1. Introduction

Since 1990, the room-temperature red photoluminescence (PL) from nanocrystallite Si prepared by chemical dissolution of anodized porous Si was preliminarily demonstrated [1]. The blue shift phenomenon of PL peak was also observed as the Si porosity increases by lengthening the immersion time in 40% aqueous HF [1]. The PL peak of the embedded Si-QD can be detuned by altering different fabrication methods and recipes with reactive pulsed laser deposition at modified the laser power, and then the Si nanostructures yields the red to infrared PL spectra from 620 to 830 nm [2]. By using Si-ion-implantation with post-annealing and changing the dosage and energy of Si⁺ ions, the PL peak around blue-green (470-550 nm) and near-infrared (710-800 nm) region can be obtained [3]. By employing the magnetron sputtering deposition and concurrently varying the temperature and ratio of the sputtering targets, the larger Si-QDs with near-infrared PL spectra between 830 and 1030 nm are usually observed [4]. For PECVD fabrication, both near-infrared and blue PL (780-800 and 410 nm) were concurrently observed by altering the fluence ratio of reactant gases and substrate temperature [5]. Therefore, it is an important issue to fine-tune the process parameter for controlling the PL at different colors. In addition, it is relatively easy to deposit lots of excessive silicon by changing the deposition recipes in PECVD system, which provides a Si-rich environment for precipitation of dense Si-QDs in the SiO_x film after high temperature annealing. In this work, we demonstrated by changing the radio frequency (RF) plasma power during the PECVD system, the wavelength tunable PL spectra can be obtained from the Si-rich SiO_x film and correlated with different Si-QD size, which corresponds to different O/Si composition ratio of the PECVD grown SiO_x sample.

2. Experiment setup

The Si-rich SiO_x film was grown on p-type (100)-oriented Si substrate by using the PECVD system at pressure of 67 Pa and the N₂O/SiH₄ fluence ratio of 5:1.1. The N₂O fluence was controlled at 150 sccm. The samples were prepared at constant substrate temperature of 450°C. There were six SiO_x samples deposited under RF plasma power varying from 20 to 70 W at increment of 10 W. All of the samples were grown for 120 seconds. Afterwards, the Si-rich SiO_x films were annealed in a quartz furnace with flowing N₂ gas at 1100°C for 10 min helps to precipitate the Si-QD. Subsequently, the room-temperature continuous-wave PL was performed using the He-Cd laser at wavelength and average power of 325 nm and 50 mW, and the working distance between the focusing lens and the sample was fine-tuned to maximize the PL intensity. The thickness of Si-rich SiO_x layer was determined by α-step. The PL ranging from 300 to 900 nm was resolved by a monochromator (CVI DK240) in connection with a photomultiplier (Hamamatsu R928) and a digital multimeter (HP 34401A). The O/Si composition ratio of the Si-rich SiO_x film and the existence of the Si-QD are investigated by transmission electron microscopy X-ray energy-dispersive spectrometer (TEM-XEDS, Link ISIS 300) and high resolution transmission electron microscopy (HRTEM, JEOL 2010), respectively. The infrared vibration modes of the Si-O and Si-H bonds in the SiO_x film were measured by the Fourier Transform Infrared Spectrometer (FTIR, ThermoNicolet NEXUS 470).

3. Results and Discussion

Figure 1 shows the PL patterns of Si-rich SiO_x with emitting color changing at different RF plasma powers. The color of PL patterns are blue-shifted (red, orange, white, green, blue) with Si-QD average size revealing a decreasing trend (4.5 ± 0.2, 3.5 ± 0.2, 2.5 ± 0.1, 2.1 ± 0.1 and 1.7 ± 0.1 nm obtained from HRTEM image) when increasing RF plasma power (20, 30, 50, 60, and 70 W). Figure 2 (a) reveals the PL with their intensities normalized to the SiO_x film thickness when growing the SiO_x film at different RF plasma powers. With RF plasma power increasing from 20 to 70 W, the peak wavelength of PL decreases from 780 to 380 nm and the normalized intensity of PL increases from 0.81 to 5.02 count/nm. For Si-QD size below 3 nm, the HRTEM analyzed average size of Si-QD correlates well with its PL wavelength predicted with a theoretical formula of λ = 1.24/(1.12 + 3.73/d^1.39) reported by Delerue et al. [6], where λ is the peak wavelength and d is the Si-QD diameter. It is expected that the Si-H bonds are reduced in the SiO_x film grown at high plasma powers, such that the Si-QD are surrounded by fewer hydrogen atoms, which somewhat violates the Delerue’s assumption that all the dangling bonds at Si-QD surface are passivated with hydrogen atoms [6]. Therefore, such a disparity about average Si-QD size obtained from experimental data and theoretical simulation can be well elucidated. Recently, Pi et al. attributed the blue emission to defect states but not to the quantum confinement effect of Si-QDs [7]. The blue PL emission is correlated with the oxidation-induced defects at the surface of Si-NCs. Moreover, the Si-QD surrounding also plays a role on the PL spectrum since the oxidation of the QD surface induces a strong red-shifted PL signal, which has been shown and modeled by Wolkin et al. [8]. The recombination mechanism in the oxidized Si-QD is different from that of the original Si-QD under hydrogen passivation. The red-shift of PL obtained from the porous Si sample after oxidation is mainly related to recombination involving a trapped electron or exciton within Si-O bonds. As stated by Wolkin et al. [8], the originally oxygen-free porous Si sample kept in Ar atmosphere shows blue PL at wavelength around 400 nm. After exposure by oxygen, the PL is significantly shifted to 600 nm with a 3-dB spectral linewidth greatly broadening from 100 to 200 nm. The FTIR analysis clearly indicates that the absorption at 1070 cm⁻¹ and 850 cm⁻¹ related to Si-O-Si and Si-O-H bonds exist in the oxidized porous Si sample with their intensities enlarging and gradually saturating by lengthening the exposing duration in oxygen environment up to 24 hrs. The PL is dominated by carriers trapped in Si = O bonds after oxidation. However, such a red-shifted PL phenomenon becomes less distinct in the oxygen invasive porous Si sample with Si-QD size larger than 2 nm (for the sample with green PL), while the wavelength shift greatly reduces to 100 nm or less. If the Si-QD size further enlarges to 2.42 nm (for the sample with yellow PL), such a PL wavelength shift completely diminishes no matter the porous Si sample is exposed in oxygen atmosphere or not.

 

Fig. 1 The PL patterns and HRTEM micrographs of Si-QDs buried in post-annealed Si-rich SiO_x grown at RF plasma power of 20, 30, 50, 60, and 70 W, respectively. (From left to right).

Download Full Size | PDF

 

Fig. 2 (a) The PL peak wavelength and intensity (normalized to film thickness) of the post-annealed Si-rich SiOx film versus RF plasma power during PECVD growth. (b) The PL peak wavelength versus Si-QD size

Download Full Size | PDF

In contrast to the aforementioned work, the Si-QDs are originally surrounded by oxide matrix of the Si-rich SiO_x sample in our case. Therefore, the oxidation effect never occurs again for such Si-rich-SiO_x samples since they have already experienced the oxidation process at very beginning. That is, the PL of the Si-QDs embedded in the SiO_x sample is very stable with a constant emission wavelength. In addition, these works never show a PL peak at energies larger than 2-2.5 eV for oxygen passivated Si QDs and were lack of discussions on the origins of oxygen defect related luminescence. In principle, there two oxygen related defects including weak-oxygen bond (WOB) and neutral oxygen vacancy (NOV) which can contribute to the PL at wavelengths of 415 nm [9] and 455 nm [10], respectively. The PL wavelengths of these defects remain constant and should never vary with different sample recipes. In our case, we have observed the shift of PL wavelength (from PL analysis) in samples with different Si-QD size (from HRTEM analysis), and the proportionality between the PL wavelength and Si-QD is coincident with the theoretical simulation. The PL peak wavelength blue-shifted from 760 to 380 nm when decreasing the Si-QD size from 4.5 to 1.7 nm, which is in a good agreement with the theoretical model established by Delerue and associates. The contributions of weak oxygen bond (WOB) and neutral oxygen vacancy (NOV) with PL peak wavelengths at 415 and 455 nm, respectively, are not significantly observed in the SiO_x:Si-QD samples due to a relatively stable host oxide matrix. In addition, the HRTEM also shows a wider distribution of Si-QD size in PECVD-grown SiO_x film, corroborating the size dispersion induced PL spectral broadening with enlarging plasma power [11].

The thickness of Si-rich SiO_x film as a function of the RF plasma power during PECVD growth is shown in Fig. 3 . When increasing RF plasma power from 20 to 70 W, the thickness of the Si-rich SiO_x film exponentially increases from 80 to 160 nm. As the RF plasma power increases in the PECVD system, the SiH₄ and N₂O molecules obtain more kinetic energy to decompose themselves in PECVD chamber, thus accelerating the surface reaction and increasing the deposition rate to increase SiO_x film thickness at higher RF plasma powers. The inset of Fig. 3 shows the PL spectra of Si-rich SiO_x film grown at RF plasma powers from 20 to 70W with their intensities normalized to the SiO_x film thickness (in the unit of count/nm). All of the as-grown samples exhibit PL spectra around 400 to 500 nm due to E’_δ and weak oxygen bond (WOB) defects [3,9]. After annealing at 1100°C, the PL intensities at visible and near-infrared bands can be greatly enhanced, which is due to the precipitation of Si-QDs in the SiO₂ host matrix by heat treatment. The PL blue-shifts its peak from 780 to 380 nm by raising RF plasma power, indicating that the Si-QD in SiO_x film greatly shrinks its size. Note that the PL peak centered at 760 nm for the SiO_x sample grown at 70 W appears as an artificial signal due to the second order harmonics of the blue PL peak wavelength at 380 nm, which is caused by the grating of the monochromator. From the normalized PL analyses of the SiO_x samples prepared at different RF plasma powers shown as below, it is clearly seen that the PL spectral linewidth elucidates a less confined size distribution of Si-QDs within the SiO_x grown at lower RF plasma power. When the SiO_x is in an extremely Si-rich condition with high density of excessive Si atoms, the Si-QDs with versatile size can be precipitated as a lot of Si atoms are distributed within the diffusion range at a specific annealing temperature. Except the sample grown at RF plasma power of 70 W with a left-side cut blue PL spectrum due to the 325-nm laser line filter, we also observe that enlarging the RF plasma power from 20 to 60 W during SiO_x growth apparently shrinks the PL spectral linewidth from 280 to 210 nm. Such a narrowing PL linewidth directly corresponds to a varied standard deviation of Si-QD size from 22% to 16% with shrinking Si-QD size. We did not take into account the sample with blue PL as the observed spectrum is somewhat distorted due to the high-pass laser line filter in front of the monochromator, which induces a sharp cut around the wavelength of 350 nm.

 

Fig. 3 (a) Thickness of the post-annealed Si-rich SiO_x film synthesized under different RF plasma powers. Inset: the normalized PL of Si-QDs within Si-rich SiO_x film.

Download Full Size | PDF

Figure 4 (a) shows the O/Si composition ratios of the as-grown and annealed SiO_x samples obtained from TEM-XEDS. In more detail, there are distinct deviation on the density of Si and O atoms dissociated from SiH₄ and N₂O at low RF plasma powers, since the dissociation energy of SiH₄ (75.6 kcal/mol) is much smaller than that of N₂O (112 kcal/mol), yielding more Si than O to facilitate the synthesis of Si-rich SiO_x. Thus, a higher concentration of the excessive Si atoms can easier be obtained with lower RF plasma power during PECVD growth. Because the oxygen is unable to be decomposed from the N₂O molecule under a low-plasma power deposition, and therefore the largest concentration of the excessive Si atom can be expected in the SiO_x grown under RF plasma power of 20 W, in which more excess Si atoms can contribute to the formation of Si-QD. With the XEDS analysis built-in with the TEM system, the O/Si composition ratio can be quantitatively calculated by using the Cliff-Lorimer equation [12]. The weight percentages of elements (C_A and C_B) in a binary system and their XEDS measured signal intensities can be correlated with C_A/C_B = k_AB × (I_A/I_B), where the term k_AB is the Cliff-Lorimer factor, I_A and I_B are the TEM/XEDS measured signal intensities [12]. The absolute values of C_A and C_B are obtained at a specific k_AB factor while assuming A and B atoms constitute 100% of the specimen (i.e. C_A + C_B = 100%). After subtracting the background noise and integrating the peak intensity, the individual compositions in terms of atomic% or weight fraction can be obtained [12]. Such a calculation procedure is similar with those used in XPS or RBS analyses. The O/Si ratio of the as-grown SiO_x is linearly increased from 1 to 1.6 as RF plasma power enlarges from 20 to 70 W. Similar increasing trend stands for the SiO_x sample after annealing. Apparently, the SiO_x film also enlarges its O/Si composition ratio by at least Δ(O/Si) = 0.08 after 10-min annealing. A larger deviation up to Δ(O/Si) = 0.15 between the as-grown and annealed SiO_x grown at lower RF plasma powers is mainly attributed to the enhanced Si-QD precipitation procedure in these cases. There is still a partial pressure of oxygen existed in the furnace even though the SiO_x samples were annealed under flowing N₂ ambient. The excess Si atom density for the SiO_x grown at RF plasma power of 20 W is the highest one among all samples, such that the residual oxygen atoms can easily invade into SiO_x films with lower O/Si composition ratios during furnace annealing. With increasing RF plasma power, the N₂O can be dissociated more completely to promote more free radicals of O atoms during the reaction on Si substrate surface, which involves the replacement of Si-H bonds by Si-O bonds to increase O/Si composition ratio. When annealing at higher temperatures, the excessive interstitial Si atoms obtain larger kinetic energy to accumulate and to form bigger Si-QDs. Since there are numerous Si and insufficient O atoms, the Si atoms can simply move and accumulate without restrain by Si-O bonds. That is, it could be easier to synthesize Si-rich SiO_x film embedded with larger Si-QDs if the PECVD-grown SiO_x film has lower O/Si composition ratio to preserve dense excessive Si atoms congenitally. In contrast, only small Si-QDs can be precipitated by the aggregation of fewer Si atoms in PECVD-grown SiO_x film with high O/Si composition ratios. These few Si atoms must travel a long distance within SiOx matrix and restrain from more Si-O bonds. The inset of Fig. 3 also shows that the size of Si-QD is concurrently decreased from 4.5 ± 0.2 nm to 1.7 ± 0.1 nm as the RF plasma power increases from 20 to 70 W. Moreover, the TEM-estimated volume density of Si-QDs within the SiO_x grown at RF plasma power from 20 to 70 W is increased from 3.2 × 10¹⁷ to 3 × 10¹⁸ cm⁻³ in Fig. 4 (b). Finite at a specific annealing temperature constrain the moving range of interstitial Si atoms, thus providing the precipitated Si-QD with smaller size and higher density in the high-O/Si-ratio SiO_x grown with larger RF plasma power.

 

Fig. 4 (a) TEM-XEDS measured O/Si composition ratios of as-grown and 10-min post-annealed SiO_x film versus RF plasma power. (b) The RF plasma power dependent size and volume density of Si-QDs in post-annealed SiO_x samples.

Download Full Size | PDF

To realize the correlation between Si atom diffusion and nucleation, the effective diffusion characteristics of interstitial Si atom in sub-stoichiometric SiO_x is simulated. The Si-QDs can eventually be found by moving the Si atom directly from one interstitial site to other interstitial sites following the diffusion equation of J = -D[∂C/∂r] + <v>_FC under a driving force F, where J is the diffusion flux (mol cm⁻²sec), D is the diffusion coefficient (cm² sec⁻¹) of the Si atom in SiO_x, and C is the concentration of the interstitial Si atoms, as based the Si diffusion species (mol cm⁻³), ∂C/∂r is the concentration gradient along the diffusion axis, and <v>_F is the atomic velocity of the excess Si atom under a certain driving forces. The definition of driving forces, the diffusion equation can be rewritten as a function of Gibbs free energy by J = -D[∂C/∂r] + [CDΔG_m/akT] under a driving force of F = ΔG_m/a, where T and k are the temperature and Boltzmann constant, respectively, ΔG_m is the change of Gibbs free energy, a is the distance between neighboring lattice planes, and the F is the driving force of atomic diffusion. In our case, the negative diffusion flux caused by Si atom nucleation from low-concentration to a high-concentration region, occurs by ΔH<0 and ΔS>0 at a given annealing temperature, which can produce the ΔG_m<0 condition and favor the interstitial Si atom diffusion and nucleation. It is observed that the Si-QD diameter enlarges with increased Si atomic content, and when annealed at the given annealing temperature and time. The growth velocity of Si-QD with a radius (r) in SiO_x host matrix is expressed as [dr/dt]_ro = [D/(C^b-C^a)](∂C/∂r)_ro, where r_o is the initial Si-QD radius, D is the Si atomic diffusion coefficient in SiO_x host matrix, C^a is the composition ratio of SiO₂, C^b is the composition ratio of Si-QD, and C is the composition ratio of the oxide matrix near the Si-QD. By combining above equations, the D* can be rewritten as D* = {[(r₂ ²-r₁ ²)/2t]⋅[(C^b-C^a)/(C^m-C^a)]⋅exp(-Q/kT)}. Note that the equation not only expresses the proportionality of Si-QD radius to SiO_x composition ratio, but also correlates the composition ratio with the activation energy of the SiO_x host matrix at a given annealing temperature and time. According to Fick's first and second Laws, the diffusion length of (4Dt)^0.5 is proportional to the square root of the time and diffusion coefficient. Therefore, the diffusion length is larger than the matrix size, finite Si atoms can thus self-aggregate to form the Si-QDs.

To characterize the relationship between the Si-O-Si bonds with the O/Si composition ratio, Fig. 5 (a) reveals a series of FTIR absorption spectra for as-grown SiO_x films synthesized at different RF plasma powers. The absorption at 630 cm⁻¹ is attributed to the Si-H₂ wagging vibration [13] and another absorption around 2170 cm⁻¹ denotes the Si-H stretching vibration of H-(SiH₂)_n and H-Si-(Si_3-nO_n) with integer n ranging between 1 and 2 [13]. It was indicated that the absorption at around 745 cm⁻¹ represents the vibration mode with strongly coupled Si-H and Si-O-Si motions [14]. Significantly, the wavenumber of the Si-O-Si stretching vibration mode is sensitive to the O/Si composition ratio, which enables the determination on the stoichiometry of SiO_x film. The empirical expression on the O/Si composition ratio dependent wavenumber of the Si-O-Si based FTIR absorption peak is given by ν = 965 + 50x, which can simply be employed to corroborate the oxygen content x obtained from the experiments [15]. With increasing RF plasma power, the linear relationship obtained in Fig. 5(c) confirms that the Si-O-Si vibrational mode increases its wavenumber from 1016 to 1044 cm⁻¹ as the corresponding O/Si ratio increases from 1 to 1.6. In more detail, it is clearly observed that the FTIR absorption peak at 745 cm⁻¹ appears in all as-grown SiO_x samples due to the existence of Si-H vibration-mode dependent absorption peak by H-Si-(Si₂O) bonds is diminished after annealing. Concurrently, the other Si-H stretching-vibration-mode related absorption peaks contributed by H-Si-(Si_3-nO_n) and Si-(Si-H₂)_n bonds also reveal a strong correlation with the RF plasma power during PECVD growth. The hydrogenation of excessive Si atoms can still be remained in the Si-rich SiO_x:H film grown at lower RF plasma power even after high-temperature annealing, whereas the Si-H bonds are seriously dissipated if RF plasma power during PECVD growth is greatly enlarged to achieve a complete decomposition of SiH₄ molecules.

 

Fig. 5 (a) FTIR absorption spectra of as-grown SiO_x films. (b) 10-min-annealed SiO_x films. (c) Wavenumber of Si-O bonds and correlated O/Si composition ratio of as-grown SiO_x films versus RF plasma power.

Download Full Size | PDF

4. Conclusion

In conclusion, we have demonstrated the RF plasma power controlled deposition of SiO_x with manipulated Si-QD size for tailoring PL wavelength. When increasing RF plasma power from 20 to 70 W, the PL blue-shifts its peak from 780 to 380 nm and increases its peak intensity from 0.8 to 5.0 count/nm. Colorful PL of red, orange, white, green, and blue patterns with varying RF plasma power is observed. The TEM-XEDS and FTIR for SiO_x samples reveal that the increasing O/Si composition ratio of SiO_x film from 1 to 1.6 essentially leads to the shrinkage on Si-QD size after high-temperature annealing. Higher O/Si composition ratio in SiO_x film fails to promote the movement of the excessive Si atoms, thus providing smaller Si-QDs with higher volume density. The HRTEM analysis reveals that the average Si-QD size decreases from 4.5 to 1.7 nm with its volume density increasing from 3.2 × 10¹⁷ to 3.0 × 10¹⁸ cm⁻³ as RF plasma power arises up to 70 W. The absorption of Si-O-Si vibrational mode shifts its wavenumber from 1022 to 1070 cm⁻¹, indicating that the SiO_x behaves more like a stoichiometric SiO₂ with complete phase separation between the oxide matrix and the Si-QDs. The dehydrogenation process is observed in SiO_x samples with higher O/Si composition ratios, in which the bending Si-O vibration mode related absorption peak at 650 cm⁻¹ becomes more pronounced than the Si-H₂, Si-(Si₃O) and H-Si-(Si₂O) dependent peaks at 630, 720, and 2170 cm⁻¹. The dissipation of hydrogen atoms from SiO_x:H film of higher O/Si ratios necessary for synthesizing dense and small-size Si-QDs with strong blue PL is concluded.