Large-scale synthesis and exciton dynamics of monolayer MoS₂ on differently doped GaN substrates

2.1 Synthesis of MoS₂/GaN heterostructures

GaN was grown by metal–organic chemical vapor deposition (MOCVD) on c-plane sapphire substrates, during which the SiH₄ gas and Cp₂Mg sources were used for n-type and p-type doping, respectively. The specific epitaxial structures of differently doped GaN are shown in Figure S1, including Si doping concentration around 4 × 10¹⁸ cm⁻³ for n-GaN, Mg doping concentration around 2 × 10¹⁹ cm⁻³ for p-GaN and unintentional doping for u-GaN. The as-prepared GaN substrates were ultrasonically cleaned in acetone, absolute ethanol, and deionized water successively for 15 min, then blown dry with a nitrogen gas gun before the growth process of MoS₂.

As for the growth of MoS₂, sodium molybdate (Na₂MoO₄) was employed as a precursor owing to its low melting point, which can easily transform into a molten liquid intermediate state at a relatively lower temperature. Meanwhile, sodium hydroxide (NaOH) was involved as a promoter to further facilitate the mobility and lateral growth of the molten liquid intermediates [28]. 155 mg Na₂MoO₄ (99 %, aladdin) and 30 mg NaOH (99 %, aladdin) were firstly dissolved in 100 mL de-ionized water to obtain mixture solutions. These solutions were then spin-coated (1000 rpm for 30s, followed by 2000 rpm for 90 s) on the GaN substrates pretreated by O₂ plasma, aiming to achieve a homogeneous distribution of precursor, which is important for obtaining large-area and uniform MoS₂ nanosheets. Notably, the O₂ plasma treatments were maintained for 300 s at a power of 50 W and an O₂ gas-flow rate of 30 sccm using a Plasma Cleaner (Tonson Tec, YZD08-10C). After being completely dried, they were put into a 3-inch three-zone furnace, 50 cm downstream from a ceramic boat carrying 400 mg S powder (99.5 %, aladdin), as shown in Figure 1(a). The three-zone furnace temperature was subsequently risen up to 190/350/850 °C as depicted in Figure 1(b) (with Zone 2 acting as moderate temperature zone to avoid the mutual influence of zone 1 and zone 3) and maintained for 5 min with 250 sccm N₂ flow in an atmospheric pressure for the growth. Meanwhile, the vapor–liquid–solid (VLS) growth occured, during which Na–Mo–O liquid droplets formed and adsorbed sulfur vapor continuously. These droplets diffused laterally on the substrate and precipitated out MoS₂ when saturated, finally achieved large-size MoS₂, as depicted in Figure S2.

Figure 1:

Growth of monolayer MoS₂ on GaN(0001)/sapphire substrates. (a) Schematic illustration of the CVD setup for the growth of monolayer MoS₂ on GaN(0001)/sapphire substrates. (b) Temperature programming process of S source and Mo source. (c) Optical microscopy image of the as-grown MoS₂ nanosheets. (d) AFM image of the MoS₂ nanosheet. Inset is a height profile along the white line. (e) The maximum size and growth rate of monolayer MoS₂ on GaN synthesized by direct CVD growth as reported in other articles [15, 23, 31–37]. The magnified optical microscopy image of a large single crystal MoS₂ nanosheet is presented as the inset (the scale bar is 100 μm).

2.2 Material characterization and optical measurement of MoS₂/GaN heterostructures

The morphology of as-grown MoS₂ nanosheets on GaN substrates was characterized using an optical microscope (Sunny Optical Technology, RX50M). The atomic force microscope (AFM) images were characterized with the aid of a Bruker Dimension AFM system (Bruker Dimension Edge). The Raman spectra and PL spectra were carried out by a laser confocal micro-Raman spectrometer (Horiba Jobin-Yvon, LabRAM HR800). Raman spectra was recorded with two accumulations with 15 s accumulation time under the excitation of a 532 nm excited line for all samples. PL spectra was collected under the excitation of a 532 nm and a 325 nm excited line with the integration time of 5 s. The Raman and PL mapping were recorded by a confocal microscope spectrometer (Alpha300 Raman, WITec). The XPS spectra was performed by an X-ray photoelectron spectroscopy (Kratos AXIS SUPRA+), in which the binding energies of the photo-emission spectra was calibrated against the C 1s peak of adventitious carbon at 284.8 eV. The ultrafast exciton dynamics characterization was performed on a transient absorption spectrometer (HELIOS, Ultrafast Systems) pumped by a Ti:sapphire regenerative amplifier (Legend Elite, Coherent) operating at 5 kHz with fundamental wavelength of 800 nm and pulse width of ∼40 fs. For measurements of MoS₂/GaN heterostructures, the probe beam with a wavelength range spanning over 450–750 nm was focused to a spot with a diameter of ∼3 μm, and the pump beam with a wavelength of 400 nm was focused to a spot with a diameter of ∼5 μm.

3.1 Large-scale synthesis of monolayer MoS₂ on GaN substrates

The typical procedure for synthesizing monolayer MoS₂ nanosheets on GaN substrates was conducted by a NaOH assisted vapor-liquid-solid CVD method, with more details about the specific growth parameters and steps described in the Experimental Section. The morphological characteristics of as-grown MoS₂ nanosheets on u-GaN substrates were analyzed by optical microscopy (OM) and AFM. From its magnified OM image displayed in Figure 1(c), uniform color contrast can be clearly seen, indicating the uniform thickness and in-plane continuity of the MoS₂ nanosheets. Large triangular MoS₂ crystal domains can be distinguished with an average edge length over 200 μm, reaching the surface coverage ratio of more than 90 %. Meanwhile, it is found that the orientation of these MoS₂ domains is randomly distributed, which can be attributed to the large amount of NaOH introduced [28]. The addition of Na element and hydroxide ions has been proven to reduce the nucleation density, inhibit the vertical crystal growth and facilitate the lateral mobility of the Na–Mo–O liquid droplets during VLS growth, leading to monolayer MoS₂ domains with large size and random arrangements [25, 29, 30]. The OM images of MoS₂ nanosheets grown on other doped GaN substrates are shown in Figure S3, indicating similar size and morphology. Figure 1(d) shows the AFM topography image of the synthesized MoS₂ nanosheet. The height profile provided as inset (extracted from the white line in Figure 1(d)) shows a thickness of 0.78 nm, corresponding to a monolayer. It is also worth mentioning that the maximum side length of triangular MoS₂ domains can reach up to 403 μm, together with rapid growth rate over 80 μm/min, which surpass the values of previously reported MoS₂ domains grown on GaN substrates, as compared in Figure 1(e) [15, 23, 31–37].

The composition and crystal quality of the MoS₂/GaN heterostructures were characterized with Raman, PL, and XPS spectra. Two characteristic Raman modes of MoS₂ can be observed in the spectra, as shown in Figure 2(a). The A _1g mode corresponds to the out-of-plane vibration of sulfur atoms and the E 2 g 1 mode associates with the in-plane vibration of Mo and sulfur atoms [38]. For the MoS₂ nanosheets on u-GaN substrates, the two Raman characteristic peaks are fixed at 385.8 cm⁻¹ ( E 2 g 1 ) and 405.8 cm⁻¹ (A _1g), with a frequency difference (Δk) ∼20.0 cm⁻¹, justifying the monolayer nature of the MoS₂. Furthermore, the full width at half maximum (FWHM) of the E 2 g 1 peak is 2.3 cm⁻¹, which is even better than high-quality MoS₂ samples reported previously [39]. As for GaN, two typical Raman modes are observed at 578.7 cm⁻¹ and 735.6 cm⁻¹, representing the E ₂ high and A ₁ longitudinal optical (LO) phonon modes, respectively. Meanwhile, both the broad Raman mode at 451.0 cm⁻¹ and the surface optical (SO) phonon of GaN at 636.6 cm⁻¹ can be attributed to the interlayer electron–phonon coupling of the heterostructure [16, 40]. In addition, the PL spectrum in Figure 2(b) demonstrates a sharp peak at 673.0 nm with relatively high intensity and a narrow FWHM (16.8 nm), suggesting the good crystallinity of the monolayer MoS₂ nanosheets on GaN (the PL spectrum of the substrate is provided in Figure S4 as a comparison). In order to further investigate the uniformity of monolayer MoS₂ nanosheets on GaN substrates, PL mapping integrated by the intensity of A exciton emission, as well as Raman mappings with the intensity of E 2 g 1 mode and A _1g mode were conducted, as displayed in Figure 2(c). Clear color contrast across the entire area is manifested, which indicates the uniform thickness and excellent crystallinity of the samples with few structural defects or disorder at the edge.

Figure 2:

The composition and quality characterizations of the MoS₂/GaN heterostructures. (a, b) Raman (a) and photoluminescence (b) spectra of monolayer MoS₂ grown on GaN, respectively. The inset in (a) is a close-up view for the characteristic Raman modes of MoS₂. (c) Optical microscopy image, PL intensity mapping for A-exciton, and Raman intensity mapping for E 2 g 1 and A _1g peak of a representative MoS₂ single crystal. (d, e) Core level XPS spectra and peak fits from (d) Mo 3d and (e) S 2p regions for the MoS₂/GaN heterostructure. Components from S 2s (MoS₂) and Mo 3d (MoS₂ and MoO _x ), together with S 2p doublet (MoS₂) and Ga 3s (GaN) are shown in (d) and (e), respectively.

XPS spectra of the heterostructure are depicted in Figure 2(d), (e) and Figure S5. The Mo⁴⁺ 3d _5/2 and 3d _3/2 peaks are located at 229.5 eV and 232.7 eV, while the S²⁻ 2p _3/2 and 2p _1/2 peaks are located at 162.3 eV and 163.5 eV, respectively, consistent with the corresponding binding energies reported in the literature for MoS₂ single crystal [41]. Also, the stoichiometric ratio between S and Mo from the intensities of the respective XPS peaks could be calculated as S: Mo = 1.92: 1, further confirming the high quality of MoS₂ nanosheets with few sulfur vacancies. In addition, the Ga 3s peak is observed at 160.5 eV (Figure 2(e)), together with Ga 2p _1/2 and Ga 2p _3/2 peaks at around 1145.4 and 1144.8 eV (Figure S5(a)), which are in line with the XPS spectra of GaN reported previously, further indicating the stability of GaN during the growth of MoS₂ [42]. Moreover, there is no signature of any chemical state associated with Mo or S in the Ga 3d core level spectra (Figure S5(b)), suggesting atomically sharp heterointerface, which is clear evidence of a van der waals interaction between MoS₂ and GaN.

3.2 Doping effect and exciton dynamics of monolayer MoS₂ on differently doped GaN

According to previous reports, 2D layered MoS₂ is sensitive to the surrounding environment, especially the substrates, that may dope the upper monolayer and thereby affect the exciton dynamics [43, 44]. To further reveal the doping effect and charge transfer properties of MoS₂ in the heterostructures, monolayer MoS₂ has been grown on differently doped GaN under the same condition and compared by Raman spectra. The optical resonance effects associated with the substrate’s geometrical features were largely ignored, since the structures were completely identical. Figure 3(a) depicts a comparison of the Raman spectra measured for monolayer MoS₂ on differently doped GaN, in which the MoS₂ on sapphire is also given as a reference. For the MoS₂ on GaN, reduced frequency difference (Δk) between two Raman characteristic modes is observed, accompanied with blue-shift of the E 2 g 1 peak, which indicate less tensile strain in MoS₂ due to perfect lattice matching [16, 45]. In addition, the Raman intensity ratio of the E 2 g 1 to A _1g mode was further investigated in Figure 3(b), as it can be regarded as an indicator of doping levels [37, 46]. The observed lower ratio of monolayer MoS₂ on unintentionally doped u-GaN than the counterparts on sapphire may be ascribed to the higher crystalline quality with fewer sulfur vacancies which generally behave as deep donors. Moreover, when the doping type of GaN is converted from n-type to p-type, the E: A ratio is largely reduced, suggesting a much lower n-doping level in MoS₂. This modulation of the n-doping level of MoS₂ can be mainly attributed to substrate-induced doping effect, which also results in the substantial blue-shift of A _1g mode [47].

Figure 3:

Effect of the differently doped GaN on the Raman modes of monolayer MoS₂. (a) Raman spectra of monolayer MoS₂ nanosheets on differently doped GaN and Sapphire. Inset is the frequency difference (Δk) between E 2 g 1 and A _1g modes of all samples. (b) Intensity ratio of the E 2 g 1 and A _1g peak and the frequency of the A _1g mode as a function of the doping type of GaN.

To confirm the above findings, the PL spectra (excitation wavelength: 532 nm) of the monolayer MoS₂ on all substrates were further analyzed by Lorentzian fitting of the experimental spectra in order to extract the individual excitonic transitions. Figure 4(a) shows three fitted exciton peaks of the monolayer MoS₂ on different substrates, which are identified as neutral A and B excitons as well as negative trions (A−). The two neutral exciton emission peaks are ascribed to the splitting of valence-band maximum owing to the broken spatial-inversion symmetry in monolayer MoS₂, whereas the trions (A−) are formed by the binding of a free electron to a neutral A exciton, represented as e + A → A− [48]. It is noteworthy that a larger PL intensity ratio of A exciton versus B exciton (A/B) can be observed in MoS₂ on GaN, illustrating the excellent quality, as this ratio is considered an effective method to evaluate the crystal quality of MoS₂ qualitatively [49]. Besides, the PL of monolayer MoS₂ on p-GaN substrate is strongest among all the substrates we studied, around five times higher than that on sapphire substrate, while that on n-GaN substrate is the worst. These features indicate that the MoS₂ systhesized on these substrates are differently n-doped, which will modulate the relative concentration of trions and neutral excitons [44]. Specifically, this relative concentration can be further analyzed by spectrum weight ratio I _A−/I _A, as displayed in Figure 4(b), where the I _A and I _A− are the integrated emission intensity of neutral excitons and trions, respectively [50, 51]. As it can be seen that, the I _A−/I _A ratios of MoS₂ on p-GaN and u-GaN are much smaller than that on n-GaN, indicating the concentration of trions A-is lower in MoS₂ on p-GaN and u-GaN. Therefore, it can be concluded that p-GaN can attract electrons from monolayer MoS₂ and weaken its intrinsic n-doping while n-GaN provides electrons and enhance the n-doping level. Meanwhile, since trion emission has lower efficiency and broader spectrum compared to the exciton emission [44, 51], the normalized PL intensity of monolayer MoS₂ decreases when the substrate changes from p-GaN, u-GaN to n-GaN.

Figure 4:

Effect of the differently doped GaN on the photoluminescence of monolayer MoS₂. (a) PL spectra (excitation wavelength: 532 nm) and corresponding exciton peak fitting of monolayer MoS₂ on p-GaN, u-GaN, n-GaN and sapphire. The neutral A exciton (red), B exciton (blue) and negative A-trion (green) fit are indicated together with the resulting cumulative spectrum (black). (b) The spectrum weight ratio I _A−/I _A and normalized PL intensity as a function of the doping type of GaN. (c) PL spectra (excitation wavelength: 325 nm) of MoS₂/u-GaN heterostructure and bare u-GaN. Inset shows the schematic of the charge transfer at the MoS₂/GaN heterointerface under such excitation.

To further elucidate the charge transfer process in MoS₂/GaN heterostructures, different excitation sources were employed to excite GaN and MoS₂/GaN heterostructures, respectively. Figure 4(c) displays the PL spectra of the MoS₂/u-GaN heterostructures under the excitation of a 325 nm laser excitation. The PL spectrum of u-GaN consists of its band-edge emission centered at 360 nm as well as a weak defect emission centered around 450 nm. However, these two emissions are remarkably weakened in MoS₂/u-GaN heterostructures, owing to the type I band alignment at the heterointerface, which facilitates the transfer of the photocarriers in GaN to monolayer MoS₂ [16, 33]. More details about the band alignment of our samples are described in Supplementary Information (Note 1), based on the Kelvin probe force microscopy (KPFM) measurements (Figure S6), UPS spectra (Figure S7) and the established band diagram (Figure S8). Notably, the PL intensity of MoS₂ is observed to be relatively weak, which can be attributed to the dominant non-radiative recombination of the hot carriers excited by such high photon energy [52]. In addition, the PL spectra of the heterostructures formed by MoS₂ and other doped GaN is shown in Figure S9. Faint band-edge emission and defect emission of GaN in those heterostructures can also be observed, further verifying the efficient charge transfer process in MoS₂/GaN heterostructures.

To understand the ultrafast exciton dynamics in the MoS₂ on differently doped GaN, pump-probe TA measurements were performed (details are provided in the Experimental Section). The transient absorption spectra decay from different samples after photoexcitation of 400 nm with density of 100 μJ/cm² are depicted in Figure 5(a–c). The spectra of monolayer MoS₂ on differently doped GaN at different pump–probe delay times all feature the same three prominent negative peaks A, B and C, which are ascribed to the ground-state bleaching (GSB) due to Pauli blocking or state filling of the excitonic resonances. The lower energy peaks A at ∼ 670 nm and B at ∼ 615 nm correspond to excitonic transitions located at the K point. The higher energy peak C near 450 nm can be assigned to band nesting transition in different regions of the Brillouin zone, especially at the Γ point [53, 54]. The two positive peaks A* (685 nm) and B* (640 nm), at the right side of bleaching peaks A and B, respectively, can be attributed to the band gap renormalization and broadening photoinduced absorption [53]. It is also worth mentioning that both A and B excitonic bands of MoS₂ on GaN show a red-shift compared to those on sapphire (Figure S10), which could be caused by the exciton–phonon coupling in the heterostructures [16].

Figure 5:

Effect of the differently doped GaN on the exciton dynamics of monolayer MoS₂. (a–c) Transient absorption spectra of the monolayer MoS₂ on p-GaN (a), u-GaN (b) and n-GaN (c) at different delay times, respectively, under excitations with fixed pump density of 100 μJ/cm². (d) The kinetics traces of the A exciton state extracted from the TA results of samples with three doping types. The inset is a zoom-in view of the early 100 ps. The open circles are experimental data, the solid lines are the exponential fitting described in the main text, with consideration of the instrument response function (∼100 fs). (e) Schematic of the exciton dynamics in monolayer MoS₂ on GaN.

In Figure 5(d), we summarize the normalized kinetics of A exciton of the three samples extracted from the TA spectra, so that we are able to quantitatively analyze the dynamics. All relaxation dynamics exhibit a multi-exponential decay trend, and can be fitted well by a tri-exponential function of [55, 56]:

where A ₁, A ₂ and A ₃ represent the respective amplitude of lifetime τ ₁, τ _2, and τ ₃. The fitting parameters are tabulated in Table 1. The carrier relaxation processes for the A excitons are illustrated in Figure 5(e). The fast decay components (see the inset of Figure 5(e)) τ ₁ of ∼2 ps are assigned to the formation process of exciton from free electron-hole pairs under nonresonant excitation [57, 58], with consistent proportion of the whole decay process in a range of 40–50 % for all three samples. The slow components τ ₂ are in the same time scale of ∼50 ps for all samples with differently doped GaN substrates. The decay with τ ₂ may be attributed to the exciton–phonon scattering process, since the excitons generated in MoS₂ are strongly coupled to GaN phonons via the 2D material–substrate interaction, then lose their energy to the crystal lattice by emitting phonons in GaN [16, 59]. Remarkably, the long decay components τ ₃ associated to the radiative recombination of excitons [56, 60], exist a strong correlation to the doping type of GaN. For the MoS₂/p-GaN sample, the τ ₃ process takes 2800 ps, apparently longer than those in samples of MoS₂/u-GaN (1756 ps) and MoS₂/n-GaN (1680 ps). The prolonged exciton lifetime during radiative recombination in MoS₂/p-GaN can be attributed to the increased concentration of exciton with longer lifetime and reduced concentration of trion with shorter lifetime caused by substrate doping, which is consistent with the lowest spectrum weight ratio I _A−/I _A shown above [44].