A more ecient Rigorous Coupled-Wave Analysis algorithm

M.G.M.M.v.Kraaij1 and J.M.L.Maubach2

Technische Universiteit Eindhoven (TU/e), Den Dolech 2, Eindhoven, The Netherlands. [email protected] Technische Universiteit Eindhoven (TU/e), Den Dolech 2, Eindhoven, The Netherlands. [email protected]

Summary. We present a modication of a well-known mathematical model based on the Rigorous Coupled-Wave Analysis (RCWA) that can be used to solve optical diraction problems on periodic structures (both 1-D and 2-D gratings with approximated layer-structure). The algorithm calculates the reected and transmitted eld which in turn determine the diraction eciencies for all reected and transmitted orders. Results created with a Matlab implementation of the modied RCWA algorithm (MSolver) show excellent overlap with other published and measured data.

Key words: Rigorous Coupled-Wave Analysis, RCWA, Diraction grating

1 Introduction
Lithography often uses gratings for various metrology tasks such as alignment, overlay metrology and CD metrology. With the tightening requirements on metrology accuracy it becomes increasingly more important to understand the behaviour of the grating in the metrology application using a rigorous mathematical diraction model. In order to understand the complexity of the grating problem, it is necessary to realize that nowadays gratings have complex proles and consist of all kinds of dierent materials. Real-life lithography does not produce symmetric proles with sinusoidal, rectangular or trapezoidal grooves for very high and very low groove frequencies. Moreover in the visible region and for shorter wavelengths the nite conductivity complicates the grating response and requires more complex mathematical models. The RCWA algorithm is often used because of its good convergence and relatively simple implementation. The algorithm uses a layered structure to approximate the grating prole but for the material properties no approximations are used.

M.G.M.M.v.Kraaij and J.M.L.Maubach

This paper presents a modied version of the RCWA algorithm which solves optical diraction problems better for the case of highly conducting materials. For a full derivation of the discrete equations from Maxwells equations, see [1, 2]. The modication in section 3 from equation (6) on (based on material from [3]) makes the RCWA algorithm converge faster. Presented numerical results show that the modied method converges much faster, especially for metallic gratings.

2 The model
First consider the diraction problem in Fig. 1 which leads to Maxwells equations which are the basis for the RCWA algorithm. This is the standard model with the standard assumptions: A linearly polarized electromagnetic eld with angle is obliquely incident at an arbitrary angle of incidence and at an azimuthal angle upon a dielectric or lossy grating. The grating is assumed to be innitely long in the periodic x-direction with a period . The grating grooves along the y-direction are also assumed to be innitely long. In the example below only two dierent media are present with refraction indices nI and nII .
k E m e d iu m I
in c

y q

m e d iu m

II

f L

y z

Fig. 1. One-dimensional periodic grating in R3

Fig. 2 shows how a general grating is approximated by a multilayered grating. Note that all calculations can be restricted to only one period. In each layer the material constants only depend on the horizontal x-coordinate and are independent of the vertical z-coordinate. Furthermore the dierent media are assumed to be homogeneous, linear and isotropic.

A more ecient RCWA algorithm

z = 0 = D D D

1 2

m e d iu m
2

n 2

1 x z

z = D = D

K K + 1

K + 1

m e d iu m x = L 2

II

K + 2 x =

K + 1 L 2
] 2

Fig. 2. Layered one-dimensional grating in [ , 2

(, )

3 The equations and boundary conditions

All the electric eld components Ei can be eliminated so that for each layer only one equation for the y-component of the magnetic eld Hi remains: i (x) 2 2 i (x) Hi,y (x, z) = k0 Hi,y (x, z) z 2 0 0 x 0 Hi,y (x, z) . (2) i (x) x

On the left and right of the domain the pseudo-periodic boundary conditions are applied and above and below the grating Rayleighs radiation condition is used. Finally at the layer interfaces the continuity of the tangential electromagnetic eld components are preserved. This can be reformulated in the following set of equations for the magnetic eld: Hi,y (x, Di ) = Hi+1,y (x, Di ), 1 1 Hi,y (x, Di ) = Hi+1,y (x, Di ). i (x) z i+1 (x) z (3a) (3b)

3 Only TM polarized incident light in a planar diraction case is considered here. For more details on other diraction cases see [4]

M.G.M.M.v.Kraaij and J.M.L.Maubach

The complex permittivity i (x) within each layer is expanded in a Fourier inc series. The incident magnetic eld Hy is assumed to be one plane wave and the magnetic eld expansions for each grating layer are given by equation (4): 0 0 0 0 0 0 0 0
1 2

inc Hy (x, z)

= n1

ejk0 n1 (x sin +z cos ) ,

(4a) (4b)

1 2

H1,y (x, z) = n1

1 2

Hi,y (x, z) = n1

Ui,n (z)ejkxn x ,
n

(4c)

1 2

HK+2,y (x, z) = n1

Tn ej(kxn x+kK+2,zn (zD)) .

(4d)

Note that these expansions already satisfy the pseudo-periodic boundary condition and Rayleighs radiation condition. Substituting these expansions into (2) and truncating the equations results in: d2 U i (z ) = Ei (Kx Pi Kx I) U i (z ). dz 2 (5)

However equation (5) does not uniformly preserve the continuity of the appropriate eld components across the discontinuities in one layer of the complex permittivity function. This is caused by the way in which the Fourier series are used in the truncated equations. We propose to use the truncated equations: d2 U i (z ) = P1 Kx E1 Kx I U i (z ). i i dz 2 (6)

This proposal is based on [3] which suggests that these truncations are better when there are discontinuities in one layer of the permittivity function. Equation (6) is not derived from (2) but from the basis equations (1) after multiplying (1b) with 1/i (x). So instead of rst eliminating the electric eld components, substituting the expansions and truncating the equations, we now start with substituting the expansions in (1), truncating the equations and then eliminating the electric eld components. Finally the complex reected and transmitted eld amplitudes are determined by calculating eigenvalues and eigenvectors of equation (6) and using the boundary conditions (3) at the layer interfaces. An enhanced transmittance matrix approach is used to calculate the reected and transmitted eld amplitudes in a stable way [2].

A more ecient RCWA algorithm

4 Numerical results
In this example a simple binary gold grating with TM polarized incident light is used. All the important grating parameters can be found in Fig. 3. The convergence of the original RCWA algorithm is compared with the modied RCWA algorithm in Fig. 4. Here the diraction eciency of the 0th reected order versus the total number of orders retained in the expansions is plotted. Note that the diraction eciency is just the amount of energy relative to the incident eld. Clearly the modied RCWA algorithm converges much faster.
1 0.9 0.8

l = 1 e -6

q = 6

Diffraction efficiency

p n
1

0.7 0.6 0.5 0.4 0.3 0.2

= 1

D = 1 e -6

5 e -7 L = 1 e -6
n
2

= 0 .2 2 - i* 6 .7 1

0.1 0 20 40 60 80 Orders 100

Original RCWA Modified RCWA 120 140

Fig. 3. Grating parameters

Fig. 4. Convergence 0th order

5 Conclusions
The original RCWA algorithm is worked out in detail for all diraction cases and is extended with Fourier factorization rules from [3] which also improved convergence for the C-method. The modied RCWA algorithm now also performs well for TM polarized light on metallic gratings. A Matlab implementation of the modied RCWA algorithm (MSolver) shows good overlap with other published and measured data.

References
1. Max Born and Emil Wolf. Principles of Optics, chapter 13: Optics of Metals. Pergamon Press, 6th edition, 1980. 2. M. G. Moharam, Drew A. Pommet, and Eric B. Grann. Stable implementation of the rigorous coupled-wave analysis for surface-relief gratings: enhanced transmittance matrix approach. J. Opt. Soc. Am. A, 12(5):10771086, May 1995. 3. Lifeng Li. Use of Fourier series in the analysis of discontinuous periodic structures. J. Opt. Soc. Am. A, 13(9):18701876, September 1996. 4. M.G.M.M.v.Kraaij. A more Rigorous Coupled-Wave Analysis (MSolver). Masters thesis, Technische Universiteit Eindhoven, 2004.