Advanced Gate Concepts for Sub 45nm Devices

The scaling of the gate length down in CMOS devices increases the drivecurrent performance and the density of devices per chip area, improvingthe chip functionality and the cost per function. As the gate length and lateral dimensions are reduced, all the other vertical dimensions haveto follow the scaling too. The junction depths, effective oxide thickness or EOT and the other dimensions have all to follow the gate scaling. The EOT is one of the most important parameters, it represents the equivalent thickness of oxide needed to achieve a desired capacitance. Unfortunately, as the physical oxide thickness or EOT is reduce under 2nm, additional effects start to harm the device performance, namely polydepletion and direct tunneling currents. Polydepletion appears as an additional, not desired, increase in EOT. That happens because of the semiconductor nature of polysilicon. For small EOTs, a part of the gate bias is dropped in the polysilicon electrode charge region. This bias loss implies on a lower channel charge and therefore a higher EOT. The use of metal electrodes will eliminate the polydepletion and further decrease EOT down. However, the threshold voltage in these devices is determined with the metal workfunction. The workfunction is a material property and is very difficult to be tuned by ion implantation or alloying. This makes very difficult to integrate two metal electrodes in one wafer. As the oxide physical thickness is reduced, direct tunneling leakage currents arise. To avoid the tunneling currents, high-k dielectrics are the most natural choice. It is possible to achieve the same EOT with a much thicker dielectric due to the high permittivity. The large physical thickness eliminates the tunneling leakage currents. Although the high-k dielectrics can solve the tunneling leakage currentproblems, carrier mobility degradation is commonly reported in MOS transistors. The cause for the mobility degradation is attributed to either remote Coulomb scattering or remote phonon scattering, but not well understood so far. Threshold voltage control with high-k dielectrics is alsobad due to the large fixed charge in the dielectric layers or interfacereactions with the gate electrode. Both the threshold voltage control and the mobility degradation harm the drive current performance of the CMOS devices. To study the mobility degradation, a new model for the inversion layer was developed based on the variational principle. The model considers trial wave functions for each subband separately. The dielectric barriers and the inversion layer wave function permeation into the barriers are also included in the model. The inversion layer model was further extended to calculate the tunneling leakage from the resonant subband states. Good agreement between measurements and simulations was found. The electron mobility was calculated using the relaxation time approximation. A new model for remote Coulomb scattering in high-k layers was developed for an arbitrary number of dielectric layers in the stack. The model is a new extension of the transfer matrix method for the non-homogenous differential equation. Good agreement between the model and experimental results was found. The electron mobility degradation in high-k layers is mostly due to remote Coulomb scattering from charged impurities in the dielectric layers. The simulation results show that high dielectric permittivity values are can reduce the mobility degradation. Metal gate electrodes can also screen some of the mobility degradation as compared with polysilicon. Transistors with metal electrodes and high-k dielectrics were fabricated with gate lengths down to 90nm. Polydepletion was eliminated and higher drive currents were obtained with metal/high-k than polysilicon/high-kstacks. High drive current levels were achieved with EOTs around 1.2nm. The inversion layer and the leakage model can be extended to new devicestructures like SOI or Finfets. The model can also be used to unite themobility and leakage together. As the leakage current is increased, small subband energy dispersion appears. This dispersion will increase the scattering rates and therefore changing the resulting mobility.status: publishe