Surface Modification for III-V Selective Area Molecular Beam Epitaxy of Non-Selective Mask Materials

This study demonstrates that depositing a sub-1 nm silicon dioxide capping layer enables selective area molecular beam epitaxy of III-V semiconductors on highly reactive or non-selective mask materials like $TiO_2$ and $HfO_2$, thereby overcoming the optical limitations of traditional masks without degrading their spectral performance.

The Gist

Imagine you are a master chef trying to bake a perfect, single-layer cake (a crystal) on top of a specific patch of a baking sheet, while leaving the rest of the sheet completely empty. This is essentially what scientists do when they build advanced light-based devices using a technique called Molecular Beam Epitaxy (MBE). They want to grow a crystal only where they want it, and they use a "mask" (like a stencil) to cover the areas where they don't want the crystal to grow.

For a long time, chefs have only used two types of stencils: Silica (SiO₂) and Silicon Nitride (Si₃N₄). These are great because they are "inert," meaning the hot crystal ingredients don't stick to them; they just slide right off. However, these old stencils have a problem: they are like dark sunglasses that block too much light. If you want to build devices that work with specific infrared light (like the kind used in night vision or high-speed data), these old stencils absorb the light and ruin the design.

The scientists in this paper asked: *"Can we use different, clearer stencils made of materials like Aluminum Oxide (Al₂O₃), Titanium Dioxide (TiO₂), or Hafnium Oxide (HfO₂)?"*

Here is what they found, broken down simply:

1. The "Try-Out" Phase: Testing New Stencils

They tried growing the crystal on these new materials to see if the crystal would stick to the mask or slide off.

• Aluminum Oxide (Al₂O₃): This was the star of the show. It behaved very much like the old trusted Silica stencil. At the right temperatures, the crystal ingredients slid right off it, allowing for clean growth. It's a promising new option.
• Hafnium Oxide (HfO₂): This one was a disaster. It was like a sticky trap. The crystal ingredients stuck to it immediately, no matter how hot they made the oven. Instead of a clean crystal, they got a messy, crumbly pile of crystals (polycrystalline material) all over the mask.
• Titanium Dioxide (TiO₂): This one was even worse. It didn't just get sticky; it reacted chemically with the ingredients. It was like the stencil itself started melting or changing when the hot ingredients hit it.

2. The "Why": It's All About the Surface

The scientists looked closely at the surface of these materials. They found that the "stickiness" wasn't because the stencils were rough (they were all smooth). It was about the chemistry of the surface.

• The "bad" stencils had tiny, hungry spots (called oxygen vacancies or hydroxyl groups) that grabbed onto the crystal ingredients.
• The "good" stencils (like Silica) had a calm surface that didn't want to grab anything.

3. The Magic Trick: The "Silica Cap"

Since they really wanted to use the new materials (because they are clearer and better for light), they needed a way to stop the "sticky" ones from grabbing the crystals.

They came up with a clever solution: The Thin Coat.

Imagine you have a very sticky piece of tape (the bad mask). You can't use it directly, but if you put a very thin, non-sticky sheet of plastic (a layer of Silica) over it, the tape underneath can't grab anything anymore.

• The Experiment: They took the sticky TiO₂ and the reactive Si₃N₄ masks and covered them with a microscopic layer of Silica (just a few nanometers thick—thinner than a human hair).
• The Result: Suddenly, the sticky masks behaved exactly like the perfect Silica mask! The crystal ingredients slid right off. Even a layer as thin as 0.9 nanometers (less than 10 atoms thick) was enough to change the surface chemistry completely.

The Bottom Line

This paper shows that we don't have to be stuck using the old, light-blocking stencils.

1. Aluminum Oxide is already a great alternative.
2. For the other materials that are too "sticky" or reactive, we can simply paint them with a microscopic layer of Silica.

This trick turns any material into a "Silica-like" surface, allowing scientists to use a wider variety of materials to build better, clearer, and more advanced light-based devices, without ruining the growth process.

Technical Summary

Technical Summary: Surface Modification for III-V Selective Area Molecular Beam Epitaxy of Non-Selective Mask Materials

Problem Statement
Selective-area embedded regrowth of III-V semiconductors via Molecular Beam Epitaxy (MBE) enables the seamless integration of metals and dielectrics into crystalline material, offering a unique design space for optoelectronic and photonic devices. However, current device designs are constrained by the limited library of mask materials, primarily Silicon Dioxide (SiO₂) and Silicon Nitride (Si₃N₄). While these materials are chemically inert, they possess high extinction coefficients in the 8–12 µm wavelength range, prohibiting the design of embedded high-contrast photonics in the long-wave infrared. Furthermore, the development of embedded metasurfaces requires mask materials with diverse refractive indices and low extinction coefficients.

The challenge lies in the fact that alternative dielectric materials (such as Al₂O₃, TiO₂, and HfO₂) often lack the deposition selectivity of SiO₂. In MBE, poor selectivity leads to the adsorption of adatoms on the mask surface, resulting in the parasitic formation of polycrystalline material rather than selective growth on the exposed semiconductor windows. Historically, achieving all-MBE selective-area regrowth with non-inert masks has been difficult, limiting the expansion of III-V embedded oxide capabilities.

Methodology
The study evaluated the deposition selectivity of emerging mask materials—Al₂O₃, TiO₂, and HfO₂—compared to established SiO₂ and Si₃N₄ masks.

• Film Fabrication: SiO₂ and Si₃N₄ were deposited via Plasma-Enhanced Chemical Vapor Deposition (PECVD). Al₂O₃, HfO₂, and TiO₂ were deposited via Atomic Layer Deposition (ALD). All films underwent UV-Ozone irradiation to remove residual carbon.
• MBE Growth: Samples were baked under an atomic hydrogen flux and then subjected to co-deposition of Gallium (Ga) and Arsenic (As₄) at substrate temperatures ranging from 570 °C to 630 °C. A growth rate of 0.25 µm/hr was used to deposit 100 nm of GaAs.
• Characterization: Polycrystalline surface coverage was quantified using Scanning Electron Microscopy (SEM) and MATLAB processing. Surface morphology was analyzed via Atomic Force Microscopy (AFM), and surface chemistry/bonding states were probed using X-ray Photoelectron Spectroscopy (XPS).
• Surface Modification: To address non-selective behavior, a "surface modification" technique was employed. This involved depositing a thin SiO₂ capping layer (via PECVD or ALD) onto non-selective or reactive films (Si₃N₄, TiO₂, and Al₂O₃) to alter surface chemistry without significantly degrading optical response.

Key Results

1. Selectivity Survey of Unmodified Films:

   • Al₂O₃: Exhibited promising selective growth characteristics. At 630 °C, surface coverage approached near-zero values, suggesting high adatom mobility. The trend was similar to SiO₂ but required higher temperatures, likely due to Ga preference for surfaces containing Group III atoms.
   • HfO₂: Demonstrated high non-selectivity. Surface coverage remained above 0.8 up to 630 °C, and polycrystalline rings appeared in Reflection High-Energy Electron Diffraction (RHEED) after only 2–3 nm of deposition. XPS revealed oxygen vacancies and suboxide states, indicating a reactive surface.
   • TiO₂: Proved highly reactive. Growth resulted in disparate nuclei morphologies and a rapid apparent temperature increase (~50 °C) during deposition, indicative of interfacial reactions between Ga and the TiO₂ surface.
   • Si₃N₄: Showed lower temperature sensitivity than oxides but still required higher minimum temperatures or lower growth rates than SiO₂ for selectivity. XPS indicated the presence of SiO₂ and SiON groups and dangling bonds acting as Ga bonding sites.

2. Surface Roughness vs. Chemistry:

   • AFM analysis showed that all emerging films had sub-nanometer root-mean-square (RMS) roughness, comparable to SiO₂. The study concluded that surface morphology was not the driver of selectivity differences; rather, surface composition and bonding states were the primary factors.

3. Surface Modification via SiO₂ Capping:

   • Applying a 15 nm SiO₂ capping layer to Si₃N₄ and TiO₂ films successfully mitigated polycrystalline deposition under conditions that previously failed. The capped films exhibited nuclei morphology and surface coverage nearly identical to pure SiO₂.
   • Thickness Dependence (Al₂O₃ System): Varying the SiO₂ cap thickness on Al₂O₃ from 0.9 nm to 9.4 nm revealed that sub-1 nm caps significantly influenced mask surface chemistry. While selectivity improved linearly with thickness from ~1 to 10 nm, a sub-nanometer cap was sufficient to decouple the selective growth properties from the underlying film's optical response. This allows for "SiO₂-like" selectivity on any mask material without degrading its optical performance.

Significance and Claims
The paper claims to demonstrate a viable pathway for expanding the library of mask materials for III-V MBE selective-area growth. The primary contribution is the validation that highly non-selective or reactive dielectric materials (like TiO₂ and HfO₂) can be rendered compatible with standard GaAs/SiO₂ growth regimes through a thin SiO₂ surface modification layer.

This technique allows researchers to leverage the desirable optical properties (refractive index, low extinction coefficient) of alternative oxides while maintaining the high optical quality and band-engineering capabilities associated with the inert SiO₂ growth environment. The study suggests that by modifying surface chemistry rather than bulk material properties, it is possible to achieve seamless integration of diverse dielectrics into crystalline III-V structures, enabling advanced designs for embedded photonics, metasurfaces, and optoelectronic devices that were previously limited by the optical constraints of traditional SiO₂ and Si₃N₄ masks.