*3.2. DFT Simulations of the Metal-Carboxylate Interactions*

Zeta potential measurements and FT-IR analysis clearly implicate metal-carboxylate binding. DFT simulations of the metal-carboxylate complex were performed to examine this hypothesis in more detail. In particular, we sought to answer three questions: (1) since a metal cation should be able to displace protons from carboxylic acid, why is alkali treatment required to activate the hydrochar? (2) Why do potassium salts outperform sodium salts? (3) What is the geometry of the binding site?

Simulating cation-carboxylate binding requires recreating a plausible local environment. The hydrochar molecule, pictured in Figure 3a, was created as a composite structure based on previous literature. Titirici et al. [93] demonstrated via NMR that the majority of the aromatic functionality of hydrochars synthesized from glucose at temperatures less than 200 ◦C can be attributed to furan groups. Latham et al. [63] supported this via NEXAFS while showing that carbonyl groups are also important. The previously mentioned IR spectra also indicate the presence of carbonyl groups and are in agreement with the model proposed by Latham et al. [63]. Accordingly, we recreated the local adsorption environment as a furanic dimer configuration to be consistent with published hydrochar structures [61]. The carboxylic acid/carboxylate group resides as a side chain on the alkyl linker between adjacent furan rings, consistent with the observation using FT-IR and the importance

of carboxylate groups inferred from sorption capacity measurements presented here. The local environment experienced by a metal cation during adsorption also includes water solvent molecules. Here, we recreated the water solvation effect using an implicit cavity model of the appropriate dielectric constant (taken as 78). Future work can improve the accuracy of our calculations by including explicit water molecules in the simulation.

**Figure 3.** Carboxylate-containing hydrochar structures optimized using Density Functional Theory (DFT). (**a**) Model of the base hydrochar molecule. The different structures in (**b**) involve carboxylate binding with hydrogen, potassium, and sodium, respectively, with the carboxylate, which make up the reactants of the modeled adsorption reactions. (**c**) Illustrates the interaction between Cu(II) and the carboxylate group. Estimated adsorption energies are provided as shown. Legend: carbon; hydrogen; oxygen; potassium; sodium; copper.

We then simulated a series of possible cation-carboxylate structures, starting with the hydrochar model shown in Figure 3a. The focus of these calculations was to answer the aforementioned questions that focus on elucidation of trends, rather than quantitative energy estimates. We then simulated cation binding, shown stoichiometrically in Figure 3b, by replacing either H+, K+, or Na<sup>+</sup> with the Cu(II) cation to form the final structure shown in Figure 3c.

Consistent with experimental observations (Table 1), we find that replacing H<sup>+</sup> with Cu(II) is energetically unfavorable, whereas replacing K<sup>+</sup> and Na<sup>+</sup> is energetically favorable. The simulated energies are consistent with the observation that glucose hydrochar requires alkali treatment prior to activation. Moreover, DFT simulations predict that replacing K<sup>+</sup> is energetically more favorable than replacing Na<sup>+</sup> by 10.72 kJ mol<sup>−</sup>1, which is consistent with the observation that KOH is a more effective activating salt than NaOH and K2CO3 is more effective than Na2CO3. That stated, the calculated energy difference between K<sup>+</sup> and Na<sup>+</sup> substitution is relatively modest, which is again consistent with experimental observation. Note that for these reactions the cations in solution may not be properly modeled by implicit solvation, which is why for instance the replacement of a hydrogen by Cu(II) is so endothermic. Nonetheless the trends in cation exchange are captured by the DFT calculations.

Figure 3c shows the optimized geometry of a Cu-hydrochar structure. Here, the Cu-carboxylate bond length is approximately 1.85 Å, slightly longer than that associated with the distance between the proton and carboxylate group in carboxylic acid. The longer bond is consistent with the size of the Cu(II) ion compared with the proton [37].

The DFT simulations summarized in Figure 3 explain that alkali treatment removes the proton to activate the sorption capacity of glucose hydrochar. Physically, the proton is more tightly bonded to the carboxylate group than the metal cations, owing to the differences in ionic radii and the strong effect of ion-ion distance on the strength of electrostatic interactions [94]. Similarly, the differences observed between potassium and sodium can be ascribed to their relative ionic radii.

Interestingly, alkali treatment is not always reported as a necessary step for observation of hydrochar sorption capacity. This may be due to differences in the reaction mixture pH for different precursors and/or the presence of alkali salts in many hydrochar starting materials [95,96]. Accordingly, subtle differences in the reaction mixture and the composition of the precursor may decide whether or not alkali treatment is required to activate a given hydrochar for metal adsorption. Alternatively, the alkali step may not be uniformly reported, even when it is required. We recommend more consistent reporting of alkali treatment and reaction mixture pH in future work in this area.

#### *3.3. Custom-Synthesis of Hydrochar for Heavy Metal Adsorption*

Experiments and DFT simulations clearly implicate the importance of metal-carboxylate interactions in hydrochar adsorption of Cu(II). Accordingly, our next step was custom synthesis of a hydrochar for heavy metal adsorption. Following the work of Demir-Cakan et al. [56], we elected to synthesize a hydrochar by co-processing glucose and acrylic acid. Acrylic acid possesses a polymerizable double bond, which can form covalent linkages with the alkyl linker groups in the hydrochar structure, thereby increasing the density of carboxylate groups in the resulting material. We term the resulting material acrylic acid-hydrochar, or simply AA-hydrochar. Demir-Cakan et al. [56] reported synthesis of a series of AA-hydrochars, starting with different amounts of acrylic acid in the precursor mixture. Here, we selected a precursor mixture with composition similar to the optimum reported by Demir-Cakan et al. [56] as a proof of concept.

Table 2 provides the sorption capacity and surface area measurements for AA-hydrochar. As expected from DFT simulations, without activation the Cu(II) sorption capacity of AA-hydrochar is negligible. Interestingly, Demir-Cakan et al. [56] did not report alkali activation of their materials, which might be attributable to their study of Pb(IV) and Cd(II) whereas we studied Cu(II) or the aforementioned impact of subtle differences in reaction mixture pH on hydrochar protonation and subsequent sorption capacity. Regardless, after alkali activation, sorption capacity increases by at least an order of magnitude for the AA-hydrochar, and strong bases are again more effective than weak bases. AA-hydrochar capacity for Cu(II) sorption is greater than that observed for standard glucose hydrochar (50 <sup>±</sup> 4 compared with 40 <sup>±</sup> 4 mg g<sup>−</sup>1). Again, the effect is not as pronounced as reported by Demir-Cakan et al. [56], but it is consistent with the design concept.


**Table 2.** Adsorption capacity and surface area of custom-synthesized hydrochar and ion exchange resins.

The effect of AA and glucose co-processing to produce hydrochar was consistent with our expectations, but consistency does not imply confirmation and we considered alternative hypotheses. Table 2 shows that the surface area of AA-hydrochar was similar to the glucose hydrochar, eliminating surface area changes as a major difference between these materials. To understand further, we studied the OFGs of AA-hydrochar using FT-IR. Figure 4 provides the FT-IR spectra obtained for AA-hydrochar before and after KOH treatment. The FT-IR of glucose hydrochar is included in Figure 4 for direct comparison to show that the AA-hydrochar spectrum exhibits much more intense bands associated with carboxylic acids at 1720 (carbonyl) and 1200 cm−<sup>1</sup> (C-O stretch) than glucose hydrochar. In fact, the carboxylic acid bands dominate the AA-hydrochar spectrum and appear as the most prominent features. The band at 1600 cm<sup>−</sup>1, which is characteristic of furans and arenes, appears only as a minor, though distinct, feature in the AA-hydrochar spectrum. In comparison, the furan/arene band is one of the most prominent features in the glucose hydrochar spectrum. Similarly, after alkali treatment, the carbonyl band shifts to approximately 1550 cm−<sup>1</sup> and becomes the most prominent feature in the AA-hydrochar spectrum. Correspondingly, the C-O stretch feature shifts and broadens. Taken together, these observations clearly indicate that AA-hydrochar has abundant carboxylic acid groups that deprotonate after alkali treatment.

**Figure 4.** FT-IR spectra of AA-hydrochar before and after KOH treatment. G-hydrochar is synthesized entirely of glucose precursor, shown before alkali treatment. AA-hydrochar synthesized from co-feed of acrylic acid and glucose, shown before and after alkali treatment using KOH.

The performance of hydrochar sorbents is often compared with activated carbon [25,39]. However, our findings indicate that hydrochar adsorption is mediated by metal-carboxylate binding interactions that are more similar to what occurs on an ion exchange resin, rather than activated carbon. Accordingly, we measured Cu(II) sorption capacity of two commercial ion exchange resins, Amberlyst®-15 and AG® 50W-X4. Capacity results for these resins are provided in Table 2. Interestingly, these resins far outperform activated carbon (Table 1) and outperform by about a factor of two the AA-hydrochar. Moreover, the ion exchange resins did not require alkali activation, unlike the hydrochars. Since the resins outperform hydrochar, even AA-hydrochar, we sought to understand the differences between the resins and the hydrochar as part of our rational design approach.

An obvious potential difference between the resins and the hydrochar is surface area. Table 2 provides N2 sorption-based BET surface areas for Amberlyst®-15 and AG® 50W-X4. Interestingly, Amberlyst®-15 exhibits much greater surface area than any of the hydrochars, which could explain its superior performance. However, the surface area measured for AG® 50W-X4 was less than any of the other materials (<1 m2 g<sup>−</sup>1, the instrument detection limit), meaning that surface area considerations alone cannot explain the performance of the resins–at least not the surface area measured by N2 sorption and estimated by BET analysis of the isotherm. In fact, the swelling behavior of ion exchange resins has been studied carefully in water and other solvents [98–100]; swelling in the presence of water likely opens the pore structure of AG® 50W-X4 (and possibly the other sorbents), accounting for its sorption capacity despite negligible N2 sorption surface area. Understanding the effects of hydrochar swelling on surface area available for cation sorption is an area that should be studied in the future.

The binding site in both Amberlyst®-15 and AG® 50W-X4 is a sulfonate group [101], whereas the findings presented here indicate that carboxylate groups are mainly responsible for binding in glucose hydrochar and especially AA-hydrochar. The sulfonic acid group is at least 1000× stronger than the carboxylic acid group, meaning that this difference could explain sorption behavior and the need for alkali activation. Accordingly, we modified the acrylic acid synthesis procedure for incorporation of a sulfonate group into the hydrochar structure by co-processing glucose and vinyl sulfonic acid. Like acrylic acid, vinyl sulfonic acid possesses a polymerizable double bond that can be incorporated in the hydrochar-alkylated backbone. Unlike acrylic acid, though, vinyl sulfonic acid can introduce a sulfonate group into the hydrochar instead of the carboxylic acid introduced by acrylic acid. Accordingly, we term this new char vinyl sulfonic acid-hydrochar, or VSA-hydrochar.

Table 2 provides the Cu(II) cation sorption capacity of VSA-hydrochar. Interestingly, despite the strength of the vinyl sulfonic acid precursor (pKa < 1 compared with 4.35 for acrylic acid) [98,102], we observed negligible Cu(II) sorption capacity for VSA-hydrochar before treatment with alkali. After treatment with KOH, the Cu(II) capacity of VSA-hydrochar increased substantially to <sup>51</sup> <sup>±</sup> 3 mg g<sup>−</sup>1. Interestingly, NaOH was much less effective at increasing Cu(II) sorption capacity than KOH, consistent with the aforementioned trend observed and simulated for carboxylate binding.

As before, the measured BET surface area of VSA-hydrochar was in the same range as the other hydrochars and <10 m2 g−1. Similarly, Figure 5 shows the FT-IR spectrum of VSA-hydrochar. Unlike carboxylic acid and carboxylate groups that have intense and well-differentiated vibrational bands, sulfonic acid and sulfonate give rise to weak and broad bands that are not easily differentiated from other features [103]. That stated, the FT-IR VSA-hydrochar spectrum contains bands in the range expected for sulfonic acid (1100–1300 cm<sup>−</sup>1). The carboxylate/carboxylic acid bands are less intense in VSA-hydrochar than glucose hydrochar, indicating substitution of the weak acid in AA-hydrochar for the strong acid in VSA-hydrochar.

Cu(II)-sulfonate structures were simulated using DFT methods, similar to those previously presented for carboxylate binding. Figure 6a shows the sulfonic acid- hydrochar geometry, which consisted of two furan groups bonded to a sulfonic acid group. As before, binding was simulated as an exchange of Cu(II) for H+, K+, and Na+. Despite the strength of the sulfonic acid, DFT calculations found that replacing H<sup>+</sup> with Cu(II) was thermodynamically unfavorable, consistent with the need to activated VSA-hydrochar with alkali. Figure 6b summarizes this result. Similarly, the distance of the Cu−O bonded to sulfonate (shown in Figure 6c) is 1.95 Å, somewhat greater than the Cu−O bond in carboxylate hydrochar (1.85 Å). As before, the cations in solution may not be properly modeled by implicit solvation, which is why some energies may be so large, despite DFT identifying the trends in cation exchange.

**Figure 5.** FT-IR spectrum of VSA-hydrochar, before and after KOH treatment. Vertical lines mark the carboxylic acid and carboxylate vibration bands. The region where sulfonic acid vibrations appear is indicated. The spectrum of glucose hydrochar is reproduced from Figure 4 as a point of reference.

Surface area, FT-IR, and DFT simulations provide further evidence of cation-sulfonate binding in the VSA-hydrochar, but do not explain why the performance of neither VSA-hydrochar nor AA-hydrochar can match the commercial ion exchange resins. As a final hypothesis, we quantified the density of surface acids present on the various sorbents, with the expectation that differences in the density of surface acids might explain observed differences in sorption capacity. For these experiments, hydrochars were first treated with strong acid (HCl) to protonate fully all available acid groups. Then, the acid group density was measured of the protonated sorbent using Boehm titration methods [29,71,72].

Table 3 summarizes the carboxylic acid site density measurements. As expected, the density of acid functional groups on the glucose hydrochar is much greater than on the activated carbons considered here, consistent with the different adsorption mechanisms for the two materials (primarily electrostatic vs. primarily π-cation). The ion exchange resins have much greater acid concentrations than any of the other sorbents, consistent with their superior performance and indicating that the AAand VSA-hydrochars function as designed, albeit with fewer acid binding groups than are available on the ion exchange resins tested here. Nonetheless, the Cu(II) adsorption performance of the designer hydrochars is comparable to the ion exchange resins (to within a factor of two) and superior to activated carbon, meaning that strategies to increase acid functional group density can be effective for synthesis of task-specific hydrochar sorbents.

Table 3 provides qualitative evidence of the importance of acid group density on sorption performance and permits analysis of a critical parameter: the binding stoichiometry of the metal-acid complex formed during adsorption. Binding stoichiometry is important for quantifying sorbent performance since the ideal absorbent will possess high density of binding sites and utilize them as efficiently as possible. Simultaneously achieving high binding site density and binding site utilization may not be possible, since densely spaced binding sites may promote bidentate binding instead of monodentate binding, which is less efficient binding site utilization. We analyzed the sorption and acid site density data to evaluate these effects in hydrochar, activated carbon, and ion exchange resins.

**Figure 6.** Sulfonate-containing hydrochar structures optimized using DFT. The structure in (**a**) is the sulfonated hydrochar molecule. (**b**) Depicts the reactants which are the interactions between the sulfonate group and either hydrogen, potassium, or sodium, respectively. Structure (**c**) shows binding of Cu(II) to the sulfonated hydrochar molecule. Adsorption energies are provided as shown. Legend: carbon; hydrogen; oxygen; sulfur; potassium; sodium; copper.



<sup>a</sup> the estimated acid site detection limit. <sup>b</sup> based on estimated BET surface area.

To use Table 3 data to understand stoichiometry, we plotted Cu(II) sorption capacity as a function of measured acid group density, converting both to molar quantities, as shown in Figure 7. For comparison, lines of constant ion-binding site stoichiometry (two Cu ions per acid, 1:1, and 1:2) are shown. Data for the activated carbons cluster around the origin and fall entirely off the stoichiometric trend lines, as expected given that the sorption mechanism to activated carbons is likely cation-π interactions and is, therefore, independent of acid site density. In contrast, sorption for the hydrochars falls between the 1:1 and 1:2 stoichiometry lines, indicating that—on average—each acid group binds approximately 0.75 Cu ions. This again is further quantitative evidence of the importance of electrostatic interactions for binding to hydrochar.

depicts activated carbon, ion exchange resins, hydrochar.

The stoichiometry inferred from Figure 7 shows that the designer hydrochars outperform Amberlyst®-15 on a per acid site basis. This is an important finding since increasing binding site utilization efficiency is an effective means of increasing sorption capacity, along with increasing the density of binding sites themselves. Interestingly, AG® 50W-X4 far exceeds all other sorbents on effectiveness per acid site, with nearly two Cu ions associated with every acid site (Figure 6). The difference between AG® 50W-X4 and Amberlyst®-15 is noteworthy as both sorbents are described in the literature as polymerized styrene backbones with periodic sulfonic acid group substitution [101]. The difference in their performance must be due either to (1) the ability of the sorbent to hold charge, which could be saturated for Amberlyst®-15 limiting its sorption capacity, (2) differences in acid site accessibility in the swollen resins and hydrochars, or (3) differences in the spatial proximity of the acid binding groups in the different sorbents. The performance of AG® 50W-X4 suggests further engineering of the hydrochar structure to optimize sorption capacity.

To understand the origins of stoichiometry between Cu(II) ions and carboxylate or sulfonate groups, we performed simulations to compare monodentate with bidentate binding of Cu(II) to carboxylate and sulfonate groups. To make the calculation accessible using DFT, we simplified the structure previously used in Figures 3 and 6 to remove the furan groups. Figure 8a shows the optimized geometry for the monodentate binding structures, and Figure 8b shows the optimized geometry for the bidentate binding structures. As expected, bidentate binding is much more energetically favorable than binding to a single acid functional group. For the sulfonate site, bidentate binding is more stable by 201 kJ mol<sup>−</sup>1, and for the carboxylate site, bidentate binding is more stable by 126 kJ mol−1. These values indicate a clear thermodynamic preference for bidentate binding. As measured by the Cu−O distance, the Cu(II) ion is roughly equidistant between the two sulfonate groups; as a result, the Cu−O distance in bidentate binding complex is greater than found in the geometry optimized for single Cu-acid stoichiometry (shown previously in Figure 6). This clearly shows that Cu(II) (and presumably other double charged cations) will prefer bidentate binding, when such an option is available. Since bidentate binding is a less efficient use of sites than monodentate binding, rational design of hydrochars should attempt to achieve uniform acid spacing to minimize acid-acid interaction and the ability of cations to bind simultaneously to multiple acid sites.

**Figure 8.** Optimized geometries of monodentate (**a**) and bidentate (**b**) binding of Cu(II) to hydrochar carboxylate and sulfonate groups simulated using DFT. Oxygen-Cu(II) distances are shown for reference. Legend: carbon; hydrogen; oxygen; sulfur; copper.

When functional group precursors with polymerizable double bonds are co-fed to the HTC reactor with glucose, the functional group bearing molecules will polymerize primarily with each other, rather than with groups present in the hydrochar. Because they are formed by co-feeding glucose and vinyl groups, VSA-hydrochar is not engineered to achieve the desired spacing, which may explain why it falls short of the desired 1:1 Cu-binding site stoichiometry. More uniform spacing of the binding groups has potential to improve binding site utilization by forcing binding to occur via the preferred monodentate arrangement rather than via the thermodynamically preferred bidentate geometry. Furthermore, utilization of vinyl sulfonic acid as a source of binding groups detracts from the renewable and green characteristics of hydrochar. Accordingly, future work in this area should seek to utilize feeds that are naturally abundant in anionic binding sites and/or functional groups that are converted into anionic binding sites during HTC. Questions of binding site access and cooperative effects should be addressed for hydrochars synthesized from renewable or waste resources, using similar methods as shown here for rational sorbent design.

As a final analysis, we evaluated cation-π binding to the furan backbone itself in the absence of acid groups, as a comparison with the arene backbone present in commercial exchange resins. By providing a secondary stabilizing interaction, optimizing the cation-π binding interaction can potentially improve the utilization efficiency of the anionic binding sites—a desired goal as explained previously. In particular, we were interested to understand the effect of locating the cation between nearby rings as compared with interacting with a single ring individually—in the absence of

anionic binding groups, such as sulfonate or carboxylate. Figure 9 provides the results of these calculations; Figure 9a,b depict arene binding and Figure 9c,d depict furan binding, respectively. In both cases, locating the Cu(II) between two nearby aromatic rings (either furan or arene) is more stable than interaction with a single aromatic ring. Interestingly, the energy difference is greater for furan-cation interactions (Figure 9c,d) than arene-cation interactions (Figure 9a,b), by approximately 23 kJ mol<sup>−</sup>1. Accordingly, a final design consideration for custom-synthesis of hydrochar sorbents is inclusion of geometries, which permit formation of furan "pockets" for optimized cation-π interaction. When combined with electrostatic interactions, cation-π interactions can provide a secondary stabilizing force to optimize hydrochar sorption capacity.

**Figure 9.** Optimized geometries calculated for arene, (**a**,**b**), and furan, (**c**,**d**), cation-π binding, in the absence of anionic binding groups. Legend: carbon; hydrogen; oxygen; sulfur; copper. Differences in energy are shown directly in the Figure.

Figures 3 and 6–9 describe a combined experimental and simulation approach for rational design of hydrochar sorbents to exploit electrostatic interactions between anionic functional groups and metal cations. Maximizing the effectiveness of each functional group can be achieved by spacing them uniformly throughout the material, thus emphasizing monodentate binding over bidentate binding. Presumably, highly effective hydrochar sorbents, such as those reported by Demir-Cakan et al. [56], who first demonstrated the acrylic acid co-HTC approach, and Xue et al. [16], who activated peanut hull hydrochar using hydrogen peroxide, already exploit these principles. Likewise, the accuracy of the computational approach in particular will benefit as hydrochar structure is further resolved, especially for materials produced from precursors other than glucose. The result of the rational design approach will be hydrochars with maximized value; thus, making them as competitive as possible with sorbents obtained from non-renewable resources. Although not within the scope of this work, computational modeling should be appropriate for guiding selection of conditions for hydrochar regeneration, for example by using alkali solutions to remove the heavy metal adsorbates. Applying our adsorption capacity results directly to metals other than Cu(II) is not recommended; however, the combined experimental and computational approach should be amenable to any metal cation of interest. Similar analysis can be applied in the future to understand the adsorption of organic substances to hydrochar, as organic pollutants will exhibit different hydrochar interactions than metal cations [104].
