See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/223229453

Effects of surface roughness on light scattering by small particles

Article  in  Journal of Quantitative Spectroscopy and Radiative Transfer · November 2004

DOI: 10.1016/j.jqsrt.2004.05.016

CITATIONS READS
48 261

3 authors, including:

Changhui li George W Kattawar

Peking University Texas A&M University
89 PUBLICATIONS   2,560 CITATIONS    318 PUBLICATIONS   7,026 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Slip-ring-based multi-transducer photoacoustic tomography system View project

Acoustically penetrable optical reflector for photoacoustic tomography View project

All content following this page was uploaded by Changhui li on 18 May 2018.

The user has requested enhancement of the downloaded file.

 ARTICLE IN PRESS

Journal of Quantitative Spectroscopy &

Radiative Transfer 89 (2004) 123–131
www.elsevier.com/locate/jqsrt

Effects of surface roughness on light scattering

by small particles
Changhui Lia, George W. Kattawara,, Ping Yangb
a
Department of Physics, Texas A&M University, College Station, TX, USA
b
Department of Atmospheric Sciences, Texas A&M University, College Station, TX, USA

Abstract

Roughness exists on almost all natural particle surfaces, such as spores, aerosols, cells, etc. In this paper,
we study the effects of surface roughness on light scattering for spherical shaped particles. A 3D finite-
difference time domain model is used to solve for all the optical properties of the scatterer. We will
introduce a roughness parameter and show its effect on the light scattering properties. Our results show that
surface roughness can indeed play an important role on the light scattering pattern under certain
conditions.
r 2004 Elsevier Ltd. All rights reserved.

Keywords: FDTD; Scattering; Phase matrix; Aerosols

1. Introduction

The scattering of light by small particles is essential to atmospheric radiative transfer [1],
biological aerosol detection [2], remote-sensing technology [3] and other disciplines. The surface
texture of a scattering particle, in addition to the overall particle geometric shape, is an important
morphological factor in determining the optical properties of the scatterer. In the past two
decades, the effect of asphericity of a particle on its single-scattering parameters (e.g., phase
function and cross sections) has been extensively investigated (e.g., [4,5] and references cited

Corresponding author. Tel.: +1-979-845-1180; fax: +1-979-845-2590.

E-mail address: kattawar@tamu.edu (G.W. Kattawar).

0022-4073/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jqsrt.2004.05.016
 ARTICLE IN PRESS

124 C. Li et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 89 (2004) 123–131

therein). However, only a handful of studies have investigated the effect of surface texture or
roughness on particle optical properties. For particles with size parameters in the geometric
regime, Macke et al. [6] and Yang and Liou [7] employed the principles of geometric optics to
study the optical properties of ice crystals with surface roughness. It is quite challenging to
compute the single-scattering parameters of irregular particles with size parameters in the
resonant region (i.e., where particle size is of the same order as the incident wavelength) where the
applicability of geometric optics breaks down. In cases where the roughness scale is much smaller
than both the particle size and the wavelength, Schiffer used an approximation approach [8,9] to
study particles from small size parameters to very large size parameters. Also, many studies were
done on deformed Gaussian spheres and spheroids using different methods such as the volume-
integral equation approach by Lumme et al. [10], the second-order perturbation series by
Nousiainen et al. [11], the finite-difference time domain (FDTD) method by Sun et al. [12]. Those
kinds of roughness surfaces, on deformed Gaussian particles, have locally smooth surfaces
without large slopes over most of the surface area. The roughness style in this paper is very
different. The surface becomes more locally ‘‘roughened’’ like the surface of a cell with microvilli.
Several research works have been done on this kind of roughness [13,14] using discrete-dipole-
approximation method. Most recently, Sun et al. [15] used FDTD method to investigate the effect
of surface roughness in two-dimensions for size parameters in the resonant regime. Since the
scattering feature of a three-dimensional object is quite different from its idealized counterpart in
a two-dimensional space, there is a need to investigate the roughness effect associated with a
three-dimensional particle whose size parameter is in the resonant regime. It is noteworthy that
surface roughness is often observed for natural particles such as ice crystals in cirrus clouds,
biological spores in air or water, and animal tissue cells. Thus, the effect of particle surface
roughness on optical properties is a subject that is of both theoretical and practical importance.
To simulate the optical properties of a roughened particle, one needs to define the surface
roughness. Unlike methods used in [13,14] by randomly removing dipole elements on surface, in
this study we employed a well-controlled roughness model to determine the surface roughness,
which specifies not only the total roughness area [14] but also the micro scale features of the
roughness (width and depth) in a straightforward manner. Furthermore, we employ the FDTD
method to compute the phase matrix of roughened spheres. This paper is organized as follows: In
Section 2, the basic principle of the FDTD method is briefly recaptured; presented in Section 3 is
the method used to specify particle surface roughness; in Section 4 the phase matrix of roughened
spheres, computed from the FDTD method, is discussed in comparison with their counterparts
for smooth spheres, and finally, the conclusions of this study are given in Section 5.

2. Finite difference time domain method for light scattering computation

Numerous methods have been developed to deal with the scattering of light by irregular
particles, as recently reviewed by Mishchenko et al. [4], Wriedt [5], Kahnert [16]. In this study we
employ the FDTD method to investigate the optical properties of roughened particles. The
FDTD method originally developed by Yee [17] in 1966 is suitable for computing the optical
properties of a particle with a complex morphology. The basic principle of this method is to
compute the near-field from the two Maxwell curl equations. In practice, the time-dependent
 ARTICLE IN PRESS

C. Li et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 89 (2004) 123–131 125

Fig. 1. Comparison of reduced phase matrix elements between FDTD and Lorenz–Mie calculations.

Maxwell equations are discretized within a finite computational domain bounded by proper
absorbing layers. The electric and magnetic fields within the domain are simulated by the finite
difference analog of Maxwell’s equations. The presence of a scattering particle in the
computational domain is specified by properly assigning permittivity values at grid points. In
this manner, the FDTD method is flexible for simulating various scattering processes involving
complex scatterer morphologies and configurations.
Since the computational domain is spatially limited, the implementation of the FDTD method
requires an absorbing boundary condition (ABC) to absorb outgoing scattered waves so that the
fields inside the interior domain are the same as the field in an unbounded space. The perfectly
matched layer (PML) [18] and Uniaxial PML (UPML) [19], among all the ABCs that have been
developed, produce minimal artificial boundary reflection. In this study, we use a three-
dimensional FDTD model developed by Yang et al. [20], which has been enhanced by using the
UPML boundary condition. The accuracy of the FDTD method is conventionally validated by a
comparison of the FDTD results with the exact Lorenz-Mie solution for spheres.
Fig. 1 illustrates the present implementation of the FDTD method in the case of spheres. The
size parameter for the computation is x ¼ 10 with a refractive index of m ¼ 1:53 þ 0:008i, a value
typical of the aerosol refractive index. Evidently, the FDTD results agree well with the analytic
Lorenz–Mie solution, although slight errors are noticed in side scattering directions. It should be
noted that even though P11 agrees well with the FDTD solution, the reduced matrix elements
(elements normalized to P11 ) P12 , P33 , and P34 are more sensitive to the staircasing effect induced
by the use of a cartesian grid and the aliasing induced by doing discrete numerical computations
based on a finite difference technique.

3. Model of pseudo-random roughness of the particle surface

In this study, we investigated the effect of surface roughness for particles whose overall
morphological shapes are spherical. In reality, the surface roughness of a natural particle is a
 ARTICLE IN PRESS

126 C. Li et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 89 (2004) 123–131

morphological feature with random nature. To model a roughened surface, we use a simple
mathematical scheme to define pseudo-random surface roughness. First we uniformly randomly
choose a number of points on the sphere surface. At each point, the region around this point
2 2
is roughened: a ‘‘spike’’ given by the Gaussian distribution aAeDy =2s centered at the point
is applied. Then the radial distances of the points in the vicinity of the selected point are in the
form of
2 2
rðy; jÞ ¼ R0 ½1 þ aAeDy =2s ; ð1Þ
where y, j are the zenith and azimuth angles of neighboring points; R0 is the original radius of the
sphere; A is a random number in the range ð1:0pAp1:0Þ, so the ‘‘spike’’ can be either outward
or inward; 0pap1 is a preset parameter; Dy is the relative angle between the radial vectors
pointing to center of the Gaussian spike and its reference point; and s is the standard deviation of
the Gaussian distribution used to control the size of the spike, which is important for the FDTD
calculation.
To specify the surface roughness effect in light scattering computations, we define the degree of
roughness by introducing a parameter: Z ¼ 2pjmjm=l, in which jmj is the absolute value of
refractive index, l is the wavelength, m is the standard deviation of the radial distance obtained
from standard statistical procedures: uniformly choosing NðN 1Þ sample points on the
roughened sphere surface, we then calculate their radial distance ri ði ¼ 1; 2; . . . ; NÞ and their
mean value r, finally
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u 1 X N
m¼t ðri  rÞ2 :
N  1 i¼1

In our pseudo-roughness sphere, r is approximately equal to original sphere radius R0 .

To ensure the validity of the FDTD method when it is applied to a particle with surface
roughness, the scale of the roughness spike needs to be larger than the spatial grid used in the
FDTD computation. In this paper, we ensure that the scale of each roughness spike is at least
three times larger than the grid size.
Fig. 2 shows the morphologies of roughened spheres with various roughness conditions. Note
that a perfectly smooth sphere is a special case of a roughened sphere when the roughness
parameter Z ¼ 0. The present method for specifying the roughness has several advantages in the
sense that random numbers are used to determine the position and height of the roughness points.
Thus, the final roughness surface has no symmetry. Additionally, the size of the roughness spikes
can be controlled to ensure the FDTD method is reliable in simulation. Furthermore, the
parameter Z can be adjusted by changing the number of roughness points and their amplitudes.
The parameter s used to control the size of the angle in this paper is very small (approximately
between 1
and 2
) to get the final surface structures shown in Fig. 2. To get similar roughness
patterns using the Gaussian random sphere model, the correlation angle would have to be very
small, and in that case, we could not guarantee the final roughness spikes were larger than the
FDTD grid size, which is the reason we did not use the Gaussian deformed sphere model in our
calculations.
 ARTICLE IN PRESS

C. Li et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 89 (2004) 123–131 127

Fig. 2. Samples of surface roughness for spheres with increasing values of the roughness parameter Z.

4. Phase matrix of particles with surface roughness

The roughened spheres studied in this paper are homogenous with refractive index
m ¼ 1:53 þ 0:008i, typical of some aerosols. Different values of the parameter Z are used for
each size parameter. This parameter is not arbitrary because we must choose it in a reasonable
range to keep the roughness relatively small compared with the original shape, which implies that
large values of Z can only occur for the large size parameter cases. To satisfy the requirement that
the size of the spatial grid in the FDTD is less than the roughness spike as described in the
previous section, we chose the grid size as Dx ¼ l=40 for x ¼ 5, Dx ¼ l=30 for x ¼ 10,15 and 20,
respectively.
For a smooth sphere, the phase matrix elements depend only on the zenith angle (angle between
the scattered light and incident light); however, for roughened spheres, it will also depend on the
azimuth angle. To compare our result with spheres, we choose a fixed scattering plane (fixed
incident direction and azimuth angle) to compare the results and we averaged the roughened
sphere results over random orientations.
Fig. 3 compares the reduced phase matrix elements P11 , P12 , P33 and P34 for the case of size
parameter x ¼ 5. To keep the overall morphological shapes close to spherical, Z is relatively small
for these small particles. The first thing to note is that P11 agrees very well with the smooth sphere
except at angles close to backscattering and there it reduces the glory value. The other reduced
matrix elements show substantial differences for a large range of scattering angles.
Figs. 4–6 compare the same elements shown in Fig.3; however, they are for size parameters of
x ¼ 10, 15 and 20, respectively. With the increase of size, a larger degree of roughness can be
achieved without changing the overall morphological shape. Just as in Fig. 3, the results for small
Z values (dotted lines in figures) keep the deviation of the phase function ðP11 Þ from its spherical
counterpart very small but the deviation gets larger with increasing size parameter. The other
 ARTICLE IN PRESS

128 C. Li et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 89 (2004) 123–131

Fig. 3. Comparison between the reduced phase matrix elements computed from the roughened sphere and Lorenz–Mie calculations
for a sphere with a size parameter x ¼ 5.

Fig. 4. Same as Fig.3 except for size parameter x ¼ 10.

elements can deviate by very large amounts; however, what is noteworthy is the fact that the
maxima and minima still keep their relative positions but the amplitudes are in general reduced.
The scattered light also becomes more unpolarized; and this effect has also been noted in [15]
using a 2D model. It is clear that when Z41, the spherical approximation can not be used any
more, otherwise substantial errors will be introduced.
Another interesting point to note is that the reduced matrix element P44 ¼ P33 for a sphere of
any size, and this relationship held up remarkably well for the roughened spheres as well although
the plot is not shown.
A good test of asphericity can be seen in the element P22 . We show this element in Fig. 7 for all
the size parameters used in this paper. The asphericity can be detected in this element for
scattering angles in the backward direction.
 ARTICLE IN PRESS

C. Li et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 89 (2004) 123–131 129

Fig. 5. Same as Fig.3 except for size parameter of x ¼ 15.

Fig. 6. Same as Fig.3 except for size parameter of x ¼ 20.

5. Conclusions

In this paper, we have studied the surface roughness effects for different size spheres. To specify
the roughness, we introduced a degree of roughness parameter Z, which is a measure of the
relative size of the roughness compared with the wavelength. Our results show that for Zo1 the
effect of surface roughness on light scattering is relatively small as far as the phase function ðP11 Þ
is concerned; however, for the other phase matrix elements, substantial errors can be incurred at
some scattering angles. For Z41 and for the larger size particles, such as x415, even the phase
function can show substantial deviations from its spherical counterpart. The upshot of this
analysis is that using spherical approximations for roughened particles can lead to large errors in
the results especially when one is doing Mueller (phase) matrix imaging.
 ARTICLE IN PRESS

130 C. Li et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 89 (2004) 123–131

Fig. 7. Comparison of the reduced phase matrix element P22 (the element sensitive to asphericity) with a sphere which is always unity
for all scattering angles.

Acknowledgements

This research was partially supported by the Office of naval Research under contract N00014-
02-1-0478. This research is also supported by National Science Foundation (ATM-0239605) and
research grants from NASA Radiation Sciences Program managed by Dr’s. Donald Anderson
and Hal Maring.

References

[1] Liou KN, Takano Y, Yang P. In: Mishchenko MI Hovenier JW, Travis LD., (editors.) Light scattering by
nonspherical particles: theory, measurements, and applications. New York: Academic Press; 2000. p.417– 449
[chapter 15].
[2] Pan YL, Aptowicz KB, Chang RK, Hart M, Eversole JD. Opt Lett 2003;28:589–91.
[3] Wielicki BA, Barkstrom BR, Baum BA, Charlock TP, Green RN, Kratz DP, Lee RB, Minnis P, Smith GL, Wong
T, Young DF, Cess RD, Coakley JA, Crommelynck DAH, Donner L, Kandel R, King MD, Miller AJ,
Ramanathan V, Radall DA, Stowe LL, Welch RM. IEEE Trans Geosci Remote Sensing 1998;36:1127–41.
[4] Mishchenko MI, Hovenier JW, Travis LD, editors. Light scattering by nonspherical particles. San Diego, CA:
Academic Press; 2000.
[5] Wriedt T. A review of elastic light scattering theories. Part Part Syst Charact 1998;15:67–74.
[6] Macke A, Muller J, Rascke E. J Atmos Sci 1996;53:2813–25.
[7] Yang P, Liou KN. Control Atmos Phys 1998;71:223–48.
[8] Schiffer R. J Opt Soc Am 1989;A6:385–402.
[9] Schiffer R. Appl Opt 1990;29:1536–50.
[10] Lumme K, Rahola J. JQSRT 1998;60:439–50.
[11] Nousiainen T, Muinonen K, Avelin J, Sihvola A. JQSRT 2001;70:639–61.
[12] Sun W, Nousiainen T, Muinonen K, Fu Q, Loeb NG, Videen G. JQSRT 2003;79–80:1083–90.
[13] Chamaillard K, Lafon JP. JQSRT 2001;70:519–28.
[14] Perrin JM, Sivan JP. Astrophys J 1991;247:497.
 ARTICLE IN PRESS

C. Li et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 89 (2004) 123–131 131

[15] Sun W, Loeb NG. In: Seventh Conference on Electromagnetic and Light Scattering by Nonspherical Particles:
Theory, Measurements and Applications. Germany: Bremen; 2003. p. 354–7.
[16] Kahnert FM. JQSRT 2003;79–80:775–824.
[17] Yee KS. IEEE Trans Antenna Propag 1966;AP-14:302–7.
[18] Berenger JP. J Comput Phys 1994;114:185–200.
[19] Sacks ZS, Kingsland DM, Lee R, Lee JF. IEEE Trans Antenna Propag 1995;43:1460–3.
[20] Yang P, Liou KN, Mishchenko MI, Gao BC. Appl Opt 2000;39:3727–37.

View publication stats