in-situ XPS Investigation of the Si-Nanocrystal Self-Assembly

Figure 5.2 shows the evolution of the O 1s and Si 2p XPS spectra during the a-Si deposition on SiO2/Si and the following in-situ annealing. All peaks were referenced to the Si⁴⁺ 2p (Si-O) peak at binding energy of EB = 103.3 eV. The energy separations between the O 1s and Si⁴⁺ 2p peaks remained the same (ΔE^B = 429.3 eV) throughout the process, indicating no chemical or structural changes in the tunnel oxide layer [118]. The initial spectrum (line 1 in Fig. 5.2) features the O 1s and Si⁴⁺

2p peaks for the tunnel-oxide and a smaller Si⁰ 2p peak at EB = 99.1 eV for the Si substrate. In the next spectrum (line 2) recorded with a 3.5-nm a-Si overlayer, the O 1s and Si⁴⁺ 2p peaks attenuated while the Si⁰ 2p signals increased. The reduction of the oxide signals was caused by the photoelectron scattering in the top a-Si layer which also provided additional Si⁰ 2p signals.

The attenuation in O 1s signals can be weighed to assess the Si deposition uniformity. In the following calculations a straight-line approximation is adopted, in which the XPS signal attenuation through a material is described by the exponential decay [49]. This assumption has been confirmed experimentally [120] and theoretically by a Monte Carlo simulation [81]. Considering a two-layer a-Si/oxide structure on a Si substrate, the XPS intensity of signals from the oxide interlayer can be given by the following equation [121]:

I = ΦAσyDT(E)cos(θ)xNλox

×{1-exp[-T^ox/λ^oxcos(θ)]}×exp[-T^a-Si/λ^a-Sicos(θ)]. (5.1) In this equation, Φ is the x-ray dose at the surface of the sample; A is the asymmetry parameter, dependent on the angle between the axis of the analyzer and the x-ray source; σ is the cross section for the photoelectric effect; y is the efficiency

of production in the photoelectric process to give photoelectrons of normal energy in the tunnel-oxide layer; DT(E) is the detector efficiency and transmission function for the kinetic energy corresponding to the specific chemical state (x-ray photoelectron transition); x is the fraction of atoms in the tunnel-oxide layer responsible for the photoelectron transition; N is the atomic density in the oxide interlayer; λox and λa-Si

are the photoelectron effective attenuation lengths in the tunnel-oxide and in the a-Si, respectively; Tox and Ta-Si represent the tunnel-oxide thickness and the a-Si thickness, respectively; and θ is the angle between the axis of the analyzer and the normal to the surface [121].

Considering that only the a-Si thickness would vary during its deposition, equation 5.1 can be simplified into the following form:

I = I 0×exp[-T^a-Si/λ^a-Sicos(θ)], (5.2)

or

I/I 0 = exp[-Ta-Si/λ^a-Sicos(θ)]. (5.3)

In these equations, I0 represents the intensity of XPS signals acquired before the a-Si top-layer deposition. Equation 5.3 is derived based upon the assumption that the a-Si layer is continuous and uniform without any pinholes or hillocks. If pinholes or hillocks do exist in the film, the XPS intensity will be an integral of signals from localized areas on which the a-Si thickness varies and can not be represented by a single term.

Figure 5.3(a) shows the XPS core-level O 1s spectra recorded during the a-Si deposition on the tunnel-oxide/Si substrate. The O 1s peak intensity, which is obtained

thin films [122]. The cross-sectional TEM micrograph in Fig. 5.3(c) confirms the uniformity of a 3.5-nm as-deposited Si layer. AFM measurements made on the a-Si surface have shown a root-mean-squared roughness of 0.12 nm, which is close to the original tunnel-oxide surface roughness (~0.1 nm). Therefore the e-beam evaporation of Si has resulted into smooth and uniform a-Si layers on top of the tunnel-oxide/Si substrates.

After in-situ RTA of the ultrathin a-Si layer, however, the XPS intensity of the oxide-interlayer signals was partially recovered, as indicated by the last three spectra (3–5) in Fig. 5.2. After in-situ annealing a 3.5-nm a-Si layer on the tunnel-oxide/Si substrate at 850 °C for 1 minute, the O 1s and Si⁴⁺ 2p intensities rebounded while the Si⁰ 2p signal decreased (line 3 in Fig. 5.2). This observation can be explained by the Si coalescence and subsequent nanoparticle formation which, as a result, partially exposed the tunnel-oxide surface and hence increased the O 1s and Si⁴⁺ 2p photoelectron yields. The decreased Si⁰ 2p photoelectron yield was due to the reduced surface coverage by Si and the signal attenuation in taller Si islands. SEM analyses later confirmed the formation of separated Si nanoparticles, as shown in Fig. 5.4.

Samples annealed for different durations (1–10 min) had nearly identical O 1s and Si⁴⁺ 2p peaks in the spectra 3–5 of Fig. 5.2, implying few changes in the nanoparticle topology. Indeed their SEM images revealed similar dot sizes and densities. For example, Figs. 5.4(a) and 5.4(b) present SEM images of samples annealed in situ at 850 °C for 1 min and 10 min, respectively. The dot densities and base radii extracted by analyzing SEM micrographs are summarized in Figure 5.5. In correspondence to the in-situ XPS results, different annealing periods have resulted into similar nanoparticle densities of approximately 8×10¹⁰ cm^-2 with average base radii around 11 nm. Therefore it is believed that most of the silicon agglomeration has completed after the first minute of annealing. It should be noted that in Fig. 5.2(b) a larger Si 2p

chemical shift was observed between the Si⁴⁺ and Si⁰ peaks after the in-situ annealing.

This is possibly due to the differential charging between the Si nanoparticles and the tunnel-oxide surface during the x-ray irradiation [118]. The broadened Si⁰ 2p peaks along with the change of the Si⁰ 2p peak shape (for example a shoulder at the lower EB side) are also indicative of possibly nonuniform charging among the Si nanoparticles.

5.3.2 Effects of the Annealing Temperature on the Silicon Agglomeration and