You are on page 1of 16

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259911087

Plane-Strain Propagation of a Fluid-Driven Crack in a Permeable Rock with


Fracture Toughness

Article in Journal of Engineering Mechanics · September 2010


DOI: 10.1061/(ASCE)EM.1943-7889.0000169

CITATIONS READS
98 866

2 authors, including:

Dmitry Garagash
Dalhousie University
79 PUBLICATIONS 2,278 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Near tip hydraulic fracture processes View project

Physics of Induced Seismicity View project

All content following this page was uploaded by Dmitry Garagash on 28 January 2014.

The user has requested enhancement of the downloaded file.


Plane-Strain Propagation of a Fluid-Driven Crack in a
Permeable Rock with Fracture Toughness
J. Hu, M.ASCE1; and D. I. Garagash, M.ASCE2

Abstract: A solution to the problem of a plane-strain fluid-driven crack propagation in elastic permeable rock with resistance to fracture
is presented. The fracture is driven by injection of an incompressible Newtonian fluid at a constant rate. The solution, restricted to the case
of zero lag between the fluid front and the fracture tip, evolves from the early-time regime when the fluid flow takes place mostly inside
the crack toward the large-time response when most of the injected fluid is leaking from the crack into the surrounding rock. This
transition further depends on a time-invariant partitioning between the energy expanded to overcome the rock fracture toughness and the
energy dissipated in the viscous fluid flow in the fracture. A numerical approach is used to compute the solution for the normalized crack
length and crack opening and net-fluid pressure profiles as a function of two dimensionless parameters: the leak-off/storage evolution
parameter and the toughness/viscosity number. Relation of this solution to the various available asymptotic solutions is discussed.
Obtained mapping of the solution onto the problem parametric space has a potential to simplify the tasks of design, modeling, and data
inversion for hydraulic fracturing treatments and laboratory experiments.
DOI: 10.1061/共ASCE兲EM.1943-7889.0000169
CE Database subject headings: Cracking; Porous media; Leakage; Numerical analysis; Rocks.
Author keywords: Fracture mechanics; Hydraulic fracture; Porous media; Leak-off; Asymptotic solution; Numerical analysis.

Introduction tion at the temporal and spatial scales of interest and 共ii兲 bench-
marking against mathematically rigorous reference solutions.
Hydraulic fracturing 共HF兲, nearly a century-old technology, is Modeling of hydraulic fractures of simple planar or radial 共penny-
widely used to stimulate oil and natural gas production 共Econo- shaped兲 geometry, which would result from a line or a point fluid
mides and Nolte 2000兲. In this technique, a fracture created by source under homogeneous in situ conditions, respectively, helps
injection of pressurized fluid, when completed and propped, pro- to address both of the above points of interest in numerical mod-
vides a highly conductive pathway for hydrocarbons toward a eling. Recent investigations 共Detournay 2004; Detournay and Ga-
producing well. The conditions under which HF treatments are ragash, “General scaling laws for fluid-driven fractures,” in
carried out in situ 共e.g., reservoir rock properties, in situ stresses, preparation; Garagash and Detournay 2005; Garagash 2006a,b,c兲
reservoir and fracturing fluids, proppants, etc.兲 vary widely. There helped to establish consistent scaling framework and parametric
is also no existing reliable method to accurately measure fracture space for hydraulic fractures of simple geometry and characterize
geometry during and after the process. Hence, modeling and nu- limiting fracture propagation regimes. A number of corresponding
merical simulation of the HF process has special importance for rigorous semianalytical solutions have been also developed 共e.g.,
the design and control of this technique. Spence and Sharp 1985; Lister 1990; Savitski and Detournay
Adachi et al. 共2007兲 gave an overview and current state of the 2002; Adachi and Detournay 2002; Garagash and Detournay
art of computational models for hydraulic fracture simulation in 2005; Garagash 2006a,b,c; Bunger et al. 2005; Mitchell et al.
the petroleum industry. The emphasis of the industrial numerical 2007兲. These asymptotic solutions generally correspond to frac-
models is on the ability to simulate fracture propagation under ture propagation under limiting condition共s兲 when either one of
realistic nonuniform in situ conditions often resulting in fractures the energy dissipation mechanisms associated with fluid viscosity
of complicated geometry, driven by fracturing fluids with compli- and rock toughness, respectively, can be neglected and/or one of
cated rheology, etc. As such, construction of these numerical the two fluid storage mechanisms, associated with fluid storage in
models calls for 共i兲 appropriate understanding and implementation the fracture and fluid storage in the surrounding rock 共leak-off兲,
of physical processes which govern hydraulic fracture propaga- dominates. From this perspective, emphasis of this work is on
construction of a general, yet mathematically rigorous, solution
1 which addresses evolution of a hydraulic fracture of a simple
Project Engineer, AMEC Geomatrix, Inc., 2101 Webster St., Oak-
land, CA 94612. 共planar兲 geometry between the appropriate asymptotic regimes
2
Associate Professor, Dept. of Civil and Resource Engineering, discussed above.
Dalhousie Univ., 1360 Barrington St., Halifax NS, Canada B3J 1Z1 共cor- Two alternative numerical methods with a moving mesh have
responding author兲. E-mail: garagash@dal.ca been suggested recently to track the evolution in time and space
Note. This manuscript was submitted on January 29, 2009; approved
of a hydraulic fracture of simple geometry. An explicit finite dif-
on March 18, 2010; published online on May 6, 2010. Discussion period
open until February 1, 2011; separate discussions must be submitted for ference scheme with a moving mesh 共Adachi and Detournay
individual papers. This paper is part of the Journal of Engineering Me- 2008兲 has been used to solve the problem of hydraulic fracture
chanics, Vol. 136, No. 9, September 1, 2010. ©ASCE, ISSN 0733-9399/ propagation in permeable rock of zero resistance to fracture
2010/9-1152–1166/$25.00. 共toughness兲 when the solution evolves from the storage-

1152 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
the leakage time via a lumped leak-off coefficient Cᐉ. Under con-
ditions when the fluid lag 共defined as the distance between the
fracture tip and the fluid front兲 is negligible, the solution of this
x
pf (x, t) w(x, t) problem corresponds to finding the fracture aperture w, the net-
fluid pressure p 共the fluid pressure p f minus the far-field confining
(t)
stress ␴0兲, and the crack half-length ᐉ as a function of the position
x along the crack 共w and p only兲, time t, and five problem param-
eters. These parameters are the injection rate Qo and four material
Fig. 1. Sketch of a plane-strain fluid-driven fracture parameters

dominated to the leak-off dominated asymptotic limit along the


E⬘ =
E
1 − ␯2
; K⬘ = 4 冉冊2

1/2
KIc ; ␮⬘ = 12␮; C⬘ = 2Cl 共1兲

“zero-toughness trajectory” in the parametric space. This method with the meaning of 共plane-strain兲 rock modulus, rock fracture
has been recently extended to obtain a general numerical solution toughness, fluid viscosity, and leak-off coefficient, respectively.
under a zero fluid lag assumption for a penny-shaped fracture 关Numerical factors were introduced in Eq. 共1兲 to simplify the
driven by a Newtonian fluid 共Madyarova 2003兲. Another numeri- form of the governing equation as will be discussed later.兴
cal method, based on the method-of-lines numerical approach, We acknowledge that the presence of a fluid lag 共a nonwetted
has been used to solve the problem of hydraulic fracture propa- zone at the fracture tip兲 is a physical requirement of the problem
gation driven by a power law non-Newtonian fluid in imperme- in order to avoid infinite suction which develops at the wetted
able rock 共Garagash 2006c兲. 共In this case, the solution evolves fracture tip as the result of the coupling between the nonlocal
from the toughness-dominated to the viscosity-dominated regime elastic response of the crack and the viscous fluid flow inside the
along the “zero-leak-off trajectory.”兲 The method-of-lines ap- crack 关e.g., Rubin 共1993兲兴. However, solutions obtained ignoring
proach uses finite discretization of the solution in space 共over a the fluid lag can be considered as intermediate asymptotic solu-
moving mesh兲 while the time evolution of the solution at the tions, i.e., solutions that capture the dominant behavior down to a
space discretization nodes is reduced to a system of ordinary dif- certain small distance from the tip and that should match with a
ferential equations, which is solved continually. We use the latter boundary layer solution encompassing the lag 共Garagash 2006b兲.
approach, where the temporal evolution is handled by a standard We herein restrict our investigations to the latter case when the
ordinary differential equation 共ODE兲 solver, to obtain a complete fluid lag is negligible, which requires either large enough confin-
numerical solution for the outlined model of a plane-strain hy- ing stress and/or slow enough fracture propagation 共Garagash
draulic fracture propagation. 2006b; Garagash and Detournay 2000兲.
We start with the mathematical formulation of the problem Using the problem symmetry about the crack inlet, the gov-
followed by the review of the scaling considerations, which de- erning equations and boundary conditions for this problem are
fine the problem: nondimensional parametric space and relevant presented in the following over a half of the crack extent 共0 ⬍ x
asymptotic regimes of the fracture propagation. The correspond- ⬍ ᐉ兲.
ing asymptotic solutions are then discussed in some detail to out- • Local fluid continuity equation
line the changing nature of the solution in the problem parametric
space and to guide the development of the general numerical ⳵w ⳵q
+g+ =0 共2兲
solution. Following the introduction of the numerical method, the ⳵t ⳵x
results of the numerical solution are presented in dimensionless
describes the balance between the rate q共x , t兲 of the unidirec-
form and contrasted to the various previously established
tional fluid flow in the deformable crack and the local rate
asymptotic limiting solutions. The ranges of problem parameters
g共x , t兲 of the fluid leak-off. The former is given by Poiseuille’s
where solution is approximated by appropriate asymptotic limit-
law 共Batchelor 1967兲
ing solutions are established. The paper concludes with the dis-
cussion of the temporal dependence of the solution, as relevant to w3 ⳵ p
q=− 共3兲
the crack length inversion from the crack inlet value of the net- ␮⬘ ⳵ x
fluid pressure, and the summary of the main results.
and the latter is approximated by the law of Carter 共1957兲
C⬘
g共x,t兲 = t ⬎ t0共x兲 共4兲
Mathematical Formulation 冑t − t0共x兲 ,
Mathematical models of a hydraulic fracture propagating in a Here t0共x兲 is the arrival time of the fracture front at the posi-
permeable rock have to account for the primary physical mecha- tion x, therefore, given in implicit form by x = ᐉ关t0共x兲兴. Phe-
nisms involved, namely, deformation of the rock, fracturing or nomenological law 共4兲 was originally suggested by Carter for
creation of new surfaces in the rock, flow of viscous fluid in the fracturing fluids, which, at initial stages of the leak-off pro-
fracture, and leak-off of the fracturing fluid into the permeable cess, create a semi-impermeable “cake” 共a layer of plugged
rock. In this framework, we consider the case of a plane-strain rock pore space兲 encasing the leaking fracture walls. Recently,
fracture propagation in a homogeneous linear elastic rock charac- this law has been justified for a regular diffusion process
terized Young’s modulus E, Poisson’s ratio ␯, and the rock tough- 共without cake building兲 at intermediate distances near the
ness KIc. The propagation is driven by injection of an crack tip under condition that the characteristic fluid diffusion
incompressible Newtonian fluid with the viscosity ␮ at the crack length scale is much smaller than the extent of the crack 共Ga-
inlet at a constant rate Qo 共Fig. 1兲. The specific rate of leak-off of ragash, “Plane-strain propagation of a fluid-driven fracture in a
the fracturing fluid into the permeable rock is related by the phe- permeable medium: Leak-off dominated regime,” Int. J. Solids
nomenological law of Carter 共1957兲 to the inverse square root of Struct., submitted; Y. Kovalishen and E. Detournay, personal

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010 / 1153

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
communication, 2008兲. In view of the conditions at the crack 
M Km
f leak-off edge Mk̃ 
K
tip, where the crack opening is zero and fluid velocity coin-
cides with the tip velocity, respectively,
Sm
f Sk̃

Km -trajectory
q dᐉ
x = ᐉ: w = 0; = 共5兲
w dt viscosity toughness
edge edge
an integral of continuity 共2兲 can be obtained in the following
form:
Cm Ck
冕 冕
ᐉ ᐉ

wdx + gdx = q 共6兲
⳵t x x M Km storage edge Mk K
• Global fluid continuity equation Fig. 2. Problem parametric space and a solution trajectory param-

冕 冕冕

etrized by, e.g., toughness number Km and evolution leak-off param-
t ᐉ共t⬘兲
eter Cm
Q ot = 2 w共x,t兲dx + 2 g共x,t⬘兲dxdt⬘ 共7兲
0 0 0

is the result of evaluating continuity integral 共6兲, the time in- totics, t0共x兲 = t − 关ᐉ共t兲 − x兴 / 共dᐉ / dt兲, when x → ᐉ. Using the above to
tegration, and the use of the flux boundary condition, q共x evaluate the left-hand side 共LHS兲 of Eq. 共6兲 and substituting Eqs.
= 0+ , t兲 = Qo / 2. Eq. 共7兲 simply states that the total volume of 共3兲 and 共12兲 result in the asymptotic expression for the net-
fluid injected over time t is equal to the sum of the volume of pressure gradient. Integrating the latter gives
fluid stored in the crack and the volume of fluid leaked into the
porous solid. Substituting Eq. 共4兲 into Eq. 共7兲, reversing the
order of integration of the double integral, and integrating by
p=
␮⬘E⬘2 dᐉ
K⬘2 dt
冋+ 2C⬘
E⬘ dᐉ
K⬘ dt
冉 冊册冉 冊
1/2
ln 1 −
x

+ O共1兲, ᐉ−xⰆᐉ

parts allow a simpler form of the latter 共13兲

冕 冕冑
ᐉ t
ᐉ共t⬘兲 Consequently, net-fluid pressure p has a negative logarithmic sin-
Q ot = 2 w共x,t兲dx + 2C⬘ dt⬘ 共8兲 gularity at the fracture tip.
0 0 t − t⬘
• Elasticity equation 共Bilby and Eshelby 1968兲



Scaling
E⬘ ⳵ w ds
p共x,t兲 = − 共9兲
−ᐉ ⳵ s s − x
4␲ A scaling for the problem under study can be formulated in the
view of the following two considerations: on the one hand, the
relates the net pressure p = p f − ␴0 to the crack opening w. The scaling should reflect the physical processes taking place during
inverse form of this equation is given by 共Sneddon and Lo- fracture propagation; on the other hand, whenever possible it
wengrub 1969兲 should reduce the problem variables to dimensionless quantities

冕 冉 冊
ᐉ of order O共1兲. Propagation of a hydraulic fracture is governed by
4 x x⬘ two competing dissipative processes associated with the fluid vis-
w共x,t兲 = G , p共x⬘,t兲dx⬘ 共10兲
␲E⬘ 0 ᐉ ᐉ cosity and solid toughness and two competing components of the
fluid balance associated with fluid storage in the fracture and fluid
where the kernel G is defined as storage in the surrounding rock 共leak-off兲. Following Detournay

G共␰,␰⬘兲 = ln 冏 冑冑 1 − ␰2 + 冑1 − ␰⬘2
1 − ␰2 − 冑1 − ␰⬘2
冏 共11兲
and Garagash, “General scaling laws for fluid-driven fractures,”
in preparation, limiting fracture propagation regimes based on the
above basic mechanisms can be conveniently conceptualized in a
• Fracture propagation condition 共Irvin 1957兲 rectangular phase diagram MM̃K̃K 共Fig. 2兲.
The vertices of the diagram correspond to the four primary
K⬘ asymptotic solutions when only one of the two fluid storage
w= 共ᐉ − x兲1/2 + O关共ᐉ − x兲3/2兴, ᐉ−xⰆᐉ 共12兲
E⬘ mechanisms and one of the two dissipation processes are present:
prescribes the opening near the crack tip in the framework of storage-viscosity M solution 共C⬘ = K⬘ = 0兲; storage-toughness K
linear elastic fracture mechanics 共LEFM兲 under the assump- solution 共C⬘ = ␮⬘ = 0兲; leak-off-viscosity M̃ solution 共C⬘ → ⬁ and
tion of mobile equilibrium 共stress intensity factor of a propa- K⬘ = 0兲; and leak-off-toughness K̃ solution 共C⬘ → ⬁ and ␮⬘ = 0兲.
gating crack is equal to the material toughness兲. These four asymptotic solutions provide approximations to the
The corresponding asymptotic behavior of the net-pressure corresponding primary fracture propagation regimes dominated
gradient for a nonzero toughness and/or leak-off can be deduced by one of the fluid storages and one of the energy dissipation
from the analysis of the leading term in lubrication 共6兲 with mechanisms. For example, in the storage-viscosity-dominated re-
Eqs. 共3兲 and 共12兲 as follows. The tip asymptotics of the first gime 共M regime兲, the fluid leak-off is small compared to the fluid
integral term in Eq. 共6兲 can be evaluated as ⳵ / ⳵t兰ᐉ共t兲 x w共x , t兲dx fracture storage and the energy expended in fracturing of the rock
= 共dᐉ / dt兲w共x , t兲 when x → ᐉ for arbitrary monomial form of the is small compared to the viscous dissipation. As the result, solu-
crack opening tip asymptotics, i.e., w共x , t兲 = A关ᐉ共t兲 − x兴␣ with arbi- tion in this regime can be approximated by the asymptotic M
trary constants A and ␣ of which asymptotics 共12兲 is a particular solution.
example. The asymptotic expression for the other 共leak-off兲 inte- The edges of the phase diagram correspond to the four second-
gral in Eq. 共6兲 is easily evaluated using the arrival time asymp- ary asymptotic solutions when either one fluid storage mechanism

1154 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
Table 1. Four Different Scalings of the Problem: Scaling Prefactors ␧ and L and Corresponding Dimensionless Governing Parameters
Scaling ␧ L Energy partition parameter Fluid storage partition parameter
M
冉 冊
␮⬘
E⬘t
1/3
冉 冊
E⬘Q3o 1/6 2/3
␮⬘
t Km = 冉 冊
K ⬘ E⬘
E ⬘ ␮⬘Q o
1/4
Cm = 冉冊t
t

*
1/6
, tⴱ =
␮⬘Q3o
E ⬘C⬘6

冉 冊 冉 冊 冉 冊
−4
K K⬘4 1/3
E⬘ Q o 2/3 Mk = K m t 1/6
−2/3
t2/3 Ck = = Km Cm
E⬘4Qot K⬘ 4
Km tⴱ

冉 冊
C⬘2 ␮⬘
E⬘ Q o t
1/4
Qo 1/2
C⬘
t
Km̃ = Km
Sm̃ = 冉 冊 tmm̃
t −1/4
−3/2
= Cm


冉 冊
K⬘4C⬘2
E⬘ 4
Q2ot
1/4
Qo 1/2
C⬘
t
M˜k = M k
S˜k = 冉 冊4
Km tⴱ
t −1/4
= Ck−3/2

or one dissipation process is vanishing: storage MK solution • Global fluid balance

冕 冕
共C⬘ = 0兲; leak-off M̃K̃ solution 共C⬘ → ⬁兲; viscosity MM̃ solution 1 1
L共␪t兲 ␥共␪t兲
共K⬘ = 0兲; and toughness KK̃ solution 共␮⬘ = 0兲. These solutions pro- 1 = 2S␥2 ⍀̄d␰ + 2C d␪ 共18兲
0 0 L共t兲 冑1 − ␪
vide approximations to the corresponding secondary propagation
regimes dominated by one of the fluid storages or one of the Here ␪共␰ , t兲 = t0共x兲 / t is the normalized arrival time of the frac-
energy dissipation mechanisms. ture front at location x = ␰ᐉ共t兲. It can be defined for a given
Further scaling analysis allows us to characterize time/ scaling 共14兲 in the implicit form
parametric dependence of the solution in the MM̃K̃K phase dia-
gram and to find corresponding parametric groups which quantify ␥共␪t兲L共␪t兲 = ␰␥共t兲L共t兲 共19兲
the solution departure from or transition to the respective limiting • Elasticity equation
regimes as recounted below. In the general scaling framework,


1
normalized crack half-length ␥, crack opening ⍀, and net-fluid 4
pressure ⌸ can be introduced as follows 共Detournay 2004兲: ⍀̄共␰,t兲 = G共␰,␰⬘兲⌸共␰⬘,t兲d␰⬘ 共20兲
␲ 0
ᐉ共t兲 = L共t兲␥共t兲
• Propagation condition
w共x,t兲 = ␧共t兲L共t兲⍀共␰,t兲 ⍀̄ = K␥−1/2共1 − ␰兲1/2, 1−␰Ⰶ1 共21兲

p共x,t兲 = ␧共t兲E⬘⌸共␰,t兲 共14兲 A particular scaling 关i.e., a choice of the scaling factors ␧ and
L in Eq. 共14兲兴 can be defined by setting two of the four dimen-
where ␧ = small dimensionless parameter and L = fracture length sionless groups 关Eq. 共16兲兴 to unity while the remaining two di-
scale. The normalized solution is a function of the scaled crack mensionless groups provide two dimensionless parameters
coordinate ␰ = x / ᐉ 共0 ⱕ ␰ ⱕ 1兲 and time. We will also make use of governing the solution in the respective scaling. A choice of the
an alternative crack opening scaling in order to simplify the form two dimensionless groups that are set to unity can be physically
of resulting normalized governing equations motivated by a particular regime of the fracture propagation char-
⍀共␰,t兲 acterized by a prevailing fluid storage and energy dissipation
⍀̄共␰,t兲 = 共15兲 mechanisms. For example, it is to be expected that the solution to
␥共t兲 a fracture, which propagation is dominated by the fluid storage
Governing equations scaled according to Eqs. 共14兲 and 共15兲 within the crack and by the viscous dissipation, corresponds to the
are formulated below in terms of the following four dimension- O共1兲 values of the respective dimensionless groups S and M,
less groups: while the other two groups, C and K, which quantify that the
effects of the leak-off 共i.e., the storage in the porous rock兲 and the
␧L2 C ⬘L Q o␮ ⬘ K⬘ toughness dissipation, respectively, vanish.
S= , C= , M= , K= 共16兲
Q ot Qot1/2 ␧ 4L 2E ⬘ ␧E⬘L1/2 Four different physically motivated scalings are identified as
the storage-viscosity M scaling 共S = M = 1兲, storage-toughness K
with the meaning of dimensionless fluid storage in the fracture,
leak-off, viscosity, and toughness, respectively. scaling 共S = K = 1兲, leak-off-viscosity M̃ scaling 共C = M = 1兲, and
• Lubrication equation 关local fluid balance 共6兲 and Poiseuille’s leak-off-toughness K̃ scaling 共C = K = 1兲. Corresponding expres-
law 共3兲兴 sions for the scaling factors and the two governing dimensionless

S 冋冉 t␧˙

tL̇
+2 +2

t␥˙

冊冕 ␰
1
⍀̄d␰ + 冉 冊 冕 册
tL̇ t␥˙
L ␥
+ ␰⍀̄ +
1


˙
t⍀̄d␰
parameters for these four scaling schemes are summarized in
Table 1. Governing equations in each of the four scalings can be
readily obtained from Eqs. 共18兲–共21兲 using expressions from
Table 1. The resulting sets of normalized equations are given in

冕冑
1
C d␰ ⍀̄3 ⳵ ⌸ Appendix I.
+ =− 共17兲 The four scalings are naturally related to the corresponding
␥ ␰ 1−␪ M ⳵␰
vertices of the MM̃K̃K diagram, as the two governing dimension-
where the overdot stands for the partial derivative in time at a less parameters in a given scaling quantify the departure of the
fixed ␰, 共⳵ / ⳵t兲␰ = 共⳵ / ⳵t兲x + 共ᐉ̇ / ᐉ兲␰共⳵ / ⳵␰兲. solution from the corresponding vertex solution 共Fig. 2兲. For ex-

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010 / 1155

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
ample, normalized solution in the M scaling, which can be ex- Asymptotic Solutions
pressed as a function of scaled coordinate ␰ and two governing
dimensionless parameters Cm and Km In this section we review available solutions corresponding to the
vertices and edges of the parametric space. As discussed earlier,
Fm共␰,Cm,Km兲 = 兵⍀m共␰,Cm,Km兲,⌸m共␰,Cm,Km兲,␥m共Cm,Km兲其 these solutions constitute various asymptotic limits of the sought
general solution to a fluid-driven crack in a permeable solid, and
共22兲
as such are used to guide the development of the general solution
共where the lower-case index m identifies the scaling兲 naturally and to benchmark it. For example, the storage MK-edge and the
reduces to the self-similar M solution in the limit of zero leak-off leak-off M̃K̃-edge solutions correspond to “the beginning” and
Cm = 0 and zero toughness Km = 0 : Fm0共␰兲 = Fm共␰ , Cm = 0 , Km = 0兲. “the end” of the fracture evolution and, therefore, provide the
Similarly for other self-similar vertex solutions in their corre- initial value problem and the large-time benchmark to the sought
sponding scalings we have M̃ solution Fm̃0共␰兲 = Fm̃共␰ , Sm̃ = 0 , Km̃ solution, respectively. Similarly, the viscosity MM̃-edge solution
= 0兲, K solution Fk0共␰兲 = Fk共␰ , Ck = 0 , M k = 0兲, and K̃ solution and the toughness KK̃-edge solution can be used to check the
F˜k0共␰兲 = F˜k共␰ , S˜k = 0 , M˜k = 0兲. numerical algorithm at small and large values of toughness pa-
Evolution of the general solution in the phase diagram MM̃K̃K rameter, respectively, and to provide 共to a certain degree of accu-
is readily assessed from the time dependence of the governing racy兲 the small-toughness and large-toughness approximations of
dimensionless parameters 共Detournay and Garagash, “General the general solution, respectively. Thereby, numerical solution in-
scaling laws for fluid-driven fractures,” in preparation兲. The vis- side the parametric rectangle MM̃K̃K would have to be computed
cosity M’s and toughness K’s parameters are time-independent only in the finite domain of parameters 共excluding very small or
constants, while the leak-off C’s and storage S’s parameters are large values of governing parameters兲. Outside of this finite do-
increasing and decreasing power laws of time, respectively. Con- main, solution is to be given with a certain degree of accuracy by
sequently, general solution of hydraulic fracture starts out on the the edge solutions.
storage MK edge with its starting position specified by a given
value of, e.g., Km and evolves along the corresponding Km trajec-
tory toward the leak-off M̃K̃ edge 共Fig. 2兲. Either C or S in a Vertices
given scaling can be used to track evolution along a Km trajectory
since the time-derivative terms in lubrication 共17兲 can be ex- Vertex solutions are self-similar since both governing parameters
pressed according to in the appropriate scaling are zero, and evolution terms in local
balance 共17兲 are zero. Self-similarity implies that normalized ver-
⳵ C ⳵ S ⳵ tex solutions for the crack opening and net-fluid pressure are
t = =− 共23兲 functions of scaled coordinate only, and the normalized half-
⳵t 6 ⳵C 4 ⳵S
length of the crack is a time-independent constant. Dependence of
Further, owing to the time invariance of M and K, the partition of corresponding dimensional opening, pressure, and length 关Eq.
energy between the dissipation in the solid and the fluid, respec- 共14兲兴 on time and problem parameters is given by that of the
tively, is invariant. This behavior is particular to the plane-strain scaling factors ␧ and L 共Table 1兲. For example, the fracture half-
fracture driven by a Newtonian fluid since either different fracture length ᐉ is given by that of the corresponding length scale L
geometries, e.g., penny-shaped fracture 共Savitski and Detournay 共Table 1兲: ⬃t2/3 in the storage-dominated 共M and K兲 vertices and
2002兲, or different fracturing fluid rheologies 共e.g., Garagash ⬃t1/2 in the leak-off dominated 共M̃ and K̃兲 vertices.
2006c兲 generally result in time-dependent energy dissipation par- The M-solution Fm0共␰兲 = 兵⍀m0共␰兲 , ⌸m0共␰兲 , ␥m0其 is characterized
titioning and, consequently, solution trajectories in the corre- by zero-toughness Km = 0 and zero-leak-off Cm = 0 parameters and
sponding MM̃K̃K space evolving between pairs of vertices, i.e., is associated with the storage-viscosity-dominated regime of frac-
M → K̃ or K → M̃, respectively. ture propagation. The solution is given by Carbonell et al. 共1999兲
As the fracture propagation regime evolves with time, the di- and Adachi and Detournay 共2002兲 in the form of series expansion
mensional solution dependence on time and problem parameters of the opening ⍀m0共␰兲 and the net pressure ⌸m0共␰兲 over a class of
evolves from that given by an “early-time” scaling 共M or K scal- base functions with prescribed tip asymptotics. Since fracture
toughness and associated energy release at the crack tip are null,
ing兲 to that by a “large-time” scaling 共M̃ or K̃ scaling兲. The fol- the latter is no longer defined by the classical LEFM “square-
lowing relationships established from Eq. 共14兲 and Table 1 root” asymptote 关Eq. 共12兲兴 or in normalized form 关Eq. 共21兲兴. Dif-
ferent “storage-viscosity” tip asymptote emerges 共Spence and
⌸m ⍀m ␥m Sharp 1985; Lister 1990; Desroches et al. 1994兲
= Km4/3 ; = Km2/3 ; = Km−2/3
⌸k ⍀k ␥k

⌸m ⍀m ␥m
⍀m0 = ␤m 冉 冊
2 3

3 m0
1/3
共1 − ␰兲2/3 ; ⌸m0 = ␦m 冉冊2
3
1/3
共1 − ␰兲−1/3,
= Cm1/2 ; = Cm−1/2 ; = Cm−1
⌸m̃ ⍀m̃ ␥m̃ 1−␰Ⰶ1 共25兲

共with ␤m = 3.147, ␦m = −3.302兲 as a result of the coupling between


⌸m ⍀m ␥m the lubrication flow under storage-dominated conditions and the
= Cm1/2Km ; = Cm−1/2Km ; = Cm−1 共24兲
⌸˜k ⍀˜k ␥˜k elastic deformation of the crack faces. The corresponding singu-
larity at the crack tip is weaker than the classical LEFM, which is
will, therefore, be useful to convert the normalized field variables consistent with the zero-toughness constraint. The dimensionless
from one scaling to another. crack half-length value and the inlet values of the crack opening

1156 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
and the net pressure are ␥m0 = 0.6152, ⍀m0共0兲 = 1.1260, and 共Garagash and Detournay 2005兲. As the result, the solution out-
⌸m0共0兲 = 0.5450, respectively. side of the boundary layer is still closely approximated by the M
The M̃-vertex solution Fm̃0共␰兲 = 兵⍀m̃0共␰兲 , ⌸m̃0共␰兲 , ␥m̃0其 is char- solution for Km ⬍ 1.
acterized by zero-toughness Km̃ = 0 and zero-storage Sm̃ = 0 param- The M̃K̃ edge corresponds to the case when most of the in-
eters and is associated with the leak-off-viscosity-dominated jected fluid leaks into the rock, which according to the sense of
regime of fracture propagation. The solution 共Adachi and Detour- the time-dependent leak-off dimensionless parameters, corre-
nay 2008兲 is based on a similar expansion method 共as in the spond to the large-time asymptotic limit. The semianalytical
M-vertex solution兲, which incorporates the “leak-off-viscosity” M̃K̃-edge solution 共Garagash, “Plane-strain propagation of a
tip asymptote 共Lenoach 1995兲 fluid-driven fracture in a permeable medium: Leak-off dominated
regime,” Int. J. Solids Struct., submitted兲 depends only on the
⍀m̃0 = ␤m̃共2␥m̃0 兲 共1 − ␰兲5/8 ;
6 1/8
⌸m̃0 = ␦m̃共2␥m̃0 兲 共1 − ␰兲−3/8,
−2 1/8
dimensionless toughness Km̃ or the dimensionless viscosity M˜k.
1−␰Ⰶ1 共26兲 As in the case of the tip behavior along the MK edge discussed
above, the dominant tip behavior along the M̃K̃ edge switches
with ␤m̃ = 2.534, ␦m̃ = −0.164. This asymptote arises from the cou-
from the zero-toughness asymptote ⍀ ⬃ 共1 − ␰兲5/8, ⌸ ⬃ 共1 − ␰兲−3/8
pling between the lubrication flow under leak-off-dominated con-
to the nonzero toughness asymptote ⍀ ⬃ 共1 − ␰兲1/2, ⌸ ⬃ ln共1 − ␰兲
ditions and the elastic crack deformation in the vicinity of the
crack tip and automatically satisfies the zero-toughness constraint. when the solution departs away from the M̃ vertex. This transition
The dimensionless crack half-length value and the inlet values of in the tip behavior is accommodated by the tip boundary layer
the crack opening and the net pressure are ␥m̃0 = 1 / ␲, ⍀m̃0共0兲 with thickness ⬃Km̃8 ᐉ. The boundary layer solution possesses the
= 0.8165, and ⌸m̃0共0兲 = 0.7515, respectively. toughness asymptote ⍀ ⬃ 共1 − ␰兲1/2 in the immediate vicinity of
the tip and the viscosity asymptote ⍀ ⬃ 共1 − ␰兲5/8 at intermediate
The K- and K̃-vertex solutions are characterized by zero fluid
distances from the tip.
viscosity M k = M˜k = 0 and zero leak-off Ck = 0 or storage S˜k = 0,
respectively, and are associated with the toughness-dominated The MM̃ edge and the KK̃ edge correspond to the viscosity-
propagation regime under storage and leak-off dominated condi- dominated and the toughness-dominated hydraulic fracture evolu-
tions, respectively. Zero fluid viscosity implies that the gradient of tions from the storage MK edge to the leak-off M̃K̃ edge. The
pressure is null, and the solution corresponds to a special case of MM̃ edge corresponds to the fracture propagation along a preex-
a uniformly pressurized Griffith’s crack. The K- and K̃-vertex isting discontinuity 共K⬘ = 0兲; i.e., there is only viscous dissipation.
solutions for the opening and pressure can be expressed in terms The numerical solution 共Adachi and Detournay 2008兲 evolving
of dimensionless fracture length in appropriate scaling 共Garagash from the storage-dominated M vertex to the leak-off dominated M̃
2006a; Bunger et al. 2005兲 vertex has been obtained using an explicit numerical algorithm
1/2 that combines a conventional finite difference scheme for the fluid
⍀k0,k˜0 = 2−1/2␥k0,k˜0共1 − ␰2兲1/2 flow with a first-order displacement discontinuity method 共Crouch
and Starfield 1983兲 for the elastic solid deformation. The algo-
−1/2
⍀k0,k˜0 = 2−5/2␥k0,k˜0 共27兲 rithm discretizes the crack length using a moving or stretching
mesh with constant number of regularly spaced nodes. The tip
where respective values of the dimensionless fracture length are behavior changes from storage asymptote ⍀ ⬃ 共1 − ␰兲2/3, ⌸ ⬃ 共1
␥k0 = 2 / ␲2/3 and ␥˜k0 = 1 / ␲. − ␰兲−1/3 to leak-off asymptote ⍀ ⬃ 共1 − ␰兲5/8, ⌸ ⬃ 共1 − ␰兲−3/8 once
the leak-off parameter Cm is not zero. The change in the tip be-
havior results in the boundary layer of the O共Cmᐉ兲 thickness near
Edges the crack tip 共Adachi and Detournay 2010; Garagash et al. 2010兲,
which accommodates the leak-off asymptote immediately near
The MK edge corresponds to the case of a crack propagating in an the tip and the storage asymptote at the intermediate distances
impermeable medium 共C⬘ = 0兲. The normalized solution depends from the tip, and the transition between the two asymptotes. This
on a single parameter: the dimensionless toughness Km 共in the M boundary layer solution is implemented into a tip element logic of
scaling兲 or the dimensionless viscosity M k 共in the K scaling兲. The the numerical algorithm. The utilization of the tip element in this
original solution due to Spence and Sharp 共1985兲 and its further numerical algorithm effectively provides evolution of the tip
improvement 共Adachi 2001兲 corresponds to an expansion over a boundary condition in accord with evolving propagation regime.
class of functions 共of the similarity variable ␰兲 with the LEFM The KK̃ edge corresponds to the case of zero fluid viscosity.
near tip asymptote. Fast deterioration of the accuracy of their Semianalytical KK̃-edge solution 共Bunger et al. 2005兲 evolves
numerical method with dimensionless toughness Km decreasing
from the K vertex to the K̃ vertex, i.e., from the small- to the
below 1 has prompted Garagash and Detournay 共2005兲 to inves-
large-time asymptotic solution. In this problem, the pressure is
tigate further the small-toughness case. The departure of the so-
uniform along the crack and the tip behavior is of the form of
lution from the zero-toughness M vertex with increasing
⍀ ⬃ 共1 − ␰兲1/2. There is no change in the tip asymptote along this
toughness Km is driven by change in the dominant tip behavior
edge due to the fact that the tip solution is independent of the
from storage-viscosity asymptote 共25兲, ⍀ ⬃ 共1 − ␰兲2/3, ⌸ ⬃ 共1
balance between storage and leak-off when the viscosity of the
− ␰兲−1/3, to the toughness asymptote, ⍀ ⬃ 共1 − ␰兲1/2, ⌸ ⬃ ln共1 − ␰兲
fluid is zero, ␮⬘ = 0.
共where the negative logarithmic singularity of the net-fluid pres-
sure at the crack tip is the consequence of the zero fluid lag
assumption兲. This change is accommodated within a thin bound- General Solution
ary layer near the tip 共with thickness of ⬃Km6 ᐉ兲, which possesses
⍀ ⬃ 共1 − ␰兲1/2 behavior in the immediate vicinity of the tip and Asymptotic solutions discussed above correspond to the vertices
⍀ ⬃ 共1 − ␰兲2/3 behavior at the intermediate distances from the tip and edges of the rectangular parametric space MM̃K̃K. As such,

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010 / 1157

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
they provide a guide for the general numerical solution inside the Using representations 共31兲 and 共32兲 for the net pressure and
MM̃K̃K rectangle considered in this study. The solution inside the opening, respectively, the solution for W共␰ , Cm兲, P共␰ , Cm兲, and
parametric rectangle 共0 ⬍ Km ⬍ ⬁ , 0 ⬍ Cm ⬍ ⬁兲 has to utilize the ␥共Cm兲 of equations in the M scaling, Eq. 共37兲 with Eq. 共23兲, and
correct dominant tip behavior. Comparison of numerical solutions Eq. 共20兲 is sought via the adaptation of the numerical method of
with available analytical and semianalytical asymptotic solutions lines 共Liskovets 1965; Nilson and Griffiths 1983; Garagash and
indicates that it is essential to embed the correct tip behavior in Detournay 1997; Garagash 2006a兲. This method approximates the
numerical codes in order to obtain the relevant solution. Imposing above set of governing equations by a coupled system of ordinary
a tip behavior which is incompatible with the dominant physical differential equations governing evolution of the values of
process 共such as imposing the classical LEFM square-root asymp- P共␰ , Cm兲 at the fixed grid points over the space interval represent-
tote when the dissipation is essentially viscous兲 causes large er- ing the crack. This history-dependent system of the first-order
rors in prediction. Garagash and Detournay 共2005兲 and Garagash, ordinary differential equations with the initial conditions specified
“Plane-strain propagation of a fluid-driven fracture in a permeable in the next paragraph is coded in Visual Fortran V6.6 and solved
medium: Leak-off dominated regime,” Int. J. Solids Struct., sub- using the fifth-order adaptive Runge-Kutta method 共Press et al.
mitted, showed that in the zero and large leak-off limits, the zero- 1996兲. Details of this numerical method are given in Appendix II.
toughness solution provides adequate approximation to the The initial state, Cm = 0, of the sought solution is given by the
general solution for dimensionless toughness values as high as 1. self-similar MK solution, which corresponds to a stationary state
This is the result of a weak dependency of the small-toughness of the studied dynamical system 关all evolution terms in governing
solution on the dimensionless toughness: O共Km3.18兲 under the zero Eq. 共37兲 evaluate to zero at Cm = 0兴. In order to facilitate the
leak-off and O共Km̃4.44兲 under the large leak-off conditions, respec- numerical solution, a perturbation to this stationary state is intro-
tively. Consequently, the general numerical solution is considered duced by setting the transient solution equal to the MK-edge so-
in the finite toughness range, Km ⱖ 1, which allows for implemen- lution at small 共but finite兲 value Cm = 共Cm兲ini ⬎ 0.
tation of the toughness tip asymptote and, therefore, avoid the The accuracy of the numerical solution is estimated by the
complications of the changing nature of the dominant tip asymp- following quadratic error measure:
tote associated with small values of the toughness number.
冕冋 册
1
RHS共␰,Cm兲 2
e共Cm兲 = 1− d␰ 共34兲
0 LHS共␰,Cm兲
Numerical Method where RHS共␰ , Cm兲 and LHS共␰ , Cm兲 are the right- and left-hand
sides of the local fluid balance equation, second equation in Eq.
The asymptotic tip behavior of the net-fluid pressure in the case 共37兲.
of nonzero toughness 共13兲 can be expressed in the M scaling 关see
Eq. 共14兲 and Table 1兴 as
⌸m = Km−2Am ln共1 − ␰2兲, 1−␰Ⰶ1 共28兲 Dimensionless Numerical Results
with
Numerical solution inside the MM̃K̃K parametric space was ob-

冤 冢 冣冥
1/2
2 tained for various values of the “trajectory number,” the dimen-
+ ⌳m
2 3 sionless toughness Km = 兵1 , 1.5, 2 , 2.5, 3 , 4其. The initial value of
Am = ␥m + ⌳m + 2CmKm−1 共29兲 the leak-off evolution parameter Cm was set at 共Cm兲ini = 0.01 and
3 ␥m
the calculations were carried out up to Cm = 4.64. A uniformly
The term ⌳m = ⌳m共Cm兲 in Eq. 共29兲 is defined by spaced mesh of M = 20 elements was used, resulting in global
error 共34兲 bounded by 10−8.
t␥˙ m 1 Cm d␥m
⌳m共Cm兲 = = 共30兲 Figs. 3–6 show evolution of several fracture solution dimen-
␥m 6 ␥m dCm sionless variables with respect to the dimensionless evolution pa-
Since the net pressure and its gradient at the fracture tip are rameter 共leak-off or storage兲 along various solution trajectories
singular, we choose to present solution for the net pressure in the 共corresponding to various fixed values of Km兲. Fig. 3 shows the
form evolution of the dimensionless crack half-length in 共a兲 M scaling
and 共b兲 K scaling. The greater Km value represents trajectory fur-
⌸m = Km−2Am⌸共1兲共␰兲 + P共␰,Cm兲 共31兲 ther away from the MM̃ edge 共the viscosity-dominated regime兲
where the first term with ⌸共1兲共␰兲 = ln共1 − ␰2兲 = singular tip asymp- and closer to the KK̃ edge 共the toughness-dominated regime兲.
tote 关Eq. 共28兲兴 and P共␰ , Cm兲 = remaining “regular” part of net pres- Corresponding zero-toughness 关Km = 0 共Carbonell et al. 1999;
sure 共its gradient at the fracture tip is finite兲. Upon substitution of Adachi and Detournay 2002兲兴 and zero-viscosity edge solutions
Eq. 共31兲 into elasticity 共20兲, we obtain corresponding representa- 关Km Ⰷ 1 共Bunger et al. 2005兲兴 are also shown in Figs. 3共a and b兲,
tion for the opening respectively. At large time 共leak-off parameter兲, solution con-
verges toward the leak-off dominated regime 关the M̃K̃ edge 共Ga-
⍀̄m = Km−2Am⍀̄共1兲共␰兲 + W共␰,Cm兲 共32兲 ragash, “Plane-strain propagation of a fluid-driven fracture in a
where permeable medium: Leak-off dominated regime,” Int. J. Solids


1
Struct., submitted兲兴 shown by the dashed-dotted line. The conver-
4 gence of the numerical solution toward the asymptotic edge solu-
⍀̄共1兲共␰兲 = G共␰,␰⬘兲⌸共1兲共␰⬘兲d␰⬘ ;
␲ 0 tions in appropriate parametric ranges, e.g., M̃K̃-edge asymptote


1 in the large leak-off range and KK̃-edge asymptote in the large-
4
W共␰,Cm兲 = G共␰,␰⬘兲P共␰⬘,Cm兲d␰⬘ 共33兲 toughness range, is apparent in Fig. 3 and provides an indepen-
␲ 0 dent validation of the numerical method.

1158 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
Sm
f
Km = 4


MM-edge

Ωm (0)
 K-edge
M  Km = 1 
MM
γm -e dge


KK-edge
Km = 1, 1.5, 2, 2.5, 3, 4
 
MK-edge

Cm
(a) Cm Fig. 4. Fracture opening at the inlet versus the evolution leak-off
parameter in the M scaling for various values of the toughness num-
Sk̃ ber Km 共as in Fig. 3兲. The corresponding edge asymptotic solutions
are shown as indicated.


KK-edge
Fig. 6 shows the evolution of the efficiency ␩, an important
parameter in HF treatments, defined as the ratio of the fluid vol-
ume contained in the crack to the total volume of injected fluid. In
either M or K scaling, the efficiency can be evaluated as ␩
= ␥兰−1
1
⍀d␰. Efficiency evolves from the unit value at early time
γk  K-edge
 共when the leak-off is negligible兲 to the eventually zero value
M
共leak-off dominant regime when most of the fluid is leaked into
the rock兲. This efficiency transition is slower for higher values of
Km = 4, 3, 2.5, 2, 1.5, 1 the dimensionless toughness parameter since higher values of
toughness correspond to shorter crack length and, consequently,
smaller crack face area exposed to the leak-off. The goal of any
HF treatment is to remain within the area of “high efficiency”
during the whole procedure. The plot shows that the drop of ef-
ficiency is relatively sharp 共this is partially concealed by the loga-
(b) Ck rithmic scale in time兲. For instance, an efficiency of 50% is
reached at about Cm ⯝ 0.5 for Km = 1 and Cm ⯝ 0.9 for Km = 4.
Fig. 3. Fracture half-length versus evolution leak-off or storage pa-
rameter for various values of the toughness number, Km
= 兵1 , 1.5, 2 , 2.5, 3 , 4其: results in 共a兲 M scaling and 共b兲 K scaling. The
corresponding asymptotic solutions for the MM̃ edge 共Km = 0兲, KK̃

KK-edge
edge 共Km Ⰷ 1兲, and M̃K̃ edge 共C Ⰷ 1兲 of the parametric space 共Fig. 2兲
 
MK-edge
are also shown as indicated.

Figs. 4 and 5 show the evolution of the dimensionless crack


Πm (0)

opening and net pressure at the crack inlet in the M scaling. The Km = 4
same asymptotic edge solutions as in the case of the crack half-
length 共Fig. 3兲 are shown in Figs. 4 and 5 with the only distinction
that the toughness KK̃-edge asymptote is rescaled to the M scal-
ing using Eq. 共24兲, ⍀m = Km2/3共⍀k兲兩Km=⬁兩 and ⌸m = Km4/3共⌸k兲兩Km=⬁兩,
and evaluated at Km = 4. Clearly, the Km = 4 solution trajectory is

MM-edge
Km = 1
closely approximated by the rescaled KK̃-edge asymptote.
Notice that the dimensionless crack half-length and the inlet
crack opening decrease and the dimensionless inlet pressure in-
creases with Cm 共or with time兲. The time dependence of dimen- Cm
sional opening and pressure depends on corresponding scaling
factors ␧共t兲 and L共t兲 关see scaling 共14兲 and Table 1兴 and is actually Fig. 5. Net fluid pressure at the inlet versus the evolution leak-off
opposite, i.e., as expected, the dimensional crack length and open- parameter in the M scaling for various values of toughness number
ing increase and the net pressure decreases with time, as further Km 共as in Fig. 3兲. The corresponding edge asymptotic solutions are
discussed in the next section. shown as indicated.

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010 / 1159

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
Km = 4

Km = 1

M
η


M
-e
dg
e

KK-edge

Cm

Fig. 6. Efficiency versus the evolution leak-off parameter in M scal-


ing for various values of toughness number Km 共as in Fig. 3兲. The
corresponding edge asymptotic solutions are shown as indicated.

Figs. 7 and 8 show profiles of the dimensionless crack opening


and net pressure in the M scaling along the crack, respectively,
corresponding to the set of increasing values of the leak-off evo-
lution parameter Cm along four different solution trajectories with
共a兲 Km = 1, 共b兲 Km = 2, 共c兲 Km = 3, and 共d兲 Km = 4. Solutions for the
crack opening 共Fig. 7兲 with 共a兲 Km = 1 and 共d兲 Km = 4 are con-
trasted to the MM̃- and the KK̃-edge asymptotes, respectively.
The MK-edge and the M̃K̃-edge asymptotes are shown on each
plot to underline the convergence of the numerical solution to the
zero and large leak-off asymptotic limits, respectively. 关The
M̃K̃-edge asymptote in Figs. 7 and 8 is rescaled to the M scaling
using Eq. 共24兲, ⍀m = Cm−1/2共⍀m̃兲兩Cm=⬁兩 and ⌸m = Cm1/2共⌸m̃兲兩Cm=⬁兩, and
evaluated at Cm = 4.64.兴
Figs. 3–8 show good agreement between the numerical solu-
tion and the respective, independently obtained, asymptotic limits
corresponding to the edges of the parametric rectangle MM̃K̃K.
The parametric ranges where the solution is given by the respec-
tive edge or vertex asymptotes can be further quantified by con-
sidering Fig. 9, which shows four regime boundaries in the
parameteric space 共Km , Cm兲 corresponding to the 5% departure of
the solution for the net-fluid pressure at the crack inlet 共Fig. 5兲
from the corresponding edge asymptotes. The corners of this fig-
ure 共as marked兲, therefore, correspond to the parametric ranges
where solution is approximated by the corresponding vertex as-
ymptotes. Fig. 9 can serve as a practical guide to the applicability
of the asymptotic solutions to the plane-strain fluid-driven frac-
ture and, further, indicates the intermediate parametric range 共the
central curvilinear rectangle formed by the four curved bound-
aries兲 where the general numerical solution has to be used. For
Fig. 7. Profiles of fracture opening in the M scaling for different
example, the zero-toughness 共MM̃-edge兲 and zero-viscosity values of leak-off evolution parameter Cm = 兵0.1, 0.215,
共KK̃-edge兲 solutions approximate the general finite toughness so- 0.464, 1 , 2.15, 4.64其 and various fixed values of the toughness num-
lution in the ranges Km ⱕ 1.2 and Km ⱖ 3.8, respectively, while the ber: 共a兲 Km = 1; 共b兲 Km = 2; 共c兲 Km = 3; and 共d兲 Km = 4. Solutions for 共a兲
zero-leak-off 共MK-edge兲 and zero-storage 共M̃K̃-edge兲 solutions the lower and 共d兲 the higher values of the toughness number are
provide satisfactory approximations when Cm ⱕ 0.1 and Cm ⱖ 4 contrasted with the corresponding asymptotic solutions for the MM̃
共in the shown rage of Km兲, respectively. Dashed lines in Fig. 9 edge 共Km = 0—dotted lines兲 and the KK̃ edge 共Km Ⰷ 1—dashed lines兲,
correspond to the asymptotics of the four regime boundaries, respectively. Also, the zero 共Cm = 0兲 and large 共Cm Ⰷ 1—
which are obtained from the various available perturbed vertex dashed-dotted line兲 leak-off asymptotic solutions, as correspond to
solutions as explained next. Along the Km-axis, the two dashed the MK and M̃K̃ edges, respectively, are shown on each plot.
lines correspond to the 5% departure of the solution from the

1160 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
(a) Km = 1 10.00
 
K-corner
5.00 M-corner

4.64 1.00
0.50
Cm
Πm
0.10
1 0.05
M-corner K-corner
0.1 0.01
Cm = 0 0 1 2 3 4 5
Km

Fig. 9. Parametric boundaries corresponding to the 5% deviation of


the net-fluid pressure at the inlet from the corresponding asymptotic
(b) Km = 2 edge solutions. Dashed lines correspond to the boundaries’ asymptot-
ics obtained from the available perturbed solutions near the vertices
4.64
共see details in the text兲.

Πm
1
M-vertex and the K-vertex solutions along the MK edge, respec-
tively, obtained from the small-toughness 共Garagash and Detour-
0.1
nay 2005兲 and large-toughness 共Garagash 2006a兲 zero-leak-off
Cm = 0
solutions, respectively. Specifically, in the small-toughness solu-
tion in the M scaling, ⌸m共0兲 = 0.54495+ 0.0995E共Km兲, where the
first term is the M-vertex solution for the inlet value of the net
4.64
pressure and E共Km兲 = 0.1076Km3.168 is a small-toughness parameter,
gives Km ⯝ 1.343 for 5% deviation of the solution from the M
vertex. In the large-toughness solution in the K scaling, ⌸k共0兲
= 0.1831+ 1.7752Km−4, where the first term is the K-vertex solution
and the second term is a small viscosity 共M k = Km−4兲 perturbation,
gives Km ⯝ 3.732 for 5% deviation of the solution from the K
Πm 1 vertex. Similar considerations are used to obtain the Km thresh-
olds in the large leak-off case 共dashed lines along the second
Km-axis in Fig. 9兲 from the zero-storage asymptotic solutions of
0.1
Cm = 0 Garagash, “Plane-strain propagation of a fluid-driven fracture in a
permeable medium: Leak-off dominated regime,” Int. J. Solids
Struct., submitted. Along the second Cm-axis in Fig. 9 共corre-
(c) Km = 3
sponding to large Km兲, the two dashed lines correspond to the 5%
departure of the solution from the K vertex 共Ck ⯝ 0.0542兲 and
4.64 from the K̃ vertex 共Ck ⯝ 1.421兲 along the KK̃ edge 共Bunger et al.
2005兲. 关Note that above threshold values for Ck correspond to
Cm = CkKm2/3 共Table 1兲 in the 共Km , Cm兲 space in Fig. 9.兴

1
Time Dependence of the Solution
Πm
The previous section has discussed the solution in the various
0.1
Cm = 0
scalings defined by Eq. 共14兲 and Table 1, where the crack open-
ing, the net-fluid pressure, and the crack length are normalized by
corresponding time power laws expressed in terms of ␧共t兲 and
(d) Km = 4 L共t兲 共Table 1兲. These scalings arise naturally when various
asymptotic limits of the solution are considered and take advan-
ξ tage of the solution self-similarity at the vertices and some of the
edges of the parametric space 共Fig. 2兲. It is of interest to reexpress
Fig. 8. Profiles of net-fluid pressure in the M scaling for different some of the results in the form which exposes the time depen-
values of the leak-off evolution parameter Cm = 兵0.1, 0.215, dence of the dimensional solution. Specifically, Figs. 10 and 11
0.464, 1 , 2.15, 4.64其 and various fixed values of the toughness num- show the time evolution of the crack half-length and the inlet
ber: 共a兲 Km = 1; 共b兲 Km = 2; 共c兲 Km = 3; and 共d兲 Km = 4. The zero 共Cm value of the net pressure for various values of the dimensionless
= 0兲 and large 共Cm Ⰷ 1—dashed-dotted line兲 leak-off asymptotic so- toughness parameter, respectively, scaled by the corresponding
lutions are also shown on each plot. characteristic time-independent parameters

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010 / 1161

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
10 关Eq. 共35兲兴 scale the solution at the intermediate times 共t ⬃ tⴱ兲
tex when the effects of the fluid storage and leak-off are comparable
er 共Cm ⬃ 1兲. Figs. 10 and 11 can be conceivably used for the inver-
 -v
1 M sion of the downhole 共crack inlet兲 pressure data to obtain the
γm )

crack length in a HF treatment of a well.


/∗ (= Cm

Table 2 shows an example evaluation of the treatment terminal


4

0.1 efficiency, the crack half-length, and the net-fluid pressure at the
crack inlet from Figs. 6, 10, and 11, respectively, at the injection
time of t = 1 h and for a set of material parameters’ values given
0.01 by ␮⬘ = 1 Pa s, E⬘ = 30 GPa, and KIc = 1 MPa; injection rate Qo
Km= 0, 1, 2, 3, 4
= 10−3 m3 / s / m of the crack height 共unbounded in this plane-
strain model兲. Results in Table 2 are recorded for three cases with
103 x regard to the value of the leak-off coefficient, namely, C⬘
te = 10−5 m / s1/2 共“low leak-off”兲, C⬘ = 10−4 m / s1/2 共“moderate leak-
er
-v off”兲, and C⬘ = 4 ⫻ 10−4 m / s1/2 共“high leak-off”兲. 关The latter two
M
104 values were reported by Constien 共1989兲 for rock samples with
106 104 0.01 1 100 intrinsic permeabilities of 1 and 10 mD, respectively.兴 Note that
t/t∗ (= Cm
6
) even in the relatively low leak-off case, fluid losses have an im-
portant effect on the crack propagation—the corresponding value
Fig. 10. Evolution of the half-length of the crack for various values of the treatment efficiency of 89% at injection time which is still
of the dimensionless toughness parameter. 关The characteristic length small compared to the characteristic leak-off time 共t / tⴱ = 10−7 or
lⴱ and time tⴱ are defined in Eq. 共35兲.兴 Time power laws correspond- Cm = 0.07兲 is an indication that the fracture tends to leave the
ing to the M and M̃ vertex solutions are shown by dashed lines. “storage-dominated” regime very rapidly. This is consistent with
Fig. 6 where the decline rate of the efficiency is a maximum at

tⴱ =
␮⬘Q3o
E ⬘C ⬘6
; ᐉⴱ = 冉 冊
␮⬘Q5o
E ⬘C ⬘8
1/2
; pⴱ = E⬘
C ⬘2
Qo
共35兲
early time and decreases thereafter 共note the logarithmic scale of
the Cm-axis in this figure兲.
Let us now define an apparent toughness of a fluid-driven
Note that, according to Table 1, Cm = 共t / tⴱ兲1/6 and, consequently, fracture as p共0兲冑␲ᐉ. This is the value of the fracture toughness
characteristic values of time, crack length, and fluid net pressure that would be inferred from the downhole value of the net pres-
sure 共i.e., value at the crack inlet兲 if a uniform pressure distribu-
100 tion along the propagating crack is assumed. The latter
assumption is strictly true only in the toughness-dominated re-
Km= 0, 1, 2, 3, 4
gime 共KK̃ edge兲 when the apparent toughness, as defined above,
m(0))

M
-v is equal to the rock fracture toughness, p共0兲冑␲ᐉ = KIc. When the
)

er
Πm

te dissipation in the viscous fluid flow in the crack is not negligible,


x
m Π
−2

10 the apparent toughness is expected to be larger than the rock


−2
p/p∗∗ (= CCm

toughness value. Fig. 12 shows the time evolution of the apparent


fracture toughness scaled by characteristic time-independent
value
Kⴱ = pⴱ冑ᐉⴱ = E⬘3/4␮⬘1/4Q1/4
p(0)/p

o 共36兲
1
along various solution trajectories 共corresponding to the specified
set of values of the dimensionless rock toughness Km = K⬘ / Kⴱ兲.
Note that since K⬘ = 4冑2 / ␲KIc 关Eq. 共1兲兴, the apparent toughness is

M
-ve
rte indeed larger than the rock toughness, as in the normalized form
x
0.1 p共0兲冑␲ᐉ / K* ⬎ KIc / K* = 0.3133Km, for all trajectories shown in
106 104 0.01 1 100 Fig. 12.
t/t∗ (= 6
Cm ) The striking feature of the plots in Fig. 12 is the very weak
dependence of the apparent toughness on time in the entire range
Fig. 11. Evolution of the net-fluid pressure at the inlet for various 共from the early-time storage-dominated regime to the large-time
values of the dimensionless toughness parameter. 关Characteristic leak-off-dominated regime兲. In other words, p共0兲冑␲ᐉ can be ap-
pressure pⴱ and time tⴱ are defined in Eq. 共35兲.兴

Table 2. Examples of the Solution for the Crack Half-Length 共Fig. 10兲, the Net Fluid Pressure at the Inlet 共Fig. 11兲, and the Treatment Efficiency 共Fig.
6兲 at t = 1 h and a Set of Problem Parameters ␮⬘ = 1 Pa s, E⬘ = 30 GPa, KIc = 1 MPa, the Two-Dimensional Injection Rate Qo = 10−3 m2 / s, and Three
Values of the Fluid Leak-Off Parameter C⬘ as Indicated

C⬘ ᐉⴱ t ᐉ ␩
tⴱ pⴱ 共=Cm
6
兲 p共0兲
共m / s1/2兲 共h兲 共m兲 共MPa兲 t
* Km 共m兲 共MPa兲 共%兲
10−5 共low leak-off兲 9.3⫻ 106 1.8⫻ 107 0.003 10−7 0.25 238 0.36 89
10−4 共moderate leak-off兲 9.3 1,826 0.3 0.11 0.25 127 0.49 36
4 ⫻ 10−4 共high leak-off兲 0.0023 7.13 4.8 442 0.25 45 0.82 8

1162 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
1.285 when all dissipation is in the breaking of the rock at the crack tip,
Km= 4 KIc Ⰷ Kⴱ兲. The practical importance of Fig. 13 is that it allows for
1.28 the crack half-length ᐉ to be recovered directly from the value of
the downhole pressure p共0兲 共without reference to the fracture
1.275 propagation time兲 and two problem parameters: characteristic
109 107 105 0.001 0.1 10 1000 toughness value Kⴱ 关which is defined by the fluid and rock prop-
1.02
erties and injection rate 共36兲兴 and the rock fracture toughness KIc.
Km= 3
p(0) π /K∗

1.015
Summary
1.01
109 107 105 0.001 0.1 10 1000

0.825 A complete numerical solution of the fracture propagation driven


by a Newtonian fluid in an elastic permeable rock under the plane
0.82 Km= 2 strain and the zero fluid lag conditions is obtained in the space of
two dimensionless problem parameters. These parameters are re-
lated to the time-invariant partition between the dissipation in
0.815
109 107 105 0.001 0.1 10 1000 breaking the rock 共rock fracture toughness兲 and the viscous losses
0.76 in the fluid flow in the crack and time-dependent partition be-
Km= 1 tween the fluid leak-off into the permeable rock and the fluid
0.755 storage in the crack. The numerical solution converges to the
Km= 0 various available asymptotic solutions in the corresponding limits
0.75 of the two dimensionless parameters. These asymptotic solutions
109 107 105 0.001 0.1 10 1000 approximate fracture propagation in the corresponding limiting
t/t∗ physical regimes 共where, for example, one of the dissipation
mechanisms is negligible and/or one of the fluid storage mecha-
Fig. 12. Evolution of the apparent fracture toughness p共0兲冑␲l 关nor- nisms is negligible兲. The parametric boundaries of these limiting
malized by characteristic value Kⴱ; Eq. 共36兲兴 with time 关normalized regimes are established in this paper 共Fig. 9兲. Solution tabulation
by characteristic value tⴱ; Eq. 共35兲兴 for various values of the dimen- in the reduced 共dimensionless兲 parametric space allows for pos-
sionless rock toughness Km = K⬘ / Kⴱ sible inversion of the HF treatment data. It is also observed that
the inversion of the fracture length from the downhole pressure
data in a HF treatment can be greatly simplified due to the fact
proximated as a function of the dimensionless toughness param- that the apparent fracture toughness p共0兲冑␲ᐉ 关where p共0兲 is the
eter Km only, given by either the early-time 共MK-edge兲 or the downhole, i.e., the crack inlet, value of the net-fluid pressure, and
large-time 共M̃K̃-edge兲 asymptotic solutions 共see Fig. 13兲. The ap- ᐉ is the half-length of the crack兴 is approximately time invariant
parent toughness is approximately equal to 共0.755⫾ 0.005兲Kⴱ in and 关when normalized by a characteristic value Kⴱ
the small rock toughness limit, Km Ⰶ 1 共i.e., when all dissipation is = 共E⬘3␮⬘Qo兲1/4兴 is a tabulated function of the dimensionless rock
in the viscous fluid flow, KIc Ⰶ Kⴱ兲, and increases toward the rock toughness Km = K⬘ / Kⴱ 共measure of the solid-fluid dissipation par-
toughness value in the large rock toughness limit, Km Ⰷ 1 共i.e., titioning兲.

Acknowledgments
1.5
ge The writers thank Dr. José I. Adachi for insightful discussions on
ed
K- the numerical methods. Acknowledgment is made to the donors
K
of The Petroleum Research Fund, administered by the American
Chemical Society, for partial support of this research under Grant
1.0 MK-edge
√ (t/t∗ → 0)
No. ACS-PRF 35729-G2.
p(0) π
 K-edge

K∗ M
(t/t∗ → ∞)

MM-edge Appendix I. Normalized Governing Equations
0.5
Normalized global continuity equation 关Eq. 共18兲兴 with Eq. 共19兲,
lubrication 关Eq. 共17兲兴, and propagation condition 关Eq. 共21兲兴 are
specialized below for each of the four scalings defined in Table 1.
0.0 The form of elasticity 共20兲 is identical in all four scalings.
0 1 2 3 4 5 • In the M scaling, normalized solution Fm共␰ , Cm , Km兲
Km = K  /K∗ = 兵⍀̄m , ⌸m , ␥m其 is governed by

Fig. 13. Apparent fracture toughness p共0兲冑␲l versus the rock frac-

1
1 = 2␥m2 ⍀̄md␰
ture toughness K⬘ normalized by characteristic value Kⴱ 0
= E⬘3/4␮⬘1/4Q1/4
o 关Eq. 共36兲兴. The early-time 共MK-edge兲 and the large-


1
time 共M̃K̃-edge兲 asymptotic solutions demonstrate that the apparent ␥m共␪1/6Cm兲
+ 2Cm ␪2/3 d␪ 关␪2/3␥m共␪1/6Cm兲 = ␰␥m共Cm兲兴
toughness is approximately time invariant. 0
冑1 − ␪
JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010 / 1163

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
冉 1+2
t␥˙ m
␥m
冊冕 ␰
1
⍀̄md␰ + 冉 2 t␥˙ m
+
3 ␥m
␰⍀̄m +冊 冕 ␰
1
˙
t⍀̄md␰ S˜k 冋冉 3
4
+2
t␥˙ ˜k
␥˜k
冊冕 ␰
1
⍀̄˜kd␰ + 冉 冊 冕 册
1 t␥˙ ˜k
+
2 ␥˜k
␰⍀̄˜k +
1


˙
t⍀̄˜kd␰

冕冑 冕冑
1
Cm d␰ ⳵ ⌸m 1
3
⍀̄˜k ⳵ ⌸˜k
+ = − ⍀̄m3 1 d␰
␥m 1−␪ ⳵␰ + =−
␰ ␥˜k ␰ 1−␪ M˜k ⳵ ␰

−1/2
⍀̄m = Km␥m−1/2共1 − ␰兲1/2, 1−␰Ⰶ1 共37兲 ⍀̄˜k = ␥˜k 共1 − ␰兲1/2, 1−␰Ⰶ1 共40兲

and the elasticity equation. and the elasticity equation.


• In the K scaling, normalized solution Fk共␰ , Ck , M k兲 The time-derivative terms in the lubrication equation, as ex-
= 兵⍀̄k , ⌸k , ␥k其 is governed by pressed above in the four different scalings, can be expressed in
terms of the derivative in corresponding dimensionless evolution


1 parameter 关Eq. 共23兲兴.
1 = 2␥m2 ⍀̄kd␰
0


1 Appendix II. Numerical Method of Lines
␥k共␪1/6Ck兲
+ 2Ck ␪2/3 d␪ 关␪2/3␥k共␪1/6Ck兲 = ␰␥k共Ck兲兴
0
冑1 − ␪ A general method-of-lines numerical approach 共Liskovets 1965;
Nilson and Griffiths 1983; Garagash and Detournay 1997兲 is used

冉 冊冕 冉 冊 冕
herein. This method essentially corresponds to discretization of
1 1
t␥˙ k 2 t␥˙ k ˙ the solution in space, but not in time, such that governing equa-
1+2 ⍀̄kd␰ + + ␰⍀̄k + t⍀̄kd␰
␥k ␰ 3 ␥k ␰
tions reduce to a system of coupled ordinary differential equations
in time governing evolution of the values of the solution at the

冕冑
1
Ck d␰ ⳵ ⌸k
⍀̄3k discrete set of points in the space interval along the crack.
+
␥k
=−
Mk ⳵ ␰ First, a set of fixed grid points 兵␰i其 , i = 1 , . . . , M with ␰1 = 0 and
␰ 1−␪
␰ M = 1 is selected. The regular function P共␰ , Cm兲 关Eq. 共31兲兴 is then
approximated with order of error O共⌬␰2兲 共⌬␰ denoting the grid
step兲 by a piecewise linear function defined by its values 兵Pi共Cm兲其
⍀̄k = ␥−1/2
k 共1 − ␰兲 ,
1/2
1−␰Ⰶ1 共38兲 at grid points 兵␰i其
and the elasticity equation. Pi+1 − Pi
• In the M̃ scaling, normalized solution Fm̃共␰ , Cm̃ , M m̃兲 P共␰,Cm兲 = Pi + 共␰ − ␰i兲, ␰ 苸 关␰i,␰i+1兲
␰i+1 − ␰i
= 兵⍀̄m̃ , ⌸m̃ , ␥m̃其 is governed by
Corresponding expression for W共␰ , Cm兲 is then given by


1 M−1
1 = 2Sm̃␥m̃2
0
⍀̄m̃d␰ W共␰,Cm兲 = 兺
i=1
关Ia共␰,␰⬘兲ai共Cm兲 + Ib共␰,␰⬘兲bi共Cm兲兴␰⬘=␰
␰ =␰

i+1
i

冕 ␥m̃共␪−1/4Sm̃兲
1
In the above expression ai共Cm兲 and bi共Cm兲 are given by
+2 ␪1/2 d␪ 关␪1/2␥m̃共␪−1/4Sm̃兲 = ␰␥m̃共Sm̃兲兴
0
冑1 − ␪
Pi+1 − Pi Pi+1 − Pi
ai = Pi − ␰i ; bi =

冋冉 冊 冕 册
␰i+1 − ␰i ␰i+1 − ␰i

Sm̃
3
4
+2
t␥˙ m̃
␥m̃
冊冕 1


⍀̄md␰ + 冉 1 t␥˙ m̃
+
2 ␥m̃
␰⍀̄m̃ +
1


˙
t⍀̄m̃d␰ and function Ia,b共␰ , ␰⬘兲 is defined after Garagash 共2006a, Appen-
dix D兲

+
1
␥m̃
冕冑
1


d␰
1−␪
= − ⍀̄m̃3
⳵ ⌸m̃
⳵␰
Ia共␰,␰⬘兲 =
4

冕 ␰⬘

0
G共␰,␰⬘兲d␰⬘ =
4


2冑1 − ␰2arcsin ␰⬘

⍀̄m̃ = Km̃␥m̃−1/2共1 − ␰兲 , 1/2


1−␰Ⰶ1 共39兲
+ ␰⬘ ln 冏 冑冑 1 − ␰⬘2 + 冑1 − ␰2
1 − ␰⬘2 − 冑1 − ␰2

and the elasticity equation.
• In the K̃ scaling, normalized solution F˜k共␰ , S˜k , M˜k兲
+ ␰ ln 冏 ␰ − ␰⬘ 1 + ␰␰⬘ + 冑1 − ␰⬘2冑1 − ␰2
␰ + ␰⬘ 1 − ␰␰⬘ + 冑1 − ␰⬘2冑1 − ␰2
冏冊

= 兵⍀̄˜k , ⌸˜k , ␥˜k其 is governed by
Ib共␰,␰⬘兲 =
4
冕 ␰⬘
G共␰,␰⬘兲␰⬘d␰⬘ =
4
− 冑1 − ␰⬘2冑1 − ␰2
冕 ␲ ␲
1
2 0
2S˜k␥˜k ⍀̄˜kd␰

冏 冑冑 冏冊
1=
0 ␰ ⬘2 − ␰ 2 1 − ␰⬘2 + 冑1 − ␰2


+ ln
␥˜ 共␪−1/4S˜k兲 1 − ␰⬘2 − 冑1 − ␰2
1
1/2 k
2
+2 ␪ d␪ 关␪1/2␥˜k共␪−1/4S˜k兲 = ␰␥˜k共S˜k兲兴
0
冑1 − ␪ Furthermore, indefinite integral of W共␰ , Cm兲 can be written as

1164 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
冕 = t0共␰ j+1/2␥m兲 / t = Cm0 共␰ j+1/2␥m兲 / Cm6 . To make use of this relation,
M−1 6
1


W共␰,Cm兲d␰ = 兺 兵关Ja共1,␰⬘兲 − Ja共␰,␰⬘兲兴ai共Cm兲 + 关Jb共1,␰⬘兲 the values Cm0 must be interpolated from a 兵Cm , ␥m其 table as there
i=1 is no relationship known a priori between these two quantities.
This makes it necessary to store data points along the 兵Cm , ␥m其
− Jb共␰,␰⬘兲兴bi共Cm兲其␰⬘=␰i+1
␰ =␰
共41兲
⬘ i curve as determined by the evolving solution of the problem. The
technique of book keeping of data points is used for this purpose.
where Ja,b共␰ , ␰⬘兲 = indefinite integrals of Ia,b共␰ , ␰⬘兲 over ␰, which
An additional ODE is provided by Eq. 共29兲 in the form
lengthy closed-form expressions are omitted here for brevity.

冉 冊
Substitution of Eq. 共30兲 into the lubrication equation in Eq.
共37兲 yields d␥m 2Cm2 + AmKm2 − 2Cm冑Cm2 + AmKm2 2 6␥m
= − 共43兲

冋冉
Km2 ␥m
冕 冊 冕
1 1
dCm 3 Cm
⳵ ⍀̄m 6 2 Cm
d␰ = − + ⌳m ␰⍀̄m + 共1 + 2⌳m兲 ⍀̄md␰ + ⌫m Two additional equations are provided by the global balance
⳵ Cm Cm 3 ␥m
␰ ␰ equation in Eq. 共37兲 in the view of Eq. 共41兲

+ ⍀̄m3
⳵ ⌸m
⳵␰ 册 Km−2Am 冕
1
⍀̄共1兲共␰兲d␰ +
M−1


i=1
兵关Ja共1,␰⬘兲 − Ja共0,␰⬘兲兴ai + 关Jb共1,␰⬘兲
0
Evaluation of this equation at the midpoints ␰i+1/2 = 共␰i + ␰i+1兲 / 2


1
共i = 1 , . . . , M − 1兲 of the grid elements yields system of M − 1 1 ␪2/3␥m共␪1/6Cm兲
− Jb共0,␰⬘兲兴bi其␰⬘=␰i+1 = ␥m−2 − ␥m−2Cm
␰ =␰
d␪
coupled ordinary differential equations ⬘ i 2 冑1 − ␪


0


1 M−1
dAm dai 共44兲
Km−2
dCm
⍀̄共1兲共␰兲d␰ + 兺 关Ja共1,␰⬘兲 − Ja共␰,␰⬘兲兴
dCm

再冉
␰ i=1 where the integral on the RHS can also be computed using Sim-

+ 关Jb共1,␰⬘兲 − Jb共␰,␰⬘兲兴
dbi
dCm
冎 ␰⬘=␰i+1

␰⬘=␰i
=−
6
Cm
2
3

+ ⌳m ␰⍀̄m
pson’s rule and the stored data table 兵Cm , ␥m其 and the tip condition
in Eq. 共37兲

冕 冋
M−1
d⌸共1兲共␰兲
兺 关Îa共␰⬘兲ai + Îb共␰⬘兲bi兴␰⬘=␰
1 ␰ =␰
Cm Km␥m−1/2 = Km−2Am␬1 + i+1
共45兲
+ 共1 + 2⌳m兲 ⍀̄md␰ + ⌫m + ⍀̄m3 Km−2Am ⬘ i
␥m d␰ i=1

册冎

where functions Îa,b共␰⬘兲 = lim关Ia,b共␰ , ␰⬘兲 / 共1 − ␰兲1/2兴 are as follows:


+ bi , ␰ = ␰i+1/2 共42兲 ␰→1

8冑2 8冑2
in M + 2 unknowns: P1共Cm兲 , . . . , P M 共Cm兲 , ␥m共Cm兲 , Am共Cm兲, where, Îa共␰⬘兲 = arcsin ␰⬘ ; Îb共␰⬘兲 = − 冑1 − ␰⬘2
according to Eq. 共29兲 ␲ ␲

冕冑 and ␬1 is given by
1
1
⌫m共␰;Cm兲 = d␰ and
␰ 1 − ␪0共␰;Cm兲
⍀̄共1兲共␰兲
= − 8冑2ln 2
2Cm2 + AmKm2 − 2Cm冑Cm2 + AmKm2
␬1 = lim
2 ␰→1 共1 − ␰兲1/2
⌳m共Cm兲 = −
Km2 ␥m 3 Using two algebraic relations 共44兲 and 共45兲 to reduce the num-
Special care must be given to this numerical integral owing to the ber of unknowns in Eqs. 共42兲 and 共43兲 by 2 共e.g., by expressing
singular nature of the kernel near 1. The integral can be split into P M−1 and P M in terms of ␥m , Am , P1 , . . . , P M−2兲 yields the system
two parts: the regular part which can be integrated using Simp- of M history-dependent first-order ordinary differential equations
son’s quadrature rule and the singular part which can be approxi- in M unknowns 共e.g., ␥m , Am , P1 , . . . , P M−2兲
mated by leading term of asymptote behavior of ␪0 near ␰ = 1


␰ M−1/2
d␥m
1 = f ␥共Cm,␥m,Am, P1, . . . , P M−2兲
⌫m共␰i+1/2 ;Cm兲 = dCm
␰i+1/2
冑1 − ␪0共␰;Cm兲 d␰


1
1 dAm


+ d␰ = f A共Cm,␥m,Am, P1, . . . , P M−2兲
␰ M−1/2 ᐉ dCm
共1 − ␰兲
tᐉ̇

冉兺 冊
dPi
M−1
1 = f i共Cm,␥m,Am, P1, . . . , P M−2兲, i = 1, . . . ,M − 2
= ⌬␰ Sij dCm
j=i
冑1 − ␪0共␰ j+1/2 ;Cm兲
These equations are solved using a fifth-order Cash-Karp

+2 冑冉 2
3

+ ⌳m 共1 − ␰ M−1/2兲
Runge-Kutta method with adaptive step size 共Press et al. 1996兲.
We used M = 20 uniformly spaced spatial nodes for all of the
numerical runs, which results are reported in this study. These
where Sij = matrix that is constructed according to Simpson’s runs required 107 – 108 steps to cover the range of the evolution
quadrature rule 共Press et al. 1996兲. According to its definition, we parameter, 0.01ⱕ Cm ⱕ 4.64, with the typical run time on a Pen-
can express ␪0 in the leak-off term as ␪0共␰ j+1/2 ; Cm兲 tium 4 PC not exceeding several hours.

JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010 / 1165

Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org
References ture during injection and shut-in: Asymptotics of large toughness.”
Eng. Fract. Mech., 73, 456–481.
Adachi, J. I. 共2001兲. “Fluid-driven fracture in permeable rock.” Ph.D. Garagash, D. I. 共2006b兲. “Propagation of a plane-strain fluid-driven frac-
thesis, Univ. of Minnesota, Minneapolis. ture with a fluid lag: Early-time solution.” Int. J. Solids Struct., 43,
Adachi, J. I., and Detournay, E. 共2002兲. “Self-similar solution of a plane- 5811–5835.
strain fracture driven by a power-law fluid.” Int. J. Numer. Analyt. Garagash, D. I. 共2006c兲. “Transient solution for a plane-strain fracture
Meth. Geomech., 26, 579–604. driven by a shear-thinning, power-law fluid.” Int. J. Numer. Analyt.
Adachi, J. I., and Detournay, E. 共2008兲. “Plane-strain propagation of a Meth. Geomech., 30, 1439–1475.
hydraulic fracture in a permeable rock.” Eng. Fract. Mech., 75, Garagash, D. I., and Detournay, E. 共2005兲. “Plane strain propagation of a
4666–4694. fluid-driven fracture: Small toughness solution.” ASME Trans. J.
Adachi, J. I., Siebrits, E., Peirce, A., and Desroches, J. 共2007兲. “Computer Appl. Mech., 72, 916–928.
simulation of hydraulic fractures.” Int. J. Rock Mech. Min. Sci., 44, Garagash, D. I., Detournay, E., and Adachi, J. I. 共2010兲. “Multiscale tip
739–757. asymptotics in hydraulic fracture.” J. Fluid Mech., to be published.
Batchelor, G. K. 共1967兲. An introduction to fluid dynamics, Cambridge Irvin, G. R. 共1957兲. “Analysis of stresses and strains near the end of a
University Press, Cambridge, U.K. crack traversing a plate.” ASME Trans. J. Appl. Mech, 24, 361–364.
Bilby, B. A., and Eshelby, J. D. 共1968兲. “Dislocations and the theory of Lenoach, B. 共1995兲. “The crack tip solution for hydraulic fracturing in a
permeable solid.” J. Mech. Phys. Solids, 43共7兲, 1025–1043.
fracture.” Fracture, an advanced treatise, H. Liebowitz, ed., Vol. 2,
Liskovets, O. A. 共1965兲. “The method of lines.” J. Diff. Eqns., 1, 1308–
Academic, New York, 191–311.
1323.
Bunger, A. P., Detournay, E., and Garagash, D. I. 共2005兲. “Toughness-
Lister, J. R. 共1990兲. “Buoyancy-driven fluid fracture: The effects of ma-
dominated hydraulic fracture with leak-off.” Int. J. Fract., 134, 175–
terial toughness and of low-viscosity precursors.” J. Fluid Mech.,
190.
210, 263–280.
Carbonell, R. S., Desroches, J., and Detournay, E. 共1999兲. “A comparison
between a semi-analytical and a numerical solution of a two- Madyarova, M. 共2003兲. “Propagation of a fluid-driven penny-shaped frac-
dimensional hydraulic fracture.” Int. J. Solids Struct., 36, 4869–4888. ture in permeable elastic medium.” Master’s thesis, Univ. of Minne-
Carter, E. 共1957兲. “Optimum fluid characteristics for fracture extension.” sota, Minneapolis.
Mitchell, S. L., Kuske, R., and Peirce, A. P. 共2007兲. “An asymptotic
Drilling and production practices, G. Howard and C. Fast, eds.,
framework for the analysis of hydraulic fractures: The impermeable
American Petroleum Institute, Tulsa, Okla., 261–270.
Constien, V. 共1989兲. “Fracturing fluid and proppant characterization.” case.” ASME Trans. J. Appl. Mech., 74, 365–372.
Nilson, R. H., and Griffiths, S. K. 共1983兲. “Numerical analysis of
Reservoir stimulation, 2nd Ed., M. Economides and K. Nolte, eds.,
hydraulically-driven fractures.” Comput. Methods Appl. Mech. Eng.,
Prentice-Hall, Englewood Cliffs, N.J., Chap. 5.
36, 359–370.
Crouch, S., and Starfield, A. 共1983兲. Boundary element method in solid
mechanics, Unwin Hyman, London. Press, W., Teukolsky, S., Vetterling, W., and Flannery, B. 共1996兲. Numeri-
Desroches, J., et al. 共1994兲. “The crack tip region in hydraulic fracturing.” cal recipes in Fortran 90, 2nd Ed., Cambridge University Press, Cam-
Proc. R. Soc. London, Ser. A, 447, 39–48. bridge, U.K.
Detournay, E. 共2004兲. “Propagation regimes of fluid-driven fractures in Rubin, A. 共1993兲. “Tensile fracture of rock at high confining pressure:
impermeable rocks.” Int. J. Geomech., 4, 35–45. Implications for dike propagation.” J. Geophys. Res., 98共B9兲, 15919–
Economides, M. J., and Nolte, K. G. 共2000兲. Reservoir simulation, 3rd 15935.
Ed., Wiley, Chichester, U.K. Savitski, A. A., and Detournay, E. 共2002兲. “Propagation of a fluid-driven
Garagash, D., and Detournay, E. 共1997兲. “An analysis of the influence of penny-shaped fracture in an impermeable rock: Asymptotic solu-
the pressurization rate on the borehole breakdown pressure.” Int. J. tions.” Int. J. Solids Struct., 39, 6311–6337.
Solids Struct., 34共24兲, 3099–3118. Sneddon, I., and Lowengrub, M. 共1969兲. Crack problems in the classical
Garagash, D., and Detournay, E. 共2000兲. “The tip region of a fluid-driven theory of elasticity, Wiley, New York.
fracture in an elastic medium.” ASME Trans. J. Appl. Mech., 67, Spence, D. A., and Sharp, P. W. 共1985兲. “Self-similar solution for elasto-
183–192. hydrodynamic cavity flow.” Proc. R. Soc. London, Ser. A, 400, 289–
Garagash, D. I. 共2006a兲. “Plane-strain propagation of a fluid-driven frac- 313.

1166 / JOURNAL OF ENGINEERING MECHANICS © ASCE / SEPTEMBER 2010

View publication stats Downloaded 18 Aug 2010 to 129.173.211.101. Redistribution subject to ASCE license or copyright. Visit
http://www.ascelibrary.org

You might also like