Electron beam-formed ferromagnetic defects on MoS2 surface along 1 T phase transition

1 T phase incorporation into 2H-MoS2 via an optimal electron irradiation leads to induce a weak ferromagnetic state at room temperature, together with the improved transport property. In addition to the 1T-like defects, the electron irradiation on the cleaved MoS2 surface forms the concentric circle-type defects that are caused by the 2 H/1 T phase transition and the vacancies of the nearby S atoms of the Mo atoms. The electron irradiation-reduced bandgap is promising in vanishing the Schottky barrier to attaining spintronics device. The simple method to control and improve the magnetic and electrical properties on the MoS2 surface provides suitable ways for the low-dimensional device applications.


Results and Discussion
We found a certain condition to increase and induce the Hall mobility and the diamagnetic to ferromagnetic phase transition, after electron irradiation on the cleaved MoS 2 surfaces by changing the electron dose and the acceleration energy, respectively: the electron dose of 300 kGy (6.70 × 10 14 electrons/cm 2 ) and the acceleration energy of 0.7 MeV 34,35 . As shown in the temperature dependence of the Hall mobility (Fig. 1a), the electron irradiation of this condition improves the Hall mobility of the pristine MoS 2 , but slightly reduces a crossover temperature (T C , as indicated by arrow) of the pristine MoS 2 (200 K) to 175 K. Above and below T C , the mobility is mainly subject to the phonon and impurity scatterings, respectively 36,37 . On the other hand, the higher electron dose of 600 kGy (1.39 × 10 15 electrons/cm 2 ) increases the T C over room temperature. Such shift of T C implies that the mobility is limited dominantly by charged impurities, while the phonon scattering plays a minor role.
Figures 1b shows the magnetizations as a function of the magnetic field strength (H) up to ± 50 kOe at low (5 K) temperature. In comparison with the diamagnetic susceptibility 11 of the pristine MoS 2 , the electron dose of 300 kGy induces the diamagnetic to a ferromagnetic phase transition. Interestingly, along the out-of-plane (the c-axis) direction, the diamagnetic behavior still remains for higher magnetic fields than ± 10 kOe. The saturated magnetizations along the in-plane (the ab-plane) and out-of-plane directions are 0.057 emu/g (1.634 × 10 −3 μ B / Mo ion) and 0.030 emu/g (8.60 × 10 −4 μ B /Mo ion) at the H = 35 kOe and 1 kOe, respectively. These weak ferromagnetic states persist up to room temperature, but the saturated magnetizations of 5 K (Fig. 1b) are significantly reduced to 0.011 emu/g (0.315 × 10 −3 μ B /Mo ion) and 0.008 emu/g (0.229 × 10 −3 μ B /Mo ion) at the H = 2 kOe along the in-plane and out-of-plane directions, respectively (Fig. 1c). The coercivities (0.2 kOe) of both directions at 5 K are also reduced to 0.1 kOe at room temperature.
On the other hand, the higher electron dose of 600 kGy induces the diamagnetic to a paramagnetic phase transition along the in-plane direction while the out-of-plane direction still remains diamagnetic (Fig. 1b). Especially along the in-plane direction, the diamagnetic state also retains over the magnetic field of ± 40 kOe, similarly to the case of the out-of-plane direction for the sample irradiated at 300 kGy. At room temperature, however, the temperature-dependent paramagnetic state disappears, while the relatively temperature-insensitive diamagnetic state remains (Fig. 1c) 10,11 . It is evident from the time-of-flight secondary ion mass spectroscopy measurements (not shown) that the electron irradiation of the current condition 34,35 has influences on a few top layers of the cleaved MoS 2 single crystals. Furthermore, the different magnetic states due to the different electron doses are elucidated in Fig. 1(d,e) of the atomic and magnetic force microscopy (AFM and MFM) images taken at room temperature. Similarly to shown in the previous study 11 , undulating magnetic domains representing the ferromagnetic state are clearly observed in the MFM image of 300 kGy (Fig. 1d), whereas the magnetic domains get much weakened in that of 600 kGy (Fig. 1e). This confirms that the electron dose of 300 kGy efficiently induces the ferromagnetic state on the MoS 2 surface.
To elucidate the magnetic domains in more detail, atomic structures on the electron-irradiated surface of 300 kGy were investigated by high-resolution transmission electron microscopy (HRTEM) after the sample was exfoliated using ultrasonic. The fringes of the HRTEM image ( Fig. 2a) indicate that the thickness of the MoS 2 layers is about mono-or bi-layers 38 . With the lack of honeycomb lattices, two types of defects are dominantly found; 1T-phase-like defect [24][25][26] (P1) and concentric circle-type vacancies (P2). The inset of Fig. 2a shows the fast Fourier transform (FFT) image, where the inner and outer hexagons correspond to the (100) and (110) planes 39,40 . The two defects lead to having two (yellow and cyan) hexagons with a twist angle 24° at each plane, respectively. The electron irradiation-induced 1T-phase-like defect (Fig. 2b) is in good agreement with a previous study 26 and supported by the negligible intensities between the main peaks of the line profiles 24 (Fig. 2d). Additionally, the structural difference of the 1T-phase-like defect is confirmed by comparing of the TEM image with the 1 H phase (Fig. 3a), which is half of the unit cell of bulk 2H-MoS 2 . It is notable that the total energy of 1T-MoS 2 is much higher than that of 1H-MoS 2 by 0.8 eV 41 . However, the 1 H to 1 T phase transition can be driven by lowering the energy barrier via the charge injection of electron irradiation 42 . Now, we focus on the concentric-circle-type defect of Fig. 2c. Compared to the line profiles of Fig. 2d, the profiles of Fig. 2e indicate that the central and nearby atoms within the circle (Fig. 2c) correspond to the Mo and S atoms, respectively. In the (dotted) circle of Fig. 3b, the upper three (orange) S atoms of the 1 H phase glide as indicated by the arrows and form partly the 1 T phase. Then such the 1 T phase (Fig. 3b) is turned further into the concentric-circle-type phase (Fig. 3c) after pushing away the inner three S atoms denoted as dotted circles in Fig. 3d. Thus, we will refer to the latter phase as a 1T-3V S defect from the vacancies of the inner three S atoms. The estimated lattice constant of the 1T-phase-like defect (Fig. 2d, a = 3.15 Å) is slightly reduced in the 1T-3V S defect ( Fig. 2e, a = 3.11 Å). The details are compared to the calculated results as described in the methods. However, the Raman signals of the characteristic 1 T phase 43 observed in the chemically exfoliated MoS 2 are not found in the electron-irradiated sample because of the finite thickness of the defect depth. Figure 4a shows two strong Raman peaks at 383 and 408 cm −1 , which correspond to the E 1 2g and A 1g modes, respectively. The Raman spectrum of the electron-irradiated sample is nearly identical to that of the pristine MoS 2 .
On the other hand, the calculations reveal that the 1T-3V S defect doped bilayer MoS 2 in a ferromagnetic state is more stable by energy difference of 0.420 eV per formula unit (fu) than a nonmagnetic one and has a magnetic moment of 0.084 μ B /Mo. Notably, the magnetic moment of the 1T-3V S defect doped monolayer is 0.168 μ B /Mo. The thickness-dependent magnetic moment manifests that, compared to the calculated magnetic moments, the significantly reduced magnetic moments of the electron-irradiated samples (Fig. 1b,c) are attributed to the diamagnetic states of the subsurface layers remaining in the non-defective status. Furthermore, compared to the 1T-3V S doped monolayer, the 1T-phase doped monolayer MoS 2 (Fig. 3b) is more favored by the difference of . However, it is notable that 1T-3V S defects were only obtained at the specific condition of 300 kGy, while the higher electron dose of 600 kGy mainly produced the 1T-phase-like defects (not shown) and induced the diamagnetic to the paramagnetic phase transition instead of the ferromagnetic phase. It is contrary to a simple consideration that the higher electron dose may produce more 1T-3V S defects than 1T-phase-like defects. Thus, the 1T-phase-like defects are considered to be closely related to the V S2 -like defects 33 , where each Mo atom lacks six nearby S atoms (3V S2 ). In other words, the remained S atoms of 1T-3V S defects in Fig. 3c may be pushed away or moved further into the vdW gap at the higher electron dose. The first-principles calculations of 3V S2 (not shown) indicate that, similar to the V S2 -doped monolayer MoS 2 , the 1 × 1 bilayer MoS 2 is nonmagnetic (after removing the topmost S layer) [20][21][22] , and the 2 × 2 bilayer MoS 2 is more likely to be antiferromagnetic than ferromagnetic (after removing the S layers at the top layer). Therefore, the undulating magnetic domains of the MFM image due to the ferromagnetic state (Fig. 1d) are considerably related to the 1T-3V S defects. More interestingly, these 1T-3V S defects can be obtained on the sliding surfaces 44 and the large-area CVD trilayer-MoS 2 film with the plasma treated substrate 45 . However, there are no 1T-phase-like or V S2 -like defects on both samples. In the former case, 1T-3V S defects are simulated by rotation of the two single hexagonal lattices by a misfit angle of 30°. The calculated density of states (DOSs) of the 1T-3V S defect doped bilayer MoS 2 show that the bandgap is closed at the top layer while it is open at the bottom layer (Fig. 3e). In order to investigate the bandgap change due to the electron irradiation, surface-sensitive measurements were performed. Figure 4b-d show that the x-ray photoelectron spectroscopy (XPS) spectra of the electron-irradiated sample shift toward the low binding energy side compared to those of the pristine MoS 2 . However, the stoichiometry of the electron-irradiated sample estimated from the respective integrated peak area of the Mo 3d and S 2p XPS core levels (Fig. 4b,c) retains the ratio (1:2) of the pristine MoS 2 . Deconvolution fits 46 (Fig. 4b,c) elucidate that both Mo 3d (d 5/2 , 229.77 eV) and S 2p peaks (p 3/2 , 162.58 eV) of the pristine MoS 2 (component C1 of the 2 H phase) are found to consist of two components after electron irradiation. The intensity ratio of C1 to C2 is estimated to be 0.5. The electron irradiation-induced peaks (component C2 of the 1T-3V S phase) are located at lower binding energies of 229.59 eV (Mo 3d 5/2 ) and 162.29 eV (S 2p 3/2 ), respectively. It is similar to the 1T-phase doped monolayer 47 . The valence-band maximum (VBM) also moves toward the Fermi energy (E F ) from 0.99 eV to 0.77 eV as indicated by arrows (Fig. 4d). Notably, the influence of the oxygen, which is inevitable during the electron irradiation, is considered to be negligible from the lack of the change at the weak peak of 236.20 eV (Fig. 4b), corresponding to Mo 6+ oxidation state of Mo. In addition to the shift of XPS spectra toward E F , more complementary measurements such as the spectroscopic ellipsometry and optical absorption were measured. Figure 5a,b show the refractive index n and extinction coefficient k of the spectroscopic ellipsometry, respectively. The sharp feature of Fig. 5b, denoted by E 0 , corresponds to the direct-gap transition at the K point with following by the E 0 + Δ 0 peak, which corresponds to the spin-orbit splitting of the valence band at the same K point 3 . These two features of the direct gap, designated as A and B excitons by PL measurements 2,3 , are not responsive to the electron irradiation, while the indirect bandgap shows the oscillating features below 1.5 eV after electron irradiation. The optical absorption results (Fig. 5c) confirm the reduction of the bandgap energy (E g ) by using the relation: α = A/hv(E − E g ) n , where A is the constant, hν is the incident photon energy, and the exponent n depends on the kind of optical transition 48 . The electron irradiation leads to decrease the indirect E g of the pristine MoS 2 by approximately 0.12 eV (Fig. 5d), while the change of direct E g is insensitive to electron irradiation as revealed in the spectroscopic ellipsometry. The bandgap reduction is promising in vanishing the Schottky barrier to attaining spintronics device 49 .

Conclusions
The electron irradiation with the electron dose of 300 kGy (6.70 × 10 14 electrons/cm 2 ) and the acceleration energy of 0.7 MeV creates the 1T-phase-like (V S2 ) and 1T-3V S defects on the MoS 2 surface. These defects reduce the bandgap and improve the transport property. The undulating magnetic domains of the MFM image due to weak ferromagnetic state are considerably related to the 1T-3V S defects. This optimal electron irradiation to improve the magnetic and transport properties at the atomic-layer scale is a key step for the successful integration of 2D TMDs into possible device applications.

Methods
Sample preparation. The natural-single crystalline MoS 2 samples (SPI) were snipped from a large piece and, after a several exfoliation to take the clean surface, irradiated with different exposure times at the electron acceleration energy (ELV-8 linear accelerators) of 0.7 MeV and 2.0 MeV, respectively, in ambient conditions at room temperature. The area of the electron irradiation at the specific point of 400 ± 50 mm was of width 600 ± 20 × length 20 ± 5 mm 2 with beam diameter of 25~35 mm. The stability of the beam energy and dose was less than ± 5%. The electron dose was checked by the dosimeter films.
Characterization. The dc magnetic and hysteresis loop measurements (ca. area of 3 × 3 mm 2 and thickness of ~100 μ m) were performed from 2 to 300 K using a SQUID magnetometer (MPMS XL-7). The MFM measurements were performed with non-contact mode AFM (Bruker-Nano N8 Neo) at room temperature. For the MFM measurements, conductive Pt tips with a radius of ~25 nm were used after Co coating. The MFM images were obtained with a distance of 80 nm between the tip and the sample surface. The electrical conductivity, carrier density, and the Hall mobility were measured as a function of temperature from 100 K to 350 K with a fixed magnetic field of 0.5 T using a Hall measurement system with Au contacts (HMS 5000). HRTEM (JEM-2100F) images for the exfoliated samples by sonication in methanol were taken at an energy of electron beam (200 keV) 40 . The electron irradiation-induced defects are supposed to be hardly influenced by the TEM measurements. From the depth profiles obtained by time-of-flight secondary ion mass spectroscopy, the possible (magnetic) impurities such as O, C, H, and Fe, were found to mostly exist at the electron-irradiated surfaces. There was a negligible reduction of S intensity compared to the Mo intensity on the sample of 300 kGy. Micro-Raman spectroscopy was operated with an Ar ion laser at 514.5 nm. The excitation laser beam of an average power less than 2.5 mW was focused onto samples of interest. The XPS measurements were performed with an Al Kα X-ray source in the vacuum of 1 × 10 −10 torr (ESCALAB 250XI). The energy calibrations were referenced to adventitious carbon at 284.50 eV with eliminating the charging of the sample during analysis. In fitting of Mo 3d and S 2p core-level spectra 46 , the Gaussian width was fixed at the instrumental resolution of 0.65 eV and 0.60 eV, respectively. The values of the spin-orbit splitting and the branching ratios [I (3d 5/2 )/I(3d 3/2 ) and I(2p 3/2 )/I(2p 1/2 )] were 3.17 eV and 0.67 for Mo 3d peaks and 1.18 eV and 0.5 for S 2p peaks, respectively. The refractive index n, extinction coefficient k, and optical absorption spectra were measured by using spectroscopic ellipsometer (UVISEL) and UV-Vis-NIR spectrophotometer (Cary 5000) in the 300-1600 nm wavelength range at room temperature. Optical absorption coefficient were obtained from the transmission mode at room temperature. The thickness of the pristine MoS 2 and electron-irradiated sample is ca. 20 and 50 μm, respectively. Theoretical calculations. First-principles calculations based on density functional theory (DFT) were performed using the Vienna ab initio simulation package (VASP) 50 . For the exchange-correlation potential, the generalized gradient approximation (GGA) was adopted 51 . Wave functions were expanded by a plane-wave basis set with a cut-off energy of 400 eV. The k mesh in the Brillouin zones sampling is 3 × 3 × 1. We account for bilayer MoS 2 , and a large spacing of between two-dimensional unit cells (15 Å) was employed to avoid interlayer interactions (Fig. 3d). To simulate the 1T-3V S doped bilayer MoS 2 (Fig. 3c,d), we adopted a 6 × 6 supercell. The in-plane lattice parameter of the bilayer MoS 2 was used to be the experimental bulk value 46 of 3.160 Å and the atomic positions were fully relaxed. After gliding three S atoms of the topmost layer along the 2 H to 1 T pathway as indicated by the arrows of the (dotted) circle in Fig. 3b, the concentric-circle-type pattern was constructed by removing the inner three S atoms as shown in Fig. 3c. The bond length and the projected distance (d Mo-S = 2.41 Å and 1.82 Å) of the 2 H phase in Fig. 3a increases to 3.72 Å and 3.64 Å in Fig. 3b and to 3.60 Å and 2.74 Å in Fig. 3c, respectively. In Fig. 3d, the bond length (2.41 Å) and bond angle (θ S-Mo-S = 80.70°) decrease to 2.38 Å and 79.18° with increasing the d Mo-S of the 1T-3V S defect to 2.43 Å and 80.9°.