
Biography
Alexander Toifl was born in St. Pölten, Austria, in 1993. He studied at the TU Wien where he received the BSc. degree in Electrical Engineering (2016) and the Diplomingenieur degree in Microelectronics and Photonics (2018). He joined the Institute for Microelectronics in August 2018 as a research assistant. Alex's scientific interests include computational modeling (postimplantation annealing of GaN and SiC) and high performance numerical approaches (nonplanar epitaxy) for process TCAD.
LevelSet Based Anisotropic Etching Simulations
Anisotropic etching of silicon (Si) with wet etchants (e.g. potassium hydroxide (KOH) and tetramethylammonium hydroxide (TMAH)) is an important semiconductor processing technique that utilizes the crystalline nature of the material. While KOH etching is known mainly for its application in the production of microelectromechanical systems, the technique plays an important role for embedded silicon germanium (eSiGe) in source/drain (S/D) engineered sub28 nm node metaloxidesemiconductor fieldeffect transistors (MOSFETs). By employing a combination of dry and wet etching, an S/D cavity formed by the characteristic {111}planes can be produced. The exact geometry of the resulting sigmashaped cavity determines the uniaxial strain in the MOSFET channel after epitaxial growth of SiGe. Thus, it is very important to control the critical design variables (i.e. tip depth, channelcavity distance and cavity depth).
In order to optimize the fabrication steps for sigmashaped cavities, process technology computeraided design (TCAD) is very valuable. We employ the levelset method, where the wafer surface is described by the zero levelset of the function and the time evolution is determined by the levelset equation. The levelset equation assumes the form of a HamiltonianJacobi equation, which critically depends on a speed function defined by the highly anisotropic etch rates. The speed function is constructed using a linear interpolation, which defines a spatially strongly varying function and thus leads to a numerically demanding nonconvex Hamiltonian. In order to enable stable and physically relevant numerical solutions, we employ a novel dissipative local LaxFriedrichs scheme, which is based on information about the local geometry and the nature of the speed function (Fig. 1).
The dissipation scheme allows us to simulate the sigmacavity by Qin et al. (Microelectron. Eng., vol. 181, pp. 22–28, 2017), which employs a twostep reactive ion etching step and a sequential wet etching step (Fig. 2). The characteristically small etch rate of {111} planes gives rise to a final profile consisting of two {111} planes that define a sharp corner at a certain position relative to the channel (sigmacavity tip). The associated design parameters, tip depth and channelcavity distance (Fig 3.), are accurately predicted by the dissipation scheme. Furthermore, we compare our scheme with an elemental approach, which only takes the local speed function and levelset normal into account. The elementary approach results in insufficient dissipation beneath the spacer and gate stack, which results in an artificially high undercut rate. In contrast, the proposed scheme reproduces the physical undercut and thus is able to correctly predict the geometry of the sigmashaped cavity.
Fig. 1: The proposed dissipation scheme is based on normals and the associated speed functions, which are calculated for a stencil consisting of the central grid point and its immediate neighbors. The inset visualizes the highly anisotropic etch rate distribution associated with TMAH.
Fig. 2: The temporal evolution of the etch profile. During the selflimiting wet etch, the slowmoving {111} planes form the sigmashaped cavity.
Fig. 3: The etch profiles presented by Qin et al. are accurately reproduced using the proposed scheme. In contrast, the elementary scheme starting from the same dry etch profile results in artificially high undercut.