Uncovering the properties of homo-epitaxial GaN devices through cross-sectional infrared nanoscopy

This paper demonstrates that combining mid-infrared and terahertz scattering-type scanning near-field optical microscopy (s-SNOM) enables high-resolution, non-destructive characterization of homo-epitaxial GaN p-i-n diodes by successfully disentangling carrier and lattice properties and detecting sub-surface defects with superior sensitivity compared to traditional metrologies like micro-Raman and KPFM.

The Gist

Imagine you are trying to fix a very complex, high-tech engine, but you can't take it apart. You need to know exactly where the fuel is flowing, where the metal is stressed, and if there are tiny cracks hiding deep inside.

This paper is about a new, super-powerful "X-ray vision" tool that scientists used to look inside a specific type of engine part called a Gallium Nitride (GaN) diode. These are the super-charged components used in next-generation electronics, like fast chargers and electric vehicles.

Here is the story of how they did it, explained simply:

1. The Problem: The "Black Box"

Gallium Nitride is a superstar material for electronics because it handles high power and heat better than old-school silicon. But to make it perfect, scientists need to grow it in layers, like a very precise cake. Sometimes, the "batter" isn't mixed right, or there are tiny cracks (defects) hidden deep inside the layers.

Traditional tools for looking at these materials have flaws:

• X-rays can see the structure but are too blurry to see tiny defects.
• Microscopes can see the surface but can't see what's buried underneath.
• Standard lasers are like trying to read a book through a foggy window; they can't get close enough to see the fine print.

2. The Solution: The "Super-Tip" Flashlight

The researchers used a technique called s-SNOM (scattering-type scanning near-field optical microscopy).

Think of this like a super-sensitive, ultra-fine paintbrush that is also a flashlight.

• Instead of shining a big beam of light (which spreads out and gets blurry), this tool uses a needle so sharp it's like a single atom.
• It taps the surface of the material very quickly, creating a tiny "bubble" of light right at the tip.
• Because the bubble is so small, it can see details as tiny as a virus, far smaller than what normal microscopes can see.

3. The Secret Sauce: Two Different Colors of Light

The real magic in this paper is that they didn't just use one color of light; they used two different "frequencies" (like two different radio stations) to get a complete picture.

• Station A (The "Traffic" Light - Terahertz):
  Imagine this frequency is like a traffic camera that only sees cars (electrons). It tells you exactly how many cars are driving on the road (carrier density) but ignores the condition of the road itself.

  • What it found: It showed where the "traffic" was heavy (highly doped areas) and where it was light.

• Station B (The "Road" Light - Mid-Infrared):
  This frequency is like a mechanic inspecting the road surface (the crystal lattice). It cares about how the road is built and if it's bumpy or cracked. However, it also sees the cars, so the picture gets a bit mixed up.

  • What it found: It showed the layers of the device clearly, but it also picked up weird "bumps" that looked like traffic jams but weren't.

The "Aha!" Moment:
By comparing the two stations, the scientists realized something amazing.

• If the "Traffic Light" and "Road Light" agreed, it was just a change in how many electrons were there.
• If they disagreed (e.g., the Road Light saw a huge bump, but the Traffic Light saw normal traffic), they knew it wasn't a traffic jam. It was a crack in the road (a lattice defect or strain).

This allowed them to separate "how many electrons" from "how broken the material is," which is something no other tool could do so clearly.

4. The Results: Finding the Hidden Flaws

Using this dual-light approach, the team looked at a slice of the GaN device (like looking at a slice of a layered cake).

• They mapped the layers: They could clearly see the different layers of the device, even though they were buried deep inside.
• They found the ghosts: They spotted tiny, line-like defects in the substrate (the base material) that other tools missed. These were like invisible cracks in the foundation of a building.
• They proved it works: They compared their results to older methods (like Raman spectroscopy and KPFM). The older methods were like looking at a map from 10 miles away; they could see the big cities (layers) but missed the potholes (defects). The new s-SNOM tool was like walking the streets with a magnifying glass.

The Takeaway

This paper shows that by using a "super-tip" flashlight and listening to two different "radio stations" (light frequencies) at the same time, scientists can finally see the hidden secrets inside high-tech electronics.

It's like having a detective who can not only count the people in a room but also tell you if the floorboards are rotting underneath them. This helps engineers build better, more reliable, and faster electronics for the future.

Technical Summary

Here is a detailed technical summary of the paper "Uncovering the properties of homo-epitaxial GaN devices through cross-sectional infrared nanoscopy."

1. Problem Statement

The development of high-performance Gallium Nitride (GaN) power devices relies heavily on homoepitaxial growth on native GaN substrates to reduce dislocation densities and improve reliability. However, current GaN substrate technology is less mature than Silicon Carbide (SiC), leading to defects, doping variations, and inconsistent device performance.

Characterizing these devices presents significant challenges:

• Limitations of Existing Techniques:
  • X-ray Topography: Lacks the spatial resolution to detect individual point/line defects.
  • Electron Microscopy (TEM/EBIC): Provides nanoscale defect visualization but offers poor contrast for doping variations and requires destructive sample preparation.
  • Atomic Force Microscopy (AFM/KPFM): KPFM maps surface potential (doping) but lacks information on local strain and lattice properties.
  • Raman Spectroscopy: Can probe carrier density and strain via phonon-plasmon coupling (LOPC) but is diffraction-limited (spatial resolution ~500 nm) and suffers from depth sensitivity issues, making it difficult to resolve thin buried layers or localized defects in cross-sections.
• The Core Challenge: Distinguishing between changes in carrier density (doping) and lattice properties (strain/defects) is difficult because both affect the optical response in polar semiconductors like GaN. Furthermore, analyzing buried heterostructures in cross-section requires techniques with high spatial resolution and depth sensitivity that exceed the diffraction limit.

2. Methodology

The authors employed Scattering-type Scanning Near-field Optical Microscopy (s-SNOM) across a broad spectral range, from Terahertz (THz) to Mid-Infrared (MIR), to characterize a cross-sectioned GaN p-i-n diode.

• Sample Preparation: A vertical GaN PIN diode (grown on an N-type GaN substrate) was embedded in epoxy, cross-sectioned, and polished to a sub-micron RMS roughness (~0.36 µm) to expose the epitaxial layers (n++ contact, 8µm drift layer, and p-type cap).
• s-SNOM Setup:
  • Utilized a free-electron laser (FELBE) tunable from 40 cm⁻¹ to 2000 cm⁻¹.
  • An AFM tip oscillated at ~256 kHz, interacting with the sample to generate evanescent near-fields.
  • Signals were demodulated at the 3rd harmonic to enhance surface sensitivity and suppress far-field background.
  • Measurements were performed at specific wavenumbers: THz range (115 cm⁻¹, 288 cm⁻¹) and MIR range (535 cm⁻¹, 683 cm⁻¹, 695 cm⁻¹, 712 cm⁻¹).
• Theoretical Modeling: A Point-Dipole Model (PDM) was used to simulate the tip-sample interaction. This model combined a Drude model (for free carriers) and a TO-LO model (for lattice phonons) to predict how carrier density and lattice strain influence the near-field scattering amplitude.
• Validation: Results were cross-validated against Micro-Raman mapping (532 nm excitation) and Kelvin Probe Force Microscopy (KPFM) on the same sample regions.

3. Key Contributions

• Multi-Spectral Disentanglement: The study demonstrates that combining THz and MIR s-SNOM allows for the separation of carrier density effects from lattice/strain effects.
  • THz Range: Primarily sensitive to free carrier concentration (plasma frequency) with minimal lattice interaction.
  • MIR Range: Sensitive to both carrier density and lattice properties via the Longitudinal Optical Phonon-Plasmon Coupling (LOPC) effect near the LO phonon resonance (~730–750 cm⁻¹).
• Sub-Surface Defect Detection: The technique successfully identified localized line defects and strain fields within the substrate that are invisible to conventional Raman and KPFM.
• Cross-Sectional Nanoscopy: Successfully mapped the vertical doping profile of a buried p-i-n structure with ~20 nm spatial resolution, overcoming the diffraction limit of far-field optics.

4. Key Results

• Layer Resolution:
  • THz (115 cm⁻¹): Showed contrast between the highly doped substrate and the drift layer, consistent with carrier density differences.
  • MIR (695–712 cm⁻¹): Provided superior contrast near the LO phonon resonance. The 712 cm⁻¹ scan fully resolved the n++ interface, the 8µm drift layer, and the substrate.
• Defect Identification:
  • At 712 cm⁻¹, "line-like" features were observed in the substrate.
  • Contradictory Signals: These features appeared as low-signal regions in the MIR (suggesting high carrier density or strain) but as high-signal regions in the THz (suggesting low carrier density).
  • Resolution: The authors attributed this to lattice distortion/strain altering the phonon resonance, which affects the MIR response differently than the THz response. This distinction was only possible by comparing the two spectral ranges.
• Comparison with Other Techniques:
  • Raman: Confirmed the general doping profile (substrate vs. drift) but failed to resolve the n++ interface clearly and could not detect the localized line defects due to diffraction limits and peak broadening.
  • KPFM: Mapped the surface potential difference between layers (~0.2–0.8 V) but failed to detect the localized defects or lattice strain variations seen in s-SNOM.
• Sensitivity: The s-SNOM technique resolved carrier density changes on the order of 10¹⁸ cm⁻³ and detected sub-surface defects that other non-destructive techniques missed.

5. Significance

• Advanced Device Characterization: This work establishes s-SNOM as a superior tool for characterizing wide-bandgap (WBG) semiconductor devices, particularly for identifying failure mechanisms like localized strain and defects that degrade device performance.
• Methodological Framework: It provides a generalizable framework for analyzing polar semiconductors (e.g., AlN, Ga₂O₃) where distinguishing between doping and strain is critical.
• Non-Destructive Depth Profiling: The ability to image buried heterostructures in cross-section with nanoscale resolution offers a powerful alternative to destructive electron microscopy or low-resolution optical methods.
• Optimization of Growth: By revealing subtle variations in homoepitaxial layers and substrate defects, this technique can guide the optimization of GaN substrate growth and device fabrication processes, leading to more reliable power electronics.

In conclusion, the paper proves that broadband s-SNOM is uniquely capable of decoupling carrier and lattice dynamics in GaN devices, offering a level of detail and sensitivity unattainable by traditional metrologies.