Report

Electrically Switchable Chiral Light-Emitting Transistor

See allHide authors and affiliations

Science  16 May 2014:
Vol. 344, Issue 6185, pp. 725-728
DOI: 10.1126/science.1251329

Controlling Chiral Light Emission

Circularly polarized light plays important roles in a number of applications such as displays, communication, and sensing. Thus, the ability to produce compact and readily controllable polarized light sources is important, and dichalcogenide materials such as tungsten diselenide may provide a route to such sources. Zhang et al. (p. 725, published online 17 April; see the Perspective by Zaumseil) formed an electric-double-layer transistor structure with WSe2 and used a gated ionic liquid to control the carrier density. Electrical control of the output light was achieved with the polarization being switched by reversing the polarity of the applied field and injected charge.

Abstract

Tungsten diselenide (WSe2) and related transition metal dichalcogenides exhibit interesting optoelectronic properties owing to their peculiar band structures originating from the valley degree of freedom. Although the optical generation and detection of valley polarization has been demonstrated, it has been difficult to realize active valley-dependent functions suitable for device applications. We report an electrically switchable, circularly polarized light source based on the material’s valley degree of freedom. Our WSe2-based ambipolar transistors emit circularly polarized electroluminescence from p-i-n junctions electrostatically formed in transistor channels. This phenomenon can be explained qualitatively by the electron-hole overlap controlled by the in-plane electric field. Our device demonstrates a route to exploit the valley degree of freedom and the possibility to develop a valley-optoelectronics technology.

Circularly polarized light is used in various fields, ranging from three-dimensional displays to technologies involving effective spin sources in spintronics (1) and information carriers in quantum computation (2), as well as to induce exotic quantum phenomena such as the Floquet topological state (3, 4). There is a strong demand for circularly polarized light sources having both compactness for achieving a high level of integration and electrical controllability of the polarization. However, currently available light sources are incapable of meeting these requirements. For example, optical filters (composed of a polarizer and a quarter-wave plate) require mechanical movement to change the polarization; spin light-emitting diodes (LEDs) (5) do not work without an external magnetic field; and metamaterials (6) or chiral polymer LEDs (7) can produce only a fixed circular polarization, with a specific material configuration. Circularly polarized luminescence from solids is known to be a consequence of direct-gap interband transitions of electrons from the conduction band to the valence band with a total angular momentum shift of ±1. Thus, in order to achieve electrical controllability of the polarization of the luminescence, the minimum requirement is that these transitions must be associated with a degree of freedom that responds to an electric field.

Transition-metal dichalcogenides (TMDs), especially in the monolayer limit, is a promising candidate for such devices because two interband transitions are associated with not only the spin degree of freedom but also the momentum degree of freedom (8), which is indeed related to the electrical current. Monolayers of TMDs consist of a triangularly aligned transition metal [typically molybdenum (Mo) or tungsten (W)] plane sandwiched by two triangularly aligned chalcogen [typically sulfur (S) or selenium (Se)] planes. Each metal atom is surrounded by six chalcogen atoms, forming a trigonal prismatic structure (Fig. 1A). Monolayer TMD shares many properties with graphene-like materials with different A- and B-site atoms, in which inversion symmetry is broken (9). In monolayer TMD, electron and hole valleys exist at K and K’ points in the hexagonal Brillouin zone, leading to a direct-gap semiconductor nature (Fig. 1B). Because the electronic states of the two valleys have different chiralities because of the inversion asymmetric crystal structure, interband transitions at K and K’ points are allowed for σ+ and σ circularly polarized light, respectively (valley circular dichroism) (8). These features are suppressed in bulk single crystals because the inversion symmetry is recovered while the band structure changes from direct gap to indirect gap (10).

Fig. 1 EDLTs in action.

(A) Crystal structure of monolayer TMD. (B) Illustration of the band structure of monolayer TMD. (C) Device structure of TMD EDLT under ambipolar charge accumulation. (D) Schematic band structure of EDLT-induced p-i-n junction under equilibrium. (E) Transfer curve (IDS versus VG) of monolayer WSe2 EDLT. (F) Output curve (IDS versus VDS) of WSe2 EDLT. (Inset) Typical optical micrograph of the device. (G) IV characteristics of WSe2 EDLT at 220 K and 160 K, which are higher and lower than the glass transition temperature of N,N-diethyl-N-(2-methoxyethyl)-N-methylammonium bis(trifluoromethylsulphonyl-imide (DEME-TFSI), respectively. Illustrations show charging status at 220 K. (H) Log |IDS| versus |VDS| plot of the blue curve in (G).

Circularly polarized photoluminescence (PL) from TMDs has been experimentally observed for monolayers (1115) and for biased bilayers (16) when they were pumped with circularly polarized incident light. Here, we report current-induced circularly polarized electroluminescence (EL) from p-i-n junctions in monolayer and multilayer tungsten diselenide (WSe2) (17). Moreover, the polarization was found to be electrically controllable by changing the current direction.

We formed p-i-n junctions using an electric-double-layer transistor (EDLT) structure with WSe2 thin flakes having thicknesses of 0.6 to 2.6 nm as the channel material (18). Because TMD–field-effect transistors (FETs) show ambipolar transistor behavior (1924), controlling the potential difference between the gate, source, and drain electrodes enables the simultaneous accumulation of both electrons and holes inside the channel (Fig. 1C), leading to light emission from the p-i-n junction, as demonstrated in organic ambipolar FETs (25, 26). EDLT-induced carriers in the channels can be electrostatically fixed when the device is cooled down under such bias conditions, forming a bias-independent p-i-n junction (Fig. 1D) (19).

EDLT has an additional benefit of eliminating the strong limitation of monolayers (2730). Naïvely thinking, multilayers (bulk material) are not suited for realizing circularly polarized EL. The restored inversion symmetry suppresses valley circular dichroism. In addition, multilayers are indirect-gap semiconductors, in which the conduction band minimum and the valence band maximum locate at T and Γ points, respectively (10). EDLTs can overcome these difficulties. First, the gate electric field breaks the inversion symmetry (16, 24). This recovers the peculiar features of monolayers, including valley circular dichroism (16). Second, because EDLTs produce an electric field one order of magnitude stronger than do conventional FETs, drastic band modulations occur: The valence band top is replaced by K points, and the energy difference between T and K points of the conduction band gets smaller (24). By combining these band structure changes with the large carrier accumulation capability of EDLTs (~1 × 1014 cm−2), not only holes but also electrons can populate the K points. Therefore, multilayers under an ELDT operates in a similar manner as monolayers (24).

Typical transistor characteristics of our WSe2 EDLTs are displayed in Fig. 1E (transfer curve) and Fig. 1F (output curve), showing clear ambipolar operation. The carrier density and carrier mobility are estimated to be of the order of 1013 cm−2 and 102 cm2 V−1 s−1, respectively. As demonstrated in (19), we controlled gate voltage (VG) and source-drain voltage (VDS) to drive our transistor into the ambipolar region in the output curve measurement (Fig. 1G, red curve) and cooled it down to 160 K in order to freeze the gate dielectric, forming a bias-independent p-i-n junction. Source-drain current (IDS) showed a contrasting response at 160 K, with a clear rectifying operation (Fig. 1G, blue curve). The rectification was further confirmed with a log-scale plot (Fig. 1H).

When the WSe2 p-i-n junction was forward-biased, clear EL was observed (Fig. 2A). To enhance the EL efficiency, all spectra were recorded at 100 K unless otherwise specified. Two features should be emphasized. First, all EL spectra showed net circular polarization. The degree of circular polarization, defined as η = [I+) – I)]/[I+) + I)], reached as high as 45% (device 1 at 40 K), which is comparable with that of PL from monolayers (1214). Second, such circularly polarized EL was clearly observed even in multilayer samples (devices 2, 3, and 5), as explained above (18). The EL intensity increased with increasing bias voltage (Fig. 2B). The total intensity depended linearly on the forward current (Fig. 2C), whereas η and the external quantum efficiency (EQE) showed smaller current dependence (Fig. 2, D and E). The EQE value depended on the samples and ranged from 2 × 10−5 to 6 × 10−4, which is larger than that of the emissions from contact Schottky diodes (27), possibly because of the location of the emission zone in our devices, which is away from the contacts.

Fig. 2 Circularly polarized electroluminescence.

(A) Polarized EL spectra from five devices. Spectra of device 1 were measured at 40 K. (Inset) Optical micrograph of device 2 and charge-coupled device image of EL. The yellow dotted curves denote WSe2 thin flakes. (B) Voltage dependence of EL spectra of device 5. (C) Current dependence of total EL intensity extracted from Fig. 2B. (D) Current dependence of EL polarization extracted from Fig. 2B. (E) Current dependence of external quantum efficiency extracted from Fig. 2B. (F) PL spectrum of monolayer WSe2 pumped with a He-Ne laser (1.97 eV).

The EL spectral shape depended on the samples (Fig. 2A), and all of them can be considered as a superposition of exciton emission around 1.66 eV (by comparing with the PL spectrum in Fig. 2F) and trion emission at lower energies, which have been observed in PL from electrostatically charged TMDs (14, 15, 31). The variation of the trion emission energy can be ascribed to the difference in doping concentrations among the samples (14). The emission type is expected to be controlled by means of carrier injection into the i-region (18).

The EL spectra of a single device for two opposite current directions is shown in Fig. 3A. Because the ionic liquid freezes at 100 K, the source and drain biases were exchanged after the device was warmed up to 220 K, followed by another cooling before the next EL measurement (32). The circular polarization was reversed when the source-drain bias was exchanged. This is the first demonstration of electrical control of circularly polarized luminescence. The EL spectral shape was also changed from an exciton-dominant one to a trion-dominant one, possibly because of the difference in doping profiles between the two bias directions (18).

Fig. 3 Electrical control of circularly polarized emission.

(A) Circularly polarized EL spectra from device 4 for two opposite current directions schematically indicated in the top illustration. Bottom illustrations represent the contribution to EL from two valleys. (B) Schematic illustrations of electron and hole distributions shifted by the electric fields originating from the built-in potential. Red and blue lines represent contours of electron and hole distributions, respectively. Orange and green areas represent the electron-hole overlap for K and K’ valleys, respectively.

According to the band structure of monolayer TMDs (Fig. 1B), circularly polarized luminescence only occurs when K and K’ valleys make inequivalent contributions. Here, we suggest a possible model from the viewpoint of semiclassical transport theory (33), which gives information on the carrier distribution under an electric field. Under a static electric field, the hole distribution shifts parallel to the field in the momentum space, whereas the electron distribution shifts to the opposite direction. As a result, the electron-hole overlap shown in Fig. 3B differs between the two valleys [valley overlap polarization (VOP)] because the carrier distributions in TMDs are not isotropic owing to trigonal warping (34). The shift discussed here is much larger than the Fermi surface shift in metals because the carrier density in the i-region is substantially low and the field strength is much stronger. VOP provides an intensity difference in the luminescence from two valleys, leading to circularly polarized luminescence. The above scenario is widely applicable to TMDs that have two isolated anisotropic valleys with different chiralities. In addition to the observations in WSe2, circularly polarized EL has also been observed in MoSe2 and MoS2 (18).

A numerical calculation was performed according to this model (18). To understand the carrier distribution in the momentum space, we combined the Boltzmann equation (33) and the charge distribution equation for a p-n junction (for simplicity, a p-n junction model was used rather than a p-i-n junction) (35). The effects of excitons and trions were ignored because we are only interested in the asymmetry. As shown in Fig. 4A, the band dispersions, especially that of the valence band, is anisotropic owing to trigonal warping (34). When an in-plane electric field is applied, the charge distribution is shifted from its equilibrium position (Fig. 4B). The consequent electron-hole overlap differs between K and K’ points, giving rise to VOP. Simulated luminescence spectra from K and K’ points, corresponding to σ+ and σ, respectively, are shown in Fig. 4C. The obtained polarization is qualitatively consistent with the experimental results.

Fig. 4 Simulation of circularly polarized electroluminescence.

(A) Schematic band structure around K point. Momentum along the x direction (kx) and along the y direction (ky) are measured from K point. (B) Color plot of charge distribution for electron at conduction band (fc) and hole at valence band (fv) around K point. Dotted lines denote kx = 0 and ky = 0. (C) Simulated EL spectra from K point (σ+) and K’ point (σ). (D) η versus electric field direction. The angle θ is measured from the kx direction. VOPs are shown in top illustrations for three electric field directions along Γ-K (θ = 0°), Γ-M (90°), and Γ-K’ (180°).

Because trigonal warping is related to the crystal symmetry (Ĉ3), this scenario indicates that η and the luminescence intensity are quite sensitive to the relative angle between the crystal orientation and the field direction, exhibiting threefold symmetry with respect to the rotation of the electric field direction, as clearly shown in the simulated results in Fig. 4D. Hence, if the current path is not perpendicular to the contact, η does not completely reverse its sign when the electrodes are exchanged, as shown in fig. S6. On the other hand, this feature will enable fine tuning of η by using a multiterminal device to apply biases in directions that are nonparallel to the p-i-n junction (fig. S7). To maximize the luminescence intensity, the degree of polarization (value of η), and the polarization tunability (tunability of η), it will be necessary to improve the device fabrication processes in order to accurately align the crystal orientation and electrode configuration.

Circularly polarized EL is a direct indication of the breakdown of valley symmetry. In our experiment, this was induced purely by the electric field, and we had to go beyond linear response theory in order to understand this phenomenon; for example, we were not able to explain this with the valley Hall effect (8). Therefore, we proposed a model that incorporates nonlinear processes from the carrier distribution shift, which becomes asymmetric from trigonal warping, leading to VOP.

Owing to their spin-valley coupled band structure (8), TMD-based spin LEDs can also exhibit circularly polarized EL. However, such devices cannot work without an external magnetic field (1). Our results, on the other hand, provide a new direction in the quest for electrically switchable circularly polarized light sources and extend the functionality of valley-based optoelectronics.

Supplementary Materials

www.sciencemag.org/content/344/6185/725/suppl/DC1

Materials and Methods

Supplementary text

Fig. S1 to S7

References (3640)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. This process is not essential for circularly polarized EL and polarization reversal. Also, electrical switching can be achieved with solid gate devices, in principle.
  3. Acknowledgments: We thank Y. Saito, H. Shioya, Y. Kasahara, T. Hatano, and Y. Segawa for experimental help and R. Akashi, R. Arita, T. Takenobu, H. Aoki, and B. P. Zhang for discussions. Y.J.Z. is supported by the Japan Society for the Promotion of Science (JSPS) through a research fellowship for young scientists. This research was supported by the Strategic International Collaborative Research Program (SICORP-LEMSUPAR) of the Japan Science and Technology Agency, Grant-in-Aid for Specially Promoted Research (25000003), and the “Funding Program for World-Leading Innovative R&D on Science and Technology (FIRST Program)” from JSPS.
View Abstract

Stay Connected to Science

Navigate This Article