Electroluminescence from μLED without external charge injection.

Stable electroluminescence from micro-pixelated light-emitting diode (μLED) occurs when electrons and holes are continuously injected from external electrodes. Different from the general recognition, in this work, μLED works in an operation mode, namely, non-electrical contact and non-carrier injection mode, and can be 'wirelessly' lit up without external charge injection, which is different from the general recognition. Inherent holes and electrons in μLEDs can provide sufficient carriers for radiative recombination under alternating-current electric field. A possible model related to the diffusion of majority carrier and the drift of minority carrier in μLED was proposed, which is further confirmed by the employment of a 'carrier pump'. Finally, the intrinsic characteristics of the device-in-capacitor, namely, self-protection against electrical breakdown, were discussed. This work demonstrates a new device configuration and an alternative operating mode for μLED and provides a research manner to obtain advanced μLED-based technology.


Results
A polyethylene terephthalate (PET, 125 μm thickness) layer s used as an insulating medium between the μLED and an external indium tin oxide (ITO) electrode, which can completely suppress external carrier injection. The stacked PET/μLED/PET component is sandwiched between two ITO electrodes, as schematically illustrated in Fig. 1a (the fabrication process is shown in Fig. S1). An input AC voltage is applied to the ITO electrodes to light the μLED. Our devices have the following advantages compared with conventional μLED: (1) The μLEDs used 1 College of Physics and Information Engineering, Institute of Optoelectronic Technology, Fuzhou University, Fuzhou, 350108, China. 2 Fujian Science & Technology Innovation Laboratory for Optoelectronic Information of China, Fuzhou, 350108, China. 3 These authors contributed equally: Kun Wang and Ye Liu. ✉ e-mail: nnnwcx@foxmail.com; xtzhou@fzu.edu.cn; gtl_fzu@hotmail.com open in this work have a simple structure of u-GaN/n-GaN/MQW/p-GaN (Fig. 1b). Conventional μLED requires a transparent contact layer and an upper p-electrode to achieve efficient carrier injection (Fig. S2) 12,13 . However, the transparent contact layer, p-electrode, and n-electrode were eliminated in this work because external carrier injection was avoided. Their removal allows a more cost-effective mass production process for μLED-based electronics. (2) Our device does not require charge injection from external materials; therefore, the elaborate design of energy band alignment for electron (hole) injection layer, transport layers, and emissive layers was eliminated 14,15 . (3) The ready-made μLEDs have higher environmental stability and better performance compared with the light-emitting materials used in organic (quantum-dot) LEDs. (4) The output performances of our device are independent of the μLED orientation as shown in Fig. 1c,d. Furthermore, we can observe the emitted light on both sides of the device because of the transparency of the 7 μm-thick μLED (Fig. S3).
The possible operation processes of our device are schematically illustrated in Fig. 1e, We simplify the μLED to a p-n junction and ignore the effect of MQW on the carrier motion in the following description. Electrons diffuse into the p-GaN region and holes diffuse into the n-GaN region under thermodynamic equilibrium to form a space charge region and a built-in field 16 . Electrons (majority carriers) in the n-GaN region move toward the MQW by diffusion under the forward bias. A similar process is performed on the p-side of the junction. Radiative recombination of electrons and holes occurs in the MQW region ( Fig. 1e-I). However, light emission under the forward bias is transient, and only one flash can be observed when a constant bias is applied (Video S1). The reason is that the drift of majority carriers would cause the formation of a depletion region (or an inversion layer under a strong field) and generate an induced electric field to shield the external field ( Fig. 1e-II). The shielding effect tends to limit the majority carrier's diffusion and finally terminates the radiative recombination process.
The previously induced field would be removed due to the drift of majority carriers when the polarity of the applied bias was inverse, as schematically shown in Fig. 1e-III. In addition, due to the strong built-in field, a net www.nature.com/scientificreports www.nature.com/scientificreports/ current flowing in a direction from n-GaN to p-GaN is generated due to the strong built-in field. Simultaneously, the drift of the majority carriers in the bulk regions finally forms accumulation layers at the ends of p-GaN and n-GaN, respectively, and generated an induced electric field to shield the external field, as illustrated in Fig. 1e-IV. The device can maintain its periodic electroluminescence under AC voltage; therefore, we can infer that the total amount of charge transferred through the p-n junction under forward bias is the same as that under reverse bias. We propose a strategy based on this inference to enhance the electroluminescence, and this strategy will be discussed later. Figure 2a shows the time-resolved electroluminescence. Electroluminescence occurs only in the positive half of the period, which is consistent with our proposed model. In addition, luminescence intensity rapidly increases to its peak value and then gradually decrease in the rest time of the positive bias, because the gradual formation of an induced field under positive bias would shield the external field and gradually reduce the electroluminescence intensity.
From the perspective of circuit analysis, a μLED is equivalent to resistor-capacitance (RC) parallel circuits, where R LED is the resistance of bulk semiconductor and MQW, and C LED is the capacitor related to p-n junction 16 . Furthermore, our device can be regarded as an RC circuit due to the presence of insulating PET layers, as shown in Fig. 2b. Therefore, the device is sensitive to the frequency of the applied AC voltages. The recorded current greatly depended on the operating frequency, and two corner frequencies (f 1 and f 2 ) are observed, as shown in Fig. 2c. The external capacitor (C EX ) related to PET layers plays a major role in the impedance change at operating frequencies below f 1 , and the recorded current increase with frequency. When the frequency is slightly higher than f 1 , the capacitive reactance of C EX is almost zero, and the current is independent of frequency. However, www.nature.com/scientificreports www.nature.com/scientificreports/ further increase in frequencies makes C LED play a major role in impedance change. The capacitive reactance of C LED decreases with increasing frequency when the operating frequency is slightly lower than f 2 , and the current increases with frequency. When the frequency is higher than f 2 , the capacitive reactance of C LED is almost zero when the frequency was higher than f 2 , and the current approaches saturation again.
Electroluminescence intensity is also sensitive to operation frequency. Figure 2d presents the luminancefrequency characteristics of the device under different applied voltages. Electroluminescence is too weak to be observed in the low frequency range (<1k Hz). The electroluminescence intensity dramatically increases at a turn-on frequency of about 1k Hz (defined as f on ) and continues to increase with frequency and finally reaches the maximum at the saturation frequency (f sat ).
The corresponding differential current-frequency, current-frequency and electroluminescence-frequency characteristics are provided in Fig. 2e to show the relationship between the recorded current and the electroluminescence intensity clearly. For the differential current-frequency curve, the frequency (first peak) where the current changes rapidly is the same as f on . Thus, we can conclude that the increase in frequency (near f on ) dramatically increases the operating current and causes more radiative recombination in the MQW region. According to circuit theory, the capacitive reactance of C EX decreases as the frequency increases and thus increases the operating current and the voltage component applied to the p-n junction. Thus, electroluminescence intensity increases with frequency. However, when the frequency is further increased to a range between f sat to f 2 , the C LED plays the major role in the current change, and the electroluminescence intensity decreases with increasing frequency. From circuit theory, the decrease in C LED capacitive reactance would decrease the current flowing through the R LED and thus result in a decrease in the number of recombined carriers in the MQW region. Because the transit-time of the majority carriers from the bulk semiconductor to the MQW is constant, as the operating frequency increases, the number of majority carriers that can diffuse into the MQW during the positive half of the period decreases. In other words, the carrier transport cannot keep up with the changes in the applied field and will result in a decrease in electroluminescence intensity 17 . An extremeness case is that all majority carriers do not have enough time to diffuse to the MQW when the frequency increases to f 2 . As a result, no electroluminescence can be observed, as shown in Fig. 2e.
We calculate the current-frequency relationship of the μLED operating in NEC&NCI mode by using phasor analysis. Note that phasor analysis does not apply to nonlinear circuit elements. However, the electrical properties of a μLED operating in the NEC&NCI mode are different from that of a conventional μLED under DC voltage. As will be discussed later, the total charge carriers transmitted through the p-n junction under forward bias should be the same as that under reversed bias. Thus, the rectification characteristics of the device are not obvious. Furthermore, the measured current curve has a sinusoidal shape ( Figure S4). Therefore, we believe that phasor analysis has certain reference value for understanding the electrical performances of our device. As shown in Fig. 2f, the theoretical values are in good agreement with the experimental values, which clearly shows the effectiveness of our model (Note S1).
In addition, the current-voltage (I-V) relationship at a fixed frequency is demonstrated, as shown in Fig. 2g. Interestingly, the current is almost linearly related to the voltage in the measure range. This result is different from the I-V relationship of LED in normal operation mode. The possible reason is that the vector sum of resistance and capacitive reactance can be considered constant at a fixed frequency. The linear I-V curve can further prove the effectiveness of our RC equivalent mode. The wall-plug efficiency (η) of this device can be calculated as: where P LED is the output luminous power of LED, P AC is the input power of the AC supply and P R is the power consumption of a current-limiting resistance. As shown in Fig. 2g, the estimated η is 0.37% when the device reaches the maximum brightness at 30 kHz (Note S2 and Figure S5). It is well known that the electroluminescence intensity depends on the number of radiation-recombination carriers (Fig. 1e-I). Radiation-recombination carriers include the electrons diffused from n-GaN and the holes diffused from p-GaN 14 . Note that the diffusion of majority electrons and holes are similar; thus, we only consider the electrons diffused from n-GaN in the following discussion, as shown in Fig. 3. The net diffused electrons (N forward ) in a forward-biased p-n junction that can be divided into two parts: (1) The electrons that recombine with holes (N forward_recom ) that contribute to electroluminescence and (2) the leaked electrons (N forward_leak ) that directly diffuse into the p-GaN region without recombination (Fig. 3a). This ratio can be considered an approximative constant because the ratio of N forward_recom /N forward mainly depends on the μLED structure. Thus, to enhance the electroluminescence, the N forward should be increased.
The total carriers transmitted through the p-n junction under forward bias should be the same as that under reverse bias, because the device can maintain its periodic electroluminescence under AC voltage. Thus, the number of the net minority electrons (N reverse ) in the reverse-biased p-n junction must be the same as the N forward . However, very few electrons drifted through the p-n junction under reverse bias (Fig. 3b). As a result, the N forward that contributes to electroluminescence is clamped to the N reverse. In other word, we can infer that electroluminescence intensity is limited due to the low reversed current under reverse bias.
Based on the above inference, wWe propose a strategy to increase electroluminescence density by employing an antiparallel μLED pair (μLED_I and μLED_II), as schematically presented in Fig. 3c. For the antiparallel μLED pair, the p-GaN terminal of μLED_I is electrically connected to the n-GaN terminal of μLED_II, and the n-GaN terminal of μLED_I is electrically connected to the p-GaN terminal of μLED_II. We only consider the electroluminescence from μLED_I in the following discussion.
The net electrons (N' forward ) driven from terminal_I to terminal_II include two parts when the antiparallel μLED pair is forward-biased (defined as μLED_I is under forward bias): (1) Electrons diffuse through the Scientific RepoRtS | (2020) 10:8059 | https://doi.org/10.1038/s41598-020-65092-z www.nature.com/scientificreports www.nature.com/scientificreports/ forward-biased μLED_I (N' I_forward ), which predominate the N' forward ; (2) Electrons drift through the reverse-biased μLED_II (N' II_reverse ). In this case, the electroluminescence from μLED_I can be observed. Analogously, when the antiparallel μLED pair is under reverse bias, electrons (N' reverse ) will be driven from terminal_II to terminal_I. As shown in Fig. 3d, the N' reverse also includes two parts: (1) Electrons diffuse through the forward-biased μLED_II (N' II_forward ), which makes the most contribution to N' reverse ; (2) Electrons drift through the reverse-biased μLED_I (N' I_reverse ). Obviously, the N' reverse is the same as the N' forward , and N' forward and N' reverse are much larger than the N reverse of the device based on single μLED. Therefore, a large number of electrons can be driven from the p-GaN terminal of μLED_I to the n-GaN terminal through μLED_II even under reverse bias. In other words, μLED_II acts as a highly efficient 'carrier pump' for the reverse-biased μLED_I. A strong electroluminescence can be observed in μLED_I under forward bias. The electroluminescence density and operating current of the antiparallel μLED pair-based device are obviously increased compared with the device based on a single μLED, as shown in Fig. 3e,f. The proposed μLED operating in the NEC&NCI mode can be equivalent to an electronic system of a μLED embedded in a capacitor, namely, device-in-capacitor (DIC). In this DIC, two parallel ITO electrodes and insulator layers constitute a capacitor, and a μLED is inserted into the insulator layers. Therefore, the capacitor connected to external AC power can be considered an equivalent power source for the embedded μLED, as schematically presented in Fig. 4a. Notably, no carrier is injected from the equivalent power source to the embedded μLED, and the equivalent power source exhibits different output characteristics. As shown in Fig. 4b, the output current, voltage and power are highly dependent on load resistance. In this measurement, the external AC power (f = 30 kHz, V pp = 70 V) is applied, and a variable resistor is connected as the external load. Output peak current dramatically decreases as the load resistance increases and the output peak voltage increases with the load resistance. These events are different from those in a normal power source. Consequently, the maximum output power can be obtained at a specific load resistance. Output power as low as 11 μW can be observed even though a high voltage (35 V) is applied. www.nature.com/scientificreports www.nature.com/scientificreports/ Interestingly, a self-protection function naturally exists in the DIC to prevent electrical breakdown due to the low output power. The resistance of nonlinear electronic element, such LED, depends on the applied voltage, as shown in Fig. 4c. The resistance of the LED will decrease when the LED is operating under a high positive bias (point A in Fig. 4d) and therefore reduces operation bias to avoid overcurrent (point B in Fig. 4d). The resistance dramatically decreases when the LED is operated under high negative bias close to the breakdown voltage threshold. As a result, the negative bias applied to the LED would also decreases to avoid breakdown. This reason explains why the μLEDs in our device can operate normally under such high voltage.

Discussion
One of the disadvantages of our device is the high operating voltage and AC frequency. Since the μLED employed in this work is used to fabricate conventional μLED display, it may be not suitable for working in the NEC & NCI mode. The operating voltage and frequency can be reduced by optimizing the structure of the μLED, including the thickness of the p-GaN and n-GaN and the impurity concentration.
Because the electrical contact between the external electrode and the μLED is avoided, the proposed NEC & NCI mode operation mode may be applied to a conventional μLED display. As well known, a large-scale precise electrical bonding technology, which might reduce the rate of working pixels and increase manufacture cost 1,18-20 , must be used for the hybrid integration of μLED arrays with separate driving circuits. The inevitable contact resistance would increase the threshold voltage and heat generation and decrease the device's operating lifetime 21,22 . No electrical contact is required between external electrode and the μLED in our device. Thus, the current complicated microelectronic fabrication process for precise electrical bonding in μLED display might be simplified or avoided when this technology is further developed in the future. Furthermore, the 'wireless' drive technology might promote the applications of μLED, such as in the biomedical field.
In summary, periodic electroluminescence from μLEDs operating in NEC&NCI mode is successfully demonstrated. Electroluminescence without external carrier injection are realised by applying an AC electric field, which can generate the periodic oscillation of inherent carriers in μLED: The forward field causes the diffusion of majority carrier and subsequent radiative recombination in the MQW; the reverse field drifts the carriers to their original state to prepare for the next electroluminescent process. The performances of the μLEDs operating in the NEC&NCI mode are highly sensitive to the applied AC frequency. Maximum electroluminescent intensity can be obtained at approximately 30 kHz. Furthermore, an antiparallel μLED acting as 'carrier pump' is employed to increase the luminescence intensity, which experimentally confirms our proposed theoretical model. Finally, we demonstrate that the load-depended output of DIC leads to the device's intrinsic self-protection against electrical breakdown. Our work presents an operating mode for μLED without external carrier injection and presents a device configuration without electrical contact, which offers a new research manner to obtain advanced μLED-based technology.  separate the 7 μm-thick μLED layer from the sapphire substrate. A PET layer (125-μm thick) was attached on ITO glass to suppress the injection of external carriers completely from ITO. The μLED was first transferred onto the surface of the PET-coated ITO glass. Then, another PET-coated ITO glass was vertically stacked on the former one ( Figure S1). Thus, the μLED was positioned between two vertically stacked PET layers.
For the antiparallel μLED pair, a patterned ITO film (0.1 mm × 5 mm) was pre-formed on a PET layer, and the μLEDs were transferred to the ITO surface. The p-GaN terminal of μLED_I was in contact with the ITO, and the n-GaN terminal of the μLED_II was in contact with the ITO. Another patterned ITO/PET layer was deposited on the μLED pair. Finally, the as-assembled device was sandwiched between the ITO electrodes.
For the fabrication of the DIC, a PET layer (5 mm × 5 mm) was first deposited on ITO glass. Then, a copper layer was deposited on the PET surface. Two layers of ITO glass/PET/Cu were stacked opposite each other, and a 125 μm-thick PET film was inserted between the two copper layers. AC voltage was applied to the ITO electrodes, and the output signals were extracted from the two copper films.
Characterisation. AC sinusoidal waveform voltages were generated from a waveform generator (RIGOL, DG4162) and an amplifier (Agitek, ATA-122D). AC currents were measured by using an oscilloscope (RIGOL, DS1102E), and luminance was measured through a silicon photodiode (EVERFINE, FLA2000). Micrograph images were obtained via a 3D measuring laser microscope (Olympus, OLS4100). All measurements were performed in air at room temperature.