Bright electroluminescence in ambient conditions from WSe₂ p-n diodes using pulsed injection

Monolayer transition metal dichalcogenides (TMDCs) are a class of single-molecule-thick direct-bandgap semiconductors that show great potential for next-generation electronics^1–5 as well as optoelectronic devices such as light-emitting diodes (LEDs) and photodetectors.^6–13 Past works on TMDC LEDs have shown high electroluminescence (EL) quantum efficiencies of ∼1%–5%.^6,7,14,15 However, these are generally reported under high-vacuum conditions, where environmental factors such as moisture and oxygen are partially avoided. On the other hand, most practical applications require operation under ambient conditions. A recent work on vertical light-emitting WS₂ tunneling structures showed sample degradation after only a few minutes of continuous operation in an ambient environment,¹⁶ highlighting the importance of improving the stability under such conditions. In this work, we show that lateral-junction TMDC LEDs under ambient conditions exhibit fast exponential decay in light emission and current over a period of seconds, similar to past reports on current decay in TMDC field-effect transistors.^17,18 Inspired by a report on bright electroluminescence from pulsed TMDC capacitors,¹³ we propose pulsed bias as a way to overcome this problem in light-emitting diodes, and show that it yields bright and stable EL from monolayer WSe₂ LEDs under ambient conditions. The EL quantum efficiency is ∼1%, close to that of photoluminescence (PL). This implies that the EL efficiency in our devices is primarily limited not by charge carrier confinement, but by intrinsic material quality. In addition, time-resolved electroluminescence under pulsed injection reveals fast rise and fall times of ∼15 ns, indicating strong potential for high-speed light modulation. Under pulsed bias, light emission is stable over hours. Our work is a key step toward realizing fast and efficient optical communication on the nanoscale using TMDC light emitters.

Our device design [Fig. 1(a)] is a back-gated WSe₂ FET structure operated as a p-n diode. The gate stack is either 50 nm SiO₂ on a p++ Si substrate, or 20 nm Al₂O₃ on a transparent ITO/glass substrate. Devices on both substrates showed bright light emission. WSe₂ was chosen due to its ambipolar nature, required for bipolar carrier injection. In general, most devices were ambipolar or slightly p-type in their I_d-V_g characteristics (supplementary material Figs. 1 and 7). Identical structures using monolayer MoS₂ showed only n-type I_d-V_g and no light emission (supplementary material Fig. 2). For diode operation, one contact is designated as the P contact with alternating the voltage between V_p and V_g, while the other (N) contact alternates between V_n and V_g [Fig. 1(b)]. Light is emitted only during the on period t_on. Figure 1(c) shows the EL spectrum of the device overlaid with the photoluminescence (PL) spectrum of the same device. The clear peak at ∼1.65 eV shows that EL and PL are both due to the usual recombination of A excitons and confirms the monolayer nature of the flakes. PL quantum efficiency vs pump intensity shows a drop at high pump intensity due to exciton-exciton recombination, similar to past work¹⁹ (supplementary material Fig. 3). Figure 1(d) shows the light emission overlaid on an image of the device, confirming emission comes from the channel region.

To illustrate the need for pulsed bias, we measure EL vs time and current vs time for both the DC (step) voltage and the pulsed voltage at 5 kHz. Bright emission is observed under both DC and pulsed biases. Note that the emission mechanism in our devices is clearly distinct from the pulsed light-emitting capacitors reported previously,¹³ since continuous light emission is seen on a scale of ∼seconds, as opposed to the ∼10 ns pulses that occur only during voltage transitions. However, under DC bias, both emission and current rapidly decay by orders of magnitude within a few seconds [Figs. 2(a) and 2(b)]. Past reports on current decay in MoS₂ FETs show roughly comparable time constants ∼10 s.^17,18 The ratio of light emission to current, which is proportional to the efficiency, remains roughly constant over time, showing that the decay in current is responsible for the decay in light emission. In contrast, pulsed bias yields extremely stable light emission and current over >1000 s [Figs. 2(c) and 2(d)]. Most devices under pulsed bias showed no decrease or a small decrease (<2×) in light emission in the first few minutes, then remained stable for the remaining duration of the applied bias, with the longest test performed being >3 h (supplementary material Fig. 4). The frequency response is shown in Fig. 2(e). Here, the emission intensity is defined as the average intensity over 10 s, starting 5 s after the pulsed bias is applied. Above ∼1 kHz, the emission intensity is relatively stable with frequency, while below ∼1 kHz, current decay causes EL to drop very quickly. However, note that the stability also depends on the duty cycle, which is fixed at 50% here. For a very low duty cycle (t_off ≫ t_on), the emission intensity can still be stable at low frequencies ∼0.1 Hz, i.e., a quasi-DC regime (supplementary material Figs. 10 and 11), indicating no permanent device degradation at DC bias (see the supplementary material for details on the causes of current decay).

Next, we study the emission characteristics under varying bias to extract the optimal bias conditions. The relative injection level of electrons and holes, and thus the exciton formation efficiency, depends on the voltages (V_p − V_g) and (V_n − V_g) applied to the contacts. If |V_p − V_g| ≫ |V_n − V_g|, holes will be predominantly injected and will simply diffuse across the channel to the opposite contact without forming excitons, and vice versa for |V_p − V_g| ≪ |V_n −V_g| [Fig. 3(a)]. We determine the optimal bias condition by keeping (V_p − V_n) constant and sweeping V_g from V_n to V_p while tracking the EL and current. Figure 3(b) shows the EL intensity vs gate voltage along with the current and EL efficiencies, where V_p and V_n are fixed at V_p = 5 V and V_g = −5 V. Each point corresponds to one on period t_on = 0.5 s, and the current and emission intensity are defined as the time-average during the on period. An off time of t_off = 10 s is used to recover the device between pulses (supplementary material). The brightest emission occurs near V_g = 0 V, where both holes and electrons are injected. The minimum current point at V_g ≈ 1.8 V corresponds to the most bipolar injection and coincides with the highest internal quantum efficiency (IQE) of ∼1%. |V_p − V_g| < |V_n − V_g| at this point, indicating slightly p-type characteristics typical for monolayer WSe₂. We obtain an L-I curve under optimal bias conditions by first setting $Vg=1.25$ V, approximately the point of maximum efficiency above which the current is also not too low. Then, we increase the source-drain bias, while keeping |V_p – V_g|/|V_n – V_g| constant at ∼0.6. For comparison, we also take an L-I curve with V_g = 0 V (i.e., equal |V_p − V_g| and |V_n − V_g|). L-I and IQE are plotted in Fig. 3(c), along with the corresponding PL L-L curves. PL shows a roughly constant IQE of ∼1.6%, while EL at an optimal bias shows a peak of ∼1.3% at low current and decreases slightly at higher currents, possibly due to variation in optimal bias conditions with voltage. In contrast, EL at equal P and N biases is low (∼0.3%) at low current and stays below ∼0.8% throughout.

Next, we study the modulation speed of the device by performing time-resolved measurements using time-correlated single-photon counting (TCSPC). Supplementary material Fig. 5 shows the measurement setup and Fig. 4 shows time-resolved electroluminescence at 1 MHz, together with the P and N voltage pulses measured on an oscilloscope. Emission occurs only when both P and N voltages are applied, confirming bipolar carrier injection. The emission intensity is nearly constant during the entire P/N voltage overlap period. The rise and fall times shown in the inset are ∼12 ns and ∼18 ns, respectively. We note that this rise/fall time is ∼20× faster than monolayer LEDs using vertical tunnel injection heterostructures,²⁰ which is due to the lower capacitance of the lateral injection scheme used here. Tunnel junctions require extremely thin oxide layers (∼1–2 nm) for current injection, leading to high capacitance. The rise and fall times are limited by the function generator used, as can be seen in the voltage pulses measured on the oscilloscope (Fig. 4, top). Therefore, the actual device speed is likely faster, ultimately limited by the intrinsic radiative lifetime of ∼few nanoseconds.²¹ 

In this work, we have shown that the hysteresis and current decay commonly seen in TMDC transistors also play an important role in light-emitting diodes. Pulsed injection is an effective way to circumvent this issue, yielding bright and stable EL using a simple back-gated FET structure, with efficiency near that of PL. We show how to extract the optimal bias conditions for efficient bipolar injection and study the high frequency behavior of light emission. A fast ∼15 ns rise/fall time is observed, indicating strong potential for high speed light modulation. Pulsed emission is stable over hours of operation. Further improvements in efficiency will come from contact optimization to enable lower voltage operation, as well as advances in chemical vapor deposition (CVD) growth to improve intrinsic material quality. Higher speed can be obtained by coupling to an optical cavity to enhance spontaneous emission rate.^20,22–24

See the supplementary material for additional tests on the causes of current decay, description of IQE calculation, and further device and material characterization data.

Device fabrication and characterization were funded by the Center for Energy Efficient Electronics Science through a grant from the National Science Foundation (NSF Award No. ECCS-0939514). K.H. was supported by the Department of Defense (DoD) through the National Defense Science & Engineering Graduate Fellowship (NDSEG) Program. WSe₂ growth was supported by the Electronic Materials Program funded by the Director, Office of Science, Office of Basic Energy Sciences, Materials Sciences and Engineering Division of the U.S. Department of Energy, under Contract No. DE-AC02-05Ch11231.