As efficiency is one of the bottlenecks of device simulation, we propose to employ deep neural networks to generate two-dimensional electrostatic potential profiles for efficiency. Supervising with previously obtained simulation results for various BJT devices, we train deep neural networks to generate an electrostatic potential profile as an initial guess for a non-equilibrium condition with estimating carrier densities by the frozen field simulation. With the generated potential profiles, we significantly reduce the number of Newton iterations without loss of accuracy.
|Title of host publication||2020 International Conference on Simulation of Semiconductor Processes and Devices, SISPAD 2020|
|Publisher||Institute of Electrical and Electronics Engineers Inc.|
|Number of pages||4|
|Publication status||Published - 2020 Sep 23|
|Event||2020 International Conference on Simulation of Semiconductor Processes and Devices, SISPAD 2020 - Virtual, Kobe, Japan|
Duration: 2020 Sep 3 → 2020 Oct 6
|Name||International Conference on Simulation of Semiconductor Processes and Devices, SISPAD|
|Conference||2020 International Conference on Simulation of Semiconductor Processes and Devices, SISPAD 2020|
|Period||20/9/3 → 20/10/6|
Bibliographical noteFunding Information:
ACKNOWLEDGEMENT This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (NRF-2019R1A2C1086656). This work was also supported by Institute for Information & communications Technology Promotion(IITP) grant funded by the Korea government(MSIT) (No.2019-0-01351, Development of Ultra Low-Power Mobile Deep Learning Semiconductor With Compression/Decompression of Activation/Kernel Data).
© 2020 The Japan Society of Applied Physics.
All Science Journal Classification (ASJC) codes
- Electrical and Electronic Engineering
- Computer Science Applications
- Modelling and Simulation