The work on photoresist fundamentals aims to tackle generic material related issues in advanced photolithography. Many of these are applicable to multiple imaging wavelengths. The goal of the work is to increase the understanding on the root causes of these issues and demonstrate proof of concept of innovative solutions or workarounds. This work is done in close collaboration with the technology specific (double patterning and EUV) lithography sub-programs.
Line width roughness
In lithography, line width roughness (LWR) does not scale with decreasing feature size. As a consequence the impact of line width roughness on the overall non-uniformity budget is increasing. It is unclear how the LWR specifications in the international technology roadmap for semiconductors can be met for the (sub-)22nm node C21080. In order to control this issue, it is important to understand the factors in the lithography process that are the major contributors for LWR. In order to tackle these issues, first of all a method for reliable measurement of LWR has been developed. With our metrology approach a long term LWR repeatability of better than 0.3nm 3σ has been demonstrated C18983. On top of this, our method is based on off-line analysis of scanning electron microscope images and has the ability to not only determine the magnitude of LWR, but also its frequency distribution. The contribution of LWR in the absorber pattern on reticle to the overall LWR on wafer has been estimated. This is done by determining how programmed LWR is transferred in the high k1 EUV lithography process. It was found that low frequency roughness (below the cut-off of the optical system) on mask is transferred into low frequency roughness on wafer. Intriguingly, high frequency roughness on mask is also transferred into low frequency roughness on wafer because of degradation of the aerial image C21467.
Figure 1: Power spectral density analysis of LWR for the same resist under different imaging conditions. The area under the graph is proportional to LWR2. Improving the image quality by going from annular to dipole illumination improves roughness over the entire frequency range.
Another option for meeting the LWR target is to work with the LWR that is currently available and use a dedicated treatment of the resist after (or during) development to smooth the pattern. Several approaches have been proposed and are available for demonstration purposes. From a logistical point of view it is most attractive to do such a smoothing process either immediately after the litho step in the track C20388 or immediately before to the etch process C20626. However, also other approaches are of interest since they may offer unique capabilities. A detailed evaluation of the state of the art of various of such smoothing techniques has been made P20475.
EUV resist modeling
Even though the resolution limit of chemically amplified photoresists has greatly improved over the past years, the expected resolution based on the optical systems is still not achieved in resist C21417. With increasing numerical aperture of EUV optics and implementation of off-axis illumination, this gap in resist resolution limits will only increase if resists do not further improve. Quantification of this gap is quite difficult, since it requires a good description of the imperfections of the entire imaging system. In projection optics this is a very complex exercise, but in EUV interference lithography the optical path is much simpler. Based on this imaging technique a method has been proposed to quantify the resist induced contrast loss as a function of resolution C20385. For chemically amplified resists, typically there is virtually no resist induced contrast loss in large pitch gratings. As the pitch decreases the resist induced contrast loss starts to increase. These curves are easily described by a single-parameter model where the blur parameter is believed to be closely related to the acid diffusion length of the chemically amplified resist.
Combination of experimental data with a full physical resist model is a more intensive, but very powerful method for gaining better understanding on the limiting parameters for resolution. Moreover, this approach allows to predict what level of improvement needs to be achieved for certain parameters. The model has been built based on exposing a single resist formulation at three imaging wavelengths: 13.5nm (EUV), 193nm (ArF) and 248nm (KrF). The model has been fitted such that acid generation kinetics are allowed to vary across the imaging wavelengths, but deprotection kinetics are required to be identical C18973. These boundary conditions severely restrict the optimization space and increase the confidence in the resulting physical parameters. In line with expectations the quantum yield of acid formation is identical at KrF and ArF wavelength, but 8X higher at EUV C20316. When EUV sensitizer is added, the quantum yield is further increased to 12X higher compared to KrF. Further analysis of the data indicates that the acid diffusion length is limited to 8nm or less. These results indicate that the intrinsic resolution of the resist material is no longer the limiting factor, but that the features cannot be defined during the development process P20838. In view of these results efforts have now moved to get a more detailed description of the development process. For this purpose a development rate monitor has been installed and methods have been established to obtain the development rate as a function of deprotection level and film depth of an EUV resist. These results will be ported into the full resist models C21603.
Figure 2: Experimentally determined development rate as a function of polymer deprotection and film depth for an acrylate (left) and hybrid (right) EUV resist.
Alternative resists and development processes
Resolution of EUV resist materials at sub-30nm half pitch is very challenging. For many polymeric materials pattern collapse is a resolution limiting factor ate these dimensions. Tetra-butyl ammonium hydroxide has been proposed as an alternative developer material for pattern collapse mitigation. This material appears to be effective for collapse reduction of acrylate-based resist materials, but shows no improvement for hybrid materials C20871. Towards 22nm half pitch and below few instruments currently allow screening of materials. Among them EUV interference lithography is a particularly interesting approach, since the image quality from this technique has been proven down to very small dimensions C21615. At these dimension both chemically amplified and non-chemically amplified approaches are currently explored.
Figure 3: Pattern collapse seriously limits resolution at sub-30nm half pitch. TBAH developer allows to mitigate pattern collapse when compared to conventional TMAH developer.
The work on EUV resist outgassing has primarily focused on having the infrastructure available for qualification of resist materials for exposure on the EUV scanner (see section EUV lithography). Besides this also effort has been done to get better understanding into the material components that are of primary importance for EUV optics contamination. This work is important in order to allow intelligent design of resist formulations that cause minimal optics contamination. It has been shown that the protecting groups contribute significantly to outgassing, but this results in minimal optics contamination C19009. Decomposition products of the photo-acid generator (PAG) do cause significant reflectance loss of the optics. The chemical composition of the PAG does play an important role for contamination. Very high and very low molecular weight PAGs typically result in little contamination, but medium size materials contaminate much more C20877.
Polysulfones have been identified as 'super-outgassers'. In collaboration with the University of Queensland a series of polysulfones has been defined and synthesized where the outgassing species are varied under controlled conditions. In these tests the impact of halogen atoms, silicon containing organics and aromatics on the contamination growth has been studied. Especially the silicon containing materials prove to be a high risk for optics contamination C21607.top