We present a GA (Genetic Algorithm) application for process/device model calibration. Although Technology CAD (TCAD) is already well-established as an indispensable tool for minimizing the development time and costs involved in new LSI processes and devices, maximum TCAD utilization is heavily dependant on the calibration of relevant model parameters. With the increasing complexity of TCAD models, however, calibration becomes more and more labor and time intensive. To overcome this difficulty, we propose a distributed GA-based calibration technique combined with the traditional local optimization algorithm, which reduces calibration time considerably. Experimental results on a PC cluster system (eight PCs) demonstrate that the calibration of 144 parameters can be completed within a few minutes, although this would typically take a human expert a few days. GA can, therefore, be a practical and robust tool for process/device calibration.
|
|
|
|
Calibrated profiles by our method for an ion implantation model;X-axis represents depth (um) from the surface of Silicon, Y-axis represents Boron concentration (cm-3), and implantation energies are from 2keV to 60keV.
|
|
| Relational Information |
|
|
AIST Today Vol. 4, No.10 (2004) 21
|
No.15 Winter 2005
back
> 
