The impact of point defects in n-type GaN layers on thermal decomposition of InGaN/GaN QWs

The aim of this paper is to give an experimental evidence that point defects (most probably gallium vacancies) induce decomposition of InGaN quantum wells (QWs) at high temperatures. In the experiment performed, we implanted GaN:Si/sapphire substrates with helium ions in order to introduce a high density of point defects. Then, we grew InGaN QWs on such substrates at temperature of 730 °C, what caused elimination of most (but not all) of the implantation-induced point defects expanding the crystal lattice. The InGaN QWs were almost identical to those grown on unimplanted GaN substrates. In the next step of the experiment, we annealed samples grown on unimplanted and implanted GaN at temperatures of 900 °C, 920 °C and 940 °C for half an hour. The samples were examined using Photoluminescence, X-ray Diffraction and Transmission Electron Microscopy. We found out that the decomposition of InGaN QWs started at lower temperatures for the samples grown on the implanted GaN substrates what provides a strong experimental support that point defects play important role in InGaN decomposition at high temperatures.

InGaN/GaN quantum wells (QWs) are used in blue (wavelength emission: λ > 450 nm) and green (λ > 500 nm) Light Emitting Diodes (LEDs) and Laser Diodes (LDs) manufactured for different application like lighting, displays, communication, as well as in the niche markets of Quantum Technologies, medicine or environmental protection. However, the technology of InGaN is far from being mature and there are still some properties of this material not well understood. InGaN quantum wells are grown at relatively low temperatures. For higher In content in green emitters, it is of only 700 °C. Additionally, InGaN has a large lattice mismatch to GaN: 10% between GaN and InN. Those two factors cause a high defect density in InGaN layers: In-content fluctuations 1 , high concentration of point defects 2 and misfit dislocations for the layers above the critical thickness 3 . Such defects contribute to so called "green gap" i.e., low luminousity of the emitters in the green region of the spectrum. Last few years problem of point defects migration during the growth from n-type layers towards QWs and their influence on optical properties is widely discussed 4,5 .
Additional problem is an InGaN decomposition in active region during high temperature growth of p-type GaN and AlGaN layers above the QWs. This problem has been reported by a number of researchers [6][7][8] . In our research to date, we have observed that the decomposition of the multi QWs stack usually begins at the deepest, i.e., first grown QW. As there is no significant structural difference between the first and other quantum wells, the interaction between the InGaN QWs and the layers below should play a role. Since Si-doped n-type GaN layer grown in MOVPE method (under the N-rich conditions) contains a high concentration of Ga-vacancies 9 we have posed a hypothesis that point defects migration could be responsible for the InGaN QWs decomposition at high temperatures.
In order to check this hypothesis, we have performed the experiment consisting of three steps: (i) implantation of GaN:Si layer grown on sapphire template with He-ions to increase significantly the point defect concentration 10 , (ii) growth of five-fold InGaN/GaN QWs on the unimplanted and implanted GaN layers at temperatures 730 °C-QWs and 810 °C-quantum barriers (QBs), (iii) annealing of these samples at temperatures 900 °C, 920 °C and 940 °C.

Experiment
GaN layer was grown using MOVPE (Metalorganic Chemical Vapour Phase Epitaxy)-AIXTRON CCS 3 × 2″ reactor. The sample was grown on sapphire in a sequence: 0.1 μm of LT (low temperature) GaN nucleation layer, 1.5 μm of undoped GaN at 300 mbar and 1050 °C and 1 μm of GaN:Si (Si concentration of 2.7 × 10 18 cm −3 ) at 150 mbar and 1000 °C. Trimethylgallium (TMGa) was used as Ga precursors, ammonia as nitrogen source and H 2 as a carrier gas. Silicon doping was realized using SiH 4 .
Such sample was divided into two parts: one of them was implanted with He-ions at room temperature (RT), second one, remained unimplanted. In order to obtain uniform implantation defect distribution Heions were implanted with three energies: E 1 = 25 keV, E 2 = 75 keV, E 3 = 160 keV and three different fluences: I 1 = 1.9 × 10 14 cm −2 , I 2 = 2.7 × 10 14 cm −2 , I 3 = 5.0 × 10 14 cm −2 , respectively. Before implantation, GaN:Si layer was covered at low temperature by plasma-enhanced chemical vapor deposition (PECVD) with 20 nm thick SiN layer to avoid surface contamination. This layer was removed after implantation. Figure 1 presents TRIM simulations of impurity and defect profiles. One can see that ion implantation with three fluences and three energies of the He ions give quasi-uniform depth distribution of the vacancies. The surface morphology of both samples was investigated using Atomic Force Microscopy (AFM), using NanoScope Dimension 3100 in tapping mode with standard Si tip. High Resolution X-Ray Diffraction (HRXRD) and Scanning Transmission Electron Microscopy (STEM) methods were used for structural sample characterization. Measurements of HRXRD were performed using Empyrean (Malvern Panalytical) X-ray diffractometer operating at the CuKα 1 wavelength, equipped with the hybrid 2bounce monochromator and a threefold Ge (220) analyzer. For each sample, 2Θ/ω scan of (0002) symmetrical reflection was measured and lattice parameters change was calculated. STEM images were taken using a FEI TECNAI G2 F20 S-TWIN microscope operating at 200 kV. The cross-sectional STEM specimens were prepared in a conventional way by using mechanical polishing followed by argon ion-milling at 4 keV. The optical properties of the samples were determined using photoluminescence (PL). Spectra were measured at room temperature (RT) and in a liquid nitrogen cryostat, excitation was obtained by an 325 nm He-CD laser, worked at continuous-wave operation mode. PL was recorded using Horiba iHR 320 Imaging Spectrometer and CCD camera as a detector. www.nature.com/scientificreports/ In the next step, five InGaN/GaN QWs were grown using MOVPE method, on unimplanted and implanted GaN:Si layer. Before regrowth, both samples, were annealed for 90 s at 950 °C and 150 mbar in NH 3 atmosphere in the MOVPE reactor . After that, a 50 nm thick GaN layer was grown at 950 °C and 150 mbar with N/III ratio of 2000. Then, five nominally 2.5 nm thick InGaN QWs with 18% of indium content and 6.5 nm thick GaN QBs, were grown at 730 °C and 810 °C, respectively. These multi QWs/QBs were covered with a 56 nm GaN cap layer grown at 850˚C. Triethylgallium (TEGa) and trimethylindium (TMIn) were used as Ga and In precursors, ammonia as nitrogen source. QBs were grown using H 2 as a carrier gas, whereas QWs were grown using N 2 . The epi-structure is schematically shown in Fig. 2.
Finally, the InGaN/GaN samples grown on unimplanted and implanted GaN:Si layers were divided into four pieces. Six samples (three pieces of MQWs grown on unimplanted layer and three pieces of MQWs grown on implanted layer) were annealed at temperatures of 900 °C, 920 °C and 940 °C in NH 3 + H 2 atmosphere, for 30 min in the MOVPE reactor.
All annealed samples and as grown MQWs samples were once more characterized using AFM, STEM, HRXRD and PL. Additionally, using Epitaxy software (Malvern Panalytical) it was possible to calculate theoretical diffraction pattern. The calculated patterns were fitted to the measured data, allowing determination of QWs and QBs thickness, indium content in QWs, and additionally to assess the structural quality of the samples.

Results and discussion
GaN:Si layers: unimplanted and implanted. Morphology. To be sure that InGaN/GaN MQWs would grow on substrates of the same morphology AFM scans were made. Comparison of the surface morphology before and after implantation is presented in Fig. 3a,b. Figure 3a shows an AFM scan of the surface of the unimplanted GaN:Si layer. It has revealed terraced structure with atomic layer steps. Similar AFM scan after ion implantation and removing of the SiN cap layer, is shown in Fig. 3b. Comparison of both scans shows that the SiN cap layer preserves the low surface roughness and atomic steps are still visible after ion implantation.  Comparison of 2theta/omega scans of the layers prior and after ion implantation is shown in Fig. 4. As a result of ion implantation, a peak at lower angle with respect to GaN (0002) appeared. Such an effect was observed also by other authors for GaN implantated with Mg, Ar, Be or Ca ions [19][20][21][22] . At low ion fluences it is attributed to GaN lattice expansion caused by displacement of atoms and/or creation of small dislocation loops. The estimated change of the lattice parameter c is Δc/c = 0.2%. The ion fluence applied was too low to produce plastic deformation of the implanted layer 20,23 what was confirmed by STEM measurements as shown in Fig. 5. Additionally STEM images show that the He ions at energies and fluences used do not produce extended defects-the implanted region is still crystalline with a well-defined atomic ordering.
Optical characterization. The next step was to examine the impact of implantation on PL. Comparison of the PL spectra for implanted and unimplanted GaN:Si layers is presented in Fig. 6. In the case of unimplanted layer, standard near-band emission (NBE), phonon replicas (LO), blue luminescence (BL) and yellow luminescence (YL) are visible. In the case of layer after implantation, the PL signal was quenched in the entire spectrum. This effect is attributed to the creation of points defects, e.g., V Ga , V N , Ga i or N i , which act as non-radiative recombination centers 11,12,17,24 . InGaN/GaN MQWs grown on unimplanted and implanted GaN:Si layers. Morphology. The surface morphology of the MQWs grown on unimplanted and implanted GaN:Si layers was studied using AFM. Figure 7 shows AFM scans of the MQWs surfaces grown on the unimplanted (Fig. 7a) and the implanted layers (Fig. 7b). In both cases, atomic steps and similar amount of "V-pits" (typical for InGaN 25 ) are clearly visible.
Structural characterization of InGaN/GaN MQWs. A set of HRXRD 2theta/omega scans of (0002) GaN reflection of both grown structures is shown in Fig. 8. Based on measured HRXRD patterns, using Epitaxy software,   Table 1.
It can be stated that in both cases, the good quality MQWs were grown with similar thickness and also with similar In content. It is worth to emphasize that the additional peak at lower angle which appeared as result of ion implantation of GaN:Si layer (Fig. 4), disappeared after growth of the MQWs structure (see insert in Fig. 8). It means that strain due to the defects presence in implanted GaN has been removed during growth of the MQWs 19 . However, intensities of the higher order oscillations (4th and 5th) are lower in case of MQWs grown on implanted layer (Fig. 8), actually 5th order oscillation is not observed-it means that the crystallographic quality of the MQWs/QB S grown on implanted GaN:Si is somewhat worse as compared to growth on unimplanted layer. It can be due to diffusion of point defects from implanted layer towards MQWs during their growth.
Optical characterization of InGaN/GaN MQWs. Optical properties of the MQWs structures were examined at RT and at 77 K and their comparison is shown in Fig. 9a and b. In the case of the MQWs grown on implanted layer, similarly to the implanted layer without MQWs, (Fig. 6), there is no peak for NBE of GaN both at RT and 77 K. It indicates that the strain removal observed in the HRXRD measurements was not accompanied by the full recovery of the implantation-induced point defects during MQWs growth.
Fitting parameters of the Gaussian function to PL spectra are presented in Fig. 9. At RT (Fig. 9a) MQWs grown on the implanted layer are characterized by a red shift (475 nm vs 468 nm) and about 25% lower intensity   Table 1). Lower intensity and higher FWHM can be caused by diffusion of point defects to the QWs from the implanted layer below. These point defects start to act as non-radiative recombination centers and in consequence PL intensity decrease. PL measurements at 77 K show (Fig. 9b) that emission intensities of MQWs grown on unimplanted and implanted layers are similar (difference is only of 2.5% compared to 25% at RT) because at 77 K non-radiative recombination centers are almost "switched off ". Additionally, in both cases, FWHM are exactly the same, what strongly indicates that there is no more difference between these two samples, besides amount of point defects.   Detailed analysis of the intensities and FWHM of the Pendeloesung fringes is shown in Fig. 11a and b. Intensities data (Fig. 11a) were normalized to the intensity of GaN (0002) reflection. Initial small differences in intensities and FWHM between as grown samples ( Fig. 11a and b-first column) started to increase after annealing at 900 °C ( Fig. 11a and b-second column). In case of MQWs grown on unimplanted layer, Pendelloesung fringes 5 th order disappear and 4 th order were too weak to be analyzed. For the implanted sample, the intensity of the lower order peaks decrease and simultaneously FWHM increase compared to MQWs grown on unimplanted layer. It can be concluded from XRD data that some structural changes take place after annealing at 900 °C.
STEM pictures (Fig. 12-second row) reveal that the decomposition of the first QW after annealing at 900 °C in both cases can be observed. Trapezoid-shape objects appear in the first grown QW. These objects consist of the metallic In precipitation and voids. Around such voids an InGaN shell with a very high In content of about 80% is formed (calculated by tetragonal distortion method). STEM pictures also indicate that larger voids appear in the first QW grown on the implanted GaN layer what is consistent with the XRD data (more pronounced changes, especially of FWHM of Pendelloesung fringes in that case shown in Fig. 11b-second column).  www.nature.com/scientificreports/ Significantly larger differences between both structures are observed after annealing at 920 °C. In both cases, i.e., for the InGaN/GaN MQWs grown on unimplanted and implanted layers, HRXRD pattern ( Fig. 10a and b-blue lines) revealed further changes as compared to the as-grown samples ( Fig. 10a and b-black line). Detailed analysis of the intensities and FWHM ( Fig. 11a and b-third column) indicates that more pronounced structural changes took place in the MQWs grown on the implanted layer: intensities decrease and FWHM of Pendeloesung fringes increase compared to the MQWs grown on the unimplanted layer. In a series of STEM pictures ( Fig. 12-third row), we observed that all five QWs (70% of specimen) or four QWs (30% of specimen) got decomposed in the case of MQWs grown on implanted layer in contrast to just one decomposed QW on the unimplanted layer as a result of annealing at 920˚C.
Further changes of diffraction patterns can be observed ( Fig. 10a and b-magenta line) after annealing at 940 °C-it suggest further structural changes of the MQWs in both cases but more pronounced in case of the MQWs grown on the implanted GaN:Si layer. Comparison of the detailed analysis of the Pendeloesung fringes for both samples ( Fig. 11a and b-fourth column) shows that after annealing at 940 °C, differences in intensities and FWHM values increase. It suggests widening difference in the decomposition stage of the MQWs grown on both types of GaN:Si layer. STEM images (Fig. 12-last row) confirmed that further increase of annealing temperature up to 940˚C led to a full decomposition of all 5 QWs (100% of specimen) grown on the implanted GaN:Si layer. In the case of the sample on unimplanted GaN:Si layer, we observe an enlargement of the voids as compared to the lower annealing temperature. In that case, the first, second or third QW got decomposed, but never all at the same time (in Fig. 12 we observe decomposed third QW only).
Optical characterization of InGaN/GaN MQWs after annealing. Optical properties of the MQWs structures grown on the unimplanted and implanted GaN:Si layers after annealing at 900 °C, 920 °C and 940 °C were examined at RT and the PL spectra are shown in Fig. 13a and b. It can be observed that annealing of the samples resulted in decreasing intensities and increasing FWHM of the emitted spectra in both cases. Detailed analysis of the measured intensities in both annealed samples is shown in Fig. 14. In case of MQWs grown on unimplanted GaN:Si layer changes of the intensities up to 920 °C are almost linear (it corresponds to one decomposed QW and enlargement of the voids) and after annealing at 940 °C decrease of the measured intensity is larger (it corresponds to decomposed different QWs in the structure, but never simultaneously). In the case of MQWs grown on implanted layer, after annealing at 900 °C change of the intensity is similar to that observed for unimplanted layer (it correspond to one decomposed QW), then after annealing at 920 °C the change is much more pronounced (it correspond to 4-5 decomposed QWs). Annealing at 940 °C results in small changes of the intensities (all 5 decomposed QWs and enlargement of the voids). In conclusion, the optical changes follow the structural changes in the samples, i.e., their stages of the decomposition.

Conclusions
Complementary analytical techniques were applied to study the role of point defects migrating at elevated temperatures from the GaN:Si layers on the decomposition of the InGaN QWs grown above. Point defects were introduced into the GaN:Si layer using He-ion implantation. Photoluminescence optical studies indicate that these defects were not fully removed during MQWs growth on implanted layer. The InGaN/GaN MQWs grown on the implanted GaN:Si are of almost the same crystallographic quality as those grown on the unimplanted GaN:Si, but two observations indicate that they contain point defects diffused from the implanted region: (i) PL spectra parameters are slightly worse, (ii) intensities of the higher order Pendeloesung fringes in XRD are lower.
Structural studies (HRXRD and STEM results) have shown that the thermal decomposition of InGaN/GaN QWs begins from the lowest well after annealing at 900 °C in both cases. After annealing at 920 °C, HRXRD and STEM results indicate that in the case of the MQWs grown on the implanted layer, rapid decomposition of almost all the QWs occurs, in contrast to the QWs grown on the unimplanted GaN layer for which only one quantum well was decomposed. After annealing at 940 °C, we observe larger voids and decomposed all QWs grown on the implanted GaN layer and still decomposed only one (first, second or third) QW grown on the unimplanted GaN layer.
The experimental data show that the migration of point defects introduced by He ion-implantation into the GaN:Si layer contributed to the decomposition of all quantum wells at a lower temperature than it was the case for the QWs grown on the unimplanted layer.  www.nature.com/scientificreports/ The most probable defects responsible for the InGaN QWs decomposition are gallium (and/or indium) vacancies V Ga (or V In ) and In Ga V Ga clusters. Since such defects are present in GaN:Si and their energy barrier for migration is very low 26 , the decomposition begins from the first grown InGaN QW.
The issue of the diffusion of point defects into the QWs must be further explored. One can see that the diffusion rate must be significant as in our experiment there was 56.5 nm thick GaN layer between InGaN MQWs and the implanted region. The diffusion is enhanced by the presence of dislocations (high density of more than 10 8 cm −2 ), by electric field and strain. These factors should be varied in further experiments.
An interesting additional issue is the point defect annihilation during the InGaN/GaN MQWs growth. Some of these defects are annihilated what can be seen in the HRXRD scan (the peak related to damaged region disappears), but some not, as the PL signal is not recovered.