Operando Spectroscopy of the Gas-Phase Aldol Condensation of Propanal over Solid Base Catalysts

Complementary to the characterization data previously reported for these samples (Table 1 [15, 17, 19]), the intrinsic acid strength of the Brønsted (BAS) and Lewis acid sites (LAS) was probed by FT-IR analysis after the adsorption and desorption of CO and pyridine as probe molecules. The amphoteric nature of CO in addition provides further insight into the basicity of the samples [26]. Figure 1 shows the FT-IR spectra of the OH (Fig. 1a, c, e, g) and CO (Fig. 1b, d, f, h) stretching region collected for the four catalyst materials during CO adsorption.

For Cs-X (Fig. 1a), the weak interactions of silanol groups [27, 28] with CO, at ~3716 cm^− 1, demonstrates its lack of BAS. This absence of Brønsted acidity is a result of the efficiency of cation exchange during the material’s synthesis. Expectedly, the CO stretching region revealed the presence of strong LAS for this sample (Fig. 1b). The Cs⁺ Lewis acid sites are characterized by spectral features in the region 2264–2240 cm^− 1. Also, the band at 2121 cm^− 1 might be caused by the coordination of CO to the electrophilic Cs⁺ centers via its C atom (O=C…Cs⁺) [29]. At high CO pressure, absorption bands at 2141 and 2132 cm^− 1 can be observed, attributed to physisorbed CO, while the shoulder at 2149 cm^− 1 is associated to CO on coordinatively unsaturated Al³⁺ Lewis acid sites [30].

For Ca-HA (Fig. 1c) the OH stretching area is virtually unperturbed by the CO adsorbed, in line with the non-acidic character of the OH groups present [19]. The sharp band at 3573 cm^− 1 is due to contributions from surface and bulk OH groups [31, 32]. Some minor perturbations are observed in the 3600–3700 cm^− 1 spectral range and these features can be attributed to P-OH groups acting as weak basic sites. Other evidence for basicity are bands formed at 1447 and 1409 cm^− 1 (inset Fig. 1c), which correspond to carbonate species formed by strongly basic OH groups [32]. Carbonation, initially present in the beginning of the test, could be due to contamination of the hydroxyapatite with CO₃ ²⁻ ions coming from atmospheric CO₂ when synthesized in air or storage [32]. Meanwhile the more intense bands at 1579 and 1547 cm^− 1 are related to surface (PO_x)_s-carbonates formed by weak PO₃ ⁴⁻ sites of the hydroxyapatite material [31]. In addition, the bands detected at 2179 and 2169 cm^− 1 in the CO stretching region (Fig. 1d) indicate the presence of some weak LAS [33].

For Na-USY (Fig. 1e), the free silanol groups observed at 3747 cm^− 1 before CO exposure, partially disappear upon CO adsorption and shift to 3654 cm^− 1. These hydroxyl species are known to be located in the supercages of zeolite Y and can be considered as isolated and very weak BAS, as the small shift Δν = 94 cm^− 1 indicates [34, 35]. The Δν shift is proportional to the acid strength of the OH group [34]. The band at 2172 cm^− 1 (Fig. 1f) can be attributed to CO adsorbed on Na⁺ Lewis acid sites [28]. More intense bands at 2166 and 2159 cm^− 1 are associated with CO adsorbed on silanol groups, meanwhile the weak band at 2179 cm^− 1 is thought to derive from the extinguished BAS of the parent USY zeolite. Physisorbed CO was also observed at 2142 and 2134 cm^− 1 [35].

For the Rb-USY sample (Fig. 1g) the free external silanol groups, centered at 3751 cm^− 1, show almost no acidic character [34‐37]. Only weak bands, at high coverages of CO (p _CO > 3 mbar), appear in the CO spectral region (Fig. 1h). The band at 2157 cm^− 1 appears as a consequence of weak CO interactions with external silanol groups, with the band at 2140 cm^− 1 corresponding to physisorbed CO. Note that for neither of the two USY samples under study an absorption band at ~2260 cm^− 1 was detected, which is known to be associated with strong LAS. Consequently, in these samples (and in contrast to the Cs-X sample) the cations must be located in more inaccessible positions, e.g. the sodalite cages [38].

Figure 2 shows the FT-IR spectra collected for the four catalyst materials during pyridine adsorption at room temperature and subsequent desorption.

The FT-IR spectra after pyridine adsorption and subsequent desorption further corroborate the findings obtained with CO as probe molecule. Indeed, the lack of a band at ~1545 cm^− 1 [39, 40] confirms that there are practically no strong BAS on any of the catalyst materials. Along these lines, the weak perturbations at ~1555 and 1595 cm^− 1, associated to the ring vibration 8a mode of pyridine (according to the nomenclature introduced by Kline et al. [41]) are seen for the alkaline metal-grafted USY zeolites and could be due to hydrogen-bonded pyridine with weakly acidic silanol groups [27, 42]. On the contrary, LAS can be observed in every catalyst material. In the case of Cs-X (Fig. 2a) bands at 1597 and 1440 cm^− 1 are thought to correspond to pyridine adsorbed on Lewis acidic Cs⁺ cations [27]. By progressive outgassing the materials, these absorption bands decrease in intensity and were no longer detected after outgassing at 150 °C, in the case of Na- and Rb-USY catalysts, and 300 °C (not displayed) in the case of Cs-X and Ca-HA, suggesting that these acid sites are not strong [39]. For Ca-HA (Fig. 2b) bands emerged at 1600, 1573, 1488 and 1445 cm^− 1 [33], and decrease with increasing outgassing temperature indicating the weakness of the LAS. Very weak LAS are detected for the Na-USY zeolite material (Fig. 2c) as characterized by the band at 1443 cm^− 1 [28, 39] which desorbs at low temperature (<150 °C). Even weaker are the LAS in the Rb-USY zeolite material (Fig. 2d), as evidenced by the absorption band at 1441 cm^− 1, which practically disappeared upon increasing outgassing temperature.

The amount of LAS for each catalyst material under study was calculated based on Eq. 1, derived from Beer´s law applied to dδ (cm^− 1) [39]. In Eq. 1 A (cm^− 1) represents the integral under the curve delimited by dδ (cm^− 1) (band at 1440–1445 cm^− 1), ε is the extinction coefficient at 150 °C, previously determined as 2.22 cm/μmol and reported by Emeis [39], and ρ is the mass (mg) of the wafer per cm² through which the beam is sent (effective cross-section). Py:LAS stoichiometry was assumed to be 1:1, i.e., only one Py molecule is adsorbed per accessible Lewis acid site. The results of this quantification procedure are included in Table 1, providing clear differences in the amount of acid sites; i.e., Cs-X >>> Ca-HA > Na-USY > Rb-USY.

The time-resolved operando FT-IR spectra obtained for the four catalyst materials, described in Table 1, are shown in Fig. 3. The peak assignments are listed in Table 2.

The time-resolved FT-IR spectra obtained for the Cs-X zeolite (Fig. 3a) shows negative intensities at 3732 and 3572 cm^− 1 suggesting interactions of reagent/products with the few silanol groups of the sample (Fig. 1a). Within the first minutes of reaction, bands in the range 2963–2718 cm^− 1 emerge and are ascribed to aliphatic C–H stretching (ν_C−H) vibrations. Although the assignment of these bands is not straightforward, the band at 2718 cm^− 1 can be unambiguously assigned to a Fermi resonance structure of the CH stretching vibrations from aldehydes [43, 45, 46]. After two minutes of reaction, a ν_C−H band arises at 3047 cm^− 1, most likely originating from aromatic species [43]. This band correlates with the increase in absorption of a strong band centered at ~1590 cm^− 1 attributed to C=C stretching (ν_C=C) of carbonaceous species (coke) [48‐52]. In the carbonyl stretching (ν_C=O) region, the propanal absorption at 1759 cm^− 1 is almost unnoticed compared to the other ν_C=O band at 1700 cm^− 1, assigned to an α,β-unsaturated compound, i.e., the aldol dimer product, 2-methyl-2-pentenal, as the experimental bands derived from its adsorption onto the catalyst indicate (Fig. S3a [44, 47]). The ν_C=C band of the enol group of the aldol dimer quickly develops at 1687 cm^− 1 (Fig. S3a). Another ν_C=O band arises at 1714 cm^− 1, which is assigned to the side product 3-pentanone (Scheme 1). Other absorption bands appearing in the fingerprint region (ν < 1500 cm^− 1) correspond to C–H bending vibrations (σ_C−H) of ethyl (1467 cm^− 1) and methyl (1384 cm^− 1) groups. After 40 min of desorption, the spectrum profile still shows FT-IR features at 3047, 1590 and 1366 cm^− 1, providing evidence for the presence of remaining carbonaceous deposits.

The time-resolved IR spectra for Ca-HA (Fig. 3b) show surface P-OH and lattice OH⁻ groups (3670 and 3566 cm^− 1) involved in the adsorption of species and/or catalysis, in a similar way to the interactions of CO with OH stretching bands indicated in the Cs-X zeolite. After 2 min of reaction, bands in the C–H stretching region ascribed to aliphatic aldehyde (i.e. propanal) (2972–2720 cm^− 1) and to aromatic/olefinic species (3058 cm^− 1) are observed. In the carbonyl stretching region and immediately after starting reaction, the ν_C=O of propanal emerges with a doublet at 1760 and 1734 cm^− 1, indicative of both gas phase and chemisorbed propanal, respectively [45]. With increasing time-on-stream both absorption bands increase till saturation levels (Fig. S7), with the 1734 cm^− 1 band associated to adsorbed propanal increasing more than the 1760 cm^− 1 band. The ν_C=O and ν_C=C bands, ascribed to the aldol dimer, arise within the first 2 min of reaction at 1705 and 1670 cm^− 1 accompanied by its σ_C−H bands at 1404 and 1379 cm^− 1. Bands at 1544 and 1441 cm^− 1 are identified as carbonate species, initially present. The carbonation of the catalyst is a consequence of the adsorption of species on strong basic sites, leading to a rearrangement between Ca²⁺ cations of the Ca-HA and lattice OH⁻ groups [31]. Nevertheless, poisoning of the basic sites by CO₂ only affects the active sites working at the beginning of the reaction, those still working at steady state being not impacted. With increasing time-on-stream another ν_C=C band arises at 1640 cm^− 1, likely due to the formation of the aldol trimer [46, 47]. As for the Cs-X catalyst material, most absorption bands disappear upon desorption, except for the ν_C=C band associated with carbonaceous deposits. The ν_C−H band assigned to carbonaceous species (3058 cm^− 1) is significantly weaker than for Cs-X, though. It could mean either that coke forms more extensively on the Ca-HA catalyst relative to the Cs-X material.

The FT-IR spectra for Na-USY are represented in Fig. 3c. The negative intensity seen for the unconstrained hydroxyls [53] at 3732 cm^− 1 indicates its involvement in the adsorption and/or catalysis. Within the first moments of the catalytic reaction, propanal spectral features emerge at 2976, 2940, 2812 and 2720 cm^− 1, for the ν_C−H vibrations, as well as 1763 and 1735 cm^− 1 for the ν_C=O vibrations. After 5 min of reaction the ν_C=O stretching bands of the aldol dimer weakly appear at 1706 cm^− 1, accompanied by the conjugated ν_C=C vibration at 1690 cm^− 1 (Fig. S5a). The absorption band emerging at 1717 cm^− 1 is again associated with the formation of 3-pentanone. Coke formation is noted by the appearance of a weak stretching band at 3063 cm^− 1 and the stronger ν_C=C signal at 1575 cm^− 1. These spectral features are retained after 40 min of desorption, and consequently the absorption band associated with external silanol groups does not fully recover, indicating that carbonaceous species remain anchored.

Within the first moments of the catalytic reaction, the FT-IR features of the Rb-USY sample (Fig. 3d) are very similar to those of the Na-USY sample (Fig. 3c). Nevertheless, the early appearance of the absorption bands at 1709 and 1698 cm^− 1 of the ν_C=O and ν_C=C vibrations of the aldol dimer, suggest a higher activity in the aldol condensation reaction, which is in line with previously reported catalytic activity tests [17]. In addition to the absorption bands corresponding to the formation of the aldol dimer, a weaker absorption band at 1688 cm^− 1 could be ascribed to the ν_C=C vibration of the aldol trimer. The ν_C−H and ν_C=C bands emerging at 3078 and 1573 cm^− 1 reveal that this catalyst material is also prone to the formation of aromatics and coke. After catalytic reaction, the Rb-USY catalyst material recovers faster than any other catalyst material under study. After 20 min of desorption signs of any remaining organics are less evident than for the three other discussed catalyst materials. However, compared to Na-USY, more aromatics are observed for the Rb-USY (as measured by the ratio ν_C=C coke/ν_C=O product). Nevertheless, the Rb-USY catalyst material presents better recoverability, which might be related to the more moderate acid character of the free silanol groups.

It is important to note that the FT-IR spectra look very different for the alkaline metal-grafted USY zeolites compared to Cs-X and Ca-HA. For both Na- and Rb-USY catalysts (Fig. 3c, d) the FT-IR features of the reagent dominate over those of the product, underlining that the surface reaction is limiting [54] in this case. Previously, it has been suggested that propanal condensation follows a single-site Eley–Rideal mechanism and the observed reaction order of 1 for propanal was suggested to be due to either a surface reaction or the desorption step being rate-limiting [55]. In this case, the spectra suggest it would be the latter. For the different steps involved in the surface reaction, i.e. enolization of propanal, C–C coupling and dehydration, C–C coupling would be then be rate-determining. For Cs-X and Ca-HA (Fig. 3a, b) the spectra suggest the desorption step to be the reaction rate-limiting, given the fact that FT-IR features of product adsorbed onto catalyst surface dominate over those of the reagent. Focusing on the surface reaction, it is meaningful that for every catalyst material under study, except for the Cs-X material, the ν_as(CH₃–) and σ_CH3 bands dominate over the ν_as(–CH₂–) and σ_CH2 bands. For Cs-X instead, the higher intensity of the bands ν_as(–CH₂–) at 2923 cm^− 1 and σ_CH2 at 1465 cm^− 1 indicates a higher presence of more linear products, i.e., 3-pentanone over the aimed aldol dimer product.

During the catalytic reaction, the initially white catalyst wafers darkened as a result of the formation of conjugated (aromatic) species, ones that might be involved in deactivation that is observed for the different catalyst materials. The operando UV–Vis Diffuse Reflectance Spectroscopy (DRS) data on this process are shown in Fig. 4, while the absorption band assignments are summarized in Table 3.

For all catalyst samples under study an absorption band centered at ~260–270 nm emerged soon after exposing the solids to propanal, associated with the weak n → π* electronic transition of the carbonyl group in the aliphatic aldehyde substrate (Table 3). Correlating with the changes seen in the time-resolved FT-IR analyses, after 1 min of reaction, for Cs-X (Fig. 4a) and Ca-HA (Fig. 4b), and 2 min in the case of alkaline metal-grafted USY zeolites Fig. 4c, d, two more absorption bands emerge at ~240 and ~300 nm respectively, which are associated with π → π* and n → π* transitions of the C=C and C=O groups of the aldol dimer [45, 55] (Fig. S3b). The shift from the theoretical position of the C=C band of the dimer (at ca. 220 nm [46]) to the experimental one (ca. 240 nm) might be due to the detection limit of the equipment, which is in the range of 230 < λ < 1000 nm. After 2 min the absorption band at ~300 nm, ascribed to the dimer, quickly increases in intensity at the expense of the propanal absorption band centered at ~270 nm. This absorption band then progressively redshifts with increasing time-on-stream, which might be due to an increase in the alkyl substituents, in line with the C–C coupling reaction and subsequent formation of the aldol trimer. According to estimations based on the Woodward-Fieser rules for predicting λ_max positions for π → π* transitions of α,β-unsaturated aldehydes [46], the position of the C=O absorption for the aldol trimer should be placed at ~376 nm. For the Cs-X (Fig. 4a) and Ca-HA (Fig. 4b) catalyst materials, after 2 min of reaction the UV–Vis spectra broaden and absorb light in the entire wavelength range. This broad absorption feature developed later for the Na-USY zeolite (Fig. 4c) and its Rb-counterpart (Fig. 4d). These electronic transitions are ascribed to the fast development of aromatic ring structures and carbonaceous species that could cause catalyst deactivation [52]. For the Cs-X catalyst material though, the absorption bands in the visible region tail less towards longer wavelengths compared to the other three catalyst materials. This observation could be related to a softer/less condensed composition of the coke developed over the Cs-X material, which correlates well with the rich-in-hydrogen carbonaceous deposits observed for Cs-X as measured by FT-IR (Fig. 3a). The noted signal saturation, which relates to the rate of coking, is more pronounced for Cs-X and Ca-HA (after 10 min of reaction) than for the two other samples; i.e., 15 min for Rb-USY and 40 min for Na-USY.

After extensive desorption by N₂ flushing the intensities of the absorption bands at longer wavelengths considerably decrease, which indicates a certain release of volatile aromatic species. After 40 min of outgassing, a large fraction of the large aromatics (i.e. species absorbing at the highest wavelengths, λ > 500 nm) were released from all catalyst materials. Additional carbonyl containing species and mono-olefinic species, which absorb light at ~300 and <250 nm respectively, still linger. In line with the FT-IR results (Fig. 3d), the decrease in absorption upon desorption is slower and less complete for Na-USY as compared to Rb-USY (Fig. 4d).

The differences seen in the operando FT-IR and UV–Vis DRS profiles of the various catalyst materials under study are also reflected in their catalytic performances, as monitored by on-line mass spectrometry (MS). The catalytic performance data of the different catalysts under study are summarized in Fig. 5. Ion currents were recorded for the m/z values of the propanal reactant, the aldol reaction intermediates, as well as the products water and the aldol condensation dimer and trimer products and the by-products 3-pentanone and propyl propionate. The MS fragments of the species considered are summarized in Table S1 (see also Schemes S1–S3). Selection of these m/z values was based on the database [44], if available. For species not present in the database (e.g., the aldol intermediate product and the aldol trimer) fragmentation patterns were predicted, based on a methodology described elsewhere [46] to determine the most reliable ion current (Scheme S1–S3). Original ion current values obtained for each species were normalized by the corresponding initial fragmentation of each species directly bypassed to the mass spectrometer, by the value of the signal of the carrier gas (N₂, m/z = 28) and by the catalyst weight.

Figure 5 shows that shortly after commencing the reaction propanal conversion reaches a quasi-steady state with increasing time-on-stream. As expected, the aldol dimer (m/z = 41) is observed to be the dominant product for all the samples. Beyond the high selectivity towards the aldol dimer, the presence of side products, such as 3-pentanone (m/z = 86) and its precursor propyl propionate (m/z = 75), is higher for Cs-X than for the other three catalysts (Fig. 5a), as previously indicated during the catalytic tests [16] and confirming the validity of the combined operando spectroscopy setup used. The aldol trimer is seen most with Ca-HA, increasing at the expense of the dimer (Fig. 5b), with its high activity for consecutive aldol condensation again being in line with reported catalytic tests run in a standard micro-reactor [19].

The high sensitivity of the MS instrument, also allowed for the detection of various other species, including the possible aldol intermediate, 3-hydroxy-2-methylpentanal (m/z = 116); the latter compound was hardly detected for the alkaline-grafted USY zeolites (Fig. 5c, d), but was seen for the Cs-X and Ca-HA materials (Fig. 5a, b). The detection of the aldol intermediate for Cs-X and Ca-HA lead us to think that dehydration is the rate-determining step of the surface reaction, which is further supported by the low intensity of water signal during the whole catalytic reaction. In contrast for Na- and Rb-USY, for which—besides the absence of 3-hydroxy-2-methylpentanal signal—the water signal is released at substantially larger scale during reaction, the C–C coupling is considered as the rate-limiting step of the surface reaction.

Alcohol dehydrations, i.e. 3-hydroxy-2-methylpentanal converting into the aldol dimer product in the case at hand, are more favorable performed over acidic than over basic catalyst materials because of the preference for abstracting the H⁺ of the C attached to the OH substituent (over the H⁺ abstraction at C-1) [13]. Therefore, the dehydration step over the more basic catalyst materials, i.e., Cs-X and Ca-HA, might become more suppressed than over the two alkaline metal-grafted USY zeolites.