Interesting things in the world of electronics served as easy to digest articles with a touch of wit.

GaN HEMT is widely used in power and RF electronics. Optimizing the structure for the required application is a crucial part of the development process. One parameter that has to be optimized is the Aluminium composition in Al_xGa_1-xN in the barrier layer. Let us see the effects of increasing aluminium composition in GaN HEMT in Silvaco.

Chosen structure

The GaN HEMT stack is a modified version of the stack used in ganfetex07.in, i.e., example 7 in GaNFET tab of the inbuilt example library. The code was put on a loop with the x.comp parameter in the region line increasing with every iteration.

The aluminium composition will have three variations, 0.3, 0.45, and 0.6. The breakdown simulations were done when the transistors were off, with the gate voltages determined by the transfer characteristics.

Energy Band Diagrams and Electron Concentration

GaN HEMTs do not require modulation doping like GaAs HEMTs due to the prevalence of polarization. As we discussed in that post, a few important foemulae were used:

the polarization induced sheet charge density is given by:

And finally the 2DEG sheet carrier concentration is:

For more information on the formation of the triangular quantum well check out : High Electron Mobility Transistors (HEMT)

The main idea is that we can clearly see that increasing lattice constants and bandgaps will lead to an increase in the number of electrons in the 2DEG.

Increasing aluminium in the barrier layer will increase the conduction band offset which leads to a higher electron concentration in the 2DEG. This is what theory says and the cutline taken from gate to channel layer proves it in figure 3.

The minimum conduction band energies were given as:

• x = 0.3: Ec = -0.0775 eV
• x = 0.45: Ec = -0.1315 eV
• x = 0.6: Ec = -0.1789 eV

This allows us to infer that the electron concentration trend will follow 0.6>0.45>0.3, and sure enough figure 4 proves our assumption.

The output in the terminal is given below.

Conduction band: x = 0.3 | Ec(surface) = 0.0000 eV | min Ec = -0.0775 eV at 0.1950 um
Conduction band: x = 0.45 | Ec(surface) = 0.0000 eV | min Ec = -0.1315 eV at 0.1950 um
Conduction band: x = 0.6 | Ec(surface) = 0.0000 eV | min Ec = -0.1789 eV at 0.1950 um
Electron conc: x = 0.3 | peak = 1.0057e+19 cm^-3 at 0.1662 um | n_s = 5.1442e+12 cm^-2
Electron conc: x = 0.45 | peak = 2.0337e+19 cm^-3 at 0.1662 um | n_s = 1.0471e+13 cm^-2
Electron conc: x = 0.6 | peak = 3.1542e+19 cm^-3 at 0.1662 um | n_s = 1.6318e+13 cm^-2

The terminal output confirms the peak values we observed in the graphs. The electron concentration has doubled from 0.3 to 0.45, and tripled if we compare 0.3 to 0.6.

Comparison of device characteristics

A dense 2DEG forming without the assistance of the gate will give us a very low threshold voltage and a really high drain voltage when the gate is unbiased .

The threshold voltage has decreased dramatically when increasing the aluminium composition.

As expected, the increased electron concentration leads to an increased amount of current.

where 𝑣_𝑒𝑓𝑓 is the effective electron velocity in the 2DEG channel and 𝑊_𝐺 is the gate width.

The gate leakage reduces considerably when we increase the Al composition. The question we must now ask is if increasing Al composition is a better choice or not? The biggest con till now is the need to give a large negative voltage to switch the device off. This will increase circuit complexity considerably. The biggest problem with this device will be the off condition drain leakage and very low breakdown voltage.

As can be seen on figure 9, the breakdown voltage is very low for 0.45 and 0.6 when we compare it to 0.3. All of the breakdown simulations were run when the transistors were given their off state voltages. At on state, the impact selb parameter was not working correctly. Even with a large off voltage, the devices exhibit low breakdown voltages and large drain leakage.

The output in the terminal for the code running characteristics is given below.

x = 0.3: Vth = -1.7306 V
x = 0.45: Vth = -3.4780 V
x = 0.6: Vth = -5.3894 V
Breakdown: x = 0.3 | V_BR = 617.0632 V | slope = 2.710327e-03 A/V | R^2 = 0.96458 | fit window = 0.2-0.6 of Imax | fit range = 625.0000 to 640.0000 V
Breakdown: x = 0.45 | V_BR = 186.5981 V | slope = 2.713723e-04 A/V | R^2 = 0.99727 | fit window = 0.15-0.2 of Imax | fit range = 275.0000 to 295.0000 V
Breakdown: x = 0.6 | V_BR = 57.0207 V | slope = 3.563762e-04 A/V | R^2 = 0.97803 | fit window = 0.2-0.5 of Imax | fit range = 75.0000 to 95.0000 V

Conclusion

Increasing the amount of current carriers does not increase device performance. A large number of current carriers decreases the voltage at which impact ionization occurs at an uncontrollable scale which leads to an avalanche breakdown. The Al composition in a GaN HEMT must be accurately controlled for best device performance.

Code source: Arjun’s Github

Related Posts

Frequently Asked Questions (FAQ)

Ques: What is a HEMT barrier layer?

The barrier layer is the wider-bandgap semiconductor layer placed above the channel, such as AlGaN in an AlGaN/GaN HEMT. It creates the heterojunction and, together with polarization effects, helps form the high-density 2DEG at the AlGaN/GaN interface.

Ques: What is a HEMT transition layer?

The transition layer is a layer placed between the barrier/channel structure and the substrate. It helps accommodate the lattice and thermal mismatch between the GaN device layers and the substrate, while reducing defects and improving material quality. It is also commonly called a buffer layer when its main purpose is defect isolation and electrical isolation.

Ques: What does Al composition in a GaN HEMT do?

The Al composition $x$ in AlₓGa₁₋ₓN changes the material’s bandgap, lattice constant, and polarization. Increasing $x$ generally increases the polarization difference between AlGaN and GaN, which can increase the 2DEG sheet charge density. However, too much Al can increase strain and defect-related problems.

Ques: Why are GaN HEMTs depletion mode transistors?

In a conventional AlGaN/GaN HEMT, the polarization-induced electric field naturally creates a 2DEG even when no gate voltage is applied. Therefore, the device is normally ON at $V_G=0$, giving it a negative threshold voltage and making it a depletion-mode transistor. Enhancement-mode GaN HEMTs require additional gate structures to suppress the 2DEG at zero gate bias.

Ques: Where do the electrons in the 2DEG come from?

The electrons forming the 2DEG can originate from several sources, depending on the HEMT structure: intentional donor impurities such as Si in a doped AlGaN barrier, donor-like surface states in an undoped AlGaN barrier, and defects such as nitrogen-vacancy-related donor states. Other proposed mechanisms involve unintentional impurities and defects within the heterostructure. In the undoped AlGaN/GaN structure, surface donor states have been experimentally and theoretically identified as an important electron source [Reference: J.P. Ibbetson (2000)]. The polarization-induced electric field then drives these electrons toward the AlGaN/GaN interface, where they accumulate to form the 2DEG.