You are on page 1of 41

A critical survey of approximate scattering wave

theories from random rough surfaces


Tanos Mikhael Elfouhaily and Charles-Antoines Guerin
Centre National de la Recherche Scientique, IRPHE, Marseille, France
Institut Fresnel, Faculte des Sciences de Saint-Jer ome, Marseille, France
Abstract.
This review is intended to provide a critical and up-to-date survey of the analytical
approximate methods that are encountered in scattering from random rough surfaces.
The underlying principles of the dierent methods are evidenced and the functional
form of the corresponding scattering amplitude or cross-section is given. The reader
is referred to the original papers in order to obtain the explicit expressions of the
coecients and kernels. We have tried to identify the main strengths and weaknesses
of the various theories. We provide synthetic tables of their respective performances,
according to a dozen important requirements a valuable method should meet. Both
scalar acoustic and vector electromagnetic theories are equally addressed.
PACS numbers: 42.25.Fx, 92.10.Hm, 92.10.Cg
CONTENTS 2
Contents
1 Introduction 3
2 Notation and Denitions 3
3 Small Perturbation Method 5
3.1 Denition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2 The historical method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3 The reduced-Rayleigh equations . . . . . . . . . . . . . . . . . . . . . . . 6
3.4 Formal Perturbation Methods . . . . . . . . . . . . . . . . . . . . . . . . 7
3.5 Rigorous derivations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4 Kirchho Approximation 9
4.1 Tangent Plane Approximation . . . . . . . . . . . . . . . . . . . . . . . . 9
4.2 High-frequency regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4.3 Curvature corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.4 Nonlocal corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5 Unifying theories 12
5.1 Meecham-Lysanov Method . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.2 Phase-perturbation method . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.3 Small-Slope Approximation . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.4 Operator Expansion Method . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.5 Tilt-Invariant Approximation . . . . . . . . . . . . . . . . . . . . . . . . 16
5.6 Local Weight Approximation . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.7 Weighted Curvature Approximation . . . . . . . . . . . . . . . . . . . . . 17
5.8 Wiener-Hermite Approach . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.9 Unied Perturbation Expansion . . . . . . . . . . . . . . . . . . . . . . . 18
5.10 Full-Wave Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
5.11 Improved Greens Function Methods . . . . . . . . . . . . . . . . . . . . 20
5.12 Volumic methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
5.13 Integral Equation Method . . . . . . . . . . . . . . . . . . . . . . . . . . 23
6 Connections between functional forms 24
6.1 Local and non-local models . . . . . . . . . . . . . . . . . . . . . . . . . 24
6.2 Gauge transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
6.3 Reduction of nonlocal models . . . . . . . . . . . . . . . . . . . . . . . . 25
7 Synthesis and Conclusion 26
CONTENTS 3
1. Introduction
Wave scattering by rough surfaces is an important issue in diverse areas of science
such as measurements in Medical, Optics, Acoustics, Geophysics, Communications, and
terrestrial or extraterrestrial remote sensing. Approximate models are still a necessity
due to the insurmountable numerical complexity of realistic scattering problems. Even
todays machines cannot cope with the enormous amount of computing demanded
in the case of rigorous numerical calculations of the most general three-dimensional
electromagnetic wave scattering from dielectric or conducting multi-scale surfaces.
This necessity has sprung the development of a myriad of approximate models
with more or less careful derivations from rst principles such as Maxwell equations.
Even alerted users might nd themselves lost in the meanders of more than twenty
approximate models with no or little guidance on which must be applied under what
conditions and for how long it holds until it breaks. The purpose of this review is to
help the practitioner nding his way among this plethora of models through a critical
and, we hope, exhaustive, survey. We have attempted to take a census of the existing
methods and to classify them in generic families. We adopted a strategy of dissecting
all identied scattering models in the literature by providing their functional form along
with their general performances from an analytical point of view. Our readers will
not be overwhelmed by pages of geometric and arithmetic functions describing the
coecients and kernels of each model. The reader is referred back to the original
publications for explicit expression of the kernels. It is worth mentioning at this point of
our contribution that former general reviews can be found in the classical monographs
[1, 2, 3, 4, 5, 6, 7, 8, 9, 10].
Our review is only concerned with the analytical approximate scattering models.
Some other complementary reviews can be found in the recent literature. Saillard
and Sentenac [11] reviewed all rigorous numerical techniques for boundary integral
methods and large linear systems when short-range interactions dominate. Warnick and
Chew [12] listed several numerical approaches on scattering approximations, dierential
equations and surface integral equation methods. More specic reviews can be found
in [13, 14] dealing mainly with two-scale models applied to large rough surfaces such as
the ocean surface.
After introducing the notation and normalization adopted for this paper, we present
the classical approximate models. Then we present recent scattering theories along with
their interconnections before concluding with recapitulative tables.
2. Notation and Denitions
A rough surface separates the vacuum (upper medium) from a homogeneous medium
(lower medium) with dierent index (optical or acoustical). We chose the right Cartesian
coordinate ( x, y, z) system with z-axis directed upward and assume is represented
by a Cartesian equation z = h(r) = h(x, y), where h is the realization of a random
CONTENTS 4
stationary process. For an arbitrary vector a, the notation a will refer to its norm
and a to its direction. The three-dimensional position vector is R = (r, z). A
downward propagating electromagnetic plane wave with wave-vector K
0
= (k
0
, q
0
)
and wavenumber K = 2/ is incident on the surface and give rise to up-going scattered
wave-vectors in directions K = (k, q
k
). The vectors k
0
and k are the horizontal
components of the incident and scattered waves, respectively, and q
0
, q
k
are the vertical
(positive) components. They are related by the relation k
2
+ q
2
k
= k
2
0
+ q
2
0
= K
2
. The
vector Q = K K
0
is the so-called momentum transfer [15] and plays an important
role in scattering theory. We will denote Q
H
= k k
0
and Q
z
= q
k
+ q
0
its horizontal
and vertical components, respectively. Equivalently, we introduce W = K + K
0
which is inspired from Dashen and Wurmser [16]. Analogously we dene its horizontal
W
H
= k +k
0
and vertical W
z
= q
k
q
0
components.
The scattered eld above and far away from the surface is related to the incident
one through the scattering operator which reads in dyadic notation:
E
s
(R) = S(r, z)

E
0
=

e
ikr+iq
k
z
q
k
S(k, k
0
)dk

E
0
, (2.1)
or equivalently at R ,
E
s
(R) = 2i
e
iKR
R
S(k, k
0
)

E
0
, (2.2)
which is a direct consequence of the Weyl representation of the Greens function. Note
that all involved eld quantities are scalar in the acoustic case. In the electromagnetic
case, the scattered eld is usually decomposed over a polarization basis (e.g. circular,
linear horizontal and vertical etc...). Once expressed in the chosen polarization basis,
the tensor S(k, k
0
) becomes a 2x2 matrix called the scattering amplitude (SA, S(k, k
0
)).
Our normalization of the scattering amplitude (SA) diers by a trivial geometrical factor
from some other conventions. For instance, Voronovich [8] uses the prefactor

q
0
/q
k
instead of 1/q
k
in the dening equation (2.1). The scattering amplitude in our denition
and normalization takes the dimension of wavelength (meters).
The scattering amplitude satises some fundamental geometrical properties that
can be used to check the a priori correctness of an approximate method. The reciprocity
expresses the time reversal invariance of the wave in the harmonic regime:
S(k, k
0
) = S
T
(k
0
, k), (2.3)
where the superscript T stands for the transposed dyad (or matrix). The shift invariance
refers to the phase-shifting (delays in the time domain) that results from horizontal and
vertical translations of the surface:
S(k, k
0
)|
h(rd)+D
= e
iQ
H
d
e
iQ
z
D
S(k, k
0
)|
h(r)
(2.4)
The tilt invariance expresses the fact that the scattering amplitude should not depend
on the choice of the reference plane and the related coordinate system. Precisely,
S(k, k
0
)|
R()
= S(

k,

k
0
)|

, (2.5)
CONTENTS 5
where R is any rotation of the surface and the tilted vectors are the horizontal
components of the (inversely) rotated wavenumber vectors R

(K), R

(K

). This
is a very stringent condition and in general an approximate method is only required to
be tilt invariant to rst-order, namely:
S(k, k
0
)|
h+ar
= S(

k,

k
0
)|
h
+O(a
2
), (2.6)
where

k = k+q
k
a and

k
0
= kq
0
a. Given an approximate expression of the scattering
amplitude, it is sometimes impossible to verify the last property analytically, and one
often contents oneself with checking this in the limit of small roughness.
Some more quantities must be dened to better describe the measurables, such as
the scattered power and the ensemble averaged moments in the case of random surfaces.
The rst order moment is also called the coherent scattered amplitude and dened as
V (k, k
0
) = S(k, k

). (2.7)
The incoherent second order or the scattering cross-section of the rough surface is
= |S(k, k

)|

|S(k, k

)|

. (2.8)
The norm and power in the previous equation must be applied to each element of the
SA matrix. For distributed targets (innitely large surfaces of area), a normalized radar
cross-section must be used where the previous denition is normalized by the total
surface area. This latter denition is conventionally called sigma naught (
0
) and used
in remote sensing of rough surfaces such as those of oceans and soils.
3. Small Perturbation Method
The Small Perturbation Method (SPM) is the oldest and perhaps most popular method
in scattering from rough surfaces. It is a perturbative expansion of the scattering
amplitude with respect to a small height parameter (in wavelength units). It was
rst introduced by Rayleigh [17] for sound waves on sinusoidally corrugated surfaces,
then adapted by Fano [18] to optical gratings. Rice [19, 20] obtained explicit rst-
(SPM1) and second-order formulas (SPM2) for 1D conducting surfaces and computed
the rst-order in horizontal polarization for 1D dielectric rough surfaces. Later Peake
[21] derived rst-order backscattering cross-sections for both polarizations. Valenzuela
completed the calculation at second-order for 2D perturbations of levelled [22] and
tilted [23] planes. The range of validity of SPM has been studied in detail via numerical
simulations, essentially in the 2D acoustic case [24, 25, 26, 27] and for Gaussian spectra.
3.1. Denition
The scattering amplitude for SPM is functionally written as a Taylor-Volterra expansion
in surface height (h) as:
S(k, k
0
) =
1
Q
z
B(k, k
0
)(Q
H
) iB(k, k
0
)h(Q
H
)
Q
z

B
2
(k, k
0
; )h(k )h( k
0
)d +. . . , (3.9)
CONTENTS 6
where h() is the Fourier transform of the surface elevation. B(k, k
0
) and B
2
(k, k
0
; )
are the rst and second order coecients, respectively. For explicit expressions and
normalization of these coecients the reader is referred to the monograph by Voronovich
[8]. Our normalization diers slightly from that of Voronovich in that one needs to
multiply by 2q
k
q
0
the SPM1 coecient B and by q
k
q
0
/Q
z
the SPM2 coecient B
2
.
Please do also pay attention to the fact that the expressions of the kernels depends
on the orientation of the vertical axis: some authors (in particular Voronovich) chose
to illuminate the surface from below. Transposition to our convention requires formal
replacements q, q

q, q

and h h. Both B and B


2
have the dimension of
wavenumber squared K
2
in our normalization. Explicit expressions of these coecients
can be found in Elfouhaily et al [28] for Dirichlet, Neumann, perfect conducting, and
dielectric boundary conditions.
3.2. The historical method
Historically, the perturbative expansion to an arbitrary order is termed the Rayleigh
method (also referred to as Rayleigh-Rice or Rayleigh-Fano procedure), which relies on
the Rayleigh hypothesis. It is assumed that the representation of the electromagnetic
eld in terms of solely outgoing waves is still valid on the surface, whereas this holds
a priori only above its maximum excursion. This makes it possible to rewrite the
boundary condition for the eld at the surface in terms of scattering amplitude and
simplies considerably the identication of the perturbative orders. This method has
been used independently by several authors to derive the second-order perturbative
term, which is necessary to account for depolarization eects in the incidence plane
[22], the occurrence of surfaces plasmons [29, 30, 31, 32] and the shift of Brewster angles
[33, 34] in p-polarization. More recently, Voronovich [35, 8] has derived from this method
a compact matricial formulation of SPM2 in the general vectorial dielectric case and
Johnson et al have pushed the calculation to third- [36] and fourth- [37] orders to explore
the backscattering enhancement and the Brewster eect. The analytical formulas at
successive orders become increasingly complicated and dissuasive beyond the second-
order, and attracted, as a matter of fact, little attention in the literature. An ecient
numerical recursive scheme based on the Rayleigh method has been proposed by Bruno
and Reitich [38, 39, 40]. This allows in principle a computation of the perturbation series
at arbitrary order and with the additional use of Pade approximants for an extension
of the perturbative domain and for a gain in accuracy. However, the method seems
well suited to gratings with few Fourier coecients but we are little dubious on its
tractability for 2D random rough surfaces.
3.3. The reduced-Rayleigh equations
An interesting renement of the Rayleigh method is the so-called method of reduced
Rayleigh equations, rst proposed in [41]. The equations obtained via the classical
Rayleigh method couple the unknown scattering amplitudes of the reected and
CONTENTS 7
transmitted waves that propagate in the two homogeneous media surrounding the
surface. Using simple but astute mathematical manipulations, the equations for
the scattering amplitude in reection can be uncoupled, leading to a closed integral
equation expression for the latter. The derivation of the perturbative series becomes
straightforward. Proceeding this way, Soubret et al [42] has found a more compact
formulation than Johnson et al [36] for the third-order scattering amplitude and tailored
the technique to the case of a slab with rough boundary. Reduced Rayleigh equations
have also be used in another context to obtain an expansion in permittivity contrast
rather than surface elevation [43].
3.4. Formal Perturbation Methods
The reduced Rayleigh equations make the scattering amplitude a solution of an integral
equation which is formally similar to the Lippmann-Schwinger equation (i.e a second-
kind integral equation), well-known in potential scattering. This has inspired a new
formulation of the perturbative problem [44, 45] based on the mathematical formalism
of Quantum Mechanics, which guarantees a priori the reciprocity and unitarity of the
scattering operator. This approach allows to use formal perturbation methods based
on diagrammatic techniques [46]. This leads to an improved perturbative scheme of
the coherent and incoherent scattering amplitudes [47], by relying on Dysons and
Bethe-Salpether equations. This statistical approach is sometimes referred to as self-
energy perturbation theory [48] and has been intensively used to study the backscattering
enhancement phenomenon [49, 50], which rst shows up at the fourth perturbative order
in height for the backscattered intensity (and thereby requires the third-order scattering
amplitude) or the shift of Brewster angle for 1D surfaces [51].
Another statistical approach based on formal perturbation methods is the so-called
smoothing method, which has been developed primarily in the context of continuous
random media and discrete scatterers [52, 53], and which can be applied to any random
quantity that satises a second-kind integral equation. The smoothing method consists
in decomposing the eld into a mean and uctuating part, and to nd a second-
kind integral equation for each component, starting from the equation satised by the
deterministic eld. It is equivalent to the diagrammatic method as it leads to the same
Dysons equation for the mean eld [46]. This closed equation implies a complicated
kernel (the so-called mass operator), which is formally given by an innite series, the rst
term of which is usually retained to nd tractable solutions. The resulting approximation
for the mean eld is known as rts-order smoothing, bilocal or Bourret approximation
[54]. This is equivalent to performing the summation of an innite subseries in the
perturbation series of the averaged eld. This perturbative approach has a larger
domain of validity than the conventional one (i.e the one based on iteration, truncation
and averaging of the deterministic perturbation series). The main reason is that only
innite subseries are capable of handling properly secular terms in the pertubation series
(see the excellent discussion by Frisch [46] and e.g. the numerical comparisons in [55]).
CONTENTS 8
The smoothing method has been applied to surface scattering by Wentzel [56], Watson
and Keller [57, 58], Ishimaru [55], Berman and Dacol [59] and Brown [60, 61]. Most
of these studies are concerned with the coherent eld only. Watson and Keller were
the rst to work out the statistical cross-section with the smoothing method, in the
acoustical case:

0
= [1 +A(k, k
0
; )]
1
|B(k, k
0
)|
2
(k k
0
), (3.10)
where is the surface power spectrum, B the SPM1 kernel and A the corrective term
to the SPM1 cross-section. By comparison with the conventional perturbative method,
this formula can be seen to bring higher-order corrections to the SPM1 cross-section.
The kernel is, however, not reciprocal. Berman and Dacol combined their manifestly
reciprocal scattering amplitude formalism with the smoothing method to derive a
similar formula with a reciprocal kernel. The term smoothing approximation is also
employed to denote the rst-term approximation of the intensity operator in Bethe-
Salpether equation, that is found by diagrammatic techniques. Ishimaru et al [55] have
revisited recently the result of Watson-Keller with this approximation and also found a
similar formula with a modied, reciprocal, kernel.
3.5. Rigorous derivations
For a long time, there has been a controversy as to the Rayleigh hypothesis leads to
the correct perturbative expansion. There is, however, a rigorous way to derive this
expansion, that does not resort to any a priori assumption on the eld. It is based on
the so-called Extinction Theorem (also called Extended boundary condition or Ewald-
Oseen), which is a variant of Greens theorem. The method yields, in principle, the exact
height expansion but is much more involved, as witnessed for example by the complexity
of the calculations presented in [5]. It has been used to reinvestigate the second-order
expansion for 1D [62] and 2D conducting [63, 64] surfaces, as well as 1D [41] and 2D
dielectric surfaces [65]. It has also been employed to derive the intensity at fourth order
in height [66, 67] in the scalar Dirichlet case (i.e hard-pressure release acoustic problem)
and for 1D conducting surfaces [26]. The overall conclusion [41, 63, 68, 26] of these
studies is that Rayleigh-Fano procedure and the Extinction Theorem lead to identical
expansions in accuracy, even though the higher-order formulas are sometimes dicult
to compare as they do not have unique expressions. More recently, there has been given
general mathematical arguments [69, 35] asserting that the Rayleigh hypothesis should
yield the correct expansion at arbitrary order (at least for analytic proles), even though
its domain of applicability is very restrained.
Another original approach common to some authors [70, 71, 72, 73] consists in
expanding the eld and its derivatives at the surface in perturbative orders. The
resulting expansion is then injected into the boundary conditions, from which eective
tangential elds [70] (resp. eective surface currents [71, 72]) are deduced. The radiated
eld in space is then found by using Kirchho-Helmholtz formulae [70] (resp. dyadic
CONTENTS 9
Greens functions [71, 72]). This procedure can be used for perturbation of non-planar
(for instance cylindrical) surfaces.
The main limitation of SPM is its restricted domain of validity, as it is valid
for small RMS height/wavelength ratios. In the limit of large wavelengths, however,
this approximation tends to the true solution of the scattering problem. Thus, SPM
constitutes the reference for any approximate method in the low-frequency limit.
4. Kirchho Approximation
Together with SPM, the Kirchho approximation (KA), also known as Tangent Plane
Approximation (TPA) in its low frequency or the Physical Optics (PO) in its high
frequency forms, is the oldest and most employed method. It addresses a complete
dierent scattering regime than SPM, since it is valid for large curvature radii or locally
smooth surfaces [1].
4.1. Tangent Plane Approximation
We refer to the monographs by Ogilvy [7] or by Voronovich [8] for excellent historical
reviews on the Tangent Plane Approximation (TPA). In the Kirchho approximation,
the eld on the surface is assimilated to the eld that would be produced by a tangent
plane at the same point. Thus, it depends only on the Fresnel reection coecient at the
local incidence angle. KA is a local approximation, in that the supposed eld at a point
of a surface does not depend on the surface elsewhere, and thus does not account for
multiple scattering. A second obvious limitation is that it does not consider curvature
eects. Hence this approximation applies a priori to gently undulating surfaces and
incidence angles at which shadowing and multiple scattering eects are negligible. The
KA relies on the physical intuition of the tangent plane. It can, however, be obtained in
a mathematical way. Meecham [74] was the rst to recognize the Kirchho current as
the zeroth iteration in the second-kind integral equation governing the surface current,
under scattering of scalar acoustic waves from soft or hard surface or scattering of vector
electromagnetic waves from perfectly conducting surfaces.
The KA was rst introduced by Brekjovskikh [75, 76] and extended by Isakovich
[77] to the statistical case. It was later treated in English in the acoustic [78] and
electromagnetic [79] cases but the common reference for the western community is the
book by Beckmann and Spizzichino [1], which treats the 2D dielectric and 3D acoustic
problem. There have been several vector formulations of the PO proceeding from the
exact Stratton-Chu equations [80, 81, 82, 83, 84, 85]. The Tangent Plane approximation
(TPA) of KA will also be termed the low-frequency Kirchho approximation (KA-LF).
The TPA takes the following functional form:
S(k, k
0
) =
1
Q
z

K(k
0
; h) e
iQ
z
h(r)
e
iQ
H
r
dr, (4.11)
where the (obviously non-reciprocal) kernel is dependent on the local slope as expected.
CONTENTS 10
It takes the dimension of wavenumber squared K
2
similar to the coecients of SPM in
(3.9).
Rodriguez [15, 86] has proposed a dierent approach to the KA, namely the
Momentum transfer expansion (MTE). He establishes a perturbative series for the
surface current with respect to a novel small parameter Q
H
/Q
z
, which is the ratio
of the horizontal and vertical components of the momentum transfer. KA is recovered
at rst order while the higher orders bring curvature corrections.
4.2. High-frequency regime
The complicated surface-dependence on the surface current via the local Fresnel
reection coecient make statistical formulas for the cross-section complicated in
addition to a deciency in the fundamental reciprocity property. However, resorting to
some reasonable additional assumptions, Stogryn [84] establishes a simplied statistical
formula for the cross-section on Gaussian dielectric surfaces using Taylor expansions
about zero of the correlation function and related quantities. In the 2D acoustic
case, Dacol [87] also extracts the local dependence on the Fresnel coecient via Taylor
expansions about zero. Kodis [85] establishes simplied formula by using the stationary
phase approximation (SPA) before averaging the intensity, while Barrick [88] shows that
the result is independent of the order in which the averaging process and the stationary
phase approximation are applied. The Physical Optics (PO) form of KA is also termed
the high-frequency Kirchho approximation (KA-HF), which simplies to:
S(k, k
0
) = K(k, k
0
)
1
Q
z

e
iQ
z
h(r)
e
iQ
H
r
dr, (4.12)
where the SA now is reciprocal. Hence, by taking the high-frequency limit of TPA both
statistical simplicity and reciprocity property are repaired. In the literature, one might
nd sometimes a small nuance about PO and KA-HF when shadowing is included or
not. We will not make any distinction between KA-HF and PO since shadowing eects
are not presently studied and can be added later on.
In the high-frequency regime (i.e large wavenumber), the KA amplitude reduces
to the probability density function of slopes evaluated at specular points, as was rst
noted by Barrick [89]. This is called the Specular Point Theory (SPT), alternatively the
Geometrical Optics (GO) limit or the Ray-Optics approximation [78, 79, 1, 90, 3]. Its
most common form for single reections is:

0
=

K(k, k

)
Q
2
z

2
P

k k
0
q
k
+q
0

, (4.13)
which is expressed in terms of the normalized radar cross-section where P is the
probability density function of surface slopes. GO has been mainly used for sea surface
scattering [91, 88, 92, 93, 94] and when convoluted with SPM1 it leads to the well-known
Two-Scale Model (TSM) [95, 13, 14] (TSM = SPM1 * GO1).
CONTENTS 11
4.3. Curvature corrections
Lynch [96] uses the variational principle in the acoustical case with the KA as a
trial function to improve the tangent-plane approximation. He obtains local curvature
corrections to the Kirchho surface current. This method has been subsequently referred
to as Lynch Variational Method (LVM), and has been rened and numerically tested
in [97, 98]. Another way to take curvature eects into account is the so-called Local
Parabolic Approximation (LPA), rst introduced by Belobrov and Fuks [99, 100] and
subsequently rened in [101, 102]. It is based on a quadratic, rather than linear,
approximation of the surface height together with an appropriate representation of the
free-space Greens function. The SA in both the LVM and LPA has the functional form
of a KA-LF (4.11).
The radius of curvature and the scattering angles are the sole criterions for the
validity of KA [75, 14, 3, 96] whenever nonlocal eects are negligible. For surfaces with
Gaussian spectra the key parameter is the correlation length, which must remain much
larger than the wavelength [103, 104, 25, 105]. Under multiple scattering conditions,
surface shadowing must be included in the KA approximation in order to ensure energy
conservation, among other things.
4.4. Nonlocal corrections
For large surface slopes and elevations the non-local eects can no longer be neglected. A
possibility to include multiple-scattering phenomena is to consider further iterations of
the surface current equations [74, 106, 107]. However, this results in n-fold integrals that
are dicult to evaluate, whose convergence is questionable and which are not adapted
to numerical computations especially in terms of ensemble averaged formulae.
Originally, Jin and Lax [108, 109] developed the iterative high-frequency KA even
with shadowing eects to second order (KA2-HF). Their model has the general form:
S(k, k
0
) = K
1
(k, k
0
)

e
iQ
z
h(r)
e
iQ
H
r
dr
+

K
+
2
(k, k
0
; ) e
i(k)r
1
i(q
k
+q

)h(r
1
)
e
i(k
0
)x
2
i(q
0
q

)h(r
2
)
ddr
1
dr
2
,
+

2
(k, k
0
; ) e
i(k)r
1
i(q
k
q

)h(r
1
)
e
i(k
0
)x
2
i(q
0
+q

)h(r
2
)
ddr
1
dr
2
, (4.14)
where the second order kernel is a combination of the product of two rst order
Kirchho approximations updated with some shadowing functions. Statistically
averaged shadowing corrections, suggested by Ray-Optics, have been introduced by
Beckmann [110], Wagner [111], and Smith [112] to account for non-local eects. They
have been widely used in conjunction with rst- [113, 114] and second-order KA to
elaborate a multiple scattering theory valid for large elevation RMS [108, 109, 115, 116,
117, 118, 119, 120, 121, 122]. Second-[123] and multiple-[124] reections have also been
considered in this framework of Ray-Optics to explain qualitatively phenomena such as
the mechanism of backscattering enhancement.
CONTENTS 12
Nonlocal corrections are complex from an analytical as well as numerical point of
view and all analytical attempts have been limited to the second-order KA, based on
one iteration of the surface current [116, 9, 125, 126, 127]. General non-local models
will be detailed in later sections but one could mention that the high frequency limit
of these models is expected to reproduce the second-order Geometrical Optics (GO2)
dened as a correction to the GO1 cross-section by:

s=1

K
s

(k, k

, ) P

k
q
k
+sq

,
k

sq

d, (4.15)
where shadowing could be included as an extra weighting function along side the nonlocal
kernel.
For nonlocal models (4.14-4.15) based on a second-order KA, there is an ambiguity
on the integration domain of the spectral variable. A cut-o over evanescent waves (i.e
|| > K) is implicit, even though the spectral integral stems from the Weyl expansion
of the Greens function and should run over an innite domain. The reason for doing
this is that the occurrence of both signs iq

h in the integrand phase would lead to


exponentially increasing terms for evanescent waves (for which q

is positive imaginary)
and cause the integral to diverge. This ad hoc assumption seems to compensate for
another error introduced in removing the cumbersome absolute values |h(r
1
) h(r
2
)|
originally present in (4.14).
5. Unifying theories
5.1. Meecham-Lysanov Method
The Meecham-Lysanov Method (MLM) is the oldest unifying theory and dates back to
the fties. It was suggested independently by Meecham [128] and Lysanov [129], for
the Dirichlet case. The unknown surface current at the surface is sought via its rst
kind (i.e single-layer potential) integral equation. The Weyl representation is used for
the corresponding Greens function, together with a key small-slope assumption: the
distance between two points on the surface is approximated by the horizontal lag. This
turns the surface integral into a convolution that can be Fourier inverted and provides
an approximate solution for the surface current. The resulting expression for the far-
eld scattering amplitude is a triple integral (two space and one frequency variables)
with a simple kernel:
S(k, k
0
) =

(k, k
0
; )e
i(k)r
1
iq
k
h(r
1
)
e
i(k
0
)r
2
iq
0
h(r
2
)
ddr
1
dr
2
. (5.16)
The KA and SPM limit are recovered [8] and non-local eects are taken into
account, owing to the two-fold space integration. Yarovoy [130] has generalized the
method to the 3D pseudo-dielectric case (polarization eects are ignored since the eld
is assumed scalar), using also an integral equation of the rst kind for the surface
elds. However, the inversion procedure that is used in the Meecham-Lysanov method
to estimate the surface current seems not adaptable to the vectorial case, since the
CONTENTS 13
corresponding integral operator is no longer invertible (essentially because it implies a
projection on the tangential direction). As a matter of fact, we are not aware of any
vectorial formulation of the method. Recently, however, a purely numerical approximate
method inspired from Meecham-Lysanov has been proposed, the so-called Small Slope
Integral Equation (SSIE) [131]. Starting from the Magnetic Field Integral Equation,
it makes the same approximation on the Greens function as Meecham-Lysanov. This
transforms the integral operator implying the unknown surface eld into a convolution
operator, that can be implemented eciently via FFT. This numerical method has in
principle the same accuracy as a Meecham-Lysanov procedure and is much faster than
a rigorous computation. The lack of analytical expression, however, prevents the use of
statistical formulae.
5.2. Phase-perturbation method
The Phase-perturbation method (PPM) originates from an idea by Shen and Maradudin
[132] but was developed by Winebrenner [133, 134] for the acoustic case. It relies on an
Ansatz rst proposed by Lynch [96] and Rytov et al [6]: the surface current on the rough
surface is written as a multiplicative (unknown) phase correction to that of a reference
plane. It is then this phase, rather than the eld itself, that is sought in the form of
a perturbative series. The corresponding expansion is found by imposing consistency
with SPM at all orders in the small roughness limit. The PPM functional form is:
S(k, k
0
) = B(k, k
0
)
1
Q
z

e
iQ
z
(k,k
0
;[h])
e
iQ
H
r
e
iQ
z
h(r)
dr (5.17)
where the functional is to rst order in surface height
(k, k
0
; [h]) =

(k, k
0
; )h()e
ir
d + (5.18)
This method has been tested, discussed and compared with other approximations by
several authors [135, 136, 137, 138, 139, 140, 141, 8], essentially in the 2D Dirichlet
case. To rst-order, it has been shown [135] to recover KA and SPM in the appropriate
limits, and to have a wider validity domain than the standard approximations. It is also
appropriate to establish statistical formulas via the use of cumulants [133]. Contrary
to both SPM and KA, PPM fails to satisfy reciprocity. A reciprocal variant of PPM
[142] sharing the same properties was, however, later proposed in the statistical case.
PPM has in our opinion two main drawbacks. First, it is numerically demanding since
it involves a double integral already at rst order. Second, the method seems to have
been worked out in the 2D and/or acoustic case only and we are not aware of a vectorial
formulation.
5.3. Small-Slope Approximation
The Small-Slope Approximation (SSA) was proposed by Voronovich [143, 35, 8, 144] as
a unifying theory that could reconcile SPM and KA. It starts from a structure Ansatz
for the scattering amplitude, that makes it formally similar to a KA integral with an
CONTENTS 14
unknown multiplicative correction of the integrand. This guarantees the geometrical
properties that a scattering amplitude should satisfy a priori. By construction, SSA
respects the proper shifting that follow from horizontal and vertical translation of the
surface (in that it is already superior to SPM), as well as reciprocity. The unknown
correction to a reference plane is sought in a functional Taylor expansion of roughness.
This is done by requiring consistency with SPM at all orders in the limit of small
roughness, together with some intuitive (as it has later turned out, [145, 146]) choice
of gauge to remove certain arbitrariness in the identication of coecients. In practice,
the expansion is performed at the lowest two orders only, for higher terms become too
involved. The functional form of SSA at second order (SSA2) is:
S(k, k
0
) =
1
Q
z

e
iQ
z
h(r)
e
iQ
H
r

B(k, k
0
) iQ
z

M(k, k
0
; ) h()e
ir
d

dr, (5.19)
where:
M(k, k
0
; ) =
1
2
(B
2
(k, k
0
; k ) +B
2
(k, k
0
; k
0
+) B(k, k
0
)) . (5.20)
The rst integral taken alone in (5.19) is the rst-order SSA (SSA1). Here B and B
2
are the rst- and second-order kernels of SPM, respectively.
Contrary to most perturbative approaches, the expansion is not performed with
respect to a well-determined small parameter, but is realized as a functional series
of surface roughness. Dimensional arguments have lead Voronovich to identify the
successive iterations as orders in roughness slopes (whence the terminology). However,
further discussions and numerical tests have clearly demonstrated that the expansion
cannot be recast in powers of a single parameter [147, 148, 149]. SSA at rst- (SSA1)
and second- (SSA2) order has been abundantly discussed and tested against rigorous
methods, essentially in the 2D case [27, 150, 151, 152]. Its qualitative predictions for
scattering on anisotropic sea spectra have been studied [153, 36, 154], but only recently
was it compared to 3D rigorous numerical [148] and experimental [155, 156] results. It
turns out from these experiments that SSA1 considerably extends the validity domain
of SPM, but remains outperformed by KA for large roughness in the domain where this
last method is known to be accurate. SSA2, however, seems to be remarkably accurate
and possesses some desirable properties than SSA1 does not display: it reduces exactly
(in the conducting 3D case) or approximatively (in the dielectric 3D case) to KA in the
high-frequency limit [157], accounts properly for the large-scale tilting eects [158] and
can be incorporated advantageously in a two-scale model [159]. The main limitation
of SSA2 is its numerical complexity. It involves a double integral (one space and one
frequency variable) with an oscillating complex-valued kernel that exhibits branch cuts
and singularities. For this reason SSA1, which is given by a single integral as KA-
HF, remains mostly employed, and 3D numerical implementation of SSA2 have been
achieved only recently [160, 156] in the statistical case.
SSA2 can be used to produce a manifestly reciprocal version of PPM, including the
dielectric vectorial case. This can be done easily by noting that the functional (5.19)
is obtained from the PPM functional (5.17) by mere linearization of the exponential.
CONTENTS 15
This was rst suggested by Berman and Dacol [59] in the 2D Dirichlet case. The same
observation can be used to produce a statistically tractable version of SSA2, by making
the reverse transformation (that is turning SSA2 to a PPM functional form), as was
done for instance in [155].
While Voronovichs procedure proceeds very much from astute guesswork, several
authors have tried rigorous mathematical approaches to SSA. Tatarskii [145, 161]
has developed the Quasi-slope expansion (QSE) and McDaniel [162] the Extended
Small-Slope Approximation (ESSA) in the acoustic case. They proceed from the
extinction theorem or the Meecham-Lysanov method, respectively. Then, they seek
the surface single-potential in a functional Taylor series of height and nd recursively
the coecients. They resort to a small-slope condition at some point to render the
formulas tractable (actually at lowest-two orders). Some a priori assumptions, such
as reciprocity at every order, and gauge arbitrariness are relaxed, but the obtained
approximations are close variants to SSA. ESSA does not reduce to SPM2 due to the
presence of an asymmetric kernel in addition to that of SSA2. Another attempt to
derive SSA from rst principles has been made by Elfouhaily et al [147, 163] in the
3D conducting case. Inspired by an earlier work of Holliday [164] in the backscattering
case, they proceeded through iterations of the surface current equation together with
rened small-slope approximations on the phase of the integrand. They obtained the
scattering eld in a Kirchho functional form, which was only later[149] recognized to
coincide with SSA1.
Even though SSA2 requires a double integration, it does not take spatial non-
local eects into account (this can be seen in the high-frequency limit where it reduces
to the mere rst order KA or GO1). To remedy this limitation, an improvement
of SSA was proposed by Voronovich himself [165], namely the Non-local Small-Slope
Approximation, NLSSA. Its Ansatz is identical to that of Meecham-Lysanov Method
(MLM). This method is in principle superior to SSA, but is dicult to use in practice.
It implies a triple oscillating integral which is unsuited to numerical applications. The
only numerical simulations we are aware of have been performed in the 2D Dirichlet case
for specic spectra [166, 167, 168]. More recently, Elfouhaily et al [28, 169] also adopted
the MLM and NLSSA Ansatz for the Non-local Curvature Approximation (NLCA).
5.4. Operator Expansion Method
The Operator expansion Method (OEM) originated from Hydrodynamics [170, 171, 172]
and was rst adapted to rough surface scattering by Milder for the 2D acoustic problem
[173]. It was subsequently extended to the 3D acoustic [174] case and to the conducting
[175] and dielectric vectorial problems [176]. The principle of OEM is to express the
scattered eld as a Green-Helmholtz surface integral involving the so-called Dirichlet-
to-Neumann Operator (DNO). The DNO relates the value of the eld at the surface
to the value of its normal derivative, one of which (Dirichlet or Neumann problem)
or a combination of which (dielectric case), is prescribed by the boundary condition.
CONTENTS 16
The DNO is explicit for plane waves and can be sought in a functional series of height,
whose successive orders are determined recursively. It is only the DNO, and not the
eld itself, that is expressed pertubatively and thus it is only a partial height expansion.
This makes the validity domain of OEM much wider than SPM. In the 2D acoustic
case [173], OEM has actually been shown to reproduce analytically the SPM and KA
expressions in the appropriate limits. The functional form of OEM is:
S(k, k
0
) =

e
ikriq
k
h(r)
N(k, k
0
; [h]) e
ikriq
0
h(r)
dr, (5.21)
where the kernel operator N(k, k
0
; [h]) is formally expanded in powers of surface
elevation h. In the acoustical case, the OEM kernel N does not depend on the scattering
geometry and thus the scattering amplitude is manifestly reciprocal. This is not the
case under general dielectric conditions.
The OEM iterations can be formally written as a succession of Fourier
multiplicators, that can be eciently implemented by FFT. At the rst level (actually
zero-th order for the DNO), the scattering amplitude is already a non-local triple integral
coinciding with Meecham-Lysanov method (at least in the Dirichlet case). The OEM is
very well adapted to the scalar problem but the vectorial formulation remains obscure
and has been in fact of little used. Numerical experiments have been performed for the
2D [173, 177] and 3D [174] Dirichlet case and for the vectorial case [175, 176]. For 1D
surfaces with Gaussian and Pierson-Moskowitz spectra [177], the OEM at rst-order
appears superior to both KA and SPM, and to extend their added domains of validity.
For sinusoidal gratings [173, 175, 176] it has also been shown to remain accurate for
RMS heights that go well beyond the perturbative domain and in any case superior to
SSA1. Statistical formulas have been given in the 2D acoustic case and for the coherent
eld only [178].
5.5. Tilt-Invariant Approximation
An important property that an approximate method should satisfy to cope with large
scales is the so-called tilt-invariance, as was recalled in the rst section. It is the ability
to treat correctly perturbations of a slightly tilted plane. While this is not satised by
SSA1 and was only checked recently for SSA2 [158], Charnotskii and Tatarskii [179]
constructed a method that would primarily account for the tilting eect, namely the
Tilt-Invariant Approximation (TIA), which they presented for the Dirichlet problem
only. They chose to write the surface current as a multiplicative unknown correction
to the Kirchho current, which is exact for a tilted plane. The unknown function is
sought, as usual, in a perturbative expansion that would be consistent with SPM. The
lowest order of the expansion leads back to KA, while SPM1 is recovered with the rst
two terms:
S(k, k
0
) =
K(k, k
0
)
Q
z

e
iQ
z
h(r)
e
iQ
H
r
dr i

T
1
(k
0
; )h()e
iQ
z
h(r)
e
i(Q
H
)r
ddr
CONTENTS 17
+

T
2
(k
0
;
1
,
2
)h(
1
)h(
2
)e
iQ
z
h(r)
e
i(Q
H

2
)r
d
1
d
2
dr. (5.22)
TIA is however not reciprocal in its correction to KA. Note that the rst two integrals
have the functional form of SSA2. Elfouhaily et al [149] actually demonstrated that
one could operate a simple gauge function change on SSA2 which transforms it into
TIA decomposition. The functional form of TIA motivated the nding of a new general
model for dielectric surfaces which is termed the local curvature approximation LCA
[180, 28]. LCA can be seen as a generalization of TIA to the dielectric case. The issue
of gauge arbitrariness and possible connection between models will be discussed in more
details in a later section.
5.6. Local Weight Approximation
In a series of papers [16, 181, 182], Dashen and Wurmser have developed a novel
formalism for scattering from rough surfaces, which is intermediate between variational
and perturbative methods. They investigated the change of scattering amplitude
resulting from a innitesimal variation of the rough surface itself (and not of a reference
plane) and deduced some interesting properties for the latter. In the 3D acoustic and
conducting case, they showed that, to rst order in curvature, the scattering amplitude
can be written as a low-frequency Kirchho integral with a local kernel, which is the
solution of a dierential equation that is solved explicitly. They consequently obtained
an expression of the same complexity as the KA-LF (or TPA), that takes into account
curvature eects. LWA functional form is:
S(k, k
0
) =
1
Q
z

G(k, k
0
; Q
z
h) e
iQ
z
h(r)
e
iQ
H
r
dr. (5.23)
In addition, this approximation recovers the SPM and KA limits, is reciprocal and
tilt-invariant. The authors did not give a name to their new method, but it has been
later referred to by Milder [173] and subsequently others as Local Weight Approximation
(LWA). The main limitation of LWA is its inapplicability to the dielectric case, since
the dierential equation for the local kernel becomes inextricable (due to the complex
slope dependence of the Fresnel coecients). Albeit powerful, this approach remained
relatively unknown, perhaps because it has been developed by theoretical physicists
outside the radar community.
5.7. Weighted Curvature Approximation
Recently, Elfouhaily et al [180, 28] have derived a model inspired from Dashen and
Wurmser Ansatz, which they named the Weighted Curvature Approximation (WCA).
They arrived at a scattering amplitude which is a correction to SSA1 that has the
structure of LWA, namely a Kirchho integral with a local unknown kernel, and
construct such a kernel with all desirable properties: reciprocity, compatibility with
SPM1 and KA limits and shift- and tilt-invariance. This kernel includes curvature
CONTENTS 18
eects since it is quadratic in its lowest order when expanded about zero. It is based on
a combination of SPM1 and KA kernels, evaluated at local angles. WCA is written as:
S(k, k
0
) =
1
Q
z

{B(k, k
0
) T(k, k
0
; Q
z
h)} e
iQ
z
h(r)
e
iQ
H
r
dr, (5.24)
where
T(k, k
0
; ) = B

k +k
0
+
2
,
k +k
0

k +k
0
+
2
,
k +k
0

. (5.25)
Contrary to LWA, WCA has a dielectric statistical formulation. The method has
been recently tested [183] and has shown satisfactory results.
5.8. Wiener-Hermite Approach
Nakayama (acoustic case) and Eftimiu (electromagnetic case) have used a probabilistic
approach, that we will refer to as Wiener-Hermite Approach (WHA) to the scattering
problem. They consider the scattered eld as a nonlinear stochastic functional of the
random surface, that can be expanded on a basis of orthogonal processes. This is done
via a functional Wiener-Hermite expansion. The unknown deterministic coecients are
then found either by imposing consistency with SPM [184, 185, 186, 187, 188, 189] or
via the surface current equation [190, 191, 192]. Eftimiu has later rened this approach
with an Ansatz similar to TIA: the surface current is written as an amplitude correction
to the Kirchho current and sought in a functional Wiener-Hermite expansion. This
last method has been termed Phased Wiener-Hermite approach by his author [193, 194].
The SPM1 limit is recovered at rst order, but not the KA limit. The diuse scattering
cross-section of WHA can be written as follows:

0
(k, k
0
) = e
Q
2
z
(0)

Q
H
F

(k, k

) +

(k, k

; )()d (5.26)
+

(k, k

)(

)(

)d

,
where (r) and () are the surface autocorrelation function and the surface power
spectrum, respectively. WHA is structurally equivalent to SPM in its statistical
formulation.
WHA is in principle well-adapted to the statistical case as it is readily available
in ensemble averaged form, but in practice the analytical derivation of the Wiener-
Hermite functional expansion and the resulting expressions for the surface currents
become inextricably complicated beyond the rst-order.
5.9. Unied Perturbation Expansion
In the Unied Perturbation Expansion (UPE), Rodriquez et al [195, 196, 197] pursued
the former idea of the momentum transfer expansion (MTE) but changes the small
parameter from Q
H
/Q
z
to W
z
h(r). SPM and KA are recovered with the lowest two
CONTENTS 19
orders of the corresponding perturbative expansion, and SPM2 with the third order.
UPE in our notation becomes:
S(k, k
0
) =
K(k, k
0
)
Q
z

Q
H
i

F
1
(k
0
; )h()e
iQ
z
h(r)
e
i(Q
H
)r
ddr (5.27)
+

F
2
(k
0
;
1
,
2
)h(
1
)h(
2
)e
iQ
z
h(r)
e
i(Q
H

2
)r
d
1
d
2
dr.
This model is not reciprocal by construction since the expanded current is function
of the incident wave alone. Another limitation of UPE is that it is available for perfect
conducting surfaces only.
5.10. Full-Wave Approach
The Full-Wave Approach (FWA) has been introduced and developed over many years by
Bahar in the 2D case [198, 199, 200, 201, 202, 203], then modied to include local slope
dependence and extended to the 3D case [204, 205, 206, 207]. The surface is seen as the
lower boundary of a semi-innite waveguide and the scattered eld is decomposed over
a basis of local modes, with unknown coecients. These local spectral components are
shown to satisfy a dierential equation of propagation along the lateral direction, the
so-called telegraphist equation for cable and waveguide transmissions. Under further
approximations such as a small-slope assumption, the equation can be solved iteratively
by the method of subsequent approximations. In the 2D Dirichlet case, the rst iteration
was recognized by Voronovich [8] to coincide with SSA1. Tractable expressions for
subsequent iterations can only be obtained via stationary phase or the steepest decent
approximations.
Ever since FWA was introduced by Bahar in early seventies, it has been
continuously evolving at a rate of more than one paper a year on average, and it is
therefore dicult to accomplish a thorough review of this model in a short paragraph.
The functional form of FWA can be written as :
S(k, k
0
) =

D
1
(k, k
0
; Q
z
h) e
iQ
z
h(r)
e
iQ
H
r
dr
+

D
+
2
(k, k
0
; , [h])e
i(k)r
1
i(q
k
+q

)h(r
1
)
e
i(k
0
)r
2
i(q
0
q

)h(r
2
)
ddr
1
dr
2
+

2
(k, k
0
; , [h])e
i(k)r
1
i(q
k
q

)h(r
1
)
e
i(k
0
)r
2
i(q
0
+q

)h(r
2
)
ddr
1
dr
2
, (5.28)
which encompasses all published forms of FWA over the years.
The FWA is essentially and originally a 2D method, as it relies on the telegraphist
equations, and the attempts that have been made by the author to extend it to
higher dimensions call on additional simplifying assumptions which are not always
mathematically justied. In an excellent critical review on the FWA, Collin [208, 209]
has given a clean and self-contained derivation of the 3D case and made comparisons
with other approximate methods. FWA is shown to reproduce SPM in the small-
roughness limit, but the KA limit is only recovered via the half-sum of the co-polarization
coecients. FWA has been constantly improved and rened by his author and some 3D
CONTENTS 20
comparisons with experimental data have been made [210], where it is found in better
agreement with data than KA and SPM. However, our subjective opinion is that FWA is
analytically as well as numerically very involved while the resulting gain in accuracy over
other unifying methods is not established. Moreover, the long series of mutually referring
papers and the successions of dierent versions of FWA that have been developed by
Bahar over thirty years render its understanding and numerical implementation dicult
to even advised users.
Two other scattering models were derived in a similar manner to FWA. The rst
is the Local Spectral Expansion Method (LSEM) developed by Garca-Valenzuela and
Collin [211, 212] for 1D surfaces but for both conducting and dielectric cases. Their
rst-order model is formed by the sum of two single integrals. The rst integral is the
high-frequency Kirchho while the second is an original form where W
z
= q
k
q
0
replaces
the standard vertical component of the momentum transfer Q
z
= q
k
+ q
0
. The second
order functional form is similar to that of SSA2 [35] and it helps the model retrieve the
low-frequency SPM2 limit. The expression of the total LSEM is:
S(k, k
0
) =
K(k, k
0
)
Q
z

e
iQ
z
h(x)
e
iQ
H
x
dx +
B(k, k
0
) K(k, k
0
)
W
z


e
iW
z
h(x)
1

e
iQ
H
x
dx
i

F(k
0
; )h()

e
iq
k
h(x)
e
iq
k
h(x)

e
i(Q
H
)x
ddx, (5.29)
where both SPM1 and SPM2 are reached. The Kirchho limit is, however, not formally
reached away from specular or backscatter where W
z
= 0. LSEM fails also some other
fundamental properties such as reciprocity, shift and tilt invariances.
The other method based on a modication of FWA derivations is the Correction
Current Method (CCM) by Schwering et al [213]. The model has a single integral form
similar to the tangent plan approximation (TPA). The kernel of CCM is formed in its
constant part by that of SPM1 in the spirit of SSA1 and FWA. The variable part of the
kernel is however original and depends simply on a quadratic form of the surface slope.
CCM is hence very structurally similar to WCA [28] with the inclusion of less orders of
surface slope. CCM exists only for 1D perfect conducting surfaces and reads:
S(k, k
0
) =
B(k, k
0
)
Q
z


1 +h

2
(x)

e
iQ
z
h(x)
e
iQ
H
x
dx. (5.30)
CCM does not reach the Kirchho approximation or the Physical Optics under the
high frequency limit. The authors did not check properly this limit by applying the
stationary phase condition, they instead concluded that KA is reached because the
half-sum of both Dirichlet and Neumann scattering amplitudes equaled KA. CCM does
however reproduce the SPM1 low-frequency limit and stays reciprocal and simple to
implement.
5.11. Improved Greens Function Methods
Another variety of unifying theories can be grouped under the common denomination
of Improved Greens Function Methods (IGFM). It is based on an improved choice of
CONTENTS 21
the Greens function in the Green-Helmholtz surface integral expressing the scattered
eld. Kodis [214, 215] and Krill and Andreo [216] combine the variational principle
with the use of half-space Greens functions. Berman and Perkins [217] use the Half-
space Greens function in conjunction with the Kirchho current in the scalar case, a
procedure which was extended by Shaw and Dougan to the 3D conducting [218] and
dielectric [219] case. They obtain a hybrid formula for the scattering amplitude, which
is written as the sum of a KA term and an SPM-like term:
S(k, k
0
) = K(k, k
0
)
1
Q
z

e
iQ
z
h(r)
e
iQ
H
r
dr i[B(k, k
0
) K(k, k
0
)]h(Q
H
). (5.31)
This expression is compatible with both SPM and KA in the appropriate limits to within
a small bias in the high-frequency limit due to the corrective term in h(Q
H
). This model
does not preserve the required shift and tilt invariance properties.
5.12. Volumic methods
There is a category of approximate methods which we call volumic because they consider
the case of a rough interface between two homogeneous media as a particular space
distribution of permittivity. The most famous is the Born approximation (BA) which
dates back to the early works of Quantum Mechanics (e.g. [220]) and has rst been
applied to general homogeneous rough surfaces in the context of X-ray and neutron
scattering [221]. The BA refers to a family of methods in which the unknown eld
satises a Lippmann-Schwinger equation, that is essentially an integral equation of
second kind, with a potential term and a Greens function in the integrand. This
is the usual representation of solution of the Schr odinger equation, but the electric eld
satisfying Maxwells equations can also be put in this form, using the dyadic Greens
functions and the permittivity contrast as potential. The solution can then formally be
expressed as an iteration series (the so-called Born- or Neumann series). The zeroth
iteration is the free eld of the reference problem. The Born approximation corresponds
to the rst iteration. In the context of scattering by rough surfaces, the incoherent
scattering amplitude takes the form:
S(k, k
0
) = U(k, k
0
)

e
iQ
H
r
e
iQ
z
h(r)
dr. (5.32)
This expression is structurally identical to SSA1 and KA-HF, but coincides with neither
of them. A consequence is that it does not recover SPM1 in the limit of small heights.
For small contrast, however, the kernel becomes identical to that of SSA1. Precisely, U
can be obtained from the SSA1/SPM1 kernel B by making the replacements q

0
, q

q, q
0
and q q. The Born approximation is valid when the permittivity contrast between
the two media is small, regardless of the roughness. Therefore the BA is mostly used
in X-ray surface scattering, a regime in which the contrasts are extremely weak. The
main weakness of the BA is that it does not take properly into account the reection
by the surface and therefore fails as one approaches the so-called critical angle, at
which total external reection occurs [222, 223]. Another (related) shortcoming of
CONTENTS 22
the method is its inconsistency with SPM1 in the limit of small heights. A great
improvement of BA in this respect can be obtained by choosing a more adapted reference
problem, namely the at surface between two media. In that case the free space Greens
function is replaced by the half-space Greens function, that depends explicitly on the
Fresnel reection coecient at the plane interface . This is the so-called Distorted-
wave Born Approximation (DWBA), a method which takes again its roots in Quantum
Mechanics [220] and was developed in the eighties for X-ray scattering by rough surfaces
[222, 224, 223]. The incoherent scattering amplitude obtained by retaining the rst-
iteration takes the following form:
S(k, k
0
) =

s,s
0
=1
U
s,s
0
(k, k
0
)

e
iQ
H
r
e
i(sq+s
0
q
0
)h(r)
dr, (5.33)
where the sum runs over all possible signs s, s
0
(four possibilities) and hence induces
dierent kernels. This method is known to provide accurate results as long as

Q
z
h <
1, where is the contrast of permittivity. The analytical expression (5.33) can be
further expanded in height to obtain SPM1 (it had in fact be shown well before [225]
that a perturbative height expansion based on a Lippmann-Schwinger equation with
half-space Greens function is consistent with SPM1). With a further iteration, the
method has been shown recently [226] to reproduce SPM2 in a compact formula that
can also be used for heterogeneous surfaces.
The Mean-eld theory (MFT) [227, 228] is a volumic and statistical approach to
rough surface scattering inspired from DWBA. The roughness interface is seen as a
permittivity uctuation in the strip limited by the maximum excursions of the surface.
A mean dielectric permittivity is obtained for each elevation by averaging over the
horizontal direction. Sampling over dierent elevations, a reference problem is dened
as a multi-layer medium whose Greens function can be computed numerically. The
DWBA is then applied to this reference problem and an ensemble average is performed
to obtain the rst two statistical moments of the electric eld. This leads readily to the
mean diuse intensity and the corresponding expression depends in a simple manner
on the reference eld and Greens function and the surface height-correlation function.
The method is semi-analytical inasmuch as the multi-layer Greens function has to be
rst computed numerically. The corresponding cross-section is of the form:

0
=

U(k, k

, z

)U

(k, k

, z

)dz

dz

e
iQ
H
r
L(r, z

, z

; )dr (5.34)
where L is an explicit functional with respect to the surface autocorrelation function
(r) and
0
= (0). U is related to the multi-layer eld and Greens function and
depends only on the surface via its RMS height. The key parameter for the validity
of the method is the permittivity contrast, which must remain moderate ( < 2).
Contrary to practically all surfacic approaches, the MFT becomes more accurate at small
correlation lengths, at a given level of (moderate but nonperturbative) RMS roughness
and permittivity. It is thus able to cope with surface parameters that do not fall in the
(known or supposed) validity domain of any surfacic method, typically for correlation
CONTENTS 23
length and RMS height that are one third of wavelength. The reason is that fast
permittivity uctuations are in favor of the homogenization process that is done at each
elevation. This has been conrmed by numerical [227, 228] as well as experimental [229]
tests. The MFT has also been numerically shown in excellent agreement with SPM1
for small roughness and moderate permittivity. Its main limitations are its restriction
to small and moderate roughness and the inability to predict cross-polarization in the
incidence plane.
5.13. Integral Equation Method
In the late 1980s, the Integral Equation Method (IEM) model was developed to bridge
the gap between SPM and PO models [230]. Initially an IEM version covering the
scattering from perfectly conducting surfaces was published [231, 232]. Subsequently
the IEM model was extended to include the case of scattering from a dielectric rough
interface [233, 9]. IEM is essentially a second iteration of the iterative Kirchho
approximation [106]. It provides an iterative solution of the pair of integral equations for
the tangential components of the electric and magnetic elds, at the dielectric interface,
developed by Poggio and Miller [234]. The expressions of the tangential surface elds are
sought as the superposition of the Kirchho surface elds and corrective terms originated
by the complementary surface currents. Hence, IEM has this functional form:
S(k, k
0
) =

F
1
(k, k
0
; Q
z
h) e
iQ
z
h(r)
e
iQ
H
r
dr (5.35)
+

F
+
2
(k, k
0
; , [h])e
i(k)r
1
i(q
k
+q

)h(r
1
)
e
i(k
0
)r
2
i(q
0
q

)h(r
2
)
ddr
1
dr
2
+

2
(k, k
0
; , [h])e
i(k)r
1
i(q
k
q

)h(r
1
)
e
i(k
0
)r
2
i(q
0
+q

)h(r
2
)
ddr
1
dr
2
.
One drawback of this approach is that the conditions for the convergence of the
iterative series are not known a priori. For instance, the parameter which should be
considered as small for this approach depends on the statistics of the rough surface
under study. To obtain the scattering coecient expressions of 2D rough surfaces in a
relatively simple form several additional approximations are made in the IEM model
(for a critical review see [126]). In the low frequency region (i.e. Ks < 2, K being
the impinging wavenumber and s the surface RMS heights), the scattering coecient
consists of single and multiple-scattering contributions. The single scattering terms
(including corrections due to complementary elds) are expressed as a series of algebraic
terms, whereas the multiple scattering terms are given as 2D three-folds integrals (quite
demanding in terms of numerical computation). Some of the assumptions made in the
IEM development have subsequently been recognized as simplistic by the same original
authors. This is the case of the spectral representation of the Greens function (as
well as of its gradient) where one of the phase terms was arbitrarily dropped leading
to wrong results for the computation of cross-polarized components (i.e. multiple-
scattering contributions). In addition, for slightly rough surfaces, the IEM fails to
CONTENTS 24
reproduce SPM in the general bistatic geometry [147]. An improved version of IEM
was released in 1997 [235], then a further corrected version was published in 2000 [236].
While the expressions of co-polarized backscattering coecient (i.e. single scattering
contribution) are not aected by these revisions, the general bistatic IEM expressions
(including multiple-scattering terms) have been continuously amended until recently
[237, 238]. Extensions of the IEM model to take into account multi-scale roughness
statistics have also been suggested [239, 240]. The IEM accuracy should deteriorate
at large incidence angle and for small radius of curvature of the rough surface. The
IEM predictions of co-polarized backscattering coecient rough surfaces with moderate
slopes and heights have been widely investigated through numerical and experimental
studies (see for instance [9, 241, 242]), where reasonable agreement has been found.
6. Connections between functional forms
6.1. Local and non-local models
In the course of this review we used several times the terminology local and nonlocal
for the dierent models. Although these concepts are rather intuitive in terms of
scattering process, it is dicult to give a general mathematical denition. We propose
to term a model local if the corresponding scattering amplitude can be written as a
single space integral with a kernel that depends only on the surface and a nite number
of derivatives (in general zero, one or two) at a given point. The generic form of a local
model is:
S(k, k
0
) =

F[k, k
0
; r; h(r); (h)(r); ..; (
n
h)(r)]dr (6.36)
SPM1, KA-LF or WCA are examples of local models. A model will be referred to
as non-local of degree (p, q) if the associated SA is a functional in h(r) and h() that
requires at least p +q integrations, with p 0 spatial and q 1 spectral variables. The
generic form for S(k, k
0
) is:

F[k, k
0
; r
1
; ..; r
p
;
1
; ..;
q
; h(r
1
); ..; h(r
p
); h(
1
); ..; h(
q
)]dr
1
..dr
p
d
1
..d
q
(6.37)
Note that possible derivatives of h can be absorbed in the Fourier integrals. The
dependence of the kernel on the Fourier variables must not be solely polynomial,
otherwise the integral can be reduced to a local form. For instance, SPM2, PPM, MLM
and TIA are non-local (0, 1), (1, 1), (2, 1) and (1, 2), respectively. Whenever a model
has a purely statistical formulation (such as GO), we will apply the same terminology
to its scattering cross-section.
The relevant question one might ask now is whether there are any possible
connections between these models. What is the transformation that reduces a nonlocal
model to a local one? Positive answers to these questions clarify most models and help
understanding not only the functional forms but also the corresponding kernels. Indeed,
most previously listed models share roughly the same functional form but with major
CONTENTS 25
dierences in their coecients or kernels. Building bridges across models could also be
seen as a criterion of cross-validation of kernels in use. Connection between models can
be established in two ways: by a gauge transformation (which preserves the degree of
non-locality) or by reduction of spatial integrals.
6.2. Gauge transformation
The main dierence between SSA (5.19) and TIA or LCA (5.22) is that the rst single
integral is multiplied by the SPM1 kernel instead of the Kirchho one. This minor
dierence has a fundamental repercussion on the second order kernel in the double
integral. It also indicates that there is some arbitrariness in the separation of the
scattering amplitude in the sum of a single and double integrals. The possibility of gauge
arbitrariness was already discussed by several authors including [8, 145, 162, 180, 28].
In this prospect, we provide the reader with a general gauge function which transforms
SSA2 (5.19) to TIA/LCA-form (5.22) (discarding the third integral). This gauge
function changes the rst- and second- orders SSA according to
B(k, k
0
) K(k, k
0
), (6.38a)
M(k, k
0
; ) M(k, k
0
; ) [B(k, k
0
) K(k, k
0
)]
W
H
W
z


Q
z
. (6.38b)
This transformed second-order kernel does not necessarily coincide with that of LCA or
TIA. However, the gauge transformation makes the functional forms agree.
6.3. Reduction of nonlocal models
Two major families of nonlocal models can be identied in our previous extensive list of
functional forms. The rst is based on the Meecham-Lysanov form (5.16). The second
family regroups all iterative Kirchho models as in (4.14). While it seems dicult
to establish links between these families, it is possible to compare their respective
transforms after a reduction process from a double spatial integral to a single one.
This reduction process is not trivial but can be outlined succinctly as follows. Let us
employ the triple integral in (4.14) as a general form which includes MLM (5.16) if the
s = sign is set to zero and the kernel to . We rst start by dening the change
of variable inspired from [162], (r
1
, r
2
) (u
1
+ u
2
, u
1
(1 )u
2
) where is an
arbitrary positive parameter (0 1). This change of variable translates the fact
that the mathematical limit of the nonlocal model when r
1
tends to r
2
depends on how
the two vectors approach each other. This arbitrariness in taking the limit is reduced
in our derivation to the sole arbitrary parameter . The reduction process yields a
constant coecient and a kernel which will be placed in the single and double integrals
as in (5.19), respectively. These reduced coecients are

B(k, k
0
) = K
s
2
(k, k
0
, k

), (6.39a)
CONTENTS 26

M(k, k
0
, ) = (q
k
+sq

)K
s
2
(k, k
0
,

) (q
k
+q
0
)K
s
2
(k, k
0
, k

)
+(q
0
sq
+

)K
s
2
(k, k
0
, +

) (6.39b)
where k

= k + (1 )k
0
and

= k

. These transformations can be applied


with particular choices of . All the previously listed nonlocal models can be reduced
to the SSA (5.19) and LCA (5.22) functional forms by setting to 0 or 1 and q
0
/Q
z
,
respectively. A perfect counter example of the SSA-LCA decomposition is when one
takes = 1/2 in the reduction of the nonlocal models. In this case, the reduced model
preserves all the limits reached by its nonlocal parent even though Kirchho and SMP1
kernels are not readily apparent in the analytical decomposition contrary to SSA or
LCA.
7. Synthesis and Conclusion
We provided a review of approximate scattering wave theories from random surfaces
in a unied notation that put the emphasis on the functional form of the scattering
amplitude. We have classied the models in three families, that follow approximatively
the historical evolution. The low-frequency models (SPM and variants), the high-
frequency approximation (Kirchho and variants) and the so-called unied methods,
that aim at bridging the gap between the former two. We have tried to outline the
main principles of the methods, the list of which is recapitulated in Table 7. We have
attempted to evaluate the dierent methods according to a dozen of general criteria
which we think are fundamental. The dierent performances are synthesized in Tables
3 and 4. We have made a distinction between local and non-local models, for which
we have proposed a precise denition. Some horizontal and vertical links between
nonlocal models have been developed in order to explain their apparent dierences.
At the present time, there does not seem to be a universal method that is to be
preferred systematically. All the methods present a compromise between versatility,
simplicity, numerical eciency, accuracy and robustness, with a dierent weighting
in these various elds. At a rst glance at the tables, it appears obvious that no
approximate model has fullled all listed criteria. Moreover, most models did not
even satisfy half of the requirements. One should also mention that a fair comparison
can only be accomplished between models that address the same scattering problem
(acoustic, conducting, dielectric, 2D, 3D, etc ...). There is still room for improvement
in the development of approximate scattering methods, and we hope the check-list of
performances we have proposed in Tables 3 and 4 will help future researchers.
References
[1] P. Beckmann and A. Spizzichino, The scattering of electromagnetic waves from rough surfaces.
Pergamon Press,Oxford, England, 1963.
[2] A. Ishimaru, Ed., Wave Propagation and Scattering in Random Media. New York: Academic,
1978.
CONTENTS 27
[3] F. G. Bass and I. M. Fuks, Wave scattering from statistically rough surfaces. New York:
Pergamon, 1979.
[4] F. T. Ulaby, R. K. Moore, and A. K. Fung, Microwave remote sensing: Active and Passive.
Dedham MA: Artech House, 1982.
[5] L. Tsang, J. A. Kong, and R. Shin, Theory of Microwave remote sensing. Wiley-Interscience,
New-York, 1985.
[6] S. M. Rytov, Y. A. Kravtsov, and V. I. Tatarskii, Principles of Statistical Radiophysics.
Berlin,Heidelberg: Springer, 1987.
[7] J. A. Ogilvy, Theory of wave scattering from random rough surfaces. Adam Hilger, 1991.
[8] A. G. Voronovich, Wave scattering from rough surfaces, ser. Springer Series on Wave Phenomena.
Springer, 1994.
[9] A. K. Fung, Microwave scattering and Emission Models and their Applications. Artech House,
Norwood, Massachusetts, 1994.
[10] L. Tsang, J. A. Kong, and K. H. Ding, Scattering of electromagnetic waves, ser. Wiley series in
remote sensing. Wiley-Interscience, New-York, 2000, vol. I,II,III.
[11] M. Saillard and A. Sentenac, Rigorous solutions for electromagnetic scattering from rough
surfaces, Waves in Random Media, vol. 11, pp. R103R137, 2001.
[12] K. F. Warnick and W. C. Chew, Numerical simulation methods for rough surface scattering,
Waves in Random Media, vol. 11, pp. R1R30, 2001.
[13] D. E. Barrick and W. Peake, A review of scattering from rough surfaces with dierent roughness
scales, Radio Sci., pp. 865868, 1968.
[14] G. R. Valenzuela, Theories for the interaction of electromagnetic waves and oceanic waves. A
review, Bound. Layer Met., pp. 6185, 1978.
[15] E. Rodriguez, Beyond the Kirchho approximation, Radio Sci., vol. 24, pp. 681693, 1989.
[16] R. Dashen and D. Wurmser, A new theory for scattering from a surface, J. Math. Phys., vol. 32,
no. 4, pp. 971985, 1991.
[17] L. Rayleigh, The theory of sound. MacMillan, London, 1896, 3rd Ed.
[18] U. Fano, The theory of anomalous diraction gratings and of quasi-stationary waves on metallic
surfaces (Sommerfelds waves), J. Opt. Soc. America, vol. 31, pp. 213222, 1941.
[19] S. O. Rice, Reection of electromagnetic waves from slightly rough surfaces, Commun. Pure
Appl. Math., vol. 4, pp. 351378, 1951.
[20] , Reection of EM from slightly rough surfaces. Interscience, New-York, 1963.
[21] W. H. Peake, Theory of radar return from terrain, IRE National Convention Record 7, vol. Pt.
1, pp. 2741, 1959.
[22] G. R. Valenzuela, Depolarization of EM waves by slightly rough surfaces, IEEE Trans.
Antennas Propag., vol. 15, pp. 552557, 1967.
[23] , Scattering of electromagnetic waves from a tilted slightly rough surfaces, Radio Sci.,
vol. 13, no. 11, p. 1057, 1968.
[24] E. I. Thorsos and D. R. Jackson, The validity of the perturbation approximation for rough
surface scattering using a Gaussian roughness spectrum, J. Acoust. Soc. America, vol. 86, pp.
261277, 1989.
[25] M. F. Chen and A. K. Fung, A numerical study of the regions of validity of the Kirchho and
small-perturbation rough surface scattering models, Radio Sci., vol. 23, pp. 163170, 1988.
[26] J. M. Soto-Crespo, M. Nieto-Vesperinas, and A. T. Fridberg, Scattering from slightly rough
random surfaces: A detailed study on the validity of the small perturbation method , J. Opt.
Soc. America, vol. 7, pp. 11851201, 1990.
[27] D. H. Berman, Simulations of rough surface scattering, J. Acoust. Soc. America, vol. 89, no. 2,
pp. 623636, 1991.
[28] T. Elfouhaily, S. Guignard, R. Awadallah, and D. R. Thompson, Local and non-local curvature
approximation: a new asymptotic theory for wave scattering, Waves in Random Media, vol. 13,
no. 4, pp. 321338, 2003.
CONTENTS 28
[29] E. Kroger and E. Ktretschmann, Scattering of light by slightly rough surfaces or thin lms
including plasma resonance emission, Z. Phys., vol. 237, no. 1, 1970.
[30] J. M. Elson and R. H. Ritchie, Photon interaction at a rough metal surface, Phys. Rev. B,
vol. 4, p. 4129, 1971.
[31] V. Celli, A. M. Marvin, and F. Toigo, Light scattering from rough surface, Phys. Rev. B, vol. 11,
pp. 17791786, 1975.
[32] A. Marvin, F. Toigo, and V. Celli, Light scattering from rough surfaces, Phys. Rev. B, vol. 11,
p. 2777, 1975.
[33] R. Schier, Reectivity of slightly rough surfaces, Appl. Opt., vol. 26, pp. 704712, 1987.
[34] M. Saillard, A characterization tool for dielectric random rough surfces: Brewsters
phenomenon, Waves in Random Media, vol. 2, pp. 6779, 1992.
[35] A. G. Voronovich, Small-Slope Approximation for electromagnetic wave scattering at a rough
interface of two dielectric half-spaces, Waves in Random Media, vol. 4, pp. 337367, 1994.
[36] J. T. Johnson, Third order small perturbation method for scattering from dielectric rough
surfaces, J. Opt. Soc. America, 1999.
[37] M. A. Demir and J. T. Johnson, Fourth- and higher-order small-perturbation solution for
scattering from dielectric rough surfaces, J. Opt. Soc. America, vol. 20, no. 12, pp. 2330
2337, 2003.
[38] O. P. Bruno and F. Reitich, Numerical solution of diraction problems: A method of variation
of boundaries, J. Opt. Soc. America, vol. 10, pp. 11681175, 1993.
[39] , Numerical solution of diraction problems: A method of variation of boundaries ii. Finitely
conducting gratings, pade approximants, and singularities, J. Opt. Soc. America, vol. 10, pp.
23072316, 1993.
[40] , Numerical solution of diraction problems: A method of variation of boundaries. iii Doubly
periodic gratings, J. Opt. Soc. America, vol. 10, pp. 25512562, 1993.
[41] F. Toigo, A. Marvin, V. Celli, and N. R. Hill, Optical properties of rough surfaces : General
properties and the small roughness limit, Phys. Rev. B, vol. 15, pp. 56185626, 1977.
[42] A. Soubret, G. Berginc, and C. Bourrely, Application of reduced Rayleigh equations to
electromagnetic wave scattering by two-dimensional randomly rough surfaces, Phys. Rev. B,
vol. 63, no. 20, pp. 245 411245 420, 2001.
[43] T. A. Leskova and A. Maradudin, X-ray scattering from a randomly rough surface, Waves in
Random Media, vol. 7, pp. 395334, 1997.
[44] G. C. Brown, V. Celli, M. Coopersmith, and M. Haller, Unitary and reciprocal expansions in
the theory of light scattering from a grating, Surf. Sci., vol. 129, pp. 507515, 1983.
[45] G. C. Brown, V. Celli, M. Haller, and A. M. Marvin, Vector theory of light scattering from
rough surface: Unitary and reciprocal expansions, Surf. Sci., vol. 136, p. 381, 1984.
[46] U. Frisch, Wave propagation in random Media, in Probabilistic Methods in Applied
Mathematics, A. T. Bharucha-Reid, Ed. New York: Academic, 1968.
[47] G. C. Brown, V. Celli, M. Haller, A. Maradudin, and A. Marvin, Resonant light scattering from
a randomly rough surface, Phys. Rev. B, vol. 31, pp. 49935005, 1985.
[48] J. A. Sanchez, A. A. Maradudin, and E. R. Mendez, Limits of validity of three perturbation
theories of the specular scattering of light from one-dimensional, randomly rough, dielectric
surfaces, J. Opt. Soc. America, vol. 12, pp. 15471557, 1995.
[49] A. V. Shchegrov, Enhanced backscattering of surfaces polaritons on a randomly rough surface,
Phys. Rev. B, vol. 57, no. 7, p. 4132, 1998.
[50] A. R. McGurn and A. A. Maradudin, Perturbation theory results for the diuse scattering of
light from two-dimensional randomly rough surfaces, Waves in Random Media, vol. 6, pp.
251267, 1996.
[51] A. A. Maradudin, R. E. Luna, and E. R. Mendez, The brewster eect for a one-dimensional
rough surface, Waves in Random Media, vol. 3, pp. 5160, 1993.
[52] J. Keller, Wave propagation in random media, ser. Proc. Symp. in Appl. Math. Providence, RI:
CONTENTS 29
Am. Math. Soc., 1962, vol. 13, pp. 227246.
[53] , Stochastic equations and wave propagation in random media, ser. Proc. Symp. in Appl.
Math. Providence, RI: Am. Math. Soc., 1964, vol. 16, pp. 145170.
[54] R. Bourret, Bilocal theory of randomly perturbed elds, Hugues Research Labs, Malibu, CA,
Tech. Rep., 1961.
[55] A. Ishimaru, J. D. Rockway, and Y. Kuga, Rough surface greens function based on the rst-order
modied perturbation and smoothed diagram methods, Waves in Random Media, vol. 10, pp.
1731, 2000.
[56] A. Wenzel, Smoothed boundary conditions for randomly rough surfaces, J. Math. Phys., vol. 15,
pp. 317323, 1974.
[57] J. Watson and J. Keller, Reection, scattering, and absorption of acoustic waves by rough
surfaces, J. Acoust. Soc. America, vol. 74, p. 1887, 1983.
[58] , Rough surface scattering via the smoothing method, J. Acoust. Soc. America, vol. 75, p.
1705, 1984.
[59] D. H. Berman and D. K. Dacol, Manifestly reciprocal scattering amplitudes for rough interface
scattering, J. Acoust. Soc. America, vol. 87, pp. 202432, 1990.
[60] G. S. Brown, Application of the integral equation method of smoothing to random surface
scattering, IEEE Trans. Antennas Propag., vol. AP-32, pp. 13081312, 1984.
[61] , The method of smoothing applied to random volume dans surface scattering, in Scattering
in Volume and Surfaces, M. Nieto-Vesperinas and J. C. Dainty, Ed. B. V. North-Holland:
Elsevier Science Publishers, 1990.
[62] D. Maystre, O. M. Mendez, and A. Roger, A new electromagnetic theory for scattering from
shallow rough surfaces, Opt. Act., vol. 30, pp. 17071723, 1983.
[63] M. Nieto-Vesperinas, Depolarization of EM waves scattered from slightly rough random surfaces
: A study by means of the extinction theorem, J. Opt. Soc. America, vol. 72, pp. 539547,
1982.
[64] J. P. Rossi and D. Maystre, A perturbation theory for scattering from three-dimensional rough
surfaces, J. Opt., vol. 56, pp. 219228, 1987.
[65] G. S. Agarwal, Interaction of electromagnetic waves at rough dielectric surfaces, Phys. Rev. B,
vol. 15, pp. 23712383, 1977.
[66] N. Garcia, V. Celli, and M. Nieto-Vesperinas, Exact multiple scattering of waves from random
rough surfaces, Opt. Comm., vol. 30, pp. 279281, 1979.
[67] M. Nieto-Vesperinas and N. Garcia, A detailled study of thec scatering of scalar waves from
random rough surfaces, Opt. Act., vol. 28, pp. 279281, 1981.
[68] D. R. Jackson, D. P. Winebrenner, and A. Ishimaru, Comparison of perturbation theories for
rough-surface scattering, J. Acoust. Soc. America, vol. 83, pp. 961969, 1988.
[69] P. M. Van den Berg and J. T. Fokkema, The Rayleigh hypothesis in the theory of reection by
a grating, J. Opt. Soc. America, vol. 69, pp. 2731, 1979.
[70] F. G. Bass and V. G. Bocharov, Theory of electromagnetic wave scattering from a statistically
rough surface, Radio Eng. and electronics, pp. 251258, 1958.
[71] B. F. Kuryanov, The scattering of sound at a rough surface with two types of irregularities,
Soviet Phys.-Acoustics, vol. 8, pp. 252257, 1963.
[72] K. M. Mitzner, Eect of small irregularities on electromagnetic scattering from an interface of
arbitrary shape, J. Math. Phys., vol. 5, pp. 17761786, 1964.
[73] , Theory of the scattering of waves by irregular surfaces, Ph.D. dissertation, California
Institute of Technology, 1964, downloadable at http://resolver.caltech.edu/CaltechETD:etd-
10182002-092308.
[74] W. C. Meecham, On the use of the Kirchho approximation for the solution of reection
problems, J. Rational Mech. Anal., vol. 5, pp. 323334, 1956.
[75] L. M. Brekhovskikh, Difrakciya voln na nerovnoj poverhnosti: 1. obschaya teoriya, Zh. Eksp.
Teor. Fiz., vol. 23, pp. 275288, 1952, (in Russian).
CONTENTS 30
[76] , Difrakciya voln na nerovnoj poverhnosti: 2. Prilozheniya obschej teorii. Zh. Eksp. Teor.
Fiz., vol. 23, pp. 289304, 1952, (in Russian).
[77] M. A. Isakovich, Wave scattering from a statistically rough surface, Zh. Eksp. Teor. Fiz., vol. 23,
pp. 305314, 1952, (in Russian).
[78] C. Eckart, The scattering of sound from the sea surface, J. Acoust. Soc. America, vol. 25, no. 3,
pp. 66570, 1953.
[79] H. Davies, The reection of electromagnetical waves from rough surfaces, Proc. IEE (London),
vol. 101, no. Part IV, pp. 209214, 1954.
[80] D. E. Barrick, A more exact theory of backscattering from statistically rough surfaces, Ph.D.
dissertation, Ohio state University, 1965.
[81] B. Semenov, Scattering of electromagnetic waves from restricted portions of rough surfaces with
nite conductivity, Radioteckhnika Elektronikha, vol. 10, no. 11, pp. 5260, 1965.
[82] T. Hagfors, Backscattering from an undulating surface with application to radar returns from
the moon, J. Geophys. Res., vol. 69, no. 18, pp. 37793784, 1964.
[83] , Electromagnetic wave theory. Pergamon Press, NY, 1966, vol. 2, ch. Scattering and
transmission of electromagnetic waves at a statistically rough boundary bewteen two dielectric
media.
[84] A. Stogryn, Electromagnetic scattering from rough nitely conducting surfaces, Radio Sci.,
vol. 2, no. 4, pp. 415428, 1967.
[85] R. D. Kodis, A note on the theory of scattering from irregular surfaces, IEEE Trans. Antennas
Propag., vol. 14, no. 1, p. 77, 1966.
[86] E. Rodriguez, Beyond the Kirchho approximation II. Electromagnetic scattering, Radio Sci.,
vol. 26, pp. 121132, 1991.
[87] D. K. Dacol, The Kirchho approximation for acoustic scattering from a rough uid-elastic solid
interface, J. Acoust. Soc. America, vol. 88, p. 978, 1990.
[88] D. E. Barrick, Rough surface scattering based on the specular point theory, IEEE Trans.
Antennas Propag., vol. 16, no. 4, 1968.
[89] , Relationship between slope probability density function and the physical optics
approximation, Proceedings of the IEEE, pp. 17281729, 1968.
[90] A. K. Fung, Theory of radar scatter from rough surface, J. Geophys. Res., vol. 69, pp. 1063
1073, 1964.
[91] C. Cox and W. Munk, Measurement of the roughness of the sea surface from photographs of
the suns glitter, J. Opt. Soc. America, vol. 44, pp. 838850, 1954.
[92] T. Berger, Satellite altimetry using ocean backscatter, IEEE Trans. Antennas Propag., vol. 20,
pp. 295309, 1972.
[93] A. Stogryn, The apparent temperature of the sea at microwaves frequencies, IEEE Trans.
Antennas Propag., vol. 15, pp. 278286, 1967.
[94] P. J. Lynch and R. J. Wagner, Rough surface scattering: Shadowing, multiple scatter and energy
conservation, J. Math. Phys., vol. 11, pp. 30323042, 1970.
[95] J. W. Wright, A new model for sea clutter, IEEE Trans. Antennas Propag., vol. AP-16, pp.
217223, 1968.
[96] P. J. Lynch, Curvature corrections to rough surface scattering at high frequency, J. Acoust.
Soc. America, vol. 47, pp. 804825, 1970.
[97] O. I. Yordanov, K. Ivanova, and A. Michalev, On the Lynch variational method for scattering
of waves by randomly rough surfaces, J. Acoust. Soc. America, vol. 91, no. 5, pp. 21042110,
1991.
[98] K. Ivanova, M. A. Michalev, and I. Yordanov, Numerical study of scattering by rough surfaces
with intermediate and large scale roughness, Radio Sci., vol. 26, no. 2, pp. 505510, 1991.
[99] A. Belobrov and I. Fuks, Short-wave asymptotic analysis of the problem of acoustic wave
diraction by a rough surface, Sov. Phys. Acoust., vol. 31, pp. 442445, 1985, (in Russian).
[100] , Depolarization of radar signal backscattered from random surface, Izv. VUZ Radioz.,
CONTENTS 31
vol. 29, pp. 10831089, 1986, (in Russian).
[101] K. Ivanova and S. L. Broshat, The method of the local parabolic approximation for rough surface
scattering, J. Acoust. Soc. America, vol. 94, no. 4, pp. 23262333, 1993.
[102] I. M. Fuks and A. G. Voronovich, Wave diraction by a concave statistically rough surface,
Waves in Random Media, vol. 9, pp. 501520, 1999.
[103] H. L. Chan and A. K. Fung, A numerical study of the Kirchho approximation in horizontally
polarized bakscattering from a random surface, Radio Sci., vol. 13, no. 5, pp. 811818, 1978.
[104] E. I. Thorsos, The validity of the Kirchho approximation for rough surface scattering using a
Gaussian roughness spectrum, J. Acoust. Soc. America, vol. 83, pp. 7892, 1988.
[105] A. G. Voronovich, The limits of applicability of Kirchhos approximation in problems of
scattering of sound on a rough free surface, Dokl. Akad. Nauk SSSR, vol. 273, pp. 830834,
1983, (in Russian).
[106] E. G. Liszka and J. J. McCoy, Scattering at a rough boundaryextensions of the Kirchho
approximation, J. Acoust. Soc. America, vol. 71, no. 5, pp. 10931100, 1982.
[107] D. F. McCammon and S. T. McDaniel, Surface reection: on the convergence of a series solution
to a modied Helmholtz integral equation and the validity fo the Kirchho approximation, J.
Acoust. Soc. America, vol. 79, pp. 6470, 1986.
[108] Y.-Q. Jin, Multiple scattering from a randomly rough surface, J. Appl. Phys., vol. 63, no. 5,
pp. 12861292, 1988.
[109] Y.-Q. Jin and M. Lax, Backscattering enhancement from a randomly rough surface, Phys. Rev.
B, vol. 42, no. 16, pp. 98199829, 1990.
[110] P. Beckmann, Shadowing of random rough surfaces, IEEE Trans. Antennas Propag., vol. 13,
pp. 384388, 1965.
[111] R. J. Wagner, Shadowing of randomly rough surfaces, J. Acoust. Soc. America, vol. 41, no. 1,
pp. 138147, 1966.
[112] B. G. Smith, Geometrical shadowing of a random rough surface, IEEE Trans. Antennas
Propag., vol. 15, pp. 668671, 1967.
[113] P. J. Lynch and R. J. Wagner, Energy conservation for rough surface scattering, J. Acoust.
Soc. America, vol. 47, no. 3, p. 816, 1970.
[114] M. I. Sancer, Shadow-corrected electromagnetic scattering from a randomly rough surface,
IEEE Trans. Antennas Propag., vol. 17, pp. 577585, 1969.
[115] A. Ishimaru and J. S. Chen, Scattering from very rough surfaces based on the modied second-
order Kirchho approximation with angular and propagation shadowing, J. Acoust. Soc.
America, vol. 88, no. 4, pp. 18771883, 1990.
[116] , Scattering from very rough metallic and dielectric surfaces: A theory based on the modied
Kirchho approximation, Waves in Random Media, vol. 1, pp. 2134, 1991.
[117] J. S. Chen and A. Ishimaru, Numerical simulation on the second-order Kirchho approximation
from very rough surfaces and a study of backscattering enhancement, J. Acoust. Soc. America,
vol. 88, no. 4, pp. 18461850, 1990.
[118] C. Bourlier, J. Saillard, and G. Berginc, Eect of correlation between shadowing and shadowing
points on the Wagner and Smith monostatic one-dimensional shadowed functions, IEEE
Trans. Antennas Propag., vol. 3, pp. 437446, 2000.
[119] C. Bourlier, G. Berginc, and J. Saillard, Bistatic scattering coecient from one- and two-
dimensional random surfaces using the stationary phase and scalar approximation with
shadowing eect: comparisons with experiments and application to the sea surface, Waves in
Random Media, vol. 11, pp. 119147, 2001.
[120] C. Bourlier, J. Saillard, and G. Berginc, Theoretical study of the Kirchho integral from two-
dimensional randomly rough surface with shadowing eect-application on the backscattering
coecient for a perfectly, Waves in Random Media, vol. 11, pp. 91118, 2001.
[121] , Bistatic scattering coecient from one- and two-dimensional random surfaces using
the stationary phase and scalar approximation with shadowing eects. comparisons with
CONTENTS 32
experiments and application to the sea surface, Waves in Random Media, vol. 11, pp. 119147,
2001.
[122] , One- and two- dimensional shadowing functions for any height and slope stationary
uncorrelated surface in the monostatic and bistatic congurations, IEEE Trans. Antennas
Propag., vol. 50, no. 3, 2002.
[123] J. J. Embrechts, Simulation of rst and second order scattering by rough surfaces with a sound
ray formalism, J. Sound Vib., vol. 229, no. 1, pp. 6587, 2000.
[124] C. Macaskill, Geometric optics and enhanced backscatter from very rough surfaces, J. Opt.
Soc. America, vol. 8, pp. 8896, 1991.
[125] E. Bahar and M. El-Shenawee, Enhanced backscatter from one-dimensional random rough
surfaces: stationary-phase approximations to full-wave solutions, J. Opt. Soc. America, vol. 12,
no. 1, pp. 151161, 1995.
[126] J. L. Alvarez-Perez, An extension of the IEM/IEMM surface scattering model, Waves in
Random Media, vol. 11, pp. 307329, 2001.
[127] P. Tran and A. A. Maradudin, Scattering of a scalar beam from a two-dimensional randomly
rough hard wall: Enhanced backscattering, Phys. Rev. B, vol. 45, pp. 39363939, 1992.
[128] W. C. Meecham, Fourier transform method for the treatment of the problem of the reection of
radiation from irregular surfaces, J. Acoust. Soc. America, vol. 28, pp. 370377, 1956.
[129] Y. P. Lysanov, About one approximate solution of the problem of acoustic wave scattering by a
rough interface, Sov. Phys. Acoust., vol. 2, pp. 190198, 1956, Engl. Transl.
[130] A. G. Yarovoy, Small-slope iteration technique for wave scattering from a rough interface of two
media, IEEE Trans. Antennas Propag., vol. 44, no. 11, pp. 14131437, 1996.
[131] M. Saillard and G. Soriano, Fast numerical solution for scattering from rough surfaces with small
slopes, IEEE Trans. Antennas Propag., 2004, to appear.
[132] J. Shen and A. A. Maradudin, Multiple scattering of waves from random rough surfaces, Phys.
Rev. B, vol. 22, pp. 42344240, 1980.
[133] D. P. Winebrenner and A. Ishimaru, Application of the phase perturbation technique to
randomly rough surface, J. Opt. Soc. America, vol. 2, pp. 22852294, 1985.
[134] , Investigation of a surface eld perturbation technique for scattering from rough surface,
Radio Sci., vol. 20, pp. 161170, 1985.
[135] S. L. Broshat, L. Tsang, A. Ishimaru, and E. I. Thorsos, A numerical comparison of the phase
perturbation technique with the classical eld perturbation and Kirchho approximation for
random rough surface scattering, J. Electromagn. Waves and Appl., vol. 2, pp. 119136, 1988.
[136] S. L. Broshat, E. I. Thorsos, and A. Ishimaru, The phase perturbation technique versus an
exact numerical method for random rough surface scattering, J. Electromagn. Waves and
Appl., vol. 3, pp. 237256, 1989.
[137] K. Ivanova, M. A. Michalev, and I. Yordanov, Numerical study of the phase perturbation
technique for scattering by rough surfaces, C. R. Bulg. Acad. Sci., vol. 42, no. 1, pp. 64
67, 1989.
[138] , Study of the phase perturbation technique for scattering of waves by rough surfaces at
intermediate and large values of the roughness parameter, J. Electromagn. Waves and Appl.,
vol. 4, no. 5, pp. 401414, 1990.
[139] S. L. Broshat, E. I. Thorsos, and A. Ishimaru, A heuristic algorithm for the bistatic radar cross
section for random rough surface scattering, IEEE Trans. Geosci. and Remote Sens., vol. 28,
no. 2, 1990.
[140] A. Schatzberg and A. J. Devaney, Rough surface inverse scattering within the Rytov
approximation, J. Opt. Soc. America, vol. 10, no. 5, pp. 943950, 1993.
[141] T. Yang and S. L. Broshat, Acoustic scattering from a uid-elastic-solid interface using the small
slope expansion, J. Acoust. Soc. America, vol. 96, no. 3, pp. 17961804, 1994.
[142] R. M. Fitzgerald and A. A. Maradudin, A reciprocal phase-perturbation theory for rough-surface
scattering, Waves in Random Media, vol. 4, pp. 275296, 1994.
CONTENTS 33
[143] A. G. Voronovich, Small-Slope Approximation in wave scattering from rough surfaces, Sov.
Phys. JETP, vol. 62, no. 1, pp. 6570, 1985.
[144] , Non-classical approaches in the theory of wave scattering from rough surfaces, in
Mathematical and numerical aspects of wave propagation (Golden, CO, 1998). Philadelphia,
PA: SIAM, 1998, pp. 8488.
[145] V. I. Tatarskii, The expansion of the solution of the rough-surface scattering problem in powers
of quasi-slopes, Waves in Random Media, vol. 3, no. 2, pp. 127146, 1993.
[146] S. T. McDaniel, A small-slope theory of rough surface scattering, J. Acoust. Soc. America,
vol. 95, pp. 185864, 1994.
[147] T. Elfouhaily, D. R. Thompson, D. Vandemark, and B. Chapron, A new bistatic model for
electromagnetic scattering from perfectly conducting random surfaces, Waves in Random
Media, vol. 9, no. 3, pp. 281294, 1999.
[148] G. Soriano and M. Saillard, Scattering of electromagnetic waves from two-dimensional rough
surfaces with impedance approximation, J. Opt. Soc. America, vol. 18, no. 1, pp. 124133,
2001.
[149] T. Elfouhaily, M. Joelson, S. Guignard, and D. R. Thompson, Analytical comparison between
the surface current integral equation and the second-order small-slope approximation, Waves
in Random Media, vol. 13, no. 3, pp. 165176, 2003.
[150] E. I. Thorsos and S. L. Broshat, An investigation of the Small-Slope Approximation for scattering
from rough surfaces. part I. Theory, J. Acoust. Soc. America, vol. 97, no. 4, pp. 20822093,
1995.
[151] , An investigation of the Small-Slope Approximation for scattering from rough surfaces.
part II. Numerical studies, J. Acoust. Soc. America, vol. 101, no. 5, pp. 26152625, 1997.
[152] S. T. McDaniel, A numerical study of rough surface scatterer, Journ. Comput. Acoust., vol. 8,
no. 3, pp. 443458, 2000.
[153] , Acoustic and radar scattering from directional seas, Waves in Random Media, vol. 9, pp.
537549, 1999.
[154] , Small-slope predictions of microwave backscatter from the sea surface, Waves in Random
Media, vol. 11, pp. 343360, 2001.
[155] A. G. Voronovich and V. U. Zavorotny, Theoretical model for scattering of radar signals in Ku-
and C-bands from a rough sea surface with breaking waves, Waves in Random Media, vol. 11,
pp. 247269, 2001.
[156] G. Berginc, Small-slope approximation method: a further study of vector wave scattering from
two dimensional surfaces and comparison with experimental data, in PIER 2003, 2003, pp.
251287.
[157] C. A. Guerin and M. Saillard, On the high-frequency limit of the second-order small-slope
approximation, Waves in Random Media, vol. 13, pp. 7588, 2003.
[158] A. G. Voronovich, The eect of the modulation of Bragg scattering in Small-Slope
Approximation, Waves in Random Media, vol. 12, pp. 341349, 2002.
[159] , A two-scale model from the point of view of the Small-Slope Approximation, Waves in
Random Media, vol. 6, pp. 7383, 1996.
[160] M. S. Gilbert and J. T. Johnson, A study of the higher-order small-slope approximation for
scattering from a gaussian surface, Waves in Random Media, vol. 13, pp. 137143, 2003.
[161] V. I. Tatarskii and V. V. Tatarskii, Statistical description of rough surface scattering using
the quasi-small approximation for random surfaces with a gaussian multivariate probability
distribution, Waves in Random Media, vol. 4, pp. 191214, 1994.
[162] S. T. McDaniel, An extension of the small-slope approximation for rough surface scattering,
Waves in Random Media, vol. 5, pp. 201214, 1995.
[163] T. Elfouhaily, D. R. Thompson, D. E. Freund, D. Vandemark, and B. Chapron, A new bistatic
model for electromagnetic scattering from perfectly conducting random surfaces: Numerical
evaluation and comparison with SPM, Waves in Random Media, vol. 11, no. 1, pp. 3344,
CONTENTS 34
2001.
[164] D. Holliday, Resolution of a controversy surrunding the Kirchho approach and the small
perturbation method in rough surface scattering theory, IEEE Trans. Antennas Propag.,
vol. 35, pp. 120122, 1987.
[165] A. G. Voronovich, Non-local Small-Slope Approximation for wave scattering from rough
surfaces, Waves in Random Media, vol. 6, pp. 151167, 1996.
[166] S. L. Broshat and E. I. Thorsos, A preliminary study of the nonlocal Small-Slope
Approximation, J. Acoust. Soc. America, vol. 100, p. 2702, 1996.
[167] S. L. Broshat, Reection loss from a Pierson-Moskowitz sea surface using the nonlocal Small-
Slope Approximation, IEEE Trans. Geosci. and Remote Sens., vol. 37, no. 1, pp. 632634,
1999.
[168] H. Kim and J. T. Johnson, Radar images of rough surface scattering: Comparison of numerical
and analytical models, IEEE Trans. Antennas Propag., vol. 50, no. 2, pp. 94100, 2002.
[169] T. Elfouhaily, S. Guignard, and D. R. Thompson, Formal tilt invariance of the non-local
curvature approximation and its connection to the integral equation method, IEEE TGARS
Letters, 2004, in press.
[170] K. M. Watson, A transport-equation description of nonlionear ocean surface wave interactions,
J. Fluid Mech., vol. 70, pp. 815826, 1975.
[171] J. W. West, A new numerical method for surface hydrodynamics, J. Geophys. Res., vol. 92,
no. C11, pp. 11 80311 824, 1987.
[172] D. M. Milder, The eects of truncation on surface-wave Hamiltonians, J. Fluid Mech., vol. 216,
pp. 249262, 1990.
[173] , An improved formalism for wave scattering from rough surface, J. Acoust. Soc. America,
vol. 89, pp. 529541, 1991.
[174] D. M. Milder and T. Sharp, An improved formalism for wave scattering from rough surface:
Numerical trials in three dimensions, J. Acoust. Soc. America, vol. 91, no. 5, pp. 26202626,
1992.
[175] D. M. Milder, An improved formalism for electromagnetic scattering from a perfectly conducting
surface, Radio Sci., vol. 31, pp. 13691376, 1996.
[176] R. A. Smith, The operator expansion formalism for electromagnetic scattering from rough
dielectric surfaces, Radio Sci., vol. 31, pp. 13771385, 1996.
[177] P. J. Kaczkowski and E. I. Thorsos, Application of the operator expansion method to scattering
from one-dimensional moderately rough Dirichlet random rough surfaces, J. Acoust. Soc.
America, vol. 96, pp. 957971, 1994.
[178] D. M. Milder, An improved formalism of coherent forward scatter from random rough surfaces,
Waves in Random Media, vol. 8, pp. 6778, 1999.
[179] M. I. Charnotskii and V. I. Tatarskii, Tilt-invariant theory of scattering by rough surface,
Waves in Random Media, vol. 5, pp. 36180, 1995.
[180] T. Elfouhaily, S. Guignard, and D. R. Thompson, Formal tilt invariance of the local curvature
approximation, Waves in Random Media, vol. 13, no. 4, p. L7, 2003.
[181] R. Dashen and D. Wurmser, Approximate representations of the scattering amplitude, J. Math.
Phys., vol. 32, no. 4, pp. 986996, 1991.
[182] , Applications of the new scattering formalism: the dirichlet boundary condition, J. Math.
Phys., vol. 32, no. 4, pp. 9971003, 1991.
[183] C. A. Guerin, G. Soriano, and T. Elfouhaily, Weighted curvature approximation: numerical tests
on 2d dielectric surfaces, Waves in Random Media, 2004, submitted.
[184] J. Nakayama, H. Ogura, and B. Matsumoto, A probabilistic theory of electromagnetic waves
scattering from a slightly rough random surface, Radio Sci., vol. 15, pp. 104957, 1980.
[185] J. Nakayama, M. Sakata, and H. Ogura, A probabilistic theory of electromagnetic waves
scattering from a slightly rough random surface: Horizontal polarization, Radio Sci., vol. 16,
pp. 831845, 1981.
CONTENTS 35
[186] , A probabilistic theory of electromagnetic waves scattering from a slightly rough random
surface: Vertical polarization, Radio Sci., vol. 16, pp. 847853, 1981.
[187] J. Nakayama, K. Mitzutani, and M. Tsuneoka, Scattering of electromagnetic waves from a
perfectly conductive slightly random surface: Depolarization in backscatter, J. Math. Phys.,
vol. 27, no. 5, pp. 14351448, 1986.
[188] J. Nakayama, L. Gao, and Y. Tamura, Scattering of a plane wave from a periodic random
surface: a probabilistic approach, Waves in Random Media, vol. 7, no. 1, pp. 6578, 1997.
[189] K. Hattori, J. Nakayama, and H. Matsuoka, Wave scattering from a periodic random surface
generated by a stationary binary sequence, Waves in Random Media, vol. 11, no. 1, pp. 120,
2001.
[190] C. Eftimiu, First-order Wiener-Hermite expansion in the electromagnetic scattering by
conducting rough surfaces, Radio Sci., vol. 23, no. 5, pp. 769779, 1988.
[191] , Second-order Wiener-Hermite expansion in the electromagnetic scattering by conducting
rough surfaces, Radio Sci., vol. 23, no. 6, pp. 10751084, 1988.
[192] , First-order Wiener-Hermite expansion in the electromagnetic scattering by dielectric rough
surfaces, Radio Sci., vol. 25, no. 1, pp. 18, 1990.
[193] , Modied Wiener-Hermite expansion in rough surface scattering, J. Opt. Soc. America,
vol. 6, no. 10, pp. 15841594, 1989.
[194] , Scattering by a rough dielectric interface: A modied Wiener-Hermite expansion
approach, J. Opt. Soc. America, vol. 7, no. 5, pp. 875884, 1990.
[195] E. Rodriguez and Y. Kim, A unied perturbation expansion for surface scattering, Radio Sci.,
vol. 27, pp. 7993, 1992.
[196] E. Rodriguez, Y. Kim, and S. L. Durden, A numerical assessment of rough surface scattering
theories: Vertical polarization, Radio Sci., vol. 27, pp. 515527, 1992.
[197] E. Rodriguez and Y. Kim, A numerical assessment of rough surface scattering theories:
Horizontal polarization, Radio Sci., vol. 27, pp. 497513, 1992.
[198] E. Bahar, Radio wave propagation in stratied media with nonuniform boundaries and varying
electromagnetic parameters: Full wave analysis, Canad. J. Phys., vol. 50, pp. 31323142,
1972.
[199] , Electromagnetic wave propagation in inhomogeneous multilayered structures of arbitrarily
varying thickness-generalized eld transforms, J. Math. Phys., vol. 14, pp. 10241029, 1973.
[200] , Electromagnetic wave propagation in inhomogeneous multilayered structures of arbitrarily
varying thickness-full wave solutions, J. Math. Phys., vol. 14, pp. 10301036, 1973.
[201] , Depolarization of electromagnetic wave excited by distributions of electric and magnetic
sources in inhomogeneous multilayered structures of arbitrarily varying thickness-generalized
eld transforms, J. Math. Phys., vol. 14, pp. 15021509, 1973.
[202] , Depolarization of electromagnetic wave excited by distributions of electric and magnetic
sources in inhomogeneous multilayered structures of arbitrarily varying thickness-full wave
solutions, J. Math. Phys., vol. 14, pp. 15101515, 1973.
[203] E. Bahar and G. G. Rajan, Depolarization and scattering of electromagnetic waves by irregular
boundaries for arbitrary incident and scatter angles full wave solutions, IEEE Trans.
Antennas Propag., vol. Ap, no. 27, pp. 214225, 1979.
[204] E. Bahar, Full-wave solutions for the depolarization of the scattered radiation by rough surfaces
of arbitrary slope, IEEE Trans. Antennas Propag., vol. 29, pp. 443454, 1981.
[205] E. Bahar and M. A. Fitzwater, Depolarization and backscatter enhancement in light scattering
from random rough surfaces: comparison of full wave theory with experiment, J. Opt. Soc.
America, vol. 6, no. 1, pp. 3343, 1989.
[206] E. Bahar, Full wave solutions for the scattering of acoustic waves excited by arbitrary source
distributions in irregular layered media, Wave Motion, vol. 12, no. 4, pp. 301314, 1990.
[207] , Scattering of acoustic waves in irregular layered media: full wave solutions, Wave Motion,
vol. 12, no. 1, pp. 5365, 1990.
CONTENTS 36
[208] R. E. Collin, Electromagnetic scattering from perfectly conducting rough surfaces (a new full
wave method), IEEE Trans. Antennas Propag., vol. 40, no. 12, pp. 14661477, 1992.
[209] , Full wave theories for rough surface scattering: An updated assessment, Radio Sci.,
vol. 29, no. 5, pp. 12371254, 1994.
[210] E. Bahar, Radar scatter cross-sections for two-dimensional random rough surfaces. full wave
solutions and comparisons with experiments, Waves in Random Media, vol. 6, pp. 123, 1996.
[211] A. Garca-Valenzuela and R. E. Collin, The local spectral expansion method for scattering
from perfectly condution surfaces rough in one dimension, J. Electromagn. Waves and Appl.,
vol. 11, pp. 3769, 1997.
[212] , The local spectral expansion method for scattering from one dimensional dielectric-
dielectric rough interfaces, J. Electromagn. Waves and Appl., vol. 11, pp. 775805, 1997.
[213] F. K. Schwering, G. M. Whitman, and A. A. Triolo, New full-wave theory for scattering from
rough metal surfaces by the correction current method: the TE-polarization case, Waves in
Random Media, vol. 14, pp. 2360, 2004.
[214] R. D. Kodis, An introduction to variational techniques in electromagnetic scattering, J. SIAM,
vol. 2, p. 89, 1954.
[215] , Variational principles in high-frequency scattering, Proc. R. Soc. Lond. A., vol. 54, p.
512, 1958.
[216] J. A. Krill and R. H. Andreo, Vector stochastic variational principles for electromagnetic wave
scattering, IEEE Trans. Antennas Propag., vol. 28, p. 770, 1980.
[217] D. H. Berman and J. S. Perkins, The Kirchho approximation and rst-order perturbation
theory for rough surface scattering, J. Acoust. Soc. America, vol. 78, p. 1045, 1985.
[218] W. T. Shaw and A. J. Dougan, Half-space Greens functions and applications to scattering of
scattering of electromagnetic waves from ocean-like surfaces, Waves in Random Media, vol. 5,
pp. 341359, 1995.
[219] , Greens function renement as an approach to radar backscatter: General theory and
application to LGA scattering form the ocean, IEEE Trans. Antennas Propag., vol. 46, no. 1,
pp. 5766, 1998.
[220] L. Schi, Quantum Mechanics. McGraw-Hill, 1968.
[221] P. Wong and A. Bray, Scattering from rough surfaces, Phys. Rev. B, vol. 37, pp. 7751, 1988.
[222] G. Vineyard, Grazing-incidence diraction anf the distorted-wave approximation for the study
of surfaces, Phys. Rev. B, vol. 26, no. 8, pp. 41464159, 1982.
[223] S. K. Sinha, E. B. Sirota, and S. Garo, X-ray and neutron scattering from rough surfaces,
Phys. Rev. B, vol. 38, pp. 22972311, 1988.
[224] U. Mohanty and S. Rice, X-ray total external reectance diraction as a probe of the structure
of the liquid-vapor interface, J. Chem. Phys., pp. 24822487, 1983.
[225] A. A. Maradudin and D. L. Mills, Scattering and absorption of electromagnetic radiation by a
semi-innite medium in presence of surface roughness, Phys. Rev. B, vol. 11, p. 1392, 1975.
[226] C. A. Guerin and A. Sentenac, Second-order perturbation theory for scattering from
heterogeneous surfaces, J. Opt. Soc. America, 2004, to appear.
[227] A. Sentenac and J.-J. Greet, Mean-eld theory of light scattering by one-dimensional rough
surfaces , J. Opt. Soc. America, pp. 528532, 1998.
[228] O. Calvo-Perez, A. Sentenac, and J.-J. Greet, Light scattering by a two-dimensional, rough
penetrable medium : A mean-eld theory, Radio Sci., pp. 311335, 1999.
[229] O. Calvo-Perez, Diusion des ondes electromagnetiques par un lm rugueux heterog`ene, Ph.D.
dissertation, Ecole Centrale Paris, 1999.
[230] A. K. Fung and G. W. Pan, An integral equation method for rough surface scattering, in
Proceedings of the Intern. Symposium on multiple scattering of waves in random media and
random surfaces, 1986, pp. 701714.
[231] , A scattering model for perfectly conducting random surface: I. Model development, Int.
J. Remote Sens., vol. 8, no. 11, 1987.
CONTENTS 37
[232] G. W. Pan and A. K. Fung, A scattering model for perfectly conducting random surface: II.
Range of validity, Int. J. Remote Sens., vol. 8, no. 11, 1987.
[233] A. K. Fung, Z. Li, and K. S. Chen, Backscattering from a randomly rough dielectric surface,
IEEE Trans. Geosci. and Remote Sens., vol. 30, no. 2, pp. 356369, 1992.
[234] A. J. Poggio and E. K. Miller, Integral Equation Solution of Three Dimentional Scattering
Problems. Computer Techniques for Electromagnetics, Pergamon, New York, 1973.
[235] C. Y. Hsieh, A. K. Fung, G. Nesti, A. J. Siber, and P. Coppo, A further study of the IEM surface
scattering model, IEEE Trans. Geosci. and Remote Sens., vol. 35, no. 4, pp. 901909, 1997.
[236] K. S. Chen, T. D. Wu, and A. K. Fung, A note on multiple scattering in an IEM model, IEEE
Trans. Geosci. and Remote Sens., vol. 38, no. 1, pp. 249256, 2000.
[237] A. K. Fung, Z. Li, and K. S. Chen, An improved IEM model for bistatic scattering from rough
surfaces, J. Electromagn. Waves and Appl., vol. 16, no. 5, pp. 689702, 2002.
[238] K. S. Chen, T. D. Wu, L. Tsang, Q. Li, J. C. Shi, and A. K. Fung., Emission of rough surfaces
calculated by the integral equation method with comparison to three-dimensional moment
method simulations. IEEE Trans. Geosci. and Remote Sens., vol. 41, no. 1, pp. 90101, 2003.
[239] K. S. Chen, T. D. Wu, and A. K. Fung, A study of backscattering from multi-scale rough
surface, J. Electromagn. Waves and Appl., vol. 12, no. 7, pp. 961979, 1998.
[240] F. Mattia, Backstattering properties of multiscale rough surfaces, J. Electromagn. Waves and
Appl., vol. 13, no. 4, pp. 493527, 1999.
[241] A. K. Fung, Z. Li, A. Shah, and M. Tjuatja, Numerical-simulation of scattering from 3-
dimensional randomly rough surfaces, IEEE Trans. Geosci. and Remote Sens., vol. 32, no. 5,
pp. 986994, 1994.
[242] G. Macelloni, G. Nesti, P. Pampaloni, S. Sigismondi, D. Tarchi, and S. Lolli, Experimental
validation of surface scattering and emission models, IEEE Trans. Geosci. and Remote Sens.,
vol. 38, no. 1, pp. 459469, 2000.
CONTENTS 38
Acronym Full name main references
BA Born Approximation [221, 223]
CCM Correction Current Method [213]
DWBA Distorted-wave Born Approximation [222, 223]
ESSA Extended Small-Slope Approximation [162]
FWA Full Wave Approach [198, 210]
GO Geometrical Optics [78, 79]
KA Kirchho Approximation [1, 3]
KA-LF Low-Frequency Kirchho Approximation [81, 82]
KA-HF High-Frequency Kirchho Approximation [84, 85]
IEM Integral Equation Method [9, 126]
IGFM Improved Greens Function Method [217, 219]
LPA Local Parabolic Approximation [99, 102]
LSEM Local Spectral Expansion Method [211, 212]
LVM Lynch Variational Method [96, 98]
LCA Local Curvature Approximation [149, 180]
LWA Local Weight Approximation [181, 182]
MFT Mean-Field Theory [227, 228]
MLM Meecham-Lysanov Method [128, 129]
MTE Momentum Transfer Expansion [15, 86]
NLCA Non-Local Curvature Approximation [28, 169]
NLSSA Non-Local Small-Slope Approximation [165]
OEM Operator Expansion Method [175, 176]
PO Physical Optics [84, 85]
PPM Phase Perturbation Method [133, 134]
QSE Quasi Slope Expansion [145, 161]
SPM Small Perturbation Method [19, 20]
SSA Small-Slope Approximation [8, 35]
TIA Tilt-Invariant Approximation [179]
TPA Tangent Plane Approximation [81, 82]
TSM Two-Scale Model [3, 14]
UPE Unied Perturbation Expansion [195, 197]
WCA Weighted Curvature Approximation [28, 183]
WHA Wiener-Hermite Approach [184, 190]
Table 1. List of acronyms (alphabetical order) together with two main
references for the corresponding method.
CONTENTS 39
Method 1 2 3 4 5 6 7 8 9
SPM (Rayleigh) - x x x x - x - -
SPM (Rigorous) x - x x x - - x -
KA-LF (TPA) x - x - - - - - -
KA-HF (PO) - x - - - - - - x
GO (SPT) - x - - - - - - x
TSM - x x x x - x - x
MTE x - x - - - - x -
LVM x - x - - - x - -
LPA x - x - - - x - -
MLM x - - x - x - x -
PPM x - x x x - - - -
SSA - x x x x - - - -
QSE x - - x - - - x -
ESSA x - - x - - - x -
NLSSA - x x x x - x - -
NLCA - x x - x - - - x
OEM x - x x - - - - -
TIA x - x x - - - x -
LCA - x x - x - - - x
LWA - x x - - - - - x
WCA - x x - x - - - x
WHA x - x x - - - x -
UPE x - x x - - x - -
FWA x - x x - - x - -
LSEM x - x x - - x - -
CCM x - x x - - x - -
IGFM x - - - - x - - -
BA - - - - - - x - -
DWBA - - - - - x x - -
MFT - - - - - x x - -
IEM x - - - - - x - -
Table 2. Main principles of the dierent methods, in their order of appearance. A
x in a row indicates that the method uses the corresponding rule.
1. Proceeds via estimation of a surface current.
2. Works directly on the scattering amplitude
3. Makes a structure Ansatz for the surface current or the scattering amplitude.
4. Uses a functional expansion of some quantity in surface height.
5. Requires consistency with SPM for the identication of some unknown series.
6. Uses a particular choice of Greens function.
7. Uses second-kind Fredholm integral equation.
8. Uses rst-kind Fredholm integral equation.
9. Uses the stationary phase approximation.
CONTENTS 40
Property 1 2 3a 3b 3c 4 5 6 7 8a 8b
SPM1
KA-LF
KA-HF
GO1 -
LVM
LPA ?
MTE
SSA1
IGFM
LWA
WCA
BA
DWBA
MFT
CCM
Table 3. Properties of the local models.
1. All types of surfaces (dielectric, conducting, acoustic)
2. Full two-dimensional surfaces.
3a. Reciprocal (manifestly)
3b. Shift invariant (to arbitrary order in shift)
3c. Tilt invariant (at least to rst order in tilt)
4. Numerically fast and stable while easy to implement
5. Statistical formulae already available or easily derivable
6. Not restricted to large correlation lengths
7. Not restricted to small surface height
8a. SPM1 limit
8b. GO1 limit
The signication of the symbols is as follows:
=satised by construction,
=satised upon inspection,
=satised under special conditions
(specular, backscattering, conducting case, etc...),
? =unknown or not tested.,
=not satised,
-=irrelevant.
CONTENTS 41
Property 1 2 3a 3b 3c 4 5 6 7 8a 8b 9a 9b 10 11
SPM2 - - (0, 1)
GO2 - - (0, 1)
WHA ? - (0, 2)
PPM - (1, 1)
SSA2 (1, 1)
QSE - (1, 1)
ESSA - (1, 1)
LSEM - (1, 1)
LCA (1, 1)
UPE (1, 2)
TIA - (1, 2)
KA2-HF - (2, 1)
MLM - (2, 1)
NLSSA (2, 1)
OEM1 (2, 1)
FWA ? (2, 1)
IEM (2, 1)
OEM2 ? ? (3, 2)
Table 4. Properties of the non-local models.
1. All types of surfaces (dielectric, conducting, acoustic)
2. Full two-dimensional surfaces.
3a. Reciprocal (manifestly)
3b. Shift invariant (to arbitrary order in shift)
3c. Tilt invariant (at least to rst order in tilt)
4. Numerically fast and stable while easy to implement
5. Statistical formulae already available or easily derivable
6. Not restricted to large correlation lengths
7. Not restricted to small surface height
8a. SPM1 limit
8b. GO1 limit
9a. SPM2 limit
9b. GO2 limit.
10. Can predict correct cross-polarization in the plane of incidence
11. Degree of non-locality (p, q)
The signication of the symbols is as follows:
=satised by construction,
=satised upon inspection,
=satised under special conditions
(specular, backscattering, conducting case, etc...),
? =unknown or not tested.,
=not satised,
-=irrelevant.
We have classied the models according to their degree of non-locality.