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

6.2 Effect of light on MIS structures

We next investigate the effect of monochromatic illumination on the impedance characteristic of MIS structures. For comparison, we use both SiC and GaN–based devices. They have similar bandgap energy, namely 3.23 eV (equivalent to a wavelength of 384 nm) for 4H–SiC and 3.39 eV (or 365 nm) for GaN. We expect the largest impact to happen when the photon energy equals or slightly exceeds that of the bandgap, which is large enough for the photons to generate minority carriers. At smaller energies only photoionization takes place, which means that electrons trapped in defect states can be emitted by means of optical radiation.

6.2.1 Exposed CV characteristics of SiC devices

For the first investigations about the impact of light onto MIS devices, we use 4H–SiC MIS samples structured in two different ways. Regular devices have a planar geometry, with an aluminum gate contact as the top layer above the MIS structure. Other devices instead are designed differently, without the thick metal layer above the active area of the device. As we will prove in this Section, these devices are optimized to maximize the absorption of optical radiation in the SiC.

In Fig. 6.3a we can see the CV curves of the optimized SiC device under exposure to monochromatic light. In this case, when the photon energies match that of the bandgap, minority carriers are created. In this way, at 100 kHz we can measure the high frequency curve limit instead of the deep depletion curve  [38]. At 100 Hz the number of generated minority carriers is enough to build the inversion layer at the SiC/dielectric interface, thus shielding the depletion region in the bulk and making the capacitance rise to . We note that a photon energy lower than the bandgap does not allow for the creation of enough minority carriers, although it will still impact capture and emission processes.

The regular devices instead do not show the same results: the inversion layer cannot be formed by photogeneration, even if the photon energy exceeds that of the bandgap and the frequency is low. We can see in Fig. 6.3b that a certain amount of minority carriers is created, but the effect is considerably smaller. The result does not change even by using the low resolution grating, which allows a higher light power by increasing the bandpass of the monochromator. This is due to the presence of the gate contact on top of the device, an aluminum layer with a thickness slightly smaller than 1 µm, which absorbs almost completely the incident radiation. Experiments with different angles of radiation did not change the outcome considerably. We must conclude that such contacts unfortunately block most of the light, therefore shielding the active area of the device from illumination.

6.2.2 Exposed CV characteristics of GaN devices

Next, we investigate the impedance characteristics of GaN/SiN MIS samples under exposure to light. The test structures have a planar design and a circular shape with an aluminum contact that does not cover entirely the device, thus leaving a certain amount of SiN surface exposed to the optical radiation. In this way, even if the thickness of the aluminum layer is the same as for the SiC regular devices, we can see the impact of light on the capacitance– and conductance–voltage characteristics, as shown in Fig. 6.4.

Differently than in SiC MIS structures, the light causes a negative shift of the impedance curves, and just a very slight increase of capacitance at reverse bias. The increase of capacitance can be explained by the shrinking of the depletion region, due to the presence of minority carriers at the GaN/SiN interface. However, it seems that the major effect is the recombination of the generated holes with the electrons in trapping centres, which results in a large negative drift  [40]. In other words, optical radiation causes the photoionization of defects that would not emit the trapped electron at the applied gate bias in dark conditions, because they are located below the Fermi level.

We also note that, at energies lower than that of the bandgap, a ledge appears in the CV curve and a second peak in the conductance characteristics. These are probably related to a specific defect type, whose photoionization energy matches the wavelength used. The fact that the bias range where this effect is visible is in depletion suggests that these trap states are located in the buffer. We study the effect of exposure to monochromatic light at different wavelengths in more detail with the DLOS experiment in the next section.