Mutlicolor electroluminescent Si quantum dots embedded in SiO<sub>x</sub> thin film MOSLED with 2.4% external quantum efficiency

Chih-Hsien Cheng; Yu-Chung Lien; Chung-Lun Wu; Gong-Ru Lin

doi:10.1364/OE.21.000391

1. Introduction

The Si-rich SiO_x film with buried silicon quantum dots (Si-QDs) has attracted much attention in the past decade due to its potential application in metal-semiconductor-oxide light emitting diodes (MOSLEDs) with the CMOS compatible processes for all-Si based optical interconnect applications. The controversy on mechanisms of the light emission from Si-QDs has been solved due to the observation of the quantum confinement effect (QCE) and its coincidence with the modeling by the non-phonon assistant carrier recombination rate. Koch et al. and Prokes reported that the light emission of Si-QDs is based on the radiative recombination occurred at surface defect states surrounding the Si-QD core [1, 2]. Another possible mechanism has attributed the luminescence to the band-to-band transition with Si-QDs with QCE by Delley et al. [3]. The latter one has been supported by the Si-QD size-dependent luminescence [4–6]. Park’s group employed a theoretical formula of E(eV) = 1.56 + 2.40/d² to obtain the relationship between the photoluminescence (PL) peak wavelength and the Si-QD size for Si-QDs embedded in the Si-rich SiN_x film [4]. In addition, a similarly theoretical formula of E(eV) = 16.89 + 23.9/d² for the PL of the Si-QDs embedded in the SiN_x matrix with varying the Si-QDs size also be reported by Nguyen et al [5]. In addition, de Boer and associates change the largest average size of Si-QD from 5.5 nm to 2.5 nm to make the PL detuned from 990 nm to 750 nm [6]. Later on, the nearly coincident photoluminescence (PL) and electroluminescence (EL) of n⁺-Si/SiO_x:Si-QDs/p⁺-Si LEDs were reported [7]. The Si-QD related EL is more complicated due to the interactions among the confinement geometry, the valley degeneracy, the crystallographic orientations, and the quantum Stark effect on intraband transitions. Instead of the Si-QDs, the carriers trapped in defects may also contribute to the PL and EL at different wavelengths [8–10]. The weak-oxygen bond and neutral oxygen vacancy related defects cause the PL at 415-455 nm [8]. In addition, the Si = O bonds at the interfaces between Si-QDs and SiO_x matrix also contribute to the PL at 700-710 nm [9]. The radiative oxide defects for Eu³⁺ doped in the SiO_x films offer the PL ranging between 400 and 800 nm [10]. Nevertheless, the quantum efficiency of the light emission from Si-QDs or radiative defects is strongly correlated with the carrier transport through SiO_x layer [11–13]. The Si-QDs embedded in SiO₂ layer could dominate the charge retention performance [14], which could be strongly affected by annealing [15, 16].

Recently, the research has been emphasized on the SiO_x based MOSLEDs with broadband tunable EL wavelengths in visible region. Furthermore, the MOSLEDs with different matrices, such as SiN_x, SiO_xN_y, etc, are demonstrated to enhance the carrier injection because these matrices have the lower barrier height [17–20]. The external quantum efficiency (EQE) of MOSLEDs suffers from a limited tunneling probability of carriers passing through the SiO_x matrix. The EQE of Si-QD embedded in MOSLEDs of up to 0.2% was reported by several groups [21–24]. Marconi’s group employed the nanocrystalline-Si/SiO₂ multilayers to fabricate the LEDs. The EQE of these devices is 0.2% with an operating voltage of 36 V [22]. Lin and associates used Si nanopillars to enhance the Fowler-Nordeim tunneling and reduce the effective barrier height, providing that the EQE of MOSLEDs is up to 0.1% [23]. In addition, they also detuned the size of Ni nanodots as the metal mask to form the different-sized Si nanopillars which control the PL wavelength ranging from 826 nm, to 874 nm [24]. Under the pulse operation, Nishimura and associates reported a maximal EQE of up to 0.8% [25]. Gelloz et al. further observed that the EQE of LEDs with the buried Si-QDs under the continuous-wave operation can be enhanced up to 1.1% by the post-anodized electrochemical oxidation of porous silicon [26]. The EQE of up to 0.2% for the graded Si-QD LEDs with the Si-QD/SiO₂ multilayer is achieved by Anopchenko et al. [27]. Per’alvarez’s group observed that the EQE of MOSLEDs with the Si-rich SiO_x film grown by ion implantation and PECVD are 10⁻³% and 0.1%, respectively [28]. In addition, the recent work further reported that pointed that the higher barrier height at the ITO/SiO_x:Si-QD interface and the lower capacity of carrier storage within Si-QDs for MOSLEDs could be prompt their power conversion ratio to 2.35 × 10⁻²% [29]. Lin and associates further employed the nano-structures, such as nano-pillars, nano-pyramids, etc., to enhance the EL intensity of MOSLEDs [30, 31]. The EQE of MOSLEDs with an n-ZnO/SiO₂-Si QDs-SiO₂/p-Si heterostructure is improved up to 4.3 × 10⁻²% by Sun’s group [32]. More recently, the EQE of Si-QD related EL device has been improved to 8.6% by using the organic layer as a host matrix [33, 34].

By properly tuning RF plasma powers under plasma-enhanced chemical vapor deposition (PECVD) and lengthening the annealing duration, the multicolor EL of Si-QD embedded in Si-rich SiO_x MOSLEDs with variable Si-QD sizes are demonstrated, and a maximal EQE for the MOSLED containing smallest Si-QDs is reported. In this work, the significantly shortened PL lifetime and the mechanism related to the enhanced EL power are elucidated. Besides, the carrier tunneling and their correlation between the Si-QD sizes and the barrier height at ITO/SiO_x interface are studied.

2. Experiment setup

The 450-nm thick SiO_x films were deposited on (100)-oriented p-type Si substrates with a resistivity of 1 × 10⁻³ Ω-cm by a PECVD system under constant SiH₄ and N₂O fluences of 40 sccm and 125 sccm, respectively, at a chamber pressure of 100 Pa and a deposition temperature of 200°C. The RF plasma power is detuned from 20 to 50 W at the 10 W increment for changing the Si-QD sizes and the PL/EL wavelengths [35]. Afterwards, the SiO_x samples were encapsulated annealed in a quartz furnace with flowing N₂ atmosphere at 1100°C for 2.5 or 90 min to induce Si-QDs precipitation. To fabricate a MOSLED, a 100-nm-thick indium tin oxide (ITO) film with a contact diameter of 0.8 mm was sputtered on the top of the SiO_x surface, and a 200-nm-thick Al film was evaporated at the bottom of the p-Si substrate. In our previous work, the resistivity of the ITO films is improved to 1.2 × 10⁻⁴ Ω-cm after annealing at 450°C for 10 min [36]. Therefore, all samples were post-annealed in a quartz furnace at 450°C under N₂ ambient for 10 min to improve the ITO resistivity [36, 37]. Under the illumination with a 1-ns gain-switched GaN laser diode at wavelength of 405 nm, the full-band time-resolved PL (TRPL) analysis with detection wavelengths ranging from 410 nm to 900 nm was resolved by a photomultiplier (Hamamatsu R928) with switching response of 2-3 ns. The current-voltage (I-V) analysis of MOSLEDs was performed by using a programmable electrometer (Keithley, model 237). The room-temperature EL was measured by driving the MOSLEDs above the Fowler-Nordheim (F-N) tunneling threshold voltage. The EL power detection ranging from 300 nm to 900 nm in silicon integral sphere head (ILX, OMH-6703B) connected with power multimeter (ILX, OMM-6810B) was performed. The schematic diagram of SiO_x MOSLEDs with buried Si-QDs is shown in Fig. 1 . By taking the SiO_x sample with smallest Si-QDs as an example, the buried Si-QD size has a relatively uniform distribution within Si-rich SiO_x films. The average size and the standard deviation of the Si-QD embedded in SiO_x film decrease from 4.2 ± 0.4 to 1.8 ± 0.2 nm with the RF plasma power increasing from 20 to 50 W, respectively.

Fig. 1 The schematic diagram of a MOSLED made by SiO_x film with buried Si-QDs (b) HRTEM image for the 1.7 ± 0.2 nm large Si-QDs embedded in the SiO_x film grown at an RF plasma power of 50 W. (c) Size distribution of Si-QDs embedded in the SiO_x film grown at an RF plasma power of 50 W.

Download Full Size | PDF

3. Results and discussion

Under forward bias, the electrons and holes inject into Si-QDs by F-N tunneling through the Si-rich SiO_x host matrix from ITO and p-Si substrate, respectively. With the RF power increasing from 20 to 50 W, the turn-on electric field increases from 3.2 to 9.2 MV/cm and from 2.6 to 8.8 MV/cm after annealing for 2.5 and 90 min, respectively. The increasing Si-QD size could lead to an enhancement of the tunneling current. That is, the lower turn-on electric field is observed in MOSLEDs with enlarged Si-QDs. The compliance voltage dropped by lengthening the annealing time of SiO_x film before fabricating MOSLEDs, which originates from the thickness shrinkage of SiO_x film after long-term annealing. By defining a transmission coefficient of T(E_x) as a function of the energy E_x incident on the barrier at metal/oxide interface, and then summing up over all possible energies using the Wentzel-Kramers-Brillouin (WKB) approximation in the absence of Schottky effect with x_ti = 0Å (with ignored temperature variation and barrier shrinkage), the F-N tunneling current density, J_F-N, can be expressed as [38]:

\begin{matrix} J_{F - N} = \frac{4 π q m_{0}}{η^{3}} \int_{E_{x}} e^{- 2 {\int_{x_{t i}}^{x_{t x}} (\frac{2 m_{o x}}{η^{2}})}^{1 / 2} [{(q Ψ (x) - E_{x})}^{1 / 2}] d x} d E_{x} \int_{E_{x}}^{+ \infty} f (E_{x}, T) d E \\ = \frac{q^{3} (m_{0} / m_{o x})}{8 π h ϕ_{B}} E^{2} \exp (- \frac{8 π \sqrt{2 m_{o x} ϕ_{B}^{3}}}{3 q h E}) \\ = 1.54 \times 10^{- 6} \frac{(m_{0} / m_{o x})}{ϕ_{B}} E^{2} \exp (- 6.83 \times 10^{7} \frac{\sqrt{2 (m_{o x} / m_{0}) / ϕ_{B}^{3}}}{E}), \end{matrix}

where m_o denotes the free electron mass, T the temperature, f (E_x,T) the Fermi Dirac function, E the electric field in SiO_x layer, and Φ_m the potential barrier at metal/oxide interface, m_ox the effective mass of electron in SiO_x (m_ox = 0.5m₀), h the reduced Planck constant, qΨ(x) the potential barrier in SiO_x layer at x abscissa, q the electron charge, and x_tx-x_ti the tunnel distance in SiO_x layer. As a result, the turn-on electric field and barrier height shown in Fig. 2 can be extracted from the ln(J_F-N/E²) vs. E plot, in which the intersection point of two fitted lines represents the turn-on electric field and the barrier height can be calculated from the descending slope.

Fig. 2 (a) The F-N plot of ln(J_G/E²) dependent electric field as a function of RF power with their SiO_x annealing at 2.5 (upper) and 90 min (lower). (b) Barrier height (black line) and turn-on electric field (blue line) of the MOSLEDs with 2.5-min (square patterns with linked dashed line) and 90-min (circle patterns with linked solid line) annealed SiO_x samples grown at different RF plasma powers.

Download Full Size | PDF

With a forward (positive) bias applied from Al contact (p-type Si) to ITO contact (SiO_x gate), electrons are injected from ITO gate contact and holes are injected from Al contact of the MOSLED device operated at an accumulation condition. With a relatively large bias added at p-type Si side, the band diagram of MOSLEDs is analogous to that of the p-type MOS diode operated at the accumulation state, as shown in Fig. 3 . With increasing RF plasma powers, the enhancing turn-on electric field accompanied with the enlarging interfacial barrier height from 1 eV to 3.6 eV is observed no matter annealing the SiO_x at short or long duration, as shown in Fig. 2. The existence of buried Si-QDs leads to a decreased turn-on voltage of the F-N tunneling and creates a tunneling path for carriers from Si substrate to ITO contact. A larger RF plasma power suppresses the Si-rich condition to shrink the Si-QDs in more stoichiomatric SiO_x ilm, which provides a larger interfacial barrier height for carrier tunneling from ITO to SiO_x.

Fig. 3 The band diagram of MOSLEDs grown at an RF plasma power of 50 W. Inset: The variations on the band diagram of the MOSLEDs grown at different RF plasma power.

Download Full Size | PDF

For F-N tunneling, the calculated barrier height is effective barrier height at the interface between ITO gate and the whole SiO_x:Si-QD film. Therefore, the effective barrier height is determined by the material characterization of SiO_x film. Even though, the barrier height of SiO_x/ITO junction is smaller than that of a pure SiO₂/ITO junction (3.7 eV), indicating that the O/Si composition ratio is still below 2.0. The band diagrams of the MOSLEDs with Si-QDs embedded in the SiO_x grown with different RF plasma powers are illustrated in Fig. 3, where the barrier height between ITO and SiO_x grown with different RF plasma powers are depicted in the inset of Fig. 3. For example, the barrier heights of MOSLEDs with buried Si-QDs with sizes of 4.2 nm and 1.8 nm are 1.05 and 3.62 eV, respectively. It is straightforward that the band diagram bends more in smaller size of Si-QDs, indicating that a larger external electric field is needed to turn on the F-N tunneling mechanism in the MOSLEDs with smaller Si-QDs embedded in the SiO_x. In our case, the O/Si composition ratio of the SiO_x film enlarges with increasing RF plasma power. In addition, the excess Si concentration has a significant effect on the turn-on voltage of MOSLEDs [39]. The less excess Si in the SiO_x film provides fewer tunneling paths to make the current enhance. The SiO_x film also enlarges its resistance with increasing O/Si composition ratio because the SiO_x film gradually becomes non-stoichiometric to approach a pure SiO₂ matrix. Therefore, the carriers need a higher turn-on voltage to be injected into Si-QDs. Moreover, the effective barrier height is varied with changing Si-QD size. That is because the smaller Si-QDs decrease the effective dielectric constant and enhance the barrier height of F-N tunneling to degrade overall tunneling probability [40].

Pavesi and Turan have simulated the individual states of the electron and hole energy levels with increasing Si-QD size. Their results have shown that the variation of conduction band is more than that of valence band for Si-QD size of smaller than 6 nm, and the quantum confinement effect of Si-QDs larger than 6 nm gradually diminishes [41]. These results clearly shown that the Fermi level in Si-QD moves upward with decreasing Si-QD size. The conduction band moves upward with shrinking Si-QD size, whereas the valence band moves downward with a smaller shift. Their simulation supports an upward movement of the Fermi level for smaller Si-QDs. Without bias, the larger bending degree of energy band at the Si-QD/SiO₂ interface is observed for samples with smaller Si-QDs grown at higher RF plasma powers.

Alternatively, there is another simulation work attributing the increasing interfacial barrier height of Si-QDs with decreasing size to the down-shifted Fermi level, from which the difference between the Fermi level of Si-QDs and that of ITO decreases making the bending degree of the energy band less significant without external bias. The Fermi level would down-shift by shrinking Si-QD size because the asymmetrical shift of the conduction band and the valence band of Si-QDs. Due to the quantum confinement effect, the up-shift of the conduction band toward the vacuum band is less significant (actually, this shift can be ignored) than the down-shift of the valence band when decreasing the Si-QD size [42]. This results in an enlarged energy band (E_g) with down-shifted Fermi level for smaller Si-QDs, providing an enlarged interfacial barrier height by decreasing the Si-QD size. With the decrease of the size of Si-QDs by enlarging the RF power or lengthening the post annealing time, the Fermi level of Si-QDs down-shifts and the difference between the Fermi level of Si-QDs and that of ITO decreases making a less significant bending degree of energy band. Correspondingly, the effective barrier height Ф_B increases to assist carriers F-N tunneling from ITO to Si-QDs through SiO₂ matrix. This inevitably induces the increased turn-on electrical fields (E_turn-on), as shown in Fig. 2.

Note that both models can successfully elucidate the decrease of the effective barrier height by enlarging Si-QD size and density results in a better carrier injection (it also depends on the resistance of the films). The overlap between electron and hole wave functions in both real and momentum space will increase in smaller Si-QDs with a higher probability of non-phonon assisted radiative recombination. Furthermore, the density of Si-QDs with smaller sizes is larger than that with larger sizes. Hence, the number of the active Si-QDs contributing to the luminescence is also increased as the size of Si-QDs reduces. The aforementioned mechanisms explain why a lower carrier injection as well as smaller current could lead to a larger external quantum efficiency and higher EL (see Fig. 5) observed from smaller Si-QDs even with a larger turn-on electrical field E_turn-on, as shown in Fig. 2.

The full-band TRPL from 410 nm to 900 nm for all SiO_x:Si-QDs have been analyzed. In principle, the full-band TRPL trace of the SiO_x:Si-QDs with a continuously distributed radiative lifetime is contributed by all Si-QDs with different sizes. Because of the inhomogeneous broadening luminescence of Si-QD, the full-band TRPL decay traces of SiO_x:Si-QDs should be analyzed by using a multi-exponential decay function shown in Eq. (1). However, the discrete multi-exponential decay function cannot perfectly fit the TRPL trace due to the continuous Si-QD size distribution. Therefore, the stretched exponential function is employed to evaluate the lifetime dispersion effect of the SiO_x:Si-QDs sample, as given by

I (t) = \sum A_{n} e^{- (\frac{t}{τ_{n}})} \Leftrightarrow I (t) = I (0) e^{- {(\frac{t}{τ})}^{β}},

where β is lifetime dispersion factor showing a reciprocal correlation with the Si-QD size distribution, i.e. b is inverse proportional to the full-width-at-half-maximum of size distribution (ΔD). The stretched exponential decay function can precisely fit the experimental TRPL trace to obtain the average lifetime of Si-QDs and the corresponding lifetime dispersion factor with error bars of less than 3%. The TRPL lifetime of Si-QD related luminescence for 2.5min annealed SiO_x samples grown by increasing RF plasma powers from 20 to 50 W significantly shortens from 5 μs to 0.31 μs, as shown in Fig. 4(a) . After lengthening the annealing time up to 90 min, the TRPL lifetime is lengthened by 2-3 times due to the enlarged Si-QD size after precipitation. Although the oxygen-related defects also induce another nanosecond blue-band emission mechanism, it can be excluded since its TRPL lifetime is as short as 10 ns [8, 43]. The shortening of TRPL lifetime by two orders of magnitude is strongly correlated with the shrinking Si-QD size from 4.2 to 1.8 nm. In smaller Si-QDs, the TRPL lifetime is significantly shorter because of a better overlap between wave functions of electron and hole states due to quantum confinement effect. A non-phonon assisted electron-hole recombination procedure within narrower momentum overlapped region occurs to provide higher recombination rate with enhanced probabilities of the direct transition and the increased EQE [44]. This explains the observation of the higher EQE in blue-light MOSLED. This evidence further supports the direct radiative recombination in Si-QD core with a strong quantum confinement effect. For the SiO_x:Si-QDs samples grown by increasing RF plasma powers, the lifetime dispersion factor (β) increases from 0.63 to 0.75 and from 0.61 to 0.7 after 2.5-min and 90-min annealing, respectively. The enhanced lifetime dispersion phenomenon occurs with enlarging Si-QD size distribution, thus providing a reducing value of the lifetime dispersion factor b. For example, the dispersion factor of 0.63 for SiO_x samples grown with the RF plasma power of 20 W has a radiative lifetime ranged between 7 and 24.5 μs. In contrast, the lifetime dispersion shrinks to range between and ns for the SiO_x samples grown with the RF plasma power of 50 W.

Fig. 4 The experimental (solid line) and fitted (dash line) TRPL traces of the (a) 2.5-min and (b) 90-min annealed SiO_x films with buried Si-QDs. (c) TRPL lifetime of the 2.5-min and 90-min annealed SiO_x film with buried Si-QDs as a function of RF plasma power.

Download Full Size | PDF

When increasing RF plasma powers from 20 to 50 W, the maximal EL power of the Si-QDs embedded in SiO_x MOSLEDs made by the 2.5 min annealed SiO_x samples slightly enlarges from 7.5 to 75 nW. In contrast, the EL power significantly enlarges by one order of magnitude when lengthening the annealing time of SiO_x films from 2.5 to 90 min. As a result, the maximal EL power can be enlarged up to 0.7 μW if the RF plasma power increases to 50 W and the annealing time lengthens to 90 min. The corresponding power-current slope also increases from 0.05 to 73.87 and 0.09 to 5.7 mW/A for the MOSLEDs with SiO_x films annealing at 2.5 and 90 min, respectively. Such an enhancement can be attributed to either the improved carrier transport under an enhanced tunneling environment, or the increasing Si-QD density in SiO_x films grown at larger RF plasma and annealed at lengthened duration. To further discriminate the main contribution from two possibilities, the time resolved PL has been performed for obtaining the lifetime and recombination rate of carriers in the SiO_x samples prepared at different recipes [45]. The increasing trend of optical power-current slope is strongly correlated with the increment of Si-QD density, which is proportional to the normalized PL intensity per thickness, indicating the increasing number of Si-QDs. However, the turn-on currents of MOSLEDs with the 2.5 min and 90 min annealed SiO_x samples decrease from 53.9 to 0.2 μA and 357.9 to 8.6 μA, respectively. The electric field across the SiO_x film beyond turn-on is dominated by F-N tunneling mechanism. The smaller Si-QD provides a deeper quantum well to enlarge the barrier height between the conduction band of Si-QDs and oxide which leads to a better confinement for carriers, yet the recombination rate is still limited by the less overlapped wave-functions between electron and hole. The P-I curves also show the degradation of current endurance in Si-QDs based MOSLEDs grown with the decreasing RF power, which is related to the high turn-on nearly the breakdown edge of SiO_x. By defining the power conversion ratio (PCR) as η_PCR = P_opt/I_bias × V_bias with V_bias, I_bias, and P_opt denoting the biased voltage, the biased current and the constant optical output power, respectively. The PCR of these devices increases from −67 dB to −38dB after annealing.

In more detail, the EQE of the MOSLEDs grown with different RF plasma powers are plotted as a function of current density as shown in Fig. 5 , which is defined as the ratio of the output photon number to the input electron number,

η_{e x t} = \int_{λ_{1}}^{λ_{2}} \int_{t_{0}}^{t_{1}} \frac{P_{E L} (t, λ)}{I (t)} \frac{e}{h ν} d t d λ = \frac{\int_{λ_{1}}^{λ_{2}} P_{E L} (λ) λ d λ}{1.24 I_{b i a s}},

where P_EL defines the optical output power, λ defines the peak wavelength, and I_bias defines the biased current. The maximal EQE of MOSLEDs with the 2.5-min annealed SiO_x samples grown by enlarging RF plasma power from 20 to 50 W significantly increases from 1.9 × 10⁻⁵ to 2.4%. The MOSLEDs with SiO_x films grown under RF power 50 W and after 2.5 min annealing demonstrate the best PCR and the highest EQE of 2.4% than ever. After annealing up to 90 min, the same devices slightly degrade its maximal EQE by one order of magnitude. For low RF power grown (20 and 30 W) MOSLEDs, the P-I slope, PCR and maximal EQE of long-term annealed devices are better than those of short-term annealed devices, but high RF power grown (40 and 50 W) devices show opposite results. The reason for such phenomenon is that the over-annealing contributes to the enlarged Si-QD size and the attenuated power at short wavelength region. There is a trade-off between the energy transforming efficiency and the operation reliability of these SiO_x MOSLEDs with buried Si-QDs. The EQE of MOSLEDs has a slightly degraded phenomenon under the higher biasing current, as shown in Fig. 5. That is because the MOSLEDs under an operation of higher current injection easily contribute to a rise on device temperature. In addition, the impact ionization of hot carriers for devices easily occurs under an operation of the higher electric field [45]. The additional kinetic energy makes the electron-hole pairs separate from but not recombine with each other. Therefore, this phenomenon also contributes to the EQE degradation of MOSLEDs under the higher biasing current and electric field. In addition, the EQE of MOSLEDs also has a reduced phenomenon when the recombination mechanism is dominated by the defect or Auger recombination. The dominated recombination could be determined by the Z-parameter analysis [46, 47]. The Eq. (3) is based on the hypothesis of the Boltzmann statistics of carriers and an absence of leakage currents [48]. The Z-parameter could be obtained by the P-I curve of MOSLEDs, which is described as

Fig. 5 EQE (solid line) and EL power (dashed line) of MOSLEDs with 2.5-min (left) and 90-min (right) annealed SiO_x samples grown at different RF plasma powers.

Download Full Size | PDF

Z \equiv \frac{\ln (I)}{\ln (P^{0.5})} .

The Z-parameter varies between 1 and 3 for different recombination mechanisms. Typically, the Z parameters are 1, 2, and 3 for the defect, radiative, and Auger related recombinations, respectively. The non-integer Z-parameter results from the mixed contribution of two different recombination mechanisms. For example, the Z-parameter between 2 and 3 represents the mixed contribution of the radiative and Auger related recombinations. When Z-parameter gradually approaches the specific integers, it represents that the specific recombination mechanism of MOSLEDs is gradually dominated. Figure 6(a) shows the Z-parameter as a function of biased currents for the MOSLEDs with 2.5-min annealed SiO_x samples grown at different RF plasma powers, respectively. At lower bias currents, the dominated recombination for MOSLEDs with 2.5-min annealed SiO_y grown at different RF plasma powers is radiative recombination with a corresponding Z-parameter of approximately 2. However, the Z-parameter gradually approaches 3 at extremely high bias, indicating that the Auger recombination becomes to dominate MOSLEDs. The Z-parameter for MOSLEDs with 90-min annealed SiO_x samples has a similar trend, as shown in Fig. 6(b). In particular, the Auger recombination becomes more significant for the MOSLEDs with 90-min annealed SiO_x grown at different RF plasma powers. The Auger recombination even starts at lower biased currents. Therefore, the EQE of MOSLEDs with 90-min annealed SiO_x samples is apparently lower than that of devices with 2.5-min annealed SiO_x samples.

Fig. 6 The Z-parameter vs. biased current of MOSLEDs with (a) 2.5-min and (b) 90-min annealed SiO_x samples grown at different RF plasma powers.

Download Full Size | PDF

The annealing time dependent EL spectra blue-shifts with enlarging RF plasma power, as shown in Fig. 7 . The biased current of MOSLEDs with SiO_x films grown at larger RF plasma powers is decreased, because the O/Si composition ratio of the SiO_x layer is increased dramatically to make the host matrix seriously isolated. The required turn-on current significantly reduces from 53.9 to 0.2 μA as the RF plasma power enlarges from 20 W to 50 W. In opposite, the EL intensity is enlarged by 40 times for the blue-light MOSLED even biases at smaller current. This is strongly correlated with the optical power-current slope discussed previously. For the MOSLED made by low-plasma grown SiO_x with buried Si-QDs, two peak wavelengths of 480 and 681 nm are observed. After 90 min annealing, the latter peak has a red-shifted phenomenon due to the QCE. These two peaks compete and evolute each other as the biased current increases. The growth of SiO_x films at higher RF plasma powers enables the precipitation of smaller Si-QDs in SiO_x, which provides the EL color shifted from red to blue color even with short-term annealing. In particular, the long-term annealing makes Si atoms obtain more energies to diffuse a longer length to form larger Si-QDs, hence the corresponding EL spectrum slightly broadens toward long wavelengths and attenuates its EL intensity at shorter wavelengths accordingly.

Fig. 7 EL spectra of MOSLEDs with SiO_x grown at RF plasma powers of (a) 20 W, (b) 30 W, (c) 40 W and (d) 50 W under a biased current density of 0.1 mA/cm².

Download Full Size | PDF

The carrier transport and recombination through small Si-QDs are less efficient than those through large Si-QDs, because the less Si-rich SiO_x host environment inevitably degrades the tunneling of carriers and the probability of tunneling through small Si-QDs is significantly reduced. With enlarging biased current, the EL contributed by larger Si-QDs gradually saturates and the overflowed carriers lead to the enhanced recombination in smaller Si-QD, as shown in Fig. 8 . The inset of Fig. 8 depicts the linear relationship between the EL intensity and biasing currents. The output power as a function of biased currents can be written as P_EL = η_ext × I_bias × (1.24/λ_EL). For MOSLEDs with buried Si-QDs of different sizes, η_ext and λ_EL are varied with the Si-QD sizes. The same phenomenon between EL intensity and biased voltages has been reported by Irrera et al [49]. The large-size Si-QD dependent EL grows and saturates earlier at lower biased condition, yet the EL power linearly increases before its saturation at higher biases.

Fig. 8 Normalized EL spectra obtained under different biased currents for the MOSLED with SiO_x film grown at an RF plasma power of 40 W. Inset: the EL intensity as a function of biased current.

Download Full Size | PDF

The color of EL pattern shown in Fig. 9 varies from red to blue by increasing RF plasma powers during PECVD growth. Growing the SiO_x at low RF power, the EL pattern is dark because of a low EL power for larger Si-QDs due to the greatly reduced wave-function overlap and recombination rate of electrons and holes. In contrast, the EL patterns become brighter with steeper P-I slope for MOSLEDs made on the SiO_x prepared with lengthening annealing time. The long-term annealing provides brighter patterns due to the endurance of these MOSLEDs under higher current injection. The EL pattern of a 40-W grown SiO_x based MOSLED reveals slightly green at lower bias but becomes a white-light pattern at higher current due to the EL spectral broadening and blue-shift effects. The brightest EL is observed from the MOSLEDs with smallest Si-QDs embedded in SiO_x. The MOSLED with the 2.5-min annealed SiO_x film shows a purely blue color but that with the 90-min annealed SiO_x film varies to a light blue pattern due to the contribution of the slightly larger Si-QDs.

Fig. 9 EL patterns (under a biased current density of 0.1 mA/cm²) of MOSLEDs with 2.5-min (upper) and 90-min (lower) annealed SiO_x samples grown by changing RF plasma power from 20 to 50 W (from left to right).

Download Full Size | PDF

4. Conclusion

The carrier recombination rate and external quantum efficiency enhanced multicolor emission from SiO_x based MOSLEDs with buried Si-QDs are demonstrated. With the RF plasma power increasing from 20 to 50 W during PECVD growth, the Si-rich condition is controlled to shrink the Si-QD size in non-stoichiomatric SiO_x films and to provide a large interfacial barrier height for carrier confinement within small Si-QDs. The MOSLED made on a 2.5-min annealed SiO_x film with buried Si-QDs slightly enlarges its maximal EL power from 7.5 to 75 nW. In contrast, the EL power significantly enlarges by one order of magnitude to 0.7 μW when the annealing time lengthens up to 90 min. The TRPL analysis reveals that the PL lifetime shortens from 5 µs to 0.31µs with shrinking the Si-QD size from 4.2 to 1.8 nm. This is attributed to a larger overlap between electron and hole wave functions, which essentially leads to a faster non-phonon-assisted direct carrier recombination in smaller Si-QDs. The turn-on currents of MOSLEDs after annealing for 2.5 and 90 min decrease from 53.9 to 0.2 μA and from 357.9 to 8.6 μA, respectively. The smaller Si-QD provides a deeper quantum well to enlarge the barrier height at ITO/SiO_x interface, which leads to a better carrier confinement for carriers at cost of a less electron-hole wave-function overlap. The EQE of the Si-QD embedded in SiO_x MOSLED grown with enlarging RF plasma powers significantly increases from 1.9 × 10⁻⁵ to 2.4% after annealing for 2.5 min. After annealing up to 90 min, the same devices slightly degrade the EQE by one order of magnitude but enlarge the EL power due to the enhanced carrier injection. The growth at higher plasma powers enables the precipitation of smaller Si-QDs in SiO_x, which provides the EL color shifted from red to blue even with short-term annealing. In particular, the long-term annealing makes Si atoms obtain more energies to diffuse the longer length to form larger Si-QDs so that the EL spectrum slightly broadens toward long wavelengths and attenuates its EL intensity at shorter wavelengths accordingly. Such an all Si-based multicolor MOSLED with enhanced on IQE and EQE is fully compatible with current Si fabrication process, which could be used as an on-chip transmitter in the next-generation optical interconnect network to improve the chip-to-chip transmission performance.

Acknowledgment

This work was partially supported by the National Science Council, Taiwan, R.O.C. and the Excellent Research Projects of National Taiwan University, Taiwan, R.O.C., under grants NSC 100-2221-E-002-156-MY3, NSC 101-2622-E-002-009-CC2, NSC 101-ET-E-002-004-ET and 99R80301.

References and links

1. F. Koch, V. Petrova-Koch, and T. Muschik, “The luminescence of porous Si: the case for the surface state mechanism,” J. Lumin. 57(1-6), 271–281 (1993). [CrossRef]

2. S. M. Prokes, “Light emission in thermally oxidized porous silicon: Evidence for oxide-related luminescence,” Appl. Phys. Lett. 62(25), 3244–3246 (1993). [CrossRef]

3. B. Delley and E. F. Steigmeier, “Quantum confinement in Si nanocrystals,” Phys. Rev. B Condens. Matter 47(3), 1397–1400 (1993). [CrossRef] [PubMed]

4. N.-M. Park, C.-J. Choi, T.-Y. Seong, and S.-J. Park, “Quantum Confinement in Amorphous Silicon Quantum Dots Embedded in Silicon Nitride,” Phys. Rev. Lett. 86(7), 1355–1357 (2001). [CrossRef] [PubMed]

5. P. D. Nguyen, D. M. Kepaptsoglou, Q. M. Ramasse, and A. Olsen, “Direct observation of quantum confinement of Si nanocrystals in Si-rich nitrides,” Phys. Rev. B 85(8), 085315 (2012). [CrossRef]

6. W. D. A. M. de Boer, D. Timmerman, K. Dohnalová, I. N. Yassievich, H. Zhang, W. J. Buma, and T. Gregorkiewicz, “Red spectral shift and enhanced quantum efficiency in phonon-free photoluminescence from silicon nanocrystals,” Nat. Nanotechnol. 5(12), 878–884 (2010). [CrossRef] [PubMed]

7. J. Linnros and N. Lalic, “High quantum efficiency for a porous silicon light emitting diode under pulsed operation,” Appl. Phys. Lett. 66(22), 3048–3050 (1995). [CrossRef]

8. G.-R. Lin, C.-J. Lin, C.-K. Lin, L.-J. Chou, and Y.-L. Chueh, “Oxygen defect and Si nanocrystal dependent white-light and near-infrared electroluminescence of Si-implanted and plasma-enhanced chemical-vapor deposition-grown Si-rich SiO₂,” J. Appl. Phys. 97(9), 094306 (2005). [CrossRef]

9. M. Wang, D. Yang, D. Li, Z. Yuan, and D. Que, “Correlation between luminescence and structural evolution of Si-rich silicon oxide film annealed at different temperatures,” J. Appl. Phys. 101(10), 103504 (2007). [CrossRef]

10. D. Li, X. Zhang, L. Jin, and D. Yang, “Structure and luminescence evolution of annealed Europium-doped silicon oxides films,” Opt. Express 18(26), 27191–27196 (2010). [CrossRef] [PubMed]

11. E. H. Snow, “Fowler-Nordheim tunneling in SiO₂ films,” Solid State Commun. 5(10), 813–815 (1967). [CrossRef]

12. M. Lenzlinger and E. H. Snow, “Fowler-Nordheim tunneling into thermally grown SiO₂,” J. Appl. Phys. 40(1), 278–283 (1969). [CrossRef]

13. Z. A. Weinberg, “On tunneling in metal-oxide-silicon structures,” J. Appl. Phys. 53(7), 5052–5056 (1982). [CrossRef]

14. G. Chakraborty, S. Chattopadhyay, C. K. Sarkar, and C. Pramanik, “Tunneling current at the interface of silicon and silicon dioxide partly embedded with silicon nanocrystals in metal oxide semiconductor structures,” J. Appl. Phys. 101(2), 024315 (2007). [CrossRef]

15. D. Comedi, O. H. Y. Zalloum, J. Wojcik, and P. Mascher, “Light emission from hydrogenated and unhydrogenated Si-nanocrystal/Si dioxide composites based on PECVD-grown Si-rich Si oxide films,” IEEE J. Sel. Top. Quantum Electron. 12(6), 1561–1569 (2006). [CrossRef]

16. D. Comedi, O. H. Y. Zalloum, E. A. Irving, J. Wojcik, T. Roschuk, M. J. Flynn, and P. Mascher, “X-ray-diffraction study of crystalline Si nanocluster formation in annealed silicon-rich silicon oxides,” J. Appl. Phys. 99(2), 023518 (2006). [CrossRef]

17. M. Wang, J. Huang, Z. Yuan, A. Anopchenko, D. Li, D. Yang, and L. Pavesi, “Light emission properties and mechanism of low-temperature prepared amorphous SiN_x films. II. Defect states electroluminescence,” J. Appl. Phys. 104(8), 083505 (2008). [CrossRef]

18. G.-R. Lin, Y.-H. Pai, C.-T. Lin, and C.-C. Chen, “Comparison on the electroluminescence of Si-rich SiN_x and SiO_x based light-emitting diodes,” Appl. Phys. Lett. 96(26), 263514 (2010). [CrossRef]

19. F. Wang, D. Li, D. Yang, and D. Que, “Enhancement of light-extraction efficiency of SiN_x light emitting devices through a rough Ag island film,” Appl. Phys. Lett. 100(3), 031113 (2012). [CrossRef]

20. D. Li, F. Wang, D. Yang, and D. Que, “Electrically tunable electroluminescence from SiN_x-based light-emitting devices,” Opt. Express 20(16), 17359–17366 (2012). [CrossRef] [PubMed]

21. N. Lalic and J. Linnros, “Characterization of a porous silicon diode with efficient and tunable electroluminescence,” J. Appl. Phys. 80(10), 5971–5977 (1996). [CrossRef]

22. A. Marconi, A. Anopchenko, M. Wang, G. Pucker, P. Bellutti, and L. Pavesi, “High power efficiency in Si-nc/SiO₂ multilayer light emitting devices by bipolar direct tunneling,” Appl. Phys. Lett. 94(22), 221110 (2009). [CrossRef]

23. G.-R. Lin, C.-J. Lin, and H.-C. Kuo, “Improving carrier transport and light emission in a silicon-nanocrystal based MOS light-emitting diode on silicon nanopillar array,” Appl. Phys. Lett. 91(9), 093122 (2007). [CrossRef]

24. G.-R. Lin, C.-J. Lin, H.-C. Kuo, H.-S. Lin, and C.-C. Kuo, “Anomalous microphotoluminescence of high-aspect-ratio Si nanopillars formatted by dry-etching Si substrate with self-aggregated Ni nanodot mask,” Appl. Phys. Lett. 90(14), 143102 (2007). [CrossRef]

25. K. Nishimura, Y. Nagao, and N. Ikeda, “High external quantum efficiency of electroluminescence from photoanodized porous silicon,” Jpn. J. Appl. Phys. 37(Part 2 No. 3B), L303–L305 (1998). [CrossRef]

26. B. Gelloz and N. Koshida, “Electroluminescence with high and stable quantum efficiency and low threshold voltage from anodically oxidized thin porous silicon diode,” J. Appl. Phys. 88(7), 4319–4324 (2000). [CrossRef]

27. A. Anopchenko, A. Marconi, M. Wang, G. Pucker, P. Bellutti, and L. Pavesi, “Graded-size Si quantum dot ensembles for efficient light-emitting diodes,” Appl. Phys. Lett. 99(18), 181108 (2011). [CrossRef]

28. M. Perálvarez, J. Barreto, J. Carreras, A. Morales, D. Navarro-Urrios, Y. Lebour, C. Domínguez, and B. Garrido, “Si-nanocrystal-based LEDs fabricated by ion implantation and plasma-enhanced chemical vapour deposition,” Nanotechnology 20(40), 405201 (2009). [CrossRef] [PubMed]

29. B.-H. Lai, C.-H. Cheng, and G.-R. Lin, “Multicolor ITO/SiO_x/p-Si/Al light emitting diodes with improved emission efficiency by small Si quantum dots,” IEEE J. Quantum Electron. 47(5), 698–704 (2011). [CrossRef]

30. G.-R. Lin, Y.-H. Pai, and C.-T. Lin, “Microwatt MOSLED using SiO_x with buried Si nanocrystals on Si nano-pillar array,” J. Lightwave Technol. 26(11), 1486–1491 (2008). [CrossRef]

31. Y.-C. Lien, Y.-H. Pai, and G.-R. Lin, “Si nano-dots and nano-pyramids dependent light emission and charge accumulation in ITO/SiO_x/p-Si MOS diode,” IEEE J. Quantum Electron. 46(1), 121–127 (2010). [CrossRef]

32. E. Sun, F.-H. Su, Y.-T. Shih, H.-L. Tsai, C.-H. Chen, M.-K. Wu, J.-R. Yang, and M.-J. Chen, “An efficient Si light-emitting diode based on an n-ZnO/SiO2-Si nanocrystals-SiO2/p-Si heterostructure,” Nanotechnology 20(44), 445202 (2009). [CrossRef] [PubMed]

33. K.-Y. Cheng, R. Anthony, U. R. Kortshagen, and R. J. Holmes, “High-efficiency silicon nanocrystal light-emitting devices,” Nano Lett. 11(5), 1952–1956 (2011). [CrossRef] [PubMed]

34. D. P. Puzzo, E. J. Henderson, M. G. Helander, Z. B. Wang, G. A. Ozin, and Z. Lu, “Visible colloidal nanocrystal silicon light-emitting diode,” Nano Lett. 11(4), 1585–1590 (2011). [CrossRef] [PubMed]

35. B.-H. Lai, C.-H. Cheng, Y.-H. Pai, and G.-R. Lin, “Plasma power controlled deposition of SiO_x with manipulated Si Quantum Dot size for photoluminescent wavelength tailoring,” Opt. Express 18(5), 4449–4456 (2010). [CrossRef] [PubMed]

36. W.-L. Hsu, Y.-H. Pai, F.-S. Meng, C.-W. Liu, and G.-R. Lin, “Nanograin crystalline transformation enhanced UV transparency of annealing refined indium tin oxide film,” Appl. Phys. Lett. 94(23), 231906 (2009). [CrossRef]

37. C.-D. Lin, C.-H. Cheng, Y.-H. Lin, C.-L. Wu, Y.-H. Pai, and G.-R. Lin, “Comparing retention and recombination of electrically injected carriers in Si quantum dots embedded in Si-rich SiN_x films,” Appl. Phys. Lett. 99(24), 243501 (2011). [CrossRef]

38. R. H. Fowler and L. Nordheim, “Electron emission in intense electric fields,” Proc. R. Soc. Lond., A Contain. Pap. Math. Phys. Character 119(781), 173–181 (1928). [CrossRef]

39. Z. H. Cen, T. P. Chen, L. Ding, Z. Liu, J. I. Wong, M. Yang, W. P. Goh, and S. Fung, “Influence of implantation dose on electroluminescence from Si-implanted silicon nitride thin films,” Appl. Phys., A Mater. Sci. Process. 104(1), 239–245 (2011). [CrossRef]

40. G. Chakraborty, S. Chattopadhyay, C. K. Sarkar, and C. Pramanik, “Tunneling current at the interface of silicon and silicon dioxide partly embedded with silicon nanocrystals in metal oxide semiconductor structures,” J. Appl. Phys. 101(2), 024315 (2007). [CrossRef]

41. L. Pavesi and R. Turan, Silicon Nanocrystals: Fundamentals, Synthesis and Application (WILEY-VCH Verlag GmbH & Co. KGaA, 2010), Chap. 2, pp. 25.

42. T. Li, F. Gygi, and G. Galli, “Tailored nanoheterojunctions for optimized light emission,” Phys. Rev. Lett. 107(20), 206805 (2011). [CrossRef] [PubMed]

43. G.-R. Lin, C.-J. Lin, and K.-C. Yu, “Time-resolved photoluminescence and capacitance-voltage analysis of the neutral vacancy defect in silicon implanted SiO₂ on silicon substrate,” J. Appl. Phys. 96(5), 3025–3027 (2004). [CrossRef]

44. V. A. Belyakov, V. A. Burdov, R. Lockwood, and A. Merdrum, “Silicon nanocrystals: fundamental theory and implications for stimulated emission,” Advances in Optical Technologies, review article ID 279502 (2008).

45. C.-J. Lin, C. K. Lee, E. W. G. Diau, and G.-R. Lin, “Time-resolved photoluminescence analysis of multidose Si-ion-implanted SiO₂,” J. Electrochem. Soc. 153(2), E25–E32 (2006). [CrossRef]

46. C.-H. Cheng, C.-L. Wu, C.-C. Chen, L.-H. Tsai, Y.-H. Lin, and G.-R. Lin, “Si-rich Si_xC_1-x light-emitting diodes with buried Si quantum dots,” IEEE Photonics J. 4(5), 1762–1775 (2012). [CrossRef]

47. E. P. O’Reilly and M. Silver, “Temperature sensitivity and high temperature operation of long wavelength semiconductor lasers,” Appl. Phys. Lett. 63(24), 3318–3320 (1993). [CrossRef]

48. L. L. Goddard, S. R. Bank, M. A. Wistey, H. B. Yuen, Z. Rao, and J. S. Harris Jr., “Recombination, gain, band structure, efficiency, and reliability of 1.5-µm GaInNAsSb/GaAs lasers,” J. Appl. Phys. 97(8), 083101 (2005). [CrossRef]

49. A. Irrera, D. Pacifici, M. Miritello, G. Franzo, F. Priolo, F. Iacona, F. Sanfilippo, G. Di Stefano, and P. G. Fallica, “Electroluminescence properties of light emitting devices based on silicon nanocrystals,” Phys. E 16(3-4), 395–399 (2003). [CrossRef]

Mutlicolor electroluminescent Si quantum dots embedded in SiO_x thin film MOSLED with 2.4% external quantum efficiency

Abstract

1. Introduction

2. Experiment setup

3. Results and discussion

4. Conclusion

Acknowledgment

References and links

Cited By

Figures (9)

Equations (4)

Optics Express