An Evaluation of Glycerol Acetalization with Benzaldehyde over a Ferromagnetic Heteropolyacid Catalyst

Abstract

Tungstophosphoric acid (H₃PW₁₂O₄₀) supported on silica-coated magnetite nanoparticles has been prepared and used as a heterogeneous acid catalyst (Fe₃O₄@SiO₂@HPW) in the condensation of benzaldehyde (B) with glycerol (Gly) for the production of cyclic acetals. Physicochemical techniques, including scanning electron microscopy (SEM), transmission electron microscopy (TEM), Fourier-transform infrared spectrometry (FTIR), X-ray diffraction (XRD), thermogravimetric analysis (TGA), and N2 physisorption were used to characterize the prepared catalyst. The effect of glycerol/benzaldehyde molar ratio (1/1 to 1/1.2), temperature (80–120 °C), and catalyst amount (1–5%) on glycerol conversion and the selectivity in main reaction products: benzoic acid, 2-phenyl-1,3-dioxolan-4-yl)methanol, 2-Phenyl-1,3-dioxan-5-ol, 2-phenyl-1,3-dioxolane, and methyl 2-hydroxy-3-phenylpropanoate was studied to evaluate the catalytic activity of the prepared Fe₃O₄@SiO₂@HPW catalyst. The optimization of these process parameters was conducted using Box–Behnken design (BBD). Using the BBD methodology, the optimal parameters (120 °C, 1:1.15 glycerol: benzaldehyde, 5% catalyst) were determined, for a glycerol conversion of and cyclic acetals yield of 85.95% and 78.36%, respectively. The catalyst also exhibits excellent activity for glycerol acetalization with other aromatic aldehydes such as cinnamaldehyde, p-methyl-benzaldehyde, p-hydroxy-benzaldehyde, and vanillin.

1. Introduction

The valorization of glycerol produced by the biodiesel industry is important for environmental issues and biodiesel process economics. It is crucial for many industries where glycerol is used as an intermediate compound to obtain added-value products by biochemical, chemical, thermochemical, or catalytic routes. This mainly focuses on the dehydration of glycerol to acrolein (an important chemical intermediate in many industries) [1,2,3], polymerization of glycerol to polyglycerols (surfactants used in the food, detergents, and cosmetics industries) [4,5,6,7], esterification to acylated esters -glyceryl diacetate, glyceryl triacetate (oxygenated fuel additives) [8,9,10], hydrogenolysis of glycerol to propandiol (an important intermediate in the polymer industry) [11,12,13], condensation with carbonyl groups to acetals and ketals (antiknock additives in combustion engines, oxygenated fuel additive, surfactant, and flavoring agent) [14], carboxylation to glycerol carbonate (electrolyte and solvent in batteries) [15,16], and the steam reforming of glycerol to hydrogen (fuel) [17,18,19]. Among them, the condensation of glycerol with carbonyl compounds (ketones and aldehydes) is one of the common glycerol valorization routes described in the literature because valuable cyclic acetals/ketals for various industries are obtained with the formation of H₂O as a non-toxic co-product. They are employed as scents or flavors in the food and cosmetic industries [20], additives for diesel fuel [21,22,23,24,25], green solvents [26,27], bases for surfactants [28], and many other applications with potential use in industry [29].

Traditionally, these reactions are performed in the presence of acidic homogenous catalysts [22], but because of their environmental issues (they cannot be recovered from the reaction system), various heterogeneous acid catalysts have been investigated. Acidic exchange resin [30,31], zeolites [32,33,34,35], carbon-based catalysts [36,37], transition or noble metals and metal oxides [38,39], functionalized mesoporous silica [40], heteropolyacids [41,42] or metal-organic frameworks [43,44] are promising versatile heterogeneous acid catalysts for cyclic acetal production from glycerol and aldehydes or ketones. Generally, the use of heterogeneous catalysts in industry and their efficiency has increased, but while in some cases the recovery of these catalysts is successful (usually by filtration), in other cases, it still remains challenging.

Catalysts supported on a magnetic material (i.e., magnetite) are an interesting alternative to heterogeneous catalysts. Due to their unique properties: non-toxic, low cost, environmentally friendly, and abundant, a wide variety of molecules of magnetite can be easily immobilized and supported on the catalyst surface. Furthermore, easy recovery by an applied magnetic field is a good choice for developing supported catalysts [45,46,47,48].

Tungstophosphoric acid (HPW) has enabled enormous progress in various acid-catalyzed reactions because of their structural mobility, super acidity, low volatility, low corrosivity, and high activity and selectivity for several reactions when compared to conventional mineral acids [49,50]. It has already been proven to be an efficient catalyst in acetalization reactions. Ferreira et al. [51] reported a study about the acetalization reaction between acetone and glycerol using heteropolyacids immobilized onto silica as catalysts. The highest catalytic activity was achieved by tungstophosphoric acid with more than 99% conversion of glycerol with 97% selectivity toward solketal (optimized conditions: glycerol to acetone = 1:6; temperature = 70 °C; catalyst loading = approximately 5 wt% wrt. Glycerol). The efficiency of the catalyst was attributed to the acidic strength of the catalyst, with an initial electrode potential Ei = 290 mV, determined by potentiometric titration with n-butylamine. The catalyst also showed good stability, demonstrated by consecutive runs. Good results for glycerol acetalization with formaldehyde over tungstophosphoric salts were also reported by Chen et al. [41,52]. A series of tungstophosphoric acid catalysts supported on metal oxides (TiO₂, CeO₂, ZrO₂) were designed to study the conversion of glycol and benzaldehyde to acetals in work by Li et al. [53]. Among them, 20 wt% H₃PW₁₂O₄₀/TiO₂ catalysts have demonstrated the highest activity, with more than 90% acetal yield. Recently, Castanheiro et al. [54] reported the use of tungstophosphoric acid encapsulated on a NaY zeolite in the acetalization of glycerol with an aromatic aldehyde (namely cinnamaldehyde), yielding a glycerol conversion of 89%, after 5 h of reaction at 100 °C and a molar ratio of glycerol:cinnamaldehyde of 1:2.25. Furthermore, the catalyst NaY-tungstophosphoric acid exhibited good catalytic stability.

A magnetite-based functionalized acid catalyst for acetal production using ethylene glycol and benzaldehyde was studied by Hosseini and Masteri Farahani [55]. They designed a catalyst with double acid type groups (sulfonic acid and phosphotungstic) that proved to be efficient in the reaction of benzaldehyde with ethylene glycol in the reaction condition: benzaldehyde to ethylene glycol molar ratio = 1/3, solvent: cyclohexane, 6.66 g cat/mol benzaldehyde at 90 °C. After 2 h of reaction, the benzaldehyde conversion reached 97%. Catalyst reuse studies were also performed, and the reported results indicated that the catalyst maintained its stability after at least four cycles of reaction. This paper presents important findings, but the studies are limited to fixed parameters without optimization of the process parameters.

Rajkumari and coworkers [56] studied the ultrasound-assisted synthesis of solketal using glycerol and acetone over a sulfonic acid-functionalized magnetic catalyst. After synthesis and characterization of the catalyst, they also performed optimization studies on the reaction to improve the five-membered solketal ring yield, varying the catalyst amount, glycerol:acetone molar ratio, reaction time, agitation conditions, and temperature. They reached a high conversion of glycerol (97%) with 100% solketal selectivity, using ultrasound-assisted synthesis, at room temperature, in a relatively short reaction time (15 min) with a glycerol:acetone molar ratio of 1:5 and a catalyst amount of 5 wt% wt. Glycerol.

Given the versatility of heteropolyacids as catalysts, the unique properties of magnetite as potential catalyst support, as well as the importance of glycerol valorization, a goal of the present work is to evaluate the catalytic activity of tungstophosphoric acid immobilized on silica coated magnetite for the synthesis of aromatic glycerol acetals—an important class of fine chemicals derived from glycerol. The reactions were conducted using aromatic aldehydes; the parameters selected for the optimization of the process were: reaction temperature, reactant molar ratio, and catalyst amount. Another goal of this work is to optimize the process parameters of aromatic acetals synthesis over the prepared catalyst using the Box–Behnken experimental design.

2. Results and Discussions

2.1. Catalyst Characterization

The SEM images of Fe₃O₄, Fe₃O₄@SiO₂, and Fe₃O₄@SiO₂@HPW are shown in Figure 1a–c. The catalyst and support are composed of non-uniform crystals with an irregular surface and particles of various sizes, which are composed of fine particles that are strongly agglomerated (due to the attractive forces between the magnetic particles). The presence of W and P (from HPW) in the structure of the catalyst was confirmed by EDS analysis (Figure 2) and by energy dispersive X-ray spectroscopy (EDX) (Figure 3 and

Table 1. EDS analysis of the Fe₃O₄@SiO₂@HPW catalyst.

Element	Weight %	Atomic %
O K	59.90	77.81
SiK	22.16	16.39
P K	0.11	0.07
FeK	14.32	5.33
W L	3.51	0.40

) obtained from SEM analysis of the catalyst.

Figure 4 shows the TEM images of the Fe₃O₄@SiO₂@HPW catalyst. The sample is composed of a large number of congeries with a near-spherical shape. In addition, the TEM images confirm the formation of the core-shell structure of the catalyst. TEM images of the Fe₃O₄@SiO₂@HPW catalyst showed that the size of the particles was less than 50 nm.

FTIR spectra of the synthesized materials are indicated in Figure 5. The FTIR spectrum of the synthesized Fe₃O₄ nanoparticles presents two peaks at 630 cm⁻¹ and 590 cm⁻¹ corresponding to the bonding of Fe-O [57,58] and two more bands in the wavenumber of 3400 cm⁻¹ and 1630 cm⁻¹ corresponding to the stretching and bending of H₂O molecules respectively [59]. The presence of a silica layer on the surface of magnetite is demonstrated in FTIR spectra by the vibration bands of Si-O bonds and silanol groups at wavenumbers 770 cm⁻¹ and by the broad band at 1090 cm⁻¹ [60,61,62]. The main FTIR bands of HPW are at about 1100 cm⁻¹ (stretching frequency of P-O in the central PO₄ tetrahedron), 990 cm⁻¹ (terminal bands for W=O in the exterior WO₆ octahedron), 890 cm⁻¹ and 800 cm⁻¹ (bands for the W-Ob-W and W-OcW bridge). In the FTIR spectrum of the Fe₃O₄@SiO₂@HPW catalyst, some bands characteristic of the Keggin structure cannot be observed because they overlap with the bands of silica. All the bands that appeared in the FTIR spectra are in agreement with those reported in the literature [63,64,65].

The Fe₃O₄@SiO₂@HPW (fresh) and S-Fe₃O₄@SiO₂@HPW (spent) FTIR spectra are similar; the characteristic bands of the heteropolyacid are still observed in the FTIR spectrum of the S- Fe₃O₄@SiO₂@HPW catalyst, indicating that the Keggin structure of HPW is maintained after catalyst testing.

The Brønsted or Lewis nature of the acidic sites was determined by FTIR pyridine adsorption (Figure 6). The infrared spectra of pyridine at room temperature exhibits three absorption bands at 1580, 1480, and 1438 cm⁻¹. The Brønsted acid sites correspond to the 1580 cm⁻¹ band (pyridinium ion), and the Lewis acid sites to the 1438 cm⁻¹ band, attributed to coordinatively bounded pyridine [66]. The peak around 1480 cm⁻¹ corresponds to the overlapping vibrations of pyridine adsorbed on both Lewis and Brønsted acid sites. Hence, the concentration of both the Brønsted and Lewis acid sites was calculated from their corresponding band intensities and extinction coefficients of each type of site using the equation reported by Isernia [67]. The results are presented in

Table 2. The concentration of acidic sites.

Catalyst	C Lewis (CL) mmol/g	C Lewis (%)	C Bronsted (CB) mmol/g	C Bronsted (%)	Total Concentration (CL + CB)	CL/CB Ratio
					mmol/g
Fe3O4@ SiO2@HPW	39.92	63.08	23.36	36.92	63.28	1.7
S-Fe3O4@ SiO2@HPW	38.35	67.57	18.41	32.43	56.76	2.08

.

The XRD patterns of the catalyst and magnetic support were prepared in 2θ ranging from 10 to 80° (Figure 7a,b—magnified from 10 to 30°). The Fe₃O₄ sample exhibits the main peaks at 2θ = 30.1, 35.4, 43.1, 53.4 56.9, 62.5, and 74.9° characteristic for magnetite [68,69]. Similar peaks may be observed in the XRD diagrams of Fe₃O₄@SiO₂ support and Fe₃O₄@SiO₂@HPWcatalyst, confirming that the cubic phase of magnetite is kept after silica coverage and HPW molecule immobilization. The silica layer did not change the structure of Fe₃O₄, except for the presence of a broad weak peak at 2θ range between 20° and 25°, indicating the presence of amorphous silica in the structure of the Fe₃O₄@SiO₂ and Fe₃O₄@ SiO₂@HPW samples [70,71,72] (Figure 7b). The pattern for the Fe₃O₄@SiO₂@HPW catalyst exhibits supplementary broad signals, characteristic to the lines of HPW structure at around 2θ = 15.1°, 18.3°, 21.5°, 23.2°, 26°, and 28° (Figure 7b), in agreement with published data [73,74,75].

Figure 8 displays the N₂ adsorption-desorption isotherms for the Fe₃O₄, Fe₃O₄@SiO_2, and Fe₃O₄@SiO₂@HPW catalysts. In the silica-coated magnetite samples, Fe₃O₄@SiO₂ and Fe₃O₄@SiO₂@HPW, the nitrogen adsorption-desorption isotherms present type IV isotherms with an H₃ hysteresis for materials with crack-type pores or flat particles (aggregates), according to the International Union of Pure and Applied Chemistry (IUPAC) classification [76]. Fe₃O₄ presents an H₁ hysteresis type, which is indicative of a material with cylindrical and uniform pores.

From the BJH pore-size distribution (Figure 9), multimodal type of pores as meso- (35 nm) and macro-pores (80–125 nm) are present in the Fe₃O₄@SiO₂ and Fe₃O₄@SiO₂@HPW samples. The bare Fe₃O₄ exhibits uniform pore size distribution with the unimodal mesoporous structure (10–40 nm mesopores).

Table 3. Textural characteristics of the bare Fe₃O₄, Fe₃O₄@SiO₂ support, and Fe₃O₄@SiO₂@HPW catalyst.

Sample	Specific Area (BET) m2/g	Pore Volume 2–300 nm (BJH Desorption) cm3/g	Mean Diameter 2–300 nm (BJH Desorption) nm
Fe3O4	73.84	0.36	17.21
Fe3O4@SiO2	56.09	0.32	34.82
Fe3O4@SiO2@HPW	40.86	0.24	36.4

presents the values of the parameters that define the textural properties of the catalyst, namely the BET-specific area, the meso- and macro-pore volume, and the mean diameter. In the Fe₃O₄@SiO₂@HPW catalyst, the BET surface area (40.86 m²/g) and pore volume (0.24 cm³/g) were decreased due to immobilization HPW molecules and structural rearrangement after catalyst calcination through -OH group elimination.

The thermal stability of the catalyst was evaluated using thermogravimetric analysis to determine whether the Keggin structure of the active phase of the catalyst was preserved after testing. The TGA profile shown in Figure 10, corresponding to the unsupported HPW, indicates that the main weight loss (approx. 6.3%) occurred until 200 °C. The DTG (derivative thermogravimetry) curve revealed four peaks of weight loss. In the first stage corresponding to peaks at 76 °C and 145 °C, the loss of physically absorbed water is observed. At 195 °C, crystallization water loss is observed, and in the last stage, corresponding to a peak of 540 °C, the Keggin structure starts decomposing.

The TGA profile corresponding to the fresh Fe₃O₄@SiO₂@HPW catalyst and spent S- Fe₃O₄@SiO₂@HPW catalyst, presented in Figure 11, exhibits similar behavior, with the difference that the DTG curve of the spent catalyst revealed a higher mass loss in the stage corresponding to water loss, thus demonstrating that the catalyst structure is still preserved after testing.

2.2. Effect of Process Variables

Instead of using the OFAT (one-factor at a time) approach, the influence of process variables was studied using a Box–Behnken experimental design. The desired major products in glycerol acetalization with benzaldehyde are the cyclic six-membered atoms (cis/trans-2-phenyl-1,3-dioxan-5-ol) and cyclic five-membered atoms (2-phenyl-1,3-dioxolan-4-yl) methanol and 2-phenyl-1,3-dioxolane). Moreover, the formation of non-cyclic acetal (methyl-3-hydroxy-3-phenylpropanoate), benzoic acid, and other minor products may also be observed (Figure 12). Yamamoto and co-workers [72] presented the reaction mechanism between benzaldehyde and glycerol and evidenced this possible reaction pathway with the formation of 2-phenyl-1,3-dioxolane-4-methanol which, in the presence of water and benzaldehyde, could form 2-phenyl-1,3-dioxolane as well as benzoic acid and methanol [72].

A similar mechanism was proposed by Castanheiro [78] for the acetalization of glycerol with citral over a heteropolyacid immobilized on KIT-6. According to this study, it may be initiated by an interaction between an oxygen atom of a carbonyl group and Bronsted acid sites. Therefore, the formation of a hemiacetal in the reaction of benzaldehyde with glycerol may be followed by eliminating a water molecule and forming a carbocation. The latter suffers an attack involving the inner or terminal hydroxyl group of a glycerol molecule to ensure the formation of cyclic acetals. A similar mechanism was evidenced by Patel and Pithadia [79].

In our study, the presence of these chemical compounds in the chemical mixture was proven by GC-MS analyses (see Supplementary Information).

The determination of proper operating conditions is essential when dealing with condensation reactions. In the current case, the goal is to improve the yields in the cyclic acetals. At low temperature (80 °C) and a catalyst amount of 3%, the increase in the initial B/Gly mole ratio has a small effect both on glycerol conversion (61.11% at B/Gly = 1 vs. 60.41% at B/Gly = 1.2) and on yields in cyclic acetals (see experiments 5 and 12 in

Table 4. Box–Behnken design matrix.

Experiment	X1	X2	X3	Y1	Y2	Y3	Y4	Y5	Y6
1	1	0	−1	1.49	16.54	27.37	24.97	8.13	78.51
2	0	−1	1	2.25	23.5	15.27	14.73	11.97	67.72
3	0	1	−1	2.89	27.38	17.18	17.58	15.13	80.16
4	1	0	1	1.04	11.78	31.64	29.58	6.27	80.31
5	−1	0	1	2.64	12.44	20.03	19.01	6.3	60.41
6	0	−1	−1	2.94	21.62	15.75	15.59	11.56	67.46
7	0	0	0	2.81	25.39	16.85	20.01	12.05	77.11
8	0	0	0	3	24.44	18.42	18.23	13.13	77.22
9	0	0	0	3.1	25.02	18.07	18.04	13.67	77.91
10	−1	−1	0	2.83	9.05	20.59	19.01	4.69	56.18
11	−1	1	0	2.8	14.41	17.71	18.22	8.53	64.67
12	−1	0	−1	3.45	10.66	20.84	20.31	5.84	61.11
13	0	1	1	2.63	28.22	16.72	17.03	15.84	80.44
14	1	1	0	1.06	13.2	33.81	31.72	7.25	87.05
15	1	−1	0	1.21	15.69	26.28	24.12	7.68	75.5

). By increasing the temperature to 120 °C for a catalyst amount of 3%, an increase of B/Gly from 1 to 1.2 has a larger influence on process performance in terms of 6-member acetal (an increase from 27.34% to 31.64%) and 5-member acetal (an increase from 24.97% to 29.58%). The increase in the B/Gly ratio at high temperatures also reduces the selectivity in secondary products (benzoic acid and non-cyclic acetal) (see experiments 1 and 4 in Table 4). This behavior was also emphasized by Kulkarni and Arvind [80] over a sulfated ceria-zirconia catalyst.

At 120 °C and a B/Gly ratio of 1.1, the increase of the catalyst amount in the mixture is leading, as expected, to a higher glycerol conversion and an increased selectivity in the cyclic acetals (for the 6-member one, from 26.28% to 33.81% whereas for the 5-member one from 24.12% to 31.72%) (see experiments 14 and 15 in Table 4).

The temperature increase has, as expected, an important effect on the increase of glycerol conversion, leading to a decrease in by-product selectivity and an increase in 5-member and 6-member acetals (see pair of experiments 4–5 and 11–14 in Table 4). This is explained by the increase of reaction rates with temperatures towards the formation of cyclic acetals.

2.3. Modeling of Process Performance

In order to evaluate the influences of the main factors on glycerol conversion and selectivity in acetals, a polynomial regression model expressed by Relation (1) was implemented, and its accuracy was evaluated using the analysis of variance techniques.

$$y={\beta }_0+{\displaystyle \sum _{i=1}^k{\beta }_i{X}_i}+{\displaystyle \sum _{i=1}^k{\beta }_{ii}{X}_i^2+{\displaystyle \sum _{i=1}^{k-1}{\displaystyle \sum _{j=2}^k{\beta }_{ij}{X}_i{X}_j}}+\epsilon }$$

(1)

The adequacy of the model was assessed by comparing the calculated Fisher test value (F-value) with the theoretical one (Ft(p-1, n-p)) for a chosen risk of 5% for a number of experiments of n and p terms in the model [81]. As mentioned in the literature regarding the statistical analysis, the parameter p-value parameter should be below 5% to consider the fact that each of the independent variables in the model is useful in predicting the system’s response [82]. The coefficients of the models as well as statistical tests were evaluated using Design-Expert Version 8.0.7.1.

For each response, the analysis of variance data calculated based on the Box–Behnken design matrix is presented in

Table 5. ANOVA statistical parameters evidencing the adequacy of the models.

Model	Dependent Variable	Determination Coefficient, R2	Adjusted R2	Calculated F-Value	p-Value
Model 1	Y1	0.9793	0.9678	85.08	0.001
Model 2	Y2	0.9803	0.9656	64.49	0.001
Model 3	Y3	0.9788	0.9630	61.65	0.001
Model 4	Y4	0.9189	0.8865	28.33	0.001
Model 5	Y5	0.9639	0.9494	66.72	0.001
Model 6	Y6	0.9911	0.9844	148.39	0.001

. The estimated coefficients for each selectivity (Y₁ to Y₅) as well as for glycerol conversion (Y₆), together with their corresponding confidence intervals, are given in Relations (2)–(7).

$$\begin{array}{l}{Y}_1=2.94\left(1\pm 0.054\right)-0.865\left(1\pm 0.133\right)\cdot X_1+0.0188\left(1\pm 6.132\right)\cdot X_2-0.2763\left(1\pm 0.417\right)\cdot X_3\\ -0.754\left(1\pm 0.224\right)\cdot X_1^2-0.2319\left(1\pm 0.729\right)\cdot X_2^2\end{array}$$

(2)

$$\begin{array}{l}{Y}_2=25.08\left(1\pm 0.043\right)+1.33\left(1\pm 0.757\right)\cdot X_1+1.67\left(1\pm 0.604\right)\cdot X_2-0.0325\left(1\pm 31\right)\cdot X_3\\ -1.96\left(1\pm 0.729\right)\cdot X_1\cdot X_2-1.64\left(1\pm 0.865\right)\cdot X_1\cdot X_3-12.11\left(1\pm 0.122\right)\cdot X_1^2\end{array}$$

(3)

$$\begin{array}{l}{Y}_3=16.89\left(1\pm 0.058\right)+4.99\left(1\pm 0.184\right)\cdot X_1+0.941\left(1\pm 0.979\right)\cdot X_2+0.315\left(1\pm 2.92\right)\cdot X_3\\ +2.6\left(1\pm 0.5\right)\cdot X_1\cdot X_2+1.27\left(1\pm 1.02\right)\cdot X_1\cdot X_3+7.89\left(1\pm 0.171\right)\cdot X_1^2\end{array}$$

(4)

$$\begin{array}{l}{Y}_4=17.32\left(1\pm 0.081\right)+4.23\left(1\pm 0.309\right)\cdot X_1+1.39\left(1\pm 0.943\right)\cdot X_2+\\ +2.1\left(1\pm 0.88\right)\cdot X_1\cdot X_2+6.05\left(1\pm 0.31\right)\cdot X_1^2\end{array}$$

(5)

$$\begin{array}{l}{Y}_5=13.34\left(1\pm 0.051\right)+0.496\left(1\pm 1.29\right)\cdot X_1+1.36\left(1\pm 0.47\right)\cdot X_2+\\ -1.07\left(1\pm 0.85\right)\cdot X_1\cdot X_2-6.5\left(1\pm 0.144\right)\cdot X_1^2\end{array}$$

(6)

$$\begin{array}{l}{Y}_6=77.41\left(1\pm 0.019\right)+9.88\left(1\pm 0.094\right)\cdot X_1+5.68\left(1\pm 0.163\right)\cdot X_2+0.205\left(1\pm 4.52\right)\cdot X_3\\ -5.21\left(1\pm 0.262\right)\cdot X_1^2-1.35\left(1\pm 1.01\right)\cdot X_2^2-2.12\left(1\pm 0.641\right)\cdot X_3^2\end{array}$$

(7)

The main, binary, and quadratic interactions between the process factors can be evaluated based on the estimated coefficients and their confidence intervals in Relations (2)–(7). The statistical significance of the coefficients was evaluated through the p-value for each coefficient, and the ones that did not undergo proper significance were eliminated. Even so, few of the non-significant coefficients have been preserved in order to keep the hierarchy of the model (as they are involved in the binary/quadratic interactions).

Based on preliminary experiments, the ranges for the main factors influencing the process were chosen between 80 and 120 °C for temperature, a catalyst amount between 1 and 5% (by weight), and an initial benzaldehyde to glycerol ratio between 1 and 1.2. The coded values of the factors corresponding to a Box–Behnken experimental design matrix together with the process responses are given in Table 4.

The experimental design matrix consists of 15 experiments, among which 3 are in the center of the experiment (experiments 7 to 9 in Table 4). The natural and coded independent variables together with their maximum and minimum limits are given in

Table 6. Natural and coded values of the independent variables in the experimental design.

Factor	Value/Level
	Low		Medium		High
	Natural (Z)	Coded (X)	Natural (Z)	Coded (X)	Natural (Z)	Coded (X)
Temperature (°C), X1	80	−1	100	0	120	+1
Catalyst amount (wt %), X2	1	−1	3	0	5	+1
Initial B/Gly ratio (mol/mol), X3	1	−1	1.1	0	1.2	+1

.

As mentioned in the experimental design literature [83], when working with coded variables, the magnitude of the coefficients gives an idea about the influence of the factor on the process response. Moreover, a positive coefficient means an increase in response to the factor, whereas a negative coefficient leads to a decrease in response to the increase of the factor. The models proposed in Relations (2)–(7) are rendered adequate, based on the correlation coefficients given in Table 5 (with values close to unity) and confirmed by the visual representation of the calculated vs. experimental data in the parity diagram in Figure 13. All these arguments prove a reasonably good representation of the process factors proposed by the mathematical models.

One aspect that must be emphasized is that the amount of benzoic acid obtained by hydrolysis of benzaldehyde in the reaction environment is relatively low. Figure 14 shows that the increase in temperature reduces the selectivity in benzoic acid (Y₁), a selectivity below 1% being obtained using a 5% catalyst and a B/Gly initial ratio of 1.2.

The values of the factors X₁–X₃ in

Table 7. The optimized parameters for maximum values of glycerol conversion and selectivity in acetals.

X1	X2	X3	Y1	Y2	Y3	Y4	Y5	Y6	Cycl
1	1	0.5	0.961	13.15	33.13	31.08	7.62	85.95	1
			1.11	12.46	32.85	30.12	6.38	82.91	2
1	1	0.5	1.23	12.25	31.85	29.87	6.08	81.31	3

have been obtained using the proposed mathematical models. The goal of this optimization study is the simultaneous maximization of glycerol conversion and selectivity in the 5-member and 6-member acetals (Y₂, Y₃, and Y₄), together with the simultaneous minimization of by-products (benzoic acid (Y₁) and non-cyclic acetals (Y₅)). The process factors ensuring the maximum values of glycerol conversion and selectivity in cyclic acetals are listed in Table 7.

A maximum glycerol conversion (85.95%) was obtained at 120 °C together with a 5% catalyst amount and a 1.15 molar ratio between glycerol and benzaldehyde. These parameters give a selectivity in cyclic acetals (Y₂, Y₃, and Y₄) of 78.56%.

In order to prove the reusability of the catalyst, three subsequent runs were carried out in the optimized conditions given in Table 7. These values show a slight decrease in glycerol conversion with no significant change in glycerol acetals. Similar results were reported by Yamamoto et al. [77] on a cationic acidic resin and by Samudrala et al. [84] over a mesoporous MCM-41 supported iron (III) chloride. The stability of the catalyst was proven by carrying out the FTIR analysis for pure, supported, and reused catalysts in the three cycles. The results are presented in Figure 15. The presence of immobilized HPW in the Fe₃O₄@ SiO₂@HPW catalyst after each cycle of utilization was proven mainly by the presence of the characteristic peaks of HPW at the 990 cm⁻¹ (terminal bands for W=O in the exterior WO₆ octahedron), 890 cm⁻¹, and 800 cm⁻¹ (bands for the W-Ob-W and W-OcW bridge), suggesting that the kegging structure is still maintained. The peaks in the FTIR spectra corresponding to 1400–1600 cm⁻¹ are from the solvent used to wash the catalyst after each run.

The maximum value of glycerol conversion obtained in our study was about 86%, being among the highest values reported in the literature (

Table 8. Acetalization of glycerol with benzaldehyde over other solid acid catalysts.

Catalyst	Reaction Conditions	Conversion/Yield (%)	Selectivity		Ref.
			Dioxolane	Dioxan
p-toluenesulfonic acid	140 °C, 15 min, microwave (600 W), Gly/B = 1/1	67% (conv.)	47	53	[85]
20 % MoO3/SiO2	100 °C, toluene as a solvent, 8 h, mcat = 10 wt %, Gly/B = 1.1/1	72% (conv.)	40	60	[86]
MoOx/TiO2–ZrO2	100 °C, solventless, 30 min, mcat = 5 wt %, Gly/B = 1/1	72% (conv.)	49	51	[87]
Amberlyst-36	61.2 °C, 0.1 MPa, refluxing chloroform as a solvent, 4 h, mcat = 0.1 g, Gly/B =1.1/1	94% (yield)	37	57	[88]
sulfonated graphene (GR-SO3H)	100 °C, 14 h, mcat = 25 mg, Gly/B = 1/10	92% (yield)	62	38	[89]
Ta/W (heteropoly acid)	25 °C, 1 h, 0.01 mmol (based on the Bronsted acid sites on the catalyst), Gly/B = 1/2	99% (conv.)	72	28	[90]

), and the selectivity in cyclic acetals was also reasonably good.

Given the activity of our catalyst, the synthesized Fe₃O₄@ SiO₂@HPW catalyst was also tested in the acetalization of glycerol with other aromatic aldehydes: cinnamaldehyde and p-methyl-benzaldehyde, and the detected conversion of glycerol was 78% and 85%, respectively; the results are summarized in

Table 9. Acetalization of glycerol with varied aromatic aldehydes on the catalyst.

Entry	Aldehyde	Conversion (%)	Selectivity
			5-Member Ring	6-Member Ring
1	Cinnamaldehyde	78	57	30
2	p-methyl-benzaldehyde	85	59	32
3	p-hydroxy-benzaldehyde	-	-	-
4	vanillin	2	-	-

. However, in these conditions, the catalyst has a low activity for the acetalization of glycerol with vanillin and 4-hydroxybenzaldehyde (entries 3 and 4 in Table 9), which might be attributed to steric hindrance and mass transfer limitation. The selectivity in 5 and 6-member ring products for cinnamaldehyde and p-methyl-benzaldehyde is above 87%.

Reaction conditions: 100 °C, heptane as solvent, reaction time of 120 min, m_cat = 5 wt%, glycerol/aldehyde = 1/1

3. Materials and Methods

3.1. Materials

FeCl₃ × 6H₂O (Carl Roth, Karlsruhe, Germany, min. 98%), FeCl₂ × 4H₂O (VWR Chemicals, Vienna, Austria, min. 99%), TEOS (tetraethyl orthosilicate) -VWR Chemicals, GPR Reactpur, NH₄OH 25% (Chimreactiv, Bucharest, Romania), Glycerol 99% extra pure (Merck KGaA, Darmstadt, Germany), Benzaldehyde 98%+ (Merck KGaA), Ethanol 96% (Chimreactiv), and Tungstophosphoric acid hydrate (Merck KGaA).

3.2. Catalyst Preparation

Preparation of magnetite nanoparticles (Fe₃O₄): The magnetite nanoparticles (Fe₃O₄) were synthesized by chemical co-precipitation of chloride salts of Fe³⁺ and Fe²⁺ (in a 2:1 molar ratio) [91,92,93]. Specifically, 27.033 g of FeCl₃·6H₂O and 9.94 g of FeCl₂·4H₂O were dissolved in 350 mL of distilled water. The solution was stirred at room temperature, and 35 mL of ammonia solution (25 wt%) was added dropwise quickly. The black nanoparticles immediately precipitated, and the agitation of the mixture continued for 30 min. After completion of the reaction, the black nanoparticles were separated from the reaction system with a magnet and washed with distilled water until a negative chloride reaction (tested with AgNO₃) was produced. Finally, the obtained black nanoparticles of Fe₃O₄ were washed with ethanol and dried at 100 °C.

Synthesis of silica-coated magnetite nanoparticles (Fe₃O₄@SiO₂): Nanoparticles of Fe₃O₄ were coated with silica using the modified Stöber method [94], which implies adding the Fe₃O₄ nanoparticles into a TEOS solution. First, the TEOS solution was prepared by adding 10 mL TEOS into a 200 mL mixture of ethanol-water (3:1 vol%), and the pH was fixed between 11 and 12 with an ammonia solution (25%). After approximately 20 min of mixing, when the solution started to become cloudy, 4 g of Fe₃O₄ nanoparticles were added to the solution, and the stirring continued for 4 h. All steps were performed at room temperature. The produced brown-grey sediment was separated from the solution with a magnet, washed with ethanol, and dried at 100 °C in air.

Immobilization of HPW: The HPW was immobilized onto the Fe₃O₄@SiO₂ using the impregnation method. To a solution of 5 g of Fe₃O₄@SiO₂ particles well dispersed in 50 mL of ethanol was added 1 g of HPW, and the mixture was continuously stirred for 24 h at room temperature. The resultant Fe₃O₄@SiO₂@HPW particles were dried at 100 °C and calcinated at 300 °C for 3 h (heating rate of 10 °C/min). The calcination temperature was selected based on the thermal stability of the HPW to preserve the HPW Keggin structure [95].

3.3. Catalyst Characterization

The catalyst was characterized using different analytical methods: FTIR spectroscopy, SEM and XRD analyses, acidity evaluation, N₂ adsorption/desorption, and thermogravimetric analysis. Meanwhile, FTIR spectroscopy, acidity evaluation, and thermal analysis were employed for the spent catalyst to confirm its stability.

FTIR spectroscopy was performed on an IR Tracer-100 (Shimadzu Europa GmbH) Shimadzu Fourier-Transform Infrared Spectrophotometer using the KBr pellet technique, and the scanning wavelength range was 4000–400 cm⁻¹. The acidic properties of the catalyst (the concentrations of Brønsted and Lewis acid sites) were evaluated through the FTIR analysis of pyridine adsorption by the ring vibration of pyridine detected in the frequency range of 1400–1600 cm⁻¹.

The XRD measurements were run at room temperature using a Shimadzu 6000 XRD diffractometer (Shimadzu Corporation, Kyoto, Japan) with a monochromator of graphite for the Cu-Kα radiation. The source power was operated at a voltage of 40 kV, an electrical current of 30 mA, and the scan speed was 2°/min. The PDF2 reference library was used for qualitative identification of the crystalline phases.

The surface area, pore volume, and pore size were measured by nitrogen adsorption-desorption isotherms at 77 K using a Micromeritics Tri Star II 3020 instrument (Unterschleissheim, Germany). Firstly, the sample was decontaminated under nitrogen flow at 100 °C for 24 h. The surface area was determined using the BET (Brunauer–Emmett–Teller) method. The micropore area and volume were obtained by applying the t-plot method. For the determination of meso- and macro-pore area and volume, we applied the BJH method on the isotherm desorption branch.

Thermogravimetric analyses (TGA) of the catalyst were recorded on a TGA/-IST (Thermal Analysis System TGA 2, METTLER TOLEDO, Greifensee, Switzerland) in the 25–600 °C temperature range, in a nitrogen atmosphere, with a heating rate of 10 °C/min.

External surface, morphology, and structure of catalysts were visualized by Scanning Electron Microscopy (SEM) using a QUANTA 133 Electron Microscope (FEI Company, Hillsboro, OR, USA) and a transmission electron microscopy (TEM) device coupled with an EDX detector (FEI Tecnai F20G2 TWIN Crio-TEM and X-MaxN 80T detector).

3.4. Catalyst Testing Procedure

A typical acetalization of glycerol with benzaldehyde was carried out in a 50 mL stirred three-neck flat-bottom flask, connected to a condenser in order to condense and reflux all the vapors keeping the reaction volume approximately constant. The reaction temperature was controlled by an external thermostat containing a thermocouple placed inside the reacting mixture. In a typical experiment, the flask was first charged with 0.1 moles glycerol, 0.1 moles benzaldehyde, 10 mL of solvent (hexane, heptane, or toluene), 0.46 g catalyst, and heated to the desired temperature. After the expected reaction time, the flask was cooled down to room temperature, and the solid catalyst was separated from the reaction mixture with a magnet, washed with distilled water and ethanol, and then kept for future recycling use.

The azeotropic distillation used in the esterification reaction leading to water elimination by creating an azeotrope mixture with an appropriate solvent is a technique widely used in the literature [23]. This was the reason for using different solvents in our experiments.

The reaction system formed biphasic layers (if toluene was used as solvent) or triphasic layers (if hexane/heptane was the used solvent). Owing to the self-separation characteristics, the upper/middle organic liquid layer containing products and reactant (mainly aldehyde) was easy to separate from the unreacted glycerol (the bottom layer).

The products from the reaction mixture were analyzed with a GC/MS 7000 Triple Quad MS (Agilent Technologies, Santa Clara, CA, USA) system equipped with an HP-FFAP (30 m, 250 mm, 0.25 mm) column and He as the carrier gas, with a volumetric flow of 1 mL/min.

3.5. Evaluation of Catalyst Performance Using Experimental Design Techniques

The performance of the acetalization of glycerol with benzaldehyde over the ferromagnetic catalyst was evaluated using a Box–Behnken experimental design, where 3 factors (independent variables) were considered: reaction temperature (X₁), catalyst amount (X₂), and the molar ratio between benzaldehyde and glycerol in the initial mixture (X₃). The reaction time considered for this evaluation was 120 min. The responses were glycerol conversion (Y₆) and the selectivity in the main reaction products: benzoic acid (Y₁), 2-phenyl-1,3-dioxolan-4-yl)methanol (Y₂), 2-Phenyl-1,3-dioxan-5-ol (Y₃) 2-phenyl-1,3-dioxolane (Y₄) and methyl 2-hydroxy-3-phenylpropanoate (Y₅).

4. Conclusions

In this study, a ferromagnetic heteropolyacid catalyst was successfully synthesized by immobilization of tungstophosphoric acid (H₃PW₁₂O₄₀; HPW) on a silica-coated magnetite, obtained by co-precipitation method. The characteristics of the catalyst were proven through SEM-EDX, XRD, and FT-IR analysis, which present the chemical and structural information. The EDS showed a good distribution of the active species on the catalyst surface, whereas TEM analysis showed an HPW particle size of 50 nm. The concentration of Brønsted and Lewis acid sites in the catalyst was 63.28 mmol/g, and the specific surface area was 40.86 m²/g. The so-prepared catalyst possesses not only easy separation property due to the magnetic core but also a superior catalytic activity for glycerol acetalization with benzaldehyde, with a maximum glycerol conversion of 96.1% and a cumulated acetal selectivity of about 78.36 %, in the optimized conditions (T = 120 °C, 5% wt catalyst and 1.15 glycerol/benzaldehyde molar ratio). The reuse in three subsequent runs proved the stability of the catalyst, with a slight decrease in glycerol concentration and no significant change in the acetals selectivities. The catalyst also exhibits excellent activity for glycerol acetalization with other aromatic aldehydes, such as cinnamaldehyde, p-methyl-benzaldehyde, p-hydroxy-benzaldehyde, and vanillin.