1. Introduction

In the last decades, silicon-rich silicon nitride (SiN_x, x<1.33) films have attracted a great research interest in a wild range of fields, including the memory medium in nonvolatile metal-insulator-semiconductor (MIS) devices [1], passivation layers and antireflection coatings for current Si solar cells [2, 3], optical waveguides [4], and light-emitting devices (LEDs) [5–7]. For the application on LEDs as a candidate of Si-based light sources, enormous efforts have been devoted to SiN_x films fabricated by plasma-enhanced chemical vapor deposition (PECVD) technique due to its conspicuous and tunable room-temperature photoluminescence (PL) with a promising electroluminescence (EL) efficiency [5, 8, 9]. Besides the SiN_x matrix, recent progress had also been focused on the Si quantum dots (QDs) based LEDs [10–12] and the Ge-on-Si emitters with the band of Ge engineered by tensile strain and n-type doping [13, 14] as the alternative Si-based light sources. Generally, the formation of Si QDs should experience a high-temperature thermal annealing procedure (usually >1000 °C) [11, 12], and the Ge-on-Si lasers would suffer an extremely high threshold current density and poor efficiency [14]. Moreover, compared with its congener silicon-rich silicon oxide (SiO_x) which is also commonly used for Si-based light sources [15], SiN_x provides a lower carrier injection barrier that may realize a higher external quantum efficiency (EQE) [16]. Meanwhile, the fabrication of SiN_x-based LEDs was a complementary metal-oxide semiconductor (CMOS) compatible process. By employing a SiN_x/SiN_y multilayer structure, Huh et al. achieved an EQE of 10⁻³% for SiN_x-based LEDs [17]. This value is still too low to achieve the demand of Si-based light sources for the application in optical interconnections. The partial reason for this insufficient EL efficiency may result from its unbalanced carrier injection barrier heights. The injection barrier height for electrons is ~3.0 eV (from ITO electrode to SiN_x matrix), which is much higher than that for holes on the p-type Si side (1.9 eV) [5]. Besides, the strong nonradiative recombination derived from the band tails and defect states in SiN_x due to the structural disorder also constrains its EL efficiency [5].

The inherent problems mentioned above in SiN_x films prevent it from being used in actual applications as Si-based light sources. Consequently, various methods were proposed to improve the EL performance of SiN_x-based LEDs in recent years. For the purpose of improving the current injection, Kim et al. used the Ni/Au contact placed on SiN_x layer, obtaining an EQE of (~3.3 x 10⁻³)% [18]. Meanwhile, by employing the n-type SiC layer to balance the carrier injection levels accompanied with a periodic micron-scale rugged surface pattern produced by optical lithographic and a standard Si etching technique to increase the light extraction from SiN_x film, Kim et al. achieved a 2.8 times enhancement of the light output power [19]. On the other hand, the nonradiative recombination can be decreased significantly by the reduction of SiN_x/Si interfacial state density or the passivation of nonradiative defects in the SiN_x films using NH₃ plasma pretreatments, which can further increase the EL efficiency of SiN_x-based LEDs [20]. Besides, the EL efficiency can be enhanced by depositing Ag or Au nanostructures underneath the luminescence layer, which is originated from the coupling between the excitons and localized surface plasmons (LSPs, the collective oscillations of excited free electrons occurring at the interface between a metal and a dielectric layer, which is confined to the surrounding of metallic nanostructures) [7, 21].

Recently, the application of LSPs on LEDs has been received a striking attention not only for the improvement of carrier injection resulted from the enhancement of the inhomogeneous local electric fields at the interface between metals and luminescence matrixes [22], but also for the increase of radiative efficiency originated from the strong coupling between exciton dipole moments and local electric field of LSPs [7]. The use of surface plasmon coupled to the excitons in active matrixes had been reported in (In)GaN-based LEDs for achieving the improved spontaneous emission rate from Purcell factor enhancement [23–25]. As have been investigated, this coupling is stronger at the frequency near its plasmon resonance affected by the size, shape, and dielectric environment of metal particles [21]. Among all the metals with plasmon resonances at visible frequencies, which are near the wavelength of EL peak for SiN_x-based LEDs, Ag provides the lowest absorption losses [26] and the strongest resonances [27].

In our previous work, a Ag nanostructures layer was deposited onto or underneath the SiN_x matrix to improve the light extraction efficiency (LEE) via the roughening of the ITO electrode, and an almost one order of magnitude enhancement of EQE for SiN_x-based LEDs was achieved [28, 29]. Besides, we found that the improvement of the LEE via embedding the Ag nanostructures underneath the luminescence layer was not only originated from the surface roughening of the front electrode, but also originated from the increased back-scattering due to the LSP resonance coupling [30]. Moreover, we fabricated the Ag nanostructures with various sizes and surface coverage onto the luminescence matrix by the modulation of the thermal annealing procedure, and obtained that the density of Ag nanostructures has a significant influence on the improvement of the PL intensity for the SiN_x films with different N/Si ratios [31]. Furthermore, the coupling between LSPs and excitons as well as the determination of the average position of excitons in SiN_x matrix was demonstrated via the Purcell effect [32]. Further investigations of the other effects of LSP resonance on the improvement of the EQE and/or the evolution of EL from SiN_x-based LEDs are also in progress. In this paper, we demonstrate the optimization process of the EL efficiency from SiN_x-based LEDs via depositing a Ag nanostructures layer with different dimensions and morphologies onto the SiN_x matrix by employing a simple and COMS compatible fabrication process. The separation analysis of the three contributions of the LEE, carrier injection efficiency, and radiative efficiency enhancement on the improvement of the EL efficiency is provided. The metal nanostructures with highly-uniform close-packed two-dimensional arrays or with highly ordered patterns can be fabricated by using the diblock-copolymer lithography templates [33, 34] or using a rapid convective deposition method [35–37]. In this work, the Ag nanostructures layer was formed by rapid thermal annealing (RTA) the sputtered Ag film, and the dimension distribution of Ag nanostructures is relative random and large.

2. Experimental

Boron-doped p/p⁺-Si (100) wafer (with the thickness and resistivity of the epitaxial layer were 18.5 ± 1 µm and 9 ± 1.5 Ω·cm, respectively) was employed as a substrate for the fabrication of SiN_x-based LED. A PECVD system was used to deposit SiN_x films, in which N₂-diluted 10% SiH₄ and NH₃ were used as the reactant gas sources. The flow rate ratio of SiH₄ and NH₃, substrate temperature, chamber pressure, and r.f. power density for the deposition were maintained at 1:1, 400 °C, 0.1 Torr, and ~1 kW/m², respectively. After the deposition of SiN_x films, the samples were annealed at 800 °C for 1 hour in N₂ ambience. To investigate the effect of Ag nanostructures on the EL efficiency of SiN_x-based LEDs, a Ag layer was deposited by magnetron sputtering onto the SiN_x films. The thickness of Ag was regulated by the sputtering time, which was also used to label the samples, e.g. Ag20 refers to the sample with the sputtering time of 20 s. After the post-treatment of rapid thermal annealing (RTA) in argon at 500 °C for 1 min, Ag nanostructures with various dimensions and surface morphologies were formed. For the investigation of the electrical properties of SiN_x-based LEDs with and without Ag nanostructures, a top transparent ITO circular electrode (~100 nm) with the diameter of 1.0 cm was deposited by magnetron sputtering using a circular mask and an Al metal contact of about 200 nm was evaporated on the backside of the Si substrates. The fabrication process and the final device structures of the SiN_x-based LEDs are shown in Fig. 1. A SiN_x-based LED without the addition of Ag nanostructures was also fabricated for comparison, which was labeled as Ag0. Meanwhile, we also fabricated the quartz/SiN_x/Ag structures for the measurement of the extinction spectra of the Ag nanostructures with different sizes and morphologies.

 

Fig. 1 Fabrication process diagrams of the SiN_x-based LEDs.

Download Full Size | PDF

The thickness of SiN_x films was around 50 nm measured by ellipsometry, J. A. Woollam M-2000D. The scanning electron microscopy (SEM, Hitachi S-4800) was used to characterize the Ag nanostructure size and the surface morphology. A Veeco Digital Instruments Innova atomic force microscopy (AFM) was employed for the characterization of the surface morphology of the ITO electrode with and without the addition of Ag nanostructures. Both the photoluminescence (PL) and EL signals of SiN_x-based LEDs were recorded by an Acton SpectraPro-2500i monochrometer coupled to a photomultiplier tube (PMT). The PL signals were excited by a 325 nm He-Cd laser, while the EL ones were driven by a DC power source. The current-voltage (I-V) characteristics of the devices were measured by a Keithley 4200 SCS semiconductor parameter analyzer. A HITACHI U-4100 spectrophotometer was employed to measure the extinction spectra of the SiN_x films with and without Ag nanostructures.

3. Results and discussion

Figure 2 shows the SEM images and the size distributions of Ag nanostructures formed by the post-treatment of RTA with the sputtering time ranging from 20 s to 80 s. The Ag nanostructures reveal a good dispersion with the average dimension (D) from ~40 nm to ~240 nm in diameter for the sputtering time from 20 s to 80 s. The aggregation of particles becomes more and more serious with the increase of the average sizes. The primary shapes are circles with a little trend of being ellipses, which is gradually noticeable as the average diameter increases. Moreover, with the increase of the average diameter of Ag nanostructures, the dipolar resonance peak in extinction spectra gradually red-shifts, as shown in Fig. 3. The intensity of the extinction spectra for Ag nanostructures is significantly larger than that for the SiN_x films [28], and the relative intensity of the dipolar resonance peak is gradually decreased with the increasing size of Ag nanostructures (not shown here). An obvious extinction valley can be observed at short wavelength, which also gradually red-shifts with the increasing size of Ag nanostructures. Interestingly, both the extinction peak and valley for Ag80 blue-shift a little compared with that for Ag60, which may result from the Ag nanostructures with smaller sizes due to the nonuniformed distribution of the dimensions for Ag80 or from the minor axis of ellipsoidal-like Ag nanostructures, as shown in Fig. 2(d). The quadrupolar resonance peak can be observed clearly in the extinction spectra of Ag80, as shown in Fig. 3.

 

Fig. 2 (a)-(d) SEM images and (e)-(h) the size distributions of Ag nanostructures with sputtering time (a) and (e) 20 s, (b) and (f) 40 s, (c) and (g) 60 s, and (d) and (h) 80 s.

Download Full Size | PDF

 

Fig. 3 Extinction spectra of Ag nanostructures with different sputtering time.

Download Full Size | PDF

The effect of Ag nanostructures with different sizes on the EL efficiency of SiN_x-based LEDs was investigated at room temperature for further improving the luminescence performance of the devices. As can be observed in Figs. 4(c)-4(f), two EL peaks can be resolved for the devices with the addition of Ag nanostructures, with the longer wavelength peaks in the range of 620-780 nm and the shorter ones at 500-600 nm. While for the sample without the addition of Ag nanostructures (Ag0), only one shorter wavelength peak (~600 nm) can be observed, as shown in Figs. 4(a) and 4(b). Meanwhile, the peak with longer wavelength can be resolved hardly for the sample with the addition of Ag nanostructures of larger sizes (Ag80), as shown in Figs. 4(e) and 4(f). We speculate that this phenomenon may be due to its poor injection of electrons as the barrier height of electrons on ITO side (3.0 eV) is much higher than that of holes on p-Si side (1.9 eV) [5]. We attributed all these peaks to the recombination of electrons and holes at defect levels or tail states lie in SiN_x bandgap in the range from ~400 nm to ~780 nm [5, 38, 39]. The peak at the shorter wavelength (~500-600 nm) is due to the radiative recombination between the K center (≡Si^-) and the valence band tail levels composed by N dangling bands ( = N^-) and Si-Si bonding states, which are in the range of ~350-650 nm [38, 39]. Meanwhile, we attribute the one at the longer wavelength to the recombination of electrons at conduction band tail states and holes trapped in the center of ≡Si⁰ [38, 39]. The weakening of the longer wavelength EL peak, as shown in Fig. 4(f), may be originated from the insignificantly improved electron injection on ITO side, where only an extremely small amount of electrons can be injected into the conduction band and contribute to the EL.

 

Fig. 4 EL spectra of SiN_x-based LEDs for (a) Ag0, (c) Ag40, and (e) Ag80 measured at different injected current. Their Gauss fittings are also provided on the right side, (b) for Ag0, (d) for Ag40, and (f) for Ag80.

Download Full Size | PDF

For the investigation of the size and morphology modulation effect of Ag nanostructures on the improvement of EL efficiency, the integrated EL intensity of SiN_x-based LEDs with and without the addition of Ag nanostructures is compared, as shown in Fig. 5. Obviously, a maximal enhancement of EL efficiency of about 6.5 times is achieved for the device with the average size of Ag nanostructures ~100 nm in diameter by comparing the integrated EL intensity of Ag40 to that of Ag0 under the almost same input power. Three aspects including the increase of LEE, internal quantum efficiency (IQE), and carrier injection efficiency should be considered to determine the reasons for the optimal EL performance of Ag40.

 

Fig. 5 Integrated EL intensity of the devices with different sputtering times of Ag nanostructures layer vs. the input power.

Download Full Size | PDF

The improvements of IQE for Ag20-Ag80 can be estimated semiquantitatively by simply comparing the enhancement factor of integrated PL intensity [28], which is defined by the ratio of the integrated PL intensity of the devices with Ag nanostructures to that of Ag0, as shown in Fig. 6(a). A maximum improvement of ~1.8 times is achieved for Ag40. We attribute this improvement mainly to the enhancement of the spontaneous emission rate due to the coupling between the excitons in SiN_x and LSPs [7, 32, 40], which is confirmed by the consistence between the enhancement factor of EL intensities and that of PL ones, as shown in Fig. 6(b). The enhancement of the flow of light emitted from the active layer due to the minimum absorption of Ag nanostructures also has a contribution to this improvement, as shown in Fig. 2 and Fig. 6(b). Obviously, the improvement of IQE is not the main contribution to the enhancement of EL efficiency in our SiN_x-based LEDs by further considering the surface coverage of Ag nanostructures (η_s), as shown in Fig. 6(c) left. The serious backscattering induced by the high η_s (>40%) of Ag nanostructures will make it difficult for light to escape from the active layer. The contribution of this improvement to the EL efficiency will be more insignificantly by considering the absorption of Ag nanostructures further.

 

Fig. 6 (a) Enhancement factor of the integrated PL intensity for Ag20-Ag80 comparing to Ag0, (b) EL and PL enhancement factor for Ag40, and (c) The surface coverage of Ag nanostructures (left axis) and the average distance between Ag nanostructures (right axis) for Ag20-Ag80 with the model for this calculation inset.

Download Full Size | PDF

Interestingly, the same tendencies of the enhancement factor of integrated PL intensity and the η_s with the increase of D are observed, as shown in Figs. 6(a) and 6(c). The improvements of IQE for Ag60 and Ag80 are higher than that for Ag40 theoretically by simply comparing the enhancement factors of integrated PL intensity and the η_s. Considering the origin of the enhancement of IQE, the distance between Ag nanostructures should also be taken account of, which will affect the spontaneous emission rate significantly [21]. A simple model is adopted for the calculation of the average distance (d) between Ag nanostructures, as shown in the inset of Fig. 6(c). The value of d can be obtained from the formula:

The roughening of ITO electrode surface via the introduction of Ag nanostructures, as shown in Fig. 7, can decrease the inner-reflection of light and improve the LEE of SiN_x-based LEDs significantly [30, 41]. The surface of ITO electrode for the device Ag0 is rather smooth with the area root mean square (RMS) roughness (δ) only ~2.0 nm, as shown in Fig. 7(a). After the addition of Ag nanostructures, this value of δ increases monotonically with the increased D of Ag nanostructures. The contribution of this surface roughening on the improvement of the LEE for our LEDs can be estimated semiquantitatively based on the total integrated scatter (TIS), which is defined as the total power scattered into a hemisphere divided by the incident power [42]. This value of TIS is proportional to the square of δ under the assumption that the scattering mainly occurs at the regular reflection direction [42]. An obvious enhancement of LEE can be achieved by this roughening of ITO electrode surface even considering the surface coverage and the absorption of Ag nanostructures on the suppression of light extraction. The significant increase of LEE also confirms that the improvement of IQE is not the main contribution to the enhancement of EL efficiency of our SiN_x-based LEDs, as have been mentioned above. Besides, the increased δ of ITO electrode would also enhance the nonradiative recombination due to its increased interface state between SiN_x matrix and ITO. Consequently, there is a tradeoff between the improvement of LEE and the enhancement of nonradiative recombination via the roughening of ITO electrode surface, and it is difficult to differentiate them. Due to the extremely large δ of ITO electrode for the sample Ag80, the overall effect of this roughening on the improvement of EL efficiency might be limited. Better improvement may be achieved by the addition of Ag nanostructures with smaller size and suitable roughness. Hence, the maximum enhancement of LEE via the surface roughening of ITO electrode may also be achieved for Ag40 among our devices. Further improvement of EL efficiency can be achieved by the optimization of the surface roughness of ITO electrode.

 

Fig. 7 AFM images of ITO for (a) Ag0, (b) Ag40, and (c) Ag80.

Download Full Size | PDF

The effect of the improvement of carrier injection efficiency on the optimization of EL efficiency for our devices is also checked via J-V characteristics, as shown in Fig. 8. A significant enhancement of carrier injection for the devices Ag20 and Ag40 is achieved comparing to the device Ag0. While for the device Ag60 and Ag80, this enhancement is not so much obviously, as shown in Fig. 8. We attribute the improvement of the carrier injection to the increased surface roughness due to the addition of Ag nanostructures, by which a large enhancement of the localized electric field at the interface can be obtained [21]. Due to the rather short distance between Ag nanostructures and high surface coverage of Ag nanostructures for Ag20 and Ag40, as shown in Fig. 2 and Fig. 6(c), the transport of carriers between Ag nanostructures may be more serious for Ag20 and Ag40 comparing to that for Ag60 and Ag80, which have a longer particles’ distance and a lower surface coverage. Better conductance is achieved for the devices Ag20 and Ag40 than that for Ag60 and Ag80. However, the accumulation of carriers at the interface states between the contact electrode and the active material will also be more serious for Ag20 and Ag40 than that for Ag60 and Ag80, from which the limited effective injection will be more obvious due to the screened applied electric field [43]. Interestingly, there is a similar relation between the relative intensity of EL peak with longer wavelength and the improvement of carrier injection. For the sample Ag40, the improvement of carrier injection, especially for electrons, is significant, as shown in Fig. 8, and the EL peak with longer wavelength can be resolved distinctly, as shown in Figs. 4(c) and 4(d). While for the sample Ag80, this improvement of electron injection is not so much obvious and the longer wavelength EL peak can hardly be resolved, as shown in Figs. 4(e) and 4(f). Consequently, the evolution of the EL peak with longer wavelength is originated from the changes of the carrier injection levels, as have been speculated above. From the discussion above, the improvement of the carrier injection may also have an instructive contribution to the enhancement of EL efficiency.

 

Fig. 8 Current density (J) vs. applied voltage (V) plot for Ag0-Ag80 with the values of J for Ag0, Ag60, and Ag80 multiplied by 10 times.

Download Full Size | PDF

4. Conclusion

In summary, an enhanced EL efficiency of SiN_x-based LED is observed by depositing a Ag nanostructures layer on its luminescence layer. By optimizing the size and morphology of Ag nanostructures, a maximal 6.5 times enhancement of integrated EL intensity is achieved at the same input power. Two distinct EL Gauss peaks originated from the radiative recombination between its corresponding band tails and trapped levels can be resolved after the addition of Ag nanostructures. We attribute the enhancement of EL efficiency mainly to the improvement of light-extraction efficiency caused by the surface roughening of ITO electrode, as well as the improved carrier injection efficiency. The improvement of internal quantum efficiency might also have an instructive contribution to this enhancement. Further detailed studies are required to quantify the various contributions of the individual efficiency enhancements. Nevertheless, the current work pointed out the importance of the dimensions of the Ag nanostructures on the optimized EL efficiency of the SiN_x-based LEDs. Moreover, our results may provide an alternative approach towards the fabrication of SiN_x-based or Si-based LEDs with high EL efficiency. Meanwhile, we also provide the possible methods on further improving the EL efficiency of SiN_x-based LEDs, which can be achieved by the optimization of the surface coverage (η_s) of Ag nanostructures, the average distance (d) between Ag nanostructures, and/or the RMS roughness (δ) of ITO surface.