Metrology, 61 Comparison of Techniques to Measure the Point Spread Function due to Scatter and Flare in EUV Lithography Systems Manish Chandhok,, Sang H. Lee, Christof Krautschik, Guojing Zhang, Bryan J. Rice, Michael Goldstein, Eric Panning, Robert Bristol, Alan Stivers, Melissa Shell October 1, 2003 Acknowledgements: ETS team (SNL), E. Gullikson (LBNL), P. Davids (Intel) 1
Source of flare in EUV optics Flare is caused by the surface roughness of optical mirrors inside the PO MSFR roughness of ~0.25 nm/mirror results in >20% flare in the ETS Flare scales as 1/ 2 For the same roughness flare is ~200X larger for a 13.5 nm system relative to 193 nm. 75% of the flare is contained within 100 m EUV flare has shorter range than DUV flare Across field intrinsic flare is uniform for EUVL tools Mirror with Surface Roughness I inc I Ref 2
Motivation Flare results in decrease in contrast and hence resolution and process window Identifying the flare level in a stepper is critical Are we underestimating the flare in EUVL systems using the 2m 2 m line to measure flare? Do we need a narrower linewidth (for e.g., 0.5 m) to measure flare? Point Spread Function due to scatter (PSF( sc ) is needed to predict CD variation and application of Flare Variation Compensation (FVC) How do we verify that the PSFsc is accurate to be used in Flare Variation Compensation (FVC)? 3
Measured flare on the ETS Vs feature size Measured flare - Kirk method (%) 30.00 25.00 20.00 15.00 Measured flare Vs line size - ETS Jun-03 Sep-02 All data taken off vertical lines, 25 um from the end Sep-02 data: refer S. Lee et. al., Proc. SPIE vol. 5037, 2003 10.00 0.00 1.00 2.00 3.00 4.00 5.00 Vertical line size (um) Flare measured by the 2 m m line (18%) is lower than flare measured with a 0.5 m m line (25%) by a factor of 1.4 We would like the measured flare values to be as close to the intrinsic flare as possible 4
Why is flare measured by a 2 m m line less than flare measured by a 0.5 m m line? Open frame flare within 2 m = 4.3%whereas that within 0.5 m m = 0.04% The PSF sc drops by an order of magnitude from 0.5 m m to 2 m The flare is calculated by convolving the PSF sc with the layout and hence any chrome in the pattern will reduce the flare The chrome from the 2 m line appears to be locally reducing the measured flare PSF(r) 1.8x10-7 1.6x10-7 1.4x10-7 1.2x10-7 1.0x10-7 8.0x10-8 6.0x10-8 4.0x10-8 2.0x10-8 0.0 PSFsc for POB2 0.5 m 2 m -5000-4000 -3000-2000 -1000 0 1000 2000 3000 4000 5000 r (nm) 5
Flare measurement Kirk test using spoke pattern Across the field 24 mm (12 copies) 250 m 4 m, 2 m, 1 m, 0.5 m wide lines Print test result Spokes are at 15 degree increments Flare = E 0 /E clear where E clear is the dose to clear a given feature 6
Calculating the flare with spoke patterns by convolution with the PSF sc from mirror roughness Flare along cut line Intensity @ Location 0.240 0.212 New Image - 4, Y = 600.0000 Inte ns ity 0.184 0.156 0.128 0.100 0.0 400.0 800.0 1200.0 1600.0 2000.0 Lo c a tio n A factor of 1.3 difference between calculated flare below the 2 m line and the 0.5 m m line Reasonably good agreement with measurements Line size (um) Units are um Calculated flare % 4.0 14.7 2.0 17.4 1.0 19.9 0.5 21.8 7
Need smaller line size to measure EUV flare Covered by Figure (Zernike 37 terms) EUV flare range Resolution of the system Not covered by 2 m line Covered by 2 m line test Using 2 m line to measure flare may neglect a large portion of the scattering contribution 8
Case to make 0.5 m m line as the standard for measuring flare in EUVL steppers Does the 0.5 m m line meet the requirements to be used for measuring flare? k 1 = 3.75 1 Line size >> the diffraction limit Resist thickness 120 nm. Line size > 4X resist thickness and hence not impacted by lateral dissolution rate Line size small enough to not mitigate flare, hence more representative of the intrinsic flare Provided the line length > 10 m, the error in the flare contributed from the Airy disc PSF is negligible 1 For = 13.4, NA = 0.1 9
Method of verification of the PSF sc from roughness Calculating flare variation Test pattern: 75 nm line, 1 m pitch, surrounded by sea of chrome 100 m Convolution with the Point Spread Function from roughness Cutline showing flare variation Calculated flare variation 12 m from edge, 15.8% flare 1 m from edge, 10.5% flare Inte ns ity 0.170 0.152 0.134 0.116 0.098 Intensity @ Location 0.080 0.000 0.020 0.040 0.060 0.080 0.100 Loc ation New Image - 1, Y = 0.0509 Location of 0.5 m line to measure flare 10
Validation of modeled CD variation using flare variation modeled with PSF sc from roughness Modeled CD variation = Flare variation (from previous slide) X measured CD sensitivity to flare DCCD (nm) 110 105 100 95 90 85 80 75 70 65 60 CD Vs Flare (EUV 2D resist) CD sensitivity to flare = 1.7 nm/%flare y = -1.7157x + 112.39 R 2 = 0.9143 Flare reticle 2 data Vertical Linear (Flare reticle 2 data Vertical) 0 5 10 15 20 25 Modeled flare (%) DCCD (nm) Comparison between expected trend and experimental data CD Vs distance from chrome edge for 75 nm line 114 112 Experiment 110 Model 108 106 104 102 100 98 96 94 0 1 2 3 4 5 6 7 8 9 10 11 12 Distance from edge Flare range = 6%, CD range = 13 nm, 1.7 nm/%flare measured CD sensitivity to flare Since the model using the PSF sc from roughness matches the experimental data reasonably well, the PSF sc from roughness is suitable to use for Flare Variation Compensation (FVC) 11
Alternative method of extracting the PSF from sc the MTF - Theory For an incoherently illuminated object, the Fourier transform of the PSF is the Optical Transfer Function of the system (in our case stepper and resist) In 1D, the Fourier transform of the Line Spread Function (LSF) is given by Hence, if M(k x ) can be measured, then the LSF can be obtained by taking the inverse Fourier Transform of M(k x ) LSF has contributions from aberrations (high spatial frequencies), resist, and scatter (low spatial frequencies) LSF = LSF F LSF( x) M( k )exp[ i( k )] x LSF aberr+resist LSF scatter x where M(k x ) is the Modulation Transfer Function (MTF), and (k x ) is the Phase Transfer function (PTF) Note: For simplicity, we will refer to the LSF as the PSF 12
Experiment to measure the PSF Equal line space patterns (100 nm to 20 m pitches) Only the pitches above ~5/NA (670 nm) contribute to flare. The MTF can be measured using the clearing dose (D c ) and appearing dose (D( a ) of line-space gratings for a particular spatial frequency M ( k x ) [ D D c c ( k ( k x x ) ) D D a a ( k ( k x x )] ) Note: The PTF (k x ) can be measured by locating the center of the lines and compare the center shift with respect to a reference - but was not measured here 13
Results comparison of measured MTF Vs aerial image model MTF Vs Flare for POB2 1.2 1 Contrast degradation due to defocus, higher order aberrations, and resist effects MTF (Aerial image) 0.8 0.6 0.4 0.2 0 0.1 1 10 Flare = 0 Flare = 5% Flare = 10% Flare = 15% Flare = 20% Flare = 25% ETS data Spatial frequency (1/um) Contrast measured at low spatial frequencies is higher than expected from ETS (> 20% flare) because of flare mitigation by 1:1 chrome features 14
MTFsc Results comparison of PSF sc 0.85 0.8 0.75 0.7 0.65 0.6 0.55 Measured MTFsc of ETS (including effects of resist and aberrations) Trend in PSF sc from Modulation Transfer Function (MTF) is similar to that from mask roughness Structures used to measure the (MTF) mitigate the flare resulting in a lower measured PSF sc y = 0.0068x 3-0.039x 2 + 0.0034x + 0.7879 R 2 = 0.986 0.5 0.1 1 10 Spatial frquency (1/um) sc Recommend limiting chrome size to 0.5 m, pitch varying PSFsc (1/um^2) Point spread function for the ETS obtained from the inverse Fourier transform of the MTF 1.E+00 1.E-01 1.E-02 1.E-03 1.E-04 1.E-05 1.E-06 1.E-07 0 1 2 3 4 5 6 7 8 9 10 Distance (um) PSFsc from MTF PSFsc from mirror roughness 15
Summary It is recommended that the 0.5 m m line be used as the standard to measure flare in the field for EUVL tools since it is closer to the intrinsic flare value Based on the experiments on the ETS, the PSF sc from roughness is sufficiently accurate to do Flare Variation Compensation (FVC) However, the PSF sc trend needs to similar from tool to tool for FVC to be feasible for HVM Lithographic measurement of the PSFsc using the resist clearing method has been demonstrated on the ETS. The MTF method to extract the PSF sc needs to be modified for EUVL tools to minimize the flare mitigation by large chrome features 16