During the last decade, silicon nanowires (Si NWs) have been studied extensively to be employed in the modern electronic industry in the direction of the size reduction and efficiency boost of the devices 1 . Because of the high surface to volume ratio, Si NWs’ properties depend firmly on their surface conditions and surface terminations, in particular. The oxidation of Si NWs, when exposed to ambient air, is believed to have a detrimental effect on their electrical properties due to the low quality of the oxide, giving rise to the uncontrolled interface states and enhanced carrier recombination rates 2 . This necessitates protection of Si NWs’ surfaces against oxidation via termination by various chemical moieties (i.e., alkyls and alkenyls) 3 4 . However, to better prevent oxide formation, a deeper understanding of the Si NW’s oxidation mechanisms and kinetics is essential. For planar Si, the widely known Deal-Grove (DG) model considers the interfacial oxidation reaction and oxidant diffusion as the major rate-determining reaction steps for short and long oxidation times, respectively 5 . DG model has undergone a number of modifications due to imprecise prediction of the oxidation behavior at low temperatures (T ≤ 700°C) in convex/concave surfaces and for very thin oxide layers 6 7 8 . Specifically, in sufficiently small Si NWs (d ≤ 44 nm), oxidation can be completely retarded by the compressive stress normal to the oxide/NW interface 9 10 . Nevertheless, the studies on the oxidation mechanisms of Si NWs have been focused mostly on the formation of thick oxide layers at relatively high temperatures and long times, overlooking the early stages of oxidation which involve removal of surface functionalities and suboxides formation.

In this article, thermal stability of hydrogen-terminated Si NWs of 85-nm average diameter was investigated by means of the surface-sensitive X-ray photoelectron spectroscopy (XPS) for a variety of temperatures and times. H-terminated surfaces are of importance since they are considered as the starting surfaces for further functionalization of Si NWs 11 12 13 14 15 . The different kinetic behavior for the three transient silicon suboxides and SiO₂ has been shown. Growth regimes were mainly addressed by four different phenomena including Si-Si backbond oxidation, surface bond propagation, suboxide growth site formation, and self-limited oxidant diffusion. A preliminary oxidation mechanism, elucidating the influence of time and temperature on the role of latter factors, is outlined.

The VLS-grown Si NWs used in this study were randomly oriented with average diameter and length of 84.96 nm and 3.508 μm, respectively. The pristine Si NWs are covered by a native oxide layer of 1 to 4 nm. SEM and transmission electron microscopy (TEM) micrographs of the pristine Si NWs are depicted in Figure 1. Residual gold nanoparticles were removed by rinsing the Si NWs into HNO₃ solution preventing its catalytic effect on oxidation.

As an illustrative Si 2p spectrum of oxidized Si NWs, the Si 2p spectrum of the H-terminated Si NWs annealed at 500°C for 60 min is depicted in Figure 2. By formation of even very thin silicon dioxide layers, the Si 2p XPS survey of Si NWs changes, showing a peak between the binding energies of 102 to 104 eV. To quantitatively evaluate the oxidation process, Si 2p spectral decomposition was conducted on the spectra after Shirley background subtraction, through a curve-fitting procedure using Gaussian-Lorentzian functions 16 . Consequently, the Si 2p spectra can be divided into six different sub-peaks including two silicon spin-splitting peaks as Si 2p _1/2 and Si 2p _3/2, three silicon sub-stoichiometric oxides (known as suboxides) peaks as Si₂O, SiO and Si₂O₃, and the silicon dioxide (SiO₂) peak. The chemical shifts (Δ) of the sub-peaks obtained in Figure 2 relative to the Si 2p _3/2 (at 99.60 ± 0.02 eV) are as follows: Si 2p _1/2 (Δ = 0.60 eV), Si₂O (Δ = 0.97 eV), SiO (Δ = 1.77 eV), Si₂O₃ (Δ = 2.50 eV), and SiO₂ (Δ = 3.87 eV).

The amount of each of suboxides, relative to the amount of intact silicon, can be calculated by dividing the integrated area under the suboxide’s peak (A _SiOx) by the sum of the integrated area under Si 2p _1/2 and Si 2p _3/2 peaks (A _{Si 2p1/2} + A _{Si 2p3/2}). The resulting value is called suboxide intensity, shown by I _SiOx. In addition, total oxide intensity (I _ox) can be calculated as the sum of all the four suboxide intensities (I _ox = I _Si2O + I _SiO + I _Si2O3 + I _SiO2). Oxide intensity can also be expressed in number of monolayers, regarding the fact that each 0.21 of oxide intensity corresponds to one oxide monolayer 17 . The total oxide intensity, besides suboxide intensities for the Si NWs specimens annealed at 150°C and 400°C, is listed in Table 1. Except SiO₂, all the suboxide intensities for both of the annealing temperatures are comparable and more or less show very slight variations over the annealing time. However, at 150°C, suboxides hold a larger share of the total oxide intensity whereas at 400°C, SiO₂ mainly contributes to the overall oxide amount detected.

Variation in the total oxide intensity (I _ox) for all the six temperatures over oxidation time up to 60 min is shown in Figure 3. For both the high temperature (T ≥ 200°C) and low-temperature oxidation (T < 200°C), the oxide intensity reaches a saturation level beyond which the oxide amount grows negligibly. However, in low-temperature oxidation, the time to reach 80% of the saturation levels (defined as Γsat) is in the range of 20 to 30 min, whereas in high-temperature oxidation it ranges from 8 min to 12 min. Average Γsat for high- and low-temperature oxidation are marked in Figure 3 by dashed and dotted lines, respectively. This indicates roughly both similarities and differences between the underlying oxidation mechanisms in these two temperature ranges. The presence of the saturation levels reveals the fact that a mechanism is hindering further oxide growth after formation of a certain oxide level. On the other hand, the growth rates differ between the two temperature ranges, revealing the existence of mechanisms with different thermal activation energies.

The growth of oxide in planar silicon in thick layers and at high temperatures has been successfully expressed by the Deal-Grove model. However, it breaks down in very thin oxide layers and has been modified considering the suboxides as nucleation sites (or oxide growth sites) that are necessary for oxide build-up 6 . Through high-temperature oxidation, silicon suboxides exhibit relatively constant values after a sharp increase in their intensities. Therefore, in the early stages of Si NWs oxidation, formation of the growth sites composed of suboxides can be taken into account as the major mechanism.

Further oxidation and rise of the flat tail indicate existence of a second mechanism, which is impeding oxide formation at the suboxide growth sites. In Si NWs, such retarded oxidation behaviors have mostly been attributed to their geometry and presence of compressive stresses normal to the silicon/silicon oxide interfaces that limit further oxide growth and its expansion 8 10 . Nevertheless, compressive stresses are more expected for NWs of diameter below 44 nm which is far below the average diameter of the Si NWs studied here 9 . Additionally, comparison between Si NWs and planar Si(100) oxidation behavior in the same time and temperature ranges showed similar flat tails of oxide 18 . Therefore, the retarded oxidation in Si NWs, in analogy with planar silicon, can be attributed to the self-limited oxidation caused by the act of firstly formed oxide layer as a diffusion barrier 19 . The two mechanisms are summarized in Figure 4.

At lower temperatures, increase of the total oxide intensity is accompanied by the rise in the intensity of suboxides with amounts comparable to SiO₂ intensity (Table 1). Backbond oxidation can be considered as the primary mechanism causing formation Si-O-Si bonds below the surface-terminating Si-H bonds. The backbonds can be oxidized in different oxidation states and can finally form the full oxide layer atop. Compared to planar samples, Si NWs exhibit faster backbond oxidation, indicating the effect of circumferential tensile stresses on the stability of Si-Si bonds 18 . For longer oxidation times, upon formation of a larger number of oxidized backbonds, isolated Si-OH bonds start to form upon interaction of Si-H and Si-O bonds in the oxidized backbond 20 . By completion of the backbond oxidation, besides the Si-OH formation, remaining Si-H surface bonds start to rupture and hydrogen propagation begins. Low-temperature oxidation mechanism is summarized in the scheme illustrated in Figure 5. It should be noted that the hydroxyl groups shown in Figure 5 represent both the isolated hydroxyl groups formed throughout the oxidation and after completion of oxidation as frequent for SiO₂ in H₂O-containing environments 21 .

In conclusion, the growth kinetics of the suboxides and silicon dioxide is highly dependent to temperature and time. At lower temperatures, oxidation is first controlled by backbond oxidation. After full oxidation of the backbonds, Si-H bond rupture dominates the process kinetics. At higher temperatures, suboxide nucleation sites (known as oxide growth sites) control the early stages of oxidation. After complete formation of the very first oxide monolayers, further oxidation is self-limited as the oxidant’s diffusion through the oxide layers is impaired. These findings suggest a perspective on more efficient methods to stabilize Si NWs against oxidation over the long term.

SEM: Scanning electron microscopy; Si NWs: Silicon nanowires; VLS: Vapor–liquid-solid; XPS: X-ray photoelectron spectroscopy.

The authors declare that they have no competing interests.

MYB and KS carried out the experiments and wrote the article. JR and SHC conceived of the study and participated in its design and coordination. All authors read and approved the final manuscript.