next up previous
Next: 3. The Physical Model Up: 2. The State-of-the-Art of Previous: 2.4.2 Silicon and

2.5 State-of-the-Art of III-V RF Device Modeling

In the silicon industry, process [31], device [134], and interconnect-simulation tools form a continuous row of tools ranging from material analysis, device analysis to chip design and even system design [293]. Since the database available for composition dependent III-V semiconductor modeling has been restricted for a long time, Monte Carlo (MC) simulation approaches for material analysis have prevailed to determine a larger number of material properties, especially for high field transport. The approach of Fischetti and Laux in [91] with some minor modifications is still widely used, though full band MC codes are also available [94,142]. III-V device simulation mainly focuses on device and circuit aspects [208]. For heterojunction devices, due to the extensive number of technological process steps, device simulation mainly is used for inverse modeling and process control, e.g. of geometry or doping. A common problem is the lack of a thorough approach to III-V semiconductor materials modeling. Material modeling of Al$ _{x}$Ga$ _{1-x}$As, In$ _{x}$Ga$ _{1-x}$As, In$ _x$Al$ _{1-x}$As, and In$ _x$Ga$ _{1-x}$P is restricted to slight modifications of GaAs or even Si material properties. Another very severe problem is the limited feedback from technological state-of-the-art process development to device simulator development. Table 2.2 gives an overview of the simulators available.


Table 2.2: Comparison of different device simulators, (DD) drift-diffusion, (ET) energy transport, (HD) hydrodynamic transport, (TE) thermionic emission, (TFE) thermionic field emission.
Simulator Dimension Model Features Remarks
POSES 1D - Schrödinger-Poisson solver -
Leeds quasi 2D HD Schrödinger equation 1D current equations
- - Electrothermal model interfaces
Fast Blaze quasi 2D HD - 1D current equations
- - - interfaces
ATLAS 2D DD,ET TE heterojunction model no tunneling
- - - modeling
SIMBA 2D,3D DD Schrödinger equation no ET (HD)
PISCES 2D DD,ET III-V models -
G-PISCES 2D DD full set III-V models no ET(HD)
MEDICI 2D DD,HD anisotropic properties mixed-mode
- - - interfaces
MINIMOS-NT 2D DD,HD full set III-V models -
- - TE/TFE model -
DESSIS 2D,3D DD,HD trap modeling, TFE model III-V modeling


The device simulator MEDICI has been used for the simulation of AlGaAs/GaAs HBTs [21] and for evaluation of expected GaN HBTs [224] properties. For the device simulator ATLAS from SILVACO [262] the simulation capability of AlGaAs/GaAs and pseudomorphic AlGaAs/InGaAs/GaAs HEMTs was announced. The two-dimensional device simulator PISCES [31] incorporates a set of III-V models and examples for GaAs-based, AlGaAs/InGaAs, InAlAs/InGaAs MESFETs and HEMTs. Further AlGaAs/GaAs, InP/InGaAs, and InGaP/GaAs HBTs are simulated. G-PISCES, a development by Gateway modeling, also demonstrated the simulation of AlGaN/GaN HEMTs [15]. The two- and three-dimensional device simulator DESSIS [134] provides an extended set of physical models for device simulation. The capabilities to model Si and SiC are extended by a heterojunction framework to III-V materials [167]. Issues such as extensive trap modeling are solved. At quantum level, a one-dimensional Schrödinger-Poisson solver POSES for charge analysis in HEMTs for process control is offered by Gateway Modeling. In the program SIMBA [220] a link between a one-dimensional Schrödinger solver and a two-dimensional Poisson solver is demonstrated. SIMBA also provided drift-diffusion transport simulation of GaN HEMTs [275]. The advance of device simulation tools further allows precise physics-based small-signal extraction, where the principal approaches were summarized in [160].

Using a one-dimensional current equation, quasi-two-dimensional approaches are demonstrated in several publications by the University of Leeds [187]. This computation time effective approach has also been verified against MC simulation for some examples for gate-lengths down to 50 nm [186].

A similar version is called Fast Blaze and presented by SILVACO. The quasi two-dimensional approach demonstrated a software interface between the quasi-two-dimensional device model and the compact Root large-signal model within the Microwave Design System (MDS and ADS). It can be combined with the Advanced Design System (ADS) delivering an interface towards the microwave circuit simulator. Even extraction with subsequent multitone excitation calculations are presented [140]. For the large-signal modeling of devices [237] showed the combination of a harmonic balance simulation and a device simulation using PISCES for a LDMOS device. For distortion analysis in a MESFET the same approach was used in MESFETs  [244]. The crucial drawback of a physics based large-signal extraction is related to the typical problems of compact large-signal models themselves. This particularly especially the accurate treatment of parasitic elements, e.g. inductances for multi-finger devices, the thermal problem, which is generically three-dimensional, and the frequency dispersion due to the fast traps in III-V semiconductors.

Various other combinations of physical representation have been proposed, such as HBT compact models [269]. The determination of thermal boundary conditions and the verification of temperature distributions are performed in agreement with three-dimensional thermal chip simulations, e.g. [311]. Simplified approaches for thermal modeling can be found in [139]. Introducing thermal modeling within a compact large-signal model, [248] provided a large-signal model for devices used in the course of this work. A global three-dimensional compact model is introduced by Batty et al. in [28].


next up previous
Next: 3. The Physical Model Up: 2. The State-of-the-Art of Previous: 2.4.2 Silicon and
Quay
2001-12-21