The ternary alloy GaAs1–xBix is a potentially important material for infrared light emitting devices, but its use has been limited by poor optical quality. We report on the synthesis of GaAs1–xBix epi-layers that exhibit narrow, band edge photoluminescence similar to other ternary GaAs based alloys, e.g., InyGa1–yAs. The measured spectral linewidths are as low as 14 meV and 37 meV at low temperature (6 K) and room temperature, respectively, and are less than half of previously reported values. The improved optical quality is attributed to the use of incident UV irradiation of the epitaxial surface and the presence of a partial surface coverage of bismuth in a surfactant layer during epitaxy. Comparisons of samples grown under illuminated and dark conditions provide insight into possible surface processes that may be altered by the incident UV light. The improved optical quality now opens up possibilities for the practical use of GaAs1–xBix in optoelectronic devices.

Bismuth has become an intriguing alloying element for III–V semiconductors because of its ability to substantially change their electronic band structures. For example, the addition of bismuth to GaAs causes its band gap energy to drop by 88 meV/% and the spin orbit splitting energy to increase by nearly the same amount.1–3 The remarkable tunability of these parameters with only a small change in composition makes GaAs1−xBix a potentially useful material for many optoelectronic applications including: next generation multijunction solar cells,4 long wavelength light emitters,5,6 transistors,7,8 and spintronic devices.3 Despite these promising characteristics, difficulty in synthesizing epi-layers with high optical quality has thus far tempered the technological impact of this material. The linewidths of its low temperature (<150 K) photoluminescence (PL) spectra are typically exceedingly broad (>60 meV), and band edge emission is often suppressed by recombination at low energy states.9–11 This behaviour is caused in part by an imperfect homogeneous crystal and is undesirable for high performance devices. The growth of GaAs1–xBix and related alloys must therefore be substantially improved in order enable their use in technological applications.

GaAs1–xBix was only first synthesized by metal organic vapour phase epitaxy (MOVPE) by Oe and Okamoto in 1998 and by molecular beam epitaxy (MBE) by Tiedje et al. in 2003 long after other III–V semiconductor alloys had been well established.1,12,13 Bismuth's large atomic radius and relatively high vapor pressure cause it to act as a natural surfactant and inhibit it from readily incorporating into the As sub-lattice. Consequently, reduced growth temperatures (<400 °C) and near-stoichiometric V/III flux ratios are necessary to achieve bismuth concentrations on the order of an atomic percent.1,9,14,15 Epitaxy of dilute bismide alloys must also strike a delicate balance between promoting bismuth incorporation into the lattice and preventing the build-up of bismuth on the surface. Too little bismuth flux coupled with low substrate temperatures and high growth rates leads to additional arsenic incorporation and high defect densities. The presence of a bismuth wetting layer on the growth surface alternatively serves the purpose of providing a bismuth reservoir for incorporation, and its surfactant effect helps to reduce defect densities.16,17 Growth conditions that allow the formation of a full Bi wetting layer have been shown to be effective in growing dilute bismide alloys, but this practice also results in broad luminescence and additionally runs the risk of allowing Bi droplets to form.16,19 Instead, it appears best to take advantage of the surfactant properties of bismuth while suppressing full wetting layer formation.

In this paper, we show that high optical quality GaAs1–xBix epi-layers can be produced under an intermediate growth regime utilizing a partial bismuth wetting layer. This approach was used to synthesize alloys with substantially reduced defect densities compared to previous reports and low temperature PL linewidths < 15 meV, which is 4 times narrower than the best previously reported films.9,10,14 The nature and advantages of this growth regime will be discussed. We also show that above bandgap light stimulation of the growth surface can potentially widen the narrow window of this growth regime. Our approach now makes it possible to fabricate high performance GaAs1–xBix-based optoelectronic devices.

Samples were grown in an Omicron EVO25 MBE system on GaAs (001) substrates under irradiation from a pulsed UV laser. Dual filament effusion cells were used for the gallium and bismuth sources, and As2 flux was produced by a two-zone valved cracker source. The growth chamber is fitted with a 4.5 in. bakable sapphire viewport on a 10 in. bottom flange for direct optical access to the growth surface. This UV transparent viewport is frequently baked to remove any arsenic build up that would decrease optical transmission. Above band gap light was sourced from a 248 nm KrF excimer laser with a 25 ns pulsewidth, operated at a repetition rate of 10 Hz and focused into a 7 × 7 mm spot size aligned to the center of the 142in. substrate. After losses, the laser energy at the illuminated area of the sample was approximately 20–30 μJ/pulse. Due to the relative absence of UV light at the edges of the substrate, these areas were used as a comparison to dark growth conditions.

Beam equivalent pressures (BEPs) were measured using an in-situ ion gauge, where gauge sensitivities were taken into consideration when calculating flux ratios.20 Growth was carried out at nearly stoichiometric As:Ga flux ratios. For each sample, Reflection High Energy Electron Diffraction (RHEED) was used to determine the arsenic valve position at which the surface reconstruction changed from a 2 × 4 to 1 × 1 pattern indicating a stoichiometric V:III flux ratio.21 For the growth of each GaAs1−xBix epi-layer, the valve position was then set to achieve a V/III flux ratio slightly above unity. The relative change in As2 flux as a function of valve position is not affected by the variations in the source flux. This method eliminates any inconsistencies due to day-to-day fluctuations in the source fluxes and results in reproducible bismuth concentrations in deposited epi-layers. Therefore, a RHEED based approach to setting the As2 flux is recommended over simply comparing measured BEP's, especially because BEP ratios at the surface of the as-growing epi-layer can easily be incorrectly estimated.

Samples were grown at a rate of 1 μm/h, which corresponds to a gallium flux of 3.5nm2s1. Gallium flux and growth rate were further verified with RHEED intensity growth oscillations, as well as by epi-layer thickness values obtained from the Pendellosung fringes in ex-situ X-ray diffraction (XRD) measurements on specially prepared separate samples. Substrate temperatures were monitored using a k-Space BandiT band edge thermometry system with an accuracy of ±1.0 °C, which was calibrated using the amorphous arsenic surface layer and native oxide desorption temperatures identified with RHEED. Note that this calibration procedure can result in an overall absolute error of up to 10 °C from the actual substrate temperature, but it does not contribute to any relative error in substrate temperatures between samples. The growth temperature quoted was measured at the central, illuminated spot of the sample. It is assumed that the band edge absorption thermometry measures the bulk substrate temperature and not just the temperature at the surface.

Prior to growth the substrates were outgassed at 300 °C in an ultra-high vacuum (UHV) buffer chamber and the oxide was thermally desorbed (Tsubstrate > 610 °C) in the growth chamber under As2 overpressure. A GaAs buffer layer greater than 300 nm thick was grown at a nominal growth temperature of 580 °C. The buffer layer growth was interrupted to find the valve position for stoichiometric As2 flux under dark conditions. Growth was then resumed until a second interruption, which was used to lower the substrate temperature to the desired value for the growth of the GaAs1–xBix epi-layer. The arsenic overpressure was also reduced to nearly stoichiometric levels during this time. Growth of the GaAs1–xBix epi-layers was first initiated with 15–20 s of bismuth flux before the gallium shutter was opened. The laser was also turned on at approximately this time. The resulting GaAs1–xBix epi-layers all had a nominal thickness of 300 nm and were post growth annealed in the chamber under As2 overpressure at a temperature of 600 °C for 10 min in order to remove any remaining bismuth on the surface. Samples grown under the same conditions without post-growth annealing did not show surface metallic, bismuth, and/or gallium, droplets. No GaAs cap layer was added after the growth of the GaAs1–xBix epi-layers.

Epi-layer compositions were determined by fitting symmetric XRD scans of the (004) GaAs peak assuming a GaBi lattice constant of 6.33 Å.1 All epi-layers were found to have excellent crystalline quality based on epi-layer XRD full width half max (FWHM) and the observation of strong Pendellosung fringes. No variation in growth rate was observed between illuminated and dark sample areas. Photoluminescence measurements were performed using a 532 nm continuous wave laser and an Andor Shamrock 0.5 m spectrometer calibrated with a neon discharge lamp. Samples were cooled to 6 K in an ARS closed cycle helium optical cryostat.

Full details of the samples grown as part of this study are given in Table I. Samples 1–4 were grown under the same conditions, except for a variation in substrate temperature, while sample 5 was grown with increased bismuth flux and samples 6 and 7 were grown with varied As2 overpressure. At the outset of the discussion, it is important to note that a single bismuth flux was chosen for the majority of the study. Its value was selected to achieve between 0.1% and 1%, where subtle differences between samples could more easily be distinguished in the PL spectra. The actual bismuth concentrations in the samples were observed to vary slightly as a function of substrate temperature, arsenic overpressure, and illumination conditions. Low temperature PL spectra from these samples are shown in Fig. 1. The dark and illuminated areas of all samples had comparable PL intensities compared to a 10× InGaAs quantum well structure;18 the brightest sample was 100× more intense. A similar structure has been used previously to benchmark GaAs1−xBix PL intensity and the PL of our samples is greater than those by more than an order of magnitude.18 However, the illuminated areas of samples 3–7 were considered high optical quality due to the presence of narrow band edge emission and the relative lack of below band edge emission from bound states. Illuminated samples grown at or above 358 °C emit band edge PL with linewidths less than 30 meV at low temperatures, which is much narrower than previously reported values of 60 meV or greater.10,11,19,23 The narrowest linewidth observed was 14 meV. This is comparable to other III–V GaAs based semiconductor alloys.22 The PL spectra of the dark sample areas, on the other hand, are dominated by below band edge luminescence. This emission differs from previously reported low temperature PL in GaAs1−xBix, in both the peak energy and the presence of structure in the lineshape rather than a broad featureless peak.11,24 In particular, the structure of this emission includes peaks at several discrete energies independent of bismuth composition, indicated by the vertical dashed lines. Such behaviour is similar to GaNzAs1−z and suggests that these features are possibly associated with recombination at bismuth pair and/or cluster states.25–27 Interestingly, the dark sample area of sample 6 (grown with an increased arsenic flux) shows band edge emission, and the illuminated area shows both band edge and below band edge emission. The origin of this behaviour will be discussed below.

TABLE I.

Details of the GaAs1−xBix samples shown in Fig. 1.

TgrowthFlux (nm−2 s−1)[Bi] (±0.01%)6 K PL
#(°C)FAsFBiIll.DarkFWHM (meV)
348 3.9 0.022 1.31 1.31 N/A 
355 3.9 0.022 1.25 1.25 N/A 
358 3.9 0.022 0.4 0.72 27 
361 3.9 0.022 0.16 0.8 17 
358 3.9 0.035 0.27 0.6 14 
360 4.1 0.022 0.11 0.25 17 (ill.) 
      22 (dark) 
358 3.8 0.022 0.23 0.67 20 
TgrowthFlux (nm−2 s−1)[Bi] (±0.01%)6 K PL
#(°C)FAsFBiIll.DarkFWHM (meV)
348 3.9 0.022 1.31 1.31 N/A 
355 3.9 0.022 1.25 1.25 N/A 
358 3.9 0.022 0.4 0.72 27 
361 3.9 0.022 0.16 0.8 17 
358 3.9 0.035 0.27 0.6 14 
360 4.1 0.022 0.11 0.25 17 (ill.) 
      22 (dark) 
358 3.8 0.022 0.23 0.67 20 
FIG. 1.

Low temperature (6 K) normalized PL comparison of GaAs1−xBix samples grown as a function of substrate temperature. Full sample details are given in Table I. Narrow band edge luminescence is observed for illuminated samples grown above Tgrowth ≃ 358 °C. Dark samples show predominantly broad low energy PL emission. The discrete energies of the below band gap states are indicated by the vertical dashed grey lines.

FIG. 1.

Low temperature (6 K) normalized PL comparison of GaAs1−xBix samples grown as a function of substrate temperature. Full sample details are given in Table I. Narrow band edge luminescence is observed for illuminated samples grown above Tgrowth ≃ 358 °C. Dark samples show predominantly broad low energy PL emission. The discrete energies of the below band gap states are indicated by the vertical dashed grey lines.

Close modal

Figure 2 shows the bismuth composition of a series of GaAs1−xBix samples grown under the same incident As2 flux conditions but with different substrate temperatures. The incorporated bismuth concentration decreases sharply at substrate temperatures above 358 °C due to the relatively high vapour pressure of bismuth and the associated increase in the surface desorption rate.28 Note that this temperature at the inflection of the curve corresponds to the temperature that produces narrow band edge PL. The lack of a significant difference in the trends of the dark and illuminated samples suggests that illumination does not heat the sample. Our previous work with UV light assisted epitaxy of GaAs also indicated that changes in the growth induced by illumination are not simply due to incidental sample heating but can be attributed to electronic effects caused by the presence of photo-generated carriers.29,30

FIG. 2.

Bismuth composition of a series of GaAs1−xBix samples grown with the same bismuth flux as a function of growth temperature which was measured at the illuminated spot of the sample.

FIG. 2.

Bismuth composition of a series of GaAs1−xBix samples grown with the same bismuth flux as a function of growth temperature which was measured at the illuminated spot of the sample.

Close modal

The bismuth concentration for a set of GaAs1−xBix samples grown under varied As2 overpressure (with all other growth conditions held constant) is shown in Fig. 3. In all cases, the V/III ratio was varied above unity, calibrated using the change in the RHEED pattern. As noted above, this method minimizes any issues with day-to-day fluctuations in source flux. Nonetheless, the samples in this subset were also grown over a short period of time to further mitigate unwanted changes. The decreasing trend of bismuth incorporation with increasing As2 flux for both illuminated and dark sample areas is in agreement with Lewis et al.16 Additionally, illumination causes an overall reduction in the bismuth concentration compared to the dark sample areas.

FIG. 3.

Bismuth composition for a set of GaAs1−xBix samples grown under the same growth conditions where only the As2 overpressure was varied. The trend is in agreement with Lewis et al.16 The illuminated area of the samples has lower bismuth concentrations than their dark sample counterparts.

FIG. 3.

Bismuth composition for a set of GaAs1−xBix samples grown under the same growth conditions where only the As2 overpressure was varied. The trend is in agreement with Lewis et al.16 The illuminated area of the samples has lower bismuth concentrations than their dark sample counterparts.

Close modal

The temperature dependence of the PL spectra for sample 3 is shown in Fig. 4. The measured bismuth concentrations within the illuminated and dark areas of the sample were 0.4% and 0.72%, respectively. Emission from the dark area is dominated by a broad low energy PL band at low temperatures (<200 K), and band edge emission only emerges above 100 K. The presence of low energy emission is generally consistent with previous reports for GaAs1−xBix.11,31 The illuminated sample area, on the other hand, emits from near the band edge over the entire temperature range. At temperatures below 40 K, the slightly lower emission energy reflects strong carrier localization at potential fluctuations associated with the incorporated bismuth atoms.11 Above 40 K, the high energy peak realigns to the band edge, and at intermediate temperatures (50–150 K) broad emission at lower energy (<1.4 eV) is also observed. This indicates the presence of some low energy states that can be populated once carriers have enough thermal energy to overcome localization at potential fluctuations. Phonon replicas with an energy shift close to the value of GaAs are also present just below the primary PL peak at temperatures below 25 K.22 Phonon replicas are typically not observed in GaAs1−xBix, and this is another indication of the excellent optical quality of these samples. At room temperature, both sample areas show band edge emission in agreement with the measured bismuth content. However, the PL peak from the dark area of the sample has a prominent low energy tail consistent with a higher density of shallow bound states. The 300 K FWHM for the illuminated and dark sample areas are 37 meV and 41 meV, respectively, and these spectra are of comparable intensity. Both of these values are narrower than previously reported room temperature FWHM values for GaAs1−xBix, which are close to 80 meV.9,10,32

FIG. 4.

Temperature dependence of the PL emission for GaAs1−xBix sample, the illuminated area has a bismuth concentration of 0.4% and the dark area has a bismuth concentration of 0.72% (sample 3 shown in Fig. 1). The illuminated area shows narrow band edge over the entire range of temperatures, whereas the dark area only shows band edge emission at high temperatures.

FIG. 4.

Temperature dependence of the PL emission for GaAs1−xBix sample, the illuminated area has a bismuth concentration of 0.4% and the dark area has a bismuth concentration of 0.72% (sample 3 shown in Fig. 1). The illuminated area shows narrow band edge over the entire range of temperatures, whereas the dark area only shows band edge emission at high temperatures.

Close modal

Figure 5 shows the temperature dependence of the peak position and FWHM of the PL spectra displayed in Fig. 4. Values for the dark sample area are only given above 150 K, where the band edge peak is clearly observable. The peak energies follow the Varshni model for both regions of the sample. The fit uses α and β constants for GaAs and low temperature band gap energies, Eg(0 K), of 1.485 eV and 1.450 eV for the illuminated and dark sample areas, respectively. These energy values agree with the bismuth compositions measured by XRD taking into account the well known band gap shift with bismuth incorporation.1 The minor S-shape in the peak PL energy of the illuminated sample below 40 K is due to strong carrier localization at potential fluctuations, similar to GaNzAs1−z.33 The PL peak FWHM values measured in the illuminated sample area decrease over the entire measured temperature range with a minimum value of 11 meV at 50 K. At temperatures below 50 K, the peak is actually comprised of two peaks, which broadens the FWHM. The FWHM values of the PL peak in the dark sample area show a weaker temperature dependence and are broader than the illuminated sample area.

FIG. 5.

Temperature dependence of the PL peak emission and FWHM for GaAs1−xBix sample 3. The measured PL peak energies agree well with the Varshni model for GaAs with Eg(0 K) in agreement with the bismuth concentration from XRD for both the illuminated and dark sample areas.

FIG. 5.

Temperature dependence of the PL peak emission and FWHM for GaAs1−xBix sample 3. The measured PL peak energies agree well with the Varshni model for GaAs with Eg(0 K) in agreement with the bismuth concentration from XRD for both the illuminated and dark sample areas.

Close modal

To understand the origin of the improved optical quality in our GaAs1−xBix samples, particularly those grown under illumination, we begin by considering the Lewis/Tiedje model for GaAs1−xBix growth.16 This model describes the bismuth atoms at the surface as either incorporated into the epi-layer by being bonded to an exposed gallium atom or segregated from the surface in a surfactant layer. The bismuth incorporation is given as16 

dxdtθGaθBia1xFGaa2xe(U1kBT),
(1)

where x is the amount of incorporated bismuth, θk is the coverage of the as-growing surface by a given atomic species k, FGa is the flux of gallium atoms in nm−2 s−1, U1 is the activation energy of bismuth moving from an incorporated site to the surfactant layer, T is the substrate temperature, and a1 and a2 are constants. The three terms in Eq. (1) represent the following surface processes: a bismuth atom finding an exposed gallium atom, a gallium atom covering an incorporated bismuth atom, and the thermal ejection of an incorporated bismuth atom to the surfactant layer. This surfactant layer is further assumed to interact only weakly with the growth surface. Bismuth atoms incorporated into as-growing epi-layer are added to the incorporated amount, x, while bismuth atoms segregated to the surfactant layer add to the overall surface coverage, θBi. These two layers are assumed to be in thermal equilibrium, where bismuth atoms readily move from incorporated sites to the surfactant layer and vice versa. For the purposes of this discussion we will assume that desorption only occurs from the surfactant layer and any other desorption may be treated as an overall reduction in the in-coming flux. A steady state solution to Eq. (1) yields the expected incorporation for a given set of growth conditions.

The surface coverage of the bismuth surfactant layer can be described as a Langmuir isotherm17 

θBi=bFnete(UokBT)1+bFnete(UokBT),
(2)

where the constant b is related to bismuth site area and desorption attempt frequency, Uo is the bismuth desorption energy with a value of 1.8 eV, and Fnet is the net bismuth flux arriving at the surface.17 Here, Fnet is simply considered to be the incoming flux from the source, FBi, less the amount incorporated into the epi-layer, xFGa. This can be thought of as the supply and demand for incorporated bismuth in the epi-layer. The supply is the in-coming flux of bismuth atoms and the demand is the expected bismuth incorporation based on a steady state solution to Eq. (1), which is set by the growth conditions.

When the supply is less than the demand, a surfactant layer does not form, i.e., θ = 0. This growth regime is shown schematically in the upper left hand corner in Fig. 6. The Lewis/Tiedje model assumes a surface coverage of bismuth atoms is necessary for bismuth incorporation, i.e., incorporation comes from the surfactant layer. However, upon closer consideration of the surface processes that the equation attempts to encapsulate, it is possible to achieve bismuth incorporation under this model with zero bismuth coverage. This is due to the fact that the model delineates bismuth atoms present at the surface into two groups: bonded (incorporated) or segregated to the surfactant layer. The absence of the surfactant layer does not negate incorporation, and bismuth incorporation can result directly from the atomic bismuth flux. One example of GaAs1−xBix growth under this regime is the work by Ptak et al. in Ref. 15, where no bismuth surface coverage is expected based on Eq. (2) and the reported growth conditions. In this growth regime, the demand of incorporation is greater than the supply, thus the incorporated amounts of bismuth will be linearly dependent on the net bismuth flux, as was experimentally found to be the case.15 Such conditions are advantageous in that they do not allow metallic surface droplets to form when incorporating small amounts of bismuth, and they lead to reproducible bismuth concentrations. The growth conditions reported in Ref. 9 by Bastiman et al. also result in zero bismuth surface coverage due to the alternate reason that the high (≃400 °C) substrate temperatures that were used promote bismuth desorption.

FIG. 6.

Schematic diagram of the three regions of GaAs1−xBix growth based on the Lewis/Tiedje model: θBi = 0 (upper left), 0 < θBi < 1 (shaded area), excess bismuth (lower right). The conditions for sample 3 with narrow, band edge PL are shown.

FIG. 6.

Schematic diagram of the three regions of GaAs1−xBix growth based on the Lewis/Tiedje model: θBi = 0 (upper left), 0 < θBi < 1 (shaded area), excess bismuth (lower right). The conditions for sample 3 with narrow, band edge PL are shown.

Close modal

If supply is greater than demand, we expect surface droplets to form. The presence of which is well known and great efforts were made to avoid.16 This growth regime occurs under lower growth temperatures where desorption is negligible (marked in the lower right hand corner of Fig. 6). Note that Ptak et al. detected surface droplets when large amounts of bismuth are incorporated due to significant amount of bismuth segregation back to the surface. This additional source of bismuth atoms depends on the incorporated amount and appears not to be an issue for the growth of GaAs1−xBix epi-layers with low (<5%) bismuth concentration.

There does exist a small region where supply is equal to demand. This is achieved by balancing the incoming flux, incorporation, and desorption to produce a surfactant layer of partial bismuth surface coverage, i.e., 0 < θBi < 1. The non-negligible desorption from the surfactant layer due to the elevated substrate temperature prevents build up of bismuth at the surface; bismuth desorption is thus treated as a further demand. Such a regime is marked as the shaded region in the center of Fig. 6. In this case, we expect that bismuth incorporation results from bismuth atoms moving from the surfactant layer to incorporation sites, as described by Lewis/Tiedje.16 Our results show that such conditions can produce GaAs1−xBix epi-layers with narrow, band edge PL emission. As an example, the growth conditions and incorporated amount for sample 3 (illuminated) are indicated in Fig. 6. These growth conditions yield a surface coverage of close to 50%, and Eq. (2) suggests an expected bismuth incorporation in agreement with the experimentally measured value.

To confirm the above conjecture that the growth conditions of the samples discussed above produced a partial bismuth surfactant layer, the bismuth surface coverage, θBi, was directly probed as a function of substrate temperature. Percent coverage was determined by measuring the variation of the RHEED spectral spot intensity as a function of the substrate temperature under incident bismuth and As2 flux for both illuminated and dark conditions. Uncovered surfaces show a strong RHEED spectral intensity which is reduced in proportion to the amount of bismuth on the surface. A surface was considered to be completely covered when the RHEED spectral intensity no longer decreased with decreasing substrate temperature. Partial surface coverage is determined by the relative drop in the RHEED spectral intensity between the initial high value for an uncovered surface and the minimum value for a completely covered surface. As2 overpressure was varied between V/III ratios of greater than 5:1 to 1.1:1 to emulate GaAs1−xBix growth conditions, and the bismuth was flux set to the same value used for the actual sample growth. The results are shown in Fig. 7 along with the Langmuir isotherm from Eq. (2).17 The experimental values follow the expected general theoretical trend, although the model does not capture the steepness of the experimentally measured transition under these specific conditions. This discrepancy is possibly due to the fact that the experiment was carried out at lower substrate temperatures and bismuth fluxes compared to the work by Young et al. in Ref. 17, from which the value of Uo was determined. Bismuth adatoms may also interact more strongly at the present growth conditions, which would affect the desorption activation energy. Regardless, a surface coverage of close to 50% was found to occur at substrate temperatures close to 358 °C, in agreement with our above assumption of a partial surfactant layer during the growth of the GaAs1−xBix epi-layers. We find no observable difference between the surface coverage, and by inference the bismuth adatom desorption rate, between the illuminated and dark conditions, neither do we see any effect with varied As2 overpressure. Since bismuth desorption is highly temperature dependent in this range, the negligible difference in the desorption rates for illuminated and dark conditions again verifies that there is no change in the surface temperature at the location of the incident UV light relative to the measured bulk substrate temperature. It subsequently confirms the earlier assumption that the resulting differences in the optical quality of our GaAs1−xBix samples are not simply due to incidental substrate heating.

FIG. 7.

Bismuth surface coverage, θBi, as a function of substrate temperature as measured by RHEED, where a bismuth flux of 0.022 nm−2 s−1 was used. The measured surface coverage is consistent with the trend of bismuth incorporation and the discussed model of GaAs1−xBix growth. The dashed curve indicates the expected Langmuir isotherm based on Young et al.17 

FIG. 7.

Bismuth surface coverage, θBi, as a function of substrate temperature as measured by RHEED, where a bismuth flux of 0.022 nm−2 s−1 was used. The measured surface coverage is consistent with the trend of bismuth incorporation and the discussed model of GaAs1−xBix growth. The dashed curve indicates the expected Langmuir isotherm based on Young et al.17 

Close modal

Our results indicate that the presence of a partial bismuth surfactant layer is critical for achieving high optical quality. Bismuth coverage was not measured directly for the samples discussed above, but based on the data in Fig. 7 we see a strong correlation between the presence of bismuth on the surface and narrow band edge PL. The effect of a surfactant layer is well known to improve the optical quality of other alloys, including GaNzAs1−z and AlwGa1−wAs.17,34 This approach therefore appears to be preferable to direct incorporation from incident bismuth flux without a surfactant layer (i.e., supply less than demand). However, a full bismuth surface coverage is also not ideal as it can lead to the formation of surface droplets (i.e., supply greater than demand). From the standpoint of achieving a partial surface coverage, the substrate temperature plays a critical role, that is the substrate temperature must be set such that there exists a partial surfactant layer for a given bismuth flux and not simply be low enough to prevent significant desorption (i.e., supply equal to demand). This conclusion can be drawn by comparing samples grown at different temperatures and under dark conditions to remove any additional effects of the illumination. Consider sample 3 (0.72% Bi, growth T = 358 °C) and sample 4 (0.8% Bi, growth T = 361 °C). Sample 4 was grown at a higher temperature resulting in a lower bismuth coverage than sample 3 and it exhibits PL from deeper energy bound states than does sample 3, suggesting that the presence of the deep states is not necessarily correlated with higher bismuth concentrations. A further comparison can be made to sample 5 (0.6% Bi, growth T = 358 °C), which was grown at the same temperature as sample 3 and therefore should have the same coverage. It exhibits the same PL spectrum as sample 3. The end result is that samples 3 and 5, grown at the same temperature with roughly the same partial bismuth surface coverage, both exhibit PL associated with higher optical quality than sample 4 which was grown with a lower bismuth surface coverage. This suggests that the substrate temperature and therefore the bismuth surface coverage is the most important variable for optical quality.

We also want to understand how illumination helps further improve the optical quality of the samples. Several possibilities exist. In general, the illuminated sample areas have a lower incorporated amount of bismuth than the dark areas for a given As2 and Bi flux. This result indicates that the incident UV light is potentially changing the dynamics of how arsenic and bismuth incorporate into the lattice. In the Tiedje/Lewis model, bismuth adatoms attach to gallium atoms that are exposed due to the low arsenic coverage and these incorporated bismuth atoms are in thermal equilibrium with the surfactant layer. One possible mechanism is that energy imparted by the illumination and/or the presence of photogenerated carriers reduces the activation energy of an incorporated bismuth adatom to escape to the surfactant layer and/or the likelihood of an incorporated bismuth atom to be replaced by an incoming arsenic atom. It is our general observation that a higher As2 flux resulted in lower bismuth concentration under dark conditions (see Fig. 3). This is expected based on the Tiedje/Lewis model because increased arsenic flux reduces the number of potential bismuth incorporation sites. The illuminated sample grown under the lowest As2 flux has a similar bismuth concentration as the dark sample grown under the highest As2 flux (see Fig. 3). This too may be an indication that UV light may be effecting how bismuth adatoms bind to the exposed gallium. Both mechanisms lead to higher bismuth surface coverage, either from ejected bismuth atoms from the incorporated layer or from arsenic preventing bismuth incorporation in the first place. This higher bismuth surface coverage then helps to improve the optical quality of the samples due to the surfactant effect. Finally, we note that it is also possible that the presence of photogenerated carriers affects the preferred configuration in which bismuth atoms reconstruct on the surface or incorporate as neighbours.35 This mechanism would affect the concentration of bismuth pair states, which appear in the low temperature PL spectra.

In conjunction with the above discussion, we note that the dark sample grown under an increased As2 flux (sample 6) also exhibited narrow band edge PL emission at low temperatures. This particular occurrence might be the result of achieving ideal Bi/As or As/Ga flux ratios within the small growth process window present under dark conditions. It may also be the result of some small amount of stray light. In any case, we have only observed such low temperature band edge PL in this one sample grown under dark conditions, suggesting that the window of ideal growth conditions is small. This result indicates that incident UV light may not be strictly necessary, but instead it may help to achieve the partial bismuth surface coverage and therefore only enlarges the small range of growth conditions under which high optical quality GaAs1−xBix samples can be grown. However when the breadth of work on the growth of GaAs1−xBix is considered with no prior reports of samples of similar optical quality, we believe that the incident UV irradiation is playing some crucial role.

To further explore the possible roles of illumination during growth changes the concentration of deep traps, several devices fabricated on a pair of n-type and p-type GaAs1−xBix samples with approximately 0.3% bismuth were measured by Deep Level Transient Spectroscopy (DLTS). These layers were grown under identical conditions to those of sample 3. The sample structure is the same as those in Ref. 36 with a 500 nm thick GaAs1−xBix layer doped either p-type or n-type to observe either hole or electron traps, respectively. No significant differences in deep trap (not observed in PL) concentrations or thermal emission activation energies were detected between devices fabricated within the illuminated and dark sample areas. This work is on-going and will be published at a later date.

We have shown for the first time that it is possible to produce GaAs1−xBix samples with narrow, band edge photoluminescence over the temperature range of 40–300 K. Overall the improvement in the optical quality and reduction in the density of shallower defect states compared to previously reported samples may now allow GaAs1−xBix to be incorporated into practical optoelectronic applications. This is especially true in cases where emission intensity, linewidth, or carrier extraction are important factors. Reported PL linewidths as low as 14 meV are less than a quarter of previously reported values. Our results indicate that the presence of a partial bismuth surfactant layer is critical for achieving such high optical quality material. While the precise role of the incident UV light plays remains unclear, light stimulation does not appear to affect the surface desorption rate of bismuth or the density of deep defect states. Based on the additional effect that As2 flux has on bismuth incorporation, we propose that the UV light affects the adatom incorporation/ejection mechanisms. Specifically, it could influence the movement of bismuth atoms to the surfactant layer due to an exchange process with an arsenic atom and/or the probability that bismuth atoms will incorporate as neighbours.

We acknowledge the financial support of the Department of Energy, Office of Science, Basic Energy Sciences under DE-AC36-O8GO-28308. We thank M. C. Tarun and P. M. Mooney at Simon Fraser Univ. for the preliminary DLTS results.

1.
S.
Tixier
,
M.
Adamcyk
,
T.
Tiedje
,
S.
Francoeur
,
A.
Mascarenhas
,
P.
Wei
, and
F.
Schiettekatte
,
Appl. Phys. Lett.
82
,
2245
(
2003
).
2.
T.
Tiedje
,
E. C.
Young
, and
A.
Mascarenhas
,
Int J. NanoTech.
5
,
963
(
2008
).
3.
B.
Fluegel
,
S.
Francoeur
,
A.
Mascarenhas
,
S.
Tixier
,
E. C.
Young
, and
T.
Tiedje
,
Phys. Rev. Lett.
97
,
067205
(
2006
).
4.
F.
Dimroth
,
A.
Howard
,
J. K.
Shurtleff
, and
G. B.
Stringfellow
,
J. Appl. Phys.
91
,
3687
(
2002
).
5.
Y.
Tominaga
,
Y.
Kinoshita
,
K.
Oe
, and
M.
Yoshimoto
,
Appl. Phys. Lett.
93
(
13
),
131915
(
2008
).
6.
R. B.
Lewis
,
D. A.
Beaton
,
X.
Lu
, and
T.
Tiedje
,
J. Cryst. Growth
311
,
1872
(
2009
).
7.
P. M.
Asbeck
,
R. J.
Welty
,
C. W.
Tu
,
H. P.
Xin
, and
R. E.
Welser
,
Semicond. Sci. Technol.
17
(
8
),
898
906
(
2002
).
8.
I.
Hase
, U.S. patent 7,009,225 (2006).
9.
F.
Bastiman
,
A. R. B.
Mohmad
,
J. S.
Ng
,
J. P. R.
David
, and
S. J.
Sweeney
,
J. Cryst. Growth
338
,
57
61
(
2012
).
10.
A. R.
Mohmad
,
F.
Bastiman
,
C. J.
Hunter
,
J. S.
Ng
,
S. J.
Sweeney
, and
J. P. R.
David
,
Appl. Phys. Lett.
99
,
042107
(
2011
).
11.
G.
Pettinari
,
A.
Polimeni
,
M.
Capizzi
,
J. H.
Blokland
,
P. C. M.
Christianen
,
J. C.
Maan
,
E. C.
Young
, and
T.
Tiedje
,
Appl. Phys. Lett.
92
,
262105
(
2008
).
12.
K.
Oe
and
H.
Okamoto
,
Jpn. J. Appl. Phys., Part 2
37
,
L1283
(
1998
).
13.
K.
Oe
,
Jpn. J. Appl. Phys., Part 1
41
,
2801
(
2002
).
14.
X.
Lu
,
D.
Beaton
,
T.
Tiedje
,
R.
Lewis
, and
M. B.
Whitwick
,
Appl. Phys. Lett.
92
,
192110
(
2008
).
15.
A. J.
Ptak
,
R.
France
,
D. A.
Beaton
,
K.
Alberi
,
J.
Simon
,
A.
Mascarenhas
, and
C.-S.
Jiang
,
J. Cryst. Growth
338
(
1
),
107
110
(
2012
).
16.
R. B.
Lewis
,
M.
Masnadi-Shirazi
, and
T.
Tiedje
,
Appl. Phys. Lett.
101
,
082112
(
2012
).
17.
E. C.
Young
,
S.
Tixier
, and
T.
Tiedje
,
J. Cryst. Growth
279
,
316
(
2005
).
18.
E. C.
Young
, Ph.D. thesis,
Univ. of British Columbia
,
Vancouver, Canada
,
2006
.
19.
X.
Lu
,
D. A.
Beaton
,
R. B.
Lewis
,
T.
Tiedje
, and
Y.
Zhang
,
Appl. Phys. Lett.
95
,
041903
(
2009
).
20.
V. V.
Preobrazhenskii
,
M. A.
Putyato
,
O. P.
Pchelyakov
, and
B. R.
Semyagin
,
J. Cryst. Growth
201/202
,
170
(
1999
).
21.
M.
Masnadi-Shirazi
,
D. A.
Beaton
,
R. B.
Lewis
,
X.
Lu
, and
T.
Tiedje
,
J Cryst. Growth
338
(
1
),
80
84
(
2012
).
22.
S.
Adachi
,
Properties of Group-IV, III-V and II-VI Semiconductors
(
Wiley
,
2005
).
23.
A. R.
Mohmad
,
F.
Bastiman
,
J. S.
Ng
,
S. J.
Sweeney
, and
J. P. R.
David
,
Appl. Phys. Lett.
98
,
122107
(
2011
).
24.
S.
Imhof
,
A.
Thränhardt
,
A.
Chernikov
,
M.
Koch
,
N. S.
Köster
,
K.
Kolata
,
S.
Chatterjee
,
S. W.
Koch
,
X.
Lu
,
S. R.
Johnson
,
D. A.
Beaton
,
T.
Tiedje
, and
O.
Rubel
,
Appl. Phys. Lett.
96
,
131115
(
2010
).
25.
X.
Liu
,
M.-E.
Pistol
,
L.
Samuelson
,
S.
Schwetlick
, and
W.
Seifert
,
Appl. Phys. Lett.
56
,
1451
(
1990
).
26.
Y.
Zhang
,
A.
Mascarenhas
,
J. F.
Geisz
,
H. P.
Xin
, and
C. W.
Tu
,
Phys. Rev. B
63
,
085205
(
2001
).
27.
G.
Ciatto
,
E. C.
Young
,
F.
Glas
,
J.
Chen
,
R.
Alonso Mori
, and
T.
Tiedje
,
Phys. Rev. B
78
,
035325
(
2008
).
28.
R. D.
Richards
,
F.
Bastiman
,
C. J.
Hunter
,
D. F.
Mendes
,
A. R.
Mohmad
,
J. S.
Roberts
, and
P. R.
David
,
J. Cryst. Growth
390
,
120
124
(
2014
).
29.
D. A.
Beaton
,
C.
Sanders
, and
K.
Alberi
,
J. Cryst. Growth
413
,
76
80
(
2015
).
30.
C.
Sanders
,
D. A.
Beaton
, and
K.
Alberi
,
Appl. Phys. Lett.
106
,
182105
(
2015
).
31.
D. A.
Beaton
,
A. J.
Ptak
,
K.
Alberi
, and
A.
Mascarenhas
,
J. Cryst. Growth
351
,
37
40
(
2012
).
32.
J.
Li
,
K.
Forghani
,
Y.
Guan
,
W.
Jiao
,
W.
Kong
,
K.
Collar
,
T.-H.
Kim
,
T. F.
Kuech
, and
A. S.
Brown
,
AIP Adv.
5
,
067103
(
2015
).
33.
S.
Mazzucato
,
R. J.
Potter
,
A.
Erol
,
N.
Balkan
,
P. R.
Chalker
,
T. B.
Joyce
,
T. J.
Bullough
,
X.
Marie
,
H.
Carrère
,
E.
Bedel
,
G.
Lacoste
,
A.
Arnoult
, and
C.
Fontaine
,
Phys. E
17
,
242
244
(
2003
).
34.
S. R.
Johnson
,
Y. G.
Sadofyev
,
D.
Ding
,
Y.
Cao
,
S. A.
Chaparro
,
K.
Franzreb
, and
Y. H.
Zhang
,
J. Vac. Sci. Technol., B
22
,
1436
1440
(
2004
).
35.
M.
Wu
,
E.
Luna
,
J.
Puustinen
,
M.
Guina
, and
A.
Trampert
,
Appl. Phys. Lett.
105
,
041602
(
2014
).
36.
P. M.
Mooney
,
K. P.
Watkins
,
Z.
Jiang
,
A. F.
Basile
,
R. B.
Lewis
,
V.
Bahrami-Yekta
,
M.
Masnadi-Shirazi
,
D. A.
Beaton
, and
T.
Tiedje
,
J. Appl. Phys.
113
,
133708
(
2013
).