Deep-ultraviolet integrated photonic and optoelectronic devices: A prospect of the hybridization of group III–nitrides, III–oxides, and two-dimensional materials

While the doping process and its effects on group III–nitride semiconductors are not fully understood yet, we assert that AlN is the hardest to dope when compared to GaN or indium nitride (InN) and their alloys (In_xGa_1−xN, 0 < x < 1 ) ^[89]. The primary difficulty is efficient doping of grown films with aluminum-rich clusters; the problem is particularly severe for p-type doping, which is essential for the realization of Ohmic contacts in optoelectronic devices. This can be attributed to the empirical fact that magnesium impurities do not incorporate well into AlN and its high-aluminum-content alloys because of the high energy required to activate dopants. In particular, a high density of compensating point defects can be generated in these materials^{[153, 154]}; such defects include nitrogen vacancies (V_N)^[155], substitutional magnesium (Mg_Ga)–V_N complex and other defect complexes^[156-158], and magnesium interstitials^{[159, 160]}. Therefore, high-conductivity p-type aluminum-rich Al_xGa_1−xN films are difficult to obtain because of the low doping efficiency caused by the high activation energies of the acceptor dopant atoms. Using experimental data obtained by different research groups^[161-165], the activation energy of silicon (Si) as an n-type dopant was observed to begin increasing dramatically when the incorporated aluminum content increased in Al_xGa_1−xN beyond 40%, and reaches its maximum in pure AlN at around 280 meV, as shown in Fig. 2. This translates to an undesirable outcome: When the activation energy of silicon is high, only a few percent of silicon impurities are activated in Al_xGa_1−xN. While increasing the silicon impurity incorporation percentage seems an option, there is a self-compensate effect when the concentration of silicon is too high, as shown in Fig. 3. High silicon concentrations were observed to compromise the crystal integrity of Al_xGa_1−xN, increasing its dislocation density and internal aluminum and/or V_Ns. This situation reciprocates for the p-type doping process, where the activation energy of magnesium ( ${ E_\text{A}^{\text{Mg}}}$) as a p-type dopant in AlN was estimated to be between 465 and 758 meV^{[115, 161, 162, 166-168]}.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The most common p-type dopant for Al_xGa_1−xN is magnesium through the introduction of highly-pure bis(cyclopentadienyl)magnesium (MgCp₂) into a metalorganic vapor-phase epitaxy (MOVPE) reactor^[169]. While group II elements and transition metals could, in theory, constitute p-type dopants for group III–nitride compound semiconductors, the conductivity of magnesium-doped Al_xGa_1−xN (specifically for high x values) is relatively low at room temperature due to the high $E^{ \text{Mg}}_ {\text{A}}$ values, as emphasized earlier. Using Hall-effect measurements, $E^{\text{Mg}}_ {\text{A}}$ values of magnesium-doped AlN were found to be around 510 meV^[115] and 630 meV^[162]. Hole concentration of magnesium-doped AlN, $p$, can be roughly estimated after making the following assumptions: (1) an acceptor concentration, N_A, of 10²⁰ cm^–3 and (2) the energy of electrons in the valence band follow the Maxwell–Boltzmann distribution,

$$p(T) = N_ {{\rm A}}{\rm{e}}^{-E^{ \rm {Mg}}_ {{{\rm A}}}/kT}, \tag{ 1 }$$

$p$ can be determined to be between 2.4 × 10⁸ and 3.9 × 10¹⁰ cm^–3 at room temperature, which results in magnesium-doped AlN being an insulator. Taniyasu, Kasu, and Makimoto observed a $p$ value of as low as 10¹⁰ cm^–3 in p-type AlN at room temperature^[162]. Nevertheless, to realize efficient DUV optoelectronic devices, conductive p-type Al_xGa_1−xN layers are necessary. With limited success, there were several attempts to improve the conductivity and efficiency of p-Al_xGa_1−xN-based materials and devices through the incorporation of the following schemes: tunnel junctions (TJs)^[170], Al_xGa_1−xN/GaN superlattices (SLs)^[171-173], p-type graded Al_xGa_1−xN polar heterostructures^[174-178], Al_xGa_1−xN-delta-GaN QW structures^{[179, 180]}, and co-doping^{[181, 182]}. In 2017, Tran et al. claimed to have experimentally accomplished remarkably high free hole concentrations of up to 6 × 10¹⁷ cm^–3 in magnesium-doped AlN nanowires at room temperature^[183], a value that is several orders of magnitude greater than those of conventional AlN epitaxial layers which are in the range of 10¹⁰ cm^{–3[115, 162, 167]}, which in turn are several orders of magnitude lower than what is necessary to realize efficient LED and LD operation (10¹⁷–10¹⁹ cm^–3). They attributed this strikingly unusual high carrier concentration in AlN to the efficient hole hopping conduction in the magnesium impurity band. This was compelled by the significantly improved magnesium-dopant incorporation in the almost defect-free AlN nanostructures^{[183, 184]}. It is worth noting that although no research group was able to achieve conductive beryllium (Be)-doped Al_xGa_1−xN, in 2018 beryllium-doped AlN was found to exhibit a lower acceptor ionization energy of 330 meV^[185] (in good agreement with effective-mass theory^[168] and density functional theory [DFT]^[186] studies) compared to that of magnesium-doped AlN. The values of p were found to be around 10¹⁴cm^–3, taking into account the aforementioned assumptions for estimating p, which demonstrates three orders of magnitude improvement compared to magnesium-doped Al_xGa_1−xN. However, later in 2010, Szabó et al. estimated the ionization energy of beryllium acceptors in AlN to be around 0.97 eV^[187]. In 2018, Soltamov et al. studied the diffusion characteristics of beryllium in AlN single crystals (vapor phase at 1850 °C). They demonstrated that beryllium diffusion led to the quenching of visible (450 nm, 2.75 eV) and DUV (265 nm, 4.7 eV) optical absorption bands simultaneously with the triggering of an absorption band peak at around (248 nm, 5 eV). They also found that the introduction of beryllium impurities compensated the donor type paramagnetic centers by examining the doped semiconductor using electron paramagnetic resonance (EPR)^[188]. They observed a shift of the Fermi level toward the lower energy levels of the AlN energy bandgap, reasoning that the beryllium atoms that incorporated into AlN through the diffusion process acted predominantly as acceptor impurities. However, this shift of the Fermi level resulted in recharging of the deep level defects in the AlN bandgap, causing the observed quenching of the visible and UV absorption bands. The commercial and scientific use of beryllium is impeded by its cost and the toxicity of inhaled beryllium-containing dusts^{[189, 190]}. Furthermore, the use of foreign substrates, such as silicon and sapphire, and the short diffusion lengths of aluminum adatoms on growth surfaces cause deteriorated Al_xGa_1−xN crystalline quality, especially for aluminum-rich Al_xGa_1−xN films. These generated dislocation defects result in high leakage currents and suppression of the radiative recombination efficiency of carriers in Al_xGa_1−xN-based device active regions, which negatively impact device reliability and quantum efficiencies of optoelectronic devices^{[191, 192]}.

Islam et al. demonstrated tunable DUV LEDs using ultrathin GaN quantum dots (QDs) in the device active regions^[193], by manipulating the GaN layer thicknesses with monolayer (ML) precision and applying a polarization-induced doping scheme for both n- and p-type carrier injection regions to enhance the electrical injection of carriers into the active regions^{[194, 178]}. The surface-emitted light from GaN ML quantized structures is predominantly transverse electric (TE)-polarized^[195], enhancing light extraction process when compared to Al_xGa_1−xN-based active regions^[196]. We note that Reich et al. employed ${{k}} \cdot {{p}}$ theoretical model calculations to demonstrate an optimized Al_xGa_1−xN MQW active region design, yielding increased TE polarization in bottom-emitting DUV LEDs with peak emission wavelengths of as short as 239 nm^[197]. Compared to a previously reported value of 243 nm (E ≈ 5.1 eV)^[198], the shortest electroluminescence (EL) emission wavelength Islam et al. achieved in their work was 232 nm (E ≈ 5.34 eV), a record short wavelength DUV LED in structures that simultaneously comprise binary GaN active regions and polarization-induced doping strategy. Their work advanced the incorporation of ultrathin GaN QWs and QDs as alternatives to Al_xGa_1−xN, constituting the light emission source material^[198-200]. Through the reduction of GaN bulk layers having a direct energy bandgap of 3.4 eV ( $\lambda =$ 365 nm), to a few ML-thick quantum confined structures that are sandwiched between AlN barriers, tunable DUV photoemission from 234 to 274 nm (from about 5.3 to 4.5 eV) has been observed^[198]. This relatively large blueshift in peak emission caused by quantum confinement effects was possible by virtue of the large conduction band offset (CBO) between GaN and AlN of about 1.8 eV^[201-205]. According to first-principles calculations based on DFT^[206], and many-body perturbation theory and experimental results^[207], the emission wavelength can be further lowered to 224–228 nm (E ≈ 5.44–5.54 eV) for 1 ML GaN QWs, and additionally reduced to 222 nm (E ≈ 5.59 eV) with 1–2 ML QDs^[208]. Local compositional disorder- and defect-free and fully strained ML-thick GaN quantum structures offer enhanced IQE characteristics over Al_xGa_1−xN-based active regions: ultrathin quantized active regions can potentially enhance the IQE^{[208, 209]}, while the quantum-confined Stark effect (QCSE) is suppressed because of the negligible voltage drops across thin layers and the resulting improved overlap integral value between the electron and hole wavefunctions^[210]. Also, because of three-dimensional (3D) quantum confinement of carriers in QDs, the injected carriers in QD-based LED structures are strongly delocalized from nonradiative recombination channels at TDs, further enhancing the IQE^{[211, 212]}.

By growing quasi-one-dimensional (quasi-1D) group III–nitride nanostructures using plasma-assisted molecular beam epitaxy (PA-MBE)^{[30, 138, 213-216]}, dislocation- and piezoelectric polarization-free Al_xGa_1−xN-based light-emitting nanowire structures can be realized. Nanowires have the advantage of allowing for the growth of lattice-mismatched foreign substrates, such as silicon and sapphire^[217-221], in addition to achieving nearly defect-free crystals^{[222, 223]}, and therefore, opportunity for fabricating high efficiency optoelectronic devices, including LEDs and lasers, because of the highly effective lateral stress relaxation associated with the nanowire large surface-area-to-volume ratios^[224-228]. However, because of the higher surface-area-to-volume ratios in lower-dimen sional semiconductor structures and the subsequent formation of deep-level surface trap energy states, the intrinsic properties of d-dimensional semiconductor structures for all d $\in$ {0,1,2} are directly influenced by their surface condition^[229-234].

Previous studies on aluminum-based group III–nitride materials have shown desirable structural and electronic properties with tunable direct bandgap that can cover the entire UV spectral band. These studies highlighted the increase in the optoelectronic active area of devices, where UV light emission could be enhanced fourfold by incorporating boron, but they lack any direct empirical demonstration^[235]. While Liu et al. have theoretically investigated the spontaneous polarization and piezoelectric constants of wurtzite B_xGa_1−xN and B_xAl_1−xN (0 ≤x ≤ 1) ternary alloys and concluded that they are not piezoelectric for boron concentrations of 75% and 87%, respectively^[236], the miscibility of gallium and aluminum in BN are still under study. However, they are projected to have very low miscibility in BN at normal growth temperatures as the boron miscibility in AlN and GaN is evidently very low as well (a maximum of 14.4% boron content in single-phase wurtzite B_xAl_1−xN films was reported by Li et al.)^[237-242]. For both unstrained ternary B_xGa_1−xN and B_xAl_1−xN alloys, Teles et al. used first-principles calculations to observe very high critical temperatures of roughly 9000 and 9500 K that cause broad miscibility gaps between BN and binaries GaN and AlN, which is attributed to the large lattice mismatch between these compounds^{[240, 241]}.

Hexagonal boron nitride (h-BN) has attracted a considerable amount of attention because of the unique combination of properties it exhibits^[243], including its wide bandgap energy isostructural with graphite^{[141, 244-248]}, low relative permittivity^[249-251], high thermal conductivity^{[252, 253]}, and chemical inertness. It typically exhibits p-type behavior because of acceptor-like vacancy formation^[254]; when grown on sapphire substrates, oxygen impurities diffuse from the substrate during high temperature growth, acting as substitutional donors^[255]. The presence of these donors causes donor-acceptor pair (DAP) recombination involving carbon deep level acceptors^{[256, 257]}. Despite that and its simple crystal structure, the nature of its bandgap configuration (i.e., direct vs. indirect) remains debatable^[258-274]. In 2016, Cassabois, Valvin, and Gil provided evidence that h-BN has 5.955 eV indirect energy bandgap^[275]. Through phonon-assisted two-photon absorption measurements, they demonstrated that h-BN exhibits phonon-assisted optical transitions that arise from the observation of a thermal distribution of excitons in the high-energy tail of the different emission lines. More evidence of the indirect nature of h-BN energy bandgap is the observation of different phonon replicas of the free exciton at 5.76 and 5.86 eV, which agree with the calculation results. Although the authors demonstrated the indirect bandgap nature of h-BN, there are many uncertainties that remain to be addressed more deeply and clearly. One of them is the strong emission line at 5.76 eV, at which even stimulated emission has been observed; this seems infeasible if the material has an indirect bandgap. Significantly more investigation need to be conducted to verify the optical transition of h-BN and the mechanism of the phonon-assisted emission if h-BN is an indirect material^[275]. In 2017, Laleyan et al. demonstrated that the critical challenges that hindered the development of efficient DUV photonic devices with aluminum-rich Al_xGa_1−xN can be addressed by magnesium dopant-free Al_xGa_1−xN/h-BN nanowire heterostructures^[276]. They observed enhanced emission from AlN (at around 210 nm [ $E\approx$ 5.76 eV], with an IQE of up to 80% at 20 A/cm²), which can be attributed to surface passivation effects from the h-BN shell. In 2019, Schué et al. observed a thermally stable bright luminescent emission at 215 nm ( $E \approx$ 5.76 eV) from single crystal bulk h-BN, with the lowest-energy exciton having a binding energy of around $300 \pm 50$ meV (determined using ab initio calculations)^[258] . Moreover, the dispersion of excitons in h-BN has revealed the presence of direct exciton at about 100 meV above the indirect one, which is responsible for the maximum of absorption in bulk h-BN, illustrating the origin of the Stokes shift observed in h-BN^{[277, 278]}. These results are consistent with highly efficient phonon-assisted luminescence.

In 2018, Pierucci et al. demonstrated the controlled growth of h-BN on graphite using MBE via van der Walls epitaxy^[279]. Fig. 4 illustrates the obtained film with scanning electron microscopy (SEM) images. They found that this type of heterostructure produces sharp bands, while h-BN and graphite mostly retained their original electronic band structure.

Zoom In Zoom Out Reset image size

Of all groups of III–nitride compound semiconductors, devices based on Al_xGa_1−xN are ideally suited for UV and DUV device design and fabrication because of the tunability of photoemission and photodetection wavelengths from about 200 to 364 nm by adjusting the AlN mole fraction. Because GaN crystal structures can be assimilated as cubic if the hexagonal planes are slided, the crystal lattice constants of group III–nitride semiconductors can be determined considering that each crystal forms a cubic lattice; hence, the lattice constant can be determined using the following expression:

$$a_{{\rm {cubic}}} = \sqrt[3]{\sqrt{3} a_ {{\rm w}}^2 c_ {{\rm w}}}, \tag{ 2 }$$

where $a_ {{\rm w}}$ and $c_ {{\rm w}}$ are the a-axis and c-axis, respectively, of the lattice of the wurtzite structure. Fig. 5 shows the bandgap energies of III–V semiconductor materials as functions of their lattice constants. The Al_xGa_1−xN equilibrium structure is the wurtzite where all atoms are tetrahedrally coordinated with atoms of the opposite type.

Zoom In Zoom Out Reset image size

The wurtzite structure is represented in Fig. 6(a) and is made of two hexagonal close-packed lattices perpendicular to the substrate (c-direction). The lattice constants are a = b $\neq$ c and the lattice angles are α = β = 90° and γ = 120°. For the case of GaN and AlN a = 3.1880 and 3.1127 Å, respectively, while c = 5.1856 and 4.9816 Å, respectively^[285]. It is also noteworthy that in such compounds, the ratio c/a varies from the ideal value, denoting some degree of distortion. Fig. 6(b) shows the different crystallographic planes of the wurtzite structure. Most of group III–nitride-based LEDs and lasers have c-orientation (polar); however, research has also focused on semipolar (a- and r-planes)^[286] and non-polar (m-plane) because of a reduction in the polarization field and QCSE^[287]. Two types of polarity are present in the wurtzite crystal structure, as it is not invariant with the inversion alongside the c-axis. This implies that the nitrogen atoms and the metal atoms exchange each other, causing crystallographic polarity, where the c-direction is the polar direction. Fig. 6(a) shows the $[0001]$ and $[000\bar{1}]$ directions that can be described as parallel to the Ga–N bond pointing toward the gallium and nitrogen atoms. Therefore, films grown along the $[0001]$ or $[000\bar{1}]$ directions are gallium- or nitrogen-polar, respectively. This is of essential significance for III–V nanowire growth and device design.

Zoom In Zoom Out Reset image size

Spontaneous polarization in III–V semiconductors has detrimental effects through band bending^{[289, 290]}. It results from the partially ionic bond between nitrogen anions and metal cations, although the chemical bonds in III–V materials are primarily covalent. Polarization vector points toward the [000 $\bar{1}]$ direction that is parallel and anti-parallel to the growth direction of nitrogen-polar and gallium-polar films. A summary of group III–nitride physical properties is provided in Table 2.

Table 2.  Crystallographic^[291–294], mechanical^{[291, 292, 295–300]}, thermal^{[291, 293, 294, 301–314]}, and optical properties^{[305, 315, 316]} of select wurtzite group III–nitride semiconductors at room temperature.

Property	AlN	GaN	InN
Lattice constant (Å)	a = 3.1127 ± 0.0003	a = 3.1880 ± 0.0001	a = 3.53–3.548
	c = 4.9816 ± 0.0005	c = 5.1856 ± 0.0005	c = 5.69–5.76
Energy bandgap (eV)	6.2	3.44	0.69
Poisson's ratio	0.287 and 0.216a	0.37 and 0.33b	0.14–0.20c
Thermal expansion coefficient (10–6K–1)	$\alpha_\parallel$ = 4.2d	$\alpha_\parallel$ = 5.59f	$\alpha_\parallel$ = 3.1
	$\alpha_\perp$ = 5.3e	$\alpha_\perp$ = 3.17g	–
Thermal conductivity (W/(m·K))	319	230	800
Specific heat (J/(g·K))	0.6	0.49	0.32
Young's modulus (GPa)	C11 = 410 ± 10, C12 = 149 ± 10, C13 = 99 ± 4, C33 = 389 ± 10, C44 = 125 ± 5, C66 = 120 ± 10	C11 = 390 ± 15, C12 = 145 ± 20 C13 = 106 ± 20, C33 = 398 ± 20 C44 = 105 ± 10, C66 = 123 ± 10	C11 = 190 ± 7, C12 = 104 ± 3 C13 = 121 ± 7, C33 = 182 ± 6 C44 = 10 ± 1
Index of refraction	2.1–2.2h	2.4–2.7i	2.05–3.06j
Melting point (K)	3100	2538l	1800m

One of the principal causes of low IQE values in group III–nitride-based optoelectronic devices is the significant polarization fields built-in by the non-centro-symmetric nature of the atomic bonds. In a multilayer heterostructure, the spatial gradient of the polarization across the interfaces causes the formation of fixed charges, resulting in band bending and in turn, separation of carriers^[319]. Hence, polarization fields are generally undesirable in light-emitting devices, as they hamper the integrity of the electronic and optical properties of the devices, reducing their performances. In the wurtzite crystal structure, spontaneous and piezoelectric polarization fields are caused by the partly ionic bond that induces a small crystal distortion and the presence of strain, respectively. In gallium-polar films the spontaneous polarization (P_SP) is parallel to growth direction (positive z-direction), whereas in nitrogen-polar films, P_SP is antiparallel to growth direction. Most of the group III–nitride planar crystals are grown gallium-polar, while most of the nanowires are grown nitrogen-polar. However, this depends on the growth conditions, and polarity can be easily interchanged.

Because the piezoelectric polarization field (P_PE) is related to the internal crystal strain ( $\varepsilon$), it can be described as the nonzero elements of the crystal strain tensor as follows:

$$\varepsilon_{xx} = \varepsilon_{yy} = \frac{a-a_0}{a_0}, \quad \varepsilon_{zz} = -2\frac{C_{13}}{C_{33}}\varepsilon_{xx}, \tag{ 3 }$$

where a, a₀, C₁₃, and C₃₃ are the equilibrium and strained values of the in-plane lattice constant and the elastic constants^[320]. The piezoelectric polarization is then expressed as:

$$P_{ \rm{PE}} = \pm 2 \varepsilon_{xx}\left(\varepsilon_{31}-\varepsilon_{31}\frac{C_{13}}{C_{33}}\right), \tag{ 4 }$$

where, depending on the metal or nitrogen polarity, we have the plus or minus sign. For the specific cases of GaN, AlN, and InN in metal-polar condition, the term within the brackets is negative; therefore, P_PE is negative (polarization vector points the substrate); while in nitrogen-polar condition it is positive, as summarized in Fig. 7

Zoom In Zoom Out Reset image size

Changes in the normal component of the polarization field, ΔP, cause the formation of fixed charges at the heterointerfaces with density σP = –ΔP and the subsequent carrier separation due to band bending. In the case of MQW structures, the continuity equation of the displacement flux through the different interfaces can be expressed as follows:

$$\varepsilon_{ {{\rm w}}}F_{ {{\rm w}}} + P_{ {{\rm w}}} = \varepsilon_{ {\rm b}} F_ {\rm b} + P_ {\rm b}, \tag{ 5 }$$

where w and b stand for well and barrier, respectively, and ε, F and P are the permittivity of the material, the built-in electric field, and the sum of spontaneous and piezoelectric polarization fields, respectively. F_w and F_b are given by the following:

$$F_ {{\rm w}} = \frac{(P_ {\rm b}-P_ {{\rm w}})L_ {\rm b}}{\varepsilon_{\rm b}L_ {{\rm w}} + \varepsilon_ {{\rm w}}L_ {\rm b}}, \quad F_ {\rm b} = \frac{(P_ {{\rm w}}-P_ {\rm b})L_ {{\rm w}}}{\varepsilon_{\rm b}L_ {{\rm w}} + \varepsilon_ {{\rm w}}L_ {\rm b}}, \tag{ 6 }$$

solved in the limit of an infinitely periodic structure andwhen voltage drop across layers is equal to zero(F_wL_w+F_wL_w = 0, Lis the layer thickness). Fig. 8 shows the MQW band bending and the misaligned electron–hole wavefunction that imply lower absorption and emission properties. Large field values of few MV/cm are usually present in c-plane III–nitride structures, causing the QCSE to dominate as the gap between conduction band minima and valence band maxima are reduced (red-shifted) by F_w.

Zoom In Zoom Out Reset image size

Although In_xGa_1−xN-based devices demonstrate a remarkable resilience to high dislocation densities^[321], dislocations are believed to be a significant factor limiting the efficiency of Al_xGa_1−xN-based LEDs^[322-325]. In_xGa_1−xN/GaN-based light-emitting devices can be operational even with defect densities of up to 10¹¹ cm^–2 because of the effects of carrier localization^{[95, 321, 326, 327]}. In any case, these defects are detrimental for optimal device performance, as they can increase leakage current densities (especially in visible wavelength LEDs)^{[328, 329]}, they act as nonradiative recombination centers, they reduce the IQE^[330], and they decrease the lifetimes of LDs^[331]. There are three types of dislocations in group III–nitrides that can be identified: (1) edge (a-type), mixed ([a+c]-type), and screw (c-type) dislocations^[332], and each type of dislocation can be expressed using a Burgers vector (b) as is illustrated in the following^{[333, 334]}:

$$b_{\rm{edge} }= \frac{1}{3}\langle 11\bar{2}0\rangle, \quad b_ {\rm{mixed}} = \frac{1}{3}\langle 11\bar{2}3\rangle, \quad b_ {\rm{screw}} = \frac{1}{3}\langle 0001 \rangle. \tag{ 7 }$$

The defect density across a structure is directly influenced by epitaxial growth conditions of the materials on designated substrates. Because of its optical and thermal properties as well as its relatively low cost when compared to free-standing GaN substrates, sapphire is commonly employed as a substrate for the growth of group III–nitride light-emitting devices^[335-337]. However, large lattice and thermal mismatches hamper their optoelectronic performances. For instance, sapphire and AlN exhibit a 13% lattice mismatch that causes high density of TDs at the interface between a sapphire substrate and an AlN buffer layer^{[338, 339]}, propagating toward the device active regions, and also causes lower IQEs^{[340, 341]}.

Moreover, due to the higher sticking coefficient and lower surface mobility of aluminum adatoms^[342-344], Al_xGa_1−xN epitaxial growth results are more challenging, compared to the GaN counterpart, having larger adatom mobility. For this reason, the growth of Al_xGa_1−xN material results in 3D islands as aluminum adatoms are incapable of moving from their point of impact to energy favorable places. As the islands merge with each other, high-density defects such as grain boundaries and dislocations grow (mosaic model)^[345], with densities in the range of 10¹⁰–10¹¹cm^–2 for AlN, much higher compared to GaN (10⁸cm^–2)^[346]. Another theory, the nucleation model, that has been supported by transmission electron microscopy (TEM) studies and atomic force microscopy (AFM) studies^[347], suggests that the TDs nucleate in the low-temperature nucleation layer extend in subsequent layers. Regarding the techniques used to reduce the TDs in the device active region, we can find superlattices growth^[348], insertion layer growth^[349], growth of graded layers^[350], and temperature- and nitrogen-controlled growth^{[349, 351]}.

Efficiency droop is a phenomenon mainly associated with the reduction in In_xGa_1−xN/GaN-based LED efficiencies as the injection current densities increase. Despite the fact that this phenomenon is very pronounced in blue and green LEDs, it also affects UV devices to a lesser degree^{[352, 353]}. Fig. 9 demonstrates the droop in quantum efficiency with increasing injection current for three LEDs emitting at different wavelength regimes, namely UV, blue, and green LEDs. The decrease in quantum efficiency is caused by the increase in bias voltages and subsequent device heating^[354-358], which in turn reduces quantum efficiencies. Such efficiency droops were attributed to carrier delocalization^{[359, 360]}, electron leakage attributed to polarization mismatches^{[357, 361, 362]}, drift-leakage mechanism^[363] tunneling leakage currents^[364], poor carrier injection^{[365, 366]}, low hole mobility in heavily p-doped aluminum-rich Al_xG_1−xaN^[367], and Auger recombination^{[122, 123, 368, 369]}. However, there is still no conclusive evidence to pinpoint the origin of this problem^[370]. For the special case of UV LEDs, the droop is supposedly less prominent, which can be attributed to the significantly reduced Auger losses in wide-bandgap semiconductors because of the substantial reduction in the coupling between electron and hole bands with increasing bandgap. However, the Coulomb interaction that assists in the Auger transitions is significantly stronger in these materials because of the smaller dielectric constants and the subsequent larger exciton binding energies, rendering the importance and strength of this argument inconspicuous^[371-373].

Zoom In Zoom Out Reset image size

The IQE of a light-emitting device takes into account how efficiently holes and electrons can be extracted from their respective injection layers and the percentage of carriers that recombine radiatively to emit photons, and is expressed as^[104]

$${{\rm{IQE}}} = \frac{ \text{rate of photon generation}}{ \text{rate of carrier injection into active region}}. \tag{ 8 }$$

The IQE can also be expressed as the ratio of the injected current that leads to radiative recombination in the device active region to the total injected current (I_tot) as follows:

$${\rm{IQE}} = \frac{I_ {{\rm r}}}{I_ {\rm{tot}}} = \frac{I_ {{\rm r}}}{I_ {{\rm r}}+I_ {\rm{nr}}}, \tag{ 9 }$$

where I_nr is the current that is dissipated in nonradiative processes and transitions. When I_nr dominates (i.e., I_nr>I_r), efficiency droop takes place. I_nr includes the carrier losses attributed to SRH recombination, Auger recombination, and carrier leakage outside the QWs. Hence, I_tot can be expressed as

$$I_ {\rm{tot}} = I_ {{\rm r}} + I_ {\rm{SRH}} + I_ {\rm{Auger}} + I_ {\rm{leakage}}. \tag{ 10 }$$

Inside the QWs, the carrier recombination current can be expressed as

$$I_ {\rm{QW}} = I_ {{\rm r}} + I_ {\rm{SRH}} + I_ {\rm{Auger}} = qV_{ {\rm{QW}}}(AN + BN^2 + CN^3), \tag{ 11 }$$

where q is the elementary charge value and V_QW is the total volume of the QW region. A, B, C are the three recombination coefficients of SRH, radiative, and third-order processes (such as Auger losses), respectively, and N is the excess density of carriers involved in the respective recombination processes. The leakage current can be represented by^[366]

$$I_ {\rm{leakage}} = \alpha(I_ {\rm{QW}})^2. \tag{ 12 }$$

By combining Eq. (9) with Eq. (12), we can express the IQE as follows

$${\rm{IQE}} = \frac{\eta_ {\rm{inj}}+BN^2}{AN + BN^2 + CN^3}, \tag{ 13 }$$

where $\eta_ {\rm{inj}}$ is the carrier injection efficiency. For In_xGa_1−xN QWs, the A, B, and C coefficients are in the order of 10⁷s^–1, 10⁻¹¹ cm³/s, and 10^–33 cm⁶/s, respectively, from first-principle study by Delaney, Rinke, and Van de Walle^{[374, 375]}. Researchers have been seeking alternatives to solve issues related to low IQEs; one approach is the employment of nanostructured devices, such as nanowire LEDs. Because of the reduced polarization fields stemming from effective strain relaxation and low dislocation densities, as well as reduced Auger recombination, group III–nitride nanowires can potentially represent a solution to solve the aforementioned device efficiency issues^{[58, 376]}.

In the case of group III–nitride semiconductors, it is broadly assumed that the photocarrier relaxation dynamics consist of a fast initial decay process in the subpicosecond range, followed by a slower decay described by the ultrafast carrier thermalization dynamics, the carrier trapping by surface states, and the slower carrier cooling effects, respectively^[377]. This process allows for radiant refrigeration, suggesting that they are solid-state radiative heat pumps^[378] as illustrated by David et al. for the case of group III–nitride LEDs^[379]. Hence, a more in-depth understanding of the opto-electrothermal properties of group III–nitride semiconductors through an entropic point of view is necessary to further understand their applicability and operational and thermal stability^[380] in multifunctional applications^[381]. In 2017, we employed temperature-dependent photoluminescence to examine the photoinduced entropy of an ensemble of thick Al_0.18Ga_0.82N nanowires^[382] and In_0.32Ga_0.68N/GaN p–i–n double-heterostructure nanowire photodiodes^[383], shown in Fig. 10, by correlating the energy exchange during the photoexcitation and photoemission processes of the light–solid reaction and the generation of photoinduced entropy of the nanowires using temperature-dependent (6 to 290 K) photoluminescence. We defined the photoinduced entropy as a thermodynamic quantity that represents the unavailability of a system's energy for conversion into work due to luminescence refrigeration^[384]. We also studied the ultrafast dynamics of photocarriers using wavelength-integrated time-resolved photoluminescence down to the subnanosecond regime. We investigated the thermodynamic behavior alongside the photocarrier dynamics in GaN-based nanowires because an in-depth understanding of the factors that govern energy transfer processes, and the transport of free carriers through them, is crucial to realizing creative designs of reliable and efficient high-power electronics and LEDs. Strong exciton localization in metal-rich clusters, carrier trapping by surface defect states, and thermodynamic entropy effects were examined and related to the photocarrier dynamics.

Zoom In Zoom Out Reset image size

In In_0.32Ga_0.68N/GaN nanowires, we observed a rising trend in the amount of generated photoinduced entropy of the system above 250 K, while a fluctuating trend in the generated entropy of the system below 250 K was observed. The fluctuations in the generated entropy stabilized between 200 and 250 K. It was supposed that the amount of generated photoinduced entropy of the In_0.32Ga_0.68N active region increase as more nonradiative channels became activated, and more shallowly localized carriers settle into deeply localized states (allowed for through the strong localization in indium (In)-rich clusters)^{[385, 386]}; thereby, additional degrees of uncertainty related to the energy of states involved in thermionic transitions were attained. For the Al_0.18Ga_0.82N nanowires, we observed a fluctuating trend in the generated entropy of the system below 200 K, with a fluctuation frequency that was significantly lower than what we had previously observed in In_0.32Ga_0.68N. In contrast to the sharp increase in generated entropy at temperatures close to room temperature in In_0.32Ga_0.68N, an insignificant increase was observed in Al_0.18Ga_0.82N, indicating lower degrees of disorder-induced uncertainty in the wider bandgap semiconductor. We conjectured that the improved atomic ordering in Al_xGa_1−xN^[387-390] induced lower degrees of disorder-related uncertainty related to the energy of states involved in thermionic transitions; in keeping with this conjecture, we observed lower fluctuation frequency below 200 K.

3.1.1. Molecular beam epitaxy

In 2016, while addressing challenges encountered during growth of aluminum-rich Al_xGa_1−xN nanowires for DUV optoelectronic devices, Zhao et al. demonstrated that such nanowires with significantly enhanced compositional uniformity can be realized through a new growth prototype other than the conventional nitrogen-rich growth conditions. They argued that they can a achieve precise control on the optical bandgap energy of the ternary Al_xGa_1−xN nanowires by employing a GaN nanowire template and varying the substrate temperature while improving the aluminum/gallium compositional uniformity. They demonstrated Al_xGa_1−xN nanowire LEDs, with emission wavelengths spanning from 236 to 280 nm^[400]. Their key results are summarized in Fig. 11. The improved growth paradigm revolves around the idea that low nitrogen flow rates would enhance the surface migration of aluminum adatoms, causing a more uniform aluminum/gallium compositional distribution. This resulted in an accurate control on the emission wavelengths of ternary Al_xGa_1−xN nanowires, which can be accomplished by altering the substrate temperature, instead of the conventional epitaxial way of varying aluminum and gallium beam equivalent pressures (BEPs)^{[99, 401-403]}.

Zoom In Zoom Out Reset image size

Vuong et al. studied the optoelectronic properties of h-BN grown by PA-MBE. By combining AFM, spectroscopic ellipsometry, and photoluminescence spectroscopy in the DUV regime, they compared the quality of h-BN grown on sapphire and highly oriented pyrolytic graphite substrates. They were able to demonstrate DUV emission in h-BN, with an emission spectra peak at 235 nm, indicating the high optical quality of the MBE-grown h-BN. The h-BN epitaxial layers grown on highly oriented pyrolytic graphite have shown superior performance in the DUV regime down to 210 nm, compared to the same films grown on sapphire^[404].

Liu et al. have demonstrated large-area AlN nanowall 214 nm LEDs grown on a sapphire substrate. Through temperature-dependent and power-dependent photoluminescence measurements and rate equation analysis, a relatively high internal quantum efficiency of about 60% was determined for the AlN nanowall structures at room temperature. A consistent blueshift in the emission wavelengths was observed with decreasing nanowall widths because of the reduced distribution in tensile strain. The LEDs exhibited excellent current–voltage I–Vcharacteristics, including a turn-on voltage of 7 V and current densities of greater than 170 A/cm² at 12 V^[405].

3.1.2. Metalorganic vapor phase epitaxy

While Al_xGa_1−xN-based structures show significant importance in applications requiring DUV light sources, the absorption of UV light by p-type layers results in these devices suffering from low light extraction efficiencies (LEEs)^{[114, 136, 406-410]}. To solve this issue, various methods have been proposed by a large body of research^{[114, 411, 412]}. In this context, Jo, Maeda, and Hirayama utilized a transparent p-Al_0.70Ga_0.30N layer that was synthesized primarily for enhancing the UV LEE, paving a promising path toward zero absorption property of output UV light^[413]. The transparent p-Al_0.70Ga_0.30N layer served as a contact layer and led to a higher LEE by using reflective metal electrodes. Conventionally, blue and UV LED structures incorporate p-type GaN contact layers, but these layers are not suitable for DUV LEDs because GaN layers absorb most of the output UV emission. Higher aluminum mole fraction Al_xGa_1−xN epitaxial layers with highly transparent and highly conductive aluminum metal layers are crucial to obtain excellent DUV devices, but the fabrication process of realizing high quality p-type Al_xGa_1−xN layers is difficult due to issues associated with the large activation energy of acceptors, resulting in low hole concentrations and low formation energies of compensation defects. Nevertheless, two 260 nm LEDs were fabricated on sapphire substrates by MOVPE, with one employing a nickel (Ni)/gold (Au) as metal electrode, while the other employed a Ni/aluminum metal electrode^[413]. Both LEDs utilize p-type Al_0.70Ga_0.30N layers with a magnesium concentration of 5 × 10¹⁹ cm^–3 and have been thermally annealed at 950 °C in N₂ ambience. Regardless of the low hole concentration < 10 ¹⁶ cm^–3, by using highly transparent (> 95% transmittance) p-type Al_0.70Ga_0.30N contact layer, the LED demonstrated an efficiency of 2%, similar to that of the LED employing p-type GaN contact layers. Furthermore, an enhancement of 3.4% in EQE was obtained in the LED employing a Ni/aluminum metal electrode compared to that employing a Ni/Au electrode. The aforementioned improved LED performance was attributed to hopping conduction by acceptor sites^{[184, 414, 415]}. To date, numerous efforts have been made to improve the conductivity property of the p-type Al_xGa_1−xN grown layers by increasing the hole concentration in magnesium-doped Al_xGa_1−xN and decreasing its acceptor activation energy. In spite of several approaches developed to overcome this obstacle, including magnesium-delta doping^{[180, 416]}, co-doping^[417], and polarization-induced doping^[418], these structures involve wurtzite crystallographic orientations and suffer from high polarization effects, which result in even larger acceptor activation energies. During the Al_xGa_1−xN growth process, applying nitrogen-rich conditions with large V/III ratios is favorable to achieve better magnesium incorporation and smoother morphology structures.

Metal-source flow-rate modulation epitaxy (FME) provides unique opportunities to realize high hole concentrations in magnesium-doped Al_xGa_1−xN. Luo et al. were able to demonstrate a hole concentration of about 2.3 × 10¹⁷ cm^–3 at room temperature, nearly ten times higher than conventional growth procedures, with resistivities as low as 12.7 Ω·cm by implementing metal-source FME^[419]. Owing to its smooth step-flow growth, improved surface morphology and crystalline quality were observed as a result of employing FME capability in enhancing the adatoms migration as seen in Figs. 12(a) and 12(b). The working principle of the FME technique is based on the interrupted flow rate of metal while a nitrogen source is continuously suppling, ensuring high nitrogen-rich conditions and keeping maximum V/III flow ratio. A slow growth rate of about 4 nm/min was observed by employing FME interrupted metal flow rate, and hence the Al(Ga) adatoms had sufficient time to migrate on terraces toward the steps. As a result, a highly efficient magnesium doping, with acceptor activation energies ranging between 20 and 22 meV, and a smooth surface morphology, with AFM root mean square (RMS) roughness of about 1 nm, compared to 5.4 nm by conventional growth method, were obtained. This can enhance quantum efficiency of DUV LEDs. The FME principle was also utilized to grow Al_xGa_1−xN/AlN distributed Bragg reflectors (DBR) with a reflectivity of 97% and a stopband of 6–9 nm at DUV wavelengths ranging between 220 nm and 250 nm^[420]. The AlN molar fraction was regulated to obtain the required bandgap wavelength associated with the peak reflective wavelength of the Al_xGa_1−xN/AlN DBRs. The grown DBRs of optical quarter wavelength of the desired peak wavelength were grown using MOVPE at a temperature of 1130 °C and a pressure of 100 mbar, while assisted through the FME technique by continuously supplying trimethylaluminum (TMAl) and NH₃, while frequently interrupting the supply of trimethylgallium (TMGa). The AlN molar fractions in two samples were chosen to be 0.48 and 0.61 by controlling the growth thickness of the respective Al_xGa_1−xN layer to be 400 nm and by varying the molar flowrate ratio TMAl/III between 0.35 and 0.44, respectively. For the FME-Al_xGa_1−xN, the AlN molar ration was controlled by tuning TMGa and TMAl period thicknesses and TMAl single period thickness. As a result, the atomic steps of DBR structure was fully strained to the AlN layer and became a defect-free surface and, hence, improved DBR reflectivity was attained. In this regard, we also reported the significant role of yttrium oxide (Y₂O₃)-doped hafnium oxide (HfO₂) on silicon dioxide (YDH/SiO₂) DBR having a high reflectivity of 99.9% at a wavelength of 240 nm with a stopband of 50 nm^[421]. The YDH/SiO₂ DBRs were deposited by RF magnetron sputtering (RF power of 75 W and argon gas flow of 25 sccm) on UV-grade sapphire substrates. The YDH/SiO₂ DBR thicknesses were kept at 30 nm with RMS roughness of about 1.2 nm and 1.07 nm for YDH and SiO₂, respectively. Because of its higher bandgap energy resulting from doping HfO₂ by Y₂O₃, YDH films are semi-transparent in the UV-C band, rendering them prime candidates for use in DUV devices to enhance optical performance.

Zoom In Zoom Out Reset image size

Instead of employing p-type Al_xGa_1−xN layers, several efforts have been made to substitute these low conductivity layers with BN^[422]. One of the advantages of epitaxially grown BN is its small magnesium acceptor activation energy of approximately 31 meV^[423] compared to about 150–300 meV in mixed-phase BN^[424-426], and approximately 170–510 meV in Al_xGa_1−xN^{[162, 167, 169, 427]}. Imura et al. conducted a detailed investigation on boron incorporation on the growth of high-quality AlN layers (Figs. 12(c)–12(g))^[154]. They have confirmed a homogeneous boron concentration of 2 × 10²¹ cm^–3 (≈ 2% of the atomic composition) in the MOVPE-grown AlN epitaxial layers. A conventional AlN/sapphire sample exhibited different emissions, namely near-band-edge emission at 208.4 nm and likely defect-related emissions at 315 nm and 400–450 nm originating from complexes of aluminum vacancies ((V ${}_ {{\rm {Al}}}$–complex)^2–) and/or oxygen impurities (V $^{\small -3}_ {\small {\rm{Al}}}$–O) with dot-shaped defects along the $\{11\bar{2}3\}$ facet oxygen impurity incorporation into AlN^[428-430]. The boron-doped AlN layers demonstrated weak and broad emission at 400–450 nm, as depicted in Fig. 12(g). The quenched emissions in these layers were mainly attributed to the fact that analysis of the incorporation of boron into AlN layers at the nanoscale revealed that the boron atoms were not integrated in the AlN crystal, forming amorphous^[431] and/or disordered precipitates that prevented the formation of perfect B_xAl_1−xN alloys by suppressing the 2D layer-by-layer synthesis of AlN.

Hakamata et al. investigated the combination of sputtering and MOVPE to study surface quality and optical properties of MOVPE-grown AlN and Al_xGa_1−xN epitaxial layers on sputtered and annealed AlN/sapphire templates^[128]. Sputtering is known for its capability to deliver high quality and large size AlN layers through inexpensive means by adopting annealing at high temperatures. The cause of the high crystallinity in sputtered films is post annealing, which enhances the coalescence of AlN columnar structure and reduces possible dislocations in sputtered layers^[432-435]. In their study, an AlN thin film was sputtered on c-plane sapphire substrates and then annealed (SP-AN) at 1700 °C. Then, the MOVPE growth processes of AlN and Al_xGa_1−xN epitaxial layer on SP-AN AlN/sapphire were detailed. Homogeneous epitaxial layers (homoepilayers) of AlN were grown in two stages. Stage I occurred when growth temperature reached and kept at 1250 °C for three minutes in H₂ and NH₃ ambience. Stage II occurred when a 1- $\mu$m-thick AlN layer was grown at 1250 °C. The AFM roughness measurements (Figs. 13(a)–13(c)) show significant enhancements in the RMS roughness (from 0.54 to 0.13 nm) as the fabrication process continued from the high temperature annealing through stages I and II. Fig. 13(c) shows high surface quality step-terrace structure of the grown AlN homoepilayer on SP-AN AlN template. As mentioned previously, this improvement was attributed to the enhancement of the AlN epitaxial layer coalescence. Moreover, oxide or oxynitride column crystals observed at the surface that contained aluminum were removed due to the high temperature annealing process in an NH₃ ambience. On the other hand, the Al_xGa_1−xN heterogeneous epitaxial layer (heteroepilayer) growth process resulted in high- and low-AlN concentrations because of reactor replacement during growth process on the plane- and the 1- $\mu$m-thick AlN regrowth layers on SP-AN AlN/sapphire template. Fig. 13(d) depicts the morphology of the grown Al_xGa_1−xN samples with high and low AlN mole fractions, with and without the AlN regrowth layer. By embedding the regrown AlN layer, dislocations were observed on the surface as 3D growth that assisted in relaxing the resultant compressive strain. Conversely, without the AlN regrowth layer, 3D growth arises because of the roughness of surfaces at grown layer interface, and hence the majority of dislocations were bent^[436-438]. It was detected that while growing thick Al_xGa_1−xN layers, the growth process followed 2D behavior, which led to a flat surface structure.

Zoom In Zoom Out Reset image size

Sun et al. have conducted AlN growth experiments by MOVPE to investigate the effects of TMAl pretreatment on the surface roughness and crystalline quality of grown AlN layers on c-plane sapphire substrates^[439]. In their study, AFM results revealed that precise pretreatment time is the most important factor to realize TD-free smooth surfaces on AlN films. The pretreatment time of five seconds showed the smoothest surfaces, whereas samples without pretreatment or those that underwent a pretreatment time of greater than ten minutes suffered from TDDs, pits, and rougher surfaces. Mixed nitrogen- and aluminum-polarity exhibited in the AlN films was believed to be responsible for low surface roughness on the untreated sample because of the presence of Al_xO_yN interlayer between AlN nucleation layer and the sapphire substrate^{[440, 441]}. For 0–5 s of TMAl pretreatment, RMS roughness values of less than 1 nm indicated excellent morphologies because the nitrogen-polarity areas vanished by inhibiting the oxygen diffusion from sapphire substrate^[442]. By increasing the pretreatment time beyond 40 s, RMS roughness values increased as nanocolumns were observed at the surface as a result of the formation of 3D islands at the nucleation layer and the presence of 3D growth mode^[443]. Finally, the MOVPE growth of single-crystalline AlN-on-sapphire-based ring resonators with high optical Q factors were demonstrated by Bruch et al.^[36]. A 1.1 $\mu$m-thick AlN grown on sapphire substrate was fabricated using a single masking lithography process, where a FOx-16 hydrogen silsesquioxane (HSQ) mask and then 300 nm poly(4-styrenesulfonic acid) (PSSA) were spun before sputtering a 10 nm-thick Au to transfer the pattern to the AlN via an inductively coupled plasma (ICP) etching process. Compared to the two-step fabrication process reported earlier by Liu et al.^[444], this proposed fabrication process enabled high etching selectivity that led to etching the entire AlN structure, and subsequently, more confinement of the visible mode was achieved.

In this section, we discuss various methodologies to enhance the performance of DUV devices. Most challenging to realizing growth of group III–nitride semiconductors are the high TDDs resulting from the lattice mismatch and/or thermal expansion mismatch between foreign substrates such as sapphire or silicon and the grown AlN epitaxial layers. On the other hand, the growth process of AlN on native substrates such as AlN or GaN is limited due to the high cost and unavailability of inexpensive commercial substrates. Various approaches have been discussed to address the sapphire substrates' TDDs^[137]. Patterned sapphire substrates (PSS) are one of these promising approaches to suppress TDs where nanometer-to-micrometer-sized patterns are transferred to the sapphire substrate by additive or subtractive techniques. The main advantage of the PSS is its ability to minimize TDDs through selective area growth and by improving the photon scattering by uneven sapphire surface. Recently, crack-free DUV LED epitaxial layers were obtained on a four inch c-plane sapphire substrate by epitaxial lateral overgrowth (ELO) on periodic air-voids-incorporated nanoscale patterns as shown in Figs. 14(a)–14(c)^[445]. Since the ELO technique demonstrates unique properties in hindering TDD effects and increasing the output power of UV LEDs, it has been widely utilized in micro-patterned substrates. However, it is considered an expensive growth process due to first, higher temperature (above 1300 °C) requirements, which are necessary for the migration of aluminum adatoms, and second, long epitaxial time necessary to realize coalesces thick of 10 $\mu$m of AlN. In their experiment, Lee et al. conducted an ELO of AlN on NPS at a relatively lower growth temperature of 1050 °C, which resulted in a faster coalescence time than previous micro-patterned substrates at much lower cost. During the MOVPE growth process, the air-voids patterns helped to relax the tensile strength issues and minimize TDD effects, and subsequently, its output power reaches 67%, compared to a reference UV LED (Fig. 14(d))^[445].

Zoom In Zoom Out Reset image size

Another method to reduce the fabrication cost of the ELO process of AlN is decreasing the coalescence thickness via nanosphere lithography (NSL). In this regard, Dong et al. successfully utilized nanopatterned sapphire substrates (NPSS) that coalesce after only 3 μm of growth, compared to 10 μm for microstructure patterning methods, and hence, reduce fabrication time^[323]. They were able to enhance the performance of 282-nm Al_0.40Ga_0.60N/Al_0.50Ga_0.50N-based DUV LED by NSL technique, where TDs in the vicinity of nanovoids are suppressed as they bend in the direction of voids sidewalls as depict in Figs. 15(a)–15(c). The EQE was achieved around 3.45% with an enhancement of approximately 98%, compared with conventional at sapphire substrate, by adopting NSL and wet etching mechanisms. In another report, Zhou et al. demonstrated the effect of different sizes of cone-shaped-PSS on performance of UV LED emitting at 375 nm^[446]. They noticed that TDDs are in a decreased trend associated with increasing pattern sizes and with the PSS filling factor. The measured light output power (LOP) of the UV LED grown on a large PSS with a fill factor of 0.71 was 131.8% higher than that of the UV LED grown on a small PSS with a fill factor of 0.4. This behavior can be attributed to the reduction of GaN island density and to an increased lateral overgrowth area of GaN, which leads to higher optical performance of the device. The growth process of GaN epitaxial layers on larger PSS including AlN nucleation layer (15 nm of thickness) increases the transition time from 3D island to 2D coalescence, which resulted in a better IQE performance. In a subsequent report, a light-extraction mechanism based on an Rh mirror electrode and a transparent p-type Al_0.65Ga_0.35N:Mg contact layer was studied by Takano et al.^[114]. They achieved a 20% enhancement in EQE at 275 nm of wavelength. In their design, a conventional Al_xGa_1–xN-based LED is modified into a UV LED structure by utilizing the following features: (1) crack-free AlN on the PSS, (2) Al_0.65Ga_0.35N:Mg p- type contact layer, (3) mirror electrode, and (4) silicon encapsulation resin. The encapsulation resin helped to reduce optical loss and reached 89% transmission. Implementing Rh mirrors into the LED structure results in an improvement in output power from 3.9 to 18.3 mW at 20 mA, and in the EQE from 4.3% to 20.3%, in comparison with conventional LEDs. Increased multireflection of light in the chip and absorption reduction effectively improved the device performance (Figs. 15(d) and 15(e)).

Zoom In Zoom Out Reset image size

Fig. 16 shows a moth-eye micro-arrays which was fabricated on the back of sapphire platform to significantly increase the LEE by weakening the total internal reflection (TIR)^[447]. Consequently, a high degree of optical polarization (DOP) of 81.8% is obtained for DUV LED utilizing moth-eye micro-arrays, compared to a standard DUV LED of DOP of 64.7%. The high performance of LEE (Fig. 16(f)) is due to the impact of moth-eye micro-arrays on improving the TE/transverse magnetic (TM) modes.

Zoom In Zoom Out Reset image size

Similar to sapphire substrate integration in the DUV LED fabrication process, growing AlN epitaxial layers on silicon substrate is of critical importance due to the availability, suitability, and low-cost of silicon substrates. Nevertheless, the lattice mismatches between silicon (111) and AlN (0001) are approximately 19% in the $[11\bar{2}0]$_AIN and $[1\bar{1}00]$ _AIN directions; however, they are around 0.74% between the $[1\bar{1}00]$_AIN direction and silicon (110), which leads to enhancement in device efficiencies^[448]. Although several reports discussed the growth process of AlN layers on different direction-oriented silicon substrates, few reported high quality AlN on silicon (110). In 2017, Shen et al. reported a V-profile pit free high quality AlN surface with clear atomic steps grown on silicon (110) orientation substrate that exhibits high resolution structure via high resolution X-ray diffraction measurements as illustrated in Fig. 17^[448]. The key factor in achieving high resolution structure is by controlling growth rates and AlN layer thicknesses.

Zoom In Zoom Out Reset image size

Micro-circle-patterned silicon substrate (mPSIS) is another configuration technique to enhance the quality of grown AlN template on silicon substrate^[449]. Different mPSIS sizes were fabricated to grow thick AlN template with the help of NH₃ pulsed-flow and ELO in the MOVPE growth process. During the lithography process, the ICP bias powers were modified to obtain different mPSIS sizes. MPSIS pattern sizes of 1.7 and 2.0 $\mu$m were fabricated on two silicon substrate samples by selecting ICP power of 10 W for 50 min of etching and ICP power of 5 W for 40 min of etching, respectively. During the growth process, the temperature decreased from 1390 to 1380 °C while NH₃ increased from 5 to 6 sccm to enhance the AlN surface. The sample with larger pattern size (i.e 2.0 $\mu$m) showed better results in term of AFM roughness of 1.6 nm, compared to 3.5 nm for the other sample with 1.7 $\mu$m pattern size. Also, TEM images showed smoother surface for the large size pattern sample as an atomic step with high quality terraces observed. The TDs become less pronounced in the large size pattern sample than in the small pattern size sample, as the dislocations bend and terminate at sidewalls.

Al_xGa_1−xN-based DUV LEDs have been dominating research in the last few years^[450-462]. Most notably, using the growth paradigm described earlier, Zhao et al. were able to fabricate defect-emission-free Al_xGa_1−xN nanowire LEDs operating in the UV-C regime, and attributed the enhanced current conduction characteristics to the improved acceptor magnesium dopant incorporation in the Al_xGa_1−xN nanowires and the subsequent acceptor impurity band conduction^{[184, 400, 463]}. Fig. 18(a) shows the device schematic, while the I–V characteristics of an Al_xGa_1−xN nanowire LED with emission wavelength around 240 nm are depicted in Fig. 18(b). Fig. 18(c) depicts the room-temperature EL spectra measured from several devices. The output power measurements (L–I–V) of a UV LED device are shown in Fig. 18(d).

Zoom In Zoom Out Reset image size

Taking advantage of high absorption coefficients (>10⁵ cm^–2) and wide bandgap energy, group III–nitride semiconductor alloys have attracted remarkable attention for UV photon detection owing to their prospective applications in solar radiation monitoring^[464], missile plume detection^[465-467], biological treatment^[468], and non-line-of-sight (NLOS) communication^[469]. Compared to other wide bandgap materials (e.g., ZnO, SiC, and Ga₂O₃), the large bandgap tunability of group III–nitrides also allows the absorption or cut-off wavelength to be engineered by simply changing the alloy composition, an easy tailoring for demanding applications that require high signal-to-noise ratios (SNR) and spectrum selectivity. Apart from that, high thermal and chemical stabilities of group III–nitride material system stands out in harsh environments, compared to conventional silicon-based detectors, which have limited operating temperature range and require extra protective packaging^[470-473]. In addition, silicon-based photodetectors typically have low photoresponsivity in the UV regime of operation due to low penetration depth of UV-wavelength photons in silicon^[474]. Photomultiplier tubes (PMTs) were also suggested as alternative devices in UV photon detection^[475]; however, the associated bulky volume and high operating voltage hinders further applications of PMTs^[476]. Following the development of a more mature epitaxial growth for group III–nitride compounds, Al_xGa_1−xN-based photodetectors (PDs), which can cover the wavelength range from 200 to 360 nm depending on the aluminum composition^[477], show promise for achieving highly reliable UV photodetectors and in alleviating the aforementioned issues, and are the subject of active research in UV and DUV photon detection, opening up a plethora of potential applications in the future.

The early work on GaN-based PDs started in in 1992 by Khan et al., where a photoconductive UV detector with high spectral responsivity (R) of 2000 A/W (200 to 365 nm) was demonstrated^[478]. This is the first reported UV sensor based on as-deposited single-crystal insulating GaN films. Since then, various types of Al_xGa_1−xN-based PDs, particularly for photon detection in the solar-blind region (i.e., 230 to 280 nm), such as Schottky-type PDs^[479-481], metal–semiconductor–metal (MSM) PDs^[482-486], HEMTs^[487], and p–i–n heterojunction PDs^[488-491], have been demonstrated. Among these, MSM-based PDs attract noteworthy attention, owing to their ease of fabrication, low stray capacitances, and high switching speed for monolithic integration on photonic circuits^{[480, 492, 493]}. Another notable advantage of MSM-based PDs is that p-type doping of Al_xGa_1−xN is not necessary, putting it in dominance when compared to p–i–n structured PDs. In addition, depending on the device design, MSM-based PDs can be illuminated through either front or bottom illumination. For example, Wang et al. compared performance differences in terms of responsivity and EQE of Al_0.40Ga_0.60N-based MSM UV PD, in both illumination modes^[494]. For a photodetector, the EQE can be determined using

$${{\rm {EQE}}} = \frac{I_{ {\rm PC}}}{P_{ {\rm {IL}}}} \frac{hc}{q\lambda} = {R} \frac{hc}{q\lambda}, \tag{ 14 }$$

where I_PC is the generated photocurrent, P_IL is the illuminated optical power, h is Planck's constant, c is speed of light, and $\lambda$ is operation wavelength. Fig. 19(a) shows the transmission spectrum of a Ni/Au interdigitated electrode commonly used in MSM-based PDs, where low transparency values of less than 45% are observed in the UV wavelength region. In the case of front-illumination, fully transparent or semi-transparent interdigitated electrodes are deemed pivotal in order to increase light absorption efficiency. In such cases, this requirement may complicate the fabrication process because of the necessity for highly UV transparent electrodes, such as graphene^{[142, 495, 496]}, carbon nanotubes^{[497, 498]}, and metal–oxide composites^{[499, 500]}. Compared to the front-illumination mode, higher EQEs are typically observed with bottom-illumination configuration owing to first, the smaller refractive index of sapphire substrates, compared to Al_xGa_1−xN layers, and second, higher absorption efficiency because of the absence of light shadowing metal electrodes^[494] . Fig. 19(b) shows the higher responsivity (inset) and corresponding EQE (around 50%) obtained in the case of a bottom-illuminated PD, compared to a front-illuminated PD. A detailed examination of the influence of AlN buffer layers and Al_0.50Ga_0.50N absorber layer for a bottom-illuminated MSM-based PD was presented by Brendel et al.^[484]. A threefold enhancement of EQE (up to 20% at 20 V) was observed for a 100 nm-thin Al_0.50Ga_0.50N-based bottom-illuminated MSM PD. At the end of 2015, the same group reported an EQE of up to 67% at 50 V in a bottom-illuminated 500 nm-thick Al_0.50Ga_0.50N absorber layer of an MSM-based PD^[485]. Apart from the influence of absorbance layers, the effects of asymmetric electrodes on the EQE of bottom-illuminated Al_0.40Ga_0.60N- and Al_0.50Ga_0.50N-based MSM PDs have also been discussed^{[483, 501]}. The dependence of the EQE on carrier lifetime and transit time in Al_0.25Ga_0.75N-based MSM PD at different operation voltages has been reported by Schlegel et al.^[502]. To produce high performance MSM-based PDs with low leakage currents and high responsivities, various counter-move investigations include correlation between lateral leakage current and screw dislocation density ( $\rho_ {\rm {screw}}$)^[503]. Edge TDs and suppression of leakage current using a surface passivation layer in MSM-based PDs have also been widely discussed^{[504, 505]}. Other than that, carrier collection mechanisms and influence of core threading dislocation in MSM-based PDs have recently been reported by Walde et al.^[506]. This work provides insightful observations to the advantage of open-core threading dislocation in Al_0.50Ga_0.50N layers for improving the EQE of MSM-based PDs, in which dislocation is typically held accountable for high leakage currents in Al_xGa_1−xN-based LEDs. Prior works present valuable device designs and optimization methods for high-performance solar-blind MSM-based PDs. Recently, Yoshikawa et al. reported a true solar-blind Al_0.60Ga_0.40N/Al_0.50Ga_0.50N-based MSM PD with a cut-off wavelength at 280 nm and a high rejection ratio of 10^6[507]. The improved photocurrent and high rejection ratio achieved in this work was attributed to the control of carrier density in the 2D electron gas (2DEG) layer through a decreased thickness of Al_0.60Ga_0.40N barrier layer. Although a high responsivity of up to 10⁶ A/W, comparable to values achieved with PMTs, was achieved; nevertheless, slow response time due to persistent photoconductivity effect was simultaneously observed.

Zoom In Zoom Out Reset image size

Despite the advantages of bottom-illuminated configuration PDs, the thermal and lattice mismatches between conventional sapphire substrates and group III–nitride materials still persist. While paving the way toward the solar-blind regime of operation, high aluminum composition of greater than 40% are required to provide the aforementioned applications in the solar-blind region. High TDDs and thermal structural cracking developed during the epitaxial growth, especially at increased aluminum contents, hinder the device performance of Al_xGa_1−xN-based PDs, which requires low charge generation rates and stray capacitances. In order to alleviate these issues, various methods have been reported, including selective area growth^[509], pulse atomic epitaxial technique (PALE)^[510], and crack-free AlN template layer use. Cicek et al. demonstrated a solar-blind Al_0.40Ga_0.60N-based p–i–n PD with a high EQE of up to 89% (at 5 V reverse bias voltage) and a responsivity of around 176 mA/W (at zero bias voltage) by using various optimized growth and structural design methods^[511]. Improvements in structural characteristics were attributed to the use of a crack-free AlN template layer, a highly conductive silicon-indium co-doped^[512] Al_0.50Ga_0.50N, and an improved magnesium-doped Al_0.38Ga_0.62N layer grown using MOVPE. In a recent work, Han et al. proposed the use of metal–insulator–semiconductor (MIS) structures in solar-blind Al_0.45Ga_0.55N-based PDs, as a competitive alternative to the conventional p–i–n structure^[513]. The latter is impeded by growth of high-aluminum-content p-doped Al_xGa_1−xN layers. The demonstrated bottom-illuminated MIS-based structure exhibited a high EQE of 70.6% (at 3 V reverse bias voltage) at 270 nm with a relatively fast rise time of 1.4 ns. The significance of optimized AlN template in growing high quality Al_xGa_1−xN PD structures has been substantially highlighted. Chen et al. reported a solar-blind Al_xGa_1−xN-based p–i–n PD grown and fabricated on a high quality and low residual compressively stressed AlN template^[514]. They attributed the improved performance of the device (EQE at 68.8% and response speed of 6.5 ns) to the insertion of a mesothermal AlN (MT-AlN) interlayer to assist in strain relief along the growth direction. Similar work on optimized growth conditions for producing high quality and crack-free Al_xGa_1−xN layers have also been reported. High specific detectivity ( $D^*$) of 1 × 10¹⁴ Jones (cm $\cdot\sqrt{ {\rm Hz}}$/W) was calculated for the solar-blind p–i–n-based PD, similar to that of a PMT, but without the need of a high voltage supply. While undertaking the effort to improve Al_xGa_1−xN-based PDs, high PD response speeds up to to 1.7 ns (rise time) and 4.5 ns (fall time) p–i–n PDs were demonstrated by Albrecht et al. through the optimization of layer growth conditions^[515]. In an earlier work, an Al_xGa_1−xN-based MSM PD with a 3-dB bandwidth of as high as 5.4 GHz was also demonstrated^[516]. More recently, Muhtadi et al. reported an Al_0.64Ga_0.36N/Al_0.34Ga_0.66N MQW-based PD with an RC-limited response speed of 2 ns^[517], undoubtedly opening up a vast application opportunity of solar-blind Al_xGa_1−xN-based PDs for high speed operation and monolithic integration.

Other than planar structures, Al_xGa_1−xN-based PDs constructed using lower-dimensional structures, such as nanowires and nanorods, have also received significant attention^[518]. In 2017, Kang et al. reported on a two-step growth process for single-crystalline silicon-doped Al_0.45Ga_0.55N nanorods as a solar-blind PD^[519]. The vertically-aligned and compositionally uniform nanorod structure displayed high responsivity of up to about 115 mA/W and high sensitivity of approximately 64% in the UV-C region^[519]. Later, the same group demonstrated improved photoresponsivity of up to 0.72 A/W, fabricated using n-type Al_0.45Ga_0.55N nanowires with flower-like morphology, also called as nanoflowers^[508]. Figs. 19(c) and 19(d) show improved photocurrent by close to one order of magnitude and high responsivity of up 0.72 A/W in Al_0.45Ga_0.55N nanoflower-structured PD, compared to that of nanowire-structured PD (0.13 A/W) and nanorods (0.09 A/W). The improved performance of nanostructured PDs was largely attributed to improved photon absorption due to high aspect ratios and compact density distribution^[138]. These works paved the way for high performance low-dimensional Al_xGa_1−xN-based PDs in various applications^[520]. Furthermore, group III–nitride-based avalanche photodetectors (APD), which show promise in providing higher photocurrent gains combined with high speeds and low operating voltages, are also of contemporary interest in the community^[521-529]. Moreover, the use of Geiger-mode operation in group III–nitride-based APDs has also received significant attention for its potential employment in high sensitivity single-photon counting for various applications, such as quantum computing and astrophysics^{[530, 531]}. Compared to UV-enhanced silicon-based APDs, Al_xGa_1−xN-based APDs promise high wavelength tunability and selectivity. Low leakage currents and high avalanche gains (5 × 10⁵) in an Al_0.05Ga_0.95N APDs grown on free-standing GaN substrates were demonstrated by Kim et al. in 2015,^[532] highlighting a significant improvement in APD performance, compared to other APD structures grown on sapphire templates. Later in 2016, Wu et al. reported a true solar-blind Al_0.40Ga_0.60N-based APD structure with zero-bias EQE of up to 52.7% with an avalanche gain of higher than 2 × 10⁴ at a relatively high voltage bias of $-140$ V^[533]. More recently, Hahn et al. demonstrated a relatively low dark current of 1 pA at up to 60 V reverse voltages in an Al_0.69Ga_0.31N-based APD grown on sapphire substrate, with multiplication gain of around 5500 at 84 V reverse bias^[534]. Another special type of Al_xGa_1−xN-based APDs, based on a p–i–n–i–n separate absorption and multiplication (SAM) structure, has also been studied for higher gain and lower avalanche breakdown voltage, compared with their conventional counterpart^{[522, 535]}. Fig. 20(a) shows a back-illuminated p–i–n–i–n SAM structure while Fig. 20(b) shows a demonstration of its capability to achieve record high gain of up to 1.2 × 10⁴ at a reverse bias of 84 V.

Zoom In Zoom Out Reset image size

However, to tailor PDs for the DUV wavelength regime of operation, high quality and crack-free high-aluminum-content layers are pivotal. Up until now, no solution has been found to address high defect densities associated with the incorporation of Al_xGa_1−xN layers with high aluminum contents. Walker et al. was one of the earliest to demonstrate Al_xGa_1−xN-based PDs operating in DUV wavelengths with a responsivity of 0.11 A/W at 232 nm (5 V reverse voltage) and an IQE of 90%^[536]. In their device structure, high-aluminum-content Al_xGa_1−xN layers of up to x = 70% were successfully grown using low-pressure chemical vapor deposition (LPCVD) on sapphire substrates. Subsequently, in 2007, Gökkavas et al. demonstrated Al_0.75Ga_0.25N-based with relatively high responsivity of up to 0.53 A/W at 222 nm^[537]. The high responsivity obtained in this work can be attributed to the high-quality Al_0.75Ga_0.25N layer, demonstrating a low dark current of less than 100 pA at very high bias voltages of up to 350 V. Alternative nitride-based materials, such as AlN and BN, have also been widely discussed^{[477, 538-542]}. AlN-based photodetectors received significant attention in the early 2000s because of their high bandgap energies (around 6.1 eV) and high SNRs with the suppression of visible background for DUV detection. The demonstration of AlN-based MSM PD for operation below 230 nm was reported by Li et al.^[477]. A sharp cut-off wavelength was observed at 207 nm with peak responsivity of 0.4 A/W (at 100 V voltage bias) at 200 nm, with a rejection ratio of four orders of magnitude. The insertion of superlattice structures after the AlN nucleation layer have also been introduced by Nikishin et al. to reduce the inversion domains (IDs) in AlN-based PDs^[543]. The fabricated large-area MSM PD exhibited a low dark current of 50 fA up until 30 V voltage bias with the insertion of AlN/GaN short-period superlattices (SPSLs), with a peak responsivity of 0.08 A/W at 202 nm. Following these discoveries, large area (about 3.1 mm²) DUV PDs based on AlN films were demonstrated by Barkad et al. with titanium nitride (TiN) as a Schottky contact^[544]. The cut-off wavelength was reported to be around 203 nm, the shortest ever cut-off wavelength reported in AlN-based PDs, with a rejection ratio of up to three orders of magnitude. Other than the above, the first surface acoustic wave (SAW)-based DUV sensor based on AlN films for photodetection below 200 nm has also been reported^[545]. In recent years, BN with similar bandgap energy (around 6 eV) as AlN have also received increasing attention for DUV applications. In 2018, Soltani et al. demonstrated DUV PD with cut-off wavelength below 200 nm using cubic BN (c-BN)^[546]. The MSM-based PD demonstrated a very short cut-off wavelength at 193 nm because of the wide bandgap of c-BN (about 6.4 eV), while the peak responsivity appears at 32 mA/W at 180 nm with 35 V voltage bias. Although the achieved responsivity is still considered relatively low, the cut-off wavelength was the shortest at that time, and a high rejection ratio of up to four orders of magnitude was also recorded. Later in 2012, Li et al. explored h-BN epitaxial layers as UV photodetectors with sharp cut-off wavelengths at around 230 nm with high absorption coefficients (approximately 7 × 10⁵ cm^–1) than that of AlN (around 2 × 10⁵ cm^–1)^[547]. In addition to their sufficient performance as thermal interface materials^[548], optically transparent (down to 210 nm) BN nanosheets (BNNSs) have also been examined for UV detection below 300 nm^[549]. This study showed that the BNNSs-based DUV PDs were blind to photons with energies of less than 4 eV, with a PD output power of around 2 $\mu$W. However, the output electrical power to the input optical power ratio was about 1%. Until recently, DUV PDs based on 3 nm-thin h-BN layers were demonstrated by Liu et al.^[550]. The devices exhibited a low responsivity of 0.1 mA/W at 212 nm, as shown in Fig. 20(c), but also exhibited a rejection ratio of over three orders of magnitudes between 212 and 284 nm. The low responsivity obtained was ascribed to the limited thickness and non-uniformity of the h-BN layers. In addition, as shown in Fig. 20(d), relatively slow response times of 0.32 s (rise time) and 0.63 s (fall time) were observed, thus limiting their relevance in high speed detection applications. Most recently, we demonstrated a novel heterogeneous integration of group III–oxides and silicon to fabricate solar-blind DUV (Al_0.28Ga_0.72)₂O₃ Schottky MSM and metal–insulator–metal (MIM) photodetectors. Apart from achieving a high peak responsivity of 1.17 A/W at 230 nm, this is also the first demonstration of DUV Ga₂O₃-based photodetector grown on a silicon substrate, paving the way for potential monolithic integration of future DUV oxide-based photodetectors on the matured silicon platform^[551].

As discussed earlier, further improvements in the performance of group III–nitride-based PDs were largely inhibited by the low crystal quality and low conductivity characteristics originating from the growth of PD structure materials. To alleviate these issues, hybridization of GaN-based PDs with 2D materials, such as graphene, have been proposed. Phototransistors based on graphene/GaN structures with relatively high responsivity of 0.36 A/W at 325 nm were reported by Tian et al. and shown in Fig. 21(a). Their reported device characteristic values were significantly higher than those realized without the integration of graphene layers (up to 700-fold improvement); the high achieved gain can be attributed to high carrier mobility effects exhibited in graphene, which allow for multiple carrier transport to generate high photocurrent densities^[552]. Without the graphene layer, it can be expected that generated carriers would be trapped in the lower-mobility GaN layer, limiting the achievable photocurrents and gains in PD devices. Similar work was later published by Tian et al. They demonstrated a significantly enhanced responsivity of up to 5.83 A/W with illumination at 325 nm and an EQE of up to 2200%, which they accredited to multiple carrier circulation and long carrier lifetimes in the graphene layer using a similar hybrid phototransistor structure^[553]. However, it was also emphasized that responsivity considerably decreased with increasing power of incident light, as shown in Fig. 21(b), because of dampening of the net built-in electric fields by opposite electric fields formed by accumulated photogenerated carriers. Nevertheless, both reports provide insight into improving the performance of UV phototransistors without stringent requirements on the crystal quality of light-absorbing layers. Notably, a similar strategy was employed in DUV PD devices to improve photoconductive gains, with graphene layers achieving transparencies greater than 80% within the DUV regime^[554]. The incorporation of BNNSs and 1D cellulose nanofibers has also been demonstrated as flexible solar-blind PDs by Lin et al. Although their demonstrated paper-based device exhibited a relatively low responsivity of 0.05 mA/W at 185 nm illumination, the device displayed near-zero responsivity between UV-B (300 nm) up to NIR (1000 nm) wavelengths, manifesting its true solar-blind detection capabilities^[52]. Furthermore, taking advantage of the high thermal conductivity of h-BN (about 146 W/mK), the paper-based device was capable of functioning at high temperatures of up to 200 °C, which represents a significant step forward in reaching the goal of flexible and wearable electronic devices. Table 3 provides a summary of key performance characteristics of state-of-the-art group III–nitride-based and hybrid UV and DUV PDs.

Table 3.  Performance comparison of select group III–nitride-based and hybrid UV and DUV PDs at room temperature (^* Indicates an EQE value we calculated using Eq. (14)).

Device structure	Material system	Bias (V)	Peak responsivity (A/W)	EQE	Rise/Fall time (s)	Year	Ref.
MSM	(Al0.28Ga0.72)2O3	2.5	1.17 (230 nm)	631%	–	2019	[551]
MSM	$\beta$-Ga2O3/TiN	15	277.08 (250 nm)	1.37 × 105%	–	2019	[555]
MSM	$\beta$-Ga2O3	10	8.41 (254 nm)	4.11 × 103%*	18 s/1.6 s	2019	[556]
MSM	$\beta$-Ga2O3	20	150 (254 nm)	7.40 × 104%	1.8 s/0.3 s	2018	[557]
MSM	Al0.50Ga0.50N	20	0.094 (270 nm)	43.2%*	<20 ms/ <20 ms	2018	[503]
MSM (flexible)	h-BN	10	0.00005 (185 nm)	0.03%*	0.267 s/0.393 s	2018	[52]
MSM	h-BN	20	0.0001 (212 nm)	0.06%*	0.32 s/0.63 s	2018	[550]
MSM	Al0.50Ga0.50N	10	5 (264 nm)	2.35 × 103%*	0.52 s/44 s	2017	[558]
MSM	Al0.50Ga0.50N	5	106 (250 nm)	4.96 × 108%*	0.2 s/1000 s	2017	[507]
MSM (SiO2-passivated)	Al0.25Ga0.75N/GaN	10	0.27 (365 nm)	91.78%*	–	2015	[504]
MSM	Al0.25Ga0.75N	5	<0.01 (310 nm)	<4%*	–	2013	[502]
MSM	Al0.42Ga0.60N	20	0.045 (270 nm)	21.5%	–	2013	[494]
			0.085 (270 nm)	39%	–
MSM	TiN/AlN	100	0.002 (200 nm)	1.24%*	–	2010	[544]
MSM	AlN	10	0.08 (202 nm)	49.14%*	–	2009	[543]
MSM	c-BN	35	0.032 (180 nm)	22.06%*	–	2008	[546]
MSM	Al0.75Ga0.25N	50	0.53 (222 nm)	> 250%	–	2007	[537]
p–i–n	Al0.46Ga0.54N	0	0.15 (271 nm)	68.8%	6.5 ns/0.95 $\mu$s	2018	[514]
p–i–n	Al0.34Ga0.66N	0.5	0.1 (250 nm)	50%	0.4 $\mu$s/0.4 $\mu$s	2017	[517]
p–i–n	AlxGa1–xN	0	0.21 (360 nm)	70%	1.7 ns/4.5 ns	2014	[515]
p–i–n	Al0.40Ga0.60N	5	0.192 (275 nm)	89%	–	2013	[511]
p–i–n	Al0.40Ga0.60N	5	0.129 (279 nm)	57%	–	2013	[509]
p–i–i	Al0.55Ga0.45N	5	0.075 (254 nm)	37%	–	2012	[510]
p–i–n	Al0.70Ga0.30N	5	0.11 (232 nm)	58.83%*	–	2000	[536]
p–i–n–i–n	Al0.40Ga0.60N	15	0.15 (280 nm)	50%	–	2014	[522]
p–i–n APD	Al0.40Ga0.60N	0	0.114 (278 nm)	52.7%	–	2016	[533]
p–i–n APD	Al0.05Ga0.95N	0	0.043 (354 nm)	16%	–	2015	[532]
MIS	Al0.45Ga0.55N	3	0.154 (270 nm)	70.6%	1.4 ns/55 $\mu$s	2018	[513]
MIM	(Al0.28Ga0.72)2O3	2.5	0.4 (230 nm)	218%	–	2019	[551]
p–n	PtSe2/GaN	0	0.193 (265 nm)	90.36%*	0.172 $\mu$s/284 $\mu$s	2019	[559]
p–n	MoS2/GaN	0	0.187 (265 nm)	87.55%*	46 $\mu$s/114 $\mu$s	2018	[560]
Heterojunction	graphene/Ga2O3 nanowires	5	0.185 (258 nm)	88.96%*	9 ms/8 ms	2018	[561]
Heterojunction	graphene/ $\beta$-Ga2O3	6	12.8 (254 nm)	6.68 × 103%	1.5 ms/2 ms	2018	[562]
Heterojunction	graphene/h-BN/GaN	1	1915 (245 nm)	9.69 × 105%*	8 s/6 s	2016	[563]
Schottky	graphene/ $\beta$-Ga2O3/graphene	10	9.66 (254 nm)	4.72 × 103%*	0.96 s/0.81 s	2017	[564]
Schottky	graphene/Al0.25Ga0.75N/GaN	2	0.56 (300 nm)	231.60%*	–	2016	[565]
Phototransistor	graphene/GaN	10	0.361 (325 nm)	138%	3.2 ms/1.2 ms	2018	[552]
Phototransistor	graphene/GaN	10	5.83 (325 nm)	> 2.20 × 103%	2.7 ms/4.6 ms	2018	[553]
Nanorods	Al0.45Ga0.55N	3	0.115 (250–276 nm)	53.84%*	–	2017	[519]
Nanowires (Schottky)	GaN	1	25 (357 nm)	8.69 × 103%*	–	2013	[518]
Nanoflowers (MSM)	Al0.45Ga0.55N	2	0.72 (265 nm)	340%	–	2018	[508]

Zoom In Zoom Out Reset image size

Poor current conduction has also prevented the achievement of electrically pumped QW lasers operating in the UV-B (280–315 nm) and UV-C (200–280 nm) bands^[566-570]. Because they are still being extensively investigated in research laboratories around the globe, the development of DUV LDs is still in its infancy stage, as these researchers almost exclusively attempt to exploit the Al_xGa_1−xN material system^{[571, 572]}. To realize an LD, energy transfer into the gain medium of the laser from an external source to produce excited states in the atoms (referred to as the laser pumping process) is necessary. The laser pumping method can be classified into two categories:

1. Electrical pumping: LDs are naturally pumped using electric currents. Researchers have demonstrated electrically pumped lasers at relatively long emission wavelengths, such as 330 nm^[573] and 342 nm^[574], and also at 275 nm^[575]. The former two operate within the UV-A region, while the latter emits a UV-C light suitable for germicidal applications. The challenges for realizing electrically injected DUV LDs using Al_xGa_1−xN MQW structures include the high densities of dislocations in the lattice structure, large spontaneous and piezoelectric polarization fields, and poor p-type doping characteristics. These issues hinder improvement in the DUV LD EQE and the output power of electrically injected Al_xGa_1−xN MQW-based DUV LDs. Some researchers have focused their attention on nanowire-based LDs because of the potentially high crystalline quality that they can achieve, the high LEEs, and the low strain-induced polarization exhibited in nanowire structures^[576]. Random lasers based on nanowire arrays have been demonstrated at peak lasing wavelengths of 239 nm^[577] and 262.1 nm^[578]. However, their optical output powers are still considerably low, while peak emitting wavelengths are mostly unstable because the laser cavities are defined through random multiple scattering.

2. Optical pumping: Pumping LDs using an optical pumping source is an artificial way to operate a laser, although dye lasers are optically pumped^[579-583]. DUV LDs emitting at the sub 250 nm regime have been demonstrated^[584-586], but producing light using this method would require another optical power source with higher optical power outputs and shorter excitation wavelengths, rendering this mechanism impractical in real-life applications.

Given these hindrances, conventional optically pumped DUV LDs and electrically pumped random DUV LDs are impossible to practically exploit in vital applications, such in the study of geology and climate in outer space. An alternative way to drive an LD active material is by using an electron beam, through which electron–hole pairs are generated by the energy transfer of high-energy electrons emitted from an electron gun. One advantage of this approach is that the incorporation of p- and n-type doped layers is not necessary, avoiding the undesired optical absorption commonly associated with these layers. In 2015, Klein et al. demonstrated high-power green and blue electron-beam-pumped LDs, with a peak optical powers of 5.9 W ( $\lambda =$ 350 nm) and 3.3 W ( $\lambda =$ 462 nm) at moderate electron energies of 42 keV and 37 keV, respectively^[587]. In 2010, Oto et al. achieved DUV emission from electron-beam-excited AlN-based QW optical source with a peak emission at 240 nm, peak optical power of 100 mW, and pumping power and current of 10 kV and 45 $\mu$A, respectively^[588]. The success shown in these reports manifests a great potential of electron-beam-pumped LDs in achieving high power and short wavelength DUV laser devices.

Kataoka, Funato, and Kawakami developed 3D Al_0.50Ga_0.50N QWs on MOVPE-grown AlN templates for polychromatic DUV LEDs^[591]. They attempted a maskless regrowth of AlN on a trench-patterned AlN template procedure to create the AlN multi-facets of (0001), { $1\bar{1}01$}, and vicinal (0001) with bunched steps^[592]. Eventually, multi-faceted p–i–n structures, including three periods of Al_xGa_1−xN/Al_yGa_1−yN active region, were grown on the AlN template, which led to different emission wavelengths as demonstrated in Fig. 22(a). Under DC injection current at room temperature, two polychromatic emission bands at 250 and 275 nm were demonstrated, with a tendency similar to those observed by Funato et al. for 3D In_xGa_1−xN polychromatic emitters^[593]. Without any patterning or further synthesis of high-level-features AlN template, Kaneda et al. used the continuous high-temperature (CHT) growth method before growing Al_xGa_1−xN-based DUV LEDs^[594-597]. With increased miscut angles of 0.2°, 0.3°, and 10 for sapphire substrate, the MOVPE AlN template exhibited lower TDDs and denser macrosteps as illustrated in Fig. 22(b). The flip-chip bonded DUV LEDs grown on a miscut angle of 1.0° demonstrated 20% higher output power, compared with the devices on 0.3° miscut. They also experimentally proved that the enhanced IQE performance was caused by the reduction in nonradiative recombination channels at the MQWs and the compositional modulations in the MQWs by virtue of the macrosteps of the AlN template^{[598, 599]}. As an alternative to MOVPE grown AlN templates, Susilo et al. and Miyake et al. investigated sputtered AlN templates with high temperature annealing (HTA)^{[600, 434]}. Sputtered AlN provide an avenue to achieve cost-efficient AlN templates with low TDDs^{[432, 601-611]}. In addition, HTA can provide beneficial effects on the grain of buffer layers at the AlN layer/substrate interfaces by eliminating tilt components between buffer layer grains, achieving TDDs down to 4.7 × 10⁸ cm^–2 for 2.3 $\mu$m-thick MOVPE-grown AlN layers^[433]. In addition to their lower production costs and reduced substrate curvature, Susilo et al. showed that sputtered thin HTA AlN (≤750 nm) template exhibited lower TDDs (7.2 × 10⁸ cm^–2) as compared to epitaxially laterally overgrown (ELO) AlN/sapphire using MOVPE (1.1 × 10⁹ cm^–2). Their DUV LED devices (fabricated on the sputtered AlN templates) demonstrated DUV emission at around 268 nm, with output powers in the range of 0.65–0.72 mW at 20 mA, which is comparable to DUV LEDs fabricated on MOVPE-grown AlN templates^[600]. Several MOVPE complex growth approaches were attempted, including lateral overgrowth of nitrogen-polar AlN domains with aluminum-polar AlN^[612] and high-V/III-ratio-induced roughening^[613]. Typical AlN layer TDD values were around 1 × 10⁹ cm^–2, which are desirable but not adequate for the realization of high-efficiency DUV LEDs^[409].

Zoom In Zoom Out Reset image size

Chang et al. demonstrated the growth of high-quality AlN templates on NPSSs via graphene-assisted quasi-van der Waals epitaxy^[614-617]. This process allowed for rapid film coalescence, reducing growth times. Because of the presence of graphene, AlN films tend to be 2D laterally expanded on the NPSS, resulting in reduction of TDD (from 5.5 × 10⁹ cm^–2 to 1.5 × 10⁹ cm^–2 on the surface for layer thickness of 1.5 $\mu$m) as well as strain release in the AlN epitaxial layer, which were attributed to air voids above the nanopattern that were formed during the epitaxial growth process. The 2D growth nature of AlN in this case caused the TDs near the air voids to bend and ultimately annihilate at the end of the void. EL measurements of the fabricated DUV LEDs utilizing the graphene interlayer have shown considerable improvement in emission characteristics at 280 nm.

Gallium oxide and its ternary alloys ((In_xAl_yGa_z)₂O₃, x + y + z = 1 ) are wide-bandgap oxides of post-transitional metals with ionic bonds. With superior breakdown fields of around 8 MV/cm compared with SiC (3 MV/cm) and GaN (3.8 MV/cm), their intrinsic band gap spans from 4.9 to 5.3 eV; there are five known polytypes of Ga₂O₃, α, β, γ, δ, and ε^[618]. Among these, β-Ga₂O₃ is the most stable structure with a superior melting point of up to 1820 °C; it became available as a wafer substrate by using conventional melt growth techniques. Despite its large bandgap ranging from 4.7 to 5.0 eV, its electrical properties can be tuned from an insulator to an n-type semiconductor with carrier concentration of up to 10¹⁹ cm^–3 for DUV device applications^{[619, 620]}. In contrast, $\alpha$-Ga₂O₃ has not yet been studied as a metastable phase of Ga₂O₃. With a highest energy bandgap of around 5.3 eV, its rhombohedral corundum structure allows for its utilization on conventional sapphire substrates as opposed to using it as a wafer. In addition, alloys of $\alpha$-Al₂O₃, $\alpha$-Ga₂O₃, and $\alpha$-In₂O₃ can make multinary oxide compounds from each other. Eventually, they can make unique physical properties with other corundum systems of transition metal oxides ( $\alpha$-M₂O₂; M = Fe, Cr, V, Ti, Rh, and Ir)^[281].

High n-type conductivity and transparency characteristics reveal that Ga₂O₃ is an ideal candidate for the realization of DUV devices as a transparent conducting oxide (TCO) substrate. Even though Ga₂O₃ does not have a hexagonal structure, it is relatively easy to induce a hexagonal atomic arrangement for optoelectronic device applications. In 2015, epitaxial growth of planar GaN layer on (100) a-plane $\beta$-Ga₂O₃ was first demonstrated^[621]. Low-temperature GaN buffer layer was first introduced, and a 1 $\mu$m-thick GaN layer was subsequently grown at 1070 °C by MOVPE. From the X-ray diffraction measurement, $\theta$–2 $\theta$ scan confirmed a wurtzite GaN layer with c-plane orientation, and it showed 1200 arcsec full width at half maximum (FWHM) rocking curve. Room temperature photoluminescence measurements showed an emission peak wavelength at around 362.6 nm, and eventually, a blue LED device was successfully fabricated. In 2006 and 2007, GaN layers grown on $\beta$-Ga₂O₃ were also demonstrated by ammonia- and RF-MBE, respectively^{[622, 623]}. In MBE epitaxial growth, nitridation was a crucial factor for planar GaN layer growth at the interface with $\beta$-Ga₂O₃ substrate. In ammonia-MBE, nitridation was processed under NH₃ condition inside the growth chamber to reconstruct the surface from a twofold symmetry a-plane $\beta$-Ga₂O₃ to a sixfold symmetry of c-plane GaN. Effective nitridation step helped to obtain a transparent and mirror-like GaN surface layer on Ga₂O₃; however, GaN has a rough surface and tends to peel off from the substrate when the nitridation step is skipped^[622]. The surface of $\beta$-Ga₂O₃ was also successfully controlled by nitridation in RF-MBE^[623]. At the interface, GaN phase was transformed from cubic to wurtzite by increasing the nitridation time. This thin GaN layer served as the basis for growth of subsequent epitaxial GaN layers on the $\beta$-Ga₂O₃ substrate. In 2012, another research group further investigated c-plane GaN layers on (100) a-plane $\beta$-Ga₂O₃ that could easily be peeled off from the substrate^[624]. Initially NH₃-treated 150 $\mu$m-thick GaN layers showed limited adhesion at their interfaces. As a result, the self-separation of GaN layer was achieved at room temperature, directly after the hydride vapor phase epitaxy (HVPE) growth process.

As part of a GaN epitaxial layer growth study on $\beta$-Ga₂O₃ substrates, an Al_xGa_1−xN layer was demonstrated by MOVPE, and its crystalline quality was compared with that of grown on sapphire. The Al_0.08Ga_0.92N and GaN growth processes were carried out on low-temperature buffer layers^[625]. An optimized facet-controlled method and thermal annealing process resulted in high-quality GaN and Al_0.08Ga_0.92N layers on $\beta$-Ga₂O₃ substrates with enhanced morphology, photoluminescence intensity, and reduced FWHM of the X-ray diffraction rocking curve. Other than the (100) plane, the growth of GaN/Al_0.20Ga_0.80N MQWs on ( $\bar{2}01$)-oriented $\beta$-Ga₂O₃ was also investigated by Ajia et al.^[626]. While GaN layer growth on (100) $\beta$-Ga₂O₃ resulted in an uncontrollably peeled off film from the substrate, Ajia et al. demonstrated that GaN/Al_0.20Ga_0.80N growth on ( $\bar{2}01$)-oriented $\beta$-Ga₂O₃ can provide a stable layer for the more operationally stable vertical device applications. With a 2 nm-thick low-temperature AlN buffer layer, a 100 nm-thick n-Al_0.75Ga_0.25N layer was grown at 1020 °C, followed by a 900 nm thick n-Al_0.30Ga_0.70N layer at 1120 °C. Subsequently, 3× GaN/Al_0.20Ga_0.80N MQWs were grown for their study. As compared to films grown on sapphire substrates, the GaN/Al_0.20Ga_0.80N MQWs on the ( $\bar{2}01$)-oriented $\beta$-Ga₂O₃ showed higher structural and optical crystalline quality as well as lower density of nonradiative recombination centers. These results hold promise of unleashing the potential of UV transparent and conductive Ga₂O₃ substrates as a suitable alternative to group III–nitride-based substrates for the realization of DUV device applications.

Polymeric materials, such as optically isotropic and stable amorphous fluorine resin^{[627, 628]} and graphene oxide (GO)-based fluoropolymer composites^[629], may constitute a suitable substitute to other materials for the mass-production and packaging of Al_xGa_1–xN-based UV and DUV LEDs because of their decomposition characteristics below 330 nm^{[410, 457, 627, 628]}. DUV LED dies are generally flip-chip bonded on submounts with gold bumps, and polymer-based materials can be used between the submount and dies as underfilling materials, as well as for the rest of entire surfaces (as an encapsulation). Yamada et al. revealed that instead of conventional resins, C-F-based polymers, so called S-type amorphous fluorine resins (terminated with a trifluoromethyl), are ideal candidates for the underfilling and encapsulating materials for the DUV LEDs^[410]. Fig. 22(c) shows the temperature-programmed desorption mass spectroscopy analysis for S- and A-type resin with an inset indicating the schematics of flip-chip LEDs with underfilling and encapsulation. From this experiment, S-type resin tests showed no peaks after 261 nm light irradiation. However, A-type resin showed CO, CO₂, and H₂O peaks after DUV irradiation, which means decomposition of materials. With the optically isotropic characteristics of S-type resin for encapsulation, it showed no degradation over 3000 h of operation time.

Increasing demand of miniature-sized LD are of importance for the optical communications and data storage industries. One of the possible approaches for the realization of nanoscale lasers is using surface plasmon polaritons, which contain strong localized fields at the dielectric/metal interfaces. Recently, Shen et al. demonstrated DUV lasers with hyperbolic metamaterial (HMM) including multiple stacks of dielectric (20 nm of MgF₂)–metal (20 nm of aluminum) interfaces^{[630, 631]}. Those specially designed hyperbolic metacavity arrays helped to achieve a high-quality photonic density of state (PDOS), and eventually all excited plasmon oscillations coupled in one lasing mode. Focused ion beam (FIB)-fabricated 200 nm² hyperbolic metacavity arrays were formed on MOVPE-grown high aluminum-content Al_xGa_1–xN MQW active regions, and the top of the structures were capped with annealed aluminum films as reflectors. Using a 266 nm pulsed laser as an exciting source, plasmonic nanolaser was demonstrated at lasing peak of 289 nm with narrow linewidth of 0.8 nm as depicted in Fig. 22(d). The lasing threshold is shown in the inset of Fig. 22(d).

In addition to UV light-emitting devices, researchers have also focused their efforts on advancing the integration of new materials for the realization of DUV photodetection. As a new approach for this application, the interesting characteristics exhibited at the interfaces of 2D TMDs and 3D group III–nitride semiconductors have recently emerged as the central theme of various studies^[632-636]. The fabrication of 2D platinum diselenide (PtSe₂)/GaN heterojunctions for self-powered DUV photodetectors has also been reported^[559]. Group III–nitride semiconductor materials and 2D TMDs can be bonded through the weak out-of-plane van der Waals interactions in the absence of surface dangling bonds on their surfaces^[637], providing an easy path for the integration of MX₂/group III–nitride interfaces for the optoelectronic applications, where M = a transition metal (Mo, W, Nb, Ta, Ti, Re) and X = S, Se, Te^{[638, 639]}. Specifically, the direct bandgap^[640-642] 2D layered molybdenum disulfide (MoS₂) is the most promising candidate for this purpose and widely explored for PD applications^{[560, 638, 643-649]}. With a relatively small lattice mismatch of 0.8% between MoS₂ and GaN layer, free-dangling bond surface of MoS₂ provides excellent heterojunctions^[637]. Zhuo et al. investigated p–n MoS₂/GaN heterojunctions as self-powered PDs and developed a detection device for 265 nm DUV light operating at zero bias voltage. The naturally formed band bending at the MoS₂/GaN interface results in the alignment of the Fermi levels of MoS₂ and GaN^[650], inducing a built-in electric field that enables the device operation at zero bias voltage. Moreover, the strong interaction of the MoS₂ layer with incident light manifested excellent photoresponsivity to DUV light, with demonstrated high responsivity of 187 mA/W, high specific detectivity of 2.34 × 10¹³ Jones, and high ON/OFF current ratio of over 10^5[560].

The state and band parameters at semiconductor heterointerfaces are essential to the effective design of active regions in contemporary electronic and optoelectronic devices^{[331, 635, 651]}. Band discontinuities present at heterojunctions, formed by adjoining two different semiconductor materials, have unequal bandgap energies that act as potential barriers, which play an important role in controlling the charge carrier transport properties^[652]. The feasibility of charge carrier injection across such heterointerfaces relies on the interface quality and the band-offset parameters (i.e., CBOs and valence band offsets (VBOs))^{[653, 654]}. Based on these parameters, heterostructures are classified based on three alignment types: straddling gap (referred to as type I), staggered gap (type II), and broken gap (type III) junctions^[655]. In general, type I band alignments with high refractive index contrasts are the most favorable for electron and optical mode confinement, thereby facilitating the carrier recombination process in the active regions of Al_xGa_1–xN-based optoelectronic devices. The other two alignment types are necessary to realize other electronic devices, such as hot electron transistors, provided they exhibit reasonably low CBOs, which allows for improved tunneling probabilities and lower turn-on voltages^[656]. Researchers have extensively employed X-ray photoelectron spectroscopy (XPS) to determine band offset parameters and interface alignment types formed by group III–nitrides using their bandgap energy values. Several studies can also be found on the band offset parameters of dissimilar heterojunctions formed between compound III–nitrides and other families of semiconductor materials, such as oxides and 2D materials^{[635, 657]}. For instance, King et al. determined the band alignment type for InN/AlN and InN/GaN heterojunctions to be type I heterointerfaces using high-resolution XPS (HRXPS) measurements and the bandgap properties of the constituent materials^{[658, 659]}. The corresponding band alignment schematic is shown in Fig. 23. Table 4 summarizes a thorough literature survey of band offset parameters for compound III–nitrides-based and hybrid heterostructures. It is noteworthy that irrespective of the polarity and crystal structure, all group III–nitride semiconductors exhibit type I heterojunctions, which satisfies the fundamental requirement to design optoelectronic devices. However, although the nature of type I heterojunctions at wide bandgap compound III–nitrides interfaces facilitates carrier and optical mode confinement, there are other bottleneck issues, such as line defects, structural cracking, and low p-type doping efficiencies, hindering the overall device efficiencies.

Zoom In Zoom Out Reset image size

Table 4.  Summary of group III–nitride, III–oxide, and hybrid material heterointerface properties.

Heterojunction structure	VBO (eV)	CBO (eV)	Heterojunction type	Measurement method	Ref.
Wurtzite GaN/AlN	0.7 ± 0.24	–	I	XPS	[660]
Cubic GaN/AlN	0.5 ± 0.1	1.4 ± 0.1	I	Theory (ab initio)	[661]
Cubic AlN/GaN (zincblende (001) SLs)	1.02	0.68	I	Theory (Quasiparticle)	[204]
Wurtzite InN/AlN	1.52 ± 0.17	4.0 ± 0.2	I	XPS	[659]
	1.81 ± 0.2	–	I	XPS	[660]
a-plane GaN/AlN	1.33 ± 0.16	–	I	XPS	[662]
a-plane AlN/GaN	0.73 ± 0.16	–	I	XPS	[662]
Semi-polar AlN/GaN	0.7 ± 0.2	2.1 ± 0.2	I	XPS	[663]
InN/GaN	0.58 ± 0.08	2.22 ± 0.1	I	HRXPS	[658]
	0.85	1.82	I	Internal photoemission	[664]
	1.04, 0.54	–	I	XPS	[665]
	1.05 ± 0.25	–	I	XPS	[660]
	0.5 ± 0.1	2.2 ± 0.1	I	XPS and PL	[666]
	1.07	1.68 ± 0.1	I	Photocurrent spectroscopy	[667]
	0.72 ± 0.28	–	–	XPS	[668]
InAl0.83N/GaN	0.2 ± 0.2	1	I	XPS	[669]
In0.17Al0.83N/GaN	0.15	–	I	XPS	[670]
In0.25Al0.75N/GaN	0.1 ± 0.2	0.4	I	XPS	[399]
In0.3Al0.7N/GaN	0.0 ± 0.2	0.2	I	XPS	[399]
AlxGa1–xN/GaN	0.3x	–	I	Theory (tight binding)	[671]
B0.14Al0.86N/GaN	0	2.1 ± 0.2	I	HRXPS	[672]
B0.14Al0.86N/Al0.70Ga0.30N	0.4 ± 0.05	0.1 ± 0.05	II	HRXPS	[673]
InN/( $\bar{2}01$) $\beta$-Ga2O3	–0.55 ± 0.11	–3.35 ± 0.11	I	HRXPS	[674]
ITO/( $\bar{2}01$) $\beta$-Ga2O3	–0.78 ± 0.3	–0.32 ± 0.13	I	HRXPS	[675]
ITO/(010) $\beta$-(Al0.14Ga0.86)2O3	–1.18 ± 0.2	0.32 ± 0.05	I	HRXPS	[676]
Al2O3/(010) $\beta$-(Al0.14Ga0.86)2O3	0.23 ± 0.04	1.67 ± 0.3	I	HRXPS	[677]
Wurtzite InN/h-BN	–0.3 ± 0.09	4.99 ± 0.09	II	XPS	[678]
GaN/single-layer MoS2	1.86 ± 0.08	0.56 ± 0.1	II	HRXPS	[636]
In0.15Al0.85N/MoS2	2.08 ± 0.15	0.6 ± 0.15	I	HRXPS	[635]

From the ABC model, the IQE is directly proportional to $\eta_ {\rm {inj}}$, where $\eta_ {\rm {inj}}$ is influenced by the efficiency of electron and hole injection^{[679, 680]}. Considering the effective masses of charge carriers (m_e,h) in group III–nitride materials, it is evident that electrons are more mobile in the crystal lattice than holes^[681-683]. The traditional approach of impurity-induced doping of III–nitrides suffers from the large acceptor activation energy of magnesium, thereby resulting in low p-type doping efficiency in Al_xGa_1−xN. Hence, electrons leak into the p-type Al_xGa_1−xN layers of the optoelectronic devices, causing reduced IQE^{[162, 684, 685]}. Strain-induced piezoelectric fields, along with spontaneous polarization fields, can have values in the order of several MV/cm, which lead to separation of electron and hole wavefunctions. Consequently, increased carrier separation results in the prevailing of QCSE, which has a detrimental effect on device efficiencies by reducing the radiative recombination^{[287, 686, 687]}. Nonetheless, using the polarization-induced charges, researchers reported graded heterojunctions of Al_xGa_1−xN, which establishes a 3D slab of bound charges, either an n- or p-type conducting layer, depending on the direction of the aluminum concentration gradient and the polarity of the material^[688]. Jena et al. showed the polarization-induced bulk electron doping in III–nitride semiconductors by grading the Al_xGa_1−xN/GaN (0 ≤x≤ 0.3) heterojunctions over a distance of ≈ 100 nm^[689]. The polarization-induced bulk charge, also referred to as 3D electron slap (3DES), mimics a local donor with zero activation energy, overcoming the localized optical transitions associated with the defect centers of donor atoms. Such carriers are not rendered thermally inert at low temperatures, unlike the shallow donor-doped bulk carriers. Thus, absence of ionized impurity scattering results in high mobilities of charge carriers; consequently, improved carrier injection efficiency can be achieved. Subsequently, Simon et al. reported ionization of the magnesium acceptor dopants utilizing built-in polarization fields in bulk wurtzite semiconductors^[178]. These field-ionized holes can oppose thermal freeze-out effects, resulting in major improvement in p-type electrical conductivity. This doping scheme helped improve optical emission efficiency in UV LED structures. Thus, polarization-induced doping renders a solution to enhance carrier injection efficiency in wide-bandgap Al_xGa_1−xN semiconductors. In contrast, the incorporation of an Al_xGa_1−xN layer as an electron-blocking layer (EBL) results in polarization-induced electron gas at the Al_xGa_1−xN/GaN interface, leading electrons to leak toward the p-type GaN layer. To overcome this leakage issue, polarization inverted EBL was proposed, but inverting the polarity during the growth remained challenging^[690]. Later, Kuo et al. replaced the GaN last quantum barrier (QB) with an In_0.01Ga_0.99N/GaN heterojunction, which aided in the development of an upward band bending at the last GaN barrier, resulting in an effective electron blocking by the EBL^[691]. Despite the difficulty in growing In_xAl_1−xN^[692], In_0.20Al_0.80N acting as an EBL that was well lattice-matched with GaN reduced the piezo-polarization effect and electron leakage^[693].

With experimental evidence, Zhang et al. computationally demonstrated that silicon step-doping of quantum barriers would lower polarization-induced fields and reduce energy barriers for hole transport^[694]. They compared three devices differing solely in terms of doping in barrier layers. Device I is a standard LED with unintentionally doped barriers, while devices II and III consist of silicon-doped layers and step-doped barriers, respectively, as shown in Fig. 24(a). Figs. 24(b) and 24(c) respectively show the experimentally measured and numerically simulated optical output power and EQE as functions of injection currents for devices I, II, and III. Considering the entire current range, device III outperformed I and II. The silicon step-doping suppressed QCSE and enhanced hole injection, consequently improving LED performance.

Zoom In Zoom Out Reset image size

With the advancement of technology, there is a colossal demand for high power and highly efficient UV optoelectronic devices because of their diverse application (water/air/food sterilization, surface disinfection, free-space non-line-of-sight communication, epoxy curing, counterfeit detection, fluorescence or Raman identification of biological/chemical agents, and various diagnostic and therapeutic medical functions)^{[695, 402]}. With such a diversity in potential applications, as exciting as it may sound, UV optoelectronic devices are far from perfected. Thus, there is a plenty to improve in materials system (choice of material, synthesis/growth), device structure (fabrication), and packaging.

EQE of a light-emitting device (Fig. 25(a)) is the product of electrical injection efficiency (EE), IQE, and LEE, and typical values are shown in Fig. 25(b)^[696]. EQE for the UV LEDs and LDs has been improved over time by tweaking materials or structures; however, it is still noticeably lacking^[402]. One of the major bottlenecks for abysmal EQE is poor current injection. In this section, we highlight the techniques utilized to improve current injection of UV optoelectronic devices.

Zoom In Zoom Out Reset image size

Current injection is affected by numerous factors: metal contacts, interfaces in between contact material and electron/hole supplier layers, concentration of electron/hole in electron/hole supplier layers, electrical conductivity, and electric field (piezoelectric field) developed across the active region. Both p- and n-type doping became challenging in DUV Al_xGa_1−xN-based devices due to high aluminum concentrations^[697]. High ionization energy of magnesium and lower mobility worsen the hole injection, which creates imbalance in charge carrier injection^{[698, 91]}. Thus, research problems, achievements, and prospects on the enhancement of hole injection are worthy of discussion. The key factors that require attention for the enhancement of hole injection are outlined in Fig. 26.

Zoom In Zoom Out Reset image size

When a device is electrically biased, the very first hurdle that a hole encounters is the interface of the p-type metal contact and the hole supplier layer^[699]. In 1999, Ho et al. demonstrated a low specific contact resistance to p-type GaN where a Ni/Au thin film was deposited on p-type GaN and annealed at 500 °C in air to form NiO and obtain specific contact resistances as low as 4 × 10^–6 $\text {&#x3a9;}$ ^[700]. This technique is one of the popular techniques to obtain low specific contact resistances to date. Other metal combinations utilized in enriching p-type Ohmic contacts include zinc-nickel (ZnNi)/indium tin oxide (ITO)^[701], ruthenium (Ru)/Ni/iridium (Ir)^[702], and Ti/Au^[699]. In addition, contact interfaces can be improved by manipulating the polarization effect and adopting superlattice structures^[703] and tunnel junctions^[704-706].

In the majority of UV LED devices, a hole supplier layer consists of p-type GaN and Al_xGa_1−xN layers. As a hole reaches the hole supply layer, it encounters a high discontinuity in the valance band at the heterojunction, which makes the hole transport challenging. Graded heterojunctions^[707], Al_xGa_1−xN/ Al_yGa_1−yN (x > y) superlattice structures^[708-710], polarized induced electric fields^{[178, 711]} (as shown in Fig. 27), and magnesium-delta doping^{[416, 180]} are some of the major techniques employed to enhance the hole transport within the hole supplier layer, lowering the activation energy of magnesium dopants to as low as 17 meV. In addition, drift velocity can be significantly increased utilizing the strong field produced at the interface (for aluminum-rich p-Al_xGa_1−xN and p-GaN) of heterojunction in the hole supply layer^[712].

Zoom In Zoom Out Reset image size

Electron leakage under high bias currents is inevitable. An EBL is not only thought to be capable of minimizing electron leakage, but also hindering hole transport because of the discontinuity of the valance band^{[652, 707, 713]}. Mehnke et al. implemented AlN electron blocking heterostructures (EBHs) in Al_xGa_1−xN-based MQW LEDs with peak emission below 250 nm^[714]. Owing to the EBHs, the emission from the QW was enhanced, whereas long wavelength parasitic luminescence was significantly reduced. Consequently, adequate hole injection was maintained by optimizing thickness of the EBH. Using such device structure, UV-C LED with peak emission at 234 nm, optical power of 14.5 $\mu$W, and EQE of 0.012% at current density of 18.2 A/cm² was demonstrated. AlN is optically transparent for DUV (> 210 nm) and often used as a QB layer in Al_xGa_1−xN-based MQW LEDs. However, carrier injection is severely hindered by poor conductivity of AlN. Thus, there is a common practice of doping the QBs. Tsai et al. revealed a DUV LED device structure that utilizes an intrinsic QB on high-quality AlN template^[715]. It is confirmed that the output power was improved as a result of restricted diffusion of magnesium tunneling into the MQW region and suppression of the sub-band parasitic emissions. To exploit the intraband-tunneling assisted hole injection, Zhang et al. proposed a p-type EBL with a very thin insertion layer having smaller energy bandgap^[712].

Holes in the active regions of visible LEDs tend to accumulate close to the interface between the MQW region and the p-type EBL; however, this is not the case in DUV LEDs^{[716, 717]}. Furthermore, hole distribution in the active region can be modified through manipulation of aluminum composition in Al_xGa_1−xN QWs and QBs. It has been observed that band offset is reduced, which is beneficial for hole transport across the active region^[717]. Enhancement of hole concentration in each QW appears more easily accomplished than uniform hole distribution across an entire active region. The thicknesses of the QWs and QBs also affect hole concentration. Fewer than a couple nanometers of QW thickness and less than 8 nm for QB are the thicknesses recommended for the high hole concentrations and excellent spatial overlap of the electron–hole wavefunctions^[718-721].