Observation of Long-Lived Interlayer Excitons in Monolayer MoSe2-WSe2 Heterostructures

Two-dimensional (2D) materials, such as graphene1, boron nitride2, and transition metal dichalcogenides (TMDs)3-5, have sparked wide interest in both device physics and technological applications at the atomic monolayer limit. These 2D monolayers can be stacked together with precise control to form novel van der Waals heterostructures for new functionalities2,6-9. One highly coveted but yet to be realized heterostructure is that of differing monolayer TMDs with type II band alignment10-12. Their application potential hinges on the fabrication, understanding, and control of bonded monolayers, with bound electrons and holes localized in individual monolayers, i.e. interlayer excitons. Here, we report the first observation of interlayer excitons in monolayer MoSe2-WSe2 heterostructures by both photoluminescence and photoluminescence excitation spectroscopy. The energy and luminescence intensity of interlayer excitons are highly tunable by an applied vertical gate voltage, implying electrical control of the heterojunction band-alignment. Using time resolved photoluminescence, we find that the interlayer exciton is long-lived with a lifetime of about 1.8 ns, an order of magnitude longer than intralayer excitons13-16. Our work demonstrates the ability to optically pump interlayer electric polarization and provokes the immediate exploration of interlayer excitons for condensation phenomena, as well as new applications in 2D light-emitting diodes, lasers, and photovoltaic devices.


Main Text
The recently developed ability to vertically assemble different 2D materials heralds a new realm of device physics based on van der Waals heterostructures 7 . The most successful example is the vertical integration of graphene on boron nitride. Such novel heterostructures not only dramatically enhance graphene's electronic properties 2 , but also give rise to super-lattice structures demonstrating exotic physical phenomena 6,[8][9] . A fascinating counterpart to gapless graphene is a class of monolayer direct bandgap semiconductors, namely transition metal dichalcogenides (TMDs) [3][4][5] . Due to the large binding energy in these 2D semiconductors, excitons dominate the optical response, exhibiting strong light-matter interactions which are electrically tunable [17][18] . The discovery of excitonic valley physics [19][20][21][22][23] and strongly coupled spin and pseudospin physics [24][25] in 2D TMDs opens up new possibilities for device concepts not possible in other material systems.
Monolayer semiconductor TMDs have the chemical formula MX2 where the M is tungsten (W) or molybdenum (Mo), and the X is sulfur (S) or selenium (Se). Although these TMDs share the same crystalline structure, their physical properties, such as bandgap, exciton resonance, and spin-orbit coupling strengths, can vary significantly. Therefore, an intriguing possibility is to stack different TMD monolayers on top of one another to form 2D heterostructures. Recent theoretical works have predicted type II band-alignment of such monolayer TMD heterojunctions 10-12 , where the conduction band minimum and the valence band maximum are located in different layers. Remarkably, the Coulomb binding energy in 2D TMDs is much stronger than in conventional semiconductors, making it possible to realize interlayer excitonic states in van der Waals bound hetero-bilayers with type II band-alignment. These interlayer excitons are composed of bound electrons and holes that are localized in different layers. Such interlayer excitons have been intensely pursued in bilayer graphene for possible exciton condensation 26    Comparison of the three spectra shows that both intralayer and − exist in the heterostructure with emission at the same energy as from isolated monolayers. This clearly demonstrates the preservation of intralayer excitons in the heterostructure region. The new spectral feature (XI) is more pronounced at low temperature, having an intensity comparable to the intralayer excitons.
The PL spectra imply type II band-alignment for the WSe2-MoSe2 heterojunction, consistent with several independent theoretical predictions 10-12 . This four-level band structure, as shown in The observation of bright interlayer excitons in monolayer semiconducting heterostructures is of central importance and the rest of this paper will focus on their physical properties.
Our PL measurements show that the intralayer exciton emission at the heterostructure region is quenched compared to isolated monolayers. This observation implies that interlayer charge transfer is fast compared to the intralayer exciton recombination rate. Importantly, there is a distinct difference in the quenching ratio at room temperature and low temperature. At room temperature, the PL of intralayer excitons XMo and XW is quenched by at least an order of magnitude ( Figure S2), while at low temperature the intralayer exciton PL is only slightly quenched (Fig. 1d). This implies that the interlayer carrier hopping rate strongly depends on the temperature and is reduced at lower temperature.
Moreover, at low temperature (20 K), we find the spectrally integrated exciton PL intensity is conserved between isolated monolayers and the heterojunction, while at room temperature the integrated PL intensity from the heterojunction is an order of magnitude smaller than the summation of isolated monolayers. This observation implies that at low temperature the quenched population of intralayer excitons is transferred to form interlayer excitons, which then relax mainly via radiative recombination. In contrast, at room temperature, a non-radiative relaxation channel dominates the interlayer exciton relaxation, giving rise to the quenching of the spectrally integrated heterojunction PL intensity.
To investigate the coupling between the interlayer and intralayer excitons, we perform PL excitation spectroscopy (PLE). A narrow bandwidth (<50 kHz) tunable laser is swept in energy from 1.6 eV to 1.75 eV (across the resonance of intralayer excitons) while monitoring interlayer exciton PL. Figure 2b shows an intensity plot of XI emission as a function of photo-excitation energy from Device 2. We clearly observe interlayer emission enhancement when the excitation energy is resonant with intralayer exciton states (Fig. 2c). We attribute the PLE resonances to the excitation of intralayer excitons, which then relax through fast interlayer charge transfer to form the energetically favorable interlayer excitons (Fig. 2a).
Furthermore, the emission energy of the interlayer exciton is highly electrically tunable.  We attribute the doublet feature of the interlayer exciton to the spin-splitting of the MoSe2 conduction band [28][29] (Fig. 4b), which is supported by the evolution of the relative strength of the two peaks with increasing excitation power, as shown in Fig. 4a Figure 4d. A fit to a single exponential decay yields an interlayer exciton lifetime of about 1.8 ns. This time scale is much longer than the intralayer exciton lifetime, which is on the order of tens of ps [13][14][15][16] .
In summary, our results demonstrate the feasibility of band engineering by creating monolayer TMD van der Waals bound heterostructures. Photoluminescence spectroscopy of interlayer and intralayer excitons with electrical control reveals the type II band-alignment of the heterojunction, crucial for many optoelectronic applications. The interlayer exciton in these heterostructures is an analog of the spatially indirect exciton in GaAs double well structures which have been intensely studied for exciton BEC phenomena 26,30 . Here, we have identified spatially indirect interlayer excitons displaying extended lifetimes and repulsive interactions, both key ingredients for the realization of exciton BEC. The long-lived interlayer exciton may also lead to population inversion, a critical step toward realizing 2D heterostructure lasers.
During the final preparation of the paper, we became aware of relevant work studying the photovoltaic properties [31][32][33] and interlayer coupling 34 of heterostructures.      Figure 3.