Systematic study of high performance GeSn photodiodes with thick absorber for SWIR and extended SWIR detection

This paper presents a systematic empirical study of high-performance GeSn photodiodes with thick absorbers (up to 2630 nm) and varying tin content (2–8%), demonstrating high responsivity and extended SWIR detection capabilities while analyzing the impact of doping design and defects to propose optimization strategies for commercial-grade devices.

The Gist

Imagine you are trying to build a super-sensitive camera that can "see" in the dark, specifically in a range of light called Short-Wave Infrared (SWIR). This isn't just for night vision; it's the kind of vision that lets self-driving cars "see" through fog, smoke, or dust, and helps environmental sensors spot pollution from miles away.

Currently, the best cameras for this job use a material called InGaAs. It's great, but it's like trying to build a Ferrari engine inside a Toyota factory—it's expensive, hard to make, and doesn't play well with the standard silicon chips used in almost all our electronics today.

The scientists in this paper are trying to build a better, cheaper alternative using a new material called GeSn (Germanium-Tin). Think of GeSn as a "super-Ge" (Germanium) that has been upgraded with a little bit of Tin. Adding Tin is like adding a special spice to a recipe; it changes the material's properties so it can detect longer, invisible wavelengths of light that standard Germanium misses.

The Big Challenge: The "Thick Cake" Problem

To catch these invisible light waves, the camera needs a thick "sponge" (called an absorber) to soak up the light. The thicker the sponge, the more light it catches, and the better the camera works.

However, growing a thick layer of GeSn is like trying to build a skyscraper on a shaky foundation. As the layer gets thicker, it starts to crack and develop defects (like tiny cracks in a wall). These cracks act like holes in a bucket, letting electricity leak out (called dark current), which creates "noise" and ruins the picture.

Most previous attempts used very thin sponges (layers) to avoid these cracks, but that meant they couldn't catch enough light to see the farthest distances. The goal of this paper was to figure out how to build a thick, high-quality sponge without it falling apart.

The Experiment: Two Different Blueprints

The team built two different types of cameras to see which design worked best. Imagine a sandwich:

1. The "Buried Junction" Design (P-i-N):

   • The Setup: They put the "active" part of the sandwich (the junction where the magic happens) deep at the bottom, far away from the top surface.
   • The Analogy: Imagine hiding a sensitive microphone in a soundproof basement. Even if the roof (the top surface) is leaky and noisy, the microphone stays quiet because it's far away from the noise.
   • The Result: This design was excellent at keeping the "noise" (dark current) low. The camera was very quiet and clear. However, because the light had to travel all the way down to the bottom to be caught, some of it got lost or recombined on the way. It was like trying to catch a ball thrown from the roof while you are standing in the basement; you might miss some.

2. The "Surface Junction" Design (N-i-P):

   • The Setup: They moved the active part right up to the top surface.
   • The Analogy: Now the microphone is right on the roof. It catches the sound (light) immediately, so it's very efficient.
   • The Result: This design caught more light (higher responsivity), making the camera brighter. But, because it was right on the "leaky roof," it also picked up a lot of noise (higher dark current). It was like a loud, clear microphone that also picked up every car horn and bird chirp nearby.

The Breakthroughs

By testing these designs with different amounts of "Tin spice" (from 2% to 8%), the team discovered some amazing things:

• Seeing Further: They successfully built cameras that could see light up to 2.5 micrometers long. This is a huge leap, pushing the boundaries of what these materials can do.
• The Sweet Spot: The 5% Tin version with the "Buried Junction" design was the winner. It managed to be both quiet (low noise) and bright (high sensitivity). It was the perfect balance.
• The Trade-off: As they added more Tin (8%), the material got better at catching long wavelengths, but it also got "crackier" (more defects), leading to more noise. This taught them that to go even further, they need better ways to grow the material without cracks.

What's Next? The Recipe for Success

The paper concludes with a "recipe" for making the perfect GeSn camera in the future:

1. Protect the Roof: If you put the sensor near the surface, you need a better "roof" (a thicker, transparent contact layer) to stop the noise from leaking in.
2. Widen the Net: The "net" (junction) that catches the light needs to be wider. Right now, it's too narrow because the material has too many natural impurities. They need to "compensate" for this to make the net bigger.
3. Better Foundation: Most importantly, they need to grow the material more carefully. Just like a skyscraper needs a perfect foundation, these cameras need a thicker, cleaner "buffer layer" underneath to stop the cracks from forming in the first place.

In a Nutshell

This paper is a major step forward in making infrared cameras that can be built cheaply on standard computer chips. They proved that by carefully designing where the "sensor" sits inside the material, they can create devices that are quiet, bright, and capable of seeing through the fog. It's a roadmap for turning a difficult, experimental material into the next generation of "super-vision" technology for our phones, cars, and satellites.

Technical Summary

1. Problem Statement

While Germanium-Tin (GeSn) alloys offer a promising pathway for monolithic integration of Short-Wave Infrared (SWIR, 0.9–1.7 µm) and extended SWIR (e-SWIR, up to 2.5 µm) detectors onto Silicon CMOS platforms, significant performance gaps remain compared to established Indium Gallium Arsenide (InGaAs) technology.

• The Core Challenge: High-quality GeSn growth is hindered by lattice mismatch with Germanium (Ge) buffers, leading to defects (threading dislocations) as layer thickness increases.
• Current Limitation: Most existing GeSn photodiodes utilize thin absorbers (100–500 nm) to minimize dark current. However, thin absorbers limit responsivity (due to incomplete light absorption) and restrict the detection cutoff wavelength (due to residual compressive strain).
• The Gap: There is a lack of systematic understanding regarding the performance limits of thick absorbers (1–5 µm) required for high-sensitivity, long-wavelength detection, specifically concerning the trade-offs between junction position, defect density, and carrier transport mechanisms.

2. Methodology

The authors conducted a systematic empirical study of GeSn photodiodes with thick absorbers (950 nm to 2630 nm) containing 2% to 8% Tin (Sn).

• Material Growth: GeSn layers were grown on relaxed Ge buffers (1200 nm) using Reduced Pressure Chemical Vapor Deposition (RPCVD).
• Device Architectures: Two distinct doping profiles were compared to analyze the impact of junction position:
  • P-i-N: A p+ top contact, intrinsic (undoped/p-type background) absorber, and n+ bottom contact. This positions the junction deep within the absorber, far from the top surface.
  • N-i-P: An n+ top contact, intrinsic absorber, and p+ bottom contact. This positions the junction near the top surface.
• Characterization: Devices were tested from 77 K to 300 K using I-V, C-V, and spectral responsivity measurements. Key metrics included dark current density ($J_d$), responsivity ($R$), cutoff wavelength ($\lambda_c$), and specific detectivity ($D^*$).
• Analysis: The study utilized Arrhenius plots to determine activation energies (identifying Shockley-Read-Hall vs. Trap-Assisted Tunneling mechanisms) and C-V measurements to extract junction widths and background carrier concentrations.

3. Key Contributions

• Systematic Thick-Absorber Study: This is one of the first comprehensive studies comparing P-i-N and N-i-P structures with absorber thicknesses exceeding 1 µm across a wide Sn content range (2–8%).
• Junction Position Analysis: The paper definitively isolates the impact of junction depth on device physics, demonstrating how proximity to surface defects vs. bulk defects dictates dark current and carrier collection efficiency.
• Optimization Pathway: The authors propose specific strategies to overcome current bottlenecks, including compensated doping and advanced buffer layer engineering, to enable commercial-grade GeSn detectors.

4. Key Results

A. Performance Metrics

• Responsivity: High responsivity was achieved, reaching 0.59 A/W at 1.55 µm and 0.43 A/W at 2 µm.
• Dark Current: Low dark current densities were recorded for lower Sn content:
  • 2% Sn (P-i-N): $1.6 \times 10^{-2}$ A/cm².
  • 5% Sn (P-i-N): $2.4 \times 10^{-2}$ A/cm².
  • Note: 8% Sn devices showed higher dark current ($4.7 \times 10^{-1}$ A/cm²) due to increased defect density.
• Cutoff Wavelength: Extended detection ranges were achieved:
  • 5% Sn: Up to 2.08 µm at 300 K.
  • 8% Sn: Up to 2.49 µm at 300 K.
• Detectivity ($D^*$): At 300 K, $D^*$ values were comparable to or better than commercial InAs and PbSe detectors, particularly in photovoltaic (0 V bias) mode.

B. Physics Insights (P-i-N vs. N-i-P)

• P-i-N Structure (Buried Junction):
  • Advantage: The junction is buried deep in high-quality material, distancing it from surface defects. This resulted in lower dark current for 2% and 5% Sn samples.
  • Mechanism: Photocarrier transport is dominated by diffusion.
  • Limitation: Responsivity at 1.55 µm was slightly lower than N-i-P because carriers generated near the surface must diffuse to the junction, risking recombination.
• N-i-P Structure (Surface Junction):
  • Advantage: The junction is near the surface, enabling drift-dominated transport for carriers generated at the top. This yielded a consistent 0.08–0.09 A/W increase in 1.55 µm responsivity compared to P-i-N.
  • Disadvantage: Proximity to the unpassivated surface amplified leakage currents, leading to higher dark current densities (especially at 2% and 5% Sn).
• High Sn Content (8%):
  • Both structures suffered from high dark current, suggesting that bulk defects (threading dislocations) became the dominant noise source, overshadowing surface effects.
  • Activation energy analysis revealed a transition from Shockley-Read-Hall generation-recombination (SRH GR) to Trap-Assisted Tunneling (TAT) as Sn content increased.

C. Material Quality Constraints

• Background Doping: Unintentional p-type background carrier concentrations ($10^{15}–10^{16}$ cm$^{-3}$) significantly reduced the depletion width (to 250–600 nm), which is much smaller than the absorber thickness.
• Diffusion Length: In P-i-N devices, the stable responsivity suggests minority carrier diffusion lengths of at least 1–2 µm in 2% and 5% Sn samples, indicating high crystalline quality in the upper absorber layers.

5. Significance and Future Optimization Strategies

This work establishes that GeSn photodiodes can achieve performance metrics competitive with InGaAs for e-SWIR applications, provided specific optimization strategies are implemented:

1. Top Contact Optimization: For N-i-P structures, the top contact layer must be thickened (200+ nm) and made of transparent materials (Si, Ge, SiGeSn) to act as a buffer against surface defects while maintaining drift collection.
2. Compensated Doping: To counteract high background doping that shrinks the junction width, compensated n-doping in the absorber is proposed to extend the depletion region, ensuring efficient collection of carriers generated deep in thick absorbers.
3. Growth Quality Improvement:
   • Thicker Graded Buffers: Adopting strategies from SiGe/InGaAs growth (grading rates of 0.2–0.4%/µm with total buffer thickness >5 µm) to manage lattice mismatch.
   • Advanced Techniques: Exploring Aspect Ratio Trapping (ART) to geometrically confine threading dislocations.

Conclusion: The study demonstrates that GeSn is a viable candidate for CMOS-integrated e-SWIR detectors. The primary path forward involves balancing junction depth (to manage surface vs. bulk defects) and improving material quality to support thick absorbers required for high responsivity and long-wavelength detection.