DSA Defects Continue Downward Trend

Recent research provides reason for optimism and concern about directed self-assembly.


As previously discussed, the majority of defects in early directed self-assembly (DSA) processes were due to particles and other contaminants, and could be attributed to the immaturity of the process and materials. As manufacturers consider whether to incorporate DSA into specific technology nodes, they need to assure themselves that production-worthy yields can be achieved. Recent research at IMEC, presented at the SPIE Advanced Lithography conference in February, offers reasons for both optimism and concern.

The IMEC group focused on the organization’s chemoepitaxial Liu-Nealey (LiNe) process flow. Defect source analysis traced the majority of defects to the neutral cross-linked polystyrene (XPS) layer that serves as the guide material for a PS-b-PMMA block co-polymer (BCP). Most of the observed defects were “bright spots,” due to incompatibility between the XPS layer and the underlying SiN surface. These can be mitigated by plasma or other surface treatments. Of the remaining defects, many were due to residues from the etch and resist strip processes, and therefore not specific to the DSA process.

Some DSA-specific defects did appear, however. The group found both dislocations and bridges between adjacent lines after the BCP segregation anneal step. Both types of defects, a design-of-experiments study found, were strongly dependent on the CD of the guide pattern. This behavior is to be expected: each BCP material has a characteristic domain size depending on the molecular weights of the co-polymers. Ideal behavior is seen when the guide patterns leave space for integer multiples of this domain size; CD variation can make the guide pattern too large or too small to accommodate the desired number of domains. Unfortunately, though both bridge and dislocation defects depend on the guide pattern CD, they follow opposite trends: the process conditions that minimize dislocations maximize bridges, and vice versa. As the researchers explained, this example shows how critical co-optimization of the materials and the process flow is for successful DSA integration. The surface energy and segregation behavior of the BCP can be adjusted through the polymer chemistry, independent of the process parameters. Conversely, both material properties and process parameters must be strictly controlled to achieve a stable DSA process.

DSA defects also depended on the BCP anneal time and temperature. Again, this behavior is expected: segregation of the co-polymer must be allowed to reach an equilibrium state. Increasing the temperature allows the system to reach equilibrium more quickly. It is here that we find cause for both optimism and concern. The IMEC golden process as of January 2014 had 200 defects/cm², while the process reported in February achieved a much more robust 24 defects/cm². Unfortunately, the segregation anneal still required 2.5 hours at 255ºC to achieve that defect level. The good news is that earlier processes required anneals of twelve hours or more. The bad news is that production-worthy lithography steps require much shorter cycle times.

Overall, the study found no insurmountable obstacles to production use of DSA. As the process and materials continue to mature, the defect levels discussed here should continue to improve. Though lengthy anneal times raise important cost issues, remember that alternatives like EUV and quadruple patterning are also very expensive.