IPTC 11718

Semi-analytical Solution for Multiple Layer Reservoir Problems with Multiple Vertical,
Horizontal, Deviated and Fractured Wells
J. Phillip Gilchrist, Abingdon Techno Centre, and Geoff Busswell, Raj Banerjee, Jeff Spath, and R.K. Michael
Thambynayagam, Schlumberger

Copyright 2007, International Petroleum Technology Conference

This paper was prepared for presentation at the International Petroleum Technology
Conference held in Dubai, U.A.E., 46 December 2007.
This paper was selected for presentation by an IPTC Programme Committee following review
of information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the International Petroleum Technology Conference
and are subject to correction by the author(s). The material, as presented, does not necessarily
reflect any position of the International Petroleum Technology Conference, its officers, or
members. Papers presented at IPTC are subject to publication review by Sponsor Society
Committees of IPTC. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the International Petroleum Technology
Conference is prohibited. Permission to reproduce in print is restricted to an abstract of not
more than 300 words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, IPTC, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.

Abstract
We present new semi-analytical solutions to the multiple layer
reservoir problem, both in real-time and Laplace space.
Assuming a vertically stacked system of layers, an analytic
solution within each layer can be derived using a method of
integral transforms. We fully account for crossflow between
layers by coupling these analytic solutions together and
solving Fredholm integral equations to obtain the flux field at
the layer interfaces. The time evolution of these fluxes is
governed by a Volterra integral equation. Also, the form of
our equations allows us to stop a model execution and then
restart from the exact terminated state.
Our solutions are applicable to partially penetrating vertical,
horizontal, deviated and fractured wells and take into account
superposition effects in multi-well and multi-rate scenarios.
Notably, regarding fractured wells, we are able to accurately
model the case of a finite conductivity fracture with nonDarcy flow as well as those of infinite conductivity. Also,
using solutions in Laplace space we are able to model
naturally fractured reservoirs, wellbore storage, non-Darcy Dfactors as well as constant well pressure production, all within
a full field multi-well scenario.
We present a comparison of our solutions with that generated
using a commercial finite difference simulator for a variety of
problems. We find amazing accuracy with massive gains in
model building and execution times for fracture problems in
particular.
Introduction
Reservoir simulation is an essential tool for the management
of oil and gas reservoirs. Prediction of pressure-production
behaviour under various operating conditions allows, among
other benefits, proper investment decisions to be made. In

order to make such a prediction one must construct a reservoir

model. History matching observed behaviour of the reservoir
must validate the parameters of this model.
Ideally, finite difference numerical simulators are used to
construct reservoir models. However, in order to make full use
of such a tool a large amount of reliable data is required. Also
a full study, including a history-matching analysis, may take
months to carry out. Therefore, there is a need for an
alternative tool that honours the physics of fluid flow and at
the same time offers a solution many orders quicker.
Analytical solutions are fast and provide a broad
understanding of the reservoir dynamics.
The equations applicable to laminar flow of fluids in a porous
medium were the results of Darcys experimental study of the
flow characteristics of sand filters. This combined with the
equation of continuity and an equation of state for slightly
compressible fluid gives the diffusivity equation, which is the
equation for pressure diffusion in a porous medium.
Solution of the diffusivity equation under different boundary
condition forms the basis for prediction of bottom hole
pressure response of a producing well. These analytical
solutions are generally applicable for a single well and used
widely in the area of well testing. The efficiency of analytical
models is generally judged by accuracy and speed.
The application of integral transform techniques to solve
physical problems involving linear partial differential
equations is well known. The theory has been extensively
developed and given by Sneddon1, Churchill2,3 and Tranter4. A
common practice in solving linear partial differential
equations is to use the classical Fourier methods in the spacevariables after removing the time variable by Laplace
transformation. The classical methods often require at their
outset a correct form of the solution that satisfies the
governing differential equation. Hence, with these methods,
solutions are developed to provide answers to specific
problems. The integral transform techniques on the other hand
are direct and can be applied to a wide range of general class
of problems. Titchmarsh5 and Korner6 presented rigorous
mathematical treatment of the theory of integral transforms.
Thambynayagam7 provides practical and elegant solutions to a
varity of configurations thet occur in oild field problems,
including subdivided quadrant layer, octant layer, and cuboids
with mixed boundary conditions. in diffusion by the use of

www.petroman.ir

IPTC 11718

successive integral transforms. In this paper we apply a

portion of Thambynayagams work to the interpretation of
generalized multiple layer, multiple well problem in singlephase hydrocarbon reservoirs. The analytical solutions
presented here are taken directly from Thambynayagam7. In
the case of gas, the partial differential equations have been
linearlized by application of real gas pseudo-pressure as
described by Al-Hussainy et al8. At low pressures linearization
was improved by using Agarwals pseudo-time9 along with
pseudo-pressure.
Formulation of the Problem
We consider a multiple layer reservoir model represented by a
series of N vertically stacked cuboids. The reservoir is
penetrated by multiple wells which can be vertical, horizontal
or deviated. The wells may be fractured or unfractured. A
representation of the reservoir model is presented in Figure 1.

p j ( x, b, z, t)

(4)

= 0

Fluid is able to flow between the layers and we therefore insist

that the pressure and flux are continuous across the layer
interfaces so that:
p j ( x, y , d j , t)
j (x, y,t) =

z
kz j
(5)
p j 1 ( x , y , d j , t )

=

z
k z j 1

p j (x, y, d j ,t) = p

(6)

j 1

( x, y, d j , t)

Assuming a slightly compressible fluid with compressibility

co , the pressure p j ( x, y, z , t ) in layer j can be shown to
satisfy the diffusivity equation:

j ct

(7)

p j

= k xj

2 pj
2x

+ k yj

2 pj
2y

+ k zj

2 pj
2z

j = 0,..., N 1 where ct = co + cR is the total

compressibility of the system assuming a rock compressibility,
cR . See for example Aziz and Settari10 for a derivation of the
diffusivity equation using the assumptions we have outlined.

Figure 1: Multiple layer reservoir with multiple vertical,

horizontal and deviated completions
The reservoir is bounded by the planes passing through x=0,
x=a; y=0, y=b; z=0, z=d. Layer j has porosity j and
permeability

k xj , k yj , k zj in the x , y and z directions

respectively. We first consider the case of a single vertical

well or line source completed in layer j between the
coordinates ( x0 v , y0 v , z01v ) and ( x0 v , y0 v , z02 v ) producing
fluid at constant rate

(2)
(3)

p j (0, y , z , t )
x
p j (a , y , z, t)
x
p j ( x, 0, z, t)

(8)
p j ( x, y, t) =

q for t > t0 . The outer boundaries of

each layer are sealed and we therefore impose no-flow,

Neumann type conditions so that, for t > 0 , they satisfy the
following conditions:
(1)

Single Vertical Partially Penetrating Well

To solve this problem we will utlilize the method of
Thambynayagam7, which is outlined in Busswell11, to
calculate the pressure due to a single, vertical, partially
penetrating well in a single layer reservoir. If in this single
layer reservoir the upper and lower boundaries are assumed to
have arbitrary flux conditions and the outer boundaries have
no-flow conditions then the pressure averaged over the
completion length of the well is given by the expression:

= 0
= 0

= 0

t t0

U (t t0 ) d
2 a bc t

z 02 v z 0 1 v

q (t t0

S x ( x , x 0 v , ) S y ( y , y 0 v , ) S z ( z 0 1 v , z 0 2 v , z 0 1 v , z 02 v , )
+

c t a bd z 02 v z 0 1 v

Bz

( x , y , z 0 1 v , z 02 v , )d + p I

We refer to the first term as the source term which describes

the pressure contribution from the producing line source. The
second term we call the boundary term since it describes the
pressure contribution due to the fluid flow across the
boundaries. Finally the third term, pI , is the initial term

www.petroman.ir

d +

IPTC 11718

which dictates the pressure contribution due to the initial

conditions of the reservoir, which is in this case is assumed to
be uniform everywhere. The source functions,

S x ( x, x0 v , ) , S y ( y, y0 v , ) , S z ( z01v , z02 v , z01v , z02 v , ) ,

and boundary function

z ( x , y , z01v , z02v , ) are defined in

Appendix C of this paper. A detailed description of the

solution derivation of equation (8) is described in Appendix A
of Busswell1.

Note that the function Bz ( x , y , z01v , z02v , ) contains a

double sum over the double Fourier transformed interfacial
flux functions. It is these flux functions that we must solve for
using a numerical scheme. Once these are known Equation 8
can be used for each layer to find the pressure at any point in
space and time.

Deviated Well
This problem also yields a system of integral equations of the
form seen in equation (10). The expressions for the
coefficients Aj , B j , C j and j are different however and
are given in Appendix A
Fractured Well
This problem also yields a system of integral equations of the
form seen in equation (10). The expressions for the
coefficients Aj , B j , C j and j are different however and
are given in Appendix A
Numerical Solution Prescription
For each of the problems so far seen in this paper we showed
that we must solve a system of Volterra integral equations
which have the form:
(11)

We take the cosine Fourier transform of equation (8) in the x

and y directions and then insist that these double transformed
pressures are equal on the layer interfaces so that:

p j (n, m , d j , t) = p

(9)

j 1

j ( n , m , ) A j ( n , m , t ) d
0
t

(n, m , d j , t)

j +1

( n , m , ) B j ( n , m , t ) d

j 1

( n , m , ) C j ( n , m , t ) d

0
t

j = 0,..., N 1 , n = 0,..., and m = 0,..., . By

+
0

doing this we obtain a three-point inhomogeneous recurrence

relationship of the interfacial flux functions in the form:

+ j ( n , m , t ) = 0

(10)

j = 0,..., N 1 , n = 0,..., and m = 0,..., . The

j ( n , m , ) A j ( n , m , t ) d

solution of this system of equations will enable us to obtain

the Fourier components of the flux functions at the layer
interfaces. Depending of whether the problem at hand
involves a vertical, horizontal, deviated or fractured well
determines the form of the coefficients Aj , B j , C j and j .

0
t

j +1

( n , m , ) B j ( n , m , t ) d

j 1

( n , m , ) C j ( n , m , t ) d

0
t

The full expressions for these coefficients are given in

Appendix A for each problem type.

+ j ( n , m , t ) = 0

j = 0,..., N 1 , n = 0,..., and m = 0,..., .

Equation 10 describes a system of Volterra integral equations
in time which we must solve to obtain to obtain the amplitudes
of the Fourier transformed flux functions at each order
n, m and for each of the N layers. The expressions describing
the coefficients Aj , B j , C j and

j are given in Appendix

A. See the numerical solution section for a prescription of how

to solve a system of integral equations of this kind.
Horizontal Well
This problem also yields a system of integral equations of the
form seen in equation (10). The expressions for the
coefficients Aj , B j , C j and j are different however and
are given in Appendix A

In order to solve the system of Volterra Integral equations

given in equation (11) we first approximate the integrals
over by a sum of smaller integrals, where the j are
assumed constant over each interval so that:
(12)
N

i =1

( n , m , )

ti

A j ( n , m , t )d

t i 1
N

+
i =1
N

+
i =1

j +1 ( n , m , )

B j ( n , m , t )d

t i 1
j 1 ( n , m , )

+ j ( n , m , t ) = 0

www.petroman.ir

ti

ti

t i 1

C j ( n , m , t )d

IPTC 11718

j = 0,..., N 1 , n = 0,..., and m = 0,..., . We

solve this in a step by step fashion so for the N th step:

(13)

j ( n , m , t N )

tN

A j ( n , m , t N )d

t N 1

j +1

( n , m , t N )

tN

B j ( n , m , t N )d

t N 1

j 1

( n , m , t N )

tN

C j ( n , m , t N )d

t N 1

N 1

ti

i =1

t i 1

j ( n , m , t i ) A j ( n , m , t i )d
N 1

+
i =1

ti

j + 1 ( n , m , t i ) B j ( n , m , t i )d

i =1

We see a number of possible uses of our semi analytical

model. It could be used for production evaluation, as a
proxy to numerical simulation, integrated with real-time
software and as a generalized well test model.

t i 1
ti

N 1

By coupling together analytic single layer solutions

we were able to construct a system of integral
equations. The form of the coefficients within these
integral equations depends on whether we have a
vertical, horizontal, deviated or fractured well.
A numerical prescription was outlined to solve for the
Fourier components of the flux functions at the layer
interfaces. We then showed how these Fourier
components could be used with the appropriate
analytic expression to find the pressure at a given
well.
We illustrate a series of problems where we show that
our algorithm achieves excellent accuracy with
massive speed gains over numerical finite difference
based simulators.
Combined with a gradient-based optimizer algorithm
our model also proves a powerful tool for performing
very fast history matching studies.

j 1

( n , m , t i ) C j ( n , m , t i )d
t i 1

+ j ( n , m , t ) = 0

j = 0,..., N 1 , n = 0,..., and m = 0,..., , which

describes a linear system of equations where the solution
vector 0 ( n , m , t N ),..., N 1 ( n , m , t N ) for each n, m is
obtained by inverting a tridiagonal matrix.
At each time level, even for varying time step lengths, all the
integrals on the right hand of the system of equations in (13)
can be re-used and only one integral need be calculated afresh
(when i = N 1 ) . The integrals that do need to be calculated
can be performed analytically ensuring extremely fast
computation. The analytic expressions for these types of
integrals are given in Appendix B.
Once the amplitudes of the Fourier transformed Flux functions
have been obtained then the appropriate analytic expression
can be used for any chosen layer to evaluate the pressure at
each well over time. For example, for a vertical well then
equation (8) would be used.
Validation and Application
In order to validate our mathematics and solution method we
have used the ECLIPSE reservoir simulator to build equivalent
numerical models. In the final version of this paper we will
present these comparisons
Conclusion
In this paper we have outlined the mathematics of a
generalized multiple layer, semi analytical model. The
following results were outlined:

Nomenclature
a = reservoir length
b = reservoir width
b f = fracture thickness

C = wellbore storage coefficient

co = fluid compressibility
cR = rock compressibility
ct = total formation (fluid + rock) compressibility
d = reservoir height
k x = permeability in the x -direction
k y = permeability in the y -direction
k z = permeability in the z -direction
k f = fracture permeability
pI = initial well/reservoir pressure
p f ( x, t ) = fracture pressure as a function of position and time

pv ( x, y, t ) = average pressure of a vertical well at point

( x, y ) at time t
pv ( x, y, s ) = Laplace transform of the average pressure of a
vertical well at point ( x, y )
qc ( x, t ) = cumulative fluid flow rate passing point x at time t
qcj (t ) =cumulative fluid flow rate across the interface
between fracture segment j and

j 1

q fjk = flow rate into fracture segment j at time interval k

www.petroman.ir

IPTC 11718

q j = difference between the flow rate at time j and j 1

s = Laplace variable corresponding to elapsed time t
s j = Laplace variable corresponding to elapsed time t t j

x0 p = x coordinate of a point source

x f = fracture half-length
x01h = lower x coordinate of a horizontal well
x02 h = upper x coordinate of a horizontal well
x0 v = x coordinate of a vertical well
y0 v = y coordinate of a vertical well
y0 h = y coordinate of a horizontal well
y0 p = y coordinate of a point source
z0 h = z coordinate of a horizontal well
z01v = lower z coordinate of a vertical well
z02v = upper z coordinate of a vertical well
z0 p = z coordinate of a point source

= non-Darcy Beta factor

= fluid viscosity
= porosity
f = fracture porosity
( x, y, z ) = initial reservoir pressure as a function of position
x = diffusivity constant in the x direction
y = diffusivity constant in the y direction
z = diffusivity constant in the z direction
f = fracture conductivity

fD = dimensionless fracture conductivity

References
1.

Sneddon, I.N.: The Use of Integral Transforms,

McGraw-Hill Book Company, New York (1972).
2. Churchill, R.V.: Extensions of Operational
Mathematics, Proc. Conf. Diff. Eqns., University of
Maryland Bookstore (1955).
3. Churchill, R.V.: Operational Mathematics, second
edition, McGraw-Hill Publishing Company, New York
(1958).
4. Tranter, C.J.: Integral Transforms in Mathematical
Physics, John Wiley & Sons, Inc, New York, (1951).
5. Titchmarsh, E.C.: Fourier Integrals, Clarendon Press,
Oxford, (1962).
6. Korner, T.W.: Fourier Analysis, Cambridge University
Press, Cambridge, (1988).
7. Thambynayagam, R.K.M.: Diffusion: A Compendium of
Analytical Solutions, TBP.
8. Al-Hussainy, R., Ramey, H.J. Jr., Crawford, P.B.: The
Flow of Real Gases Through Porous Media, Trans. SPE
of AIME, 18, (1966) 624.
9. Agarwal, R.G.: Real Gas Pseudo-Time A New
Function For Pressure Buildup Analysis of MHF Gas
Wells, paper SPE 8279 presented at SPE Annual
Technical Conference and Exhibition, Las Vegas,
Nevada, 23-26 Sep 1979.
10. Aziz, K., Settari, A.: Petroleum Reservoir Simulation,
Published by Khalid Aziz and Antonin Settari, (1979).
11. Busswell, G.S., Generalized Analytical Solution for
Single Layer Problems with Multiple Wells and Boundary
Conditions, paper SPE 99288 presented at SPE
Intelligent Energy Conference, Amsterdam, 8-11 April ,
2006

www.petroman.ir

IPTC 11718

Appendix A
Analytic Expressions for Coefficients in the Systems of Integral Equations
In this paper we have seen that we must solve systems of Volterra Integral equations in time in order to obtain Fourier components of
the interfacial flux functions in our multiple layer scenario. In the final version of this paper we will show the exact expressions for
the coefficients Aj , B j , C j and j in equation (11) for each problem type.

Appendix B
Analytic Expressions for Integrals
When performing the time stepping methodology outlined in the Numerical Solution Prescription section of this paper it was
mentioned that the integrals in equation (13) could be performed analytically. In the final version of this paper we will show the exact
expressions for these types of integrals.

Appendix C
Function Definitions
(42)

U (t t0 ) =

1 t > t0

0 t < t0

(43)

erf ( x ) =

u2

du

(44)

S x ( x , ) = 3
( x x0 ), e a
2a

(45)

S y ( y , ) = 3
( y y 0 ), e b
2b

(46)

S z ( z , ) = 3
( z z 0 ), e d
2d

+ 3 ( x + x0 ), e a

2a

+ 3 ( y + y 0 ), e b

2b

+ 3
( z + z 0 ), e d

2d

(47)

S z ( z 0 1 , z 0 2 , z , ) = 3

2d

2d

( z z 01 ), e

z
d

( z z 02 ), e

z
d

z
+ 3
( z + z 02 ), e d

2d

z
d

3
( z + z 01 ), e

2d

www.petroman.ir

IPTC 11718

(48)

S x ( x 01 , x 02 , x 03 , x 04 , ) = 3

+ 3

2a

2a

2a

2a

x
a

( x 04 x01 ), e

x
a

( x 04 x 02 ), e

x
a

( x 03 x01 ), e

x
a

( x 03 x02 ), e

+ 3

+ 3

2a

2a

2a

2a

( x04 + x02 ), e

x
a

( x04 + x01 ), e

x
a

( x03 + x02 ), e

x
a

( x03 + x01 ), e

x
a

(49)

S z ( z 01 , z 02 , z 03 , z 04 , ) = 3
( z 04 z 01 ), e d
2d

2d

2d

( z 04 z 02 ), e

+ 3

( z 04 + z 02 ), e d
2d

+ 3

( z 03 + z 01 ), e d
2d

z
d

( z 03 z 01 ), e

z
d

+ 3
( z 03 z 02 ), e d
2d

2d

2d

( z 04 + z 01 ), e

B x ( x , y , z , ) =

m =0 l=0

x ( t )

( z 03 + z 02 ), e

(
,
,
)

,e a

m
l
ayz
4
2a

x ( t )

(51)

B y ( x , y , z , ) =

n=0

l=0

2
2
n
l ( t )
x +

z
a
d

n x
l z
n l cos
c
o
s

e
a
d

y
(n , l , ) 3
,e b
x0z
2b

y ( t )

y
(n , l , ) 4
,e b
xbz
2b

www.petroman.ir

2
m 2

l
y +
z ( t )
b
d

z
d

m y
l z
m l cos
cos
e
b
d

(
,
,
)

,e a

m
l
0 yz
3
2a

z
d

(50)

y ( t )

IPTC 11718

(52)

B z ( x , y , z , ) =

n=0

2
n 2
m ( t )
x +

y
a
b

n x
m y
n m c o s a c o s b e
m =0

n
m

e
(
,
,
)

,
xy 0
3
2d

( t )

xyd

(n , m , ) 4
,e d
2d

( t )

(53)

B x ( x , y , z 0 1 , z 0 2 , ) =

m l

m =0 l=0

d
l z 0 2
l z 0 1
m y
cos
s in
s in
e
l
b
d
d

m
l

(
,
,
)

,e a
0 yz
3
2a

x ( t )

2
2
m

y +
z ( t )
d
b

m
l

(
,
,
)

,e a
4
ayz
2a

x ( t )

(54)

B y ( x , y , z 0 1 , z 0 2 , ) =

n=0

l=0

2
2
n
l ( t )
x +

z
a
d

d
l z 0 2
l z 0 1
n x
n l
cos
s in
s in
e
l
a
d
d

n
l

(
,
,
)

,e b
x0z
3
2b

y ( t )

n
l

(
,
,
)

,e b
xbz
4
2b

y ( t )

(55)

B z ( x , y , z 0 1 , z 0 2 , ) = 2 d

n=0 m =0

n m

2
2
n
m ( t )
x +

y
a
b

n x
m y
cos
c
o
s

e
a
b

(n , m , ) z
,e d
xy 0
2d

(56)

I x (u , x ) = 3
( x u ), e a
2a

xt

(57)

I y (v, y ) = 3
( y v ), e b
2b

yt

(58)

I z (w, z) = 3
( z w ), e d
2d

( t )

xyd

(n , m , ) z
,e d
2d

+ 3 ( x + u ), e a

2a

xt

+ 3 ( y + v ), e b

2b

yt

zt

+ 3
( z + w ), e d

2d

www.petroman.ir

zt

( t )

IPTC 11718

(59)

I z ( w , z 01 , z 02 ) = 3
( z 02 w ), e d
2d

3
( z 01 w ), e d
2d

0 yz ( m , l , ) =

0 yz

0 0

(m , l , ) =
ayz

ayz

0 0

a d

(62)

(n , l , ) =
x0z

x0 z

0 0

a d

(63)

xbz ( n , l , ) =

xb z

0 0

(n , m , ) =
xy 0

xy 0

a b

xyd

(n , m , ) =

xyd

0 0

0 yz

(m , l, s) =

ayz

(m , l, s) =

l z
m y s
( y , z , ) c o s
cos
e d dzdy
d
b

ayz

0 0 0

a d

(68)

x0z

(n, l, s) =

x0z

0 0 0

a d

(69)

xbz

(n, l, s) =

xbz

0 0 0

l z
n x s
( x , z , ) c o s
cos
e d dzdx
d
a

l z
n x s
( x , z , ) c o s
cos
e d d zd x
d
a

a b

(70)

xy 0

(n, m , s) =

xy 0

0 0 0

a b

(71)

xyd

(n, m , s) =

0 0 0

l z
m y s
( y , z , ) co s
cos
e d dzdy
d
b

0 yz

b d

zt

m y
n x
( x , y , ) c o s
cos
dydx
b
a

0 0 0

(67)

( z 01 + w ), e d
2d

m y
n x
( x , y , ) c o s
cos
dydx
b
a

b d

(66)

l z
n x
( x , z , ) co s
cos
dzdx
d
a

0 0

(65)

zt

zt

l z
n x
( x , z , ) c o s
cos
dzdx
d
a

a b

(64)

( z 02 + w ), e d
2d

l z
m y
( y , z , ) co s
cos
dzdy
d
b

b d

(61)

+ 3

l z
m y
( y , z , ) c o s
cos
d zd y
d
b

b d

(60)

zt

xyd

m y
n x s
( x , y , ) c o s
cos
e d dydx
b
a
m y
n x s
( x , y , ) co s
cos
e d dydx
b
a

www.petroman.ir

10

IPTC 11718

a b d

(72)

l z
m y
n x s
p ( x, y, z, t) cos
cos
cos
e d dzdydx
d
b
a

p (n, m , l, s) =

0 0 0 0

Many of the analytical solutions we present in this paper use elliptical theta functions. We define the basic elliptical theta functions as
in Abramowitz & Stegun24, although the reader should be aware that there are alternative definitions e.g. Spanier & Oldham25.
Despite these functions consisting of infinite sums they converge very rapidly indeed for all parameter input ranges. Our definitions
are as follows:
(73)

3 ( x , e t ) =
2

2
1

c o s ( 2 n x ) e t >

2
2
e t 1
e (x+n) /t

1 + 2 en
n =1

n =

2t

(74)

4 ( x , e t ) =
2

2
1

c o s ( 2 n x ) e t >

2
2
e t 1
e ( x + 0 .5 + n ) / t

1 + 2 ( 1) n e n
n =1

n =

2t

(75)
x

3 ( x , e t ) =

( x ' , e t )d x '
2

2 1
e t >

1
1
x+n
n
t

erf e

erf

2 n = t
t

1 n2 2t
sin(2n x)
e
n =1 n

x+

(76)

4 ( x , e t ) =
2

( x ' , e t )d x '
2

2 1
e t >

x + 0.5 + n
1
0.5 + n 2t 1

erf

erf

2 n =
t
t

2 2
1
(1) n e n t sin(2n x)
n =1 n

x+

(77)
2
3 ( x , e t ) =

( x ' , e 2 t )d x '

x2
1
+ 2
2 2
1
2

n =

n
n =1

e n t [1 cos(2n x) ]
2

2
x+n
( x + n )2 / t
e n / t
+ t e
t

( x + n)erf

www.petroman.ir

2t 1
>
e

1
n
2t

erf n e