Research On Noise Models And Their Applications In Nano-Scale Semiconductor Devices

A full understanding of generating mechanisms and precise models is vital to both of the research and the applications to the electrical noise in nano-scale semiconductor devices.Compared with the noise in large scale devices, the noise in nano-scale semiconductor devices shows some new characteristics. Though, the general mechanisms and models for some noise sources in nano-scale semiconductor devices, such as thermal noise and low-frequency noise are the same with those in micrometer devices. The thermal noise generates from the thermal agitation of carriers, while the low-frequency noise may originate from the slow processes related to the particles and structures. However, in nano-scale semiconductor devices, shot noise is induced by the granularity of currents and the random movement of carriers, and further is modulated by the correlations between carriers in transport. As a result, the shot noise phenomenon in nano-scale semiconductor devices is more complicated and useful than that in micrometer devices. Monte Carlo simulations are performed to study the generating mechanisms of shot noise.According to the characteristics of nano-scale semiconductor devices, some of the existing noise models are revised while others are directly used. In terms of these models, the noise performances are evaluated and the characterizations of device performance based on the noise are carried out. Combined with the scientific research items, three different kinds of nano-devices are chosen, which include quantum point contacts (QPC), spin transistors, and high-k gate stack MOSFETs. Device physics and mesoscopic physics are employed to research the noise models of these devices and their application. These cases are instructive to design these related nano-devices, and build noise models in other semiconductor nano-devices.Buttiker's transmission coefficient formula is simple in form, and is usually adopted to calculate the transport performance of QPC under zero temperature. The effect of the temperature is added to Buttiker's transmission coefficient formula, and combined with the current model and the noise model in Landauer's picture, temperature-dependent conductance and noise can be estimated. The expression of intrinsic sensitivity is proposed, which is proportional to the square root of noise and the capacitance electrostatically coupled with environment, and inversely proportional to the differential transconductance. In terms of the capacitance range of QPC, the intrinsic sensitivity is estimated. It is clear that the sensitivity of the present QPC charge sensors still has large margin to improve. Non-equilibrium Green's function method (NEGF) is used to numerically compute the transmission coefficient, and this method gives a more reasonable and relatively exact bias-dependent simulations.When the current with up spin orientation and that with down spin orientation are independent, spin polarization can be used to characterize the polarization degree of spin. But, Rashba spin-orbit coupling is the fundamental of spin transistor, continuous spin precession or instantaneous spin reversal makes currents with different spin orientation no longer independent, and thus spin polarization needs to be derived again. With the frame of scattering theory, spin-density matrix is used to deduce the expression of spin polarization, and the formulas of spin-resolved current and spin-resolved shot noise are obtained. The analytical deduction for single channel is operated, and the simulation for multiple channels is done based on NEGF techniques. With the varying bias, central conductor length and the Rashba spin-orbit coupling coefficient, the simulations for multiple channels are completed to study the relation between spin polarization and Fano factor, and to supple some references for the promising all-electrical measurements.The multi-stack unified noise (MSUN) model in high-k gate stack MOSFETs is a widely accepted analytical noise model at present. But, the trap distribution in high-k gate stack is usually complex, and is not always the exponential distribution assumed by MSUN model. The thought in MSUN model, that is the total noise as the summation of noise in different dielectric layers, is extended to layers divided by trap distribution. When the layers of trap distribution are increased, a simplified method associated with the considered frequency range is introduced. The improved low-frequency noise model could be applied for obtaining useful qualitative conclusion to guide the design of high-k gate stack, and used in noise simulation and the characterization of trap distribution based on noise fitting.At last, full counting statistics (FCS) is employed to explore inverse problem in mesoscopic structures, which is different from the usual route of calculating the high-order cumulants with FCS. FCS theory is an important way to deal with noise, but noise is only a second-order relation in FCS. Based on the exclusion models in FCS theory, a scheme to judge the mechanism and extract the parameters in a single-barrier structure is presented, and furthermore, the scheme is validated with Monte Carlo techniques. A simple but illuminating exploration of transport inverse problem is accomplished.