Introduction

The Device Modelling Group, which is the world leader in the physical simulation of statistical variability in nano-CMOS devices, offers simulation services of statistical variability with our 3D “atomistic” statistical simulator, GARAND, to semiconductor manufacturers and chip design houses. Our simulations can accurately predict the statistical MOSFET variability at the device and technology design stage.

Fig.1a Fig.1b
(a) (b)
The GARAND simulation tool can accurately predict the statistical variability found in a modern MOSFETs. Shown (a) are the combined effect of LER and random dopants on the electron distribution in 35 nm MOSFET (GARAND) and (b) the measured electron distribution in a similar device (Fukutome, Fujitsu).

Sources of Variability

We can include individually and in combination all major sources of statistical variability in the simulations.

  • Random Discrete Dopants
  • Line Edge Rougness
  • Poly-Si Granularity
  • High-k Granularity
  • Interface Roughness

Why you should use our service

We can help you to design your next generation devices with reduced variability. Your designers will know accurately the variability in your next technology generation before the technology is ready and the variability is measured. Using our cluster and grid technology we can provide in a week results for you which, on a workstation, will take several years to run.


How the service works

  • We will introduce the structure of your device obtained from your TCAD process simulation in GARAND.
  • Using the continuous doping profile we will accurately match the results of your TCAD device simulations with GARAND simulations.
  • Using our computing clusters, we will carry out 3D simulations of a statistical sample (typically 200) of your devices. In these simulations we could include individual of combined sources of variability as specified by you.
  • According to your request we can carry out simulations for threshold voltage variability, off/on-current variability or full I-V characteristics of the statistical sample.
  • If you provide a compact model based on your continuous TCAD simulation we will extract a small set of compact model parameters that accurately capture the simulated variability.

For an example of this workflow please follow this link.


Contact

For more information please contact Professor A. Asenov.