Nitrogen-displacement-related recombination centers generated by electron beam irradiation in n-type and p-type homoepitaxial GaN layers

Recombination centers originating from point defects generated by the displacement of N atoms in n-type and p-type GaN were investigated by analyzing Shockley–Read–Hall (SRH) recombination currents in homoepitaxial GaN p–n junctions. These defects were intentionally generated by electron beam (EB) irradiation at 137 keV. The net doping concentrations in p+–n junction diodes were not changed following irradiation although the levels in p–n+ junction diodes decreased as the EB fluence was increased. The SRH recombination current also increased with increases in the fluence. This work additionally evaluated the relationship between recombination lifetimes and trap concentrations obtained by deep level transient spectroscopy.


G
allium nitride (GaN) exhibits many useful properties, including a high critical electric field 1) and good electron mobility, 2) and so is one of the most promising materials for the construction of next-generation power devices.In the case of GaN-based vertical power FETs, [3][4][5][6] the p-n junction is a vital component. Tis structure is essential as a means of controlling the breakdown voltage 7) and reducing electric field crowding at the device edge formed by ion implantation.8,9) During the fabrication of such devices via processes such as ion implantation and reactive ion etching, various point defects are introduced.Unfortunately, the characteristics of such defects, which can modify the effects of doping and reduce carrier mobility, remain poorly understood.
Various techniques have been employed to evaluate the properties of point defects in GaN, including deep level transient spectroscopy (DLTS), [10][11][12] positron annihilation spectroscopy, 13,14) photoluminescence, 15) cathodoluminescence 16,17) and carrier lifetime measurements.[18][19][20] Among these, carrier lifetime measurements are best suited to assessing point defects acting as recombination centers (RCs).In the early studies, carrier lifetimes in GaN were reported to be limited due to the high threading dislocation density (TDD) values in GaN crystals. Hence,investigations of the correlation between carrier lifetime and point defect concentrations were difficult. Recent proress in growth techniques has enabled the fabrication of high-quality epitaxial GaN on GaN free-standing substrates with low TDD values, so it has become possible to assess such relationships. 18) Onapproach to determining carrier lifetimes is to evaluate the Shockley-Read-Hall (SRH) recombination current in the depletion region of a p-n junction diode.21,22) The SRH lifetime, which is the square root of the product of hole and electron carrier lifetimes (t tt = SRH n p ), can be extracted from the SRH recombination current with an ideality factor of 2. Previous research determined SRH lifetimes based on the SRH recombination currents obtained from homoepitaxial GaN p + -n and p-n + junction diodes following irradiation with 4.2 MeV protons as a means of investigating the effect of intrinsic defects.23) This prior work found that the SRH lifetime was reduced with increasing doses of proton irradiation and that V Ga -related defects appeared to function as RCs.However, because proton irradiation at this energy level can displace both N and Ga atoms, it was not possible to conclusively identify the origins of the RCs formed by intrinsic defects.
In the present study, intrinsic defects associated with N atom displacement were intentionally introduced into homoepitaxial GaN p-n junction diodes through electron beam (EB) irradiation.This technique was utilized because EB irradiation at energies in the range from 125 to 450 keV can displace only N atoms. 24,25)In these trials, the concentration of intrinsic point defects could be controlled by changing the EB irradiation fluence.In our previous work, we studied the majority carrier traps originating from N-displacement-related defects introduced by EB irradiation in n-type and p-type GaN using DLTS.Two electron traps, EE1 and EE2, were observed in n-type GaN and one hole trap, EHa, was found in p-type GaN. 25,26)However, there have been no reports concerning the ability of these traps and N-displacement-related defects to act as RCs.In addition, the effects of the conduction type (that is, the Fermi level) on the properties of these defects is still unclear.In the work reported herein, RCs in GaN were investigated based on assessing SRH recombination currents obtained from forward current-voltage characteristics in homoepitaxial GaN p + -n and p-n + junction diodes exposed to EB irradiation at an energy level of 137 keV.
In preparation for this work, two GaN p ++ /p/n structures were grown by metalorganic vapor phase epitaxy on freestanding GaN (0001) substrates.These substrates were produced using hydride vapor phase epitaxy and each had a TDD of 3 × 10 6 cm −2 .The thickness and Mg concentration of the top p ++ contact layer were 0.1 μm and 2 × 10 20 cm −3 , respectively, while the thicknesses of the p-type and n-type layers in the p + -n junction diodes were 0.5 and 3.0 μm, respectively.In the case of the p-n + junctions, these values were 3.0 and 0.5 μm, respectively.Secondary ion mass spectrometry was used to ascertain the Si and Mg concentrations.The Mg concentration in the p-type GaN layer and the Si concentration in the n-type GaN layer of the p + -n junctions were 4.6 × 10 18 and 5.0 × 10 16 cm −3 , respectively, while the values in the p-n + junctions were 1.6 × 10 17 and 5.2 × 10 18 cm −3 , respectively.A vertical mesa structure was fabricated by inductively coupled plasma reactive ion etching to a depth of 4 μm from the surface of the p ++ /p/n structure.Metallic Ni/Au and Ti/Al/Ni layers were deposited on the frontside and backside, respectively, as ohmic contacts.The specimens were subsequently sintered at 773 K for 10 min under O 2 .
The fabricated diodes were exposed to 137 keV EB radiation with fluence values of 2.0 × 10 16 , 2.0 × 10 17 , 4.9 × 10 17 , 9.8 × 10 17 or 2.0 × 10 18 cm −2 and an electron current density of 1.6 × 10 −5 A•cm −2 .The p + -n junction was exposed to additional EB irradiation with fluence values of 4.9 × 10 15 and 9.8 × 10 15 cm −2 and electron current densities of 9.7 × 10 −6 and 1.3 × 10 −5 A•cm −2 , respectively.Monte Carlo simulations 27) of electron trajectories were performed for the entire Ni/Au/GaN diode structure and the energy distributions of electrons passing through the p-n junctions were determined.In the case of the p + -n junctions, the proportion of electrons with an energy above 125 keV (the threshold energy for N atom displacement) 25) relative to the total number of electrons that were supplied was found to be almost 100%.A lower value of 95.4% was determined for the p-n + junctions due to the thicker p-type layer.These results indicate negligible reductions in the EB energy and fluence prior to the p-n junctions.The energy levels and concentrations of traps introduced by the EB irradiation were evaluated by isothermal capacitance transient spectroscopy (ICTS) analyzes using a PhysTech HERA FT1230 instrument.These assessments employed the constant capacitance mode, which allowed trap concentrations to be measured with a high level of precision even in the case of high concentrations.In these experiments, the ICTS signal was acquired in conjunction with deep level transient Fourier spectroscopy. 28)igure 1(a) shows the forward current density-voltage (J-V ) characteristics and the ideality factor values for GaN p + -n junction and p-n + junction diodes before and after EB irradiation with a fluence of 4.9 × 10 17 cm −2 at RT (301 K).All diodes both before and after exposure to the EB were found to have an ideality factor of 2 over the voltage range of 2.0-2.5 V.These results indicate that the SRH recombination current, J SRH , was dominant in the subthreshold voltage range.The value of J SRH can be calculated as 22) s s Here, n i is the intrinsic carrier density, k is the Boltzmann constant, T is the measurement temperature, F 0 is the electric field at the position at which the recombination rate is maximized, N T is the RC concentration, s n and s p are the electron and hole capture cross-sections, and v th,n and v th,p are the thermal velocities of carriers (which are, in turn, dependent on the effective masses).The present study used the electron and hole density-of-state effective masses of 0.2 m 0 29) and 1.7 m 0 , 10) respectively.
Capacitance-voltage (C-V ) measurements were also performed to ascertain net doping concentrations, N net = N a N d /(N a + N d ), before and after EB irradiation.Here, N d and N a are the net donor and acceptor concentrations in a p-type layer and an n-type layer, respectively.The changes in the net doping concentration (ΔN net = N net,after -N net,before ) obtained from C-V data are presented in Fig. 1(b).Interestingly, the ΔN net values determined for the p + -n junction diodes were negligible compared with N net,before (4.2-4.5 × 10 16 cm −3 ).The magnitude of ΔN net (|ΔN net |) for the p-n + junction diodes increased with increases in the fluence and values were comparable to N net,before (1.7-2.4 × 10 17 cm −3 ).Irradiation of the p-n + junction diodes at a fluence of 2.0 × 10 18 cm −2 resulted in high resistivity, indicating that all the acceptors in the p-type layer were compensated for by deep donor-type traps induced by the irradiation.Hence, the threshold fluence required to impart high resistivity to the p-n + junctions was much lower than that for the p + -n junction.These results are notably consistent with the trends observed in prior research concerning RCs induced by proton irradiation that displaced both Ga and N atoms. 23)The difference in the change in N net observed for the p + -n and p-n + junctions can possibly be attributed to variations in the  Figure 2 summarizes the effect of the junction diameter on the current density extrapolated to the vertical axis (J SRH,0 ) before and after irradiation at a fluence of 4.9 × 10 17 cm −2 .Note that, because the initial J SRH,0 values were so small, they have been increased by a factor of 20 in this plot.Here, J SRH,0 is seen to have increased with increases in the ratio of the periphery to the area (PA −1 ), indicating the existence of a surface recombination current.The bulk recombination current as considered in this work could be obtained from the J SRH,0 value at a PA −1 value of 0 while the surface recombination current at the mesa sidewall equaled the slope of the J SRH,0 data when plotted.
The bulk recombination current and the surface recombination current generated by the diodes prior to irradiation were both small.However, the J SRH,0 values associated with the p-n + junction diodes were increased significantly following exposure to the EB, suggesting increases in the bulk recombination current.This current increased along with the PA −1 value, meaning that the surface recombination current was also enhanced by the irradiation.The J SRH,0 values for the p + -n junction diodes were also increased significantly after irradiation.Even so, these values were unaffected by PA −1 , suggesting that the bulk current was dominant.An analysis of the relationship between J SRH and the junction diameter allowed the bulk recombination current to be determined and t SRH to be estimated using Eq.(1).
Figure 3 shows the values of t SRH and s s N T n p obtained from calculations of J SRH as functions of the EB fluence.Note that a bandgap of 3.43 eV was used for the calculation of n i . 30)The values of t SRH for both diode types were found to decrease with increases in the fluence, reflecting the formation of RCs as a consequence of the irradiation.The values of s s N T n p for the p + -n junctions exhibited a linear correlation with the fluence whereas those associated with the p-n + junction showed a super-linear increase.The product of the RC production rate (PR) and s s n p ( s s =PR n p ) was determined from the slope of the data as a function of fluence.A value of 2.8 × 10 −15 cm was determined for the p + -n junction.In the case of the p-n + junction, s s N T n p increased linearly with a slope of 4.2 × 10 −15 cm for fluence values of less than 2.0 × 10 17 cm −2 .However, at higher fluences, non-linear increases were observed, suggesting that the formation of RCs via N atom displacement involved at least two components.
The s s PR n p values obtained in this study were compared with s s PR n p calculated using the parameters related to N-displacement-related traps acquired through DLTS trials. 25,26)he results are summarized in Table I, which provides the activation energy, majority carrier capture cross-section and PR data generated on the basis of the DLTS analyzes.A minority carrier capture cross-section of 1 × 10 −13 cm 2 was assumed.Note that this is a very large value that is commonly observed in the case of large defects such as divacancies.Calculated s s PR n p values are also included in Table I.It should be noted that the EE2 parameters for n-GaN (that is, an activation energy of 1.1 eV, PR of 0.063 cm −1 and s n of 1.6 × 10 −14 cm 2 ) were newly obtained from ICTS measurements using the p + -n junction diodes fabricated in this study.
With regard to the p + -n junction diodes, a very large value of s p was assumed and the calculated s s PR n p for EE1 traps was four orders of magnitude smaller than that obtained from Fig. 2. The effect of junction diameter on the current density as extrapolated to the vertical axis (J SRH,0 ) for GaN p-n junction diodes before and after EB irradiation at a fluence of 4.9 × 10 17 cm −2 .The J SRH,0 values for the specimens prior to EB irradiation have been increased by a factor of 20 to assist in making comparisons.Table I.Various parameters of N-displacement-related traps generated in this work as obtained from DLTS 25,26) ) and constant capacitance ICTS data.Calculated s s PR n p values generated with the assumption of a minority carrier capture cross-section of 1 × 10 −13 cm 2 are also provided.

Trap type
Activation energy/eV s majority /cm 2 PR/cm −1 s s PR n p /cm EE1 25) 0.13 2 × 10 −18 However, while the s s N T n p values for the p + -n junctions increased together with increases in the EB fluence, as shown in Fig. 3, the EE2 concentrations (N EE2 ) determined using ICTS exhibited a sub-linear relationship with fluence, as can be seen in Fig. 4.These findings suggest that EE2 traps were not serving as RCs in the present devices or there was another origin for RCs that became significant at higher fluences.The origins of RCs were not determined using DLTS in previous studies and so may be located deeper than the present EE2 traps, which were at 1.1 eV.In the case of the p-n + junctions, the calculated EHa value was close to the experimentally obtained value of 4.2 × 10 −15 cm, so this could be the origin of RCs at low fluences.However, the s s N T n p value for the p + -n junctions increased rapidly (super-linearly) at higher EB fluences, as shown in Fig. 3. Hence, other defects such as vacancy clusters may also act as RCs.
In conclusion, this work studied the RCs formed in homoepitaxial GaN p + -n and p-n + junctions in response to EB irradiation at an energy of 137 keV, based on the selective displacement of N atoms.The net doping concentrations in the p + -n junction diodes were not changed by this irradiation.However, the concentrations in the p-n + junction diodes decreased as the EB fluence was increased.An analysis of the recombination currents associated with GaN p-n junctions demonstrated that N-displacement-related defects acted as RCs.
s s PR n p values of 2.8 × 10 −15 and 4.2 × 10 −15 cm for n-type and p-type GaN, respectively, were obtained from plots, confirming the fluence dependence of the recombination current.Comparing this parameter with the trap parameters obtained from previous studies established that the formation of RCs in both p + -n and p-n + junctions can be partly explained on the basis of N-displacement-related traps previously detected using DLTS.However, other unknown traps should also be considered.The introduction of N-displacement-related defects in these junctions induced two phenomena.Firstly, N net was constant in the p + -n junctions but decreased in the p-n + junctions and secondly the value of s s PR n p in the p-n + junctions was larger than that in the p + -n junctions.The results of the present work provide crucial information that improves our understanding of the point defects introduced during the fabrication of GaN-based vertical FETs with p-n junctions consisting of both p-type and n-type regions.

Fig. 1 . 2 ©
Fig. 1.(a) Forward J-V characteristics and the ideality factors for as-grown GaN p + -n and p-n + junction diodes before and after exposure to an EB fluence of 4.9 × 10 17 cm −2 at 301 K.The calculated recombination current fitted to the experimental values is also shown as the dashed lines.(b) The effect of fluence on the change in N net after EB irradiation (ΔN net = N net,after −N net,before ) based on data obtained from C-V analyzes of GaN p + -n and p-n + junction diodes.The N net,before varied over the range from 4.2 × 10 16 to 4.5 × 10 16 cm −3 for the p + -n junction diodes and from 1.7 × 10 17 to 2.4 × 10 17 cm −3 for the p-n + junction diodes.

Fig. 3 .
Fig. 3.The t SRH a n d N T (σ n σ p ) 1/2 values obtained from recombination current data acquired at 301 K as functions of the EB irradiation fluence.The t SRH values for both n-GaN and p-GaN decrease with increases in fluence.

Fig. 4 .
Fig. 4. N EE2 as a function of the EB irradiation fluence for an energy level of 137 keV.The inset plots N EE2 as a function of fluence for lower fluence values.The production rate of 0.063 cm −1 obtained in the low fluence region is shown as a red dashed line in both plots.
2024The Author(s).Published on behalf of The Japan Society of Applied Physics by IOP Publishing Ltd the SRH analysis (2.8 × 10 −15 cm).Therefore, EE1 traps are not thought to be RC candidates in this research.Conversely, the value calculated for EE2 traps was close to that generated by the SRH analysis and these traps could possibly act as RCs. ©