Thin Solid Films Volume 515, Issue 14, 23 May 2007, Pages 5620-5623 The Ninth International Conference

on Surface X-Ray and Neutron Scattering - 9SXNS Structural characterization of thin amorphous Si films D. Grozdania, B. Rakvinb, B. Pivacb,
a ,

, P. Dubekb, N. Radib and S. Bernstorffc

Faculty of Electrical Engineering and Computing, University of Zagreb, Zagreb, Croatia R. Bokovi Institute, P.O. Box 180, HR-10000 Zagreb, Croatia

Sincrotrone Trieste, SS 14, km 163.5, Basovizza (TS), Italy

Available online 23 January 2007.

Abstract
We present a study of structural changes occurring in thin amorphous silicon (a-Si). The a-Si films were deposited on single-crystalline Si substrates held at room temperature or 200 C by magnetron sputtering of a Si target in pure Ar atmosphere, and therefore the films were hydrogen-free. All samples were annealed in vacuum and subsequently studied by EPR and GISAXS. A strong decrease in the dangling bonds content at lower annealing temperatures, and then an increase of it at around 550 C, suggested significant structural changes. In parallel the samples were studied by GISAXS which confirmed changes at the nanometric scale attributed to voids in the material. A nice correlation of the results of the two techniques shows advantages of this approach in the analysis of structural changes in a-Si material. Keywords: Amorphous silicon; Nanostructures; Small angle X-ray scattering; Electron paramagnetic resonance

Article Outline
1. Introduction 2. Experimental 3. Results and discussion 4. Conclusion References

1. Introduction
The intrinsic morphology of pure amorphous silicon (a-Si) on a nanometric scale has been an intriguing issue for more than two decades. This is mainly due to the fact that the

actual morphology depends on several parameters such as: the method of preparation and the thermal history of a particular sample, as well as on many other structural properties [1], [2], [3] and [4]. It is known that a-Si deposited either by vacuum evaporation or sputtering almost inevitably contains nm-sized voids, which may or may not be connected. The existence of such structures within the a-Si has been confirmed by electron diffraction [5] and has been inferred from clustering of dangling bonds [2]. Still structural changes in a-Si during the course of thermal annealing are not completely characterized especially since the starting structure of a-Si is very much dependent on the preparation conditions. Different techniques are employed for such studies and one of them electron paramagnetic resonance (EPR) describes very well the changes during annealing [6] and [7]. Dangling bonds as a spin probe in the lattice provide very useful information on relaxation and structural changes. As they represent defects in the structure, these defects may act as the hopping sites for a variable range hopping conductivity, which is of interest as a phenomenon of disordered semiconductors. Moreover, such studies are of interest to understand if and how impurities, if present in the bulk, affect the structural changes evolving upon thermal treatment. However, not all unsaturated bonds are paramagnetically active, and therefore we shall complement such information with structural measurements performed with grazing incidence small-angle X-ray scattering (GISAXS). The small-angle X-ray scattering (SAXS) technique is a well established technique for the detection of void distribution in a-Si and it has been applied with success to both a-Si and a-Si:H [8] and [9].

2. Experimental
Intrinsic amorphous silicon thin films were deposited onto single-crystalline Si (100) Czochralski wafers by rf magnetron sputtering in a CMS18 sputtering system (Kurt J. Lesker Co., Clairton, PA, USA). During deposition, the substrate holder rotated at 10 RPM in order to avoid the formation of a built-in stress preferential direction. The base pressure in the process chamber was 10 610 5 Pa, and the working gas was argon of 5 N purity in a continuous flow. The working pressure during the deposition was 1.3 Pa and an ultrapure Si target (99.9999%) of 75 mm in diameter was used. The deposition rate was about 4 nm/min at 250 W rf magnetron power, and the final film thickness was about 1000 nm. During the deposition the substrate was held at a floating potential, and either at room temperature (RT) or at 200 C. After the deposition, the samples were divided into several pieces. Each piece was then annealed in vacuum at a selected fix temperature for 1 h, and then cooled and measured at RT. EPR measurements were performed employing a Varian E-109 spectrometer operating in the X-band (9.4 GHz). Several EPR were recorded in steps of 10 of the rotation of the magnetic field with respect to the Si (011) plane in order to reveal the dependence of the paramagnetic centers on the magnetic field orientation. In order to improve the signal-tonoise ratio, signal averaging ( 10 traces) was applied. The exact position of the g-factor and quantitative spin concentration was determined according to DPPH standard, and the detection limit of 1011 spins/cm3 was achieved.

The GISAXS experiments were carried out at the synchrotron facilities of Elettra, Trieste, Italy at the SAXS beamline [10], using synchrotron radiation with wavelength = 0.154 nm (photon energy of 8 keV). The grazing angle of incidence was selected in the range 0.3 < i < 1.4 for which the effective area of the beam foot print was smaller than the sample surface area (20 mm 20 mm); and the grazing angle i was larger than the critical angle for total external reflection on the silicon substrate crit(Si) = 0.23. A two-dimensional CCD detector with 1024 1024 pixels, positioned perpendicular to the incident beam at a detector to sample distance L = 2000 mm, was used to record the SAXS intensity. A thin Al-strip was placed in front of the 2D detector to avoid its saturation in the specular plane direction where the usually much stronger surface scattering is present. The spectra were corrected for background intensity and detector response.

3. Results and discussion

Fig. 1 shows the variation of the spin density as a function of the annealing temperature for samples deposited either at RT or at 200 C. As shown in the figure the spin density does not vary much until about 250 C for the sample deposited at RT, or until about 350 C for the sample deposited at 200 C. For the sample deposited at RT annealing at temperatures above 250 C produced a monotone increase in the spin density which passes through a maximum for annealing between 600 and 700 C where it decreases again. On the other hand the spin density for the sample deposited at 200 C increases strongly for annealing temperatures in the range 350450 C. Then it decreases slightly until about 650 C, after which it increases again strongly for higher temperatures. A similar behavior was observed by Brodsky et al. [11], and they explained their results with cracking of newly formed crystals. In our case this increase in spin density upon high temperature annealing is however, much more pronounced. Due to a specific method of deposition, i.e. magnetron sputtering it is very likely that our material is loaded with voids that might coalesce at higher temperatures. Moreover, since it has been shown that at about 600 C crystallization occurs [11] it is very likely that we are observing nanocrystals formation at even somewhat lower temperatures. These nanocrystals increase in number and size upon annealing at higher temperatures. Unlike our results, the results of Thomas et al. [6] show a continuous decrease of the spin density upon annealing up to 600 C where it vanishes completely.

Full-size image (29K) Fig. 1. Spin density versus annealing temperature of a-Si films deposited on substrates held at room temperature (open squares) or at 200 C (black dots), respectively. After deposition, each sample was divided into several pieces, which were then annealed each

at only one fixed temperature for 1 h in vacuum. Thus the data points shown were collected on different sample pieces (see also Experimental section).

In order to verify this behavior of film inhomogeneity we have employed GISAXS which is very sensitive to density variations in the nanometer size range. As a typical result, Fig. 2 shows a GISAXS pattern obtained on a non-annealed sample, deposited at room temperature. Apart from the rather strong contribution of the surface scattering that is present in the vicinity of the specular plain (qy = 0), there is a diffuse, semicircular contribution from the bulk scattering. In the range qz < 0.55 nm 1 no scattered intensity can be seen due to X-ray absorption in the sample. On the other hand, the strongest surface scattering contribution is located in the specular plain and is significantly stronger than the bulk scattering. Therefore, the central part of the detector was covered by a partly transparent aluminum absorber that prevented detector saturation and thus enabled the detection of the weak bulk scattering signal outside of the specular plain. The shadow of the absorber is visible in Fig. 2 as an apparent lack of signal for approximately 0.05 nm 1 < qy < 0.05 nm 1.

Full-size image (47K) Fig. 2. Contour plot of GISAXS pattern obtained from a non-annealed sample, deposited at room temperature. The central vertical stripe is due to a thin, partly transparent Alabsorber in front of the 2D detector used to avoid its saturation.

Since the bulk signal is isotropic, i.e. the intensity is a function of the total scattering angle only, it is sufficient to investigate the intensities along a certain direction in reciprocal space. Thus we investigated the intensities as a function of the scattering angle parallel to the sample surface in order to minimize the influence of refraction effects on the results. A linear scan of the intensity was taken from Fig. 2 at qz = 0.7 nm 1 and a Guinier approximation was fitted to the measured intensity in order to determine the typical size of the inhomogeneities that are present in the sample. Fig. 3 shows such scans which were obtained from a complete series of samples (produced by sputtering on a substrate held at room temperature, and subsequent annealing in vacuum of the different sample pieces at different, fix temperatures in the range RT (for the lowest curve) up to 750 C (highest curve). The intensities for qy > 0.3 nm 1 have been used to determine the

sizes of the inhomogeneities by curve fitting. The so obtained results for both sample series (i.e. with substrates held at room temperature or at 200 C during sputtering of the layers) are displayed in Fig. 4.

Full-size image (47K) Fig. 3. 1D intensity cuts taken from two dimensional scattering patterns for samples sputtered at room temperature, and annealed at a fixed temperature of RT, 150, 250, 350, 450, 550, 650, and 750 C (from bottom to top).

Full-size image (31K) Fig. 4. Guinier radius of inhomogeneities vs. annealing temperature for the samples sputter deposited at room temperature (empty circles) or at 200 C (full circles).

In the case of the samples sputtered at room temperature, the typical size of the inhomogeneities decreases with annealing temperature up to 150 C, where this trend is reversed, and the size increases up to 2.4 nm at the highest annealing temperature. For the sample sputtered at 200 C, the sizes are smaller but the temperature behavior is similar. The inhomogeneities size is significantly reduced upon annealing to 250 C, and above this temperature it increases with annealing temperature, showing a small dip at about 550 C. These results are in good agreement with the variation of the spin density as a function of the annealing temperature (compare Fig. 4 to Fig. 1). Now the question is what do we detect with GISAXS as an inhomogeneity in our samples. Having in mind that we have a very simple system, i.e. amorphous silicon deposited with no hydrogen presence we may have either voids or nanocrystalline silicon inclusions in amorphous matrix. The occurrence of both structures would give a similar contribution, but the presence of voids due to the larger difference in the electron density of void versus the surrounding amorphous matrix would be easier to detect. Raman measurements that we cannot present here for space limitations, however,

indicated that the as-deposited samples and the samples annealed at lower temperatures up to 400 C are completely amorphous. This supports our assumption that GISAXS measurements detect the presence of voids in the matrix. Our GISAXS results on annealed samples have shown that upon annealing up to about 250 C, the a-Si structure relaxes and thus produces a reduction in void size. However, upon annealing at higher temperatures voids grow or coalesce forming larger ones supported very likely by nanocrystal growth. We have shown that the results of both techniques correlate very well. This correlation we understand in the way that a number of voids present in the a-Si matrix relaxed this matrix so that the dominant contribution of the dangling bonds comes from the void region. Thermal treatment produced changes in the void size, as shown by GISAXS, which also reflects in the spin density, as shown by EPR. Further studies with X-ray reflectivity and diffraction are in progress.

4. Conclusion
We studied magnetron sputtered a-Si films deposited on single-crystalline Si wafer held at room temperature or at 200 C and their structural changes upon subsequent annealing. We measured changes in spin density by EPR and void parameters by GISAXS. Both techniques have shown that annealing at low temperatures caused a relaxation in the a-Si matrix producing a decrease in spin density and void size. Further annealing at higher temperatures caused again an increase in spin density and an increase in void size which probably coalesce. The nice correlation between EPR and GISAXS results on annealed samples shows that those two techniques could very nicely complement each other in studying structural changes in a-Si.

References
[1] S. Roorda, W.C. Sinke, J.M. Poate, D.C. Jacobson, S. Drierker, B.S. Dennis, D.J. Eaglesham, F. Spapen and P. Fuoss, Phys. Rev., B 44 (1991), p. 3702. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (165) [2] D.L. Williamson, S. Roorda, M. Chicoine, R. Tabti, P.A. Stolk, S. Acco and F. Saris, Appl. Phys. Lett. 67 (1995), p. 226. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (43) [3] H. Touir, K. Zellama and J.-F. Morhange, Phys. Rev., B 59 (1999), p. 10076. APS full text | Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (16) [4] A.H. Mahan, J. Yang, S. Guha and D.L. Williamson, Phys. Rev., B 61 (2000), p. 1677. APS full text | Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (56) [5] S.C. Moss and J.F. Graczyk, Phys. Rev. Lett. 23 (1969), p. 1167. Full Text via CrossRef [6] P.A. Thomas, M.H. Brodsky, D. Kaplan and D. Lepine, Phys. Rev., B 18 (1978), p. 3059. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (29)

[7] B. Pivac, B. Rakvin and R. Reitano, Nucl. Instrum. Methods Phys. Res., B Beam Interact. Mater. Atoms 147 (1999), p. 132. Article | PDF (97 K) | View Record in Scopus | Cited By in Scopus (6) [8] J. Shinar, H. Jia, R. Shinar, Y. Chen and D.L. Williamson, Phys. Rev., B 50 (1994), p. 7358. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (13) [9] P. Dubek, B. Pivac, S. Bernstorff, R. Tonini, F. Corni and G. Ottaviani, Nucl. Instrum. Methods Phys. Res., B Beam Interact. Mater. Atoms 200 (2003), p. 110. Article | PDF (140 K) | View Record in Scopus | Cited By in Scopus (2) [10] H. Amenitsch, S. Bernstorff and P. Laggner, Rev. Sci. Instrum. 66 (1995), p. 1624. Full Text via CrossRef | View Record in Scopus | Cited By in Scopus (89) [11] M.H. Brodsky, R.S. Title, K. Weiser and G.D. Pettit, Phys. Rev., B 1 (1970), p. 2632. Full Text via CrossRef