Topological 1T -phases patterned onto few-layer semiconducting-phase MoS2 by laser beam irradiation

Topological 1T -phases patterned onto few-layer semiconducting-phase MoS2 by laser beam irradiation H. Mine 1, A. Kobayashi 1, T. Nakamura 2, T. Inoue 3, J. J. Palacios 4, E. Z. Marin 4, S. Maruyama 3, S. Katsumoto 2, A. H. MacDonald 4, J. Haruyama 1,2* 1 Faculty of Science and Engineering, Aoyama Gakuin University, 5-10-1 Fuchinobe, Sagamihara, Kanagawa 252-5258, Japan. 2 Institute for Solid State Physics, The University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa, Chiba 277-8581, Japan. 3 Dept Mechanical Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan. 4 Department of Physics, The University of Texas at Austin, Austin, Texas 78712, USA. * To whom correspondence should be addressed. E-mail: J-haru@ee.aoyama.ac.jp The introduction of a two-dimensional (2D) topological insulating (TI) state is crucial for voltage-controlled spintronic devices with low power dissipation. However, their experimental observation is rare. A 2D material family of transition metal dichalcogenides (TMDCs) has been recently predicted and experimentally reported as a new class of 2D TI material. Here, we pattern the 1T'-phase onto thin 2H-semiconducting phase molybdenum-disulfide (MoS 2 ), a TMDC family, by laser beam irradiation. The integer fractions of resistance quantum in resistance peaks observed for the two-different 1T patterns indicate the presence of helical edge spin modes of quantum spin Hall effect in 2D TI states at the 1T'/2H phase interface. This result is supported by scanning tunneling spectroscopy spectra and theoretical calculations. The present observation opens the door to the patterning of topological phases into non-topological phases of TMDCs and their application to topological quantum computation. Three-dimensional (3D) topological insulating (TI) states, in which the bulk is an insulator with an energy band gap while the surface is a gapless conductor, are attracting significant attention (1,2). In a two-dimensional (2D) TI state, the quantum spin Hall effect (QSHE) emerges with a bulk energy gap but gapless helical edge states protected by time-reversal symmetry, in which opposite spin states form a Kramers doublet counterpropagate (3-5,22). 2D TI states had been theoretically predicted for graphene (6-8), while they have been experimentally reported in only a few systems, such as monolayer graphene subjected to a very large magnetic field angled with respect to its in-plane (19) and low-coverage Bi 2 Te 3 nanoparticle-decorated graphene (10). However, control of the QSHE remains a challenge. Recently, a 2D material family of transition metal dichalcogenides (TMDCs) has been predicted to be a new class of 2D TI (11-14,21), and experimentally observed in, for example, monolayer WTe 2 (11,20) and 1T'-WTe 2 (12). Measuring various two-probe electrical properties showed that monolayer WTe 2, at temperatures up to about 100 K, is insulating in its interior while the edges still conduct (11). Latest observation of a half-integer resistance quantum (R Q /2 = h/2e 2 = 12.9 k, where h is Planck s constant and e is the charge on the electron) of edge resistance up to 100K in atomically flat monolayer WTe 2 realized by fully encapsulating it with hexagonal boron nitride reconfirmed presence of a QSHE through selective doping of the flake using a combination of global top and local bottom gates (20). 1

On the other hand, in 1T'-WTe 2, the observation of clear signatures of topological phase due to band inversion with bandgap opening ( 55 mev), supported by angle-resolved photoemission spectroscopy (ARPES) and scanning tunneling spectroscopy (STS) spectra, was reported (12). Typically, a TMDC (MX 2 ) has three phases, i.e., 2H, 1T, and 1T'. 1T-MX 2 is composed of three hexagonally packed atomic layers in an ABC stacking. The metal M atoms are octahedrally coordinated to the chalcogen X atoms. This phase is not stable in free-standing form and undergoes a spontaneous lattice distortion into the 1T' phase via a doubling of the periodicity in the x direction. X atoms are dislocated from the original octahedral positions to form a zigzag chain in the y direction. The lattice distortion from the 1T phase to the 1T' phase induces band inversion and causes 1T'-MX 2 to become a TI state. There are various methods to create 1T or 1T' phases. We formed a 1T phase into few-layer 2Hsemoconducting MoS 2, a member of the MX 2 family, by electron beam irradiation and revealed the presence of a few-atomic-layer pinning-free Schottky junction at the interface of the 2H/1T phases (18). Alternatively, laser beam irradiation can generate the 1T' phase through thermal activation of crystal. When this 1T' phase is within a 2D TI state, one can pattern TI states, embedding them into the 2H semiconductor phase, using such laser beam irradiation. This method has many advantages in TI engineering. For instance, the interface between one-dimensional (1D) TIs and superconductors can create Majorana fermions (MFs), which are their own anti-particle and can realize topological quantum computation (15-17). For their application, it is indispensable to form multiple MFs (i.e., patterning of multiple 1D TI states/superconductor junctions) and control those positions as required. However, formation of the multiple 1D TI state at desired locations is difficult. Laser beam patterning of the 1T' phase resolves this problem and enables topological quantum computation using multiple and mobile MFs (e.g., based on non-abelian statistics). In the present experiments, thin MoS 2 flakes are formed by mechanical exfoliation of the bulk material onto an SiO 2 /Si substrate. Layer thicknesses of about 20 nm have been confirmed by atomic force microscopy (AFM) and optical microscopy (OM). Example OM and AFM images of a flake irradiated by a laser beam (532-nm wavelength) are shown in Figs. 1A and 1B, with irradiation spots created with different irradiation times (points 1, 2, 3, 4, 5 correspond to 500 s, 50 s, 30 s, 20 s, 10 s, respectively, under a constant power of 50 mw) and different powers (50 mw for points 1-5, for 25 mw with 50 s duration for point 6). In the OP image (Fig. 1A), the color of the individually irradiated points drastically changes to semi-transparent with irradiation. In contrast, an irradiation time below 10 s and a power below 25 mw result in no changes in the color. The AFM image (Fig. 1B) reveals a decrease in the layer thickness of about 10 nm at one of the irradiated positions. These observations are consistent with previous reports, which demonstrated that transition to the 1T' phase and its crystal structures decreased the interlayer distance, and thus reduced the total thickness. When only the upper layers are transformed into the 1T' phase by laser irradiation, leaving the lower layers semiconductive, the total thickness of the flake is reduced (Fig. 4H). Typical Raman spectra, (recorded using laser irradiation with a power of 0.82 mw and a wavelength of 532 nm), are shown in Fig. 1C. The laser-irradiated (50 mw for 50 s) sample shows a significant peak around 230 cm -1, which is absent from the spectrum of the sample that was not irradiated and can be unique to 1T' phase. Photo luminescence (PL) signals of the sample imaged in Fig. 1A, (recorded using laser beam irradiation with a power of 0.82 mw), are shown in Figs. 1D and 1E. These reveal that the peak positions shifted to lower wavelengths (i.e., higher energies) and peak intensities decreased with laser-irradiation time. This can be also understood by the abovementioned interpretation. When the upper layers become the 1T' phase, these layers cause no PL signals, whilst the lower semiconducting layers yield the PL signal (Fig. 4H), and the peak heights reduce and the peak positions shift to higher energy due to the decrease in the total layer thickness for increased laser 2

irradiation time. These results strongly suggest the transition of the upper layers to the 1T' phase by laser beam irradiation. Indeed, X-ray photoelectron spectroscopy (XPS) of the laser-irradiated (50 mw for 50 s) sample demonstrates hybridization of Mo 3d orbitals of the 1T' and 2H phases (Fig. 1F). Based on these experiments, two different 1T'-phases are patterned onto thin 2H-MoS 2 flakes by laser beam (17-mW power, 532-nm wavelength) irradiation to clarify the presence of the 2D TI state. Figure 2 shows OM examples of 1T rectangular pattern and 1T'-Hall-bar pattern, which consists of four branches that together form an H-letter (3, 10). Au/Ti electrodes are in contact with four corners of the rectangular pattern (Fig. 4A) and each branch of the H-letter like patter (Fig. 4B). For the rectangular pattern, two-terminal resistance between electrodes 1,3 and 2,4 is measured as a function of back gate voltage (V bg ) by flowing a constant current between electrodes 1,3 and 2,4 (Fig. 4A). The results of two different samples are shown in Figs. 3A and 3B. In Fig. 3A, two R peaks of R 34 R Q and R Q /2 are confirmed at V bg -5V and +25V, respectively. In Fig. 3B, two different R peaks are confirmed at V bg -20V and +10V with R 12 3R Q /2 and R Q /2, respectively. For the H-letter like pattern, when a constant current flows between electrode pair 1-2, non-local resistance (R NL ) between electrode pair 3-4 (R 34 ) is measured as a function of V bg (Fig. 4B). Figure 3C demonstrates the results; R NL peak of R 34 R Q /2 at V bg +5V and R NL plateau of R 34 R Q /4 at V bg +20-25V. These quantized R values can be understood by some scenarios. One of them is explained as below, considering the description for a QSHE and its dephasing (3, 10). The R peak values R Q /2 confirmed for the 1T rectangular patterns (V bg +25V in Fig. 3D and V bg +10V in Fig. 3B) coincide with the R value observed in a helical edge mode of a QSHE without dissipation. As mentioned above, when helical edge modes exist without dissipation, the two counterpropagating channels can be preserved at two different quasi-chemical potentials between the electrodes, leading to a net current flow along the edges with R equal to R Q /2 based on the Landauer-Büttiker (LB) formalism (Fig. 4A) (19,20). Indeed, the R plateau R Q /2 was reported in the two-terminal resistance in a rectangular graphene with four electrodes at four corners, which is the same structure as ours, and conformed as evidence for a helical edge spin modes in a QSHE derived by applying high magnetic fields (19). Moreover, it should be noticed that the R NL peak values R Q /2 observed in Figs. 3A and 3B are slightly larger than R Q /2. This agrees well with the R plateau value of 0.56R Q observed in Ref.19. In contrast, once the helical edge spins enter the voltage electrodes, they interact with a reservoir containing infinitely many low energy degrees of freedom, and the time-reversal symmetry is effectively broken by the macroscopic irreversibility. Consequently, the two counterpropagating channels equilibrate at the same chemical potential, determined by the voltage of the metallic electrodes, leading to dissipation and, thus, emergence of integer fractions of R Q. The R NL plateau value R Q /4 existing at V bg +20-25V in Fig. 5C corresponds to this in an H-letter like pattern with four metal electrodes (Fig. 4B), which exactly agrees with those in Refs. 3 and 10. Moreover, R > R Q was observed in the two-terminal measurements in Ref. 3, when a constant current flowed along a main Hall-bar with four branches attached to metal electrodes, and the twoterminal R was measured at the two ends of the Hall-bar (Fig. 4C). Because R Q /2 was conserved along the edges of the main Hall bar between each of the two branches with electrodes but dephasing occurred in the metal electrodes, the two-terminal R resulted in R Q /2 (three inter-electrode regions) = 3R Q /2. The R peak values except for R Q /2 (i.e., R Q, 3R Q /2, and R Q /2) observed in Figs. 3A- 3C, respectively, can be interpreted by this model, assuming presence of one or two defects existing at edges of 1T /2H interface (Fig. 4D). When the defects act as dephasing centers like metal electrodes, the two-terminal measurement leads to R Q /2 (two or three inter-defect regions) = R Q or 3R Q /2, respectively, for Figs. 3A and 3B, and R Q /4 (two inter-defect regions) = R Q /2 for Fig. 3C. Indeed, large defect-like boundary (i.e., topological/non-topological state junction) can be observed 3

at some points of the 1T'/2H junction (Figs. 4E and 4F), which are very different from clean edges of other parts (Fig. 4G), in the most surface layer. This mean that at least the most surface layer has edge defects in individual samples, because the laser beam is directly irradiated to it, resulting in the most significant 2H-1T phase transition with yielding defects. Here, individual samples for Figs. 3A-3C have two R peaks, which can be understood by the abovementioned individual interpretations for a QSHE. This suggests that at least two layers showing such R peaks exist in the multiple 1T -layers with 10 nm thickness of individual samples (Fig. 4H; e.g., the second or third surface layer for Fig. 4A and the most surface layer for Fig. 4D). When such two layers independently yield individual the R peaks, the two R peaks can be observed in one figure as a result of those hybridization. In fact, Fig. 3D, which is a result of the sample with 1T -rectangular pattern formed by reduced-power laser-beam irradiation (10mW), demonstrates presence of only one R peak with 3R Q /2 value. This suggests that only the most surface layer can have Fig. 4D-structure, while the layers below the second layer have no topological phase. This supports that the laser beam irradiation with 17-mW power can cause the abovementioned two topological layers for Figs. 3A-3C. Scanning tunneling spectroscopy (STS) spectra of the H-letter-like 1T region are shown in Fig. 5. In the bulk region (near the center of left vertical bar of the H-letter), an STS gap of 35 mev is confirmed, while it disappears at an edge (near the boundary at 1T'/2H). Although this bulk gap value is smaller than the 45 mev gap reported in 1T'-WTe 2 /bilayer-graphene/sic (12), it is appropriate for 1T'-MoS 2. These behaviors are consistent with those in QSHEs and support the presence of the helical edge mode observed in Fig. 3. The results for band structures by first principle calculation are shown in Fig. 6. The calculated structure consists of 2H/1T /2H regions with very narrow 1T area (three unit cells) as shown in (A). Figures 6B and 6C show the calculated band structures in individual regions of (a) and axisdependence of 1T phase of (B) under spin-orbit interaction (SOI), respectively. Clear topological gaps cannot be confirmed around Fermi level (E F ) in these figures. Figure 6D shows expansion of the band structure for x-axis in (C) (red-dashed rectangle of the left panel of (C)). It suggests that two energy bands (indicated by two arrows) exist around E F at where a SOI gap of 50 mev should exist if they disappear. This SOI gap value is in good agreement with that observed in STS (Fig. 5). These two bands originate from the 2H regions existing at both sides of the 1T region, because of the very narrow 1T region. In the experimental structure, width of the 1T region of the H-letter is 1 m (Fig. 2A), which is much larger than this. Therefore, disappearance of these two bands crossing into the SOI gap is expected in the actual sample, resulting in the observed TI states. In conclusion, we patterned the two-different 1T'-phases onto thin 2H-phase MoS 2 by laser beam irradiation. The quantized R peaks observed for the rectangular pattern and the H-letter like pattern indicated the presence of helical edge modes of a QSHE in 2D TI states at the 1T'/2H phase interface with defects. This result was supported by STS spectra and first principle calculations. The present observation opens the door to the patterning of 2D (or 1D) topological phases onto non-topological phases of TMDCs and their application to topological quantum computation, particularly realized by controlling the positon and operation of multiple MFs combining with superconductor electrodes. 4

References and Notes 1. H. Zhang et al., Nat. Phys. 5, 438 (2009). 2. Y. Xia et al., Nat. Phys. 5, 398 (2009). 3. A. Roth et al., Science 325, 294 (2009). 4. C. Brune et al., Nat. Phys. 8, 485 (2012). 5. L. J. Du et al., Phys. Rev. Lett. 114, 096802 (2015). 6. C. L. Kane & E. J. Mele, Phys. Rev. Lett. 95, 226801 (2005). 7. J. Hu et al., Phys. Rev. Lett. 109, 266801 (2012). 8. H. Jiang et al., Phys. Rev. Lett. 109, 116803 (2012). 9. T. Nanba, J. Haruyama et al., Appl. Phys. Lett. In press 10. K. Hatsuda, J. Alicea, J. Haruyama et al., Science submitted 11. Z. Fei et al., Nat. Phys. 13, 677 (2017). 12. S. Tan et al., Nat. Phys. 13, 683 (2017). 13. X. Qian et al., Science 346, 1344 (2014). 14. L. Kou et al., J. Phys. Chem. Lett. 8, 1905 (2017). 15. V. Mourik et al. Science 336, 1003 (2012). 16. S. M. Albrecht et al., Nature 531, 206 (2016). 17. M. T. Deng et al., Science 354, 1557 (2016). 18. Y. Katagiri, J. Haruyama et al., Nano Lett. 16, 3788 (2016). 19. A. F. Young et al., Nature 505, 528 (2014). 20. S. Wu et al., Science 359, 76 (2018). 21. C. Liu et al., Phys. Rev. Lett. 100, 236601 (2008). ACKNOWLEDGEMENTS The authors thank S. Tang, Z.-X. Shen, Y. Shimazaki, T. Yamamoto, S. Tarucha, T. Ando, M. Dresselhaus, P. Herrero, and P. Kim for their technical contributions, fruitful discussions, and encouragement. The work carried out at Aoyama Gakuin University was partly supported by a grant for private universities and a Grant-in-Aid for Scientific Research (15K13277) awarded by MEXT. 5

Intensity (a.u.) Intensity (a.u.) Count (a.u.) A 1 1 2 6 3 4 5 B 500nm 500nm C Irradiation Non-irradiation 10nm 10n 5 m D Irradiation Irradiation time time E Time F Time 150 200 250 300 Raman shift (nm -1 ) 1T 2H a 520 560 600 640 680 Wavelength (nm) 640 660 680 b Wavelength (nm) 700 235 230 225 Binding energy (ev) Fig. 1 Room-temperature sample characterizations. (A) Optical microscopy image of a thin MoS 2 flake with seven dark points caused by laser beam irradiation (532-nm wavelength, 1- m diameter) with various irradiation times (500 s, 50 s, 30 s, 20 s, and 10 s corresponding to points labelled 1, 2, 3, 4, and 5, respectively, under a constant power of 50 mw) and different powers (50 mw for points 1-5, and 25 mw with 50 s duration for point 6). (B) AFM image of a cross section of the irradiated point labelled in (A) (red dashes line), for which the irradiation conditions were the same as those for point 1. (C) Example Raman spectra for laser beamirradiated point 2 in (A) and a non-irradiated point (recorded using a laser power of 0.82 mw and wavelength of 532 nm). (D,E) Photo luminescence (PL) spectra of the laser-beam irradiated points illustrated in (A) (recorded using a laser with 0.82-mW power) plotted for wavelengths (D) 530-710 nm and (E) 640 700 nm. The numbers on the graph are the irradiation time for each plotted line and are common to both (D) and (E). (F) X-ray photoelectron spectroscopy (XPS) of the sample after laser irradiation by a 50-mW beam for 50 s. Red and blue lines are data fits for spectra of the 1T' and 2H phases, respectively. 6

A 1um B Fig. 2 Optical microscopy images of a thin MoS 2 flake patterned by laser beam irradiation with and without metal electrodes. (A) 1T'-phase rectangular pattern ( 4 5 m 2 ) and H- letter-like pattern ( 1 m channel width), on a thin 2H-MoS 2 flake formed by laser beam irradiation with 17-mW power and 532-nm wavelength, for the resistance measurements shown in Fig. 3. (B) The flake shown in (A) connected by Au/Ti electrodes, corresponding to Figs. 4A and 4B. 7

R (K ) R (K ) R (K ) Laser power 17 mw Laser power 17 mw T=1.5K T=1.5K A B Laser power 10 mw R34 (K ) C Laser power 17 mw T=1.5K D T=1.5K Fig. 3 Resistance measurements as a function of back gate voltage (V bg ). (A,B) For the 1T - rectangular pattern on different two samples; two-terminal resistance measured between electrodes 1,3 and 2,4 as a function of V bg by flowing a constant current between electrodes 1,3 and 2,4 (Fig. 4A). (C) For the 1T' H-letter like pattern; Non-local resistance (R NL ) observed for electrode pairs 3-4 as a function of V bg, when a constant current flows between electrode pairs 1-2 (Fig. 4B). (D) For the rectangular pattern formed by laser beam irradiation with lowered power (10mW). 8

E F G H Fig. 4 (A-D) Schematic views of layer structure for helical edge modes and metal electrodes. (A) 1T -phase rectangular pattern with four metal electrodes on each corners for a helical edge spin mode in two-terminal measurement. (B) 1T' H-letter like pattern with four metal electrodes on each branche in four-terminal measurement. Dephasing occurs in metal electrodes. (C) Hall bar pattern with four metal electrodes in two-terminal measurement. Dephasing occurs in metal electrodes. (D) A Pattern (a) with defects, B which act as dephasing centers. (E-G) AFM images of 1T region and 1T /2H interface with a defect at edge (E,F) and with clean edges (G). (F) is an expansion of the white rectangle position of (E). (H) Schematic cross section of the 1T phase formed into 2H phase by laser beam irradiation. 9

C 300 mk Vbg=25V Fig. 5 Scanning tunneling spectroscopy (STS) spectra of the H-letter-like pattern. The bulk signal is measured near the center of left vertical bar of the H-letter, while the edge signal is measured near the boundary of the 1T'/2H phases. 10

E (ev) A B C D Fig. 6 (A) Schematic MoS 2 structure consisting of 2H/1T /2H regions with very narrow 1T area (three unit cells). (B-D) Band structures of (A) calculated from first principle calculation under SOI; (B) Individual region s bands of (A), (C) Axis dependence of 1T region of (B), and (D) Expansion of red-dashed rectangle in left panel of (C). Two arrows mean mixed two bands arising from the two 2H regions. 11