‘me- Sooq

Propagation of a Vertical Hydraulic Fracture

R. P. NORDGREN
I
. SHELL DEVELOPMENT
HOUSTON, TEX.
CO.

ABSTRACT fracture length and width.

A different approach to the determination of
This paper treats the propagation of hydraulic fracture width was taken by Perkins and Kern. G
fractures o/ limited vertical extent and elliptic They considered a vertically limited fracture under
cross-section with the effect of fluid loss included. the assumption of plane strain in vertical planes
Numerical and asymptotic approximate solutions in perpendicular to the fracture plane. The cross-
dimensionless jorrn give the lracture length and section of the fracture is found to be elliptical, and
width at any value of time /or any set of physical the maximum width decreases along the fracture
parameters. The insight provided by tl]e dimension- according to a simple formula that contains the
less results and approximate solutions should be fracture length. In the derivation of this formula, G
..-.. /..1 :-
U> CJUL L/l
●L- .3--:
L~JC UC>t~TI
-- -1
U!
l--_a ----
!! ULLUIC
a-__ -.---a-
LICULIIIC1l L>.
c!..:> AU>>
lLUIU
1 ---- -_-l
dIIU
[---------
lld LLUIC
_- l.._..-
VUIUI1l
-l -----
C Cllilllf+
--- _.. -l---->
tll C llt’/$1CCLCU

in the continuity equation and the fracture length is

INTRODUCTION not determined. In a subsequent applicarion,7 a
“reasonable” fracture length was assumed. Carter’s
The theory and practice of hydraulic fracturing
formulaz for length and the width formula of Perkins
has been reviewed by Howard and Fast} Therefore,
and Kern G are both cited by Howard and Fast, 1 and
we confine our discussion of previous investigations
combined use of the two formulas is believed to be
to those pertinent to the present study of the
common practice.
propagation of vertical fractures.
The present theoretical investigation is
.4n important theoretical result is Carter’s
concerned with vertically limited fractures of the
formulr? for the area of a fracture of constant widtil
type studied by Perkins and Kern. G However, we
formed by injection at constant rate with fluid lost
include the effecrs of fluid loss and fracture volume
to the formation. For a vertical fracture of constant
change in the continuity equation. Consequently,
height, Carter’s formula gives fracture length as a
fracture length is determined as part of the solution.
function of time. In general, Carter’s assumption of
General results for the variation of fracture width
constant width is not realistic. However, at large
and length with time are obtained in dimensionless
vaiues of time the effect of this assumption
form by-a numerical method. In addition, asymptotic
becomes insignificant since the effect of fluid loss
solutions are derived for large and small values of
dominates. time. The small-time solution is also the exact
The width of a vertical fracture was first
solution for ‘the case of no fluid loss to the
;-., -.-.: --*. J h., ul..:o*:-.. n.r; .- 7hal,,..*3
-,1> ..” “ .~=~~~
~a~

111“=L1~aL=u Uy . . . . ..a.. a., u.. u

formation. For large values of time our asymptotic
the assumption that the width does not vary in the
formula for fracture length is identical with Carter’s
vertical direction. Thus, a state of plane strain
formula2 at large time. Our large-time formula for
prevails in horizontal planes and the width can be
fracture width differs from the formula of Perkins
determined as the solution of a plane elasticity
and KernG by a numerical factor that varies along
problem. An approximate solution is found in Ref. 3
the fracture. In comparison with our formula, this
upon neglect of fluid loss, fracture volume change,
formulaG overestimates the width by 12 percent at
a?.d y..
m.,=c=t,.,=
.,.,...- var;nri,-tn
. . . . . . . . .. ~!~,n>o ~~~ f~~~~~~~. ~~~
the well and 24 percent at the midiength of the
fracture length is determinedbby the condition of
fracture. At earIy times Perkins and Kern’s
finite stress at the fracture tip. Baron et al.4 and
formulaG for width in terms of length is again a fair
Geerrsma and de Klerk5 have included the effect of
approximation to OUr resu1t. However> OUr formula
fluid loss in the approach of Ref. 3. Geertsma and
for length differs from Carter’s formula, 2 which is
de Klerk 5 give simple approximate formulas for
not applicable since the neglected width variation
is important at early times.
Original manuscript received in Society or Petroleum Engineers
office June 1, 1970. Revised manuscript received March 20, 1972. The results for the width of a vertically limited
Paper (SPE 3009) was presented at SPE 45th Annual Fall fracture as obtained here and in Ref. 6 differ from
Meeting, held in Houston, Oct. 4-7, 1970. @ Copyright 1972
American Institute of Mining, Metallurgical, and Petroleum the results for vertically constant fractures.3-5 In
Engineers, Inc.
particular, our large-time solution for width differs
]Rererences given at end of PaPer.
from that of Geertsma and de Klerk5 by a factor
This paper will be printed in Transactions volume 253, which
will cover 1972. proportional to the fourth root of fracture height

sOCIETY Ok’ PET II OLEUM EN CINDERS JOURNAL

divided by fracture length. This difference may be solution. In view of the expected slow variations
significant in practical applications, and we will with x, an approximate state of plane strain
discuss it further. prevails in vertical planes perpendicular to the
In the next section we formulate the problem of fracture (planes x = const.).
vertically limited fractures along the lines of Ref. Before fracturing, a compressive stress (.S) acts
6. The underlying assumptions are discussed in on planes parallel to the fracture in the reservoir.
some detail. The basic equations then are cast in Different values for S in the rock surrounding the
dimensionless form suitable for numerical solution, reservoir again should have a small effect. The
which is carried out in Appendix A. The last section fluid in the fracture is under a pressure S + .Ap, and
of the paper gives a comparison with previous we neglect variation of Ap with z so that ~ =
investigations and a discussion of the applicability Ap (x, t). Then, by the plane-strain solution for a
of the present results. crack under constant pressure,8 the fracture is
elliptical and the width, w(x, t), is given by
FORMULATION
(1 - ‘v)
We consider a vertical fracture propagating in a w= (h2 - /}@ )1/2 Ap ,
straight line away from the well (Fig. 1). As in Ref.
G
6, the vertical height of the fracture is assumed to
be limited to a constant distance, h, by layers of
fracture resistant rock. The fracture resistance may
be due to higher horizontal tectonic stress or higher
fracture strength in the rock surrounding the
W=o> 121+ . . ...(1)
reservoir.
where G and v are the bulk shear modulus and
Let x, y, z be a system of rectangular Cartesian
Poisson’s ratio for the formation. The plane-strain
coordinates with x axis in the direction of fracture
solution neglects the effect of pore-pressure
propagation, z axis parallel to the well axis, and
variation on elastic response which is a small
origin at the well face (Fig. 1). Thus, the fracture
effect in fracture problems. 9
lies in the x-z plane with O <x< L and IzI <%L.
Next, we consider the flow of fluid in the fracture.
Following Ref. 6, we assume that an isotropic
The continuity equation for flow of an incompressible
homogeneous elastic material surrounds the
fluid in the fracture can be written as
fracture. The homogeneity assumption may not be
fully consistent with our earlier assumption that
aq AA
fracture resistant rock bounds the fractured
reservoir. However, on the basis of St. Vennant’s —+q~+—= o’”””” ““””(2)
* at
principle in the theory of elasticity, the elastic
behavior near the fracture is controlled mainly by where q (x, t) is the volume rate of flow through a
the reservoir rock and moderately different surround- cross-section (x = const. ) of the fracture, qf(x, t) is
ing rock should have a small effect on fracture the volume rate of fluid loss to formation per unit
width. We confine attention to the stage of fracture length of fracture, and A (x, t) is the cross-sectional
propagation when the length, L, is much greater area of the fracture.
.--. .—- UIaL
we assume .L-. ~ iS f~i~t~(i -----..--
than the height, b, and aii quantities are expected tO & ~IC>auIc

tO vary siowi~ with x alone -.–. ..O most

---- Of the fraccure, gradient by the classical solution 10 for laminar
This assumption will be verified after obtaining the flow of a Newtonian viscous fluid in an elliptical
tube of semiaxes %h and ZWmax with b >> Wmax,
namely,
~w’”

~~h aAp
q= -— W=wmax’””(s)
—J
64p ax

where p is the viscosity of the fracturing fluid. The

range of validity for laminar flow has been
discussed in Ref. 6. By Eq. 1,

(1 - v)h
w=q=o = ~ AP” “ “ o ‘4)

i .--– ---=.
and Eq. 3 becomes
d I ITG
t =. L(@). . . . (5)
z q
FIG, 1 — FRACTURE GEOMETRY. 256(1 - v)~ ax

.ALGL ST. 1972 S07

The cross-sectional area of the elliptical fracture Other boundary conditions can be treated by the
is given by numerical method given in Appendix A.
Initially, the fracture is closed, i.e.,
h/2 11
A= f wd Z =- Wh . . . . ...(6) /19\
iJ(x, Oj= O...... .“..”~lLJ
‘-h/2 4
Furthermore, the fracture remains closed for x >
The fluid-loss rate, qf, is related to the loss L (t), i.e.,
rate, Uf, per unit area of fracture surface by
W(x, t) = o , X z L(t) . . (13)
h/ 2
~J=2
J -h/2
adz,
i
. . . . . . . (7)
where the fracture length L (t) is to be determined
as part of the solution.
Upon the substitution of Eqs. 5, 6, 8 and 9 into
and if u p is independent of z, then
Eq. ~, we ~btaifi

qi =2t-lu . . . . . . . . . . . .(8) G a’ 8C ~w
1 +—
— (w’) =
Experiments 1,11 suggest that for many fracturing 64(1 - V )Wh ~X2 m’m(a ~ t

fluids the loss rate can be taken as

O< X<L, t>o. .(l4
c
=— . . . . . . . . . . . . (9) where 7(x) is the time at which the fracture opened
at x, i.e.,
‘L m
where C is the fluid-loss coefficient and 7 is the
T[. L(t’)] = t’ , o<t’5t <15)
time at which flow begins. Coefficient C depends
on various fluid and formation parameters, and it is
In other words, ~(x) is the function inverse to L (t).
usually determined experimentally. For some fluids,
Substitution of Eq. 5 into Eq. 11 gives
Eq. 9 should be modified to include the so-called
“spurt loss” at t = r. The effect of spurt loss on

1
fracture length is found to be small in general 256(1 - v) p
(Appendix E). We adopt Eq. 9 in the present study, ~ (w’) = qi “
although the conditions under which it was verified ax = ?_TG
experimentally are not strictly satisfied. Namely, -[ x o
the pressure difference between fracture pressure . . . . . . . . . . . . . . (16)
S + & and formation pore pressure pl, i.e.,
Now our problem is the solution of the nonlinear
partial differential equation (Eq. 14) for IV(X, t)
s+ Ap-pf . . . . . . . . . . . .(10)
subject to initial condition (Eq. 12) and boundary
contains Ap, which varies with x and t. In some conditions (Eqs. 13 and 16) with L (t) also to be
cases, Ap << S – p so that the assumption of determined.
constant pressure {inferences should lead to To minimize the amount of computation, it is
meaningful results. Furthermore, we will find that convenient to cast the governing equations into
Ap varies slowly with t and x and use of an average dimensionless form. Thus, if we set
value for Ap in Eq. 10 may be acceptable. The
X=a
condition of one-dimensional flow is not strictly % ‘ ‘=aLD
satisfied either, although in view of the slow t = BtD , W = eWO
variation of Ap with x, the resulting error should be
small. Theoretical refinements in the determination where
of q ~ would require additional experimental research
on filter cake behavior under time-dependent
pressure. Also, consideration of two-phase flow
would be necessary if the fracturing fluid differs
significantly from the pore fluid.
At the well (x =0) a boundary condition involving
g and p must be specified. This condition will
depend on the q(p) relation for pumping equipment
used in the fracturing process. Here we treat the
case of constant inflow rate
1/3
16(1 - ~)vq . . . (17)
q(O, t) = qf = const. . . . . . . (11) e =
[ 1
c~ Gil J

308 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 n. .-.7 ------ -- T. -m.,. --
then Eqs. 12 and i6 become UDL. U>MUN ur KESUI., 1s

Results have been obtained by three different

~Lcc~o& . Tha ,.,,,..,,.:-.1
*A.L *,IA --J. -4
U, L.. LCL’S ,.$
,,C.AL”” “A As--.-.A:-
.,ppc L,u&* ‘,A k,’aa
1.. .

been applied to the dimensionless equation (Eq..

18). The fracture length LD(tD) and the fracture
width at the well W(O,tD) vs time tD are shown in
Figs. 2 and 3.
An approximate solution in elementary functions
-~ (Wt)lx=o = 1 ~ W(XD, 0) = 0 is derived in Appendix B upon neglect of the term
axD D
dA/Jt in Eq. 2. This approximation is valid for
sufficiently large time as indicated in Appendix B
x~>L~j and confirmed by comparison with the numerical
w~ (XD,t~) = O for
results in Figs. 2 and 3. For an error in the
approximation of less than 10 percent tD > 1.o.
The case of no-fluid loss out of the fracture is
treated in Appendix C. A similarity transformation
in combination with the numerical method of
. . . . . . . . . . . . . . (18) Appendix A furnishes a solution for any value of
time. The results of Appendix C also constitute an
Eqs. 18 have been solved for WD(XD, tD) by the
approximate solutlon for the case of fluid 10SS out
numerical method of Appendix A. of the fracture at sufficiently small time, BY
comparison with numerical results in Figs. 2 and 3,
for an error in the approximation of less than 10
percent, tD <0.01. Also, tD must be large enough
t 1 that L >> h, as discussed below.
10 Fig. 4 shows the variation of maximum fracture
LARGE FLUID L05s width along its length, x, in dimensionless form for
( LARGC
to APPROXIMATION
05 the approximate solutions of Appendices B and C.
.D(lO1
=0637ID—
,/, These two width variations are remarkably similar.
11 /
Widths obtained by the numerical method lie close
to the width curves in Fig. 4 and are not shown.
‘oI:E%G-77RCAL
01
/, For any particular problem the fracture length
and width at any time can be obtained from Figs. 2

FE
●
Lo[lD)=
132,0
through 4 by means of Eq. 17. One merely computes
005
the value of tD by Eq. 17, reads L.D(tD) and WD(O,
Y tD) from Figs. 2 and 3, and computes L(t) and
W(O, t) from Eq. 17. Then, the approximate variation
t of maximum fracture width with distance from the
001 , I , t I , I well follows from Fig. 4. Also, the pressure Ap is
0,01 005 01 05 10 50
‘D obtained from W by Eq. 4.
The approximate solutions of Appendices B and
FIG. 2 — DIMENSIONLESS FRACTURE LENGTH VS
DIMENSIONLESS TIME.
C provide considerable insight into the effect of

,
i- -i
LARGE FLUID LOSS

1~
(LARGE t~ APPROXIMATION)
NO FLUID LOSS
0,5 (SMALL tD APPROXIMATION) wD(o, tD) =0.798 t:
NUMERICAL

‘wD(o, tD) = 1.00 tD~

$r ‘

0.1 1 I 1 I 1 I I 1 1 I I I 1 I I I 1 I I I I I 1 I
0003 0.005 0.01 0.02 0.05 0.1 0.5 1,0
‘o

AUGUST, 1972 S09

 ~~~:~ ~a~
ym~=,rc;ral
, .=.L.-. Ra. al’n’ara.
~-.m,,,L.L.a =’ ~~ $.’=.- . . ..-
LL-v.u.= .c,’~..l,
1---*I. Carter’s i0rmuia,2 since the fluid loss effect now
pressure. For the case of large fluid-loss rate or is dominant.
large time (Appendix B), the fracture length and the Detailed studies of effects of parameter variations
maximum width at the borehole are given by Eqs. can be made through Eq. 17 and the dimensionless
B-5 and B-9 as results in Figs. 2 and 3 which should be useful in
the design of fracture treatments. For example, the
qi tllz fluid-loss coefficient, C, required to achieve a
L(t) = — desired fracture length after a specified pumping
time can be determined from these results. Pumping
rrCh
pressures required to achieve a desired injection
rate can be determined. Also, the fracture treatment
can be designed so that the fracture opening admits
F( 21- V)wqf 1~/* , a certain size of proppant.
w(~,~) G Q .— t“ P .
L
GCh 113 J When applying- the- foregoing results, the basic
assumptions of the formulation must be kept in
. . . . . . . . . . . . . . . (19) mind. Probably the most important assumptions are
that fracture height is limited vertically and that
fracture length is much greater than height (L >> b).
As seen from Fig. 4, the latter assumption assures
slow variation of width, w, with x which validates
the basic assumption of plane strain. If the condition
L >> b, in particular immediately after fracture
L(t) = 0.68
initiation, is not fulfilled or if the fracture is not
[,1 I,~h411’5t4’5 vertically limited, then other theoretical models of
1/5 fracture propagation are more appropriate. For a

W(O, t) = 2.5
[
(1 - ~)pq

Gh
1 tl/5 “
vertical fracture with L << b and injection over all
of the height, the horizontal plane strain model of
Refs. 3 through 5 appears to be more appropriate.
If injection is limited to a short interval, then a
. . . . . . . . . . . . . . . (20)
radially propagating vertical fracture should be a
As seen from Eqs. 19 and 20. the fracture length suitable model as suggested by Perkins and Kern. c
and width grow faster with time in the no-loss This model has been treated approximately by
(small-time) case than in the large-loss (large-time) Geertsma and de Klerk 5 and others. There are
case. The manner in which other ~arameters affect significant differences between these solutions 3-6
fracture size can also be seen from Eqs. 19 and and the present one. In particular, for a linearly
20. For example, L is proportional to qi in Eq. 19 propagating (one-sided) vertical fracture, the
and qi315 in Eq. 20. In Eq, 19, L is not affected by approximate fracture width at the well as given by
G or u. Geertsma and de Klerk (for v = 0.25) as
Comparison is made with the approximate
solution of Perkins and Kernc in Appendix D. It is
found that, if Eqs. 19 and 20 are cast in the form
W(O, t) as a function of L (by elimination of t), then
6 ~~ f~~~!y
.. ,.,(E@
TJi’n)
\”, )’”>.

2Gh
Ehe approxirn.are ~Q!lJ~~~t) accmrare
.. -s...-.- . 1 I

However, at early times Eq. 20 for L(t) differs from

which differs from our large-time solution (Eq. 19)
Carter’s formula, 2 which was based on the
by the factor 0.67(L/h) ~. This factor is large when
assumption of constant width. At large times, Eq.
L/h << 1 and L/h >> 1. As stated above, our
19 for L(t) agrees with the large-time form of
solution is not valid for L/b << 1. On the other
hand, we do not believe that the approach of Refs.
3 through 5 is valid for L/h >> 1, since the
, .––
—“1 assumption of vertically constant width in the
1 vertical &i~C<i-il pi~C!iid~S fi2Ctiii~ CIGSiii~ at t!i~
($..., T]”, .,.,0,, ”., !0., 7
top and bottom. The error thereby induced is
CL.*GC r,., AP.ROX4.. T!ON I
roughly proportional to L and, thus, the error
becomes large for L >> b. In contrast the error
q:,,
,,
:
induced in our solution by rapid variations near the
fracture tip is roughly proportional to h and this
error becomes small for L >> h. For L .= h a three-
1 dim.e~~i~na! Se!utietl is ..“e,-accanr
-., s.,..le. Y ? h,.,
“--
lcJeL; rl*
‘--=
.=*
L&.& ‘k
,t.—-—.L.._.L
o
J—
c.,
.
. . . . .--J .. . . . J,0
present. Such a solution also would conclusively
L
resolve the ranges of applicability of our results
and those of Refs. 3 through 5.
FIG. 4 — VARIATION OF MAXIMUM FRACTURE WIDTH
WITH DISTANCE FROM WELL.

S.lo SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 NOMENCLATURE International Society of Rock Mechanica, Lisbon
(1966) 585-594.
A= cross-sectional area of the fracture 10. Lamb, H.: Hydrodynamics, 6th Ed., Dover Publica-
c. fluid-loss coefficient tions, Inc., New York (1932).

G= bulk shear modulus of formation 11. Hall, C. D., Jr. and Dollarhide, F. E.: “Performance
of Fracturing Fluid Loss Agents Under Dynamic
h= thickness of fractured layer Conditions, ” J. Pet. Tech. (July, 1968) 763-769.
K= constant (defined by Eq. C-3) 12. Ames, W. F.: Nonlinear Partial Differential Equations
L= length of fracmre in En.eineerinp, Academic Press, Inc., New York
(1965): -
LD = dimensionless L
h.= 1.,..- -.nn.l.,..a ;..! $n.in-,;,-.?l
f’! . tJv’~ tJ.~---LL . .. .“...,’-..”.,
APPENDIX A
q= volume rate of flow across a cross-section
of the fracture
NUMERICAL SOLUTION
qt = fluid-loss rate per unit length of fracture
qi = constant flow rate into fracture at well Here we derive a finite-difference analog of Eqs.
c–
d— ..- . t,,a.
,,”. . ..”1..,....... -.--;.,-vc =.0..-0.
L“nnny. caa.
. . 6.. -.... - ,.In”aL
Uc=a “n’ LL-QLLALL y.=.
12 through 16 and give a numerical method of
before fracturing solution. In order to obtain finite-difference
equations which satisfy an over-all fluid balance
t= time
condition, we begin the derivation with the
tD = dimensionless t continuity equation (Eq. 2) together with Eqs. 8, !3
fft = fiuid-ioss rate per unit surface area and 11, i.e.,
v= fluid velocity
w. fracture width at center of fracture (wm,ax) aq 2hC 3A
WD . dimensionless W + —= o, O<X<L
W/. fracture width ~+~ ~’
. . . . . . . . . . . . . . (A-1)
X,y, z = rectangular Cartesian coordinates (Fig. 1)
XD . dimensionless x
q(o, t) = qi . . . . . . . . . . (A-2)
A#J = pressure in fracture minus S
p. viscosity of fracturing fluid Integration of Eq. A-1 with respect to time from tm
v. bulk Poisson’s ratio of formation to tn+~ with use of the trapezoidal integration
T(x) = time at which fluid loss begins at distance formula gives
x from well
Atm a
REFERENCES —. (q”l + q’)
1. Howard, G. C. and Fast, C. R.: Hydraulic Fracturing, 2 ax
Monograph Series, Society of Petroleum Engineers, t8+l .(A-3)
+A]tm =07””
Dallas (1970) Vol. 11,
2. Carter, R. D.: “Derivation of the General Equation where
for Estimating the Extent of the Fractured Area, ”
Appendix to: “Optimum Fluid Characteristics for
Fracture Extension, ” by G. C. Howard and C. R. Atm = tm+l - ‘, , qQ = q(x, tn)
Fast, Drill, and Prod. Prac., API (1957) 261-270.
3. Khristianovic, S. A. and Zheltov, Y. P.: “Formation Integration of Eq. A-3 with respect to ~ from xi~A
of Vertical Fractures by Means of Highly Viscous to xi+y, with use of Eq. 6 and the central value for
Liquid, “ Proc., Fourth WorldPet. Cong., Rome (1955) the integral gives
Sec. 2, 579-586.
4. Baron, G. et al.: ,, Fra~turatiOn hydraulique; bases
~h& -. -=---, --. --. —
Clrirlllr-s. Iahoratorie.
-- —..- . . . . . . essais
6st11rles sur &

champ, “ Proc., Seventh World Pet. Cong., Mexico

n: ... ,.,-.c-f. “-- -1 a?.
Ll[y {lYOJJ Dec. a, al 1.

5. Geertsma, J. and de Klerk, F.: “A Rapid Method of

Predicting Width and Extent of Hydraulically Induced
rr.m-+,..ae
. ,.z,_Lu, =c, , ,, ~1, -p=:. ~ec,~. fnc..-
\- ._,.., 1060)
A>”<, 1<71-I<Q1
A“. .– .-u..

6. Perkins, T. K. and Kern, L. R.: “Widths of Hydraulic +

Fracture,” J, Pet. Tech (Sept., 1961) 937-949.
7. Kern, L. R., Perkins, T. K. and Wyant, R. E.:
“Designing Aluminum-Pellet Fracturing Treatments:’
Drill. and Prod. Prac., API (1961).
m--,g,d,,
% U,, --~ u, ‘4. F.. ~~Ld Gi~~il, A. u
-.: tfc,.-a
UULL.= T!+.c=
Dimensional Punch and Crack Problems in Classical
Elasticity, ” Proc., Cambridge Phil. Sot. (1963) Vol.
59, 489.
9. Geertsma. T.: “Problems of Rock Mechanics in
Petroleum Engineering, ” Proc., First Congress

AUGUST, 1972 ml
 Ax = Xi+l - xi x
2- -b ‘ dx

w: = W(xi,tm) ,
q=qi-zhc

Since there is no flow out of the end of the fracture

J 0 J’-
. . (B-3)

[q(~,t) = @ by ~q. ~-~

L
= :(Xi+* + x,
‘1 dx q~
2 -+) “ .—. . . . .(B-4)

If we take x% = O, then by Eq. A-2, I o ~- 2hC

The solution of Eq. B-4 with Eq. B-2 is

q~=q, . . . . . . . . . . . .. (A-3)
1
~hCx 2 qi t=
and by Eq. 2,
T=— s L= —Jo - .(B-5)
$h
W;= O . . . . . . . . . . . . ..(A-6) () q~

which was obtained previously 2,S*9 in a somewhat

A finite-difference analog of Eq. 5 is
different manner. By Eq. B-5, Eq. B-3 becomes

2 x
q=qil-–sin-lc
[ 1 , C=-
.
‘l-f L
L 1
-(w; )4] . . . . . . . . ..(A-7) . . . . . ... .O . . . . . . .(B-6)

The result of substituting Eq. A-7 in Eq. A-4 is a By Eqs. 5 and B-6 and the condition W(L, t) = O, we
nonlinear system of equations for W~rn+l with W? obtain
considered known. In order to solve this system we
write *

[1
2(1 - v) pq;
m+l W=4 t*@(~) , . .(B-7)
Wi =w; +@ . . . . . . . .. (A-8)
#CGh

and take where

[1
(W:+1)4 ~
(~ )4 + 4(W; )3AW; . . .(A-9) W(x, t)
d(c) = —
On substitution of Eqs. A-6, A-8 and A-9 into Eq. W(o, t)
A-4, we obtain a linear tridiagonal system, which -’X=LC
:- s,. ..:1.. SUl”cu
--1 -.-J L- Awm -.
A= caa.ly 1“1 Al””i ac ~ = ~ ~, -.. . +

l-r
APPENDIX B = [ g sin-l ~ + (1 - ~2)* - -P 1

SOLUTION FOR LARGE FLUID-LOSS RATE L 2 ‘J

Here we assume that the term dA/dt in Eq. 2 is

. . . . . . . . . . . . . . , .(B-8)
$J@g~~!~ -,w?lhara~
-V...y... L. m“,~~lh,
.h - .,{1..U.”-.
..1>
:J– I----
”3=
fate
~~. The function ~ (<) is shown in Fig. 4. By Eqs. 4,
Then by Eqs. 8 and 9, Eq. 2 becomes B-5 and B-7, the maximum fracture width and
pressure at the origin are given by
M 2hC
—=- . .(B-l) &

ax J-’”””””
W(o, t) = 4-[
2(1 - v) pqf
1 4

where r depends on x such that

L @GCh
J
.
~[L(t’)] = t’ , O<t’<t .(B-2)
1 2G3 @ 1+

Ap(O,t) = 4 @ . .(B-9)
Integration of Eq. B-1 with Eq. 11 yields
#(l - v)3Ch5
L -1

912 SOCIETY OF PETROLEUM ENGINEERS JOURNAL

 ,

Having the solution, we can examine the term

.
dA/dt, which was neglected in obtaining it.By Eqs.
6, 8, 9 and B-9, we have

. . . . . . . . . . . . . (B-lo)
-[ 1
%’) =; >
d~
5=.
limg =0
. . . . . . . . . . . . . . .(c-@

Thus, dA/c?t is negligible compared with qe at x with

= O (and probably over most of the fracture) for
sufficiently large t. An equation of the same form
as Eq. B-10 can be derived for

1 4
m=-, n=--- (c-7)
APPENDIX C 5 5

NO FLUID LOSS The function # (~) can be obtained numerically by

Eq. C-6. Alternately, the original problem, Eqs.
In this section we consider the case of no-fluid
C-2, C-4, 12 and 13, can be solved numerically for
flow out of the fracture, i.e.,
W, and then ~ (~) can be obtained from Eqs. C-5
and C-7. By the latter approach, using the method
. . . . . . . . .(C-l)
‘k = o“””” of Appendix A, we find ~ (~) and

Then,
gives
substitution of Eqs. 5, 6 and C-1into Eq. 2
*(E,*) = O at 5* = 1.01

a~ aw *(O) =0.830 .””” ””o(C-8)

. . . . . . . . .(C-2)
K— (w)=—’
# at
where 1
Gq:

H
14
G
. . . . . . . .(C-3) L = 0.68 t~ . .(C-9)
K=
64(1 - ~)~h
(1 - V) ph’

By Eq. 5, the boundary condition, Eq. 11 becomes with

x s
-=- .
[1
a 4qi
.— (F) =—. . . . . . .(C-4)
L 5*
ax nKh
~=~ The function # (flxt “j~/# (0) = W(x, t)/ W(O,t) is
displayed in Fig. 4. By Eqs. 4, C-3, C-5 and C-7,
The nonlinear partial differential equation ( Eq. C-2)
the maximum width and pressure at the origin are
together with Eqs. 12, 13 and C-4 could be soIved
k.. t.- -.. -...: --1 -_-~_J ..t A---- J:.. h
UY LI1=lluIIIcLl&=LlllCL1lUU U1 Il~~CU”Iii n.

In order to minimize computations and provide

insight into the effect of various parameters, a W(O, <) = 2.5 P -“)”qT ~
similarity solution will be obtained. (Solutions of
this type are discussed by Ames.12) We seek a
solution of the form
t+
w = Qq(g)tm , g = @xt” , . .(C-5)

. (c-lo)

AUGUST. 1972 Sls

—
 aw approximation at early time. However, length must
— - t-: . . . . . . . . . . . (C-13) be determined from Eq. C-9 and not Carter’s
formula. 2
at

Thus, at t- O, the term dW/dt dominates the 8C-term APPENDIX E

in Eq. 14 except near x = L. Therefore, the
foregoing solution with C = O should be a good EFFECT OF SPURT LOSS ON
-.’. . ----
approximation LQ ~h~ s~!~~~~fi foi ~ hOat r~a~i uRE LENGTH
sufficiently small times.
We replace Eq. 9 by

c
APPENDIX D

PERKINS AND KERN APPROXIMATION

For ease of comparison, we record here the where V~ is the spurt loss (volume/area) occurring
approximate solutlon of Perkins and Kern, G Upon in~tatltafie~ii~ly ~t i = ~ and 8([ - r) is the Dirac
neglect of fluid-loss and rate of fracture volume delta symbol. We follow the method of Appendix B
change in the continuity equation (Eq. 2), the and obtain instead of Eq. B-4
right-hand side of Eq. 14 is zero and integration
with Eqs. 15 and 16 yields

(+ )]’4.
~=4[(1-:)VqL Eq. E-2 can be solved by the method of Laplace

............... .(D-l)
transform with the result

v,
This equation contains fracture length, L, as an
unknown. Further, the flow rate, q(x), is equal to qi
everywhere including x = L, which is unrealistic.
For comparison, our large-timesolution on
elimination of C between Eqs. B-> and B-7 gives
. . . . .,
—(Lb2t
2C
e=f~ b<

. . . . . . . . . . . .(E-3)
_

1
1) ,

where
1
n~c
b=—
v,
. . . . . . . . . . . . . . . . . .(D-2)
since
where q5(L”) is given by Eq. B-8. Thus, Eq. D-1
differs from Eq. D-2 at x = O by the factor 0.89.
At small times, an approximate solution for L
can be obtained from the over-all continuity
equation obtained by integration of Eq. 14 with C
= O and dW/dt from Eq. D-1. In this manner, we
eb2t erfc b$ =

3-~+””
”)t-’
~

. . . . . . . . . .(E-4)

obtain Eq. C-9 with the numerical factor replaced

the V~ term is negligible in Eq. E-3 for
by 0.60. Further, a correction to Eq. D-1 can be
obtained from Eq. 14 with dW/dt calculated from Eq.
D-1. At x = O the corrected value of W differs from t>>(v,/c)a”....... “**(E-5)
Eq. D-1 by the factor 0.94. Thus, Perkins and Kern’s
result (Eq. D-1) for width in terms of length is a fair which is a small time for many fracturing fluids .1’ 11
***

914 SOCIETY OF PETROLEUM ENGINEERS JOURNAL