Subsections

5.1.5 InAlN/GaN Devices

InAlN/GaN HEMTs have been proposed to provide higher polarization charges without the drawback of high strain [62]. Several groups have demonstrated devices based on InAlN/GaN [390,391] with maximum current capabilities surpassing those of AlGaN/GaN structures.

Optimization of the structures has been carried out by using analytical models [62]. In order to fully develop the potential of the device, proper modeling of the materials is required. Based on the material parameters discussed in Chapter 4, a simulation study is conducted. The HEMT structures described in [391] and [392] are used to benchmark the DC and AC simulation results against measured data. A schematic layer structure of the investigated In$ _{0.2}$Al$ _{0.8}$N/GaN device [392] is shown in Fig. 5.40. All layers are non-intentionally doped.

Figure 5.40: Schematic layer structure of the investigated device.
\includegraphics[width=14cm]{figures/sim/inaln/InAlN3.eps}

5.1.5.1 Simulation Results

A band gap bowing factor of 3 eV is assumed for InAlN as discussed in Section 4.3.1. This yields a band gap of 4.58 eV for In$ _{0.2}$Al$ _{0.8}$N at 300 K (Fig. 5.41). The values for the band offsets are $ \Delta$E $ _\ensuremath{\mathrm{C}}$=0.66 eV and $ \Delta E_\ensuremath{\mathrm{V}}$=0.59 eV, corresponding to a 53%/47% setup.

Figure 5.41: Band alignment of the heterointerface.
\includegraphics[width=8.7cm]{figures/sim/inaln/Banddiag2.eps}

The calculated dielectric permittivity of In$ _{0.2}$Al$ _{0.8}$N is $ \varepsilon_\ensuremath{\mathrm{r}}$=9.86, which is in a good agreement with the value listed in [392]. The barrier height of the gate Schottky contact is 1 eV. The value of the sheet charge density at the InAlN/GaN interface induced by the polarization effects is found to be 3.3$ \times $10$ ^{13}$ cm$ ^{-2}$ from the DC characteristics (simulation results for different values are given in Fig. 5.42).

Figure 5.42: Transfer characteristics for different values of the polarization charge density at the InAlN/GaN interface and $ -10^{13}$ cm$ ^{-2}$ at the InAlN top surface ( $ V_\ensuremath {\mathrm {DS}}$=8.0 V).
\includegraphics[width=9.5cm]{figures/sim/inaln/Transc_lc2.eps}

A commensurate negative surface charge (as the device is not passivated, a low value of 10$ ^{13}$ cm$ ^{-2}$ is assumed) at the top of the InAlN surface is also considered in the simulation. Simulation results for the transfer characteristics assessing different charge values are shown in Fig. 5.43.

Figure 5.43: Transfer characteristics for different values of the total charges (polarization and traps) at the InAlN top surface and 3.3$ \times $10$ ^{13}$ cm$ ^{-2}$ at the InAlN/GaN interface ( $ V_\ensuremath {\mathrm {DS}}$=8.0 V).
\includegraphics[width=9.5cm]{figures/sim/inaln/Transc_uc.eps}

Self-heating effects are accounted for by using a thermal resistance of R $ _\ensuremath{\mathrm{th}}$=3 K/W at the substrate thermal contact. Fig. 5.44 compares transfer characteristics without self-heating effects and with different values of R $ _\ensuremath{\mathrm{th}}$. This value lumps the thermal resistance of the nucleation layer and the sapphire substrate, and possible three-dimensional thermal effects.

Figure 5.44: Comparison of simulated transfer characteristics for different values of the thermal resistance and experimental data ( $ V_\ensuremath {\mathrm {DS}}$=8.0 V).
\includegraphics[width=10cm]{figures/sim/inaln/Transc2.eps}

The simulation exhibits good agreement with the experimental data under consideration of Ohmic contact resistances R $ _\ensuremath{\mathrm{S}}$=R $ _\ensuremath{\mathrm{D}}$=1.3 $ \Omega$mm. The simulated output characteristics show a good agreement with the experimental data (Fig. 5.45).

Figure 5.45: Comparison of simulated output characteristics and experimental data.
\includegraphics[width=10cm]{figures/sim/inaln/Outputc2.eps}

By AC analysis of the device a cut-off frequency $ f_\ensuremath {\mathrm {t}}$$ \approx$7 GHz is obtained. This low value can be explained with the conservative design of the device and the low carrier mobility in the channel ($ \mu$=230 cm$ ^2$/Vs). Downscaled devices are analyzed ( $ L_{\mathrm{g}} =0.5/0.25 \mu\ensuremath{\mathrm{m}}$) and the effect of higher quality GaN material on the AC performance is studied. In our simulation of a device with $ L_{\mathrm{g}} =0.25 \mu\ensuremath{\mathrm{m}}$ reported in [391] (carrier mobility $ \mu$=530 cm$ ^2$/Vs) $ f_\ensuremath {\mathrm {t}}$=36 GHz is reached.


S. Vitanov: Simulation of High Electron Mobility Transistors