This is the accepted version of the article: Francaviglia L., Tütüncüoglu G., Martí-Sánchez S., Di Russo E., Escobar Steinvall S., Segura Ruiz J., Potts H., Friedl M., Rigutti L., Arbiol J., Fontcuberta I Morral A.. Segregation scheme of indium in AlGaInAs nanowire shells. Physical Review Materials, (2019). 3. 023001: - . 10.1103/PhysRevMaterials.3.023001. Available at: https://dx.doi.org/10.1103/PhysRevMaterials.3.023001 1 Segregation scheme of Indium in AlGaInAs nanowire shells 1 1 2 3 1 2 Luca Francaviglia, Gözde Tütüncüoglu, Sara Martí-Sánchez, Enrico Di Russo, Simon Escobar Steinvall, Jaime 4 1 1 3 2, 5 1, 6 , ∗ 3 Segura Ruiz, Heidi Potts, Martin Friedl, Lorenzo Rigutti, Jordi Arbiol, and Anna Fontcuberta i Morral 1 4 Laboratoire des Matériaux Semiconducteurs, Institut des Matériaux, 5 Ecole Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland 2 6 Catalan Institute of Nanoscience and Nanotechnology (ICN2), 7 CSIC and BIST, Campus UAB, Bellaterra, 08193 Barcelona, Catalonia, Spain 3 8 Groupe de Physique des Matériaux, Université de Rouen, Saint Etienne du Rouvray, 76801, France 4 9 European Synchrotron Radiation Facility, Grenoble, 38043, France 5 10 ICREA, Pg. Lluís Companys 23, 08010 Barcelona, Catalonia, Spain 6 11 Institute of Physics, École Polytechnique Fédérale de Lausanne, CH-1015, Switzerland 12 (Dated: November 26, 2018) Quaternary alloys enable the independent optimization of dierent semiconductor properties, such as the separate tuning of the bandgap and the lattice constant. Nanowire core-shell structures should allow a larger range of compositional tuning as strain can be accommodated in a more eective manner than in thin lms. Still, the faceted structure of the nanowire may lead to local segregation eects. Here, we explore the incorporation of In in AlGaAs shells up to 25 %. In particular, we show the eect of In incorporation on the energy shift of the AlGaInAs single-photon emitters present in the shell. We observe a redshift up to 300 meV as a function of the group-III site fraction of In. We correlate the shift with segregation at the nanoscale. We nd evidence of the segregation of the group-III elements at dierent positions in the nanowire, not observed before. We propose a model that takes into account the strain distribution in the nanowire shell and the adatom diusion on the nanowire facets to explain the observations. This work provides novel insights on the segregation phenomena necessary to engineer the composition of multidinary alloys. 13 Introduction Quaternary alloys enable to indepen- 42 particular, we nd consistent evidence of the In segrega- 14 dently tune the semiconductor lattice constant and 43 tion in novel wedge-shaped In-rich features. We correlate 15 bandgap by careful composition engineering. However, 44 these observations with the 3-fold polarity of the NW 16 the controlled deposition of quaternary alloys with ran- 45 cross section and the migration of the adatoms on the 17 domly distributed composition can be challenging. On 46 NW sidewalls. In addition to the role of the crystal direc- 18 one hand, miscibility gaps and diusion may segregate 47 tions of the NW facets [1820], we consider strain to cause 19 the dierent species that compose the alloy, and limit the 48 the peculiar In segregation; the presence and distribu- 20 possible compositions [1, 2]. On the other hand, strain 49 tion of strain are further analyzed by nite-element sim- 21 may build up when quaternary semiconductors epitaxi- 50 ulations as well as Raman and photoluminescence (PL) 22 ally grow on substrates of dierent lattice constant. To 51 spectroscopy. 23 avoid plastic relaxation [3, 4], the theoretical range of 52 The material system: core-shell NWs The GaAs- 24 available compositions is signicantly narrowed in real 53 AlGa(In)As core-shell NWs are grown in a high-mobility 25 applications. Due to their reduced diameter, semiconduc- 54 molecular-beam-epitaxy system (MBE, DCA P600). We 26 tor nanowires (NWs) provide a suitable growth platform 55 use Si(111) substrates and self-catalyzed growth [23, 24] 27 to minimize plastic relaxation, [510] permitting lattice- ◦ 56 at a substrate temperature of 640 C. When the NWs are 28 mismatched material combinations not achievable by pla- 57 about 10µm long, we stop the axial growth by interrupt- 29 nar schemes [1114]. 58 ing the Ga supply and consuming the catalyst. We then 30 In this work, we explore the range in which In can be ◦ 59 lower the substrate temperature to 460 C to promote ra- 31 incorporated in AlGaAs shells of core-shell GaAs-AlGaAs 60 dial growth of an AlGa(In)As shell on the NW sidewalls 32 NWs. The goal is to understand how much the emission 61 [25, 26]. We start with an Al 33 Ga67 As shell and then 33 energy of single-photon emitters, spontaneously forming 62 incorporate In without modifying the Ga or Al rates. 34 in the shell, can be redshifted [1517]. In particular, we 63 For this, we varied the In partial pressure from 4.5 X 35 want to dissociate the bandgap engineering from addi- −9 −7 64 10 Torr to 2.5 X 10 Torr, while the Al and Ga pres- 36 tional segregation eects that may occur due to the pres- 65 sures were kept constant. Unless stated, the samples are 37 ence of In. Consequently, we characterize the optical 66 identied by the group-III site fraction of In measured 38 emission and correlate it with the incorporation of In us- 67 through various techniques and expressed as percentage 39 ing several techniques at dierent length scales. We nd 68 in the notation Al x Gay In1−x−y As (Al, Ga, and In sum 40 that group-III elements in the shell segregate to dierent 69 to 100%) : 0%, 1%, 2%, 3%, 4%, 15%, and 25%. An 41 positions and generate regions of dierent bandgaps. In 70 outer GaAs shell of 5 nm prevents the oxidation of the 71 inner AlGa(In)As. 72 Previous reports show that the composition of an ∗ anna.fontcuberta-morral@ep.ch 73 AlGaAs shell deposited around a GaAs NW exhibits 2 a) Core-shell QDs GaAs 74 nanoscale uctutations [15, 16, 2730]. This phenomenon 75 depends on the dierent surface mobility of Ga and Al NW 76 during growth and the presence of surface-potential gra- 77 dients (µ) on the NW sidewalls. The formation of Al-rich 78 planes at the ridge between two NW facets was reported 79 [19]. Furthermore, Ga-rich clusters can form quantum- 80 dots (QDs) that behave as single-photon emitters in the AlGa(In)As 81 NW shell [15, 16, 29]. A sketch of a core-shell GaAs- 82 AlGaAs NW and its cross section is presented in g- Al-rich plane 83 ure 1a, illustrating the distribution of Al-rich planes and 84 QDs made from Ga-rich nanoclusters. b) 10x laser 85 Light-emission properties First we address the op- 250x10-9 12K power 86 tical functionality of the core-shell NWs and embedded Torr 87 QDs. We used a single-frequency optically pumped semi- 88 conductor laser with a wavelength of 532 nm and power 2 86x10-9 89 of 100 W/cm focused in a spot of less than 1 µm in di- Torr 90 ameter to measure NW samples at 12 K using a helium In 91 cryostat. The PL signal is collected into a spectrome- d 92 ter and dispersed by a 300 l/mm grating onto a Peltier- 25x10-9 iu 93 cooled CCD. The QDs in the AlGaAs shell emit bright m Torr 94 and narrow PL lines (linewidth below 100 µeV) [15] be- in c 95 tween 1.7 eV and 1.9 eV [27]. The green spectrum at the reas 96 bottom of gure 1b illustrates the PL emission of these 0 Torr 97 structures; the sharp peaks are attributed to the pres- e 98 ence of QD single-photon emitters [15, 31]. Figure 1b 99 also contains PL spectra of the shell-QD emission for 100 increasing In fraction, revealing several sharp peaks at 101 dierent energies. The emission-energy range redshifts c) 102 with increasing In fraction in the shell. One can also 103 qualitatively determine that the QD emission linewidth 104 broadens in samples of higher In fraction. The sample g2(τ) 105 with the highest In fraction showed a decreased PL in- EgShell 1.0 106 tensity, compensated for by increasing the laser power 107 (see gure 1b). 0.5 108 For a statistical analysis on large ensembles of NWs 109 and QDs, gure 1c shows the median emission energy X-X 4K 110 of the QDs as a function of the measured In pressure 0.0 111 in the MBE chamber (bottom axis) or as a function of -4 -2 0 2 4 Delay τ (ns) 112 the In fraction (top axis) measured in the shell by X-ray 113 uorescence (XRF) combined with atom probe tomogra- 114 phy (APT) for 1 % to 4% of In and scanning transmis- 115 sion electron microscopy X-ray energy-dispersive spec- 116 troscopy (STEM EDS) for 15 % and 25% In (supplemen- GaAs Eg 117 tary information). We measured 25 NWs from each sam- 118 ple. The QD emission lines were identied by an auto- 119 matic routine [27]. The dash-dot line indicates the ex- 120 pected shell bandgap as a function of shell composition 121 [21]. The dashed line corresponds to the GaAs bandgap. FIG. 1: (a) Sketch of a core-shell NW and shell QDs. (b) 122 Both lines are corrected for the simulated strain [22] that Micro-PL spectra of shell QDs grown at increasing In pressures. 123 arises from the core-shell lattice mismatch, as further dis- (c) Ensemble medians of the PL QD emission energy vs the In 124 cussed in the manuscript. MBE pressure (bottom abscissa) or shell fraction (top abscissa) given by APT, XRF, STEM EDS and growth-rate calibrations. 125 The inset in gure 1c shows a Hanbury-Brown-Twiss 1st and 3rd quartile as error bars. The dash and the dash-dot lines 126 autocorrelation measurement (g 2 (τ )) of the exciton line are the strained core and shell bandgaps, respectively [21, 22]. 127 of a QD from a sample with 3 % In. The power-dependent Inset: single-photon g2 of a QD exciton (3% In). 128 PL showing the exciton nature of the emission line is 129 reported in the supporting information. The sample is 130 measured at 4.2 K in an Attodry 700 closed-cycle cryostat 131 and is excited in continuous wave by a HeNe laser at 3 132 632.8 nm through an objective with NA = 0.81, which 191 fraction decreases to about 1.6 %. In gure 2b and c we 133 also collects the QD signal from the cryostat. A 1200 192 can also observe that the shell in proximity of the Al-rich 134 l/mm grating is used to select the QD line of interest 193 planes is slightly depleted in Al and enriched in Ga. 135 that is sent to a 50:50 beam splitter. The two paths 194 To study the segregation of Al, Ga, and In in more 136 of the beam splitter are coupled to single-mode optical 195 extended regions, we studied the NW cross-sections by 137 bers that send the signal to two single-photon avalanche 196 high angular annular dark eld (HAADF) STEM and 138 diodes, one for each path. The dip in the autocorrelation 197 EELS in an aberration-corrected (AC) TEM microscope 139 function in the inset of gure 1c is below 0.5, which is the 198 (FEI Titan) operated at 300 keV. We used the sample 140 signature of the single-photon emission. Several factors 199 with 15% In. We prepared NW cross sections to directly 141 increase the count at zero delay, including background 200 map the shell composition: the NWs were spread on a 142 counts from QD lines spectrally close to the chosen one. 201 Si substrate and the cross sections were FIB-cut per- 143 The eect of the shell-composition engineering on the 202 pendicularly to the NW growth direction (<111>). The 144 QD emission energy is signicant. From gure 1c, the 203 cross-section HAADF micrographs are acquired along the 145 median QD emission redshift is visible as a function of 204 <111> zone axis, with the NW sidewalls corresponding 146 the increasing In incorporation, up to about 300 meV. 205 to the {110} family (gure S4 in the supporting informa- 147 In a random-distribution alloy, the addition of In to Al- 206 tion). Figure 2d reports the HAADF micrograph of one 148 GaAs lowers the shell bandgap and the QD emission en- 207 NW cross section. The HAADF contrast depends on the 149 ergy. Since these QDs form as lower-bandgap nanoscale 208 sample thickness and composition: the higher the atomic 150 clusters in the AlGa(In)As matrix, their emission is ex- 209 number (Z) of the species in the sample, the brighter 151 pected to be redshifted with respect to the shell-matrix 210 the HAADF signal. Through high-precision FIB cut, the 152 bandgap (dash-dotted line in gure 1c). While this is true 211 NW cross sections have negligible thickness variations: 153 in the samples with a low In fraction (up to 4 %), the me- 212 the contrast in gure 2d depends on the average Z at 154 dian QD-emission shifts less signicantly from the shell 213 dierent positions across the sample. In this gure, the 155 bandgap in the samples with 15 % and 25% of In. Con- 214 hexagonal shape of the NW cross section is clearly visible 156 comitantly, the overall QD emission-energy range broad- 215 and the core is distinguishable from the shell. The dark 157 ens in the shells with high In fractions (gure S1 in the 216 stripes that form in the shell along the ridges between two 158 supplementary information and error bars in gure 1c). 217 NW facets ({112} crystalline directions) are due to local 159 Together with the linewidth broadening and brightness 218 enrichment with a light element, such as Al, as already 160 quenching previously commented (gure 1b), these obser- 219 observed by APT and in agreement with the literature 161 vations suggest an increasing alloy disorder as a function 220 [15, 1820]. It is possible to observe that the thickness of 162 of the In incorporation. 221 the Al-rich planes follows a 3-fold symmetry as observed 164 163 Compositional analysis at the nanoscale So far, 222 by Zheng et al. [18]. We obtain a thickness of 3 nm and 165 we have reported on the average In incorporation (g- 223 1.7 nm for the thick and thin planes, respectively (more 166 ure 1c). In NWs the composition can vary signicantly 224 details in gure S4 in the supporting information). Sim- 167 at the nanometer scale due to their faceted nature, as 225 ilarly, the HAADF contrast shows that the thick planes 168 dierent facets exhibit dierent sticking coecients [30]. 226 are richer in Al than the thin ones. Comparing our obser- 169 Here we provide measurements with spatial resolution 227 vations with the literature [18], we assign the thin planes 170 down to the atomic scale: APT and STEM-based elec- 228 to the A polarity and the thick ones to the B polarity. 171 tron energy loss spectroscopy (EELS) and EDS. 229 Accordingly, we labelled the two polarities in gure 2d as 172 We performed laser-assisted APT measurements on 230 A (orange) and B (blue). 173 the sample with 2 % In. The specimen is cooled to 80 K 231 With the exception of the Al-rich planes, the HAADF 174 and irradiated with UV laser light (343 nm wavelength) 232 micrograph in gure 2d shows little compositional con- 175 in 2-nJ pulses; the detection rate is 0.0025 events/pulse. 233 trast in the shell. By carefully inspecting gure 2d 176 The evaporated NW volume is a cylinder with diameter 234 and gure S4 in the supporting information, a slightly 177 of 64 nm and length of 90 nm; in gures 2a - c, it is shown 235 brighter contrast is visible in proximity of the Al-rich 178 as a 2D projection on a plane perpendicular to the NW 236 planes: it corresponds to a local increase in the fraction of 179 growth axis. Figures 2a - c show respectively the In, Al 237 elements with higher Z, such as Ga and In. To clarify this 180 and Ga fractions. It is possible to distinguish the GaAs 238 observation, in gures 2e - h we report the EELS maps of 181 core by the absence of In and Al. In the shell, radial seg- 239 the upper half of the NW cross section shown in gure 2d: 182 regation of Al along the ridges between two facets of the 240 the selected region, indicated by the dashed rectangle in 183 hexagonal NW core is visible as the three Al-rich stripes 241 gure 2d, includes an A-polar and two B-polar ridges as 184 in the reconstructed NW volume, in agreement with pre- 242 well as part of the {110} NW sidewalls. The Ga, In, Al, 185 vious works [15, 16, 18, 19, 29, 30]. In addition to the 243 and As distributions are shown in gure 2e, f, g, and h 186 Al-rich planes, APT reveals that the shell distribution 244 respectively. Principal component analysis (PCA) was 187 of the other group-III atoms is not perfectly random. A 245 used to enhance the signal-to-noise ratio in these maps 188 slightly higher In fraction (up to 3 %) is visible in prox- 246 [32, 33]. The NW core is clearly distinguished from the 189 imity of two of the Al rich planes indicated as 1 and 3 247 shell by the absence of Al and In and a thin Ga-rich 190 in gure 2a. Farther from these positions, the average In 248 layer, corresponding to the GaAs capping, surrounds the 4 a) Indium b) Aluminum c) Gallium 3% 70% 70% 2 1 50% 2% 45% 30% 3 1% 10% 20% d) A e) Ga f) In B B g) Al h) As A A B i) A Composition (atom %) 60 112(B) Ga 50 (112)A B (112)B 110 40 30 (110) Al 20 In A 10 112(A) 0 20 40 60 80 100 120 140 160 180 200 220 240 B Position (nm) FIG. 2: (a), (b), and (c) APT of In, Al, and Ga fractions (2 % In). (d) HAADF STEM micrograph of a NW cross section (15 % In) with A and B polarity indicated. (e-h) EELS Ga, In, Al, and As maps of the dashed rectangle in (d). (i) Left: sketch of shell segregation. Right: shell Ga (green), Al (blue), and In (red) atomic group-III percentages vs position along the circumference in the sketch. Scale bars: (a-c) and (e-h) 10 nm, (d) 20 nm. 249 shell (gure 2e). As is randomly distributed through- 259 In agreement with the localized In enrichment shown 250 out (gure 2f ), as opposed to the distributions of the 260 by APT (gure 2a), STEM EELS shows in gure 2f 251 group-III elements: the presence of Al-rich planes along 261 In-rich features surrounding the Al-rich planes at the 252 the hexagon ridges is conrmed (gure 2g), together with 262 <112>B directions: interestingly, moving from the core 253 the Ga and In depletion at the same positions (gure 2e 263 to the outer shell, the In-rich regions become broader 254 and f ). Reports on GaAs-InGaAs and GaAs-InAlAs core- 264 with a wedge-like shape on each side of the Al-rich 255 shell NWs [34] also show Ga and In depletion along the 265 {112}B planes. Ga has a more random distribution in 256 <112> direction. The brightness and thickness of the 266 the shell, although a slight increase is visible around the 257 Al-rich planes agree with the 3-fold polarity-driven seg- 267 A-polar {112} nanofacet in gure 2e. This Ga-rich fea- 258 regation previously discussed. 268 ture is not as pronounced as the In-rich ones surrounding 5 269 the {112}B planes; it also has a relatively constant thick- 328 the corner. In smaller proportions, the In distribution is 270 ness around the {112}A planes, while the In-rich features 329 opposed to the Al trend: for the same positions, In rst 271 develop an unusual wedge-shaped prole. One may also 330 increases (from 14 % to 19%) and then decreases to 12 % 272 observe that, in proximity of the {112} Al-rich planes 331 at the corner. Ga is almost constant at 61 % and 60% and 273 of both polarities, the Al fraction slightly decreases (g- 332 then decreases to 28 % at the corner. Very similar trends 274 ure 2g). Although the contrast is not sharp, this would 333 in Al/Ga/In fractions are reproduced at the six corners 275 agree with the local increase in the Ga and In fractions 334 of the hexagonal NW cross section and are modulated 276 at the same positions. EELS maps on whole NW cross 335 according to a 3-fold symmetry: the compositional vari- 277 sections conrm the 3-fold symmetry of the shell seg- 336 ations are consistently more pronounced in proximity of 278 regation and the In enrichment only around the three 337 the <112>B directions. 279 <112>B directions (see the supporting information). On 338 Strain analysis and growth model We turn now 280 the left of gure 2i, a scheme summarizes the main com- 339 to the understanding of the element distribution in the 281 positional segregations observed in the NW cross sections 340 shell by taking strain into account. Figure 3a shows the 282 (the scheme is aligned with gures 2d - h): the red-shaded 341 PL spectra of the GaAs core at 12K for core-shell NWs of 283 areas indicate the wedge-shaped In-rich segregation and 342 increasing In fraction. The band-edge GaAs PL redshifts 284 blue stripes indicate the alternatively thicker and thinner 343 from 1.51 eV down to 1.37 eV and 1.29 eV for shells con- 285 Al-rich planes. The polarity of the <112> nanofacets is 344 taining 15% and 25% In, which we attribute to the tensile 286 labelled. 345 strain imposed by the shell. Tensile strain in the core is 287 We acquired STEM-EDS maps of NW cross sections 346 also conrmed by Raman spectroscopy (supplementary 288 from the same samples as the one used for EELS. We use 347 information). 289 a FEI Tecnai Osiris TEM operated in STEM mode at 348 To gain insight in the strain distribution imposed by 290 200 kV with a probe current exceeding 1 nA. The X-ray 349 the lattice-mismatched core-shell NW, we simulated the 291 signal is collected by four silicon drift detectors under 350 strain given the lattice mismatch, dimensions and geome- 292 a solid angle of 0.9 srad. The NW cross sections are 351 try of the structure. We use the methodology of Boxberg 293 prepared by embedding the as-grown sample into epoxy. 352 et al. [36], adapted to hexagonal GaAs-AlGaInAs NWs 294 After hardening, the epoxy with embedded NWs is de- 353 at low temperature [21] in the software Nextnano [37, 38]. 295 tached from the growth substrate and is mounted into 354 The strain magnitude and distributions in the core and 296 an ultramicrotome and cut into 80-nm-thick slices trans- 355 shell are claculated by minimizing the elastic energy due 297 ferred on a TEM grid. The EDS maps (gure S3 in the 356 to the lattice mismatch between the two. Figure 3b maps 298 supplementary information) for the Ga, Al, and In distri- 357 of the hydrostatic strain  in NWs with 25 % In in hydro 299 butions in the core-shell cross sections conrm the pres- 358 the shell.  hydro is dominated by the principal component 300 ence of the same features observed in the EELS maps. 359 along the NW longitudinal axis, as reported elsewhere 301 Our observations are particularly robust: three indepen- 360 [39]. Shear strain components are one order of magnitude 302 dent techniques (APT, EELS and X-ray EDS) conrm 361 smaller than  hydro , in agreement with previous reports 303 the formation of novel 3-fold wedge-shaped In-rich segre- 362 [39]. As expected from the core-shell lattice mismatch, 304 gates in NWs with dierent average In fractions; together 363  is on average negative in the shell (compressive) hydro 305 with the clear symmetry of these features, the agreement 364 and positive in the core (tensile). We nd about 2.2 % 306 of the three techniques excludes artifacts due to sample 365 of tensile strain (position A in gure 3b), softly modu- 307 preparation. 366 lated following a 6-fold symmetry. The shell is subject to 308 We now turn to a quantitative analysis of the distribu- 367 a maximum compression of -0.35 % in the middle of the 309 tion of the dierent species in the NW cross section: the 368 NW facet (position C of gure 3b). The shell strain in- 310 EDS-based quantication is more accurate than the one 369 tensity is modulated by a 6-fold symmetry in a more pro- 311 based on EELS [35]. Figure 2i shows on the right a plot of 370 nounced way. The outer corners are relaxed (minimum 312 the Ga (green curve), Al (blue curve), and In (red curve) 371 strain of 0% at position B in gure 3b). Importantly, 313 atomic percentage extracted from a STEM EDS map as 372 the regions of smaller strain in the shell expand around 314 a function of the position along the dashed circumference 373 the NW sidewall ridges, becoming larger the farther they 315 shown in the sketch on the left (i.e. a circular linescan 374 are from the core. The experimental (PL) [40] and simu- 316 of 240nm in length). The data points at 0 nm and 240 375 lated strain values are plotted in gure 3c: the simulated 317 nm correspond to the same position on the circumference 376 compressive strain in the shell (position C) and the the 318 (arrow in the sketch). The shaded areas are colored in 377 experimental and simulated tensile strain in the core in- 319 orange and blue with the same coding as in gures 2g to 378 crease linearly with the In fraction. The theoretical and 320 distinguish the A- and B-polarity of the (112) directions 379 experimental values agree within 15 %. 321 respectively. The data points acquired in the middle of 380 A growth model that explains the non-random dis- 322 the at 323 {110} facets have a white background. 381 tribution of In and Al in the shell is presented. The 324 We focus on the most prominent uctuation around 382 expected behavior of Ga and In adatoms at the NW 325 the <112>B direction at position 100 nm. Moving from 383 facets is sketched in gure 3d. We start by addressing 326 the {110} plane to the {112}B nanofacet, the Al fraction 384 the Al distribution. Al tends to incorporate more ef- 327 rst decreases (from 25 % to 20%), then rises to 60 % at 385 ciently at the vertexes of the hexagonal cross section 6 ω a) b) Core Shell In In In hydro 25% 15% 0% 2.22% 0% normalized PL 80 150 GaAs-core meV meV C A Core B PL 1.20 1.30 1.40 1.50 Emission energy (eV) 50 nm 2.20% -0.35% c) d) In, Ga C In 2 A B A In, Ga (%) Ga hydro 1 Experimental ω Simulations 110 B 0 112(A) C 112(B) 0.00 0.05 0.10 0.15 0.20 0.25 Indium fraction FIG. 3: (a) GaAs PL of NWs with In shell fraction of 0 %, 15%, and 25%. (b) Hydrostatic strain simulation in core-shell NWs (25 % In) as in ref. [36]. (e) Simulated and average experimental (1 σ error bar) hydrostatic strain vs In fraction. (f) Sketch of the Ga and In diusion on the NW sidewalls. 386 ({112} nanofacets) [15, 18, 19]. This is the consequence 408 low miscibility of In with Al-containing alloys [43]. At 387 of a larger sticking coecient on those facets [41, 42]. For 409 the {112} nanofacets In competes with Al for the avail- 388 a similar reason, it has been shown that the incorpora- 410 able binding sites. This is particularly relevant for the 389 tion is higher on B- than A-polar facets [18]. This results 411 (112)B-polar directions because Al tends to accumulate 390 in a three-fold symmetry of the Al distribution. 412 more in these positions. 391 We now turn to the incorporation of In. According 413 Conclusion In conclusion, we have demonstrated a 392 to our simulations, there is a compressive strain in the 414 signicant (300 meV) redshift of the emission energy of 393 middle of the facets (position C in the drawings). Since 415 single-photon emitters embedded in the shell of core- 394 AlGaInAs exhibits a larger lattice constant than AlGaAs, 416 shell GaAs-AlGaInAs NWs. The redshifting mechanism 395 we expect In to incorporate favorably in the relaxed cor- 417 is based on the In incorporation in the AlGaAs shell alloy 396 ners. In adatoms exhibit high mobilities, allowing dif- 418 to form a quaternary semiconductor of lower bandgap. 397 fusion to occur at the scale of the NW facets. Strain 419 The spatial distribution of dierent species in the shell is 398 relaxation should support an In ux towards the more 420 determined by several high-resolution techniques of com- 399 relaxed {112} corner nanofacets, while the {110} facets 421 positional analysis. In addition to the well-known Al and 400 exhibit a signicantly lower In fraction in the centre (ar- 422 Ga segregation, we nd evidence of novel wedge-shaped 401 rows leaving this position in gure 3d). We also observe 423 In-rich segregation. We explain the distribution of the 402 that the In segregation seems to be enhanced by polarity. 424 dierent species in the quaternary alloy with their dif- 403 Regions with a markedly higher In fraction coincide with 425 fusion on the NW sidewalls driven by crystal-phase, po- 404 the {112}B vertexes. We note that the polarity-driven in- 426 larity and strain. Finite-element simulations provide in- 405 corporation of In has not been demonstrated in the past. 427 sight in the role of strain to drive the segregation of In. 406 Finally, we should note that In is poorly incorporated 428 These ndings advance the understating of the segrega- 407 at the highly Al-rich stripes. We attribute this to the 429 tion phenomena in quaternary alloys, as required to take 7 430 full advantage of the additional degrees of freedom that 444 ish MINECO project ENE2017-85087-C3. ICN2 ac- 431 they oer. 445 knowledges support from the Severo Ochoa Programme 446 (MINECO, Grant no. SEV-2013-0295) and is funded 432 Acknowledgments LF would like to thank the 447 by the CERCA Programme / Generalitat de Catalunya. 433 Nextnano team for the prompt support with the sim- 448 Part of the present work has been performed in the 434 ulations and JSR the European Synchrotron Radiation 449 framework of Universitat Autònoma de Barcelona Ma- 435 Facility (ESRF) for the beamtime allocated on the beam- 450 terials Science PhD program. Part of the HAADF- 436 line ID16B. LF, GT, SES, HP, MF, and AFiM thank 451 STEM microscopy was conducted in the Laboratorio de 437 SNF for funding through NCCR QSIT, IZLRZ2_163861 452 Microscopias Avanzadas at Instituto de Nanociencia de 438 and BSCGI0_157705 and H2020 for funding through 453 Aragon-Universidad de Zaragoza. Authors acknowledge 439 grant LIMQUET. EDR and LR thank Labex EMC3 454 the LMA-INA for oering access to their instruments and 440 AQURATE (French ANR). SMS acknowledges fund- 455 expertise. This work has received funding from the Eu- 441 ing from "Programa Internacional de Becas "la Caixa"- 456 ropean Union's Horizon 2020 Research and Innovation 442 Severo Ochoa. JA and SMS acknowledge funding from 457 Programme under grant agreement No. 654360 NFFA- 443 Generalitat de Catalunya 2017 SGR 327 and the Span- 458 Europe. The authors declare no nancial interest. 459 [1] K. Onabe, Japanese Journal of Applied Physics 21 , L323 502 sevier Inc., 2016) pp. 159184. 460 (1982). 503 [18] C. Zheng, J. Wong-Leung, Q. Gao, H. H. Tan, C. Ja- 461 [2] G. B. Stringfellow, Journal of Crystal Growth 65 , 454 504 gadish, and J. Etheridge, Nano Lett. 13 , 3742 (2013). 462 (1983). 505 [19] N. Sköld, J. B. Wagner, G. Karlsson, T. Hernán, 463 [3] J. W. MATTHEWS and A. E. BLAKESLEE, Journal of 506 W. Seifert, M.-E. Pistol, and L. Samuelson, Nano Lett. 464 Crystal Growth 27 , 118 (1974). 507 6, 2743 (2006). 465 [4] M. Leyer, J. Stellmach, C. Meissner, M. Pristovsek, and 508 [20] D. Rudolph, L. Schweickert, S. Morkötter, L. Hanschke, 466 M. Kneissl, Journal of Crystal Growth 310 , 4913 (2008). 509 S. Hertenberger, M. Bichler, G. Koblmüller, G. Abstre- 467 [5] R. B. Lewis, L. Nicolai, H. Küpers, M. Ramsteiner, 510 iter, and J. J. Finley, New Journal of Physics 15 (2013), 468 A. Trampert, and L. Geelhaar, Nano Letters 17 , 136 511 10.1088/1367-2630/15/11/113032. 469 (2017). 512 [21] E. H. Li, Physica E: Low-Dimensional Systems and 470 [6] O. Salehzadeh, K. L. Kavanagh, and S. P. Watkins, Jour- 513 Nanostructures 5, 215 (2000). 471 nal of Applied Physics 114 (2013), 10.1063/1.4816460. 514 [22] I. Vurgaftman, J. R. Meyer, and L. R. Ram- 472 [7] S. Raychaudhuri and E. T. Yu, Journal of Vacuum Sci- 515 Mohan, Journal of Applied Physics 89 , 5815 (2001), 473 ence & Technology B: Microelectronics and Nanometer 516 arXiv:1206.0723. 474 Structures 24 , 2053 (2006). 517 [23] C. Colombo, D. Spirkoska, M. Frimmer, G. Abstreiter, 475 [8] T. E. Trammell, X. Zhang, Y. Li, L. Q. Chen, and E. C. 518 and A. i Morral, Phys. Rev. B 77 , 155326 (2008). 476 Dickey, Journal of Crystal Growth 310 , 3084 (2008). 519 [24] T. V. Hakkarainen, A. Schramm, J. Mäkelä, P. Laukka- 477 [9] K. L. Kavanagh, Semiconductor Science and Technology 520 nen, and M. Guina, Nanotechnology 26 , 275301 (2015). 478 25 (2010), 10.1088/0268-1242/25/2/024006. 521 [25] M. Heigoldt, J. Arbiol, D. Spirkoska, J. M. Rebled, 479 [10] M. De La Mata, C. Magén, P. Caro, and J. Arbiol, 522 S. Conesa-Boj, G. Abstreiter, F. Peiró, J. R. Morante, 480 Nano Letters 14 , 6614 (2014). 523 and A. i Morral, J. Mater. Chem. 19 , 840 (2009). 481 [11] F. Glas, Physical Review B - Condensed Matter and Ma- 524 [26] A. i Morral, D. Spirkoska, J. Arbiol, M. Heigoldt, J. R. 482 terials Physics 74 , 2 (2006). 525 Morante, and G. Abstreiter, Small 4, 899 (2008). 483 [12] F. Glas, in Lattice Engineering , edited by S. Wang (Tay- 526 [27] L. Francaviglia, Y. Fontana, S. Conesa-Boj, G. Ttncoglu, 484 lor and Francis group, 2012) Chap. 5, pp. 189228. 527 L. Duchne, M. B. Tanasescu, F. Matteini, and A. Fontcu- 485 [13] T. Kuykendall, P. Ulrich, S. Aloni, and P. Yang, Nature 528 berta i Morral, Applied Physics Letters 107 , 033106 486 Materials 6, 951 (2007). 529 (2015). 487 [14] M. De La Mata, R. Leturcq, S. R. Plissard, C. Rolland, 530 [28] L. Francaviglia, G. Tütüncüoglu, F. Matteini, and 488 C. Magén, J. Arbiol, and P. Caro, Nano Letters 16 , 531 A. Fontcuberta i Morral, Nanotechnology (2018). 489 825 (2016). 532 [29] N. Nari Jeon, B. Loitsch, S. Morkoetter, G. Abstreiter, 490 [15] M. Heiss, Y. Fontana, A. Gustafsson, G. Wüst, C. Ma- 533 J. Finley, H. J. Krenner, K. G., and L. J. Lauhon, ACS 491 gen, D. D. O'Regan, J. W. Luo, B. Ketterer, S. Conesa- 534 Nano 9, 8335 (2015). 492 Boj, A. V. Kuhlmann, J. Houel, E. Russo-Averchi, J. R. 535 [30] N. Jeon, D. Ruhstorfer, M. Döblinger, S. Matich, 493 Morante, M. Cantoni, N. Marzari, J. Arbiol, A. Zunger, 536 B. Loitsch, G. Koblmueller, and L. J. Lauhon, Nano 494 R. J. Warburton, and A. i Morral, Nature Mater. 12 , 537 Letters , acs.nanolett.8b02104 (2018). 495 439 (2013). 538 [31] Y. Fontana, P. Corfdir, B. Van Hattem, E. Russo- 496 [16] L. Mancini, Y. Fontana, S. Conesa-Boj, I. Blum, 539 Averchi, M. Heiss, S. Sonderegger, C. Magen, J. Arbiol, 497 F. Vurpillot, L. Francaviglia, E. Russo-Averchi, M. Heiss, 540 R. T. Phillips, and A. i Morral, Phys. Rev. B 90 , 75307 498 J. Arbiol, A. Fontcuberta I Morral, and L. Rigutti, Ap- 541 (2014). 499 plied Physics Letters 105 , 243106 (2014). 542 [32] P. Trebbia and N. Bonnet, Ultramicroscopy 34 , 165 500 [17] L. Francaviglia, Y. Fontana, and A. Fontcuberta i Mor- 543 (1990). 501 ral, Semiconductors and Semimetals , 1st ed., Vol. 94 (El- 544 [33] I. Jollie, Principal Component Analysis (Springer, 8 545 2002). 560 ELECTRON DEVICES 54 , 2137 (2007). 546 [34] L. Balaghi, G. Bussone, R. Grifone, R. Hübner, J. Gren- 561 [39] J. Grönqvist, N. Søndergaard, F. Boxberg, T. Guhr, 547 zer, and A. Krasheninnikov, arXiv:1803.10873 (2018). 562 S. Åberg, and H. Q. Xu, Journal of Applied Physics 548 [35] For the elements here considered, the EELS peaks over- 563 106 (2009), 10.1063/1.3207838. 549 lap with an important background that is absent in the 564 [40] P. Pfeer, I. Gorczyca, and W. Zawadzki, Solid State 550 EDS spectra. The EELS background does not prevent 565 Communications 51 , 179 (1984). 551 the localization of the elements in the maps, but may 566 [41] S. Koshiba, Y. Nakamura, M. Tsuchiya, H. Noge, 552 reduce the accuracy of the quantication. 567 H. Kano, Y. Nagamune, T. Noda, and H. Sakaki, Journal 553 [36] F. Boxberg, N. SØndergaard, and H. Q. Xu, Advanced 568 of Applied Physics 76 , 4138 (1994). 554 Materials 24 , 4692 (2012). 569 [42] Y. Kim, H. J. Joyce, Q. Gao, H. H. Tan, C. Jagadish, 555 [37] A. Trellakis, T. Zibold, T. Andlauer, S. Birner, R. K. 570 M. Paladugu, J. Zou, and A. A. Suvorova, Nano Letters 556 Smith, R. Morschl, and P. Vogl, Journal of Computa- 571 6, 599 (2006). 557 tional Electronics 5, 285 (2006). 572 [43] P. Ballet, J. B. Smathers, and G. J. Salamo, Applied 558 [38] S. Birner, T. Zibold, T. Andlauer, T. Kubis, M. Sabathil, 573 Physics Letters 75 , 337 (1999). 559 A. Trellakis, and P. Vogl, IEEE TRANSACTIONS ON