Strain-related recombination mechanisms in polar InGaN/GaN MQWs on amorphous Si<sub>x</sub>C<sub>1-x</sub> buffers

Tao Lin; Fangze Wang; Chih-Hsien Cheng; Shuai Chen; Zhe Chuan Feng; Gong-Ru Lin

doi:10.1364/OME.8.001100

1. Introduction

InGaN/GaN multiple quantum well (MQW) light-emitting diodes (LEDs) have attracted widespread attention for their potential applications in semiconductor laser devices, solid-state lighting and panel displays. To further enhance the device performances, it is crucial to understand the intrinsic photoluminescence (PL) dynamics and carrier recombination mechanisms as well as how they determine internal quantum efficiency (IQE). Particularly, most of present MQW LED structures are fabricated on substrates like sapphire, SiC, ZnO or crystalline silicon, all of which only allow polar orientation growth of GaN layers along c-axis. Thus, strain-induced piezoelectric fields strongly influence the carrier recombination processes, which is recognized as quantum confinement Stack effect (QCSE), making PL dynamics of MQW very complicated. Examination to these PL characters is ordinarily done by time-resolved (TR) PL spectrum measurements, which directly reveal the decay of PL that provides profound understanding to the recombination processes of carriers. There are generally two categories of mechanisms for MQW LEDs, which lead to totally different interpretations of the TRPL decay results: One is based on localized exciton recombination [1, 2] meanwhile another is based on bimolecular recombination [3, 4]. Exciton localization model was originally hypothesized for explaining the extraordinarily high IQE of blue MQW LEDs on sapphire substrates with high dislocation defect densities [5, 6]. The PL decay was described as $I (t) / I_{0} \propto k_{l o c} N \propto - d N / d t$ , in which $k_{l o c}$ represented exciton localization rate and $N$ represents carrier concentration [2]. A single-exponential decay of $I (t)$ was then deduced from this dynamic function. Unfortunately, in most case the as-measured PL decays were found to deviate from single-exponential decay [7–10]. Thus, various additional hypothesises that ascribing these to the complexity of recombination paths, and different modified fitting models, like “pseudo-DAP” recombination [11], bi-exponential decay model [12] and stretched-exponential decay model [13] were proposed to better describe the PL decay data in different works, but most of them were only appropriate for their specific fabrication conditions. Unlikely, someone used a totally different dynamic function so-called ABC model [14] to interpret the carrier recombination processes. In this model, total recombination rate $k = A N + B N^{2} + C N^{3}$ and $I (t) / I_{0} \propto k_{r} N^{2}$ , which provided a possible explanation to non-exponential decay of $I (t)$ . But these definitions imply that the radiative recombination process, which is called bimolecular recombination, occurs only between completely uncorrelated electrons and holes, thus ignoring carrier localization and the possibility of excitonic recombination [15], which means the QCSE effect shall strong enough to spatially isolate the electrons and holes from forming excitons. In another word, it has difficulty in explaining the high IQE of MQW LED on polar substrate. Base on the above, there is still a controversial debate on the exact mechanism behind the non-exponential decay nature.

In our previous work [16], MQW LED structures were successfully grown on amorphous SiC buffers with variable C/Si compositions. The devices showed bright electroluminescence (EL) and external quantum efficiencies (EQEs) as high as 42.3%. Strain relaxation led by increasing C composition was proved to be the main reason for EQE enhancement. This provided a group of excellent objects for us to investigate the real recombination mechanisms in polar MQW LEDs in different strain conditions, because the average lattice constant of SiC buffer, which sensitively determined strain in GaN/SiC interface, could be isolatedly changed by controlling C/Si ratio that cannot be realized by using versatile substrates as sapphire. Therefore, in this work, PL decays of three InGaN/GaN MQW LEDs grown on Si_xC_1-x/SiO₂/Si substrates with different x were measured for exploring the strain-related carrier recombination mechanism in polar MQWs. The simplest fitting model was used to extract the PL lifetimes which associating with full-scale PL dynamic information in MQWs. Then relative photoluminescence (PL) efficiencies and recombination lifetimes (rates) for each sample were extracted from temperature-varied steady-state (SS) PL spectra and time-resolved (TR) PL spectra. The results turned out that, strain releasing process suppressed the occurrences of nonradiative recombination, which accorded with room temperature EL and IV results. But it surprisingly slowed down the radiative recombination. This was attributed to the suppressed formations of fast localized recombination centers related to well thickness variations.

2. Experimental

The fabrication and characterization methods of InGaN/GaN MQW LED on Si_xC_1-x/SiO₂/Si substrates have been described previously [16]. For buffered substrates preparation, 500-nm Si_xC_1−x film with different C/Si composition ratios was grown upon the 1-μm thermal SiO₂ coated Si substrate by using a plasma enhanced chemical vapor deposition (PECVD) system. C/Si composition ratios in as-prepared thin films were controlled by fluence ratio of gas-phased precursors and examined by X-ray photoelectron spectroscopy (XPS). Three substrate samples with x = 0.35, 0.49 and 0.66 were tested in this work, which were named as “C-rich SiC”, “stoichiometric SiC” and “Si-rich SiC”, respectively. Then similar InGaN/GaN MQW LED structures were grown on these three Si_xC_1−x/SiO₂/Si substrates by using a commercial metal organic chemical vapor deposition (MOCVD) system, which consisted of 2-μm n-type GaN bottom layers, 0.2-μm InGaN/GaN MQW active layers (10 periods with 4-nm i-In_0.18Ga_0.82N and 16-nm i-GaN for each), and 0.3-μm p-type GaN top layers.

A Zolix-750 system was used for PL spectra tests, in which a 375 nm Picoquant pulsed laser with pulse width less than 300 ps worked as the excitation light source, and an Andor Newton CCD with 0.09 nm resolution worked as the photodetector. In TRPL measurements, the PL decays were recorded by a time-correlated single-photon counting system in 10-300 K. The excitation output power was fixed at 6.32 mW in all PL experiments.

3. Results and discussion

As seen in the SSPL spectrum for each tested MQW sample at 13K (Fig. 1(a)), a clean and sharp PL peak centered at blue range of 2.80~2.83 eV with ~0.06 eV full width at half maximum (FWHM) was excited by 375 nm (3.31 eV) laser beam. It is generally accepted that this main peak originates from radiative recombination on InGaN QWs, while the series located at lower energy side with an interval of ~90 meV are due to its longitudinal optical (LO) phonon replicas [17]. The blue PL peak tends to slightly blueshift with increasing C composition. Considering that strain-induced QCSE results in redshift of PL position, this phenomenon indicates the strain releasing effects. SSPL intensities tend to decrease monotonously following the testing temperature increasing toward 300 K for the defreezing of nonradiative recombination, as well as respective SSPL position shows slight redshift on account of bandgap shrinkage with increasing temperature (Fig. 1(b) is the SSPL for stoichiometric SiC sample at different temperature as an example). Based on these observations, the detected wavelength (photon energy) in subsequent temperature-varied TRPL measurement was kept at each peak maximum varying with temperature for accuracy.

Fig. 1 (a) SSPL spectra for InGaN/GaN MQW LEDs on Si_xC_1-x/SiO₂/Si with different C/Si ratio excited by a 375-nm laser and tested at 13 K. (b). Evolution of SSPL spectrum with temperature for InGaN/GaN MQW LEDs on stoichiometric SiC buffer. (c). Evolution of TRPL spectrum with temperature for InGaN/GaN MQW LEDs on stoichiometric SiC buffer. The detected photon energy was kept at each maximum that shown in (b). (d) Single-exponential decay fitting for Stoichiometric SiC sample at 13K and 300K.

Download Full Size | PDF

It is critical to test the PL decay properties of MQW structures for understanding their recombination nature. As shown in Fig. 1(c), the PL intensity of MQW on SiC sample (Stoichiometric SiC as an example) decays continuously faster with increasing temperature. All of the decay curves tend to deviate from single-exponential, and this trend becomes stronger with temperature rising. ABC-Model-based equation provided by Xing et al [3]. was firstly tried to fit the non-exponential decaying data. The results turned to be illogical as that obtained non-radiative recombination rate values were negative in most cases. Meanwhile, the obtained radiative recombination rates tended to continuously increase with rising temperature. Both of these results were against general SSPL observations, which indicated that the prerequisite for ABC model, electrons and holes were spatially isolated without formation of excitons and the whole PL originated from bi-molecular recombination, was not satisfied in these cases. Therefore, exciton localization effect was taken into our consideration.

Here, we used a single-exponential decay function to approximatively fit the data (as shown in Fig. 1(d)), in which we defined the average PL lifetime as the time period during which the PL intensity reduced from beginning to 1/e of initial intensity for each curve. This treatment was for avoiding introducing any additional assumption to the recombination behaviors. The average PL lifetime value provides an overall evaluation to the contribution from all kinds of recombination paths in the system. The obtained lifetime values of three different samples are collected in Fig. 2(a). Then total recombination rate $k$ is defined as $k = 1 / τ$ , which connects the PL dynamic results to the recombination probabilities. Spots of respective $k$ versus temperature are also shown in Fig. 2(b). It is seen that evolutions of $k$ with rising temperature are not monotonous, especially the one for Si rich sample. Inflexions appear at low temperature regions of 25 K~50 K. Note that $k = k_{r} + k_{n r}$ , in which $k_{r}$ represents radiative recombination rate and $k_{n r}$ represents nonradiative recombination rate. It is expected that $k_{n r}$ increases with temperature rises as it always relates to energy exchange with heat. So, the $k$ results at high temperature side indicate that growth of C composition, or the strain releasing process, has positive influences on suppressing nonradiative recombination in MQWs throughout reducing dislocation defects or cracks that have major influence on $k_{n r}$ . This result well accords with the previous observations of structure, EL and IV measurements at room temperature [16]. However, non-radiative recombination is frozen and $k_{r}$ becomes nonnegligible on low temperature condition. So, the bending of $k$ at low temperature side definitely relates to increasing $k_{r}$ with decreasing temperature. According to this discussion, the $k$ results at low temperature side imply a conclusion that may overrides our expectation: the MQW structure with stronger strain and lower crystalline quality has higher radiative recombination rate at low temperature. But this assertion is not solid enough at present because $k_{n r}$ is not completely negligible even at the temperature range as low as 13 K. Therefore, additional information and further analysis are needed to explain the behavior of $k_{r}$ at low temperature side.

Fig. 2 (a) Average PL lifetimes $τ$ , (b) total recombination rates, and (c) relative PL efficiencies for InGaN/GaN MQW LEDs on Si_xC_1-x/SiO₂/Si with different C/Si ratio, in which $τ$ was defined as the time period during which the PL intensity reduced from beginning to 1/e of initial, $k = 1 / τ$ and $η = I (T) / I_{0}$ obtained from SSPL. (d) Demonstration of total recombination rate $k$ , radiative recombination $k_{r}$ and nonradiative recombination $k_{n r}$ for MQW sample on C-rich SiC buffer, in which $k = k_{r} + k_{n r}$ .

Download Full Size | PDF

SSPL spectrum on various temperature condition for each sample was applied for further splitting $k_{r}$ and $k_{n r}$ from each $k$ value. The intensity for each PL maximum corresponding to detected wavelengths on previous TRPL tests were recorded as I(T). After that, relative PL efficiency was defined as $η = I (T) / I_{0}$ , in which $I_{0}$ represented PL intensity at 0 K. The obtained PL efficiencies for three samples were put together in Fig. 2 (c). All three efficiencies are nearly exponential to temperature, which are normally described by Arrhenius equation. The slope of $ln η$ reflects the activation energy needed for a nonradiative recombination process. It can be estimated that the activation energy increases, and the chances for occurrence of nonradiative recombination decrease with growing C composition. It is known that only radiative recombination contributes to PL intensity, therefore this relative PL efficiency reflects the ratio of radiative recombination rate in total recombination rate:

η = k_{r} / (k_{r} + k_{n r}) = k_{r} / k .

Thus, it is capable to resolve

k_{r} = k η

and

k_{n r} = k (1 - η)

from the TRPL results combined with

η

. The respective demonstration of

k_{r}

and

k_{n r}

for C-rich SiC sample was shown in Fig. 2(d) as an example. The results turned out that even for MQW LEDs on C-rich SiC buffer, which have best GaN crystal quality and smallest strain in the three samples, nonradiative recombination rate is larger than radiative recombination rate until reaching a very low temperature down to 50 K. Their difference becomes larger with increasing temperature and

k_{n r}

is two orders of magnitude larger than

k_{r}

at 300 K.

Based on the above calculation, $k_{r}$ and $k_{n r}$ versus temperature for all three samples were summarized in Fig. 3(a) and (b), respectively. First of all, the low temperature behavior of $k_{r}$ (marked in dash circle in Fig. 3(a)) doubtlessly verifies the previous hypothesis: stronger strain and lower crystalline quality lead to higher radiative recombination rate at low temperature. Secondary, All the three radiative recombination rates $k_{r}$ are found decline monotonously with growing temperature, which does not agree with typical radiative recombination behavior of free electron-hole pair recombination. In that case, $k_{r}$ shall be free from temperature or increase with temperature. However, it is reasonable if the PL process is dominant by localized exciton recombination: In this case, free electron-hole pairs tend to combine into excitons in very high rate. Then these free excitons are captured by localized states in a localization rate $k_{l o c}$ . The subsequent recombination of localized excitons is very fast, therefore the observed recombination rate $k_{r}$ is mainly determined by exciton localization rate $k_{l o c}$ . Here, excitons have higher opportunity to delocalize in higher temperature range, which leads to the decline of localization rate [1]. It is found that the decline of $k_{l o c}$ for Si-rich sample with growing temperature is relatively steeper than the ones for stoichiometric SiC or C-rich SiC. This indicates that its average depth of localized states is relatively smaller than the other ones, making the exciton easier to delocalize (There is positive correlation between the activation energy for delocalization processes and the average depth of localized states). In other words, the average depths of localized states in samples without strain releasing are smaller than the ones with strain releasing. Based on the previous works [18], the localized radiative recombination centers in InGaN/GaN MQWs often originate from structural defects in InGaN well layers, like well thickness variations and indium rich clusters, in which well thickness variations offer shallow states that lead to fast decays as well as indium rich clusters offer deeper states with slower decays [18]. Here the results of $k_{r}$ can be ascribed as that strong strain on MQW interfaces may improve the formation of radiative shallow structural defects, so the average depth of localized states for Si-rich sample is smaller and the average $k_{r}$ at low temperature is higher (relative radiative decays are faster). Unfortunately, this positive effect was not reflected in IQEs (see $η$ values in Fig. 2(c)) in high temperature range because the negative effects led by strain and increased nonradiative defect density are much more dominant to IQE: As mentioned above, $k_{n r}$ is two orders of magnitude larger than $k_{r}$ at 300 K, where $k_{r}$ for three samples descend toward the same level. On another hand, as seen in Fig. 3(b), $k_{n r}$ increases with growing Si composition, and this trend becomes stronger following the rising temperature (dash circle in Fig. 3(b)). Therefore, according to Eqs. (1), nonradiative recombination rate is the dominant factor for IQE, and IQE reduces with growing Si composition.

Fig. 3 (a) Radiative recombination $k_{r}$ and (b) nonradiative recombination $k_{n r}$ for InGaN/GaN MQW LEDs on Si_xC_1-x/SiO₂/Si with different C/Si ratio.

Download Full Size | PDF

4. Conclusion

In summary, PL decay properties for InGaN/GaN MQW LEDs grown on Si_xC_1-x/SiO₂/Si substrates were tested for investigating the recombination nature of polar MQW structures, in which the existing stain is straightly controlled by C/Si ratio of SiC buffers. The non-exponential-like PL decays were incapable to be described by ABC model based on bi-molecular recombination. Instead, simplest model based on exciton localization was used to evaluate full-scale properties of the PL dynamics. The obtained nonradiative recombination rate was found decrease with growing C composition that was ascribable to strain releasing process and suppression of nonradiative recombination in MQWs throughout reducing dislocation defects or cracks. On the contrary, the radiative recombination rate was found increase with Si composition which implied that stronger strain and lower crystalline quality led to higher radiative recombination rate especially at low temperature. This was attributed to the enhanced formations of fast localized recombination centers related to well thickness variations. All the above results indicate that ABC model based on bi-molecular recombination is not a general solution for polar MQW LEDs. The correlation of electron and hole cannot be ignored especially on the condition that numbers of structure-imperfect-related localized states existing in the system.

Funding

NSFC (61504030, 11533003, and U1731239); Guangxi Natural Science Foundation (2015GXNSFCA139007); Special Funding for Guangxi Distinguished Professors (Bagui Rencai & Bagui Xuezhe).

References and links

1. M. S. Minsky, S. Watanabe, and N. Yamada, “Radiative and nonradiative lifetimes in GaInN/GaN multiquantum wells,” J. Appl. Phys. 91(8), 5176–5181 (2002). [CrossRef]

2. S. Chichibu, T. Onuma, T. Sota, S. P. DenBaars, S. Nakamura, T. Kitamura, Y. Ishida, and H. Okumura, “Influence of InN mole fraction on the recombination processes of localized excitons in strained cubic InxGa1-xN/GaN multiple quantum wells,” J. Appl. Phys. 93(4), 2051–2054 (2003). [CrossRef]

3. Y. Xing, L. Wang, D. Yang, Z. Wang, Z. Hao, C. Sun, B. Xiong, Y. Luo, Y. Han, J. Wang, and H. Li, “A novel model on time-resolved photoluminescence measurements of polar InGaN/GaN multi-quantum-well structures,” Sci. Rep. 7, 45082 (2017). [CrossRef] [PubMed]

4. T. J. Badcock, M. Ali, T. Zhu, M. Pristovsek, R. A. Oliver, and A. J. Shields, “Radiative recombination mechanisms in polar and non-polar InGaN/GaN quantum well LED structures,” Appl. Phys. Lett. 109(15), 151110 (2016). [CrossRef]

5. H. Jeong, H. J. Jeong, H. M. Oh, C.-H. Hong, E.-K. Suh, G. Lerondel, and M. S. Jeong, “Carrier localization in In-rich InGaN/GaN multiple quantum wells for green light-emitting diodes,” Sci. Rep. 5(1), 9373 (2015). [CrossRef] [PubMed]

6. M. A. Sousa, T. C. Esteves, N. B. Sedrine, J. Rodrigues, M. B. Lourenço, A. Redondo-Cubero, E. Alves, K. P. O’Donnell, M. Bockowski, C. Wetzel, M. R. Correia, K. Lorenz, and T. Monteiro, “Luminescence studies on green emitting InGaN/GaN MQWs implanted with nitrogen,” Sci. Rep. 5(1), 9703 (2015). [CrossRef] [PubMed]

7. T. H. Ngo, B. Gil, P. Valvin, B. Damilano, K. Lekhal, and P. D. Mierry, “Internal quantum efficiency in yellow-amber light emitting AlGaN-InGaN-GaN heterostructures,” Appl. Phys. Lett. 107(12), 122103 (2015). [CrossRef]

8. P. Dawson, S. Schulz, R. A. Oliver, M. J. Kappers, and C. J. Humphreys, “The nature of carrier localisation in polar and nonpolar InGaN/GaN quantum wells,” J. Appl. Phys. 119(18), 181505 (2016). [CrossRef]

9. J. Wang, L. Wang, W. Zhao, Z. Hao, and Y. Luo, “Understanding efficiency droop effect in InGaN/GaN multiple-quantum-well blue light-emitting diodes with different degree of carrier localization,” Appl. Phys. Lett. 97(20), 201112 (2010). [CrossRef]

10. H. Fu, Z. Lu, and Y. Zhao, “Analysis of low efficiency droop of semipolar InGaN quantum well light-emitting diodes by modified rate equation with weak phase-space filling effect,” AIP Adv. 6(6), 065013 (2016). [CrossRef]

11. A. Morel, P. Lefebvre, S. Kalliakos, T. Taliercio, T. Bretagnon, and B. Gil, “Donor-acceptor-like behavior of electron-hole pair recombinations in low-dimensional (Ga,In)N/GaN systems,” Phys. Rev. B 68(4), 045331 (2003). [CrossRef]

12. Y. Iwata, R. G. Banal, S. Ichikawa, M. Funato, and Y. Kawakami, “Emission mechanisms in Al-rich AlGaN/AlN quantum wells assessed by excitation power dependent photoluminescence spectroscopy,” J. Appl. Phys. 117(7), 075701 (2015). [CrossRef]

13. D. C. Johnston, “Stretched exponential relaxation arising from a continuous sum of exponential decays,” Phys. Rev. B 74(18), 184430 (2006). [CrossRef]

14. S. Karpov, “ABC-model for interpretation of internal quantum efficiency and its droop in III-nitride LEDs: a review,” Opt. Quantum Electron. 47(6), 1293–1303 (2015). [CrossRef]

15. T. Langer, A. Chernikov, D. Kalincev, M. Gerhard, H. Bremers, U. Rossow, M. Koch, and A. Hangleiter, “Room temperature excitonic recombination in GaInN/GaN quantum wells,” Appl. Phys. Lett. 103(20), 202106 (2013). [CrossRef]

16. C.-H. Cheng, A.-J. Tzou, J.-H. Chang, Y.-C. Chi, Y.-H. Lin, M.-H. Shih, C.-K. Lee, C.-I. Wu, H.-C. Kuo, C.-Y. Chang, and G.-R. Lin, “Growing GaN LEDs on amorphous SiC buffer with variable C/Si compositions,” Sci. Rep. 6(1), 19757 (2016). [CrossRef] [PubMed]

17. D. M. Graham, A. Soltani-Vala, P. Dawson, M. J. Godfrey, T. M. Smeeton, J. S. Barnard, M. J. Kappers, C. J. Humphreys, and E. J. Thrush, “Optical and microstructural studies of InGaN/GaN single-quantum-well structures,” J. Appl. Phys. 97(10), 103508 (2005). [CrossRef]

18. Z. Li, J. Kang, B. W. Wang, H. Li, Y. H. Weng, Y.-C. Lee, Z. Liu, X. Yi, Z. C. Feng, and G. Wang, “Two distinct carrier localization in green light-emitting diodes with InGaN/GaN multiple quantum wells,” J. Appl. Phys. 115(8), 083112 (2014). [CrossRef]

Strain-related recombination mechanisms in polar InGaN/GaN MQWs on amorphous Si_xC_1-x buffers

Abstract

1. Introduction

2. Experimental

3. Results and discussion

4. Conclusion

Funding

References and links

Cited By

Figures (3)

Equations (1)

Optical Materials Express