2.1. Compositional Analysis of Crude Cardoon Samples
The chemical compositions of the two investigated biomasses, the un-treated defatted cardoon residue (C) and the same biomass recovered after a steam-explosion pretreatment (E), are reported in
Figure 1.
The steam-explosion pretreatment is a high-pressure and high-temperature physical process that has been widely used for biomasses in the last decades [
29], since it allows increased availability of cellulose of the raw lignocellulosic matrix for further processing. The possible use of steam-exploded pretreated biomass is on one hand the production of bioethanol or more in general energy carriers [
30] and on the other hand the possibility to obtain high value biochemical, such as the nanocrystalline cellulose [
5] and other products useful for several chemical or pharmaceutical applications. The input matrices to the steam-explosion pretreatment section can be lignocellulosic softwoods, hardwoods and spruces but also residues obtained from agroindustrial activities, such as tomatoes, vineyard pruning or residues of infesting species, such as the
Cynara cardunculus that can be considered third-generation biomass of growing interest for the production of sustainable biomaterials [
31]. The process “intensity” is generally described by the severity factor (R0) that considers the process parameters, such as the temperature and the pressure. In the case of lignocellulosic materials, the process yields can be significantly increased using a double-step process of both solid and liquid fraction after the pretreatment [
30]. The steam-explosion facility used for this research is available at the University of Perugia (CRB/CIRIAF) and it is composed of: (i) vapor generator; (ii) charging section for raw biomass; (iii) expansion valves; (iv) high-pressure reactor; (v) postexplosion tank and (vi) exploded liquid recovery section [
30].
Both C and E samples show very different content of cellulose, about 38 and 65 wt %, respectively. In fact, after the steam-explosion pretreatment, the hemicellulose amount decreased from about 17 to 4 wt %, whereas extractives and ash were removed, as expected. On the other hand, lignin increased from about 17 up to 30 wt %, due to the reduction of the content of other components. The above-reported composition was determined on the dry biomasses and the enrichment of cellulose content in the steam-exploded sample proved very important from an industrial point of view because at equal biomass loading, it enables the processing of a higher amount of cellulose, which results in marked increases of target product concentrations in the reaction mixture with subsequent ease for the successive work-up, separation and purification processes. The available defatted raw sample C had a humidity amount of 5.9 wt %, while, as expected, the steam-exploded sample E had a humidity level of 73.6 wt %. Both wet as-received samples and dried ones were tested in the catalytic runs.
2.2. MW-Assisted Hydrolysis of C and E Cardoon Samples
A preliminary study on the effect of the catalyst types, their amount and humidity on LA formation was performed. The hydrolysis of dry cardoon, both C and E samples, was carried out using two acid catalysts, HCl and H
2SO
4. The amount of these mineral acids was calculated to have the same concentration of hydronium ions in the starting mixtures.
Table 1 reports the results for the experiments performed employing a biomass loading of 10 wt % (on dry basis), working at 190 °C for 20 min in the MW reactor.
In the presence of HCl, both dry samples (runs C1 and E1,
Table 1) achieve comparable LA yields, 49.3 and 48.4 mol % respectively. Adopting the same dry samples, when H
2SO
4 was used (runs C2 and E2), lower LA molar yields were obtained, 41.5 and 40.7 mol % for C and E cardoon samples, respectively. As reported in previous studies, HCl enhances LA hydrolysis via a one-pot mechanism, since Cl
− ions catalyze the intermediate 5-hydroxymethyl-2-furaldehyde (HMF) rehydration reaction, while SO
42− ions are responsible for an inhibitor effect [
32]. In all these runs, low amounts of unconverted glucose were detected, while no intermediates, such as furfural and HMF, were observed. On the other hand, appreciable amounts of formic acid (FA), which is another commodity chemical [
8], were coformed by the hydrolysis of both C and E cardoon samples. Indeed, FA is coproduced during biomass hydrolysis [
33] and its use in several fields has further encouraged the interest in LA synthesis. In our experiments, FA formation seems to be slightly facilitated when HCl is employed (compare runs C1 with C2 and runs E1 with E2) and for this reason, only HCl was selected for the subsequent investigation regarding the humidity grade. To investigate this parameter, both the cardoon samples C and E were adopted as-received with the humidity grade of 5.9 and 73.6 wt % respectively (runs C3 and E3), using the same experimental conditions of runs C1 and E1. As expected, dry and wet cardoon samples show similar LA molar yields, which are 49.3 and 50.7 mol % for dry and wet C (runs C1 and C3), respectively, and 48.4 and 49.7 mol % for dry and wet E (runs E1 and E3), respectively. Based on these results, since wet biomass can be directly used, only as-received wet cardoon samples were employed for a more detailed investigation. In the perspective of industrial application, the adoption of a high biomass loading is to be preferred, thus applying the high-gravity approach to achieve the highest products concentration. In other words, this is the same concept already applied regarding the cellulose content (on dry biomass) with the steam-explosion pretreatment. The high-gravity method presents several advantages for an industrial perspective: it increases the concentration of crude products, reduces the costs for their purification and waste-water treatment. On the other hand, in an over-loaded reactor, a more difficult physical agitation of the reaction slurry could lead to a significant decrease in the reaction rate.
Table 2 reports the results of hydrolysis experiments carried out by increasing the biomass loading from 10 up to 20 wt % on dry basis, keeping constant the other reaction parameters (190 °C, 20 min, HCl as catalyst). In the same
Table 2, the effect of the substrate to catalyst ratio was investigated at 20 wt % of biomass loading, for both wet C and E cardoon samples.
For both C and E samples, no significant decrease of the yields was observed increasing the biomass loading. On the other hand, higher LA concentrations, from 15.9 up to 34.6 g/L for C samples, and from 25.5 to 58.9 g/L for E ones, were reached. Considering that lower concentrations of HCl allow minimizing environmental impact and process costs,
Table 2 reports the results of experiments performed with lower HCl amounts, increasing the substrate/catalyst ratios from 0.9 up to 2.0 mol/mol, run C5, C6, C7 and runs E5, E6 and E7, for C and E samples respectively. These runs evidenced that LA molar yield is not affected by the reduction of the catalyst content, at least in the range of the investigated substrate to catalyst ratios. For both C and E samples, promising values of LA molar yields were obtained, as well as no significant formation of side-products. In the case of E biomass hydrolysis, even acetic acid was not observed, due to the low content of acetyl groups in this feedstock. As expected, higher LA concentrations were achieved from the cellulose-rich steam-exploded cardoon (E sample) rather than from that untreated (C sample), justifying the adopted pretreatment in the industrial perspective: the maximum reached value was 59.0 g/L for E and 34.6 g/L for C. On the other hand, it is interesting that the reactivity of the cellulosic fraction of the exploded biomass E is very similar to that of the untreated sample C, thus suggesting that pretreatment does not significantly modify the accessibility of the cellulose fibers. Regarding FA, its formation is almost equimolar with respect to LA for runs C3–E3 of
Table 2, as expected from the overall hydrolysis reaction mechanism, whilst this correspondence does not fully fit for the remaining runs E4–E7 of
Table 2 (lower FA molar concentration than the theoretical one), probably due to a combined effect of the type of feedstock, which has been previously steam-exploded (resulting more reactive than the crude sample to the hydrolysis), and of the high loading used for these runs, both leading to its more appreciable thermal degradation to CO
2 and H
2 in the liquid phase.
At the end of every run, a solid residue was recovered, accounting for about 30 wt % respect to the starting dry biomass. Due to the ponderal relevance of these residues, their characterization and applicative perspectives will be discussed later.
Regarding the presence of by-products, the reaction mixture certainly includes also low amounts of soluble impurities, which have not been considered up to now and which should be otherwise better characterized, to get information about the proper work-up procedures, although the concentration of any single by-product in the hydrolyzates is under the limit of detection of the HPLC analysis by routine refractive index detector (0.1 g/L) and, for this reason, not inserted in
Table 1 and
Table 2. To detect trace amounts of impurities, higher sensitivity mass and/or UV detectors should be used. For this purpose, first of all, the crude hydrolysate obtained from run E7 of
Table 2 was extracted by diethyl ether and the recovered extract was analyzed by the GC-MS technique. The analysis revealed the presence of oxygenated C5 compounds and of aromatics of lignin sources, such as guaiacol (2-methoxy phenol) and syringol (2,6-dimethoxy phenol) (see
Figure S1, Supplementary Section,). The dehydration of C5 and C6 carbohydrates causes the formation of furanic monomers and soluble precursors of solid humins [
34]. On the other hand, the above phenolic derivatives show characteristic absorptions in the ultraviolet region, in particular at 284 nm, due to carbonyl n π* transitions [
35], while stronger carbonyl π π* transitions occur at a lower wavelength, at about 205 nm. Based on these statements, the HPLC-UV analysis of the crude hydrolysate was carried out at 284 and 205 nm, thus better differentiating furanic impurities from other carbonyl species, such as aliphatic carboxylic acids. Regarding the UV-HPLC chromatogram at 284 nm (
Figure S2, Supplementary Section), it shows the presence of many furanic/aromatic species, which elute after 20 min, whereas at 205 nm aliphatic carbonyl species prevail (
Figure S3, Supplementary Section). To obtain more in-depth information, the crude liquor was analyzed by HPLC-MS and total ion current (TIC) chromatogram related to all ions of all detected masses and UV chromatograms at 280, 250 nm and 205 nm are reported in the
Supplementary Section, together with the chromatographic data of the main detected compounds including, where possible, the best matched chemical formula (
Figure S4 and Table S1, respectively). At this level of investigation, given the high sensibility of this technique, the TIC signal due to the crude hydrolysate is very complex, including many impurities, besides LA (compound n° 17 of
Table S1, Supplementary Section), which by far represents the main product of interest, as previously stated,. The cross-comparison with the GC-MS data confirms the presence of methoxyphenol and 2,6-dimethoxyphenol (compounds n° 36 and n° 38 of
Table S1 of Supplementary Section, respectively), as aromatics of lignin source. To obtain further useful information for the following discussion, some typical compounds deriving from biomass hydrolysis have been sought [
36], paying particular attention to some selected carboxylic acids of interest [
37]. On this basis, after having excluded the contribution of the LA as the main component, extracted ion chromatogram (XIC) of some typical organic acids deriving from biomass hydrolysis treatment were acquired (
Figure S5/Table S2 and Figure S6/Table S3 of Supplementary Section), thus ascertaining the presence, in low abundance, of many of these, such as tartaric, malic, succinic, lactic, butyric, itaconic, maleic, pyruvic, glutaric, adipic, 2-hydroxy-2-methylbutyric, gluconic and citric acid. In addition, 5-HMF is still present, as well as its dimeric/trimeric derivatives, as confirmed by the corresponding XIC processing (
Figure S7, Supplementary Section). Instead, furfural, the furanic compound deriving from the acid-catalyzed conversion of the hemicellulose fraction, has not been found, thus indirectly confirming the effectiveness of the steam-explosion treatment, which has allowed the preliminary removal of this biomass component. Lastly, glucose has been identified only in traces (
Figure S5/Table S2 of Supplementary Section ), confirming the effectiveness of the acid-catalyzed conversion of the C6 fraction, as highly desired by our approach. Lastly, in order to demonstrate the cost-effectiveness of the hydrolysis process, the purification of the crude hydrolysate deriving from run E7 of
Table 2 was carried out on laboratory scale adopting 2-methyltetrahydrofuran as extraction solvent, followed by subsequent fractional distillation of the corresponding extract, according to the experimental procedure reported in the Materials and Methods section. By this way, the final isolated LA yield of 20.3 wt % was ascertained, similar to that obtained from the HPLC analysis of the corresponding crude hydrolysate (run E7 of
Table 2). The ascertained LA purity grade was 93%, estimated by both the GC and HPLC techniques, including FA, AA and angelica lactone as residual impurities. The latter compound originates from the acid-catalyzed LA dehydration, followed by ring closure, which typically occurs upon LA heating, during the purification procedure.
Steam-exploded cardoon E was selected for further studies regarding the effect of temperature on the hydrolysis reaction and
Table 3 reports the results achieved at the lower temperature of 180 °C and the upper one of 200 °C, in comparison with the run at the temperature of 190 °C previously used. The obtained results confirm that 190 °C is the best temperature for the production of LA, as highlighted in the literature for the hydrothermal treatment of many lignocellulosic biomasses [
13].
Finally, the effect of the reaction time on LA formation was investigated (runs E7, E10 and E11,
Figure 2). As reported in
Figure 2, experiments were carried out on E sample at 20, 40 and 60 min of MW irradiation, under the same reaction conditions of temperature (190 °C), catalyst (HCl, 1.4 wt %), substrate-to-catalyst molar ratio (2.0 mol/mol) and biomass loading (20 wt %).
Only a weak time effect on the hydrolysis efficiency was observed, achieving the best LA yield of 49.9 mol % with the highest LA concentration of 62.1 g/L after 40 min of heating, adopting HCl in a low amount, 1.4 wt %.
2.4. MW-Assisted Alcoholysis of E Cardoon Sample
In the second part of our investigation, the one-pot alcoholysis reaction of wet E cardoon in
n-butanol to BL was studied in the MW reactor and the results are reported in
Table 5. The presence of a significant amount of introduced water, due to the high humidity of the sample, is an unprecedented approach respect to the up to now reported alcoholysis studies, performed in the presence of alcohol alone as reactant/reaction medium. This procedure allowed us to convert efficiently the starting biomass, without the initial drying step, thus at the same time-saving time and significant resources.
In the presence of low contents of the acid catalyst H
2SO
4 (1.3 and 2.4 wt %) with the same substrate to catalyst molar ratio, and applying only 15 min of heating, the butanolysis led to BL molar yields of 42.5 and 22.1 mol %, using respectively the biomass loading of 8 and 15 wt % on dry basis, runs AE1 and AE2 (
Table 5). When the lower catalyst loading of 1.3 wt % was maintained with the biomass loading of 15 wt %, a significant decrease of BL yield (19.8 mol %) was ascertained, essentially due to the above-mentioned mass transfer issues. Taking into account that the best BL yield, as well as a good BL concentration (27.5 g/L), were obtained in the run AE1, the effect of duration was studied adopting the same conditions but prolonging the reaction time to 30 and 45 min (runs AE4 and AE5, respectively). The achieved results show that the prolonging of the reaction time does not improve the BL production, whereas slightly lower BL yields and BL concentrations were obtained, demonstrating the efficacy of MW irradiation within a short reaction time. Finally, to better evaluate the impact of introduced water on the ascertained performances, an explorative run adopting the dry cardoon E sample was performed employing the same reaction conditions of run AE1: in this case, the BL concentration of 17.9 g/L was achieved with BL ponderal and molar yields of 26.4 wt % and 38.3 mol % respectively, proving similar to that ascertained on wet cardoon. In all the above runs only traces of levulinic acid were detected, thus confirming that under an excess of bioalcohol the hydrolysis reaction is irrelevant and also when a certain humidity is introduced with the wet biomass.
At the end of the reaction for every run, a significant amount of solid residue was also ascertained, whose characterization and potential valorization will be discussed below.
The results of this preliminary study demonstrate that direct BL production can be performed not only from conventional starting materials (LA, disaccharides, polysaccharides and furfuryl alcohol [
24]) but also directly, using raw steam-exploded defatted cardoon as the starting feedstock, opening the way to the direct conversion of the cellulosic fraction of this cheap, residual biomass in a valuable intermediate/bio-fuel.
2.5. Characterization of Postreaction Solid Residues
The solid residues recovered from the best MW-assisted hydrolysis and alcoholysis reactions to LA and BL, runs E7 and C7 for hydrolysis and run AE1 for alcoholysis, amounted respectively 30.6, 31.4 and 27.4 wt % of the starting biomass, calculated on a dry basis. All these samples were analyzed by elemental analysis and
Table 6 reports the obtained results, compared with the C and E dry cardoon samples starting feedstocks.
The above data related to the starting feedstocks C and E confirm only a limited beginning of carbonization for the latter sample, as shown by the slight increase in its carbon content, occurred as a consequence of the mild steam-explosion treatment, aimed at the breakdown of the biomass matrix (cross-linking lignin), the bulk solubilization of the hemicellulose fraction and the removal of smaller hydrocarbon molecules (volatiles and gases) [
38]. Instead, more advanced carbonization has occurred as a consequence of the acid-catalyzed hydrothermal treatment. H/C and O/C molar ratios of both solid residues at the end of runs C7 and E7 fall within the range reported in the literature for the hydrochars (H/C: ~0.8–1.4 and O/C: ~0.3–0.5) [
39] and also in agreement with our previous work [
40]. Therefore, the differences in the carbon content of the two different starting feedstocks (cardoon C and E, both as starting feedstocks in
Table 6) have been attenuated by the acid-catalyzed hydrothermal treatment (solid residue of run C7 and E7, respectively, in
Table 6), demonstrating the progressed carbonization occurred with this technology. Lastly, the postalcoholysis residue (solid residue of run AE1 in
Table 6) does not show significant compositional differences respect to the posthydrolysis ones (solid residues of run C7 and E7 in
Table 6), thus highlighting the similarity between the performed hydrolysis/alcoholysis treatments. These conclusions are also evident from the Van Krevelen diagram shown in
Figure 4 which plots the molar ratios of the H/C and O/C for both the starting biomasses, cardoon samples C and E, and for the solid residues at the end of runs C7 and E7 for hydrolysis and run AE1 for butanolysis.
The positions of the starting feedstocks and the corresponding chars in the Van Krevelen diagram confirm that, in all cases, dehydration is the main allowed path, leading to the formation of carbonaceous material and the H/C and O/C ratios of all the obtained residues fall within the range reported in the literature for this kind of biomaterial (H/C: ~0.8–1.4 and O/C: ~0.3–0.5) [
40,
41]. In addition, the higher heating value (HHV) was calculated from elemental analysis (
Table 6): the ascertained values significantly increase going from the starting feedstock to the corresponding chars, thus resulting comparable with that of the traditional lignite coal, in agreement with the conclusions gathered from the Van Krevelen plot [
41].
Furthermore, FT-IR characterization of the postreaction solid residues was performed and
Figure 5 shows the FT-IR spectra registered in ATR mode both for the starting crude cardoon, both C and E samples, and for the solid residues recovered after hydrolysis reactions, runs E7 and C7.
The spectrum of crude sample C shown in
Figure 5a shows characteristic signals of the three biopolymers, cellulose, hemicellulose and lignin: a broad band at about 3330 cm
−1 assigned to vibration mode of O–H bonds and a band at 2910 cm
−1 assigned to C–H stretching of methyl and methylene groups. The peak at 1734 cm
−1 is due to C=O stretching in the acetyl group and carboxylic acid of hemicellulose, while the signals at 1600 and 1510 cm
−1 are attributed to aromatic C=O stretching and C=C vibration of lignin [
41,
42]. Weak peaks are observed at 1422, 1375 and 1318 cm
−1 that can be assigned to the C–H asymmetric mode of CH
2 in cellulose, aromatic C–H and C–O in lignin, respectively [
43,
44]. The band at 1239 cm
−1 is due to the C–O stretching of alcoholic, phenolic and ether groups [
44]. The peak at 1155 cm
−1 is assigned to C–O–C asymmetric stretching in cellulose and hemicellulose, the intense peak at 1030 cm
−1 to C–O–H stretching and the one at 900 cm
−1 to anomeric vibration at the β-glycosidic linkage, while the band at 590 cm
−1 can be due to aromatic C–H bonds [
40,
45]. As expected, the peaks at 1734 and 1600 cm
−1 are not noticeable in the spectrum of crude sample E showed in
Figure 5b, since the steam-explosion pretreatment degrades hemicellulose and lignin. The FT-IR spectroscopy in ATR mode allows the characterization of functional groups on the material surface and the analysis is in agreement with the bulk composition determined by the NREL procedure for the samples of raw cardoon. In the case of the solid residues recovered after the posthydrolysis reactions 7E and 7C, reported again in
Figure 5a,b respectively, peaks characteristic of cellulose, such as the intense peak at about 1030 cm
−1 and the peak at about 900 cm
−1, are strongly decreased in accordance with the fact that cellulose was efficiently converted to LA. However, signals that may be assigned to byproducts such as humins appear: a peak at about 1700 cm
−1 due to the C=O stretching, another one at about 1600 cm
−1 due to stretching vibration of C=C bonds of furanic rings and the last one at about 795 cm
−1 due to aromatic bending off the plane of the C–H bond [
38,
44] are evident in the spectrum of both post-hydrolysis residues. Broad absorbance in the 1300–1100 cm
−1 region could be ascribed to multiple C–OH stretching bonds. However, the intense signal at 1208 cm
−1 evident in the posthydrolysis residues 7C and 7E could arise also from the presence of ether bonds [
46]. Humins can result from condensation reactions between sugars, HMF and intermediates during the dehydration of carbohydrates [
47,
48,
49,
50] and their formation can be competitive with the rehydration of HMF to LA. Moreover, humins are complex and recalcitrant carbonaceous materials that can cover the substrate surface, making it less accessible to acid attack.
Figure 6 reports the FT-IR spectra of the solid residue recovered after the butanolysis reaction AE1 together with the starting crude E sample.
In
Figure 6, the FT-IR spectrum of the postalcoholysis char is shown, resulting very similar to that obtained after hydrothermal processing (
Figure 5). In particular, the absorption bands at 1700 and 1600 cm
−1 (C=O and C=C stretching vibrations, respectively) are visible also in the postalcoholysis char, as well as that at 1030 cm
−1 (C–O–H stretching in cellulose), even if in this case of lower intensity respect to the crude E feedstock, thus confirming the occurred cellulose decomposition, as previously stated. In addition, the absorption bands at 1208 cm
−1 (C–O stretching), and 795 cm
−1 (C–H bending off the plane) further confirm the similarity between the chars produced by hydrolysis/alcoholysis. At this stage of the investigation, the severity of the performed treatments dampens any differences deriving from the use of different reaction solvents, otherwise reported by some authors very recently [
51].
Moreover, the recovered residues and the starting biomasses were also characterized by thermogravimetric analysis and weight loss and weight loss thermograms are depicted in the
Supplementary Section (Figure S8).
Regarding the devolatilization behavior of the starting feedstocks (
Figure S8, thermograms (a) and (b)), the first peak is found below 100 °C and is due to the loss of humidity. The degradation of the cardoon starts at higher temperatures. In particular, a shoulder is found at about 250 °C, present only in the C cardoon starting feedstock, ascribed to hemicellulose fraction, overlapped with that of cellulose, in the range 300–350 °C [
52,
53]. Lastly, lignin degradation is very slow, occurring for the whole temperature range of the thermogravimetric analysis [
52]. The comparison between the thermograms of the starting feedstocks, C and E cardoon samples ((a) and (b)), confirms the effectiveness of the steam-explosion treatment, which has allowed the selective removal of the hemicellulose fraction. Moreover, the E cardoon starting feedstock shows a very weak shoulder at about 400 °C, which is due to the lignin component [
52], which proved more accessible and reactive as a consequence of the occurred steam-explosion pretreatment. On the other hand, the three residues deriving from hydrolysis/alcoholysis treatments ((c), (d), (e)) show analogous thermal profiles, thus confirming their chemical similarity, also in agreement with the previous characterization data. Again, the humidity loss of these samples occurs below 100 °C, then a degradation step was found at about 200 °C, due to the release of some organic compounds (such as LA and FA), trapped into this porous bio-material [
54]. The absence of the degradation steps of hemicellulose and cellulose fractions confirms the effectiveness of the performed hydrolysis/alcoholysis treatments, whereas the main peak at about 400 °C is attributed to the decomposition of volatile lignin/furanic structures [
54,
55]. In all cases of the synthesized residues, a final weight loss of ~50 wt % was ascertained, thus revealing increased thermal stability above the starting biomass and demonstrating similar carbonization occurred [
40], also in agreement with the previous characterization data.
The characterization of the obtained solid residues opens the way to their uses. Regarding this aspect, certainly the most immediate use is the combustion for the energy recovery, but this choice is currently considered as the last option, preferring, when possible, its reuse within the scopes of the circular economy [
56]. For agricultural uses, the application of biochar to the soil can mitigate climate change by promoting carbon sequestration and decrease greenhouse gas emissions [
57]. Moreover, char has been advantageously proposed as a growing medium, to be used in combination with other components (vermicultite, clays, etc.) to improve physicochemical soil properties, such as increase in cation exchange capacity, water holding capacity, available water, improvement of soil structure, reduction in soil acidity, microbiological activity, quality, and yield of the crops [
57]. Use of the char as a soil amendment for the recovery of contaminated soils, including stabilization of organic and inorganic contaminants, has been proposed and seems attractive [
57]. Lately, more added-value char-based products are under development in many research fields, such as adsorption, catalysis and electrochemical energy storage (lithium-ion batteries, lithium-sulfur batteries, sodium-ion batteries and supercapacitors), after having properly tuned its porosity and functional groups, by choosing the appropriate starting feedstocks and optimizing the reaction conditions [
58].