Optical signatures of charge- and energy transfer in TMDC/TMDC and TMDC/perovskite heterostructures

While graphene and the prototypical monolayer TMDCs such as MoS₂ and WSe₂ have a similar honeycomb crystal structure, their optoelectronic properties are vastly different [17]. These properties are determined by the specifics of their electronic bandstructures, which differs significantly between graphene and the TMDCs. The difference roots in two fundamental properties: (i) The heavier transition metals in TMDCs induce strong spin–orbit coupling (SOC), and (ii) the specific two-atom system in the MX₂ (M: Mo, W, X: S, Se) structure breaks inversion symmetry. These fundamental properties of TMDCs result in the opening of a bandgap and the splitting of the K and K' valleys in the electronic bandstructure. Furthermore, a direct-to-indirect bandgap transition results upon stacking several TMDC layers due to specific orbital contributions of the metal and chalcogen atoms, see figure 1(b). Specifically, there are three valleys of interest in the electronic bandstructure of TMDCs: at the Γ, the K and the K' point in the Brillouin zone. A major distinction between the Γ and the K-valleys is in the aforementioned relative contributions of the metal and chalcogen orbitals, which cause the direct-to-indirect bandgap crossover due to the more out-of-plane p-orbital character of the bands at the Γ and Λ points [32–34]. With increasing number of layers, the states at the Γ point are strongly affected, as shown in figure 1(b). As a result, the prototypical MoS₂, MoSe₂, WS₂ and WSe₂ have room-temperature bulk electronic bandgaps of 1.2, 1.1, 1.4 and 1.2 eV, respectively, while their respective monolayers have 2.5, 2.3, 2.6 and 2.2 eV [29, 35]. However, the monolayer values are very sensitive to the measurement geometry and generally, it is more difficult to extract electronic band energies rather than optical bandgap energies, which are related to the excitonic states in these materials.

The distinction between K and K' valleys is due to the combined effect of SOC and broken inversion symmetry, causing opposite spins of states in these two valleys [33, 36]. Furthermore, the SOC causes splitting of valence and conduction bands in predominantly these K and K' valleys, as shown in figure 1(c) [31]. It was expected that these effects cause perfect spin-valley polarization protected by the above symmetry considerations [37]. However, in practice, the electron–hole exchange interaction causes rapid valley depolarization, even at cryogenic temperatures and in high-quality samples [38, 39]. This suggests that the realization of the often-discussed spin-valleytronics, which aims to use the spin- and valley-degrees of freedom next to the electric charge for solid-state applications, is not very practical in TMDC monolayers. Quantitatively, the energy offsets due to splittings of the valence bands at the K points are predicted to be 148, 183, 426 and 456 meV, respectively, for MoS₂, MoSe₂, WS₂ and WSe₂ monolayers, while those for the conduction bands are, respectively, 16, 35, 3 and 3 meV [32, 40]. This spin-splitting can render the lowest energetic transition between the valence and conduction band at the K-points spin-forbidden, leading to dark states [40]. The existence of these dark states has been shown [41], and it turns out that molybdenum-based dichalcogenides have a spin-allowed lowest transition. In contrast, the lowest transition for tungsten-dichalcogenides is spin-forbidden. This affects the optical properties to a great degree [42, 43].

Another important aspect of the transition from bulk to monolayer is the dielectric screening of electric charges by the material. This screening is drastically reduced from bulk to monolayer, resulting in more strongly bound states of electrons and holes, called excitons [45, 46]. While these bound states also exist in the bulk, their binding energy is enhanced in monolayers, where it can significantly affect the energetics of the electronic bandstructure. In this case, a simple band model of free carriers is not applicable any more, and ultimately, many of the optoelectronic properties of TMDCs become dominated by the properties of excitons [27]. Experimentally, this manifests as sharp resonances in the optical spectra. At the same time, the symmetry considerations causing the spin-split states still apply. As a result, TMDCs display excitons at the valence and conduction band spin-split states, termed A and B excitons, respectively, for the lower and higher energy states of the direct ($K\rightarrow K$) bandgap transition [26]. These A and B excitons represent the bright states. Besides these bright states, there might be spin-dark exciton states with energies above or below the bright states, depending on the exchange- and spin–orbit interactions [40]. Moreover, momentum-indirect transitions ($\Gamma \rightarrow\Lambda$, $K \rightarrow\Lambda$) in the bandstructure contribute additional bound states, but optical transitions to these states have much lower oscillator strengths due to the momentum mismatch [16], so they are much less frequently observed. An overview of the possible transitions is given in figure 1(c).

Because of the high exciton binding energies, excited states of the excitons can be experimentally observed. As the excitons are composed of a positively-charged hole and a negatively-charged electron similar to a hydrogen-like atom, their excited states are commonly modelled by a simple Rydberg-like formula. The excited exciton states are then labelled 1s, 2s, 2p, etcetera, following hydrogenic atoms, and they can contribute to the optical spectra. These states lie between the 1s excitonic ground state and the continuum band, where the energy difference between these two extremes is called the exciton binding energy. Due to selection rules, however, only the s states can be optically excited or probed in the absence of any external fields [47, 48]. Excitingly, recent work suggests that TMDC monolayers show signs of band topology, which would alter the optical selection rules [49]. Mapping out the excited-state energies is one of the common ways of deducing exciton binding energies, as these energies asymptotically approach the continuum band [50].

Finally, in the presence of extra free charges or other excitons, exciton complexes such as trions or bi-excitons can form resulting in extra observable states. These extra free charges can occur due to defects, doping or high excitation fluences [51–54], resulting in binding energies in these complexes on the order of 10s of meV, much lower than those of the excitons, yet in some cases high enough to be observed at room temperature. Besides these common excitonic states, many more states can show up in the optical spectra, such as bulk-specific excitons [55], phonon replicas [56, 57], polaritons [58], and defect emission [59], among others. Many of the mentioned phenomena have by now been established in the prototypical Mo- and W-based TMDCs; yet, even in these materials, the richness of the excitonic states makes it difficult to uniquely assign them to spectroscopic features. For the less established TMDCs, compositional changes, such as in the group IV TMDCs (M: Hf, Zr; X: S, Se), alter the bandstructure [60]. Furthermore, reduced crystal symmetries might cause additional degrees of freedom to appear, such as horizontal linear optical polarization in the case of group VII ReX₂ (X: S, Se) compounds [61]. Especially in these less-studied materials, there is still much to discover regarding the nature of their excitonic properties.

While there are many excitonic states present, typically only one of them, the lowest allowed transition, is observed in PL spectra at room temperature, see for example figure 1(d). Nevertheless, there are numerous radiative and non-radiative recombination pathways in TMDC monolayers that can affect the average decay rate (figure 2). The dominant pathways can vary depending on the temperature, density and type of defect states present, as well as the applied power density [62]. Under low excitation intensities, exciton–phonon scattering and radiative coupling are the dominant recombination processes. Under higher excitation intensity, exciton–exciton scattering also becomes prominent. One common way of characterising material quality is through PL quantum yield (PLQY) measurements, where low PLQY values indicate that non-radiative dominate over radiative decay channels. Attaining high PLQY in these materials remains a challenge, as a significant fraction of excitons typically recombines non-radiatively via intrinsic trap states, mostly due to chalcogen vacancies [63–66]. This limits the PLQY to an order of 1% (0.1% in MoS₂, 6% in WS₂) and radically minimises the feasible optoelectronic applications of TMDC monolayers [67]. On the other hand, the combined effects of confinement and reduced screening tend to increase light absorption at the exciton lines and enhance the PLQY in monolayers compared to the bulk [16, 27, 44, 68]. Consequently, these 2D materials are strong candidates for applications in next-generation optoelectronic devices [69–72].

Zoom In Zoom Out Reset image size

Depending on the application in mind, quantum-yield limitations might be overcome by stacking a TMDC monolayer with a material with high quantum efficiency—for example, a perovskite semiconductor. TMDC monolayers can also be stacked with other TMDC monolayers, which can alter the overall excitonic properties of the materials. Naturally, the abundance of phenomena already present in monolayers provides a very versatile platform for material tunability once these monolayers are combined into HSs. For example, the valley polarization, which is lifted in monolayers due to the electron–hole exchange interaction, can be more protected in HSs due to small wavefunction overlap for the IE [73–75]. Before using these sophisticated effects in applications, however, much work remains to be done to elucidate the structure–property relations of HSs. In the following sections, we briefly touch on the very rich field of TMDC–TMDC HSs before going more deeply into recent results of mixed HSs.

The relative energy levels of the electronic bands of the stacked materials give rise to characteristic band alignment, which determines much of the properties of the HS. One can distinguish three types of alignments, as summarised in figure 4(a): in a type I alignment, the valence band maximum (VBM) and conduction band minimum (CBM) of material A are sandwiched by the valence and conduction band extrema of material B. In this case, both electrons and holes tend to diffuse into the material with the smaller bandgap, and thus energy transfer can occur from material B to A. In a type II alignment, the bands are staggered, meaning that both the VBM and the CBM of one material lie above those of the other material. This favours charge separation at the interface of the two materials, where electrons and holes experience opposite driving forces, and thus get transferred into opposite layers. This alignment is the one required for photodiodes or solar cells to separate photo-generated charges. A rare case is the type III HS where the energies of both valence and conduction band extrema of one material lie above the CBM of the other, creating a so-called broken gap. Because electrons and holes now experience a large potential barrier at the interface, charge transport is essentially only possible due to tunnelling, and thus this type of band alignment can be used to create tunnelling junctions [84–86].

Zoom In Zoom Out Reset image size

The band alignment is generally calculated by density functional theory (DFT) or by the single-shot perturbative G₀W₀ approximation to determine the work functions [87]. The G₀W₀ is also referred to as the dynamically screened interaction; in this approximation, the multi-body electron–electron interactions are captured in a quantity called the electron self-energy and are calculated as a product of Green's function (G) with the screened interaction (W) [90–92].

Experimental methods such as microbeam scanning tunnelling microscopy (STM), high-resolution x-ray photoelectron spectroscopy (XPS), ultraviolet-visible (UV–VIS) spectroscopy and angle-resolved photoemission spectroscopy can then be used to extract the valence band and conduction band offsets and the electronic band gap of the material [93–95].

From the obtained energy values, the band alignment of the HS can be determined [96, 97]. Figure 4(b) shows an overview of band energies for the most common TMDCs, calculated using the $G_0W_0$ method [87], and PMs, obtained through experimental [88], and DFT techniques [89].

TMDC HSs were first reported by Fang and co-workers in 2014, who fabricated a WSe₂/MoS₂ structure [13] by mechanical exfoliation and subsequent stacking of the exfoliated monolayers on top of each other. This technique is now common throughout the literature and allows for convenient identification of the separate monolayers, and the HS region where they overlap, using techniques like room-temperature optical microscopy, Raman spectroscopy, TEM and photoemission electron microscopy. With absorption measurements, the authors showed that the HS region exhibited an absorption profile that is roughly a linear combination of the individual monolayers. Contrastingly, the PL measurements showed quenching of the main exciton peaks of both materials and an additional lower-energy peak. These results suggested that the extra PL peak is not due to a shift of the main exciton lines, but rather due to the HS. The authors attributed this peak to the relaxation of electron–hole pairs bound across the interface; they showed that the PL signal of this feature decreases as hexagonal boron nitride (hBN) layers are introduced between the WSe₂ and MoS₂ layers, supporting the idea that this PL feature originates from interface-separated bound charges. The interfacial electron–hole pairs from this interpretation came to be called IEs and result from HSs of a type II alignment. Subsequent work on a MoSe₂/WSe₂ HS similarly identified a lower-lying PL peak, see figure 5, and found that this state is long-lived, with a lifetime of 1.8 ns [14], much longer than the typical PL lifetimes of excitons and trions, which are on the order of 10 ps for TMDC monolayers [98]. Such slow radiative recombination has also been observed in MoS₂/MoSe₂ HSs [99], indicating that the increased average lifetime measured for TMDC/TMDC HSs is a general property of IEs in group-VI TMDCs. Since then, similar PL signatures have been observed in several other works [100–103], and the measurements were expanded to temperature-dependent PL [102]. In addition, electrostatic gating was introduced to tune the energy of the IE, as shown in figure 6. The vertically separated charges across the stacked layers are associated with an electric dipole moment along the out-of-plane direction, which can be modulated by an electric field induced by an electrostatic gate. Such a static dipole moment is not expected from defect states, thus the observation of electrical modulation of the PL feature provides important support to the existence of IEs [102].

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Crucial theoretical insight came from studies on a MoS₂/WSe₂ HS, where PL spectroscopy was combined with DFT calculations to show that the lower-energy IE feature in the PL spectra could be explained by a $\Gamma \rightarrow K$ transition, showing that the exciton is not just characterized by a lower spatial overlap of the charges due to the layer separation, but its recombination also involves a momentum-indirect transition [104]. Similarly, another study attributed spectroscopic features of TMDC HSs to momentum-indirect transitions [105], although a MoSe₂/WSe₂ HS was suggested to have interlayer features comprised of both direct and indirect transitions [106]. Such momentum-indirect transitions are consistent with the direct-to-indirect transition in bilayer homostructures as the orbital character of TMDCs at the Γ and Λ points have significant out-of-plane character [32–34]. These points in the electronic bandstructures of the HSs are sensitive to both interlayer distance and interlayer stacking angles and can thus be altered by stacking. These features support the observations of predominantly momentum-indirect IEs and are consistent with their long PL lifetimes [74].

Detailed studies on the formation and decay kinetics of excitons in TMDC/TMDC HSs were first performed by Hong and co-workers [107]. They used pump-probe spectroscopy on the femtosecond time scale along with PL measurements to study the ultrafast dynamics of charge transfer across the interface, inspired by previous studies, which demonstrated ultra-fast charge transfer kinetics in organic heterostructures that were also bound together by Van der Waals forces [108, 109]. Hong's work focused on MoS₂/WS₂ layers that exhibit a type II band alignment. Low-temperature (77 K) and room-temperature PL measurements showed strong emission intensities in the separated monolayers, but PL quenching in the combined HS, indicating possible charge transfer. When resonantly pumping at the energy of the MoS₂ A exciton and probing the absorption of the HS, the authors found a pump-induced decrease of absorption in both materials, even though the A and B exciton lines of WS₂ had energies far above the pump energy. This pump-induced decrease of absorption is indicative of charge transfer since the generated charge carriers after transfer occupy states which reduce the absorption. Confirming that the TA signal resurges in isolated WS₂ then consolidated the interpretation of charge transfer from MoS₂ to WS₂. Additionally, by varying the delay time of the probe, the authors determined that holes from the MoS₂ layer had migrated into the WS₂ layer within 50 fs after irradiation by the pump. These kinetics were observed both at 77 K and at room temperature. Other works, such as that of Ceballos and co-workers, yielded similar results for a MoS₂/MoSe₂ heterobilayer [110]. These examples indicate how optical methods can provide insight into charge transfer processes in the TMDC/TMDC heterobilayers [107, 110, 111].

IEs typically recombine at a much slower rate than intralayer excitons as a result of the reduction in the spatial overlap of their wavefunctions [112–114]. The average IE lifetime spans from the order of 50 ps to 1.5 ns, depending on the stacking nature of the layers and the nature of exciton recombination [115]. Hence, recombination lifetimes can be prolonged not only by varying the stacking order but also by measuring at lower temperatures, measuring with a different twist angle, increasing spatial separation of the monolayers and synthesising the HS with a different technique [116, 117]. Furthermore, differences in the photoluminescence spectra and kinetics between TMDC HSs and their monolayers can also be ascribed to surface defect states and other radiative trap states [63, 113, 118–120]. Therefore, single-experiment characterisation, using PL or TRPL alone, is generally insufficient to conclude on the formation of IEs. An additional powerful technique is cathodoluminescence (CL), in which electrons are used to locally excite charge carriers to obtain insight into their local radiative decays, achieving superior spatial resolution in photoemission mapping. While this technique has been applied to IEs to identify and isolate them from intralayer excitons [121, 122], comprehensive work including both steady-state and time-resolved CL is limited, but could provide further insights on the nature of IEs.

One of the critical tuning parameters in stacked TMDC HSs is the relative angle of orientation between two layers. This twisting angle plays a pivotal role in shaping the electronic properties of the resulting structure. For example, in bilayer graphene, the stacked material undergoes a transformative shift into a superconducting state at a highly specific twisting angle, commonly termed the 'magic angle' [123]. The intricate electronic effects induced by the twisting angle create a Moiré pattern—a superlattice formed by the periodic interference between the two layers. Under a specific stacking angle, this leads to the emergence of a flat electronic band, intensifying electron correlations and facilitating superconductivity in the specific case of bilayer graphene [124, 125]. In TMDC HSs, several works have explored the effects of variable stacking orientation and twisted layers. This exploration has revealed a new class of quasiparticles known as Moiré excitons [126–128]. The Moiré excitons can be interlayer or intralayer in nature. These twist-induced excitonic states exhibit enhanced binding energies and, provided that the Bohr radius of the Moiré exciton is smaller than the Moiré pattern itself, long-range spatial periodicity—properties which are distinct from excitons found in individual TMDC layers [129]. In addition to their formation, the precise behaviour of the Moiré excitons is significantly affected by the difference in the twist angle between the TMDC bilayers [130, 131]. For example, with an increasing twist, regions of constructive interference in the Moiré pattern become more localised, which in turn affects the spatial distribution of the excitons themselves. Furthermore, changes in the electronic band structure due to variations in the Moiré pattern will directly impact the absorption and emission characteristics of the Moiré excitons—an increase in twist angle will lead to shifts in the optical signatures associated with Moiré excitons [132, 133].

We note that PL and TRPL are not the only optical techniques applied to study the precise properties of IEs. Reflection spectroscopy has been used as an additional assessment tool to provide further insight into the excitonic nature. As an example, optical spectra, measured at 10 K, were employed to study the twist-angle dependent properties of the IE in the WSe₂/WS₂ TMDC HS under different angles of orientation, after the presence of an IE peak was established from PL [126]. The authors showed that a near-zero twist angle contributes to a much stronger enhancement of the interlayer excitonic peak than a larger angle. This is attributed to the Moiré pattern at the near-zero twist angle generating distinct Moiré exciton states, which are then visible in the reflectance spectra. Notably, the reflectance spectra were directly compared to the PL excitation (PLE) spectra, and found to yield the same excitonic features. With properties such as the ones discussed above, Moiré excitons in TMDC heterobilayers have expansive potential applications in the fields of 'twistronics' (utilising the twist-dependent electronic properties of the system), nano-photonics and quantum information [134, 135].

Over the past few years, alloying has emerged as a powerful technique to tune the structure and optoelectronic properties of TMDCs. In this fabrication technique, layers of $\mathrm{M}_ {\mathrm{y}} \mathrm{M}^{^{\prime}}_{1-{\mathrm{y}}}\mathrm{X}_{2}$ or $\mathrm{MX}_{2{\mathrm{y}}} \mathrm{X}^{^{\prime}}_{2(1-{\mathrm{y})}}$ are synthesised by varying the composition (y) of the metal ($M/M^{\prime}$) or chalcogen ($X/X^{\prime}$) atoms [136–138]. By controlling the alloying composition in each individual layer, fundamental properties such as bandgap and effective mass can be effectively modified [136, 139, 140]. Both theoretical and experimental studies have shown composition-dependent bandgaps in the TMDC alloys and the obtained results have opened possibilities of bandgap engineering [136, 141–143]. Since the alloyed layers can be used for stacking, this further expands the space of HS possibilities for the TMDCs.

Interesting examples of alloyed layers are Janus monolayers, where the controlled mixing of chalcogen atoms leads to bottom and top chalcogen planes of different composition. This results in an out-of-plane dipole, as for the layered HSs [144]. However, due to the complete absence of any real stacking, there is no possibility of momentum mismatch, and the optical transitions tend to be at energies in between those of the pure compounds [145]. One study fabricated WSSe/MoSSe and WSe₂/WSSe HSs by a combination of plasma-induced vacancy formation coupled with a CVD process [145]. The PL spectra of the HSs clearly showed that the HSs exhibit PL from both layers, yet no clear conclusions were drawn on charge transfer effects. Another study on stacking Janus monolayers investigated the effects of the built-in field from the different chalcogen planes on charge transfer rates in a WSSe/WS₂ HS and found that the Janus-induced electric field can function as an additional potential barrier, prohibiting charge transfer, even when the type II band alignment suggests its possibility [146]. This makes the charge transfer rate directional and dependent on the stacking sequence. A further alloying possibility is the alloying of the metallic component, followed by HS formation, creating layers with different metallic constituents. This has been explored briefly in early work where HS alloys were directly grown by CVD [147]. While the HS composition was indeed verified through Raman and PL techniques, however, no clear analyses were performed on charge transfer, leaving it as an interesting future research direction. These examples show that, already within the TMDC family, there is a lot of design space for fabricating HSs with diverse functionality. This design space can be further expanded by stacking TMDC monolayers with completely different compounds. This direction will be discussed in the following section, focusing on TMDC/perovskite HSs.

We will primarily focus on HSs consisting of the conventional perovskite with the ABX₃ crystal structure, wherein A corresponds to a small cation, such as Cs⁺ or methylammonium (MA⁺), B corresponds to a large cation, such as Pb²⁺ and Sn²⁺, and X is a halide anion or some mix of halide anions. Much has been written on the chemical and optoelectronic properties of these perovskites and we refer the reader to some in-depth reviews [154–158]. Briefly, semiconducting inorganic CsPbX₃ materials, where X is a halide atom, have been known to crystallise in the perovskite structure since the 1950s, [159], and the prototypical organic–inorganic compound of methyl-ammonium lead halide (MAPbX₃) was subsequently described in the 1970s [160]. Differences in A-site cations (MA versus Cs) not only result in structural differences and relative differences in stability (CsPbBr₃, which is the more stable of the two, is typically found in the orthorhombic phase at room temperature, while MAPbBr₃ is found in the cubic phase) but also results in optical and optoelectronic changes in the material—for example, the band gap of MAPbBr₃ tends to be lower than CsPbBr₃ [161, 162]. Going from bulk to confined systems, the 2D-layered, Pb-based, perovskites were first synthesised in the 1980s [163], and basic optical properties were described in the early 1990s [164]. The main motivation for studying these layered crystals was that they act as natural quantum well systems, due to the Van der Waals interactions between the layers, as opposed to the heavily designed GaAs-based wells. Subsequently, perovskites experienced a resurgence as high-efficiency emitters due to the landmark paper by Kovalenko's group on the synthesis of CsPbX₃ nanocrystals in 2015 [165]. Clearly, TMDCs and PMs overlap in the motivation for studying systems with strong excitonic behaviour, and it is no surprise that after these low-dimensional systems were shown to be stable in practice, the field quickly moved to combine these two classes of semiconductors.

What makes the PMs interesting candidates for HSs with TMDCs, is, next to their bandgap tunability by composition, the fact that carrier transport in the perovskites has been shown to be effective enough for the migration of charges to the HS interface even if the perovskite layer is 200 nm thick [166, 167]. This might be in part due to a process called photon recycling in PM that effectively enhances carrier diffusion lengths by repeated absorption and emission events [168]. The thickness of these perovskite layers can be increased without significantly changing the electronic bandstructure, while for TMDCs clear changes in bandstructure and excitonic properties occur with increased thickness. Therefore, the addition of a perovskite layer might enhance flexibility and tunability in HS properties without sacrificing potentially desired TMDC monolayer properties. In the following, we discuss multiple dimensionalities of perovskites, from bulk to nanocrystals. This is an important distinction as nanocrystals, for example, are typically capped with stabilising ligands [165], which can potentially affect the HS interface. Thus, we will specify the perovskite dimensionality for each discussed work.

The first reports of TMDC/PM HSs date from 2016, where MAPbI₃ thin films (10s–100s nm thick) were used as the perovskite layer [169–172]. While MoS₂ had been used before as a hole-transport layer in organic solar cells, [173], in the TMDC/perovskite HS, it was used for the first time as an active absorbing layer, next to the perovskite layer. To tune the charge and energy transfer between the perovskite and TMDC, modulation of the band alignment between these two materials is essential. In this context, the work by Peng and co-workers is particularly interesting, who analysed charge transfer processes in a MoS₂/MAPbI₃ perovskite HS, where the MAPbI₃ was 30 nm thick [171]. They experimentally analysed the excitonic states using pump-probe spectroscopy in a similar manner to what has been described in section 3.2. The work reported that the band edge alignment can be tuned by defect engineering: the type I heterojunction could be converted to a type II heterojunction using oxygen plasma treatment of MoS₂. As a result of the band-alignment engineering, competitive charge transfer and energy transfer mechanisms could be varied. This work demonstrated highly efficient charge transfer in which 83% of the holes generated in the perovskite were transferred to the MoS₂, on a timescale of 300 fs after photoexcitation. The fact that defects can play a strong role in the performance of these mixed HSs was explicitly investigated in another work, in which defects were passivated in a CVD-grown WSe₂ monolayer through laser-induced oxidation. The photo-responsitivities of the resulting HS with a bulk-like film of MAPbI₃ device with the treated monolayer were higher than the device with the untreated monolayer [172]. The more diverse band alignments in these mixed HSs are a main advantage over the typical TMDC/TMDC HSs, which generally form type II band alignments, as shown in figure 4. However, as the previous works show, the role of TMDC defects can not be neglected in the characterisation of the HSs.

In the following years, more works on the device performance of these HSs were published (see [176–181], for a selection, with various perovskite compositions and morphologies). While most of these works rely on the electronic charge- or energy transfer, it is worth pointing out at least one study that used the excellent emissive properties of a 37 nm thick layer of CsPbBr₃ nanocrystals and subsequent radiative energy transfer to effectively boost the absorption of the underlying MoS₂ layers [182]. Other works also went more deeply into the processes underlying the energy and charge transfer in the HSs. In this regard, the work from Bauer et al focused on a 100 nm thick film of MAPbI₃ perovskites but with WS₂ as the TMDC in the HS [174]. Akin to TMDC/TMDC HS analysis, they used microscopic ultra-fast pump-probe transient spectroscopy measurements combined with room-temperature PL spectroscopy, to determine the kinetics within the HS. This work presented a somewhat unusual finding: despite a type I band alignment, for which energy transfer and hence PL emission would be expected (figure 7(a)), the PL contributions from the individual constituents were quenched (figure 7(b)). As the PL was quenched in both materials, the authors concluded that energy transfer was not the dominant transfer mechanism. Rather, the strong binding energy of the excitons in the TMDCs and the small energy difference between the VBMs of the two materials limited the transfer of holes to the perovskites. While the holes from the B-exciton band could still transfer, the effective hole transfer rate was smaller than the electron transfer rate, for which both A-exciton band and B-exciton band electrons could contribute in the transfer from the TMDC to the perovskite. This resulted in an HS dominated by charge, not energy, transfer. This suggests that the peculiar spin–orbit physics of TMDCs has far-reaching effects for their functionality in HSs, opening up sophisticated ways of manipulating charge transfer by comparing A-exciton and B-exciton dissociation rates.

Zoom In Zoom Out Reset image size

Extending these band alignment considerations further, a HS of monolayer MoS₂ and 40 nm thick CsPbBr₃ nanosheets, with a type II alignment was reported [176], while another work showed a type I alignment between CsPbBr₃ nanowires and monolayer MoS₂ [183]. Similarly, a WS₂ monolayer/CsPbBr₃ 180 nm thick microplate HS exhibited a type II alignment in one study [101], while in another study, WS₂ with CsPbBr₃ QDs (average size of 11 nm)—rather than sheets—resulted in a type I alignment [184]. In both examples, the calculated or measured bandgaps of the PMs did not change significantly; rather, the morphology and dimensionality of the perovskites seem to have caused the shift in band alignments. In terms of material dimensions, it was shown that the diffusion length of charges in a CsPbBr₃ film is sufficiently long that perovskite films from 8 nm up to 200 nm thick can still efficiently transfer charges to TMDC monolayers [166, 167]. These examples indicate that the relative ease of solution processing of the perovskites combined with the spin–orbit split bands in the TMDCs allows a wide variety of band alignments to be engineered. As pointed out before, these spin–orbit split bands cause alignments that are neither precisely type I nor type II but create competing transfer processes [184].

Naturally, adjusting the perovskite bandgap by composition is a powerful way of tuning band-alignments. In 2018, a comparative study on the differences in the band alignments for inorganic and organic perovskite QDs, namely MAPbBr₃ and CsPbBr$_{3-{x}}$I$_{{x}}$, with average diameters of 7 nm and 15 nm respectively, in MoS₂ HSs was published [175]. Using ultra-violet photoelectron spectroscopy measurements, the work showed that organic MAPbBr₃/MoS₂ heterobilayers formed a type I band structure, while the inorganic CsPbBr$_{3-{x}}$I$_{{x}}$/MoS₂ HS formed a type II structure (figure 7(c)). In line with the general PL quenching trend in type II structures, the CsPbBr$_{3-{x}}$I$_{{x}}$/MoS₂ showed PL quenching in both layers, while for the type I HS, the PL intensity was quenched only for the MAPbBr₃ QDs, but enhanced for the MoS₂, highlighting the role of band alignment in the excitonic charge and energy transfer. The PL spectra obtained from this study are shown in figure 7(d). In addition to the PL and TRPL analyses, the authors also studied the optoelectronic performance of these HSs by separately incorporating them into photofield-effect transistors (PFETs). In both devices, the HSs were fabricated on top of back-gate SiO₂/p-type Si substrates, and Au/Cr contacts were evaporated onto the substrates as the source and drain contacts. The performance of each system was analysed by current–voltage sweeps in the dark and under illumination. The measured photocurrent of the all-inorganic type II HS was more than four times that of the organic–inorganic type I HS. The superior optoelectronic properties of the all-inorganic HS was attributed to trapped long-lifetime carriers in the surface states or interface trap states, particularly arising from the MoS₂ interface. The Cs-based system's ability to transfer photoexcited electrons while accumulating photogenerated holes resulted in a strong photogating effect, contributing to the overall enhanced optoelectronic performance. Conversely, both of the photogenerated electrons and holes were transferred to the MoS₂ layer in the MA-based system, where they subsequently recombined; this ultimately led to a relatively poor performance in the organic–inorganic device. More generally, the authors argue in this work that the advantageous optoelectronic properties coupled with the higher overall stability of CsPbBr₃ perovskite QDs may lead to more favourable TMDC/perovskite HS systems and devices in the future.

As shown by the Kovalenko group, the bandgaps of the all-inorganic perovskite nanocrystals change with the halide element, from around 3 eV in CsPbCl₃ via 2.4 eV for CsPbBr₃ to 1.8 eV for CsPbI₃ [165]. This change in bandgaps is accompanied by a change in band energies [88], and it will be interesting to see if the facile change in halides can induce a change from a type II to a type I HS. However, no systematic experimental study exists yet on this halide-dependent band alignment in an HS with TMDCs. Clearly, the facile exchange of cations and anions in semiconducting PMs promises a wealth of tunability for HSs in the near future.

Finally, further opportunities for interfacial engineering in TMDCs/perovskite HSs arise through the surface chemistry of the perovskites, or by inserting spacer layers between the two materials. Using room-temperature, TRPL spectroscopy, Liu et al showed that the introduction of hBN spacer layers between all-inorganic perovskite nanocrystals, with average sizes of 8.5 nm and monolayer MoS₂ altered the PL lifetime of the perovskite nanocrystals, with shorter lifetimes for thinner spacers [80]. They showed that the charge transfer rates can be effectively modulated by introducing hBN flakes of up to 7 nm thick. Using ligand engineering in the same HS, albeit with slightly larger nanocrystals of around 15 nm, interfacial charge transfer was enhanced by removing as much of the nanocrystal ligands as possible without compromising the stability of the perovskite nanocrystals [185]. With this treatment they improved the photo-responsitivity of their photodetector devices by a factor of 15. Another approach to ligand engineering exists for quantum-confined nanocrystals, where the choice of surface ligands can affect the band alignments, without changing the bandgap. This has been shown before for QD HSs [186], but to our knowledge, not yet for mixed TMDC–perovskite HSs.

The above examples demonstrate that elemental composition, dimensionality, morphology, defects, layer distances and surface chemistries of a TMDC/perovskite HS can all alter the band alignment of the HS, which in turn, determines the extent of charge or energy transfer between the layers. As a consequence, the simple band alignment picture is not always indicative of charge or energy transfer processes, as can be seen by comparing the panels of figure 7. Rather, a combination of experimental methods such as (time-resolved) PL and TA spectroscopy is required to determine the charge kinetics in TMDC/perovskite HSs.

Therefore, the same combination of techniques that already provided insight into the charge dynamics of TMDC/TMDC HSs is useful in elucidating the carrier dynamics in the TMDC/perovskite HSs, as will be discussed in the following sections.

Steady-state PL spectroscopy, particularly PL spectroscopy at low temperatures, is a convenient technique to optically characterise the TMDC/perovskite HSs. In Wu's comparative analysis, the PL quenching in the CsPbX₃/MoS₂ HS indicated full energy transfer, while the PL quenching of the organic perovskite along with the PL enhancement of MoS₂ indicated charge transfer [175]. This argument is generally applied in most TMDC/perovskite reports [101, 184, 187].

Besides that, low-temperature PL measurements can provide insights into the kinetics of the biexcitons and charged excitons within a TMDC/perovskite HS, which otherwise would not be observed. For example, one study reported the PL of a WS₂/microplate CsPbBr₃ structure and its pristine components at 3.3 K to resolve the excitonic emission bands [101]. As the density of the biexcitons varies quadratically with the number of excitons in the pristine TMDC monolayer, small fluctuations in the exciton density, which may be undetectable in many PL systems, can be detected from changes in the biexciton emission (figure 8). Additionally, since the binding energies of trions and biexcitons are lower than those of neutral excitons, the driving force for charge separation at the interface will be more effective for these exciton complexes. At low temperatures, the authors observed full quenching of the biexciton emission in the HS compared to the monolayer, indicating that charge transfer occurred within the sample.

Zoom In Zoom Out Reset image size

Adhikari et al further studied the excitonic room-temperature PL spectra of a CVD-grown monolayer MoS₂ and 11 nm sized CsPbBr₃ nanocrystals HS and noted that each of the neutral A and B excitons and the negatively charged A⁻ exciton red-shifted in the HS [188]. Their argument was further strengthened by showing more dominant trion emission, independent of laser power, and related shifts and power dependence in TA measurements. Furthermore, by constructing gated devices with and without the QDs, the work also found a downshift in the threshold voltage in the HS compared to the pristine device. Both indicate that the MoS₂ layer's electron concentration was higher in the hybrid than in the pristine sample. In turn, this implied charge transfer had occurred in the HS, with electrons transferred from the QDs to the MoS₂. Overall, this study highlights the difficulty of characterising charge transfer by exciton lines, as both the exciton binding energy as well as the electronic bandgap can shift with the influx of charges.

Before we continue towards alternative signatures of energy transfer, we briefly exemplify a lead-free system, namely the evaporated CsSnBr₃/WS₂ (grown by CVD) HS as recently fabricated and characterised in the work of Mu et al [189] The key novel concepts summarised from this work are two-fold: first, as it has been argued that toxicity of lead will limit the extent of future applications for any perovskite optoelectronic device, a substitution of lead by another element such as tin is necessary. Second, contrary to the majority of published works using mostly solution-processed methods such as mechanical exfoliation or vacuum deposition to prepare the perovskite, in this study, precise tuning of the perovskite thickness (varying between 10 nm and 70 nm) was achieved using thermal evaporation. The optical properties were characterised using UV–VIS absorption spectroscopy, steady-state PL and TRPL. Using UV–VIS measurements supplemented with XPS, the HS was found to exhibit type-II band alignment. The observed PL quenching in this system together with the differences in TRPL lifetime was attributed to charge carrier transfer across the system. This work by Guo et al is a pioneer study on the use of lead-free perovskites in TMDC-based HSs. Their findings are naturally not as encompassing as those from the multitude of studies done on lead-based perovskite HSs. However, their study presents an important step towards sustainable implementation of TMDC/PM HSs and it will be interesting to see future work going deeper into the comparison of lead-based and lead-free perovskite/TMDC HSs.

PL lifetimes can also yield insight into the possible formation of IEs and whether charge or energy transfer occurs in a TMDC/perovskite HS. For example, room-temperature TRPL and fluorescence lifetime imaging microscopy measurements of organic–inorganic heterostructures found that the average PL lifetime of the CsPbBr$_{3-{x}}$I$_{{x}}$ QDs was 16.7 ns, which reduced to 6.2 ns in the MoS₂/ CsPbBr$_{3-{x}}$I$_{{x}}$ HS, while the average PL lifetime of MAPbBr₃ reduced from 11.2 ns in the pristine sample to 2.9 ns in the HS [175]. This reduction was attributed to the charge extraction kinetics of the HS.

Further insights into interfacial dynamics can be obtained by varying the interfacial width, and evaluating its effect on the charge transfer dynamics. By either decreasing the thickness of MoS₂ or removing insulating hBN layers between the materials [80], the PL lifetime of the perovskite QDs decreases (figure 9), indicating the presence of an additional relaxation channel. Panel (c) of figure 9 shows the TRPL data for varying MoS₂ layer thickness, and panel (d) the extracted thickness-dependent relaxation rate, which saturates for thicknesses of around 10 MoS₂ layers, approaching the value of the isolated QDs. Panels (e) and (f) reveal a similar behaviour for the thickness of an hBN spacer layer between the QDs and an MoS₂ monolayer. This decrease of the QD PL lifetime suggests the emergence of additional non-radiative relaxation channels that become activated for lower energy barriers or when the band-alignment of the two compounds becomes more favourable due to thin MoS₂ as opposed to thicker MoS₂. This thickness-dependent PL lifetime thus points to the presence of charge transfer across the HS.

Zoom In Zoom Out Reset image size

Ultrafast spectroscopic characterisation thus presents a powerful tool to study the kinetics and determine rates of charge- and energy transfer in TMDC/perovskite HSs. It is therefore interesting to ask whether similar to TMDC/TMDC HSs, there have been signs of spatially separated IEs in mixed HSs as well. The studies towards this topic are few in the mixed TMDC/PM HS. Nevertheless, an interesting study was presented by Li et al, who performed a full analysis of the charge dynamics across a WS₂/CsPbBr₃ QD HS using a combination of femtosecond TA spectroscopy and room-temperature steady-state PL [184]. The TA signal showed photobleaching (PB) proportional to the exciton fluence and photoabsorption (PA) due to biexcitons and other pump-induced absorbing states, see figure 10(a). The quenching of the PB signal suggests that few excitons remain in the QDs after excitation. Additionally, the PB signal of the A-exciton resonance was enhanced, suggesting that the excitons generated in the perovskite were immediately transferred to the TMDC. To pin down the competitive processes behind the PB and PA features, 'zero-points' (ZP) were defined at probe wavelength positions where the photo-induced absorption vanished (vanishing ΔA in figures 10(b) and (c). In the HS, these ZPs shift over time towards an increased PA feature, indicating the formation of a built-in electric field due to IEs across the HS, which in turn altered the resonances of the excitons in the WS₂ layer. This concept had previously been described by He and co-workers for a WS₂/graphene HS [190], and here could be developed into a full model of the recombination dynamics and fitted to the TA signal for fast and slow recombination processes as depicted in figure 10(d).

Zoom In Zoom Out Reset image size

In addition to the above examples of thin films or QDs of lead halide perovskites with ABX₃ structure, 2D Ruddleston–Popper perovskites, with a structure of A$_{{n}+1}$B$_{{n}}$X$_{3{n}+1}$ are also widely employed as one of the layers in TMDC/perovskite HS [187, 191–193]. As an example, excitonic energy transfer has been observed in the quasi-hybrid perovskite C₆H₅C₂H₄NH₃ (PEPI)/WS₂ HS, with the PEPI being around 100 nm thick and monolayer WS₂ [187]. Experimental evidence for the energy transfer was presented using PL, PLE and reflectance spectroscopy techniques. In the PLE measurements performed at 110 K, the authors observed that the A exciton PL from WS₂ is enhanced at the excitation energy corresponding to the exciton of PEPI, indicating energy transfer between the two compounds had occurred. Overall, while many initial works have been performed on these mixed HSs, important details of the fundamental processes are yet to be uncovered. For example, many tuning parameters have been discussed above, but it is not always clear how these tuning parameters affect the time dynamics of charge transfer. To gain further insight, the optical experimental methods described above should be supplemented by further optical measurements such as reflectance spectroscopy, and other physical characterisation methods such as x-ray diffraction, TEM, and atomic force microscopy, and by theoretical analyses. In this regard, a theoretical study using time-dependent DFT points to yet another factor in band alignments [194]. The authors found that the exact crystal facet of CsPbBr₃ nanocrystals exposed to the TMDC can alter the band alignment and suggested that this can be an alternative explanation for mixed band-alignment, besides the spin–orbit splitting mechanism proposed by Bauer et al [174]. This interpretation is still discussed in the follow-up theoretical work by Liao et al, who found a type II alignment for both crystal facets in the same system [195]. Subsequently, Xu et al stressed that the different crystal facets of CsPbBr₃ do result in different band alignment types with MoS₂, as long as the SOC is included to describe the properties of the perovskite [196]. Similar work has been undertaken on the CsPbBr₃/MoSe₂ system, where both perovskite facets show a type II alignment with the TMDC [197]. Finally, also in the case of the CsPbI₃/MoS₂ and CsPbI₃/WS₂ systems, the alignment turns out to be of type II, independent of the perovskite facet exposed to the TMDC. This shows that for some specific combinations, such as CsPbBr₃/MoS₂, the band alignment is so close between type I and type II that the specifics of crystal facets, and SOC start to become important [198]. It will be interesting to see how these effects, and the aforementioned tuning parameters, overlap in less-than-ideal macroscopic samples and practical situations.

Before we address the alternative processes that may be occurring in TMDC/TMDC and TMDC/PM HSs, we summarise the key experimental reports that we have previously discussed regarding TMDC/PM HSs in table 1.

Table 1. Summary of the experimental works discussed in this literature review which characterise various TMDC/PM heterostructure materials using various key optical methods. These tabulated works also explicitly identify the heterostructure type and describe evidence for either charge transfer, energy transfer, or both. Abbreviations: BA = butylamine, N.D. = not described, Abs. = UV–VIS absorption spectroscopy, TRPL = time-resolved photoluminescence spectroscopy, TAS = transient absorption spectroscopy, DRS = differential reflection spectroscopy.

TMDC/perovskite	Perovskite type	HS type	Transfer type	Key optical techniques	Year	Reference
WS2/CsPbBr3	Microplates	Type II	Charge	Abs., PL, Raman	2018	[101]
MoSe2/CsPbBr3	Nanocrystals	Type I	Energy	Abs., PL, TRPL	2022	[153]
WSe2/CsPbBr3	Thin film	Type I	Energy	Abs., PL, TAS	2021	[166]
MoS2/CsPbBr3	Thin film	Type II	Charge	Abs., PL, TAS, TRPL	2022	[167]
WS2/CsPbBr3	Thin film	Type II	Charge	Abs., PL, TAS, TRPL	2022	[167]
MoS2/MAPbI3	Thin film	Type II	Charge	PL, TAS, Abs., Raman	2016	[171]
WS2/MAPbI3	Thin Film	Type I	Charge	Abs., PL, TAS, DRS	2018	[174]
MoS2/CsPbBr$_{3-{x}}$I$_{{x}}$	Quantum dots	Type II	Charge	Raman, PL, TRPL	2018	[175]
MoS2/MAPbBr3	Quantum dots	Type I	Energy	Raman, PL, TRPL	2018	[175]
MoS2/CsPbBr3	Nanosheets	Type II	Charge	Abs., PL, TRPL	2018	[176]
WS2/CsPbI3	Nanocrystals	Type II	Charge	Abs., PL, TRPL, Raman	2023	[180]
MoS2/CsPbI$_{3-{x}}$Br$_{{x}}$	Quantum dots	Type II	Charge	Raman, Abs., PL, TRPL	2018	[181]
MoS2/Cs0.85FA0.15PbBr3	Quantum dots	N.D.	Charge	Raman, Abs., PL	2022	[182]
MoS2/CsPbBr3	Nanowire	Type I	Energy	Raman, PL, TRPL, TAS	2018	[183]
WSe2/CsPbBr3	Nanowire	Type II	Charge	Raman, PL, TRPL, TAS	2018	[183]
WS2/CsPbBr3	Quantum dots	Hybrid I/II	Charge	TAS, PL, Raman	2018	[184]
MoS2/CsPbI$_{3-{x}}$Br$_{{x}}$	Quantum dots	Type II	Charge	Abs., PL, TRPL, Raman	2019	[185]
WS2/(C6H5C2H4NH3)2PbI4	Quasi-2D flakes	Type II	Energy	DRS, PL, PLE, Raman	2020	[187]
MoS2/CsPbBr3	Quantum dots	Type II	Charge	Abs., PL, TAS, Raman, TRPL	2020	[188]
WS2/CsSnBr3	Thin Film	Type II	Charge	Raman, Abs., PL, TRPL	2023	[189]
WSe2/(C4H9NH3)$_ 2$PbI4	2D Microplates	Type II	Charge	PL, Raman	2020	[191]
WS2/(C6H5C2H4NH3)2PbI4	2D single crystals	Type II	Charge	DRS, PL	2020	[192]
WS2/(BA)2(MA)3Pb4I13	2D bulk crystals	Type I	Charge	Raman, PL, PLE, TRPL	2019	[193]