Hazar Guesmia,
Robert Grybos*b,
Jarosław Handzlikc and
Frederik Tielens*d
aCNRS-UM2-ENSCM-UM1, UMR 5253, Institut Charles Gerhardt Montpellier, Ecole Nationale Supérieure de Chimie de Montpellier, 8 rue de l'Ecole Normale, 34296 Montpellier, France
bJerzy Haber Institute of Catalysis and Surface Chemistry, Niezapominajek 8, Kraków, Poland
cFaculty of Chemical Engineering and Technology, Cracow University of Technology, ul. Warszawska 24, 31-155 Kraków, Poland
dSorbonne Université, UPMC Univ Paris 06, UMR 7574, Laboratoire Chimie de la Matière Condensée, Collège de France, 11 place Marcelin Berthelot, 75231 Paris Cedex 05, France. E-mail: frederik.tielens@upmc.fr
First published on 13th April 2016
A DFT based characterization of tungsten oxide supported on amorphous hydroxylated silica is presented. The different molecular organizations are investigated on the surface topology and tungsten oxygen coordination. The presence of mono- and di-grafted species is discussed and rationalized, using an atomistic thermodynamic approach. The presence of WO groups are preferred over W–OH groups and the grafting coordination is dominated by the degree of hydration of the silica surface. At room temperature di-oxo digrafted and mono-oxo-tetragrafted species are in competition regulated by the ambient degree of hydration which also affects the silanol density of the silica support. A comparison between tungsten and the other group VI elements confirms a greater chemical difference with Cr than with Mo.
Tungsten oxide supported on silica is mainly used as an industrial catalyst for olefin metathesis.4,5 Originally, it was used to transform propene into ethene and butene in the triolefin Phillips process.6 At present, since the world demand for propene is rapidly increasing,7 tungsten oxide catalyzes a reverse reaction. Other uses include selective oxidation of, among others, methane,8–10 styrene,11 propylene12 and methanol.13 It is also used for photocatalytic water splitting.14
The structure and character of WOx/silica catalyst has been studied by a wide range of experimental techniques.3,15,16 Also, models of isolated oxo-tungsten species were synthesized and characterized.17 A combined DFT-NMR study found that the terminal W–OH groups are weakly acidic and very stable.18 On high surface area catalysts, a well-dispersed and reduction-resistant layer of tungsten oxide covers the active sites of silica surface.8
The catalytic activity is generally attributed to the well-dispersed surface forms of tungsten oxide rather than the crystalline WO3 phase.11,13,14,19–25 At low coverage, the surface sites are usually found to be isolated tungsten monomers, either mono-oxo3,26 or di-oxo,3,26 or dimers27 and oligomers,28 at least under dehydrated conditions. Discrimination between different types of monomers is not trivial, however. Vibrational spectrum obtained by in situ Raman spectroscopy shows dominant band in the range 975–991 cm−1 associated with WO stretching vibration.3,15,26,28–30
It was at first ascribed to the mono-oxo form due to lack of the asymmetric component, characteristic for the di-oxo form.15,29 However, when a weak signal at 968 cm−1 was discovered, it was identified as originating from the asymmetric stretching vibration of the di-oxo monomer and the dominant band was re-interpreted as a proof that the tungsten oxide monomers on the silica surface are mainly di-oxo.3,26 Existence of the mono-oxo form was tied to another band at 1014–1015 cm−1.3,26,28
In one case the same band was interpreted as originating from polymeric WO5/WO6 forms.30 Co-existence of mono-oxo and di-oxo forms on the surface was confirmed by 18O–16O isotope exchange experiments.3 Comparison of Raman peak intensities show that the mono-oxo/di-oxo ratio depends on temperature and at higher temperatures the di-oxo form dominates.26
To the best of our knowledge, no theoretical ab initio studies on realistic tungsten oxide forms on amorphous silica support are available. To fill this gap, we have performed an extensive density functional theory (DFT) study to find the most stable structures of isolated monomeric tungsten oxide species and to characterize them by vibrational analysis.
Vibrational spectra have been calculated for selected surface species within the harmonic approximation. Only the tungsten center and its first and second neighbors (O–Si and OH groups) are considered in the Hessian matrix. This matrix is computed by the finite difference method followed by a diagonalization procedure. The eigenvalues of the resulting matrix lead to the frequency values. The assignment of the vibrational modes is done by inspection of the corresponding eigenvectors. A scaling factor of 0.9659 was used according to Halls et al.41
![]() | ||
Fig. 1 Side view of the unit cell of the amorphous silica surface model on which di-oxo, W(VI) oxide cluster is grafted (W atom in blue). |
The supercell dimensions are 12.77 Å × 17.64 Å × 25.17 Å and it contains 27 silicon atoms. Without water, the formula can be written as pure silicon oxide – (SiO2)27. In the presence of water, the surface of silicon oxide becomes hydroxylated. Silanol OH groups are formed through hydrolysis of the siloxane Si–O–Si bridges according to the reaction:
–(Si–O–Si)– + H2O → –Si–OH, HO–Si– | (1) |
Therefore the hydroxylated surface of amorphous silica can by represented by a formula (SiO2)x(H2O)y, where two surface silanol groups are formed for each H2O molecule. Specifically, the supercell of our model has an overall formula Si27O67H26 which can be written as (SiO2)27(H2O)13.
Thus, it is empirically shown that a high temperature and dehydration conditions are necessary to obtain the multi-grafted tungsten oxide species.
The catalyst precursor is modeled by a WO2(OH)2 molecule. One such species is added to the silica unit cell resulting in a coverage of 0.44 monomers per nm2, a typical coverage found in working catalysts.50
Grafting of WO2(OH)2 molecule can result in several different structures with different number of: (i) tungstenyl WO oxygen atoms, (ii) –OH hydroxyl groups and (iii) W–O–Si bridges to the surface. Theoretically up to four silanols may be involved in the reaction yielding different modes of grafting: mono, di, tri and tetra. Structures involving different silanol types: isolated (Si–OH), vicinal (HO–Si–O–Si–OH), geminate (HO–Si–OH) and non-vicinal (two Si–OH groups not directly connected) on the surface were considered. In every case, however, the oxidation state of tungsten remains +VI. Creation of the W–O–Si link can be represented by a following simplified reaction:
–Si–OH, HO–Si– + W → –Si–OH, W–O–Si– + H | (2) |
Thermodynamic analysis, described below, is used to explore the stability of various monomer surface forms under varying conditions. However, in order to probe and compare the grafting sites present on the amorphous silica surface, a complete and systematic series of tungsten monomer models are investigated (see Fig. 2), in analogy to chromium and molybdenum.33–35
![]() | ||
Fig. 2 Different geometries as a function of its hydration state for the supported W oxide grafted on amorphous silica. |
In order to establish the stable tungsten surface monomer under different conditions an atomistic thermodynamic analysis has to be performed. Possible forms of mono- (A, B, C), di- (D, E, F), tri- (G, H) and tetra-grafted (I) W(VI) monomer on the surface are presented in Fig. 2. To take into account deviations in surface composition and the presence of gas phase, one introduces appropriate chemical potentials to calculate an approximation of the Gibbs free-surface energy. Assuming that the surface is in thermodynamic equilibrium with the gas phases, the chemical potentials are related to a given temperature T and pressure p. This procedure enables to extend the 0 K and zero pressure DFT results to experimentally relevant environments, thereby bridging the gap between ultra-high vacuum like conditions, and temperatures and gas phase pressures that are applied in realistic catalytic conditions. The grafting process can be described as a reaction between the surface and the precursor, with consumption or liberation of water. Appropriate equations for each model is given below (eqn (3)):
![]() | (3) |
The W(VI)/silica system is considered to be in contact with a gaseous water reservoir. Starting from a calculated electronic energy, the free energy of water/W(VI)/silica interface under known thermodynamic conditions may be estimated following the approximations used by Digne et al.,52 as originating from Kaxiras et al.53 and Qian et al.54 Main assumptions are that there is no variation of the chemical potential of the surface upon adsorption and that the gas phase as a perfect gas. In the proposed scheme, the free energy of water (including the ZPE correction) in the gas phase is:
ΔG(H2O) = E(H2O) + ((ΔHG − TΔSG(T)) + RT![]() | (4) |
Using the above mentioned formalism, the approximated free energy of the grafting reactions for the formation of the mono-, di-, tri and tetra-grafted W(VI) monomers at equilibrium conditions, can be expressed as (eqn (5)):
![]() | (5) |
Only the ΔG of water depends on T and p, therefore energies of models B and D will not depend on the temperature. In principle, increasing temperature will tend to remove water from the monomer and at high temperatures model I should be most stable. Note, however, that it requires four links to the surface, i.e. four hydroxyl groups in a specific arrangement. Such sites are much less common on the surface than the ones for mono- or bi-grafted monomers. Therefore, in the real system, the amount of tetra-grafted monomers will be less than predicted from thermodynamic analysis.
In this approach, we consider that the energies of the different types of grafting transitions are independent of the degree of hydration of the silica surface. It is known experimentally that silanols are stable at silica surfaces until 673 K. Above this temperature, silanols begin to condensate into siloxane bridges.56 Thus, our model with 5.8 OH nm−2 corresponding to conditions of a hydroxylated surface, remains valid until the temperature of 673 K.
In Table 1 we present grafting energies for all possible grafting possibilities (germinal (g), vicinal (v), and non-vicinal (n) on the silica slab considered for structure F) – negative values indicate thermodynamically stable structures. The different grafting sites are shown in Fig. 3 in ref. 35. Even for thermodynamically unstable structures, the monomer stays at the surface, due to a kinetic barrier. When two water molecules are available, monomers can detach from the surface as WO2(OH)2 and leave two silanol groups behind. When only one H2O is present, the monomer leaves as WO3. If we allow for the possibility of creating Si–O–Si siloxane bridges, then two more situations can be envisioned: (i) with one water present the monomer leaves as WO2(OH)2, (ii) with no water, the monomer leaves as WO3.
WO2(OH)2 | WO3 | |
---|---|---|
Geminal | ||
g1 | 1.79 | −1.65 |
g2 | 1.72 | −1.72 |
g3 | 1.50 | −1.93 |
g4 | 1.79 | −1.63 |
![]() |
||
Vicinal | ||
v1 | 0.64 | −2.81 |
v2 | 0.41 | −3.04 |
v3 | 0.71 | −2.73 |
v4 | −0.03 | −3.50 |
v5 | 0.38 | −3.07 |
v6 | 0.49 | −2.97 |
v7 | 0.22 | −3.23 |
v8 | 0.85 | −2.61 |
v9 | 0.31 | −3.18 |
v10 | 0.38 | −3.07 |
![]() |
||
Non-vicinal | ||
n1 | −0.25 | −3.72 |
n2 | −0.11 | −3.58 |
n3 | −0.12 | −3.58 |
n4 | −0.05 | −3.51 |
n5 | 0.05 | −3.41 |
n6 | 0.26 | −3.20 |
n7 | −0.02 | −3.48 |
n8 | 0.32 | −3.13 |
n9 | 0.00 | −3.45 |
n10 | −0.14 | −3.60 |
n11 | 0.03 | −3.44 |
n12 | 0.09 | −3.38 |
n13 | −0.20 | −3.66 |
For anchoring, geminate sites are energetically disfavored due to their high rigidity. In abundance of water, the monomers are barely stable at the surface. The adsorption energy is between 1.50 and 1.79 eV indicating an unstable structure. Vicinal sites are more flexible and can accommodate a W monomer with adsorption energies between −0.03 and 0.85 eV. Monomers are more stable on the flexible non-vicinal sites – all adsorption energies are below 0.32 eV. Few structures show stable adsorption, although only barely stable as the best adsorption energy is only −0.25 eV. It seems that structures with more hydrogen bonds are more stable, but no clear correlation could be found.
As noted above, without large excess of water vapor available, anchored monomers can only detach from the grafting sites as WO3 species. This reaction still requires one water molecule to recreate two surface hydroxyl groups. The monomer stability with respect to gaseous WO3 in the presence of traces of water is very high – energies required to remove a monomer from the surface range from 1.63 to over 3.72 eV. Without water, WO3 can still be removed if the anchoring Si centers are close enough to create a siloxane bridge. However, this reaction pathway is expected to require even more energy.
The geometry of the monomer does not vary much between different grafting sites (see Table 2). The tungstenyl WO bonds have lengths between 1.71 and 1.74 Å, depending on the number of hydrogen bonds between surface and the monomer. If no such hydrogen bonds are present, both W
O bonds are 1.74 Å. The formation of hydrogen bonds with one tungstenyl oxygen atom has a weaker effect on the bond length than the coordination of the W atom, which is in line with the idea that the W
O dipole moment is smaller than for the two other group VI elements. This decrease in iconicity was already observed for the group V elements.55,57 The angle between W
O bonds is usually 108–109° with some deviations induced by hydrogen bonds.
Model | d(W–OH) | d(W–OSi) | d(W![]() |
A(OWO) |
---|---|---|---|---|
A | 1.914 | 1.921 | ||
1.928 | ||||
1.928 | ||||
1.937 | ||||
1.99 | ||||
B | 1.899 | 1.961 | 1.751 | |
1.908 | ||||
1.917 | ||||
C | 1.907 | 1.882 | 1.734 | 108.23 |
1.739 | ||||
D | 1.889 | 1.934 | ||
1.903 | 1.942 | |||
1.961 | ||||
1.965 | ||||
E | 1.943 | 1.933 | 1.722 | |
1.912 | 1.912 | |||
F | 1.890 | 1.737 | ||
1.893 | 1.738 | 108.19 | ||
G | 1.849 | 1.938 | ||
1.934 | 1.938 | |||
1.946 | 2.030 | |||
H | 1.915 | 1.890 | 1.733 | |
1.904 | ||||
2.052 | ||||
I | 1.916 | 1.715 | ||
1.929 | ||||
1.954 | ||||
1.973 |
The lengths of W–OSi bonds anchoring the monomer to the surface range from 1.84 to 2.05 Å, but for the di-oxo digrafted situation a W–OSi bond distance of 1.89 Å is calculated. The angle between those bonds vary between 128° and 137° (up to 155° in the B-structure). The W coordination is 4 in structures C and F (see OWO angles in Table 2), 5 in structures B, D, E, H and I, and 6 in structures A and G.
Grafting of tungsten oxide species on amorphous silica surface has a relatively small effect on the silica framework, and is comparable with what has been found in our previous study on the grafting of oxides on silica.31–35
The hydrogen bond network, on the other hand is heavily affected by grafting. Depending on their local density, silanols on a clean surface interact with their neighbors forming an H-bond network. The grafting process perturbs the local H-bond network in two ways: (i) surface hydroxyl groups are removed upon grafting; (ii) the W oxide units might also form hydrogen bonds with the silica support. In the models studied, the W–OH groups bind to surface silanols stabilizing the structure while the WO groups do not form hydrogen bonds.
Considering the reaction energy ΔEreact of Table 3 calculated according to the eqn (5) for the best grafting modes as a function of hydration rate. For W oxide values of −2.71, −2.46, −2.35, −1.49 and −1.56 eV for +1, 0, −1, −2, and −3 water molecules are obtained, corresponding to the models A, D, C, F, and I, respectively. All reaction energies are exothermic for grafting WO2(OH)2, indicating that the adsorption of the MO2(OH)2 with M = Cr, Mo, or W, is favored for W compared with Mo and Cr, with respect to the initial situation (hydroxylated silica and H2MO4 in the gas phase). Interesting to note is that following this reaction scheme the MO2(OH)2 transition metal precursor reacts more favorably when going down the column of the group VI elements (Cr to W) in the periodic Table.
Modela | ΔEreact | ΔEreact35 | ΔEreact34 |
---|---|---|---|
Group VI metal | W | Mo | Cr |
a See Fig. 2. | |||
A: surface + MO4H2 + 1H2O | −2.71 | −0.49 | −0.26 |
B: surface + MO4H2 | −2.18 | −0.26 | −0.89 |
C: surface + MO4H2–1H2O | −2.35 | −0.76 | −2.09 |
D: surface + MO4H2 | −2.46 | −0.18 | 0.13 |
E: surface + MO4H2–1H2O | −2.34 | −0.39 | −0.80 |
F: surface + MO4H2–2H2O | −1.49 | 0.02 | −1.33 |
G: surface + MO4H2–1H2O | −1.26 | 1.04 | 1.72 |
H: surface + MO4H2–2H2O | −0.85 | 1.06 | 0.34 |
I: surface + MO4H2–3H2O | −1.56 | 0.38 | 0.42 |
Another point which is revealed by this reaction energy analysis, is that the most favorable models associated to the different degrees of hydration are different between Cr and Mo on one hand, and W on the other. For Cr and Mo model B is favored against model D for W, the other structures' relative reaction energy differences do not change the overall stability trend between the models. This might be interpreted by the higher coordination chemistry for W compared with Mo and Cr (see Fig. 2).
Nevertheless, it should be noted that the results in Table 3 report electronic energies only, which are identical to the free energy at 0 K. Under given temperature T and pressure p, the contributions of entropy and chemical potentials have to be taken into account in the free energies.
Fig. 3 shows the surface free energy Γ, defined as the free energy per surface area (the reaction free energy of grafting that was approximated divided by the surface of our slab, or multiplied by surface coverage of W), of the mono-, di-, tri- and tetra-grafted W(VI)-complexes on the silica surface as a function of temperature (T) for a water partial pressure (p) equivalent to the ambient air water partial pressure (pw = 1500 Pa).58 At these conditions, the mono-grafted model A is the most stable until T = 100 K, followed by the mono-grafted models C in the temperature range of 100–220 K and finally at T > 215 K the tetra-grafted complex (model I) is found as the most stable configuration. It should be noted that the tetra grafted species (mono-oxo species), although observed experimentally between 473–1073 K is statistically disfavored due to the low probability to have four silanols in one nest. Taking this constraint into account it is the F structure (di-grafted di-oxo species) that appears to dominate at T > 400 K. These results are fully consistent with the experimental procedure used in the synthesis of W(VI)-supported catalysts by grafting methods,3 where samples are heated and annealed at high temperatures to obtain W di-oxo surface structures. Note that such species correspond to completely dehydrated conditions. In hydrated conditions (high water pressure or low temperature) mono-grafted model with W–OH group could be stabilized. Hydroxylated di-oxo species are predicted to be stable in the range of 100–400 K depending on the silanol density at the surface. However, at ambient conditions on low silanol density silica surfaces the di-oxo WO species are expected to dominate completely the W oxide supported geometries.
![]() | ||
Fig. 3 Phase diagram (surface energy vs. temperature) showing the stability ranges for the different grafted W oxide geometries. |
In summary, the three grafted W(VI) species may exist on a silica surface depending on the experimental conditions. They are supposed to reversibly interconvert in the presence of water, and they might coexist on the surface.
Comparing the theoretical frequencies with the experimental ones,3,26 we can conclude that the model containing the most similarities with the experiment are models F and I (di- and mono- oxo tungsten species, respectively). Interesting to note is that Chauvin et al.28 propose oligomers, which might correspond to model H monomers.
This was concluded with the use of a scaling factor for the frequencies, independently from the type of bond and normalized on the well-known silanol vibration. This approach has been used with success in former studies31,32,34,35 According to Lee et al.3,26 the surface tungsten oxide species on the supported WO3/SiO2 catalyst are present as both dioxo (O)2W(–O–Si)2 and monoxo O
W(–O–Si)4 surface species, giving rise to Raman bands for νsym (W(
O)2) at 985 cm−1 and νsym (W
O) at 1014 cm−1. The corresponding asymmetric νasym (W(
O)2) vibration appears as a shoulder at 968 cm−1 and the bending δ(O–W–O) mode at 346 cm−1.3,26,59,60
The calculated vibrational frequencies are tabulated in Table 4. Vibrational analysis shows that the WO bond vibrations are not as cleanly decoupled from others as was in the case for chromium.34 Only the frequency of asymmetric W
O vibration can be easily given. The calculated values (not show in the table) range from 973 to 1001 cm−1 due to various hydrogen bonds arrangements, which in turn depend on the hydration level. In effect, a wide band cantered around 991 cm−1 should appear in the spectrum. From these results combined with the thermodynamic calculations, one can conclude that the vibrational frequencies calculated for model structures F and I confirm the experimental results presented in ref. 3 and 26, which supposes a competition between mono-oxo and di-oxo species, originating from the availability of the silanol groups on the silica surface. The grafting of these oxide species does not only depend on the concentration or coverage of silanol groups on the surface but also their distribution. In other words, in order to have mono-oxo-species one needs have 4 silanols grouped close together, whereas for the di-oxo species only two silanols close to each other is sufficient. The proportion of the mono/di-oxo species is thus not only dependent on the degree on hydration or the silanol coverage.
W![]() |
W![]() |
Exp.3,26 | |
---|---|---|---|
B | 1008 | 968 | |
C(sym) | 1037 | 995 | |
C(asym) | 987 | 947 | |
E | 1039 | 997 | |
F(sym) | 1024 | 983 | 985 |
F(asym) | 981 | 942 | 968 |
H | 1033 | 991 | |
I | 1055 | 1013 | 1014 |
As a general conclusion compiling the results from our former studies we have calculated that in comparison with the other two group VI elements (Mo and Cr), supported W oxide species shows similar geometrical properties with supported Mo oxide species, with both elements having di-oxo species being the dominant one above 100 K at ambient vapor pressure, whereas for Cr the di-oxo species become dominant only above 400 K. M–OH groups are predicted to be more common for W follow by Cr. Mo has the least affinity to have a M−OH groups.
This journal is © The Royal Society of Chemistry 2016 |