Formation of silicon quantum dots by RF power driven defect control

Abstract

We studied the structural and optical properties of silicon nitride (SiN_x) films synthesized by changing the applied radio frequency (RF) power in plasma-enhanced chemical vapor deposition. By decreasing the RF power from 100 W to 40 W, the photoluminescence (PL) of SiN_x becomes stronger and the central PL peak position shifts from 2.41 eV to 1.78 eV. The size of the silicon grains becomes bigger as the applied RF power decreases, and eventually nano-sized amorphous Si quantum dots (nsa-Si QDs) are formed in the sample fabricated using 40 W RF power. By analyzing the chemical states of the Si 2p X-ray photoelectron spectroscopy core-level spectra, we found that the formation of nsa-Si QDs is considerably enhanced due to the disappearance of N₃≡Si–Si≡N₃ defects when the applied RF power reaches a certain point between 60 W and 40 W. From these results, we conclude that the type and the number of defects in SiN_x play crucial roles in the initial formation of nsa-Si QDs in SiN_x. In addition, this investigation paves the way for controlling the formation of quantum structures from defects to nsa-Si QDs simply by tuning the RF power. We expect that this work will contribute to the realization of Si-based full-color LEDs or tandem solar cells in the near future.

---

Introduction

Silicon is an indirect band-gap material in which the bottom of the conduction band is not aligned with the top of the valence band in k-momentum space. Because of this, light emission from silicon is believed to be a phonon-mediated process with a low electron–hole recombination probability, which gives silicon poor optical properties.¹ Despite this intrinsic optical limitation in bulk silicon, many attempts have been made to obtain light emission using silicon.

In previous studies, it was found that silicon quantum dots (Si QDs) generated by phase separation in silicon-rich silicon oxide or silicon nitride matrices present strong visible luminescence at room temperature (RT), and that this visible emission is dependent on the QD size, appearing primarily at wavelengths greater than 600 nm.^2–6 Secondly, the various defect states in silicon oxide or silicon nitride matrices have been reported as good emission centers. In particular, the defect states in silicon nitride provide multi-fold energy levels within the band gap, which allow for various optical transitions.^7,8 The visible emissions from defect states in silicon nitride primarily cover the blue and green visible spectrum. However, color control is difficult due to the rigid energy levels of the defect states.

Furthermore, visible light emission via surface functionalization on freestanding Si QDs that are not surrounded by a host matrix has recently been studied.^9,10 Depending on the surface groups used in surface functionalization, the light emissions from freestanding Si QDs can be tuned in the visible range, which shows that the engineering of QD surface states and QD size control are important in tuning freestanding Si QD emissions. However, an additional acid etching process is necessary to liberate Si QDs from the silicon oxide matrix, and the transfer of Si QDs from solution to solid substrates is still difficult to solve.

We notice that previous studies on light emission from silicon have focused only on either QDs or defect originated emissions, limiting applications such as Si-based full-color LEDs and tandem solar cells which would require light emission across the full visible range.^3,4,7,8 Therefore, it would be very helpful to control the formation mechanism between quantum structures originating from defects or QDs by changing only one experimental variable.

The previous studies reported that radio frequency (RF) power influenced the dissociation of N₂ gas more strongly than that of silane gas among the reactant gases used in plasma-enhanced chemical vapor deposition (PECVD).^11,12 This gives a way to readily control the stoichiometry of silicon nitride films by adjusting only the RF power, without changing any other parameters. Therefore, we chose the RF power as an efficient variable to cross qualitatively different luminescence origins from QDs to defects by controlling the chemical phases of the host material in silicon nitride films. Since this method allows not only the tuning of the emission wavelength over the full visible range but also offers very simple and eco-friendly fabrication, we expect to realize Si-based LEDs and solar cells in the near future.

In this work, we study the structural and optical properties of silicon nitride films synthesized by changing the applied RF power during PECVD and find that a certain RF power between 60 W and 40 W is the critical condition needed to cross the luminescence origin from defect-dominated to QD-dominated transition regimes. Furthermore, we investigate the crucial roles played by the type and the quantity of defects in silicon nitride in the initial formation of Si QDs. Controlling the defects by tuning the RF power provides a way to readily control the formation of quantum structures from defects to Si QDs, which is expected to facilitate future Si-based full-color LEDs and tandem solar cells that require multiple absorption and emission processes.

Experimental

Amorphous silicon nitride (a-SiN_x) films were fabricated by PECVD in a glow discharge system with a conventional parallel plate radio frequency of 13.56 MHz. We used He-diluted 5% silane and additional N₂ of 6 N (99.9999%) grade as the reactant gas sources. The silicon substrates used in this work were boron-doped and had (001) orientation with a resistivity of 0.02 Ω cm. When preparing the a-SiN_x films, we fixed the flow rates of SiH₄ (30 sccm) and N₂ (20.0 sccm) and varied the applied RF power to the reactant gas in the range of 100 W to 40 W. We denote the samples studied as 100 W, 60 W, and 40 W according to the decreasing order of the applied RF power. For the RF power range lower than 40 W, the effect of the RF power strength on the optical properties and film growth behavior has been reported in depth in our previous work.¹² We noted that the visible luminescence of the samples studied in the lower RF power range is mainly attributed to Si QDs embedded in SiN_x films.¹² The working pressure and the growth temperature were maintained at 7.5 × 10⁻¹ Torr and 350 °C, respectively.⁸

After the growth of the films, thermal annealing was carried out at 700 °C for 10 min under N₂ which is a somewhat lower temperature condition than the conventional thermal treatments for silicon related nanostructures. To measure the photoluminescence (PL) and photoluminescence excitation (PLE), we used a HORIBA® PL spectrometer (Fluorolog®-3) at RT with a He–Cd laser (325 nm, 25 mW) and a Xe lamp, respectively. The chemical states of the samples were analyzed using an X-ray photoelectron spectrometer (XPS) (VG Microtech) with a triple channel electron analyzer (CLAM2). The non-monochromatic Al Kα line (hν = 1486.6 eV) was used as an excitation source, and the analyzer pass energy was fixed at 20 eV. The binding energy was referenced using the peak position of the adsorbed hydro-carbon (C 1s = 284.6 eV). The overall resolution of the spectrometer was 1.3 eV, as determined by the full-width half maximum (FWHM) value of Au 4f_7/2 for a clean gold plate. Bright-field transmission electron microscopy (TEM) cross-section micrographs were obtained with a JEM-ARM 200F instrument operated at 200 kV.

Results and discussion

Fig. 1 shows the PL spectra for the annealed a-SiN_x samples (100 W, 60 W, and 40 W) in the spectral range from 2.41–1.78 eV. Decreasing the applied RF power from 100 W to 40 W causes the PL intensity to become stronger and the central PL peak position to shift toward a lower energy. Our group recently confirmed that the quantity and the type of defects in a-SiN_x can be controlled by varying the applied RF power. We also noted that visible light emission due to radiative defect states can be tuned between ∼1.9 eV and ∼2.4 eV.⁸ Here, the 100 W and 60 W samples represent PL spectra corresponding to the defect related emissions mentioned above. However, the PL peak at 1.78 eV in the 40 W sample is a new discovery. According to our previous study, no possible transition state corresponding to 1.78 eV exists among the energy levels for optical transitions of a-SiN_x. This raises a question as to the luminescence origin of the PL peak at 1.78 eV. In order to investigate the origin of this PL band at ∼1.78 eV, PLE experiments were performed.

Fig. 1 PL spectra for the samples fabricated using 100 W, 60 W, and 40 W RF power.

For the PLE measurements shown in Fig. 2(a)–(c), we first selected the emitted photon energies from the prominent peak positions of the PL spectra corresponding to each sample. Since the PL peaks for the 60 W and 40 W samples show strong luminescence over a wide spectral range, those PLE experiments were carried out by selecting two emitted photon energies as indicated in Fig. 2(b) and (c). Because the actual energy range of the Xe lamp light source is at most 4.6 eV, we mainly considered the transitions between the valence band top-edge and the intermediate gap states, which can be excited by incident photons from the Xe lamp.

Fig. 2 PLE spectra of the samples: (a) 100 W, (b) 60 W, and (c) 40 W.

The PLE spectra for the 60 W and 100 W samples show two peaks, while the spectrum for the 40 W sample displays only a single peak (Fig. 2). This indicates that the origin of PL at ∼1.78 eV in the 40 W sample is different from that of typical defect-related PL.

In order to confirm the structural variations brought about by changing the applied RF power in a-SiN_x films, all samples were subjected to additional TEM analysis. Fig. 3 depicts cross-sectional bright-field TEM images for the 100 W, 60 W, and 40 W samples. The TEM image of the 100 W sample in Fig. 3(a) shows tiny grains (black dots) with a uniform size distribution and placement. However, the grains in the 60 W sample are larger as shown in Fig. 3(b). When the applied RF power was set at 40 W, the formation of nano-sized amorphous Si QDs (nsa-Si QDs) with a size of about 2.8–3 nm was observed. According to the literature, PL energy corresponding to a nsa-Si QD size of ∼3.0 nm is about 1.8 eV,¹³ which is in good accord with the PL result shown in Fig. 1. From these TEM results, we attribute the strong visible RT-PL at 1.78 eV to the nsa-Si QDs in a-SiN_x.

Fig. 3 Cross-sectional bright-field TEM images of the samples: (a) 100 W, (b) 60 W, and (c) 40 W.

To better understand the formation process of the nsa-Si QDs in a-SiN_x samples, XPS analysis was also conducted on these samples. Fig. 4 represents the Si 2p XPS core-level spectra for all samples. The Si 2p peak is broad for all samples, indicating the presence of multiple components as expected. Thus, we deconvoluted the core-level spectra with the appropriate components according to the peak positions previously reported in the literature, as shown in Fig. 4.^8,14–16

Fig. 4 Si 2p XPS core-level spectra for the samples 100 W, 60 W, and 40 W.

We used the Doniach–Sunjic curves combined with Lorentzian and Gaussian functions after subtracting the background signals due to inelastic scattering by Shirley background subtraction procedure.^14,15 The Si 2p core-level spectrum shows five chemical states as shown in Fig. 4. The five peaks are located at 103.5 ± 0.1 eV, 102.2 eV, 101.4 eV, 100.8 eV, and 99.7 eV (denoted as S₍₁₎, S₍₂₎, S₍₃₎, S₍₄₎, and S₍₅₎, respectively). The S₍₁₎, S₍₂₎, and S₍₅₎ components are attributed to well-known chemical states of the native surface SiO₂, Si₃N₄, and bulk-Si, respectively.¹⁶ Subsequently, we identified the S₍₃₎ and S₍₄₎ components as the chemical states of ˙Si≡N₃ (silicon dangling bond) and N₃≡Si–Si≡N₃ (Si–Si bond) defects as discussed previously.⁸

Fig. 5 compares the integrated intensities of the Si 2p peaks of all the samples as a function of the applied RF power. This shows that the S₍₂₎ component decreases as the RF power decreases to 60 W from 100 W. This means that as the applied RF power decreases in this range, the amount of the stoichiometric SiN_x phase decreases. In addition, both the S₍₃₎ and S₍₄₎ components tend to increase as the applied RF power decreases to 60 W, with the enhancement of the S₍₄₎ component being especially remarkable. By decreasing the applied RF power from 100 W to 60 W, the S₍₄₎ component corresponding to the N₃≡Si–Si≡N₃ defect rises due to the deficiency of nitrogen. Our previous work reported that the energy level of the S₍₄₎ component is located above ∼4.2 eV from the valence band top-edge.⁸ Since the energy of the PL source we used is lower than the energy used to excite an electron from the valence band top-edge to S₍₄₎, the emission of 1.9 eV from N₃≡Si–Si≡N₃ to ˙Si≡N₃ could not be observed, although the excitation from the valence band top-edge to N₃≡Si–Si≡N₃ could be observed. As we can see from the PLE data from the 60 W sample in Fig. 2(b), the large emission at an excitation energy of ∼4.2 eV is consistent with the enhancement of the S₍₄₎ component. Once a sufficient quantity of S₍₄₎ component is present, the circumstances are suitable for the formation of nsa-Si QDs. In other words, when the number of N₃≡Si–Si≡N₃ defects in a-SiN_x crosses a certain value, Si clustering is favourable. By dropping the RF power further to 40 W, the S₍₄₎ component corresponding to N₃≡Si–Si≡N₃ defects decreases again and the appearance of S₍₅₎ corresponding to bulk-Si is obvious. These results illustrate the influence of the RF power on the formation mechanism of nsa-Si QDs in the early stages. First, we note that a lack of nitrogen in the a-SiN_x film stimulates the formation of ˙Si≡N₃ and N₃≡Si–Si≡N₃ defects as the applied RF power decreases to 60 W.¹⁷ When the applied RF power reaches a certain point between 60 W and 40 W, N₃≡Si–Si≡N₃ defects begin to change into nsa-Si QDs. The nsa-Si QDs in the 40 W sample are the source of the light emission at 1.78 eV. This indicates that the type and the quantity of defects are crucially related to the early stages of nsa-Si QD formation in a-SiN_x. In addition, at RF powers lower than 40 W, larger nsa-Si QDs seem to be formed in the SiN_x matrix due to enhancement of the Si-rich environment, which could have been hypothesized from the Si 2p core-level and PL spectra in our previous work.¹² Thus, we can readily select an appropriate combination of phases among quantum structures of defects and nsa-Si QDs in a-SiN_x by tuning only the RF power.

Fig. 5 Relative integrated intensities of the various chemical states of the Si 2p peak for the samples 100 W, 60 W, and 40 W. (See the text for a description of the components S₍₁₎, S₍₂₎, S₍₃₎, S₍₄₎ and S₍₅₎.)

Conclusions

We investigated the effect of applied RF power on optical and structural changes of a-SiN_x films. By decreasing the applied RF power from 100 W to 40 W, the PL from a-SiN_x becomes stronger and the central PL peak position shifts from 2.41 eV to 1.78 eV. In TEM analysis, as the applied RF power decreases, the size of the Si grains in a-SiN_x becomes larger, eventually forming nsa-Si QDs in the 40 W sample. By analyzing the chemical states of the Si 2p XPS core-level spectra, we found that N₃≡Si–Si≡N₃ defects begin to convert into nsa-Si QDs when the applied RF power reaches a certain point between 60 W and 40 W. This means that a certain RF power between 60 W and 40 W is a critical condition which changes the luminescence origin from defect-dominated to QD-dominated transition regimes. In addition, this indicates that the type and the quantity of defects in a-SiN_x play a crucial role in the initial formation of Si QDs in a-SiN_x. We propose that defect control through RF power tuning readily controls the formation mechanism of quantum structures from defects to Si QDs, and could facilitate the application of Si-based full-color LEDs or tandem solar cells using silicon nitride multilayers with multiple absorption and emission features.

Acknowledgements

We acknowledge support from the National Research Foundation of Korea under Grant No. NRF-2015M2B2A4032245 and NRF-2015R1D1A1A01060381.

References

1. D. Liang and J. E. Bowers, Nat. Photonics, 2010, 4, 511–517.
2. S. Jang, B. S. Joo, S. Kim, K. Kong, H. Chang, B. D. Yu and M. Han, J. Mater. Chem. C, 2015, 3, 8574–8581.
3. M. Zacharias, J. Heitmann, R. Scholz, U. Kahler, M. Schmidt and J. Bläsing, Appl. Phys. Lett., 2002, 80, 661–663.
4. L. Tsybeskov, K. D. Hirschman, S. P. Duttagupta, M. Zacharias, P. M. Fauchet, J. P. McCaffrey and D. J. Lockwood, Appl. Phys. Lett., 1998, 72, 43–45.
5. A. Zelenina, A. Sarikov, S. Gutsch, N. Zakharov, P. Werner, A. Reichert, C. Weiss and M. Zacharias, J. Appl. Phys., 2015, 117, 1753031–1753038.
6. G. Fu, X. Wang, H. Feng, X. Yu, W. Dai, W. Lu and W. Yu, J. Alloys Compd., 2013, 579, 284–289.
7. L. Zhang, H. Jin, W. Yang, Z. Xie, H. Miao and L. Ana, Appl. Phys. Lett., 2005, 86, 0619081–0619083.
8. S. Jang and M. Han, J. Alloys Compd., 2014, 614, 102–106.
9. M. Dasog, G. B. De los Reyes, L. V. Titova, F. A. Hegmann and J. G. C. Veinot, ACS Nano, 2014, 8, 9636–9648.
10. M. Dasog, Z. Yang, S. Regli, T. M. Atkins, A. Faramus, M. P. Singh, E. Muthuswamy, S. M. Kauzlarich, R. D. Tilley and J. G. C. Veinot, ACS Nano, 2013, 7, 2676–2685.
11. L. S. Zambom, R. D. Mansano and R. Furlan, Vacuum, 2002, 65, 213–220.
12. S. Jang, S. Jung, E. Choi, M. Han and J. Lee, J. Korean Phys. Soc., 2011, 59, 2334–2337.
13. N. Park, C. Choi, T. Seong and S. Park, Phys. Rev. Lett., 2001, 86, 1355–1357.
14. S. Doniach and M. Sunjic, J. Phys. C: Solid State Phys., 1970, 3, 285–291.
15. D. A. Shirley, Phys. Rev. B: Solid State, 1972, 5, 4709–4714.
16. P. Cova, S. Poulin, O. Grenier and R. A. Masut, J. Appl. Phys., 2005, 97, 0735181–07351810.
17. C. D. Valentin, G. Palma and G. Pacchioni, J. Phys. Chem. C, 2011, 115, 561–569.

---

Footnote

† Present address: Chemical Infrastructure Division, Korea Research Institute of Chemical Technology (KRICT), Daejeon, 305-343, Republic of Korea.

---