|Ph.D Student||Eitan Shauly|
|Subject||A Study of the Two-Dimensional Diffusion of Dopants in CMOS|
|Department||Department of Materials Science and Engineering||Supervisor||Professor Emeritus Komem Yigal|
Simulation of semiconductor device structures requires modeling the underlying physics of point defects and impurity interactions. This work deals with the simulation of two-dimensional (2-D) impurity diffusion in CMOS silicon devices. The primary goal is the development of the Reverse Modeling concept. It is based on statistical comparison between 2-D process and device simulations and various electrical and physical measurements of sub-micron devices.
The model focused on the enhanced diffusion that occurs, after boron implantation, during oxidation of the silicon surface. This enhancement takes place due to injection of silicon interstitials from the oxidized interface into the bulk, leading to excess of interstitials that interact with the impurities. The Reverse Modeling method was used to determine the diffusion coefficient (DI), surface recombination rate of defects (KI) and the characteristics of the injecting source.
The study was initiated by tuning the process and device simulators by 1-D process modeling adjustment using results of 1-D SIMS, thickness, and sheet resistance measurements. 2-D electrical modeling for the mobility and threshold voltage was compared to sub-micron device measurements and data from the literature.
The bulk of the work included experiments performed under different process conditions namely heat treatment under N2 or dry O2 atmospheres, at different temperatures. 2-D process and device simulations were executed using SUPREM-IV and MEDICI, respectively. In some cases, comparison was made with SIMS results of similar experiments performed on non-patterned samples.
Analysis showed similarity between DI in 2-D system compared with the value obtained from non-patterned samples. The diffusivity DI in EPI samples was found to be faster by a factor of 3 compared to CZ samples. This was explained by interactions of interstitials with traps located in the bulk. The activation energy for KI was calculated. The dependence of KI on interface type, was explained by stress induced from the upper layers, into the silicon substrate. Experiment showed that KI of samples having a strained Si3N4/Si layer, was half an order of magnitude lower compared to samples with SiO2/Si interface. Finally, the equilibrium concentration of interstitials and the parameters that control the generation rate of interstitials were calculated. This gave the ability to present a full set of parameters needed for simulation of 2-D diffusion of boron in silicon.