The specific surface area of the supports before and after modification and its point of zero charge are presented in Table 1.

Figure 1 presents the pore volume distribution as a function of pore diameter for the bare and modified support. For the bare support, a bimodal distribution was observed with an average pore size of 67 Å. The modification of the surface by phosphoric acid provokes a general increasing of the average size of the pores to 110 Å which results in higher pore volume but in lower surface caused by the blocking of some mesopores.

The strength of the interaction of a Lewis base (PO₄³⁻) with the hard Lewis acid sites characteristic for alumina depends on the electron density and the polarizability of the base [26]. Lower polarizability and higher electron density lead to stronger interaction. The strength of interactions of the bases is as follows: Phosphate > fluoride > citrate > sulfate > acetate > formate > nitrate > chloride

The biggest effect on the surface and, which is more important, the biggest possibility to enhance interaction modifier/alumina, should be observed when the phosphoric acid is applied and this was confirmed by the observation of significant decrease in the point of zero charge (Table 1).

The characterization by IR spectroscopy of the modified support compared to bare unmodified alumina (Figure 2(A,B)) reveals a new band at 1100 cm⁻¹ attributed to the P–O groups [27,28].

As claimed in the literature [29], the anions can be adsorbed in an oxide–water interface in an inner- or outer-sphere mode. Inner-sphere surface complexes are created as a result of direct chemical bond formation between adsorbing anion (Lewis base) and the metal ion at the surface (Lewis acid). In the case of anions with more than one donor atom (e.g., phosphate), several coordination geometries are possible: monodentate, chelating bidentate, or bridging bidentate. Other -sphere surface complexes are formed as a result of electrostatic attraction of opposite charges. Busca et al. [30] reported a selective interaction of the phosphoric acid on the basic sites which excludes the formation of H-bonded phosphate species, preventing the formation of the two types of bidentates. The formation of only ionically bonded species can be envisaged (Scheme 1, row 1).

One can suppose that in this case, the biggest part of the phosphates is chemically bonded but the formation of the hydrogen bonded phosphates cannot be excluded. However, the phosphoric acid treatment does not provoke any chemical changes and the only detectable phase by XRD was γ-Al₂O₃ and no formation of aluminium phosphate was observed (Figure 3(A,B)).

However, the acid treatment of the support provokes changes in its morphology as could be seen from the SEM images (Figure 4(A,B)).

The bare alumina presents a heterogeneous morphology, composed of a mixture of grains and plates, which became rather laminar and ordered after acidic treatment but seemed to lose a part of its crystallinity.

The deposition of the ionic liquids depends on the specific surface of the support [9,31]. A high specific surface area provides a large number of adsorption sites and will increase the quantity of the adsorbed ionic liquid. De Castro et al. [9] have found a 20 weight percent maximal value for the quantity of deposited ionic liquids adsorbed on the gamma alumina surface. However, in this study the modification of alumina provides additional adsorption sites resulted in 50 wt% maximal ionic liquid loading. Nevertheless, this sample, was not be used since the possibility of the formation of multiple layers of ionic liquid on the support are higher with the increase of the deposited amount. The lower the quantity of ionic liquid adsorbed, the higher the possibility of a good anchoring and subsequent fine dispersion of the gold nanoparticles. That is why a sample with 20 wt% deposited ionic liquid was chosen for the study.

The specific surface area changes dramatically when the ionic liquid is supported, a value of one tenth being obtained (Table 1). However, the outgas treatment provokes ionic liquid melting and the pores filling, which results in a lower surface area and a quasi inexistent pore volume.

As the ionic liquid did not possess a crystalline structure, no structural changes have been observed (Figure 3). However, heterogeneity of the surface was observed. In some places, the morphology seems to change and to recover the initial morphology of the non-treated alumina (Figure 4(C)) while, in other parts, it keeps the morphology of the modified alumina (Figure 4(D)). This interesting observation led us to suggest that the part of the support surface is liberated from the weakly bonded phosphates through the H-bond and which are now bonded to the ionic liquids as a counter ion.

The gold amount was fixed to 0.8 wt% and verified by X-ray fluorescence. The specific surface area and its pore volume of the gold prepared samples are listed in Table 1. When citrate was used for the gold reduction, the sample recovers the specific surface area of the initial alumina indicating the complete liberation of the surface by ionic liquid loss during the deposition procedure.

The use of NaBH₄ as a reducing agent at room temperature in gold solutions, no matter the solvent, increases the surface and pore volume compared to those of the supported ionic liquid system demonstrating the presence of the ionic liquid in the samples. As expected, the lower surface area of the sample prepared in acetone solution indicates the higher amount of the ionic liquid present in the sample.

In aqueous media, no matter the reducing agent, the diffractions of metallic gold have not been observed (Figure 3(D,E)). This could be provoked either by the lower particle size or by the low quantity of the gold present in the sample (less than 1%).

Nevertheless, the sample prepared in acetone solution clearly presents the diffraction lines attributed to metallic gold suggesting far bigger particle size. In addition, the diffraction lines of the gamma alumina become more stretched and intensive and a new diffraction peak appears at 2Θ = 32° (Figure 3(F)). This line has been reported by De Yu et al. [32] for calculated gamma alumina XRD pattern, although, to our knowledge, no experimental XRD pattern has yet been reported. The utilization of acetone in combination with a strong reducing agent seems to influence the crystallization of gamma alumina.

The gold particles size has been studied by Transmission Electron Microscopy (Figure 5).

In aqueous solutions the change of the reducing agent from citrate to NaBH₄ did not seem to influence the gold particle size. However, it should be noted that the restrictions of the microscope did not permit observation of particles smaller than 3 nm. The use of water as a solvent resulted in lower gold particle sizes at around 6.5 nm for both samples. The change to acetone provokes an increase of the particle size to 9.5 nm in average but the agglomerates with average size higher than 40 nm have been observed and excluded from the particle size distribution. For the samples prepared in the gold solutions at room temperature (Figure 5(B,C)), the presence of a fine organic layer can also be observed, which confirms the presence of IL.

The IR study of the supported ionic liquid compared to the IR spectra of the synthesized unsupported ionic liquid is presented in Figure 2(C,D). For the IL only, the bands at 3146, 3090 and 3017 cm⁻¹ are attributed to the ν(CH) vibrations of C₄, C₅ and C₂ (as shown in Scheme 1, on the right) of the imidazolium ring and those at 2962, 2935 and 2874 cm⁻¹ corresponds to ν(CH) of butyl and methyl groups bonded to the ring [33,34]. On the base of theoretical calculations of IR spectra of an isolated gas phase methyl-ethyl imidazolium ion, the ν(C-H) stretching of the C₂ atom appears at 3118 cm⁻¹ [34]. Moreover, it was evidenced that the frequency of this band strongly depends on the counter ion nature, being the charge to volume ratio, the most important factor influencing the shift of this band. For instance, a shift from 3118 cm⁻¹ to 3049 cm⁻¹ was detected when chloride acts as a counter ion, while this band appears at 3063 cm⁻¹ if chloride is replaced by bromide. In our case, the counter ion is methanesulfonate, with oxygen atoms able to interact with the imidazolium. Taking into account this information and the higher charge to volume ratio, the produced shift of this band from the theoretical value (3118 cm⁻¹) to 3017 cm⁻¹ is not surprising.

After the ionic liquid deposition onto the support, the maximum of the bands positioned at 3146 and 3090 cm⁻¹ ν(C₄H C₅H) suffer a slight shift to a higher wave number (being practically constant in relative intensity). In addition, the band at 3017 cm⁻¹ disappears, while the bands due to the methyl and butyl groups' vibration keep the same position. This fact is also observed in the bending region (not shown because of the low intensity). These observations suggest that the interaction with the phosphate-modified support proceeds via the C₂ atom of the imidazolium ring (disappearance of the 3017 cm⁻¹ band), being the frequencies of vibration of the substituting groups the least affected (Scheme 1, second row). Therefore, we can conclude that the immobilization process of the imidazolium ring, initially equilibrated by methanesulfonate, proceed now through interaction with the phosphate groups from the support, which provoke the disappearance of the ν(C₂-H) band.

When gold is attached to the support/ionic liquid system from aqueous solutions, no trace of ionic liquid vibrations could be observed and the obtained spectrum is practically identical to the IR spectra of the modified alumina. This result shows that the formed bond between the support and IL is strongly electrostatic (attraction between negatively charged alumina surface and the imidazolium ring). However, in acetone solution the vibrations due to the IL are still present, which indicates the occurrence of this phase either as bonded species or as a physical mixture.

In addition, Raman spectroscopy was used to study the IL/support interaction. Figure 6 shows the C-H bonding region for all the samples and Table 2 the assignments of the bands, as reported previously for butyl methyl imidazolium based ionic liquids [33,35,36].

After the deposition of IL on the alumina support, all Raman lines of the bare IL are preserved, except the vibration band at 3107 cm⁻¹ due to the vibration of CH bond between the two nitrogen atoms in the imidazolium ring (the C₂H bond, as shown in Scheme 1). This observation is not completely unexpected and agrees with the IR data, and confirms that the bonding of the IL goes through electrostatic attraction between the positive charge at this place in the imidazolium ring and the negatively charged alumina surface. The bands caused by the vibrations of butyl and methyl radicals bonded to the imidazolium ring, did not change after the deposition on the support, which was expected for the short chain radicals where the possibility of steric attraction or repulsion is close to zero.

After gold deposition in acetone solution (Figure 6(E)), a weak signal was observed indicating the presence of ionic liquid. However, all Raman lines were red shifted signalizing a perturbation of the imidazolium ring by the gold addition. In aqueous solution, no signals were detected either by masking of the signals by the strong fluorescence or by the ionic liquid removal from the support (Figure 6(C,D)).

The most interesting results have been obtained when submitting the catalysts on CO oxidation reaction (Figure 7). The sample prepared in acetone solution does not show any activity in the employed range of temperatures (from room temperature to 300 °C). As observed, the gold particle size was rather big, but still not enough to explain this result, so there should be another reason for the complete inactivity of this sample, no matter the temperature. The dark field transmission study on this sample presented in Figure 5(D), shows that the reduced gold particles are defect free resulting in a perfect crystal. The absence of defects, i.e., no active sites for oxygen activation and CO oxidation, could explain the absence of activity for this sample [37,38]. This interesting result also confirms the importance of the presence of water in the preparation of the gold catalyst and gives the opportunity to hypothesize that the gold active sites are formed and stabilized not only by the presence of support but as well by the complicated network of hydrogen bonding. Actually, this result could explain why the small quantities of water are always beneficial for CO oxidation, as reported previously [39,40], not only because of the participation of water in the CO oxidation mechanism, but because of the active sites electronic distribution stabilization.

The samples prepared in aqueous solutions show a good activity in CO oxidation and high dependence on gold reduction conditions. When milder reduction conditions are applied, (Au/20IL/Al₂O₃-citrate) the catalyst shows a normal behavior, an increase of the activity with the increase of the temperature. No room temperature activity for this sample was observed. However, the catalytic behavior changes dramatically when rapid reduction was applied (NaBH₄ instead of citrate). As described above, a stabilization of the gold particles by the ionic liquid could be envisaged. The catalyst was not influenced by temperature and shows an important activity in the whole temperature range.

This big difference in the catalytic comportment of the two samples prepared in aqueous media, very similar in terms of gold particle size, leads to the conclusion that the mode of stabilization of the gold active sites for the two samples does not even match. When the milder conditions of reduction are applied, the stabilization of the gold particles was prioritized on the surface of the support free of ionic liquid, that is why the catalyst presents a “normal” light off curve already observed for gold catalyst supported on alumina [41,42].

For the rapid reduction sample, a different mode of stabilization of the gold particles can be imagined. Gold can be stabilized, at the moment of the colloid nucleation and formation, by the creation of micelle-like aggregations in the low concentrated ionic liquid aqueous solutions; such formation being eminent and easy [43]. In this case, the as produced IL stabilized gold colloids will not interact directly with the support but through the imidazolium ring. Thus the delocalization of the electrons of gold on the imidazolium ring could allow a different mode of activation of oxygen (step limiting stage) not influenced by the temperature, which could explain the unusual catalytic behavior. However, this can be applied only for the so called “inactive” supports [44] for which the interaction gold-support is very weak. However, our recent study on the IL templated titania–gold systems also shows a different catalytic behavior, for which the presence of the ionic liquid was strongly responsible [12].

However, every excess of this IL (1-butyl-3-methylimidazolium methanesulfonate) could be detrimental for the catalytic activity, because of the preferential deposition of the imidazolium ring in the plane [45], thus completely covering the gold active sites. The latter can provide an additional explication for the complete inactivity of the acetone prepared samples in which the quantity of the IL is higher. Nevertheless, this inconvenience can be avoided by the application of IL with another counter ion, like Cl⁻, BF₄⁻, etc. for which the geometry of deposition is different.

The activity measurements, as described elsewhere [37,38] were carried out in a flow U-shape glass reactor at atmospheric pressure. The catalyst (80 mg, φ < 200 μm) was placed into the glass reactor and a reactive flow (3.4% CO and 21% O₂ balanced by helium) was passed through it. The total flow was 42 mL min⁻¹ and the quantitative analysis was carried out with a Balzers Omnistar Bentchop mass spectrometer. The catalyst was tested in the flow during 1 h at room temperature. Then, the system was heated to 300 °C and stabilized at this temperature for 30 min.

In the reaction of carbon monoxide oxidation the samples were employed after drying without any additional treatment.

X-ray diffraction (XRD) analysis was performed on a Siemens diffractometer D500. Diffraction patterns were recorded with Cu Ka radiation (40 mA, 40 kV) over a 2θ-range of 10 to 70° and a position-sensitive detector using a step size of 0.05° and a step time of 1 s.

The point of zero charge (PZC) of the support was determined by measuring (Malven Zetamaster) electrophoretic mobilities of aqueous dispersions as a function of pH, at constant ionic strength (10⁻² mol L⁻¹ NaCl). The pH was varied by adding adequate amounts of 0.01 mol L⁻¹ HCl or NaOH.

The textural properties were studied by N₂ adsorption measurements at liquid nitrogen temperature. The experiments were carried out on a Micromeritics ASAP 2010 equipment. Before analysis, the samples were degassed for 2 h at 150 °C in vacuum.

Scanning electron micrographs (SEM) were taken using a JEOL JSM-5400 electron microprobe and Link energy dispersive X-ray (EDX) analyser.

Transmission electron microscopy (TEM) observations were carried out in a Philips CM200 microscope operating at 200 kV. The samples were dispersed in ethanol by sonication and dropped on a copper grid coated with a carbon film.

The IR spectra were recorded on a Nicolet 380 Spectrometer at a spectral resolution of 2 cm⁻¹ and accumulation of 100 scans. The pellets were prepared from the dried sample powder and KBr (1/20).

The Raman spectra were recorded on a dispersive Horiba Jobin Yvon LabRam HR800 microscope, with a 300 mW He-Ne red laser (784.56 nm) without filter and with a 600 g mm⁻¹ grating, 15 acquisitions (10 s/scan). The microscope used a 50× objective with a confocal pinhole of 500 μm.

The elemental analysis of gold was performed by X-Ray Fluorescence in a Panalytical AXIOS PW4400 sequential spectrophotometer with Rh tube as source of radiation. The measurements were performed onto pressed pellets containing 6 wt% of wax.