Geophys. J. Int. (2011) 184, 897–906 doi: 10.1111/j.1365-246X.2010.04885.x

GJI

Sei

smol

ogy

Stress-induced anisotropy in brine saturated shale

C. Delle Piane,1 D. N. Dewhurst,1 A. F. Siggins2 and M. D. Raven3

1CSIRO Earth Science and Resource Engineering, Australian Resources Research Centre, 26 Dick Perry Avenue, Kensington, WA 6151, Australia.E-mail: claudio.dellepiane@csiro.au2CSIRO Earth Science and Resource Engineering, Private Bag 10, Clayton South, VIC 3169, Australia3CSIRO Land and Water, Waite Rd, Urrbrae, SA 5064, Australia

Accepted 2010 November 7. Received 2010 September 22; in original form 2010 July 13

S U M M A R YThis paper reports the intrinsic and crack-induced anisotropic properties of a set of preserved,brine-saturated shale samples and their response to external stresses. We used undrainedmultistage triaxial tests to evaluate how the ultrasonic wave velocities and their anisotropychanged with increasing isotropic and differential stress conditions. In addition, the impactof stress orientation with respect to fabric orientation was evaluated. An array of ultrasonictransducers allowed to measure five independent wave velocities which were used to calculatethe elastic properties of the shale. Results indicate that in this shale P- and S-wave velocitiesvary with stress in a different manner dependent on the maximum principal stress orientationwith respect to the fabric. Where the maximum stress is normal to bedding, Vpv and Vs1

increase monotonically with increasing effective stress. However Vph and Vsh decrease duringindividual loading stages but increase from stage to stage as confining pressure increases.The reverse occurs when the microfabric is parallel to the maximum principal stress. Wherethe maximum stress is bedding normal, velocity anisotropy decreases as differential stressincreases; when maximum stress is fabric parallel, anisotropy increases. Elastic anisotropy isrelated to the initial composition and the spatial distribution of the different minerals (fabric)in the sediment and the presence of microfractures, while changes in elastic anisotropy resultfrom the applied stresses, their orientation with respect to the rock fabric and the degree ofstress anisotropy.

Key words: Geomechanics; Microstructures; Body waves; Seismic anisotropy; Wave prop-agation; Acoustic properties.

I N T RO D U C T I O N

Shales are the most abundant sedimentary rock type and play animportant role in oil and gas drilling operations, where they makeup for long portions of the drilled section and act as seals for hydro-carbon reservoirs. Furthermore, several countries are consideringshales as possible host lithologies for radioactive waste reposito-ries and as natural seals in CO2 geo-sequestration projects. Hence,shales have a primary civil and economic importance but never-theless our understanding of their properties and behaviour underdifferent physical conditions is very limited. Despite many yearsof research and their important industrial implications, the elasticproperties of shales are not well understood, often the result ofpoorly preserved samples. In particular, the monitoring of elasticwave velocities and anisotropy in preserved shales under loadingis relatively uncommon, although a better understanding of suchresponses would be helpful in resolving ambiguities in seismic pro-file interpretation and the seismic signature of subsurface fluids.This relative rareness is caused partly by: (1) the difficulty of ob-

taining well-preserved samples from the subsurface, and (2) thetime consuming nature of laboratory tests resulting from the lowpermeability of shales.

Most of the experimental determinations of elastic wave veloc-ities on shale samples have been performed under isotropic stressconditions (Johnston & Toksoz 1980; Jones & Wang 1981; Lo et al.1986; Johnston 1987; Johnston & Christensen 1995; Hornby 1998;Stanley & Christensen 2001). In addition, elastic wave velocitymeasurements on shale are reported (Sarout et al. 2007, Sarout &Gueguen 2008) under triaxial and polyaxial loading (Yin 1992),and under uniaxial loading (Podio et al. 1968). Many of the labo-ratory studies are conducted without control of pore pressure or inundersaturated conditions not reflecting the in situ settings of thesediments and raising concerns about the known impact of less than100 per cent saturation on P-wave velocity. Only a few studies (e.g.Jakobsen & Johansen 2000; Dewhurst & Siggins 2006; Sarout et al.2007) report elastic wave velocity data on saturated shale samplesunder triaxial loading. A collection of experimental data obtainedfrom dry and wet shale samples is presented by Sayers (1999) to

C© 2010 CSIRO 897Geophysical Journal International C© 2010 RAS

Geophysical Journal International

898 C. Delle Piane et al.

introduce a theoretical formulation of the stress-dependent seismicanisotropy of shales.

This study investigates the dynamic elastic response of satu-rated Australian Northwest Shelf shale with the main aim to ex-amine changes resulting from the application of both isotropic andanisotropic stress fields. P- and S-wave velocities were measuredat ultrasonic frequencies with different propagation direction withrespect to the sample geometry; dynamic elastic moduli could thenbe calculated and used to estimate the anisotropy evolution duringmultistage triaxial tests

The anisotropic nature of shales mainly results from the align-ment of anisotropic plate-like clay minerals, which in turn is aconsequence of sedimentation and subsequent mechanical and di-agenetic compaction. At the microscale shales, can therefore beapproximated as transversely isotropic (TI) materials, with the axisof rotational symmetry normal to bedding. Questions may arise aswhether this degree of symmetry is maintained during anisotropicloading as microfractures and dilation occur inside and outside theplane of isotropy while strain is accommodated. An evaluation ofthe symmetry evolution of the samples with respect to the loadingconfiguration during triaxial tests is also discussed in the paper.

Finally the shale was also characterized in terms of composi-tion and microstructure as well as main physical and petrophysicalproperties (porosity, bulk density, etc.).

M E T H O D S

General characterization

The experimental programme was run on a series of plugs recoveredfrom cores drilled on the Australian Northwest Shelf, from a depthbelow 2000 m. Cylindrical core plugs of nominally 76 mm in lengthand 38 mm in diameter were taken for geomechanical testing. Thespecimens were drilled both parallel (horizontal plugs) and normal(vertical plugs) to the macroscopic bedding of the sediment. Theplugging was carried out using a mineral oil as cooling fluid soas not to alter chemistry and the water content of the rock. Allthe specimens were then stored under a light process oil to avoidevaporation of the pore fluids.

Offcuts from the drilling operations were used for general sam-ple characterization (e.g. composition, porosity and density) andmicrostructural evaluation. Quantitative X-ray diffraction (XRD)analyses were performed on six specimens of Northwest Shelf shale(Table 1). Details of the sample preparation and XRD methodologyare given in the Appendix.

Table 1. Average composition of the bulk rock, from 0.2to 2 μm and <0.2 μm fraction of the Northwest Shelfshale obtained from X-ray diffraction analyses based onsix specimens. I-S, mixed layer illite-smectite.

Bulk 0.2–2 μm <0.2 μm

Quartz 39.8 9.1Siderite 12.5 6.0Orthoclase 3.8 0.8Kaolin 21.8 50.8 34.7Mica 2.1 3.2Illite/Smectite 19.2 29.3 65.3Pyrite 0.4 0.2Anatase 0.3 0.7Calcite 0.1Percentage illite in I-S 95

Polished thin sections were analysed under optical and scanningelectron microscopes (SEM) to visualize the main microstructuralfeatures, such as mineral phases and porosity distribution, grain-to-grain contacts and particle alignment.

Geomechanical testing

Consolidated undrained multistage triaxial tests (Fjaer et al. 1992)were performed on preserved shale samples using a Terratek testingmachine (Dewhurst & Siggins 2006). The equipment comprises ahigh stiffness load frame, a pressure vessel and three independentstepping motor pumps for cell and pore pressure control, as wellas for axial load. The operational limits of the rig are 70 MPaconfining pressure (oil used as a confining medium), 70 MPa porepressure and 400 MPa axial stress on a 38 mm diameter sample.All experiments were conducted at room temperature; data loggingand pump control are based on a LabVIEW program. The sampleassembly includes the following:

(1) a cylindrical sample mounted between top and base platens;(2) an impermeable Viton membrane (0.75 mm thick), isolating

the specimen from the confining fluid and housing the radial andoff-axis ultrasonic P-wave and S-wave transducers;

(3) two steel platens housing ultrasonic P and S transducers withprovision for pore pressure measurements placed at both ends ofthe specimen;

(4) two diametrically opposed linear variable differential trans-formers (LVDT) clamped on the top and bottom platens to measureaxial displacements;

(5) a load cell placed underneath the bottom platen.

This assembly configuration allows the full stiffness tensor to bemeasured using a single specimen (Wang 2002; Dewhurst & Siggins2006) assuming the tested rock is a TI body (Fig. 1).

Once the sample is installed in the rig, an initial confining pres-sure of 15 MPa is applied, while pore pressure is simultaneouslyraised to 5 MPa at a rate of 0.5 MPa per minute. A saline solution(3.5 per cent NaCl) was used as a pore fluid. The sample is left tosaturate and equilibrate for about a week until strain and ultrasonicmeasurements reach a steady value. Axial load is then applied ata strain rate of ∼10−7 per second under undrained conditions untilwithin approximately 10 per cent of the peak strength, after whichthe load is decreased to 1 kN, the pore pressure allowed to drainto 5 MPa and confining pressure increased to the next level underdrained conditions. During the equilibration time a constant load of1 kN (corresponding to a differential stress of <1 MPa) is applied tomaintain a good coupling between the sample and the steel platensand ensure the transmission of axial ultrasonic impulses. Stagesof confining pressure for each sample were increased in steps of10 MPa from 15 to 65 MPa. At the last increment of confiningpressure (65 MPa) the sample is loaded until failure. Ultrasonicmeasurements are taken at increasing values of axial load for eachstage, starting from the isotropic stress point (i.e. zero axial load).The use of multistage tests has both pros and cons. The main reasonsfor using this type of testing are twofold, (1) shortage of enoughsingle samples to get a decent failure envelope and (2) heterogene-ity in properties between core plugs. This can be a significant issuein shales, especially for rock physics properties. The downside ofmultistage testing is potentially fatiguing samples from repeatedcycling above the elastic yield point of the material. However, therelative paucity of shale samples and likely heterogeneity has ledus to prefer the multistage test for gathering rock physics data onshales.

C© 2010 CSIRO, GJI, 184, 897–906

Geophysical Journal International C© 2010 RAS

Anisotropy in shales 899

Figure 1. Schematic diagram illustrating the ultrasonic transducers’ configuration with respect to the cylindrical sample and the axial load. For a samplecut normal to bedding, the microfabric and microfractures are normal to the maximum principal stress (left-hand side); for a sample cut parallel to bedding(right-hand side), microfabric and any associated microfractures are parallel to the maximum principal stress. Dashed lines represent bedding planes, arrowsindicate the direction of applied maximum principal effective stress (σ ′

1). The ultrasonic velocities measured tests are: P wave normal to bedding plane (Vpv), Pwave parallel to bedding plane (Vph), S wave normal to bedding with particle motion parallel to bedding (Vs1), S wave parallel to bedding with particle motionparallel to bedding (Vsh) and quasi-P at 45◦ from the sample axis (qVp45). Double headed arrows in the figure indicate the plane of polarization of the S waves.

The stress–strain data from multistage triaxial tests were usedto determine the specimen’s Mohr–Coulomb failure envelope. Thefailure envelope can be determined from the stress (σ 1, σ 3, porepressure) conditions at the time of failure following the equations:

σ ′n = 0.5(σ ′

1 + σ ′3) − 0.5(σ ′

1 − σ ′3) cos 2θ, (1)

τ = 0.5(σ ′1 − σ ′

3) sin 2θ, (2)

where the prime indicates effective stress values and θ is the anglebetween σ 1 and the failure plane. Stress–strain relationships fromthe performed tests were also used to determine the Young’s modulusof the Northwest Shelf shale and its evolution under changing stressconditions.

Ultrasonic measurements

Ultrasonic velocity measurements were made following the pulsetransmission technique described by (Birch 1960) using the trans-ducer configuration described by Dewhurst & Siggins (2006). Themethod consists of measuring the traveltime of an elastic pulsethrough a rock sample of known length. The nominal excitationfrequencies used in this study were 0.5 MHz for P waves and 0.5and 0.25 MHz for S waves. Using two frequencies for the S wavesallowed to compare two waveforms recorded along the same travelpath and reduced the uncertainty in the arrival time discrimination,which for S waves is often obscured by the coda of the precursoryP waves. The ultrasonic propagation system consists of a pulser-receiver and a digital oscilloscope recording the signals. Arrivaltimes were manually picked from the digitized waveforms; eachtrace consisting of 1000 points over a time range of 1 × 10−4 s.

Raw data were corrected for the platen thickness and converted toultrasonic velocity using the length of the specimen at the corre-sponding pressure conditions. Considering the uncertainties in thearrival times (±0.1 μs for P waves, and ±0.2 μs for S waves) andthose related to the dimension of the samples and the positioning ofthe transducers, the corresponding ultrasonic velocity errors can beestimated to sum up to approximately ±2 per cent for P waves and±4 per cent for S waves (Sarout & Gueguen 2008).

Vertical plugs

The ultrasonic velocities measured during triaxial tests were asfollows: P wave normal to bedding plane (Vpv), P wave parallel tobedding plane (Vph), S wave normal to bedding with particle motionparallel to bedding (Vs1), S wave parallel to bedding with particlemotion parallel to bedding (Vsh) and quasi-P at 45◦ from the sampleaxis (qVp45). The measurement of five velocities is justified by thefact that shales plugs cored normal to bedding can be convenientlydescribed as TI rocks (e.g. Sarout et al. 2007). Transverse isotropicelasticity ranks second to isotropic elasticity in the degree of expres-sion of elastic symmetry in the material behaviour. Media exhibitingtransverse isotropy include laminated materials and stratified rocks.In the latter case, all lines lying in the plane of bedding are axesof elastic symmetry. The only other axis of elastic symmetry is thenormal to the plane of isotropy.

Velocities and elastic coefficients are related via the followingequations:

C11 = ρV 2ph → Vph = (C11/ρ)1/2, (3)

C33 = ρV 2pv → Vpv = (C33/ρ)1/2, (4)

C© 2010 CSIRO, GJI, 184, 897–906

Geophysical Journal International C© 2010 RAS

900 C. Delle Piane et al.

C44 = ρV 2s1 → Vs1 = (C44/ρ)1/2, (5)

C66 = ρV 2sh → Vsh = [(C11 − C12)/2ρ]1/2 , (6)

C13 = −C44 + [(C11 + C44 − 2ρV 2

p45

) (C33 + C44 − 2ρV 2

p45

)]1/2,

(7)

where ρ is the bulk density of the sample.Elastic coefficients derived from the measured velocities and

sample density, were used to calculate the anisotropy factors ε andγ (Thomsen 1986) defined as

ε = C11 − C33

2C33

(8)

and

γ = C66 − C44

2C44. (9)

Horizontal plug

To verify the TI assumption in the case of loading with maximumprincipal stress parallel to bedding and in an effort to track the evo-lution of elastic anisotropy during loading six ultrasonic velocitieswere measured during triaxial testing. We extracted the P arrivalfrom ultrasonic waveforms recorded at three orthogonal directionswith respect to the sample axis. Therefore the recorded P waveswere Vph (with propagation direction parallel to the sample axis andto the bedding), Vph2 (with propagation direction parallel to bed-ding and across the cylinder diameter) and Vpv (with propagationdirection normal to bedding across the diameter of the cylinder). Itshould be noted that only two P-wave transducers are available inour experimental configuration (Vpv and Vph), and the precursory Pwave recognized in the S waveform was used for the arrival pick-ing of the third P-wave velocity Vph2. In addition S wave travellingnormal to bedding with particle motion parallel to bedding (Vs1),S wave parallel to bedding with particle motion parallel to bedding(Vsh) and quasi-P at 45◦ from the sample axis (qVp45) were alsomeasured.

R E S U LT S

Rock description

The X-ray diffraction results indicate a fairly homogeneous mineral-ogy for the Northwest Shelf shale samples. The shale is composed of

dominant quartz, kaolin, illite/smectite and siderite as major con-stituents and minor amounts of orthoclase, mica, pyrite, anatase(titanium oxide) and calcite (Table 1). No discrete illite was iden-tified using the XRD technique; analysis of the <0.2 μm fractionrevealed a bimineral composition with 35 per cent of kaolin and65 per cent of illite/smectite. The largest part (95 per cent) of theillite/smectite is composed of illite. The clay fraction, defined as theweight percent of the <2 μm content, is ∼65 per cent. Clay content(sum of weight percent clay minerals) is approximately 40 per cent.

Several thin sections were prepared from offcuts of the North-west Shelf shale: all sections show similar features and exhibita macroscopic primary foliation (bedding) defined by alternatinglayers of coarse (∼100 μm in equivalent diameter), well sorted,angular grains and dark, fine-grained material (Fig. 2). The shapeof the coarse grains is round to elongate with low aspect ratios.The elongated grains are preferentially aligned parallel to the foli-ation. In the clay-rich foliations, the coarse grains are not in con-tact with one another but are dispersed and surrounded by a fine-grained, light coloured matrix. Large particles are often surroundedby a fine-grained, high birefringence material, which is likely to becalcite.

The optical characteristics of the coarse grains allow identify-ing them as quartz, K-feldspar, glauconite and chlorite. Occasionalgrains of pyrite are also found. Quartz grains are often decoratedwith fluid inclusions arranged in randomly oriented trails. The fine-grained layers are composed of mixed materials too fine to be re-solved under the optical microscope. The thickness of these foliais around 500 μm. Thin, straight, parallel fissures are often foundwithin the fine-grained layers and are oriented parallel to the mainfoliation and are probably due to stress relief.

Under the SEM, the shale is composed of alternating layers ofcoarse, quartz-rich nature and fine-grained material exhibiting avast heterogeneity in terms of mineralogical content (Fig. 3). Someporosity can be observed at the grain contacts within the coarse-grained layers; here precipitated mineral phases such as carbonatestend to occlude the pores (Fig. 3); in the fine-grained layers, smallelongated fissures are present oriented parallel to the laminations(Fig. 3). Fabric anisotropy can be recognized in terms of weak shapepreferred orientation of the elongated particles with the long axisof the grains oriented subparallel to the macroscopic lamination(Fig. 3).

Quartz and feldspar grains are floating in a fine-grained ma-trix although some grain-to-grain contact can be observed. Quartzgrains are generally intact and some are decorated with fluid in-clusions; feldspar grains are often fractured and/or altered withfractures radiating from point-contacts with other minerals (Fig. 3).

Figure 2. Thin section photographs of the NWS shale. (A) Alternating bright (silt-rich) and dark (clay-rich) layers define the bedding, that is, plane ofanisotropy. (B) Note the difference in grain size between the layers.

C© 2010 CSIRO, GJI, 184, 897–906

Geophysical Journal International C© 2010 RAS

Anisotropy in shales 901

Figure 3. SEM images in backscattered electron mode, contrast in grey level corresponds to contrast in chemical composition. (A) Silty layer: the coarsegrains are floating in a matrix of clay, only few grain-to-grain contacts are noted. White arrow indicates open porosity around a fragmented quartz grain. (B)Clay layer, the white arrow indicates thin stress-relief fractures oriented subparallel to bedding. (C) Grain contacts between different minerals. Q = quartz;M = muscovite; S = siderite; C = clay aggregate. (D) Kinked muscovite grain, note the secondary porosity created along the basal plane (arrow). A patch oforganic matter can be recognized to the left of the white arrow.

Diffusive mass transfer has resulted in sutured grain boundariesbetween different mineral phases and precipitation of minerals attriple junctions and along grain boundaries. The backscattered SEM(BSEM) images reveal the presence of abundant heavy minerals(pyrite, titanium oxide) in the fine-grained layers.

Mechanical data

Stress–strain relationships from geomechanical tests were usedto determine the Young’s modulus of the Northwest Shelf shale(Fig. 4). The tangent Young’s modulus was calculated as the slopeof the axial stress-axial strain curve between 40 and 60 per cent,of the peak strength at each stage of the triaxial test. The peakstress values also allowed the determination of the failure envelopeand of the Mohr–Coulomb failure parameters of the sediment. TheNorthwest Shelf Shale is rather weak with a cohesive strength ofapproximately 4 MPa and a friction coefficient of 0.3. The rockstrength parameters and elastic moduli are summarized in Table 2as a function of effective confining pressure.

The mechanical response from the three different specimens isrelatively uniform; the highest peak strength at failure is exhib-ited by the sample with the loading oriented normal to the bedding(Fig. 4). Mechanical anisotropy is very low when peak stress val-ues are compared at the same condition of isotropic pressure (seeTable 2).

Rock physics data

Rock physics data were collected during the deformation cycles.The calculated ultrasonic velocities derived for each experiment areshown in Figs 5–7. Unfortunately we could not record all five ve-locities for each test due to signal transmission issues and/or poor

signal-to-noise ratio. The figures illustrate the response of the vari-ous ultrasonic velocities to changing stress conditions. SpecificallyP- and S-ultrasonic velocities are plotted as a function of meaneffective pressure defined as

σ ′m =

(σ ′

1 + σ ′2 + σ ′

3

3

), (10)

where σ ′1, σ ′

2 and σ ′3 are the maximum, intermediate and minimum

principal effective stresses, respectively. Effective stress (σ ′) is de-fined as total stress (σ) minus the pore fluid pressure (Pp). In thestandard axisymmetric triaxial tests used here, σ ′

2 = σ ′3. Mean ef-

fective stress was used as it gives the best synthesized representationwhen both axial stress and confining pressure are changed during thetest; moreover, published studies (e.g. Atkinson & Bransby 1978)indicate this parameter as the governing one for porosity evolutionof clays and shales during burial and compaction.

Vertical plug

Fig. 5(a) shows a velocity-mean effective stress plot for sample V1,a core plug cut normal to bedding. For this sample, we were unableto record Vsh due to connection issues with the transducer after thetest commenced. For the other velocities calculated (Fig. 5a), Vpv in-creases relatively monotonically both within each of the six stages ofthe test and also between stages from ∼3150 m s–1 to ∼3500 m s–1

across the mean effective stress range of 10–73 MPa; Vs1 behaves ina similar fashion, rising from ∼1800 m s–1 to ∼2050 m s–1 over thesame stress range. However, Vph remains approximately constantduring the early parts of individual stages (low differential stress)but then decreases towards the end of individual axial loading stagesas differential and mean effective stresses increase. From stage tostage, increasing confining pressure, Vph increases at any given

C© 2010 CSIRO, GJI, 184, 897–906

Geophysical Journal International C© 2010 RAS

902 C. Delle Piane et al.

Figure 4. (a) Stress–strain behaviour recorded from the three multistage triaxial experiments. Differential stress is defined by the difference between themaximum (σ ′

1) and minimum (σ ′3) principal effective stresses; (b) Evolution of static Young’s modulus as a function of effective confining pressure.

Table 2. Summary of the test condition and results. CP, confining pressure.

Effective Max σ 1–σ 3 Young’s Max σ 1–σ 3 Young’s Max σ 1–σ 3 Young’sCP (MPa) (MPa) (GPa) (MPa) (GPa) (MPa) (GPa)

10 21.31 6.56 13.04 5.08 14.73 5.1920 29.91 8.83 26.76 7.66 27.35 8.8130 34.29 8.44 36.05 8.59 35.17 9.0940 38.95 9.57 40.25 9.78 39.20 10.4150 42.27 10.98 44.28 10.77 41.46 9.9360 60.72 10.81 54.16 9.46 41.53 11.54Sample V1 H2 H1

level of differential stress. The off axis P-wave velocity (qVp45)behaves in an intermediate fashion, mostly increasing monotoni-cally but with occasional decreases in velocity within individualstages.

Anisotropy of P-wave velocity (ε) only (no Vsh measurement) isshown in Fig. 5(b) and the six individual stages of the multistage tri-axial test are delineated. ε decreases during each individual loadingstage from values of 0.11–0.14 to values of 0.06–0.09 as differential

C© 2010 CSIRO, GJI, 184, 897–906

Geophysical Journal International C© 2010 RAS

Anisotropy in shales 903

Figure 5. Ultrasonic velocities (a) and anisotropy parameters (b) as a function of mean effective stress for sample V1, where the core plug is cut normal tobedding. Note moderate-to-low P-wave anisotropy which decreases within individual axial loading stages.

stress increases within each stage. The effect of isotropic stress (i.e.confining pressure) can be assessed from the first point in each stageand this does not vary much or systematically as stress increases.

Horizontal plugs

Tests run with bedding parallel loading exhibit a different behaviour.In this case, Vph tends to increase within each loading cycle andbetween loading cycles, while Vpv decreases within individual stagesbut generally increases between stages at any given differentialstress (Fig. 6). Two multistage tests were run on separate shale coreplugs taken parallel to bedding. In experiment H1 (Fig. 6a), Vph

rises consistently from 3426 to 3820 m s–1 over a mean effectivestress range of 10 to ∼70 MPa; however, while Vpv increases from3420 to 3690 m s–1 across the six stages, within each individualstage, decreases in velocity are noted of between 35 and 130 m s–1,with the drop more pronounced at lower effective stress. Vsh remainsrelatively constant across the entire stress range, while Vs1 showsvery slight decreases. Unfortunately, the electrical connection forthe off axis P wave was lost early in the experimental cycle and wewere unable to record these waveforms. A second set of experimentswas run on a horizontal core plug due to the transducer failure. Inthis test (H2; Fig. 6b), data consistent with the previous experimentwere recovered. Vph increases from 3444 to 3759 m s–1 across asimilar stress range and again increases within each stage of axial

loading. Vph2 shows values of velocity almost overlapping with Vph

at low mean effective stress but diverges from Vph as stress valuesincrease. Overall the Vph2 changes from 3390 to 3520 m s–1 alongthe range of mean effective stress applied to the sample; moreoverVph2 slightly decreases within each axial loading stage.

Vpv increases from a low value of 3133 m s–1 to a high of3582 m s–1 over the applied mean effective stress range but de-creases within each axial loading stage by between 30 and 100 m s–1.The off axis P-wave velocity lies close to Vpv and varies between3288 and 3528 m s–1. Fig. 6(b) also shows significant differencesbetween Vs1 and Vsh, with the former lying between 1584 and1706 m s–1 between 10 and 70 MPa mean effective stress andthe latter between 1971 and 2193 m s–1. Vsh gently increases withinand between stages, whereas Vs1 slightly decreases within stagesbut increases from one stage to the next.

The velocity versus pressure trend observed in sample H2 forVph and Vph2 indicates that as mean effective stress increases thebedding plane does no longer represent a plane of isotropy as at thelowest stress levels. It is therefore not correct to treat the sampleas a TI body. In the following values of anisotropy extracted fromvelocity measurements on the horizontal plugs will be referred toas AVP and AVS for P- and S-wave anisotropy, respectively. Theanisotropy in these cases is calculated as

AVP = (Vph − Vpv)

(Vph)× 100, (11)

C© 2010 CSIRO, GJI, 184, 897–906

Geophysical Journal International C© 2010 RAS

904 C. Delle Piane et al.

Figure 6. Ultrasonic velocities as a function of mean effective stress for two shale core plugs (a = H1; b = H2) cut parallel to bedding. Note that in (b) twoVph and Vph2, represents the two orthogonal P-wave velocity measured within the isotropy plane (oriented parallel to σ1 during the triaxial loading). Vph andVph2 are overlapping at low effective mean stress conditions, but start diverging as the stress conditions change, indicating loss of TI symmetry.

AVS = (Vsh − Vs1)

(Vsh)× 100. (12)

An additional parameter is introduced to characterize the azimuthalanisotropy (within the bedding plane) of sample H2.

AVP−Az = (Vph − Vph2)

(Vph)× 100. (13)

The two horizontal specimens, both loaded parallel to bedding,are moderately anisotropic and the anisotropy parameters are sen-sitive to changes in stress configuration and magnitude (Fig. 7).Sample H1 shows a P-wave anisotropy scattering around zero andslightly increasing within each loading cycle. S-wave anisotropyhas an initial value of approximately 30 per cent that is graduallyreduced to around 20 per cent with increasing confining pressure(Fig. 7a). As observed for the AVP parameter, the AVS value tendsto increase within each individual loading cycle (Fig. 7a). Simi-lar overall behaviour is observed for sample H2 (Fig. 7b), thoughthe S-wave anisotropy shows much larger scatter and both P- andS-wave anisotropies show steeper increment during loading at lowlevels of mean effective stress. Moreover the measurement indicatesan initial isotropy within the bedding plane (AVP−Az close to zero)which is gradually lost as stress conditions increase and damage isintroduced in the specimen and AVP−Az approaches values of 8 percent (Fig. 7b).

Reproducibility of the data is high, with velocities measuredon different samples with the same orientation with respect to thedeviatoric stress are very similar indicating reasonable homogeneityof the different specimens.

D I S C U S S I O N

Tectonic loading may lead to generation and propagation of crackswithin rocks which in turn result in a decrease in seismic wavevelocities. The crack pattern would be mainly influenced by

(i) macroscopic lamination in the sediment, and(ii) magnitude and orientation of the applied stress system.

As already described by (Dewhurst & Siggins 2006) for the Mud-erong Shale and (Popp & Salzer 2007) for the Opalinus Clay, devi-atoric loading induces marked changes in the physical properties ofthe tested shale: bedding-parallel loading favours dilatancy alongsedimentary surfaces and pre-existing fractures therefore increas-ing elastic anisotropy, while bedding normal loading tends to closesuch bedding parallel fissure and reduce the elastic wave anisotropy.

On average both P- and S-wave velocities increase by approxi-mately 10 per cent of their initial value over an increment of meaneffective stress of around 60 MPa. It has been argued that S-wavesvelocities in shales may be strongly affected by the stress-inducedloss of interlayer water from smectite (Dewhurst & Siggins 2006);however, these Northwest Shelf shale samples are poor in smectiteand therefore do not show such pressure sensitive controls on theshear velocities.

In terms of absolute values of ultrasonic velocities, our experi-mental results are comparable to other laboratory measurements onshale of different composition and age (e.g. Dewhurst & Siggins(2006) for illite-smectite rich; Popp & Salzer (2007) for kaoliniterich samples and Sarout & Gueguen (2008) for illite-chlorite richspecimens).

C© 2010 CSIRO, GJI, 184, 897–906

Geophysical Journal International C© 2010 RAS

Anisotropy in shales 905

Figure 7. P-wave and S-wave anisotropy factors as a function of mean effective stress for two shale core plugs cut parallel to bedding (a = H1; b = H2). Notethat two anisotropy parameters are used in the case of plug H2: one (AVP) denoting the difference between Vpv and Vph and one (AVP−az) characterizing theanisotropy of P waves travelling along two orthogonal directions within the bedding plane.

In terms of elastic anisotropy, the very low P-wave anisotropyseems to reflect the weak preferred orientation of elongated mineralsseen through microstructural investigations (Figs 2 and 3), and maybe linked the mineralogy of the sample in that the Northwest Shelfshale is rich in rigid grains and kaolinite, neither of which arestrongly intrinsically anisotropic in nature. While the relatively highvalue of S-wave anisotropy can be correlated with the presenceof aligned microfractures caused by differential stress during theexperiments and stress relief due to the core extraction.

In first-order observations P-wave velocity is a measure of bulkmodulus K, and the S-wave velocity of the shear modulus, G (Scholz2002). Under the experimental P-T conditions explored one can as-sume that Krock ≈ Kwater whereas Grock>>Gwater; therefore the intro-duction of aligned saturated cracks in undisturbed sediments (likein the case of undrained triaxial loading test on saturated shales)will have little effect on the velocity of compressional elastic wavespropagating in different directions across the sample and on theiranisotropy. On the other hand, the presence of iso-oriented, thin,fluid filled fissures will significantly reduce Vs travelling parallelto and with particle motion normal to the cracks long axis, whichstrongly affects S-wave anisotropy. Given the undisturbed rock ma-trix has little intrinsic anisotropy due to high silt content and weakalignment of the clay fraction; the observed S-wave anisotropy isdirectly linked to the presence of cracks appearing as a result ofthe applied anisotropic stress field. Such behaviour is not specificto shale but has been observed on other crustal lithologies tested

in the laboratory in the presence of fluid saturated microracks (i.e.Gao & Crampin 2003).

The interplay between microcrack closure and fissure openingduring axial loading at elevated effective pressure is likely responsi-ble for the observed evolution of elastic anisotropy. Opposite trendsin anisotropy as a function of mean effective stress are observed inorthogonally oriented samples with respect to the differential stress.These results emphasize that anisotropy estimation for seismic in-terpretation should take into account both vertical (overburden) andhorizontal (tectonic) stresses. Moreover the experimental results in-dicate that anisotropic stress field may alter the elastic symmetry ofan initially TI sediment.

C O N C LU S I O N S

Ultrasonic velocities and dynamic elastic anisotropy were measuredon different samples of a Northwest Shelf shale during undrainedmultistage triaxial tests. Samples were chosen to have different bed-ding orientation with respect to the direction of applied differentialstress. The shale is moderately anisotropic, with both velocities andanisotropic parameters sensitive to changes in differential stress (i.e.stress anisotropy). Velocities increase from stage to stage with in-creasing mean effective stress suggesting a progressive increase ofstiffness of the sediment associated with crack closure. However,the orientation of bedding with respect to the loading direction is

C© 2010 CSIRO, GJI, 184, 897–906

Geophysical Journal International C© 2010 RAS

906 C. Delle Piane et al.

critical when interpreting the velocity and anisotropy evolution ofthe shale with stress. Bedding parallel loading favours dilatancy ofpre-existing fractures therefore increasing anisotropy, while bed-ding normal loading tends to close such bedding parallel fissuresand reduce the anisotropy. These results indicate that anisotropicelastic behaviour under anisotropic stress fields should be consid-ered when estimating elastic wave velocities and their anisotropyfor pore pressure prediction and fluid identification in field sur-veys. Accounting for all acting stresses and rock anisotropy wouldimprove seismic interpretation and pore pressure prediction.

A C K N OW L E D G M E N T S

This contribution was performed as part of the IPETS (IntegratedPredictive Evaluation of Traps and Seals) project and was financiallysupported by the CSIRO Wealth from Oceans Flagship and spon-sored by Santos, Woodside, Schlumberger, Chevron, Anadarko,PIRSA and Origin Energy. Bruce Maney, Shane Kager, LeighKiewiet, Ian Penny and Peter Knowles are thanked for the mainte-nance of the experimental apparatus and their technical expertisein CSIRO’s rock mechanics laboratory. Joel Sarout, Julien Bourdetand Matthew Josh are thanked for their comments and suggestionson an earlier version of the manuscript. Two anonymous reviewersand Editor Xiaofei Chen are thanked for their efforts in improvingthe quality of the paper.

R E F E R E N C E S

Atkinson, J.H. & Bransby, P.L., 1978. The Mechanics of Soils-An Introduc-tion to Critical State Soil Mechanics, McGraw-Hill, London, 375 pp.

Birch, F., 1960. The velocity of compressional waves in rocks to 10 kbars,J. geophys. Res., 65, 1083–1102.

Dewhurst, D.N. & Siggins, A.F., 2006. Impact of fabric, microcracks andstress field on shale anisotropy, Geophys. J. Int., 165, 135–148.

Fjaer, E., Holt, R.M., Horsrud, P., Raaen, A.M. & Risnes, R., 1992.Petroleum Related Rock Mechanics, Vol. 33, Elsevier, Amsterdam.

Gao, Y. & Crampin, S., 2003. Temporal variations of shear-wave splittingin field and laboratory in China, J. appl. Geophys., 54, 279–287.

Hornby, B.E., 1998. Experimental laboratory determination of the dy-namic elastic properties of wet, drained shales, J. geophys. Res., 103,29 945–29 964.

Jakobsen, M. & T.A. Johansen, 2000. Anisotropic approximations for mu-drocks: a seismic laboratory study, Geophysics, 65, 1711–1725.

Johnston, D.H., 1987. Physical properties of shales at temperature and pres-sure, Geophysics, 52, 1391–1401.

Johnston, J.E. & Christensen, N.I., 1995. Seismic anisotropy of shales, J. geo-phys. Res., 100, 5991–6003.

Johnston, D.H. & Toksoz, M.N., 1980. Ultrasonic P and S wave attenuationin dry and saturated rocks under pressure, J. geophys. Res, 85, 925–936.

Jones, L.E.A. & Wang, H.F., 1981. Ultrasonic velocities in cretaceous shalesfrom the Williston basin, Geophysics, 46, 288–297.

Lo, T.W., Coyner, K.B. & Toksoz, M.N., 1986. Experimental determinationof elastic anisotropy of Berea sandstones, Chicopee shale, and Chelmsfordgranite, Geophysics, 51, 164–171.

Podio, A.L., Gregory, A.R. & Gray, K.E., 1968. Dynamic properties of 16dry and water-saturated green river shale under stress, J. Soc. PetroleumEng. (SPE), 8(4), 389–404.

Popp, T. & Salzer, K., 2007. Anisotropy of seismic and mechanical propertiesof Opalinus clay during triaxial deformation in a multi-anvil apparatus,Phys. Chem. Earth, 32, 879–888.

Sarout, J. & Gueguen, Y., 2008. Anisotropy of elastic wave velocities indeformed shales, part 1: experimental results, Geophysics, 73, D75–D89.

Sarout, J., Molez, L., Gueguen, Y. & Hoteit, N., 2007. Shale dynamic prop-erties and anisotropy under triaxial loading: experimental and theoreticalinvestigations, Phys. Chem. Earth, 32, 896–906.

Sayers, C.M., 1999. Stress-dependent seismic anisotropy of shales, Geo-physics, 64, 93–98.

Scholz, C.H., 2002. The Mechanics of Earthquake and Faulting, 2nd edn,Vol. 1, Cambridge University Press, Cambridge.

Stanley, D. & Christensen, N.I., 2001. Attenuation anisotropy in shale at ele-vated confining pressures, Int. J. Rock Mech. Mining Sci., 38, 1047–1056.

Thomsen, L., 1986. Weak elastic anisotropy, Geophysics, 51, 1954–1966.Wang, Z., 2002. Seismic anisotropy in sedimentary rocks, part 2: laboratory

data, Geophysics, 67, 1423–1440.Yin, H., 1992. Acoustic velocity and attenuation of rocks: isotropy, intrinsic

anisotropy, and stress-induced anisotropy, PhD thesis, Stanford Univer-sity Stanford, USA.

A P P E N D I X

Sample preparation and XRD analysis

Bulk samples were pre-ground for 15 s in a tungsten carbide me-chanical mill to pass through a 0.5 mm sieve. A 1 g subsample wasfurther ground for 10 min in a McCrone micronizing mill underethanol. The resulting slurry was oven dried at 60 ◦C then thor-oughly mixed in an agate mortar and pestle before being lightlypressed into stainless steel sample holders for X-ray diffractionanalysis.

An additional 10 g of the remaining bulk samples were shaken for10 min with 10 ml of a solution of 1M sodium chloride and deionisedwater. The slurries were repeatedly dispersed and centrifuged torecover all of the <0.2 μm and 0.2–2 μm fractions. The suspensionswere treated with acetic acid to remove carbonate minerals, calciumsaturated twice (using 1M CaCl2), washed with water then alcoholand oven dried at 60 ◦C. The resulting powders were lightly pressedinto stainless steel sample holders to achieve random orientation ofthe mineral particles for XRD analysis.

40 mg subsamples of the <0.2 μm powders were redispersedin deionised water and the suspensions sucked under vacuum ontocellulose nitrate filter discs to produce maximum orientation of theplaty clay mineral particles. These oriented samples were saturatedwith glycerol to aid identification of the clay mineral componentsby XRD.

XRD patterns were recorded with a PANalytical X’Pert Pro mul-tipurpose X-ray diffractometer using Fe filtered Co Ka radiation,1/4◦ divergence slit, 1/2◦ antiscatter slit and X’Celerator Si stripdetector. The diffraction patterns were recorded in steps of 0.016◦

2θ with a 0.4 s counting time per step, and logged to data files foranalysis.

Quantitative analysis was performed on the XRD data using thecommercial package SIROQUANT from Sietronics Pty Ltd. Thedata were first background subtracted and calibrated for the auto-matic divergence slit. The results are normalized to 100 per cent,and hence do not include unidentified or amorphous materials.

The percentage of illite layers in the interstratified illite-smectitewas determined by comparing the oriented and glycerolated XRDpatterns with patterns calculated by the NEWMOD for Windowsprogram.

C© 2010 CSIRO, GJI, 184, 897–906

Geophysical Journal International C© 2010 RAS