Dissolution–precipitation growth of doped monolayer molybdenum disulfide through double-faced precursor supply

Two-dimensional (2D) materials have received much attention in recent years^1,2 because of their unique physical properties.^3,4 Among 2D materials, transition metal dichalcogenides (TMDCs) exhibit a wide range of properties and applications, including optical,⁵ magnetic,⁶ electronic,^7,8 and energy conversion.^9–12 As one of the most studied TMDCs, molybdenum disulfide (MoS₂) has been used to fabricate integrated circuits, memristors, photodetectors, and sensors.^13–16 However, the performances of MoS₂ in the above fields still need to be improved. For example, MoS₂ exhibits n-type transport behavior, which is not enough to construct complex integrated circuits where both n-type and p-type semiconductors are needed.¹⁷ In addition, the low room-temperature photoluminescence (PL) quantum yield of MoS₂ limits its performance in optoelectronics.¹⁸ Motivated by these limitations, a lot of research studies have been conducted to develop approaches for modifying the physical properties of MoS₂ and endowing it with new functionalities.^19–23 Compared with methods such as surface adsorption doping and substrate induced doping, substitutional doping can change the physical properties of MoS₂ in a more stable and effective way by directly introducing foreign metal atoms into the host lattice.²⁴ Previous reports have used this doping strategy to modulate the property and performance of MoS₂ in many fields. For example, Zhang et al. have used V dopants to enhance the S-2p_z orbital coupling between adjacent layers of bilayer MoS₂, giving rise to an ambipolar transport behavior in the monolayer region and p-type transport behavior in the bilayer region.²⁵ Zhou et al. showed that doped metal atoms can activate the inert sulfur sites and enhance the hydrogen evolution reaction activity of MoS₂ in electrocatalysis.²⁶ Usually, mixed powders of metal precursors composed of the Mo source and dopant source are carried by gas flow into the furnace center to grow doped MoS₂. During the growth and doping process, the two metal precursors sublimate unevenly along the diffusion path, which typically results in poor controllability and repeatability. In addition, for some transition metal atoms (such as Co and Fe), it is difficult to apply the above approach to incorporate them into the MoS₂ lattice because of the low vapor pressure of their powdered dopant precursors during growth. To solve these problems, Zhang et al. have developed a liquid phase precursor assisted method by replacing powdered metal precursors with aqueous solution of metal precursors and improved the controllability of growth.²⁷ Along this direction, if one can separate the diffusion paths of different precursors, it may lead to precise control of dopants and, consequently, improve the substitutional doping of 2D materials.

Recently, we invented a dissolution–precipitation (DP) method to grow TMDCs where the chalcogen source and metal source are independently supplied through the vapor phase and underlying glass substrates.²⁸ Based on this new growth method, here we developed a double-faced precursor supply DP method to control the doping of metal atoms into the MoS₂ lattice, in which the Mo source and dopant precursor were supplied from the bottom and upper surface of the glass substrate, respectively. Due to the separated diffusion paths of the Mo source and dopant precursor, the doping concentrations can be facilely tuned by adjusting the amount of the precursor. Taking this strategy, we successfully grew V-doped MoS₂ (V–MoS₂) with tunable doping concentrations, which exhibited different electrical transport properties. Moreover, this new doping method is universal in growing some other metal doped MoS₂, including Fe doped MoS₂ (Fe–MoS₂), Cr doped MoS₂ (Cr–MoS₂), and Re doped MoS₂ (Re–MoS₂).

By providing the Mo source underneath the glass growth substrate while the S source from the gas phase, the DP method restricts the growth of MoS₂ to the top surface of the glass. Our previous work has proved that the concentration of the diffused Mo source in the DP method can guarantee the formation of monolayer MoS₂ with a clean surface.²⁸ Thus, the DP method provides a reliable way to control the Mo source. Here, as a modification of the DP method, we further developed the double-faced precursor supply DP method for the growth of doped MoS₂. A schematic illustration of the double-faced precursor supply DP method is shown in Fig. 1(a). We first embedded the Mo source between two pieces of glasses with a thin top glass (0.15 mm) and thick bottom glass (2 mm), forming a sandwiched structure. In this way, the whole structure will serve as the growth substrate and Mo source together. Next, the aqueous solution containing transition metal dopant precursors (NaVO₃, Na₂CrO₄, NaReO₄, FeCl₃, etc.) was spin-coated on the surface of the top glass and dried in an oven to scatter the dopant precursor on the substrate. Then, the whole sandwiched structure was placed at the center of the horizontal furnace where the temperature was increased to 730 °C under Ar, and sulfur powder was put upstream. During the growth, the molten Mo source diffused across the upper glass to its surface and reacted with sulfur to form monolayer MoS₂. At the meantime, the scattered dopant precursor on the glass surface served as the local source of doped atoms to initiate doping. Therefore, different from the traditional method to dope MoS₂, which is based on powdered metal precursors [Fig. 1(b)], the double-faced precursor supply DP method provides the Mo source and dopant precursor from two separated paths [one from the bottom, while the other from the top, Fig. 1(c)], leading to better controllability of the doping process. Moreover, for those dopant precursors scattered on top of the glass, they supply doped metal atoms from sites near the growing MoS₂, making the control of dopant concentration easy. By tuning the amount of dopant precursor spin-coated on the surface of the glass, we can change the concentration of this precursor during growth and, consequently, control the concentrations of doped atoms in MoS₂.

Next, we characterized the V-doped MoS₂ grown by the above proposed double-faced precursor supply DP method. Figure 2(a) is a representative optical microscope image of these V-doped flakes, with sizes of ∼10 μm. The atomic force microscopy (AFM) topography indicates a height of ∼0.91 nm for the V–MoS₂ flake [Fig. 2(b)], confirming its monolayer nature.²⁹ The high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) characterization of the as-grown V–MoS₂ [Fig. 2(c)] proves that the doping concentrations of V in the flakes can reach up to 6 at. %. The detailed HAADF-STEM image shows that substitutional V dopants replace Mo atoms at the cation sites, as highlighted by red circles in Fig. 2(d). The intensity profile extracted from a magnified area further reveals that the intensity ratio between V and Mo is around 0.4, consistent with the square ratio of atomic number Z [V (Z = 23), Mo (Z = 42)] and the corresponding simulated results [Figs. 2(e) and 2(f)]. We did not observe a Na atom in the MoS₂ lattice via HAADF-STEM characterization. We speculate that Na is a main group element with only one valence electron and has a radius much smaller than those transitional metals such as Mo, which makes it hard to be incorporated stably into the lattice of MoS₂. X-ray photoelectron spectroscopy (XPS) measurements show that four peaks belong to the V 2p signal [Fig. 2(g)], where the peaks at 520.4 and 512.8 eV are assigned to V⁴⁺ 2p_1/2 and V⁴⁺ 2p_3/2, respectively. These two peaks indicate the existence of V atoms in the hexagonal lattice of MoS₂,^30,31 consistent with STEM results. The signals of V⁵⁺ 2p_1/2 and V⁵⁺ 2p_3/2 at binding energies of 523.9 and 516.5 eV can be attributed to the residues of the doped V precursor (NaVO₃) on the substrate.³² The downward shift of the binding energy of both Mo 3d and S 2p by around 1.5 eV [Figs. 2(h) and 2(i)] indicates the shift of the Fermi level toward the valence band and the decrease in electron concentration of V–MoS₂, both of which agree well with the electron acceptor feature of V dopants.³³ These results together confirm the incorporation of V atoms in MoS₂.

Then, we investigated the advantages of the double-faced precursor supply in controlling doping concentrations of MoS₂. By tuning the amount of dopant precursor, we obtained two types of V-doped MoS₂: One is lightly doped V–MoS₂ (L-V–MoS₂) with a doping concentration of around 2 at. % and the other is heavily doped V–MoS₂ (H-V–MoS₂) with a doping concentration of around 6 at. %. Figure 3(a) shows the Raman spectra of pristine undoped MoS₂ and the two types of V-doped MoS₂. For the L-V–MoS₂, two main characteristic peaks E_2g and A_1g are still observed. Compared with undoped MoS₂, the Raman shift between E_2g and A_1g peaks of L-V–MoS₂ is 2 cm⁻¹ wider, i.e., 19 cm⁻¹ in L-V–MoS₂ vs 17 cm⁻¹ in undoped MoS₂. This can be attributed to the V dopant or defect induced strain effect.³⁴ As for the H-V–MoS₂, its Raman spectrum has many differences with undoped MoS₂. The Raman spectrum of H-V–MoS₂ shows two new characteristic peaks at 325 cm⁻¹ and 350 cm⁻¹ and a group of peaks in the range of 100 to 250 cm⁻¹, which can be attributed to the disorder effect caused by high-concentration V dopants.³⁵ In addition, doping also has effects on the PL of MoS₂ [Fig. 3(b)]. The A exciton peak of L-V–MoS₂ shows around 30 nm blueshift compared with undoped MoS₂. Because V has one less valence electron than Mo, doping of V atoms will decrease the electron concentration in the initial MoS₂ system. Therefore, the contribution of exciton to A peak will increase and the contribution of trion will decrease. This feature results in the blueshift of A peak in L-V–MoS₂.^36,37 In contrast, a profound PL quenching effect can be observed for high-concentration V-doped samples, which may originate from high-concentration impurity scattering centers caused by dopants in the MoS₂ lattices.³⁸ 

We further studied the modulation of the electrical transport properties of MoS₂ after doping. The above XPS and PL results have indicated that the doped V atoms can act as p-type dopants in MoS₂. We investigated the electronic properties of the undoped and lightly (2 at. %) and heavily (6 at. %) V-doped MoS₂ based field effect transistors (FETs). The undoped MoS₂ FET shows n-type transfer characteristics with a threshold voltage around −2.0 V [Fig. 3(c)]. The threshold voltage of the L-V–MoS₂ FET shifts to the positive side at around 12.6 V due to the decrease in electron concentration in it. In this work, we extracted V_th by the extrapolation in the linear region (the ELR method) based on the linear scale I_ds − V_g curves. Interestingly, the H-V–MoS₂ FET exhibits a p-type transfer characteristic [Figs. 3(c) and 3(d)], where the drain current increases, with the decrease in gate voltage. This result indicates that the drain current of the H-V–MoS₂ FET is dominated by holes. This change in the electrical transport of H-V–MoS₂ can be attributed to that V dopants served as electron acceptors and caused the Fermi level of MoS₂ to lower down to the valence band, resulting in a smaller hole Schottky barrier than electrons. We calculated the carrier concentration of undoped MoS₂, L-V–MoS₂, and H-V–MoS₂ based on the above measurements. The concentration of carriers can be calculated as $n=Gq×μ$⁠, where q is the electron charge, G is the channel conductance, and μ is the mobility of the FET device. As shown in Fig. 3(e), the lightly doped V atoms decreased the electron concentration of MoS₂ from 3.86 × 10¹² to 3.45 × 10¹² cm⁻². Meanwhile, the hole concentration in H-V–MoS₂ is around 1.20 × 10¹² cm⁻². Note that the electron and hole concentration of undoped MoS₂ and L-V–MoS₂ are calculated under a gate voltage of 80 V and the hole concentration of H-V–MoS₂ is calculated under a gate voltage of −80 V. The lightly doped V atoms also induced an intriguing electric property in MoS₂. We found that the drain current kept increasing when the gate voltage swept forward and backward from −50 to −40 V for 5 cycles [Fig. 3(f)]. This phenomenon shows the potential of using the L-V–MoS₂ sample to fabricate memtransistors.³⁹ The above results proved the ability of the double-faced precursor supply DP method in tuning the electronic properties of monolayer MoS₂.

Finally, we examined the universality of the double-faced precursor supply DP method in growing 2D MoS₂ doped by other metal atoms. We found that this method can readily be extended to grow other types of metal-atom doped monolayer MoS₂ (Fig. 4). Figure 4(a) shows an optical microscope image of Fe–MoS₂ with a triangle shape. As for the microstructure, the corresponding HAADF-STEM image in Fig. 4(b) reveals the substitutional Fe atoms at the Mo sites. Due to the different atomic Z-numbers between Fe (Z = 26) and Mo (Z = 42), the intensity of the Fe atom occupied at the cation site shows around 40% lower image contrast than the neighboring Mo atom, as revealed by the intensity profile in Fig. 4(c). Due to the low concentration of Fe dopants, the two main characteristic peaks E_2g (385.6 cm⁻¹) and A_1g (405.2 cm⁻¹) are the same compared with the undoped MoS₂ [Fig. 4(d)]. The inset of Fig. 4(d) is the PL spectrum of Fe–MoS₂ showing a main peak at 658 nm. The optical, atomic-resolved imaging and spectroscopic characterizations of other metal doped MoS₂ samples were also investigated, as shown in Figs. 4(e)–4(h) (Cr–MoS₂), Figs. 4(i)–4(l) [lightly doped Re–MoS₂ (L-Re–MoS₂)], and Figs. 4(m)–4(p) [highly doped Re–MoS₂ (H-Re–MoS₂)]. For L-Re–MoS₂, Re atoms are uniformly incorporated in 2H phase MoS₂, and its Raman spectrum has the same two prominent peaks (382 and 403 cm⁻¹) with the undoped MoS₂. In addition, it shows some new peaks at the low frequency region (151, 220, and 307 cm⁻¹), which are consistent with the peak position of ReS₂.²¹ The inset of Fig. 4(i) is the PL spectrum of L-Re–MoS₂, which exhibits an emission peak of the A exciton at 668 nm and is similar to the undoped MoS₂. When further increasing the Re concentration, the lattice changes from the trigonal prismatic 2H to the 1T′ phase with distorted octahedral units [Fig. 4(n)]. The Raman spectrum of H-Re–MoS₂ [Fig. 4(p)] shows characteristic peaks (151, 220, 275, 303 cm⁻¹, etc.) belonging to 1T′ ReS₂. Overall, the double-faced precursor supply DP method is universal in growing different metal doped MoS₂ and shows potential for phase engineering of MoS₂.

We have developed a double-faced precursor supply DP method for substitutional doping of different metal atoms into monolayer MoS₂. In this new doping method, the diffusion paths of the Mo source and dopant precursor are separated to improve the controllability of the process. As a result, we incorporate different concentrations of V atoms into monolayer MoS₂ to modulate its optical and electrical properties. Extending this doping method to grow some other metal doped MoS₂ is also realized, including Fe–MoS₂, Cr–MoS₂, and Re–MoS₂, showing the universality of this method. Our results can further be used to grow novel materials such as Janus 2D materials⁴⁰ and extend the applications of 2D materials.

First, a drop of Na₂MoO₄ solution [4 μl, 1 mol/l in de-ionized (DI) water] was dropped on the 2 mm thick soda-glass and then dried in an oven. Then, a thinner (0.15 mm) glass was put on top of the above thick glass and the two glasses were sealed together under 660 °C for 30 min to form a glass/metal precursor/glass sandwiched structure. Next, a drop of NaVO₄ solution was spin-coated (20 μl, 1 mol/l in DI water, 3000 rpm for lightly doped V–MoS₂ and 30 μl, 1 mol/l in DI water, 2000 rpm for heavily doped V–MoS₂) on top of the sandwiched structure and dried in an oven. Finally, this sandwiched structure was put into the center of the horizon furnace and S powder (100 mg) was put in the upstream. During the growth, the furnace was heated to 730–750 °C at 50 °C/min and S powder was heated to 150 °C. Ar was used as the carrier gas at a flow rate of 80 SCCM. For growth of other types of doped MoS₂, the metal precursor spin-coated on the substrate was replaced by Na₂CrO₃, NaReO₄, and FeCl₃ solution. For the growth of pristine undoped MoS₂, only Na₂MoO₄ was used.

Optical microscope images were taken by Carl Zeiss Microscopy. AFM measurements were performed with Cypher ES, Asylum Research. Raman and PL spectra were collected through Horiba LabRAM HR Evolution by using 532 nm laser excitation with a beam size of ∼1 μm. XPS spectra were measured by monochromatic Al Kα x rays with a 1486.6 eV, PHI VersaProbe II. HAADF-STEM analyses were carried out using a FEI Titan Themis G2 double aberration corrected TEM with a field emission gun at 60 kV. The convergence angle was about 25 mrad, and the collection range of the HAADF-STEM imaging was set around 52–200 mrad.

The as-grown samples were attached by polyethylene terephthalate (PET) at 65 °C. Then, they were transferred to target substrates (SiO₂/Si or TEM grids) under 90 °C to be attached to substrates tightly. Next, the PET/sample/target substrate was immersed into dichloromethane solution to get rid of the PET, leaving the samples on the target substrate.

FET devices were fabricated using a laser writing system (Aresis Dell, ZKS). In brief, the source and drain areas were patterned by photo-lithography followed by development, metal deposition, and lift-off procedure. The electrodes of the MoS₂ FET and L-V–MoS₂ FET were made of 5 nm Cr and 50 nm Au, and the electrodes of the H-V–MoS₂ FET were made of 5 nm Pd and 50 nm Au. The devices were measured by Lake Shore TTPX under vacuum (10⁻⁵ mbar).

The authors acknowledge support from the National Natural Science Foundation of China (Grant Nos. 51991340, 51920105002, and 51991343), the Guangdong Innovative and Entrepreneurial Research Team Program (Grant No. 2017ZT07C341), the Bureau of Industry and Information Technology of Shenzhen for the “2017 Graphene Manufacturing Innovation Center Project” (Grant No. 201901171523), and the Shenzhen Basic Research Project (Grant Nos. JCYJ20200109144620815 and JCYJ20200109144616617). They also acknowledge the assistance of SUSTech Core Research Facilities, especially technical support from Pico-Centre.

The data that support the findings of this study are available from the corresponding author upon reasonable request.