Device fabrication and characterization of epitaxial heterostructures. To design an effective unipolar barrier APD device, both interfacial lattice compatible and energy band offsets need careful considerations. By taking account of our previous experimental results28 (Supplementary Fig. S1), MgO was integrated with Ga2O3 to form a unipolar barrier structure possessing a large conduction band offset△EC and nearly zero valance band offset △EV. We fabricated an nBn unipolar barrier APD based on the as-prepared Ga2O3/MgO/Nb:STO heterostructure using laser molecular beam epitaxy (LMBE) assisted by in-situ reflection high-energy electron diffraction (RHEED) (Fig. 1a, growth details in Methods Section). A Ga2O3/Nb:STO heterostructure was also fabricated for APD devices performance comparison. The evolution of the RHEED specular spot intensity enables us to monitor the quality and the thickness of both MgO and Ga2O3 layers at atomic scale. Figure 1b shows the typical RHEED patterns for a Ga2O3/MgO/Nb:STO heterostructure. The streaky RHEED patterns observed at the end of the deposition for both MgO and Ga2O3 layers verify flat interface and surface. Atomic force microscopy (AFM) indeed reveals a surface topography with a root-mean square (rms) roughness values less than 1.3 nm, as shown in Supplementary Fig. S2. Such uniform and well-defined interface could therefore minimize the interfacial defects/dislocations and benefit the carrier flow cross the heterostructure. The thickness of Ga2O3 (200 nm) and MgO (25 nm) layers were confirmed by both ellipsometer and cross-sectional scanning electron microscopy (Fig. 1c). The X-ray diffraction (XRD) θ-2θ scan manifests sharp (l00) diffraction peaks of β-Ga2O3 and MgO, indicating that both Ga2O3/MgO/Nb:STO and Ga2O3/Nb:STO heterostructures are of epitaxial form (Fig. 1d). And the out-of-plane relationship is β-Ga2O3[600]//MgO[200]//Nb:STO[200]. Off-specular Φ-scan was then conducted to investigate the in-plane film-substrate alignment. As shown in Fig. 1e, the oblique Ga2O3 \(\left\{710\right\}\), MgO \(\left\{220\right\}\) and Nb:STO \(\left\{220\right\}\) Bragg reflections appear at coinciding angles and separated azimuthally by 90◦, revealing that the Ga2O3 lattice is rotated by 45o on the MgO (100) surface. Thus, the in-plane relationship of the unipolar heterostructure could be assigned to β-Ga2O3[001]//MgO[011]//Nb:STO[011]. For monoclinic structured β-Ga2O3, the oxygen atoms along [010] and [001] directions are arranged approximately as equilateral squares, with spacing of 0.304 nm and 0.29 nm, respectively29. On the other hand, the oxygen atoms spacing along [011] and [01_1] directions in cubic structured MgO and Nb:STO have the same squared arrangement with spacing of 0.298 nm and 0.276 nm, respectively. Therefore, when the Ga2O3/MgO/Nb:STO nBn unipolar heterostructure is constructed, Ga atoms in the β-Ga2O3 (100) plane could bond to the oxygen atom layer in the MgO/Nb:STO (100) plane with 45o rotation. A schematic atomic picture of the cross-sectional heterostructure and the interfacial oxygen-atom arrangements are depicted in Fig. 1f and 1g. Thus, the exquisite control offered by the LMBE assisted by in-situ Rheed has enabled the direct integration of epitaxial Ga2O3/MgO heterostructure on Nb:STO.
Unipolar barrier calculations. We conducted X-ray photoelectron spectroscopy (XPS) to explore the band diagram of the Ga2O3/MgO/Nb:STO and the Ga2O3/Nb:STO heterostructures (Fig. 2a and Supplementary Fig. S3)30. According to Kraut’s method, the △EV could then be determined by analyzing the binding energy difference between the valence band maximum (VBM) and the core-level shifts (see Method Section)28. The obtained △EV for the Ga2O3/Nb:STO is 0.46 eV, while the △EV for the Ga2O3/MgO/Nb:STO are \(\varDelta {E}_{V-{Ga}_{2}{O}_{3}/MgO }\)= 0.12 eV and \(\varDelta {E}_{V-MgO/Nb:STO}\) = 0.45 eV, respectively. Then, considering the bandgap (Eg) values for Ga2O3 (4.94 eV), MgO (7.83 eV), and Nb:STO (3.2 eV), the calculated conduction-band offset (△EC) for the Ga2O3/Nb:STO is 1.28 eV. Importantly, the obtained valance band arrangements for Ga2O3/MgO/Nb:STO are \(\varDelta {E}_{C-{Ga}_{2}{O}_{3}/MgO}\) = 3.01 eV and \(\varDelta {E}_{C-MgO/Nb:STO}\) = 4.18 eV, respectively. Therefore, the band diagrams for both the Ga2O3/MgO/Nb:STO and Ga2O3/Nb:STO heterostructures in equilibrium condition are illustrated in Fig. 2b and Supplementary Fig. S4. The challenging issue for building Ga2O3-based heterojunctions is the intrinsic large bandgap of Ga2O3, which limits the selection of dielectrics as the designed barrier. The bandgap of MgO can reach up to 7.83 eV, which is much larger than that of Ga2O3 (4.94 eV), favoring the construction of a rectifying junction with large band offset. More intriguingly, with an elaborate design, the as-prepared nBn-type heterojunction exhibits desired unipolar barrier characteristics, i.e., an extremely high △EC and a negligible △EV across the Ga2O3/MgO/Nb:STO heterointerfaces. The carrier transport behaviors of the heterostructures are controlled by the energy band structure. The general purpose of unipolar barrier design is blocking one type of carrier while allowing the flow of the other. In the case of APDs presented here, the large △EC can effectively block the electron transfer from Nb:STO to Ga2O3, which is able to solve the bottleneck of dark current and further improve the performance of APDs. APDs work in a high reverse bias condition and therefore avalanche multiplication is successfully occurred as designed when the excess carriers are formed upon the UV photon impact. The band alignment facilitates the seperation and migration of photoexcited carriers. With increasing the reverse bias, photon-induced charge carriers are accelerated and undergo cascade amplifacations through impact ionization in the depletion layer. Ionization coefficients of both electrons and holes increase along with the electric field in the depletion region. Therefore, higher reverse voltage generally contribute to higher gain. The large breakdown electric field (~ 8 MV/cm) endows Ga2O3 a born figure-of-merit (FOM) as excellent potential material for APDs. However, so far the breakdown electric field of reported Ga2O3-based APDs is limited by the barrier height of the rectifying junction, which is much less than the often quoted theoretical breakdown field of Ga2O3. Very recently, we constructed an amorphous Ga2O3/ITO APDs with enlarged barrier height (~ 2.07 eV), resulting in dramatically improved reverse bias voltage and photoresponse21. Herein, the enlarged △Ec between MgO and Nb:STO is further estimated up to 4.18 eV, promising larger tolerance of higher reverse bias. Next, we consider the distinctive role of unipolar barrier design influencing the avalanche progress under working condition (Fig. 2c). Note that owing to the electron concentration in Nb:STO (~ 1020 cm-3) is much larger than that in unintensitional doped Ga2O3 (1016-1017 cm-3), the depletion region is mainly in Ga2O3 side. Under UV light irradiation, charger carriers are acceleraed under high electric fields and hence more electron-hole pairs are ionized (Fig. 2d). Under avalanche breakdown, the extensive electrons are swept to the anode, while a negligible △EV across the heterojunctions facilitates the generated holes flowing unimpeded. The ideal APDs should separate electrons and holes as efficiently as possible, with minimal relaxation of charge carriers from the avalanche multiplication. Unipolar barrier construction filters out photocurrent components on demand, rather than aggregating them at the interfaces, which can reduce the adverse recombination rate. Furthermore, the holes aggregation would give rise to an electric field counteracts the externally applied field, and impair the avalanche process.
Unipolar barrier APDs. Figure 3a schematically presents the device of Ga2O3/MgO/Nb:STO nBn unipolar barrier APD and Ga2O3/Nb:STO n-n isotype APD. As the nBn unipolar barrier heterostructure is designed to suppress dark current and enhance avalanche gain, carrier transport behaviors in dark were first investigated, as shown in Fig. 3b. To prevent unrecoverable breakdown in both APDs, 5 µA corresponging to 500 µA/mm2 is set as the limit current. Obvious rectifying behaviors are observed in the current-voltage (I-V) curves of both two types of heterostructures. Consistent with the proposed transport mechanism as discussed above, the dark current in the nBn unipolar barrier heterostructure is much smaller than that in the n-n isotype heterostructure, especially under a reverse bias. Note that the dark current in the Ga2O3/MgO/Nb:STO heterostructure remained below 0.06 nA (0.006 µA/mm2) under 50 V reverse bias, where the dark current in the Ga2O3/Nb:STO already reached 5 µA (500 µA/mm2). Thus, the large conduction band barrier exhibits significant inhibition of dark current over more than one order of magnitude, which provides great advantages for improving the internal gain and detectivity of APD. Figure 3c demonstrates the dark current and photocurrent (under 0.1 µW/cm2 254 nm light irradiation), as well as the calculated gain versus reverse voltage for the two types of APDs. Here, the unmultiplied current at 1 V was assigned as the unity gain reference. It is found that the avalanche breakdown threshold voltages (calculated onset avalanche electric fields) are 31 V (1.55 MV/cm) and 71 V (3.16 MV/cm) for the Ga2O3/Nb:STO and the Ga2O3/MgO/Nb:STO heterostructures, respectively. The avalanche gain (M) is determined using the following relation: M = [IPh(V)-Id(V)]/[IPh(0)-Id(0)], where IPh(V) and Id(V) are the multiplied photocurrent and dark current, whereas IPh(0) and Id(0) are the unmultiplied photocurrent and dark current, respectively23. With increasing the reverse voltage, the avalanche gain values keep increase exponentially, as shown in Fig. 3c (right axis). Compared to the maximum M value of 5.30 × 104 at 43.6 V (2.18 MV/cm) in the Ga2O3/Nb:STO n-n isotype APD, the maximum M value for the Ga2O3/MgO/Nb:STO nBn unipolar barrier APD increased by one order of magnitude and reaches as high as 5.86 × 105 at 78.1 V (3.47 MV/cm), which is the record-high avalanche gain value reported in Ga2O3-based solar-blind APDs. Graphically, the “performance squarensee” defined by three key parameters of APD including dark current, reverse workind voltage and gain is significantly amplified for unipolar barrier APD. Thus, order of magnitude improvements in both dark current and gain fullfill our unipolar barrier design intention. Temperature dependent threshold voltages investigation presents a positive temperature coefficient of 0.026 V/°C, indicating that the underly mechanism of the breakdown is attributed to avlanche effect instead of Zener tunneling effect (Supplementary Fig. S5). To access the ultraweak light detection ability of the APD device, several critical FOMs, specific detectivity (D*) and linear dynamic range (LDR) were later quantitatively evaluated using the following equations: D* = RS1/2/(2eId)1/2 and LDR = 20×log(IPh/Id), respectively31,32. Here, R represents the responsivity21, which could be calculated by R = (IPh-Id)/(P×S). P, S and e represent incident light intensity, effective irradiation area and electron charge, respectively. As displayed in Fig. 3d, the maximum D* increases from 1.22 × 1016 Jones to 3.28 × 1017 Jones, and the maximum LDR increases from 21.7 dB to 81.8 dB for Ga2O3/Nb:STO and Ga2O3/MgO/Nb:STO heterostructures, respectively. The enhancement of D* and LDR values indicate that the MgO layer introduced through deliberate lattice and band engineering play a major role in the Ga2O3/MgO/Nb:STO nBn unipolar barrier APDs, leading to excellent ultraweak light detection with low noise and linear responsivity for a wide range of light intensities32.
The detailed photoresponse performance of the Ga2O3/MgO/Nb:STO nBn unipolar barrier APD was then systematically investigated. Figure 4a displays the I-V curves of the device measured under different light intensity under reverse bias, all revealing outstanding photoresponse behaviors. The R and D* as functions of light intensities and applied bias are presented in Fig. 4b and Fig. 4c, respectively. It is found that with an increase in light intensity, both R and D* of the device decrease, which could be explained by the increased probability of electron-hole recombination under high intensity light illumination. The responsivity can reach up to 4.46 × 105 A/W under 0.1 µW/cm2 light irradiation at -78.1 V, which is among the best performance of Ga2O3-based solar-blind photodetectors2,6,12. The photoresponse performance of the Ga2O3/Nb:STO n-n isotype APD was also studied for comparison (Supplementary Fig. S6-S8). The spectral response of the device in the range of 200–700 nm was studied, as shown in Fig. 4d. The responsivity reaches its maximum value at ~ 260 nm with a cutoff edge at ~ 280 nm, revealing an excellent solar-blind region spectral selectivity. Response speed is another important parameter for the APD. To assess the temporal response speed of the APD, the transient photoresponse signal was measured using a coherent KrF 248 nm pulse laser combined with a 500 MHz oscilloscope, as depicted in Fig. 4e. By fitting the transient response curves using \(I={I}_{0}+{A}_{1}{e}^{-t/{\tau }_{1}}+{A}_{2}{e}^{-t/{\tau }_{2}}\) Eq. 6, the obtained rise/decay time τr/τd are 12.4 ns and 41.7 µs, respectively. So fast response speed could be understood by the rapid electron-hole separation and transport owing to the enhanced avalanche electric fields and dispelling holes at the interface. The stability and reliability of the nBn unipolar barrier APD were further examined by periodic turn on/off the UV lamp. As displayed in Fig. 4f, the APD retained reliable photoresponse characteristics even after 104 on/off cycles.