You are on page 1of 13

Transp Porous Med (2007) 67:487499

DOI 10.1007/s11242-006-9038-7
ORI GI NAL PAPER
On the problem of uid leakoff during hydraulic
fracturing
Michael J. Economides Dmitry N. Mikhailov
Victor N. Nikolaevskiy
Received: 21 March 2006 / Accepted: 07 July 2006 / Published online: 28 September 2006
Springer Science+Business Media B.V. 2006
Abstract While a hydraulic fracture is propagating, uidowandassociatedpressure
drops must be accounted for both along the fracture path and perpendicularly, into
the formation that is fractured, because of uid leakoff. The accounting for the leakoff
shows that it is the main factor that determines the crack length. The solved prob-
lem is useful for the technology of hydraulic fracturing and a good example of mass
transport in a porous medium. To nd an effective approach for the solution, the thin
crack is represented here as the boundary condition for pore pressure spreading in the
formation. Earlier such model was used for heat conduction into a rock massif from a
seam under injection of hot water. Of course, the equations have other physical sense
and mathematically they are somewhat different. The new plane solution is devel-
oped for a linearized form that permits the application of the integral transform. The
linearization itself is analogous to the linearization of the natural gas equation using
the real gas pseudo-pressure function and where the ux rates are held constant and
approximations are introduced only into the time derivatives. The resulting analytical
solution includes some integrals that can be calculated numerically. This provides rig-
orous tracking of the created fracture volume, leakoff volume and increasing fracture
width. The solutions are an advance over existing discreet formulations and allow
ready calculations of the resulting fracture dimensions during the injection of the
fracturing uid.
Keywords Hydraulic fracture Pressure diffusion Mass transport Porous media
Leakage Integral transform Non-steady uid ows Special boundary condition
Heat conductivity
M. J. Economides
University of Houston, Houston, TX, USA
D. N. Mikhailov V. N. Nikolaevskiy (B)
Institute of Earth Physics,
Russian Academy of Sciences, Bolshaya Gruzinskaya 10,
Moscow 123995, Russian Federation
e-mail: victor@ifz.ru
488 Michael J. Economides et al.
1 Introduction
Hydraulic fracturing has been the stimulation/completion method of choice for the
vast majority of petroleum wells throughout the world, estimated to cover more than
90% of gas wells and 70% of oil wells.
Historically, it has been for lowpermeability reservoirs (e.g., k < 1 millidarcy). This
situation is no-longer true. Nowfracturing for high permeability reservoirs is common.
Companies have fractured even 2 darcy reservoirs. For low-permeability reservoirs,
obtaining the indicated length of the hydraulic fracture has always been the key ele-
ment in the execution of the stimulation treatments; for high-permeability reservoirs
increasing the fracture conductivity (width multiplied by fracture permeability) is
important.
Hydraulic fracturing is done by a fast increase in the bottomhole uid pressure
that can result in the cracking of the formation matrix with simultaneous injection of
proppant (often sand but, lately, man-made materials) mixed with a special gel, which
is a polymer solution usually with water as the carrying medium (Economides et al.
2002).
To model the fracture propagation there have been two classic and original
attempts, both deriving fromlinear elastic fracture mechanics and accepting the vision
of Grifths (1920) and Sneddon (1973) as the fundamental basic element of fracture
morphology. A necessary assumption has also been a plane-strain condition, whether
the resulting strain disappears in an innite number of horizontal or vertical planes.
This has resulted in two distinct geometries that are most currently accepted, both
vertical but one, where the fracture height is much longer than the length and one
where the length is much larger than the height.
Christanovich et al. (1959) and Zheltov (1975) were the rst to attempt to track
the problem, and assumed inadvertently the horizontal plane-strain condition and
thought that the main problemis the overcoming of toughness (as in the metal fractur-
ing) and rock pressure at the crack tip. Perkins and Kern (1961) and later augmented
by Nordgren (1972) solved a similar problem but assumed vertical plane-strain, which
provides for far longer fractures. The Christianovich and Zheltov and Perkins and
Kern works have formed the basis for much of the modeling work currently in use in
petroleum applications.
There are of course lingering problems in the original simplications.
First, there is the issue of accounting for uid leakoff, an often-dominating phe-
nomenon during fracturing. Second, often the net pressure during fracturing cannot
be accounted for by the original assumptions.
It was found (see, Garagash and Detournay 2000) that, because the crack tip is
spreading faster than the injected uid, the zone of vacuum was appearing between
them. Detournay and Garagash (2003) had assumed that the uid from permeable
ambient formation lled this zone, which played the role of a sink for the formation.
Our point, which is used below, consists of assumption that the zone is lled by
fracturing mixture at the rst moments after the overcoming of initial toughness of
the formation and the crack tip is converting into the uid source as well as the total
crack length. This can be deduced from the typical behavior of bottom hole pressure.
If one looks at the bottom hole pressure graphs (Fig. 1), the indication of sudden
drop in pressure immediately after initial crack appearance can be seen. It happens
due to lling of empty part of initial crack opening. Usually its volume is small in com-
parison with the total volume of mixture contained in the crack space that determines
On the problem of uid leakoff during hydraulic fracturing 489
Fig. 1 Bottom hole pressure curve shows fast decrease due to empty opening at the tip and slow
fracture growth
the length of the crack. The main part of the crack growth is happening smoothly and
consists (as we assume) of overcoming of rock pressure during the proppant injection
process along a whole crack length. The difference between the injected volume and
the contained one is the uid leakoff into the formation and has to be accounted for
carefully.
In the book (Economides et al. 2002) practically the method of stationary states
was developed for corresponding design of crack lengths. In many papers, see for
example (Nordgren 1972), the sink term with time of ltration was added into the
uid mass balance equation.
Here, we suggest considering the differential balance of uxes, concentrated along
the crack line, as the boundary condition for the pressure diffusion in the ambient
formation. That is, we use the method, developed much earlier for heat conduction in
strata from hot water masses, injected into a single layer (Rubinshtein 1972).
The approximations of our approach are: one-dimensional action of rock pressure
on the crack element, one-dimensional uid leakage only in the counter direction and
the linearization of uid balance equation along the crack. These lead to the analyti-
cal problem solution in the form of integrals that need only numerical parameters to
apply in computer calculations.
2 Mathematical problem formulation
Consider the fracture in its rst approximation where a cross-section of the fracture
is rectangular with xed height h
f
. The width of a crack in a cross-section is changing
continuously w = w(x, t).
Assume that the mixture of proppant and pad uid is modeled by Newtonian uid
with an effective viscosity
eff
.
Thus, the following are in effect:
1. The cross-sectional area is A(x, t) = h
f
w(x, t);
2. Using the velocity through a slot, the volumetric ow rate across A(x, t) is:
q(x, t) = A(x, t) (x, t) =
h
f
w
3
(x, t)
12
eff

x
p
f
(x, t) (1)
490 Michael J. Economides et al.
Here, p
f
(x, t)the uid pressure inside a crack; (x, t)the velocity of the ow along
the crack.
The width of a crack in a cross-section can be estimated approximately in accor-
dance with (67) part of the book (Economides et al. 2002)
w(x, t) =
2h
f
E

p
n
(x, t) ; E

=
E
1 v
2
(2)
Here, E

the plane elasticity modulus; Ethe Young modulus; vthe Poisson


coefcient; p
n
(x, t)net pressure.
By denition, net pressure is the difference between pressure p
f
(x, t) of the frac-
turing uid and the rock stress (x, t), that is, it coincides absolutely with the Terzaghi
effective stress (1996), and in the case under consideration it is arching the crack and
allows it to grow:
p
n
(x, t) = p
f
(x, t) (3)
Assume further that the rock stress is constant (and related to the vertical litho-
spheric load at the given depth). Then, we have for increments:
dp
n
= dp
f
dp (4)
The mass balance is used in the following form for the uid (see, 621 part of
Economides et al. (2002))

t
A(x, t) +

x
q(x, t) q
L
(x, t) = 0 (5)
Here q
L
uid owout because of leakoff into a porous formation through the
lateral walls of the crack.
To describe the pore pressure changes in the formation along axis Y, we can use a
linear equation of pressure one-dimensional diffusion (pieso-conductivity
Nikolaevskiy 1996):
p
n
t
=

2
p
n
y
2
; =
k
B
(6)
Here coefcient of pieso-conductivity; Beffective compressibility of the for-
mation (in a simple form, a sum of compressibility of uid B
f
and rock material
(1 )B
s
), is viscosity of a uid part of the fracturing mixture that penetrates into
the ambient formation.
So, the resulting mathematical model, describing the growth of a hydraulic fracture
has the following form:

t
p
n
(x, t)
p
n
y
=

2
x
2
p
4
n
(x, t) (y = 0; t > 0, x > 0) (7)
p
n
t
=

2
p
n
y
2
(y > 0; t > 0, x > 0) (8)
=
k

h
f
; =
1
12
eff
h
2
f
E
2
; =
k
B
=
h
f
E

B
On the problem of uid leakoff during hydraulic fracturing 491
p
n
= p
0
n
, (x = 0; t > 0) p
n
= p
0
, (t = 0, x, y)
p
n
= p
0
, (t; x, y ).
The maximum crack width w(x, t) is at the injection point (at the crack contact with
a well), that is, at x = 0. Let this value be w(x = 0, t) = w
0
.
This problem differs with the scheme of heat losses at hot water injection into a
seam(solved by Lauwerier and Rubinstein (1972)) mainly by the non-linearity of pore
pressure diffusion along the axis X. The non-linear term p
4
n
(x, t) could describe the
nite length of the crack growing in a virgin formation as in the case of the non-linear
Bussinesq equation for shallow ground water ow (see Polubarinova-Kochina 1977).
3 Linearization
In accordance with (2), the value w
0
corresponds to maximal effective pressure (net
pressure)
p
0
n
= p
n
(x = 0, t) =
E

2h
f
w
0
. (9)
Let us linearize the set of equations. For this goal, we introduce a new variable
u = p
4
n
Then, we have
p
n
= u
1/4
, dp
n
=
1
4 u
3/4
du
1
4 u
3/4
0
du, u
0
= p
4
0
and Eqs. 7 and 8 are transformed into the following ones:
u
t

u
y
=


2
u
x
2
(y = 0; t > 0, x > 0) (10)
u
t
=

2
u
y
2
(y > 0; t > 0, x > 0). (11)
Here,

= 4 u
3/4
0
and let u

= uu
0
to have zero initial conditions. We shall omit
the upper symbol though.
We introduce now new dimensionless variables
=

x; =

y; =

2

t. (12)
Then,
u

=

2
u

2
, ( > 0; > 0, > 0) (13)

u(, , )

u(, , ) =

2

2
u(, , ) , ( = 0; >0, >0) (14)
492 Michael J. Economides et al.
u( , ) = u( > 0, = 0) = 0, u( = = 0; > 0) = u
0
(15)
The used linearization is known after L.S. Leibenson for non-linear pressure diffu-
sion process and introduces a small error in the leakage process when the zone of ow
is innite (Nikolaevskiy 1996). Here, we have the same geometry. At the same time,
as it is known (Polubarinova-Kochina 1977) for the Bussinesq equation, the convex
prole typical for non-linear solution changes its form to the concave one after any
linearization.
4 Analytical solution
The system (13)(15) can be solved, applying the LaplaceCarson transformation
(Ditkin and Kuznetzov 1951; Sneddon 1961) U u relatively to variable . That is,
let introduce
U (, , ) =
_

0
e

u(, , ) d (16)
Then, transform U(, , ) of function u(, , ) will satisfy the following set of
equations:

2
U (, , ) = U (, , ) ( > 0; > 0, > 0) (17)
U = u
0
, = = 0; U = 0 ( =; =) (18)

2
U (, = 0, ) +

U (, = 0, ) U (, = 0, ) = 0 (19)
The solution of (17) with account for the condition at innity (18)has a form
U (, , ) = B(, ) e

(20)
The boundary condition (19) leads to a new differential equation:

2
B(, ) = ( +

)B(, ) (21)
that has the solution:
B = u
0
e

(22)
because of the condition u( = 0, = 0, t) = u
0
.
Therefore, U(, , ) has a form
U (, , ) = u
0
e

(23)
The solution in a full formcorresponding to (23) can be found if we recall the following
theorem (Ditkin and Kuznetzov 1951)
1

F () ()

_
0
f ( s) (s) ds (24)
On the problem of uid leakoff during hydraulic fracturing 493
where F(), () are transforms of functions f (t) and (t).
Let, use also the relation (Ditkin and Kuznetzov 1951)

F
_
+

t
_
0
1

t
exp
_


2
4 (t )
_
f () d (25)
and represent (23) as
U (, , ) = u
0
1

_
+

_
+

_
e

_
_
+

(26)
The following correspondences are known

exp
_

exp
_

2
4
_
(27)
exp
_

s
3
exp
_

2
4s
_
z = +

(28)
Now, using the theorems (24) and (25) one can get the double integral:
U (, , ) = u
0

_
0
1

( s)
exp
_


2
4 ( s)
_
ds

s
_
0
1

(s )
exp
_


2
4 (s )
_

2

3
exp
_

2
4
_
d (29)
If one changes the order of integration and use the integral (Rubinshtein 1972)
I () =
1

exp
_


2
4 ( s)


2
4 (s )
_
ds

( s)

(s )
= erfc
_
+
2

( )
_
(30)
where the tabulated errata co-function is used:
erfc y = 1 erf y =
2

_
z
e
y
2
dy
The resulting formula for a growing hydraulic fracture has the form
u(, , ) =
1
2
u
0

_
0
1

3
erfc
_
+
2


_
exp
_

2
4
_
d (31)
This integral permits easy numerical calculations. The illustration of behavior of width
of a crack in a cross-section is shown in Fig. 2.
One can see that, due to the linearization of (7), the crack tip position cannot
be determined exactly by special behavior of pressure non-linear diffusion along the
494 Michael J. Economides et al.
w
*
=w/w
0
0
0.2
0.4
0.6
0.8
1
0 2 4 6 8 10
1
2 3
Fig. 2 The width of hydraulic fracture along its axis w

() in the different time moments: 1. = 0.1;


2. = 0.5; 3. = 1.50
crack path. Therefore we consider the necessary opening of a tip as equal to the
diameter of single proppant particle and this determines the crack tip stop.
5 Cumulative leakoff
Its rate through walls of two parts at time is determined in accordance with (6):
Q
L
= 2h
f
k

_
0
p
f
(x, y, t)
y
y=0
dx = 2h
f
_

1
4
u
3/4
0

_
0
u(, , )

=0
d (32)
u(, , ) =
1
2
u
0

_
0
1

3
erfc
_
+
2


_
exp
_

2
4
_
d
We take into account also that
dp
n
=
1
4 u
3/4
du
1
4 u
3/4
0
du =
1
4 p
3
0
du
Because

erfc
_
+
2


_
=
1

( )
exp
_

1
4
( +)
2

_

_
0
exp
_

2
4
_
d = 2
(33)
the combination of the previous expressions gives:
Q
L
() =
h
f
p
0
2
_

_
0
1

( )
exp
_

1
4
( +)
2

_
d (34)
On the problem of uid leakoff during hydraulic fracturing 495
Because variable is changing from zero to , it is positive, and with account of
(39), the nite expression is
Q
L
() =
p
0
h
f
k
2
_

_
0
1

( )
exp
_

1
4
( +)
2

_
d (35)
According to [1], the main factor is the volume V of the mixture instantly containing
in a crack. This can be determined by the next step integration of Fig. 2: Consequently,
we shall use V instead of time.
The volume of proppantuid mixture in a crack of a constant height at time is
accounted for by the integral of w = w(x, t) as follows
V = 2h
f

_
0
w(x, t) dx = 2h
f

_
0
w(, ) d; (36)
w(, ) =
2h
f
E

p
n
(, ) p
n
= u
1/4
The illustrations of scaled cumulative rate of leakoff
Q

L
() = p
0
_
2Q
L
()
_
h
f
_ __

_
k
__
_

(37)
as well as scaled volume of the crack
V

() = V () E

__
4h
2
f
_
(38)
are given in Figs. 3 and 4 for arbitrary time moment.
6 The case of small leakoff
If leakage or time is small ( 0), the approximate solution can be found easily.
Using the Taylor expansion for (23) when >> 1, we get:
U (, , ) = u
0
exp
_

_
1 +
1

_
(39)
and
u(x, y, t) = u
0
exp(
1
2

) erfc
_
1
2

t
_
x

+
y

__
(40)
The rate of leakoff now is the following one:
Q
L
(t) = 2h
f
k

_
0
p
f
(x, y, t)
y
dx =
1
2
h
f
k

p
0
exp
_

t
_
erfc
_

t
_
(41)
We have accounted for that

y
p
f
(x, y, t) =
1
4
p
0

t
exp
_

x
_
exp
_

x
2
4

t
_
;
496 Michael J. Economides et al.
0
200
400
600
800
0 638 816 956 1077 1187 1290
2
V, m
3
V,V
L
, m
3

1
V,V
L
, m
3

0
200
400
600
800
0 808 963 1085 1193 1291 1384
2
V, m
3
1
V,V
L
, m
3

V, m
3
0
200
400
600
800
1000
0 806 898 967 1022 1076 1131
1
2
a
b
c
Fig. 3 The crack volume V(1) and the leaked volume V
L
(2) as functions of the injected volume at:
a. k = 10
12
m
2
; b. k = 10
13
m
2
; c. k = 10
14
m
2
0
200
400
600
800
0 1000 2000 3000 4000 5000 6000 7000 8000
V, m
3
1
2
3
t, sec
Fig. 4 The injected volume at in time at: 1. k = 10
12
m
2
; 2. k = 10
13
m
2
; 3. k = 10
14
m
2
On the problem of uid leakoff during hydraulic fracturing 497

_
0
exp
_

x
_
exp
_

x
2
4

t
_
dx =
_

t exp
_

t
_
erfc
_

t
_
The Taylors expansion of the right-hand part of (43) gives
Q
L
(t) =
1
2
h
f
k

p
0
_

_
1
2

t +

2

t
_
(42)
Expression (42) is valid for t 4/(). We recall that = (k/)(E /h
f
).
One can see that initially the point of injection and crack initiation are playing the
role of a source but differs from a source of pore pressure diffusion. The deviations
are specially seen in the next form of the solution
u(, , ) = u
0
e

/
2
erfc
_
1
2
+

_
(43)
7 Resulting formula and numerical samples
Let rewrite the formulae for the crack volume (36) and leakage rate (35) using the
initial coordinates x, y, t (see (12)). Changing the variable of integration to new one
=
_

2
_
, we get
Q
L
(t) =
1
2
h
f
p
0
k

t
_
0
1

(t )
exp
_

1
4

2
t
_
d (44)
V = 2h
f

_
0
w(x, t) dx; w(x, t) =
2h
f
E

p
n
(x, t) =
2h
f
E

u
1/4
(x, t) (45)
Here
u(x, t) =
1
2
u
0
x

t
_
0
1

3
erfc
_

t
_
exp
_

x
2
4
_
d
In Fig. 3, the dependencies of leaked volume and crack volume on the injected
volume are given for different values of permeability. One can see that in the case
of high formation permeability the leakage is prevailing that makes the traditional
concept of hydraulic fracturing meaningless.
The injected volume is represented as a function of time in Fig. 4. The length of
crack is shown in Fig. 5 for the same argument. The criterion for its opening is selected
as the equality of its width to the proppant radius. In the calculation the latter is 2 mm.
Other process parameters are represented in the table.
498 Michael J. Economides et al.
0
200
400
600
800
1000
0 1000 2000 3000 4000 5000 6000 7000 8000
t, sec
L, m
1
2
3
Fig. 5 The crack length (kept by proppant) in time at: 1. k= 10
12
m
2
; 2. k = 10
13
m
2
;
3. k = 10
14
m
2
Table of parameters
Crack height h
f
120 ft 37 m
Plane elastic modelus E

1.5 10
10
Pa
Effective mixture viscosity (proppant +uid)
eff
2 Pa s
Effective compressibility of the seam B 5 10
9
MPa
1
Seam porosity 0.3
Viscosity of leaking uid 0.07 Pa s
Net pressure p
0
n
at injection point 21 M Pa
The proppant particle diameter 4 mm
Permeability of the seam k 10
12
10
14
m
Note that w
0
1 cm for selected parameters. Besides t

= 4/() 26050 s if k = 10
14
m
2
;
2605 s if k = 10
13
m
2
and 260.5 sec if k = 10
12
m
2
. Let recall that approximation (42) can be
used for times t t

.
8 Conclusion
The solution developed here describes two simultaneous and interacting processes
of hydraulic fracturing the removal of rock mass along the fracture direction due
to crack growth and ltration of the fracturing uid from the crack into the porous
medium, perpendicularly to the fracture.
It is based on some simplications such as neglecting of a nite thickness of a
crack and correspondingly of lter cake effect, the assumption of one-dimensional
formation displacements and lter ow both orthogonal to the crack, exclusion of
non-linear effect when nite opening at the crack tip is determined only by a proppant
grain size. But these hypotheses, permitting the analytical solution, still save the main
features of the process overcoming of ow resistance along the crack and of lateral
rock pressure by leaky crack walls.
References
Economides, M.J., Oligney, R., Valko, P.: Unied Fracture Design. Orsa Press, Alvin, Texas (2002)
Grifth, A.A.: The phenomenon of rupture and ow in solids. Phil. Trans. Roy, Soc. A221, 163198
(1920)
Sneddon, I.N.: Integral transform methods. In Mechanics of fracture, Part 1, Methods of analysis and
solution of crack problems. (ed.) Sih G.C. Nordhoff Int. Leyden (1973)
On the problem of uid leakoff during hydraulic fracturing 499
Christianovich, S.A., Zheltov, Yu. P, Barenblatt, G.I., Maximovich, G.K.: Theoretical principles of
hydraulic fracturing of oil strata. The 5th World oil Congress, Section II, p. 23 (1959)
Zheltov, Yu. P.: Mechanics of Oil/Gas Seam. Moscow, Nedra (1975)
Perkins, T.K., Kern, L.R.: Width of the hydraulic fractures. J. Petrol. Technology. 937949, Espt, (1961)
Nordgren, R.P.: Propagation of a vertical hydraulic fracture. J. Petrol. Technology. 253, 306314 (SPE
3009) (1972)
Garagash, D., Detournay, E.: The tip region of a uid-driven fracture in an elastic medium. J. Appl.
Mech. 67, March, 183192 (2000)
Detournay, E., Garagash, D.I.: The near tip region of a uid-driven fracture propagating in a
permeable elastic solid. J. Fluid Mech. 494, 132 (2003)
Rubinshtein, L.I.: Temperature Fields in Oil Formations. Nedra, Moscow (1972)
Nikolaevskiy, V.N.: Geomechanics and Fluidodynamics (with Applications to Reservoir Engineering).
Kluwer, Dordrecht (1996)
PolubarinovaKochina, P.Ya.: Theory of Ground Water Flows. 2nd ed. Nauka, Moscow (1977)
Ditkin, V.A., Kuznetzov, P.I.: Handbook on Operation Calculus. Gostekhizdat, Moscow (1951)
Sneddon, I.: Fourier Transforms. New York (1961)

You might also like