Low Al-composition p-GaN/Mg-doped Al0.25Ga0.75N/n+-GaN polarization-induced backward tunneling junction grown by metal-organic chemical vapor deposition on sapphire substrate.

Low Al-composition p-GaN/Mg-doped Al0.25Ga0.75N/n(+)-GaN polarization-induced backward tunneling junction (PIBTJ) was grown by metal-organic chemical vapor deposition on sapphire substrate. A self-consistent solution of Poisson-Schrödinger equations combined with polarization-induced theory was used to model PIBTJ structure, energy band diagrams and free carrier concentrations distribution. The PIBTJ displays reliable and reproducible backward tunneling with a current density of 3 A/cm(2) at the reverse bias of -1 V. The absence of negative differential resistance behavior of PIBTJ at forward bias can mainly be attributed to the hole compensation centers, including C, H and O impurities, accumulated at the p-GaN/Mg-doped AlGaN heterointerface.

Inter-band tunneling junction (n class="Chemical">TJ) is a type of inpan>terestinpan>g verticn class="Chemical">al quantum transport device, which presents a negative differential resistance (NDR) behavior under forward bias and an ohmic characteristic rather than a rectification characteristic under reverse bias1. NDR behavior in TJ enables a variety of attractive applications in low-power memory, logic circuit, high-speed switching and high-frequency oscillation up to terahertz range234. Backward diode, a modification of TJ, presents an excellent frequency response and can be used in microwaves detection, small signals rectification and frequency mixing567. In addition, TJ is a critical component in electrically connecting multiple active regions of devices such as multi-color light emitting diodes (LEDs), laser diodes (LDs), photo-detectors and tandem solar cells (SCs)89.

Owing to their high thermn class="Chemical">al stability, high breakdownpan> voltage, high radiationpan> n class="Disease">hardness and high electron saturation velocity, III–V nitrides have great potential in fabricating vertical quantum transport devices, including the inter-band TJ10. However, efficient inter-band TJ is still a challenging issue for wide-band-gap GaN and AlGaN materials, mostly due to the large width and height of tunneling barrier. Moreover, the Mg dopant solubility limits, low acceptor ionization efficiency and severe self-compensation effect of Mg-doped III–V nitrides at room temperature (RT) makes it hard for the realization of degenerate hole concentration11. Recently, the investigations of AlGaN/GaN high electron mobility transistors (HEMTs) have demonstrated that two-dimensional electron gas (2DEG) with degenerate concentration can be induced at the AlGaN/GaN heterointerface, due to the effect of polarization charges1213. Furthermore, during the theoretical study of metal-AlGaN/GaN heterostructure, Singh et al. predicted that the large band bending and degenerate 2DEG induced by the polarization charges at the undoped AlGaN/GaN heterointerface makes possible a new type of TJ with tailored tunneling probabilities and current-voltage (I–V) characteristics1415. Theoretically, for a GaN/AlGaN/GaN heterostructure, a two-dimensional hole gas (2DHG) also can be induced at the upper GaN/AlGaN heterointerface with the similar carrier concentration as that of 2DEG at the lower AlGaN/GaN heterointerface16. The simultaneous appearance of degenerate polarization-induced 2DEG and 2DHG in GaN/AlGaN/GaN heterostructure can enable interband tunneling process. Based on this concept of polarization-induced theory, III–V nitrides inter-band polarization-induced TJ (PITJ) in the form of forward NDR device and backward diode has been successfully verified experimentally and theoretically by some research groups using Al-polar AlN or N-polar In-rich InGaN as a polarization-induced layer171819202122. Simon et al. introduced a Al-polar GaN/AlN/GaN PITJ, demonstrating backward diode characteristic with reverse current density of 750 mA/cm2 at 0.8 V19. Alternatively, Krishnamoorthy et al. achieved PITJ with forward NDR effect and backward tunneling characteristics using high In-composition N-polar InGaN202122. These results are very encouraging for the design and fabrication of III–V nitrides TJ with eliminating the need of degenerate p-type doping, which is a general requirement for traditional TJ and also a significant problem for wide-band-gap III–V nitrides. However, due to the high barrier height, the wide band gap of AlN in GaN/AlN/GaN PITJ reduces tunneling probability greatly, resulting in low current density and high tunneling resistance1920. While narrow band gap InGaN in GaN/InGaN/GaN PITJ leads to additional tunneling band discontinuities at the GaN/InGaN heterojunction, which is expected to reduce the overall tunneling probability2122. In addition, the optical absorption loss of InGaN layer impedes the application of InGaN-based PITJ into optoelectronics device like tandem SCs and UV emitters2122. Furthermore, the above mentioned PITJs were all grown on GaN templates by molecular beam epitaxy (MBE) using Al-polar AlN or N-polar In-rich InGaN. Though most commercial III–V nitrides optoelectronic devices are grown on sapphire substrates by metal-organic chemical vapor deposition (MOCVD) nowadays, the growth of III–V nitrides PITJ by MOCVD using Al-polar AlN and N-polar In-rich InGaN is not an easy task due to the difficult control of surface morphology, crystalline quality and impurities incorporation of materials23. In view of the tradeoffs of Al-polar AlN-based and N-polar In-rich InGaN-based PITJ, it may be expected that Ga-polar AlGaN with low Al-composition, owing lower tunneling barrier height and better optical transparence, is an ideal candidate for the efficient inter-band tunneling in III–V nitrides material, since the growth and doping of Ga-polar AlGaN with low Al-composition (<30%) on sapphire substrates by MOCVD are easy and commercial-available with smooth surface morphology and excellent crystalline quality2425. Especially, in our recent report, using Ga-polar low Al-composition Mg-doped Al0.15Ga0.85N as a polarization-induced layer, an obvious NDR behavior with high peak-to-valley current ratio of about 9 and peak current of 22.4 mA (~current density of 11.4 A/cm2) at 3.67 V was realized in p-GaN/Mg-doped Al0.15Ga0.85N/n+-GaN heterostructure grown by MOCVD on sapphire substrate26. Though the backward tunneling current density was found to be low, it still reveals the possibility of growing Ga-polar AlGaN-based PITJ with low Al-composition on sapphire substrate by MOCVD.

In this study, un class="Chemical">sing n class="Chemical">Mg-doped Al0.25Ga0.75N interlayer combined with n+-GaN, a low Al-composition p-GaN/Mg-doped Al0.25Ga0.75N/n+-GaN PITJ was grown by MOCVD on planar sapphire substrate using conventional GaN epitaxial processing. The physical mechanism of p-GaN/Mg-doped Al0.25Ga0.75N/n+-GaN PITJ was presented through calculations based on self-consistent solution of Poisson-Schrödinger equations combined with polarization-induced theory. Repeatable and efficient inter-band tunneling was achieved with a current density of 3 A/cm2 at the reverse bias of −1 V. The influence of hole compensation centers accumulated at the p-GaN/Mg-doped Al0.25Ga0.75N interface on the forward current-voltage (I-V) characteristic of PITJ was also discussed.

Results

During their investigation of Al0.28Ga0.72n class="Chemical">N/GaN heterostructure, Singh et al. revealed that there was a compromise between the thickness of polarization-induced layer and the tunneling probability1415. The density of states (DOS) in polarization-induced charges at the heterointerface increases with the increase of the thickness of AlGaN. However, the tunneling probability through the barrier is low for a thick polarization-induced layer according to the Wentzel-Kramers-Brillouin (WKB) model27. Thus, considering the low Al-composition of AlGaN used in our study, the thicknesses of AlGaN was fixed at 5 nm which is thick enough to allow high DOS in polarization-induced charges and thin enough to allow high tunneling probability. Fig. 1(a) and Fig. 1(c) is the schematic diagram of the p-GaN/Mg-doped Al0.25Ga0.75N/n+-GaN PITJ, named sample A and the reference sample, a p-GaN/n+-GaN junction, named sample B. To investigate the physical mechanism of PITJ, simulations of sample A and sample B were performed using self-consistent solution of Poisson-Schrödinger equations combined with the polarization-induced theory. The material parameters adopted during the simulations can be found in the Ref. 28.

Figure 1

(a) Schematic diagram of sample A and (b) spatial distributions of fixed polarization charges, ionized dopants and free mobile carriers gas in sample A and (c) schematic diagram of sample B.

Acn class="Chemical">cordinpan>g to polarizationpan>-inpan>duced theory of III–V n class="Chemical">nitrides, a thin strained Mg-doped Al0.25Ga0.75N interlayer surrounded by a GaN p-n+ junction in sample A will induce a polarization dipole layer with high densities of fixed polarization charges ±σ (~1013/cm2) at the upper GaN/AlGaN and lower AlGaN/GaN heterointerface, respectively29. As shown in the Fig. 1(b), these high densities of fixed polarization charges combined with ionized dopants constitute the total space charges in the depletion region of sample A. To balance these ionized dopants and the fixed polarization charges, 2DHG and 2DEG are accumulated with degenerate concentration at the upper GaN/AlGaN and lower AlGaN/GaN heterointerface, respectively. The calculated equilibrium band diagram of sample A shown in the Fig. 2(a) verifies that there exists potential-wells for electrons at the upper GaN/AlGaN and holes at the lower AlGaN/GaN heterointerface, respectively. The inset of Fig. 2(a) shows that the potential-well depths for 2DEG and 2DHG are 263 and 127 meV, respectively. Furthermore, as shown in Fig. 2(b), the concentrations of 2DEG and 2DHG are 5.85 × 1019 and 6.6 × 1019 cm−3, respectively, which meet the requirements of degenerate carrier concentrations of III-V nitrides TJ30. In contrast, the concentrations of electrons and holes of sample B shown in the Fig. 3(b) are less than 1018 and 1019 cm−3, respectively, which is far lower than those of sample A. Moreover, the polarization electric field E created by the polarization dipole in sample A can assist the built-in electric field E (in the same direction of E) created by ionized dopants dipole, which introduces a total electric field as high as ~106 V/cm19. This strong total electric field lead to extremely large band bending and would enable the alignment of energy bands over a very narrow depletion region across the GaN p-n junction. The inset in Fig. 2(a) proves that the giant electric field shrinks the depletion region width to a thickness less than 8 nm closing to the thickness of AlGaN interlayer, which is lower than the depletion region width of sample B, presenting a thickness of 20 nm as shown in the inset of Fig. 3(a). Therefore, it can be found that the carrier concentrations and depletion region width of sample B are far from satisfactory for tunneling process, due to the lack of polarization-induced effect of AlGaN interlayer.

Figure 2

(a) Energy band diagrams and (b) carrier distributions of sample A at equilibrium state. The inset shows the detail view of band diagram at the tunneling region.

Figure 3

(a) Energy band diagrams and (b) carrier distributions of sample B at equilibrium state. The inset shows detail view of band diagram at the space charge region.

Because of the strong quantum-n class="Chemical">confinpan>ement effect of two potentin class="Chemical">al wells in sample A, 2DEG and 2DHG are distributed in the confined potential-well states. At the thermal equilibrium state, there is an overlap of energy levels between the conduction band on the n-side and the valence band on the p-side. If the electrons energy levels of 2DEG in the potential well coincide with one of the holes energy levels of 2DHG, carriers can tunnel between conduction band and valence band, leading to the formation of tunneling current31. When a forward-bias voltage is applied to the sample A, the concentrations of degenerate carrier gases increase and electrons inject from the n-side to the p-side, causing the increase of tunneling current. With further increasing forward bias to 0.3 V (as shown in the Fig. 4(a)), the Fermi level E under thermal equilibrium conditions splits into two different quasi-Fermi levels and for electrons and holes, respectively. Conduction and valence bands are no longer aligned and NDR effect is expected to happen. The concentrations of 2DEG and 2DHG at 0.3 V shown in the Fig. 4(b) are 5.9 × 1019 and 6.7 × 1019 cm−3, respectively. Moreover, the inset in Fig. 4(a) shows that the potential-well depths of 2DEG and 2DHG at 0.3 V are 272 and 133 meV, respectively. The deeper potential-wells at forward bias are attributed to the increasing accumulation of 2DEG and 2DHG. At a higher forward bias, tunneling current ceases and diffusion current becomes the dominant. On the other hand, when a reverse-bias voltage of −0.3 V is applied to the sample A, the conduction band edge on the n-side is pulled below the valence band edge on the p-side, as shown in the Fig. 5(a). Valence band electrons from occupied states on the p-side can tunnel into the unoccupied states in the conduction band of n-side, leading to the great increase of reverse tunneling current. Though high densities of valence band electrons keep tunneling from p-side to n-side, electrons and holes are quickly swept out of the potential wells due to the driven effect of reverse-bias electric field. Therefore, as shown in the Fig. 5(b), the concentrations of 2DEG and 2DHG at −0.3 V are 5.7 × 1019 and 6.4 × 1019 cm−3, respectively, which are a little lower than those at thermal equilibrium state. In addition, the inset in Fig. 5(a) shows that the potential-well depths of 2DEG and 2DHG at −0.3 V are 260 and 125 meV, respectively.

Figure 4

(a) Energy band diagrams and (b) carrier distributions of sample A at +0.3 V. The inset shows the detail view of band diagram at the tunneling region.

Figure 5

(a) Energy band diagrams and (b) carrier distributions of sample A at −0.3 V. The inset shows the detail view of band diagram at the tunneling region.

Fig. 6 presents the ω-2θ XRD scans of the n class="Chemical">Mg-doped n class="Chemical">Al0.25Ga0.75N/GaN SLs. The distinct satellite peaks up to second order indicate that the SLs are in good crystalline quality and abrupt interface. According to the formula from Ref. 32, the Al-composition and thickness of AlGaN interlayer are calculated to be 25% and 5 nm, respectively, indicating a good agreement with the simulation conditions during the self-consistent solution of Schrödinger and Poisson equations. The root-mean-square (RMS) roughness of sample A, as shown in the inset of Fig. 6 is 0.42 nm over a scan area of 2 × 2 μm2 with an obvious atomically flat terraces-and-steps structure, demonstrating a smooth surface morphology.

Figure 6

The ω-2θ XRD scans of p-AlGaN/GaN SLs.

The left inset shows the surface morphology of sample A.

Discussion

Fig. 7 shows RT I–V curves of sample A and sample B taken by n class="Chemical">consecutively forward sweepinpan>g from −4 up to 4 V while measurinpan>g the current through the device. The positive bias is defined when applying a positive voltage on the p-GaN layer. For the I–V curve shown in the inset of Fig. 7, the sample B presents a usual diode rectifying behavior with a very low reverse current. On the other hand, the I–V curves of sample A show much higher reverse current than that in the forward direction with a strong nonlinearity at zero bias, which is a typical behavior of a backward diode. Backward diode is a type of TJ in which the forward current (including the NDR region) is suppressed greatly compared with reverse current33. Therefore, the I–V characteristics comparison between two samples demonstrate that tunneling process is actually generated by inserting Al0.25Ga0.75N strain layer into p-GaN/n+-GaN junction. We have to point out that, just like the phenomenon in metal-semiconductor Schottky contact, as a heterostructure with multiple dielectric layers for sample A, there is a possibility that the image effect of 2DEG may also impose some influence on the tunneling process343536. However, according to the study of A Minj et al., image force and image charges only play a major role when the size or electrode contact area of device is at the nanoscale, such as in the size of a few square nanometers35. Considering the device size of sample A (~200 × 150 μm2), the effect of image charges can be neglected in this study. Moreover, as pointed out by A. A. Grinberg, the image force induced by 2DEG is quite different from that in metal-semiconductor contact based on the following two points36. Firstly, the image force lowering of the heterointerface barrier by 2DEG is smaller than it would be in a metal-semiconductor contact. Secondly, the influence of image force induced by 2DEG on the interface barrier remains finite even though their distance from the surface or interface goes to zero. Therefore, tunneling process can't be realized only through image force or image potential, since the image charges alone are not enough to decrease the tunneling height and width in our work. From the above discussions, it can be concluded that the polarization-induced effect of AlGaN interlayer plays a dominant role in the tunneling process and the generation of 2DHG and 2DEG, which also excludes the possibility of Schotty tunneling between the electrode and p-GaN.

Figure 7

The I–V curves of sample A under consecutive scans.

The right inset shows the I–V curves of sample B.

As shown in the Fig. 7, the reliability and reproducibility of PIn class="Chemical">TJ were studied via twenty scans of I–V characteristics with no n class="Chemical">significant degradation of the device performance, in contrast to our previous report of NDR device26. The enhanced reproducibility with consecutive scans can be attributed to the improved crystalline quality of PITJ, especially the AlGaN interlayer, which decreases stress relaxation and thereby dislocations densities. Though slight degradation of reverse tunneling current are still happened after twenty scans, a more stable I–V characteristic can be obtained through the further optimization of layers structure and material growth conditions. However, the I–V characteristics of sample A just present backward diode behavior without NDR phenomenon at forward bias, which seems to contradict with the theory of PITJ and our above simulation results. Generally, tunneling current of TJ with generate carrier concentrations in forward and backward directions increases, as the electric field increases and depletion width decreases. The backward tunneling current only involves normally filled valence band electrons from the p-side tunneling into normally empty conduction states on the n-side. However, the forward tunneling current depends on the carrier concentrations for providing available conduction electrons in the n-side and available empty states in the p-side7. Hence, it can be concluded that the p-side of sample A is not quite degenerate as the n-side, resulting in the decrease of available empty states in the p-side, which may be associated with the increase of hole compensation centers. Generally speaking, the decrease of carrier concentrations is mainly due to the compensation effect of defects, including dislocations and impurities. As reported by Simon and Krishnamoorthy, similar I–V characteristics without negative differential resistance (NDR) phenomenon at forward bias have also been observed in the PITJ with Al-polar GaN/AlN/GaN and N-polar GaN/InGaN/GaN structures grown on GaN templates by MBE1920. In view of the lower dislocation densities and better interface quality of their samples than those of our sample A grown on planar sapphire substrate by MOCVD, the still absence of NDR phenomenon in their samples indicates that dislocations and interface states don't play a dominant role in the compensation of hole concentration. Moreover, since the thickness of AlGaN interlayer in sample A was less than the critical thickness, the influence of strain relaxation and the related dislocations and interface states can be neglected in this study. Therefore, the reason for the similarity of I–V characteristic between our work and Simon and Krishnamoorthy's samples may lie in the distribution of impurities, since AlN and N-polar III–V nitrides are easily susceptible to the incorporation of impurities.

To verify our reasoning, SIMS was taken to ann class="Chemical">alyze the distribution of impurities in sample A. In our study, the minimum size of SIMS sputtering-spot was about 100 × 100 μm2, which was at the same level of device size as that of sample A. The impurities distributions at a sputtering point can actually provide the average information of spatial distributions of impurities in a sputtering plane. Therefore, the SIMS data can be used to analyze the influence of the planar distributions of impurities on the sheet charges. As shown in Fig. 8, in the SIMS study of p-GaN/Mg-doped Al0.25Ga0.75N hetrointerface, an accumulation of C, H and O impurities happens near the p-GaN/Mg-doped Al0.25Ga0.75N heterointerface. The C, H and O concentrations at the heterointerface are 6 × 1018, 1.5 × 1019 and 5 × 1017 cm−3, respectively. The high levels of H can be explained by the formation of Mg-H complexes, since H incorporation is in tandem with Mg doping37. The incorporation of O originates largely from NH3, easily forming strong chemical bond with Al38. The possible source of C impurities are from the decomposition of metal-organic precursors, CO and CO2 contaminants in the NH3, and hydrocarbons resulting from H2 etching of SiC coated graphite susceptors39. Moreover, the accumulation of C, H and O impurities at the heterointerface can be attributed to the growth interruption after the growth of Mg-doped Al0.25Ga0.75N to adjust growth condition for p-GaN4041. During the growth interruption, the residual Mg and Al in the reactor are accumulated on the surface of Mg-doped Al0.25Ga0.75N, which will promote the adsorption of impurities404142. In addition, as can be seen in the SIMS data, the C, H, O impurities show the similar variation trend as the change of depth, which illustrates the spatial diffusion or distribution traces of impurities. Impurities may diffuse along the dislocations and vacancies and gather at the heterointerface, which also increase the impurities incorporation in the heterointerface43. Though SIMS data alone can't tell us whether the impurities are interstitial or substitutional, the most favorable sites for impurities in material depend on their formation energies as a function of growth stoichiometry and Fermi level, which can be calculated by first-principles theory or, to be specific, density-functional theory (DFT)3944. Firstly, C (C substituted at a Ga site) is found to be a shallow donor while C (C at an interstitial site) exhibits deep donor behavior in p-type GaN39. Using DFT, A. F. Wright presented that the formation energies of C and C are lower that of substitutional C (C substituted at a N site) in Mg-doped p-type GaN, making it likely that both C and C are preferred site in p-GaN39. However, C is favored under Ga-rich growth conditions while C is preferred under N-rich growth conditions3944. In light of the N-rich conditions during the growth of p-GaN and AlGaN by MOCVD, C would be more preferred3944. Secondly, DFT calculations indicate that Mg-H complex and isolated H+ should be the dominant H species in p-type Mg-doped GaN45. Mg-H complex was known to play a crucial role in passivating the Mg acceptors. The calculations by Jörg Neugebauer et al. present that H acts as a donor (H+) in p-type GaN, and as an acceptor (H−) in n-type material46. Thirdly, in a DFT study of O impurities in GaN, A. F. Wright indicated that the formation energies of O (O substituted at a Ga site) and O (O at an interstitial site) were higher than that of O (O substituted at a N site) for all values of the Fermi level, suggesting that O, being a single donor, is the dominant and most stable O state in p-type GaN47. Moreover, O can either compensate a Mg acceptor or combine with it to form (Mg complex. Therefore, the above discussions indicate that C, H and O impurities in the p-GaN and AlGaN behave dominantly as donor in the form of C, Mg-H complex or H+, and O, respectively.

Figure 8

SIMS depth profiles in sample A for C, H and O atomic concentrations at p-GaN/Mg-doped Al0.25Ga0.75N hetrostructure interface.

The enhanced inn class="Chemical">corporationpan> of C, H and O impurities, behavinpan>g as n class="Species">donors in p-type material, will increase the background concentrations of electrons and hole compensation in the p-side. As a result, a high degree of hole compensation is expected to be happened in the region around 2DHG, which reduces the degenerate degree of 2DHG. Therefore, when a reverse bias is applied, electrons could easily tunnel from the valence band of the p-side into the conduction band of the n-side, which leads to the increase of a tunneling current in the backward direction. However, under forward bias, as the band overlapping is not significant considering the different degree of degeneration of 2DEG and 2DHG, the tunneling vanishes in the forward direction and NDR effect will not appear. Forward diffusion current across the p-n junction becomes the dominant process with depressed tunneling current. Due to the relatively high barrier height of Al0.25Ga0.75N for carrier transition, the forward current of sample A is lower than that of sample B at the same bias value, which also leads to the increase of turn-on voltage of sample A. Though the impurities compensation can be considered as the dominant compensation mechanism, the influence of other impurities, such as N vacancies, are not discussed in the study. Therefore, more detailed investigations are needed to further explore the hole compensation centers and theirs effect on the performance of PITJ, which are beneficial for the freely modulating the generate degree of 2DHG at the p-side to conveniently fabricate different device characteristics of GaN-based PITJ for forward NDR behavior or backward tunneling property.

The PIn class="Chemical">TJ studied here exhibits current density in the 3–30 A/cm2 range at the reverse bias between −1 V and −4 V, which is low for III-nitrides optical emitters such as LEDs and LDs. To further increase the tunneling current, decreasing heterointerface hole compensation centers, adjusting the thickness of polarization dipole layer and optimizing material epitaxial processing are needed. Moreover, the performance improvement by hybrid using low Al-composition AlGaN and low In-composition InGaN layers (to grow structures like n-GaN/p-AlGaN/p-InGaN and n-InGaN/p-AlGaN/p-InGaN) can also greatly increase the polarization charges profiles and thus the tunneling current.

Conclusions

In summary, the den class="Chemical">sign of PIn class="Chemical">TJ using low Al-composition AlGaN interlayer combined with n+-GaN was confirmed by theoretical simulations and experimental results. The PITJ displays reliable and reproducible backward tunneling with a current density of 3 A/cm2 at the reverse bias of −1 V. The absence of NDR behavior of PITJ at the forward bias can be attributed to the hole compensation centers, such as C, H and O impurities, accumulated at the p-GaN/Mg-doped AlGaN heterointerface. The PITJ demonstrated in this work can balance optical transparence and tunneling probability simultaneously, which is also compatible with the currently commercial III–V nitrides epitaxial processing. Low Al-composition AlGaN-based PITJ opens up new opportunities for the performance improvement of III–V nitrides optoelectronics devices in multi-color LEDs, LDs, and tandem SCs, and also holds much promise for future applications in GaN-based microwave detectors, inter-band vertical transport quantum devices, and terahertz oscillators.

Methods

Samples fabrication

For the experimentn class="Chemical">al studies, sample A and sample B were grownpan> onpan> planar c-plane sapphire substrates un class="Chemical">sing conventional GaN epitaxial method by MOCVD with an Aixtron close coupled showerhead system. Trimethylgallium (TMGa), trimethylaluminum (TMAl), and ammonia (NH3) were used as Ga, Al, and N precursors, respectively. Biscyclopentadienyl magnesium (Cp2Mg) and silane (SiH4) were used as p-type and n-type GaN dopants, respectively. The structure of sample A consists of a 2 μm undoped GaN, 2 μm GaN:Si (Si ~ 4 × 1018 cm−3), 15 nm GaN:Si (Si ~ 4 × 1019 cm−3), 5 nm Al0.25Ga0.75N:Mg (Mg ~ 5.5 × 1019 cm−3), 150 nm GaN:Mg (Mg ~ 6 × 1019 cm−3). The sample B has the similar structure as sample A but without Al0.25Ga0.75N interlayer. After growth, the samples were annealed in the O2 ambient at 750°C for 15 min. Diodes with a size of 200 × 150 um2 were fabricated on the whole 2-inch wafers. Ni/Au (20/150 nm) and Ti/Al/Ni/Au (20/100/40/50 nm) stacks were evaporated using standard optical lithography on the p-GaN and n-GaN layers, respectively.

Measurements

n class="Chemical">Al-composition and epilayer thickness determination of sample A were investigated and calibrated using high resolution X-ray diffraction (HRXRD from Bede D1) by growing 20 period Mg-doped Al0.25Ga0.75N/GaN (5 nm/7 nm) SLs at the same growth condition, thickness and Al-composition as sample A. The surface morphology of sample A was viewed by atomic force microscope (AFM from Veeco D3100) with a tapping mode. I–V electrical properties of two samples were characterized at RT with I–V analyzer (Keithley Model 2611A). A secondary ion mass spectroscopy (SIMS from CAMECA IMS-7f) was used to probe the concentrations of element in the samples.

Author Contributions

K.X.Z. carried out the experiments den class="Chemical">sign, the samples growth, theoreticn class="Chemical">al simulations and data analysis. W.P.G. carried out the samples' XRD and AFM measurements. K.X.Z. wrote the manuscript with suggestions from H.W.L. K.X.Z. and H.W.L. revised the manuscript. Y.L., R.S.S., W.P.G., D.S.W., X.C.X., P.C.T., C.Y., Y.M.L. and G.T.D. gave scientific advices. All the authors contributed through scientific discussion and reviewed the manuscript.