Christoph Lenz’s research while affiliated with TU Wien and other places

Fluorocarbon dry etching of vertical silica-based structures is essential to the fabrication of advanced complementary metal-oxide-semiconductor and dynamic random access memory devices. However, the development of etching technology is challenged by the lack of understanding of complex surface reaction mechanisms and by the intricacy of etchant flux distribution on the feature-scale. To study these effects, we present a three-dimensional, TCAD-compatible, feature-scale modeling methodology. The methodology combines a level-set topography engine, Langmuir kinetics surface reaction modeling, and a combination of reactant flux evaluation schemes. We calibrate and evaluate our model to a novel, highly selective, etching process of a SiO2$\mathrm {SiO_2}$ via and a Ru$\textrm{Ru}$ hardmask by CF4/C4F8$\mathrm {CF_4/C_4F_8}$. We adapt our surface reaction model to the novel stack of materials, and we are able to accurately reproduce the etch rates, topography, and critical dimensions of the reported experiments. Our methodology is therefore able to prototype and study novel etching processes and can be integrated into process-aware three-dimensional device simulation workflows.

The level-set method is widely used in expanding front simulations in numerous fields of computational research, such as computer graphics, physics, or microelectronics. In the latter, the level-set method is employed for topography simulations of semiconductor device fabrication processes, being driven by complicated physical and chemical models. These models tend to produce surfaces with critical points where accuracy is paramount. To efficiently increase the accuracy in regions neighboring these critical points, automatic hierarchical domain refinement is required, guided by robust feature detection. Feature detection has to be computationally efficient and sufficiently accurate to reliably detect the critical points. To that end, we present a fast parallel geometric feature detection algorithm for three-dimensional level-set functions. Our approach is based on two different, complementary curvature calculation methods of the zero level-set and an optimized feature detection parameter to detect features. For performance reasons, our algorithm can be in principal linked to different curvature calculation methods, however, as will be discussed, two particularly attractive options are available: (i) A novel extension of the standard curvature calculation method for level-set functions, and (ii) an often disregarded method for calculating the curvature due to its purported low numerical accuracy. We show, however, that the latter is still a viable option, and that our algorithm is able to reliably detect features on geometries stemming from complicated, practically relevant geometries. Our algorithm and findings are applicable to other fields of applications such as surface simplification.

Thin material layers are common structures in modern semiconductor device fabrication and are particularly necessary for light-emitting diodes and three-dimensional NAND memory devices. Such layers are not only deposited on the flat wafer surface but are also partially removed during subsequent etching steps. Level-set based process TCAD simulations are capable of representing flat thin material layers, such as those occurring after deposition, with sub-grid accuracy. However, topographical changes during etching processes modeled via Boolean operations expose the low underlying grid resolution, leading to detrimental artifacts. We present a novel algorithm that analyzes the thickness of all material layers and derives a refined target resolution for local regions of thin layers affected by the etching process. This allows to accurately represent topographical changes in thin layers by hierarchically refining the grid without unnecessary refinement in unaffected regions of the domain. We simulate the fabrication of a light-emitting diode device using our algorithm to automatically predict the optimal resolution for all etched material layers. Our algorithm selects efficient refinement factors to obtain the local target resolutions of the hierarchical grids, and achieves a three times faster computation time than a benchmark refinement algorithm based on topographical features.

We present a feature detection method for adaptive grid refinement in hierarchical grids used in process technology computer-aided design topography simulations based on the local curvature of the wafer surface. The proposed feature detection method enables high-accuracy simulations whilst significantly reducing the run-time, because the grid is only refined in areas with high curvatures. We evaluate our feature detection method by simulating selective epitaxial growth of silicon-germanium fins in narrow oxide trenches. The performance and accuracy of the simulation is assessed by comparing the results to experimental data showing good agreement.

In etching and deposition simulations of a semiconductor fabrication process the calculation of the surface rates of particles is an essential but also the computationally most demanding step. A promising approach is to preprocess the simulation domain by simplifying the surface. We thus propose a new surface mesh simplification method that takes advantage of geometric domain-specific surface properties that are prevalent in topography simulations. We compare our method to a suitable reference algorithm and show that our method maintains higher geometric accuracy and accordingly maintains the original geometry in great detail. Furthermore, the evaluation of the simplified meshes show an enhanced performance of the particle surface rate calculation.