Comparison of aqueous and native oxide formation on Cu 111

Transcription

1 JOURNAL OF CHEMICAL PHYSICS VOLUME 110, NUMBER MARCH 1999 Comparison of aqueous and native oxide formation on Cu 111 Y. S. Chu and I. K. Robinson Department of Physics, University of Illinois, Urbana, Illinois A. A. Gewirth Department of Chemistry, University of Illinois, Urbana, Illinois Received 10 July 1998; accepted 4 December 1998 We present the results of an x-ray diffraction investigation of the formation of oxide on electropolished Cu 111 surfaces, both in situ at ph 4.5 and in air. In both cases the oxide is found to be crystalline cuprite and epitaxially aligned with the substrate, but with two possible epitaxial orientations. We followed the progress of oxidation by monitoring the shapes of the diffraction peaks for the two orientations as a function of time and potential. There is a narrow potential region where the oxide is a single monolayer thick. Beyond that, only one of the two oxide orientations becomes thicker, and does so in an inhomogeneous manner, thickening in narrow regions before it spreads American Institute of Physics. S I. INTRODUCTION Fundamental understanding of the metal oxidation processes is an important problem in many disciplines of physical science. In particular, oxidation of copper has attracted attention because of a strong interest to understand the corrosion of this widely used metal. Dry oxidation of copper has been studied extensively using a number of techniques such as transmission electron microscopy TEM, 1,2 low-energy electron diffraction LEED, 3,4 reflection high-energy electron diffraction RHEED, 3 electron energy loss spectroscopy EELS, 5 scanning tunneling microscopy STM, 4,6 and extended x-ray absorption fine structure EXAFS. 7 In these studies, the copper surfaces were cleaned under vacuum and oxidized by dosing oxygen. The growth of cubic cuprous oxide cuprite, Cu 2 O) on low index copper surfaces was found to be epitaxial with the substrate, and a wide array of growth morphologies was reported, depending on the growth conditions. By comparison, there exists only a small number of reports on the oxidation of copper in the presence of water concerning the structure and growth properties of the aqueous oxides on copper surfaces. You et al. investigated the interfacial roughening of copper film due to oxidation in borate buffer solution at ph Gewirth et al. reported a c 2 2 oxygen adlayer on Cu and 2 1 and 3 1 oxygen chain structures on Cu in dilute acid solutions ph 2to3, using in situ atomic force microscopy AFM, which were considered as precursors to bulk oxide growth. Ikemiya et al. investigated the structures of aqueous oxide on Cu 111 and Cu 100 in alkaline solutions at ph 13 and reported epitaxial Cu 2 O structures on both surfaces. 11 The AFM measurements have provided valuable information on the structure and morphologies of the aqueous oxide. Yet, since this is a prototypical thin-film system, the questions of epitaxial strain and growth properties are important yet still largely unknown. We have previously studied the structure of Cu 111 under electrolyte at ph 1 and found this to consist of a specifically adsorbed oxygen layer. 12 The thermodynamics of the copper/water interface is explained by the Pourbaix diagram, 13 which specifies the allowed bulk phases at each ph and potential value. When ph 3.5, Cu metal, Cu in solution and bulk oxide are all stable within certain potential ranges. At ph 3.5 the Cu metal phase is adjacent to Cu in solution, so the formation of bulk oxide is strictly forbidden. Since oxidation is impossible in the bulk, any oxygen binding here must occur as monolayers or very thin films, analogous to the under-potential deposition UPD process seen in metals. The Pourbaix diagram for the Cu/water system is discussed further below, in the context of our results. In this study, we investigate using x-ray diffraction, the structure of the aqueous oxide on Cu 111 in a very dilute acid solution at ph 4.5. The focus is to understand how the structure of the epitaxial Cu 2 O changes with the thickness of the oxide film. In addition, we examine the structure of the native oxide, found upon exposure to air on the freshly electropolished Cu 111 surface, to understand the differences and similarities between these two oxides formed under different conditions. II. EXPERIMENTAL METHODS Electrochemical x-ray diffraction experiments were carried out in an in situ thin layer cell described previously. 14 The sealed teflon cell contained the copper sample, a double junction Ag/AgCl reference electrode and a gold counterelectrode immersed in a 15 cm 3 volume of electrolyte connected to an external feed supply. X-rays reached the sample through a 6 m polypropylene window which could be inflated with electrolyte to create a bulk liquid environment for detailed electrochemical characterization. The membrane could then be sucked down onto the face of the sample, by lowering the level of the electrolyte feed, to leave only a thin capillary layer which is almost transparent to the x-ray beam. An argon environment provided on the outside of the membrane prevented inopportune oxygenation of the electrolyte. The measurements reported here were made by sweep /99/110(12)/5952/8/$ American Institute of Physics

2 J. Chem. Phys., Vol. 110, No. 12, 22 March 1999 Chu, Robinson, and Gewirth 5953 FIG. 1. Structure of cuprite. Each copper ion solid shading has two oxygen ions light shading as its nearest neighbors while each oxygen ion is surrounded by a tetrahedron of copper atoms. The top view of the 111 surface is constructed by a cross section containing three oxygen atoms. The crystallographic 111 unit cell has twice as many atoms as the primitive cubic unit cell, with the hexagonal base of length, a 2a o, and a height, c 3a o. In the surface view, all the copper atoms at the ( 4, 1 1 4, 1 4) positions see the 3D view are omitted for clarity. ing the applied potential in the sucked down configuration; the sweep rate was slow enough that electrochemical equilibrium was apparently maintained. Electrolyte stock solutions were made from ultrapure reagents and deoxygenated by bubbling argon prior to their use. The supporting electrolyte was 0.1 M NaClO 4, adjusted to ph 4.5 with HClO 4. Cu 111 samples were oriented within 0.5 and mechanically polished with 0.05 m alumina paste. They were electropolished in 43% phosphoric acid supersaturated with Cu at 1.0 to 1.5 V and 30 to 60 ma/cm Considerable caution was used to avoid excessive oxidation after electropolishing, because the electrolyte used was much less acidic. When visible native oxide formed after electropolishing, a drop of 0.1 M HClO 4 was used to remove it. Then the sample was rinsed with Millipore water. Otherwise, the same experimental procedures and the sample preparations methods described previously were employed. 14 The x-ray diffraction measurements were carried out at 8.5 kev ( 1.46 Å, sufficiently below the Cu K-edge to avoid fluorescence. Beamline X16C at the National Synchrotron Light Source NSLS was used with a saggitallyfocused monochromator and high-speed Kappa diffractometer. 15 The instrumental resolution was set by 1 mm entrance slits and 2 2 mm slits in front of a scintillation detector. The crystal was aligned by centering its bulk Bragg peaks and the surface diffraction pattern was indexed relative to these. As is conventional in surface x-ray diffraction experiments, a hexagonal coordinate system was used whose third, L-index represents the component of momentum transfer perpendicular to the surface. In this account, the suffix hex denotes the use of these coordinates. III. STRUCTURE OF BULK CUPRITE FIG. 2. Side view of the atomic model of cuprite attached to the Cu 111 substrate. The oxygen and copper ions of the oxide are represented by gray and black spheres, respectively, while the metallic copper atoms are shown as light spheres with close-packing radius. Both aligned and reversed oxides have their 111 axis parallel with the substrate 111 axis but with different in-plane epitaxial relations. In the aligned cuprite, 110 Cu2 O 110 Cu(111), while in the reversed cuprite, 110 Cu2 O 110 Cu(111). In order to understand its epitaxial relationship with the substrate, the structure of cubic cuprite, Cu 2 O, is shown in Fig. 1. The primitive unit cell is a cube with a lattice constant, a 4.27 Å, and contains two oxygen and four copper atoms. 16 The oxygen atoms form a body centered cubic structure, while the copper atoms partially occupy the interstitial positions in an alternating pattern. Each copper atom has only two nearest oxygen neighbors, while each oxygen atom is surrounded by a tetrahedron of copper atoms. The copper atoms are located midway between two oxygen atoms with a Cu O bond length of 1.85 Å. The Cu Cu separation is 3.01 Å. The Cu 2 O 111 surface is constructed in Fig. 1. In hexagonal coordinates, its crystallographic unit cell is defined by a hexagonal base of length, a 2a o, and a height, c 3a o and contains twice as many number of atoms as the cubic unit cell. The diffraction pattern of this structure has cubic characteristics, with three-fold symmetry about the 111 perpendicular axis. IV. RESULTS A. Native oxides on Cu 111 The freshly electropolished dry surfaces of four different Cu 111 samples were examined with x-ray diffraction as a function of time after polishing during exposure to ambient air. After removal from the electropolishing bath, samples were rinsed in deionized water and dried by blowing nitrogen gas over the surface. On all four samples, epitaxial oxides with two distinct azimuthal orientations were identified by their diffraction peak positions. The diffracted intensities from the two oxide orientations were about equal, which resulted in a six-fold symmetric diffraction pattern. The structure of the oxide was identified by the symmetry and lattice dimensions derived from its diffraction pattern to be the cubic cuprite structure (Cu 2 O described above. The 111 perpendicular axis was parallel with the 111 substrate axis. However, the two orientations had different in-plane epitaxial relations with the substrate, as illustrated in Fig. 2. In the aligned cuprite (Cu 2 O structure, the 11 0 oxide axis was parallel with the 11 0 substrate axis, while in the reversed cuprite the 11 0 oxide axis was anti-parallel with the 11 0 substrate axis. Another way to appreciate the difference between the two structures is that they are related with a 180 rotation about the axis perpendicular to the plane in Fig. 2. The details of the native oxide structures were investigated as a function of the film thickness during the subsequent growth in the ambient laboratory air. Conditions defining native, such as humidity, were not carefully

3 5954 J. Chem. Phys., Vol. 110, No. 12, 22 March 1999 Chu, Robinson, and Gewirth FIG. 3. X-ray diffractometer scans along the 0, 0.85, L direction, taken at different elapsed times after electropolishing the sample; 9 hours triangles, 12 hours circles and 23 hours squares. Peaks A and B are the oxide Bragg peaks indexed in Fig. 4. Peak C is due to the diffuse scattering from 0, 1, 2 substrate Bragg peak. The lines are multiple Gaussian fits, used to determine the exact peak locations. controlled, but were not abnormal. There was significant variation among the samples of the initial oxide film thickness, measured within a few hours after electropolishing, but all four samples displayed a similar growth trend subsequently. Figure 3 shows the profiles of the cuprite Bragg peaks at different elapsed times since electropolishing. Peak A and peak B, located in substrate coordinates at 0, 0.85, 0.85) hex and 0, 0.85, 1.7) hex, correspond to the positions of the reversed 1,1,1 and aligned 0,0,2 cuprite Bragg peaks. The diffraction patterns of the aligned and reversed cuprite, along with the substrate, are displayed in the reciprocal space map of Fig. 4. The diffraction patterns of the two structures are symmetric with respect the 00L axis. About 9 hours after electropolishing, the cuprite Bragg peaks were very weak and wide indicating a thin layer. However, about 23 hours after polishing, the intensity of the 0, 0.85, 0.85) hex peak increased by about four times, and its width decreased by about 50%. The peak positions changed slightly with time, suggesting changes in the epitaxial strain of the oxide film. Dependence of the lattice parameters of the native oxide on oxide film thickness is displayed in Fig. 5. For comparison purposes, the quantities a/ 2 parallel to the interface and c/ 3 perpendicular are shown, which both adopt the value a Å for an unstrained film. Although the natural misfit between cubic Cu 2 O (a Å and copper FIG. 4. Reciprocal space diagram in the hexagonal coordinate system of the Cu 111 substrate defined in Fig. 1. The peaks A and B, located approximately at 0, 0.85, 0.85 and 0, 0.85, 1.7, are the observed positions of the reversed and aligned oxide Bragg peaks seen in Fig. 3. FIG. 5. In-plane and surface-normal lattice parameters of the cuprite oxide on Cu 111 as a function of their peak intensities. For comparison purposes, the measured values are converted into cubic lattice parameters by plotting a/ 2 circles and c/ 3 squares on the same vertical axes. As the oxide grew thicker, the in-plane lattice parameter increased from a value close to the 3 7 commensurate spacing 4.22 Å toward the bulk cuprite spacing 4.27 Å. The opposite trend was observed for the surface-normal lattice parameter. (a Å is 15.3%, the in-plane spacing of Cu 111 surface a 2.56 Å and Cu 2 O 111 surface a 6.04 Å are such that the coincidence misfit for three Cu 2 O 111 unit cells over seven Cu 111 unit cells is only 1.22% when the two lattices are aligned. The previous studies on the dry oxide 1 and the AFM study on the aqueous oxide at ph have proposed this 3 7 commensurate structure. Our results show that when the film is thin, the in-plane oxide spacing is close to this 3 7 commensurate spacing, but as the oxide grows thicker, the spacing approaches the bulk cuprite spacing instead. Meanwhile the perpendicular oxide spacing decreases toward the bulk spacing from much larger initial values, showing the opposite trend. These results can be understood in general terms as a reduction of the compressive epitaxial strain at the interface while conserving the unit cell volume. B. Voltammetry of aqueous oxidation Cyclic voltammetry CV ofacu 111 single crystal in 0.1 M NaClO 4 ph 4.5 is shown in Fig. 6. These measurements were made on a sample installed in the electrochemical cell but with the membrane inflated to expose it to bulk electrolyte. Major features of the voltammetry are marked with labels, A, B, and C. The anodic peak, A, occurs during the oxidation part of the cycle, while the cathodic peak, B, corresponds to reduction. The tail of the voltammogram, labeled as C, is due to reduction of 2H to H 2. A well defined oxidation peak exists because the surface becomes passivated after initial oxidation in a well deoxygenated electrolyte. The charge transferred during the oxidation or reduction, estimated by integrating the voltammogram in Fig. 6, is 150 C/cm 2. However, the exact amount of charge depends upon certain kinetic variables: scan rate, polarization history, limits of the potential cycle, as well as the thickness of the electrolyte film.

4 J. Chem. Phys., Vol. 110, No. 12, 22 March 1999 Chu, Robinson, and Gewirth 5955 FIG. 6. Cyclic voltammetry of Cu 111 single crystal in 0.1 M NaClO 4 at ph 4.5 measured at 10 mv/s. Potentials were measured with respect to Ag/AgCl 3 M KCl and have been converted to the normal hydrogen electrode NHE scale. Peak A and peak B are due to oxidation and reduction of the Cu 111 surface, respectively. Peak C is due to reduction of protons (H ) to hydrogen gas (H 2 ). C. Formation of aqueous oxides We investigated the interfacial structures corresponding to these electrochemical signatures by in-situ x-ray diffraction. Our study revealed immediately that the electrochemical oxidation of the Cu 111 surface in 0.1 M NaClO 4 at ph 4.5 resulted in the formation of the same two orientations of epitaxial oxides, aligned and reversed, identified by their Bragg peak positions in Fig. 4, as the native oxidation. These were the only oxide peaks found; scans were made along other high-symmetry in-plane directions to look for additional peaks without success. In particular, positions corresponding to superstuctures of the substrate lattice, such as the ( 3 3) position known to pertain to chloride contamination, were checked. We then studied the growth kinetics by tracking simultaneously the intensity of both the reversed and aligned diffraction peaks as the potential was slowly scanned with the cell membrane sucked down. Simultaneous observation of two diffraction peaks, separated far apart in reciprocal space, was practical because of the high speed of the custom-designed Kappa diffractometer. 15 As shown in Fig. 7, the growth of aqueous cuprite on Cu 111 has a strong dependence on the crystallographic orientation of the interface. After the potential was increased FIG. 7. Diffraction intensity of the reversed open symbols and aligned solid symbols oxides as a function of potential. Potentials were measured with respect to Ag/AgCl and have been converted to the normal hydrogen electrode NHE scale. The arrows indicate the sweep direction. from the reduced state ( 0.4 V relative to NHE to the potential denoted A ( 0.23 V, both reversed and aligned structures began to form on the surface. Between points A and B at about 0.12 V, the intensities of both oxide orientations remained the same and roughly constant. As the potential was increased further, the intensity of the reversed cuprite began increasing rapidly, while the intensity of the aligned cuprite remained constant. The scan direction was reversed at 0.15 V and the intensity of the reversed oxide still continued to increase. It did not begin to decrease until the potential once again reached the reducing potential region ( 0.2 V on the anodic sweep. At its maximum level, the intensity ratio, which roughly represents the proportions of aligned and reversed oxides, was about 50:1. D. Structural details of aqueous oxides Monitoring the diffraction peak intensities as a function of the potential has thus provided useful kinetic information on the quantity of the oxide present at different stages of growth. However, it did not provide much information about the structure. The electrochemically oxidized Cu 111 surface was examined by x-ray diffraction at various potentials after trapping the intermediates formed in the oxidation process. In the first step of our trapping procedure, the potential was cycled with the membrane inflated. Once the potential reached its extreme negative value sample completely reduced, the membrane was deflated and the potential was increased slowly at 1 mv/s into the oxidation potential region. After ramping to the desired potential, the scan direction was reversed to compensate for hysteresis, in order to prevent further oxidation. This reverse potential ramping was stopped when the current became slightly negative again. During the x-ray measurements, the current density remained small, typically a few A/cm 2. In this way, the extent of oxidation on the Cu 111 surface could be controlled accurately and reliably. This trapping method was possible because of the small hysteretic irreversibility associated with the oxidation/reduction process of the Cu 111 surface: it requires some overpotential to reduce the surface after it is oxidized, and vice versa. Without precautions, oxidation would have continued during the measurements, and have led to time dependence in the data. In practice, complete trapping was not possible over a long time because of the slight background current and the motion of the diffractometer. Figure 8 compares the shapes of the diffraction peaks from the aligned and reversed oxides for a thin film, by growing a thin oxide film and applying the trapping procedure to stabilize the film during the measurements. They are strikingly different: the reversed oxide exhibits a well defined peak in both the q -direction parallel to the surface and q -direction normal to the surface. This indicates that the reversed cuprite already has a three-dimensional crystalline structure. Conversely, the diffraction peak from the aligned oxide is sharp only in the surface-parallel direction and almost without a maximum in the surface-normal direction. The true line shape in the surface-normal direction is difficult to extract due to the broad diffused scattering from the nearby 1,0,1 substrate Bragg peak. This very wide

5 5956 J. Chem. Phys., Vol. 110, No. 12, 22 March 1999 Chu, Robinson, and Gewirth FIG. 8. Profiles of the aqueous oxide Bragg peaks, a along the in-plane top panel and b along the surface-normal bottom panel direction. The data were measured while holding the potential at 0.23 V, after cycling up to around 0.05 V for a short time so that a thin oxide appears. The diffraction peak from the reversed oxide is well defined in both directions, while the diffraction peak from the aligned oxide is sharp only in the surface-parallel direction. The broad diffused scattering from the nearby substrate Bragg peak make it difficult to extract the true surface-normal profiles of the aligned diffraction peak. surface-normal profile implies that the aligned cuprite is very thin, probably not much more than a single atomic layer. To verify that this was not a homogeneity effect, the measurements were reproduced at other locations on the sample, showing that the preferential growth of the reversed cuprite occurred throughout the entire surface. The diffraction patterns from both epitaxial oxides disappeared completely when the surface was reduced and reappeared again in the same proportions upon re-oxidation. Next we investigated how the lattice spacing of the reversed cuprite varied with the potential in Fig. 9, again applying the trapping procedure at each potential in the cycle. Both the in-plane lattice spacing, a/ 2, and the surface-normal lattice parameter, c/ 3, depended very sensitively on the potential. At 0.2 V the first potential in the cycle, the diffraction peak was very wide in the L-direction, and its position could not be determined accurately. At this potential, the reversed cuprite film was probably no thicker than a few monolayers, and its in-plane lattice spacing was under significant compressive strain about 1.9%. Asthe potential increased from 0.2Vto 0.14 V, both the inplane and the surface-normal lattice spacings increased. At 0.14 V, the in-plane lattice parameter of the reversed cuprite was still under about 1.3% strain. After the potential scan direction was reversed at 0.14 V, the relaxation of the in-plane lattice parameters still continued but at much lower rate, because the growth rate of the oxide had become reduced significantly by this point see Fig. 7. In the potential range beyond 0.05 V, no further changes occurred until about 0.30 V when a slight increase in the in-plane lattice parameter and considerable increase in the surface-normal FIG. 9. Lattice spacings of the reversed cuprite versus potential. For comparison purposes, the measured values are converted into cubic lattice parameters by plotting a/ 2 filled symbols and c/ 3 open symbols on the same axis. The start and the end point of the potential cycle are indicated in the plot. The diffraction peak at 0.2 V the first data point of the series was very broad in the surface-normal L direction, so its position could not be determined accurately. For reference, the bulk cuprite and 3 7 commensurate spacings are indicated with dotted lines. The insert shows the percentage deviation of the unit cell volume from that of bulk cuprite versus the potential. lattice parameter were observed at the onset of reduction of the oxide. At this potential, the oxide had probably become unstable as it was being dissolved while the measurements were made. During the entire cycle, the in-plane lattice parameter never became fully relaxed; even while the oxide was being reduced, it was under about 1% strain. The relaxation behavior of the aqueous cuprite film as it thickens was very unusual, especially when compared with the relaxation behavior of the native cuprite film see Fig. 5. Both the in-plane and surface-normal spacing relaxed in the same direction, indicating a substantial change in the unit cell volume. The volume, shown inset in Fig. 9, has undergone a monotonic increase from a value about 4% smaller than the bulk volume. At 0.1 V, it was almost completely relaxed about 99.7% of the bulk volume, while both inplane and surface-normal lattice parameters remained significantly strained from the bulk values. After the oxide became unstable below 0.3 V; arrow in Fig. 9 insert, the unit cell volume became slightly larger than the bulk volume. This unusual nonelastic behavior is most likely due to changing stoichiometry of the oxide, perhaps due to a large vacancy concentration upon its initial formation, as seen during passive oxidation of iron. 17 E. Regimes of growth Since the oxide tends to thicken gradually during the measurement, there is expected to be some contribution of time dependence among the data plotted versus potential in Fig. 9. For example, it was found that there were small variations in the peak positions depending on the rate at which the measurements were made. One way this effect can be eliminated is to consider the inverse L-width FWHM of the dif-

6 J. Chem. Phys., Vol. 110, No. 12, 22 March 1999 Chu, Robinson, and Gewirth 5957 FIG. 11. Pictorial representation of the growth of the aqueous oxide in the four growth regimes. I. Oxygen adsorption on the surface to form a monolayer oxide. II. Initial growth of nonuniform oxide film showing the formation of narrow but thick nuclei of highly strained and compressed oxide film. III. Lateral oxidation of the regions in between the nuclei with no significant increase of the film thickness. IV. Uniform growth of oxide and subsequent passivation. FIG. 10. The in-plane spacing, a/ 2, of the reversed oxide top panel and the integrated intensity of the diffraction peak bottom panel are plotted against the apparent film thickness. Four distinct growth regimes are identified by the labels I, II, III, and IV. The dashed line in the upper figure is the fit to the data using a simple elastic equilibrium theory see text, while the dashed line in the lower figure is the linear behavior expected for uniform growth. fraction as the independent variable rather than the potential. In the simplest Scherrer diffraction model with no resolution correction, the thickness of the film, h, is related to the width FWHM of the diffraction peak in the surface-normal direction, L as h c L, where c is the unstrained surface-normal lattice parameter of cuprite and h is the apparent thickness of the film. Use of this model ignores the possibility that inhomogeneities of the lattice parameter can also contribute to the peak width and thus cause the value of h to be slightly underestimated. Ignoring instrumental resolution causes a similar result. In both cases, the distortion is negligible for small values of h, and can be estimated to be significant only when h reaches 100 Å, based on the variation in surface-normal lattice parameter seen in Fig. 9. The combined data collected for several repeats of this experimental procedure including the data shown in Fig. 9 are plotted in Fig. 10 as a function of the apparent thickness, h. The results of Fig. 10 can be divided into four different regimes, according to the way the integrated intensity and the in-plane lattice spacing depend on the apparent thickness. In regime I, no diffraction pattern from the oxide was observed. In regime II, the in-plane lattice spacing remained constant, while the integrated intensity was very small and did not change much with the thickness. The diffraction pattern in this regime was characterized by very wide and rapidly changing diffraction profiles in the L-direction, and the oxide film was under high epitaxial strain in-plane ( 2.0 to 1.5%. In regime II, the extracted values of the apparent film thickness are subject to large systematic errors because the L-width of the diffraction peak depended strongly on the peak fitting methods and the background subtraction methods. The data shown in this regime were obtained in the potential range about 0.23 to 0.2 V. Throughout the growth regimes II and III, the inplane lattice parameter increases systematically. This behavior can be compared with Van der Mervwe s equilibrium theory of epitaxy, which predicts that a thin film will be commensurate with its substrate, but will become energetically unfavorable beyond a certain critical thickness. 18 According to this model, the epitaxial in-plane lattice parameter, a, is found to depend on the thickness of the film, h, in the following way: a a 0 h c h a s a 0, for h h c, where a 0 is the bulk lattice parameter, a s is the substrate lattice parameter, and h c is the critical thickness. The theoretical curve in the upper panel of Fig. 10 is a fit of this functional dependence to the data, which yielded, a s 4.20 Å, h c 26.8 Å. This curve explains the trend of the data, but the in-plane lattice spacing below the critical thickness, a s 4.20 Å, cannot be explained by any obvious epitaxial matching between the substrate and cuprite, because it does not correspond to any simple substrate periodicity. The boundary between the regimes II and III is marked by remarkable change in the rate of increase of the integrated intensity with the film thickness. This change is not due to reversing the potential scan direction, which occurred much later at h 35 Å. Since the integrated intensity of the diffraction peak is proportional to the number of the scatterers, the integrated intensity should be proportional to the thickness if the film is uniform. The obvious deviation of this linear behavior in regime I and II indicates that the growth of the oxide film is not uniform, while a uniform growth of the oxide film was observed in regime III. The slight deviation from the uniform growth for the last three data points is due to peak broadening during the reduction, which slightly increased the integrated intensity. Based on the diffraction from the oxide film, the growth of the aqueous oxide can be summarized with the schematic model in Fig. 11. The panels of the figure correspond to the

7 5958 J. Chem. Phys., Vol. 110, No. 12, 22 March 1999 Chu, Robinson, and Gewirth four distinct regimes of Fig. 10. In regime I, the diffraction pattern was found to be strictly two-dimensional, with no sign of bulk cuprite. During this interval, adsorption of oxygen onto the surface led to the formation of a monolayer oxide film with a smaller in-plane lattice constant and some diffusion into the substrate might have also occurred. The adsorbed layer could become the nucleus for the oxide. In growth regime II, the diffraction pattern of cuprite structure was first observed. Highly nonuniform film up to 27 Å thick was observed, possibly involving formation of narrow oxide pits into the copper substrate. The oxide film was highly strained with average lateral strain of 1.7% and the oxide unit cell volume compressed by about 4%. The diffraction intensity was very weak and did not increase much with the film thickness. In growth regime III, the intensity of the oxide diffraction peak increased rapidly, yet the film thickness has not increased significantly, probably because of lateral spreading of the oxide so that the regimes between the oxide nuclei were becoming converted to oxide. The epitaxial strain of the film became somewhat reduced, and the unit cell volume of the oxide relaxed rapidly toward the bulk value. In the final phase of the growth, denoted IV, the film was growing uniformly. Before the passivation, the inplane lattice spacing of the oxide film was still strained by about 1%, but the unit cell volume had almost completely relaxed. V. DISCUSSION Our x-ray diffraction study demonstrates that the same epitaxial oxides are formed during both native and aqueous oxidation at ph 4.5. Although their chemical structure and epitaxial relationship with the substrate are identical, the native and aqueous oxides are different in many important respects. The most interesting contrast was found in their growth properties. During the native oxidation, both parallel aligned and anti-parallel reversed cuprites were observed to grow three-dimensionally 3D. The same results have been reported for oxide grown under vacuum at various temperatures in the range 150 to 800 C. 1,2,19 However, during the aqueous oxidation at ph 4.5, only the reversed cuprite was found to develop into a 3D bulk phase, while the aligned cuprite remained as a very thin film, no more than a monolayer thick. In an experiment performed at ph 13 data not shown, we found only the reversed cuprite was stable: the initial aligned cuprite of the native oxide, disappeared completely upon the first reduction cycle and did not reappear. Therefore, the preferential growth of the reversed orientation over the aligned orientation appears to be a characteristic of the aqueous oxidation, which is fundamentally different from the dry oxidation. An obvious reason for the preferential growth of the reversed orientation is that the reversed cuprite might have a more energetically favorable interface with the substrate. However, the fact that both orientations were observed in the dry oxidation suggests that the energy difference cannot be very large. Another mechanism that might lead to oriented growth is preferred nucleation of oxide along the surface steps, which, because of miscut or free-energy considerations, might themselves have a preferred crystallographic orientation. During dry oxidation in vacuum, the pyramidal growth of oxide along 110 and 11 0 directions observed on Cu 111 was attributed to the preferred nucleation along the steps. 1 In addition, regions of oxide with the same orientation have been found to form islands with sizes around 0.1 m. 1,2,20 Since the islands of opposite orientations are too dissimilar to have originated from the same initial nucleus by means of tilt or slip, it was proposed that the oxide orientation might be determined at the time of nucleation at the steps. 20 This model of oxide nucleation at steps has some appeal as it explains naturally how the nucleating oxide can sense the stacking sequence of the copper without requiring influence of second nearest neighbors in the lattice. We can rule out the role of crystal miscut alone as we observed the orientational preference effect on more than one sample with different directions of miscut. Regardless of the exact reason, the preferential growth of the reversed oxide still provides valuable insights into the aqueous oxidation process. If the oxidation were to occur at the surface i.e., at the interface between the oxide film and the electrolyte, the crystallographically preferred oxide growth would be very difficult to maintain because the possible dislocations and defects in the oxide film would interfere with the perfect selection of one epitaxial orientation over the other. For this reason, we conclude that the growth must occur at the interface between the Cu 111 substrate and the oxide film: the oxide grows inward into the substrate rather than the outward into the electrolyte. We note that this peculiar oxide growth property could have significant implications in corrosion/oxidation inhibition of copper surfaces. Since the aligned cuprite does not grow beyond a monolayer, once this layer is in place, the surface would not be oxidized further; the aligned oxide acts as a corrosion barrier. Further differences between the native and aqueous oxides were also found in their epitaxial strain and relaxation behavior. When the native oxide film was thin, the observed in-plane spacing was very close to the 3 7 commensurate spacing, while that of the surface-normal was much larger than the bulk spacing. As the film grew thicker with time, both in-plane and the surface-normal spacings approached the bulk value. These results for the native oxide are in reasonable agreement with the findings on the oxide grown in vacuum. 1 On the other hand, the aqueous oxide film displayed both much larger strain and very different relaxation behavior. Here the observed values for the in-plane lattice spacing for the thin aqueous oxide film were much smaller than the 3 7 commensurate spacing see Fig. 9. We believe this deviation from normal elastic behavior is a signature of changing stoichiometry of the oxide, due to the incorporation of lattice vacancies during nucleation. Analogous behavior has been reported during passivation of iron. 17 Even when the aqueous oxide film became considerably thick, its in-plane spacing remained below the 3 7 commensurate spacing, while that of the native oxide was always greater. Furthermore, the trend in the surface-normal lattice spacing was exactly opposite to that of the native oxide. Consequently, the volume of the aqueous oxide unit cell was significantly compressed about 4% at the beginning of the oxidation and eventually relaxed toward the bulk value. Al-

8 J. Chem. Phys., Vol. 110, No. 12, 22 March 1999 Chu, Robinson, and Gewirth 5959 between the native and aqueous oxides. Our results show that the aqueous oxidation process is fundamentally different from the formation of the native oxide. In the aqueous oxidation experiments, we also discovered a new oxide phase at more negative potentials than the bulk oxide. By the symmetry of its diffraction we identified this as a monolayer oxide phase. We can include the new phase in the Pourbaix diagram as shown in Fig. 12. Its range of stability in potential was found to extend from 0.23 V to 0.15 V in Fig. 7. It therefore lies just below the potential for formation of bulk Cu 2 O, which is 0.15 V at ph 4.5. Monolayer phases at chemical potentials outside the stable range for the bulk phase occur widely in surface science, for example in adsorption of monolayers of gas on solids at temperatures higher than bulk condensation 21,22 or layer-bylayer freezing of liquid crystals. 23 FIG. 12. Modification to the Pourbaix diagram for the copper/water system to include the new monolayer phase identified in this work. This is a phase diagram representing the equilibrium state of Cu as a function of ph and potential Ref. 13. The dashed lines indicate the potential range of stability of H 2 O versus formation of H 2 and O 2, as labeled. The Cu concentration dependence is not shown and is fixed at 10 2 M. Potentials are referenced to the normal hydrogen electrode NHE. The arrow and dotted line shows the ph used in these experiments. The shaded region in the center of the figure denotes the range of stability in potential of the monolayer oxide at this ph. most complete relaxation 99.7% of the bulk volume was achieved after passivation by a thick oxide layer. The very different relaxation behavior of the aqueous versus the native oxide film must be attributed to differences between the chemical species involved at the microscopic level. Native oxidation in air is presumably due to reaction of O 2 and/or H 2 O molecules, while aqueous oxidation under the conditions we consider involves just H 2 OorOH species. Since oxidation takes place at the interface between the growing film and the bulk metal, the differences may arise from different diffusion mechanisms of the respective species through the oxide. There may also be a structural difference between the aqueous metal/oxide interface and the native one, which we have not yet been able to detect. The relevance of this to the general chemistry of oxidation highlights its importance and invites further investigation. VI. CONCLUSION The in situ x-ray diffraction technique was used to study the aqueous oxidation of a Cu 111 single crystal surface in 0.1 M NaClO 4 at ph 4.5. The observed native oxide and the aqueous oxide exhibited the same cuprite structure and the same epitaxial relationship with the Cu 111 substrate, in agreement with the results of the previous studies of the oxide grown under vacuum. Many details of the growth and strain relaxation behavior were found to differ dramatically ACKNOWLEDGMENTS Support for this research has been provided by the Metals and Ceramics division of the Department of Energy DOE through the University of Illinois Materials Research Laboratory under Grant No. DEFG02-96ER The NSLS is supported by the DOE under Contract No. DEAC02-98CH I. H. Ho and R. W. Vook, J. Cryst. Growth 44, D. A. Goulden, Philos. Mag. 33, G. W. Simmons, D. F. Mitchell, and K. R. Lawless, Surf. Sci. 8, F. Jensen, F. Besenbacher, E. Laegsgaard, and I. Stensgaard, Surf. Sci. 259, L L. H. Dubois, Surf. Sci. 119, F. Jensen, F. Besenbacher, E. Laegsgaard, and I. Stensgaard, Phys. Rev. B 42, U. Döbler, K. Baberschke, J. Stö, and D. A. Outka, Phys. Rev. B 31, H. You, C. A. Melendres, Z. Nagy, V. A. Maroni, W. Yun, and R. M. Yonco, Phys. Rev. B 45, B. J. Cruickshank, D. D. Sneddon, and A. A. Gewirth, Surf. Sci. 281, L J. R. LaGraff and A. A. Gewirth, Surf. Sci. 326, L N. Ikemiya, T. Kubo, and S. Hara, Surf. Sci. 323, Y. S. Chu, I. K. Robinson, and A. A. Gewirth unpublished. 13 M. Pourbaix, Atlas of Electrochemical Equilibria in Aqueous Solution Pergamon, New York, Y. S. Chu, I. K. Robinson, and A. A. Gewirth, Phys. Rev. B 55, I. K. Robinson, H. Graafsma, Å. Kvick, and J. Linderholm, Rev. Sci. Instrum. 66, R. W. G. Wychoff, Crystal Structures, 2nd ed., Vol. 1 Wiley, New York, M. F. Toney, A. J. Davenport, L. J. Oblonsky, M. P. Ryan, and C. M. Vitus, Phys. Rev. Lett. 79, J. H. van der Merwe, Surf. Sci. 31, D. F. Mitchell and K. R. Lawless, J. Paint Technol. 38, R. H. Milne and A. Howie, Philos. Mag. A 49, D. M. Zhu and J. G. Dash, Phys. Rev. Lett. 60, M. Bienfait, Europhys. Lett. 4, B. D. Swanson, H. Stragier, D. J. Tweet, and L. B. Sorensen, Phys. Rev. Lett. 62,