Quantitative investigation of the influence of carbon surfactant on Ge surface diffusion and island nucleation on Si(100)

We investigated the surface diffusion and island nucleation of Ge on Si(100) in presence of a submonolayer coverage of carbon as surfactant by using scanning Auger microscopy and atomic force microscopy. Ge stripes have been deposited and lithographically etched on a Si substrate and used as sources for the surface diffusion of Ge promoted by annealing at 600, 650, and $700°\text{C}$. The diffusion coefficient has been determined by fitting the postannealing coverage profiles measured by Auger microscopy with a one-dimensional continuous model. The carbon coverage has been spatially modulated on a single sample, allowing the measurement of the diffusion coefficient as a function of the C thickness at $600°\text{C}$. We show that the reduction in the diffusion coefficient while increasing the surfactant coverage is described by a linear dependence of the diffusion activation energy on the C coverage. This dependence is discussed in terms of the chemical interactions among Si, C, and Ge, of the surface roughness and the local strain field induced by the C surfactant. Spontaneous nucleation of SiGe islands coexists with the continuous surface diffusion of Ge. The transition of the island nucleation as a function of the carbon coverage is observed to be continuous from the Stranski-Krastanov mode to the Volmer-Weber regime. We propose a consistent scenario correlating diffusion and nucleation parameters within a diffusion limited growth regime and show the existence of a threshold for C coverage below which no effect is observed.

Figure 1

(Color online) (a) SEM image of the stripe after annealing at $600°\text{C}$ for 10 min showing a bilateral diffusion (gray shaded area). Surface contaminations have been removed by using an isotropic ion sputtering (as schematically shown in the inset in the bottom-left corner). In the top-left corner is shown the SEM image of the stripe before the annealing. (b) Ge LMM and Si LMM Auger lines measured at different distances from the stripe as indicated by the solid black arrows in the panel (c). (c) Overlayer thickness as a function of the distance from the source as determined by SAM analysis (black squares). The green continuous line is the best fitting of the experimental data using the analytical solution of a 1D diffusion model (see text). Inset: schematic of the detection geometry and of the diffusion region as represented inside the discrete layer model.

Figure 2

(Color online) (a) $L^2/4$, with $L$ diffusion length, plotted as a function of the annealing time, $\tau$, for different temperatures (600, 650, and $700°\text{C}$). Linear fittings of the experimental data were used to extract the values of the diffusion coefficient, $D$, at different temperatures $\left(L^2=4D\tau \right)$. (b) The diffusion coefficient has an exponential temperature dependence given by an Arrhenius law: diffusion constant $D_0=6.08\times {10}^{-2}{\text{cm}}^2/\text{s}$, activation energy $E_A=1.24\text{eV}$.

Figure 3

(Color online) Diffusion regions for two different stripes after annealing for 10 min at $600°\text{C}$ in presence of a wedged C coverage. Transversal sputtering has been used in this case (see inset on top). For both stripes, in the C-rich region (at the bottom) the diffusion is inhibited, while in the C-free region (at the top) the diffusion is favored with a continuous transition between these two extremes. The limit of the diffusion region as a function of the carbon coverage is outlined by the green contour (left panel) and red contour (right panel, where the rescaled green contour is also reported as a comparison). The regions with low, medium, and high C coverage used in Sec. 3C are shown by red shaded areas.

Figure 4

(Color online) (a) Auger mapping of the Ge diffusion profiles as a function of the C coverage. (b) Quantitative correlation between the Ge diffusion coefficient and the carbon coverage (linear plot): the green continuous line is the best fit of the experimental data (black squares) using a model where the activation energy for surface diffusion linearly depends on the carbon coverage (see text). The free parameters $D_0^{\ast }$ and $E_A^1$ result from the fitting to be $5.68\pm 0.33\times {10}^{-2}{\text{cm}}^2/\text{s}$ and $0.29\pm 0.04\text{eV}$, respectively. Inset: logarithmic plot of the diffusion coefficient vs the carbon coverage.

Figure 5

(Color online) (a) Island density and (b) overlayer thickness vs distance, $x$, from the stripe in regions with different C coverages. The vertical blue, green, and black solid lines are the position of the onset for island nucleation in case of low C, medium C, and high C coverages, respectively (gray shaded zones represent the uncertainty). The comparison of the two graphs allows for the direct experimental determination of the critical OL thickness for nucleation with different C coverages.

Figure 6

(Color online) (a) Carbon concentration $c$ (black squares) in the topmost layer of the C-alloyed Si substrate as a function of the C coverage $\vartheta$; green line is the best linear fitting of the data. [(b)–(d)] Volume histograms of the SiGe islands nucleated in areas with different carbon coverages. Smaller islands are preferentially nucleated in the C-rich region [panel (b)] while bigger islands grow at low C-coverage zones [panels (c) and (d)]. The insets in the panels are AFM images in gradient modes of islands nucleated for different C coverages. The dashed-dotted green lines shown in each panel represent the average volume of the corresponding distribution.

Figure 7

Average (a) volume, (b) area, (c) aspect ratio, and (d) and density of the islands nucleated in areas with different C coverages correlated with the diffusion coefficient.