JOSA COMMUNICATIONS

Communications are short papers. Appropriate material for this section includes reports of incidental research results,
comments on papers previously published, and short descriptions of theoretical and experimental techniques.
Communications are handled much the same as regular papers. Proofs are provided.

On the zero-thickness model of diffractive

optical elements

Markus Testorf
University of Massachusetts-Lowell, Department of Electrical and Computer Engineering, One University Avenue,
Lowell, Massachusetts 01854

Received October 28, 1999; accepted March 10, 2000

In a recent paper [J. Opt. Soc. Am. A 16, 113 (1999)] a thin-element approximation of diffractive optical ele-
ments was used to describe diffraction of oblique incident wave fronts. This expression motivated by a ray
optical analysis is shown to be incorrect. I discuss how the thin-element approximation can be generalized to
arbitrary diffraction geometries. This includes an intuitive interpretation of the results. © 2000 Optical So-
ciety of America [S0740-3232(00)01406-X]
OCIS codes: 050.1940, 050.1970.

This communication was prompted by several private
communications, as well as by a recent publication on the
comparison between rigorous and approximate diffraction
models.1 The problem that is addressed here is the ap-
proximation of diffractive optical elements (DOE’s) as
thin diffraction screens, which is valid for paraxial wave
fronts only.2 This approximation is referred to as the
thin-element approximation (TEA), or zero-thickness
model. In diffractive optics this model is also termed sca-
lar approximation.1 In particular, how the phase-only
transmission function relates to the profile of a surface-
relief DOE in the case of an oblique incident wave front is
discussed.
The surface-relief structure can be described as a depth
profile z ⫽ h(x), with x and z being the lateral and lon-
gitudinal coordinates, respectively. The TEA yields the
corresponding transmission function as a phase-only
modulation of the incident complex amplitude distribu-
tion. The phase distribution is given as
2␲
␾ 共 x兲 ⫽ h 共 x兲 ⌳,
␭ proximation for reflecting boundaries,6 the EKA does not
with ␭ being the wavelength of the monochromatic signal.
For dielectric interfaces and an incident plane wave
propagating along z, one finds ⌳ ⫽ n 1 ⫺ n 2 , where n 1
and n 2 are the refractive indices on either side of the in-
terface. For reflecting surfaces it is ⌳ ⫽ ⫺2, where the
sign accounts for the fact that a larger height h causes a
smaller phase delay. Equation (1) can be interpreted as
a projection of the surface profile scaled with the optical
path-length difference in wavelength units.
In Fig. 1 the geometry of a reflecting surface with two
different modulation depths h 1 and h 2 is shown. The
phase distribution with which an incident wave parallel
to the z coordinate is modulated can be calculated as the
optical path length of the projections AB and CD from a
reference plane z 0 to the diffractive surface and of the
projections BA and DC from the surface back to the ref-
erence plane. The choice of z 0 corresponds to a constant
phase in the transmission function, and Eq. (1) can be ob-
tained for z 0 ⫽ 0.
If the angle between the direction of the incident wave
and the optical axis z exceeds the paraxial domain, the
phase modulation changes. A recent publication1 derives
the modified expression for the phase modulation by cal-
culating the optical path-length distribution from projec-
tions A 1 BA 2 and C 1 DC 2 , respectively. This yields a
phase distribution that is characterized by substitution of
⌳ ⫽ ⫺2/cos ␣ into Eq. (1). Although this idea on how the
TEA has to be modified in the case of an oblique incidence
seems to be rather widespread, it is incorrect neverthe-
less.
A formal way to derive the correct expression for the
TEA is provided by the Born approximation (BA) of
surface-relief elements.3,4 For reflecting boundaries, an
equivalent expression is provided by the extended Kirch-
(1) hoff approximation (EKA).5 Unlike the Kirchhoff ap-
assume surface fields to vanish in geometrical shadow re-
gions. Both the BA and the EKA are not limited to
paraxial wave fronts. In fact, the TEA can be derived as
a small-angle approximation of the BA,3 which makes it
well suited to explore the validity of the TEA. Assuming
a two-dimensional geometry and a plane-wave incident at
an angle ␣, BA and EKA predict the plane-wave spectrum
of the diffracted wave as
ũ 共 ␯ 兲 ⫽ C 关 ␤ 共 i 兲 , ␤ 兴 冕
⫺⬁
⬁
exp(⫺i2 ␲ 兵 关 ␤ ⫺ ␤ 共 i 兲 兴 h 共 x, y 兲
⫹ 关 ␯ ⫺ ␯ 共 i 兲 兴 x 其 )d x, (2)

0740-3232/2000/061132-02$15.00 © 2000 Optical Society of America

JOSA Communications
Fig. 1. Geometry of diffraction at a reflecting step profile.
where the factor C 关 ␤ ( i ) , ␤ 兴 depends on the type of diffrac-
tive structure. The incident wave is specified by a spatial
frequency ␯ ( i ) ⫽ sin ␣/␭ and a propagation constant ␤ ( i )
⫽ cos ␣/␭. Equally, each diffracted plane-wave compo-
nent is described by a frequency ␯ and a propagation con-
stant ␤. To retrieve the TEA from Eq. (2), we assume the
propagation constant to be equal for all diffraction angles,
i.e., 关 ␤ ⫺ ␤ ( i ) 兴 ⬇ ⫺2 ␤ ( i ) for reflecting surfaces. Equation
(2) can be recognized as the Fourier transformation of a
phase-only function. A comparison of this phase function
with Eq. (1) yields
⌳ ⫽ ⫺2 cos ␣ ,
where we have assumed ␤ ⬍ 0. A similar analysis can
(i)
be carried out, if the diffractive structure changes the ori-
entation of the main propagation direction, for instance,
by use of a high-frequency diffraction grating. If all dif-
fracted plane-wave components propagate with a direc-
tion that is well approximated by ␤ ⬇ cos ␣⬘/␭ we find
⌳ ⫽ ⫺共 cos ␣ ⬘ ⫹ cos ␣ 兲 .
Finally, the thin-element approximation for dielectric
surface profiles is derived by appreciating that the propa-
gation constants are ␤ ( i ) ⫽ n 1 cos ␣/␭ and ␤ ⫽ n 2 cos ␣⬘/␭
before and behind the interface, respectively. The corre-
sponding phase distribution is defined by
⌳ ⫽ n 1 cos ␣ ⫺ n 2 cos ␣ ⬘ .
Note that essentially the same relation for the phase
modulation was derived independently for use in a wide-
angle beam propagation method7 and can be found in
textbooks as well.2
It seems worthwhile to discuss why the heuristic deri-
vation from the projection of the ray optical path does not
Vol. 17, No. 6 / June 2000/ J. Opt. Soc. Am. A 1133
provide the correct expression. To this end, it is impor-
tant to realize that the TEA provides a transmission func-
tion that allows calculation of the diffracted field from a
pointwise multiplication with the incident wave front.
For instance, to obtain the diffracted wave at point A the
incident wave at point A is multiplied with a transmission
function evaluated at the same point. By calculation of
the path A 1 BA 2 , in Fig. 1, however, the diffracted wave
at point A2 is calculated from the incident field at point
A1, which is not in agreement with the concept of a trans-
mission function. Instead of analyzing the propagation
of optical rays, one has to consider the corresponding
plane wave propagating along A 1 BA 2 . The lines AP 1
and AP 2 mark planes of constant phase of the incident
and the reflected wave, respectively, both intersecting
with point A. Therefore the phase delay at point A is
given as the optical path length necessary to propagate
the wave front from P 1 to B and from B to P 2 . This
again yields the correct result ⌳ ⫽ ⫺2 cos ␣.
In summary, I have generalized the relation between
the profile of surface-relief DOE’s and the phase modula-
tion predicted by the TEA. An analysis based on the
EKA shows that a simple evaluation of the ray optical
path-length distribution may lead to incorrect results. It
should be pointed out that the use of the correct expres-
sion for the TEA does not only influence the accuracy of
the predicted diffraction pattern. More significant, if the
thin-element model is used to calculate the required etch
depth for a photolithographic fabrication of DOE’s, it di-
rectly impacts the performance of the element and its
function in an optical system.
The author can be reached at the address on the title
(3) page or by telephone at 978-934-3306, by fax at 978-934-
3027, or by e-mail at testorf@galileo.eng.uml.edu.
REFERENCES
1. J. M. Bendickson, E. N. Glytsis, and T. K. Gaylord, ‘‘Metal-
lic surface-relief on-axis and off-axis focusing diffractive cy-
(4) lindrical mirrors,’’ J. Opt. Soc. Am. A 16, 113–130 (1999).
2. B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics
(Wiley, New York, 1991), Chap. 2, pp. 55–63.
3. W. Singer and K.-H. Brenner, ‘‘Transition of the scalar field
at a surface in the generalized Kirchhoff diffraction theory,’’
J. Opt. Soc. Am. A 12, 1913–1919 (1995).
4. M. Testorf, ‘‘Perturbation theory as a unified approach to
design diffractive optical elements,’’ J. Opt. Soc. Am. A 16,
(5) 1115–1123 (1999).
5. P. C. Waterman, ‘‘Scattering by periodic surfaces,’’ J.
Acoust. Soc. Am. 57, 791–802 (1975).
6. A. Ishimaru, Electrodynamic Wave Propagation, Radiation,
and Scattering (Prentice-Hall, Englewood Cliffs, N. J.,
1991), Chap. 6.3, pp. 156–157.
7. K.-H. Brenner and W. Singer, ‘‘Light propagation through
microlenses: a new simulation method,’’ Appl. Opt. 32,
4984–4988 (1993).