*2.5. Computational Modeling*

DFT simulations were performed to study cation-hydrochar binding energies and geometries. The proposed hydrochar structures were modeled using DFT with the Gaussian 09 program [74] run via WebMO [75]. All the geometries were optimized using the Becke, 3-parameter, Lee–Yang–Parr (B3LYP) [76,77] hybrid functional, with the 6–311+G(d,*p*) basis set. Because the physical experiments included a water solvent, the polarizable continuum model (PCM) [78] was chosen to implicitly include the solvent in these calculations, assuming its dielectric constant was equal to 78. Model adsorption reactions were modeled and compared with one another to determine their energies and relative favorability. For each adsorption reaction, the products were hydrochar with a bound copper atom, and the respective ion, either hydrogen, sodium, or potassium, in solution. Multiple initial

configurations were studied to confirm that the final geometry captured a global minimum, rather than a local one.

#### **3. Results and Discussion**

The objective of this study development of rational methods to design hydrochar for heavy metal adsorption at the molecular level [79]. The study consists of several components: (1) measurements of sorption capacity of glucose hydrochar and comparison with activated carbon; (2) characterization and DFT modeling of glucose hydrochar; (3) custom-synthesis of hydrochars with tailored heavy metal capacity.

#### *3.1. Glucose Hydrochar Sorption Capacity and Characterization*

As a starting point, hydrochar was synthesized from a glucose precursor solution. The base structure of glucose hydrochar was recently reconciled between several models [61–63], with the finding that it consists of a furan-rich polymer decorated with OFG groups. Accordingly, and following previous reports, we hypothesized that the OFG groups would encourage heavy metal sorption. Unfortunately, glucose hydrochar exhibited negligible Cu(II) sorption capacity, <3 mg g<sup>−</sup>1, as shown in Table 1. The negligible Cu(II) capacity measured for glucose hydrochar seemingly contradicted previous literature descriptions of hydrochar absorption capacity [80–82], thus, motivating further examination.


**Table 1.** Adsorption Capacity and Surface Area of Hydrochar and Activated Carbon. BET = Brunauer–Emmett–Teller model.

Sun et al. [37] reported that alkali treatment increased the sorption capacity of hydrochar by 2–3 times, motivating the study of alkali treatment in the current study. As shown in Table 1, alkali treatment greatly increased the Cu(II) capacity, by at least an order of magnitude compared with the original glucose hydrochar. Several different bases were evaluated, with the finding that strong bases (hydroxides) outperformed weak ones (carbonates) and that bases featuring the potassium cation outperformed ones possessing sodium.

We compared the capacity of alkali-activated glucose hydrochar with several different activated carbons (Table 1), selected to cover a range of properties [83]. Interestingly, the activated carbons exhibited much greater sorption capacity than glucose hydrochar without activation, but less capacity than their alkali-activated forms. Alkali treatment was evaluated for two of the activated carbons, and it was found that the treatment either had no effect (Norit® SX1) or even negative effect (Nuchar®) on sorption capacity. The different response to alkali treatment observed for activated carbon and hydrochar clearly points to differences in the mechanism that must be understood for molecular-level hydrochar design.

To understand the adsorption results presented in Table 1, surface areas were measured using N2 sorption and the BET isotherm fitting method. Consistent with previous reports [29,37,43], the measured BET surface area of hydrochar was <10 m2 g−1. Alkali treatment had no effect on the measured hydrochar surface area, allowing us to reject the hypothesis that the effect of the treatment was to open up the hydrochar pore structure [82,84]. Similarly, consistent with previous reports [83], the BET surface areas of the activated carbons were >800 m<sup>2</sup> g−<sup>1</sup> and not affected by the dilute alkali treatment. Accordingly, while the capacity of glucose hydrochar is comparable to activated carbon on a mass basis, on a surface area basis the capacity is orders of magnitudes greater. This observation clearly points to a specific hydrochar-sorbate interaction that can be engineered to maximize adsorption.

The strongest common sorbate-hydrochar interaction is electrostatic [85], which can be understood as the interaction between the positively charged metal cation and negatively charged functional groups on the hydrochar surface. Accordingly, as a way to understand and quantify hydrochar surface charge, we measured hydrochar zeta potential before and after alkali activation and over a wide range of pH, from 2–12. Figure 1 presents the results, showing that zeta potential of alkali activated hydrochar was much more negative in the pH range of interest (pH < 7) than the parent hydrochar. Under strongly alkali pH, the zeta potential of the parent and the alkali treated material are the same to within the limits of experimental uncertainty, which is consistent with expectations given that the alkali treatment is simply immersion in an alkali solution with pH > 9.

**Figure 1.** Comparison of zeta potential of different sorbent materials as a function of pH.

Zeta potential measurements support electrostatic interaction as the primary basis of cation sorption to the hydrochar, providing a valuable clue for rational design. For further comparison, we measured the zeta potential of one of the aforementioned activated carbons (Norit® SX-1) and include these data in Figure 1. The zeta potential of the activated carbon was much less negative than glucose hydrochar, even before alkali treatment. Again, this points to a qualitatively different sorption mechanism for activated carbon compared to hydrochar, with cation sorption to activated carbon likely occurring due to cation-π interactions, which appear to be less important than electrostatic interactions for cation binding to hydrochar [86].

Rational sorbent design requires understanding the molecular binding sites. Figure 1 clearly implicates the importance of groups that ionize on alkali treatment, which naturally suggests carboxylic acids, acid anhydrides, and strongly acidic aromatic alcohols, such as phenol [16,24,73]. The fact that strong bases were more activating than weak ones (Table 1) seems consistent with de-protonation of weak acids but does not provide sufficient molecular detail for sorbent design. Accordingly, glucose hydrochar was analyzed using FT-IR for identification of ionizable OFGs. Figure 2 provides FT-IR spectra divided into the fingerprint region, 1000–2000 cm−<sup>1</sup> (Figure 2a) and the C−H and O−H stretching region, 2400–4000 cm−<sup>1</sup> (Figure 2b). Prior to alkali treatment, glucose hydrochar exhibits bands attributable to carbonyl (1720 cm−1) and hydroxyl (3200–3400 cm−1) groups associated with carboxylic acids. Other features at 2900, 1600, and 1200 cm−<sup>1</sup> are attributable to C−H stretches, arenes/furans breathing modes, and C−O stretches, respectively [65] These spectroscopic attributions are broadly consistent with the structural models previously inferred from NMR [65], Raman [61], and NEXAFS [62–64,66]. Interestingly, the C−H and O−H stretches are sufficiently differentiated to suggest that carboxylic acid exists in its free, rather than dimerized, state [87].

**Figure 2.** FT-IR spectra of glucose hydrochar as synthesized and after activation by various bases. (**a**) the fingerpint region (1000–2000 cm<sup>−</sup>1) (**b**) the O−H and C−H stretch region (2500–4000 cm<sup>−</sup>1).

After alkali treatment, the intensities of the hydroxyl band at 3200–3400 cm−<sup>1</sup> and carbonyl band at 1720 cm−<sup>1</sup> become much less intense. Simultaneously with these changes, the intensity of the band at 1600 cm−<sup>1</sup> increases and the feature broadens noticeably. The C−O stretch present at approximately 1200 cm−<sup>1</sup> becomes less intense and broader after alkali treatment. The effects are more noticeable for treatment with the hydroxides than the carbonates, consistent with their relative basicities and with the observed effects on sorption capacity, noted in Table 1.

All of the aforementioned changes observed in the FT-IR spectra of glucose hydrochar after alkali treatment are attributable to deprotonation of carboxylic acid groups to form carboxylates [37,88]. Specifically, deprotonation involves a shift of the main carbonyl band from approximately 1700 to about 1600 cm<sup>−</sup>1; [84] a reduction of intensity of the C−O stretch at 1200 cm<sup>−</sup>1; and a reduction of the intensity of the O−H stretch at 3200–3400 cm<sup>−</sup>1. The last of these is consistent with partial removal of the H atoms involved with O−H stretches, as expected for de-protonation. The carboxylate feature at 1600 cm−<sup>1</sup> overlaps with the furan/arene breathing mode that is characteristic of hydrochar [61,63,89]. The fact that alkali treated hydrochar still exhibits an O−H stretching band is consistent either with incomplete de-protonation of acid groups or with the presence of multiple forms of O−H in the structure (i.e., alcohol groups that are not sufficiently acidic to be deprotonated).

Since the carboxylic acid groups present in glucose hydrochar appear to be primarily in their protonated forms (Figure 1) and since the pH of the HTC reaction mixture is about 3 [90], the pKa of these acid groups must be greater than approximately 3–otherwise, they would be present in hydrochar in their deprotonated forms. Alkali treatment then deprotonates these groups, resulting in formation of the alkali carboxylate. Because hydrochar is a complex material and because localized induction and steric effects can influence pKa [91], carboxylic acids present in hydrochar likely possess a range of pKa's. In fact, this assertion is supported from the broad zeta potential curve observed for glucose hydrochar and shown in Figure 1. Treatment with carbonates may therefore deprotonate only the strongest carboxylic acids present in hydrochar, while treatment with hydroxides deprotonates both strong and weak carboxylic acids.

We considered the possibility of alternative ionizable groups, aside from carboxylic acid. Treatment with hydroxide would partially deprotonate any strongly acidic alcohol groups (e.g., phenols) present in the hydrochar structure; however, the FT-IR spectra show no direct evidence to support the formation of phenolate ions, nor do reported structural models suggest the presence of phenol in hydrochar [24,61,63]. Accordingly, metal-carboxylate binding appears to be the primary cation adsorption mechanism underlying glucose hydrochar sorption, providing a clear target for molecular simulation.

Hydrochar is thought to be composed of furan/arene polymers connected by alkyl spacers. Mild alkali treatment is insufficient to break or form covalent bonds present in this structure [82], which is consistent with the negligible change in surface area associated with alkali treatment (see Table 1). That stated, Mihajlovic et al. [92] proposed that hydrolytic degradation of OFGs can sometimes occur during alkali treatment, and re-arrangement of the hydrochar structure from its hypothesized form would complicate attempts to model the binding site. Accordingly, we searched for evidence of bond breaking in the spectra shown in Figure 2. Inspection of the FT-IR spectra suggests that important hydrochar features [28,29] at 1020 cm−<sup>1</sup> (C−OH alcohol and/alkyl-substituted ethers groups), 1600 cm−<sup>1</sup> (furan/arene breathing modes), and 2900–3000 cm−<sup>1</sup> (C−H stretch) are not affected by alkali treatment, consistent with the main effect of alkali treatment being confined to deprotonation rather than making and/or breaking of covalent bonds. This observation permits use of published hydrochar structural models to recreate the local environment of the metal-carboxylate binding site for DFT simulation.
