Boosting the electronic and catalytic properties of 2D semiconductors with supramolecular 2D hydrogen-bonded superlattices

The electronic properties of two-dimensional semiconductors can be strongly modulated by interfacing them with atomically precise self-assembled molecular lattices, yielding hybrid van der Waals heterostructures (vdWHs). While proof-of-concepts exploited molecular assemblies held together by lateral unspecific van der Waals interactions, the use of 2D supramolecular networks relying on specific non-covalent forces is still unexplored. Herein, prototypical hydrogen-bonded 2D networks of cyanuric acid (CA) and melamine (M) are self-assembled onto MoS2 and WSe2 forming hybrid organic/inorganic vdWHs. The charge carrier density of monolayer MoS2 exhibits an exponential increase with the decreasing area occupied by the CA·M unit cell, in a cooperatively amplified process, reaching 2.7 × 1013 cm−2 and thereby demonstrating strong n-doping. When the 2D CA·M network is used as buffer layer, a stark enhancement in the catalytic activity of monolayer MoS2 for hydrogen evolution reactions is observed, outperforming the platinum (Pt) catalyst via gate modulation.


Table of contents
displays the surface morphology a freshly cleaved MoS2 bulk crystal coated with a bi-component CA·M film, processed via drop-casting followed by the solvent evaporation at room temperature. The AFM image shows that the molecular film is characterized by several sub-μm domains with boundaries appearing as dark lines for non-annealed samples. These fragmented small domains merge forming larger, micrometer-scaled structures upon thermal annealing at 430K for 12h in vacuum. As illustrated in Fig. S3b, a more uniform and less defective CA·M supramolecular network is obtained over a 400 µm 2 area, with only sporadic non-coated surface areas observed. The height profile performed along the coated MoS2 surface (see white line on Fig. S3b) indicates that the CA·M film is monolayer, with thickness of ca. 4.8 Å. Moreover, the roughness of this CA·M supramolecular network decreased from 0.433 nm to 0.109 nm upon annealing. In order to quantify the thickness of the CA·M monolayer more accurately, we performed AFM topographical imaging of sub-monolayer thick films. Fig. S3c displays the statistics on the CA·M layer thickness executed by sampling data from 20 different sample's regions; the resulting average value amounts to 0.397 ± 0.103 nm. Moreover, the bare MoS2 surface has a roughness of about 0.1 nm, as depicted in Fig. S3d. Noteworthy, the distance between the two planar sheets in CA·M crystal has been previously estimated by AFM imaging as 0.317 nm. The difference in the CA·M monolayer thickness is mostly caused by the local environment. More precisely, while in our work the CA·M monolayer was physisorbed on MoS2 surface hence it encountered strong moleculesubstrate interactions, in the melamine-cyanuric acid crystal, the investigated CA·M layers are stacked via van der Waals forces. The use of different substrates can also lead to various adsorption distances, as our simulated vertical distance between CA·M molecular layer and MoS2 surface is about 0.33 nm, while the simulated average distance between CA·M molecular layer and Au (111) surface has been reported being 0.371 nm.

S4
The same CA M preparation was carried out on a mechanically exfoliated monolayer MoS2 flake (1L) transferred onto the SiO2 surface. The topographical images of pristine monolayer MoS2 are displayed in Fig. S3e and S3f, which reveals a smooth surface (roughness 0.164 nm) having a thickness of 6.1 Å. The phase image indicates that the exfoliated MoS2 surface is uniform and free of debris. After the CA·M SAM deposition, the MoS2 flake surface is completely coated (roughness = 0.133 nm), as exhibited in Fig. S3g. The CA·M network also assembles outside the MoS2 flake limits, i.e., on SiO2, producing a uniform molecular film that covers the whole sample area, as shown in Fig. S3h. The STM image of 2H phase MoS2 crystal surface is shown in Fig, S3i, and it displays a lattice constant of a = b = 3.14 Å and γ = 120°, in good accordance with theoretical lattice parameters. The E 1 2g and A1g modes of WSe2 are observed at approximately 250 cm -1 . These two modes are almost degenerate and cannot be easily resolved in the case of thin-layer WSe2

3
. As seen in Fig. S6a, we only find a single E 1 2g / A1g peak for the clean monolayer and bilayer flakes, where the peak position is slightly blue-shifted for bilayer WSe2 in respect to monolayer WSe2. Moreover, a broad side maximum at 260 cm -1 (2LA(M)) is observed for the two samples. Finally, a small signature at 309 cm -1 (B 1 2g) is evident in the Raman spectrum of bilayer WSe2 (Fig. S6c).
This can be assigned to the normally inactive B2g mode. Upon the CA·M SAM formation on the WeS2 surface, the E 1 2g / A1g band redshifts by 1.3 cm -1 and 1.0 cm -1 for monolayer and bilayer samples, respectively. Doping-induced Raman shift in WSe2 flakes can be explained by the change of electron density. The n-type doping increases the electron-phonon scattering due to a higher electron concentration. Upon CA M coating, significant quenching is observed in the PL spectra of both monolayer and bilayer flakes, as seen in Fig. S6b and 6d. The PL peak is strongly suppressed and redshifted as a result of electron doping caused by the CA·M supramolecular lattice formed on the WSe2 surface 4 . All these optical spectra were measured at room temperature in ambient. FETs based on bilayer WSe2 have been produced to evaluate the CA·M doping. The asprepared clean device (channel length / width is 2.31 µm / 3.49 µm) exhibits ambipolar transfer characteristics with slightly higher p-channel conductance, corresponding to close to mid-gap Schottky barriers at the source-drain Au contacts 5 . Upon deposition of the CA·M film, the bilayer WSe2 device shows unipolar n-type transfer characteristics, with a VTH shift towards a more negative voltage (Fig. S7b). Applying a VDS of 1V, the field-effect mobility (μ) of bare WSe2 is found amounting to 17.9 cm Devices prepared via photolithography typically require further high-temperature annealing to remove the residual solvent and photoresist, which can cause surface defects on WSe2 S8 surface 6 . Based on our experience, WSe2 is much easier to be oxidized than MoS2. Thus, to avoid this and obtain high-quality, stable p-type monolayer WSe2 devices, dry transfer method of Au nanomembrane electrodes is a viable route 7 . Au electrodes (90 nm-thick) were pre-deposited on a Si / SiO2 wafer and mechanically laminated on the TMDC crystal surface, resulting in an atomically flat metal-crystal interface 8 . Fig. S9a displays the topographic characteristics of the as-prepared device containing the transferred Au source and drain electrodes. The FET channel length (L) / width (W) is 1.6 µm / 2.6 µm for reported WSe2 monolayer device. After depositing the CA·M mixture, the coated device is annealed at lower temperatures (373K rather than typical 430K) for 12h to remove adsorbed water and oxygen. The AFM phase image (Fig. S9e) indicates though that 373K is not high enough to form a continuous 2D crystalline CA M supramolecular film. On the boundaries of co-assembled molecular domains, some exposed areas are evident. We also observed that the surface roughness of WSe2 monolayer exhibits a slight increase from 0.43 nm to 0.88 nm after partial coating with CA M. Back-gated monolayer WSe2 FETs operated at VDS of 1V change their ptype behavior to ambipolar one (Fig. S9b), where the hole μ decreases from 8.
. The n-type transport after CA·M coating is characterized by electron μ of 7.14 cm 2 V -1 s -1 . For monolayer WSe2 FETs, one expects a more evident n-doping than for fewlayer WSe2 devices. Here, our results do not meet such expectations because of the poor crystallinity and coverage exhibited by CA·M SAM formed at lower annealing temperatures (373K) annealing that weakens the doping effect to some extent. Figure 10: Raman (a,c) and PL (b,d) spectra acquired for monolayer MoS2 flakes before (black curve) and after (red curve) coating with individual M and CA S9 molecules, measured at room temperature in ambient.

Supplementary
Both individual CA and M molecules can dope (at a certain degree) the underlying monolayer MoS2 flakes, as shown in Fig. S10. Raman spectroscopy measurements revealed a shortening of the distance between E 1 2g and A1g peaks from 18.3 cm -1 to 17.9 cm -1 for M coating ( Figure   S10a), and to 16.9 cm -1 for the CA coating ( Figure S10c). Meanwhile, a marked quenching and redshift on the PL A exciton peak also confirm the n-doping of MoS2 upon functionalization with the individual CA and M molecules ( Fig. S10b and S10d). As seen in Fig. S11a and S11b from AFM measurements, M molecules arrange themselves into small, sub-μm domains after casting and solvent evaporation, which is difficult to image by STM at ambient conditions. When annealed at 430K in vacuum, such a film structure is destroyed and the M molecules aggregate into smaller particles, as shown in Fig. S11c and S11d. The intramolecular hydrogen bonds between M molecules are disrupted at such elevated temperatures which does not occur for the network formed from the bi-assembled CA·M structure. The morphology of the monolayer MoS2 device before and after M molecule deposition is shown in Fig. S11e-h. The surface roughness of the monolayer MoS2 flake in the device channel increases significantly after coating, from 0.62 nm to 3.05 nm. Some exposed MoS2 surface (darker regions in the AFM phase image) indicate that the on-surface coverage and the molecular film quality are significantly poor compared to the CA·M bi-assembled network.   S12a and S12b show the AFM height and phase of bulk MoS2 surface after coating with individual CA molecules, which is characterized by small, sub-μm aggregates. After vacuum annealing (430K, for 12h) the aggregates grow into islands, as seen in Fig. S12c and S12d. Onto the monolayer MoS2 surface, CA molecules aggregate densely after vacuum annealing ( Fig. S12g and S12h), leaving more than half of the device channel area exposed, i.e., noncoated. The surface roughness increases to 9.59 nm after coating with CA molecules, compared to 0.49 nm of the pristine MoS2 surface. s -1 after coating with individual CA molecules (Fig S13b). The negative shift of the FET transfer curves indicates that both M and CA molecules induce an n-doping effect to the underlying monolayer MoS2. From the FET transfer characteristics (Fig. S13b and  , see main text) is approximately an order of magnitude larger than that for MoS2 functionalization individual M molecules, and almost two orders of magnitude superior to that calculated for coatings with individual CA molecules.  The WF can be estimated from the measured SP difference between TMDC flake and Au electrode. As discussed in the main text, here we assume that the WF of polycrystalline Au at ambient conditions is WFAu = 4.85 eV, as determined via Kelvin probe force microscopy on macroscopic Au thin films produced via thermal evaporation. The WF of our flakes are calculated by using the equation: WFFLAKE = WFAu + e ΔVCPD 9 . ) with a continuous CA M monolayer produced via drop-casting is not feasible. In this sense, multilayer CA·M films are preferred to serve as a buffer layer between SiO2 substrates and the transferred TMDC flakes. Here, multilayer CA M is obtained by drop-casting 100 µL of 1 X 10 -6 M CA and M mixture on a preheated SiO2 wafer (430 K), and further annealed at 450K for 12h. Transferred monolayer MoS2 flake adheres well on CA·M film by mechanical exfoliation, forming a CA·M / MoS2 hybrid structure. As seen in AFM height image (Fig. S19a), the SiO2 / CA·M surface is smooth with roughness ca. 0.29 nm, while the attached monolayer MoS2 flake exhibits roughness of 0.28 nm. The flake thickness is measured as 0.77 nm, confirming its monolayer nature. AFM phase image reveals an excellent contrast between CA M and MoS2 (Fig. S19b). Fig. S19c shows the normalized the Raman spectra for monolayer MoS2 onto CA·M film and onto SiO2. The MoS2 A1g peak shows a slightly redshift. The underlayer CA·M film is able to n-dope the upper layer monolayer MoS2, likewise the doping effect observed when CA M monolayer is deposited on top of the MoS2 flake. The n-doping effect is also confirmed by the normalized PL spectroscopy, as shown in Fig. S19d. Compared with data from MoS2 transferred to bare SiO2, the PL spectra of monolayer MoS2 on CA·M film reveals a marked redshift at A excitation peak and a wider peak width.  = ! 2 | | " ⁄ The capacitance remains constant within the analyzed frequency range. After the deposition of 5-10 nm CA·M molecular film, the capacitance of the dielectric layer is slightly decreased to 16.1 nF/cm 2 , which is a neglectable change. As the thickness of the CA·M thin film is about 5 nm (10 layers), the capacitance of the 270 nm SiO2 substrate has almost no change with and without the CA·M film attachment. The HER polarization curves for monolayer MoS2 onto SiO2 devices under VGS varying from 0 to 5 V are shown in Fig. S23a. The curves shift towards smaller potentials while increasing VGS, indicating that the sample catalytic property is largely improved upon gate bias. A relatively low overpotential (η) of 255 mV for VGS = 5V is achieved with respect to the scenario where VGS is null (η = 347 mV). Fig. S23b shows the corresponding Tafel slopes that decrease from 110 mV dec -1 to 60 mV dec -1 upon VGS application. Fig. S23c and S23d display the field-tuned polarization curves and the related Tafel slopes for CA·M / MoS2 hybrid surface. The catalytic property of the hybrid device is largely boosted when VGS increase from 0 V to 4.3 V. At VGS = 4.3V, η reaches as low as 14 mV, which is comparable with that of Pt. Because of the rapid discharge process on the CA·M / MoS2 hybrid surface, it achieves smaller Tafel slopes (29 mV dec -1 ) than Pt for VGS > 4V. All η and Tafel slopes are summarized in Table S2 and S3.

VGS(V)
Tafel slope (mV dec To have a better understanding of the n-doping mechanism of MoS2 by CA·M layer attachment, ultraviolet photoelectron spectroscopy (UPS) and X-ray photoelectron spectroscopy (XPS) measurements are performed to study the interfacial electronic structure at the monolayer CVD MoS2 / CA·M heterostructure interface. The comparison of UPS spectra before and after CA·M film (about 2 nm thick) deposition is shown in Fig. S24a. By linearly extrapolating the low kinetic energy onset, the work function of monolayer MoS2 sharply decreases from 4.0 eV to 1.6 eV (Fig. S24b), originating from substantial interfacial electron transfer from CA·M film to MoS2. At the low energy region, the work function reduction is accompanied by a downward band-bending of the valence band of MoS2 toward higher binding energy. In the low binding energy region of UPS spectra ( Fig. S24c and Fig. S24d), the binding energy at valence band maximum upshift by 0.33 eV (from 1.85 eV to 2.18 eV) away from Fermi level (EF). Fig. S24e and Fig. S24f show the Mo 3d and S 2p core levels evolved with the increased CA·M coverage. At the initial stage, with sub-monolayer coverage To assess the behaviors of molecular film in the charge transfer processes, core level spectra (XPS) of monolayer CVD MoS2 with sub-monolayer molecular coverage are analyzed (Fig.  S25). Note, the N 1s level is related to CA and M molecules. Compared with molecular film deposition on SiO2, the N 1s core-level shift to higher energy for all three interfaces, which dominantly represents the process of losing electrons from CA and M molecules owing to the interfacial charge transfer 11 . The MoS2 / CA·M interface has the most obvious shift of N 1s by 0.7 eV, compared with 0.4 eV for MoS2 / M interface and 0.35 eV for MoS2 / CA interface. In case of CA, the N 1s XPS spectra could be deconvoluted into two peaks, as seen in Fig. S25c. The peak around 400.5 eV is assigned to the nitrogen of CA tautomer in the form of triketone, whereas the peak around 398.7 eV relates to the nitrogen of CA tautomer in form of triacid. The charge transfer is more obvious when the CA molecule is adsorbed on MoS2 surface in triacid form. Therefore, for all three systems, the N 1s core-level spectra consistently confirm the occurrence of electron transfer from molecule layer to MoS2.