1 July 2006 Concise formula for the Zernike coefficients of scaled pupils
Author Affiliations +
Abstract
Modern steppers and scanners have a projection lens whose numerical aperture (NA) can be varied so as to optimize the image performance for certain lithographic features. Thus a variable fraction of the aberrations is actually involved in the imaging process. In this letter, we present a concise formula for the NA scaling of the Zernike coefficients. In addition, we apply our results to the Strehl ratio.
Janssen and Dirksen: Concise formula for the Zernike coefficients of scaled pupils

1.

Introduction

It is well known that different lithographic features have different optimal settings of the numerical aperture (NA) and illumination condition.1, 2 Typically the depth of focus or the process latitudes are optimized. To achieve the optimal settings, modern steppers and scanners have a projection lens with a variable numerical aperture, where the NA value can be reduced to about 70% of its maximum value. As a consequence, a smaller fraction of the lens pupil and a corresponding smaller fraction of the phase aberrations is actually involved in the imaging process.

In the literature (see Ref. 3 for a survey) the problem of computing the aberration Zernike coefficients of scaled pupils has been discussed at various places. The recent result of Dai3 gives the Zernike coefficients of the scaled pupil in terms of those of the unscaled pupil in analytic form [see Eq. 18].

In this letter we give an alternative expression for Dai’s result. Our main result in Eq. 4 agrees mathematically with Dai’s Eq. 18 but has the advantage that it is very simple and direct in terms of Zernike polynomials. Thus it leads to the simple and explicit results in Eqs. 7, 9 from which the sensitivity of the aberration coefficients and Strehl ratios can be assessed when NA is close to its maximum value. The short and elegant proof of our main result in Eq. 4 is given in Appendix A.

2.

Zernike Coefficients of Scaled Pupils

We consider a pupil function

1

P(ρ,θ)=exp{iΦ(ρ,θ)},
on a unit disk 0ρ1 with real phase Φ , and we assume that Φ is expanded as a Zernike series according to:

2

Φ(ρ,θ)=n,mαnmRnm(ρ)cos(mθ),
where m and n are integers with nm being even and 0 , and Rnm(ρ) denotes the Zernike radial polynomial4 of azimuthal order m and degree n . For simplicity we only consider cosin terms. Scaling to a smaller pupil with relative size ϵ=NANAmax1 means that we have to expand the scaled phase Φ(ϵρ,θ) into a Zernike expansion:

3

Φ(ϵρ,θ)=n,mαnm(ϵ)Rnm(ρ)cos(mθ),0ρ1,0θ2π.
The problem is how to express the Zernike coefficients αnm(ϵ) of the scaled pupil function in terms of the Zernike coefficients αnm of the unscaled pupil function. In the Appendix A we prove our main result:

4

αnm(ϵ)=nαnm[Rnn(ϵ)Rnn+2(ϵ)],n=m,m+2,,
where the summation is over n=n,n+2, , and where we use the convention that Rnn+20 . We note that (1) the entries Rnn(ϵ)Rnn+2(ϵ) do not depend on m ; we consider them only for n,n=m,m+2, , and (2) αnm(ϵ=1)=αnm since all involved Zernike polynomials, except Rnn+2(ϵ)0 , equal 1 at ϵ=1 .

3.

Examples and Sensitivity Analysis

Figure 1 shows an example where we scale low-order coma α31(ϵ) with α31=2π(0.016) and α51=2π(0.016) . For this special case, Eq. 4 and Dai’s Eq. 18 become

5

α31(ϵ)=α31R33(ϵ)+α51[R53(ϵ)R55(ϵ)]
and

6

α31(ϵ)=ϵ3[α31+4α51(ϵ21)],
respectively. Both equations give an identical result, as R33(ϵ)=ϵ3 and R53(ϵ)R55(ϵ)=4ϵ3(ϵ21) .

Fig. f1

Scaling low-order coma α31(ϵ) when α31=2π(0.016) and α51=2π(0.016) . The solid line represents Eq. 4, the dashed line is the tangent line given by Eq. 7, and Eq. 18 is represented by circles.

030501_1_1.jpg

A consequence of Eq. 4 is that

7

αnmϵϵ=1=nαnm+2(n+1)[αn+2m+αn+4m+].
Equation 7 shows a significant sensitivity of αnm(ϵ) when ϵ is slightly below its nominal value of 1, especially when high-order aberrations are present. In our example we obtain a relative sensitivity 1α(αϵ) of 11, indicating that the NA value should be specified to 103 accuracy when the required coefficient accuracy is 1%.

The Strehl ratio4 is approximated as:

8

S1n,m(αnm)2γm(n+1)
in which γ0=1,γ1=γ2==2 and the term with n=m=0 is omitted. A consequence of Eq. 7 is that

9

Sϵϵ=1=2n,m(αnm)2γm(n+1)2m1γm(nαnm)2 =2[Φdisk2¯Φrim2¯],
where the two averaging operations are over the whole disk 0ρ1 and the rim ρ=1 , respectively. Figure 2 shows the dependence on ϵ of the Strehl ratio in the case of a mixture of low- and high-order spherical aberration: α40=2π(0.02) and α60=2π(0.02) , where the scaled Strehl ratio S(ϵ) is calculated by combining Eqs. 8, 4. This result compares well with the numerical calculation obtained from the lithography simulator SOLID-C.5 Intuitively it is expected that the Strehl ratio increases when the NA is decreased. However, in general this is not true. In the special case shown in Fig. 2, the Strehl ratio decreases when the NA is decreased from its maximum value. This result can be seen as follows: as in our example nαnm=0 for each m value, the phase aberration Φ(1,θ) at the rim of the pupil equals 0. From Eq. 9 it then follows that the slope at ϵ=1 is positive.

Fig. f2

Scaling the Strehl ratio when α40=α60=2π(0.02) . The solid line represents the combination of Eqs. 8, 4, the circles represent SOLID-C numerical calculation, and the dashed line represents the tangent line given by Eq. 9.

030501_1_2.jpg

Appendices

Appendix A: Proof of the Main Result, Eq. 4

By decoupling per azimuthal order m=0,1, , and normalization and orthogonality4 of Rnm(ρ),n=m,m+2, , the Zernike coefficients αnm(ϵ) of Φ(ρϵ,θ) and the Zernike coefficients αnm of Φ(ρ,θ) are related by

10

αnm(ϵ)=2(n+1)nαnmMnnm(ϵ),n=m,m+2,.
The summation in Eq. 10 is over n=m,m+2, , and

11

Mnnm(ϵ)=01Rnm(ρ)Rnm(ρϵ)ρdρ,n,n=m,m+2,.
We shall show that

12

Mnnm(ϵ)=12(n+1)[Rnn(ϵ)Rnn+2(ϵ)];
in particular, it follows that Mnnm(ϵ)=0 when n<n and that Mnnm(ϵ) does not depend on m , except that in Eq. 10 we only use n,n=m,m+2, . For this we use6

13

Rlk(ρ)=(1)(lk)20Jl+1(r)Jk(ρr)dr,0ρ<1,
when k,l are integers 0 with same parity; in the case of kl> 0 , the right-hand side of Eq. 13 vanishes, which is consistent with the convention that then Rlk0 . We use Eq. 13 in Eq. 11 to rewrite Rnm(ρϵ) and interchange integrals to get

14

Mnnm(ϵ)=(1)(nm)20Jn+1(r)[01Rnm(ρ)Jm(ρϵr)ρdρ]dr.
To the inner integral we apply the result

15

01Rnm(ρ)Jm(ρv)ρdρ=(1)(nm)2[Jn+1(v)v]
from the Nijboer-Zernike theory,4 and we get

16

Mnnm(ϵ)=(1)(n+n2m)20Jn+1(r)Jn+1(ϵr)ϵrdr.
Next we use the identity7

17

Jn+1(ϵr)ϵr=Jn(ϵr)+Jn+2(ϵr)2(n+1),
and use Eq. 13 to rewrite the resulting two integrals in terms of Zernike polynomials. This gives Eq. 12.

Appendix B: Dai’s Formula

We reproduce Dai’s formula3 for the scaling of Zernike coefficients:

18

αnm(ϵ)=ϵn[αnm+(n+1)i=1(Nn)2αn+2imj=0i(1)i+j(n+i+j)!(n+j+1)!(ij)!j!ϵ2j]
with N being the maximum n value.

References

1.  B. J. Lin, “Signamization of resist images,” Proc. SPIE0277-786X 3051, 620–628 (1997). Google Scholar

2.  A. Engelen, R. Socha, E. Hendrickx, W. Scheepers, F. Nowak, M. van Dam, A. Liebchen, and D. Faas, “Optimization on step & scan systems,” Proc. SPIE0277-786X 5377, 1323–1333 (2004). Google Scholar

3.  G. M. Dai., “Scaling Zernike expansion coefficients to smaller pupil sizes: a simpler formula,” J. Opt. Soc. Am. A0740-3232 10.1364/JOSAA.23.000539 23, 539–543 (2006). Google Scholar

4.  M. Born and E. Wolf, Principles of Optics, 7th ed., Sec. 9.2, Cambridge University Press, Cambridge, MA (2001). Google Scholar

5. SOLID-C (release 6.3.0), SIGMA-C GmbH, Thomas-Dehlerstrasze 9, D-81737 Munich, Germany. Google Scholar

6.  R. J. Noll, “Zernike polynomials and atmospheric turbulence,” J. Opt. Soc. Am.0030-3941 66, 207–211 (1976). Google Scholar

7.  M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions, p. 361, Dover, New York (1970). Google Scholar

Augustus J.E.M. Janssen, Peter Dirksen, "Concise formula for the Zernike coefficients of scaled pupils," Journal of Micro/Nanolithography, MEMS, and MOEMS 5(3), 030501 (1 July 2006). https://doi.org/10.1117/1.2345672
JOURNAL ARTICLE
3 PAGES


SHARE
Back to Top