Characterization of electrically active defects at III-N/dielectric interfaces

 
« PreviousUpNext »Contents
Previous: 6.3 Effect of light on GaN/AlGaN MIS-HEMTs    Top: Home    Next: Outlook

Conclusions

In this work we have addressed the problem of characterizing the defects at the interface between the AlGaN barrier and the dielectric in composite GaN/AlGaN/SiN MIS–HEMTs. Such states act as trapping centres for electrons, causing severe threshold voltage shifts that compromise correct device operation.

The first result of this thesis is that most of the well–known, established methods for interface trap characterization based on the analysis of the impedance characteristics are inappropriate for GaN/AlGaN devices. In fact, while these techniques are used successfully on silicon structures, several approximations do not hold true for GaN–based samples. This is mainly due to the response of the barrier in the composite GaN/AlGaN stack.

Secondly, we have developed an accurate method for characterizing the defects at the interface with the dielectric, based on the study of the device response to a forward bias DC stress applied to the gate. The fundamental approach is the same as that already in use for bias temperature instability investigations on silicon MOSFETs. We have optimized such techniques for our test structures, taking into account their high level of instability and their sensitivity to the measurement parameters. A series of experiments of this kind at various temperatures has allowed us to extract the density of interface traps as a function of their activation energy for electron trapping.

We have found that the GaN/AlGaN/SiN devices with a standard interface have a rather high density of trap levels (1013/(cm2 eV)). The most important feature of such defects is their broadly distributed activation energy. In fact, the density of traps per unit area and energy does not fall below 5 × 1012/(cm2 eV) in a continuous range from 0.1 eV to 0.8 eV, with a slight increase at the lower extrema. This results can be explained by the disordered nature of the AlGaN/SiN interface, that is characterized by a non–stoichiometric, nanometer–thick interlayer between the two materials. Here, the point defects might have a number of slightly different microscopical arrangements, giving rise to a broad distribution of activation energies for charge trapping.

On the other hand, MIS–HEMTs that undergo a plasma cleaning process involving fluorine of the AlGaN surface before the deposition of the dielectric exhibit unique charge trapping behavior. In fact, in this case we have measured a larger density of traps of 8 × 1013/(cm2 eV), with activation energies normally distributed around 0.32 eV. As a consequence, the trapping transients are very fast and threshold voltage shifts saturate within one second. This results in a stable device behavior that has never been observed before. The fundamental change with respect to the structures with a standard interface suggests that the fluorination process passivates the defects at the AlGaN surface and substitutes them with another kind of trap level, characterized by a well–defined activation energy and a larger density per unit area and energy.

Finally, we have established a setup and developed an improved design for test structures for combined electrical and optical experiments, aimed to investigate trap levels with large activation energies. The first results indicate the presence of several trap states that have been observed previously by other authors, mainly related to vacancies and impurities in the GaN or the AlGaN. Furthermore, we have observed a defect level with optical activation energy of 2.3 eV that could be located at the AlGaN/SiN interface, as is visible only after the application of a positive gate bias stress.

« PreviousUpNext »Contents
Previous: 6.3 Effect of light on GaN/AlGaN MIS-HEMTs    Top: Home    Next: Outlook