Design automation methodology and rf/analog modeling for rf CMOS and SiGe BiCMOS technologies

IBM Journal of Research and Development, Mar/May 2003 by Harame, David L, Newton, Kim M, Singh, Raminderpal, Sweeney, Susan L, Et al

2. Predictive modeling

A fundamental difference between digital technology development and analog/rf technology development is the sensitivity of analog/rf circuits to many manufacturability/performance tradeoffs that must be made. Thus, while feedback between IC designers and technologists during technology definition is critical for timely product development, it is difficult to realize in practice, because to an IC designer the technology is the design kit-a complete and accurate set of compact models for both active and passive devices as well as interconnects. It is expected to reflect a stable and well-characterized process, while to the technologist the technology is a process recipe that is roughly characterized, often in flux, and, in practice, often not actually realized until the final technology qualification stage. The role of predictive modeling is to utilize detailed process and device simulation, or technology CAD (TCAD), in place of hardware to facilitate the feedback loop between circuit designer and process technologist until definitive hardware data is finally incorporated into the design kit. This feedback path provides timely notification to the technologist of potential shortcomings in the targeted technology design point, thus optimizing the use of the available experimental wafer budget. For the circuit designer, it provides more design turns with the technology and thus a greater likelihood of meeting circuit performance targets.

Technology CAD

Semiconductor TCAD originated in the early 1960s [7] with efforts to understand and optimize bipolar transistors. This effort continues today, with the increases in computing power available leveraged to understand and engineer devices with higher operating speeds and fabricated with more complex processes. The TCAD paradigm is applied to all conceivable types of active and passive devices, and intensive TCAD studies are now part of all semiconductor technology development efforts.

The TCAD paradigm can be described as follows: Detailed process simulation creates one-, two-, or threedimensional device representations, consisting of structural (film thicknesses and shapes) and impurity concentrations used as input for device simulation. Device simulation produces the dc and ac characteristics of interest, which are in turn used to define compact models for use in a prototype design kit for the technology. Thus, process options can receive circuit-design feedback before expending the time and budget to define these options in silicon. While TCAD is typically used to assist technology development, it is leveraged to its fullest extent when combined with compact model development to provide early technology access for circuit design.

Process simulation

Physical process simulation is the critical component in a predictive TCAD capability. Research and development of existing process simulation capabilities are due to the last decade's worth of logic CMOS scaling. Ever more sophisticated process simulation capabilities are being developed as semiconductor processing capabilities, driven by an extremely competitive microelectronics industry, continually progress. However, despite intensive efforts to bring higher-level modeling capabilities such as molecular dynamics and kinetic Monte Carlo codes into practical use, continuum codes based on silicon process physics are still the primary platform for semiconductor process simulation and are focused on here. The critical silicon process operations are impurity implantation and diffusion, oxidation, and material deposition and etching. Silicon and silicon-germanium epitaxy, an increasingly critical silicon process step, is treated via a series of material depositions and diffusions.