Using molecular beam epitaxy, self-assembled AlGaN nanowires are grown directly on Ta and Ti foils. Scanning electron microscopy shows that the nanowires are locally textured with the underlying metallic grains. Photoluminescence spectra of GaN nanowires grown on metal foils are comparable to GaN nanowires grown on single crystal Si wafers. Similarly, photoluminescence lifetimes do not vary significantly between these samples. Operational AlGaN light emitting diodes are grown directly on flexible Ta foil with an electroluminescence peak emission of ∼350 nm and a turn-on voltage of ∼5 V. These results pave the way for roll-to-roll manufacturing of solid state optoelectronics.
Nanowire based electronics and optoelectronics are attracting more attention due to their successful applications in transistors,1 resonant tunneling diodes,2 light emitting diodes (LEDs),3 lasers,4 and photodetectors.5 The formation of dislocations in conventional thin film devices due to lattice mismatch strain restricts the choice of substrate and heterointerface. On the other hand, the large strain accommodation capability of nanowires due to large surface to volume ratio not only permits large lattice mismatched heterostructures without dislocation formation6 but allows the formation of unconventional heterostructures which are otherwise impossible in conventional planar films.7
III-Nitride nanowires have been shown to be more useful in several fields, such as red and deep ultraviolet lasers,8,9 photo-electrochemical water splitting,10 and sensors,11 compared to their planar film counter parts. However, both conventional thin-film and nanowire devices are primarily grown on single crystalline substrates. Recently, there has been interest on the growth of GaN nanowires and related materials on low-cost, scalable metal substrates. The use of metallic substrates is also useful for efficient thermal management, as well as enhanced optical extraction for top emitting devices. Wölz et al.12 recently demonstrated the growth of GaN nanowires on Ti films deposited on sapphire. Later, the first operational LEDs on metal films were shown by Sarwar et al.,13 demonstrating LEDs emitting in the ultraviolet to green band of AlGaN LEDs grown on Mo films on Si wafers. Subsequently, Zhao et al.14 fabricated red nanowire LEDs on Ti coated bulk (0.5 mm) polycrystalline Mo substrates. This previous work indicates that, surprisingly, high quality GaN based nanowires can be reproducibly grown on a variety of metal substrates and can be electrically integrated for optoelectronic device applications. The previous work focused on the growth of nanowires on rigid substrates, whereas very large scale manufacturing, for example, roll to roll processing, often requires a flexible substrate, such as thin metal foils.
Here, we utilize molecular beam epitaxy (MBE) to demonstrate the growth of AlGaN nanowire LEDs on flexible Ta and Ti foils. The morphologies of nanowires grown on metal foils are similar to nanowires grown on single crystal Si wafers as revealed through scanning electron microscopy (SEM). Cryogenic temperature micro-photoluminescence (μ-PL) measurements show a dominant neutral donor-bound A exciton recombination in nanowires grown on Ta and Ti foils. Time-resolved μ-PL measurements show similar PL decay times for nanowires grown on metal foils compared to nanowires on Si wafers indicating comparable optical quality. Finally, a nanowire LED is grown directly on flexible Ta foil emitting in the ultraviolet band.
The Ta and Ti metal foil substrates are 100-μm thick with purities of 99.9% and 99.6%, respectively, and cut into 1 inch squares. The foils are cleaned with standard solvents before vacuum introduction, but no other surface preparation is performed. They are vacuum baked at 600 °C for 1 h before introduction to the growth reactor. Self-assembled catalyst-free GaN nanowires are grown on the foils using plasma-assisted MBE in a VEECO GEN 930 system equipped with a RIBER N2 plasma source. The Ga beam equivalent pressure is 6.20 × 10−8 Torr measured using a beam flux monitoring (BFM) ion gauge. A nitrogen flow rate of 7.5 sccm is used with a plasma power of 500 W, which gives a N-limited growth rate of ∼670 nm/h. A III-V ratio of 0.18 is used during growth. The substrate is rotated away from the plasma source when striking to avoid nitridation of the surface. The nanowires are grown employing a two-step growth method15 allowing for separate control over the density and height of the structures. For growth on Si, Ta, and Ti, GaN nanowires are first nucleated at 750 °C for 5 min then growth proceeds at 800 °C for 2 hours. Reflection high energy electron diffraction (RHEED) for the starting foils is difficult to obtain due to their unpolished, polycrystalline surface. However, once nanowires form, RHEED reveals a spotted ring pattern (inset Fig. 1(a)) that is similar to that of nanowires grown on single crystalline Si. This spotted ring configuration arises as a superposition of patterns from the single crystalline nature of the wires (spots) and the various orientations to the substrate (rings).
An image of GaN nanowires grown on a flexible Ti foil is shown in Figure 1(a). SEM measurements were carried out on a FEI/Philips Sirion for GaN nanowires grown on each substrate. Figure 1(b) shows the nanowires grown on a single crystalline Si substrate. The nanowires on Si grow along the c-axis approximately perpendicular to the substrate. SEM images of wires grown on multiple grains of the Ti foil are shown in Figure 1(c), and an image of the nanowires grown on Ta foil is shown in Figure 1(d). The nanowires on metal foils are uniformly tilted with respect to the surface normal within the individual grains of the foils, Figures 1(c.1)–1(c.3). The wires have similar heights and radii, and are evenly distributed throughout the region. However, the density and tilt direction vary between different grains which suggest an epitaxial relationship with the metals directly below the nanowire. However, the details of such a relationship are not yet completely understood and subject to future study. The Ta foil gives less variation between regions, and the individual regions are larger than those compared to the Ti foil (SEM images not shown).
Figure 2 shows normalized μ-PL spectra of the as-grown nanowires on Si (black), Ta (red), and Ti (blue) at 27.6 K on linear and semi-logarithmic scales. The nanowires were optically excited using a third harmonic (267 nm) of a mode-locked Ti:sapphire oscillator (Coherent Chameleon Ultra II) operating at 800 nm and 80 MHz. The samples were illuminated with an average power of 125 μW and focused on the sample surface through a 0.5 NA 36× reflective objective which results in a beam diameter of ∼10 μm. The emission from the samples was collected through a 300 nm long pass filter and passed to a 0.5 m spectrometer (Princeton Instruments SP2500i) equipped with a UV–VIS CCD (Princeton Instruments PIXIS100) and a 1200 g/mm diffraction grating. All the samples show dominant ∼358 nm (∼3.472 eV) neutral donor bound A exciton (D0, XA) recombination.12,16–18 A high energy shoulder is also observed at ∼357.5 nm (∼3.477 eV), attributed to free A exciton (XA) recombination.12,17 A relatively broad and weak peak at ∼362.7 nm (∼3.427 eV) is observed. It has been previously attributed to surface related excitons,19 or exciton bound to structural defects such as I1 stacking faults in GaN nanowires,16 which could originate from sidewall coalescence of the nanowires17,20 or from defects near the bottom surface of the nanowires.21 No long wavelength defect peaks (yellow luminescence) are observed in any of the samples.
Time-resolved PL was carried out using time-correlated single photon counting spectroscopy using a micro-channel plate (MCP) photomultiplier tube (PMT) detector coupled to a 0.15 m spectrometer. Figure 3 shows the time-resolved PL at the 358 nm (D0, XA) peak (data points) along with the instrument response function (IRF) (dashed line) to show the time resolution of our setup. The PL decay curves are well fitted to a biexponential model (solid lines) with a short lifetime and long lifetime component. For nanowires grown on Ta foil, both the short and long decay components are almost identical to those of nanowires grown on Si. On the other hand, the nanowires on Ti foil show a reduced short decay constant but enhanced long decay constant. Even with this variation, the relatively similar recombination characteristics for all the samples indicate that the nanowires grown on metal foils are of similar optical quality to those grown on single crystalline Si substrates.
Having established the high quality of GaN nanowires grown on metal foils, we grew an AlGaN LED on flexible Ta foil. The Ta foil was chosen due to the higher degree of uniformity compared to the Ti foil, as mentioned previously. The LED heterostructure follows a tunnel junction (TJ) integrated design that was previously demonstrated on Si.22 The n-GaN wires were nucleated at 750 °C, then a 100 nm n-GaN base was grown at 790 °C followed by a polarization engineered InGaN TJ. A p-type region was formed through composition grading from GaN to AlN over 100 nm, taking advantage of the polarization doping. Then, an active region consisting of 3× AlGaN quantum wells (QWs) with AlN barriers was grown at 840 °C before growth of the n-type layer achieved through grading back from AlN-GaN. Details of polarization induced doping in nanowires can be found in Refs. 23–25. The devices were fabricated using the same process as fabrication of devices on Si. First, a HCl etch was performed to remove any surface oxide present on the nanowires, followed by deposition of a 10/20 nm Ti/Au semi-transparent metal top contact. A bottom contact was formed by mechanical removal of the nanowires and a diffused In dot directly on Ta.
The current–voltage curve (Fig. 4 inset) shows good diode characteristics with a turn on voltage around 5 V, comparable to the same heterostructure grown on n-Si. The electroluminescence (EL) spectrum (Fig. 4) shows ultraviolet emission with a peak at ∼350 nm. The blue shift with increasing current injection in the EL is a result of the screening of the quantum confined Stark effect by the polarization field in the QWs.23 It is noted that the EL intensity is about 16× lower compared to similar devices on Si. A potential large loss mechanism is observed through the substantially higher reverse bias current, compared to Si devices. This is likely from the inhomogeneous distribution in nanowire tilt and density. Such variations could result in metal deposition on the sidewalls of the nanowires and substrate during fabrication of the top contact. Consequently, a leakage pathway is formed, reducing the number of active nanowires, thus decreasing the overall EL intensity. The nanowire ensemble uniformity will be addressed in future work by improving nanowire density control through exploration of growth conditions as well as through planarization of the processed devices.
In summary, self-assembled GaN nanowires were grown by MBE on flexible Ta and Ti foils with optical quality comparable to devices grown on Si. Using this method, nanowire LEDs were directly grown and electrically integrated on flexible Ta foil. In this study, only two types of metal foils were tested, but a wider variety of metals may be possible to utilize as long as they are compatible with the growth temperature. The realization of operational nanoLEDs grown directly on flexible metal foils provides a first step towards scalable roll-to-roll manufacturing of nanomaterial based solid-state optoelectronics.
This work was supported by the Army Research Office (W911NF-13-1-0329) and by the National Science Foundation CAREER Award No. DMR-1055164.