Medium-energy ion-scattering study of strained holmium silicide nanoislands grown on silicon (100)

We have used medium-energy ion scattering (MEIS) to quantitatively analyze the structure of holmium silicide islands grown on the Si(100) surface. Structure fitting to the experimental data unambiguously shows that the tetragonal silicide phase is present and not the hexagonal phase, which is associated with the growth of nanowires at submonolayer coverages. Islands formed with a lower holmium coverage of 3 ML are also shown to be tetragonal, which suggests that the hexagonal structure is not a low coverage precursor to the growth of the tetragonal phase. MEIS simulations of large nanoislands, which include the effects of lateral strain relief, have been performed and these compare well with the experimental data.

Figure 1

(a) A top down schematic illustrating the lattice matching of the hexagonal defect-${\text{AlB}}_2$ structure to the Si(100) surface. The anisotropy in lattice mismatch creates the high aspect ratio nanowires commonly observed for several RE metals. (b) A side-view schematic of the ${\text{AlB}}_2$ structure.

Figure 2

A side-view schematic of the tetragonal ${\text{ThSi}}_2$ structure. The orthorhombic ${\text{GdSi}}_2$ structure of ${\text{HoSi}}_2$ is essentially the same but with $a=4.03\text{Å}$ and $b=3.94\text{Å}$, a difference of just 2.25%.

Figure 3

View from above of the bulk terminated Si(100) surface showing the two incident-beam directions used in the MEIS experiments. For the $\left[\overline{1}\overline{1}0\right]/\left[1\overline{1}0\right]$ geometry, the angle with respect to the sample normal was $45°$ and for the $\left[\overline{1}\overline{1}\overline{1}\right]/\left[2\overline{1}\overline{1}\right]$ geometry, the polar angle was $54.74°$. The scattering plane through the tetragonal silicide is shown for the $\left[\overline{1}\overline{1}0\right]/\left[1\overline{1}0\right]$ geometry in Fig. 9.

Figure 4

(Color online) (a) A comparison of the simulations of the hexagonal structure with two to six layers of Ho to the experimental data for the $\left[\overline{1}\overline{1}0\right]/\left[1\overline{1}0\right]$ geometry. (b) The fit obtained for six Ho layers in the hexagonal structure with a Ho-Ho layer separation of $3.28\text{Å}$.

Figure 5

(Color online) (a) Comparison of simulations of the tetragonal structure with thicknesses of two to four ${\text{ThSi}}_2$ cells in height to the experimental data for the $\left[\overline{1}\overline{1}0\right]/\left[1\overline{1}0\right]$ geometry. (b) The fit obtained for the $\left[\overline{1}\overline{1}0\right]/\left[1\overline{1}0\right]$ geometry when the silicide is two ${\text{ThSi}}_2$ cells in height. (c) The fit obtained for the $\left[\overline{1}\overline{1}\overline{1}\right]/\left[2\overline{1}\overline{1}\right]$ geometry when the silicide is two ${\text{ThSi}}_2$ cells in height.

Figure 6

(Color online) A plot showing how the determined $c$ axis is dependent on the size of the $a$ axis of the silicide. Fitting a straight line that passes through the origin yields a $c/a$ ratio of $3.42\pm 0.01$.

Figure 7

View from above of the $10\times 10\times 2$ unit cell showing the lateral strain.

Figure 8

(Color online) Comparison of simulated strain in large $10\times 10\times 2$ unit cells with experiment. The lateral lattice constant of each unit cell that makes up the nanoisland was allowed to relax from $3.84\text{Å}$ [interface lattice matching with Si(100)] to $3.99\text{Å}$ on the surface (bulk lattice constant). The four simulations plotted here correspond to fixed values of the two $c$ axes in the $10\times 10\times 2$ unit cell. The first unit cell $\left(c1\right)$ is nearest the surface and the second $\left(c2\right)$ is nearest the interface.

Figure 9

Schematic slice showing the scattering plane for the $\left[\overline{1}\overline{1}0\right]/\left[1\overline{1}0\right]$ geometry and the atoms responsible for the principal blocking dips observed. Filled circles represent Ho atoms and empty circles represent silicon atoms.

Figure 10

(Color online) Showing the detail around the major blocking dips in Fig. 8. With the gradual introduction of larger, more physically reasonable $c$-axis values into the strained cells, the fit is improved until it is better than that for the fully optimized but nonstrained cell.

Figure 11

(Color online) A comparison of the 6 ML coverage experimental data with data recorded after a sample was grown under identical growth conditions using a lower holmium coverage for the $\left[\overline{1}\overline{1}0\right]/\left[1\overline{1}0\right]$ geometry. The vertical scales have been artificially offset to aid the eye.