Self-compensation by silicon $DX$ centers in ultrawide-bandgap nitrides

This study demonstrates that silicon $DX$ centers induce significant self-compensation in ultrawide-bandgap nitrides like AlN by stabilizing a negative charge state, which severely limits free electron concentrations and renders them largely independent of doping levels, although higher carrier densities may be achievable in AlGaN alloys or cubic boron nitride where the $DX$ level is closer to the conduction band.

The Gist

Imagine you are trying to build a super-fast highway for electricity (electrons) through a very tough, narrow tunnel made of special materials like Aluminum Nitride (AlN) or cubic Boron Nitride (c-BN). These materials are "ultra-wide-bandgap" semiconductors, which means they are incredibly strong and can handle high power and heat, making them perfect for next-generation electronics.

To make electricity flow, you need to add "passengers" (electrons) to the highway. You do this by adding a special ingredient called Silicon (the dopant), which acts like a ticket booth handing out free rides.

However, this paper reveals a frustrating problem: The ticket booths are turning into toll booths that block the road.

Here is the breakdown of what the scientists found, using simple analogies:

1. The "Double-Grab" Problem (The DX Center)

In a normal world, a Silicon atom gives up one electron to the highway and stays positively charged (like a happy ticket seller). But in these tough materials, Silicon behaves strangely. It acts like a greedy magnet called a DX center.

Instead of just giving away an electron, the Silicon atom grabs two electrons from the highway and holds onto them tightly.

• The Result: The Silicon atom becomes negatively charged (a "toll booth" that demands money).
• The Consequence: It cancels out the positive charge of other Silicon atoms. This is called self-compensation. It's like hiring 100 people to build a road, but 50 of them immediately start digging holes to undo the work. No matter how many workers you hire, the road never gets built.

2. The AlN Tunnel: A Dead End

The researchers looked at Aluminum Nitride (AlN) first.

• The Situation: In AlN, the Silicon "greedy magnet" is very strong. It sits deep in the tunnel, far away from the main highway.
• The Outcome: Even if you dump a massive amount of Silicon into the material (heavy doping), the Silicon atoms just pair up: one gives an electron, the other grabs it back.
• The Limit: The number of free electrons stays stuck at a very low level (about 300 trillion per cubic centimeter), regardless of how much Silicon you add. It's like trying to fill a bucket with a hole in the bottom; adding more water doesn't help.

3. The Solution: Adding a "Lubricant" (Alloying with Gallium)

The scientists asked, "Can we fix this?" They tried mixing in a little bit of Gallium to create an alloy (AlGaN).

• The Analogy: Imagine the tunnel has a steep slope that makes the Silicon atoms want to grab electrons. Adding Gallium is like lowering the slope.
• The Result: The "greedy magnet" (the DX level) moves closer to the highway. It becomes harder for Silicon to grab those two electrons.
• The Payoff: Now, the Silicon atoms act like normal ticket sellers again. They give up their electrons and stay positive. By adding just 9% Gallium, the number of free electrons jumped by 1,000 times compared to pure AlN.

4. The c-BN Shortcut: A Better Tunnel

Finally, they looked at cubic Boron Nitride (c-BN).

• The Situation: In this material, the "greedy magnet" isn't as strong as in AlN, but it's not as weak as in the Gallium alloy. It's somewhere in the middle.
• The Result: You can get more electrons than in AlN, but if you try to add too much Silicon, the greedy magnet wakes up again, and the road gets blocked.
• The Sweet Spot: You can get good conductivity, but you have to be careful not to over-dope it. Light to moderate amounts of Silicon work best here.

5. The Temperature Factor

The paper also checked what happens when the materials get hot (like in a real engine or power grid).

• The Good News: Heat helps shake the electrons loose, making it easier for them to get on the highway.
• The Bad News: Even with heat, if you are in pure AlN, the "greedy magnet" is still too strong. The road remains blocked. But in the Gallium alloy and c-BN, heat helps even more, allowing for better performance at high temperatures.

The Big Takeaway

If you want to build super-fast, high-power electronics using these tough materials:

1. Don't just dump Silicon into Aluminum Nitride (AlN). It won't work well because the Silicon will cancel itself out.
2. Mix in Gallium. This "tricks" the Silicon into behaving normally, allowing electricity to flow freely.
3. Consider Boron Nitride (c-BN). It's a good middle ground, but you still need to be careful with how much Silicon you add.

In short: The Silicon atoms in these materials have a bad habit of hoarding electrons. By changing the recipe (adding Gallium) or choosing a different material (c-BN), scientists can stop the hoarding and finally build the super-highways needed for the future of electronics.

Technical Summary

1. Problem Statement

Ultrawide-bandgap (UWBG) nitride semiconductors, such as Aluminum Nitride (AlN) and cubic Boron Nitride (c-BN), are critical for high-power, high-frequency, and extreme-environment electronics. However, achieving efficient n-type conductivity in these materials remains a significant challenge.

• The Core Issue: While silicon (Si) is the most effective n-type dopant candidate, it often fails to provide high free electron concentrations.
• The Mechanism: Instead of acting as effective-mass donors, Si impurities in these materials frequently exhibit DX center behavior. A DX center captures two electrons, transitioning from a neutral state ($d^0$) to a negatively charged state ($DX^-$).
• Self-Compensation: This process leads to a reaction $2d^0 \rightarrow d^+ + DX^-$, where the dopant creates both a positive donor and a negative compensating center. This "self-compensation" pins the Fermi level deep within the bandgap, severely limiting the number of free carriers, regardless of how much silicon is added.
• Knowledge Gap: Previous studies often attributed low carrier concentrations to extrinsic defects (impurities or native defects). This paper investigates whether the DX behavior of the dopant itself is the primary limiting factor, even in the absence of other defects.

2. Methodology

The authors employed a multi-scale computational approach combining first-principles data with thermodynamic modeling:

• Input Data: They utilized existing first-principles calculations for Si donor transition levels in AlN, AlGaN, and c-BN (specifically the $(+/−)$, $(+/0)$, and $(0/−)$ levels).
• Temperature-Dependent Band Structure:
  • Standard transition levels are calculated at 0 K. To model real-world device operation, the authors calculated the temperature dependence of band gaps and band-edge positions using Density Functional Theory (DFT).
  • They used the r2SCAN meta-GGA functional within the VASP code and the Quasi-Harmonic Approximation (QHA) to account for thermal expansion and electron-phonon interactions.
  • This allowed them to track the shift of the Conduction Band Minimum (CBM) relative to the defect levels as temperature increases (up to 700 K).
• Charge Neutrality Modeling:
  • Using the SC-FERMI code, they solved the charge neutrality equation.
  • They modeled systems with varying Si concentrations ($10^{16}$ to $10^{20}$ cm$^{-3}$) and temperatures (200–700 K).
  • Assumption: The model assumes an idealized "best-case scenario" where Si is the only defect species, isolating the effects of DX self-compensation from extrinsic impurities.

3. Key Contributions

• Isolation of DX Self-Compensation: The study definitively demonstrates that DX centers alone are sufficient to cause severe self-compensation in Si-doped AlN, even without other defects.
• Temperature-Dependent Analysis: By explicitly calculating the shift of the CBM with temperature, the authors provide a more accurate prediction of carrier concentrations at operating temperatures (e.g., 300 K and 500 K) compared to static 0 K models.
• Comparative Material Analysis: The paper systematically compares three materials:
  1. AlN: The baseline with the deepest DX level.
  2. Al$_{0.91}$Ga$_{0.09}$N: An alloy where Ga incorporation moves the CBM closer to the DX level.
  3. c-BN: A material with a DX level closer to the CBM than AlN but further than the alloy.

4. Key Results

A. Silicon-Doped AlN

• Severe Limitation: The Si $(+/−)$ transition level is $\sim$271 meV below the CBM.
• Carrier Saturation: Free electron concentrations are strictly limited to approximately $3.4 \times 10^{14}$ cm$^{-3}$ at room temperature, regardless of whether the Si doping is light ($10^{16}$ cm$^{-3}$) or heavy ($10^{20}$ cm$^{-3}$).
• Mechanism: As Si concentration increases, the Fermi level is pinned near the $(+/−)$ level. Additional Si atoms simply convert into equal amounts of $Si^+$ (donors) and $Si^-$ (compensating acceptors), resulting in a near-zero activation ratio (<0.001% for heavy doping).
• Temperature Effect: While heating to 500 K increases activation (nearly 100% ionization for light doping), the carrier concentration remains capped by the self-compensation mechanism.

B. Silicon-Doped Al$_{0.91}$Ga$_{0.09}$N

• Alloying Benefit: Adding 9% Ga shifts the CBM such that the Si $(+/−)$ level coincides with the CBM.
• High Activation: This eliminates the energy barrier for self-compensation.
  • Light Doping: Nearly 100% of donors are activated.
  • Heavy Doping: Free electron concentrations scale with doping. At $10^{20}$ cm$^{-3}$ Si, the material achieves $4.2 \times 10^{18}$ cm$^{-3}$ free electrons.
• Trade-off: While conductivity improves, the authors note that alloy scattering may drastically reduce electron mobility (potentially to <20% of pure AlN values).

C. Silicon-Doped c-BN

• Intermediate Performance: The Si $(+/−)$ level is 110 meV below the CBM (closer than AlN, further than the alloy).
• Results:
  • Light Doping: High activation ($\sim$30$\times$ higher than AlN).
  • Heavy Doping: Carrier concentrations saturate at $\sim$1.5 $\times 10^{17}$ cm$^{-3}$ (at 300 K) even with $10^{20}$ cm$^{-3}$ Si doping.
  • Temperature: At 500 K, heavy doping yields a significant boost ($1.6 \times 10^{18}$ cm$^{-3}$), suggesting c-BN is a viable candidate for n-type devices if extrinsic compensation is minimized.

5. Significance and Conclusions

• Fundamental Limit: The study confirms that for AlN, simply increasing Si doping concentration is futile due to intrinsic DX self-compensation. The "best-case" free carrier concentration is capped at $\sim 10^{14}$ cm$^{-3}$.
• Strategic Pathways:
  • AlN: Only light doping is viable to avoid massive ionized impurity scattering and mobility collapse.
  • AlGaN: Alloying with Ga is the most effective strategy to overcome the DX barrier, enabling high carrier densities, though mobility trade-offs must be managed.
  • c-BN: Offers a promising middle ground, potentially supporting moderate n-type doping levels, especially at elevated temperatures.
• Implication for Device Design: To realize high-performance UWBG devices, researchers must either utilize alloying (AlGaN) to align the Fermi level or accept that pure AlN cannot support high-density n-type conduction via Si doping alone. The findings explain the non-monotonic behavior of conductivity in experimental studies and provide a theoretical ceiling for future device optimization.