Professional Documents
Culture Documents
A Plane Multilayered Medium: Theoretical and Numerical Green's Function Field Solution in
A Plane Multilayered Medium: Theoretical and Numerical Green's Function Field Solution in
An explicit generalform is derived for the depth-dependentGreen's function occurringin the integral
solution to the Helmholtz wave equationfor range-independent layered media. This representation
permits arbitrary locationof the sourceand receiver.In addition, a technique,the Fast Field Program
(FFP), for the evaluation of the integral solution is delineated.Examplesof the use of both the
formulationand the FFP to the problemof modelingunderwateracousticpropagationlossversusrange,
where the source/receiverare in air/water, in water/bottom, and in a cross-layersurface duct, are
discussed.
INTRODUCTION
the former is straightforward, since the index of re-
fraction is independent of range. The result for the
Wave motion in a plane stratified medium has been depth-dependent Green's function is complicated due to
studied in many diverse fields. The techniques devel- the piecewise nature assumed for the index of refrac-
oped in one field, however, are often applicable to tion.
others. This is not surprising, since the waves (wheth-
Eigenfunction expansion solutions to this integral
er they are electomagnetic, acoustic, seismic, etc).
have common mathematical denominators. Phenomena havebeenformulated.
1-6 Kutschale
5 givesa complete
wave solution in terms of normal modes and branch
associated with wave propagation in a medium with non-
line integrals for homogeneous
layers, andStickler?
uniform characteristic properties are of particular in-
provides the complete solution when the square of the
terest. Although they are, in general, a function of all
index of refraction in each layer is a linear function of
spatial coordinates and time, the case of arbitrary var-
depth. There are times, however, when it is computa-
iation in only one spatial direction is a sufficiently ac-
tionally easier to deal directly with the integral solu-
curate assumption for many applications.
tion, especially as new numerical integration tech-
It is easy to express the formal solution for the field niques become available. Solutions of this nature re-
produced by a point monochromatic source imbedded in quire the explicit formulation provided in Sec. I. One
sucha mediumas a Fourier-Bessel transformt or, such method is the Fast Field Program (FFP) conceived
equivalently,a Green's functionconvolution.
2 However, by Marsh8 in 1967andextended
by DiNapoli,9whichhas
the explicit general form for the kernel, which results proved useful in applications to underwater acoustic
when the index of refraction is a piecewise continuous propagation. This technique is briefly summarized and
function and, in addition, the source and field point used in conjunction with the results of Sec. I to evaluate
depths are arbitrary, has not appeared in the literature several examples.
to our knowledgeø Previous treatments have considered
either a specific variation in the index of refraction or In Sec. II, the depth-dependent Green's function so-
lution is derived when the index of refraction is a piece-
have used source and field points that were within the
wise continuous function of z. The Green's function is
samelayer. Wait,3 for example,usestheFourier-Bes-
constructed from fundamental solutions of the classical
sel method to generalize the Sommerfield problem to
Sturm-Liouville differential equation with discontinuous
the case of m-homogeneous layers. Harkrider4 and
Kutschale 5 havealso givenintegral solutionsfor this coefficients by use of the matrizant (propagator) method.
case. Felsen andMarcuvitz,2Wait,3 andBrekhovskikh 6
give excellent accounts of the electromagnetic field I. FORMAL SOLUTION
within general stratified media, but their explicit re-
A given stratified medium is approximated by N- 1
sults hold only when the source and field points are
,inhomogeneous layers sandwiched between two homo-
within the same layer.
geneous half-spaces as shown in Fig. 1. A point source
The formal G reen's function integral to the reduced of angular frequencyw [time factor of exp(-iwt) has
wave equation in cylindrical coordinates is given in beensuppressed]is locatedat (0,Zs) in the LSth layer,
Sec. I for the case of piecewise layered media, and ar- where 0•< LS •<N. The field point is located at (r,z) in
bitrary source and receiver depths. The stated for- the LRth layer, where 0•< LR •<N. It is assumed that
malism pertains to the acoustic case but can be used to z >•Zs. When z •<zs, reciprocity is used.
describe other types of wave propagation when the
The Helmholtz reduced wave equation for a point
change in variables describes the characteristic pro-
source located at {s in an inhomogeneous medium is
perties of the medium. The integrand is defined as a
given by
product of a range, r-dependent Bessel function, and a
depth, z-dependent Green's function. The derivation of V2G.--I--
•2eff(g)(.•
= -/.O,),D1
/ 2(z$)S•(•(•'
- •'$). (1)
92 J. Acoust.
Soc.Am.67(1),Jan.1980 92
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
IMPEDANCE
LOCATION
J'•'C• 0(SEMI-INFINITE)
DEPTH
[o(z)
ffz
(o-•
• ) +F(z)-
•"]
O(z,z•;
•)
1 Zl-'O
z2
1 = - i•o(z•)•(z - z•), (•)
where •2 is the separationconstant. G,(r, •) is givenby
iSa/4timesthe HankelfunctionH•)(•r), andG(z,zs;•)
z3 represents the depth-dependent Green's function.
,,.,k
The integral solution to the above boundary value pro-
Z•zZkI:-I zk
z k+l
blem can be given by the Fourier-Bessel transform,
S••• G(z,z•;•)Jo(•r)•d•,
m(r,z,z•)=• (8)
first usedby Sommerfeld
ll andLambix or by Green's
function convolution,
ZN -1
'"'N
_.•_zZ
i-1 N-1 zrl • f•.. G(z,z•;•)n•)(•r)•d•
v(r,z,z•)=4v ' (9)
first describedby Titc•arsh. •
N (SEMI-INFINITE)
FIG. 1. Environmental description of stratified media.
Equations (8) and (9) represent the formal solution to
our problem. Either representation can be used, since
they may be transformed into one another.
The'effective
wavenumber
keff(g
) iSdefined
in termsof There are many ways by which Eqs. (8) and (9) can
the usual wavenumberk(z•=w/sound speedvariation be integrated. We will discuss only three. The first
c(z)] according to and most popular is the so-called residue method. In
this method, Cauchy's residue theorem is used to eval-
- az•.([p(z)]•/•
)' (•)
uate the basic integral. The poles are given by the
zeros of the Wronskian of the independent solutions used
where, for acoustics, p(z) is the density. In electro- to construct G(z,z,; •). By using the residue theorem,
magnetics, p(z) may be either the dielectric parameter the basic integral can be written as a summation over
e(z) or the permeability •(z); in seismology, p(z) is the the residue contributions minus line integrals around
Lam• parameter•-•(z), for SH prop•ation. any branch points. The residue contriubtions yield the
eigenfunction expansion usually referred to as normal
By the simpletransformation
•=•, Eq. (1) may modes. This method has been very valuable in many
be rewritten as '
wave propagation problems;for example, Lambt• and
V•&+k2(z)&-Vp'•/p=-i•p(z• •5(•-L) , (3) SommerfeldTMusedit in their pioneeringworks, and
additionalinsighthas beenprovidedby Brekhovskikh,•
where • represents the pressure at some point r due to Tolstoy,TM
Wait,• and FelsenandMarcuvitz• in their
the point source, and S• is the source strength defined books.
by Morse andIngard.•ø If cylindrical coordinates
(r, •, z) are assumed with azimuthal symmetry, the fol- The multipath expansion technique is another method
lowing equation is obtained: used to evaluate the basic integral solution. Van der
Pol andBremmer •s seem to have beenthe first to study
the expansion
in detail, althoughDebye• hadusedit
far • +p(z) • +•(z)• earlier. Wait • extended the method to allow for caustic
regions and more recently further extensions to the
=-i•p(z•)
2vr
s•(r)•(z- z•). (4) basic work of Van der Pol and Bremmer have been made
by BatorskyandFelsen,•sLeibiger,•sandWeinberg.
xø
The vertical depth coordinate z varies from -• • z The latter workers have also developed computer pro-
+ • •d the range coordinate r varies from 0 < r < •. grams utilizing the multipath expansion.
The boundary conditions imposed on Eq. (4) are that
must satisfy radiation conditions for In the multipath expansion method, the denominator
of G(z,zs; f•) is expanded in a Taylor series which allows
r-• and z- •, (Sa)
for the basic integral to be broken up into a series of
simpler integrals which sometimes can be identified
.
zwp& and & (Sb) with certain ray paths. These ray path integrals are
then generally evaluated by using stationary phase tech-
must be continuous across all interfaces which are
niques or numerical schemes. The multipath method
sumed to be parallel in the r direction. Equation (4) can represents an approximate method when the series is
be separated as truncated, and the effect of layering is only partially
allowed for in the integration.
• +• •(r,•)- • (•) The last method for evaluating the basic integral will
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
be the Fast Field Program (FFP). The FFP was de- satisfied at the source depth zs.
velopedby Marsh andDiNapoli8'9andmakesuseof the The notation used in this article was instituted to con-
Fast Fourier Transform (FFT) to directly integrate
serve spaceø Although compact, it is unorthodox, and
Eq. (8). The first step in this method consists of re-
will be explained whenever it is felt that confusion might
placing the Bessel function in the integral solution of
arise. The symbol U without a subscript or superscript
Eq. (8) with the Hankel function associated with outward
is an abbreviation for the depth-dependent particle ve-
propagation.[The integralinvolvingH•2)(•r) doesnot
locity U=(dF/dz)/[i•op(z)], for z > zs. When it appears
contribute significantly and is neglected. An example
with a subscriptandsuperscript,e.g., U•q the sub-
comparing the results from Eq. (8) with those obtained
script (k) refers to the layer in which F is defined. The
from usingonlyH•1)is providedin Appendix
B, Ex- superscript (k+l) refers to the depth (zkq, see Fig. 1)
ample B2.] If the first term in its asymptotic expan-
at which the operation is evaluated. Thus
sion,
H•l
)(•r)•_
(•.•)l
/2<•.)]/2, (10) k --icopn(z)dz
,
d(•1 d)+q(z)
' where
dz
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
where/4(z,z •) is the fundamentalmatrix (matrizant, The Green's function solution to Eqo (16) can now be
propagator, state transition matrix, etc.) which satis- constructed from the following two matrizant equations:
fies the matrix equation.
dM(z,zt)'-A(z
)]•(z,zi)' (21)
dz (z)/[ioo•(z
) =•(z
)=•O(z
)
suchtha• M(z•,z•)=I, whereI is the identitymatrix. =•(v',v'•)3(z•), (z•-• z < z.),
Whenthe entire interval [z 1,zN] is consideredas a
single layer the matrizant can be calculated by various
techniques, such as Volterra's product integral,
23,24
or (z)/[
iwp(z
) =•(z
)=LU(z
the Peano-Bakerexpansion. 21'25An alternatemethod
for evaluating]•r is to subdividethe total interval[zl,zN] =•(•',•'•') •(z•), (z, < z • z•), (2•)
intoN - 1 subintervals,(zl < z2<... < z• < z•.t <. ßß< z•)
where z• and zn represent the lower and upper bound-
as in Fig. 1o Within each subinterval k(z) is replaced
aries (see Fig. 1), respectively, of •o homogeneous
by a continuousfunctionk•(z) which is a goodapproxi-
half-spaces between which the index of refraction is
mationto k(z) over [zt,zt4]. The functionkt(z) is cho-
low ed be a piecewise continuous function of z. The ini-
sen such that fundamental solutionsf(y •) and g(y •) of
tial values are •hus known up to some muRiplicative
œ(')=0 over [zt,ztq] canbe expressedin terms of
constants, a0 and an, respectively, since
known solutions. The known solutions satisfy a linear
second-order differential equation whose independent
variable 7s[=y(z)] is a functionof z. The matrizant •(z)=ao - -• P(z), z<z
e-•(•q), U(z)-wOo(Z)
M(z [Q(¾
,z•)--M(v•,y
q)--LS•,•'•')P(Y',Y"))]
(V,,Vq)
R(V,,V
q (22) F(z)=ane • ) F(z), z>z
•-'•'• , U(z)-wO•,(z
over somearbitrary subinterval[z,za] may then be ex- where g0 and • are the z components of the wavenum-
(•)
(25)
[••] •,V
=
•) M
o(
b,a
[i •-•
• s•tv•,v•)
' ' +R•(v•,
•= -1
(30)
MdyL¾• •"'•[
, II M•(a
,b
(31)
N-I
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
at theupperandlow- v•.•[• •.•s,(•,•)+n,(•, •)]
wherea =•,•, b=•,•+lare evaluated
er boundaries of the pth layer, respectively. The sym- , LS+ITLS•
bols a, b will be used as abbreviations for the arguments _Ur.
--sS.•r.s.•Sr.
L$sO'r.s
, ' ø (38)
throughout, except where explicitly noted to the con- I •.iS•(b,a).Rr(b,a)]
p--•+l
trary. The continuity of impedance at the boundaries
between zl and z•. is automatically satisfied from the Similarly, if continuity of F. and U. is to hold every-
imposed condition that M(zi,zi) be the identity matrix. where below the source,
•(z)
=M•(•,•q)
'
••*•][O•_l(•_,S,(b,a) + R•(b,a)]• (34)
la
The superscript S denotesthat the evaluation occurs
L•ewise the matrix •(z) below the source may be ex- the source depthz s. One finds that the aboveconditions
pressed. as will be satisfied if the constants are defined according
to
• _ •+1 ' +1 TT•+I
½•(z)=•(y•,y• ) •.•, (•)
U•S*•=a•./W
S+I ,
If continuity
of• and• is toholdever•hereabove Thus G(z ,zs) in any layer can be foundby using these
thesource,then½,(z,)=½,.x(z
•) or, equivalently, constantsin Eqs. (36) and (37). If both the source and
[•.,S,(b,a)+a,(b,a)]=b•
'•. Therelation
tothecon- receiver are in the same layer (i.e., LS = LR), the
stant in the source layer is for 1 • k • LS- 1. Green's functionfor Zr.s•<Z•<zs and 1 •<LS•<N-1 is
(45)
•LS+I -- S
96 J. Acoust.
Soc.Am.,Vol.67, No.1, January
1980 F.R. DiNapoliandR. L. Deavenport:
Green's
functionfieldsolution 96
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
The solutionvalid for zs•< z •<ZLs+•is obtainedby interchangingz and z s in the above equation.
If thereceiveris belowthe sourceandin a differentlayer (i.e., LR>LS), the Gre,en's
functionfor zz•<z •<zz•+•,
LS + I •< LR •< N-1 is
/•+1 ,
G(z
zs)= •'•)+p•(y•
[Z•.•Qm(yiR z •
•'•)][•'•Q•s(yfs
y•]'•)+P•s(y•%,•
'•) • (46)
where
The form of the solution for G provided above is the each of the L equispaced, discrete L values •m(rn -0, 1,
familiar way of expressing the Green's function solu- 2,... ,L - 1) of the horizontal component of the wave-
tion. FelsenandMarcuvitz2 andalso Coddington and number. This calculation represents the major portion
Levinson26providesolutionswhichare identicalinLR form of the required execution timeø Thus the particular
to Eq. (46), except for the absence of the factor manner in which fk,gk depend upon • will have a strong
This factor is absent if both the source and receiver bearing upon the ease with which G can be found as a
are in the same layer (LS =LR), as can be seen by ex- function of •m. Consideration of a few possible choices
amining Eq. (45). Thus the above result represents an for the fundamental solutions will illustrate the point.
extension of this earlier work in that the source and re- Trigonometric
functions:Ck(z)=a•,z•< z•< z•q
ceiver may be in different layers in a piecewise layered
media. The results for either LS or LR equal to zero f •(9'•, • •) = sin9,•, gk(T' • •) = cost f
or N are special cases from a notational standpoint
only. The Green's function for these cases is provided where
in Appendix A.
2- ]'/2.
The Green's function in.the LRth layer may be ex-
pressed in various alternative forms, using the rela- The evaluation of a square root and a trigonometric
tions provided in this section. One such form which function is required for each of the L values of
displays the dependence on the known terminal impe-
dancesZ•: and1• explicitlyis foundto be Airy function' C•(z)=an + bk(zk-z) ,zk•<z •<z•+,
-- Z o)/•1 'rLS'rLR.'LR+
1 f•(v•, •)=A•(T•, gk(v•,•,•)= Ba(V•),
G(z ,zs)= , (47) where
(1_,•,[)K•'-],
[g'[
] I,.1
I
- LSk'•LS
• • S
•TLS
LS+I
C•(Z)=a•f½•-•)/Hk;Z••<z•< z•q, 1•< k•<N-1 ,
where a• is the sound speed at the top of the kth layer,
LR • TLR
which need not equal C•_•(z•), and H• is an arbitrary
scale factor. The fundamental solutions are then found
to be '
•,1•,. (48)
f•(9'•' •(9'• g•(9'•;•) =V'(. (49)
The expression for G(z,Zs) valid above the source is that is, cylindrical functions of complex order
obtained from Eq. (47) by interchanging LS and.LR,
also z s and z. (v•)m-•( •,nHn- i(•H•) ,
III. NUMERICAL CONSIDERATIONS where a represents the value for attenuation in the me-
dium and real argument
The only restriction imposed uponthe matrizant of
Sec. II was that the soundspeed variation Ck(z) associ-
ated with the known solutions be a good approximation 9'•=•; av;
_ '
az--H• (50)
to the given soundspeedvariation within the subinterval
zk•< z•< zaq. The solutionfor the pressure field by The dependence upon • can then be efficiently found by
meansof the FFT [Eq. (14)] necessitatesthe evaluation evaluating Eq. (49) once for a pair of starting values,
of the depth-dependent
Green's function[Eq. (47)] for say (v•)0, (•)•, and utilizing the recurrence relations
97 J.Acoust.
Soc.Am.,Vol.67,No.1,January
1980 F.R.DiNapoli
andR.L.Deavenport:
Green's
function
fieldsolution 97
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
Since the minimum value of • should be zero and we
can showthat G(z,zl; •) decayswhen •> co/Crotn,
where
C,•ta is the minimum sound speed within the region zl
•< z •< z$, we find that the number of sample points L
needed is approximately
L ~ (•/C mi.)I-'Ima.
xß (55)
a5
TypicallyHma•
'• 10.5; hencenot onlydoesthe numberof
calculations increase at high frequencies but also the
=- ++ available core storage may be exceeded. A technique
hasbeenimplemented
28for rigorouslycircumventing
+[(v
+1)•]P,+,,
P• Pb (51) the core storage problem when it arises. Although the
execution time and storage requirements increase with
where p•=i•H•(z•), pb=i•H•p•(z•+l), and the argu- increasing frequency, the solution remains equally val-
id and accurate.
ments of the ma•rizant elements, which have been sup-
pressed for notational convenience, are as in Sec. II, The only remaining point involves the accuracy of re-
a =y , b =y• . placingthe Hankelfunction
H•(•r) withthefirst term in
its asymptotic expansion, which was introduced without
The recurrence relations [Eq. (51)] are stable except justification in the Introduction. A discussion of this
whenthe ]v [ falls betweenthe argumentsa, b. A dis- approximation is provided in Appendix B.
c•sion of their use in this situation is provided by
D• apoli and Powers.•• IV. EXAMPLES
The wavenumber domain sampling distance In this section, the results for the pressure field,
found to be plotted in terms of propagation loss versus range, are
provided for three examples. The source and receiver
(52)
locations have been purposely chosen at unusual depths
u•n noting that the recurrence relations require that in two of the examples to illustrate the generality of
the change in u be unity. The fact that the FFT requires the formalism given in Sec. II. For the third example,
equispacedvalues of •m is in conflict with Eq. (52), the source is in a surface duct; the receiver is below
since H• is different for each of the (N - 1) possible lay- the duct; and a relatively high frequency has been cho-
ers. Then let Af equal the reciprocal of the largest sen to illustrate the general applicability of the FFP
scale [actor algorithm.
k = 1,2,... ,N - 1. (54)
This restriction limits the ability to approximate the 338 1480 1510 1540
-lOO.OO
given sound speed variation arbitrarily dose but leaves
enough flexibility to model most cases of interest in
o.oo
lol .94
--F/ I AIR
I I
•de•ater acoustics. Then if the discrete values of 123.30 •
256.83
are given by
( - + m = 0,1,2,... - 1, 1001.20
l•k•N-1.
1550 1650
The required computer memory for each layer with I _
1799.54
exponential sound speed is then 2p• complex values of
BOTTOM
P(,)• andp• complexvaluesfor bothQ(,)• and
[Thevaluesfor S(,)•canbeobtained
fromEq. (26).] 2099.54
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
TABLE I. Environmental and related FFP values, examples 1-3.
Depth (m) Depth (rn) Sound speed Sound speed Sound speed type
top of bottom (m/s) (m/s) (1--exponential Hk or pk
Layer layer of layer top of layer bottom of layer 2= constant) [Eq. (54)] Density
Table I. The density is assumed constant within each slightly less than that associated with the sound speed
layer; hence, n,•,(•)
eft reduces to the usual depth-dependent in air.
z• : k(z2)
s,, lo'"{'
P•(z)C•(z!i
) -
z 36
k (i5)
• 24 I
A. Example 1--Air to water
12
A 16-Hz cw source was located at -100 m in air
0
(z <Zo, LS=0) andthe receiver was put at 15.2 m (zo ko i i i i i i
•<z •<z•, LR = 1). The pertinentparameters for the o.o.o, .oh.o; .ol .o, .,;, .o,' .,, .,, '.;,
FFP were WAVENUMBER
The solitary peak to the right occurs at a wavenumber FIG. 3. Magnitude of kernel versus wavenumber for example 1.
99 J. Acoust.
Soc.Am.,Vol.67, No.1, January
1980 F.R. DiNapoliandR. L. Deavenport'
Green's
functionfieldsolution 99
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
AIR TO WATER
Z s=-100 rnZ =15.2rn
f= 16Hz
110
150
170 I I I I I
0 15 30 45 60 75 90 105 120
RANGE (km)
situation, except for the high initial loss due to the C. Example 3--Cross layer
mismatch in impedance between air and water.
The source has once again been placed in the surface
B. Example 2-Water to bottom , duct (LS = 1) at zs= 15.2 m, but in this instancethe re-
ceiver is located just below the duct (LR = 2) at z = 110
In this case the 16-Hz cw source is located in the rn; the frequency has been increased to 1000 Hz. The
surface duct zs TM15.2 m, LS = 1. The receiver is placed resulting propagation loss versus range is given in
in the bottom at z = 1850 m, LR = 6 where a sound speed Fig. 7, where the unusually low propagation loss in the
gradient exists. bottom bounce region is due to the unrealistic nature of
the bottom at this frequency.
The distinguishing feature of the kernel versus
wavenumber plot, Fig. 5, is the absence of the peak
correspondingto the wavenumber in air. Figure 6 offers
an examination of the propagation loss versus range v. SUMMARY
for this source and receiver configuration. The figure
An explicit general representation for the depth-de-
reveals no significant difference from what would be
pendent Green's function occurring in the integral solu-
obtained if the receiver were in the water column,
tion to the wave equation has been derived. This rep-
with the exception of the high initial loss at close
resentation permits the monochromatic point source
ranges.
and receiver to be located at any two depths in a plane
multilayered medium. An example of its use to under-
water acoustics is provided by modeling the sound
WATER TO BOTTOM
Z s = 15.2 Z = 1850 rn
speed within the air, water, and bottom with a combi-
4ø
I f=16Hz nation of layers which have either constant or ex-
ponential sound speed variation. In addition, a tech-
nique for the evaluation of the integral solution, the
Fast Field Program (FFP), has been delineated and
applied to obtain answers for the propagation loss ver-
sus range resulting when:
z 24 (a) The source is in the air and the receiver is in
the water;
u. 20
0 (b) The source is in the water and the receiver is in
the bottom;
• 16
z (c) Both source and receiver are in the water in a
(D 12- cross-layered juxtaposition.
ACKNOWLEDGMENT
100 J. Acoust.Soc.Am., Vol. 67, No. 1, January1980 F.R. DiNapoliand R. L. Deavenport:Green'sfunctionfield solution 100
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
40- WATER TO BOTTOM
Zs-15.2mZ =1850m
f= 16Hz
lOO
120[ I I I
0 15 30 45 60 75 90 105 120
RANGE (km)
APPENDIX A: SPECIAL CASES FOR G(z, zs) NOT Case 3. LS=O, LR=N,(N> I)
COVERED BY EQ. (46)
Case 4. LS = LR = O
ceiver in the zero or Nth half-space. These results
are derived in the same manner used to obtain Eq. (46)
but that result does not reduce to the equations below
because of an indexing inconsistency. In each instance,
G(z,z•)=
k •o [ie"•o•s][Z•cos(•oZ)-•
(Z•-2•)/(iwpo) '
Case 5. LS = LR = N
it is assumed that z >zs and the quantities are as de-
fined in Sec. II. .
a(•'•,)=[2•.•cos•(•-•)+
(z•-2•.,)/(i•p•)
•s•(•,-•)]
Case 1. LS = O, I •<LR •<N-1.
L•--'/.,,R
+!Qz•(a, t•)+Pz•(a, t•
fie-,ao
c(z,z•)=L •o•s
1 (z,•_2•)/(iWpo) ß
X •N '
whered=7•s , e=7•s , andb=-•s• in the abovespecial
Case 2. I•LS•N-2,LR=N c•ses.
G(zz,)=• •s Q•s(e
b)+P•s(e,b)]e•"m'"•)Z•a
•
%• LS+z -- • LS
N •S+•
60
FIG. 7. Propagationloss
8O versus range for example 3.
100
15 3O 45 6O 75 9O 105 120
RANGE (km)
101 J. Acoust. Soc. Am., Vol. 67, No. 1, January 1980 F. R. DiNapoli and R. L. Deavenport' Green's function field solution 101
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
transforms was first made by H. W. Marsh in 1967. In and v is a theoretically arbitrary constant, except for
his approach, the Hankel function is replaced by its the restriction that Re(v)> 0. Next, we introduce the
large argument (•r) asymptotic expansion. Small val- Fourier Integral Transform pairs
ues of r correspond to close proximity to the source.
Small values of • may result from either a very low A(z ,z 2; X)= õ(z,z 2; •) e'•"'ud•
frequency or from angles that are close to 90ø, mea-
sured from the horizontal. and (B2)
Recently, Tsang et al.,2ø evidently unaware of the õ(z,z 2; •)= A(z ,z2; X)e•"'udX.
work of Marsh, have proposed a different scheme for
the evaluationof t•esseltransforms,whichalsoutili- The solution for the pressure field is then
zes the FFT algorithm. The advantage of their ap-
proach is that it is exact for all values of •r. The ma-
jor disadvantage is that it requires considerably more (•(r,z
2,z
)=•.•A(z
,z2;X)I(v,
X,r)dX, (B3)
(•(r,z,z2)•ax
(0)(v,.+r2)3/,.
+A [(v+i•LaX)2+r"]
3/ aX [A(naX)I(v,r,naX)+A(L-n•aXP/"I*(
(e•)
where the asterisk denotes the complex conjugate. The depth-dependent Green's function is
Equation (B4) is valid for any r, but for each different
r it must be recalculated.
G(z,z 2; •)= ie•'•2•/2• ,
where
1. Example BI' Inverse solution for direct path (infinite
cw)
102 J. Acoust.Soc.Am., Vol. 67, No. 1, January1980 F.R. DiNapoliand R. L. Deavenport:Green'sfunctionfield solution 102
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
180 DIRECT PATH INVERSE SOLUTION
200
o 60 __•4
, F.FP
I
.[ II millill
II II1 IIIIIIII
i: rx -- 'A' 'h f
• I •, ,• It I1'IIIIIIIII - u. O0
x
--•b•j '• ii I III II IIIIIil
z •,, • • • II ! I]
-50
o_ 20 . x,c, .... t _-__ t-, ,,,,,,,, --
• I I I. 'l' II I IIIIIIl!
400
\ II i 1' ',11
I III I!1111
0 -20
, ,
\
,;
/i
i
,, ,,, ,,,,till[
\ill
k/!
'
' !]1i '•
I tli !
Ill II
]•1 Ilfill]
- -"-'
-150
-20½
<
x
-60
.......
•, !'!
ß • i I ? 'lJ !• IIIIIII
•. '. '.". ILl I!!!!.1![ -- O0 .010 .020 .030 .040 .050 .O60
I '• ' I I I ', [ I , ' ! i I' I IIII!i t
I ', ' ] ' ] ! ] I ; I,'' I ! I!11', HORIZONTAL COMPONENTS OF WAVE NUMBER
-lOO ' ' ' i ! [ I •, [ [ ' ! • I I III11
FIG. B2. Direct path inverse solution.
-14o , ,
-18o , , l,i ,I i, ., If vr> •R , the Fourier Integral Transform for A(z ,zs,X)
0 .01 .02 .03 .04 .05 .06 .07 .08 will not converge. On the other hand, the numerical
HORIZONTAL COMPONENT OF WAVENUMBER results become unstable if v is extremely small. For
FIG. B1. Comparison of FFP and Tsang algorithms for evalu-
this case, a value of v= c•/2 worked nicely. Equation
ation of Eq. (B6). (B4), with r and • interchanged, was evaluatedat •m
= mA• to obtain a comparison at more than one point.
The answers for various intervals of •m are plotted in
ted by boththe FFP technique[Eq. (14)] and the Tsang Fig. B2, along with the exact answers. Excellent
algorithm [Eq. (B4)]. For the FFP evaluation, the agreement is observed for all values of •, including
following input values were used' those close to • = 0 as was expected.
L=8192, ko=0.062831853(ft'•),
2. Example B2: Direct solution with multipath (infinite
cw)
a: 10-qN•
ft '
Af=0.306
79616x 10-4(ft-•),
Both approaches are used to evaluate the received
z-zs:lOOft(30.48m), Ar=25 ft(7.62m), pressure field [Eq. (B1)] for a 35-Hz source locatedat
80 ft (24.38 m), LS= 1 and a receiver at 450 ft (137.16
f=50Hz, •o=A•, ro=Ar.
m), LR--2. A pressurereleasecondition,
•--0, has
The FFP answers are plotted along with the exact been imposed at z•= 0. The pertinent environmental .
answers in Fig. B1. It is evident that excellent agree- and related FFP values are listed in Table B1. The
ment exists everywhere, except for small values parameter values for the FFT are
(•< 0.01) of •. This error is due to the assumption
that the large argumentasymptoticapproximationwas L=8192, A•=2/H•.=0.277... xlO'Sx(2),
reasonable for all values of •. Ar=69.029135, •o=A•, r o=Ar/2.
The same values for the parameters listed above The FFP propagation loss (reference to 1 ft)versus
were used in the evaluation of the Tsang algorithm, range,is shownin Fig. B3. The propagationloss ob-
with the exception that r o and •o were set equal to zero. tained by the Tsang approach wiLh v=200 for Lwo range
The FFT provides results for A(z,zs,nAX) at the points intervals is shown in Figs. B4 and B5, where the FFP
X,: hah, where •= (L•r) '•. The proper selectionfor answers for the same range points have also been pro-
Depth (ft) Depth (ft) Sound speed Sound speed Sound speed type
top of bottom of (ft/s) top (ft/s) bottom (1= exponential Hk or pk
Layer layer layer of layer of layer 2=constant) [Eq. (54)] Density
103 J. Acoust. Soc. Am., Vol. 67, No. 1, January 1980 F. R. DiNapoliand R. L. Deavenport'Green'sfunctionfield solution 103
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
50-
MED FFP
60 FREQUENCY--35
70
• SREU•FEi/•CE•
•)•LpOTCHI__T4Y;•)
Oft40
ft/s,
80
90
lOO
u. 110 _
n- 120
FIG. B3. Propagation loss versus range
o 130 for example B-2/FFP method.
0 3 6 g 12 15 18 21 24 27 30 33 36 39 42 45 48
0
-- 80
DENSITY RATIO=1.5
M--13
0
,• 90 2 LAYERS
0
n- 100
110
120
130
140 I i i i i t i i i 1
48 51 54 57 60 63 66 69 72 75 78 81 84 87 90 93 96
RANGE (KYD)
vided for comparison. between the two solutions. As the range increases, the
two rays merge and a shadow zone develops, which
The results for Fig. B4 start at the value to= 69.029 accounts for the monotonic increases in propagation
ft (21.04 m). The dominant paths, as given by ray
loss. The pattern is interrupted at the range where
theory at the beginning of the range interval, would be
significant bottom bounce energy reaches the receiver.
a direct and a surface-reflected path. The destructive
This energy is associated with steep angles (i.e.,
interference of these two paths is responsible for the
small values of •) and a difference is evident between
null in the pattern at very close ranges. The angles
the two results. This small difference most likely can
associated with these paths are small as measured
be attributed to the introduction of the asymptotic ex-
from the horizontal. Thus, excellent agreement exists
pansion in the FFP approach.
The results in Fig. B5 correspond to a range interval
FFP AND TSANG et aI.(SUMMER MED) where bottom bounce energy predominates. However,
50
70
FFP AND TSANG et aI.(SUMMER MED)
60
80
• 7o
90-
"' 80
100 -
O
.• 90
110 -
lOO
120 -
110. i I i I j I l 130
0 i 2 3 4 5 6 7 23 24 25 26 27 28 29 30
FIG. B4. Propagation loss versus range for example B-2/FFP FIG. B5. Propagation loss versus range for example B-2/FFP
and Tsang methods. and Tsang methods.
104 J. Acoust.
Soc.Am.,Vol.67, No.1, January
1980 F.R. DiNapoliandR. L. Deavenport:
Green's
functionfieldsolution 104
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26
the corresponding values of • would be larger than in crete andBranc• Line Contributions,"J. Acoust.Soc.Am.
the previous example; thus, the agreement between the 57, 856-861 (1975).
two results should be closer. An examination of the 8H. W. Marsh and S. R. Elam, Internal Document, Raytheon
Co., Marine Research Laboratory, New London, CT (1967).
figure reveals that this is the case.
9F. R. DiNapoli, "A Fast Field Program for Multilayered
The execution time (UNIVAC 1108) required for each Media," NUSC Technical Report 4103 (1971).
scheme is summarized below. 10p.M. Morse and K. U. Ingard, Theoretical Acoustics (Mc-
Graw-Hill, New York, 1968), pp. 309-312.
llA. Sommerfeld,"'l•berdie Ausbreitungder Wellenin der
Execution time (s) drahtlosen Telegraphie, "Ann. Physik 28, 665-736 (1909).
Function Tsang et al. FFP 12H.Lamb, "Onthe Propagationof Tremors Over the Surface
of an Elastic Solid," Phil. Trans. R. Soc. London Ser A 203,
Generation of input 101.78 104.64 1-42 (1904).
for FFT (8192)points 13E.C. Titchmarsh, "A Relation Between Green's Functions,"
J. London Math. Soc. 26, 31-36 (1951).
Calculationof 3.41/point 4.23
14I.Tolstoy, WavePropagation(McGraw-Hill, New York,
propagation loss (8192 points)
1973).
Total time 140.6914 x 10 •' 108.87
15B.Van der Pol and H. Bremer, "The Diffraction of Electro-
(4096) points magnetic Wavesfrom an Electrical Point SourceRounda
Finitely ConductingSphere, with Applications to Radiote,le-
The only approximation made in deriving the FFP is graphy arid the theory of the Rainbow," Phil. Mag. 24, 141-
the use of the large argument asymptotic expansion for 176 and 825-863 (1937).
the Hankel,function. This will result in some discrep- 16p.Debye, "Daselektromagnetische
Feld um einen Zylinder
und die Theorie des Regenbogens," Phys. Z. 9, 775-778
ancies at very close ranges to the source for some cw
(1908).
applications, but not for all. Such factors as source
l?j. R. Wait, "A Diffraction Theory for LF Sky-WavePropa-
and receiver depths, sound speed variation, frequency, gation," J. Geophys. Res. 66, 1713-1730 (1961).
and boundary losses have a significant bearing on 18D.V. Batorsky and L. B. Felsen, "Ray-Optical Calculation
whether or not the discrepancy will exist. For very of Modes Excited By Sources and Scatterers in a Weakly In-
close range cw applications, the uncertainty could be homogeneousDuct," Radio Sci. 6, 911-923 (1971).
removed by using the formulation of Sec. II in conjunc- 19G.A. Leibiger andD. Lee, "Applicationof Normal Mode
tion with Eq. (B4). This approach should be taken for Theory to Convergence Zone Propagation, "Vitro Laboratory
Res. Memorandum VL-8512-12-0, Vitro Laboratories, West
close range pulse propagation problems, such as
Orange, NJ (30 November 1968).
multichannel seismic reflection schemes, in which 2øH.Weinberg, "Applicationof Ray Theoryto AcousticPropa-
near-normal incident returns can be separated in time gation in Horizontally Stratified Oceans," J. Acoust. Soc.
from the shallow angle water refracted and reflected Am. 58, 97-109 (1975).
arrivals. 21H.F. Baker, "On the Integrationof Linear Differential Equa-
tions," Proc. London Math. Soc. Set. 1 35, 333-378 (1902).
22R.A. Frazer, W. J. Duncan, andA. R. Collar, Elementary
Matrice• (Cambridge U.P., London, 1960).
1j. E. Freehafer, "Physical Optics," in Propagationof Short 23F.R, Gantmacher, The Theory of Matrices (Chelsea, New
Radio Waves, edited by D. E. Kerr, MIT Radiation Labora- York, 1959), Vol. 2, pp. 125-135.
tory Series (McGraw-Hill, New York, 1951), Vol. 13, pp. 24F. Gilbert and G. E. Backus, "Propagator Matrices in Elas-
58-7O. tic Wave and Vibration Problems," Geophysics 31, 326-332
2L. B. Felsen and N. Marcuvitz, Radiation and Scattering of (1966).
Waves (Prentice-Hall, Englewood Cliffs, NJ, 1973). 25G.Peano,'qnt•gration
parSeriesdes•quations
Diff•rentiel-
3j. R. Wait, ElectromagneticWavesin Stratified Media (Per- les Lin•aires," Math. Ann. 32, 450-456 (1888).
gamon, New York, 1970). 26E.A. Coddington
andN. Levinson,Theoryof OrdinaryDif-
4D. G. Harkrider, ',Theoretical and ObservedAcoustic-Gravity ferential Equations (McGraw-Hill, New York, 1955), pp.
Waves from Explosive Sources in the Atmosphere," J. Geo- 246-252.
phys. Res. 69, 5295-5321 (1964). 27F.R. DiNapoli andM. R. Powers, 'q•ecursive Calculationof
5H. W. Kutschale, "Further Investigation of the Integral Solu- Products of Cylindrical Functions," NUSC Technical Memo-
tion of the Sound Field in Multilayered Media: A Liquid-Solid randum No. PA-83-70 (1970).
Half Space With a Solid Bottom," Technical Report CU-6-71, 28F.R. DiNapoli, "The CollapsedFast Field Program (FFP),"
Lamont-Doherty Geological Observatory of Columbia Uni- NUSC Technical Memorandum No. TAll-317-72 (1972).
versity, Palisades, NY (March 1972). 29L.Tsang, R. Brown, J. A. Kong, and G. Simmons,'•Numeri-
6L. M. Brekhovskikh, Wavesin Layered Media (Academic, cal Evaluation of Electromagnetic Fields Due to Dipole An-
New York, 1960). tennas in the Presence of Stratified Media," J. Geophys. Res.
?D.C. Stickler, '%lormal-Mode Program with Both the Dis- 79, 2077-2080 (1974).
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 138.162.0.39 On: Tue, 10 Nov 2015 16:19:26