Zhiyong
Wang‡
a,
Keke
Zhao‡
ad,
Hong
Li‡
b,
Zheng
Liu
c,
Zujin
Shi
*a,
Jing
Lu
*b,
Kazu
Suenaga
*c,
Soon-Kil
Joung
c,
Toshiya
Okazaki
c,
Zhaoxia
Jin
d,
Zhennan
Gu
a,
Zhengxiang
Gao
b and
Sumio
Iijima
c
aBeijing National Laboratory for Molecular Sciences, State Key Lab of Rare Earth Materials Chemistry and Applications, College of Chemistry and Molecular Engineering, Peking University, Beijing, 100871, P. R. China. E-mail: zjshi@pku.edu.cn; Fax: +86-10-62751708; Tel: +86-10-62751495
bState Key Laboratory of Mesoscopic Physics and Department of Physics, Peking University, Beijing, 100871, P. R. China. E-mail: jinglu@pku.edu.cn; Fax: +86-10-62751615; Tel: +86-10-62756393
cNanotube Research Center, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba, 305-8565, Japan. E-mail: suenaga-kazu@aist.go.jp; Fax: +81-29-861-4806; Tel: +81-29-861-6850
dDepartment of Chemistry, Renmin University of China, Beijing, 100872, P. R. China
First published on 20th October 2010
Layered tungsten disulfide nanostructures are of both fundamental and technological interest. The widths of currently synthesized WS2 ribbons are in the microscale. By using single-walled carbon nanotubes and double-walled carbon nanotubes as templates, we fabricate WS2 nanoribbons with smooth zigzag edges and uniform widths down to 1–3 nm and layer numbers down to 1–3, dependent on the nanotube diameter. Although bulk WS2 is a nonmagnetic semiconductor, the ultra-narrow free-standing zigzag-edged WS2 nanoribbons turn out to be magnetic or nonmagnetic metals depending on the edge passivation way according to our first-principles calculations, whereas the ultra-narrow armchair-edged WS2 nanoribbons remain nonmagnetic semiconductors with a narrow gap.
In this paper, we report the synthesis of ultra-narrow WS2 nanoribbons inside carbon nanotubes, following the same procedure to synthesize ultra-narrow MoS2 nanoribbons. The synthesized WS2NRs have smooth edges, and even have widths down to 1–3 nm, and layer numbers down to 1–3 layers, depending on the inner diameter of the carbon nanotube. In contrast to nonmagnetic semiconducting bulk WS2 and magnetic semiconducting or metallic ultra-narrow zigzag-edged MoS2NRs, first-principles calculations demonstrate that the ultra-narrow zigzag-edged WS2NRs are always metallic and can be magnetic or nomagnetic, dependent on the passivation method.
![]() | ||
Fig. 1 Raman spectra of SWCNT, WS2@SWCNT, and WS2@DWCNT. The E2g1 and A1g modes are from WS2NRs. |
![]() | ||
Fig. 2 (a) Annular dark-field image of WS2NRs in SWCNTs. The identification of nanoribbons is based on the fact that the width of this encapsulated quasi-one-dimensional structure is obviously larger than the height (for example, see Fig. 5c). We found that all the nanoribbons inside SWCNTs consist of one WS2 layer and the formation of multi-layer WS2NRs only occurs in DWCNTs with larger diameters. Thus the crystals in this dark-field image are ascribed to single-layer WS2NRs rather than multi-layer WS2NRs or nanowires. (b) Micro EDX signals obtained from the area denoted by the rectangle in (a), confirming the presence of W and S elements. Signals of Cu come from the supporting foil. |
The diameter of SWCNTs used for encapsulation of WS2NRs is mainly in the range of 1.3–1.6 nm. Only single-layer WS2NRs are observed inside SWCNTs. Fig. 3a and 3b show the high-resolution transmission electron microscopy (HR-TEM) images of encapsulated single-layer WS2NRs that are parallel to the incident electron beam. The thick lines correspond to the WS2NRs (denoted by arrows), and the adjacent thinner lines are the nanotube walls. Fig. 3c shows a structural model of WS2NRs encapsulated inside SWCNT. A WS2 layer consists of three layers of atoms, i.e., a layer of W atoms in the center and two layers of S atoms at two sides. The dark lines denoted by arrows in HR-TEM images actually represent three layers of atoms, however the W and S layers are not distinguishable. When using DWCNTs with inner diameters of 1.5–4.0 nm as templates, double-layer WS2NRs are generated. Typical HR-TEM images are shown in Fig. 3d and 3e, in which the two parallel dark lines correspond to the WS2 layers. The distance between the two WS2 layers is 0.62 nm, in agreement with that in hexagonal phase of WS2 (0.618 nm). Besides double-layer WS2NRs, single-layer and triple-layer WS2NRs encapsulated in DWCNTs are occasionally observed (see ESI, Fig. S1†). The lengths of WS2NRs inside SWCNTs and DWCNTs are usually tens of to over one hundred nanometres.
![]() | ||
Fig. 3 (a) HR-TEM image of a single-layer WS2NR (denoted with an arrow) encapsulated in SWCNT. (b) Two adjacent SWCNTs filled with single-layer WS2NRs. (c) Schematic model of a single-layer WS2NR encapsulated in SWCNT. (d, e) HR-TEM images of double-layer WS2NRs encapsulated in DWCNTs. (f) Schematic model of a double-layer WS2NR encapsulated in DWCNT. In model, blue ball: W; yellow ball: S; grey ball: C. |
When the WS2NR basal plane is nearly perpendicular to the incident electron beam, the structure of the WS2NR can be resolved at atomic level by HR-TEM. Fig. 4a shows a WS2NR with a width of ∼1.5 nm. From the hexagonal pattern in the image, the WS2NR is unambiguously assigned to a zigzag-edged nanoribbon. Simulation of single-layer WS2NR with different tilting angles under an electron beam (Fig. 4b and Fig. S2†) reveals that the position of W atoms is darker than that of S atoms, and the contrast difference becomes more notable with increasing tilting angle x and y. Accordingly, the dark spots in Fig. 4a are ascribed to W atoms, as illustrated in Fig. 4c, and the position of S atoms are then identified. Following the convention of GNRs,18 we use N-ZWS2NR (N-AWS2NR) to denote a zigzag-edged and (armchair-edged) WS2NR with N zigzag chains (dimer lines) across the ribbon width. The WS2NR in Fig. 4a is identified as a 6-ZWS2NR. The distance between the adjacent W atoms in Fig. 4a is 0.32 nm, which is very close to the theoretical distance of 0.315 nm. From the annular dark-field image in Fig. 2a, we can see three WS2NRs. According to the arrangement manner of the W atoms, the WS2NR in the top-left corner of the image is identified as a 5-ZWS2NR and the other two nanoribbons are identified as 4-ZWS2NRs with width of 1.0 nm, which is the smallest width measured for the WS2NRs in our sample. In fact, all the plane views of the WS2NRs in our sample have zigzag edges, indicating that zigzag-edged WS2NRs are more stable than armchair-edged ones.
![]() | ||
Fig. 4 (a) Atomic-resolved HR-TEM image of a single-layer WS2NR encapsulated in a SWCNT. The yellow hexagons help with the identification of the atom structure, and this WS2NR is identified as 6-ZWS2NR. (b) HR-TEM simulation and structural model of a single-layer WS2NR at 4 different tilting angles. The tilting angle is x = 0° and y = 0° when the WS2NR plane is perpendicular to the incident electron beam. More simulated images can be found in the ESI.† (c) An enlarged view of the image in (a). The dark spots correspond to W atoms. After assignment of the W atoms, the position of the S atoms are identified. (d) Schematic model of the WS2NR in (a). In the model, blue ball: W; yellow ball: S. |
Besides WS2NRs parallel or perpendicular to the electron beam, WS2NRs in other orientations are also observed. Fig. 5a shows a WS2NR with a tilting angle of 0° < x < 90° and 0° < y < 90°. Fig. 5b is the structural model. Due to the confinement of carbon nanotubes, the width of the encapsulated WS2NRs is close to the inner diameter of the carbon nanotubes. However, the accurate determination of the width of WS2NRs are sometimes difficult because of the inclination of the nanoribbons. The measured maximum width of WS2NRs is ∼3 nm in our experiments.
![]() | ||
Fig. 5 (a) HR-TEM image and (b) schematic model of a tilted WS2NR encapsulated in DWCNT. (c) HR-TEM image showing a twisted single-layer WS2NR encapsulated in SWCNT. In the schematic model, blue ball: W; yellow ball: S; grey ball: C. |
WS2NRs are often found to be twisted inside carbon nanotubes, and the host carbon nanotubes are radially deformed to a large extent. One example is shown in Fig. 5c. The twist of both the nanoribbon and the SWCNT can be clearly seen. As the schematic model shows, the cross section of the SWCNT expands in the direction of the nanoribbon plane as a consequence of the strong interaction between the nanoribbon edge and the carbon nanotube wall, and meanwhile, shrinkage of the cross section in the perpendicular direction occurs. The minoraxis of the SWCNT cross section is measured to be 1.0 nm in Fig. 5c. Based on measurements of many WS2NR@SWCNTs, the minoraxis of the distorted SWCNT cross section is in the range of 1.0–1.1 nm and the distance between the basal plane of the WS2 layer and the nanotube wall is 0.50–0.55 nm. Compared with the initial diameter range of 1.3–1.6 nm, it is apparent the SWCNTs have undergone severe radial deformation. For DWCNTs filled with double-layer WS2 nanoribbons, the distance between the basal plane of the WS2 layer and the innertube wall is also found to be 0.50–0.55 nm. These results indicate that a stable hybrid structure is formed when the distance between the nanoribbon and the nanotube gets into this range and the radial deformation of the nanotubes tends to satisfy this rule in the formation process of the nanoribbons.
We perform an intensive DFT calculation for the free-standing 1 nm wide 4-ZWS2NRs and 7-AWS2NR to identify the details of the edge and explain why only the zigzag edge is observed. 0%, 50% and 100% S-saturated W edges and 100% and 50% saturated S edges are considered in our 4-ZWS2NR model. We use a symbol of Wp%S-Sq% to denote the edge configuration of the 4-ZWS2NR with a p% S-saturated W edge and a q% saturated S edge. Only the intact 7-AWS2NR model is considered in this paper. Supercells containing two and one unit cells are built for the 4-ZWS2NRs and 7-AWS2NR, respectively.
The optimized structures of the WS2NRs are shown in Fig. 6a–6g. The triple-layer networks are kept at both the 4-ZWS2NRs and 7-AWS2NR after full relaxation, but the thickness of the marginal triple-layer changed by −0.18 Å to 0.16 Å, comparing to the bulk and single-layer WS2, whereas the central triple-layer was nearly unchanged. The fully bare and fully S-saturated W edges of the 4-ZWS2NRs shift inward notably by 0.16–0.21 Å, while the half S-saturated W edges shift outward slightly by 0.01–0.05 Å. The fully and half-saturated S edges shift outward by 0.01–0.06 Å and 0.05–0.09 Å, respectively. The neighboring S atoms of the half-saturated S edges significantly approach each other in the vertical direction by 1.24–1.35 Å. We find that the W atoms at the half S-saturated W edge are dimerized with an amplitude of 0.11–0.54 Å. Similar phenomenon was found in their structural analogue ZMoS2NRs.22 The edge W atoms of the 7-AWS2NR shift inward by 0.39 Å.
![]() | ||
Fig. 6 (a–g) Optimized structures of the 4-ZWS2NRs with different edge configurations (see text) and the 7-AWS2NR. (h–k) Ground-state spatial spin density distributions arranged by descending δG of the 4-ZWS2NRs. The nonmagnetic 4-ZWS2NRs with W(100%S)–S(50%) and W(50%S)–S(50%) edge configurations are not shown here. The parallel and antiparallel spin orientations between the two edges in (i) are degenerate in total energy. Red and green are used to denote the spin-up and spin-down orientations, respectively. The isovalue is 0.01 a.u. Blue ball: W; yellow ball: S. |
Direct comparison of cohesive energy per atom of WS2NRs is not a suitable way because the relative stability depends upon the chemical potentials of their constituents, which in turn represent environmental conditions.25 Following previous works,25–27 we adopt the molar (per atom) Gibbs free energy of formation δG to investigate the relative stability of the WS2NRs. We define δG for the WS2NRs as
δG = ERibbon + nWμW + nSμS |
The calculated δG values are given in Table 1. The 4-ZWS2NRs are more stable than the 7-AWS2NR with the exception of the 4-ZWS2NR with the W(0%S)–S(50%) edge configuration. The calculated higher relative stability of the 4-ZWS2NRs over the 7-AWS2NR explains why the zigzag-edged WS2NRs are observed. The relative stability of the 4-ZWS2NRs ascends in the order of: W(0%S)–S(50%) < W(0%S)–S(100%) < W(100%S)–S(50%) < W(100%S)–S(100%) < W(50%S)–S(50%) < W(50%S)–S(100%). The most stable passivation method of 4-ZWS2NRs is the same as that of 4-ZMoS2NRs.22 The half S-saturated W edge turns out to be the most stable via comparing δG of the 4-ZWS2NRs with one identical edge, the fully S-saturated W edge is slightly higher in δG (0.04 eV/atom), and the fully bare W edge is significantly higher in δG (0.17–0.20 eV/atom). This trend is similar to the structural analogue 4-ZMoS2NRs,22 and we speculate that the half and fully S-saturated W edges probably coexist in sulfidation conditions. The half-saturated S edge is 0.02–0.05 eV/atom higher in δG than the fully saturated S edge.
Structure | 4-ZWS2NRs | 7-AWS2NR | ||||||
---|---|---|---|---|---|---|---|---|
W(0%S)–S(50%) | W(0%S)–S(100%) | W(100%S)–S(50%) | W(100%S)–S(100%) | W(50%S)–S(50%) | W(50%S)–S(100%) | |||
a The parallel and antiparallel spin coupling states between the two edges are degenerate in total energy. b The ground state is nonmagnetic. c M: Metallic. d S: Semiconducting. e A: Antiparallel. f P: Parallel. g one edge is nonmagnetic. | ||||||||
δG/eV | −0.40 | −0.45 | −0.56 | −0.58 | −0.60 | −0.62 | −0.44 | |
ΔE/meV | −27 | −86a | ——b | −53 | ——b | −60 | ——b | |
Electronic type | M c | M | M | M | M | M | S d | |
Edge Spin coupling | A e | P f | A | —— | ——g | —— | ——g | —— |
M tot (μB) | 2.56 | 3.04 | 1.28 | 0.00 | 1.00 | 0.00 | 1.16 | 0.00 |
ξ | —— | —— | —— | 53% | —— | 55% | —— |
The 4-ZWS2NRs with W(0%S)–S(50%), W(0%S)–S(100%), W(100%S)–S(100%) and W(50%S)–S(100%) edge configurations turn out to be magnetic, while the other two 4-ZWS2NRs as well as the 7-AWS2NR are nonmagnetic. As given in Table 1, the magnetic ground states of the 4-ZWS2NRs are 27–86 meV per supercell lower in total energy than their respective nonmagnetic states. These differences are smaller than the corresponding values of 81–197 meV in 4-ZMoS2NRs.22 Taken together, the magnetism of the ZWS2NRs is weaker than that of ZMoS2NRs. The ground-state spin distributions of the four magnetic 4-ZWS2NRs are displayed in Fig. 6h–6k. The bare W edge and fully saturated S edge are found to be always ferromagnetic with a localized magnetic moment of MW(edge) = 1.00–1.16 μB and MS(edge) = 0.26–0.32 μB, respectively. Fully and half S-saturated W edges and half-satutared S edges completely quench MW(edge) and MS(edge), respectively. The energy differences between the parallel and antiparallel spin orientations between the two edges of the 4-ZWS2NR with W(0%S)–S(100%) edge configuration is less than 1 meV per supercell, implying that the spin coupling between the bare W edge and fully saturated S edge is negligible.
The ground-state band structures for the WS2NRs are presented in Fig. 7a–7g. The 4-ZWS2NRs, whether magnetic or not, are all metallic. The orbital wave functions of the two metallic states of the 4-ZWS2NR with W(50%S)–S(100%) edge configuration are analyzed by using a double numerical atomic orbital basis set plus the polarization function (DNP) as implemented in the DMol3 package28 and are found to be chiefly localized on the W and S edges (Fig. 7h–7i). Hence, the spin current travels along the two edges of this nanoribbon. A similar phenomenon was found in the structural analogue 4-ZMoS2NR.22 The 7-AWS2NR is a semiconductor with a band gap of 0.39 eV.
![]() | ||
Fig. 7 (a–g) Ground-state band structures of the 4-ZWS2NRs arranged by descending δG and the band structure of the 7-AWS2NR. Red and blue stand for the spin-up and spin-down states, respectively. The Fermi level is set to zero. (h–i) Isosurfaces of the orbital wave functions of the two conduction bands in the spin-down channel at the Γ point for the 4-ZWS2NR with W(50%S)–S(100%) edge configuration. The isovalue is 0.05 a.u. Red and green are used to indicate the positive and negative signs of the wavefunctions, respectively. Blue ball: W; yellow ball: S. |
We perform transport calculations of two 1.9 nm long truncated 4-ZWS2NRs with W(100%S)–S(100%) and W(50%S)–S(100%) edge configurations, which are the two most stable magnetic nanoribbons. The former has three and two bands across Ef in the spin-down and spin-up channels, respectively, whereas the latter has two and one bands across Ef in the spin-down and spin-up channels, respectively. The optimized two-probe models with truncated 4-WS2NRs chemically bonded to a quasi-one-dimensional Li electrodes are shown in Fig. 8. We define the spin polarization ξ of the electron current as
![]() | ||
Fig. 8 Optimized two-probe models and transmission spectra for the 1.9 nm long truncated 4-ZWS2NRs with W(100%S)–S(0%) and W(50%S)–S(0%) edge configurations using the lithium (100) surface as electrodes. Inset plots show surfaces of the constant spin-resolved local DOS evaluated at the Fermi energy. Red and blue stand for the spin-up and spin-down states, respectively. The Fermi level is set to zero. Blue ball: W; yellow ball: S; purple ball: Li. |
Finally, we study the properties of the WS2NR@SWCNT composite. The length of four periods of ZWS2NR is nearly equal to that of three periods of zigzag SWCNT, and we therefore construct a supercell model consisting of four unit cells of the 4-ZWS2NR and three unit cells of the zigzag (16,0) SWCNT. Two 4-ZWS2NRs with W(0%S)–S(100%) and W(50%S)–S(100%) edge configurations are considered. Fig. 9a and Fig. 10a show the optimized structures of 4-ZWS2NRW(0%S)-S(100%)@(16,0)SWCNT and 4-ZWS2NRW(50%S)-S(100%)@(16,0)SWCNT, respectively. The host (16,0) SWCNT is radially deformed with deformation degrees of 40% and 57% measured by aspect ratio. The magnetism of the two inner 4-ZWS2NRs is completely quenched by the host (16,0) SWCNTs. The 4-ZWS2NRW(0%S)-S(100%)@(16,0)SWCNT composite is a semimetal and the 4-ZWS2NRW(50%S)-S(100%)@(16,0)SWCNT composite is a metal (Fig. 9b and 10b). The two deformed host SWCNTs have a small 0.13 eV and zero eV direct band gap (Fig. 9c and 10c, respectively), much smaller than thet band gap of 0.59 eV of pristine (16,0) SWCNTs (Fig. 9b and 10b). The electronic structures of the two 4-ZWS2NR@(16,0)SWCNT composites are quite similar to those of their 4-ZMoS2NR@(16,0)SWCNT counterparts.22 The conductivity of the two 4-ZWS2NR@(16,0)SWCNT composites is mainly contributed to by the inner 4-ZWS2NRs based on the partial density of states analysis (Fig. 9e and 10e).
![]() | ||
Fig. 9 (a) Side and top views of the optimized 4-ZWS2NRW(0%S)-S(100%)@(16,0)SWCNT composite. (b–d) Band structures of the 4-ZWS2NR W(0%S)-S(100%)@(16,0)SWCNT, intact (16,0) SWCNT, and deformed host (16,0) SWCNT. (e) Total density of states of the 4-ZWS2NRW(0%S)-S(100%)@(16,0)SWCNT and partial density of states of its inner bare 4-ZWS2NR and deformed host (16,0) SWCNT. Blue ball: W; yellow ball: S; gray ball: C. |
![]() | ||
Fig. 10 (a) Side and top views of the optimized 4-ZWS2NRW(50%S)-S(100%)@(16,0)SWCNTcomposite. (b–d) Band structures of the 4-ZWS2NRW(50%S)-S(100%)@(16,0)SWCNT, intact (16,0) SWCNT, and deformed host (16,0) SWCNT. (e) Total density of states of the 4-ZWS2NRW(50%S)-S(100%)@(16,0)SWCNT and partial density of states of its inner 4-ZWS2NR with W(50%S)–S(100%) edge configuration and deformed host (16,0) SWCNT. Blue ball: W; yellow ball: S; gray ball: C. |
HR-TEM: Hitachi H-9000NAR was operated at 100 kV to characterize the structure of the products. JEM-2010F equipped with a CEOS post-specimen spherical aberration corrector (Cs corrector) was operated at 120 kV for atomic resolution imaging. A Gatan 894 CCD camera was used for digital recording of the HR-TEM images. A sequence of HR-TEM images (20 frames) was recorded, with a 0.5 s exposure time for each. After drift compensation, some frames can be superimposed to increase the signal-to-noise (SN) ratio for display. Cs was set in the range 0.5–2 μm, and the HR-TEM images were recorded under a slightly under-focus condition in order to enhance the contrast of the encapsulated materials with respect to the graphene network of carbon nanotubes; this ensured that the contrast of carbon nanotubes could be minimized by fabricating the desirable contrast transfer function (CTF). Image simulations were carried out by using MacTempas software package.
EDX: The JEM-ARM200F incorporating a CEOS pre-specimen spherical aberration corrector (Cs corrector) was operated at 80 kV for EDX analyses. The electron probe, after its aberrations were corrected, featured a current density level higher by an order of magnitude than conventional transmission electron microscopes and was capable of atomic level analysis.
Raman: Jobin Yvon HR-800 spectrometer was used to record the Raman spectra of the samples excited with a laser of 632.8 nm. F1g mode of silicon at 520.7 cm−1 was used to calibrate the spectra.
Footnotes |
† Electronic supplementary information (ESI) available: HR-TEM images of single-layer and triple-layer WS2NR encapsulated in DWCNTs, HR-TEM simulation of single-layer WS2NR at ten different tilting angles, and the spin distributions of the 4-ZWS2NRs. See DOI: 10.1039/c0jm02821e |
‡ These authors contributed equally to this work. |
This journal is © The Royal Society of Chemistry 2011 |