Scanning internal photoemission microscopy measurements of n-GaN Schottky contacts under applying voltage

We applied scanning internal photoemission microscopy (SIPM) to characterize the degradation of GaN Schottky contacts formed on a thick n-GaN layer grown on a freestanding GaN substrate by in situ applying reverse bias voltage (Vbias) down to −45 V. For most of the contacts, uniform distribution of the photocurrent was observed over the electrode with the visible lasers. Irregular-shape regions with 5%–25% larger photocurrent appeared with the near UV laser by applying Vbias, but the I–V characteristics were stable. On the other hand, for the contacts with a slightly larger reverse current, the photocurrent distribution was also uniform at Vbias = 0 V, but over Vbias = −36 V, the photocurrent was intensively increased at small spots. After the SIPM measurements, the I–V characteristics became leaky, and the same spots were observed in the microscope image. These results indicate that SIPM is useful for in situ monitoring of the initial stage of the degradation under applying reverse bias voltage.


Introduction
GaN based electron devices have been employed in a variety of commercial RF-power-amplifier applications, [1][2][3][4][5] because of their characteristics of high-breakdown field and high electron velocity. Recently, the development of high-quality freestanding GaN substrates [6][7][8] has accelerated the development of GaN power switching devices. Freestanding GaN substrates allow epitaxial growth of the thick drift layers with low dislocation density, which is indispensable for the diodes to achieve highbreakdown voltages. To achieve GaN Schottky and p-n diodes with very high-breakdown voltages (over 1 kV), [9][10][11] a free carrier concentration of 1 × 10 16 cm −3 or lower is required in the n-GaN thick drift layer. Freestanding GaN substrates allow the homo-epitaxial growth of thick GaN layers without wafer curvature or cracks, which have been a typical problem for lattice-mismatched substrates. However, controlling the carrier concentration in the thick drift layer is still difficult, and the effects of residual C acceptors and other point defects must be taken into account. 12,13) We also reported uniformity of carrier concentration in conjunction with surface morphology over the wafer. 14) The surface off-angle of the GaN substrate affected incorporated C concentration during the growth, which resulted in uniformity of the carrier compensation in a low-carrier n-GaN drift layer. 15) On the other hand, we have developed a new two-dimensional mapping characterization termed scanning internal photoemission microscopy (SIPM) to verify the electrical inhomogeneity of metal-semiconductor (M/S) interfaces. 16,17) Thus far, we have demonstrated the mapping of characteristics in interfacial reactions, degradation under applying voltage stress and surface damage in Si, GaAs, GaN, IGZO, and SiC Schottky contacts. [18][19][20][21][22][23][24][25] We also demonstrated the characterization of crystal quality of wafer-bonded Si/SiC heterointerfaces. 26) We confirmed that macroscopic mapping investigation for the entire electrodes by using this technique is a powerful tool for developing high-power wide-bandgap electron devices.
We came up with applying SIPM to characterize the initial stage of the degradation of the Schottky diodes that were formed on a low-carrier thick n-GaN layer grown on a freestanding GaN substrate. In this paper, we propose two new functions in the SIPM measurement. We have been using visible lights for probing M/S interfaces. We adopted near ultra violet (NUV) light, which can enable us to observe not only M/S interfaces but also depletion layers. In addition, we applied reverse bias voltage during the measurements, which can provide in situ observation of the degradation. In our previous report, we conducted demonstration of SIPM measurements for GaN Schottky contacts using the NUV light in Ref. 27. In this paper, we report in more detail the photocurrent transport mechanism including spectroscopic analysis associated with the bias voltage. Figure 1 shows the device structure of the Ni/GaN Schottky contact. Freestanding GaN substrates were prepared by the unique void-assisted separation (VAS) technology of SCIOCS, which is based on the hydride vapor phase epitaxy method. [6][7][8] The dislocation density was 3 × 10 6 cm −2 , and the donor concentration was 2 × 10 18 cm −3 in the substrates. We grew 2 μm n-GaN layers, doped with 2 × 10 18 cm −3 silicon, as an access region, followed by 12 μm thick n-GaN drift layers, where the target free carrier density was controlled to 1 × 10 16 cm −3 . After a HCl surface treatment, circular 100 nm thick Ni Schottky contacts (200 μm in diameter) were deposited by the electron beam evaporation. After that, InGa ohmic contacts were deposited on the back surface.

Device fabrication and measurement
We measured I-V characteristics of the fabricated devices by using a semiconductor parameter analyzer (HP4142B). We determined the Schottky barrier height (qf B ) and an ideality factor (n-value) by using the thermionic emission model: 28) where A * is the effective Richardson constant (26.4 A cm −2 K −2 for n-GaN based on A* = 4πm*qk 2 /h 3 and m* = 0.22 m 0 ), T is the temperature, q is the electron charge, k is the Boltzmann constant, and V is the applied voltage.
Prior to the SIPM measurements, we conducted conventional photoresponse (PR) measurements where a photon energy (hν) was continuously scanned from 1.2 to 4.0 eV, in order to confirm the shape of the PR spectrum of the Ni/n-GaN contact. In the PR measurements, a monochromatic light from a monochrometer was used and irradiated to the entire contact area. As shown in Fig. 2, when a monochromatic light with hν that is greater than qf B is incident on a metal/n-GaN interface, carriers in the metal can surmount the Schottky barrier and a photocurrent may be generated, which is called the internal photoemission effect. The relationship between hν and the photoyield (Y), which corresponds to the photocurrent per the number of incident photon, is given by Fowler's equation: 29,30) The qf B can be determined from Eq. (2). In the SIPM measurements, we focused and scanned the laser beam over the interface of the electrode to obtain a two-dimensional image of Y. One mapping measurement takes about 1 h. The Y imaging measurements were repeated with different wavelengths to obtain an estimate of qf B by Eq. (2). In this case, all the information is from the M/S interface.   When hν is close to the fundamental absorption edge, a large photocurrent always flows because of the generation of electron-hole pairs as in a solar cell, and the linear relationship that is defined in Eq. (2) no longer holds. We can observe both M/S interface and depletion region. In this SIPM study, we used red (λ 1 = 659 nm, hν 1 = 1.88 eV) green (λ 2 = 517 nm, hν 2 = 2.40 eV) and blue (λ 3 = 447 nm, hν 3 = 2.77 eV) light lasers for the internal photoemission, and NUV (λ 4 = 375 nm, hν 4 = 3.31 eV) for the fundamental absorption. The beam spot diameter was estimated as 2 μm. In addition, the SIPM measurements were conducted with applying reverse bias voltage (V bias ) down to −45 V to the contact for the in situ observation of the degradation. Conventional PR measurements were conducted, where a monochromatic light was illuminated over the entire contact for Dot 1. Figure 4 shows the entire PR spectra under application of different V bias from 0 to −45 V. As mentioned in Sect. 2, the square root of Y linearly increased up to approximately 3.2 eV. We noticed that the PR spectra are independent of the V bias values in this hν range. It is likely because excited electrons in the vicinity of the interface on the metal side can surmount the Schottky barrier and contribute to the photocurrent. We assumed that the Y obeys Eq. (2). Therefore, the values of qf B were obtained to be 1.20 eV, which is close to the qf B value obtained from the I-V measurement. Y steeply increased from approximately hν = 3.2 eV; then, when hν was larger than 3.3 eV, Y dramatically decreased because of the fundamental absorption in the GaN epitaxial-layer and substrate as explained in Sect. 2. Even though hν is slightly smaller than E g , it can be considered that the excitation is possible via impurities and excitons. In this higher hν range, the PR signal significantly increased as V bias decreased. We speculate that the depletion layer width became wider under large |V bias |, and the number of electron-hole pairs which can contribute photocurrent increased. Therefore, we confirmed that we can investigate both M/S interface and the depletion layer in this higher hν range. Basically, the same PR results were obtained from Dot 2. Judging from these results, we can estimate that the SIPM results with the red, green, and blue lasers are based on the internal photoemission, and that with the NUV laser includes the effect of the fundamental absorption.   Figure 5 shows typical Y maps measured with the (a) red, (b) green, (c) blue, (d) NUV lasers and (e) qf B maps of Dot 1 when V bias = 0 V and V bias = −45 V. In the Y maps with the visible lasers, uniform distribution of the photocurrent was observed over the electrode. In the qf B map calculated from these Y maps, uniform distribution was obtained as well. As expected in the PR results, the Y signals are independent of V bias . The averaged qf B value is 1.24 eV. We confirmed that our device fabrication process provided properly uniform M/ S interfaces. On the other hand, as for the NUV map, the Y intensity increased from V bias = 0 V to V bias = −45 V as expected in the PR results. In addition, irregular-shape regions with 5%-25% larger Y appeared by applying V bias .

Mapping characterization
We also conducted the SIPM measurements with the NUV laser with changing V bias for more detailed investigation. As shown in Fig. 6, the number and area of such regions increased as V bias decreased. It can be considered that this inhomogeneity is located in the semiconductor side of the M/ S interface. The origin of the inhomogeneity is not clear at this moment, but cluster of the threading dislocations may be possible. The I-V curves are the same before and after the SIPM measurements, because the applied |V bias | is much smaller than the typical catastrophic breakdown voltage of |−500| V. Even though the inhomogeneity existed, Dot 1 was stable.
We conducted the SIPM measurements for Dot 2 in the same manner for Dot 1. Basically, the same uniform results were obtained with the visible lasers as those of Dot 1. However, during the NUV measurement with changing V bias , two spots appeared within the large Y regions at V bias = −36 V as shown in Fig. 7. The maximum intensity of Y at the spots is more than one-order-of-magnitude larger than that of the surrounding region. It is likely that these spots   were formed by the voltage stress. We succeeded in in situ monitoring of the initial stage of the degradation. Figure 8 shows the optical microscopy images of Dot 2 (a) before and (b) after the SIPM measurements. The same spotted pattern as the Y map at V bias = −45 V in Fig. 7 is observed in the image after the measurements. In the spots, the electrode surface became rough. As well as in the Y maps with the green laser at V bias = 0 V (a) before and (b) after the SIPM measurements as shown in Figs. 9(a) and 9(b), the same results were obtained. Unfortunately, the qf B value was not determined in the spots as shown in Fig. 9(c), because the photocurrent was so small and unstable. Figure 10 shows the I-V characteristics before and after the SIPM measurements. In the forward I-V characteristics, slight increase of the current is observed up to +0.8 V, which indicates slight decrease of qf B . In the reverse characteristics, the current level is the same down to −4 V, but the leakage current increased to 200 pA at −50 V. We can speculate that the degraded region is not uniform and consists of two interfaces. One interface is reacted by the voltage application, which is not flat and shows leaky characteristics especially under a large reverse bias voltage. The other one is the burned-out interface by more intensive interfacial reaction, which induces a loss of the active area of the electrode. These two regions are microscopically mixed in the degraded region under the spatial resolution of SIPM. As for the SIPM result with the green laser without a bias voltage, as the area loss is dominant, the signal was small in the degraded region. On the other hand, with the NUV laser, as the carrier generation was enhanced by induced defects upon the interfacial reaction in the vicinity of the interface, a large signal was obtained. Therefore, we demonstrated that SIPM clearly visualized an initial stage of the degradation sensitively in conjunction with the electrical characteristics.
One possible model of the degraded mechanism might be existence of bunches of dislocations. We used the VAS technology for the substrate growth, but the distribution of the dislocation is not completely uniform. The large Y regions may include bunches of dislocations, which can induce largestructure defects. For further study, cathode luminescence and X-ray topography measurements of the GaN surfaces are required to clarify the origin of such regions.

Conclusions
SIPM was applied to characterize the initial stage of the degradation of Ni/n-GaN Schottky contacts formed on a thick n-GaN layer grown on a freestanding GaN substrate with applying V bias down to −45 V.
Most of the contacts showed small reverse biased current below the noise floor at −50 V, and uniform distribution was observed in the Y maps over the electrode with the visible lasers. Irregular-shape regions with 5%-25% larger Y appeared with the NUV laser by applying V bias , but the I-V characteristics were stable.
On the other hand, for the contacts with a slightly larger reverse current of around a pA order at −50 V, the Y maps were also uniform at V bias = 0 V, but over V bias = −36 V, two spots appeared within the large Y regions at V bias = −36 V. After the SIPM measurements, the I-V characteristics became leaky, and the same spots were observed in the microscope image of the electrode. We demonstrated that SIPM clearly visualized an initial stage of the degradation as an in situ monitoring in conjunction with the electrical characteristics.