Simulations of ordered structure of adsorbates and fragment-

Saturation of Si(100) by HCl consists of a sequence of kinetic events. To analyze the pair correlation functions extracted from the STM images, random adsorption simulations were performed. In the simulations, on a Si(100)-2×1 lattice, a dangling bond that is hit by an HCl molecule, which dissociates, is randomly selected; one of the two separated atoms (referred to as the first atom “A”) is chemisorbed. If a randomly selected site at non-zero coverage already has a previously adsorbed H or Cl adatom, then no adsorption is allowed and a new site is chosen randomly. The second atom (referred to as “B”) is assumed to land with equal probability at one of three immediately neighboring vacant sites on the same dimer row, because the correlation of occupancy between sites in the same row is markedly stronger than that in adjacent row, as discussed in Sec. IVA. The simulation based on this assumption is

s = 1–4. If the immediate neighboring sites are all occupied, then Program I (II) randomly selects one vacant site among the second nearest neighbors s = 4–8 (5–8). At high enough coverage, all eight sites (s = 1–8) can be occupied. In this case, both programs assume that an A-abstraction reaction occurs and B cannot be adsorbed. Without the abstraction reaction,

~10% of surface dangling bonds will be left unoccupied,[75] in contradiction with the STM and photoemission observations. Based on these assumptions, Programs I and II generate the arrangement of H- and Cl-sites, as displayed in Fig. 5.3(a). The corresponding g’s can thus be calculated from the simulated arrangement, and the results are as displayed in Fig. 5.3 (a).

Figure 5.3(a) shows that the final Cl coverage is 0.5, even though the simulation programs indicate that about 8.3% of adsorption reactions are abstractive, because the programs include no atomic selectivity for H and Cl. Program I (II) results in lower values of g’ for s = 1–3 (1–4), because the complementary fragment B atom (H) initially seeks dangling bonds that are immediate neighbors to A (Cl). The occupancy of all other sites is approximately 0.5, indicating no correlation between sites 0 and s. These results from the simulation clearly agree poorly with the experimental data.

Once formed, H- and Cl-sites on the silicon substrate are immobile at RT or below, because of large diffusion barriers. Thus, the correlation between Cl adsorption sites is established during the adsorption process and before the formation of Cl-Si bonds. Hence, a force is already present before the separated Cl atom is adsorbed. Based on the assumption of the fragment-adsorbate interactions are short-range, the forces between a fragmented Cl atom at sites 0 and a Cl-site at s = 1–4 can be described by positive interaction potentials, V01, V02, V03, and V04, respectively. Between Cl- and H-sites, the interaction energy is negligible. The effective substrate potential Veff experienced by the cleaved Cl atom that arrives at a DB (s = 0) is the sum of the four energies V0i (i = 1–4). When an HCl molecule impinges on site 0, the Cl occupation probabilities should be Boltzmann distributed according to

kT

where RCl indicates the initial probability of Cl adsorption on a DB and is 0.5, assuming no atomic selectivity. Equation (5.2) thus suggests a reduced probability from RCl of Cl abstraction at a dangling bond site which has one to four nearest neighboring Cl-sites and, therefore, a lower coverage θand a lower Cl occupancy at s = 1–3.

Since the g’s for s = 1–3, which are located in the same dimer row, are similar, Program I equalizes the potentials Vintra = V01 = V02 = V03 for simplicity, where Vintra denotes intra-row interaction. In the simulation, the reduced factor PCl is also applied to the adsorption of the complimentary B (Cl) atom if H-abstraction occurs initially. All simulations of HCl passivation processes were conducted on an area with 300×150 dimers (300×300 unit cells).

In a quantitative analysis, the g’s obtained from the STM measurement is compared with that from the computer simulations, in terms of the standard deviation σ, which is defined as,

20 respectively, for site s and n = 20 is the total number of calculated sites. A smaller σ indicates a closer match with the simulation result and the corresponding experimental measurement.

Figure 5.5 shows an intensity plot of standard deviation σ using a gray scale for 110 K adsorption, as functions of the repulsion energies Vintra and Vinter = V04 , where Vinter denotes inter-row interaction. The darker shades indicate lower σ and therefore a better fit. As Fig. 5.5 indicates, the repulsion energies of about V_intra = 8.5 ± 1 meV and V_inter = 3.5 ± 1 meV give the best fit. The intra-row energy of fragment-adsorption interaction is about 1/3 of that of adsorbate-adsorbate interaction,[76] perhaps because that the Cl fragment is a neutron radical and not yet negatively charged. Using these energies, the distributions of Cl- and H-sites are simulated for adsorption at 110 K, 300 K, and 450 K (not shown) and displayed in Figs. 5.4(b) and 5.4(c). Figures 5.3(a), 5.3(b) and 5.3(c) plot the corresponding g’s. As Figs. 5.3(b) and 5.3(c) show, simulations reproduce well the overall trends in the g’s. Despite the successful reproduction of g’s for Ts < 300 K, the simulation results for adsorption at elevated temperature do not agree with the experimental ones, as shown in Fig. 5.3(d). Since H-sites and Cl-sites can diffuse on Si(100) via a couple of pathways with rather low energy barriers and the simulation programs herein do not take into account the effect of diffusion, the simulated adsorbate distributions cannot be compared with the experimental ones. Therefore, the disagreement in g’ at 450 K does not exclude the suggested model that is used in the simulation and is accurate at Ts < RT.

PCl that a fragment Cl atoms bond with a DB. However, reducing RCl can lower PCl as well.

As mentioned above, the simulation programs assume that H or Cl is adsorbed with equal probability on a DB that is randomly struck by a HCl molecule, such that RCl = 0.5.

Simulations using various RCl and Veff = 0 give rather featureless g’s and large θ, as shown in Fig. 5.6 and, therefore, fail to predict the measurements accurately. Giving the set of the repulsion energies found above, the simulation reproduces the unnormalized pair correlation function g’ reasonably well over a large range of RCl. These findings further explain the presence of the repulsive forces between fragment atoms and adsorbates as well as the repulsion energies found above.

Figure 5.3 Unnormalized pair correlation function g’ obtained from simulation (a) Program I (squares) and Program II (diamonds) with a zero fragment-adsorbate energy of interaction;

(b-d) simulations with fragment-adsorbate energy of interaction and STM images (filled circles) of the samples in Figs. 5.2(a-c). The repulsive fragment-adsorbate energies used are V_intra = 8.5 meV and V_inter = 3.5 meV. The insets in (a) show the lattice of Si(100)-2×1:HCl;

each circle corresponds to an H- or Cl-site. Numbers are site index s for sites that neighbor to the central adsorbed Cl atom (s = 0). Insets in (b-c) are grey-scale representations of g’.

Figure 5. 4 Simulated distributions of coadsorbed H (circles) and Cl (filled circles) sites on Si(100)-2×1 at (a-b) 110 K and (c) 300 K. Coverages of Cl-adsorbed sites in monolayer are given. Arrows indicate selective patches of 2×2 structure. In (a) all interaction energies are 0.

In (b-c), the repulsive energies are Vintra = 8.5 meV and Vinter = 3.5 meV.

Figure 5.5 Contour representation of standard deviation σ between the simulation and STM result (110 K) as functions of repulsive interacting energies V_intra and V_inter. Simulated results more closely match STM measurements in darker areas where σ is smaller

Figure 5.6 Unnormalized pair correlation function g’ obtained by Program I, which uses different R_Cl values with and without effective substrate potential V_eff as labeled.

5.5 Conclusion

This work elucidates the kinetics of HCl adsorption on the Si(100)-2×1 surface by integrating a spectroscopic measurement of core-level photoemission, atomic resolved STM imaging, and computer simulations. Experimental results indicate that the adsorption proceeds by the combination of a dissociative mechanism and surface abstraction in favor of H fragments over Cl. During dissociative chemisorption on surfaces, the fragment-adsorbate interactions are present and change the ordered structure of the adsorbates. Reliable values of the fragment-adsorbates potentials can be extracted by comparing the simulated and measured occupancy pair correlation functions. If a Cl fragment is in same dimer row as another Cl adsorbate, then the energy of their interaction energy (8.5 meV) exceeds that (3.5 meV) when they are in adjacent rows.