You are on page 1of 3

December 1987 / Vol. 12, No.

12 / OPTICS LETTERS 979

Monte Carlo simulations for scattering of electromagnetic waves


from perfectly conductive random rough surfaces

M. Nieto-Vesperinas and J.M. Soto-Crespo


Instituto de Optica, Consejo Superior de Investigaciones Cientificas, Serrano 121, 28006-Madrid, Spain

Received June 8, 1987; accepted July 30, 1987


Numerical calculations of mean scattered intensities by simulation of one-dimensionalperfectly conductive random
rough surfaces are presented. Results relative to backscattering enhancement and more accurate criteria for the
validity of the Kirchhoff approximation are obtained. This method can also be used for assessing perturbative
theories and for further experiments.

The scattering of light and other electromagnetic surfaces are generated by following a procedure used
waves from rough surfaces has long been a subject of in Ref. 18, that is, a sequence of random numbers, with
interest. Recently much attention has been focused normal statistics, zero mean, and variance unity is
on this subject in connection with the discovery of the constructed from another series of random numbers
weak localization of photons in random media. This uniformly distributed in (0, 1) directly generated by
is manifested by an enhancement of intensity in the the computer. Then the former sequence is rescaled
backscattering direction when multiple scattering to the desired variance, and the result, say, lykl, is
takes place.1 -4 This phenomenon has been observed correlated with a Gaussian, (2/4rT)1 exp[-2 (k/t) 2],
12

recently in extremely rough random surfaces with to get the appropriate surface profile tzsl with a Gauss-
large conductivity. 5' 6 However, none of the current ian correlation function.
reported theories has predicted this effect sofar. This For a given incident linearly polarized plane wave
may be because multiple-scattering models are pertur- a exp[iko (x sin 00 - z cos 00)] (ko = 27r/X, a is a unit
bative, based either on the Rayleigh hypothesis7' 8 or vector specifying the polarization, and 00 is the inci-
on the extinction theorem9 - 1 3 as formulated in Ref. 14. dence angle) upon a certain surface sample, the corre-
(A good review may be found in Ref. 15.) Therefore sponding electric current J(x) is determined from the
all calculations so far have been restricted to small extinction theorem boundary condition, namely, Eqs.
roughness, a <<X. A perturbative analysis established (15) and (17) of Ref. 14 for the electric field for s waves
in Ref. 16 accounts for the existence of backscattering and Eqs. (18) and (20) of Ref. 14 for the magnetic field
peaks for small roughness when there is interaction for p waves. (Note that only the electric current term
with surface polaritons but does not apply to the ob- will contribute for perfect conductors. Also, since the
servations of Refs. 5 and 6. surface is 1-D, no depolarization occurs.) Then the
The purpose of this Letter is to show that the formu- electric current is introduced into the far-zone expres-
lation based on the extinction theorem (which in prin- sion for the scattered field, and the angular distribu-
ciple constitutes an exact approach to the problem) tion of scattered intensity for that sample is obtained.
can account for the enhanced backscattering phenom- This intensity is normalized to its total area, so that
enon if one uses it without perturbative series approxi- the result is independent of the total incident energy,
mations. As a matter of fact, that method permits illuminated sample length, and other constants ap-
numerical calculations that may be performed for al- pearing in the asymptotic far-zone expressions. For
most any roughness a and correlation length T of the each incident wave, the intensities for 200 samples
surface profile within the limitations of today's com- were calculated, and then their average was taken; 220
puters (which require that the surface model be one points were taken in each surface sample to perform
dimensional). This is important in our opinion, since the x integrals. Calculations were done on a CDC
it permits the prediction of interesting results in quite Cyber 180/855 computer. Small dispersion in the ar-
different situations, including those of large a or large eas of the calculated nonnormalized intensities was
incidence angles, not yet considered in multiple-scat- considered a criterion of numerical consistency of the
tering theories. Further, it can yield an exhaustive results. The statistical bias of the mean intensities
collection of scattering data with the possibility of produced asymmetric curves for 00 = 00, which were
establishing more-accurate criteria of validity of the artificially symmetrized by averaging every two values
well-known Kirchhoff approximation (KA).17 at scattering angles 0 and -0.
We present Monte Carlo computer simulations for For brevity we present only a few representative
the scattered intensities from samples of one dimen- results. Figs. 1-3 show the angular distribution of
sional (1-D) perfectly conductive random surfaces, z = mean scattered intensities for s waves (solid lines) and
D(x), with rms a, normal statistics, and a Gaussian p waves (dotted lines) for T = 0.2X, X,and 1.8X, respec-
correlation function exp(-r 2/T 2 ), r = x - x'. The tively, and for different values of a-,(also in units of X).

0146-9592/87/120979-03$2.00/0 © 1987, Optical Society of America


980 OPTICS LETTERS / Vol. 12, No. 12 / December 1987

INCIDENC ANZGL - 70 DeGREES INCIDENCE ANGLE 70 DEGREES


broadly scattered in all directions, and its intensity
7
distribution is skewed toward the backscattering di-
I- I
4
rection, its backscattering peak being sharper as 00
..... ... ........... l7 lt l
decreases. For the same value of T and larger u [Fig.
§ -7 4 .a . 7s .s tl
I (b)], both s and p waves yield a peak in the backscat-
Ob..... LIon .ngl. RN 8.0080 0
-o EglE
tering, although at larger 00the specular peak is still
INCIDNCt ANGEU 40 DEGREES INCIDENCEA7OGLE. 40 DEGREES noticeable for s waves. In this connection, a sort of
competition between the specular and backscattering
I .2
-1
directions exists for different values of T and a. This
can even shift the maximum with respect to either
I direction [see, e.g., Fig. 3(b) at 00 = 401 for-p waves].
Ob ..... t00 .. Sngl.
- .. I, -11..
- .. 11 .. .' .. ..
.. II. Figures 2(b) and 3(b) also illustrate some aspects of
tNEID[NEC ANGLE 0 DEGUIS MOVE- .. - D ECREEI
this. Incidentally, these two figures are qualitatively
.j equivalent to Fig. 2 of Ref. 5, which corresponds to an
.2
experiment with 00= 20', T = 2.2X,and a = 1.6X.
.0- It was remarked in Ref. 19 that the backscattering
..i . I I I . peak could not be observed unless ensemble averages
\_,j

Ob ..... II..
were made. This is clearly confirmed in our results.
0.2
( a)
cr=
To= .2 (b)
cr=
T = 0.2
0.5
Figure 4 (a) shows the scattered intensity for T = X,a =
1.5X, and 00 = 40' for one surface sample (p waves
Fig. 1. Mean scattered intensity for (a) T = 0.2,N,a = 0.2,N, above, s waves below). Large speckle fluctuations are
(b) T = 0.2X, a = 0.5X, and several angles of incidence. Solid exhibited that swamp the backscattering and specular
curves, s waves; dotted curves, p waves.

INIDENCE AN.L. -. 70rDECR. I111:..E111T CL. - 1. DEGI.I.

INCIDEN" -LE - 7. D-l" I.L. - 1. D-EE.


I.I - "I,\

'\
7 -
Eg ....
......... . - ...........
-- a. o - .6 -17 .. .1 .. 1i 1. ...
.. - .. - .. -11 I. .1 .. . 1. .. ..
Ob ... 0I).-.1i.. -'I.
SI . . . . - . . . . . .I . I
4. i -"".i ", .. , ." .. , .'. .. ,, .. .'. ..
Ob..... II.. .. gI. b ...- ... ... 1, ... I: 11.1r. - 1. INIEC NLE . o0 DECREES
Z
I
INCIENE .. CLE 1. DECREE5 INCIEXCE: 11CLE. 1o DEGIEES
i
2

\ I _' ' _/ 'I... ' '"


. I
... .. - ' - '. ' '.Ob .. .. gl.
..
Ob.-.Ii..n .. gl.
11.1r. - D E.'- 11"Gtr. 0 DIGI-1

INCDENI
.. XC.L . O DECGlES .."'IDE11E G1r 0 DEr.11
I

4-

-. ,I 1.
;. ........... .. .1 Ob ..... ..
O.2
-67-;'. 'I' I% .II. ( 1) ) T,
Cr- 0.2
( a ) T = 1.0

Fig. 3. Same as Fig. I for (a) 1.8X, UN; (b) T 1.8X,


Fig. 2. Same as Fig. 1 for (a) T = X,a = 0.2X; (b) T = X,a = a = 1.5X.
1.5X.
W.D... -L. - 1. D...I 111I.D.11C. GL. - 10 - ...

.S' . . I

It is seen that the specular peaks for both s and p 1. : -1

waves are dominant as a << X and T increases with


respect to a and X [Figs. 2(a) and 3(a)]. (Of course
.1`F_

- ..
11

Iv - I
"". "
- - I

' -X__ I.,


I,
k4 111. .11. ..
.. g.
these peaks are larger with respect to the background
as 00increases.) Conversely,the backscattering effect 1.11D.1111 G. - .. ...- -1DI111t GL. . 1. DIGRM

is dominant as a increases, specificallyfrom values of a -I


similar to or greater than T [Figs. 1(b), 2(b), and 3(b)]. 2: !! . I

Note that in Figs. 2(a) and 3(a) there is an immersion I i 0A VV\_


of the specular peak into the background halo for K\_\
11...
-
.. .. .".
lower 00because of its diffraction width, as the illumi- Ob-1.1i.. .. SI. , = , .5 -11 11:111.
1111 1111
nated sample is finite. An interesting phenomenon (a) T = I .0 (b) T = ' 'O

appears at low T; see Fig. 1(a). There a is comparable Fig. 4. (a) Scattered intensity from one sample with T = X,
with T, and thus small a still gives rise to a large a = 1.5Xfor s waves (lower curve) and p waves (upper curve).
specular peak for the s component. However, the p (b) Mean scattered intensity from ten samples with T = X,a
component has a drastically different behavior: it is = 1.5X for s waves (lower curve) and p waves (upper curve).
December 1987 / Vol. 12, No. 12 / OPTICS LETTERS 981

maxima. Speckle tends to disappear as one starts References


averaging. Figure 4(b) shows the same scattered in-
tensity as Fig. 4(a) but for averaging over 10 samples. 1. Y. Kuga and A. Ishimaru, J. Opt. Soc. Am. A 1, 831
This figure and the average in Fig. 2(b) at Oo = 400, (1984).
which involves 200 samples, confirms the above state- 2. L. Tsang and A. Ishimaru, J. Opt. Soc. Am. A 1, 836
ment in the case of rough surfaces. (1984).
3. H. P. Van Albada and A. Lagendijk, Phys. Rev. Lett. 55,
Assessments of the validity of the KA can also be 2692 (1985).
made from these calculations. Figures 2(a) and 3(a) 4. P. E. Wolf and G. Maret, Phys. Rev. Lett. 55, 2696
show almost identical distributions for s and p waves (1985).
for 00up to about 40° in Fig. 2(a) and for any 00 in Fig. 5. E. R. M6ndez and K. A. O'Donnell, Opt. Commun. 61,91
3(a). We have made calculations with the KA and (1987).
obtained identical results (the KA does not account 6. K. A. O'Donnell and E. R. Mendez, J. Opt. Soc. Am. A 4,
for depolarization). We have obtained an approxi- 1194 (1987).
mate agreement even for T = 0.2X and a- = 0.05X. If 7. S. 0. Rice, Comm. Pure Appl. Math. 4, 351 (1951).
one examines these cases one observes that they all 8. G. R. Valenzuela, IEEE Trans. Antennas Propag. AP-
correspond to values a-/(T)cos 0o < 0.2, regardless of 15, 552 (1967).
9. F. Toigo, A. Marvin, V. Celli, and N. R. Hill, Phys. Rev.
the value of T/X. We propose to adopt this as a more B 15, 2371 (1977).
accurate criterion for the validity of the KA in random 10. J. Shen and A. Maradudin, Phys. Rev. B 22,4234 (1980).
surfaces. [It remains to be seen whether the criterion 11. M. Nieto-Vesperinas, J. Opt. Soc. Am. 72, 539 (1982).
also holds for a/X very large, namely, a-/(Xcos Go)> 10, 12. D. P. Winbrenner and A. Ishimaru, J. Opt. Soc. Am. A 2,
and for periodic surfaces.] This value is considerably 2285 (1985).
more refined than the one (T > 2X and a-T < 0.05) 13. D. Maystre and J. P. Rossi, J. Opt. Soc. Am. A 3, 1276
proposed in Ref. 20 and explains the agreement of the (1986).
KA with actual experiments at T X, a- 0.2X,and So 14. D. N. Pattanayak and E. Wolf, Opt. Commun. 6, 217
= 20° found in Ref. 6. (1972).
15. G. S. Brown, Wave Motion 7, 195 (1985).
Finally, we believe that these calculations may be
16. V. Celli, A. A. Maradudin, A. A. Marvin, and A. R.
useful in checking other results, such as the range of McGurn, J. Opt. Soc. Am. A 2, 2225 (1985).
validity of perturbative theories, scattering from non- 17. P. Beckmann and A. Spizzichino, The Scattering of
Gaussian surfaces, and further experiments. E.M. Waves from Rough Surfaces (Macmillan, New
We thank J. C. Dainty, E. R. M6ndez, and K. A. York, 1963).
18. A. K. Fung and M. F. Chen, J. Opt. Soc. Am. A 2, 2274
O'Donnell for sharing their experimental data with us. (1985).
This research was supported by the Comisi6n Asesora 19. S. Etemand, R. A. Thompson, and M. Andrejco, Phys.
de Investigaci6n Cientffica y T6cnica under grant Rev. Lett. 57, 575 (1986).
2502/83. The research of J. M. Soto-Crespo is sup- 20. M. Nieto-Vesperinas and N. Garcia, Opt. Acta 28, 1651
ported by the Ministerio de Educaci6n y Ciencia. (1981).

You might also like