J. Phys. Chem. 1995,99, 17219-17225
17219
Atomistic Simulation of the Effect of Dissociative Adsorption of Water on the Surface Structure and Stability of Calcium and Magnesium Oxide N. H. de Leeuw," G. W. Watson, and S. C. Parker School of Chemistry, University of Bath, Claverton Down, Bath BA2 7AY, United Kingdom Received: May 30, 1995; In Final Form: September 12, 1995@
Hydroxylation of the {loo}, ( 1lo}, { 11l}, and (310) surfaces of MgO and CaO by dissociative adsorption of water has been studied using atomistic simulation techniques. We found that unstable surfaces with high surface energies tend to microfacet into steps of stable (100) planes, creating two possible sites for hydroxylation. As found experimentally, the perfect MgO (100) surface is shown not to be amenable to hydroxylation. However, on stepped {loo} surfaces, modeled by the (310) and faceted { 110) surfaces, adsorption is energetically favorable and preferentially occurs at low-coordinated sites. The calculated hydration energies for CaO and MgO are in good agreement with experimental values where available.
Introduction
In this paper we investigate the effect of the dissociative adsorption of water molecules on the surface structure and stability of the ceramic oxides MgO and CaO using atomistic simulation techniques. MgO is a basic oxide which catalyzes the H2 and D2 exchange reaction and the dehydrogenation of formic acid or methanol,' and as such understanding the surface structure and reactivity is of clear importance. It is also widely used as a support for metal catalysts and high-temperature superconductors, and water is known to be an effective catalyst for the sintering of MgO aggregates2 In addition, many oxide surfaces are hydroxylated, which may influence properties such as their mechanical response3 and catalytic b e h a ~ i o r . ~As a consequence, the structure and hydration of MgO surfaces, especially the {loo) surface, have been the subject of much research, both e~perimental'.~-~ and t h e o r e t i ~ a l , ~and - ~all ~ *agree ~ ~ that on the perfect { 100) surface water is only physisorbed while chemisorption is energetically unfavorable. However, a perfect surface cannot be obtained experimentally, and every real surface contains defects: kinks, comers, and edge^.^.^ It has been ~uggested~.'*-'~ that these sites, consisting of three- or fourcoordinated atoms rather than the five-coordinated atoms of the pure ( 100) surface, are the sites where chemical reactions take place, and although no conclusive quantitative experimental evidence has been shown, there are some qualitative experimental results available. Dunski et a1.,4 upon dehydroxylating the MgO {loo) surface by temperature-programmeddesorption (TPD), found three energetically different sites for probable chemisorption of water, and they ascribe these to three-, four-, and five-coordinated sites. Onishi et a1.* used UPS and XPS to probe the adsorption of water on the (100) and (111) surfaces and found that chemisorption of water does take place leading to hydroxylation of surface atoms. They also suggested that this adsorption takes place at surface defects on the {loo} plane and on the low-coordinated edge sites of the (111) surface. Jones et aL6 found that when MgO smoke particles were eroded, electron micrographs showed considerable surface roughening, increasing the number of low-coordinated sites, and infrared spectroscopic measurements showed a resulting increase in surface hydroxylation.
* TelOl225 - 826505; Fax 01225 - 826231; E-mail
[email protected]. @
Abstract published in Advance ACS Abstracts, November 1, 1995.
0022-365419512099-17219$09.00/0
In this paper we have calculated surface structures, energies, and hydration energies on dissociative adsorption of water on the fourpure surfaces (loo), (110), { l l l ) , and (310) andon the microfaceted ( 110) surface as a function of coverage. By considering various partial coverages, we can compare the results directly with microcalorimetry and TPD, and by calculating the energies for adsorption both at the normal fivecoordinated surface atoms and at four-coordinated atoms, we can identify the energetically favored configurations to investigate the assumption that low-coordinated sites should be more amenable to hydroxylation. CaO, which has the same rock salt structure as MgO, has been studied alongside MgO, because it is easily hydrated, and microcalorimetric studies by Fubini et al.I5 have identified the hydration energies of CaO after dosing with controlled amounts of water vapor. Theoretical Methods
The properties of CaO and MgO were modeled using atomistic simulation techniques based on the Born model of solids. As we wish to model the hydration energies as a function of coverage, atomistic simulation techniques are currently more appropriate over full electronic structure calculations because of the number of ions treated (up to 2200 in the extreme case). In addition, by treating a larger portion of crystal we can be confident that the energies are independent of crystal size. However, the Born model of ionic solids assumes that the ions interact via long-range electrostatic forces and short-range forces which can be described using simple analytical functions that need to be tested using, for example, the electronic structure calculations. The components of the short-range forces include both the repulsions and the van der Waals attractions between neighboring electron charge clouds.I6 The electronic polarizability of the ions is described by the shell model of Dick and Overhauser" in which each polarizable ion, in our case the oxygen ion, is represented by a heavy core and a massless shell, connected by a spring. The polarizability of the model ion is then determined by the spring constant and the charges-of the core and shell. These are usually obtained by fitting to experimental dielectric data when available. We employed the simulation code METADISE,I9 designed to model dislocations, interfaces, and surfaces, to calculate the surface energies of the two oxides. In this code, following the approach of Tasker,20the crystal comprises two regions which are periodic in two dimensions. In this study, region I contains 0 1995 American Chemical Society
de Leeuw et al.
17220 J. Phys. Chem., Vol. 99,No. 47,1995
the surface layer and a few layers immediately below, while region I1 represents the bulk of the crystal. The ions in region I are allowed to relax to their mechanical equilibrium, but those in region I1 are kept fixed at their bulk equilibrium positions. The computer code CHAO$' was used to calculate the energies of isolated point defects at the surfaces. Like METADISE, this code employs a two-region principle, region I now being a hemisphere centered on the surface defect with region I1 lying outside region I. This code uses the Mott-Littleton method for treating charged defects. In both codes regions I and II need to be sufficiently large for the energy to converge. The simulation of the hydroxylated surfaces was achieved by replacing a surface oxygen by two hydroxyl groups, i.e., in Kroger-Vink notation:
+
H 2 0 0;
- (OH); + (OH);
One difficulty is that there are many possible surface configurations that must be considered, particularly as the coverage is varied. The energies quoted in the later sections refer to the energetically most stable configuration obtained. The reliability of the energies obtained in this way depends on the potential model used. The long-range Coulombic interactions are calculated by the Parry technique22whereas the shortrange interactions are described in the form of a Buckingham potential:
Lr);
V(r)= A exp - - -
The forces between the atoms of the OH group are modeled by a Morse potential using fractional charges for the hydrogen and hydroxyl oxygen. As a result of these fractional charges, the oxygen-cation Buckingham potential needs to be modified for the interaction between cations and hydroxyl oxygen ions. For the pure crystals we have used the potential parameters derived empirically by Lewis and Cat10w.~~The potential employed for the interaction of the hydroxide ion with the rest of the crystal was developed by Leslie24and successfully used by Wright et al.25to study the hydrogarnet defect in grossular. The potential model parameters for the hydroxide ion itself were modified by Baram et a1.26to include a polarizable oxygen ion and applied in their study of hydroxide formation at the surfaces of quartz and zeolites. The potential parameters are given in Table 1. Thus, by using potential models which are derived independently of the hydroxylated surfaces, the results should be predictive, and this work will provide a framework for considering the interaction of water with other materials, e.g., titania. Results
Both MgO and CaO have a face-centered cubic structure where each cation and each oxygen atom is six coordinate. The crystals were cut to obtain four different surfaces, {loo},{ 1lo}, {lll}, and {310}, and their relaxed surface structures and energies were calculated. The surface energy is defined as the difference in energy between the surface and an equivalent number of bulk ions. The most stable surface has the lowest positive surface energy. Pure Surfaces. The pure { 100) surfaces are the most stable surfaces considered. As expected from previous calculat i o n ~ , ~they ~ - *have ~ the lowest surface energies and are the dominant surfaces in the crystal morphologies. Table 2 shows the calculated surface energies for the MgO and CaO surfaces considered in this paper.
TABLE 1: Potential Parameters Buckingham Potential (short-rangecutoff: 9.25 A) A (eV) P (A) c (evA6) 1090.40 1428.50 22764.00 311.97 777.27 941.50 22764.00 22764.00 311.97
0 0 27.88 0 0 0 13.94 6.97 0
0.34370 0.29453 0.14900 0.25000 0.34370 0.29453 0.14900 0.14900 0.25000
Morse Potential 7.0525
02-
core-shell interaction Mg core-shell interaction Ca 0 core charge 0 shell charge core-shell interaction 0 core charge
014-
a (A-1)
De (eV)
ion pair H04+-014-
3.1749
ro (A) 0.94285
54.80 eV A-2 47.96 eV A-2 1.oo -3.00 74.92038 eV +0.86902 -2.29502
+
0 shell charge
TABLE 2: Surface Energies of the Pure (Unhydroxylated) Surfaces of CaO and M e 0 ~~
~~
surface enerw/J m-2
surface (100) (110) (110) facet 1 (110}facet2 CaO
MgO
0.77 1.25
1.95 3.02
1.31 2.09
1.14 1.87
{ill} (310)
2.47 3.86
1.15 1.84
The (110) surface has a much larger surface energy than the {loo} surface (Table 2 ) . However, previous simulations have assumed that the surface formed by cleaving the crystal is flat (Figure la). We have found that the (110) surface microfacets into steps of {loo} planes, thereby reducing the surface energy drastically, tending toward the surface energy of the (100) surface. Parts b and c of figure 1 illustrate the microfaceting of the MgO { 1lo} surface. The faceted { 1lo} surface, especially facet 2 (Figure IC),can also be viewed as a collection of stepped {loo} planes. The fact that faceting of the { 1lo} surface produces the very stable {loo} surface with steps is probably the reason for the lowering of the surface energy. This result also illustrates the potential difficulty with static simulations, namely that unless every significant surface configuration is considered (including facets) it is not possible to be confident that the global minimum has been found. The atoms at the surface of the { 1lo} and those on the edges of the faceted (1 lo} surfaces are all four-coordinated. As such, these sites should be more amenable to chemical reactions than the five-coordinated surface atoms of the {loo} For the purpose of hydroxylating the { 11l} surface, we were interested in the termination with a surface plane of oxygen atoms. The unit cell of the cut required to create the (111) surface terminating at a full oxygen plane has a dipole, and as such the energy is divergent with crystal size.29 The surface was stabilized by shifting one-half of the oxygen atoms from the top to the bottom of the repeat unit, thus canceling the dipole.30 As shown in Table 2 the surface energy of the { 11l} surface thus created is very high, and as such we should not expect this surface to exist without further modification. Indeed, there is experimental evidence from LEED pattems8x3'that, in the absence of water, microfaceted { 11l} surfaces exist rather than the unstable polar { 11l} plane. The (310) surface is less stable than the dominant (100) surface, and not unlike the faceted (110) surface, it can be
Adsorption of Water on CaO and MgO Surfaces
J. Phys. Chem., Vol. 99,No. 47, 1995 17221 TABLE 3: Surface Energies of the Planar Hydroxylated Surfaces and Hydroxylated Microfacted { 110) Surfaces of CaO and MgO as a Function of Coverage (a) Planar Hydroxylated Surfaces surface energylJ m-2 surface CaO { 100) MgO { 100) CaO { 110) MgO { 110) CaO { 1 1 1)
MgO{lll) CaO (310) edge valley MgO { 3 10) edge valley
0% 0.77 1.25 1.95 3.02 2.47 3.86 1.15 1.15 1.84 1.84
12.5% 0.75 1.33 1.62 2.56" 2.14 3.42 1.05 1.10 1.74 1.79
25% 50% 75% 0.72 0.78 0.38 1.30 1.79 1.20 1.30 0.70 0.35 2.12" 1.26 1.24 1.83 1.05 0.39 3.02 1.82 0.91 0.96 0.76 0.61 1.04 0.98 0.89 1.63 1.42 1.23 1.67 1.58 1.75
100% 2.76 3.81 1.58 2.32 -0.28 0.04 0.45 0.80 1.03 1.48
(b) Hydroxylated Microfaceted { 110) Surfaces surface energylJ m-2 surface CaO { 110) facet 1 edge valley edge facet 2 edge valley edge MgO { 110) facet 1 edge valley edge facet 2 edge valley edge
+ valley + valley
11101 0
4
Oxygen
L [iio]
Magnesium
+ valley
Figure 1. Microfaceting of the MgO { 110) surface: (a) planar surface, (b) facet 1, faceting affecting the surface layer only, and (c) facet 2, faceting affecting the surface layer and two layers below, showing the exposed { 100) planes.
described as steps of { 100) planes, in effect mimicking edges consisting of low-coordinated atoms in a { 100) surface. Hydroxylated Surfaces. Every surface was hydroxylated in a series of partial coverages from infinite dilution to a full monolayer coverage, consisting of a dissociated water molecule adsorbed on every surface cation-oxygen pair. For every partial coverage four or five different configurations of adsorbed water molecules were tested, and the surface energies gathered in Table 3 are those of the most stable configurations. In each case the adsorption process was modeled by dissociating the water molecule such that the OH was bonded to a surface cation and the H atom to a surface oxygen atom. The hydration energies for the MgO and CaO surfaces were also calculated and are shown in Figure 2a-f. The calculation of the hydration energies required a value for the energy of dissociation of a water molecule:
+
2H20 O:gi
-
20Hi)
(3)
However, this requires the second electron affinity of oxygen?* which is material dependent. This can be overcome by using experimental heats of formation for the reaction (4)
Lattice energies and experimental enthalpies, given in Table 4, can then be used to calculate the dissociation energies for the reaction in eq 3. These were -617.5 kJ mol-' and -643.6
kJ mo1-l for CaO and MgO, respectively. As noted above, the discrepancy between the two values will be due to the difference in the second electron affinity but may also include uncertainties in the fit of calculated and experimental data.
+ valley
a
0% 1.95 1.31 1.31 1.31 1.14 1.14 1.14 3.02 2.09 2.09 2.09 1.87 1.87 1.87
25% 1.30 1.18 1.21
50% 0.70 1.08 1.10 1.00 1.05
2.12" 1.97 1.96
1.26 1.87 1.81 1.73 1.74
100% 1.58 1.16 0.99 0.39 0.99 0.93 0.78 2.32 1.88 1.59 1.1 1 1.68 1.57 1.39
Energy has not converged.
(100) Surface. The hydration energies of the partial coverages of the (100) surface show reasonable Langmuir behavior until full monolayer coverage, which was particularly unfavorable for both CaO and MgO. Other coverages show MgO to be slightly endothermic and CaO slightly exothermic. Examination of the favored surface structures showed that the surface configurations depend on the surface coverage. At low surface coverages of up to 12.5% in both MgO and CaO the water molecules prefer to adsorb as far apart as possible. However, at a coverage of 25% the dissociated water molecules begin to show a clear tendency to form pairs, giving rise to a surface cluster containing four OH groups. This trend is continued through to the 50% coverage. At 75% partial coverage, extensive surface rearrangement takes place (Figure 3) with cations from the surface layer relaxing upward by 1.34 in the case of MgO and 1.29 A in that of CaO. The hydroxyl groups adsorbed on the cations tilt by 15" and 20.3" for MgO and CaO, respectively, relative to the surface normal, forming a second bond to the cations which have moved upward. As a result this surface is the most stable configuration as is shown from the surface energies in Table 3a, more stable than even the pure unhydroxylated surface; cf. the surface energy of pure CaO surface of 0.77 J m-2 with 0.38 J m-2 for the 75% covered surface. In MgO the surface energies of the pure and 75% partially covered surfaces are comparable. This result of maximum stability at 75% coverage is not dissimilar to that of Oliver et al.?O who found in their study of the effect of Ni3+ holes on the morphology of NiO, which also has the rock salt structure, that a minimum surface energy occurred at a Ni3+ hole coverage of 75%. As expected, in both CaO and MgO the hydration energies are also at their most negative value for the 75% coverage. In
de Leeuw et al.
17222 J. Phys. Chem., Vol. 99, No. 47, 1995 D
0
20
60
40
80
Coverage (%)
+MgO edge
100
0 2 0 4 0 6 0 8 0 1 0 0 Coverage (%)
* MgO valley
-120
4 -
0 2 0 4 0 6 0 8 0 1 0 0 Coverage(%)
0
20
40
60
100
80
Coverage (%)
I
I _
A
-320
2 %
:-360 ' 03
E
3 0
20
40
60
80
Coverage (%)
100
0 2 0 4 0 6 0 8 0 1 0 0 Coverage (%)
Figure 2. Energies of hydration of the hydroxylated surfaces as a function of coverage for (a) { 100) surface, (b) planar { 1 10) surface, (C) faceted { 110) surface; facet 1, (d) faceted { 110) surface; facet 2, (e) { 1 1 1) surface, and (f) (310) surface.
TABLE 4: Lattice Energies and Enthalpies of Formation
CaO(S, Ca(OH)2(s, MgO(s, Mg(OH)2(s,
H20w
lattice energym mol-'
enthalpy of fonnationkl mol-'
-3468.46 -29 16.25 -3984.80 -3378.26
-635.10 -986.10 -60 1.70 -924.50 -285.83
the case of MgO it is only at 75% coverage that hydroxylation of the surface is energetically favorable. { 110) and Faceted { 110) Surfaces. For all partial coverages of the planar { 110) surface there is a strong tendency for the water molecules to adsorb onto the surface in pairs. Like the { 100) surface above, this may be due to attraction between the adsorbed hydrogen atoms and hydroxyl oxygen atoms. When in addition the adsorbed hydroxyl groups and the adsorbed hydrogen atoms are able to tilt toward each other, the surface energy is lowered even further. This is shown in the case of MgO by the partial coverage of 50% where the hydroxyl group tilts toward the hydrogen atom by 22.7" and the hydrogen toward the hydroxyls by 28.3'. The resultant surface energy is about two-thirds of the surface energy of the next most stable configuration. CaO at all coverages and MgO at coverages from about 75% behave in a different way. Here the oxygen of the hydroxyl group on relaxation moves toward the site where a surface oxygen would be were the crystal one layer thicker, forming bonds with two cationsin the surface layer. The adsorbed hydrogen tilts toward an empty cation site in the next layer.
Hydrogen
Q
Oxygen
e
Magnesium
[I001
A
i[ O l l ]
Figure 3. Relaxed surface configuration for the 75% hydroxylated MgO 1100) surface, showing tilted hydroxides and raised surface magnesium ions.
Figure 4 illustrates the configuration. In effect, the adsorbed water molecules form a microfacet on the planar { 110) surface.
J. Phys. Chem., Vol. 99, No. 47, 1995 17223
Adsorption of Water on CaO and MgO Surfaces
09
Q Q
0
Hydrogen
0
Oxygen
0
Magnesium
bi,
P101
A
L[iiol
Figure 4. Minimum energy configuration for the 75% hydroxylated MgO planar { 110) surface, illustrating bridging hydroxide groups and alternating hydrogens.
As discussed above, microfaceting of the { 110) plane stabilizes the surface, and it is therefore not surprising that microfaceting of the hydroxylated surfaces also lowers the surface energy compared to that of the flat surface. In the case of the partial coverages of the faceted (110) surfaces, only adsorption on the edges or in the valleys was considered. The partial coverages are relative to these sites only. Adsorption on the edges is comparable to adsorption at a lowcoordinated site on a { 100) plane, for instance, on the edge of a step defect. The position in the valleys can be compared to adsorption of water molecules inside the step of a step defect on a { 100) plane. Figure 5 shows the two different positions. We separately adsorbed water molecules either on the edges or in the valleys, and we did not consider the adsorption sites on the (100) planes exposed by the faceting process. We also looked at a surface where all sites both on the edges and in the valleys were covered by hydroxyl groups. This configuration was found to have a particularly low surface energy (Table 3b). When adsorbing on the edges of the faceted surface, the hydroxyl group and proton do not tilt toward each other when the edges are fully covered, unlike at lower coverages where the tilt angle for CaO, for example, is 26.4" for the hydroxyl group and 29.4" for the hydrogen atom. In the valleys the hydroxyl groups can never tilt, as the oxygen atom from the dissociated water molecule is placed in an oxygen site of an unfaceted surface and is therefore coordinated to two cations instead of one, as in adsorption on the edge. The proton from the dissociated water molecule is bonded to a surface oxygen atom on the (100) plane inside the valley. Langel et a1.: in their Car-Parinello simulation of water adsorption on a microfaceted { 1 IO} surface of MgO, calculated that an initially undissociated water molecule placed inside the step, with the oxygen in the regular lattice position of an unfaceted surface, would dissociate, and the proton would indeed form a bond with a surface oxygen on the { 100) plane. Simulations of hydroxylation of both facet 1 and facet 2 of CaO reveal that at low coverages of 0-50% of the adsorption sites, the configuration with water molecules adsorbed on the edges is more stable than adsorption in the valleys of the facet. However, at the higher surface coverage of 100% of the edge or valley sites, water adsorbs preferentially in the valleys.
3 0
Hydrogen Oxygen Calcium
t
[iio]
Figure 5. Minimum energy configuration for the hydroxylated CaO { 110) facet 2: (a) adsorption on all edge sites giving rise to linear hydroxide groups, and (b) adsorption on all valley sites giving rise to bridging hydroxide groups and alternating hydrogens.
The simulations on facet 1 of MgO show that the configurations with water molecules adsorbed in the valleys are more stable and the hydration energies lower for all partial coverages than the configurations with water adsorbed on the edges. However, at infinite dilution the hydration energy of the edge configuration is somewhat lower than the valley configuration (Figure 2c). This would indicate that at low partial water pressure, water molecules initially adsorb on the low-coordinated edge atoms, rather than in the valley. Hydration in the valley then becomes more attractive with increased coverage. Facet 2 of MgO resembles the faceted surfaces of CaO: hydration on the edges is more favorable at low partial coverage while at higher coverages the adsorption in the valleys is preferred. Thus it seems that on the faceted { 110) surfaces water adsorbs at low-coordinated sites only at low partial coverages whereas at higher coverages the position in the valley, where the hydroxyl oxygen is highly coordinated, is preferred. { 111) Surface. On the { 111) surface the oxygen atoms of the OH groups of the added water molecules were adsorbed in the oxygen vacancies left when half the oxygen atoms of the surface are moved to remove the dipole as described earlier. The hydrogen atoms of the water molecules were bonded to the remaining surface oxygen atoms. A full monolayer coverage would then result in a smooth surface plane of hydrogen atoms with a full oxygen layer below. Not suprisingly, this is a very stable surface for both CaO and MgO and is structurally similar to the basal plane of Ca(OH)2 and Mg(OH)2. In CaO the water molecules tend to adsorb in pairs again as on the { 100) and planar { 110) surface. At full monolayer
de Leeuw et al.
17224 J. Phys. Chem., Vol. 99, No. 47, 1995
coverage, the surface energy becomes negative (Table 3a), indicating that the crystal is not stable and would energetically prefer to form the hydroxide at the { 111) surfaces. In MgO, instead of adsorbing in pairs the water molecules prefer to adsorb in rows. This is possibly due to the smaller cation-oxygen distance in MgO, which d u c e s the requirement of the hydroxyl groups to tilt to the hydrogen atom, and all the adsorbed hydrogen and oxygen atoms stabilize each other. The hydration energies for the various partial coverages of the { 111) surface in both CaO and MgO are highly exothermic compared to those of the other surfaces, and the { 111) surface clearly prefers to be hydroxylated. (310) Surface. As noted earlier, the { 3 10) surface represents a simple step structure and allows us to model adsorption both on the edge and inside the step of the { 100) plane. The steps are 15.2 and 13.3 8, apart in CaO and MgO, respectively, sufficiently well separated that we do not anticipate significant interactions between steps. We found that the adsorbed oxygen atom of the OH group upon relaxation always takes up the same position, essentially adding to the step and coordinating to two cations, one cation next to the step on the { 100) plane and the other at a low-coordination site on the edge of the step. There are, however, two configurationsfor the proton from the water molecule which are similar in energy, certainly at low coverage. These are on the top of the step, Le., the “edge” site or inside the step, i.e., the “valley” site. When the hydrogen atom is adsorbed at a surface oxygen on the { 100) plane inside the step, Le., at the valley site, it relaxes toward the position where a cation would have been had the step been extended. In effect, the adsorbed water molecules seem to build on to the step. When adsorbed on the edge, i.e., at the top of the step, however, the hydrogen atom relaxes toward the oxygen atom of the adsorbed hydroxyl group. Figure 6 shows the different configurations. At all coverages the surface energy of the edge configuration is lower than that of the valley configuration although they are quite similar (Table 3a). In the valley configuration at low coverages, up to 25%, the water molecules tend to adsorb separately at isolated positions. At higher coverages, however, the water molecules show a definite tendency to adsorb in full rows rather than spaced equally over the surface, resulting in alternating fully hydroxylated and empty edges. In this way, the adsorbed molecules form a smooth edge rather than kinks and corners. The same trend is shown in both CaO and MgO. Contrary to the valley configuration, in the position on the step the water molecules prefer not to adsorb in full rows. They tend to adsorb in a straight line from step to step rather than equally spaced over the surface. The hydration energy of adsorption on the edge is more favorable than that in the valley for both CaO and MgO (Figure 20, and thus water should preferentially adsorb on the four-coordinated atoms on the edges as inferred by rather than on the five-coordinated { 100) atoms inside the step. I
Discussion The lowest hydration energies of the most stable of the CaO surfaces studied range from -36.5 kJ mol-’ on the (100) surface to -178.5 kJ mol-’ on the edges of the (310) surface. Fubini et al.15 in their experimental study of the reactivity of CaO with water vapor found an initial heat of adsorption of approximately -170 kJ mol-’ at very low coverage, decreasing to about -145 kJ mol-’ at higher coverages. In their study, no particular surface was prepared or studied, and we can expect
.\
A
n 0
0
Hydrogen
0
Oxygen Magnesium
Figure 6. Minimum energy configuration of the MgO { 310) surface: (a) unhydroxylated surface, (b) adsorption on the edge giving rise to bridging hydroxide groups with the edge hydrogens directed between the hydroxide groups, and (c) adsorption in the valley, again giving rise to bridging hydroxide groups with the valley hydrogens directed between the hydroxide groups.
the hydration energy to reflect the properties of the most stable { 100) surface with defect features such as steps. The edges of the { 3 10) surface are our most accurate model for four-coordinated sites of chemisorption on a { 100) plane. Our calculated hydration energies for this surface agree very well with those obtained experimentally by Fubini et al.I5 At infinite dilution the hydration energy is -178.5 kJ mol-’, decreasing to an average of approximately -160 kJ mol-’ from about 12.5% coverage. This indicates that the real experimental { 100) surface containing defects is, as expected, more accurately modeled, both qualitatively and quantitatively, by the { 310) plane than by the perfect { 100) plane. In MgO the hydration energies range from -3.9 kJ mol-’ on the { 100) surface to -175.9 kJ mol-’ on the edges of the { 3 10) surface. The hydration follows a very similar pattern to that of CaO although hydration of the dominant { 100) surface in MgO is not energetically favorable, in agreement with theoretical work by Scamehorn et a1.l0 Beruto et al.,’ when investigating water chemisorption on MgO at elevated temperatures under a vapor pressure of 613 Pa, found an enthalpy of adsorption of water between -189 kJ mol-’ at a partial coverage of 25% and - 113 kJ mol-’ at a partial coverage of about 70%, assuming that only the (100) surfaces were present and that chemisorptiontook place on fixed sites. If we assume that those fixed sites are the low-coordinated surface atoms, including fourcoordinated atoms which are comparable to the atoms on the { 110) surface and (3 10) edges, rather than the five-coordinated
Adsorption of Water on CaO and MgO Surfaces
J. Phys. Chem., Vol. 99, No. 47, 1995 17225
atoms of the perfect (100) plane, then these values again agree quite well with the hydration energy calculated for the (310) surface, which is -176 W mol-' at very low coverage and decreases to an average value of - 139 W mol-' from a coverage of about 25%. As in CaO, the highly stepped (310) surface of MgO provides a better model for the { 100) surface of a real defective rock salt oxide. In addition to c o n f i i n g the sites of chemisorption, we found in our simulation that surfaces which in their planar forms have high surface energies are often stabilized by microfaceting into steps of the dominant (100) plane. LEED pattems show that microfaceting of the ( 110) surface indeed O C C U ~ S . ~ Moreover, ' when hydrated, the initially planar ( 110) surface shows facets created by the adsorbed hydroxyl groups (Figure 4).
aqueous conditions. In the near future we aim to have infrared modeling capabilities which would allow further comparison with experiment. Another area for further research is in molecular dynamics simulations of the faceting of the { 110) and ( 111] surfaces. A microfaceted ( 111) surface contains three-coordinated comer atoms and may be even more amenable to hydration than the four-coordinated edge atoms studied in this paper and could be used as a model for kink sites on the (100) surface.
Conclusion
References and Notes
We have used atomistic simulation to model dissociative adsorption of water on the {loo), planar and microfaceted (110), ( l l l ) , and (310) surfaces of CaO and MgO. The calculations are in accord with the available experimental data, Le., they show that the low-coordinate sites represent the most energetically favorable sites for chemisorbed water. Specific predictions include the following: (1) The pure MgO { 100) surface is generally not amenable to dissociative adsorption but the pure CaO (100) surface is. (2) There is a particularly stable surface reconstruction at a coverage of 75% on the (100) surface, particularly for CaO, that would be worthy of future investigation. (3) The microfaceted surfaces are more stable than the planar surface, suggesting that these will be the surfaces observed experimentally. (4)Microfaceted surfaces energetically favor chemisorption at the low-coordinated edge positions at low partial coverages for both MgO and CaO, but at higher coverages the sites in the valley are preferred. ( 5 ) Chemisorption at the pure { 111) surfaces of both MgO and CaO is very favorable, in agreement with Onishi et except that at high coverages the crystal becomes thermodynamically unstable and would rather form the ( 111) hydroxylated surface. (6) On the (310) surfaces, again two possible sites for chemisorption were compared, on the edge of the step and inside the step (valley site). The hydration energies of the fourcoordinated edge sites are more exothermic than their counterparts at the five-coordinated sites inside the step, and dissociative adsorption of water will therefore preferentially take place at the low-coordinated atoms of the edge on top of the step. From our study we can make some general predictions about the mode of dissociative adsorption, namely that there is a tendency at low coverages for the dissociativelyadsorbed water to form dimers. Furthermore, in CaO the OH groups often tilt so as to form stronger hydrogen bonds on the surface. This does not occur for MgO, presumably due to a smaller lattice spacing in MgO. After the success in modeling the hydrated surfaces, the next step is to study the physisorption of water molecules at the different crystal surfaces and hence model the surface under
Acknowledgment. We thank EPSRC for the provision of computer time. N.H.dL. and S.C.P. thank EPSRC for the EPSRC quota and Biosym Technologies Inc. for the use of Insight 11.
(1) Wu, M.-C.; Estrada, C. A,; Comeille, J. L.; Goodman, D. W. J. Chem. Phys. 1992, 96 ( 9 , 3892. (2) Beruto, D.; Searcy, A. W.; Botter, R.; Giordani, M. J . Phys. Chem. 1993, 97, 9201. (3) Langel, W.; Parrinello, M. Phys. Rev. Lett. 1994, 73 (3). 504. (4) Dunski, H.; Jozwiak, W. K.; Sugier, H. J . Catal. 1994, 146, 166. (5) Duriez, C.; Chapon, C.; Henry, C. R.; Richard, J. M. Surf. Sei. 1990, 230, 123. (6) Jones, C. F.; Reeve, R. A.; Rigg, R.; Segall, R. L.; Smart, R. St. C.; Tumer, P. S. J . Chem. Soc., Faraday Trans. I 1984, 80, 2609. (7) Coluccia, S.;Barton, A,; Tench, A. J. J . Chem. Soc., Faraday Trans. I 1981, 77, 2203. (8) Onishi, H.; Egawa, C.; Aruga, T.; Iwasawa, Y. Surf. Sei. 1987, 191, 479. (9) Coluccia, S.; Tench, A. J.; Segall, R. L. J . Chem. Soc., Faraday Trans. I 1979, 75, 1769. (10) Scamehom. C. A.: Hess. A. C.: McCarthv. M. I. J . Chem. Phvs. 1993,99, 2786. (1 1) Scamehom, C. A.; Harrison, N. M.; McCarthy, M. I. J . Chem. Phys. 1994, 101 (2). 1547. (12) Ito, T.; Tashireo, T.; Kawasaki, M.; Watanabe, T.; Toi, K.; Kobayashi, H. J. Phys. Chem. 1991, 95, 4476. (13) Kobayashi, H.; Yamaguchi, M.; Ito, T. J . Phys. Chem. 1990, 94, 7206. -~~
(14) Colboum, E. A.; Mackrodt, W. C. Surf. Sei. 1982, 117, 571. (15) Fubini, B.; Bolis, V.; Bailes, M.; Stone, F. Solid State Ionics 1989, 32/33, 258. (16) Bom, M.; Huang, K. Dynamic Theorj of Crystal Lattice; Clarendon: Oxford, 1954. (17) Dick, B. G.; Overhauser, A. W. Phys. Rev. 1958, 112, 90. (18) Colboum, E. A. Surf. Sei. Rep. 1992, 15, 281. (19) Watson, G. W.; Kelsey, E. T.; de Leeuw, N. H.; Harris, D. J.; Parker, S. C. J . Chem. Soc., Faraday Trans. I , in press. (20) Tasker, P. W. Philos. Mag. 1979, 39A, 119. (21) Duffy, D. M.; Tasker, P. W. Philos. Mag. 1984, 50A. 143. (22) P q ,D. E. S u f f Sei. 1975,49,433. Parry, D. E. S u ~Sei. . 1976, 54, 195. (23) Lewis, G. V.; Catlow, C. R. A. J . Phys. C: Solid State Phys. 1985, 18, 1149. (24) Leslie, M. SERC Daresbury Laboratory Report. DUSCI-TM3 1T; 1981. (25) Wright, K.; Freer, R.; Catlow, C. R. A. Phys. Chem. Miner. 1994, 20, 500. (26) Baram, P. S.; Parker, S. C. Philos. Mag. A, in press. (27) Tasker, P. W.; Colbourn, E. A,; Mackrodt, W. C. J . Am. Ceram. Soc. 1985, 68, 74. (28) Tasker, P. W.; Duffy, D. M. Surf. Sei. 1984, 137, 91. (29) Tasker, P. W. J . Phys. C: Solid State Phys. 1979, 12, 4977. (30) Oliver, P. M.; Parker, S. C.; Mackrodt, W. C. Modell. Simul. Mater. Sei. Eng. 1993, 1, 755. (31) Henrich, V. E. Surf. Sei. 1976, 57, 385. (32) Harding, J. H.; Pyper, N. C. Philos. M a g . Lett. 1995,72 (2), 113.
JP95 1479Y
