Geometry of α-Cr2O3(0001) as a Function of H2O Partial Pressure

Surface X-ray diffraction has been employed to elucidate the surface structure of α-Cr2O3(0001) as a function of water partial pressure at room temperature. In ultra high vacuum, following exposure to ∼2000 Langmuir of H2O, the surface is found to be terminated by a partially occupied double layer of chromium atoms. No evidence of adsorbed OH/H2O is found, which is likely due to either adsorption at minority sites, or X-ray induced desorption. At a water partial pressure of ∼30 mbar, a single OH/H2O species is found to be bound atop each surface Cr atom. This adsorption geometry does not agree with that predicted by ab initio calculations, which may be a result of some differences between the experimental conditions and those modeled.


■ INTRODUCTION
The presence of a passive surface film is key to the exceptional corrosion resistance of stainless steel alloys. 1,2 Consequently, much effort has targeted the characterization and enhancement of this protective layer, which is often composed, at least partially, of chromia. 1,2 Such work includes fundamental studies of single crystal surfaces of α-Cr 2 O 3 to gain atomic scale insight into pertinent properties, e.g., refs 3−10. To date, however, most of these measurements have been conducted in ultra high vacuum (UHV), limiting their relevance with regard to mechanistic understanding of corrosion performance in engineering environments. Targeting this omission, the current study is concerned with determining the surface structure of α-Cr 2 O 3 (0001) in the presence of H 2 O vapor through acquisition of surface X-ray diffraction (SXRD) data; water is an essential ingredient for many corrosion phenomena.
The structure of α-Cr 2 O 3 (0001) as a function of both H 2 O partial pressure and temperature has previously been explored by Costa et al. through ab initio modeling. 6 As a starting point for these calculations, a clean surface terminated by a single layer of 3-fold coordinated chromium atoms was assumed, as depicted in Figure 1A; this surface termination is labeled Cr− O 3 −Cr− on the basis of its first three atomic layers (the subscript indicates the average number of atoms in each 1 × 1 unit cell). Near room temperature, it was concluded that two other terminations become energetically favorable in the presence of H 2 O. At lower H 2 O partial pressures, dissociative adsorption was proposed to be the most likely scenario with each surface Cr becoming decorated with two hydroxyls (OH), i.e. (OH) 2 Figure 1, parts B and C. These theoretical structures are consistent with experimental characterization performed under UHV conditions of thin films of α-Cr 2 O 3 (0001) exposed to H 2 O, in that at room temperature, dissociative adsorption is evident. 3,4 Moreover, the existence of two distinct OH species, as well as interadsorbate hydrogen bonding, is apparent from vibrational data. 3 Here, the validity of the theoretical study of Costa et al. 6 is explored experimentally. Analysis of SXRD data acquired at a H 2 O partial pressure of ∼30 mbar indicates that the α-Cr 2 O 3 (0001) surface is decorated by OH, but not exactly as predicted. This work builds on a previous SXRD study examining the impact of O 2 on the surface structure of α-Cr 2 O 3 (0001). 7

■ EXPERIMENTAL METHODS
Experimental work was carried out at the Diamond Light Source (DLS) synchrotron facility, employing the Surface Village's off-line UHV chamber for sample preparation, and beamline I07 for SXRD measurements. In situ cleaning of the single crystal α-Cr 2 O 3 (0001) sample (supplied by PI-KEM Ltd.) involved repeated cycles of Ar + bombardment and annealing in UHV to approximately 1200 K. Low energy electron diffraction (LEED), and auger electron spectroscopy (AES) facilities were employed for sample characterization; LEED and AES data can be found in Supporting Information. It should be noted that following acquisition of LEED and AES data, the sample underwent a further cycle of Ar + bombardment and annealing to minimize the possibility of surface damage due to electron beam impingement.
Following completion of surface preparation, the sample was exposed to ∼1000 L (Langmuir) of H 2 O vapor. Prior to dosing, the H 2 O had been degassed through repeated freeze−pumpthaw cycles. The sample was then transferred under vacuum to I07's diffractometer in EH1, using a custom-built vacuumsuitcase and UHV baby chamber combination. The latter (base pressure ∼1 × 10 −9 mbar) incorporates a dome shaped X-ray transparent beryllium window suitable for undertaking SXRD measurements. Once located on the beamline the sample was exposed to a further ∼1000 L of H 2 O vapor; henceforth this surface will be referred to as Cr 2 O 3 −H 2 O UHV . The purpose of dosing H 2 O prior to commencing diffraction measurements was to mitigate the risk of any surface contamination during the sample transfer process. SXRD data were collected at an incidence angle of 1°with the substrate at room temperature, using a photon energy of hv =17.7 keV and a 2D Pilatus photon detector. Initially, a systematic series of X-ray reflections was acquired from Cr 2 O 3 − H 2 O UHV . More specifically, for a given (h,k)-integer, data were measured as a function of l to facilitate generation of so-called crystal truncation rods (CTRs); fractional-order rods (FORs) were also surveyed. h, k, and l are the reciprocal lattice vectors, and are defined with reference to the real space (1 × 1) unit cell of the α-Cr 2 O 3 (0001) surface, described by lattice vectors (a 1 , a 2 , a 3 ) which are parallel to the [100], [010], [001] directions, respectively. The magnitudes of these lattice vectors are a 1 = a 2 = a = 4.957 Å, and a 3 = c = 13.592 Å, 7 where a and c are the bulk lattice constants.
Subsequent to compiling surface diffraction data from Cr 2 O 3 −H 2 O UHV in UHV, the H 2 O partial pressure was increased in a stepwise fashion by appropriate backfilling of the baby chamber with H 2 O. We note that above 1 × 10 −4 mbar a static volume of H 2 O was employed rather than obtaining an equilibrium pressure through balancing the rates of H 2 O inflow and pumping, i.e. the baby chamber was no longer continuously pumped. For each H 2 O partial pressure, the intensity of the (1, 0, 2.9) reflection was monitored to identify changes in the α-Cr 2 O 3 (0001) surface structure. Selection of this reflection was based upon its sensitivity to such variation as a function of O 2 partial pressure. 7 On the basis of these measurements (see below), a further systematic series of X-ray reflections was acquired from α-Cr 2 O 3 (0001) at a H 2 O partial pressure of ∼30 mbar; henceforth this surface will be referred to as Cr 2 O 3 −H 2 O 30mbar . It should be noted that as 30 mbar of H 2 O is equivalent to ∼100% relative humidity with the substrate at room temperature, one would expect the surface to be submerged beneath multiple monolayers of H 2 O in this environment. 11 To facilitate fully quantitative structure determination, the raw diffraction data acquired at UHV and p(H 2 O) ∼ 30 mbar were integrated and corrected 12 to enable plots of structure factor versus perpendicular momentum transfer for each CTR to be compiled. This procedure resulted in a total of 1054 (1142) nonequivalent reflections from six CTRs for Cr 2 O 3 − H 2 O UHV (Cr 2 O 3 −H 2 O 30mbar ). Concerning FORs, no evidence for any surface unit cell other than 1 × 1 was found.
For surface structure determination, we adopted the usual approach of generating simulated SXRD data for a series of potential model structures, and iteratively refining structural (and nonstructural) parameters to find the overall best fit between experiment and theory. The ROD software was employed for this purpose. 13 Reduced χ 2 was used to evaluate the goodness of the fit; this is defined as follows: The Journal of Physical Chemistry C Article N is the number of measured structure factors, P the number of parameters optimized during fitting, and F i exp (hkl) and F i th (hkl) are the experimental and theoretically calculated structure factors, respectively. σ i exp (hkl) is the uncertainty associated with F i exp (hkl). χ 2 behaves such that a value of 1 indicates that experiment and theory are essentially coincident, with agreement decreasing with increasing χ 2 . Values of χ 2 significantly less than 1 suggest that the magnitudes of experimental uncertainties have been overestimated. The quoted precision of each fitted parameter is determined by systematically varying the parameter about its optimal value, and for each step optimizing all other parameters, until χ 2 has increased by 1/(N−P) from its minimum value. 14 ■ RESULTS Figure 2 displays the intensity of the (1, 0, 2.9) reflection as a function of increasing H 2 O partial pressure; please note, as described above, the sample had already been dosed with ∼2000 L of H 2 O prior to acquisition of these data. Upon exposure of the sample to ∼30 mbar of H 2 O vapor, there is an increase of ∼20% in the signal. This increase is fully reversible (i.e., there is an ∼20% decrease in intensity upon reducing the pressure down to ∼8 × 10 −7 mbar), which indicates that the change occurring at 30 mbar of H 2 O is not maintained at lower vapor pressures. The inset in Figure 2 compares rocking scans acquired in UHV (∼1 × 10 −9 mbar) and ∼30 mbar, demonstrating the significance of the variation in reflection intensity. Furthermore, this comparison shows that there is no appreciable variation in the width of the reflection, indicating that terrace size is not significantly influenced by the presence of H 2 O. These data suggest that the presence of ∼30 mbar of H 2 O vapor leads to a modification of the surface structure of α-Cr 2 O 3 (0001). This supposition will be confirmed below, through analysis of the CTR data sets acquired from Initially, attention was focused upon the diffraction data acquired from Cr 2 O 3 −H 2 O UHV . To begin the search for a structural solution, the clean surface structure (Cr 2 O 3 − clean UHV ) determined in recent quantitative LEED (LEED-IV) 15  . Given that the current measurements were undertaken following exposure to ∼2000 L of H 2 O, terminations with surface Cr atoms bonded to one or more OH/H 2 O species were tested. It should be noted that H atoms were not explicitly included during generation of simulated of SXRD data, due to their negligible X-ray scattering, i.e. only an oxygen atom was added for each OH/ H 2 O. Refinement of these OH/H 2 O decorated structures, including atomic coordinates, site occupation, and a surface roughness parameter (β), resulted in χ 2 values of 1.7, 2.1, and 1.8 for Cr atoms bound to one, two, or three OH/H 2 O species, respectively. For completeness, a similar structural refinement was undertaken without any adsorbed OH/H 2 O. Optimization of this structure resulted in a χ 2 of 1.2, indicating that the SXRD data provide no substantive evidence for adsorbed OH/H 2 O under the prevailing experimental conditions. Figure 3 displays a comparison of the experimental CTRs acquired from Cr 2 O 3 −H 2 O UHV with the best-fit theoretical simulations. To achieve this fit 41 parameters were optimized, i.e. 35 atomic coordinates, a scale factor, a surface roughness  The Journal of Physical Chemistry C Article parameter, and fractional occupancy factors for Cr(1), Cr (2), O(1), and Cr(3) (these atoms are identified in Figure 4); Debye−Waller factors for all atoms were maintained at bulk values, i.e., 0.5 Å 2 . As may be expected for a χ 2 value of 1.2, there is a good level of agreement between theory and experiment. In a number of regions away from Bragg peaks, however, the uncertainty in the experimental structure factor is relatively large and can encompass zero. Given this situation, which may lead one to question the reliability of the optimum structure, a further structure refinement was undertaken excluding all data points where the error in the structure factor includes zero. Employing this more limited data set did not result in any significant changes in the structural solution, and so was not considered further.
The surface geometry emerging from analysis of the data acquired from Cr 2 O 3 −H 2 O UHV is illustrated in Figure 4. Corresponding atomic coordinates are listed in Table 1; all nearest neighbor Cr−O interatomic distances are physically reasonable. A topmost partially occupied Cr double layer is maintained in the optimized structure, although the fractional occupancies of both Cr sites are less than those determined previously for Cr 2 O 3 -clean UHV ; 7,15 layer occupancies determined from analysis of SXRD data from Cr 2 O 3 −H 2 O UHV (the present study) and Cr 2 O 3 -clean UHV 7 are indicated in Figure 4. Atomic layer spacings perpendicular to the α-Cr 2 O 3 (0001) surface derived from both the present results and the earlier measurements 7,15 are listed in Table 2. As with the fractional occupancies, there are variations in these values, which are not negligible. As mentioned in ref 7, a plausible explanation for these discrepancies in surface structure is that they arise from small variations in sample preparation methods, e.g. anneal temperature. Furthermore, it should be remembered that the latest SXRD data were recorded after exposure to 2000 L of  a Fractional occupancy is indicated by a non-integer subscript in the "atom" column; the overall occupancy of oxygen atoms in the layer containing O(1) is 1.11 ± 0.12, as there are three symmetry equivalent oxygen atoms per (1 × 1) unit cell. Atomic coordinates for the bulkterminated Cr−Cr−O 3 -structure are also listed. Figure 4 provides a key to the identity of the atoms, and the axes x, y, and z. An asterisk ( * ) indicates that the parameter has been held constant during optimization. x and y coordinates not optimized due to symmetry constraints are italicized. off atop adsorption was also tested, but found to increase χ 2 . A comparison of the experimental CTRs with the best-fit theoretical simulations is shown in Figure 5. Figure 6 depicts the surface structural model employed to obtain the best-fit displayed in Figure 5, in which the oxygen atoms of adsorbed OH/H 2 O species are labeled with 1′, 2′, 3′, and 4′. As illustrated, the best fit was obtained with OH/H 2 O (O(1′) − O(4′)) located atop Cr(1) and Cr(2), as well as above any Cr(3) and Cr(4) atoms available for bonding due to fractional occupation of the topmost oxygen layer (O(1)). It should be borne in mind that the presence of O(3′) and O(4′) does not result in unphysical interatomic distances, as the fractional occupancy of these atoms is governed by fractional occupancy of O(1). Optimum atomic coordinates are listed in Table 3. Here, again, all nearest neighbor Cr−O interatomic distances are physically reasonable. During fitting 35 atomic coordinates were varied. In addition, as above, a scale factor, a surface roughness parameter, and fractional occupancy factors for Cr(1), Cr   Figure 4 indicates the identity of the atomic layers.   a Fractional occupancy is indicated by a non-integer subscript in the "atom" column; the overall occupancy of oxygen atoms in the layer containing O(1) is 1.2 ± 0.09, as there are three symmetry equivalent oxygen atoms per (1 × 1) unit cell. Atomic coordinates for the bulkterminated Cr−Cr−O 3 -structure are also listed. Figure 6 provides a key to the identity of the atoms. An asterisk ( * ) indicates that the parameter has been held constant during optimization. x and y coordinates not optimized due to symmetry constraints are italicized.   Figure 2, i.e., would expect increase in diffracted signal to occur at lower p(H 2 O). Concerning the structure determined at p(H 2 O) ∼ 30 mbar, it can be seen that it is inconsistent with the predictions emerging from the calculations of Costa et al. 6 They conclude that a (OH)−Cr−O 3 − termination is energetically unfavorable, contradicting our experimental structure determination. It may be argued that this discrepancy is again due to the influence of the X-ray beam. However, in our opinion, that the X-ray beam would selectively desorb OH/H 2 O species from each surface Cr leaving one remaining is doubtful. One more plausible reason for this difference, as stated in the Experimental Methods, is that the SXRD data have been acquired from a surface submerged beneath multiple monolayers of H 2 O, due to the relative humidity being ∼100%. 11 In contrast, Costa et al. do not explicitly include multiple layers of water in their modeling, 6 which may be the origin of the variation in interfacial structure. An alternative explanation for the divergence between the ab initio predictions and experiment is that the initial clean substrate termination used for the calculations is significantly different to that determined by diffraction. It could be that the more disordered surface found in the experiment hinders the formation of an extended network of OH/H 2 O hydrogen bonding, which emerges from the first-principles modeling, resulting in Cr binding to multiple OH/H 2 O species becoming energetically unfavorable. Another possible reason for the difference is that there is a significant energy barrier to the attachment of additional OH/H 2 O to each Cr, and so cannot be realized with the substrate at room temperature. Further ab initio modeling, which more closely mimics both the experimentally determined α-Cr 2 O 3 (0001) termination and the measurement environment, is required to test these hypotheses. For example, ab initio molecular dynamics could be employed to better simulate the submerged substrate. 16 Regarding   6 On this basis, one might propose that at 30 mbar of H 2 O, Cr 2 O 3 (0001) is decorated with adsorbed molecular H 2 O rather than OH. However, this deduction must be regarded with caution due to the approach employed for fitting the diffraction data, i.e. to limit the number of parameters optimized, the possibility for each substrate atom to adopt more than one location was not incorporated, even where there were variations in local environment. For example, regarding the Cr 2 O 3 − H 2 O 30mbar data, this approach results in employing only two atoms (Cr(3) and Cr(4)) to describe the first subsurface double layer of Cr atoms, even though some are bound to O(1) and others to O(3′) or O(4′) atoms. Given that this change in local coordination is likely to lead to somewhat different atomic coordinates, the Cr-(OH/H 2 O) inter atomic distance (2.09 Å) is less well-defined than indicated above. Consequently, we conclude that it is not possible to uniquely identify the adsorbate (OH or H 2 O) present on Cr 2 O 3 −H 2 O 30mbar from the current SXRD data set. However, the weight of other evidence 3,4,6 suggests that dissociative adsorption is more likely, i.e. the adsorbate is OH.
Comparing the present results to those obtained for the interaction of H 2 O with other (0001) surfaces of corundumtype metal oxides, of particular interest are near ambient pressure photoemission data from α-Fe 2 O 3 (0001). 17 In that study, it was concluded that as the partial pressure of H 2 O increases, the surface becomes increasingly decorated with surface OH, attaining a maximum coverage of ∼1 monolayer at ∼10 −4 mbar. Adsorbed H 2 O is also observed, suggested to be located above the OH layer, i.e. a three layer Fe 2 O 3 (0001)/ OH/H 2 O interface is formed. One might also expect similar layering on α-Cr 2 O 3 (0001) at 30 mbar of H 2 O, but only a single OH/H 2 O layer is evident from analysis of the SXRD data. However, it should be remembered that SXRD is sensitive to adsorbed layers displaying order both parallel and perpendicular to the surface plane of the substrate. Hence, the analysis presented here should not be interpreted as indicating that only a single layer of OH/H 2 O is present on Cr 2 O 3 (0001) at 30 mbar H 2 O, but that only this layer is sufficiently ordered to be apparent in SXRD. Again, 30 mbar of H 2 O is equivalent to ∼100% relative humidity with the substrate at room temperature, and so the surface is expected to be submerged beneath multiple monolayers of H 2 O. 11 Another result worthy of mention is the local adsorption geometry of OH obtained from photoelectron diffraction (PhD) measurements in UHV performed following exposure of V 2 O 3 (0001) to H 2 O. 18 In contrast to the current study, no evidence for OH atop surface V atoms was found. Instead, only the surface oxygen atoms (equivalent to O(1) in Figures 4 and 6) are hydroxylated through attachment of H, presumably derived from H 2 O dissociation; the location of the dissociated OH fragment is not identified. This difference may be due to the initial V 2 O 3 (0001) surface being terminated by vanadyl groups (VO), rather than under-coordinated V atoms. 18 Of further interest is a brief consideration of the previous SXRD study probing the surface structure α-Cr 2 O 3 (0001) as a function of oxygen partial pressure. 7 It is pleasing to note that away from UHV (p(H 2 O) ∼ 30 mbar (current study), and p(O 2 ) = 1 × 10 −2 mbar 7 ) the optimum surface structures are not identical, i.e. data have not simply been acquired from similarly contaminated surfaces due to extrinsic components of the ambient environment. For both H 2 O and O 2 , adsorption occurs atop under-coordinated surface Cr atoms, but the Cr−O bond distance is significantly shorter in the presence of O 2 (1.57 ± 0.03 Å). This shorter Cr−O distance is consistent with the formation of surface chromyl (CrO) groups. 7 Furthermore, unlike H 2 O/OH decorated α-Cr 2 O 3 (0001), the chromyl terminated surface remains intact following reduction of the O 2 partial pressure.
The Journal of Physical Chemistry C Article ■ CONCLUSIONS To summarize, SXRD data have been acquired from α-Cr 2 O 3 (0001) as a function of H 2 O partial pressure. In UHV, after exposure to ∼2000 L of H 2 O, the surface is terminated by a partially occupied double layer of chromium atoms; the lack of adsorbed OH/H 2 O is concluded to be most likely a result of either adsorption only at defects, or X-ray induced desorption. This surface geometry is largely consistent with those determined in recent LEED-IV 15