Hello everyone, welcome to the new post today. In this article, we present the simulation of 4H-SiC and GaN p-n diodes using the modified parameters as adjustments in Medici.

In the field of power semiconductors, wide bandgap p-n junction devices such as silicon carbide (SiC) and gallium nitride (GaN) have gained enormous attention. This is mainly due to their ability to provide a high critical electric field that sustains elevated breakdown voltage. On one hand, silicon carbide and its substrates offer us exceptionally high-voltage SiC devices whereas GaN devices provide better efficiency and cost saving, eliminating channel mobility issues.
Ideally, it is quite easy to simulate Sic and GaN devices in Medici with normal models and parameters. However, simulation for reverse characteristics of WBG p-n junction using Medici is difficult. This is because the ideal equilibrium minority carrier concentration is too short to give acceptable results on impact ionization carrier generation.

Therefore, in this article, we present the simulation of 4H-SiC and GaN p-n diodes using the modified parameters as adjustments in Medici. Moreover, we compare the obtained results with calculated results to establish the strength of predictive capability for reverse characteristics of WBG p-n junction devices.

Modern Sic high-power devices are still lacking in efficiency, mainly due to their defects and imperfect substrate quality. The commonly seen defects in SiC epilayers are surface and crystalline defects. Surface defects like scratches and growth pits affect shallow junction devices, whereas crystalline defects like elementary screw dislocations and micro pipes affect p-n junctions adversely.

Another important defect is the dislocation effect, which is a key factor in varying  SiC diode reverse characteristics. The existence of this effect leads to the generation of trap states within band gaps which increases the impact ionization process. In this article, we have found the dislocation levels to be at (EC - 0.3 eV) and (EV + 0.4 eV), which is included as a key physical parameter in the simulation of reverse characteristics.

In order to simulate the reverse characteristics of the SiC p-n junction in Medici, we have to account for all the above defects using the TRAP model. If the trap state is CHARGED, then we can apply the following Poisson equation 1:

Equation 1

Here, N+D and N-A  are the ionized impurity concentrations, Ψis the intrinsic Fermi potential, ρs interface charge density and Nti is the total number of traps for the i-th energy level.

It is known that Gallium Nitride (GaN) technology is the next big thing in the field of electronics, however, its simulation for reverse characteristics is somewhat difficult.

The selection of impact ionization coefficients of electrons and holes is the key factor in the simulation of GaN breakdown voltage. Figure 1 shows the impact ionization coefficients as a function of inverse electric field for holes and electrons.

Figure.1 Ionization coefficient vs. inverse electric field

It is clearly seen that the impact ionization coefficients are bound to unite within a range of 5×104 and 3×105 per cm when the field is larger than 4 MV/cm.

Ⅰ. Simulation results and discussion

The convergence that results from the incredibly low intrinsic carrier concentration is the main issue with the simulation of SiC device breakdown. In order to eliminate this issue, we introduce photo generation in Medici which increases minority carrier concentration artificially.
For ease of simulation, we set up a single constant parameter called AI, this constant represents uniform carrier generation which decides the photo-generation rate. Ideally, the value of AI is set to be 10 % of the doping concentration of the drift layer.  

Moreover, trap density is another parameter that is related to thickness and drift layer concentration as follows in equation 2;

Equation 2

Here α and β are the parameters of trap density which is incited by implantation and growth respectively, N is the drift layer doping concentration and W is the drift layer thickness. In this simulation, we have set the value of  α  as 0.06675, whereas β is 0.89×1015. These above values are then compared with the values obtained in the simulation for verification shown in figure 2.

Figure.2 Simulation acquired values for different parameters

The vulnerability of breakdown voltage of p-n junction diode with trap density is shown in figures 3 and 4 with varying doping concentration and width W.

                  Figure.3 Trap density for thick low doped drift region

             Figure.4 Trap density for highly doped drift layer

It is observed that Medici simulations are fairly accurate for devices with thin drift layer, showing that it used the trap model in equation 2 successfully to simulate the SiC p-n junction realistically. Moreover, it can be proved that trap density is more notably effective at thick and lightly doped drift regions.

Leakage currents which are simulated with adjusted AI are shown in Figures 5 and 6.

Figure.5 Leakage current with more drift layer concentration

Figure.6 Leakage current with lesser drift layer concentration

The value of diodes for Figures 5 and 6 with respective drift layer concentrations of 9.7×1015 cm-3 and 3×1015 cm-3 is 2×1015 cm-3 and 1×1014 cm-3 respectively. As the AI values do not vary with breakdown voltage simulation, we can safely select its optimum value around 10 % of drift layer doping concentration to find reverse current.

Figure 7 shows the capability to predict SiC p-n junction leakage current via an ideal simulated leakage current with experimental data which is reported as shown.

Figure.7 Leakage current with experimental data(optimized)

Here optimized A1 of 1×1014 cm-3  is simulated with the drift layer concentration of 1×1015cm-3.

We know that for the simulation of GaN p-n junction breakdown, we need adjustments in impact ionization coefficients. The said adjustments are shown in figure 8 below:

Figure.8 Adjustments made in parameters for better simulation results

From the simulation results, it can observe that we achieve breakdown voltage at 500 V with the adjusted coefficients, which is in the acceptable range. Moreover, more investigation on the quantitative expression of Ec against drift region thickness and defect density will be carried out.  

Ⅱ. CONCLUSION

SiC and GaN p-n junction devices offer high efficiency and power which makes them a critical factor in the development of semiconductor devices, therefore it is crucial to simulate them correctly.
We did this by choosing appropriate optimum trap and carrier photo-generation density parameters established on the doping concentration and drift layer thickness related to defect density. The selection of all parameters with their adjustment is used to find both the breakdown voltage and leakage current of simulated SiC p-n diodes. This methodology clearly establishes a relationship between defects and junction breakdown characteristics.

For GaN p-n junction simulation the impact ionization coefficients are varied within the specified critical field zone which is within the calculated values, Good concurrence is found by using the adjusted coefficients.