WATER RESOURCES RESEARCH, VOL. 40, W03502, doi:10.

1029/2003WR002717, 2004

Spontaneous switching of permeability changes in a limestone

fracture with net dissolution
Amir Polak,1,2 Derek Elsworth,1 Jishan Liu,3 and Abraham S. Grader1
Received 1 October 2003; revised 12 December 2003; accepted 12 January 2004; published 2 March 2004.
[1] Results are reported for water flow-through experiments conducted on an artificial
fracture in limestone at room temperature and under ambient confining stress of 3.5 MPa.
Tests are concurrently monitored for mineral mass loss or gain and for changes in
differential pressure between the inlet and outlet, throughout the 1500-hour duration of the
experiment. Periodic imaging by X-ray computed tomography augments the fluid and
mineral mass balance data and provides a third independent constraint on dissolution
processes. The sample is sequentially circulated by water of two different compositions
through the 1500-hour duration of the experiment, the first 935 hours by sampled
groundwater (pH
8), followed by 555 hours of distilled water (pH
6). Large changes
in the differential pressure are recorded throughout the experiment, for the constant flow
rate of 2 cm2/m; these are used as a proxy for recorded changes in fracture permeability,
under invariant effective stress conditions. Mass of Ca and Mg were net-removed
throughout the experiment. During the initial circulation of groundwater, the differential
pressure increased almost threefold and is interpreted as a net reduction in permeability as
the contacting asperities across the fracture are removed and the fracture closes. With the
circulation of distilled water, permeability initially reduced threefold and ultimately
increased by 2 orders of magnitude as a wormhole developed in the sample. This
spontaneous switch from net decrease in permeability to net increase occurred with no
change in experimental conditions of flow rate or applied effective stress, and Ca was net
dissolved throughout. This behavior is attributed to the evolving localization of mass
removal, triggered as free-face dissolution outcompetes stress-mediated dissolution at the
asperity contacts. INDEX TERMS: 5114 Physical Properties of Rocks: Permeability and porosity; 1829
Hydrology: Groundwater hydrology; 5139 Physical Properties of Rocks: Transport properties; 8010 Structural
Geology: Fractures and faults; KEYWORDS: fracture permeability, dissolution, precipitation, geochemistry,
channeling
Citation: Polak, A., D. Elsworth, J. Liu, and A. S. Grader (2004), Spontaneous switching of permeability changes in a limestone
fracture with net dissolution, Water Resour. Res., 40, W03502, doi:10.1029/2003WR002717.

1. Introduction result of mineral precipitation [e.g., Tenthorey et al., 1998;

[2] The natural development of karst, the purposeful Aharonov et al., 1998], and in fluid flow through dissolution
acidization of carbonate petroleum reservoirs, and the channels with connection to the Earth mantle [Kelemen et
proposed high-volume injection and recovery of surface al., 1995]. These processes may transfer large quantities of
waters via wells in the Florida Everglades are three prom- mass between the fluid and the solid mineral phases
inent examples where nonequilibrium chemical and me- [Domenico and Schwartz, 1998].
chanical processes will drive changes in system porosity. [3] Changes in effective stresses, driven by injection and
The processes of dissolution and precipitation in porous dewatering, and changes in reactive chemistry, driven by
media exert considerable influence in groundwater hydrol- nonequilibrium chemical potentials, drive changes in po-
ogy [White, 2002], groundwater chemistry [Lasaga, 1984], rosity. For fractures as analogs of the flow system, the
the enhancement of oil and gas production from petroleum evolution in porosity is modulated by initial distributions of
reservoirs [e.g., Hoefner and Fogler, 1988; Rege and fracture aperture, and evolves in an apparently unpredict-
Fogler, 1989], changes in hydraulic conductivity as a able manner as a function of the paths of stress, and of
chemical potential, followed by the system. Importantly, for
1 carbonate systems, close to equilibrium, the evolution of
Energy Institute and Department of Energy and Geo-Environmental
Engineering, Pennsylvania State University, University Park, Pennsylvania, permeability is not single-sensed (see Figure 5b), and where
USA. stress paths are involved, fractures may either gape or close
2
Now at Department of Civil and Environmental Engineering, Technion- with net dissolution; permeability may increase as fast-flow
Israel Institute of Technology, Haifa, Israel. channels are etched or may decrease as dissolution from
3
School of Oil and Gas Engineering, University of Western Australia,
Nedlands, Western Australia, Australia. bridging asperities brings fracture walls closer together.
[4] The competition between flow channel formation due
Copyright 2004 by the American Geophysical Union. to dissolution and channel destruction by precipitation
0043-1397/04/2003WR002717 [Rege and Fogler, 1989; Singurindy and Berkowitz, 2003]
W03502 1 of 10
W03502 POLAK ET AL.: SPONTANEOUS SWITCHING OF PERMEABILITY CHANGES W03502

industrial scanner (Universal systems, Inc.). The scanner

produces two-dimensional slices through the sample with a
thickness of 50 mm and an in-plane pixel resolution of
47 mm. The X-ray energy level used in the experiment was
195 kV at 800 mA. The core holder was removed from the
CT scanner after each scanning procedure. In order to
ensure proper registration of the CT images, two small
(1  1  1 mm) stainless steel targets were inserted into
the outer flank of the fracture to act as fiducial marks. In
Figure 1. The limestone sample with its nonconnecting each scanning sequence the CT system rotated the sample
fractures and scanning areas. 360 while generating X rays that are detected by a digital
camera that acts as a target. In each rotation the CT system
takes 41 horizontal slices from the sample. Detailed descrip-
tions on the use and limitations of X-ray CT in core analysis
illustrates the important roles of chemistry and flow rates in
are given by Stock [1999], Ketcham and Carlson [2001],
defining the potential for wormhole formation in carbonate
and Wildenschild et al. [2002].
rocks. The limited studies on fractures [Durham et al.,
2001; Lin et al., 1997; Moore et al., 1994; Morrow et al., 2.2. Experimental Procedure
2001] suggest an increased sensitivity of their transport 2.2.1. Core Packing and Vacuum Saturation
properties to stresses and reactive chemical processes, over [9] After coring, the core sample was fractured using a
porous medium flows. Brazilian-like test [Vukuturi et al., 1974] to produce a
[5] This study examines the processes of dissolution and longitudinal plane fracture. The sample exterior was sealed
precipitation that progress within an induced fracture in with epoxy resin to prevent flow short-circuiting, sealed
limestone, specifically examining the evolution in perme- within a Viton1 sleeve, and inserted into the core holder
ability that develops under net dissolution as driven by with hydraulic access via end-plugs. The confining fluid was
changes in composition of the inlet solute. Detailed mea- added and pressured to 3.5 MPa, and the sample
surements of the driving fluid pressure, effective stresses, was evacuated to a pressure of 9 Pa for 24 hours. The
flow rates, influent and effluent fluid compositions, and experiment was initiated by vacuum saturation with water.
the imaged interior structure of the fracture provide 2.2.2. Scanning Procedure
constraint to important processes controlling dissolution [10] Following core packing, a total of four scans were
and precipitation. taken at different stages of the experiment. All scanned
sections are perpendicular to the core axis. A single rotation
2. Method collects 41 slices, each of a thickness of 50 mm. Succes-
sive rotations are completed by translating the sample
[6] The flow experiment was conducted on a cylindrical
relative to the static X-ray source and detector, enabling
core (50 mm diameter  108 mm length) of Bellefonte
the scanning of larger volumes of the core. The first scan
limestone containing a single diametral fracture intersecting
was carried out after saturation to determine whether the
an axial fracture as shown in Figure 1. The core was
fracture was completely open throughout its length and to
confined within a Viton1 sleeve with end-to-end flow tests
serve as a baseline for the rest of the experiment. Three
conducted within an industrial X-ray computed tomography
scans followed, all at an effective stress of 3.5 MPa, and
(CT) scanner. For the applied steady flux, measured axial
represented progress of the experiment at 938, 1173, and
pressure differential yields a continuous record of the
1492 hours from the start of the experiment. Each scan
permeability constrained by intermittent X-ray scans.
2.1. Experimental Apparatus
2.1.1. The Coreholder and Flow System
[7] The fluid flow system and the coreholder are shown
in Figure 2. The core holder is an 80-cm-long triaxial
pressure cell of 9.5-cm outside diameter made of carbon
fiber-reinforced aluminum tube (Temco Inc.) with stainless
steel end plugs that are pierced by fluid supply tubes.
Controlled flow rates and regulation of the confining
pressure were prescribed by twin opposing-stroke Quizix
(5400) pumps with an applied flow rate of 2 cm2/min during
the entire length of the experiment. The sample is sequen-
tially circulated by water of two different compositions
through the 1500-hour duration of the experiment, the first
935 hours by sampled groundwater, followed by 555 hours
of distilled water, with an estimated residence time of the
order of tens of seconds. The inlet water was pumped from a
6-L reservoir during both parts of the experiment.
2.1.2. The CT System
[8] X-ray imaging of two-dimensional slices perpendic- Figure 2. A schematic description of the experimental
ular to the core axis was completed by a custom-made setup.
2 of 10
W03502 POLAK ET AL.: SPONTANEOUS SWITCHING OF PERMEABILITY CHANGES W03502

sampled volumes at three different axial locations in the aperture, as the evolving flow conduit was tubular rather
core (610, 646, and 665, representing distances of 30, 66, than planar.
and 85 mm, respectively, from the core inlet). The first [16] Rates of aperture reduction, db/dt, recovered from
location (610) had nine scanning rotations each of 41 slices, recorded effluent concentrations, may be compared with
representing an axial length of 18.45 mm (50 mm for each those independently recovered from the hydraulic test. For
slice). The second location (646) had three rotations, each of dissolution from beneath the contact area between two
41 slices, representing an axial length of 6.15 mm, and the fracture surfaces, of area Ac[L2], the rate of fracture closure,
third location had a single rotation with an axial length of db/dt [L T1], may be defined as [Polak et al., 2003]
2.05 mm. " #
2.3. Experimental Analysis db Q CCACO3 CCaMgCO3 2
2
[11] Concurrent X-ray imaging, indirect monitoring of dt Ac rCACO3 rCaMgCO3 2
the evolution of hydraulic aperture (via differential pres-
sure measurement), and monitoring of solute mass, were where Q is the volumetric flow rate [L3 T1], C and r are
completed. the concentration and density [M L3] of CaCO3 and
2.3.1. Chemical Analysis CaMg(CO3)2 (the two components that were detected in the
[12] During the 1500 hours of the experiment, 73 water sample using the X-ray diffraction system) in the effluent,
samples were collected for chemical analysis of Al, Ba, Ca, respectively, calculated from the concentration measure-
Fe, K, Mg, Mn, Na, P, Si, and Sr using an inductively ments of Ca and Mg. These calculations enable the change
coupled plasma-atomic emission spectrophotometer with an in aperture to be determined from effluent concentrations
estimated relative error of 5%. Fourteen samples were and measured flow rates.
collected from the inlet side and 59 were collected from the 2.3.3. CT Image Analysis
outlet side. The sample solutions were also analyzed for [17] The redistribution of mineral mass within the frac-
their pH values using a ROSS combination pH electrode ture was examined throughout the test by subtracting
(Thermo Orion). sequential thresholded CT images. Sequential CT image
[13] X-ray diffraction of the sample was performed using pairs were selected with prescribed threshold CT-number
a Rigaku Geigerflex system with Cu radiation. The results (1670 and 1600, for the third and fourth time sequences,
show a strong calcite peak at d = 3.03 A , with a much respectively) representing the transition from solid to pore
smaller dolomite peak at d = 2.89 A . No magnesite was space. Pixel counts define the volumetric transition from
detected in the sample. rock to pore space within the sample and enable quantifi-
[14] In order to check the ratio between calcium and cation of mineral mass redistribution. This procedure was
magnesium in the rock sample, three powder-crushed preformed on 12 pairs from different areas along the core.
samples were weighed and immersed into 5% HNO3
(1:1000 solid/liquid ratio) for 24 hours, to allow full
dissolution of the solid phase. The samples were then 3. Results and Discussion
analyzed by ICP for the ion concentrations. The results [18] Two distinct behaviors exist within the test. The first
from this batch test show that only 8% of the sample is is net reduction in aperture in the initial test period, when
magnesium, by weight, and the rest is calcium. Fe, Na, Sr, asperities are net removed, and the confining stress closed
Si, and K were also detected in the samples but with the facture. This reduction in aperture accelerates as the
concentrations of less then 0.5 ppm. The presence of these circulating fluid is first switched to distilled water. The
ions is attributed to the small amount of clays that was second is a net increase in permeability resulting from
found in the sample residue. the widening of a single dominant solution channel, termed
2.3.2. Fracture Aperture Analysis a wormhole, apparent in the differential pressure record,
[15] Differential pressure between the sample inlet and and corroborated by imaging. Contrasting models are
outlet were recorded throughout the duration of the flow- developed for these two situations.
through tests with a resolution of 40 Pa. Labview recorded [19] The ensemble balances of fluid flow and mineral
two Quartzdyn QS transducers at six cycles/min. The low mass efflux, and CT imaging provide independent measure-
porosity of the limestone matrix enabled flow rates to be ments of the process effecting mineral redistribution within
converted directly to equivalent hydraulic aperture via the the sample. Fluid flows constrain hydraulic apertures un-
parallel plate approximation [e.g., Witherspoon et al., 1980; equivocally, and cumulative measurements of mineral mass
Silliman, 1989; Piggott and Elsworth, 1993] removal may be converted to changes in mechanical aper-
ture if contact areas are known or estimated and an
DP 3 appropriate model for permeability enhancement assumed.
Q b W 1
12mL Similarly, imaging defines changes in fracture porosity if
resolution is sufficiently high. In the following we link
where m is temperature-dependent dynamic viscosity of the observations of fluid and mineral flux, and imaging, with
fluid [M L1 T1], L is length of the sample [L], W is a consistent view of dominant processes and sites of
diameter of the sample [L], Q is flow rate [L3 T1], DP is dissolution, summarized in Table 1.
the differential pressure [M T2 L1], and b is the hydraulic
aperture of the fracture [L]. The aperture was calculated 3.1. Mineral Mass Balance
only during the first 1080 hours of the experiment; for the [20] The influent and effluent concentrations of Ca and
remainder of the experiment, the changes in the differential Mg are shown in Figure 3. Only three ions were detected in
pressure are converted to an approximate equivalent fracture the inlet and outlet samples: Ca, Mg, and Na. The absence
3 of 10
W03502 POLAK ET AL.: SPONTANEOUS SWITCHING OF PERMEABILITY CHANGES W03502

Table 1. Estimated Volumes of Mineral Removed Throughout the Test as Sourced From Asperities or From Free-Face Dissolution
Estimate of Volume Removed (mm3) and Source of
Evaluation

Concentration Thin X-Ray Hydraulic

Permeant Time, hours (Net Efflux) Section CT [Aperture Change, mm]

Groundwater 0 935 360 (180) 35a [6]

Distilled water 935 1100 23 (3) 18a [6]
Distilled waterb 1100 1500 108 (6) 220c 170c 170 (50 mm)c [+32]
53c (10 mm)
20c (2 mm)
a
Sourced from asperities.
b
Magnitudes of mass removed for hydraulic data are consistent with observed pressure drop and flow rates, and for flow slot widths of 50, 10, and 2 mm.
Negative aperture changes denote closing.
c
Sourced from free-face dissolution.

of Na in concentrations above 0.5 ppm in the batch (Figure 4c) in the early test and a larger and more coherent
experiment suggested its presence was introduced only change in the later test period. However, the identification
through the influent fluid, and at relatively high concen- of two periods when significant mass redistribution may
trations (20 ppm). occur is clear (70 190 and 640 890 hours), and these will
[21] The chemical composition of the influent water was be used to evaluate the mineral volume removed from the
switched after the initial 935 hours of circulation. For the system.
initial circulation of groundwater, in near-chemical equilib- [23] Mass removal rate is evaluated from the product of
rium with the limestone (0 935 hours), the inlet and outlet volumetric flow rate and species concentration, to yield
water chemistry is similar. Ca concentrations are of the cumulative mass removed. These calculations produce 265
order 45 ppm, and Mg is of the order 20 ppm for the first and 137 mg for Ca and Mg, respectively, during the first
stage, with minor variations resulting from a changing water
source. For the second stage of the experiment (935
1490 hours), distilled water is circulated, with influent Ca
and Mg concentrations under the detection limit. Effluent
concentrations of Mg remain below 0.1 ppm, and Ca
concentrations asymptote to 4 ppm. This sudden elimination
of Mg raises the suspicion that the measured inlet and outlet
concentrations of Mg were primarily as a result of the inlet
ground water and that the interaction with the rock matrix
played a minor role. Measurement error bars show that only
in two parts of the first experiment (180 and 800 hours) is
there a substantial difference between the inlet and outlet
concentrations.
[22] The variations between the differential concentra-
tions (inlet minus outlet) are shown in Figure 4, where the
positive numbers indicate net precipitation within the sam-
ple. The first part of the experiment is characterized by
concentration differences that reach 22 and 7 ppm for Ca
and Mg, respectively, with precipitation of mass during the
first 100 hours followed by short episodes thereafter. The
main period of active dissolution occurs after 500 hours
when large amounts of Ca are removed (Figure 4a), fol-
lowed by the delayed removal of smaller amounts of Mg
beginning at about 700 hours. Error bounds on these
measurements are relatively large, and hence the timing
and delay in removal of Mg may not be significant.
Absolute magnitudes of concentrations are of the order of
50 ppm, and relative errors are of the order of 5%, as
measured on concentration magnitudes 50 ppm. These
correspond to error magnitudes of 5 ppm on the differen-
tial concentrations between inlet and outlet (Figure 3) in the
early test and 0.25 ppm in the late test. These error
observations suggest that the early-test data represent little Figure 3. Chemical concentrations for the inlet and outlet
net mass removal, while dissolution in the later test is real. samples for (a) Ca and (b) Mg. The vertical line at 935 hours
This is substantiated by the observation of only minor signifies the change in the injected water chemistry, and the
increases in pH as the fluid transits from inlet to outlet error bars were created for a relative error of 5%.
4 of 10
W03502 POLAK ET AL.: SPONTANEOUS SWITCHING OF PERMEABILITY CHANGES W03502

representing a volume of 160 mm3 and bringing the total

mass removed to 520 mm3.
[25] The solubility of calcium carbonate is affected by the
pH of the solution [Krauskopf and Bird, 1995]. The overall
process is represented by the reaction

CACO3 H2 CO3 , Ca2 2HCO

3 3

The rate of dissolution of calcite is not pH-dependent above

pH 5.5 [Plummer et al., 1978]. However, the precipitation
rate of calcite increases with increasing pH above 5.5;
therefore the net rate of dissolution in expected to be less in
the groundwater than in the distilled water. The pH values
are shown in Figure 4c. The differences in the pH values
between the two stages of the experiment are explained by
the change in the inlet water. During the first stage of the
experiment, the inlet groundwater had an average pH of
8.15 and the outlet water had pH values from 8.15 to 8.53.
During the second stage of the experiment, the inlet
(distilled) water had an average pH of 6 while the pH of
the outlet water was in the range 6 7. Influent and effluent
pH magnitudes differ little throughout the experiment,
suggesting that the extent of reaction in the core was low
enough that pH was little changed.
3.2. Fluid Flux
[26] Under constant injected flow rate, the differential
pressure recorded between the inlet and the outlet acts as a
direct indicator for change in hydraulic aperture of the
fracture. Differential pressure magnitudes are in the range
500 10 kPa for the duration of the experiment, with
transducer resolution of the order of 80 Pa; these are well
resolved. The change in differential pressure with progress
of the test, and corresponding change in equivalent aperture
magnitude, are illustrated in Figure 5a. In the first stage of
the experiment (0 935 hours) the pressure differential
across the sample rises from 67 to 172 kPa, representing
hydraulic apertures reducing from 24 to 18 mm. Aperture
reduces at a rate of 0.0125 mm per hour for the first 500 hours
Figure 4. Concentration changes with time between the before asymptoting to a final aperture of 20 mm. A power
inlet and the outlet for (a) Ca and (b) Mg, at inlet and outlet. failure that stopped the injection pumps for a few minutes at
Positive values correspond to mass left in the fracture, and 600 hours initiated a sharp drop in the aperture at a rate of
negative values correspond to mass taken out of the system. 0.015 mm per hour for 120 hours, but again asymptoting to
Note the large error bars that are the result of the combined an equilibrium aperture of 18 mm over the final 200 hours of
error of each sample. (c) Change in measured pH the first test stage. This differential pressure record corre-
magnitudes at inlet and outlet, with time. sponds to an initial decrease in permeability of 1 order of
magnitude, followed by a spontaneous increase of 2 orders
of magnitude (Figure 5b). Permeabilities were still increas-
part of the experiment. In order to calculate the mineral ing at the time the test was terminated.
volume that was removed from the system, the mass of each [27] The second stage of the test, initiating with the
ion was multiplied by the ratio of its origin mineral to give a circulation of distilled water, comprises two distinct patterns
corrected mass of 760 and 278 mg for CaCO3 and in flow behavior. Initially, differential pressures jump from
CaMg(CO3), respectively (while Ca content was distributed 84 to 520 kPa over a period of 139 hours, corresponding to
92% as CaCO3 and 8 % as CaMg(CO3)). The total mineral a reduction in fracture aperture from 18 to 12 mm (Figure 5).
volume that was removed from the system (using the Subsequently, with no change in the composition of the
density values of CaCO3 and CaMg(CO3) as 2900 and influent water, pressure drops sharply from 520 to 11 kPa in
2840 kg/m3, respectively) was 262 and 98 mm3 for the 400 hours, corresponding to a 100-fold increase in perme-
first part of the experiment, bringing the total volume to ability and an approximately fourfold increase in equivalent
360 mm3. hydraulic aperture to 43 mm. This occurs without a signif-
[24] During the second part of the experiment only Ca icant change in the mass removal rate (Figure 6). Aperture
was detected in the effluent (Figures 3 and 4). The total increases at a rate of 0.075 mm per hour over these final
measured mass of Ca was 189 mg, or 470 mg of CaCO3, 200 hours. Independent of the interpretation of the recorded
5 of 10
W03502 POLAK ET AL.: SPONTANEOUS SWITCHING OF PERMEABILITY CHANGES W03502

of fracture voids. The absolute difference in the CT-numbers

recorded for the matrix, in the two successive scans, is a
calibration difference and does not indicate a change in any
chemical or physical parameters of the sample.
[29] Two vertical fractures and one horizontal fracture are
visible in the profile taken upstream of the prior section
(Figure 8). Location 610 is 30 mm from the core inlet
(Figure 1). The two vertical fractures are oblique to the
section, and their apparent apertures are consequently en-
larged over their true magnitude. These fractures are not
hydraulically connected and extend only a few millimeters
along the core axis, as apparent in adjacent scanned images.
Consequently, these nonflowing fractures exhibit little
change in apparent aperture in the time-lapse sequence of
images shown in Figure 8. Conversely, time-lapse changes
in aperture are apparent within the flowing horizontal
fracture where a dissolution tube opens close to the core
center, upstream of the feature observed at location 665. The
CT-number profile shows a similar signature of the fracture
to that observed downstream, where a trough of 500
below a plateau of 1800 enables the delineation of the
evolving dissolution feature. This signature may be used
both to delineate features that extend more than one pixel in
size and to calibrate features at the subpixel level.

4. Process-Based Models for Permeability

Change
Figure 5. (a) Change in differential pressure (solid line) [30] Two different behaviors are apparent during the test:
and corresponding hydraulic fracture aperture (dotted line) Apertures first reduce, under presumed net dissolution, and
with time for continuous flow at 2 cm2/min through the then gape as net dissolution is more positively apparent. Net
fracture. Scanning sequences and the composition of dissolution is confirmed throughout the test by two inde-
injected water are noted. (b) Related change in permeability pendent indicators: that inlet pH is always lower than outlet
over the duration of the test. pH and that net efflux of mineral mass is recorded through-
out. Mechanistic models are proposed to replicate these
behaviors. These define permeability reduction as mineral
mass is net removed from contacting asperities, and perme-
differential pressures, the CT imaging indicates the devel- ability enhancement as free-face dissolution localizes along
opment of a through-going wormhole, possibly invalidating a solution channel. If mineral matter is uniformly removed
this simple conceptualization as a parallel sided fracture. from either the face of the conducting fracture, or the
bridging asperities, then measured effluent mass removal
3.3. CT Imaging
[ 28 ] The results of the first (0 hours) and fourth
(1492 hours) scanning sequences at locations 610 and 665
(see Figure 1) are shown in Figures 7 and 8. The evolution
of a wormhole is apparent in the time-lapse imaging
sequence at location 665 (Figure 7), 85 mm downstream
from the core inlet. Two horizontal fractures are visible,
with the lower fracture noncontinuous out of the section
plane; flow occurs in only the upper fracture. Changes in
aperture are shown to localize on a narrow segment of the
fracture that opens in response to net dissolution. An
enlarged segment shows that even in the first scanning
sequence the fracture appears closed, with an aperture lower
than the spatial resolution of the CT imaging. At early time,
the CT-number profile across the fracture shows a decrease
in the CT numbers at the fracture (2000) compared with
the CT number of the matrix (2200). As the fracture
widens, the trough in the CT-number profile deepens from
1800 to 1300, as individual pixels fully register the
signature of the water-filled void, rather than a mixed Figure 6. Change in mg of Ca per hour (dM/dt) (dotted
sample of void and matrix. This CT-number differential of line) and the corresponding differential pressure (solid line),
500 may be used to calibrate for the presence and extent with time during the second part of the experiment.
6 of 10
W03502 POLAK ET AL.: SPONTANEOUS SWITCHING OF PERMEABILITY CHANGES W03502

monotonically increased. The subsequent rapid growth of

permeability is attributed to the development of a single
primary solution channel, with this feature apparent in the
imaging data. This self-organized switch in behavior, from
net permeability reduction to net increase, with no change in
test conditions, is constrained in the following with the
independent measurements of flow rate, mineral flux, and
the scanning data.
4.2.1. Net Reduction in Permeability
[33] With the initial switch in permeant, an accelerating
rate of permeability reduction correlates with a slowly
increasing rate of net Ca removal (Figure 6). This increas-
ing rate of mass removal is indexed by the rise of Ca
concentrations from near background levels (0.001 ppm)
to a steady peak magnitude of 5 ppm, as apparent in
Figure 4. The measured rate of mass removal correlates
with the measured change in differential fluid pressure, as
illustrated in Figure 6. Since flow rate remains constant
throughout the test, a consistent model links accelerating
dissolution from asperities as a feasible source of reduction
in fracture permeability, represented by the proxy increase
in driving pressure. The measured 23 mm3 of mineral mass
removed during the period 935 1100 hours correlates with
a net reduction in fracture aperture of 8 mm if the contact
area of the asperities is of the order 50% (equation (2)).
This assumed fracture contact area is at the upper limit
of those measured where hydrochemical effects are not
Figure 7. Scanned images for the core at longitudinal apparent [e.g., Pyrak-Nolte et al., 1987; Zimmerman
location 665, 85 mm downstream of the inlet, at (a) test and Bodvarsson, 1996; Polak et al., 2003], but the evaluated
initiation (0 hours), and (b) after 1492 hours.

rates should scale directly with measured changes in frac-

ture aperture.
4.1. Circulation of Groundwater: Net Dissolution and
Permeability Reduction
[31] The 6.4-mm reduction in fracture aperture, measured
during the first 935 hours of the test (Figure 5), represents
a total volume of 35 mm3 if distributed over the entire
5500-mm2 (51  108 mm) surface area of the fracture. This
represents a predicted upper bound on the volume removed
from asperities, as contact area must be less than 100%.
However, even this upper bound is an order of magnitude
lower than the net volume evaluated from the chemical
efflux data, of 360 mm3. This suggests either that the
material removed from the sample was drawn from a
broader source than the asperities of a single fracture or
that the contacting asperities and free-face surface of the
fracture both participated in the dissolution. The resolution
of the imaging is unable to definitively discriminate
between these two scenarios, although the source volume
for the excess dissolved mineral mass would be the non-
connecting fractures visible in the scans of Figure 8, for
example.
4.2. Circulation of Distilled Water: Net Dissolution and
Spontaneous Permeability Switching
[32] The changes in permeability that occurred within the
second portion of the test may be correlated with the
growing and coherent net removal of mineral mass from
the sample, as apparent following the initial circulation of Figure 8. Scanned images for the core at longitudinal
distilled water at hour 932. With the introduction of the location 610, 36 mm downstream of the inlet, at (a) test
distilled water, permeability first dropped abruptly and then initiation (0 hours), and (b) after 1492 hours.
7 of 10
W03502 POLAK ET AL.: SPONTANEOUS SWITCHING OF PERMEABILITY CHANGES W03502

shows two thin sections recovered approximately 29 and

46 mm downstream of the core inlet that contain a through-
going fracture of approximately 150 mm in aperture, with an
embedded central flow conduit. Notably, the observed mean
fracture aperture of 150 mm is larger than that predicted
from the hydraulic measurements (Figure 5) and is influ-
enced by the sampling process; the unfused halves of the
core were removed from the core holder at the conclusion of
the test and have sustained displacements. Constraining the
separation of the thin-section fracture walls to the hydrau-
lically measured minimum aperture (12 mm, as apparent in
Figure 5), then the two measured wormhole sections are of
the order of 350 mm in aperture and 1800 mm in width.
These are representative of the remaining five thin sections
and account for a volume of 220 mm3 when extended for a
through-going conduit. This closely matches the 160 mm3
of mass removal recorded from the mineral efflux data. The
concave inward surfaces of the widened conduit are not
apparent within the main portion of the fracture, indicating
that the flow was concentrated on the widened portion
of the conduit and explaining the declining Ca concentra-
tions recorded in the late test following flow localization
(Figure 3).
4.2.2.2. CT-Image Subtraction
[36] Conduit morphology and volume may be indepen-
dently estimated by image subtraction for CT images taken
both before (scan 3) and after the creation of the wormhole
(scan 4). A total of 12 images were selected, representing
30% of the entire sample. The two sets of images are shown
in Figure 10, at positions 36 and 84.5 mm from the core inlet.
The binary pixel map is constructed for a threshold CT
number of between 1600 and 1625 to represent the transition
between rock and void. The threshold changes, depending on
the calibration for each scan location. The upper pair of
images (36 mm from the inlet) represents an area of 25 
10 mm that straddles the fracture and the lower set (84.5 mm
from the inlet) an area of 17  6 mm. The scattered pores of
sequence 3 precede the evolution of the wormhole and
represent total cross-sectional flow areas of 0.87 mm2
Figure 9. A thin-section image of a wormhole taken by (391 pixels) and 0.26 mm2 (116 pixels) at the upstream
SEM (magnified by 25) at two different locations and downstream locations, respectively. After the formation
approximately (a) 29 mm and (b) 46 mm from the core inlet. of the wormhole, some preexisting pores have disappeared,
consistent with the observed reduction in permeability
8-mm reduction correlates closely with the 6-mm reduction in that directly followed the introduction of distilled water
fracture aperture recovered from the hydraulic data. as the permeant. Subsequently, a wormhole has formed
4.2.2. Net Increase in Permeability (sequence 4) from apparently unremarkable starter pores.
[34] Flow localization initiated after distilled water had The summed fracture areas evolve to 3.73 mm2 (1733 pixels)
been circulated within the fracture for 100 hours, as appar- and 0.79 mm2 (366 pixels) for upstream and downstream
ent in the differential pressure record, and continued for the locations. This represents a roughly fourfold increase in
following 300 hours. To constrain processes contributing to cavity sectional area and indicates the variable geometry of
this behavior, we attempt to constrain the geometry of the the conduit along its length. Repeating this procedure for
evolving dissolution feature. The evolution of the wormhole the remaining 10 scan locations, and weighting the volu-
may be constrained by four independent measurements: metric coverage from 30% to 100%, results in a total
direct measurement of wormhole geometry from thin sec- estimated wormhole volume of 170 mm3.
tions, nondestructive imaging by X-ray CT, and constraint 4.2.2.3. Hydraulic Measurements
placed by proxy measurements of pressure drop and net [37] The measured differential pressure may be used to
mineral loss from the sample, with progress of the test. estimate the average aperture of a high aspect-ratio slit,
Constraints placed by these four suites of independent corresponding to the thin-section geometry of Figure 9. The
measurements are compared. observed change in pressure drop from 520 to 11 kPa
4.2.2.1. Thin-Section Analysis corresponds to an increase in an equivalent aperture of
[35] A total of seven thin sections were taken perpendic- 31 mm (increasing from an aperture of 12 mm at hour
ular to the sample axis, and at distributed locations. Figure 9 1088 of the experiment). When extended across the full

8 of 10
W03502 POLAK ET AL.: SPONTANEOUS SWITCHING OF PERMEABILITY CHANGES W03502

[39] Overall, these four complementary but independent

approaches yield consistent estimates of the volume of the
evolving solution feature, save the hydraulic estimates. The
variable fracture aperture and the distribution of the contact
area determine the shape and volume of the mass that will
be removed in the dissolution process. The measurement of
the differential pressure at the inlet and outlet sides of the
fracture gives only an average magnitude of aperture, which
does not reveal the true distribution along the entire flow
path. By using the CT system we can, within the limitations
of resolution, follow the heterogeneity of the fracture and its
evolution in time.

5. Conclusions
[40] The results from this experimental study examine
the controls exerted by dissolution and precipitation on
flow inside an artificial fracture in limestone. Processes
accompanying the continuous flow of water at ambient
temperatures for 1500 hours are constrained by the con-
tinuous measurement of differential pressure and chemical
concentration of the inflow and outflow, together with
periodic imaging by X-ray CT. The flow of groundwater
in the first part of the experiment caused the impedance
to increase monotonically as contacting asperities are
removed and the fracture walls close. Measured net
mineral efflux magnitudes are an order of magnitude larger
than those sensibly required to generate the measured
change in impedance, if the source of the mass is the
contacting asperities, alone. Correspondingly, the fractured
interior of the sample must contribute to the measured
mass removed, despite the significant errors attached to the
measured net efflux.
[41] The switch in permeant from groundwater to distilled
Figure 10. Two sets of CT images taken from two water caused accelerating changes in flow impedance; an
different locations, (a) 36 mm and (b) 84.5 mm from the initial rapid increase in impedance (closure in aperture),
core entrance. In each set of images the upper image is from subsequently halted, and spontaneously transformed into
the third (before the creation of the wormhole) scanning an exponentially decreasing impedance, corresponding to
sequence and the lower image from the fourth (after the the observed growth of a wormhole. The initial measured
creation of the wormhole) scanning sequence. In each increase in impedance is consistent with closure of the
scanning sequence a threshold value was set to signify the fracture resulting from the fast removal of asperities from
transition between the matrix and the fracture. contact areas. This rapid increase in impedance correlates
with a slow rise in effluent concentrations of Ca to an
equilibrium magnitude of 5 ppm. The cumulative mineral
sample, this represents an equivalent removed volume of volume removed during this stage (23 mm3) corresponds
170 mm3. Alternatively, for the approximately 2000 mm closely with that predicted from the hydraulic data (23 mm3),
(2 mm) wide slot observed in thin section (Figure 9), the assuming a 50% contact area across the fracture walls. Stated
same pressure drop corresponds to an increase in aperture differently, the hydraulically sampled reduction in fracture
of 91 mm, representing a total volume removed of only aperture of 6 mm corresponds closely with the 8-mm reduc-
20 mm3. A slot of 10 mm corresponds to an aperture change tion evaluated from the mineral-mass efflux data, assuming a
of 53 mm and a total volume removed of 53 mm3. 50% fracture contact area.
4.2.2.4. Mineral Mass Removal [42] With continued circulation of distilled water at
[38] Finally, measurements of the mineral mass removed constant flow rate, the sample reaches a maximum imped-
following the switching of permeants may also constrain ance as the concentration of the effluent approaches a steady
total changes in porosity. The total mass removed during the maximum of 5 ppm Ca. Following this, the impedance
permeability-enhancement stage of the test is 300 mg of sharply drops with the observed development of a dominant
CaCO3 (Ca was the only ion found during that time), flow channel. This occurs with no changes in the other
representing a total volume of 108 mm3. This volume is experimental conditions, or in the resulting concentration of
smaller than that estimated from the other approaches. the effluent, recording net removal of mass in the sample.
Notably, the CT and the thin-section results are a direct Estimates of cumulative mass removed, evaluated from
measurement of the final wormhole size, while estimates the aqueous efflux data, correspond closely to estimates
from hydraulic measurements, and from cumulative mass from destructive thin-section and nondestructive X-ray CT
removed, are indirect. measurements. Estimates of conduit geometry from the
9 of 10
W03502 POLAK ET AL.: SPONTANEOUS SWITCHING OF PERMEABILITY CHANGES W03502

supporting hydraulic data are more ambiguous. Through- Lin, W., J. Roberts, W. Glassley, and D. Ruddle (1997), Fracture and matrix
going conduits, the dimension of the observed wormholes, permeability at elevated temperatures, paper presented at the Workshop
on Significant Issues and Available Data: Near-Field/Altered-Zone
are considerably more conductive than the observed pres- Coupled Effects Expert Elicitation Project, Dep. of Energy, San Francisco,
sure drops suggest. Correspondingly, conduit volumes eval- Calif., November .
uated from the hydraulic data alone are much smaller than Moore, D. E., D. A. Lockner, and J. A. Byerlee (1994), Reduction of
the consistent predictions from mineral efflux data and permeability in granite at elevated temperatures, Science, 265(5178),
1558 1560.
measured dimensions from thin sections and X-ray CT. Morrow, C. A., D. E. Moore, and D. A. Lockner (2001), Permeability
Since consistent predictions exist between the ensemble reduction in granite under hydrothermal conditions, J. Geophys. Res.,
suite of concentration, thin section, and X-ray CT data, 106(B12), 30,551 30,560.
these models suggest that flow constrictions along the Piggott, A. R., and D. Elsworth (1993), Laboratory assessment of the
length of the wormhole present large impedances that are equivalent apertures of a rock fracture, Geophys. Res. Lett., 20(13),
1387 1390.
not readily accommodated using aggregate conduit cross- Plummer, N. L., T. M. L. Wigley, and D. L. Parkhurst (1978), The kinetics
sectional dimensions. In the absence of imaging data, it of calcite dissolution in CO2-water systems at 5 to 60C, and 0 to 1 atm
would not be possible to remove the ambiguity between the CO2, Am J. Sci., 278(2), 179 216.
mass efflux data and the observed change in hydraulic Polak, A., D. Elsworth, H. Yasuhara, A. S. Grader, and P. M. Halleck
(2003), Permeability reduction of a natural fracture under net dissolution
impedance of the sample. This observed switching in by hydrothermal fluids, Geophys. Res. Lett., 30(20), 2020, doi:10.1029/
permeability under net dissolution has implications in the 2003GL017575.
rate of formation of karst and in potential changes in Pyrak Nolte, L. J., L. Myer, N. G. W. Cook, and P. A. Witherspoon (1987),
porosity and permeability that may accompany the injection Hydraulic and mechanical properties of natural fractures in low perme-
and recovery of surface waters in the proposed restoration ability rock, in Proceedings of the International Congress for Rock
Mechanics, pp. 225 231, Int. Soc. for Rock Mech., Lisbon.
of the Florida Everglades. These observed changes in Rege, D. S., and H. S. Fogler (1989), Competition among flow, dissolution,
porosity affect the rates of channel evolution in fractures, and precipitation in porous media, AIChE J., 35(7), 1177 1185.
their resulting morphology, and their consequences in Silliman, S. E. (1989), An interpretation of the difference between aperture
developing fast flow pathways, depressed entry pressures, estimates derived from hydraulic and tracer tests in a single fracture,
Water Resour. Res., 25(10), 2275 2283.
enhanced immiscible phase mobility, and suppressed mech- Singurindy, O., and B. Berkowitz (2003), Evaluation of hydraulic conduc-
anisms for retardation and attenuation of aqueous and tivity by precipitation and dissolution in carbonate rock, Water Resour.
particulate contaminants alike. Res., 39(1), 1016, doi:10.1029/2001WR001055.
Stock, S. R. (1999), X-ray microtomography of materials, Int. Mater. Rev.,
44(4), 141 164.
[43] Acknowledgments. This work is a result of partial support under Tenthorey, E., C. H. Scholz, and E. Aharonov (1998), Precipitation sealing
grants DOE-BES-DE-FG02-00ER15111 and ARC DP0209425. This sup-
port is gratefully acknowledged. The authors also thank Ozgen Karacan for and diagenesis: 1. Experimental results, J. Geophys. Res., 103(B10),
help in the CT scanning, and Susan Brantley for comments on an earlier 23,951 23,967.
version of the manuscript. Vukuturi, V. S., R. D. Lama, and S. S. Saluja (1974), Handbook on
Mechanics Properties of Rocks, vol. 1, Trans Tech Publ., Enfield, N. H.
References White, W. B. (2002), Karst hydrology: Recent developments and open
questions, Eng. Geol., 65, 85 105.
Aharonov, E., E. Tenthorey, and C. H. Scholz (1998), Precipitation sealing Wildenschild, D., J. W. Hopmans, C. M. P. Vaz, M. L. Rivers, D. Rikard,
and diagenesis: 2. Theoretical analysis, J. Geophys. Res., 103(B10), and B. S. B. Christensen (2002), Using X-ray computed tomography in
23,969 23,981. hydrology: Systems, resolutions, and limitatations, J. Hydrol., 267, 285
Domenico, P. A., and F. W. Schwartz (1998), Physical and Chemical 297.
Hydrogeology, 2nd ed., John Wiley, Hoboken, N. J. Witherspoon, P. A., J. S. Y. Wang, K. Iwai, and J. E. Gale (1980), Validity
Durham, W. B., W. L. Bourcier, and E. A. Burton (2001), Direct observa- of cubic law for fluid flow in a deformable rock fracture, Water Resour.
tion of reactive flow in a single fracture, Water Resour. Res., 37(1), 1 Res., 16(6), 1016 1024.
12. Zimmerman, R. W., and G. S. Bodvarsson (1996), Hydraulic conductivity
Hoefner, M. L., and H. S. Fogler (1988), Pore evolution and channel for- of rock fractures, Transp. Porous Media, 23(1), 1 30.
mation during flow and reaction in porous media, AIChE J., 34(1), 45
54.
Kelemen, P. B., J. A. Whitehead, E. Aharonov, and K. A. Jordahl (1995),
Experiments on flow focusing in soluble porous media, with applications 

to melt extraction from the mantle, J. Geophys. Res., 100(B1), 475 496. D. Elsworth and A. S. Grader, Energy Institute and Department of
Ketcham, R. A., and W. D. Carlson (2001), Acquisition, optimization and Energy and Geo-Environmental Engineering, Pennsylvania State Univer-
interpretation of X-ray computed tomographic imagery: Applications to sity, 403 Academic Activities Building, University Park, PA 16802, USA.
the geosciences, Comput. Geosci., 27, 381 400. ([email protected])
Krauskopf, K. B., and D. K. Bird (1995), Introduction to Geochemistry, J. Liu, School of Oil and Gas Engineering, University of Western
McGraw-Hill, New York. Australia, 35 Stirling Highway, Western Australia 6009, Australia.
Lasaga, A. C. (1984), Chemical kinetics of water-rock interactions, J. Geo- A. Polak, Department of Civil and Environmental Engineering,
phys. Res., 89(B6), 4009 4025. Technion-Israel Institute of Technology, Haifa 32000, Israel.

10 of 10