Environmental Catalysis - American Chemical Society

Chapter 11

Adhesion of Rh, Pd, and Pt to Alumina and NO Reactions on Resulting Surfaces Model Calculations 1

2

3

Downloaded by GEORGETOWN UNIV on August 25, 2015 | http://pubs.acs.org Publication Date: February 23, 1994 | doi: 10.1021/bk-1994-0552.ch011

Thomas R. Ward , Pere Alemany , and Roald Hoffmann

Department of Chemistry and Materials Science Center, Cornell University,Ithaca,NY14853-1301

We report here approximate molecular orbital computations of adhesion and NO reduction in the Three Way Catalyst, modeled by a monolayer of either Rh, Pd or Pt on the (0001)O and (0001)Al faces of α-Al O . Platinum and palladium form stable interfaces with both oxygen and aluminum faces. Only the aluminum interface is stable with rhodium. Depending on the nature of the interface, the Fermi level of the composite systems varies dramatically. This, in turn, affects the adsorption mode, molecular or dissociative, of nitric oxide. From our study, it appears that an oxygen-platinum interface is best suited for both dissociative adsorption of NO as well as the coupling of two adsorbed nitrosyls to form a reduced dinitrosyl species with significant N-N double bond character. 2

3

Although nitric oxide was recently named molecule of the year for some of its physiological properties, it remains a major pollutant. In this respect, the 1990 amendments to the Clean Air Act set some stringent standards for the automotive industry (7).To meet some of these by the end of the millennium will take much work, one in which chemistry must participate. We believe that quantum mechanical modeling may aid in this effort. Since 1981, rhodium has been a major component in automotive catalysts in order to meet the standards for NO emissions (2). Although the nitrosyl reduction mechanism has not been fully elucidated, it is generally accepted that nitrosyls adsorbed on group 9 metal surfaces (3-6) dissociate more readily than ongroup 10 ones (7-13). This is often believed to be a key factor in determining the greater efficiency of Rh vs. Pd or Pt in the nitrosyl reduction process in the Three Way Catalyst (T WC). There is strong experimental evidence for the dissociative adsorption of NO on Ir, Mo, Al, Pt, Rh, Ru Ni, Ti, and Zn (8,14). In addition to linear M(NO)+, bent M(NO)- (75x

ûrrent address: ICMA, Université de Lausanne, Place du Château 3, CH-1005 Lausanne, Switzerland Current address: Departamento de Quimica Fisica, Universidad de Barcelona, Barcelona, Catalunya, Spain Corresponding author

2

3

0097-6156/94/0552-0140$08.00/0 © 1994 American Chemical Society In Environmental Catalysis; Armor, J.; ACS Symposium Series; American Chemical Society: Washington, DC, 1994.


11.

Adhesion ofRh, Pd, and Pt to Alumina

WARD ET AL.

141

19), and atomic Ν and Ο adsorption, surface dinitrosyl species, M(NO)2, have been reported on Pd(100)(20), Pd(l 11) (21) and Pt(l 11) surfaces (22,23). What if these dinitrosyl species were reaction intermediates in the nitric oxide reduction mechanism? In such a case, the formation of an N-N bond would preceed the NO bond cleavage. This appears to be the case for many gas phase reactions involving NO. As early as 1918, Bodenstein postulated the presence of a nitrosyl-dinitrogen dioxide preequilibrium to account for the third order rate law of many gas phase reactions involving NO (24,25). The actual TWC systems consist of metals supported on oxides. Much experimental work has been done on Cr-Si02 (26,29) as well as Ni-SiO2(30,31). In both cases dinitrosyl species, M(NO)2, have been detected. Such dinitrosyl entities are also present with transition metals supported on alumina, e.g. Rh (32,33), Cr, Mo, W (34) and Co (35,36). More recently, aluminosilicates (zeolites) have received increasing attention as supports. Again, dinitrosyl species recur in this field, i.e., Cr (37), Co (38), Fe (40) and Rh (41,42) exchanged into X or Y zeolites, as well as Cu exchanged into ZSM-5 zeolites (43). Adsorbed dinitrosyl entities have received little attention as possible intermediates in the transition metal mediated reduction of nitric oxide. Their frequent occurrence, especially with supported catalysts, triggered our curiosity. Therefore, we set out to analyze the possible role of dinitrosyl entities as possible intermediates in the nitric oxide reduction mechanism, by means of the extended Hiickel methodology. To do this we shall study the adsorption of nitric oxide and a dinitrosyl moiety on bare metals, rhodium, palladium and platinum, as well as on metals supported on alumina. 1

Geometry Since our main purpose here is the comparison of NO and N2O2 adsorption on bare and supported metal catalysts, the transition metals were modeled by a slightly modified three layer slab with the same geometry for all three metals. Rh, Pd and Pt crystallize in the fee structure with cell parameters 3.804, 3.891 and 3.924 Â respectively (50). In this study, we used the value of 3.803 Â for all three metal model slabs. This value leads to a triangular (hexagonal close-packed) mesh with an M-M distance of 2.70 À for each layer. The inter-layer separation is 2.20 Â for the (111) face. The results obtained using this geometry are in close agreement with a previous study where the correct cell parameters were chosen (57). The extended Hiickel method does not allow reliable optimization of interatomic distances, so we must fix these. It is necessary to keep distances constant, so as to obtain a relative ordering of bond strengths for the systems considered. With this in mind, we set the N-O distance to 1.17 Â and the M-N distance to 1.8 Â for all calculations. Since the coupling of two adsorbed nitrosyls will be considered, yielding a dinitrogen dioxide, we include two NOs per unit cell. This simplifies the analysis, since both reactant 1 and coupled product 2 have the same number of FMO's. We shall focus on the interaction of the frontier orbitals of the adsorbate with the surface. For the adsorbate's FMO's, we introduce labels which hint at their origin, derived from 5aand 2π. In the starting geometry, both nitrosyls lie parallel to each other, 2.70 Â apart. At this separation, the in- and out-of-phase linear combinations of the FMOs show practically no splitting. Bringing the two nitrosyls together in the yz plane, as depicted above, lifts the degeneracy of the π and ny orbitals. As the nitrogens approach, the FMO's with N-N bonding character are stabilized, and the splitting between the in- and out-of-phase combinations becomes substantial. In 2, the N-N distance was set to 1.44 Â (57). A Walsh diagram for the coupling of two NO molecules with the formation of an N-N bond is presented in Figure 1. χ

In Environmental Catalysis; Armor, J.; ACS Symposium Series; American Chemical Society: Washington, DC, 1994.


142 ENVIRONMENTAL CATALYSIS


11. WARD ET AL.


143


Before considering specific adsorption geometries for both 1 and 2, let us compare the molecular orbital energies of the adsorbates to the DOS of the metal slabs. Figure 2 depicts the DOS of the three three-layer slabs RI13, Pd3 and Pt3, as well as the MO levels for a nitrosyl and those of a dinitrogen dioxide 2. Clearly, the band widths differ significantly from slab to slab, rhodium having the widest d-band. In our previous contribution, we showed that for nitrosyl chernisorption, the energy term E° - E°j in the second order perturbation expression (1) dominates (57). Independent of adsorption geometry, we found that the energy match controls the stabilization energy.

For a nitrosyl molecule, the 2π* level is occupied by one electron. The computed N-O overlap population for a pair of nitrosyls 1 is 1.13. Upon interaction with a surface, if this level is located above the Fermi level, electrons will be dumped from the 2π* level into the slab, resulting in a strengthening of the N-O bond, Figure 3a. If the 2π* level is located below the Fermi level, the electron flow is reversed, weakening the N-O bond, Figure 3b. This latter situation corresponds to backbonding in the Blyholder model, the bonding being provided to a great extent by the 5σ Metal interaction (52). Therefore, by inspecting the electron occupation in the 2π* level once the interaction with the slab is turned on, it is possible to predict whether the N-O bond is strengthened or weakened upon interaction. On the M(l 11) face of the metals there are three common adsorption geometries of a linear nitrosyl adsorbate to consider: on-top M(NO) 3a, bridging-M(NO) 3b and capping M(NO) 3c. However, we must bear in mind that we need two nitrosyls in close proximity in order to couple them. The various adsorption geometries of linear M(NO) are depicted below: From these starting geometries, we may consider a number of coupling paths to afford adsorbed N 2 O 2 2. Three of these are depicted below: We shall discuss in detail the on-topM(NO) 3a geometry and the corresponding coupled product bridging Μ ( Ν 2 θ 2 ) 4a. Keeping in mind that the energy term in (1), which is independent of adsorption geometry, dominates, similar results are expected for these geometries. The reader should refer to a the complete account of this work for details on other geometries (53). We may now turn on the interaction between the adsorbate and the slab. From Figure 2a, we predict a good interaction for all 5σ and 2π levels of NO 2 adsorbed on Rh(lll). As all these levels are all located "in the d band", a good interaction is expected. The palladium situation is quite different, Figure 2b. Since the d band is significantly narrower for palladium than for rhodium, the 2π levels are expected to interact only weakly with the d band. The 5σ levels lie at the top of the d band, and their interaction with the palladium slab is expected to be good. The platinum situation is intermediate between rhodium and palladium, Figure 2c. Again, the width of the d band allows prediction of the strength of interaction: the 5σ levels interact very strongly, whereas the 2π levels, lying slightly above the d band, should interact moderately with the platinum slab. From our computations, it appears the interaction between 5σ levels and the slab is more or less independent of the nature of this latter. By comparing both 5σ electron occupations after interaction for all metal situations, these numbers vary little. Of the two electrons present before interaction, ~0.25 electrons are donated to the slab. On the other hand, the 2π electron occupation varies dramatically. For this FMO, the resulting electron occupation is 0.90, 0.09 and 0.28 for on-top Rh(NO), on-top Pd(NO) and on-top Pt(NO) respectively. Upon interaction with either Pd or Pt, the electron present in the 2π FMO has been dumped into the metal slab (Figure 3): the computed N-O overlap populations are 0.91, 1.32 and 1.23 for on-top Rh(NO), on-top Pd(NO) and on-top Pt(NO) respectively.


144

ENVIRONMENTAL CATALYSIS

π—•—J— —r 1

Rh

-4

-τ—'—ι—•—ι—'—r

Pt

Pd


-6 ~-1CH

·κ«

jwAvi!

e n

g -12

H

„:ÎM«Ï.".:""

LU

-U

-16

H

-18

Figure 2 a-c). Total DOS of a three layer metal slab M3 where a) M = Rh; b) M = Pd; c) M = Pt d) MO levels of NO ; e) MO levels of N2O2 2.


Adhesion of Rh, Pd, and Pt to Alumina

145


11. WARD ET AL.

bridging

4a

M(N 0 ) 2

2

on-top

M(N 0 )

4b

2

2

perp

M(N 0 )

4c


2

2

146

ENVIRONMENTAL CATALYSIS

Since the N-O Overlap Population of free nitric oxide is 1.13, it appears that the strength of interaction between the 2π FMO and the slab determines the electron flow and therefore the adsorption mode: dissociative for Rh and molecular for Pd and Pt. The same analysis can be carried out for the coupled product adsorbed on the various metal slabs, Figure 2e. The 2πγ level is shifted up in energy to 3.68 eV upon coupling. Its interaction with either metal slab will be insignificant. For the rhodium slab, all other levels are expected to interact strongly with the metal surface. Since for the palladium case only the 5σ+ level lies within the d band, no other level will interact very significantly with the slab. Finally, for N 2 O 2 2 adsorbed on platinum, except for the 2 π and 2%-, all other levels lie in the d band, and, therefore, their interactions with the metal slab are expected to be good. In the case of adsorbed N 2 O 2 2, we should focus both on the N-O and N-N overlap population. Indeed, we wish to have a coupled product with a strong N-N bond and an N-O bond as weak as possible. For comparison, we may consider two limiting structures, dinitrosyl 2 and hyponitrite 5, and their respective N-O and N-N overlap populations. The hyponitrite 5 corresponds to a reduced dinitrosyl with a 6π electron system. The computed N-N and N-O overlap populations, as well as the Fermi levels for bridging Μ ( Ν 2 θ 2 ) 4a are summarized in Table I. Again in this case, rhodium stands out. The N-O overlap population points towards that of a single bond as in 5 whereas the N-N overlap population is intermediate. +

+


ν

Table I. Overlap Populations and Fermi Level for bridging Μ ( Ν 2 θ 2 ) bridging Rh(N202) bridging Pd(N0?) bridging Pt(N202) -9.57 E/(Ev) -8.57 -9.54 N-O OP 0.84 1.11 1.03 N-N OP 0.70 0.61 0.62 Next we shall study the adhesion of these metals on alumina, and its consequences on the nitric oxide reduction. Models Computational constraints prevents us from using the γ-Αΐ2θ3 modification as the support. Instead, we used C1-AI2O3, which has a considerably smaller unit cell, and has been shown to be usable as the support in the TWC (54,55). C1-AI2O3 has a corundum structure with all aluminum atoms in an octahedral coordination geometry, while γ-Αΐ2θ3 has a vacant spinel structure with aluminum atoms both in octahedral and tetrahedral holes of the structure (56). From our previous experience, adhesion properties are not strongly affected by the coordination geometry of the surface aluminum atoms (57,58). The most important factor is the surface Al/O ratio for each face. Thus, it seems reasonable to assume that adhesion strength on the (0001)0 and the (0001)A1 faces of C1-AI2O3 will give lower and upper bounds for the adhesive properties of transition metals on different forms of alumina. The main features of the DOS for γ-Αΐ2θ3 will closely resemble those of C1-AI2O3; only the magnitude of the interaction of surface states with the metal bands may differ. The choice of γ-Αΐ2θ3 as a support is due to a great extent to its high surface area. This allows a high dispersion of the precious metal. In our models, we must assume coverages of 1ML (MonO and MonA\) or 2/3ML ( M 2 A I ) , while in industrial catalysts the metal coverage used is never higher than 0.1 ML. The particle size of the noble metal rafts in a fresh TWC is less than 50 Â. This can increase to 1000 Â, or more, at the high temperatures of vehicle operation. Since most of the catalytic activity is preserved (59), we



11. WARD ET AL.


147

may consider our metal coverage of LOML as approximating such large rafts. However, we cannot simulate interactions between the adsorbates and the support. The formation of rafts or islands thicker than one layer is also not incorporated in our models. This phenomenon can lead to adsorption of NO on a different face of the metal or to simultaneous interactions of the adsorbate with metal and support. Other effects that have been omitted in our models are modifications of metal support interactions due to dopants in the oxide phase (NiO, Ce02). Nickel oxide is added to the system to inhibit the diffusion of the precious metal in the spinel (59). Cerium oxide is added to prevent sintering of the catalyst and to increase the dispersion of the noble metal. Finally, physical characteristics of the support (polycrystalline, rough surfaces, diffusion of metal into oxide and of Al or Ο into the metal) were not included in our simulations either. Our model of the interface consists of the superposition at a fixed distance (2.0 Â) of two two-dimensionally infinite slabs of oxide and metal, respectively. Keeping this interface distant constant allows us to compare trends in adhesion energies. Comparable results were obtained with a 2.5 Â separation. In our previous study (57,58), the metal geometry was adjusted to fit the rigid CC-AI2O3 slab. Since our main purpose here is the comparison of NO and N2O2 adsorption on bare and supported metal catalysts, a different approach was followed. The transition metals were modeled by a slightly modified three layer slab with the same geometry for all three metals (vide infra). The oxide slab has been modified to provide epitaxy with the rigid metal slab. Our model consists of four layers of close-packed oxygen atoms with interlayer O-O distance of 2.70 Â. Three layers of aluminum cations, occupying two-thirds of the octahedral holes between oxygen layers, and a passivating hydrogen layer on the back surface of the slab, complete our model for the (0001)0 face. In our model slab all aluminum atoms between two oxygen layers are located in a single plane, centering octahedral holes. All Al-O distances have the value 1.903 Â. The (0001)A1 face is modeled by a slab obtained by removal of both the topmost oxygen layer and one of the aluminum atoms on the first aluminum layer. Figure 4 shows a schematic representation of two views of the (0001)0 surface model slab. Adhesion Let us review briefly the results of our earlier study of the interfaces formed between first row transition metals and different faces of C1-AI2O3 (57,58). Two main interactions were found to dictate the adhesive properties: a) Net destabilizing O-M interactions (See schematic diagram 6) are responsible for the weak (or even repulsive) adhesion energies found for oxide surfaces with high oxygen concentration. Filling of an interface M-O* band leads to a highly unfavorable situation in the case of high d-electron counts (late transition metals). b) Aluminum cations on the oxide surface are capable of forming strong charge transfer bonds with the metal. These bonds result from interaction of empty aluminum dangling bond states with the d band of the metal 7. A balance of both interactions determines the adhesion of a particular transition metal to C1-AI2O3. The most important factor is the ratio of Al/O atoms directly exposed to the metal surface. Aluminum-rich surfaces are predicted to form strong interfaces with almost all first transition row metals, while oxygen-rich surfaces do not result in strong interfaces, especially with late transition metals. The coordination geometry of atoms at the surface does not seem to be a major factor in determining adhesion properties for different faces. For instance, calculations of the adhesion energy for first row transition metals on different faces of C1-AI2O3 surfaces indicates that the (0001)0 and the (0001)A1 surfaces give upper and lower bounds for the adhesion energies. Surfaces with intermediate Al/O ratios yield interfaces with intermediate energies.


148

ENVIRONMENTAL

Θ


%

Ν

//0.95

Ν

0.61

\ = Ν

/

θ

·

6

Θ

2

0.80

5 " ff

. ,ϋοϊο] [ 2< r

[0001],

[Μ00]0

-5ι

ιδ,

CATALYSIS

•[Ν20]

Dloo]

[ooo ι] Gr

Ο

Oxygen ( 1st layer)

•

Oxygen (2nd layer)

Aluminum

Δ

Aluminum (1st layer)

Oxygen (2nd layer)

A

Aluminum (2nd layer)

•

Hydrogen

Oxygen (1st layer)

Octahedral holes

Figure 4. Top and side views of the (0001)0 surface model slab of C1-AI2O3.


Adhesion of Rh, Pd, and Pt to Alumina


11. WARD ET AL.


149

150

ENVIRONMENTAL

CATALYSIS

These findings agree with recent experimental results by Ealet and Gillet for the alumina (-1012) surface (60) as well as Gorte et al work on alumina (0001) (61,62). Results


Table II shows the adhesion energy values obtained for both the (0001)0- and the (0001)A1-Metal interfaces. For simplicity, we abbreviate these as M3CWO and M3^nAl, respectively. M3 symbolizes a three layer metal slab. Surprisingly, we obtain stable interfaces for both Pd and Pt with the (0001)0 surface, PàônO and VXônO. Such stabilization was not observed for Ni in our previous work (57,58). The charge transfer, AqM(e")> (Table II) also indicates the magnitude of the charge transferred to the metal (positive values) or to the ceramic (negative values). Table II. Adhesion energies and charge transfer for the (0001)0 and the (0001)A1 metal interfaces, M30/2O and MônAl respectively ionic contribution Môn 0 M30/1AI eV/uc J/m Aqvi(e") eV/uc J/m AqM(e-) eV/uc J/m Rh 0.483 0.409 +0.11 -3.616 -3.059 -0.32 -0.874 -0.739 Pd -0.921 -0.779 +0.16 -1.999 -1.691 +0.11 -0.135 -0.114 Pt -0.324 -0.274 +0.16 -2.508 -2.122 -0.24 0.000 0.000 uc refers to unit cell AqM( ") refers to the charge transfer 2

2

2

e

For the interaction with the aluminum surface a different factor, ionic charge transfer, has to be included in our discussion. This term arises from the electrons being transferred from the slab with the highest Fermi level to the other slab. It gives an additional stabilizing factor for the interface. In a one-electron calculation, such as the extended Hiickel one we use, this ionic component is overestimated. In order to assess the relative magnitude of ionic charge transfer and covalent interactions we performed two separate calculations with interfacial separations of 2.0 and 10.0 Â, respectively. At an interfacial distance of 10.0 Â all overlap terms between orbitals of both slabs, metal and ceramic, are zero, and the entire stabilization energy comes from the ionic term. The last entry of Table II shows that the magnitude of ionic charge transfer is important for Rh, moderate for Pd, and non-existent for Pt. Analysis of the changes in average charge on the metal atoms leads to the qualitative diagram shown in Figure 5. From this figure we can see that the interface with Rh is unstable for the (0001)0 surface, RI130/ÎO, but strongly stabilized when interacting with the (0001 )A1 surface, RhônAl. Important ionic charge transfer to the empty aluminum dangling states and the good energy match of these with the d band of the metal provide a substantial adhesion energy in this case. The case of palladium, with stable interfaces for both PâônO and Pd3

Environmental Catalysis - American Chemical Society

Recommend Documents