Modeling CVD effects in Atomic Layer Deposition on the Feature Scale
Simulation of atomic layer deposition requires the coupling of complex surface chemistry to ballistic transport in geometrical structures. The calculation furthermore has to be transient with chemical process models which have transients on different time
- PDF / 565,385 Bytes
- 4 Pages / 439.37 x 666.14 pts Page_size
- 58 Downloads / 160 Views
Modeling CVD effects in Atomic Layer Deposition on the Feature Scale W. Jacobs, A. Kersch, G. Prechtl, G. Schulze Icking-Konert Infineon Technologies, Balanstrasse 73, 81541 Munich, Germany alfred.kerschCiV,infineon.com
Abstract Simulation of atomic layer deposition requires the coupling of complex surface chemistry to ballistic transport in geometrical structures. The calculation furthermore has to be transient with chemical process models which have transients on different time scales. The paper shows a solution of this problem with a Monte Carlo based model implemented in the gene ral purpose feature scale simulator TOPSI 3D . The chosen example is HiD 2 deposition with HfCl 4 / H 20 in a trench which may show a nonconformal film profile.
1 CVD Effects in Atomic Layer Deposition Conformal deposition into high aspect ratio structures is an important technological achievement enabled by atomic layer deposition (ALD). For example the suitability of ALD Aha] as node dielectric for trench capac itors with excellent conformality and thermal stability was demonstrated [1]. For HfO, ALD, however, CVD effects have been reported [2,3] which may lead to nonconformal deposition in trenches and nonuniformity over the wafer. In this paper we present a feature scale model which is able to explore a sophisticated, well calibrated chemistry model [4] on the feature scale and to identify the root cause for such effects.
2 Feature Scale Model We have developed a 3D feature scale simulator capable of treating high aspect ratios as well as arb itrary particle transport and chemical kinetics models [5].The level set front propagation module achieves high accuracy and efficiency by using a narrow band discretization and a fast marching algorithm. The local velocity is calculated on the extracted front segments from the particle fluxes and is extended to the level set grid nodes. The physical fluxes at the front are calculated using ballistic transport from user -defined particle sources. Thermally and kinetically activated surface reactions of arbitrary order are consistently calculated as a function of sampled flux and surface coverages. A new algorithm developed for proper description of ALD processes allows the calculation of transient surface coverages with the ballistic Monte-Carlo based flux model. While in conventional film growth the time discretization is determined by the maximum speed, in ALD simulations it is controlled by the rate of change of the surface coverages. After temporal evolution of the chemistry over a time step, the G. Wachutka et al. (eds.), Simulation of Semiconductor Processes and Devices 2004 © Springer-Verlag Berlin Heidelberg 2003
138 ballistic particle transport is repeated using the newly calculated surface reactivity. With this approach essentially arbitrary surface reaction mechanisms can be coupled to the particle transport. In a previous work, the propagation of the film front in a trench has been calculated with a reactive sticking coefficient chemistry model [6].
3
ALD Chemistry Model
In
Data Loading...
