Supercapacitor Performance of Magnetite Nanoparticles Enhanced by a Catecholate Dispersant: Experiment and Theory

The full potential of Fe3O4 for supercapacitor applications can be achieved by addressing challenges in colloidal fabrication of high active mass electrodes. Exceptional adsorption properties of catecholate-type 3,4-dihydroxybenzoic acid (DHBA) molecules are explored for surface modification of Fe3O4 nanoparticles to enhance their colloidal dispersion as verified by sedimentation test results and Fourier-transform infrared spectroscopy measurements. Electrodes prepared in the presence of DHBA show nearly double capacitance at slow charging rates as compared to the control samples without the dispersant or with benzoic acid as a non-catecholate dispersant. Such electrodes with active mass of 40 mg cm−2 show a capacitance of 4.59 F cm−2 from cyclic voltammetry data at a scan rate of 2 mV s−1 and 4.72 F cm−2 from galvanostatic charge–discharge data at a current density of 3 mA cm−2. Experimental results are corroborated by density functional theory (DFT) analysis of adsorption behaviour of DHBA and benzoic acid at the (001) surface of Fe3O4. The strongest adsorption energy (ca. −1.8 eV per molecule) is due to the catechol group of DHBA. DFT analysis provides understanding of the basic mechanism of DHBA adsorption on the surface of nanoparticles and opens the way for fabrication of electrodes with high capacitance.


Introduction
Surface modification of materials using molecules from the catechol family is emerging as a new area of technological and scientific interest [1,2]. Catecholate molecules can achieve extremely strong adhesion to inorganic surfaces by utilizing a similar adsorption mechanism to that of mussel proteins when bonded to different surfaces [2][3][4]. Very fast and very strong adsorption to various inorganic substrates is a necessity for mussels to avoid damage by sea waves. This adhesion mechanism is attributed to bonding of the catechol ligands of the mussel proteins to metal atoms on material surfaces [4][5][6][7][8].
Polymer molecules were modified with catechol ligands for the fabrication of adherent polymer and composite films [4,[9][10][11][12]. Catecholate molecules were used as charged dispersants for electrophoretic deposition of inorganic and organic materials [5,13,14]. Catecholate molecules were also utilized as particle transfer vehicles for liquid-liquid extraction [15], which prevented nanoparticle agglomeration. In this technology, inorganic particles were modified with catechol ligands in the synthesis medium or at the liquidliquid interface and transferred from the synthesis medium directly to the device processing medium [15]. Catechol molecules facilitated efficient and versatile extraction [16] of MnO 2 , Mn 3 O 4 , FeOOH, and ZnO for energy storage and sensor applications. The phase transfer of the particles resulted in reduced agglomeration, which allowed for improved electrolyte access to the particle surface [16] and improved device performance.
Recent studies highlighted benefits of catecholate molecules for surface modification of materials for applications in dye-sensitized solar cells and photoelectrochemical sensors [17][18][19]. For dye-sensitized solar cells, a staggered gap heterojunction can occur at a application of density functional theory (DFT) [57] for the analysis of the DHBA adsorption mechanism. Fe 3 O 4 has emerged as a promising supercapacitor material due to its low cost and environmentally benign nature [58]. Here, we aimed at improving electrochemical performance of fabricated Fe 3 O 4 supercapacitor anodes and selected DHBA as a catechol molecule to play the role of a capping and dispersing agent. The chemical structure of DHBA contains adjacent phenolic OH functional groups on the 3rd and 4th carbons on the benzene ring (Figure 1), providing strong adsorption to the particle surface, while a carboxyl group can facilitate dispersion via electrostatic repulsion [59,60]. These adsorption and dispersion properties are critical to prevent agglomeration of nanoparticles in solution, particularly materials with magnetic properties, such as Fe 3 O 4 . Benzoic acid (BA) is chosen as a negative control in our experimental design. The chemical structures of DHBA and BA differ only by the adjacent phenolic OH groups present in DHBA but not in BA (Figure 1). To corroborate the understanding of the adsorption processes seen in our experimental results, we used DFT to model DHBA and BA adsorption on the surface of Fe 3 O 4 and assess their binding affinities and mechanisms.
Molecules 2023, 28, x FOR PEER REVIEW 3 of 18 synthesis of Fe3O4 particles, fabrication, and testing of supercapacitor electrodes and application of density functional theory (DFT) [57] for the analysis of the DHBA adsorption mechanism. Fe3O4 has emerged as a promising supercapacitor material due to its low cost and environmentally benign nature [58]. Here, we aimed at improving electrochemical performance of fabricated Fe3O4 supercapacitor anodes and selected DHBA as a catechol molecule to play the role of a capping and dispersing agent. The chemical structure of DHBA contains adjacent phenolic OH functional groups on the 3rd and 4th carbons on the benzene ring (Figure 1), providing strong adsorption to the particle surface, while a carboxyl group can facilitate dispersion via electrostatic repulsion [59,60]. These adsorption and dispersion properties are critical to prevent agglomeration of nanoparticles in solution, particularly materials with magnetic properties, such as Fe3O4. Benzoic acid (BA) is chosen as a negative control in our experimental design. The chemical structures of DHBA and BA differ only by the adjacent phenolic OH groups present in DHBA but not in BA (Figure 1). To corroborate the understanding of the adsorption processes seen in our experimental results, we used DFT to model DHBA and BA adsorption on the surface of Fe3O4 and assess their binding affinities and mechanisms.

Experimental Results
Dispersing properties of BA and DHBA were investigated by analysing the suspension stability of Fe3O4 particles in water ( Figure 2). Fe3O4 suspensions without additives and containing BA showed rapid precipitation. However, the suspension prepared with DHBA as a dispersing agent remained stable, further confirming the role of the phenolic OH groups of the molecule for adsorption, which are functional groups that BA lacks. The synthesis of Fe3O4 was conducted at pH = 9, which is above the isoelectric point of Fe3O4 (pH = 6.5) [61]. Therefore, the as-prepared Fe3O4 particles were negatively charged. The adsorption of anionic DHBA on the negatively charged Fe3O4 particles indicated a strong adsorption power of the catecholate DHBA molecules.

Experimental Results
Dispersing properties of BA and DHBA were investigated by analysing the suspension stability of Fe 3 O 4 particles in water ( Figure 2). Fe 3 O 4 suspensions without additives and containing BA showed rapid precipitation. However, the suspension prepared with DHBA as a dispersing agent remained stable, further confirming the role of the phenolic OH groups of the molecule for adsorption, which are functional groups that BA lacks. The synthesis of Fe 3 O 4 was conducted at pH = 9, which is above the isoelectric point of Fe 3 O 4 (pH = 6.5) [61]. Therefore, the as-prepared Fe 3 O 4 particles were negatively charged. The adsorption of anionic DHBA on the negatively charged Fe 3 O 4 particles indicated a strong adsorption power of the catecholate DHBA molecules.
Molecules 2023, 28, x FOR PEER REVIEW 3 of 18 synthesis of Fe3O4 particles, fabrication, and testing of supercapacitor electrodes and application of density functional theory (DFT) [57] for the analysis of the DHBA adsorption mechanism. Fe3O4 has emerged as a promising supercapacitor material due to its low cost and environmentally benign nature [58]. Here, we aimed at improving electrochemical performance of fabricated Fe3O4 supercapacitor anodes and selected DHBA as a catechol molecule to play the role of a capping and dispersing agent. The chemical structure of DHBA contains adjacent phenolic OH functional groups on the 3rd and 4th carbons on the benzene ring (Figure 1), providing strong adsorption to the particle surface, while a carboxyl group can facilitate dispersion via electrostatic repulsion [59,60]. These adsorption and dispersion properties are critical to prevent agglomeration of nanoparticles in solution, particularly materials with magnetic properties, such as Fe3O4. Benzoic acid (BA) is chosen as a negative control in our experimental design. The chemical structures of DHBA and BA differ only by the adjacent phenolic OH groups present in DHBA but not in BA (Figure 1). To corroborate the understanding of the adsorption processes seen in our experimental results, we used DFT to model DHBA and BA adsorption on the surface of Fe3O4 and assess their binding affinities and mechanisms.

Experimental Results
Dispersing properties of BA and DHBA were investigated by analysing the suspension stability of Fe3O4 particles in water ( Figure 2). Fe3O4 suspensions without additives and containing BA showed rapid precipitation. However, the suspension prepared with DHBA as a dispersing agent remained stable, further confirming the role of the phenolic OH groups of the molecule for adsorption, which are functional groups that BA lacks. The synthesis of Fe3O4 was conducted at pH = 9, which is above the isoelectric point of Fe3O4 (pH = 6.5) [61]. Therefore, the as-prepared Fe3O4 particles were negatively charged. The adsorption of anionic DHBA on the negatively charged Fe3O4 particles indicated a strong adsorption power of the catecholate DHBA molecules.   The adsorption of DHBA on Fe3O4 particles was confirmed by results of Fourier Transform Infrared (FTIR) spectroscopy studies. Figure 3 compares FTIR spectra of Fe3O4 prepared without and with DHBA and pure DHBA. The FTIR spectrum of Fe3O4 prepared without DHBA (Figure 3a) showed absorptions at 1355, 1475, and 1630 cm −1 , which can be attributed [13,62] to stretching of surface OH groups and vibrations of adsorbed CO2. The broad absorption centred at 3260 cm −1 resulted from the O-H bending vibrations of adsorbed water [62]. Similar absorptions were observed in the spectrum of Fe3O4 prepared using DHBA ( Figure 3b). However, the spectrum of Fe3O4 prepared using DHBA showed additional absorptions due to adsorbed DHBA. The absorptions at 1097 and 1122 cm −1 resulted from bending C-H vibrations [63,64]. Similar absorptions were observed in the spectrum of DHBA (Figure 3c). The C-O stretching and bending vibrations [7,64] of DHBA resulted in a broad absorption ( Figure  3c) centred around 1245 cm −1 and corresponding absorptions at 1218 and 1270 cm −1 for Fe3O4 prepared using DHBA (Figure 3b), which were not observed in the spectrum of Fe3O4 prepared without DHBA (Figure 3a). FTIR studies also revealed additional peaks related to the C-C vibrations in the range of 1300-1600 cm −1 in the spectrum of Fe3O4 prepared using DHBA (Figure 3b), which were not observed in the spectrum of Fe3O4 prepared without DHBA (Figure 3a). Such peaks, related to C-C stretching of aromatic rings, were observed in the spectrum of pure DHBA (Figure 3c). Electrokinetic measurements showed that the adsorption of anionic DHBA resulted in increasing absolute value of negative zeta potential from −28.9 mV for Fe3O4 prepared without DHBA to −38.7 mV for Fe3O4 prepared using DHBA. Therefore, the results of sedimentation tests, zeta potential measurements, and FTIR spectroscopy indicated that DHBA adsorbed on Fe3O4 and facilitated the formation of suspensions with improved colloidal stability.
In this investigation, MWCNT were used as a conductive additive for the fabrication of composite Fe3O4-MWCNT electrodes with enhanced electronic conductivity.  The FTIR spectrum of Fe 3 O 4 prepared without DHBA (Figure 3a) showed absorptions at 1355, 1475, and 1630 cm −1 , which can be attributed [13,62] to stretching of surface OH groups and vibrations of adsorbed CO 2 . The broad absorption centred at 3260 cm −1 resulted from the O-H bending vibrations of adsorbed water [62]. Similar absorptions were observed in the spectrum of Fe 3 O 4 prepared using DHBA ( Figure 3b). However, the spectrum of Fe 3 O 4 prepared using DHBA showed additional absorptions due to adsorbed DHBA. The absorptions at 1097 and 1122 cm −1 resulted from bending C-H vibrations [63,64]. Similar absorptions were observed in the spectrum of DHBA (Figure 3c). The C-O stretching and bending vibrations [7,64] of DHBA resulted in a broad absorption (Figure 3c) centred around 1245 cm −1 and corresponding absorptions at 1218 and 1270 cm −1 for Fe 3 O 4 prepared using DHBA (Figure 3b), which were not observed in the spectrum of Fe 3 O 4 prepared without DHBA (Figure 3a). FTIR studies also revealed additional peaks related to the C-C vibrations in the range of 1300-1600 cm −1 in the spectrum of Fe 3 O 4 prepared using DHBA (Figure 3b), which were not observed in the spectrum of Fe 3 O 4 prepared without DHBA (Figure 3a). Such peaks, related to C-C stretching of aromatic rings, were observed in the spectrum of pure DHBA (Figure 3c). Electrokinetic measurements showed that the adsorption of anionic DHBA resulted in increasing absolute value of negative zeta potential from −28.9 mV for Fe 3 O 4 prepared without DHBA to −38.7 mV for Fe 3 O 4 prepared using DHBA. Therefore, the results of sedimentation tests, zeta potential measurements, and FTIR spectroscopy indicated that DHBA adsorbed on Fe 3 O 4 and facilitated the formation of suspensions with improved colloidal stability.
In this investigation, MWCNT were used as a conductive additive for the fabrication of composite Fe 3 O 4 -MWCNT electrodes with enhanced electronic conductivity. Figure 4 presents experimental results for the CV analysis of the Fe 3 O 4 -MWCNT composite electrodes. Electrodes were fabricated with synthesized Fe 3 O 4 and MWCNT in the presence of DHBA, BA, and no dispersing agent to investigate the impact of a dispersant on capacitive behaviour. Electrodes prepared using DHBA showed significant improvement in capacitance at scan rates 2-50 mV s −1 . The role of the phenolic OH groups of DHBA is evident when comparing CVs and obtained capacitances of electrodes fabricated with DHBA and BA as dispersants. The capacitance of 4.59 F cm −2 for DHBA at 2 mV s −1 scan rate represents significant improvement, compared to the 2.38 F cm −2 and 2.80 F cm −2 obtained when using no dispersing agent and BA dispersant, respectively. The slightly lower capacitance of electrodes prepared using DHBA at a high scan rate of 100 mV s −1 can be attributed to the insulating properties of adsorbed DHBA, which affect the capacitance at such high scan rate.
Molecules 2023, 28, x FOR PEER REVIEW 5 of 18 of DHBA, BA, and no dispersing agent to investigate the impact of a dispersant on capacitive behaviour. Electrodes prepared using DHBA showed significant improvement in capacitance at scan rates 2-50 mV s −1 . The role of the phenolic OH groups of DHBA is evident when comparing CVs and obtained capacitances of electrodes fabricated with DHBA and BA as dispersants. The capacitance of 4.59 F cm −2 for DHBA at 2 mV s −1 scan rate represents significant improvement, compared to the 2.38 F cm −2 and 2.80 F cm −2 obtained when using no dispersing agent and BA dispersant, respectively. The slightly lower capacitance of electrodes prepared using DHBA at a high scan rate of 100 mV s −1 can be attributed to the insulating properties of adsorbed DHBA, which affect the capacitance at such high scan rate.  Figure 5 compares the galvanostatic charge-discharge curves (A-C) for different electrodes, with corresponding capacitances (D) calculated from the discharge data at different current densities. GCD was performed in a voltage window of −0.8-0 V vs. SCE. Nearly symmetrical and triangular charge-discharge curves are seen, confirming the pseudocapacitive behaviour. Electrodes prepared with DHBA showed significant increase in discharge time and significant improvement in capacitance when compared to electrodes prepared using BA or no dispersant. A capacitance of 4.72 F cm −2 was obtained for electrodes made with DHBA at a current density of 3 mA cm −2 and decreased to 3.16 F cm −2 with increasing current density. This capacitance shows significant improvement compared to the capacitances of 2.76 F cm −2 and 2.37 F cm −2 obtained for the electrodes prepared with BA as a dispersing agent and no dispersing agent, respectively, at the current density of 3 mA cm −2 . It should be noted that capacitance decreases with increasing charge-discharge time. The difference in the capacitance values obtained using CV and GCD data can result from the different time scale ranges used in both experiments. The Coulombic efficiency at 3 mA cm −2 was 98.0%; it decreased with current density to the value of 97.7% at a current density of 40 mA cm −2 . The electrodes prepared using DHBA  Figure 5 compares the galvanostatic charge-discharge curves (A-C) for different electrodes, with corresponding capacitances (D) calculated from the discharge data at different current densities. GCD was performed in a voltage window of −0.8-0 V vs. SCE. Nearly symmetrical and triangular charge-discharge curves are seen, confirming the pseudocapacitive behaviour. Electrodes prepared with DHBA showed significant increase in discharge time and significant improvement in capacitance when compared to electrodes prepared using BA or no dispersant. A capacitance of 4.72 F cm −2 was obtained for electrodes made with DHBA at a current density of 3 mA cm −2 and decreased to 3.16 F cm −2 with increasing current density. This capacitance shows significant improvement compared to the capacitances of 2.76 F cm −2 and 2.37 F cm −2 obtained for the electrodes prepared with BA as a dispersing agent and no dispersing agent, respectively, at the current density of this investigation for negative electrodes is important for the development of asymmetric devices operating in Na 2 SO 4 electrolyte [55,56]. The obtained capacitance is higher than the capacitance of carbon electrodes of the same mass operating in the negative potential range [54]. It is comparable with capacitance of advanced positive electrodes [55,56]. It should be noted that catechol exhibits redox active properties in a positive potential range [25,65]. Specifically, the redox activity is associated with the transition of phenolic groups of DHBA to a quinone-like structure [66,67]. In this investigation, a catecholate type DHBA dispersant was used for the fabrication of negative electrodes. Therefore, redox reaction contribution of this dispersant to the to the electrode capacitance in the negative potential range is not expected. However, adsorbed catecholate molecules can facilitate charge transfer [21,35,36].
Molecules 2023, 28, x FOR PEER REVIEW 6 of 18 showed significantly higher capacitance, compared to electrodes prepared with BA and without dispersant in the current density range of 3-40 mA cm −2 . The high areal capacitance obtained in this investigation for negative electrodes is important for the development of asymmetric devices operating in Na2SO4 electrolyte [55,56]. The obtained capacitance is higher than the capacitance of carbon electrodes of the same mass operating in the negative potential range [54]. It is comparable with capacitance of advanced positive electrodes [55,56]. It should be noted that catechol exhibits redox active properties in a positive potential range [25,65]. Specifically, the redox activity is associated with the transition of phenolic groups of DHBA to a quinone-like structure [66,67]. In this investigation, a catecholate type DHBA dispersant was used for the fabrication of negative electrodes. Therefore, redox reaction contribution of this dispersant to the to the electrode capacitance in the negative potential range is not expected. However, adsorbed catecholate molecules can facilitate charge transfer [21,35,36]. The analysis of impedance data provides evidence of improved performance of the electrode fabricated using DHBA as a dispersing agent. Figure 6A shows significant increase in the low frequency capacitance (Cs') for the electrode fabricated with DHBA as a dispersing agent compared to the electrodes where BA and no dispersing agent was used. The analysis of frequency dependences of CS" shows an increase in the relaxation frequency for both electrodes prepared using DHBA and BA compared to the electrode prepared with no dispersing agent ( Figure 6B). The Nyquist plot, presented in Figure 6C, shows relatively low real part of impedance (Z'), indicating low resistance. Furthermore, the large slope shown in the Z" vs. Z' curves indicate good capacitive behaviour. Modelling of impedance spectroscopy data using equivalent circuits is an important strategy for the analysis of supercapacitor electrodes [68]. In this investigation, an equivalent circuit developed for high active mass bulk electrodes [69] was used. The simulation results The analysis of impedance data provides evidence of improved performance of the electrode fabricated using DHBA as a dispersing agent. Figure 6A shows significant increase in the low frequency capacitance (Cs ) for the electrode fabricated with DHBA as a dispersing agent compared to the electrodes where BA and no dispersing agent was used. The analysis of frequency dependences of C S shows an increase in the relaxation frequency for both electrodes prepared using DHBA and BA compared to the electrode prepared with no dispersing agent ( Figure 6B). The Nyquist plot, presented in Figure 6C, shows relatively low real part of impedance (Z ), indicating low resistance. Furthermore, the large slope shown in the Z vs. Z curves indicate good capacitive behaviour. Modelling of impedance spectroscopy data using equivalent circuits is an important strategy for the analysis of supercapacitor electrodes [68]. In this investigation, an equivalent circuit developed for high active mass bulk electrodes [69] was used. The simulation results presented in Figure 6D showed good agreement with experimental data and indicated that the electrodes exhibited low charge transfer resistance (R T ), which was found to be 0.08 Ohm.
presented in Figure 6D showed good agreement with experimental data and indicated that the electrodes exhibited low charge transfer resistance (RT), which was found to be 0.08 Ohm.
Fe3O4 is a promising negative electrode material for the development of asymmetric supercapacitors [52] with enlarged voltage window for operation in mild electrolytes, such as K2SO4 or Na2SO4. However, the lower specific capacitance of the negative electrodes [52], compared to the capacitance of positive electrodes, limits the development of such devices. Another difficulty is related to significant decrease in specific capacitance with increasing active mass [70]. However, high active mass loading is necessary for practical applications. The approach developed in this investigation allowed for improved utilization of capacitive properties of Fe3O4. Despite the high active mass of 40 mg cm −2 , the capacitance of 114.8 F g −1 (4.59 F cm −2 ) was obtained from the CV data, which is higher than the reported capacitance [52] of 75 F g −1 at active mass of 8.8 mg cm −2 .

DFT Modeling of Adsorption on Fe3O4 (001) Surface
Next, we performed DFT modelling of adsorption of molecules studied experimentally at the surface of Fe3O4. For this purpose, we established the most stable surface of Fe3O4 and its reconstruction. Relaxation of a bulk cubic Fe3O4 unit cell was conducted. The corresponding lattice parameter, bond lengths, and magnetic moments were in good agreement with values obtained in a prior DFT+U Fe3O4 study [71] and the experimental values found in literature. Table 1 summarizes these results, where differences between the two DFT studies may be attributed to the selection of a slightly different value for Ueff, Fe 3 O 4 is a promising negative electrode material for the development of asymmetric supercapacitors [52] with enlarged voltage window for operation in mild electrolytes, such as K 2 SO 4 or Na 2 SO 4 . However, the lower specific capacitance of the negative electrodes [52], compared to the capacitance of positive electrodes, limits the development of such devices. Another difficulty is related to significant decrease in specific capacitance with increasing active mass [70]. However, high active mass loading is necessary for practical applications.
The approach developed in this investigation allowed for improved utilization of capacitive properties of Fe 3 O 4 . Despite the high active mass of 40 mg cm −2 , the capacitance of 114.8 F g −1 (4.59 F cm −2 ) was obtained from the CV data, which is higher than the reported capacitance [52] of 75 F g −1 at active mass of 8.8 mg cm −2 .

DFT Modeling of Adsorption on Fe 3 O 4 (001) Surface
Next, we performed DFT modelling of adsorption of molecules studied experimentally at the surface of Fe 3 O 4 . For this purpose, we established the most stable surface of Fe 3 O 4 and its reconstruction. Relaxation of a bulk cubic Fe 3 O 4 unit cell was conducted. The corresponding lattice parameter, bond lengths, and magnetic moments were in good agreement with values obtained in a prior DFT+U Fe 3 O 4 study [71] and the experimental values found in literature. Table 1 summarizes these results, where differences between the two DFT studies may be attributed to the selection of a slightly different value for U eff , as well as the Van der Waals interaction, accounted for in our work but not in the prior study [71]. For the purposes of this study, adsorption investigations were performed exclusively on the (001) surface of the Fe 3 O 4 crystal. The abundance of the (001) surface on Fe 3 O 4 nanoparticles has been confirmed experimentally [74]. The Fe 3 O 4 nanoparticle surface was modelled as a slab of √ 2 × √ 2 R45 • supercell of bulk cubic magnetite. The exact structure of (001) surface of magnetite is a delicate topic [75,76], sensitive to the chemical potential of species involved. We selected a stochiometric surface terminated with tetrahedrally coordinated Fe atoms (Fe tet ), which is in line with previous computational studies [77][78][79] that identified this (001) surface and its termination as the most energetically favourable.
Slabs of varying thicknesses (9, 15, and 23 layers) separated from its periodic image by a 25 Å vacuum layer were constructed to test for convergence of the surface energy ( Figure 7). The resulting slabs were constrained by fixing the positions of the middle three layers of atoms at their relaxed bulk positions (the region enclosed by dashed lines in Figure 7) to simulate the bulk phase of Fe 3 O 4 , while the remaining outer layers of atoms were allowed to relax during optimization to simulate the surface reconstruction. The surface energy was calculated as where E slab is the total energy of the constructed surface slab, N is the equivalent number of Fe 3 O 4 formula units in the slab, E bulk is the total energy per formula unit of the bulk Fe 3 O 4 , and A is the area of the surface slab. The calculated surface energies were compared to the surface energy found in a prior DFT study of Fe 3 O 4 (001) surface [79]. These results can be seen in Table 2, where surface energy and Fe-O bond length are compared. The bond length converges quickly as a function of the number of layers, while the surface energy converges very slowly. Due to computational constraints, we selected the 9-layer surface to investigate the physicochemical nature of the catechol adhesion. Figure 7 shows the final 9-layer force relaxed surface of which DHBA and BA adsorption will be modelled using DFT. One can see that the tetrahedrally coordinated Fe atoms on the top and bottom of the surface (shown by arrows) moved towards the bulk during relaxation, which is in agreement with literature [77]. R PEER REVIEW 9 of 18 modelled using DFT. One can see that the tetrahedrally coordinated Fe atoms on the top and bottom of the surface (shown by arrows) moved towards the bulk during relaxation, which is in agreement with literature [77].

DFT Analysis of DHBA and BA Adsorption at the (001) Surface
Previous adsorption studies [79][80][81] of single adatoms and water molecules on the surface of Fe3O4 have found favourable adsorption sites and have reported adsorption energies for the various species. Ni adatoms were found to adsorb favourably onto the (001) surface and reported an adsorption enthalpy of −3.21 eV [79]. This study also reported the incorporation of Ni and Ti into the subsurface by replacing Fe tetrahedral and octahedral sites, respectively. Incorporation energies of −3.39 eV for the Ni and −8.29 eV for the Ti were reported [79]. Gargallo-Caballero et al. [76] reported co-adatom adsorption as well as incorporation at octahedral/tetrahedral sites with energies of the order of -5.5 eV for the magnetite (001) surface. Additionally, adsorption enthalpies of −0.76 eV for H2O molecules adsorbing to surface Fetet atoms and −0.85 eV for single H atoms adsorbing to surface O atoms were reported [80,81]. For adsorption of DHBA onto the Fe3O4 surface, bonding to both tetrahedrally and octahedrally coordinated Fe atoms were considered, in a manner that maintained coordination seen in the bulk crystal.
Adsorption of DHBA is modelled on the (001) surface in four different configurations, seen in Figure 8. Two H atoms are cleaved from phenolic OH groups, in the case of Figure 8A,B,D, and are accommodated on surface O atoms. In the cases where the carboxyl group is responsible for adsorption ( Figures 8C and 9A), only one H atom is cleaved and accommodated by Fe3O4 forming a surface OH group. Adsorption strength of molecules to the surface of Fe3O4 is evaluated by the calculation of an adsorption enthalpy (Hads), which represents the difference between the total energy (Etot) of the adsorbed and nonadsorbed states

DFT Analysis of DHBA and BA Adsorption at the (001) Surface
Previous adsorption studies [79][80][81] of single adatoms and water molecules on the surface of Fe 3 O 4 have found favourable adsorption sites and have reported adsorption energies for the various species. Ni adatoms were found to adsorb favourably onto the (001) surface and reported an adsorption enthalpy of −3.21 eV [79]. This study also reported the incorporation of Ni and Ti into the subsurface by replacing Fe tetrahedral and octahedral sites, respectively. Incorporation energies of −3.39 eV for the Ni and −8.29 eV for the Ti were reported [79]. Gargallo-Caballero et al. [76] reported co-adatom adsorption as well as incorporation at octahedral/tetrahedral sites with energies of the order of -5.5 eV for the magnetite (001) surface. Additionally, adsorption enthalpies of −0.76 eV for H 2 O molecules adsorbing to surface Fe tet atoms and −0.85 eV for single H atoms adsorbing to surface O atoms were reported [80,81]. For adsorption of DHBA onto the Fe 3 O 4 surface, bonding to both tetrahedrally and octahedrally coordinated Fe atoms were considered, in a manner that maintained coordination seen in the bulk crystal.
Adsorption of DHBA is modelled on the (001) surface in four different configurations, seen in Figure 8. Two H atoms are cleaved from phenolic OH groups, in the case of Figure 8A,B,D, and are accommodated on surface O atoms. In the cases where the carboxyl group is responsible for adsorption ( Figures 8C and 9A), only one H atom is cleaved and accommodated by Fe 3 O 4 forming a surface OH group. Adsorption strength of molecules to the surface of Fe 3 O 4 is evaluated by the calculation of an adsorption enthalpy (H ads ), which represents the difference between the total energy (E tot ) of the adsorbed and nonadsorbed states (2) tion is the least favourable. These trends can be explained by examining the resulting bond angles. In the case of the chelating bidentate adsorption of DHBA via phenolic OH groups ( Figure 8A), the Fetet atom is pulled away from the surface. This causes the bond angle between the surface Fetet atom and the surface O atom to change from 157.26° to 103.79°, seen in Figure 9A, returning the bond angle to a value much closer to what occurs in the bulk crystal (109.47°). This configuration also forms a bond angle with O atoms in the DHBA molecule of 94.64° ( Figure 9B).  In the case where DHBA is adsorbed via carboxyl group, we again see the bond angles changing, this time to 115.45° between the surface O and Fetet and 69.40° between the O atoms in the carboxyl group and Fetet, shown in Figure 9C,D, respectively. This less drastic change in the bond angle towards what is seen in the bulk crystal may contribute The nonadsorbed state is modelled as the surface slab with the respective molecule in a fixed, nonadsorbed state 10 Å above the surface in the vacuum layer. The calculated adsorption enthalpies are shown within Figure 8. The DFT study reveals that the three adsorbed configurations are energetically favourable, while the Fe tet -detached configuration ( Figure 8D) is not. Figure 8A,C,D include adsorption of DHBA to a surface Fe tet atom where Figure 8B captures adsorption of DHBA to two surface octahedrally coordinated Fe atoms (Fe oct ). The chelating bidentate configuration is the most energetically favourable, closely followed by the bridging bidentate; the carboxyl bidentate configuration is the least favourable. These trends can be explained by examining the resulting bond angles. In the case of the chelating bidentate adsorption of DHBA via phenolic OH groups ( Figure 8A), the Fe tet atom is pulled away from the surface. This causes the bond angle between the surface Fe tet atom and the surface O atom to change from 157.26 • to 103.79 • , seen in Figure 9A, returning the bond angle to a value much closer to what occurs in the bulk crystal (109.47 • ). This configuration also forms a bond angle with O atoms in the DHBA molecule of 94.64 • ( Figure 9B).
In the case where DHBA is adsorbed via carboxyl group, we again see the bond angles changing, this time to 115.45 • between the surface O and Fe tet and 69.40 • between the O atoms in the carboxyl group and Fe tet , shown in Figure 9C,D, respectively. This less drastic change in the bond angle towards what is seen in the bulk crystal may contribute to the less favourable adsorption energy seen when adsorption occurs via the carboxyl group. Another consideration is that less energy may be required during the cleaving/adsorption process of the H atom in one of the phenolic OH groups ( Figure 8A) compared to breaking the double bonded O in the carboxyl group ( Figure 8C) to bond to the Fe tet atom. We see similar changes in bond angle for adsorption via carboxyl group for BA to that of adsorption via carboxyl group for DHBA, this time changing to 114.33 • in the case of Figure 9C and to 65.12 • in the case of Figure 9D. This further explains the results seen experimentally, describing the role of phenolic OH groups for adsorption onto the Fe 3 O 4 surface. In all configurations, the DHBA adsorption restores the coordination of the metal ions involved in bonding.
Two different configurations of BA on the surface of Fe 3 O 4 can be seen in Figure 10. In the chelating bidentate configuration ( Figure 10A) adsorption energy is similar to that of the chelating bidentate adsorption via carboxyl group seen with DHBA ( Figure 8C). In both cases where the carboxyl group is responsible for adsorption, the adsorption energy is much less than that of the chelating bidentate adsorption via phenolic OH groups. Figure 10B shows the optimized position of BA when it is positioned above the surface so that the H atoms point toward the surface. We can see that in this case adsorption is very weak. To explain the clear difference seen in the experimental results, where dispersion and capacitive properties are greatly improved with the use of DHBA, and the use of BA compares similarly to no dispersing agent at all; it is the electrostatic repulsion that must differ and not adsorption. For DHBA, adsorption utilizing the two adjacent OH groups results in the most energetically favourable adsorption configuration. Then, the carboxyl group is available to facilitate the electrostatic repulsion needed for good dispersion. However, in the case of BA, where the carboxyl functional group is involved in adsorption to the surface, there are no functional groups to provide the charge essential for electrostatic repulsion.
During the modelling of adsorption of both DHBA and BA, the presence of a solvent was neglected. Since the synthesis of Fe 3 O 4 was conducted in a water medium, we would expect adsorption energies to reduce by approximately a factor of two [82] without changing trends. Thus, the presence of a solvent in our calculations would not affect the main conclusions, since we are interested in the trend of adsorption energy across different configurations and molecules, and not in the magnitude of adsorption energy seen. static repulsion.
During the modelling of adsorption of both DHBA and BA, the presence of a solvent was neglected. Since the synthesis of Fe3O4 was conducted in a water medium, we would expect adsorption energies to reduce by approximately a factor of two [82] without changing trends. Thus, the presence of a solvent in our calculations would not affect the main conclusions, since we are interested in the trend of adsorption energy across different configurations and molecules, and not in the magnitude of adsorption energy seen.
Synthesis of Fe3O4 was performed by a chemical precipitation method [41,53] using aqueous solutions of iron (II) chloride and iron (III) chloride. The molar ratio of FeCl2 to FeCl3 in the solutions was 1:2. A similar procedure was carried with DHBA to analyse the influence of DHBA on the electrode performance, as well as with BA to gain insight into the dispersion and adsorption mechanisms involved. For the synthesis of Fe3O4 in the presence of MWCNT, a 3 g L −1 MWCNT suspension, containing 1.5 g L −1 DHBA or BA, was initially prepared. Preparation included ultrasonication of the MWCNT suspensions using a high energy Cole-Parmer (VCX 500, Cole-Parmer, Quebec City, QC, Canada) ultrasonic processor. Iron (II) chloride tetrahydrate and iron (III) chloride hexahydrate were added to the suspension, allowing for an Fe3O4:MWCNT mass ratio of 4:1. The pH of the solutions was adjusted to pH = 9 by ammonium hydroxide. Chemical precipitation was performed at 50 °C at continuous stirring. Obtained suspensions were filtrated, washed to remove non-adsorbed dispersant and reaction by-products, and dried overnight in an oven at 60 °C.
Electrodes were prepared by impregnation of Ni foam current collectors with slurries, containing Fe3O4, MWCNT and PVB binder. The mass ratio of Fe3O4:MWCNT:PVB was 80:20:3. The mass of the impregnated material after drying was 40 mg cm −2 . The impregnated Ni foams were pressed to 30% of their original thickness in order to improve electrical contact of the impregnated material and current collector.
A Bruker Vertex 70 spectrometer (Bruker, Milton, ON, Canada) was used for the Fourier Transform Infrared Spectroscopy (FTIR) experiments. Zeta potential measurements were performed using a dynamic light scattering instrument (DelsaMax Pro: Beckman
Synthesis of Fe 3 O 4 was performed by a chemical precipitation method [41,53] using aqueous solutions of iron (II) chloride and iron (III) chloride. The molar ratio of FeCl 2 to FeCl 3 in the solutions was 1:2. A similar procedure was carried with DHBA to analyse the influence of DHBA on the electrode performance, as well as with BA to gain insight into the dispersion and adsorption mechanisms involved. For the synthesis of Fe 3 O 4 in the presence of MWCNT, a 3 g L −1 MWCNT suspension, containing 1.5 g L −1 DHBA or BA, was initially prepared. Preparation included ultrasonication of the MWCNT suspensions using a high energy Cole-Parmer (VCX 500, Cole-Parmer, Quebec City, QC, Canada) ultrasonic processor. Iron (II) chloride tetrahydrate and iron (III) chloride hexahydrate were added to the suspension, allowing for an Fe 3 O 4 :MWCNT mass ratio of 4:1. The pH of the solutions was adjusted to pH = 9 by ammonium hydroxide. Chemical precipitation was performed at 50 • C at continuous stirring. Obtained suspensions were filtrated, washed to remove non-adsorbed dispersant and reaction by-products, and dried overnight in an oven at 60 • C.
Electrodes were prepared by impregnation of Ni foam current collectors with slurries, containing Fe 3 O 4 , MWCNT and PVB binder. The mass ratio of Fe 3 O 4 :MWCNT:PVB was 80:20:3. The mass of the impregnated material after drying was 40 mg cm −2 . The impregnated Ni foams were pressed to 30% of their original thickness in order to improve electrical contact of the impregnated material and current collector.
A Bruker Vertex 70 spectrometer (Bruker, Milton, ON, Canada) was used for the Fourier Transform Infrared Spectroscopy (FTIR) experiments. Zeta potential measurements were performed using a dynamic light scattering instrument (DelsaMax Pro: Beckman Coulter, Brea, CA, USA). Electrochemical studies were performed in aqueous 0.5 M Na 2 SO 4 electrolyte using PARSTAT 2273 potentiostat (AMETEK, Berwyn, PA, USA) for cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS). Galvanostatic charge-discharge (GCD) investigations were performed using Biologic VMP 300 potentiostat (BioLogic, Seyssinet-Pariset, France). Testing was performed using a 3-electrode electrochemical cell containing a working (impregnated Ni foam), counter-electrode (Pt mesh, MiliporeSigma, Oakville, ON, Canada), and a reference electrode(SCE, saturated calomel electrode, AMETEK, Berwyn, PA, USA). Mass and area normalized capacitances were calculated from the corresponding CV and GCD data, as described by previous studies [83,84]. The capacitances calculated from the CV and GCD data represented integral capacitances measured in a potential window of −0.8-0 V versus SCE. The capacitances calculated from the EIS data represented differential capacitances measured at an open circuit potential at voltage amplitude of 5 mV. CV results were obtained at 2, 5, 10, 20, 50, and 100 mV s −1 scan rates with EIS measurements performed afterwards. GCD results were obtained at 3, 5, 7, 10, 20, 30, and 40 mA cm −2 current densities.

Computational
The first-principles electronic structure calculations were performed in the framework of DFT [57] using Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation [85] for the exchange correlation functional, augmented by the DFT-D3 correction with Becke-Johnson damping [86,87] to capture van der Waals interactions. The Vienna ab initio simulation program (VASP) (version 5.4.4, University of Vienna, Vienna, Austria) [88][89][90] and projector augmented-wave potentials [91] were used, where the p semi-core states were treated as valence states for Fe potentials in all calculations. Standard potentials were used for all other elements. The cut-off energy for a plane wave expansion of 500 eV was used for the relaxation of the bulk and surface structures to achieve an accurate stress tensor, and the cut-off energy of 400 eV was used in the case of adsorption calculations. During bulk relaxation, atomic position, cell shape, and cell volume were relaxed, whereas only atomic positions were relaxed during the surface and adsorption calculations. We included on-site Coulomb interaction to treat the highly correlated Fe 3d-electrons in the framework of Dudarev et al. [92] using an effective Hubbard energy of U = 3.7 eV [79]. Collinear spin-polarized calculations were performed for all structures. Magnetic moments were initialized with opposing spin orientations of magnitude 4.0 µ B for tetrahedrally and octahedrally coordinated Fe atoms [93]. Full self-consistent structural optimization was performed for the bulk, while only forces were relaxed for surface models of Fe 3 O 4 with additional constrains (see Section 2.2 for more details). The structure was considered as optimized when the magnitude of Hellmann-Feynman forces acting on atoms dropped below 50 meV/Å and components of the stress tensor did not exceed 1 kbar. The groundstate energy was calculated using first order Methfessel-Paxton smearing with a width of 0.02 eV. A blocked-Davidson algorithm with high precision is used during the relaxation of the bulk and surface Fe 3 O 4 structures. The Brillouin zone was sampled with a Γ-centered k-mesh generated automatically with a linear density of 15 divisions per 1 Å −1 of the reciprocal space. Adsorption calculations utilize a preconditioned conjugate gradient with normal precision and automatic Γ-centered k-mesh with 30 divisions per 1 Å −1 . Construction of the Fe 3 O 4 (001) surfaces is discussed in the results and discussions section. All structure files and VASP input files used in this work can be found in the Supporting Information section. Structure files can be visualized in VESTA [94].

Conclusions
We have investigated the role DHBA has on the electrochemical performance of composite Fe 3 O 4 -MWCNT anode material for use in supercapacitors. DHBA's exceptional adsorption and dispersing properties, utilized during synthesis of Fe 3 O 4 , contribute to the significant increase in capacitance at low values of impedance. With a mass loading of 40 mg cm −2 , a capacitance of 4.59 F cm −2 was obtained from CV data at a scan rate of 2 mV s −1 and 4.72 F cm −2 from GCD data at a current density of 3 mA cm −2 . DFT surface and adsorption study of DHBA and BA on the Fe 3 O 4 (001) surface sheds light on the adsorption and dispersion mechanisms seen experimentally. The largest magnitudes of adsorption enthalpies, −1.8 and −1.7 eV, are seen for DHBA chelating bidentate and DHBA bridging bidentate configurations, respectively. Lower adsorption enthalpies were calculated for adsorption of DHBA and BA via bidentate bonding of the carboxyl group, where we see comparable adsorption enthalpies of −1.2 and −1.3 eV, respectively. One can therefore conclude that, due to favourable and comparable adsorption enthalpies of DHBA and BA utilizing a carboxyl functional group, the difference in dispersion and electrochemical performance seen when comparing these two molecules as dispersing agents is dependent on the means of electrostatic repulsion. We see that the more favourable adsorption configurations for DHBA are those that rely on the phenolic OH groups for adsorption, allowing the carboxyl group to facilitate electrostatic repulsion. The Fe 3 O 4 -MWCNT anode obtained with the use of DHBA as a dispersing agent, showing improved electrochemical properties, is favourable in the development of advanced energy storage devices.