COMMUNICATIONS IN NUMERICAL METHODS IN ENGINEERING

Commun. Numer. Meth. Engng 2007; 23:703719

Published online 19 September 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/cnm.921
An efcient Greens function for acoustic waveguide problems
J. A. F. Santiago
1, ,
and L. C. Wrobel
2,
1
COPPE/UFRJ, Programa de Engenharia Civil, Caixa Postal 68506, 21945-970 Rio de Janeiro, Brazil
2
School of Engineering and Design, Brunel University, Uxbridge, Middlesex UB8 3PH, U.K.
SUMMARY
Efcient implementations of the boundary element method for underwater acoustics should employ Greens
functions which directly satisfy the boundary conditions on the free surface and the horizontal parts of the
bottom boundary. In the present work, these Greens functions are constructed by using either eigenfunction
expansions or Ewalds method. This method is discussed in detail, including an attempt to optimize the
value of the parameter b, which splits the integral employed in Ewalds representation.
A numerical analysis of the innite series used in the two-dimensional acoustic waveguide problem is
described, considering the following simplications: the source of acoustic disturbance is time-harmonic,
the velocity of sound is constant and the medium in the absence of perturbations is quiescent. Copyright
q 2006 John Wiley & Sons, Ltd.
Received 10 May 2006; Accepted 31 July 2006
KEY WORDS: shallow water acoustics; wave propagation; Ewalds method
1. INTRODUCTION
A popular alternative to improving the convergence of the series generally used to represent
the Greens function for shallow water acoustics is to construct it in the form of eigenfunction
expansions, the normal mode solution, which involves the sum of an innite number of terms;
however, if the evanescent modes are ignored and only the propagating modes are retained, the
number of terms in the series becomes nite [1]. In spite of that, convergence problems still remain
when the source and eld points are located along the same vertical line [2, 3].

Correspondence to: J. A. F. Santiago, COPPE/UFRJ, Programa de Engenharia Civil, Caixa Postal 68506, 21945-970
Rio de Janeiro, Brazil.

E-mail: [email protected]

E-mail: [email protected]
Copyright q 2006 John Wiley & Sons, Ltd.
704 J. A. F. SANTIAGO AND L. C. WROBEL
Recent papers by Linton [4, 5] and Papanicolaou [6] discuss numerous mathematical techniques
for accelerating slowly convergent series. They show that one powerful technique is the method
of Ewald [7], which is capable of providing dramatic improvements in the speed of convergence.
This method has been successfully implemented in the BEM context by Venakides et al. [8], for
the calculation of electromagnetic scattering of photonic crystals.
In the present article, a two-dimensional model is studied as representative of coastal regions,
which have little variation in the long shore direction. The Ewalds method was derived and imple-
mented for speeding-up the calculations of the eigenfunction expansion of the Greens function.
The accuracy of both forms of the Greens function obtained by the techniques above mentioned
are compared with respect to the number of iterations, particularly close to singularities. The
comparison accounts for source points located at a xed position and eld points moving along
vertical lines.
2. GOVERNING EQUATIONS OF THE PROBLEM
Consider the problem of acoustic wave propagation in a region of innite extent, considering
sections with constant depth and sections with irregular seabed topography, according to
Reference [3].
If the medium in the absence of perturbations is quiescent, the velocity of sound is constant and
the source of acoustic disturbance is time-harmonic, the problem is governed by the Helmholtz
equation [9]

2
+k
2
=
Nes

=1
B

o(E

, S) in (1)
where is the velocity potential, B

is the magnitude of the sound source E

located at (x
e

, y
e

),
S is the source point, located at (u, v), Nes is the number of sound sources, o(E

, S) is the Dirac
delta generalized function and k =c/c is the wavenumber, with c the natural frequency and c the
velocity of sound in the medium.
The problem is subject to the following boundary conditions:
The Dirichlet condition at free surface
F
:
(X) =0 (2a)
The Neumann boundary condition at bottom
B
:
*
*n
(X) =0 (2b)
Sommerfeld radiation condition at innity:
*
*n
(X) =ik(X) (2c)
in which X is the eld point, located at (x, y), n is the outward normal and i =

1.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
EFFICIENT GREENS FUNCTION FOR ACOUSTIC WAVEGUIDE PROBLEMS 705
x
y
free surface
bottom
(
F
)
(
B
)
I1
I2 I 3
B2
1
2
3
4
source

B1

B3

B4
Figure 1. Regions dividing a typical shallow water section.
According to Greens second identity, Equation (1) can be transformed into the following
boundary integral equation [10]:
C(S)(S) =
_

G(S, X)
*
*n
(X) d
_

*G
*n
(S, X)(X) d +
Nes

:=1
B
:
o(E
:
, S) (3)
where is equal to
F

B
and G(S, X) is the Greens function. The function */*n(X) represents
the normal derivative of the velocity potential. The coefcient C(S) depends on the boundary
geometry at the source point S. It is noted that the Greens function implicitly satises the
Sommerfeld condition, therefore no discretization of the boundary at innity is necessary.
General propagation problems with irregular seabed topography can be dealt with by the sub-
regions technique. In this case, the domain is divided into several regions, as depicted in Figure 1.
Two types of subregions can be seen in this gure, for which different fundamental solutions are
employed. In the rst, the region has irregular bottom (
2
and
4
), while the other has constant
depth (
1
and
3
).
Instead of using the fundamental solution of the Helmholtz equation for the free-space, Greens
functions which directly satisfy the boundary conditions either on free surface (
F
) or on free
surface and horizontal parts of the bottom (
B1
and
B3
) are adopted. Therefore, only the irregular
parts of the bottom boundaries (
B2
and
B4
) and interfaces (
I1
,
I2
and
I3
) need to be
discretized.
For the rst subregion type, the Greens functions are developed using the method of images
and have the following form:
G
f
(S, X) =i
1
4
[H
(1)
0
(kr) H
(1)
0
(kr
(1F)
)] (4)
where H
(1)
0
( ) is the Hankel function. The Greens function G
f
(S, X) satises the boundary
condition on the free surface.
For the second subregion type, the Greens functions satisfy the boundary conditions on free
surface and horizontal part of the bottom, and they have the form of innite series, which are
presented in the next section.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
706 J. A. F. SANTIAGO AND L. C. WROBEL
3. GREENS FUNCTIONS
3.1. Eigenfunction expansion
The Greens function G
M
(S, X) exactly satises the boundary conditions on the free surface and
the bottom boundaries. This series, in terms of normal modes, can be written as [9]
G
M
(S, X) =
1
H

m=1
_
sin[k
ym
(Y
F
v)] sin[k
ym
(Y
F
y)]
e

k
xm
|xu|

k
xm
_
(5)
where H is the depth of the free surface (Y
F
Y
B
); Y
F
and Y
B
are the y co-ordinate of the free
surface and bottom, respectively, as can be seen in Figure 2. The parameters k
xm
and k
ym
are
horizontal and vertical wavenumbers, respectively:
k
ym
=
_
m
1
2
_

H
(6a)
k
xm
=k
2
k
2
ym
(6b)
3.2. Ewalds method
3.2.1. Expression obtained for the Greens function. An alternative form of representing
Equation (5) is [11]
G
E
(S, X) =
2
H

m=1
sin[k
ym
(Y
F
v)] sin[k
ym
(Y
F
y)]
_

0
e
k
xm
t (xu)
2
/4t

4t
dt (7)
x
y
free surface
bottom
H
Y
B
Y
F
j

Ij-1

Ij
Figure 2. Geometric constant used in series.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
EFFICIENT GREENS FUNCTION FOR ACOUSTIC WAVEGUIDE PROBLEMS 707
The underlying idea of Ewalds representation is to split the integral in Equation (7) into two
parts [6]
G
E1
(S, X) =
2
H
_
b
2
0
1

4t

m=1
e
k
xm
t (xu)
2
/4t
sin[k
ym
(Y
F
v)] sin[k
ym
(Y
F
y)] dt (8a)
G
E2
(S, X) =
2
H

m=1
sin[k
ym
(Y
F
v)] sin[k
ym
(Y
F
y)]
_

b
2
e
k
xm
t (xu)
2
/4t

4t
dt (8b)
where the parameter b, which divides the integral in Equation (8), is chosen appropriately, taking
account of the position of the source and eld points.
Therefore, the nal expression of Ewalds representation of the Greens function is
G
E
(b, S, X) =
1
4

n=0
C
n
(b, S, X) +

m=1
cos(m)
4

n=0
C
mn
(b, S, X)
+

m=1
sin[k
ym
(Y
F
v)] sin[k
ym
(Y
F
y)]
2H

k
xm

e
(xu)

K
xm
erfc
_
b
_
k
xm
+
(x u)
2b
_
+e
(xu)

K
xm
erfc
_
b
_
k
xm

(x u)
2b
_

(9)
where E
n+1
is the exponential integral function, erfc(z) is the complementary error function and
the coefcients C
n
(b, S, X) and C
mn
(b, S, X) are dened as
C
n
(b, S, X) =
k
2n
b
2n
n!
_
E
n+1
_
a
1
b
2
_
E
n+1
_
a
2
b
2
__
(10a)
C
mn
(b, S, X) =
k
2n
b
2n
n!
_
4

j =1
(1)
j
E
n+1
_
a
j m
b
2
_
_
(10b)
the coefcients a
1
, a
2
and a
j m
being dened in the form
a
1
=
(x u)
2
+(v + y)
2
4
(11a)
a
2
=
(x u)
2
+(Y
F
+v + y)
2
4
(11b)
a
1m
=
(x u)
2
+(2Hm 2Y
F
+v + y)
2
4
(11c)
a
2m
=
(x u)
2
+(2Hm +v y)
2
4
(11d)
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
708 J. A. F. SANTIAGO AND L. C. WROBEL
0.2
0.4
0.6
0.8
1
b
0
2
4
6
8
10
y
-10
-5
0
5
10
C
1
(
y
,

b
)
0.2
0.4
0.6
0.8
b
0 0.2 0.4 0.6 0.8 1
-7.5
-5
-2.5
0
2.5
5
7.5
10
b
y=1
y=3
y=6
y=9
Infinity
Infinity
Figure 3. C
1
(b, S, X) for source point located at position (1, 2) and eld point
moving along the same vertical line.
a
3m
=
(x u)
2
+(2Hm 2Y
F
+v + y)
2
4
(11e)
a
4m
=
(x u)
2
+(2Hm v + y)
2
4
(11f)
The rst and second terms of expression (9), denoted by C
1
(b, S, X)/4 and C
2
(b, S, X),
respectively, are presented in Figures 3 and 4 for a source point located at position (1, 2) and a
eld point moving along the same vertical line, considering the following values:
Y
F
= H =10 m
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
EFFICIENT GREENS FUNCTION FOR ACOUSTIC WAVEGUIDE PROBLEMS 709
0.2
0.4
0.6
0.8
1
b
0
2
4
6
8
10
y
-10
-5
0
5
10
C
2
(
y
,

b
)
0.2
0.4
0.6
0.8
b
Infinity
0 0.2 0.4 0.6 0.8 1
-7.5
-5
-2.5
0
2.5
5
7.5
10
b
y=1
y=2
y=6
y=9
Figure 4. C
2
(b, S, X) for source point located at position (1, 2) and eld point
moving along the same vertical line.
Y
B
=0
x =1 m, 0y10
u =1 m, v =2 m
f =1000 Hz, c =1500 m/s
n =0 to 5, m =1 to 5
A study of the behaviour of these terms is important as they inuence the choice of the parameter
b, since they may assume values that are very close to zero or rapidly tend to innity, even with
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
710 J. A. F. SANTIAGO AND L. C. WROBEL
a few iterations. Therefore, the choice of the value of the parameter b may lead the series to have
a fast rate of convergence, or to diverge.
3.2.2. Function for the parameter b. (a) Source and eld points located on the same verti-
cal. An appropriate function to dene the parameter b was selected based on the following
considerations.
The parameter b should vary between 0 and 1 (0<b<1) to accelerate the decay of the term
containing b
2n
.
The parameter b should have a very small value near the singularity points, to avoid the
slow convergence to zero of the product b

k
xm
in the complementary error function.
This function tends to zero more rapidly as its argument increases. This does not affect the
convergence of the series involving the summation in n, there is, C
n
(b) and C
mn
(b), due to
the rapid decay of the exponential integrals E
n
(z).
For the variation of the eld point X, with co-ordinates (x, y), in a vertical line containing the
xed source point S, with co-ordinates (u, v), b([) can be dened in the form:
b([) =
:
2
[cos([) +1] + (12)
where : and are constants and [ is a y-dependent function, dened in what follows.
For that, the Greens function and the variables r and a
2
can be re-written as
G
E
(b, S
v
, X
v
) =
1
4

n=0
C
n
(b, S
v
, X
v
) +

m=1
cos(m)
4

n=0
C
mn
(b, S
v
, X
v
)
+

m=1
sin[k
ym
(Y
F
v)] sin[k
ym
(Y
F
y)] erfc[b

k
xm
]
H

k
xm
(13)
r =|y v| (14)
a
2
=
(v + y Y
F
)
2
4
(15)
where S
v
and X
v
indicate that source and eld points are located along the same vertical line.
The co-ordinate y of the eld point varies from Y
B
to Y
F
(Y
B
yY
F
). Thus, at most two
singular points may occur at the following positions:
y =v r 0 (16a)
y =Y
F
v a
2
0 (16b)
An analysis of the coefcients a
1
, a
2
and a
j m
showed that only a
2
(Y
F
, S, X) has a singu-
larity at a position different from the above, which appears when the eld and source points
coincide.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
EFFICIENT GREENS FUNCTION FOR ACOUSTIC WAVEGUIDE PROBLEMS 711
y
x
H
Y
B
Y
F
v
YF-v
S
a
0
0
=2
=2
=0
=0
0
2
0
2
x=u
X
y
F B
Y y, v Y
free surface
bottom
Function b ( (y))
Figure 5. Function used as parameter b, considering source and eld points
located on the same vertical line.
The y-dependent function [(y) is depicted in Figure 5, considering the general form (two
singular points) and can be dened as
[ =

1 +
r
[
0
for v [
0
yv +[
0
1 +
a
2
[
0
for Y
F
v 2
_
[
0
yY
F
v +2
_
[
0
0 for y outside the above intervals
(17)
with [
0
= H/0.
In the present paper the constants :, and 0 are taken to be 0.1, 0.000001 and 20, respectively.
The corresponding function b([) obtained with these values is shown in Figure 6.
(b) Source and eld points located on different vertical lines. To verify the inuence of the
position of the eld point X along a vertical line which does not contain the source point S (x =u),
three different positions of point S were considered: v =1m (S11), v =5m (S15) and v =8m (S18).
The co-ordinate u was kept constant at the position u =1 m. The position of point X varies along
a vertical line (Y
B
=0yY
F
=10 m), for the values of x listed in the tables below.
Keeping the source point xed and assuming the eld point moves along vertical lines indicated
by their abscissas, the behaviour of the following variables were studied:
(1) Parameter b: averages of the values producing the lowest errors along a vertical line. Values
from 0.1 to 0.95 in intervals of 0.05 were used.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
712 J. A. F. SANTIAGO AND L. C. WROBEL
0
0.02
0.04
0.06
0.08
0.1
0.12
0 1 2 3 4 5 6 7 8 9 10
y
P
a
r
a
m
e
t
e
r

b

(

(
y
)
)
0
0.02
0.04
0.06
0.08
0.1
0.12
P
a
r
a
m
e
t
e
r

b

(

(
y
)
)
0 1 2 3 4 5 6 7 8 9 10
y
(a)
(b)
Figure 6. Parameter b for H =Y
F
=10 m and Y
B
=0: (a) source and eld points placed at (1, 8) and
(1, y) m, respectively; and (b) source and eld points placed at (1, 5) and (1, y) m, respectively.
(2) Error: highest values along a vertical line.
(3) Number of iterations: average values along vertical lines.
The variable error, employed in this article, refers to the absolute value of the difference between
the Greens functions G
M
(S, X) and G
E
(S, X), obtained at the points S and X.
The mean values of parameter b, the highest values of error and the mean values of number
of iterations are shown in Figures 79, respectively, for source points located at three different
positions and eld points moving along the vertical lines, dened by their abscissas.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
EFFICIENT GREENS FUNCTION FOR ACOUSTIC WAVEGUIDE PROBLEMS 713
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
x
p
a
r
a
m
e
t
e
r

b
S11
S15
S18
Figure 7. Behaviour of the mean value of parameter b for source points located at three different positions
and eld points moving along the vertical lines, dened by their abscissas.
-0.05
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
x
E
r
r
o
r
S11
S15
S18
Figure 8. Behaviour of the error for source points located at three different positions and eld points
moving along the vertical lines, dened by their abscissas.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
714 J. A. F. SANTIAGO AND L. C. WROBEL
0.0
20.0
40.0
60.0
80.0
100.0
120.0
140.0
160.0
180.0
1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
x
N
u
m
b
e
r

o
f

i
t
e
r
a
t
i
o
n
s
Gm S11
Gm S15
Gm S18
Ge S11
Ge S15
Ge S18
Figure 9. Mean value of the number of iterations for source points located at three different positions and
eld points moving along the vertical lines, dened by their abscissas.
Table I. Parameter b for source and eld points on different vertical lines.
Mean Standard deviation
x S11 S15 S18 S11 S15 S18
1, 05 0, 20000 0, 26100 0, 18350 0, 10987 0, 13662 0, 09130
1, 10 0, 21050 0, 25400 0, 17400 0, 12377 0, 13571 0, 07961
1, 20 0, 21600 0, 25600 0, 19000 0, 10822 0, 12498 0, 08762
1, 30 0, 22600 0, 27400 0, 20050 0, 11381 0, 12091 0, 07604
1, 40 0, 24600 0, 30400 0, 22200 0, 10933 0, 13040 0, 07259
1, 60 0, 27100 0, 33050 0, 25950 0, 09723 0, 11479 0, 07167
1, 80 0, 31350 0, 37200 0, 28800 0, 09235 0, 11378 0, 06671
2, 00 0, 33950 0, 39300 0, 31850 0, 09301 0, 10372 0, 06099
3, 00 0, 45900 0, 50200 0, 44250 0, 07500 0, 09015 0, 04994
4, 00 0, 50650 0, 54400 0, 51050 0, 12606 0, 14129 0, 10524
5, 00 0, 50250 0, 56200 0, 50850 0, 13284 0, 15327 0, 12372
6, 00 0, 53700 0, 55800 0, 52050 0, 10240 0, 13812 0, 09926
7, 00 0, 64300 0, 66000 0, 63350 0, 04925 0, 07071 0, 04441
8, 00 0, 84150 0, 84500 0, 82850 0, 05461 0, 05436 0, 07291
9, 00 0, 85800 0, 84750 0, 85600 0, 03674 0, 04227 0, 02954
10, 00 0, 94100 0, 93900 0, 93550 0, 02695 0, 02805 0, 02955
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
EFFICIENT GREENS FUNCTION FOR ACOUSTIC WAVEGUIDE PROBLEMS 715
Table II. Number of iterations for source and eld points on different vertical lines.
Mean Standard deviation
GM GM GE GE GM GM GE GE
x S15 S18 S15 S18 S15 S18 S15 S18
1, 05 274, 93 276, 92 35, 89 43, 16 17, 427 22, 105 11, 639 12, 177
1, 10 158, 76 161, 11 36, 44 43, 34 8, 639 10, 717 11, 497 11, 517
1, 20 92, 49 93, 80 34, 89 39, 75 3, 093 4, 765 9, 819 9, 879
1, 30 67, 72 68, 22 31, 86 35, 74 2, 712 3, 087 6, 937 6, 915
1, 40 54, 79 55, 09 29, 56 32, 68 1, 336 2, 147 5, 238 4, 535
1, 60 41, 43 42, 07 26, 84 28, 36 1, 289 2, 066 3, 116 2, 448
1, 80 34, 58 34, 38 24, 98 26, 08 1, 296 1, 052 1, 990 1, 807
2, 00 30, 37 29, 73 23, 89 24, 45 0, 812 0, 510 1, 363 0, 744
3, 00 22, 78 23, 36 20, 85 20, 63 0, 416 0, 718 0, 435 0, 506
4, 00 20, 87 20, 68 19, 86 19, 97 0, 338 0, 469 0, 876 0, 948
5, 00 19, 97 19, 97 19, 81 20, 01 0, 171 0, 171 1, 269 1, 480
6, 00 19, 85 19, 87 18, 88 18, 92 0, 386 0, 367 0, 327 0, 367
7, 00 19, 14 19, 32 18, 88 18, 88 0, 551 0, 601 0, 327 0, 327
8, 00 18, 88 18, 88 18, 88 18, 88 0, 327 0, 327 0, 327 0, 327
9, 00 18, 88 18, 88 18, 88 18, 90 0, 327 0, 327 0, 327 0, 362
10, 00 18, 88 18, 88 19, 11 19, 23 0, 327 0, 327 0, 737 0, 827
The parameter b along the x-co-ordinate has a similar behaviour for the three positions of the
source points (see Figure 7). The standard deviation of the errors given in Table I shows there
is no dispersion, particularly for values along vertical lines far from the one which contains the
source points (u = 1 m).
It can be seen in Figures 8 and 9 that the error tends to zero and the number of iterations
tends to around 20 rapidly as eld and source points are going away in x-direction. As before,
the standard deviation of the number of iterations given in Table II shows there is no dispersion,
conrming the validity of the analysis.
Thereby, the series obtained by means of Ewalds method should be employed for the vertical
lines close to source points, considering the function b([) presented in the item a. On the other
hand, both Greens functions can be used for vertical lines far from source points.
4. EXAMPLE
A problem of acoustic wave propagation from a region of depth h =10.0 m was studied in order
to verify the performance of the function b([) in the Ewald series. Different situations were
considered, in which the position of the source point S was xed and the eld point X varied
along a vertical line. The sound velocity and frequency are taken to be 1500 m/s and 1000 Hz,
respectively (see Reference [11]).
Figures 1012 show the behaviour of the real part of the functions G
M
[Re(GM)] and
G
E
[Re(GE)], in a region of constant depth of 10.0 m, for source points placed at the xed
positions (1.0, 1.0), (1.0, 5.0) and (1.0, 8.0) m, respectively. The eld points are located along the
vertical line passing through the source point.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
716 J. A. F. SANTIAGO AND L. C. WROBEL
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 1 2 3 4 5 6 7 8 9 10
y
G
r
e
e
n

s

f
u
n
c
t
i
o
n
s
Re(GM)
Re(GE)
Figure 10. Real part of the functions G
M
(S, X) and G
E
(S, X) along a vertical line for S
located at position (1.0, 1.0) m.
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 1 2 3 4 5 7 8 9 10
y
G
r
e
e
n

s

f
u
n
c
t
i
o
n
s
Re(GM)
Re(GE)
6
Figure 11. Real part of the functions G
M
(S, X) and G
E
(S, X) along a vertical line for S
located at position (1.0, 5.0) m.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
EFFICIENT GREENS FUNCTION FOR ACOUSTIC WAVEGUIDE PROBLEMS 717
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 1 2 3 4 5 6 7 8 9 10
y
G
r
e
e
n

s

f
u
n
c
t
i
o
n
s
Re(GM)
Re(GE)
Figure 12. Real part of the functions G
M
(S, X) and G
E
(S, X) along a vertical line for S
located at position (1.0, 8.0) m.
0
5000
10000
15000
20000
25000
0 1 2 3 4 5 6 7 8 9 10
y
N
u
m
b
e
r

o
f

i
t
e
r
a
t
i
o
n
s
NiGM
NiGE
Figure 13. Number of iterations of the functions G
M
(S, X) and G
E
(S, X) along a vertical line for S
located at position (1.0, 5.0) m.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
718 J. A. F. SANTIAGO AND L. C. WROBEL
0
5000
10000
15000
20000
25000
0 1 2 3 4 5 6 7 8 9 10
y
N
u
m
b
e
r

o
f

i
t
e
r
a
t
i
o
n
s
NiGM
NiGE
Figure 14. Number of iterations of the functions G
M
(S, X) and G
E
(S, X) along a vertical line for S
located at position (1.0, 1.0) m.
It is noticed that, for the present case, the real part of the functions G
M
(S, X) and G
E
(S, X)
produced virtually the same results, conrming the validity of Ewalds representation.
Figures 13 and 14 present the number of iterations necessary for the functions G
M
(NiGM) and
G
E
(NiGE) to converge, for a source point S located at the positions (1.0, 1.0) and (1.0, 5.0) m,
and a eld point X moving along the same vertical line in which the source point is located, from
y =0.0 to y =10.0 m. It can be seen that a much faster convergence is produced by the function
G
E
(number of iterations around 60).
5. CONCLUSIONS
The Ewalds method was derived in this paper for speeding-up the calculations of the eigenfunction
expansion of the Greens function for underwater acoustic analysis by the BEM. It was shown
that Ewalds representation is an accurate and efcient method when the source and eld points
are located along the same vertical line or otherwise. The inuence of the parameter b used to
split the innite integral was investigated, as well as the singular integral generated by the innite
series obtained by the Ewalds method.
According to this preliminary study, the proposed function b([) employed as parameter in
the Ewalds method presented a good performance. Higher values of b improve the speed of
convergence of the series for source and eld points placed along the same vertical line, while the
accuracy near singularities is improved by using lower values of b.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm
EFFICIENT GREENS FUNCTION FOR ACOUSTIC WAVEGUIDE PROBLEMS 719
REFERENCES
1. Wu TW. On computational aspects of the boundary element method for acoustic radiation and scattering in a
perfect waveguide. Journal of the Acoustical Society of America 1994; 96:37333743.
2. Dawson TW, Fawcett JA. A boundary integral equation method for acoustic scattering in a waveguide with
nonplanar surfaces. Journal of the Acoustical Society of America 1990; 87:11101125.
3. Santiago JAF, Wrobel LC. A boundary element model for underwater acoustics in shallow waters. Computer
Modeling in Engineering and Science 2000; 1:7380.
4. Linton CM. The Greens function for the two-dimensional Helmholtz equation in periodic domains. Journal of
Engineering Mathematics 1998; 33:377402.
5. Linton CM. Rapidly convergent representations for Greens functions for Laplaces equation. Proceedings of the
Royal Society of London, Series A 1999; 455:17671797.
6. Papanicolaou V. Ewalds method revisited: rapidly convergent series representation of certain Greens functions.
Journal of Computational Analysis and Applications 1999; 1:105114.
7. Ewald PP. Die Berechnung optischer and elektrostatischen Gitterpotentiale. Annalen Der Physik 1921; 64:253268.
8. Venakides S, Haider MA, Papanicolaou V. Boundary integral calculations of two-dimensional electromagnetic
scattering by photonic crystal FabryPerot structures. SIAM Journal on Applied Mathematics 2000; 60:16861706.
9. Kinsler LE, Frey AR, Coppens AB, Sanders JV. Fundamentals of Acoustics (3rd edn). Wiley: New York, 1982.
10. Wrobel LC, Aliabadi MH. The Boundary Element Method. Wiley: Chichester, 2002.
11. Santiago JAF, Wrobel LC. Modied Greens functions for shallow water acoustic wave propagation. Engineering
Analysis with Boundary Elements 2004; 28:13751385.
Copyright q 2006 John Wiley & Sons, Ltd. Commun. Numer. Meth. Engng 2007; 23:703719
DOI: 10.1002/cnm