Surface grafting as a route to modifying the gas-sensitive resistor properties of semiconducting oxides: Studies of Ru-grafted SnO₂

Abstract

The phenomenon of electrical conductivity of an oxide being controlled by the chemical state of a surface grafted reactive centre, resulting in a room-temperature gas response, is demonstrated. Surface grafting of Ru centres onto SnO₂ by reaction of [(η⁶-C₆H₆)Ru(acetone)₃][BF₄]₂ with surface OH, followed by thermal decomposition in either H₂/N₂ or air resulted in additional electronic states in the SnO₂ band gap associated with surface Ru species, revealed by XPS and correlated with resistance increase of the material. The samples were characterized by EXAFS to confirm the structure of the surface Ru species, TPD, UV–VIS spectroscopy, XPS and electrical measurements. Thermal decomposition in H₂/N₂ resulted in isolated Ru centres bound to SnO₂. The chemical state of these Ru centres varied as a result of gas chemisorption (oxygen, water vapour, carbon monoxide and nitric oxide). An electronic interaction between grafted Ru centres and the SnO₂ support was manifested in room-temperature conductivity being controlled by the surface state of the Ru and by gas adsorption onto these centres. Decomposition in air resulted in small surface-bound clusters of RuO₂ and did not result in a gas-sensitive preparation. DFT molecular cluster calculations successfully illustrated the main features of the experimental results, supporting the interpretation that the conductivity changes were caused by gas interaction with surface Ru centres altering the trapping of conduction electrons on these centres. Oxygen, NO and CO were non-dissociatively adsorbed. The Ru centres slowly oxidised in air and hydroxylated in the presence of water vapour. Carbon monoxide and NO in ppm concentrations in air displaced oxygen from the Ru centres but the electrical response was strongly affected by the presence of water vapour in the atmosphere. In principle, surface-grafted reactive centres can be chosen to be specific to a particular gas, providing a route to new types of gas detector tailored for a particular application.

---

Introduction

There have been a number of reports concerning the modification of gas sensor properties of tin dioxide by dispersion over the surface of sufficiently small particles of a second phase, usually a precious metal but also another oxide.^1–3 One model for understanding these modifications is that the charge carrier concentration in tin dioxide is controlled by an electronic equilibrium with the supported second phase, such that the conductivity of the tin dioxide support reflects the electronic state of the supported particles, which is itself altered by interaction with the gas phase. The behaviour of semiconducting oxides as gas-sensitive resistors is due to variations in surface concentration of electronic trap states, these states being identified with adsorbed oxygen species.^4,5 It is a natural extension of these ideas to consider engineering a surface trap state, lying in energy below the adsorbed oxygen species so that the engineered state, rather than the oxygen state, controls the conductance, and choosing this surface species so that it might have a specific interaction with the gas to be detected. Then, a change in electrical conductivity of the support would signal the occurrence and extent of the interaction. In this paper, we describe some steps towards this goal, using surface grafting of Ru onto SnO₂ as the model. We support the interpretation of the experimental results with density functional theory (DFT) calculations on model Ru–SnO₂ clusters.

Canevali et al.⁶ have recently described an alternative approach, encapsulating Ru(III) as Ru(acac)₃ within a tin oxide gel then generating isolated Ru(IV) centres within and on SnO₂ by thermal decomposition in air. They used EPR spectroscopy to show that the Ru(III) could be regenerated by reduction in CO(5%)/Ar at 773 K, that there were four different environments for the Ru species of which three (judged by their subsequent evolution upon exposure to air at room temperature) lay on or near the surface, and that there was a synergistic interaction of Ru(III) and Sn(IV) with oxygen in dry air at room temperature, expressed as:

Impregnation of solid SnO₂ with Ru(acac)₃ followed by thermal decomposition in air then reduction with CO/Ar in contrast produced many fewer Ru(III) centres, showing instead clusters of reduced Ru, and the EPR signal of V_O^•. These authors demonstrated an electronic interaction between the Ru centres and the oxide but did not explore consequences for the electrical conductivity or the electrical response of their materials to the presence of traces of reducing gases in the atmosphere, at room temperature.

Experimental methods

Preparation of SnO₂–0.2%Sb

SnO₂ was prepared by calcining metastannic acid (Keeling and Walker) at 900°C for 3 h in air. The material was then doped with 0.2 atom Sb, by mixing for 24 h with Sb₂O₅ in a ball mill with acetone and ceramic beads, separating the beads, drying and firing at 1000°C for 12 h in air. The doping was to increase the electrical conductivity to a range easily measurable at room temperature: if the dopant level is too low then resistances are too high for convenient measurement; and if too high, then the conductivity is so high that changes in surface trap state concentration have negligible effect. Scanning electron microscopy and transmission electron microscopy of materials dispersed onto a grid from a slurry in CCl₄ showed that the resultant material was in the form of agglomerates, typically 200 nm in dimension, of primary crystallites typically 50 nm in dimension. The surface area (Kr adsorption at 77 K) was 19 m² g⁻¹ implying an average particle diameter of 46 nm.

SnO₂–0.2%Sb support pretreatment

To obtain reproducible initial conditions for the SnO₂–0.2%Sb support surface, the powder was first dehydrated by calcination at 400°C under vacuum for 2 h, and then oxidized at 400°C for a further 2 h in pure oxygen (BOC). To rehydrate the surface, the material was exposed overnight to flowing synthetic air (21% O₂, 79% N₂, purity 99.99%, BOC) saturated with water at room temperature. Finally, to remove molecular water from the surface but retain surface –OH, the powder was dried at 250°C for 2 h in vacuum.⁷

Reactive solution preparation

[{(η⁶-C₆H₆)RuCl(μ-Cl)}₂] was prepared by literature methods.^8,9 The compound [{(η⁶-C₆H₆)RuCl(μ-Cl)}₂] was stirred with Ag[BF₄] (1:5 mole ratio) in anhydrous acetone (10 cm³) for 1 h then filtered through celite four times to remove the AgCl precipitate. The orange solution of [(η⁶-C₆H₆)Ru(acetone)₃][BF₄]₂^10,11 generated in this way (Ru precusor solution) was used in further preparations.

Grafting procedure

The route employed for surface grafting was to react the SnO₂–0.2%Sb support with the Ru precursor solution, in order for the latter to react with surface –OH. The remaining ligands could subsequently easily be removed from the Ru by pyrolysis. Thus, preconditioned powder (200 mg), was mixed into the bright orange Ru²⁺ precursor solution (10 cm³) without exposure to ambient air, and stirred overnight at room temperature. Unreacted precursor solution was removed carefully by repeatedly subjecting the mixtures to centrifugal force, removing the solution and adding acetone, then finally filtering and drying. Absorption spectroscopy of the Ru aryl acetone precursor solution (λ_max = 394 nm; absorption coefficient, ε_λ = 450 dm³ mol⁻¹ cm⁻¹) was used to measure the consumption of the precursor and hence the extent of surface reaction, by comparison with a blank procedure identical in all respects except for the introduction of the SnO₂ . The surface loading measured in this way was 1.25 × 10⁻¹⁰ mol cm⁻². The decomposition of the surface-bound complex was observed by mass-spectrometric measurement of the evolved benzene during heating in ultra-high vacuum (UHV) over the temperature range 100–300°C (Fig. 1). Uniformity of surface loading and the absence of large (μm-scale) agglomerates of Ru species was confirmed by electronic probe microanalysis.

Fig. 1 TPD of the SnO₂–0.2%Sb supported Ru aryl complex: mass 78(C₆H₆⁺) signal.

X-ray photoelectron spectroscopy (XPS)

The method utilised an ESCALAB 220i-XL instrument; base pressure <10⁻⁹ torr; monochromated Al-Kα X-rays; 300 μm spot size; analyser pass energy 20 eV; spectra acquired with 0.l eV steps; sample charging controlled with 3 eV electron flood gun; spectrum quantification and curve fitting carried out using a Shirley background and sensitivity factors obtained from Wagner et al.¹² The adventitious hydrocarbon C 1s peak could not be used for satisfactory Fermi level referencing as the Ru 3d signal partially overlaps the C 1s signal. Therefore, the Sn 3d_5/2 peak was used as a binding energy reference and assigned a value of 486.7 eV.¹² No evidence for chemical shifts of the Sn 3d_5/2 was found since only one peak of constant FWHM (1.16 ± 0.05 eV) was required to fit the region and the binding energy difference (ΔE_b) between the O 1s and the Sn 3d_5/2 peaks was constant (ΔE_b = 44.0 ± 0.1 eV) for all samples. However, C 1s obstructions complicate both the determination of the exact Ru peak position and the quantification of the Ru peak areas. Thus a curve fitting procedure was applied in the Ru 3d region to enable separation of spectra into constituent peaks and therefore allow assignment. This, in the case of Ru 3d, was carried out by setting the 3d_5/2:3d_3/2 peak ratio to 3:2 and the (3d_5/2–3d_3/2) binding energy difference to 4.1 eV.¹³ RuO₂ was used as a standard. The spectrometer was not fitted with a sample pretreatment chamber. For experiments which required a controlled atmosphere transfer (see Results) the sample loading port was covered with a nitrogen-filled bag.

To estimate the degree of surface hydroxylation of SnO₂–0.2%Sb support the curve fitting procedure was applied in the O 1s region of the spectra of unmodified support. Two oxygen species were identified on the SnO₂–0.2%Sb support surface with binding energies of 530.5 eV for the lattice oxygen contribution¹² and 531.8 eV for the hydroxy group oxygen.¹⁴ The qualitative contribution of the hydroxy group oxygen to the total surface oxygen was found to be 23.6 atom%.

Extended X-ray absorption fine structure (EXAFS)

Ru K-edge X-ray absorption spectra were collected at CCLRC Daresbury Laboratory Synchrotron Radiation Source, operating with a beam energy of 2 GeV and an average current of ca. 200 mA, on station 9.2, equipped with a double crystal Si(220) water-cooled monochromator which was offset to 50% of maximum intensity for harmonic rejection. Data was collected at room temperature and in a fluorescent mode for the supported catalysts using a Canberra 13-element Solid State detector. Spectra for the Ru metal and RuO₂ standards and the precursor material were collected at room temperature and in a transmission mode. Ru K-edge data was analysed using a suite of programs available at Daresbury Laboratory, namely EXCALIB (for converting the raw data to Energy [italic v]s. absorption coefficient), EXBROOK (for pre-edge subtraction, extraction normalized XANES data and post-edge background subtraction to extract EXAFS data) and EXCURV92 (for detailed curve-fitting analysis of the EXAFS data).

Electrical conductivity measurements

Connection for electrical measurements was made by pressure contact of gold foils to pellets previously formed into discs 13 mm diameter, 1–2 mm thick by compression under 1000 kg force. Three pellets and electrodes for concurrent measurement were mounted in series, separated by alumina spacers, inside a tube furnace inside a silica tube, with flowing gas atmosphere controlled by mass-flow controllers. Two-terminal dc resistance measurements used an auto-ranging digital multimeter. Although there are difficulties of electrode polarisation in some circumstances for such 2-terminal measurements, experience has shown them to be reliable for comparison of materials.^15–17. In the present work, at least one of the pellets being measured simultaneously was an un-ruthenised blank. To check for electrode contact artifacts, measurements were also made in ambient air using a Wayne-Kerr precision ac bridge operating at 1 kHz, with powder compressed within an insulating die in a press at constant force of 300 kg. Gases were certified mixtures or pure gases obtained from BOC plc.

Gas adsorption and temperature-programmed desorption (TPD) measurements

These were performed in a UHV chamber (base pressure <10⁻⁸ torr) fitted with a quadrupole mass spectrometer. Powder (0.1 g) treated with the precursor but not decomposed was placed in a quartz tube (2 mm id) on which was wound a tungsten heater filament. Sample temperature was monitored using a Pt/Pt–Rh thermocouple positioned half-way into the tube. A linear temperature ramp controlled using this thermocouple was obtained using a computer-controlled power supply applied to the heater element. The sample was heated under vacuum to decompose the precursor (Fig. 1), then cooled to room temperature before gas adsorption and TPD.

Computational methods

All calculations were performed with the Amsterdam density functional (ADF) program suite.^18,19 This code employs uncontracted Slatertype orbital valence basis sets; the quality of the bases used in the present calculations is described in Results, Section (v). Frozen cores have been employed for all atoms bar H—O 1s, C 1s, Sn 4p, Ru 3d. The local density parametrisation of Vosko, Wilk and Nusair²⁰ was employed, and was sometimes supplemented with Becke's gradient correction²¹ to the exchange part of the potential and the correlation correction due to Perdew²² (see Results, Section (v) for details). Mulliken population analyses were performed.²³ The calculations were performed on a Silicon Graphics Indigo² workstation and the EPSRC's “Columbus/Magellan ” central computing facility.

Results

(i) Electrical behaviour in response to CO, NO and H₂O at low concentration in the gas phase

The change of electrical resistance of the Rugrafted SnO₂–0.2%Sb in response to change in the gas atmosphere depended upon the preconditioning of the pellet, which was carried out in the test rig immediately prior to application of the test gas mixtures. There were two pretreatments: (a) reduction in 10% H₂/N₂ at 300°C for 1 h to decompose the surface Ru complex, followed by cooling to room temperature under H₂/N₂ then change of the gas atmosphere to the test mixture. This pretreatment is called “reduction ” throughout the following; (b) heating in dry synthetic air at 300°C for 1 h followed by cooling in dry air to room temperature. This pretreatment is called “air decomposition” throughout the following. Fresh samples were taken for each new set of measurements. The main features of the electrical behaviour were:

(a) The “reduced” Ru-grafted SnO₂–0.2%Sb showed an electrical response at ambient temperature to small concentrations of CO and NO in both dry air and dry N₂ (Fig. 2 and 3). In dry N₂, in contrast to the behaviour in dry air, the recovery of resistance after removal of the trace gas was extremely slow. The recovery after exposure to NO was slower than that after CO exposure. Ungrafted material and the air-decomposed material showed no electrical response to these gases apart from a very small and irreversible effect of CO in air on the “ reduced” ungrafted material (Fig. 4). The electrical resistance of the “reduced” Ru-grafted SnO₂ steadily increased with time of exposure to air, to values much greater than that of the untreated blank. Neither the untreated blank, nor the “air decomposed” sample showed this effect. The resistivity in air of the “reduced ” samples and blanks was much less than that of the “air-decomposed” samples and blanks (Table 1).

Fig. 2 Time profile of the room temperature (20°C) resistance of the reduced Ru-modified SnO₂–0.2%Sb pellet as a function of exposure to ppm concentrations of (a) NO and (b) CO in nitrogen.

Fig. 3 (a) Time profile of the room-temperature resistance of the reduced Ru-modified SnO₂–0.2%Sb pellet as a function of water vapour pressure and exposure to ppm concentrations of CO in dry and wet (50% relative humidity, 25°C ) air. (b) Time profile of the room-temperature resistance of the dry reduced Ru-modified SnO₂–0.2%Sb pellet as a function of exposure to ppm concentrations of CO in dry air—(part labelled ‘B’ in Fig. 3(a)). (c) Time profile of the room-temperature resistance of the dry reduced Ru-modified SnO₂–0.2%Sb pellet as a function of exposure to ppm concentrations of NO in dry air.

Fig. 4 Time profile of the room-temperature resistance of the ungrafted reduced SnO₂–0.2%Sb pellet as a function of water vapour and exposure to CO in dry and wet (50% relative humidity) air.

Table 1 Resistivity, ρ/Ω mm of pressed porous pellets, in dry air and air with 1240 Pa H₂O (50% relative humidity at 20°C: “wet air ”); 20°C

Dry air	Wet air
a Continuous upward drift, see Fig. 1. b Resistivity up to 30 times that of the ungrafted blank was observed, depending on the time of exposure to ambient air.
Sn(Sb)O2, “reduced”	100	30
Sn(Sb)O2, “air decomposed”	2 × 10^5	1 × 10^3
Ru-Sn(Sb)O2, “reduced”
(a) freshly prepared (Fig. 1)	150–300^a	50–150^b
(b) after exposure to ambient air	1 × 10^3
Ru–Sn(Sb)O2, “air decomposed”	3 × 10^5	1.2 × 10^3

---

(b) All the preparations showed an electrical response to water vapour (Table 1). The effect of exposure to water vapour was to diminish significantly the response of the “reduced” Ru-grafted material to traces of CO (Fig. 2(a)). This effect was not reversed by changing the atmosphere back to dry air.

(ii) EXAFS characterisation

All the samples characterised by EXAFS had been subject to prolonged exposure to ambient air prior to measurement. Fig. 5 shows the Fourier transform of the Ru K-edge EXAFS oscillations. Table 2 summarizes structural parameters for the RuO₂ standard and the two Ru-modified SnO₂–0.2%Sb samples (reduced and air decomposed) used to obtain the best complete fit to the total EXAFS spectrum (R = 44.78% for the reduced sample and 58.76% for the air decomposed sample). The air-decomposed material conformed well to the spectrum of RuO₂, with intense peaks corresponding to the first (Ru–O, 1.925 ± 0.006 Å) and second (Ru–Ru, 3.515 ± 0.008 Å) nearest neighbour distance. Evidently, bulk RuO₂ species were formed as a result of air decomposition. However, the reduced material showed no evidence of Ru–Ru next nearest neighbours. There was only one peak in the Fourier transform, cor responding to a Ru–O distance of 2.047 Å, longer than the Ru–O distance in RuO₂ (1.989 Å) and less than the Ru–O distances quoted for Ru supported on silica (2.18 Ų⁴) or titania (2.19 Ų⁵). The average oxygen coordination number for the reduced material was 4.3 compared to 6 for the bulk RuO₂. Evidently, isolated surface-grafted Ru species were formed by decomposition of the surface-bound complex in H₂/N₂.

Fig. 5 Fourier transforms of the Ru K-edge EXAFS oscillations of: (a) Ru/SnO₂ decomposed in air, and (b) Ru/SnO₂ reduced.

Table 2 Curve-fitting results for Ru K-edge EXAFS data (complete fit to the total EXAFS spectrum)

Sample	Atom type	N ^a	r ^b/Å	2σ^2^c/Å^2
a N: number of atoms in shell. b r: Distance of shell from absorbing atom (Ru). c σ ^2: Debye–Waller factor.
RuO2	O	6	1.989	0.008
	Ru	12	3.436	0.010
Ru/SnO2	O	4.3	2.047	0.008
Reduced	Ru	—	—	—
Ru/SnO2	O	6	1.925	0.006
Air decomposed	Ru	2	3.079	0.008
		8	3.515	0.008

---

(iii) XPS characterisation

For the [(η⁶-C₆H₆)RuCl₂]₂ precursor and for the surface-bound, undecomposed Ru aryl complex, the Ru 3d_5/2 fragment within the XP Ru 3d region was fitted by a single narrow peak (FWHM≈1.4 eV) corresponding to a single formal oxidation state (Ru²⁺). The binding energy was slightly increased (from 281.8 to 282.6 eV) for the surface-bound complex (Table 3).

Table 3 XPS data for surface-grafted Ru on Sn(Sb)O₂

Sample	Ru 3d5/2 peak binding energy maximum/eV	Ru 3d5/2 FWHM/eV	Ru/(Ru + Sn + Sb) (atom%)
(η^6-C6H6RuCl2)2	281.8	1.4
(η^6-C6H6-Ru)-SnO2	282.6	1.45	2.8
“Reduced”—as prepared	2.4
—dry air	2.7
—moist air	282.2	2.5
“Air decomposed”—as prepared	282.2	1.1
—moist air	282.2	1.4

---

Because the electrical response results showed that exposure both to dry and wet air changed the state of the “ reduced” surface in some way, on a timescale on the order of 10 min (Fig. 2(a)), exposure to air and moisture was controlled before the XPS measurement and the effects of such exposure studied. Samples were reduced in a small tube furnace and cooled under N₂/H₂, then either transferred into the spectrometer [italic v]ia a N₂-filled bag which covered the sample loading port or pretreated with either dry or moist air within the furnace tube before transfer into the spectrometer. Fig. 6 compares the XP spectra for the reduced material, the reduced material exposed to dry air and to moist air and the air-decomposed material. The C 1s signal overlaps Ru 3d_3/2′ but the Ru 3d_5/2 signal can be seen clearly. After reduction of the surface-bound complex, the formal oxidation state of the surface Ru species became very ill-defined and could not be characterised by a single binding energy: there appeared to be at least three Ru surface species, with the major species at lowest binding energy: 280.6 eV, 1.2 eV lower than that for the original dimer. Exposure to dry air caused a slight partial oxidation of the surface Ru. Further exposure to water vapour in air caused a further slight partial oxidation, as indicated by a shift of the Ru 3d_5/2 peak maximum to higher binding energy: the major species were now formed with binding energy approximately 282.2 eV. Table 3 illustrates that the surface Ru concentration remained unchanged as a consequence of these treatments. The air-decomposed samples also showed the Ru 3d_5/2 peak as a broad feature, with maximum around 282.2 eV, very similar to that of the reduced sample following exposure to moist air. In this case, on subsequent exposure to moist air, no significant change in either peak intensity or position was observed. The apparent Ru surface concentration was only half that before decomposition (Table 3).

Fig. 6 XP spectra of the Ru 3d region for SnO₂–0.2%Sb supported Ru recorded following reduction (1–3): 1, freshly reduced; 2, dry air exposed; 3, subsequently exposed to air of 50% relative humidity (20°C); and decomposition in dry air (4, 5): 4, freshly decomposed in dry air; 5, subsequently exposed to air of 50% relative humidity (20°C). (a) Ru 3d_5/3 peak fit for spectrum (1), peaks at A, B and C correspond to different chemical states of the surface Ru; (b) 3d_5/3 peak fit for spectrum (3), peaks B and C correspond to different chemical states of the surface Ru.

Fig. 7 compares the XP valence band spectrum of the reduced Ru-modified sample following exposure to ambient (moist) air with that of the Sn(Sb)O₂ sample pretreated in an identical way. New electronic states tailing into the band gap and a broadening of the valence band edge are shown for the Ru-modified sample.

Fig. 7 XP valence band spectra of pure SnO₂–0.2%Sb (a); and Ru modified SnO₂–0.2%Sb (b) Both samples were subjected to identical pretreatment. Note increase in intensity at ca. 1.5 eV below the Fermi edge for Ru-modified sample (marked with arrow) and broadening of the valence band edge towards the lower binding energy.

(iv) Studies by TPD of adsorption sites for CO, NO and O₂

The surface-bound Ru aryl complex was decomposed by heating in UHV then the resultant surface adsorption sites probed by adsorption followed by TPD of CO, of O₂ followed by CO and of NO. Probing the surface with CO showed (Fig. 8) at least 3 different CO binding sites, labelled A (75°C), B (150°C), C (270°C) in order of increasing desorption temperature: CO₂ desorption was associated predominately with CO binding to type C sites. Pre-adsorption of O₂ significantly diminished CO adsorption onto type A sites and caused a significantly increased desorption of CO₂, associated with CO adsorption onto type B and C sites. A blank measurement on unmodified Sn(Sb)O₂ showed only a very small desorption of CO, more than ten times less than that from the Ru-modified samples and barely above the vacuum background. NO preadsorbed onto the Ru sites desorbed (Fig. 9) from two sites: a molecular adsorption site with desorption temperature ∽125°C, and a dissociative adsorption site with desorption temperature ∽230°C.

Fig. 8 TPD of (——) CO and (– – –) CO₂ obtained after exposure of the reduced SnO₂–0.2%Sb supported Ru to: (a) excess CO (8 × 10⁻⁶ torr for 5 min) at room temperature; and (b) excess O₂ (8 × 10⁻⁶ torr for 5 min) followed by excess CO (8 × 10⁻⁶ torr for 5 min) at room temperature. Heating rate 30°C min⁻¹.

Fig. 9 TPD of (——) NO and (– – –) N₂ obtained after exposure of the reduced SnO₂–0.2%Sb-supported Ru to excess NO (8 × 10⁻⁶ torr for 5 min) at room temperature. Heating rate 30°C min⁻¹.

(v) Computational studies of model molecular clusters

DFT calculations were undertaken to probe the general qualitative trends in electronic structure modifications caused by Ru grafting and subsequent gas chemisorption onto supported Ru, with the aim of shedding light on the resulting experimentally observed room temperature gas sensitive electrical behaviour. These calculations employed the cluster approach, in which a small part of the surface of the solid is represented by a molecule. Although analogy between molecular cluster models and the properties of solid surfaces is limited (since any collective electronic and structural characteristics of the crystal lattice may have no counterparts among molecular models containing relatively few atoms) the cluster approach has been successfully employed to tackle local phenomena, e.g. to describe active surface sites^26–29 or adsorbates on the surface^29–31 as well as supported surface systems.^32,33

SnO₂ has the rutile crystal structure.³⁴ In the experimental parts of this work, polycrystalline SnO₂ was used with particle sizes around 50 nm. Rutile of this dispersion is believed to expose predominantly (60–85%) the thermodynamically most stable (110) crystal plane.³⁵ It is therefore reasonable to take as our computational molecular cluster model a small segment of the SnO₂ (110) surface, which we have assumed to be fully hydroxylated in agreement with our experimental results (see Discussion, Section (i)).

The stoichiometric hydroxylated SnO₂ (110) surface was modelled with molecular clusters of increasing size Sn_nO_mH_x. In all cases the valencies of the peripheral oxygen atoms of the cluster were satisfied by hydrogen atoms. Three different clusters were considered, ball and stick pictures of which are shown in Fig. 10. The smallest cluster used was [Sn₂O₁₀H₁₀]²⁻ (A), which represents a “ two-fold” tin dioxide adsorption site consisting of two tin atoms and a full nearest-neighbour coordination cell of five lattice oxygens and one terminal OH group per tin atom. The second cluster was [Sn₄O₁₈H₁₈]²⁻ (B), which has two more tin atoms and corresponding oxygen atoms and terminal OH groups. Finally the cluster was extended to [Sn₁₀O₄₀H₃₄]⁶⁻ (C), which contains rows of bridging OH groups (standing out of the Sn–O plane), and which maintains the full oxygen coordination around each added tin atom. In all cases the initial cluster geometries were based on the metric parameters for bulk SnO₂.³⁶ The initial r(O–H) distance was chosen to be 0.970 Å and the hydrogens were placed along the O–Sn direction. Note that for the purpose of the present calculations no distinction is made between the hydrogen saturators used to truncate the cluster and those used to hydroxylate the surface.

Fig. 10 Cluster models taken to represent surface adsorption sites on SnO₂ (110). Numbered atoms are referred to in the text.

To address the effects of adding one or more ruthenium atoms to the “surface” sites we have investigated the addition of Ru to clusters A and C. For cluster A we have replaced the hydrogen atoms of the two “ surface” hydroxy groups with a Ru atom to generate an O–Ru–O bridge (a two-fold Ru coordination site) while for cluster C we have added two ruthenium atoms. These Ru-modified clusters are shown schematically in Fig. 11. For each cluster the Ru atoms were initially placed at an Ru–O distance of 2.047 Å, this being the value obtained from the EXAFS experiments (see Results, Section (ii)).

Fig. 11 Cluster models for surface-grafted Ru on SnO₂(110). Numbered atoms are referred to in the text.

Fully stoichiometric clusters should have a 2:1 O:Sn ratio, as in bulk SnO₂. As our cluster models are non-stoichiometric an electronic charge must be added to compensate for the missing neighbour atoms. The number of electrons in each cluster (and hence the total cluster charge) is determined by assuming that tin, oxygen, hydrogen and ruthenium (bridging 2 oxygens) have formal charges of + 4, − 2, + 1 and + 2 respectively.

Varying levels of geometry optimisations were performed from the initial atomic positions outlined above. Local density approximation (LDA) calculations on hydroxylated clusters (Fig. 10 A–C) were performed optimising the positions of all of the hydrogen atoms and oxygen atoms 9 and 10 (cluster A—the two hydroxy group oxygens related by C_2v symmetry), 12, 13, 21 and 22 (cluster B) and 47–50 (cluster C), all other atomic positions being fixed. For the case of the Ru-modified clusters (Fig. 11) the calculations using the LDA were carried out optimising the positions of the Ru atoms and the “surface ” oxygens to which they are bound within the constraint of C_2v symmetry, all other atomic positions being fixed (the positions of the terminating hydrogen atoms were fixed as those for unmodified clusters). Double-zeta basis sets (ADF Type II) were used for all atoms for these LDA calculations.

For more sophisticated generalised gradient approximation (GGA) calculations the methodology was improved by the addition of gradient corrections to the LDA together with an increase in the quality of the basis sets to triple-zeta plus a single polarisation function (ADF Type IV) for all atoms (note that no polarisation functions are available for transition metals within ADF). Furthermore, the positions of all the atoms were optimized within the constraints of C_2v symmetry. The significantly increased computational cost of this approach meant that the calculations were confined to the smallest cluster, A (hydroxylated and Ru-modified).

(a) LDA. The computed LDA energy levels for the three clusters A–C are shown in Fig. 12. We concentrate on the electronic structure of the levels around the highest occupied (HOMO) and lowest unoccupied (LUMO) molecular orbitals (MOs), the atomic orbital (AO) compositions of which are given in Table 4.

Fig. 12 Computed LDA MO levels for: [Sn₂O₁₀H₁₀]²⁻, (a); [Sn₄O₁₈H₁₈]²⁻, (b); and [Sn₁₀O₄₀H₃₄]⁶⁻, (c). Double-headed arrows indicate the “ band gap” of the clusters.

Table 4 LDA eigenvalues and atomic compositions for the highest occupied and lowest unoccupied molecular orbitals of [Sn₂O₁₀H₁₀]²⁻, [Sn₄O₁₈H₁₈]²⁻ and [Sn₁₀O₄₀H₃₄]⁶⁻. C_2v symmetry notation is used to label the orbitals. The superscripts on the atomic symbols correspond to their location in the cluster from Fig. 10, and imply all atoms connected by C_2v symmetry

Cluster	Orbital	Energy/eV	Composition (%)
9a2	3.117	65 O2p^10, 22 O2p^3, 10 O2p^11
A	14b2 (HOMO)	3.437	83 O2p^10, 9 O2p^3
[Sn2O10H10]^2−	17a1 (LUMO)	5.453	62 H^19, 24 H^13
12b1	6.449	50 H^19, 27 Sn5p^1, 15 H^13
19b1	0.908	44 O2p^15, 29 O2p^19, 26 O2p^21
B	17a2 (HOMO)	0.909	44 O2p^15, 30 O2p^19, 25 O2p^21
[Sn4O18H18]^2−	30a1 (LUMO)	2.829	41 H^29, 42 H^33
28b2	3.688	74 H^33, 10 H^23
50b1	10.449	82 O2p^33, 9 O2p^45, 8 O2p^49
C	45a2 (HOMO)	10.450	82 O2p^33, 9 O2p^45, 8 O2p^49
[Sn10O40H34]^−	61a1 (LUMO)	11.620	31 H^77, 30 H^73, 6 Sn5s^3
51b1	12.280	27 H^77, 19 H^73, 7 H^69, 6 Sn5s^3, 5 H^57

---

The highest occupied levels in all three clusters are predominantly of oxygen 2p character, while the lowest unoccupied orbitals are a mixture of tin and hydrogen. In as much as the separation between the HOMO and LUMO of a molecule can be equated with the band gap of a solid, Table 4 indicates that the “band gap ” of clusters A and B is approximately 2 eV, while that of cluster C is significantly less, at 1.17 eV. (Note that LDA is notorious for underestimating band gaps.³⁷ Given the approximation of the calculations no deductions are drawn from the obtained values of the band gaps.)

Fig. 13 presents the valence MO energy levels for the two Ru-modified clusters. Filled circles indicate the electrons in the highest occupied level. We identify the “band gap” as the energy difference between the top of the occupied “valence states” (HOMO) and the bottom of “conduction states ” (20a₁/21a₁ for [Sn₂O₁₀H₈Ru]²⁻ and 65a₁/67a₁ for [Sn₁₀O₄₀H₃₀Ru₂]⁶⁻). As with the unmodified clusters, the “conduction states” are found to be of mixed tin–hydrogen AO character. It is notable that there are several unoccupied molecular orbitals in the “band gap” and that in both Ru-modified clusters the LUMO is now only 0.25–0.35 eV above the HOMO. The composition of these gap orbitals are given in Table 5.

Fig. 13 Computed LDA MO levels for: [Sn₂O₁₀H₈Ru]²⁻, (a); and [Sn₁₀O₄₀H₃₀Ru₂]⁶⁻, (b). Double-headed arrows indicate the “band gap” of the clusters.

Table 5 LDA eigenvalues and atomic compositions for the low-lying unoccupied “band gap” orbitals of [Sn₂O₁₀H₈Ru]²⁻ and [Sn₁₀O₄₀H₃₀Ru₂]⁶⁻ clusters. C_2v symmetry notation is used to label the orbitals. The superscripts on the atomic symbols correspond to their location in the cluster from Fig. 11, and imply all atoms connected by C_2v symmetry

Cluster	Orbital	Energy/eV	Composition (%)
[Sn2O10H8Ru]^2−	10a2	3.786	61 Ru4d, 35 O2p^11
[Sn10O40H30Ru2]^6−	53b1	10.213	59 Ru4d^81, 22 O2p^49, 13 O2p^47
48a2	10.283	60 Ru4d^81, 20 O2p^49, 16 O2p^47

---

Note that in both cases the lowest unoccupied gap orbitals arise due to mixing of Ru 4d and surface oxygen 2p orbitals, and are largely localized on the surface Ru atoms. Given the similarity in the composition of the 10a₂ MO of [Sn₂O₁₀H₈Ru]²⁻ and the 53b₁ and 48a₂ MOs of [Sn₁₀O₄₀H₃₀Ru₂]⁶⁻, we conclude that, qualitatively speaking, increasing in the cluster size does not significantly alter the modifications in electronic structure caused by the attached Ru.

(b) GGA. Before examining the electronic structure of GGA-calculated cluster A, we turn briefly to its geometric structure. Table 6 presents the optimised Sn–O distances from our calculation, together with those of Gillan et al.³⁷ (gradient-corrected DFT) on an SnO₂ lattice. Our calculation shows three Sn–O distances which are very similar to one another (Sn–O₃, Sn–O₁₀ and Sn–O₁₁; see Fig. 10(a) for atom numbers) and one significantly longer value (Sn–O₁₁). This difference is presumably due to oxygens 3, 10 and 11 being terminally bound to tin while oxygen 5 is bridging between two tin atoms. Comparison of our data with those of Gillan et al. reveals very good agreement for bulk SnO₂, with the exception of the Sn–O₅ value which is much shorter in the Gillan calculation. The surface data of Gillan reveal a shortening of all of the Sn–O bonds, and the agreement with our calculation is slightly worse than for the bulk data. Given the differences in computational approach, however, we feel that our cluster is an acceptable geometric model for part of the SnO₂ (110) surface, although we note that the bridging Sn–O distances are not as well reproduced as the others.

Table 6 Optimised Sn–O₂ distances (Å) from GGA-calculated cluster A and previous work by Gillan et al.³⁷ Atom numbers refer to Fig. 10(a) and imply all atoms connected by C_2v symmetry

	Cluster A (present work)	Bulk SnO2^37	SnO2 (110) surface^37
Sn–O3	2.088	2.054	2.029
Sn–O5	2.245	2.054	2.029
Sn–O10	2.114	2.088	2.000
Sn–O11	2.078	2.088

---

The GGA electronic structure around the HOMO–LUMO gap of the geometry optimised cluster A is given in Fig. 14(a). Table 7 summarizes the composition of those orbitals around the “band gap”. As with the LDA results described above, the highest occupied “ valence” orbitals are composed largely of O 2p levels and the unoccupied “conduction ” orbitals are of mixed tin-hydrogen character. The HOMO–LUMO “band gap” is 3.9 eV.

Fig. 14 Computed GGA MO levels for: [Sn₂O₁₀H₁₀]²⁻, (a); and [Sn₂O₁₀H₈Ru]²⁻, (b). Double-headed arrows indicate the “band gap” of the clusters.

Table 7 GGA eigenvalues and atomic compositions for the energy levels around the “band gap” of cluster A. C_2v symmetry notation is used to label the orbitals. The superscripts on the atomic symbols correspond to their location in the cluster from Fig. 10, and imply all atoms connected by C_2v symmetry

Orbital	Energy/eV	Composition (%)
9a2	0.999	91 O2p^3
11b2 (HOMO)	1.005	77 O2p^3, 11 O2p^10, 6 O2p^5
17a1 (LUMO)	4.897	27 H^13, 25 H^15, 25 H^21, 13 H^19, 9 Sn5s^1
15b1	5.779	46 Sn5p^1, 38 H^13, 14 H^19

---

Thus far we have only considered molecular cluster models for the stoichiometric hydroxylated SnO₂ surface, whereas in practice SnO₂–0.2%Sb was used as a support. The latter exhibits extrinsic n-type electrical conductivity at room temperature owing to the pentavalent Sb. We have attempted to imitate the extrinsic electrical behaviour of Sb-doped SnO₂ by introducing an extra charge to cluster A. GGA calculations were performed on [Sn₂O₁₀H₁₀]³⁻, allowing full cluster geometry relaxation within the constraints of C_2v symmetry. The excess “conduction electron” was found to occupy the bottom of the “conduction states”, i.e. the former LUMO of mixed tin–hydrogen character (MO 17a₁ in Fig. 13(a)), and the “band gap” between the “valence states” and now partially occupied “conduction states ” was found to be 3.3 eV.

The calculated electronic structure of the fully optimised [Sn₂O₁₀H₈Ru]²⁻ cluster is shown in Fig. 14(b). If we once again define the “band gap ” to be between the HOMO and the LUMO of mixed tin–hydrogen character (18a₁/16b₁), the GGA calculation yields a number of unoccupied gap orbitals induced by Ru–O interaction. The 10a₂ LUMO is now within the band gap, 0.25 eV above the top of the “ valence” orbitals. Table 8 summarizes the composition of the unoccupied gap orbitals lying in the bottom half of the “band gap” (just above the HOMO). The low-lying unoccupied MOs are of Ru–O(surface) character and are largely localized on the surface Ru atom. The results of the gradient corrected calculation are qualitatively similar those of the LDA calculations described above.

As with the unmodified clusters we have introduced an extra charge into the [Sn₂O₁₀H₈Ru]²⁻ system to imitate an extrinsic “conduction electron”. A series of calculations was performed on [Sn₂O₁₀H₈Ru]³⁻, in which the excess charge was first considered to occupy the Ru–O gap orbitals and then the lowest unoccupied tin–hydrogen orbital (as would be the case in the absence of Ru). The total molecular bonding energies (BE) were computed for each electronic configuration. The minimum energy configuration was found to be that with the excess electron occupying the 10a₂ Ru–O LUMO in the gap and, as shown in Table 8, this orbital is largely localized on the ruthenium atom.

Table 8 GGA eigenvalues and atomic compositions for the low-lying unoccupied “band gap” orbitals of [Sn₂O₁₀H₈Ru]²⁻. C_2v symmetry notation is used to label the orbitals. The superscripts on the atomic symbols correspond to their location in the cluster from Fig. 11, and imply all atoms connected by C_2v symmetry

Orbital	Energy/eV	Composition (%)
10a2	2.679	63 Ru4d, 34 O2p^10
15b1	3.595	60 Ru4d, 33 O2p^10

---

(c) The interaction of O₂, CO or OH with the Ru atom of [Sn₂O₁₀H₈Ru]²⁻. The results described above demonstrate that both the LDA and the more sophisticated GGA approach yield qualitatively similar information on the electronic structures of unmodified clusters A–C, and also on the changes to the electronic structures that are induced by the attachment of ruthenium atoms to the clusters. This conclusion, together with the fact that the chemical interaction of an adsorbate molecule with surface Ru is a local interaction involving largely the Ru atom–gas molecule pair, lead us to believe that LDA calculations of the interaction of O₂ , CO or OH moieties with the smallest Ru-modified cluster ([Sn₂O₁₀H₈Ru]²⁻) should be sufficient to describe qualitatively the changes in the electronic structure that are induced by gas adsorption. The calculations described in this section were performed with double-zeta basis sets for all atoms (ADF Type II), and the positions of the surface oxygen atoms bound to the Ru as well as the positions of the Ru atom itself and adsorbent species were reoptimized in each case, all other atomic positions being fixed. In the following descriptions, the Ru atom and the species adsorbed onto it are referred to as the “surface fragment ”, and the surface oxygens bound to Ru are considered separately.

In the first set of calculations, an O₂ molecule with O–O bond distance of 1.2 ų⁸ was placed over the Ru atom with the O–O axis oriented along the x direction as shown in Fig. 15(a). The positions of all atoms of the surface fragment and the surface oxygens were then optimized. The oxygen molecule was then forced to dissociate (Fig. 15(b), the O–O distance was initially set to 3.2 Å) and the minimum energy geometry was computed, again optimising the positions of all atoms in the surface fragment and surface oxygens. In both cases the initial distance from the Ru to the O–O centroid was set to 2.0 Å.

Fig. 15 Schematic diagram showing the interaction of a surface-grafted Ru atom with (a) molecular O₂ and (b) dissociated O₂ (two O atoms).

Table 9 summarizes the LDA bond distances and total molecular bonding energies obtained in the two cases. Interaction of an O₂ molecule with the attached Ru centre gave a local energy minimum with a partially-dissociated geometry in which the O–O distance increased from the 1.2 Å starting value to 1.564 Å. When the initial O–O distance was set to 3.2 Å, the geometry optimised to an O–O distance of 3.29 Å, with a shorter Ru–O distance than in the partially dissociated case. The fully dissociated geometry was found to be 292 kJ mol⁻¹ more stable than that involving partially dissociated O₂ .

Table 9 Bond distances and total bond energies obtained in LDA calculations of O₂ adsorption onto [Sn₂O₁₀H₈Ru]²⁻

O2 arrangement	Ru–O (O2)/Å	O–O (O2)/Å	BE/kJ mol^−1
Partially dissociated	1.944	1.564	−10591
(Fig. 15(a))
Fully dissociated	1.7813	3.294	−10883
(Fig. 15(b))

---

A second series of calculations was then performed using the ADF “lineartransit” facility. In these calculations the O–O distance was initially set to 1.56 Å and the Ru–O₂ centroid distance to 2.0 Å. The latter distance was then gradually reduced, the positions of the Ru and the four O atoms to which it is attached being allowed to optimise at each step. The total molecular bonding energy is plotted against optimised O–O distance in Fig. 16, giving an approximate measure of the energy profile of O₂ dissociation over the surface Ru. Note that the energy of the complex at an O–O distance of 1.66 Å is some 135 kJ mol⁻¹ higher than at O–O = 1.56 Å, strongly suggesting that O₂ dissociation is an activated process. For room-temperature interaction (as in this work) we feel that it is sensible to assume at most partially dissociated adsorption of O₂.

Fig. 16 LDA energy profile for the reaction of O₂ dissociation over the attached Ru atom. Bond energy, referenced to the free atoms, is plotted against O–O distance. Calculations in the region between the points 1 and 2 were not attempted and this region is therefore represented by a dashed line.

Finally, the interactions of CO and OH groups with the Ru centre were studied. Two arrangements were considered in the case of CO. First a single CO molecule was placed over the Ru atom “atop” with the C end down, and secondly two CO molecules were arranged over the surface Ru in the xz plane (Fig. 17(a)). The two CO groups were then replaced by OH units (Fig. 17(b)). Electronic structures were computed for the optimized geometry in each case where the positions of all atoms of the surface fragment and surface oxygens were relaxed.

Fig. 17 Schematic diagram showing the arrangement of (a) CO adsorbent molecules and (b) hydroxy groups over the attached Ru.

The LDA Ru–O bond distances, obtained for the [Sn₂O₁₀H₈]²⁻ cluster with the attached Ru(OH)₂ surface fragment, are compared in Table 10 to those measured by EXAFS for the hydroxylated Ru modified reduced sample (see Results, Section (ii)). The LDA Ru–O bond distances are clearly lower than the EXAFS experimental value. This discrepancy may well reflect the well-known tendency of the LDA to overbinding,³⁷i.e. to produce bond lengths which are short compared with experimental values.

Table 10 Ru–O bond distances obtained by LDA calculations on the {[Sn₂O₁₀H₈]Ru(OH)₂}²⁻ cluster and measured by EXAFS for the hydroxylated Ru-modified reduced sample

Probe	LDA Ru–O (surface)	LDA Ru–O (OH)	EXAFS Ru–O (average)
Ru–O/Å	1.846	1.961	2.047 ± 0.008

---

Calculations on the Ru-modified cluster with either CO, OH or O₂ species adsorbed onto the Ru centre yielded in each case an electronic structure characterized by a “ band gap” (between the highest occupied energy level and the first unoccupied energy level of mixed tin–hydrogen character) of about 1.5–2 eV and a number of MOs lying in the gap. Table 11 summarizes the atomic compositions of the lowest unoccupied orbitals, positioned in the gap Δ eV above the HOMO, for each considered structure of the surface fragment. The surface oxygen atom is marked in bold for convenience. Note that as a result of adsorption of a single CO molecule onto the bare Ru the contribution of the surface oxygens to the lowest unoccupied MO in the “band gap” increased, compared with the CO-free case. Adsorption of molecular oxygen onto the bare Ru site, on the other hand, resulted in further localization of the LUMO on the Ru–O₂ surface fragment.

Table 11 LDA eigenvalues and atomic compositions of the lowest unoccupied energy level of the [Sn₂O₁₀H₈Ru]²⁻ cluster following adsorption of various species onto Ru. C_2v symmetry notation is used to label the orbitals. The non-adsorbate oxygen atoms bound to Ru are indicated in bold

Adsorbent	LUMO	Δ/eV	Composition (%)
Species
None	10a2	0.37	35 O2p, 61 Ru4d
CO	10a2	0.54	42 O2p, 50 Ru4d
CO + CO	17b1	0.29	26 O2p, 29 Ru4d, 7 Ru5p, 7 O2p^adsorbate, 18 C1s, 9 C2p
O2	23a1	0.04	21 O2p, 52 Ru4d, 7 Ru5s, 4 Ru5p, 9 O2p^adsorbate
OH + OH	23a1	0.27	14 O2p, 40 Ru4d, 10 Ru5s, 17 H1s^adsorbate, 7 O2p^adsorbate

---

Table 12 summarizes the bonding energies for the interaction of the surface Ru with each adsorbent species. The bonding energies of the free adsorbent molecules are also given. It follows from Table 12 that, on thermodynamic grounds, one CO molecule should not substitute for an adsorbed O₂ molecule:

Table 12 Total LDA molecular bonding energies (BE) for [Sn₂O₁₀H₈Ru]²⁻ cluster following adsorption of various species onto Ru. The BE for free adsorbent molecules are also given

Adsorbent species	BE/kJ mol^−1	Free species	BE/kJ mol^−1
None	−10243	O2	−828
O2	−11397	CO	−1347
CO	−11788	OH	−695
2CO	−13413
2OH	−12368

---

However an adsorbed O₂ molecule might be substituted by two CO species:

Table 11 indicates that the surface oxygen atoms' contribution to the lowest unoccupied eigenstate slightly increases on substitution of O₂ with two CO.

Two OH groups adsorbed onto the Ru centre would not be displaced, as the arrangement appears to be energetically favoured with respect to O₂ and CO:

Referring again to Table 11, note that in the case of the two OH species bound to the surface Ru centre the maximum degree of localization of the lowest unoccupied gap state on the Ru/adsorbate fragment is obtained.

Discussion

(i) Surface amount of Ru

It is interesting to estimate the fraction of the SnO₂–0.2%Sb surface hydroxy groups which reacted with the Ru complex. The SnO₂ (110) surface was used as a model for this purpose. The unit cell parameters were obtained from Camargo et al.³⁶ as a = b = 4.5 Å and c = 3.0 Å. On this surface, if every exposed oxygen was hydroxylated, the number of surface hydroxy groups would be 1 × 10¹⁵ cm⁻². Assuming that three atomic layers were probed by XPS, a depth of about 10 Å, the total number of lattice oxygen atoms probed by the measurement would be 6 × 10¹⁵ cm⁻². Therefore the surface hydroxy contribution to the total oxygen signal should be about 14%, which is in fact slightly lower than that measured (23%). Therefore, given the approximation in the estimation, it is reasonable to assume a full surface hydroxylation. Thus, the measured surface loading of Ru≈8 × 10¹³ atom cm⁻², means that about 10% of the total surface OH groups had reacted with the Ru complex.

We can also check the correspondence between XPS measurements and the estimate of Ru coverage obtained from the amount of complex reacted, since the number of Sn atoms probed by XPS, based on the (110) model surface, would be 1 × 10¹⁵ cm⁻² per layer. Assuming a measurement depth of three atomic layers implies 3 × 10¹⁵ Sn cm⁻². Based on the amount of complex reacted, assuming that the Ru is only on the top atomic layer, the fraction which is Ru of the total metal atoms probed by XPS therefore would be about 3%, close to the measured value for the initial grafting and for the ‘reduced’ preparation (Table 2). Therefore, substantially all of the Ru atoms were visible to the XPS probe implying that none of the Ru was present in agglomerates more than about 3 atomic layers thick.

(ii) Surface state of the Ru and speculations on structure

(a) Surface-bonded precursor . Assignment of the Ru oxidation state based on the XPS binding energy is highly complex and is largely dependent on the type of ligands, the conditions of the experiment, state of the surface, degree of hydration etc.^39–41 However, Citrin⁴² demonstrated that it is possible to distinguish between Ru(II) and Ru(III) species, based on the XPS line width. He pointed out that the 3d level of the Ru(III) component is broadened by exchange splitting between the 3d and 4d electrons, while exchange splitting is absent for the case of the Ru(II).

We did not detect any 3d peak broadening for the case of the supported Ru species prior to decomposition of the surface-bound precursor. The structure of the Ru 3d core level spectra for the supported complex was almost identical to that of the original dimer [{(η⁶-C₆H₆)RuCl₂}₂]. Furthermore, the binding energy shift of 0.7 eV recorded for the supported species is too small to suggest formation of the Ru(III) surface species. The thermal decomposition studies showed that the supported Ru retained the original aryl unit and that the grafted complex was rather stable, as its decomposition in vacuum took place only on heating to 230°C. It is suggested therefore that the surface grafted Ru species remained in the + 2 oxidation state due to the formation of two direct Ru–O–Sn links, leaving the aryl unit intact. The binding energy shift would then be due to a substitution of the chlorines in the original dimer for more electronegative oxygens in the supported complex.

(b) Surface complex decomposed under H₂. On decomposition of the Ru surface bound complex under H₂, some degree of reduction of the Ru(II) species occurred, recorded as a decrease in the XPS binding energy and a peak broadening. The signal characteristic of bulk Ru metal was not detected and therefore we suggest the formation of small surface bound unsaturated Ru clusters with varying nuclearity. Since only 10% of surface OH groups were used in Ru–O–Sn link formation and Ru was shown to be uniformly distributed over the surface, those clusters should indeed be very small and the formation of monomeric surface bound species cannot be excluded. The assumption of monomeric surface-bound Ru allows interpretation of the TPD results for NO. This gas is known to adsorb mostly dissociatively onto Ru metal surfaces with only a small amount desorbing as molecular nitric oxide.⁴³ A reasonable explanation for undissociated NO on the reduced supported Ru species is therefore that possibly two adjacent sites are required for dissociative adsorption. If the Ru species anchored onto SnO₂–0.2%Sb following the reduction of the surface complex were close to monomeric, then NO could be chemisorbed in an undissociated state, which is what was observed.

(c) Effect of ambient air on the reduced supported Ru. Exposure to dry air caused a formal oxidation of the Ru species, and exposure to water vapour in air caused a further oxidation, as detected by XPS. The main oxidation state for the Ru surface species after water vapour exposure was recorded by XPS as similar to hydrated RuO₂·xH₂O.⁴⁰ Thus, on exposure of small surface clusters to moisture in air the cleavage of any Ru–Ru bond took place, resulting in the formation of hydrated surface bound Ru⁴⁺ centres. Small oxidized Ru centres were revealed by EXAFS on the surface of the reduced sample, followed by long exposure to air at normal ambient relative humidity. Peaks other than the nearest neighbour Ru–O were not observed and the coordination number was lower than that for bulk RuO₂. Thus those oxidized Ru species were obviously formed by oxidation of very small metal clusters or monomeric centres. Furthermore, the Ru–O bond distance of 2.047 Å, measured by EXAFS, was longer than the Ru–O interatomic distance in bulk RuO₂ (1.989 Å). The EXAFS experiment gives averaged information on the local structure. For bulk RuO₂ on the surface, the metal–oxygen (support) interaction would be hidden by strong Ru–O interaction in an oxide particle. In our case, however, the cluster size appears small indeed, which would increase the proportion of the Ru atoms bound to the support oxygen. We believe that the measured elongated Ru–O distance signifies the presence of chemical bonding between Ru species and the support oxygen.^24,25

For the reduced Ru-modified samples exposed to ambient air, XPS demonstrated that Ru forms energy states within the tin dioxide band gap and caused a slight broadening of the valence band edge. Furthermore the resistance in ambient air was shown to be higher than that for an unmodified sample: direct evidence of an electronic interaction between Ru species and tin dioxide support, which could be expected on formation of Ru–O–Sn links to the surface.

(d) Effect of decomposition in air . The XPS spectrum of the supported Ru species formed after the decomposition of the surface complex in air was very similar to that recorded for the reduced moisture treated sample: that is, a formal oxidation state of Ru⁴⁺. Furthermore, a lowering in the Ru surface content was found by XPS, signifying an agglomeration of the supported Ru. Formation of RuO₂ particles whose dimension was greater than the XPS probe depth would reduce the apparent Ru surface area observed by XPS and therefore the evaluated Ru concentration. Formation of RuO₂ particles was confirmed by EXAFS.

(iii) Electrical measurements and correlation with surface state

On exposure of the in situ reduced sample to dry air, a positive base line resistance drift was recorded, characteristic only for the Ru modified sample. XPS studies of the reduced sample revealed partial oxidation of the surface grafted Ru on exposure to dry air recorded as a binding energy shift of the Ru 3d core level. The observed increase in the resistance in dry air is therefore a monitor of variation in the oxidation state of the surface Ru caused by oxygen adsorption. Oxidation of the supported Ru evidently was slow, given the rather gradual resistance drift in dry air.

A further resistance increase was recorded in moist air. The XPS and EXAFS studies showed the formation of fully oxidized Ru⁴⁺ centres after this treatment. Higher resistance values in dry air were recorded for the sample after its exposure to moist air. This can be interpreted as due to the formation of stable Ru species in + 4 oxidation state. Ru(IV) lies deep in the band gap and is a very effective trap state, as is shown by its effect on electronic conductivity.

The gas sensitivity demonstrated for the newly reduced supported Ru samples in nitrogen is believed to arise directly from coordination of gaseous molecules with unsaturated surface bound Ru centres. We have demonstrated that both CO and NO indeed adsorb onto newly decomposed Ru species at room temperature. Three types of adsorbed CO species were identified by TPD, the species desorbing at 75°C being the major type. Desorption of NO was recorded at the higher temperature of 125°C. A resistance decrease was recorded on exposure to both CO and NO in nitrogen. In the case of NO the effect appeared to be almost irreversible: when pure nitrogen was reapplied no increase in the resistance was recorded. For the more weakly chemisorbed CO, however, a partial resistance recovery was observed in nitrogen.

Mizushima et al.⁴⁴ studied CO adsorption onto Al₂O₃ supported Ru clusters by EXAFS and IR spectroscopy and found that the electronic state of the supported Ru was significantly perturbed on CO chemisorption. They observed three different CO absorption bands in IR spectra and assigned those to linearly bound CO, and to symmetric and antisymmetric stretching vibrations of Ru–(CO)₂. They reported the average value for the Ru/CO ratio as 1.3, and demonstrated that the band due to the linearly bound CO disappeared on heating to 90°C. Following Mizushima et al.⁴⁴ it seems that the linearly bound species were the major type of adsorbed CO species on SnO₂–0.2%Sb supported Ru in our study.

On the very first exposure of the reduced supported Ru sample to dry air the loss in gas sensitivity was over 50%. Evidently the reactive gas adsorption was obstructed by molecular oxygen. Chemisorption of molecular oxygen was confirmed by XPS, as a partial oxidation of surface Ru, and also by TPD studies, as an increase in CO₂ evolution.

We postulate that, at room temperature, the electronically active Ru site, responsible for the observed gas sensitive electrical conductivity, is the least strongly bound state labeled “ A” in CO TPD. We propose that oxygen adsorption onto site “A” caused the observed resistance increase, and that CO substitution for oxygen caused the resistance decrease, measured as a gas response. To support this interpretation, we recall that for oxygen-precovered Ru species, the CO desorption signal “A” was significantly diminished, signifying that CO substitution for oxygen is a chemically activated process. Interaction of the grafted Ru centres with moisture resulted in the formation of a stable Ru⁴⁺ hydrated species, thus blocking the CO adsorption site. Consistent with this observation, after exposure to moisture the sample gas sensitivity was lost irreversibly.

Decomposition in air of the surface bound complex resulted in the formation of bulk RuO₂ particles. In this case, the gas sensitivity of electrical conductivity was lost. The interpretation is that, in the bulk particles, the formal oxidation state of Ru was locked in Ru⁴⁺ and the Ru was coordinatively saturated with tightly bound oxygen, giving no means for gas adsorption to perturb the electronic states of the surface-attached particle. The behaviour was that of the unmodified support material.

(iv) Comparison of experimental with computational results

Before beginning this discussion, we emphasise that the molecular cluster calculations are only intended to probe the general qualitative trends in electronic structure modifications caused by Ru grafting and subsequent gas chemisorption onto supported Ru. Nevertheless, we believe that our calculations are useful aids in interpreting the experimental data.

It appeared that in both LDA and GGA calculations and independently of the cluster size, the electronic structure of an unmodified cluster displayed occupied “valence” orbitals built primarily from oxygen 2p atomic orbitals, separated by a HOMO–LUMO energy gap of over 1.5 eV from unoccupied “conduction” orbitals of mixed hydrogen–tin character. Our representation of the electronic structure of the hydroxylated SnO₂(110) surface is quite simplistic, but in qualitative terms it is clearly adequate to describe the absence of the electrical conductivity at moderate temperatures, due to the high excitation energy required to promote the charge carriers from the “valence ” orbitals into the “conduction ” orbitals across a substantial energy gap.

The general features of the electronic structures of the unmodified clusters were retained on Ru grafting. Thus both GGA and LDA calculations on the Ru-modified clusters, independent of the size of the basis set or the cluster size, yield an energy gap of over 1.5 eV between occupied “valence ” orbitals and unoccupied “conduction” orbitals. However, all calculations also gave extra unoccupied MOs positioned in the gap, which are the result of mixing between Ru 4d orbitals and surface O 2p levels. All calculations give qualitatively similar electronic structures within the “band” gap, both in terms of orbital energies and compositions, placing the Ru–O LUMO low in the bottom half of the “band” gap.

We have demonstrated, by introducing an extra charge to the unmodified clusters, that the excess electron occupies the lowest “conduction band” tin-hydrogen based orbital. When the Ru atoms are introduced, the extra charge occupies the lowest MOs in the “ band gap”. These MOs are largely of Ru d character, and it can be said that electrons occupying these orbitals are mainly localized on the attached Ru. The depletion of the “conduction” orbitals of the charge carriers is expected to result in lower electrical conductivity of the model system.

The electronic interaction between the attached Ru and the molecular cluster occurs through surface oxygen atoms [italic v]ia direct Ru–O(–Sn) chemical bonding. It should be noted that although the “band gap ” Ru–O MO is largely localized on the attached Ru atom, it also has partial surface oxygen character, and consequently some of the character of the supporting oxide. Changes in the electrical conductivity of the material could be then predicted by examining the extent of localization of the Ru–O orbitals on the attached Ru. If the contribution of the surface oxygen atoms to the gap MOs increases, the degree of localization of the excess charge (trapped in this orbital) on the grafted Ru would decrease. As a result, charge redistribution over the Ru–O gap orbital would be expected, with partial charge donation from the attached Ru to the surface oxygen atoms. This charge donation would be experienced by the surface Sn cations bound to those surface oxygens, and is expected to be reflected in an electrical conductivity increase. LDA calculations on the Ru-modified cluster showed that gas chemisorption onto Ru centres induced variations in the composition of the Ru–O orbital positioned in the gap, implying charge redistribution. Thus a conductivity increase is expected on adsorption of a single CO molecule onto the clean Ru centre, as the degree of localization of the excess charge on the surface Ru decreases. This is in agreement with the experimental results, where a strong resistance response to CO was recorded in oxygen free conditions.

Dissociation of molecular oxygen over the supported Ru was found to be an activated process with an activation energy of over 1.5 eV. It is believed, therefore, that at ambient temperatures oxygen adsorbs non-dissociatively onto surface Ru centres. This is in agreement with our XPS results, where only slight partial oxidation was recorded on exposure of the supported Ru samples to oxygen in dry air. Our calculations also predict an increase in the localisation of the Ru–O gap MO on the surface fragment on molecular oxygen chemisorption. This would be expected to cause a conductivity decrease, as indeed was observed for the Ru modified material in dry air.

From thermodynamic considerations it appears that an O₂ molecule adsorbed onto the supported Ru would be replaced by two, not one, CO molecules. As a result of such substitution the weight of the lowest Ru–O gap orbital on the surface oxygen atoms was shown to increase slightly. An excess charge trapped onto this state would therefore be a little less localized on the surface fragment, resulting in a small conductivity increase. The resistance response to CO in the presence of oxygen in the dry air, measured experimentally, was actually quite small, compared to that in oxygen-free conditions. Weakening of the resistance response to CO in the presence of oxygen, where Ru adsorption sites are precovered with chemisorbed molecular oxygen, would also be expected from steric considerations. Substitution of the adsorbed O₂ with two CO species would most probably proceed in two stages. In the first instance a single CO would be expected to adsorb onto a vacant Ru coordination site. This should cause some bending of the Ru–O₂ bond due to repulsion of adsorbate species and also weakening of the Ru–O₂ interaction. The second CO could then substitute for O₂. The process would evidently require substantial activation in the first instance and is expected to be rather slow.

As a result of interaction of OH groups with the attached Ru the Ru–O gap MO became strongly localized on the surface fragment. An even greater decrease in the electrical conductivity would therefore be expected, compared to that caused by adsorbed molecular oxygen. Our calculations yield that the adsorption of OH groups onto supported Ru was energetically favoured over molecular oxygen and CO. The adsorbed OH groups therefore inhibit subsequent chemisorption of either O₂ or CO. Experimentally we have shown that exposure to moisture of the Ru modified material resulted in an irreversible resistance increase and the loss of CO resistance response. Computationally, the hydroxylated surface fragment was found to be the most stable arrangement. This arrangement corresponds to a Ru coordination number of 4, with oxidation state of the supported Ru being + 4. This is in agreement with our XPS and EXAFS findings, where a coordination number of 4.3 was found with the supported Ru in the + 4 oxidation state, following contact with moisture in air.

Conclusions

We believe that we have succeeded in surface grafting of unsaturated Ru centres onto tin dioxide [italic v]ia a surface organometallic reaction and that the resultant modification of the electronic structure and electrical conductivity is also characteristic of the formation of chemical bonds. We have demonstrated that changes in electrical conductivity of the sample correspond to variations in the chemical state of the surface-bound Ru species caused by gas chemisorption. We have shown that if the unsaturated surface centres are destroyed by thermal oxidation, the electrical effects are also destroyed. We have demonstrated that quantum mechanical calculations on simple cluster models can be used to interpret the experimentally determined electrical effects.

Acknowledgements

We are grateful to the UK's Computational Chemistry Working Party for a grant of computing time on the EPSRC's “ Columbus/Magellan” central computing facility. This work was supported by EPSRC and Capteur Sensors and Analysers Ltd. Great assistance with EXAFS was rendered by Dr Richard Oldroyd and Dr Gopinathan Sankar (Royal Institution, UK).

References

1. J. F. McAleer, P. T. Moseley, J. O. W. Norris and D. E. Williams, J. Chem. Soc., Faraday, Trans. 1, 1988, 84, 441.
2. N. Yamazoe and N. Miura, in Chemical Sensor Technology, ed. S. Yamauchi, Kodansha, Tokyo, 1992, vol. 4, pp. 19–42..
3. G. S. Henshaw, R. Ridley and D. E. Williams, J. Chem. Soc., Faraday Trans., 1996, 92, 3411.
4. D. E. Williams, in Solid State Gas Sensors, ed. P. T. Moseley and B. C. Tofield, Adam Hilger, Bristol, 1987..
5. D. E. Williams, Sens. Actuators B, 1999, 57, 1.
6. C. Canevali, N. Chiodini, F. Morazzoni and R. Scotti, J. Mater. Chem., 2000, 10, 773.
7. J. F. McAleer, P. T. Moseley, J. O. W. Norris and D. E. Williams, J. Chem. Soc., Faraday Trans. 1, 1987, 83, 1323.
8. R. A. Zelonka and M. C. Baird, J. Organomet. Chem., 1972, 35, C43.
9. T. Arthur and T. A. Stephenson, J. Organomet. Chem., 1981, 208, 369.
10. M. A. Bennett, T. W. Matheson, G. B. Robertson, W. L. Steffen and T. W. Turney, J. Chem. Soc., Chem. Commun., 1979, 32.
11. M. A. Bennett and T. W. Matheson, J. Organomet. Chem., 1979, 175, 87.
12. C. D. Wagner, in Practical Surface Analysis, ed. D. Briggs and M. P. Seah, Wiley, Chichester, 1990, vol. 1..
13. P. H. Citrin, J. Am. Chem. Soc., 1973, 95, 6472.
14. J. B. Metson, M. M. Hyland, A. Gillespie and M. Hemmingsen-Jensen, Colloids Surf. A, 1994, 93, 173.
15. G. S. Henshaw, L. Morris, L. J. Gellman and D. E. Williams, J. Mater. Chem., 1996, 6, 1883.
16. G. S. Henshaw, V. Dusastre and D. E. Williams, J. Mater. Chem., 1996, 6, 1351.
17. G. S. Henshaw, L. J. Gellman and D. E. Williams, J. Mater. Chem., 1994, 4, 1427.
18. G. te Velde and E. J. Baerends, J. Comput. Phys., 1992, 99, 84.
19. ADF〈2.3〉; Department of Theoretical Chemistry, Vrije Universiteit, Amsterdam, 1997..
20. S. H. Vosko, L. Wilk and M. Nusair, Can. J. Phys., 1980, 58, 1200.
21. A. Becke, Phys. Re[italic v]. A, 1988, 38, 3098.
22. J. P. Perdew, Phys. Re[italic v]. B, 1986, 33, 8822.
23. R. S. Mulliken, J. Chem. Phys., 1955, 23, 1833; R. S. Mulliken, J. Chem. Phys., 1955, 23, 1841; R. S. Mulliken, J. Chem. Phys., 1955, 23, 2338; R. S. Mulliken, J. Chem. Phys., 1955, 23, 2343.
24. F. W. Lytle, G. H. Via and J. H. Sinfelt, J. Chem. Phys., 1977, 67, 3831.
25. K. Asakura, K.-K. Bando and Y. Iwasawa, J. Chem. Soc., Faraday Trans., 1990, 86, 2645.
26. R. Sekine and Y. Iwasawa, J. Mol. Catal., 1993, 78, 327.
27. K. Hermann, A. Michalak and M. Witko, Catal. Today, 1996, 32, 321.
28. K. Hermann, M. Witko and A. Michalak, Catal. Today, 1999, 50, 567.
29. M. Witko, K. Hermann and R. Tokarz, Catal. Today, 1999, 50, 553.
30. T. Bredow, A. M. Marquez and G. Pacchioni, Surf. Sci., 1999, 430, 137.
31. T. S. Rantala, V. Lantto and T. T. Rantala, Phys. Scr. T, 1994, 54, 252.
32. T. Bredow and G. Pacchioni, Surf. Sci., 1999, 426, 106.
33. T. B. Shay, L. Y. Hsu, J. M. Basset and S. G. Shore, J. Mol. Catal., 1994, 86, 479.
34. T. Seyller, D. Borgmann and G. Wedler, Surf. Sci., 1998, 400, 63.
35. F. Rittner, R. Fink, B. Boddenberg and V. Staemmler, Phys. Re[italic v]. B, 1998, 57, 4160.
36. A. C. Camargo, J. A. Igualada, A. Beltrön, R. Llusar, E. Longo and J. Andrös, Chem. Phys., 1996, 212, 381.
37. J. Goniakowski, J. M. Holender, L. N. Kantorovich and M. J. Gillan, Phys. Re[italic v]. B, 1996, 53(4), 957.
38. F. Zhang, X. H. Liu, Y. J. Mao, N. Huang, Y. Chen, Z. H. Zheng, Z. Y. Zhou, A. Q. Chen and Z. B. Jiang, Surf. Coat. Technol., 1998, 104, 146.
39. G. Lemay, S. Kaliaguine, A. Adnot, S. Nahar, D. Cozak and J. Monnier, Can. J. Chem., 1986, 64, 1943.
40. K. S. Kim and N. Winograd, J. Catal., 1974, 35, 66.
41. P. H. Citrin and A. P. Ginsberg, J. Am. Chem. Soc., 1981, 103, 3673.
42. P. H. Citrin, Phys. Re[italic v]. B, 1973, 8, 5545.
43. R. Klein and A. Shih, Surf. Sci., 1977, 69, 403.
44. T. Mizushima, K. Tohji, Y. Udagawa and A. Ueno, J. Phys. Chem., 1990, 94, 4980.

---