In recent decades, the compensated mechanisms of the polar surface of ionic crystals, which are closely related to its chemical and physical properties as well as the mechanisms of film epitaxy [1, 2], are a fundamental issue in surface science. So far, five possible compensated mechanisms accounting for the stability of polar surfaces have been proposed, namely electron transfer [3–5], adsorption of species [6–9], surface reconstructions [10], formation of neutral terrace [11, 12], and complementary stabilization based on surface disorder [13]. The surface-reconstruction mechanism reflects that the polar surfaces probably present richly large area atomic reconstructions, as observed in some multiple element ionic compounds [e.g., SrTiO3(011)] or weak ionic crystals [e.g., TiO2(101)] systems [14–16]. These large-scale atomic-reconstruction surfaces often have widespread applications in multiple fields such as serving as substrates, offering platforms for understanding the catalytic mechanism, and contributing to the preparation of high-quality polar-to-polar interface of functional oxides which have been predicted to emerge topological interface states [17]. However, the polar surfaces in strong ionic crystals such as many binary bulk-terminated ionic crystals [e.g. ZnO (0001)] do not follow this expected behavior and have disorder or neutral plane structures demonstrated by both experimental and theoretical studies [7, 11, 12, 18]. Such restriction of the polar surface for bulk-terminated ionic crystals greatly limits its potential applications. Therefore, a key question is naturally raised: can a large area of atomic reconstruction in bulk-terminated ionic crystals be experimentally prepared? If this problem is overcome, another key question is the observation of the complex evolution pathways from the initial disorder state to the well-order phase with atomic reconstructions, which is centrally important to understand the chemical and physical properties of ionic crystals as well as the mechanisms of corresponding film epitaxy.
To fundamentally address these challenges, here we take inspiration from covalent-crystal systems, in which the polar surfaces tend to exhibit rich reconstructions [19, 20], as suggested by Table S1, which lists the reconstruction details of several materials ranging from covalent to ionic crystals. Generally speaking, the important distinctions between covalent and ionic crystal polar surface are mainly reflected in two aspects: 1) with respect to the area charge density in every layer, atomic net charge is small for the covalent system but large for the ionic system, because the valence electrons are shared by multiple atoms in the former case but almost griped by the strong oxidized atoms (forming ions and cations) in the latter case; 2) with respect to the feature of the dangling bond states, the covalent surfaces have the partially filled-state images while the ionic surfaces have the fully empty-state or filled-state images [21]. Generally, it is difficult to change the area charge density for a compound system with stable chemical valences. Thus, an approach to realize the preparation of the large scale atomic reconstructions is proposed by the authors by, presumably, modifying the occupied states of dangling bonds by the surface electron doping. In our design, we use ZnO(0001), a prototypical polar surface widely used in optoelectronic devices, methanol synthesis, photocatalysis, and hydrogen sensing [22–24], as a model system for proving the above-mentioned conjecture as well as tracking the dynamic-evolutionary pathways from initially disorder to well-order phase. The surface electron doping, mainly referring to the transition between empty states and partially occupied states of dangling bonds, could be realized by doping Ga atoms with electronegativity and ionic radius close to those of Zn atoms onto the surfaces [25].
Electron doping the surface system is an efficient strategy to modulate the surface electronic structure, which is commonly achieved via low-dose atom deposition, e.g.,doping Sn onto the Si(111) surface [26] and doping Pb onto the Cu-based superconducting surface [27]. However, the atom-deposition method is not suitable for ZnO surface system, because it often leads to the emergence of cluster structures [28, 29], but not of desired ordered atomic structures at the surface using this method. Therefore, it is necessary to search for alternative and better preparation methods. Here, for the first time, we propose an Ga atom surface segregation approach [illustrated by a schematic diagram in Fig. 1(a)] to prepare plentifully atomic reconstructions on the ZnO surface. Surface segregation is neglected previously by investigators because it probably destroy the photoelectronic performance of semiconductor materials, but indeed it is ubiquitous phenomena in alloys at high temperatures where foreign atom seems unable to diffuse to surface from interior [30, 31]. In the present study, a commercial Ga-doped ZnO crystal (ppm = 0.005%, the material information and experimental details are provided in Methods section) is chosen to investigate the atomic reconstructions and dynamic-evolutionary pathways using the Ga atom surfaces segregation approach in combination with scanning tunneling microscopy (STM) and density functional theory (DFT).
First, we discuss the Ga atom segregation at the surface after treatment. Figure 1(b) shows the X-ray photoelectron spectroscopy (XPS) spectra of Ga-2p from the cleaned sample and the treated sample upon Ar + etching and annealing at 780°C. There are no Ga-2p peaks in the cleaned sample (black line) upon Ar + etching and annealing at 350°C for 30 min. After Ar + etching and annealing at 500°C for 10 min, very weak Ga-2p peaks appear (red line) [32]. Treating the sample again at 780°C for 3–5 min results in the appearance of strong Ga-2p peaks (blue line). These results show that some Ga atoms gradually diffuse to the surface from the body after treatment, producing Ga atom segregation at the surface, although the Ga content in bulk ZnO is as low as approximately 0.005%. According to the XPS spectra of Ga- and Zn-2p obtained at different depths [Fig. S1(a) and (b)], we obtained the Ga and Zn element contents [Fig. S1(c)]. The segregated Ga atoms stay at the surface with a depth of less than 7 nm after treatment [See the discussion in Fig. S1]. Besides, the distribution of Ga concentration from the topmost surface to the body obeys an exponential function, as illustrated by the red line in Fig. S1(c). This result is further confirmed by the energy dispersive spectroscopy (EDS) [Fig. 1(c) and Fig. S2], showing the distribution of Ga atoms (blue line) at the surface layer within a depth of about 7 nm. The schematic views of the Ga atom segregation from interior to subsurface is shown in Fig. 1(a). Besides, the increase in the doping concentration induces a transformation from the pure semiconductor to the n-type surface, as demonstrated by the dI/dV spectrum [Fig. S3]. This result shows that Ga atom segregation is able to realize surface electron doping, like the H adsorption at the subsurface of the pure ZnO(0001) surface [33].
Next, we discuss the finally well-order phase structure on the electron doping polar surface. As in pure ZnO [34, 35], the polar surface presents numerous pits and cavities [Fig. S4] on straight and zigzag steps with a height of c/2 after cleaning via Ar + bombardment and annealing at 500°C for 10 min. At this stage, the measured Ga concentration is less than 1%. After treating the sample by Ar + bombardment and annealing at 720°C–780°C for 3–5 min, the surface presents the final structure with perfectly column-like atomic reconstructions, as shown in high-resolution STM images in Fig. 2(a), which comprises two sets of the same reconstruction (labeled by the red and green circles) with slightly opposite slits parallel to the column direction (labeled by the red and green arrows), forming a zigzag atomic arrangement (illustrated by the light blue scattering line). Moreover, in the reconstruction, the nearest atomic distances are equal to 2.28 nm (≈ 4asin60°) and 0.63 nm (≈ 2a) along the directions vertical and parallel to the column, respectively, as marked by the red double arrows. This result indicates that the surface is a (2 × 8) reconstruction. Combining the simulation results and the experimental data [See the discussion in Fig. S5] obtained by the scanning tunneling spectroscopy mapping, a useful tool for the discrimination of surface atoms, it is verified that the larger (green and red circles) and smaller (violet circles) spots originate from the surface states of Zn and O atoms respectively, revealing the arrangement of Zn and O atoms in the reconstruction. The simulated result of STM image in Fig. 2(b) further confirms the (2 × 8) reconstruction. Based on these results, the atomic model can be drawn, as shown in Fig. S5(h).
Figure 2(c) shows the morphology of the reconstructed surface, and the inset shows the green profile line. The step height becomes several times of ~ c/2, different from that of the disordered surface morphology [Fig. S4(a)]. Besides, the reconstructed surface displays many defects and adsorbed species, as labeled by the red and green circles in Fig. 2(d). More interestingly, the (2 × 8) reconstruction has a two-fold symmetry, different from the symmetry of ZnO(0001). To maintain the same symmetry with ZnO(0001), the (2 × 8) reconstruction surface was divided into three distinct regions with the same normal but with a rotation of 60° from each other, as illustrated by the light blue regular triangle in Fig. 2(d). This phenomenon is similar to the incommensurate epitaxy of the film, such as the ZnO film on the MgO(001) substrate or the SrTiO3 film on the ZnO(0001) substrate [36, 37].
Next, we address the complex dynamic-evolutionary pathways from initial structure to the (2 × 8) reconstruction as well as the other atomic reconstructions appeared in the evolution process. Although STM images demonstrate no time resolution, the comparison of the STM images obtained at various stages from different regions of the surface offers us to reveal the different routes of the surface dynamic evolution. Fully disordered structures are chosen as the initial stage, as discussed in Fig. S4 and Fig. 3(a). At the next series of stage evolution, the STM images [Fig. 3(a)–(i)] show two dynamic-evolutionary pathways from disorder to the (2 × 4) surface phase. One pathway involves the formation of the (2 × 2) atomic reconstruction with numerous clusters at the localized regions or step edges [Fig. 3(b)], followed by its gradual growth throughout the whole terrace through these clusters, as illustrated in Fig. 3(c). The inset shown in Fig. 3(c) corresponds to the amplified image, showing that the clusters are a six-membered ring, as labeled by the blue circles. At the following stage, the whole region becomes a (2 × 2) reconstruction with few clusters [Fig. 3(d)]. These STM images [Fig. 3(a)→Fig. 3(d)] indicate that the dynamic-evolutionary pathway from disorder to (2 × 2) atomic reconstruction involves replacement of the disorder surface structures by a huge of random six-membered ring clusters, and then formation of (2 × 2) reconstruction gradually via numerous atom desorption. Following the sample treatment, we found that the (2 × 4) reconstruction appears in some localized regions via the evaporation of Zn atoms [Fig. 3(e)]. The (2 × 4) reconstruction phase extends gradually until the whole region becomes the (2 × 4) phase, as illustrated by the inset in Fig. 3(h) [Fig. 3(e) → Fig. 3(f) → Fig. 3(h)]. This result can be seen clearly obtained from the analysis of Fig. 3(f), which shows four different regions (region Ⅰ, region Ⅱ, region Ⅲ, and region Ⅳ). The former three regions (region Ⅰ, region Ⅱ, and region Ⅲ) have a column-like structure with three different directions, and the angle between the adjacent column area is 60°, as marked by the right blue arrows, similar to the observation shown in Fig. 2(c). Figure 3(g) shows the amplified image of the area, as labeled by the red rectangle in Fig. 3(f). The green dashed line shows the boundary between region Ⅱ and region Ⅲ. Region Ⅳ maintains the (2 × 2) structure with a lattice length of ~ 0.65 nm [Fig. 3(g)], and yet the formation of region Ⅱ stems from the disappearance of one row of atoms in every two rows, as illustrated by the white dashed circles, forming (2 × 4) reconstruction. The STM images of Fig. 3(e)→Fig. 3(h) reflect that the dynamic evolution from (2 × 2) to (2 × 4) reconstruction is realized by one of row atom missing gradually in every adjacent row atoms. After sample retreatment, the comparison of the surface atomic structure between Fig. 3(h) and Fig. 3(i) reflects that the surface involves two sets of the (2 × 8) reconstruction through a slight slit of the adjacent two columns of atoms and subsurface atomic movement, forming the final surface structure [Fig. 3(h) → Fig. 3(i)]. This result shows the formation of (2 × 8) reconstruction via atomic motions of Zn and O.
The other pathway involves the gradual transformation of the surface structure to the (2 × 4) phase under or around amorphous phase islands, the disappearance of the amorphous phase, and the gradual extension of the (2 × 4) structure until the surface phase becomes a single (2 × 4) structure [Fig. 3(j) → Fig. 3(k) → Fig. 3(m)]. These STM images hint that the (2 × 2) atomic reconstruction is able to form under the amorphous structures. On the other hand, the disappearance of these amorphous structures occur simultaneously accompanying with the formation of (2 × 4) structure.
The third pathway involves the transformation of atomic-scale hexagonal structures at the step edge at localized regions from the disordered phase [Fig. 3(a)] after annealing the sample at 780°C without the 500 ºC treatment, as shown in Fig. 4(a) and Fig. S6. Figure 4(b) shows the amplified area labeled by the green rectangle in Fig. 4(a). The lattice distance is calculated to be ~ 0.55 nm, as marked by the green double arrow in the amplified image [Fig. 4(b)]. This value is close to the value of \(\sqrt{3}\)aZnO (0.56 nm), indicative of the formation of a (\(\sqrt{3}\) × \(\sqrt{3}\)) reconstruction. The green arrows label the azimuth directions of the sample surface. By comparing the azimuth directions with the atomic arrangement, we obtain that the atom arrangement in Fig. 4(b) presents a clockwise rotation of 30° along the normal of the sample surface, as illustrated by the blue line. This result indicates a (\(\sqrt{3}\) × \(\sqrt{3}\))R30° reconstruction in Fig. 4(a). And then the sample surface becomes atomically flat throughout a large area with many defects, as labeled by white circles in Fig. S7. Figure 4(c) shows the high-resolution atomic-scale STM image. The triangular defect results from the absence of a Zn atom, as illustrated by the white dashed and light blue circles. The surface has a hexagonal structure with a lattice length of 0.68 nm (≈2aZnO = 0.65 nm), as labeled by the green double arrow in Fig. 3(c). This result indicates that the sample surface becomes a (2 × 2) reconstruction. These results indicate that the formation of (2 × 2) reconstruction in the third evolutionary pathway is first gradual formation of (\(\sqrt{3}\) × \(\sqrt{3}\))R30° reconstruction as a transition structure via numerous atom desorption of topmost surface, and then form (2 × 2) reconstruction, which is different from the two pathways above.
Following sample retreatment, the hexagonal structure becomes a columnar structure with the zigzag arrangement [Fig. 4(d)]. This surface structure change [Fig. 4(c) → Fig. 4(d)] stems from the atomic movement. At the next stage, the atomic movement increases further, and the two closest rows of atoms get closer together, changing the structure of the initial zigzag arrangement, as indicated in Fig. 4(e). After treating the sample further, we obtain the final surface structure [Fig. 4(f)]. Comparing the surfaces shown in Fig. 4(f) and 4(e), we notice the evaporation of one of the two closest rows of atoms, as well as the movement of subsurface atoms, as marked by the circles in red rectangles shown in Fig. 4(e) and (f), leading to two sets of (2 × 8) reconstructions.
To see the dynamic evolution of the surface structure from (2 × 2) to (2 × 8) reconstruction more clearly, we observed atomic arrangement in localized regions for comparison. These atomic arrangements have similar zigzag structures, as marked by the light blue solid balls in Fig. 4(c)–(f). The angle of the zigzag arrangement gradually increases [120º → 145º → 158º → 171º], while the lattice distances slightly decrease [(2.39, 0.68) → (2.32, 0.65) → (2.29, 0.63) → (2.21, 0.61)], as labeled by the green double arrows. This comparison shows that the third pathway of evolution from (2 × 2) to (2 × 8) involves atomic movement, followed by the combination of the subsurface atomic movement of O atoms and the evaporation of Zn atoms. This result indicates that the dynamic-evolutionary pathway from (2 × 2) to (2 × 8) reconstruction is realized via atomic movement of O and Zn as well as evaporation of Zn atoms.
The STM, EDS, and XPS results show that the ZnO(0001) surface with Ga surface segregation leads to plentifully atomic reconstructions. In addition, these reconstructions occur only at the surface, and the underneath phase structure maintains the lattice parameters of single-crystal ZnO, as demonstrated by the scanning transmission electron microscopy results [See the discussions for Fig. S8]. It is known that such large area and high-resolution STM images have not been reported previously on the ZnO(0001) surface. Next, we briefly discuss the compensated mechanisms of polarity and transformation from disorder to various atomic reconstructions. At the initial stage with lower Ga atom segregation [Fig. 3(a)], the disordered phase with numerous vacancies plays a key role in the compensation of polarity as in pure ZnO(0001) surface [7]. For this surface structure comprising both disorder and atomic reconstructions [Fig. 3(b), Fig. 3(j), Fig. 3(k), and Fig. 4(a)], the disordered phase, absent Zn atoms, and doped electrons by Ga atom segregation compensate the polarity. For the case of atomic reconstructions, absent Zn atoms, adsorbed species [Fig. 2(c)], and electron doping are responsible for the compensation of polarity.
The transformation from disorder to rich atomic reconstructions stems from the change in the charge-occupied states of dangling bonds. To demonstrate this point, we calculated the charge numbers of Zn and O atoms at different layers. The calculation method is provided in method section. Fig. S9 shows the calculated models. Fig. S10 shows the valence electron distributions of Zn and O atoms at different doping concentrations and surfaces. Here, green dashed lines denote the charge number of the bulk. One can see that the charge number of Zn atoms at the topmost surface is always larger than the underneath layers. Especially, for the doping below the topmost surface, the increase in the charge number is more obvious. Besides, this charge number increases with the increase in the Ga-doped concentration, while the charge number around O atoms presents little change. These results indicate that numerous electrons accumulate at the topmost surface driven by the internal drive for surface segregation. These accumulated electrons at the topmost surface on the one hand play a prominent role in compensating for the polarity of ZnO(0001). On the other hand, these charges fill the dangling bonds of the surface, as a result of a transition from fully empty states to partially occupied states. In fact, previous reports showed a very obscure and localized (1 × 3) on metallic ZnO(0001) surface [33], which indicates the dangling bonds of surface are at partially occupied states. Similar phenomena had been observed on (001)KTaO3 polar surface [3]. According to the principles of semiconductor surface reconstruction [20], the partially filled dangling bonds of surface undergo reconstruction spontaneously for the reduction of charge number and surface energy, like the covalent crystal system such as SiC(0001) and GaAs(100) [38–40]. That is why we observed large area atomic reconstructions on the polar surface of ionic crystals.
In conclusion, although the polar surfaces of ionic crystal systems with ordered reconstructions have extensive application value, they are difficult to be prepared, which limits their practical applications. Here, we demonstrated that the atom surface segregation can be an effective method to prepare rich reconstructions on ZnO (0001) surface as a prototypical ionic-crystal polar surfaces. And our studies show the fascinating atomic reconstructions involving (\(\sqrt{3} \times \sqrt{3}\))R30°, (2 × 2), (2 × 4), and (2 × 8), as well as other transitory surface structures, which differ substantially from pure ZnO(0001) surfaces. The key reason for our success is that the chemical valence states of dangling bonds can be changed from fully empty states to partially occupied states under the internal field by Ga atom surface segregation, and these occupied dangling bonds spontaneously induce a large area atomic reconstruction from disordered state, similarly to covalent surface system. More importantly, three distinct evolution-dynamic pathways from disorder to (2 × 8) reconstruction are observed by using STM. Our findings provide a general model for preparing a large area of atomic reconstructions and revealing dynamic-evolutionary pathways on complex polar surfaces of ionic crystals.