You are on page 1of 12

Flagello et al.

Vol. 13, No. 1 / January 1996 / J. Opt. Soc. Am. A

53

Theory of high-NA imaging in homogeneous thin films


Donis G. Flagello
ASM Lithography, 1110 de Run, 5503 LA Veldhoven, The Netherlands

Tom Milster
Optical Sciences Center, University of Arizona, Tucson, Arizona 85721

Alan E. Rosenbluth
IBM T. J. Watson Research Center, P.O. Box 218, Yorktown Heights, New York 10598
Received May 23, 1994; accepted July 26, 1995; revised manuscript received July 28, 1995
A description is given of a modeling technique that is used to explore three-dimensional image distributions
formed by high numerical aperture sNA . 0.6d lenses in homogeneous, isotropic, linear, and source-free thin
films. The approach is based on a plane-wave decomposition in the exit pupil. Factors that are due to
polarization, aberration, object transmittance, propagation, and phase terms are associated with each planewave component. These are combined with a modified thin-film matrix technique in a derivation of the
total field amplitude at each point in the film by a coherent vector sum over all plane waves. One then
calculates the image distribution by squaring the electric-field amplitude. The model is used to show how
asymmetries present in the polarized image change with the influence of a thin film. Extensions of the model
to magneto-optic thin films are discussed. 1996 Optical Society of America

1.

INTRODUCTION

In the field of microphotolithography the demand for


highly integrated electronic circuits has motivated investigation into better lens resolution. Traditionally, a
highly corrected optical system is used to record images
in a thin photoresist layer s1 mmd that is coated over
electronic thin-film layers and a substrate such as silicon.
The developed photoresist forms a stencil pattern that is
transferred into underlying films. Since resolution improves as the numerical aperture (NA) of the optical system increases, high-NA systems sNA . 0.6d are desired.
In this paper we discuss a theory for estimating the irradiance distribution that results from high-NA imaging
in thin films.
Since photolithographic optical systems are highly corrected, the achievable density and resolution of printed
patterns are limited by diffraction. For low-NA systems
sNA , 0.6d the Fresnel approximation1 can be used to predict accurately the aerial image. The initial work in this
area was done with a crude model of periodic structures,
based on a modulation transfer function approximation.2
The irradiance distribution throughout the photoresist
was assumed to be proportional to the aerial image.
Subsequent enhancements to this simple theory were
based on the Fresnel approximation and the incorporation of partial coherence theory.3 Although the use of the
Fresnel approximation and paraxial propagation of the
aerial image into the photoresist does not adequately describe high-NA imaging, it has proven invaluable for the
study of related photoresist chemical reactions.
There is much literature on the subject of vector highNA diffraction.4 8 These works tend to discuss aerial image formation in terms of the point spread function (PSF),
or Airy disk. Also, the description of a focused electro0740-3232/96/010053-12$06.00

magnetic distribution within a medium has been examined extensively by Ling and Lee,9 based on methods by
Gasper et al.10 However, the description of high-NA volume imaging into layered media with the use of extended
objects, such as is the case in microlithography, is not
discussed at any length. It was not until Yeung that a
high-NA vector model for photoresist emerged.11 Yeung
based his imaging theory on the work of Richards and
Wolf 4 and used a numerical solution to derive the threedimensional (3-D) image as a weighted sum of plane
waves propagating into the photoresist. Yeung simplified his model by assuming that the object is a periodic
grating, thus reducing the problem to a two-dimensional
(2-D) solution for S or P polarization. He treated propagation into a bleaching photoresist as a one-dimensional
(1-D) electromagnetic grating problem, using differential
methods to calculate the fields. After realizing that the
bleaching effects do not substantially alter the image,
Yeung presented a further simplification12 by assuming
homogeneous and linear thin films. This allowed the use
of standard thin-film matrix techniques for calculation of
the electromagnetic fields within the photoresist. Yuan
derived a 1-D vector imaging model with a 2-D waveguide scattering model to simulate high-NA images that
can accommodate nonplanar substrates.13 However, like
Yeungs work, it is limited to polarization in either S or P
and cannot describe the more general 3-D imaging case.
The intent of this work is to present a model for vector
image formation within the volume of the first film of a
general thin-film stack.14 18 The primary film of interest is photoresist, but the theory is sufficiently general to
accommodate other applications. The films are considered to be homogeneous, isotropic, linear, and source free.
The model is based on the generalized Debye approach19
in which the vector image field is characterized as a plane 1996 Optical Society of America

54

J. Opt. Soc. Am. A / Vol. 13, No. 1 / January 1996

wave decomposition of each component of the electric


field. This approach is similar to Yeungs in that the vector diffraction from the exit pupil of the imaging system is
based on the concepts of Richards and Wolf; however, the
analysis here extends the development into a matrix formalism of the imaging within a Cartesian coordinate system. This formalism lends itself to easy implementation
of the numerical computations for 3-D simulation of photolithographic applications. Also, we use assumptions of
isoplanarity for object and image fields, which gives us
a broader scope of validity. This allows us to construct
the imaging from the object using plane-wave decomposition as described by Hopkins.20 Since the first film of the
film stack of interest is assumed to be located at or near
focus, the amplitude and the phase of each plane wave,
weighted by factors that are due to polarization, aberration, and diffraction, can be used as input into a thin-film
matrix routine that calculates the local field within the
film, with the final image distribution being proportional
to jEj2 . The result is a matrix formalism that views highNA imaging within a thin film as the output of a linear
system.
In Sections 2 and 3 we discuss the basic methodology
and the assumptions required for the derivation. In Section 4 we proceed to derive the general scalar form of
high-NA imaging in free space. The transition to vector
theory in thin films is shown in Section 5. The use of the
vector theory for gyro-optic media is outlined in Section 6.
In Section 7 we present the results of the vector theory
for a PSF and an extended object imaged into a film. We
conclude in Section 8 with a summary.

2.

METHODOLOGY

Our model simulates a Kohler optical projection system with polarized coherent illumination. The projector
forms an image of the object in the vicinity of a thin-film
assembly, as shown in Fig. 1. The polarization just before the object is decomposed into components along the
x axis and the y axis. The imaging lens is assumed to
be lossless, with an image-side NA typically $ 0.6 but an
object-side NA , ,0.2. The imaging lens is represented
by an entrance pupil reference sphere whose center of
curvature is at the object and by an exit pupil reference
sphere whose center of curvature is at the geometrical
focus. Aberrations are defined by small deviations from
the exit pupil sphere in terms of wave-front phase errors.
The imaging is stationary (or shift invariant),21,22 which
implies that an x y translation of the object field results
in only a likewise translation of the image. This assumption results in isoplanatic conjugate fields, such that the
aberrations across the fields are constant. Isoplanatic
imaging imposes strong constraints on the image-forming
properties of the system, allowing one to model it without
considering in detail the propagation of light between the
entrance and exit pupil surfaces. Hopkins used this assumption to show that the entrance pupil must exist in
the Fraunhofer region of the object, whereas the image
exists in the Fraunhofer region of the exit pupil.
The propagation direction of a general plane wave
through the optical system is described by a geometrical
ray with a direction given by the propagation vector k, as
shown in Fig. 2. The polarization direction of the elec-

Flagello et al.

tric field remains perpendicular to the geometrical ray


and is represented by a polarization vector. The reduction ratio is assumed large enough that the diffraction
from the object to the entrance pupil can be modeled
as a scalar process. The polarization vectors for objectside propagation rays are then approximately parallel
to the object plane. Calculation of the exact polarization mapping from the entrance pupil to the exit pupil
by means of polarization ray tracing requires an exact
lens prescription23,24 ; however, in the absence of a specific lens prescription, further simplifying assumptions
can be made that are suitable for investigating general
features of high-NA imaging. If the isoplanatic patch is
large enough to include the axial field position, all rays
can be assumed meridional without loss of generality. In
such circumstances the angle that the polarization vector
makes with the meridional plane (the plane containing
the propagation vector and the z axis) will remain unchanged during propagation through the lens if polarizing
effects from the lens coatings and the like are assumed
negligible. A general meridional plane is illustrated in
Fig. 3. The S-polarization state is assumed perpendicular to the meridional plane, and the P-polarization state is
assumed parallel to this plane. The polarization vector
of each plane wave has global Cartesian components that
are decomposed into local projections on the S and P axes.
Debye has shown with a stationary phase argument
that one can decompose the electric-field amplitude converging on the geometrical focus into a simple spectrum of
plane waves.25 Subsequently, Luneburg26 showed that a
separate expansion can be made for each Cartesian component of the electric field, E . For this paper we assume that the cone of light emerging from the exit pupil
of an optical system defines the angular subtense of the
spectrum.
The conditions for the validity of the Debye approach27
require that the image region be small compared with the
pupil diameter and that the Fresnel number be large, i.e.,

Fig. 1. Kohler illuminated projection system focusing to a


thin-film stack.

Fig. 2. Propagation vector k .

Flagello et al.

Fig. 3.

Vol. 13, No. 1 / January 1996 / J. Opt. Soc. Am. A

Meridional plane through a high-NA optical system.

p
the NA must be large compared with lyr 0 , where r 0 is
the distance from the exit pupil to the image. In this
work a local field size less than 10 mm about the axis is
used,
with r 0 . 500 mm and l 0.442 mm.
p This gives
p
0
lyr , 0.03, which satisfies the condition lyr 0 , NA.
We calculate the E fields in the volume of the film by
summing forward and backward plane waves within the
film and satisfying boundary-value conditions at the interfaces. The E fields are derived through a modification
of the thin-film matrix technique outlined by Macleod,28
which defines additional reflection and transmission coefficients with respect to the interface between the first
film and the second film. Since the intent here is to calculate the total vector fields inside a film, the solution of
the axial szd electric-field component is required, in addition to the transverse components on which traditional
thin-film matrices operate. Finally, the image distribution within the first film of a thin-film stack is given by
Joules heat term, Q, defined by Poyntings theorem and
is proportional to jEj2 for an isotropic medium.

3. PRELIMINARY DEFINITIONS
AND NOTATIONS
A plane wave with a propagation vector k at a position
r is given by
U sx, y, zd U0 sx, y, zdexpsik ? rd
U0 sx, y, zdexpf2i2pNsax 1 by 1 gzdg , (1)
where
r l0 sxx 1 yy 1 zzd ,
!

ky
kz
2pN kx
x1
y 1
z
k
l0
k
k
k

2pN
fscos f sin udx 1 ssin f sin udy 1 scos udzg
l0

2pN
sax 1 by 1 gzd .

l0

(2)

55

Traditionally, the magnitude of the time-averaged


Poynting vector has been used to calculate the image
irradiance; however, it is more appropriate to use the
absorbed energy, i.e., local Joule heating per unit volume,
in calculations of the image field within a thin film of
photoresist. We estimate the Joule heating by considering the conservation of energy using Poyntings theorem,
where the time-averaged divergence of the Poynting vector is shown to represent the power absorbed within the
volume. In this paper we define
Q jk= ? sE 3 Hdlj

1
2

sjEj2 k0 Y nkjEj2 ,

(3)

where s is the conductivity, Y is the free-space admittance, and k0 2pyl0 . If we assume that only Ohmic
losses are being considered, Q represents the form of the
recorded image distribution within an absorbing medium
with units of power per volume. It has been interpreted
by Stratton29 as the power dissipated by thermochemical
activity. In general, for photosensitive films, a fraction
of the energy absorbed is used in a photochemical reaction
with the remainder being dissipated as heat. Therefore
the recorded image within a film is proportional to jEj2 .

4. SCALAR HIGH-NA IMAGING


IN FREE SPACE
This section describes general image formation through
an optical system in the absence of a thin-film stack,
where the object and image spaces are defined in air.
The derivation is based on the scalar imaging concepts
introduced by Goodman,30 where the image distribution is
formed by propagation of plane waves sequentially from
object to entrance pupil, entrance pupil to exit pupil, and
exit pupil to image.
Harvey31 has shown that a direction cosine spectrum
of plane waves can be used to describe the propagation
of a scalar field in free space. His geometry consisted of
a diffracting aperture illuminated by a converging beam.
The diffracted electric-field amplitude is observed on a
hemisphere in the far field, i.e., the Fraunhofer region.
Harveys results may be applied to the propagation between the object and the entrance pupil reference surface. Figure 4 displays the diffraction geometry, where
r is the distance from the object to the entrance pupil.
The entrance pupil surface is described in direction cosine space, where locations on the surface are referenced
by coordinates sa, bd. The limit of the pupil is a circular
ring defined by the NA, that is,
p
NA sin umax amax2 1 bmax 2 ,

Here and throughout this paper, scalar distances x, y,


z, and r are normalized to the vacuum wavelength l0 .
U0 sx, y, zd is defined as the complex amplitude [the harmonic time dependence of the form expsivtd is implicitly
assumed], and N is the complex index of refraction given
by N n 2 ik. k defines the length of the propagation
vector, and a, b, and g are the direction cosines, related
by
a2 1 b2 1 g2 1 .

Fig. 4. Diffraction geometry from object to entrance pupil.

56

J. Opt. Soc. Am. A / Vol. 13, No. 1 / January 1996

Flagello et al.

dS 0 r 02 dV 0 r 02

Fig. 5.

Circular pupil defined in direction cosine space.

where umax is the maximum marginal ray angle defined


in object space. The direction cosine coordinates in the
entrance pupil reference surface are shown in Fig. 5.
The diffraction to the entrance pupil, based on the work
of Hopkins, uses the Huygens Fresnel principle. It is
written with the use of a Fourier transform relationship
and is given by

exps2i2prd
F hOsx, ydj ,
r

hNb h0 N 0 b 0 ,

(6)

which is an important relationship that will be used in


the development below.
Again, from the work of Hopkins, we calculate the
image field by recognizing that it is proportional to the
inverse Fourier transform of the spherical exit pupil surface S and thus is a weighted integration of plane waves.
The solution is consistent with Wolf s generalized Debye
approximation and is given by

h0
a
b
0 0.
h
a
b

The field distribution in the exit pupil is derived from


scaling Eq. (5) and is written as
1
0
E 0 sa 0 , b 0 d 2 Esma
, mb 0 d
m

(5)

where a transmission function T sa, bd specifies the size


and the shape of the entrance pupil and hence umax .
T sa, bd has a value of 1 on the entrance pupil and 0 otherwise. Equations (4) and (5) are rearranged to give

r Esa,
bd ig exps2i2prdT sa, bdF hOsx, ydj ,

h ,
h

where h and h0 refer to normalized object and image size.


From the Abbe sine condition,

(4)

where the tilde is used to define a transformed object.


Without any loss of generality the object is placed at
z 0. The distribution of the electric-field amplitude on
the entrance pupil is represented by

Esa,
bd Osa,
bdT sa, bd ,

and since N N 0 1, the transverse magnification is


given as

3 expfi2psax 1 bydgdxdy
ig

and the primed notation is used to represent the imageside geometry, as shown in Fig. 6. r 0 has been written inside the transform for reasons that will be obvious shortly.
E 0 sa 0 , b 0 d is the scalar complex amplitude distribution in
the exit pupil. Note that the exponential term in z0 represents an image focus term. The geometrical focus, located at z0 0, is at the vertex of the solid angle subtended by the exit pupil.
A simple magnification scaling is assumed between entrance and exit pupils. Any aberrations within the lens
are treated with an aberration term that modifies the
phase of the wave front at the exit pupil reference surface. The transverse magnification is defined as

hNa h0 N 0 a 0 ,

exps2i2prd Z ` Z `

Osa,
bd ig
Osx, yd
r
2`
2`

da 0 db 0
g0

1
Osma 0 , mb 0 dT sma 0 , mb 0 d .
m2

The relationship between the magnitudes of the entrance pupil fields and the exit pupil fields must be taken
into account. Since there is a physical difference in size,
i.e.,
T 0 sa 0 , b 0 d T sma 0 , mb 0 d ,

(9)

and T 0 sa 0 , b 0 d is the exit pupil transmission function, conservation of energy must be ensured. Figure 7 illustrates
the differential areas on the entrance and exit pupil sur-

E 0 sx0 , y 0 ; z0 d
expsi2pr 0 d ZZ 0 0 0
E sa , b dexps2ik0 ? r0 ddS 0
2i
r0
S0
8
9
< 0 0 0 0
0 0 =
0
21 r E sa , b dexps2i2pg z d
, (7)
2i expsi2pr dF
:
;
g0
where an element of wave front dS 0 is defined by

(8)

Fig. 6.

Diffraction geometry from exit pupil to image focus.

Flagello et al.

Vol. 13, No. 1 / January 1996 / J. Opt. Soc. Am. A

57

5. VECTOR HIGH-NA IMAGING


IN THIN FILMS
The transition to a vector field is first accomplished by
establishing the initial state of polarization that illuminates the entrance pupil. A polarization state matrix Mi

is defined such that multiplication with Osa,


bd gives the
description of polarization in the entrance pupil, i.e.,

Osa,
bd Osa,
bdMi ,
Fig. 7.

Mapping geometry between entrance and exit pupils.

faces. An energy balance yields

jEsa,
bdj2 da jE 0 sa 0 , b 0 dj2 da0 ,
where the differential areas on each pupil are given by
da r 2 dV r 2

dadb ,
g

da0 r 02 dV 0 r 02

For example, x-polarized illumination results in


!
!
O
1
x

Osa,
bd
Osa,
bd
Osa,
bdMi .
0
0
Similarly,
0
0

, mb 0 d Osma
, mb 0 dMi .
Osma

da db .
g0

The electric fields must satisfy the conservation equation


s
p
g 0 dadb
0 0
0
0

p
r jE sa , b dj rjEsa, bdj
g da 0 db 0
s
g0 .

m
(10)
rjEsa,
bdj
g
As stated above, lens aberrations are given by a wavefront phase error W sa 0 , b d and are represented by a scalar
term, expf2i2pW sa 0 , b 0 dg. In the absence of any film
stack the scalar electric-field distribution, given by the
combination of Eqs. (7), (8), and (10), is
E 0 sx0 , y 0 ; z0 d
2i expsi2pr 0 d
8
9
<r 0 E 0 sa 0 , b 0 dexps2i2pg 0 z0 dexpf2i2pW sa 0 , b 0 dg=
3 F 21
:
;
g0
(
expsi2pr 0 d 21
0

2i
F
, mb 0 dexps2i2pg 0 z0 d
r Esma
m
)
1
0
0
3 expf2i2pW sa , b dg p 0 .
(11)
gg
Substitution of Eq. (6) and the application of Eq. (8) yield
E 0 sx0 , y 0 ; z0 d

where Osa,
bd and Mi are 2 3 1 matrices. Mi has elements given by Sl that may be complex, i.e.,
!
Sx .
Mi
Sy

expfi2psr 0 2 rdg
m
21
0 , b 0 ; z0 dj ,
3 F hOsma 0 , mb 0 dCsa

A vector analogy to Eq. (12) is derived by first weight0

0 , b 0 ; z0 d,
ing the plane-wave amplitudes, Osma
, mb 0 dCsa
by the polarization amplitudes for each Cartesian component of the electric field. Figure 8 illustrates a propagation vector, k0 , with a direction sa 0 , b 0 d emerging from the
exit pupil. The polarization vector remains at a constant
orientation with respect to k0 and the meridional plane
defined by the plane of incidence with the film stack, and
the component amplitudes are projected onto the local S
or P coordinate system. The component amplitudes have
been defined by Mansuripur 8 in terms of the propagation
vector direction and the initial x or y polarizations from
P sa 0 , b 0 d
the object. It is given here as a 5 3 2 matrix M
with matrix elements Plmn , where l gives the initial object
polarization, m refers to the global sx, y, zd coordinates,
and n is the local S or P coordinate. Hence
3
2
2a 0 b 0
b 02
7
6
6 1 2 g 02 1 2 g 02 7
7
3 6
2
7
6 0 02
6 ga
PxxS PyxS
a0b0 g0 7
7
6
7 6
6
6PxxP PyxP 7 6 1 2 g 02 1 2 g 02 7
7
7 6
6
7
7
.
P sa 0 , b 0 d 6
6PxyS PyyS 7 6
2a 0 b 0
a 02 7
M
7
6
7 6
6
6PxyP PyyP 7 6 1 2 g 02 1 2 g 02 7
7
5 6
4
7
6 0 0 0
0 02 7
PxzP PyzP
gb
7
6abg
7
6
6 1 2 g 02 1 2 g 02 7
5
4
2a 0
2b 0
(14)

(12)

0 , b 0 ; z0 d is a function that contains only the


where Csa
scalar elements of the optical system and is defined by
0 , b 0 ; z0 d T 0 sa 0 , b 0 dexps2i2pg 0 z0 d
Csa
r
g .
3 expf2i2pW sa 0 , b 0 dg
g0

(13)

0 , b 0 ; z0 d represents the complex ampli, mb 0 dCsa


Osma
tude distribution of plane waves for a given z0 . The image irradiance is proportional to jE 0 sx0 , y 0 ; z0 dj2 .

Fig. 8.
pupil.

Rotation of the polarization vector emerging from exit

58

J. Opt. Soc. Am. A / Vol. 13, No. 1 / January 1996

Flagello et al.

By simple orthogonal decomposition any polarization


state can be defined by Eq. (14). As an example, consider a propagation vector direction given by sa 0 , b 0 , g 0 d
s0.5, 0.5, 0.707d. The corresponding matrix is
3
0.5
20.5
7
6
0.35 7
6 0.35
7
6
.
P 6
M
0.5 7
7
620.5
7
6
4 0.35
0.35 5
20.5
20.5
2

the substitution
Fsz0 d 2pN1 g1 sd1 2 z0 1 z0 d
2psd1 2 z0 1 z0 dfN1 2 2 sin2 scos21 g 0 dg1/2

iI sa 0 , b 0 ; z0 d M
P sa 0 , b 0 dU
sa 0 , b 0 ; z0 d
A
0

P sa 0 , b 0 dOsma
, mb 0 dCsa 0, b 0 ; z0 d ,
M

(15)
iI is a 5 3 1 matrix with elements given by
where A

sAiI dmn . The distance z0 is an offset term that determines the initial phase for each plane wave incident on
interface I. Variation of z0 is equivalent to moving the
top film surface in and out of focus.
The derivation of the electric-field image amplitude
within the first layer of a thin-film stack is based on
the thin-film matrix techniques presented by Macleod.28
Figure 9 shows the geometry used for the calculation.
The vector field for any point within the first film is found
by summation of the downward and upward plane waves
for each S and P contribution, n, of the Cartesian component, m. The derivation presented here departs from the
typical thin-film method by deriving this field in terms
of the field at interface II, the bottom of the film, and
then relating the result to the incident field at interface I
through the matrix formalism. If the first film has a normalized thickness of d1 , the incident and reflected fields
at interface II are
A 1 sa1 , b1 d AiII expfi2pN1 g1 sd1 2 z0 1 z0 dg
(16)

A 2 sa1 , b1 d ArII expf2i2pN1 g1 sd1 2 z0 1 z0 dg


3 expf2i2pN1 sa1 x0 1 b1 y 0 dg ,

(18)

yields the total field in the first film as

The resultant amplitude projections onto the x, y, z axis


are simply the sum of the S and P contributions. If
the initial object polarization is along the x axis, the
amplitudes are Pxx 0.854, Pxy 20.147, and Pxz 20.5.
The complex amplitudes of the plane-wave components
located at the top surface (interface I) of a thin-film stack
at z0 z0 are now written in matrix form as

3 expf2i2pN1 sa1 x0 1 b1 y 0 dg ,

the incident medium with


# 1/2
"
sin2 scos21 g 0 d
,
gj 1 2
Nj 2

(17)

A1 sa 0 , b 0 ; z0 d expf2i2psa0 x0 1 b 0 y 0 dg
3 fAiII exp iF 1 ArII exps2iFdg ,

(19)

where it is understood that F is a function of z0 .


The following derivation shows that AiII and ArII in
Eq. (19) are related to the incident field through the transverse sx, yd and axial szd transmission and reflection coefficients. Since the fundamentals of thin-film matrix
techniques are well known, only the departures from the
standard formalism will be presented. It is assumed that
the film in question is the first film on an arbitrary film
stack composed of homogeneous, linear films. A matrix
can be defined for each film by
#
"
cos dj
i sin dj yhj ,
Mj
(20)
ihj sin dj
cos dj
where hj is defined as the tilted optical admittance for S
and P polarization in free-space units and dj is the phase
for a normalized film thickness d1 . Hence
(
S polarization
Nj gj
,
hj
Nj ygj P polarization
dj 2pNj gj dj .
The characteristic matrix for a film assembly with q layers is defined by
1
! 0 q
!
B
BY
C 1 ,
@ Mj A
(21)
C
hm
j1
where hm is the substrate admittance. A subassembly
characteristic matrix is defined without the first film as
1 !
! 0 q
Y
Bs
B
C 1 .
@ Mj A
(22)
hm
Cs
j 2
With reference to Fig. 9, transverse and axial reflection and transmission coefficients are defined for the full

where A 1 is the downward field and A 2 is the upward


field. It is understood in Eqs. (16) and (17) that the complex amplitudes AiII and ArII represent mn matrix elements. The numerical subscripts denote the underlying
media of interest in image space, and the quantity z0 2 z0
ranges from 0 at interface I to d1 at interface II.
The following relationships are used from Snells law
in direction cosine notation (assume an incident medium
of N 0 1):
a 0 Nj aj ,

b 0 Nj bj .

Since for any film j the z-direction cosine is referenced to

Fig. 9. Thin-film stack with incident plane waves and electricfield amplitudes.

Flagello et al.

Vol. 13, No. 1 / January 1996 / J. Opt. Soc. Am. A

Table 1. Reflection and Transmission Coefficients


Component

Full Assembly
ArI
AiI
Atm
t
AiI
ArIz
rz
AiIz
Atmz
tz
AiIz
r

Transverse reflection
Transverse transmission
Axial reflection
Axial transmission

Subassembly
ArII
AiII
Atm

AiII
ArIIz

AiIIz
Atmz

AiIIz

rII
tII
rIIz
tIIz

film assembly and subassembly in Table 1. The original


assumption of film homogeneity and linearity has been invoked to allow the equality of x and y coefficients. Since
the transverse components of E and H are required to
be continuous across all interfaces, the coefficients can be
expanded in B and C terms for the total assembly and in
Bs and Cs terms for interface II as
2h 0
,
Bh 0 1 C
2h1
,
tII
Bs h1 1 Cs
t

Bh 0 2 C ,
Bh 0 1 C
Bs h1 2 Cs ,
rII
Bs h1 1 Cs
r

59

total vector electric image field within the film is calculated by the summation of all plane waves. Therefore
1 sa 0 , b 0 ; z0 dj
E1 sx0 , y 0 ; z0 d c0 F 21 hA
iI sa 0 , b 0 ; z0 dj
F sa 0 , b 0 ; z0 dA
c0 F 21 hM
F sa 0 , b 0 ; z0 dM
P sa 0 , b 0 d
c0 F 21 hM
0 , b 0 dCsa
0 , b 0 ; z0 dj .
3 Osa

(30)

Equation (30) suggests that image formation within a thin


film can be regarded as the output of a linear system.
The polarization, film, and scalar lens terms behave as
transfer functions. The film stack in this context must
then be treated as an integral part of the image formation
1 sa 0 , b 0 ; z0 d besystem. If the object is a delta function, A
comes the vector version of the system transfer function,
and the vector point spread function (PSF) is given by its
Fourier transform.
Alternatively, E1 sx0 , y 0 ; z0 d can be written in terms of a
convolution of the transformed elements:

(23)

E1 sx0 , y 0 ; z0 d c0 MF sx0 , y 0 ; z0 d MP sx0 , y 0 d


!

x0 , y 0
Csx0 , y 0 ; z0 d .
O
m m

(24)

(31)

where h 0 is the incident admittance. The axial szd coefficients are not continuous across the boundaries. We derive them here by requiring that propagating plane waves
in the first film be solutions to Maxwells equations in
a source-free medium. In particular, the divergence of
the field must be 0 in the incident medium and the film;
hence, for a general plane wave,

This formalism further implies that the image shape is


a function of the film as well as of the polarization and
scalar terms.
Finally, the image distribution within the first film is
given by

= ? E k ? E 0.

The advantage of Eq. (30) is that the field within the


film is represented in a concise format that lends itself to computer programming. The interaction of the
downward- and upward-traveling waves in the film is succinctly represented by one film function per component in
F sa 0 , b 0 ; z0 d.
the matrix M

Solving this equation for the electric field in the incident


and reflected media results in the following relationships
between the transverse and axial reflection and transmission coefficients:
tz
t N 0g0 ,

rIIz
tII N1 g1

rIIz 2rII .

(25)

Equations (23) (25) are now used to express general


plane-wave amplitudes in the film in matrix notation as
1 sa 0 , b 0 ; z0 d M
iI sa 0 , b 0 ; z0 d ,
F sa 0 , b 0 ; z0 dA
A

(26)

F sa 0 , b 0 ; z0 d is called the film function matrix. It


where M
has dimensions 3 3 5 with elements given by Fn such that
2
3
FS FP 0
0
0
6
7
F sa 0 , b 0 ; z0 d 4 0
0 FS FP
0 5,
s27d
M
0
0
0
0 FzP
!
t
Fn
fexpsiFd 1 srII dn exps2iFdg , s28d
tII n
!
N 0g0
t
fexpsiFd 2 srII dP exps2iFdg .
FzP
N1 g1 tII P
(29)
1 sa 0 , b 0 ; z0 d has dimenThe matrix amplitude function A
sions 3 3 1 pertaining to the Cartesian coordinates. The

Qsx0 , y 0 ; z0 d k0 Y n1 k1 jE1 sx0 , y 0 ; z0 dj2 .

6.

(32)

GYRO-OPTIC MEDIA

Unfortunately, Eq. (30) is not easily used with gyro-optic


media, for which it is traditional to use 2 3 2 matrices
with strictly S- and P-polarization decomposition. In this
section we derive an alternative form of Eq. (30) by maintaining the local S and P coordinate system through the
film, then using an xyz projection matrix to map to a
global coordinate system.
As explained in Section 2, the polarization vector
remains perpendicular to the propagation vector with
components parallel (P) and perpendicular (S) to the
meridional plane. It is easily shown that the amplitudes
of the S and P components in the entrance pupil are
given in matrix form by
"
# !
sin f 2 cos f Ox .
MSP O
(33)
Oy
cos f
sin f
Since the direction cosine relationships give
cos f p

a0
,
1 2 g 02

sin f p

b0
,
1 2 g 02

60

J. Opt. Soc. Am. A / Vol. 13, No. 1 / January 1996

Flagello et al.

e
6
e 4iqe
0

2iqe
e
0

3
0
7
05.
e

c relates to the Faraday rotation in the material, and


q 0 for pure dielectrics.
Likewise, the upward field is
2 sa 0 , b 0 d M2 sa 0 , b 0 ; z0 dMSP sa 0 , b 0 dOsa 0 , b 0 d
A
F
3 Csa 0 , b 0 ; z0 d ,

(37)

where
"
0
0
0
M2
F sa , b ; z d exps2iFd

cos c
sin c

2 !
6 t rIIS
6 t
6 II S
36
6
4
0

2 sin c
cos c
0
!

trII
tII

#
3

7
7
7.
7
7
rIIP 5

(38)

Fig. 10. PSF for NA0 0.95 given by values of normalized jEj2 :
(a) isoimage contours, ( b) profiles along x and y.

Table 2.

Parameters for PSF Simulation

Film Assembly

Optics

N0 1
N1 1.656 2 i0.004, d1 1 mm

NA0 0.95, l 0.442 mm


NA ,, 1
x-polarized

Nm 1.656 2 i0.004

MSP is written as
2
MSP

b0
6p
6 1 2 g 02
6
6
4 p a0
1 2 g 02

Osx, yd dsx, yd
W 0, z0 0

2a 0
1 2 g 02
b0
p
1 2 g 02
p

3
7
7
7.
7
5

(34)

Fig. 11. Q distributions of PSF at interface I with NA0 0.95:


total Q distribution on top surface with x, y, and z components.

The downward field for each local S and P wave through


the film is now written as
1 sa 0 , b 0 d M1 sa 0 , b 0 ; z0 dMSP sa 0 , b 0 dOsa 0 , b 0 d
A
F
3 Csa 0 , b 0 ; z0 d ,

(35)

1 sa 0 , b 0 d is a 2 3 1 matrix corresponding to the


where A
0
0
0
S and P fields and M1
F sa , b ; z d is defined as
"
#
cos c
sin c
1
0
0
0
MF sa , b ; z d expsiFd
2 sin c cos c
2 !
3
t
6
0 7
6 t
7
6 II S ! 7 ,
7
(36)
36
6
7
t
4 0
5
tII P
where c pNqsd1 2 z0 1 z0 d and q is the off-diagonal
permittivity matrix coefficient given by32

Fig. 12. Normalized PSF distribution in film: (a) meridional


slices, ( b) comparison of x and y polarizations at interface I and
interface II.

Flagello et al.

Vol. 13, No. 1 / January 1996 / J. Opt. Soc. Am. A

61

that the total field through the film is now written as


0
0 1
E1 sx0 , y 0 ; z0 d c0 F 21 hM1
xyz sa , b dA sa1 , b1 d
2
2 sa1 , b1 dj .
sa 0 , b 0 dA
1 Mxyz

7.

(40)

RESULTS

A. Point Spread Function


It is useful to examine first the normalized image of a
delta function object that produces a constant electricfield distribution across the entrance pupil reference
surface. The resultant image in the film, Qsx0 , y 0 ; z0 d,
becomes the point spread response function (PSF). Sub 1 sa 0 , b 0 ; z0 d the system transfer funcsequently, we call A
tion within the exit pupil domain. Note that the film is
incorporated in this function.
Table 3.

Parameters for Tribar Simulation

Film Assembly
1
N1 1.656 2 i0.004, d1 1 mm
N0

Optics
NA0

0.95, l 0.442 mm
NA 0.048

Osx, yd tribar, 3 bars


of unit amplitude on
an opaque field
Nm 1.656 2 i0.004

W 0, z0 0, m 0.05

Fig. 13. Magnitudes of normalized system transfer function


1 sa 0 , b 0 ; z0 0d for x, y, and z components given in exit pupil
A
coordinates for NA0 0.95. The magnitudes are normalized to
the maximum x-component value.

After they emerge from the exit pupil, it can be shown


that the projections of the S and P components onto the
x, y, and z axes are given by a matrix Mxyz in terms of
the direction cosines of the propagation vector:
2
3
b0
a0g0
p
p
6
7
6 1 2 g 02
7
1 2 g 02
6
7
6 2a 0
7
0 0
b
g
6
7
6
p
q
6
7,
Mxyz 6
(39)
02
7
1 2 g 02
6 12g
7
6
7
6
7
4
5
N 0 g0 p
0
6
1 2 g 02
N1 g1
where the rows represent the x, y, and z components and
the columns represent the S and P components. The necessary correction to the z component has been incorporated into downward and upward projection matrices such

Fig. 14. (a) Tribar object with scaled image dimensions,


( b) entrance pupil electric-field distribution normalized to unit
magnitude.

62

J. Opt. Soc. Am. A / Vol. 13, No. 1 / January 1996

Flagello et al.

that the x slice is slightly wider with fewer interference


effects, such as ringing, than the y slice. The differences
are more apparent 1 mm from the focal plane.
Figure 13 shows the magnitude of the normalized sys 1 sa 0 , b 0 ; z0 0d, at interface I.
tem transfer function, A
The maximum value of the x component is used for normalization. Since the initial polarization is x polarized,
the x component transfers the bulk of the power. The
obliquity terms tend to counter the polarization term, producing a top hat shape. The y and z transfer functions
are mostly dominated by the polarization terms. The
quadrupole symmetry for y does not give a significant
magnitude as compared with that for z, where the greatest effect is at the edge of the pupil and along the direction of polarization. The interaction of the film reduces
the overall magnitude of the z transfer function relative

1 sa 0 , b 0 ; z 0 0d components, with the


Fig. 15. Magnitude of A
object, scalar lens, polarization, and film terms at interface I for
a tribar object and NA0 0.95.

Figure 10 illustrates the effect of the film stack by comparing aerial PSF images (no film stack) to the PSF at
interface I of a thin-film stack, where jEj2 is normalized to
its maximum value. The parameters set for this example
are given in Table 2. The index of the first thin film represents a mildly absorbing photoresist. The substrate is
silicon. Clearly, interaction with the film reduces asymmetry in the PSF. This is caused by the reduction in the
cone angle of the transmitted beam through Snells law
as well as the angular dependence of reflectivity. The
image width decreases along the direction parallel to the
polarization and slightly increases along the perpendicular direction.
The image at interface I is further illustrated in Fig. 11,
where the PSF is shown for each xyz component and
the image, which is the component sum. The x component contains the most power and is similar to a scalar
response. The z component contains substantial power
that is distributed along the x axis. It is the primary
contributor to image asymmetry. The y component has
an interesting quadrupole symmetry but contains little
power and has an insignificant effect on the overall image. We further analyze PSF film images by considering
image contours in the film shown in Fig. 12(a) for meridional slices along the x and y axes. A slight image asymmetry exists along the film depth. Figure 12( b) shows

Fig. 16. Simulation of Q distribution of tribar object at interface I with NA0 0.95: Q distribution at top surface with x, y,
and z components.

Fig. 17. Q simulation of tribar image with NA0 0.95 and


z0 0: (a) x z meridional plane, ( b) profiles comparing x and
y polarizations at interface I and interface II.

Flagello et al.

Vol. 13, No. 1 / January 1996 / J. Opt. Soc. Am. A

63

High-NA effects are clearer at the bottom surface. The


phenomenon of spurious resolution is evidenced by the
three maxima at interface I becoming four maxima at
interface II. Although the differences between polarizations appear small at the top surface, they are much
larger at the bottom. This increase of the polarization
differences is again seen in Fig. 18, with z0 0.45 mm.
The focus offset is chosen to reproduce the tribar image
at interface II. The respective profile difference at the
bottom of the film is much larger than that in Fig. 17.

8.

Fig. 18. Q simulation of tribar image with NA0 0.95 and


z0 45 mm: (a) x z meridional plane, ( b) profiles comparing
x and y polarizations at interface I and interface II.

to x, thus reducing the image asymmetry of the PSF as


compared with the aerial image.
B. Extended Object
This study simulates the image of an object with finite
dimensions using a system magnification of 0.05. The
object-side NA is 0.048, so that g 1 with an image-side
NA0 0.95. Other parameters are given in Table 3. A
clear tribar object is used with scaled image dimensions of
0.25 mm lines and 0.25 mm spaces parallel to the x axis,
as shown in Fig. 14(a). The tribar object has clear bars
with unit amplitude on an opaque background field. The
normalized entrance pupil distribution is presented in
Fig. 14( b). Since there is substantial magnitude at the
edge of the pupil, the object dimensions are approximately
at the limit of resolution.
1 sa 0 , b 0 ; z0 0d components for
The magnitudes of the A
an initial x polarization are shown in Fig. 15, where the
functions are normalized to the maximum magnitude of
the x component. These functions now include the amplitude contributions of the object. Most of the energy is
contained in the x component. The y and z components
have a small amount of energy concentrated at the extremes of the pupil. The z component has a maximum
of ,30% compared with that of the x component, whereas
the y component has a maximum of ,2%. The resulting
Q distributions just after interface I are shown in Fig. 16.
Most of the energy is in the x component. The maximum
value of the z component is approximately 3.7% of the
maximum image magnitude. The contribution of the y
component is negligible, at approximately 0.01% of the
maximum image magnitude.
The asymmetry that is due to polarization is shown
in Fig. 17(a) as iso-Q contours, where the x z meridional plane is shown for initial x and y polarizations
with z0 0. A comparison of the two polarizations is
shown for two film depths in Fig. 17( b). The x-polarized
slice has slightly more power at the top surface for x
60.25 mm and is broader than the y-polarized slice.

SUMMARY

This work presents a vector description of high-NA imaging in homogeneous thin films. The resultant derivation
gives a form for the image distribution within the volume of the first film of a thin-film stack. It is based on a
generalized Debye plane-wave decomposition of the radiation that propagates from the pupil. Each plane wave
is weighted by polarization, aberration, input amplitude,
and phase terms. We combine this decomposition with
a thin-film matrix technique to derive the electric fields
within the film.
The model is limited to imaging systems that have
isoplanatic object and image fields. In general, this requires that the optical system be well corrected, i.e., the
wave-front aberrations are below 0.25l. This requirement further implies that the electric-field polarization
is not substantially altered upon propagation through
the optical elements of the imaging lens. The polarization maintains a constant orientation with respect to the
propagation vector and the meridional plane.
We also present a vector formalism of image formation
as a linear, shift-invariant system. This gives rise to a
vector representation of transfer functions that are due to
the lens, the polarization, and the thin-film stack. We
further show a method to incorporate gyro-optic media
into the imaging.
Finally, examples were presented showing asymmetries
that are due to vector imaging components. The model
also accurately predicts classic spurious resolution as the
focal depth is increased.

REFERENCES
1. J. W. Goodman, Introduction to Fourier Optics (McGrawHill, San Francisco, 1968).
2. F. H. Dill, A. R. Neureuther, J. A. Tuttle, and E. J. Walker,
Modeling projection printing of positive photoresist, IEEE
Trans. Electron Devices ED-7, 456 464 (1975).
3. R. S. Herchel, Partial coherence in projection printing, in
Developments in Semiconductor Microlithography III, D. R.
Ciarlo, J. Dey, K. Hoeppner, and R. L. Ruddell, eds., Proc.
Soc. Photo-Opt. Instrum. Eng. 135, 24 29 (1978).
4. B. Richards and E. Wolf, Electromagnetic diffraction in
optical systems. II. Structure of the image field in an
aplanatic system, Proc. R. Soc. London Ser. A 253, 358 379
(1959).
5. R. Kant, A general numerical solution of vector diffraction
for aplanatic systems, Tech. Rep. TR02.1713.B (IBM, San
Jose, Calif., 1991).
6. T. D. Visser and S. H. Wiersma, Spherical aberration and
the electromagnetic field in high aperture systems, J. Opt.
Soc. Am. A 8, 1404 1410 (1991).
7. J. J. Stamnes, Waves in Focal Regions (Hilger, Bristol, UK,
1986).

64

J. Opt. Soc. Am. A / Vol. 13, No. 1 / January 1996

8. M. Mansuripur, Distribution of light at and near the focus


of high numerical aperture objectives, J. Opt. Soc. Am A 3,
2086 2093 (1986).
9. H. Ling and S. W. Lee, Focusing of electromagnetic waves
through a dielectric interface, J. Opt. Soc. Am. A 1, 965
973 (1984).
10. J. Gasper, G. C. Sherman, and J. J. Stamnes, Reflection
and refraction of an arbitrary wave at a plane interface,
J. Opt. Soc. Am. 66, 955 961 (1976).
11. M. Yeung, Modeling high numerical aperture optical lithography, in Optical/Laser Microlithography, B. J. Lin, ed.,
Proc. Soc. Photo-Opt. Instrum. Eng. 922, 149 167 (1988).
12. M. Yeung, Photolithography simulation on non-planar substrate, in Optical/Laser Microlithography III, V. Pol, ed.,
Proc. Soc. Photo-Opt. Instrum. Eng. 1264, 309 321 (1990).
13. C. Yaun and A. Strojwas, Modeling optical microscope images of integrated-circuit structures, J. Opt. Soc. Am. A 8,
778 790 (1991).
14. D. G. Flagello and T. Milster, Three-dimensional modeling
of high numerical aperture imaging in thin films, in Design, Modeling, and Control of Laser Beam Optics, Y. Kohanzadeh, G. N. Laurence, J. G. McCoy, and H. Weichel, eds.,
Proc. Soc. Photo-Opt. Instrum. Eng. 1625, 246 261 (1992).
15. D. G. Flagello, A. E. Rosenbluth, C. Progler, and J. Armitage, Understanding high numerical aperture optical
lithography, Microcircuit Eng. 17, 105 108 (1991).
16. D. G. Flagello and A. E. Rosenbluth, Lithographic tolerances based on vector diffraction theory, J. Vac. Sci. Technol. 10, 2997 3003 (1992)
17. D. G. Flagello and A. E. Rosenbluth, Vector diffraction
analysis of phase-mask imaging in photoresist films, in Optical/Laser Microlithography VI, J. D. Cuthbert, ed., Proc.
Soc. Photo-Opt. Instrum. Eng. 1927, 395 412 (1993).
18. D. G. Flagello, High numerical aperture imaging in homogeneous thin films, Ph.D. dissertation (University of
Arizona, Tucson, Ariz., 1993).

Flagello et al.
19. E. Wolf, Electromagnetic diffraction in optical systems. I.
An integral representation of the image field, Proc. R. Soc.
London Ser. A 253, 349 357 (1959).
20. H. H. Hopkins, Image formation with coherent and partially coherent light, Photogr. Sci. Eng. 21, 114 122 (1977).
21. D. S. Goodman, Stationary optical projectors, Ph.D. dissertation (University of Arizona, Tucson, Ariz., 1979).
22. J. D. Gaskill, Linear Systems, Fourier Transforms, and Optics (Wiley, New York, 1978).
23. R. A. Chipman, Polarization aberrations, Ph.D. dissertation (University of Arizona, Tucson, Ariz., 1987).
24. R. A. Chipman, The mechanics of polarization ray tracing,
in Polarization Analysis and Measurement, R. A. Chipman
and D. H. Goldstein, eds., Proc. Soc. Photo-Opt. Instrum.
Eng. 1746, 62 75 (1989).

25. P. Debye, Das Verhalten von Lichtwellen in der Nahe


eines
Brennpunktes oder einer Brennlinie (Behavior of light
waves in the proximity of a focal point or focal line), Ann.
Phys. (Leipzig) 30, 755 776 (1909).
26. R. K. Luneberg, Mathematical Theory of Optics (U. of California, Berkeley, Calif., 1944); reprint: (Brown U., Providence, R.I., 1964).
27. E. Wolf and Y. Li, Conditions for the validity of the Debye
integral representation of focused fields, Opt. Commun. 39,
205 210 (1981).
28. H. A. Macleod, Thin Film Optical Filters (McGraw-Hill, New
York, 1989).
29. J. A. Stratton, Electromagnetic Theory (McGraw-Hill, New
York, 1941).
30. D. S. Goodman, Topics in image formation, short course
presented at Optical Society of America Annual Meeting,
Orlando, Fla., October 19, 1989.
31. J. E. Harvey, Fourier treatment of near-field scalar diffraction theory, Am. J. Phys. 47, 974 980 (1979).
32. R. Hunt, Magneto-optical scattering from thin solid films,
J. Appl. Phys. 38, 1652 1671 (1967).

You might also like