Air Spacer MOSFET Technology
Jemin Park and Chenming Hu
Samsung and ERSO Chair fund
In a 20 nm-gate MOSFET with oxide spacer, 77% of the gate charge is due to the gate-to-plug/diffusion capacitances. Reducing these capacitances will be an increasingly important way to improve the device speed and switching energy/power at 20 nm and beyond. Compared to an air-spacer inverter, a conventional nitride-spacer inverter has 82% longer delay and 85% larger switching energy (power consumption). Even a pure-oxide-spacer inverter has 41% longer delay and 48% larger switching energy than the air-spacer inverter.
High density memories employ the SAC technology that requires the use of nitride spacers. This significantly raises the gate-to-plug/diffusion capacitance and increases the delay and switching energy by about 60%. A novel air-spacer SAC device can preserve the 35% area benefit of a SAC device while reducing the delay and power by over 75% to levels even better than the non-SAC conventional device. It also reduces the bit-line and word-line capacitances. The result is increased DRAM and SRAM speed, reduced power, and reduced chip size. These air spacer technologies are a promising key technology for the 20 nm generation and beyond.
Figure 1: MOSFETs constructed with 3D simulator. In (b) and (c) part of ILD is removed to show the outlines of SAC. Lgate = 20 nm, nitride/oxide/air spacer thickness = 12 nm. (a) Non-SAC MOSFET (oxide spacer); (b) SAC MOSFET (nitride spacer); and (c) SAC MOSFET (air spacer)
Figure 2: The mixed-mode simulation of inverter delay. The delay of the air-spacer is decreased by 45% and 30% compared with the nitride-spacer and oxide-spacer, respectively.
- M. Togo, A. Tanabe, A. Furukawa, K. Tokunaga, and T. Hashimoto, "A Gate-Side Air-Gap Structure (GAS) to Reduce the Parasitic Capacitance in MOSFETs," Symposium on VLSI Tech. Dig., p. 38, 1996.
- TSUPREM4 User Manual, Synopsys, Mountain View, CA.
- SENTAURUS User Manual, Synopsys, Mountain View, CA.