An avalanche-and-surge robust ultrawide-bandgap heterojunction for power electronics

Avalanche and surge robustness involve fundamental carrier dynamics under high electric field and current density. They are also prerequisites of any power device to survive common overvoltage and overcurrent stresses in power electronics applications such as electric vehicles, electricity grids, and renewable energy processing. Despite tremendous efforts to develop the next-generation power devices using emerging ultra-wide bandgap semiconductors, the lack of effective bipolar doping has been a daunting obstacle for achieving the necessary robustness in these devices. Here we report avalanche and surge robustness in a heterojunction formed between the ultra-wide bandgap n-type gallium oxide and the wide-bandgap p-type nickel oxide. Under 1500 V reverse bias, impact ionization initiates in gallium oxide, and the staggered band alignment favors efficient hole removal, enabling a high avalanche current over 50 A. Under forward bias, bipolar conductivity modulation enables the junction to survive over 50 A surge current. Moreover, the asymmetric carrier lifetime makes the high-level carrier injection dominant in nickel oxide, enabling a fast reverse recovery within 15 ns. This heterojunction breaks the fundamental trade-off between robustness and switching speed in conventional homojunctions and removes a key hurdle to advance ultra-wide bandgap semiconductor devices for power industrial applications.

The built-in potential (Vbi) is determined to be 2.1 V from the intercept of 1/C 2 -V plot on the x-axis [6]. The depth profile of net doping concentration (ND -NA) in the n-Ga2O3 drift layer is extracted from C-V curve, and the average net (ND -NA) is determined to be about 1.7 × 10 16 cm -3 , which was used in the TCAD simulation.  (GW INSTEK), and a printed circuit board for different tests. The high-voltage probe PCB5500B (5 kV, 500 MHz) from Zhiyong Electronics with a 100-fold voltage attenuation has been employed to record the voltage waveforms. Both the coaxial shunt (SSDN-414-10) and current probes (TCP0030A: 120 MHz, 0 ~ 30 A and TCP0150: 20 MHz, 0 ~ 150 A) have been used to measure the current waveforms. To maintain measurement accuracy, the current probe was degaussed before measurements.

Avalanche characterizations
The device under test (DUT) was reversely biased and connected in parallel with a high-BVAVA SiC MOSFET [7]. In the test, the load inductor (LUIS) was first charged by the power supply (VCC) with the ONstate SiC MOSFET. Once the SiC MOSFET was turned OFF, the energy stored in LUIS was forced to go through the DUT, driving it into the breakdown state, whereby the voltage and current waveforms were captured by the oscilloscope. By controlling a programmable pulse generator to adjust the ON/OFF state of the SiC MOSFET, the DUT could be driven into breakdown once-only or repeatedly. The SiC MOSFET is 1.7-kV rated (Wolfspeed, C2M0045170D); it did not avalanche due to its higher avalanche breakdown voltage. LUIS with values from 50 µH to 30 mH were varied to modulate IAVA and avalanche energy. The DC bus voltage (VDC) was set to be 50~300 V.

Surge current characterizations
The DUT was forwardly biased and connected in series with a 72-A rated SiC MOSFET (Wolfspeed, C2M0045170D) [8]. In the test, a 10-ms-wide half-sinusoidal surge current was generated by the resonant circuit module and passed through the DUT. The resonant circuit module consists of a load inductor (Lsurge) and a capacitor (Csurge). The corresponding surge current and voltage waveforms were recorded by the oscilloscope. The peak current (Ipeak) can be tuned by adjusting the VCC, Csurge and Lsurge, in terms of the relationship of Ipeak = VCC (Csurge / Lsurge) 1/2 [9].  with inductive loads is widely used to measure the reverse recovery characteristics of diodes [10]. During the test, a SiC MOSFET [7] is first turned on to build up the linear inductor current (i.e., the desired forward current, IF) in phase I, and IF can be changed by adjusting the Lload, Vbus and t1, according to the relationship of IF = (Vbus × t1) / Lload, where t1 is the first pulse width of the double pulse signal, Lload is the load inductor. When the SiC MOSFET is turned off (phase II), the DUT is under forward bias [11]. As the SiC MOSFET is turned on again (phase III), the diode switches to a reverse blocking condition with a short period of time (i.e., reverse recovery time, trr) usually present to remove the stored charge (i.e., reverse recovery charge).

Reverse recovery characterizations
The trr is defined as the time when the reverse current recovers to 10% of its peak value (Irr) ( Fig. S4 (b)) [12]. Note that the second pulse width (phase III) is shorter than the first pulse width (phase I) to avoid overheating of the device.

Supplementary Section S4 -Physics-based Simulation of NiO/Ga2O3 Heterojunction
The physics-based Technology Computer Aided Design (TCAD) simulation is performed using Silvaco [13]. Table S1 lists the basic material parameters used in the simulation for the NiO/Ga2O3 heterojunction diode.

Impact Ionization Coefficient Calculation
The Chynoweth model is used for the impact ionization (I. I.) coefficient of the electron (αn) and hole (αp), which formulates the I. I. coefficient (α) as Eqn. (S1), where E is electric field, An, Bn, Ap and Bp are constants related to the material and the subscript "n" and "p" represent electron and hole, respectively. Figure S5. Schematic of p-NiO/n-Ga2O3/n + -Ga2O3 structure. Figure S5 shows the schematic of the simplified 1-D model for the p-NiO/n-Ga2O3/n + -Ga2O3 HJD. The relation between α and multiplication coefficient (M) is given by Eqn. (S2).

Supplementary
where W is the width of depletion region. Considering that the acceptor concentration (NA > 10 18 cm -3 ) in p-NiO is much higher than the donor concentration (ND ~ 1.7 × 10 16 cm -3 ) in the n --Ga2O3 drift layer, the avalanche breakdown occurs in the punch-through condition, where n --Ga2O3 is fully depleted. The punchthrough voltage (Vpunch) is calculated as where εGaO is the permittivity of Ga2O3. Note that the experimental avalanche voltage is ~1600 V, being higher than Vpunch. With V ≥ Vpunch, the E-field E(x,V) can be written as: Considering x = W and V ≥ Vpunch and substituting Eqn. (S1) and (S4) into Eqn. (S2), the multiplication coefficient can be calculated as: The multiplication coefficient M can be extracted from the experimental avalanche I-V characteristics ( Fig.  2(a)), as defined by the ratio between the avalanche current (Itotal) and background current (Ibackground) before avalanche [22]. As shown in Figure S6, a polynomial fit between 0 and 1400 V was extended to the avalanche voltage range to obtain Ibackground. The avalanche multiplication effect causes the background electrons/holes to multiply, making the Itotal. Using this method, M~V characteristics can be obtained from experimental I-V characteristics.
The values of An and Bn are theoretically predicted in Ref. [23], which are used as the initial numbers in the fitting. Ap and Bp can be obtained by the method of least squares to fit the M-V characteristics derived from the experimental I-V characteristics. As no Ap and Bp have been reported, we choose the initial values of Ap and Bp to be similar to An and Bn. Figure S7 shows the calculated and experimental results of M, revealing a good agreement. Note that the slight deviation of the model at high avalanche currents is due to nonidealities related to the material and device structure (e.g., series resistance). The extracted key parameters for the Chynoweth model are summarized in Table S2.  Table S2.   Figure S8 shows the simulated profile of (a) electron and (b) hole mobilities and (c) E-field as a function of vertical depth in the HJD under an IAVA of 30 A using the models and parameter in Table S2. In the drift region, carrier mobilities increase from the heterojunction towards the substrate due to the reduced E-field. In the Ga2O3 substrate, the E-field is nearly zero, leading to constant carrier mobilities.   Table S2. The I. I. coefficients and generation rates show strong dependence on E-field, both peaking at the heterojunction. This is consistent with the analytical calculation presented above. In addition, Fig. S9(a) suggests that the I. I. coefficient of hole is higher than that of electron in Ga2O3 drift region, which is consistent with the cases in SiC and GaN materials [26]. Note that the I. I. coefficient is influenced by multiple material properties including the material bandgap, electric field, optical phonon energy, mean free path, electron mobility and electric field, as suggested by the widely-used Thornber model [27]. Fig.  S9(b) shows that the I. I. generation rate in Ga2O3 is much higher than in NiO, confirming the initiation of I. I. in Ga2O3.

Surge simulations
The basic parameters and key minority carrier mobility used in the simulation are shown in Table S1 and S2, respectively. Other physical models and parameters related to minority carrier transport include the carrier recombination, minority carrier mobility, and minority carrier lifetime models. The Shockley-Read-Hall (SRH) recombination model is employed for the recombination at the heterojunction interface. All minority carrier lifetime parameters are extracted at room temperature and low electric field. The hole lifetime (τh,GaO) of 6.2 ns in n-Ga2O3 and the electron lifetime in p-NiO (τe,NiO) of 124.0 ns in p-NiO, respectively, were extracted from microscopic EBIC characterizations. These values are employed in the simulation. These results were measured in the DC mode of the B1505 power device analyzer. In the subthreshold region, the ideality factor of the device is close to 2, suggesting that the interfacial SRH recombination mechanism is dominant at low forward bias. The increased subthreshold current at higher temperature is a result of high temperature-induced lowering of the barrier height.  Figure S14. Ron,sp versus BV benchmark for Ga2O3 SBDs, HJDs and JBS diodes [4][5][6], [12], [29][30][31][32][33][34][35][36][37][38]. Note that ampere-level large-area power devices are required for industrial applications, so their performance usually represents the application prospects of a power device technology. For the emerging Ga2O3 power technology, there is still a large gap between the performance of large-area devices and small-area devices, possibly due to the non-uniformity in fabrication process and wafer material properties. In addition, the breakdown voltage measured under the switching circuit tests instead of static I-V tests represents the true overvoltage margin of devices in practical power electronics applications. Note that, in this benchmark, the BV of devices in this work is the only one that has been validated in the circuit tests. The robust breakdown capability achieved in this work is attributed to the effective edge termination that incorporates the small-angle beveled junction termination extension (JTE) and high-k field plate.