Characterizing the geometric and electronic structure of defects in the “29” copper surface oxide

The characterization of the geometric and electronic structure of solid surfaces broadly impacts the advancement of many fields including electronic devices, photovoltaics, photocatalysis, electrocatalysis, and heterogeneous catalysis. Specifically, the electronic structure of Cu₂O surfaces is of interest for a variety of applications. The 2.17 eV bandgap of Cu₂O is optimal for solar applications, such as solar cells, electrochromic devices, and photocatalytic water splitting.^1–3 Cu₂O has also shown promise for the electrochemical reduction of CO_2.^4,5 Furthermore, Cu-based heterogeneous catalysts are extensively used for many important industrial reactions including water-gas shift, methanol synthesis, and methanol oxidation.^6–8 Partially oxidized Cu is of growing catalytic interest because it is believed that Cu₂O regions of Cu are the most active sites in oxidation reactions.^9–12 Understanding intermediate oxide structures is fundamental not only to catalysis but also to other important technologies such as oxide growth and corrosion. This has motivated many surface science studies investigating the oxidation of Cu, and a variety of different oxide structures have been discovered on Cu(111).^13–23 In general, the oxidation of Cu(111) leads to the formation of Cu₂O(111)-like structures on the surface.¹⁴ At elevated substrate temperatures and sufficient oxygen exposures, highly ordered Cu₂O-like structures form a commensurate homogeneous single layer film across the surface.^13–15 In this study, the surface of interest is one of these highly ordered Cu₂O-like structures, referred to as the “29” oxide, which has been previously characterized at the atomic level.^23,24 This surface serves as a stable intermediate in the oxidation of Cu²³ and, therefore, an excellent model for studying defect properties in Cu redox mediated processes.¹¹ On all surfaces, and in particular oxide surfaces, defects play a vital role in the performance of the material in its applications.^25–30 Defects are typically the most reactive sites on an oxide surface and nucleation sites for metal particles.^29,30

By using scanning tunneling microscopy/spectroscopy (STM/S), the local density of states (DOS) of the substrate can be measured with nano-scale spatial resolution and compared to density functional theory (DFT) calculations.^31,32 Since the structure of the “29” oxide has been well characterized,^23,24 we can more carefully probe the surface DOS to deliver geometric-electronic structure correlations. Furthermore, the naturally occurring defects can be probed to the same degree. In this study, we investigate the electronic structure of the Cu₂O(111)-like “29” oxide and characterize two types of naturally occurring defects by STM/S and DFT.

The DFT calculations performed here utilized the Vienna Ab Initio Simulation Package (VASP) code^33,34 where the core electrons were treated with the projector augmented wave (PAW) method.^35,36 The valence electrons for all systems were described using the generalized gradient approximation (GGA) with the Perdew–Burke–Ernzerhof (PBE) functional.³⁷ The energy cutoff for the plane wave basis set was set to 500 eV, and electron smearing was described by the Gaussian smearing method with a width of 0.2 eV. All surface calculations were performed using a (1 × 2 × 1) Monkhorst-Pack k-points mesh.³⁸ The details of the construction and verification of the surface model of “29” oxide support were thoroughly investigated elsewhere, and the resulting “29” oxide support model is used here.^23,24 Each ground-state optimization calculation was considered converged when the total energy changed by less than 10⁻⁶ eV, and the forces between atoms were smaller than 0.02 eV/Å. The DFT-based simulated STM images were generated using the method discussed in our previous studies.^23,24 The DOS was calculated using the method of Dronskowski and co-workers,^39,40 which reconstructs the VASP calculated total wave function of the system from a set of Slater-type local orbitals, allowing us to obtain accurate local information from plane wave calculations. The use of this projection scheme resulted in an error of less than 2% for the states of interest (i.e., those below the Fermi level).

The “29” Cu_xO/Cu(111) surface structure was modeled as in our previous studies^23,24 by putting a Cu₂O-like layer on a 4-layer thick Cu(111) surface with a √13R46.1° × 7R21.8° supercell. The bottom 2 layers of the slab were kept fixed in their bulk positions, with a lattice constant of 3.635 Å that is consistent with the reported computational value of 3.634 Ź⁷ and the experimental value of 3.62 Å.⁴¹ The Cu_xO layer is made from fused hexagonal rings, each with 6 Cu atoms and 6 O atoms. There are 6 of these hexagonal rings per “29” oxide unit cell, which has 18 Cu oxide atoms and 12 O oxide atoms in total (Fig. 1). There are also O adatoms in the center of 5 of the 6 rings, which adsorb at hollow sites of the Cu(111) surface where they are bound most strongly.⁴² 

In the STM experiments, the sample was prepared in a preparation chamber (P = 2 × 10⁻¹⁰ mbar). The Cu(111) crystal was cleaned by Ar⁺ sputtering and annealing to 750 K. The “29” oxide film was formed by exposure to O₂ gas (Airgas, USP grade) at a pressure of 5 × 10⁻⁶ mbar for 3 min at a sample temperature 650 ± 20 K. The sample was then transferred in UHV to the STM chamber (P = 1 × 10⁻¹¹ mbar) and into the pre-cooled STM stage at 80 K using a low-temperature Omicron NanoTechnology STM. The STM was equipped with a preamplifier set with a low-pass filter of 80 kHz. Images were acquired using an etched W tip, biases were reported with respect to the sample, and color maps were applied to STM data using Gwyddion.⁴³ All dI/dV measurements used a lock-in amplifier. For constant-current dI/dV spectroscopy, the lock-in amplifier was set with a sinusoidal modulation frequency of 8 kHz and an amplitude of 0.06 V added to the sample bias with a 3 ms time constant. The spectra were recorded with the feedback loop closed at −0.5 V and 1 nA, and the voltage was swept from −0.5 to −6 V over 300 points with a T-raster of 10 ms. Reported dI/dV curves are an average of 10 (red—1), 8 (orange—2), 6 (green—3), 6 (light blue—4), 7 (dark blue—5), and 5 (magenta—6) individual spectra in Fig. 2, and an average of 4 (dim defect) and 6 (bright defect) individual spectra in Fig. 6. In the case of constant-height dI/dV spectroscopy in Fig. 3(b), the spectra were recorded with the feedback loop open and blanked at 1.1 V and 0.5 nA, and the voltage was swept from 2.25 to −3.5 V over 300 points with a T-raster of 30 ms. The lock-in amplifier was set with a modulation frequency of 6.75 kHz and an amplitude of 0.018 V with a 30 ms time constant. The presented data are averaged over 6 (dim defects), 12 (bright defects), and 10 (dark and bright rows) individual spectra. The lines in the inset in Fig. 3(b) are smoothed using a Savitzky-Golay filter (2nd-order polynomials fit to centered 0.3 V windows) overlaid with a scatter plot of the original data points. The lines in the inset are also offset for visibility. All dI/dV images in Fig. 4 were recorded at a 2 nA set point current with a 5 ms T-raster using a lock-in amplifier set with a modulation frequency of 800 Hz, with an amplitude of 0.009 V, and with a 10 ms time constant.

As shown in Fig. 1, the previously determined atomic model of the “29” oxide²³ is presented along with the DFT calculated DOS. Four different categories of surface species are shown in Fig. 1(a), namely, the Cu metal, Cu oxide, O oxide, and O adatoms. A Bader charge analysis⁴⁴ of the “29” oxide surface finds that these species have an average charge state of approximately 0.2, 0.7, −1.0, and −1.0 electrons per atom, respectively, revealing that even the Cu in the Cu(111) near surface layer is partially oxidized. As shown in Fig. 1(b), the total DOS of the surface is broken-down into the projected s-, p-, and d-states arising from each of the various surface species, and the lines are offset to make them each visible. As expected of a transition metal surface, the d-states dominate the total DOS, which are centered near −2 eV below the Fermi level. As a whole, the d-states of the surface are almost entirely filled, which generally leads to weak binding of adsorbates to the surface⁴⁵ and is consistent with previous studies finding weak binding with CO.²⁴ 

In order to experimentally examine the DOS of the “29” oxide down to lower energy levels (deeper filled states), constant-current tunneling spectra were recorded, as described by Berndt and co-workers.⁴⁶ The background shape of constant-current dI/dV spectra is such that the signal increases as the sample bias approaches the Fermi level.⁴⁶ As shown in Fig. 2(a), spectra were recorded at six different STM tip positions over the “29” oxide, and the corresponding dI/dV curves are shown in Fig. 2(c). It is clear that the d-states around −2 eV from the DFT calculated DOS in Fig. 1(b) are not seen in the experimental dI/dV spectra. It has been established that STS is relatively insensitive to d-states due to their localization near the nuclear cores, whereas s- and p-states project farther out into the vacuum.^47,48 Therefore, the s- and p-states are used to compare to the experimental dI/dV curves. In Fig. 2(b), the tip positions of the spectra are mapped onto the atomic model, and the DOS at those positions is shown in Fig. 2(d). To approximate the radius of curvature of the STM tip apex, the DOS was summed over the surface species within a circular area 1 nm in diameter.^49,50 These areas are highlighted in Figs. 2(e)–2(j), with the center of the circular area fixed at the determined tip position in the DFT model. It can be seen that the model sampling region of the oxide by STS is roughly half of the area of the entire “29” oxide unit cell. This explains why the same electronic states are found in the spectra at different tip positions, with only different relative intensities. The contributions to the DOS from the surface species in each area are presented in Figs. 2(e)–2(j), and the totals, plotted in black, are the curves presented in Fig. 2(d). The DOS at each position in Fig. 2(d) is normalized to the same y-scale value at −0.5 eV to match the initial conditions of STS data.

There is good qualitative agreement between the theory and experiment in this analysis despite the large number of potential sources of variation in the STS measurements and DOS calculations. The calculated DOS reveal that the state around −1 eV is primarily due to surface O species, both in the oxide rings and as adatoms. Additionally, the states around −4.5 and −5.5 eV are largely due to O atoms within the oxide and the O adatoms species, respectively. For example, the increase in the −5.5 eV state in dI/dV spectrum 6 (magenta) can also be seen by the relative increase in the projected DOS at low energy at site 6. There are variations in the dI/dV spectra in Fig. 2(c) around −2.5 V and −3.5 V which are not well reproduced by the s- and p-states of the calculated DOS, which indicates that there may be some small contribution from the Cu d-states to the dI/dV signal. Due to the agreement between the experimental and the DFT-obtained results, the analysis suggests that the O species project their electronic states much further away from the surface, and into the vacuum, than the Cu d-states.

The STM and STS combination is ideal for studying the geometric and electronic structure of the surface. There are two different kinds of defects that have been identified, which are referred to as bright and dim defects according to their appearance in STM images, as seen in Fig. 3(a), which will be discussed in detail later in the text. Also labeled in Fig. 3(a) are the dark and bright rows of the “29” oxide. Constant-height dI/dV spectroscopy through the Fermi level was recorded over each of the surface sites, as is shown in Fig. 3(b). The background shape of constant-height dI/dV spectra is such that the signal increases as the sample bias is increased.⁵¹ The onset of the conduction band edge can be seen at all surface features around 0.5 V. However, this state is slightly shifted toward the Fermi level at the dark rows with respect to the bright rows. On the other side of the Fermi level, the two forms of spectroscopy (constant-current and constant-height) can be linked. There is a small peak around −1 V, which is the same electronic state as was seen by the constant-current dI/dV spectra in Fig. 2(c); it is just more intense in the constant-current spectroscopy mode. The −1 V peak is highlighted in the inset of Fig. 3(b) in order to enhance its visibility. While the peaks are very small, the signal is strongest at the bright rows. In referencing back to the constant-current dI/dV spectra in Fig. 2(c), the −1 V peak is also stronger over sites 4 and 5, which constitute the bright regions of the oxide. Probing at lower energy in the dI/dV spectra in Fig. 3(b), there is another state around −2.5 V, which is present at the dark rows. As previously mentioned, the constant-current dI/dV spectra in Fig. 2(c) also show variation around −2.5 V, with moderate increase in signal at points 1, 2, and 3. These are precisely the darkest imaging points in the oxide that make up the dark rows in the STM images shown in Figs. 3(a) and 2(a).

A pseudo-bandgap of ∼1.5 eV can be seen by further analysis of the dI/dV spectra shown in Fig. 3(b). This exists, where the dI/dV signal is low and relatively flat, between the conduction band onset at 0.5 V and the s- and p-states of O atoms at −1 V. The “29” oxide has no true bandgap, as there are energetically continuous electronic states present near the Fermi level, but there is a lower density of states within the mentioned energy range. The ∼1.5 eV width of the measured pseudo-bandgap of the “29” oxide is reduced with respect to the 2.17 eV bandgap of bulk Cu₂O, which is also consistent with other studies on similar thin Cu oxide films.^52,53 From a fundamental point of view, it is interesting to track the growth and emergence of the Cu₂O bandgap, as it is a result of the hybridization of the valance states of Cu and O.³ In the case of the “29” oxide, the atoms making up the surface layer oxide are still coupled to the metallic Cu(111) substrate, preventing the formation of a true bandgap. Therefore, based on the present data, the emergence of the bandgap from Cu to Cu₂O would be expected to be a relatively smooth evolution as a function of the degree of oxidation.

The defects in the oxide can be identified in the STM image shown in Fig. 3(a) and are denoted as bright and dim defects according to their appearance in STM images; the color bar provided indicates the relative apparent height in the STM image. The bright defects image as protrusions situated on the dark rows of the oxide, seen as green features on dark blue rows on the lower terrace (right-hand side of the image) and red features on yellow rows on the upper terrace (left-hand side of the image). The dim defects image as depressions situated on the bright rows at the sample bias in Fig. 3(a) and can be seen as black features on the green rows on the lower terrace and yellow features on red rows on the upper terrace. From this STM image, it is clear that the dim and bright defects are always located at preferred sites on the oxide. The defects also show differences in dI/dV spectroscopy, shown in Fig. 3(b). The bright defect has a strong state around −2.5 V, which is present at the dim defect as well. In reference to Fig. 2, this state has previously been proposed to be due to a contribution from the Cu d-states. At positive sample bias, the defects display partial quenching of the conduction band edge around 1 V.

In order to spatially map the states near the Fermi level, a series of dI/dV images were recorded simultaneously with typical constant-current topographic STM imaging, shown in Fig. 4. There are two dim defects and one bright defect in the probed area that serve as landmarks for seeing changes in the oxide imaging. Working from left to right, first the sample bias was set to −1.25 V, which is sensitive to the s- and p-states of O atoms previously discussed. The appearance of the bright rows and dark rows in the topography image is matched in the dI/dV image. However, both defects flip from appearing as protrusions in the topographic image to depressions in the dI/dV image, which indicates a local reduction in the DOS at −1.25 V and, therefore, a lack of O character at the defects. This is highlighted by the inset in Fig. 3(b), in which the dI/dV spectra show which surface sites give rise to a peak, and hence the presence of an electronic state. The bright rows have the strongest peak above the background of the spectra at −1.25 V, followed by the dark rows, dim defects, and bright defects (in decreasing order). This same trend is matched in the dI/dV image at −1.25 V in Fig. 4. Next are the set of images at 0.6 V, which is at the base of the conduction band edge in the dI/dV spectra in Fig. 3(b). In moving from the −1.25 V to the 0.6 V dI/dV image, the relative brightness of the bright and dark rows has switched, which we have also shown previously in DFT simulated STM images.²³ This is made clear by looking at the defects as landmarks, where at −1.25 V, the bright defect is centered on a bright row, but at 0.6 V, it is centered on a dark row. The same observation holds at 0.7 V; however, it is noteworthy that the topographic images at 0.6 and 0.7 V are quite different for such a small change in the sample bias. The appearance of the defects as depressions in the dI/dV images at 0.6 and 0.7 V is further evidence of quenching of the conduction band edge, as seen by dI/dV spectroscopy shown in Fig. 3(b). It is also made visually clear that the electronic structure of the defects is very different from that of the intact “29” oxide.

Thus far, the defects have been imaged as protrusions in the topography image and as depressions in the dI/dV image. Continuing to move from left to right in Fig. 4, the images at a sample bias of 0.9 V break this trend; the dim defects appear as depressions in the topography image, while the bright defect still appears as a protrusion. In the dI/dV image, the dim defects appear as depressions; however, the dim defects are surrounded by locally bright features in the dI/dV image. This indicates a higher DOS at 0.9 V surrounding the dim defects, and as a result, the local electronic structure is disturbed over a ∼3 nm area by the defect. Another change in the imaging at 0.9 V sample bias is that the bright rows in the topography image appear as the dark rows in the dI/dV image. This is explained by analyzing the dI/dV spectra in Fig. 3(b) at 0.9 V, as the dI/dV signal is strongest at the dark rows. Finally, at 1.25 V, which is beyond the conductance band edge peak, the dim defects appear as depressions in the topography image but are protrusions in the dI/dV image. Conversely, the bright defect appears as a protrusion in the topography image but as a depression in the dI/dV image. The bright rows and dim defects are very close in the magnitude of their dI/dV signal at 1.25 V, and this is reflected in the dI/dV image, as the bright rows and dim defects both appear to have a very similar brightness. The collective set of images in Fig. 4 also infers structural information about the defect. For the bright defect, all sets of images find it as a protrusion in topography images but as a depression in dI/dV images, meaning that the bright defect is lacking electronic states for tunneling at each of these sample biases. Finding this consistently over a range of biases suggests that the bright defect is a physical protrusion on the surface.

The proposed atomic-scale structures of the defects were determined through a combination of DFT and STM/S. As a structural starting point, it is well established that common defects in oxide surfaces are oxygen vacancies.^28,29 As shown in Fig. 5(a), high-resolution STM imaging allows for the determination of the location of the defects within the “29” oxide unit cell. As previously discussed, the evidence of Cu d-states in the dI/dV spectroscopy of the defects in Fig. 3(b) indicates that there is more Cu character at these sites relative to the intact “29” oxide. In other words, the defect sites are more reduced than the surrounding surface. With this in mind, the removal of each O atom in the “29” oxide unit cell was structurally optimized by DFT, the change in energy was calculated with reference to the intact “29” oxide structure, and the results are presented in Fig. S1 and Table S1 of the supplementary material. Previous work has found that the reduction of ordered Cu₂O films on Cu(111) proceeds through a variety of intermediate structures,^17,21,22 and commonly O adatoms are the most easily removed. This was found to be true in a general sense on the “29” oxide as well, as the O adatoms were among the least energetically costly O atoms to remove from the surface (see Table S1 in the supplementary material). Furthermore, in previous work, we found that changing the O adatom coverage on the “29” oxide had relatively small effects on the surface energy of the structure.²³ However, the most favorable O to remove in the “29” oxide is an O oxide atom, and the optimized structure is shown in Fig. 5(b). Shown in Fig. 5(d) is a DFT simulated STM image at the same sample bias (−1.25 V) as the STM image shown in Fig. 5(a), in which the O removed structure is in the red unit cell surrounded by the intact “29” oxide. A new protrusion in the right half of the unit cell in the simulated STM image can be seen with respect to the intact “29” oxide shown, which is consistent with the brightness and location of the dim defect in experimental STM images. Furthermore, the simulated STM image of the dim defect at 0.9 V, shown in Fig. 5(f), is consistent with the experimental STM data showing a reduction in brightness with respect to the −1.25 V image. As such, we propose that the appearance of the dim defect is due to the local reduction of the oxide.

To determine a plausible structure of the bright defect, other O vacancy structures were considered. However, none of the projected DOS matched the dI/dV data showing a loss in O character, as argued in Fig. 3. Therefore, we conclude that the bright defect must be due to a different structural modification. Thus, the next logical structural modification to test was the addition of Cu adatoms to the “29” oxide surface, which would be expected to exhibit more Cu character in dI/dV spectroscopy and be a topographic protrusion. A Cu atom atop an O adatom would create the stoichiometric Cu₂O surface, which has been proposed in other Cu₂O thin films studies.^5,53–55 The most stable structure for a Cu adatom which maintained the structural integrity of the “29” oxide is shown in Fig. 5(d). The adsorption energy at other tested sites on the “29” oxide is shown in Table S1 of the supplementary material. A simulated STM image of this structure is shown in Fig. 5(e) in the highlighted red unit cell, which is also surrounded by the intact “29” oxide. This structure reproduces a bright feature in the bottom-left region of the unit cell relative to the “29” oxide, in excellent agreement with the location of the bright defect in Fig. 5(a). The simulated STM image of the bright defect at 0.9 V, shown in Fig. 5(g), displays a reduction in the brightness of the defect relative to the −1.25 V image, which is also seen in the experimental STM images shown in Fig. 4. This leads us to propose that the bright defect is due to the presence of a Cu adatom.

The electronic structure of the defects below the Fermi level was probed with constant-current dI/dV, shown in Fig. 6(a), just as the intact “29” oxide in Fig. 2. Similar to the analysis presented in Fig. 2, the corresponding DFT calculated projected DOS of the defects is shown in Fig. 6(b). The breakdown of the various states contributing to the DFT calculated DOS is shown in Figs. 6(c) and 6(d), applying the same assumption of a circular 1 nm diameter sampling area. The state seen around −2 V in the dI/dV spectra is consistent with a d-state as previously identified in Fig. 3(b). The relative position of the −2 V state at the dim defect is at a lower energy with respect to the bright defect, which is reproduced by DFT, offering good qualitative agreement between experiment and theory. The measurement of d-states in the STS at defects is explained by the presence of lifted Cu oxide atoms, presumably putting Cu atoms in closer proximity to the STM tip and allowing for some overlap between the Cu d-states and tip wave functions. In the intact “29” oxide, the linear O–Cu–O bonds make it so that the Cu oxide species always has a neighboring O oxide species above it,²³ leading to the strong presence of O states in STS measurements. These lifted Cu atoms created by the defects are highlighted by yellow circles in Figs. 6(c) and 6(d), and their d-states are included along with the s- and p-states of all of the other surface species to yield the total defect DOS. In the case of the dim defect, shown in Fig. 6(c), the highlighted lifted Cu atom is the highest Cu atom in the structure by approximately 0.2 Å and has been raised by 0.44 Å with respect to its position in the intact “29” oxide. In the case of the bright defect, shown in Fig. 6(d), the 3 highlighted atoms are all within 0.08 Å of each other by height and have been raised by approximately 0.4 Å with respect in the intact “29” oxide. Additional states at deeper filled state energies are also seen. A peak at −3 V in the dI/dV spectra at the bright defect is most closely matched by a state in the DOS at −2.75 eV, which is due to a shoulder in the d-band. It is known that dI/dV spectroscopy can induce a shift in the measured peak with respect to the true energy of the state,³² which may explain the differences seen here. However, the STM tip DOS and potential sources of error in the DFT calculations may likely also play a role. The last prominent state in the dI/dV spectra is centered around −4.5 V and is present at both defects. These states are in the same energy range as the O s- and p-states seen in the intact “29” oxide. The −4.5 V state is broader at the dim defect with respect to the bright defect, which is qualitatively captured by the DFT calculated DOS. The agreement between experiment and theory in the electronic structure of the surface provides a high level of understanding of the system.

Structural defects in oxide surfaces are often associated with electronic charge redistribution.²⁸ In addition to the thorough characterization of the DOS, the Bader charge of the defects was also calculated.⁴⁴ As shown in Fig. 7, any atoms with a charge difference greater than 0.1 electrons with respect to the intact “29” oxide surface are highlighted with Cu oxide atoms in yellow and Cu metal atoms in blue. The creation of the dim defect results in a large number of atoms with a significant change in the charge state, shown in Fig. 7(a) and quantified in Table I. The dim defect results in a net reduction of the surface; however, there is also a significant charge transfer between surface species across the “29” oxide unit cell due to the large structural changes to the oxide layer, leading to an overall delocalization of charge away from the defect site. This is consistent with the dI/dV imaging results in Fig. 4 finding a large area around the dim defect with modified DOS. As opposed to the dim defect, the introduction of a Cu adatom and creation of the bright defect results in a more localized redistribution of charge in the Cu metal and Cu oxide species, as shown in Fig. 7(b). In this case, only two atoms have their charge state significantly modified, and atom “i” is the new Cu adatom. The localization of charge around the bright defect suggests that this site could be more active for adsorption processes than the dim defect.²⁹ 

To further investigate the reactivity of the defects with respect to the intact “29” oxide, the position of the Cu oxide d-states can be examined. The DFT calculated projected DOS of the lifted Cu atoms, shown in Figs. 6(c) and 6(d), finds that the Cu d-states in the dim and bright defects are centered at approximately −1.50 and −1.42 eV, respectively. The d-states of the defects are closer to the Fermi level than the d-states of the intact “29” oxide, which is centered around −1.65 eV for the Cu oxide atoms, as seen in Fig. 1(b). Therefore, the d-band model and scaling relations would suggest that the defect sites would have stronger binding and more reactive properties than the intact “29” oxide, as the center of the d-band of the Cu atoms affected by the defects are closer to the Fermi level.⁴⁵ Furthermore, an interesting relationship between the DOS and charge state is born out from this analysis, as coincidentally the bright defect consisting of an extra Cu adatom exhibits a slightly greater shift in the d-states toward the Fermi level and greater localization of charge redistribution relative to the dim defect.

In conclusion, we have experimentally characterized the electronic and geometric structure of the “29” oxide Cu₂O/Cu(111) film and two types of commonly occurring defects with atomic-scale details with both STM/S and DFT. Understanding the nature of defects is important, as the increased chemical reactivity of defect sites on both metals and oxides has been well established. In this study, we have determined how structural defects affect local electronic properties that make the defect sites more likely to be reactive, and these same correlations may be applicable to similar defects on other thin film oxide surfaces. We propose that one of the defects is an O vacancy, while the other is an additional Cu atom, both of which stoichiometrically create a reduced area on the surface. In the case of the two defects characterized here, there are clear signatures in the dI/dV spectroscopy indicating new DOS profiles at these sites. Interestingly, the defects present their d-states in STS curves, whereas d-states cannot be appreciably detected on the rest of the “29” oxide surface. Furthermore, charge redistribution occurs at the defect sites and the d-states shift toward the Fermi level, particularly for the Cu adatom defect which creates a unique surface site with respect to the rest of the surface for strong binding to, and charge transfer with, adsorbates.

See supplementary material for the description of the “29” oxide model and tested O vacancy formation energies and Cu adatom adsorption energies.

The work at Tufts was supported by the Department of Energy BES under Grant No. DE-FG02-05ER15730. M.D.M. thanks Tufts Chemistry for an Illumina Fellowship. Financial support to Washington State University was provided by the National Science Foundation (NSF) EAGER program under Contract No. CBET-1552320 and the NSF CAREER program under Contract No. CBET-1653561. Our thanks also go to the donors of the American Chemical Society Petroleum Research Fund. A portion of the computer time for the computational work was performed using EMSL, a national scientific user facility sponsored by the Department of Energy’s Office of Biological and Environmental Research and located at PNNL. PNNL is a multi-program national laboratory operated for the US DOE by Battelle.