Enhancement of the Electroluminescence from Amorphous Er-Doped Al₂O₃ Nanolaminate Films by Y₂O₃ Cladding Layers Using Atomic Layer Deposition

Abstract

Amorphous Al₂O₃-Y₂O₃:Er nanolaminate films are fabricated on silicon by atomic layer deposition, and ~1530 nm electroluminescence (EL) is obtained from the metal-oxide-semiconductor light-emitting devices based on these nanofilms. The introduction of Y₂O₃ into Al₂O₃ reduces the electric field for Er excitation and the EL performance is significantly enhanced, while the electron injection of devices and the radiative recombination of doped Er³⁺ ions are not impacted. The 0.2 nm Y₂O₃ cladding layers for Er³⁺ ions increase the external quantum efficiency from ~3% to 8.7% and the power efficiency is increased by nearly one order of magnitude to 0.12%. The EL is ascribed to the impact excitation of Er³⁺ ions by hot electrons, which stem from Poole-Frenkel conduction mechanism under sufficient voltage within the Al₂O₃-Y₂O₃ matrix.

1. Introduction

Rare earth (RE) ions are generally efficient luminescence centers in various matrices. Nowadays diverse RE-doped insulating materials have been developed for the applications in solid state lasers and phosphors [1,2]. Erbium (Er) ions are one of the most researched luminescence centers due to their near-infrared (NIR) 1.53 μm emission which coincides with the window of optical telecommunication [3,4] Aiming for the realization of Si-integrated optoelectronics, the 1.53 μm electroluminescence (EL) from Er³⁺ ion has been researched extensively in many materials, including SiO_x, SiN_x, TiO₂ and ZnO [5,6,7,8]. However, the efficiencies of the devices based on these aforementioned materials are still far from practical application, due to the limitations in doping tolerance and excitation efficiency. Y₂O₃ is one of the attractive doping hosts for RE ions as the substitution of other RE³⁺ ions in Y₂O₃ is quite easy without charge compensation and severe lattice distortion. In addition, Y³⁺ ions are not luminescent and Y₂O₃ has a large bandgap (5.8 eV) and high stability [9,10]. In our previous study, Al₂O₃ has been proved to be a suitable matrix for the excitation of RE³⁺ ion to realize the EL emissions but the doping concentration is still limited [11,12,13]. Therefore, using Y₂O₃ as a cladding layer in Er-doped Al₂O₃ could utilize the merits of both oxides, the Er-clustering and resultant concentration quenching could be reduced while the optical-active Er³⁺ ions can be excited more effectively [14].

In this work, we fabricate the metal-oxide-semiconductor light-emitting devices (MOSLEDs) based on the amorphous Al₂O₃-Y₂O₃:Er nanolaminate films, which are deposited using atomic layer deposition (ALD). Due to the unique growth mechanism based on the successive self-limiting gas-surface reactions, ALD realizes the precise control of the thickness of different compositions with excellent homogeneity [15,16]. By alternating deposition sequence of Al₂O₃ and Y₂O₃, nanolaminate Al₂O₃-Y₂O₃:Er films with interlayers of different thicknesses are fabricated. Under sufficient forward bias, such devices exhibit ~1530 nm emissions originating from the infra-4f transitions of Er³⁺ ions. Inserting of the Y₂O₃ cladding layers increases the external quantum efficiency (EQE) from 3% to 8.7% and almost upgrades the power efficiency (PE) by one order of magnitude, while the excitation and recombination of the Er³⁺ ions are not affected. We believe that this work contributes to the development of silicon-based light sources for integrated optoelectronic applications.

2. Experimental

The luminescent Al₂O₃-Y₂O₃:Er nanolaminates were grown on <100>-oriented n-type silicon (2–5 Ω·cm) using the thermal ALD system (NanoTech Savannah 100, Cambridge, MA, USA). The growth chamber was first evacuated to a base pressure of 0.3 Torr. Trimethylaluminum [TMA, Al(CH₃)], Y(THD)₃ and Er(THD)₃ (THD = 2,2,6,6-tetramethyl-3,5-heptanedionate) were used as the precursors for Al₂O₃, Y₂O₃ and Er₂O₃, respectively, with ozone acting as the oxidant. During the ALD process, the Al precursor was maintained at room temperature (RT), while Y and Er precursors were maintained at 180 °C and 190 °C, respectively. The precursor delivery lines were heated at 190 °C. N₂ was used as the carrier and purge gas with a flow rate of 20 sccm. The pulse time for Al and RE precursors are 0.015 s and 2 s, respectively. One growth cycle consists of one precursor pulse, the 5 s N₂ purge, a 1.8 s ozone pulse, and the 9 s N₂ purge. Based on the former research, the Er dopant cycles are fixed at 2, which are preferable concerning both the efficient doping and the absence of RE clustering [11,13,17,18]. The substrates were maintained at 350 °C, and the growth rates for the Al₂O₃, Y₂O₃ and Er₂O₃ films are calibrated to 0.79, 0.2 and 0.23 Å/cycle respectively, which agree well with the previous reports [19]. During the deposition, the dopant Er₂O₃ atomic layers were sandwiched in two cladding Y₂O₃ layers of designed thickness, and then the Al₂O₃ interlayers with certain thickness and these Y₂O₃-Er₂O₃-Y₂O₃ composite nanolaminates were deposited repeatedly to achieve the nanolaminates with the deposition sequence of Al₂O₃-Y₂O₃-Er₂O₃-Y₂O₃. In order to explore the Al₂O₃-Y₂O₃:Er nanofilms, firstly for the Al₂O₃-Y₂O₃:Er nanofilms of different Y₂O₃ cladding layers, the thickness of Al₂O₃ interlayers was fixed at 3 nm and the two Y₂O₃ cladding layers (x nm) in each supercycle were changed from 0 to 0.2, 0.5 and 1.0 nm (with their growth cycles varied from 10 to 50), the same repeat numbers of 16 for the supercycles resulted into the total thickness of 48.6, 55.0, 64.6, and 80.6 nm for the Al₂O₃-Y₂O₃(x nm):Er nanolaminates. The calculated nominal doping concentrations of Er are 0.51–0.34 at%. Secondly, for the Al₂O₃-Y₂O₃:Er nanofilms of different Al₂O₃ interlayers, the thickness of Y₂O₃ cladding layers were fixed at 0.2 nm and the Al₂O₃ interlayers (y nm) in each supercycle were changed from 0.5 to 1, 2, 3 and 5 nm. To achieve the Al₂O₃-Y₂O₃:Er nanofilms of the total thickness of ~65 nm, the repeat numbers of the supercycle were changed from 69 to 45, 27, 19 and 12 for the Al₂O₃(y nm)-Y₂O₃:Er nanolaminates. The calculated nominal doping concentrations of Er are 4.28–0.29 at%. Here the deposition velocities and the growth cycles in recipes, and the densities of oxides (Al₂O₃, Y₂O₃, Er₂O₃) are used to calculate the corresponding dopant amount of Er³⁺ ions. After the deposition, the films were annealed at 800 °C in N₂ atmosphere for 1 h to enable activation of the dopants. Subsequent device procedures were as previously mentioned [12,13,17,18], resulting in the multilayer-structured MOSLEDs of ZnO:Al/TiO₂-Al₂O₃/Al₂O₃-Y₂O₃:Er/Si/Al. The top ZnO:Al electrodes were lithographically patterned into 0.5 mm circular dots, while the TiO₂-Al₂O₃ nanolaminates were used to enhance the operation stability of the devices.

The film thickness was measured by an ellipsometer with a 632.8 nm He-Ne laser at an incident angle of 69.8°. The phase and the crystal structure of the films were identified by an X-ray diffractometer (XRD, D/max 2500/pc, Rigaku) using the Cu Kα radiation. To activate EL from the MOSLEDs, appropriate forward bias was applied with the negative voltage connecting to the n-Si substrates. EL and Current-Voltage (I–V) characteristics were recorded by a Keithley 2410 SourceMeter. The EL signal was collected by a 0.5 m monochromator and detected by an InGaAs detector connected to a Keithley 2010 multimeter. The absolute EL power from the device surface was measured using a calibrated Newport 1830-C optical power-meter with an 818-IR Sensor. All measurements were performed at RT.

3. Results and Discussion

The XRD patterns of all the Al₂O₃-Y₂O₃:Er films annealed at 800 °C confirm that the nanolaminates are amorphous, one representative XRD pattern from the nanolaminate using 3 nm Al₂O₃ interlayers and 0.2 nm Y₂O₃ cladding layers is shown in Figure 1. The Al₂O₃ layers are not crystalized at such a relatively low temperature of 800 °C that beneficial for the EL performance from RE-doped Al₂O₃ films, while the crystallization of the sub-nanometer Y₂O₃ layers is restricted [17,20]. The amorphous nanolaminate films are quite smooth under the observation of scanning electron microscope, with a root-square roughness of only 0.56–0.7 nm scanned by the atomic force microscopy (AFM, Dimension Icon, Bruker) [21].

Figure 2a illustrates the schematic diagram for the MOSLEDs and the structure and deposition sequence of the luminescent nanolaminates. Figure 2b shows the NIR EL spectra of MOSLEDs based on the Al₂O₃-Y₂O₃:Er films of different Y₂O₃ cladding layers (with the thickness of x nm). The EL peaks centered at ~1530 nm correspond to the infra-4f ⁴I_13/2→⁴I_15/2 transitions of the Er³⁺ ions. The presence of other shoulder peaks is ascribed to the splitting levels associated with the Stark effect [22]. These EL peaks are similar in positions and sharps in the Al₂O₃-Y₂O₃:Er nanofilms with different Y₂O₃ cladding layers, thus the incorporation of Y₂O₃ cladding layers imposes no apparent effect on the Er³⁺ intra-4f transitions. In comparison with the Er-emissions from different matrices, the spectra also confirm that the Al₂O₃-Y₂O₃:Er films are amorphous due to the absence of companion peaks [20,23].

Figure 3a presents the dependence of the 1530 nm EL intensities and the injection currents on the applied voltages for the Al₂O₃-Y₂O₃:Er MOSLEDs with different Y₂O₃ cladding layers (with the thickness of x nm). These EL–V and I–V curves are similar with our previous reports on the MOSLEDs based on RE-doped oxides, with the typical characteristic of MOS structures [13,18,24,25]. Beneath the threshold electric field, the defect states contribute to the low background currents. In the working voltage region, the currents increase exponentially until breakdown. The difference on the current injection will be discussed afterwards concerning the conduction mechanism. All the EL intensities also present an exponential relationship with the applied voltages until reaching saturation. The MOSLED with 0.2 nm Y₂O₃ cladding layers presents the highest EL intensity, with the lowest threshold voltage and the highest injection current. The devices with thicker Y₂O₃ layers underperform in EL intensities and the injection currents are restricted. Despite the uncertainty brought about by the device preparation, Y₂O₃ cladding layers with suitable thickness can effectively enhance the current injection and promote the EL emissions from these Al₂O₃-Y₂O₃:Er MOSLEDs. As previously reported, the incorporation of Y³⁺ ion makes the crystal field around Er³⁺ ions less symmetric and introduces distortion in the crystal field, moreover the Er³⁺ ions are dispersed to suppress the concentration quenching, resulting in the enhanced radiation probability [26]. Therefore, the 0.2 nm Y₂O₃ layers act as cladding layers that inhibit the Er-clustering, while the thicker Y₂O₃ layers inhibit the electron injection, which is ascribed to the higher dielectric index of Y₂O₃ and the disruptive interfaces among Al₂O₃ and Y₂O₃ interlayers.

Figure 3b shows the dependence of EL intensities on the injection currents for the Al₂O₃-Y₂O₃:Er MOSLEDs with different Y₂O₃ cladding layers. The threshold currents for all the devices are ~0.2 μA, the EL intensities and the injection currents present linear relationship. In comparison, the devices with different Y₂O₃ cladding layers exhibit similar EL, which increases more prominently than that based on the Al₂O₃:Er film. Y₂O₃ also lessens the saturation of EL intensities at higher injection currents. Considering the thick interlayers among RE layers (the Al₂O₃ interlayers with the thickness of at least 3 nm), the acceleration distance for hot electrons is sufficient; therefore the enhanced EL should result from more optical-active Er dopants as the Er³⁺ ions disperse into the Y₂O₃ layers and the Er-clustering is suppressed.

In our previously reported MOSLEDs based on RE-doped Al₂O₃, the RE-related EL is triggered by the direct impact excitation of the RE ions by the hot electrons accelerated under sufficient bias voltages [12,25]. As the I–V characterization are accordingly comparable, it is rational to ascribe the NIR EL from these Al₂O₃-Y₂O₃:Er MOSLEDs to the same mechanism. Considering the high bandgap of the matrix materials and the barrier for electrons to be injected from the Si substrates into the conduction band of the oxides, the current conduction of these MOSLEDs has been ascribe to the Poole-Frenkel (P-F) mechanism, in which the electrons hop via the defect-related trap states under sufficient electrical field [11,26,27]. In simplicity, the plot of the ln(J/E) versus E^1/2 presents linear relationship in P-F conduction mechanism, where J and E are the current density and the electric field, respectively [28,29]. Figure 3c shows the plots of the I–V characteristics derived from Figure 3a, the electrical fields across the luminescent films are roughly calculated in terms of electrostatics [30], and the well-defined linearity is established for all the MOSLEDs in the EL-enabling region. Thus the electron transport through the Al₂O₃-Y₂O₃:Er nanofilms is governed by the P-F mechanism. These electrons tunnel into the conduction band of oxides and transport by hopping among trap states in the Al₂O₃-Y₂O₃ nanolaminates under sufficient electric field. Certain parts of the electrons are accelerated therein and become hot electrons that excite the Er³⁺ ions by inelastic impact, the subsequent recombination gives rise to the characteristic EL emissions. Apparently, the Y₂O₃ cladding layers decrease the working electric field prominently. As mentioned in the discussion on the I–V characteristics, the Y₂O₃ cladding layers increase both the injection currents and EL intensities, we conclude that ultrathin Y₂O₃ layers introduce defect sites within Al₂O₃, via which electrons transport by the P-F hopping mechanism; therefore the injection currents are enhanced. Since the accelerated electrons collide with the doped Er³⁺ ions and contribute to the NIR EL, adding the aforementioned crystal field distortion and cluster dispersion effects of the Y₂O₃ on Er³⁺ ions, the EL performance are greatly enhanced by the Y₂O₃ cladding layers in Al₂O₃ films. However, the Y₂O₃ cladding layers should be thin enough to not impact the carrier transport which could be ascribed to the formation of distinct Al₂O₃-Y₂O₃ interfaces when using thicker Y₂O₃ cladding layers.

In evaluation of the thickness of Al₂O₃ interlayers on the EL performance, the dependence of the EL intensities from each dopant cycle on the injection currents for the Al₂O₃-Y₂O₃:Er MOSLEDs with different Al₂O₃ interlayers (with the thickness of y nm) are shown in Figure 4a, the thickness of Y₂O₃ cladding layers is the optimal 0.2 nm. Again, the EL intensities increase almost linearly with the injection currents. The difference on the EL–I–V characteristics among these MOSLEDs with different Al₂O₃ layers are small (not shown herein), and the increase in EL intensity with the Al₂O₃ thickness could be ascribed to the less concentration quenching of doped Er³⁺ ions together with the longer acceleration distance. When the thickness of Al₂O₃ declines, the EL intensity decreases greatly due to the cross-relaxation of Er³⁺ between adjacent dopant layers when the inter-distance (the Al₂O₃ thickness) is smaller enough, and the limited acceleration length for the hot electrons to gain energy to excite the Er³⁺ ions [18,21,31,32,33]. Cross-relaxation is a common phenomenon that occurs among the same ions or different ions of similar energy intervals. One ion in the excited state (⁴I_13/2 in the case of Er³⁺ ion) transfer the energy to another one (in the ground state of ⁴I_15/2 in this case of Er³⁺ ions), excite the latter to higher energy levels (⁴I_13/2) while relaxing itself to lower energy levels (⁴I_15/2) without radiation. The interaction of energy transfer by cross-relaxation could finally disperse the excitation energy through phonons instead of luminescent emissions.

In the RE-doped Al₂O₃ MOSLEDs, the Al₂O₃ sublayer thickness affects the cross relaxation between excited RE ions, and the acceleration distance for injected electrons. Figure 4b shows the dependence of the integrated 1530 nm EL intensity per Er cycle on the thickness of Al₂O₃ interlayers under different injection currents. Under all these injection currents, with the increase in the thickness of Al₂O₃ interlayers, the contribution of single Er cycle to the EL intensity firstly increases and then saturates as the Al₂O₃ interlayer thickness reaches 3 nm. This is still in consistency with the common characteristic for the luminescent RE³⁺ ions in Al₂O₃ matrix that the distance for the presence of non-radiative interaction and adequate electron acceleration is around 3 nm [11,12,21,33].

Considering the total EL intensity from the MOSLEDs with Al₂O₃ interlayers of different thicknesses (marked as y nm here) shown in Figure 5a, the device using 3 nm Al₂O₃ interlayers presents the optimal emission intensity in the operation range, with the highest power density of 4.6 mW/cm². External efficiency is widely used to evaluate LED performance. Figure 5b shows the EQE and PE of these MOSLEDs based on different Al₂O₃-Y₂O₃:Er nanolaminate films. These EL efficiencies sustain a broad maximum, and fall down at higher currents. Generally, the EQE of the devices with 2–3 nm Al₂O₃ interlayers are the highest. As aforementioned, this phenomenon could be ascribed to the sufficient distance for electron acceleration and suppression of the cross-relaxation among adjacent Er₂O₃ dopant layers. The Y₂O₃ cladding layers somewhat decrease this critical distance which is beneficial for higher doping concentrations. The optimal device with 3/0.2 nm Al₂O₃/Y₂O₃ interlayers achieves the maximum EQE of 8.7% and a corresponding PE of 0.12%. These values are comparable to our Yb₂O₃:Er MOSLEDs but with lowered working voltages. In comparison, the control Al₂O₃:Er MOSLED presents only an EQE of 3% and a PE of 0.014%, much lower than the Al₂O₃-Y₂O₃:Er MOSLEDs. The Y₂O₃ cladding layers with suitable thickness enhance the efficiencies from the MOSLEDs to a great extent. We have found that by using a thicker luminescent layer, the efficiency of the Al₂O₃:RE MOSLED might be further increased to higher than 10% [25,34]. These efficiencies are superior to that from Si-based EL devises in literature, thus further optimization of the luminescent Al₂O₃-Y₂O₃:Er nanolaminates would supply potential light source for the applications in Si-based optoelectronics.

4. Conclusions

In summary, significantly enhanced ~1530 nm NIR EL emissions are achieved from the MOSLEDs based on the amorphous Al₂O₃:Er nanolaminate films by the insertion of cladding Y₂O₃ sub-nanolayers, which are fabricated by ALD on Si substrates. The Y₂O₃ cladding layers reduce the threshold electric field for excitation and increase the radiative possibility of doped Er³⁺ ions, resulting in improved EL performance. The Al₂O₃-Y₂O₃:Er MOSLEDs with 0.2 nm Y₂O₃ and 3 nm Al₂O₃ interlayers present an EQE of 8.7% and a corresponding PE of 0.12%, which are much higher than that of the counterpart without Y₂O₃ cladding layers. The incorporation of Y₂O₃ does not change the electron injection mode under sufficient electric field that conforms to P-F mechanism, the resultant energetic electrons trigger the impact-excitation of Er³⁺ ions and subsequent EL emissions. The strategy of Y₂O₃-cladding by ALD can be employed to improve the EL performance from LEDs based on RE-doped oxides.