February 2, 2012 — Extreme ultraviolet lithography (EUVL) is a leading option for semiconductor technology scaling below 22nm. EUVL adoption depends not only on the instrumentation, but on image-correction software that can correct new distortions seen with EUVL.
Generally speaking, EUVL requires less optical proximity correction (OPC) than 193nm lithography. In fact, device feasibility studies were carried out with limited or no correction. However, a rigorous optical correction strategy and an appropriate electronic design automation (EDA) infrastructure are critical to face the challenges of the 22nm node and beyond, and EUV-specific effects such as flare and shadowing have to be fully integrated in the correction flow and properly tested.
The EDA industry has developed computational lithography solutions for handling all known EUVL imaging issues, including the modeling and correction of mask shadowing effects. In a previous article, we discussed EUV flare effects. In this article, we explore the current capabilities of model-based OPC software to model and correct for the shadowing distortions unique to EUVL. We describe the distortions, and explain why a full-chip, model-based OPC solution is the best approach at sub-22nm nodes.
What are mask shadowing effects?
EUV scanners use an off-axis, all-reflective imaging system that is non-telecentric on the mask side. This means that the rays striking the reflective mask have an incidence angle of 6-degrees, and a secondary azimuthal component that varies across the scanner slit (Figure 1).
 |
| Figure 1. Source of the mask shadowing effect in EUV lithography. At the mask side, the chief rays are incident at a constant 6-degree angle, with an azimuthal angle which varies across the scanner slit from approximately 68 degrees to 112 degrees. |
The impact of such an imaging system, when combined with the topography of the mask, is an orientation-dependent shadowing effect, which imparts a directional CD bias and shift. The magnitude of this effect is on the order of several nanometers [1].
Since the effect is too large to be ignored, it must be simulated and corrected in OPC. It is generally preferable to compensate for imaging effects through model-based OPC, rather than rules-based schemes. The infrastructure for model-based compensation of mask shadowing effects exists within the domain decomposition method (DDM) model in the Calibre nmOPC tool from Mentor Graphics [2].
Mask shadow compensation experiments
To fully study the challenges inherent in creating a full-chip, model-based OPC solution for EUV, we set up the experiment using the following conditions:
- λ = 13.5nm, NA = 0.32, Sigma = 0.5/0.25 (annular);
- 10nm image diffusion, CTR resist model;
- Simulation and OPC performed with Calibre nmOPC;
- MaskSim, Domain Decomposition 3D Mask Model;
- Mask film stack: 70nm total absorber thickness, 40 bilayers;
- Test layout = 16nm node design rule “torture test”
-45nm minimum pitch — vary pitch, vary aspect-ratio (2D patterns)
-Varying pattern types
-Patterns in both horiz/vertical orientation
-Patterns arrayed across the slit (x-axis)
-Patterns arrayed with varying surrounding density in scan direction (y-axis)
We set the mask focal plane in our DDM 3D mask model to -160nm, where we find the image shift in a 25nm dense contact induced by the non-telecentric optics to be zero [3] (Figure 2).
 |
| Figure 2. Mask focal plane setting optimization to reduce the image shift due to the EUV system’s non-telecentricity. |
Figure 2 shows that a mask focal plane setting of -160nm minimizes pattern shifts. This setting was used in the rest of the simulation study.
Plots of the CD error and shift for horizontal and vertical patterns for each test structure at the center of the slit (phi=90 o) are shown in Figure 3 (for measurements on the image) and Figure 4 (for measurements made of the post-OPC mask displacement).
 |
| Figure 3. CD image error and pattern shift differences for horizontal and vertical oriented test patterns. |
 |
| Figure 4. CD mask displacement bias after OPC for horizontal and vertical oriented test patterns. |
There are several conclusions we can draw from analyzing these results. First, there is almost no shift (overlay error) in the image or in the post-OPC mask shape. This seems to indicate that our centering of the mask focal plane is reasonably correct. Second, there is a rather pronounced variation in the CD H-V image bias depending upon the pattern type and pitch. The total range of measured CD H-V biases of ~14nm would seem to suggest the difficulty of implementing a simple rules-based mask shadowing compensation scheme. However, when observing the associated mask displacement needed to neutralize the H-V CD bias during OPC, the range is much smaller: only 1.3nm.
Can mask shadowing be adequately compensated by applying a simple bias to the horizontal CDs of 2.33nm (the average mask bias in Figure 4)? This was attempted, with results shown in Figure 5. The data seems to indicate that by treating mask shadowing with a constant mask bias, we will not be able to adequately compensate the densest pitches, leaving a residual 3.7nm of uncorrected CD error.
 |
| Figure 5. Results of an attempt at rules-based compensation of the mask shadowing effect with a constant bias to the horizontal edges of -1.15nm. |
Conclusion
Computational lithography infrastructure to model and correct for EUVL distortions is evolving along with the exposure tools. No known show-stoppers in computational lithography have yet been encountered, and solid foundations seem to exist for handling all known EUV imaging issues.
Mask shadow compensation is likely ready, awaiting stable and accurate wafer data to fully validate the solution. It is clear that rules-based solutions are inadequate, and only a model-based approach, as seen with the domain decomposition method (DDM) is able to compensate for EUVL mask shadowing effects. Work is ongoing to study the feasibility of decoupling across-reticle effects from the OPC, with the goal of reducing the complexity and runtime of the overall flow.
References:
[1] G. McIntyre, et. al. “Modeling and Experiments of Non-Telecentric Thick Mask Effects for EUV Lithography,” SPIE 2009, vol. 7271.
[2] K. Adam et. al. “Methodology for accurate and rapid simulation of large arbitrary 2D layouts of advanced photomasks,” Proc. of SPIE 2002, Vol. 4562
[3] Lorusso, et. al. “EUV at IMEC: shadowing compensation…strategy,” JVST-B 2007, Nov/Dec vol. (25/6)
James Word received his BSEE from the University of New Mexico and is a product manager for OPC at Mentor Graphics. He can be contacted at [email protected].
Christian Zuniga received his BS, MS, and PhD in electrical engineering at the University of Texas at Austin and is a technical marketing engineer at Mentor Graphics. He can be contacted at [email protected].
Read the first article in this mini-series: EUV lithography flare distortion correction