From Waste to Resource: Utilizing Sweet Chestnut Waste to Produce Hydrothermal Carbon for Water Decontamination

: Carbonaceous materials are a highly appealing class of adsorbents, owing to their excep-tional properties, such as high surface area and thermal and chemical stability. These materials have found successful applications in water puriﬁcation. Sweet chestnut ( Castanea sativa ) cupules are disposed of as waste. Valorization of these residues is a step forward in terms of circular economy and sustainability. Meanwhile, per- and poly-ﬂuoroalkyl substances (PFASs) pose signiﬁcant concerns due to their persistence, bioaccumulation, and toxicity, emerging as contaminants of concern for human health and the environment. This study focuses on preparing carbonaceous material by hydrothermal carbonization from chestnut cupules, followed by their use as adsorbents for PFAS removal from polluted water. The cupule waste material was crushed, ground, sieved, and subjected to hydrothermal treatment at temperatures ranging from 180–200 ◦ C to produce hydrothermal carbons. The adsorbents obtained were characterized by various techniques such as nitrogen adsorption isotherm, porosimetry, point of zero charge, Fourier-transform infrared, scanning electron microscopy, and thermal, elemental, and energy dispersive X-ray analyses. Surface area (S BET ) values of 42.3–53.2 m 2 · g − 1 were obtained; pH PZC ranged from 3.8 to 4.8. This study also determined the adsorption kinetics and isotherms for removing perﬂuorooctanoate-contaminated water. The equilibrium was established at 72 h and q e = 1029.47 mg · g − 1 . To summarize, this research successfully valorized a biomass residue by transforming it into hydrothermal carbon, which was then utilized as an adsorbent for water decontamination.


Introduction
Carbonaceous materials constitute an attractive group of adsorbents due to their properties, such as large surface area and thermal and chemical stability. In addition, activated carbons with different pore sizes can be obtained, thus adapting them to the functions they are designed for. On the other hand, functionalizing the surface of activated carbons is possible, giving them specific chemical properties adapted to a particular purpose. It is also of great importance that carbonaceous materials are considered safe, accessible, and affordable [1]. Various types of biomass residues, due to their low cost and abundance, have been evaluated to obtain functional carbonaceous materials. Many methods provide materials whose physicochemical properties can be tuned at will, thus obtaining materials that have shown promising results in different fields, such as adsorption or energy storage [2]. For this reason, they are considered suitable candidates for obtaining sustainable materials. These materials have been successfully applied in catalysis, water C 2023, 9,57 3 of 16 has the advantage of being a more environmentally friendly manufacturing process, as it is carried out in hermetically sealed containers and at a lower temperature. This allows for easier management of possible waste and lower energy consumption. Additionally, despite its limited specific surface area, its high proportion of functional groups can give rise to adequate adsorption capacity. We have been working on the obtention and use of hydrothermal carbons from chestnut cupules for the last few years. Our preliminary results (unpublished) show that these carbons are suitable adsorbents in the decontamination of heavy metals, such as thorium, from water. Given these results, we thought it would be apt to check for the adsorption capacity of apolar organic molecules such as PFAS. Adsorption onto carbonaceous materials has been established as a promising technology for PFAS removal from water [20,[35][36][37][38][39][40][41][42].
The objective of this work is to investigate the use of chestnut cupule waste as sustainable biomass raw material to obtain hydrothermal carbons-hence, adding value to it-and its use as adsorbent material for the removal from water of sodium perfluorooctanoate (PFONa), the anion of perfluorooctanoic acid, one of the most significant PFASs.

Materials
Sodium perfluorooctanoate 97%, CAS 335-95-5 ( Figure 1), was purchased from Thermo Fisher. Milli-Q water was used as a solvent in all determinations. Sweet chestnut (Castanea sativa) cupules were collected from chestnut groves in Valle del Jerte, in Cáceres, Spain. C 2023, 9, x FOR PEER REVIEW 3 of 18 This has the advantage of being a more environmentally friendly manufacturing process, as it is carried out in hermetically sealed containers and at a lower temperature. This allows for easier management of possible waste and lower energy consumption. Additionally, despite its limited specific surface area, its high proportion of functional groups can give rise to adequate adsorption capacity. We have been working on the obtention and use of hydrothermal carbons from chestnut cupules for the last few years. Our preliminary results (unpublished) show that these carbons are suitable adsorbents in the decontamination of heavy metals, such as thorium, from water. Given these results, we thought it would be apt to check for the adsorption capacity of apolar organic molecules such as PFAS. Adsorption onto carbonaceous materials has been established as a promising technology for PFAS removal from water [20,[35][36][37][38][39][40][41][42].
The objective of this work is to investigate the use of chestnut cupule waste as sustainable biomass raw material to obtain hydrothermal carbons-hence, adding value to it-and its use as adsorbent material for the removal from water of sodium perfluorooctanoate (PFONa), the anion of perfluorooctanoic acid, one of the most significant PFASs.

Synthesis of the Hydrothermal Carbons
The chestnut cupules were allowed to dry naturally. They were subsequently ground and sieved, collecting three fractions: large size (1-2 mm), medium size (0.5-1 mm), and small size (less than 0.5 mm). Then, they were subjected to hydrothermal treatment in a duralumin-coated Teflon™ reactor.
To obtain HC-A, in a hydrothermal reactor, 100 mL of distilled water and 15 g of small-size sieved cupule were added. The reactor was closed, placed in an oven previously programmed at 180 °C, and kept for 24 h. The reactor was then removed from the oven and, once cold, opened. The final product was filtered and washed with plenty of distilled water (ca. 200 mL) on a filter. Subsequently, it was dried in an oven at 110 °C for 8 h. The other carbons were prepared similarly, according to the data in Table 1.

Synthesis of the Hydrothermal Carbons
The chestnut cupules were allowed to dry naturally. They were subsequently ground and sieved, collecting three fractions: large size (1-2 mm), medium size (0.5-1 mm), and small size (less than 0.5 mm). Then, they were subjected to hydrothermal treatment in a duralumin-coated Teflon™ reactor.
To obtain HC-A, in a hydrothermal reactor, 100 mL of distilled water and 15 g of small-size sieved cupule were added. The reactor was closed, placed in an oven previously programmed at 180 • C, and kept for 24 h. The reactor was then removed from the oven and, once cold, opened. The final product was filtered and washed with plenty of distilled water (ca. 200 mL) on a filter. Subsequently, it was dried in an oven at 110 • C for 8 h. The other carbons were prepared similarly, according to the data in Table 1. Textural characterization was performed by N 2 adsorption isotherms at 77 K in a Quantachrome Quadrasorb Evo (Quantachrome Ltd., Hook, UK), and the specific surface area was calculated by applying the BET method [43]. Porosimetry was performed in a Quantachrome Poremaster 60 (Quantachrome, Hook, UK). Thermal analysis was carried out in an STA 449 F3 Jupiter thermobalance (Netzsch, Selb, Bavaria, Germany). Fourier-transform infrared spectra were recorded in a Bruker Vertex70 spectrometer in the 400-4000 cm −1 range, with a DLaTGS detector and a 0.4 cm −1 spectral resolution. A KBr tablet was prepared with 1/250 dilution (Bruker, Billerica, Massachusetts, USA). Elemental analysis (C, H, N, S, O) was performed using a Leco CHMS-932 elemental analyzer (Leco, St. Joseph, MI, USA). C, H, N, and S were analyzed, and the difference was assigned to ash (measured from proximate analysis) and oxygen content. The point of zero charge (PZC) values were determined using the method proposed by Valente Nabais and Carrott [44]. To carry out this measurement, a 0.1 M solution of sodium nitrate was prepared, and 7% in weight of the carbon was added, followed by stirring for 48 h at 25 • C in a thermostatic bath. Subsequently, it was filtered, and the pH of the filtered solution was measured with a pH meter. This pH value corresponded to the point of zero charge. A field emission scanning electron microscope (SEM) Quanta 3D FEG (FEI Company, Hillsboro, OR, USA) was used to explore the surface morphological characteristics of all samples. The sample analysis was performed under high vacuum with a secondary electron. The surface of the samples was coated with gold due to their low conductivity. Energy dispersive X-ray analyses (EDX) were recorded with the same equipment.

Adsorption Experiments
The conjugate base sodium salt (PFONa), instead of the perfluorooctanoic acid, was used to carry out the adsorption experiments, since the acid has a K a ca. -1 and dissociates completely in water.

Adsorption kinetics
To determine the PFONa adsorption over time, the following procedure was used: A 200 mg/L PFONa water solution was prepared by dissolving 10 mg of PFONa in a 50 mL volumetric flask. It was then transferred to an Erlenmeyer flask, and 50 mg of hydrothermal carbon (HC) was added, so the HC to PFONa ratio was 5:1. All flasks were thermostated at 25 • C and magnetically stirred at 400 rpm. Assays were performed in duplicate; 2 mL samples were taken at 2, 4, 24, 48, 72, and 96 h, then filtered with a 0.22 µm pore filter, and, finally, aliquots were analyzed for PFONa concentration.

2.
Adsorption isotherm Different PFONa solutions of 40, 70, 100, 130, 165, and 200 mg/L were prepared in 10 mL volumetric flasks. They were then transferred to separate Erlenmeyer flasks, and 10 mg of hydrothermal carbon was added to each one, so the HC to PFONa mass quotient was 25.0, 14.3, 10.0, 7.7, 6.1, and 5.0, respectively. All flasks were thermostated at 25 • C and magnetically stirred at 400 rpm. Assays were performed in duplicate; 2 mL samples were taken at 96 h, then filtered with a 0.22 µm pore filter, and, finally, aliquots were analyzed for PFONa concentration.

Analytical Determinations
The samples were analyzed using liquid chromatography (Agilent 1290 Infinity II LC (Agilent Technologies, Santa Clara, California, USA)) coupled to a triple quadrupole mass spectrometer (Agilent 6460 triple quad LC/MS (Agilent Technologies)). Aliquots of 3 µL were injected, and elution was carried out by operating in gradient mode at 0.4 mL/min with a reverse phase HPLC Zorbax Extend C18 column (3 µm, 100 mm × 1.8 mm) at 30 • C. As the mobile phase, 0.1% formic acid in ultrapure water (A) and 0.1% formic acid in acetonitrile (B) were used, as shown in Table 2.
Ionization was carried out in negative mode. The equipment parameters were gas temperature 225 • C, gas flow 10 L/min, nebulizer 45 psi, sheath gas temperature 350 • C, sheath gas flow 11 L/min, capillary voltage 3.6 kV, nozzle voltage 1500 V, and delta EMV 400. The MassHunter B.07.00 (Agilent) software was used for data acquisition and equipment control. Multiple Reaction Monitoring (MRM) was used, with the transitions sought to be 412.9 → 368.9 and 412.9 → 169, the former used in the quantification of the compound. The fragmentation and collision energy voltages were 86 V and 5 V, respectively.

Characterization of the Hydrothermal Carbons
The nitrogen adsorption isotherms ( Figure 2) corresponded to type III, according to the IUPAC classification [45,46]; therefore, the interaction between the N 2 molecules and the surface must have been weak. The surface obtained by the BET method produced values that were higher than those obtained by the DFT method but in the same order of magnitude (Table 3) [43,47]. Nevertheless, since the isotherms were of Type III, the BET model was not to be considered highly reliable. All three carbons showed a minimal surface area, low microporosity, and a considerably greater volume of large pores, i.e., mesopores and macropores (Figures 3-5). The HC-A coal had the least-developed porosity. The DFT method was used to study the distribution of micropores and mesopores. The results show that the three carbons presented a similar pore distribution. A high volume of narrow mesopores was detected, whereas no micropores were present. The HC-B carbon had a larger volume of narrow mesopores, while the HC-C showed more volume when the pores widened. The HC-A carbon showed a more significant development of wide porosity than the other two adsorbents. It should be noted that HC-B and HC-C carbons showed a more developed narrow mesoporosity than HC-A. However, the mesopore and macropore structure was very similar in the three materials.     1 Brunauer, Emmett, and Teller model [43]. 2 Dubinin and Radushkevich model [48,49]. 3 Density Functional Theory [47].  1 Brunauer, Emmett, and Teller model [43]. 2 Dubinin and Radushkevich model [48,49]. Functional Theory [47].
The DFT method was used to study the distribution of porosity in micro an pores, and the results are shown in Figure 3. The cumulative volume plot obtain porosimetry is shown in Figure 4, and the width pore distribution plot, also obtain porosimetry, is shown in Figure 5.        The measurements of the PZC show that all three materials were acidic (Table 4 is usual in hydrothermal carbons. It must be considered that basic oxygenated functi groups (carbonyl) are more labile than those that provide acidic properties (alcohol carboxylic), so it is to be expected that the former will degrade in a more significant portion than the latter. In addition, as the degree of graphitization was low, as obse from the elemental composition with low carbon content, there was not a substantial tribution of basicity by the aromatic structures. The Fourier-transform infrared (FTIR) spectra obtained are shown in Figure 6. merous observable bands in the FTIR spectra belonging to many functional groups cate that the degree of graphitization of these materials was not high. The most int band at 3400 cm −1 corresponded to either alcohol or carboxylic O-H bonds. The ban 2935 cm −1 was assigned to C-H bonds, but its low wavenumber suggested they The DFT method was used to study the distribution of porosity in micro and mesopores, and the results are shown in Figure 3. The cumulative volume plot obtained from porosimetry is shown in Figure 4, and the width pore distribution plot, also obtained from porosimetry, is shown in Figure 5.
The measurements of the PZC show that all three materials were acidic (Table 4), as is usual in hydrothermal carbons. It must be considered that basic oxygenated functional groups (carbonyl) are more labile than those that provide acidic properties (alcohol and carboxylic), so it is to be expected that the former will degrade in a more significant proportion than the latter. In addition, as the degree of graphitization was low, as observed from the elemental composition with low carbon content, there was not a substantial contribution of basicity by the aromatic structures. The Fourier-transform infrared (FTIR) spectra obtained are shown in Figure 6. Numerous observable bands in the FTIR spectra belonging to many functional groups indicate that the degree of graphitization of these materials was not high. The most intense band at 3400 cm −1 corresponded to either alcohol or carboxylic O-H bonds. The band at 2935 cm −1 was assigned to C-H bonds, but its low wavenumber suggested they were mainly aliphatic bonds, since the aromatic ones give a signal at higher values. The 1710 cm −1 band was due to C-O double bonds that can be carbonyl or carboxylic, whereas the neighboring band of 1610 cm −1 was characteristic of C-C conjugated double bonds. Several bands between 1550 cm −1 and 1300 cm −1 can be attributed to the carbon backbone of these materials, for instance, those at 1515 cm −1 and 1460 cm −1 ; however, the presence of O-H and C-O bands in this region made assigning them somewhat problematic. Between 1300 cm −1 and 100 cm −1, the bands were mainly due to oxygenated or nitrogenous functional groups (C-O or C-N bonds) [50]. cm −1 band was due to C-O double bonds that can be carbonyl or carboxylic, whereas the neighboring band of 1610 cm −1 was characteristic of C-C conjugated double bonds. Several bands between 1550 cm −1 and 1300 cm −1 can be attributed to the carbon backbone of these materials, for instance, those at 1515 cm −1 and 1460 cm −1 ; however, the presence of O-H and C-O bands in this region made assigning them somewhat problematic. Between 1300 cm −1 and 100 cm −1, the bands were mainly due to oxygenated or nitrogenous functional groups (C-O or C-N bonds) [50]. The scanning electron microscopy images of the samples in which hydrothermal carbonization had been carried out (Figures 7-9) showed a more defined structure than in the original material ( Figure 10), in which the cells of the lignocellulosic material were more clearly observed. The treatment, therefore, removes matter from the original material, increasing the porosity. In some images, small spheres on the surface could be seen; it is known that obtaining hydrothermal coals from carbohydrates gives rise to spherical particles, so they could have come from dissolved substances of the chestnut cupule that underwent hydrothermal carbonization. Another option is that they were wax deposited on the surface of the biochar; given the similarity of composition in both cases, it is difficult to determine which of the processes gave rise to these structures. The scanning electron microscopy images of the samples in which hydrothermal carbonization had been carried out (Figures 7-9) showed a more defined structure than in the original material (Figure 10), in which the cells of the lignocellulosic material were more clearly observed. The treatment, therefore, removes matter from the original material, increasing the porosity. In some images, small spheres on the surface could be seen; it is known that obtaining hydrothermal coals from carbohydrates gives rise to spherical particles, so they could have come from dissolved substances of the chestnut cupule that underwent hydrothermal carbonization. Another option is that they were wax deposited on the surface of the biochar; given the similarity of composition in both cases, it is difficult to determine which of the processes gave rise to these structures.      The thermal analysis program can be found in the Supplementary Material (T S1). Thermograms (Figure 11) show that the moisture of these materials was low, as as the amount of residual ash, which was less than 1% in the three carbons. Regarding volatile matter (from 50 to 80 min approximately), it followed the order HC-A > HC HC-C, and the fixed carbon followed the opposite order. Therefore, they can be con ered simply as carbonaceous materials with a low content of mineral matter. The thermal analysis program can be found in the Supplementary Material (Table S1). Thermograms ( Figure 11) show that the moisture of these materials was low, as well as the amount of residual ash, which was less than 1% in the three carbons. Regarding the volatile matter (from 50 to 80 min approximately), it followed the order HC-A > HC-B > HC-C, and the fixed carbon followed the opposite order. Therefore, they can be considered simply as carbonaceous materials with a low content of mineral matter. The results of the organic elemental analysis (C, H, N, S, and O) and the ash content (Table 5) show that the samples prepared at a higher temperature (HC-B and HC-C) presented more carbon, nitrogen, and ash, and less hydrogen and oxygen. This indicates that the carbonization process was more intense when the temperature was raised. The size did not seem to be crucial, though, since the smaller the particle size, the more intense the effect of pressurized water should have been, and this was not observed in the results. The data in the table agree with the fixed carbon content, which tended to increase with elemental carbon content, and the volatile matter, which tended to increase with oxygen and hydrogen content. The results of the organic elemental analysis (C, H, N, S, and O) and the ash content (Table 5) show that the samples prepared at a higher temperature (HC-B and HC-C) presented more carbon, nitrogen, and ash, and less hydrogen and oxygen. This indicates that the carbonization process was more intense when the temperature was raised. The size did not seem to be crucial, though, since the smaller the particle size, the more intense the effect of pressurized water should have been, and this was not observed in the results. The data in the table agree with the fixed carbon content, which tended to increase with elemental carbon content, and the volatile matter, which tended to increase with oxygen and hydrogen content. On the other hand, EDX analyses were performed on all three hydrothermal carbons and the raw chestnut cupule (Table 6). This determination, which studied a more superficial area than the elemental analysis, confirmed the same trend. All carbons had been enriched in carbon and depleted in oxygen. It is worth noticing that the HC-C sample contained the most significant mineral matter. This effect is larger than the one found in the elemental analysis, which indicates that the hydrothermal treatment affected the surface more than the interior of the material particles.

Adsorption Experiments
The kinetic study of adsorption processes makes it possible to relate the amount of adsorbate adsorbed per gram of adsorbent with the elapsed contact time between adsorbent and adsorbate. The variation in the excess PFONa concentration over time data is shown in Table 7. At 96 h, equilibrium can be considered established, having reached the maximum adsorption. Our preliminary studies show that HC-A and HC-B have adsorption of around 15%, while HC-C adsorbs 26%; therefore, further decontamination studies presented herein focused on the most promising adsorbent, HC-C. applied it, we found it to be inappropriate for our data. On the other hand, the pseudosecond-order adsorption kinetic model assumes that the rate-limiting step is the chemical adsorption process, and it can be expressed mathematically, as shown in Equation (1): where t is the time, q t is the amount of solute adsorbed at time t, q e is the amount of solute adsorbed at equilibrium, and k is the rate constant. The data are shown in Table 8, whereas the graph of t/q t vs. t is plotted in Figure 12.  The data points follow a straight line, suggesting that the pseudo-second-order adsorption kinetic model applies to this system. The slope was 9.7137 × 10 −4 , and the intercept was 1.1612 × 10 −3 , so we can calculate the rate constant (k = 8.126·10 −4 h −1 ·mg −1 ·g) and the amount of solute adsorbed at equilibrium (qe = 1029.47 mg·g −1 ).
Experiments were conducted with different HC-C to PFONa ratios to determine the carbon adsorption isotherm. The values obtained are shown in Table 9    The data points follow a straight line, suggesting that the pseudo-second-order adsorption kinetic model applies to this system. The slope was 9.7137 × 10 −4 , and the intercept was 1.1612 × 10 −3 , so we can calculate the rate constant (k = 8.126·10 −4 h −1 ·mg −1 ·g) and the amount of solute adsorbed at equilibrium (q e = 1029.47 mg·g −1 ).
Experiments were conducted with different HC-C to PFONa ratios to determine the carbon adsorption isotherm. The values obtained are shown in Table 9 and plotted in Figure 13.  We found that the data obtained did not mathematically fit the Langmuir adsorption model, i.e., the adjustment of the model to the experimental data was inconsistent. On the other hand, the Freundlich model fit these results. Data for the Freundlich model are shown in Table 10. We found that the data obtained did not mathematically fit the Langmuir adsorption model, i.e., the adjustment of the model to the experimental data was inconsistent. On the other hand, the Freundlich model fit these results. Data for the Freundlich model are shown in Table 10.
The consistency with the Freundlich model indicates some degree of heterogeneity on the surface (not all active sites are identical) and the possibility that more than one layer can be adsorbed. Since the material used as adsorbent was amorphous, this relative heterogeneity can be expected. Moreover, the value of n indicates that the interaction between adsorbent and sorbate is mainly of the chemisorption type. Our results are promising, with a q e = 1029.47 mg·g −1 ; interestingly, they suggest that the maximum adsorption capacity had not been achieved at high concentrations. To put these results in perspective, we can refer to the excellent review by Liang et al.; in it, the authors compared the different adsorption methods for removing PFAS from water [37]. Analyzing the results obtained in water at pH 7, we found that, for granular activated carbon (GAC), a value of q e = 2.34-102.7 mg·g −1 is reported; and for powdered activated carbon (PAC), q e values of 16.9-500 mg·g −1 are found. On the other hand, for biochar derived from hardwood sawdust, the value obtained for q e at pH 7.2 is 2.4 mg·g −1 , and for pinewood at the same pH, q e is 2.1 mg·g −1 . Liang et al. found that ion exchange resins (q e = 525-1500 mg·g −1 ) and carbon nanotubes (q e = 140-1110 mg·g −1 ) show the highest adsorption capacities, yet they interestingly concluded that low-cost adsorbents with high adsorptive capacity are highly recommended in place of commercial AC due to their local availability, technical feasibility, and cost-effectiveness. The mechanisms affecting PFAS adsorption have been reviewed by Liang et al. in an excellent review [37]. Diffusion, electrostatic interaction, hydrophobic interaction, ion exchange, and hydrogen bonds all can play a role in the adsorption process and are affected by the physicochemical characteristics of the adsorbent, such as pore size, pore distribution, and functional groups. In our case, HC-C is the carbon with the largest volume of mesopores; since PFONa is a large PFAS, it is essential for diffusion and adsorption that there are large numbers of mesopores. It is also the most graphitized hydrothermal carbon, i.e., the least polar; this also plays a vital role in the PFONa adsorption mechanism onto HC-C. Because of the hydrophobic nature of PFAS chains, the sorption of anionic PFAS can be dominated by hydrophobic interactions between the negatively-charged hydrophobic fluorinated tail and the surface of hydrophobic adsorbents, regardless of the electrostatic repulsion [51,52]. Particularly at high pH, hydrophobic interaction, which is not sensitive to changes in pH, should be the primary PFAS adsorption mechanism [53,54].
Given these results, and due to the advantages of the hydrothermal carbon herein reported, further work is underway to gain insight into the full potential of this adsorbent in water decontamination.

Conclusions
The preparation conditions of the hydrothermal carbons greatly influence the result; this allows for obtention of coals with tuned properties by optimizing the obtention process.
Carbonaceous material obtained by hydrothermal carbonization of sweet chestnut (Castanea sativa) cupule residues is a cost-effective, environmentally benign adsorbent, which effectively adds value to this biomass waste.
The carbons thus obtained show a limited surface area and a low microporosity while having a large volume of mesopores and macropores. All of them are acidic, with a low graphitization index, minimal moisture, and slight residual ash content; they are carbonaceous materials with a low mineral matter content.
It is possible to purify water contaminated with perfluorooctanoate by adsorption on this type of hydrothermal carbon. The adsorption process follows pseudo-second-order kinetics. Additionally, it fits the Freundlich model, showing that more than one layer can be adsorbed, which is expected in such an amorphous adsorbent.