Introduction

The ubiquity of light-emitting diodes (LEDs) in our everyday life is undeniable, and micro-LEDs are now being rapidly developed to usher in a new wave of display technologies. One obstacle to realizing these technologies is the lack of an efficient red LED based on GaN. Currently, most full-color LED displays are made by combining green and blue GaN-based LEDs with traditional red GaAs-based LEDs using the pick-and-place technique1. A long-standing issue with fabricating GaN-based red LEDs is the large Indium content required in the InGaN layers, which introduces strain and reduces efficiency due to the quantum-confined Stark effect and indium phase separation2,3,4,5,6. For micro-LEDs in general, InGaN faces two additional challenges associated with the reduction in device size: a sharp decrease in external quantum efficiency (EQE) due to leakage current at the device walls and significant blue shifts in peak emission wavelength due to band-filling effects7. The leakage current increases for smaller devices8,9, due to the non-radiative recombination at the sidewalls of the micro-LED structures. Despite these challenges, blue and green GaN-based micro-LEDs with EQEs of 13% and 25%, respectively, for an active area of 1 µm2 were reported10,11. However, the EQE of red micro-LED remains below 5% for active regions smaller than 100 µm211,12,13,14. Once these challenges are overcome, micro-LEDs will facilitate the development of innovative technologies such as augmented reality and transparent displays15.

Eu-doped GaN-based LEDs represent an alternative path towards full-color monolithic displays and offer favorable properties over InGaN LEDs, especially for micro-LED applications. Figure 1 shows monolithically grown blue (InGaN) and red (Eu-doped GaN) micro-LEDs with a 20 µm width. For Eu-doped GaN LEDs, the red emission at ~620 nm originates from intra-4f transitions within the Eu3+ ions16. As with all rare earth (RE) ions, Eu3+ ions exhibit sharp and spectrally stable emission regardless of the host system or excitation conditions. Additionally, Eu-doped materials have been shown to be less influenced by non-radiative sidewall defects due to their short carrier diffusion lengths17,18. They have also been used to realize monolithically stacked full-color LEDs on a single chip19. These properties make Eu-doped GaN a promising candidate for micro-LED applications.

Fig. 1: Monolithically grown two-color micro-LED.
figure 1

a Schematic for monolithically grown blue (InGaN) and red (Eu-doped GaN) micro-LEDs. b Prototype of a monolithically grown two-color micro-LED array with 20 µm wide pixels.

Full size image

The output power of Eu-doped GaN devices now exceeds 1.2 mW with EQEs as high as 9.2%. This was achieved using a superlattice structure consisting of alternating GaN and Eu-doped GaN layers, resulting in a 25-fold increase16,19. While this performance far exceeds that of devices grown using single active layers of Eu-doped GaN, the exact origin of the increased output power and EQE remains unknown16. Using atomic force and transmission electron microscopy, it was shown that the size and density of threading dislocations were significantly reduced in superlattice samples, which would reduce leakage current and could partially explain the improved electrical properties of the superlattice structure20. In addition, these results also showed that the lattice expanded within the Eu-doped layers relative to the undoped layers. A similar enhancement in luminescence and device performance was reported for superlattice devices consisting of alternating Si/Er-Si layers, which also outperformed monolayer-based devices by over an order of magnitude, at 80 K21,22,23,24,25. In this case, the enhancement has simply been explained as an increase in the efficiency of the energy transfer between the host and the RE ions or the selective formation of highly efficient defect centers26. However, there may be a deeper underlying mechanism for this enhancement.

The origin of these enhancements may be related to reports in other RE-doped semiconductor systems used for different applications24,27,28,29,30,31,32,33,34. For example, the doping of RE ions, such as Eu, Nd, Er, Tb, and Sm into semiconductor nanoparticles such as TiO2 and ZnO has been used for nearly two decades to modify the bandgap of the nanoparticles themselves, which makes the host materials more suitable for certain applications, such as photovoltaics27,28,30,31. Dopant concentrations typically exceed 2%; however, this behavior has also been observed for dilute dopant levels of ~0.5%. Since the optical band gap and index of refraction are related quantities35, it is not surprising that several other groups have also reported an increase in the index of refraction due to dilute doping25,36, which could be used to fabricate integrated waveguides for telecommunication applications36,37. However, there are no reports, to the best of our knowledge, on the simultaneous measurement of bandgap and index of refraction changes due to RE doping, or the use of superlattice structures to enhance optoelectronic properties of RE-based devices due to the carrier confinement and waveguiding that should result. To this end, probing the carrier behavior in the Eu-doped GaN superlattice structure is necessary to understand the device performance and further optimize these devices.

Here, we employed terahertz (THz) emission spectroscopy (TES) to study the dynamics of photoinduced carriers in the Eu-doped superlattice structures as compared to single-layer Eu-doped GaN samples. When photocarriers are excited in semiconductors by femtosecond (fs) optical pulses, they are accelerated by a built-in electric field. This acceleration generates THz radiation that reflects the carrier movement within the first few picoseconds after excitation. The waveforms and amplitudes can provide physical information on a wide range of device materials and structures, such as the semiquantitative estimation of semiconductor surface/interface potentials, in a non-contact and non-destructive manner38,39,40,41. This technique is particularly useful for wide bandgap semiconductor evaluation as it allows for the direct measurement of the material’s bandgap41,42. In many cases, measuring the bandgap energy using room-temperature photoluminescence (PL) is difficult in semiconductors with a high concentration of impurities or dopants. However, TES solely results from excitation to the conduction band and is not influenced by impurity levels and provides a new platform to discuss the ultrafast photocarrier dynamics of wide bandgap semiconductors43,44,45,46,47. Thus, TES is an ideal tool to study the mechanism behind the enhanced properties of dilutely doped RE-based superlattice optoelectronic devices, and Eu-doped GaN serves as an example system for the utility of these properties.

Results

Bandgap reduction due to Eu doping

The structures of all samples used in this study are shown in Fig. 2. Measurements were taken on undoped GaN (Fig. 2a), a thick Eu-doped GaN sample (Fig. 2b), GaN doped with low and high Eu concentrations (Fig. 2c), and a superlattice structure consisting of 40 pairs of alternating Eu-doped GaN (3 nm) and GaN (10 nm) layers (Fig. 2d). These samples are labeled ud-GaN, GaN:EuTDS, GaN:Eulow and GaN:Euhigh, and superlattice, respectively.

Fig. 2: Structures for the samples studied in this work.
figure 2

a ud-GaN, (b) GaN:EuTDS, (c) GaN:Eulow,high, (d) superlattice. All structures were grown on the low-temperature-grown (LT)-GaN, which were grown on c-plain sapphire substrates.

Full size image

The bandgap of a semiconductor material can be determined with meV precision by evaluating the excitation wavelength dependence of the emitted THz signal within 10 ps of excitation43,48. This behavior was used to determine the influence of dilute Eu doping on the bandgap of GaN. Figures 3a–c shows the time evolution of the THz emission obtained from the ud-GaN, GaN:Eulow, and GaN:Euhigh samples, where the THz signals are shown for different excitation wavelengths, at room temperature. Since the bandgap of ud-GaN is ~361 nm at room temperature49,50, the THz signals should be strong for excitation wavelengths of <361 nm, for all samples. On the other hand, the THz signal was clearly visible even at 365 nm excitation for GaN:Euhigh. We have previously shown that polarity of the THz emission was dependent on the GaN conducting type. This indicates that the THz emission from the GaN surface is caused by the photocurrent in the surface electric field41,47.

Fig. 3: THz emission and time-domain spectroscopy for bulk GaN samples.
figure 3

THz waveforms emitted from (a) ud-GaN, (b) GaN:Eulow, and (c) GaN:Euhigh with various excitation wavelengths at room temperature. Each color of the waveforms represents excitation at 359–368 nm. d Peak THz amplitudes at t = 10 ps of the waveforms shown in the (a–c) plotted as a function of excitation wavelengths. Black, blue, and red solid circles correspond to ud-GaN, GaN:Eulow, and GaN:Euhigh, respectively. Data for ud-GaN and GaN:Euhigh are enlarged by a factor of 5. Solid lines are fits to the THz amplitudes using Eq. (2). e Refractive index n (black solid line) and extinction coefficient k (red dashed line) in Eu-doped GaN for a range of THz frequencies as measured by THz time-domain spectroscopy.

Full size image

Figure 3d shows the peak amplitudes of the THz emission at ~10 ps as a function of excitation wavelength for all three samples. The black, blue, and red data points correspond to ud-GaN, GaN:Eulow and GaN:Euhigh, respectively. (See Supplementary Note 1 and Supplementary Fig. 1 for details on the superlattice structure measurements). Among these samples, GaN:Eulow exhibited the highest emission amplitude. This is attributed to changes in surface potential and carrier mobility due to the Eu doping. Here, the THz emission amplitude can be described by

$$|{E}_{{{{{{\rm{THZ}}}}}}}|\propto \mu {E}_{{{{{{\rm{Max}}}}}}}{I}_{{{{{{\rm{p}}}}}}}\propto \mu \frac{{V}_{{{{{{\rm{D}}}}}}}}{{{\lambda }}_{{{{{{\rm{L}}}}}}}}{I}_{{{{{{\rm{p}}}}}}},$$
(1)

where µ is the carrier mobility, \({I}_{{{{{{\rm{p}}}}}}}\) is the injection photon flux, \({V}_{{{{{{\rm{D}}}}}}}\) is the diffusion voltage, and \({\lambda }_{{{{{{\rm{L}}}}}}}\) is the penetration depth of excitation pulses41,48. Eu-doped GaN has shown a larger surface potential and smaller carrier mobility compared to ud-GaN17,51. These changes cancel each other in Eq. (1). Consequently, GaN:Euhigh showed a similar amplitude to ud-GaN. However, in GaN:Eulow, carrier mobility might be higher than that in GaN:Euhigh due to a lower concentration of Eu doping. As a result, the emission from GaN:Eulow was larger than the others. To gain further insight, we need to obtain the doping concentration dependence of the penetration depth, surface potential, and carrier mobility accurately. However, these discussions are beyond the scope of the present study and are left for future work.

The THz emission by the optical rectification has shown to be not dominant in the emission from c-plane GaN bulk samples41,47. We also explored the influence of optical rectification on the emission from the superlattice sample as shown in Supplementary Fig. 2, which suggested the less/no nonlinear effect. Therefore, we can rule out the nonlinear effect as the THz emission mechanism in this work (See Supplementary Note 2 and Supplementary Fig. 2 for information about the THz emission property). This means that the THz emission is only generated when photocarriers are excited by photons with an energy larger than the bandgap of the material. Once the excitation energy falls below the bandgap, the THz emission amplitude will go to zero, and we can use this drop-off in amplitude to determine the bandgap energy. However, to extract the bandgap from the curves shown in Fig. 3d, we must take into account the spectral width of the fs excitation pulses, which have a broad wavelength profile due to the relationship between energy and time. To do this, we assume that the spectral profile of the fs pulses follows a hyperbolic secant distribution. Under this assumption, the THz emission amplitude \({E}_{{THz}}\) can be described by,

$$|{E}_{{{{{{\rm{THz}}}}}}}|{\propto } \int_{0}^{({\lambda }_{{{{{{\rm{g}}}}}}})}{\sec}{{{{{\rm{h}}}}}}^{2}(A(\lambda -{\lambda }_{{{{{{\rm{p}}}}}}}))d\lambda$$
(2)

where \(A\) is a fitting parameter, \({\lambda }_{{{{{{\rm{g}}}}}}}\) is a wavelength corresponding to the bandgap energy, \({\lambda }_{{{{{{\rm{p}}}}}}}\) is a center wavelength of fs excitation pulses41. The solid curves in Fig. 3d are fits of the data using Eq. (2). From these fits, we obtained bandgap energies (and corresponding wavelengths) of 3.430 \(\pm\)0.003 eV (361.5 \(\pm\)0.3 nm), 3.419 \(\pm\)0.002 eV (362.7 \(\pm\)0.2 nm), and 3.393 \(\pm\)0.002 eV (365.5 \(\pm\)0.2 nm) for ud-GaN, GaN:Eulow, and GaN:Euhigh, respectively (we showed the stability and validity of this measurement in Supplementary Note 3 and Supplementary Figs. 3–5). The determined bandgap energy of ud-GaN was consistent with previously reported values at room temperature41,49,50, while the bandgap energy of the Eu-doped samples was found to decrease with increasing Eu concentration. The total bandgap difference between ud-GaN and GaN:Euhigh was found to be ~37 meV.

In this calculation, the absorption difference under different excitation wavelengths was not considered because the difference is negligible within the energy range above the bandgap52. Although the absorption difference is larger in the wavelength range below the bandgap, the THz emission should be much smaller compared to the above bandgap excitation. In addition, we disregarded the influences of the interfaces between ud-GaN/low-temperature-grown(LT)-GaN or GaN:Eu/ud-GaN in this work. Since the fermi level in the n-type GaN materials is nearly independent on the carrier concentration53,54, we expect the fermi level in the GaN:Eu to be almost the same as the ud-GaN (Carrier density in the GaN:Eu will be obtained in the later section). Consequently, it seems that no significant band bending was formed between these layers. In addition, the depletion layer thickness w can be described as

$$w=\sqrt{\frac{2{\varepsilon }_{{{{{{\rm{r}}}}}}}{\varepsilon }_{0}{V}_{{{{{{\rm{D}}}}}}}}{{{{{{{\rm{e}}}}}}}^{2}{N}_{{{{{{\rm{D}}}}}}}}},$$
(3)

where \({\varepsilon }_{{{{{{\rm{r}}}}}}}\) is the dielectric constant of GaN (\({\varepsilon }_{{{{{{\rm{r}}}}}}}\) = 9.555), \({\varepsilon }_{0}\) is the vacuum permittivity, e is the electron charge, and \({N}_{{{{{{\rm{D}}}}}}}\) is the donor concentration. Since obtaining the activated donor density is often challenging, we approximated it using the measured carrier density N. w at GaN:Eu surface was estimated to be ~0.53 µm using \({V}_{{{{{{\rm{D}}}}}}}\) = 1.61 eV41,51, N = \(6\times {10}^{17}\) cm-3 (mentioned in the next section), and \({\varepsilon }_{{{{{{\rm{r}}}}}}}\) = 9.555. Since this value is smaller than the thickness of the GaN:Eu layer, the GaN:Eu/ud-GaN interface should not influence the THz emission from the surface depletion layer. Additionally, the penetration depth for an excitation wavelength of 360 nm is ~0.8-1 µm in GaN41. This suggests that the excitation pulses could not reach the ud-GaN/LT-GaN interface. Therefore, it seems that the interfaces in GaN:Eu samples did not influence the optical properties and THz emissions in this work.

Index of refraction and extinction coefficient in Eu-doped GaN

Terahertz time-domain spectroscopy (THz-TDS) was used to obtain the refractive index (n), extinction coefficient (k) and free carrier density (N) in GaN:EuTDS samples by measuring the THz pulses that were transmitted through the sample and sapphire substrate. The complex spectra of the transmitted pulses, \({E}_{{{{{{\rm{o}}}}}}}\left(\omega \right)\), were compared to reference spectra, \({E}_{{{{{{\rm{r}}}}}}}\left(\omega \right)\) using the relationship,

$$\frac{{E}_{0}\left(\omega \right)}{{E}_{{{{{{\rm{r}}}}}}}(\omega )} \sim 1+\frac{{{{{{\rm{j}}}}}}\omega d}{c\left({n}_{{{{{{\rm{air}}}}}}}+{n}_{{{{{{\rm{sub}}}}}}}\right)}\left[-{n}_{{{{{{\rm{air}}}}}}}+{n}_{{{{{{\rm{air}}}}}}}{n}_{{{{{{\rm{sub}}}}}}}+{\left(n+{{{{{\rm{j}}}}}}k\right)}^{2}-{n}_{{{{{{\rm{sub}}}}}}}\right],$$
(4)

where \(d\) is the GaN:Eu layer thickness, c is the speed of light, \({n}_{{{{{{\rm{air}}}}}}}\) and \({n}_{{{{{{\rm{sub}}}}}}}\) are the refractive indexes of the air and sapphire substrate56. From this, n and k were obtained at different THz frequencies as shown in Fig. 3e. As example, we find that the index of refraction at 1 THz is ~3.5, while that previously reported for n-type GaN at 1 THz was ~357. This represents a 20% increase in the refractive index due to dilute Eu doping. In addition, the free carrier concentration can be extrapolated from this data and was found to be N = \(6\times {10}^{17}\) cm-3. See Supplementary Note 4 and Supplementary Fig. 6 for more details.

Potential barrier energy in the superlattice structure

The observed bandgap reduction due to Eu doping results in bandgap differences between the undoped and Eu-doped layers within the superlattice structure. This bandgap difference plays an important role in device performance because it can induce potential barriers resulting in a carrier confinement effect similar to that observed in multiple quantum wells (MQWs)58,59,60,61. Therefore, it must be determined how much of this reduction is due to a change in the conduction band as opposed to the valence band relative to the GaN barrier layers, as this will impact the carrier dynamics differently. Since THz spectroscopy directly probes electrons within the conduction band, we can use the temperature dependence of the THz emission to determine the influence of the potential barriers within the conduction band on the carrier dynamics within the superlattice structures. We performed temperature-dependent THz spectroscopy on the GaN:Euhigh and the superlattice samples, as shown in Fig. 4. Here, the peak amplitudes are plotted against temperature (See Supplementary Note 5 and Supplementary Fig. 7 for more details on peak amplitudes determination). Note: The shortest excitation wavelength available was 355 nm, which limited our ability to probe the samples below 200 K50.

Fig. 4: THz emission peak amplitudes from GaN:Euhigh (blue circle) and superlattice (red circle) measured at different temperatures.
figure 4

Blue and red solid lines are fits for GaN:Euhigh and superlattice using Eqs. (5) and (7), respectively.

Full size image

The peak amplitude of the THz emission from the GaN:Euhigh sample decreased continuously with increasing temperature within the probe range. On the other hand, the emission from the superlattice first increased between 200 K and 250 K and then decreased continuously at higher temperatures. As described by Eq. (1), the THz emission electric field is proportional to the electron mobility. Consequently, our results suggested that the electron mobility in the GaN:Euhigh sample increased continuously with decreasing temperature. The mobility behavior in superlattice structure, however, resembles that observed in GaN/InGaN MQWs and is related to an induced potential barrier for the electrons between the GaN and InGaN layers62.

To understand this behavior quantitatively, we note that the amplitude of the THz emission in Fig. 4 is proportional to the electron mobility, as described in Eq. (1). We employed a model for the temperature dependence of the electron mobility in bulk GaN to evaluate the mobility in GaN:Euhigh63. In this model, we replace the carrier mobility in GaN with that in Eu-doped GaN, which has been previously measured at room temperature to be \({\mu }_{{{{{{\rm{Eu}}}}}}}\left(300\right)=53{{{{{{\rm{cm}}}}}}}^{2}\,{{{{{{\rm{V}}}}}}}^{-1}\,{{{{{{\rm{s}}}}}}}^{-1}\) 17. Fitting the data for GaN:Euhigh in Fig. 4 to this model yields an expression for the temperature dependence of the carrier mobility in Eu-doped GaN, \({\mu }_{{{{{{\rm{Eu}}}}}}}(T)\). To evaluate the impact of the superlattice structure on the carrier mobility between the doped and undoped layers and determine the barrier potential for the electrons within this structure, we use \({\mu }_{{{{{{\rm{Eu}}}}}}}(T)\) in a mobility model that takes into account the barrier height between layers62,64. A detailed description of the carrier mobility models is given in the Methods section. After fitting the superlattice structure data in Fig. 4, we obtained a potential barrier value of \(\sim 11\pm 4\) meV for electrons. Given that the growth condition for Eu-doped layer in the superlattice is the same as the condition for GaN:Euhigh, the bandgap difference between GaN and GaN:Eu layers seems to be ~37 meV, consistent with the difference between ud-GaN and GaN:Euhigh. Consequently, we can obtain the energy discontinuity on the valence band as ~26 \(\pm 4\) meV. This barrier height induces confinement for ~63% of holes at room temperature as determined from the thermal emission, \(\exp (-q{V}_{{{{{{\rm{b}}}}}}}/{k}_{{{{{{\rm{B}}}}}}}T)\), where \({k}_{{{{{{\rm{B}}}}}}}\) is the Boltzmann constant, and \(q{V}_{{{{{{\rm{b}}}}}}}\) represents the barrier height. In addition, this barrier height is low enough as to not impede the hole transport, which can cause poor LED performance as mentioned by Xie et al.65 Therefore, we suggest that the enhanced Eu luminescence efficiency within the superlattice structure is the result of a small but effective carrier confinement for both electrons and holes, which does not significantly impede their mobility allowing for favorable device performance under current injection.

Discussion

The concept of using RE ions to modify the bandgap of a semiconductor host is well-established for certain applications. An example would be the use of RE doping to tune the bandgap of TiO2, SnO2, and ZnO nanocrystals to increase photocatalytic activity in the visible range or for wavelength matching in dye-sensitized solar cells32,33. Several origins of the bandgap modification have been proposed. A prominent explanation is that the tensile strain that results from the larger ionic radii of the RE ion relative to the atoms they replace leads to a change in the intra-atomic s − d exchange interaction, which is believed to reduce the conduction band edge energy29,30,31. Previous calculations suggest that the introduction of RE ions introduces 4f electron states just below the conduction band, which forms a new lowest unoccupied level and modifies the bandgap energy, such as in Nd-doped TiO228. However, this argument seems to be unsuitable for Eu-doped GaN, where the lowest unoccupied Eu 4f levels are much deeper (and not ~40 meV below the CBM)66,67,68. Alternatively, a significant expansion of the lattice in the Eu-doped layers of the superlattice structure has been observed20. Since the nature of the bandgap is correlated with the lattice constant49,69,70,71, it is expected that even 0.1% doping will result in a decrease in the bandgap due to the lattice expansion in Eu-doped GaN. In addition, increases in the index of refraction have also been reported in a number of dilutely doped RE-doped systems such as Er-doped GaN, and Er-, Sm- or Eu-doped InGaAs, and has been suggested as a simple mechanism to develop optically active waveguide devices24,36,37,72,73,74,75. The increase in index of refraction and decrease of the bandgap have been computed simultaneously using first-principles methods68,73. For example, Maskar et al. predicted these behaviors for GaN doped with >6.25% Eu, Sm and Pm68. Experimentally, however, these quantities are not measured or reported simultaneously.

Experimental results that demonstrate a bandgap modification are usually obtained by observing a red-shift in the absorption spectra of the host material, where impurity states and phonon modes can make the analysis complex27,28,29,76. In addition, the primary reason for the RE doping is usually to enhance the properties of the host, and not to enhance the emission from the RE ions themselves28,32,33. Conversely, the index of refraction is generally determined by fitting the results of reflectivity or ellipsometry measurements, for example in Er-doped GaAs, Er-doped GaN or Er-doped Si24,36,37,72. The benefit of using THz measurement techniques to evaluate RE-doped semiconductor systems, as was shown here, is that the bandgap, refractive index, and carrier concentration of RE-doped semiconductors can be determined, as well as barrier potentials for electrons within superlattice structures, using only THz techniques.

Regarding the performance of RE-doped GaN-based optoelectronics devices, the observed bandgap reduction and refractive index increase have significant benefits, particularly when a dilutely doped superlattice structure is implemented. While Er-doped Si superlattice structures have been shown to exhibit a higher performance than bulk Er-doped Si, the underlying principle behind this behavior has not been addressed in detail21,22,23,24,25. The results from this work could provide new insight into the origin of the enhancement due to the superlattice structure. In addition, RE doping has been shown to increase the refractive index of host materials experimental approaches in Er-doped GaN36,37,74,75 or through first-principles calculations68. The optical bandgap and index of refraction are related quantities35, which suggests that RE doping will reduce the bandgap of host semiconductors. In this case, the bandgap reduction in the RE-doped superlattice structure induces the carrier confinement within the active layers, which enhances the carrier recombination and luminescence efficiency16,60,61. Additionally, the higher index of refraction will result in a waveguiding effect37,72,75, increasing the extraction efficiency. These benefits are achieved even at a very low doping concentration of ~0.1%, which also facilitates high crystal-quality growth. Moreover, the reduction of threading dislocations due to the alternating strain between Eu-doped layers and GaN layers was already observed20. Taking all of these effects into consideration, it seems that introducing a superlattice structure with dilute Eu doping naturally enhances all of the properties needed to produce high brightness, high-efficiency red GaN LEDs. Critically, we argue that these enhanced optical and electrical properties are not limited to the Eu-doped GaN system. Er-doped GaN, for example, is known to have a higher refractive index than GaN, and GaN/GaN:Er waveguides and optical amplifiers that take advantage of this have already been reported36,37,74,75. By introducing multiple alternating layers, an efficient Er-doped GaN LED operating at the telecom C-band could be achieved.

The use of carrier confinement by introducing a MQW structure is the standard way to increase carrier recombination in commercial LEDs1,2,3,4,5,6. However, although introducing an MQW structure consisting of alternating Eu-doped GaN or Er-doped GaN layers and AlGaN barrier layers was shown to improve the PL intensity and the excitation efficiency of devices77,78,79,80,81; the simple superlattice devices have consistently been found to significantly outperform the MQW devices16,77,78. One reason could be that the optimal growth conditions of the AlGaN barrier layers needed to form intentional QWs are substantially different from those optimized for the RE-doped GaN layers. Therefore, a sacrifice must be made regarding the quality of either the RE-doped GaN layers or the AlGaN layers. In addition, Xie et al. reported that large potential barriers in MQW structures can interfere with hole transport and cause poor LED performance65. For the dilutely doped superlattice structure, the potential barriers for the electrons and holes were found to be <27 meV each, and are therefore not expected to impact carrier transport, while still offering the benefits of carrier confinement. Moreover, the growth conditions between Eu-doped GaN and the GaN layers are close enough, that high crystal quality can be achieved. Figure 5 shows a schematic of the THz and PL emission in a superlattice structure with the determined barrier heights.

Fig. 5: Schematics of a carrier confinement and band diagram in the superlattice structure.
figure 5

a Schematic of the carrier dynamics and THz emission in the superlattice. Photocarriers are captured into Eu-doped layers due to the bandgap difference between undoped and Eu-doped layers, leading to a higher carrier recombination probability and increased red luminescence. b Energy diagram near the surface in superlattice. Ec and Ev represent the energy levels for conduction and valence bands, respectively. Bandgap energies (\({E}_{{{{{{\rm{g}}}}}}}\)) for undoped and Eu-doped layers are 3.430 and 3.393 eV, respectively. The potential barriers in the conduction and valence band edge were estimated to be 11 meV and 26 meV, which capture the carriers within Eu-doped layers efficiently without interfering with their transports. Enhanced carrier recombination within the active layers and waveguiding effect due to the refractive index differences between undoped and Eu-doped layers significantly increases the luminescence efficiency in LEDs, as shown in the (a) and ref. 16.

Full size image

Conclusion

Overall, this study confirms that dilute doping of REs into semiconductors produces significant modifications in the optical properties of the host material. By utilizing a superlattice structure, these modified properties naturally produce both carrier confinement and waveguiding effects, thereby enhancing the energy transfer efficiency due to improved carrier recombination, but at a low enough level so as to not impede carrier mobility, as well as increasing the light extraction efficiency. We assert that this is also the origin for the enhanced LED properties previously observed in Er-doped Si superlattice structures, especially given the more recent reports on the increase in the index of refraction and reduction of the bandgap in Er-doped Si. Thus, our findings have answered long-standing questions in the field of RE-based optoelectronic devices as well as helped to unify previous observations and calculations for RE-doped semiconductor systems used for a variety of different applications within numerous fields. Given the universality of bandgap reduction and refractive index increase in a myriad of host semiconductor systems (e.g., GaN, ZnO, TiO2, SnO2, GaAs, InGaAs, InP, and Si) with different RE elements (e.g., Eu, Er, Tm, Sm, Gd, Tb), we propose that the superlattice structure should be used as a general approach to enhance the performance of RE-based optoelectronic devices. Furthermore, in the present work, TES is shown as an ideal tool allowing for direct characterization of ultrafast carrier dynamics within RE-doped semiconductors in non-contact/non-destructive manner. Such measurements will be needed to accelerate research on next-generation RE-based optoelectronic devices based on the dilutely doped superlattice structure as TES can yield insight on the bandgap, index of refraction, potential barriers, and carrier density, allowing for a quick and comprehensive analysis.

Methods

Sample fabrication

Bulk undoped GaN (ud-GaN), bulk Eu-doped GaN, and superlattice samples were grown by organometallic vapor-phase epitaxy on c-plane sapphire substrates, as shown in Fig. 2. The ud-GaN sample had a 5 μm ud-GaN layer, and one Eu-doped GaN sample had 700 nm thick GaN:Eu layer on a 30 nm low-temperature-grown GaN (LT-GaN) buffer layer, which were named ud-GaN and GaN:EuTDS (Fig. 2a, b). The other Eu-doped GaN and superlattice samples were initiated with a 30 nm LT-GaN buffer layer followed by a 2 µm ud-GaN layer (Fig. 2c, d). These Eu-doped GaN samples had 300 nm Eu-doped layers followed by 10 nm GaN cap layers. To explore the relationship between Eu doping density and the optical properties, we prepared two GaN:Eu samples with different Eu flow rates, 1.5 slm and 0.5 slm, during GaN:Eu layer growth, which were labeled as GaN:Euhigh and GaN:Eulow, respectively. GaN:Euhigh is expected to have a doping density of \(3\)-\(5\times {10}^{19}\,{{{{{{\rm{cm}}}}}}}^{-3}\) 16,19,82,83. The superlattice sample consisted of 40 pairs of 10-nm-thick GaN layers and 3-nm-thick GaN:Eu layers, which were grown with an Eu flow rate of 1.5 slm. Further details of the sample growth have been reported elsewhere16,19,82,83.

Optical measurements

THz emission measurements were performed using a TES set-up illustrated in Fig. 6. Femtosecond pulses obtained from a Ti:sapphire laser (center wavelength: 710–740 nm, pulse width: ~100 fs, repetition rate: 80 MHz) were converted into UV pulses through a Beta Barium Borate (BBO) crystal to excite GaN. Samples were excited with a pump power of 6 mW at an incident angle of 45° to obtain the large signal under a phased-array effect84. The inset of Fig. 6 shows the geometrical configuration for excitation and emission. The beam spot diameter was approximately 300 μm. Generated THz waves were focused onto a detector by two parabolic mirrors and detected using a dipole-type photoconductive antenna (PCA) fabricated on a low-temperature-grown gallium arsenide. The samples were mounted on a liquid nitrogen cryostat to measure the temperature dependence. Terahertz time-domain spectroscopy (THz-TDS) was used to obtain the free carrier density in GaN:Eu samples. The measurements were performed using a transmission type THz-TDS system (TR-1000; Otsuka Electronics Co. Ltd.) at 23 °C and relative humidity of 16%. Figure 7 shows the experimental configuration. Data were acquired using TR-1000 measurement software (Otsuka Electronics Co. Ltd.).

Fig. 6: Diagrams of the TES systems.
figure 6

The system employed a wavelength-tunable fs laser and a cryostat for wavelength and temperature dependent measurements. THz emissions are detected by a photoconductive antenna (PCA). The inset shows the excitation geometry and transient displacement of carriers which generates the THz emission.

Full size image
Fig. 7: Experimental configuration of THz time-domain spectroscopy (THz-TDS).
figure 7

Transmission type THz-TDS system (TR-1000; Otsuka Electronics Co. Ltd.) was employed. The measurements were performed at 23 °C and relative humidity of 16%.

Full size image

Determination of barrier heights in the superlattice structure

To describe the relationship between the THz amplitude and the carrier mobility in GaN:Euhigh, we employed the electron mobility model reported in ref. 63 and replaced the mobility in GaN with that in Eu-doped GaN, \({\mu }_{{{{{{\rm{Eu}}}}}}}\), as,

$${E}_{{{{{{\rm{THz}}}}}}}\left(T\right)\propto {\mu }_{{{{{{\rm{Eu}}}}}}}\left(T\right)\propto {\mu }_{{{{{{\rm{Eu}}}}}}}\left(300\right)\frac{{{{{{\rm{B}}}}}}\left(N\right){\left(\frac{T}{300}\right)}^{\alpha }}{1+{{{{{\rm{B}}}}}}\left(N\right){\left(\frac{T}{300}\right)}^{\alpha +\beta }},$$
(5)
$${{{{{\rm{B}}}}}}\left(N\right)=\frac{{\mu }_{\min }+{\mu }_{\max }\left(\frac{{N}_{{{{{{\rm{g}}}}}},{{{{{\rm{i}}}}}}}}{N}\right)}{{\mu }_{\max }-{\mu }_{\min }},$$
(6)

where \(\alpha\) and \(\beta\) are fitting parameters related to the impurity scattering at low temperatures and the lattice scattering at high temperatures, respectively. The values of \({\mu }_{\min },{\mu }_{\max }\) and \({N}_{{{{{{\rm{g}}}}}},{{{{{\rm{i}}}}}}}\) are estimated from the carrier density dependence of the mobility in bulk GaN, reported in ref. 63, and \(N\) is the carrier density. In our calculations, we used \({\mu }_{{{{{{\rm{Eu}}}}}}}\left(300\right)=53{{{{{{\rm{cm}}}}}}}^{2}\,{{{{{{\rm{V}}}}}}}^{-1}\,{{{{{{\rm{s}}}}}}}^{-1}\) for the mobility in Eu-doped GaN as reported in ref. 17. We also used the previously determined free carrier density was for GaN:Euhigh, \(N \sim 6\times {10}^{17}\,{{{{{{\rm{cm}}}}}}}^{-3}\). By fitting the temperature dependent THz amplitudes in Fig. 4 with Eq. (5), we obtained \(\alpha \sim -0.2\pm 0.2,\beta \sim 5.3\pm 0.2\). The value of \(\alpha\) has a large relative error; however, since the contribution of the impurity scattering is significantly less in the measured temperature range, the exact value for α should have a negligible impact on the results63.

Using these parameters for GaN:Euhigh, we can now evaluate the impact of the superlattice structure on the carrier mobility between the doped and undoped layers and determine the barrier potential for the electrons within this structure by using a mobility model that takes into account the barrier height, reported in ref. 62. Although this model includes some transport processes, the mobility due to the thermal emission process, \({\mu }_{{{{{{\rm{th}}}}}}}\), should be dominant in superlattice structure due to the 10 nm thick barrier layers62. With these considerations, the THz emission amplitude within the superlattice structure can be described as:

$${E}_{{{{{{\rm{THz}}}}}}}\left(T\right)\propto {\mu }_{{{{{{\rm{th}}}}}}}\left(T\right)\propto \left(1+\frac{{\mu }_{3{{{{{\rm{D}}}}}}}\left(T\right)}{{qL}}\sqrt{8{m}_{3{{{{{\rm{D}}}}}}}^{* }q{V}_{{{{{{\rm{b}}}}}}}}\right)\exp \left(-\frac{q{V}_{{{{{{\rm{b}}}}}}}}{{k}_{{{{{{\rm{B}}}}}}}T}\right),$$
(7)

where \({\mu }_{3{{{{{\rm{D}}}}}}}\) and \({m}_{3{{{{{\rm{D}}}}}}}^{* }\) are the mobility and effective mass of the electrons, \(q\) is the elementary charge, \(L\) is the sum of GaN and Eu-doped GaN layer thicknesses, \({k}_{B}\) is the Boltzmann constant, and \(q{V}_{{{{{{\rm{b}}}}}}}\) represents the barrier height62. Here, we use \({\mu }_{3{{{{{\rm{D}}}}}}} \sim {\mu }_{{{{{{\rm{Eu}}}}}}}\) and \({m}_{3{{{{{\rm{D}}}}}}}^{* } \sim {m}_{{{{{{\rm{GaN}}}}}}}^{* }\): the effective mass of electron in the bulk GaN64, assuming there is no significant difference between these values in the Eu-doped layer within the superlattice structure and bulk GaN:Euhigh. It should be noted that the calculated value of \(q{V}_{{{{{{\rm{b}}}}}}}\) changes by less than 1 meV even when calculated with \(2{m}_{{{{{{\rm{GaN}}}}}}}^{* }\), justifying this assumption.