WSe₂-contact metal interface chemistry and band alignment under high vacuum and ultra high vacuum deposition conditions

All WSe₂ samples originated from the same bulk crystal purchased from HQ Graphene [33]. The procedures employed for sample preparation, mounting, metal source outgassing, determining metal deposition rates (0.3–0.6 nm min⁻¹ depending on the metal), metal deposition under both HV and UHV conditions, and subsequent XPS analysis are identical to that employed for an analogous investigation of interface chemistry between Au, Ir, Cr, and Sc and bulk MoS₂ described elsewhere [29]. Exfoliation is performed under laboratory ambient conditions, as is typical in device fabrication. Under UHV conditions, metals were deposited via electron beam in a chamber, base pressure ~1  ×  10⁻⁹ mbar, attached to a UHV cluster tool described in detail elsewhere [34]. Following metal deposition, samples were transferred, without breaking vacuum, to a separate chamber through a UHV transfer tube held at 1  ×  10⁻¹⁰ mbar where XPS was performed. Under HV conditions, metals were deposited in an elastomer-sealed Temescal BJD-1800 e-beam evaporator with base pressure ~1  ×  10⁻⁶ mbar in the cleanroom facility [35]. Samples were then transferred ex situ as quickly as possible to the analysis chamber under UHV conditions for all XPS characterization (~5 min air exposure between removal from cleanroom deposition tool and loading into UHV cluster tool). All Au, Ir, and Cr films were deposited on WSe₂ with a target thickness of 1.5 nm.

The method by which the Au 4f, Ir 4f, and Cr 2p core level reference spectra were obtained is described elsewhere [29]. The W metal reference was obtained by sputtering 50 nm of W under an N₂ flow of 30 sccm onto a Si(1 1 1) wafer cleaned in a 100:1 HF bath. The base pressure of the interconnected sputtering chamber is ~2  ×  10⁻⁹ mbar. Prior to transferring the sample into the sputtering chamber for W deposition, the W target was outgassed for 1 h with identical parameters as those employed during deposition. Following outgassing, the chamber was pumped to a pressure of 2  ×  10⁻⁹ mbar before moving the sample inside from UHV transfer tube. The sputtering tool is employed solely for the purpose of obtaining the W reference spectrum, and is attached to the same UHV cluster tool as was mentioned earlier. Prior to obtaining the W 4f reference spectrum, the W film obtained by sputter deposition was bombarded with Ar⁺ ions accelerated under 5 kV and 25 mA for a total of 30 min to ensure a clean, oxide-free surface. The reference W 4f spectrum was obtained in situ via XPS under UHV conditions immediately following Ar⁺ ion treatment. The corresponding O 1s and C 1s core level spectra were also obtained to confirm a clean W film within the photoelectron sampling depth.

XPS characterization of as-exfoliated WSe₂ and WSe₂ following metal deposition was performed via a monochromated Al Kα source and Omicron EA125 hemispherical analyzer with  ±0.05 eV resolution. In addition, a takeoff angle of 45°, acceptance angle of 8°, and pass energy of 15 eV were employed during spectral acquisition. The analyzer was calibrated with Au, Ag, and Cu foils according to ASTM E1208 [36]. Spectra were deconvolved using AAnalyzer [37], a curve fitting software.

Reactions between Au and WSe₂ are not thermo-dynamically favorable. The persistence of WSe₂ is far more energetically favorable than the reduction of WSe₂ by Au to form AuSe as $\Delta \text{G}_{\text{f},\text{WS}{{\text{e}}_{\text{2}}}}^{{}^\circ}$  =  −132.0 kJ mol⁻¹, whereas the $\Delta \text{G}_{\text{f},\text{AuSe}}^{{}^\circ}$  =  −31.99 kJ mol⁻¹ [38]. The same presumably applies for other gold selenide stoichiometries. Figure 1(a) displays the Se 3d, W 4f, and W 5p_3/2 core level spectra obtained from bulk WSe₂ following exfoliation and subsequent Au deposition under UHV and HV conditions.

Zoom In Zoom Out Reset image size

Following exfoliation, the Se 3d_5/2 and W 4f_7/2 core levels are detected at binding energies of 55.08 and 32.85 eV, respectively, corresponding with a peak separation of 22.23 eV, typical of exfoliated, bulk WSe₂ [28, 39]. Following Au deposition under UHV conditions, no new chemical states are detected in either the Se 3d or W 4f core levels. This indicates Au forms a van der Waals interface with WSe₂, which has been predicted by DFT and verified in previous experiments [8, 23, 40]. Au deposition on WSe₂ under HV conditions also results in a van der Waals interface. In contrast with Au deposition under UHV conditions, the formation of WO_x via WSe₂ oxidation is observed, evidenced by the additional chemical state appearing at a binding energy of 35.25 eV in the corresponding W 4f core level spectrum. WSe₂ oxidation is thermodynamically favorable in the presence of gaseous oxidizing species typically present in the atmosphere (e.g. OH⁻, H₂O, CO⁻, CO₂) as $\Delta \text{G}_{\text{f},\text{W}{{\text{O}}_{\text{2}}}}^{{}^\circ}$  =  −533.9 kJ mol⁻¹ and $\Delta \text{G}_{\text{f},\text{W}{{\text{O}}_{\text{3}}}}^{{}^\circ}$  =  −764.1 kJ mol⁻¹, respectively [41]. Recently, DFT calculations predicted a particularly low kinetic energy barrier to oxygen adsorption on WSe₂ as a passivation mechanism under ambient conditions [42]. Addou et al and Park et al experimentally observed oxidation of WSe₂ domain edges, which are the most reactive sites within a WSe₂ domain [28, 43, 44]. Despite mild WSe₂ oxidation following Au deposition under HV conditions, the Se:W ratio remains virtually identical with that detected prior to and following Au deposition under UHV conditions (2.1:1). [The method employed in calculating stoichiometries reported here is discussed in further detail in the supporting information (stacks.iop.org/TDM/4/025084/mmedia).] The formation of WO_x rather than WO_xSe_y would result in complete loss of Se from WSe₂ and corresponding complete oxidation of W^x+ formed from WSe₂ reduction. Any Se–Se or Se–O bonds resulting from the liberation of Se from WSe₂ via oxidation is below the limit of detection due to the significant difference between atomic sensitivity factors of the W 4f and Se 3d core levels (W 4f $\gg$ Se 3d, see Supporting Information). WO_xSe_y formation would be evidenced by the appearance of chemical states in the W 4f, Se 3d, and O 1s core level spectra in addition to chemical states associated with bulk WSe₂ and WO_x (see Cr–WSe₂ formed under HV conditions). A change in WSe₂ stoichiometry resulting from WO_x formation via Au deposition under HV conditions is not expected and, if it occurs in this case, is below the detection limit of XPS.

The O 1s core level spectra obtained from the as-exfoliated WSe₂ surface and the Au–WSe₂ interface are displayed in figure 1(b). Oxygen-containing contaminants are below the XPS detection limit following exfoliation as there are no chemical states distinguishable from baseline noise in corresponding O 1s core level spectrum. A small concentration of oxygen related to adventitious species from background gases within the UHV cluster tool are detected following Au deposition. The doublet in the HV W 4f core level spectrum found at high binding energy indicating the presence of WO_x is accompanied by a chemical state in the O 1s core level spectrum detected at a binding energy of 530.79 eV, which is in agreement with previous XPS analysis of analogous WO_x species [45].

Adventitious carbon is detected on all as-exfoliated samples as expected due to brief (~5 min), unavoidable air exposure following ex situ exfoliation and prior to loading into UHV. However, there are no carbon-related chemical states detected in any of the W 4f, Se 3d, Au 4f, Ir 4f, Cr 2p, or corresponding C 1s core level spectra which would suggest the formation of metal-carbon bonds of any kind. Our recent work demonstrated similar inert behavior of adventitious carbon at the metal-MoS₂ interface [29] and therefore carbon at the metal-WSe₂ interface will not be discussed further here. However, a typical procedure employed to fabricate TMD based devices will involve exposing the channel material to air prior to patterning and metal deposition. Therefore, it is reasonable to assume the interface between metal and TMD channel will contain carbon, which could impact contact resistance.

The Au 4f core level spectra obtained from a Au reference film (~25 nm thick), and following Au deposition on WSe₂ under UHV and HV conditions are displayed in figure 1(c). A single chemical state is detected in each of the Au 4f core level spectra following Au deposition under UHV and HV conditions at a binding energy of 84.00 eV in both cases, evidenced by the near perfect overlap with reference Au 4f spectrum when normalized (figure S1). This chemical state is indicative of metallic Au (Au–Au bonding) due to the asymmetric tail to the high binding energy side typical of core hole screening effects in metals [46, 47]. This also indicates Au forms a van der Waals interface with WSe₂ regardless of deposition chamber ambient. Considering the attenuation of the W 4f and Se 3d core levels following Au deposition under UHV conditions, the thickness of the Au film is estimated to be 6 Å. This represents an estimate of the average film thickness. However, images obtained by atomic force microscopy (AFM) ex situ following Au deposition under UHV and HV conditions (figure S2) demonstrate Volmer–Weber (VW) type growth of Au on WSe₂ regardless of deposition chamber ambient. These results are consistent with a similar study alternatively employing scanning tunneling microscopy (STM) to investigate Au growth on WSe₂ [48]. Therefore, the estimate of Au thickness based upon the attenuation of the bulk WSe₂ chemical states is not representative of the Au film morphology in the initial stages of growth on WSe₂, specifically the height and diameter of Au clusters.

The binding energies of the Au 4f_7/2 core level obtained from Au films deposited on WSe₂ under UHV and HV conditions are 0.25 eV higher than that of the Au reference (83.75 eV). This suggests that the Au deposited under either UHV or HV conditions exhibits a smaller Φ than expected of bulk, polycrystalline Au, consistent with the size-dependent Φ of Au nanoparticles on a Si(1 1 1) substrate [49]. It is possible that a disagreement in the binding energies of reference Au and deposited Au indicates additional Au chemical states corresponding with reactions between Au and WSe₂. However, when the Au 4f core level spectra obtained following Au deposition under UHV and HV conditions are shifted 0.25 eV to lower binding energy and normalized with the reference Au 4f core level spectrum (figure S1), there are no unique spectral features resolvable which would indicate a chemical state of Au other than Au⁰. Chemical state perturbation at the Au–WSe₂ interface formed under UHV conditions, if present, is below the detection limit. However, the 0.25 eV Au 4f binding energy shift from reference Au and concomitant 0.25 eV shift to lower binding energy of the bulk WSe₂ chemical states in the W 4f and Se 3d core levels following Au deposition under UHV conditions suggests an interface dipole forms upon deposition of an incomplete Au film (see table S1 for binding energies of initial bulk WSe₂ chemical states). A similar interface dipole formation mechanism is predicted when Au is deposited on another layered material, GaSe [50]. Significant Schottky barrier height lowering has been reported when high Φ metal nanoparticles (e.g. Au, Pt) are present at the interface between a semiconductor and another substantially lower Φ metal contact (e.g. Ti) [51–53]. Therefore, the interface dipole mechanism between an incomplete coverage Au film (effectively Au nanoparticles) and WSe₂ could be analogously employed in conjunction with a substantially lower Φ electrode metal (e.g. Ti, Sc) as a Schottky barrier lowering technique for electron contacts on WSe₂-based devices.

Regardless of deposition chamber ambient, Ir reacts with WSe₂ to form sub-stoichiometric tungsten selenide (WSe_x, x  <  2) and corresponding IrSe_x via reduction of WSe₂. Figure 2(a) shows the Se 3d and W 4f core level spectra obtained following Ir deposition under UHV and HV conditions.

Zoom In Zoom Out Reset image size

Following deposition under UHV conditions, the bulk WSe₂ chemical states detected in the W 4f_7/2 and Se 3d_5/2 core level spectra at binding energies of 32.29 and 54.50 eV, respectively, have shifted 0.49 eV to lower binding energy from as-exfoliated WSe₂. This indicates that interaction between Ir and WSe₂ results in a FL shift towards the valence band maximum (VBM), which is expected as the FL of Ir (Φ_Ir  =  5.27 eV) [31] should reside deep in the band gap of WSe₂. The chemical state at high binding energy in the Se 3d core level spectrum obtained following Ir deposition under UHV and HV conditions indicates the formation of IrSe_x at the Ir–WSe₂ interface under either base pressure. The IrSe_x and WSe_x chemical states in the Se 3d, W 4f, and Ir 4f core level spectra exhibit binding energies which are seemingly dependent upon the deposition chamber ambient. This could be related to the difference in Ir film thickness achieved under UHV and HV as well as slight differences in reaction product stoichiometry. Reduction of WSe₂ to form sub-stoichiometric WSe_x would result in a simultaneous reduction and oxidation of the W⁴⁺ and Se²⁻ chemical states in WSe₂, respectively. The additional chemical states detected in the W 4f and Se 3d core level spectra at binding energies of 32.05 and 54.73 eV, respectively (figure 2(a)), are in agreement with the proposed reaction mechanism.

Figure 2(b) displays the O 1s core level spectra obtained from as-exfoliated WSe₂ and following Ir deposition under UHV and HV conditions. Adventitious oxygen present in O–C and O–C–H adsorbates is detected prior to Ir deposition due to the unavoidable brief duration of air exposure prior to loading into UHV. Following Ir deposition under UHV conditions, there are no chemical states detected in the O 1s core level spectrum below 532.0 eV that would indicate metal or chalcogen oxidation. However, following deposition under HV conditions, two additional chemical states appear with binding energies of 529.47 and 530.55 eV suggesting the formation of IrO_x and WO_x, respectively. Reduction of WSe₂ by Ir results in available W atoms free to react with residual oxidizing species. The electronegativity of Ir (2.20) is less than that of W (2.36) [54], which indicates iridium oxide should exhibit a chemical state in the O 1s core level spectrum with a lower binding energy than an analogous tungsten oxide species, consistent with the chemical states detected in this work. Formation of IrO_x and WO_x in the presence of excess oxidizing gases is thermodynamically favorable as the Gibbs free energies of formation of Ir₂Se₃ and WSe₂ are considerably more positive than that of IrO₂ and WO₂ or WO₃ [29].

Figure 2(c) displays the Ir 4f core level spectra obtained from an Ir reference film (see Methods section for more details) and following Ir deposition on WSe₂ under UHV and HV conditions. Both peaks comprising the reference Ir 4f doublet exhibit a significant degree of asymmetry, consistent with core-hole screening commonly observed in metallic species [46]. The intense asymmetric doublet detected with Ir 4f_7/2 core level at a binding energy of 60.78 eV following Ir deposition under UHV and HV conditions indicates that metallic Ir (Ir–Ir bonding) comprises the majority of the deposited film. However, a chemical state in addition to metallic Ir⁰ is detected in the Ir 4f core level spectra following Ir deposition under UHV and HV conditions with a binding energy of 61.40 and 61.67 eV, respectively, indicating the formation of a small concentration of IrSe_x. Unique to Ir deposition under HV conditions, a third chemical state is detected at high binding energy (62.59 eV) from the IrSe_x state, indicating the formation of a small concentration of IrO_x.

Thickness estimations using the attenuation of the total intensities of the Se 3d and W 4f core level spectra suggest that the thickness of the Ir film deposited under UHV conditions is 20 and 25 Å thick, respectively. This indicates Se diffusion into Ir in addition to Ir–Se interfacial bonding rather than Ir–Se bonding solely across the Ir–WSe₂ interface, similar to previous reports of sulfur out-diffusion when Ir is deposited on MoS₂ [29]. AFM images, which are discussed in greater detail in the Supporting Information, indicate Ir grows uniformly on WSe₂ regardless of deposition chamber ambient (figure S2). It is therefore reasonable to directly compare W 4f and Se 3d chemical state attenuation following Ir deposition under either base pressure to evaluate the Ir film thickness even at a thickness less than 2 nm. Intermetallic species comprised of the contact metal and chalcogenide at the metal-WSe₂ interface, such as IrSe_x in this case, may permit control of interface resistivity, similar to various silicides employed in MOSFET and CMOS technology [55]. The potential benefits of Ir and/or IrSe_x present at the contact-TMD interface are discussed further in the Supporting Information.

The Se 3d, W 4f, and W 5p_3/2 core level spectra obtained following Cr deposition on WSe₂ under both UHV and HV conditions are displayed in figure 3(a).

Zoom In Zoom Out Reset image size

Bulk WSe₂ is detected in the W 4f_7/2 (32.40 [UHV] and 32.58 eV [HV]) and Se 3d_5/2 (54.65 [UHV] and 54.84 eV [HV]) core level spectra obtained following Cr deposition under both UHV and HV conditions despite reactions between Cr and WSe₂ in both cases. An asymmetric doublet is detected in the W 4f core level spectra following Cr deposition under UHV and HV conditions at binding energies of 31.24 and 31.14 eV, respectively, indicating the formation of metallic W⁰ (W–W bonding). The binding energies and degrees of asymmetry of both the metallic W⁰ chemical state in the W 4f_7/2 and W 4f_5/2 core levels are in close agreement with those detected from a W reference (see Methods section for details regarding the W reference). In addition, Se scavenging by Cr results in the formation of CrSe_x regardless of deposition chamber base pressure, evidenced by the chemical states detected in the Se 3d core level spectra at binding energies of 54.35 (UHV) and 54.05 eV (HV). It is important to note that CrSe₂ is a member of the TMD family of layered materials with metallic transport properties and therefore CrSe_x formation could be advantageous for charge transport across the interface [56]. Complete reduction of WSe₂ by Cr to form metallic W and CrSe_x at the interface is thermodynamically favorable as $\Delta \text{G}_{\text{f},\text{C}{{\text{r}}_{\text{2}}}\text{S}{{\text{e}}_{3}}}^{{}^\circ}$ (−175.1 kJ mol⁻¹) [57] is more negative compared with $\Delta \text{G}_{\text{f},\text{WS}{{\text{e}}_{2}}}^{{}^\circ}$. However, when Cr is deposited under HV conditions, intermediate WSe_x (x  <  2) that presumably forms as Cr scavenges Se is not stable in an atmosphere containing an excess of oxidizing gases (i.e. under HV or in air). As a result, reduced tungsten selenide (WSe_x) oxidizes to form WO_xSe_y, as indicated by the chemical states detected in the W 4f and Se 3d core level spectra following Cr deposition under HV conditions at binding energies of 32.38 and 54.58 eV, respectively. A decrease in binding energy of the WO_xSe_y chemical state from that of bulk WSe₂ in the W 4f core level spectrum indicates a decrease in W oxidation state from WSe₂ (W⁴⁺) to WO_xSe_y (W³⁺). It may be intuitive to expect an increase in W oxidation state resulting from oxidation of WSe_x because O is significantly more electronegative than Se [54]. Cr reacting with Se atoms from WSe₂ will result in a reduced W oxidation state from W⁴⁺ in WSe₂. In the event that each Se atom involved in the formation of CrSe_x is not replaced by an oxygen atom during the formation of WO_xSe_y, it would be possible for W to undergo a net decrease in oxidation state from W⁴⁺ in WSe₂ to W³⁺ in WO_xSe_y. Cr exhibits a significantly smaller electronegativity (1.66) [54] than W. This suggests the Se²⁻ chemical state in a CrSe_x compound would appear at lower binding energy from an analogous WO_xSe_y compound, which is consistent with the chemical states reported here. In addition, a nearly undetectable concentration of metallic W is converted to WO_x, evidenced by the chemical state at high binding energy (35.15 eV) in the W 4f core level spectrum obtained following Cr deposition under HV conditions. As was mentioned earlier, the formation of WO_x is highly thermodynamically favorable.

Figure 3(b) displays the O 1s core level spectra obtained from exfoliated WSe₂ as well as WSe₂ following Cr deposition under UHV and HV conditions. Small concentrations of adventitious species, likely O–C and O–C–H, are detected in the O 1s core level spectrum between 532 and 535 eV following WSe₂ exfoliation. Concomitant chemical states corresponding with C–O and C–O–H are detected in the as-exfoliated WSe₂ C 1s core level spectrum (not shown here) between approximately 285 and 286 eV. This is expected due to unavoidable air exposure prior to introduction into UHV. Following Cr deposition under UHV conditions, oxygen is below the XPS detection limit. Photoelectrons originating from oxygen-based species initially present on the WSe₂ surface are no longer detectable due to attenuation by the Cr film. It is also possible that Cr reacts with adventitious species at the interface, however these reaction products are also below the limit of detection as there is no evidence of chromium carbide or chromium oxide in the UHV Cr 2p core level spectrum [58]. Following Cr deposition under HV conditions, a significant concentration of oxygen-based species are detected, including O–C–H (534.46 eV), O–C (532.95 eV), and WO_xSe_y (531.1 eV), similar to the binding energy previously reported from an analogous tungsten oxychalcogenide (WO_xS_y) chemical state in the O 1s core level spectrum [29]. A chemical state is also detected at 531.85 eV corresponding with the formation of Cr(OH)₃ according to the intensities of the related chemical states corrected with appropriate atomic sensitivity factors, which is in reasonably close agreement with previous XPS analysis of a bulk Cr(OH)₃ sample [59]. The most intense chemical state detected at 530.58 eV in the HV O 1s core level spectrum consists of a convolution of Cr_xO_y and WO_x chemical states. Transition metal oxide species can exhibit a range of binding energies between 528.0 and 531.0 eV depending on both the transition metal and its oxidation state [60].

There are substantial differences in Cr 2p core level spectra for different reactor base pressures. The Cr 2p core level spectra obtained from a Cr reference and following Cr deposition on WSe₂ under UHV and HV conditions are displayed in figure 3(c). The reference Cr 2p_3/2 and Cr 2p_1/2 core levels (see Methods section for more details) exhibit binding energies of 574.10 and 583.41 eV, respectively, FWHMs of 0.92 and 1.45 eV, respectively, and significant degrees of asymmetry [61]. Following Cr deposition on WSe₂ under UHV conditions, metallic Cr⁰ (Cr–Cr bonding) and CrSe_x are detected at binding energies of 574.05 and 575.74 eV [62], respectively. We only fit the Cr 2p_3/2 core level spectrum obtained following Cr deposition under HV conditions due to parameter constraints in the fitting software. A much smaller concentration of metallic Cr is detected following Cr deposition under HV conditions coinciding with the formation of a substantial concentration of both Cr_xO_y and Cr(OH)₃ as evidenced by the chemical states detected at 575.99 and 577.21 eV, respectively, in the HV Cr 2p core level spectrum [56, 59, 60]. The four peaks which appear at higher binding energy from the Cr_xO_y chemical state in the Cr 2p_3/2 core level spectrum (each denoted by a downward pointing arrow) arise from multiplet splitting common in transition metal ions with unpaired d-electrons [59]. Multiplet splitting observed in the 2p core levels of various first row transition metal ions (Cr³⁺ in Cr_xO_y in this case) is predicted theoretically [63] and the origins of which are discussed in further detail elsewhere [56, 59]. The binding energies of the chemical states associated with Cr_xO_y and Cr(OH)₃ are in much closer agreement with those of reference samples [59] when the four multiplet peaks are included in fitting the Cr 2p spectrum obtained following Cr deposition under HV conditions compared with binding energies of the same chemical states when the multiplet peaks are omitted. In addition, the stoichiometry of Cr_xO_y is reasonably close to Cr₂O₃, the most stable oxide of chromium, when multiplet peaks are included in fitting the HV Cr 2p core level spectrum. If multiplet peaks are omitted from the HV Cr 2p spectrum, the stoichiometry of the chromium hydroxide species is far from Cr(OH)₃, which is the most stable hydroxide of chromium. Cr(OH)₃ is localized to the outermost surface as is evident when comparing normalized O 1s core level spectra obtained at two different take off angles (figure S3). The binding energy of the Cr(OH)₃ chemical states in the O 1s and Cr 2p_3/2 core level spectra are consistent with previous XPS analysis of a Cr(OH)₃ reference sample and its formation has been observed in other transition metals via passivation upon air exposure [59]. The conversion of Cr to Cr_xO_y is highly thermodynamically favorable as $\Delta \text{G}_{\text{f},\text{C}{{\text{r}}_{2}}{{\text{O}}_{3}}}^{{}^\circ}$  =  −353 kJ mol⁻¹ and $\Delta \text{G}_{\text{f},\text{C}{{\text{r}}_{3}}{{\text{O}}_{4}}}^{{}^\circ}$  =  −383 kJ mol⁻¹ [41]. Unlike Sc, which completely oxidizes under HV conditions and following brief air exposure [29], Cr oxidation is not as thermodynamically favorable. This permits preservation of CrSe_x at the Cr–WSe₂ interface despite substantial oxidation of remaining metallic Cr. The location of CrSe_x at the interface formed with WSe₂ under HV conditions beneath metallic Cr, Cr_xO_y, and Cr(OH)₃ serves as a kinetic barrier to oxidation.

WO_x is detected at all metal-WSe₂ interfaces formed under HV conditions investigated in this work. It is therefore necessary to consider the presence of WO_x at the metal-WSe₂ interface as a typical device fabrication process will involve contact metal deposition under HV conditions. The binding energy exhibited by the W^x+ chemical state in WO_x following Au, Ir and Cr deposition under HV conditions suggests x  =  5 as these binding energies are between that previously detected from the W⁶⁺ and W⁴⁺ chemical states in WO₃ and WO₂, respectively [45]. In addition, WO_x detected at each metal-WSe₂ interface exhibits O:W ratios between 2.6:1 and 2.7:1. Preferential WSe₂ domain edge oxidation in the presence of excess oxidizing species has been predicted by DFT calculations and recently experimentally verified with STM and Scanning Tunneling Spectroscopy (STS) [44]. The same work theoretically predicted and experimentally verified metallic behavior at oxidized WSe₂ domain edges. XPS obtained from WSe₂ grown by chemical vapor deposition with oxidized edges indicates an O:W ratio between 2.6:1 and 2.7:1, consistent with oxygen deficient WO₃ studied elsewhere [44] and also with WO_x detected in this work. WO_x detected in this work is expected to be metallic even considering the error associated with stoichiometry calculation by XPS (±0.2). Photoelectron spectroscopy has demonstrated, both previously [64] and in this work (see supporting information), the formation of shallow occupied gap states (figure S4(d)) and concomitant Fermi level shift towards the conduction band edge upon the formation of oxygen vacancies in WO₃ (figure S4(a)). This results in metallic n-type conduction in amorphous WO_x (a-WO_x, x  <  3), the structure and electronic properties of which are likely analogous with WO_x detected on WSe₂ following metal deposition under HV conditions. We find that the surface of amorphous WO₃ (a-WO₃) exhibits an ionization energy (FL position) of 9.77 (6.90) eV, respectively, as indicated by ultra violet photoelectron spectroscopy (UPS) and XPS. However, generation of oxygen defects to form a-WO_x (x  <  3) results in a FL shift 0.20 eV towards the conduction band and corresponding appearance of additional chemical states at lower binding energy from the W⁶⁺ oxidation state in the W 4f core level spectrum (W⁵⁺, W⁴⁺) and higher binding energy from the intense oxide state in the O 1s core level spectrum. The high vacuum work functions of both a-WO₃ and a-WO_x we report here are in agreement with previous reports of WO_x serving as a p-type interlayer between channel and contact metal in WSe₂-based devices [18]. In addition, the high vacuum work functions of a-WO₃ and a-WO_x indicate that both have potential to serve as Ohmic hole contacts in FETs employing a number of different TMDs as the channel material. In addition, the degree of p-type Fermi level shift in WSe₂ can be controlled from  <0.1 eV to  >0.25 eV by varying WO_x coverage on the WSe₂ surface (data not shown here). This work highlights the potential for enhanced hole transport in WSe₂-based devices by hole injection (likely also applicable with other TMDs) via WO_x inserted at the contact-WSe₂ interface. It also suggests predominant hole transport in WSe₂-based devices could be a product of the interface chemistry formed between contact metal and WSe₂ under HV conditions. A more detailed discussion of the expected band alignment and potential implications on electrical transport between a-WO₃, a-WO_x, and WSe₂ can be found in the Supporting Information.

Theories attempting to explain the FLP mechanism between a 3D semiconductor and metal include, but are not limited to, the formation of gap states by decaying metal wavefunction into the semiconductor [65, 66] and also by defects and/or disorder at the metal/semiconductor interface [67]. Regardless of the mechanism, the formation of interface gap states play a critical role in FLP between metal and semiconductor [32]. McDonnell et al found that defects present in MoS₂ provide preferential conduction pathways across the metal-MoS₂ interface [27]. It is therefore possible that the defective nature of TMDs plays a critical role in FL realignment upon metallization of a TMD. In addition, strong orbital hybridization of the transition metal in the TMD and contact metal near the metal-TMD interface has been predicted by DFT [10, 32]. This facilitates charge redistribution across the interface and therefore influences the position of the pinned FL relative to the underlying bulk TMD. The Mo d-orbital contributes the majority of density of states near the conduction band minimum (CBM) of MoS₂ [32], hence the tendency for FLP near the MoS₂ CBM as strongest d-orbital overlap occurs there. Analogous FLP may occur in WSe₂ near the VBM as the W d-orbital contributes the majority of the density of states near the VBM of WSe₂ [10].

Figure 4 displays representative band diagrams and corresponding FL position of as-exfoliated WSe₂ (Φ_Initial, purple) and subsequent FL shift following metal deposition under UHV conditions (ΔΦ, light blue). The position of the FL prior to metal deposition is determined from the valence band spectra (figure S4), which are discussed further in the Supporting Information. The degree of FL shift resulting from metal deposition is determined by the corresponding binding energy shift of the bulk WSe₂ chemical states in the W 4f and Se 3d core level spectra following metal deposition. A FL shift towards the VBM will be denoted as positive. The FL is positioned 0.97, 0.87, and 0.89 eV from the VBM within the band gap of WSe₂ prior to Au, Ir, and Cr deposition, respectively. Assuming bulk WSe₂ exhibits an ionization energy of 5.33 eV [68], these values indicate Φ_Initial of 4.36, 4.46, and 4.44 eV prior to Au, Ir, and Cr deposition, respectively, and are consistent with n-type behavior [28]. The error associated with the Fermi level position following metal deposition is calculated to be  ±0.11 eV (95% confidence interval). This is derived by considering the error propagation associated with assuming the magnitudes of the electron affinity and band gap according to previous work and also from deducing the Fermi level position prior to and following metal deposition by XPS.

Zoom In Zoom Out Reset image size

Thus, Au deposition should result in a  +0.77 eV (p-type) FL shift considering Φ_Au  =  5.13 eV [31], however a  +0.27 eV shift is detected. Less severe band bending induced by Au deposition under either UHV or HV conditions is consistent with VW-type Au growth on WSe₂ results in Au clusters with significantly smaller Φ than expected of bulk, polycrystalline Au. Au nanoparticles on the order of 5 nm in diameter suspended on a Si(1 1 1) substrate can exhibit a Φ as small as 3.4 eV, with Φ approaching Φ_Au,bulk as particle size increases [49]. It is also possible that orbital hybridization near the interface accompanied by charge redistribution results in FLP and therefore an anomalously low Φ_Au. The 0.53 eV electron Schottky barrier measured in this work is in relatively close agreement with previous DFT predictions of electron Schottky barrier magnitude for Au contacts to monolayer (0.58 eV) and bilayer (0.66 eV) WSe₂ [8]. Nonetheless, this highlights the complex nature of FL realignment even at a van der Waals interface.

Similar to Au, Ir causes p-type FL shift (+0.50 eV), 0.24 eV smaller than expected considering the initial FL position of WSe₂. Unlike Au, the binding energy of the chemical state corresponding with metallic Ir⁰ in the Ir 4f core level spectra obtained following deposition under UHV conditions is in close agreement with that of the Ir reference. Therefore, additional gap states presumably arising from IrSe_x contribute to the resulting FL position and larger than expected hole Schottky barrier for an Ir contact on WSe₂.

Contrary to the electron Schottky barrier expectedfor a pristine Cr contact on WSe₂ (0.40 eV, Φ_Cr  =  4.50 eV [31]), Cr deposition under UHV conditions results in a  +0.41 eV FL shift consistent with Φ  =  4.92 eV and a substantial hole Schottky barrier. It is possible that a mixture of metallic W (Φ_W  =  4.55 eV) [31] and Cr_xSe_y (Φ_Cr2Se3  =  5.15 eV) [68] present at the Cr–WSe₂ interface could result in a similar FL position. Although no Cr_xO_y is detected following Cr deposition under UHV conditions, the electronic structure of chromium oxide varies substantially with composition [68–70] and therefore is likely to appreciably affect the FL position when Cr is deposited under HV conditions. Recently, n-type conduction has been reported for Cr contacts (HV) to multi-layer WSe₂-based devices in contrast with the band alignment obtained in this work between WSe₂ and Cr deposited under UHV conditions [71]. This highlights the need to understand the role of the interface on contact performance in TMD-based devices.

It is important to reiterate that experiments discussed here were performed on bulk WSe₂ crystals. However, single or few layer TMDs are most commonly employed as the channel material in device structures. The band structure of mono and few layer TMDs, and therefore the band alignment formed with various contact metals, can be quite different from that of the bulk analog. Monolayer (bilayer, bulk) WSe₂ exhibits a band gap of 2.2 (1.8, 1.3)  ±0.1 eV as measured by STS [44]. Angle-resolved photoemission spectroscopy indicates the VBM of monolayer (bilayer, bulk) WSe₂ reside 1.8 (1.5, 1.1) eV from the Fermi level [72]. Error in photoemission measurements manifests in both the resolution of the spectrometer and also the method by which the VBM is extracted, typically on the order of  ±0.03–0.04 eV [73]. Yeh et al [72] report a spectrometer resolution of  ±0.1 eV but none associated with the VBM extraction method employed. Therefore, an error of  ±0.13–0.14 eV associated with the ARPES derived VBM according to WSe₂ layer number is suspected. Together, these STM and ARPES measurements suggest a relatively unperturbed conduction band minimum (~0.3 eV offset from Fermi level) and increasingly large ionization energy and therefore bandgap with decreasing number of WSe₂ layers. The magnitude of electron Schottky barriers between the same contact metals deposited under analogous conditions as were studied in this work and mono or few-layer WSe₂ are therefore expected to be similar to those reported here. However, associated hole Schottky barriers are likely to be much greater in magnitude in the mono, bi-, and tri-layer cases due to substantial increase in associated ionization energy.

DFT predicts a direct proportionality between the degree of contact metal-TMD d-orbital overlap and interface transport properties [10, 23, 24]. Parameters such as interatomic distance and bond energy between contact metal and chalcogen are proposed as predictive factors for chemical and electrical properties of various metal-TMD interfaces. However, additional compounds arising from reactions in the vicinity of the interface will perturb orbital overlap further, resulting in different chemical and electrical interface properties. The FL shifts reported here provide evidence that interfacial reaction products in addition to the inherently defective nature of WSe₂ contribute to anomalous FL positions detected for various contact metals on WSe₂, and therefore must be considered in engineering low resistance Ohmic contacts.