| Summary: | 碩士 === 國立交通大學 === 電子工程學系 電子研究所 === 102 === With the technology generation down-scaling trend of
metal-oxide-semiconductor field effect transistors (MOSFETs), random telegraph signals (RTS) turn into an essential issue in device development. These signals take place through the carrier capture-emission process via a defect at the silicon/oxide interface or in the oxide layer. The capture-emission behavior of the defect causes the witching of source/drain current between high and low level. The switching magnitude may have negative impact on device performance of MOSFETs, and it can be displayed by calculating ΔId/Id as an index to analyze the characteristic of RTS.
In the experiment, it is not easy to observe the two-level switching magnitude of source/drain current. Besides, it usually appears in terms of more than two levels. Multiple levels would make the issue more complicated, which is not the major issue in this thesis. Most important of
all, dealing with all the parameters, which may affect RTS phenomenon, such as the substrate doping concentration, the device size, and the IV position of defect into control through experiment, are quite difficult. However, we can easily modify these parameters by building all kinds of
device characteristics in TCAD simulations. A conventional RTS magnitude model was derived in flat potential
distribution across the whole channel, but now fails in subthreshold region. The reason is that the potential barrier is highly localized in the middle of channel under a small gate voltage. In this thesis, we establish a new model taking the local barrier into account, which is shown as
conduction band energy along channel length and channel width direction. To obtain RTS index ΔId/Id, we run two cases in device simulations: one of a defect at the silicon/oxide interface or in the oxide layer, and one of
no defect. Then, corresponding drain terminal currents are simulated to determine ΔId/Id.
In our previous work, the switching magnitudes of source/drain current measured in both subthreshold and above-threshold regions were separately transformed into the effective size of the affected region and the effective area of the percolation path. In this work, many of well-known RTS models are considered, by taking into account the defect
position factor ηs1 and the random discrete doping coefficient (percolation effect) ηs2 and ηc.
|
|---|
