pubs.acs.org/NanoLett

Ultraviolet Photodetector Based on GaN/AlN Quantum Disks in a Single Nanowire L. Rigutti,*,† M. Tchernycheva,† A. De Luna Bugallo,† G. Jacopin,† F. H. Julien,† L. F. Zagonel,‡ K. March,‡ O. Stephan,‡ M. Kociak,‡ and R. Songmuang§ †

Institut d’Electronique Fondamentale, University of Paris Sud XI, UMR 8622 CNRS, 91405 Orsay, France, ‡ Laboratoire de Physique des Solides, University of Paris Sud XI, UMR 8502 CNRS, 91405 Orsay, France, and § CEA-CNRS group “Nanophysique et Semiconducteurs”, Institut Ne´el, 25 Rue des Martyrs 38042, Grenoble Cedex 9, France ABSTRACT We report the demonstration of single-nanowire photodetectors relying on carrier generation in GaN/AlN QDiscs. Two nanowire samples containing QDiscs of different thicknesses are analyzed and compared to a reference binary n-i-n GaN nanowire sample. The responsivity of a single wire QDisc detector is as high as 2 × 103 A/W at λ ) 300 nm at room temperature. We show that the insertion of an axial heterostructure drastically reduces the dark current with respect to the binary nanowires and enhances the photosensitivity factor (i.e., the ratio between the photocurrent and the dark current) up to 5 × 102 for an incoming light intensity of 5 mW/cm2. Photocurrent spectroscopy allows identifcation of the spectral contribution related to carriers generated within large QDiscs, which lies below the GaN band gap due to the quantum confined Stark effect. KEYWORDS Nanowires, quantum structures, photodetector, nitrides

T

he recent progress in the controlled synthesis of wideband-gap nanowires has enabled the development of nanoscale photonic devices in the visible-UV spectral range. Single-wire optically pumped lasers,1 nanolight emitting diodes covering the whole visible spectrum,2 and photovoltaic devices3 have been demonstrated using core-shell InGaN/GaN multiquantum well nanowires grown by catalystassisted MOCVD. UV photodetectors based on ZnO or GaN nanowires have been also demonstrated.4-7 The small nanowire size and the high photoconductive gain demonstrated in nanowire photodetectors4,6 are very promising for the fabrication of focal plane arrays with diffraction-limited spatial resolution and high responsivity. Numerous studies have been devoted to the photoconduction properties of GaN nanowires. It has been pointed out that the nanowire surface has a strong impact on the wire photoconduction.8 Due to the Fermi level pinning, the bands bend close to the lateral surface and create a region depleted of electrons, which changes the nanowire conduction and also induces a Franz-Keldysh effect responsible for sub-band-gap photocurrent.9 All these studies focus on binary GaN nanowire; no nanowire photodetectors based on QDisc III-N heterostructures have been reported so far. In the present work we report the demonstration of an ultraviolet photodetector based on a single GaN nanowire containing GaN/AlN QDiscs. The doped extremities of the nanowires provide an electrical access allowing probing of the carrier photogeneration in the QDisc region. Two heterostructured nanowire samples containing QDiscs of dif-

ferent thickness are analyzed and compared to a reference binary n-i-n GaN nanowire sample. The responsivity of a single wire QDisc detector has been measured to be as high as 2 × 103 A/W at λ ) 300 nm at room temperature. We show that the insertion of an axial heterostructure drastically reduces the dark current with respect to the binary nanowires and enhances the photosensitivity factor (i.e., the ratio between the photocurrent and the dark current) up to 5 × 102 at 5 mW/cm2 illumination power density. Photocurrent spectroscopy allows identifying the spectral contribution related to carriers generated within large QDiscs, which lies below the GaN band-gap due to quantum confined Stark effect (QCSE). The study of photocurrent at different temperatures shows that photogenerated carriers escape from QDiscs mainly by thermal activation. Catalyst-free GaN nanowires were grown by radio frequency plasma-assisted molecular beam epitaxy (PA-MBE) on Si(111) substrates under N-rich atmosphere at 790 °C, according to the technique described in ref 10. The nanowire length is approximately 1.2 µm and the diameter is 25-80 nm. For nanowire heterostructures, a stack of 20 AlN/GaN QDiscs was inserted in the middle of Si-doped GaN nanowires, as schematically shown in Figure 1a. Quantum disks have been formed by switching from Ga to Al flux without any growth interruption. A reference sample containing n-i-n GaN binary nanowires was grown using similar conditions. The length of the nominally undoped segment is 100 nm. The STEM analyses (Figure 1b,c) show that the thickness of AlN barriers is 2-3 nm for both samples and the QDisc thickness is 3-5 nm (1-3 nm), for sample 1 (sample 2). As seen from STEM images, the QDisc thickness increases progressively toward the nanowire top. An AlN shell of about

* Corresponding author, [email protected]. Received for review: 03/29/2010 Published on Web: 07/09/2010 © 2010 American Chemical Society

2939

DOI: 10.1021/nl1010977 | Nano Lett. 2010, 10, 2939–2943

FIGURE 2. Room-temperature PL spectra of the nanowire ensembles of the reference n-i-n sample (black solid line), of sample 1 (red dashed line) and of sample 2 (blue dash-dotted line).

tured samples 1 and 2. The excitation was performed at normal incidence to the substrate. The main peak at 3.41 eV observed in all samples corresponds to the GaN nearband edge (NBE) emission at 300 K. On the low-energy side, a broad peak is observed at 2.75-2.9 eV (3.19 eV) for sample 1 (sample 2). This low-energy peak is attributed to the QDisc emission. Its energy is shifted below the GaN band gap due to the QCSE, which has been observed in GaN/AlN heterostructures in nanowires.13 The QCSE stems from the internal electric field of piezoelectric and pyroelectric origin present in the QDiscs. In ref 13, the dependence of the emission energy of GaN/AlN QDiscs vs QDisc thickness has been studied, and the results show a good agreement with our observations. The QDisc luminescence peak is inhomogeneously broadened because of the size and strain dispersion. The upward shift of the PL peak energy is related to the QDisc size decreasing from sample 1 to sample 2. Thus, it excludes the attribution of this peak to either blue band14,15 or donor-acceptor pair transitions. Single nanowires were dispersed on Si/SiO2 templates, their position with respect to the alignment marks was located, and the contact pattern was defined by e-beam lithography. Ti(5 nm)/Al(25 nm)/Ti(15 nm)/Au(100 nm) metalization was deposited on the nanowire-doped extremities by e-beam evaporation. Figure 3 displays a scanning electron microscopy (SEM) image of a single contacted nanowire. The nanowire current-voltage characteristics were tested using a probe station and a Keithley 2636 source-meter. The measurements were performed at RT in vacuum (P < 10-4 mbar). The I-V characteristics of single nanowires from the reference sample and the heterostructured samples 1 and 2 are displayed in Figure 3c-e). The electrical parameters of the analyzed nanowires are summarized in Table 1. The dark nanowire resistance strongly differs in the three studied samples. The average value of zero-bias dark resistance increases from ∼0.4 MΩ in the reference n-i-n sample, to ∼30 MΩ in sample 1 and reaches ∼1 TΩ in sample 2. This trend is related to the presence of the heterostructure, as schematically reported in Figure 3c-e). In the GaN n-i-n

FIGURE 1. (a) Schematic representation of the three analyzed nanowire samples. (b) STEM image of nanowires from sample 1. Inset displays a high-angle annular dark field (HAADF) STEM image showing a detail of the external GaN shell. (c) HAADF STEM image of a nanowire from sample 2.

5-10 nm is formed around the lower nanowire part due to the lateral growth.11 For sample 1, an additional GaN shell of 5-15 nm surrounding the QDisc region and the lower nanowire part was formed (Figure 1b). The formation of the GaN shell was avoided in sample 2 (Figure 1c) by increasing the growth temperature by 20 °C during the deposition of the GaN cap. Previous studies performed on MBE-grown GaN nanowires have shown that the lateral growth can be reduced by increasing the substrate temperature.12 It should be noted, however, that the formation of the lateral GaN shell depends on the diffusion of Ga adatoms which is affected by the growth temperature but also by the strain state of the underlying AlN layer. Detailed investigations of the growth conditions allowing control of the lateral growth of GaN are underway and will be the object of a dedicated study. The optical properties of nanowire ensembles were characterized by photoluminescence (PL) spectroscopy. The spectra were measured at 4 K and at room temperature (RT) using a Jobin Yvon HR460 spectrometer equipped with a CCD camera. The excitation was provided with a continuous wave frequency-doubled Ar2+ laser (λ ) 244 nm). Figure 2 displays the RT PL spectra of the nanowire ensembles of the reference sample and of the heterostruc© 2010 American Chemical Society

2940

DOI: 10.1021/nl1010977 | Nano Lett. 2010, 10, 2939-–2943

of the electron concentration in the dark. The photosensitivity factor increases in sample 1 to Iph/Idark ∼ 2 and in sample 2 it reaches 5 × 102. In sample 1, the generated photocurrent has the same order of magnitude as the dark current. It could be related to the carrier generation in the QDiscs but also in the outer GaN shell and in the GaN extremities. In nanowires from sample 2 the photocurrent is about 2 orders of magnitude higher than the dark current, which illustrates the efficiency of the heterostructures as an electron blocking layer. Considering the nanowire surface exposed to the light as the active device area, the responsivity is in the range Rref ) 8 × 104 to 4 × 105 A/W for the reference sample, R1 ) 1 × 104 to 6 × 104 A/W for sample 1, and R2 ) 1 × 102 to 2 × 103 A/W for sample 2. Finally, it should be noted that for all samples the illumination with visible light (λ ) 500 nm) does not produce photocurrent. The nanowire photocurrent spectra were measured using a tunable vis-UV light source, consisting of a Xe lamp coupled with a Jobin Yvon Triax 180 spectrometer. The spectral resolution of the system used in this study is ∼40 meV. The illumination conditions were kept the same for all analyzed nanowires. The RT photocurrent spectra of nanowires from the three studied samples are reported in Figure 4. The spectra are normalized by the optical response of the setup. The photoresponse of the reference n-i-n nanowire basically reproduces the GaN bulk absorption. The photocurrent appears at ∼3.25 eV starting with a steep onset followed by a gradual increase above the GaN band gap. The photocurrent onset below the GaN band gap can be explained either by the contribution of sub-band-gap band tails due to Si doping or by the Franz-Keldysh effect due to the lateral band bending.9 No photocurrent associated with either blue or yellow band is observed within experimental accuracy. The photocurrent spectrum of nanowires from sample 1 has a similar behavior but shows a change of slope above the GaN band gap. This slope change could be related to the contribution of the strained GaN lower nanowire part. Because of the presence of the GaN shell around the QDisc region, the photocurrent contribution of the QDiscs is masked by a high background current and cannot be identified. A different behavior is observed for nanowires from sample 2. Indeed, the nanowires present a significant photocurrent contribution under the GaN band gap. The photocurrent signal starts at ∼2.6-2.8 eV, goes through a maximum, and then once again increases close to the GaN band gap. The low-energy contribution is attributed to the funda-

FIGURE 3. (a) SEM image displaying in the metal pads for macroscopic electrical access. (b) The single contacted nanowire. The I-V characteristics at T ) 300 K of the nanowire from (c) reference sample, (d) sample 1, and (e) sample 2. The blue dotted lines correspond to the dark current; the red solid lines correspond to illumination with the continuous 300 nm light with 5 mW/cm2 power density. The sketches on the lower right-hand side of each plot depict the conduction regime for each sample.

nanowires, the conduction takes place through the entire nanowire cross section except for a narrow depletion region close to the lateral surface.8 The presence of QDiscs increases the wire resistance in sample 1; however the GaN shell provides a leakage channel. The GaN shell is absent in sample 2 and the current is forced to flow through the multiple AlN barriers, which drastically increases the nanowire dark resistance. In addition, in sample 2 the resistance of the contact to the base nanowire side covered by AlN shell should be higher. The dark I-V characteristics are compared to the I-V characteristics under UV illumination at λ ) 300 nm with a power density of 5 mW/cm2. The reference n-i-n sample shows only a weak dependence on the illumination, its photosensitivity factor at -1 V (defined as the ratio of the photocurrent over the dark current) is Iph/Idark ) 0.05. Indeed, the photogenerated carriers represent only a small fraction

TABLE 1. Electrical Parameters of Nanowires from the Reference Sample, Sample 1, and Sample 2 nanowire sample parameter dark current at -1 V, Idark (A) dark resistance at 0 V, R0,dark (Ω) photosensitivity factor at -1 V, Iph/Idark © 2010 American Chemical Society

ref 6.2 × 10 4 × 105 <0.05

1 -6

2941

1.7 × 10 3 × 107 <2

2 -7

5.2 × 10-14 ∼1012 5 × 102

DOI: 10.1021/nl1010977 | Nano Lett. 2010, 10, 2939-–2943

FIGURE 5. (a) Photocurrent spectra of a nanowire from sample 2 collected at RT (black dashed line) and at 10 K (green solid line). (b) RT photocurrent spectra of the same nanowire under -3 V (blue dashed line) and +3 V (red solid line) applied bias. (c) Simulation of the conduction band of the QDiscs in the framework of a onedimensional effective mass approximation for different applied bias (-3 V, blue line; 0 V, black line; +3 V, red line). Only 10 QDiscs are shown for clarity.

gap should be mainly generated by the fundamental absorption in narrow QDiscs and by excited transitions in large QDiscs, for which carriers have a stronger extraction probability. However, a contribution of the GaN nanowire extremities cannot be excluded. The electron-hole pairs generated in the QDiscs can contribute to the photocurrent by thermal activation over the barrier or by tunneling. The hypothesis of thermal escape is confirmed by comparison of photocurrent spectra at different temperatures. Figure 5a displays the photocurrent spectra at RT and at 10 K of a nanowire from sample 2. The photocurrent is reduced by 1 order of magnitude when lowering the temperature. The spectral shape is also changed. The photocurrent onset is shifted to higher energies at low temperature according to the band gap increase. The subband-gap contribution of the large QDiscs is strongly reduced and is almost masked by the background noise at 10 K. This strong temperature dependence demonstrates that the photogenerated carrier extraction is thermally assisted. Figure 5b reports the photocurrent spectra of the same nanowire for -3 V and +3 V applied bias. For positive bias the photocurrent over the whole spectral range is weaker than that for negative bias. In particular, the sub-band-gap contribution of large QDiscs is reduced by almost 1 order of magnitude. This asymmetry can be understood based on a simulation of the band profile. Figure 5c shows the conduction band diagram along the nanowire axis calculated in the framework of a one-dimensional effective mass approximation. It shows that at zero bias there is a band tilting across the QDisc region. For negative bias the external voltage drop

FIGURE 4. Normalized photocurrent spectra collected at RT under bias Vb ) -3 V from a single nanowire from (a) reference sample, (b) sample 1, and (c and d) sample 2. (e) Schematic band diagram illustrating the photocurrent generation mechanism in the QDiscs.

mental hh1-e1 absorption in the large QDiscs, as illustrated in Figure 4e. It must not be confused with defect-related photogeneration mechanisms, such as the so-called blue band, which lies at lower energies.15 The low-energy QDisc contribution is peaked at 2.9-3.2 eV depending on the investigated nanowire because of the fluctuation of the QDisc size from wire to wire. The onset of the photocurrent is in good agreement with the QDisc PL energy of the subband-gap PL of sample 2 at RT (Figure 2). This gives further evidence that the mechanism responsible for the sub-bandgap photocurrent is the interband absorption in the QDisc region. The broadening of the low energy peak is due to the contribution of QDiscs of different thicknesses within the same wire but also to the contribution of absorptions implying light holes and crystal-field split holes. The responsivity at 2.9 eV corresponding to the contribution of large QDiscs is estimated to be 13 A/W (assuming that the active detector area is equal to the exposed surface of the QDisc region). This value is weak compared to the responsivity at high energies (2 × 103 A/W at 4.1 eV). This is due to the low efficiency of the carrier extraction from the ground state of large QDiscs, which lies at a much lower energy than the barriers. Photocurrent at energies higher than the GaN band © 2010 American Chemical Society

2942

DOI: 10.1021/nl1010977 | Nano Lett. 2010, 10, 2939-–2943

Acknowledgment. This work was supported by the French ANR agency under the programs ANR-08-NANO-031 BoNaFo and ANR-08-BLAN-0179 NanoPhotoNit.

adds up with the band tilting which enhances the extraction and the transport of photogenerated carriers through the QDisc region. At the same time, the band tilting creates an additional barrier which reduces the dark current. For positive bias, the external voltage is opposed to the band tilting so that the effective voltage drop across the heterostructure region is reduced, thus lowering the extraction of carriers from the QDiscs. The band tilting across the QDisc region at zero bias is expected to produce a photovoltaic effect under UV illumination. This was experimentally verified by probing the photocurrent at zero bias. The short-circuit photocurrent equals 0.2 pA under UV illumination at zero bias, corresponding to a responsivity of 0.1 A/W when normalized to the exposed nanowire surface. In conclusion, we have demonstrated a single-nanowire photodetector relying on the carrier generation in GaN/AlN QDiscs. The nanowire properties have been analyzed by scanning transmission electron microscopy, photoluminescence spectroscopy, electrical characterizations, and photocurrent spectroscopy. The photoluminescence studies showed that the QDiscs emission energy is lower than the GaN band gap as a consequence of quantum confined Stark effect. The sub-band-gap QDisc contribution could also be identified in the photocurrent spectra, demonstrating localized photodetection in the heterostructure region. The QDisc based photodetectors showed a strong reduction of the dark current with respect to a reference GaN sample. The responsivity at λ ) 300 nm and at -1 V applied bias was measured as high as 2 × 103 A/W when normalized by the photodetector size.

© 2010 American Chemical Society

REFERENCES AND NOTES (1) (2) (3) (4) (5)

(6) (7) (8) (9) (10) (11)

(12) (13) (14) (15)

2943

Gradecˇak, S.; Qian, F.; Li, Y.; Park, H.-G.; Lieber, C. M. Appl. Phys. Lett. 2005, 87, 173111. Qian, F.; Gradecˇak, S.; Li, Y.; Wen, C.-Y.; Lieber, C. M. Nano Lett. 2005, 5 (11), 2287–2291. Dong, Y.; Tian, B.; Kempa, T. J.; Lieber, C. M. Nano Lett. 2009, 9 (5), 2183–2187. Soci, C.; Zhang, A.; Xiang, B.; Dayeh, S. A.; Aplin, D. P. R.; Park, J.; Bao, X. Y.; Lo, Y. H.; Wang, D. Nano Lett. 2007, 7, 1003. Prades, J. D.; Jimenez-Diaz, R.; Hernandez-Ramirez, F.; Fernandez-Romero, L.; Andreu, T.; Cirera, A.; Romano-Rodriguez, A.; Cornet, A.; Ramon Morante, J.; Barth, S.; Mathur, S. J. Phys. Chem. C 2008, 12, 14639. Chen, R. S.; Chen, H. Y.; Lu, C. Y.; Chen, K. H.; Chen, C. P.; Chen, L. C.; Yang, Y. J. Appl. Phys. Lett. 2007, 91, 223106. Han, S.; Jina, W.; Zhanga, D.; Tanga, T.; Lia, C.; Liua, X.; Liua, Z.; Leia, B.; Zhou, C. Chem. Phys. Lett. 2004, 389, 176. Calarco, R.; Marso, M.; Richter, T.; Aykanat, A. I.; Meijers, R.; Hart, A. v.d.; Stoica, T.; Lu¨th, H. Nano Lett. 2005, 5 (5), ]?>), 981–984. Cavallini, A.; Polenta, L.; Rossi, M.; Stoica, T.; Calarco, R.; Meijers, R. J.; Richter, T.; Lu¨th, H. Nano Lett. 2007, 7, 2166. Songmuang, R.; Landre´, O.; Daudin, B. Appl. Phys. Lett. 2007, 91, 251902. Tchernycheva, M.; Sartel, C.; Cirlin, G.; Travers, L.; Patriarche, G.; Harmand, J.-C.; Dang, Le Si.; Renard, J.; Gayral, B.; Nevou, L.; Julien, F. H. Nanotechnology 2007, 18, 385306. Meijers, R.; Richter, T.; Calarco, R.; Stoica, T.; Bochem, H.-P.; Marso, M.; Lu¨th, H. J. Cryst. Growth 2006, 289, 381. Renard, J.; Songmuang, R.; Tourbot, G.; Bougerol, C.; Daudin, B.; Gayral, B. Phys. Rev. B 2009, 80, 121305. Reshchikov, M. A.; Morkoc, H. J. Appl. Phys. 2005, 97, 061301. Cavallini, A.; Polenta, L.; Rossi, M.; Richter, T.; Marso, M.; Meijers, R.; Calarco, R.; Luth, H. Nano Lett. 2006, 6, 1548.

DOI: 10.1021/nl1010977 | Nano Lett. 2010, 10, 2939-–2943