Detection of Bi_Ga hetero-antisites at Ga(As,Bi)/(Al,Ga)As interfaces

The last few years have seen significant advances in the research and development of the so-called dilute-bismide alloys, which are based on III–V semiconductor compounds after the isoelectronic substitution of group-V atoms (As, P, or Sb) by bismuth, the largest group-V element. Their interest lies in the peculiar electronic properties exhibited by these compounds upon the incorporation of a few percent of Bi, such as a large reduction in the bandgap, large enhancement of the spin–orbit splitting, and the reported almost temperature insensitive bandgap dependence,^1,2 which prompt potential applications in high efficiency solar cells, infrared optoelectronic devices, and spintronics. Ga(As,Bi) is the most studied compound of the dilute III–V-bismide semiconductors family. Additionally, Ga(As,Bi) can be considered as an exemplary case of a highly mismatched alloy due to the atomic-mismatch in size and/or electronegativity between the host group-V atom (As) and Bi. This atomic-mismatch, which provides Ga(As,Bi) with its peculiar and attractive properties, imposes at the same time serious challenges in the alloy synthesis due to a large miscibility gap and the extremely low solubility of Bi into GaAs.³ In order to successfully incorporate Bi into GaAs and to achieve homogeneous layers, extreme growth conditions and innovative strategies are required, for instance, using molecular beam epitaxy (MBE). In this context, low substrate temperatures (T_s) and careful control of the As/Ga flux ratio close to stoichiometry are crucial parameters in the process.^1,2

While it seemed that the microstructure and compositional properties of Ga(As,Bi) had been sufficiently studied and the main characteristic structural features, such as clustering, droplet formation, phase separation, and atomic ordering^4–11 identified, we report here on a groundbreaking yet expected finding: the detection of Bi_Ga hetero-antisites. More specifically, we report on the detection of Bi_Ga at the interfaces of Ga(As,Bi)/(Al,Ga)As quantum well (QW) structures.

The use of (Al,Ga)As barriers was earlier proposed in the context of Ga(As,Bi)-based laser structures.^12,13 This strategy, however, has regained interest in recent years due to the improved performance of the QWs as demonstrated by Butkutė et al. on the growth of Ga(As,Bi) QW with AlAs barriers,¹⁴ by Pan et al. with Ga(As,Bi)/(Al,Ga)As-based structures emitting beyond 1.2 μm with 5.8% Bi,¹⁵ or by the recent achievements by Pūkienė et al. and Jokubauskaitė et al. reporting enhanced photoluminescence emission from Ga(As,Bi) QWs with parabolic graded (Al,Ga)As barriers.^16,17 It is argued that the presence of Al might suppress the well-known Bi surface segregation, although the exact role of Al is unclear. As a matter of fact, Ga(As,Bi)/(Al,Ga)As heterostructures are comparatively under-studied.

During routine investigation of the interface properties in Ga(As,Bi)/(Al,Ga)As QWs using (scanning) transmission electron microscopy (S)TEM techniques, we were very struck by the radically different appearance of the layers depending on the specific imaging mode used to inspect the layers. In particular, whereas STEM micrographs showed at first glance the expected sequence of layers, diffraction-based chemically sensitive dark-field TEM (DFTEM) images taken with g = 002 revealed the striking presence of dark-lines delimiting Ga(As,Bi)/(Al,Ga)As interfaces. As will be discussed here, the dark contrast at the Ga(As,Bi)/(Al,Ga)As interfaces in g₀₀₂ DFTEM micrographs cannot be simply explained. The aim of the present work is thus very straightforward: attempt to identify the origin of the unknown dark contrast at the Ga(As,Bi)/(Al,Ga)As interfaces when these are visualized using g₀₀₂ DFTEM. In the search for an explanation for this anomaly, we recognize that Bi incorporation at the group-III element position, i.e., the presence of Bi_Ga hetero-antisites has a remarkable impact decreasing the diffracted intensity I_002, and 1% Bi_Ga would explain the observed contrast. Furthermore, revised STEM measurements and analytical STEM procedures reveal Ga depletion and Bi accumulation at the interface, consistent with Bi_Ga at this location. The realization of a growth interruption together with a Bi surface pre-treatment may have favored the formation of Bi_Ga at this location, which demonstrates that bismuth incorporation into the lattice is far from being well understood. On the other hand, this work attests that diffraction-based g₀₀₂ DFTEM imaging is positioned as a very powerful technique for the detection of point defects in materials with the zinc-blende crystal structure.

All investigated samples were grown by solid source molecular beam epitaxy (MBE) on GaAs(001). A schematic representation of the sample structure is displayed in Fig. 1. Since the samples were designed for transport measurements,¹⁸ the structures include n- and p-type modulation doped (Al,Ga)As barriers and a doped GaAs cap layer (cf. Fig. 1). In the investigation, we focus on two samples, Samples A and B, which are identical in their design and growth conditions except for the different character of the doped regions, namely, n-type in Sample A and p-type in Sample B. respectively. As shown in this work, both samples exhibit the very same features in terms of microstructure and interface properties, being basically indistinguishable, hence we consider them indistinctly in our work. The nominal 7 nm thick Bi-containing Ga(As,Bi) QWs were grown at 370 °C, whereas T_s was raised to 580 °C for the growth of the barriers. During MBE growth, there was a growth interruption (GI) before and after the QW to adjust T_s and the atomic As/Ga flux ratio. During GI at the bottom Ga(As,Bi)-on-(Al,Ga)As interface, only As was open during the ramp down from 580 °C to 370 °C. After this ramp, the Bi shutter was opened for about 79 s, at this point, only As and Bi are open. The purpose of the Bi treatment was to try to make the QW structure more uniform by increasing the Bi surface coverage before the QW growth, which started when the Ga shutter was opened. The atomic As/Ga flux ratio was intentionally adjusted during the GI to very slightly below one for Ga(As,Bi) QW growth to ensure Bi incorporation. Finally, there was a second GI after QW growth to ramp T_s up to 580 °C with only As open, after which the final (Al,Ga)As and GaAs capping layers were grown. The Bi concentration in the QW was determined as 4% using x-ray diffraction (XRD).¹⁸ Reference samples (Samples C and D) composed of GaAs/(Al,Ga)As layers grown at 580 °C, i.e., without Bi in the QW and with no GI, were also investigated for comparison.

The samples were investigated on a JEOL JEM-3010 microscope operating at 300 kV equipped with a GATAN slow-scan CCD camera for g₀₀₂ DFTEM and on a JEOL 2100F microscope operated at 200 kV for STEM, equipped with a bright-field (BF) detector, high-angle annular dark-field (HAADF) detector, and a JEOL JED-2300 spectrometer for energy-dispersive x-ray spectrometry (EDS) studies. Investigations were performed on both conventionally prepared TEM specimens and on lamellas prepared using a focus ion beam (FIB) system JEOL JIB-4501. In conventional preparation, the cross-sectional TEM foils were prepared in the [110] and $[ 1 ¯10]$ projections using mechanical thinning followed by Ar-ion milling. In order to minimize sputtering damage, the Ar-ion beam energy was reduced to ∼1.5 keV for final milling. A total of eight TEM samples were analyzed, probing different areas. These include measurements on six TEM specimens (by conventional preparation) and on two FIB lamellas. All samples exhibit the very same characteristics. The electron beam direction used for observation was the $⟨ 110 ⟩$ zone axis for all samples.

Figure 2 shows representative HAADF-STEM overview micrographs with so-called atomic number Z–contrast of the samples. The Ga(As,Bi) QWs are clearly identified in the images due to their brightest contrast as a result of the higher average atomic number compared to the adjacent (Al,Ga)As barriers and GaAs layers (Z_Bi = 83, Z_As= 33, Z_Ga = 31, and Z_Al= 13). The layers grow pseudomorphically on the GaAs substrate and no extended defects or dislocations are detected. The sequence of layers as revealed by HAADF-STEM is in agreement with the nominal design, as is the thickness of the QW (∼7.3 nm) and composition (∼4% Bi). Bi content is roughly estimated from the analysis of I_HAADF intensity assuming I_HAADF∝ Z^1.7 following the methodology described in Refs. 19 and 20, which includes noise-correction from the annular dark-field detector and background subtraction to compensate for thickness variations in the TEM specimen.^19,20

We found that the appearance of the layers dramatically changed when these are imaged using two-beam diffraction-based chemically sensitive g₀₀₂ DFTEM conditions, as shown in Fig. 3. As opposed to HAADF imaging, the contrast in the g₀₀₂ DFTEM “structure-factor imaging mode” mainly arises from the difference in the atomic scattering factors between the group-III and group-V elements.²¹ Notably, g₀₀₂ DFTEM micrographs reveal the striking presence of two “dark-lines” at both Ga(As,Bi)-on-(Al,Ga)As and (Al,Ga)As-on-Ga(As,Bi) interfaces, as evidenced in the representative images in Fig. 3. The dark-lines are precisely positioned at the location of GI, delimiting the interfaces. They are relatively broad, about ∼2 nm for the bottom Ga(As,Bi)-on-(Al,Ga)As interface and about ∼1 nm for the upper (Al,Ga)As-on-Ga(As,Bi) interface. Both samples exhibit exactly the same features. We observe slight differences in the contrast and homogeneity of the bottom and upper dark-lines. The bottom dark-line at the Ga(As,Bi)-on-(Al,Ga)As interface is well defined and has a regular thickness with slight lateral corrugations on a scale of few tens of nm. Fluctuations in the QW thickness and in the thickness of the upper (Al,Ga)As-on-Ga(As,Bi) dark-line are also detected, as can be seen in Figs. 4(a) and 4(b). In particular, the upper dark-line at the (Al,Ga)As-on-Ga(As,Bi) interface is more irregular than the one at the Ga(As,Bi)-on-(Al,Ga)As interface as demonstrated by the fact that, locally, there are regions where the upper dark-line is barely visible, cf. Fig. 3 with Figs. 4(a) and 4(b). Conversely, the dark-line at the Ga(As,Bi)-on-(Al,Ga)As interface is ubiquitous and is characterized by its regular thickness of about 2.1 nm. Due to the solid appearance of the dark-line at the Ga(As,Bi)-on-(Al,Ga)As interface, in this work, we will mainly focus on this interface. Chemically sensitive g₀₀₂ DFTEM micrographs of reference samples, i.e., without Bi in the QW, show the expected contrast without any relevant feature, as displayed in Figs. 4(c) and 4(d).

In III–V alloys, the diffracted intensity I₀₀₂ for the 002 reflection is proportional to the square of the structure factor F₀₀₂, which in turn depends on the difference in the atomic scattering factors of the alloy components (f_III and f_V for the group-III and group-V elements, respectively), $I 002∝ F 002 2∝| f III− f V | 2$⁠. Hence, when two-beam 002 imaging conditions are properly set up (the specimen is tilted ∼10° from the $⟨ 110 ⟩$ zone axis toward the [100] pole, while keeping the interface edge-on), compositional information can be directly extracted following the procedure proposed by Bithell and Stobbs, based on analysis of the DFTEM image contrast and assuming kinematical approximation.²¹ In order to avoid measuring absolute intensities, the diffracted intensity of the layer I₀₀₂^layer is normalized to that of an adjacent reference layer of known composition, in this case, I₀₀₂^GaAs, R₀₀₂ = I₀₀₂^layer/I₀₀₂^GaAs. In the analysis, we assume the validity of Vegard’s law as well as the substitutional incorporation of Bi atoms in the group-V lattice [for Ga(As,Bi)] and of Al atoms in the group-III lattice [for Al(Ga,As)]. The calculation uses the atomic scattering factors given by Doyle and Turner.²² Analysis of the experimental R₀₀₂ at the QW location yields the same Bi content in both samples, [Bi]_g002∼ 3.9%, which is in agreement with the Bi content extracted from XRD and estimations from HAADF-STEM measurements, about 4% Bi, and verifies that Bi incorporation in the Ga(As,Bi) QW is substitutional at group-V element position as demonstrated in earlier experiments.²³ The estimated Al content in the (Al,Ga)As barriers is about [Al]_g002∼ 12%, as deduced from the analysis of g₀₀₂ DFTEM micrographs. In contrast to previous investigations of Ga(As,Bi) QWs with GaAs barriers,^11,24 we found no indications of Bi surface segregation in the Ga(As,Bi)/(Al,Ga)As QWs, a result which agrees with recent works suggesting that the presence of Al reduces Bi segregation due to Al blocking Bi atoms.^14,15

Noteworthy, the most noticeable observation regarding g₀₀₂ DFTEM investigation of the samples is the presence of the dark-lines at the interfaces, precisely at the GI positions. Experimental g₀₀₂ DFTEM contrast associated with the dark-lines ranges between R₀₀₂^exp ∼ 1.2–1.35 for the dark-line at the Ga(As,Bi)-on-(Al,Ga)As interface. Furthermore, this contrast seems not to depend on defocus, on the imaging conditions to reach two-beam g₀₀₂ DFTEM or on the microscope used in the investigation (the presence of the dark-line has been cross-checked with additional g₀₀₂ DFTEM measurements using the JEOL 2100F microscope).

The lower contrast associated with the dark-lines cannot be simply explained. EDS does not reveal the presence of contamination or unintentional incorporation of elements during GI (at this level of detection). The formation of the quaternary compound (Al,Ga)(As,Bi) at the interface due to the gradual transition in composition from Al(Ga,As) to Ga(As,Bi) is anticipated²⁵ and may cause the features. However, theoretical estimations of the g₀₀₂ diffracted intensity for (Al,Ga)(As,Bi) reveal a much brighter contrast than observed experimentally. In particular, taking into consideration the estimated compositions of ∼12% Al in the (Al,Ga)As barrier and ∼4% Bi in the Ga(As,Bi) QW, a realistic composition at the interface would be (Al_0.06Ga_0.94)(Bi_0.02As_0.98) that yields R₀₀₂∼ 1.8, which clearly deviates from the experimental value R₀₀₂^exp ∼ 1.2–1.35. In the calculation, Bi and Al are incorporated substitutionally at group-V and group-III element positions, respectively. Interestingly, as will be shown in this work, Bi incorporation at group-III element position, i.e., the presence of Bi_Ga hetero-antisites, has a remarkable impact decreasing I₀₀₂ and, hence, R₀₀₂. This is evidenced by a simple estimation of the g₀₀₂ DFTEM contrast (R₀₀₂) assuming a z-molar-fraction of Bi_Ga hetero-antisites, i.e., Bi at the group-III element position, (Al_xGa_1−x−zBi_z)(As_1−yBi_y) leading to R₀₀₂ = (1 + Ax + By + Cz)² with A = (f_Ga − f_Al)/(f_As − f_Ga), B = (f_Bi − f_As)/(f_As − f_Ga), and C = (f_Ga − f_Bi)/(f_As − f_Ga), where f_i with i = Al, Ga, As, Bi are the corresponding atomic scattering factors. Substituting values yields to R₀₀₂ = (1 + 2.85x + 8.2y − 9.2z)². Atomic static displacements are not considered in the calculations. The presence of 1% Bi_Ga, i.e., (Al_0.06Ga_0.93Bi_0.01)(Bi_0.01As_0.99) yields R₀₀₂∼ 1.35, while 1.5% Bi_Ga (Al_0.06Ga_0.925Bi_0.015)(Bi_0.01As_0.99) results in R₀₀₂∼ 1.24, which are in agreement with experimental observations. Hence, the presence of Bi_Ga quantitatively explains the striking contrast at the interfaces.

One may also consider a high concentration of point defects such as As antisites (As_Ga) and Ga vacancies (V_Ga) as is typical in low temperature (LT) grown GaAs. We note that the presence of As_Ga and V_Ga alone (without Bi_Ga) does not explain the experimental contrast. Such a structure should be $( A l x G a 1 − x − u − w V w Ga A s u)( A s 1 − y B i y)$⁠, with R₀₀₂ = (1 + Ax + By + Dw − u)² and D = f_Ga/(f_As − f_Ga), i.e., R₀₀₂ = (1 + 2.85x + 8.2y + 8.47w − u)². Taking into consideration the typical concentration of point defects in LT-GaAs is in the order of 10¹⁹–10²⁰ cm⁻³ (i.e., w, u ∈[0.0001, 0.001]),^26–30 assuming no Bi_Ga but Bi and Al substitutional incorporation at group-V and group-III element position, respectively, $( A l 0.06 G a 0.94 − u − w V w Ga A s u)( B i 0.02 A s 0.98)$⁠, yields R₀₀₂∼ 1.78, which is far from the experimentally determined value. In the present samples, we expect less As_Ga and V_Ga than in LT-GaAs since the Ga(As,Bi) QWs were grown close to stoichiometry, even slightly below the stoichiometric As/Ga flux ratio, i.e., under growth conditions that substantially reduce excess As incorporation.^31,32 Furthermore, As_Ga defect formation in Ga(As,Bi) is largely suppressed with increasing T_s as shown, for instance, by Dagnelund et al. who observed a significant reduction in As_Ga density when increasing T_s from 270 to 315 °C,³³ temperatures that are far below T_s= 370 °C used to grow the present Ga(As,Bi) QWs. In any case, assuming a high density of As_Ga and V_Ga as in LT-GaAs in addition to Bi_Ga, the full structure should be (Al_xGa_{1−x−z−u−w}Bi_zV^Ga_wAs_u)(As_1−yBi_y), the estimated contrast being R₀₀₂ = (1 + Ax + By + Cz + Dw − u)², i.e., R₀₀₂ = (1 + 2.85x + 8.2y – 9.2z + 8.47w − u)². From this expression, it is clear that the presence of Bi_Ga and As_Ga explains the experimental contrast, albeit the impact of Bi_Ga in decreasing R₀₀₂ is significantly larger than that of As_Ga. g₀₀₂ DFTEM imaging in Ga(As,Bi) is, therefore, highly sensitive to the presence of Bi_Ga. Although our results demonstrate that the Bi_Ga hetero-antisite explains the experimental contrast at the Ga(As,Bi)/(Al,Ga)As interfaces, it is realistic to contemplate the presence of As_Ga and V_Ga in addition to Bi_Ga, not only as point defects but as part of defect complexes. On the other hand, as mentioned earlier, we verified the Bi incorporation in the approximately 7 nm Ga(As,Bi) QW is substitutional at group-V element position, as demonstrated in earlier experiments.²³ 

In the light of these findings, we decided to repeat the HAADF-STEM measurements. Revised HAADF-STEM at higher magnification than those used in the images in Fig. 2 reveals the subtle presence of a Bi-enriched region below the QW, at the location of the dark-line at the Ga(As,Bi)-on-(Al,Ga)As interface in g₀₀₂ DFTEM micrographs. An example is shown in Fig. 5, which displays representative HAADF-STEM micrographs as well as the intensity profiles obtained from area-scans in the regions indicated in the figure. Careful inspection of the I_HAADF profiles reveals that there is a subtle bump right below the QW. Due to its low signal which overlaps with the main contribution arising from the QW itself, its presence is overlooked in quick scans or in lower magnification images as those used in routine STEM examination (cf. Fig. 2), Instead, the feature is invariably detected in thorough HAADF-STEM measurements at higher magnification specifically aimed at the detection of any anomaly at the location of the dark-lines. Due to the chemically sensitive character of HAADF-STEM, the bump is indicative of the local presence of a Bi-rich area at the onset of the QW and extending over ∼2 nm. Note that HAADF-STEM imaging is particularly sensitive to the presence of Bi due to its high atomic number, Z_Bi = 83. Furthermore, we observe that the location and spatial extension of the Bi-rich area exactly agrees with the precise location and dimensions of the dark-line at the Ga(As,Bi)-on-(Al,Ga)As interface as observed in g₀₀₂ DFTEM.

The presence of a Bi-rich area below the QW is further validated using EDS. Routine EDS elemental maps as those shown in Fig. 6 reproduce the expected element distribution of the nominal structure, with signatures of a decreased Al and As content at the QW position, concomitant with an increased Bi and Ga content at the QW region, as expected. Bi elemental profiles obtained from area-scans perpendicular to the QW on the acquired spectrum imaging (SI) EDS elemental maps disclose compelling information as shown in Fig. 7. On the one hand, we find that Bi EDS profiles exactly reproduce the shape of the I_HAADF profiles, which are highly sensitive to the presence of Bi. In particular, the Bi EDS profile demonstrates again the emergence of a higher Bi content bump below the QW at a location, which agrees with the position of the dark-line. Furthermore, the bump extends over ∼2.1 nm, distinctly matching the thickness of the dark-line at the Ga(As,Bi)-on-(Al,Ga)As interface, as shown in the representative profile in Fig. 7(b) extracted from the Bi map in Fig. 7(a). Outstandingly, Ga EDS profiles obtained from SI x-ray mapping provide even more significant information than the Bi profiles. Figure 7(d) displays the Ga profile obtained from analysis of the same SI map, leading to the Bi profile in Fig. 7(b). As observed in the figure, we find that in the region where Ga signal increases after growth of the first (Al,Ga)As barrier, there is an abrupt drop in the Ga signal at the onset of the Ga(As,Bi) QW. The location of the sudden Ga drop exactly coincides with the precise location of the Bi-rich band in the Bi EDS profile. More specifically, both features Bi accumulation and Ga depletion occur at the same spot of the dark-line at the Ga(As,Bi)-on-(Al,Ga)As interface in g₀₀₂ DFTEM micrographs. Notably, the feature accounting for the drop in Ga signal has a certain thickness, which strikingly matches that of the Bi-rich area (bump in Bi EDS profile and in I_HAADF profiles), as well as the thickness of the dark-line at the Ga(As,Bi)-on-(Al,Ga)As interface in g₀₀₂ DFTEM micrographs. It could be argued that the bump in the Bi profile is due to delayed Bi incorporation in the layers, as has been reported for Ga(As,Bi)/GaAs QWs grown by metal-organic vapor phase epitaxy (MOVPE).^34,35 Such a delayed Bi incorporation has also been observed during MBE growth^36,37 and is often attributed to the strong Bi surface segregation.^11,36 There are several strategies to reduce Bi delayed incorporation, for instance, with the deposition of a Bi pre-layer before opening the Ga shutter for Ga(As,Bi) growth,^37–39 which is the approach used in the present samples. Furthermore, we find no indications of Bi surface segregation. These facts, together with the remarkable match in width, dimensions, and location of the Bi bump, of the Ga-depleted region, and of the dark-line in g₀₀₂ DFTEM micrographs, suggest that the features in the HAADF profiles [Figs. 5(b) and 5(d)] and in the Bi EDS profile in Fig. 7(b) have a different origin than the features observed in HAADF-STEM profiles of MOVPE-grown Ga(As,Bi)/GaAs superlattices.^34,35 The representative maps in Figs. 7(a) and 7(c) and the corresponding profiles are obtained from Sample B using Bi M-line and Ga K-line, respectively. Same results are obtained mapping Ga L-line. Furthermore, similar features are observed from single EDS line measurements. Whereas EDS maps displayed in Figs. 6 and 7 are obtained from measurements on two TEM lamellas from Sample B, the very same results are obtained from EDS on conventionally prepared specimens of both Sample A and Sample B. Moreover, due to the extreme sensitivity of EDS mapping to external disturbances and sample drift, EDS measurements have been repeated several times, on lamellas as well as on conventionally prepared samples, with identical results in all cases: detection of a Bi-rich area and of a drop in the Ga signal below the QW, coinciding with the precise position of the dark-line at the Ga(As,Bi)-on-(Al,Ga)As interface in g₀₀₂ DFTEM micrographs. Likewise, there are faint evidence of a Bi bump and slight drop in Ga signal also at the upper (Al,Ga)As-on-Ga(As,Bi) interface, although these features are not as visible as those in the bottom interface. The Ga-depleted Ga(As,Bi)-on-(Al,Ga)As interface can be explained by the absence of Ga supply during the GI and subsequent Bi treatment. Hence, analytical STEM expose a Ga deficiency and a Bi enrichment at the Ga(As,Bi)-on-(Al,Ga)As interface, which is compatible with the presence of Bi_Ga hetero-antisites at this location.

The present investigation of Ga(As,Bi)/(Al,Ga)As QW structures relies on the analysis of micrographs obtained using different (S)TEM techniques, the corresponding contrast arising from different physical mechanisms, ranging from incoherent (quasi) elastic Rutherford scattering in HAADF-STEM to diffraction-contrast based two-beam g₀₀₂ DFTEM.⁴⁰ In addition to the information provided by each method, each technique probes the interface on a different length scale. All the results presented here align with the presence of Bi_Ga antisites at the Ga(As,Bi)/(Al,Ga)As interface. Notably, we found no indications on the presence of Bi_Ga inside the Ga(As,Bi) QW.

The sensibility and potential of g₀₀₂ DFTEM imaging for point defects detection in compound semiconductors with the zinc-blende structure were already demonstrated by Glas et al. in their seminal article on the determination of the local concentration of Mn-interstitials and antisite defects in (Ga,Mn)As layers.⁴¹ Additionally, we had previously noticed using g₀₀₂ DFTEM that there is a change in contrast between LT-GaAs and GaAs-buffer in Ga(As,Bi) structures and between GaSb and LT-GaSb in Ga(Sb,Bi)/GaSb QW structures, respectively.⁴² We tentatively attributed this condition to the sensitivity of g₀₀₂ DFTEM to detect local variations in point defects density, as had been reported by Glas et al.⁴¹ Likewise, the results presented here highlight the relevance of g₀₀₂ DFTEM imaging in unveiling the presence of point defects, in this case, the widely anticipated presence of Bi_Ga in dilute bismides and, more specifically, in Ga(As,Bi) compounds.^43,44 Hence, our work is among the scarce experimental evidence of the presence of Bi_Ga in dilute bismide compounds, more specifically at the interfaces of Ga(As,Bi)/(Al,Ga)As QW structures.

Although theoretical works have widely anticipated the presence of Bi_Ga in dilute bismide compounds as single point defect or as Bi_Ga-related complex defects,^43,44 the fact is that the experimental detection of Bi_Ga has mostly remained elusive. A bismuth hetero-antisite defect was previously identified by electron-spin resonance and magnetic-circular-dichroism absorption technique in lightly Bi-doped GaAs grown by the liquid encapsulation Czochralski technique.⁴⁵ The results by Kunzer et al. unambiguously demonstrated the existence of Bi_Ga in semi-insulating GaAs doped with Bi, and it is the soundness of this work what has served as starting point for further discussions on Bi_Ga in Ga(As,Bi). For instance, Bi_Ga has been recurrently suggested in the literature as the reason for a decrease in hole concentration with increasing Bi content⁴⁶ or for the anomalous compositional dependence of electronic properties in Ga(As,Bi),⁴⁷ unfortunately without proof of existence. In this sense, it is worth noting the efforts by Ciatto et al. to detect Bi_Ga in Ga(As,Bi) using extended x-ray absorption fine structure (EXAFS) spectroscopy⁴⁸ or those by Dagnelund et al. using optically detected magnetic resonance techniques.³³ In both cases, they could not find evidence of the presence of Bi_Ga in their GaAs_1−xBi_x layers. Even for In(P,Bi) alloys where Bi_In hetero-antisites formation is energetically favored compared to Bi_P,⁴⁹ detection of Bi hetero-antisites is not trivial as demonstrated in the elaborated investigation by Krammel et al. who identified potential Bi_In antisites in In(P,Bi) using a combination of cross-sectional scanning tunneling microscopy with density functional theory calculations.⁵⁰ 

Recent investigations by Gelczuk et al. of deep-level defects in MBE grown n-type GaAs_1−xBi_x with 0 < x < 0.023 using deep level transient spectroscopy (DLTS) suggest the presence of at least two Bi-related traps, which the authors attribute to the Bi-related pair defects (V_Ga+ Bi_Ga)^−/2− and (As_Ga+ Bi_Ga)^0/1−.⁵¹ For the defect identification, the authors combine the results of their experimental DLTS measurements with ab initio calculations by Luo et al.⁴⁴ Additionally, previous DLTS measurements of GaAs_1−xBi_x with x ∼ 0.01 − 0.03 by other authors found trap signatures, which were also discussed in connection to point defects and defect complexes related to As_Ga and Bi_Ga antisites.^52,53 Noteworthily, most studies suggesting the presence of Bi hetero-antisites do not specify their precise location, whether inside the layers or at interfaces.

Incidentally, we noted that the Ga(As,Bi) epilayers investigated by Gelczuk et al. were grown under remarkably similar conditions to the ones used for the QWs of this study: the growth was interrupted to decrease T_s to 378 °C before Ga(As,Bi) growth with a reduced As/Ga flux ratio of approximately 1 and, more importantly, Ga(As,Bi) growth was preceded by deposition of a Bi pre-layer for 30 s, prior to opening the Ga shutter and, hence, start growing Ga(As,Bi).⁵¹ Note that our experimental observations are consistent with the presence of Bi-related pair defects (V_Ga+ Bi_Ga)^−/2− and (As_Ga+ Bi_Ga)^0/1− as suggested by Gelczuk et al.

Based on our experience from previous investigations of Ga(As,Bi)/GaAs QWs and heterostructures and in connection to the work of Gelczuk et al., we can conclude that the realization of GI alone does not justify the creation of Bi_Ga. This conclusion relies on previous studies of Ga(As,Bi)/GaAs QW structures and epilayers grown applying similar GI at the interfaces and that exhibit the expected contrast at the interface when investigating them using g₀₀₂ DFTEM, i.e., the absence of dark-lines delimiting the interfaces.²⁴ The close similarity of our results with those from Gelczuk et al.⁵¹ indicates that the Bi pre-layer may play a more relevant role than previously anticipated. On the other hand, it becomes clear that the dark-lines at the bottom Ga(As,Bi)-on-Al(Ga.As) interface and at the upper Al(Ga,As)-on-Ga(As,Bi) interface show differences in contrast and homogeneity, see Figs. 3 and 4. Growth conditions associated with each GI are also different, not only in terms of the presence/absence of a Bi procedure but regarding the temperature regime. While GI at the bottom interface aims to decrease T_s and As/Ga flux ratio and to apply a Bi treatment before the QW, GI at the upper interface aims to increase T_s and As/Ga flux ratio toward the favorable conditions for (Al,Ga)As growth. During GI at the upper interface, any possible excess Bi after QW growth will likely be desorbed during the increase in T_s. The exact role of Al and of Bi accumulation on the surface (intentional due to a planned treatment before QW growth or unintentional after QW growth) and/or any interrelationship between them is still unclear and is the subject of current investigations. Interestingly, based on first-principles methods, Luo et al. already predicted that the usual assumption that Bi incorporates primarily substituting As does not necessary hold under certain growth conditions, prompting for further experimental verification of Bi substitution sites depending on growth conditions.⁴⁴ In this respect, we have not only given evidence of the validity of such predictions but demonstrated the presence of Bi_Ga. More importantly, we are able to track their exact location, in this case at the interfaces of the present Ga(As,Bi)/(Al,Ga)As QW. Note that, in general, studies discussing Bi_Ga hetero-antisites in dilute bismides do not specify their precise location in the layers (i.e., whether inside the layers/QWs or at interfaces). Hence, our work paves the way for future investigations aiming to disentangle the still intricate incorporation process of Bi in dilute bismide layers.

The results presented here provide a significant contribution in the identification of point defects in dilute bismides, which is fundamental to improve material quality. In particular, we report on the detection of Bi_Ga hetero-antisites, a largely anticipated defect in Ga(As,Bi) yet challenging to detect, at the interfaces of Ga(As,Bi)/(Al,Ga)As QW structures. Whereas the presence of any anomaly at the Ga(As,Bi)/(Al,Ga)As interfaces may go unnoticed in routine low magnification HAADF-STEM imaging, unexpected features are remarkably visible in g₀₀₂ DFTEM observations. Detailed analysis of the data demonstrate that the presence of Bi_Ga hetero-antisites at this specific location explain the experimental contrast. Hence, the use of g₀₀₂ DFTEM paves the way for further studies of the complex Bi incorporation into the lattice as well as its dependence on the growth conditions. Since calculations predict a large relaxation of the Bi-As interatomic distances when Bi substitutes Ga atoms,⁴⁸ taking into consideration the large concentration of Bi_Ga at the interfaces of the present samples, we envisage the detection of the local perturbations using advanced microscopy techniques and dedicated simulation tools.

The authors thank Sabine Krauß, Doreen Steffen, and Christopher Matzeck for preparation of TEM specimens and FIB lamellas, respectively, and Max Litschauer for TEM support. Special thanks are due to Achim Trampert, Roman Engel-Herbert, and John Freeland (Argonne National Laboratory) for insightful discussions. The authors thank Achim Trampert for a critical reading of the manuscript.

The authors have no conflicts to disclose.

Esperanza Luna: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Investigation (lead); Methodology (lead); Validation (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (equal). Janne Puustinen: Methodology (supporting); Writing – review & editing (equal). Joonas Hilska: Methodology (supporting); Writing – review & editing (equal). Mircea Guina: Supervision (supporting); Writing – review & editing (equal).

The data that support the findings of this study are available within the article. Additional data are available from the corresponding author upon reasonable request.