As EUV lithography extends to smaller pattern dimensions, photomask fabrication must support features that are both finer and more consistent. Optical proximity correction (OPC) and inverse lithography techniques can introduce sub-resolution assist features (SRAFs) to improve lithographic process latitude, but producing these small mask features reliably is challenging. A revised EUV mask fabrication process using a thinner resist and a chromium-based hard mask demonstrated absorber patterning for 30 nm SRAF lines and 24 nm spaces, alongside improved line-width roughness (LWR).
The results are relevant to semiconductor process and mask teams evaluating pattern resolution, critical dimension (CD) performance, and roughness as connected—but distinct—manufacturing requirements. The process improved several measures, but did not meet every stated target.
Why EUV masks need smaller, more controlled features
For 0.33 numerical aperture (NA) EUV lithography targeting patterning beyond the 15 nm half-pitch range, mask features must support both the primary pattern and the assist features used in OPC. The study set three mask targets: a 44 nm minimum primary feature, LWR of 2.8 nm or less for the primary feature, and minimum SRAF dimensions of 30 nm line and 24 nm space.
Conventional processing used resist as the mask during absorber etching. Making the resist thinner can help resolve smaller patterns and reduce the risk of resist pattern collapse. However, the conventional process faced challenges associated with backscatter and damage to the top of the absorber pattern. Its LWR also remained above the specified target. These limitations motivated a process change rather than simply further adjustment of the conventional resist-mask approach.
How the revised EUV mask fabrication process works
The revised approach changed the mask blank structure and the pattern-transfer sequence. It used thinner, lower-sensitivity positive-tone chemically amplified resist (p-CAR) to support fine patterning, together with a chromium-based hard mask for absorber etching. In this arrangement, the resist defines the pattern in the hard mask; the hard mask then serves as the patterning layer during absorber etching.
The process sequence was:
- Electron-beam writing: The resist was patterned using an electron-beam writer.
- Hard-mask etching: The written pattern was transferred into the chromium-based hard mask.
- Resist removal and absorber etching: After resist removal, the absorber was etched using the hard mask.
- Hard-mask removal: The hard mask was removed using a developed process designed to avoid damage to the ruthenium (Ru) layer.
Etching conditions were optimized for the new blank because conventional dry-etch processes were not suitable for its structure. This detail underscores that changing mask materials or layer thicknesses can require coordinated development across the patterning and etch steps, rather than an isolated resist adjustment.
Resolution and CD results for primary features and SRAFs
Top-view and cross-section scanning electron microscope images showed that the absorber patterns met their target sizes, including dense lines, isolated spaces, and isolated lines. The reported resolution reached 30 nm for SRAF patterning, with the process demonstrating the stated 24 nm space and 30 nm line dimensions.
For primary-feature CD linearity, the measured result was 2.2 nm, meeting the study’s target of 2.8 nm or less for a 44 nm primary feature. CD uniformity was measured across a 104 mm by 132 mm mask area as a three-sigma deviation. Both hp64 nm and hp32 nm dense-line patterns measured 2.2 nm, as did the evaluated 24 nm isolated-space and 30 nm isolated-line SRAFs. The results indicated similar CD uniformity across the tested pattern sizes and types.
That measured uniformity was above the stated target of 1.8 nm or less. The authors estimated that global CD uniformity correction on the electron-beam writer could bring the result below 1.8 nm, but the reported measurements were obtained without that correction. For manufacturing decisions, this distinction matters: a projected outcome after correction should not be confused with an achieved measurement.
Lower LWR, with a remaining target to address
The revised process reduced LWR for primary features from approximately 5.5 nm with the conventional process to approximately 3.1 nm. This represents an improvement of more than 40%, and the study found no dependence on pattern size for primary-feature LWR in the new process. LWR also improved for isolated SRAF lines, while isolated-space LWR was reported as approximately the same as that of the primary feature.
The LWR measurement used averaged line widths from 40 nm segments and was defined as three times the standard deviation of those widths. The study also explained why spatial frequency matters: for the 0.33 NA EUV system considered, lower-frequency mask roughness can transfer more strongly to the wafer, while higher-frequency roughness is suppressed by the scanner’s numerical aperture. Mask LWR therefore needs to be evaluated with its potential wafer impact in mind.
Despite the substantial improvement, the new process’s primary-feature LWR of about 3.1 nm did not meet the 2.8 nm-or-less target. Further assessment was identified as necessary to determine the wafer impact and whether additional roughness reduction was required.
Manufacturing implications
The study demonstrates how a thinner resist and hard-mask transfer strategy can extend EUV absorber patterning to small SRAFs while improving roughness relative to a conventional process. It also shows the importance of reviewing resolution, CD linearity, CD uniformity, and LWR independently. Meeting a feature-size target does not automatically mean that uniformity or roughness requirements have been met.
For mask process development, the practical takeaway is to validate the complete pattern-transfer flow on the intended blank structure, include the effect of any planned correction method in CD-uniformity qualification, and assess roughness in relation to its potential wafer transfer. The reported work established a path to 30 nm SRAF lines and 24 nm spaces, but left primary-feature LWR and defect evaluation as areas for further study before the process could be considered fully qualified for high-volume production.