In this work we present a study of small-angle x-ray scattering (SAXS) and nuclear magnetic resonance (NMR) on hydrogen-implanted amorphous silicon (a-Si) and standard device-quality plasma-grown hydrogenated a-Si (a-Si:H), both having a concentration of 11 at.%. The modifications of the short- and medium-range structural order induced by annealing are investigated. We find that annealing causes the formation and growth of nanoscale H complexes in both materials. However, the volume content of the H nanoclusters remains at least a factor of three lower in a-Si:H than in H-implanted a-Si at all annealing temperatures. The comparison of the results obtained from H-implanted a-Si, characterized by a high disorder and defect density, with those from plasma-grown a-Si:H, which exhibits a low defect state, sheds light on the process of H precipitation during annealing and its influence on the initial structural disorder and on the atomic-scale distribution of H in the matrix.

Fig.1.

Figure 1 shows the SAXS intensity (in e/a, electron units) as a function of the momentum transfer q=(4 / )sin (2 =scattering angle), for H-implanted a-Si, referred to as the MH sample. The scattering intensity for the sample in the as-implanted state and after annealing up to and including 250 C (not shown) is essentially independent of the momentum transfer q. It follows that (1) the signal contains only diffuse SAXS intensity, which originates from atomic-scale electron density fluctuations; (2) nanoscale inhomogeneities, if present, are well below the present SAXS sensitivity of 0.1 vol.%. A similar conclusion can be drawn from the analysis of the SAXS intensities for the as-deposited state of the a-Si:H sample, made by plasma enhanced chemical vapor deposition (PECVD) and referred to as the GD sample, as displayed in Fig. 2.

Fig.2.

We therefore conclude that the two materials have an homogeneous nanoscale structure in the initial state. Thermal treatments at temperatures higher than 250 C for the MH sample and than 300 C for the GD sample modify the structure of the samples. The scattering intensities exhibit now a q dependence, corresponding to the formation of nanoscale low-density structural features. The systematic increase of the SAXS intensity with increasing annealing temperature demonstrates an increase of the volume content of the inhomogeneities.

Fig.3.

Figure 3 shows the variation of the volume fraction f2 of the nanoscale inhomogeneities as a function of the annealing temperature for the MH and the GD samples. The f2 values are obtained from SAXS data. The value of f2 of the inhomogeneities for the GD sample remains lower than that of the implanted material in the whole temperature regime. In addition, the annealing temperature at which the nanoparticles are detected is different in the two cases. In the implanted material, nanoparticles appear at temperatures higher than 250 C, while temperatures higher than 300 C are necessary for the GD sample. Thus, despite the qualitative resemblance of the structural evolution of the MH and GD samples, the magnitude of the H nanoclustering differs considerably in the two cases. Considering the similar initial H content, this difference is to be related to the initial defect structure, which is drastically different in the two cases.
Previous results1 on hydrogenated amorphous silicon made by ion implantation showed that introducing H into a-Si at a concentration higher than ~4 at.% leads to a thermal instability of the alloy structure. This upper limit for thermal stability of the H-implanted a-Si structure was associated with the number of defect-related trapping sites for H in the matrix and in fact defines the H solubility. The H-implanted a-Si sample studied in this work contains 12 at.% H and thus has an unstable network structure. The evolution upon annealing of the alloy structure of plasma-grown a-Si:H is qualitatively similar to that of H-implanted a-Si with H concentration exceeding solubility.
However, the degree of H clustering is considerably lower than that observed in H-implanted a-Si. Together with the solubility limit for a-Si:H, this would imply a larger solubility of PECVD a-Si:H than in the implanted material.

Fig.4.

Figure 4 shows NMR spectra for H-implanted a-Si and PECVD a-Si:H, in the as-prepared state. These two samples are also referred to as the MH and the GD samples because they were made under same conditions as those studied by SAXS. Displayed are the Fourier transform of the free induction decay (FID) at the proton resonance. A two-component spectral line, which is typical for hydrogenated a-Si, is observed in the two samples. The spectral components identify two local environments for bonded hydrogen in a-Si. The narrow spectral line is attributed to bonded hydrogen essentially randomly distributed in the matrix and the broad line is associated with hydrogen bonded in a more clustered phase, i.e., the mean distance between H atoms is smaller than about 7 . Experimental spectral lines were fitted with the sum of a Lorentzian for the narrow line and a Gaussian for the broad line. For the H-implanted a-Si, the fitting indicates that 63% of the total H is in the narrow component. Taking as average H content 7.4±1 at.% (from SIMS), and assuming that the amount of molecular H2 present is the sample can be neglected, results indicate that on average 4.7±1 at.% of H is in the narrow line and 2.7±1 at.% is in the broad line. The concentration of H present in the narrow line, which corresponds to the more randomly distributed protons, is fully consistent with the concentration of 4 at.% that defined the H solubility in ion-implanted a-Si.1 Therefore, these results indicate a correspondence between the fraction of H atoms that can evolve upon annealing while remaining dissolved in the matrix as Si-H bonds and the concentration of H atoms more randomly distributed on a local scale.
From the fitting of the a-Si:H sample (Fig. 4, bottom), which contains H 11±1 at.% H, we infer that 7±1 at.% H is the narrow component and 4±1 at.% is in the broad component. Present findings are hence in qualitative agreement with SAXS results, which have suggested a higher H solubility in the a-Si:H sample compared to the implanted material. Therefore, despite the initial similar H concentration, the initial defect state and the degree of disorder determines the initial configurations in the network and thereby their evolution upon annealing.

[1]	S. Acco, D.L. Williamson, P.A. Stolk, F.W. Saris, M.J. van der Boogaard, W.C. Sinke, W.F. van der Weg, S. Roorda, and P.C. Zalm, Phys. Rev. B 53, 4415 (1996).

[BACK] [MAIN] [FURTHER]