pubs.acs.org/NanoLett
Spatial Distribution of Defect Luminescence in GaN Nanowires Qiming Li, and George T. Wang* Sandia National Laboratories, P.O. Box 5800, Albuquerque, New Mexico 87185 ABSTRACT The spatial distribution of defect-related and band-edge luminescence from GaN nanowires grown by metal-organic chemical vapor deposition was studied by spatially resolved cathodoluminescence imaging and spectroscopy. A surface layer exhibiting strong yellow luminescence (YL) near 566 nm in the nanowires was revealed, compared to weak YL in the bulk. In contrast, other defect-related luminescence near 428 nm (blue luminescence) and 734 nm (red luminescence), in addition to band-edge luminescence (BEL) at 366 nm, were observed in the bulk of the nanowires but were largely absent at the surface. As the nanowire width approaches a critical dimension, the surface YL layer completely quenches the BEL. The surface YL is attributed to the diffusion and piling up of mobile point defects, likely isolated gallium vacancies, at the surface during growth. KEYWORDS GaN, nanowire, cathodoluminescence, defects, yellow, III-nitride.
G
aN-based nanowires have attracted considerable attention as potential nanoscale building blocks for future optical, electrical, or optoelectronic devices. Even though GaN nanowire-based devices, such as light emitting diodes,1 high electron mobility transistors,2 photovoltaics,3 and field emission transistors,4,5 have been demonstrated, understanding and optimizing the electrical and optical properties of GaN-based nanowires are critical for realizing their full potential in device applications. GaN-based nanowires are known to be generally free of threading dislocations due to their nanometer to submicrometer scale in their lateral dimensions, high aspect ratio, and efficient strain relaxation.6 However, point defects are still incorporated7-10 and detrimentally impact the performance of the GaN-based nanowire devices. In GaN thin films and bulk materials, various defects, including Ga vacancies and carbon and oxygen impurities, are believed to cause radiative transitions from the conduction band (or a shallow donor) to deep acceptors and give rise to defect-related luminescence bands.11 In GaN nanowires grown by various techniques, the defect related yellow-luminescence band6,12-16 and blue band13,14 have also been observed and have been suggested to have similar origins as in GaN thin films or bulk materials.6,12,15-18 However, the identity and spatial distribution of the point defects in GaN nanowires has not been fully established. In particular, a link of the yellow luminescence (YL) near 560 nm with surface states in GaN thin films19 and bulk materials20 has been previously suggested, which could have significant implications for GaN nanowires given their high surface-to-bulk ratio. In addition, previous studies on the dependence of the electrical and optical characteristics on the
nanowire dimensions suggest a nonhomogenous distribution of defects in the GaN nanowire.15,16,21 To further elucidate nature of defects in GaN nanowires, we present here a detailed study on the spatial distribution of defect-related luminescence in GaN nanowires using spatially resolved cathodoluminescence (CL) imaging and spectroscopy. CL can achieve much higher resolutions compared to other optical probe techniques such as microphotoluminescence (better than 20 nm spatial resolution have been recently reported using a scanning transmission electron microscopybased CL setup),22 which is of particular importance in studying nanostructures. Two inch (1-102) r-plane sapphire wafers were used as substrates for GaN nanowire growth. The lattice alignment between r-plane sapphire and GaN results in primarily vertically aligned nanowires with a [11-20] growth orientation and isosceles-triangle cross sections with one {0001} and two {11¯01} side facets.17,23,24 A 2 nm thick Ni film was first deposited onto the sapphire wafers using electron beam evaporation as a catalyst for vapor-liquid-solid-based GaN nanowire growth.25 Unintentionally doped n-type GaN nanowires were subsequently grown on the Ni-coated sapphire at 900 °C in an Emcore D125 metal-organic chemical vapor deposition (MOCVD) reactor. The flow rate of trimethylgallium and NH3 is set to 1.1 and 348 millimol per minute, respectively. The typical nanowire growth rate at these growth conditions is 2 µm per min. The growth time for the nanowires was 7.5 min. Further details on the growth conditions, structure, and properties of GaN nanowires grown by our group by this method have been previously published,7,8,16,17,23,25-29 and the nanowires were determined to be unintentionally n-type doped via electrical characterization.8,16 To prepare nanowire cross sections for study by CL, a layer of indium metal was melted onto asgrown samples on a hot plate at 160 °C. A large number of GaN nanowires were imbedded in the indium layer during
* To whom correspondence should be addressed. E-mail:
[email protected]. Received for review: 10/20/2009 Published on Web: 04/14/2010 © 2010 American Chemical Society
1554
DOI: 10.1021/nl903517t | Nano Lett. 2010, 10, 1554–1558
FIGURE 1. (a). SEM image showing the triangular cross-section of a GaN nanowire; (b-e) monochromatic CL images of the same nanowire at 366, 428, 566, and 734 nm; (f). composite image of (b) and (d).
the cooling of the melt. After cooling, the indium layer was peeled off of the sapphire substrate, revealing exposed nanowire cross sections and was subsequently mounted onto a Si wafer with the exposed nanowire side facing upward. The indium-embedded nanowire samples were immediately loaded into the CL chamber equipped with a Zeiss Gemini scanning electron microscope (SEM), a Gatan MonoCL3 system, and a high-sensitivity, water-cooled GaAs: Cs photomultiplier tube. The CL experiments were performed at room temperature and ultrahigh vacuum conditions, with the impinging electron beam energy set to 5 KeV. A representative SEM image of a GaN nanowire with exposed cross-section is shown in Figure 1a. CL spectra of the nanowire were taken showing emission peaks centered around 366 nm (band-edge luminescence, BEL), 428 nm (blue luminescence, BL), 566 nm (YL), and 734 nm (red luminescence, RL). Monochromatic CL images were subsequently acquired from the same nanowire at these four wavelengths and are shown in Figure 1b-e. In Figure 1b, intense BEL (366 nm) is seen at the nanowire cross section as expected. A monochromatic CL image taken at the center of the BL band at 428 nm is shown in Figure 1c, where the 428 nm emission is seen to distribute uniformly over the cross-section of the nanowire. This result suggests that the BL related defects are uniformly distributed in the nanowire. In dramatic contrast to the BEL and BL, the YL at 566 nm is observed to be highly localized near the surface of the GaN nanowire as shown in Figure 1d. Further examples of this phenomenon can be repeatedly observed for additional nanowires as shown in the Supporting Information. From the 566 nm CL image, YL surface layer thicknesses of ∼80 and ∼60 nm are measured on two different facets, which may possibly be related to the different {11¯01} and {0001} facets of the nanowire. The thickness of the surface YL on the third facet cannot be measured as it cannot be resolved from the YL from the nanowire sidewall. The thickness of the surface YL layer was defined as the full width at half © 2010 American Chemical Society
maximum of the CL intensity profile. In the bulk region of the nanowire, the monochromatic CL image at 566 nm appears relatively dark; however, the YL intensity near the center of cross section of the nanowire is nonzero. We observed a relatively weak YL intensity using the spot mode CL, which will be discussed in the following section, and which indicates weaker YL originating from the nanowire bulk. An additional defect related emission at 734 nm corresponding to the red luminescence (RL) band in GaN is also observed and has a uniform distribution in the nanowire bulk, as shown in Figure 1e. Interestingly, we note that the 566 nm monochromatic CL image (Figure 1d) shows an elevated background (YL outside the nanowire) as compared with the other monochromatic CL images. We believe this elevated YL background may result from the strong YL surface component on the GaN nanowires. When the impinging electron beam scans to a location adjacent to a nanowire, some scattered electrons impact the nearby GaN nanowire surface. These reduced-energy, scattered electrons primarily excite the surface region of the GaN nanowire, which leads to a YL background signal in the image. The indium/Si substrate was verified as not being the source of the background YL by scanning an area distant from any nanowires. We also note that the observed CL intensities of the BEL appear higher in the lower left corner/edge of the nanowire, but not for the BL. The exact origin of the observed BEL intensity variation in the nanowire cross-section has not yet been determined, although it could possibly be due to the presence and distribution of nonradiative defects,7,29 which as a competing carrier recombination pathway could locally reduce BEL. The lower observed CL intensity at the left edge of the nanowire is likely an edge effect whereby the electron beam interaction volume is reduced by the concave interior angle. The spatial dependence of the nanowire optical emissions was further investigated by taking CL spectra while focusing the electron beam at designated locations across a nanowire 1555
DOI: 10.1021/nl903517t | Nano Lett. 2010, 10, 1554-–1558
FIGURE 3. The measured thickness of the surface YL layer as a function of the nanowire dimension (width). The experimental data (open circles) is fitted by a linear function.
defects giving rise to BL and RL are not surface-related and also that BEL is quenched in the YL surface region of the nanowire. Figure 3 shows the measured thicknesses of the surface YL layers as a function of the nanowire dimension (width) obtained from monochromatic CL mapping at 566 nm of the cross sections of 17 different GaN nanowires. Interestingly, the thickness of the surface YL is observed to increase with increasing nanowire dimension in a near-linear fashion. The linear function is expressed by y ) 0.07x + 45 nm, where y represents the thickness of surface YL and x is the dimension of the nanowires. The term “0.07x” represents the increase rate of the surface YL thickness and the constant term “45 nm” suggests there is a critical nanowire dimension (dc) defined as a nanowire dimension when the surface YL completely “pinches off” the BEL. We note that due to some degree of nanowire tilt from the surface normal when imaged, the measured YL thickness represents a “projection” of the true YL thickness, which leads to an underestimate. For the measurements in Figure 3, the large majority of nanowires were estimated from cross-section profiles to have tilt angles less than 30° from normal, which leads to an upper-bound underestimation of the true YL of 13%. Assuming a “worst-case” 13% underestimation error for all points in Figure 3 leads to a adjusted linear fit expressed by y ) (0.075 ( 0.005)x + (48 ( 3) nm. We further measured dc by studying tapered GaN nanowires, the shape of which results from competing nanowire sidewall growth. The SEM image in Figure 4a represents a typical tapered nanowire and its monochromatic CL images for BEL and YL are given in Figure 4b,c, respectively. A salient feature is that when the dimension of the nanowire is less than ∼170 nm the BEL is below the detection limit whereas YL can be clearly seen as the nanowire diameter decreases below 170 nm. By moving the electron beam along the center line of the nanowire axis, we obtained a series of CL spectra (Figure 4d) as a function of the nanowire dimension. These spectra confirmed that the BEL can no
FIGURE 2. (a) A monochromatic CL image obtained at 566 nm of a GaN nanowire. (b) CL spectra acquired while focusing the electron beam at various locations on the nanowire cross-section as shown in (a).
cross-section. Figure 2a shows a 566 nm monochromatic CL image of a nanowire, where a ∼100 nm thick YL surface layer is clearly observed. When the electron beam is focused at the outer edge of this surface YL layer (Spot a), strong YL is seen to dominate the CL spectrum while the BEL near 366 nm is barely detectable (Spectrum a, Figure 2b). As the electron beam is moved to the center region of the surface YL layer (Spot b), the height of the BEL peak is estimated to increase by ∼3 times while the YL intensity remains roughly constant (Spectrum b). When the electron beam is focused at the inner edge of the surface YL layer (Spot c), the BL and RL peaks appear and the BEL intensity continues to increase, while the YL intensity still remains roughly constant. As the electron beam is targeted at the center of the nanowire (Spot d), the BL, RL, and BEL peak intensities saturate while the YL decreases dramatically but does not go to zero (Spectrum d). This nonzero YL intensity was further confirmed at the centers of larger nanowires with dimensions (widths) on the order of 1 µm. We therefore rule out the possibility that the YL detected in the bulk (core) is caused by excitation of the surface YL layer because the nanowire dimension is larger than electron energy dissipation volume (∼0.32 µm diameter) of the excitation beam, as estimated using a Monte Carlo simulation. The data from Figures 1 and 2 thus indicate that the BEL, BL, and RL emissions are absent from the surface region exhibiting strong YL. This indicates that the © 2010 American Chemical Society
1556
DOI: 10.1021/nl903517t | Nano Lett. 2010, 10, 1554-–1558
can be assumed to be primarily comprised of VLS-grown “core” material rather than sidewall-growth material, since the axial VLS growth rate is substantially faster than the radial sidewall growth rate. Although this tip region is seen to be dominated by YL in Figure 4, it is difficult to determine whether this is a result of YL-related defects being incorporated through the VLS growth mechanism (e.g., catalyst tip) or by YL resulting from surface dangling bonds combined with the finite resolution of the technique. The thickness of the surface YL layer is shown in Figure 3 to increase with the dimension (width) of the nanowire, which suggests at first glance a mechanism whereby YL-related point defects are incorporating through sidewall growth, since the increase in nanowire dimension is primarily due to increased sidewall growth. (The diameters of the VLS-grown core of the nanowires are observed to be under ∼100 nm (based on the nanowire diameter adjacent to the catalyst tip) and should be constant throughout the length of any given nanowire.) However, the observed thickness of the surface YL layer at a given nanowire dimension (Figure 3) is much less than the expected thickness of the nanowire sidewall growth, given a typical nanowire core dimension of ∼100 nm or less. We propose that this suggests that YL-related defects may initially form or incorporate through sidewall growth or a combination of sidewall and VLS (catalyst) growth and then subsequently redistribute toward the surface during nanowire growth via a diffusion-related mechanism. Isolated gallium vacancies (VGa), which have been linked to YL,11 were reported to become mobile at 500-600 K with an estimated migration energy of 1.5 eV30 and thus represent a potential defect source for the surface-related YL. The weaker bulk related YL may arise from other less-mobile defects linked to YL, such as C and O impurities,11 or from the same defects responsible for the surface YL (e.g., VGa) but at lower concentrations. BL and RL are observed to be fairly evenly distributed in the nanowire bulk without a distinct surface component, which suggests that the BL and RL related defects may be relatively stable and do not migrate toward the surface during growth. Potential candidates for the BL related defect include VGa-related complexes, specifically VGa-ON,31,32 which has been reported to have a relatively high diffusion barrier of 2.2 eV,30,33 and thus may be relatively immobile. The RL band is less common and less studied in GaN than the YL and BL bands.11 In GaN thin films, RL was reported in carbon-doped GaN and attributed to have an origin from threading dislocations.34,35 However, since these GaN nanowires are free of threading dislocations,17 a different type of defect must be considered in this case. Hofmann et al. observed a RL at 1.8 eV in MBE grown GaN and suggested that VGa-ON or VGa-SiGa could be responsible for the RL.36 Since the nanowires in this study were not intentionally doped with Si, the VGa-SiGa complex is unlikely to represent the source of the RL here. These complexes are also expected to be relatively immobile, consistent with our hy-
FIGURE 4. (a) SEM image showing a tapered GaN nanowire. The scale bar is 0.5 um. (b,c) The corresponding monochromatic CL images at 366 and 566 nm. (d). Spot-mode CL spectra obtained along the center line of the nanowire are plotted as a function of the nanowire dimension (width).
longer be detected as the dimension of the nanowire decreases below a critical nanowire dimension dc, which in this case is estimated to be ∼175 nm. On the basis of the geometry of an equilateral triangle, a close approximation of our nanowire geometry, a surface YL thickness that corresponds to this dc is ∼50 nm, which agrees reasonably well with the value of 57 nm as extracted from the linear function in Figure 3. Above the critical nanowire dimension, BEL and BL intensities are seen in Figure 4d to increase with increasing nanowire diameter (and hence increasing volume probed) up to ∼300 nm, where the dissipation volume becomes the limiting factor. The YL intensity follows a similar trend based on increased volume probed as well as the increase in YL layer thickness with increasing nanowire dimension observed in Figure 3. YL can also be observed at the catalyst tip region in Figure 4c, which is the likely result of GaN surrounding the catalyst as confirmed through energy dispersive X-ray mapping. Point defects may be incorporated into the nanowire though the vapor-liquid-solid (VLS) growth mechanism at the Ni catalyst tips or though the competing lateral (“sidewall”) growth on the nanowire side walls. The nanowire surface itself may also be a source of defects (e.g., dangling bonds). In Figure 4, the nanowire region directly below the Ni catalyst tip has a diameter of approximately 60 nm and © 2010 American Chemical Society
1557
DOI: 10.1021/nl903517t | Nano Lett. 2010, 10, 1554-–1558
(9)
pothesis. Regardless of their respective origins, the absence of surface-related BEL, BL, and RL indicates that YL dominates the other radiative recombinative pathways in the surface region. In summary, defect-related luminescence in GaN nanowires was investigated using spatially resolved CL. The results reveal the presence of a strong YL (566 nm) surface layer and a much weaker YL component in the nanowire bulk. In contrast, BEL at 366 nm and other defect-related luminescence at 428 nm (BL) and 734 nm (RL) are observed only in the nanowire bulk and not at the surface, where YL dominates. The strong surface-related YL layer may be caused by the diffusion and piling up of mobile point defects, such as Ga vacancies, at the surface during growth. As the nanowire dimension decreases to a critical dimension, the nanowire is pinched-off by the high-defect-density surface layer, entirely quenching BEL.
(10) (11) (12)
(13) (14) (15)
(16) (17) (18)
(19)
Acknowledgment. We acknowledge support from the DOE Office of Basic Energy Sciences and Sandia’s Laboratory Directed Research and Development program. We thank Andrew Armstrong for insightful discussions. Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energy’s National Nuclear Security Administration under contract No. DE-AC04-4AL85000.
(20) (21) (22) (23) (24) (25) (26)
Supporting Information Available. Monochromatic images of YL at 566 nm for sixteen GaN nanowire cross sections showing a strong YL surface component. All scale bars represent 500 nm. This material is available free of charge via the Internet at http://pubs.acs.org.
(27)
(28)
REFERENCES AND NOTES (1) (2) (3) (4) (5) (6) (7) (8)
(29)
Qian, F.; Li, Y.; Gradecak, S.; Wang, D. L.; Barrelet, C. J.; Lieber, C. M. Nano Lett. 2004, 4, 1975. Li, Y.; Xiang, J.; Qian, F.; Gradecak, S.; Wu, Y.; Yan, H.; Blom, D. A.; Lieber, C. M. Nano Lett. 2006, 6, 1468. Dong, Y.; Tian, B.; Kempa, T. J.; Lieber, C. M. Nano Lett. 2009, 9, 2183. Yu, H.; Xiangfeng, D.; Yi, C.; Lieber, C. M. Nano Lett. 2002, 2, 101. Vandenbrouck, S.; Madjour, K.; Theron, D.; Yajie, D.; Yat, L.; Lieber, C. M.; Gaquiere, C. IEEE Electron Device Lett. 2009, 30, 322. Hersee, S. D.; Sun, X.; Wang, X. Nano Lett, 2006, 6, 1808. Armstrong, A.; Li, Q.; Bogart, K. H. A.; Lin, Y.; Wang, G. T. J. Appl. Phys. 2009, 106, No. 053712. Armstrong, A.; Wang, G. T.; Talin, A. A. J. Electron. Mater. 2009, 38, 484.
© 2010 American Chemical Society
(30) (31) (32) (33) (34) (35) (36)
1558
Ji, H.; Kuball, M.; Burke, R. A.; Redwing, J. M. Nanotechnology 2007, 18. Furtmayr, F.; Vielemeyer, M.; Stutzmann, M.; Laufer, A.; Meyer, B. K.; Eickhoff, M. J. Appl. Phys. 2008, 104, 074309. Reshchikov, M. A.; Morkoc, H. J. Appl. Phys. 2005, 97, 61301. Chen, C. C.; Yeh, C. C.; Chen, C. H.; Yu, M. Y.; Liu, H. L.; Wu, J. J.; Chen, K. H.; Chen, L. C.; Peng, J. Y.; Chen, Y. F. J. Am. Chem. Soc. 2001, 123, 2791. Li, J. Y.; Chen, X. L.; Qiao, Z. Y.; Cao, Y. G.; Lan, Y. C. J. Mater. Sci. Lett. 2001, 20, 757. Cheng-Shan, X.; Ying-Ge, Y.; Hong-Lei, M.; Hui-Zhao, Z.; Jin, M. Chin. Phys. Lett. 2003, 20, 568. Mastro, M. A.; Maximenko, S.; Gowda, M.; Simpkins, B. S.; Pehrsson, P. E.; Long, J. P.; Makinen, A. J.; Freitas, J. A., Jr.; Hite, J. K.; Eddy, C. R., Jr.; Kim, J. J. Cryst. Growth 2009, 311, 2982. Talin, A. A.; George, T. W.; Elaine, L.; Richard, J. A. Appl. Phys. Lett. 2008, 92, 093105. Wang, G. T.; Talin, A. A.; Werder, D. J.; Creighton, J. R.; Lai, E.; Anderson, R. J.; Arslan, I. Nanotechnology 2006, 17, 5773. Dhara, S.; Datta, A.; Wu, C. T.; Lan, Z. H.; Chen, K. H.; Wang, Y. L.; Chen, Y. F.; Hsu, C. W.; Chen, L. C.; Lin, H. M.; Chen, C. C. Appl. Phys. Lett. 2004, 84, 3486. Shalish, I.; Kronik, L.; Segal, C.; Rosenwaks, Y.; Shapira, Y.; Tisch, U.; Salzman, J. Phys. Rev. B 1999, 59, 9748. Reshchikov, M. A.; Morkoc, H.; Park, S. S.; Lee, K. Y. Appl. Phys. Lett. 2001, 78, 3041. Calarco, R.; Marso, M.; Richter, T.; Aykanat, A. I.; Meijers, R.; Hart, A. V.; Stoica, T.; Luth, H. Nano Lett. 2005, 5, 981. Lim, S. K.; Brewster, M.; Qian, F.; Li, Y.; Lieber, C. M.; Gradecak, S. Nano Lett. 2009, 9, 3940. Li, Q.; Wang, G. T. J. Cryst. Growth 2008, 310, 3706. Li, Q.; Lin, Y.; Creighton, J. R.; Figiel, J. J.; Wang, G. T. Adv. Mater. 2009, 21, 2416. Li, Q.; Wang, G. T. Appl. Phys. Lett. 2008, 93, No. 043119. Arslan, I.; Talin, A. A.; Wang, G. T. J. Phys. Chem. C 2008, 112, 11093. Prasankumar, R. P.; Choi, S. G.; Wang, G. T.; Picraux, S. T.; Taylor, A. J. Ultrafast Carrier Dynamics in Semiconductor Nanowires. Proceedings of IEEE Lasers and Electro-Optics Society (LEOS), New York, NY, May 4-9, 2008; p 1796. Westover, T.; Jones, R.; Huang, J. Y.; Wang, G.; Lai, E.; Talin, A. A. Nano Lett. 2009, 9, 257. Upadhya, P. C.; Li, Q.; Wang, G. T.; Fischer, A. J.; Taylor, A. J.; Prasankumar, R. P. Semicond. Sci. Technol. 2010, 25, No. 024017. Saarinen, K.; Suski, T.; Grzegory, I.; Look, D. C. Physica B 2001, 308, 77. Toth, M.; Fleischer, K.; Phillips, M. R. Phys. Rev. B 1999, 59, 1575. Yang, H. C.; Lin, T. Y.; Chen, Y. F. Phys. Rev. B 2000, 62, 12593. Tuomisto, F.; Saarinen, K.; Paskova, T.; Monemar, B.; Bockowski, M.; Suski, T. J. Appl. Phys. 2006, 99. Klein, P. B.; Freitas, J. A., Jr; Binari, S. C.; Wickenden, A. E. Appl. Phys. Lett. 1999, 75, 4016. Klein, P. B.; Binari, S. C.; Ikossi, K.; Wickenden, A. E.; Koleske, D. D.; Henry, R. L. Appl. Phys. Lett. 2001, 79, 3527. Hofmann, D. M.; Meyer, B. K.; Alves, H.; Leiter, F.; Burkhard, W.; Romanov, N.; Kim, Y.; Kruger, J.; Weber, E. R. Phys. Status Solidi A 2000, 180, 261.
DOI: 10.1021/nl903517t | Nano Lett. 2010, 10, 1554-–1558