Limited Diffusion of Silicon in GaN: A DFT Study Supported by Experimental Evidence

This study combines first-principles DFT calculations with ultra-high-pressure annealing experiments to demonstrate that silicon diffusion in gallium nitride is extremely limited due to prohibitively high activation barriers, thereby confirming the material's stability for precise doping in advanced electronic applications.

The Gist

Imagine Gallium Nitride (GaN) as a high-tech, ultra-durable city built for the future of electronics. It's the material that powers our bright LED lights and fast internet connections. To make this city work, engineers need to add "citizens" called Silicon (Si) atoms to specific neighborhoods. These Silicon atoms act as the electricity carriers (donors) that make the devices turn on.

The big question the researchers asked was: Once we place these Silicon citizens in their homes, do they stay put, or do they wander off?

In many materials, atoms are like restless tourists; if you heat them up, they start packing their bags and moving to new spots. This "wandering" (diffusion) is bad for electronics because it blurs the precise lines between different parts of a chip. The team wanted to know if Silicon in GaN is a "homebody" or a "traveler."

Here is what they found, explained simply:

1. The "Empty Seat" Theory (How Atoms Move)

To move from one spot to another in a crystal city, an atom usually needs an empty seat (a vacancy) next to it to jump into.

• The Study: The scientists used powerful computer simulations (like a super-accurate video game) to watch a Silicon atom try to jump into an empty seat.
• The Result: They found that the "stairs" the Silicon atom has to climb to make that jump are incredibly high.
  • Moving sideways (along the city streets) requires climbing a 3.2 eV wall.
  • Moving up or down (vertical) requires climbing a 3.8 eV wall.
  • Moving diagonally across the city is even harder, requiring a 10 eV wall.

The Analogy: Imagine trying to push a heavy boulder up a mountain. Even if you give the boulder a massive shove (heating the material to extreme temperatures), it barely moves because the mountain is just too steep.

2. The "Direct Swap" and "Group Dance" Failures

The researchers also checked if Silicon could move by swapping places directly with a neighbor or by doing a complex "group dance" with three atoms at once.

• The Result: These methods were even more impossible. The energy required was like trying to jump over a skyscraper (over 12 eV).
• Conclusion: Silicon is stuck. It simply won't move unless it finds a very specific, empty seat, and even then, the climb is too steep.

3. The "Extreme Heat" Test (The Experiment)

Computer models are great, but the team wanted real-world proof. They took actual GaN crystals, implanted Silicon into them, and then subjected them to Ultra-High-Pressure Annealing (UHPA).

• The Setup: Think of this as putting the crystals in a pressure cooker that is also a furnace. They heated them to over 1300°C (hotter than a pizza oven) and squeezed them with immense pressure (1 GPa) for 30 minutes to 3 hours.
• The Test: They used a special microscope (SIMS) to take a "before and after" photo of where the Silicon was.
• The Result: The Silicon didn't budge. The "before" and "after" photos looked exactly the same. Even after being cooked and squeezed, the Silicon stayed exactly where they put it.

4. Why This Matters

The paper concludes that Silicon in Gallium Nitride is an extremely loyal citizen.

• No Wandering: Unlike some other materials where atoms get restless and blur the lines when heated, Silicon in GaN stays put.
• Precision: This means engineers can create very sharp, precise boundaries in their electronic devices without worrying that the heat of the manufacturing process will smear the design.
• Consistency: It doesn't matter if the crystal was grown on a sapphire floor or a GaN floor, or if the Silicon was implanted lightly or heavily; the Silicon just refuses to move.

In a Nutshell:
The researchers proved that Silicon in Gallium Nitride is like a stone statue in a hurricane. No matter how hot or how much pressure you apply, it stays exactly where it belongs. This makes GaN a perfect, stable foundation for building the next generation of fast, powerful, and precise electronic devices.

Technical Summary

Technical Summary: Limited Diffusion of Silicon in GaN: A DFT Study Supported by Experimental Evidence

Problem Statement
Silicon (Si) serves as the primary intentional donor for fabricating n-type Gallium Nitride (GaN), a material critical for high-power and high-frequency optoelectronic and electronic devices. While Si doping is routinely achieved via in-situ incorporation during epitaxial growth or ion implantation, precise control over the lateral spatial distribution of doped regions remains a challenge. Although ion implantation offers precise doping, it requires high-temperature annealing to repair lattice damage and activate dopants. A critical concern in this process is the potential for Si diffusion, which could blur sharp doping profiles essential for advanced device architectures.

Existing literature presents conflicting reports regarding Si diffusion in GaN. Some studies suggest measurable diffusion at high temperatures with low activation energies (e.g., 0.89–1.55 eV), while others indicate negligible diffusion or exceptionally high activation enthalpies (e.g., ~10 eV) mediated by specific defect pairs. Furthermore, the behavior of Si under Ultra-High-Pressure Annealing (UHPA) conditions—used to stabilize GaN at temperatures exceeding typical growth limits—requires clarification. This study aims to resolve these discrepancies by investigating the diffusion mechanisms of Si in GaN using first-principles calculations and validating them with experimental UHPA data.

Methodology
The research employs a dual approach combining Density Functional Theory (DFT) calculations with experimental characterization.

• Theoretical Framework:

  • Computational Setup: Calculations were performed using the SIESTA software package within the Generalized Gradient Approximation (GGA) using the PBE potential. A supercell model consisting of 324 atoms (3a√3 × 3a√3 × 3c GaN unit cells) was utilized.
  • Defect Modeling: The study modeled a substitutional Si atom on a Ga site ($Si_{Ga}$) and a Gallium vacancy ($V_{Ga}$). Both neutral and n-type (triply negatively charged $V^{3-}_{Ga}$) scenarios were simulated to reflect experimental doping conditions.
  • Diffusion Mechanisms: The Nudged Elastic Band (NEB) method was used to determine Minimum Energy Paths (MEPs) for Si migration. Three crystallographic directions were analyzed: a-direction [11$\bar{2}$0], c-direction [0001], and m-direction [$\bar{1}$100]. Both vacancy-mediated mechanisms and direct exchange (D-Ex) or ring-migration mechanisms were investigated.
  • Thermodynamics: Phonon calculations were conducted to determine vibrational free energy contributions ($\Delta F_{vib}$), allowing for the calculation of temperature-dependent effective energy barriers and diffusion coefficients ($D$) based on transition state theory.

• Experimental Validation:

  • Samples: Si-implanted GaN samples were prepared using both Metal-Organic Vapor Phase Epitaxy (MOVPE) and Hydride Vapor Phase Epitaxy (HVPE) on various substrates (Ammono-GaN and sapphire) with varying threading dislocation densities (TDD).
  • Process: Samples underwent UHPA at temperatures up to 1300°C (and higher for specific durations) under 1 GPa of N$_2$ pressure.
  • Characterization: X-Ray Diffraction (XRD) was used to assess crystalline recovery, while Secondary Ion Mass Spectrometry (SIMS) was employed to measure Si concentration profiles before and after annealing to detect any diffusion.

Key Results

1. Energy Barriers and Mechanisms:

   • Vacancy-Mediated Diffusion: The primary diffusion mechanism is vacancy-mediated. The calculated energy barriers are high: approximately 3.2 eV along the a-direction and 3.8 eV along the c-direction for n-type systems. The m-direction barrier is exceptionally high (~10 eV), making direct long-range migration along this axis highly improbable.
   • Alternative Mechanisms: Direct exchange and ring-migration mechanisms without vacancies were found to have prohibitively high energy barriers (11.8–14.8 eV), rendering them unlikely pathways.
   • Temperature Dependence: While vibrational free energy reduces the effective barrier slightly (by ~0.25 eV at 1800 K for neutral systems), the barriers remain high. In n-type systems, the reduction is negligible.
   • Diffusion Coefficients: The calculated diffusion coefficients ($D$) do not exceed $10^{-9}$ cm$^2$/s even at 1800 K. When accounting for realistic vacancy concentrations (typically $<10^{18}$ cm$^{-3}$), the effective diffusion rate is expected to be at least three orders of magnitude lower.

2. Experimental Observations:

   • Structural Recovery: UHPA successfully removed implantation-induced damage and restored high-quality crystalline structures in MOVPE and HVPE samples.
   • Si Profile Stability: SIMS depth profiles showed no detectable broadening or shift in Si concentration after UHPA, regardless of the growth method (MOVPE vs. HVPE), substrate type, TDD (ranging from $5 \times 10^4$ to $10^8$ cm$^{-2}$), or implantation parameters (energy up to 300 keV, dose up to $1 \times 10^{15}$ cm$^{-2}$).

Significance and Claims
The paper claims to resolve previous inconsistencies in the literature regarding Si diffusion in GaN. By combining rigorous DFT calculations with high-pressure experimental validation, the authors demonstrate that:

• Si diffusion in GaN is intrinsically limited due to high energy barriers, particularly for vacancy-mediated processes.
• The diffusion is anisotropic, with lateral (a-direction) diffusion being faster than vertical (c-direction) diffusion, yet both remain negligible under standard and extreme processing conditions.
• Si profiles remain stable even under UHPA conditions (temperatures >1100°C, high pressure), contradicting reports of measurable diffusion in similar ranges.
• Factors such as threading dislocation density, growth method, and substrate type do not significantly enhance Si mobility.

The study concludes that Si doping in GaN is stable for device fabrication applications requiring precise n-type doping, as the risk of diffusion-related broadening of doped regions is minimal. This provides a comprehensive theoretical and experimental basis for understanding Si behavior in GaN, supporting the development of reliable GaN-based electronic and optoelectronic devices.