Michel
Sassi
* and
Kevin M.
Rosso
Physical and Computational Sciences Directorate, Pacific Northwest National Laboratory, Richland, Washington 99352, USA. E-mail: michel.sassi@pnnl.gov
First published on 18th December 2023
Developing a better understanding of water ordering and hydroxylation at oxide mineral surfaces is important across a breath of application spaces. Recent vibrational sum frequency generation (vSFG) measurements on MgO(100) surfaces at ambient conditions showed that water dissociates and hydroxylates the surface yielding a non-hydrogen bonded hydroxyl species. Starting from previously determined water hydroxylation patterns on MgO(100), we performed ab initio thermodynamic calculations and vibrational analysis to compare with the vSFG observations. At ambient conditions (i.e., T = 298.15 K and pH2O = 32 mbar), the most thermodynamically favorable surface hydroxylation is found to be p(3 × 2) – 8H2O, involving a dissociation of 25% of the adsorbed water. Analysis of the vibrational density of states for this hydroxylation configuration yielded three different hydrogen bonding environments with the frequency of the peaks in very good agreement with the vSFG measurements. However, the non-H-bonded spectral feature on this surface is predicted to be similar to that expected for Mg(OH)2, a thermodynamically downhill alteration of the surface that must be independently ruled out before one can be fully confident in the apparent theory/vSFG agreement. Our study provides more insights into the ordering and structure of water monolayer at MgO(100) surface at ambient conditions and completes previous theoretical and experimental analysis performed at low temperature and ultra-high vacuum conditions.
It is therefore unsurprising to find a lack of consensus in the literature regarding the hydration of MgO(100) when contacted by water. Using single crystal surfaces, early experiments using ultraviolet photoemission spectroscopy (UPS)33 and infra-red spectroscopy (IR)34 could not detect if water dissociates to form hydroxyl groups on the surface. In agreement with this experimental uncertainty, early density functional theory (DFT) simulations using the Car–Parrinello method concluded that water physisorbs on defect-free MgO(100) terraces, and that it only dissociates at surface defect sites such as steps.35 A few years later, ab initio investigations indicated that water strongly reacts with MgO(100) terraces and spontaneously dissociates at a ratio of two out of six water molecules adsorbed.13,15 This partial dissociation of water on well-ordered and low-defect MgO(100) surfaces was later experimentally confirmed.19–21
At low coverage, ultra-high vacuum (UHV) studies11 indicate formation of two ordered water monolayer phases, namely c(4 × 2), detected in the temperature range 100–180 K, and p(3 × 2), found at temperatures of 185–221 K. A more complete structural model of the p(3 × 2) phase was obtained from additional investigations using quantitative low energy electron diffraction (LEED) aided by semiempirical potential calculations performed at 0 K.36 For low temperature and UHV conditions, Włodarczyk et al.25 performed ab initio simulations and used a genetic algorithm to identify eight hydroxylation configurations for the p(3 × 2) and c(4 × 2) symmetries. Their predictions agreed well with the observed temperature dependence of surface phase changes and water desorption. The findings indicated that the low-temperature (T < 200 K) phase is composed of ten water molecules per unit cell with c(4 × 2) symmetry while at higher temperature (200 K ≤ T ≤ 202 K) the structure containing six water molecules per unit cell with p(3 × 2) symmetry is the most stable phase. At temperature higher than 202 K, water desorption from the surface was found to be more thermodynamically favorable. The most stable hydroxylated structure found for c(4 × 2) and p(3 × 2) respectively yielded 1/5 and 1/3 of the water molecules being dissociated. Extending this work to ambient conditions and for more than a monolayer of water coverage, Ončák et al.32 found that a fully hydroxylated surface is not energetically favorable while a reconstructed surface, involving hydrated/hydroxylated Mg2+ ions above the surface, are more stable.
Despite this comprehensive work, direct validation of the hydroxylation and hydration structure of MgO(100) at ambient conditions remains experimentally challenging. Recently, vibrational sum frequency generation (vSFG) spectroscopy in the OH stretching region showed a predominant non-hydrogen bonded peak at ∼3700 cm−1 in addition to a broad spectrum of states at lower frequencies consistent with an H-bonded network of surface hydroxyls and adsorbed water.31 Although the vSFG features alone do not unambiguously inform on the specific structure at the surface, they do provide a basis for comparison to theoretical vibrational spectra predicted for specific hydroxylation/hydration patterns.
To make this comparison, here we report ab initio thermodynamics (AIT) calculations to examine the vibrational spectra for specific surface configurations of adsorbed water and hydroxyls on MgO(100), and to relate these characteristics to the relevant state variables that define the chemical potential of water, namely temperature and the partial pressures of O2 and H2. To do so we revisit the series of eight hydroxylated structures found by Włodarczyk et al.25 and expand the work performed by Ončák et al.32 for ambient conditions who limited their investigation to the two most stable hydroxylation structures for c(4 × 2) and p(3 × 2) identified by Włodarczyk et al.25 for low temperature and UHV conditions. By using the AIT approach, we investigate how the chemical potential of oxygen and hydrogen can affect the relative phase stability of different hydroxylated configurations. While four hydroxylation configurations have a favorable surface Gibbs free energy, we found that the most thermodynamically stable one under ambient conditions is different to that identified by Włodarczyk et al.25 for low temperature and UHV conditions. The predicted low energy hydroxylated configurations yield vibrational spectra that compare well to the vSFG results of Adhikari et al.31 However, we also show that the prominent non-H-bonded feature at ∼3700 cm−1 is overlapping with the spectrum predicted for a bulk-like Mg(OH)2 termination, a structure that is based entirely on non-H-bonded hydroxyls. Therefore, to unambiguously identify the hydroxylated surface structure in experiments requires careful elimination of the prospect of an Mg(OH)2-like termination.
In order to assess the accuracy of the PBEsol functional for the calculation of surface energies, some tests on surface energy of clean (i.e., defect-free) MgO(100) surface and slab size were performed as shown in Table 1. A comparison with the surface energy calculated using the Perdew, Burke, and Ernzerhof (PBE) functional42 with Grimme (D3) dispersion corrections43 and the Becke–Johnson (BJ) damping function44 (i.e., PBE + D3 (BJ)) is also provided. The calculated surface energy of MgO(100) with PBEsol is in very good agreement with the experimental value of 1.04 J m−2,45 while the PBE + D3 functional overestimates the surface energy by about 30%. Depending on the functional used, previous theoretical calculations estimated the MgO(100) surface energy in the range of 0.83 to 1.29 J m−2.46
Clean (100) Surface | γ (J m−2) | ||
---|---|---|---|
(1 × 1) 8 Layers | p(3 × 2) 4 Layers | c(4 × 2) 4 Layers | |
PBEsol | 0.99 | 1.00 | 1.00 |
PBE + D3 (BJ) | 1.35 | 1.35 | 1.35 |
Expt.45 | 1.04 |
The surface Gibbs free energy dependance to oxygen and hydrogen chemical potentials for various hydroxylation configurations of the MgO(100) surface has been evaluated within the ab initio thermodynamic framework. In these calculations, the surface Gibbs free energy is given by:
(1) |
Esurface = ETotalsurface + EZPEsurface + Δμ(T,P)surface | (2) |
EBulkMgO = ETotalMgO + EZPEMgO + Δμ(T,P)MgO | (3) |
(4) |
(5a) |
(5b) |
A comparison of calculated and NIST-JANAF experimental48 Δμ(T,P) for bulk Mg and MgO crystals is shown in Fig. 1. The agreements between calculated and experimental values are within 2 kJ mol−1 across the temperature range.
Fig. 1 Comparison of experimental and calculated chemical potentials for Mg and MgO bulk materials as function of temperature at a pressure of P = 1 bar. |
In order to account for the atomization errors induced by GGA for O2, H2, and H2O, the DFT total energies of these species have been corrected by using their experimental, zero-point energy corrected, atomization energies.42,49 Therefore, the total energy of O2, H2, and H2O molecules were respectively corrected by 1.483 eV, −0.104 eV, and −0.515 eV.
In the cases where water dissociates (into H+ and OH−) at the surface, Fig. 2 shows that OH groups are formed at the surface with a hydrogen atom forming a covalent bond with a surface O atom, while the remaining hydroxyl anion is only interacting with the water monolayer by forming hydrogen bonds with nearby water molecules. In the case where water does not dissociate, the water hydrogen of the first water monolayer forms a hydrogen bond with the oxygen atoms of the MgO(100) surface. The other water hydrogen of the first water monolayer is involved in hydrogen bonding with the water oxygen of the second water monolayer. The formation of such hydrogen bond network between first and second water monolayers helps prevent water dissociation.
The relative stability of each hydroxylation configuration has been evaluated with ab initio thermodynamics, in which the surface Gibbs free energy was calculated as function of the oxygen and hydrogen chemical potentials at T = 298.15 K. A negative surface Gibbs free energy indicates that the surface is energetically and thermodynamically favorable to form while a surface with positive Gibbs free energy suggests that the formation of the surface is not favorable. As shown in Fig. 3, the most thermodynamically favorable surface configuration for μO ≤ −3.52 eV is the clean MgO(100) surface. For μO values between −3.52 eV and −2.72 eV, the surface hydroxylation configuration c(4 × 2) with 10H2O is found to be more thermodynamically favorable than the clean surface. For μO ≥ −2.72 eV, the surface hydroxylation configuration p(3 × 2) with 8H2O is found to be more thermodynamically favorable than c(4 × 2) with 10H2O.
Beside the influence of oxygen chemical potential, the chemical potential of H also impacts the relative stabilities. Interestingly, the clean MgO(100) and the c(4 × 2) – 10H2O surfaces are energetically competitive only when hydrogen is in equilibrium with H2. For hydrogen in equilibrium with H2O, the surface configuration p(3 × 2) with 8H2O is the most energetically favorable surface. More detailed surface Gibbs free energy values for H in equilibrium with H2O (i.e., μO ≥ −2.72 eV) at T = 298.15 K are provided in Table 2. Of the nine hydroxylation configurations investigated, four have a negative surface Gibbs free energy, p(3 × 2) – 8H2O, c(4 × 2) – 10H2O, p(3 × 2) – 12H2O, and p(3 × 2) – 7H2O with a surface energy of −6.191 meV Å−2, −5.546 meV Å−2, −3.328 meV Å−2, and −0.888 meV Å−2 respectively. While the bilayer p(3 × 2) – 12H2O configuration, for which no water dissociation occurs, is energetically favorable (γ(μO, μH) < 0) at room temperature, two monolayer surface configurations for which water dissociation is predicted to occur are lower in energy, indicating that water dissociation should be preferrable at ambient conditions (i.e., T = 298.15 K and P = pH2O = 32 mbar). Interestingly, Fig. 3 and Table 2 show that all nine hydroxylated surface configurations investigated are more energetically favorable than clean MgO(100) surface when H is in equilibrium with H2O. This is in good agreement with MgO being a hydrophilic mineral and thus strongly interacting with water.
Surface configuration | γ(μO, μH) (meV Å−2) |
---|---|
Clean MgO(100) | 60.426 |
p(3 × 2) – 5H2O | 15.736 |
p(3 × 2) – 6H2O | 6.844 |
p(3 × 2) – 7H2O | −0.888 |
p(3 × 2) – 8H2O | −6.191 |
p(3 × 2) – 12H2O | −3.328 |
c(4 × 2) – 7H2O | 14.452 |
c(4 × 2) – 8H2O | 8.119 |
c(4 × 2) – 9H2O | 0.670 |
c(4 × 2) – 10H2O | −5.546 |
Previous thermodynamic calculations by Włodarczyk et al.25 carried out for similar hydroxylated surface configurations as function of temperature and at P = 10−10 mbar, indicated that c(4 × 2) – 10H2O and p(3 × 2) – 6H2O were the most stable surfaces configurations for temperatures below 202 K. For temperatures above 202 K, a complete desorption of water from the MgO(100) surface was predicted. For the thermodynamic conditions selected (i.e., low temperature and pressure), these theoretical results were in good agreement with ultra-high vacuum (UHV) experiments performed at 85 K during water adsorption. The work performed in our study completes the theoretical analysis of Włodarczyk et al.25 by focusing on the hydroxylation of MgO(100) surface at ambient conditions (i.e., T = 298.15 K and P = pH2O =32 mbar). At those conditions, we found that p(3 × 2) – 8H2O is the most thermodynamically stable surface configuration.
In connection with the recent vSFG study performed by Adhikari et al.31 at ambient laboratory conditions, we calculated the phonon dispersion relationship using the Phonopy code47 for the two most thermodynamically favorable surface hydroxylation configurations. The vibrational density of states (VDOS) for p(3 × 2) – 8H2O and c(4 × 2) – 10H2O are shown in Fig. 4 along with the experimental frequency range of different hydrogen bonding environments, as detected by vSFG. While a direct comparison of the calculated VDOS and experimental vSFG spectral intensity is not possible because the VDOS does not have any selection rule and it would require the computation of more complex quantities, such as the second order susceptibility which can be calculated from molecular dynamic simulations,51 one can still compare the frequencies of the peaks. From the calculated VDOS for p(3 × 2) – 8H2O, three types of hydrogen bonds can be distinguished with frequencies higher than 3300 cm−1. Those correspond to hydrogen bonded surface hydrogen (labelled H-bonded Hs), non-hydrogen bonded surface hydrogen (labelled non-H-bonded Hs), and non-hydrogen bonded hydrogen of the hydroxyl anion (labelled non-H-bonded HOH) with frequency peaks located at 3390 cm−1, 3642 cm−1, and 3778 cm−1 respectively.
Fig. 4 Vibrational density of state of the two most thermodynamically stable hydroxylated configurations of MgO(100). Vertical blue, pink, and green bands delimit the experimental frequency range of different H bonding environments, as detected by vSFG.31 The red dashed line indicates the calculated frequency of OH stretch in bulk Mg(OH)2. |
In the case of surface hydroxyl groups (OHs) originating from water dissociation, only the Hs located right under a hydroxyl anion are not bonded to any surrounding water molecules as those water molecules preferentially form hydrogen bonds with the hydroxyl anions. This feature gives rise to two types of hydrogen bonding environments for surface hydroxyl (OHs) groups, with the hydrogen bonded ones having a lower frequency than the non-hydrogen bonded ones. The frequencies associated with the three types of hydrogen bonding environments are in good agreement with the vSFG frequency peaks detected experimentally,31 as shown by the colored zones in Fig. 4. While c(4 × 2) – 10H2O is the second most thermodynamically stable hydroxylation surface configuration, only two peaks can be distinguished for frequencies higher than 3300 cm−1. These peaks correspond to non-hydrogen bonded OHs (3630 cm−1) and to hydrogen bonded hydroxyl anions (3754 cm−1). In this hydroxylated surface configuration, there are no hydrogen bonded OHs, and thus, lower frequencies in the range from about 3350 cm−1 to 3550 cm−1 are not present. This suggests that under ambient conditions, the MgO(100) surface is hydroxylated similarly to p(3 × 2) – 8H2O, which is the most thermodynamically favorable surface configuration and shows a vibrational spectra with frequency positions above 3300 cm−1 in good correlation with vSFG experiments.
For the calculated VDOS with frequencies lower than 3300 cm−1, which would mainly correspond to the hydrogen bond network between water molecules, we note that the experimental vSFG has intensity in this frequency region. However, the vSFG peaks are at least 100 times less intense than the probed OH stretches, which have a strong vSFG signal. The small vSFG intensity of the hydrogen bonded water frequencies can be observed in Fig. S4 and S7 (ESI†) of the study by Adhikari et al.31
However, conceptually there is another way to produce non-H-bonded hydroxyl spectral signal at the hydrated MgO surface – by converting it wholly or in part to more thermodynamically favorable Mg(OH)2.32 For example, using the same DFT methods we calculated that the thermodynamic driving force to convert bulk MgO to Mg(OH)2 at ambient conditions (i.e., T = 298.15 K and pH2O = 32 mbar) is −0.71 eV (or −68.42 kJ mol−1). In the bulk structure of Mg(OH)2, every OH group can be considered non-H-bonded, as shown in Fig. S1 (ESI†), as they each point toward the center of a triangle made of three octahedral Mg atoms. The frequency associated to the OH stretch is calculated to be 3636 cm−1 and has been reported in Fig. 4 by a red dashed line. As this frequency overlaps with the non-H-bonded Hs range of frequencies, chemical transformation of the MgO surface to an Mg(OH)2-like termination would likely yield a prominent non-H-bonded feature (pink band) in the vSFG spectrum. The remaining hydroxylation/hydration spectral weight at lower frequencies would in turn depend on the structural details of the Mg(OH)2-like termination onto which water adsorbs and reacts at a given temperature and pressure, which is beyond the scope of the current study. Nonetheless, these results suggest that although our findings predict a good correspondence between the p(3 × 2) – 8H2O hydroxylated MgO(100) configuration, and that this configuration is low in energy, a contribution from an Mg(OH)2-like termination cannot yet be entirely ruled out.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3cp04848a |
This journal is © the Owner Societies 2024 |