EUV Mask Defects: What Can We Do About Them?
by Abbas Rastegar, SEMATECH Fellow
Although mask makers have made masks for decades, reducing their defectivity remains a challenge. Why? The simple answer is their reflectivity and the process of making EUV mirrors.
Almost all known materials will partially absorb EUV light (λ = 13.5 nm); therefore, a single-layer reflective mirror cannot be used to reflect EUV light. Instead, Bragg multilayer mirrors with 40 to 50 bilayers of Mo and Si and a peak reflectivity between 64% to 72% at a centroid wavelength of λ = 13.5 nm are employed. Although building an EUV mirror with reasonably high reflectance is not a challenge, building one that meets the surface flatness, roughness, and defectivity required for the 16 nm half-pitch and beyond is proving difficult.
As early as 2000, SEMATECH recognized the importance of building defect-free EUV mask blanks. Consequently, in 2003, it launched the EUV Mask Blank Development Center (MBDC) in Albany, New York, with the mission to determine the risk associated with manufacturing EUV mask blanks and to enable commercial mask blank suppliers to provide defect-free EUV mask blanks. Much of the current knowledge about EUV defects, including their formation, printability, characterization, and removability, was derived from work by scientists and engineers at SEMATECH. In joint programs with different suppliers, SEMATECH developed many prototype tools including the very first EUV mask blank inspection tool, EUV actinic inspection tool (at LBNL), and EUV micro exposure tools. In addition to building expertise and infrastructure, SEMATECH began working with EUV substrate and blank suppliers (e.g., Corning, Asahi Glass Co, HOYA, and Schott Lithotec) to ensure the availability of low thermal expansion material (LTEM) glass substrates with the required surface quality. Despite these efforts and investments, the defectivity of EUV mask blanks has remained a challenge.
Achieving zero defect density on the whole quality area of the mask (142×142 mm2) is the greatest hurdle. Determining the critical defect size for an EUV mask is more complex than for its optical counterparts. Figure 1 shows a schematic of the mask defects that are a combination of EUV substrate, blank, and absorber patterning defects. Defects both on top of as well as inside and under the multilayer are important as they alter the reflective properties of the multilayer film. During EUV mask blank fabrication, the size and shape of particles and pit (void) defects on the substrate change when multilayers are deposited. This change, commonly known as defect decoration, depends on the MoSi deposition conditions and the number of bilayers. In addition to substrate defects, particles are added during (i.e., in-film particles) or after the multilayer deposition (top particles). Finally, defects are also added during absorber layer deposition and patterning. A defect becomes critical if it can be printed on the wafer. Defect printability depends on many parameters including defect size, shape, proximity to an absorber line, coherence of the EUV source, and the illumination angle and focus condition of the exposure tool. For defects on the top or close to the top of the multilayer, defect composition is also crucial. See Figure (below).
Many substrate defects including scratches, pits, and embedded particles are created by chemical mechanical polishing (CMP) of the mask substrates. CMP processes are optimized to achieve a surface roughness of 0.08 nm RMS (note that this is close to the Bohr radius of hydrogen atoms at 0.053 nm) and surface flatness of 50 nm peak-to-valley on the whole quality area of the mask.
Because substrates are square with no rotational symmetry, the distribution of the applied load force during CMP is not uniform across the area of the mask. Glass is amorphous with a non-crystalline structure; therefore, it lacks a crystalline direction by which to obtain an atomically well arranged surface. Since substrates are thicker (6.4 mm) than wafers (~1.2 mm), controlling the surface profile by applying back pressure during CMP is more difficult. Hence, CMP of EUV mask substrates is totally different from wafer CMP. Few have the proper expertise to achieve these challenging surface requirements. Currently, EUV mask substrate CMP is more an art than a science with a dearth of artists worldwide.
In addition, since the market for masks is considerably less than the market for wafers, few CMP tool manufacturers offer tools for masks. Moreover, mask CMP tools tend to be less flexible than wafer CMP tools. This complexity and lack of proper resources have slowed the rate of progress in this field.
Besides meeting surface quality requirements, CMP processes must be optimized for defect reduction. Slurry agglomeration, pad material and quality, particles in chemicals and DI water, and CMP process parameters all need modifications.
Post-CMP and subsequent cleans will also contribute to t pit formation on the substrate surface. Common brush cleaning to remove slurry particles after CMP creates scratches. Megasonic cleaning creates nano-scale pits. Because pit defects currently make up ~85% of the EUV mask blank defects, SEMATECH is spearheading extensive research in reducing these defects.
Pit defects are usually created on the surface by cavitation collapse near the surface. Hence, part of SEMATECH’s research is focused on controlling cavitation by controlling acoustic power, liquid gas content, and other megasonic cleaning parameters to prevent transient cavitation. Other research centers on material parameters such as the bulk and surface hardness of LTEM substrates. Under similar megasonic cleaning conditions LTEM substrates are more prone to pit formation than quartz substrates. We have previously shown that on the very top surface (~100 nm thick) of substrates, the surface layers for quartz and LTEM are different and their thickness depends on the polishing processes. Although the compositions of these layers have been determined, less is known about their molecular structure. In particular, we are interested in determining why quartz surfaces are more resistant to pit formation than LTEM even though LTEMs have a composition similar to quartz ( i.e., current LTEM substrates are Ti-doped synthetic quartz).
Along with preventing pit formation, SEMATECH is investigating pit mitigation techniques such as pit smoothing or pattern shifts. Pit smoothing, changes the aspect ratio (height/width) of a pit from a higher to lower value. The relationship of the pit aspect ratio before and after smoothing, called the smoothing power, is indicative of the capability of a certain technique. To be successful, a smoothing technique should not only have high smoothing power but also not add non-removable defects to the surface. Although SEMATECH has invested extensive effort into different smoothing methods, none of the high smoothing power techniques has yet resulted in low defectivity substrates. On the other hand, techniques with less smoothing power, such as a SEMATECH-developed non-isotropic etch/clean smoothing process and oblique deposition of multilayers, have enjoyed relative success. Further research using spin-on materials for smoothing is in progress.
The pattern shift method is currently used by mask makers to build EUV masks. In this technique, the locations of defects are determined by inspection of the EUV blank so that the absorber pattern can be positioned to cover as many of the blank defects possible. To ensure success, an actinic blank inspection tool is needed to determine the location of printable defects. SEMATECH initiated the EUV mask infrastructure (EMI) organization to bring the industry together to collaborate on ensuring that EUV blank inspection tools and an actinic defect review tool (AIMSTM ) will be developed in time for the 16 nm half-pitch node. Currently, all key tools needed for EUV mask manufacturing are being developed by different suppliers.
For the pattern shift technique to work, information about the exact location of defects is crucial. Errors in finding the defect location should be less than 30 nm; however, this is a daunting and expensive challenge since the inspection tools require costly substrate stages.
Blank defects resulting from the multilayer deposition likewise pose problems. Currently wide ion beam deposition (IBD) is commonly used to deposit Mo, Si and Ru layers. The main sources of defects during such deposition include the ion source, sputtering from shields and/or targets, handling inside vacuum, or while transferring plates from atmospheric pressure to vacuum. SEMATECH is unique for its extensive program to reduce IBD defects. We have determined sources of most added particles during deposition and are actively working with different component suppliers to mitigate them.
The challenges of removing particles on top of EUV masks and those resulting from patterning are similar to the challenges of removing particles from optical masks. In some respects, EUV patterned masks are less prone to structural damage by megasonic cleaning since they lack sub-resolution assist features (SRAFs) and their absorber stacks have better mechanical properties. However, their optical properties can be seriously impacted by repetitive cleaning, which is unavoidable given their lack of pellicles.
A typical EUV mask should be cleaned at least 50 times during its effective lifetime. However, the Ru cap on the multilayer is oxidized by the cleaning chemistries, diminishing EUV reflectivity. Cleaning also adds pits, some of which are printable. According to our studies, the same cleaning processes create more pits on the surface of Ru-capped surfaces than on quartz, indicating that the Ru cap is softer than the quartz. Finally, cleaning processes can also etch the absorber lines, which will alter the critical dimensions of patterns.
In summary, despite progress in reducing EUV mask defects, many challenges must still be overcome for the successful implementation of EUV. Mask blank defectivity, especially substrate defectivity, is the main near-term hurdle. CMP and post-CMP cleaning should be optimized to reduce pits while the cleanliness of the cleaning tool and process should be improved to reduce particles. Multiple cleaning of the Ru-capped multilayer should not compromise EUV reflectivity. Finally, cleaning chemistries should be modified to prevent etching of the TaN absorber, and carbon contamination should be removed.
Biography:
Abbas Rastegar is a Fellow at SEMATECH, leading mask cleaning activities, in Albany New York. Abbas has more than 20 years of R&D experience in different physics, nano-science, and nano-technology disciplines with more than 130 published papers in technical journals and conferences.
He has Ph.D. in Physics from University of Ljubljana Slovenia in collaboration with International Center for Theoretical Physics (ICTP) in Trieste, Italy. He has a diploma in Advanced Condensed Matter theory from ICTP. Abbas received his M.Sc. in Solid State Physics and B.Sc. in Applied Physics from Ferdowsi University, Mashhad Iran.
Tags: Sematech