Effects of the illumination NA on EUV mask inspection with
coherent diffraction imaging
Ricarda Nebling, Hyun-su Kim, Uldis Locans, Atoosa Dejkameh, Yasin Ekinci, and Iacopo
Mochi
Paul Scherrer Institut, 5232 Villigen PSI, Switzerland
ABSTRACT
RESCAN is a coherent diffraction imaging based APMI microscope prototype. A complex image of the EUV
reticle is reconstructed from diffraction patterns collected on a CCD detector. With the next upgrade of the
tool, the resolution will be enhanced from the current 34 nm down to 20 nm on mask. Also the illumination NA
value will change from the current range of 0.002 to 0.02 to a value of 0.035. Here, we study how a change of
the illumination NA affects the EUV mask inspection in simulation. We observe a better image quality, lower
object error and higher defect sensitivity with increasing illumination NA.
Keywords: EUV lithography, actinic patterned mask inspection, coherent diffraction imaging, ptychography,
illumination NA
1. INTRODUCTION
RESCAN (reflective-mode EUV mask scanning lensless microscope) is an APMI microscope prototype developed
at the Paul Scherrer Institut. It is based on a coherent diffraction imaging
1
technique called ptychography,
2, 3
where the reticle is scanned with spatially-confined, coherent EUV illumination in overlapping positions. On a
CCD detector in the far-field, the diffraction patterns for each scan position are collected. The (complex) image
of the reticle is reconstructed from the intensity data by using a phase-retrieval algorithm.
M
1
Detector
M
2
Aperture
stop
EUV beam
EUV reticle sample
Pellicle
(a)
(b)
Figure 1. (a) Schematic of the current RESCAN setup. (b) Rendering of the upgraded RESCAN setup, with a Fourier
synthesis illuminator.
The current RESCAN tool (the optical layout is depicted in fig. 1 (a)), has a resolution of 34 nm on mask. The
EUV beam is focused onto the sample with an angle of incidence of 6
o
and the illumination numerical aperture
(NA) value ranges from 0.002 to 0.02.
4
An upgrade of the tool is under construction: an image of the new design
is shown in fig. 1 (b). The two mirror system will be replaced by a Fourier synthesis illuminator, which will
Further author information:
E-mail: ricarda.nebling@psi.ch
Extreme Ultraviolet Lithography 2020, edited by Patrick P. Naulleau, Paolo A. Gargini, Toshiro Itani, Kurt G. Ronse,
Proc. of SPIE Vol. 11517, 115170W · © 2020 SPIE · CCC code: 0277-786X/20/$21 · doi: 10.1117/12.2573181
Proc. of SPIE Vol. 11517 115170W-1
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Nov 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
enhance the resolution on mask down to 20 nm.
5
A visible light microscope will facilitate the navigation on the
sample. Furthermore, the illumination NA changes to a larger value of 0.035.
In this paper, we study in simulation how a change in the illumination NA influences the EUV mask inspection
with coherent diffraction imaging.
2. ILLUMINATION NA SIMULATION
In this study, we are performing defect inspection on simulated data sets with illumination NA values ranging
from NA
illum
= 0.005 to NA
illum
= 0.050. The lowest NA value corresponds to the typical illumination NA
of the current RESCAN setup, the range is chosen to cover the illumination NA for the RESCAN upgrade
(NA
illum
= 0.035). As our current RESCAN tool is limited to a maximum illumination NA of 0.02, we study
the influence of the illumination NA on EUV mask inspection in simulation only.
The EUV mask sample is generated from a mask design file with programmed absorber defects, varying in
size from 200 nm to 20 nm. We assume the sample to be binary, and add a phase shift to the absorber layer, that
corresponds to 70 nm TaBN under 6
o
angle of incidence at a wavelength of 13.5 nm. The complex illumination
function (also called probe) is simulated matching the characteristics of the current RESCAN setup (fig. 1 (a)).
It is generated as an image of the beam shaping aperture on the reticle, with a wavelength of 13.5 nm (EUV).
The illumination NA is determined by the aperture stop diameter and by its distance from the sample. The
EUV reticle is moved through focus to maintain a constant probe diameter of 10 µm on object for all data sets.
Images of the probe magnitudes are shown in fig. 2.
NA = 0.005
NA = 0.015
NA = 0.025
NA = 0.035
NA = 0.050
1e6
1.0
0.8
0.6
0.4
0.2
0.0
Figure 2. Simulated illumination magnitudes without noise on the object plane. From left to right and top to bottom
increasing illumination NA.
To generate the diffraction patterns, the probe function is multiplied with the object region, that corresponds
to the current scan position. The product corresponds to the exit wave and is propagated to the detector
in the far-field, using a Fourier transform and an appropriate phase term.
6
The squared absolute value of
the propagated exit wave gives the noise-free diffraction pattern. To get the full ptychographic data set, this
procedure is repeated for all scan positions. The average detector count per data set is scaled according to real
RESCAN data. We add Poisson noise to each diffraction pattern. The sample is scanned in a circular pattern
with a step size of 1 µm to avoid regular grid pathology.
7
Proc. of SPIE Vol. 11517 115170W-2
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Nov 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
3. EUV RETICLE IMAGE RECONSTRUCTION
We reconstruct a (complex) image of the EUV mask for each illumination function using the difference map
algorithm.
7
The algorithm is run for 250 iterations, while updating the probe function from the 100th iteration
on. The initial object is generated following a random uniform distribution of values between zero and one, the
initial probes are the complex illumination functions as simulated.
To compare the reconstructions quantitatively, we introduced the object error
8
E
obj
=
P
r
P
i,j
||O
ref
ij
| − |
ˆ
O
ij
||
2
P
r
P
i,j
1
. (1)
The absolute value of the reconstructed object
ˆ
O is compared at each iteration to a reference object O
ref
from the sample layout. The error is normalized to the total number of pixels. The object error for all NAs is
shown in fig. 3. After the probe update starts, at the 100th iteration, the error is decaying fast to a steady level
for all the illumination NAs. Fluctuations around that level, best visible for the lower NA curves, are inherent
to the difference map algorithm, that typically reaches a steady state close to the optimal solution.
7
To find the
optimal image reconstruction, one can average the solutions of the last few iterations,
7
or run for more iterations
with another algorithm (for example from the PIE-family
9
) that is known to converge more likely to the global
minimum. We observe a larger object error for the two smallest NAs, and a smaller object error for the largest
three illumination NAs. The lowest error is observed for the largest NA (NA
illum
= 0.050). From fig. 3, we
observe a trend for a lower object error and hence a more accurate reconstruction of the sample, with larger
illumination NA.
Figure 3. Object error for the illumination NAs.
4. DEFECT DETECTION
The simulated EUV mask contains an area with planned defects ranging in size from 200 nm to 20 nm. A
schematic of the region is shown in fig. 4 (c). Each of the nine crosses contains line intrusions and extrusions,
one corner defect and one pin-dot defect with respective size. The critical dimension of the sample is 200 nm.
To detect the defects, we perform a die-to-database comparison for each of the reconstructed mask images.
4
A
die with defects, depicted in fig. 4 (a), is compared to the defect free reference shown in 4 (b).
Proc. of SPIE Vol. 11517 115170W-3
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Nov 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
200 nm 80 nm 40 nm
150 nm
60 nm
30 nm
100 nm
50 nm 20 nm
750 nm
(a)
(b)
(c)
(d)
(e) (f)
Figure 4. Defect maps for three different illumination NAs. (a) Schematic of a sample area with defects and (b) the
respective reference from the database. (c) Layout of the sample design with defect sizes ranging from 200 nm to 20 nm.
(d) Shows the defect map overlaid on the reconstructed image magnitude for an illumination NA of 0.005, (e) for an
illumination NA of 0.025, and (f) from an illumination NA of 0.050, from a die-to-database comparison.
In fig. 4 (d), (e), and (f), the defect maps for three different illumination NA values (NA
illum
= 0.005,
NA
illum
= 0.025, and NA
illum
= 0.050) are shown overlaid with the reconstructed EUV mask image magnitudes.
From a first look, we observe a better image quality for the larger illumination NA reconstructions. The features
are better resolved and the pattern is well visible for the largest illumination NA in fig. 4 (f). Considering all
three defect maps, we see that more defects are detected down to smaller defect sizes for a larger illumination
NA. It is important to note that this is a simulation study. For real RESCAN data we demonstrated defect
detection down to 50 by 50 nm
2
with an illumination NA of 0.002.
4
To look more closely on the influence of the illumination NA on EUV mask inspection with coherent diffraction
imaging, we listed the detected defect signals for different illumination NA and defect size in a table for each
defect type. The tables for line intrusions (left) and extrusions (right) are shown in fig. 5. For both line defect
types, smaller defects are detected with increasing illumination NA. Please note, that the defect signals in the
table are not corrected for false positives.
In fig. 6 the same tables for corner and pin-dot defects are shown. For all defect types, we observe a higher
defect sensitivity with larger illumination NA.
Proc. of SPIE Vol. 11517 115170W-4
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Nov 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
200
150
100
80
60
50
40
30
20
defect size [nm]
NA = 0.005
NA = 0.015
NA = 0.025
NA = 0.035
NA = 0.050
200
150
100
80
60
50
40
30
20
defect size [nm]
Figure 5. Table containing the detected defect signals for the different illumination NAs with changing defect size. On
the left, a table with line intrusion defects is shown, while on the right, a table with line extrusion defects is shown.
200
150
100
80
60
50
40
30
20
defect size [nm]
NA = 0.005
NA = 0.015
NA = 0.025
NA = 0.035
NA = 0.050
200
150
100
80
60
50
40
30
20
defect size [nm]
Figure 6. Left: Table showing the corner defect signals for different NAs and changing defect size. On the right, the same
table with pin-dot defects is shown.
5. CONCLUSIONS AND OUTLOOK
In this paper, we studied the effects of the illumination NA on EUV reticle inspection with coherent diffraction
imaging. We first simulated ptychographic data sets for several illumination functions with illumination NA
values ranging from NA
illum
= 0.005 to NA
illum
= 0.050, and reconstructed the (complex) EUV mask image
using the difference map algorithm.
We observed that larger illumination NA values yield lower object errors. The reconstructed image quality
is better for a larger illumination NA and the pattern is better resolved. We furthermore observe a higher
defect sensitivity with larger illumination NA values. More and smaller defects are detected in a die-to-database
comparison for the large illumination NA data sets.
We expect that with the RESCAN upgrade to a Fourier synthesis illuminator and an increased illumination
NA, we will enhance the resolution to 20 nm on mask and get a higher defect sensitivity.
REFERENCES
[1] Chapman, H. N. and Nugent, K. A., “Coherent lensless X-ray imaging,” Nat. Photonics 4(12), 833–839
(2010).
[2] Hoppe, W., “Beugung im inhomogenen Prim¨arstrahlwellenfeld. III. Amplituden- und Phasenbestimmung bei
unperiodischen Objekten,” Acta Crystallogr. Sect. A 25(4), 508–514 (1969).
[3] Rodenburg, J. M., “Ptychography and related diffractive imaging methods,” Adv. Imaging Electron
Phys. 150(07), 87–184 (2008).
Proc. of SPIE Vol. 11517 115170W-5
Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 16 Nov 2020
Terms of Use: https://www.spiedigitallibrary.org/terms-of-use