In order to cope with explosive costs, reduced time-to-market figures, and to
deal with high competition, Technology Computer-Aided-Design (TCAD) is more
and more applied during development and production of new devices
[162]. The basic motivation for applying numerical simulation
tools is to reduce the number of test wafers, to evaluate a vast number of
device variations by numerical means and to optimize devices and processes.
Optimization of geometry, doping, materials, and material compositions targets
high output power, high breakdown voltage, high speed (high
and
),
low leakage, low noise, and low power consumption. This is a challenging task
that can be significantly supported by device simulation. While steady-state
simulations are sufficient for optimization targets such as breakdown voltages,
turn-on voltages, or leakage currents, small-signal simulations are required
for many performance and noise issues [162].
The continuously increasing computational power of the average workstation computer and available cluster computing technologies enables the large-scale application of TCAD software. Several commercial device simulators [4,42,200,20,214,111,215,216] company-developed simulators like [30,148], and university-developed simulators [116,62,207,203,104,237,105,153,38,205] have already been used for supporting device design and development. The major differences between these simulators are
On the circuit simulation level the approaches mentioned above including the various macroscopic transport models require too much computational resources, especially in terms of time and memory, if large circuits consisting of thousands of transistors have to be simulated. For that reason, circuit analysis tools like the well-known SPICE [168] simulator are based on compact models (see Section 3.6.1). They try to provide a closed form description of the electrical behavior of the devices, which results in far less equations than the approaches based on discretized partial differential equations. However, especially for advanced semiconductor devices the extraction of the various parameters is a cumbersome task. Furthermore, calibration against given experimental data is required as extrapolation from one generation to the next is rarely possible. A possible replacement of normally expensive experimental data can be simulation results from device simulations.
Tools for Technology Computer Aided Design have been essential for the development of more advanced and sophisticated devices. In the area of device simulation, the major player in the commercial market for simulators is the company Synopsis at the moment. Since the acquisition of Integrated Systems Engineering AG (ISE) in November 2004, the company provides the well-known TCAD software packages MEDICI/DAVINCI [216,214] and DESSIS [111]. Competition comes for example from Silvaco with the ATLAS framework [200] and Crosslight (formerly Beamtek Software) with the LASTIP/PICS3D/APSYS simulators for optoelectronics and advances physical modeling [42]. In addition, several general frameworks as presented in Section 4.2.1 are provided for the solution of partial differential equations, which can also be employed for semiconductor modeling and simulation.
S-PISCES of Silvaco as part of the ATLAS simulation
framework [200] calculates steady-state, small-signal AC, and
transient solutions for general non-planar two-dimensional silicon device
structures. The related simulator for compound semiconductors is BLAZE2D/3D, which provides a library including also ternary and quaternary
materials. The calculated small-signal characteristics are the cut-off
frequency
, S-, Y-, H-, and Z- parameters, the maximum available gain
(MAG), the maximum stable gain (MSG), the maximum frequency of
oscillation (
) and the stability factor.
The multi-dimensional simulator DESSIS [111] provides related features. The small-signal capabilities are incorporated in the mixed-mode, which supports electrothermal netlists with mesh-based device models and SPICE circuit models.
MEDICI of Synopsis [216] (a former Avant! product) is a two-dimensional simulator. A small-signal analysis can be performed to calculate frequency-dependent capacitances, conductances, admittances, Y-, S-, and H-parameters. DAVINCI [214] is the related three-dimensional device simulation program with a similar set of features. The approach is also based on [127], including the numerical split of real- and complex-valued part.
Applications of advanced RF devices must often be seen in a circuit related context [208]. For that reason, circuit simulation programs such as Spice [147], Agilent's ADS [4], or HSPICE of Synopsis [215] are employed. Whereas these circuit simulators are based on compact models, device simulators with distributed modeling (solving of a system of partial differential equations) of the transistors offer so-called mixed-modes. Realistic dynamic boundary conditions imposed by a circuit allow to extract circuit-related figures of merit. Although this approach is limited by performance and memory considerations, the highly sophisticated models required for today's advanced device structures can be directly employed for transient or small-signal circuit simulations [231].
It is a well-known fact, that correct steady-state modeling is an important prerequisite for any kind of subsequent simulations. Thus, the advanced simulators incorporate drift-diffusion and advanced transport models such as energy transport models [89] and provide several advanced mobility models. In addition, models for recombination, band-gap narrowing, impact ionization, band-to-band tunneling, hot carrier injection, Schottky contacts, and floating gates have to be included to account for the properties of advanced device structures.
MINIMOS-NT [105] is a general-purpose semiconductor device simulator providing steady-state, transient, and small-signal analysis of arbitrary two- and three-dimensional device structures. In addition, mixed-mode device/circuit simulation [88] is offered to embed numerically simulated devices in circuits with compact models. The devices can be connected both electrically and thermally.
The simulator deals with different complex structures and materials, such as Si, Ge, SiGe, GaAs, AlAs, InAs, GaP, InP, their alloys and non-ideal dielectrics. In combination with this comprehensive material database, the state-of-the-art set of physical models enables the user to simulate all kinds of advanced device structures, such as MOS devices of the sub-100nm technology, silicon-on-insulator devices, and heterostructures. The implemented physical models take all important physical effects such as bandgap narrowing, surface recombination, transient trap recombination, impact ionization, self-heating, and hot electron effects into account.
Besides the basic semiconductor equations [193], several different types of transport equations can be solved. Among these are the energy-transport equations which capture hot-carrier transport [209,22], a six moments transport model [85], the lattice heat-flow equation to cover thermal effects like self-heating [227]. Furthermore, various interface and boundary conditions are taken care of, which include Ohmic and Schottky contacts, thermionic field emission over and tunneling through various kinds of barriers.
MINIMOS-NT is equipped with a very powerful control language called Input-Deck Programming Language (IPL) [118]. For the specification of a simulation various keywords must be set. These keywords can contain arbitrarily nested expressions and can depend on the current simulation status. With the IPL, the user is able to customize a simulation by creating input-deck files written in plain ASCII text. MINIMOS-NT provides various default input-deck files with standard settings.
MINIMOS-NT is the successor of MINIMOS [195]. Whereas the latter is restricted to simple MOS structures, MINIMOS-NT can be employed for arbitrary device structures with unstructured grids.