Wet chemical preparation and isotope exchange process of H/D-terminated Si(111) and Si(110) studied by adsorbate vibrational analysis

Before proceeding to the vibrational and geometrical analyses, we review the surface atomic arrangement on hydrogen-terminated Si(111) and Si(110) surfaces based on knowledge accumulated so far.^{7–9,18–20,22–27)}

Figure 1 shows model structures of hydrogen-terminated Si(111) and Si(110). The H:Si(111) surface inherits its periodicity, three fold rotational symmetry (two-dimensional space group p6mm), and unit-cell size from the (111) bulk-truncated plane of the diamond-type Si crystal. The first-nearest-neighbor distance of a terminating hydrogen atom is 0.384 nm.

Zoom In Zoom Out Reset image size

The unit-cell size of H:Si(110) is also identical to that of bulk-truncated (110). Along the (110) plane of the diamond-like lattice, a glide-plane symmetry is introduced. H:Si(110) belongs to the two-dimensional p2mg space group, with two hydrogen atoms accommodated in each unit cell. The glide planes are parallel to the $[1\bar{1}0]$ direction. According to our previous estimation,⁷⁾ the first-nearest-neighbor distance of a hydrogen atom is 0.299 nm and the second-nearest-neighbor distance is 0.365 nm, which are significantly shorter than those of Si(111).

The test pieces of H:Si(111) or H:Si(110) were prepared in 40% NH₄F/H₂O [with a small amount of (NH₄)₂SO₃] to prepare microscopically flat, completely H-terminated surfaces. Each piece was immersed in saturated KF/D₂O (with a small amount of K₂SO₃) for the desired period (0–180 min) at room temperature (297 K). Then the test piece was subjected to HREELS recording. The primary energy of the incident electron beam (E_i) used in HREELS was 6.0 eV, while its incident angle (θ_i) and scattering angle (θ_s) were 60° with respect to the surface normal.

Figure 2 shows a series of HREELS spectra on H/D:Si(111) recorded along the sagittal plane parallel to $[11\bar{2}]$ of Si(111). At time = 0 (without immersion in KF/D₂O), two distinct peaks are observed at 258.2 ± 0.5 and 78.5 ± 0.5 meV, which are assigned to the H–Si stretching and bending modes, respectively. These vibration energies were previously studied by infrared absorption spectroscopy (IR) and HREELS. By IR, Higashi et al. observed the stretching mode at 2083.7 cm⁻¹ (258.2 meV)¹⁷⁾ and Caudano et al. observed the bending mode at 626.7 cm⁻¹ (77.6 meV).³²⁾ By HREELS, Stuhlmann et al. observed the stretching mode at 2085 cm⁻¹ (258.3 meV) and the bending mode at 631 cm⁻¹ (78.2 meV).³³⁾ Kato et al. observed the two peaks at 258.5 and 78.1 meV, respectively.²⁰⁾

Zoom In Zoom Out Reset image size

With increasing immersion time in KF/D₂O, the two peaks are weakened and two new peaks appear at 187.3 ± 0.5 and 51.8 ± 1.0 meV. They are assigned to the D–Si stretching mode and sagittal-plane bending mode, denoted "B₃",²¹⁾ respectively. Kato et al. observed the stretching mode at 186.9 ± 1.0 meV, the B₁ mode at 65.7 ± 0.6 meV, and the B₃ mode at 50.9 ± 0.6 meV by HREELS.²¹⁾ By IR, Watanabe measured the stretching mode at 1515.6 cm⁻¹ (187.8 meV) and the two bending modes at 536.8 cm⁻¹ (66.5 meV) and 415.4 cm⁻¹ (51.5 meV).^23,28) Ferguson et al. observed the stretching mode at 1516 cm⁻¹ (187.8 meV) and the bending mode at 537 cm⁻¹ (66.5 meV).³¹⁾

Assuming a H/D–Si vibrator with an infinite Si atom mass, the "reduced mass" of the D–Si vibrator should be exactly twice that of H–Si, and then the D–Si frequencies should be $1/\sqrt{2}$ times the H–Si frequencies. However, compared with the expected values, there are +2.4 and −6.7% shifts in the HREELS spectra energy of the D–Si stretching mode and B₃ mode, respectively. This is due to the motion of Si atoms, or, more specifically, to coupling with the phonon modes below the surface, as discussed previously.²¹⁾

Other peaks are assigned to oxygen and hydrocarbon contamination.³⁴⁾ The peaks at 103 and 129 meV are associated with Si–O–Si bonds. The background intensity in the vicinity of these two peaks is contributed by the C–H bending modes in adsorbed hydrocarbons. The adsorption of hydrocarbons is indicated in the peak at 360 meV associated with the C–H stretching mode. After 180 min etching, we obtained a D:Si(111) surface with zero H content. The amount of SiO₂ on the present D:Si(111) surface is below 0.001 ML.^21,35)

Figure 3 shows the KF/D₂O immersion time dependence of H:Si(110) HREELS spectra. The incident energy and configuration of HREELS were the same as those in the case of the Si(111) surface, with the sagittal plane along $[\bar{1}10]$ of Si(110). The H/D exchange on Si(110) proceeded markedly faster than that on Si(111), and 30 min.

Zoom In Zoom Out Reset image size

At time = 0, two distinct peaks are observed at 258.8 ± 0.5 and 77.6 ± 0.5 meV, which are assigned to the H–Si stretching and bending modes of H:Si(110), respectively. The D–Si peaks appearing at 187.7 ± 0.5 and 53.7 ± 0.5 meV were assigned to the D–Si stretching and bending modes of D:Si(110), respectively. Generally, on Si(110), the level of contamination by hydrocarbons and oxygen was higher than that on Si(111). This might be due to the corrugated microscopic structure. The low-coverage D signals on Si(110) often became unrecognizable in the presence of hydrocarbon signals.

For H:Si(110), Matsushita et al. observed the stretching mode at 259.1 ± 0.5 meV and the bending mode at 77.4 ± 0.5 meV by HREELS,⁷⁾ and Clark et al. observed the stretching mode at 2089.1 cm⁻¹ (258.8 meV).²⁶⁾ The reduced-mass argument applies to H/D:Si(110) similarly to Si(111). A blue shift of 2.8% for the stretching mode and a red shift of 3.2% for the bending mode for the D–Si vibrations from $1/\sqrt{2}$ times the H–Si frequencies are observed, anticipated to be related to the subsurface phonon modes. We are currently investigating the phonon dispersion on D:Si(110).

From Figs. 2 and 3, we can extract two pieces of important information on the H/D exchange reaction. One is the H/D peak intensity and the other is the H/D peak frequency, both of which vary as functions of the KF/D₂O immersion time. In this section, we use the H/D peak intensity and analyze the kinetics of the H/D exchange reaction on Si(111)/(110) in contact with KF/D₂O solution. The peak areas and positions were calculated by fitting every loss peak to a Gaussian function. Here, we only consider H/D–Si stretching peaks because of their accountability for background subtraction. Then each of the HREELS curves is associated with the H/D intensity ratio.

As the Si(111)/(110) surfaces were always completely terminated by H or D during the isotope exchange process, the surface coverages of H and D, θ_H and θ_D, respectively, must always satisfy

$$\begin{equation} \theta_{\text{H}} + \theta_{\text{D}} = 1, \end{equation} \tag{ 1 }$$

and therefore θ_H and θ_D will both be fixed if the ratio θ_D/θ_H is fixed. Owing to the isotope effect on the electron scattering, the oscillator strength of the D–Si stretching mode is 51.4% that of the H–Si stretching mode.^28,29,36) In this study, we multiplied this factor and converted all HREELS peak intensity ratios into θ_D values.

Figure 4 shows the time courses for θ_D on Si(111) and Si(110). On Si(111) [Fig. 4(a)], θ_D increased linearly up to the full surface coverage. The rate of exchange, defined as dθ_D/dt, is almost constant at 7.5 × 10⁻⁵ monolayer/s at room temperature in KF/D₂O. This is extremely low compared with 0.0067 monolayer/s, which is the exchange rate from D to H reported by Luo.²⁹⁾ This constant isotope exchange rate can be associated with the step-flow mechanism of etching Si(111) involving the removal of surface Si atoms into the etching solution. We initially fabricated H:Si(111) before its immersion in KF/D₂O, and the removal of the sacrificial SiO_x layer was completed before its immersion. The process in KF/D₂O only involved the removal of Si atoms from the Si substrate. The step-flow mechanism was proposed by Hessel et al.³⁷⁾ and Hines³⁸⁾ on the basis of their ex situ STM observations during Si etching. Once the pre covering SiO_x has been removed, the density of step lines on the surface will be constant as long as the removal of Si atoms only takes place at the step lines. This is the mesoscopic mechanism for a constant isotope exchange rate under a constant supply of etching reagent molecules and constant temperature.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 5 shows ultrahigh-vacuum STM images and cross-sectional height profiles of H:Si(111). In Figs. 5(a) and 5(b), the step and terrace structures are clearly observed. The average step height is 0.314 nm, matching the monoatomic step height of Si(111). Figure 5(c) exhibits the hexagonal Bravais lattice, and Fig. 5(d) indicates that the first-nearest-neighbor distance of each H atom is 0.39 ± 0.02 nm, which is close to the ideal distance of 0.384 nm. These images and the height profiles provide supporting evidence of the step-flow mechanism, proposed previously,^37,38) taking place in our case. In Fig. 5(a), it can be observed that the step lines along the $[\bar{1}10]$ direction are almost straight as a result of the ideal step-flow etching process.^37,38) These $[\bar{1}10]$ step lines are terminated by monohydride species (H–Si) and inert in the etching reaction. The kinks on the $[\bar{1}10]$ step lines are H₂–Si sites, captured by STM as a frozen structure upon emersion from the etching solution. The overall density of the $[\bar{1}10]$ step lines is temporarily constant, and the constant isotope exchange rate can be revealed. The same process is anticipated in KF/D₂O.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In contrast, in Fig. 4(b), Si(110) exhibits a rapid initial increase in θ_D that slows down with time. The initial rate of H/D exchange in KF/D₂O at room temperature [the same as that for Si(111)] is calculated to be (dθ_D/dt)_t=0 = 1.4 × 10⁻³ monolayer/s, which is 1 order of magnitude higher than that on Si(111). The isotope exchange of H/D on Si(110) must also be accompanied by surface Si etching.

Ye et al.¹⁹⁾ first observed the electrochemical etching process of H:Si(110) in aqueous NH₄F solution by in situ electrochemical STM and revealed many stripe-shaped islands with the longitudinal direction along $[1\bar{1}0]$. The longitudinal edges of these islands were terminated by monohydride H–Si, as determined from the crystallographic structure of Si(110), and therefore the etching on these edges was expected to be slow. The main active sites were at the ends of these stripes, which are terminated with dihydride H₂–Si and were assumed to be the locations of Si atom removal. Since a stripe island was removed from the ends along the stripes, this was called a "zipper reaction". Ye et al. actually observed the rapid retraction of island ends by in situ STM.¹⁹⁾ In the initial stage of surface etching, the surface was rough and the stripe ends were densely populated; therefore, the rate of etching was high. As the etching proceeded, the surface became flat, the number of ends decreased, and the total etching rate decreased.

This mechanism can account for the rapid H/D exchange in the initial stage and the slowing down in the later stage. We also performed a similar observation of the striped structure on H:Si(110) by ultrahigh-vacuum STM after etching in the same 40% NH₄F as in Ref. 7. The etching time in Ref. 7 was shorter than 15 min and roughly matches the time scale of Fig. 4(b). Furthermore, we are now attempting to investigate the zipper reaction of H:Si(110) in a longer time range by mesoscopic morphological observation.

Figure 6 displays the stretching vibrational energies for H–Si and D–Si on Si(111) and Si(110) as functions of the deuterium-exchange ratio (θ_D). On every HREELS spectrum that can clearly resolve the H–Si and D–Si stretching peaks, the peak energy losses were read out, and the deuterium-exchange ratio θ_D was calculated for each.

Zoom In Zoom Out Reset image size

On Si(111), the H–Si stretching peak was observed to be almost constant at approximately 258.2 meV and the D–Si stretching peak was scattered at approximately 187.3 meV within the readout error of the peak position, calculated in the procedure of curve fitting. A similar result was obtained on H/D:Si(111) by attenuated total reflection infrared spectroscopy by Luo and Chidsey.²⁹⁾ To theoretically estimate the vibrational energy shift, attempts have been made to involve the dipole–dipole interaction of H–Si vibrators. The energy shift derived from the dipole–dipole interaction was estimated to be as small as 0.6 meV by Jakob et al.³⁹⁾ and 0.2 meV by Miyauchi et al.⁴⁰⁾ It is difficult to measure an energy shift of this scale in the present experiment. The nearest-neighbor distance for H–Si on (111) is 0.384 nm, a relatively long distance between surface adsorbates in general. Most adsorbate–adsorbate interactions probably do not exhibit visible changes.

On Si(110), the H–Si stretching energy remains constant at approximately 258.8 meV then starts to decrease at θ_D = 0.7. This energy is regarded to correspond to the symmetric stretching mode of H–Si at 259.1 meV.^7–9) At θ_D = 0.76, the peak position decreases to 257.3 meV, which is almost exactly the energy of the antisymmetric stretching mode of 257.2 meV.^8,9) The present data set for the D–Si stretching energy does not involve distinctive changes, and is scattered at approximately 187.7 meV within the peak readout error. The D–Si stretching frequency exhibits an apparent shift near θ_D = 0. Because of the low D–Si peak intensity near this end and the strong background due to contaminant overlapping with the D–Si stretching position, we could not obtain concrete values of D–Si peak energies.

It is clear that a certain interaction induces a red shift of the H–Si stretching frequency on Si(110). This is natural in regard to Fig. 1(b), in which the H–H nearest-neighbor distance is 0.299 nm, notably shorter than that on Si(111). Actually Matsushita et al. demonstrated that, on H:Si(110), a visible HREELS profile occurs between H atoms owing to the shorter nearest-neighbor distance of H atoms than that on H:Si(111) by measuring the surface phonon dispersions.^7–9) A surface phonon is a coherent vibrational excitation of surface atoms propagating along the surface, and a perfect array of H–Si is required for this propagation. The present result indicates that the vibrational coherence is decreased for θ_D larger than 0.7 with random arrangement of H and D on the chains.

We are now confident that we can control the perfection of H-terminated Si(111) and Si(110) by adjusting the terminating H/D isotope ratio as desired. The observation of the D:Si(110) phonon dispersion is planned, and we will be able to double the data for the adsorbate–adsorbate interaction. This will essentially add more information for discussing the interactions on theoretical and computational bases. This will provide a driving force for improving the precision of quantum-mechanical and molecular-dynamic predictions of surface vibrational motion on semiconductor surfaces in general.

It is rational to anticipate that the geometrical structures of D:Si(111) and D:Si(110) are respectively the same as those of H:Si(111) and H:Si(110), because the atomic radii of H and D are much smaller than the two-dimensional Si unit cells on the surface and no steric hindrance should occur between H/D adsorbates. Actually, our previous report provided a rough confirmation of the (1×1) structure on D:Si(111) obtained by LEED observation.²¹⁾ This time we performed a more careful LEED observation on both D:Si(111) and D:Si(110), confirmed to be 100% deuterated by HREELS, to verify that the two-dimensional periodicities are practically identical to those on H:Si(111) and H:Si(110), respectively.

For the D-terminated samples, we employed the KF/D₂O etching periods of 180 and 30 min to saturate the Si(111) and Si(110) surfaces with D, respectively. Figure 7 shows typical LEED images of H:Si(111) and D:Si(111). Regular three fold rotationally symmetric patterns were observed. The LEED patterns for H and D exactly match each other at incident electron energies of 40 and 110 eV. There is no difference in the unit cell constants for H:Si(111) or D:Si(111) detectable by LEED. The spots are sharper than those in the D:Si(111) LEED pattern observed previously.²¹⁾ The new method of D:Si(111) preparation is simpler and more convenient than the previous method.²¹⁾

Zoom In Zoom Out Reset image size

Figure 8 shows LEED images of the D:Si(110) surface obtained for various incident electron energies ranging from 30 to 200 eV. The rectangular diffraction pattern indicates the presence of reflection symmetry along the [001] direction and two fold rotational symmetry axes, which are perpendicular to the surface as shown in Fig. 1(d). The aspect ratio of the rectangular unit cell is $1:\sqrt{2}$, which corresponds to the bulk Si(110)-(1×1) structure (11.6 nm⁻¹: 16.4 nm⁻¹). Our previously reported photographs of H:Si(110) were used⁷⁾ for comparison. The LEED patterns again match the previous ones almost exactly. The characteristic extinction of specific spots for the p2mg symmetry is also common to D:Si(110) and H:Si(110). Note that the spots corresponding to (10), $(\bar{1}0)$, (30), and $(\bar{3}0)$ are missing for all energies, in accordance with the two-dimensional extinction rule.^8,41,42)

Zoom In Zoom Out Reset image size