BaTiO₃/Al_0.58Ga_0.42N lateral heterojunction diodes with breakdown field exceeding 8 MV/cm

The Al_xGa_1−xN material system has garnered significant research interest due to its attractive material properties such as an extremely high critical breakdown field, predicted to be above 9 MV/cm for x ≥ 0.5, good low-field transport properties, and high electron saturation velocity comparable to GaN.^1–4 As a result of these properties, Al_xGa_1−xN has a higher predicted Johnson Figure of Merit,

compared to GaN (v_sat is the saturated electron velocity and F_BR is the critical breakdown field). This can enable ultra-scaled devices, which can surpass the state of the art GaN-based amplifiers in terms of current density, breakdown voltage, and output power density at mm-wave and THz frequencies.^2–4 In addition, the large critical breakdown field of AlGaN could also enable more favorable device characteristics for normally off devices used in power switching applications.⁵ 

While various figures of merit suggest the excellent potential of these materials for future device technology, approaching the breakdown limit in an actual device remains a standing challenge. Material breakdown limits for wide-band gap semiconductors have been reached previously only with PN junctions.⁶ Achieving the theoretical material breakdown field in lateral diodes and field effect transistors is challenging due to non-uniform electric field distribution in the depletion region (which causes electric field peaking) and is further limited by tunneling breakdown at the Schottky gate or anode electrode. These effects are more significant in the case of ultra-wide band gap semiconductors, which have average breakdown fields estimated to be higher than 7 MV/cm. AlGaN devices that employ metal–semiconductor based Schottky anodes or gates are limited by the breakdown strength of the Schottky diodes, which typically have barrier heights of less than 2 eV. Breakdown fields up to 3.9 MV/cm have been demonstrated for Al_0.70Ga_0.30N/Al_0.50Ga_0.50N high electron mobility transistors (HEMTs) and 2.86 MV/cm for Schottky gate electrodes for Al_0.70Ga_0.30N metal semiconductor field effect transistors (MESFETs).^7,8 However, these numbers are still significantly lower than the estimated material breakdown field (11 MV/cm) at this composition. Conventional dielectrics such as SiO_X and Al₂O₃ inserted between the anode or gate metal and semiconductor do not offer a significantly better performance since they are limited by the breakdown of the metal–dielectric junction.

A solution to this problem, proposed recently, is to use extreme permittivity oxides inserted between the metal and semiconductor layers and between the gate and drain region.^9,10 Extreme permittivity dielectrics enable improved breakdown due to two reasons. First, a high dielectric constant layer greatly reduces the peaking or non-uniformity in the lateral electric field between the gate to drain for lateral transistors or anode to cathode for lateral diodes and results in a higher breakdown voltage and higher average breakdown field due to a more uniform electric field distribution. This is illustrated in Fig. 1, which shows a comparison of the total electric field and potential distribution profiles between the anode and cathode of a heterojunction diode in the semiconductor (channel) near the semiconductor–dielectric interface (inset of Fig. 1) with a 25 nm low-k dielectric (ε_r = 10) and a high k dielectric (ε_r = 100) for an applied anode to cathode voltage of 150 V and an anode to cathode spacing of 0.2 μm. The channel material assumed here is AlGaN with a sheet charge density, n_s, of 1.5 × 10¹³ cm⁻². The addition of the high permittivity dielectric expands the depletion region due to the negative polarization charge induced by the field. Therefore, the voltage drop occurs over a larger gate-drain depletion width, thereby decreasing the field in the gate-drain region. In addition to reducing peak fields due to depletion, the high permittivity material below the gate or anode metal helps to reduce leakage, which suppresses gate or anode leakage related breakdown. This can be understood by considering the continuity of the displacement field across the dielectric/semiconductor junction, which suppresses the electric field within the high dielectric constant material.

Barium titanate, BaTiO₃, a perovskite oxide displaying a high dielectric constant, is a suitable candidate for such applications. In this report, we demonstrate a >8 MV/cm average breakdown field for BaTiO₃/Al_0.58Ga_0.42N (predicted critical breakdown field of Al_0.58Ga_0.42N ∼9.7 MV/cm) heterojunction diodes, which represents an important step toward achieving a true material breakdown field in lateral devices. In comparison, control Schottky diodes displayed an average breakdown field of ∼4 MV/cm, which is significantly lower than the material breakdown field limit and representative of typical observations for Schottky-based devices.^7,8

Low-pressure (LP) Metal-Organic Chemical Vapor Deposition (MOCVD) was used to grow the active layers for the reported device structures on 3 μm thick high quality AlN-(0001) sapphire templates. These template layers were also deposited by LPMOCVD at growth temperatures close to 1250 °C. Their RMS surface roughness was measured using an atomic-force microscope (AFM) to be 1.44 nm, as shown in Fig. 2(a). The epilayer consisted of a 500 nm thick undoped i-Al_0.58Ga_0.42N buffer layer grown pseudomorphically over the entire 2-in. diameter AlN template at a growth temperature close to 1100 °C. This was followed by the growth of a 60 nm thick [Si⁺]-doped n-type Al_0.58Ga_0.42N layer, with a doping concentration of 4 × 10¹⁸ cm⁻³ on top of the buffer layer as reported elsewhere.¹¹ High-resolution x-ray diffraction (XRD) spectra (BEDE D1 High Resolution Triple Axis X-ray Diffraction System) [Fig. 2(b)] were used to determine the composition.

Selective area regrowth by molecular beam epitaxy (MBE) was used to achieve Ohmic contact to the MOCVD grown epilayer. 500 nm of silicon dioxide (SiO₂) was deposited and patterned on the MOCVD-grown channel layer to act as a hard mask for MBE contact regrowth. Heavily [Si⁺]-doped Al_0.58Ga_0.42N (50 nm) was then regrown in the selectively formed pits using MBE to completely fill the etched regions. This was capped with a heavily [Si⁺]-doped reverse Al-composition graded AlGaN layer of thickness of 50 nm and a doping of 1 × 10²⁰ cm⁻³. The purpose of the reverse Al-composition graded n⁺⁺ AlGaN layer is to minimize abrupt conduction band offsets and facilitate low resistance non-alloyed Ohmic contact formation.^7,12

Following MBE contact regrowth, the SiO₂ regrowth mask was removed using a diluted buffered oxide etch (BOE) solution, and a metal stack of Ti/Al/Ni/Au (20/120/30/100 nm) was deposited on the regrown contact regions via an e-beam evaporator to form Ohmic contacts. This was followed by device isolation by an inductively coupled plasma-reactive ion etching (ICP-RIE) system with an RIE power of 30 W and a pressure of 5 mTorr. BaTiO₃ was deposited via RF sputtering at 630 °C in an oxygen ambient. The dielectric constant of the BaTiO₃ films was estimated to be around 60. The dielectric constant of BaTiO₃ is a function of the grain size of the deposited film. For the deposition conditions used, the deposited film is expected to be amorphous or nano-crystalline.¹³ The observed dielectric constant of the deposited films is consistent with previous reports.¹⁴ For the fabrication of the control samples, BaTiO₃ was etched away completely using SF₆ plasma ICP-RIE etching. For the heterojunction diodes, BaTiO₃ was etched away in the contact regions only. Anodes were formed by the deposition of a Pt/Au (60/100 nm) metal stack via e-beam evaporation for both the control (Schottky) and the heterojunction diodes. The schematic of the fabricated heterojunction diode is shown in Fig. 3(a). Figure 3(b) shows the equilibrium energy-band diagrams calculated using a self-consistent one-dimensional Poisson–Schrödinger solver, under the anode, for the BaTiO₃ sample.¹⁵ The conduction band offset and Schottky barrier height were estimated from the known electron affinity and work function values for the materials.^16,17 The schematic and energy band-diagram of the control (Schottky) device, under the anode, are shown in Figs. 3(c) and 3(d), respectively.

Hall measurements were performed on four-terminal ungated van der Pauw (VDP) structures without BaTiO₃ and indicated a sheet resistance of 6.2 kΩ/□, a Hall mobility of 65 cm²/V s, and a sheet charge density of 1.55 × 10¹³ cm⁻². Electrical characterization of the devices was performed using an Agilent B1500 parameter analyzer. The inset of Fig. 4 shows the capacitance voltage characteristics measured on a 16 μm × 100 μm heterojunction diode. The integrated charge density, n_s, was found to be 1.06 × 10¹³ cm⁻². The measured charge density and the simulated charge density are shown in Fig. 4. The extracted doping density is found to be 4 × 10¹⁸ cm⁻³, as expected, near the top of the structure, and the lower density at larger depletion depths is likely due to errors introduced by capacitor loss.

Figure 5(a) shows the forward characteristics of a Schottky diode and a heterojunction diode with an anode to cathode spacing of 700 nm. The turn-on voltages of the Schottky diode and the heterojunction diode are 0.9 V and 1.5 V, respectively. Both these devices display a significantly lower turn-on voltage compared to what is typically observed for AlGaN PN/PIN diodes. This suggests that the BaTiO₃/AlGaN interface presents relatively small barriers to electron transport, in contrast to conventional metal-oxide-semiconductor junctions where a large voltage would be needed to achieve turn-on of the current. The differential on-resistance (R_ON), calculated from the forward IV characteristics, was 31 mΩ cm² for the Schottky diode and 56 mΩ cm² for the heterojunction diode. The increase in R_ON of the heterojunction device is likely due to the presence of the BaTiO₃ layer, which is nominally undoped. Breakdown characteristics were evaluated for several randomly selected Schottky and heterojunction diodes. Figure 5(b) shows the breakdown characteristics for a Schottky diode and a heterojunction diode, with anode to cathode spacings of 0.19 μm and 0.18 μm, respectively, as measured using a scanning electron microscope (SEM). While the Schottky diode breaks down at an anode-to-cathode voltage of 42 V, the heterojunction diode breaks down at a much higher voltage of 155 V, despite both devices having a similar anode to cathode spacing. The improvement in the average breakdown field is attributed to a reduced peak field and the suppression of anode leakage current due to the BaTiO₃ layer. Assuming a total depleted sheet charge density of 1.55 × 10¹³ cm⁻², the peak vertical field (see the supplementary material for details) was estimated to be 3.15 MV/cm. Assuming a pinch-off voltage of −12 V, the average lateral field of the Schottky diode was estimated to be 1.58 MV/cm (see the supplementary material), which gives a total average field of 3.5 MV/cm (see the supplementary material). The peak vertical field and the average lateral field of the heterojunction diode were estimated to be 3.15 MV/cm and 7.94 MV/cm, respectively, leading to an average total breakdown electric field of 8.5 MV/cm. The vertical field should be the same in both devices since the total number of ionized donors in the depletion region should be the same for both devices. Multiple randomly chosen devices were measured for both the Schottky diodes and the heterojunction diodes, and the results are shown in Fig. 6(a). In general, for the heterojunction diodes, for an anode to cathode spacing in the range of 0.2–0.3 μm, the average breakdown fields observed were in the range of 6–8 MV/cm. Figure 6(b) shows the highest reported average breakdown fields for various semiconductor materials as a function of the bandgap. The value reported in this work, 8.5 MV/cm, is the highest reported experimental average breakdown field for any semiconductor material.

The breakdown voltage achieved for the diode with an average breakdown field of 8.5 MV/cm was 155 V, which is lower than what is expected from the theoretically estimated maximum breakdown field of 9.7 MV/cm and a flat electric field distribution. Therefore, there exists room for optimization of this BaTiO₃/AlGaN heterostructure design including optimization of the BaTiO₃ thickness and growth conditions together with the exploration of potential integration of field plate structures.

One important benefit of using BaTiO₃ as a gate or anode dielectric is that it enables the fabrication of ultra-scaled devices while meeting the breakdown voltage requirements due to the much higher breakdown fields achievable in these structures. This is particularly beneficial for low mobility materials like AlGaN and Ga₂O₃, which require scaled devices for high performance RF electronics to compensate for the higher sheet resistance compared to GaN. The presence of the high-k dielectric also leads to scaling of the gate to channel distance, which can result in lower output conductance and higher transconductance.¹⁰ However, it should be noted that in such a heterojunction with a high-k dielectric, the drain depletion region is longer than the case with a low-k dielectric and a correspondingly higher gate to channel capacitance due to the high-permittivity region between the gate and drain. As a result, the peak cutoff frequency is expected to be lower than a device with a low-k barrier.¹⁰ It is therefore important to properly optimize the thickness of the high-permittivity region to manage the trade-off between field management and gate–drain capacitance. Scaling, enabled by the high-k heterostructure devices, is also expected to play an important role in next generation power electronics, whereby the on-resistance of the device can be significantly reduced due to shortened device dimensions while at the same time delivering higher output power more efficiently.

In conclusion, we have demonstrated Pt/BaTiO₃/Al_0.58Ga_0.42N lateral heterojunction diodes with enhanced breakdown characteristics. By using BaTiO₃ as an anode and an anode to cathode region dielectric, the peak field at the anode edge was significantly reduced and an average breakdown field exceeding 8 MV/cm was achieved for devices with an anode to cathode spacing of <0.2 μm. In contrast, Pt/Al_0.58Ga_0.42N control Schottky diodes displayed an average breakdown field of ∼4 MV/cm for devices with similar dimensions. The use of a high-k dielectric can more effectively utilize the high breakdown fields in ultra-wide bandgap materials by proper electric field management. This demonstration thus provides a framework to realize ultra-scaled lateral devices with improved breakdown characteristics.

See the supplementary material for a description of the estimation method of the average breakdown electric field.

The authors acknowledge funding from the Air Force Office of Scientific Research (AFOSR Grant No. FA9550-17-1-0227, Program Manager Kenneth Goretta) and the DARPA DREaM program (No. ONR N00014-18-1-2033, Program Manager Dr. Young-Kai Chen, monitored by the Office of Naval Research, Program Manager Dr. Paul Maki).