Vrouw, Jul 2011
Vrouw, Jul 2011
Vrouw, Jul 2011
ABSTRACT
We report Raman and time-resolved photoluminescence spectroscopic studies of multiwalled BN and BxCyNz nanotubes. The Raman spectroscopy shows that the as-grown BxCyNz nanotubes are radially phase separated into BN shells and carbon shells. The photoluminescence decay process is characterized by two time constants that are attributed to intra- and inter-BN sheet charge recombination, respectively. A comparison of the photoluminescence of BN nanotubes to that of hexagonal BN is consistent with the existence of a spatially indirect band gap in multiwalled BN nanotubes as predicted by theory.
Hexagonal boron nitride (h-BN) is isostructural with graphite and is the normal phase of BN that is stable at room temperature and ambient pressure. It is a wide band gap semiconductor with potential applications in optoelectronic devices. First principles local-density calculations have shown that the lowest band gap (4.07 eV) is indirect, located near the Brillouin-zone edges. However, because of the quasi-2D nature of the hexagonal structure, the lowest direct band gap is predicted to be close by at 4.2 eV.1 A wide range of lowest gap energies, such as 3.8,2 4.3,3 5.2,4 and 5.9 eV,5 have been reported by experimental studies using various techniques. When a hexagonal BN sheet is rolled up into a singlewalled BN nanotube (NT), its electronic structure is very different from that of carbon nanotubes because of the wide gap of BN sheets as opposed to the zero gap in graphite. In contrast to carbon nanotubes, which have diverse electronic properties depending on their structure, the band gaps of BN NTs are predicted theoretically to be approximately equal to that of h-BN, nearly independent of their radius and chirality, except for a band gap narrowing in thin NTs (diameter < 0.8 nm) as a result of sp3 () hybridization
* Corresponding author. E-mail: [email protected]. Materials Sciences Division, Lawrence Berkeley National Laboratory. Department of Physics, University of California. Molecular Foundry, Lawrence Berkeley National Laboratory. | Department of Materials Science and Engineering, University of California. 10.1021/nl049862e CCC: $27.50 Published on Web 03/18/2004 2004 American Chemical Society
induced by the high curvature.6 In this context, the predicted structural independence of the electronic properties and thermal stability of BN NTs open promising opportunities for applications in nanoelectronic and nanophotonic devices. The optical properties of BN NTs are still not well understood because of poor sample quality and low lightemission efficiency. Recently, it was demonstrated that multiwalled BN and boron carbon nitride (BxCyNz) NTs can be synthesized in large quantities using a carbon nanotube-substituted reaction.7 Transmission electron microscopy (TEM) shows that highquality, nearly perfect layer-structured BN or BxCyNz NTs can be grown in either individual or bundled form. In this letter, we present the results of optical spectroscopy studies of such BN and BxCyNz NTs. Raman spectroscopy and continuous-wave and time-resolved photoluminescence have been carried out, and the results are compared to recent theoretical calculations. BxCyNz nanotubes were synthesized via the carbon nanotube-confined reaction using chemical vapor depositionderived carbon NTs together with B2O3 and N2 at 1600 C for half an hour. Subsequent thermal oxidation treatment at 650 C efficiently transformed the BxCyNz NTs into BN NTs. Details of sample synthesis and oxidation processes have been published elsewhere.7 The resulting product is a mixture of 60-70% multiwalled BxCyNz (or BN after oxidation) NTs and 30-40% BxCyNz (or BN) fullerene-like particles
Figure 2. Raman spectra of BN and BxCyNz nanotubes and bulk h-BN recorded at room temperature. Inset shows the main peak near 1366 cm-1.
Figure 1. TEM image of (a) BxCyNz nanotubes, (b) BN nanotubes, and (c) h-BN powder. Inset is the high-resolution TEM image.
(mostly), nanoparticles, and nanowires. Electron energy loss spectroscopy (EELS) shows that, prior to oxidation, the C/B atomic ratio is between 0.2 and 0.4 and that after oxidation it is reduced to <0.02. Most NTs have outer diameters below 8 nm and about two to eight walls, as demonstrated by the high-resolution TEM images shown in Figure 1a and b. After the removal of carbon by oxidation from the BxCyNz NTs, the average number of NT layers is reduced, and the walls of the NTs still appear to be free of large defects, suggesting that the original BxCyNz NTs are radially phase separated into BN shells and carbon shells. Similar results were reported by Suenaga et al. on BxCyNz NTs made by the arc-discharge method.8 The Raman scattering spectra were excited with a frequency-doubled (244-nm) argon laser and recorded with a liquid-nitrogen-cooled CCD camera. The continuous-wave photoluminescence (PL) spectra were generated in the backscattering geometry by excitation with the same laser and were detected by a photomultiplier tube. The timeresolved PL was excited with 200-fs pulses from a frequencytripled (267-nm) mode-locked Ti:sapphire laser, dispersed with a 0.25-m single-grating monochromator, and detected with a synchroscan streak camera. The system spectral and temporal resolution were 0.05 nm and 11 ps, respectively. For all of the experiments, commercial hexagonal BN powder (grain size, several micrometers; see Figure 1c) was also measured as a comparison. All experiments were performed at room temperature. Figure 2 shows the Raman spectra of h-BN, BxCyNz, and BN NTs in the wavenumber range of interest. The dominant peak near 1366 cm-1 that shows up intensely in all three samples is attributed to the well-known zone-center, counterphase B-N vibrational mode (E2g) within BN sheets.4,9 As
648
shown in the inset of Figure 1, this E2g mode is broadened and shifted to higher wavenumbers in the BN and BxCyNz NTs. The peak shift is about 2.1 cm-1, and the FWHM increases from 15.6 cm-1 in h-BN to 17.9 cm-1 in the BN and BxCyNz NTs. Nemanich et al. have reported a correlation between the width and peak energy of this mode and the average crystal grain size of the h-BN sample.10 The shift and broadening were explained in a formulation of the Raman cross section for scattering from nanocrystals in which the wave-vector uncertainty of the phonons is related to the crystal grain size. The shift of this peak from its bulk value can be described by 380 10-8/L - 0.29 cm-1 and its FWHM can be described by 1417 10-8/L + 8.7 cm-1, where L is the average crystal size.10 From these quantitative relationships, we estimate the effective crystal size to be 16 nm for our BN and BxCyNz NTs. This value is close to the average perimeter of the NTs in our sample. The signal may also include contributions from similarly sized hexagonally structured nanoparticles. As compared to the Raman spectrum of BN NTs, the BxCyNz sample shows a strong peak near 1585 cm-1 with a FWHM of 48 cm-1. This peak is clearly due to the E2g stretching mode of a 2D graphite sheet or carbon shells in the BxCyNz sample.11,12 The higher energy of this mode compared to that of the same mode of BN reflects the softening of the bond as a result of ionicity in BN. The large width may be a result of a superposition of multiple peaks due to differently structured carbon shells because this mode is chirality- and diameter-dependent in carbon nanotubes. The emergence of this C-C E2g peak and the absence of any other new peaks associated with B-C or N-C bonds from the BxCyNz sample (as compared to the BN sample) confirm the conclusion drawn from the TEM studies that the BxCyNz NTs are structured in the form of radially separated BN shells and carbon shells. The fact that the carbon shells do not survive the oxidation process indicates that the majority of them form the outer or the inner layers of the multiwalled structure rather than being sandwiched between BN shells.
Nano Lett., Vol. 4, No. 4, 2004
Figure 3. Continuous-wave photoluminescence spectra obtained at room temperature. For the BxCyNz NTs, the dashed curve shows the spectrum recorded when the excitation laser was focused on a dark grain in the sample. Inset shows the phonon replica energies plotted against the number of phonons emitted.
There are also some other weaker peaks in the spectra. A broad peak appears at 920 cm-1 for all three samples. Because in a perfect h-BN crystal the only other Ramanactive mode (in-phase E2g mode) is located far away (near 50 cm-1)4, this peak at 920 cm-1 is probably due to disorderinduced Raman activity or contamination. A sharp peak is observed near 880 cm-1 for the BxCyNz and BN NTs. This peak has been reported in studies of BN single-walled NTs,13 but its origin has not been identified. The room-temperature PL spectra of these samples are shown in Figure 3. All samples show a wide emission peak between 3.5 and 4.2 eV, tailing off toward low energy. The signal is about 100 times stronger for the h-BN sample than in the other samples. Phonon replica features visible on the broad PL peaks are especially pronounced for the h-BN sample. In the inset of Figure 3, we plot the peak of each phonon replica against the number of peaks. The slope of the linear dependence is 0.17 eV (1370 cm-1) per phonon, which is close to the zone-center E2g phonon energy of the BN sheet. The extrapolation of the linear dependence in the inset of Figure 3 leads to a zero-phonon line (ZPL) of 4.07 eV for the h-BN and 4.02 eV for the BN and BxCyNz NTs. This value is close to the direct band gap (4.2 eV at zero temperature) predicted in ref 1. We have also studied the excitation power dependence of the PL peaks. A linear increase in the emission intensity was observed for all three samples when the excitation laser power was changed by more than 3 orders of magnitude (not shown here). The absence of saturation in emission intensity at strong excitation and the zone-center phonon replica feature indicate that the emission is indeed caused by band-to-band optical transitions across the direct band gap in the samples. Both the BN NT and the BxCyNz NT sample appear to be white in general, but large, darker grains (size 0.5 mm) are visible to the naked eye in the BxCyNz sample. When the excitation laser is focused on these darker grains, the sample exhibits a stronger carbon E2g Raman peak and a much
Nano Lett., Vol. 4, No. 4, 2004
weaker PL peak, as shown by the dashed curve in Figure 3. Therefore, it can be concluded that the BxCyNz NT sample is inhomogeneous and that the darker areas contain more carbon. Because of the chirality-independent band gap of singlewalled BN nanotubes and the weaker coupling between adjacent BN shells, the electronic structure of multiwalled BN nanotubes should be also less diverse than that of multiwalled carbon NTs. However, the - hybridization in high-curvature BN shells that reduces the band gap of small-diameter single-walled BN NTs also plays an important role in multiwalled BN NTs. Local density approximation calculations by Okada et al. have shown that in double-walled BN NTs both the conduction and valence bands of the inner shell with larger curvature are negatively shifted (type-II offset) from those of the smaller-curvature outer shell as a result of their stronger hybridization with remote bands.14 Consequently, in the double-walled structure a spatially indirect band gap is developed in which the conduction-band minimum is located on the inner shell and the valence-band maximum is on the outer shell. The fundamental band gap is thus smaller than that of the inner shell by an amount equal to the band offset. For example, the band gaps of (5, 0), (7, 0), (9, 0), and (15, 0) single-walled BN NTs are calculated to be 2.4, 3.4, 3.8, and 4.4 eV, respectively. When these NTs are encapsulated in a (15, 0) BN shell, the band gaps are reduced to 1.7 eV for (5, 0)@(15, 0), 2.9 eV for (7, 0)@(15, 0), and 3.6 eV for (9, 0)@(15, 0). The effect is weaker for larger-diameter NTs. The slight downward shift (50 meV) of the ZFL in Figure 3 from h-BN to BN NTs could be caused by these curvature-induced effects in the multiwalled NTs. The shift is weak, which could be due to the large diameter of the NTs in our sample: the maximum outer-shell diameter of 8 nm corresponds to a (90, 0) zigzag NT or a (50, 50) armchair NT. To study the recombination mechanism of the emission, we have carried out time-resolved PL by excitation with a laser line at 267 nm. Figure 4a shows the PL decay profile of the BN NTs at wavelengths of 315 nm (3.94 eV), 335 nm (3.70 eV), and 355 nm (3.49 eV). Although a stretched exponential function can also fit the spectra, we chose to fit these spectra using the physically simpler double-exponential decay function over the entire time range of detection (2 ns). The PL decay process is characterized by a fast decay at the beginning with a time constant of 1 200 ps followed by a slower decay with 2 > 1500 ps. In Figure 4b, the two lifetimes obtained are plotted as a function of the emission energy. It can be seen that both lifetimes increase on the lower-energy shoulder of the PL peak. This is typical band-edge luminescence behavior in which the shorter lifetime of higher-energy carriers is attributed to the additional relaxation channel via thermalization to lower-energy states.15 The temporal behavior of the emission from the BxCyNz NTs is quantitatively similar to that of the BN NTs (not shown here). As a comparison, Figure 5 shows the lifetime plot of the h-BN sample. The plot shows qualitatively similar behavior but with several differences. Lifetime 2 is longer in the BN
649
sheets and between different BN shells/sheets, respectively. As the h-BN sheets form multiwalled NTs, 1 is reduced because point defects induced on BN shells act as additional recombination centers. Lifetime 2 is increased as a result of the spatial separation of electrons and holes. In both the BN NT and the h-BN samples, the intensity of the 1-related emission is larger by a factor of 5-10 than the 2-related emission. This indicates that the charge relaxation and recombination processes are predominantly occurring within individual hexagonal BN shells in the multiwalled NT structure. This behavior is consistent with the recent observation of strongly delocalized plasmons along the tube axis in multiwalled BN NTs as studied by EELS.16 In summary, we have observed the strong photoluminescence of multiwalled BN and BxCyNz nanotubes at room temperature. A temporal analysis of the PL signals reveals a charge recombination process dominated by the fast recombination occurring within individual BN sheets. The slower decay process is attributed to the charge transfer and recombination across different BN sheets and shows behavior that is consistent with the predicted spatially indirect band gap of multiwalled BN NTs. Results of Raman spectroscopy confirm that the multiwalled BxCyNz NTs are structured into radially phase-separated BN and carbon shells.
Figure 4. (a) Time-resolved photoluminescence of BN NTs. The wavelength window for each curve is 5 nm. The solid curves show the best fit using a double-exponential function. (b) Lifetimes obtained from the fitting with the double-exponential decay function.
Acknowledgment. This work is supported by the Director, Office of Science, Office of Basic Energy Sciences, Division of Materials Sciences and Engineering of the U.S. Department of Energy under contract no. DE-AC0376SF00098. W.-Q. H . and A.Z. acknowledge support from the Molecular Foundry, LBNL. References
(1) Xu Y.; Ching, W. Y. Phys. ReV. B 1991, 44, 7787-7798. (2) Rand M. J.; Roberts, J. F. J. Electrochem. Soc. 1968, 115, 423427. (3) Zupan J.; Kolar, D. J. Phys. C 1972, 5, 3097-3100. (4) Hoffman, D.; Doll G. L.; Eklund, P. C. Phys. ReV. B 1984, 30, 60516056. (5) Tarrio C.; Schnatterly, S. E. Phys. ReV. B 1989, 40, 7852-7859. (6) Rubio, A.; Corkill, J. L.; Cohen, M. L. Phys. ReV. B 1994, 49, 50815084. Blase, X.; Rubio, A.; Louie, S. G.; Cohen, M. L. Europhys. Lett. 1994, 28, 335-340. (7) Han, W. Q.; Mickelson, W.; Cumings, J.; Zettl, A. Appl. Phys. Lett. 2002, 81, 1110-1112. (8) Suenaga, K.; Collix, C.; Demoncy, N.; Loiseau, A.; Pascard, H.; Willaime, F. Science 1997, 278, 653-655. (9) Popov, V. N. Phys. ReV. B 2003, 67, 085408. (10) Nemanich, R. J.; Solin, S. A.; Martin, R. M. Phys. ReV. B 1981, 23, 6348-6356. (11) Rao, A. M.; Richter, E.; Bandow, S.; Chase, B.; Eklund, P. C.; Williams, K. A.; Fang, S.; Subbaswamy, K. R.; Menon, M.; Thess, A.; Smalley, R. E.; Dresselhaus, G.; Dresselhaus, M. S. Science 1997, 275, 187-191. (12) Eklund, P. C.; Holden, J. M.; Jishi, R. A. Carbon 1995, 33, 959972. (13) de la Concha, R. A.; Wirtz, L.; Mevellec, J. Y.; Lefrant, S.; Rubio, A.; Loiseau, A. AIP Conf. Proc. 2003, 685, 384-388. (14) Okada, S.; Saito, S.; Oshiyama, A. Phys. ReV. B 2002, 65, 165410. (15) Kim, H. S.; Mair, R. A.; Li, J.; Lin, J. Y.; Jiang, H. X. Appl. Phys. Lett. 2000, 76, 1252-1254. (16) Fuentes, G. G.; Borowiak-Palen, E.; Pichler, T.; Liu, X.; Graff, A.; Behr, G.; Kalenczuk, R. J.; Knupfer, M.; Fink, J. Phys. ReV. B 2003, 67, 035429.
Figure 5. Lifetimes of h-BN obtained from fitting with the doubleexponential function.
NTs than in the h-BN sample by more than 50% in the main peak region. This effect is tentatively attributed to the spatially indirect nature of the band gap in the multiwalled BN NTs. Similar to type-II offset planar quantum wells, the recombination probability of electrons and holes is reduced as a result of the spatial separation of their wave function, which increases the lifetime of the PL. However, lifetime 1 is shorter (by 40%) in the BN NTs. Therefore, we can speculate that the decay processes associated with lifetimes 1 and 2 involve recombination within single BN shells/
650
NL049862E
Nano Lett., Vol. 4, No. 4, 2004