Performance comparison between current-mode signaling and voltage-mode signaling for multilayer graphene nanoribbon (MLGNR) interconnects

Graphene nanoribbon (GNR) is emerging as a superior material for nanometer-scale interconnects, offering superior performance compared with traditional copper materials. To date, most research on GNR interconnects has focused on voltage-mode signaling (VMS) scheme, with little study on current-mode signaling (CMS) scheme. In this paper, we propose an equivalent circuit model of two-wire coupled multilayer graphene nanoribbon (MLGNR) interconnects using VMS and CMS schemes. Moreover, the model takes into account influence of temperature effect, coupling capacitive and mutual inductive. Performance of victim wire in two-wire coupled MLGNR and Copper (Cu) interconnects using VMS and CMS signaling schemes is investigated by applying the decoupling approach and ABCD parameter matrix method at local, intermediate, and global levels, respectively. In addition, the performance of MLGNR and Cu interconnects employing VMS and CMS systems is thoroughly compared and examined in this research. The results reveal that interconnects adopting the CMS scheme have less output voltage swing, less crosstalk delay, greater 3-dB bandwidth, and better signal integrity, compared to interconnects applying the VMS scheme, under the same conditions. With respect to noise, we observe that the CMS scheme has lower noise amplitude, smaller noise peak, and smaller noise width, resulting in greater noise immunity. Moreover, it is manifested that crosstalk delay, noise width, and 3 dB bandwidth are all temperature-dependent. As the temperature rises, both the delay and noise width increase, while the bandwidth decreases. In addition, the results indicate that MLGNR interconnects exhibit lower crosstalk delay, narrower noise width, larger bandwidth, and smaller dynamic power consumption compared to Cu interconnects under the same conditions. Furthermore, we discuss performance optimization methods for interconnects using both VMS and CMS schemes. Also, it is discovered that there is great agreement between the results of HSPICE simulations and those produced by the ABCD parameter matrix technique.