Analysis of optical elements with the local

plane-interface approximation

Albrecht v. Pfeil, Frank Wyrowski, Andreas Drauschke, and Harald Aagedal

The local plane-interface approximation 共LPIA兲 is a method for propagating electromagnetic fields
through the inhomogeneous regions 共e.g., elements兲 of an optical system. The LPIA is the superclass of
all approximations that replace the usually curved optical interfaces with local tangential planes.
Therefore the LPIA is restricted to smooth optical surfaces. A maximum radius of curvature of the
optical interface of the order of a few wavelengths is a rough estimate for the validity of the LPIA. Two
important approximation levels of the LPIA are the thin-element approximation 共TEA兲 and a geometric-
optical version of the LPIA 共LPIAray兲. The latter combines the wave-optical propagation of an electro-
magnetic field in the homogeneous region of an optical system with a ray-tracing step in the
inhomogeneous region. We discuss the regions of validity of the LPIA in general and the approximation
levels LPIAray and TEA in detail. © 2000 Optical Society of America
OCIS codes: 350.4600, 050.1940, 120.5710, 050.1970, 350.3950.

1. Introduction shown in Fig. 1. Ray tracing does not distinguish

In physical optics system design and analysis, the between homogeneous and inhomogeneous media, as
problem of free propagation of light in a homogeneous rays are used to propagate the electromagnetic field
medium has been solved,1,2 though some numerical through the whole optical system. The effects of dif-
problems still exist in the nonparaxial case. How- fraction are neglected. However, especially in
ever, the process of propagation of light through an micro-optics, there are many optical systems for
inhomogeneous medium is still a major topic of in- which, because of the small dimensions of the com-
vestigation. Figure 1 shows schematically the basic ponents, the wave nature of light cannot be ignored or
geometry of an important case of inhomogenity. is even utilized to produce the desired optical func-
Two dielectric media with different real refractive tion. Therefore we want to take a wave-optical ap-
indices are separated by an arbitrarily shaped inter- proach to propagate the electromagnetic field
face. The resultant three regions are formally sep- through the inhomogeneous region.
arated by planes zI and zII. In this paper we discuss In paraxial-wave optics the inhomogeneous regions
approaches to analysis of optical elements, the optical of an optical system are usually analyzed by use of
function of which is realized by a single modulated the thin-element approximation 共TEA兲.4 As dis-
interface of this type. Furthermore, we restrict our- played in Fig. 2共a兲, the field outside the inhomoge-
selves to a monochromatic incident wave, Uin. neous region is propagated by wave optics. In the
In optical engineering, ray tracing3 is a common inhomogeneous region the optical path is taken into
tool for handling the optical effects of an interface as account without consideration of the direction of in-
put wave Uin or the change of direction by refraction
at the optical interface. Thus the use of the TEA is
A. v. Pfeil 共[email protected]兲 is with the Fraunhofer Institut, An- restricted to interfaces and to incident waves for
gewandte Optik und Feinmechanik, Schillerstrasse 1, D-07745 which the effects of refraction can be neglected, that
Jena, Germany. F. Wyrowski and A. Drauschke are with the is, for paraxial incident and transmitted waves. As
Institut für Angewandte Physik, Friedrich-Schiller-Universität is discussed in more detail below, the TEA can
Jena, Max-Wien-Platz 1, D-07743 Jena, Germany. H. Aagedal is
be understood as the first approximation level of
with Kongsberg Defence & Aerospace AS, Avd. Asker, P.O. Box
265, N-1372 Asker, Norway. the local plane-interface approximation 共LPIA兲,
Received 8 December 1999; revised manuscript received 17 April 共LPIA1st ⫽ TEA兲.
2000. In recent years various manufacturing techniques
0003-6935兾00兾193304-10$15.00兾0 for obtaining optical elements with heights of several
© 2000 Optical Society of America micrometers that can perform large-angle deflection

3304 APPLIED OPTICS 兾 Vol. 39, No. 19 兾 1 July 2000

 version plane, the rays are transformed back into an
electromagnetic field.
2. Fundamentals of the Local Plane-Interface
Approximation
In this section we give a closer definition of what we
call the LPIA. Figure 2共b兲 shows the geometric-
optical version of the local plane-interface approach,
LPIAray. At zI the field is sampled equidistantly.
Each sampling point is a starting point of a ray. The
ray carries information on its position, direction, in-
tensity, phase, and polarization. The direction of
the ray can be determined in either of two ways.
Fig. 1. Schematic illustration of the typical inhomogenity consid- One possibility is by use of the local plane-wave ap-
ered in this paper. It is an arbitrarily shaped interface separated proximation. In this approximation the wave front
by two homogeneous dielectrics. of the incident field is locally approximated as a plane
wave. Each of these plane waves is represented by
a ray, which can be traced through the inhomoge-
have been invented and improved.5–10 Several neous medium. If the local plane-wave approxima-
kinds of so-called multi-order elements with large tion is not valid, e.g., in the case of large curvatures
structure heights have been discussed and in the wave front, the incident wave can be decom-
fabricated.11–15 For many of these kinds of element posed into a spectrum of plane waves.2 With the
the TEA is not a valid approximation. LPIA each of these plane waves can be propagated
In this paper we investigate the second level of through the inhomogeneous medium. For the sake
approximation of the LPIA. This model combines of simplicity we restrict ourselves to plane input
the wave-optical propagation of light in homogeneous waves at normal incidence.
media with a ray-tracing step in the inhomogeneous Within the inhomogeneous region each ray is
region 共LPIA2nd ⫽ LPIAray兲 and is illustrated in Fig. traced to the modulated optical interface. In the re-
2共b兲. An input field Uin is propagated to a plane zI by gion about the intersection point of a ray with the
wave optics. In this plane, which we call the first interface, the interface is replaced by a local tangen-
conversion plane, the electromagnetic field is trans- tial plane. The influence of this locally plane inter-
formed into rays, which then are traced through the face on the ray is considered.
inhomogeneous region. In plane zII, the second con- In the second conversion plane, zII, the optical path
difference 共OPD兲 of the rays yields the phase Arg
关Uout共x, y, zII兲兴. The intensity and the change of ray
density at zI and zII, respectively, account for the
amplitude distribution 兩Uout共x, y, zII兲兩.
We specify that all approximations that replace the
optical interface with local tangential planes fall into
the superclass of the LPIA. In its most accurate
geometric-optical version the direction of incident
wave Uin, the refraction and reflection of the rays at
the optical interface according to Snell’s law,
Fresnel’s formulas, and the OPD of the rays are con-
sidered. Here we use a slightly lower-level version
of the LPIA, as we regard only sequential ray tracing
in the transmission mode of operation. This approx-
imation is LPIAray. If we subsequently drop some
features of the LPIA, we come the lowest-level ver-
sion, for which only the OPD of the rays is considered.
This first approximation level of the LPIA is usually
referred to as the TEA.
The model of LPIA follows the ideas of Beck-
mann.16 Swanson17 applied a similar theory, which
he called the “extended scalar theory,” to optimize the
depth of blazed gratings. Noponen et al.18 and Rossi
et al.19 used the method to predict the diffraction
efficiency of blazed gratings and diffractive lenses,
Fig. 2. Illustration of the analysis of an optical element by two respectively. Stamnes and Heier20 gave a detailed
methods. In both cases the field is propagated by wave optics in description of a computational technique based on
the homogenous region. In the inhomogeneous region 共a兲 the the “combined method of ray tracing and diffraction,”
TEA 共a兲 and 共b兲 LPIAray are applied. a version of the LPIA. Our aims in this study are to

1 July 2000 兾 Vol. 39, No. 19 兾 APPLIED OPTICS 3305

Fig. 3. Curvature C of sinusoidal gratings. The gratings have a
height of ␭. r ⫽ 1兾C ⫽ 1.3␭, ␭ ⫽ 632.8 nm.

discuss general limits of the LPIA and to investigate

the regions of validity for various approximation lev-
els of the LPIA.
3. Limitations to the Curvature of the Interface
The name of the LPIA already implies that the optical
interface is considered to be locally plane. For ex-
ample, for LPIAray the region about the intersection
point of a ray with the interface is approximated by a
tangential plane. To estimate which kinds of sur-
face can be analyzed with the LPIA, we assume that Fig. 4. Rigorously calculated amplitudes of fields behind the pe-
a ray has an extension of the order of the wavelength. riods of sinusoidal gratings with height ␭, refractive index 1.5, and
The interface should be plane in this region, which grating periods 共a兲 2.5 ␮m and 共b兲 10 ␮m. The gratings are illu-
restricts the application of the LPIA to smooth sur- minated with plane waves with wavelength ␭ ⫽ 632.8 nm under
faces. A measure for the smoothness is the curva- normal incidence.
ture of the optical interface. Curvature C of a
function y共x兲 can be defined as follows21:
with the output field obtained with the TEA, the first
y⬙共x兲 level of the LPIA. With this approximation, the am-
C共x兲 ⫽ , (1) plitude of the transmitted field is constant. The am-
关1 ⫹ y⬘共x兲2兴3兾2
plitude of the rigorously calculated field behind the
where y⬘共x兲 and y⬙共x兲 are the first and second deriv- grating with a period of 2.5 ␮m shows large varia-
atives, respectively, of y共x兲. tions 关Fig. 4共a兲兴. These variations are due to reso-
Figure 3 illustrates the curvature of sinusoidal nance in the grating and cannot be simulated with
gratings with periods of 10 and 2.5 ␮m and a height the TEA. For the grating with a period of 10 ␮m, the
of ␭ ⫽ 632.8 nm. To relate the curvature of the variations in amplitude are much smaller 关Fig. 4共b兲兴.
gratings to the wavelength, we have included the This calculation agrees much better with the TEA.
curvature of a surface with a radius of curvature r ⫽ The phase of the field exhibits the same behavior as
1兾C ⫽ 1.3␭ in Fig. 3. Compared with this surface, the amplitude.
the grating with a period of 2.5 ␮m has a larger, and By using the curvature of an optical interface as a
the grating with a period of 10 ␮m a smaller, maxi- measure of the suitability of the LPIA, one cannot
mum curvature. Although the curvature cannot properly analyze surfaces with vertical steps 共e.g.,
provide an exact measure of the kinds of surface relief blazed gratings or diffractive lenses兲 with this ap-
that are suitable for the LPIA, the comparison of the proximation. In theory, at the steps the curvature is
minimum radius of curvature with a radius of curva- infinite. Figure 5 shows the rigorously calculated
ture of 1.3␭ can give a rough estimate. The correla- amplitudes of the output fields of modulo 2␲ blazed
tion between the curvature of the interface and the gratings with refractive indices n1 ⫽ 1.5 and n2 ⫽ 1.0.
suitability of the LPIA for the analysis of this inter- The amplitude behind the grating with a period of 2.5
face is further illustrated in Fig. 4. The amplitudes ␮m has large variations 关Fig. 5共a兲兴, which cannot be
of output fields of sinusoidal gratings with periods of reproduced with the TEA. The transmitted ampli-
2.5 and 10 ␮m and refractive indices n1 ⫽ 1.5 and tude of the grating with a period of 10 ␮m has only
n2 ⫽ 1.0, obtained by a rigorous approach, are dis- small variations in the center of the prism grooves
played. For clarity we compare the rigorous results 关Fig. 5共b兲兴. However, at the vertical steps, where the

3306 APPLIED OPTICS 兾 Vol. 39, No. 19 兾 1 July 2000

 Fig. 6. First-order diffraction efficiencies of sinusoidal gratings as
a function of period. The gratings have a height of ␭ ⫽ 632.8 nm
and are illuminated with on-axis plane waves in TE polarization.
They are analyzed by the TEA, LPIAray, or electromagnetic anal-
ysis 共dotted curve兲.

of a sinusoidal grating with a height of ␭ as a function

of the grating constant. With the TEA, the diffrac-
tion efficiency remains constant. Calculations with
LPIAray are also shown. With decreasing grating
constants the diffraction efficiency diminishes. The
effects that account for the difference between LPI-
Aray and TEA are discussed in Section 5 below.
Down to grating constants of about 3.2 ␮m the rig-
orous calculations 共also shown in Fig. 6兲 can be re-
produced very well with LPIAray. This grating
period corresponds to a minimum radius of curvature
Fig. 5. Rigorously calculated amplitudes of the fields behind the of rmin ⫽ 1.3␭. For smaller grating constants the
periods of blazed gratings with height 2␭, refractive index 1.5, and deviations between LPIAray and the rigorous ap-
grating periods 共a兲 2.5 and 共b兲 10 ␮m. The gratings are illumi- proach increase because of the restrictions on the
nated with plane waves with wavelength ␭ ⫽ 632.8 nm under curvature of the optical interface discussed above.
normal incidence.
Simulations of sinusoidal gratings with various re-
fractive indices show that the validity of the LPIA
depends not only on the maximum curvature Cmax of
LPIA is not valid, the variations in amplitude are the interface but also on the difference of the refrac-
large and cannot be simulated with the TEA. Again, tive indices ⌬n ⫽ 兩n1 ⫺ n2兩. It turns out that the
the phase of the field shows the same behavior as the curvature of the product h共x兲⌬n is a good measure for
amplitude. the validity of the LPIA, where h共x兲 is the height
So far we have had a qualitative discussion of the profile of the interface. If the minimum radius of
general limits on a LPIA for modeling the optical curvature of this product is smaller than rmin ⫽ 2.5␭,
effect of an interface. To this end we combined rig- the LPIA is not a valid approximation.
orous calculations with the lowest approximation The first-order diffraction efficiency of a blazed
level of the LPIA, that is, the TEA. Now we turn to grating with a height of 2␭ is shown in Fig. 7. For a
the second approximation level of LPIA, that is, LPI- perfect grating the first-order diffraction efficiency
Aray, to get more information about what kinds of calculated with the TEA has a constant value of 1.
interface can be analyzed with the LPIA. We use The simulations performed with LPIAray reproduce
LPIAray to analyze gratings with different grating the rigorous results much better than the calcu-
constants. These results are compared with rigor- lations carried out with the TEA. However, for
ous calculations. To illustrate the differences be- grating constants as large as 25 ␮m, LPIAray over-
tween versions of the LPIA, we also perform the estimates the first-order diffraction efficiency by 2%.
calculations with the TEA. The first-order diffrac- The reason for this is the presence of vertical steps of
tion efficiency of the gratings is used as a merit func- the blazed grating where the LPIA is not valid in
tion. The refractive indices of the grating are n1 ⫽ general.
1.5 and n2 ⫽ 1.0. The illumination wave is TE po- It can be summarized that the LPIA is restricted to
larized and has a wavelength ␭ ⫽ 632.8 nm. surfaces with moderate curvature. If the curvature
Figure 6 shows the first-order diffraction efficiency is very large, e.g. in the case of vertical steps, the

1 July 2000 兾 Vol. 39, No. 19 兾 APPLIED OPTICS 3307

Fig. 7. First-order diffraction efficiencies of blazed gratings as a
function of period. The gratings have height 2␭ and are illumi-
nated with on-axis plane waves in TE polarization with wave- Fig. 8. Maximum distance of propagation dmax of a Gaussian
length ␭ ⫽ 632.8 nm. They are analyzed by the TEA, LPIAray, or intensity distribution with a 1兾e2 full width w0, assuming that
an electromagnetic analysis. w共dmax兲兾w0 ⫽ 1.05. The wavelength of the Gaussian beam is ␭ ⫽
632.8 nm.

LPIA is not valid. However, if the region about a

vertical step accounts for only a small portion of the the LPIA results in local planes with extensions typ-
interface 共e.g., in the case of large grating periods兲 the ically not larger than w0 共in fact, that is the necessary
error in the approximation can be neglected. A more grid period of the starting points of rays in the first
detailed investigation of the limits of the TEA for conversion plane兲 the distance from the first to the
elements with vertical steps is given by Pommet et second conversion plane will not exceed dmax共w0兲.
al.22 Thus the smallest feature of the interface limits the
thickness of the inhomogeneous region, which can be
4. Estimation of the Influence of Diffraction in the analyzed with LPIAray. Obviously, a restriction of
Inhomogeneous Region the thickness is well known for the lowest approxi-
LPIAray uses ray tracing to propagate the electromag- mation level of the LPIA, that is, the TEA. Because
netic field through the inhomogeneous region of the LPIAray takes the refraction effect into account, it is
optical system. During this ray-tracing step the ef- suitable for analysis of thicker elements. However,
fects of diffraction are neglected. In this section we the neglect of diffraction in the inhomogeneous region
give a rough estimation of this approximation. still mandates careful consideration of the depen-
In our discussion we assume that the optical inter- dence of the thickness restrictions on the features of
face locally produces a Gaussian amplitude distribu- the interface.
tion with a plane phase instead of a ray. As a
Gaussian wave retains its shape during its propaga- 5. Effects of Different Approximation Levels of a Local
tion, it is appropriate to use this wave to estimate the Plane-Interface Approach
effect of diffraction. The broadening of a Gaussian In the preceding sections we investigated some gen-
beam with a 1兾e2 full width of w0 that propagates a eral limitations of the LPIA and LPIAray. Now we
distance d can be expressed by23 turn to a discussion of differences between two dif-

冋 冉 冊册 2 1兾2
ferent versions of the LPIA, i.e., LPIAray and the
w共d兲 4d␭ TEA.
⫽ 1⫹ , (2)
w0 ␲w02 LPIAray considers three effects which are neglected
by the TEA: change in the optical path as a result of
where ␭ is the wavelength of the Gaussian beam and refraction, amplitude modulation of the output field
w0 is the width of the beam at its waist, which in our owing to the various densities of the rays in the sec-
is model the width of the local region. The neglect of ond conversion plane, and amplitude modulation of
diffraction seems to be justified if the width of the the output field according to Fresnel’s formulas.
beam after propagation w共d兲 does not significantly To get a basic understanding of the influence of
exceed w0, because then the light from one interface these effects we should examine the example of a
location does not seriously interfere in the second prism with varying width d and height h. As illus-
conversion plane with the light from another inter- trated in Fig. 9, varying d and h results in varying the
face location. A bit arbitrarily, let us assume that angle of refraction ␸. The prism is a quite general
the width after propagation is not larger than 105% of example of an interface, because it can be interpreted
the width on the interface. Then the resultant max- as the local plane-interface segment of the modulated
imum distance of propagation dmax as a function of w0 interface of an optical element, as illustrated in Fig.
is as displayed in Fig. 8. 10. According to the definition of the LPIA, each
This result can be interpreted in the following way: region about an intersection point of a ray rn with the
If the application of the geometric-optical version of optical interface is treated as a plane surface. This

3308 APPLIED OPTICS 兾 Vol. 39, No. 19 兾 1 July 2000

 Fig. 11. Difference in optical path between TEA and LPIAray as a
function of ␸. The dashed line indicates an error in the optical
path of ␭兾10.

A. Change in the Optical Path Caused by Refraction

The TEA neglects refraction at the optical interface.
Fig. 9. Ray tracing through a prism with base d and height h, As shown in Fig. 9, this results in two effects that
which results in deflection angle ␸. The illuminated region be- yield errors of the phase of Uout. The length of the
hind the prism has width d̃. s is the distance that a ray with the optical path to zII is changed, and the intersection
starting point 共x, y, zI兲 has to be traced to hit the wave front, which point of the ray with zII is not calculated correctly.
runs through 共x, y, zII兲. For plane output waves Uout, as are present in the
example of the prism, the two effects can be treated
together. The TEA is as good as LPIAray if it yields
the same wave front at zII as the one specified by the
surface can also be interpreted as a prism with height direction of the rays that is due to application of
hn. Therefore the results for a prism should also be LPIAray. This is the case if
applicable for a more general interface. Hence we
also consider the aspherical interface of a lens with a ⌬s ⫽ n2共h ⫺ s兲 ⬍⬍ ␭, (3)
numerical aperture 共N.A.兲 of 0.4, refractive indices
where ␭ is the wavelength of the input wave. There-
n1 ⫽ 1.5 and n2 ⫽ 1.0, and a diameter of 500 ␮m to
fore the TEA is a valid approximation in the frame-
discuss the effects mentioned above.
work of LPIAray if the condition

⌬s hn2
⫽ 共1 ⫺ cos ␸兲 ⬍⬍ 1 (4)
␭ ␭
is satisfied. That is, prism height h and angle of
refraction ␸ have to be small. Let us consider that a
bit more quantitatively. In Fig. 11 ⌬s兾␭ is displayed
as a function of ␸. ⌬s兾␭ is calculated for three prism
heights h. ⌬s ⫽ ␭兾10 indicates a rule of thumb for
the applicability of the TEA. With this criterion, the
TEA is useful up to large angles of refraction if the
height of the prism is only 0.4 ␭ 共e.g., a first-order
blazed silicon grating兲. If the height is 2␭ 共e.g., a
first-order blazed glass grating兲, prisms with deflec-
tion angles up to ⬃18° can be analyzed with the TEA.
A micro-prism with a height of 20 ␭ has to be ana-
lyzed with LPIAray if the angle of refraction exceeds
6°. Figure 12 serves to give an idea for which kinds
of interface LPIAray instead of the TEA has to be
Fig. 10. An optical interface can be thought to be composed of applied. Using the assumption that the optical path
approximately local prisms. The prism angle is defined by the should be calculated with an accuracy of at least
tangential plane on the interface at the intersection of a ray rn. ␭兾10, we display the height of the prism as a function
The height of the prism hn is the distance from the intersection of of the angle of refraction. All elements located in the
rn to zII. shaded region of Fig. 12 can be analyzed with the

1 July 2000 兾 Vol. 39, No. 19 兾 APPLIED OPTICS 3309

Fig. 12. Overview of prisms expressed by ␸ and the height of the
prism. The locations of the elements that can be considered with Fig. 14. Quotient ITEA兾ILPIAray of the intensities in the second
the TEA or have to be analyzed with LPIAray are shown. The conversion plane of a prism obtained with LPIAray and the TEA,
calculations are based on a maximum error in the optical path of calculated for two refractive indices n of the prism.
␭兾10.

tical steps, such as blazed gratings and Fresnel

TEA. For all other elements the refraction at the lenses, this width reduction manifests itself as a re-
interface cannot be ignored. duction of the maximum intensity in the far field,
Now we discuss the phase errors of the TEA for a which is referred to as the shadowing effect.24 The
lens with a N.A. of 0.4. Figure 13共a兲 shows the OPD intensity modulation in the second conversion plane,
behind the lens analyzed with LPIAray and the TEA. zII, of the prism can be expressed as a function of the
Close to the optical axis, both approximations give angle of refraction ␸:
the same results. Near the margins of the lens the
difference in the optical path increases. This result ITEA d̃ n2 sin ␸
is due to the growing angle of refraction and the ⫽ ⫽1⫺ tan ␸, (5)
ILPIAray d n1 ⫺ n2 cos ␸
increasing height of the prism segments of which the
lens can be thought to be composed. Using the cri- where n1 and n2 are refractive indices of the prism
terion that the error in the optical path should not and outside the prism, respectively. From Eq. 共5兲 it
exceed ␭兾10, we can analyze lenses with diameters of becomes obvious that ITEA兾ILPIAray is independent of
as much as 140 ␮m by using the TEA. For larger the height of the prism. It decreases with increasing
diameters, LPIAray should be applied. That is in angles of refraction, and the smaller the difference of
good agreement with Fig. 13共a兲. n1 and n2 becomes, the larger the deviation from 1 is.
This relation is illustrated in Fig. 14. The quotient
B. Modulation of the Amplitude Owing to Refraction of the intensities ITEA and ILPIAray is expressed for two
Considering the refraction at the optical interface indices of refraction n1. For example, for the glass
yields a varying beam density at zII that corresponds prism 共n1 ⫽ 1.5兲 with an angle of refraction of 20°, the
to a modulation of the amplitude in the output field. intensity of Uout共x, y, zII兲 in the illuminated region of
For the example of the prism in Fig. 9, the width of zII is underestimated by more than 20% by the TEA.
the illuminated part of the output field is reduced Figure 13共b兲 shows the modulation of the intensity
from d to d̃. For special cases of surfaces with ver- of the output field behind the lens that is due to the

Fig. 13. Cross section of the field in the second conversion plane of the aspherical interface of a lens with a N.A. of 0.4, refractive index
1.5, and diameter 500 ␮m, illuminated with a plane wave with wavelength ␭ ⫽ 632.8 nm. The effects that can be observed with LPIAray
are compared with calculations performed with the TEA. 共a兲 OPD, 共b兲 modulation of the intensity of the field owing to the refraction at
the optical interface, and 共c兲 modulation of the intensity of the field according to the Fresnel coefficients. In 共c兲 the dashed and the dotted
curves represent simulations with LPIAray. In 共b兲 and 共c兲 the intensities are scaled with the intensity value obtained with the TEA.

3310 APPLIED OPTICS 兾 Vol. 39, No. 19 兾 1 July 2000

 Fig. 16. Three lenses with the same surface profile apart from the
corresponding phase wrapping. 共a兲 The phase of the modulo 2␲
lens is wrapped such that the phase difference between two zones
is ␭. 共b兲 The modulo 4␲ lens has twice the height of the modulo 2␲
lens. 共c兲 The lens is smooth, that is, the phase is completely
Fig. 15. Transmitted intensity of a prism with a single interface. unwrapped.
The transmission is calculated for TE and TM polarization with a
refractive index of 3.5.
6. Analysis of the Focal Region of a Lens
We analyze the focal region of the interface of a cy-
varying density of the rays in the second conversion lindrical lens with the TEA and LPIAray to enable us
plane. For the TEA the intensity of the output field to discuss further the influence of the effects that
has the same amplitude as the input field. If LPI- were discussed separately in Section 5. The surface
Aray is applied, the intensity is modulated. The profile of the lens was designed with the TEA. That
edges of the lens are shadowed, and the intensity is is, the desired cylindrical phase ␸共x兲 is directly trans-
no longer uniform. ferred to h共x兲 ⬃ ␸共x兲. The height of the surface pro-
file can be expressed as a function of the distance
C. Modulation of Amplitude According to Fresnel’s from the optical axis x:
Formulas
For interfaces with large deflection angles the f ⫺ 冑x 2 ⫹ f 2
h共x兲 ⫽ , (7)
Fresnel coefficients of refraction25 cannot be ne- n1 ⫺ n2
glected. Figure 15 shows the transmitted intensity
of a prism with a refractive index n1 ⫽ 3.5 as a where f is the focal length of the lens and n1 and n2
function of the angle of refraction. The input fields are the refractive indices of the lens and outside the
are linearly polarized in the TE or the TM direction. lens, respectively. To discuss effects that depend on
The solid curve represents the quotient of the trans- the distance between zI and zII, we vary the height of
mitted intensities of TE and TM polarized input fields the lens by wrapping the phase function starting at a
and can be calculated according to modulo 2␲ lens. As shown in Fig. 16共a兲, this is a lens
with a height of ␭兾共n1 ⫺ n2兲, where ␭ is the wave-
length for which the lens is designed, and is usually
ITE called a diffractive lens. Now the height of the lens
⫽ 共cos ␸兲2. (6) is doubled, yielding a modulo 4␲ lens 关Fig. 16共b兲兴, and
ITM
so on, until the phase function is completely un-
wrapped 关Fig. 16共c兲兴. The lenses used for the eval-
The quotient ITE兾ITM is independent of the refractive uations have a N.A. of 0.4, refractive indices of n1 ⫽
index of the prism. It can be seen from Eq. 共6兲 that 1.5 and n2 ⫽ 1.0, and a width of 500 ␮m. A plane
with a growing angle of refraction the contributions of input wave Uin with a unit amplitude in TE polar-
TE and TM polarization in the transmitted field be- ization propagates through the lens by use of either
come increasingly more asymmetric. If, for exam- LPIAray or the TEA. In the homogeneous region
ple, a lens is illuminated with linearly polarized light from zII to the focal region, Uout共x, y, zII兲 is propa-
with uniform intensity, the output field has a rota- gated by application of the Rayleigh–Sommerfeld
tionally asymmetric intensity distribution. wave propagation integral.
This asymmetry can be observed in Fig. 13共c兲. It Figure 17 shows the intensity of the field calculated
shows the modulation of the intensity of the output on the optical axis as a function of distance z from zII.
field of an aspherical lens that is due to the Fresnel The thicker solid curve represents the field propa-
coefficients. With the TEA the output field has the gated from the lens, analyzed with the TEA. As the
same intensity distribution as the input field. If the surface profile of the lens was designed by use of this
Fresnel formulas are considered, the transmitted in- approximation, this curve has the highest intensity.
tensity is decreased. Close to the center of the lens All other curves are normalized to the maximum in-
共small angles of refraction兲 the intensities for TE and tensity of this curve. As the TEA neglects refraction
TM polarization are nearly the same. At the edge of at the optical interface, the results are independent of
the lens 共large angles of refraction兲 the intensities for the height of the lens as long as the phase steps
TE and TM polarization differ by almost 15%. between the zones of the lens are multiples of 2␲.

1 July 2000 兾 Vol. 39, No. 19 兾 APPLIED OPTICS 3311

 gential planes for analysis of the optical effect of the
interface fall into the superclass of the LPIA.
The approximation of the interface with local tan-
gential planes restricts the use of the LPIA to sur-
faces with a moderate curvature. Surfaces with
large curvatures, e.g., vertical steps, are not suited
for the LPIA unless the steps can be neglected be-
cause they are sparsely distributed in the whole in-
terface area.
Depending on the accuracy of the ray-tracing step,
there will be different approximation levels of the
LPIA. The thin-element approximation, the first
approximation level, considers the optical path dif-
ference of the rays and neglects refraction at the op-
tical interface. Amplitude modulation by refraction
Fig. 17. Intensity on the optical axis in the focal region of cylin- and Fresnel’s formulas are not taken into account at
drical lenses analyzed with the TEA as well as with LPIAray. The all. Disregarding these effects yields three impor-
lenses are illuminated with a plane wave with a wavelength of ␭ ⫽ tant sources of errors in the TEA. The change in the
632.8 nm in TE polarization. Intensity is plotted as a function of optical path that is due to refraction is directly pro-
the distance from the second conversion plane of the lens. The portional to the height of the structures and increases
lenses were designed with a TEA with a N.A. of 0.4, refractive with large angles of deflection. The displacement of
index 1.5, and width 500 ␮m.
the rays owing to refraction results in a modulation of
the intensity of the output field. In the case of in-
terfaces with vertical steps, this modulation is called
the shadowing effect, which is independent of the
The thinner solid curve was also calculated by use of
height of the element. Also, the Fresnel coefficients
the TEA. However, in addition, the Fresnel coeffi-
yield a modulation of the intensity of the output field.
cients at the optical interface of the lens are consid-
The ratio of the transmitted intensities of input fields
ered. The maximum intensity of this curve
in TE and TM polarization is solely a function of the
decreases to 91%. All other curves were calculated
angle of refraction. The examples in this paper
by LPIAray with increasingly thick lenses, starting
demonstrate the need for use of the geometric-optical
with a modulo 2␲ lens and going to a modulo 8␲ lens,
version LPIAray instead of the TEA in various situa-
until the phase function was completely unwrapped.
tions that are of basic interest in wave-optical engi-
The focus of the modulo 2␲ lens has a maximum
neering. In this paper we considered plane
intensity of 76%. The difference in intensity com-
illumination waves under normal incidence. For
pared with that of the lens analyzed with the TEA
oblique illumination the effects discussed are even
plus Fresnel’s coefficients can be attributed mainly to
more dominant, as the TEA does not consider the
the shadowing effect. If the thickness of the lens is
direction of input wave Uin at all.
increased to modulo 8␲, the maximum intensity on
We have addressed a definition and a better un-
the optical axis is 63%. As, according to Eq. 共5兲, the
derstanding of the validity of the LPIA in general and
shadowing effect is independent of the height of the
considered two levels of approximation within the
element, the loss of intensity is caused by the change
LPIA model. On that basis and following the nu-
in the optical path that is due to the refraction at the
merical implementation of LPIAray, the next step is
optical interface. According to Eq. 共4兲, change of the
the design of optical elements on the basis of a
optical path is directly proportional to the height of
geometric-optical local plane-interface approach in
the element. The evaluation of the smooth lens
wave-optical engineering.26
yields a maximum intensity of only 19%. In this
case the effect of the change in the optical path be-
comes absolutely prominent. In analyzing the
References
smooth lens, LPIAray and the TEA give completely
different results. 1. J. J. Stamnes, Waves in Focal Regions. Propagation, Diffrac-
tion and Focusing of Light, Sound and Water Waves 共Adam
Hilger, Bristol, UK, 1986兲.
7. Conclusions 2. L. Mandel and E. Wolf, Optical Coherence and Quantum Op-
The geometric-optical version of the local plane- tics 共Cambridge U. Press, Cambridge, 1995兲.
interface approximation 共LPIAray兲 combines the 3. M. Born and E. Wolf, Principles of Optics 共Pergamon, New
wave-optical propagation of an electromagnetic field York, 1991兲.
4. J. W. Goodman, Introduction to Fourier Optics 共McGraw-Hill,
in the homogeneous regions of an optical system with
New York, 1968兲.
a ray-tracing step in the inhomogeneous regions. In 5. E.-B. Kley, F. Thoma, U. Zeitner, L. Wittig, and H. Aagedal,
the vicinity of the intersection of a ray with the opti- “Fabrication of micro-optical surface profiles by using gray
cal interface the interface is approximated as a local scale masks,” in Miniaturized Systems with Micro-Optics and
tangential plane. All approximations that decom- Micromechanics III, R. Goering and M. Motamedi, eds., Proc.
pose a usually curved optical interface into local tan- SPIE 3276, 254 –262 共1997兲.

3312 APPLIED OPTICS 兾 Vol. 39, No. 19 兾 1 July 2000

 6. M. Gale, M. Rossi, J. Pedersen, and H. Schütz, “Fabrication of elements,” Tech. Rep. 914 共Massachusetts Institute of Tech-
continuous-relief micro-optical elements by direct laser writing nology, Cambridge, Mass., 1991兲.
in photoresist,” Opt. Eng. 33, 3556 –3566 共1994兲. 18. E. Noponen, J. Turunen, and A. Vasara, “Electromagnetic the-
7. G. Blough and M. Morris, “Hybrid lenses offer high perfor- ory and design of diffractive-lens arrays,” J. Opt. Soc. Am. A
mance at low cost,” Laser Focus World 31共11兲, 67–74 共1995兲. 10, 434 – 443 共1993兲.
8. M. T. Gale, “Direct writing of continuous-relief micro-optics,” 19. M. Rossi, C. G. Blough, D. H. Raguin, E. K. Popov, and D.
in Micro-Optics, H. P. Herzig, ed. 共Taylor & Francis, London, Maystre, “Diffraction efficiency of high-NA continuous-relief
1997兲, pp. 87–126. diffractive lenses,” in Diffractive Optics and Microoptics, Vol. 5
9. M. B. Fleming and M. C. Hutley, “Blazed diffractive optics,” of 1996 OSA Technical Digest Series 共Optical Society of Amer-
Appl. Opt. 36, 4635– 4643 共1997兲. ica, Washington, D.C., 1996兲, pp. 233–236.
10. D. Daly, R. F. Stevens, M. C. Hutley, and N. Davies, “The 20. J. J. Stamnes and H. Heier, “Scalar and electromagnetic dif-
manufacture of microlenses by melting photoresists,” J. Meas. fraction point-spread functions,” Appl. Opt. 37, 3612–3622
Sci. Technol. 1, 759 –766 共1990兲. 共1998兲.
11. D. W. Sweeney and G. E. Sommargren, “Harmonic diffractive 21. I. N. Bronstein and K. A. Semendjajew, Taschenbuch der
lenses,” Appl. Opt. 34, 2469 –2475 共1995兲. Mathematik 共Verlag Harri Deutsch, Thun und Frankfurt兾
12. T. R. Sales and G. M. Morris, “Diffractive-refractive behavior Main, Germany, 1989兲.
of kinoform lenses,” Appl. Opt. 36, 253–257 共1997兲. 22. D. A. Pommet, M. G. Moharam, and E. B. Grann, “Limits of
13. S. Sinzinger and M. Testorf, “Transition between diffractive scalar diffraction theory for diffractive phase elements,” J. Opt.
and refractive micro-optical components,” Appl. Opt. 34, Soc. Am. A 11, 1827–1834 共1994兲.
5970 –5976 共1995兲. 23. H. Kogelnik and T. Li, “Laser beams and resonators,” Proc.
14. M. Rossi, R. E. Kunz, and H. P. Herzig, “Refractive and dif- IEEE 54, 1312–1329 共1966兲.
fractive properties of planar micro-optical elements,” Appl. 24. M. T. Gale and M. Rossi, “Continuous-relief diffractive lenses
Opt. 34, 5996 – 6007 共1995兲. and microlens arrays,” in Diffractive Optics for Industrial and
15. D. Faklis and G. M. Morris, “Spectral properties of multiorder Commercial Applications, J. Turunen and F. Wyrowski, eds.
diffractive lenses,” Appl. Opt. 34, 2462–2468 共1995兲. 共Akademie Verlag, Berlin, 1997兲, pp. 103–145.
16. P. Beckmann, “Scattering of light by rough surfaces,” in 25. J. D. Jackson, Classical Electrodynamics 共Wiley, New York,
Progress in Optics, E. Wolf, ed. 共North-Holland, Amsterdam, 1962兲.
1967兲, Vol. 6, pp. 53– 69. 26. A. v. Pfeil and F. Wyrowski, “Wave-optical structure design
17. G. J. Swanson, “Binary optics technology: theoretical limits with the local plane-interface approximation,” J. Mod. Opt. 共to
on the diffraction efficiency of multilevel diffractive optical be published兲.

1 July 2000 兾 Vol. 39, No. 19 兾 APPLIED OPTICS 3313