Impact of MoS₂ layer transfer on electrostatics of MoS₂/SiO₂ interface

Since in the Van der Waals bonded stacks of two-dimensional (2D) layers the lattice matching requirements are relaxed, artificial heterojunctions combining dissimilar materials would be permitted with no detrimental mismatch-induced defects [1–3] promising a broad variety of electronic and optoelectronic device applications [4–17]. For example, few-monolayer (ML) MoS₂ has already been suggested to enable improvements of metal-insulator-semiconductor transistors including channel downscaling, better electrostatic control, and steep sub-threshold slope [18–26]. However, realization of practical devices is hampered by the fact that direct growth of the Van der Waals bonded stacks is problematic because the synthesis of the most common 2D materials, e.g. graphene or MoS₂, requires exposure of the few-ML film surface to the reactive ambient at elevated temperatures. As a result, the key technology enabling 2D stacking remains solely based on synthesis of a 2D layer on a sufficiently stable substrate (Cu, Pt, SiC, SiO₂, sapphire), exfoliation of the grown film, and subsequent layer transfer (or, alternatively, ink printing) to the desired target substrate. This transfer approach represents nowadays the only technically feasible way to fabricate Van der Waals heterojunctions and, as a result, is universally used. However, it is by far not benign, e.g. it may introduce organic contamination which impairs electrical properties of the transferred layers [27]. What is not realized yet is that the 2D layer transfer also has considerable impact on the electrostatic potential distribution at the 2D/substrate interfaces which may significantly affect electron transport in the 2D layers and their stacks [28], in particular, in tunneling-based devices [4, 26].

In the present work, by analyzing transport of optically excited electrons from MoS₂ films synthesized on top of Si/SiO₂ substrates to the same layers transferred on top of SiO₂ after being grown on another substrate (Si/SiO₂ or sapphire), we reveal that the MoS₂ transfer strongly influences the electrostatics of the MoS₂/SiO₂ interface. The observed significant (≥0.5 eV, up to 1 eV in extreme cases) increase of the electron barrier height upon layer transfer indicates the presence of a dipole in the interface region—an effect ascribed to interaction of water molecules with the oxide surface. This result obviously pertains to transfer of other 2D materials on SiO₂, exposing an important violation of electroneutrality at the interface which may lead to electron scattering and threshold voltage instabilities in electronic applications. This perturbation of electrostatics at the interface is probably responsible for the much degraded electron mobilities in 2D-semiconductor films transferred on top of SiO₂ as compared to theoretical expectations. At the same time, thanks to the sufficient thermal stability of silanols, the revealed effect may be considered for adjusting threshold voltages beneficial for low-voltage circuit applications. As a technological advancement, we reveal that application of post-transfer annealing in H-containing ambient at least partially restores the electrical neutrality of the MoS₂/SiO₂ interface.

Thanks to the possibility of direct synthesis on top of SiO₂ and the good chemical stability of MoS₂ prepared by metallic Mo sulfurization, the latter represents an optimal 'test vehicle' to examine electrical impact of the layer transfer. The technologies of synthesis and transfer of few-ML thin MoS₂ films used in this work have been described in previous publications [29–34]. In short, the synthesis starts from thermal evaporation of few-Å thin Mo films on 8''-diam. SiO₂(50 or 90 nm)/Si or 6''-diam. c-plane sapphire substrates under high vacuum (10⁻⁶ mbar) at a deposition rate of 0.01 nm s⁻¹. Then the samples were extracted from the deposition chamber and loaded (with air-break) into a lamp-heated annealing system pumped to a base pressure of 10⁻² mbar before starting the annealing process in a pure H₂S atmosphere at a pressure of 10 mbar at 800 °C (Si/SiO₂ substrates) or 1000 °C (sapphire substrates). Some of the grown MoS₂ films were then transferred onto 50 or 90 nm thick SiO₂ layers thermally grown on Si substrates. To this purpose, we used a water intercalation-based tape assisted transfer method [34, 35] which starts from the delamination of the MoS₂ from the growth substrate and followed by subsequent bonding of MoS₂ on the target SiO₂ surface. Initially a poly(methyl methacrylate) layer (PMMA) is spin-coated over the entire MoS₂ surface. A rigid thermal tape (REVALPHA 3195 V) is placed over the desired area (several cm²) of the PMMA/MoS₂/substrate stack and then immersed in deionized water at 80 °C in a beaker. The beaker is collocated in a heated ultrasound bath to facilitate water intercalation. Next, the tape/PMMA/MoS₂ stack is slowly peeled off and separated from the substrate. The stack is then removed from water, blow-dried with N₂ and then placed on the target surface. Next, the sample is put on a hot plate at temperature of 125 °C or 140 °C to allow the thermal tape to be peeled off from the protective PMMA layer. To remove PMMA the samples were wet-cleaned using isopropyl alcohol and hot acetone. Fifteen-min long post-transfer annealing treatments were carried out at 370 °C under a H₂S pressure of 10 mbar.

Composition and structural properties of the synthesized MoS₂ layers were analyzed by several techniques as reported elsewhere [29–33]. These include Rutherford backscattering (RBS) spectroscopy [31], x-ray photoelectron spectroscopy (XPS) [29, 31, 32], in-plane and cross-sectional transmission electron microscopy [16–20], atomic force microscopy [29, 31–33], grazing incidence x-ray diffraction [29], as well as Raman [29–32] and photoluminescence spectroscopy [31, 39]. These methods reveal that Mo sulfurization in H₂S results in the formation of semiconducting polycrystalline 2H-MoS₂ layers, 1–5 ML thick depending on the thickness of initially evaporated metallic Mo, with good uniformity across the sample area. More results on structural characterization can be found in the supporting information, which is available online at stacks.iop.org/NANO/30/055702/mmedia.

In the present work we focus on the analysis of electron states at the MoS₂/oxide interface by XPS using Al K_α excitation and by internal photoemission (IPE) spectroscopy. In the latter case, electrical contacts to few-ML thick MoS₂ layers were prepared by thermal evaporation of optically non-transparent (100 nm thick) Au or Al pads (0.01 mm² area) through a shadow mask leaving the MoS₂ surface intact. The IPE current was measured under negative bias applied to the MoS₂ and illuminating the sample by light with known photon energy hν [30]. To ensure the absence of substantial degradation of the ultrathin sulfurization-grown MoS₂ films in air, IPE measurements after initial probing right after synthesis, have been repeated over extended period of sample storage (>20 month after synthesis). The experiments indicate absence of any substantial MoS₂/SiO₂ barrier height variation (within an accuracy limit of ±0.1 eV) after storage as opposed to the effect of layer transfer processing discussed in the next section.

An initial indication of the MoS₂ transfer influence on the electrostatic potential distribution at the interface with the underlying SiO₂ is provided by the XPS data shown in figures 1(a), (b): upon transfer of a 2ML-MoS₂ film from sapphire to SiO₂ (red spectra) the Mo3d_3/2−3d_5/2 and S2p_1/2−2p_3/2 core doublets shift towards lower binding energy by ≈1 eV as compared to the as-synthesized film (black spectra) while no signs of Mo or S oxidized states can be noticed. Annealing in H₂S (blue and green spectra) partially restores the peak energies. Surprisingly, lowering of the thermal tape release temperature from 140 °C to 125 °C helps to return the binding energy closer to the initial (prior to the transfer) value despite the same high-temperature (370 °C) post transfer anneal finally applied in both cases. At the same time, neither annealing in H₂S nor variation of the tape release temperature seems to affect the O1s binding energy, as evidenced in figure 1(c); only a reduction in the intensity of the broad low-energy shoulder after H₂S annealing can be noticed. However, association of this chemical state with oxidized sulfur (SO_x) is not supported because of the absence of the corresponding features that would be expected to appear in the binding energy range 168–170 eV in the S2p spectra shown in panel (b). Probably then, the binding energy shift in atoms pertaining to the transferred MoS₂ layer reflects the transfer-induced variation of electrostatic potential across the MoS₂/SiO₂ interface. One also may notice a slight broadening of the Mo Mo3d_3/2−3d_5/2 spectral lines after transfer (from ≈0.8 eV to ≈0.95 eV) while the linewidth of the S2p_1/2−2p_3/2 spectra is barely affected. This effect points towards some site-specific variation of the electrostatic potential along MoS₂ film surface. Hypothetically, it might be caused by adsorption (or intercalation) of the MoS₂ film by polar molecules, e.g. H₂O, during layer transfer.

Zoom In Zoom Out Reset image size

In order to exclude the influence of x-ray induced charging effects caused by ionizing radiation in the oxide layer [36, 37], we used IPE that probes the interface potential barrier for electrons between the valence band top of MoS₂ and the conduction band bottom of SiO₂ in the most straightforward way [29, 30, 38]—a procedure not affected by the radiation-induced charges. Comparison of the electron IPE spectra from MoS₂ directly-synthesized on SiO₂ with those of MoS₂ transferred onto identical SiO₂ layers (figure 2) indicates a considerable increase of the IPE spectral threshold, with attendant weaker field dependence, after the transfer, which correlates with the XPS results. Apparently, the film transfer induces a stepwise electrostatic potential variation at the SiO₂/MoS₂ interface. Moreover, it is noticed that in some MoS₂ samples transferred from sapphire, an even larger upshift of the IPE spectral threshold (up to ≈1 eV in some extreme cases as illustrated in figure 3) can be found near the sample edge despite the fact that Raman spectra (not shown) indicate good uniformity of the initial MoS₂ film itself over a length scale of ≈5 cm. At the same time, simple coating of the MoS₂ surface by PMMA (without bake) followed by removal in unheated acetone bath do not lead to any measurable variation of the IPE spectra. These observations point to the SiO₂ surface as the possible source of changes in electrostatic potential occurring during MoS₂ transfer.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

IPE spectra taken from the samples subjected to the post-transfer annealing in H₂S indicate the electron barrier to become somewhat lower (see figure 3), approaching its pre-transfer value (figure 2). However, the field dependence of the IPE spectra threshold remains much weaker than in the reference (no transfer) MoS₂/SiO₂ samples as can be seen from the Schottky plots shown in figure 4. This behavior may be explained by incorporation at the SiO₂/MoS₂ interface during MoS₂ layer transfer of interface dipoles and/or negative charges that are partially 'repaired' during post-transfer annealing in H₂S. Taking into account that the surface of SiO₂ is exposed to air moisture with water molecules remaining at the oxide surface during thermal tape release (130 °C–140 °C), it may prove worth to address interactions of H₂O with the oxide [39] as the possible reason of the electroneutrality violation at MoS₂/SiO₂ interface.

Zoom In Zoom Out Reset image size

Silica surfaces in air are known to be covered by a layer of silanols (SiOH groups) of various geometries and water molecules [40]. These remain present even after transfer of a 2D-layer on top as is suggested to happen in the case of graphene on SiO₂ [41]. Potentially, an array of polar O–H bonds in silanol groups with preferential orientation along the normal to the surface can form a dipole layer [42]. If assuming an effective charge transfer of 0.452 electron per H atom along the 0.12 nm long OH bond [42], the formation of a 1 eV dipole would require a SiOH entity density in the range of 1–4 nm⁻², depending on the chosen static permittivity value (1 for vacuum or 3.9 for SiO₂). This density agrees well with the density of silanols found at the silica surface varying from ca.3 to ca.0.8 nm⁻² as the temperature increases from 200 °C to 700 °C [43], suggesting that silanols are realistic candidates to account for the interface barrier increase upon MoS₂ layer transfer onto SiO₂.

On the other hand, it is long known that water permeates into thermally grown SiO₂ at relatively low temperature (<150 °C), leading to the formation of electron trapping centers [44, 45]. Both silanol (O₃Si–OH) groups and interstitial water molecules have been identified as electron traps in SiO₂ [46]. Trapping of electrons will result in a plane of negative charges which would 'pin' the top of the potential barrier at the MoS₂/SiO₂ interface, thus reducing its sensitivity to the externally applied electric field observed in the IPE spectra. Annealing in forming gas (N₂ + 10%H₂) at 400 °C eliminates the silanol groups due to reaction with atomic hydrogen generated at the metal/SiO₂ interface [46] providing a technology-compatible process to restore the original oxide charge neutrality. However, since the application of forming gas anneal may lead to a sulfur loss from MoS₂ [47], we used annealing in H₂S instead. This processing is thought to be capable of 'repairing' both possible S-deficiency in MoS₂ and remove silanols from the oxide surface. The annealing temperature of 370 °C was chosen to ensure thermal desorption of water and other adsorbates from the sample surface but still within the typical 'back-end-of-the-line' thermal budget. Indeed, upon such treatment, the IPE threshold shifts back towards the ≈4–4.2 eV range. However, the threshold field dependence is still much weaker than that in control (no transfer) samples (also shown in figure 4) suggesting that not all negative charges are removed from the MoS₂/SiO₂ interface.

In order to distinguish between barrier height modifying contributions from interface dipole and interface charges, we conducted electrical conductivity measurements on MoS₂ layers using backside gating. The current–voltage curves shown in the Supporting Information for the MoS₂ films transferred on 90 nm thick SiO₂ layer can be compared to the previously reported results for the case of directly-synthesized MoS₂ on SiO₂ [16], revealing the absence of any significant shift along the backside gate voltage (V_g) axis. At the same time, these measurements affirm good lateral uniformity of the transferred MoS₂ layer at the length scale of several cm, which supports the earlier conclusion that the macroscopic inhomogeneity of the interface barrier (figure 3) is probably related to the surface of the SiO₂ underlayer. On this basis we suggest that an interface dipole rather than a net charge in SiO₂ is responsible for the interface barrier changes upon transfer of MoS₂ layers on SiO₂.

More detailed electrical analysis of back-gated sulfurization-grown MoS₂ transistors as well as patterned structures used in the transfer-line method (TLM) has been presented in earlier publications [29, 32, 33]. The results indicate that the electron mobility increases from ≈ 0.2 cm² V⁻¹ s⁻¹ in layers synthesized at 800 °C on SiO₂ to ≈ 3 cm² V⁻¹ s⁻¹ in layers grown at 1000 °C on sapphire (see table 1 in [29]) which is ascribed to better crystallinity of MoS₂. At the lower end, these values are comparable to those reported for single-crystal bi-layer MoS₂ flakes obtained by defoliation if probed in air (0.12 cm² V⁻¹ s⁻¹) [48] and to the mobilities of <0.1 cm² V⁻¹ s⁻¹ reported for large-area MoS₂ films prepared by similar metal sulfurization process [49]. Removal of adsorbates from un-encapsulated MoS₂ flakes in vacuum leads to significant mobility improvement [48, 50] but it remains well below the theoretical estimate of ≈410 cm² V⁻¹ s⁻¹[51]. In line with these observations is the significant improvement of electron mobility (up to 12 cm² V⁻¹ s⁻¹) observed in the sulfurization-grown MoS₂ layers studied in this work upon post-transfer capping by chemical-vapor deposited (CVD) SiO₂ [28] which can also be a result of the Si-OH group removal effect due to atomic hydrogen release during the oxide CVD. Similar mobility values, close to 10 cm² V⁻¹ s⁻¹ for 1 ML and to 20 cm² V⁻¹ s⁻¹ for a bi-layer MoS₂, were recently reported for the films grown by CVD from ther Mo(CO)₆ precursor [52].

It is also noticed that Ni/MoS₂ contact resistance (Rc) values measured in TLM structures are sensitive to the air and water exposure [28]. Typical values of Rc for the studied sulfurization-grown MoS₂ films transferred onto SiO₂ are in the range 25–100 kΩ μm with a trend to increase at low back gate voltage (see upper inset in figure 3(a) in [33]). This behavior may be explained by the formation of a depletion layer in n-type MoS₂ due to the considerable (≈1 eV) barrier between the Fermi level of Ni (3.7 eV below the CB bottom of SiO₂, see table 1 in [53]) and the CB of the sulfurization-grown MoS₂, see band diagram shown in figure 5(b) in [30].

The revealed violation of electroneutrality of the MoS₂/SiO₂ interface upon layer transfer processing may have a significant impact on electron transport in the MoS₂ film because the dipole potential of polar silanol groups may result in additional carrier scattering. The high density of these scattering centers, in the order of ≈1 nm² as suggested by the observed electron barrier changes, may explain the inferior transport properties of 2D films transferred on SiO₂ as compared to other substrate materials such as BN [27]. Though the beneficial effect of the post-transfer anneal in H₂S (at 370 °C) is obvious, the anneal temperature should probably be increased to facilitate more efficient silanol removal. This, however, might cause a problem if non-sulfide materials, such as graphene, are used in the 2D stack since theyse can be damaged during anneal [54]. It is also worth of adding here that similar violation of interface electroneutrality can also be expected in the case of the ink-printing approach to 2D-heterojunction device fabrication [55]. In other words, not only stability of the transferred 2D semiconductor layer matters but, also, that of the target substrate as well. To resolve this issue, non-oxide substrates with more stable surfaces might help to maintain the electrostatic neutrality.

In conclusion, experiments reveal significant, up to ≈1 eV, interface barrier changes caused by transfer of MoS₂ layers onto thermally grown SiO₂. The results suggest that the formation of an interface dipole layer is responsible for the observed barrier instability which is tentatively associated with the presence of silanol groups at the surface of the SiO₂ layer below the MoS₂ film. Post-transfer annealing in H₂S appears to be useful in reducing the interface barrier perturbation suggesting that electroneutrality of the interface is restored at least partially.

This work was supported by the Flanders Innovation & Entrepreneurship [2Dfun (2D functional MX₂/graphene hetero-structures), an ERA-NET project in the framework of the EU Graphene Flagship] and by the KU Leuven Internal Fund (project C14/16/061).