Physics equation-based semiconductor device modeling is accurate but time and money consuming.The need for studying new material and devices is increasing so that there has to be an efficient and accurate device model...Physics equation-based semiconductor device modeling is accurate but time and money consuming.The need for studying new material and devices is increasing so that there has to be an efficient and accurate device modeling method. In this paper, two methods based on multivariate rational regression（MRR） for device modeling are proposed. They are single-pole MRR and double-pole MRR. The two MRR methods are proved to be powerful in nonlinear curve fitting and have good numerical stability. Two methods are compared with OLS and LASSO by fitting the SMIC 40 nm MOS-FET I–V characteristic curve and the normalized mean square error of Single-pole MRR is 3.02 × 10^-8 which is 4 magnitudes less than an ordinary least square. The I–V characteristics of CNT-FET and performance indicators（noise factor, gain, power） of a low noise amplifier are also modeled by using MRR methods. The results show MRR methods are very powerful methods for semiconductor device modeling and have a strong nonlinear curve fitting ability.展开更多
The optical loss in the bent region is one of the key features for bent-waveguide superluminescent diodes that affects the device performance greatly under some conditions. For the purpose of device fabrication and op...The optical loss in the bent region is one of the key features for bent-waveguide superluminescent diodes that affects the device performance greatly under some conditions. For the purpose of device fabrication and optimization, it will be helpful if this bend loss can be estimated. In this letter, we have derived an analytical formula which can be used to get the bend-loss coefficient by fitting the P-I curves of the devices. It is proved that the formula is successful in estimating the loss coefficients from the P-I curves simulated from a complicated quantum-dot device model. We expect this method could also be valid in estimating bend losses of actual devices.展开更多
This work presents a self-consistent two-dimensional(2-D) simulation method with unified physical models for different operation regimes of charge trapping memory. The simulation carefully takes into consideration the...This work presents a self-consistent two-dimensional(2-D) simulation method with unified physical models for different operation regimes of charge trapping memory. The simulation carefully takes into consideration the tunneling process, charge trapping/de-trapping mechanisms, and 2-D drift-diffusion transport within the storage layer. A string of three memory cells has been simulated and evaluated for different gate stack compositions and temperatures. The simulator is able to describe the charge transport behavior along bitline and tunneling directions under different operations. Good agreement has been made with experimental data,which hence validates the implemented physical models and altogether confirms the simulation as a valuable tool for evaluating the characteristics of three-dimensional NAND flash memory.展开更多
文摘Physics equation-based semiconductor device modeling is accurate but time and money consuming.The need for studying new material and devices is increasing so that there has to be an efficient and accurate device modeling method. In this paper, two methods based on multivariate rational regression（MRR） for device modeling are proposed. They are single-pole MRR and double-pole MRR. The two MRR methods are proved to be powerful in nonlinear curve fitting and have good numerical stability. Two methods are compared with OLS and LASSO by fitting the SMIC 40 nm MOS-FET I–V characteristic curve and the normalized mean square error of Single-pole MRR is 3.02 × 10^-8 which is 4 magnitudes less than an ordinary least square. The I–V characteristics of CNT-FET and performance indicators（noise factor, gain, power） of a low noise amplifier are also modeled by using MRR methods. The results show MRR methods are very powerful methods for semiconductor device modeling and have a strong nonlinear curve fitting ability.
基金Project supported by the National Natural Science Foundation of China(Nos.61274072,60976057)
文摘The optical loss in the bent region is one of the key features for bent-waveguide superluminescent diodes that affects the device performance greatly under some conditions. For the purpose of device fabrication and optimization, it will be helpful if this bend loss can be estimated. In this letter, we have derived an analytical formula which can be used to get the bend-loss coefficient by fitting the P-I curves of the devices. It is proved that the formula is successful in estimating the loss coefficients from the P-I curves simulated from a complicated quantum-dot device model. We expect this method could also be valid in estimating bend losses of actual devices.
基金supported by National Natural Science Foundation of China (Grant No. 91230107)National Basic Research Program of China (973) (Grant No. 2013CBA01604)National High Technology Research and Development Program of China (863) (Grant No. 2015AA016501)
文摘This work presents a self-consistent two-dimensional(2-D) simulation method with unified physical models for different operation regimes of charge trapping memory. The simulation carefully takes into consideration the tunneling process, charge trapping/de-trapping mechanisms, and 2-D drift-diffusion transport within the storage layer. A string of three memory cells has been simulated and evaluated for different gate stack compositions and temperatures. The simulator is able to describe the charge transport behavior along bitline and tunneling directions under different operations. Good agreement has been made with experimental data,which hence validates the implemented physical models and altogether confirms the simulation as a valuable tool for evaluating the characteristics of three-dimensional NAND flash memory.
