Kinetic simulation tools for nano-scale semiconductor devices

Alex Fedoseyev ( yuiv-AT-ibmh-DOT-msk-DOT-su.gif ), V. Kolobov, R. Arslanbekov, A. Przekwas
CFD Research Corporation, 215 Wynn Drive, Huntsville, Alabama 35805, USA

The progress in LSI technologies has resulted in scaling down semiconductor devices to nano-scale dimensions where kinetic and quantum effects in the carrier transport become crucial for the device performance. Designing an efficient simulation tool for quantum devices involves a trade-off between efficiency and accuracy.

We are developing a novel simulation tool for design of quantum devices based on deterministic solution of the classical Boltzmann equation with quantum corrections proposed in Ref [1]. Using two-term spherical harmonic expansion in velocity space reduces the Boltzmann equation for electrons to a Fokker-Planck equation in a four-dimensional space (3 spatial and 1 energy dimensions). The Fokker-Planck equation is coupled to the Poisson and hole-continuity equations as described in Ref. [2].

We shall describe the details of numerical implementation of the 4-dimensional Fokker-Planck solver using kinetic and total energy domains. The effect of quantum corrections on spatial distribution of carriers will be analyzed. This solver provides a substantially lower computational cost than the Monte Carlo method, and resolves all points in phase space with equal accuracy. We will present simulation results for an ultra-small MOSFET.

We work on software that will provide a practical simulation tool for nano-scale semiconductor devices. Two numerical methods are employed, the finite volume technique with time-splitting factorization of spatial and energy space transport, and a meshless multiquadric technique [3].

References:
  1. H. Tsuchiya and T. Miyoshi, IEICE Trans. Electron E82-C, (1999) 880.
  2. W. Liang, N. Goldsmann, I. Meyergoyz, P.J. Oldiges, IEEE Trans. Electron Devices, 44, (1997) 257.
  3. E.J. Kansa, Comput. Math. Applic., 19, No.8/9 (1990), 147-161.
Figure 1. I-V curves for 2.5 nm GaAs/AlGaAs single tunneling barrier (2.5 nm 0.22V single barrier in 40 nm GaAs/AlGaAs device, T = 300K, doping density 1018 cm-3 , bias up to 0.35 V). Classical and quantum BTE solutions with CFD-ACE+ compare well with ones from [1](upper diagram). Electric field: classical and one with quantum correction for bias 0.2 V (lower diagram).