5.1.4 Optimization Techniques

A significant improvement of the device performance has been achieved by adopting the field plate technique [68]. With its origins in the context of high-voltage p-n junctions [10] the main purpose of the field plate is to reshape the electric field distribution in the channel and to reduce its peak value on the drain side of the gate edge. The benefit is an increase of the breakdown voltage and a reduced high-field trapping effect. Overall the power density could be increased from 10 W/mm [389] to 40 W/mm [65]. Although sharing the same principle with FETs/MESFETs the effect of the field plate on HEMTs has to be studied extensively with account of the different transport mechanisms in the later devices.

5.1.4.1 Device Structures

The devices share a similar design with the already simulated structures (especially to Device A, excluding the field plate). The gate is so far e-beam defined with a gate length of $L_{\mathrm{g}} =150 \ensuremath{\mathrm{nm}}$, 300 nm, and 600 nm. Device isolation is achieved by mesa isolation. Fig. 5.31 gives an example of a typical fully planar AlGaN/GaN HEMT. An Al$_{0.3}$Ga$_{0.7}$N/GaN heterointerface is grown on top of a thick insulating GaN buffer. All layers are non-intentionally doped, except the $\delta$-doping which is introduced in the AlGaN supply layer to provide additional carriers and to improve the access resistances. The maximum drain current density is larger than 900 mA/mm and the transconductance is larger than 200 mS/mm at $V_\ensuremath {\mathrm {DS}}$=7 V. The current gain cut-off frequency $f_\ensuremath {\mathrm {t}}$ is well beyond 30 GHz for devices with $L_{\mathrm{g}} =300 \ensuremath{\mathrm{nm}}$. The density of the positive charge at the channel/barrier interface is found to be $\approx$1.1$\times$10$^{13}$cm$^{-2}$ (see Fig. 5.32).

Figure 5.31: A schematic layer structure of single heterojunction AlGaN/GaN HEMTs with field plates.

Figure 5.32: Transfer characteristics for different polarization charge densities at the AlGaN/GaN heterojunctions.

5.1.4.2 Simulation Results

The critical variables associated with the field plate for a given gate length $L_\ensuremath {\mathrm {g}}$ are the field plate length $L_\ensuremath {\mathrm {FP}}$ and the SiN thickness (see Fig. 5.31). While the gate length is crucial for the transit time, the field plate length is the major factor for the size of the electrical field-reshaped region [371]. The nitride thickness controls the onset voltage but has also significant influence on the maximum electric field.

Calibration and Field Plate Optimization of $L_{\mathrm{g}} =600 \ensuremath{\mathrm{nm}}$ HEMTs:

AlGaN/GaN HEMTs with $L_{\mathrm{g}} =600 \ensuremath{\mathrm{nm}}$ nm have been used for device analysis and model calibration. Fig. 5.33 shows measured (solid lines) and simulated (dashed lines) output characteristics of $L_{\mathrm{g}} =600 \ensuremath{\mathrm{nm}}$ HEMTs. Modeling issues remain for $V_\ensuremath {\mathrm {GS}}$=1 V.

Figure 5.33: Comparison of measured (solid lines) and simulated (dashed lines) output characteristics of $L_\ensuremath {\mathrm {g}}$= $L_\ensuremath {\mathrm {FP}}$=600 nm HEMTs.

Fig. 5.34 compares measured (symbols) and simulated (lines) transfer characteristics of HEMTs without and with field plate. Very good agreement between simulation and measured electrical data is achieved for both devices. The difference caused by the field plate is better demonstrated by the electrical field distribution in the channel as shown in Fig. 5.35. A $50\%$ reduction of the maximum electric field, located at the drain side of the gate edge, is achieved by the introduction of the field plate. A second peak occurs at the field plate edge.

Figure 5.34: Comparison of measured (symbols) and simulated (lines) transfer characteristics of HEMTs with and without field plate.

Figure 5.35: Simulated electric field along the channel of $L_\ensuremath {\mathrm {g}}=600$ nm HEMTs with and without field plate for $V_\ensuremath {\mathrm {GS}}$=0 V and $V_\ensuremath {\mathrm {DS}}$=7 V.

The choice of an optimum field plate length is made with respect to the desired electrical properties [130]. Here, the aim is the highest reduction of the peak electric fields in the channel for $V_\ensuremath {\mathrm {GS}}$= $V_\ensuremath{\mathrm{FPS}}$=$-$7 V, $V_\ensuremath {\mathrm {DS}}$=60 V, thus securing a reliable device operation up to this bias. The optimum field plate length $L_\ensuremath {\mathrm {FP}}$= $L_{\mathrm{g}} =600 \ensuremath{\mathrm{nm}}$ is found after variation of $L_\ensuremath {\mathrm {FP}}$in the range 200 nm-800 nm (see Fig. 5.36).

Figure 5.36: Simulated electric field along the channel for various field plate lengths $L_\ensuremath {\mathrm {FP}}$ ( $L_\ensuremath {\mathrm {g}}=600$ nm).

Field Plate Optimization of Shorter HEMTs:

Fig. 5.37 shows simulation results for HEMTs with shorter gate lengths down to 150 nm. While such devices are economically favored, the simulation results show that for small $L_{\mathrm{g}}$ the maximum electric field increases by up to $60\%$, thus requiring a mechanism to compensate the negative impact. The field plate length is kept equal to $L_{\mathrm{g}}$. The optimum value for the thickness of the nitride passivation in $L_{\mathrm{g}} =600 \ensuremath{\mathrm{nm}}$ devices is further preserved.

Figure 5.37: Simulated electric field along the channel for various gate lengths ( $L_\ensuremath {\mathrm {FP}}$= $L_\ensuremath {\mathrm {g}}$).

The next simulations involve devices with $L_{\mathrm{g}} =300 \ensuremath{\mathrm{nm}}$ (Fig. 5.38). Increasing $L_\ensuremath {\mathrm {FP}}$ is advantageous up to $L_\ensuremath {\mathrm {FP}}$=500 nm regarding the peak electrical field. Fig. 5.39 shows electrical field distribution in the channel of $L_{\mathrm{g}} =150 \ensuremath{\mathrm{nm}}$ HEMTs. $L_\ensuremath {\mathrm {FP}}$ varies from 100 nm to 400 nm. Compared with the previous simulations a clear improvement for longer field plates is observed. The increase of $L_\ensuremath {\mathrm {FP}}$ from 150 nm to the optimum of about 350 nm helps to decrease the peak electrical field by up to $40\%$.

Figure 5.38: Simulated electric field along the channel for various field plate lengths $L_\ensuremath {\mathrm {FP}}$ ( $L_\ensuremath {\mathrm {g}}=300$ nm).

Figure 5.39: Simulated electric field along the channel for various field plate lengths $L_\ensuremath {\mathrm {FP}}$ ( $L_\ensuremath {\mathrm {g}}=150$ nm).