Introduction

Since the discovery of two-dimensional materials in 20041, this emerging material category has captured the attention of scientists owing to their array of superlative properties. Hexagonal boron nitride (hBN) is a forerunner for a range of applications in relation to its superior thermal conductivity, chemical stability, high remote optical phonon energies, and flat surface without dangling bonds2,3,4. For these reasons, hBN is considered at the very least to be a suitable next-generation dielectric applications, providing high resistance and enhanced thermal dissipation while exerting minimum reduction in channel mobility for semiconductor devices ranging from Si logic to GaN HEMT power devices that are known to particularly suffer from joule heating effects5.

Research groups are also now exploring whether hBN can be used as an active material beyond dielectric and thermal dissipation applications. Recently it has been shown that hBN can be p-type doped using Mg (Ea~30 meV), and recently, n-type has been demonstrated through the use of Ge-O doping, and later combined with p-GaN forming a pn heterojunction device6. Furthermore, hBN is a efficient DUV emitter ×500 greater than AlN which suffers from a low internal quantum efficiency (< 3%) at < 250 nm and is used for commercial solid-state DUV emissive devices7,8,9. For example, hBN has a reported room-temperature quantum efficiency of ∼60% at UV-C wavelengths10.

Along with AlGaN, diamond, and boron nitride, gallium oxide (Ga2O3) is also one of the fourth-generation ultra-wide bandgap (UWBG) materials with a 4.8 eV band gap and is a promising candidate for several classes of power electronics, solar cells, solar-blind UV photodetectors, and even gas sensors11,12,13,14,15.

Specifically, Ga2O3 has as high critical electric field (Ecrit >8 MV/cm) and an impressive Baliga figure of merit value of 3124 (relative to Silicon) permitting Ga2O3-based power devices in achieving reduced power losses16 compared to conventional semiconductor materials. Importantly, device quality Ga2O3 can be grown heteroepitaxially on sapphire substrate facilitating technology scaling17,18,19. Unfortunately, Ga2O3 performance is hindered by low thermal conductivity, resulting in faster device degradation. Like many wide bandgap materials, Ga2O3 suffers from asymmetric doping where n-type is achievable, however p-type doping is non-trivial14. Though β-Ga2O3 is an indirect bandgap material, it possesses a direct gap only 29 meV higher, suggesting DUV emission maybe possible if high quality pn junctions can be fabricated20.

This naturally leads one to consider the formation of hBN/Ga2O3 heterostructures where the aforementioned challenges can be overcome owing to their complementary material properties.

In this article, we show that Mg doped hBN (ambient wet-transfer process) on n-type Ga2O3 can lead to the formation of high quality stable pn junction diodes. To the best of our knowledge, this is the first report of a pn heterojunction formed between multilayer p-type hBN and n-type Ga2O3.

Methods

hBN growth

A homemade atmospheric pressure chemical vapor deposition (APCVD) system with three heating zones was used for hBN growth (Figure S2a). The system includes two internal quartz tubes, with each tube hosting ammonia borane (99.7% purity) in Zone 1 and Mg3N2 dopant precursor (99.6% purity) in Zone 2. Cu foil (99.7% purity, 30 μm thick) was electrochemically polished using a standard process21. Next, Cu was annealed in an Ar/H2 ambient (300 SCCM, 4 N purity) for 1 h at 1000 °C for enlarged Cu grains at 93.3 kPa. Figure S1 illustrates the time-temperature process in each reaction zone during growth. Vacuum base pressure was 7\(\:\:\times\:{10}^{-2}\) Pa prior to growth.

During hBN growth, ammonia borane (30 mg) is held at 90 °C and a zone 3 temperature of 1050 °C is used. The doping precursor (30 mg) was rapidly ‘turned-on’ reaching 850 °C after \(\:\sim\)5 minutes of hBN growth (from a set point of 500 °C). Growth duration was 1 h, then Ar/H2 was used as a protective gas and the system was allowed to naturally cool to room temperature. hBN film thickness is 10 nm. Thereafter, we apply a standard wet chemical process for hBN transfer to Ga2O322.

n-type Ga2O3growth

The heteroepitaxial Ga2O3 layer on c-plane sapphire is 400 nm thick grown by metal organic chemical vapour deposition. Tetraethoxysilane (TEOS) gas was used for in-situ n-type doping during growth, further details are published here17. The electron carrier concentration is \(\:\sim\)1 \(\:\times\:\) 1018 cm−3 determined by Hall measurement.

Device fabrication

The pn junction is fabricated through a sequence of lithographic, etch and lift-off process steps. Inductiviely coupled reactive ion etching (ICPRIE) was performed using Ar gas. A Ti/Al/Ni/Au metal stack with 20/300/40/50 nm thicknesses were deposited by E-Gun on an n-type Ga2O3 epilayer followed by rapid thermal anneal to obtain an Ohmic contact. Then, Pt was deposited on hBN similarily to obtain the p-electrode contact (Fig. 3a). The fabrication process flow is shown in Fig. 1a, device structure and final layout are shown in Fig. 1b,c.

Fig. 1
figure 1

(a) pn junction diode fabrication process, (b) schematic of device structure and (c) the optical microscope (OM) image of the fabricated device (the scale bar is 400 \(\:\mu\:m\)).

Full size image

Physical analysis and electrical test

Morphological and thickness measurements were performed by scanning electron microscopy (SEM), optical microscopy (OM) and atomic force microscope (AFM). X-ray photoelectron spectroscopy (XPS) was used to determine B: N stoichiometric ratio. Raman spectra (JadeMat DPL 532 nm laser) was obtained with a 100 mW laser and 20 s integration time. X-ray diffraction (XRD, (Cu Kα, λ = 1.54 Å)) was used to confirm hBN crystallinity. Ultraviolet photoelectron spectroscopy (UPS) was performed using a ULVAC-PHI 5000 Versaprobe II system (He UV light source) with an energy resolution of 0.02 eV (beam diameter 300 μm) to locate Fermi level position.

The Mg acceptor level was identified through temperature dependent Hall effect analysis on a reference sample transferred onto sapphire substrate. Capacitance-Voltage (C-V) and Current-Voltage (I-V) analysis is performed by a Solartron 1260 A impedance system and a HP4140B, respectively. An ezHEMS system with a 1 T field switching capability was employed for Hall effect analysis (Van der Pauw configuration).

Results and discussion

In Fig. 2a–d, XRD, XPS, Raman and UV-Vis data is presented. Figure 3a shows a singular hBN diffraction peak from the (002) plane indicating good crystallinity (bottom panel). For the Mg doped sample, (top panel) a similar result is obtained. For hBN doped and undoped material, diffraction peaks occur at 26.0 ° and 26.19 ° respectively. This may be linked to induced crystal strain as Mg atoms occupy B sites22,23,24. Full width half maximum (FWHM) values of intrinsic and doped hBN register at 1.4 ° and 1.45 ° respectively. Intrinsic hBN possesses triangular domains (lateral edge) that vary from 2 to 30 \(\:\mu\:m\)25,26,27 depending on growth conditions25. In Fig. S2b Mg doped hBN domains are similarily triangular in nature and have an \(\:\sim\)1.5 \(\:\mu\:m\) lateral edge, in other regions values up to 6 \(\:\mu\:m\) are seen. A nucleation density of \(\:\sim\)104/mm2 is calculated. These values are \(\:\times\:10\:\)higher than our undoped hBN material. SEM analysis was performed after hBN doping on Cu foil (reference sample), confirming a clean surface devoid of particles (Fig. S2c). In Fig. S2d an AFM image illustrates the Mg doped hBN layer on Ga2O3 substrate. Surface roughness is approximately 0.8 nm. Residual photoresist is also observed.

Fig. 2
figure 2

(a) XRD spectrum of hBN (b) Raman spectrum (c, d) XPS spectra of B1s and N1s of Mg doped hBN (e) UV-VIS Tauc plot (f) UPS analysis of Mg doped hBN.

Full size image
Fig. 3
figure 3

(a) I-V measurement of Mg doped sample (the insert shows transferred Mg doped hBN on sapphire for Hall effect analysis). (b) Carrier concentration versus 1000/T of Mg doped hBN on sapphire.

Full size image

To confirm Mg incorporation, Raman spectroscopy was performed. In Fig. 3b, (bottom panel) a single E2g vibrational mode for undoped hBN is recorded at 1370 cm− 1. For the Mg doped sample (top panel) a 2nd mode appears at 1357 cm− 1. This may relate to Mg incorporation on B sites as similar secondary peaks at 1347 cm− 1 and 1363 cm− 1 are reported in22,28.

In Fig. 2c,d the XPS spectra of Mg doped hBN is presented. The B1s and N1s binding energies are 190.5 and 398.4 eV, in agreement with published values and the N/B ratio is 1.0622. UV-VIS absorption is used to evaluate bandgap and impurity incorporation. Figure 2e shows a clear band edge absorption peak for both samples at \(\:\sim6.0\:\)eV. In addition, a shoulder appears that we ascribe to Mg impurity incorporation.

In Fig. 2f, UPS demonstrates that EF is 1.8 eV from valence band edge after a 5 s 1 kV Ar sputter taken from Mg doped hBN reference sample on Cu foil grown under indentical conditions. The EF position is deeper than expected, and is clearly p-type, compared to our previous works where EF is midgap for intrinsic hBN material26. This maybe linked to doping induced defects or Fermi level pinning (FLP) effects which are known to occur in 2D materials. For example, EF in n-type 2D hBN is 1.9 eV from Ec as determined by UPS including a built-in voltage of 3.11 V when combined with p-GaN6.

Electrical measurements (I-V) were performed to confirm sample conductivity change after doping, shown in Fig. 3a. Intrinsic hBN has a resistance \(\:>\)109 Ω cm29 reducing to 0.36 Ω cm after Mg incorporation (Four point probe method). An Arrhenius plot is used in Fig. 3b to study the influence of temperature on carrier concentration and to extract an activation energy of 60 meV, similar to Ea ’s reported in7,22,30. A carrier concentration of 7 \(\:\times\:\:\)1017 cm−3 and a hole mobility of 24.8 cm2/Vs is measured at 300 K including a positive Hall voltage.

The I-V characteristics of the p-hBN/n-Ga2O3 junction measured from 300 to 475 K can be observed in Fig. 4a. A rectification ratio (RR) of 2.6\(\:\times\:\)104 at 300 K and \(\:\sim\)400 at 475 K are recorded and are among the highest values seen for ultra-thin hBN based pn junctions. A cut-in voltage (Vcut) of 2.05 V is extrapolated. Leakage current under reverse bias (−10 V ) is 0.2 µA at 300 K rising to 2\(\:\times\:\)10−4 A at 475 K, increasing with reverse bias voltage. This suggests that a current leakage path exists owing to the presence deep defect levels in the depletion layer. In Fig. 4b we plot saturation current I0 (-5 V) versus 1000/T where a 0.44 eV activation process is found, which may correlate with defects within hBN31. At forward bias, all I-V curves display a high RON originating from either a high hBN contact resistance, which is not optimized or a lateral current pathway in in n-type Ga2O3 (ND\(\:\sim\) 1\(\:\times\:\)1018 cm−3) which typically has low mobility. Ga2O3 ohmic contact resistance is typically < 10−5 ohm cm218. S. Majety32 demonstrated BN/AlGaN pn junctions with comparable properties. Here, leakage current was 3 µA at -10 V, however these BN films were considerably thicker, similarly in24.

Fig. 4
figure 4

(a) Semi-log plot of I-V as a function of temperature (b) I0 plotted against 1000/T at −5 V (c) I-V measurement before and after high temperature measurements at 300 K (d) C-V at different frequencies and 1/C2 plot (secondary y-axis).

Full size image

After cool-down, the device was remeasured to confirm its stability, shown in Fig. 4c and current saturation in the linear I-V plot (> 7 V ) is owing to measurement system limitation. The pn junction remains highly functional, and rectification has increased to 2.6 \(\:\times\:{10}^{5}\). Diode RON before and after temperature ramping is 2\(\:\times\:\)103 Ω reducing to 400 Ω respectively. We attribute these improvements to reduced contact resistance. Likewise, a number of other devices were also tested at 300 K producing RRs of \(\:\sim\) 104, leakage currents < 1 \(\:\mu\:\)A and RON between 200 and 400 Ω. Owing to the large series resistance, the ideality factor is obtained using Shockley equation in the sub-threshold region only (Table 1). High values of 3.77 reducing to 2.0 at 475 K suggests thermal activation is occurring as resistance reduces at elevated temperatures. Alternatively, high values that reduce with increasing temperature have been linked to tunneling through ultra-thin barriers33.

Table 1 Extracted ideality factors as a function of temperature.
Full size table

C-V measurements are plotted in Fig. 4d from 1 to 6 MHz after temperature ramping at 300 K. The forward voltage is limited to 4 V as beyond this point diffusion capacitance dominates. The measured capacitance reduces with increasing reverse bias voltage indicating a widening of the depletion region. Under forward bias, capacitance increases owing to space charge accumulation at the depletion edge. At lower modulation frequencies total capacitance increases consecutively. This behaviour is typically associated with trapping/detrapping behaviour in the depletion region as at higher frequencies these states cannot follow the AC signal. The calculated junction capacitance is 0.18 nF in close agreement with those values depicted in Fig. 4d at zero bias.

In Fig. 4d built-in potential (Vbi) and doping concentration (NA) can be determined assuming an abrupt pn+ junction (Eq. 1). Extrapolation from 1/C2 yields a \(\:{v}_{bi}\) of 2.34 V from the linear gradient, which is close to the cut-in voltage. The 1/C2 gradient is non-linear implying a non-uniform doping profile at the pn junction interface. A net NA concentration of 7.0 \(\:\times\:\)1017 cm− 3 has been extracted from the slope of the linear segment when ND is 1.0 \(\:\times\:\)1018 cm− 3 (Ga2O3 doping level). These findings are in agreement with Hall effect measurements.

$$\:\frac{1}{{C}^{2}}=\:\frac{2({V}_{bi}-V-\:\frac{{k}_{B}T}{q})\left({\epsilon\:}_{BN}{N}_{A}+{{\epsilon\:}_{Ga}N}_{D}\right)}{q{A}^{2}{N}_{A}{N}_{D}{{{\epsilon\:}_{BN}{\epsilon\:}_{Ga}}_{\:}}_{\:}}$$
(1)
$$\:{\varnothing\:}_{B,\:eff}^{C-V}={v}_{d0}+\:{V}_{p}$$
(2)

(ΔEv < ΔEc34)

$$\:{\varnothing\:}_{B}^{\:}=\frac{{k}_{B}T}{q}\text{l}\text{n}\left(\frac{{A{A}^{*}T}^{2}}{{I}_{0}}\right)$$
(3)

Where A is the effective junction area (5.8\(\:\times\:{10}^{-8}\) m2), V is the applied voltage, \(\:\epsilon\:\) is the permittivity of hBN (\(\:{\epsilon\:}_{r}=3.5)\) and Ga2O3 (\(\:{\epsilon\:}_{r}=9.5)\) and NA is the net hole concentration. Richardson constant A* is taken as 3.12 × 105 Am−2 K−235.

For barrier height (φB) elucidation36, \(\:{v}_{d0}\) = \(\:{v}_{bi}+\:\left(\raisebox{1ex}{$kT$}\!\left/\:\!\raisebox{-1ex}{$q$}\right.\right)\) denotes the diffusion voltage at zero bias, and \(\:{v}_{p}\) = (\(\:{k}_{B}\)T⁄q)\(\:ln\left(\raisebox{1ex}{${N}_{v}$}\!\left/\:\!\raisebox{-1ex}{${N}_{A}$}\right.\right)\) represents the theoretical Fermi level position above the valence band in a neutral region of the semiconductor (Eq. 2). Taking Nv as 2.1\(\:\times\:{10}^{19}\) cm−337, φB equates to 2.47 eV in close agreement with \(\:{v}_{bi}\). φB has also been evaluated using Eq. 3, here the barrier height is recorded at a lower value of 0.86 V. Differences in barrier heights from I-V and C-V methods are well known and have been linked to effects such as interfacial barrier height inhomogeneity, deep defect levels or interfacial charge38,39.

The intrinsic hBN/Ga2O3 band alignment (direct growth) has been studied by XPS33,40 revealing a staggered type-II band alignment. Using Anderson’s method where the electron affinity of hBN and Ga2O3 are 2.0 eV and 4.0 eV26,41, respectively, we estimate the theoretical Vbi as 2.2 eV in agreement with the 1/C2 value (Fig. 5). The larger ΔEc should hinder the movement of electrons from Ga2O3 into hBN suggesting current transport may be dominated by holes moving from p-hBN to n-Ga2O3.

Fig. 5
figure 5

Band diagram of the p-hBN/n-Ga2O3 junction by Anderson’s method.

Full size image

Conclusion

Pseudo-vertical high performing hBN/Ga2O3 pn junctions have been demonstrated. Rectification ratios > 105 and leakage currents < 0.2 \(\:\mu\:\)A are measured at 300 K and devices continue to operate reliably after high temperature ramping to 475 K. C-V analysis indicates device performance follows classical pn junction behaviour as bias voltage varies. The hBN carrier concentration is \(\:\sim\)7\(\:\times\:\)\(\:{10}^{17}\) cm−3 in reasonable agreement with C-V analysis. UPS analysis suggests FLP effects are present, and the theoretical Vbi is in good agreement with the measured value. We expect further performance gains as hBN doping and growth improves.