1. Introduction

Wide band gap (3.37 eV) and high exciton binding energy (60 meV) make ZnO a promising candidate for ultraviolet (UV) light-emitting diodes (LEDs) and low-threshold lasing diodes (LDs) [1–3]. However, the difficulty in producing high-quality p-type ZnO with reliable stability and reproducibility remains the greatest challenge for the development of homojunction LEDs/LDs. An alternative strategy for achieving ZnO-based UV electroluminescence (EL) is to fabricate p-n heterojunction devices by employing other available p-type materials, such as p-GaN, p-Cu₂O, and p-type organics etc [3–5]. However, these p-n heterojunctions also suffer from some problems, for example, the EL component from p-type material will contribute to (or even dominate) the total LED spectra, and thus the advantage of ZnO excitonic emission cannot be fully utilized. To avoid this problem, another type of device structure, named “metal/insulator/semiconductor” (MIS) heterostructure, was popularly adopted to obtain relatively pure ZnO UV EL [6–9]. For instance, Zhu et al. has observed electrically pumped near-UV excitonic emission and low-threshold lasing from the Au/MgO/ZnO film heterostructure [6]. With similar device configuration, our group has also demonstrated ZnO/MgO core/shell nanowire LDs with lowered threshold current density and improved emission efficiency [7]. The EL mechanism for the MIS-type LEDs can be well understood based on the energy band alignment depicted in Fig. 1(a) . Due to the large conduction band offset between ZnO and MgO, electrons would be blocked and accumulated at the heterojunction interface under forward bias. Most applied voltages would drop on the MgO layer considering its dielectric nature, and the local electric field strength could be as high as 10⁷-10⁸ V/m therein [6]. Thus, electrons and holes can be generated through a so-called “impact-ionization process” in the insulating MgO layer. The generated holes would be driven into ZnO under forward bias and radiatively recombine with the electrons accumulated at ZnO/MgO interface, giving rise to the UV emission of ZnO. However, such an EL mechanism also seriously limits the LED external quantum efficiency (EQE), because the production of holes via “impact ionization process” is not very efficient, and the relatively high bias voltage may increase undesired heating effect [6,10]. Thus, how to improve the device efficiency is a key issue for the future development of ZnO-based MIS heterojunction LEDs.

 

Fig. 1 (a) Energy-band alignment of Au/MgO/ZnO/n-Si heterostructure under forward bias. (b) Schematic diagram of LSP-enhanced Au/Ag-NPs@MgO/ZnO MIS heterojunction LED.

Download Full Size | PDF

Recently, localized surface plasmons (LSPs) have been proved to be an effective method for improving the efficiency of light-emitting materials and devices [11–14]. When the energy of excitons/photons in a semiconductor is similar to the electron vibrational energy of metal LSPs, the energy will be coupled into the LSPs and finally scattered into free space as radiation. The formation of such additional recombination/extraction paths can improve the internal quantum efficiency (IQE) of LEDs, as well as their light extraction efficiency (LEE). Based on this method, several groups have demonstrated noticeable EL improvement in ZnO-based p-n junction LEDs via Ag nanoparticles (Ag-NPs) decoration [15–20]. Nonetheless, available reports on LSP-enhanced MIS junction LEDs are still very limited [21]. Moreover, the luminescence enhancement mechanism discussed before is mainly focused on the IQE improvement of semiconductor active region induced by exciton-LSP coupling, while the contribution from LEE improvement due to the photon-LSP coupling has not yet been paid enough attention.

In this work, the LSP-enhanced UV LEDs were demonstrated based on a Au/Ag-NPs@MgO/ZnO MIS heterostructure, where the Ag-NPs were embedded in the MgO insulating layer. A ~6-fold EL enhancement was achieved from the Ag-NPs decorated device. Time-resolved photoluminescence (TR-PL) measurements, as well as finite-difference time-domain (FDTD) analogue simulation and theoretical estimation based on experimental data, reveal that the IQE and LEE of the LED device are increased as a consequence of the introduction of Ag LSPs, indicating that both the exciton-LSP coupling and photon-LSP coupling are responsible for the observed luminescence enhancement.

2. Experiments

Figure 1(b) shows a structural diagram of our LSP-enhanced MIS junction LED. In view of integration with existing silicon-based electronics, the fabrication of LEDs on silicon wafers is of great importance. Therefore, nominally undoped ZnO films (electron concentration: 4.20 × 10¹⁶ cm⁻³, mobility: 7.50 cm²/Vs) with a thickness of ~200 nm were first grown on commercially available n-Si (100) substrates (electron concentration: 1.1 × 10¹⁸ cm⁻³, mobility: 3.47 × 10² cm²/Vs) by pulsed laser deposition (PLD). The O₂ pressure and substrate temperature were kept at 20 Pa and 600 °C during deposition, respectively. The low-resistivity n-Si wafers can serve as not only an electron injection layer for the ZnO active layer (see Fig. 1(a)), but also as a bottom contact electrode and current spreading layer to enhance the uniformity of electron injection [22,23]. Then, MgO layers with almost the same growth conditions, were subsequently deposited onto the ZnO films by PLD. Here, the high-quality MgO dielectric layers play two roles. One is the insulating layer in MIS heterostructure, responsible for the electron blocking and hole generation/injection under forward bias. This role requires the MgO layers have a certain thickness. Otherwise, the dielectric layer can be easily broken down and the electrons cannot be blocked. According to the previous findings of Chen et al. and our group, the MgO thickness in the MIS heterojunction device should be more than 25 nm [24,25]. Another role is a spacer layer between Ag-NPs and ZnO film, suppressing the undesired charge transfer and nonradiative Förster resonant energy transfer (FRET) between them, instead favoring a LSP-induced field enhancement effect. The coupling with LSPs is known as a near-field interaction, and the LSP evanescent-field intensity exponentially decays with distance from the metal NP’s surface. Here, the penetration depth (Z) of the Ag-LSP evanescent field into MgO is given by${\rm Z}=\lambda /2\pi \sqrt{\left(\epsilon {\text{'}}_d-\epsilon {\text{'}}_m\right)/\epsilon {\text{'}}_d{}^2}$, where $\epsilon {\text{'}}_d$ and $\epsilon {\text{'}}_m$ are the real part of the dielectric constants of the MgO (2.9) and Ag (−6.5), and $\lambda$ is the wavelength of incident light [11]. The calculation yields ${\rm Z}=64$ nm for the ZnO UV emission wavelength of ~380 nm. Thus, MgO spacer layer thickness should be controlled within 60 nm in our experiments. Here, MgO insulating/spacer layer with different thicknesses were prepared to evaluate the LSP-induced luminescence enhancement effect, and this will be discussed in detail in Figs. 2(c) and 2(d). Next, ~20 nm Ag-NPs with a surface density of ~6.6 × 10¹⁰ cm⁻² (as shown in Fig. 2(a)) were deposited onto the surface of MgO films by a physical sputtering method under an optimized deposition condition. The size, surface density, and surface coverage of Ag-NPs can be well controlled by adjusting the sputtering current and pressure, as has been detailedly discussed in our previous work [20]. So far, the Ag-NPs/MgO/ZnO composite films have been constructed.

 

Fig. 2 (a) Scanning electron microscope image of the Ag-NPs layer. (b) The extinction spectrum of Ag-NPs coated on MgO film (blue solid line) and a typical PL spectrum of ZnO film (red dash-dot line). (c) PL spectra of bare ZnO film, Ag/ZnO film and Ag-NPs/MgO/ZnO composite films with different MgO spacer layer thicknesses. (d) The variations of UV emission enhancement ratio with the MgO thickness. The 325 nm line of a He-Cd laser was employed as the excitation source.

Download Full Size | PDF

3. Results and discussions

Figure 2(b) gives the LSP resonance extinction spectrum of Ag-NPs coated on MgO film, which is centered at a wavelength of ~425 nm. The broad distribution of this extinction band may result from the size fluctuation of Ag-NPs and the coupling between neighboring LSPs [26]. An obvious spectral overlap is observed between the ZnO UV luminescence and Ag LSP resonance extinction, suggesting the possibility of a resonant coupling between ZnO excitons/photons and Ag LSPs. Thus, the enhanced ZnO UV emission would be expected. Figure 2(c) presents the room-temperature PL spectra of the Ag-NPs/MgO/ZnO composite films with different MgO spacer thicknesses, where all spectra were recorded under the same experimental conditions. One can see that all spectra consist of a near-band-edge (NBE) emission at ~380 nm and a relatively weak and broad deep-level (DL) emission centered at ~550 nm [27]. The ZnO UV PL is slightly enhanced when the Ag-NPs are directly deposited on the ZnO film. As mentioned above, the introduction of MgO spacer layer can suppress the undesired charge transfer and nonradiative FRET processes, and thus results in further PL enhancement. With increasing MgO spacer layer thickness, the ZnO UV emission gradually increases. A maximum ~6-fold enhancement of the PL intensity is obtained at the MgO thickness of ~30 nm. Further increase in its thickness only leads to the reduction of PL enhancement. Such a variation trend is also illustrated in Fig. 2(d). Therefore, we choose the MgO thin film with optimized thickness of ~30 nm as the insulating/spacer layer in the following LED fabrication procedure. In contrast, no obvious change was observed in the visible DL emission region.

To construct the MIS junction LED, another MgO film (~20 nm), serving as an anti-oxidation capping layer for Ag-NPs, was deposited onto the former Ag-NPs/MgO (30 nm)/ZnO composite film at room-temperature and pure argon ambient (That is, the Ag-NPs were embedded into the MgO layer, see Fig. 1(b)). Then, very thin Au electrodes were thermally evaporated on the topside, and patterned into 1 mm circular pads by shadow mask. The EL signal can be collected from these semi-transparent Au electrodes and their edges. As a comparison, a reference device without embedded Ag-NPs was also fabricated in parallel. Typical rectifying diode behaviors (Fig. 3(a) ) with slightly different turn-on voltages of 4.5 and 4.9 V are observed from both LEDs with and without Ag-NPs, respectively. As mentioned at the beginning, the EL mechanism of the MIS junction LED is based on the “impact ionized process”, in which electrons and holes are generated in the MgO insulating layer under high local electric field, and the ionized holes are driven into the ZnO active layer under forward bias. This is a typical phenomenon of space-charge-limited current conduction, which can be well described by the Mott-Gurney law of J∝(V-V₀)², as shown in Fig. 3(b). Herein, J and V are the injection current density (ICD) and applied bias voltage, respectively, and V₀ is a fitting parameter. The good matching between the experimental J-V data and fitting curve, in turn, confirms that the EL behavior and carrier transportation of MIS junction LED are indeed based on such a single carrier injection mechanism [28]. It is noteworthy that, compared with the LED without Ag-NPs, the Ag-NPs decorated device has larger forward and reverse ICD. Similar phenomena have also been observed in some Ag-NPs-inserted photonic and electronic devices, such as Si-quantum-dots LED and resistive-switching memory, and were ascribed to the enhanced conductivity of dielectric matrixes induced by the strong local electric field around embedded metal NPs [14,29].

 

Fig. 3 (a) J-V curves of the MIS junction LEDs with (blue line) and without (red line) Ag-NPs. (b) A log-log plot of the J-V characteristic of the pristine LED under forward bias, which illustrates the single carrier injection process of space-charge-limited current model.

Download Full Size | PDF

Under forward bias, the NBE excitonic emission peaked at ~385 nm was obtained from the LEDs with and without Ag-NPs, while no EL signal was detected in the visible region. The spectra of both devices were recorded under the same ICD in order to allow comparison. As shown in Fig. 4(a) , at a relatively high ICD of 44 A/cm², the device without Ag-NPs just shows a weak EL signal with very low signal-to-noise ratio. This phenomenon further indicates the extremely low EQE, as well as poor power efficiency, of pristine MIS junction LEDs. In contrast, for the Ag-NPs decorated device, obvious EL can be detected even at a lower ICD, and its EL intensity is ~6 times as large as that of the LED without Ag-NPs (see Fig. 4(a)) when the ICD is 44 A/cm². That is, a ~6-fold EL improvement is achieved from the LSP-enhanced MIS junction LED. As the ICD further increases, the EL intensities of both devices increase (Fig. 4(b)), but the EL enhancement ratio decreases rapidly and ends at ~2 as the ICD approaches 90 A/cm² (Fig. 4(c)). This phenomenon can be understood in terms of a screening effect of excess carriers under a high current injection, which weakens the exciton-LSP coupling [13]. In fact, the “residual” ~2-fold EL enhancement exactly reflects the LEE improvement of Ag-NPs decorated LED, and will be discussed in detail later.

 

Fig. 4 (a) EL spectra of the LEDs with (blue line) and without (red line) Ag-NPs under the same ICD of 44 A/cm². (b) The variations of integrated UV EL intensity with the ICD for the devices with and without Ag-NPs. (c) EL enhancement ratio of the two LEDs as a function of the ICD.

Download Full Size | PDF

As shown in the previous text, both PL and EL from Ag-NPs decorated LED show a ~6-fold improvement, suggesting that they have the same luminescence enhancement mechanism: resonant coupling between ZnO excitons and Ag LSPs. To support this argument, two kinds of verification methods, including analogue simulation and experimental measurement, were adopted. One is a 2D FDTD simulation of the electrical field distribution in LSP-enhanced LED. The simulation structure is constructed as follows: 1) a 50 nm thick MgO layer (30 nm insulating layer + 20 nm capping layer) is placed on top of a 200 nm thick ZnO layer; 2) a dipole light source is placed at the interface between the MgO and ZnO layers, corresponding to the active region for exciton recombination; 3) a 20 nm silver sphere, embedded in the MgO layer, is put right above the dipole source. The space between the silver sphere and the ZnO layer is 30 nm. Such a structure is consistent with the experimental device configuration in this work. As can be seen in Fig. 5(a) , the energy from the dipole light source can be well coupled into the silver sphere, confirming the existence of resonant coupling interaction between ZnO exciton and Ag LSP. The other is the room-temperature TR-PL measurements, which were conducted using a streak camera (temporal resolution: 2 ps) as the detector and a femtosecond pulsed laser (wavelength: 266 nm, repetition rate: 76 MHz, pulse width: 200 fs) as the excitation source. Figure 5(b) presents the TR-PL spectra of the LEDs with and without Ag-NPs, both of which can be well fitted by a bi-exponential attenuation function: $I=A_1\exp (-t/{\tau }_1)+A_2\exp (-t/{\tau }_2)+I_0$, where the I and I₀ are the PL intensity, the A₁ and A₂ are the weight factors of two decay processes, and the ${\tau }_1$ and ${\tau }_2$ represent the corresponding radiative recombination lifetime, respectively. The effective PL lifetime (${\tau }_{eff}$) can be obtained by a weighted average of two time constants (${\tau }_1$ and ${\tau }_2$) by the following formula: ${\tau }_{eff}=(A_1{\tau }_1{}^2+A_2{\tau }_2{}^2)/(A_1{\tau }_1+A_2{\tau }_2)$. The fitting results of the above parameters for the devices with and without Ag-NPs are listed in the table of Fig. 5(b). In the following discussion, all the physical quantities with and without asterisks *, respectively, correspond to the Ag-NPs decorated and undecorated samples. As shown in Fig. 5(b), the shortened ${\tau }_{eff}^{\ast }$ of 7 ps reveals that the sample with Ag-NPs decoration shows a faster spontaneous emission rate than the one without Ag-NPs (${\tau }_{eff}=25ps$). In general, the lifetime of ZnO excitons coupled with and without Ag LSPs (${\tau }_{eff}^{\ast }$ and ${\tau }_{eff}$) can be expressed as [11]:

 

Fig. 5 (a) FDTD simulation of the electrical field distribution in the LSP-enhanced LED. (b) Room-temperature TR-PL spectra of MgO/ZnO composite films with (blue circle) and without (red square) Ag-NPs; the monitoring wavelength is 380 nm, and the solid lines are the fits to a bi-exponential decay model. The inset table gives the lifetimes of fast and slow components, as well as the effective lifetimes. (c) Arrhenius plots of the normalized integrated UV-PL intensity of MgO/ZnO composite films with (blue circle) and without (red square) Ag-NPs.

Download Full Size | PDF

The increase in recombination rate can result in the improvement of IQE. In theory, the IQE for the LEDs with and without Ag-NPs (${\eta }_{IQE}^{\ast }$ and ${\eta }_{IQE}$) can be written as [11]:

 

Fig. 6 Schematic diagram showing the EL enhancement mechanism of Au/Ag-NPs@MgO/ZnO MIS heterojunction LEDs. The $k_{rad}$ and $k_{non}$ represent the radiative and nonradiative recombination processes, respectively.

Download Full Size | PDF

Finally, let us consider the ultrafast dynamic process involved in the TR-PL spectra. As mentioned above, the PL decay curve of ZnO film without Ag-NPs can be well fitted by the bi-exponential attenuation function, yielding a fast (${\tau }_1=10ps$) and a slow (${\tau }_2=60ps$) decay process. The slow component is attributed to the ZnO exciton recombination, which usually exhibits radiative lifetime of several tens to hundreds of picoseconds in ZnO films and nanostructures [34,35]. The other fast component may be associated with the formation and recombination of ZnO polaritons, which has already been detailedly discussed in our previous work [20]. Briefly speaking, strong couplings between excitons and photons confined in a microcavity can produce a novel hybrid quasiparticle-polariton, and this phenomena has been observed in ZnO nanorods and some III-V group semiconductors (e.g. GaAs and GaN) [36,37]. The ZnO polaritons may be present in our LED device because random microcavity structures can be easily formed in the textured polycrystalline ZnO film through a closed-loop multi-scattering process [38,39]. As effective luminescence centers, polaritons possess two main characters: 1) they can emit photons with similar energy to the exciton recombination; 2) the PL lifetime of polaritons is usually very short (a few ps) due to a fast photon leakage rate from the microcavity [40]. Thus, the relatively fast decay component (${\tau }_1=10ps$) is attributed to the ZnO polariton recombination, and the coupling between ZnO polaritons and Ag LSPs can also contribute to the observed luminescence enhancement. It seems that when the Ag-NPs are introduced, the radiative recombination of ZnO polaritons dominates the luminescence process because the time constants became the same for the two components, and the related reason needs further studies.

4. Conclusions

In summary, we have demonstrated the UV LED prototype devices based on Au/MgO/ZnO MIS heterostructures. A ~6-fold UV EL improvement is achieved with the introduction of Ag-NPs. Based on the analyses of FDTD simulation, TR-PL spectra and theoretical estimations, it is concluded that the observed luminescence enhancement results from the simultaneous increment of both IQE and LEE induced by exciton-LSP coupling and photon-LSP coupling, respectively. The ultrafast dynamic process found in the TR-PL spectra was also discussed in terms of interaction among exciton, polariton and LSP. The LSP-enhanced LED shows higher EL intensity at relatively low injection current, which implies its potential applications in lighting system with relatively small input current such as backlight display, indoor illumination, and automobile taillight. Though further optimization of the device structure and performance is still needed, the work presented here provides a feasible way to improve LED efficiency, and is important for the future development of high-efficiency, low threshold MIS-type LEDs/LDs.