Twelve InGaN MQW LED samples with varying well thickness grown via metal-organic chemical vaper deposition (MOCVD) are investigated. It is observed from electroluminescence (EL) measurement that at low current densities, the peak energy shifts to blue with increasing current, and when the current change by fixed increment, the peak energy shifts to blue end to different extent among samples. This blue shift was expected to be stronger when the well thickness increases, however, for well widths above 5 nm we observe a decrease in emission energy. Since no relaxation was detected from reciprocal space mapping (RSM), the deteriorated homogeneity is found to be responsible for this phenomenon. Temperature dependent photoluminescence (TDPL) results analyzed by band-tail model fitting show that the localization effect gets more prominent with increasing well thickness. It is found that elevating the growth temperature of active region from 710°C to 750°C significantly improves the homogeneity of InGaN layer.
© 2018 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
The group III-nitrides (e.g., AlN, InN, and GaN) with their ternary and quaternary alloys have contributed to solid-state lighting by enabling efficient light emitting diodes (LEDs)  and lasers , and opened up other research areas such as solar cells , power electronics , optical wireless communications , thermoelectricity , nanoribbon , photonics , acoustic filters , high electron mobility transistors , and high frequency electronic devices , in the past two decades. They are also expected to bring display industry to the next era . Among the optoelectronic applications, InGaN/GaN multiple quantum wells (MQWs) has been appealing to researchers to pursue better performance by seeking appropriate growth and structural parameters, among which the well width and growth temperature are prominent ones. In commercialized devices, the typical thickness of QW is only 2∼3 nm. On one hand, the quantum well would not be designed too thin for this is not beneficial for the well to capture the injected carriers . In addition, thin wells are more susceptible to well width fluctuation and interface roughness. On the other hand, the quantum confinement effect will decay in broader wells. Reports had shown that the PL peak emission intensity decreased dramatically with the increase in InGaN well width after 6 nm due to weakened quantum confinement effect . Besides, it is usually difficult to grow thick wells without relaxation and thus induced defects. In general, however, a larger volume of active region is preferable for higher performance of optoelectronic devices, such as high-power LEDs and laser diodes [12, 15]. In this experimental investigation on the influences of well thickness, the samples are designed to change its value in a quite wide range and with relatively small steps.
Considering the large quantity of crystallographic defects exemplified by high density of threading dislocations, the light emission efficiency of InGaN-based light-emitting devices is surprisingly high. This is, as generally accepted, resulted from the localization of free carriers in the localized potential minima, induced by well width and indium content fluctuation in InGaN QWs, which will suppress non-radiative recombination process [16–18].
Research where indium content and well width are adjusted simultaneously to achieve the same emission wavelength found that sample with higher indium content and smaller well width has a stronger carrier localization effect , however, the dependence of localization effect on well width alone is not clear yet. Other previous study on this issue showing a positive correlation between well width and localization effect yet limited by a narrow range of 1.8 nm to 3.6 nm of well width and mere PL measurements were taken .
In this study, 9 samples with well thickness ranging from 2.5 nm to 7.6 nm present a stronger localization effect for broader wells, which further explained the interesting phenomenon that the blue-shift energy, induced by current injection, rises first and then declines with increasing well thickness. In addition, in order to reduce inhomogeneity of InGaN alloy, another three samples with the same nominal well thickness and indium content as their counterparts but grown at higher temperature during MQW growth were implemented and the result was examined.
2. Experimental process
Twelve InGaN/GaN MQW samples were grown on c-plane sapphire substrates via metal-organic chemical vapor deposition (MOCVD) system with close-coupled showerhead vertical reactor. During the epitaxial growth, triethylgallium (TEGa), trimethylindium (TMIn) and ammonia (NH3) were used as precursors for Ga, In and N sources, respectively. All the samples consist of a 20 nm thick buffer layer, a 1 μm thick Si-doped GaN layer, a two-period unintentionally doped InGaN/GaN MQW active region and a 150 nm Mg-doped GaN layer. With the same 10 nm barrier layer and nominal indium concentration of 7%. For 9 of the 12 samples, both the InGaN well and GaN barrier layers were grown under 710 °C, their well thickness varied from 2.5 nm to 7.6 nm step by step through changing the growth time. Three additional samples were grown under 750 °C and have the well thickness of 2.5 nm, 4.4 nm, and 5.7 nm, respectively. A 2.2 nm cap layer was added above each well and annealed with increased temperature for 180 seconds so as to remove the In-rich surface film on InGaN layers. The whole structure is depicted in Fig. 1.
Room temperature electroluminescence (EL) spectra were measured using Ocean Optics HR2000 high-resolution spectrometer at direct current (DC) mode. Temperature-dependent PL (TDPL) measurements were carried out with the 325 nm line of a He-Cd laser at an excitation density of 0.4 W/cm2 and the temperature was controlled to change from 30 K to 300 K using a closed-cycle helium refrigerator of CTI Cryogenics. Nonsymmetric diffraction reciprocal space mapping (RSM) was performed along the (105) direction with a Rigaku X-ray diffractometer to check the relaxation of InGaN/GaN MQW active region.
3. Results and discussion
In EL measurement, it is found that under a fixed injection current (1 mA in Fig. 2), the EL peak energy decreases monotonously as the well gets broader, and then rises again abnormally after well width is over 5 nm with a gentler slope. On one hand, the optical transition has lower energy in broader quantum wells, and on the other hand, the cusps of conductive band and valence band in a triangular potential well move closer to each other when the well gets broader due to more impact from quantum confined Stark effect (QCSE), hence the electrons and holes recombine with a smaller transition energy. It is highly possible that when the well width increases to larger than 4.4 nm, the rising part of peak energy might be resulted from the relaxation in broader wells. To verify the strain condition, reciprocal space mapping (RSM) was performed. As Fig. 3 shows, the main EL peak (GaN peak) and the satellite peaks are well-aligned along Qz axis. the differences of the position of satellite peaks from the main peak position are less than 0.001 Å−1, i.e. the well layers are fully strained according to XRD RSM measurement, which suggests that the polarized electric field should not be weakened due to relaxation in thicker well samples. It means that the rising part of EL peak energy is conceivably ascribed to the difference in the carrier screening effect: Even if the broader and the narrower well samples were under the same current density injection, the carriers are captured more effectively in broader wells with a less leakage, which may result in a blue shift of the EL peak energy for samples with increasing well width.
It is also found in EL measurement that as injection current increases, the emission peak energy ascends to different extent at first, and then descends, as is shown in Fig. 4. It can be understood that at lower injection currents, the carriers accumulated in the QWs perform as screening charges over the spontaneous and piezoelectric polarized electric field in InGaN/GaN MQW. It is known that the quantum confined stark effect (QCSE) due to polarized electric field results in inclined wells and thus red-shifted EL peaks . The electric field formed by injected carriers counteract with that of polarization, therefore, the more carriers, the stronger screening effect formed by them, the less red shift caused by polarization electric field, thus the emission peak moves to higher energy with increasing current. Note that the variation of blue shift differs for different samples, which will be discussed later. However, with the injection current increases further, the band-gap shrinking process owing to thermal effect gradually takes over, so the peak energy declines with increasing well width.
As mentioned above, when the injection current increases from 1 to 30 mA, the studied samples behave differently in the blue-shift as depicted in Fig. 5. where the variations of blue shift in peak energy shown on the low current side of Fig. 4 were measured. Fig. 6 schematically demonstrates why the blue-shift energy increases with broadening well width monotonously at first, and then decreases dramatically when the well width further broadens beyond 5 nm. It is obvious that injected carriers can impact on the electric field distribution to induce a screening effect. For a fixed carrier injection increment (from 1 to 30 mA here), the net electric field in broader wells will have a larger change and a less sloped energy band consequently, hence will be less influenced by QCSE due to carrier screening. This difference is from two aspects: a broader well brings (i) a greater potential change (ε2 > ε1) and (ii) a greater change of slope of the energy band in the well region (θ2 > θ1), as schematically shown in Fig. 6.
To discuss this feature in a quantificational way, the QW structures were simulated by TCAD program from Crosslight Software Inc. Data extracted from the energy band diagram present larger slope change for thicker wells. The exact values for the thinnest (2.5 nm) and the thickest (7.6 nm) well samples are listed in Table 1 as the current increases from 2 mA to 22 mA.
The slope changes are illustrated as θ1 and θ2 for thinner and thicker wells respectively in the schematic diagram in Fig. 6. A larger slope change accompanying with broader well width geometrically results into a larger potential elevation of well bottom, i.e. ε2 > ε1 in Fig. 6, thus more significant screening effect and less red shift due to QCSE.
Based on the above frame work, the variation of blueshift should have ascended all the way as the well gets wider. However, it dropped when the well width increases beyond 5 nm. One plausible explanation is that the electric field distribution, either the polarized field or the screening field, is perturbed in thicker wells. Reminded that RSM diagram [Fig. 3] had shown no detectable relaxation, it is reasonable to believe that the perturbation is from the screening process, exemplified by localized states in InGaN due to inhomogeneity of Indium molar fraction and well thickness. According to QCSE, wave functions of electrons and holes are spatially separated to opposite sides of the well, and the screening electric field would theoretically cover the whole layer of the QW along c-axis. However, localized states weakened this separation and results into partial or incomplete screening, rendering a stronger QCSE compared to theoretical screening effect. Hence the thick wells with stronger localization effect shift less to high energy end as depicted in the dropping segment in Fig. 5. Temperature dependent photoluminescence was also performed to consolidate this inference. The peak positions in dependence of temperature for varied well width are illustrated in Fig. 7.
As the signature of localization effect [22, 23], temperature dependent “S-shape” curve of luminescence peak energy can be observed. As shown in Fig. 7(b), it is more conspicuous in middle-width well. The S-shape PL peak curves are generally composed of three segments. I: red shift, II: blue shift, III: red shift. The curves manifest themselves as follows: at very low temperature, e.g. 10 K, the photogenerated carriers are “frozen” spatially at where they are excited due to low mobility, so they distribute randomly in the whole QW region and recombine over a correspondingly broad range of photon energy. With temperature rising, the carriers migrate off to lower energy and recombine at the potential minima on the energy landscape, presented as the red shift of PL peak (segment I). On further raising temperature, carriers are now able to migrate out of the potential minima and redistribute in a relatively wide range of energy. This shift can give rise to a temperature-induced blueshift of the emission peak when the rate of the shift overcomes the rate of temperature-induced band-gap shrinkage (segment II). As temperature steps close to room temperature, potential minima cannot trap carriers anymore due to increased thermal energy and band-gap shrinking process prevails, therefore the PL peak moves to red end again as predicted by the well-known Varshni equation .
For the sample with narrower well [Fig. 7(a)], the third segment, where red shift is an obvious tendency with increasing temperature, implies that the emitting energy is subject to band gap shrinking process. Whereas in sample with broader well [Fig. 7(c)], the blue-shift segment lasts so long that the third red-shift segment hardly begins, indicating that, compared with the other two samples, the energy of localized states is deeper into the forbidden band and the movement of carriers out of these potential pits competes with band gap shrinking. so that even at 300 K the band gap shrinking does not occur yet. In other words, the temperature where the third segment of red shift begins becomes higher than 300 K in thicker wells.
To quantitatively compare the depth of localized states in the broad and narrow wells, the experimental data were fitted with band-tail model, which approximates localized-state ensemble (LSE) model in high temperature [25,26]. In InGaN QWs, localization states are treated as band tail states that statistically distributed lower than the band edge. As signified by Eliseev et al., in the band-tail model, the temperature-dependent emission peak energy can be given by the following expression [27,28]:
The fitting parameter shows that the localized-state energy is 11.47 meV, 18.37 meV and 45.63 meV for 2.5 nm, 4.4 nm and 5.7 nm wells respectively, which implies that the spreading of localized states are broadened with greater well width.
Based upon the above discussions on localized-state energy accompanying with the PL results, it is suggested that the homogeneity of InGaN QWs are worsened with increasing well thickness. During the step-flow growth of the active layer, any imperfection of layers laid below may accumulate up towards to upper layers, resulting in a gradually serious inhomogeneity of indium content and well thickness. One of the factors that deteriorate active layers is that the step-flow growth is disturbed by small islands and adatoms that are unable to diffuse to the edge of terraces. This may attribute to tepid temperature for InGaN growth. Generally, the temperature of InGaN QW growth is 680 °C to 780 °C. 710 °C, as in our experiment, is relatively low, thus the adatoms have insufficient kinetic energy to diffuse, which favors the formation of islands, clusters, and further results in gradually lumpy surface with increasing growth time.
From this perspective, we grew another three samples with the growth temperature of QW risen to 750 °C. Their nominal In molar content remained the same as previous samples, i.e., 7%, and well thicknesses are 2.5 nm, 4.4 nm and 5.7 nm, corresponding to the thin, middle and thick QW in the former set of samples respectively.
A comparison between the low growth temperature set (L25, L44, L57) and high growth temperature set (H25, H44, H57), as in Fig. 8, shows that within both sets, the localization effect is enhanced with increasing well thickness according to the S-shape curve explanation, however, the homogeneity of high temperature set is significantly improved: different from L25, the PL peak of sample H25 shifts little to higher energy in middle temperature range and begins to red shift till 170 °C; the third section of S-shape curve of H44 begins at ∼130 K, much earlier than L44 (at 175 K) and it shifts to red end much faster than L44, indicating the change of peak energy is band-shrinking predominated; for H57 the long-lasting blue shift is suppressed in comparison with L57, which is also a sign of better homogeneity. The σ value (from the band-tail model) of high growth temperature set samples is reduced to 6∼7 meV as well. Note that the peak energy positions of high temperature set are higher than those of low temperature set. For one reason, this may be due to slightly reduced indium incorporation because of higher growth temperature, which is supported by data from XRD ω−2θ scan, thus the emission energy recovers influenced by weker QCSE. Additionally, the concentration difference is greater for broader samples. This trend is consistent with data in Fig. 8. However, the band gap changes calculated from the Vegard’s law (at 30 K) are only ∼18% of the actrual difference, thus another contributor can be the reduced depth of localized states.
In summary, two sets of InGaN MQW samples with varying well width are grown by MOCVD and investigated by EL, XRD and TDPL. From EL and PL results, it turns out to show a deterioration of homogeneity of InGaN layer with increasing well thickness. This explains the anomalous reduction of blue-shift variation of EL peak energy when a fixed increment of current injection was applied. Otherwise, the blue shift should have risen all the way up as the well gets broader. However, if a thicker well must be achieved, higher growth temperature of active region growth can be adopted to mitigate the excessively strong localization effect. In other words, to achieve a better homogeneity, narrower well layer and higher temperature of InGaN/GaN MQW growth will be the more beneficial choice.
National Key R&D Program of China (2016YFB0401801, 2016YFB0400803); National Natural Science Foundation of China (61674138, 61674139, 61604145, 61574135, 61574134, 61474142, 61474110, 61377020, 61376089); Science Challenge Project (TZ2016003); Beijing Municipal Science and Technology Project (Z161100002116037).
References and links
1. S. Nakamura, T. Mukai, and M. Senoh, “Candela-class high-brightness InGaN/AlGaN double-heterostructure blue-light-emitting diodes,” Appl. Phys. Lett. 64, 1687–1689 (1994). [CrossRef]
2. S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, Y. Sugimoto, and H. Kiyoku, “Ridge-geometry InGaN multi-quantum-well-structure laser diodes,” Appl. Phys. Lett. 69, 1477–1479 (1996). [CrossRef]
3. N. Lu and I. Ferguson, “Iii-nitrides for energy production: photovoltaic and thermoelectric applications,” Semicond. Sci. Technol. 28, 074023 (2013). [CrossRef]
4. B. J. Baliga, “Gallium nitride devices for power electronic applications,” Semicond. Sci. Technol. 28, 074011 (2013). [CrossRef]
5. C. Du, C. Jiang, P. Zuo, X. Huang, X. Pu, Z. Zhao, Y. Zhou, L. Li, H. Chen, W. Hu, and Z. L. Wang, “Piezo-phototronic effect controlled dual-channel visible light communication (PVLC) using InGaN/GaN multiquantum well nanopillars,” Small. 11, 6071–6077 (2015). [CrossRef] [PubMed]
6. A. Sztein, J. Haberstroh, J. E. Bowers, S. P. DenBaars, and S. Nakamura, “Calculated thermoelectric properties of InxGa1−xN, InxAl1−xN, and AlxGa1−xN,” J. Appl. Phys. 113, 183707 (2013). [CrossRef]
7. R. Chandiramouli and S. Sriram, “First-principles investigation on band structure and electronic transport property of gallium nitride nanoribbon,” Nano. 9, 1450020 (2014). [CrossRef]
8. N. Vico Triviño, U. Dharanipathy, J.-F. Carlin, Z. Diao, R. Houdre, and N. Grandjean, “Integrated photonics on silicon with wide bandgap GaN semiconductor,” Appl. Phys. Lett. 102, 081120 (2013). [CrossRef]
9. H. Yang, W. Wang, Z. Liu, and G. Li, “Homogeneous epitaxial growth of aln single-crystalline films on 2 inch-diameter si (111) substrates by pulsed laser deposition,” CrystEngComm. 15, 7171–7176 (2013). [CrossRef]
10. W. Wang, W. Yang, H. Wang, and G. Li, “Epitaxial growth of GaN films on unconventional oxide substrates,” J. Mater. Chem. C 2, 9342–9358 (2014). [CrossRef]
11. M. A. Khan and V. Adivarahan, “Fabrication technique for high frequency, high power group iii nitride electronic devices,” (2013). US Patent 8,476,125.
12. J. Lingrong, L. Jianping, T. Aiqin, C. Yang, L. Zengcheng, Z. Liqun, Z. Shuming, L. Deyao, M. Ikeda, and Y. Hui, “Gan-based green laser diodes,” J. Semicond. 37, 111001 (2016). [CrossRef]
13. M. Aumer, S. LeBoeuf, B. Moody, and S. Bedair, “Strain-induced piezoelectric field effects on light emission energy and intensity from AlInGaN/InGaN quantum wells,” Appl. Phys. Lett. 79, 3803–3805 (2001). [CrossRef]
14. T. Ko, T. Lu, T. Wang, J. Chen, R. Gao, M. Lo, H. Kuo, S. Wang, and J. Shen, “Optical study of a-plane InGaN/GaN multiple quantum wells with different well widths grown by metal-organic chemical vapor deposition,” J. Appl. Phys. 104, 093106 (2008). [CrossRef]
15. D. Zhao, J. Yang, Z. Liu, P. Chen, J. Zhu, D. Jiang, Y. Shi, H. Wang, L. Duan, L. Zhang, and H. Yang, “Fabrication of room temperature continuous-wave operation GaN-based ultraviolet laser diodes,” J. Semicond. 38, 051001 (2017). [CrossRef]
16. Y. Narukawa, Y. Kawakami, M. Funato, S. Fujita, S. Fujita, and S. Nakamura, “Role of self-formed InGaN quantum dots for exciton localization in the purple laser diode emitting at 420 nm,” Appl. Phys. Lett. 70, 981–983 (1997). [CrossRef]
17. F. C. Massabuau, P. Chen, M. Horton, S. Rhode, C. Ren, T. O’Hanlon, A. Kovács, M. J. Kappers, C. J. Humphreys, R. Dunin-Borkowski, and R. Oliver, “Carrier localization in the vicinity of dislocations in InGaN,” J. Appl. Phys. 121, 013104 (2017). [CrossRef]
18. W. Liu, D. Zhao, D. Jiang, D. Shi, J. Zhu, Z. Liu, P. Chen, J. Yang, F. Liang, S. Liu, X. Yao, L. Zhang, W. Wang, M. Li, Y. Zhang, and G. Du, “Effect of carrier transfer process between two kinds of localized potential traps on the spectral properties of InGaN/GaN multiple quantum wells,” Opt. Express 26, 3427–3434 (2018). [CrossRef] [PubMed]
19. C. Li, Z. Ji, J. Li, M. Xu, H. Xiao, and X. Xu, “Electroluminescence properties of InGaN/GaN multiple quantum well-based leds with different indium contents and different well widths,” Sci. Reports 7, 15301 (2017). [CrossRef]
20. X. Li, D.-G. Zhao, D.-S. Jiang, J. Yang, P. Chen, Z.-S. Liu, J.-J. Zhu, W. Liu, X.-G. He, X.-J. Li, F. Liang, J.-P. Liu, L.-Q. Zhang, H. Yang, Y.-T. Zhang, G.-T. Du, H. Long, and M. Li, “Analysis of localization effect in blue-violet light emitting InGaN/GaN multiple quantum wells with different well widths,” Chin. Phys. B 26, 017805 (2017). [CrossRef]
21. Y.-H. Kuo, Y. K. Lee, Y. Ge, S. Ren, J. E. Roth, T. I. Kamins, D. A. Miller, and J. S. Harris, “Strong quantum-confined stark effect in germanium quantum-well structures on silicon,” Nature 437, 1334 (2005). [CrossRef] [PubMed]
23. F. Wang, Z. Ji, Q. Wang, X. Wang, S. Qu, X. Xu, Y. Lv, and Z. Feng, “Green and blue emissions in phase-separated InGaN quantum wells,” J. Appl. Phys. 114, 163525 (2013). [CrossRef]
24. Y. P. Varshni, “Temperature dependence of the energy gap in semiconductors,” physica. 34, 149–154 (1967). [CrossRef]
25. Q. Li, S. Xu, M. Xie, and S. Tong, “Origin of the S-shaped temperature dependence of luminescent peaks from semiconductors,” J. Physics: Condens. Matter 17, 4853–4858 (2005).
26. Q. Li, S. Xu, M. Xie, and S. Tong, “A model for steady-state luminescence of localized-state ensemble,” EPL Europhys. Lett. 71, 994 (2005). [CrossRef]
27. P. G. Eliseev, P. Perlin, J. Lee, and M. Osiński, “blue temperature-induced shift and band-tail emission in InGaN-based light sources,” Appl. Phys. Lett. 71, 569–571 (1997). [CrossRef]
28. P. G. Eliseev, M. Osinski, J. Lee, T. Sugahara, and S. Sakai, “Band-tail model and temperature-induced blue-shift in photoluminescence spectra of InxGa1−xN grown on sapphire,” J. Electron. Mater. 29, 332–341 (2000). [CrossRef]