We have examined the influence of bismuth (Bi) and nitrogen (N) fluxes on N and Bi incorporation during molecular-beam epitaxy of GaAs_{1-x-y}N_{x}Bi_{y} alloys. The incorporation of Bi is found to be independent of N flux, while the total N incorporation and the fraction of N atoms occupying non-substitutional lattice sites increase with increasing Bi flux. A comparison of channeling nuclear reaction analysis along the [100], [110], and [111] directions with Monte Carlo-Molecular Dynamics simulations indicates that the non-substitutional N primarily incorporate as (N-As)_{As} interstitial complexes. We discuss the influence of Bi adatoms on the formation of arsenic-terminated [110]-oriented step-edges and the resulting enhancement in total N incorporation via the formation of additional (N-As)_{As}.

Due to the significant bandgap narrowing induced by dilute fractions of N in GaAs, dilute nitride alloys are attractive for a variety of applications, including long-wavelength lasers and detectors, ultra-high-efficiency solar cells, and high performance heterojunction bipolar transistors.^{1} However, N-related point defects often lead to degraded minority carrier transport properties and optical efficiencies.^{2,3} Co-alloying GaAsN with larger elements such as indium (In), antimony (Sb), and/or Bi allows lattice-matching to GaAs or Ge substrates and is expected to lead to significant bandgap narrowing using a substantially lower N fraction, with a correspondingly lower fraction of N-related defects.^{4,5} Bi incorporation into GaAs at a mole fraction of 0.01 leads to a bandgap reduction of ∼84 meV, much larger than the reduction for similar fractions of Sb (21 meV) or In (16 meV).^{6} In addition, the strong spin-orbit splitting induced by Bi is expected to lead to reduced nonradiative Auger recombination for Bi fractions in excess of 0.105.^{7} Thus, GaAsNBi is promising for optoelectronic applications operating in the near-infrared range.

For GaAsN and related alloys, bismuth is often reported to surface segregate without incorporation. However, the presence of a Bi flux has been reported to increase or decrease the incorporation of N. For example, metalorganic vapor-phase epitaxy (MOVPE) with a Bi flux has been reported to decrease the N fraction,^{8,9} while molecular-beam epitaxy (MBE) with a sufficiently high Bi flux has been reported to increase the N fraction.^{10–12} To date, the atomistic mechanisms for the influence of Bi flux on N incorporation remain unknown. Here, we examine the influence of Bi flux on N incorporation in GaAsNBi alloys. Using a combination of Rutherford backscattering spectrometry (RBS), nuclear reaction analysis (NRA), and high resolution x-ray rocking curves (HRXRC), we show that both the total N content and the fraction of N atoms occupying non-substitutional sites increase with increasing Bi flux. A comparison of channeling NRA measurements along the [100], [110], and [111] directions with Monte Carlo-Molecular Dynamics (MC-MD) simulations indicates that the non-substitutional N primarily occur as (N-As)_{As} interstitial complexes. We discuss the influence of Bi adatoms on the formation of arsenic-terminated [110]-oriented step-edges and the resulting enhancement in total N incorporation via the formation of additional (N-As)_{As}.

The GaAs_{1-x-y}N_{x}Bi_{y} alloy films were grown on semi-insulating (001) GaAs substrates by molecular-beam epitaxy using solid Ga, As, and Bi sources and a radio frequency nitrogen plasma source. The cracking zone of the As source was maintained at a relatively low temperature such that predominantly As_{4} was supplied.^{13,14} After an initial 500-nm thick GaAs buffer layer grown at 580 °C, the substrate temperature was held at 580 °C for a 5 min anneal and then lowered to the GaAsNBi growth temperature and held for another 5 min anneal.^{15} 100 nm thick GaAs_{1-x-y}N_{x}Bi_{y} films were then grown at 345 ± 15 °C with an As_{4}/Ga beam equivalent pressure (BEP) ratio of ∼20 and a growth rate of 1 *μ*m/h. As will be discussed below, excess As incorporation is expected to be [As_{Ga}] < 9.3 × 10^{18} cm^{−3}.^{14} A series of films with a range of N and Bi fractions were achieved by independently varying the Bi BEP from 0 to 1 × 10^{−7} Torr (with the N_{2} flow rate fixed at 0.25 sccm) and the N_{2} flow rate from 0.17 to 0.35 sccm (with Bi BEP fixed at 5.7 × 10^{−7} Torr), which we will refer to as the “Bi flux series” and the “N flux series,” respectively.

To monitor the influence of Bi and N fluxes on surface reconstruction, reflection high-energy electron diffraction (RHEED) patterns were collected along the [$110$] and [$11\xaf0$] directions for the Bi flux series and the N flux series. For all films, a (2 × 4) RHEED pattern [Figs. 1(a) and 1(b)] was observed during growth of the GaAs buffer layer at 580 °C; the pattern transitioned to a (2 × 3) pattern [Figs. 1(c) and 1(d)] as the temperature was ramped under As overpressure to 345 ± 15 °C. During GaAsNBi growth with low Bi fluxes, a (1 × 3) pattern [Figs. 1(e) and 1(f)] was observed, consistent with the Bi-induced (1 × 3) reconstruction reported by others for GaAsN growth in the presence of a Bi flux.^{10,12} At higher Bi flux, above BEP ∼5.7 × 10^{−8} Torr, a (2 × 1) reconstruction was observed instead [Figs. 1(g) and 1(h)]. For the N flux series, with the Bi flux kept constant at 5.7 × 10^{−8} Torr, all films showed either a (1 × 3) or a (2 × 1) pattern, suggesting that this Bi flux is near the threshold between the (1 × 3) induced by lower Bi fluxes and the (2 × 1) induced by higher Bi fluxes. Growth of GaAsN without a Bi flux resulted in a (2 × 1) pattern similar to Figs. 1(g) and 1(h).

For both the N and Bi flux series of GaAsNBi films, the surface morphologies and compositions were examined using a combination of atomic force microscopy (AFM), HRXRC, RBS, and NRA. AFM images were collected using a Bruker Dimension Icon in both tapping and contact modes. HRXRC measurements were performed using Cu Kα_{1} radiation. A series of Δω scans were collected near the GaAs (004) and GaAs (224) reflections. RBS and NRA were performed using a NEC tandem accelerator with a 4.46 MeV He^{+} beam. The RBS (NRA) detector was placed at 167° (135°) with respect to the incident beam to detect the backscattered He^{+} ions.^{16} We used the nuclear reaction ^{14}N(α,p)^{17}O to detect the nitrogen atoms.^{17} RBS and NRA measurements were performed in [100], [110], and [111] channeling and non-channeling conditions achieved by oscillating the specimen ±4° around the channeling condition in both phi and theta directions during spectra collection. We note that both RBS and NRA data are analyzed using the simulation of nuclear reaction analysis (SIMNRA) code, an analytical simulation program in which multiple small-angle scattering events are treated as energy broadening.^{18} To simulate the channeling NRA spectra, we use a combined Monte Carlo-Molecular Dynamics (MC-MD) approach. More details about the simulation and analysis are provided elsewhere.^{16}

HRXRC of the Bi flux series and N flux series are presented in Figs. 2(a) and 2(b). Due to their compressive and tensile misfit with respect to GaAs, the GaAsBi and GaAsN diffraction peaks appear on the low-angle and high-angle sides of the GaAs substrate peak, respectively. For the Bi flux series, the GaAsN(Bi) peak is shifted from the high-angle side with negligible Bi incorporation to the low-angle side as the Bi incorporation is increased. In addition, for the N flux series, the GaAs(N)Bi peak is shifted toward the high-angle side as the N incorporation is increased. Due to the absence of a distinct diffraction peak associated with non-stoichiometric GaAs, often termed low-termperature-grown GaAs (LT-GaAs), we estimate an upper bound for [As_{Ga}] using the full-width-half-maximum of the GaAs peak, i.e., Δ*a*/*a*_{GaAs} = 1.24 × 10^{24} × [As_{Ga}], such that [As_{Ga}] < 9.3 × 10^{18} cm^{−3}, at least two orders of magnitude lower concentration than standard LT-GaAs.^{14} AFM images for the Bi flux series and the N flux series are presented in Figs. 3(a)–3(d) and 3(e)–3(h), respectively. For both the Bi and N flux series, the surfaces appear featureless, with rms roughness <0.5 nm, consistent with observations of layer-by-layer growth of GaAsN.^{19,20}

For the N flux series, shown in Fig. 2(a), the N mole fraction, *x*, and Bi mole fraction, *y*, determined from analyses of NRA and RBS data, respectively, are indicated on the HRXRC plots. As the N_{2} flow rate increases, *x* increases monotonically. Meanwhile, *y* is unchanged with increasing N_{2} flow rate, indicating that Bi incorporation is unaffected by co-incorporation with N. For the Bi flux series, shown in Fig. 2(b), the *x* and *y* fractions determined from analyses of NRA and RBS are also indicated on the HRXRC plots. As the Bi flux increases, *y* increases monotonically. Furthermore, although the N flux is fixed, *x* also increases with Bi flux, up to a saturation value of ∼0.018.

We now discuss the RBS spectra for the Bi flux series and the NRA spectra for the N flux series. For the Bi flux series, the RBS yield vs. backscattered ion energies is plotted in Fig. 4(a); an enlarged portion is plotted in Fig. 4(b) with random spectra data overlaid with SIMNRA fitted spectra. Due to the similar atomic masses of Ga and As, the energies of the He ions backscattered from Ga and As are similar, ∼3.6 MeV, in Fig. 4(a), and similar RBS yields associated with both atoms are also observed due to their similar atomic numbers. Due to the high atomic mass of Bi, the energies of the He ions backscattered from Bi are at higher energy, ∼4.1 MeV. Both the random and channeling RBS yields associated with Bi increase with increasing Bi flux. SIMNRA fits to the random RBS spectra reveal *y* ranging from 0 to 0.059. In addition, SIMNRA fits, assuming a uniform Bi depth profile, produce Gaussian-shaped RBS yields that match with the experimental spectra, suggesting a uniform incorporation of Bi throughout the GaAsNBi film. For the channeling spectra, the distinct peaks (corresponding to He scattering from Ga and As) near ∼3.6 MeV and the asymmetric peak (associated with He scattering from Bi) at ∼4.1 MeV are due to preferential scattering from exposed surface atoms.^{21}

For the N flux series, the NRA yield vs. reaction-emitted proton energy is plotted in Fig. 4(c). Both random and channeling NRA yields associated with N increase with increasing N flux. SIMNRA fits to the random NRA spectra, using a GaAsN standard, reveal *x* ranging from 0 to 0.017; assuming a uniform N depth profile, SIMNRA fits suggests a uniform incorporation of N throughout the GaAsNBi film. The fraction of substitutionally incorporated N atoms, *f _{N-sub}*, was calculated according to

where *χ(N)* is the ratio of the channeling to the non-channeling NRA yield and *χ _{min}(GaAs)* is the ratio of the channeling to the non-channeling RBS yield for the GaAs reference film. To determine the fraction of Bi atoms incorporating substitutionally in GaAs(N)Bi films,

*f*, we used an analogous equation with

_{Bi-sub}*χ(N)*replaced by

*χ(Bi),*defined as the ratio of the channeling to the non-channeling Bi-related RBS yield. In Fig. 5, the total,

*x*, substitutional,

*x*, and interstitial,

_{sub}*x*, mole fractions of N are plotted for the Bi flux series. As discussed earlier,

_{int}*x*increases with increasing Bi BEP. In the GaAsN film (Bi BEP = 0 Torr), the fraction of N occupying substitutional sites,

*f*/

_{N-sub}= x_{sub}*x*, is 0.81, consistent with other literature reports for GaAsN.

^{15,17,22}As the Bi BEP is increased,

*f*decreases, indicating that the fraction of N atoms occupying non-substitutional sites increases with increasing Bi flux. For all Bi-containing films,

_{N-sub}*f*is ≥0.90, independent of Bi BEP or N

_{Bi-sub}_{2}flow rate.

To determine the N interstital complex configuration, the measured NRA channeling data are compared with MC-MD simulations of a 3 × 3 unit cell of GaAsNBi with a N-to-Bi incorporation ratio of 1-to-2. Within each cell, each N is positioned at the center of the group V site as either substitutional N, N_{As}; (N-N)_{As}, with N_{2} aligned along the [111] direction; or (N-As)_{As}, with the N-As pair aligned along the [010] direction.^{23–26} Within each cell, each Bi is positioned at the center of the group V site as substitutional Bi, Bi_{As}. Due to the large size of Bi atoms, we also include atomic displacement of nearest-neighbor Ga.^{27} In Fig. 6, we present a comparison of the simulated and measured NRA yields for GaAsNBi. For both the N_{Sub} and (N-N)_{As} interstitial complexes, the NRA simulations predict the highest (lowest) yields in the [100] ([111]) directions, as shown in Figs. 6(a) and 6(b), leading to a yield trend of Y_{[100]} > Y_{[110]} > Y_{[111].} In contrast, for the interstitial pair (N-As)_{As}, NRA simulations predict the highest (lowest) yield in the [111] ([100]) directions, as shown in Fig. 6(c), with a yield trend of Y_{[111]} > Y_{[110]} > Y_{[100]}. As shown in Fig. 6(d), the measured yield trend is Y_{[111]} > Y_{[110]} > Y_{[100];} this particular yield trend is predicted only for the case where (N-As)_{As} is the dominant interstitial complex. Therefore, our combined computational-experimental approach suggests that (N-As)_{As} is the dominant interstitial complex in GaAsNBi alloys, consistent with other reports for GaAsN and related dilute nitride alloys.^{28–33}

We next discuss a mechanism for Bi-induced enhancement of N incorporation based upon Bi adatom induced disordering of $1\xaf10$-oriented step edges during growth. Standard MBE growth of GaAs, with a (2 × 4) reconstruction, typically results in long terraces with step edges oriented along the $1\xaf10$ direction.^{34} Dimroth *et al.* showed that N incorporation is suppressed on (111) A offcut surfaces upon which the density of $1\xaf10$-oriented step edges is increased.^{8} On a (2 × 4) reconstructed GaAs surface, it has been shown that $1\xaf10$ step edges may be disrupted by exposure to Bi, resulting in (1 × 3) or (4 × 3) reconstruction consisting of smaller islands with a higher density of [110]-oriented step edges.^{35} Thus, for GaAsN, the surface Bi adatoms, which induce the (1 × 3) reconstruction shown in Figs. 1(e) and 1(f), may increase the density of [110] step edges, allowing increased incorporation of N atoms. Furthermore, the [110] step edge consists of As dangling bonds, such that a N atom incorporating on a [110] step edge would have an increased likelihood of forming a (N-As)_{As} interstitial complex.^{28–33} Consequently, the Bi adatom induced enhancement in N incorporation would be accompanied by an increased fraction of interstitial N, consistent with our observations.

In summary, we have investigated the influence of Bi flux on N incorporation in GaAsNBi alloys. Increasing the Bi flux results in an increase in both the total N content and the fraction of N atoms occupying non-substitutional sites, likely as (N-As)_{As} interstitial complexes. The enhancement in both N incorporation and N interstitial formation may be linked to a Bi-adatom- induced increase in the fraction of $110$-oriented step edges with As dangling bonds. This insight provides a pathway to tailored N incorporation in GaAsNBi and related alloys.

This work was supported by the National Science Foundation (Grant No. DMR 1410282) and the U.S. Department of Energy Office of Science Graduate Student Research (SCGSR) Program. Partial support was also provided by the Center for Integrated Nanotechnologies (CINT), a DOE nanoscience user facility jointly operated by Los Alamos and Sandia National Laboratories.