Bidirectional LED as an AC-driven visible-light source

doi:10.21203/rs.3.rs-2741889/v1

Download PDF

Article

Bidirectional LED as an AC-driven visible-light source

https://doi.org/10.21203/rs.3.rs-2741889/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 20 Nov, 2023

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

The GaN-based light emitting diodes (LEDs) brought a revolution in the lighting market by becoming the most energy-efficient light sources. However, the power grid, i.e. electricity delivery system, is built based on alternating current (AC), which raises problems for directly driving LEDs that require direct current (DC) to operate effectively. In this paper, we demonstrate a proof of concept device that addresses this fundamental issue – a GaN-based bidirectional light emitting diode (BD LED). The BD LED structure is symmetrical with respect to the active region, which, depending on the negative or positive bias, allows for the injection of either electrons or holes from each side. It is composed of two tunnel junctions that surround the active region. The optical and electrical properties of such BD LEDs under DC and AC conditions are investigated. We find that the light is emitted in both directions of supplied current, contrary to the standard LEDs, hence BD LEDs can be considered as a semiconductor light source powered directly with AC. In addition, we show that BD LEDs can be stacked vertically to multiply the optical power achieved from a single device.

Physical sciences/Optics and photonics/Lasers, LEDs and light sources/Inorganic LEDs

Physical sciences/Physics/Electronics, photonics and device physics/Photonic devices

The III-nitride light emitting diodes (III-N LEDs) coated with phosphors are nowadays the most efficient white light sources for general lighting^1–4. The most efficient violet III-N LEDs reach around 80% of both external quantum efficiency and wall plug efficiency^5–8. However, there is one crucial disadvantage of using LEDs for general lighting. The electric power in the electrical grid is distributed as alternating current (AC), whereas LEDs emit light only when the junction is forward biased. The AC is therefore converted into direct current (DC) in order to power LEDs. The high quality AC/DC converters are bulky, complex in design and always lead to some inevitable power loss^9,10. It was shown that the absence of AC/DC converter saves up to 20% of electrical power under specific conditions. Great effort was made to fight this inconvenience^11–15. Different approaches to obtain directly AC driven light emitting devices can be gathered in three groups.

The first group of AC LEDs is based on connecting a matrix of several LEDs both in series and parallel. The LEDs can be connected in an anti-parallel configuration, or in a more sophisticated way, to form a rectifier^11,12. The advantage of this solution is that highly efficient III-N LEDs can be used. However, under AC conditions only a part of all LEDs used in the circuit emits light at the same time, which effectively reduces the surface power density received from the chips.

Another group of AC LEDs consists of devices based on organic emissive materials¹³. They are predominantly characterized by a symmetrical structure, in which an emitting layer is sandwiched between insulating dielectric layers and light is generated either by the hot-electron impact excitation mechanism or the radiative recombination of excitons. The disadvantage of this solution is a limited power efficiency, because achieving a proper balance between injected electrons and holes is difficult. In addition, they require a high operating voltage.

The last group of AC-driven devices involves specially designed structures based on III-nitrides, in which the phenomenon of carriers tunneling takes places. The first example of this group is an device presented by Hartensveld et al.¹⁴. The structure contains an InGaN quantum well (QW) and a thin Al₂O₃ barrier on the top of it. There were no p-type layers, thus the holes were injected into the QW by means of tunneling through Al₂O₃ barrier. The authors have demonstrated a planar device with two identical NiO/Al₂O₃ contacts, which were able to inject holes or electrons into the QWs depending the bias condition. The second example is a device presented by Sadaf et al.¹⁵. They have shown that the integration of two oppositely biased III-N LEDs can be realized directly during the epitaxy process by using selective area growth method. Firstly, a standard upper p-type was grown on half of the surface of a silicon wafer. Next, on the second half of the wafer the growth was initiated with a TJ and then an LED structure with inverted p-n junction was grown. This way in the TJ-based LED the junction was in the opposite direction and the current flow was reversed. The two above mentioned devices worked under AC, however in both of these approaches only half of the device area emits light at the same time, because they are equivalent to two LEDs connected in an anti-parallel configuration.

In this article, we propose an alternative way to obtain AC driven device with embedded TJs, which allows to use the full area of the device for lighting. We will refer to this device as a bidirectional light emitting diode (BD LED), because the light is emitted from the same active region both in the forward and reverse current directions. BD LED consists of single III-N QW surrounded by two oppositely oriented TJs. This design is advantageous, as it combines the simplicity of symmetrical structure with high quantum efficiency characteristic for devices based on III-nitrides. Furthermore, several BD LEDs can be stacked vertically to multiply optical power obtained from a single device. The optical and electrical properties of such BD LEDs under DC and AC are investigated.

It should be emphasized here that during the last decade III-nitride TJs have reached maturity and a high tunneling current under low operating voltage^8,16−18. Efficient TJs are used to control the arrangement of piezoelectric fields in the active region^19–22, to stack several LEDs or laser diodes (LDs) for multiplying the optical power or obtaining multi wavelength optical spectrum^23–27, to control current flow in micro LEDs^28–31, in distributed-feedback LDs³² or in tunneling field-effect transistors³³.

BD LEDs structure

The structures of a single and a stack of two BD LEDs are presented in Fig. 1a and 1b, respectively. Samples were grown by plasma-assisted molecular beam epitaxy (PAMBE) on Ga-polar (0001) bulk GaN crystals. The structures are symmetrical with respect to the QW, when it comes to the arrangement of epitaxial layers and their doping levels. The growth was initiated with 200 nm thick GaN:Si (Si: 1x10¹⁹ cm^-3) followed by heavily doped bottom n-p TJ, which consist of 20 nm In_0.02Ga_0.98N:Si (Si: 1x10²⁰ cm^-3) and 10 nm In_0.02Ga_0.98N:Mg (Mg: 1x10²⁰ cm^-3). Then, a 35 nm thick p-type In_0.02Ga_0.98N:Mg region is placed, in which the Mg concentration decreases linearly from 1x10²⁰ cm^-3 to 2x10¹⁹ cm^-3. In the middle of BD LED structure an unintentionally doped (UID) active region is located, which consists of a 10.4 nm wide In_0.17Ga_0.83N quantum well (QW), surrounded by 30 nm In_0.08Ga_0.92N barriers from each side. A relatively wide QW was chosen due to a higher internal quantum efficiency than thin QWs ^34,35. Above the active region, symmetrically with respect to the QW, the same layer stack, but in the inverted order, is placed. Layers consist of 35 nm In_0.02Ga_0.98N:Mg region with the Mg concentration increasing linearly from 2x10¹⁹ cm^-3 to 10²⁰ cm^-3, followed by the top p-n TJ. The top TJ has the same chemical compositions like bottom one. The structure is then capped with 100 nm GaN:Si (Si: 3x10¹⁹ cm^-3). In case of the stack of two BD LED the sequence of layers repeats.

In the case of structures with buried p-type layers, PAMBE is more suitable, because the p-type region is active immediately after growth. In contrast, for structures grown by metalorganic vapor phase epitaxy (MOVPE), the Mg-doped regions must be thermally activated by hydrogen removal whose vertical diffusion through the n-type region is suppressed^36,37. Nevertheless, recently two methods were proposed, which allow for activation of buried p-type layers grown by MOVPE, but creates some challenges for processing: activation by side walls and with selective area growth method^31,38.

The epitaxial structure of stack of two BD LEDs is presented in cross-section TEM image shown in Fig. 1c. Low number of extended defects and sharp interfaces shows that the device stacking process can be used for a larger number of BD LEDs. Such a good crystalline quality results from the earlier investigations of highly doped TJs¹⁷. In that work we found that the doping level up to 2x10²⁰ cm^-3 in TJ does not degrade the crystalline quality and give good electrical performance.

BD LEDs operating principle

We discuss the operating principle of BD LED by analyzing its band diagram. A simplified structure of single BD LED is presented in Fig. 2a, whereas Figs. 2b, 2c and 2d show the band diagrams without, under positive and under negative biases, respectively. Figures 2e and 2f show the magnification of the band structures in the crucial QW region. The bottom TJ is located on the left-hand side, while the top TJ on the right-hand side. Both TJs are able to operate in two modes: carrier tunneling, when polarized reversely, or carrier diffusion, when polarized forwardly, depending on the overall bias of the BD LED.

Under the positive bias of BD LED (Fig. 2c) the top TJ is reversely polarized, interband tunneling of carriers occurs and thus the holes can be injected into the QW from the right-hand side. Simultaneously, the bottom TJ is polarized in the forward direction, so electrons are injected into the QW from the left-hand side. Under the negative bias of BD LED (Fig. 2d) the bottom TJ is reversely, while the top TJ is forwardly polarized. Therefore, the holes are injected from the left-hand side and the electrons from the right-hand side. In other words, when the BD LED is negatively biased the sides from which the carriers are injected into the QW are interchanged. Importantly, in both bias cases (positive and negative) the carriers are injected into the same active region. Therefore, in our concept, the BD LED will emit light from the entire surface of the device under the AC conditions. This distinguishes the proposed device from the devices published in the literature in which only part of the device can emit light at the same time^11,12,14,15.

BD LED properties under DC

The current-voltage (IV) and optical power characteristics of examined devices are presented in Fig. 3a and 3b, respectively. For both positive and negative biases, the IV characteristics resemble the IV curves for forwardly biased standard LED and more importantly, the light is emitted above the opening voltages in both biasing conditions. The opening voltages are about 5 V and − 5.5 V for the single and 9.4 V and − 11.2 V for the stack of two BD LEDs. For current densities of 1 kA/cm² and − 1 kA/cm² the operating voltages are 6.9 V and − 8 V for the single and 12.5 V and − 13.3 V for the stack of two BD LEDs, respectively. As can be seen, the operating voltage is about 1 V lower for positive bias than for negative bias.

In the case of the single BD LED, the collected light power grows approximately linearly with the positive and negative supply currents. However, the luminous power is about 10 times higher for negative than for positive bias at high current densities. Under the low bias condition (below 0.2 kA/cm²) this difference is even greater, what is presented in logarithmic scale in the inset of Fig. 3b. In the case of stack of two BD LEDs the luminous power is roughly 1.8 times higher for all current densities under the negative bias. The situation is different in the case of positive bias. Below 0.6 kA/cm² the light output from the stack is lower than from single structure, but above it raises strongly and finally for a current density of 1 kA/cm² is 1.5 times higher than the value obtained for single BD LED.

The optical spectra of examined devices were measured in steps of 0.2 kA/cm² up to 1 kA/cm² and they are shown in Figs. 3c-f. In the case of the single BD LED for negative bias, we observe only a single peak at λ = 447 nm related to recombination in the QW (Fig. 3c). However, for positively biased BD LED a second peak appears at λ = 428 nm. This parasitic recombination dominates spectra at low current densities, whereas at higher currents the QW peak prevails. In the case of the stack of two BD LEDs, the parasitic peak also occurs under the positive bias (at λ = 419 nm), but its contribution to the spectrum is much smaller. Importantly, under the negative bias a single peak originating from the QW is present(at λ = 454 nm), similarly to the single BD LED. The wavelength of the peaks originating from the QWs in the case of single and stack of BD LEDs is slightly different, due to a small variation between the actual and the intended indium content in the QW during the epitaxial growth. Insets of Figs. 3c-f presents the real color pictures of devices taken under the current density of ± 1 kA/cm². In all cases, the light is emitted from the entire surface of BD LED, but it is slightly more uniform for negative bias. The dark square in the middle is the top metallization, which blocks part of the outgoing light.

We have noticed three main differences in operation between the positive and negative biases: (1) the operating voltage is higher at negative bias, (2) the optical power is over 10 times lower at positive bias and (3) there is parasitic recombination at λ = 428 nm (at λ = 419 nm for the stack of two BD LEDs) at positive bias only. We attribute all of these differences to the presence of built-in electric fields in the III-nitride structures grown on polar substrates, which breaks the symmetry of the band structure. This problem will be analyzed in more detail in the Discussion section.

BD LEDs properties under AC

The BD LED was tested under the sinusoidal voltage condition to check its capability as a directly AC-driven source of light. The single BD LED was powered with 12.6 V peak to peak alternating voltage with a frequency of 50 Hz. Simultaneously, the current passing through the device and light intensity was registered. All data are presented in Fig. 4a. The BD LED stays turned on for more than half of the total pulse period and importantly it emits light in both half-cycles. Although the light intensity is not continuous, we did not observe any flickering with the naked eye. Due to these unique properties under AC, different than in other well-known semiconductor devices, we propose a special electronic circuit symbol for BD LEDs, which is shown in the insets of Fig. 4a.

Additionally, we performed a time-resolved electroluminescence measurement to experimentally determine the device response time from positive to negative operation modes. The single BD LED device was biased alternately with positive + 100 A/cm² and negative − 60 A/cm² pulses, both 5 µs long. Firstly, the optical spectra were measured in the middle of both pulses and are shown in Fig. 4b. The measured spectra resemble the ones observed under DC conditions. For current density of -60 A/cm² a single peak at λ = 447 nm is observed, whereas for 100 A/cm² two peaks appear at λ = 428 nm and λ = 447 nm. The time evolution of light intensity together with measured current density are presented in Fig. 4c. In Figs. 4d-e we present in detail the region in which the direction of current flow is changing. As can be observed, the current pulses are not ideally rectangular in shape. Indeed, current spikes are present in both cases and reach ± 500 A/cm² in peak.

Firstly, we will analyze the intensity of electroluminescence coming from carrier recombination in the QW at λ = 447 nm. When the current switches from 100 A/cm² to -60 A/cm² (Fig. 4d), the intensity of QW peak raises immediately. Otherwise, when current switches back again to 100 A/cm² (Fig. 4e) the intensity of λ = 447 nm line decreases. However, the electroluminescence from QW never goes away and the BD LED switches very quickly. The response time from positive to negative pulse, and vice versa, are 60 ns and 30 ns, respectively, and they are comparable with time needed to stabilize the current equal to 50 ns. Therefore, it seems that the presence of p-type layers surrounding the active region from both sides do not disturb the carrier transport. We will analyze this subject theoretically in the Discussion section.

By examining the luminescence intensity of parasitic emission at wavelength of λ = 428 nm, it can be seen that when the current goes to negative values, its intensity drops (see Fig. 4d), while when the current switches to positive values, the light intensity peaks (see Fig. 4e). Although, after stabilizing the current the intensity of electroluminescence at λ = 428 nm is greater for − 60 A/cm² than for 100 A/cm², one can notice that during the negative pulse it does not form a distinct peak in the spectrum. This is just a tail of the QW peak, so in Fig. 4 we have left the dashed line for this range.

Influence of QW polarization field on carrier injection

Utilization of two TJs, each on every side of the active region, makes the BD LED structure symmetrical. However, there are some noticeable differences in the band alignment, evident even for unbiased structure in Fig. 2b. Indeed, the growth on (0001) GaN introduces an asymmetry in the arrangement of the built-in spontaneous and piezoelectric fields in the QW³⁹. In the conduction band the left-hand side barrier of the QW is higher than the right-hand side, which can lead to electron escape from the QW to the right-hand side barrier. When it comes to the valence band it is the opposite: the right-hand side barrier is higher. Therefore, taking into account the direction of the carrier injection, the band alignment in the active region is more favorable for reverse bias (see Fig. 2f) and leads to high injection efficiency. In the case of positively biased BD LED, the barrier for electron escape from the QW is small (Fig. 2e) and the resulting injection efficiency is low. In addition, electrons, which overshoot the QW under forward bias, can recombine with holes outside of it giving rise to parasitic recombination. In this work, we decided to keep the structure symmetric, when it comes to the doping and composition of subsequent layers. This choice allows to directly link the differences in band alignment, due to existence of internal electric fields, with the electro-optical properties of BD LEDs under both DC biasing conditions. It should be emphasized that so far it has not been possible to study this effect for the same QW in the active region.

In fact, the positive and negative bias conditions of the BD LED can be directly compared to III-N LEDs grown on Ga-polar and N-polar GaN substrates, respectively. In such LEDs, the differences in the injection efficiencies for structures with various polarities have been widely discussed^{19–22, 40-42}. In the case of Ga-polar structures, low injection efficiency and electron overflow above the QW is a commonly known issue. It occurs due to large difference in electron and hole mobility. This problem is obviated by introduction of an electron blocking layer (EBL) on the p-side right after the QW¹. Heavily doped AlGaN:Mg layer is most commonly used as the EBL, which due to its higher bandgap, forms an energetic barrier for electrons. When the EBL is not present or is not working properly (e. g. under cryogenic temperatures) electrons that pass over the QW recombine in the p-type region. This leads to the parasitic recombination in the spectral range of λ = 420–430 nm^22,42. In literature this peak is often identified as transition from the unknown deep donor down to Mg acceptor level^43,44. In case of N-polar LEDs there is a favorable arrangement of built-in electric fields that provides high injection efficiency^{20–22, 40,41}.

In view of these well-known properties of III-N LEDs, one can find that presented here electro-optical properties of BD LED under positive and negative biasing conditions correspond to Ga-polar and N-polar III-N LEDs, respectively. The injection efficiency is much higher for negatively biased BD LEDs, which leads to above 10 times higher optical power in this case. Moreover, for positively biased BD LED, electrons overflow above the QW and recombine with holes in the In_0.02Ga_0.98N:Mg layer (in between 235 nm and 270 nm in the band diagram shown in Fig. 2c) as evidenced by the registered parasitic peak at λ = 428 nm and λ = 419 nm for single and stack of two BD LEDs, respectively. The electron overflow problem, has been even more pronounced for pulse measurements presented in Fig. 4, where the intensity of parasitic luminesce follows current spikes, especially when BD LED was switched from negative to positive bias conditions. It is also possible that part of the electrons that overflow QW reaches the n-type layers through the top TJ, which could be the reason for lower operating voltages for positively biased BD LEDs. Considering the presented here BD LEDs as directly AC-driven light source, efforts should be made to increase the optical power obtained under positive bias. In principle, an AlGaN:Mg EBL placed on the right-hand side of the active region, should increase the injection efficiency and prevent electron overflow. However, under the negative bias, it could also provide a barrier for electrons that are injected into QW and thus disrupt this process. Therefore, in order to implement the EBL into the BD LED polar structure it is necessary to take a broader look at its band diagram.

It should be pointed out that the asymmetry in operation of presented here BD LEDs is solely due to the existence of built-in electric fields specific to III-N structures grown on polar substrates. Fabricating a BD LED on a non-polar GaN substrate or in a different material system, in which the issue of embedded piezoelectric fields does not exist, should give a symmetric light-current-voltage (LIV) characteristics, regardless of the current flow direction. Then, when powered directly by AC, one will get twice as much light from a single BD LED than from a single standard LED, which will be a step towards greater energy efficiency. It is worth to highlight the ability of stacking multiple BD LEDs during a single epitaxial process, in order to multiply the optical power obtained at the same current density. By growing a cascade of BD LEDs, high light output can be provided when operating at low peak current and high voltage, such as in an electrical grid²⁵.

Electron transport through p-type layers in BD LEDs

For standard LEDs, electrons are injected into the active region directly from the n-type doped region. However, in the case of BD LED, electrons firstly have to pass through a thin p-type region placed below and above the active region for positive and negative biases, respectively. Electrons that flow though the p-type region, on the path to the QW, may cause the ionization of neutral magnesium acceptors (\({R}_{A}^{-}\)) following the formula \({e}^{-}+{N}_{A}^{0}\to {N}_{A}^{-}\), which is shown in Fig. 5. Additionally, there are two other processes that are always present in p-type region and occur between the acceptor level (N_A) and the valence band (E_V) (see Fig. 5). Neutral magnesium acceptors are thermally ionized that promotes holes to the valence band (\({R}_{T}\)), which is \({N}_{A}^{0}\to {N}_{A}^{-}+{h}^{+}\). In the opposite process ionized acceptors are neutralized by holes (\({R}_{A}^{0}\)). \({R}_{T}\) and \({R}_{A}^{0}\) are in thermal equilibrium under constant bias, providing a locally constant concentration of holes over time. However, if the rate of \({R}_{A}^{-}\) process is much larger than \({R}_{T}\) and \({R}_{A}^{0}\), then all of the incoming electrons will firstly ionize neutral acceptors and only then will reach the QW. Hence, it may be a source of unwanted loss of injected electrons and cause a delay in switching the device from negative to positive bias and vice versa.

The rates of \({R}_{A}^{-}\), \({R}_{T}\), \({R}_{A}^{0}\) processes can be determined form the following formulas⁴³:

\({R}_{A}^{-}={C}_{nA}{N}_{A}^{0}n\) (1),

\({R}_{T}={N}_{A}^{0}{C}_{pA}{N}_{v}{g}^{-1}exp\left(-{E}_{A}/kT\right)\) (2),

\({R}_{A}^{0}={C}_{pA}{N}_{A}^{-}p\) (3),

where \({C}_{nA}\) is the electron capture coefficient (3.2x10^-12 cm³/s)⁴³, \({N}_{A}^{0}\) is the concentration of neutral magnesium acceptors, \(n\) is the concentration of electrons in the conduction band, \({C}_{pA}\) is the electron capture coefficient (5x10^-7 cm³/s)⁴³, \({N}_{v}\) is the effective density of states in the valence band, \(g\) is the degeneracy factor of acceptor level, \({E}_{A}\) is the ionization energy of magnesium acceptor, \(k\) is the Boltzmann constant, \(T\) is temperature, \({N}_{A}^{-}\) is the concentration of ionized magnesium acceptors and \(p\) is the concentration of holes in the valence band.

In general, considering all three processes, the kinetic equation describing the in-time changes of concentration of neutral acceptors can be written as:

\(\frac{d{N}_{A}^{0}}{dt}={R}_{A}^{0}-{R}_{T}-{R}_{A}^{-}\) (4).

However, while there is no electron in the p-type Eq. (4) reduces to:

\(\frac{d{N}_{A}^{0}}{dt}={R}_{A}^{0}-{R}_{T}\) (5).

We will now consider the situation, when a positive bias is suddenly applied to unbiased or negatively biased BD LED, following the experiment, to check if there is any significant delay of electroluminescence caused by \({R}_{A}^{-}\). In other words, the in-time stable Eq. (5) changes to Eq. (4). Now we need to check if the \({N}_{A}^{0}\) is still governed by thermodynamic processes \({R}_{T}\) and \({R}_{A}^{0}\). As an example, we will estimate their rates locally for a point in the middle of considered p-type region at 142.5 nm. We extract the carrier concentrations in the structure using One Dimensional Drift-diffusion Charge Control solver (1D DDCC) made by Y-R Wu^45–47 and use constants in Eqs. (1–3) from Reshchikov et al.⁴³.

Firstly, we calculate the rates of \({R}_{T}\) and \({R}_{A}^{0}\) for unbiased structure, in which no electron current is flowing through p-type region. Under these conditions \(p\), \({N}_{A}^{0}\) and \({N}_{A}^{-}\) were locally 1.0x10¹⁸ cm^-3, 6.5x10¹⁹ cm^-3 and 1.0x10¹⁸ cm^-3, respectively, which gives the rates of \({R}_{T}\) and \({R}_{A}^{0}\) equal to 2.8x10²⁹ cm^-3s^-1 and 5.3x10²⁹ cm^-3s^-1, respectively. The order of magnitude of both process is the same, whereas small inequality of \({R}_{A}^{-}\) and \({R}_{T}\) values might results from the different carrier ionization energies assumed in the simulations (0.178 eV) and in article of Reshchikov et al (0.15 eV)⁴³, which affects the calculated concentration of holes. Then, with applying the forward bias of 100 A/cm² we get \(n\) of 1.94x10¹⁴ cm^-3, which gives the rate of \({R}_{A}^{-}\) equal to 4.06x10²² cm^-3s^-1.

The above calculations show that the rate of \({R}_{A}^{-}\) is seven orders of magnitude smaller than \({R}_{A}^{0}\) and \({R}_{T}\). This means that the \({R}_{A}^{-}\) process is too weak to effectively ionize the neutral magnesium acceptors. \({N}_{A}^{0}\) does not change even if electrons are passing through p-type. Therefore, we do not observe any significant time delay, due to the electron-acceptor ionization process \({R}_{A}^{-}\), from switching the current direction of BD LED to the emission from QW. Otherwise, if the \({R}_{A}^{-}\) process were dominant, we should observe delays of hundreds of nanoseconds, which is the time needed to supply a matching number of electrons equal to the total number of neutral magnesium acceptors in the p-type region in between 120 nm and 165 nm.

Nevertheless, the relatively small value of \({R}_{A}^{-}\) process does not mean that it can be completely ignored, when it comes to the loss of electron current (\({j}_{loss}\)) passing through Mg-doped layers. Current losses can be estimated from the equation:

\({j}_{loss}=q{\int }_{{x}_{1}}^{{x}_{2}}{R}_{A}^{-}\left(x\right)dx\) (6),

where \(q\) is elementary charge, \({x}_{1}\) and \({x}_{2}\) are start and end point of considered p-type region (120 nm and 165 nm) and \({R}_{A}^{-}\left(x\right)\) is the rate of space depend electron-acceptor ionization process. From Eq. (6) we get \({j}_{loss}\) equal to 0.03 A/cm², which is, however, negligibly small comparing to the total current of 100 A/cm². From the above discussion, it is clear that the presence of p-type layers surrounding the active region has no significant impact on electron transport to the QW, which has been also confirmed by the time-resolved electroluminescence experiment. This makes BD LEDs feasible even in applications requiring fast switching times on the order of tens of nanoseconds.

This paper provides new insights into directly AC-driven semiconductor light sources. Here we demonstrated a proof-of-concept device – the GaN-based bidirectional light emitting diode (BD LED), which consists of single III-N wide QW surrounded by two oppositely oriented TJs. This allows to inject electrons and holes into the QW during both positive and negative biases. We have successfully shown that the BD LEDs are able to operate and emit light in both directions of supplied current, especially in both half-cycles of AC power supply. No delay in electroluminescence was observed after switching the current direction, which indicates that thin p-type layers surrounding the active region do not affect the transport of electrons. Moreover, we have presented that the BD LEDs can be vertically stacked in order to multiply the light emitted from single device and to withstand high-voltage application.

In addition, BD LED can be considered as a novel platform for investigation of the influence of the built-in electric fields on electro-optical properties of III-N devices. Indeed, we observed different optical powers under the positive and negative biases, even though recombination occurred from the same QW. This reflects different carrier injection efficiencies for different current directions. Furthermore, in order to make III-N BD LED grown on Ga-polar substrate more attractive as a directly AC-driven light source, we recommend to break the symmetry of the structure. We propose to increase the optical power for the positive bias by introduction of an electron blocking layer, known from standard GaN-based LEDs. We believe that with the unique construction of the BD LED, it will be possible to design cutting-edge experiments to better understand the physical processes occurring in III-N LEDs in general.

Structures were grown in metal-rich conditions by plasma-assisted molecular beam epitaxy (PAMBE) on the Ga-polar (0001) hydride vapor phase epitaxy bulk GaN crystals. The growth was performed in a VG Semicon V90 reactor. The GaN and InGaN layers were grown at 730°C and 650°C respectively. Details of PAMBE growth can be found elsewhere⁴⁸.

Samples were processed into devices with dimensions of 100x100 µm² by standard photolithography. For each sample a mesa was etched through the whole structure by reactive ion etching. We used n-type metallization Ti/Al/Ni/Au (30/60/40/75 nm) sequence for top and bottom contacts. Upper contact was placed on the top of mesa in 90x90 µm² squares and it was annealed at 750°C for 1 minute in N₂ atmosphere. The bottom metallization was applied onto entire surface of the sample and it remained not annealed. For both samples several dozens of devices were measured and representative characteristics were chosen for further analysis.

The electro-optical properties of BD LEDs under DC conditions were examined in a home-made setup, in which probe station is coupled with an Si photodiode power meter (Thorlabs S130VC), a spectrometer (Ocean Optics, USB4000) and source meter unit (Keithley, SMU2400). We would like to emphasize here that the measured optical power should be treated qualitatively, because we did not use an integration sphere.

The AC time-resolved electroluminescence measurement was performed in another home-made experimental setup, which combines the arbitrary waveform generator (Siglent, SDG2042X), oscilloscope (Tektronix, TDS3032B) and spectrometer (SPEX Industries, 1000M) coupled with photomultiplier tube (Hamamatsu, R636-10) and photon counter (Stanford Research Systems, SR400). The BD LED was powered by arbitrary waveform generator (Siglent SDG2042X), which was the source of sinusoidal and rectangular voltage pulses. Current, which was flowing though device was in time controlled with the oscilloscope by checking the voltage drop on 50 Ohm resistor connected in series with the BD LED. Spectrophotometer operated in two modes: spectra measurement in the specific moment in time with 1 nm resolution or real time light intensity measurement for specific wavelength in 10 ns time windows. Obtained data were smoothed with weighted adjacent-averaging method and the delay time was corrected based on the photomultiplier tube response time and the optical path of the light beam.

Acknowledgements

This work was supported partially by TEAM-TECH POIR.04.04.00-00-210C/16-00 and HOMING POIR.04.04.00-00-5D5B/18-00 projects of the Foundation for Polish Science co-financed by the European Union under the European Regional Development Fund and the Polish National Centre for Research and Development Grants LIDER/29/0185/L-7/15/NCBR/2016 and LIDER/35/0127/L-9/17/NCBR/2018 and National Science Center Poland within grant No. 2019/35/D/ST3/03008.

Conflict of interest

A patent describing the BD LEDs in this work has been filed to the Polish Patent Office (application number P.438744) and to the European Patent Office (application number EP22461527.8). M. Żak, G. Muzioł, M. Siekacz, K. Nowakowski-Szkudlarek, H. Turski and C. Skierbiszewski are the co-inventors. The remaining authors declare no competing interests.

Nakamura, S., Senoh, M., Iwasa, N. & Nagahama, S.-i. High-Brightness InGaN Blue, Green and Yellow Light-Emitting Diodes with Quantum Well Structures. Japanese Journal of Applied Physics 34, L797, doi:10.1143/JJAP.34.L797 (1995).
Krames, M. R. et al. Status and Future of High-Power Light-Emitting Diodes for Solid-State Lighting. Journal of Display Technology 3, 160–175, doi:10.1109/JDT.2007.895339 (2007).
Pimputkar, S., Speck, J. S., DenBaars, S. P. & Nakamura, S. Prospects for LED lighting. Nature Photonics 3, 180–182, doi:10.1038/nphoton.2009.32 (2009).
De Almeida, A., Santos, B., Paolo, B. & Quicheron, M. Solid state lighting review – Potential and challenges in Europe. Renewable and Sustainable Energy Reviews 34, 30–48, doi:10.1016/j.rser.2014.02.029 (2014).
Narukawa, Y., Ichikawa, M., Sanga, D., Sano, M. & Mukai, T. White light emitting diodes with super-high luminous efficacy. Journal of Physics D: Applied Physics 43, 354002, doi:10.1088/0022-3727/43/35/354002 (2010).
David, A. et al. High light extraction efficiency in bulk-GaN based volumetric violet light-emitting diodes. Applied Physics Letters 105, 231111, doi:10.1063/1.4903297 (2014).
Hurni, C. A. et al. Bulk GaN flip-chip violet light-emitting diodes with optimized efficiency for high-power operation. Applied Physics Letters 106, 031101, doi:10.1063/1.4905873 (2015).
Yonkee, B. P., Young, E. C., DenBaars, S. P., Nakamura, S. & Speck, J. S. Silver free III-nitride flip chip light-emitting-diode with wall plug efficiency over 70% utilizing a GaN tunnel junction. Applied Physics Letters 109, 191104, doi:10.1063/1.4967501 (2016).
Arias, M., Vázquez, A. & Sebastián, J. An Overview of the AC-DC and DC-DC Converters for LED Lighting Applications. Automatika 53, 156–172, doi:10.7305/automatika.53-2.154 (2012).
Kim, H., Ryu, M., Baek, J. & Jung, J. High-Efficiency Isolated Bidirectional AC–DC Converter for a DC Distribution System. IEEE Transactions on Power Electronics 28, 1642–1654, doi:10.1109/TPEL.2012.2213347 (2013).
Cho, J. et al. Alternating-current Light Emitting Diodes with a Diode Bridge Circuitry. Japanese Journal of Applied Physics 46, L1194-L1196, doi:10.1143/jjap.46.l1194 (2007).
Yen, H.-H., Kuo, H.-C. & Yeh, W.-Y. Characteristics of Single-Chip GaN-Based Alternating Current Light-Emitting Diode. Japanese Journal of Applied Physics 47, 8808–8810, doi:10.1143/jjap.47.8808 (2008).
Wang, L., Xiao, L., Gu, H. & Sun, H. Advances in Alternating Current Electroluminescent Devices. Advanced Optical Materials 7, 1801154, doi:10.1002/adom.201801154 (2019).
Hartensveld, M., Melanson, B., Vangipuram, V. T. & Zhang, J. 450 nm Gallium Nitride Alternating Current Light-Emitting Diode. IEEE Photonics Journal 12, 1–6, doi:10.1109/JPHOT.2020.3032382 (2020).
Sadaf, S. M., Ra, Y. H., Nguyen, H. P. T., Djavid, M. & Mi, Z. Alternating-Current InGaN/GaN Tunnel Junction Nanowire White-Light Emitting Diodes. Nano Letters 15, 6696–6701, doi:10.1021/acs.nanolett.5b02515 (2015).
Akyol, F., Zhang, Y., Krishnamoorthy, S. & Rajan, S. Ultralow-voltage-drop GaN/InGaN/GaN tunnel junctions with 12% indium content. Applied Physics Express 10, 121003, doi:10.7567/apex.10.121003 (2017).
Żak, M. et al. Tunnel Junctions with a Doped (In,Ga)N Quantum Well for Vertical Integration of III-Nitride Optoelectronic Devices. Physical Review Applied 15, 024046, doi:10.1103/PhysRevApplied.15.024046 (2021).
Young, E. C. et al. Hybrid tunnel junction contacts to III–nitride light-emitting diodes. Applied Physics Express 9, 022102, doi:10.7567/APEX.9.022102 (2016).
Turski, H., Bharadwaj, S., Xing, H. & Jena, D. Polarization control in nitride quantum well light emitters enabled by bottom tunnel-junctions. Journal of Applied Physics 125, 203104, doi:10.1063/1.5088041 (2019).
Bharadwaj, S. et al. Enhanced injection efficiency and light output in bottom tunnel-junction light-emitting diodes. Opt. Express 28, 4489–4500, doi:10.1364/OE.384021 (2020).
Turski, H. et al. Nitride LEDs and Lasers with Buried Tunnel Junctions. ECS Journal of Solid State Science and Technology 9, 015018, doi:10.1149/2.0412001JSS (2020).
Chlipala, M. et al. Nitride light-emitting diodes for cryogenic temperatures. Opt. Express 28, 30299–30308, doi:10.1364/OE.403906 (2020).
Siekacz, M. et al. Stack of two III-nitride laser diodes interconnected by a tunnel junction. Opt. Express 27, 5784–5791, doi:10.1364/OE.27.005784 (2019).
Siekacz, M. et al. Vertical Integration of Nitride Laser Diodes and Light Emitting Diodes by Tunnel Junctions. Electronics 9 (2020).
Akyol, F., Krishnamoorthy, S. & Rajan, S. Tunneling-based carrier regeneration in cascaded GaN light emitting diodes to overcome efficiency droop. Applied Physics Letters 103, 081107, doi:10.1063/1.4819737 (2013).
Akyol, F., Krishnamoorthy, S., Zhang, Y. & Rajan, S. GaN-based three-junction cascaded light-emitting diode with low-resistance InGaN tunnel junctions. Applied Physics Express 8, 082103, doi:10.7567/APEX.8.082103 (2015).
Li, P. et al. Fully transparent metal organic chemical vapor deposition-grown cascaded InGaN micro-light-emitting diodes with independent junction control. Opt. Express 29, 22001–22007, doi:10.1364/OE.430694 (2021).
Malinverni, M., Martin, D. & Grandjean, N. InGaN based micro light emitting diodes featuring a buried GaN tunnel junction. Applied Physics Letters 107, 051107, doi:10.1063/1.4928037 (2015).
Gibasiewicz, K. et al. InGaN blue light emitting micro-diodes with current path defined by tunnel junction. Opt. Lett. 45, 4332–4335, doi:10.1364/OL.394629 (2020).
Slawinska, J. et al. Ion implantation of tunnel junction as a method for defining the aperture of III-nitride-based micro-light-emitting diodes. Opt. Express 30, 27004–27014, doi:10.1364/OE.458950 (2022).
Li, P. et al. Size-independent low voltage of InGaN micro-light-emitting diodes with epitaxial tunnel junctions using selective area growth by metalorganic chemical vapor deposition. Opt. Express 28, 18707–18712, doi:10.1364/OE.394664 (2020).
Muziol, G. et al. Distributed-feedback blue laser diode utilizing a tunnel junction grown by plasma-assisted molecular beam epitaxy. Opt. Express 28, 35321–35329, doi:10.1364/OE.405994 (2020).
Chaney, A. et al. Gallium nitride tunneling field-effect transistors exploiting polarization fields. Applied Physics Letters 116, 073502, doi:10.1063/1.5132329 (2020).
Muziol, G. et al. Beyond Quantum Efficiency Limitations Originating from the Piezoelectric Polarization in Light-Emitting Devices. ACS Photonics 6, 1963–1971, doi:10.1021/acsphotonics.9b00327 (2019).
Gardner, N. F. et al. Blue-emitting InGaN–GaN double-heterostructure light-emitting diodes reaching maximum quantum efficiency above 200A∕cm2. Applied Physics Letters 91, 243506, doi:10.1063/1.2807272 (2007).
Nakamura, S., Iwasa, N., Masayuki Senoh, M. S. & Takashi Mukai, T. M. Hole Compensation Mechanism of P-Type GaN Films. Japanese Journal of Applied Physics 31, 1258, doi:10.1143/JJAP.31.1258 (1992).
Czernecki, R. et al. Hydrogen diffusion in GaN:Mg and GaN:Si. Journal of Alloys and Compounds 747, 354–358, doi:10.1016/j.jallcom.2018.02.270 (2018).
Neugebauer, S. et al. All metalorganic chemical vapor phase epitaxy of p/n-GaN tunnel junction for blue light emitting diode applications. Applied Physics Letters 110, 102104, doi:10.1063/1.4978268 (2017).
Bernardini, F., Fiorentini, V. & Vanderbilt, D. Spontaneous polarization and piezoelectric constants of III-V nitrides. Physical Review B 56, R10024-R10027, doi:10.1103/PhysRevB.56.R10024 (1997).
Masui, H. et al. Luminescence Characteristics of N-Polar GaN and InGaN Films Grown by Metal Organic Chemical Vapor Deposition. Japanese Journal of Applied Physics 48, 071003, doi:10.1143/JJAP.48.071003 (2009).
Akyol, F., Nath, D. N., Gür, E., Park, P. S. & Rajan, S. N-Polar III–Nitride Green (540 nm) Light Emitting Diode. Japanese Journal of Applied Physics 50, 052101, doi:10.1143/JJAP.50.052101 (2011).
Grzanka, S. et al. Role of the electron blocking layer in the low-temperature collapse of electroluminescence in nitride light-emitting diodes. Applied Physics Letters 90, 103507, doi:10.1063/1.2711765 (2007).
Reshchikov, M. A., Ghimire, P. & Demchenko, D. O. Magnesium acceptor in gallium nitride. I. Photoluminescence from Mg-doped GaN. Physical Review B 97, 205204, doi:10.1103/PhysRevB.97.205204 (2018).
Eckey, L. et al. Photoluminescence and Raman study of compensation effects in Mg-doped GaN epilayers. Journal of Applied Physics 84, 5828–5830, doi:10.1063/1.368853 (1998).
Wu, Y.-R., Singh, M. & Singh, J. Gate leakage suppression and contact engineering in nitride heterostructures. Journal of Applied Physics 94, 5826–5831, doi:10.1063/1.1618926 (2003).
Wu, Y. R., Chiu, C., Chang, C. Y., Yu, P. & Kuo, H. C. Size-Dependent Strain Relaxation and Optical Characteristics of InGaN/GaN Nanorod LEDs. IEEE Journal of Selected Topics in Quantum Electronics 15, 1226–1233, doi:10.1109/JSTQE.2009.2015583 (2009).
Wu, Y.-R. & Singh, J. Metal piezoelectric semiconductor field effect transistors for piezoelectric strain sensors. Applied Physics Letters 85, 1223–1225, doi:10.1063/1.1784039 (2004).
Skierbiszewski, C. et al. Nitride-based laser diodes grown by plasma-assisted molecular beam epitaxy. Journal of Physics D: Applied Physics 47, 073001, doi:10.1088/0022-3727/47/7/073001 (2014).

Yes there is potential Competing Interest. A patent describing the BD LEDs in this work has been filed to the Polish Patent Office (application number P.438744) and to the European Patent Office (application number EP22461527.8). M. Żak, G. Muzioł, M. Siekacz, K. Nowakowski-Szkudlarek, H. Turski and C. Skierbiszewski are the co-inventors. The remaining authors declare no competing interests.

Download PDF

Journal Publication

published 20 Nov, 2023

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

Bidirectional LED as an AC-driven visible-light source

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Results

BD LEDs structure

BD LEDs operating principle

BD LED properties under DC

BD LEDs properties under AC

Discussion

Influence of QW polarization field on carrier injection

Electron transport through p-type layers in BD LEDs

Conclusions

Materials And Methods

Declarations

References

Additional Declarations

Status:

Journal Publication

Version 1