ARTICLES PUBLISHED ONLINE: 20 APRIL 2014 | DOI: 10.1038/NNANO.2014.64
Atomic mechanism of the semiconducting-tometallic phase transition in single-layered MoS2 Yung-Chang Lin1, Dumitru O. Dumcenco2†, Ying-Sheng Huang2 and Kazu Suenaga1 * Phase transitions can be used to alter the properties of a material without adding any additional atoms and are therefore of significant technological value. In a solid, phase transitions involve collective atomic displacements, but such atomic processes have so far only been investigated using macroscopic approaches. Here, we show that in situ scanning transmission electron microscopy can be used to follow the structural transformation between semiconducting (2H) and metallic (1T) phases in single-layered MoS2 , with atomic resolution. The 2H/1T phase transition involves gliding atomic planes of sulphur and/or molybdenum and requires an intermediate phase (a-phase) as a precursor. The migration of two kinds of boundaries (b- and g-boundaries) is also found to be responsible for the growth of the second phase. Furthermore, we show that areas of the 1T phase can be controllably grown in a layer of the 2H phase using an electron beam.
F
In situ observation of 2H/1T phase transition Here, we provide in situ observations of the transformation process between 2H and 1T phases in single-layered MoS2 at high temperatures. To monitor the phase transition in situ, we operated an aberration-corrected scanning transmission electron microscope (STEM) at 60 kV to visualize the dynamic process of the atomic motions in single-layered MoS2. This technique has already been used and verified while studying another ideal two-dimensional material, graphene, for dislocations15,16, grain boundaries17–19 and a
b
2H
1T
c S plane gliding (2H → 1T)
3.16 Å
or several centuries, molybdenum disulphide (MoS2) has been widely used as a practical solid lubricant1,2. MoS2 crystal is composed of stacks of atomic layers bound by van der Waals forces, with each layer constructed from S–Mo–S′ triple atomic planes with strong in-plane bonding. Recently, single-layered MoS2 , a direct-bandgap quasi-two-dimensional semiconductor, has shown its great potential for applications in electrical and optoelectronic devices3–5. Interestingly, one of the unique features of MoS2 is polymorphism, with its distinct electronic characteristics. Depending on the arrangement of its S atoms, single-layered MoS2 appears in two distinct symmetries: the 2H (trigonal prismatic D3h) and 1T (octahedral Oh) phases (Fig. 1a,b). The two phases should exhibit completely different electronic structures, with the 2H phase being semiconducting and the 1T phase metallic6–8. The two phases can easily convert one to the other via intralayer atomic plane gliding, which involves a transversal displacement of one of the S planes. The 1T phase was first reported to transform from 2H-MoS2 by Li and K intercalation6,9, with restacked 1T phases in LiMoS2 and KMoS2 confirmed by electron diffraction10,11, and it is also known to be stabilized by substitutional doping of Re, Tc and Mn atoms, which serve as electron donors12. However, 1T-LiMoS2 is thermodynamically unfavourable, and has been observed (by Raman spectroscopy13) to gradually transform to the 2H phase at room temperature. The phase transitions between 2H and 1T or 2H′ phases due to atomic plane gliding are presented in Fig. 1c,d. Note that 2H′ is a 608 (or 1808) rotational phase of 2H. Although the coexistence of metallic and semiconducting phases has indeed been reported in chemically exfoliated MoS2 by Eda and colleagues14, the actual dynamical process of the transformation between 2H and 1T phases involving intralayer atomic plane gliding has never been experimentally proven, nor has the atomic process of the phase transition been investigated in situ. If one is to consider the possibility of intentionally introducing the phase transition in single-layered materials in a controllable manner, the atomic process of this phase transition as well as its boundary structures must be corroborated in order to reliably design future low-dimensional devices.
d Mo plane gliding (2H → 2H’)
AB
ABC S Mo S’
Figure 1 | Polymorphs of single-layered MoS2. a,b, Schematic models of single-layered MoS2 with 2H (a) and 1T (b) phases in basal plane and cross-section views. Mo, blue; top S, orange; bottom S′ , purple. The incident electron beam transmits from top to bottom. The 2H phase shows a hexagonal lattice with threefold symmetry and the atomic stacking sequence (S–Mo–S′ ) ABA. The 1T phase shows the atomic stacking sequence (S–Mo–S′ ) ABC, with the bottom S′ plane occupying the hollow centre (HC) of √ a 2H hexagonal lattice. c, The S plane glides over a distance equivalent to a/ 3 (a ¼ 3.16 Å). and occupies the HC site of the 2H hexagon, which results in a 2H 1T phase transition. d, Gliding of the Mo plane results in a 2H 2H′ transition. The shadow atomic model shows the original 2H-MoS2 structure. The three planes (Mo, S and S′ ) in single-layer MoS2 can glide individually to give different transitions.
1 National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba 305-8565, Japan, 2 Department of Electronic Engineering, National Taiwan University of Science and Technology, Taipei 10607, Taiwan. † Present address: Electrical Engineering Institute, Ecole Polytechnique Federale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland. * e-mail: [email protected]
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
1
ARTICLES
NATURE NANOTECHNOLOGY
a
0s
b
c
100 s
110 s
d
DOI: 10.1038/NNANO.2014.64
220 s
β
α 60°
2H
Re
α
α
1 nm
e
g
f
h
β
α
γ
1T
1T
2H
α
α j
i
γ
1T
1T
α1
2H
k 2H
Mo 2H
β
S
γ 1T
α2
α2
Figure 2 | Atomic movements during 2H 1T phase transformation in single-layered MoS2 at T 5 600 8 C. a, Single-layered MoS2 doped with Re substitution dopants (indicated by arrowheads) has the initial 2H phase of a hexagonal lattice structure with a clear HC. b, At t ¼ 100 s, two identical intermediate (precursor) phases (denoted a) form with an angle of 608, and consist of three constricted Mo zigzag chains. c, At t ¼ 110 s, a triangular shape indicating the 1T phase (1.08 nm2) appears at the acute corner between the two a-phases. The 1T phase provides noticeable contrast because of the S atoms at the HC sites (Supplementary Fig. 1). d, At t ¼ 220 s, the area of the transformed 1T phase is enlarged to 8.47 nm2. Three different boundaries (a, b and g) are found at the three edges between the 1T and 2H phases. e–h, Simple schematic illustrations of the 2H 1T phase transition corresponding to the ADF images in a–d, respectively. i, Atomic model of a-phase formation by the constriction of three Mo zigzag chains. j, Nucleation of the 1T phase (triangular) with the Mo þ S (or S′ ) atoms gliding in the directions indicated by blue and pink arrows. k, b-Boundary formation at the growth frontier side. The a1-phase transforms to a g-boundary, and the a2-phase becomes wider (Supplementary Fig. 2).
the dynamics of defect movement20,21. In the case of MoS2 , few studies have been carried out, except for those that study defects and the native grain boundary between two MoS2 domains22–24. A MoS2 specimen doped with 0.6 at% Re was exfoliated and transferred to a microgrid25,26. To promote the phase transition, the specimen was heated to 400–700 8C in a microscope to provide thermal activation energy for atom displacement. An example of the phase transition is provided in Fig. 2a–d as sequential annular dark-field (ADF) images, where the step-by-step progress of MoS2 phase transformation at T ¼ 600 8C is represented (see also Supplementary Movie 1). Figure 2e–h presents schematics correlating with the ADF images in Fig. 2a–d to illustrate the structural changes in the MoS2 lattice. A corresponding model of the atomic movements in the 2H 1T phase transition is presented in Fig. 2i–k. The Re dopants (indicated by arrowheads in Fig. 2a) tend to substitute at the Mo sites and display brighter contrast26. The initial MoS2 lattice (Fig. 2a) exhibits the 2H phase, with a honeycomb structure consisting of three Mo atoms and three overlapped S pairs arranged in a hexagon. At t ¼ 100 s, two identical band-like structures (labelled a in Fig. 2b) gradually form along two zigzag directions. This a-phase is a precursor that basically consists of three to four constricted MoS2 zigzag chains (Fig. 2i). When two non-parallel a-phases are in contact, the atoms at the corner formed by the local acute angle are very densely packed, triggering them to glide towards the area with less atomic concentration to release 2
the stress. As a consequence, a triangular 1T phase forms, as shown in Fig. 2c. The 1T phase, with its different S contrast, can be unambiguously discriminated from the 2H phase in the ADF image (Supplementary Fig. 1). After continuous electron-beam scanning, the size of the 1T phase can be gradually enlarged to 8.47 nm2 (Fig. 2d). Two new phase boundaries (b and g in Fig. 2d,k) are found at the edges of the 1T phase. The structures and dynamic behaviours of these boundaries will be discussed in detail in the following. The phase transformation occurs only in the area scanned by the electron beam, and no atom loss is needed to explain this phase transition. For more detailed discussions about the mechanism of phase transition see Supplementary Figs 2 and 3. The phase transformation in single-layered MoS2 involves numerous atomic displacements besides the simple atomic plane gliding. We investigated the atomic process of phase transitions for more than 100 cases using in situ STEM, for cases involving 2H 1T, 1T 2H, 2H 2H′ and 1T 1T′ phase transitions (Table 1). We then tried to categorize these transitions into three important elemental steps: (i) nucleation (formation) of the a-phase as a precursor or an intermediate state, and (ii, iii) migration of the b and g boundaries. Figure 3 shows these three steps with independent sequential ADF images. See also Supplementary Figs 4 and 5 for structure models and image simulations of the a-phase and b- and g-boundaries.
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
NATURE NANOTECHNOLOGY
constriction is probably connected to the S out-of-plane displacement induced by the electron beams. These local strains must be released by changing the Mo–S bond angles, although our STEM observation was not capable of proving the exact strained structure of this a-phase. Figure 3d presents a schematic of the in-plane constriction (green arrows, left) and a structure model of the a-phase (right). The S out-of-plane displacement can propagate in the zigzag direction and so elongate the a-phase (Fig. 3b). Interestingly, the a-phase has a strong tendency to nucleate at the vicinity of Re substitution dopants (Supplementary Fig. 3 and Movie 3). Presumably, the initial out-of-plane protuberance of the Re–S bond could help the S out-of-plane displacement and thus the formation of the a-phase26. Note here that the a-phase always consists of three or four MoS2 zigzag chains and does not expand in width (Supplementary Fig. 6 and Movie 4). There is no atom loss during the formation and elongation (directional growth) of the a-phase. In Fig. 3c, during migration of the central a-phase, the left part of the bottom a-phase indeed disappears and reverts to the original 2H structure (indicated by an arrowhead). Such a reversible phase transformation between the 2H and a-phase proves that there is no massive atom loss during a-phase formation. Even though the electron beam is certainly required to displace the S atoms out-of-plane, no atom is kicked out by the knock-on effect. In the Supplementary Information we show another scenario in
Table 1 | Summary of phase transformation in single-layered MoS2. Phase transition 2H 1T
Gliding plane Mo þ S (or S′ ) S (or S′ ) Mo S þ S′ Mo þ S (or S′ ) S (or S′ ) Mo þ S (or S′ )†
2H 2H′ 1T 2H 1T 1T′
Boundary structures aþbþg a þ g or g* aþbþg g* aþbþg aþg a þ b þ g†
*The special case where no a-phase is formed before the phase transition (Supplementary Fig. 9 and Movies 8 and 9). †The 1T′ phase, a 608 (or 1808) rotation of the 1T phase, has not yet been found. Distinguishing the 1T′ and 1T phases from one another is not straightforward because they have the same symmetry in the STEM image.
a-Phase formation The a-phase is a precursor structure that is essential before phase transition. It is an intermediate state, but forms a stable structure at high temperatures under electron-beam irradiation. In the a-phase, Mo atoms do not show a trigonal arrangement, but align as zigzag chains. The Mo–Mo distance is locally compressed along these zigzag chains, resulting in a local strain in the MoS2 lattice (Fig. 3a, Supplementary Movie 2). The Mo–Mo in-plane
α-Phase formation
a
α
Re
b
16 s
Mo–Mo distance shrinkage
c
117 s
d
110 s
h
α α
e β-Boundary migration
0s
ARTICLES
DOI: 10.1038/NNANO.2014.64
0s
f
β
60 s
Mo + S (or S’) gliding
g
2H 2H
1T
1T
S (or S’) gliding γ-Boundary migration
i
0s
j
γ
20 s
k
30 s
l
2H 1T
1T
2H
Figure 3 | Three elemental steps responsible for phase transitions in single-layered MoS2 (T 5 600 8 C). a–d, a-Phase (three or four zigzag chains) formation. a, Nucleation of an a-phase at an angle of 608 with the other a-phases. The a-phase shows three or four constricted zigzag MoS2 chains. Three white lines highlight the distance between the zigzag chains, with the in-plane constriction in the a-phase being 15% that of the original MoS2. b, Growth of the a-phase. c, At t ¼ 117 s, the a-phase begins to migrate rightward. The left side of the bottom a-phase (indicated by an arrowhead) disappears and reverts to the initial MoS2 lattice. Re dopants are marked by green circles. d, Constriction (green arrows) induces strain in-plane (left), and the model a-phase forms with a reduced Mo–Mo distance (right). The S atoms in the a-phase are also misaligned vertically. e–h, b-Boundary migration. e, Singlelayered MoS2 with 2H phase. The orientation of the initial 2H phase is indicated by the blue triangle. f, At t ¼ 60 s, the b-boundary (highlighted by yellow shading) appears in the middle of the 2H-MoS2. The left-hand side of the b-boundary demonstrates the 1T phase. g, The b-boundary migrates rightward and the 1T phase is enlarged. h, Schematic model before (top) and after (bottom) gliding of the Mo þ S (or S′ ) atoms, which causes b-boundary migration. In e–g the ADF images are filtered by a local two-dimensional Wiener filter to enhance the contrast. i–l, g-Boundary (two zigzag chains) migration. i, Singlelayered MoS2 with 1T phase (a-phase is also visible). j, At t ¼ 20 s, a g-boundary (highlighted by purple shading) appears in the middle. The left-hand side of the g-boundary demonstrates the nucleated 2H phase. k, The g-boundary migrates rightward and has a non-straight structure. l, Schematic model before (top) and after (bottom) gliding of top S (or S′ ) atoms, which drives g-boundary migration. Scale bars, 1 nm. NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
3
ARTICLES
NATURE NANOTECHNOLOGY
Transformed area (nm2)
a 30
20
d
b γ
2H
400 °C-0119 400 °C-0120 400 °C-0121 500 °C-0100 500 °C-0101 500 °C-0102 500 °C-0106 500 °C-0107 600 °C-0020 600 °C-0269 700 °C-0006
DOI: 10.1038/NNANO.2014.64
1T
S α
(II) Semiconductor
Mo QW
α
Metal S
e S γ
β
10
c (I)
β M
M α
β S
M
f
0
M β
β 0
50
100
150
S
Electron dose (Me nm−2)
γ
Figure 4 | Time dependence of phase transformation process and fabrication of nanodevices in single-layered MoS2. a, Area of transformed phase as a function of total electron dose (instead of time). The dispersion is divided into two regions by the threshold electron dose (grey dotted line). Region I comprises the initial step to create the intermediate structures, a-phase. In region II, the phase transition is initiated and the transformed area increases with increasing electron dose. Data sets were recorded at various temperatures ranging from 400 8C to 700 8C. The four digit number after the temperature in the legend indicates the experiment batch number. b–f, Attempts to create prototypes of the nanodevices: 2H and 1T heterostructure with a g-boundary, as a Schottky diode (b); 2H sandwiched between two 1T phases with two b-boundaries, as a Schottky barrier nanotransistor (c); single Mo hexagon chain formed on top of the 2H-MoS2 , as a metallic quantum wire (d) (this is an exceptional case and its formation mechanism is still unclear); 1T phase embedded in a 2H phase, as an embedded metallic quantum dot (e); 2H phase embedded in a 1T phase, as an embedded semiconducting quantum dot (f). Scale bars, 1 nm.
which prolonged electron-beam irradiation occasionally leads to the loss of a MoS2 zigzag chain and the formation an agglomerated structure on the MoS2 surface (Supplementary Fig. 7 and Movie 5). The a-phase barely forms at room temperature (not shown).
b-Boundary migration
The b-boundary between the 2H/2H′ interfaces is a twin boundary containing the Mo–S four-membered rings, which in principle is the same as that recently observed at the boundary of two 608 rotated MoS2 domains (that is, 2H and 2H′ ) synthesized by chemical vapour deposition23. The S atoms in the b-boundary are four-coordinated despite all the other phases having three-coordinated S atoms. The 2H 2H′ phase transition requires Mo plane gliding and generates a b-boundary (Supplementary Fig. 8 and Movie 6). The b-boundary can also be found between 2H and 1T phases (Fig. 2d); in this case, it is no longer a twin boundary. For example, Fig. 3e shows a typical ADF image of MoS2 in the 2H phase with a threefold symmetry (orientation described by a blue triangle). At t ¼ 60 s, the b-boundary (highlighted by yellow shading) appears in the middle of the 2H-MoS2 (Fig. 3f ). The left-hand side of the b-boundary becomes 1T phase. Note that the b-boundary shows up when a Mo plane and a S plane both glide during the 2H 1T transition. Another simpler transition from 2H to 1T with only one S-plane glide results in the formation of a g-boundary between the phases (Table 1 and Supplementary Movie 9). The Mo þ S (or S′ ) atoms gliding across the b-boundary (Fig. 3h, left) then drive the b-boundary to migrate (Fig. 3h, right). Figure 3g shows the b-boundary migrating rightward at t ¼ 110 s. For a detailed structure and gliding model, see Supplementary Fig. 11.
g-Boundary migration The g-boundary consists of two constricted MoS2 zigzag chains. The a-phase is made of exclusively three or four MoS2 zigzag chains, but the g-boundary always has two chains. The S atoms at the g-boundary remain three-coordinated to the Mo atoms. Figure 3i presents an ADF image of MoS2 in the initial 1T phase. 4
The two S planes are misaligned vertically in the 1T phase. At t ¼ 20 s, a g-boundary (highlighted by purple shading) appears between the initial 1T phase and the nucleated 2H phase (Fig. 3j). The left-hand side of the g-boundary transforms from 1T to 2H by S plane gliding. The schematic model shown in Fig. 3l (left) illustrates the S atoms sequentially gliding towards the g-boundary, and the 2H phase then increasing with g-boundary migration (Fig. 3l, right). The g-boundary migrates rightward step by step (the non-straight boundary showing the atomic step is indicated by an arrow in Fig. 3k; Supplementary Movie 7). For a detailed structure and gliding model see Supplementary Fig. 11. The distinct features of the boundary structure were also corroborated by electron energy loss spectroscopy (EELS). Recent literature has reported single impurity atoms of Si in a graphene lattice discriminated in three- and four-coordinated configurations27,28. Accordingly, the bonding states of the S atoms in the newly discovered boundaries are intriguing and would be most prominent in the S L-edge. The electron energy loss near-edge structures (ENLES) for the S L23-edge and Mo M45-edge were recorded in the b- and g-boundary regions, and are shown in Supplementary Fig. 12. The phase transformation presented here involves complicated dynamic processes. The gliding planes decide the relationship of the initial and final phases, as well as the correlated phase boundaries. Table 1 catalogues the results of a systematic investigation of the discovered phase transformations in single-layered MoS2 , and all the corresponding schematic models and detailed discussions are presented in the Supplementary Information. In an attempt to control the phase transformation by the electron beam, we continuously recorded the size of the transformed area as a function of time. The data points plotted in Fig. 4a shows the relation between the electron dose and the area of the transformed phase in different thermal environments (400 8C , T , 700 8C). We used the dose instead of time, because the data were normalized by the dose rate for the unit area. The phase transformation area A increases as A / es(D−D0 ) , where D is the electron dose, D0 ≈ 40 MeV nm22 is the threshold before triggering the phase transformation, and s ≈ 0.028–0.061. The relation between the transformed area and
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
NATURE NANOTECHNOLOGY
DOI: 10.1038/NNANO.2014.64
the electron dose can be divided into two regions by D0. In region I, D , D0. , the phase transformation does not start until the electron dose creates the intermediate-state structures, the a-phases. In region II, D . D0 , the phase transition can be triggered in the local area in the temperature range 400–700 8C, and the phase transformation starts to enlarge with the increasing dose. Previous studies have suggested that the 2H 1T transformation is triggered by a high doping concentration12. In our experiment, Re is an n-type dopant26, but the doping rate is relatively small (,1 at%). Accordingly, we can reasonably infer that the continuous electron-beam irradiation may play an electronic role in accumulating negative charge to trigger the phase transition.
Single layers with semiconducting and metallic domains Because the electron beam scanning area and the irradiation time can be controlled easily in a STEM, we can intentionally introduce the phase transition in a chosen area with a predetermined size. Because the 1T and 2H phases have distinct electronic properties, this controllable local phase transition may enable bottom-up processes in the fabrication of nanoelectronics. To explore these possibilities, in Fig. 4 we demonstrate several attempts to produce prototypes of nanodevices. Figure 4b shows a serial junction of semiconductor and metallic phases, which can be regarded as a Schottky diode. A local semiconductor region sandwiched between two metallic electrodes forms a nanoscale transistor (Fig. 4c). A metallic wire can be embedded in the semiconducting matrix as a quantum lead (Fig. 4d). Finally, quantum dots (in a triangular shape) can be stably produced in the initial phase, with a metallic quantum dot embedded in semiconductor (Fig. 4e), or vice versa (Fig. 4f ). To date, these structures have been fabricated only in an electron microscope and their functions have not yet been confirmed experimentally. The transfer process and the prevention of surface contamination remain obstacles to be overcome. The relatively stable single-layered structures of this system are, however, very promising in the quest to obtain the first single-layered electronic device. Although the low-dimensional nanodevice was first proposed using nanotubes of metallic and semiconducting components, it has turned out to be difficult to realize nanotube composites with controlled chiralities. Patterning single layers is definitely a more promising approach towards the realization of nanodevices. Indeed, it would be extremely intriguing to explore similar phenomena in MoWS2 alloys with tunable bandgaps29 and in n- and p-type doped dichalcogenides26.
Methods Material synthesis and specimen preparation. A single crystal of Re-doped MoS2 was grown by a chemical vapour transport method using Br2 as a transport agent at 950 8C (ref. 25). Mo, S and Re elements (99.99% purity, 10 g in total) containing Br2 (5 mg cm23) were cooled in a quartz ampoule with liquid nitrogen and sealed in vacuum (1 × 1026 torr). Single-crystalline Re-doped MoS2 flakes (3 × 3 mm2 surface area, 0.5 mm thickness) were mechanically exfoliated by Scotch tape and transferred to the surface of a Si substrate with 300 nm thermal oxide. The target single-layer flakes were transferred to a Mo quantifoil grid with 2-propanol and cleaned with chloroform26. The specimens were heated in the TEM chamber (vacuum of 1.7 × 1025 Pa) overnight at 550 8C in a JEOL heating holder to remove residual contamination. ADF-STEM. ADF images were obtained using an aberration-corrected JEOL-2100F cold field-emission gun electron microscope equipped with a DELTA corrector and operated at an accelerating voltage of 60 kV. The convergence semi-angle and inner acquisition semi-angle were 35 and 79 mrad for the ADF imaging. The electron-beam current was 10–15 pA. The dwell time per pixel was 38–76 ms and the pixel size 0.0139–0.026 nm for sequential images (movies), corresponding to an electron dose of 0.7–1.8 × 107 e nm22. Electron beam damage on the MoS2 was observed when the total electron dose exceeded 5 × 108 to 11 × 108 e nm22. Surface cleanliness was critical to preventing damage, and hydrocarbon contamination on the surface enhanced damage development. False-colour images and the alignment of sequential images (movies) were performed using ImageJ. Figure 3e–g presents filtered images using a local two-dimensional Wiener filter to enhance the contrast.
ARTICLES
EELS. Data were recorded using a Gatan Quantum camera. The EEL image spectra in Fig. 4 consist of 12 × 12 pixels, obtained using a 0.1 nm probe with 0.05 nm increments for each step. Each spectrum was acquired in 0.5 s and was summed in the vertical direction to increase the signal-to-noise ratio. The energy dispersion for the recorded spectra was 0.25 eV and the zero-loss width of the incident electron beam was 0.35 eV. The EELS collection semi-angle was 79 mrad. Structure modelling and image simulations. MoS2 polymorphous and phase boundary models were constructed using CrystalMaker and geometry optimizations using HyperChem. ADF image simulations were carried out using QSTEM with a probe size of 1 Å (spherical aberration coefficient, Cs ¼ 1 mm, Scherzer defocus ¼ 24 nm).
Received 21 September 2013; accepted 4 March 2014; published online 20 April 2014
References 1. Winer, W. O. Molybdenum disulphide as a lubricant: a review of the fundamental knowledge. Wear 10, 422–452 (1967). 2. Holinski, R. & Gansheimer, J. A study of the lubricating mechanism of molybdenum disulphide. Wear 19, 329–342 (1972). 3. Radisavljevic, B., Radenovic, A., Brivio, J., Giacometti, V. & Kis, A. Single-layer MoS2 transistors. Nature Nanotech. 6, 147–150 (2011). 4. Wang, Q. H., Kalantar-Zadeh, K., Kis, A., Coleman, J. N. & Strano M. S. Electronics and optoelectronics of two-dimensional transition metal dichalcogenides. Nature Nanotech. 7, 699–712 (2012). 5. Chhowalla, M. et al. The chemistry of two-dimensional layered transition metal dichalcogenide nanosheets. Nature Chem. 5, 263–275 (2013). 6. Matthesis, L. F. Band structure of transition-metal–dichalcogenide layer compounds. Phys. Rev. B 8, 3719–3740 (1973). 7. Wypych, F. & Scho¨llhorn, R. 1T-MoS2 , a new metallic modification of molybdenum disulfide. J. Chem. Soc. Chem. Commun. 1386–1388 (1992). 8. Bissessur, R., Kanatzidis, M. G., Schindler, J. L. & Kannewurf, C. R. Encapsulation of polymers into MoS2 and metal to insulator transition in metastable MoS2. J. Chem. Soc. Chem. Commun. 1582–1585 (1993). 9. Py, M. A. & Haering, R. R. Structural destabilization induced by lithium intercalation in MoS2 and related compounds. Can. J. Phys. 61, 76–84 (1983). 10. Heising, J. & Kanatzidis, M. G. Structure of restacked MoS2 and WS2 elucidated by electron crystallography. J. Am. Chem. Soc. 121, 638–643 (1999). 11. Wypych, F., Solenthaler, C., Prins, R. & Weber, Th. Electron diffraction study of intercalation compounds derived from 1T-MoS2. J. Solid State Chem. 144, 430–436 (1999). 12. Enyashin, A. N. et al. New route for stabilization of 1T-WS2 and MoS2 phases. J. Phys. Chem. C. 115, 24586–24591 (2011). 13. Sandoval, S. J., Yang, D., Frindt, R. F. & Irwin, J. C. Raman study and lattice dynamics of single molecular layers of MoS2. Phys. Rev. B. 44, 3955–3962 (1991). 14. Eda, G. et al. Coherent atomic and electronic heterostructures of single-layer MoS2. ACS Nano 6, 7311–7317 (2012). 15. Hashimoto, A., Suenaga, K., Gloter, A., Urtia, K. & Iijima. S. Direct evidence for atomic defects in graphene layers. Nature 430, 870–873 (2004). 16. Kotakoski, J., Krasheninnikov, A. V., Kaiser, U. & Meyer, J. C. From point defects in graphene to two-dimensional amorphous carbon. Phys. Rev. Lett. 106, 105505 (2011). 17. Huang, P. Y. et al. Grains and grain boundaries in single-layer graphene atomic patchwork quilts. Nature 469, 389–392 (2011). 18. Kurasch, S. et al. Atom-by-atom observation of grain boundary migration in graphene. Nano Lett. 12, 3168–3173 (2012). 19. Kim, K. et al. Grain boundary mapping in polycrystalline graphene. ACS Nano 5, 2142–2146 (2011). 20. Warner, J. H. et al. Dislocation-driven deformations in graphene. Science 337, 209–212 (2012). 21. Lehitnen, O., Kurasch, S., Krasheninnikov, A. V. & Kaiser. U. Atomic scale study of the life cycle of a dislocation in graphene from birth to annihilation. Nature Commun. 4, 2098 (2013). 22. Van der Zande, A. M. et al. Grains and grain boundaries in highly crystalline monolayer molybdenum disulfide. Nature Mater. 12, 554–561 (2013). 23. Zhou, W. et al. Intrinsic structural defects in monolayer molybdenum disulfide. Nano Lett. 13, 2615–2622 (2013). 24. Najmaei, S. et al. Vapor phase growth and grain boundary structure of molybdenum disulphide atomic layers. Nature Mater. 12, 754–759 (2013). 25. Tiong, K. K., Huang, Y. S. & Ho, C. H. Electrical and optical anisotropic properties of rhenium-doped molybdenum disulphide. J. Alloys Comp. 317–318, 208–212 (2001). 26. Lin, Y. C. et al. Properties of individual dopant atoms in single-layer MoS2: atomic structure, migration, and enhanced reactivity. Adv. Mater. http://dx.doi.org/10.1002/adma.201304985 (2014).
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
5
ARTICLES
NATURE NANOTECHNOLOGY
27. Zhou, W. et al. Direct determination of the chemical bonding of individual impurities in graphene. Phys. Rev. Lett. 109, 206803 (2012). 28. Ramasse, Q. M. et al. Probing the bonding and electronic structure of single atom dopants in graphene with electron energy loss spectroscopy. Nano Lett. 13, 4989–4995 (2013). 29. Chen, Y. et al. Tunable band gap photoluminescence from atomically thin transition-metal dichalcogenide alloys. ACS Nano 5, 4610–4616 (2013).
Acknowledgements The authors from AIST acknowledge support from the JST Research Acceleration Programme. D.O.D. and Y.S.H. acknowledge the support of the National Science Council of Taiwan (projects NSC 100-2112-M-011-001-MY3 and NSC 101-2811-M-011-002).
6
DOI: 10.1038/NNANO.2014.64
Author contributions Y.C.L. performed experiments and analysed data. D.O.D. and Y.S.H. grew materials. K.S. and Y.C.L. designed experiments. Y.C.L. and K.S. co-wrote the manuscript.
Additional information Supplementary information is available in the online version of the paper. Reprints and permissions information is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to K.S.
Competing financial interests The authors declare no competing financial interests.
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
Atomic mechanism of the semiconducting-tometallic phase transition in single-layered MoS2 Yung-Chang Lin1, Dumitru O. Dumcenco2†, Ying-Sheng Huang2 and Kazu Suenaga1 * Phase transitions can be used to alter the properties of a material without adding any additional atoms and are therefore of significant technological value. In a solid, phase transitions involve collective atomic displacements, but such atomic processes have so far only been investigated using macroscopic approaches. Here, we show that in situ scanning transmission electron microscopy can be used to follow the structural transformation between semiconducting (2H) and metallic (1T) phases in single-layered MoS2 , with atomic resolution. The 2H/1T phase transition involves gliding atomic planes of sulphur and/or molybdenum and requires an intermediate phase (a-phase) as a precursor. The migration of two kinds of boundaries (b- and g-boundaries) is also found to be responsible for the growth of the second phase. Furthermore, we show that areas of the 1T phase can be controllably grown in a layer of the 2H phase using an electron beam.
F
In situ observation of 2H/1T phase transition Here, we provide in situ observations of the transformation process between 2H and 1T phases in single-layered MoS2 at high temperatures. To monitor the phase transition in situ, we operated an aberration-corrected scanning transmission electron microscope (STEM) at 60 kV to visualize the dynamic process of the atomic motions in single-layered MoS2. This technique has already been used and verified while studying another ideal two-dimensional material, graphene, for dislocations15,16, grain boundaries17–19 and a
b
2H
1T
c S plane gliding (2H → 1T)
3.16 Å
or several centuries, molybdenum disulphide (MoS2) has been widely used as a practical solid lubricant1,2. MoS2 crystal is composed of stacks of atomic layers bound by van der Waals forces, with each layer constructed from S–Mo–S′ triple atomic planes with strong in-plane bonding. Recently, single-layered MoS2 , a direct-bandgap quasi-two-dimensional semiconductor, has shown its great potential for applications in electrical and optoelectronic devices3–5. Interestingly, one of the unique features of MoS2 is polymorphism, with its distinct electronic characteristics. Depending on the arrangement of its S atoms, single-layered MoS2 appears in two distinct symmetries: the 2H (trigonal prismatic D3h) and 1T (octahedral Oh) phases (Fig. 1a,b). The two phases should exhibit completely different electronic structures, with the 2H phase being semiconducting and the 1T phase metallic6–8. The two phases can easily convert one to the other via intralayer atomic plane gliding, which involves a transversal displacement of one of the S planes. The 1T phase was first reported to transform from 2H-MoS2 by Li and K intercalation6,9, with restacked 1T phases in LiMoS2 and KMoS2 confirmed by electron diffraction10,11, and it is also known to be stabilized by substitutional doping of Re, Tc and Mn atoms, which serve as electron donors12. However, 1T-LiMoS2 is thermodynamically unfavourable, and has been observed (by Raman spectroscopy13) to gradually transform to the 2H phase at room temperature. The phase transitions between 2H and 1T or 2H′ phases due to atomic plane gliding are presented in Fig. 1c,d. Note that 2H′ is a 608 (or 1808) rotational phase of 2H. Although the coexistence of metallic and semiconducting phases has indeed been reported in chemically exfoliated MoS2 by Eda and colleagues14, the actual dynamical process of the transformation between 2H and 1T phases involving intralayer atomic plane gliding has never been experimentally proven, nor has the atomic process of the phase transition been investigated in situ. If one is to consider the possibility of intentionally introducing the phase transition in single-layered materials in a controllable manner, the atomic process of this phase transition as well as its boundary structures must be corroborated in order to reliably design future low-dimensional devices.
d Mo plane gliding (2H → 2H’)
AB
ABC S Mo S’
Figure 1 | Polymorphs of single-layered MoS2. a,b, Schematic models of single-layered MoS2 with 2H (a) and 1T (b) phases in basal plane and cross-section views. Mo, blue; top S, orange; bottom S′ , purple. The incident electron beam transmits from top to bottom. The 2H phase shows a hexagonal lattice with threefold symmetry and the atomic stacking sequence (S–Mo–S′ ) ABA. The 1T phase shows the atomic stacking sequence (S–Mo–S′ ) ABC, with the bottom S′ plane occupying the hollow centre (HC) of √ a 2H hexagonal lattice. c, The S plane glides over a distance equivalent to a/ 3 (a ¼ 3.16 Å). and occupies the HC site of the 2H hexagon, which results in a 2H 1T phase transition. d, Gliding of the Mo plane results in a 2H 2H′ transition. The shadow atomic model shows the original 2H-MoS2 structure. The three planes (Mo, S and S′ ) in single-layer MoS2 can glide individually to give different transitions.
1 National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba 305-8565, Japan, 2 Department of Electronic Engineering, National Taiwan University of Science and Technology, Taipei 10607, Taiwan. † Present address: Electrical Engineering Institute, Ecole Polytechnique Federale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland. * e-mail: [email protected]
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
1
ARTICLES
NATURE NANOTECHNOLOGY
a
0s
b
c
100 s
110 s
d
DOI: 10.1038/NNANO.2014.64
220 s
β
α 60°
2H
Re
α
α
1 nm
e
g
f
h
β
α
γ
1T
1T
2H
α
α j
i
γ
1T
1T
α1
2H
k 2H
Mo 2H
β
S
γ 1T
α2
α2
Figure 2 | Atomic movements during 2H 1T phase transformation in single-layered MoS2 at T 5 600 8 C. a, Single-layered MoS2 doped with Re substitution dopants (indicated by arrowheads) has the initial 2H phase of a hexagonal lattice structure with a clear HC. b, At t ¼ 100 s, two identical intermediate (precursor) phases (denoted a) form with an angle of 608, and consist of three constricted Mo zigzag chains. c, At t ¼ 110 s, a triangular shape indicating the 1T phase (1.08 nm2) appears at the acute corner between the two a-phases. The 1T phase provides noticeable contrast because of the S atoms at the HC sites (Supplementary Fig. 1). d, At t ¼ 220 s, the area of the transformed 1T phase is enlarged to 8.47 nm2. Three different boundaries (a, b and g) are found at the three edges between the 1T and 2H phases. e–h, Simple schematic illustrations of the 2H 1T phase transition corresponding to the ADF images in a–d, respectively. i, Atomic model of a-phase formation by the constriction of three Mo zigzag chains. j, Nucleation of the 1T phase (triangular) with the Mo þ S (or S′ ) atoms gliding in the directions indicated by blue and pink arrows. k, b-Boundary formation at the growth frontier side. The a1-phase transforms to a g-boundary, and the a2-phase becomes wider (Supplementary Fig. 2).
the dynamics of defect movement20,21. In the case of MoS2 , few studies have been carried out, except for those that study defects and the native grain boundary between two MoS2 domains22–24. A MoS2 specimen doped with 0.6 at% Re was exfoliated and transferred to a microgrid25,26. To promote the phase transition, the specimen was heated to 400–700 8C in a microscope to provide thermal activation energy for atom displacement. An example of the phase transition is provided in Fig. 2a–d as sequential annular dark-field (ADF) images, where the step-by-step progress of MoS2 phase transformation at T ¼ 600 8C is represented (see also Supplementary Movie 1). Figure 2e–h presents schematics correlating with the ADF images in Fig. 2a–d to illustrate the structural changes in the MoS2 lattice. A corresponding model of the atomic movements in the 2H 1T phase transition is presented in Fig. 2i–k. The Re dopants (indicated by arrowheads in Fig. 2a) tend to substitute at the Mo sites and display brighter contrast26. The initial MoS2 lattice (Fig. 2a) exhibits the 2H phase, with a honeycomb structure consisting of three Mo atoms and three overlapped S pairs arranged in a hexagon. At t ¼ 100 s, two identical band-like structures (labelled a in Fig. 2b) gradually form along two zigzag directions. This a-phase is a precursor that basically consists of three to four constricted MoS2 zigzag chains (Fig. 2i). When two non-parallel a-phases are in contact, the atoms at the corner formed by the local acute angle are very densely packed, triggering them to glide towards the area with less atomic concentration to release 2
the stress. As a consequence, a triangular 1T phase forms, as shown in Fig. 2c. The 1T phase, with its different S contrast, can be unambiguously discriminated from the 2H phase in the ADF image (Supplementary Fig. 1). After continuous electron-beam scanning, the size of the 1T phase can be gradually enlarged to 8.47 nm2 (Fig. 2d). Two new phase boundaries (b and g in Fig. 2d,k) are found at the edges of the 1T phase. The structures and dynamic behaviours of these boundaries will be discussed in detail in the following. The phase transformation occurs only in the area scanned by the electron beam, and no atom loss is needed to explain this phase transition. For more detailed discussions about the mechanism of phase transition see Supplementary Figs 2 and 3. The phase transformation in single-layered MoS2 involves numerous atomic displacements besides the simple atomic plane gliding. We investigated the atomic process of phase transitions for more than 100 cases using in situ STEM, for cases involving 2H 1T, 1T 2H, 2H 2H′ and 1T 1T′ phase transitions (Table 1). We then tried to categorize these transitions into three important elemental steps: (i) nucleation (formation) of the a-phase as a precursor or an intermediate state, and (ii, iii) migration of the b and g boundaries. Figure 3 shows these three steps with independent sequential ADF images. See also Supplementary Figs 4 and 5 for structure models and image simulations of the a-phase and b- and g-boundaries.
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
NATURE NANOTECHNOLOGY
constriction is probably connected to the S out-of-plane displacement induced by the electron beams. These local strains must be released by changing the Mo–S bond angles, although our STEM observation was not capable of proving the exact strained structure of this a-phase. Figure 3d presents a schematic of the in-plane constriction (green arrows, left) and a structure model of the a-phase (right). The S out-of-plane displacement can propagate in the zigzag direction and so elongate the a-phase (Fig. 3b). Interestingly, the a-phase has a strong tendency to nucleate at the vicinity of Re substitution dopants (Supplementary Fig. 3 and Movie 3). Presumably, the initial out-of-plane protuberance of the Re–S bond could help the S out-of-plane displacement and thus the formation of the a-phase26. Note here that the a-phase always consists of three or four MoS2 zigzag chains and does not expand in width (Supplementary Fig. 6 and Movie 4). There is no atom loss during the formation and elongation (directional growth) of the a-phase. In Fig. 3c, during migration of the central a-phase, the left part of the bottom a-phase indeed disappears and reverts to the original 2H structure (indicated by an arrowhead). Such a reversible phase transformation between the 2H and a-phase proves that there is no massive atom loss during a-phase formation. Even though the electron beam is certainly required to displace the S atoms out-of-plane, no atom is kicked out by the knock-on effect. In the Supplementary Information we show another scenario in
Table 1 | Summary of phase transformation in single-layered MoS2. Phase transition 2H 1T
Gliding plane Mo þ S (or S′ ) S (or S′ ) Mo S þ S′ Mo þ S (or S′ ) S (or S′ ) Mo þ S (or S′ )†
2H 2H′ 1T 2H 1T 1T′
Boundary structures aþbþg a þ g or g* aþbþg g* aþbþg aþg a þ b þ g†
*The special case where no a-phase is formed before the phase transition (Supplementary Fig. 9 and Movies 8 and 9). †The 1T′ phase, a 608 (or 1808) rotation of the 1T phase, has not yet been found. Distinguishing the 1T′ and 1T phases from one another is not straightforward because they have the same symmetry in the STEM image.
a-Phase formation The a-phase is a precursor structure that is essential before phase transition. It is an intermediate state, but forms a stable structure at high temperatures under electron-beam irradiation. In the a-phase, Mo atoms do not show a trigonal arrangement, but align as zigzag chains. The Mo–Mo distance is locally compressed along these zigzag chains, resulting in a local strain in the MoS2 lattice (Fig. 3a, Supplementary Movie 2). The Mo–Mo in-plane
α-Phase formation
a
α
Re
b
16 s
Mo–Mo distance shrinkage
c
117 s
d
110 s
h
α α
e β-Boundary migration
0s
ARTICLES
DOI: 10.1038/NNANO.2014.64
0s
f
β
60 s
Mo + S (or S’) gliding
g
2H 2H
1T
1T
S (or S’) gliding γ-Boundary migration
i
0s
j
γ
20 s
k
30 s
l
2H 1T
1T
2H
Figure 3 | Three elemental steps responsible for phase transitions in single-layered MoS2 (T 5 600 8 C). a–d, a-Phase (three or four zigzag chains) formation. a, Nucleation of an a-phase at an angle of 608 with the other a-phases. The a-phase shows three or four constricted zigzag MoS2 chains. Three white lines highlight the distance between the zigzag chains, with the in-plane constriction in the a-phase being 15% that of the original MoS2. b, Growth of the a-phase. c, At t ¼ 117 s, the a-phase begins to migrate rightward. The left side of the bottom a-phase (indicated by an arrowhead) disappears and reverts to the initial MoS2 lattice. Re dopants are marked by green circles. d, Constriction (green arrows) induces strain in-plane (left), and the model a-phase forms with a reduced Mo–Mo distance (right). The S atoms in the a-phase are also misaligned vertically. e–h, b-Boundary migration. e, Singlelayered MoS2 with 2H phase. The orientation of the initial 2H phase is indicated by the blue triangle. f, At t ¼ 60 s, the b-boundary (highlighted by yellow shading) appears in the middle of the 2H-MoS2. The left-hand side of the b-boundary demonstrates the 1T phase. g, The b-boundary migrates rightward and the 1T phase is enlarged. h, Schematic model before (top) and after (bottom) gliding of the Mo þ S (or S′ ) atoms, which causes b-boundary migration. In e–g the ADF images are filtered by a local two-dimensional Wiener filter to enhance the contrast. i–l, g-Boundary (two zigzag chains) migration. i, Singlelayered MoS2 with 1T phase (a-phase is also visible). j, At t ¼ 20 s, a g-boundary (highlighted by purple shading) appears in the middle. The left-hand side of the g-boundary demonstrates the nucleated 2H phase. k, The g-boundary migrates rightward and has a non-straight structure. l, Schematic model before (top) and after (bottom) gliding of top S (or S′ ) atoms, which drives g-boundary migration. Scale bars, 1 nm. NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
3
ARTICLES
NATURE NANOTECHNOLOGY
Transformed area (nm2)
a 30
20
d
b γ
2H
400 °C-0119 400 °C-0120 400 °C-0121 500 °C-0100 500 °C-0101 500 °C-0102 500 °C-0106 500 °C-0107 600 °C-0020 600 °C-0269 700 °C-0006
DOI: 10.1038/NNANO.2014.64
1T
S α
(II) Semiconductor
Mo QW
α
Metal S
e S γ
β
10
c (I)
β M
M α
β S
M
f
0
M β
β 0
50
100
150
S
Electron dose (Me nm−2)
γ
Figure 4 | Time dependence of phase transformation process and fabrication of nanodevices in single-layered MoS2. a, Area of transformed phase as a function of total electron dose (instead of time). The dispersion is divided into two regions by the threshold electron dose (grey dotted line). Region I comprises the initial step to create the intermediate structures, a-phase. In region II, the phase transition is initiated and the transformed area increases with increasing electron dose. Data sets were recorded at various temperatures ranging from 400 8C to 700 8C. The four digit number after the temperature in the legend indicates the experiment batch number. b–f, Attempts to create prototypes of the nanodevices: 2H and 1T heterostructure with a g-boundary, as a Schottky diode (b); 2H sandwiched between two 1T phases with two b-boundaries, as a Schottky barrier nanotransistor (c); single Mo hexagon chain formed on top of the 2H-MoS2 , as a metallic quantum wire (d) (this is an exceptional case and its formation mechanism is still unclear); 1T phase embedded in a 2H phase, as an embedded metallic quantum dot (e); 2H phase embedded in a 1T phase, as an embedded semiconducting quantum dot (f). Scale bars, 1 nm.
which prolonged electron-beam irradiation occasionally leads to the loss of a MoS2 zigzag chain and the formation an agglomerated structure on the MoS2 surface (Supplementary Fig. 7 and Movie 5). The a-phase barely forms at room temperature (not shown).
b-Boundary migration
The b-boundary between the 2H/2H′ interfaces is a twin boundary containing the Mo–S four-membered rings, which in principle is the same as that recently observed at the boundary of two 608 rotated MoS2 domains (that is, 2H and 2H′ ) synthesized by chemical vapour deposition23. The S atoms in the b-boundary are four-coordinated despite all the other phases having three-coordinated S atoms. The 2H 2H′ phase transition requires Mo plane gliding and generates a b-boundary (Supplementary Fig. 8 and Movie 6). The b-boundary can also be found between 2H and 1T phases (Fig. 2d); in this case, it is no longer a twin boundary. For example, Fig. 3e shows a typical ADF image of MoS2 in the 2H phase with a threefold symmetry (orientation described by a blue triangle). At t ¼ 60 s, the b-boundary (highlighted by yellow shading) appears in the middle of the 2H-MoS2 (Fig. 3f ). The left-hand side of the b-boundary becomes 1T phase. Note that the b-boundary shows up when a Mo plane and a S plane both glide during the 2H 1T transition. Another simpler transition from 2H to 1T with only one S-plane glide results in the formation of a g-boundary between the phases (Table 1 and Supplementary Movie 9). The Mo þ S (or S′ ) atoms gliding across the b-boundary (Fig. 3h, left) then drive the b-boundary to migrate (Fig. 3h, right). Figure 3g shows the b-boundary migrating rightward at t ¼ 110 s. For a detailed structure and gliding model, see Supplementary Fig. 11.
g-Boundary migration The g-boundary consists of two constricted MoS2 zigzag chains. The a-phase is made of exclusively three or four MoS2 zigzag chains, but the g-boundary always has two chains. The S atoms at the g-boundary remain three-coordinated to the Mo atoms. Figure 3i presents an ADF image of MoS2 in the initial 1T phase. 4
The two S planes are misaligned vertically in the 1T phase. At t ¼ 20 s, a g-boundary (highlighted by purple shading) appears between the initial 1T phase and the nucleated 2H phase (Fig. 3j). The left-hand side of the g-boundary transforms from 1T to 2H by S plane gliding. The schematic model shown in Fig. 3l (left) illustrates the S atoms sequentially gliding towards the g-boundary, and the 2H phase then increasing with g-boundary migration (Fig. 3l, right). The g-boundary migrates rightward step by step (the non-straight boundary showing the atomic step is indicated by an arrow in Fig. 3k; Supplementary Movie 7). For a detailed structure and gliding model see Supplementary Fig. 11. The distinct features of the boundary structure were also corroborated by electron energy loss spectroscopy (EELS). Recent literature has reported single impurity atoms of Si in a graphene lattice discriminated in three- and four-coordinated configurations27,28. Accordingly, the bonding states of the S atoms in the newly discovered boundaries are intriguing and would be most prominent in the S L-edge. The electron energy loss near-edge structures (ENLES) for the S L23-edge and Mo M45-edge were recorded in the b- and g-boundary regions, and are shown in Supplementary Fig. 12. The phase transformation presented here involves complicated dynamic processes. The gliding planes decide the relationship of the initial and final phases, as well as the correlated phase boundaries. Table 1 catalogues the results of a systematic investigation of the discovered phase transformations in single-layered MoS2 , and all the corresponding schematic models and detailed discussions are presented in the Supplementary Information. In an attempt to control the phase transformation by the electron beam, we continuously recorded the size of the transformed area as a function of time. The data points plotted in Fig. 4a shows the relation between the electron dose and the area of the transformed phase in different thermal environments (400 8C , T , 700 8C). We used the dose instead of time, because the data were normalized by the dose rate for the unit area. The phase transformation area A increases as A / es(D−D0 ) , where D is the electron dose, D0 ≈ 40 MeV nm22 is the threshold before triggering the phase transformation, and s ≈ 0.028–0.061. The relation between the transformed area and
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
NATURE NANOTECHNOLOGY
DOI: 10.1038/NNANO.2014.64
the electron dose can be divided into two regions by D0. In region I, D , D0. , the phase transformation does not start until the electron dose creates the intermediate-state structures, the a-phases. In region II, D . D0 , the phase transition can be triggered in the local area in the temperature range 400–700 8C, and the phase transformation starts to enlarge with the increasing dose. Previous studies have suggested that the 2H 1T transformation is triggered by a high doping concentration12. In our experiment, Re is an n-type dopant26, but the doping rate is relatively small (,1 at%). Accordingly, we can reasonably infer that the continuous electron-beam irradiation may play an electronic role in accumulating negative charge to trigger the phase transition.
Single layers with semiconducting and metallic domains Because the electron beam scanning area and the irradiation time can be controlled easily in a STEM, we can intentionally introduce the phase transition in a chosen area with a predetermined size. Because the 1T and 2H phases have distinct electronic properties, this controllable local phase transition may enable bottom-up processes in the fabrication of nanoelectronics. To explore these possibilities, in Fig. 4 we demonstrate several attempts to produce prototypes of nanodevices. Figure 4b shows a serial junction of semiconductor and metallic phases, which can be regarded as a Schottky diode. A local semiconductor region sandwiched between two metallic electrodes forms a nanoscale transistor (Fig. 4c). A metallic wire can be embedded in the semiconducting matrix as a quantum lead (Fig. 4d). Finally, quantum dots (in a triangular shape) can be stably produced in the initial phase, with a metallic quantum dot embedded in semiconductor (Fig. 4e), or vice versa (Fig. 4f ). To date, these structures have been fabricated only in an electron microscope and their functions have not yet been confirmed experimentally. The transfer process and the prevention of surface contamination remain obstacles to be overcome. The relatively stable single-layered structures of this system are, however, very promising in the quest to obtain the first single-layered electronic device. Although the low-dimensional nanodevice was first proposed using nanotubes of metallic and semiconducting components, it has turned out to be difficult to realize nanotube composites with controlled chiralities. Patterning single layers is definitely a more promising approach towards the realization of nanodevices. Indeed, it would be extremely intriguing to explore similar phenomena in MoWS2 alloys with tunable bandgaps29 and in n- and p-type doped dichalcogenides26.
Methods Material synthesis and specimen preparation. A single crystal of Re-doped MoS2 was grown by a chemical vapour transport method using Br2 as a transport agent at 950 8C (ref. 25). Mo, S and Re elements (99.99% purity, 10 g in total) containing Br2 (5 mg cm23) were cooled in a quartz ampoule with liquid nitrogen and sealed in vacuum (1 × 1026 torr). Single-crystalline Re-doped MoS2 flakes (3 × 3 mm2 surface area, 0.5 mm thickness) were mechanically exfoliated by Scotch tape and transferred to the surface of a Si substrate with 300 nm thermal oxide. The target single-layer flakes were transferred to a Mo quantifoil grid with 2-propanol and cleaned with chloroform26. The specimens were heated in the TEM chamber (vacuum of 1.7 × 1025 Pa) overnight at 550 8C in a JEOL heating holder to remove residual contamination. ADF-STEM. ADF images were obtained using an aberration-corrected JEOL-2100F cold field-emission gun electron microscope equipped with a DELTA corrector and operated at an accelerating voltage of 60 kV. The convergence semi-angle and inner acquisition semi-angle were 35 and 79 mrad for the ADF imaging. The electron-beam current was 10–15 pA. The dwell time per pixel was 38–76 ms and the pixel size 0.0139–0.026 nm for sequential images (movies), corresponding to an electron dose of 0.7–1.8 × 107 e nm22. Electron beam damage on the MoS2 was observed when the total electron dose exceeded 5 × 108 to 11 × 108 e nm22. Surface cleanliness was critical to preventing damage, and hydrocarbon contamination on the surface enhanced damage development. False-colour images and the alignment of sequential images (movies) were performed using ImageJ. Figure 3e–g presents filtered images using a local two-dimensional Wiener filter to enhance the contrast.
ARTICLES
EELS. Data were recorded using a Gatan Quantum camera. The EEL image spectra in Fig. 4 consist of 12 × 12 pixels, obtained using a 0.1 nm probe with 0.05 nm increments for each step. Each spectrum was acquired in 0.5 s and was summed in the vertical direction to increase the signal-to-noise ratio. The energy dispersion for the recorded spectra was 0.25 eV and the zero-loss width of the incident electron beam was 0.35 eV. The EELS collection semi-angle was 79 mrad. Structure modelling and image simulations. MoS2 polymorphous and phase boundary models were constructed using CrystalMaker and geometry optimizations using HyperChem. ADF image simulations were carried out using QSTEM with a probe size of 1 Å (spherical aberration coefficient, Cs ¼ 1 mm, Scherzer defocus ¼ 24 nm).
Received 21 September 2013; accepted 4 March 2014; published online 20 April 2014
References 1. Winer, W. O. Molybdenum disulphide as a lubricant: a review of the fundamental knowledge. Wear 10, 422–452 (1967). 2. Holinski, R. & Gansheimer, J. A study of the lubricating mechanism of molybdenum disulphide. Wear 19, 329–342 (1972). 3. Radisavljevic, B., Radenovic, A., Brivio, J., Giacometti, V. & Kis, A. Single-layer MoS2 transistors. Nature Nanotech. 6, 147–150 (2011). 4. Wang, Q. H., Kalantar-Zadeh, K., Kis, A., Coleman, J. N. & Strano M. S. Electronics and optoelectronics of two-dimensional transition metal dichalcogenides. Nature Nanotech. 7, 699–712 (2012). 5. Chhowalla, M. et al. The chemistry of two-dimensional layered transition metal dichalcogenide nanosheets. Nature Chem. 5, 263–275 (2013). 6. Matthesis, L. F. Band structure of transition-metal–dichalcogenide layer compounds. Phys. Rev. B 8, 3719–3740 (1973). 7. Wypych, F. & Scho¨llhorn, R. 1T-MoS2 , a new metallic modification of molybdenum disulfide. J. Chem. Soc. Chem. Commun. 1386–1388 (1992). 8. Bissessur, R., Kanatzidis, M. G., Schindler, J. L. & Kannewurf, C. R. Encapsulation of polymers into MoS2 and metal to insulator transition in metastable MoS2. J. Chem. Soc. Chem. Commun. 1582–1585 (1993). 9. Py, M. A. & Haering, R. R. Structural destabilization induced by lithium intercalation in MoS2 and related compounds. Can. J. Phys. 61, 76–84 (1983). 10. Heising, J. & Kanatzidis, M. G. Structure of restacked MoS2 and WS2 elucidated by electron crystallography. J. Am. Chem. Soc. 121, 638–643 (1999). 11. Wypych, F., Solenthaler, C., Prins, R. & Weber, Th. Electron diffraction study of intercalation compounds derived from 1T-MoS2. J. Solid State Chem. 144, 430–436 (1999). 12. Enyashin, A. N. et al. New route for stabilization of 1T-WS2 and MoS2 phases. J. Phys. Chem. C. 115, 24586–24591 (2011). 13. Sandoval, S. J., Yang, D., Frindt, R. F. & Irwin, J. C. Raman study and lattice dynamics of single molecular layers of MoS2. Phys. Rev. B. 44, 3955–3962 (1991). 14. Eda, G. et al. Coherent atomic and electronic heterostructures of single-layer MoS2. ACS Nano 6, 7311–7317 (2012). 15. Hashimoto, A., Suenaga, K., Gloter, A., Urtia, K. & Iijima. S. Direct evidence for atomic defects in graphene layers. Nature 430, 870–873 (2004). 16. Kotakoski, J., Krasheninnikov, A. V., Kaiser, U. & Meyer, J. C. From point defects in graphene to two-dimensional amorphous carbon. Phys. Rev. Lett. 106, 105505 (2011). 17. Huang, P. Y. et al. Grains and grain boundaries in single-layer graphene atomic patchwork quilts. Nature 469, 389–392 (2011). 18. Kurasch, S. et al. Atom-by-atom observation of grain boundary migration in graphene. Nano Lett. 12, 3168–3173 (2012). 19. Kim, K. et al. Grain boundary mapping in polycrystalline graphene. ACS Nano 5, 2142–2146 (2011). 20. Warner, J. H. et al. Dislocation-driven deformations in graphene. Science 337, 209–212 (2012). 21. Lehitnen, O., Kurasch, S., Krasheninnikov, A. V. & Kaiser. U. Atomic scale study of the life cycle of a dislocation in graphene from birth to annihilation. Nature Commun. 4, 2098 (2013). 22. Van der Zande, A. M. et al. Grains and grain boundaries in highly crystalline monolayer molybdenum disulfide. Nature Mater. 12, 554–561 (2013). 23. Zhou, W. et al. Intrinsic structural defects in monolayer molybdenum disulfide. Nano Lett. 13, 2615–2622 (2013). 24. Najmaei, S. et al. Vapor phase growth and grain boundary structure of molybdenum disulphide atomic layers. Nature Mater. 12, 754–759 (2013). 25. Tiong, K. K., Huang, Y. S. & Ho, C. H. Electrical and optical anisotropic properties of rhenium-doped molybdenum disulphide. J. Alloys Comp. 317–318, 208–212 (2001). 26. Lin, Y. C. et al. Properties of individual dopant atoms in single-layer MoS2: atomic structure, migration, and enhanced reactivity. Adv. Mater. http://dx.doi.org/10.1002/adma.201304985 (2014).
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
5
ARTICLES
NATURE NANOTECHNOLOGY
27. Zhou, W. et al. Direct determination of the chemical bonding of individual impurities in graphene. Phys. Rev. Lett. 109, 206803 (2012). 28. Ramasse, Q. M. et al. Probing the bonding and electronic structure of single atom dopants in graphene with electron energy loss spectroscopy. Nano Lett. 13, 4989–4995 (2013). 29. Chen, Y. et al. Tunable band gap photoluminescence from atomically thin transition-metal dichalcogenide alloys. ACS Nano 5, 4610–4616 (2013).
Acknowledgements The authors from AIST acknowledge support from the JST Research Acceleration Programme. D.O.D. and Y.S.H. acknowledge the support of the National Science Council of Taiwan (projects NSC 100-2112-M-011-001-MY3 and NSC 101-2811-M-011-002).
6
DOI: 10.1038/NNANO.2014.64
Author contributions Y.C.L. performed experiments and analysed data. D.O.D. and Y.S.H. grew materials. K.S. and Y.C.L. designed experiments. Y.C.L. and K.S. co-wrote the manuscript.
Additional information Supplementary information is available in the online version of the paper. Reprints and permissions information is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to K.S.
Competing financial interests The authors declare no competing financial interests.
NATURE NANOTECHNOLOGY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturenanotechnology
© 2014 Macmillan Publishers Limited. All rights reserved.
