Choosing a suitable support for Co 3 O 4 as an NH 3 oxidation catalyst †

Co 3 O 4 as an NH 3 -oxidation catalyst may transform reversibly to CoO under reaction conditions, even in the presence of excess oxygen. The use of alumina may then result in the formation of cobalt aluminate rendering the catalyst inactive. The formation of cobalt aluminate can be avoided by using ZnAl 2 O 4 as a support.


Introduction
Ammonia oxidation yielding NO/NO 2 forms the core of the nitric acid production plant. This process typically uses a platinum-rhodium gauze as a catalyst for ammonia oxidation, since it is active and highly selective. 1 However, the lifetime of the platinum gauze is rather short (up to 12 months for operation at atmospheric pressure, and the lifetime decreases with increasing reaction pressure). 2 Thus a variety of transition metal oxides for the oxidation of ammonia yielding nitrogen oxides have been studied. 3 Co 3 O 4 is utilized in some nitric acid plants 4 and seems to have the added advantage of lower yield of the undesirable greenhouse gas N 2 O and longer catalyst lifetime.
Ammonia oxidation is thought to proceed via a Mars-van Krevelen mechanism utilizing lattice oxygen. 5,6 Hence, the catalytically active phase in Co 3 O 4 -catalyzed ammonia oxidation may be reduced to CoO, and this phase transformation may result in catalyst deactivation. The more stable perovskite LaCoO 3 has been investigated [6][7][8] as an alternative. This may however result in a lower intrinsic catalytic activity, due to its resistance to the release of lattice oxygen. Another alternative would be to adapt the reaction conditions (oxygen partial pressure/temperature/crystallite size of Co 3 O 4 -vide verde) to minimize the reduction of Co 3 O 4 under the operating conditions.
The removal of lattice oxygen and re-oxidation are crucial steps in the Mars-van Krevelen mechanism. 5,6 The completion of the catalytic cycle requires the oxidation of CoO, viz. 3CoO + 1 2 O 2 -Co 3 O 4 . The thermodynamic feasibility of this process is dependent on temperature, partial pressure of oxygen and the crystallite size of Co 3 O 4 in the starting material (due to the difference in the surface energy contribution 9 ) and in the absence of sintering the thermodynamic feasibility of the re-oxidation process can be expressed as: In industrial operation, Co 3 O 4 as an ammonia oxidation catalyst is typically used in the form of catalyst pellets resulting in severe mass transfer limitations. 15 The use of smaller pellets in conjunction with the high linear velocity typically employed in the ammonia oxidation would result in a high pressure drop. Hence, it is desirable to incorporate the catalytically active component in a structured reactor (e.g. monolith). 8,16 The strong hydrothermal conditions prevalent during ammonia oxidation may result in the formation of the Co-support compound. 17,18 The formation of such compounds would represent a thermodynamic sink. In this communication, we evaluate the activity of Experimental g-Al 2 O 3 (Degussa; d pellet = 3 mm; S BET = 225 m 2 g À1 ; V pore = 0.63 cm 3 g À1 ; d pore = 9 nm) and zinc aluminate were used as support materials. Zinc aluminate was prepared by co-precipitation. 19 The resulting product was pressed into pellets with an average size of 3 mm. An industrial, unsupported Co 3 O 4 catalyst (r pellet = 3.54 g cm À3 ; S BET = 0.3 m 2 g À1 ; V pore = 0.12 cm 3 g À1 ) containing an irregular distribution of Co 3 O 4 crystallites ranging from 200 nm to 5 mm, with the smaller crystallites attached to the large crystallites, was used for comparison.
The support pellets were impregnated with a solution containing Co(NO 3 ) 2 Á6H 2 O in deionised water. 20 The catalyst precursor was aged at room temperature for 20 minutes and dried at 120 1C for 2 hours. Subsequently, the dried precursor was calcined in air (180 ml (STP) min À1 ) at 350 1C for 2 hours (heating rate: 5 1C min À1 ). Catalyst pellets were crushed to a size between 125 and 212 mm for the activity test.
The Co 3 O 4 loading was verified using AAS-ICP. The BET surface area and micro-pore volume was determined using N 2 adsorption-desorption at 77 K using a Micromeritics Tristar 3000. The phase composition on the catalyst and the average crystallite size of the various phases in the catalyst pellet were determined using X-ray diffraction (Bruker D8 Advance laboratory X-ray diffractometer; source Co-K a,1 ; voltage: 35 kV; current: 40 mA) equipped with a position sensitive detector (VANTEC-2000, Bruker AS). The obtained diffraction patterns were fitted using Rietveld refinement as employed in TOPAS 4.2 (Bruker AXS), such that the weighted profile factor (R WP ) was o10 and the Bragg factor (R Bragg ) was o5.
The catalytic activity of the catalysts in the ammonia oxidation was determined in a quartz, fixed bed reactor (d inner = 9 mm; d outer = 12 mm, l = 410 mm) with an isothermal zone of 20 mm (DT o 5 1C). The catalyst (d p = 125-212 mm; m = 3 mg) was diluted with ca. 300 mg silicon carbide (d p o 75 mm) and loaded into the isothermal zone. The high degree of dilution (m diluent / m catalyst = 100) may result in an inhomogeneous distribution of the catalyst and bypassing of the reactants, 21 however the use of fines as a diluent mitigates this effect, due to extensive radial mixing. A bed of silicon carbide (d p = 425-600 mm; l = 125 mm) served as a pre-heating zone. A quartz thermo-well (d outer = 4 mm) was inserted into the catalyst bed to monitor the reaction temperature. A pre-mixed gas containing 7.4% NH 3 , 19.4% O 2 and the balance He was passed over the catalyst bed at a volumetric flow rate of 100 ml (NTP) min À1 . Ammonia oxidation was performed at atmospheric pressure in the temperature range 450-800 1C. The reactor temperature was raised to 450 1C and after 30 minutes the ammonia conversion was determined by bubbling the effluent gas for 20 minutes through deionised water ensuring complete absorption of ammonia. Subsequently, the reactor temperature was raised to investigate the activity at higher temperatures, viz. 580 and 740 1C. In order to test the reversibility of the observed deactivation the reactor temperature was subsequently decreased to 580 1C and 450 1C. The concentration of ammonia in the ammonia trap was determined at each reaction temperature spectro-photometrically using the Nessler method 22 (Jenway 6405 UV/Vis spectrometer; l = 450 nm). The repeatability of the experiments was good and deviations in the ammonia conversion on repeat runs were less than 2%.
The small amount of catalyst used to evaluate the activity does not allow for a post-mortem analysis of the spent catalyst. Hence, the catalyst was used in the form of pellets to enable the characterization of the 'spent' catalyst. The use of pellets introduces mass transfer limitations, 15 which should be kept in mind when interpreting the data. The catalyst pellets (300 mg) were loaded in the isothermal zone of the reactor. The void spaces were filled with ca. 300 mg silicon carbide (d p o 75 mm) to ensure sufficient radial mixing.

Results and discussion
The unsupported Co 3 O 4 as a powder shows the normal increase in the ammonia conversion with increasing temperature (see Fig. 2). A further increase in the ammonia conversion is obtained upon the subsequent decrease in temperature. In order to explore this phenomenon better an unsupported Co 3 O 4 pellet was loaded (which showed significant internal mass transfer limitations 15 ). The ammonia conversion increases with increasing temperature up to 640 1C, after which a decrease in the activity is observed. XRD analysis on the catalyst pellet after exposure to a reaction  crystallites with an average size of ca. 8.4 nm (see Table 2). The ammonia conversion decreases with increasing temperature even going from 450 to 580 1C. The NH 3 conversion at a reaction temperature of 740 1C was equal to the ammonia conversion in the absence of the catalyst. Lowering the reaction temperature did not restore the catalyst activity.
The spent catalyst, both in pellet and in powder form, was dark blue in colour indicating the presence of cobalt aluminate in the sample. The formation of cobalt aluminate cannot be confirmed beyond doubt from the XRD analysis of the spent catalyst due to strongly overlapping diffraction peaks. Tentatively, ca. 10% of cobalt is present as XRD-visible cobalt aluminate. Furthermore, ca. 20% of cobalt has been transformed into XRD-invisible cobalt (possibly amorphous cobalt aluminate).
The formation of cobalt aluminate requires the presence of CoO, 17,18 which may react with alumina under hydrothermal conditions yielding cobalt aluminate. According to our thermodynamic analysis, neglecting the effect of crystallite size (which may become substantial in this size range 24  The rapid diffusion of divalent cobalt into the alumina phase as demonstrated for g-Al 2 O 3 may lead to a cobalt-poor aluminate, 17 i.e. formation of XRD-amorphous cobalt aluminate with cobalt to aluminum ratio of less than 0.5. The location of the cobalt aluminate phase is currently unknown, but it may be associated with the cobalt oxide phase as suggested for the cobalt silicate formation, 25 since the reaction will also take place at the interface between CoO and alumina. This is further substantiated   The amount of XRD-invisible cobalt in the catalyst pellet is drastically reduced with ca. 10% of the cobalt remaining XRD-invisible. This is accompanied by an increase in the average crystallite size of Co 3 O 4 . The increase in the crystallite size can partially be explained by the reduction in the amount of XRD-invisible cobalt. Well-dispersed cobalt may migrate resulting in crystal growth. However, the change in the average crystallite size cannot solely be explained by the reduction in the amount of XRD-invisible cobalt indicating that some sintering is taking place as well. The use of large crystallites of Co 3 O 4 on a support would allow operation at a higher reaction temperature without the formation of CoO. High temperature may lead to a reversible type of deactivation (as observed for unsupported Co 3 O 4 and pellets of Co 3 O 4 /ZnAl 2 O 4 -see Fig. 1), since the formation of cobalt aluminate from CoO and zinc aluminate is thermodynamically not allowed, 26 even for nano-sized CoO crystallites if the surface energy of CoO is smaller than the surface energy of ZnO (g ZnO 27 = 1.42 J m À2 ) assuming that the exchange reaction is not associated with simultaneous crystallite growth. The lack of incorporation of cobalt into the support makes zinc aluminate an ideal material as a washcoat to be applied in structured reactors or on foams, on which nano-crystallites of Co 3 O 4 can be deposited. The formation of cobalt aluminate is intrinsically linked to the ability to form CoO under reaction conditions. 18 The thermodynamic driving force for the formation of this phase can be linked to the crystallite size of the catalytically active phase. At this stage it is not clear whether the reduction or the re-oxidation is the size dependent step. This can only be evaluated using catalysts with well-defined crystallite size distributions. 28 Supporting these well-defined crystallites on the selected support materials would allow further insights into the deactivation behavior of this type of catalysts.

Conclusions
Co 3 O 4 -containing ammonia oxidation catalysts may undergo a reduction to CoO in the presence of oxygen at high temperature. Nano-sized Co 3 O 4 crystallites are more susceptible to the reduction reaction in the presence of oxygen, if the surface energy of CoO is lower than the surface energy of Co 3 O 4 . The reduction of Co 3 O 4 represents a reversible deactivation, since lowering the reaction temperature will result in the re-oxidation of CoO. However, CoO can further react under the hydrothermal conditions of ammonia oxidation with support materials, such as alumina, yielding cobalt aluminate, which may adhere to cobalt oxide reducing the catalytically active surface area accessible to the reactants. The reaction between CoO and ZnAl 2 O 4 is thermodynamically not allowed, making zinc aluminate an ideal support material for Co 3 O 4 as a catalyst for ammonia oxidation.