You are on page 1of 20

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

net/publication/341837205

A Retrospective of Roland Shack’s Global View of Diffraction

Article  in  Applied Optics · June 2020


DOI: 10.1364/AO.392365

CITATION READS
1 120

1 author:

James Harvey
University of Central Florida
199 PUBLICATIONS   2,970 CITATIONS   

SEE PROFILE

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

Solar UltraViolet Imager (SUVI) View project

All content following this page was uploaded by James Harvey on 23 September 2020.

The user has requested enhancement of the downloaded file.


Review Vol. 59, No. 22 / 1 August 2020 / Applied Optics G185

Retrospective of Roland Shack’s global view of


diffraction [Invited]
James E. Harvey
Photon Engineering, LLC, 330 S. Williams Blvd., Suite 222, Tucson, Arizona 86711, USA (jimh@photonengr.com)

Received 17 March 2020; revised 28 May 2020; accepted 1 June 2020; posted 2 June 2020 (Doc. ID 392365); published 9 July 2020

In April of 1972, Professor Roland Shack presented a series of four colloquium talks at the Optical Sciences Center
at the University of Arizona in which he reformulated scalar diffraction theory in terms of the direction cosines of
the propagation vectors of the angular spectrum of plane waves described by the Fourier integral transform of the
diffracting aperture. The fourth lecture, entitled Radiometry and Lambert’s Law, described diffuse reflectance and
surface scatter phenomena as merely a diffraction phenomenon caused by random phase variations in the system
pupil function. In 1974, he elegantly condensed these four lectures into a single colloquium talk entitled A Global
View of Diffraction. This paper is intended to provide a compilation showing the further development of that work
over the last 46 years. © 2020 Optical Society of America
https://doi.org/10.1364/AO.392365 Provided under the terms of the OSA Open Access Publishing Agreement

1. INTRODUCTION One such application is wide-angle surface scatter, which is


The traditional linear systems formulation (Fourier treat- merely a non-paraxial diffraction phenomenon that degrades
ment) of scalar diffraction theory as presented by Bracewell the performance of many advanced optical systems. The classi-
[1], Goodman [2], and Gaskill [3] results in the irradiance cal Rayleigh–Rice surface scatter theory [4] is a rigorous (vector)
distribution on a plane in the far field (Fraunhofer region) of perturbation technique which requires an explicit smooth-
a diffracting aperture being given by the squared modulus of surface approximation. The classical Beckmann–Kirchhoff
the Fourier transform of the complex amplitude distribution surface scatter theory [5] is valid for rougher surfaces, but
emerging from the diffracting aperture contains a moderate angle limitation and assumes a Gaussian
surface autocovariance (ACV) function. These two surface
1  2
scatter theories are thus complementary, but not all-inclusive.
E (x 2 , y 2 ) = 2 2 F Uo+ (x 1 , y 1 ) ξ = x 2 ,η= y 2 . (1)

λ z λz λz Neither of them, nor the combination of them, are adequate to
Here, Uo+ (x 1 , y 1 ) = Uo− (x 1 , y 1 )t1 (x 1 , y 1 ) is the complex calculate scattered light effects from surfaces of arbitrary rough-
amplitude distribution emerging from the diffracting aperture ness, arbitrary incident, and/or scattered angles or arbitrary
of complex amplitude transmittance t1 (x 1 , y 1 ). Uo− (x 1 , y 1 ) surface ACV functions.
is the complex amplitude distribution incident upon the dif- In April of 1972, Professor Roland Shack presented a series
fracting aperture, and F { } denotes the Fourier transform of four colloquium talks at the Optical Sciences Center at the
operation University of Arizona in which he reformulated scalar diffrac-
Z ∞Z ∞ tion theory in terms of the direction cosines of the propagation
vectors of the angular spectrum of plane waves described by
F Uo+ (x 1 , y 1 ) = Uo+ (x 1 , y 1 )

−∞ −∞ the Fourier integral transform of the diffracting aperture.
The fourth lecture, entitled Radiometry and Lambert’s Law,
× exp [−i2π (x 1 ξ + y 1 η)] dx 1 dy 1 . (2)
described diffuse reflectance and surface scatter phenomena
The spatial frequencies ξ and η are the reciprocal variables in as merely a diffraction phenomenon caused by random phase
Fourier transform space. Also, the Fresnel diffraction integral variations in the system pupil function. In 1974, he elegantly
is given by the Fourier transform of the product of the aperture condensed these four lectures into a single colloquium talk
function with a quadratic phase factor [2,3]. Implicit in both the entitled A Global View of Diffraction [6].
Fresnel and Fraunhofer approximations is a paraxial limitation
that restricts their use to small diffraction angles and small angles
of incidence [2]. It is this paraxial limitation that has prevented 2. RETROSPECTIVE
the conventional Fourier treatment of scalar diffraction theory Professor Roland Shack’s global view of diffraction removed
to be readily applied to many practical non-paraxial optical the inherent paraxial limitation in the conventional linear
engineering applications. systems formulation (or Fourier treatment) of scalar diffraction

1559-128X/20/22G185-19 Journal © 2020 Optical Society of America


G186 Vol. 59, No. 22 / 1 August 2020 / Applied Optics Review

field parameter for several different geometrical configurations


of incident beam and observation space. These aberrations,
which are inherently associated with the diffraction process, are
precisely the effects ignored when making the usual Fresnel and
Fraunhofer approximations. Furthermore, these aberrations
have the same functional form as the familiar aberrations caused
by the refraction or reflection process in imperfect imaging
systems. All third-order aberrations are present when a plane
wave is incident upon the aperture and the observation space
is a plane. Defocus and all orders of spherical aberration are
eliminated by changing the incident beam into a spherical wave
converging to the observation plane. Distortion, field curvature,
and all orders of piston error vanish when the observation space
is a hemisphere centered upon the diffracting aperture as shown
in Fig. 1 [7,8]. This leaves only coma and astigmatism (through
third-order aberrations), as shown in Fig. 2.
Fig. 1. Illustration of the physical situation for the case of a spheri- The Fresnel diffraction integral is valid only if all third-
cal wavefront incident upon a diffraction grating, and converging to a
and higher-order aberrations are negligible, and it therefore
hemispherical observation space (adapted from Ref. [7]).
merely describes a defocused Fraunhofer diffraction pattern.
This treatment thus provides more accuracy than the Fresnel–
Kirchhoff theory and more insight and understanding than the
theory as presented in the Introduction. It thus represents the
Rayleigh–Sommerfeld theory.
most significant advance in the understanding of scalar diffrac-
Case#2, Wide-angle Diffraction Is Shift-invariant in
tion theory since Sommerfeld removed the inconsistencies in
Direction Cosine Space (1979): Another early paper stimu-
Kirchhoff’s boundary conditions in 1896, stimulating a series lated by Shack’s keen physical insight shows that scaling the
of advances in the understanding of scalar diffraction that is still spatial variables by the wavelength of light [9]
going on today. Nine separate cases, applying to different appli-
cations, will be briefly described and illustrated in the following x̂ = x /λ, ŷ = y /λ, ẑ = z/λ, etc. (3)
retrospective. References of peer-reviewed journal articles are
cited for each case to provide access to detailed derivations and causes the reciprocal variables in Fourier transform space to
explanations. be precisely the direction cosines α and β of the propagation
Case#1, Aberrations of Diffracted Wave Fields (1978): vectors of the plane wave components in the angular spectrum
Perhaps the first new insight stimulated by Professor Shack’s of plane waves discussed by Ratcliff, [10] Goodman, [2] and
Global view of diffraction was that of showing that the Gaskill [3]:
Rayleigh–Sommerfeld diffraction integral is equivalent to α = x̂ /ˆr , β = ŷ /ˆr , and γ = ẑ/ˆr . (4)
the Fourier transform integral of a generalized pupil func-
tion which includes a term that represents phase errors in the A very simple and elegant derivation of the Rayleigh–
aperture [7,8]. This term can be interpreted as describing a Sommerfeld diffraction integral can now be obtained by the
conventional wavefront aberration function. The resulting direct application of Fourier transform theory. We merely
aberration coefficients can be calculated and expressed in terms assume that the Fourier transform of the complex amplitude
of the aperture diameter, observation distance, and appropriate distribution emerging from the diffracting aperture in Plane

Fig. 2. Aberrations inherent to the diffraction process are illustrated here as magnified diffraction orders from a 10-line per millimeter Ronchi
Ruling placed in a converging cone of light (r is the radius of the observation hemisphere, d is the aperture diameter in which grating is placed, and β
is the sin of diffracted angle) (adapted from Ref. [7]).
Review Vol. 59, No. 22 / 1 August 2020 / Applied Optics G187

Po in Fig. 3 exists; and likewise, for the complex amplitude Consider now a unit circle in the α−β plane of direction
distribution in the observation plane, P . This is not a severe cosine space as shown in Fig. 4. Inside the unit circle, γ is real
restriction, as Bracewell [1] points out that physical possibility is and the corresponding part of the disturbance will propagate
a sufficient condition for the existence of a Fourier transform. and contribute to the wave field in plane P . However, those
We can thus write the following Fourier transform pairs for components of the direction cosine spectrum which lie outside
planes Po and P : the unit circle have imaginary values of γ and represent that part
Z ∞Z ∞ of the disturbance which experiences a rapid exponential decay.
A o (α, β; 0) = Uo (x̂ , ŷ ; 0) exp[−i2π(α x̂ + β ŷ )]dx̂ d ŷ , This is the part of the disturbance which is commonly referred
−∞ −∞
to as the evanescent wave.
Z ∞Z ∞ The convolution theorem of Fourier transform theory
Uo (x̂ , ŷ ; 0) = A o (α, β; 0) exp[i2π(α x̂ + β ŷ )]dαdβ, requires that a convolution operation exists in the domain of real
−∞ −∞
(5) space that is equivalent to Eq. (8) [1–3]. We thus have the alter-
native method of expressing the complex amplitude expression
Z ∞Z ∞
A(α, β; ẑ) = Uo (x̂ , ŷ ; ẑ) exp[−i2π(α x̂ + β ŷ )]dx̂ d ŷ , in the observation plane by the convolution of the original dis-
−∞ −∞ turbance with a system impulse response. The impulse response,
Z ∞Z ∞
obtained by taking the inverse Fourier transform of the transfer
U (x̂ , ŷ ; ẑ) = A(α, β; ẑ) exp[i2π(α x̂ + β ŷ )]dαdβ. function in Eq. (9), is given by [7,9]
−∞ −∞
(6) ẑ exp(i2π rˆ )
 
1
h(x̂ , ŷ ; ẑ) = F −1 exp(i2π γ ẑ) =

−i .
We will now call A o and A the direction cosine spectrum of 2π rˆ rˆ rˆ
plane waves in planes Po and P , respectively, in analogy to the (11)
angular spectrum of plane waves obtained in the conventional
The above equation is an exact mathematical expression for a
linear systems formulation of scalar diffraction theory. This is
Huygens’ wavelet that is valid right down to the initial disturb-
consistent with a more general treatment which is not restricted
to small angles. ance emerging from the diffracting aperture. However, for rˆ 
1, it reduces to the familiar expression for a spherical wave with
In the scaled coordinate system, ∇ˆ 2 = λ2 ∇ 2 and
a cosine obliquity factor, ẑ/ˆr , and a π/2 phase delay [7–9]. The
k̂ = λ2 k 2 = (2π )2 . Hence, the Helmholtz equation becomes
2
convolution of this impulse response with the initial disturbance
[∇ˆ 2 + (2π )2 = λ2 ∇ 2 ]U (x̂ , ŷ ; ẑ) = 0. (7) emerging from the diffracting aperture yields
Z ∞Z ∞  
Now, by utilizing Eq. (6) and requiring the individual plane 1
U (x̂ , ŷ ; ẑ) = Uo (x̂ , ŷ ; 0)
0 0
−i
wave components to satisfy the Helmholtz equation, we −∞ −∞ 2π `ˆ
find that
ẑ exp(i2π `)
ˆ
A(α, β; γ ) = A o (α, β; 0)e i2π γ ẑ , where γ = 1 − α 2 − β 2 . dx̂ 0 d ŷ 0 ,
p
× (12)
(8) `ˆ `ˆ

A system transfer function is defined as the ratio of the system


output spectrum to the system input spectrum. Eq. (8) can thus
be rewritten as the transfer function of free space, H(α, β; ẑ):

H(α, β; ẑ) ≡ A(α, β; ẑ)/A o (α, β; ẑ) = e i2π γ ẑ . (9)


We have thus far applied no restrictions on γ and two regions
are apparent: that for real values of γ and that for imaginary
values
for(α 2 + β 2 ≤ 1 γ is real
p 
γ = 1−α −β 2 2
for(α 2 + β 2 ≥ 1 γ is imaginary.
(10)

Fig. 4. Unit circle in direction cosine space. The planewave compo-


nents inside the circle will propagate, and the planewave components
Fig. 3. Geometrical configuration of Planes Po and P (adapted outside the circle do not propagate and constitute the evanescent wave
from Ref. [7]). (adapted from Ref. [7]).
G188 Vol. 59, No. 22 / 1 August 2020 / Applied Optics Review

Fig. 5. (a) Geometric relationship between the normally incident beam, diffracting aperture, and observation hemisphere. (b) Geometric configu-
ration when the incident beam strikes the diffracting aperture at an oblique angle (adapted from Ref. [7]).

which is identical to the general Rayleigh–Sommerfeld diffrac- of the observation point from the undiffracted beam in direction
tion formula [7,9]. cosine space.
We have thus derived the general form of the Rayleigh– As a specific example, suppose we have an incident beam of
Sommerfeld diffraction integral from two basic assumptions: light striking a diffraction grating at an angle θo , as illustrated
(i) that the Fourier transform of the optical disturbance exists, in Fig. 5(b). This is the situation usually referred to as conical
and (ii) that in propagating, each of the plane wave components diffraction, where the diffracted orders do not lie in the plane of
obeys the Helmholtz wave equation. incidence, but instead the diffracted orders lie on the surface of
When a spherical wave is incident upon the diffracting aper- a cone centered on the z axis with its apex at the intersection of
ture and the observation space is a hemisphere (as shown in the z axis and the grating. The diffracted orders will strike the
Fig. 5(a), this scaled coordinate system results in the extremely observation hemisphere in a cross section which is not a great
powerful result that the diffracted wave field on the obser- circle but instead a latitude slice, as shown in Fig. 6(a).
vation hemisphere is given directly by the Fourier transform Thus, for large angles of incidence, the various orders appear
of the aperture function multiplied by a spherical Huygen’s to lie in a straight line only if they are projected onto the α−β
wavelet [7–9]: plane in direction cosine space. It is therefore clear that varying
the angle of incidence merely shifts the diffracted wave field in
U (α, β; rˆ ) = γ [exp(i2π rˆ )/(i rˆ )]F {Uo (x̂ , ŷ ; 0)}. (13) the direction cosine space without changing its functional form.
Figure 6(a) is the “globe” from whence Shack’s global view of
Furthermore, this Fourier transform relationship is valid diffraction comes.
not merely over a small region about the optical axis, but over For a reflection grating, Fig. 6(b) shows the location of the
the entire hemisphere [7,9]. Adding a linear phase variation incident beam, and the diffracted orders are displayed in direc-
causes the incident spherical wavefront to strike the diffracting tion cosine space when the grating grooves are parallel to the y
aperture at an arbitrary angle θo , as shown in Fig. 5(b). Applying axis. Note that the diffracted orders are always exactly equally
the shift theorem of Fourier transform theory to Eq. (13), we spaced and lie in a straight line parallel to the α axis in direction
find that the complex amplitude distribution in direction cosine cosine space. Those diffracted orders that lie inside the unit
space is a function of β − βo : circle are real and propagate, whereas the diffracted orders that
U (α, β − βo ; rˆ ) = γ [exp(i2π rˆ )/(i rˆ )]F {Uo0 (x̂ , ŷ ; 0) lie outside the unit circle are evanescent (and thus do not propa-
gate). For reflection gratings, the undiffracted zero order always
× exp(i2π βo ŷ )}, (14) lies diametrically opposite the origin of the α−β coordinate
system from the incident beam. As the incident angle is varied,

where Uo0 (x̂ , ŷ ; 0) = γo Uo (x̂ , ŷ ; 0). the diffraction pattern (size, shape, separation, and orientation
Here, β is the direction cosine of the position vector of of diffracted orders) remains unchanged but merely shifts its
the observation point in Fig. 5(b), and βo is the direction position, maintaining the just-described relationship between
cosine of the position vector of the undiffracted beam. Note the undiffracted order and the incident beam [11].
that the direction cosines are obtained by the mere projection of Figure 7 illustrates the direction cosine diagram for a beam
the respective points on the hemisphere back onto the plane of obliquely incident (αi = −0.3, βi = −0.4) upon a reflection
the aperture and normalizing to a unit radius. grating for different orientations of the grating. Note that in all
The complex amplitude at an arbitrary point on the observa- cases the zero order is diametrically opposite the origin from the
tion hemisphere can now be said to be a function of the distance incident beam and the diffracted orders remain equally spaced
Review Vol. 59, No. 22 / 1 August 2020 / Applied Optics G189

Fig. 6. (a) Illustration of the position of the diffracted orders in real space and in direction cosine space, showing that the diffracted orders strike
the observation hemisphere in a latitude slice, not a great circle (adapted from Ref. [7]). (b) Relative position of the diffracted orders and the incident
beam (for a reflection grating) in direction cosine space. Diffracted orders outside the unit circle are evanescent (adapted from Ref. [11]).

Fig. 8. Schematic representation of reflectance from a rough surface


(adapted from Ref. [7]).

Fig. 7. Direction cosine diagrams for four different orientations of It was customary to present angular scattering data as scat-
a diffraction grating with period d = 3λ illuminated with an obliquely
tered intensity (radiant power per unit solid angle) versus
incident beam (αi = −0.3, βi = −0.4) (adapted from Ref. [11]).
scattering angle. The resulting scattered light profiles change
shape drastically with angle of incidence, becoming quite
skewed and asymmetrical at large angles of incidence, as shown
in a straight line, but this line is rotated about the zero order in Fig. 9(a). However, if we divide each data point by the cosine
such that it is always perpendicular to the grating grooves. This of the scattering angle (thus converting scattered intensity to
simple shift-invariant behavior of “conical” diffraction from scattered radiance), then re-plot in direction cosine space, the
linear gratings, when expressed in direction cosine space, pro- five curves with the incident angle varying from 0 to 60◦ coin-
vides understanding and insight not provided by most textbook cide almost perfectly. This shift-invariant behavior, illustrated
treatments of this topic [11]. in Fig. 9(b), drastically reduces the quantity of data required
Case#3, The Original Harvey–Shack (OHS) Surface to completely characterize the scattering properties of the
Scatter Theory (1976): When light is reflected from an surface [7,12].
imperfect optical surface, the reflected radiation consists of Scattered light data is usually collected on an observa-
a specularly reflected component and a diffusely reflected com- tion hemisphere with some type of goniometric instrument.
ponent, as illustrated in Fig. 8. Professor Shack described surface Figure 10 shows an incident beam striking a scattering surface at
scatter as merely a diffraction phenomenon caused by random some angle of incidence (θi = −θo ), a specularly reflected beam
phase variations in the system pupil function [6]. This suggested striking the observation hemisphere at the position indicated,
that surface scatter should also exhibit shift-invariant behavior and the scattered intensity distribution being sampled at an
in direction cosine space. arbitrary angle, θ.
G190 Vol. 59, No. 22 / 1 August 2020 / Applied Optics Review

Fig. 9. (a) Experimentally measured scattered intensity plotted


versus scattered angle. (b) Scattered radiance plotted versus β − βo in
direction cosine space (adapted from Ref. [7]).

Fig. 11. Geometrical configuration of the two principal planes


in which the scattered light distribution was sampled (adapted from
Ref. [7]).

In 1976, based upon the empirical experimental observation


that scattered radiance is shift-invariant in direction cosine
space, Harvey and Shack reported upon a linear systems for-
mulation of surface scatter phenomena in which the scattering
behavior is characterized by a surface transfer function expressed
Fig. 10. Illustration of the geometrical configuration for scatter
measurements (adapted from Ref. [7]).
mathematically as [7]

Hs (x̂ , ŷ ) = exp{−(4π σ̂s )2 [1 − C s (x̂ , ŷ )/σs2 ]}, (15)


The scattered intensity distribution illustrated on the obser- where C s (x̂ , ŷ ) ≡ C 12 is the two-dimensional autocovariance
vation hemisphere becomes very skewed and asymmetrical as function of the surface, and σs is the rms surface roughness.
the angle of incidence is increased; however, when this data Insight into the scattering process was inferred by consider-
is divided by the cosine of the scattering angle and projected ing the nature of the surface transfer function and its Fourier
onto the plane of the scattering surface, as indicated in Fig. 10, transform, the angle spread function (ASF) [7]. Note that this
the scattered radiance distribution does not change shape with ASF is scattered radiance, not irradiance or intensity. This was
incident angle. It merely shifts in direction cosine space while consistent with the fact that the bidirectional reflectance dis-
maintaining the symmetrical shape indicated in the figure [7]. tribution function (BRDF) was defined as reflected radiance
Scatter measurements outside of the plane of incidence were divided by incident irradiance [13]. An expression very similar
confined to a plane passing through the specular beam and to Eq. (15) had been derived by Hufnagel for the atmospheric
orthogonal to both the plane of incidence and the plane of the structure function in the presence of random turbulence, [14]
scattering sample; i.e., a latitude slice (not a great circle) on the and by Chandley and Welford for a reformulation of Beckmann
observation hemisphere, as illustrated in Fig. 11 [7]. scattering for rough surfaces [15].
Extensive experimental surface scatter measurements were Case#4, The Modified Harvey-Shack (MHS) Surface
made from optical surfaces polished by various techniques on Scatter Theory (1988, 2006): The transfer function charac-
different optical materials. Note in Fig. 12(a) that experimen- terization of scattering surfaces was modified in the late 1980s
tal data taken in the plane of incidence (α = 0) for different to include grazing incidence effects in x-ray telescopes, and
incident angles and plotted on a log–log scale as a function “mid” spatial frequency surface errors that span the gap between
of distance from the specular beam in direction cosine space “figure” and “finish” errors [16]. This allowed an understanding
(β − βo ) exhibits not only a shift-invariant behavior, with of image degradation due to scattering effects from residual
respect to incident angle, but an inverse power, or fractal, optical fabrication errors on NASA’s Chandra Observatory and
behavior as well [7]. NOAA’s Solar X-ray Imager (SXI) [17,18].
The behavior exhibited in Fig. 12(b) is strong experimen- Figure 13 shows a ray incident upon a scattering surface at
tal evidence that (for isotropically random rough surfaces) an arbitrary angle of incidence θi . The optical path difference
the angular size of the scattering function remains fixed, and (OPD) experienced by a ray reflected from the surface in the
rotationally symmetric, in direction cosine space [7]. specular (θo = −θi ) direction is given by
Review Vol. 59, No. 22 / 1 August 2020 / Applied Optics G191

Fig. 12. (a) Illustration of shift-invariance, relative to incident angle, when displayed in direction cosine space. (b) illustrates that the forward,
backward, left, and right surface scatter profiles (for a fixed 45◦ incident angle) are also superposed (adapted from Ref. [7]).

A(γi ) = exp[−(4π γi σ̂s )2 ], B(γi ) = 1 − exp[−(4π γi σ̂s )2 ],


(20)
θi θs
Surface Height

and
exp[(4π γi )2 C s (x̂ , ŷ )] − 1
G(x̂ , ŷ ; γi ) = . (21)
exp[(4π γi σ̂s )2 ] − 1
h
Figure 14 illustrates the very substantial difference in both
ŷ the peak value and the shape of the relative intensity profile (in
the plane of incidence) predicted by the original Harvey–Shack
and the modified Harvey–Shack surface scatter theories as the
Fig. 13. Illustration of the OPD for a specularly reflected ray incident angle increases. A Gaussian surface autocovariance
(adapted from Ref. [37]). function with rms roughness of 2.27 µm; autocovariance width,
`c , of 20.9 µm; and a wavelength of 10.6 µm was assumed for
these calculations.
OPD = (γi + γo )h(x̂ , ŷ ) = 2γi h(x̂ , ŷ ), (16) The difference between the two theories is striking, and the
where γi = cos θi and γo = cos θo . The corresponding phase O’Donnell–Mendez experimental data [21] agrees well with
variation is given by the MHS predictions for the large incident angle of 70◦ . Note
that the total integrated scatter (TIS) calculated from the MHS
2π theory decreases with increasing incident angle, dropping dra-
φ(x̂ , ŷ ) = OPD = 4π γi ĥ(x̂ , ŷ ). (17)
λ matically for 50◦ < θi < 70◦ . A more detailed description of
Provided that the scattering angles are small relative to the the MHS surface scatter theory and the calculations resulting in
angle of specular reflection, the phase function of Eq. (17) Fig. 14 is available in Section 3.4 of Ref. [22].
describes the phase variations introduced upon reflection from a Clearly the surface scatter process is no longer shift-invariant
scattering surface for a wavefront incident at an arbitrary angle. with respect to incident angle, as Eq. (18) can be interpreted as a
Krywonos has shown that the surface transfer function for an one-parameter family of surface transfer functions, i.e., a differ-
arbitrary incident angle (assuming small angle scattering) can be ent surface transfer function is required for each incident angle.
expressed as [19] This is analogous to imaging in the presence of field–dependent
aberrations, where a different optical transfer function (OTF) is
Hs (x̂ , ŷ ; γi ) = exp{−(4π γi σ̂rel )2 [1 − C s (x̂ , ŷ )/σ̂s2 ]}, (18) required for each field angle.
Case#5, Diffracted Radiance, Fundamental Quantity in
where γi = cos θi and σ̂rel is the relevant band-limited rms sur- Scalar Diffraction Theory (1999): Before proceeding to show
face roughness (see Ref. [20]). that radiance is the fundamental quantity that is shift-invariant
This can again be written in the form with respect to incident angle in a non-paraxial scalar diffraction
Hs (x̂ , ŷ ; γi ) = A(γi ) + B(γi )G x̂ , ŷ ; γi ,

(19) theory, we will briefly review the definitions of a few radiometric
quantities. In the past, scientists have generally used the word
where intensity to mean the flow of energy per unit area per unit time.
G192 Vol. 59, No. 22 / 1 August 2020 / Applied Optics Review

Fig. 14. Comparison of scattered intensity predictions from the OHS and the MHS surface scatter theories for different incident angles.
Experimental data is also displayed for θi = 70◦ (this figure previously published as Fig. 3–23 in Ref. [22]).

However, by international, if not universal agreement, that term Source Collector


13
is slowly being replaced by the word irradiance, [23] θs θc
∂P ∂ As r ∂Ac
Irradiance ≡ E − (watts/area). (22)
∂A c Fig. 15. Geometrical configuration used to demonstrate the fun-
damental theorem of radiometry from which the quantity radiance is
Irradiance is defined as the radiant power density incident obtained.
upon a collecting surface (thus the subscript c ). Radiant inten-
sity, on the other hand, carries the units of watts per steradian
[24]. Intensity is properly used when describing the radiation Rearranging the just-given defining equation for radiance
emanating from a point source (or a source that has negligible and noting that ∂ωc = ∂A c cos θc /r 2 and ∂ωs = ∂A s cos θs /r 2 ,
area compared to the square of the viewing distance) [25,26] we obtain the double differential
∂P ∂ 2 P = L(θs , φs , x , y )∂A s cos θs ∂ωc
Radiant Intensity ≡ I = (watts/steradian). (23)
∂ωc
= L(θs , φs , x , y )∂A s cos θs ∂A c cos θc /r 2
And radiance, the radiometric analog to the more familiar
photometric term brightness, is defined as radiant power per = L(θs , φs , x , y )∂ωs ∂A c cos θc . (25)
unit solid angle per unit projected source area. The quantity
radiance is used to characterize an extended source, that is, one Here, r is the distance between the source and the collector as
that has appreciable area compared to the square of the viewing illustrated in Fig. 15. This equation can be considered to be the
distance [25,26]. In differential form, fundamental theorem of radiometry, as it describes the radiant
∂2 P power transfer between an elemental source and an elemental
Radiance ≡ L = (watts/steradian projected area). collector.
∂ωc ∂A s cos θs
(24) Most textbook treatments of the propagation of diffracted
wave fields include a paraxial (small angle) assumption.
The radiance of a source is, in general, a function of position Although this assumption may be reasonable for imaging
on the source and a function of the two angular variables θs and systems with clear apertures, it is clearly inappropriate when
φs in conventional spherical coordinates. dealing with diffraction gratings (inherently wide-angle devices)
Review Vol. 59, No. 22 / 1 August 2020 / Applied Optics G193

or with scattering surfaces, both of which are frequently used


with large oblique incident angles.
Recall that in Case#2 above, we showed that wide-angle
scalar diffraction phenomena are shift-invariant with respect to
changes in incident angle when formulated in direction cosine
space. Furthermore, when a spherical wave is incident upon
the diffracting aperture and the observation space is a hemi-
sphere, this Fourier transform relationship is valid not merely
over a small region about the optical axis, but over the entire
hemisphere.
In References [27] and [28], it has been shown that the
squared modulus of the Fourier transform of the complex
amplitude distribution emerging from the diffracting aperture Fig. 16. Illustration of a moderately broad Gaussian diffracted
yields diffracted radiance, not irradiance or intensity: radiance distribution. (a) Normal incidence, (b) 64◦ incident angle
(adapted from Ref. [27]).
λ2  2
L(α, β) = F Uo (x̂ , ŷ ; 0) . (26)
As
radiance distribution of unit height and half-width 1β = 0.4, as
Furthermore, this equation, unlike the more familiar Eq. (1),
is not restricted to small diffraction angles. From the shift theo- illustrated in Fig. 16(a). An incident angle of 64◦ (βi = 0.9) will
rem of Fourier Transform theory, [3] it is clear that changes cause 26.5% of the shifted radiance function, L(α, β − βo ),
in the angle of incidence of the radiation illuminating the dif- to fall outside of the unit circle in direction cosine space. The
fracting aperture will merely result in a shift of the radiance evanescent (imaginary) waves created produce no scattered
function in direction cosine space and an attenuation by the light. This requires a renormalization factor of K = 1.36, as
factor γo = cos θo shown in Fig. 16(b) [27,28].
In either case, evanescent waves are produced, and the expres-
λ2 sion for radiance must be renormalized. This is not done in
L(α, β − βo ) = γo |F {Uo (x̂ , ŷ ; 0) exp(i2πβo ŷ )}|2 . a heuristic manner merely to conserve energy, but is a direct
As
(27) result of Parseval’s theorem from Fourier transform theory.
Since the functional form does not change, the diffracted The well-known Wood’s anomalies that occur in diffraction
radiance is shift-invariant in direction cosine space, and the grating efficiency measurements are entirely consistent with
simple Fourier techniques that have proven to be so useful in this predicted renormalization in the presence of evanescent
paraxial applications can be used for non-paraxial applications waves.
as well. Case#6, Predicting Rayleigh Anomalies with Non-
In Ref. [27], after an egregious error was corrected in an errata paraxial Scalar Diffraction Theory (2006): The application
(Ref. [28]), sound radiometric principles were incorporated of the preceding non-paraxial scalar diffraction theory to sinus-
into this linear systems formulation of non-paraxial scalar oidal phase gratings is shown in Fig. 17 and compared to the
diffraction theory which resulted in the following expression classical Beckmann–Kirchhoff theory, the standard parax-
replacing Eq. (1): ial scalar diffraction theory and rigorous vector theory [29]

2 2
L 0 (α, β − βo ) = K γo λAs F Uo (x̂ , ŷ ; 0) exp(i2π βo ŷ ) for α 2 + β 2 ≤ 1,

(28)
L 0 (α, β − βo ) = 0 for α 2 + β 2 > 1,

where K is a renormalization factor required for extreme cases


where the diffracted radiance distribution function extends
beyond the unit circle in direction cosine space. This can occur The nonparaxial scalar diffraction theory provides remark-
due to high spatial frequency content in the diffracting aperture ably good agreement with rigorous integral electromagnetic
resulting in very large diffracted angles, or due to large incident theory, not merely in the paraxial regime and the smooth
angles that shift a radiance distribution function of modest surface (shallow grating) regime, but over the entire range of
width such that a portion of it extends beyond the unit circle in λ/d . Figure 17 not only shows good agreement between our
direction cosine space, as shown in Fig. 16:
non–paraxial scalar theory and rigorous calculations, it also
R∞ R∞
L(α, β − βo )dαdβ demonstrates that our non-paraxial scalar diffraction theory
α=−∞ β=−∞
K= √ ≡ Normalization Factor. indeed predicts the Rayleigh anomalies [30,31] that occur
R1 1−α 2
L(α, β − βo )dαdβ
R
α=−1
√ when a propagating order goes evanescent. Note the abrupt
β=− 1−α 2
(29) increase in diffraction efficiency at λ/d = 0.67. This is precisely
the value of λ/d at which the −1 and +2 diffracted orders go
Hence, it is diffracted radiance (not intensity or irradiance) evanescent.
that is shift-invariant in direction cosine space. Figure 16 The development of this non-paraxial scalar diffraction the-
illustrates an application with an on-axis diffracted Gaussian ory greatly expands the range of parameters over which accurate
G194 Vol. 59, No. 22 / 1 August 2020 / Applied Optics Review

Fig. 17. Diffraction grating efficiency of the first order of a perfectly conducting sinusoidal phase grating in the Littrow condition (for two differ-
ent values of h/d ). Courtesy of OSA, first appeared in Ref. [29].

predictions can be made with simple Fourier techniques, and That the Fraunhofer criterion is usually given as an axial dis-
is a crucial step in the development of the GHS surface scatter tance much greater than some amount relative to the maximum
theory. dimension of the aperture is rather vague and unsatisfying to
Case#7, Tolerance on Defocus Precisely Locates the Far the optical engineer wanting to know exactly how far he must
Field (2002): [32] The Fresnel and Fraunhofer diffraction be from a given diffracting aperture in order to safely utilize the
integrals are obtained by making explicit approximations to Fraunhofer diffraction integral.
the more general Rayleigh–Sommerfeld diffraction integral. It was mentioned in Case#1 that the Rayleigh–Sommerfeld
These approximations restrict the region of space over which the diffraction integral can be written (without making any explicit
respective integrals are valid, as shown in Fig. 18 [3,4] approximations) as a Fourier transform integral of a generalized
Note that the Fraunhofer region is contained in the Fresnel pupil function containing phase variations which are identi-
region, and both the Fraunhofer and Fresnel regions are con- fied with conventional wavefront aberrations [7,8]. Near-field
tained in the Rayleigh–Sommerfeld region. The far field is diffraction patterns are thus aberrated Fraunhofer diffraction
widely understood to be synonymous with the Fraunhofer patterns. Fraunhofer diffraction is therefore aberration-free, as it
region; however, there is less agreement in the literature upon is precisely the reference from which our aberrations are defined,
the definition of the near field. We define the near field to be that and Fresnel diffraction can be described as merely a defocused
region of space that does not satisfy even the Fresnel criterion. Fraunhofer diffraction pattern [8,32].
Hence, the near field and far field are mutually exclusive, but not Recall that the wavefront aberration function is expressed
all-inclusive. as a power series expansion of the appropriate field and pupil
parameters [8,32–34]:
o
Ŵ = Ŵ000 (Piston Error) Zero-order
)
+ Ŵ200 ρ + Ŵ020 â + Ŵ111 ρ â cos(φ − φ)
2 2 0
2nd-order
Piston Defocus Lateral Magnification Error

+Ŵ400 ρ 4 + + Ŵ131 ρ â 3 cos(φ 0 − φ)



Ŵ040 â 4 

Piston Spherical Aberration Coma
4th-order
+Ŵ222 ρ 2 â 2 cos2 (φ 0 − φ) + Ŵ220 ρ 2 â 2 + Ŵ311 ρ 3 â cos(φ 0 − φ) 

Astigmatism Field Curvature Distortion

+ Higher-order Terms, (31)


where ρ is the normalized field position of the observa-
tion point, and a is the normalized pupil height. The
The Fraunhofer criterion specifies the minimum distance of “hats” in these equations merely indicate that those quan-
the observation plane from a diffracting aperture to assure the tities are also normalized by the wavelength of light,
validity of the Fraunhofer diffraction integral [3]
i.e., ẑ = z/λ, â = a /λ, andŴ020 = W020 /λ.
z  (k/2)(x 12 + y 12 )max . (30) That a Fresnel diffraction pattern is merely a defocused
Fraunhofer diffraction pattern is evidenced by the fact that
Review Vol. 59, No. 22 / 1 August 2020 / Applied Optics G195

Fig. 18. Illustration of the axial regions of validity for the Rayleigh–Sommerfeld, Fresnel and Fraunhofer diffraction integrals. Courtesy of OSA;
this figure previously published as Fig. 1 in Ref. [32].

π 2
Table 1. Fraunhofer Criterion for Different Tolerances φ= d (radians). (34)
upon Defocus 4λz
W020 z This measure of quality is used in many areas of physical
optics and antenna measurement. Finally, we would comment
λ/2 0.80 km
that the preceding analysis approach is applicable to far-field
λ/4 1.60 km
measurements made using Fourier transform lenses to bring the
λ/8 3.20 km
λ/10 4.00 km
measurement plane to the focal plane of the transform lens. In
λ/20 8.00 km this case, lens aberrations and positioning errors relative to the
focal plane also affect the far-field quality.
The new expression for the Fraunhofer criterion given by
the two diffraction integrals differ only by a quadratic phase Eq. (33) should be particularly useful to optical designers and
factor [3,4]. For a plane wave incident upon a circular aperture engineers that routinely deal with such tolerances.
of diameter d , the aberration coefficient for defocus is given Case#8, The Modified Beckman–Kirchhoff Surface
by [32] Scatter Model (2007): [35] A detailed experimental inves-
tigation of light scattering from well-characterized random
!2 surfaces was reported by O’Donnell and Mendez in 1987
ẑ d̂
Ŵ020 = . (32) [21]. Several puzzling effects were observed when comparing
2 2ẑ experimental scattered intensity measurements with classical
Beckmann–Kirchhoff scattering theory. The surfaces were
Solving for z, we obtain made by exposing photoresist to a laser speckle pattern and
then processing and coating with gold. The surface autoco-
d2 variance (ACV) function was almost a perfect Gaussian. The
z= . (33)
8W020 measured rms surface roughness was σs = 2.27 µm, and the
measured ACV length (e −1 half width of the ACV function) was
Equation (33) serves as a new expression for the Fraunhofer
`C = 20.9 µm. For the small slopes represented by these param-
criterion, based upon a specified tolerance for the amount of
eters, multiple-scattering effects are negligible and conventional
defocus allowed for a given application. For a circular aperture high-angular-resolution scattering data were measured for
4.0 cm in diameter and for a wavelength of 0.5 µm, we obtain two wavelengths, λ = 0.6328 µm and λ = 10.6 µm, and two
from Eq. (30) that an observation distance much greater than angles of incidence, θi = 20◦ and θi = 70◦ . At these wavelengths
2.5 km is required for the Fraunhofer diffraction integral to be and incident angles, this surface is far too rough to satisfy the
valid. We now see from Table 1 that, if we are willing to toler- Rayleigh–Rice smooth-surface criterion.
ate one-half wave of defocus, an observation distance of only For a wavelength of 10.6 µm and modest angles of incidence,
800 m is required. On the other hand, if we tighten the allowable the Beckmann–Kirchhoff solution agreed quite well with the
tolerance upon defocus to λ/20, 8 km is required! experimental data; however, a persistent tendency for the data to
The Fraunhofer criterion, as usually stated, says that the be narrower than the theory was observed as shown in Fig. 19(a).
amount of defocus allowed must be much less than λ/2π . This was non-intuitive, as the presence of experimental error
Clearly there are many practical applications where this toler- sources (jitter, turbulence effects, etc.) would tend to make the
ance can be relaxed. Also, by multiplying Eq. (32) by 2π , we experimental curve broader than the theoretical curve. There is
obtain the following equation for the maximum phase error no discernable specular beam for an incident angle of 20◦ , as the
across the aperture: total integrated scatter (TIS) is almost unity.
G196 Vol. 59, No. 22 / 1 August 2020 / Applied Optics Review

Fig. 19. Non-intuitive surface scatter effects. Solid lines represent classical Beckmann–Kirchhoff scatter theory, circles indicate experimental mea-
surements (reprinted with permission from Ref. [21]).

For a wavelength of λ = 10.6 µm and an incident angle of For example, Fig. 21(a) shows a comparison between scat-
θi = 70◦ , there is a specular beam containing more than 40% of tered radiance and scattered intensity as predicted by the MHS
the reflected radiant power; however, the data points influenced surface scatter theory for a beam normally incident upon the
by it have been omitted from the experimental data shown in O’Donnell and Mendez scattering surfaces. This is strikingly
Fig. 19(b). Note that the peak of the scattering function does not similar to Fig. 19(a).
lie in the specular direction; instead, it lies approximately 10◦ Likewise Fig. 21(b), is very similar to Fig. 19(b). Note that
inside the specular beam. Likewise, the Beckmann–Kirchhoff O’Donnell and Mendez have labeled both of the curves in
theory no longer exhibits symmetry about the specular direction Fig. 19 as scattered intensity versus scattering angle. Since the
and shows a similar shift of the peak of the diffuse component of solid angle subtended by the collecting aperture of their scat-
the scattered light distribution. Furthermore, the Beckmann– terometer is constant as they scan the observation hemisphere,
Kirchhoff profile exhibits a discontinuity at −90◦ , which is the voltage signal received from their instrument is indeed
a nonphysical effect if the profile represents radiant intensity proportional to scattered intensity (W/sr). The discontinu-
(proper definitions, units, and terminology of radiometric ity in the solid curve (representing the Beckmann–Kirchhoff
theory) at −90◦ in Figs. 19(b) and 19(c) is unphysical if the
quantities presented in Case#5). The authors of Ref. [21] offer
quantity being plotted is radiant intensity (W/sr); however, such
no explanation for these nonintuitive effects.
a discontinuity would be allowed if plotting scattered radiance.
Finally, Fig. 19(c) illustrates the data for a wavelength of
Their problem is in misinterpreting the radiometric quantity
λ = 0.6328 µm and an angle of incidence of θi = 70◦ . The predicted by the Beckmann–Kirchhoff theory.
experimental data are highly asymmetrical about the specular It should be pointed out that Beckmann and Spizzichino [5]
direction (note that there is no specular beam at this wavelength) do not use conventional radiometric terminology. O’Donnell
and drop smoothly to zero at −90◦ . The Beckmann theory and Mendez used Eq. (35) on p. 86 of Ref. [5], a closed-
predicts a symmetrical intensity distribution about the specular form solution valid only for Gaussian surface autocovariance
direction, though it is unphysical in that it exhibits a large dis- functions, which is written as
continuity at −90◦ . The authors of Ref. [21] suggest that the

reason the Beckmann theory fails at this angle of incidence may π `2c F 2 exp(−g ) X g m
D {ρ} = exp[−(vxy `c /4m)],
2 2
be primarily shadowing and multiple scattering effects, and state As m!m
m=1
that they are unaware of any available theory that compares to (35)
their measured data. where A s is the illuminated surface area, `c is the surface ACV
When applied to the O’Donnell and Mendez scattering length, and F is a geometrical factor defined as the follow-
surface, the MHS surface scatter theory discussed in Case#4 ing function of incident and scattered angles (θ and φ are the
predicts a shift-invariant ASF (scattered radiance distribution) scattered angles in standard spherical coordinates):
in direction cosine space, as illustrated in Fig. 20(a) for nor-
mal incidence and λ = 0.6328 µm. At an incident angle of 1 + (cos θi cos θ − sin θi sin θ cos φ)
F= . (36)
70o and λ = 10.6 µm, a significant fraction of the scattered cos θi (cos θi + cos θ )
radiance distribution function falls outside the unit circle in g is a measure of the phase variation introduced by an rms sur-
direction cosine space. When truncated and renormalized in face roughness of σs :
accordance with Eq. (28), the ASF is as illustrated in Fig. 20(b).
And after applying Lambert’s cosine law to convert the scattered g = [(2π σs /λ)(cos θi + cos θ )]2 , (37)
radiance to scattered intensity, we have the scattered intensity
and
distribution illustrated in Fig. 20(c). These results qualitatively q
and intuitively explain all three of the non-intuitive surface vxy = k sin2 θi − 2 sin θi sin θ cos φ + sin2 θ . (38)
scatter effects illustrated in Fig. 19, and lead us to believe that
O’Donnell and Mendez might be inappropriately comparing The quantity D{ρ} in Eq. (35) is the time average of the
different radiometric quantities. squared modulus of the electric field vector, and is called the
Review Vol. 59, No. 22 / 1 August 2020 / Applied Optics G197

Fig. 20. (a) Predicted ASF (radiance) for normal incidence and λ = 0.6328 µm; (b) truncated and renormalized ASF for θi = 70◦ and
λ = 10.6 µm; (c) scattered intensity for θi = 70◦ and 10.6 µm (adapted from Ref. [27]).

3.5
λ = 10.6 μm
1
3 θo = 70 deg

Scattered Intensity λ = 10.6 μm


θo = 70 deg
0.8 2.5

2
0.6 Radiance Scattered Radiance
Intensity
1.5
0.4
1

Specular
0.2
0.5 Direction

0 0
-1 -0. 8 -0.6 -0.4 -0. 2 0 0.2 0. 4 0.6 0.8 1 -1 -0.9 -0.8 -0.7 -0.6 -0.5
Direction Cosine of Scattering Angle (β) Direction Cosine of Scattering Angle ( β )

(a) (b) (c)

Fig. 21. (a) Comparison of MHS prediction of scattered radiance and intensity for the O’Donnell–Mendez surface with θi = 0 and λ = 10.6 µm;
(b) predicted radiance and intensity for θi = 70◦ and λ = 10.6 µm; (c) O’Donnell–Mendez measurements compared to MHS prediction of scattered
intensity (adapted from Ref. [27]).

Z
mean scattered power by Beckmann and Spizzichino [5]. Most I (θ, φ) = L(θ, φ) cos θ ∂A s = L(θ, φ)A s cos θ (40)
physicists would interpret this quantity to be proportional to As
radiant power density on the collecting surface (and perhaps
or
inappropriately call it intensity). The geometrical factor, F ,
is perhaps an attempt to model the asymmetries that occur in ∞
X gm
measured scattered intensity profiles. I (θ, φ) = K π `2c cos θ exp(−g ) exp[−(vxy `c /4m)].
2 2
m!m
However, the new insight provided by Eqs. (28) and (29), in m=1
which diffracted radiance is presented as the fundamental quan- (41)
tity predicted by scalar diffraction theory, and the success of the Recall that for a wavelength of 10.6 µm and an incident
non-paraxial scalar diffraction theory (Case#5) in explaining the angle of 20◦ , almost all of the reflected light is scattered
non-intuitive behavior just described [27–29] leads us to try an (TIS = 0.998), and there is virtually no specular beam. In
empirical modification of the Beckmann–Kirchhoff scattering Fig. 22, scattered intensity predictions from this new modified
theory. Beckmann–Kirchhoff model expressed by Eq. (41) are superim-
Three explicit empirical modifications include: (i) eliminat- posed upon the previous curves of Fig. 19. Since we only have
ing the geometrical factor, F 2 , from Eq. (35); (ii) introducing one-dimensional experimental scattering profile data, we merely
the re-normalization factor, K , from Eq. (29) (this accounts for normalize the peak of the scattering function to unity, and label
the re-distribution of radiant energy from the evanescent waves the curve relative intensity. For the modest 20◦ incident angle of
into the propagating waves); and (iii) equating the right side Fig. 22(a), the departure of the predictions from the classical and
of Eq. (35) to scattered radiance. We thus obtain the following modified Beckmann–Kirchhoff surface scatter models are not
modified Beckmann–Kirchhoff scatter model: severe; however, the modified model does more closely match
∞ the experimental data.
π `2c exp(−g ) X g m In Fig. 22(b) we have a 70◦ incident angle and a wave-
L(θ, φ) = K exp[−(vxy `c /4m)].
2 2
As m=1
m!m length of 10.6 µm. The modified Beckmann–Kirchhoff
(39) model once again more closely matches the experimental
To obtain the scattered intensity distribution, we merely multi- measurements, and it does not exhibit the non-physical discon-
ply the radiance by cos θ and integrate over the illuminated area tinuity at a scattering angle of −90◦ exhibited by the classical
of the scattering surface [27–29] Beckmann–Kirchhoff theory.
G198 Vol. 59, No. 22 / 1 August 2020 / Applied Optics Review

Fig. 22. Predictions from the empirically Modified Beckmann–Kirchhoff model are compared to the previously non-intuitive surface scatter
effects illustrated in Fig. 17 (adapted from Ref. [35]).

For the much shorter wavelength of λ = 0.6328 µm, the Furthermore, as either the incident and/or scattered angles
O’Donnell/Mendez surface must be categorized as very become large, the modified Beckmann–Kirchhoff model con-
rough (σs /λ = 3.59), and clearly, for such a rough sur- tinues to agree with the Rayleigh–Rice theory, and the classical
face, all the incident light is scattered (no specular beam). Beckmann–Kirchhoff theory begins to break down and starts to
Equations (39) and (41) do not converge; hence, we use the exhibit markedly different behavior, including the non-physical
classical Beckmann–Kirchhoff Eq. (48) on page 87 of Ref. [5]: discontinuities at ±90◦ as shown in Figs. 23(b) and 23(c).
vxy `c In summary, the linear systems formulation of nonparaxial
" 2 2 #
π`2c F 2
D {ρ} = exp , where vz = −k(cos θi + cos θ ). scalar diffraction phenomena discussed in Case#5, indicating
A s vz2 σs2 4vz2 σs2
that diffracted radiance is a fundamental quantity predicted by
(42)
scalar diffraction theory, led to a re-examination and empirical
As before, we make an empirical modification by eliminating
modification of the classical Beckmann–Kirchhoff scattering
the geometrical factor, F 2 , adding the renormalization factor,
theory. This modified Beckmann–Kirchhoff surface scat-
K , and equating the right side of Eq. (42) to scattered radiance,
thus obtaining the following modified Beckmann–Kirchhoff ter model agrees extremely well with experimental data for
scatter model for very rough surfaces: rough surfaces, even for large incident and scattered angles.
It also exhibits excellent agreement with the well-established
vxy `c
" 2 2
#
π `2c Rayleigh–Rice vector perturbation surface scatter theory (for
L(θ, ϕ) = K exp − 2 2 . (43) TE polarization) within its domain of applicability for smooth
A s vz2 σs2 4vz σs
surfaces (comparisons performed for surfaces with Gaussian
Again converting radiance to intensity by multiplying by cos θ , autocovariance functions).
we obtain Case#9, The Generalized Harvey–Shack (GHS) Surface
Scatter Theory (2011): [19,36,37] The classical Rayleigh–Rice
vxy `c
" 2 2
#
π `2c cos θ (1951) vector perturbation theory agrees well with moderately
I (θ, φ) = K exp − 2 2 . (44)
vz2 σs2 4vz σs wide-angle (<50◦ ) experimental scatter measurements from
“smooth” surfaces. However, not all applications of interest
In Fig. 22(c), scattered intensity predictions from the modi- satisfy the smooth surface approximation. The Beckmann–
fied Beckmann–Kirchhoff model expressed by Eq. (44) are Kirchoff (1963) scattering theory is valid for rougher surfaces,
superimposed upon the previous curves in Fig. 19(c). Again, but contains a moderate-angle assumption that limits its ability
we normalize the peak of the modified Beckmann–Kirchhoff
to accurately handle wide-angle scattering and large angles of
scattering profile to unity. For this short wavelength and a 70◦
incidence. In addition, an analytical solution only exists for
incident angle, the departure between the predictions of the
surfaces with Gaussian statistics. These two theories are thus
classical and the modified theory is quite dramatic. The clas-
sical Beckmann–Kirchhoff theory indicates a very significant complementary, but not all-inclusive. Neither of them, nor
(non-physical) discontinuity at a scattering angle of −90◦ , the combination of them, can be used to predict surface scatter
and is clearly not capable of making accurate surface scatter effects for moderately rough surfaces with large incident and/or
predictions for very rough surfaces at these large incident angles; scattering angles.
however, the empirically modified Beckmann–Kirchhoff model It was the contractual requirement to perform the image
agrees extremely well with the experimental data. degradation analysis due to surface scatter effects on NOAA’s
We now compare the predictions of the modified Solar UltraViolet Imager (SUVI), a normal-incidence EUV
Beckmann–Kirchhoff surface scatter model with the classi- imaging instrument developed for NOAA’s Geostationary
cal Beckmann–Kirchhoff theory and the classical Rayleigh–Rice Orbiting Environmental Satellite (GOES) that provided the
theory for optically smooth surfaces that exhibit increasingly impetus to further generalize the MHS surface scatter theory.
large incident and scattering angles. As expected, for small Even state-of-the-art EUV mirrors did not satisfy the Rayleigh–
incident and scattering angles, all three of the scattering the- Rice smooth-surface requirement at the SUVI wavelengths, and
ories yield virtually identical results as shown in Fig. 23(a). the Beckmann–Kirchhoff surface scatter theory only provided
Review Vol. 59, No. 22 / 1 August 2020 / Applied Optics G199

Fig. 23. Predictions from the empirically Modified Beckmann–Kirchhoff model are compared to the Classical Beckmann–Kirchhoff theory and
the Rayleigh–Rice theory (adapted from Ref. [35]).

Because of the well-known π phase change experienced by an


Ac

R
θi
tua
Re

A C electromagnetic wave upon external reflection (n 2 ≥ n 1 ) from


l

Scattered by
θ i θ st
fer

Reflectance
θ sr the boundary between two dielectric media, [38] the minus sign
en

B
c e

is used when n 2 ≤ n 1 and the plus sign is used when n 2 ≥ n 1 .


Surface Height

O
n1 Following the derivation approach of the OHS and MHS
h theories, we find that a two-parameter family of surface transfer
ŷy
functions is required to characterize the scattering process for
θ st n2 arbitrary incident and scattering angles [19,35–37]
Scattered by
Transmittance
Hs x̂ , ŷ ; γi , γs

(θ st = − θ sr )

Fig. 24. Illustration of both forward (transmitted) and backward n  2  o


= exp − 2π σ̂rel (n 1 γi ∓ n 2 γs ) 1 − C s x̂ , ŷ /σs2 ,

(reflected) scattering from a moderately rough interface between two
media with arbitrary refractive indices. Courtesy of SPIE, this figure
previously published as Fig. 7 in Ref. [36]. (47)

where γi = cos θi , γs = cos θs , and C s (x̂ , ŷ ) is the surface ACV


function.
closed-form solutions for surfaces exhibiting Gaussian surface This general expression for the surface transfer function may
PSDs.
be used to model either reflective or transmissive scatter, the
It was thus necessary to develop a surface scatter theory that
latter being of interest, for example, in calculating the increased
would be valid for arbitrarily rough surfaces at arbitrary incident
efficiency of thin-film photovoltaic silicon solar cells by uti-
and scattered angles that would allow us to calculate image
lizing enhanced roughness on the TCO–Si interface [39–41].
degradation due to surface scatter from surfaces with arbitrary
However, the discussion in this paper will be restricted to appli-
surface PSDs.
cations of scattering from mirror surfaces, i.e., n 2 = − n 1 . If the
It was shown in Fig. 14 that the MHS surface scatter theory is
mirror is immersed in air (or vacuum), n 1 = 1, and Eq. (47) can
a significant improvement over the OHS theory, especially for
be written as [37]
large incident angles. However, the restriction of small scattering
angles was still very limiting. Furthermore, the OHS and MHS Hs x̂ , ŷ ; γi , γs

surface scatter theories were restricted to mirror surfaces, and
did not include the more general situation of scattering from
n  2  o
= exp − 2π σ̂rel (γi + γs ) 1 − C s x̂ , ŷ /σs2 .

a random rough interface between two media with arbitrary
refractive indices as shown in Fig. 24 [36]. (48)
The optical path difference (OPD) introduced when an
A separate surface transfer function is thus required for each
incident ray at an arbitrary angle, θi , is scattered at an arbitrary
incident angle and each scattered angle. Note that we are again
angle, θs , by a moderately rough interface between two media
representing the effective rms surface roughness with the band-
with arbitrary refractive indices can be written as
limited relevant rms roughness, σrel , used previously in Eq. (18).
OPD(x̂ , ŷ ) = −(n 1 cos θi − n 2 cos θs )h(x̂ , ŷ ), (45) The σs2 at the end of Eqs. (47) and (48) remains as the square
of the total, or intrinsic, surface roughness, as its purpose is to
where n 1 and n 2 are the refractive indices of the media before merely normalize the surface ACV function to unit height.
and after the interface, respectively. The phase variation intro- It was recognized as early as 1976 that a two-parameter fam-
duced for light scattered at an arbitrary angle is thus given by the ily of surface transfer functions would be required to accurately
expression model the scattered light behavior for an arbitrary incident and
an arbitrary scattered angle [7], but such a two-parameter fam-
φ(x̂ , ŷ ; γi , γs ) = 2π/λOPD
ily of surface transfer functions was considered too unwieldy and
computationally intensive to attempt to solve in 1976.
= −2π(n 1 cos θi ∓ n 2 cos θs ) · ĥ(x̂ , ŷ ). (46)
G200 Vol. 59, No. 22 / 1 August 2020 / Applied Optics Review

However, with the computer technology available today, it is not a function of the variables of integration, x̂ and ŷ . When a
is viable to numerically predict a two-dimensional scattering numerical solution is required, multiple transforms will indeed
function by setting γi and γs to some specific value then per- have to be performed in order to calculate the entire ASF. This is
forming a two-dimensional numerical Fourier transform of accomplished by specifying the parameters γs and γi before per-
the right side of Eq. (48) to produce the corresponding angle forming the two-dimensional Fourier transform, then repeating
spread function (ASF). However, only the one data point in the this procedure for each desired incident and scattered angle.
resulting two-dimensional numerical array that corresponds to Direction cosine space is a very convenient space in which
the fixed γi and γs is valid. New values of γi and γs can be chosen to perform the necessary calculations, as the entire observation
and the process repeated until a two-dimensional scattering hemisphere can be reduced to a circle with unit radius in the α,
function has been synthesized [37] β plane. This unit circle corresponds to θs = 90◦ for all values
Since the Fourier transform of the above surface transfer func- of φs in the spherical coordinate system. It is straightforward
tion yields the ASF for normal incidence, we can again invoke to create a numerical grid as a two-dimensional array, and then
the shift theorem of Fourier transform theory and express the perform the calculation of the scattering function, S, given by
scattered radiance distribution for an arbitrary incident angle Eq. (56), for each point in the array. For points that lie outside
(recall that θo = −θi ) by of the unit circle, we can simply assign a value of zero to S, since
γs is not a real number at those locations. The calculation of the
S(αs , βs ; γi , γs )
entire scattering function can thus be written as [37]
= F Hs (x̂ , ŷ ; γi , γs ) exp(−i2π βo ŷ ) α=αs ,β=βs . (49)
 XX
Sjk (α, β; γi ) = K (γi ) S(α j , βk ; γi , γjk ), (57)
Equation (49) corresponds to the scattered radiance in the αs , βs j k
direction. The direction cosines αs and βs are related to γs by
where
q

γs = 1 − αs2 − βs2 .
!−1
(50) Z 1 Z 1−α 2
K (γi ) = B(γi ) √
S(α, β − βo ; γi )dαdβ ,
α=−1 β=− 1−α 2
Once again, the surface transfer function can be written in the
form (58)
and
Hs x̂ , ŷ ; γi , γs = A(γi , γs ) + B(γi , γs ) G x̂ , ŷ ; γi , γs ,
 
S(α j , βk ; γi , γjk )
(51)
where = B(γi , γs )F {G(x̂ , ŷ ; γi , γjk ) exp(i2π βo ŷ )}|αs =α j ,βs =βk ,
n  2 o
A(γi , γs ) = exp − 2π (γi + γs ) σ̂rel , (52) (59)

n  2 o and
B(γi , γs ) = 1 − exp − 2π (γi + γs ) σ̂rel , (53) q
γjk = 1 − α 2j − βk2 . (60)
and

2
 The process just described is very computationally intensive
2 σrel
exp [2π (γi + γs )] σ 2 C s x̂ , ŷ − 1

since, for each scattering angle, we have to perform a two-
s
G(x̂ , ŷ ; γi , γs ) = . dimensional discrete Fourier transform of the surface transfer
exp 2π (γi + γs ) σ̂rel 2 − 1
 
function, extract the one data point corresponding to that scat-
(54) tering angle, and then repeat the process for all other desired
The ASF can then be written as the sum of a shifted delta scattering angles. If the surface roughness is isotropic, the sur-
function (specularly reflected beam) and an associated scattering face transfer function will be rotationally symmetric, and the
function, S(α, β; γi , γs ) [36]: two–dimensional Fourier transform in Eq. (59) reduces to
a Hankel transform. Since the Hankel transform operation
S(αs , βs ; γi , γs ) = [A(γi , γs )δ(α, β − βo )
is one-dimensional, this can help to reduce the computation
+ S(α, β − βo ; γi , γs )]|α=αs ,β=βs , time significantly. The process is still the same, except that
(55) at each step, a numerical Hankel transform is performed,
yielding the radial profile of the circularly symmetric interme-
where diate ASF. Using this profile, it is a simple matter to perform a
S(α, β − βo ; γi , γs ) one-dimensional interpolation to obtain the one point corre-
sponding to α j , βk . For the rough surface scattering predictions
= B(γi , γs )F G(x̂ , ŷ ; γi , γs ) exp(−i2π βo ŷ ) .

(56) that follow, we utilized a quasi-discrete Hankel transform
algorithm based on a Fourier–Bessel series expansion [42].
Equations (48) and (56) indicate that, for a given incident In 2011, the generalized Harvey–Shack (GHS) surface scat-
angle, a different Fourier transform needs to be performed for ter theory, characterized by a two-parameter family of STFs, had
each scattering angle in order to calculate the ASF. This process evolved into a practical modeling tool to calculate BRDFs from
can be avoided if the Fourier transform can be solved analyti- optical surface metrology data for situations that violate the
cally. When this is the case, γs is just treated as a constant since it smooth-surface approximation inherent in the Rayleigh–Rice
Review Vol. 59, No. 22 / 1 August 2020 / Applied Optics G201

Fig. 25. Comparison of scattered intensity predictions from the OHS, MHS, and GHS theories for different incident angles. Experimental data is
displayed for incident angles of 20◦ and 70◦ . The difference between the MHS and GHS theories is modest but significant, and the experimental data
provides excellent agreement with the GHS predictions. This figure previously published as Fig. 6 in Ref. [37].

theory and/or the moderate-angle limitation of the Beckmann– theories. Since we do not know the absolute radiometric value
Kirchhoff theory [37]. And finally, the STF can be multiplied by of the experimental data, we have normalized the peak value of
the classical OTF to provide a complete linear systems formu- the experimental data to be equal to that of the GHS theory. The
lation of image quality as degraded by diffraction, geometrical shape of the experimental curve is in excellent agreement with
aberrations, and surface scatter effects from residual optical that of the GHS theory.
fabrication errors [43]. For the 40◦ angle of incidence illustrated in Fig. 25(b), the
Using the procedure just described, we have performed a three theories still all predict that there is no significant specu-
comparison of the relative scattered intensity (in the plane of lar reflection (TIS ≥ 0.986). The MHS and GHS theories
incidence) predicted by the OHS, MHS, and GHS surface predict almost identical peak intensities (about 100% higher
scatter theories. The same surface as that used previously for than the OHS theory), and the peak of the GHS intensity is
Fig. 13 (Gaussian autocovariance function with σs = 2.27 µm centered upon the specular direction. The intensity distribu-
and `c = 20.9 µm) is used here. These comparisons are pre- tion predicted by the MHS theory is shifted to slightly smaller
sented in Fig. 25 for incident angles of 20◦ , 40◦ , 60◦ , and 70◦ . angles. The OHS scattered intensity distribution is considerably
We will initially assume that λ = 10.6 µm. At this wavelength, broader than that predicted by the other two theories.
σs /λ = 0.214, which is considered a moderately rough surface. For an incident angle of 60◦ , Fig. 25(c) shows that the peak
Fairly wide-angle scatter will be produced since the surface ACV intensity predicted by the MHS and GHS theories has increased
width, normalized by the wavelength, is given by `c /λ = 1.97. another 25% to 30%, with the MHS peak being somewhat
Experimental data from Ref. [21] will also be plotted for com- higher than that of the GHS theory. The GHS and MHS theo-
parison to the theoretical predictions for incident angles of 20◦ ries now predict a TIS equal to 0.792, which implies that ∼21%
and 70◦ . of the reflected radiation is now contained in the specular beam.
For the incident angle of 20◦ , Fig. 25(a) indicates that the The OHS intensity distribution has a peak intensity about 30%
GHS theory predicts a peak intensity that is approximately as high as the other two theories, and it is substantially broader.
27% higher than the OHS theory, and slightly higher than Finally, for an angle of incidence of 70◦ , Fig. 25(d) shows
the MHS theory. Although there is no significant specular that the peak intensities of the scattering function predicted by
reflection (TIS ≈ 1), the peak of the scattered intensity dis- the GHS and MHS theories are now substantially lower, as the
tribution lies essentially in the specular direction for all three predicted value of the TIS has dropped to 0.438. There is a more
G202 Vol. 59, No. 22 / 1 August 2020 / Applied Optics Review

of Diffraction. In this single colloquium talk, Professor Shack


removed the inherent paraxial limitation in the conventional
linear systems formulation of scalar diffraction theory. This
led to new insight and understanding, and enabled many new
applications that included:
1) Showing that near-field diffraction patterns are merely
aberrated Fraunhofer diffraction patterns, and those aber-
rations are our old friends: spherical aberration, coma,
astigmatism, etc.
2) Demonstrating that wide-angle diffraction phenomena is
shift-invariant relative to incident angle when formulated
in direction cosine space.
3) Surface scatter phenomena is merely a diffraction process
and therefore, from 2), it can be characterized by a surface
Fig. 26. Comparison of the MHS and GHS surface scatter the- transfer function; i.e., the original Harvey–Shack (OHS)
ories with the classical Beckmann–Kirchhoff scatter theory and the surface scatter theory was born.
O’Donnell–Mendez experimental data for a wavelength of 0.6328 µm 4) The OHS surface scatter was generalized to be valid for
and an incident angle of 70◦ . Courtesy of OSA, this figure previously large incident angles (grazing incidence x-ray telescopes).
published as Fig 7 in Ref. [37]. This modified Harvey–Shack (MHS) surface scatter theory
was characterized by a one-parameter family of surface
transfer functions (as are all imaging systems exhibiting
pronounced asymmetry in the scattered intensity distribution off-axis aberrations).
predicted by the MHS and GHS theories. The GHS theory 5) A linear systems formulation of non-paraxial scalar diffrac-
is an excellent fit to the experimental data, and a substantial tion theory was developed in which diffracted radiance
improvement over the MHS theory. It should also be noted that (not irradiance or intensity) was shown to be the funda-
experimental data is absent for those angles that represent the mental quantity that exhibits shift-invariant behavior when
location of the specular beam (presumably to protect the detec- displayed in direction cosine space.
tor), and the peak of both the measured and predicted scattered 6) The non-paraxial scalar diffraction theory in 5) was demon-
intensity distributions are shifted substantially (∼10◦ ) from strated to predict diffraction grating efficiencies with
the specular direction. This was one of the non-intuitive effects an accuracy usually thought to require rigorous electro-
that the authors of Ref. [21] were unable to explain. We now magnetic theory, including Rayleigh anomalies even for
understand that this shift occurs when the predicted truncated amplitude transmission gratings.
scattered radiance distribution is multiplied by the Lambert 7) The vaguely defined near boundary of the Fraunhofer
cosine function to convert it to a scattered intensity distribution. region (or the far field) was given a precise definition
In Fig. 26, we compare our predictions from the MHS and involving a tolerance upon the amount of allowable
GHS theories with the experimental measurements and the clas- defocus.
sical Beckmann–Kirchhoff theory as presented by O’Donnell 8) The radiometric units in the most popular western refer-
and Mendez, for the same surface as above, with an incident ence on scattering of electromagnetic waves from rough
angle of 70◦ and a wavelength of λ = 0.6328 µm. surfaces, Ref. [5], was clarified and validated by the (empir-
For this wavelength, the surface can be considered very rough ically) modified Beckmann–Kirchhoff surface scatter
since σs /λ = 3.59. Clearly all of the light is diffusely scattered. model.
The peak of the MHS theory is shifted substantially from the 9) A generalized Harvey–Shack (GHS) surface scatter theory
peak of the GHS theory, and is also about 40% higher. The was developed that is characterized by a two-parameter
classical Beckmann–Kirchhoff theory has its peak at approxi- family of surface transfer functions. This GHS surface
mately the same angle as that of the GHS theory, but it exhibits a scatter theory produces accurate results for rougher surfaces
nonphysical discontinuity at −90◦ . The shapes of all three theo- than the classical Rayleigh–Rice theory and (due to a more
retical profiles are thus quite different. The agreement between general obliquity factor) for larger incident and scattered
the GHS theory and the experimental data is again excellent, angles than either the classical Beckmann–Kirchhoff or
especially considering that this is a very rough surface with a Rayleigh–Rice theories [22]. The transfer function charac-
large incident angle. Figures 25 and 26 provide a convincing terization of scattering surfaces can be readily incorporated
experimental validation of the GHS surface scatter theory for into the traditional linear systems formulation of image
rough surfaces at large incident and scatter angles. formation, thus allowing a systems engineering analysis of
image quality as degraded by diffraction effects, geometri-
3. SUMMARY AND CONCLUSIONS cal aberrations, and surface scatter effects [43–45]. This, in
turn, allows us to derive the optical fabrication tolerances
In this paper, we have presented and reviewed a compilation necessary to satisfy a specific image quality requirement,
of nine separate advances in the understanding of applied which further enables the integration of optical fabrication
optics and optical engineering that were a direct result of and metrology into the optical design process [46].
Roland Shack’s 1974 colloquium talk entitled A Global View
Review Vol. 59, No. 22 / 1 August 2020 / Applied Optics G203

Disclosures. The author declares no conflicts of interest. 24. J. M. Palmer, “Getting intense about intensity,” Opt. Photon. News
30, 4 (1995).
25. E. L. Dereniak and G. D. Boreman, Infrared Detectors and Systems
REFERENCES (Wiley, 1996).
26. R. W. Boyd, Radiometry and the Detection of Optical Radiation
1. R. N. Bracewell, The Fourier Transform and its Applications (McGraw- (Wiley, 1983).
Hill, 1965). 27. J. E. Harvey, C. L. Vernold, A. Krywonos, and P. L. Thompson,
2. J. W. Goodman, Introduction to Fourier Optics, 2nd ed. (McGraw-Hill, “Diffracted radiance: a fundamental quantity in non-paraxial scalar
1996). diffraction theory,” Appl. Opt. 38, 6469–6481 (1999).
3. J. D. Gaskill, Linear Systems, Fourier Transforms, and Optics (Wiley, 28. J. E. Harvey, C. L. Vernold, A. Krywonos, and P. L. Thompson,
1978). “Diffracted radiance: a fundamental quantity in non-paraxial scalar
4. S. O. Rice, “Reflection of electromagnetic waves from slightly rough diffraction theory: eratta,” Appl. Opt. 39, 6374–6375 (2000).
surfaces,” Commun. Pure Appl. Math. 4, 351–378 (1951). 29. J. E. Harvey, A. Krywonos, and D. Bogunovic, “Non-paraxial scalar
5. P. Beckmann and A. Spizzichino, The Scattering of Electromagnetic treatment of sinusoidal phase gratings,” J. Opt. Soc. Am. A 23,
Waves from Rough Surfaces (Pergamon, 1963). 858–865 (2006).
6. R. V. Shack, Colloquium talk at the Optical Sciences Center at the 30. R. Petit, Electromagnetic Theory of Gratings (Springer-Verlag, 1980),
University of Arizona in 1974. p. 98.
7. J. E. Harvey, “Light-scattering characteristics of optical surfaces,” 31. E. G. Loewen and E. Popov, Diffraction Gratings and Applications
Ph.D. dissertation (University of Arizona, 1976). (Marcel Dekker, 1997).
8. J. E. Harvey and R. V. Shack, “Aberrations of diffracted wave fields,” 32. J. E. Harvey, A. Krywonos, and D. Bogunovic, “A tolerance on defo-
Appl. Opt. 17, 3003–3009 (1978). cus precisely locates the far field (exactly where is that far field any-
9. J. E. Harvey, “A Fourier treatment of near-field scalar diffraction the- way?),” Appl. Opt. 41, 2586–2588 (2002).
ory,” Am. J. Phys. 47, 974–980 (1979). 33. V. N. Mahajan, Optical Imaging and Aberrations: Part I (SPIE, 1998).
10. J. A. Ratcliff, “Some aspects of diffraction theory and their appli- 34. H. H. Hopkins, Wave Theory of Aberrations (Clarendon, 1950).
cation to the ionosphere,” in Reports of Progress in Physics, A. C. 35. J. E. Harvey, A. Krywonos, and C. L. Vernold, “A modified Beckmann-
Strickland, ed. (The Physical Society, 1956), Vol. XIX. Kirchhoff surface scatter model for rough surfaces with large incident
11. J. E. Harvey and C. L. Vernold, “Description of diffraction grating and scattering angles,” Opt. Eng. 46, 078002 (2007).
behavior in direction cosine space,” Appl. Opt. 37, 8158–8160 36. J. E. Harvey, N. Choi, A. Krywonos, S. Schroder, and D. H. Penalver,
(1998). “Scattering from moderately rough interfaces between two arbitrary
12. J. E. Harvey, “Surface scatter phenomena: a linear, shift-invariant media,” Proc. SPIE 7794, 77940V (2010).
process,” Proc. SPIE 1165, 87–99 (1989). 37. A. Krywonos, J. E. Harvey, and N. Choi, “Linear systems formulation
13. F. E. Nicodemus, “Reflectance nomenclature and directional of surface scatter theory for rough surfaces with arbitrary incident
reflectance and emissivity,” Appl. Opt. 9, 1474–1475 (1970). and scattering angles,” J. Opt. Soc. Am. A 28, 1121–1138 (2011).
14. R. Hufnagel, “Random wavefront effects,” Photo. Sci. Eng. 9, 38. B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics (Wiley,
244–247 (1965). 1991).
15. P. J. Chandley and W. T. Welford, “A re-formulation of some results 39. S. Fay, S. Dubail, U. Kroll, J. Meier, Y. Ziegler, and A. Shah, “Light
of P. Beckmann for scattering from rough surfaces,” Opt. Quantum trapping enhancement for thin-film silicon solar cells by rough-
Electron. 7, 393–397 (1975). ness improvement of the ZnC front TCO,” in Proc. 16th EU-PVSEC,
16. J. E. Harvey, E. C. Moran, and W. P. Zmek, “Transfer function charac- Glasgow, Scotland (2000), pp. 361–364.
terization of grazing incidence optical systems,” Appl. Opt. 27, 1527– 40. D. Domine, F. J. Haug, C. Battaglia, and C. Ballif, “Modeling of light
1533 (1988). scattering from micro- and nanotextured surfaces,” J. Appl. Phys.
17. P. Glenn, P. Reid, A. Slomba, and L. P. Van Speybroeck, “Performance 107, 044504 (2010).
prediction of AXAF technology mirror assembly using measured mir- 41. S. Schröder, A. Duparré, K. Füchsel, N. Kaiser, A. Tünnermann, and
ror surface errors,” Appl. Opt. 27, 1539–1543 (1988). J. E. Harvey, “Scattering of roughened TCO films—modeling and
18. J. E. Harvey and P. L. Thompson, “Generalized Wolter type I design measurement,” in Presented at OSA Topical Meeting on Optical
for the solar x-ray imager (SXI),” Proc. SPIE 3766, 173–183 (1999). Interference Coatings, Arizona, USA, June 7–9, 2010.
19. A. Krywonos, “Predicting surface scatter using a linear systems 42. M. Guzar-Sicairos and J. C. Gutierrez-Vega, “Computation of quasi-
formulation of non–paraxial scalar diffraction,” Ph.D. dissertation discrete Hankel transforms of integer order for propagating optical
(University of Central Florida, 2006). wave fields,” J. Opt. Soc. Am. A 21, 53–58 (2004).
20. J. E. Harvey, S. Schroder, N. Choi, and A. Duparre, “Total integrated 43. J. E. Harvey, “Parametric analysis of the effect of scattered light upon
scatter from surfaces with arbitrary roughness, correlation widths, the modulation transfer function,” Opt. Eng. 52, 073110 (2013).
and incident angles,” Opt. Eng. 51, 013402 (2012). 44. N. Choi and J. E. Harvey, “Numerical validation of the Generalized
21. K. A. O’Donnell and E. R. Mendez, “Experimental study of scat- Harvey-Shack surface scatter theory,” Opt. Eng. 52, 115103 (2013).
tering from characterized random surfaces,” J. Opt. Soc. Am. A 4, 45. N. Choi and J. E. Harvey, “Image degradation due to surface scatter
1194–1205 (1987). in the presence of aberrations,” Appl. Opt. 51, 535–546 (2012).
22. J. E. Harvey, Understanding Surface Scatter Phenomena (SPIE, 46. J. E. Harvey, “Integrating optical fabrication and metrology into the
2019). optical design process,” Appl. Opt. 54, 2224–2233 (2015).
23. J. J. Muray, F. E. Nicodemus, and I. Wunderman, “Proposed supple-
ment to the SI nomenclature for radiometry and photometry,” Appl.
Opt. 10, 1465–1468 (1971).

View publication stats

You might also like