Modulation and Refinement of In–N re-Bonding of InGaN Through in Post-Flow During a Refined Temper Fire Treatment Process

In this article, we describe a subtle method for modulating and refining the indium–nitrogen (In–N) re-bonding effect of InGaN by employing an In post-flow during temper fire (<inline-formula> <tex-math notation="LaTeX">$\Delta T = 110\,^{\circ }\text{C}$ </tex-math></inline-formula>) treatment. After optimizing the In flow rate and the temper fire treatment process, the In content in InGaN quantum wells (QWs) increased from 12.7 to 22.3% and the (102) epitaxy quality of InGaN improved, as revealed by the full-width at half-maximum (FWHM) of the X-ray diffractometry signal decreasing from 410 to 374 arcsec. In addition, the quality of a five InGaN/GaN multiple-QW epilayer surface also improved greatly when applying this technique. Merely by modulating the In post-flow rate (0, 5.6, 11.2, 16.8, 22.4, or <inline-formula> <tex-math notation="LaTeX">$28.0~\mu $ </tex-math></inline-formula>mol/min), the In<sub><italic>x</italic></sub>Ga<sub>1−<italic>x</italic></sub>N photoluminescence signal (and FWHM) changed from 449 nm (58 nm) in the absence of In post-flow during the temper fire treatment process, to 523 nm (46 nm) when the In post-flow rate was <inline-formula> <tex-math notation="LaTeX">$11.2~\mu $ </tex-math></inline-formula>mol/min, and to 534 nm (55 nm) when the In post-flow rate was <inline-formula> <tex-math notation="LaTeX">$28.0~\mu $ </tex-math></inline-formula>mol/min. This technique is, therefore, effective at improving the InGaN quality and compensating for the In–N bond desorption rate.


I. INTRODUCTION
Light emitting diodes (LEDs) have become pervasive in our daily lives since the invention of the InGaN-based blue LED (which achieves white light in conjunction with a yellow phosphor) completed the set of LEDs of the three primary colors (RGB). Nevertheless, the materials selected for the various emission wavelengths in RGB LEDs have several limitations. For example, in green LEDs, conventional isoelectronic doping with GaP:N leads-because of the indirect band gap-to a low external quantum efficiency (EQE). Upon increasing the content of aluminum in AlGaInP, the direct energy band gap is increased and the emission wavelength becomes shorter, resulting in the light emission efficiency becoming higher than that of GaP-based LEDs. Nevertheless, The associate editor coordinating the review of this manuscript and approving it for publication was Jiajie Fan . when with the aluminum content is too high (near to the green band), the weak conduction band discontinuity results in a dramatic decrease in the EQE. InGaN has been used successfully in blue-band LEDs, but increasing the In content, aiming for a longer emission wavelength, can cause the EQE to decrease, due to the lower quality of the InGaN quantum wells (QWs) and a phase separation effect in the InGaN. [1] Although nitride-based materials have provided emission wavelengths of up to 740 nm, there remain a need to improve the performance. [2] Furthermore, for micro-LEDs, the self-emission of a AlGaInP red LED has yet to be realized. If we could overcome this obstacle, InGaN would appear to be the best candidate material for RGB self-emission LEDs, allowing the realization of all nitride-based micro-LED displays. [3] Many breakthroughs have been made in extending the emission of InGaN to longer wavelengths (toward the green VOLUME 8, 2020 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ wavelength). For example, increasing the thickness and quality of the GaN buffer layer appears to be suitable for improving the performance of next-generation devices. [4]- [6] Hangleiter et al. generated V-pits in an InGaN/GaN superlattice (SL) structure to release the strain in the InGaN active region; these V-pits, which originated from threading dislocations, also enhanced carrier injection, screening dislocations, and radiative recombination. [4], [7]- [9] Common approaches for improving the crystalline quality of InGaN active layers include the use of two-step high-temperature growth and hybrid structures. Saito et al. reported that twostep high-temperature growth, with different growth temperatures for the well and the barrier layer, improved the flatness of the barrier layer and maintained the uniformity of the thickness of the QW layer in the case of multilayer growth.
[10]- [13] When the growth temperature of the barrier was higher than that of the well, the quality of the barrier improved to led to superior InGaN QWs; this approach was especially useful for preparing multiple quantum well (MQW) LEDs. Unfortunately, the In bonds undergo greater desorption during the temperature ramping process. To depress the rate of In desorption, Hubáčeka et al. developed some subtle methods for varying both the GaN and InGaN cap layer thicknesses at the InGaN QW growth temperature. [12] The optimal conditions-employing a thinner InGaN composition cap layer and no interruption during the temperature ramping process-resulted in the signal in the photoluminescence (PL) spectrum changing from 410 to 450 nm. Thus, the presence of an InGaN QW capping layer can suppress the desorption of In during the temperature ramping process. Nevertheless, the quality of the capping layer remains poor when using a low growth temperature, potentially seriously affecting the quality of the next InGaN QW.
The challenge remains to maintain the InGaN/GaN QW under compressive strain and improve the InGaN crystal quality which ensuring a high In content in the InGaN QW. The concept of a repaired weak bonding method (balancing the desorption rate and re-bonding rate) for chemical equilibrium has been developed. In this article, we demonstrate a subtle method for modulating and refining the In-nitrogen (In-N) re-bonding effect of InGaN by employing an In post-flow during the temper fire treatment process ( T = 110 • C). Under our optimal experiment conditions, we obtained high quality InGaN with a high In content in an efficient manner. Moreover, this method effectively compensated for the rate of InN desorption.

II. EXPERIMENTAL
All the samples were grown on c-plane sapphire substrates in a triple-layer horizontal-flow Nippon Sanso SR-2000 metal-organic vapor phase epitaxy (MOVPE) system under atmospheric pressure. For growth of the GaN buffer layer, hydrogen (H 2 ) was used as the carrier gas and trimethylgallium (TMGa) and ammonia (NH 3 ) were used as the Ga and N precursors, respectively. For growth of the MQWs, nitrogen (N 2 ) was used as the carrier gas and triethylgallium (TEGa) and trimethylindium (TMIn) were used as the Ga and In precursors, respectively. Fig. 1 presents the structure of the samples. A four-µm-thick GaN layer was prepared in two steps; first, a low-temperature nucleation layer was deposited and then a high-temperature buffer layer was formed on the sapphire substrate. [14]- [17] The following blue band QW layers featured two pairs of GaN layers and a lower-In-content In 0.12 Ga 0.88 N QW for strain modulation and reduce the effect of piezoelectric field on the subsequent InGaN MQW layer. Iida et al. reported that a high-In-content InGaN active layer exhibits two peak emissions, attributed to phase separation in the InGaN layers; an InGaN medium QW layer hybrid structure can not only suppress these emissions almost completely but also result in narrower full-widths at half-maximum (FWHMs) in the spectra of the InGaN active QW. [11] Finally, the active region was formed from five pairs of two-step MQWs on the strain modulation blue band, each consisting of an In 0.22 Ga 0.78 N well layer (3 nm) and a GaN barrier layer (10 nm). The growth temperature of each GaN barrier was 110 • C higher than that of the InGaN QWs, with the temper ramp time maintained at 3 min. For the structures described above, the TEGa and TMIn source flows were identical, but different TMIn supply post-flows were tested during the temperature ramping time (180 s) during the growth of the GaN barrier. Fig. 2 provides the temperature and TMIn flow profiles for the formation of the active layer. Fig. 2(a) displays the control sample A (in Table 1) flow profile; the growth temperatures for the InGaN QW and GaN barrier were 720 and 830 • C, respectively. After growth of the QW layer, the temperature was ramped to 830 • C without a TMIn source post-flow. Fig. 2(b) presents the profiles of the other treatment group samples; various TMI source post-flow rates (5.6, 11.2, 16.8, 22.4, 28.0 µmol/min) were applied during the course of temperature ramping, with all other growth conditions remaining identical.
The crystal quality of the GaN in each sample was determined from symmetric (002) and asymmetric (102)  ω-scans measured using high-resolution X-ray diffractometry (HR-XRD, PANalytical X'Pert). PL spectroscopy was used to define the QW emission wavelength and optical quality and to determine the InGaN band gap (E g ) by neglecting the InGaN potential reaction with Vegard's law. A semiconductor pulsed laser emitting at 248 nm was used for PL excitation. Finally, the epitaxy of the surface morphology and the relative In content were determined using atomic force microscopy (AFM) and secondary ion mass spectrometry (SIMS), respectively. Table 1 presents the XRD measurement data and calculated dislocation densities of all of the samples prepared with different TMI post-flow flow rates during the temperature ramping process. Fig. 3 reveals the effect of the TMIn flow on the PL peak wavelength for each sample. Within the active layer, the electron-hole radiation rate will be affected directly by the QW crystalline quality and the confined potential barrier. A two-step high-temperature growth process is used commonly to improve the LED efficiency; a higher growth temperature for the barrier can not only significantly ameliorate the flatness of the GaN barrier surface and the crystalline quality but also meet the minimum quality requirements for subsequently formed InGaN QWs. In Table 1, sample A represents the control sample. During the growth of the five QWs, the InGaN wells were prepared at 720 • C and then the temperature for GaN barrier growth was ramped up to 830 • C. During the interruption period, the flows of both TEGa and TMIn were closed while waiting for the growth temperature to reach 830 • C as the conditions for the growth of the next barrier. Notably, the device target emission wavelength was set within the green band, but the PL peak wavelength in Fig. 3 reveals that the emission wavelength of sample A was approximately 449 nm-that is, it had shifted to the blue band.

III. RESULTS AND DISCUSSION
The In content in the InGaN QW had decreased dramatically in sample A relative to other samples, presumably because of desorption of the In-N bonds during the period of temperature ramping, where the InNbonding electrons possessed sufficient energy to escape, such that a desorption mechanism occurred and the In content of InGaN decreased rapidly. Although some previous reports have indicated that the use of an InGaN low-temperature QW cap layer can suppress the rate of In desorption [12], the alloy arrangement and crystal quality of the binary GaN material were superior to those of the ternary InGaN materials. To improve the quality of InGaN MQWs, we sought to decrease the rate of In-N bond desorption and to establish an In-N re-bonding process closer to thermal equilibrium by refining the temper fire treatment process. While maintaining the temperature ramp time between the QW and the barrier at 3 min, we wished to compensate for more desorption of In-N weak bonds during the temperature ramping time with more stable In-N re-bonding. Thus, all of the other samples retained the same structure as sample A, but we employed different TMIn source post-flow rates during the period of temperature ramping as a subtle method for refining the temper fire treatment of In-N bonds. Table 1  where D is the dislocation density and b is the Burgers vector. The calculated dislocation densities (Table 1) reveal that the densities of the screw dislocations for all of our samples were similar. We suspect that the In desorption process was mostly related to the formation of edge dislocations, whose Burgers vectors were aligned along the QW-barrier interface, leaving the symmetric (002) planar spacing essentially undistorted. Because the density of edge dislocations was mostly reflected by the FWHM of the asymmetric (102) value, the effect of introducing the TMIn post-flow into the reactor chamber was manifested by the decreasing values of D edge in Table 1. For samples B-F, the contents of edge-type dislocations were 11-16% lower than that in sample A, which was prepared without the TMIn post-flow introduced into the reactor chamber. Fig. 3 also presents the relationship between the TMIn post-flow rate during temperature ramping and the PL spectral peak wavelength. The PL spectral peak wavelength shifted to substantially longer wavelength-from the blue-violet (449 nm) to the blue-green (534 nm)-when the TMIn post-flow was introduced during the period of interruption. The lateral growth mode is most determined by growth temperature. But how to control the InGaN desorption rate is very critical in InGaN. We have executed both to depress the InGaN MQW desorption rate and also increase the indium atom migration length.
In the absence of the TMIn post-flow during the interruption period, the PL spectral peak wavelength appeared at 449 nm and the corresponding In content was 12.7%. We calculated the InGaN In content by using the Schrödinger wave equation with a quantized level shift (will be published later) and following Vegard's law, estimated as follows: where E g,InGaN (x), E g,InN , and E g,GaN represents the band gaps of InGaN, InN, and GaN, respectively, at room temperature and b (equal to 2.8 eV) is the bowing parameter for InGaN. [21], [22] When the TMIn post-flow rate during the temperature ramping period was 5.6 µmol/min, the PL spectral peak wavelength increased to 515 nm and the FWHM decreased from 58 nm (sample A) to 52 nm (sample B).
In general, the FWHM of the PL spectral signal was inversely related to the PL spectral emission band gap. Sample A had an In content lower than that of sample B, but the FWHM of the PL spectral signal of sample B was less than that of sample A. Thus, through modulation and refinement of the In-N re-bonding of the InGaN by applying an In post-flow during the temper fire treatment process, we could greatly increase both the In content of the InGaN and, astonishingly, the epitaxial quality of the InGaN QW. When we increased the TMIn post-flow further, the PL spectral signal underwent a continuous shift to a longer wavelength, due to a greater In-N re-bonding effect during the ramping period. Nevertheless, the peak wavelength shift became increasingly saturated upon proceeding from sample B to sample F, indicating that not only the rate of In-N bond desorption suppressed effectively but also more compensating In-N bonds were formed. The optimal sample was that formed when the post-flow rate was 11.2 µmol/min, with the PL spectral emission peak and the FWHM being 523 and 46 nm, respectively. When the flow rate was saturated at 16.8 µmol/min, the PL spectral emission peak appeared at 530 nm, corresponding to an In content of 21.9%. The PL spectral data suggested that the introduction of the TMIn post-flow into the chamber suppressed the In desorption significantly, by at least 8%. We want to breakthrough the destiny of the InGaN growth with high indium composition temperature limited. As we know, the bounding site will turn better with the growth of increased temperature. So, in our study, we refined the indium-nitrogen (In-N) re-bonding effect of InGaN by employing an In post-flow during temper fire ( T = 110 • C) treatment. From PL data, we have to make sure the weak bonding of InGaN immediately desorption during the temper fire process and re-bonding a better bonding site of InGaN with the In-post flow rate method, which could effectively compensate for the weak bonding InGaN desorption rate and also increase the InGaN bonding energy. We used AFM to evaluate the surface morphologies of our samples. Fig. 4(a) reveals an average surface roughness (Ra) of 1.48 nm for the sample prepared without introducing a TMIn post-flow into the chamber during temperature ramping period; when the TMIn post-flow was introduced into the chamber at a rate of 22.4 µmol/min during the period of temperature ramping, the value of Ra decreased significantly, to 1.09 nm [ Fig. 4(b)]. This behavior is consistent with the XRD data, which suggested that the introduced TMIn post-flow not only decreased the dislocation density but also led to a smoother surface morphology. Finally, we used SIMS to measure the relative In contents. The SIMS data (not shown here) revealed that the In/Ga concentration ratio in the InGaN QWs increased upon increasing the depth. Thus, by applying an TMIn post-flow during a refined temper fire treatment process, we could not only compensate for the rate of In-N bond desorption but also enrich the sample with In-N bonds. In the QW region, the In content of the sample prepared with an introduced TMIn post-flow was six times higher than that of the sample prepared without the TMIn post-flow.

IV. CONCLUSION
We have prepared InGaN MQW structures by applying an In post-flow during the period of interruption, thereby effectively compensating for the desorption of In-N bonds by enriching the In-N bonds and improving the epitaxial quality of the InGaN MQWs. After optimizing the In flow rate and refining the temper fire treatment process, the In content of the InGaN QWs increased from 12.7 to 22.3% and the (102) epitaxial quality improved, with of the FWHM of the XRD signal of the InGaN decreasing from 410 to 374 arcsec. Furthermore, the PL spectral FWHM was inversely related to the PL spectral emission band gap of the InGaN MQW. Although sample A had an In content lower than that of sample B, the PL spectral FWHM of sample B was less than that of sample A. Thus, through modulation and refinement of the In-N re-bonding process of InGaN samples by applying an In postflow during a refined temper fire treatment process, we could greatly increase the In content in the InGaN samples astonishingly, and improve the epitaxial quality of the InGaN MQWs.

ACKNOWLEDGMENT
(Tsung-Yen Liu and Lung-Chien Chen contributed equally to this work.)