Evolution of photoluminescence mechanisms of Si+-implanted SiO2 films with thermal annealing

L. Ding, T. P. Chena), Y. Liu, C. Y. Ng, M. Yang, and J. I. Wong School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798 F. R. Zhu, and M. C. Tan Institute of Materials Research and Engineering, Singapore 117602 S. Fung, X. D. Chen, and Y. Huang Department of Physics, The University of Hong Kong, Hong Kong

1

ABSTRACT

The information of band structure of silicon nanocrystal (nc-Si) embedded in SiO2 thin films synthesized by Si+ implantation and subsequent thermal annealing at various temperatures has been obtained from spectroscopy ellipsometric (SE) analysis. The indirect band structure and the band gap (~1.76 eV) of the nc-Si are not affected by the annealing. In contrast, the photoluminescence (PL) spectra show a continuous evolution with the annealing. Six PL bands located at 415, 460, 520, 630, 760, and 845 nm have been observed depending on the annealing temperature. The annealing at 1100 oC yields the strongest PL band at 760 nm (~ 1.63 eV) with an intensity much higher than that of all the other PL bands. Based on the knowledge of the band structure, the 760nm-PL band can be attributed to the indirect band-to -band transition of the nc-Si assisted by the Si-O vibration of the nc-Si/SiO2 interface with the stretching frequency of ~1083 cm-1 (~ 0.13 eV). On the other hand, it is shown that the first four PL bands mentioned above could originate from different extended defects in the oxide matrix, while the 845-nm PL band could be related to the interface luminescent centers.

a) Electronic mail: [email protected]

2

Enormous interest in fabricating light emitting devices (LED) based on nanostructured silicon has been triggered since the discovery of photoluminescence (PL) from porous silicon in 1990 [1]. The development of Si-based nanostructured materials for light emitting devices (LED) has provided an opportunity to incorporate Si-based optoelectronic devices into the matured Si integrated circuit (IC) technology. Among all types nanostructured silicon, porous silicon has received the most intensive research attention since 1990. However, PSi is a fragile and environmentally sensitive type of nanostructured silicon. Therefore, it is not reliable to utilize porous silicon in optoelectronic devices because of its instability in light emission [2], structural fragility and incompatibility with mainstream CMOS technology [3]. Silicon nanocrystals (nc-Si) are considered as a preferable strategy for overcoming these challenges. Various techniques have been employed to synthesize nc-Si, including chemical vapor deposition (CVD) [4, 5], sputtering [6, 7], pulse laser deposition (PLD) [8, 9], and silicon ion implantation into SiO2 [3, 10-14]. Among all these techniques, silicon ion implantation into a SiO2 matrix followed by high temperature annealing is considerer as one of the most promising methods for producing chemically and electrically stable nc-Si.

Although efficient room temperature photoluminescence (PL) from SiO2 films embedded with nc-Si synthesized by ion implantation technique has been reported by many research groups [3, 10-14], the mechanism of PL from such materials is still under debate. For example, for the system synthesized with Si ion implantation, the visible-light (in particular the red-light) luminescence was attributed to the quantum confinement effect of the nc-Si [3, 13]. However, Liao et. al found that the short-wavelength luminescence was related to the presence of the oxygen vacancies in the Si+-implanted SiO2 films [14]. Thanks to the large amount of previous studies, it is certain that there are multiple mechanisms responsible for the visible luminescence. However, to the best of our knowledge, few investigations have

3

been carried out on the mechanism evolution of photoluminescence due to thermal annealing. On the other hand, to assert that the light emission is due to defects or quantum confinement effect, the information of band structure including the band gap of nc-Si is necessary. Many theoretical calculations on the band gap and optical properties of silicon nanocrystals have been reported [15-21], but there are only few experimental studies [22-24]. In this work, the evolution of PL of Si+-implanted SiO2 thin films under different annealing conditions is investigated. Based on the information of the nc-Si band structure obtained from spectroscopy ellipsometric (SE) analysis [22-24], the evolution of PL mechanisms with the annealing has been examined.

550-nm SiO2 films were grown on p-type Si wafers with (100) orientation by thermal oxidation in dry oxygen. A dose of 1×1017 cm-2 of Si ions were implanted into the SiO2 films at the energy of 100 keV at room temperature. Afterwards, postimplantation thermal annealing was carried out in nitrogen ambient at various temperatures ranging from 500 to 1100 oC (500, 600, 700, 850, 900, 1000, 1100 oC). No significant change in the nc-Si size under different annealing temperatures was observed from the x-ray diffraction (XRD) measurement. This is mainly due to the very low diffusion coefficient of Si in SiO2 films [25]. It was observed that a high temperature annealing at 1100 oC for 16 hours induced only an increase of 0.5 nm in the nc-Si size as compared to that of as-implanted sample [25]. Both XRD and high resolution transmission electron microscopy (HRTEM) measurements show the formation of nc-Si with a mean size of ~4.5 nm embedded in the SiO2 matrix.

As an extension of our previous works [22-24], the band gap and absorption coefficient of nc-Si under different annealing conditions were determined using spectroscopic ellipsometry. The band gap of nc-Si obtained for various annealing temperatures is shown in

4

Fig. 1. As can be seen, the nc-Si shows a large expansion in band gap as compared to the bulk crystalline silicon, i.e., a band gap expansion of ~0.6 eV is observed. It can be observed in Fig. 1 that the nc-Si band gap is actually not affected by the annealing. This could be due to the fact that the annealing at various temperatures does not cause a significant change in the nc-Si size. The nc-Si band gap obtained in this work is in very good agreement with the first-principle calculation of the optical gap of silicon nanocrystals based on quantum confinement. For the nc-Si size of 4.5 nm, the nc-Si band gap due to the quantum confinement is 1.74 eV [22], which is very close to the average band gap of 1.76 eV shown in Fig. 1. On the other hand, based on the plot of (αE)γ versus E, where α is the absorption coefficient of the nc-Si and E is the photon energy, one can examine whether the nc-Si is a direct ( γ =2) or indirect ( γ =1/2) band gap semiconductor [26]. As an example, Fig. 2(a) shows the Tauc plot ( γ =1/2) of nc-Si annealed at 600 oC, and the plot for bulk crystalline silicon is also included for comparison. With γ =1/2, a linear relationship in the photon energy range near the absorption edge is observed for both the nc-Si and crystalline silicon, as shown in Fig. 2(a). In contrast, one can see in Fig. 2(b) that the linearity is much poorer when γ = 2 for both nc-Si and bulk crystalline silicon. This suggests that the nc-Si has an indirect band gap structure. The absorption edge of the nc-Si yielded from the Tauc plot is ~1.75 eV, being consistent with the value shown in Fig. 1. Furthermore, the Tauc plot is very similar for different annealing. Therefore, one can conclude that the annealing does not change the indirect band gap structure and the band gap.

Intense visible light emission from our samples was observed with the excitation wavelength of 325 nm at room temperature. All the samples show a very broad PL spectrum, suggesting that it may be composed of several different PL bands. Figure 3 shows the PL spectra for the samples at the annealing temperatures ranging from 500 to 1100 oC. The PL 5

spectrum changes significantly when the annealing temperature increases from 500 to 1100 o

C. It is found that the PL spectra of the samples annealed at 500, 600, and 700 oC have a

similar shape while the PL spectra with annealing at above 700 oC show some different traits. Particularly, the PL peak intensity of the 1100 oC-annealed sample is almost 30 times stronger than the ones with annealing below 1100 oC. For a further investigation, as shown in Fig. 3, the PL spectra of all samples were decomposed into several Gaussian-shaped peaks centered at different wavelengths of 415 nm, 460 nm, 520 nm, 630 nm, 760 nm, and 845 nm using a fitting procedure. Figure 4 shows the annealing-temperature dependence of the integrated PL intensity for each band. As can be observed in this figure, the 415-, 460-, and 520-nm PL bands have a similar behavior, i.e., the integrated intensity first increases slightly and then decreases dramatically with the increasing annealing temperature. The integrated intensity for the 415-, 460-, and 520-nm bands reaches a maximum at 700, 600 and 700 oC, respectively. The above three bands almost disappear when the annealing temperature exceeds 900 oC. The 630-nm band always exists for annealing temperatures up to 1000 oC, and no significant change in the intensity is observed with increasing annealing temperature. In contrast, the integrated PL intensity of both the 760- and 845-nm bands continuously increases with the annealing temperature. Their integrated intensity increases slightly with the annealing temperature up to 1000 oC and then experiences a dramatic increase by approximately 30 times when the annealing temperature reaches 1100 oC.

As discussed early, the annealing in the temperature range of 500 – 1100 oC does not change the indirect band gap structure and the band gap of the nc-Si. This contrasts with the very large changes in the PL discussed above. Therefore, one may conclude that not all the PL bands observed in this study originate from the quantum confinement effect of the nc-Si and some PL bands can be ascribed to the oxide matrix. It is well known that various defects

6

in the oxide matrix can serve as visible luminescent centers. In the following discussions, the PL bands at 415, 460, 520 and 630 nm are attributed to the defects in the oxide matrix. As for the disappearance or steep decrease of these four PL bands under annealing temperature of 1100 oC, it is probably because that such a high annealing temperature can give rise to the elimination of most of the luminescent defects in the oxide.

It has been reported that the weak oxygen bond (WOB) defects in silicon oxide can contribute to the 415-nm (~3eV) PL band [27-29]. The WOB defects can be generated in the Si+ implanted SiO2 layers after suitable thermal annealing. The oxygen vacancies are induced by the displacement of oxygen caused by the Si+ implantation into a normal SiO2 network, and the oxygen interstitials, which are considered as the precursors of the WOB defects, are created

concurrently.

The

reaction

can

be

represented

by

O3 ≡ Si-O-Si ≡ O3 → O3 ≡ Si-Si ≡ O3 +Ointerstitial . The oxygen interstitials will change into the WOB

defects

immediately

after

annealing,

as

described

by

the

reaction

Ointerstitial + Ointerstitial → O-O , which is an reversible reaction under excessive thermal annealing energies. This can be indicated by the decrease of the 415-nm PL band with the increase of annealing temperature when the temperature is higher than 700 oC. The 460-nm (~2.7 eV) PL band has been studied on pure silica glass [30] and Si-rich silicon oxides [14, 27, 31]. The 460-nm PL band has been shown to be associated with the neutral oxygen vacancy (NOV) defect represented by O3 ≡ Si-Si ≡ O3 [14, 27, 31]. The NOV defects could be observed in ion-implanted or radiation-damaged SiO2 networks. As for the Si+-implanted SiO2 film, the NOV defects are actually generated in the same process as for the WOB defects, as described above. Therefore, it is not surprising that the 415- (WOB) and 460-nm (NOV) PL bands have a very similar dependence on the annealing temperature. The origin of PL at 520 nm is thought to be associated with the Eδ' defect [27, 32]. In Si+-implanted SiO2, Eδ' defects 7

can be generated by both extrinsic ion-implantation-induced dissociation process [27, 34] and intrinsic UV photon absorption in the PL measurement itself [33]. The generation of Eδ' defects caused by ion-implantation is a transformation of an NOV defect to an Eδ' center, as described by O3 ≡ Si-Si ≡ O3 → O3 ≡ Si •+ Si ≡ O3 + e− [27, 34]. As can be seen in Fig. 4, the

WOB-related PL peak at 415 nm, the NOV-related PL peak at 460 nm, and the Eδ' -related PL peak at 520 nm are observed in the same temperature range and have a similar annealing-temperature dependence. This indicates that the extrinsic ion-implantation -induced dissociation process plays and important role in the generation of Eδ' defects. The luminescence of another radiative defect, the non-bridging oxygen hole center (NBOHC) which has been observed and confirmed in both pure silica glass and ion-irradiated SiO2 [35-38], is in accordance with the 630-nm PL band observed in this study. As shown in Fig. 4(d), the 630-nm PL band intensity changes little with increase of the annealing temperature, indicating that the number of NBOHCs is very stable during thermal annealing when the annealing temperature is below 1100 oC. The steep decrease of the intensity of 630-nm PL band at the annealing temperature of 1100 oC suggests that NBOHCs can be largely reduced at the temperature.

On the other hand, the 760 and 845-nm PL bands both have a distinct annealing behavior, suggesting that these two bands have luminescence mechanisms different from the defects in the oxide matrix discussed above. Here the 760-nm PL band is attributed to the indirect band to band transition assisted by the Si-O vibration of the nc-Si/SiO2 interface. Note that there is a difference of ~0.13 eV between the energy of the 760-nm band (~1.63 eV) and the nc-Si band gap (~1.76 eV) (see Fig. 5). This energy difference is the same as the energy (~ 0.13 eV) of the Si-O vibration with a stretching frequency of ~1083 cm-1 in the

8

system of nc-Si embedded in SiO2 [39], implying the important role of the Si-O vibration at the nc-Si/SiO2 interface. Considering the energy conservation and the indirect band structure of the nc-Si, we therefore are inclined to believe that the Si-O vibration at the nc-Si/SiO2 interface provides the mean required for both the energy dissipation due to the energy conservation requirement and the momentum conservation in the PL process. In other words, the 760-nm band can be attributed to the band-to-band indirect transition of the nc-Si assisted by the Si-O vibration at the nc-Si/SiO2 interface. Fig. 5 shows the maximum intensity of each PL band. It can be seen that the intensity of the 760-nm band is more than 30 times stronger than that of the 415, 460, 530, 630 nm PL bands. For Si+-implanted SiO2, the 1100 oC annealing could lead to both ultra-fine nc-Si and the desirable nc-Si/SiO2 interface such that the probability of the band-to-band transition assisted by the Si-O vibration is large leading to a strong light emission at 760 nm. As for the 845-nm band, it mostly originates from the localized luminescent centers at the nc-Si/SiO2 interface. This is consistent with the study of Allan et.al, in which they attributed the 845-nm (~1.5-eV) PL to the luminescent surface state of nc-Si [40]. This conclusion was later supported by Zhuravlev et.al who also observed the 1.5-eV PL band from Si+ implanted SiO2 after high temperature treatment [41].

To further confirm the origin of the main PL peak (e.g. the 760 nm band) which is attributed to the band-to-band indirect transition of the nc-Si assisted by the Si-O vibration at the nc-Si/SiO2 interface, we have fabricated four samples with different nc-Si size using the Si+ implantation and subsequent annealing. The implantation details of the four samples are given as follows: (a) 3×1016 ions/cm2 at 120 keV, (b) 1×1017 ions/cm2 at 100 keV, (c) multiple implantations with 5×1015 ions/cm2 at 20 keV, 1×1016 ions/cm2 at 40 keV, and 2×1016 ions/cm2 at 80 keV, and (d) 8×1016 ions/cm2 at 5 keV. After implantation, the samples were annealed at 1100 oC for 20 min in the atmosphere of N2 to induce nanocrystallization.

9

XRD measurements show the average nc-Si sizes for the four samples are ~4 nm, ~4.5 nm, ~4.9 nm, and 6.5nm, respectively. This result can be explained by the concentration of excess Si in SiO2. The average volume fraction of excess Si throughout the region containing nc-Si were calculated to be 3%, 8%, 11%, and 80%, respectively. Excess Si with a higher concentration in SiO2 forms a larger nanocrystal size after a high temperature annealing. The band gaps of the four samples obtained from the SE study are 1.85 eV, 1.76 eV, 1.70 eV, and 1.53 eV, respectively. The room temperature PL spectra were taken using the same excitation source as mentioned previously. The PL of the four samples is similar in their spectrum shapes, but the position of the main peak is different for different samples, as shown in Fig.6. .The main peaks of the four samples are located at 720 nm (1.72 eV), 760 nm (1.63 eV), 785 nm (1.58 eV), and 890 nm (1.39 eV), respectively. This means that the PL peak shifts to a lower energy when the concentration of excess Si increases, which is consistent with the result reported in [10]. For all the four samples, the difference between the band gap and the corresponding PL peak energy is always 0.13 ± 0.01 eV. This energy difference is actually equal to the Si-O vibration energy. Although both the band gap and the PL peak energy changes with the nc-Si size, the energy difference remains unchanged. This strongly supports the suggestion that the main PL peaks of the annealing at 1100 oC is due to the band-to-band indirect transition of the nc-Si assisted by the Si-O vibration at the nc-Si/SiO2 interface.

In summary, the band structure of nc-Si embedded in SiO2 matrix at different temperatures has been studied with spectroscopy ellipsometry. It is found that the annealing does not change the indirect band structure and the band gap (~1.76 eV) of the nc-Si. In contrast, the photoluminescence shows a continuous evolution with the annealing. Six PL bands located at 415, 460, 520, 630, 760, and 845 nm are observed depending on the

10

annealing temperature. The annealing at 1100 oC yields the strongest PL band (i.e., the main PL peak) at 760 nm (~ 1.63 eV) with intensity much higher than that of all other PL bands. Both the band gap and the main PL peak change with the nc-Si size. However, the difference between the band gap and the PL peak energy is always 0.13 ± 0.01 eV, being equal to the Si-O vibration energy. This strongly suggests that the main PL peak is due to the band-to-band indirect transition of the nc-Si assisted by the Si-O vibration at the nc-Si/SiO2 interface. On the other hand, the 415-, 460-, 520- and 630-nm bands are ascribed to the WOB, NOV, Eδ' and NBOHC defects, respectively, while the 845-nm band is proposed to be related to the interface luminescent centers.

11

References

[1] L. T. Canham, Appl. Phys. Lett. 57, 1046 (1990). [2] M. A. Tichler, R. T. Collins, J. H. Stathis, and J. C. Tsang, Appl. Phys. Lett. 60, 639 (1992). [3] P. Mutti, G. Ghislotti, S. Bertoni, L. Bonoldi, G. F. Cerofolini, L. Meda, E. Grillo, and M. Guzzi, Appl. Phys. Lett. 66, 851 (1994). [4] J. F. Tong, H. L. Hsiao, and H. L. Hwang, Appl. Phys. Lett. 74, 2316 (1999). [5] N. -M. Park, T. –S. Kim, and S. J. –Park, Appl. Phys. Lett. 78, 2575 (2001). [6] S. Furukawa and T. Miyasato, Jpn. J. Appl. Phys. 27, L2207 (1988). [7] Q. Zhang, S. C. Bayliss, and R. A. Hutt, Appl. Phys. Lett. 66, 1977 (1995). [8] T. Yoshida, S. Takeyama, Y. Yamada, and K. Mutoh, Appl. Phys. Lett, 68, 1772 (1996). [9] X. Y. Chen, Y. F. Lu, Y. H. Wu, B. J. Cho, M. H. Liu, D. Y. Dai, and W. D. Song, J. Appl. Phys. 93, 6311 (2003). [10] T. S-Iwayama,, N. Kurumado, D. E. Hole, and P. D. Townsend, J. Appl. Phys. 83, 6018 (1998). [11] H. M. Cheong, W. Paul, S. P. Withrow, J. G. Zhu, J. D. Budai, C. W. White, and D. M. Hembree, Appl. Phys. Lett. 68, 87 (1996). [12] S. Guha, M.D. Pace, D.N. Dunn, and I. L. Singer, Appl. Phys. Lett. 70, 1207 (1997). [13] T. Komoda, J. Kelly, F. Cristiano, A. Nejim, P. L. F. Hemment, K. P. Homewood, R. G.william, J. E. Mynard, and B. J. Sealy, Nucl. Instrum. Methods Phys. Res. Sect. B. 96, 387 (1995). [14] L. S. Liao, X. M. Bao, X. Q. Zheng, N. S. Li, and N. B. Min, Appl. Phys. Lett. 68, 850 (1996). [15] L.-W. Wang and Alex Zunger, Phys. Rev. Lett. 73, 1039 (1994). [16] C. Delerue, G. Allan, and M. Lannoo, Phys. Rev. B 48, 11024 (1993).

12

[17] H.-Ch. Weissker, J. Furthmüller, and F. Bechstedt, Phys. Rev. B 65, 155327 (2002). [18] H.-Ch. Weissker, J. Furthmüller, and F. Bechstedt, Phys. Rev. B 67, 165322 (2003). [19] H.-Ch. Weissker, J. Furthmüller, and F. Bechstedt, Phys. Rev. B 65, 155328 (2002). [20] I. Vasiliev, S. Ogut, and J. R. Chelikowsky, Phys. Rev. Lett. 86, 1813 (2001). [21] S. Ogut, J. R. Chelikowsky, and S. G. Louie, Phys. Rev. Lett. 79, 1770 (1997). [22] L. Ding, T. P. Chen, Y. Liu, C. Y. Ng, and S. Fung, Phys. Rev. B 72, 125419 (2005). [23] L. Ding, T. P. Chen, Y. Liu, C. Y. Ng, Y. C. Liu, and S. Fung, Appl. Phys. Lett. 87, 121903 (2005). [24] L. Ding, T. P. Chen, Y. Liu, C. Y. Ng, and S. Fung, Nanotechnology, 16, 2657 (2005). [25] M. Lopez, B. Garrido, C. Bonafos, A. Perez-Rodriguez, and J. R. Morante, Solid-State Electronics. 45, 1495 (2001). [26] J. I. Pankove, “Optical process in semiconductors,” (Dover, New York, 1975). [27] Chun-Jung. Lin, Chao-Kuei Lee, E. Wei-Guang Diau, and Gong-Ru Liu, J. Electrochem. Soc. 152, E25 (2006). [28] H. Nishikawa, E. Watanabe, D. Ito, M. Takiyama, A. Leki, and Y. Ohki, J. Appl. Phys. 78, 842 (1995). [29] J. C. Cheang-Wong, A. Oliver, J. Poiz, J. M. Hernandez, L. Rodriguez-Fernandez, J. G. Morales, and A. Crespo-Sosa, Nucl. Instrum. Methods Phys. Res. B 175, 490 (2001). [30] R. Tohmon, Y. Shimogaichi, H. Mizuno, and Y. Ohki, Phys. Rev. Lett. 62, 1388 (1989). [31] H. S. Bae, T. G. Kim, C. N. Whang, S. Im, J. S. Yun, and J. H. Song, J. Appl. Phys. 91, 4078 (2002). [32] H. Nishikawa. R. E. Stahlbush, and J. H. Stathis, Phys. Rev. B. 60, 15910 (1999). [33] J. H. Stathis and M. A. Kastner, Phys. Rev. B. 29, 7079 (1984). [34] H. Nishikawa, E. Watanabe, D. Ito, M. Takiyama, A. Leki, and Y. Ohki, J. Appl. Phys. 78, 842 (1995).

13

[35] L. Skuja, Solid State Commun. 84, 613 (1992). [36] S. Munekuni, T. Yamanaka, Y. Shimogaichi, R. Tohmon, Y. Ohki, K. Nagasawa, and Y. Hama, J. Appl. Phys. 68, 1212 (1990). [37] T. Bakos, S. N. Rashkeev, and S. T. Pantelides, IEEE Transaction on Nuclear Science, 49, 2713 (2002). [38] M. Ya. Valakh, V. A. Yukhimchuk, V. Ya. Bratus, A. A. Konchits, P. L. F. Hemment, and T. Komoda, J. Appl. Phys. 85, 168 (1999). [39] Y. Liu, T. P. Chen, Y. Q. Fu, M. S. Tse, P. F. Ho, J. H. Hsieh, and Y. C. Liu, J. Phys. D 36, L97 (2003). [40] G. Allan, C. Delerue, and M. Lannoo, Phys. Rev. Lett. 76, 2961 (1996). [41] K. S. Zhuravlev, A. M. Gilinsky, and A. Yu. Kobitsky, Appl. Phys. Lett. 73, 2962 (1998).

14

Figure captions:

FIG. 1. Band gap of nc-Si embedded in SiO2 versus annealing temperature.

FIG. 2. (a) Plots of (α E )

1/ 2

versus photon energy (E) for both the nc-Si under the annealing

at 600 oC and bulk crystalline silicon. (b) Plots of (α E ) versus photon energy (E) for both 2

the nc-Si under the annealing at 600 oC and bulk crystalline silicon.

FIG. 3. Decomposition of PL spectra for the samples annealed at (a) 500 oC, (b) 600 oC, (c) 700 oC, (d) 850 oC, (e) 900 oC, (f) 1000 oC, and (g) 1100 oC.

FIG. 4. Integrated intensity of each PL band as a function of annealing temperature.

FIG. 5. Maximum integrated intensity for each PL band.

FIG. 6. Normalized PL spectra for the following samples which were annealed at 1100 oC for 20 minutes: (a) 3×1016 ions/cm2 at 120 keV, (b) 1×1017 ions/cm2 at 100 keV, (c) multiple implantations with 5×1015 ions/cm2 at 20 keV, 1×1016 ions/cm2 at 40 keV, and 2×1016 ions/cm2 at 80 keV, and (d) 8×1016 ions/cm2 at 5 keV.

15

2.0 (a)

Band gap (eV)

1.8 1.6 1.76 eV (nc-Si embedded in SiO2) 1.4 1.2 1.0 1.12 eV (bulk crystalline silicon) .8 500

600

700

800

900

1000 o

Annealing temperature ( C) FIG. 1.

16

1100

140 (a) 120

1/2

1/2

(αE) (eV cm

-1/2

)

nc-Si, γ=1/2 Bulk crystalline Si, γ=1/2

100 80 60 40

1.12 eV

1.75 eV

20 0

1.0

1.2

1.4

1.6

1.8

2.0

2.2

1.8

2.0

2.2

X Data

nc-Si, γ=2 Bulk crystalline Si, γ=2

-2

(αE) (eV cm )

20x106

(b)

2

2

15x106

10x106

5x106

0 1.0

1.2

1.4

1.6

Photon energy (eV) FIG.. 2.

17

3x106

(a) 500 oC

2x106 1x106 0 (b) 600 oC

3x106 2x106

PL intensity (arb. units)

1x106 0 3x106

o

(c) 700 C

2x106 1x106 0 2x106

(d) 850 oC

1x106 0 2x106

(e) 900oC

1x106 0 3x106

o

(f) 1000 C

2x106 1x106 0 100x10

(g) 1100 oC 6

50x106 0 400

500

600

700

Wavelength (nm) FIG. 3.

18

800

900

(a)

415 nm

Integrated PL intensity (arb. units)

(b)

460 nm (c)

520 nm (d)

630 nm 760 nm

500

600

700

800

o 900 1000 ( C)

845 nm

500

500

600

700

600

800

(e)

(f)

o 900 1000 ( C)

700

800

900

1000

Annealing temperature (oC) FIG. 4.

19

1100

Max. intensity of PL band (arb. units)

band gap of nc-Si (1.76 eV)

3 10

102

415 nm (700 oC)

460 nm (600 oC)

0.13 eV

10 3.0

2.8

845 nm (1100 oC)

630 nm (700 oC)

520 nm (700 oC)

760 nm (1100 oC)

2.6

2.4

2.2

2.0

1.8

Photon energy (eV) FIG. 5.

20

1.6

1.4

Normalized PL intensity (arb. units)

(a)

1.0

(b) (c)

(d)

.8 .6 .4 .2

0.0 600

700

800

900

Wavelength (nm)

FIG. 6.

21

1000