ETE:S diffraction pattern indexed along the [1120] axis( Fig. Id, inset).a Second, these data show that the overall In gaN/Gan shell structure on the two (1101) facets is thicker, 65 nm, than on the 10001) facet, 10 nm. High-magnification data( Fig. Ih and Supplementary Information, Fig. S1) show clearly the 26-period InGaN/GaN MOw In gan quantum-well structure on the (1101) facets, whereas uantum-well contrast variation was observed on the (0001) facet. Third, EDS measurements( Supplementary Information, Fig. S2) further confirm this difference in shell thickness with a well-defined bur gnal recorded from the MQW shell on the (1101) facets contrast variation can also be observed at the intersection between Figure 2 High-resolution TEM of nanowire Mows. a, Bright-field TEM image the (1101) facets in some MQW nanowire samples. Preliminar of a typical 26Maw nanowire cross-section. The dash high-resolution TEM analysis of a 26MQW sample( Supplementary heterointerface between the core and shell. The scale bar is 10 nm. b Enlarge Information, Fig. S3)suggests that these features arise from image of MOws in a White arrows highlight In GaN quantum-well positions. The stacking faults at this intersection, although further studies will be dashed line outlines the interface between an InGaN quantum well and adjacent GaN equired to understand fully this interesting structural feature quantum barriers. The scale bar is 2 nm. Inset: Two-dimensional Fourier transforms To understand further the MQW growth differences on of the entire two(1101 lateral facets versus the (0001) facet of the nanowire, we carried out convergent-beam electron diffraction(CBED) Comparison of experimental and simulated(see the Methods In, 16 Gao. N and In 2 Ga,77N, are consistent with that expected section) CBED patterns(Fig. le)suggests that the (0001) bottom based previous studies of In gan shell growth. The peak facet of the MQW nanowire is an N-polarity face. Interestingly, broadening with In% is consistent with In gan planar structures previous studies of planar growth have reported that an N-face has Second, the photoluminescence images show strong emission a much slower growth rate than a Ga-face, and could thus explain from the nanowire bodies where excitation occurs, and also from the difference in shell growth for our nanowires. Given the more the nanowire ends, which indicates good waveguiding in the apid and controlled growth on the [1101) nanowire facets, we have core/MQW-shell structure. Third, the images also show that most focused on growth of multiple-quantum-well structures on these In-rich nanowires exhibited spontaneous bending. We attribute facets to explore limits of synthetic control and to prepare nanolaser this behaviour to the non-symmetrical shell coating, where large structures in which the gain medium is tuned. Notably, by varying strains are generated on (1101) versus(0001 )facets. It may the growth time and temperature and total repeat periods during interesting in the future to quantify the relationship between the MQW deposition (see the Methods section), we were able In composition and bending in these structures, and thus the to vary the In gan/gan quantum-well unit thicknesses, the In possible strain relaxation through mechanical bending. Last,we composition and total quantum-well numbers. Cross-sectional note that no impurity or defect-related emission was observed at STEM images of three distinct MQW nanowires used in our longer wavelengths in the photoluminescence spectra, consistent ubsequent studies( Fig. If-h)show structures with resolvable 3, with the high degree of structural order determined from our 3, and 26 periods, quantum-well thicknesses of about 2.4, 3.0 and high-resolution TEM analys 1.5 nm and average gan barrier thicknesses of about 40, 10 and Excitation power-dependent studies of a representative tively. The In composition control was confirmed by 26MQW nanowire structure were also carried out, as shown optical studies to be discussed later in Fig 3c. At low power density(245 kW cm-), we observed a We have further investigated the detailed structure of the relatively broad spontaneous emission peak centred at 448 nm In GaN/Gan quantum-well repeat by acquiring high-resolution As power density increased, the end emission intensity increased bright-field images from an independent 26MQW nanowire cross- rapidly and became dominant. Above a threshold of 900 kw cm on the (1101) nanowire facet(Fig 2a)reveals homogeneous narrow peaks centred around 438 nm. The full-width at quantum-well deposition as discussed above. High-magnification half-maximum of these peaks was less than 0.8 nm, limited by data of several quantum wells(Fig. 2b)reveal that the In gaN/gan the spectral resolution of our detection system and indicative interfaces are atomically sharp, with the In gan quantum-well of longitudinal modes in the nanowire cavity. To confirm this hickness estimated to be <l nm and approaching the growth assumption, we calculated the mode spacing using AX=2/(2n,L) limit of several monolayers. In addition, two-dimensional Fourier (ref. 24), where 2 is the emission wavelength, ne is the effective transforms from the MQW region(Fig 2b, inset) exhibit no refractive index and L is the nanowire cavity length. For 2=438 nm, plitting of reciprocal lattice peaks, suggesting that there is no strain ne= 2.625(ref. 25)and L= 50 um, the predicted mode spacing relaxation at the In GaN /Gan heterointerfaces These tEm results is around 0.73 nm, which is consistent with the measured mode onfirm that we can prepare dislocation-free MQW nanowire spacing 0.72-0.82 nm. These features together with the observed heterostructures with Ingan quantum-well thicknesses similar to characteristic nonlinear spectrally integrated output power versus or smaller than the bohr radius, 3 nm, for In gan(ref. 20). pump power density(Fig 3c, inset)are clear indications of lasing We have characterized the optical properties of the behaviour Comparison of the measured output power data with new core/MQW-shell nanowire heterostructures using a fit derived from the rate equations(see the Supplementary photoluminescence and modelling studies. Photoluminescence Information)yields a spontaneous emission factor, B, of 0.03 Rferenmission positions(Fig 3a, b)reveal several points. First, the lasers". The blueshift of lasing peaks relative to the spontaneous In GaN emission peak redshifts from 382 to 440 to 484 nm(Fig. 3b) emission is a distinct feature of In Gan MQW lasers, which has ith increasing In composition. The In gan quantum-well been attributed to band filling and/or photo-induced screening of compositions estimated from the emission data, In os Gao gs N, internal electric fields naturematerialsvol7iSepTembeR2008iwww.nature.com/naturematerials @2008 Macmillan Publishers Limited. All rights reserved.LETTERS diffraction pattern indexed along the [1120¯ ] axis (Fig. 1d, inset). Second, these data show that the overall InGaN/GaN shell structure on the two {1101 ¯ } facets is thicker, ∼65 nm, than on the {0001} facet, ∼10 nm. High-magnification data (Fig. 1h and Supplementary Information, Fig. S1) show clearly the 26-period InGaN quantum-well structure on the {1101 ¯ } facets, whereas no quantum-well contrast variation was observed on the {0001} facet. Third, EDS measurements (Supplementary Information, Fig. S2) further confirm this difference in shell thickness with a well-defined In signal recorded from the MQW shell on the {1101 ¯ } facets but no detectable In signal from the thinner shell on the {0001} facet. We address further this difference below. Last, we note that contrast variation can also be observed at the intersection between the {1101 ¯ } facets in some MQW nanowire samples. Preliminary high-resolution TEM analysis of a 26MQW sample (Supplementary Information, Fig. S3) suggests that these features arise from stacking faults at this intersection, although further studies will be required to understand fully this interesting structural feature. To understand further the MQW growth differences on the two {1101 ¯ } lateral facets versus the {0001} facet of the nanowire, we carried out convergent-beam electron diffraction (CBED). Comparison of experimental and simulated18 (see the Methods section) CBED patterns (Fig. 1e) suggests that the {0001} bottom facet of the MQW nanowire is an N-polarity face. Interestingly, previous studies of planar growth have reported that an N-face has a much slower growth rate than a Ga-face19, and could thus explain the difference in shell growth for our nanowires. Given the more rapid and controlled growth on the {1101 ¯ } nanowire facets, we have focused on growth of multiple-quantum-well structures on these facets to explore limits of synthetic control and to prepare nanolaser structures in which the gain medium is tuned. Notably, by varying the growth time and temperature and total repeat periods during the MQW deposition (see the Methods section), we were able to vary the InGaN/GaN quantum-well unit thicknesses, the In composition and total quantum-well numbers. Cross-sectional STEM images of three distinct MQW nanowires used in our subsequent studies (Fig. 1f–h) show structures with resolvable 3, 13, and 26 periods, quantum-well thicknesses of about 2.4, 3.0 and 1.5 nm and average GaN barrier thicknesses of about 40, 10 and 1 nm, respectively. The In composition control was confirmed by optical studies to be discussed later. We have further investigated the detailed structure of the InGaN/GaN quantum-well repeat by acquiring high-resolution bright-field images from an independent 26MQW nanowire cross￾section sample. A lattice-resolved image of the 26MQW structure on the {1101 ¯ } nanowire facet (Fig. 2a) reveals homogeneous quantum-well deposition as discussed above. High-magnification data of several quantum wells (Fig. 2b) reveal that the InGaN/GaN interfaces are atomically sharp, with the InGaN quantum-well thickness estimated to be 61 nm and approaching the growth limit of several monolayers. In addition, two-dimensional Fourier transforms from the MQW region (Fig. 2b, inset) exhibit no splitting of reciprocal lattice peaks, suggesting that there is no strain relaxation at the InGaN/GaN heterointerfaces. These TEM results confirm that we can prepare dislocation-free MQW nanowire heterostructures with InGaN quantum-well thicknesses similar to or smaller than the Bohr radius, ∼3 nm, for InGaN (ref. 20). We have characterized the optical properties of the new core/MQW-shell nanowire heterostructures using photoluminescence and modelling studies. Photoluminescence images and spectra from individual 26MQW nanowires containing different In compositions (Fig. 3a,b) reveal several points. First, the InGaN emission peak redshifts from 382 to 440 to 484 nm (Fig. 3b) with increasing In composition. The InGaN quantum-well compositions estimated21 from the emission data, In0.05Ga0.95N, GaN InGaN/GaN MQW GaN a b Figure 2 High-resolution TEM of nanowire MQWs. a, Bright-field TEM image of a typical 26MQW nanowire cross-section. The dashed line indicates the heterointerface between the core and shell. The scale bar is 10 nm. b, Enlarged TEM image of MQWs in a. White arrows highlight InGaN quantum-well positions. The dashed line outlines the interface between an InGaN quantum well and adjacent GaN quantum barriers. The scale bar is 2 nm. Inset: Two-dimensional Fourier transforms of the entire image. In0.16Ga0.84N and In0.23Ga0.77N, are consistent with that expected based on our previous studies of InGaN shell growth22. The peak broadening with In% is consistent with InGaN planar structures23 . Second, the photoluminescence images show strong emission from the nanowire bodies where excitation occurs, and also from the nanowire ends, which indicates good waveguiding in the core/MQW-shell structure. Third, the images also show that most In-rich nanowires exhibited spontaneous bending. We attribute this behaviour to the non-symmetrical shell coating, where larger strains are generated on {1101 ¯ } versus (0001) facets. It may be ¯ interesting in the future to quantify the relationship between In composition and bending in these structures, and thus the possible strain relaxation through mechanical bending. Last, we note that no impurity or defect-related emission was observed at longer wavelengths in the photoluminescence spectra, consistent with the high degree of structural order determined from our high-resolution TEM analyses. Excitation power-dependent studies of a representative 26MQW nanowire structure were also carried out, as shown in Fig. 3c. At low power density (245 kW cm−2 ), we observed a relatively broad spontaneous emission peak centred at 448 nm. As power density increased, the end emission intensity increased rapidly and became dominant. Above a threshold of 900 kW cm−2 , this spontaneous emission peak collapsed into several well-defined narrow peaks centred around 438 nm. The full-width at half-maximum of these peaks was less than 0.8 nm, limited by the spectral resolution of our detection system and indicative of longitudinal modes in the nanowire cavity. To confirm this assumption, we calculated the mode spacing using 1l=l 2 /(2neL) (ref. 24), where l is the emission wavelength, ne is the effective refractive index and L is the nanowire cavity length. For l=438 nm, ne = 2.625 (ref. 25) and L = 50 µm, the predicted mode spacing is around 0.73 nm, which is consistent with the measured mode spacing 0.72–0.82 nm. These features together with the observed characteristic nonlinear spectrally integrated output power versus pump power density (Fig. 3c, inset) are clear indications of lasing behaviour26. Comparison of the measured output power data with a fit derived from the rate equations (see the Supplementary Information) yields a spontaneous emission factor, β, of 0.03. This value is consistent with that reported for InGaN MQW lasers27. The blueshift of lasing peaks relative to the spontaneous emission is a distinct feature of InGaN MQW lasers, which has been attributed to band filling and/or photo-induced screening of internal electric fields22,28,29 . nature materials VOL 7 SEPTEMBER 2008 www.nature.com/naturematerials 703 © 2008 Macmillan Publishers Limited. All rights reserved