Surface band bending in semimetal Td-WTe2

Recently, spontaneous out-of-plane electric polarization and ferroelectric switching were found in WTe 2 devices. On single crystal with ferroelectric property, a built-in electric eld and corresponding band bending would be expected at the surface. However, such band bending in WTe 2 hasn’t directly been observed experimentally. Here, by tting angle-dependent X-ray photoelectron spectroscopy (XPS) spectra on WTe 2 surface, we clearly observed downward band bending after slightly exposure to air, verifying appearance of surface polarization. Such band bending can’t be observed on pristine WTe 2 surface and will disappear on fully oxidized sample. It suggests strong correlation between surface band bending and oxidation. Ionized donors from oxide species pinned at surface may contribute to the formation of surface band bending and polarization. Our study here offers new insight to gure out the microscopic origin of ferroelectric in WTe 2 .

Their DFT calculation suggests polarization is pristine in such polar semimetal due to contribution from ionic cores and valence bands. 16 However, experimental study about the origin of such polarization is still seldom. Normally, out-of-plane spontaneous polarization in single crystal should induce surface charge redistribution and built-in electric elds, which result into band bending near surface. 17 Therefore, surface band bending can be a evidence for existence of polarization or ferroelectric on WTe 2 . Angle resolved photoemission spectroscopy (ARPES) measurements 18-21 on freshly cleaved WTe 2 haven't shown such band bending. If exposing WTe 2 to air for 3 hours, J. M. Woods et al. 22 observed decreased work function by 300 meV, which was claimed to be due to surface band bending.
If exposure to air, the surface of T d -WTe 2 is easy to be oxidized. 15,21−29 Investigation of its oxidation behavior is important for WTe 2 -based devices in real application, because surface oxidation will decrease conductivity, 23 suppress magnetoresistance [22][23][24][25][26] and induce weak antilocalization. 24,25 Strong oxidation happens within a few minutes, 24,27 but the WTe 2 oxidation is self-limiting process and occurs mostly on the top surface. A thin surface oxide layer (~ 2 nm thick 16,22 ) can be quite effective to protect inner bulk layer from further degradation. The reason is still unknown. According to XPS measurements, 22,23,26−28 the oxidation driving force are formation of WO x and TeO x on the surface of WTe 2 . Theoretical calculation 24,29 claimed the fast oxidation happens with the aid of H 2 O, which can decrease the reaction barrier. And charge depletion was found to near the O adsorption site, which may induce charge dipolar and even polarization.
In spite of many investigations about oxide behavior of WTe 2 , oxidation dynamic and how the oxidation affects the electron band structure are seldom to be studied. Here we combined angle-dependent X-ray photoelectron spectroscopy (ADXPS) and TOF-SIMS to investigate the exposure-time-dependent oxidation behavior and surface band bending. We found the oxidized species were WO 3 and TeO with oxygen vacancies, which protect deeper layer to be oxidized. What's more, we observed clear downward surface band bending, con rming the out-of-plane polarization on WTe 2 . Further we found the surface band bending is closely related to surface oxidation.

Results And Discussion
Among TMDs, WTe 2 crystallizes to T d phase energetically with an additional structural distortion: the tungsten atoms form quasi one-dimensional zigzag chains along a direction and Tellurium atoms form undulation morphology, as show in Fig. 1a. The thickness of Te-W-Te is ~ 0.42 nm and the space between neighboring van der Waals interacting layer is ~ 0.29 nm. Before ADXPS measurements, we have systematically studied the oxidation behavior by combined XPS, TEM and TOF-SIMS. Scanning X-ray Induced Secondary Electron Image (SXI) was adopted to locate and trace the probe area with size of ~ 100 um, as shown in the inset of Fig. 1b. Figure 1c shows the normalized XPS spectra of Te 3d core level, collected at TOA of θ = 45º. Only two Te 3d peaks (Te 3d 5/2 -W and Te 3d 3/2 -W) with splitting energy of 10.39 eV were observed after UHV cleavage. No trance of C and O signal indicates free of oxidation. After exposure to air for one minute, a small shoulder appears at about 576 eV, which is the Te 3d 5/2 -O peak.
We used the ratio of shoulder to main peak of Te 3d 5/2 -W to estimate the oxidation degree. The oxidation ratio is only ~ 2%. When increasing exposure to 5 min, the Te-O peak are pronounced and the ratio quickly increases to ~ 16.6%. Further exposures only induce slightly increased oxidation to ~ 19.2%, as shown in Fig. 1d. It suggests WTe 2 oxidizes quickly after exposure to air, within few mins, and the oxidation is selflimited process without deep oxidation, consistent with previous Raman measurements. 24,27 We found the core level of Te 3d 5/2 -W shifted after exposure, which will be discussed in Fig. 4d.
The HRTEM image in the left panel of Fig. 2a shows after long-time exposure, a thin amorphous oxide layer with slight weak intensity can be found between WTe 2 and deposited Pt protective layer. EELS line pro le in Fig. 2c and mapping image in the Fig. 2b suggest the thickness of oxidation layer is ~ 2.4 nm, agreeing with previous estimation. 16,22 It means the top four WTe 2 layers are nearly oxidized after long time exposure. We must point out that due to the integration of photoelectrons emitting from a depth of d 0 , oxidation ratio obtained from XPS measurements is ~ 19.2%. Trace of some weak layered structure still can be identi ed. Such amorphous layer is uniform and dense, which can block the entrance of O 2 and H 2 O to avoid further oxidation. Information about the proportion of elements after oxidation is helpful to understand the oxidation dynamic. To illustrate the components of oxidation layer, we performed delicate TOF-SIMS measurements according to the depth pro le of different negative fragments, as shown in Fig. 2d To gure out the band structure evolves after oxidation, we systematically performed ADXPS measurements after each exposure. Figure 3a shows a series of spectra on UHV cleaved sample at different emission angles of θ. We can't nd any trace of oxidation signal even at θ = 20º.The intensity increases with emission angles θ, while the core level of Te 3d 5/2 -W keeps almost constant at different TOAs, as shown in Fig. 3c. The photoelectron escape depth λ depends on the TOA (θ) in a simple relation of λ 0 sin(θ), where λ 0 is the inelastic mean free path of photoelectrons. The value of λ 0 is 3.25 nm for photoelectrons of Te 3d in WTe 2 , as calculated by TPP-2M method 31 in NIST's database. 32 After exposure to air for one minute, Te-O bond can be observed at each emission angle in Fig. 3d. We nd obvious binding energy shift in Fig. 3 f. After subtracting Shirley background, the spectra collected under different TOAs were tted using Eqs. (1) and (2) in Method with the tting parameters: binding energy E 0 , strength of the electrostatic potential K, the ratio of Gaussian function α, and the full width at half maximum F. The tted curves were displayed in Fig. 3b and Fig. 3e. For UHV cleaved sample, the resulted tting parameters are E 0 = 572.97 eV, K = 0.0112 eV/nm, α = 0.404 and F = 1.01 eV. The small K suggests the surface band bending can be negligible on fresh cleaved surface, consistent with previous ARPES measurements. [18][19][20][21] However, on one min-exposure sample, the strength of electrostatic potential K decreases to 0.0358 eV/nm with E 0 = 573.00 eV, α = 0.396 and F = 1.25 eV. Taking the obtained tting parameters back into Eqs. (1) and (2), we can draw the simulation curve as function of TOAs, as shown in Fig. 3c and 3 f. The negative value of K suggests the surface band is bending downward. − − When exposing to air for 5 min, the strength of electrostatic potential K further decreases to 0.0576 eV/nm, but the tting error is large according the simulation curve in Fig. 4a. Such large error is probably due to quick oxidation and nonuniform charge distribution at the rst stage of 5 min. Nevertheless, increased electrostatic potential or band bending is obvious. After exposure to air for more than 32 hours in Fig. 4b, the binding energy nearly keeps constant again at different TOAs and the value of K increases back to near zero. We summarize the strength of surface electrostatic potential as function of exposure time in Fig. 4c. Clear close correlation between surface oxidation and surface electrostatic potential or surface downward band bending is found.
According to the carrier density and dielectric constant of WTe 2 in Poisson's Eq. 2 2 , the depletion layer or spatial extent of the band bending is estimated to be 1 ~ 2 nm. We notice that previous two experimental observations of ferroelectricity were performed in glove box with N 2 protective gas. Considering the rapid oxidation behavior, the surface of WTe 2 was expected to be in analogy with the one of one-mins-exposed samples, where the strength of electrostatic potential is 0.0358 eV/nm. It means the built-in electric eld potential should be ~ 36 meV, which agrees well with the estimated changes in potential difference for polarization reverse (2δV ≈ 40 meV) in previous report. 14 Further, previous observed ferroelectric domain structure during piezo response force microscopy measurement 16 has distorted circular pro le with size of 20 ~ 50 nm and decoratively distributed on the surface. Considering the orthorhombic symmetry of T d -WTe 2 and our observation of surface built-in electric eld, the distorted circular may be oxidized defects with amorphous structure. Although we conclude the close correlation between surface polarization and surface oxidation, the detailed microscopic mechanism for ferroelectricity after oxidation in WTe 2 deserves further investigation.
We would like to discuss about how the surface oxidation affects the amount of band bending. Surface of freshly cleaved sample has the same size of electron pockets and hole pockets across fermi level (E F ) with perfect electron − hole charge compensation. As expected, negligible band bending was observed with small K in our ADXPS measurement. When the surface starts to oxidize, the oxidation species (WO x and TeO x ) will induce hole doping due to the high work function. At the beginning, oxidation happens locally (such as vacancies) with the ionized donors pinned on the surface, as shown in the schematic diagram in Fig. 4e. The pinned donor will attract the electrons and then form built-in electric eld (E) pointing to the bulk. It agrees with the previous observation of polarization on WTe 2 surface. 14,16 However, here we show the oxidation effect induces such polarization rather than its intrinsic property. Due to the pinned ionized donors, the attracted electrons below oxide layer will shift the core-level to higher binding energy. It agrees with the peak shift observation in Fig. 4d. When the surface further oxidizes after long time exposure, the oxidation layer becomes thick and nonlocalized, which can't pin the donor again. On the contrary, the oxidation layer induces nonlocal hole doping to compensate the attracted electrons, then results into weakened surface band bending. Such hole doping agrees with the shift the core-level to lower binding energy when the exposure time increase to 2 h and 32 h in Fig. 4d.

Conclusion
In summary, we have systematically studied the oxidation behavior of WTe 2 and probed the changes of surface band structure by using angle-dependent XPS. WTe 2 oxidizes quickly after exposure to air for 5 min. By combined XPS and TOF-SIMS measurements, we can identify that oxide species in amorphous layer are mainly WO 3 and TeO with oxygen vacancies. By tting the angle-dependent XPS spectra, obvious downward band bending can be observed after exposure to air. It con rms an electric eld or polarization pointing to the bulk. We attribute such behavior to the pinned ionized donors from surface oxide species, which attract electrons from bulk and form the electric eld. Surface band bending isn't found on freshly cleaved WTe 2 and fully oxidized surface. Our study here provides important insights about the surface polarization of WTe 2 and may also help to explain magnetotransport behavior after oxidation.

Methods
Sample preparation. WTe 2 crystal samples were purchased from PrMat (PrMat, China). The WTe 2 single crystal were stuck with a ceramic rod using Ag paste, and then loaded into ultra-high vacuum (UHV) chamber. The samples were cleaved in-situ in XPS chamber before ADXPS measurement. After the ADXPS measurement on fresh-cleaved WTe 2 , the samples were exposed to the air for different times (such as 1 min, 5 min, 2 h, 32 h here). After each exposure, the WTe 2 sample were immediately loaded into the XPS's UHV chamber and probed at the same location. ADXPS characterization. XPS measurements were carried out in an ultra-high vacuum (UHV, ~ 5 × 10 − 10 mbar) chamber, equipped with a hemispherical electron energy analyzer (PHI 5000, VersaProbe, ULVAC-PHI) and a monochromatic Al K α X-ray source of 1486.7 eV. X-ray. Spot size of 100 µm diameter was adopted during high-resolution XPS spectra measurement. To probe the same position on the surface after each exposure, we used X-ray beam induced secondary electron imaging (SXI). With xed hemispherical electron energy analyzer, the sample stage can be titled to collect photoelectron from different take-off-angles (TOA, θ, the elevation angle with respect to the sample surface), as shown in Fig. 1(b). The photoelectron escape depth λ equals to λ 0 sinθ, where λ 0 is the inelastic mean free path of photoelectrons. The binding energy (BE) of core-level peaks were calibrated with respect to the C-C 1 s bond (BE = 284.8 eV). For freshly cleaved samples without C adsorption, the binding energy were calibrated with Au 4f 7/2 emission line. The spectra are curve-tted after subtracting a Shirley-type background.
An internal electrostatic potential, due to surface charge pinning or polarization, will induce band bending, which changes the measured value of core-level spectra under different TOAs. Based on the apparent binding energy and line shape in ADXPS measurements, surface band bending and internal potential gradient can be calculated. 33 The principle and calculation method have been speci ed previously. 33-36 Brie y, a measured apparent core level spectrum is actually an integration of photoelectrons emitting from a depth of d 0 , and can be given by