Notes from IEDM, part 1

Typical carbon nanotube FET. Click for larger image. Source: IBM

As the largest technical conference devoted to semiconductor devices, the IEEE Electron Device Meeting, going on this week in Washington, DC, requires some time to absorb and digest. Still, plenty of noteworthy developments are already apparent. For example, IBM demonstrated carbon nanotube field effect transistors. Though such devices have been seen before, IBM claims to have improved conductance by two orders of magnitude, thanks to a new contact deposition and annealing scheme. Elsewhere in the nanoscale device world, a group at NTT Corporation proposed multi-valued logic devices combining single electron transistors with conventional MOSFETs.

Closer to the current mainstream, researchers from IBM presented two papers examining soft and hard breakdown mechanisms in thin gate oxides, and questioning the predictive value of stress-induced leakage current (SILC) measurements. Hard and soft breakdowns appear to follow the same defect paths, with hard breakdown occurring only when the energy dissipated exceeeds some critical threshold. Moreover, the Weibull slope of the time-to-breakdown (T_BD) scales with oxide thickness and is independent of stress conditions. The Weibull slope of the SILC at breakdown does not scale with oxide thickness and tends to saturate over time. The presenter of both papers, Ernest Wu, suggested that new methods are needed to assess and model reliability of very thin gate oxides, particularly for devices that can tolerate increased leakage current due to soft breakdown.

Researchers at IMEC and Infineon concurred, extending the standard percolation model of breakdown to account for SILC, pre-breakdown jumps in leakage current, and breakdown itself. For gate oxides with thicknesses in the 2.4 nm range, two charge traps must be located within 1.2 nm of each other for breakdown to occur. SILC results from single, isolated traps. Pre-breakdown jumps happen when two traps form a bridge across the oxide, but the percolation distance between them is too large for oxide breakdown. Thus, both phenomena illustrate progressive degradation of the oxide.