High voltage gallium nitride vertical PN diode

Abstract

A vertical gallium nitride (GaN) PN diode uses epitaxial growth of a thick drift region with a very low carrier concentration and a carefully designed multi-zone junction termination extension to achieve high voltage blocking and high-power efficiency. An exemplary large area (1 mm.sup.2) diode had a forward pulsed current of 3.5 A, an 8.3 m-cm.sup.2 specific on-resistance, and a 5.3 kV reverse breakdown. A smaller area diode (0.063 mm.sup.2) was capable of 6.4 kV breakdown with a specific on-resistance of 10.2 m-cm.sup.2, when accounting for current spreading through the drift region at a 45 angle.

Claims

1. A high voltage gallium nitride vertical PN diode, comprising: a high-doped n-type gallium nitride substrate or contact layer; a low-doped n-type gallium nitride drift layer epitaxially grown on the gallium nitride substrate, wherein the drift layer is greater than 10 microns in thickness and has a donor concentration of less than 210.sup.15/cm.sup.3; a p-type region comprising one or more p-type layers epitaxially grown on the drift layer, thereby forming a PN junction with the drift layer; an ohmic cathode contact deposited on the high-doped n-type gallium nitride substrate or contact layer; an ohmic anode contact deposited on the p-type region; and a step-etched multi-zone junction termination extension structure laterally surrounding the anode contact in the p-type region.

2. The high voltage gallium nitride vertical PN diode of claim 1, wherein the low-doped n-type gallium nitride drift layer is grown by metal-organic chemical vapor deposition under compensation doping conditions.

3. The high voltage gallium nitride vertical PN diode of claim 1, wherein the drift layer is equal to or greater than 50 microns in thickness.

4. The high voltage gallium nitride vertical PN diode of claim 1, wherein the area of the anode contact is greater than 0.01 mm.sup.2.

5. The high voltage gallium nitride vertical PN diode of claim 4, wherein the area of the anode contact is greater than 0.1 mm.sup.2.

6. The high voltage gallium nitride vertical PN diode of claim 1, wherein the p-type region comprises a moderately doped p-type gallium nitride layer epitaxially grown on the drift layer and a high-doped p-type gallium nitride layer epitaxially grown on the moderately doped p-type gallium nitride layer.

7. The high voltage gallium nitride vertical PN diode of claim 6, wherein the moderately doped p-type gallium nitride layer has an acceptor concentration of less than 10.sup.19 cm.sup.3.

8. The high voltage gallium nitride vertical PN diode of claim 6, wherein the high-doped p-type gallium nitride layer has an acceptor concentration of greater than 10.sup.19 cm.sup.3.

9. The high voltage gallium nitride vertical PN diode of claim 1, wherein the step-etched multi-zone junction termination extension structure comprises two or more zones.

10. The high voltage gallium nitride vertical PN diode of claim 1, wherein the PN diode has an avalanche breakdown voltage of greater than 5 kV in reverse bias.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The detailed description will refer to the following drawings, wherein like elements are referred to by like numbers.

(2) FIG. 1 is an axisymmetric schematic illustration of an exemplary vertical PN diode with a 4-zone step-etched JTE (not to scale). The JTEs were designed to ensure full depletion of the smallest zone and a uniform electric field at breakdown in the remaining zones. Anode diameters of 284 mm and 1.126 mm were fabricated.

(3) FIG. 2A is an image of a completed vertical GaN PN diodes with a 2-zone JTE design. FIG. 2B is an image of a completed vertical GaN PN diodes with a 4-zone JTE design.

(4) FIG. 3 is a graph of reverse I-V characteristics and breakdown voltage of two PN diodes with an anode area of 0.063 mm.sup.2 and 1 mm.sup.2. The noise floor change is a result of different anode areas and not due to the characterization equipment.

(5) FIG. 4A is a graph of forward I-V characteristics of the 6.4 kV diode with an anode area of 0.063 mm.sup.2. FIG. 4B is a graph of forward I-V characteristics of the 6.0 kV diode with an anode area of 1 mm.sup.2. Insets show forward current on a semi-logarithmic scale. Right y-axis shows the device specific on-resistance calculated using both the anode area and assuming a 45 current spreading through the drift region.

(6) FIG. 5A is a graph of forward I-V characteristics of a 1 mm.sup.2 device under pulsed conditions. Inset shows forward current on a semi-logarithmic scale.

(7) FIG. 5B is a graph of reverse I-V characteristics of the same 1 mm.sup.2 device.

(8) FIG. 6 is a graph of reverse I-V characteristics for 4-zone and 2-zone JTE designs where both had a total JTE width of 300 mm. A similar breakdown of 5.9 kV is achieved with both device types. Inset shows the forward I-V characteristics on a semi-logarithmic scale.

(9) FIG. 7A is an optical microscope image of a processed quarter wafer indicating the location of device CV measurements. FIG. 7B is a graph showing extracted carrier concentration of measured devices in the n-GaN drift layer. A frequency of 1 MHz was used for the CV measurements.

(10) FIG. 8 is a graph showing the temperature dependence of the reverse I-V characteristics of a PN junction diode with an anode area of 0.063 mm.sup.2.

DETAILED DESCRIPTION OF THE INVENTION

(11) A schematic cross-sectional illustration an exemplary high voltage GaN PN diode 10 is shown in FIG. 1. The diode 10 comprises a greater than 10-m thick low-doped n-type GaN voltage-blocking drift layer 13 epitaxially grown on a highly doped (e.g., a few orders of magnitude higher than the drift layer doping) n-GaN substrate 12 or contact layer. The n-type GaN drift layer 13 can be doped with a donor (e.g., Si) and have an electron concentration of less than about 210.sup.15 cm.sup.3. Next, a p-type region 14 comprising a moderately doped p-GaN layer 15 capped with a heavily doped p.sup.-GaN layer 16 is grown to assist in forming a good p-type contact. The p-GaN/n.sup.-GaN layers form the main junction of the vertical PN diode. Next, a JTE structure 17 comprising a series of stepped zones 18 is formed in the p-region layers 14 outside of the main PN junction. Finally, ohmic metal contacts are deposited on the top of the high-doped p.sup.+-GaN layer, to form the anode contact 19, and the exposed bottom side of the n-GaN substrate, to form the cathode contact 11.

(12) During forward-bias operation, charge carriers are injected from the highly doped p- and n-layers (i.e., anode and cathode). Conversely, under increased reverse bias, the depletion region and electric field increase inside the diode. Because of the low doping in the drift layer, most of the potential will drop across the drift layer in the region of the main PN junction. The variation in thickness in the JTE structure reduces field crowding at the PN junction periphery, thereby reducing the peak electric field to avoid premature avalanche breakdown under reverse bias. Avalanche breakdown will occur at a critical electric field due to runaway impact ionization from highly energetic electrons and holes.

(13) As an example of the invention, PN diodes with large anode areas were fabricated. The exemplary diodes experienced non-catastrophic breakdown and demonstrated a temperature-dependent breakdown voltage consistent with an avalanche breakdown process. This was achieved with a thick, low-doped n.sup. drift region and step-etched multi-zone JTE (MZ-JTE) design which consisted of either 2 or 4 zones. Both JTE designs resulted in device breakdowns up to or greater than 6 kV. Additionally, several fabricated devices were found to have exceptional leakage current suppression with a reverse leakage current density less than 110.sup.6 A/cm.sup.2 up to 90% of breakdown.

(14) Device Fabrication

(15) Epitaxial growth was performed via metal organic chemical vapor deposition (MOCVD) in a Taiyo Nippon Sanso Corp (TNSC) SR4000HT reactor. The substrate was a commercially available HVPE-grown laser diode grade 2 GaN substrate with nominal thickness of 400 m, threading dislocation density of 10.sup.6 cm.sup.2, and electron concentration of 210.sup.18 cm.sup.3. As shown in FIG. 1, the complete epi structure of the exemplary device consisted of a 15 nm p.sup.++ GaN contact layer (Mg210.sup.20 cm.sup.3), 86 nm p.sup.+ GaN (Mg310.sup.19 cm.sup.3), 295 nm p.sup. GaN (Mg110.sup.18 cm.sup.3), and a 50 m thick lightly-doped n.sup. GaN:Si drift layer. Trimethylgallium and ammonia were used as precursors in a mixed H.sub.2/N.sub.2 carrier gas, while silane (SiH.sub.4) and bis(cyclopentadienyl)magnesium (MgCp.sub.2) were used as n- and p-type dopants, respectively. The growth was performed at atmospheric pressure (630 torr) at a substrate temperature of 1050 C. for the drift layer, 1015 C. for the p-type anode layers, and 920 C. for the p.sup.++ contact layer. Activation of the Mg acceptors was achieved by thermal annealing at 900 C. for 15 minutes in nitrogen. The sheet resistance of the p-type anode was estimated as 68 k/sq from on-wafer CTLM measurements. A Nomarski microscope map of the 50 m thick epilayer after growth indicated sporadic deposition of large particles, but the density of particles had a minimal impact on device yield. Further characterization by atomic force microscopy confirmed an atomically smooth, step-bunch-free surface morphology with roughness <0.3 nm RMS.

(16) As described above, one of the key challenges in realizing breakdown voltages >5 kV in vertical GaN devices is the need for doping control at carrier concentrations 10.sup.15 cm.sup.3. This is quite difficult due to compensating carbon impurities that are present in the reactor from the decomposition of metalorganic precursors. Thus, to achieve very low doping levels one must have an accurate measurement of the carbon concentration in the drift layer (610.sup.15 cm.sup.3 in this exemplary device as measured by secondary-ion mass spectroscopy for a separate growth), as well as precise control of Si doping. The dopant flow used here produced a net electron concentration of 0.5-110.sup.15 cm.sup.3 with some variation across the wafer.

(17) After the epitaxial growth, the 2 wafer was divided into 4 quarters and each quarter was processed individually. One quarter was used to process devices with a 2-zone JTE and another was processed using a 4-zone JTE design. The charge required to terminate electric field lines into the JTE and ensure a uniform field profile that is distributed across multiple steps away from the anode contact was calculated based on Gauss's Law. See J. R. Dickerson et al., Simulation and Design of Step-Etched Junction Termination Extensions for GaN Power Diodes, in 2020 4th IEEE Electron Devices Technology & Manufacturing Conference (EDTM), 2020: IEEE, pp. 1-4, which is incorporated herein by reference. The thicknesses for each JTE design are shown in Table I. The anticipated critical electric field values at N.sub.dN.sub.a=110.sup.15 cm.sup.3 carrier concentration are 2.34 MV/cm and 1.84 MV/cm, based on ionization coefficients from both Ji and Maeda, respectively. See D. Ji et al., Appl. Phys. Lett. 115(7), 073503 (2019); and T. Maeda et al., J. Appl. Phys. 129(18), 185702 (2021). While other groups have reported ionization coefficients for GaN, Ji and Maeda encompass the best known upper and lower bounds in the current literature. See L. Cao et al., IEEE Trans. Electron Devices 68(3), 1228 (2021); K. Kunihiro et al., IEEE Electron Device Lett. 20(12), 608 (1999); and A. M. Ozbek, Measurement of Impact Ionization Coefficients in Gallium Nitride, North Carolina State University, 2011. The maximum electric field at breakdown is expected to span over multiple zones for the 4-zone design. For the 2-zone design, the 1.sup.st zone will come close to fully depleting for either critical field extreme and the 2.sup.nd zone would have a uniform electric field during breakdown. The goal of both JTE designs was to increase the process window for an estimated critical electrical field, to account for variations from targeted values of doping and step height, and to account for potential etch damage impacts on the top surfaces. See X. Cao et al., Appl. Phys. Lett. 75(17), 2569 (1999). An axisymmetric schematic illustration of the exemplary 4-zone JTE design and epilayer stack is shown in FIG. 1. The 2-zone JTE design maintains the total JTE width, with the thickness and width of each zone increased relative to those of the 4-zone JTE design. While not depicted in FIG. 1, t.sub.1, t.sub.2, x.sub.1, and x.sub.2 from Table 1 are the relative JTE thicknesses and step widths for the 2-zone JTE design. In addition, the mask layout consisted of four step width variations for each design. These are denoted in Table 1 as Var 1, 2, 3, and 4, resulting in total JTE widths of 150, 200, 250, and 300 m for both designs.

(18) TABLE-US-00001 TABLE 1 Thicknesses for junction termination extension designs 4-zone (nm) 2-zone (nm) t.sub.1 t.sub.2 t.sub.3 t.sub.4 t.sub.1 t.sub.2 325 251 172 90 260 137 step width (m) step width (m) x.sub.1 = x.sub.2 = x.sub.3 = x.sub.4 x.sub.1 = x.sub.2 Var 1 37.5 Var 1 75 Var 2 50 Var 2 100 Var 3 62.5 Var 3 125 Var 4 75 Var 4 150

(19) The JTE design was implemented using standard GaN processing methods. The p-GaN ohmic contact for the anode electrode was formed via e-beam evaporation of 20 nm of Pd and 320 nm of Au, followed by a rapid thermal anneal at 600 C. for 1 min in N.sub.2 ambient. The specific contact resistance of the p-GaN ohmic contact was estimated to be 5.710.sup.5 -cm.sup.2 from CTLM measurements. This is in line with previously reported results from Cho, in which they measured a specific contact resistance of 2.410.sup.5 -cm.sup.2 for a Pd/Ni/Au contact on a p-GaN layer with 210.sup.17 cm.sup.3 carrier concentration. See H. Cho et al., Solid State Electon. 49(5), 774 (2005). The isolation trench and step-etching were carried out with a BCl.sub.3/Cl.sub.2-based inductively coupled plasma reactive ion etching (ICP-RIE) with an ICP and RF power of 125 W and 10 W, respectively. Finally, a non-alloyed Ti/AI/Ti/Ni/Au (10 nm/200 nm/30 nm/50 nm/300 nm) was evaporated on the back side of the GaN substrate to form the cathode contact. Completed devices with both the 2-zone and 4-zone JTE are shown in FIGS. 2A and 2B. It should be noted that the exemplary PN diodes did not contain device passivation due to prior evidence of passivation damage induced leakage. See L. Chen et al., IEEE Trans. Device Mater. Reliab. 18(3), 359 (2018); and S. Kolli et al., J. Mater. Sci.: Mater. Electron. 29(22), 19353 (2018).

Experimental Results

(20) Devices were evaluated using a Keysight B1505A Power Device Analyzer/Curve Tracer with the N1268A ultra high voltage expander. This allows for reverse I-V characteristics up to 10 kV with a 100 pA noise floor. The fabricated PN diodes were immersed in Fluorinert (FC-70) during testing to avoid premature breakdown in air or at the device surface. Measurements were performed under dc bias unless otherwise indicated. Additional forward I-V and capacitance-voltage (C-V) measurements were performed to fully characterize the devices and epilayers. Finally, temperature dependent reverse breakdown measurements were done to demonstrate a robust avalanche capability.

(21) FIG. 3 shows the reverse voltage characteristics and breakdown for two of the best performing devices, a 0.063 mm.sup.2 device and a 1 mm.sup.2 device. Both devices consisted of a 4-zone JTE design with a total JTE width of 300 m. The 1 mm.sup.2 diode had a breakdown voltage of 6.0 kV with a leakage current of <110.sup.6 A/cm.sup.2 up to 5.4 kV (i.e., 90% of its demonstrated breakdown). The 0.063 mm.sup.2 diode was able to achieve a breakdown voltage of 6.4 kV, again with <110.sup.6 A/cm.sup.2 leakage current up to 5.6 kV, i.e., 88% of its breakdown. Note that the noise floor in FIG. 3 changes because of differing anode areas and is not due to any change in characterization equipment. In the implemented test configuration, the minimum current was limited to 100 pA. Breakdown voltages of greater than 6.0 kV were obtained from 4-zone JTEs with total widths ranging from 200 m to 300 m. The devices with the lowest leakage and largest breakdown came from the 4-zone JTE design with a 300 m total width. Ohta demonstrated a vertical GaN PN diode capable of 5.0 kV breakdown that consisted of a three-layer growth stack in which the carrier concentration was graded throughout the drift region and utilized a low doped (<210.sup.15 cm.sup.3) layer at the junction. See H. Ohta et al., Jpn. J. Appl. Phys. 57(4S), 04FG09 (2018). This was postulated to suppress the maximum electric field at the junction and aid in reducing the on-resistance of the device. The device structures reported by Ohta consisted of diodes with a 60 m anode diameter. The uniform low doped (<110.sup.15 cm.sup.3) 50 m thick drift region of the exemplary device of the present invention is capable of achieving a 6.0 kV breakdown for a device with a 1 mm.sup.2 anode area contact.

(22) The forward I-V characteristics of the two PN diodes from FIG. 3 are shown in FIGS. 4A and 4B. Both devices had a turn-on voltage of 3.1 V based on linear extrapolation. For the 6.4 kV diode, the specific on-resistance was calculated as 5.7 m-cm.sup.2 when assuming an anode area of 0.063 mm.sup.2; however, assuming 45 current spreading through the thickness of the drift layer, an adjusted specific on-resistance of 10.2 m-cm.sup.2 was calculated. Multiple models for current spreading through the drift region have been suggested, and by assuming a 45 spreading angle from the anode through the thickness of the drift region, the specific on-resistance upper limit can be bounded. See B. J. Baliga, Fundamentals of Power Semiconductor Devices, Springer Science & Business Media, 2010. This has substantial impact for the 0.063 mm.sup.2 devices with a 50 m drift layer, where the value almost doubles when current spreading is considered. The impact is less for the larger 1 mm.sup.2 device since the additional area through the 50 m drift region is much less compared to the anode contact itself. For this device, a specific on-resistance of 11.1 m-cm.sup.2 and 13.1 m-cm.sup.2 was calculated using the anode and spreading areas, respectively. The calculated on-resistance was taken at the low point of the curve since there appeared to be some device heating during the dc measurement that caused an increase in resistance at higher current densities. The specific on-resistances are higher than what has been previously reported for vertical GaN PN diodes with 3.7 kV and 5.0 kV breakdown. See I. C. Kizilyalli et al, IEEE Electron Device Lett. 35(2), 247 (2013); and H. Ohta et al., Jpn. J. Appl. Phys. 57(4S), 04FG09 (2018).

(23) Additional forward I-V measurements were carried using a pulsed measurement mode to mitigate heating effects and push the devices to higher currents. The pulsed conditions consisted of a 500 s pulse width and a 0.5% duty cycle. Under these conditions, several of the 1 mm.sup.2 devices were evaluated up to 3.5 A (350 A/cm.sup.2). It was found that there still appeared to be some device heating with the 500 s pulse width, although not as severe as with the dc measurements. FIG. 5A shows the forward I-V characteristics of a 1 mm.sup.2 device under the pulsed conditions. A reduction in the specific on-resistance is observed as compared to the dc measurement condition. The specific on-resistance was estimated as 8.3 m-cm.sup.2 and 9.7 m-cm.sup.2 for the anode and spreading areas, respectively. The reverse I-V characteristic for the same device is shown in FIG. 5B. A breakdown of 5.3 kV is measured with a reverse leakage current below 110.sup.6 A/cm.sup.2 up to breakdown.

(24) The 4-zone JTE is expected to be more robust with higher breakdown. However, near 6 kV performance was achieved with both the 2-zone and 4-zone JTE designs, as shown in FIG. 6. From multiple device measurements it appeared that discrepancies in the reverse leakage and breakdown characteristics between devices were more likely due to processing variations and/or positional dependency of the carrier concentration across the wafer.

(25) To investigate the positional dependency on carrier concentration, small PN diodes with only an isolation etch were included on the mask set. C-V measurements were performed with these devices, considering that the area of the depletion region is readily known with this device structure. The location of measured devices on the quarter is indicated in FIG. 7A with the measured N.sub.dN.sub.a carrier concentration shown in FIG. 7B. The C-V device measurements indicated that the N.sub.dN.sub.a carrier concentration was 110.sup.15 cm.sup.3 near the center of the wafer and dropped to around 5-610.sup.14 cm.sup.3 towards the wafer's edge. Because the breakdown voltage is inversely proportional to the net carrier concentration, it is critical to reduce this value to obtain such high breakdown performance. However, this does increase the drift region resistance. This trade-off has been well documented and studied elsewhere. See B. J. Baliga, J. Appl. Phys. 53(3), 1759 (1982). In the case of the exemplary PN diodes, breakdown is limited due to the thickness of the drift region, a condition known as punch-through, where the drift region is fully depleted and the maximum electric field becomes equal to the critical electric field. See B. J. Baliga, Fundamentals of Power Semiconductor Devices, Springer Science & Business Media, 2010. This type of design is often desirable as it maximizes the trade-off between high breakdown and low on-resistance when the optimum doping and thickness for a desired breakdown is employed. See J. A. Cooper and D. T. Morisette, IEEE Electron Device Lett. 41(6), 892 (2020). If, however, the doping level is significantly less than the calculated optimum level based on the critical electric field, then the drift region resistivity will increase proportional to the inverse of the doping level. For the exemplary devices, the doping level is likely too low for the drift thickness, thereby resulting in an extreme punch-through condition and subsequently an increased specific on-resistance as compared to an ideal case.

(26) As described above, knowledge of the impact ionization coefficients in GaN is essential to properly design and evaluate vertical GaN PN diodes. To evaluate the effectiveness of the exemplary JTE design, the ionization coefficients from Ji and Maeda were used to determine the expected breakdown voltage for the punch-through diodes at the measured carrier concentration. See D. Ji et al., Appl. Phys. Lett. 115(7), 073503 (2019); and T. Maeda et al., J. Appl. Phys. 129(18), 185702 (2021). For an N.sub.dN.sub.a of 610.sup.14 cm.sup.3, a 50 m drift region should be able to sustain a breakdown voltage of 9.7 kV or 7.3 kV according to Ji and Maeda, respectively. This would imply that the JTE is either 66% or 87% effective. Many of the devices experienced non-catastrophic breakdown; avalanche breakdown of a 5.6 kV device is shown in FIG. 8. A strong breakdown dependence on temperature was seen with an increased reverse leakage current prior to breakdown. This is in line with previous reports of avalanche breakdown in vertical GaN PN diodes; however, an increased temperature coefficient was observed compared to previous reports. According to a linear breakdown voltage dependence of BV(T)=BV.sub.25 C.(1+T) with a measured =1210.sup.4 K.sup.1, approximately twice the dependence that has been previously reported by most authors, and more than 3 of that reported by Ji. See I. C. Kizilyalli et al., IEEE Trans. Electron Devices 60(10), 3067 (2013); O. Aktas and I. Kizilyalli, IEEE Electron Device Lett. 36(9), 890 (2015); H. Fukushima et al., Jpn. J. Appl. Phys. 58(SC), SCCD25 (2019); and D. Ji et al., Robust avalanche in GaN leading to record performance in avalanche photodiode, in 2020 IEEE International Reliability Physics Symposium (IRPS), 2020: IEEE, pp. 1-4.

(27) The present invention has been described as a high voltage gallium nitride vertical PN diode. It will be understood that the above description is merely illustrative of the applications of the principles of the present invention, the scope of which is to be determined by the claims viewed in light of the specification. Other variants and modifications of the invention will be apparent to those of skill in the art.