Characterization of nanowire light-emitting diodes with InP/InAsP heterostructures emitting in telecom band

First, we present the results for low-temperature PL of a single NW. The PL results from the samples formed with and without annealing of the InP NW are compared and summarized in Fig. 2. In both NWs, we observed emission from InP at approximately 900 nm, and broad emission at approximately 1300 or 1200 nm originating from InAsP. In addition, we observed multiple sharp peaks at approximately 1389 nm (0.893 eV) in the sample without annealing, and at approximately 1515 nm (0.817 eV) in the sample formed with annealing. The linewidth of the peaks was 0.91 meV and 0.71 meV, respectively, for samples formed without and with annealing of InP NW. We attributed these sharp peaks to originating from QDs formed in the InP NWs. To confirm this hypothesis, we investigated the excitation intensity dependence of these emissions. The results for the sample formed with annealing are shown in Fig. 2(b). The peaks labeled as X and XX were resolved, and it was confirmed that peak X increased linearly with the excitation intensity, and XX increased proportionally to the square of the excitation intensity, as shown in Fig. 2(c). Similar results were obtained for the peak at approximately 1389 nm for the sample formed without annealing (data not shown). Thus, we conclude that the present results indicate the formation of an InAsP QD in the NWs, and X and XX correspond to the emission of excitons and biexcitons in the QDs, respectively.

Zoom In Zoom Out Reset image size

In the case of the sample formed without annealing, we believe that the QDs were formed by the fluctuation of the well width of the quantum well (QW) formed on top of the InP NW because the lateral size of the NW was not sufficiently narrow to provide lateral quantum confinement. The broad emission at higher energies is due to the QWs formed on the top of the NW. For a sample formed with annealing, a QD is likely to form at the tip of the InP NW if the tip of the InP NW is sufficiently narrow (∼20 nm), as shown in Fig. 1(b). However, because the annealed InP NWs were tapered, the growth of InAsP at the sidewall was also highly likely, which led to the formation of a QW at the sidewall of the InP NWs. In addition, QDs can be formed at the sidewall owing to the thickness fluctuation or alloy composition fluctuation of the InAsP on the NW sidewall. In other words, discrete localized states resembling QDs are formed in the sidewall QW. These parasitic QW and QDs are thought to be the origin of the emission at 1200 nm and relatively narrow peaks between 1200 and 1500 nm, respectively.

Figure 3 summarizes the results of the room-temperature characterization of the NW-LEDs formed with and without annealing the InP NWs. Considering the area and the pitch of the NW array (a = 3 μm), we estimated that approximately 300 NW were connected in parallel within the LED pattern (50 × 50 μm²) of the NW array. Although the leakage current under reverse bias was much larger than that expected in the ideal pn-junction diode, rectifying current–voltage (I–V) characteristics were confirmed, as shown in Fig. 3(a). We believe that the reverse leakage current was mainly due to the leakage current through the top and bottom electrodes, suggesting insufficient isolation between them. In fact, our recent reports on InP NW-LEDs ^35,36) show that the amount of reverse leakage varies from sample to sample. The NW-LED fabrication process must be optimized to reduce the reverse leakage current. The analysis of the forward bias characteristics gave ideality factors n of 5.4 and 8.2, and series resistances R_s of 0.133 and 0.065 Ω cm² for the samples with and without annealing, respectively. Large ideality factors indicate that the current through the pn-junction is dominated by the tunneling, as in the case of InP homojunction NW LEDs, ³⁴⁾ as discussed in the next section. In the EL spectra of Figs. 3(b) and 3(c), we confirmed a broadband emission including 1.0–1.6 μm (or 1.5 μm in samples with in situ annealing). Emissions longer than 1.6 μm were not confirmed owing to the cutoff of the detector. This emission wavelength range is on the longer wavelength side compared to the InAsP/InP NW-LEDs in our previous study, with the InAsP layer formed under p_AsH3 = 4%. ³⁰⁾ Thus, we conclude that emissions covering telecom bands using InP-based heterostructured NWs can be achieved by introducing an InAsP layer with a larger As alloy composition. Figure 3(d) shows the current density J versus integrated light intensity L (J–L) characteristics of the two samples, and both show that the integrated intensity increases nearly linearly with the current density. In this figure, the ratio η of L and the injection current I, which is a measure of the external quantum efficiency (EQE), ³⁴⁾ is also shown as a function J and is almost constant for both samples. Therefore, we conclude that efficient carrier injection into InAsP was achieved in the present heterostructured NW-LEDs at RT, and no significant carrier overflow was confirmed at the current injection level.

Zoom In Zoom Out Reset image size

The spectra and their peaks were located at lower energies for samples formed with in situ annealing of InP compared to the sample formed without annealing. The EL spectra exhibited a blueshift when the injection current was increased in both samples. This is thought to be due to band filling. The blueshift due to band filling was larger in the samples formed with in situ annealing. These results suggest that the energy distribution of the density of states (DoS) in the InAsP layer is different between the two samples and that samples with annealing have a broader energy distribution in DoS. As described in the previous section, parasitic InAsP QWs are expected to be formed at the sidewalls of the NWs in the sample formed with annealing and are the origin of the EL emission in the telecom band.

Finally, we attempted the low-temperature characterization of the NW-LED. A heterostructured NW-LED using an NW array formed with in situ annealing of InP was used for this investigation. The pitch a of the array was 1 μm and approximately 2800 NWs were measured in parallel. The characterization results are shown in Fig. 4. Figure 4(a) shows optical microscopy image of the NW-LED and its emission images at RT and low temperature (LT, ∼9 K) taken with a vidicon camera at a forward bias voltage of 2 V. Uniform emissions were confirmed in this sample at both RT and LT. Figure 4(b) compares the J–V characteristics of an NW-LED at RT and LT on a semi-logarithmic scale. The ideality factor n and series resistance were calculated to be 3.88 and 0.024 Ω cm² at RT. The current was more than two orders of magnitude smaller at LT than at RT; however, the slopes in the forward-bias regions were almost the same at both temperatures. This implies that the current is dominated by tunneling. ³⁷⁾ Figure 4(c) shows the LT EL spectra at different current densities together with the PL spectra. We obtained the emissions including the telecom band at LT for both EL and PL. It should be noted, however, that EL emissions longer than 1.4 μm were almost completely suppressed, whereas PL included a wavelength range from 1.4 to 1.6 μm. It is also clear that the emission exhibits a large blueshift with an increase in the injection current density. In Fig. 4(d), we compare the EL spectra at RT and LT at similar current injection levels. We can confirm the blueshift due to temperature, but the amount of blueshift (∼170 meV) is much larger than that expected from the temperature dependence of the band gap energy of InAsP (∼70 meV). The amount of blueshift at LT was much larger than that observed at RT. In Fig. 4(e), the integrated EL intensities L and η are plotted as a function of the current density J at RT and LT. The efficiency η is approximately 18 times larger at LT than at RT at the same injection level. This enhancement is due to the reduction in nonradiative recombination, but the value is much smaller than the ratio of the PL efficiency of the InAsP-related emission (ratio of the integrated PL intensity of the InAsP-related peak) between LT and RT (approximately 250). At RT, L is nearly proportional to J; thus η is nearly independent of the injection current, as shown in Fig. 2(e). However, at LT, L depends on J as L $\propto$ J^1.07, resulting in a monotonous increase in η with J. The reason for the slight superlinear dependence of L on J is not clear at present.

Zoom In Zoom Out Reset image size

Although emission above 1.4 μm was strongly suppressed in the LT EL, sharp lines attributable to emission from the QDs or discrete localized states were observed at LT at approximately 1.25 μm. The results are shown in Fig. 5. In this measurement, a spatial filter was inserted between the microscope objectives and spectrometer to increase the spatial resolution and limit the number of NWs to be detected in the EL spectrum. We can see narrow peaks at 1248.0 nm (unassigned), 1250. 4 nm (labeled as X), and 1252.0 nm (XX). The linewidth of the sharp peak labeled X was 0.36 meV at a very low injection level. The integrated intensity of peak X increased linearly with injection current I when the broad background emission was subtracted. Peak XX appeared at the higher-energy side of peak X and developed superlinearly with I. Note that the ratio of the peak heights of X and XX was different for I = 108 and 189 μA. The integrated intensity of the XX peak was proportional to I². These EL results are analogous to the exciton and biexciton emissions in QD observed in PL measurements (Fig. 2) and in LED with self-assembled QD. ¹²⁾ However, care must be taken that we measured multiple NWs with pn-junction in parallel and, because of possible nonuniformities between NWs, the current flowing in one NW may not be doubled when the total current is doubled. Nevertheless, we believe that the peaks X and XX originate from excitons and biexcitons in a QD, respectively, obtained by current injection. To unambiguously determine whether the intensity is proportional to I or I² for excitons and biexcitons, the measurement of current through single NWs is required.

Zoom In Zoom Out Reset image size

In the NW-LED investigated in this study, EL originating from the InAsP layer was obtained in the wavelength range of 1.2 ∼ 1.6 μm at RT. If we consider the difference of the band gap energy between RT and 0 K for InAs or InP is in the range of 63–70 meV, ³⁸⁾ the EL from the InAsP layer shows blueshift and is expected to be obtained in the wavelength range from 1.12 to 1.47 μm at LT. The low-temperature PL results in Fig. 3(c) also suggest the existence of an InAsP layer having band gap energy corresponding to 1.2 ∼ 1.6 μm wavelength range. However, the blueshift in the EL spectra was much larger when the temperature was varied. Furthermore, the amount of blueshift induced by current injection was larger at LT than at RT. As a result, emission in the wavelength range from 1.4 to 1.6 μm was strongly suppressed in the LT EL spectra. This resulted in the failure to observe emissions with narrow linewidth attributable to QDs or localized states in EL measurements at wavelengths longer than 1.4 μm.

The reason for the unexpectedly large blueshift caused by temperature and current injection is not clear at present, but we think it is related to the mechanism of carrier injection into the active InAsP layer. If the pn-junction is ideal, carrier injection from InP to InAsP expected to occur by diffusion and energy relaxation; however, carrier injection via tunneling is much more likely because the current–voltage characteristics in Fig. 4(b) indicate that the current through the pn-junction is dominated by tunneling. In our previous study, we observed radiative tunneling in homojunction InP NW-LEDs, ³⁴⁾ where the EL peak was located at qV and its height was proportional to exp(qV/E_T). Here, q is the elementary charge, V is the forward bias voltage applied across the pn-junction, and E_T is the characteristic energy of tunneling. ^34,39,40) Our present results for the heterostructured NW-LEDs did not follow the characteristic features of radiative tunneling, but we speculate that the interplay of the two types of carrier injection may influence the shape of the EL spectra. It is also possible that the layer structure of the NW-LEDs results in an inefficient carrier injection. As shown in Fig. 1(e), the present NW-LED structure is a core–shell pn-junction and heterostructure, and electrodes are formed only in the upper part of the n-InP. Based on the SEM image shown in Fig. 1, the thicknesses of the InAsP and InP shell layers were estimated to be approximately 50 nm. Assuming a donor concentration of 10¹⁸ cm⁻³ and the surface state distribution modeled in Ref. 41, the width of the depletion layer from the sidewall is estimated to be approximately 20 nm. Although this implies that the shell layers are only partially depleted, the small effective thickness of the InP shell layer and its finite resistivity significantly limit the current path. If the current injected from the top n-InP layer does not spread sufficiently to the sidewall of the n-InP NWs, carrier injection into the InAsP layer at the sidewalls does not occur. In this case, an improvement in the fabrication process is required to form ITO electrodes on the sidewalls of the NWs by exposing the n-InP sidewalls. It is also necessary to form the desired QDs only at the tip of the InP NWs. Nevertheless, it is necessary to clarify the mechanism and pathway of carrier injection in heterostructure NW-LEDs in future studies.