Combining scanning tunneling microscope (STM) imaging and local manipulation to probe the high dose oxidation structure of the Si(111)-7×7 surface

Understanding the atomistic formation of oxide layers on semiconductors is important for thin film fabrication, scaling down conventional devices and for the integration of emerging research materials. Here, the initial oxidation of Si(111) is studied using the scanning tunneling microscope. Prior to the complete saturation of the silicon surface with oxygen, we are able to probe the atomic nature of the oxide layer formation. We establish the threshold for local manipulation of inserted oxygen sites to be +3.8 V. Only by combining imaging with local atomic manipulation are we able to determine whether inserted oxygen exists beneath surface-bonded oxygen sites and differentiate between sites that have one and more than one oxygen atom inserted beneath the surface. Prior to the creation of the thin oxide film we observe a flip in the manipulation rates of inserted oxygen sites consistent with more oxygen inserting beneath the silicon surface.


Introduction
Silicon dioxide is the most commonly used insulator and gate dielectric in integrated circuits and is also an important material for integration with emerging materials [1][2][3][4]. The oxidation of the silicon surface creates states which modify charge transport through the silicon, with these interface effects becoming more prominent as devices are scaled down. The scanning tunneling microscope (STM) is an ideal tool to perform local nanoscale chemistry directly on individual O-Si bonding sites. The STM can induce molecular manipulation via local injection of electrons directly into single atoms and molecules. This manipulation may induce weakening (diffusion, switching) or breaking (desorption, dissociation, and transformation) of bonds, probing the chemistry and energetics of the surface [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20].
As well as exploring the chemistry of adsorbed oxygen, the STM can be used as an atomic-resolution imaging tool to study where oxygen molecules react with the surface. Understanding the initial oxidation of the silicon surface is important to grow thin and high quality oxide films [21][22][23]. Therefore, the structure and electronic properties of the O-Si interface have been examined with methods including X-ray diffraction [24], electron energy-loss spectroscopy [25], corelevel spectroscopy [26], photoelectron spectroscopy [27], and near edge X-ray absorption fine structure spectroscopy [28]. Although the oxidation states of SiO2 on Si can be examined with these methods, the STM can distinguish between different oxidation states spatially and directly identify the atomic positions at which reactions take place [16][17][18][29][30][31][32][33][34][35][36][37][38][39].
In this work we study the initial oxidation of the Si(111) surface as a function of oxygen exposure in ultra-high vacuum using the STM. When the oxidation of silicon is studied at high dose, the surface shows only saturated silicon adatoms and at still higher doses the formation of the disordered oxide layer [40]. There remains a dosing regime between the initial oxidation and the formation of this disordered oxide layer where several structures of inserted oxygen atoms and surface oxide are still visible at atomic resolution and can be manipulated by local injection of electrons to elucidate their nature. Previously this was done for low dose oxygen exposures of 0.05 and 0.1 L, which using measured sticking coefficients corresponded to 0.025 monolayer (ML) and 0.005 ML coverage, respectively [29]. Here we extend this to higher doses of 0.3 and 0.4 L, which using typical sticking coefficients corresponds approximately to 0.015 and 0.02 ML coverage, respectively. We observe behavior that bridges existing low dose results up to the reported high dose behavior. By combining imaging and manipulation of the initial inserted oxygen we offer clearer understanding of how these sites transform under increasing oxygen exposure.
were prepared in 2M NaOH solution, and then outgassed by direct current annealing in vacuum. Silicon (p-type boron doped, 0.01-0.02 Ω·cm, Goodfellow) was annealed in situ to achieve the Si(111)-7×7 reconstruction [16], before exposure to 0.3 and 0.4 L oxygen (99.6% purity with typical analysis of 0.4% total hydrocarbon (TCH) as CH4 < 20 volumetric part per million (vppm) and H2O < 3 vppm, Argo International Ltd). Passive STM imaging was performed before and after injection to monitor manipulation events, using tunneling conditions where no transformations of reacted sites were observed (constant current, +1.5 V, 100 pA). Drift-tracking was enabled to minimize unwanted tip displacement due to piezo creep following charge injection. All manipulations were performed with the feedback loop enabled with increasing bias voltage at a constant tunneling current of 6 nA for 50 ms.

Results
In the oxidation of the clean Si(111)-7×7 surface, oxygen atoms bond with either silicon surface adatoms or the rest-atoms sat lower in the surface reconstruction. If oxygen molecules are dissociatively adsorbed by the silicon adatoms, the surface adatom dangling bonds at ~ 0.3 eV above and below the Fermi level (EF) are removed, reducing the density of states for the STM to tunnel into [28,[40][41][42][43][44]. These "dark sites" are visible in Fig. 1(a) at +1.0 V gap voltage as less bright atoms within the usual surface reconstruction. If instead one or more oxygen atoms are inserted into the back bonds between the adatoms and rest atoms, the surface adatom and its unreacted bonds are shifted ~ 0.5 eV higher in energy compared to the unreacted adatom dangling bond site [40]. The outward relaxation of the unreacted silicon adatom also moves it physically closer to the tip [41,45]. We find these "bright sites" are not visible at the same gap voltage but when it is increased to +2.0 V in Fig. 1(b) the shifted surface states do become accessible to the STM and bright sites appear, indicated by circles. Figure 1(c) shows a unit cell of the Si(111)-7×7 reconstruction (white rhombus) with its faulted half (FH) and unfaulted half (UH) sites. Corner and center positions for the FH and UH are indicated on the image by the red and blue arrows, respectively.
While the initial stages of the oxidation of the Si(111) surface have been studied at atomic resolution before for low doses, we extend this investigation to 0.3 and 0.4 L oxygen exposure at room temperature. First we investigate the nature of the oxidation by identifying the atomic positions at which reactions take place. The number of bright and dark sites were counted and sorted by the four different possible bonding positions within the unit cell with all data normalized over a 56 nm × 56 nm area after subtraction of surface defects from the number of dark sites; presented in Table 1.
When increasing the oxygen dose from 0.3 to 0.4 L the number of dark sites increased on both the faulted and unfaulted halves. The number of bright sites decreased for almost all sites but remained the same for faulted corner sites, because of the lower number of dangling bonds. Other work has found that with low oxygen exposure of 0.05 L, bright sites are prominent mostly on the faulted side of the unit cell [30].
Combining our data with existing work by Mayne for lower dose 0.05 and 0.1 L oxygen exposure [29], we present the combined effects of increasing dose on the number of bright and dark sites observed with the STM in Fig. 2. At low doses more bright sites form than dark sites, but even in the transition from 0.05 to 0.1 L the relative increase in dark sites is higher than bright sites. We confirm that this trend increases to higher doses and observe the region where dark sites now dominate, but bright sites, which represent inserted oxygen without any surface oxygen, are still present. Beyond this coverage it has been found that only dark sites are observed prior to the formation of the disordered oxide overlayer, which we also observed when the oxygen dose was increased further [40,46,47].
Combining this with existing work on O-Si oxidation, we begin to construct a model of the oxidation pathway leading to the dark and bright sites viewed with the STM, shown in Fig. 3. While it is possible to directly form dark sites by oxygen bonding to silicon adatoms without any inserted oxygen, the higher number of bright sites at low dose oxygen exposure suggests that bright sites are predominantly formed. These represent the dominant initial state with an oxygen atom inserted in one of the back bonds, as seen in Fig. 3(b). With further oxygen exposure it is possible for more oxygen atoms to insert into a second back bond maintaining the bright site, or for oxygen to bond to the silicon adatom dangling bond creating    a dark site- Fig. 3(c). This behavior can be explored further by using the STM to inject electrons directly into the silicon adatom and determining whether the associated oxygen desorbs, dissociates or transforms. Inserted oxygen is known to be more difficult to manipulate, so we first determine the voltage threshold required for bright site manipulation by measuring the number of injection sites which undergo any such reaction as a function of gap voltage, shown in Fig. 4. We determine the threshold to be +3.8 V, where the manipulation occurs because of electronic excitation of a surface state associated with the inserted oxygen. Although increasing the bias voltage increases the tip-sample distance in constant current mode, the electric field still increases [48], which leads to the observed increased manipulation above the voltage threshold [8].
Three examples of local site manipulations for dark sites are presented in Fig. 5. In Figs. 5(a) and 5(b) the dark site is displaced to a neighboring adatom site, indicated with a white arrow. In this manipulation the surface oxygen is transferred from the original adatom to the neighbor, resulting in a displaced dark site. The original adatom would almost certainly not have any inserted oxygen in the back bonds, otherwise with the surface oxygen removed, the original atom would become bright. In Figs. 5(c) and 5(d) the dark site transforms into a bright site. This represents either the removal of the surface oxygen revealing a bright site which already had inserted oxygen, or the manipulation of the surface oxygen into a back bond. In the  The three possible manipulations of bright sites are shown in Fig. 6. In Figs. 6(a) and 6(b) the original bright site remains and a new dark site is created on a neighboring atom, described in the literature by the notation B→B+D. This manipulation can only occur if there are at least two oxygen atoms in back bonds under the original atom. The current injection causes one of them to move to the neighboring adatom, reacting with the dangling bond to create a new dark site, while at least one oxygen remains in the original back bond maintaining the bright site. In Figs. 6(c) and 6(d), a bright site is transformed into a dark site in the same position (B→D). This represents an oxygen in a back bond coming to the surface to create a dark site instead. In the last example for bright site manipulation in Figs. 6(e) and 6(f), the bright site is removed (the adatom returns to background contrast) and a new dark site is created near the injection site (B→_+D). This represents most likely a single back-bonded oxygen migrating to a neighboring adatom to create the new dark site. Thirty-eight bright site injections here at high dose yielded thirteen successful manipulation events which are shown in Table 2 combined with results from twenty-three bright site manipulations at low dose performed by Mayne [29]. At low dose the predominant observed manipulation was for a bright site to transform to a dark site (B→D) which represents an oxygen in a back bond coming to the surface to bond with the adatom instead. However, for our higher dose oxygen exposure there is almost a direct flip in the table between the proportion of this B→D manipulation and the bright site remaining plus a new dark site creation on a neighboring adatom (B→B+D). This manipulation can only occur if the injection site has at least two oxygen atoms in back bonds under this original site. The swap in proportion of these two manipulations suggests that at higher oxygen doses more oxygen atoms are inserted into back bonds on bright sites, yielding a large increase in the number of transformations that are characteristic of multiple inserted oxygen atoms, while those that do not require multiple inserted oxygens reduce in turn.

Discussion
It is now generally accepted by both experimental and density functional theory (DFT) studies that molecular oxygen on Si(111)-7×7 is unstable and spontaneously dissociates upon adsorption [41,42,49,50]. Some DFT studies assume that both of these oxygen atoms must then be incorporated in to the surface [49], however our manipulation studies show that two single-oxygen configurations are possible. Firstly, we find that dark sites can be formed where oxygen bonds only to a surface adatom, with no oxygen inserted into back bonds, usually abbreviated as the ad structure [44]. This ad structure has been considered before in experiment and DFT modeling [42,44]. While it cannot be detected with STM imaging, we find evidence for it here with local manipulation. Secondly, we are also able Table 2 Comparison of bright site manipulations via STM current injection for low dose oxygen (Mayne [29]) and higher dose (this work)

Manipulation
Low to confirm the singly-inserted oxygen "ins" structure using manipulation. Atomic force microscopy (AFM) can also differentiate between the singly-inserted (ins) and doublyinserted (ins×2) structure from the increase in the upward relaxation of the silicon adatom above [45]. Our oxidation pathway therefore starts from the assumption of spontaneous dissociation of molecular oxygen and initially considers just one oxygen atom reacting with the surface. At low dose the predominant form is for oxygen to insert in to a back bond, lifting up the relevant adatom and producing a bright site. The absolute number of these inserted-oxygen bright sites increases with dose, but imaging with the STM alone cannot determine how many oxygen atoms have been inserted into back bonds-All sites appear equally brighter. However, by studying the manipulation rates we are able to distinguish between singly-inserted (ins) and at least doubly-inserted oxygen (ins×n, n = 2, 3) sites. We find that although it is possible to have more than one oxygen atom inserted for a bright site even at low dose, the increase in the proportion of B→B+D manipulations at high dose, and the corresponding reduction of B→D manipulations, suggest that predominantly single atom insertion occurs at low dose. This agrees with DFT studies which find that the inserted oxygen structures (ins×n, n = 1, 2, 3) are the most stable and form preferentially. Therefore, our oxidation pathway shows the primary singly-inserted ins structure as the first step.
Local manipulations requiring at least two inserted oxygen atoms increase at higher dose while manipulation processes that do not require multiple oxygens reduce, suggesting that the proportion of inserted oxygen atoms increases with dose. These (at least) doubly-inserted bright sites (ins×n, n = 2, 3), shown in the bottom portion of Fig. 3(c), would be indistinguishable in STM imaging at +1.5 V from the singly-inserted oxygen bright sites (ins), but behave statistically differently during manipulation via current injection. Both forms of bright site eventually go on to accept an oxygen on the adatom which transforms them to a dark site, which forms the final steps in our pathway.
With the direct creation of ad sites, as well as the conversion of existing bright sites into dark sites (ins×n → ad-ins×n), dark site creation overtakes the number of bright sites as dose increases. Further dosing leads eventually to a surface dominated by oxygen-saturated dark sites, followed upon continued oxygen exposure to the creation of thin oxide films on which atomic resolution STM is no longer possible, as observed by us and others [40]. This is also in agreement with DFT studies which find that after inserted structures, the most favorable form energetically is for the combined inserted and adsorbed oxygen forms (ad-ins×n) [49].
As well as dose affecting which structure forms, the specific bonding site within the surface reconstruction also changes. Previous STM studies have shown a 2:1 preference for inserted bright sites to bond to the faulted half of the unit cell at low dose compared to the unfaulted half, confirmed by DFT to be energetically preferential [49,51,52]. We find the same ratio of 2.0 at 0.3 L, increasing to 2.6 at 0.4 L. We also find a preference of dark sites to form in the faulted half with a ratio to the unfaulted half of 1.6 at 0.3 L increasing to 2.3 at 0.4 L. DFT studies confirm that the faulted half is also preferential for ad-ins structures, in agreement [49]. In line with DFT and other experimental studies, we find that the inserted bright sites preferentially form at faulted corner hole sites with a ratio of 2.0 for 0.3 L increasing to 2.3 at 0.4 L [49]. We find the only exception to the corner hole preference is for dark sites at the higher 0.4 L dose, where center holes become preferential on both the faulted and unfaulted half of the unit cell. Niu et al. found that the ad-ins×3 structure preferentially forms at center holes on both the faulted and unfaulted halves [49]. The ad-ins×3 structure, which requires repeated oxidation, would be expected to form more at higher dose. We therefore match experimentally their predication of increasing center hole reactions for this high dose structure.
In summary, STM imaging alone cannot distinguish between singly-and (at least) doubly-inserted oxygen sites, nor can it determine whether a saturated dark site sits on top of inserted oxygen. However, combining STM imaging with local atomic manipulation can reveal the dose-dependent behavior of these sites and inform our understanding of how oxygen reacts with the silicon surface to form the initial oxide.
Open Access: This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.