Hyperspectral Imaging of Exciton Photoluminescence in Individual

Aug 26, 2014 - The separation of Xb and Xd is given by the zero-field exchange splitting, ... At low excitation powers and low temperature, the higher...
6 downloads 0 Views 4MB Size
Download PDF
Letter pubs.acs.org/NanoLett

Hyperspectral Imaging of Exciton Photoluminescence in Individual Carbon Nanotubes Controlled by High Magnetic Fields Jack A. Alexander-Webber,† Clement Faugeras,‡ Piotr Kossacki,‡,§ Marek Potemski,‡ Xu Wang,† Hee Dae Kim,† Samuel D. Stranks,† Robert A. Taylor,† and Robin J. Nicholas*,† †

Department of Physics, Clarendon Laboratory, Parks Road, Oxford, OX1 3PU, United Kingdom Laboratoire National des Champs Magnetiques Intenses, CNRS-UJF-UPS-INSA, F-38042 Grenoble, France, and § Institute of Experimental Physics, University of Warsaw, 00-927 Warsaw, Poland ‡

S Supporting Information *

ABSTRACT: Semiconducting carbon nanotubes (CNTs) provide an exceptional platform for studying one-dimensional excitons (bound electron−hole pairs), but the role of defects and quenching centers in controlling emission remains controversial. Here we show that, by wrapping the CNT in a polymer sheath and cooling to 4.2 K, ultranarrow photoluminescence (PL) emission line widths below 80 μeV can be seen from individual solution processed CNTs. Hyperspectral imaging of the tubes identifies local emission sites and shows that some previously dark quenching segments can be brightened by the application of high magnetic fields, and their effect on exciton transport and dynamics can be studied. Using focused high intensity laser irradiation, we introduce a single defect into an individual nanotube which reduces its quantum efficiency by the creation of a shallow bound exciton state with enhanced electron−hole exchange interaction. The emission intensity of the nanotube is then reactivated by the application of the high magnetic field. KEYWORDS: Carbon nanotube, PFO, Aharonov−Bohm effect, exciton

S

ince their discovery 1 the exceptional optical2 and electronic3 properties of single-walled carbon nanotubes (CNTs) have been of particular interest for the study of fundamental physics as well as for potential applications.4,5 The band structure of CNTs is calculated from the graphene band structure using the zone-folding approximation and may be semiconducting or metallic depending on the direction and diameter of folding (chirality).6 Photoluminescence (PL) and photoluminescence excitation (PLE) spectroscopy experiments on semiconducting CNTs allow the study of a rich variety of one-dimensional (1D) exciton physics. The narrow diameter of CNTs leads to strongly bound excitons due to the large Coulomb interactions.2 The electron and hole spin combined with the two degenerate valleys at the K and K′ points give rise to just one optically active (bright) state out of a total of 16 possible exciton states:7 12 spin-triplet states which are all optically inactive (dark) due to nonzero net spin, and a manifold of 4 spin-singlet states. Of the spin-singlet states the two intervalley (|KK′⟩ and |K′K⟩) states are also optically inactive due to nonzero momentum, and the remaining two intravalley states (|KK⟩ and |K′K′⟩) are coupled into evenparity (dark) and odd-parity (bright) states where a Coulombinduced splitting is expected between the bright state and the lower energy dark state.7 Symmetry-breaking effects, such as the application of a magnetic field, are able to brighten some © 2014 American Chemical Society

previously dark states. The presence of a number of low-lying dark states is important for potential applications of CNTs in optical devices as such states dramatically reduces the PL efficiency of an isolated CNT.8 A magnetic flux threading the axis of a CNT modulates the band gap by adding an Aharonov−Bohm (A−B) phase to the circumferential component of the electron wave function which lifts the valley degeneracy.7,9 Consequently the previously dark even parity intravalley spin-singlet exciton state becomes optically active, and the odd−even parity energy splitting can be observed directly.7 The zero-field splitting energy Δx between the dark and the bright exciton states is determined by the strength of the electron−hole exchange interaction.10 While theoretical predictions range from a few to tens of millielectron volts,7,11−13 experimental values of Δx = 1−8 meV have been reported from low-field microphotoluminescence (μPL) of individual CNTs14−16 and from high-field ensemble measurements.17−24 The strong interaction of CNTs with the local environment has led several recent studies of optical properties to focus on the role of extrinsic effects such as surface charge trapping and Received: June 8, 2014 Revised: August 20, 2014 Published: August 26, 2014 5194

dx.doi.org/10.1021/nl502016q | Nano Lett. 2014, 14, 5194−5200

Nano Letters

Letter

defects.25−31 Here we demonstrate that by studying the magnetooptical properties of individual CNTs it is possible to identify both the intrinsic and defect induced states and to create these defect states with high intensity optical excitation. Results. Individual semiconducting CNTs were selectively dispersed onto a quartz substrate after wrapping with the polyfluorine polymer PFO.32 Here we perform hyperspectral imaging which combines high spatial resolution x,y scanning with a complete high-resolution emission spectrum taken at each spatial position. The PL excitation wavelength was initially set to 720 nm to select the predominantly (8,6) nanotubes33 and isolated single nanotubes which emit at a single wavelength were selected for detailed study as described in the Supporting Information (SI). Figure 1 is a comparative study of ensemble

K) are shown in Figure 1b. The line width of the E22 absorption remains broad, ∼50 meV, due to ultrafast (130 fs34) depopulation of the E22 van Hove singularities, whereas the fwhm of E11 narrows by typically 2 orders of magnitude. The narrowing is a particular property of emission from individual CNTs despite the fact that the ensemble of CNTs with the same chirality shows strong inhomogeneous broadening.36 Using a spatial PL map taken on a 225 × 225 μm2 area we have analyzed the distribution of emission energies from 156 (8,6) CNTs. All of the individual CNTs showed single narrow emission lines, but a histogram of all the emission energies is centered around the peak observed from solution measurements with a width corresponding to an inhomogeneous distribution of ∼20 meV. A few individual tubes were selected for high resolution studies, several of which demonstrated exceptionally narrow E11 emission line widths down to a resolution limited 76 μeV fwhm (Figure 1c), with no evidence that the emission wavelength varies as the excitation is scanned along the length of the tube. Such narrow line widths correspond to an E11 exciton lifetimebroadening of over 30 ps13 and have only previously been observed recently in the highest quality suspended CVD CNTs.37 We attribute the increased lifetimes to the use of (PFO) wrapping and low dispersion densities which electronically decouple the CNTs from the local environment and each other, providing a homogeneous dielectric environment along the nanotube. The line width is improved by an order of magnitude compared to previous μPL studies of solutionprocessed CNTs14,35 and opens the way to new studies of exciton properties for applications such as quantum information processing.38,39 The very large difference between the single tube emission and inhomogeneous line widths observed in the ensemble suggests that the mechanism for light emission in CNTs must be associated with long-lived excitons in a low energy state and that any individual nanotube has sufficiently long exciton lifetimes and diffusion lengths to emit from probably only a single emission site governed by the local energy landscape of each individual tube. Introducing a magnetic flux (B) threading the CNT lifts the K, K′ degeneracy through the A−B splitting and alters the composition of the exciton states7 so that the odd-parity spin singlet bright state, Xb, mixes with the previously dark evenparity state, Xd, which also becomes bright. The A−B shift (ΔAB) is given by6 ⎛ 3πd 2B ⎞ ΔAB ⎟⎟ = ±⎜⎜ Eg ⎝ 4ϕ0 ⎠

Figure 1. PLE spectroscopy. Photoluminescence excitation spectroscopy of (a) an ensemble of (8,6) CNTs in solution at room temperature and (b) individual (8,6) CNTs at 4.2 K. (b, inset): Energy distribution of 156 (8,6) CNTs taken from a 225 × 225 μm2 scanning spatial PL map. (c) An ultranarrow example from an individual CNT with fwhm of 76 μeV collected in a 20 s acquisition. (c, inset) Experimental configuration.

(1)

where ϕ0 (=h/e) is the magnetic flux quantum, Eg (= 0.79 eV for (8,6) CNTs) is the single-particle band gap,23,40 d (= 0.966 nm for an (8,6) CNT33) is the diameter of the CNT, and ± correspond to the two valleys, respectively. The separation of Xb and Xd is given by the zero-field exchange splitting, Δx, and at high fields the two states split linearly up and down by ± ΔAB. Including the exciton exchange energy in a simple two-level model gives Eb and Ed, for the two exciton states, as14

and individual CNTs using PLE spectroscopy which probes the E22 transition in absorption and the E11 emission from the bright odd-parity spin singlet (KK + K′K′), which we label as Xb. Figure 1a is the PLE map of a solution of PFO-wrapped (8,6) CNTs in o-xylene at room temperature. The fwhm of E22 and E11 are 53 and 29 meV, respectively. For comparison, PLE maps of several individual (8,6) CNTs at low temperature (4.2

E b,d = E0 ± 5195

Δ2x + 4Δ2AB 2

(2)

dx.doi.org/10.1021/nl502016q | Nano Lett. 2014, 14, 5194−5200

Nano Letters

Letter

resolved A−B splittings in most CNTs at fields B < 1 T with a consistent value of Δx = 2 ± 0.4 meV for all of the (8,6) CNTs studied. At low excitation powers and low temperature, the higher energy Xb becomes energetically unfavorable, and the previously dark Xd is dominant for B > 3 T (Figure 2a). Ensemble measurements have deduced values for (8,6) chirality tubes of Δx = 2.6 meV from temperature-dependent intensity,21 4 meV from magnetic brightening,23 and 7.9 meV from high field PL.19 Direct observations of Δx for individual CNTs in previous measurements show large variability from tube to tube by almost an order of magnitude.14,15 It has also been shown that Δx can be strongly dependent on the local environment of the CNT with an increase of ∼30% in strained CNTs embedded in polymer films.23 We attribute much of this variability to the influence of more strongly bound excitons, as discussed below. Increasing the excitation power density to 0.5 mW/μm2 effectively heats the CNT42 to ∼30 K (see SI), enabling the bright state to be observable up to B > 8 T as shown in Figure 2b. The A−B shift follows the predicted magnetic field dependence up to 30 T, with ΔAB = ± 0.35 meV/T consistent with theoretical predictions and ensemble measurements17,23,24 although there is some evidence of an increase in Δx due to the carrier heating. The overall consistency of the data and their agreement with theory lead us to conclude that this behavior is characteristic of the free exciton states which are essentially unaffected by being confined in slowly varying local potentials. In addition to that shown in Figure 2a, many CNTs exhibit a more complex high field behavior. By 1 T the Xd exciton is brightened but as the magnetic field is increased, further low energy peaks often become activated, suggesting that they have a larger value of Δx. The magnetic field at which each peak becomes activated is linearly dependent on the energy splitting from Xb, consistent with the condition ΔAB ≃ Δx/2 as shown for one example in Figure 2c with further examples in the SI. Ando7 has predicted that Δx should be very strongly dependent on the short-range contributions to the Coulomb interaction, and the states are sufficiently sensitive that the potential and wave functions are even capable of changing the ordering of the states. Therefore, we attribute the multitude of additional emission peaks at high magnetic fields to the magnetic brightening of previously dark impurity states associated with segments where the value of Δx has been increased due to local changes in the Coulomb interaction caused by changes in dielectric function and trapped charges. The predicted magnetic field dependent energy and emission intensity are plotted in Figure 2e for Δx = 2, 10, and 15 meV and give a very similar dependence to experiment. A common feature of highresolution single CNT PL imaging is the variability of the total emission intensity along the tube, where some segments emit strongly while in other segments the emission is quenched.28,43 Raman spectroscopy and imaging44 and AFM45 on the same nanotubes have been used to attribute this variability to the presence of defects and trapped charges. In order to investigate this hypothesis further, we performed high-resolution spatial mapping in high magnetic fields and deliberately modified the nanotube environment as described below. For CNTs with a length longer than the resolution of the scanning μPL (∼1 μm) interesting variations of emission are observed along the tube. Figure 3 shows the spatial dependence of emission from what we assume to be a single (8,6) nanotube of about 10 μm in length which shows a single bright emission

shown schematically in Figure 2 with the main transfer of intensity between the two states occurring by ΔAB = Δx/2. The

Figure 2. Magnetic field dependent emission from an isolated (8,6) CNT at 4.2 K using low excitation power [0.1 mW/μm2] (a) and high excitation power [0.5 mW/μm2] (b). White dashed lines indicate the fitting with eq 2. (Inset) Cross sections showing the evolution of the two states to high fields. (c) Magnetic field dependent emission spectra from an (8,6) tube, which shows a single bright emission peak at zero magnetic field. At increased magnetic field further emission peaks are activated at low energy. The field required to activate each state is linearly dependent on the magnetic field at 0.5 T/meV (yellow dashed line). (d) Example spectra taken at 0 T and at 29 T. (e) Theoretical magnetic field dependence of the emission intensity and energy of Xb (blue) and Xd (red) with three values of Δx, where the thermal population of the two states assumes a laser-induced heating to 10 K. Note that the intensity of Xb has been increased by 10× for clarity. Schematic: Energy level diagrams for Xb and Xd with and without field, blue and orange represent optically active and inactive, respectively.

actual emission intensity of the two states is determined by the flux dependence of the oscillator strengths714 and the Boltzmann distribution (see the SI). Figure 2a shows a typical magnetic field dependence of the PL from an isolated (8,6) CNT in the Voigt geometry excited with a relatively low power density of 0.1 mW/μm2, which is thought to heat the CNT by a few degrees to ∼10 K. The minimum exciton occupancy within the CNT is estimated to be ∼2, calculated using a typical illuminated spot size of 1 μm, the CNT absorption cross section reported by Schoppler et al.41 and 30 ps exciton lifetime. At zero field a single bright emission peak is observed at ∼1.044 eV. As the magnetic field is applied, an additional peak appears and brightens at lower energy with a splitting Δx = 2.1 meV. The narrow line widths result in well5196

dx.doi.org/10.1021/nl502016q | Nano Lett. 2014, 14, 5194−5200

Nano Letters

Letter

these peaks with emission from magnetically brightened states with increased values of Δx, quenched at zero field, which now emit strongly. The creation of magnetically activated sidebands from bound excitons in previously dark segments of the tube is shown schematically in Figure 4. All segments along the tube can contribute both red and blue-shifted emission within the wavelength range observed in ensemble measurements (Figure 1). An overall picture of the life of an exciton in a CNT emerges where, with no applied magnetic field, long-lived highly mobile excitons diffuse freely through a slowly varying energy landscape before recombining radiatively as nearly free exciton states unaffected by their environment and in the central region of the nanotubes. Excitons trapped in deeper states or which reach the nanotube ends46,47 can only recombine nonradiatively. In magnetic field more strongly bound states begin to emit and several different emission sites appear along the nanotube emitting at different wavelengths, as shown schematically in Figure 4. We test the picture of multiple bound states by deliberately introducing an additional defect site into a specific nanotube with focused high intensity irradiation from the excitation laser as previously used to induce blinking,27 reduce quantum efficiency,28 and introduce additional deep levels.26,28,29,31 Laser excitation at 720 nm was increased stepwise from 0.1 to 0.6 mW/μm2, and the sample region in Figure 2a was irradiated for 1 min at each power. Emission intensity, at 10 T increased linearly with excitation power until a critical power density above 0.5 mW/μm2 at which the emission dramatically decreased (Figure 5f). When the power was reduced, an additional pair of peaks, one bright and the other dark but magnetically brightened, appear at lower energies labeled Xbirr and Xdirr, shown in Figure 5a,b. The new transitions are also well-described by eq 2 with ΔAB = 0.35 meV/T, but the exchange energy Δx has increased to 7 meV, the midpoint E0 is red-shifted by 10 meV, and the magnetic brightening of the new dark state is considerably greater (Figure 5g). The total emission intensity at high magnetic fields is comparable to the previously undamaged tube. Magnetic activation of the lower lying bound states also explains the high magnetic field (B > 20 T) suppression of intensity observed in some nanotubes23 (Figure 5e and SI), where activation of the lower lying states provides greater competition for the mobile excitons, thus reducing emission from higher lying states. The time dependence of the emission is also very different before and after laser irradiation. In the pristine tube we see blinking and spectral diffusion which suggest the dynamic modification of the exciton states associated with fluctuating local charge in the wrapping polymers. Figure 5c shows the time evolution of emission at 10 T from the pristine CNT studied in Figure 2 taken with 5 s resolution for 500 s. Xd emits at a constant energy of 1.038 eV, but an additional peak is clearly resolved at lower energy, with fluctuating intensity and spectral diffusion, ∼0−10 meV below the dark state. Figure 5e shows that the fraction of the total emission intensity for each of the two peaks is anticorrelated with a Pearson correlation coefficient of r = −0.28 (−1 < r < 1) where the negative sign indicates the anticorrelation. Such behavior has been observed by Matsuda et al.25 in single CNTs, attributed to a quantumconfined Stark effect,48 which causes a red-shifted emission peak due to fluctuating local charges which are photoinduced in the surfactant wrapping the nanotube. Once the tube has been irradiated another time series at 10 T (Figure 5d) shows that

Figure 3. Spatially dependent PL maps at 0 T (a) and 20 T (b) from the same regions on the sample. The direction of the magnetic field and excitation polarization is along the y-axis. Spectra from the map are shown taken from positions indicated by squares on the map. The highlighted wavelengths represents the wavelength selection corresponding to the color in the spatial map. The acquisition times are 20 s per pixel at 0 T and 5 s per pixel at 20 T due to the enhanced emission in the high field.

line at 0T centered in the middle of the CNT and extending about 4 μm. At 20 T a corresponding red-shifted and magnetically brightened emission is observed for the same positions, but a second magnetically activated emission is also observed with energy 5 meV higher than the first state which originates mainly from the regions either side of the zero field emission. This demonstrates that a large contribution to the magnetic brightening comes from the activation of previously dark regions probably containing spatially localized excitons with a higher exchange energy as described above. In the most dramatic example of magnetic brightening, shown in Figure 4, we study a particular (8,7) chirality CNT which only shows a single bright emission peak at 0 T, when excited at certain positions along the tube. At 17 T multiple sidebands appear around the original 0 T emission. Two peaks in particular, shown in green and blue, are spatially localized on either side of the main 0 T bright emission peak. At zero magnetic field no emission is observed when these sections are excited, whereas magnetic brightening causes them to emit more strongly than the zero field bright state. We associate 5197

dx.doi.org/10.1021/nl502016q | Nano Lett. 2014, 14, 5194−5200

Nano Letters

Letter

Figure 4. Field-activated spatial PL map showing a region which has a single bright emission peak at 0 T (a,c) [orange region] and a multitude of peaks at 17 T (b,d) [orange, magenta, blue, green]. The direction of the magnetic field and excitation polarization is along the x-axis. Example spectra are shown at the regions indicated by squares. Some of these peaks (circled) originate from spatially localized sites which previously showed no emission when excited. Bottom: Schematic energy levels and emission along the CNT with (e) and without (f) applied magnetic field. Sections of the tube where Δx is enhanced are quenched at zero magnetic field.

both peaks Xd and Xdirr still show blinking, but now the two lines have a strong positive correlation (r = 0.83) as shown in Figure 5e. This intensity correlation is strong evidence that both dark states, Xd and Xdirr, originate from the same single nanotube, which has its overall emission controlled by the fluctuations in its nonradiative emission rate due to the Stark effect. The high intensity irradiation clearly has dramatic irreversible effects on the properties of a CNT by introducing a new more strongly bound state. In addition to the predictions by Ando of enhanced Δx values due to short-range Coulomb interactions, Tomio and Suzuura have used effective mass theory for an impurity potential on a scale comparable with the tube diameter but ≳ lattice constant, to predict a new pair of exciton states both red-shifted from the bright and dark free excitons but with the same A−B splitting.49 We can attribute Xbirr and Xdirr to this new pair of bound bright and dark exciton states but with an increased zero-field exchange splitting of Δbound = 7 meV, suggesting that Δx is significantly enhanced at x local defect sites due to an increase in the electron−hole exchange interaction, possibly by the mechanisms suggested by Ando.7 This may explain the variability of splitting found in previous single tube experiments, as the lower quality of tubes studied could lead to spectra dominated by a variety of bound exciton states. The physical origin of such an impurity site may be characterized by the removal of the PFO coating along the tube or the creation of structural defects. Recently Miyauchi et al.30 have shown that deep lower lying exciton states can be induced by oxygen doping of ensembles of (7,5) CNTs, which they attribute to bound excitons located at structural defects. From the temperature dependence of the emission intensity they also deduce that a dark state should exist with an enhanced Δx of order 10 meV, consistent with our observations. Conclusions. We have shown that PFO purification yields isolated semiconducting CNTs with exceptional ultranarrow (76 μeV) emission spectra at 4.2 K. Using hyperspectral imaging we show that excitons diffuse over long distances before reaching either long-lived, shallow states confined in a slowly varying energy landscape or dark states, such as those

with enhanced electron−hole exchange energies. The weakly confined free excitons can be studied at specific locations and have a consistent zero field exchange splitting of Δx = 2 meV and an Aharonov−Bohm splitting of ±0.35 meV/T as deduced from magnetic field brightening of the spin singlet state for (8,6) CNTs. The number of radiative sites can be greatly enhanced in high magnetic fields, and both the local and spectral distribution of the emission changes greatly. Introducing local damage by laser irradiation generates additional magnetically activated bound exciton states with enhanced electron−hole exchange interaction. The long lifetimes and magnetic control of the excitonic emission, combined with the ability to create specific emission sites by optical exposure, gives us both an improved understanding of the exciton transport and dynamics and offers the potential for phase controlled ultranarrow emission sources using fields of only a few Tesla, with applications such as quantum information processing.5 This greater understanding of the origin of quenching segments could lead to maximizing the quantum efficiency of CNT based light-emitting diodes and laser diodes. Methods. A solution of predominantly (8,6), semiconducting CNTs was prepared using polymer wrapping techniques similar to those described in Nish et al.32 using chirality selective polyfluorene (PFO) and raw, dried HiPco CNTs as a starting material. After 100-fold dilution the solution was spincoated onto Quartz substrates at 2000 rpm for 90 s, and annealed in air at 180 °C, giving a distribution of spatially isolated individual CNTs. Typical CNT lengths of 1−10 μm were deduced from atomic force microscopy (AFM) as shown in the SI. Low-temperature microphotoluminescence measurements were taken with the sample immersed in helium gas at 4.2 K (Figure 1c, inset). Continuous wave resonant excitation of E22 transitions for (8,6) tubes was achieved by a tunable Ti:sapphire laser through a monomode fiber with 5 μm diameter core, in the range 700−800 nm. The excitation wavelength used in the magnetic field dependences and scanning maps was set to 720 nm corresponding to the typical 5198

dx.doi.org/10.1021/nl502016q | Nano Lett. 2014, 14, 5194−5200

Nano Letters

Letter

Figure 5. Irradiated PL spectroscopy. (a) Cross sections from (b), the magnetic field dependent emission after irradiation using low intensity excitation. White and orange dashed lines show fits to the original and irradiation induced excitons, respectively. Time series of emission at 10 T before (c) and after (d) high intensity irradiation. (e) Analysis of the emission time series before (red = Stark effect) and after (black = Xdirr) irradiation. The relative intensity of the two dominant peaks are plotted as a function of the intensity of Xd. (f) Intensity of CNT emission at increasing excitation power density; the intensity dramatically reduces above 0.5 mW/μm2. (g) The magnetic brightening before (Xd(B)/Xb, blue) and after irradiation (Xdirr(B)/Xbirr, red), including polynomial fits.

absorption peak observed in Figure 1b. Emission was collected using a 50 μm diameter core fiber and detected by a nitrogen cooled InGaAs array. After passing a beam splitter the excitation was focused on the sample surface with a spot size of ∼1 μm. The sample was manipulated with submicron precision with an Attocube x,y,z-piezostage. The piezostage was used to scan the position of the excitation (step sizes 0.15−0.5 μm), and a spectrum was taken for each pixel to create a detailed map of the CNTs on the sample. These hyperspectral images contain a great deal of information including the chirality of nanotubes, position, bundling, and specific energy levels. Continuous magnetic fields up to 30 T were provided by a 20 MW resistive-coil magnet at the LNCMI, Grenoble. The magnetic field was applied in both the Faraday geometry, where the light propagation vector k is parallel to the magnetic field and perpendicular to the surface of the substrate, and the Voigt geometry, where k is perpendicular to the direction of the magnetic field and the surface of the sample, using a mirror at 45°. In the Voigt geometry, excitation and collection polarized parallel to the magnetic field were used to select CNTs aligned on the substrate in the direction of the magnetic field using the polarization characteristics of the E11 and E22 transitions.50

Scans taken up to 30 T in the Faraday geometry showed no measurable effect on the spectra consistent with the expectation that only effects due to the Aharonov−Bohm flux penetrating the tubes should be observed.

■

ASSOCIATED CONTENT

S Supporting Information *

Large area spatial PL mapping, atomic force microscopy of CNT depositions, additional magnetic field dependent emission spectra, additional example of hyperspectral imaging at high magnetic fields, and theoretical modeling of thermal exciton state population. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Author Contributions

R.J.N. initiated the experiment. J.A.A.-W. prepared the samples with assistance from S.D.S. Preliminary imaging was performed by J.A.A.-W, X.W., H.D.K., and R.A.T. High magnetic field data was collected by J.A.A.-W., R.J.N., C.F., P.K., and M.P. in Grenoble. The data were interpreted by J.A.A.-W. and R.J.N. 5199

dx.doi.org/10.1021/nl502016q | Nano Lett. 2014, 14, 5194−5200

Nano Letters

Letter

who also wrote the first draft of the paper. All authors contributed to its revision and final interpretation.

(40) Chuang, K.-C.; Nish, a.; Hwang, J.-Y.; Evans, G.; Nicholas, R. Phys. Rev. B 2008, 78, 085411. (41) Schoppler, F.; et al. J. Phys. Chem. C 2011, 115, 14682. (42) Bassil, A.; Puech, P.; Tubery, L.; Bacsa, W.; Flahaut, E. Appl. Phys. Lett. 2006, 88, 173113. (43) Georgi, C.; Green, A. A.; Hersam, M. C.; Hartschuh, A. ACS Nano 2010, 4, 5914. (44) Freitag, M.; et al. Appl. Phys. Lett. 2007, 91, 031101. (45) Rauhut, N.; et al. ACS Nano 2012, 6, 6416. (46) Harrah, D. M.; Swan, A. K. ACS Nano 2011, 5, 647. (47) Rajan, A.; Strano, M. S.; Heller, D. A.; Hertel, T.; Schulten, K. J. Phys. Chem. B 2008, 112, 6211. (48) Empedocles, S. A.; Bawendi, M. G. Science 1997, 278, 2114. (49) Tomio, Y.; Suzuura, H. J. Phys.: Conf. Ser. 2011, 302, 012005. (50) Ajiki, H.; Ando, T. Physica B 1994, 201, 349.

Notes

The authors declare no competing financial interest.

■ ■

ACKNOWLEDGMENTS This work was supported by EuroMagNET II under the EU contract No. 228043, and by the EPSRC in the UK. REFERENCES

(1) Iijima, S.; Ichihashi, T. Nature 1993, 363, 603. (2) Avouris, P.; Freitag, M.; Perebeinos, V. Nat. Photonics 2008, 2, 341. (3) Avouris, P. Chem. Phys. 2002, 281, 429. (4) Jorio, A.; Dresselhaus, G.; Dresselhaus, M. S. Carbon Nanotubes: Advanced Topics in the Synthesis, Structure, Properties and Applications; Springer: New York, 2008. (5) Park, S.; Vosguerichianb, M.; Zhenan Bao, Z. Nanoscale 2013, 5, 1727. (6) Ando, T. J. Phys. Soc. Jpn. 2005, 74, 777. (7) Ando, T. J. Phys. Soc. Jpn. 2006, 75, 024707. (8) Charlier, J.-C.; Roche, S. Rev. Mod. Phys. 2007, 79, 677. (9) Ajiki, H.; Ando, T. J. Phys. Soc. Jpn. 1993, 62, 2470. (10) Kono, J.; Nicholas, R. J.; Roche, S. Carbon Nanotubes: Top. Appl. Phys. 2008, 111, 393−421. (11) Spataru, C. D.; Ismail-Beigi, S.; Capaz, R. B.; Louie, S. G. Phys. Rev. Lett. 2005, 95, 247402. (12) Kilina, S.; Badaeva, K.; Piryatinski, A.; Tretiak, S.; Saxena, A.; Bishop, A. R. Phys. Chem. Chem. Phys. 2009, 11, 4113. (13) Perebeinos, V.; Tersoff, J.; Avouris, P. Nano Lett. 2005, 5, 2495. (14) Srivastava, A.; Htoon, H.; Klimov, V. I.; Kono, J. Phys. Rev. Lett. 2008, 101, 87402. (15) Matsunaga, R.; Matsuda, K.; Kanemitsu, Y. Phys. Rev. Lett. 2008, 101, 147404. (16) Matsunaga, R.; Miyauchi, Y.; Matsuda, K.; Kanemitsu, Y. Phys. Rev. B 2009, 80, 115436. (17) Zaric, S.; et al. Science 2004, 304, 1129. (18) Zaric, S.; et al. Phys. Rev. Lett. 2006, 96, 016406. (19) Shaver, J.; et al. Nano Lett. 2007, 7, 1851. (20) Mortimer, I. B.; et al. Phys. Rev. B 2007, 76, 085404. (21) Mortimer, I. B.; Nicholas, R. J. Phys. Rev. Lett. 2007, 98, 027404. (22) Shaver, J.; et al. Phys. Rev. B 2008, 78, 081402. (23) Nish, A.; Nicholas, R. J.; Faugeras, C.; Bao, Z.; Potemski, M. Phys. Rev. B 2008, 78, 245413. (24) Takeyama, S.; Suzuki, H.; Yokoi, H.; Murakami, Y.; Maruyama, S. Phys. Rev. B 2011, 83, 235405. (25) Matsuda, K.; Inoue, T.; Murakami, Y.; Maruyama, S.; Kanemitsu, Y. Phys. Rev. Lett. 2008, 77, 193405. (26) Harutyunyan, H.; et al. Nano Lett. 2009, 9, 2010. (27) Georgi, C.; et al. ChemPhysChem 2008, 9, 1460. (28) Finnie, P.; Lefebvre, J. ACS Nano 2012, 6, 1702. (29) Santos, S.; et al. Phys. Rev. Lett. 2011, 107, 187401. (30) Miyauchi, Y.; et al. Nat. Photonics 2013, 7, 715. (31) Piao, Y.; et al. Nat. Chem. 2013, 5, 840. (32) Nish, A.; Hwang, J.-Y.; Doig, J.; Nicholas, R. J. Nat. Nanotechnol. 2007, 2, 640. (33) Weisman, R. B.; Bachilo, S. M. Nano Lett. 2003, 3, 1235. (34) Lauret, J.-S.; Voisin, C.; Cassabois, G.; Delalande, C.; Roussignol, P.; Jost, O.; Capes, L. Phys. Rev. Lett. 2003, 90, 057404. (35) Kiowski, O.; Lebedkin, S.; Hennrich, F.; Kappes, M. M. Phys. Rev. B 2007, 76, 075422. (36) Htoon, H.; O’Connell, M.; Cox, P.; Doorn, S.; Klimov, V. Phys. Rev. Lett. 2004, 93, 027401. (37) Hofmann, M. S.; et al. Nat. Nanotechnol. 2013, 8, 502. (38) Hogele, A.; Galland, C.; Winger, M.; Imamoglu, A. Phys. Rev. Lett. 2008, 100, 217401. (39) Laird, E. A.; Pei, F.; Kouwenhoven, L. P. Nat. Nanotechnol. 2013, 8, 565. 5200

dx.doi.org/10.1021/nl502016q | Nano Lett. 2014, 14, 5194−5200