Introduction

Heterojunctions, where two different materials meet, form the fundamental building blocks for modern functional devices1, such as light emitting diodes2, photodetectors3 and field-effect transistors4. The critical determinant of heterojunction device characteristics lies in the alignment of conduction and valence bands between the two materials5,6,7. Type-I band alignment occurs when the band gap of one material is fully encompassed within the band gap of the other material, i.e., both the conduction band minimum (CBM) and the valence band maximum (VBM) of the heterostructure reside in the same material. This straddled band alignment causes photoexcited electrons and holes to relax into the same medium8,9 (Fig. 1a, b). As a result, the excitons have larger electron-hole wavefunction overlap, higher oscillator strength, and shorter radiation lifetime, which favor applications in light emitting devices10,11. In contrast, type-II band alignment situates the CBM of the heterostructure in one material and the VBM in the other (Fig. 1c, d). This staggered alignment causes the photoexcited electrons and holes to relax to different materials. This facilitates exciton dissociation and photocarrier extraction and hence favors applications in photodetection12,13 and photocatalysis14.

Fig. 1: Illustration of electric-field-induced type-I to type-II band alignment transition in MoSe2/WS2 heterobilayer.
figure 1

a, b Atomic layers and energy band diagram of a MoSe2/WS2 heterobilayer with no electric field. The position-dependent conduction bands (CB) and valence bands (VB) of monolayer MoSe2 (blue) and WS2 (red) exhibit the type-I alignment. The photoexcited electrons and holes relax to the MoSe2 layer to form intralayer excitons (\({A}_{{Mo}}\)). c, d Atomic layers and energy band diagram under strong vertical electric field. The heterobilayer exhibits type-II band alignment. The photoexcited electrons (holes) relax to the WSe2 (MoSe2) layer to form interlayer excitons (\({IX}\)). e Measured photoluminescence (PL) spectra under zero (blue) and E = 0.25 V/nm (red) electric field. The interlayer excitons are turned on and off by the electric field.

Full size image

In the conventional scenario, the band energy offset between two materials is fixed; this imposes many constraints on device functionality. To usher in the next generation of devices, it is highly desirable to transition between different alignment types within a single heterostructure. The realization of tunable band alignment holds immense potential for unlocking unprecedented multifunctionality in device applications. However, achieving such demonstrations is challenging. While certain approaches involve material engineering through adjusting chemical compositions15,16,17,18 or layer thickness19, these methods necessitate multiple synthesis iterations, precise dopant control, or the use of different samples. A more efficient alternative envisions the capability to tune the band alignment types through straightforward electrical means in a single device, though, to the best of our knowledge, such a demonstration has not been realized.

A groundbreaking advancement in the fabrication of band-engineered heterostructures has emerged within the realm of two-dimensional (2D) van der Waals (vdW) materials20,21,22,23,24, notably including monolayer transition metal dichalcogenides (TMDs) like MoSe2 and WS2. These TMDs feature direct bandgaps, heavy carriers, robust excitons25,26,27,28,29,30,31,32, and innovative spin-valley-coupled physics33,34,35,36,37. In contrast to conventional grown heterostructures, 2D vdW heterostructures exhibit reduced dimensions, atomically sharp interfaces without dangling bonds, and high tolerance for lattice mismatch. These attributes greatly enhance their stability, tunability and overall functionality. Moreover, the continually expanding library of diverse 2D materials offers a rich palette of options for creating vdW heterostructures. Indeed, recent theoretical propositions have delved into the possibility of tuning band alignments in 2D vdW heterostructures through various means, including electric field38,39, strain38, interlayer spacing38, and twist angle40. However, experimental demonstration of such band alignment tuning is still lacking.

In this Article, we demonstrate an electric-field-induced transition between type-I and type-II band alignments in MoSe2/WS2 heterobilayers, as evidenced through photoluminescence (PL) and photocurrent (PC) spectroscopy. Figure 1 illustrates our key findings. Initially, in the absence of an external electric field, the heterojunction of monolayer MoSe2 and WS2 exhibits a type-I band alignment. In this state, the WS2 CBM is marginally higher than the MoSe2 CBM, while the WS2 VBM is considerably lower than the MoSe2 VBM (Fig. 1a, b). This configuration, with its straddled band gap, facilitates the relaxation of photo-excited electrons and holes to the same (MoSe2) layer, leading to a pronounced PL peak of intralayer excitons (\({A}_{{Mo}}\)) at 1.6 eV (blue line in Fig. 1e). However, the application of a strong vertical electric field, directed from the WS2 to MoSe2 layer, triggers a critical shift—the WS2 CBM moves below the MoSe2 CBM (Fig. 1c, d). This alteration leads to a type-II band alignment with staggered band gaps, where the photoexcited electrons and holes tend to relax to different layers—electrons to the WS2 layer and holes to the MoSe2 layer. This separation leads to the formation of interlayer excitons (\({IX}\)), which have a lower energy than their intralayer counterpart, resulting in a dominant PL peak at ~1.57 eV (Fig. 1e). Beyond the striking shift of the PL spectrum, the emergence of interlayer excitons also effectively traps electron-hole pairs at the heterojunction, thereby suppressing the interlayer photocurrent (Ipc). This results in a highly nonlinear relationship between Ipc and the applied interlayer voltage. Overall, our findings highlight the profound impact of the type-I to type-II transition on the optical and optoelectronic properties of devices. The ability to electrically control this transition opens up exciting possibilities for designing innovative multifunctional devices using vdW heterostructures.

Results and discussion

Photoluminescence measurements

Our experiment employs dual-gate MoSe2/WS2 heterobilayers encapsulated in hexagonal boron nitride (BN)41 (Supplementary Fig. S1, Fig. 2a). We use thin graphite flakes to contact the TMDs and electrodes to enhance device performance. The heterobilayers for PL measurements have twist angles of either ~\(0^\circ\) or ~ \(60^\circ\). Deviation from these angles will suppress the interlayer emission due to electron-hole momentum mismatch. Below we will present the PL results of Device 1 while the reflectance contrast results are presented in Section 4 of the Supplementary Information.

Fig. 2: Photoluminescence signature of type-I to type-II band alignment transition in MoSe2/WS2 heterobilayers.
figure 2

a The schematic of a dual-gate MoSe2/WS2 heterobilayer device encapsulated by boron nitride. b The charge-density-dependent PL map of Device 1. Equal voltages Vbg = Vtg are applied to the bottom and top gates. The charge density is proportional to the gate voltages. The \({A}_{{Mo}}\), \({A}_{{Mo}}^{-}\), \({A}_{{Mo}}^{+}\) features arise from the intralayer excitons, electron-side and hole-side exciton polarons (or trions) in the MoSe2 layer, respectively. c The electric-field-dependent PL map of Device 1. Opposite voltages Vbg = −Vtg (left axis) are applied to the bottom and top gates to induce an interlayer electric field (right axis). An interlayer exciton (\({IX}\)) feature appears at high electric field. The dashed line is a linear extrapolation of its Stark shift to zero field. All measurements were performed with 532-nm laser excitation (incident power ≈ 3 μW) at sample temperature T ≈ 6 K.

Full size image

In Device 1, the top and bottom BN have similar thickness. By applying voltages with the same sign on the bottom gate (Vbg) and top gate (Vtg), we can inject carriers into the sample without inducing any vertical electric field. The charge-density-dependent PL map of Device 1 (Fig. 2b) exhibits a pronounced line (\({A}_{{Mo}}\)) at 1.60 eV at the charge neutrality region, which matches the reported A-exciton energy in BN-encapsulated MoSe2 monolayers42,43,44,45. Upon injecting electrons or holes into the heterobilayer, the \({A}_{{Mo}}\) peak subsides and two new PL peaks (\({A}_{{Mo}}^{-}\), \({A}_{{Mo}}^{+}\)) emerge at ~30 meV below the exciton on the electron and hole side, respectively. This energy separation is close to the known exciton-polaron (or trion) binding energies in monolayer MoSe242,43,44,45. Therefore, we infer that \({A}_{{Mo}}\), \({A}_{{Mo}}^{-}\), \({A}_{{Mo}}^{+}\) originate from the intralayer exciton and exciton polarons in the MoSe2 monolayer. We note that the weak lines at higher energies than \({A}_{{Mo}}\) may arise from moiré effect or sample inhomogeneities, since they are not reproducible in other devices (Supplementary Figs. S8 and S9). Atomic reconstruction is unlikely to occur in this system due to the small moiré wavelength (~8 nm).

By applying voltages with opposite signs on the bottom and top gate (Vbg = −Vtg), we can apply a vertical electric field between the MoSe2 and WS2 layers while maintaining the heterobilayer in the charge neutrality regime. Figure 2c displays the PL map of Device 1 at varying Vbg = –Vtg (left axis), from which we extract the out-of-plane electric field (right axis) with a BN dielectric constant of 3.4 (see the details in Supplementary Information, Section 2.1). At weak electric field (E < 0.16 V/nm pointing from WS2 to MoSe2), the \({A}_{{Mo}}\) PL line remains pronounced and exhibits no Stark shift; this observation confirms its intralayer nature and supports the type-I alignment of the heterobilayer (Fig. 1b). When the electric field exceeds 0.16 V/nm, the \({A}_{{Mo}}\) line subsides and below it emerges a new PL peak (\({IX}\)). The \({IX}\) peak redshifts linearly with a slope of 44 \(\pm\) 11 meV per 0.1 V/nm of field; the \(\pm\)11 meV error is mainly due to the uncertain BN dielectric constant (2.6–4.2) with a minor \(\pm\)0.9 meV linear-fit uncertainty. At high field (E > 0.23 V/nm), \({IX}\) becomes bright and dominates the PL.

The Stark shift of \({IX}\) indicates that it has an out-of-plane dipole, a signature of interlayer excitons. By assuming that the electron and hole are localized in different layers, we deduce an electron-hole separation of 0.4 \(\pm\) 0.1 nm based on the Stark shift. This separation is comparable to the interlayer spacing (~0.6 nm) of the heterobilayer46, providing evidence for the origin of interlayer excitons. When we extrapolate the Stark shift of \({IX}\) linearly to zero electric field, we arrive at an energy of ~1.69 eV, which is ~90 meV above the \({A}_{{Mo}}\) line at 1.60 eV. Considering the different exciton binding energies between \({A}_{{Mo}}\) and \({IX}\), we further estimate that the WS2 CBM resides ~40 meV above the MoSe2 CBM, comparable to a predicted 0.03-eV band offset in ref. 47 (see Supplementary Information, Section 3).

Our observation can be readily explained using the schematics in Fig. 1. At low field, the heterobilayer exhibits a type-I alignment with the WS2 CBM lying slightly above the MoSe2 CBM. Consequently, photocarriers relax to the MoSe2 layer to form intralayer excitons (Fig. 1a, b). As the electric field (directed from WS2 to MoSe2) increases, the WS2 CBM is lowered, leading to a transition to a type-II alignment. In the type-II configuration, photoexcited electrons and holes relax to different layers to form interlayer excitons (Fig. 1c, d). We note that an opposite electric field (from MoSe2 to WS2) elevates the WS2 CBM and does not induce the type-II alignment transition. This is consistent with our observation that no \({IX}\) peak appears at negative electric field (Fig. 2c).

Photocurrent measurements

In addition to the striking PL shift, the emergence of interlayer excitons can also drastically affect the optoelectronic charge transport through the MoSe2/WS2 interface. We have measured photocurrent in another BN-encapsulated MoSe2/WS2 heterobilayer (Device 2), which has source and drain contacts with a SiO2/Si back gate (Fig. 3a). We first characterize the device by measuring the interlayer current with no optical illumination as a function of source-drain voltage (Vsd) and gate voltage (Vg) at room temperature. Vsd is applied to the WS2 flake and current is measured from MoSe2 (Fig. 3a). The interlayer current is small at negative Vg and becomes increasingly large at positive Vg (Fig. 3b). This indicates n-type transport mediated by electrons in the conduction bands. At constant positive Vg, the current is nearly zero at Vsd < 0, but increases dramatically with an exponential turn-on at Vsd > 0; such rectifying behavior is consistent with our band scheme in Fig. 1b, where the MoSe2 CBM is lower than the WS2 CBM. This further supports the intrinsic type-I alignment (see more discussions in Supplementary Information, Section 5.1).

Fig. 3: Photocurrent of MoSe2/WS2 heterobilayer.
figure 3

a The schematic of a MoSe2/WS2 heterobilayer device with silicon back gate (Device 2). b Grayscale map of interlayer current I as a function of the source-drain voltage (Vsd) and gate voltage (Vg) with no optical excitation. c Color map of interlayer photocurrent Ipc vs. Vsd and Vg and (d) Cross-cut Ipc-Vsd profiles extracted from panel c at specific Vg values, denoted by dashed lines with corresponding colors. e Color map of the log of the relative rates of multi-particle decay (Auger processes) to single electron-hole pair decay \(\gamma /(\alpha+\beta )\) as a function of Vsd and Vg over the same range as in (c).

Full size image

Afterward we measure the interlayer photocurrent (Ipc) under the excitation of an ultrafast laser. We tune the laser photon energy to be \(\hslash \omega\) = 1.49 eV (\(\lambda\) = 830 nm), which is close to the MoSe2 exciton resonant energy at room temperature (see Methods). Figure 3c displays a photocurrent color map at varying Vsd and Vg. The photocurrent occurs predominantly in forward bias (Vsd > 0 V applied to WS2), consistent with the type I band alignment shown in Fig. 1b. Notably, the photocurrent map exhibits a narrow, curved region where Ipc drops with increasing Vsd. Such an anomalous suppression of interlayer photocurrent can be seen clearly from the Ipc-Vsd line traces at different Vg (Fig. 3d). As Vsd increases, the interlayer photocurrent first increases, then drops in the range of Vsd = 1–2 V, and afterward increases again. Similar non-monotonic Ipc-Vsd characteristics are found for a wide range of Vg from −3.0 to −7.5 V, where the suppression region shifts from lower to higher Vsd values (Fig. 3c). This shift of suppression region with Vg is likely due to the change of contact resistance when the carrier density is modulated by the global silicon gate. A higher contact resistance means a higher Vsd is required to drive an equivalent interlayer voltage drop across the heterobilayer. By imaging the spatial photocurrent response, we confirm that this photocurrent suppression only occurs in the heterobilayer region (Supplementary Fig. S7).

We have extracted the interlayer electric field (E) from Vsd by modeling the device as a p-n junction and fitting its charge transport data (see Supplementary Information, Section 2.2). The anomalous suppression of photocurrent starts near E ~ 0.23 V/nm, comparable to the critical field (E ~ 0.16 V/nm) of interlayer exciton formation observed in the PL map (right axis in Fig. 2c). This suggests that the photocurrent suppression is induced by the interlayer exciton formation.

To clarify the origin of the photocurrent suppression, we have investigated its dependence on the excitation laser power P. Figure 4a shows the photocurrent Ipc at varying Vsd (top axis) and electric field E (bottom axis) under increasing laser power at Vg = −6.5 V. At low laser power, Ipc increases monotonically with increasing Vsd. At high laser power, however, Ipc drops in the range of Vsd = 1.0–1.5 V. As the laser power increases, the photocurrent suppression becomes more severe. By examining the power dependence of Ipc at varying Vsd and Vg, we find that the photocurrent generally increases sublinearly with the laser power (e.g., see the inset of Fig. 4a) and, remarkably, the degree of sublinearity is closely related to the photocurrent suppression.

Fig. 4: Photocurrent signature of type-I to type-II band alignment transition in MoSe2/WS2 heterobilayer.
figure 4

a Photocurrent (Ipc) as a function of source-drain voltage (Vsd, top axis) and interlayer electric field (bottom axis) under various excitation laser powers. The color of each curve represents the excitation power, as indicated by the color scale bar. The inset shows the photocurrent with increasing laser power at Vsd = 1.15 V (E = 0.24 V/nm). The red line is a fit based on the model described in the text. The gate voltage is Vg = −6.5 V in all measurements. b The best-fit \(\gamma /(\alpha+\beta )\) value as a function of electric field. The vertical dashed lines approximately define three regions that correspond to the three scenarios depicted by the lower insets, which, from left to right, illustrate the absence, formation and dissociation of interlayer excitons at increasing field. The measurements were conducted at 830-nm excitation wavelength at room temperature.

Full size image

The photocurrent dynamics under pulsed excitation can be captured by using a simple model. Each laser pulse instantaneously generates \({N}_{0}\) electron-hole pairs, which drops to zero population before the next pulse arrives. The photocurrent is obtained as

$${I}_{{PC}}={ef}\alpha {\int}^{\infty}_{0} N\left(t\right){{{{{\rm{dt}}}}}}$$
(1)

Here \(e\) is the elementary charge; \(f\) is the pulse repetition rate; \(\alpha\) is the carrier extraction rate; \(N\left(t\right)\) is the time-dependent population of electron-hole pairs after the pulsed excitation. \(N(t)\) decays according to the equation

$${dN}/{dt}=-(\alpha+\beta )N-\gamma {N}^{2}.$$
(2)

Here \(\beta\) is the recombination rate of the electron-hole pairs. The \(-(\alpha+\beta )N\) term describes the decrease of carrier population due to the extraction and first-order recombination of photocarriers. \(\gamma\) is the decay rate due to exciton-exciton annihilation or Auger process. When exciton density N increases at increasing laser power, the nonlinear decay term (\(-\gamma {N}^{2}\)) becomes more important, leading to faster exciton decay and photocurrent suppression.

Solving Eq. (2) gives

$$N\left(t\right)=\frac{{{N}_{0}e}^{-\left(\alpha+\beta \right)t}}{1+\tfrac{\gamma {N}_{0}}{\alpha+\beta }\left(1-{e}^{-\left(\alpha+\beta \right)t}\right)}$$
(3)

which, when combined with Eq. (1), yields an analytical expression for the photocurrent under pulsed excitation:

$${I}_{{PC}}=\frac{{ef}\left(\alpha+\beta \right)}{\gamma }{{{{\mathrm{ln}}}}}\left(\frac{\gamma {N}_{0}}{\alpha+\beta }+1\right)$$
(4)

By assuming that \({N}_{0}\) scales linearly with the laser power (P), we can use Εq. (4) to fit the Ipc—P data. The fitting is excellent for a broad range of Vsd and Vg values (e.g., see the line in the inset of Fig. 4a). From the fitting, we extract the relative rate \(\gamma /\left(\alpha+\beta \right)\) (up to a proportionality constant) between multi-particle decay (Auger-like processes) and single electron-hole pair decay at different Vsd and Vg (Fig. 3e).

Figure 4b displays the best-fit \(\gamma /\left(\alpha+\beta \right)\) value as a function of electric field at Vg = −6.5 V. We observe a striking peak of \(\gamma /\left(\alpha+\beta \right)\) in the range Vsd = 1–1.5 V, which coincides with the E-field range where the photocurrent is suppressed in Fig. 4a. To consolidate this observation, we extract the \(\gamma /\left(\alpha+\beta \right)\) value at varying Vsd and Vg (Fig. 3e) and compare it with the photocurrent map (Fig. 3c). The \(\gamma /\left(\alpha+\beta \right)\) enhancement is found to coincide well with the photocurrent suppression. As the carrier extraction rate α and the Auger-like decay rate \(\gamma\) typically evolve smoothly with electric field, the sharp peak of \(\gamma /\left(\alpha+\beta \right)\) implies a sudden decrease of the electron-hole recombination rate \(\beta\). This is consistent with the formation of interlayer excitons in a type-I to type-II transition because the interlayer excitons have much smaller recombination rate (longer lifetime) than the intralayer excitons due to the spatial separation of electrons and holes.

The insets of Fig. 4b illustrate how the interlayer exciton formation may account for the observed photocurrent behavior. At weak interlayer electric field (left inset), the heterobilayer has the type-I band alignment and hence exhibits rectifying behavior, in which the photocurrent increases monotonically with increasing interlayer field. When the electric field reaches a critical value Ec ~ 0.15 V/nm (comparable to Ec ~ 0.16 V/nm in the PL results in Fig. 2c), the heterobilayer transitions from type-I to type-II band alignment, enabling the formation of interlayer excitons (middle inset). The interlayer exciton formation traps the carriers at the interface, simultaneously suppressing the photocurrent and reducing the photocarrier decay rate \(\beta\) (i.e. boosting \(\gamma /\left(\alpha+\beta \right)\)). When the increasing electric field becomes strong enough to dissociate the interlayer excitons (right inset), the exciton effect subsides and the photocurrent resumes its normal increasing trend with increasing Vsd.

Besides Devices 1 and 2 presented above, we have also measured Devices 3 to reproduce the major PL results and Device 4 to reproduce both the major PL and photocurrent results (see Supplementary Information, Sections 1, 6 and 7).

In summary, we demonstrate controlled on/off switching of interlayer excitons in MoSe2/WS2 heterobilayers through a type-I to type-II transition, which substantially influences the optical properties and photocurrent behavior. This phenomenon stems from the closely aligned conduction band minima with field-tunable offset between monolayer MoSe2 and WS2, and it is not expected to occur in other TMD heterobilayers with large band offsets47. Our findings establish MoSe2/WS2 heterobilayers as a highly adaptable platform for excitonic research and applications. For instance, one may harness this effect for switching a hypothetical interlayer excitonic Bose-Einstein condensate, tuning exciton potential depth, and realizing depth-adjustable exciton traps within van der Waals heterostructure materials. The tuning mechanism complements other ‘live’ tunable parameters, such as strain, stress, and twist angles, and achieves ‘in-situ’ control of band-engineered exciton behaviors. This integration promises a new level of precision and adaptability in manipulating excitonic properties in these advanced materials.

Methods

Device fabrication

All MoSe2/WS2 heterobilayer devices are fabricated by applying a polycarbonate-based dry-transfer technique to stack different 2D crystals together. The substrates are silicon wafers with 300-nm-thick oxide layer. For the dual-gate Devices 1, 3, 4, we use a polycarbonate stamp to sequentially pick up a thin graphite flake (serving as the top-gate electrode), a thin BN flake (as the top-gate dielectric), monolayer MoSe2, monolayer WS2, a second thin graphite flake (as the contact electrode), another thin BN flake (as the bottom gate dielectric), and a third thin graphite flake (as the bottom-gate electrode). During the stacking process, we align the sharp edges of the MoSe2 and WS2 crystals so that the twist angles between them are expected to be close to 0° or 60°. Afterward, we deposit the stack of materials onto the Si/SiO2 substrate. Finally, we use the standard electron-beam lithography to deposit the gold contacts (70-nm thickness) onto the devices.

For single-gate Device 2 used in the photocurrent experiment, we first use electron-beam lithography to deposit the two gold contacts (as source and drain electrodes) on a Si/SiO2 substrate. Afterward, we use a polycarbonate stamp to transfer a BN flake to cover the area between the two electrodes. Upon this surface with pre-patterned electrodes, we transfer a MoSe2/WS2 heterobilayer stack by using a large thin BN flake to sequentially pick up monolayer MoSe2 and monolayer WS2. We align the sample position so that the WS2 layer contacts one electrode and the MoSe2 layer contacts the other electrode. This allows us to apply a bias voltage between the two layers.

Photoluminescence experiments

The photoluminescence (PL) experiments are performed in a closed-cycle cryostat (Montana), where the sample temperature is estimated to be T ~ 6 K. The excitation light source is a 532-nm continuous-wave laser (Torus 532, Laser Quantum). The laser is focused onto the sample with a spot diameter of 1~2 μm by an objective lens (numerical aperture 0.6). The incident laser power is P ~ 3 μW. The PL is collected by the same objective and analyzed by a spectrometer (HRS-500-MS, Princeton Instruments) equipped with a charge-coupled-device (CCD) camera. Two Keithley K2400 source meters are used to independently control the top and bottom gate voltages.

Photocurrent experiments

Photocurrent experiments are performed in vacuum in a customized Janis Research ST-3T-2 optical cryostat. Device 2 is measured at room temperature; Device 4 is measured at T = 50 K. The light source is an ultrafast Coherent Mira laser that generates pulses with 150-fs duration and 75-MHz repetition rate. The laser wavelength is tuned to either 790 nm or 830 nm, close to the optical band gap of monolayer MoSe2 as well as the energy of the interlayer exciton. The laser is focused onto the sample with a spot diameter of ~2 µm using a Thorlabs gradient-index (GRIN) lens. The laser position on the sample is controlled by a Thorlabs galvanometer. The galvo position, Vsd, and Vg are controlled by two data acquisition cards (DAQs) from National Instruments. We measure the interlayer current with a pre-amplifier (DL Instruments 1211). The optically induced current is extracted from the total current by using a lock-in amplifier (Stanford Research) and optical chopper.