1. Introduction
Ultraviolet (UV) light emission and detection have been widely studied in recent years due to their varied applicability from engineering to healthcare[1−3]. Wide-bandgap materials of the III-N family are fundamental in the design and development of high-quality deep UV (DUV) devices. Aluminum-gallium nitride (AlGaN) has been extensively analyzed by epitaxial growth techniques such as metal−oxide chemical vapour deposition (MOCVD)[4], molecular beam epitaxy (MBE)[5, 6], and liquid phase epitaxy (LPE)[7]. This high-quality growth has allowed us to understand the feasibility of Ⅲ-N properties, such as superior transport properties[8, 9] and wide bandgap tunability[10, 11]. Despite indium nitride (InN) significant thermodynamical instability compared to aluminum nitride (AlN) or gallium nitride (GaN)[12], the ternary aluminum−indium−nitrogen alloy (AlInN) pointed to new approaches for controlling the electron leakage and overflowing due to the difference between p- and n-type carrier mobilities in light emitting diodes[13].
AlInN-based layers have been widely used in the recent past due to properties that benefit confinement and strain relaxation with reduced quantum confinement Stark effect. In this case, these characteristics have resulted useful substitute for the traditional AlGaN-based quantum barriers. It has been observed that p-type doping of AlInN layers provides lower internal quantum efficiency drop, lower electric field, and higher radiative recombination increasing the light output power[14]. Moreover, the carrier injection is improved due to the higher conduction band offset compared to AlGaN[10]. The use of thin layers based on AlInN has been studied pointing out that thinner layers assisted not only the blocking of electrons but also enhanced the hole injection through tunneling[15, 16].
Despite the AlInN-based electron blocking layers (EBL) has revealed an alternative to increasing the DUV-LED efficiency by increasing the confinement and enhancing the hole injection into the active region[14, 17], the high segregation to the surface[12] and interdiffusion[18] are current problems that hinder the proper incorporation of In-atoms, making it essential to overcome these disadvantages by avoiding high growth temperatures. Indium nitride (InN) requires lower growth temperatures than aluminum and gallium nitrides due to the higher atomic mobility of In atoms, the thermodynamical instability, and its trend to segregate to the surface[19]. The high growth temperature effect on the In incorporation is commonly addressed using epitaxial techniques at a lower substrate temperature, such as migration-enhanced epitaxy (MEE) and metal-modulated epitaxy (MME), which are appropriate for lower substrate temperatures. However, these lower growth temperatures result in less energy for the atoms to displace and locate into their minimum energy sites, reducing the migration distance of the group-Ⅲ atoms[20]. On the other hand, it is well known that AlN requires higher growth temperature than the other compounds and ternary alloys of the Ⅲ-N family, where MBE-based techniques allow incorporating high Al content, more controlled nucleation, and efficient dopant activation with reduced impurity incorporation[21]. The Al−N bonds' higher cohesive energy than Ga−N and In−N bonds[22−24] drives a remarkable consideration in the effect of the In-flux, by a preference during the epitaxial growth. An excessive In-flux results in Al-based droplet formation due to In atoms increasing the mean free path passivating the surface during the growth process, allowing Al atoms to reach their most energetically favorable sites in the lattice and making that the N free bonds become more saturated preventing N-atoms form Al−N[24].
In this work, the experimental growth of AlInN-based p-type layers has been performed using short-period MME to ensure incorporation into the alloy without forming metal clusters or phase separation. Experimental results were considered to design and simulate a DUV-LED structure based on the multi-ple quantum wells (MQWs) active region with an electron blocking layer (EBL) based on Mg-doped AlInN with a finite-element solution by SILVACO Atlas. The effect of AlInN-based EBL based on around 20% maximum of In-atoms incorporation during the experimental stage is described below.
2. Materials and methods
Silicon substrates present, in general, a large lattice- and thermal expansion mismatching, including the polarity difference with Ⅲ-N semiconductors, moreover the flat surface requirement of the silicon surface before MBE growth drives to carefully prepare Si-based substrates before epitaxy obtaining plane and clean surface[25]. Despite this, the Si-compatible devices oriented to large-scale integration and their low-cost applications maintain well-established Si-based wafer devices[26, 27]. In this study, samples were grown on Si(111) substrate and prepared according to Ref. [28].
Regarding the growth procedure, it was designed for a plasma-assisted MBE using an MME technique. MME technique exhibits good incorporation of metal by the shutter modulation open−close cycle, which has overcome the formation of high defective layers caused by the use of lower growth temperature[29]. The metal modulation was implemented with short time open−close cycles of shutters, close to the migration-enhanced epitaxy (MEE) approach with a constant plasma-assisted nitrogen atmosphere. For the chemical composition analysis of the layers, a thermo scientific system K-alpha X-ray photoelectron spectroscopy (XPS) equipment with Al Kα (1486 eV) as an X-ray source was used. XPS peak fitting and quantitative analysis were carried out using Thermo Avantage software. Measurements were calibrated using the adventitious carbon of the surface, C1s located at 284.8 eV; high-resolution XPS spectra were fitted using Shirley background subtraction to deconvolute and process core level peaks. A quantification procedure was performed considering Scofield’s relative sensing factors.
3. Growth of AlInN layers
The studied structure grown by PA-MBE equipment is presented in Fig 1(a). The AlN buffer was grown on Si(111) using the growth conditions in Ref. [30]. For the p-AlInN, an MME process with short periods for shutter open and close cycles was employed to increase the Mg incorporation and reduce the In-atoms interdiffusion. Interdiffusion is anticipated due to several influencing factors. One primary factor is the growth temperature, additionally, an increase in thickness may further contribute to this observed interdiffusion[12]. Fig. 1(b) shows the short-period MME sequence employed for the p-type AlInN layer, it was designed to ensure the In-atoms incorporation and avoid metal clustering with a nominal thickness of 150 nm. Note that a continuous N-atomic flux was used during growth to reduce In-atom mobility by rapidly forming Ⅲ−N bonds.
The substrate temperature timing diagram for the whole growth process is shown in Fig. 2. The sample surface was monitored during the different stages employing reflection high energy electron diffraction (RHEED), as shown in the insets of Fig. 2. Firstly, the silicon substrate presented a 7 × 7 reconstruction, turning into a 3 × 3 reconstruction while the temperature increased up to 850 ℃. During the AlN-buffer layer, the streaky pattern indicated a 2D growth. Finally, the substrate temperature was diminished to 450 ℃ to start the short-period MME process for the p-AlInN growth named as AlInN growth stage in Fig. 2. At the end of this stage, before cooling, partially spotty RHEED patterns were noted, which indicates that increasing the thickness of the p-layer resulted in a slight roughening of the surface[31, 32]. However, for AlInN-based EBL only a few nanometers in thickness is necessary[17], for this thickness RHEED patterns indicated that flat surfaces were still conserved. The interface between the AlN buffer and AlInN layer was formed by reducing the substrate temperature from 850 to 450℃ to prepare for the growth by short-period MME. We employed an Al and In, equivalent pressure (BEP) of 1.0 × 10−7 and 2.79 × 10−7 Torr, respectively. The Mg effusion cell BEP was set according to previous work to achieve a high incorporation of Mg in the AlInN alloy, expecting a free hole concentration over 1 × 1019 cm−3[33]. Finally, the samples were processed to deposit 30 nm Au-based ohmic contacts for further measurements and characterization, with an 8-minute rapid thermal annealing at 500 ℃.
4. Results and discussions
4.1 Metal-modulated epitaxy of p-type AlInN layer
The p-AlInN layers were analyzed by X-ray diffraction (XRD), as shown in Fig. 3, where the ternary alloy reflection AlInN(0002) is observed at 2θ = 35º. Samples 694 and 695 were grown using the shutter sequence illustrated in Fig. 1(b) and the growth conditions described in the previous section. The only difference lies in the BEPMg of 1 × 10−10 and 5.19 × 10−9 Torr, for samples 694 and 695, respectively. In addition, the nitrogen plasma was turned off for approximately five minutes during the growing process of sample 694 to observe the effects of a lower nitrogen atmosphere on the samples' characteristics. Both samples presented a relatively similar Indium content, however, the interruption of N-flux during the growth process allowed the rapid metallization of In-atoms, forming In(101) clusters as was revealed by a diffraction peak at 32.85º. On the other hand, the sample labeled as 695, exhibited a more pronounced shoulder of the AlInN(0002) peak towards smaller θ angles, pointing to a nonhomogeneous layer and defects formation[34, 35].
Furthermore, composition and defects were assessed using cathodoluminescence (CL) spectra. CL measurement revealed the emission peak around 350 nm (3.5 eV), as shown in Figs. 4(a) and 4(b). On the other hand, as shown in Fig. 4(c), the c-parameter obtained from XRD diffractograms could approximate the Al content using Vegard's law with a bowing factor b = 4.97 eV[36, 37], bringing a value of 0.79−0.80 for the Al molar fraction in the AlInN alloy. It is worth noting that Mg-doped AlN has a high activation energy exceeding 500 meV[33, 34], in contrast to Mg-doped InN which displays a lower activation energy of approximately 60 meV[35, 36]. Note that AlxIn1−xN, where x is close to 0.8, emits between 4–5 eV because of its noticeable nonlinear-dependent bandgap[37], and that the acceptor activation energy of AlInN is larger than 500 meV, which is derived from the linear interpolation between Mg-doped AlN and InN. From these considerations, the main CL peak in sample 694 at about 3.5 eV can be likely attributed to conduction band-acceptor (eA0) or donor-acceptor (D0A0) emissions when the Al concentration is around 80%. For sample 695 the main peak centered at approximately 3.7 eV is also possibly related to eA0 or D0A0 transitions, but with a higher Al concentration in the AlInN alloy. The CL spectrum of sample 694 is very wide with a long tail towards longer wavelengths, in contrast, the CL spectrum of sample 695 is sharply divided into two bands. This difference could be related to the nitrogen flux interruption in the growth of sample 694, but further analysis is necessary to clarify this point. As Fig. 4(c) shows, the main CL emissions from samples 694 and 695 are near Al0.85In0.15N, marked as stars, which is close to the Al molar fraction obtained from XRD.
The XPS analysis was conducted under three distinct etching surface conditions, each bombarded by Ar+ ions at an energy of 500 eV for durations of 60, 300, and 600 s. During the third measurement, the Carbon signal was deliberately reduced by less than 1% to facilitate the study of the bulk chemical properties. High-resolution measurements uncovered signals at 73.9, 444.5, and 396.7 eV, which correspond to Al, In, and N, respectively. Specifically, the photoelectron signal at 73.9 eV is well-recognized as Al2p[40], and the signal at 444.5 eV is associated with In3d in InN[41, 42]. The N1s peak at a binding energy of 396.7 eV pertains to In−N bonds[43, 44]. Al, In, and N peaks are presented in Fig. 5 for sample 694.
The chemical state quantification was performed using Thermo Advantage software and was corroborated by Shirley's background peak analysis through manual calculations with Origin software (OriginLab). The results revealed the chemical compositions as 85% Al, and 15% In. According to Shard[45], the XPS detection limits for most elements range from about 0.1 to 1.0 at.%. For heavy elements in a light element matrix, the detection limit can be enhanced to 0.01 at.%, while for light elements in a heavy element matrix, detection limits above 10 at.% are possible. Thus, in the case of Mg in the AlInN matrix, the amount of Mg dopant is beneath the XPS detection limit, which is suitable for the doping level, while SIMS spectroscopy reveals an approximation of around an atomic concentration of 2 × 1019 cm−3. Consequently, the XPS findings indicate that the composition of the ternary alloy is Al0.85In0.15N, aligning well with XRD and cathodoluminescence observations. This result lies within the target value for an EBL layer in a UV−C LED. Finally, the estimation of the valence band maximum (VBM) was performed according to Kraut et al.[46], and the energy level difference between the Fermi level (Ef) and the valence band maximum (Ev) is depicted in Fig. 6. The samples that underwent Ar+ ion etching for durations of 60 and 600 s exhibited a significant difference in the energy between Ef and Ev. This energy difference decreased from 1.39–1.18 eV. Therefore, based on bulk measurements, the Eg can be deduced, presuming recombination from the conduction band to the acceptor state. This suggests that nitrogen vacancies and other defects might be responsible for the notably defective characteristics of the AlInN alloy[36], which aligns with the CL measurements that ranged between 3.47–3.91 eV. Finally, the quantification summary for the studied samples by XPS is documented in Table 1.
Sample | Element | Atomic (%) | x in ternary alloy AlxIn1−xN | |
694 | Al | 45.76 | 0.85 | |
In | 7.66 | 0.14 | ||
N | 46.57 | − | ||
Mg | <0.01 | − | ||
695 | Al | 44.74 | 0.83 | |
In | 9.17 | 0.17 | ||
N | 46.08 | − | ||
Mg | <0.01 | − |
Carrier properties of the p-AlInN are essential to determine the transport properties of an LED EBL. To estimate the carrier nature and mobility, the structure shown in Fig. 1(a) considered adding an AlN buffer layer to isolate the substrate and the Mg-doped AlInN layer. The hole concentration ranged from 1.23 × 1019 to 3.27 × 1019 cm−3 with mobility limits of 146.44 and 76.70 cm2/(V∙s), for samples 694 and 695, respectively.
Surface analysis (Fig. 7) was performed by AFM and SEM images confirming a low roughness of the surface around 2.6 nm for a scan area of 1 µm2. Comparing the thickness between the samples indicates, that the growth conditions are consistent since the thickness only differs around a few nanometers as shown in Figs. 7(a) and 7(d). SEM images from the surface indicate that the increase of Mg-flux, from 1 × 10−10 to 5.1 × 10−9 Torr, for samples 694 and 695, respectively, modified the surface slightly.
Figs. 7(b) and 7(e) illustrate the coalescence of the p-AlInN layer for samples 694 and 695, respectively, with some 3D structures, such as columns or walls, observed during SEM analysis. The hillock-like morphology observed in sample 695 can be attributed to variations in growth conditions, particularly due to the stress induced by higher Mg-flux. Additionally, the presence of over 13% indium content is closely associated with dislocation-induced morphology. This is likely a result of the compressive strain exerted by the AlInN barrier on the AlN buffer[47−49]. In this research, we suggest that the coalescence of plates, occurring due to a shift in growth mechanisms within the Stranski−Krastanov mode, may lead to crack-like formations. This is a consequence of lattice strain[50], which is further influenced by the thickness and Mg doping. Furthermore, the role of indium as a surfactant in the alloy, known to potentially create v-pits or incomplete coalescence in low-dimensional layers[51], cannot be overlooked. Therefore, a separate, detailed study is essential to fully comprehend the constraints of MME in the growth of AlInN-based layers. The roughness shown in Figs. 7(c) and 7(f) is quite acceptable, around 2.6 nm, even under metal-rich conditions, and doping with Mg. The 3D formation, visible in SEM images, could be due to the change from the 2D-to-3D mechanism due to the thickness increase and the metal-rich conditions. However, the MME approach allows for diminishing the effect of the stress at low growth temperatures.
4.2 Simulation of the deep-ultraviolet light-emitting diode structure
Experimental results noted the importance of thickness in the AlInN growth and doping processes, moreover, the MME approach improves the metal incorporation without clustering and benefits the incorporation of In atoms due to the lower growth temperature employed. Theoretical and numerical research has shown that the AlInN-based layer is suitable as EBL for DUV-LED development due to the increase in radiative recombination rates, electron leakage mitigation, and the resulting hole injection into the active region, using a few nanometers layer[14, 17]. In this work, the thickness of the p-type AlxIn1−xN EBL was set to 13 nm to evaluate the effect of indium content in the EBL. The maximum In content was limited to 16%, which is lower than the ~20% achieved in MME samples. The lower 16% In content ensured a case study with no significant phase separation or 3D growth issues arising from the growth conditions. This is despite the observation of non-significant phase separation and 3D growth in the MME samples, even with ~20% In content. Experimental results by XPS, XRD, and CL analysis, revealed a 0.8–0.85 of Al molar fraction to determine the maximum Al content required in the EBL. Spanning the Al content from 0.90 down to 0.84 was studied to determine the lowest In content limit, analyzing how the increasing In content affects the LED performance. The LED structure shown in Fig. 8 describes the simulated approach with a thin AlGaN-based p-contact layer of 2 nm at the top of the structure to improve the electrical performance of the LED[47, 48]. Under this thin contact layer, it was considered a 30 nm AlGaN thick p-cladding layer. The electron blocking layer is set at 13 nm of AlInN with Al content varied from 0.84 to 0.90 Al molar fraction. All the p-type contact layers doping were considered at 2 × 1018 cm−3, while the n-contact layer was set at 2 × 1019 cm−3. The active region is defined by an Al0.55Ga0.45N (1.5 nm)/Al0.70Ga0.30N (10 nm) MQW. The entire polarization charge was calculated using the dataset for lattice constant provided by SILVACO Atlas, and the bands and ternary values are calculated employing the data of the binary compounds reported in Table 2.
Parameter | AlN | GaN | InN |
Eg | 6.2 eV | 3.4 eV | 0.7 eV |
Affinity | 1.4 eV | 4.0 eV | 5.5 eV |
Permittivity | 8.5 | 8.9 | 15.3 |
Electron mobility | 110 cm/(V∙s) | 300 cm/(V∙s) | 2000 cm/(V∙s) |
Hole mobility | 10 cm/(V∙s) | 14 cm/(V∙s) | 20 cm/(V∙s)[50] |
Bowing factor | 1.348 eV | 1.775 eV | 3.678 eV |
Al molar fractions in the AlInN-based EBL were taken from 0.84 to 0.90 and simulated to obtain the electroluminescence spectra and electric field profile. We have established a baseline of over 20% for the In molar fraction, based on the In incorporation achieved using the MME method, and assessed its effect through SILVACO simulations. The hexagonal nature and its related piezoelectric effect of the material drive diminished emissions[51], where the strong piezoelectric field is well-known to affect the active region. In Fig. 9(a) the electric field in the proposed LED structure is presented, we observe that the electric field intensifies within the MQWs as the In content increases, largely due to the rise in lattice mismatch[52].
On the other hand, carrier densities were analyzed using the capture-escape model to treat their dynamics by dividing them into bulk and bond subsystems, mitigating the effect of the bulk carriers in the simulation. Finally, in-plane, self-consistency joined to Shockley−Read−Hall (SRH), Auger, and radiative recombination models were included in the simulation setup. Introducing the AlInN-EBL helps to compensate for the disparity between electron and hole mobilities, thereby reducing electron overflow. As shown in Fig. 9(b), electron density is localized within the quantum well (QW) region. So, as the Indium (In) content in the electron blocking layer decreases, the electron concentration within the barriers correspondingly increases while the electric field is diminished. As the In content in the EBL rises, the electric field in the active area leads to a decreased electron concentration in the QWs positioned after the EBL in the device layout.
Incorporating the AlInN-based EBL leads to an expanded active region. Despite the presence of QWs, there's a notable emission within the barriers, extending throughout the MQW region and even within the quantum barriers (QBs), as depicted in Fig. 10(a). While the confinement increases and the hole injection is improved using p-AlInN-based EBL[14, 15], there exists a spillover of holes into the quantum barriers, increasing the probability of recombination, even radiative, in the barriers[53]. However, the increased power spectral density was observed as illustrated in Fig. 10(b). As shown in the inset, unwanted radiative recombination rates are observed as the In content in the EBL rises attributed to excessive carrier confinement in the active region.
According to simulation, thicker layers increase the number of confined electrons in the MQW region, as a result, it also increases the electron diffusion within the barriers[15] related to an over-confinement of electrons in the active region and hole injection across the EBL via tunneling, which is depicted in Fig. 11. To curtail electron diffusion within the barriers, it's crucial to balance the EBL's thickness and In content, ensuring consistent carrier concentration across all QWs.
Finally, we make a comparison between the same LED structure using AlGaN-based EBL versus AlInN-based by using the minimum Al content used in this work. Figs. 12(a) and 12(b) show the electron confinement in the QWs and the hole injection through the EBL, respectively. While AlInN EBL creates a barrier for electron leakage to the p-region, it also provides a medium for the holes to reach the QWs. The hole accumulation in the frontier between the EBL and the first QB favors the tunneling into the active region while the more pronounced barriers allow the confinement in the QWs, provoking the spillover of the carriers. This explanation is depicted in Fig. 12(c).
5. Conclusions
In the present study, no discernible In-atoms segregation or In-cluster formation was observed, employing a continuous nitrogen flux supply, attributed to the precisely controlled short-period metal modulation. The metal modulation epitaxy method ensured a consistently low surface roughness even under metal-rich growth conditions and nitrogen flux interruptions. RHEED patterns confirm the suitability and controllability to grow films as thin as a few tens of nanometers for EBL applications. CL analyses implied the presence of point defects, potentially signaling Mg substitutional patterns. Through XRD, XPS, and CL characterizations, our experimental Al molar content resulted in higher than 80% suitable for EBL applications. CL-spectra and XPS-VBM suggest potential inter-band transitions. This information can provide deeper insight into the MME-based AlInN growth process and the role of Mg incorporation. Although earlier studies propose higher Al contents in AlInN alloys, our findings underscore the importance of capping the Al molar fraction at 0.86 in the LED structure to counteract strong polarization effects and achieve closer lattice matching. Notably, a decrease in Al molar fraction between 0.86 and 0.88 augments emissions across the QW and QB, broadening the active region. The electron diffusion into the QB requires exploring thinner EBL with Al content consistent with experimental results obtained from the MME approach. Future endeavors should emphasize optimizing the LED structure for recombination rates, investigate the interplay of EBL thickness on the QW's performance, and assess the strategic blending of the final QB into the EBL. Despite the inherent challenges of high Al content, we reported improvements in carrier properties, resulting in decreased electron overflow and enhanced power spectral density. As we forge ahead, a key focus will be discerning the ramifications of variable EBL thickness on LED performance, especially given the anticipated adjustments to bulk properties due to the reduced EBL thickness from p-AlInN integration, and future works need to be also related to the study of asymmetric carrier injection and confinement to mitigate these parasitic emissions in the active region.
Acknowledgments
The authors thank the CINVESTAV-IPN for their full support and assessment during the experimental stage and CONAHCYT for its postgraduate scholarship program. This work was supported by the TECNM/ITTG project: 18685.23-P.