1. Introduction

III-nitride nanowires (NWs) grown on silicon and quantum-confined heterostructures exhibit excellent characteristics of a high radiative recombination rate [1, 2], large electron mobility [3], small Auger coefficient [2], and tunable energy bandgap from the near-infrared (InN, 0.65 eV) to the ultraviolet (AlN, 6 eV) [1, 4]. Such advantages enable the III-nitride NWs to be exploited in various applications such as light-emitting diodes (LEDs) [1, 5–7], photodetectors [8, 9] and polariton- and photon lasers [10–14]. However, an as-grown NW on silicon has an inherent silicon nitride (SiN_x) layer at the interface between the NW and the silicon substrate, which behaves as an insulating barrier, resulting in a high turn-on voltage in NW-based LEDs. In addition, a high growth temperature (~500–800 °C for GaN and InGaN [1, 6]) does not allow the direct growth of NWs on flexible plastic substrates for implementing foldable and wearable devices. Because an organic semiconductor as a current viable platform for such devices exhibits innate weaknesses (water vapor and oxygen induce rapid degradation and instability at high temperature) [15–17], the development of dislocation-free inorganic semiconductor-based flexible devices is indispensable to achieve long-term stability and higher device performance.

In this context, we have grown a single-crystalline InGaN/GaN disk-in-NWs on a SiO₂ sacrificial layer and successfully transferred the grown NWs onto a flexible polyimide sheet using a cold-welding process and an epitaxial lift-off technique. LEDs are fabricated with both transferred NWs on a flexible polyimide sheet and as-grown NWs on silicon. The transferred nanowire LED has a lower turn-on voltage compared to that of the conventional nanowire LED because of the absence of an insulating SiN_x layer. However, both LEDs exhibit similar emission wavelengths (610–616 nm and 647 nm for transferred NW LEDs and conventional NW LEDs, respectively) and output intensities.

2. Growth

The nanowires and disk heterostructures are grown by a plasma-assisted molecular beam epitaxy (MBE). It was believed that III-nitride NWs preferentially grew on (111)Si because of a similar atomic arrangement to the c-plane of the Wurtzite crystalline structure. However, it has been observed that a spontaneous growth of single-crystalline NWs is also possible on an amorphous SiO₂ layer [18, 19]. The deposition of dielectrics at a low temperature, e.g., via plasma-enhanced chemical vapor deposition (PECVD), may induce contamination in an MBE growth chamber by any possible desorption at a higher growth temperature. Hence, in this work, wet oxidation was performed on (001)Si at 900 °C to form the 200 nm thick SiO₂ layer on which the InGaN/GaN nanowire heterostructures were grown in the same manner as those grown on silicon. First, a deposition of a few monolayers of Ga at 800 °C results in the Ga droplets from which GaN NWs were grown at a rate of ~300 nm/hour under N-rich conditions. To obtain p- and n-type GaN, Mg and Si were doped, respectively. A detailed description of InGaN/GaN disk-in-NWs is provided in previous works [1, 2].

Figure 1 schematically shows the nanowire heterostructure grown on SiO₂, which consists of a 300 nm n-doped GaN, 8 layers of InGaN disks as an active region, and a 150 nm p-doped GaN. Each InGaN disk is 2 nm thick and is separated by a 12 nm thick GaN spacer. The identical nanowire heterostructures were also grown on (111)Si for comparison.

 

Fig. 1 Schematic illustration of InGaN/GaN nanowires. Schematic representation of InGaN disks in GaN nanowires grown on a 200 nm thick SiO₂ layer. The InGaN disks are 2 nm thick and are separated by a 12 nm thick GaN spacer.

Download Full Size | PDF

3. Characteristics of as-grown nanowires on SiO₂

The structural characteristics of the as-grown NWs on SiO₂ were investigated by scanning electron microscope (SEM) and high-resolution transmission electron microscope (HR-TEM) imaging. Figure 2(a) shows an oblique-view SEM image of the NWs with a high density of ~10¹⁰ cm⁻². The cross-sectional SEM image in Fig. 2(b) shows that the NWs are vertically grown on an amorphous SiO₂ layer. The nanowire has a length of ~700 nm and an average diameter of ~50 nm. Figure 2(c) presents an HR-TEM image of a randomly selected single NW, demonstrating that the nanowires are relatively free of extended defects. The selective area diffraction pattern in the inset reveals that the NWs grow in a Wurtzite crystalline structure. For the growth of NWs on silicon, it has been reported that an amorphous SiN_x layer is formed at the interface between a NW and silicon [3]. Whereas for NWs grown on SiO₂, it was observed that a single crystalline NW region immediately starts on an amorphous SiO₂ layer (Fig. 2(d)). The dark region at the interface in the figure is attributed to the Pt behind the NW, which was deposited while preparing the sample for TEM.

 

Fig. 2 (a) An oblique view SEM image of the InGaN/GaN nanowires with a high density of ~10¹⁰ cm⁻². (b) cross-sectional SEM image of vertically aligned InGaN/GaN nanowires grown on SiO₂/Si. (c) TEM image of a randomly selected InGaN/GaN nanowire. The inset shows the selective area diffraction pattern of the nanowire, indicating that the nanowires grow in the Wurtzite crystalline structure. (d) TEM image of the interface between a GaN nanowire and SiO₂. The dark region is due to the deposited Pt behind the nanowire.

Download Full Size | PDF

To identify the amorphous layers directly below the NWs, energy dispersive X-ray spectroscopy (EDS) was performed near the interface between a GaN NW and SiO₂. Figure 3(b) shows the EDS signal variations of Ga, O, Si, and N atoms along the red line in Fig. 3(a). Whereas Ga atoms are clearly observed in the GaN NW region, Si and O are detected in SiO₂ only. The N-atom level is low and unchanged in the entire scanned region, which is due to the insufficient sensitivity of nitrogen in EDS. Nonetheless, an intermediate region, where Ga, O, and Si are intermixed, is present rather than SiN_x.

 

Fig. 3 (a) TEM image of the nanowire grown on SiO₂. Energy dispersive X-ray spectroscopy was performed along the red line in the figure. (b) The energy dispersive X-ray spectroscopy signals for Ga, O, Si, and N along the red line in (a). The shaded region is the intermediate region consisting of Ga, O, and Si.

Download Full Size | PDF

The photoluminescence (PL) map was measured at room temperature using the Accent RPM2000 with a spatial resolution of 0.5 mm. Figure 4(a) shows the measured peak wavelengths of the as-grown disk-in-NWs on SiO₂, and the sample boundary is indicated by a solid line. A variation in the peak wavelengths from 580 nm at the center to 635 nm at the edge is observed. A similar trend was also observed in the PL spectra of NWs grown on Si. This finding can be attributed to the inhomogeneous indium incorporation in disk nanostructures resulting from the slight temperature difference on the sample during growth. In Figs. 4(b)-4(f), the PL spectra of selected points in Fig. 4(a) are presented. The sharp peak at λ = 532 nm commonly appearing in the spectra is the second harmonic of the pumping laser (λ = 266 nm). As the detection region moves toward the center, the peak wavelength is blue-shifted, and the linewidth is reduced. Simultaneously, the additional emission peak starts appearing at a longer wavelength (λ = 719 nm). The spectrum measured in the center clearly decomposed into two peaks, as observed in Fig. 4(f). It appears that multiple ensembles with different indium compositions are present, which will be further studied elsewhere.

 

Fig. 4 (a) Peak wavelength map of room-temperature photoluminescence measured on as-grown nanowires on SiO₂. The peak wavelength varies from 580 nm to 635 nm depending on the measured positions; (b), (c), (d), (e), and (f) the black solid lines represent the measured photoluminescence spectra of the indicated spots in (a). The blue dash-dot and red dash line are single Gaussian fits and cumulative Gaussian fits, respectively.

Download Full Size | PDF

4. Light emitting diode fabrication and characteristics

The sacrificial layer of SiO₂ on which the NWs are grown enables the NWs to be successfully transferred to a flexible substrate using an epitaxial lift-off (ELO) technique. The LEDs were also demonstrated on the transferred NWs, and the fabrication procedure is outlined as follows. First, NWs are passivated and planarized by a spin-coating of a polymer. Excessive polymer is etched by oxygen plasma until the top of the NW is exposed. Deposition of 5/20 nm Ni/Au as a p-type Ohmic contact is followed by annealing at ~500 °C for 1 min in nitrogen ambient. Commonly used flexible substrates, such as polyester (PET) and polyimide, are not compatible with high-temperature eutectic bonding because of their low glass temperatures (T_g = 70 °C and 254–326 °C for PET and polyimide, respectively [20, 21]). Hence, cold-welding is employed to bond the NWs to a polyimide sheet [22, 23]. Then, 50/100 nm Cr/Au are deposited on both NWs and a polyimide sheet, and 4 MPa pressure is then applied at 200 °C while the deposited Cr/Au surfaces are in contact. The bonded sample is immersed in an acetone-diluted hydrofluoric (HF) acid solution to release the NW from silicon by the selective etching of SiO₂. Hydrophilic acetone reduces the surface tension of the etchant, resulting in efficient lateral etching [24]. The inherent tensile strain (1.5–2 GPa) in the sputtered Cr layer provides the force to help the NW be released from the substrate. The released NWs on a polyimide sheet are carefully cleaned by solvents. The 5/5 nm Ti/Au and 200 nm indium tin oxide (ITO) are successively deposited with a shadow mask to define the n-contact. Finally, the sample is annealed at ~300 °C for 1 h in air. Conventional NW (CNW) LEDs as a control sample are also fabricated with NWs grown on silicon. The fabrication of CNW LEDs is described in our previous work [3].

Figures 5(a) shows the current-voltage characteristics of the transferred NW (TNW) LED and the CNW LED. Whereas the CNW LED exhibits a soft turn-on voltage of approximately 5 V, the current in the TNW LED more abruptly increases from a voltage of 4 V. The smaller turn-on voltage in the TNW LED results from the direct contact of metal on n-GaN without any interfacial layer (e.g., SiN_x) present in the CNW LEDs. The interfacial layer appears to be etched away in the acetone-diluted HF acid when the NWs are released. The inset of Fig. 5(a) shows a microscope image of the forward-biased TNW LED, where bright orange color emission was clearly observed. Electroluminescence (EL) spectra of the TNW LEDs and the CNW LED were measured at room temperature using the Ocean Optics USB2000. Figure 5(b) shows the normalized EL spectra of the CNW LED and two individual TNW LEDs with an injection current of 15 mA. All the output spectra exhibit broad Gaussian emission with linewidths of 116, 123, and 130 nm for TNW LED 1, TNW LED 2, and CNW LED, respectively. Because the LEDs are made on the boundary of the sample, the spectra exhibit almost a single emission peak. The wider emission linewidth compared to a typical emission (about 50-75 nm [25]) from a planar quantum well structure is contributed to additional inhomogeneity of disks such as variation of the nanowire diameters. The maximum intensities of emission were observed at wavelengths of 610, 616, and 647 nm for TNW LED 1, TNW LED 2, and CNW LED, respectively. The slight difference in the peak wavelengths is due to the position-dependent indium incorporation in the grown sample, as described earlier. It is worth noting that despite the flexible substrates, the TNW LEDs unfortunately could not withstand a bending force because of the rigid metal contacts. Hence, flexible contacts such as a silver nanowire contact and a metal grid contact will be necessary to achieve completely flexible LEDs.

 

Fig. 5 (a) Room-temperature current-voltage characteristics of the transferred nanowire LED and conventional nanowire LED. The transferred nanowire LED exhibits a smaller turn-on voltage of approximately 4V compared with that of the conventional nanowire LED. (b) Electroluminescence of the transferred nanowire LEDs and the conventional nanowire LED with an injection current of 15 mA at room temperature. The peak wavelengths are observed at 610, 616, and 647 nm for the TNW LED 1, TNW LED 2, and CNW LED, respectively.

Download Full Size | PDF

5. Conclusion

In conclusion, we have grown III-nitride NWs and disk heterostructures on an amorphous SiO₂ layer using PAMBE. Structural characterization reveals that InGaN/GaN NWs are vertically grown even on an amorphous oxide layer. Moreover, the NWs grown on SiO₂ exhibit comparable optical characteristics to those grown on Si. It is concluded that there is no evidence of more defects in NWs when they are grown on SiO₂. With these NWs, we have successfully demonstrated the transfer of the NWs to a flexible polyimide sheet by Au-welding and epitaxial lift-off processes. The LED fabricated with the transferred NWs has a smaller turn-on voltage of approximately 4 V because of the absence of any interfacial layer. The electroluminescence measurements of the TNW LEDs and the CNW LED reveal analogous single broad peaks with slightly different peak wavelengths due to the non-uniform indium incorporation in the sample. It is worth mentioning that NWs grown on SiO₂ and transferrable to foreign substrates is applicable to various LED designs. For example, highly reflective mirrors can be integrated at the back side to achieve a better extraction efficiency. Efficient thermal release is also attained when thermally conductive metals are used.