Ir-Catalysed Nitrous oxide (N2O) Decomposition: Effect of Ir Particle Size and Metal–Support Interactions

The effect of the morphology of Ir particles supported on γ-Al2O3, 8 mol%Y2O3-stabilized ZrO2 (YSZ), 10 mol%Gd2O3-doped CeO2 (GDC) and 80 wt%Al2O3–10 wt%CeO2–10 wt%ZrO2 (ACZ) on their stability on oxidative conditions, the associated metal–support interactions and activity for catalytic decomposition of N2O has been studied. Supports with intermediate or high oxygen ion lability (GDC and ACZ) effectively stabilized Ir nanoparticles against sintering, in striking contrast to supports offering negligible or low oxygen ion lability (γ-Al2O3 and YSZ). Turnover frequency studies using size-controlled Ir particles showed strong structure sensitivity, de-N2O catalysis being favoured on large catalyst particles. Although metallic Ir showed some de-N2O activity, IrO2 was more active, possibly present as a superficial overlayer on the iridium particles under reaction conditions. Support-induced turnover rate modifications, resulted from an effective double layer [Oδ−–δ+](Ir) on the surface of iridium nanoparticles, via O2− backspillover from the support, were significant in the case of GDC and ACZ.


Introduction
N2O is a powerful greenhouse gas (global warming potential ~300) and also the dominant ozonedepleting gas of the twenty-first century [1,2]. Stationary or mobile combustion sources and certain chemical processes, are the main sources of anthropogenic N2O emissions, whose abatement is therefore of global importance. In this context, the catalytic reduction of N2O to N2 is a most promising approach [3].
A wide range of materials have been investigated as catalysts, including noble metals, transition metals, mixed oxides, perovskites and zeolites [4][5][6][7][8]. Among them, noble metals (NMs) exhibit good activity at intermediate temperatures so despite their high cost they are promising candidates for practical implementation [3,4]. Rh exhibits the best catalytic activity, although Ru and Ir, significantly cheaper, are almost as good [7,9]. However, a drawback of both Ir and Ru nanoparticle catalysts is their well known inferior thermal stability compared to Rh and Pt [10].
We have recently shown that the thermal sintering characteristics of Ir are strongly correlated with the lability of lattice oxygen in the oxide supports [11] and that metal-support interactions between Ir and oxide supports having high oxygen ion lability can strongly stabilize Ir nanoparticles at high temperatures under oxidizing conditions. It is well known that such supports can also strongly modify the electronic state of the active phase interfaced with them, and, depending on the electrophilic orphobic character of the reactants, promote or inhibit the catalytic chemistry [12,13]. Here we clarify the influence of these phenomena on the direct decomposition of N2O by means of well characterized supported Ir catalysts whose particle size and morphology were controlled by aging. Specifically, we extend the general applicability of our model for the inhibition of sintering to include the important case of Ostwald Ripening. Furthermore, we report the first observation of a counter-clockwise hysteresis phenomenon in light-off curves of Ir-catalysed N2O decomposition and discuss their origin.

Materials preparation
The support materials were γ-Αl2Ο3 (Engelhard), 8 mol% yttria stabilized zirconia, YSZ (Zirconia sales, UK Ltd.), 10 mol% gadolinia doped ceria, GDC (Anan Kasei Co, Ltd.) and laboratory made by co-precipitation ACZ (80wt% Al2O3-10wt% CeO2-10wt% ZrO2 mixed oxide). All were pre-treated in air at 800 o C for 2h. Ir-dispersed catalysts, Ir/γ-Αl2Ο3, Ir/YSZ, Ir/GDC, and Ir/ACZ, were prepared by wet impregnation of the supports at 75 o C in a solution of IrCl3·H2O (Abcr GmbH & Co.KG) followed by reduction in H2/He flow to yield 1.0 wt% nominal Ir loading. The actual metal loading was determined by induced coupled plasma optical emission spectroscopy (ICP-OES). After water evaporation, suspensions were dried at 110°C for 12 h and the resulting powders heated at 400°C for 2h under 50% H2/He flow and then at 850°C for 1h under 1% H2/He flow (hereafter denoted as fresh catalysts). This produced a very narrow Ir particle size distribution (1-2 nm), unlike other methods in which an oxidation step (instead of a reduction) is used at this stage leading to a broad Ir size distribution (from ~1 to > 150 nm) [9,14]. In addition, this prolonged H2 treatment allows the effective removal of residual chlorine resulting from the IrCl3 precursor used. In order to study the susceptibility of our catalysts toward sintering and the consequent influence on their de-N2O activity, pre-aged counterparts of the four catalysts were prepared from the fresh samples using the following protocol: two consecutive cycles of 1 h oxidation (20%O2/He, 50 cm 3 min -1 , 750 o C) followed by 0.5h reduction (2%H2/He, 50 cm 3 min -1 , 750 o C)" (hereafter referred to as aged catalysts). Catalytic testing and stability measurements were carried out in a continuous flow apparatus equipped with a single-pass, tubular, quartz, fixed-bed reactor operated in both differential and integral modes. Catalyst bed temperature was measured via a centrally located K-type thermocouple.

Materials characterization and catalytic evaluation
Reactor inlet and outlet gases were analysed by on-line gas chromatography (Shimatzu GC-2014, TC detector, He carrier, MS 5A and HayeSep D columns). Nitrogen and oxygen mass balances on the basis of the reactor inlet N2O concentration and the outlet N2 and O2 concentrations measured by online GC were found to be better than 98% in all cases. The feed composition was always 1000 ppm(v) N2O balanced with He. Evaluation of catalyst performance as a function of temperature (light-off profiles) was carried out at 300-550 o C with a total feed flow rate of Ft=150 cm 3 min -1 ; differential reactor operation for intrinsic activity evaluation (i.e., turnover frequency), was performed maintaining N2O conversion below ~15%.

Textural, structural and other characteristics of the materials
The OSCs of the four supports and associated characteristics, are shown in Table 1. In accord with the literature [15,16], γ-Al2O3 and YSZ showed negligible and very small OSC of 0 and 5.7, respectively, whereas ACZ and GDC exhibited substantial OSC of 116 and 185.5 μmol O2/g, respectively ( Table 1). The higher capacity of GDC was expected as it contains a higher proportion of CeO2 (~90%) than does ACZ (~10%). Both GDC and ACZ showed two broad peaks in their H2-TPR profiles [17]. For ACZ these peaks are located at ~630 o C (larger) and ~900 o C (smaller); For GDC the peaks appeared at the same temperatures but the high temperature peak was much larger than the low temperature one ( Table 1). Note that the redox (H2-consumption) behavior of ACZ was activated at a significantly lower temperature (~300 o C) than for GDC (~500 o C). The low temperature peak of CeO2-based materials is attributed to the reduction of small surface particles whereas the high temperature peak is assigned to the reduction of bulk CeO2 [17]. We thus conclude that in the ACZ composite the CeO2-ZrO2 species are preferentially located at the surface of the γ-Al2O3 particles as a capping layer. As we shall see, both the magnitude of the OSC and its detailed characteristics have important implications for N2O decomposition catalysed by Ir particles supported on ACZ and GDC.
Textural characteristics of the supporting materials and of the fresh and aged catalysts are summarized in Table 2. Only marginal differences in the total surface areas, SBET, resulted from incorporation of Ir onto the supports. ACZ was an exception; the pronounced sensitivity of its surface area to high temperature treatments is well known [16,18]. The supports span a wide range of SBET values, with YSZ lowest (ca. 5 m 2 /g) and γ-Al2O3 highest (178 m 2 /g). High temperature aging resulted in only marginal changes ( Table 2).   The particle sizes derived from H2-chemisoprtion were very close to those estimated from PXRD data Scherrer analysis, although in the case of very small particles Ir were not detectable by PXRD.
On the other hand, pre-and post aging particle sizes obtained from HRTEM are in broad agreement with those obtained from the other two methods; the apparent discrepancy in the cases of aged Ir/γ- The susceptibility to sintering of Ir particles on γ-Al2O3 under oxidizing conditions due to the high volatility of IrO2, is well known [10]. In contrast, metallic Ir particles are very stable during high temperature treatments in reducing atmospheres [10,11]. The key point arising from our results is that the strong resistance to sintering in an oxidising environment of Ir particles on supports with high oxygen ion lability (GDC and ACZ) is dramatically different from their vulnerable behaviour on supports with negligible or low oxygen ion lability (γ-Al2O3 and YSZ). This is clearly demonstrated in Figure 2, which depicts the percentage of sintering  We have previously proposed a plausible model to explain the effect of the support on catalyst sintering characteristics [11] summarized briefly below (Fig. 3). The model is further expanded in the context of the present work. An effective electrical double layer, [O δ--δ + ](M), forms on the catalyst nanoparticle (M) due to lattice oxygen spillover from supports with substantial oxygen ion lability [12,13]. It should be noted here that the aging protocol followed involves temperatures much greater to the Tammann temperature of IrO2, which is ~413 o C [19], and is associated with the onset of mobility of metal particles on the surface [20]. Thus small iridium nanoparticle possessing a surface oxide overlayer is expected to be, in principle, highly mobile (therefore prone to sintering) on the surface of all oxide supports investigated in this study, as indeed confirmed experimentally for iridium nanoparticles dispersed on γ-Al2O3 [10]. However, the effective double layer [O δ--δ + ], created on supports with high oxygen ions lability ( Fig. 3a), endows the particle with a net negative charge at its gas-exposed surface, leading to interparticle repulsion at short range, thus preventing particle-particle encounters followed by coalescence.
Such spillover effects are expected to be strongest for supports with the highest oxygen ion lability.
In the present case this effect could operate with both metallic Ir 0 and IrO2 nanoparticles or composites of the two, since IrO2 is an electronic conductor. In the case of very small Ir particles, labile lattice oxygen creates vacancies that act as traps which inhibit diffusion and agglomeration (Fig. 3b) under a concept consistent to that described in [21,22]. It is also expected that the O δmodified surface barrier increases the activation energies for detachment (diffusion) and re-attached of the entities whose transport would otherwise lead to continuous growth of large particles at the expense of smaller particles according to the Ostwald Ripening model (see Fig. 3). All aforementioned effects will be negligible for low-lability supports (e.g. Al2O3 and YSZ), which therefore do not prevent particles agglomeration whatever the sintering mechanism is (large particles diffusion and coalescence or Ostwald Ripening). It is worth noting that pure CeO2 has limited lattice defects concentration in comparison to Gd2O3 or ZrO2-doped CeO2. Therefore the model suggests that optimal supports against sintering are those with sufficiently high O 2ions mobility i.e. CeO2 modified with cations such as gadolinium, zirconium and others, that enhance ceria lattice defects, increasing O 2vacancies and consequently O 2mobility [16,18].  [23,24], whilst Ir is one of the most active catalysts for N2O decomposition from the Pt-group metals family [9]. Starting with a pre-reduced catalyst a significant counter-clockwise hysteresis during heating-cooling is clear for all samples.

N2O decomposition
Holding the sample at maximum temperature until a true steady state was reached caused the system to follow a higher activity, and subsequently reproducible, curve upon cooling (Fig. 4a). This hysteresis indicates changes in the catalyst state, further demonstrated in Figure 4b, which depicts the time-on-stream stability of all fresh and aged catalysts at 450 o C. Starting with pre-reduced catalysts, in every case N2O conversion increased progressively with time-on-stream, stabilizing at maximum activity after ~5-7 h (for Ir/γ-Al2O3 and Ir/YSZ) and ~10-12 h (for Ir/GDC and Ir/ACZ). Since at ~450 o C particle growth is not favoured [10], the increase in catalytic activity may be attributed to progressive oxidation of the initially metallic particles under the mildly oxidizing conditions (1000 ppm N2O), leading to formation of IrO2 or a superficial overlayer of IrO2, likely the catalytically active phase, as proposed by Ohnishi et al. [14]. To avoid pseudo steady-state, all subsequent catalytic measurements for both differential and integral modes of reactor operation were carried out on catalysts that had been stabilized after sufficient time-on-stream. Table 4 lists the temperatures for 20, 50 and 80% conversion (T20, T50 and T80, respectively) by the four catalyst pairs, obtained from the corresponding steady state light-off profiles. Note that T50 values of aged samples are close to those of fresh samples; and unexpectedly in the case of Ir/γ-Al2O3 the T50 value of the aged sample is lower than that of the fresh sample-i.e.
aged Ir/γ-Al2O3 was more active that its fresh counterpart. As we shall see, this behaviour can be rationalized by taking account of the significant particle size dependence of the intrinsic turnover activity of Ir-catalysed N2O decomposition.
Note also that the activities of the fresh catalysts, which have quite similar particle sizes, are in the order Ir/-Al2O3>Ir/YSZ>Ir/GDC>Ir/ACZ, as judged by their T50 values (Table 4). This sequence is almost the exact opposite to the OSC sequence of the supports (γ-Al2O3 < YSZ < ACZ < GDC) pointing to the presence of metal-support interactions: the higher the oxygen ion lability of the support the higher the inhibition of the de-N2O activity of the catalyst. In passing, we note that the inverted order between GDC and ACZ is understandable, because although ACZ has lower oxygen storage capacity than GDC, its redox properties are activated at significantly lower temperature (300 o C versus 500 o C; Table 1). Accordingly, in the temperature interval 350-550 o C, within which the N2O decomposition reaction was investigated, the oxygen lability of ACZ actually exceeds that of GDC.
In passing, we note that in the temperature range used all the supports were essentially inactive towards N2O decomposition. In accord with this, there was no correlation between the BET areas and catalyst activity. Differential reactor operation (<15% N2O conversion) at 420 o C provided turnover frequency (TOF) values for the fresh and aged catalysts as shown in Table 4. Striking differences between the low and high oxygen ion lability supports is obvious (Table 4). In particular, aged Ir/γ-Al2O3 exhibits a much higher TOF than its fresh counterpart (ca. 48 times higher). TOFaged/TOFfresh is also substantial for Ir/YSZ (~7), whereas only small differences were observed for the Ir/GDC and Ir/ACZ catalyst pairs. This is consistent with the aforementioned sintering behaviour of the catalysts and the known particle size dependence of Ir catalysts in N2O decomposition: large particles are more active than small ones [14] as we also find here (Table 4). This is also consistent with the view that the influence of the particle size on the de-N2O activity of Ir-based catalysts is far more important than effects due to metal-support interactions. This is apparent from comparison of the TOFaged/TOFfresh ratio values of Ir/γ-Al2O3 and Ir/YSZ with that of Ir/GDC and Ir/ACZ, suggesting that the O δlayer created on catalyst particles interfaced with intermediate or high oxygen ion lability supports acts to inhibit de-N2O activity. According to Vayenas [12] this effective O δlayer (Fig. 3) is expected to weaken the chemisorption bond of electron acceptor adsorbates such as N2Oads, and Oads arising from N2O dissociation. To a first approximation, weakening the chemisorption bonds of both N2Oads and Oads should have opposite effects on the reaction rate in terms of the well-established mechanism for N2O dissociation over noble metals: N2Oads → N2 (g) + Oads (1) 2Oads → O2 (g) Weakening both the N2Oads and Oads chemisorption bonds by oxygen spillover should inhibit reaction (1) whilst also promoting reaction (2). In the present case, the first effect predominates.

. CONCLUSIONS
Ir nanoparticles on a variety of supports show thermal sintering characteristics that are very strongly correlated with the lability of lattice oxygen in the supporting oxide material: the higher the lability of oxygen in the support, the lower the vulnerability of the nanoparticles to sintering under oxidative conditions. Oxygen back-spillover form the support to iridium particles causes this resistance to sintering. The same species are responsible for the appearance of metal-support interactions, which cause inhibiting effects on the de-N2O activity of the iridium particles. These effects are however of significantly less importance than particle size effects on the structure sensitive de-N2O reaction. The catalytically active phase in direct N2O decomposition by Ir-based catalysts is IrO2 or a superficial overlayer of IrO2.