Annealing induced atomic rearrangements on (Ga,In) (N,As) probed by hard X-ray photoelectron spectroscopy and X-ray absorption fine structure

We study the effects of annealing on (Ga0.64,In0.36) (N0.045,As0.955) using hard X-ray photoelectron spectroscopy and X-ray absorption fine structure measurements. We observed surface oxidation and termination of the N-As bond defects caused by the annealing process. Specifically, we observed a characteristic chemical shift towards lower binding energies in the photoelectron spectra related to In. This phenomenon appears to be caused by the atomic arrangement, which produces increased In-N bond configurations within the matrix, as indicated by the X-ray absorption fine structure measurements. The reduction in the binding energies of group-III In, which occurs concomitantly with the atomic rearrangements of the matrix, causes the differences in the electronic properties of the system before and after annealing.

systems 19,20 . In conventional X-ray photoemission spectroscopy using Al or Mg anodes and X-ray sources with energies of 1-1.5 keV, the escape depth of the photoelectrons is generally several nanometers. Therefore, the signals in conventional X-ray photoemission spectroscopy are strongly affected by both the surface conditions and the out-diffusion of the elements, particularly in the post-annealing case 20 . In addition, the high photon fluxes realized by undulators in third-generation synchrotron light sources and the development of the electron energy analyzer for high-kinetic-energy electrons have recently made practical high-resolution hard X-ray photoelectron spectroscopy (HAXPES) possible. The increased escape depths of photoelectrons, which range to depths of several tens of nm, with higher kinetic energies will enable nondestructive studies of bulk materials, nanoscale buried layers and their interfaces 21,22 . XAFS techniques can provide nondestructive analysis of the electronic and local structures of specific atoms. Here, we use a depth-resolved XAFS technique to analyze the effects of annealing on the sample 23 . This technique has been used to clarify the atomic structure without substrate information and provides detailed microstructural information about the buried epitaxial layers contained within the sample. These developed HAXPES and XAFS investigations can also compliment the preceding reports from Lordi et al. 's 12,13 , through the direct observation of chemical bonding in (Ga,In) (N,As) and the nearest-neighbor configurations around Ga and As due to its depth-resolved material selectivity and greater signal sensitivities. In this report, we analyze the effects of annealing on the microscopic crystal characteristics of (Ga,In) (N,As)/Ga(N,As) material systems by comparison of the electronic states and bond configurations that were obtained by HAXPES and XAFS.

HAXPES Results
In this report, we study a sample consisted of ten periods of [(Ga,In) (N,As) (8 nm)/Ga(N,As) (14 nm)] with constituent compositions of 36% In and 4.5% N for the (Ga,In) (N,As) wells and 0.8% N for the Ga(N,As) barriers, as shown in Fig. 1. To investigate the effects of annealing on the material, we made comparison for the sample before (as-grown) and after rapid thermal annealing (RTA), which was performed for 1 min under a N 2 atmosphere at 730 °C. In the HAXPES spectra in this report, when a clear fit has been obtained, the experimental data points are indicated by dots and the fitting curves are superimposed on those dots as colored lines, unless otherwise stated. Figure 1 shows the wide range HAXPES spectrum of the as-grown sample. The conditions close to the surface are also illustrated schematically. As shown in the inset in the figure, distinct In-related peaks are observed. These peaks indicate that the probing depth of this measurement is sufficient for analysis of the buried (Ga,In) (N,As) layer below the 14 nm-thick Ga(N,As) surface. In this section, we present representative HAXPES spectra to enable discussion of the findings of this study. The other HAXPES spectra, including those for the light elements of C, N, and O, along with the spectra around the Ga and In orbitals, are summarized in the Appendix. Figure 2 shows the HAXPES spectra at the As 2p 3/2 regions of the samples before and after RTA. The peaks at 1323.2 eV and 1326.55 eV can be assigned to Ga-As 24 and either As-O 24 or N-As 25,26 , respectively. Figure 3 shows the HAXPES spectra at the Ga 2p 1/2 regions of the samples before and after RTA. The main peak at 1144.6 eV appears to originate from the Ga-As bond 24,[27][28][29] . The sub-peak observed at 1145.8 eV can then be assigned to the Ga-O bond. As shown in Fig. 2, the main Ga-As peak is observed at 1323.2 eV, while a small sub-peak occurs around 1326.6 eV under as-grown conditions. The sub-peak may stem from the As-O or N-As configurations. However, this sub-peak vanished after RTA. Additionally, the intensity of the peak that was assigned to Ga-O bonding increased after RTA, as shown in Fig. 3.
These results indicate the presence of increased surface oxides after RTA, and these oxides are unlikely to decrease. Consequently, the sub-peak shown in Fig. 2 is related to the N-As configurations 30 . The group V-group V bond is a defect in itself. The as-grown sample would have N-As bonds with a concentration that is close to the detection limit. RTA diminishes the numbers of these defects, which causes their concentration to fall below the detection limit. Figure 4 shows the HAXPES spectra measured at the In 3d regions before and after RTA. The peaks shown at 452.3 eV and 444.8 eV can be assigned to In 3d 3/2 and In 3d 5/2 , respectively, and are related to the In-As bond 31 . Specifically, the peak positions of both peaks in the figure are clearly different before and after RTA. This suggests that there is a characteristic change in the electronic states around In. Further discussions involving comparison of the overall HAXPES and subsequent XAFS results will be provided later in the paper. Figure 5(a,b) show radial structure functions (RSFs) around the Ga K-and In K-edges, respectively, that were obtained via XAFS measurements. Note that In is only present in the 10-period (Ga,In) (N,As) quantum well layers, and the Ga information was extracted from approximately 40 nm of the top-most surface using depth-resolved measurements, as described in the Supplemental Material. In both RSFs, we observe a sharp peak    at approximately 2.1 Å, which corresponds to the first-nearest neighbor As bond shell. Additionally, the intensity of the main peak increases for Ga but the corresponding intensity decreases for In after RTA. When we focus on the atomic configurations around In, the results above indicate that RTA induces the reduction in the number of In-As bonds and thus also induces the corresponding increase in the number of In-N bonds. Figure 6(a) shows the polarization direction dependence of the In K-edge RSF of the sample. We have measured the spectral variations with the incident X-ray polarization directions as indicated in the figure, which shows the difference between the in-plane direction (corresponding to the incident X-ray polarizations of //[110] and [−110]) and the out-of-plane growth direction (//[001]). In the figure, the peak intensity differs between the in-plane directions and the out-of-plane [001] direction because of the atomic structure. Figure 6(b) shows the RSFs before and after RTA for incident X-rays polarized in the [110] direction. The RTA induced a reduction in the In-As peak intensity. The phenomenon is more pronounced here than in the case that was observed in Fig. 5

XAFS Results
The results shown in Figs 5 and 6 indicate the increase in the Ga-As bonding configuration and the decrease in the In-As bonding configuration that occur after RTA. Because we can obtain genuine information from the (Ga,In) (N,As) layer from the In K-edge spectra without it being affected by the Ga(N,As) barrier, we focus first on the reduction in the number of In-As configurations. The results for the Ga K-edge complement and support the findings above. The greater reduction in the In-As peak intensity that was observed in the in-plane [110] direction in Fig. 6  conclusions. The coordination number of the In atoms apparently decreases from 3.8 to the 3.5-3.6 range after RTA. In contrast, the coordination number of Ga is either maintained or increases slightly around the initial value of 3.9 after RTA. Based on the results above, the atomic configuration considered here, including the N atoms before and after RTA, is illustrated schematically in Fig. 7. As shown in the figure, the bond configurations of the N atoms that are adjacent to In increased after annealing while the corresponding bond configurations adjacent to Ga were reduced. This phenomenon can be expressed using the atomic rearrangements that were proposed by Lordi et al. 12,13 , which will be discussed further in the next section.

Discussions
Elimination of N-As bond defects and surface oxidation. We summarize the spectral peak area changes in the characteristic bond configurations that were induced by RTA in Fig. 8. As shown in the figure, the bond configurations related to the constituents, i.e., the group III to group V bonds, did not show any significant differences before and after RTA. In contrast, the N-As bond configuration that was observed in the As 2p 3/2 spectrum showed a significant reduction in intensity after RTA. In addition, the oxide related spectra from the Ga-O and H-O bonding configurations observed in the O1s spectrum showed remarkable increases in all their peak areas. The N-As bond is in itself a defect and is most likely to originate from a split interstitial consisting of a nitrogen atom and an arsenic atom on a single arsenic lattice site 30,32,33 . These split interstitial defects have been predicted theoretically as defects with energetically favorable configurations 32,33 , and these predictions have been supported experimentally [34][35][36][37] . Consequently, RTA appears to terminate these defects while concomitantly causing surface oxidation to progress 30 .  In core level shifts with atomic rearrangement. Figure 9 compiles the HAXPES spectral peak shifts that were induced by RTA. We only observe significant spectral peak shifts for the In-related peaks, indicating a reduction in the binding energy. The RTA thus has a peculiar effect on the electronic states of In. To consider the origins of this phenomenon, we first discuss the atomic configurations around In based on the XAFS results. The results shown in Figs 5 and 7 indicate that the bond configuration of In around N is changed by the annealing process. This phenomenon can be expressed in terms of the atomic rearrangements 12,13,38 . In the XAFS results, we observed possible atomic rearrangements around the majority group-V As atoms and that should have an effect on the N atoms. This indicates that there is an increase in the large number of In-N bonds for the minority group-V constituent N atoms after the annealing step. Our findings follow the results of Lordi et al. 12,13 . They found that the as-grown material consisted of near-random atomic distributions, whereas annealed samples showed segregation of the In atoms towards N 12,13 . These thermodynamically stable configurations minimize the local strain in the matrix, inducing atomic rearrangement in addition to the random distribution 12,13,38 . This is also accepted to be the origin of the blueshift in the wavelength emitted from the material. When we consider the 36% In concentration in the (Ga,In) (N,As) layer, the numbers of the configurations with one or two In atoms around the N atom would initially be roughly similar under random distribution conditions. However, the increase in the number of In-N bonds suggests that, when the most energetically favorable configurations suggested by Lordi et al. 12,13 are taken into account, the configuration with three In atoms around N could have the largest numbers within the matrix after RTA. The atomic configurations considered around N before and after RTA are illustrated schematically in Fig. 10.
As described above, RTA strongly affects the electronic status of In and this results in a chemical shift of the In peak towards the lower binding energy side in HXPES. This shift can be explained by the increase in electron charge density of the In atoms that is caused by the reduction in the interactions between In and the surrounding atoms 39 . Because the conduction band edge basically originates from the s-orbitals of the group-III elements in  the III-V semiconductor, the reduced interactions between the group-III In atoms and the surrounding atoms may induce a reduction in the group-V N-induced band gap modifications 40 . The modified electronic states of the group-III In can thus explain the blueshift that occurs in this material system after annealing.

Summary
We studied the effects of annealing on (Ga 0.64 ,In 0.36 ) (N 0.045 ,As 0.955 ) using HAXPES and XAFS. We observed surface oxidation with concomitant progress in termination of the N-As bond defects as a result of the annealing step. The characteristic chemical shift was observed solely for the photoelectron spectra related to In at energies ranging towards the lower binding energies between the adjacent atoms, and this was observed most specifically in the HAXPES spectra. The XAFS results indicate that the reduction in the binding energy around In is likely to be caused by the atomic rearrangement, which leads to increased In-N bond configurations within the matrix. The reduction of the group-III In binding energies may explain the blueshift in the emission wavelength of this material system.

Methods
We investigated (Ga,In) (N,As)/Ga(N,As) sample that was grown on semi-insulating GaAs(001) substrates by molecular beam epitaxy 41 . Conventional solid-source effusion cells were used for Ga and In, and an As-valved  cracker cell was operated in the As 4 mode. Nitrogen was supplied by an radio frequency plasma source. The substrate temperature during growth was monitored with a thermocouple, which has been calibrated by combining the GaAs oxide desorption temperature of 580 °C, the In melting point of 160 °C, and the transition points of reflection high-energy electron diffraction (RHEED) patterns 42 . Prior to the growth of the (Ga,In) (N,As)/ Ga(N,As) MQW, a 300 nm thick undoped GaAs buffer layer was grown at 560 °C. For the growth of the MQW, the fluxes of all the elements were fixed throughout this study. A low As BEP of 8 × 10 −7 Torr was used, which corresponds to a V/III BEP ratio of 5. The atomic V/III flux ratio was estimated to be 3. The growth rates were 0.24 μm/h for the Ga(N,As) barriers, and 0.4 μm/h for the (Ga,In) (N,As) quantum wells (QWs). The In content was kept at 36%. The N concentration was kept slightly higher than 4% by assuming an identical incorporation for the ternary and quaternary material. The Ga(N,As) barriers were grown by closing the shutters of both the In and the N cells in order to obtain a large band offset. However, a small amount of about 0.6-0.8% N was incorporated into the GaAs layers in our case, due to the residual amount of atomic N diffusing into the growth chamber around the closed shutter. The sample consisted of ten periods of [(Ga,In) (N,As) (8 nm)/Ga(N,As) (14 nm)] with constituent compositions of 36% In and 4.5% N for the (Ga,In) (N,As) wells and 0.8% N for the Ga(N,As) barriers, as shown in Fig. 1. Note that the In and N concentrations are among the largest concentrations of these constituents within the (Ga,In) (N,As) materials to be reported to date 41,43 . To investigate the effects of annealing on the materials, RTA was performed for 1 min under a N 2 atmosphere at 730 °C employing a halogen lamp heated rapid thermal processing system (JetFirst, SEMCO, France) after growth; these conditions appropriately enhance the luminescence efficiency of the samples 41 . During RTA, the samples were proximity-capped using a GaAs wafer to prevent evaporation of the constituent elements.
All measurements described below were performed at room temperature. The HAXPES measurements were carried out at the BL46XU beamline of SPring-8 30 . The undulator X-rays were monochromatized using a Si (111) double-crystal monochromator and were further monochromatized using a Si (111) channel-cut post-monochromator. The photon energy was set at 7.94 keV using the Si (444) reflections of the post-monochromator. The photoelectron energy analysis was performed using the VG-Scienta R-4000, and the pass energy was set to 200 eV. The X-ray angle of incidence and the photoelectron take-off-angle with respect to the sample surface were 10° and 80°, respectively. The binding energy was calibrated using the 4f core level energy that was observed from a directly deposited Au thin film on the sample surface. For qualitative analysis, the observed peak intensities were normalized with respect to the peak intensity of the dominant group-V element: the As 2p 3/2 peak. Deconvolution of the peaks that were superposed in the spectra was performed by curve fitting using a sum function of Gaussian and Lorentzian profiles with a Gaussian percentage of approximately 80%. No special surface cleaning processes were carried out before the HAXPES measurements to prevent sample damage and to obtain information about the bare sample characteristics.
The XAFS experiments were performed on the BL37XU beamline of SPring-8 23,44,45 . The X-ray beam that was monochromatized by the Si (111) double-crystal was used to irradiate the sample surface after passing through a slit with dimensions of 0.5 mm x 0.5 mm. The angle of incidence of the X-ray on the sample was controlled by considering the polarization of the X-ray beam. A two-dimensional pixel array detector (Pilatus 100 K), which was positioned parallel to the sample plane, was used to detect the emitted fluorescence. The distance between the irradiation point on the sample surface and the detector window was 172 mm. The detection angle resolution of a one-pixel array with a width of 172 μm is 1 mrad.
The fluorescence intensity that is detected at each pixel includes the different weights of contributions from the layers located at different depths that are dependent on the detection angle. Further information about the XAFS apparatus and its measurement modes is given in the Supplemental Materials. To measure the Ga K-edge, we used depth-resolved XAFS measurements; otherwise, the signal would also contain the fluorescence from the GaAs substrate. To investigate In, we performed standard fluorescence mode XAFS measurements. Because In is only included in the (Ga,In) (N,As) layer, genuine information can thus be obtained from the (Ga,In) (N,As) layer.
The extraction of the extended X-ray absorption fine structure (EXAFS) oscillation from the spectra, normalization via an edge jump, Fourier transformation, and curve-fitting analysis were carried out using the REX2000 analytical package via a standard procedure 46 . The EXAFS oscillation was analyzed using the following EXAFS oscillation function χ(k): Nf k k r kr k kr ( ) ( )exp( 2 2/ )sin(2 ( ))/ 2 2 2 , where the electron wave vector k is defined as k = [2m(E − E 0 )/ħ] 1/2 . N and r are the number of atoms and the distance between the absorbing atom and the corresponding shell, respectively; f(k) is the backscattering amplitude, σ is the Debye-Waller factor, λ is the mean free path of the photoelectrons, φ is the phase shift, m is the electron mass, and E 0 is the energy at the absorption edge. f(k) and φ(k) were calculated using the McKale database 46 . We then obtained the parameters N, E 0 , r, σ, and λ by curve fitting. To obtain the R-factor (a measure of the goodness-of-fit), we used where χ exp and χ fit are the experimentally and theoretically obtained values of χ, respectively. Figure 11 shows the HAXPES spectra in the C 1s regions. We can see that a single peak originates from C 1s at 285.7 eV 47 . We do not see any distinct differences between the features of these spectra before and after annealing. HAXPES spectra of (a) Ga 2s and (b) Ga 2p 3/2 spectral regions. The vertical lines act as guides for the eye to aid in examination of the peak shift. Figure 12 shows the HAXPES spectra in the N 1s regions. The peak at 397.8 eV originates from N 1s, which is mainly bonded to Ga 19,48,49 . The other peak at 400.1 eV is likely to be related to Ga-LMM Auger electrons 49 . Note that the HAXPES measurements are not sensitive to lighter elements. Therefore, we show this N 1s solely for the purposes of qualitative discussions or reference. We omitted the results related to this spectrum from the quantitative discussions contained in the main text.   Fig. 13, the peak at 533.1 eV originates from the atmospheric H-O bond that is commonly observed at the surface of matter 24 . The other peak at 531.7 eV results from the superposition of the oxides of the constituent elements, i.e., where the dominant component, which comes from the Ga-O configuration 50,51 , is mixed with the minor components from the In-O 52 and As-O 53 configurations. Figure 14(a) shows the HAXPES spectra in the Ga 2s regions of the sample before and after RTA. The main peak at 1302.5 eV originates from the Ga-As bond 24 . We also observed a weak peak at 1296.5 eV that induces an overall spectral line shape that is tailored to the lower energy side, although the origin of this peak is uncertain. Figure 14(b) shows the HAXPES spectra in the Ga 2p 3/2 regions. The weak signal at 1116.8 eV was attributed to Ga 27 . The distinctive peaks that occur at 1117.65 eV and 1118.5 eV can be assigned to Ga-As 28 and Ga-O bonds 27 , respectively. As shown in Fig. 3, we found an increase in the peak intensity on the oxide related peak. Figure 15(a,b) show the HAXPES spectra at the In 3p 1/2 and In 3p 3/2 regions, respectively, as measured before and after RTA. These peaks can be assigned to the In-As bond 31 . Note here that the signal-to-noise ratio is rather worse than that which originated from peaks ranging from As to Ga because of the small In content when compared with the other atoms. However, we can clearly see the peak shift before and after RTA that was also seen in the In-related peak for In 3p 1/2 in Fig. 4. Figure 16 shows the HAXPES spectra measured at the Ga 3d-In 4d regions before and after RTA. The peaks at 17.8 eV and 18.7 eV can be assigned to In-As (In 4d 5/2 ) and In-As (In 4d 3/2 ), respectively 31 . Two components that were assigned to the Ga 3d 5/2 and Ga 3d 3/2 peaks in the Ga-As bonding configuration are noticeable at 19.6 eV and 20.0 eV, respectively, because of the high energy resolution of the HAXPES measurement system that was used 50,54,55 . The peak shown at 21.0 eV is attributed to Ga-O 50,54,55 . As discussed earlier, we observed enhancement of the Ga-O related peaks, as shown in Figs 3 and 14. Most specifically, the two In-related peaks shifted obviously after RTA by more than 0.2 eV towards the lower energy side. In contrast, the shifts in the other peaks were related to the Ga marginal.