Thermal Conductivity and Temperature-Induced Band Gap Renormalization in Crystalline and Amorphous Ga$_2$O$_3$

This study employs a machine-learned moment tensor potential coupled with first-principles calculations to reveal that crystalline $\beta$-Ga$_2$O$_3$ exhibits significantly stronger temperature-induced band gap renormalization and higher lattice thermal conductivity compared to its amorphous counterpart, establishing a reliable computational framework for predicting semiconductor properties under microelectronic operating conditions.

The Gist

Imagine Gallium Oxide (Ga₂O₃) as a super-advanced, high-performance building material for the next generation of electronics. It's like the "super-concrete" used to build faster, more powerful computers and lights. But just like real concrete, this material has two very different personalities depending on how it's built:

1. Crystalline (β-Ga₂O₃): Think of this as a perfectly organized city with a grid of streets. Every atom is in its exact, pre-planned spot.
2. Amorphous (a-Ga₂O₃): This is like a chaotic pile of bricks dumped on the ground. The atoms are there, but they're jumbled up with no long-range order.

The scientists in this paper wanted to understand two critical things about these two versions of the material:

• How well do they conduct heat? (If they get too hot, the electronics break.)
• How does their "electrical personality" change when they get hot? (This is called "Band Gap Renormalization," but let's call it the "Electronic Mood Swing.")

The Problem: The "Supercomputer" Bottleneck

To figure out how atoms behave, scientists usually use a method called DFT (Density Functional Theory). Think of DFT as a master chef who cooks every single meal from scratch, tasting every ingredient to get it perfect. It's incredibly accurate, but it takes forever. If you want to simulate a whole city of atoms moving around, the chef would need to cook for a million years.

To solve this, the researchers used a Machine-Learning Potential (MTP). Think of the MTP as a brilliant sous-chef who has watched the master chef cook a million times. The sous-chef can predict exactly what the meal will taste like in a split second, with 99% of the accuracy of the master chef, but in a fraction of the time. This allowed the team to run massive simulations that would have been impossible otherwise.

Discovery 1: The "Electronic Mood Swing" (Band Gap Renormalization)

Every semiconductor has a "gap" between its energy levels. Think of this gap as a fence that electrons must jump over to conduct electricity.

• The Cold Reality: Even at absolute zero (0 Kelvin), the atoms aren't perfectly still; they are vibrating due to quantum mechanics. This "zero-point vibration" shrinks the fence by about 0.2 eV. That's a big deal! It means the material behaves differently even when it's "off."
• The Heat Effect: As the material heats up (like a computer running a heavy game), the atoms vibrate more wildly.
  • In the Crystalline (Organized) version: The fence shrinks dramatically. At 700 K (about 800°F), the gap shrinks by nearly 0.45 eV. The organized city allows these vibrations to sync up and change the material's properties significantly.
  • In the Amorphous (Chaotic) version: The fence also shrinks, but less than in the organized version. The chaotic pile of bricks dampens the vibrations. The "mood swing" is weaker.

The Takeaway: If you are designing a device, you cannot ignore how heat changes the electrical properties. The material changes its mind as it gets hot, and this change is different depending on whether the atoms are organized or messy.

Discovery 2: The Heat Traffic Jam (Thermal Conductivity)

Now, let's talk about heat. Heat moves through solids via phonons, which are like sound waves or ripples traveling through the atomic structure.

• Crystalline Ga₂O₃: Because the atoms are in a perfect grid, the ripples (phonons) travel smoothly and quickly, like a high-speed train on a straight track. This material conducts heat very well.
• Amorphous Ga₂O₃: Because the atoms are jumbled, the ripples hit a wall, get scattered, and get stuck. It's like trying to run through a crowded, messy room full of furniture. The heat energy gets trapped in place.

The Result: The amorphous (messy) version conducts heat about 10 times worse than the organized version.

• The organized version is a highway for heat.
• The messy version is a traffic jam.

This is actually a double-edged sword. If you want a heat sink (to cool things down), you want the organized version. If you want a thermal insulator (to keep heat in, like in a sensor), the messy version is great.

The Big Picture

This paper is a victory for smart computing. By using a "sous-chef" (Machine Learning) to mimic the "master chef" (Quantum Physics), the researchers could:

1. Prove that even tiny, invisible vibrations at absolute zero change how the material works.
2. Show that making the material messy (amorphous) drastically changes how it handles heat and electricity.
3. Provide a reliable recipe for engineers to design better electronics, knowing exactly how these materials will behave when they get hot.

In short: Order conducts heat well but changes its electrical mind easily when hot. Chaos traps heat but stays more stable electrically. Knowing this helps us build better gadgets for the future.

Technical Summary

1. Problem Statement

Gallium oxide ($\text{Ga}_2\text{O}_3$), particularly its $\beta$-phase, is a critical wide-bandgap semiconductor for high-power electronics, optoelectronics, and sensors. However, its performance under operating conditions is heavily influenced by two temperature-dependent phenomena:

1. Lattice Thermal Conductivity (LTC): Essential for heat dissipation in devices.
2. Band Gap Renormalization (BGR): The shift in the electronic band gap due to electron-phonon coupling and lattice vibrations, which affects device efficiency and optical properties.

The Challenge: Accurately calculating these properties using traditional ab initio Density Functional Theory (DFT) is computationally prohibitive. DFT requires large supercells and long molecular dynamics (MD) trajectories to capture anharmonic effects and temperature dependence, making it impractical for complex systems like amorphous materials or high-temperature simulations.

2. Methodology

The authors developed a hybrid computational framework combining Machine-Learned Interatomic Potentials (MLIPs) with first-principles DFT to overcome computational limitations.

• Potential Model: They utilized Moment Tensor Potentials (MTP), a type of MLIP trained on DFT data. This allows for DFT-level accuracy in predicting interatomic energies and forces at a fraction of the computational cost.
• Training & Validation:
  • $\beta$-$\text{Ga}_2\text{O}_3$ (Crystalline): MTPs were trained on DFT data from unit cell optimizations and MD trajectories. Validation confirmed high agreement in forces, unit cell parameters, and phonon band structures compared to DFT.
  • Amorphous $\text{Ga}_2\text{O}_3$ ($a$-$\text{Ga}_2\text{O}_3$): An active learning workflow was employed. Starting from random packing, the MTP was iteratively refined using MD simulations at high temperatures (3000–400 K) to accurately model the quenching process and generate physically realistic amorphous structures.
• Property Calculation:
  • Band Gap Renormalization (BGR): Calculated using the Allen-Heine-Cardona theory approach. The study distinguished between:
    • Harmonic Approximation: Using normal mode superposition (HIPHIVE package).
    • Anharmonic Effects: Using classical NVT MD trajectories (LAMMPS) to capture nuclear motion anharmonicity and lattice thermal expansion.
    • Quantum Effects: Incorporated zero-point energy corrections.
  • Lattice Thermal Conductivity (LTC): Calculated for amorphous $\text{Ga}_2\text{O}_3$ using the Green-Kubo method, which involves integrating the heat current autocorrelation function (HCACF) from MD trajectories.

3. Key Contributions

• Framework Development: Established a computationally tractable MTP-based workflow capable of predicting both thermal and electronic properties for both crystalline and amorphous semiconductors under realistic operating conditions.
• Comprehensive BGR Analysis: Provided a systematic breakdown of BGR in $\beta$-$\text{Ga}_2\text{O}_3$, quantifying the specific contributions of zero-point vibrations, classical nuclear motion, anharmonicity, and thermal expansion.
• Amorphous vs. Crystalline Comparison: Directly compared the thermal and electronic behaviors of crystalline $\beta$-$\text{Ga}_2\text{O}_3$ and amorphous $\text{Ga}_2\text{O}_3$ within the same theoretical framework, a rare feat due to the difficulty of modeling amorphous systems ab initio.

4. Key Results

A. Band Gap Renormalization (BGR)

• Crystalline $\beta$-$\text{Ga}_2\text{O}_3$:
  • Exhibits a significant zero-point renormalization (ZPR) of $\sim 0.2$ eV, comparable to materials like SiC and GaN. This highlights the necessity of including quantum nuclear effects even at low temperatures.
  • The total BGR at 700 K is approximately 0.45 eV.
  • Anharmonicity: The study found that lattice anharmonicity and thermal expansion have a minor influence on BGR in $\beta$-$\text{Ga}_2\text{O}_3$. The material behaves largely harmonically (anharmonicity measure $\sigma_A < 0.34$), meaning the temperature dependence is dominated by harmonic phonon modes.
  • The final calculated BGR ($BGR_{anh+Q}$) aligns well with experimental data.
• Amorphous $\text{Ga}_2\text{O}_3$:
  • Shows a weaker temperature dependence of BGR compared to the crystalline phase.
  • At 900 K, the BGR is 0.48 eV for amorphous vs. 0.66 eV for crystalline $\beta$-$\text{Ga}_2\text{O}_3$.
  • This indicates that structural disorder (amorphization) reduces the sensitivity of the band gap to temperature-induced lattice vibrations.

B. Lattice Thermal Conductivity (LTC)

• Amorphous $\text{Ga}_2\text{O}_3$:
  • The LTC remains nearly constant at $\sim 0.9 \text{ W m}^{-1} \text{K}^{-1}$ across the 300–700 K range.
  • This value is approximately one order of magnitude lower than that of crystalline $\beta$-$\text{Ga}_2\text{O}_3$.
• Mechanism: The drastic reduction in LTC for the amorphous phase is attributed to the loss of long-range structural order, which strongly localizes lattice vibrations (phonons), trapping thermal energy and suppressing heat propagation.
• Validation: The calculated LTC values match the order of magnitude of experimental data, validating the MTP approach for disordered systems.

5. Significance

• Design Guidelines for Microelectronics: The study confirms that for $\text{Ga}_2\text{O}_3$ devices, ignoring temperature effects leads to significant errors in predicting band gaps (up to 0.45 eV shift). Furthermore, the choice between crystalline and amorphous phases drastically alters thermal management strategies.
• Methodological Advancement: By demonstrating that MTPs can accurately model both harmonic/anharmonic electronic properties and thermal transport in disordered systems, the paper provides a scalable route for simulating next-generation semiconductor materials where DFT is too expensive.
• Material Physics Insight: The work clarifies that while $\beta$-$\text{Ga}_2\text{O}_3$ is a "harmonic" material regarding band gap renormalization, its amorphous counterpart exhibits distinct thermal transport physics driven by phonon localization, offering new insights for designing thermal barrier coatings or memory devices.