Introduction

Two-dimensional transition metal dichalcogenides (2D TMDCs) are promising next-generation semiconductor materials, owing to their superior electronic properties that originate from their low dimensionality and high symmetry1,2,3. For practical 2D TMDC-based electronic devices, e.g., thin-film transistors (TFTs)4,5, chemical sensors6,7,8, photodetectors9,10, and photovoltaic cells11,12,13,14, one of the key challenges is controlling the doping level, and thus the Fermi level (EF) position, of the 2D TMDC semiconductors without breaking their low dimensionality and high symmetry, so that the desired electrical properties are retained.

Various methods have been proposed for tuning EF within the band gap of TMDCs, including substitution with dopant atoms15, stacking with other 2D materials16, exposure to gases6,7,8,17, adsorption of alkali metals18,19, and adsorption of organic molecules20,21,22,23,24,25,26. In accordance with this approach, deposition of molecular electron acceptors or donors on TMDCs has been suggested as a very effective strategy for controlling the EF position via doping, and accordingly high-performance TFTs were realized22,24,26. However, the extent to which the changes in electrical TFT characteristics were indeed due to electron transfer between the dopant molecules and the monolayer TMDC (ML-TMDC) was not unambiguously demonstrated, and mostly changes of photoluminescence intensity of excitons versus trions were interpreted as signature of charge transfer (CT). A fundamental understanding of the CT mechanisms occurring at the interface between dopant molecules and ML-TMDCs as well as direct evidence of EF position tuning, e.g., as typically achieved by photoelectron spectroscopy, are lacking, mostly because conductive substrates for the ML-TMDC were employed. For instance, Katoch et al. 27 investigated n-type doping of WS2 by potassium atom deposition. However, degenerately doped, highly conductive TiO2 was used as substrate, from which the ML-TMDC was separated only via an ultrathin boron nitride layer, and the role of the underlying metal oxide substrate remains elusive. Regardless, renormalization of the WS2 band gap and spin-orbit splitting due to many-body effects was demonstrated.

Understanding the electronic properties of interfaces in devices such as those present in 2D TMDC-based devices is crucial for performance optimization. For instance, the TMDC/insulator interface determines the electronic properties of the active layer when a bias is applied to the gate in a TFT, while TMDC/metal interfaces serve as the electrical contacts between the semiconductor and the drain or source electrodes. In this context, it was demonstrated that the substrate dielectric properties significantly affect the electronic structure of ML-TMDC monolayers, such as band gap renormalization27,28, and the TMDC–substrate interaction influences the exciton and trion binding energies (BEs)28,29. Also of importance for devices is minimizing the Schottky barrier height between a metal contact and the TMDC, as pronounced Fermi level pinning seems to limit the range of tunability, as revealed via thermionic emission studies by Kim et al. 30. Here, the electronic coupling and chemical interaction between the TMDC and metals play vital roles. These findings show that the supporting substrate should be fully considered in studies aiming at unraveling molecular dopant/TMDC monolayer interactions, as also CT processes may be affected. Therefore, investigating molecular dopant/ML-TMDC interfaces on different types of substrates is crucial for obtaining a comprehensive understanding of the CT processes and the energy levels, which is needed for reliably tuning ML-TMDC electronic properties for device applications.

In this study, we investigate molecular acceptor/ML-TMDC/substrate systems using insulating (sapphire31), semi-metallic (highly oriented pyrolytic graphite (HOPG)), and metallic (Au) substrates. With angle-resolved ultraviolet photoelectron spectroscopy (ARUPS) and X-ray photoelectron spectroscopy (XPS), we monitor the changes of EF, the sample work function (Φ), and core levels induced by acceptor deposition. According to our observations, we identify three distinct CT mechanisms between the acceptor molecules (1,3,4,5,7,8-hexafluoro-tetracyano-naphthoquinodimethane (F6TCNNQ)) and ML-TMDC (MoS2), depending on substrate type. The mechanisms for electron transfer from the (i) MoS2 gap states to the acceptors (sapphire), (ii) substrate to acceptors with an electric field across the ML-MoS2 (HOPG), (iii) combined metal/ML-MoS2 system to acceptors. The difference between (ii) and (iii) is rooted in the different electronic coupling strength between the ML-MoS2 and the substrates. Consequently, factual doping of MoS2 proceeds only on the insulating substrates.

Results

Impact of CT on work function and valence electronic levels

To investigate how CT processes between a TMDC monolayer and the molecular acceptor F6TCNNQ, which has also been employed as p-type dopant for organic semiconductors, could be influenced by the choice of substrate, ARUPS measurements were performed for MoS2/sapphire, MoS2/HOPG, and MoS2/Au with stepwise deposition of F6TCNNQ. We note that the MoS2 monolayers used in this study consist of azimuthally disordered flakes. However, owing to the specific features of the TMDC band structure—related to the high symmetry of the TMDCs—ARUPS spectra feature a high photoemission intensity only along the Γ-K and Γ-M directions of the single-crystal Brillouin zone (BZ), thus the spectra correspond to a linear superposition of the electronic bands along the aforementioned high symmetry directions32,33 We note that, while it was shown that the electrical conductivity of sapphire can increase to about 10−9 S/m by creating a high density of defects after annealing to 600 °C and proton bombardment31, our experimental conditions ensure the insulating character of sapphire. Particularly, for samples with insufficient MoS2 monolayer coverage and thus no percolation pathways for CT, we were not able to record any photoemission spectrum due to sample charging (also after sample annealing).

The ARUPS spectra measured in normal emission, as shown in Fig. 1, correspond to the electronic states at the Γ point of the BZ. These Γ point spectra show the highest intensity in the entire BZ, with the top valence band (VB) as an intense and sharp peak, allowing any spectral energy shifts to be accurately resolved. In addition, all spectra in this study exhibit a rigid energy shift upon molecular acceptor deposition, without band distortions (within instrument resolution), indicating that any momentum line cut (1D spectra along energy) provides the same information on energy shifts (exemplified in Supplementary Fig. 1 for the Γ and K points). For clarity, we stress that the global VB maximum is not at the Γ point (VBMΓ), which is only a local maximum point in the band structure, but at the K point (VBMK). Thus, the K point should be considered for determining the MoS2 ionization energy (IE), with IE = Φ + VBMK (with all valence energy levels referenced to EF)34. VBMΓ, determined by a linear extrapolation of the emission onset toward the background from the spectra in Fig. 1, is at 1.80 eV BE for the pristine MoS2 films on sapphire and HOPG, and at 1.30 eV BE on Au. In analogy, VBMK was determined to be at 1.78 eV BE on sapphire, 1.70 eV BE on HOPG, and 1.28 eV BE on Au substrates (see Fig. 2). There is a very small substrate-dependent variation (not more than 0.08 eV) in the energy difference between VBMΓ and VBMK, which might be related to the presence of defect states of MoS2 monolayer and/or substrate-specific interactions. Moreover, the MoS2 monolayer on HOPG might be more homogeneous, leading to the slightly wider energy difference between VBMΓ and VBMK.

Fig. 1
figure 1

Angle resolved ultraviolet photoelectron spectra of F6TCNNQ/MoS2/substrate (sapphire, highly oriented pyrolytic graphite (HOPG), and Au) heterostructures. Secondary electron cutoff and valence band (VB) spectra of MoS2/sapphire (a, b), MoS2/HOPG (c, d), and MoS2/Au (e, f) for different F6TCNNQ coverages. A work function increase of ~0.6 eV is observed for all samples after acceptor adsorption. All VB spectra were recorded at the Γ point. The unit of numbers inside the plots is electron volts (eV), which indicate initial sample work function and its change upon F6TCNNQ deposition (a, c, e), and initial valence band onset and its change upon F6TCNNQ deposition (b, d, f)

Full size image
Fig. 2
figure 2

Energy distribution curves (EDCs) near Fermi level (EF). EDCs obtained by integration of the corresponding angle-resolved ultraviolet photoelectron spectra around the K point in a narrow interval (1.08–1.36 Å−1), of a MoS2/sapphire, b MoS2/highly oriented pyrolytic graphite (HOPG), and c MoS2/Au with (violet) and without (green) adsorbed F6TCNNQ. The numbers inside the plots correspond to the MoS2 valence band onset, given in electron volts (eV)

Full size image

As seen from Fig. 1a, c, and e, the SECO spectra provide sample Φ values of 4.60 eV for MoS2/sapphire, 4.60 eV for MoS2/HOPG, and 4.95 eV for MoS2/Au. With increasing coverage of F6TCNNQ, Φ changes (ΔΦ) are observed, and ΔΦ saturates at approximately +0.6 eV at a nominal acceptor layer mass-thickness of 10 Å for all the samples. Apparently, this large ΔΦ is related to electron transfer from ML-MoS2/substrate to the lowest unoccupied molecular orbital (LUMO) level of F6TCNNQ, with a possible contribution stemming from orbital polarization. Concomitantly, as shown in Fig. 1b, d, and f, VBMΓ shifts toward lower BE by 0.43 eV for MoS2/sapphire and 0.24 eV for MoS2/HOPG, while no shift is observed for MoS2/Au. Although ΔΦ values are the same within 0.05 eV for all the samples, the strongly substrate-dependent VBMΓ shifts already indicate fundamentally different CT mechanisms.

Gap states of the MoS2 monolayer and F6TCNNQ anions

To obtain detailed insight into the different CT mechanisms, high-resolution VB spectra of the individual ML-MoS2/substrate systems around the K point were measured prior and after deposition of 5 Å F6TCNNQ, as shown in Fig. 2. In Fig. 2a, a clear photoemission feature is observed just below EF, i.e., within the technically empty band gap of ML-MoS2. We suggest that this feature stems from gap states related to the presence of sulfur vacancies, as predicted by previous studies35,36,37,37, and found experimentally for bulk and liquid exfoliated MoS238. However, investigating the exact origin of these gap state would necessitate extensive additional experiments, which is beyond the scope of the present study.

Most notably, the gap states observed for MoS2/sapphire are not present for MoS2/HOPG as well as MoS2/Au (Fig. 2b, c), although the MoS2 monolayers transferred to HOPG and Au were synthesized under the same conditions compared to those on sapphire, and underwent the same cleaning procedure in UHV. Thus, the likely reason for the absence of gap states is that the (semi-) metallic substrates (HOPG and Au) withdrew the electrons from the gap states of the monolayer, resulting in now unfilled gap states not observable in an ARUPS experiment.

With nominally 5 Å of F6TCNNQ adsorbed on ML-MoS2/sapphire (Fig. 2a), two features related to the molecules (marked “A” at 0.3 eV BE and “B” at 0.8 eV BE) are observed. As shown in Fig. 2b, these features are also clearly detected for F6TCNNQ on MoS2/HOPG, with essentially the same A/B intensity ratio. According to previous reports, these states can be assigned to anionic F6TCNNQ (F6TCNNQ−1), resulting from electron transfer to the molecules39,40. This assignment is further confirmed by density functional theory (DFT) calculations as provided in the Supplementary Note 1. Specifically, these calculations show that upon adding one electron to the LUMO level of an isolated molecule, strong orbital rearrangements occur, which give rise to two gap states that are a combination of (i) the former highest occupied molecular orbital (HOMO) level and the singly filled former LUMO level (peak A), and (ii) the former HOMO and HOMO-1 levels (peak B). In addition, comparison of the spectra obtained with and without 5 Å F6TCNNQ on MoS2/sapphire reveals that the peak positions of the gap states and peak A are too similar to be distinguished. If no CT occurs, the intensity of the gap states (in the chosen momentum interval around K) should be ca. 65% of that observed for the bare ML-MoS2 because of the attenuation due to the F6TCNNQ overlayer, as deduced from the attenuation of the Mo 4d valence feature. Thus, peak A would be the sum of the gap states and peak A from F6TCNNQ−1, if the gap states still existed. However, the relative intensities of peaks A and B for F6TCNNQ−1 on MoS2/sapphire and MoS2/HOPG are almost identical, though no gap states existed for the pristine MoS2/HOPG sample. From this, we infer that the gap states are quenched, i.e., emptied, as a result of the electron transfer from MoS2 to F6TCNNQ, when using sapphire as supporting substrate. As for the pristine MoS2/HOPG no gap states are detected, the origin of the electrons transferred to the molecules must therefore be different. In contrast, much broader molecule-induced features are observed for F6TCNNQ/MoS2/Au (Fig. 2c), which yet calls for another CT mechanism as compared to when using sapphire and HOPG substrates. Consequently, three different CT mechanisms are at play for the three studied cases, which are discussed in more detail in the following.

Impact of CT on the individual heterostructure components

To assess the involvement of the individual components, i.e., ML-MoS2 versus substrate, in the CT processes, core level spectra of the three ML-MoS2/substrate systems with increasing F6TCNNQ coverage are analyzed, and displayed in Fig. 3. Note, the F 1s level is related only to F6TCNNQ molecules and the Mo 3d level to the MoS2 monolayer. The Al 2p, C 1s, and Au 4f levels are representative of the sapphire, HOPG, and Au, respectively.

Fig. 3
figure 3

X-ray photoelectron spectroscopy (XPS) core level spectra. XPS spectra for MoS2/sapphire (a–c), MoS2/highly oriented pyrolytic graphite (HOPG) (d–f), and MoS2/Au (g–i) heterostructures with different F6TCNNQ coverage, given in Angstrom (Å) in the right handside plots. The numbers inside the plots correspond to the core level shifts, given in electron volts (eV)

Full size image

First, we attend to the F 1s core level that is indicative of the oxidation state of F6TCNNQ. As apparent from Fig. 3a, d, and g, the F 1s core levels for 5 Å F6TCNNQ on all substrates are observed at 686.7 eV (±0.05 eV), which dominantly represents F6TCNNQ−1 owing to the interfacial CT. With increasing F6TCNNQ coverage, F 1s levels shift by 0.3–0.5 eV toward higher BE, due to an increasing contribution from neutral F6TCNNQ molecules that become predominant in the multilayer film. Therefore, for all three systems, the F 1s core level spectra consistently confirm the occurrence of electron transfer from the MoS2/substrate system to the acceptor.

Before discussing the details of the Mo 3d and Al 2p core levels, the energy reference in XPS is briefly reiterated. Because of the electrically insulating properties of sapphire, the position of EF within the band gap of the (grounded) MoS2 monolayer serves as the energy reference for the XPS measurements. When F6TCNNQ is adsorbed on top of the MoS2 monolayer, electrons are transferred from MoS2 to the F6TCNNQ molecules, which results in a change in EF (ΔEF) in the MoS2 monolayer (Fig. 1b). Because the ML-MoS2 is electrically connected to the detector, EF is re-aligned to 0 eV BE again, resulting in a rigid shift (same value of ΔEF) for all ML-MoS2/sapphire spectra. For this reason, the Mo 3d and Al 2p core level shifts observed for MoS2/sapphire are due to a change in EF within the MoS2 monolayer, and not due to any chemical shifts. For MoS2/sapphire (Fig. 3b), a shift of Mo 3d by up to 0.3 eV toward lower BE is observed, depending on the F6TCNNQ coverage. As explained above, this is due to the Fermi level shift, as caused by p-doping of the MoS2 monolayer. The concomitant shift of the Al 2p core level (sapphire) by the same amount further supports the interpretation of the EF shift in ML-MoS2 due to doping by the acceptors.

In contrast, for ML-MoS2 on HOPG and Au, the energy reference (EF) is that of the highly conductive (semi-) metallic substrates. In Fig. 3f, the C 1s core level at 284.4 eV can be assigned to the sp2 carbons of HOPG. Regardless of the F6TCNNQ coverage, the BE of the HOPG level remains constant, and only decreasing intensity is observed because of the signal attenuation by the F6TCNNQ overlayer. The same is observed for the Au 4f core level at 84 eV BE in Fig. 3i. Both substrate core level spectra series indicate that the energy reference does not depend on the F6TCNNQ coverage when conductive substrates are employed.

Remarkably, for ML-MoS2/HOPG, the Mo 3d core level shifts by 0.25 eV to lower BE upon acceptor deposition, the same as did VBMΓ (compare Figs. 1d and 3e). These rigid shifts support the notion of p-doping of MoS2 monolayer/HOPG. In contrast, for MoS2/Au, the Mo 3d core levels exhibit no energy shift, as shown in Fig. 3h. This evidences that the interfacial CT did not affect the electronic properties of the MoS2 monolayer, in full agreement with the constant VBMΓ in Fig. 1f. For the three investigated systems, quantitatively different Mo 3d core level shifts are observed, all being consistent with the respective VBMΓ shifts in Fig. 1. Since the Mo 3d and S 2p core levels of the MoS2 shift in parallel (see Supplementary Fig. 2), notable site-specific chemical reactions between the molecular dopant and the monolayer can be ruled out. These findings, in combination with the same ΔΦ observed for all three substrate types, further imply three fundamentally different CT mechanisms.

Identification of three CT mechanisms

All experimental findings significantly differ depending on the employed substrate, invoking different CT mechanisms, which seem to be strongly influenced by the interaction between the MoS2 monolayer and the supporting substrate. Because the heterostructures comprise strong molecular acceptors, a MoS2 monolayer, and a substrate, two different CT pathways should be considered, namely: (i) a direct CT occurring between the acceptors and the ML-MoS2, and (ii) an indirect CT between acceptors and the substrate, possibly involving the MoS2 monolayer.

We start by testing the direct CT scenario for the three substrates according to the measured energy levels. Considering the reported electron affinity (EA) of F6TCNNQ (5.6 eV)41 and the measured IE of ML-MoS2, with IE = Φ + VBMK = 4.6 + 1.78 = 6.38 eV for sapphire, 6.30 eV for HOPG, and 6.23 eV for Au, the possibility of CT from the MoS2 VB to the F6TCNNQ LUMO level can be ruled out. Another possible CT pathway is from thermally excited electrons in the MoS2 global conduction band minimum (CBMK) to F6TCNNQ. The EA of ML-MoS2/substrate is obtained from the determined IE values (see above) and the reported substrate-dependent band gaps (EA = IE − band gap), yielding EA values of 4.27 eV (on sapphire), 4.19 eV (on HOPG), and 4.33 eV (on Au). The energy difference between the EA of ML-MoS2 and the EA of F6TCNNQ supports transfer of thermally excited electrons. In this case, the electron density (ne) in the conduction band is an important factor. The ne of ML-MoS2 is estimated from42

$${n}_{\mathrm{e}} = \frac{{{g}_{\mathrm{v}}{m}_{\mathrm{e}}^ \ast {k}_{\mathrm{B}}{T}}}{{2{\mathrm{\pi }}\hbar ^2}}{\mathrm{ln}}\left( {1 + {\mathrm{e}}^{({E}_{\mathrm{F}} - {E}_{\mathrm{C}})/{k}_{\mathrm{B}}{T}}} \right),$$
(1)

where gv, me*, \(\hbar\), kB, T, and EC represent the valley degeneracy factors, the band edge effective mass of electrons, the reduced Planck constant, the Boltzmann constant, the absolute temperature, and the energy position of the CBMK, respectively. Using gv = 6, me* = 0.37m0, and EC = 0.33 eV, ne is estimated to be ~107 cm−2 (refs. 43,44).

For examining the impact of ne on eventual CT, we also estimate the density ρ (e∙cm−2) of charge transferred into the LUMO levels of F6TCNNQ molecules adsorbed on ML-MoS2 by using the measured ΔΦ and VBM shift (ΔVBM), according to the Helmholtz equation45,46:

$$\rho = \frac{{{\mathrm{eD}} \times \varepsilon _{{\mathrm{eff}}}\varepsilon _0}}{{ed_{{\mathrm{eff}}}}},$$
(2)

where \({\mathrm{eD}},\;{\varepsilon} _0,\;{\varepsilon} _{{\mathrm{eff}}},\;d_{{\mathrm{eff}}}\), and e are the interface dipole, the vacuum permittivity, the effective dielectric constant, the effective dipole distance, and the elementary charge, respectively. The eD value can be determined from eD = ΔΦ – ΔVBM, with ΔVBM being the change in VBM binding energy before and after acceptor adsorption. From our data, eD is found to amount to 0.17 eV (sapphire), 0.36 eV (HOPG), and 0.55 eV (Au)45,47,48. With eD thus being in the range of 0.17–0.55 eV and using theoretical values of 4.70 for \({\varepsilon _{{\mathrm{eff}}}}\) (corresponding to the average of \({\varepsilon _{{{{\rm{MoS}}_2}}\;{\mathrm{ML}}}} = 6.4\) and \({\varepsilon} _{{{{\rm{F}}_{\rm{6TCNNQ}}}}} = 3\)) and 3 Å for deff45,49, ρ is found to be ~1013 cm−2, providing an estimate of the density of the electrons transferred to an F6TCNNQ monolayer for all samples. Clearly, the estimated density of transferred charge ρ to the F6TCNNQ exceeds by far (ca. six orders of magnitude) the estimated available electron density in the conduction band. This allows ruling out direct CT from the conduction band as the predominant process in all three cases.

Furthermore, the estimated ne (~107 cm−2) is small compared to the reported carrier density in MoS2 monolayers (~1013 cm−2), which necessitates the existence of gap states between the CBM and VBM42,50. Gap states in 2D TMDCs fabricated via chemical vapor deposition (CVD) thus appear as critical factors for the mechanisms of interfacial CT and energy level alignment. Numerous reports have attributed gap states in ML-MoS2 on insulating substrates, e.g., SiO2 and sapphire, to S-vacancies owing to their lower formation energy as compared to Mo-vacancies35,37,37,51,52. Accordingly, gap states have been invoked as key for the energy level alignment at metal/ML-MoS2 interfaces51,53.

In the present work, for MoS2/sapphire, a clear gap state is observed near EF with an IE of 4.6 eV, which is energetically favorable for electron transfer to F6TCNNQ, and judged from the spectral weight in Fig. 2a it has a sufficient density. This gap state, which is in the present study unique to MoS2/sapphire, allows direct CT, i.e., from the MoS2 gap state to the LUMO level of F6TCNNQ, to occur, as confirmed by the quenching of this state after acceptor adsorption (see Fig. 2a and its discussion above). This is fully in line with the proposition of Wang et al.22.

In contrast, for the ML-MoS2 on HOPG and Au substrates (Fig. 2b, c), no gap states are observed, which therefore rules out the possibility of direct CT for these two cases. To explain the origin of the transferred charge, the supporting substrate should be considered as an electron reservoir, since ΦHOPG (4.7 eV) and ΦAu (5.3 eV) are both lower than the F6TCNNQ EA (5.6 eV). Therefore, electron transfer from HOPG and Au to F6TCNNQ through the MoS2 monolayer is plausible. For ML-MoS2/HOPG, electrons accumulate in the F6TCNNQ layer and their image charges in HOPG, which leads to a high electric field between HOPG and F6TCNNQ. The concomitant electrostatic potential drop across the ML-MoS2, sandwiched between acceptor layer and substrate, shifts the energy levels of the MoS2, which appears to become p-type doped via this indirect CT. A similar indirect CT process has been suggested for the C60F48/WSe2/HOPG interface, and ultra-thin organic heterojunctions on metal substrates26,54.

Finally, for ML-MoS2/Au, the absence of energy level shifts (MoS2 valence and core levels) upon acceptor deposition is caused by a strong electronic coupling of the TMDC and Au, as orbital hybridization between ML-MoS2 and Au has been previously reported55. In this case, the electrons are transferred from the hybridized ML-MoS2/Au system, which acts as an electron reservoir, to F6TCNNQ.

Discussion

Figure 4 summarizes simplified schematics of the three proposed CT mechanisms and the corresponding energy level diagrams for the investigated systems. On general grounds, owing to the low dimensionality of 2D TMDCs, the CT between 2D TMDCs and dopant molecules (acceptors and donors) is strongly influenced by the properties of the surrounding medium, i.e., the supporting substrate. Depending on the insulating, semi-metallic, and metallic properties of the substrate, as well as the coupling strength between the 2D TMDC and the substrate, three different cases can be classified:

Fig. 4
figure 4

Schematics of the mechanisms of charge transfer (CT) between MoS2/substrate and F6TCNNQ. a Direct CT (MoS2/sapphire, 2D transition metal dichalcogenides (TMDC)/insulator), b indirect CT (MoS2/highly oriented pyrolytic graphite (HOPG), 2D TMDC/weakly interacting conductor), and c orbital hybridization with CT (MoS2/Au, 2D TMDC/strongly interacting conductor). Corresponding energy level diagrams for d F6TCNNQ/ML-MoS2/sapphire, e F6TCNNQ/ML-MoS2/HOPG, and f F6TCNNQ/ML-MoS2/Au. The solid and dashed lines indicate the levels before and after F6TCNNQ adsorption, respectively. The blue and red lines denote the occupied and unoccupied states, respectively. All the values are given in electron volts (eV)

Full size image

(1) Insulating substrate: Direct CT occurs between the 2D TMDC valence levels (if energetically allowed, not the case for the materials investigated here) or gap states (observed in this work) to the LUMO level of the molecular acceptor, resulting in p-type doping of the 2D TMDC, as illustrated in Fig. 4a and d. Note that in the present case, the ML-MoS2 starts out highly n-type (due to the gap states) and doping by F6TCNNQ depletes the gap states, rendering the monolayer less n-type by 0.4 eV. For molecular donors and n-type doping, an analogous mechanism can proceed.

(2) Weakly coupled (semi-) metallic substrate: Indirect CT occurs from the substrate to the dopant molecule, through the MoS2 monolayer as shown in Fig. 4b, e. The charges accumulated in the molecular layer and the substrate lead to a high electric field across the 2D TMDC, resulting in an ostensible doping of the TMDC layer.

(3) Strongly coupled metallic substrate: CT occurs from the 2D TMDC-metal hybridized frontier level to the LUMO level of the molecular acceptor, as shown in Fig. 4c and f. Because of the strong coupling, the energy levels of the 2D TMDC are pinned to those of the metal substrate and no doping of the 2D TMDC layer proceeds despite substantial CT.

We emphasize that in all three cases above, the sample Φ increased upon F6TCNNQ deposition by virtually the same amount. Consequently, a change of the TMDC/substrate Φ induced by molecular acceptors/donors (or other dopants deposited atop the TMDC) cannot be used as a reliable indicator for doping of the semiconductor. Likewise, a shift of EF, as observed here for the HOPG substrate, can result from an electric field drop across the ML-TMDC, induced by the charges in the molecular layer and the substrate; this situation is reminiscent of a parallel plate capacitor where the TMDC is the dielectric between the two plates. Only if the substrate can be ruled out as source of charges for the dopants, factual doping of the TMDC occurs, which is crucially important when aiming to study the fundamental behavior of ML-TMDCs as a function of carrier density. Moreover, with the knowledge of the three different CT mechanisms and how they can be implemented, a wide variety of possibilities to fine-tune the electronic properties of ML-TMDCs in device heterostructures becomes available.

Methods

Sample preparation

MoS2 monolayers were grown on sapphire via CVD and transferred onto an HOPG or Au substrate. The monolayers consist of interconnected individual flakes, thus featuring high overall surface coverage and percolation paths for charge carriers, as exemplarily shown by optical micrographs in Supplementary Fig. 3. Prior to the ARUPS and XPS measurements, all MoS2 monolayer samples were annealed overnight at 300–350 °C in situ in an ultrahigh vacuum (UHV) preparation chamber (10−9 mbar) to remove carbon contaminations and residual poly(methyl methacrylate) involved during the transfer process. F6TCNNQ (Novaled) was deposited in the preparation chamber by sublimation from a resistively heated quartz crucible, and the nominal mass-thickness was determined via quartz crystal microbalance measurements.

Photoemission measurements

ARUPS and XPS spectra were obtained using a hemispherical electron analyzer (SCIENTA DA30L) using He Іα (21.22 eV) and Mg Kα (1253.6 eV) as excitation sources, respectively. The energy resolution of ARUPS and XPS of 0.13 eV and 0.75 eV was determined by the width of the Fermi edge of a clean Au sample and the full-width at half-maximum (FWHM) of the Au 4f core level, respectively. For all the shown ARUPS spectra, the contribution from the He Ιβ satellite was removed. The work function (Φ) values were determined from the secondary electron cutoff (SECO) spectra, measured with −10 V sample bias. For the sapphire substrate, proper MoS2 monolayer electrical grounding was established by contacting all sample edges with metal clips (connected to the metal sample holder), to prevent charging during the photoemission measurement. For HOPG and Au substrates, the MoS2 monolayer was connected to the ground through the conductive substrates.