Introduction

Monolayer semiconducting transition metal dichalcogenides (TMDs) have recently emerged as a promising system in fundamental physics and technology-related studies of electronics, optoelectronics, valleytronics, and twistronics. Their prominent properties are enabled by the weak dielectric screening, breaking of the inversion symmetry, and strong spin-orbit coupling1,2,3,4. Despite being direct bandgap semiconductors, most monolayer TMDs exhibit poor photoluminescence (PL) yield due to the low absolute optical absorption of <10% as well as the low external quantum yield of <1%, which limits the development of nanophotonic devices such as light-emitting diodes (LEDs), lasers, display devices, and optical on-chip networks5,6,7.

One prospective way to achieve a homogeneous enhancement of emission intensity is to deliver the excitation energy from the donor to the acceptor by nonradiative energy transfer. Conventional energy transfer comprises Förster resonance energy transfer (FRET) and Dexter-type energy transfer (DET); the former is based on dipole-dipole coupling (where the dipoles in the semiconductors are excitons), while the latter is based on charge transfer and spin conservation8,9,10,11. To evaluate the effect of energy transfer, an enhancement factor (η) is usually defined as \(\eta={I}_{{{{{{\rm{DA}}}}}}}/{I}_{{{{{{\rm{A}}}}}}}\), where IDA and IA are the PL intensities of the acceptor in the heterojunction and on the bare substrate, respectively. For both FRET and DET in 2D semiconductors, η is limited to ~2 to date8,10,12,13,14. High-performance nanophotonic devices call for a new structure or mechanism for energy transfer with a colossal η to brighten 2D semiconductors.

Alternatively, for organic-semiconductor heterojunctions with Frenkel excitons, exciton-photon polaritons due to the strong coupling of excitons with microcavity provide a powerful platform to achieve unconventional energy transfer by polariton relaxation15,16,17,18,19,20. In contrast, TMDs feature a more delocalized Wannier-Mott exciton nature21,22. Their polaritons inherit the perfect in-plane dipole orientation, strong quantum confinement, and valley polarization, which provides more degree of freedoms to engineer new functionalities in the polariton field. Thus, TMDs polaritons have attracted great attention for the study of Bose-Einstein condensates23,24,25, nonlinear optical processes26,27, valley properties28,29,30, LEDs31 as well as Moiré heterojunctions32. However, it remains elusive to realize the strong coupling of TMDs heterojunctions with microcavity for unconventional energy transfer due to the considerable challenge to avoid the PL quenching by ultrafast charge transfer and meanwhile to achieve the mode match among donor exciton, acceptor exciton, and cavity photon.

Here, by embedding the hBN/MoS2/hBN/WS2 heterojunction into a Fabry-Pérot (FP) microcavity, we have realized the strong coupling among the donor exciton, acceptor exciton, and cavity photon mode, which leads to the brightening of MoS2 with a two-order-of-magnitude enhancement of PL yield (η ≈ 440) based on the exciton-photon polariton relaxation. The custom-designed k-space transient-reflectivity spectroscopy, which features a better spectral resolution than the real-space measurement for microcavity samples, was applied to evaluate the polariton relaxation dynamics, and a characteristic time as short as ~1.3 ps was extracted. The efficient and ultrafast polariton relaxation is associated with the significantly enhanced intra- and inter-branch exciton-exciton scattering, which overcomes the hot phonon bottleneck effect. Moreover, the phase diagram has been established to correlate the polariton relaxation efficiency with the Rabi energies. This study demonstrates a new branch of microcavity-confined 2D semiconductor heterojunctions with a great potential for ultrafast highly-bright polaritonic light sources.

Results

Sample fabrication and optical characterization

For the optical microcavity-confined heterojunction, the strong coupling among the acceptor’s (Eex1) and donor’s (Eex2) excitons as well as the cavity photon mode (Ec) will form the exciton-photon polariton with three anti-crossing eigenstates denoted by the upper (UPB), middle (MPB), and lower (LPB) polariton branches (Fig. 1a)15,16,17,18,19. Under off-resonance optical excitation, the polariton tends to relax from the UPB to MPB and finally to LPB (i.e., from high energy and high in-plane momentum (k//) to low energy and low k//). The relaxation process is an unconventional energy transfer since the UPB and LPB are mostly Eex2-like and Eex1-like, respectively, as determined by the Hopfield coefficients15,16,17,18,19.

Fig. 1: Design, construction, and properties of FP microcavity-confined hBN/MoS2/hBN/WS2 heterojunction.
figure 1

a Schematic formation and relaxation of exciton-photon polariton. The black and red arrows in the right panel represent intra- and inter-branch scattering, respectively. b Schematic illustration of microcavity-confined hBN/MoS2/hBN/WS2 heterojunction. c Normalized real-space PL spectra of het@SiO2 (black) and het@cavity− (blue). The differential reflection spectrum of het@SiO2 (dashed red) is presented for reference. XA and XB are A and B exciton, respectively.

Full size image

The configuration of microcavity-confined heterojunction in this study is illustrated in Fig. 1b. The hBN/MoS2/hBN/WS2 heterojunction was fabricated by a pick-up method (Supplementary Fig. 1)32. WS2 (donor) and MoS2 (acceptor) are monolayers with a direct bandgap, while the top and middle hBN are 15 and 2 nm, respectively. The middle thin hBN layer prevents the ultrafast interlayer charge transfer from nonradiative recombination while keeping the FRET almost intact13,33,34. Then, the heterojunction was transferred onto a bottom-distributed Bragg reflector (DBR) composed of 6.5 pairs of SiO2 and TiO2 dielectric layers, with its photonic stopband covering the heterojunction exciton energies (Supplementary Fig. 2). Subsequently, a poly(methyl methacrylate) (PMMA) spacer and a silver mirror were deposited to finalize the microcavity device. The thickness of PMMA spacer was adjusted to achieve the negative (positive) acceptor exciton-photon detuning, with the microcavity heterojunction labeled as het@cavity− (het@cavity+), respectively. Supplementary Table 1 summarizes the thickness information of het@cavity−, which ensures an energy match between the cavity mode and heterojunction excitons, i.e., the heterojunction placed in the maximum of the electromagnetic field. For quantitative comparison, an identical heterojunction was fabricated on a SiO2(90 nm)-coated Si substrate to avoid the cavity effect, defined as het@SiO2 (Supplementary Fig. 3a).

Figure 1c compares the normalized real-space PL spectra of het@cavity− and het@SiO2, with the differential reflection spectrum of het@SiO2 as a reference. The real-space PL spectrum of het@SiO2 features two emission peaks, i.e., A exciton (XA) emission from WS2 at 2002 meV and MoS2 at 1890 meV, with a Stokes shift of ~10 meV relative to the corresponding reflectivity spectrum. The PL intensity of WS2 XA is much stronger than that of MoS2 XA, which results from the higher intrinsic exciton dipole oscillator strength (f) and quantum yield of the former, and the interlayer FRET (η ≈ 3), as detailed in Supplementary Fig. 3. On the contrary, the PL spectrum of het@cavity− is only composed of a single peak at 1873 meV, with the even smaller energy than the PL of MoS2 XA. The significant change of PL spectra, in both shape and energy, points out the crucial effect of the optical microcavity confinement on energy transfer.

The energy transfer in het@cavity− is further understood by means of the k-space energy-resolved reflectivity and PL mappings, in comparison with the cases of microcavity-confined hBN/MoS2/hBN (MoS2@cavity) and hBN/WS2/hBN (WS2@cavity), as summarized in Fig. 2a–f. The polariton dispersions for het@cavity− are fitted with a three coupled oscillator model15,16,17, i.e.,

$$\left(\begin{array}{ccc}{E}_{c}\left(\theta \right) & {\Omega }_{2}/2 & {\Omega }_{1}/2\\ {\Omega }_{2}/2 & {E}_{{ex}2} & 0\\ {\Omega }_{1}/2 & 0 & {E}_{{ex}1}\end{array}\right)\left(\begin{array}{c}\begin{array}{c}{\alpha }_{c}\\ {\alpha }_{{ex}2}\end{array}\\ {\alpha }_{{ex}1}\end{array}\right)={E}_{{pol}}\left(\theta \right)\left(\begin{array}{c}\begin{array}{c}{\alpha }_{c}\\ {\alpha }_{{ex}2}\end{array}\\ {\alpha }_{{ex}1}\end{array}\right)$$
(1)

where \({{E}_{c}\left(k \right)={E}_{0}(1-{\sin }^{2}(\theta )/{n}^{2})}^{-1/2}\) is the cavity photon mode energy, \(\theta\) is the angle of incidence, E0 is the cavity cut off energy, and n is the cavity effective refractive index. Eex1, Eex2 and Epol(θ) are the energies of acceptor exciton, donor exciton, and polariton, respectively. Ω1 and Ω2 are the corresponding Rabi splitting energies. αex1, αex2 and αc are the corresponding Hopfield coefficients (Supplementary Fig. 4). The angle dependence of the three eigenvalues corresponds to three dispersion branches termed UPB, MPB, and LPB, respectively (the white dotted lines in Fig. 2a, b). Similarly, the polariton dispersions for MoS2@cavity and WS2@cavity are fitted with a two-coupled oscillator model, which gives two branches, termed UPB and LPB (see Method).

Fig. 2: Comparison of polariton relaxations in het@cavity−, MoS2@cavity and WS2@cavity.
figure 2

a, b k-space energy-resolved reflectivity mapping (a), PL mapping (b) for het@cavity − . c, d The corresponding results for MoS2@cavity. e, f The corresponding results for WS2@cavity. The Rabi splitting (Ω) and exciton-photon detuning (Δ) energies are labeled in the unit of meV. Note: data range of the color bar in (d) is much smaller than that in (b) and (f). g, h The real-space PL spectra of het@cavity− and MoS2@cavity in linear (g), and log (h) scale under 10 μW excitation.

Full size image

The reflectivity mappings of MoS2@cavity and WS2@cavity show two anti-crossing branches with Rabi splitting energy (Ω) of 18 and 41 meV, respectively (Fig. 2c, e), consistent with the smaller exciton dipole oscillator strength of MoS2 than WS2 (\(\Omega \propto \sqrt{f}\), Supplementary Fig. 3b)35,36. In contrast, het@cavity− features three branches with two anti-crossing points, demonstrating the strong coupling among MoS2 exciton, WS2 exciton, and cavity photon mode, with Ω1 (Ω2) ≈ 26 (26) meV and a negative (negative) exciton-photon detuning (Δ) of −15 (−150) meV for MoS2 (WS2) (Fig. 2a). Ω1 of het@cavity− is larger than Ω of MoS2@cavity and Ω2 of het@cavity− is smaller than Ω of WS2@cavity, which straightforwardly indicates the transfer of dipole oscillator strength from Eex2 to Eex137. A similar transfer is also reflected in het@cavity+ with a Ω1 (Ω2) ≈ 42 (24) meV and a positive (negative) Δ of +13 (−132) meV for MoS2 (WS2) (Supplementary Fig. 5a). These results suggest a general feature of the transfer of dipole oscillator strength in the strongly coupled microcavity-confined heterojunctions, in sharp contrast to the negligible transfer in the cavity-free het@SiO2 counterpart (Supplementary Fig. 3b). The experimental reflectivity mappings are well-reproduced by theoretical calculation with the transfer matrix method (Fig. 2a, c, e, detailed in Supplementary part III)38,39,40.

The PL mapping of MoS2@cavity suffers from a weak polariton emission due to the weak dipole oscillator strength and low quantum yield of MoS2 (Fig. 2d)5, while that of WS2@cavity suffers from a hot phonon bottleneck effect (the broad emission centred at ~1.7 μm−1) due to the large negative Δ (−105 meV, Fig. 2f)29,41,42. On the contrary, the PL mapping of het@cavity− integrates the advantages of MoS2@cavity and WS2@cavity, i.e., inherits the narrow linewidth from MoS2@cavity and high intensity from WS2@cavity (Fig. 2b); so does the PL mapping of het@cavity+ (Supplementary Fig. 5b). The PL intensity of het@cavity− in real-space photoluminescence excitation spectrum increases monotonously without resonance peak when decreasing the excitation energy towards LPB (Supplementary Fig. 6). The preceding results confirm the energy transfer mechanism of polariton relaxation, rather than FRET or DET, in the strongly coupled microcavity-confined heterojunctions. In the linear-scale real-space PL spectrum, het@cavity− and MoS2@cavity show bright and negligible emission under 10 μW excitation, respectively (Fig. 2g). From Fig. 2b, it is learned that the polariton emission of het@cavity− comes from the LPB and MPB with |k//| < 1.6 μm−1, which is composed of MoS2 exciton and cavity photon but negligible WS2 exciton (Supplementary Fig. 4a). Hence, η is obtained to be ~440 from the ratio of \({I}_{{het@cavity}-}^{{LPB}+{MPB}}/{I}_{{{MoS}}_{2}{@cavity}}^{{LPB}+{UPB}}\). Here \({I}_{{het@cavity}-}^{{LPB}+{MPB}}\) and \({I}_{{{MoS}}_{2}{@cavity}}^{{LPB}+{UPB}}\) are the integrated polariton emission intensity from LPB + MPB (red+cyan region) of het@cavity− and LPB + UPB of MoS2@cavity (black-outlined region) in real-space, respectively (Fig. 2h, detailed in Supplementary Note 1). As summarized in Supplementary Table 2, the giant η of 440 for het@cavity− is over 146 times larger than the η of 3 with FRET for het@SiO2 counterpart, also two orders of magnitude higher than the typical η of FRET or DET processes in literature, demonstrating a powerful strategy to significantly increase the energy transfer efficiency by polariton relaxation.

Supplementary Figs. 7 and 8 show the optical characteristics of the additional microcavity-confined heterojunction samples, specifically identified as het8@cavity (8 nm hBN), het2@node@cavity (heterojunction with 2 nm hBN at the node), and het60@cavity (60 nm hBN). These diverse configurations, characterized by different heterojunction’s positions within the cavity and/or different thicknesses of hBN, give rise to distinct Rabi energies Ω1 and Ω2. Consequently, the distinct Rabi energies lead to a substantial difference in polariton relaxation efficiency. This result demonstrates the crucial role of Rabi energies in polariton-mediated energy transfer, which has not been recognized to date and will be further theoretically clarified later.

Theoretical simulation

We simulated the k-space PL mappings with the coupled rate equations43,44,45,46, which describe the dynamics of polariton populations in different branches. The different points with the different relative excitonic and photonic fractions in \(k\)-space for each branch are identified by the coupled oscillator model and the Hopfield coefficients (Supplementary Fig. 4). The coupled rate equations for het@cavity− with three polariton branches read

$$\frac{d{N}_{j}^{\Xi }}{{dt}}=\left({P}_{1}{X}_{1,j}^{\Xi }+{P}_{2}{X}_{2,j}^{\Xi }-\gamma {C}_{j}^{\Xi }\right){N}_{j}^{\Xi }+{\sum }_{{\Xi }^{{\prime} }}{S}^{\Xi {\Xi }^{{\prime} }}$$
(2)

where \({N}_{j}^{\Xi }\) is the polariton population, \(\Xi=\left\{U,M,L\right\}\) corresponds to the UPB, MPB, and LPB, 1 (2) indicates MoS2 (WS2), and index \(j\) denotes the respective point in \(k\)-space. \({P}_{1(2)}\) stands for an effective pumping for the exciton, and \(\gamma\) for the decay rate of the photon mode. Here, \({P}_{2} > {P}_{1}\) due to the much brighter polariton emission of WS2@cavity than MoS2@cavity (Fig. 2d, f). The \({X}_{1(2)}^{\Xi }\) and \({C}^{\Xi }\) denote \({|{\alpha }_{1(2)}|}^{2}\) and \({\left|{\alpha }_{c}\right|}^{2}\) (Supplementary Fig. 4). The relaxation processes, described by the last term in (2), consist of the intra- and inter-branch transitions originating from the phonon-mediated exciton scattering47,48. In this formalism, the preferred transition with a high rate happens between initial and final destinations with high \({X}_{1}^{\Xi }{X}_{1}^{{\Xi }^{{\prime} }},{X}_{2}^{\Xi }{X}_{2}^{{\Xi }^{{\prime} }}\) and phonon population \({n}_{{ph}}({E}_{{ph}})\). According to Bose distribution function, i.e., \({n}_{{ph}}=\frac{1}{\exp \left(\frac{{E}_{{ph}}}{{k}_{B}T}\right)-1}\) (here \({E}_{{ph}}=|{E}^{{\Xi }^{{\prime} }}-{E}^{\Xi }|\)), \({n}_{{ph}}\) is large for transitions with close energies. The photonic part of the numerical steady-state solution for \({N}_{j}^{\Xi }\), together with an added Gaussian broadening, agrees with our experimental result (Fig. 2b, d, f; Supplementary part III for details).

Specifically, for MoS2@cavity, P1 excites the population in the low- (high-) k// region of UPB (LPB) with high exciton fractions \({X}_{1}^{U}({X}_{1}^{L})\) (Supplementary Fig. 4b). The population will relax by the inter-branch scattering (from UPB to LPB) due to the high \({X}_{1}^{U}{X}_{1}^{L}\) at the anti-crossing point and the high \({n}_{{ph}}\) (Eph ≈ Ω ≈ 18 meV), followed by the intra-branch scattering (within LPB) due to the non-negligible \({{X}_{1}^{L}X}_{1}^{{L}^{{\prime} }}\) and the high \({n}_{{ph}}\) (flat dispersion)47,48. However, the polariton emission is considerably weak due to the weak oscillator strength and low quantum yield of MoS2 exciton (Supplementary Fig. 3). For WS2@cavity, P2 also excites the population in the low- (high-) k// region of UPB (LPB) with high exciton fractions \({X}_{2}^{U}\) (\({X}_{2}^{L}\)) (Supplementary Fig. 4c). However, only a tiny portion of the population will relax to the bottom of LPB due to the poor inter-branch scattering for the low \({n}_{{ph}}\) (Eph ≈ Ω ≈ 41 meV) and the poor intra-branch scattering for the negligible \({{X}_{2}^{L}X}_{2}^{{L}^{{\prime} }}\), presenting the troublesome hot phonon bottleneck (Fig. 2f)29,41.

Based on the above experimental and theoretical results, the polariton relaxation in het@cavity− could be well elucidated (Fig. 2b). Due to \({P}_{2}\, > \,{P}_{1}\), P2 excites the most population in the k// < 2.5 (>2.5) μm−1 region of UPB (MPB) with high \({X}_{2}^{U}\) (\({X}_{2}^{M}\)) (Supplementary Fig. 4a). The population will relax by the inter-branch scattering (from UPB to MPB) due to the high \({X}_{2}^{U}{X}_{2}^{M}\) at the top anti-crossing point and the high \({n}_{{ph}}\) (Eph ≈ Ω2 ≈ 26 meV). The subsequent intra-branch scattering within MPB is faster than that within LPB for WS2@cavity due to the non-negligible \({{X}_{2}^{M}X}_{2}^{{M}^{{\prime} }}\) and \({{X}_{1}^{M}X}_{1}^{{M}^{{\prime} }}\) for the former compared to the negligible \({{X}_{2}^{L}X}_{2}^{{L}^{{\prime} }}\) for the latter. The synergism of the two scatterings leads to a fast relaxation of almost all the population to the bottom anti-crossing point. The population will further relax towards the bottom of LPB in a way similar to that in MoS2@cavity, as reflected in Fig. 1a. Such a fast polariton relaxation dynamics leads to the bright PL of het@cavity− at the bottom of LPB (Fig. 2b).

Dynamics and phase diagram

With a custom-designed microscopic k-space transient-reflectivity spectroscopy (see Methods), we have simultaneously unravelled the evolution of the polariton population in momentum, energy and time domains. After being pumped by a 200-fs laser pulse of ~2 μW at 580 nm, the change of reflectivity \((\frac{\triangle R}{R}=\frac{{R}_{{{{{{\rm{pump\; on}}}}}}}-{R}_{{{{{{\rm{pump\; off}}}}}}}}{{{{{{{\rm{R}}}}}}}_{{{{{{\rm{pump\; off}}}}}}}})\) of the broadband white light is measured after a specific time delay, where \({R}_{{{{{{\rm{pump\; on}}}}}}}\) and \({R}_{{{{{{\rm{pump\; off}}}}}}}\) are the reflectivity mappings with or without optical pumping, respectively. Such an excitation fluence is low enough to avoid exciton-exciton annihilation49,50. The k-space transient-reflectivity spectroscopy results of WS2@cavity and het@cavity− are shown in Fig. 3, with the corresponding movies in Supplementary Material. For the WS2@cavity, the laser pulse excites the exciton reservoir to quickly form polaritons, which induces a net photobleaching signal at 0 ps at k// ≈ 2.3 μm−1 (Fig. 3a), also demonstrated in the integrated \(\triangle R/R\) at k// from 2.1 to 2.7 μm−1 (Supplementary Fig. 9). Then, a derivative signal emerges at k// ≈ 0 μm−1 region, as clarified by the integrated \(\triangle R/R\) at k// = 0 ~ 0.33 μm−1 (Fig. 3a, b and the movie in Supplementary Material). Such a signal represents the blue-shift of LPB due to the polariton-polariton repulsive interaction or the phase-space-filling effect of exciton (Supplementary Figs. 10, 11)51. The signal intensity gradually increases, reaches the maximum at ~6 ps, and slowly decays afterward (Fig. 3a). By fitting the evolution of the derivative signal with the rising function (y = A + B erf (t/τ), where \({{{{{\rm{erf}}}}}}(z)=\frac{2}{\sqrt{\pi }}{\int }_{0}^{z}{e}^{-{t}^{2}}{dt}\)), a characteristic rising time (τ) of ~2.8 ± 0.4 ps is obtained (Fig. 3c)52. In contrast, for het@cavity−, the intensity of the derivative signal at k// ≈ 0 μm−1 region increases much faster and reaches the maximum at ~2 ps (Fig. 3d, e). With the same fitting method, a much shorter τ of ~1.3 ± 0.2 ps is obtained (Fig. 3f), which approximates to the τ of 1.2 ± 0.2 ps in MoS2@cavity (Supplementary Fig. 12). Therefore, het@cavity− demonstrates the ultrafast energy transfer processes accompanied by the highest enhancement factor (440) to date (Supplementary Table 2).

Fig. 3: Time-resolved polariton relaxation dynamics.
figure 3

a–c k-space transient-reflectivity spectroscopy mapping (a), the integrated \(\triangle R/R\) in 0 ~ 0.33 μm−1 at typical time delays (b), and the intensity evolution of the derivative signal (c) of WS2@cavity. d–f The corresponding data of het@cavity−. Note: The steady-state polariton branches are shown in dashed black to guide the reading. The error bars in (c) and (f) correspond to the standard error of the data points.

Full size image

Based on the preceding qualitative and quantitative analysis, we have formulated the phase diagram that correlates the polariton relaxation efficiency and characteristic rising time with the Rabi energies Ω1 and Ω2, as shown in Fig. 4 (see Supplementary part III for details). Here we define a measure of efficiency as the sum of the photonic part of LPB population in the steady-state, i.e., \({{{{{\mathscr{N}}}}}}\left({\tau }_{s}\right)\equiv \sum {N}_{j}^{L}\left({\tau }_{s}\right){C}_{j}^{L}\) only for low-k// points (\({\tau }_{s}\) denotes the time at steady state). In general, \({{{{{\mathscr{N}}}}}}\left({\tau }_{s}\right)\) monotonously increases with decreasing \({\Omega }_{2}\) (except for \({\Omega }_{2}\to 0\)) due to the enhanced interbranch scattering between UPB and MPB. For a specific \({\Omega }_{2}\), \({{{{{\mathscr{N}}}}}}\left({\tau }_{s}\right)\) depends on the trade-off between the intra-branch scattering in MPB and the inter-branch scattering (MPB to LPB), hence a suitable \({\Omega }_{1}\) is needed for the optimal \({{{{{\mathscr{N}}}}}}\left({\tau }_{s}\right)\) as marked by the dashed-dotted curve (Fig. 4a). For the \({{{{{\mathscr{N}}}}}}\left({\tau }_{s}\right)\) of specific cases, the het@cavity − , het@cavity+ and het60@cavity feature high efficiency. In contrast, \({{{{{\mathscr{N}}}}}}\left({\tau }_{s}\right)\) is low for \({\Omega }_{2}\to 0\) or \({\Omega }_{1}\to 0\) (such as MoS2@cavity, WS2@cavity, and het2@node@cavity), or Ω1 and Ω2 are both large (such as het8@cavity), in agreement with the experimental results. Such an agreement offers a deep insight into the polariton-mediated energy transfer, which is inherently dominated by the Rabi energies.

Fig. 4: Phase diagram of the polariton relaxation dynamics versus Rabi energies.
figure 4

a, b Polariton relaxation efficiency (a) and characteristic rising time (b) with respect to Rabi energies Ω1 and Ω2. The seven specific experimental cases are marked. In a, the dashed-dotted curve indicates the optimal efficiency concerning \({\Omega }_{1}\) and \({\Omega }_{2}\), and the dashed blue boxes show regions where \({\Omega }_{1},{\Omega }_{2}\to 0\). Note: In the dark area of b, the population in the low-k// region of LPB does not rise initially.

Full size image

Different from the \({{{{{\mathscr{N}}}}}}\left({\tau }_{s}\right)\), the characteristic rising time is closely related to the scattering rate rather than the final polariton population in the low-k// region of LPB (Fig. 4b). After a Gaussian effective pump with amplitudes \(\{{P}_{1},{P}_{2}\}/50\gamma\), the evolution of the photonic part of the LPB population, i.e., \({{{{{\mathscr{N}}}}}}\left(t\right)\equiv \sum {N}_{j}^{L}\left(t\right){C}_{j}^{L}\), is fitted by the rising function. For (\({\Omega }_{1},{\Omega }_{2}\)) located in the dark region, the negligible polariton fraction in the low-k// region of LPB prevents \({{{{{\mathscr{N}}}}}}\left(t\right)\) from growing. For \({\Omega }_{1}\gtrsim 8\) meV, the rising time monotonously decreases with increasing \({\Omega }_{1}\), which can be attributed to the dominant intra-branch scattering for the flatter LPB dispersion. Although the larger \({\Omega }_{1}\) means the less transition from MPB to LPB, once the population relaxes to LPB, the strong intra-branch scattering leads to the faster relaxation to low-\(k\) points. It is worth noting that the faster rising time might not end up with a higher \({{{{{\mathscr{N}}}}}}\left({\tau }_{s}\right)\), as evidenced by the comparison in Fig. 4a and Fig. 4b. The dependence of the rising time on \({\Omega }_{2}\) is associated with the supply of population that will later end up in the low-k// region of LPB. Smaller \({\Omega }_{2}\) results in more population relaxation from UPB to MPB, which allows faster growth of \({{{{{\mathscr{N}}}}}}\left(t\right)\), as supported by the longer rising time of WS2@cavity than het@cavity− and MoS2@cavity (Fig. 4b), in agreement with the experimental result (Fig. 3).

Discussion

In summary, we have successfully designed and constructed a planar optical microcavity-confined MoS2/hBN/WS2 heterojunction, in which the middle insulating hBN prevents the charge transfer-induced PL quenching. Such a configuration realizes the strong coupling among donor exciton, acceptor exciton and cavity photon mode, which results in the unconventional energy transfer mechanism of polariton relaxation in 2D material heterojunctions for the first time. Consequently, we have brightened the MoS2 with the record-high energy transfer enhancement factor of ~440, which is two-order-of-magnitude higher than the data reported to date. A short characteristic time of ~1.3 ps is extracted for the ultrafast polariton relaxation dynamics, resulting from the significantly enhanced intra- and inter-branch exciton-exciton scattering to overcome the hot phonon bottleneck effect. The formulated phase diagram in this study establishes the correlation between the polariton relaxation dynamics and Rabi energies, which deepens the understanding on the underlying physics. This study drives the collaborative development of energy transfer and polariton fields toward the new topics of dark exciton energy transfer, valley-polarized energy transfer, and high-brightness ultrafast polaritonic light sources.

Methods

Sample fabrication

The bottom DBR was deposited on a sapphire substrate by using an e-beam evaporator (Cello, Ohmiker-50B), and was composed of 6.5 alternating pairs of titanium dioxide (TiO2, n = 2.498) and silicon dioxide (SiO2, n = 1.478). The hBN/MoS2/hBN/WS2 heterojunctions were fabricated using a dry-transfer method with a polypropylene carbonate (PPC) stamp. Monolayer WS2, monolayer MoS2, and thin hBN flakes were exfoliated onto silicon substrates with a 90-nm SiO2 layer. A PPC stamp was used to pick up the flakes in sequence by a home-built micro-transfer stage (Supplementary Fig. 1). Then, the heterojunction was released to the DBR substrate, immersed in acetone overnight to remove the PPC, and annealed in a high vacuum (<10−6 mbar) at 200 °C for 2 hours. The top PMMA layer was spin-coated and the top silver mirror was deposited by a thermal evaporator.

Two coupled oscillator model

Similar to het@cavity−, the polariton dispersion in MoS2@cavity and WS2@cavity can be fitted with two coupled oscillator model,

$$\left(\begin{array}{cc}{E}_{c}\left(\theta \right) & \Omega /2\\ \Omega /2 & {E}_{{ex}}\end{array}\right)\left(\begin{array}{c}{\alpha }_{c}\\ {\alpha }_{{ex}}\end{array}\right)={E}_{{pol}}(\theta )\left(\begin{array}{c}{\alpha }_{c}\\ {\alpha }_{{ex}}\end{array}\right)$$

where Eex is the energy of exciton, Ω the corresponding Rabi splitting energy. \({\left|{\alpha }_{c}\right|}^{2}\), \({\left|{\alpha }_{{ex}}\right|}^{2}\) are the Hopfield coefficients describing the photonic and excitonic weightings of the polaritons.

Optical characterization

The real-space steady-state PL and reflectivity spectra were conducted in a confocal spectrometer (Horiba Evolution 800) by using a CW laser (532 nm). K-space energy-resolved reflectivity and PL mapping were measured in a home-built setup with the Fourier imaging configuration with a high numerical aperture 100× microscope objective (NA = 0.9). The PL mappings are excited by a 532 nm continuous-wave (CW) laser of ~100 μW. The emission from the microcavity was collected through the narrow entrance slit of the spectrometer (Horiba iHR550) and finally onto the 2D charge-coupled device (CCD) array (Horiba, Symphony II). In the k-space transient-reflectivity spectroscopy measurement, the excitation pulsed laser was taken from a Ti:Sapphire laser equipped with an ultrafast amplifier (Spectra-Physics) and a computer-controlled Optical Parametric Amplifier (1 kHz repetition rate, with a roughly 200 fs pulse width). The output beam was split into two paths. One beam excited the sample almost at normal incidence to serve as the pump beam. The second beam went through a mechanical delay stage (Newport, M-ILS 150CC DC Servo Linear Stage) and a sapphire crystal to generate a delayed continuum light to serve as the probe pulse. The spot size of the pump and probe beams was around 2 μm and 3 μm, respectively. PLE spectra were obtained with a supercontinuum light acting as the excitation source which is coupled to a monochromator. The excitation intensity was kept below 10 μW. A 655 nm long-pass filter was used to cut off the excitation photon that backscattered from the sample. All the measurements were carried out in a reflection configuration.