Rod-like Precipitates Formed in Vapor-deposited Fe-Si Film

Semiconducting β-FeSi2 is a potential candidate material for solar cells. Various fabrication methods have therefore been proposed for smart films of this material. However, the dynamics of β-FeSi2 formation are not fully understood and require investigation. Our experimental results based on TEM and SEM observations imply that the mechanism for forming iron silicide is very complex, and exhibits strong dependence on the fabrication method. Rod-like precipitates form in a sample fabricated with double iron deposition and no precipitates form in a sample fabricated with single iron deposition on a silicon substrate.


Introduction
Semiconducting β-FeSi2 is a potential candidate material for thin film photovoltaic power generation, Si-based light emitters, and optical fiber photodetectors because it has a band gap of around 0.8 eV [1,2] and high optical absorption of 10 5 cm -1 [3].Several fabrication methods have been proposed which focus on growing β-FeSi2 films on a Si substrate, such as ion beam synthesis [4], molecular beam epitaxy [5][6], and sputter deposition [7], and the detailed dynamical mechanisms that occur on the Si substrate are thus important for obtaining a smart film of β-FeSi2.In this study, we therefore focus on the dynamics by mainly examining Fe/Si interfaces through the use of transmission electron microscopy (TEM) and scanning electron microscopy (SEM).

Experimental Procedures
We prepared two samples, samples A and B, for observation and quantitative evaluation of Fe/Si interfaces.For sample A, 49 nm of iron was first deposited onto thin Si(100) using vacuum deposition equipment as shown in Fig. 1.However, we had no specific surface treatments on Si; thus the presence of very thin SiO2 layer on Si surface can not be denied.After this deposition, the sample was annealed at 873 K for 18 ks and then 254 nm of iron was deposited onto the annealed sample again.After the second deposition of iron, sample A was annealed at 1073 K for 72 ks in vacuum again.Sample B was prepared by deposition of 304 nm of iron onto thin Si(100) followed by annealing at 1073 K for 72 ks.We utilized a high-resolution TEM (JEM2010, JEOL) and STM(JEOL JEM-ARM200F) for detailed observation of the Fe/Si interfaces, and employed a JSM-7500FA (JEOL) system for SEM observation and EDX analysis.

SEM Observations and EDX Analysis
Figure 2 shows an SEM image of sample A. The observed thickness of the iron layer is less than 300 nm and the Fe/Si interface is indistinct, which suggests active diffusion of both Fe and Si atoms because of the long annealing time (72 ks) and high annealing temperature (1073 K).Furthermore, the outward surface of the iron layer that faces away from the Si side appears relatively flat with no peculiar crystal growth.

TEM Observations
Figure 5(a) shows a TEM image of sample A. Precipitates as indicated by white ellipses are observed on the Si side and are filamentous instead of rod-like.The lengths are around 800 nm.In addition, the arrangements of these precipitates are not random, but are quite highly regular with the angles between precipitates being 0 (parallel) or about 75.As can be seen in Fig. 5(a), a noprecipitation zone is observed.This implies that precipitates are formed by a diffusion mechanism and that the diffusion distance of iron is about 1000 nm based on the distance between the Fe sites and the no-precipitation zone.Figure 5(b) shows a magnified image of Area A. From this, we estimate the width of the precipitation zone to be about 50 nm.We tried to analyze the crystalline structure of precipitates using TEM electron diffraction patterns (TED).However experiments on TED show the pattern of Si because of very small size of precipitates ( width ～50nm) ; thus it seems difficult to determine the precipitates.

011402-3
to distortion suggests a strain field, which implies coherent precipitation.○ 3 .Figure 7(b) shows a magnified view of domain ○ 3 , from which we estimate the length of prong ○ 3 to be around 40 nm.Based on the results of EDX composition analysis, domain ○ 1 is mainly iron (Fe 95.3 at%, Si 4.7 at%), domain ○ 2 is an iron-silicon compound (Fe 62.4 at%, Si 37.6 at%), probably Fe5Si3, and domain ○ 3 is almost entirely Si (Fe 4.4 at%, Si 95.6 at%).Considering that domain ○ 2 is an iron-silicon compound that forms the interface between Si and iron, the Si prong in domain ○ 3 may be formed by the reaction between iron-silicon compound (i.e., domain ○ 2 ) and the Si matrix.However it is difficult to determine the precise evaluation for each domain because TEM-EDX results are dependent on penetration depth of electron beam; thus more detailed analysis on the basis of TED will be required for the purpose of determining the atomic arrangement and/or crystalline structure of each domain.

Analysis
Focusing on the experimentally determined differences between samples A and B, no precipitation is observed in sample B while many filamentous or rod-like precipitates are observed in sample The difference between the samples is the fabrication method.Iron is deposited onto the Si surface twice in sample A, whereas it is only deposited once onto the Si surface in sample B. The iron that is deposited first in sample A thus appears to play an important role in determining the formation of precipitates.Since iron atoms are presumably thought to disperse into the Si matrix extensively during the first annealing step (873K for 18ks), this diffused iron atoms may act as nuclei for the formation of precipitates.In the second deposition of iron and subsequent annealing process, iron atoms cohere around the iron nuclei, resulting in the formation of precipitates.In contrast, as shown in Fig. 4, large Fe prongs form on the side of the Fe facing away from the Si matrix, which leads to a relative decrease in the amount of Fe in the Si matrix.The formation of precipitates appears to be difficult in sample B due to this decrease in the amount of Fe in the Si matrix.However rod-like precipitates may be observed in sample B if iron atoms are secondly deposited on the side of the Fe; thus experiments that determine the relation between the formation of precipitates in Si matrix and deposition number of times of iron will be required.
Next let us consider the formation mechanism of precipitates as shown in Fig. 7.We can easily mention intermetallic compound ε-FeSi if precipitations are formed as the result of Fe/Si interface reaction.However, the experimental result that precipitates exist in the range relatively away from Fe/Si interface suggests more complicated mechanism for the formation.Though diffusion mechanism is presumably reasonable mechanism to explain, the presence of very thin SiO2 layer on Si surface as stated in the section of experimental procedures will inhibit the motion of Fe in the annealing process of 72ks at 1073K; thus more detailed mechanism remains unknown.Consequently precise analysis determining crystalline structure and atomic arrangements seems first step for explaining observed results.

Conclusion
Through the experimental results of samples A and B using SEM, EDX analysis, and TEM, we conclude that no precipitates are found on the Si side in sample B. This difference is presumably due to the difference in fabrication methods.Sample A was treated by annealing and two sets of iron deposition whereas sample B was treated with only one annealing procedure and iron deposition.In sample A, because of the diffusion of iron atoms caused by the first annealing process, many iron atoms are thought to migrate from iron sites into the Si matrix where they act as fine nuclei thus 011402-5 forming precipitates as show in Fig. 5.This may be promoted by the second annealing process.In contrast, in sample B, the presence of Fe prongs is thought to decrease the amount of Fe in the Si matrix.Because of this, the formation of precipitates appears to be difficult in sample B. However, the reason why this rod-like precipitates with specific orientation is formed remains unknown.Further detailed analysis is required.

Figures 3 (
Figures 3(a), (b), and (c) respectively show an SEM image, Si K-line image, and Fe K-line image of the Fe/Si interface in sample A. The Fe K-line image in Fig. 3(c) indicates a large amount of iron atoms dispersed in the Si matrix.This dispersion of iron atoms is thought to be due to the long annealing time and high annealing temperature.

Figures 4 (Fig. 1 .Fig. 2 .
Figures 4(a), (b), and (c) respectively show an SEM image, Si K-line image, and Fe K-line image of the Fe/Si interface in sample B. As can be seen Fig. 4(a), sample B exhibits peculiar crystal growths (prongs) on the iron surface facing away from the Si, which contrasts with the lack of

Fig. 3 .Fig. 4 .
Figure5(a) shows a TEM image of sample A. Precipitates as indicated by white ellipses are observed on the Si side and are filamentous instead of rod-like.The lengths are around 800 nm.In addition, the arrangements of these precipitates are not random, but are quite highly regular with the angles between precipitates being 0 (parallel) or about 75.As can be seen in Fig.5(a), a noprecipitation zone is observed.This implies that precipitates are formed by a diffusion mechanism and that the diffusion distance of iron is about 1000 nm based on the distance between the Fe sites and the no-precipitation zone.Figure5(b) shows a magnified image of Area A. From this, we estimate the width of the precipitation zone to be about 50 nm.We tried to analyze the crystalline structure of precipitates using TEM electron diffraction patterns (TED).However experiments on TED show the pattern of Si because of very small size of precipitates ( width ～50nm) ; thus it seems difficult to determine the precipitates.Furthermore, as shown in this figure (see arrow), contrast due

Figure 6 (
a) shows the HAADF image of precipitate in sample A while Figure 6 (b)and (c) represent EDX-STEM images of precipitate in sample A, Fe K-line image(b) and Si-K line image(c) in Area A1.Through these figures, precipitate contains iron atoms, however quantitative estimation is difficult.TEM images of sample B are shown in Figs.7(a) and (b).No precipitation is observed in the Si matrix, which is apparently different from the results of Fig. 5(a).