Temperature-Dependent Optical Behaviors and Demonstration of Carrier Localization in Polar and Semipolar AlGaN Multiple Quantum Wells

Abstract

Semipolar AlGaN multiple quantum wells (MQWs) have unique advantages in deep ultraviolet light emitters due to the weak Quantum-Confined Stark Effect. However, their applications are hampered by the poor crystalline quality of semipolar AlGaN thin films. Different treatments were developed to improve the crystal quality of semipolar AlGaN, including a multistep in situ thermal annealing technique proposed by our group. In this work, temperature-dependent and time-resolved photoluminescence characterizations were performed to reveal the carrier localization in the MQW region. The degree of carrier localization in semipolar AlGaN MQWs grown on top of the in situ-annealed AlN is similar to that of conventional ex situ face-to-face annealing, both of which are significantly stronger than that of the c-plane counterpart. Moreover, MQWs on in situ-annealed AlN show drastically reduced dislocation densities, demonstrating its great potential for the future development of high-efficiency optoelectronic devices.

1. Introduction

AlGaN-based multiple quantum wells (MQWs) with a high Al content are key building blocks in deep ultraviolet light-emitting diodes (DUV-LEDs), which have great potential for applications in water purification, environmental protection, medical instrumentation, non-line-of-sight communications, etc. [1,2]. Almost all commercial DUV-LEDs reported so far are grown on c-plane sapphire substrates due to better crystalline quality, smoother surface morphology, and cheaper substrate prices. However, traditional c-plane MQWs suffer from strong polarization fields that result in the spatial separation of the electrons and holes, and herein a reduction in the quantum efficiency of the device, known as the Quantum-Confined Stark Effect (QCSE) [3,4]. This disadvantage can be mitigated by growing III-nitride thin films with semipolar or nonpolar orientations [5], whose spontaneous polarization field perpendicular to the quantum well is strongly reduced or fully eliminated. Unfortunately, the development of semipolar or nonpolar DUV-LED is still in its embryonic stage due to the high density of dislocations and stacking faults. Among various methods to improve the crystalline quality of semipolar or nonpolar AlN and AlGaN [6,7,8], high-temperature annealing (HTA) has emerged as a promising technique owing to its simple process steps and efficient defect annihilation capability [9,10,11,12]. During the thermal annealing process, columnar AlN with small domain boundaries in the as-prepared films coalesced into a uniform thin film with reduced mosaicity and defect densities. However, the ex situ annealing technique poses a challenge to the subsequent LED growth because large residual strains are built up. Recently, our group developed an in situ treatment technique that greatly improves the crystal quality while relaxing the strain on the semipolar AlN template, ultimately achieving a strong enhancement of the luminescence intensity [13]. The in situ treatment technique proposed is proven to be an efficient method for the preparation of high-quality semipolar AlN, showing great potential toward the realization of high-efficiency optoelectronic devices. An in-depth understanding of the optical behaviors of the semipolar DUV-LED grown on the above-mentioned in situ-annealed AlN template is strongly desired.

Optical behaviors are critically important to understand the carrier localization effect, which has long been responsible for the improved recombination efficiency in InGaN-based blue LED despite its high density of threading dislocation. Carriers can be isolated from defect-related nonradiative recombination sites due to indium clustering [14,15], random alloy fluctuation [16], and variation in QW width [17], thus drastically improving the internal quantum efficiency (IQE). Experimentally, a unique S-shaped curve of peak position versus temperature strongly indicates carrier localization and is commonly observed in InGaN material [18,19]. The S-shaped curve refers to the red–blue–red shift of QW emission energy with increasing temperature. The initial red shift is due to carrier redistribution to deeper localized states via a hopping process. With increasing temperature, the hot carriers delocalize to higher energy states, giving rise to a blue shift in peak energy. The subsequent red shift at higher temperatures is due to temperature-induced band gap shrinkage [20]. Carrier localizations are also observed in AlGaN alloys, even though the effect might be much weaker due to fully miscible Al and Ga atoms [21,22,23]. It has been reported that the strength of localization increases with Al content increasing up to 70% and then gradually decreasing. Meanwhile, the carrier localization also depends on other factors, such as composition uniformity and quantum well/barrier width [21,24,25]. However, the carrier localizations of AlGaN systems are less studied and the carrier localization behavior in AlGaN is still vague compared to InGaN. In semipolar MQWs, there are only a few reports about carrier localization behaviors. Morteza et al. reported the existence of exciton localization in (11$\stackrel{-}{2}$2)-oriented semipolar InGaN MQWs by analyzing PL peak intensity and decay times as a function of temperature and demonstrated that exciton localization benefits IQE improvement [26]. Dinh et al. compared the localization depth in the polar (0001) and semipolar (11$\stackrel{-}{2}$2) In_0.2Ga_0.8N MQWs using temperature-dependent photoluminescence (TD-PL), claiming that semipolar InGaN MQWs exhibit a stronger exciton localization compared to c-plane counterparts, which can be attributed to the anisotropic growth and the unequal distribution of dislocations in (11$\stackrel{-}{2}$2) III-nitrides [27]. Zhang et al. reported that (11$\stackrel{-}{2}$2) QWs exhibit greater localization depth than polar QWs by means of both photoluminescence excitation and time-resolved PL measurements [28]. Even though most of the reports show that the semipolar MQWs have stronger localization than the c-plane MQWs in InGaN, it remains in debate whether the situation is the same in AlGaN.

In this work, TD-PL investigations of (11$\stackrel{-}{2}$2) semipolar AlGaN MQWs grown on AlN templates subject to various thermal treatment techniques were carried out in order to clarify the underlying localized states in semipolar AlGaN MQWs. The ex situ HTA and in situ-treated semipolar AlN templates as well as the c-plane AlN template were used to grow MQWs of the same structure. Peak positions were analyzed by Varshni fitting, and the physical mechanism was revealed by comprehensively analyzing the changes in emission energy, intensity, and linewidth. Time-resolved photoluminescence (TRPL) investigations were also performed to support the conclusion obtained by TD-PL.

2. Sample Structure and Experiments

The epitaxy of UV-LED starts with AlN template. Three different AlN templates were prepared, i.e., the unannealed c-plane AlN template grown on c-sapphire, semipolar AlN template grown on m-sapphire subject to conventionally ex situ face-to-face annealing, and last of all, semipolar AlN subject to in situ treatment in Metal–Organic Chemical Vapor Deposition (MOCVD) chamber. Specifically, the in situ treatment, i.e., a multistep in situ interface modification technique, was conducted under a temperature of 1350 °C for half an hour in the NH₃ ambiance with a reactor pressure of 400 Torr, in which the flow rate of NH₃ was 1050 µmol/min, while the ex situ face-to-face annealing was carried out at 1700 °C for 3 h in N₂ ambiance with a flow rate of 700 µmol/min. All the AlN layers were grown in MOCVD with 30 kPa, a TMAl flow rate of 34 µmol/min, and an NH₃ flow rate of 850 µmol/min. DUV-LEDs were grown on different AlN templates simultaneously. Figure 1 shows the schematic structure of the samples evaluated in this study. The epitaxial structure on top of the AlN template consists of 200 nm regrown AlN, 1.5 µm thick unintentionally doped (uid) AlGaN layer with stepped composition of 80%, 70%, and 60%; 1.05 µm thick Si-doped n-AlGaN; 5 pairs of Al_0.55Ga_0.45N/Al_0.4Ga_0.6N MQWs with thickness of quantum well and barrier being 2 and 8 nm, respectively; 100 nm thick Mg-doped p-Al_0.6Ga_0.4N/p-Al_0.4Ga_0.6N layer; and 30 nm thick p-GaN cap layer. The structures were grown in an AMEC HiT3 MOCVD reactor. Trimethylaluminum (TMA), triethylgallium (TMG), and ammonia (NH₃) were used as precursors of Al, Ga, and N, respectively. It should be noted that for different layers in UV-LED, the flow rates of TMA and TMG were different, and the flow rate in this structure varied between 1 and 10 µmol/min. Meanwhile, SiH₄ and Cp₂Mg were used as precursors of Si and Mg. The n-AlGaN layers were doped with Si to a concentration of 2 × 10¹⁹ cm⁻³, and the concentration of Mg for p-AlGaN layers was 5 × 10¹⁹ cm⁻³. Crystal orientation, in-plane anisotropy of crystal grains, and dislocation densities of the epitaxial layer were characterized using a Bruker D8 Discover High-Resolution X-ray Diffraction (HRXRD) instrument with a Cu Kα 1 (λ = 0.15056 nm) source. The rocking curve scan was performed using an open detector in a double-axis configuration with a four-bounce monochromator in the incidence beam. For the TD-PL measurements, a Coherent Ar-F (193 nm) excimer laser was used to excite the sample in nonresonant mode. The average excitation power density was 0.25 W/cm². The sample was mounted in a helium closed-circuit cryostat, and the temperature varied from 20 to 280 K. The PL signal from the sample was dispersed by a Horiba iHR550 monochromator with 1200 lines/mm grating and detected by a thermoelectrically cooled Synapse CCD detector under room temperature (RT). TRPL was performed by FLS1000 laser and recorded by a Time-Correlated Single-Photon Counting System. Time-resolved PL (TRPL) experiments were performed with a Ti: sapphire laser (Spectra Physics MaiTai HP). A fourth harmonic generator was used to excite the samples with a pulse width of 305 fs and pulse repetition rate of 80 MHz. The time transients were recorded by time-correlated single-photon counting (TCSPC), integrated over the full MQW emission spectrum.

3. Results and Discussion

Low-temperature PL characterization was conducted at 20 K to suppress the influence of nonradiative recombination centers on the optical properties of the LEDs. The PL spectra are presented in Figure 2a. Similar emission profiles are observed for two AlGaN MQWs grown on semipolar AlN templates. The emission wavelengths are 293.8 nm and 292.6 nm for the samples grown on ex situ-annealed AlN and in situ-treated AlN, respectively. In comparison, the c-plane sample exhibits a much shorter emission wavelength of 288.1 nm and a broader spectrum width. Compared to the c-plane sample, semipolar MQWs are more likely to incorporate Ga atoms under the same growth condition due to a smaller Ga diffusion length and higher incorporation rate [7]. As a result, a red shift of peak emission in semipolar MQW is expected. Similar full width at half maximum (FWHM) values of 7.27 nm and 6.90 nm were calculated for two semipolar MQWs grown on ex situ and in situ AlN, respectively; meanwhile, an FWHM value of 13.35 nm was identified for the c-plane counterpart.

Unlike c-plane III-nitride, the anisotropic characteristic of the crystalline quality of semipolar AlN epitaxial layer is a variable that might influence the optical behaviors of DUV-LED and thus was revealed by HRXRD rocking curves (RCs) of (11$\stackrel{-}{2}$2) reflections of AlN at multiple azimuth angles. As shown in Figure 2b, the azimuth angle is defined as 0° when the projections of the incident and diffracted beams are along the AlN [11$\stackrel{-}{2}$3] direction, and 90° refers to the direction along the AlN [1$\stackrel{-}{1}$00] direction. Overall, the in situ thermal-treated sample shows much narrower XRD RC linewidths, suggesting a lower defect density and better crystal quality than the HTA sample. The efficient blocking and annihilation of stacking faults in the in situ-treated sample can be ascribed to the interface modification process as previously demonstrated in our work [13]. Specifically, the stacking faults were efficiently blocked due to the modification of atomic configurations at the related interfaces. Coherently regrown AlGaN layers were obtained on the in situ-treated AlN template, and stacking faults were eliminated in the postgrown AlGaN layers. At the same time, the strains between the AlGaN layers were relaxed through a dislocation glide in the basal plane and misfit dislocations at the heterointerfaces; thus, a better stress condition was obtained compared with the ex situ HTA sample. Despite that, as marked in Figure 2b, the crystal quality of semipolar AlN is still inferior to the c-plane AlN counterpart. Moreover, the 2θ-ω XRD scans of two semipolar samples are presented in Figure 2c. The peaks of the n-AlGaN and MQWs are largely overlapping, indicating the Al components of two samples are almost the same. In addition, compared to the ex situ HTA sample, the narrower linewidth and stronger intensity of the AlN and n-AlGaN peaks for the in situ-treated sample indicate a better crystal quality with the proposed thermal treatment technique. Notice in the graph that the strain-free position of the AlN (11$\stackrel{-}{2}$2) peak locates at 71.4° as shown by the black dashed line. Both samples exhibit tensile strains, but the strain level is smaller for the in situ-treated AlN, suggesting less influence from thermal mismatch and less likelihood to form cracks.

TD-PL is commonly used to characterize the optical properties of semiconductor devices, such as carrier transport dynamics and localized states, and the defect-related nonradiative processes in the MQWs were thoroughly investigated in previous work [29,30,31,32]. The TD-PL emission spectra of all three MQWs with temperatures ranging from 20 to 280 K are plotted in Figure 3.

With increasing temperature, integrated PL intensity dramatically decreases owing to the dominant contribution of the nonradiative recombination rate whose decay time becomes smaller. Figure 4a shows the Arrhenius plot of the integrated PL intensity of three MQW samples. The integrated PL intensity drops slowly with temperature during the low-temperature range, while decreasing more rapidly during the high-temperature range for the two semipolar samples. This behavior suggests two nonradiative recombination centers, corresponding to two different activation energies at these different temperature regions. The PL intensity variations cannot be described by a thermal quenching process with single activation energy. Therefore, two different thermal quenching channels with two activation energies are introduced as the following equation:

$$I\left(T\right)=\frac{1}{1+A\mathit{exp}(-\raisebox{1ex}{$E_a$}\!\left/ \!\raisebox{-1ex}{$k_{B}T$}\right.)+B\mathit{exp}(-\raisebox{1ex}{$E_b$}\!\left/ \!\raisebox{-1ex}{$k_{B}T$}\right.)}$$

(1)

where I(T) represents the normalized integrated PL intensity. The parameters A and B are two constants corresponding to the density of the nonradiative recombination center in the samples, and k_B is the Boltzmann constant. E_a and E_b are the activation energies corresponding to the nonradiative recombination process. Note that the two activation energies are different; the greater one E_a is considered as the potential barrier between the localized potential minimum and the nonradiative centers (probably due to the high dislocation density) located inside the MQWs, while the smaller one E_b can be attributed to the localized exciton binding energy or carrier localization energy [33,34]. At lower temperatures, the nonradiative recombination centers are frozen and inactive. The carriers are easily captured and restricted by the localized states. Contrastively, at higher temperatures, the carriers in localized states and nonradiative centers are thermally activated. The thermally activated carriers can be trapped by the nonradiative centers, leading to the nonradiative recombination process. All fitting values are listed and shown in

Table 1. Fitting parameters: activation energies (E_a and E_b) and constants (A and B).

Sample	A	Ea (meV)	B	Eb (meV)
Ex situ HTA	207.4	59.3	11.4	17.8
In situ-treated	184.9	69.4	11.9	16.8
c-plane	98.2	53.2	1.1	8.9

. The difference in the fitting constants A and B implies the difference in the crystal quality of the MQWs; that is, the quality of the in situ-treated sample is better than that of the ex situ HTA sample, while the c-plane sample is the best. In addition, the greater thermal activation energy E_a is quite similar among all three samples, which can be attributed to the similar character of the dislocations in all samples. The E_a values are around 50–70 meV, which are comparable with those reported for AlGaN [35,36,37]. Meanwhile, the activation energies E_b for two semipolar MQW samples are close to each other, revealing a comparable localization strength between them. The activation energies E_b are greater for two semipolar samples compared to the c-plane counterpart, suggesting a stronger localized strength. In addition, Figure 4b presents the FWHM of the PL emission peak of AlGaN MQWs at various temperatures. The FWHM of the in situ-treated semipolar MQW is the narrowest among all three samples throughout the temperature range, indicating a more uniformly distributed localized state and better optical properties for the development of high-efficiency DUV-LED. Note that a “U-shaped” temperature-dependent FWHM curve is generally observed for both resonant and nonresonant excitation conditions in the InGaN system. However, in our case, FWHM increases monotonically with increasing temperature. This can be interpreted that at a low temperature, for AlGaN systems with a generally weaker localized energy state, the narrowing of the FWHM value with increasing temperature due to the redistribution of carriers in the local state seems to be insignificant. In the moderate-temperature region (80–200 K), the enhanced carrier mobility results in a nonuniform redistribution of the carriers, followed by a slight broadening of the linewidth due to a wide dispersion of the carriers. The drastic broadening of the FWHM in the high-temperature region (200–280 K) is caused by the coupling between the carrier and acoustic phonons/longitudinal optical (LO) phonons as well as the scattering from ionized impurities and defects [38]. The primary difference in the FWHM variations among the three samples is in the high-temperature region, where the c-plane sample presents the largest broadening due to the strongest coupling strength between the carriers to acoustic phonons and LO phonons. Finally, compared with the in situ-treated sample, the ex situ HTA semipolar sample reveals a more rapid broadening, which can possibly be explained by its higher dislocation density and thus more defect-related scattering.

Through Gaussian fitting of the PL spectra, the variations of peak position as a function of temperature were calculated. As shown in Figure 5, the points present the relative shift of the peak position with respect to that at 20 K. A general trend of the red shift is revealed as the temperature increases. Due to the low carrier localization strength in AlGaN materials, the “S-shaped curve” is too weak to be revealed. The peak shift is fitted by the modified Varshni equation, which can be expressed as below:

$$E\left(T\right)=E_0-\frac{\alpha T^2}{T+\beta }-\frac{{\sigma }^2}{k_{B}T},$$

(2)

where E₀ is the band gap energy of AlGaN at 0 K, α and the characteristic Debye temperature β are the Varshni coefficients, which are strongly correlated with the specific material system, k_B is the Boltzmann constant, σ is the standard deviation of the Gaussian distribution of the band gap fluctuations due to the Al content and width of the quantum well [39]. In general, a greater σ value corresponds to a stronger localization effect. The reasonable fitting in the MQW emissions can be obtained with α = (1.1 − 1.8) × 10⁻³ eV/K and β = 1200–1800 K, and the fitting results of the localization parameter σ are also shown in Figure 5. The localization strength is 13.48 meV, 14.15 meV, and 5.41 meV for MQWs grown on the ex situ HTA AlN, in situ-treated AlN, and c-plane AlN, respectively. Stronger localization strengths of semipolar MQWs than the c-plane MQWs are revealed. Meanwhile, compared to the MQWs grown on the ex situ HTA AlN template, the MQWs with the in situ-treated AlN template have a comparable localization effect but exhibit relatively better crystal quality than the ex situ one. The in situ thermal treatment demonstrates itself as a cost-effective approach for the realization of carrier localization and thus better optical properties.

Finally, TRPL measurements were performed on all samples at room temperature to investigate the carrier dynamic properties in AlGaN MQWs. A monoexponential model was typically employed to fit the decay trace of AlGaN MQWs, and decay times were determined in the same way, i.e., by an exponential fit of the initial part of the intensity decay to 1/e of the maximum intensity. The extracted lifetimes are indicated in Figure 6. It is worth noting that there is a faint secondary excitation peak shown at 1.25 ns due to laser reflection at the sample surface/interface, but it does not interfere with the fitting process. It is clearly shown that semipolar MQWs exhibit much faster decay rates, which are attributed to a reduction in QCSE with increased wave function overlap, allowing an enhanced recombination rate [40]. At the same time, the two semipolar samples have similar decay lifetimes, which are comparable with that of the laser excitation pulse, implying that the actual decay lifetimes of the semipolar samples might be even shorter.

4. Conclusions

In summary, TD-PL of the semipolar (11$\stackrel{-}{2}$2) and c-plane AlGaN MQWs grown by MOCVD was investigated in the temperature range from 20 K to 280 K. The peak positions are directly following the thermal redistribution of carriers among the potential landscape with the Varshni empirical equation. The localization strength of the semipolar (11$\stackrel{-}{2}$2) AlGaN MQWs is considerably greater than the c-plane counterpart, which can be possibly correlated with the weaker polarization field. Furthermore, MQWs on in situ-annealed AlN template exhibit the narrowest emission linewidth due to better crystalline quality. The in situ-treated semipolar AlN template demonstrates itself as a promising material for the growth of semipolar MQWs. Our results provide deep insight into the optical properties and localization effect of semipolar AlGaN MQWs, which is of significant importance to the development of high-efficiency semipolar AlGaN-based DUV-LED.