Silver Nanowire Transparent Conductive Electrodes for High-Efficiency III-Nitride Light-Emitting Diodes

Silver nanowires (AgNWs) have been successfully demonstrated to function as next-generation transparent conductive electrodes (TCEs) in organic semiconductor devices owing to their figures of merit, including high optical transmittance, low sheet resistance, flexibility, and low-cost processing. In this article, high-quality, solution-processed AgNWs with an excellent optical transmittance of 96.5% at 450 nm and a low sheet resistance of 11.7 Ω/sq were demonstrated as TCEs in inorganic III-nitride LEDs. The transmission line model applied to the AgNW contact to p-GaN showed that near ohmic contact with a specific contact resistance of ~10−3 Ωcm2 was obtained. The contact resistance had a strong bias-voltage (or current-density) dependence: namely, field-enhanced ohmic contact. LEDs fabricated with AgNW electrodes exhibited a 56% reduction in series resistance, 56.5% brighter output power, a 67.5% reduction in efficiency droop, and a approximately 30% longer current spreading length compared to LEDs fabricated with reference TCEs. In addition to the cost reduction, the observed improvements in device performance suggest that the AgNWs are promising for application as next-generation TCEs, to realise brighter, larger-area, cost-competitive inorganic III-nitride light emitters.

Transparent conductive electrodes (TCEs), which are used as p-type spreading contacts in III-nitride light-emitting diodes (LEDs), are of particular importance because of their direct influence on light extraction efficiency, forward voltages, device reliability, and production cost [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15] . In the III-nitride LED field, one technological breakthrough in the past decade was adopting indium-tin-oxide (ITO) electrodes, because these electrodes yield a high optical transmittance (T op ) (approximately 93% at 450 nm), a low sheet resistance (R sh ) (~15 Ω /sq (commercial-grade)), and a feasible ohmic contact to p-GaN [16][17][18][19][20] . Despite ITO's good electrical and optical properties, it is important to identify an alternative to this material for the following reasons. First, the use of ITO materials is not cost-competitive owing to the rapid depletion of the elemental indium source and the complex vacuum systems required for the deposition of ITO films. Second, ITO deposition by conventional sputtering processes results in ion damage to the p-GaN surface 21,22 , degrading the ohmic contact and causing a leakage current to evolve. To overcome this damage-related issue, conventionally, an e-beam evaporator and/or a modified sputtering system are used. In addition, elaborately designed p-contact layers, such as superlattices 23,24 and tunnel junctions 25,26 , have been employed because of their reduced sensitivity to the ion damage arising from the use of an n-type conductive top layer. However, these methods suffer from the relatively poor electrical

Results
Electrical and optical properties of AgNWs. Figure 1a shows scanning electron microscopy (SEM) top views and bird's-eye views of AgNW films coated on a sapphire substrate before and after thermal annealing at 100 and 150 °C in an N 2 ambient. The as-coated sample showed a typical squashed arrangement of AgNWs with an average diameter of ~30 nm and average length of a few tens of micrometres, which is consistent with the literature [45][46][47] . After thermal annealing, AgNWs were agglomerated and disconnected, indicating that the as-coated condition was the best in terms of T op and electrical conductivity. Figure 1b shows the optical specular transmittance spectra of the as-coated and annealed AgNW films; for comparison, the transmittance spectra of reference TCEs, including the oxidized Ni/Au with a different overall thickness of 7, 10, and 14 nm, e-beam evaporated 400 nm-thick ITO films, and sputtered 100 nm-thick ITO films, are also plotted. As expected, the as-coated AgNW sample produced the highest T op among the AgNW samples. Furthermore, this AgNW sample had high T op across the entire wavelength region tested, except at 380 nm, which is associated with the transverse plasmon mode of AgNWs 42 . The T op of the ITO films was high in the visible wavelength region but began to drop significantly at wavelengths below 400 nm owing to ITO's band-to-band absorption. The T op of the oxidized Ni/Au samples showed significant negative dependence on their thickness, i.e., the thicker a Ni/Au film was, the lower its T op was. In addition, the Ni/Au samples showed relatively poor T op at wavelengths below approximately 350 nm and above 600 nm, which is directly associated with the optical absorption spectra of Au films 49 . In this study, the primary wavelength of interest was 450 nm because the developed TCEs were applied to commercial LED wafers with blue emission (as plotted in Fig. 1b), where the EL spectrum of the fabricated LEDs was acquired at 10 mA. In addition, the T op at the deep UV wavelength of 260 nm (T op@260nm ) was monitored in the study because this wavelength is of particular interest to the improvement of the extraction efficiency for deep UV LEDs 50 .
To determine the best TCEs, R sh should also be taken into consideration. Figure 1c summarizes the T op versus R sh plots of the as-coated AgNWs, oxidized Ni/Au films with the overall thicknesses of 7, 10, and 14 nm, and ITO films. Evidently, AgNW films showed the best figures of merit: T op@260 nm = 87.8%, T op@450 nm = 96.5%, and R sh = 11.7 Ω /sq. Meanwhile, the ITO film had relatively poor electrical and optical properties, i.e., T op@260 nm = 29.8/16.0/7.6%, T op@450 nm = 98.4/91.0/85.0%, and R sh = 63.6/39.2/12.7 Ω /sq for the e-beam evaporated 200/300/400 nm -thick ITO films and T op@260 nm = 20.0%, T op@450 nm = 91.5%, and R sh = 24.6 Ω /sq for the sputtered ITO films. Notably, the R sh of Ni/Au films (as well as their T op ) depended strongly on their layer thickness: R sh = 27.1/15.4/11.3 Ω /sq and T op@450 nm = 94.2/87.1/81.1% for the overall thicknesses of 7, 10, and 14 nm. These results indicate that the as-coated AgNW films are promising as TCEs for enhancing the performance characteristics of LEDs for blue and deep UV applications. In this study, blue-emitting LEDs were adopted to verify the usability of AgNW TCEs because they are regarded as the standard platform in terms of industrial mass production. In addition, Ni/Au films were used as the reference electrodes because their T op and R sh values were comparable to those of ITO films (although the properties were slightly poorer than those of commercial-grade ITO) and the ohmic contact to p-GaN was very good, as will be discussed below. The optical and electrical properties of the Ni/ Au and AgNW electrodes are presented in Table 1.

Contact properties of AgNWs.
To fabricate reliable LEDs with AgNW electrodes, it is crucial to form AgNW ohmic contact to p-GaN. To investigate the contact property, the specifically designed TLM  Ni/Au (7 nm Figure 2b shows the current-voltage (I-V) curves of the AgNWs and 7, 10, and 14 nm-thick Ni/Au TLM pads, as measured from adjacent contact pads with spacing of 10 μ m. It is evident that the I-V curves of the Ni/Au electrodes are much steeper than that of the AgNW electrode. In addition, both reference Ni/Au electrodes and the AgNW electrode produced nearly linear I-V curves (for the 10-and 14-nm-thick Ni/Au) and very slightly nonlinear I-V curves (for the 7-nm-thick Ni/Au and AgNWs), indicating the formation of ohmic contact. For example, the specific contact resistance (ρ sc ) estimated based on the measured electrical resistance (R) at 1.0 V, i.e., R = V/I = 1.0 V/I, was as low as 2.2 × 10 −3 and 3.5 × 10 −3 Ω cm 2 for the 14 nm-thick Ni/Au electrodes and AgNW electrodes, respectively. However, this prediction might be inaccurate because of the nonlinearity of the I-V curves, as will be discussed in detail below.
One of the major concerns in Fig. 2b is the different amount of current flow, which is particularly pronounced in AgNW electrodes. The possible origins for the significantly reduced I-V slopes can be inferred based on the thickness dependence of the reference Ni/Au electrodes: the higher the sheet resistance is (i.e., the thinner the electrode is), the worse the I-V slope is. Based on this tendency, the lowest I-V slope of the AgNW electrode might be due to the degraded R sh of the AgNWs. Indeed, this explanation is plausible considering that the electrical conductivity of the AgNWs might be degraded due to the use of an overlying structure of AgNWs on the bottom p-GaN surface and the Pt pads; specifically, the AgNWs at the boundary region of the p-GaN window and the 30 nm-thick probing pads might be disconnected as a result of the height difference. This possibility was confirmed by analysing the probing-position-dependent I-V curves of AgNW films ( Supplementary Fig. 1). For example, with a constant probing distance of approximately 80 μ m, the I-V curve obtained by probing only the C region was steeper than that obtained by probing both the B and C regions. Another possible origin, which appears to be a much more dominant factor, is the difference of contact area. For example, whereas the reference electrodes cover the entire p-GaN surface defined by the photographic patterning, the AgNWs occupy only a small fraction of the defined area owing to their zero-or one-dimensional NW contacts. Consistently, the measured R sh values of the p-GaN layer determined from the AgNWs and 14-nm-thick Ni/Au samples according to the TLM theory were 2.9 × 10 7 and 1.0 × 10 5 Ω /sq, respectively. Despite the use of the same wafer, the two orders of magnitude higher p-layer resistance obtained from the AgNW sample is indicative of reduced contact area.
One thing we have to remind is the nonlinearity of the I-V curves, particularly pronounced for the AgNW electrodes. Nonlinearity was even observed for the 14nm-thick Ni/Au electrodes when the V-I curve was plotted in the voltage range from − 5 to + 5 V (Supplementary Fig. 2). This suggests that a corrected method should be taken into consideration to evaluate ρ sc accurately for our samples because the conventional TLM theory is only valid when the I-V curve is perfectly linear or when perfect ohmic contact is formed. Recently, Piotrzkowski et al. 51 , while investigating corrected ρ sc values when the I-V curve was nonlinear, proposed a universal method to obtain bias-voltage-dependent or injected-current-density-dependent ρ sc values according to where a is the width of the TLM pattern and R c = V/I 0 and R c * = dV/dI 0 are the common and differential front resistances of single planar contact in TLM structure, respectively ( Supplementary Fig. 3) 51 . Using this model, reasonable ρ sc values were obtained as a function of the bias voltage and current density (J), as shown in Fig. 2c. Note that the ρ sc values of the AgNW electrodes are more significantly dependent on the bias voltage and current density than those of the Ni/Au electrodes. As a result, whereas the ρ sc value of the AgNW electrodes was higher than that of the reference 14 nm-thick Ni/Au electrodes at V < 1.95 V, those of AgNW electrodes became much lower at V > 1.95 V, e.g., ρ sc = 1.2 × 10 −2 and 6.8 × 10 −3 Ω cm 2 at V = 1.0 V and 2.5 × 10 −4 and 6.3 × 10 −4 at V = 5.0 V for the AgNW and Ni/Au electrodes, respectively. A possible cause of the stronger field dependence of the AgNW electrodes than the Ni/Au electrodes is the zero-or one-dimensional NW contact. For example, higher bias induces more current to flow through a small cross section, resulting in Joule heating. The temperature rise in such a localized area leads to a ρ sc reduction owing to enhanced thermionic emission 52 and/or improved adhesion between AgNWs and p-GaN via interfacial reaction, causing more current to flow and accelerated Joule heating. Notably, this process would be self-reinforcing, leading to a self-accelerating thermal process. Indeed, it is instructive to recall how the observed field-enhanced NW ohmic contact affects the electrical and optical performance characteristics of AgNW-LEDs, particularly at higher bias voltages. Figure 3a shows the optical microscopic top views and schematic cross-sectional diagrams of AgNW-LEDs and reference LEDs fabricated with Ni/Au electrodes. In this figure, it is shown that the AgNWs were formed on the bottom p-GaN surface and Au probing pads. In contrast, the Ni/Au electrode was first deposited on the p-GaN, on which the probing pad was formed. The length (L) and width (W) of the mesa were 200 and 500 μ m, respectively. Figure 3b shows the I-V curves of AgNW-LEDs and reference LEDs. As expected from the TLM results, the forward I-V curves of the AgNW-LEDs showed distinctive behaviour compared to the reference LEDs. For example, the current flow of AgNW-LEDs was small in the low voltage range but began to rapidly increase with increasing bias voltage, particularly above ~4.0 V. Accordingly, the series resistance of the devices (R s ) extracted according to the relation I(dV/dI) = IR s + nkT/q, where n is the ideality factor, k is the Boltzmann constant, and T is the absolute temperature, was 9.2 Ω for the AgNW-LEDs and 20.5 Ω for reference LEDs fabricated with 14 nm-thick Ni/Au electrodes (Fig. 3b inset). This finding is quite interesting because, despite the expected smaller effective active area, the AgNW-LEDs showed a 55% lower R s value compared to the reference LEDs. As discussed above, a possible mechanism for this finding might be field-enhanced carrier transport through the AgNW/p-GaN interface, as observed in Fig. 2c. Figure 3c shows the optical output power versus injection-current curves of the LEDs. Notably, the AgNW-LEDs had significantly higher optical output power than the reference LEDs. For example, the optical output powers measured at 80 mA were 14.4, 25.8, 28.3, and 44.3 mW for the LEDs fabricated with 7, 10, and 14 nm-thick Ni/Au electrodes and an AgNW electrode, respectively. In the reference LEDs, the significant drop of optical output power with decreasing Ni/Au thickness (despite its higher optical transmittance) was attributed to the degraded current spreading associated with increased R sh values, i.e., operation in the current-spreading limited regime 53 , as shown in the electroluminescence (EL) images of Fig. 3c. Specifically, a significant current crowding was observed near the p-probing pads; current crowding is regarded to originate from the poor electrical conductivity of the Ni/Au electrodes rather than that of the n-GaN layer (13.9 Ω /sq) 53 Here, the R sh value of the n-layer was estimated from an n-contact using the TLM method. Therefore, LEDs fabricated with 14 nm-thick Ni/Au electrodes are hereafter referred to as the true reference LEDs because they performed the best. Once more, note that the AgNW-LEDs showed optical output power 56.5% higher than that of the reference LEDs.

Discussion
The significantly improved optical output power of AgNW-LEDs is essentially due to the AgNW electrode having a higher T op (96.5%) than the Ni/Au electrode (81.1%). However, it is questionable whether the 18% increase in the T op can lead to the 56.5% higher optical output power observed. To investigate this question, an optical ray-tracing simulation was performed using the real device configuration and optical constants (Fig. 4a). Details of the optical ray-tracing simulations can be found elsewhere 54,55 . The calculations showed that the 18% increase in the T op could lead to ~17% higher optical output power, indicating that another factor also contributed to the enhanced output power of the AgNW-LEDs.
The efficiency droop, that is, the loss of external quantum efficiency at high injection current, was also found to significantly decrease in the AgNW-LEDs, as shown in Fig. 4b. For example, the efficiency droop values measured from the maximum peak intensity to those measured at 100 mA were 34.3 and 50.8% for the AgNW-LEDs and reference LEDs, respectively. Several studies have reported that the efficiency droop originates variously from crystallographic defects 56-59 such as point defects and threading dislocations, Auger recombination 60-62 , electron leakage 63,64 (caused by limited hole-injection efficiency 65 , ineffective electron blocking layers 66 , and incomplete carrier capture by quantum wells) 67 , current crowding 68 , and heating effects 69 . Among these possible origins, the effect of defects can be neglected because the same LED wafers were used in this study. In addition, the two representative origins of Auger recombination and electron leakage, which respond sensitively to the carrier density injected into the active region, can also be ignored because the efficiency droop of the AgNW-LEDs was greatly alleviated despite their much higher expected current density associated with the reduced active area. A reasonable explanation for the reduced efficiency droop, therefore, might be a reduction in the heating effect and/or improved current spreading, which would also be responsible for the improved optical output power. Indeed, the reduced heating effect in the AgNW-LEDs is acceptable according to the lower R s value (9.2 Ω ) and Joule's first law, i.e., Q∝ I 2 R s , where Q is the generated heat. Accordingly, the EL peak position of the AgNW-LEDs taken at the injection current of 80 mA was blue-shifted by 4 nm compared to the reference LEDs (Fig. 4c). This is consistent with the reduced heating effect explanation, considering that a temperature increase results in a red shift owing to bandgap shrinkage 70 . The blue shift of the EL spectra might also be due to the increased band filling in the AgNW-LEDs, which is associated with a higher density of injected current 7 .
To investigate another hypothesis that the current spreading is better in the AgNW-LEDs, we fabricated test structures with various lateral mesa lengths (L) ranging from 100 to 1200 μ m and collected EL images while operating the test structures at an injection current of 5 mA (Fig. 5a). The current is spread more uniformly in the AgNW-LEDs; specifically, the current spread perfectly for AgNW-LED structures with L = 400 μ m, and for reference LED structures of L = 200-300 μ m. To obtain the current spreading length (L s ), defined as the length over which the current density drops to the 1/e value at the mesa edge, the current densities J of both LEDs were plotted as a function of the forward voltage and L, as shown in Fig. 5b. Using the experimental J−L characteristics re-plotted as a function of the bias voltage (see the Supplementary Fig. 3), the L s was estimated by theoretical fitting of the experimental J−L data using the following equation: where J 0 is the current density at the mesa edge. The detailed theoretical background and the procedure to obtain the L s can be found elsewhere [71][72][73] . Figure 5c shows the L s values obtained as a function of the bias voltage. It is evident that the L s values of the AgNW-LEDs were longer than those of the reference LEDs by approximately 30 μ m, which is consistent with the EL images. Therefore, the enhanced optical output power and reduced efficiency droop of the AgNW-LEDs can also be due to the improved current spreading, which may contribute to the reduced R s value. However, the L s values (approximately 120 μ m) are likely to be underestimated. This is attributed to the lack of an exact theoretical model to describe our LED samples. Eq. (2) is only valid when the vertical p-layer resistances, including the p-contact and the p-GaN, are negligibly small [71][72][73] , whereas that of our samples, particularly for the AgNW-LEDs, is not small enough to ignore and, more importantly, has a significant bias voltage dependence. The better electrical and optical performance observed for AgNW-LEDs and their longer L s values suggest that AgNWs can be used as next-generation TCEs in large-area devices. The suitability of AgNW electrodes for large-area devices is further confirmed from their area dependence of optical output power, as shown in Fig. 5d. At a constant current injection of 80 mA, the optical output power increased with increasing L (or device area), which is essentially due to the reduced efficiency droop associated with the decreased current density. Interestingly, the optical output powers of AgNW-LEDs increased rapidly up to a high level at L ≤ 300 μ m and changed insignificantly at L ≥ 300 μ m, whereas those of the reference LEDs gradually increased with increasing L up to 1200 μ m. This result suggests that the AgNW TCEs are suitable for use in large-area devices owing to their higher optical output power and less area-dependent output efficiency, as demonstrated in Fig. 6. Figure 6 shows the EL images of 1 cm × 1 cm large-area LEDs fabricated with AgNW TCEs (at 50 mA). The devices were simply fabricated using a quick-test structure 74 and indium as both the n-and p-contacts. The LEDs showed bright light emission across the entire active region, suggesting that the large-area LEDs could be fabricated using efficient AgNW spreading contact.
So far, the field-enhanced NW ohmic contact (making a strong localized current flow underneath the NW contact), the improved current spreading, and excellent optical transmittance were found to cause the enhanced performance characteristics of AgNW-LEDs. Although improved current spreading was observed in the EL images, the images do not constitute evidence of localized current flow underneath the AgNW contact. It is also questionable whether the localized current flow through the NW contact could be large enough to explain the improved electrical properties. For this purpose, spatially-resolved EL images were observed using confocal scanning electroluminescence microscopy (CSEM) 75,76 (Fig. 7). For this measurement, the standard LED structure with surrounding n-electrodes was used ( Supplementary  Fig. 4). The dimensions of the mesa were 400 μ m × 500 μ m. EL images taken at an injection current of 10 mA revealed that the current spreading of AgNW-LEDs (Fig. 7a) was more uniform than that of the reference LEDs (Fig. 7b), even though the surrounding n-electrodes were employed. The magnified EL images from the areas indicated by targets in Fig. 7 showed brighter and more uniform light emission (along with squashed NWs) in the AgNW-LEDs than in the reference LEDs. At a similar position with identical magnification, high-resolution CSEM showed non-uniform light emission with a few very bright localized spots in the AgNW-LEDs, supporting our arguments. CSEM showed uniform emission in the reference LEDs, in which the broad emission inhomogeneity was associated with the non-uniform current spreading. In Fig. 7a, more specifically, the localized brighter emission had shapes of short dots and lines, which predicted the physical shape of the AgNW/p-GaN contact. In the CSEM images, the light emission near the bright localized spots was not much less than, or was even comparable to, the average brightness of the reference LEDs, indicating that the carrier transport around the AgNW contact was also active. However, a very dark area was also observed, which should be avoided to obtain better performance of AgNW-LEDs. Specifically, process optimization to achieve a more uniform contact between the NWs and the p-layer, which may be obtained using an intermediate contact layer, will be challenging.

Methods
Formation of AgNW films and reference electrodes. To form AgNW films, an as-received dispersion containing AgNWs (CambriosClearOhm Ink) was sonicated for 300 s, shaken well, and then spin coated on a pre-cleaned sapphire substrate for 40 s at 800 rpm [45][46][47] . The AgNW films were evaluated before and after a rapid thermal annealing at 100 and 150 °C for 1 min in N 2 ambient. For a comparative study as a reference electrode, Ni/Au bilayers with thicknesses of 3.5/3.5, 5/5, and 7/7 nm were deposited on the sapphire substrate using an e-beam evaporator, followed by rapid-thermal annealing at 550 °C for 1 min in an O 2 ambient 77,78 . Additionally, ITO films were deposited on the sapphire substrate using an e-beam evaporator or a radio-frequency (rf) magnetron sputtering system with an rf power of  100 W, deposition time of 6 min, temperature of 500 ºC, Ar:O2 gas ratio of 10:0, and working pressure of 10 mTorr, followed by rapid-thermal annealing at 550 °C for 1 min in oxygen ambient. The spin-coated AgNW films were analysed by SEM, revealing that their overall thickness was approximately 70 nm. The optical transmittance and Rsh of the AgNW films, Ni/Au, and ITO layers were measured using a UV/ VIS spectrometer (V-670EX) and a four-point probe system (CMT-SR1000N).

Evaluation of contact properties.
To investigate the contact properties, TLM patterns specifically designed to measure the AgNW contact on p-GaN by probing adjacent SiO 2 /Pt pads were fabricated, as shown in Fig. 2a. For example, 20 nm-thick SiO 2 films were deposited on the top p-layer of an LED wafer by an e-beam evaporator; subsequently, the SiO 2 layer was selectively wet etched (using buffered oxide etchants) to expose the p-layer surface for AgNW contact. Then, a 10 nm-thick Pt pad was selectively deposited on the edge of the exposed p-layers, followed by selective coating of the AgNWs onto the exposed p-layer and the Pt pads simultaneously. The photolithographic lift-off technique was used for the selective deposition of the Pt pad and AgNW contact. For a comparative study, an oxidized Ni/Au electrode was also evaluated using the standard TLM patterns. For the AgNW electrodes, the TLM pattern included 150 × 200 μ m contact pads and spacings of 10, 20, 40, and 60 μ m; for the Ni/Au electrodes, the TLM pattern included 100 × 200 μ m contact pads and spacings of 5, 10, 15, 20, 25, and 30 μ m. The reason for the use of the larger pattern size for AgNWs was the poorer patterning accuracy of AgNWs than of Ni/Au. The electrical properties of the contacts were evaluated using a parameter analyser (HP4156A).

Demonstration of AgNW-LEDs.
To fabricate LEDs with AgNW TCEs, the same method that was used to form the AgNW TLM pattern was utilized ( Supplementary Fig. 5). Specifically, the rectangular mesa was defined by dry etching to a thickness of ~1.0 μ m to expose the n-layer using an inductively-coupled plasma reactive ion etching system on which a Ti/Al/Ni/Au (30/70/30/70 nm) layer was deposited as an n-electrode by an e-beam evaporator. Rapid thermal annealing was performed at 550°C for 1 min in ambient N 2 to form an n-type ohmic contact. To form AgNW TCEs on the p-layer, a Ti/Au (20 nm/10 nm) probing pad was formed on the mesa, followed by selective AgNW coating on the exposed p-layer and Ti/Au probing pads by means of a lift-off technique. The spin-coating process was performed in the last process step to minimize the possible contamination of AgNWs by additional photolithographic processing. To implement our study, commercially available LEDs wafers were used; these were grown on c-plane sapphire substrates by a metalorganic chemical vapour deposition. The structure of the LEDs comprised 2.0 μ m of undoped GaN, 3.5 μ m of n-GaN, 5-period GaN/InGaN multiple quantum well (MQW) active regions with 450 nm-emission, a 0.024 μ m p-AlGaN electron blocking layer, and a 0.14 μ m p-GaN layer. The fabricated LEDs were evaluated using a parameter analyser connected to a photodiode (UV-818) and an optical spectrometer (Ocean Optics-USB2000+ ). To investigate the spatially-resolved electroluminescence (EL) images, CSEM 75,76 was employed.