Band alignment at interfaces of synthetic few-monolayer MoS₂ with SiO₂ from internal photoemission

Thanks to the relaxed lattice matching conditions in the Van der Waals bonded stacks, 2-dimensional (2D) layers potentially permit almost any material combination to create artificial heterojunctions without introducing detrimental mismatch-induced defects.^1,2 As a result, electronic properties of these heterostructures can be tailored “on demand” enabling the development of broad variety of electronic and optoelectronic devices.^3–7 For example, 2D transition metal dichalcogenide (TMD) semiconductors and, in particular, few-monolayer (ML) MoS₂ have already been shown to enable metal-insulator-semiconductor (MIS) transistor channel downscaling and improved electrostatic control.^8–12 Furthermore, MoS₂ MLs can be used in tunneling stacks to fabricate transistors with very steep sub-threshold slope which is mandatory for low voltage electronics.¹³ 

Obviously, the characteristics of the TMD-based devices are critically sensitive to the band alignment with other materials which determines electrostatics of the stack, including the threshold and the built-in voltages, height of the tunneling barriers, contact resistance, etc. However, band offset characterization at interfaces of few-ML two-dimensional (2D) materials represents an experimental challenge. For example, results reported for the ‘model’ MoS₂/SiO₂ interface characterization using electron photoemission in vacuum differ by ≈800 meV in the derived valence band (VB) offset value.^14,15 This discrepancy may be due to oxide charging as indicated by the vacuum level difference between bare SiO₂ and MoS₂ flakes as well as by the effect of oxide illumination by ionizing radiation.¹⁵ Also, one cannot exclude that the stoichiometry of MoS₂ used in the latter work was affected by the applied 12 h vacuum anneal at 300 °C to remove contaminants: Sulfur loss upon such treatment is reported already at 200 °C, while annealing at 300 °C affects even the MoS₂ VB photoemission spectrum.¹⁶ To avoid these surface contamination and the oxide charging artifacts, in the present work, we address the band offset issue by using the spectroscopy of Internal PhotoEmission (IPE) of electrons from the VB of 1- or 3-ML thick MoS₂ films synthesized directly on SiO₂ in the same way as described in Ref. 14. Further, we compared these results to the band alignment at the SiO₂/MoS₂ interfaces fabricated by sulfurization of metallic Mo by annealing in H₂S^17,18 revealing a significant (≈0.5 eV) electron barrier increase which is tentatively ascribed to the formation of an interface dipole due to hydrogen interaction with SiO₂.

The studied large area 1- and 3-ML thick MoS₂ films identical to those previously used in X-ray photoelectron spectroscopy (XPS) study¹⁴ were grown on top of the p-type (100)Si/SiO₂(25 nm) substrates by dc magnetron sputtering of Mo in sulfur vapor (the Mo + S process). The deposition was performed at a substrate temperature of 700 °C, a S partial pressure of 4 × 10⁻⁷ mbar, and Ar pressure 6 × 10⁻⁴ mbar.^14,19 Results of structural characterization of these layers have been reported elsewhere.^19,20 To explore the influence of hydrogen, the MoS₂ films synthesized using the H-free Mo + S route were compared to the 2- and 4-ML MoS₂ films grown on top of the (100)Si/SiO₂(50 nm) substrates by using sulfurization of sputtered metallic Mo in pure H₂S (100 mbar) at 800 °C (the Mo + H₂S process). Results of extended structural analysis of these films can be found in Refs. 17 and 18.

For IPE experiments, optically non-transparent (100-nm thick) Al or Au contact pads were thermo-resistively evaporated on top of the MoS₂ film to ensure reliable electrical contacts,¹⁸ while a blanket Al layer on the backside of Si served as the substrate contact. When illuminating the sample by photons of known energy, hν, electrons excited in the MoS₂ film at the periphery of the negatively biased contact pad are injected into SiO₂ and drift towards the Si substrate thus generating the IPE current which is measured in the external circuit as illustrated in Fig. 1. The quantum yield of IPE (Y) is defined as the photocurrent normalized to the incident photon flux and measured as a function of hν in the range hν = 3-5.5 eV with a constant spectral resolution of 2 nm.²¹ The energy dependences of the yield were analyzed at different MoS₂ bias voltages V_g to determine spectral threshold, Φ_e, of IPE corresponding to the interface barrier height between occupied electron states in the VB of MoS₂ and unoccupied states in the conduction band (CB) of the SiO₂ insulator.^21,22

The IPE yield spectra measured on samples with 1- and 3-ML MoS₂ films (Mo + S process, Au pads) are shown in Fig. 2 as semi-logarithmic plots (a) and Y^1/3-hν plots [(b) and (c)] used to determine spectral thresholds of IPE.²² As revealed earlier by the XPS analysis,⁶ the density of states in the MoS₂ VB increases approximately linearly with binding energy in the range of 1 eV below the VB top. In agreement with this result, the quantum yield of IPE increases with photon energy as ∼(hν − Φ_e)³ (Ref. 22) which corresponds to the observed linear increase on the Y^1/3-hν plots shown in (b) and (c). Any contribution to the photocurrent of electron IPE from the metal pad can firmly be excluded because changing the pad material from Au to Al has no influence on the yield spectra [cf. the inset in (b)]. The spectral distributions of electron IPE from Au are also different from those of MoS₂ samples as evident from Fig. 2(a), additionally showing the yield spectrum from a control capacitor fabricated by evaporation of a semitransparent (13 nm) Au electrode of 0.5 mm² area on top of the same Si/SiO₂(25 nm) substrate. Finally, association of the photocurrent with electron IPE from MoS₂ gains additional support from the signal enhancement observed in the case of 3-ML MoS₂ as compared to the 1-ML film (Fig. 2) which reflects the increase of the photoexcited MoS₂ volume.¹⁸ 

Spectral thresholds of electron IPE were determined using linear extrapolation of the Y^1/3-hν plots to zero yield²¹ and then plotting the inferred threshold values as a function or electric field strength (F) in SiO₂ using the Schottky co-ordinates Φ_e(F)- F^1/2, in this way enabling to account for the field-induced barrier lowering.^21,22 The results compiled in Fig. 3 for 1- and 3-ML MoS₂ (Mo + S process) samples with Au and Al contact pads suggest a barrier lowering with the image force dielectric constant ε_i ≈ 2. This value is close to ε_i ≈ n², where n = 1.46 is the refractive index of SiO₂ and corresponds to the ideal case of a charge-free interface.²³ The observed weaker field dependence of the IPE threshold for the case of the 1-ML sample with the Au pad (cf. Fig. 3) might be related to the contribution of electrons excited in MoS₂ below the metal as was earlier observed in the Mo + H₂S synthesized samples.¹⁸ Nevertheless, the major effect of the MoS₂ thickness change is clear: The VB top in 1-ML MoS₂ lies ≈400 meV below the VB top in the 3-ML film as referenced to the common reference level of the SiO₂ CB bottom. Taking into account the experimental accuracy limit of ±50 meV on the spectral threshold and ±100 meV on the zero-field barrier value, this VB shift agrees reasonably with the ≈300 meV value inferred from XPS.¹⁴ However, it is approximately twice as large as the difference of the ionization potential between 1-ML and 3-ML MoS₂ flakes,¹⁵ pointing to some additional contribution to the barrier height at the MoS₂/SiO₂ interface as opposed to the exposed MoS₂ surface. Nevertheless, if taken together, the experimentally observed VB shifts appear to be systematically smaller than the theoretical estimates^23,24 predicting an MoS₂ gap narrowing from E_g = 1.9 eV (1 ML) to E_g = 1.45 eV (3 ML) to occur through a shift in the VB edge.

There is, however, a noticeable difference in the MoS₂ gap position relative to the SiO₂ gap edges inferred from IPE and XPS experiments: For the 1-ML film, the CB offset can be estimated as ΔE_C = Φ_e(F = 0) − E_g≈ 2.1 eV from IPE compared to ΔE_C ≈ 3 eV from XPS.¹⁴ There are two possible explanations for this discrepancy. First, the samples used in the IPE experiments have been exposed to air, which contains oxygen and moisture that will form an adsorbate layer on the MoS₂ surface²⁵ while the XPS measurements were done in situ. Though the IPE measurements repeated over a period of 8 months gave results reproducible within the above indicated accuracy, the formation of an adsorbate layer during sample unloading and handling cannot be excluded. This, however, makes the IPE results more pertinent to the majority of experiments, being conducted ex situ in which MoS₂ is exposed to air or even to liquids during layer transfer.

In order to evaluate the influence of air exposure, the XPS measurements were repeated ex situ (see the supplementary material) and compared to the in situ results.^14,19 The S2p_1/2/S2p_3/2 spectra exhibit no shift or broadening indicating that the basal S-terminated surface plane of MoS₂ crystallites remains intact. The Mo3d_3/2 spectrum reveals a weak satellite at a binding energy of ≈236.8 eV attributed to the presence of Mo6+ states corresponding to the oxidized Mo.²⁶ At the same time, XPS analysis (spectra not shown) of the air-exposed MoS₂ layers grown by sulfurization of metallic Mo in H₂S reveals no significant traces of Mo or S oxidation affirming the intrinsic chemical stability of the MoS₂ basal plane. This comparison suggests that the interaction of few-ML polycrystalline MoS₂ with atmospheric oxidants occurs at the grain boundary regions. Taking into account that electron states in the VB of oxidized MoS₂ are expected to be significantly deeper in energy than in the MoS₂ VB (cf. Fig. 4 in Ref. 26), the oxidation cannot account for the lower CB offset inferred from the IPE measurements. Though the C1s signal at a binding energy of ≈285 eV was detected after air exposure in samples of both types (see the supplementary material), the absence of detectable changes in the position or lineshape of S2p and Mo3d doublets suggests that there occurs neither chemical interaction nor charge transfer. Probably then, the C1s line stems from the adsorbates on top of the MoS₂ and has no significant effect on band alignment.

More relevant seems to be the second explanation based on the differential charging effect arising from the difference in the escape depth of electrons from MoS₂ (1 ML of ≈0.65 nm) and from SiO₂ (Si 2p electron escape depth of ≈4 nm for the used Al K_a X-ray source²⁷): As indicated in Ref. 15, exposure of the SiO₂/MoS₂ system to continuous ionizing radiation (He lamp) leads to a photoemission threshold shift, indicating a non-zero electrostatic potential variation across the near-interface oxide layer. Even if using the adsorbate-related C1s signal as the energy reference, the X-ray induced positive charging of SiO₂ would cause the downshift of energy levels at the sample surface as compared to the neutral sample²⁸ which explains observation of a larger CB offset in XPS. Nevertheless, both XPS and IPE techniques appear to be consistent in evaluation of the thickness-dependent MoS₂ VB top shift.

Next, we did evaluate the interface barrier in samples prepared by the Mo + H₂S synthesis. Comparison of the Y^1/3-hν plots for 3-ML (Mo + S) and 4-ML (Mo + H₂S) samples [Fig. 4(a)] reveals a similar yield behavior indicative of close electron energy distributions near the VB top in both flavors of MoS₂. However, the thresholds in the latter case appear systematically higher than in the H-free samples. From the Schottky plot shown in Fig. 4(b), a ≈500 meV barrier increase in the Mo + H₂S sample can clearly be seen—a rather unexpected result: Theory predicts virtually no VB edge shift to occur when increasing the thickness from 3 to 4 MLs.^23,24 Indeed, the difference in the VB top energy in 2- and 4-ML thick MoS₂ layers (Mo + H₂S synthesis) is hardly noticeable in the IPE experiments.^17,18 The earlier discussed XPS results indicate no compositional differences between two flavors of MoS₂ except for the mentioned traces of oxidized Mo in the air-exposed (Mo + S) samples as compared to the (Mo + H₂S) synthesis. However, oxidation of MoS₂ would shift the VB top deeper in energy²⁶ leading to a higher IPE threshold in the (Mo + S) than in the (Mo + H₂S) case which is inconsistent with experiment (Fig. 4). Since slopes of the Schottky plots [Fig. 4(b)] in both cases are close and correspond to the ideal charge-free case, the only plausible explanation of the barrier enhancement consists in the formation of a dipole layer at the MoS₂/SiO₂ interface due to the presence of hydrogen during Mo + H₂S synthesis as schematically illustrated in the inset in Fig. 4(a). Hypothetically this dipole layer can be seen as an array of polar O-H bonds in silanol (SiOH) groups with preferential orientation along the normal to the surface, as has been suggested previously for the graphene/SiO₂ interface.²⁹ If assuming an effective charge transfer of 0.452 electron per H atom along the 0.12-nm long OH bond,³⁰ the formation of a 0.5 eV dipole would require a SiOH group density in the range of 0.5-2 nm⁻², depending on the static permittivity value (1 for vacuum or 3.9 for SiO₂). This density agrees well with the density of silanols from ca.0.8 to ca.3 nm⁻², depending on temperature, found at the silica surface.³¹ 

To conclude, the presented results of IPE analysis of few-ML MoS₂/SiO₂ interface barriers suggest their significant sensitivity not only to the thickness-dependent electronic structure of the semiconducting MoS₂ films but also to the MoS₂ synthesis route. In particular, the presence of hydrogen during Mo sulfurization is found to introduce a significant (≈0.5 eV) dipole probably by forming an array of silanol groups at the interface. Taking into account the known sensitivity of the silanol density to temperature and chemical environment, the interface barrier instability and variability may become significant issues in the fabrication of functional electron devices, mandating tight control of hydrogen-containing species during interface fabrication and further processing.

See supplementary material for XPS spectra of the air-exposed samples.

The work at KU Leuven received partial support from Flanders Innovation & Entrepreneurship [2Dfun (2D functional MX2/graphene hetero-structures), an ERA-NET project in the framework of the EU Graphene Flagship] and from KU Leuven Internal Fund (Project No. C14/16/061).