Bozidar Novakovic

Objective I:

• Improve initial guess for the first bias point by making it more accurate while still fast.

Approach:

• Selfconsistent solution of Poisson equation with semiclassical density with 'confined' effective mass (top figure).

Results:

• 'Confined' effective mass brings semiclassical guess closer to final quantum solution (bottom figure).

Impact:

• Fast and accurate initial guess that reduces the probability of divergence.

Acknowledgments: useful discussion Z. Jiang, leveraging code infrastructure by S. Steiger.

Si NW GAA MOSFET, middle line plot, quantum (blue) in sp3d5s*.

Objective II:

• Poisson/transport equations convergence in as few iterations as possible.

Approach:

• Improved semiclassical initial guess for 1st bias.
• Extrapolate initial guess for higher bias.
• Newton root-finding method with full step size (since divergence is unlikely).

Results:

• ~2x reduction in number of iterations.

Impact:

• ~2x code speedup. As a basic algorithm used in many works, for example:
1. Journal of Computational Electronics, Dec. 2013, vol. 12, iss. 4, pp 592-600. doi:10.1007/s10825-013-0509-0
2. IEEE Electron Device Lett., vol. 35, pp. 1212-1214, Dec. 2014. doi:10.1109/LED.2014.2364593
3. IEEE Journal on Exploratory Solid-State Computational Devices and Circuits, vol. 1, pp. 12-18, Apr. 2015. doi:10.1109/JXCDC.2015.2423096

Acknowledgments: leveraging code infrastructure by M. Povolotskyi.

Si NW GAA MOSFET in sp3d5s* empirical tight-binding basis.