Carbon Nanotube (CNT) transistor has been widely studied as an emerging technology for circuit applications. The recently demonstrated CNT computer serves as a major achievement of the technology. However, compared to traditional CMOS devices, the fluctuation in CNT number, separation and diameter leads to additional and major sources of device-to-device variation for CNTFETs. Quantifying the impact of these CNT distribution variations requires modeling these variations into a CNTFET device model.

The first requirement to model the CNT distribution variation is to develop a CNTFET gate capacitance model valid for any generic spacing amongst CNTs. Our proposed capacitance model use the conformal mapping approach to simplify the planar gate CNTFET geometry into well-known structures th...[ Read more ]

Carbon Nanotube (CNT) transistor has been widely studied as an emerging technology for circuit applications. The recently demonstrated CNT computer serves as a major achievement of the technology. However, compared to traditional CMOS devices, the fluctuation in CNT number, separation and diameter leads to additional and major sources of device-to-device variation for CNTFETs. Quantifying the impact of these CNT distribution variations requires modeling these variations into a CNTFET device model.

The first requirement to model the CNT distribution variation is to develop a CNTFET gate capacitance model valid for any generic spacing amongst CNTs. Our proposed capacitance model use the conformal mapping approach to simplify the planar gate CNTFET geometry into well-known structures thus deriving analytical expressions of the capacitance. The capacitance model shows improved accuracy, with maximum error of 2%, in simulating the practical CNTFET geometries. Additionally a capacitance model for conformal top-gated CNTFET structure is also proposed.

Probability density functions of the CNT distribution variations are also required to realistically estimate their impact on CNTFET drive current. Therefore, using statistical methods, we identify the probability density functions of the channel CNT separations and diameters, and verified them using experimental data. Furthermore, a new generic methodology to incorporate the CNT separation and diameter variations into available CNTFET models is also proposed. Simulations of CNTFET drive current shows over-estimated results, by more than 10%, when realistic probability distribution is ignored for uniform distribution. Our analysis also shows that the probability of having a uniform CNT channel region is negligible thus highlighting the contribution of our work. Comparison of a CNTFET against a conventional planar CMOS transistor is also done for the same mean drive current for variation. We found that the CNT process needed to be improved, both in terms of CNT density and variation, for CNTFET to challenge the CMOS technology.