Beruflich Dokumente
Kultur Dokumente
Vadim Backman
Department of Biomedical Engineering, Northwestern University, Evanston, Illinois 60208, USA
(Received 26 May 2004; published 17 November 2004)
We report the development and validation of the equiphase-sphere (EPS) approximation for calculating the
total-scattering cross-section (TSCS) spectra of inhomogeneous microparticles having complex interior struc-
tures. We show that this closed-form, analytical approximation can accurately model the TSCS of randomly
inhomogeneous spherical particles having internal refractive index variations with geometrical scales spanning
from nanometers (i.e., subwavelength) to microns (i.e., suprawavelength). Moreover, we derive an easy-to-use
criterion for the range of validity of the EPS approximation in modeling TSCS of inhomogeneous particles.
The work discussed in this paper may positively impact tissue optical imaging and diagnostic applications.
In the EPS approximation, the wavelength-dependent direction of the light-ray propagation [Eq. (5)], the formulas
TSCS spectrum of a particle is given by the sum of the given by Eqs. (1), (2), and (6) provide improved accuracy for
“edge-effect” term s共s兲共兲 and the, “volume-diffraction- calculating the TSCS spectra, particularly for particles with
effect” term s共兲共兲 [7] higher refractive indices.
Motivated by the questions whether the interference struc-
s共兲 = s共s兲共兲 + s共兲共兲. 共1兲 tures are preserved for nonspherical particles, and how their
TSCS spectra are associated with the particle size and shape
Here, when the high frequency ripples resulting from inter-
characteristics, we previously introduced the concept of the
ference of the surface waves is neglected, s共s兲共兲 can be
“equiphase sphere” [7]. Most recently, we proposed to use
approximated as [7,10] the equiphase-sphere (EPS) approximation [Eqs. (1), (2), and
s共s兲共兲 ⬇ 2S关2共3V/4兲1/3/兴−2/3 , 共2兲 (6)] to calculate the TSCS spectra of a variety of nonspheri-
cal particles [8,9], where is replaced by the equivalent
where S is the particle’s maximum cross-section area trans- maximum phase shift calculated according to the particle’s
verse to the direction of the incident light, and V is the vol- geometrical characteristics.
ume of the particle.
For a particle with 2d共n − 1兲 / Ⰷ 1 and 共n − 1兲 ⬍ 1, where
III. APPLICATION OF EPS THEORY TO THE
d is the mean diameter and n is the refractive index, the
INHOMOGENEOUS SPHERES: RANGE OF VALIDITY
volume term s共兲共兲 can be approximated using the WKB
technique [11] In this section we focus our discussion on applying the
冉冕 冕 冊
EPS approximation to spherical particles with inhomoge-
s共兲共兲 = 2 Re 兵1 − exp关i共r⬘兲兴其d2r⬘ , 共3兲 neous interior refractive index. Here, of Eq. (6) is simply
S replaced by the maximum phase shift produced by the ho-
mogeneous counterpart of the particle with n0 equal to the
where r⬘ is a position vector in the plane orthogonal to the volume-averaged refractive index of the inhomogeneous par-
direction of propagation of the incident wave and 共r⬘兲 is the ticle. Upon this substitution, Eq. (6) predicts that the oscilla-
phase shift of a light ray crossing plane S at position r⬘. 共r⬘兲 tion features in the TSCS spectrum of an inhomogeneous
is expressed as particle follow that of its homogeneous counterpart with a
共r⬘兲 = 共2/兲 冕
L共r⬘兲
关n共l共r⬘兲兲 − 1兴dl, 共4兲
volume-averaged refractive index.
In order to apply the EPS method in practice, it is impor-
tant to determine the validity conditions of this approxima-
where L共r⬘兲 is the path of the light ray crossing r⬘. For a tion. We now investigate how the internal refractive-index
homogeneous spherical particle with refractive index n0, distribution affects the validity and accuracy of the EPS ap-
proximation applied to inhomogeneous particles. The deriva-
L = d关1 − sin2 ␥共r⬘兲/n20兴1/2 , 共5兲 tion of an analytical validity condition for Eq. (3) is summa-
rized below.
where ␥ is the angle between the incident-ray propagation The validity analysis of the EPS approximation is based
direction and the radial vector pointing from the center of the on the WKB technique [Eq. (3)] from which Eq. (6) is de-
particle. After performing the integration in Eq. (3), s共兲共兲 rived. For an inhomogeneous spherical particle, the relative
for a homogeneous spherical particle is given by phase shift 共r⬘兲 can be expressed as 共r⬘兲 = 0共r⬘兲 + ␦共r⬘兲.
s共兲共兲 = 2S关1 − 2n0 sin / + 4n0 sin2共/2兲/2兴, 共6兲 Here, 0共r⬘兲 = 2共n0 − 1兲L共r⬘兲 / is the phase shift of a light
ray propagating through the homogeneous counterpart of the
where = 2d共n0 − 1兲 / is the maximum phase shift pro- particle. The term ␦共r⬘兲 accounts for the phase-shift differ-
duced by the homogeneous sphere. ence due to refractive-index inhomogeneity. If
We note Eq. (6) becomes equivalent to the van de Hulst
approximation [12] ␦共r⬘兲 ⬍ /2, 共8兲
s共兲 = 2S关1 − 2 sin / + 4 sin2共/2兲/2兴 共7兲 the exponent in Eq. (3) can be expanded to perform the in-
tegration analytically. This yields
for spheres with low refractive indexes. The most distinctive
feature that can be observed from both Eqs. (6) and (7) is the s共兲 ⬇ n共0兲 + ␦共兲 , 共9兲
“interference structure” [13], which refers to slow oscilla-
tions of TSCS as a function of wavelength with the fre- where n共兲共兲 = 2S关1 − 2n0 sin / + 4n0 sin2共 / 2兲 / 2兴 is the
0
quency of these oscillations proportional to the diameter of scattering produced by the equiphase-sphere counterpart of
the particle. With sufficiently large , the higher order term the particle, and ␦共兲 is the error term produced by
sin2共 / 2兲 / 2 can be neglected; thus the diameter of the par- refractive-index inhomogeneity. The EPS approximation is
共兲
ticle can be easily derived from the oscillation frequency by valid provided that ␦共兲 Ⰶ EPS .
d = 12 / 共2 − 1兲 / 共n0 − 1兲, where 1 and 2 are wavelengths We point out that the expansion in Eq. (9) depends on
corresponding to two adjacent maxima or minima in the condition (8). Thus, we shall examine the inequality (8) in
TSCS spectrum. In addition, by including a surface term [Eq. detail. The phase shift error ␦共r⬘兲 due to inhomogeneity is
(2)] and implicitly incorporating the refraction effect on the given by
056610-2
EQUIPHASE-SPHERE APPROXIMATION FOR LIGHT … PHYSICAL REVIEW E 70, 056610 (2004)
␦共r⬘兲 = 冕
L共r兲
2
␦n共r⬘,l兲dl, 共10兲
have more significant impact on the accuracy of the EPS
approximation.
056610-3
LI et al. PHYSICAL REVIEW E 70, 056610 (2004)
Cn共r兲 = 具关n共0兲 − n0兴关n共r兲 − n0兴典, 共13兲 Figures 1(a)–1(c) graph shows sample inhomogeneous
spherical particles with fixed n0 = 1.5 and correlation length
Lc = 400 nm, but increasing standard deviations ranging from
n = 0.05 [Fig. 1(a)] to n = 0.163 [Fig. 1(c)]. In each figure,
where r = 兩r兩. In this paper, we use the Gaussian function as the particle refractive-index distribution is depicted in the 3D
the correlation model view of a surface plot (left panel), a cross-sectional cut in the
2/共L /2兲2
x̂-ẑ plane (middle panel), and a cross-sectional cut in the
Cn共r兲 = e−r c , 共14兲 ŷ-ẑ plane (right panel). Each colormap of the particle interior
illustrates the spatial distribution of the particles’ refractive
indices. The corresponding scale of the variation is illus-
where Lc is the characteristic correlation length representing trated using the colorbars displayed on the right in each fig-
ure. We note that the exact geometry of the refractive-index
the length scale over which the correlation drops to a negli-
spatial distribution is unique for each case since the stochas-
gible level. For such a choice of correlation function, the
tic method is used in randomly generating these geometries.
statistics of the spatial distribution of n共r兲 is uniquely deter- However, since the three particles all have the same correla-
mined by the parameter Lc. If Lc → ⬁, we have Cn共r兲 ⬅ 1 and tion length Lc = 400 nm, their spatial refractive-index distri-
the resulting spatial distribution is homogeneous. Lower Lc butions have fluctuations on the same geometrical scale. It is
corresponds to refractive-index fluctuations in smaller geo- also evident that the standard deviation n determines the
metric scales. magnitude of the refractive-index fluctuation. For example,
Various methods can be used to generate realizations of for the particle shown in Fig. 1(c), the standard deviation
the GRF model. In this paper, we have adopted the turning- n = 0.163 results in refractive-index fluctuations raging ap-
band method [16], where the 3D realizations of the GRF proximately from 1.0 to 2.0.
model are generated by summing independent realizations of Figure 2 shows sample inhomogeneous spherical particles
one-dimensional random functions with directional vectors with fixed n0 = 1.5 and n ⬇ 0.1 but Lc increasing from
uniformly distributed over the unit sphere. Using this 100 nm [Fig. 2(a)] to 1.2 m [Fig. 2(c)]. These examples
method, we create geometrical models of spherical particles demonstrate the capability of the GRF model to mimic
with the refractive index having GRF distributions. refractive-index fluctuations occurring over a wide range of
056610-4
EQUIPHASE-SPHERE APPROXIMATION FOR LIGHT … PHYSICAL REVIEW E 70, 056610 (2004)
geometrical scales appropriate for simulation of natural scribed in our previous work [9], we calculated TSCS spectra
particles. of (n0 = 1.5, d = 4 m) spherical particles ranging from
slightly inhomogeneous 共1.45艋 n 艋 1.55兲 to highly inhomo-
V. NUMERICAL VALIDATION OF THE EPS geneous 共1.0艋 n 艋 2.0兲. The numerical experiments include
APPROXIMATION AND ITS EXPECTED RANGE a wid range of geometrical scales of interior refractive-index
OF VALIDITY fluctuations with Lc ranging from 50 nm to 1.2 m.
Figure 3 shows four representative results of our numeri-
We now describe our validation of the EPS approximation
cal experiments. In each example, the spatial distribution of
discussed in Secs. II and III. To this end, we have conducted the particle refractive index in one cross-sectional cut is dis-
a series of numerical experiments that compared TSCS spec- played on the left, and the TSCS spectra calculated with
tra calculated using the EPS approximation with numerical FDTD and the EPS approximation are graphed on the right.
FDTD benchmark data for a wide variety of inhomogeneous We note that although these inhomogeneous particles have a
spherical particles such as shown in Figs. 1 and 2. This com- variety of values of n and Lc, the validity condition 
parison permits us to validate the EPS approximation and to ⬅ 4冑Lcdn / ⬍ 1 is satisfied for all four cases. Indeed, the
explore the correlation between the approximation accuracy TSCS spectra calculated by the EPS approximation very well
and the geometric characteristics of the refractive-index dis- matched the benchmark data provided by FDTD for all four
tribution. cases.
The FDTD method has been shown to be a robust means As Lc or n of the interior refractive-index distribution
to numerically solve the Maxwell’s equations in studies of become greater, the accuracy of the EPS approximation is
light scattering problems [17]. We used a staircasing scheme expected to decline. This effect is illustrated in Fig. 4. For
with 25-nm resolution to sample the refractive-index spatial these four particles with increasing , the EPS-calculated
variations of interest. Following the same procedures as de- TSCS spectra progressively deviate from the FDTD data. In
056610-5
LI et al. PHYSICAL REVIEW E 70, 056610 (2004)
an extreme case shown in Fig. 4(d), where both the magni- rameters of refractive-index distribution (characterized by
tude and geometrical scale of the refractive-index inhomoge- the  factor) on the validity and accuracy of the EPS method.
neity are large (1.0艋 n 艋 2.0, Lc = 1.0 m, and  = 2.6), the We used two complimentary parameters, the rms error R and
oscillatory period of the TSCS spectrum calculated by the the correlation coefficient rc to quantify the accuracy of the
EPS approximation completely departs from the FDTD approximate EPS-calculated TSCS spectra with respect to
benchmark data. the exact FDTD benchmark data. The rms error measures the
We summarize our numerical experiments with a para- overall estimation accuracy, while the correlation coefficient,
metric study to demonstrate the impact of the statistical pa- which is defined as
measures the capability of the EPS approximation to repli- covering a wide variety of refractive index distributions (Lc
cate the oscillation characteristics of the TSCS spectrum. ranging from 50 nm to 1.2 m and n ranging from 0.02 to
Figure 5 plots the rms error R and the correlation coeffi- 0.163). In order to better illustrate the connection between
cient rc as functions of  for 26 inhomogeneous spheres the quality of the EPS approximation and the accuracy mea-
056610-6
EQUIPHASE-SPHERE APPROXIMATION FOR LIGHT … PHYSICAL REVIEW E 70, 056610 (2004)
sures, we cross-reference eight data points in Figs. 5(a) and can accurately model the TSCS of randomly inhomogeneous
5(b) with their corresponding particle geometries and TSCS spherical particles having internal refractive index variations
spectra shown in Fig. 3 and Fig. 4. with geometrical scales spanning from nanometers (i.e., sub-
We observe from Fig. 5 that when criterion (9) is satisfied wavelength) to microns (i.e., suprawavelength). An easy-to-
共 ⬍ 1兲, the EPS approximation is sufficiently accurate, i.e., use criterion for the range of approximation validity has been
rc 艌 0.9 and R ⬍ 5%. It is also evident from Fig. 5(a) that provided to guide the practical application of this method.
when  ⬎ 1, the accuracy of the EPS approximation degrades Although not limited to a single category of applications,
rapidly as  increases. This further demonstrates the impor- the work discussed here may positively impact tissue optical
tance of the  parameter in determining the validity of the imaging and diagnostic applications. It is recognized that the
EPS approximation. analysis of spectral, angular, and other characteristics of light
scattered from living tissue can provide valuable diagnostic
information [18–25]. Due to the complexity in the microar-
chitecture of biological tissue, the understanding of light
scattering by particles with complex shapes and interior
VI. SUMMARY AND DISCUSSION
structure is of great importance for the future refinement of
We have presented the development and validation of the the current optical techniques. Importantly, the development
equiphase-sphere (EPS) approximation for the total- of methods such as the EPS approximation to analyze tissue
scattering cross-section (TSCS) spectra of inhomogeneous light scattering will enable gathering new accurate informa-
spherical particles having complex interior structures. We tion about tissue organization and its alteration in disease. In
have shown that the closed-form, analytical approximation turn, these insights can be further used for disease diagnosis
056610-7
LI et al. PHYSICAL REVIEW E 70, 056610 (2004)
056610-8