Mapping Bismuth Distribution in GaAsBi/GaAs Heterostructures Using Aberration-Corrected HAADF‑STEM

Background

GaAsBi semiconductors have attracted attention for temperature‑stable, mid‑infrared devices [1]. Incorporating small amounts of Bi into the GaAs lattice reduces the bandgap dramatically, a key optoelectronic effect [2, 3]. However, the weak Ga–Bi bond, large miscibility gap, and lattice mismatch make Bi incorporation challenging, requiring non‑equilibrium growth conditions to achieve efficient incorporation. Even when successful, Bi often segregates into clusters or droplets, creating non‑radiative recombination centers that degrade device performance. The tendency for phase separation, surface droplets, atomic ordering, and nanodroplets has been documented by X‑ray absorption spectroscopy (XAS), atomic force microscopy, and XRD [11–13]. Electron spin resonance (ESR) studies show that roughly 10 % of incorporated Bi occupies Ga sites [14]. Consequently, a deep understanding of Bi incorporation and defect formation is essential for device optimization. Advanced characterization tools, notably aberration‑corrected HAADF‑STEM, enable sub‑angstrom imaging of dilute alloys, where intensity scales with the average atomic number (Z) [15–18]. qHAADF analysis further correlates HAADF intensity with atomic column composition in III–V alloys [12, 16, 18], while high‑resolution images reveal crystalline quality [19–22].

This work employs aberration‑corrected HAADF‑STEM and complementary EDX to probe Bi distribution in GaAs/GaAs_1-xBi_x/GaAs heterostructures grown by MBE at ~340 °C. We also examine the impact of Bi segregation on the nano‑ and micro‑scale structure, using qHAADF, FFT analysis, and XRD.

Methods

Two heterostructures were fabricated by solid‑source MBE on 2″ n⁺ GaAs:Si (001) wafers. Sample S1 was grown with a Bi cell temperature of 460 °C, while S2 used 505 °C. The VG V80 MBE system operated at a background pressure of ~5 × 10⁻¹⁰ mbar. Each structure comprised a 130 nm GaAs buffer, a 130 nm GaAs_1-xBi_x layer, a 5 nm GaAs spacer, and a 130 nm GaAs cap. Growth temperatures were ~580 °C for GaAs layers and ~340 °C for the GaAsBi layer under near‑stoichiometric As₄ flux. The GaAsBi layer was annealed in situ at ~580 °C for ~20 min during cap growth. No surface metallic droplets were observed. Substrate temperatures were calibrated by optical thermometry and surface reconstruction markers.

Cross‑sectional specimens were prepared by mechanical grinding and Ar⁺ ion milling using a precision ion polishing system (PIPS). Final milling employed 1.5 kV to improve surface quality, and samples were plasma‑cleaned to minimize electron‑beam deposition.

HAADF‑STEM, zero‑loss EELS, and EDX line scans were performed at 200 kV on a Titan³ Themis microscope equipped with a Cs corrector, monochromator, and Super‑X EDX detector. Secondary electron images were acquired on a FEI NOVA NANOSEM 450 at 2 kV. Bi quantification used the Bi‑M line at 2.42 keV with Bruker Espirit software. Specimen thickness was derived from spatially resolved zero‑loss EELS using Digital Micrograph (GATAN™). qHAADF analysis was conducted with the Digital Micrograph plugin, mapping integrated intensity around atomic columns. ω‑2θ (0 0 4) XRD spectra were recorded on a Bruker D8 Discover diffractometer (Cu‑Kα₁) and simulated with Bede Rads Mercury.

Results and Discussion

Figure 1 displays low‑magnification HAADF‑STEM images of the [110] zone axis for samples S1 (a) and S2 (b), along with thickness‑gradient‑corrected intensity profiles (green rectangles). No threading dislocations or stacking faults were detected. In HAADF, intensity is proportional to the average atomic number; thus, brighter regions indicate higher Bi content (Z_Bi = 83, Z_Ga = 31, Z_As = 33). Sample S1, with lower Bi, shows no pronounced contrast variations, whereas sample S2 exhibits regularly distributed brighter spots along the GaAsBi layer and interfaces, interpreted as Bi‑rich clusters. These clusters are embedded within the layer, as confirmed by low‑voltage SEM topography matching the STEM images.

Intensity profiles (Fig. 1c) reveal similar layer thicknesses (~140 nm) for both samples. The GaAsBi/GaAs interface shows a gradual intensity drop over ~10 nm, implying Bi incorporation into the GaAs cap even without a Bi flux. In the low‑Bi sample (S1), the interface is sharper than in S2, suggesting a higher Bi incorporation coefficient at lower Bi flux, consistent with kinetic‑limited growth [26].

STEM‑EDX maps (Fig. 2) confirm the HAADF observations. The average Bi atomic fraction was 1.2 ± 0.4 % in S1 and 5.3 ± 0.4 % in S2, confirming non‑uniform Bi distribution. Bi clusters arise from steric hindrance; larger Bi atoms impede Ga incorporation, creating vacancies and leading to Bi segregation. EDX elemental maps of a cluster (Fig. 2b) show simultaneous decreases in Ga and As signals where Bi is high, indicating Bi occupies both group III and group V sites.

High‑magnification HAADF‑STEM images in the [110] projection were analyzed using qHAADF. Integrated intensities of group V columns were compared to those of bulk GaAs to obtain the ratio R = I_As‑Bi/I_As. Figure 3a shows the interface for S1; the R map (Fig. 3b) displays values ranging from 1.00 to 1.27, indicating a 27 % Bi enrichment in the highest‑intensity columns. The coefficient of variation (Cv) was 1.3 % in the GaAs substrate, 2.6 % in the GaAsBi layer, and 2.6 % in the cap, reflecting Bi‑related compositional fluctuations.

For sample S2, HAADF‑STEM reveals ≈ 12 nm equiaxial clusters uniformly distributed in the GaAsBi layer and at interfaces (Fig. 4a). Low‑pass filtering (Fig. 4b) highlights {111} and (001) facets. The clusters consist of a Bi‑rich core (≈ 30 % Bi) surrounded by a lower‑Bi shell, similar to observations in GaAs_1-xBi_x/GaAs_1-yBi_y multilayers [10]. qHAADF mapping (Fig. 4c) shows a pronounced Bi gradient peaking at the cluster centre. FFT analysis of selected regions (Fig. 4d) reveals additional {102} spots attributable to rhombohedral Bi (rh‑Bi) nucleating within the zinc‑blende matrix, indicating phase coexistence and local strain relaxation.

XRD ω‑2θ scans (Fig. 5) show sharp GaAs substrate peaks and broader GaAsBi peaks at negative angles, indicating strain. In S2, a shoulder on the GaAs peak signifies tensile strain in the cap, confirming partial relaxation of the GaAsBi layer. Simulations based on EDX and HAADF data (5.8 % Bi) fit the XRD curve reasonably for S1, but for S2 the presence of Bi‑rich clusters necessitates a higher average Bi content (~6 %) to match the observed peak positions, corroborating the HAADF‑STEM findings.

Conclusions

Aberration‑corrected HAADF‑STEM, complemented by qHAADF, FFT, and XRD, provides a comprehensive view of Bi distribution in GaAs/GaAs_1-xBi_x/GaAs heterostructures. Low‑Bi samples exhibit sharp interfaces with minor Bi segregation, whereas high‑Bi samples form ≈ 10–12 nm equiaxial Bi‑rich clusters throughout the matrix. These clusters contain both rhombohedral Bi and zinc‑blende phases, inducing local lattice relaxation and partial strain relief. The observed Bi segregation and phase coexistence are key parameters for tailoring the optical and electronic properties of GaAsBi‑based devices.

Abbreviations

Ac-HAADF-STEM

Aberration‑corrected high‑angle annular dark‑field scanning transmission electron microscopy

EDX

Energy‑dispersive X‑ray

EELS

Electron energy loss spectrum

ESR

Conventional electron spin resonance

FEG

Cold field emission gun

FFT

Fast Fourier transform

HRTEM

High‑resolution transmission electron microscopy

IMEYMAT

Instituto Universitario de Investigación en Microscopía Electrónica y Materiales

MBE

Molecular beam epitaxy

qHAADF

Quantitative HAADF image analysis algorithm

SEM

Scanning electron microscope

XAS

X‑ray absorption spectroscopy

XRD

X‑ray diffraction