3.1. Characterization of heteropoly salts
The characterization of all the metal transition-exchanged catalysts used in this work was previously discussed by us in two previous papers; the first one , where the metal exchanged catalysts were used in glycerol esterification reactions, and the second one, where the Aluminum salts were used as catalysts in olefin oxidations . However, in this work, to check if the catalysts were correctly synthesized, we carried out UV-Vis and infrared spectroscopy analyses, and acidity measurements. All the results are presented in the supplemental material (Figures 1SM to 9SM; Tables 1SM).
The typical absorption bands of the Keggin anion of H3PW12O40 appeared at 1080, 990, 890, and 790 cm−1 wavenumbers, and were assigned to the vibrations υass (P–Oa), υass (W–Od), υass (W–Ob–W) and υass (W–Oc–W) bonds, respectively [46, 47]. These bands appear practically at the same wavenumbers in the FT-IR spectra of meta-exchanged salts, indicating that the structure of the Keggin anion was preserved after the protons exchange. These data agree with the literature . The absence of splitting in-band attributed to the vibration of the P-O bond assure that no lacunar anion was formed and that the primary structure remained almost intact after the synthesis. The same effect was observed when silicotungstic acid had its protons exchanged by the same actions used herein .
The main absorption bands that are typical of Keggin heteropolyanions were also observed in the infrared spectra of all the metal-exchanged phosphotungstic acids, as unequivocally demonstrated a comparison of their spectra with the phosphotungstic acid (Figure 1SM-3SM).
The UV spectra were recorded at the range of 200-750 nm, however, the main absorption bands only appeared within intervals 190-325 nm (Figure 2).
The UV spectra of these compounds display two characteristic absorption bands, attributed to the charge transfer from Oxygen atoms to Tungsten since that W atom has electronic configuration “d0” . The first, with maximum absorption placed at low wavenumber (ca. 190 nm) was assigned to the charge transfer from terminal Oxygen to Tungsten. This band was almost not affected by the presence of metal cations. Conversely, a second absorption band, noticed around 260 nm was assigned to charge transfer from bridges Oxygen atoms to W atom . The maximum absorption was slightly toward red, mainly when in the presence of M3+ cations (Figure 2b).
In the supplemental material, we compared the UV-Vis spectra of three Keggin HPAs with Copper-exchanged salts (Figs. 4SM-9SM). While the Uv-Vis spectra of phospho- or silicotungstic acid and Copper salt catalysts are similar, with two bands absorption at 190 nm and 255 nm, the Molybdenum catalysts were different. Only a strong absorption band was noticed, at 220 nm wavelength, with a shoulder at 225 nm. Although this result agrees with the literature, which attributes this band at charge transfer from O2− anions to Mo6+ cations (i.e., probably in bridges Oc-Mo-Oc, likewise in Tungsten heteropoly catalysts), we suppose that the absorption band involving charge transfer from terminal Oxygen atoms was also shifted toward lower wavelength, which is out of spectral range analyzed .
It is noteworthy that when metal-substituted polyoxometalates (POM) are synthesized removing an MO unit (M = W, Mo), a lacunar catalyst is generated, and when in this catalyst is introduced a metal cation into Keggin anion structure, metal-substituted POM salts are obtained, which are also very active photocatalysts (i.e., [PW11M2+O39]5−, M2+ = Mn, Fe, Co, Ni, Cu, Zn) . In the UV-Vis spectra of these types of doped-metal heteropoly salts, the presence of the transition metal ions in octahedral coordination leads to risen “d-d” transitions absorption bands, in the region from 450 to 780 nm . However, comparatively to the charge transfer bands, “d-d” transitions lead to very weak absorption bands. Consequently, to detect these bands highly concentrated solutions were prepared and their UV-Vis spectra are shown in Figure 3. Aiming for higher clarity, only the spectral region from 350 to 1000 nm was presented.
The absence of these bands in the spectra of phosphotungstic acid and their Aluminum phosphotungstate salt can be explained by the absence of “d” electrons. Conversely, the intensity and position depended on the metal cation. They were visible in the spectra of Ni3/2PW12O40 (380-450 nm), Cr3/2PW12O40 (530-650 nm), Cu3/2PW12O40 (650-900 nm), Co3/2PW12O40 (450-600 nm) [54–56]. The spectra of metal cations with d5 configuration (Mn2+ and Fe3+ ions) do not show bands with a detectable intensity
The potentiometric titration curves of metal-exchanged phosphotungstate are presented in Figure 4. According to Pizzio et. al, the measurements of initial electrode potential (Ei) provide the acidity strength of acid sites: Ei > 100 mV (very strong acid sites), 0 < Ei < 100 mV (strong acid sites), 100 < Ei < 0 (weak acid sites), and Ei < 100 mV (very weak acid sites) . All the synthesized catalysts presented Ei > 100 mV, indicating that they have very strong acid sites. However, they showed different values of Ei, which can be ordered as follow: FePW12O40 ≈ CrPW12O40 > AlPW12O40 > Co3/2PW12O40 ≈ Cu3/2PW12O40 > Ni3/2PW12O40 Mn3/2PW12O40. Expected, the trivalent metal cations were the highest strength of acidity.
All these catalysts have still Brønsted acidity since the Lewis metal cations can react with hydration water molecules present in the heteropolyanion. It can be confirmed by the absorption band placed at 1620 cm−1 wavenumber in the infrared spectra of these salts (Figure 1). In supplemental material, we showed the titration curves of Keggin heteropolyacids (Fig. 10SM). The titration curve of phosphotungstic acid (Figure 10SM) and their AlPW12O40 salt have a sharp decline after the addition of 0.8 mL of n-butylamine. Conversely, the other curves presented a different behaviour, some with a strong fall of potential at the beginning of titration (i.e., Ni3/2PW12O40, Mn3/2PW12O40, Cu3/2PW12O40, Co3/2PW12O40), and other with more gradual diminishing in the Ei value (i.e., CrPW12O40, FePW12O40). Nonetheless, all these curves presented two plateaus, suggesting the presence of sites with different acidity strengths. Literature has described that the presence of acidity in the neutral salts of Keggin HPAs is due to the coordination of metal cations to water molecules, which lead to a release of H+ ions. In addition, we previously demonstrated that the metal cations thyself can consume n-butylamine, due to the presence of empty orbitals.
When we compared the titration curves of Keggin HPAs with those of Cooper exchanged salts (Figs. 10SM and 11SM), we conclude that this modification remarkably modifies the profile of curves; although the Ei values have undergone a slight lowering (ca. 50 mV, at maximum), all the three curves presented two plateaus at different regions, after the Cu(II) doping, suggesting that there now acidity sites with different strength.
3.2. Catalytic tests
3.2.1. Evaluation of different metal phosphotungstate catalysts in HMF synthesis
An initial screening aiming to verify the most active and selective catalyst in the conversion of fructose to HMF was performed, using reaction conditions previously described by the group . The main results are shown in Figure 5. It is important to highlight that at this stage, the reaction conditions are not optimized to achieve the maximum yield. Our objective herein is to evaluate all the catalysts at the same conditions.
The HMF yields varied from 40 to 53%, being the reactions catalyzed by Chromium and Copper phosphotungstate salts those that demonstrate to be more efficient. Initially, we have expected that the catalysts with higher strength of acidity (i.e., with higher Ei values, Figure 4) will achieve higher HMF yields.
Interesting, although the CrPW12O40 catalyst has presented a strong acidity and its reaction has achieved also a high HMF yield, the reactions in the presence of FePW12O40 or AlPW12O40, which also showed a high acidity strength, reached only lower yields (Figure 5). Surprisingly, Cu3/2PW12O40 that had an Ei value almost equal to the Ei measured for AlPW12O40, suggesting that both have a similar acidity strength, achieved the highest yield. It suggests that probably the success of the reaction depend on the other aspects that may be beyond the acidity strength of the catalysts. As the basis of these results, we can conclude that although necessary, the Brønsted acidity strength should not be too strong, to prevent further rehydration of HMF, consequently, the strongest acid will not necessarily be the best catalyst .
It is noteworthy that this reaction occurs in an aqueous phase, therefore, we carried out pH measurements to verify the acidity of the reaction medium after adding the metal exchange salt catalyst (Table 2SM). The pH value of the solutions was 3.3 (i.e., AlPW12O40, FePW12O40), which were partially soluble, and 3.2. for all the other soluble salts. These very close pH values are a consequence of levelling effect of the water. The aqueous medium becomes acidic due to the hydrolysis of metal cations, which generate H3O+ cations (Equations 1 and 2).
3.2.2. Mechanistic insights
It is natural to suppose that the protons should be the main players in these catalytic games and therefore act as the main active sites. Therefore, in Scheme 1 we propose a reaction pathway that can be useful to explain how they have risen into reaction medium and how they can participate in steps that leads to the conversion of fructose to HMF. However, the Copper phosphotungstate has no protons. The literature describes that transition metal salts can be hydrolyzed generating hydronium cations in aqueous solutions (Equations 1 and 2). We suppose that the same occurs with Copper phosphotungstate (Scheme 1). Afterwards, the reaction probably proceeds such as the typical Brønsted -acid-catalyzed reactions (Scheme 1).
The first step begins with protonation of the hydroxyl bound to the anomeric carbon of fructose, followed by the loss of a water molecule, and the formation of the oxonium ion (intermediate I, Scheme 1). Subsequently, water removes β-hydrogen from the oxonium ion, forming an enol (intermediate II, Scheme 1) that is in tautomeric equilibrium (keto-enol) (intermediate III, Scheme 1). From the intermediate aldehyde, there is the protonation of a second hydroxyl (intermediate IV, Scheme 1), followed by the loss of another water molecule, leading to the formation of an α,β-unsaturated aldehyde (intermediate V, Scheme 1). Finally, the protonation of the secondary hydroxyl (intermediate VI, Scheme 1), followed by the loss of a dehydration e leads to the formation of the HMF molecule (Scheme 1).
Gomes et al. previously evaluated the activity of phosphotungstic acid and their pure Cesium salts or supported on matrices such as MCM-41 silica mesoporous and conclude that although H3PW12O40/ MCM-41 has been an efficient solid catalyst, it was less active than soluble acid . They attributed this highest activity the greatest strength of Brønsted acidity of HPA. Nonetheless, it is known that when a Lewis acidity site is introduced in the catalyst, a positive effect can occur. For instance, Lai et al. investigated the conversion of glucose to HMF and verified that the yield achieved in the presence of Ag3PW12O40 was higher than that in the presence of H3PW12O40 catalyst . Those authors attributed this effect to the Lewis acidity of Ag+ cations, which favour the isomerization of carbon skeletal of glucose to fructose, and consequently the HMF formation. Herein, as we are starting from the fructose, no isomerization is required.
Although the mechanism describes as the reaction proceed in the presence of H3O+ cations, remains obscure still how the Copper cations can directly participate in this reaction. As the basis of these results, we can conclude that in addition to participating in the generation of H3O+ cations, the metal cation itself should also play an essential role in the catalytic process, possibly, promoting the dehydration steps. Another possibility is the Copper heteropoly salt minimizes the formation of humins from the polymerization of HMF, increasing its yielding.
Therefore, we will proceed with our investigation for a better understanding of this catalytic system. Although the CrPW12O40 has been also very active, due to the highest HMF yield achieved in Cu3/2PW12O40-catalyzed reaction, it was selected to assess the effect of other reaction variables.
3.2.3. Effect of heteropolyanion in the activity of Copper exchanged heteropolyacid salts in fructose dehydration to synthesize HMF
The fructose dehydration in a biphasic system was carried out in the presence of different HPAs, as well as in the presence of their Copper exchanged salts (Figure 6). All the reactions were assisted by microwaves radiation.
Regardless of the Keggin anion, the Copper exchanged salts were always more efficient than their precursor acid. It means the protons exchange by Copper had a positive effect on the activity of catalysts. Measurements of pH performed in the solution after the catalyst addition showed that as similar to the observed in the reactions with the other metal exchanged phosphotungstate salts the pH was always close to 3.2, excepted when Cu2SiW12O40 was the catalyst. Once more, it is demonstrated that the presence of Copper had a key aspect, possibly more important than Brønsted acidity, since the solutions with the same pH achieved different yields (i.e., Cu3/2PMo12O40, ca. 39% against Cu3/2PW12O40, ca. 53%).
Another important conclusion is that the anion present in the Copper salt also plays a relevant role; even with an equal Copper content, the reactions in the presence of phosphotungstate or phosphomolybdate salts reached different HMF yields (ca. 39 and 53%, respectively, Figure 6). Probably, these heteropolyanions should act also to stabilize the protonated species generated during the process.
3.2.4. Effect of type of anion in the activity of Copper salts in fructose dehydration reactions to synthesize HMF
To investigate in more detail the impact of anion present in the Copper salt, reactions with different salts were performed and the main results are shown in Figure 7. The acids corresponding were also included for a better comparison.
In all the reactions, the Copper salts were more effective than respective Brønsted acids. Possibly, Brønsted acid catalysts can polymerize the HMF, reducing consequently its selectivity as well as its yield. In addition, we have found that Copper chloride or nitrate-catalyzed reactions achieved almost the same yield of HMF as Copper phosphotungstate. Although cheaper than phosphotungstate, these salts are more corrosive and can generate by-products due to the addition of chloride or nitrate ions to the HMF.
3.3. Evaluation of main reaction variables
To assess the influence of catalyst load on the HMF yield, all the other reaction parameters were kept constant (Figure 8).
An increase in the amount of catalyst had a positive effect on the yield of HMF until 10 mol %. It can be assigned to the increase of active sites in solution due to higher catalyst concentration. For catalyst load higher than 10 mol %, no beneficial effect was verified. It is possible that at these conditions the reaction equilibrium within the time interval evaluated has been achieved. Therefore, using 10 mol % of catalyst, the next step was investigating how the yield of HMF vary through time (Figure 9).
The HMF yield was gradually increasing when the reaction was carried through greater periods. However, the maximum yield was achieved in a time of 17.5 min of reaction. Reactions performed in longer periods leads to a lowering of the yield of HMF, probably due to its degradation to humins.
Another important parameter is the temperature of the reaction. It was not possible to evaluate temperatures above 413 K due to limitations of the MW reactor because when doing experiments at 423 K the ethyl acetate (boiling point is 350 K) was lost, due to high pressure in the system.
Figure 10 displays the results of the reaction performed at three heating levels (ca. 393, 403 and 413 K), in the biphasic system containing the soluble catalyst.
The yields considerably drop to 46% at 403 K and 10% at 393 K. Therefore, as the greatest yield was achieved at the highest temperature, it is evidence of the endothermic character of the reaction.
The composition of the biphasic system is an aspect that may drastically affect the reaction yield, mainly because although the reaction occurs in the aqueous phase, the organic phase acts extracting the product formed. To decrease the solubility of HMF in the aqueous solution and to favour its migration to the organic phase, the reaction is carried out in a saturated solution. Different salts can distinctly affect the ionic strength of an aqueous solution and consequently impact the reaction yield. Figure 11 show the HMF yields obtained in reactions with the aqueous phase saturated using different ionic compounds.
When no salt was used, only a poor yield of HMF was achieved (ca. 24%, Figure 11). This result can be explained by the salting-out effect [59, 60]. When a solution is saturated with ions, the solubility of other solutes is compromised. These ions interact strongly with water molecules, through ion-dipole interactions, lowering the solubility of HMF in the aqueous medium. Consequently, this makes it migrate more efficiently to the extracting organic phase. This explains why all evaluated systems containing salts have a better yield for HMF, since that it is extracted more efficiently in their presence, and to avoid also parallel reactions that can occur in the aqueous phase [15, 21].
In general, the best results are observed for systems with the salts NaCl (ca. 81%) and KCl (ca. 67%), which indicates a better HMF extraction efficiency for systems containing the cations Na+ and K+. Conversely, in the solution saturated with LiCl, only a 49% yield was reached. It is expected that cations with smaller radium will be more efficient to saturate this solution, and consequent they can favour the HMF removal toward the organic phase.
The reactions carried out in solutions saturated with CaCl2 or MgCl2 achieved yields of ca. 41% and 37%, respectively; the yields were lower and the formation of humins was observed. This lower yield can be assigned to greater ionic radium of cations Ca2+ and Mg2+, which require a higher water molecules number to be completely hydrated. As smaller the cation hydration radii, as greater the salting-out effect, which leads to greater HMF extraction efficiency and, consequently, a lower occurrence of parallel reactions in the aqueous phase [21, 59]. Interesting that this effect was not observed when compared to the HMF yield of NaCl and LiCl.
The polarity of the organic phase as well as the intermolecular forces existent can favour the removal of HMF from the aqueous phase. This effect was investigated using the conditions optimized and different organic solvents. The main results are shown in Figure 12.
The selection of the organic solvents used to compare with ethyl acetate was performed as the basis on the literature as follows: MIBK [19, 61], acetonitrile  and tetrahydrofuran [20, 63] (Figure 12).
The reactions carried out with ethyl acetate as organic phase continued to be that with the highest HMF yield (ca. 81%), followed by the acetonitrile (ca. 65%) and MIBK ones (ca. 60%). Among the organic solvents tested, tetrahydrofuran was the worst extractor, reaching the poorest yield of ca. 32%, which was also lower than the yield achieved in a system without the extraction phase (ca. 46% yield). The superior performance of ethyl acetate can be assigned to its lowest dielectric constant, and its low dipole moment (Table 4SM). Although no hydrogen bond can exist between ethyl acetate molecules and HMF molecules, their physical properties look to have a synergic effect that favours the extraction of these product molecules from an aqueous phase.
Finally, the possibility of reducing the volume of ethyl acetate used in the extraction was also evaluated. The extraction was done using different volumes (ca. 4.0, 3.0, 2.0 and 1.0 mL: Figure 13). It was found that a reduction of volume used to extract the HMF triggered a drop in yields. This can be explained by the limitation caused by decreasing the volume of the extracting phase, which restricts the amount of HMF extracted and this leads to parallel reactions in the reaction phase, such as HMF self-condensation reactions that produce humins .