Letter pubs.acs.org/NanoLett
Mixed Polarity in Polarization-Induced p−n Junction Nanowire LightEmitting Diodes Santino D. Carnevale,† Thomas F. Kent,† Patrick J. Phillips,∥ A. T. M. G. Sarwar,‡ Camelia Selcu,§ Robert F. Klie,∥ and Roberto C. Myers*,†,‡,§ †
Department of Materials Science and Engineering, ‡Department of Electrical and Computer Engineering, and §Department of Physics, Ohio State University, Columbus, Ohio 43210, United States ∥ Department of Physics, University of Illinois at Chicago, Chicago, Illinois 60607, United States S Supporting Information *
ABSTRACT: Polarization-induced nanowire light emitting diodes (PINLEDs) are fabricated by grading the Al composition along the c-direction of AlGaN nanowires grown on Si substrates by plasma-assisted molecular beam epitaxy (PAMBE). Polarization-induced charge develops with a sign that depends on the direction of the Al composition gradient with respect to the [0001] direction. By grading from GaN to AlN then back to GaN, a polarization-induced p−n junction is formed. The orientation of the p-type and n-type sections depends on the material polarity of the nanowire (i.e., Ga-face or N-face). Ga-face material results in an n-type base and a p-type top, while N-face results in the opposite. The present work examines the polarity of catalyst-free nanowires using multiple methods: scanning transmission electron microscopy (STEM), selective etching, conductive atomic force microscopy (C-AFM), and electroluminescence (EL) spectroscopy. Selective etching and STEM measurements taken in annular bright field (ABF) mode demonstrate that the preferred orientation for catalyst-free nanowires grown by PAMBE is N-face, with roughly 10% showing Ga-face orientation. C-AFM and EL spectroscopy allow electrical and optical differentiation of the material polarity in PINLEDs since the forward bias direction depends on the p− n junction orientation and therefore on nanowire polarity. Specifically, C-AFM reveals that the direction of forward bias for individual nanowire LEDs changes with the polarity, as expected, due to reversal of the sign of the polarization-induced charge. Electroluminescence measurements of mixed polarity PINLEDs wired in parallel show ambipolar emission due to the mixture of p−n and n−p oriented PINLEDs. These results show that, if catalyst-free III-nitride nanowires are to be used to form polarization-doped heterostructures, then it is imperative to understand their mixed polarity and to design devices using these nanowires accordingly. KEYWORDS: Nanostructures, nitrides, molecular beam epitaxy, light-emitting diodes, polarization, conductive atomic force microscopy
P
accommodation in nanowires, when compared to planar material, can largely avoid the formation of strain-related defects in a graded structure. Recent work by the authors demonstrates that catalyst-free III-nitride nanowires grown by plasma-assisted molecular beam epitaxy (PAMBE) can be used to form polarization-induced nanowire light-emitting diodes (PINLEDs) in compositionally graded AlxGa1−xN heterostructures that are free of extended defects.3 If one considers this result along with the fact that catalyst-free III-nitride nanowires are known to have an exceptionally high material quality,5 it is clear that catalyst-free nanowires offer the best solution to fully utilize polarization doping without sacrificing material quality.
olarization-induced doping in semiconductors addresses some of the key shortcomings of impurity-doping, namely, large activation energies in wide-band gap materials and the effects of random dopant fluctuations. Previous work with compositionally graded III-nitride heterostructures demonstrates that it is possible to use polarization-induced electric fields to either activate dopants in wide-band gap materials1 or form n- and p-type materials without the use of dopants at all.2,3 The former may solve the problem of large activation energies, and the latter may solve the problem of random dopant fluctuations. Unfortunately, compositionally graded heterostructures of lattice mismatched materials are needed to achieve polarizationinduced doping, and the strain induced by this grading will have a negative impact on material quality and device performance. For example, recent reports show a 2 orders of magnitude increase in dislocation density when GaN is graded to only Al0.3Ga0.7N in thin films.4 Thankfully, enhanced strain © XXXX American Chemical Society
Received: January 16, 2013 Revised: June 11, 2013
A
dx.doi.org/10.1021/nl400200g | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
and Mg in the top, which, as will be discussed in greater detail later, should reinforce the polarization-induced charge in Gaface nanowires and detract from the polarization-induced charge in N-face nanowires. A summary of the samples measured is provided in Table 1.
However, one drawback to using catalyst-free nanowires is controlling material polarity. In polarization-doped material the sign of the charge induced by the grading is determined by the orientation of the crystal structure (i.e., its polarity) relative to the direction of compositional grade. For example, in a layer graded from GaN to AlxGa1−xN, the resulting material is n-type if the material is grown along the wurtzite [0001] direction (Ga- or metal-face) and it is p-type if the material is grown along the [000−1] direction (N-face). In planar structures, both Ga-face and N-face substrates are commercially available, and since an epi-layer will take on the same polarity as its substrate, it is not difficult to grow layers with the desired polarity. Unfortunately the polarity of catalyst-free nanowires is more complicated. While some groups have shown catalyst-free III-nitride nanowires with a Ga-face polarity,6−8 recent studies demonstrate N-face to be more common.9−14 Most notably for the present work, there is evidence showing that both Ga-face and N-face nanowires grow in a single sample. Alloing et al. characterized the polarity of catalyst-free GaN nanowires grown by ammonia MBE using convergent beam electron diffraction (CBED) and found that 90% of the nanowires in a sample were Ga-face, the others being N-face.7 Cherns et al. grew catalystfree GaN nanowires by PAMBE on a thin layer of AlN deposited on a sapphire substrate. They found using CBED measurements that the resulting nanowires were Ga-face while a thin layer of highly defective GaN that grows between the nanowires is N-face.8 The effects of polarity in III-nitride materials are well-documented and quite varied, including effects on the incorporation of impurities during growth,15 indium adatom incorporation in InGaN films,16 transport properties in GaN/AlN nanowire devices,9 and electron overflow in green LEDs.17 While these aspects are certainly important, polarity becomes much more important in polarization-doped heterostructures because it determines the sign of polarization-induced charge. The current work provides evidence to support the mixed polarity seen in previous works, with nanowires predominately growing N-face and a small percentage of them Ga-face. Scanning transmission electron microscopy (STEM) and selective etch tests are used to characterize nanowire polarity. In addition to these previously reported techniques, two new techniques are used to determine polarity. First, conductive atomic force microscopy (C-AFM) is used to electronically determine the polarity of polarization-induced nanowire p−n junctions. Second, electroluminescence (EL) measurements of devices fabricated from ensembles of PINLEDs allows for an optical measurement of the mixed polarity of nanowires. All samples in this study are grown by PAMBE on n-Si(111) substrates. The growth conditions and two-step growth procedure used in the current study were previously reported.3,18 After the brief nucleation of plain GaN nanowires, the composition is linearly graded from GaN to AlN, then back to GaN, forming a p−n junction with polarization doping. In previous work, this grading was achieved by logarithmically changing the Ga and Al effusion cell temperatures to achieve a linear change in flux.3 In subsequent work this method of grading has been replaced with a shutter pulsing technique,19 which has led to brighter EL intensity. While all of the samples in this study have the same change in composition (graded from GaN to AlN and back to GaN), samples are either intentionally doped with Si and Mg, or not intentionally doped (or unintentionally doped, UID) with impurities. All intentionally doped samples are doped with Si in the nanowire base,
Table 1. Description of Samples Measured in This Study sample name
dopant in nanowire base
dopant in nanowire top
composition of quantum well at center of nanowire
A B
Si none
Mg none
Al0.15Ga0.85N GaN
As previously mentioned, nanowires grown by a catalyst free growth mode can have a mixed polarity. For UID nanowires, this results in the situation depicted in Figure 1A. With Ga-
Figure 1. Impact of polarity on polarization-induced p−n junctions formed using graded AlGaN nanowires. (a) Schematic of an ensemble of polarization-induced nanowire pn junctions with different polarities. (b) Energy band diagram of a polarization-induced p−n junction formed by grading a Ga-face nanowire from GaN to AlN then back to AlN. (c) Same as b, except for N-face polarity.
polar nanowires, the bottom graded section (GaN to AlN) will form n-type material, while the top graded section (AlN to GaN) will form p-type material. In N-polar nanowires the same changes in composition lead to p-type on the bottom and ntype material on top. Therefore, the as-grown sample of UID nanowires will consist of a collection of individual nanowire p− n junctions, some of which with n-type down (i.e., the Ga-face nanowires) and others with p-type down (N-face nanowires). The number of nanowires in each orientation simply corresponds to the ratio of N-face to Ga-face nanowires. Simulated energy band diagrams for the different p−n junction orientations in UID nanowires (Figure 1B and C) are calculated by solving the Poisson equation in one-dimension self-consistently taking into account the band gap variation and spontaneous polarization charge, but assuming that the B
dx.doi.org/10.1021/nl400200g | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
nanowires are fully relaxed.20 If strain-induced piezoelectric polarization is taken into account, the band diagrams are only very slightly modified due to surface mediated strain relaxation in the graded nanowires.3 The major difference between the two orientations comes at the base of the nanowires. In Ga-face (Figure 1B) the n-side is down, providing an n-GaN/n-Si junction. Assuming that the base of the nanowire is contacted through one electrode placed directly on the Si substrate and a second electrode placed on the nanowire tops, the overall device is n−n−p. This, along with a small barrier between the conduction bands at the n-GaN/n-Si interface (assuming Anderson’s Law and using published values for the electron affinities of GaN and Si21) allows for electrons to move easily into the nanowire base when the nanowire p−n junction is forward biased. The N-face situation (Figure 1C) is quite different. Because the p-type side of the junction is down in these structures, the interface at the base of the nanowire will be between p-GaN and n-Si. Assuming the electrodes are placed in the same position as described for the Ga-face nanowires, the overall device is now n−p−n or two back-to-back p−n diodes. Therefore, when the nanowire p−n junction is forward biased, the n-Si/p-GaN junction is reverse biased, restricting conduction through the substrate into the nanowires. STEM is used to directly image the c-axis bond orientation of individual nanowires (Figure 2), thereby revealing the nanowire
GaN nanowires.14 By imaging both Ga and N atoms, the stacking order in the lattice along the growth direction, and therefore the material’s polarity, is determined. Seven nanowires in total are measured in this way, and all of them demonstrate N-face polarity. While this method directly determines the polarity of the nanowires, it is quite timeconsuming to image a large number of nanowires. This makes STEM an impractical technique for determining the relative number of N-face and Ga-face nanowires in a given sample. Luckily, the selective etching of N-face and Ga-face nanowires can be used to more efficiently determine the number of Ga- and N-face nanowires in a given sample. Researchers have previously used selective etches of N-face GaN to determine the polarity of GaN nanowires.12,13 However, in this previous work, nanowires were etched for a relatively short amount of time, somewhere between 5 and 10 minutes. In this amount of time, N-face nanowires show facets on their tops from the etching. In the present work, nanowires are etched for longer amounts of time so that only Ga-face nanowires remain on the surface after etching, and the exact number of Ga-face nanowires can be determined. Scanning electron microscopy (SEM) images taken in cross-section before and after etching clearly show that at long etch times only a small number of nanowires remain (Figure 3A and B).
Figure 2. Electron microscopy characterization of nanowire polarity. (a) HAADF or Z-contrast image of a representative, single nanowire heterostructure showing GaN as the lighter and AlN as the darker regions in the image. (b) Atomic-resolution ABF image. (c) Increased magnification of highlighted area in b showing N-face polarity. Figure 3. Selective etch of N-face nanowires. Nanowires (a) before and (b) after selective etch of an as-grown sample. (c) HAADF image of a single nanowire postetch. (d) Nanowire density as a function of etch time. (e) Atomic resolution ABF image of nanowire remaining after etching.
polarity. All STEM images are acquired using a probe-corrected JEOL JEM-ARM 200CF operated at 200 kV. To prepare the nanowires for STEM work, a carbon grid is rubbed onto an asgrown sample, and individual nanowires adhere to the carbon. A high angle annular dark field (HAADF) STEM image of a single nanowire is shown in Figure 2A, which provides Zcontrast, and hence differentiates between areas with high % of Ga (brighter) compared to areas with higher % of Al (darker). To determine polarity, nanowires are imaged in ABF mode so that both group-III and N atoms are apparent (Figure 2B). This method has been previously utilized to determine polarity in
By imaging samples in plan view (see Supporting Information, Figure 1) the density of nanowires remaining after etching is found. To determine how nanowire density changes as a function of etch time, separate pieces of the same sample are etched for a series of progressively longer times (Figure 3D). The results show an initial nanowire density of roughly 350 μm−2, and at long etch times the density saturates at 30 μm−2. C
dx.doi.org/10.1021/nl400200g | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
Figure 4. Electrically determining nanowire polarity with C-AFM. (a) C-AFM measurement of Ga-face nanowires using positive applied bias (inset: C-AFM measurement with a negative bias). (b) An isocurrent map formed from image in b (inset: a schematic of the measurement setup). (c) A schematic demonstrating the combinations of polarization and impurity doping in Ga-face nanowires. (d) Comparing possible explanations of conducting sites alongside the density of conducting areas determined by C-AFM results.
The nanowires remaining after a long etch (1.5 h) are imaged using STEM (Figure 3C and E). STEM images taken in HAADF mode (Figure 3D) show that the coaxial AlN shell that forms during deposition has been removed by the KOH. The tops also appear to have been etched by a small amount; otherwise, the nanowires remain largely intact, allowing the growth polarity to be determined using atomic resolution STEM images taken in ABF mode (Figure 3E). A total of six nanowires taken from a post-etch sample have been imaged in this way, and all are found to be Ga-face, definitively showing that N-face nanowires have been removed. Additional PINLED samples are prepared and etched (data not shown), revealing that though there is some variation, the majority of the nanowires are N-face, with between 3 and 10% of nanowires being Ga-face. C-AFM is used to further characterize the polarity of asgrown samples. In this measurement a bias is applied to the tip of the C-AFM probe, and the Si substrate is grounded. As the tip scans over the surface of the sample the current running through the tip is collected and used to form an image. Given the mixed polarity of nanowires on a sample, nanowires with opposite orientations (n-p or p-n) should pass current under opposite biasing conditions. The sample measured using this technique is polarization and impurity doped (Sample A). The doping is chosen to reinforce the polarization-induced charge in Ga-face nanowires (Figure 4C); thus Si (n-type) doping is included in the graded bases (i.e., from GaN to AlN) and Mg (p-type) doping is included in the top graded section (i.e., from AlN to GaN). Given the orientation of the p−n junctions in the Ga-face nanowires (p-top), a positive voltage applied to the C-AFM tip should lead to forward-bias condition, yielding current when electron−hole recombination occurs in the depletion region of the nanowires. The bias needed to turn on the individual nanowires will be a combination of the PINLED structure itself and the voltage dropped where the tip makes contact to the nanowire tops, which is likely a Schottky barrier. The presence of this Schottky barrier should increase the turn-on voltage of the nanowires but should not affect our overall purpose, which is to selectively image Ga-face nanowires. C-AFM images taken
with a positive bias applied to the tip are provided in Figure 4A. Note that the scale in Figure 4A does not represent the absolute range of currents measured by C-AFM since in some small areas of the sample the measured current is as high as 100s of pA. A narrower range of currents is selected for clarity. To more easily interpret the results from the C-AFM images, the current values from this image are binned to create isocurrent maps (Figure 4B). When a positive bias is applied from the tip and the substrate is grounded, only a fraction of the total nanowires are seen in the image. With a bias of +10 V applied from the tip to the substrate, each of these nanowires passes a small, positive current in the pA range. This is consistent with a small number of forward-biased, Ga-face nanowires. From the isocurrent map in Figure 4B, the density of nanowires that pass a current is 32 nanowires/μm2. When compared to the density of nanowires remaining after etching determined from previously discussed SEM measurements (30 ± 5 μm−2), these independent electrical (C-AFM) and structural (SEM) measurements are remarkably consistent, supporting the conclusion that Ga-face (p-top) PINLEDs are detected with C-AFM under positive tip bias. An alternative explanation is that the conducting areas found by the C-AFM measurement correspond to areas in which nanowires have coalesced. The coalescing of catalyst-free GaN nanowires grown by PAMBE were discussed in previous works;22−26 however, we are not aware of any measurements of their electrical properties. One might hypothesize that nanowire coalescence sites would act as a source of leakage in a device, leading to strong conduction features in the C-AFM scans that could be mistaken for forward-biased Ga-face PINLEDs. However, the data clearly show that the conducting areas have a similar shape to the tops of nanowires (Figure 4A). In contrast, if coalescence sites were responsible for conduction, then the conducting regions should resemble the boundaries between nanowires (i.e., thin lines of conduction). Furthermore, a large number of the conducting areas shown in the isocurrent map exhibit a core−shell structure, consistent with the center of the nanowires providing the highest conduction pathway. This not surprising since the sidewalls of the PINLEDs are passivated with an insulating AlN shell. If D
dx.doi.org/10.1021/nl400200g | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
Figure 5. Electroluminescence (EL) measurements. (a) Log scale current density−voltage characteristics of intentionally doped and dopant-free devices (inset: linear scale plot of same data). EL from (b) intentionally doped and (c) dopant-free PINLEDs. (d) Wavelength and (e) EL intensity as a function of current density for dopant-free devices.
coalescence sites were responsible for conduction, the interfaces between nanowires would be more conducting, and a core−shell shape in the data would not be expected. Finally, the density of coalescence sites is counted from plane view SEM images and found to be five times larger than the density of conducting areas found by C-AFM (Figure 4D), further indicating that coalescence sites do not show up in the C-AFM current maps. With a negative tip bias, Ga-face PINLEDs are reverse-biased and therefore will not pass any current until they breakdown at large voltages. Given the instrument limited range of voltages possible with C-AFM (±10 V), reaching breakdown (>30 V) is not possible. Meanwhile, impurity provided charge and polarization−induced charge have opposite signs in the Nface PINLEDs. That is, the sample is intentionally doped to reinforce the specific p−n orientation of Ga-face PINLEDs, which will have the opposite effect in any N-face nanowires. Therefore the two sources of charge compensate one another in these nanowires, making the devices more resistive and reducing any possible current in C-AFM. The results of C-AFM taken with a negative tip bias are provided in the inset of Figure 4A. The measured current is more than 2 orders of magnitude lower than what was found for a positive bias and is close to the noise limit of this measurement. The lack of any significant current is consistent with compensation doped N-face PINLEDs. Given that the current values are near the noise limit of the instrument, they should be treated with some skepticism. However, the shape of the few conducting areas suggests conduction through nanowires. These data are discussed in greater detail in the Supporting Information, where it is speculated that the conducting areas under a negative tip bias could either correspond to N-face PINLEDs in which the polarization-induced charge is greater than the impurity doping, giving a p-bottom structure, or nanowires in which Si doping dominates in the base, but polarization dominates in the tops, giving an all n-type N-face nanowire. It is possible to further characterize the effects of mixed polarity through EL measurements. Devices are processed using standard photolithography and an e-beam evaporator to deposit thin metal contacts directly onto the nanowire tops that connects an ensemble of nanowires together in parallel. A second contact is made directly to the Si substrate by first scratching off a small area of nanowires and then annealing a
small piece of In directly to the bare Si. EL spectra are measured using a thermoelectrically cooled UV−vis CCD coupled to a 0.5m spectrometer. Light from the sample is collected using a fused quartz 50 mm f/2 singlet lens and subsequently focused onto the entrance slit of the spectrometer. The results of LEDs fabricated from both intentionally doped (Sample A) and UID (Sample B) graded nanowire devices are shown in Figure 5B and C, respectively. Plots of current density as a function of voltage for both devices are included in Figure 5A to provide the reader an estimate of the voltages needed to achieve the specific current densities associated with each EL spectrum. EL was taken at a range of currents under both positive and negative bias. Other than the difference in impurity-doping and quantum well active region composition (Table 1), samples A and B have identical structures. EL measurements for intentionally doped nanowires (Sample A) are taken first (Figure 5B). In this device bright EL is only observed under a positive bias applied to the nanowire tops (red line in Figure 5B), corresponding to forward bias of the Ga-face PINLEDs with p-top and n-bottom (analogous to p−n junction shown in Figure 1B). The EL peak is at 352 nm, consistent with electron−hole recombination in an Al.15Ga.85N QW with shifted EL due to the quantum confined Stark effect from built-in polarization charge.3 To reiterate, the nanowires emitting light in this case are the minority Ga-face PINLEDs with p-top and n-bottom. As discussed above, the Ga-face PINLEDs make up about 10% of the total nanowires in any device as revealed by both SEM and C-AFM measurements. Following from the negative tip bias results in C-AFM (inset of Figure 4A), significant current flow in compensation doped Nface PINLEDs cannot be achieved except at very high negative bias applied to the nanowire tops. When very large negative bias (−30 V) is applied to the top, corresponding to forward bias of partially compensated N-face PINLEDs, the EL intensity is at least 5000 times dimer than the emission under positive bias at the same current, with a peak at 264 nm as well as a broader tail extended to lower energies. The peak at 264 nm is clearly distinct from the QW emission (352 nm) indicating that electron−hole recombination occurs outside of the QW under these conditions. The large difference in intensity for the same device at identical current values is consistent with a large increase in nonradiative electron−hole recombination for the E
dx.doi.org/10.1021/nl400200g | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
results3 by demonstrating the robustness of dopant-free p−n diodes formed purely by polarization doping in either polarity (Ga-face or N-face). Finally, mixed polarity has an effect on the performance of devices fabricated using an ensemble of nanowires, with device performance changing depending on which polarity of nanowires are biased. Given that the majority of nanowires are N-face, any device utilizing polarization in these nanowires should be designed specifically for N-face nanowires. Because the nanowires in the study were originally designed to function with Ga-face polarity, future work must address how they might be used with an Nface polarity. One complication with switching the design from using Ga-face to N-face nanowires is that the p-side of N-face nanowires is in contact with the substrate, forming an n−p−n structure if grown on n-Si substrates. There is no barrier to conduction when using Ga-face nanowires because the conduction band offset is quite small between n-Si and nGaN. One could instead grow N-face PINLEDS on p-Si substrates, but this will result in a large valence band offset, which will hinder conduction through the device. It should be noted that, to our knowledge, there has not been a comprehensive study of charge transport at GaN/Si interfaces with different combinations of doping (i.e., n-GaN/n-Si, pGaN/p-Si, n-GaN/p-Si, and p-GaN/n-Si) and material polarity (i.e., Ga-face or N-face nanowires). Such studies would be useful to determine the possibilities and limitations of GaN nanowires integrated directly onto Si substrates and should be the subject of further investigation. Assuming for the time being that conduction at any p-GaN/ Si interface will be difficult, there are several ways to make pdown N-face PINLEDs. First, a section of n-GaN with a tunnel junction could be inserted at the base of the nanowires. The nGaN would have good band alignment with n-Si, and the tunnel junction will act as a carrier type inverter between the nGaN and p-GaN. Tunnel junctions have been extensively studied in planar III-nitride devices,27−30 so there are clear paths to implementing such devices in nanowires. Second, instead of maintaining the Si substrate and changing the nanowire heterostructure, one could adopt a new substrate. Because catalyst-free nanowires grow on a wide variety of substrates, including amorphous substrates,31 there are many options for substrates that could offer good electrical contact to the p-GaN nanowire base. These two separate approaches will both be pursued in future work.
compensation doped N-face PINLEDs. When the N-face nanowires are forward biased (i.e., negative bias on top), the compensated dopants (Si+ or Mg−) may act as recombination centers leading to efficient carrier recombination outside of the QW. Because hole transit in AlGaN is expected to be the limiting factor, electron−hole recombination is likely occurring in the bottom graded AlGaN (p-type) section containing Si+, implying electron overshoot of the QW. The 264 nm peak could be from radiative recombination at a specific defect site or from the band gap emission in graded AlGaN. In dopant-free nanowires (Sample B), where neither the Gaface nor N-face PINLEDs are compensated, the mixed polarity of these nanowires leads to an unusual case in which an LED shows comparable EL regardless of the sign of the applied bias because individual nanowires with opposite orientations are wired in parallel. Under a positive bias a single peak is observed at 405 nm (Figure 5C), consistent with electron−hole recombination in the GaN QW at the center of the nanowire. The emission is again red-shifted from the GaN band edge (365 nm) due to built-in polarization-induced quantum confined Stark effect. When the applied bias polarity is reversed (negative to top), the majority N-face oriented nanowires (ntop/p-bottom, Figure 1C) are forward-biased, and a single EL peak at nearly the same wavelength (390 nm) is observed. The difference between these two emission energies is most likely due to different current densities, and therefore different screening of polarization charge, in the device active region. The large difference in EL energy and intensity between samples A and B under a negative applied bias (Figure 5B and C, blue lines) demonstrates that compensation doping completely quenches the QW EL in N-face nanowires due to a charged impurity enhanced recombination. For sample B, the EL is 60 times brighter when forward biasing the N-face PINLEDs (negative to top) compared to forward biasing the Ga-face PINLEDs (positive to top) in the very same device. In Figure 5C, twice the amount of current is used in positive versus negative bias conditions. Using a higher current for the positive bias condition is necessary given that at lower currents the EL is too weak to accurately collect, implying that the N-face PINLEDs in sample B are at least 2 orders of magnitude brighter at a given current than the Ga-face PINLEDs. This is most likely due to the larger number of Nface nanowires in the sample. Furthermore, wavelength and EL intensity as a function of current density for sample B are shown in Figure 5D and E. As expected, under both positive and negative bias the EL wavelength blue shifts toward the GaN band gap. Interestingly, EL under negative bias blue shifts more rapidly with current than does the EL under positive bias. Under positive bias, the majority N-face nanowires are reversed-biased and act as a large source of leakage for the Ga-face nanowires under forward-bias, such that only a fraction of the total current will pass through Ga-face PINLEDs, leading to less screening in the QWs. Under the opposite bias, reversebiased Ga-face PINLEDs are a relatively smaller source of leakage for forward biased N-face PINLEDs. This asymmetry in the leakage current can also explain the difference in EL intensity with bias polarity. Taking the C-AFM and EL measurements together, one can reach the following conclusions. Polarization-induced p−n junctions can be formed using graded AlxGa1−xN nanowires with either Ga-face or N-face polarity. Impurity doping can be used to either augment or compensate polarization-induced charge. These measurements also expand upon our previous
■
ASSOCIATED CONTENT
S Supporting Information *
SEM images of etch study, C-AFM images of impurity doped N-face PINLEDs, and results of the discussion of the C-AFM measurements. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS This work was supported by the Office of Naval Research (N00014-09-1-1153) and by the National Science Foundation CAREER award (DMR-1055164). S.D.C. acknowledges F
dx.doi.org/10.1021/nl400200g | Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
Letter
(26) Consonni, V.; Knelangen, M.; Jahn, U.; Trampert, A.; Geelhaar, L.; Riechert, H. Appl. Phys. Lett. 2009, 95, 241910. (27) Grundmann, M. J.; Mishra, U. K. Phys. Status Solidi C - Curr. Top. Solid State Phys. 2007, 4, 2830−2833. (28) Krishnamoorthy, S.; Nath, D. N.; Akyol, F.; Park, P. S.; Esposto, M.; Rajan, S. Appl. Phys. Lett. 2010, 97, 203502. (29) Krishnamoorthy, S.; Park, P. S.; Rajan, S. Appl. Phys. Lett. 2011, 23, 233504. (30) Krishnamoorthy, S.; Yang, J.; Park, P. S.; Myers, R. C.; Rajan, S. Efficient GaN tunnel junctions using embedded GdN nanoislands. arXiv 2012, arXiv:1206.3810. (31) Goodman, K.; Protasenko, V.; Verma, J.; Kosel, T.; Xing, G.; Jena, D. J. Cryst. Growth 2011, 334, 113−117.
support from the National Science Foundation Graduate Research Fellowship Program (2011101708). The acquisition of the JEOL JEM-ARM200CF at the University of Illinois at Chicago (UIC) is supported by an MRI-R2 grant from the National Science Foundation (DMR-0959470). Support from the UIC Research Resources Center is also acknowledged.
■
REFERENCES
(1) Simon, J.; Protasenko, V.; Lian, C.; Xing, H.; Jena, D. Science 2010, 327, 60−64. (2) Jena, D.; Heikman, S.; Green, D.; Buttari, D.; Coffie, R.; Xing, H.; Keller, S.; DenBaars, S.; Speck, J. S.; Mishra, U. K.; Smorchkova, I. Appl. Phys. Lett. 2002, 81, 4395−4397. (3) Carnevale, S. D.; Kent, T. F.; Phillips, P. J.; Mills, M. J.; Rajan, S.; Myers, R. C. Nano Lett. 2012, 12, 915−920. (4) Li, S.; Ware, M.; Wu, J.; Minor, P.; Wang, Z.; Wu, Z.; Jiang, Y.; Salamo, G. J. Appl. Phys. Lett. 2012, 101, 122103. (5) Bertness, K. A.; Sanford, N. A.; Davydov, A. V. IEEE J. Sel. Top. Quantum Electron. 2011, 17, 847−858. (6) Brubaker, M. D.; Levin, I.; Davydov, A. V.; Rourke, D. M.; Sanford, N. A.; Bright, V. M.; Bertness, K. A. J. Appl. Phys. 2011, 110, 053506. (7) Alloing, B.; Vezian, S.; Tottereau, O.; Vennegues, P.; Beraudo, E.; Zuniga-Perez, J. Appl. Phys. Lett. 2011, 98, 011914. (8) Cherns, D.; Meshi, L.; Griffiths, I.; Khongphetsak, S.; Novikov, S. V.; Farley, N.; Campion, R. P.; Foxon, C. T. Appl. Phys. Lett. 2008, 92, 121902. (9) den Hertog, M. I.; Gonzalez-Posada, F.; Songmuang, R.; Rouviere, J. L.; Fournier, T.; Fernandez, B.; Monroy, E. Nano Lett. 2012, 12, 5691−5696. (10) Fernandez-Garrido, S.; Kong, X.; Gotschke, T.; Calarco, R.; Geelhaar, L.; Trampert, A.; Brandt, O. Nano Lett. 2012, 12, 6119− 6125. (11) Kong, X.; Ristic, J.; Sanchez-Garcia, M. A.; Calleja, E.; Trampert, A. Nanotechnology 2011, 22, 415701. (12) Hestroffer, K.; Leclere, C.; Bougerol, C.; Renevier, H.; Daudin, B. Phys. Rev. B 2011, 84, 245302. (13) Largeau, L.; Galopin, E.; Gogneau, N.; Travers, L.; Glas, F.; Harmand, J. Cryst. Growth Des. 2012, 12, 2724. (14) de la Mata, M.; Magen, C.; Gazquez, J.; Utama, M. I. B.; Heiss, M.; Lopatin, S.; Furtmayr, F.; Fernandez-Rojas, C. J.; Peng, B.; Ramon Morante, J.; Rurali, R.; Eickhoff, M.; Fontcuberta i Morral, A.; Xiong, Q.; Arbiol, J. Nano Lett. 2012, 12, 2579−2586. (15) Chichibu, S.; Setoguchi, A.; Uedono, A.; Yoshimura, K.; Sumiya, M. Appl. Phys. Lett. 2001, 78, 28−30. (16) Nath, D. N.; Guer, E.; Ringel, S. A.; Rajan, S. J. Vac. Sci. Technol., B 2011, 29, 021206. (17) Akyol, F.; Nath, D. N.; Krishnamoorthy, S.; Park, P. S.; Rajan, S. Appl. Phys. Lett. 2012, 100, 111118. (18) Carnevale, S. D.; Yang, J.; Phillips, P. J.; Mills, M. J.; Myers, R. C. Nano Lett. 2011, 11, 866−871. (19) Carnevale, S. D.; Kent, T. F.; Phillips, P. J.; Sarwar, A. T. M. G.; Klie, R. F.; Rajan, S.; Myers, R. C. Proc. SPIE 2012, 84670L-1− 84670L-11. (20) Grundmann, M. J. BandEng Software; University of California: Santa Barbara, CA. (21) Adachi, S. Handbook on Physical Properties of Semiconductors; Kluwer Academic Publishers: Dordrecht, 2004; Vols. 1−3. (22) Yoshizawa, M.; Kikuchi, A.; Mori, M.; Fujita, N.; Kishino, K. Jpn. J. Appl. Phys., Part 2-Lett. 1997, 36, L459−L462. (23) Songmuang, R.; Landre, O.; Daudin, B. Appl. Phys. Lett. 2007, 91, 251902. (24) Calarco, R.; Meijers, R. J.; Debnath, R. K.; Stoica, T.; Sutter, E.; Luth, H. Nano Lett. 2007, 7, 2248−2251. (25) Ristic, J.; Calleja, E.; Fernandez-Garrido, S.; Cerutti, L.; Trampert, A.; Jahn, U.; Ploog, K. H. J. Cryst. Growth 2008, 310, 4035−4045. G
dx.doi.org/10.1021/nl400200g | Nano Lett. XXXX, XXX, XXX−XXX
