Atomic short-range order control of GeSn as a new degree of freedom for band engineering

Chemical short-range order (SRO) refers to preference or avoidance between neighboring atomic species, which significantly impacts the properties of advanced alloys. However, quantifying and further controlling SRO remains a major challenge, especially for semiconductor alloys. Inspired by theoretically predicted impact of SRO on the band structure of direct-bandgap GeSn for infrared photonics, we quantify and compare SRO in GeSn grown by molecular beam epitaxy (MBE) vs. chemical vapor deposition (CVD) using atom probe tomography. Remarkably, MBE-grown GeSn exhibits a stronger preference for Sn-Sn 1st nearest neighbors and an even smaller bandgap than CVD-grown samples with 2 at.% higher Sn composition. First-principles modeling confirms that the bandgap reduction originates from differences in SRO and further indicates that these SRO variations arise from different surface terminations and growth temperatures between MBE and CVD. These findings suggest that controlling SRO during GeSn growth offers a new degree of freedom for band engineering to achieve lattice-matched, high-quality Si-based electronic/photonic devices.

💡 Research Summary

This paper presents a comprehensive experimental and theoretical investigation of chemical short‑range order (SRO) in GeSn alloys and demonstrates its powerful role as an additional degree of freedom for band‑gap engineering. The authors focus on two widely used growth techniques—molecular beam epitaxy (MBE) and chemical vapor deposition (CVD)—and compare the SRO that each method induces.

Using atom probe tomography (APT) they acquire three‑dimensional atomic coordinates for four samples: two MBE‑grown (a 20 at.% Sn thin film and a 7 at.% Sn multiple‑quantum‑well stack) and two CVD‑grown (a 14 at.% Sn thin film and a 7 at.% Sn MQW). Because APT introduces a modest positional jitter (~1 Å), the authors apply a physics‑informed Poisson‑KNN statistical reconstruction to retrieve the true K‑nearest‑neighbor (KNN) shells. They define an SRO parameter αₖᴬ⁻ᴮ = Pₖᴬ⁻ᴮ / xᴮ, where Pₖᴬ⁻ᴮ is the probability that a B atom appears in the K‑th neighbor shell of an A atom and xᴮ is the overall B concentration. For a random alloy α = 1; α > 1 indicates a preference for A‑B pairing.

Analysis of 5 nm × 5 nm × 5 nm nanocubes (≈2000 atoms each) yields the following key observations:

1. The Sn‑Sn 1st‑nearest‑neighbor (1NN) SRO parameter α₁ᴺᴺˢⁿ is ≈1.14 ± 0.03 for both MBE samples, whereas it is ≈1.01 ± 0.02 for the CVD samples. Histograms show that ~90 % of MBE nanocubes have α > 1, compared with only ~30 % for CVD. This demonstrates a clear, growth‑method‑dependent preference for Sn‑Sn adjacency in MBE‑grown GeSn.

2. The difference in α is independent of overall Sn concentration (the MBE and CVD MQWs both contain 7 at.% Sn) and persists across different reactors and laboratories, indicating that the underlying cause is intrinsic to the growth physics rather than experimental artifacts.

3. Photoluminescence (PL) measurements at 10 K reveal that the MBE MQW (7 at.% Sn) exhibits a PL peak ≈85 meV lower in energy than the CVD MQW with the same composition. This band‑gap reduction correlates directly with the higher Sn‑Sn 1NN SRO in the MBE sample.

First‑principles density‑functional theory (DFT) calculations support the experimental trend. Simulations of GeSn supercells with varying Sn‑Sn 1NN frequencies show that increasing Sn‑Sn adjacency expands the local lattice, reduces the direct Γ‑Γ band gap, and lowers the optical transition energy. The calculations also indicate that the observed SRO differences can be traced to two growth‑related factors: (i) growth temperature (MBE: 120–150 °C vs. CVD: 250–350 °C) and (ii) surface termination (Ge‑terminated surface in MBE versus Sn‑terminated surface in CVD). Lower temperature and Ge termination promote surface diffusion pathways that allow Sn atoms to cluster, whereas higher temperature and Sn termination favor Sn‑Sn repulsion.

The authors thus establish SRO as a tunable, independent variable for band‑gap engineering in GeSn, complementary to composition and strain. By selecting appropriate growth conditions—temperature, precursor chemistry, and surface termination—one can deliberately tailor the Sn‑Sn 1NN probability and achieve desired band‑gap values without altering overall Sn content or lattice mismatch.

Beyond GeSn, the Poisson‑KNN approach to APT data provides a robust framework for quantifying SRO in any multicomponent semiconductor where atomic‑scale ordering influences electronic or optical properties. The work opens a pathway toward “order‑controlled” design of Si‑compatible infrared photonic devices such as lasers, detectors, and modulators, potentially enabling lattice‑matched, high‑quality devices with precisely engineered emission wavelengths.

In summary, the paper (1) experimentally quantifies SRO in GeSn grown by MBE and CVD using advanced APT analysis, (2) demonstrates that stronger Sn‑Sn 1NN ordering in MBE leads to a measurable band‑gap reduction of ~0.08 eV, (3) validates the observation with DFT modeling that links SRO to lattice distortion and electronic structure, and (4) proposes SRO control as a new, practical degree of freedom for semiconductor band‑gap engineering, with broad implications for future Si‑integrated photonics.