interpretation of the references to color in this gure, the reader is referred to the web
version of this article.)
[m5G; v1.221; Prn:26/07/2017; 10:56] P.7 (1-10)
7
way for percolation. At this stage, permeability increases and radon
concentration increases transiently (Fig. 5), with the largest radon
concentrations coincident with rupture of the sample.
Looking at the rupture point in more detail, let us now take
into account the transit time of argon through the sample. This
characteristic time can be calculated using = L 2 /k, where
is the compressibility of argon (7.9 108 Pa1 ). We calculate
considering k value determined before rupture and k value deter-
mined after rupture. Then, we subtract from the measured time
of the experiment for radon time-series. Regardless of sample per-
meability and transit time considerations, we observe that radon
concentration increase is systematically detected before sample
rupture (Fig. 6). Furthermore, the transient radon signals produced
before rupture are observed for all the studied granite samples
(Fig. 6a,b and Supplementary Fig. S3). This shows that radon emis-
sion increases signicantly, from a half-day to a few hours be-
fore rupture, depending on sample permeability (Fig. 6 and Sup-
plementary Fig. S3), consistent with other recent studies of gas
emission during deformation of crustal rocks (Nicolas et al., 2014;
Koike et al., 2015; Bauer et al., 2016).
To explain the radon anomalies observed at the rupture point,
let us compare sample characteristics before and after macroscopic
failure. The radon source term (EC Ra ) of the samples after rup-
ture is larger than that before rupture by 14 1% to 58 4%
(42 14% on average). This increase is larger than most reported
values obtained under uniaxial compression, possibly due to our
experimental protocol that places the sample under more natural
conditions (conning pressure and pore pressure). The gradual in-
crease in the number of acoustic emissions with differential stress
increases is corroborated by observation in thin-sections (Fig. 7).
Indeed, thin sections of the fractured PL and AL samples, from the
middle of each sample near the macroscopic fracture, show a sig-
nicant increase in the number of cracks compared to the intact
JID:EPSL AID:14555 /SCO [m5G; v1.221; Prn:26/07/2017; 10:56] P.8 (1-10)
8 F. Girault et al. / Earth and Planetary Science Letters ()

samples. This supports the large permeability increase recorded at

the rupture point.
Taking into account the steady-state increase of the radon
source term after rupture is not sucient to explain the tran-
sient radon signals. In addition, the increase in acoustic emissions
(Fig. 8a) and the increase in permeability (Fig. 8b) seem correlated
with the radon anomalies. A signicant transient radon source is
necessary to explain the observed anomalies at rupture. Incorpo-
rating such a transient effect at rupture and over preceding stages
of differential stress increase (failure stages) in the advective radon
transport model described in subsection 4.1 gives satisfactory rst-
order predictions of the radon signal, as shown by the dashed
orange curves in Fig. 5. For example, in Fig. 5a, EC Ra is transiently
increased by a factor of 2 (event 1), 1.2 (event 2), and 4.5 (event 3).
In Fig. 5b, EC Ra is transiently increased by a factor of 2 (event 2),
2.5 (event 3), 7 (event 4), and 30 (event 5). This transient radon
source, dened here by the transient emanation factor, is well cor-
related with the radon anomalies (Fig. 8c).

4.5. Rupture-induced transient radon signals: from experiments to the

Earths crust

From a physical point of view, the transient radon source ap-

pears likely related to sudden dynamic liberation of radon, ini-
tially accumulated in isolated cracks at radioactive equilibrium,
when transient increases of pore and fracture connectivity oc-
cur. The higher radium-226 activity concentration of the PL sam-
ple compared with that of the AL sample (Table 1) may ex-
plain the larger radon anomalies observed for the PL sample. As
shown in Supplementary Fig. S4, the transient emanation factor
is larger as permeability increase and thus fracturing is more in-
tense. This suggests that it is directly related to the number of
cracks within the sample, and that uids can be channeled from
the top surface to the bottom surface of the sample through di-
lating cracks (Main et al., 2012). Therefore, permeability changes
(Manning and Ingebritsen, 1999; Manga et al., 2012), not readily
detectable after the experiment, are eciently monitored during
the experiment using radon and likely nucleate the radon anoma-
lies recorded in our deformation experiments. These short time
scale variations have larger amplitudes and shorter durations when
sample permeability is larger than 1018 m2 . This permeability
value is 2 to 4 orders of magnitude lower than common up-
per crustal permeability values (Ingebritsen and Manning, 2010;
Manga et al., 2012) and 5 to 6 orders of magnitude lower than Fig. 8. Experimental radon concentration anomalies obtained in this study for AL
permeability values predicted in fault zones (Miller et al., 2004). (in red) and PL (in blue) granites versus (a) the increase of the number of acoustic
However, similar permeability values have been measured in situ emissions, (b) the increase of axial gas permeability, and (c) transient increase of
in deep drilling experiments (Townend and Zoback, 2000). the radon emanation factor. Solid lines are power-law best ts ( y = axb ): in (a)
a = 56.5 and b = 0.615 (R 2 = 0.19), in (b) a = 3.61 and b = 0.686 (R 2 = 0.36), and
Considering a vertical crustal fault, one can calculate the frac- in (c) a = 0.295 and b = 0.661 (R 2 = 0.61). (For interpretation of the references to
tion of advective radon signal, i.e., the fraction of deep-originated color in this gure, the reader is referred to the web version of this article.)
radon that can be detected at the surface, and its transit time to
reach the surface (Girault and Perrier, 2014; Neri et al., 2016). Sup- 5. Conclusion
plementary Fig. S5 shows the fraction of deep-originated radon
as a function of fault length for different fault porosity and fault Based on several month-long deformation experiments per-
permeability values (15% and 1015 1011 m2 , respectively). For formed on granite samples under crustal conditions and coupled
large fault permeability, small fault porosity and/or small fault with high resolution monitoring of the concentration of radon gas,
length, the transit time of radon is reduced and the fraction of we have been able to reproduce the variety of transient radon sig-
deep-originated radon is signicantly larger. Therefore, detecting nals predicted or reported in the literature. We have shown that
deep-originated radon signals at the surface, such as transient uid pressure pulses, micro-crack closure, and failure of crustal
radon signals caused by brittle fracturing at depths of the order rocks are associated with different sorts of radon anomalies but
of 1 km, requires fault permeability around 1012 m2 (1 D), con- that all of them can be explained using a rst-order advective
sistent with previous considerations for carbon dioxide (Miller et model of radon transport. (1) Increased uid pressure is associ-
al., 2004). This high fault permeability value represents fast gas ated with a large peak of radon emission and can be related to
migration, as reported in several hydrothermal systems located in the fault-valve model. (2) The increase of isostatic pressure rst
tectonically and volcanically active regions. enhances radon emission and then diminishes it beyond a crit-
JID:EPSL AID:14555 /SCO
F. Girault et al. / Earth and Planetary Science Letters ()
ical pressure, i.e., a given depth below which radon emission is
hampered by the closure of cracks. (3) The increase of differen-
tial stress under hydrostatic conditions systematically triggers sig-
nicant radon increases before macroscopic failure is reached by
connecting previously isolated cracks. This study thus offers an ex-
perimental and physical basis to clarify reported transient radon
signals.
Coupling high-resolution monitoring of geochemical parameters
with rock mechanics experiments illuminates the complex pro-
cesses that occur during rock deformation. A radioactive uid such
as radon is easier and faster to track than stable gas species be-
cause radiation detection instruments are much more sensitive.
Radon concentration monitoring during deformation experiments
provides information on the state of the porous and fracture net-
works. All the radon signals can be explained by the three terms
of the transport model (the source term, advective transport, and
decay or losses), and thus associated with variations in pore pres-
sure, conning pressure, differential stress, or permeability. This
methodology opens promising opportunities to study other types
of geological media, in particular those having different types of
porosity (granite versus sandstone), vacuoles (basalt), or micro-
and macro-pores (volcanic tuff). Our methodology may also be
applied to study stress-induced release of radiogenic noble gases
(helium, argon; Bauer et al., 2016) and other gases detected near
active fault zones worldwide (carbon dioxide, methane, dihydro-
gen).
The detection of transient radon signals before rupture (Fig. 6)
and when differential stress increases before brittle fracturing
(Fig. 5b) is extremely encouraging. It suggests that, when perco-
lation threshold is reached, connection of initially isolated cracks
may lead to anomalous radon signals. The present study supports
long-term monitoring of concentrations and uxes of radon and
other gases in tectonically and volcanically active regions of fast
gas migration from deep or shallow crustal sources. These sites
have the potential to elucidate stress and permeability variations
in the Earths crust.
Acknowledgements
We warmly thank J. Fortin (ENS) and S. Vinciguerra (University
of Leicester, UK) for fruitful discussions. We are also thankful to
Y. Pinquier (ENS) for technical help, A. Nicolas (ENS) for provid-
ing the block of PL sample, and S. Inel (ENS) for processing the
thin-section photographs. S. Ingebritsen (USGS Menlo Park, USA)
and F. Perrier (IPGP) are thanked for their inspiring comments.
The original version of the manuscript was improved thanks to
the constructive suggestions of two anonymous reviewers. This is
a joint research laboratory effort in the framework of the CEA-ENS
Yves Rocard LRC (France).
Appendix A. Supplementary material
Supplementary material related to this article can be found on-
line at http://dx.doi.org/10.1016/j.epsl.2017.07.013.
References
Bauer, S.J., Gardner, W.P., Lee, H., 2016. Release of radiogenic noble gases as a new
signal of rock deformation. Geophys. Res. Lett. 43, 1068810694.
Belikov, V.T., Kozlova, I.A., Ryvkin, D.G., Yurkov, A.K., 2014. Origin of anomalies
of radon volume activity during failure of rocks. Russ. Geol. Geophys. 55,
12191222.
Brace, W.F., Walsh, J.B., Frangos, W.T., 1968. Permeability of granite under high pres-
sure. J. Geophys. Res. 73, 22252236.
Byerlee, J., 1993. Model for episodic ow of high-pressure water in fault zones be-
fore earthquakes. Geology 21, 303306.
Cigolini, C., Laiolo, M., Coppola, D., 2007. Earthquakevolcano interactions detected
from radon degassing at Stromboli (Italy). Earth Planet. Sci. Lett. 257, 511525.
[m5G; v1.221; Prn:26/07/2017; 10:56] P.9 (1-10)
9
Cox, M.E., Cuff, K.E., Thomas, D.M., 1980. Variations of ground radon concentrations
with activity of Kilauea Volcano, Hawaii. Nature 288, 7476.
Darot, M., Guguen, Y., Baratin, M.-L., 1992. Permeability of thermally cracked gran-
ite. Geophys. Res. Lett. 19, 869872.
Ferry, C., Richon, P., Beneito, A., Cabrera, J., Sabroux, J.-C., 2002. An experimen-
tal method for measuring the radon-222 emanation factor in rocks. Radiat.
Meas. 35, 579583.
Fischer, T., Matyska, C., Heinicke, J., 2017. Earthquake-enhanced permeability ev-
idence from carbon dioxide release following the ML 3.5 earthquake in West
Bohemia. Earth Planet. Sci. Lett. 460, 6067.
Fortin, J., Schubnel, A., Guguen, Y., 2005. Elastic wave velocities and permeability
evolution during compaction of Bleurswiller sandstone. Int. J. Rock Mech. Min.
Sci. 42, 873889.
Geller, R.J., 2011. Shake-up time for Japanese seismology. Nature 472, 407409.
Girault, F., Perrier, F., 2014. The Syabru-Bensi hydrothermal system in central Nepal:
2. Modeling and signicance of the radon signature. J. Geophys. Res., Solid
Earth 119, 40564089.
Girault, F., Bollinger, L., Bhattarai, M., Koirala, B.P., France-Lanord, C., Rajaure, S.,
Gaillardet, J., Fort, M., Sapkota, S.N., Perrier, F., 2014. Large-scale organization
of carbon dioxide discharge in the Nepal Himalayas. Geophys. Res. Lett. 41,
63586366.
Holub, R.F., Brady, B.T., 1981. The effect of stress on radon emanation from rock. J.
Geophys. Res. 86, 17761784.
Igarashi, G., Tohjima, Y., Wakita, H., 1993. Time-variable response characteristics of
groundwater radon to earthquakes. Geophys. Res. Lett. 20, 18071810.
Igarashi, G., Saeki, S., Takahata, N., Sumikawa, K., Tasaka, S., Sasaki, Y., Takahashi,
M., Sano, Y., 1995. Ground-water anomaly before the Kobe earthquake in Japan.
Science 269, 6061.
Ingebritsen, S.E., Manning, C.E., 2010. Permeability of the continental crust: dynamic
variations inferred from seismicity and metamorphism. Geouids 10, 193205.
Irwin, W.P., Barnes, I., 1980. Tectonic relations of carbon dioxide discharges and
earthquakes. J. Geophys. Res. 85, 31153121.
Kennedy, B.M., Kharaka, Y.K., Evans, W.C., Ellwood, A., DePaolo, D.J., Thordsen, J.,
Ambats, G., Mariner, R.H., 1997. Mantle uids in the San Andreas fault system,
California. Science 278, 12781281.
King, C.-Y., Luo, G., 1990. Variations of electric resistance and H2 and Rn emis-
sions of concrete blocks under increasing uniaxial compression. Pure Appl. Geo-
phys. 134, 4556.
Koike, K., Yoshinaga, T., Suetsugu, K., Kashiwaya, K., Asaue, H., 2015. Controls on
radon emission from granite as evidenced by compression testing to failure.
Geophys. J. Int. 203, 428436.
Liu, Y., Chen, T., Xie, F., Du, F., Yang, D., Zhang, L., Xu, L., 2014. Analysis of uid
induced aftershocks following the 2008 Wenchuan Ms 8.0 earthquake. Tectono-
physics 619620, 149158.
Main, I.G., Bell, A.F., Meredith, P.G., Geiger, S., Touati, S., 2012. The dilatancy-diffusion
hypothesis and earthquake predictability. Geol. Soc. (Lond.) Spec. Publ. 367,
215230.
Manga, M., Beresnev, I., Brodsky, E.E., Elkhoury, J.E., Elsworth, D., Ingebritsen, S.E.,
Mays, D.C., Wang, C.-Y., 2012. Changes in permeability caused by transient
stresses: eld observations, experiments, and mechanisms. Rev. Geophys. 50,
RG2004.
Manning, C.E., Ingebritsen, S.E., 1999. Permeability of the continental crust: implica-
tions of geothermal data and metamorphic systems. Rev. Geophys. 37, 127150.
Miller, S.A., Collettini, C., Chiaraluce, L., Cocco, M., Barchi, M., Kaus, B.J.P., 2004. Af-
tershocks driven by a high-pressure CO2 source at depth. Nature 427, 724727.
Mitchell, T.M., Faulkner, D.R., 2008. Experimental measurements of permeability
evolution during triaxial compression of initially intact crystalline rocks and im-
plications for uid ow in fault zones. J. Geophys. Res. 113, B11412.
Mollo, S., Tuccimei, P., Heap, M.J., Vinciguerra, S., Soligo, M., Castelluccio, M., Scar-
lato, P., Dingwell, D.B., 2011. Increase in radon emission due to rock failure: an
experimental study. Geophys. Res. Lett. 38, L14304.
Nazaroff, W.W., 1992. Radon transport from soil to air. Rev. Geophys. 30, 137160.
Neri, M., Ferrera, E., Giammanco, S., Currenti, G., Cirrincione, R., Patan, G., Zanon,
V., 2016. Soil radon measurements as a potential tracer of tectonic and volcanic
activity. Sci. Rep. 6, 24581.
Nicolas, A., Girault, F., Schubnel, A., Pili, ., Passelgue, F., Fortin, J., Deldicque, D.,
2014. Radon emanation from brittle fracturing in granites under upper crustal
conditions. Geophys. Res. Lett. 41, 54365443.
Perrier, F., Richon, P., Byrdina, S., France-Lanord, C., Rajaure, S., Koirala, B.P., Shrestha,
P.L., Gautam, U.P., Tiwari, D.R., Revil, A., Bollinger, L., Contraires, S., Bureau, S.,
Sapkota, S.N., 2009. A direct evidence for high carbon dioxide and radon-222
discharge in Central Nepal. Earth Planet. Sci. Lett. 278, 198207.
Pili, E., Perrier, F., Richon, P., 2004. Dual porosity mechanism for transient ground-
water and gas anomalies induced by external forcing. Earth Planet. Sci. Lett. 227,
473480.
Rice, J.R., 1992. Fault stress states, pore pressure distributions, and the weakness of
the San Andreas fault. In: Evans, B., Wong, T.F. (Eds.), Earthquake Mechanics and
Transport Properties of Rocks. Academic Press, London, pp. 475503.
Richon, P., Sabroux, J.-C., Halbwachs, M., Vandemeulebrouck, J., Poussielgue, N., Tab-
bagh, J., Punongbayan, R., 2003. Radon anomaly in the soil of Taal volcano, the
JID:EPSL AID:14555 /SCO
10 F. Girault et al. / Earth and Planetary Science Letters ()
Philippines: a likely precursor of the M 7.1 Mindoro earthquake 1994. Geophys.
Res. Lett. 30 (9), 1481.
Sakoda, A., Ishimori, Y., Yamaoka, K., 2011. A comprehensive review of radon emana-
tion measurements for mineral, rock, soil, mill tailing and y ash. Appl. Radiat.
Isot. 69, 14221435.
Sato, M., Sutton, A.J., McGee, K.A., Russell-Robinson, S., 1986. Monitoring of hy-
drogen along the San Andreas and Calaveras faults in Central California in
19801984. J. Geophys. Res. 91, 1231512326.
Scarlato, P., Tuccimei, P., Mollo, S., Soligo, M., Castelluccio, M., 2013. Contrasting
radon background levels in volcanic settings: clues from 220 Rn activity con-
centrations measured during long-term deformation experiments. Bull. Vol-
canol. 75, 751.
Scholz, C.H., 1968. Microfracturing and the inelastic deformation of rock in com-
pression. J. Geophys. Res. 73, 14171432.
Schubnel, A., Fortin, J., Burlini, L., Guguen, Y., 2005. Damage and recovery of cal-
cite rocks deformed in the cataclastic regime. Geol. Soc. (Lond.) Spec. Publ. 245,
203221.
Sibson, R.H., 2014. Earthquake rupturing in uid-overpressured crust: how com-
mon? Pure Appl. Geophys. 171, 28672885.
Steinitz, G., Begin, Z.B., Gazit-Yaari, N., 2003. Statistically signicant relation be-
tween radon ux and weak earthquakes in the Dead Sea rift valley. Geology 31,
505508.
Sugisaki, R., Ido, M., Takeda, H., Isobe, Y., Hayashi, Y., Nakamura, N., Satake, H.,
Mizutani, Y., 1983. Origin of hydrogen and carbon dioxide in fault gases and
its relation to fault activity. J. Geol. 91, 239258.
Tanner, A.B., 1964. Radon migration in the ground: a review. In: Adams, J.A.S., Low-
der, W.M. (Eds.), The Natural Radiation Environment. Univ. of Chicago Press,
Chicago, pp. 161190.
[m5G; v1.221; Prn:26/07/2017; 10:56] P.10 (1-10)
Terakawa, T., Hashimoto, C., Matsuura, M., 2013. Changes in seismic activity follow-
ing the 2011 Tohoku-oki earthquake: effects of pore uid pressure. Earth Planet.
Sci. Lett. 365, 1724.
Townend, J., Zoback, M.D., 2000. How faulting keeps the crust strong. Geology 28,
399402.
Trique, M., Richon, P., Perrier, F., Avouac, J.-P., Sabroux, J.-C., 1999. Radon emana-
tion and electric potential variations associated with transient deformation near
reservoir lakes. Nature 399, 137141.
Tuccimei, P., Mollo, S., Vinciguerra, S., Castelluccio, M., Soligo, M., 2010. Radon and
thoron emission from lithophysae-rich tuff under increasing deformation: an
experimental study. Geophys. Res. Lett. 37, L05305.
Virk, H.S., Singh, B., 1994. Radon recording of Uttarkashi earthquake. Geophys. Res.
Lett. 21, 737740.
Wakita, H., Nakamura, Y., Notsu, K., Noguchi, M., Asada, T., 1980. Radon anomaly:
a possible precursor of the 1978 lzu-Oshima-kinkai earthquake. Science 207,
882883.
Wakita, H., Igarashi, G., Notsu, K., 1991. An anomalous radon decrease in groundwa-
ter prior to an M6.0 earthquake: a possible precursor? Geophys. Res. Lett. 18,
629632.
Walia, V., Lin, S.J., Fu, C.C., Yang, T.F., Hong, W.-L., Wen, K.-L., Chen, C.-H., 2010. Soil-
gas monitoring: a tool for fault delineation studies along Hsinhua Fault (Tainan),
Southern Taiwan. Appl. Geochem. 25, 602607.
Wang, C.-Y., Manga, M., 2010. Hydrologic precursors. In: Earthquakes and Water. In:
Lecture Notes in Earth Sciences. Springer-Verlag, Berlin Heidelberg, pp. 141159.
Ziarani, A.S., Aguilera, R., 2012. Knudsens permeability correction for tight porous
media. Transp. Porous Media 91, 239260.