Bi Incorporation and segregation in the MBE-grown GaAs-(Ga,Al)As-Ga(As,Bi) core-shell nanowires

Incorporation of Bi into GaAs-(Ga,Al)As-Ga(As,Bi) core-shell nanowires grown by molecular beam epitaxy is studied with transmission electron microscopy. Nanowires are grown on GaAs(111)B substrates with Au-droplet assisted mode. Bi-doped shells are grown at low temperature (300 {\deg}C) with a close to stoichiometric Ga/As flux ratio. At low Bi fluxes, the Ga(As,Bi) shells are smooth, with Bi completely incorporated into the shells. Higher Bi fluxes (Bi/As flux ratio ~ 4%) led to partial segregation of Bi as droplets on the nanowires sidewalls, preferentially located at the nanowire segments with wurtzite structure. We demonstrate that such Bi droplets on the sidewalls act as catalysts for the growth of branches perpendicular to the GaAs trunks. Due to the tunability between zinc-blende and wurtzite polytypes by changing the nanowire growth conditions, this effect enables fabrication of branched nanowire architectures with branches generated from selected (wurtzite) nanowire segments.

alloy NWs are interesting also in the context of topological materials. Although, Ga(As,Bi) planar layers grown by MBE have been investigated for over two decades, there is a very limited number of reports on Ga(As,Bi) in the NW geometry. The main reason is quite extreme MBE growth conditions of Ga(As,Bi) (in comparison with GaAs); i.e., low growth temperature (~ 200 -350 °C) and close-to stoichiometric V/III elements flux ratio to avoid Bi surface segregation and to induce its substantial incorporation at As sites in the GaAs host lattice 14,15 . These growth conditions deviate from the GaAs NWs growth requirements, where higher substrate temperatures (> 500 °C) and high As excess are indispensable 16 . This implies that Ga(As,Bi) NWs with significant Bi content can only be obtained as low-temperature shells grown on Bifree core NWs. This is similar to NWs implementing (Ga,Mn)As dilute ferromagnetic semiconductor with akin growth requirements, demanding even lower growth temperatures [~ 100 °C lower than Ga(As,Bi)] and the same stoichiometric III/V flux ratio 17,18,19,20 . Recently the mixed phase WZ-ZB GaAs NWs exposed ex-situ to Bi vapor were studied by scanning tunneling microscopy 21 but this study has little reference to our investigations of in-situ Bi incorporation.
In this paper we thoroughly investigate the radial distribution of Bi in GaAs NWs (not studied so far to our knowledge) and elucidate the emergence of side Bi droplets and branches on the wurtzite segments of GaAs NWs with Ga(As,Bi) shells.

Samples and Experimental Methods
The core-shell nanowires were grown in a dedicated III-V MBE system. First the GaAs NW trunks were crystallized on GaAs(111)B substrates by the Au-assisted growth mode. 5 Å thick gold film was deposited on epi-ready GaAs(111)B wafers in another MBE system, and transferred (in air) to the III-V one. The growth was monitored by reflection high energy electron diffraction (RHEED) system. The substrate temperature was controlled by the MBE substrate manipulator thermocouple, calibrated using GaAs(100) surface reconstruction transition temperatures 22 . For that, a piece of GaAs(100) wafer was placed in the vicinity of Au-coated GaAs(111)B dedicated to the NWs growth. Planar Ga(As,Bi) layer grown on GaAs(100) during the Ga(As,Bi) NW shells deposition also serves as references to compare Bi incorporation into ZB and WZ GaAs phase of planar layers and NW shells, respectively (see the Supplementary   Information). After preheating the substrates to 600 °C in the MBE growth chamber resulting in the thermal desorption of native oxide and formation of AuGa eutectic droplets at random surface sites of GaAs(111)B wafer, the substrate temperature was decreased to 540 °C, and GaAs NWs have been grown for 1-3 hours, depending on the sample. The Ga flux intensity (calibrated through a test growth on GaAs(100) substrate by RHEED intensity oscillations) corresponded to the planar growth rate of 0.2 m/h, which resulted in the axial NWs growth rate of about 1 m/h.
As the source of arsenic, the valved cracker cell has been used with cracking zone temperature of 950 °C i.e., As dimers were prevailing in the As flux. The As/Ga flux ratio during this growth stage was about 10. The growth of GaAs NW trunks was completed by closing Ga and As shutters. The latter was closed one minute after the first one. After closing the Ga shutter the substrate temperature was decreased for deposition of the NW shells. In the case of one sample (sample 3) prior to the Ga(As,Bi) growth, about 30 nm (Ga,Al)As shells have been grown at the substrate temperature of 400 °C, with As and Ga flux ratios the same as used previously for the axial GaAs NWs growth. For the deposition of Ga(As,Bi) shells the substrate temperature was further decreased to about 300 °C. Ga(As,Bi) shells have been deposited in close-tostoichiometric growth conditions i.e. with V/III flux intensity ratio close to 1.

6
We investigate three different types of GaAs-Ga(As,Bi) NW samples. The Ga(As,Bi) shells were grown in similar conditions for all the samples, i.e. at the same substrate temperature (300 °C) and As/Ga flux ratio (~ 1). The main difference between the growths was the Bi flux intensity during deposition of the low temperature (LT) Ga(As,Bi) shells. Bi flux was generated from a standard Knudsen effusion cell. To obtain different Bi concentrations the Bi cell temperature was set to 540 °C or 580 °C, corresponding to low (1%) or high (2-4%) Bi content in Ga(As,Bi) shells of sample 1, (2 and 3), respectively. In each case the Ga(As,Bi) shell was finished by the deposition of 4-7 nm thick LT GaAs.   Nanolab 600 FIB as it is described in Ref. 23 . or twin boundaries (TB) are visible. We infer that these ZB tips emerge due to the residual axial growth during Ga(As,Bi) shell deposition. The pure ZB phase at the upper NW section is consistent with the reported dependence of GaAs NW phase on the NW tip-droplet contact angle 24 . The authors of Ref. 24  As can be observed in Figure 2b, the droplet contact angles for the NW tip are in the low β range,

9
90° on the left and 100° on the right side of the NW-droplet interface (β 1 and β 2 , respectively), which according to Ref. 24 should promote ZB phase of GaAs NWs.   Fig. 2f is expressed in at.%, for Ga and As, whereas the Bi concentration is normalized to the As atoms.
From the Bi distribution and STEM-HAADF profile we can estimate the NW core diameter of ~ 40 nm and the Ga(As,Bi) shell thickness of ~ 30-40 nm. The maximum concentration of Bi is found at the NW side-wall surface (the amorphous "skin" of the NWsee Fig. S5 in the Supplementary Material), but the maximum concentration of Bi/As in the shell is at the level of ~1%. Interestingly in the planar ZB Ga(As,Bi) layer grown together with sample 1, the Bi content evaluated from the Ga(As,Bi) lattice parameter (see Fig. S1 in the supplementary material) amounts to 4.6%, which proves much effective incorporation of Bi into planar ZB GaAs(100) than to WZ GaAs (11)(12)(13)(14)(15)(16)(17)(18)(19)(20) NW sidewall planes.
In order to increase the Bi content in Ga(As,Bi) NW shells, sample 2 was grown with much higher Bi flux intensity, than that used for sample 1. SEM images of sample 2 (Figure 1c   nm. From now on, we will assign the name "NW neck" to the top ZB segment of the NW, which results from the residual axial growth during the low-temperature Ga(As,Bi) shell deposition (as discussed above). The shape differences of the NW necks in samples 1 and 2 can be explained as follows. In both cases, two growth modes of the NW neck occur simultaneously: (i) axial growth and (ii) radial growth. In sample 1, the catalytic droplet is small, without any Bi, since the Bi flux is low and all impinging Bi is incorporated into the NW shells (see Fig.1a  respectively in the NW tip -NW top droplet section shown in Fig. 3a. According to the in-situ TEM NW growth investigations reported in Ref. [24], the catalyzing droplet -NW tip contact angles higher than about 125 deg. promote the ZB GaAs NW phase.
(Ga,Al)As shells, with 30% Al, had been grown on the GaAs NW core before the Ga (As,Bi) shell growth, in the case of sample 3. The (Ga,Al)As shells were deposited at 400 °C. The lighter (Ga,Al)As layer contrast in ADF-STEM images (see Fig. 4b) allows to unequivocally identify the Ga(As,Bi) shells both in the plan-view and cross sections images of the NW. EDS compositional distribution across the WZ area of the NW shown in Fig. 4f   The Ga(As,Bi) shells were grown in the same conditions (growth temperature, Bi and Ga fluxes, As/Ga flux ratio) as sample 2, but for a slightly longer time (45 min for sample 3, versus 30 min for sample 2). The 50% longer Ga(As,Bi) shell growth time under Bi excess conditions leads to a more pronounced Bi segregation forming more droplets at the NW sidewalls. Figure 4 summarizes the compositional analysis carried out for two representative NWs from sample 3.
Two different upper GaAs-(Ga,Al)As-Ga(As,Bi) core-double shell NWs are displayed. The first one shown in Fig 4 a, Similarly to the case of sample 2 the residual axial growth during crystallization of Ga(As,Bi) shell produces distinct necks (visible also in Fig 1e, and 1f). This neck part can easily be broken during the mechanical transfer of NWs to the TEM grid. A complete NW is shown in Fig. 4 d and Fig. S7 in the Supporting Information. The EDS maps (Fig. 4c) and profiles (Fig. 4e,f) reveal the inner GaAs core and (Ga,Al)As shell. This data let us conclude that during the growth of (Ga,Al)As shell also axial (Ga,Al)As growth was continued over the length of ~200 nm.
After the temperature drop necessary to grow Ga(As,Bi) shell, the gold droplet was replaced by the Bi one, and axial growth continued in both axial and slightly off-axis (random) directions, as can be seen in Fig.1e,f.
The accumulation of liquid Bi at the sidewall surface resulted in the formation of additional Birich droplets, which started to catalyze the secondary branches reproducing the trunk structuretwinning ZB parts and SF in WZ parts of the core, as shown in Fig.S6 in the supporting information. Replacing the Au droplet by the Bi one at the NW top can be explained as follows.
The temperature decrease to 300 °C for Ga(As,Bi) shell deposition caused crystallization of Au at the NW top, but after delivery of Bi in large amount, the external part of hitherto crystalized Au droplet became liquid. In the Au-Bi system, there are two eutectic points 371 °C and 241 °C, depending on the Bi concentration 6 . We infer that the 241 °C BiAu eutectic was formed and the surface of the gold nanocrystal at the NW top became liquid, which allowed the droplet to move and float down as shown in the EDS map in Fig. 4d. High magnification ADF-STEM image of the Bi-droplet catalyzed branch, corresponding to the region marked by the red square in Fig.4b, is shown in Fig.S6 in the Supporting Information.  5a). As can be seen in Fig. 5a the hexagonal shape of the GaAs core is reproduced in the successive shells. However, the last LT-GaAs outer shell does not develop sharp corners. The measured thicknesses of the (Ga,Al)As shells on all six NW sidewalls are equal to 18-20 nm.  The dark lines running from the GaAs core corners visible in Fig. 5a have already been reported 27,28,29 for ZB NWs with (Ga,Al)As shells where high Al concentration, i.e. ~50% was detected in these regions and 30% Al elsewhere. However the Bi composition profile in the NWs with Ga(As,Bi) shells was not investigated so far, since no GaAs-Ga(As,Bi) NW cross-sections were studied before, to our knowledge. At the (Ga,Al)As-Ga(As,Bi) shell interface, the very thin adjacent Ga(As,Bi) shell region with higher Bi concentration appears as brighter lines in Fig 5a (red stripes in Fig 5b and 5c). Moreover, additional, very thin bright lines are visible at the very edge of the NW (orange in Fig. 5c). It should be noted that the Ga(As,Bi) shell thicknesses are slightly different at each NW sidewall especially for the sidewalls in the closest neighborhood to other NWs, due to the shadowing effect during the MBE growth. Figure   (c) 23 about 20 nm thick cross-section of Bi-rich shell; yellow color represents Bi-rich atomic columns averaged over 20-nm thick specimen, the inset shows zoomed area enclosed by the white rectangle. Figure 6 shows a cross-sectional TEM image along the [0001] direction recorded in the WZ section of a NW from sample 3. In this case, the WZ phase can be easily identified due to fact that the only contrast maxima corresponding to …ACAC… stacking are visible, contrary to ZB part where maxima corresponding to ..ABC… stacking are visible (compare Fig 5). In this case, the hexagonal shape of the core is also preserved. More detailed discussion concerning distinctiveness of ZB and WZ phases in the TEM cross-section specimen is included in the supplementary material. The 5-6 nm thick Bi-rich shell can be unequivocally identified in Fig.6.
This shell appears homogenous for all three NW facets visible in the image. We estimate that the specimen thickness is about 90 nm, so the fine Bi fluctuations cannot be detected. The segregation of Al is clearly visible inside the (Ga,Al)As shell. We also detect Al segregation at the edges of the hexagon, never reported before for WZ NWs. However, the segregation shown in the inset to Fig. 6a, 6b looks differently than that occurring in ZB (Ga,Al)As NW sections, reported earlier 27 . In our case, the high Al concentration regions have zigzag shapes 27 . Slightly similar zig-zag corner-line shapes were observed for P-rich regions of ZB Ga(As,P)/GaAs coaxial NWs 30 . The structural details of this Al-enhanced region are shown in the filtered image (inset to Fig. 6) where the zigzag shape, Al-rich path (dashed line) does not follow solely radial

Concusions
In conclusion, we have investigated Bi incorporation into predominantly wurtzite GaAs-

Supporting Information:
XRD spectrum, SEM and TEM images of a planar Ga(As,Bi) layer grown together with GaAs-Ga(As,Bi) core shell nanowires (Sample 1); additional TEM images of NWs and NW cross-sections including EDS composition maps (Samples 2 and 3).

Characterization by X-ray diffraction and SEM
The planar layer grown on the Au-free GaAs(001) substrate together with sample 1 (GaAs -Ga(As,Bi) core-shell NWs) was measured by high resolution X-ray diffraction (XRD) X'Pert MRD with Cu tube (CuK α1 radiation (λ = 1.5406 Å)), equipped with: X-ray mirror, monochromator (asymmetrically cut Ge 4x(220)) and with two proportional detectors, one of the detectors is preceded by the analyzer 3xGe(220). Figure S1 shows results of 2/ measurement around 004 Bragg reflection of GaAs (001) substrate. The perpendicular lattice parameter obtained from the angular position of the peak corresponding to Ga(As,Bi) layer (grown simultaneously with Ga(As,Bi) shells of GaAs NWs on Au-coated GaAs(111)B) equals to: a  = 5.7126 Å. The Ga(As,Bi) layer is fully strained to the GaAs substrate, hence we can calculate the relaxed lattice parameter (assuming the Ga(As,Bi) elastic constant values C 11 and C 12 to be the same as for GaAs) which amounts to: a relaxed = 5.68473 Å.
Taking the hypothetical lattice parameter of binary GaBi equal to 6.33 Å and assuming Vegards law we obtain then 4.6% Bi content in the planar Ga(Bi,As) layer grown together with sample 1.
The distance between pendelösung fringes, confirmed by the XRD simulations yields the thickness of Ga(As,Bi) planar layer equal to 137 nm, which agrees quite well with the value assumed from the GaAs growth rate calibrations based on RHEED oscillations.
Apparently, in the planar zinc-blende (ZB) counterpart of sample 1 the incorporation of Bi is much higher than that in the WZ Ga(As,Bi) NW shell. EDS signal for Bi is low and from the average spectra we obtain values at the level of 1at %, however with large error (about the same order of magnitude, i.e. ±1%). Nevertheless, the Bi-M line in the X-ray spectrum is clearly above the noise level. Analysis of different spectra gives similar results, around 1at%.
The surface of the reference planar layer grown together with sample 1 is a bit "milky" (as observed by naked eye), suggesting the presence of some 3D objects (Bi droplets) but it still shows distinct 2D RHEED patterns, hence the concentration of the droplets is not huge.
However, for samples 2 and 3, grown with about 4 times higher Bi flux, corresponding to the temperature of Bi source T Bi = 580 °C (40 °C higher than that used during the growth of sample 1), RHEED images for reference planar layers disappeared completely upon the growth of Ga(As,Bi). This indicated the presence of amorphous surface features yielding diffused uniform RHEED background. This can be attributed to liquid Bi droplets (the growth temperature is equal 300 °C which is slightly higher than the Bi melting temperature 270 °C)). These droplets (solidified at room temperature of SEM measurements) are shown in Fig. S2, both in plan view and in cross-sectional sample orientation.

Characterization by cross-sectional TEM
More insight into the planar Ga(As,Bi) layer and individual surface Bi droplets is provided by the cross-sectional TEM image of a planar reference to sample 2. Interestingly the surface morphology of planar Ga(As,Bi) in-between segregated Bi droplets is consistent with the predictions based on the Monte-Carlo simulations of Bi droplet segregation during Ga(As,Bi) MBE growth. 1    The STEM images and EDS composition maps of the upper part of the NW collected form sample 3, with unbroken top part are shown in Figure S7. The curved NW part below the very top Bi droplet is axially grown during the Ga(As,Bi) shell deposition and has purely ZB structure, Figure S8 shows images corresponding to the thin cross-sections of a ZB part of NW selected form sample 3. The thickness of the analyzed NW cross-section is measured using position averaged convergent beam electron diffraction (PACBED) technique. This pattern is obtained using the 9.5 mrad converged electron beam scanned across the core area of the NW crosssection. The comparison of experimental and simulated PACBED patterns is very sensitive to STEM HAADF Al Ga Bi Au As local thickness determination. 2 In our case the best match is measured as l 2 -norm metric defined

Supplementary information on nanowires
Where ̃ and ̃ denote experimental and simulated PACBED patterns, and t denotes the sample thickness input to the simulation. The minimum difference is obtained for t 75 nm. With this specimen thickness only relatively larger composition fluctuations can be determined.