6. Conclusion and Outlook

THE objective of this work was to specify a profound simulation setup for the investigation of GaN-based heterostructure devices and subsequently to study a number of structures. Available material properties (both experimental and theoretical) were summarized with special emphasis on electron transport characteristics. For the latter Monte Carlo simulations were performed while accounting for recent findings (e.g. InN band structure). Based on this extensive compilation and with respect to device peculiarities new models for the electron mobility were developed and calibrated. All other model and material parameters were reexamined and reevaluated according to the current state of knowledge. Relying on this setup several generations of AlGaN/GaN HEMTs were simulated and excellent predictive results were achieved for both the DC and AC characteristics. An optimization study of the gate geometry was performed, which had major influence on the design of subsequent transistor generations. Further investigations included high temperature performance and down-scaling phenomena as well as transconductance collapse. Two different approaches towards realizing normally-off operation were examined supported by experimental data and compared: the recessed gate technique and the InGaN cap layer technique. InAlN/GaN structures were also simulated in order to estimate the advantages against common AlGaN/GaN devices.

All those studies showed that the simulation setup describes adequately numerous devices produced by different processes and featuring several material setups and geometries. The simulations proved to allow valuable insights into device physics. This knowledge can be broadened by gate leakage studies which require a modified Schottky contact model accounting for tunneling effects from the channel.

Another possible direction for future work is the addition of proper treatment of quarternary alloys to the simulator. Devices with an InAlGaN barrier have been proposed recently, as the barrier layer can be grown lattice-matched to GaN and offers superb sheet charge density. The simulation of such structures requires models for quarternary alloys to be developed and calibrated.

HEMT performance is still confined by surface traps. This phenomenon is manifested mainly by drain source current collapse or current lag in pulsed mode. The investigation of this effect will require transient simulations and some additional time-variant sheet charge at the surface.

Last but not least the presented setup allows the simulation of heterostructure bipolar transistors with some minor model expansion for hole transport. Such simulation studies can answer, whether the reduced device performance is due to intrinsic problems or flawed Ohmic contacts.

Those exciting new investigations will help to further advance Nitride device technology. While they can (and most probably will) offer excellent challenges (either due to the required additional models or through reduced computational stability), this work can serve as a basis for those focused studies.