Optimizing Quantum‑Well Width for Peak Electroluminescence in AlGaN Deep‑UV LEDs Across Temperatures

Background

AlGaN‑based deep‑UV LEDs have emerged as key components for solid‑state lighting, sterilization, and biochemical sensing. Advances in crystal quality, p‑type doping, and device architecture—such as high‑temperature annealing of AlN, sapphire‑grown AlN thick films, and defect‑engineering in homoepitaxy—have steadily improved performance [1–9]. Nonetheless, achieving high external quantum efficiency (EQE) remains a challenge, largely due to strong spontaneous and piezoelectric fields inherent to wurtzite nitrides that induce band‑tilting and carrier separation [10–17]. Prior work by Hirayama et al. demonstrated that a 1.5–1.7 nm QW width yields the brightest photoluminescence in single‑QW AlGaN LEDs, while thinner wells suffer increased non‑radiative recombination at heterointerfaces [18]. Building on these insights, we fabricated DUV LEDs with three distinct QW widths (2.0, 3.5, and 5.0 nm) to systematically examine how well thickness and operating temperature shape EL characteristics.

Methods

All devices were grown on c‑plane sapphire using metal‑organic chemical vapor deposition (MOCVD). A 1.0‑µm AlN buffer was followed by a 0.5‑µm undoped Al0.6Ga0.4N layer and a 1.0‑µm n‑Al0.6Ga0.4N template, achieving a dislocation density of ~6×10^9 cm^−2 (TEM). The active region comprised Al0.49Ga0.51N/Al0.58Ga0.42N multiple QWs with 5.0‑nm barriers. Three samples—LED A (2.0 nm QW), LED B (3.5 nm QW), LED C (5.0 nm QW)—were fabricated. Standard lithography defined 500 µm × 500 µm p‑n junctions. Reactive ion etching exposed the n‑Al0.6Ga0.4N ohmic layer. Ti/Al/Ni/Au (15/80/12/60 nm) contacts were deposited by e‑beam evaporation and annealed at 900 °C for 30 s in N2. Transparent p‑contacts used Ni/Au (6/12 nm) and were annealed in air at 600 °C for 3 min; a final Ni/Au (5/60 nm) layer completed the device. EL spectra were recorded from 5 K to RT using a Jonin Yvon Symphony UV‑enhanced, liquid‑nitrogen‑cooled CCD. Pulse injection (1 µs pulses at 0.5 % duty cycle) mitigated self‑heating effects [19].

Results and Discussion

Figure 1a displays the room‑temperature EL spectra of LEDs A, B, and C under 100 mA forward bias. Normalized to band‑to‑band emission, the peak wavelengths shift from ~261 nm (A) to ~268 nm (C), indicating a red‑shift with increasing QW width. A weak parasitic peak near 304 nm in LED A points to electron overflow [20]. Figure 1b plots the relative EQE versus pulse current, normalized to LED B’s peak EQE. LED B’s maximum EQE exceeds LED A’s by 6.8× and LED C’s by 4.8× at RT. Simulations using APSYS illustrate the underlying physics. Figures 2a–c show the band diagrams, ground‑state energies, and carrier wavefunctions for a 100 mA injection. The internal field inclines the QW band profile, and QCSE reduces electron‑hole overlap as the QW widens. Ground‑state energies are 4.733, 4.669, and 4.637 eV for LEDs A, B, and C, matching the observed wavelengths. Barrier heights rise from 0.030 eV (A) to 0.069 eV (C), yet LED A still suffers from electron overflow, while LED C’s efficiency is limited by QCSE. Thus, the 3.5 nm well balances confinement and overlap, yielding the highest EQE. At low temperature, the relative EQE curves shift markedly. Figure 3a shows that at 5 K, LED B’s maximum EQE is 8.2× and 1.6× higher than LEDs A and C, respectively. Figure 3b reveals that the current required for peak EQE decreases with temperature for all devices. The drop in hole concentration with cooling—due to Mg ionization energy in p‑AlGaN—reduces carrier availability, while the electron overflow current, expressed as J_overflow = D (ΔE/kT)^3 q B l, grows as T decreases [22]. Consequently, the optimal injection point shifts to lower currents. At 5 K, freeze‑out of non‑radiative centers improves internal efficiency, particularly benefiting LED C with the largest active volume. This explains the reduced EQE ratio of LED B to LED C at 5 K compared to RT, while the ratio to LED A increases. Overall, the data underscore the critical role of QW width, polarization fields, and temperature in dictating DUV LED performance.

a The RT EL spectra for LEDs A, B, and C under 100 mA. All spectra are normalized to band‑to‑band emission. b Relative EQE versus pulse current.

The band structure, ground‑state level, and carrier wavefunctions for (a) LED A, (b) LED B, and (c) LED C at 100 mA.

a Relative EQE at 5 K. b Current‑dependent relative EQE at various temperatures for LED B.

Hole concentrations in the active region at 100 K and 300 K for LED B under 100 mA injection.

J_overflow = D (ΔE/kT)^3 q B l, where D is a constant, ΔE is the Fermi‑level to QW band‑edge difference, k is the Boltzmann constant, T is temperature, q is electron charge, B is the bimolecular radiative recombination coefficient, and l is the MQW thickness. At low T, the T^−3 dependence causes a marked rise in J_overflow, explaining the reduced injection current for peak EQE. This, combined with decreased hole concentration, accounts for the observed trends across temperatures.

Conclusions

We systematically studied the impact of QW width on the electroluminescence of AlGaN deep‑UV LEDs from 5 K to RT. The 3.5 nm QW yielded the highest maximum EQE—6.8× (RT) and 8.2× (5 K) higher than the 2 nm device, and 4.8× (RT) and 1.6× (5 K) higher than the 5 nm device. These variations arise from a balance between reduced non‑radiative recombination, increased active‑volume, and optimal electron‑hole overlap. Temperature‑dependent measurements reveal that the required current for peak EQE decreases as T drops, driven by enhanced electron overflow and suppressed hole concentration. The findings provide a clear framework for engineering QW dimensions and operating conditions to maximize efficiency in AlGaN deep‑UV LEDs.

Abbreviations

DUV LEDs

Deep ultraviolet light‑emitting diodes

EL

Electroluminescence

EQE

External quantum efficiency

MQW

Multiple quantum well

PL

Photoluminescence

QCSE

Quantum‑confined Stark effect

QW

Quantum well

RT

Room temperature