Using a 50kV electron beam to pattern extreme-UV (EUV) masks for circuit fabrication can result in systematic errors that affect reliability in the finished product. This is because electrons scatter within the blank substrate,1, 2 leading to additional exposure of the photoresist, which drives errors of critical dimension (CDs). We regard these errors as the EUV process signature. Figure 1 shows an example of the error in drawing space features (separating trenches in the mask absorber pattern) of 80nm with varying distance, or pitch, between them. Our previous work investigated the impact of systematic CDs on the lifetime performance of 10nm devices, and identified the need for tight control of space patterns drawn on the mask.3
Figure 1. Critical dimension (CD) error for 80nm features drawn at varying distances from one another (‘space through pitch’).
It is not possible to address the specific EUV process signature using tools such as conventional proximity effect correction (where the dose of the beam is modified). Therefore, we considered altering the mask design in line with a process model to compensate the process signature, much like optical proximity correction (OPC), a technique that corrects errors by moving edges or modifying the exposure dose on a per-shot basis.4–7 In analogy to OPC, this method is called mask process correction (MPC).
We tested the mask models of various electronic design automation companies over a two-year period. We calibrated these models, using different types of mask blanks and electron-beam (e-beam) pattern generators, and then assessed the quality of each model using test cells. A corrected version of each cell was printed and processed using EUV mask blanks.
To assess the model and correction quality, we considered a defined set of test gauges for each solution, covering a wide range of CD sizes, pattern types, fill densities, and different electron scatter ranges (see Figure 2). Our analysis revealed significant variation in model quality, with only a very small number of solutions able to control lines and spaces at the same time. All other models suffered from different failure modes (see Figure 3). Models B, C, and E showed large shifts of the mean CD value for the different feature types, while others (D, J) showed systematic CD trends in the ‘two-line’ pattern (see Figure 3). Since most model shortcomings were visible in this pattern, we concluded that the models were unable to adequately cover the mid-range electron scatter specific to EUV.
Figure 2. Density for conventional 50kV long-range and extreme-UV-specific mid-range electron scatter of selected test gauges.
Figure 3. Performance of all small space patterns in the sample. CD error for 50% fill (top). Range of CD error and fill density per model and across all fills (bottom). The target was to keep CD errors close to zero and the CD range across all gauges as small as possible.
Figure 4 shows statistical analysis of the CD error distribution for each correction (line-end features omitted), indicating the overall quality of the models and corresponding corrections. Almost all models returned substantial improvements: average CD errors were generally below 1.5nm, maximum errors equal to or less than 5nm for the best models, and we achieved standard deviations below 2nm.
Figure 4. Statistical analysis of the post-etch CD error distribution for each model without line-end features (LES).
Finally, we investigated the integrity of the modified pattern by comparing the printed result to the uncorrected reference by die-to-database inspection (see Figure 5). We could not examine some of the cells completely because of a surplus of defect detections. Others were defect-free, or contained only process-related issues such as isolated spots.
Figure 5. Inspection test mask overview, including locations of inspection triggers indicated by black dots. Color code shows local pattern density. MPC: Mask process correction.
We evaluated MPC solutions from various companies to test the capabilities and shortcomings of state-of-the-art tools for correction of EUV mask process signatures. It was difficult to identify a best model in this evaluation, since none provided a satisfactory correction for all the features we considered. However, we did identify a leading group of three models that were able to provide a reasonable correction quality for at least one-dimensional features with good pattern fidelity and no inspection issues.
In conclusion, there have been great improvements in the modeling and correction achieved in MPC solutions, but work remains to further reduce CD errors. This will become increasingly challenging with the extension of EUV technology toward sub-10nm nodes. We will continue to co-operate with selected companies to drive the capabilities of the tools toward the final requirements.
AMTC would like to thank all partners that participated in this evaluation for their cooperation, dedication, and high quality of work.
AMTC Dresden GmbH
Christian Bürgel is a senior member of the technical staff in the Lithography Module. He is responsible for the introduction of new 50kV processes down to 14nm technology. He also directs the evaluation of advanced e-beam correction schemes to further improve performance of advanced 50kV processes.
Keith Standiford is a principal member of the technical staff in the Emerging Lithography and Tools Group. He has more than 25 years of experience in e-beam lithography. He previously worked at KLA-Tencor, and consulted for numerous organizations on next-generation lithography systems.
GekSoon Chua is a member of the technical staff in the Mask Technology group, and is involved in mask and optical proximity correction research activities. He has more than 25 publications and eight US patents.
1. H. Tanabe, T. Abe, Y. Inazuki, N. Hayashi, Short-range electron backscattering from EUV masks, Proc. SPIE
7748, p. 774823, 2010. doi:10.1117/12.862641
2. J. Choi, R. W. Lee, S. H. Lee, B. S. Ahn, H. B. Kim, S.-G. Woo, H. K. Cho, Degradation of pattern quality due to strong electron scattering in EUV mask, Proc. SPIE
7823, p. 78230D, 2010. doi:10.1117/12.864212
3. K. Standiford, C. Bürgel, A new mask linearity specification for EUV masks based on time dependent dielectric breakdown requirements, Proc. SPIE
8880, p. 88801M, 2012. doi:10.1117/12.2023109
4. M. M. Reddy er al., Enhancement of correction for mask process through dose correction on already geometrically corrected layout data, Proc. SPIE
8522, p. 85220W, 2012. doi:10.1117/12.964408
5. T. Lin, T. Donelly, S. Schulze, Model based mask process correction and verification for advanced process nodes, Proc. SPIE
7274, p. 72742A, 2009. doi:10.1117/12.814362
6. P. Schiavone, L. Martin, C. Browning, V. Farys, F. Sundermann, S. Narukawa, T. Takikawa, N. Hayashi, A novel mask proximity correction software combining accuracy and reduced writing time for the manufacturing of advanced photomasks, Proc. SPIE
8441, p. 84411F, 2012. doi:10.1117/12.981609
7. T. Kamikubo et al., Mask process correction (MPC) modeling and its application to EUV mask for electron beam mask writer, EBM-7000, Proc. SPIE
7823, p. 782331, 2010. doi:10.1117/12.865822