Light-matter interactions and afterglow emission of 2D transition metal dichalcogenides

Summary

Light-matter interactions underlie the fundamental physics of two-dimensional (2D) transition metal dichalcogenides (TMDs). We show that both pure and alloyed 2D TMDs can absorb considerable optical energy and slowly emit it, which we attribute to spin-forbidden electron-flip transitions (i.e., phosphorescence). We measured an optical charging time of ∼20 s, indicating that photons are the energy source for this behavior, and extracted phosphorescence quantum efficiency (PQE) values reaching ∼3%. As W atoms are heavier than Mo atoms, Mo_1–xW_xS₂ (0 < x < 1) alloys demonstrated the heavy-atom effect, showing that a high concentration of W atoms presented a stronger afterglow signal. Our samples showed long-lasting room temperature afterglow, from several minutes to 1 h, and signal quenching in regions of low atomic order and when the TMDs exhibited enhanced interactions with ambient molecules. This understanding sheds light on the fundamental light-matter interactions of TMDs and, thereby, may inspire their integration into future optical applications.

Graphical abstract

Highlights

• •Monolayer TMDs show ultra-long, room temperature phosphorescent emission
• •Atomically thin TMDs can store considerable optical energy (milli-Joules)
• •We found optical charging time of ∼20 s and quantum efficiency between 0.15% and 3%
• •We observed the heavy atom effect in TMD alloys

Physics; Optics; Materials science; Materials synthesis

Introduction

Monolayer transition metal dichalcogenides (TMDs) are atomically thin semiconducting materials with strong light-matter interactions^1,2,3 and spin-orbit coupling.^4,5 They present a direct bandgap with a minimum of the conduction band and a maximum of the valence band located at the K^± points of their hexagonal Brillouin zone.⁶ In this work, we focus on the afterglow properties of TMDs, which are often associated with the excitation of singlets and triplets.

Previous work suggested the possible formation of charged excitons in WS₂, which can be formed in the K⁺ and K⁻ valleys and generate intervalley trions—some of which form optically active triplets.⁷ These three-particle complexes, composed of an electron-hole pair and an excess charge carrier (either an electron or a hole), form triplet-state trions when their electron spins are parallel and from different valleys.^6,8 While previous work suggested that these triplets can exist at room temperature, their light emission is more efficient at lower temperatures.⁹ The emission of these triplets can be increased in several ways; for example, it was shown that the heterostructure of W-based TMDs enables stronger emission at room temperature.^10,11

Previous publications showed that several mechanisms can induce afterglow illumination in materials. These mechanisms mainly include electroluminescence,¹² chemiluminescence,¹³ and delayed fluorescence. However, in the present study, we focus on the phosphorescence of atomically thin TMDs. Phosphorescent materials emit light long after their illumination has ceased (i.e., they demonstrate afterglow). Although phosphorescence is erroneously associated with phosphorus (which glows upon chemical reaction with oxygen), it is, in fact, related to the excitation of a spin-allowed singlet state and its transition to a spin-forbidden triplet state, which yields long emission when the electrons return to their ground state.¹⁴ Most commonly, phosphorescence occurs when electrons from the first excited singlet state (S₁) conduct intersystem crossing (i.e., forbidden spin inversion) to the first excited triplet state (T₁) and light is emitted when they return to the ground singlet state (S₀), as illustrated in Figure 1A. Therefore, optically activated triplets play an instrumental role in phosphorescent materials.

Figure 1.

Description of phosphorescence

(A) Jablonski diagram describing the main photophysical process of phosphorescence.

(B) Illustration of the phosphorescence process in monolayer TMDs. Left: light emission of the TMD during laser illumination (fluorescence emission). Right: light emission of the TMD after the laser was turned off (phosphorescence emission).

Phosphorescent materials have been integrated into everyday applications, such as safety markers and glow-in-the-dark toys, as well as in scientific devices, including opto-chemical sensors,⁴ biomedical diagnostic devices,¹⁵ and more. The phosphorescence signal is sensitive to molecular aggregation,¹⁶ temperature changes,¹⁷ and the existence of oxygen¹⁸ and humidity¹⁹—all of which can enhance non-radiative interactions that diminish the phosphorescence signal (i.e., quenching), reduce the phosphorescence quantum efficiency (PQE), and, therefore, may challenge the large-scale applicability of phosphorescent materials.

Phosphorescent light emission at room temperature is highly desirable for many applications, as it simplifies their operation. However, this requirement often entails the use of rare and expensive metals,²⁰ which hinders their widespread use. Alternative materials, such as organic materials,²¹ have been suggested for use as room temperature phosphorescent materials, but some organic materials showed a low PQE due to weak spin-orbit coupling, which is insufficient for breaking the spin-prohibition of the singlet-triplet transition (i.e., the intersystem crossing). Among the reported materials, hybrid organic-inorganic perovskites have been identified as room temperature phosphorescent materials due to their efficient Dexter-type energy transfer from the inorganic to the organic components.²² Other studies emphasized how the interactions between the energetically matched donor and acceptor can enhance the phosphorescence signal.²³

Improving the longevity of the afterglow of phosphorescent materials is an additional important requirement and, indeed, numerous studies have focused on this goal. Suggested strategies included engineered materials with matched energy structures.²⁴ The incorporation of heavy atoms and heteroatoms has also been found to improve the singlet-to-triplet excitation, facilitating the intersystem crossing.²⁵ The enhancement of the phosphorescence signal has also been achieved via the incorporation of sulfur-rich materials and heavy atoms.²⁶ The existence of aromatic carbonyl was reported to improve phosphorescence since it assists the spin-orbit coupling.²⁷ Additionally, the engineering of host-guest doping was presented as an approach to improve long-lived phosphorescence afterglow,²⁸ and an H-atom aggregation approach was reported to slow down triplet decay in metal-free organic materials.²⁹

Two-dimensional (2D) TMDs are attractive materials for a wide range of optical applications. Indeed, 2D TMDs were integrated into photodetectors,³⁰ photovoltaic cells,³¹ and light-emitting diodes.³² In addition to their direct bandgap structure and strong light-matter interactions, 2D TMDs show high carrier mobility,³³ adjustable optoelectronic properties,³⁴ and other attractive physical properties, such as high mechanical stretchability^35,36,37 and piezoelectricity.³⁸ Contrary to many phosphorescent materials, 2D TMDs benefit from well-established, safe, scalable, and low-cost synthesis processes.³⁹ In their zero-dimensional form, MoS₂ quantum dots (QDs) are among the few TMD materials that present phosphorescence, exhibiting an afterglow of several seconds and high sensitivity to temperature change.²⁶ Other phosphorescent QDs have been reported, including silica nanoparticles,⁴⁰ gold nanoclusters,⁴¹ and carbon dots.⁴² Here, we study the afterglow behavior of TMDs, as illustrated in Figure 1B.

In the present work, we demonstrate the fluorescence illumination of TMDs (specifically, WS₂, MoS₂, WSe₂, and their alloys) when subjected to laser illumination (Figure 1B, left), followed by an additional light emission after we stopped the laser illumination (Figure 1B, right), which we attribute to the possible phosphorescence of these materials. Indeed, we studied important mechanisms that significantly affect the afterglow, such as the heavy atoms effect and nonradiative relaxation routes induced by ambient air molecules and lattice disorder. We also investigated the optical charging process of the TMDs, the stored and emitted energies, and the resulting PQE.

Results

CVD growth of TMD samples and material characterization

We grew monolayer TMDs in a high-temperature CVD process on Si/SiO₂ wafers. The sources of the metal and chalcogen atoms were solid powders, which we first evaporated and then crystallized on the wafers. We focused on WS₂ and MoS₂ monolayers, as shown in Figures 2A and 2B, which demonstrated the typical triangular-shaped grains. We verified their monolayer structure via Raman analysis (i.e., by calculating the difference between the Raman shifts of the A_1g and E¹_2g peaks), as shown in Figures 2C and 2D. The PL emissions that correspond to the fluorescence responses of the WS₂ and MoS₂ monolayers are shown in Figures 2E and 2F, and they demonstrated peaks at ∼646 nm (∼1.9 eV) and ∼682 nm (∼1.8 eV), respectively, as typically observed in these monolayered materials. Notably, since the PL emission is highly sensitive to many parameters, such as stresses, growth parameters, surface interactions, local defects, and more, the PL peak position may somewhat vary. For example, publications regarding CVD grown WS₂ presented PL peak locations ranging from above 2.0 eV to ∼1.8 eV.^43,44 A fluorescence mapping of monolayer WS₂ is provided in the supplemental information section (Figure S5) and also shows some spatial variation and a stronger signal close to the edges.

Figure 2.

Material characterization of CVD-grown WS₂ and MoS₂

(A and B) Optical images of a CVD-grown monolayer WS₂ (A) and MoS₂ (B). Scale bars: 10 μm.

(C and D) Raman spectra of monolayer WS₂ (C) and MoS₂ (D).

(E and F) PL emissions of monolayer WS₂ (E) and MoS₂ (F).

We conducted an XPS analysis to verify the atomic structure of the TMDs. In Figures 3A and 3B, we show the W 4f and the S 2p peaks related to WS₂. A deconvolution of the W 4f peak revealed W 4f_7/2 at ∼32.5 eV and W 4f_5/2 at ∼34.7 eV, indicating the vast existence of W–S bonds, in addition to lower peaks of W 4f_7/2 at ∼35.6 eV and W 4f_5/2 at ∼37.7 eV, which indicate that some of the metal ions were oxidized. The S 2p peaks presented an S 2p_3/2 peak at ∼162.3 eV and an S 2p_1/2 peak at ∼163.4 eV, likewise, indicating the existence of numerous S–W bonds. These findings demonstrate a hexagonal phase of the WS₂ lattice. Similarly, an XPS scan of MoS₂ presented the Mo 3d peaks (subpeaks: Mo 3d_3/2 at ∼233.6 eV, Mo 3d_5/2 at ∼230.4 eV, a lower intensity Mo⁶⁺-related peak at ∼236.8 eV, and an S 2s peak at ∼227.6eV) and the S 2p peaks (subpeaks: S 2p_3/2 at ∼163.2 eV and S 2p_1/2 at ∼164.4 eV), as shown in Figures 3C and 3D, respectively. Overall, these readings demonstrate a hexagonal-phase MoS₂ dominated by Mo–S bonds. Topographic atomic force microscopy (AFM) thickness measurements of the WS₂ and MoS₂ monolayers are shown in the supplemental information section (Figure S2).

Figure 3.

XPS measurements

(A and B) High-resolution XPS measurements of WS₂.

(C and D) High-resolution XPS measurements of MoS₂.

To investigate the heavy-atom effect, we synthesized monolayer TMD alloys of Mo_1–xW_xS₂. These samples showed varying concentrations of W atoms (quantified by x), which are significantly heavier than Mo atoms. We used the nucleation-diffusion process, as it enables a wide range of alloy compositions (0 ≤ x ≤ 1). Briefly, we first grew WS₂ on a Si/SiO₂ wafer, as explained in the "method details" section, followed by the successive nucleation of MoS₂ around the WS₂ and the diffusion of Mo atoms into the WS₂ lattice to achieve monolayer Mo_1–xW_xS₂. Further details regarding this process were reported elsewhere.⁴⁵ Next, we characterized the atomic composition of the alloys (i.e., we acquired x) by using EDS (see more EDS results of alloys in the supplemental information section, Figure S7). Previous works have shown that the nucleation-diffusion process has a minor impact on the lattice order.⁴⁵ Moreover, theoretical calculations revealed that the alloying process barely changes the lattice parameter compared to that of WS₂, further demonstrating the small impact of the alloying process on the atomic configuration.⁴⁶ Thus, below, we attribute the difference between the emission of pure materials and TMD alloys primarily to the heavy-atom effect.

An optical image of a representative alloy flake is shown in Figure 4A, and the Raman spectra of samples with varying x values are presented in Figure 4B (more Raman measurements are shown in the supplemental information section, Figure S3). Notably, the Raman spectra of the TMD alloys demonstrated the characteristic modes of both WS₂ and MoS₂, with more pronounced WS₂ modes when x→1 and more pronounced MoS₂ modes when x→0. The fluorescence responses of these samples, shown in Figure 4C, further demonstrated the shift from WS₂- to MoS₂-like behavior as x decreases.

Figure 4.

Material characterization of nucleation-diffusion CVD-grown Mo_1–xW_xS₂ alloys

(A) Optical image of a monolayer Mo_1–xW_xS₂ alloy. Scale bars: 10 μm.

(B) Raman spectra of samples with varying x values. Notably, x = 1 corresponds to WS₂ and x = 0 corresponds to MoS₂.

(C) Fluorescence emission of alloys with different compositions.

Afterglow signal acquisition

We acquired the afterglow behavior of our samples by illuminating them with a 532 nm, 10 mW laser for several tens of seconds (exceeding the charging time; see below) and then turning off the laser and acquiring the emitted light, as illustrated in Figure 1B. Each spectrum shown below corresponds to the signal accumulated during a ∼30–90 s period (see the “method details” section for experimental parameters). To ensure that the afterglow signals do not stem from the measurement equipment, we performed similar measurements on the SiO₂ layer, which lacks TMDs; indeed, in this case, only noise signals were acquired, with no afterglow peaks after turning off the laser (these measurements are provided in the supplemental information section, Figure S4).

Figure 5A shows the fluorescence signal (i.e., the emitted light during illumination) of monolayer WS₂, demonstrating a clear peak at ∼645 nm (∼1.9 eV). Following the illumination, we turned off the laser and acquired the emitted afterglow until the signal decays below the noise level; we denote this time as the afterglow lifetime. Selected spectral afterglow responses obtained after turning off the laser, shown in Figure 5B, demonstrated a higher peak wavelength (i.e., lower energy emission) than the fluorescence response, as further discussed below. We present the afterglow peak location in the supplemental information section, Figure S8, which shows no observable trend with respect to the time. The peak intensity of the emitted afterglow signals as a function of time, shown in Figure 5C, demonstrated intensity reduction with time and an afterglow lifetime of ∼1 h. We estimated the emitted energy by integrating the spectral signal (i.e., summing photon counts) and converting it to energy units (see further details regarding the calibration of the optical spectrometer in the “method details” section). The accumulated emitted energy as a function of time, shown in Figure 5D, demonstrated that the monolayer WS₂ emitted energy of ∼0.25 mJ.

Figure 5.

Light emission response of a representative monolayer WS₂ sample

(A) Fluorescence emission signal.

(B) Selected phosphorescence time-responses.

(C) Phosphorescence peak intensities as a function of time. Error bars represent measurement uncertainty. Dashed line: exponential decay fit.

(D) Accumulated emitted energy as a function of time.

(E) Accumulated emitted energy as a function of time for multilayer WS₂.

Generally, TMDs in monolayer form demonstrate direct semiconduction behavior, which shifts to indirect behavior as the number of layers increases. We found that multilayer WS₂ exhibited a phosphorescence response that was comparable to that of the monolayer. Figure 5E shows the accumulative emitted energy of multilayer WS₂ as a function of time (see Figure S1 in the supplemental information for more data). Notably, the transition of thick WS₂ samples to an indirect bandgap resulted in a weaker fluorescence signal than that of monolayer samples. Notwithstanding this transition, we show that multilayer WS₂ can still accumulate considerable optical energy, similar to that of monolayer WS₂.

We conducted similar experiments on monolayer MoS₂ and showed its fluorescence response, selected afterglow responses, phosphorescence intensities, and accumulated emitted energy in Figure 6. The afterglow peak locations are presented in the supplemental information section, Figure S8, again showing no observable trend with respect to the time. The accumulated emitted energy of monolayer MoS₂ was significantly lower than that of monolayer WS₂. In addition, the afterglow decay time of MoS₂ was shorter than that of monolayer WS₂, indicating a shorter lifetime of the phosphorescence response of MoS₂. We attribute the differences between WS₂ and MoS₂ mainly to the heavy-atom effect, as discussed below.

Figure 6.

Light emission response of a representative monolayer MoS₂ sample

(A) Fluorescence emission signal.

(B) Selected phosphorescence time-responses.

(C) Phosphorescence peak intensities as a function of time. Error bars represent measurement uncertainty. Dashed line: exponential decay fit.

(D) Accumulated emitted energy as a function of time.

The afterglow characteristics of monolayer WSe₂, including the fluorescence response, selected phosphorescence responses, peak phosphorescence intensities, and accumulated emitted energy, are presented in Figure 7 and the peak locations are shown in the supplemental information section, Figure S8. Notably, WSe₂ showed an afterglow response duration longer than 1,000 s, similar to WS₂. The accumulated emitted energy graphs of the examined TMDs (Figures 5D, 6D, and 7D) show that 1–1.5 min after turning off the laser, WS₂ emitted 20%–30% of the stored energy, and WSe₂ emitted ∼40% of the stored energy. Conversely, MoS₂ emitted significantly more (50%–60%) of the stored energy due to its shorter phosphorescence lifetime.

Figure 7.

Light emission response of a monolayer WSe₂ sample

(A) Fluorescence emission signal.

(B) Selected phosphorescence time-responses.

(C) Phosphorescence peak intensities as a function of time. Error bars represent measurement uncertainty. Dashed line: exponential decay fit.

(D) Accumulated emitted energy as a function of time.

In a different set of experiments, we studied the optical charging time of monolayer WS₂ and MoS₂. In Figure 8A, we show the emitted afterglow energy of WS₂, acquired 90 s after we turned off the laser, as a function of the laser illumination duration (i.e., the optical charging time). Below ∼20 s, the emitted energy increased with the illumination duration. The emitted energy reached its maximal value after ∼20 s of illumination. For longer illuminations, we observed saturation as the emitted energy stopped increasing, even for longer illumination durations. MoS₂ showed a similar behavior, as it exhibited an increase in emitted energy for a short illumination duration, reaching saturation after ∼20 s, although it emitted a lower amount of energy than that of WS₂ (Figure 8B). Therefore, we deduce that the TMDs were optically charged by the laser and that this optical charging process reached saturation after ∼20 s.

Figure 8.

Optical charging time

(A and B) Afterglow emitted energy 90 s after turning off the laser vs. initial illumination time for monolayer WS₂ (A) and MoS₂ (B).

Enhancing non-radiative interactions

Non-radiative interactions are among the most dominant parameters that limit the performance of phosphorescent materials. Therefore, we studied them under controlled conditions. First, we investigated the afterglow characteristics in the grain boundaries of CVD-grown monolayer WS₂ (see the inset of Figure 9A). The peak emission intensity as a function of time is shown in Figure 9A. While the afterglow response was clearly observed, Figure 9B presents a lower energy emission and faster decay, as compared with the WS₂ body (Figure 5). Grain boundaries are characterized by disorganized atomic structure and a higher density of defects compared with the grain body. Such an atomic structure enhances the relaxation of the triplet-state electrons to the ground state through non-radiative interactions, which leads to a much smaller phosphorescence response and a weaker afterglow signal.

Figure 9.

Non-radiative interactions

(A) Phosphorescence peak intensities as a function of time for the WS₂ grain boundary. Error bars represent measurement uncertainty. Dashed line: exponential decay fit. Inset: optical image of WS₂ grain boundary (emphasized by arrows). Scale bars: 10 μm.

(B) Accumulated emitted energy as a function of time for WS₂ grain boundary.

(C) Phosphorescence peak intensities as a function of time for WS₂ exposed to humidity. Error bars represent measurement uncertainty. Dashed line: exponential decay fit.

(D) Accumulated emitted energy as a function of time for WS₂ exposed to humidity.

We also explored the non-radiative interactions of WS₂ in the presence of ambient H₂O molecules. We exposed the monolayer WS₂ to high humidity, which resulted in the enhanced interactions of the monolayer with the H₂O molecules. The afterglow response is shown in Figures 9C and 9D and, again, demonstrated lower intensity and shorter emission lifetime than WS₂ illuminated under ambient conditions (Figure 5). Previous work has shown that during the first few minutes of illumination, TMDs absorb small ambient molecules, such as oxygen and H₂O molecules.⁴⁷ In addition, the exposure to H₂O molecules encourages the oxidation of the TMD lattice and, thereby, increases the defect density. Therefore, we conclude that when exposed to high humidity, H₂O molecules accumulate on the TMD surface and increase the number of defects. Thereby, humidity enhances non-radiative interactions, as more energy is dissipated to the H₂O molecules and relaxed via interactions with the atomistic defects, compared to the case of no or little exposure to ambient humidity. Therefore, the exposure to humidity resulted in a weaker phosphorescence response.

Heavy-atom effect

The atomic composition of phosphorescent materials plays a determinant role in their afterglow optical behavior. According to the heavy-atom effect, larger (heavier) atoms have larger nuclei and, therefore, they present stronger spin-orbit coupling, which facilitates the spin-forbidden intersystem crossing, where electrons swap their spin and reach the triplet state.⁴⁸ Indeed, previous publications reported that materials consisting of large atoms strengthen the spin-orbit coupling and improve the intersystem crossing rate.^49,50 Accordingly, materials composed of heavier atoms are expected to present stronger phosphorescence signals than those with smaller atoms. We investigated this effect by studying the phosphorescence emission of Mo_1–xW_xS₂ alloys. Notably, W atoms are significantly larger than Mo atoms and, therefore, as x increases, the alloy consists of a higher concentration of heavy atoms.

The fluorescence emissions of the alloys are shown in Figure 4C and demonstrate the shifting of the peak as a function of W atom concentration, x, from WS₂- (x→1) to MoS₂-like (x→0) behaviors, with intermediate behavior for alloys (0 < x < 1). The phosphorescence intensities of alloys of several different W concentrations are shown in Figures 10A–10C and their corresponding emitted energy is shown in Figure 10D. Importantly, alloys with larger values of x, namely, with a higher concentration of heavy W atoms, demonstrated a stronger phosphorescence signal with greater emitted energy, confirming the heavy-atom effect in TMD alloys.

Figure 10.

Light emission response of monolayer Mo_xW_1–xS₂ samples

(A–C) Phosphorescence peak intensities as a function of time for Mo_1–xW_xS₂ alloys of different atomic compositions, noted on the graphs. Inset: optical image of the alloy. Scale bars, 10 μm. Error bars represent measurement uncertainty. Dashed lines: exponential decay fit.

(D) Emitted energy as a function of W concentration, x.

PQE

The PQE is defined as the ratio between the phosphorescence emitted energy and the energy absorbed by the material; therefore, it quantifies the effectiveness of converting the absorbed energy into the afterglow light emission. We first calibrated our optical spectrometer to convert its photon counts to energy (see “method details” section) and then integrated the absorbed energy over the illumination duration, as follows:

$$E_{absorbed}={\phi }_{TMD}{\int }_0^{t_i}Pdt$$ (Equation 1)

where φ_TMD is the optical absorbance of WS₂ (∼6%) and MoS₂ (∼10%, both values correspond to wavelength illumination of 532 nm),^51,52,53,54 P is the power of the illumination laser, and t_i is the illumination duration. We estimated the emitted phosphorescence energy as follows:

$$E_{emitted}=A_{cal}{\int }_{t_i}^{t_e}{\int }_0^{\infty }I\left(\lambda \right)d\lambda dt$$ (Equation 2)

Here, λ is the wavelength, I(λ) is the intensity spectrum, and A_cal is the extracted calibration constant required for converting the photon counts to energy. Equation 2 indicates that we sum the spectral energy from the time that we turn off the laser (t_i) until the emission decays (t_e). Thus, we calculate the PQE as follows:

$$\Phi =\frac{E_{emitted}}{E_{absorbed}}$$ (Equation 3)

The values of PQE are shown in Figure 11A for MoS₂, WS₂, and their alloys. Although our calculation is merely a rough estimate of the PQE values, these results demonstrate the heavy-atom effect, as WS₂ shows the highest value Φ_WS2 ≅ 3.0%, while as the concentration of W atoms decreased, the PQE decreased as well. Indeed, MoS₂ presented a PQE of Φ_MoS2 ≅ 0.15% and alloys demonstrated intermediate PQE values, ranging between those of WS₂ and MoS₂. The results shown in Figure 11A are correlated to the emitted energy, Figure 10D. Therefore, we conclude that while WS₂ and MoS₂ demonstrate comparable optical absorbance, the higher concentration of heavy W atoms dictates a lattice that is capable of storing and emitting more optical energy, and as a result, it is associated with a higher PQE. A comparison with other materials shows that the PQE values that we found are relatively high, especially for WS₂. We obtained PQE values larger than those reported for metal complexes,⁵⁵ but comparable to those reported for carbon QDs embedded in a polymer.⁵⁶

Figure 11.

Emission energy and PQE

(A) PQEs of WS₂, MoS₂, and Mo_1–xW_xS₂ alloys.

(B–E) Emission peak energies of WS₂ (B), MoS₂ (C), WSe₂ (D), and Mo_1–xW_xS₂ alloys (E). Error bars represent measurement uncertainty.

(F–I) The difference between the fluorescence and phosphorescence emission energies of WS₂ (F), MoS₂ (G), WSe₂ (H), and Mo_1–xW_xS₂ alloys (I).

Discussion

We attribute the relatively high PQE of WS₂ to several parameters. First, as noted above, the heavy W atoms contribute to the strong spin-orbit coupling and increase the probability of the spin inversion process of excited electrons, leading to their increased relaxation from the triplet excited state to the ground singlet state through phosphorescence (i.e., emissive relaxation). Indeed, W atoms were reported to play an important part in the afterglow emission of other materials.⁵⁷ Similarly, chalcogen atoms were reported to contribute to the non-metallic heavy-atom effect, with an emphasis on sulfur atoms.²³ Contrary to some studies, which showed the possible energy transfer between energetically matched molecules based on donor-acceptor mechanism,⁵⁸ in our case, there is a geometric and energetic mismatch between the atomically ordered TMDs and the amorphous SiO₂. This mismatch impedes energy transfer between the TMD and the SiO₂ and, therefore, limits non-radiative relaxation, which improves the PQE and longevity of the phosphorescence response. The time decay responses of our TMD materials correspond to an exponential decay (R² > 0.95), as expected for triplet decay.⁵⁹

Several possible afterglow illumination mechanisms have been reported. Among these, chemiluminescence is associated with a chemical reaction that emits light.⁶⁰ However, we conducted our experiments in ambient conditions, where such a luminescence response is not likely to occur. For example, other studies reported the reaction of TMDs with ambient oxidation but did not observe any chemiluminescence.⁴⁷ Electroluminescence results in illumination due to interactions between the material and an electric field or charge. However, electroluminescence can be excluded for two main reasons. First, we did not apply an external electrical field to the TMDs. Second, the optical charging experiments (Figure 8) clearly indicate that the energy source of the afterglow response originated from optical laser charging. Delayed fluorescence afterglow is also characterized by long-lived electron-induced emission due to reverse intersystem crossing and light emission of singlets. In this reverse intersystem crossing process, electrons from the triplet band gain energy and return to the singlet band, from which they return to the ground state.⁶¹ Prolonged fluorescence signal emission is further enhanced when charge is trapped in lattice defects (e.g., in vacancies⁶⁰). Notably, the energy of the emitted light in such cases is derived directly from the material energy bandgap. We show, in Figures 11B–11E, the fluorescence and phosphorescence peak emission energies of WS₂, MoS₂, WSe₂, and Mo_1–xW_xS₂ alloys, respectively, and the difference between the light emission energies when the laser is turned on (i.e., fluorescence) and off (i.e., phosphorescence) (Figures 11F–11I). This energy difference represents the difference between the singlet and triplet emissions. Importantly, our samples demonstrated that the fluorescence peak energy (i.e., the emitted energy under laser illumination) is higher than the emitted afterglow peak energy, namely, the afterglow emission presents lower energy than the fluorescence emission, and, therefore, the afterglow emission is involved in some energy loss. We attribute energy reduction to the intersystem crossing process (Figure 1A). Consequently, we can also exclude delayed fluorescence singlet emission in our experiments. Additionally, previous studies indicated that the typical lifetime of singlets is in the order of nano-seconds,⁶¹ which is significantly shorter than the lifetimes presented here (hundreds and thousands of seconds) and that correspond to those of triplets.

We note that, due to the low thermal capacity of our samples, illuminating them also increased their temperature and stronger or longer illumination resulted in significant heating. In general, this effect has a broad impact on the optical behavior of TMDs due to reshaping of the band structure and possible adsorption of ambient molecules, as described elsewhere.⁴⁷ However, we illuminated the samples only for tens of seconds (Figure 8), such that these thermal effects are minor. Additionally, temperature changes the dynamics of free and bound excitons, which alter the photoluminescence signal.⁶¹ We illuminated the monolayer WS₂ shown in Figure 12A under varying intensities to understand the relationship between the intensity and afterglow. The fluorescence signals for different illumination intensities are shown in Figure 12B. We present the afterglow illumination intensities in Figure 12C (measured 1 min after we turned off the laser), demonstrating that significant illumination is required for a high-intensity afterglow. Indeed, stronger illumination resulted in a longer and stronger afterglow signal (inset of Figure 12C). Accordingly, low-intensity illumination induced a short afterglow lifetime and a low-intensity emission. For example, for a very low-intensity illumination, we were able to capture only one afterglow data point, after which the signal was lower than the noise level or completely decayed, as in the case of 10 mW (light blue data point) or 1 mW (we did not capture afterglow response) in Figure 12C. Notably, the sample investigated in Figure 12 demonstrated a lower response compared to previous samples, as the other samples showed a more significant afterglow signal when illuminated under 10 mW (see above). On the other hand, illumination that is too strong (above 70 mW, and in some cases, even above 35 mW) can result in undesired damage to the sample, Figure 12D.

Figure 12.

Measurement of varying illumination intensities performed on monolayer WS₂

(A) Optical image of the investigated WS₂ (the monolayer is purple and the multilayer is light blue). Scale bar: 10 μm.

(B) Fluorescence emissions for different illumination intensities.

(C) Phosphorescence peak intensities, as a function of time, for varying illumination intensities. Error bars represent measurement uncertainty. Dashed lines: exponential decay fits. Inset: phosphorescence emission intensity acquired 1 min after turning off the laser.

(D) Optical image of a defective WS₂ exposed to high-intensity illumination (>35 mW; the defective area is emphasized by arrows). Scale bar: 10 μm. Inset: dark-field enlargement of the defective area.

In Figure 13, we show the log-log graph of the integrated emission intensity as a function of time for WS₂, MoS₂, and WSe₂ samples. These graphs are in good agreement with an exponential time decay. A previous study indicated that such decay is characteristic of phosphorescence light emission, while delayed fluorescence demonstrates a different trend.⁵⁹ Taken together the possible reported existence of room temperature triplets, the strong spin-orbit coupling, the observed heavy-atom effect, the optical charging, the exponential time decay, and the enhanced non-radiative quenching that we observed in grain boundaries and after exposing samples to humidity, indicate that the afterglow of TMDs results from a phosphorescence response.

Figure 13.

Integrated emitted intensities

(A–D) Log-log graphs of integrated emission intensity as a function of time for monolayer WS₂ (A), multilayer WS₂ (B), MoS₂ (C), and WSe₂ (D).

To conclude, we elucidate the physics underlying TMD afterglow behavior. Among the studied materials, we found that WS₂ and WSe₂ showed unusually long-lived phosphorescence response, while MoS₂ and Mo_1–xW_xS₂ alloys showed an afterglow emission of several minutes. These materials showed an optical charging time of ∼20 s and the capability to emit energy of up to sub-millijoules. Phosphorescent materials can potentially be key components in a wide range of future applications, such as high-end chemo-optical sensors and safety systems. This work, therefore, will allow the integration of TMDs into such applications and pave the way toward new fields of science, such as atomically thick energy-storage systems and display devices.

Limitations of the study

The afterglow response of TMDs depends on the defect distribution and the interaction with the supporting substrate. Therefore, samples fabricated under different conditions may exhibit a different afterglow signal.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Assaf Ya’akobovitz (assafyaa@bgu.ac.il).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •All data reported in this paper will be shared by the lead contact upon request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We thank Nathan Samuel Shultzman for his advice, Dr. Avi Niv for the fruitful discussions, Dr. Natalya Froumin for her assistance with XPS measurements, and Roxana Golan for her assistance with EDS measurements. The authors also acknowledge the support of the Israeli Science Foundation (ISF), grant no. 226/25.

Author contributions

Conceptualization, G.A., Y.M.J., and A.Y.; methodology, G.A. and A.Y.; investigation, G.A. and A.Y.; writing – original draft, G.A. and A.Y.; writing – review and editing, G.A. and A.Y.; funding acquisition, A.Y.; resources, A.Y.; supervision, A.Y.

Declaration of interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

The authors did not use AI tools in the current publication.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
Sulfur powder	Sigma-Aldrich Products	www.sigmaaldrich.com/
MoO3 and WO3 powders	Sigma-Aldrich Products	www.sigmaaldrich.com/
Perylene-3,4,9,10-tetracarboxylic acid tetrapotassium salt (PTAS) growth promoter	2D Semiconductors	https://2dsemiconductors.com/
Software and algorithms
LabSpec6	Horiba	www.horiba.com/int/
Other
Nitrogen gas (High purity)	Maxima	maxima.co.il/home-page-en/

Method details

CVD growth

We used a three-zone tube furnace (Thermocraft) to conduct the chemical vapor deposition (CVD) processes. For MoS₂ growth, we placed ∼30 mg of sulfur and ∼3.5 mg of MoO₃ in two separate boats and set the temperature at zone 1 to 650°C–700°C, at zone 2 to 30°C, and at zone 3 to 90°C–70°C. For WS₂ growth, we used ∼50 mg of sulfur and ∼10 mg of WO₃ and set the temperature at zone 1 to 700°C–750°C, at zone 2 to 30°C, and at zone 3 to 70°C–50°C. We applied inert gases (nitrogen or argon) at a volumetric flow rate of 100 cm³ min⁻¹. In both processes, we set the system pressure to 500 mbar. We used two small Si/SiO₂ wafer pieces and, on one of them (the growth promoter), we dropped a PTAS catalyst enhancer to promote growth on the other wafer. Then, we placed both wafers upside down on top of a boat that contained MoO₃/WO₃ powder and positioned it in zone 1. In the other boat, we placed the sulfur powder and positioned it in zone 3. The distance between the boats in the tube was approximately 28 cm. The growth duration was 15–30 min. For Mo_1-xW_xS₂ growth, we placed a wafer with CVD-grown WS₂ in the CVD furnace and applied the same conditions used to grow MoS₂, such that we induced the nucleation-diffusion process and yielded alloys.⁴⁵ Finally, we cleaned the tubes inside the furnace by setting a high temperature in all zones and applying the inert gas, at the same volumetric flow rate as during the growth, for ∼2 h under a pressure of 100 mbar.

Raman, photoluminescence, and phosphorescence acquisition

We used an XploRA PLUS Horiba Raman microscope with a 532 nm laser to acquire Raman, photoluminescence (PL), and phosphorescence spectra of the samples. We used a ×100 objective with a laser intensity of 10 mW and varied the exposure time. We used gratings of 1800 grooves⋅mm⁻¹ for the Raman measurements and 600 grooves⋅mm⁻¹ for the PL and phosphorescence measurements.

Light-sensor calibration

We calibrated our optical spectrometer to convert its photon counts to energy. We obtained a PL response of the Si substrate with an exposure time of 1 s and a power of 10 mW. Notably, integrating this power over the exposure time yielded the applied energy during the Si illumination. Knowing the optical absorbance of Si under this wavelength illumination (∼0.56%), we evaluated the energy absorbed and emitted by the Si.⁶² Additionally, we measured the PL emission of Si and fitted the acquired curve to a Lorentzian function. Then, we integrated the fitted spectral signal, thereby summing the photon counts emitted by the Si substrate. By correlating the emitted energy and photon counts, we calibrated the optical spectrometer, enabling us to convert its reading to energy units.

Phosphorescence emission experiments

We started our experiments by first conducting fluorescence measurements, in which we shortly exposed the TMDs to a 532 nm laser with a power of 10 mW and acquired the emitted signal. Then, we optically charged the TMDs via their illumination using the same laser. The illumination duration was several tens of seconds to ensure that we exceeded the minimum duration required for their full charging (see above). After the optical charging, we turned off the laser and captured the afterglow signal. Each afterglow spectrum was taken after the photons accumulated for 30–90 s, depending on the signal. We stopped the measurement when the signal could no longer be detected (i.e., dropped below the noise level) and evaluated the afterglow lifetime based on this measurement.

X-ray photoelectron spectroscopy and energy-dispersive X-ray spectroscopy

We used an ESCALAB Xi+ instrument for the X-ray photoelectron spectroscopy (XPS) measurements and examined the atomic composition of the pure flakes. The minimum acquisition area of the system was limited to a 300 μm square. For the compositional analysis of specific flakes of TMD alloys (i.e., acquisition of the x value of Mo_1-xW_xS₂ flakes), we used an energy-dispersive X-ray spectroscopy (EDS) analysis in a Phenom XL G2 tabletop scanning electron microscope (SEM).

Quantification and statistical analysis

We did not perform any statistical analysis to filter out relevant data.