1. Introduction

Electroluminescence (EL) from the light-emitting devices based on rare earth (RE)-doped dielectrics and semiconductors has stimulated worldwide interest over the years, aiming at achieving optoelectronic integration on silicon [1,2]. The RE³⁺ ions incorporated into various hosts emit sharp and wavelength-stable light in a wide spectral range, originating from the intra-4f transitions [3–5]. Moreover, as the RE-related emissions are considerably insensitive to the local environment, there is no stringent requirement on the crystalline quality of host materials. Contemporary optoelectronic devices based on GaInN materials suffer from a steep drop of efficiency when the emission wavelength is designed within the green region, which is well known as “green gap” [6]. In this context, EL from the terbium (Tb)–incorporated films is of special interest because of its intense green emission at ~540 nm, which has already been widely used in organic light emitting devices. So far there have been several reports on the EL from Tb-doped materials on silicon [7,8]. Berencen et al. reported the green EL device based on the Tb-implanted SiN_x films on SiO₂/Si [9]. Rebohle et al. demonstrated strong EL from the metal-oxide-semiconductor light emitting devices (MOSLEDs) composed of SiO₂-Tb₂O₃-Al₂O₃ multi-layers with the power efficiency up to 0.15% [10]. However, these prototype devices are far from practical application.

Al₂O₃ is a technologically important material due to its relatively high refractive index and dielectric constant, which allows well light confinement in the optical waveguide [11]. In addition, the good adhesion to silicon surface and high stability make Al₂O₃ attractive in the Si microelectronics and optoelectronic applications. Up till now, there are few reports on the EL from the devices with Tb-doped Al₂O₃, mainly due to the insulating property which impede efficient carrier injection and transport. However, the realization of green EL from Tb³⁺ ions in the Al₂O₃ film can be in principle expected, which has great potential in applications such as displays and lighting [12,13]. In this work, we fabricate the efficient Al₂O₃/Tb₂O₃ MOSLEDs by atomic layer deposition (ALD). The devices mainly emit green light centered at ~543 nm, together with weak blue emissions, which origin from the impact-excited Tb³⁺ ions by hot electrons. The EL intensities, decay dynamics and excitation cross sections are studied concerning variations of the Tb dopant cycles and the Al₂O₃ sublayer thickness, by precisely controlling of the nanolaminate structures utilizing ALD technique. The influence of cross relaxation on the EL emissions originating from different excited states are explored. This work highlights the possibility to use ALD for the dopant design of nanolaminate materials for future development of silicon-based optoelectronic devices from oxides doped with RE ions.

2. Experimental details

The Al₂O₃/Tb₂O₃ films were grown on <100>-oriented phosphorous-doped silicon (n-Si) substrates with a resistivity of 2-5 Ω·cm. Prior to the growth, the Si substrates were cleaned through the standard RCA process. The ALD equipment is a 4-in. chamber system (NanoTech Savannah 100, Cambridge). Trimethylaluminum [TMA, Al(CH₃)₃] and Tb(THD)₃ (THD = 2,2,6,6-tetramethyl-3,5-heptanedionate) were used as the metal precursors for Al₂O₃ and Tb₂O₃, respectively, while ozone was used as oxidant. During the ALD process, the TMA source was maintained at room temperature, while the Tb precursor and the precursor delivery lines were heated at 190 °C, the substrates were maintained at 350 °C. N₂ was used as the carrier and purge gas with a flow rate of 20 sccm. The growth chamber was firstly evacuated to a base pressure of 0.26 Torr. One Al₂O₃ or Tb₂O₃ cycle consist of 0.015 s TMA or 2 s Tb(THD)₃ pulse, 5 s N₂ purge, 1.8 s ozone pulse, and 9 s N₂ purge.

For exploration of the luminescent Al₂O₃/Tb₂O₃ film, two different series of devices were fabricated. Firstly, the ALD growth contained 4 Al₂O₃ cycles, and various Tb₂O₃ cycles including 1, 2, 4 and 6, in order to investigate the effect of dopant quantity on the EL property. Secondly, the Tb₂O₃ cycles were fixed at 2, while the cycles of Al₂O₃ varied from 2 to 100, corresponding to a Al₂O₃ sublayer of 0.15 to 7.43 nm. The corresponding experimental parameters are summarized in Table 1. The thickness of the films was measured by an ellipsometer with a 632.8 nm He-Ne laser beam at an incident angle of 69.8°. Based on our former results, the growth velocity maintains at around 0.743 Å per cycle for Al₂O₃, while that for Tb₂O₃ is slower and susceptible to the surface condition. Thus the growth rate of Al₂O₃ and the final thickness of the films are used to calculate the Tb₂O₃ thickness and corresponding Tb doping concentration. The films were subsequently annealed at 800 °C in N₂ atmosphere for 1 h to reduce defects and enable Tb activation. X-ray diffraction patterns (D /max 2500/pc, Rigaku, Cu Kα radiation, λ = 1.5406 Å) of the annealed films exhibit no diffraction peaks (not shown herein), indicating the amorphous Al₂O₃ matrix and no apparent TbO_x segregation. Afterwards, ~100 nm TiO₂/Al₂O₃ multilayer films composed of ~1.94 nm Al₂O₃ and 8.0 nm TiO₂ sublayers were deposited by ALD onto the Al₂O₃/Tb₂O₃ films as the dielectric layer to enhance the device stability. ~100 nm thick ZnO:Al₂O₃ films were deposited onto the TiO₂/Al₂O₃ films by ALD, which were lithographically patterned into circular dots with a diameter of 0.5 mm. Then ~100 nm thick Al films were deposited on the backsides of n-Si substrates by heat evaporation, resulting in MOSLEDs with a multi-layer structure of ZnO:Al/TiO₂-Al₂O₃/Al₂O₃-Tb₂O₃/Si/Al.

Table 1. The corresponding experimental parameters for clarity

View Table

To activate EL from the MOSLEDs, appropriate forward bias was applied with the negative voltage connecting to n-Si substrates. Negative biases feature a lower EL at higher voltages. EL and Current-Voltage (I-V) characteristics were recorded by means of a Keithley 2410 SourceMeter unit. The EL signal was collected by a 0.5 m monochromator and detected by a Si detector connected to a computer controlled Keithley 2010 multimeter. The absolute EL power from the device surface was measured using a calibrated Si detector connected with a picoammeter. Photographic images were taken by a digital camera through the ocular of a microscope. The decay times of the EL emissions were measured by a SR430 multi-channel scaler (Stanford Research Systems) equipped with a waveform generator (DG5072, RIGOL) and a high-voltage amplifier. All measurements were performed at room temperature.

3. Results and discussion

The top-right inset in Fig. 1 presents a schematic diagram manifesting the structure of the luminescent Al₂O₃/Tb₂O₃ films fabricated with x Al₂O₃ sub-cycles and y Tb₂O₃ sub-cycles. Figure 1(a) shows the EL spectra from the MOSLEDs based on the Al₂O₃/Tb₂O₃ films with different Tb dopant cycles. The EL spectra mainly contain several peaks at the wavelengths of 488, 543, 587, and 623 nm, which origin from the intra-4f transitions of ⁵D₄→⁷F_J (J = 6, 5, 4, and 3, respectively) in Tb³⁺ ions [14]. The green light is easily seen by naked eyes, as shown in Fig. 1(b). These images were taken by a digital camera from the MOSLEDs in which the Al₂O₃:Tb₂O₃ sub-cycle ratio is 4:2, at different injection currents. The whitish color for higher currents is due to the saturation of the camera chip. Moreover, there are much weaker violet and blue EL peaks at 380, 415, 437 and 458 nm which are attributed to the ⁵D₃→⁷F_J transitions. Note that the EL intensity of these violet and blue peaks are multiplied by 50 for clarity. In comparison, emissions from the ⁵D₃ level monotonically decrease with the Tb doping concentration, while EL from the ⁵D₄ level shows the strongest peak at higher doping concentration (Al₂O₃:Tb₂O₃ = 4:2, 2.5 at%). Thus the higher ⁵D₃ level is more susceptible to concentration quenching.

 

Fig. 1 (a) EL spectra from the MOSLEDs based on the Al₂O₃/Tb₂O₃ films with different Tb dopant cycles, at the injection current of 50 μA. (b) The light-emitting images from the MOSLEDs in which the Al₂O₃: Tb₂O₃ sub-cycle ratio is 4:2, at different injection currents. The top-right inset shows the schematic diagram for the luminescent films.

Download Full Size | PDF

Figure 2(a) demonstrates the dependence of representative green (543 nm) and blue (437 nm) EL intensities, together with the current densities, on the applied voltages for the MOSLEDs based on different Al₂O₃/Tb₂O₃ films. All devices feature the typical I-V characteristic of a MOS structure, which start with a low background current for low electric fields, continues with approximately exponential increase of the current, and ends with a dielectric breakdown [14]. Higher Tb-doping concentration leads to less prominent increase of the current on voltage. The blue emissions are roughly three orders weaker than the green ones, also with a ~20 V larger threshold voltage. After the threshold, both the current densities and EL intensities increase exponentially with the applied voltages until the breakdown. Figure 2(b) shows the dependence of the EL intensities on the current densities for different MOSLEDs. Generally, with the increase of Tb concentration, the EL intensities gradually decrease under the same current density. The threshold currents for the green EL are around 0.1 μA (0.05 mA cm⁻²), and the blue EL from higher excited state ⁵D₃ starts at higher current of 1-10 μA (0.5-5 mA cm⁻²). The EL intensities all increase linearly with the current densities except that the green EL peaks begin to saturate above 80 V (50 mA cm⁻²), while the blue emissions continue to increase. The explanation is that at higher current density (the equivalent of higher electric field), the energy distribution of electrons in the conduction band of Al₂O₃ shifts to higher energies, which in turn alters the excitation probability in favor of higher energetic states. Green EL with an optical power up to 6.62 μW (power density of 3.37 mW cm⁻²) was measured from the MOSLED based on the Al₂O₃/Tb₂O₃ film with 2.5 at% Tb concentration, at 1.4 mA (0.71 A cm⁻²) and 88.7 V.

 

Fig. 2 (a) The dependence of representative green (543 nm) and blue (437 nm) EL intensities on the applied voltages, together with the I-V characteristics of the MOSLEDs based on different Al₂O₃/Tb₂O₃ films. (b) The dependence of the EL intensities on the current densities for these MOSLEDs.

Download Full Size | PDF

Figure 3(a) illustrates the decay curves of the 543 nm EL emissions from the MOSLEDs with different Al₂O₃/Tb₂O₃ films, all of which are nearly in the form of exponential decay function. With the increase of Tb doping concentration, the decay time of the 543 nm emission firstly decreases from 1.14 ms to 0.95 ms, then drastically drops to 0.32 ms for the device with the highest Tb-concentration. These radiative lifetimes are somewhat lower than that of 1.0-1.4 ms for Tb³⁺ ions in SiO₂ matrix [10]. It should be mentioned that the luminescent films are amorphous with no crystalline shown in the X-ray diffraction patterns. Since small agglomerates are formed in the beginning of the film growth in many ALD processes, it is speculated that excessive dopant cycles lead to the growth of Tb clustering in the deposition and annealing, which is responsible for the reduction of the luminescent Tb³⁺ ions [15,16]. Apparently the Tb-clustering due to excess Tb cycles serves as the main limitation when attempt to enhance the EL intensity by higher doping concentration via thicker Tb layer in the nanolaminate films deposited by ALD.

 

Fig. 3 (a) The decay traces of the 543 nm EL emissions from MOSLEDs based on different Al₂O₃/Tb₂O₃ films. (b) The plot of ln(J/E) versus E^1/2 (Poole-Frenkel plots of the I-V characteristics) for these MOSLEDs.

Download Full Size | PDF

As mentioned above, the blue emissions from higher ⁵D₃ energy level have a higher threshold voltage than the green ones from the ⁵D₄ level, and the green EL presents saturation while the blue one increases further with the current. Moreover, the intensity ratio between the 543 and 437 nm peaks increases with the voltage up for all the devices. These phenomena indicate that the emissions originate from the direct impact excitation of Tb³⁺ ions by hot electrons in the conduction band of Al₂O₃ matrix. The conduction mechanism has been proved to be Poole-Frenkel (P-F) conduction mode, in which the conduction mainly corresponds to the electrical field-enhanced emission of electrons from traps in the dielectric layer [17–19]. In simplicity, the plot of the ln(J/E) versus E^1/2 presents linear relationship in the EL-enabling voltages (J is the current density and E is the electric field). As presented in Fig. 3(b), well-defined linearity is established for all the MOSLEDs in the EL-enabling voltages. Therefore, the EL emissions from Al₂O₃/Tb₂O₃ are governed by the P-F mechanism: the electrons tunnel and transport into the conduction band of Al₂O₃ by virtue of the trap states under sufficient electric field. The injected electrons are accelerated by electrical field to reach an equilibrium energy distribution, which afterwards excite the Tb³⁺ ions through inelastic impaction, the following de-excitation process results in the characteristic EL.

By changing the cycle numbers of Al precursor in ALD growth, the Al₂O₃ sublayer thickness in the nanolaminate film is conveniently adjusted. Figure 4(a) presents the dependence of the 543 nm EL intensities, together with the current densities, on the applied bias voltages for the MOSLEDs based on the Al₂O₃/Tb₂O₃ films with different Al₂O₃ sublayers. The devices feature somewhat diversity in the I-V characteristics. The current density declines with lower Tb₂O₃ ratio, thus Tb dopant induces defect states into Al₂O₃ matrix which contributes to the carrier injection. The EL intensity increases with the doping concentration with the maximum at 2.5 at% Tb dopant (Al₂O₃:Tb₂O₃ = 4:2), a higher doping is adverse to the EL performance because of the cross-relaxation and concentration quenching.

 

Fig. 4 (a) The dependence of the 543 nm EL intensities, together with the current densities, on the applied voltages for the MOSLEDs based on Al₂O₃/Tb₂O₃ films with different Al₂O₃ sublayers. (b) The EL decay time for the 543 nm and 437 nm EL emissions, and the EL intensity from single Tb₂O₃ dopant layer at 50 μA (25.5 mA cm⁻²), as a function of the thickness of Al₂O₃ sublayers.

Download Full Size | PDF

For clarity, the dependence of the EL decay time and EL intensity from single Tb₂O₃ dopant layer on the Al₂O₃ sublayer thickness are compared in Fig. 4(b), for both the 543 nm and 437 nm EL at a constant current of 50 μA (25.5 mA cm⁻²). In evaluation of the contribution of a single Tb₂O₃ dopant layer on EL intensity, the green emission tends to decrease gradually as the Al₂O₃ sublayer thickness decreases and drops significantly with the thickness ranging from 3 nm to 0.15 nm, while the blue one maintains decreasing. The EL decay time of the 543 nm emission is much higher than that of the 437 nm emission, and the former stays nearly constant at 0.7-0.8 ms while the latter keeps decreasing drastically with the Al₂O₃ thickness decreasing (higher doping concentration), from 110 μs for the 0.1 at% device to less than 1 μs for the 4.5 at% one. Apparently the fast non-radiative decay of ⁵D₃ level is prominent and it leads to a higher average population of ⁵D₄ level as the Al₂O₃ sublayer thickness decreases. These phenomena verify the cross-relaxation from ⁵D₃ to ⁵D₄ levels when the distance between adjacent Tb layers is short enough to trigger this process among Tb³⁺ ions, depopulating the excited ions in ⁵D₃ level and contributing to the green EL from the ⁵D₄ level, especially in the 0.5-3 nm region. The prominent decrease of the green EL intensity when the Al₂O₃ sublayer is thinner than 1 nm should be ascribed to the concentration quenching that diminishing the luminescent Tb³⁺ ions. Similar effect has been previously reported [10,14]. Since the energy difference between these levels approximately equals that between the ⁷F₀ and ⁷F₆ levels, energy can be exchanged between excited ions in ⁵D₃ state and excite another ion in ⁷F₆ ground state. The cross relaxation process depopulates excited ions from ⁵D₃ state to the ⁵D₄ state, giving rise to a quenching of the ⁵D₃ levels and a simultaneous enhancement for the excitation of the ⁵D₄ levels in the thickness range of Al₂O₃ from 0.5 nm to 3 nm, leading to the different tendency of EL intensity between the green and blue emissions. The decay time of the 437 nm emission continues to decline while for the 543 nm emission its decay time changes slightly, thus the ions in higher ⁵D₃ level is more susceptible to the concentration quenching effect. This is also in agreement with the EL performance of the first series of devices shown in Figs. 1-3.

In comparison with the aforementioned MOSLEDs, this series of devices have the same Tb dopant cycles with the Al₂O₃ sublayer changed, and the Tb clustering should be not prominent. The cross-relaxation process occurs when the distance between adjacent Tb layers is short enough to trigger the interaction between excited Tb³⁺ ions. As the contribution of a single Tb₂O₃ dopant layer on green EL increases significantly with the thickness of Al₂O₃ ranging from 0.15 nm to 3 nm due to the cross-relaxation, and tends to saturate afterwards. Considering the possible migration of Tb³⁺ ions in annealing, this is consistent with the reported value of the critical transfer distance for the coupling RE ions [14,20]. In addition, as the thickness of Al₂O₃ decrease, there is no sufficient acceleration distance which reduces the average energy of the hot electrons under electric field, resulting in the weaker EL emissions. When the Al₂O₃ sublayer increases to more than 3 nm, the influence of cross relaxation among adjacent Tb³⁺ ions gradually declines. Moreover, sufficient acceleration under the electric field and meanwhile phonon scatterings within Al₂O₃ lattice enable the injected hot electrons to obtain the equilibrium kinetic energy, for which reasons the average EL intensity of a single Tb₂O₃ dopant layer reaches to a saturation value. From the experimental view, the distance for the prominent impact on the green EL due to cross-relaxation in Tb³⁺ ions, or the sufficient distance for electron acceleration, is around 3 nm in Al₂O₃ matrix. It is previously reported that to get an equilibrium distribution with enough hot electrons in the case of Tb³⁺ in SiO₂, high electric fields and a minimum acceleration distance of 10-20 nm are needed [21,22]. By comparison, Al₂O₃ is superior to SiO₂ as the matrix for Tb³⁺ ions in concern of the downscaling potential. The above-mentioned results also show that ALD supplies a convenient way to modify the dopant structure in the RE-doped nanolaminates. Excess atomic-layer of doped RE is inadvisable due to the clustering and consequent quenching effect. For the sublayer thickness of matrix, it should be sufficient to suppress the cross-relaxation counting against the intended emission and gain sufficient acceleration for the injected electrons.

The dependence of the EL intensity on the current density can be simulated by the solution of rate equation. At the steady state, the relationship between the EL intensity and the charge flux can be deduced to: I_EL = I_max∙[(J⁄e)/(J⁄e + 1⁄στ)], where I_max is the saturated EL intensity, σ is the cross section for excitation, and τ is the luminescence decay time accounting for both radiative and non-radiative decays [23]. The product of στ can be evaluated by fitting the dependence of EL intensity versus J/e. The decay times of the 543 nm emissions of different devices are used for calculation. The excitation cross sections of the ⁵D₄ level as a function of the Tb concentration including the two series of devices are illustrated in Fig. 5(a). In the first series, the excitation cross section keeps decreasing with the increase of Tb concentration, generally consistent with the tendency of EL intensity shown in Figs. 2(a)-2(b). This is mainly ascribed to the Tb-clustering, causing non-radiative de-excitation and the reduction of the luminescent Tb³⁺ ions at higher doping concentration. In the second series, the excitation cross section increases initially and then drop with the decrease of the thickness of Al₂O₃ sublayer. The increase again be interpreted by the cross-relaxation from the ⁵D₃ to ⁵D₄ level, proved by the sharp decline of the decay time of the 437 nm emission. With further increase of the Tb dopant, the increasing of interaction between adjacent Tb³⁺ ions serves as the main limitation and the excitation cross section declines. The device containing 0.5 at% Tb₂O₃ (1.49 nm Al₂O₃ sublayer) has the highest excitation cross section of 1.48 × 10⁻¹⁵ cm². Moreover, even though the two series of devices have somewhat different luminescent layer thickness, they present a similar tendency in the Tb concentration scale, and the calculated values are close, validating the proposed assumption and calculation.

 

Fig. 5 (a) The excitation cross sections of the EL from ⁵D₄ level as a function of the Tb concentration. (b) The dependence of external quantum efficiency and power efficiency on the current density for MOSLEDs based on different Al₂O₃/Tb₂O₃ films.

Download Full Size | PDF

In order to optimize the EL efficiency from the MOSLEDs based on Al₂O₃/Tb₂O₃ films, the thickness of luminescent layer is evaluated, together with the effect of cladding Al₂O₃ layer. By increasing Al₂O₃/Tb₂O₃ layer thickness, the EL threshold voltage increases slightly, while the injection current is strongly suppressed by several orders of magnitude. As the maximum of EL intensity is not apparently affected, a thicker Al₂O₃/Tb₂O₃ luminescent layer is more preferable in concern of the external efficiency. Moreover, the addition of the cladding layer of Al₂O₃ further enhances the abovementioned effect on the I-V characteristics, at the expense of lower EL intensity, which should be ascribed to the limited number of hot electrons to excite Tb³⁺ ions, as the current density can’t reach to a relatively high value beneath the breakdown voltage.

The power efficiency and external quantum efficiency (EQE) versus the current density are compared in Fig. 5(b). The Tb doping concentrations in the luminescent layers are around the optimal value of 2.5 at%. The EL efficiencies increase with the current densities to a maximum, and then falls down at higher current densities. The device with 70 nm Al₂O₃/Tb₂O₃ layer presents an EQE of 0.73%, and a power efficiency of 2.4 × 10⁻⁴ at ~20 mA cm⁻². The obtained EQE and optical power density are about five times higher than that reported in Ref. 9, while the relatively lower power efficiency compared with SiO₂:Tb can be mainly ascribed to the higher defect and trap density for which reason the charge transport is dominated by a P-F injection mechanism. Part of electrons are not effective for the Tb³⁺ excitations as they lost kinetic energy due to the impaction with defects and traps, which also contribute to the non-radiative de-excitations [14,22]. This is verified by the lower excitation cross section. However, it should be mentioned that the working voltages of these Al₂O₃/Tb₂O₃ MOSLEDs are 50-80 V, much lower than that of 150-220 V reported in Ref. 10. The efficiencies of the MOSLED with 50 nm Al₂O₃/Tb₂O₃ layer are 5-6 times lower than the aforementioned 70 nm one. While the MOSLED with 100 nm Al₂O₃/Tb₂O₃ in middle of 60 nm Al₂O₃ layer presents a maximum EQE of 4.6% and a power efficiency of 7.1 × 10⁻⁴ at ~3.3 mA cm⁻². Therefore, high external efficiencies can be achieved through further optimization on not only the thickness of the Al₂O₃/Tb₂O₃ luminescent layer but also the cladding layer.

4. Conclusion

In summary, intense green EL is obtained from the MOSLEDs based on nanolaminate Al₂O₃/Tb₂O₃ films fabricated by ALD. The main green peaks originate from the intra-4f ⁵D₄→⁷F_J transitions of Tb³⁺ ions. The green EL intensity benefits from the weaker blue ones because the cross-relaxation from the higher excited levels to the lower ones of Tb³⁺ ions relieves the concentration quenching of the green emissions, resulting in the optimal doping ratio of around 2.5 at% Tb. The EL is attributed to the impact excitation of the Tb³⁺ ions by hot electrons via P-F mechanism. The distance for the prominent impact on the green EL due to cross-relaxation among Tb³⁺ ions and the sufficient acceleration of hot electrons in the conduction band, is determined to be ~3 nm in Al₂O₃. The excitation cross section for the 543 nm Tb³⁺ emission is up to 1.48 × 10⁻¹⁵ cm², while the power density reaches to 3.37 mW cm⁻². An EQE of 0.73% and a power efficiency of 2.4 × 10⁻⁴ are obtained. Further optimization is achieved by adopting a thicker Al₂O₃/Tb₂O₃ luminescent layer and Al₂O₃ cladding layer, resulting in maximum EQE of 4.6% and a power efficiency of 7.1 × 10⁻⁴. These results present that ALD supplies a viable approach to modify the dopant structure in the RE-doped nanolaminates to achieve efficient emissions.