Computer simulations of optical

turbulence in the weak- and strong-

scattering regime: angle-of-arrival
fluctuations obtained from ray optics
and wave optics

David Voelz
Erandi Wijerathna
Andreas Muschinski
Xifeng Xiao

David Voelz, Erandi Wijerathna, Andreas Muschinski, Xifeng Xiao, “Computer simulations of optical
turbulence in the weak- and strong-scattering regime: angle-of-arrival fluctuations obtained from ray
optics and wave optics,” Opt. Eng. 57(10), 104102 (2018), doi: 10.1117/1.OE.57.10.104102.

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 05 Sep 2023

Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 Optical Engineering 57(10), 104102 (October 2018)

Computer simulations of optical turbulence in

the weak- and strong-scattering regime: angle-of-arrival
fluctuations obtained from ray optics and wave optics
David Voelz,a,* Erandi Wijerathna,a Andreas Muschinski,b,c and Xifeng Xiaoa
a
New Mexico State University, Klipsch School of Electrical and Computer Engineering, Las Cruces, New Mexico, United States
b
NorthWest Research Associates, Boulder, Colorado, United States
c
University of Colorado Boulder, Aerospace Engineering Department, Boulder, Colorado, United States

Abstract. It is known that certain geometrical-optics predictions often agree well with optical turbulence field
observations even though theoretical constraints for ignoring diffraction may be violated. Geometrical optics
assumptions can simplify analyses, and ray optics can significantly reduce simulation computation time.
Here, an investigation into angle-of-arrival fluctuations is presented involving wave optics and geometrical
(ray) optics computer simulations of a plane wave of visible light propagating through a turbulent refractive-
index field. The simulation and Rytov-based theory results for the variances of aperture-filtered angle-of-arrival
fluctuations generally agree well for weak scattering (Rytov variance, σ 2R ≲ 0.2), but for increasing Rytov vari-
ance, the simulation results demonstrate a positive slope that can be significantly shallower than that predicted
by the theory. For weak-to-moderate scattering regimes (σ 2R ≲ 2.67), a comparison of the ray and wave results
show they match for aperture diameters greater than about two Fresnel lengths. This result is consistent with
a previous theoretical analysis by Cheon and Muschinski. For the strongest scattering case studied (σ 2R ¼ 26.7),
the wave and ray simulations match for aperture diameters greater than about 10 Fresnel lengths. For
smaller apertures, we attribute the disparity between the wave and ray simulation results to a Fresnel filtering
effect. © The Authors. Published by SPIE under a Creative Commons Attribution 3.0 Unported License. Distribution or reproduction of this work in
whole or in part requires full attribution of the original publication, including its DOI. [DOI: 10.1117/1.OE.57.10.104102]
Keywords: atmospheric turbulence; Rytov variance; angle of arrival; wave optics; split-step simulation; ray optics.
Paper 180577 received Apr. 23, 2018; accepted for publication Aug. 29, 2018; published online Oct. 5, 2018.

1 Introduction through turbulence. The utility of geometrical and ray optics

Understanding the physics of optical wave front distortions is exemplified in a variety of publications from the last
caused by turbulence in the atmospheric refractive-index several decades. For example, Churnside and Lataitis10
field is essential for the design and operation of optical sys- presented a simple geometrical optics equation for the vari-
tems used in free-space optical communication, surveillance, ance of beam displacements caused by propagation through
navigation, remote sensing, astronomy, and directed-energy weak turbulence, and geometrical optics simulation models
technologies.1–6 If particle scattering, absorption, coupling, have been developed by Yuksel et al.11 to investigate effects
and depolarization effects are negligible then there are of aperture averaging on intensity fluctuations. More
two phenomena that determine the optical field: refraction recently, Lachinova et al.12 have compared the performance
and diffraction. The combined effects of refraction and of a brightness function-based numerical simulation method
diffraction lead to phase-gradient fluctuations (angle-of- with a wave optics-based method for incoherent imaging
arrival fluctuations) and intensity fluctuations (scintillation), through turbulence. The brightness function at the receiving
both of which limit the performance of a variety of optical pupil is estimated using ray-like trajectories that are per-
systems. turbed by the turbulent refractive index gradients. These
The analysis of propagation scenarios through both deter- authors found that the brightness approach can reduce
ministic and stochastic refractive-index fields may be sub- computation times by several orders of magnitude. Other
stantially simplified if diffraction effects can be neglected, recent examples include ray-tracing methods applied for
and considerable theoretical effort has been devoted to estab- simulating laser beam-steering effects in turbulent media,13
lishing criteria for the negligibility of diffraction effects.1–3,7,8 the derivation of turbulence strength profiles from radiometer
These theoretical constraints, however, tend to be exces- temperature measurements with a geometrical ray-tracing
sively conservative, as pointed out by Strohbehn 50 years analysis,14 and the application of ray tracing through phase
ago.9 In other words, certain geometrical-optics predictions screens for the simulation of images from optical survey
often agree well with field observations, but it remains telescopes.15
unclear, from the theoretician’s perspective, why that is so. Since the 1980s, powerful phase-screen techniques have
In addition to simplifying theoretical analyses, geometri- been developed16–19 that allow optical propagation through
cal optics and ray tracing can also substantially reduce com- atmospheric turbulence to be accurately simulated well into
putation time for numerical simulations of propagation the “strong-scattering” regime, that is, for scenarios where
the negligibility of diffraction effects and the “weak-scatter-
ing” assumption become invalid, or at least questionable.
*Address all correspondence to: David Voelz, E-mail: [email protected] Such simulations can be used to test and refine the

Optical Engineering 104102-1 October 2018 • Vol. 57(10)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 05 Sep 2023

Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 Voelz et al.: Computer simulations of optical turbulence in the weak- and strong-scattering regime. . .

traditional, often excessively conservative, criteria for the (3) to investigate the consequences of ignoring diffraction
negligibility of diffraction effects. and the applicability of ray tracing for modeling AOA
Here, we present and discuss computer simulations of fluctuations.
plane waves of visible light (λ ¼ 500 nm) propagating
through fully turbulent, homogeneous, and isotropic refrac- 2 Approach
tive-index fields. We study the optical angle-of-arrival A plane wave is propagated through a set of split-step phase
(AOA) fluctuations by means of both a wave-optics propa- screens that model atmospheric turbulence [Fig. 1(a)] and
gation simulation and a ray-tracing simulation through a similarly, a bundle of parallel rays is traced through the
turbulent path that is characterized by a sequence of phase same screens [Fig. 1(b)]. The wave optics simulation pro-
screens with Kolmogorov statistics. The path length assumed vides a coherent wave result that accounts for refraction, dif-
is L ¼ 2 km and the refractive-index structure parameter C2n fraction, and interference. On the other hand, the ray-tracing
ranges from 10−16 m−2∕3 to 10−13 m−2∕3 , which corresponds simulations provide the trajectories of a large number of
to Rytov variances σ 2R that span from 0.0267 (weak scatter- individual rays, where the phases and amplitudes are left
ing) to 26.7 (strong scattering), where σ 2R ¼ 1.23C2n k7∕6 L11∕6 undefined.
and k ¼ 2π∕λ is the optical wave number. For the two sim- The simulation parameters of interest are defined in
ulation methods, we examine the variance and histograms of Table 1 and the values used for the simulations are listed.
the aperture-filtered (aperture-averaged) AOA fluctuations The turbulence along the path is assumed to be statistically
for aperture diameters ranging from 0.6 to 13 Fresnel homogeneous and isotropic. The path is divided into equal
lengths. We compare the simulation results with theoretical length segments and a phase screen is positioned at the center
predictions20 based on the Rytov theory, which accounts for of each segment. The screens are created with the Fourier
both refraction and diffraction effects but does not capture filtering method and have Kolmogorov statistics. To model
strong-scattering effects. Our specific interests are: (1) to bet- an infinite outer scale, a random tilt component is added to
ter understand under what conditions the Rytov theory for each screen that compensates for turbulent wave-front tilts
AOA fluctuations becomes invalid, (2) to reveal the behavior with length scales greater than the grid width (see Ref. 18
of the AOA fluctuations in strong scattering conditions, and and references therein). Zero inner scale is assumed for

Fig. 1 Illustration of the split-step phase screens that model turbulence in (a) wave and (b) ray
simulations. L is the propagation path length and D is the diameter of the aperture that is applied in
the observation plane for the aperture-averaged AOA.

Optical Engineering 104102-2 October 2018 • Vol. 57(10)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 05 Sep 2023

Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 Voelz et al.: Computer simulations of optical turbulence in the weak- and strong-scattering regime. . .

Table 1 Simulation parameter values. 2.1 Wave Optics Simulation

The initial plane wave at the entrance plane, z ¼ 0, is a
Parameter Notation Value unit-amplitude, zero-phase field that is propagated in steps
between the entrance plane and the observation plane using
Grid physical size S×S 1 m×1 m a discretized form of the Fresnel diffraction integral.21 At the
Grid samples 1000 × 1000 i’th intermediate plane, the phase screen ψ i is applied in the
usual way as u 0 i ¼ ui expðjψ i Þ, where ui is the incident field
Grid spacing Δx 1 mm and u 0 i is the field that exits the plane. The discrete Fourier
transform-based simulation produces a periodicity of the
Wavelength λ 0.5 μm
electric field in the x-y plane across the grid boundaries.
Path length L 2 km In addition, the compensating random tilt component in
pffiffiffiffiffi the turbulence phase screens creates some “ringing” artifacts
Fresnel length λL 3.2 cm in the receiving plane near the grid boundaries. To minimize
unwanted periodicity and edge effects, the aperture is only
Aperture diameter D 0.02 to 0.4 m
applied to the center 0.4 × 0.4 fraction of the receiving
Turbulence spectrum Kolmogorov plane grid.
One approach to find the aperture-averaged AOA for
Structure parameter C 2n 10−16 to 10−13 m−2∕3 a given turbulence realization is to numerically compute
Rytov variance σ 2R 0.0267 to 26.7
the field phase gradients at the observation plane grid points
and average the results within the aperture region. However,
Number of phase screens N 6 (for 10−16 m−2∕3 ); direct calculation of the phase gradients at the grid points is
6 (for 10−15 m−2∕3 ); problematic for strong scattering because of branch points
10 (for 10−14 m−2∕3 ); (implying discontinuous phase changes from 2π to 0, or
22 (for 10−13 m−2∕3 )
0 to 2π) that appear in the wave front and correspond to theo-
retically infinite phase gradients.
For the results presented here, we use an approach where
the turbulence spectrum; however, the period of the highest the wave front within the aperture for a given turbulence
spatial frequency of a screen is twice the grid spacing, so the realization is numerically focused with a focus transmittance
modeled inner scale is no smaller than 2 mm. function.21 The x- and y-direction centroid positions of the
The number of turbulence screens N required along the intensity spot at the focal plane are found and the aperture-
path is an integer and should be greater than ð10σ 2R Þ6∕11 , averaged AOA is the angle defined by the ratio of the
centroid displacement (from the optical axis) and the focal
which ensures each screen corresponds to weak scattering,
length. This AOA is computed for many independent
and therefore phase-only screens are adequate.19 The mini-
turbulence realizations and the variance is calculated. This
mum value of N for the strongest scattering case is 22
approach has the advantage that it is representative of a prac-
screens. Fewer screens can be used for the weaker scattering tical AOA measurement that can be made with a lens and
cases, which reduces computational time. However, we camera system.22 It also effectively weighs the angle contri-
implemented more than the required minimum number of butions by the spatial field amplitude in the pupil, where,
screens for the weak scattering cases to provide a better for example, the amplitude is vanishing near a branch
representation of the continuous volume turbulence that is point so the sometimes extremely large gradients associated
assumed along the path. with these features are not significant for the AOA variance
The grid parameters are chosen to provide a grid spacing calculation.
of Δx ¼ λL∕S, where S is the grid side length. This sampling
choice has been shown to make artifacts that arise from the
2.2 Ray Tracing Simulation
discrete Fourier transform implementation of the Fresnel dif-
fraction integral to be sufficiently small to be negligible.21 For the ray simulation, as shown in Fig. 1(b), the source is
The resulting grid spacing Δx of 1 mm allows about 3× sam- comprised of a set of rays where each ray is positioned at
pling of the minimum Fried parameter (r0 ) value encoun- a grid point corresponding to the wave simulation and is
tered in our study. The propagation length L, wavelength initially directed parallel to the optical axis. For a given
λ, and four choices of turbulence structure parameter turbulence realization, the rays are traced from one plane
C2n (10−16 , 10−15 , 10−14 , and 10−13 m−2∕3 ) correspond to the next where at the i’th plane, the phase screen changes
the x-direction ray angle as
to Rytov variances σ 2R of 0.0267, 0.267, 2.67, and 26.7,
which range in scattering strength from weak (σ 2R ≪ 1) 1 dψ i
to strong (σ 2R > 1). Corresponding Fried diameter [r0 ¼ θ 0 i ¼ θi þ
EQ-TARGET;temp:intralink-;e001;326;183 ; (1)
k dx
ð0.423 k2 C2n LÞ−3∕5 ] values for the four cases are calculated
to be 21.3, 5.3, 1.33, and 0.33 cm. where θi and θ 0 i are the incoming and outgoing ray angles
Ultimately, our interest is to examine the variance and and dψ
dx is the phase-screen gradient in the x-direction. Ray
i

histogram of the fluctuations of the aperture-averaged AOA. angle changes in the y-direction are computed similarly
This requires running the simulation many times for using the y-direction gradients. A three-point calculation is
independent turbulence realizations, obtaining the aperture- used to compute the screen gradients, and the gradients are
averaged AOA for each realization, and then computing interpolated at the position where the ray strikes the plane.
the variance. The x- and y-direction angles for the rays that eventually fall

Optical Engineering 104102-3 October 2018 • Vol. 57(10)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 05 Sep 2023

Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 Voelz et al.: Computer simulations of optical turbulence in the weak- and strong-scattering regime. . .

within the aperture region at the observation plane are aver- 3 Theory
aged to obtain the aperture-averaged AOA. Our approach for The variance of the aperture-filtered AOA fluctuations for a
tracing rays through the split-step phase screens is similar to circular aperture and for a plane wave propagated through
the method presented by Lachinova et al.,12 however, we do homogeneous and locally isotropic inertial subrange turbu-
not compute the brightness function that is part of their lence is given by Cheon and Muschinski20 as
incoherent image simulation. 1 7 5 1
hθ2 iP ¼ γ P ðqÞC2n LD−3 ¼ 0.81γ P ðqÞσ 2R k−6 L−6 D−3 ;
EQ-TARGET;temp:intralink-;e002;326;694 (2)

Fig. 2 Observation plane examples. Left-to-right in each row: ray-density, wave-intensity, and wave-
phase cross-sections. Turblulent scattering strength increasing from top-to-bottom along columns:
(a–c) σ 2R ¼ 0.0267, C 2n ¼ 10−16 m−2∕3 , r 0 ¼ 21.3 cm; (d–f) σ 2R ¼ 0.267, C 2n ¼ 10−15 m−2∕3 , r 0 ¼ 5.3 cm;
(g–i) σ 2R ¼ 2.67, C 2n ¼ 10−14 m−2∕3 , r 0 ¼ 1.33 cm; (j–l) σ 2R ¼ 26.7, C 2n ¼ 10−13 m−2∕3 , r 0 ¼ 0.33 cm.
Ray density binning: (a) 10 × 10 bins, (d) 8 × 8 bins, (g) 3 × 3 bins, and (j) 1 × 1 bins.

Optical Engineering 104102-4 October 2018 • Vol. 57(10)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 05 Sep 2023

Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 Voelz et al.: Computer simulations of optical turbulence in the weak- and strong-scattering regime. . .

where scattering for larger apertures (e.g., σ 2R ¼ 2.67 and q ≳ 10).

pffiffiffi      1   1 The over-prediction of the AOA variances by the Rytov
3 1 8 β −3 6 π 6 1 1
EQ-TARGET;temp:intralink-;e003;63;741

γ P ðqÞ ¼ Γ Γ 1þ β 3 q3 theory is likely a result of the well-known failure of the

16 6 3 2 5 2 Rytov approximation in the strong-moderate- to strong-scat-
 5   
π 2 4 4 12 5 2 tering regime. We note that the Rytov theory is also known to
× 1þ β q sin arctan ; (3) over-predict intensity scintillation as a function of the Rytov
4 6 πβ2 q2
variance in the moderate- and strong-scattering regimes.
pffiffiffiffiffiffi In general, the ray-based and wave-based simulation var-
where D is the aperture diameter, q ¼ D∕ λL is the aperture
diameter normalized by the Fresnel length, β ¼ 0.5216 is iances compare favorably in Fig. 3. The largest discrepancy
a constant, and Γ is the gamma function. Equations (2) is a 15% higher variance for the rays in strong scattering with
and (3) were developed assuming the Rytov approximation a small aperture (σ 2R ¼ 26.7 and q ¼ 0.63). For the weak-to-
for the propagating field. This result is applicable for all moderate scattering cases [Figs. 3(a)–3(c)], the ray and wave
values of q and assumes infinite outer scale and zero results match well for q ≳ 2. That is, diffraction effects are
inner scale. negligible in the weak- and moderate-scattering regime if the
aperture diameter is larger than about twice the Fresnel
4 Results length. This result is consistent with the theoretical analysis
by Cheon and Muschinski20 who found that for weak scat-
4.1 Observation Plane Illustration tering, the AOA variance predicted by geometrical optics
Before presenting the AOA results, we first examine the deviates from that predicted by the Rytov theory by <1%
character of the rays and waves arriving at the observation if q > 1.65. For smaller q, the wave simulations predict a
plane for a single turbulence realization. Figure 2 shows sin- lower AOA variance than the ray simulations. For the strong-
gle realization examples within the central 0.4 m × 0.4 m est scattering case [Fig. 3(d)], the wave and ray variance
area for the scattering scenarios corresponding to σ 2R ¼ values converge for q ≳ 10.
0.0267, 0.267, 2.67, and 26.7. Shown for comparison are: Figure 4 shows an alternative presentation of the results
(1) ray-density fields obtained from the ray simulation where aperture-averaged AOA variances for the small aper-
and (2) intensity and phase fields obtained from the wave ture (q ¼ 0.63; D ¼ 0.02 m) and the large aperture (q ¼
optics simulation. The ray density fields were obtained by 12.6; D ¼ 0.4 m) are plotted as a function of the Rytov vari-
counting the rays within areas defined by the binning values. ance, σ 2R . The Rytov theory curves demonstrate the linear
The binning is relative to the grid spacing Δx, for example, dependence described by Eq. (2). For the small aperture
10 × 10 binning indicates an area of 10 mm × 10 mm. [Fig. 4(a)], the simulations show a positive slope as a func-
The bin size was selected to emphasize the most apparent tion of scattering strength that is significantly shallower than
characteristic feature sizes within the frame. A contour the theory line. For large aperture diameters, the Rytov-based
plot algorithm was also applied to the ray density plot to theory is expected to collapse to the geometrical optics
reduce the blockiness of the display. solution.20 This behavior is suggested in the large aperture
Common spatial features are apparent in the intensity and case [Fig. 4(b)], where the simulation results follow the
ray-density fields as shown in Fig. 2. The corresponding theory line much more closely.
features are related to the effects of refraction as ray tracing Figure 4 also shows a small difference between the
does not account for the interference and diffractive effects ray-based and wave-based results for the small aperture
provided in the wave simulation. The phase cross-sections, (q ¼ 0.63), but essentially no difference for the large aper-
displayed modulo-2π, illustrate the increasing wave front ture (q ¼ 12.6). As noted previously, geometrical optics is
complexity as a function of the Rytov variance. As noted known to be questionable for q < 1, which is consistent with
previously, after the ray and wave fields are generated at the fact that our wave-optics simulations produce smaller
the observation plane as shown in Fig. 2, a circular aperture AOA variances than the AOA variances produced by our
of diameter D is imposed and the methods described in ray-tracing simulations. We attribute this to “Fresnel filtering.”
Sec. 2 are applied to find the aperture-averaged AOA field The Fresnel-filter kernel in the wave-number representation of
for the particular turbulence realization. the AOA variance for plane waves predicted by the Rytov
theory has the form K F ðxÞ ¼ 1 þ ð2π∕x2 Þ sinðx2 ∕2πÞ,20
4.2 Variances of Aperture-Filtered Angle-of-Arrival where x ¼ κðλLÞ1∕2 is the wave number normalized by the
Fluctuations Fresnel wave number 1∕ðλLÞ1∕2. The case x ≪ 1 (Fresnel
The variances of aperture-averaged AOA fluctuations are length is small compared with κ−1 , the wave-front tilt length
computed from many independent turbulence realizations. scale under consideration) is the geometrical-optics limit
Figure 3 shows results for the theoretical [Eq. (2)] and simu- where the Fresnel length has no effect on the AOA variance.
lated aperture-averaged AOA variances as a function of the For x comparable to and >1, the Fresnel length (and therefore
normalized aperture diameter q for the four values of Cn 2 diffraction) does play a role: K F ðxÞ becomes smaller than
listed in Table 1. Each data point was estimated from K F ðxÞ in the geometrical-optics limit. Therefore, diffraction
1000 turbulence realizations. The general behavior is that causes the AOA variance predicted by wave optics to be
the AOA variance decreases as a function of increasing aper- smaller than the AOA variance predicted by geometrical optics.
ture diameter. Both simulation results generally agree well That is, a finite Fresnel length acts as a characteristic filter
with the AOA theory20 for weak scattering (e.g., σ 2R ≲ 0.2) scale similar to the filtering effects due to a finite inner tur-
but decrease significantly below the theory curve as the bulence length scale or a finite aperture diameter. Essentially,
Rytov variance increases. The results also show the agree- diffraction causes wave-front tilt components with length
ment between theory and simulation is better in stronger scales comparable to and smaller than the Fresnel length

Optical Engineering 104102-5 October 2018 • Vol. 57(10)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 05 Sep 2023

Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 Voelz et al.: Computer simulations of optical turbulence in the weak- and strong-scattering regime. . .

(a) (b)
10 -12 10 -11
2 2
Rytov theory Rytov theory
1.8 Wave (focus centroid) 1.8 Wave (focus centroid)
Ray (angle) Ray (angle)
1.6 1.6

1.4 1.4

1.2 1.2

1 1

0.8 0.8

0.6 0.6
0 2 4 6 8 10 12 0 2 4 6 8 10 12
q q

(c) -10 (d) -9

10 10
2 2
Rytov theory Rytov theory
1.8 Wave (focus centroid) 1.8 Wave (focus centroid)
Ray (angle) Ray (angle)
1.6 1.6

1.4 1.4

1.2 1.2

1 1

0.8 0.8

0.6 0.6
0 2 4 6 8 10 12 0 2 4 6 8 10 12
q q

Fig. 3 Aperture-averaged AOA simulation results and Rytov-based theory as a function of normalized
aperture size q: (a) σ 2R ¼ 0.0267, C 2n ¼ 10−16 m−2∕3 , (b) σ 2R ¼ 0.267, C 2n ¼ 10−15 m−2∕3 , (c) σ 2R ¼ 2.67,
C 2n ¼ 10−14 m−2∕3 , and (d) σ 2R ¼ 26.7, C 2n ¼ 10−13 m−2∕3 .

scales to be smoothed out. The Fresnel-filtering effect is sig- is accounted for in the wave-optics simulations but not in
nificant when the aperture size and the inner scale of turbu- the ray-tracing simulations.
lence are both comparable to or smaller than the Fresnel To examine the AOA distribution results in detail, we
length, and it is negligible otherwise. The Fresnel filtering generated histogram estimates of the probability density

(a) -9 (b) -10

10 10
2 Rytov theory 8 Rytov theory
Wave (focus centroid) Wave (focus centroid)
7
Ray (angle) Ray (angle)
1.5 6

1 4

0.5 2

0 0
0 5 10 15 20 25 0 5 10 15 20 25
2 2
R

Fig. 4 Aperture-averaged AOA simulation results and Rytov-based theory as a function of σ 2R for:
(a) q ¼ 0.63; D ¼ 0.02 m and (b) q ¼ 12.6; D ¼ 0.4 m.

Optical Engineering 104102-6 October 2018 • Vol. 57(10)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 05 Sep 2023

Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 Voelz et al.: Computer simulations of optical turbulence in the weak- and strong-scattering regime. . .

Fig. 5 Statistics of aperture-filtered AOA fluctuations for weak scattering, σ 2R ¼ 0.0267. Left-to-right in
each row: q ¼ 0.63 (D ¼ 0.02 m), q ¼ 6.3 (D ¼ 0.2 m), and q ¼ 12.6 (D ¼ 0.4 m). (a–c) pdf estimates
and (d–f) corresponding q–q plots.

functions (pdfs). Figures 5(a)–5(c) show the histogram plots quantile–quantile (q–q) plots23 shown in Figs. 5(d)–5(f).
for the weakest scattering case and for three aperture diam- The q–q plots compare the simulation sample distribution
eters. To reduce noise in the histogram, we used 10,000 real- to a standard normal (Gaussian) distribution. In all cases,
izations for these results. We also generated corresponding the ray and wave samples create straight lines that indicate

Fig. 6 Statistics of aperture-filtered AOA fluctuations for strong scattering, σ 2R ¼ 26.7. Left-to-right in
each row: q ¼ 0.63 (D ¼ 0.02 m), q ¼ 6.3 (D ¼ 0.2 m), and q ¼ 12.6 (D ¼ 0.4 m). (a–c) pdf estimates
and (d–f) corresponding q–q plots.

Optical Engineering 104102-7 October 2018 • Vol. 57(10)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 05 Sep 2023

Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 Voelz et al.: Computer simulations of optical turbulence in the weak- and strong-scattering regime. . .

Gaussian distributions. The two different slopes in the q–q References

curves in Fig. 5(d) correspond to a difference in the AOA 1. V. I. Tatarskii, Wave Propagation in a Turbulent Medium, pp. 285,
variance for the wave and ray results, which was discussed McGraw-Hill, New York (1961).
previously for small q. Histogram and q–q plots for strong 2. V. I. Tatarskii, The Effects of the Turbulent Atmosphere on Wave
Propagation, pp. 472, Israel Program for Scientific Translation,
scattering are shown in Fig. 6. These distributions are also Jerusalem (1971).
Gaussian and q–q plot slope differences in Figs. 6(d)–6(f) 3. A. D. Wheelon, Electromagnetic Scintillation I—Geometric Optics,
pp. 455, Cambridge University Press, Cambridge, United Kingdom
are consistent with the variance results in Fig. 3(d). We (2001).
expect a Gaussian distribution for the aperture-averaged 4. A. D. Wheelon, Electromagnetic Scintillation II—Weak Scattering,
AOA if the propagation through turbulence and aperture pp. 440, Cambridge University Press, Cambridge, United Kingdom
(2003).
averaging represents a large sum of independent tilt compo- 5. L. C. Andrews and R. L. Phillips, Laser Beam Propagation through
nents such that the central limit theorem applies. Random Media, 2nd ed., pp. 783, SPIE Press, Bellingham,
Washington (2005).
6. O. Korotkova, Random Light Beams: Theory and Applications, pp. 361,
CRC Press, Boca Raton, Florida (2014).
5 Conclusions 7. A. Sommerfeld and I. Runge, “Anwendung der Vektorrechnung auf die
Grundlagen der geometrischen Optik,” Ann. Phys. 340, 277–298
The simulations illustrate that after propagation through (1911).
homogeneous turbulence, wave and ray results share obvious 8. D. L. Fried, “Optical resolution through a randomly inhomogeneous
medium for very long and very short exposures,” J. Opt. Soc. Am.
intensity/ray-density and AOA characteristics related to the 56, 1372–1379 (1966).
effects of refraction. We find that both the wave and ray sim- 9. J. W. Strohbehn, “Line-of-sight wave propagation through the turbulent
atmosphere,” Proc. IEEE 56, 1301–1318 (1968).
ulation results generally predict variances of the aperture- 10. J. H. Churnside and R. J. Lataitis, “Wander of an optical beam in
averaged AOA fluctuations that agree well with the Rytov the turbulent atmosphere,” Appl. Opt. 29, 926–930 (1990).
theory for weak scattering (e.g., σ 2R ≲ 0.2), but as scattering 11. H. Yuksel, W. Atia, and C. C. Davis, “A geometrical optics approach for
modeling atmospheric turbulence,” Proc. SPIE 5891, 589109 (2005).
strength (i.e., the Rytov variance) increases, the simulation 12. S. L. Lachinova et al., “Comparative analysis of numerical simulation
results demonstrate a positive slope that can be significantly techniques for incoherent imaging of extended objects through atmos-
pheric turbulence,” Opt. Eng. 56, 071509 (2017).
shallower than that predicted by the weak-scattering (Rytov) 13. Y. Wang and W. D. Kulatilaka, “Optical ray tracing method for simu-
theory. This over-prediction by the Rytov theory is likely a lating beam-steering effects during laser diagnostics in turbulent media,”
result of the failure of the Rytov approximation in the strong- Appl. Opt. 56, E106–E115 (2017).
14. B. E. Vyhnale, “Path profiles of Cn 2 derived from radiometer temper-
scattering regime. ature measurements and geometrical ray tracing,” Proc. SPIE 10096,
Some insights related to the negligibility of diffraction 100961G (2017).
15. J. R. Peterson et al., “Simulation of astronomical images from optical
effects are gained from a comparison of the wave and ray survey telescopes using a comprehensive photon Monte Carlo
simulation results. For weak-to-moderate scattering regimes approach,” Astrophys. J. Suppl. Ser. 218, 14 (2015).
16. D. L. Knepp, “Multiple phase-screen calculation of the temporal behav-
(σ 2R ≲ 2.67), the ray and wave results match well for aper- ior of stochastic waves,” Proc. IEEE 71, 722–737 (1983).
tures where q ≳ 2. This result is consistent with the theoreti- 17. J. M. Martin and S. M. Flatté, “Intensity images and statistics from
cal analysis by Cheon and Muschinski20 who found that for numerical simulation of wave propagation in 3-D random media,”
Appl. Opt. 27, 2111–2126 (1988).
weak scattering, the AOA variance predicted by geometrical 18. J. D. Schmidt, Numerical Simulation of Optical Wave Propagation,
optics deviates from that predicted by the Rytov theory by pp. 149–182, SPIE Press, Bellingham, Washington (2010).
19. X. Xiao and D. Voelz, “On-axis probability density function and fade
<1% if q > 1.65. For the strongest scattering case studied behavior of partially coherent beams propagating through turbulence,”
here (σ 2R ¼ 26.7), the wave and ray variance values agree Appl. Opt. 48, 167–175 (2009).
well for q ≳ 10. For smaller apertures, we find the wave sim- 20. Y. Cheon and A. Muschinski, “Closed-form approximations for the
angle-of-arrival variance of plane and spherical waves propagating
ulation variances are smaller than those produced by the ray through homogeneous and isotropic turbulence,” J. Opt. Soc. Am. A
simulations, and we attribute this to a “Fresnel filtering” 24, 415–422 (2007).
21. D. G. Voelz, Computational Fourier Optics, pp. 63–96, SPIE Press,
effect associated with diffraction. Both the wave and ray Bellingham, Washington (2011).
AOA distributions are found to be Gaussian. 22. Y. Cheon et al., “Angle-of-arrival anemometry by means of a large-
aperture Schmidt–Cassegrain telescope equipped with a CCD camera,”
J. Opt. Soc. Am. A 24, 3478–3492 (2007).
23. W. L. Martinez and A. R. Martinez, Computational Statistics Handbook
Acknowledgments with MATLAB, 3rd ed., pp. 124–132, CRC Press, Boca Raton, Florida
(2016).
Earlier results of this work appeared in the conference
proceedings paper: D. Voelz, E. Wijerathna, X. Xiao, David Voelz is a professor in the Klipsch School of Electrical and
A. Muschinski, “Angle-of-arrival variance of waves and Computer Engineering at New Mexico State University. His research
rays in strong atmospheric scattering: split-step simulation interests include spectral and polarization sensing, laser beam
results,” Proc. SPIE 10410, Unconventional and Indirect propagation through atmospheric turbulence, laser communications,
imaging theory, and astronomical instrumentation development.
Imaging, Image Reconstruction, and Wavefront Sensing He earned his PhD EE degree from the University of Illinois in 1987.
2017, 104100Y (6 September 2017). This work was funded He is a fellow of SPIE.
by Air Force Office of Science Research (AFOSR) (FA9550-
17-C-0021 and FA9550-18-C-0011), Air Force Office of Erandi Wijerathna is a PhD student in the Klipsch School of Electrical
and Computer Engineering at New Mexico State University. Her
Science Research (AFOSR) Multidisciplinary University research interests include laser beam propagation through atmos-
Research Initiatives (MURI) Program (FA9550-12-1-0449 pheric turbulence, imaging theory, and polarimetric lidar. She earned
and FA9550-18-1-0009), National Science Foundation her BS (honors) in physics from the University of Colombo, Sri Lanka,
(NSF) Physical and Dynamics Meteorology Program in 2011. She received her MS degree in physics and in electrical
engineering from New Mexico State University in 2016.
(AGS-1547476), and U.S. Department of Defense (DOD)
High Performance Computing Modernization Program Andreas Muschinski (MS in physics, TU Braunschweig, Germany,
(Frontier Project FP-CFD-FY14-007). 1990; PhD in meteorology, U Hannover, Germany, 1992; habilitation

Optical Engineering 104102-8 October 2018 • Vol. 57(10)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 05 Sep 2023

Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
 Voelz et al.: Computer simulations of optical turbulence in the weak- and strong-scattering regime. . .

in meteorology, U Hannover, Germany, 1998) is an atmospheric Xifeng Xiao received her BS and MS degrees in physics from Xiamen
physicist. He was a CIRES research scientist at the University of University Fujian, China, in 1998 and 2001, respectively. She earned
Colorado at Boulder, Colorado (1998–2004) and a professor of her MS and PhD degrees in electrical engineering from New Mexico
electrical engineering at UMass Amherst (2004–2011). Since 2011, State University in 2004 and 2008, respectively. Currently, she is a
he has been a senior research scientist at NorthWest Research software engineer at Ball Aerospace, Albuquerque, NM. Her research
Associates, Boulder, Colorado. During the last 28 years, he has con- interests include simulation and modeling of free-space laser commu-
ducted research on atmospheric turbulence and atmospheric wave nication, liquid-crystal polarization, and demonstration and implemen-
propagation. tation of AOTF technologies.

Optical Engineering 104102-9 October 2018 • Vol. 57(10)

Downloaded From: https://www.spiedigitallibrary.org/journals/Optical-Engineering on 05 Sep 2023

Terms of Use: https://www.spiedigitallibrary.org/terms-of-use