NANO LETTERS
Electronic Structure of Subnanometer Diameter MoS2−Ix Nanotubes
2004 Vol. 4, No. 9 1749-1753
Anna Zimina,† Stefan Eisebitt,*,† Mirko Freiwald,† Stefan Cramm,‡ Wolfgang Eberhardt,† Ales Mrzel,§ and Dragan Mihailovic§ BESSY m.b.H., Albert-Einstein-Str. 15, 12489 Berlin, Germany, Forschungszentrum Ju¨lich GmbH, 52425 Ju¨lich, Germany, and Jozef Stefan Institut, JamoVa 39, SI-1000 Ljubljana, SloVenia Received June 17, 2004; Revised Manuscript Received August 2, 2004
ABSTRACT The local electronic structure in subnanometer diameter MoS2−Ix nanotubes is investigated by soft X-ray absorption and X-ray fluorescence spectroscopies. The nanotubes are found to be semiconducting, with a band gap of 0.4 ± 0.2 eV. The density of electronic states is altered in both the conduction and the valence band as compared to 2H−MoS2. Electronic structure calculations based on simple model structures suggest that the changes can be mainly attributed to the effect of Mo−S−Mo bond angle distortion imposed by the small diameter of the nanotubes.
The discovery of a new form of carbon in the form of fullerenes and nanotubes (NTs), jointly with the basic understanding of the origin of their remarkable properties and the potential applications in nanoelectronics, initiated many investigations in this field of material science in the past decade.1 It was found that other, carbon-free materials, which possess layered structures in their normal modification, can be synthesized in fullerene- and nanotube-like forms.2,3 MoS2 is such a material that can exist in 3D cage structure, singleand multiwalled nanotubes, and ropes. Theoretical work4 on individual MoS2 tubes with outside diameters ranging from 8 to 26 Å predicted that the electronic structure of the MoS2 NT differs from the bulk material: it depends on the chirality and the band gap is increasing monotonically toward the bulk value with increasing tube diameter. This is unlike carbon NTs, where the band gap changes much more rapidly with the structure. Zigzag MoS2 NTs are predicted to have a narrow direct gap and thus might be used for future optoelectronic devices.5 However, MoS2 NTs, with a diameter smaller than 20 Å, appeared to be unstable in theory4 due to the strong distortions when the trilayer is curved to a tube. For nanotubes to be useful for electronics, a weak dependence of the band gap on diameter is required, so that a small distribution in the diameter can still be used. Alternatively, perfectly monodisperse materials are required. Carbon NTs do not fulfill either of these requirements.6 From * Corresponding author. Phone: +49 30 6392 4884. Fax: +49 30 6392 2989. E-mail:
[email protected]. † BESSY m.b.H. ‡ Forschungszentrum Ju ¨ lich GmbH. § Jozef Stefan Institut. 10.1021/nl049068t CCC: $27.50 Published on Web 08/21/2004
© 2004 American Chemical Society
the application point of view, the situation is more favorable for MoS2 NTs, where perfect size control is not absolutely necessary, since a distribution in the diameter does not cause a large change of the band gap character (i.e., direct vs indirect). Prospects for application of NTs in electronics were improved with the reported synthesis of the subnanometer diameter MoS2-Ix nanostructures.7 The material grows in the form of bundles of identically structured molecules, with an outer diameter of 9.6 Å. Bundles can be up to several hundreds of micrometers long and consist of up to 106 individual nanotube-like objects. High-resolution transmission electron microscopy measurements along bundle axes showed that MoS2Ix NTs build a hexagonal close-packed structure. On the basis of the X-ray and electron diffraction data, a model structure of bundles consisting of S-Mo-S trilayer cylinders was suggested. 7 The samples used in this investigation were grown as described previously.7 To study the changes in the structure of small diameter MoS2-Ix NTs compared to 2H-MoS2 bulk material, we investigate these materials by soft X-ray absorption (SXA) and soft X-ray fluorescence (SXF) atomselective photon-in photon-out spectroscopy. The local partial density of the electronic states (LPDOS) in the valence band (VB) and conduction band (CB) is probed independently by SXF and SXA. With photon excitation using monochromatized synchrotron radiation, the information obtained is related to the bulk of the material under investigation. Both SXA and SXF spectroscopies are performed in a photon-in photon-out mode and hence are insensitive to the electrical charging of the nanotubes. The basic process involved in
both techniques is the excitation of a core electron into empty states in the CB and the subsequent relaxation when an electron from the VB recombines with the core hole. In our study, S 2p core holes are selectively created. Only states that can participate in dipole transitions with the S 2p core hole are probed, i.e., S s- and d-states and Mo-S molecular orbitals that have significant s- and d-character with respect to the excited S atom. The production of the core vacancy as a function of the incoming photon energy is measured in SXA and reflects the density of s- and d-symmetry unoccupied states in the CB. The energy-resolved SXF measurements of the emitted photons after the electron-hole recombination monitor the occupied states in the VB having s- and d-character. The experimental results are compared to simple model calculations exploring the effects of bond distortion on the electronic structure. 2H-MoS22 bulk material and MoS2-Ix NTs have been investigated at the BESSY UE56/1-SGM beamline. The beamline energy bandwidth was set to 150 meV. S L2,3 SXF was recorded in a Rowland-type soft X-ray fluorescence spectrometer with 600 L/mm spherical grating of 3710 mm radius and 2D detection at an appropriate off-Rowland angle to obtain the entire spectrum in a parallel mode. With the entrance slit of 100 µm width, the SXF energy resolution was 0.5 eV as measured by the energy width of the elastically reflected light from the primary monochromator. SXA was monitored in total fluorescence yield (FY) mode using a photodiode as detector. Saturation effects were avoided by using a normal-in grazing-out geometry.8 Brillouin Zone Integrating Measurements. First, we present our measurements on bulk material. MoS2 belongs to a class of semiconducting layered compounds with a trigonal prismatic coordination of atoms. In the most common 2H polytype the hexagonal unit cell consist of two S-Mo-S sandwich layers. The bonding in the trilayer has covalent character. The interaction between the adjacent S layers belonging to different sandwiches is very weak and of van der Waals type. Calculations9 demonstrated that the principal valence band structure of this compound is composed of 4d orbitals of Mo as well as 3p, 3s orbitals of S. The d states of Mo are split into a nondegenerate a1g and doubly degenerated ee bands in the ligand-field of the trigonal-prismatic surrounding and are strongly hybridized with the S p states. The lowest group in the band corresponds to the completely occupied S 3s states. The band close to the maximum of the VB is built of S 3p and Mo a1g, and the Fermi level is located at the upper edge of the Mo band. Mo 4d states hybridized with empty S 3p and 3d states dominate the bottom of the CB. Experimentally, the value of the band gap in the 2H-MoS2 is 1.2 eV10 and theoretical calculations based on the local density approximation predict an indirect band gap for this material equal to 0.89 eV.9 Turning from the bulk to the subnanometer diameter NTs, one would expect changes due to (i) different overlapping of the Mo and S states inside the tube, (ii) effects of quantum confinement and, possibly, (iii) intertube interaction within a bundle. 1750
Figure 1. SXA (red) and SXF (blue) spectra of (a) 2H-MoS2 and (b) MoS2-Ix nanotubes compared with calculated spectra for the three-atom cluster model (experiment: symbols; calculation: solid lines). For 2H-MoS2 the S-Mo-S bond angle in the model corresponds to the bulk value of 82°, while the S-Mo-S bond angle in the nanotubes corresponds to 63° as suggested in ref 7. In the energy range shown, S 3p and Mo 4d derived states are dominating. The lower valence band composed of states with predominantly S 3s-character is located below 154 eV emission energy.
In Figure 1 the measured SXA and high energy excited SXF spectra of the 2H-MoS2 and MoS2-Ix NTs are presented. The measured spectra reflect the energy range in the vicinity of the top of the VB and bottom of the CB, where states are most sensitive to the changes in the chemical bonding. The energy window in Figure 1 does not include the most intense S 3s emission, which is at lower emission energies. The most notable changes in spectra of NTs compared to the bulk can be seen in the immediate vicinity of the band edges. The peak at 162.5 eV in the absorption spectrum, which can be observed in the bulk material, disappears in the NTs. In SXF, spectral signal weight at high energies, corresponding to the states at the top of the VB, is strongly increased in the NTs. The top of the VB and the bottom of the CB are dominated by Mo states. We observed those states in the experimental spectra via their overlap with the S site. Apparently, those states are affected most drastically when going from bulk material to the NTs. Interestingly, the band gap is reduced in the tubes as compared to the bulk, as will be seen more clearly in the resonant inelastic X-ray scattering Nano Lett., Vol. 4, No. 9, 2004
(RIXS) measurements below. This is counter intuitive with respect to a simple quantum confinement model, which would not take bond angle distortion into account. Theoretical Results and Discussion. To understand the main changes in the electronic structure between bulk and MoS2-Ix NT material, we performed electronic structure calculations of model structures. As noted in ref 7, the most drastic change in the local geometric arrangement of the MoS2-Ix NTs as compared to the bulk is the S-Mo-S bond angle. Now we consider only the most simple model structure containing this bond angle as a parameter, namely a threeatom S-Mo-S cluster. Based on the electronic structure calculations, we compute the probability of electronic transitions from the 2p core level to the unoccupied states in the CB to simulate SXA spectra and from the valence band to the 2p core hole to simulate SXF spectra. The calculations were made using the StoBe software,11 based on the linear combination of Gaussian type orbitals-MO solution of the Kohn-Sham DFT equations. The bulk material was modeled with a S-Mo-S angle of 82° and a S-Mo bond length equal 2.417 Å9. The MoS2Ix NTs were modeled with the smaller S-Mo-S angle of 63° suggested in ref 7, keeping the bond length constant. The calculated spectra were convoluted with a Gaussian function with a FWHM of 0.3 eV for SXA and 0.5 eV for SXF to account for the lifetime broadening of the core hole and the experimental energy resolution. The 1.1 eV spinorbit splitting of the 2p3/2 and 2p1/2 levels was taken into account assuming a 2:1 branching ratio. The resulting theoretical model spectra are plotted in Figure 1 together with the experimental data. The calculations are in rather good agreement with the measured intensities and reproduce the main features of the experimental spectra. The simple model obviously cannot describe effects resulting from the crystalline structure of the bulk and the NTs. Also, off-center transition matrix element effects may influence the overall spectral weight of the top versus the bottom of the valence band. Nevertheless, the main trend in the SXA and SXF spectra when going from the bulk to the NTs is predicted: the SXF spectrum exhibits increased spectral weight at the top of the VB, whereas the peak at 162.5 eV is absent in the SXA spectrum. The good agreement between the calculated spectra and both the experimental SXA and SXF spectra suggests that our simple model captures the essential effects in the changes of the local electronic structure between bulk MoS2 and subnanometer MoS2-Ix NTs. We conclude that the redistribution of electron density in the tubes compared to 2HMoS2 bulk material is mainly due to the S-Mo-S bond distortion. A charge distribution analysis in the ground state shows that electron density is redistributed from the S atoms toward the Mo atom due to the changes in bond angle. In addition to S and Mo, the NTs contain iodine. As the atomic position of iodine in the nanotube bundles is not wellknown, we have conducted several calculations with iodine at different positions in our model structure. We could not detect an improved match between model calculation and Nano Lett., Vol. 4, No. 9, 2004
experiment upon the inclusion of iodine into our model structures. Momentum-Sensitive Measurements. Electronic structure information with sensitivity to the momentum of the electronic states can be obtained by resonant inelastic X-ray scattering (RIXS).12,13 Parts of the Brillouin zone can be probed selectively by RIXS, exploiting conservation of energy and momentum in the combined excitation-decay scattering process. Here, the SXF spectra are recorded as a function of the incident photon energy in the vicinity of the S 2p3/2 absorption edge. We compare the changes in the SXF spectra as a function of the incident photon energy pωin for the bulk material and MoS2-Ix NTs. The band structure of the 2H-MoS2 can be shortly described as follows: the top of the VB and the bottom of the CB are at the Γ point and at the midpoint of the ΓΚ line, respectively; along the ΓΑ axis the band dispersion at the top of the VB around the Γ point is relatively large, reflecting the extension of wave functions near the Fermi energy over several trilayers.14 In Figure 2, SXF spectra of 2H-MoS2 are shown as a function of pωin. At pωin ) 162.35 and 162.50 eV the core electron is excited into states close to the bottom of the CB, thus transitions occur resonantly from states close to the ΓΚ midpoint of the VB. At pωin ) 163.35 eV the resonant excitation moves closer toward the Κ point. Starting at pωin ) 163.70 eV states in the vicinity of the top of the VB at the Γ point and states at the Γ-Α direction appear in fluorescence spectra. Additionally, at pωin ) 163.70 eV the 2p1/2 level is excited and the measured spectrum contains additional intensity, shifted to the higher emission energies. Changes in the shape of the spectrum excited with the pωin ) 167.70 eV (above the second band gap) are related to the excitations to states above the minimum of the second Mo d-band: the contribution of the states in the vicinity of the Γ point to the fluorescence intensity increases again. In Figure 2b we present SXF spectra for MoS2-Ix NTs. In general, the shape of the spectra is significantly different from the corresponding bulk spectra, as discussed earlier for the Brillouin zone integrated spectra. Compared to the 2HMoS2 spectrum at pωin ) 163.35 eV, the peak around the 161.5 eV, close to the top of the VB, is more intense in the MoS2-Ix NTs spectrum at pωin ) 163.40 eV. The shape of the SXF spectra does not change significantly with the energy up to pωin ) 164.68 eV, when the 2p1/2 threshold is excited. The SXF spectra change when the photon energy reaches 167.75 eV, now the intensity of low energy peak around 156 eV emission energy increases. This behavior indicates the existence of a second set of bands separated by a second gap from the lower energy unoccupied states in MoS2-Ix NTs similar to the 2H-MoS2 band structure. The position of the top of the VB for the MoS2-Ix NTs can be measured from the spectrum recorded with pωin ) 164.00 eV, below the 2p1/2 threshold. The VB maximum for the MoS2-Ix NTs is observed at higher emission energies, compared to the 2H-MoS2. In conjunction with the absence of the low energy peak in the SXA spectra of MoS2-Ix NTs, 1751
Figure 2. (a) SXF spectra of bulk MoS2 recorded with different excitation energies corresponding to the specific features in the SXA spectrum (see Figure 1a). Intense peaks at high energies in spectra excited with energies close to the absorption threshold are due to elastically scattered primary photons. For pωin e 162.50 eV, these peaks have been divided by two. (b) SXF spectra of MoS2-Ix NTs recorded with different excitation energies corresponding to the specific features in the SXA spectrum (see Figure 1b). For pωin e 164.95 eV, the intense elastic peaks have been divided by 25 and clipped.
the total reduction of the band gap in the NTs compared to the bulk is measured to be 0.8 eV. From our measurements in conjunction with the bulk 2HMoS2 band gap of 1.2 eV11 we determine a band gap of 0.4 ( 0.2 eV for the MoS2-Ix NTs. To our knowledge, this is the first measurement of the band gap in this nanotube material. The size of the band gap is in agreement with the prediction of a smaller gap in the NTs as compared to the bulk material.4 The measured RIXS effects can be used to check electronic structure calculations of MoS2-Ix NTs, as RIXS spectra can be calculated on the basis of the band structure. The atomic positions within the MoS2-Ix tubes and bundles are not unambiguously determined yet. Electronic structure calculations for different NT bundle configurations as carried out in ref 15 may be used in conjunction with the experimental RIXS data presented here in order to decide between different structural models for the MoS2-Ix NTs. In summary, the local electronic structure in subnanometer diameter MoS2-Ix NTs has been investigated by photon-in photon-out X-ray spectroscopies. The nanotubes are found to be semiconducting, with a band gap of 0.4 ( 0.2 eV. Significant differences in both the conduction and valence 1752
band density of electronic states are observed as compared to 2H-MoS2. Simple model calculations suggest that the main features of the altered electronic structure can be attributed to the effect of Mo-S-Mo bond angle distortion imposed by the small diameter of nanotubes. Additional momentum selective spectra may be used to check full band structure calculations for the nanotubes and possibly to decide between different structural models. Acknowledgment. We thank Prof. L. G. M. Pettersson from University of Stockholm (Sweden) for his help with the electronic structure calculations. References (1) Dresselhaus, M. S.; Dresselhaus, G.; Avouris, Ph. Carbon Nanotubes; Springer-Verlag: Berlin, Germany, 2000. (2) Tenne, R.; Margulis, L.; Genut, M.; Hodes, G. Nature (London) 1992, 360, 444. (3) Chopra, N. G.; Luyken, R. J.; Cherrey, K.; Crespi, V. H.; Cohen, M. L.; Louie, S. G.; Zettl, A. Science 1995, 269, 966. (4) Seifert, G.; Terrones, H.; Terrones, M.; Jungnickel, G.; Frauenheim, T. Phys. ReV. Lett. 2000, 85, 146. (5) Lux-Steiner, M. Ch.; Jager-Waldau, A.; Bucher, E. Polycrystalline Semiconductors III, Physics and Technology, Solid State Phenomena; Scitec Publications: Switzerland, 1994; pp 37-38, 214. (6) Eisebitt, S.; Karl, A.; Eberhardt, W.; Fisher, J. E.; Sathe, C.; Agui, A.; Nordgren, J. Appl. Phys. A 1998, 67, 89. Nano Lett., Vol. 4, No. 9, 2004
(7) Remskar, M.; et al. Science 2001, 292, 479. (8) Eisebitt, S.; Rubensson, J.-E.; Bo¨ske, T.; Eberhardt, W. Phys. ReV. B 1993, 47, 14103. (9) Raybaud, P.; Kresse, G.; Hafner, J.; Toulhoat, H. J. Phys: Condens. Matter 1997, 9, 11085. Raybaud, P.; Kresse, G.; Hafner, J.; Toulhoat, H. J. Phys: Condens. Matter 1997, 9, 11107. (10) Kam, K. K.; Parkinson, B. A. J. Phys. Chem. 1987, 86, 463. (11) StoBe-deMon Version 1.0; Hermann, K.; Pettersson, L. G. M. Casida, M. E.; Daul, C.; Goursot, A.; Koester, A.; Proynov, E.; St-Amant, A.; Salahub, D. R. Contributing authors: Carravetta, V.; Duarte, H.; Godbout, N.; Guan, J.; Jamorski, C.; Leboeuf, M.; Malkin, V.;
Nano Lett., Vol. 4, No. 9, 2004
(12)
(13) (14) (15)
Malkina, O.; Nyberg, M.; Pedocchi, L.; Sim, F.; Triguero, L.; Vela, A. StoBe Software, 2002. Eisebitt, S.; Eberhardt, W. J. El. Spec. Rel. Phen. 2000, 110, 335. Eisebitt, S.; Lu¨ning, J.; Rubensson, J.-E.; Eberhardt, W. Phys. Status Solidi B; 1999, 215, 803. Lu¨ning, J.; Rubensson, J.-E.; Ellmers, C.; Eisebitt, S.; Eberhardt, W. Phys. ReV. B 1999, 59, 10573. Kobayashi, K.; Yamauchi, J. Phys. ReV. B 1995, 51, 17085. Verstraete, M.; Charlier, J.-C. Phys. ReV. B 2003, 68, 045423.
NL049068T
1753
