ORIGINAL PAPER

Observations of Fractures Induced by Hydraulic Fracturingin Anisotropic Granite

Youqing Chen • Yuya Nagaya • Tsuyoshi Ishida

Received: 30 April 2014 / Accepted: 22 February 2015 / Published online: 5 March 2015

� Springer-Verlag Wien 2015

Abstract To investigate how the viscosity of the frac-

turing fluid affects fracture propagation, hydraulic frac-

turing experiments using three fluids with different

viscosities (supercritical CO2, water, and viscous oil) under

the true tri-axial condition were conducted on anisotropic

granite specimens, and then the induced fractures were

microscopically observed via a fluorescent method. Frac-

tures induced by hydraulic fracturing are considerably

tortuous from a microscopic view. A higher viscosity cre-

ates a smoother fracture pattern. The tortuosity, which is

defined as the total fracture length along a pathway divided

by the direct length of the two ends of a fracture, ranges

from 1.05 to 1.13, demonstrating that the viscosity of

fracturing fluid influences the fracture propagation pattern

due to the different pathways of fracture propagation. In

addition, hydraulic fracturing can induce many derivative

pathways around the main fracture. Hydraulic fracturing

with a lower viscosity fluid forms a more complex fracture

network in rocks; the fracture induced by supercritical CO2

has the most branches along the main fracture. From these

observations, fracture propagation by hydraulic fracturing

sometimes develops by the shear fracture mode. This shear

fracturing is often observed for a low-viscosity super-

critical CO2 injection, which agrees with our results from

AE monitoring and waveform analysis.

Keywords Anisotropy � Granite � Hydraulic fracturing �Fracture � Microscopic observation � Fluid viscosity

1 Introduction

Hydraulic fracturing is commonly used to increase the

permeability of a depleted reservoir for oil and gas pro-

duction. In recent years, shale gas recovery, geothermal

energy extraction, and carbon dioxide (CO2) capture and

storage projects, have received attention as advanced ap-

plications of hydraulic fracturing.

Well stimulations using CO2 as the fracturing fluid have

been considered due to the advantages of CO2 (Sinal and

Lancaster 1987; Liao et al. 2009). In addition, CO2 injec-

tion has been considered to enhance shale gas recovery

because CO2 has a higher affinity for shale than methane

(Kalantari-Dahaghi 2010). In hot dry rock geothermal en-

ergy extraction, CO2 has been proposed as a fracturing and

circulating fluid in order to eliminate scaling in pipes and

to reduce the requirements for the circulating pumping

power (Brown 2000). Although it is important to under-

stand how the injected CO2 in CO2 capture and storage

projects infiltrates the surrounding rock mass (Xue et al.

2006; Spetzler et al. 2008), the characteristics of fracture

propagation induced with CO2 injection have yet to be

clarified.

In these projects, CO2 is usually injected into rocks at a

depth of more than 1000 m. The temperature and pressure

in this scenario are sufficient to induce the supercritical

state. The viscosity of CO2 in the supercritical state is very

low (i.e., around 5 % of that of water). Since it has been

pointed out that the viscosity may result in variations in the

induced fractures (Ishida et al. 2004), our studies consider

the viscosity of the fracturing fluid.

Y. Chen (&)

Department of Energy Science and Technology,

Graduate School of Energy Science, Kyoto University,

Yoshida-Honmachi, Sakyo-ku, Kyoto 606-8501, Japan

e-mail: chen@energy.kyoto-u.ac.jp

Y. Nagaya � T. IshidaDepartment of Civil and Earth Resources Engineering,

Graduate School of Engineering, Kyoto University, Kyoto, Japan

123

Rock Mech Rock Eng (2015) 48:1455–1461

DOI 10.1007/s00603-015-0727-9

Microseismic monitoring and analysis are generally

applied to examine fracture propagation by hydraulic

fracturing because directly observing induced fractures is

very difficult. In this study, hydraulically induced fractures

are observed by applying the fluorescent method proposed

by Nishiyama and Kusuda (1994). Fractures are detected

based on a marked difference in brightness under ultra-

violet light irradiation because they are filled with an

acrylic resin mixed with a fluorescent substance in ad-

vance. The advantages of this method are that microfrac-

tures can be quickly and accurately identified with an

optical microscope and it has been previously used to ex-

amine rock fracturing (Nishiyama et al. 2002; Chen et al.

2011).

Specifically, this study conducts hydraulic fracturing

experiments on the laboratory scale using granite speci-

mens under a true tri-axial loading condition. The induced

fractures are directly observed via the fluorescent method.

To clarify the differences in fracture propagation patterns,

the experiments use three fluids with different viscosities:

supercritical CO2, water, and viscous oil. In addition, the

features of the induced fractures are quantified by a digital

image analysis technique.

2 Experimental

2.1 Specimens

The specimens were Kurokami-jima granite, which is

coarse-grained biotite granite from Yamaguchi Prefecture,

Japan. It has been reported that Kurokami-jima granite

contains three splitting planes that are mostly perpendicular

to each other. Cubic specimens with each surface parallel

to the three splitting planes and side lengths of ap-

proximately 170 mm were prepared. The specimen had a

20-mm diameter central injection hole, and the inherent rift

plane of the granite specimen was oriented so that it cor-

responds to the YZ-plane in the Cartesian coordinate sys-

tem (Fig. 1). A total six specimens were fractured by the

hydraulic fracturing method. Table 1 summarizes the

P-wave velocities for each tested specimen measured along

the X-, Y- and Z-directions. The tested specimens corre-

spond to the names reported by Ishida et al. (2013) and Inui

et al. (2014).

2.2 Experimental Conditions

For each experiment, the specimen was set in a cylindrical

pressure cell. Four bow-shaped spacer blocks were inserted

between the specimen and the inner wall of the cylindrical

pressure cell, and four flat jacks were set between the

specimen and each spacer block to apply pressure in the X-

and Y-directions. In the Z-direction, two flat jacks were set

between the specimen and the two loading plates supported

by the end caps on the top and the bottom of the pressure

cell. In all experiments, confining pressures of 3, 6, and

4 MPa were applied in three directions (X, Y, and Z). In this

way, the fractures induced by hydraulic fracturing should

develop along the YZ-plane (i.e., the rift plane).

2.3 Fluid Injection

To inject the fracturing fluid, a packer with a 60-mm

pressurizing section was inserted and centered into the

central injection hole of the specimen. In each experiment,

the injection was terminated once hydraulic fracturing was

induced, which was indicated by a sudden pressure drop of

the injecting fluid. Three fluids with different viscosities

were used for the hydraulic fracturing experiment: CO2,

water, and viscous oil. CO2 reaches the supercritical state

when the temperature exceeds 31.1 �C at a pressure above

Fig. 1 Coordinate system, loading condition of the specimen, and

positions of the piezo elements for AE monitoring

Table 1 Anisotropy of the P-wave velocity of the tested granite

specimens

Specimen

name

Fracturing

fluid

P-wave velocity (km/s)

X-direction Y-direction Z-direction

G1102 Water 4.0 4.9 4.7

G1103 Water 4.3 5.0 4.7

G1112 Oil 4.3 4.9 4.6

G1114 Oil 4.0 4.7 4.6

G1115 CO2 4.3 4.9 4.7

G1202 CO2 5.2 5.7 5.7

1456 Y. Chen et al.

123

7.38 MPa. Supercritical CO2 has a very low viscosity,

which is around 5 % of the viscosity of water. CO2 was

discharged from the cylinder of the syringe pump at a

constant flow rate of 10 cm3/min. The cell containing the

specimen was filled with hot water (*45 �C) to prevent

the injected CO2 from cooling. Water and oil were injected

directly from the syringe pump just below room tem-

perature at a flow rate of 10 cm3/min in the same manner as

the supercritical CO2 injection, but the cell was not filled

with hot water. The oil was Mahha Super Transmission

Oil, 75W-90, by Fuji Kosan Co. Ltd., and its estimated

viscosity under the tested condition is approximately

300 mPa�s.To monitor the fracturing induced by the fluid injection,

acoustic emission (AE) events were measured using the 16

cylindrical PZT elements glued on the four side surfaces of

the specimen with four sensors on each surface (Fig. 1).

The detailed monitoring system and its conditions are de-

scribed in Ishida et al. (2012).

2.4 Microfracture Detection

The fluorescent method proposed by Nishiyama and

Kusuda (1994) was applied to identify fractures within the

tested specimens because it quickly and accurately identi-

fies microfractures with an optical microscope and it has

been applied to examine rock fracturing (Nishiyama et al.

2002; Chen et al. 2011). After filling the fractures with an

acrylic resin mixed with a fluorescent substance, the frac-

tures were detected based on a marked difference in

brightness under ultraviolet light irradiation. The detection

limit of this method is reportedly less than a few mi-

crometers (Nishiyama and Kusuda 1994).

This method can efficiently detect microfractures within

granite (Chen et al. 2001; Chen 2008). When observing

thin sections of granite, the brightest parts correspond to

fractures or pores. Under UV light, quartz and feldspar

grains are blue while other colored minerals, including

biotite, are black.

3 Observations of the Induced Fractures

All specimens were fractured by fluid injection, and the

fractures were identified on the surfaces of the cubic spe-

cimens. Table 2 summarizes the breakdown pressures and

the estimated viscosities of the fluids under the test con-

ditions. The viscosities of supercritical CO2, water, and oil

are around 0.05, 0.8, and 315 mPa�s, respectively. The

viscosity of oil is more than 6000 times that of supercritical

CO2.

To clarify the characteristics of fracture (crack)

propagation, the following procedure was used to prepare

two thin sections for each specimen. After the hydraulic

fracturing experiment, over-coring (68-mm diameter)

along the central injection hole was performed. Then the

hollow cores were prepared for observations by the

fluorescent method; that is, an acrylic resin mixed with a

fluorescent substance was penetrated and fixed into the

cores in advance. Two sections were made along the

plane with Z = 85 mm, which corresponds to the center

of the pressuring section following the two hydraulically

induced fractures (Fig. 2). In this way, secondary fractures

due to preparing the thin sections were not detected be-

cause the pre-treatment was completed before the sections

were cut.

3.1 Fracture Propagating Patterns

The induced fractures are considerably tortuous from a

macroscopic view (Fig. 3). At least in significant parts, the

fracture does not propagate along the flat plane on this

scale. Among the fracturing fluids, a higher viscosity re-

sults in a smoother fracture propagation pattern. On the

other hand, injecting a lower viscosity fracturing fluid

causes intermittent and stepwise fracture propagation.

More detailed observations show that the main fracture

path has many branches or derivative paths. Among these,

the fracture made by supercritical CO2, which has the

lowest viscosity, has the most branches.

Table 2 Summary of the hydraulic fracturing experiments

Specimen

name

Fracturing

fluid

Viscosity

(mPa�s)Breakdown pressure

(MPa)

G1102 Water 0.77 13.01

G1103 Water 0.79 13.37

G1112 Oil 316.8 25.05

G1114 Oil 314.9 23.08

G1115 CO2 0.05 9.10

G1202 CO2 0.05 10.16

Y

X

20mm68mm injection hole

reference area

induced fracture

Fig. 2 Reference area on a thin section

Hydraulic Fracturing in Anisotropic Granite 1457

123

3.2 Correlation Between Fractures and Mineral Grains

To discuss the state of fracture propagation, the correla-

tions between fracture pathways and constituent mineral

grains of specimens were observed. The induced fractures

for specimens fractured by supercritical CO2 are mainly

along the grain boundaries of the constituent minerals

(Fig. 4a). On the other hand, those in the specimens frac-

tured by water and oil often cut through the mineral grains

(Fig. 4b). It is inferred that the fractures induced by water

and oil with higher viscosities are controlled by the hidden

pre-existing microfractures that are aligned along the

fracture propagation orientation since the fractures extend

in the direction of the rift plane. Therefore, it is likely that

different fracture propagation pathways cause the varia-

tions in the tortuosity of the fracture patterns.

4 Discussion

4.1 Fracture Propagating Pathways

4.1.1 Quantification of the Complexity of the Fracture

Propagation Patterns

In hydraulic fracturing operations in the petroleum indus-

try, fractures with larger surfaces are desirable for oil and

gas production. Although the available techniques to make

desired fractures are limited, changing the viscosity of the

fracturing fluid is possible even in actual operations. From

this viewpoint, the fractures induced by fracturing fluids

with different viscosities were evaluated using two pa-

rameters and digital image analysis techniques: tortuosity

and the number of fractures.

The tortuosity is defined as the total fracture length

along a pathway divided by the direct length of the two

ends in the reference area. To measure the total fracture

length, the main pathway of the induced fractures was

deconstructed into 50-lm-long straight-line segments, and

then the sum was obtained (Fig. 5). The tortuosity in-

creases as the viscosity of the fracturing fluid decreases.

This tendency agrees well with the result stated in

Sect. 3.1, suggesting that a lower viscosity fluid can create

fractures with larger surfaces.

Fig. 3 Fracture propagation patterns of the induced fractures ob-

served on thin sections by applying a fluorescent method under

ultraviolet light. Bright and bluish-white parts correspond to fractures.

Fracture propagation direction is horizontal. Specimens fractured by

a supercritical CO2 (G1115 ?Y), b water (G1102 ?Y), and c oil

(G1112 ?Y) (color figure online)

Fig. 4 Observations of the induced fractures and mineral grains. Two

photographs captured under ultraviolet light and crossed nicols are

combined into one figure. Blue corresponds to fractures. Fracture

propagation direction is horizontal. a Fractures develop around grain

boundaries for the specimen fractured by supercritical CO2 injection

(G1115 ?Y) and b fractures cut across the grains for the specimen

fractured by oil injection (G1112 ?Y) (color figure online)

Fig. 5 Correlation between tortuosity and viscosity of the fracturing

fluid

1458 Y. Chen et al.

123

To count the number of branches, the scanning lines

were set to 500-lm intervals along the Y-direction, which is

the fracture propagation direction. Fractures induced by

supercritical CO2 have more branches than those by water

and oil (Fig. 6). Additionally, hydraulic fracturing can in-

duce many derivative pathways around the main fracture.

4.1.2 Differences in the Fracture Propagation Pattern

by Viscosity

Both the microscopic observations and the quantification of

the characteristics by image analysis show that the vis-

cosity of the fracturing fluid affects the fracture propaga-

tion pattern. The fracture propagation pathways of

supercritical CO2 fracturing are inferred to be along the

mineral grain boundaries. Around the fracture tip, super-

critical CO2 can penetrate the hidden weak planes such as

pre-existing defects and grain boundaries more easily than

a higher viscosity fluid. Consequently, fractures propagate

into the randomly oriented grain boundary, resulting in a

high tortuosity.

Many researchers have noted the fractality of the dis-

continuities in rocks, such as fractures and faults (e.g.,

Hirata et al. 1987; Sahimi et al. 1993). Because tortuosity

appears to be an essential parameter to estimate the volume

of the fracture surfaces, the present study may provide

fundamental data to determine it. However, whether the

tortuosity is significantly influenced by the size of the

constituent mineral grains must be evaluated before ap-

plying these results to larger scale estimations.

4.2 Fracture Formation by Shear

It seems reasonable to hypothesize that the major

mechanism to induce hydraulically fracturing in rocks is

tensile. Many parts of the fractures cut across the mineral

grains, which run almost parallel to the direction of fracture

propagation. The lack of relative displacement on both

sides of the fracture and a clear aperture between them

strongly suggests that these fractures are induced by the

tensile fracture mode (Fig. 7). However, this pathway is

not the case for fractures inclined toward the fracture

0

2

4

6

8

0 5 10 15 20

N(r)

r

0

2

4

6

8

0 5 10 15

N(r)

r

0

2

4

6

8

0 5 10 15 20

N(r)

r

0

2

4

6

8

0 5 10 15

N(r)

r

0

2

4

6

8

0 5 10 15 20

N(r)

r

0

2

4

6

8

0 5 10 15 20

N(r)

r

(a) (b)

(c) (d)

(e) (f)

Fig. 6 Number of fracture

pathways [N(r)] counted in

500-lm intervals along the Y-

direction (distance r). Fractures

observed in the specimen

fractured by supercritical CO2

injection [G1115 ?Y direction

(a) and -Y direction (b)], waterinjection [G1102 ?Y direction

(c) and -Y direction (d)], andoil injection [G1112

?Y direction (e) and-Y direction (f)]

Hydraulic Fracturing in Anisotropic Granite 1459

123

propagation direction, which passes along the grain

boundaries and the inside grains. Some fractures are likely

formed by the shear sense (Fig. 8) because shear stress

develops on a plane inclined to the principal stress direc-

tions, which are those of rx and ry in this experiment. It is

inferred that the fracture propagation pathways by hy-

draulic fracturing are induced by both shear and tensile

stresses. In a series that includes this study (Inui et al.

2014), the fracturing process has been analyzed by AE

monitoring. The shear fractures are predominant in speci-

mens fractured by supercritical CO2 and water, which have

relatively low viscosities. This analytical result agrees well

with the microscopic observations in this study.

On the other hand, a model for volcanic earthquake

swarms, which was proposed by Hill (1977), may help to

elucidate the mechanism of such tensile and shear frac-

tures. This model suggests that magma intrudes into one of

the weak planes that lie along the direction of the max-

imum compressive stress amongst the many weak planes in

a volcanic region. The magma intrusion forms a dike and is

accompanied by some tensile fracturing, while shear frac-

turing forms conjugate faults that connect the tips of the

dikes. The granite specimens used in the present ex-

periments should also have many small inherent defects

along the direction of maximum compressive stress asso-

ciated with its rift plane. Thus, the tensile fracturing ob-

served during the viscous oil injection should correspond to

an intrusion into such an inherent defect and the mineral

Fig. 7 Fracture cutting across the mineral grain (yellow arrow)

observed in the specimen fractured by oil injection (G1112). This

fracture is most likely induced by the tensile fracture mode. a Crossednicols and b observed under ultraviolet light. Fracture propagation

direction is horizontal (color figure online)

Fig. 8 Fracture propagating along the grain boundaries (yellow

arrow) observed in the specimen fractured by supercritical CO2

injection (G1115). This fracture is most likely induced by the shear

fracture mode. a Crossed nicols and b observed under ultraviolet

light. Fracture propagation direction is horizontal (color figure online)

1460 Y. Chen et al.

123

grains in the direction of maximum compressive stress. On

the other hand, shear fractures, which are dominant during

supercritical CO2 injection, should correspond to the frac-

ture connecting tips of inherent defects.

5 Conclusion

The results of this study are as follows:

1. Fractures induced by hydraulic fracturing are consid-

erably tortuous from a microscopic view. Comparing

the fracture propagation patterns of three fracturing

fluids with different viscosities shows that a higher

viscosity yields a smoother fracture pattern. The

tortuosity measures 1.13 when the lowest viscosity

fluid (supercritical CO2) is injected, but is 1.05 when

the highest viscosity fluid (oil) is injected, demonstrat-

ing that the viscosity of the fracturing fluid affects the

fracture propagation pattern.

2. Hydraulic fracturing induces many derivative path-

ways around the main fracture. The fracture induced

by supercritical CO2 has the most branches along the

main fracture, indicating that hydraulic fracturing with

a lower viscosity fluid forms a more complex fracture

network in rocks. These results suggest that desirable

fractures with larger surfaces can be made to improve

oil and gas production.

3. Fracture propagation by hydraulic fracturing some-

times occurs by the shear fracture mode. AE monitor-

ing of the fracturing process shows that shear

fracturing is predominant for a low-viscosity super-

critical CO2 injection. The microscopic observations

presented in this paper agree with this AE monitoring

result.

Acknowledgments This work was financially supported by the

Japan Society for the Promotion of Science Grants-in-Aid for Sci-

entific Research (B) and (A), Grand Numbers 21360446 and

25249131, as well as by the ENEOS Hydrogen Trust Fund of 2012.

References

Brown DW (2000) A hot dry rock geothermal energy concept

utilizing supercritical CO2 instead of water. Proceedings of 25th

Workshop on Geothermal Reservoir Engineering, SGP-TR-165,

Stanford, California, USA

Chen Y (2008) Observation of microcracks patterns in Westerly

granite specimens stressed immediately before failure by

uniaxial compressive loading. Chin J Rock Mech Eng

27:2440–2448

Chen Y, Nishiyama T, Ito T (2001) Application of image analysis to

observe microstructure in sandstone and granite. Resour Geol

51:249–258

Chen Y, Watanabe K, Kusuda H, Kusaka E, Mabuchi M (2011) Crack

growth in Westerly granite during a cyclic loading test. Eng Geol

117:189–197

Hill DP (1977) A model for earthquake swarms. J Geophys Res

82:1347–1352

Hirata T, Satoh T, Ito K (1987) Fractal structure of spatial distribution

of microfracturing in rock. Geophys J R Astr Soc 90:369–374

Inui S, Ishida T, Nagaya Y, Nara Y, Chen Y, Chen Q (2014) AE

monitoring of hydraulic fracturing experiments in granite blocks

using supercritical CO2, water and viscous oil. Proceedings of

ARMA 2014, Paper No. ARMA 14-7163, Minneapolis, MN,

USA

Ishida T, Chen Q, Mizuta Y, Roegiers JC (2004) Influence of fluid

viscosity on the hydraulic fracturing mechanism. J Energy

Resour Technol 126:190–200. doi:10.1115/1.1791651

Ishida T, Aoyagi K, Niwa T, Chen Y, Murata S, Chen Q, Nakayama

Y (2012) Acoustic emission monitoring of hydraulic fracturing

laboratory experiment with supercritical and liquid CO2. Geo-

phys Res Lett 39:L16309. doi:10.1029/2012GL052788

Ishida T, Nagaya Y, Inui S, Aoyagi K, Nara Y, Chen Y, Chen Q,

Nakayama Y (2013) AE monitoring of hydraulic fracturing

laboratory experiments conducted using CO2 and water. Proc

Eurock 2013:957–962

Kalantari-Dahaghi A (2010) Numerical simulation and modeling of

enhanced gas recovery and CO2 sequestration in shale gas

reservoirs: a feasibility study. Paper presented at International

Conference on CO2 Capture, Storage, and Utilization, Soc Pet

Eng, New Orleans, LA, USA

Liao S, Brunner F, Mattar L (2009) Impact of ignoring CO2 injection

volumes on post-frac PTA. Proceedings of Canadian Interna-

tional Petroleum Conference, Document ID 2009-124, Petro-

leum Society of Canada, Calgary, Canada

Nishiyama T, Kusuda H (1994) Identification of pore spaces and

micro cracks using fluorescent resins. Int J Rock Mech Min Sci

Geomech Abstr 31:369–375

Nishiyama T, Chen Y, Kusuda H, Ito T, Kaneko K, Kita H, Sato T

(2002) The examination of fracturing process subjected to

triaxial compression test in Inada granite. Eng Geol 66:257–269

Sahimi M, Robertson MC, Sammis CG (1993) Fractal distribution of

earthquake hypocenters and its relation to fault patterns and

percolation. Phys Rev Lett 70(14):2186–2189

Sinal ML, Lancaster G (1987) Liquid CO2 fracturing: advantages and

limitations. J Can Pet Technol 26:26–30

Spetzler J, Xue Z, Saito H, Nishizawa O (2008) Case story: time-lapse

seismic crosswell monitoring of CO2 injected in an onshore

sandstone aquifer. Geophys J Int 172:214–225

Xue Z, Tanase D, Watanabe J (2006) Estimation of CO2 saturation

from time-lapse CO2 well logging in an onshore aquifer,

Nagaoka, Japan. Explor Geophys 37:19–29

Hydraulic Fracturing in Anisotropic Granite 1461

123