3.1 Characterization of samples

Table 1 summarizes the main characteristics of samples used for catalytic tests. The samples belonging to the K series of montmorillonite-type acid clays (K5, K10, K20, K30), possess an increasing silica-to-alumina ratio, and a corresponding decreasing amount of iron oxide, of alkali/alkaline earth metal oxides, as a consequence of the dealumination (acid treatment) procedure. The latter also causes the increase in surface area; the pore distribution for K5 and K30 is illustrated in Fig. 1 . It is shown that the dealumination treatment modifies the pore size distribution; the mesopores distribution in the K30 sample is shifted towards the macropore region, while in the K5 sample the majority of pores falls in the 20–60-Å pore radius range. Sample FULCAT 22B possesses a composition that is intermediate between that of K5 and of K10 samples.

The XRD patterns of samples belonging to the K series are shown in Fig. 2 . Patterns are those of silico-aluminates (e.g., montmorillonite IPDF 02-0037, silica IPDF 33-1161, and tosudite IPDF 22-0956); additional reflections are observed in K5, attributable to Ca/Mg/Fe/Si mixed oxide (IPDF 13-0421), and to kaolinite Al₂Si₂O₅(OH)₄ (IPDF 06-0221). These reflections disappear in dealuminated samples.

The acidity of samples K5 and K30 was checked by means of pyridine adsorption. Fig. 3 plots the FT–IR spectra of K5 and K30 samples, after evacuation and adsorption of pyridine at room temperature, followed by heating at 200 and at 400 °C. It is shown that, after evacuation at 200 °C, sample K5 possesses a higher number of Brønsted sites than sample K30, as inferred from the higher intensity of the IR band positioned at 1545 cm^–1. After evacuation at 400 °C, the residual amount of Brønsted sites is very low, while Lewis sites are still present, as inferred from the band falling at 1455 cm^–1. The intensity of the latter band is comparable in the two samples. Flessner et al. [22] found that the overall number of surface acid sites (Brønsted + Lewis), as determined by ammonia adsorption, decreases on increasing the Si/Al ratio. The ratio between the number of surface silanols at the external layer, and of Al₂OH groups in the interlayer region also was found to follow the Si/Al ratio. Spectroscopic measurements evidenced that the highest number of Brønsted acid sites was in samples having the lowest alumina content (K30); this led to the conclusion that medium-weak Brønsted sites are associated to external silanols in the tetrahedral clay layer. The acid strength of these sites was not much affected by the Si/Al ratio. Lewis acid sites, relatively strong, were instead associated to low-coordination Al³⁺ cations [22]. The authors proposed that the progressive dealumination of the K samples, from the K5 to the K30 sample, degraded the octahedral clay layers, with a reduction of the interlayer charge and thus making internal protonic acid sites, associated to silanols, available for interaction with organic molecules. Disagreement with data reported by Flessner et al. [22] concerns the number of Brønsted sites, which in our case was found to be higher in the sample having the higher Al content (K5). This difference may be due to the fact that the basic probe used in the present work is smaller (pyridine against pivaronitrile); this made it possible better access of the probe into the intralayer spacing of the clays, in which part of hydration water remained entrapped even after thermal treatment and evacuation at high temperature. Water may be responsible for the generation of Brønsted sites.

3.2 The reactivity in resorcinol acylation: the reaction scheme

Tests were made by changing the reaction time, with the K5 and K30 samples; results are summarized in Fig. 4 , which reports the selectivity to the products, resorcinol monobenzoate (RMB) and 2,4-dihydroxybenzophenone (2,4-DHB), as functions of resorcinol conversion; the latter was varied by varying the reaction time and/or the amount of catalyst loaded. Under our conditions, no other by-products were found. It is shown that with both catalysts the only primary product was RMB, the selectivity of which was total for low resorcinol conversion. On increasing the conversion, the selectivity to RMB decreased, with a corresponding increase in the selectivity to 2,4-DHB. Therefore, the latter was exclusively formed by consecutive reaction upon RMB, via Fries rearrangement. This suggests that (i) RMB is not directly transformed, once formed, to 2,4-DHB inside catalyst porosity, since otherwise this would correspond to a direct contribution in the formation of this compound (which instead was not observed), and (ii) RMB first diffuses out of the catalyst pores, and successively is activated and transformed to 2,4-DHB. One possible explanation is that the high hydrophilic features of the clay surface, especially in samples having higher alumina content, makes the pores filled with more polar substances, such as benzoic acid and resorcinol, while less polar compounds rapidly diffuse out of the clay porosity and undergo consecutive transformation at the active sites located at the external particle surface.

Tests shown in Fig. 4 were made so as to obtain data far from the possible equilibrium; further data were obtained with the FULCAT 22B sample at 180 °C, reaction time 2 h, and by addition of 50% wt. of mesitylene as the reaction solvent. This also makes possible the continuous removal of the water formed by the O-acylation through the azeotrope mesitylene/water. Results are compared in Table 2 with those obtained (i) by carrying out the reaction in the absence of solvent, (ii) by reacting RMB alone (thus starting the reaction in the absence of resorcinol and benzoic acid), (iii) by reacting together resorcinol, benzoic acid and RMB, (iv) by reacting RMB with an equimolar amount of water. Even for reactions made starting from RMB, data have been expressed with reference to resorcinol (thus, under the hypothesis of loading an equimolar amount of resorcinol and benzoic acid), and considering RMB and 2,4-DHB as the reaction products. This makes possible the comparison of data obtained when starting from RMB with those obtained when starting from resorcinol + benzoic acid. It is shown that the distribution obtained was the same when starting from the different reacting mixture. This suggests that the reaction between resorcinol and benzoic acid to yield RMB is fully reversible, and that the hydrolysis of RMB to yield resorcinol and benzoic acid also occurs in the absence of added water. This is an indication of the role played by water retained in clays.

Our data agree with those reported by Hoefnagel and van Bekkum [5], who studied the reaction between resorcinol and benzoic acid catalysed by ion-exchange resins and zeolites. The authors proposed that the main primary reaction that occurs is the esterification to yield the (RMB); evidences for direct ring-acylation were not obtained, but a contribution of this reaction was not excluded. Besides RMB and 2,4-DHB, by-products found were resorcinol dibenzoate (RDB), and 3,6-dioxy-9-phenyl-xanthydrol, a coloured compound obtained by reaction between two moles of resorcinol with one mole of benzoic acid; these compounds were not found in our tests with the K series of catalysts, likely because of the lower conversions achieved. Evidence for the formation of low amounts of these by-products was instead obtained with the FULCAT catalyst, which reached higher resorcinol conversion due to the higher reaction temperature employed (Table 2).

The removal of water from the bulk liquid by azeotropic distillation makes it possible to considerably increase the conversion, which reaches 80.6% against 61.3% in the absence of the solvent, since the esterification reaction is favoured.

Scheme 1 summarizes the main findings.

3.3 The reactivity of the Süd-Chemie K series

Table 3 compares the catalytic performance of the K samples; the activity considerably decreased on decreasing the Al content in samples, despite the opposite trend of surface area. This clearly points out that the weak-medium Brønsted sites in samples, associated to silanols on the catalyst surface [22], do not play an important role in the activation of the reactants. On the contrary, the activity trend follows the amount of Al in samples. Therefore, water molecules present in samples and coordinated to Al³⁺ sites may act as the Brønsted active sites, through protonation of the carboxylic group; in fact, water is present in clays, which are highly hydrophilic materials. Moreover, water is the co-product of the esterification reaction, and therefore is anyway present in the reaction environment.

In order to distinguish between the possible role of Lewis-type and Brønsted-type sites, K5 and K30 samples, i.e., those samples having the highest and the lowest Al content, respectively, have been treated at increasing temperatures, in order to decrease their water content, and have been successively tested as catalysts in the benzoylation of resorcinol. Fig. 5 plots, for the two clays, the loss of weight as a function of the thermal-treatment temperature, the conversion of resorcinol and the selectivity to the products achieved for the corresponding thermally treated samples. It is shown that increasing temperature of thermal treatment led to a progressive loss of weight. Correspondingly, the conversion of resorcinol progressively decreased, and the selectivity to the primary product increased, as a consequence of the decreased level of conversion. Therefore, the data clearly point out that the catalytic activity of these materials is associated to water retained in the solid. The same occurred for FULCAT sample; a pre-treatment at 180 °C led to a decrease in the conversion from 52.0% (for the untreated sample) to 39.0% (for the treated sample).

Fig. 6 compares the selectivity to the two main products, RMB and 2,4-DHB, at the same level of conversion, i.e., 23–25%, for the whole K series; the silica-to-alumina wt. ratio is taken as the main compositional parameter of samples. The comparison at resorcinol iso-conversion makes possible to search for relationships between the catalyst chemical-physical features and the distribution of products; the latter in fact is affected by the reactants conversion when the reaction scheme consists of consecutive reactions. Moreover, a conversion close to 25% is far from that obtained under equilibrium conditions; therefore, under the conditions employed, the reaction is under kinetic control. It is shown that an increase in the silica-to-alumina ratio leads to an increase of the selectivity to RMB and to a corresponding decrease of selectivity to 2,4-DHB. Since the formation of 2,4-DHB exclusively occurs via consecutive reaction upon RMB, and since the formation of the latter compound is the only primary reaction contributing to resorcinol conversion, the observed effect points out a decrease of the rate of the consecutive intramolecular rearrangement of the ester to the ring-acylated compound occurring in dealuminated samples, with respect to the primary rate of RMB formation.

Therefore, two different phenomena are evident when the alumina content is progressively decreased in catalysts: (i) a considerable decrease of catalyst activity, and (ii) a decrease of the rate of transformation of RMB into 2,4-DHB as compared to the rate of RMB formation. The effect on selectivity may be associated to the lower clay hydrophilicity and acidity that develops for increasing Si/Al ratios. This decreases the interaction of the catalyst not only with the reactants (thus contributing to the decrease of conversion), but also with RMB, leading to a lower rate of the consecutive transformation of RMB, due to both an accelerated desorption of the product from the active surface, and a slower rate of consecutive re-adsorption of the same compound, preliminary to its intramolecular rearrangement.