A Modeling Study of Stacked Cu-CNT TSV on Electrical, Thermal, and Reliability Analysis

Abstract

In this article, we propose a position-oriented carbon nanotube (CNT) sampling method based on the Monte Carlo (MC) concept, which obtains the relative positional distribution and concentration of CNTs, providing an accurate modeling basis. The copper-CNT (Cu-CNT) through-silicon via (TSV) model is established through the transition from the unit level to the circuit level, and the circuit analysis is performed with the TSV pair. We find that Cu-CNT TSVs have better resistivity than Cu TSVs in long-distance transport under the premise of ensuring variability. But CNT does not optimize the crosstalk performance of the composite material, as expected, due to the inhibition of CNTs by Cu in the composite. Cu-CNT well reduces the negative impact of interfacial resistance on delay (<0.146%) and exhibits excellent heat dissipation and reliability. Under ideal conditions, its median time to failure (MTTF) can be as high as Cu ten times with a small current capacity loss. Moreover, we also intersperse discussions of changes in the filling ratio, CNT diameter, and aspect ratio (AR) throughout the analysis to further study the effect of CNT variability.