Sustainable Synthesis of CoFe₂O₄/Fe₂O₃ Catalyst for Hydrogen Generation from Sodium Borohydride Hydrolysis

Abstract

Hydrogen has been explored as a greener alternative for greenhouse gas emissions reduction. Sodium borohydride (NaBH₄) is a favorable hydrogen carrier due to its high hydrogen content, safe handling, and rapid hydrogen release. This work presents a novel synthesis of the catalyst CoFe₂O₄/Fe₂O₃ using nanocellulose fibers (TCNF) as reactive templates for metal adsorption and subsequent calcination. The resulting material was tested for H₂ production from basic NaBH₄ aqueous solutions (10–55 °C). The catalyst’s composition is 74.8 wt% CoFe₂O₄, 25 wt% Fe₂O₃, and 0.2 wt% Fe₂(SO₄)₃ with agglomerated spheroidal particles (15–20 nm) and homogeneous Fe and Co distribution. The catalyst produced 1785 mL of H₂ in 15 min at 25 °C (50 mg catalyst, 4.0% NaBH₄, and 2.5 wt% NaOH), close to the stoichiometric maximum (2086 mL). The maximum H₂ generation rate (HGR) reached 3.55 L min⁻¹ g_cat⁻¹ at 40 °C. Activation energies were determined using empirical (38.4 ± 5.3 kJ mol⁻¹) and Langmuir–Hinshelwood (L–H) models (42.2 ± 5.8 kJ mol⁻¹), consistent with values for other Co-ferrite catalysts. Kinetic data fitted better to the L–H model, suggesting that boron complex adsorption precedes H₂ evolution.

1. Introduction

The pursuit of greener energy resources has intensified due to the direct impact of greenhouse gas emissions on climate change mitigation [1,2,3]. Solar and wind energy hold visible potential but also face seasonality impacting consistent energy (electricity) generation [4]. Hydrogen (H₂) emerges as a promising renewable alternative, offering diverse synthesis pathways and producing only water upon oxidation. However, the storage of H₂ in either gaseous or liquid form presents significant technological challenges, and borohydrides like NaBH₄ are a potential alternative in alkali pH [5,6,7,8,9,10,11,12,13].

H₂ generation can be achieved through various environmentally friendly processes, including water splitting and hydrocarbon reforming, with the ideal scenario of H₂ oxidation yielding only water as a byproduct [5,6]. Nevertheless, the storage of H₂ is complicated by the difficulty in preventing diffusion when confined as a gas or liquid under pressure. Strategies to address this include obtaining H₂ gas through hydrocarbon separation via heating, natural gas, and the hydrolysis of metal–boron hydrides [7,8,9]. Sodium borohydride (NaBH₄) stands out as a stable source and high storage content of H₂ in basic aqueous solutions. Although H₂ evolution occurs spontaneously in NaBH₄ aqueous solutions at neutral pH, the reaction kinetics are slow at room temperature [8,10]. To inhibit H₂ formation, a strong base (e.g., NaOH) for high alkali pH that stabilizes the borohydride through the formation of boron complexes with available OH⁻ anions [11]. The controlled release of H₂ at a satisfactory rate is then achieved by introducing a suitable catalyst [12]. The global process of NaBH₄ hydrolysis in the presence of an appropriate oxide catalyst is represented by Equation (1) [13], with hydrogen evolution expected to proceed irreversibly as long as boron hydride is present, and approximately half of the hydrogen originating from water splitting [8], indicating a non-elementary reaction mechanism [8,10,12,14,15,16,17,18,19,20,21].

Catalytic hydrolysis of NaBH₄ in alkali aqueous solutions can be accomplished through homogeneous or heterogeneous catalysis. Heterogeneous catalysis offers a solution to this separation problem, as the reaction medium and solid catalyst exist in different physical states, especially when magnetic catalyst crystals like cobalt ferrite spinel (CoFe₂O₄) containing materials are used, facilitating easier separation and low-cost production and comparable performance with noble metal-based materials (which often involve considerable costs, although they have high catalytic activity) [10,14,15,16,17,18,19,20,21]. As a result, non-metal-based catalysts, such as alloys or oxides, are being explored as feasible (technically and economically) alternatives [12].

The literature has reported various cobalt-based catalysts for H₂ generation from sodium borohydride. Liu and Li synthesized Co-B nanoparticles, achieving a catalyst performance of 26 L H₂ min⁻¹ g⁻¹ [21], while Balčiūnaitė et al. [22] synthesized different metallic cobalt-containing catalysts, reporting an H₂ generation rate of 53.5 mL min⁻¹ for a CoBMo/Cu catalyst. These results reinforce the potential associated with cobalt-containing catalysts towards NaBH₄ hydrolysis [23,24,25,26,27]. Other studies have investigated Co-B-P particles [28], zirconium oxysulfate samples containing a carbonaceous phase (ZrOSO₄/C) [29], kaolin-clay composites functionalized with CoB particles [30], cobalt–nickel bimetallic nanoparticles [31], Co₃O₄ nanorods [32], and Co-ferrite anchored in graphene structures [33], all showing promising catalytic activity. Co-containing materials, including the spinel CoFe₂O₄ [28,29,30,31,32,33,34,35,36,37,38], have demonstrated potential as catalysts for H₂ production via NaBH₄ hydrolysis, with ongoing research exploring novel synthesis procedures and kinetic data for this reaction [35,36,39].

This research aims to contribute new kinetic data on NaBH₄ hydrolysis, a promising route for future H₂ storage and generation, given the current energy transition scenario. This study presents characterization results to analyze sample morphology and chemical nature, followed by kinetic analysis of H₂ evolution. The catalyst used, composed of an aggregate of spherical nanoparticles, is presented as a potential alternative to existing Co-bearing catalysts. Notably, Co-ferrite spinel was synthesized using a novel method involving a type of nanocellulose (TCNF) as a reactive template. This specific synthesis route, employing TCNF, was chosen to achieve better control over the sample’s morphology and chemical composition through the simultaneous adsorption of Co and Fe cations on the nanofibers’ surface, which is not extensively explored in the literature, which employs a type of nanocellulose (TCNF) as a reactive template [40].

The novelty of this work lies in the development of a CoFe₂O₄/Fe₂O₃ catalyst through a nanocellulose-templated synthesis route, which distinguishes it from conventional cobalt-based catalysts reported in the literature. By employing TCNF as reactive templates for simultaneous adsorption of cobalt and iron ions, the authors achieved better control over nanoparticle morphology, homogeneous elemental distribution, and strong magnetic properties that enable straightforward recovery—advantages not commonly observed in previously reported methods. Compared to other cobalt ferrite systems, the synthesized catalyst demonstrated competitive or superior hydrogen generation rates (up to 3.55 L min⁻¹ gcat⁻¹ at 40 °C), approaching stoichiometric limits under mild conditions. Furthermore, this study provides new kinetic insights, showing that the Langmuir–Hinshelwood model better describes the reaction mechanism than empirical models, which is rarely addressed in the literature. Collectively, these improvements highlight the dual contribution of this work: advancing sustainable synthesis strategies through bio-based templating and enhancing the fundamental understanding of sodium borohydride hydrolysis kinetics with cobalt ferrite catalysts.

2. Methods

2.1. Catalyst Synthesis

TEMPO-oxidized cellulose nanofibers (TCNF) were obtained as previously described [40]. Then, 15 mL of nanocellulose aqueous suspension was added to a 100 mL aqueous solution containing Co²⁺ and Fe²⁺ of concentrations, respectively, equal to 58.21 and 55.75 g L⁻¹, using Co(NO₃)₂∙6H₂O and FeSO₄∙7H₂O as cation precursors. The oxidative action of the TEMPO catalyst enables the substitution of carbon six hydroxyl units by the COOH group, which can be viewed as interesting sites for the adsorption of the desired metallic cations.

After 30 min of contact time, the solid phase was separated through vacuum filtration, and the obtained solid, TCNF fibers with cobalt and iron cations adsorbed on it (TCNF@FeCo), was calcined in a muffle furnace at 500 °C for 2 h.

2.2. Hydrogen Generation

H₂ generation experiments were performed (Figure 1) with controlled amounts (±0.005 g) of NaOH (99% IsoFar) and NaBH₄ (97% Nuclear) dissolved in deionized water, and the homogeneous solution was added to a reaction flask at a specific temperature controlled by using a thermal bath (±2 °C). Then, the desired amount of the catalyst was loaded into the reactor and set under constant stirring. H₂ volume generated was then measured as a function of time in a graduated cylinder (±2%).

The effect of catalyst mass (12, 25, and 50 mg), NaOH concentration (1.0, 2.5, 5.0, and 10% w/v), NaBH₄ concentration (0.5, 1.0, 2.0, and 4.0% w/v), and temperature (10, 25, 40, and 55 °C) on the hydrogen production rate was studied. Kinetic modeling was performed based on the predicted NaBH₄ consumption rate. To calculate the time derivative of NaBH₄ concentration as a function of time, a second-order polynomial fit of NaBH₄ of the measured data was performed.

2.3. Kinetic Modeling

During kinetic modeling (Equation (1)) [28,29], it was assumed that water was present in excess, and, therefore, its molar amount was considered constant for the calculation of NaBH₄’s instantaneous concentration. In fact, this assumption is plausible based on the initial molar amounts of borohydride employed, which varied between 0.0025 and 0.02 mol, resulting in a maximum (stoichiometric) water consumption of 0.005 and 0.04 mol, when compared to 1.1 mol of water (20 mL) used in all tests. The obtained polynomial function was then used for the analytical first-order derivative calculation, whose negative value represents the desired reaction rate at each time ($r_a$).

$${\mathrm{N}\mathrm{a}\mathrm{B}\mathrm{H}}_{4\left(\mathrm{a}\mathrm{q}\right)}+{{2\mathrm{H}}_2\mathrm{O}}_{\left(\mathrm{l}\right)}\to {\mathrm{N}\mathrm{a}\mathrm{B}\mathrm{O}}_{2\left(\mathrm{a}\mathrm{q}\right)}+{4\mathrm{H}}_{2\left(\mathrm{g}\right)}$$

(1)

In the present article, kinetic modeling was performed through both an n-th order empirical model (Equations (2) and (3)) and also the Langmuir–Hinshelwood (L–H) model, defined by Equation (4) [39]. In Equations (2) and (3), $n$ designates the reaction order, $k$ represents the kinetic constant, and $r_a$ represents the absolute value of the NaBH₄ hydrolysis rate at a specified time. In Equation (4), $C_{a,0}$ represents the initial concentration of NaBH₄ and $K$ the adsorption equilibrium constant. In the present work, the NaBH₄ consumption rate was measured in g L⁻¹ min⁻¹ and concentration in g L⁻¹.

$${-r}_a=k{C}_a^n$$

(2)

$$ln ln \left({-r}_a\right)=n ln\left(C_a\right)+lnk$$

(3)

$$kt=C_{a,0}-C_a+{\displaystyle \frac{ln\left(\frac{C_{a,0}}{C_a}\right)}{K}}$$

(4)

Regarding the empirical model, a plot of $ln \left({-r}_a\right)$ as a function of $ln\left(C_a\right)$ enables estimation of global reaction order and kinetic constant through the linear and angular coefficients, respectively. In the case of the L–H model, a MATLAB (version 2024b) function was constructed for simultaneously estimating the values of $k$ and $K$, which was based on the use of the fminsearch built-in MATLAB function.

In both cases, the process global activation energy ($E_a$) was calculated from the Arrhenius equation, according to its linearized form (Equation (5)), through evaluation of the angular coefficient of a plot of $lnk$ as a function of $1/T$. In the present work, activation energy is given in J mol⁻¹, $R$ represents the universal gas constant (8.314 J K⁻¹ mol⁻¹), and temperature is given in Kelvin. In Equation (5), A represents the Arrhenius frequency factor, which should be constant as long as no change in the mechanism takes place in the temperature range explored, as for both $E_a$ and $n$.

$$lnk=lnA-{\displaystyle \frac{E_a}{RT}}$$

(5)

Recycling tests were carried out three times at 25 °C, with a 50 mg catalyst mass and initial concentrations (% w/v) of NaOH and NaBH₄ equal to 2.5 and 2%, respectively. After each reaction cycle, the catalyst was separated from the reaction medium via centrifugation, dried at 80 °C overnight, reweighed, and finally used for the next reaction cycle. H₂ generation rate (HGR) was calculated by dividing the measured rate of H₂ production (L min⁻¹) by the catalyst mass employed (g).

2.4. Sample Characterization

X-ray diffraction (XRD) was employed to evaluate sample composition regarding present crystalline phases. Diffraction patterns were obtained employing a Bruker D8 Discover equipped with a Ni filter, Lynxeye detector, and copper tube operating at 40 mA and 40 kV. Analyses were performed in the range between 10 and 95°, with a 0.02° step and an acquisition time of 15 s. Rietveld refinement analysis was performed using TOPAS 5.0 software from Bruker (Billerica, MA, USA). Sample morphology was first studied through scanning electron microscopy (SEM) in a JEOL JSM 7100F microscope operating at 2 kV and then through transmission electron microscopy (TEM) in a JEOL JEM 2100F microscope operating at 200 kV (JEOL, Tokyo, Japan). In order to evaluate the sample chemical composition, elemental distribution EDS maps were constructed in STEM mode.

To prepare samples for TEM analysis, isopropanol suspensions containing the synthesized oxide particles were drop-casted onto carbon-coated copper grids. Magnetization measurements were performed using a Physical Property Measurement System (PPMS) DynaCool from Quantum Design (San Diego, CA, USA). Magnetization measurements as a function of the applied magnetic field were conducted at room temperature to analyze the sample’s magnetic response, as well as coercive field, saturation magnetization, and remanence magnetization evaluation. Regarding sample preparation for magnetic characterization, powder masses determined using an analytical balance (Mettler Toledo model AB204-S, Mettler Toledo, Columbus, OH, USA) were carefully wrapped in Teflon tape and positioned in the brass sample holder of the PPMS. Finally, XPS measurements were made using a SPECS system with a Phoibos 150 hemispherical analyzer (SPECS GmbH, Berlin, Germany) and an aluminum non-monochromatic X-ray gun (1486.6 eV). A survey and spectra of Fe 2p, Co 2p, C1s, and O1s were made and analyzed, removing the Shirley background and deconvolving the peaks using Voight curves. The spectra were corrected using the carbon 1s peak position. In the Cobalt 2p region, the Iron Auger LMM was used in order to perform the desired signal deconvolution.

3. Results and Discussion

3.1. Catalyst Characterization

The XRD pattern of the synthesized oxide sample can be observed in Figure 2. It was possible to identify by Rietveld analysis (Figure A1) the presence of three crystalline phases: Co-spinel ferrite (CoFe₂O₄, COD CIF file: 1535820, 74.76 wt%), hematite (Fe₂O₃, ICSD, CIF file 82902, 24.98 wt%), and Fe(III) sulfate (Fe₂(SO₄)₃, COD CIF file: 9008258, 0.276 wt%). Our analysis demonstrated a goodness of fitting (GOF) of 1.15, supporting its quantitative level. Mean crystallite sizes were obtained by integral peak breadth-based volume calculation (LVol-IB), resulting in values of 12 nm and 48 nm for Co-ferrite and hematite, respectively. The higher affinity of iron towards interaction with the active sites (-OH or COO- groups) contained in the nanocellulose molecular structure can explain the lower cobalt concentration evidenced after Rietveld analysis (all cobalt is stabilized in the cobalt ferrite lattice).

Formation of the spinel structure can be explained by the fact that both Fe³⁺ and Co²⁺ cations, once adsorbed in the TCNF nanofibers, chemically react together with O₂ present in the calcination atmosphere, thereby forming a spinel cubic lattice. On the other hand, hematite (Fe₂O₃) crystals were also produced due to Fe excess (free iron cations, which do not interact with Co²⁺) over the TCNF fibers. Finally, SO₄²⁻ anions originally present in aqueous media, also adsorbed over TCNF nanofibers, can chemically react with part of the Fe cations present, thereby forming Fe-sulfate, which should be thermodynamically stable for temperatures in the range between 400 and 800 °C (thermodynamic simulation shown in Figure A2).

Based on data from Figure A2, Fe₂(SO₄)₃ can decompose to either Fe₂O₃ or FeSO₄, and no sulfate residue is expected after 900 °C. For the employed calcination temperature (500 °C), it is then expected that most of the Fe(III)-sulfate formed during sample thermal treatment should be present in the material after the calcination step. Therefore, the low mass fraction of Fe(III)-sulfate (0.26 wt%, Figure A1) and thermodynamic data (Figure A2) strongly suggest that the amount of sulfate anions (SO₄²⁻) adsorbed in TCNF from the initial aqueous solution should be of low magnitude compared to Fe³⁺ (Fe₂O₃) or Co²⁺ (CoFe₂O₄). The distance between the atom planes (Figure 3) was measured as 0.486 nm, which corresponds to the value of the (111) plane of CoFe₂O₄. Moreover, the SAED pattern in Figure 4D was indexed as CoFe₂O₄, and the diffraction ring of the (111) plane agrees well with the HRTEM image (Figure 3C).

These results confirm the presence of cobalt ferrite spinel, as evidenced by XRD analysis (Figure 2 and Figure A1). The histogram of the particle size distribution obtained from Figure A3 indicates that the nanoparticles have an average size of 19.0 ± 0.2 nm (Figure A3). Additionally, the elemental chemical composition of a group of ferrite nanocrystals was investigated through EDS, and corresponding elemental maps and associated STEM images are depicted in Figure 4. Data clearly reveal that iron, cobalt, and oxygen are spatially homogeneously distributed, corroborating the formation of a single spinel phase (CoFe₂O₄).

Magnetization as a function of magnetic field obtained at room temperature (Figure A4) revealed the magnetic characteristics of the studied spinel Co–ferrite-containing sample. The saturation magnetization (M_s), reaching 37.2 emu g⁻¹, represents the maximum magnetization the material can achieve under the influence of a magnetic field. This value indicates a notable sample magnetic response capacity.

On the other hand, the remanence magnetization (M_r), which achieved a value of 15.7 emu g⁻¹, represents the residual magnetization that persists after the removal of the earlier applied external magnetic field. Additionally, the coercive field (H_c) of the sample, measured as 1260 Oe, indicates the material’s resistance to demagnetization, i.e., the amount of magnetic field required to completely reverse its magnetization to zero. In addition, the observed magnetic behavior (Figure A4) indicates a lower saturation magnetization in comparison with other studies conducted earlier for Fe-Co spinel crystals [41,42]. It occurs (Figure 2 and Figure A1) due to the presence of Co-ferrite spinel (magnetic) and hematite (inferior magnetic response due to antiferromagnetic nature) [43], thereby contributing to a reduction in M_s. Additionally, TEM analysis (Figure 3) evidenced the presence of nanoparticle agglomeration in the sample. The advent of agglomeration can significantly influence magnetic properties by promoting magnetic interactions between neighboring particles, thereby leading both to higher values of M_r (residual magnetization) and H_c (coercive field) and also to a hysteresis enhancement, which, for the catalyst mean particle size (19 nm), should not be expected.

3.2. XPS Analysis

Signals associated with the catalyst sample as a whole, as well as for the binding energy regions characteristic of Fe and Co, are presented, respectively, in Figure 5, Figure 6 and Figure 7, and C and O in Figure A5 and Figure A6. The powder sample was placed in the system fixed with double-sided tape. Spectra were acquired in the Fe 2p, Co 2p, O 1s, and C 1s regions, along with a survey spectrum. The background was removed using the Shirley method, and the peak fitting was performed using Voigt curves with a Gaussian–Lorentzian ratio of 40–60%. In the iron region, we identified the presence of Fe₂O₃ (Fe³⁺) only, with the 2p₃/₂ peak located around 711 eV and the 2p₁/₂ peak at 724 eV. In the case of iron (Figure 6) and cobalt (Figure 7), the black curve indicates the model and the gray curve the background. The red curves in the case of cobalt represent the deconvoluted contributions to the model. The same is valid for carbon and oxygen spectra presented in the Appendix (Figure A5 and Figure A6). In the case of iron, the red line is coincident with the black curve, and therefore, does not appear in the figure.

The powder sample was placed in the system fixed with double-sided tape. Spectra were acquired in the Fe 2p, Co 2p, O 1s, and C 1s regions, along with a survey spectrum. The background was removed using the Shirley method, and the peak fitting was performed using Voigt curves with a Gaussian–Lorentzian ratio of 40–60%. In the Fe region, we identified the presence of Fe₂O₃ (Fe³⁺) only, with the 2p₃/₂ peak located around 711 eV and the 2p₁/₂ peak at 724 eV.

In the Co region, the 2p₃/₂ and 2p₁/₂ peaks were observed at 781.1 eV and 786.2 eV, respectively, along with two characteristic satellite peaks indicative of Co–O bonding (Co²⁺) [44]. The obtained results, considering only the transition metals of relevance, indicate that Fe and Co atomic fractions should have values, respectively, equal to 71.3% and 29.7%.

A quantitative comparison between atomic fractions determined through Rietveld analysis and XPS (

Table 1. Comparison of atomic fractions from XRD and XPS analysis.

Element	XPS Atomic Fraction (%)	XRD Atomic Fraction (%)
Fe	71.3	75.1
Co	29.7	24.9

) shows that both techniques achieved very similar results, corroborating their findings and confirming the presence of the identified transition metals in the sample.

3.3. Hydrogen Generation

After synthesizing and characterizing the material, our focus shifted to investigating its catalytic behavior, depicted in Figure 8. Figure 8 shows the obtained data for H₂ generation as a function of time under varying NaOH concentration (Figure 8a), catalyst mass (Figure 8b), NaBH₄ concentration (Figure 8c), and temperature (Figure 8d). According to Figure 8, the synthesized cobalt ferrite-containing sample clearly presents a prominent catalytic activity towards hydrogen generation from sodium borohydride aqueous solution in basic media. In order to avoid the NaBH₄ self-hydrolysis, NaOH was used as a stabilizing agent [45]. In the set of experiments conducted in the absence of the catalyst, emulating the experimental conditions of the catalyzed tests, no appreciable volume of hydrogen was generated over time. Thus, the hydrolysis in a basic medium is controlled by the action of the CoFe₂O₄/Fe₂O₃ catalyst. In NaOH absence (

Table A1. H₂ generation without catalyst addition under varying NaOH initial concentrations at 25 and 55 °C.

NaOH (wt%)	NaBH4 (wt%)	Temperature (°C)	H2 Generated in 1 h (mL)
0	2	25	123.0
1	2	25	2.0
2	2	25	0.0
4	2	25	0.0
2.5	2	55	10.0

), the amount of H₂ generated at 25 °C (123 mL) was almost ten times higher in comparison with the amount produced at 55 °C for an initial concentration of NaOH equal to 2.5 wt% (10 mL). For comparison, these volumes are much lower than those observed in the presence of a cobalt ferrite catalyst for the same conditions (around 1 L, in many cases generated in shorter times).

Figure 8a presents the kinetic behavior for different NaOH initial concentrations and shows that when NaOH concentration increases from 1.0 to 2.5 wt%, a slight increase in HGR is observed, with values, respectively, equal to 1.13 L min⁻¹ g⁻¹ and 1.89 L min⁻¹ g⁻¹. The maximum hydrogen volume obtained has been shown to be equal to 956.6 mL after only 20 min of reaction using 2.5 wt% NaOH, 2.0 wt% NaBH₄, and 50 mg of catalyst at 25 °C. This volume is only 8% lower than the one expected for the consumption of all NaBH₄ present (1040 mL) based on the global reaction stoichiometry (Equation (1)). However, higher NaOH concentrations result in an appreciable reduction of process kinetics, as can be seen by the decrease in HGR values. Hydrogen production should be associated with both BH⁻⁴ complex reduction with concomitant BH₃ formation and BH₃ hydrolysis. As pH becomes higher, H⁺ concentration reduces, making BH₃ formation slower, but, on the other hand, it should contribute to the kinetics of water regeneration through B(OH)₄⁻ oxidation to B₄O₇⁻², which is kinetically interesting from the BH₃ hydrolysis point of view [30]. Moreover, under high pH, precipitation of low-soluble byproducts could also be present, for example, sodium metaborate and sodium tetrahydroxyborate, thereby blocking access to catalytic sites. Another factor that can contribute to the HGR decrease is the reaction medium viscosity increase, which could make it difficult for the NaBH₄ molecules to diffuse through the catalytic interface [46].

The H₂ generation rate is also strongly affected by the catalyst amount (Figure 8b), as demonstrated by the considerable increase when the catalyst mass increases from 12 mg to 25 mg. A further increase to 50 mg does not change maximum H₂ volume appreciably but contributes considerably to the observed kinetics, as maximum H₂ production is reached after only 15 min for 50 mg catalyst, in comparison to 35 min for 25 mg. Such an effect could be explained by the availability of active sites, both for adsorption of reactant species and/or involved catalytic reaction steps. The relevance of adsorption for the observed hydrolysis kinetics is supported by the present results achieved with the L–H model, which considers adsorption as an important step before the chemical transformation over the catalyst surface.

Next, NaBH₄ initial concentration and possible effects on H₂ production kinetics were also addressed (Figure 8c). According to the data, there is a visible positive stimulation of reaction kinetics when NaBH₄ concentration increases in the interval from 0.5 to 2.0 wt%. The H₂ generation rate increased from 0.257 to 1.89 mL min⁻¹ g_cat⁻¹ when the NaBH₄ concentration varied in this range. As NaBH₄ is the source of BH₄⁻, it is understood that an enhancement of its content should have a positive and significant contribution for promoting overall hydrolysis kinetics [30,46]. The further increase in NaBH₄ initial concentration from 2.0 to 4.0 wt% contributed to the maximum amount of H₂ generated, due to the stoichiometry of the reaction, which for 4.0 wt% reached a value of 1785 mL, higher than the one for 2.0 wt% (956.6 mL). On the other hand, it has a negative impact on kinetics, as understood by a comparison of reaction times in order to achieve maximum H₂ production—13 min (2.0 wt%) and 20 min (4.0 wt%). This could be explained through an increase in solution viscosity, resulting in mass-transfer limitations of BH₄⁻ species to the active sites available on the catalyst surface [47].

Finally, it can be observed from Figure 8d that the increase in temperature, as expected, has a significant effect, enhancing the H₂ generation rate. In this case, the same initial conditions were employed for all tests (2.5 wt% NaOH, 2.0 wt% NaBH_4, and 50 mg of catalyst); the same maximum H₂ volume is reached, although in shorter reaction times when more thermal energy is available. At 40 °C, for example, the maximum amount of H₂ produced (about 956.6 mL) is reached in approximately 10 min. The calculated specific HGR values (H₂ generation rate per gram of catalyst) as a function of time at different temperatures are presented in Figure 9. As expected, based on Figure 8, HGR decreases as time evolves, approaching zero for all three temperatures. Such behavior could be explained, for example, by the reduction in the availability of NaBH₄ by its consumption during reaction and by the chemical reaction of the catalyst with the NaBH₄, which can act as a reducing agent, resulting in phases containing Co and B in the catalyst surface, also serving as a barrier for contact with the active sites. This possibility is corroborated by the preliminary results for the catalyst after three successive reaction cycles during reuse (Figure 10).

A comparison of activity between the synthesized cobalt ferrite spinel-containing catalyst and some important Co-based catalysts reported in the literature is given in

Table 2. Comparison of the synthesized catalyst with other reported cobalt-containing catalysts.

Catalyst	Catalyst Type	NaBH4	NaOH	Solution Volume [mL]	HGR (L min−1 gcat−1)	Temperature (°C)	Ref.
CoFe2O4/oxidized graphene	Cobalt ferrite anchored in GO nanoribbons	2.0 wt%	2.0 wt%	50	3.7	35	[34]
CoFe2O4/graphene	Cobalt ferrite anchored nitrogen and sulfur co-doped graphene architecture	0.75 M	N/S	20	8.5	25	[33]
CoFe2O4/Pd	Nanospheres modified with Pd	0.1 M	N/S	10	74.8	30	[48]
CH3COOH-kaolin-CoB	CoB catalyst on acetic acid-activated kaolin support	5.0 wt%	1.0 wt%	10	1.53	30	[30]
Co-Fe	Iron and cobalt metal complex	2.5 wt%	10 wt%	10	0.69	30	[49]
AC (Ni1/Co3/AC)	Core–shell Co-Ni bimetallic oxides on activated carbon	10 wt%	10 wt%	N/S	0.66	30	[31]
Co3O4	Hollow microspheres	10 wt%	2.0 wt%	3.0	5.34	25	[50]
M2.5U10Co3O4-400	Sheet-like structure with Co3O4 anchored	1 wt%	N/S	50	2.04	25	[51]
CNSs@Pt0.1Co0.9	Pt and Co anchored in carbon nanospheres	0.945 g	1.0 g	6.0	8.94	30	[52]
CoO/CaO	Cobalt-doped catalyst from chicken eggshell powder biowaste	1.0 wt%	1 wt%	10	0.43	30	[53]
Ni-Co-B	Nickel–cobalt–boride alloy (annealed)	0.16 g	15 wt%	5.0	2.60	28	[54]
(Pd0.5-Pt0.5)n(SG)m/Co3O4	Single-alloy Pd/Pt quantum dots over Co3O4 nanoparticles	1 wt%	N/S	100	8.30	25	[55]
CoFe2O4/Fe2O3	Cobalt ferrite nanocorals	2.0 wt%	2.5 wt%	10	1.89	25	This work

. Maximum specific hydrogen generation rates for experiments conducted with 2.0 wt% NaBH₄ and 2.5 wt% NaOH at 10, 25, 40, and 55 °C were equal to 0.64, 1.89, 3.55, and 5.20 L min⁻¹ g_cat⁻¹, respectively, suggesting a significant and even better catalytic activity when compared with some earlier reported data (see Table 1). It is worthwhile to mention that the synthesized catalyst (Figure 2 and Figure A1) is actually a physical mixture of a spinel phase (CoFe₂O₄) and hematite (Fe₂O₃). Similar catalytic tests were conducted using a pure hematite sample produced through the same methodology as described before, but with only iron cations (Fe³⁺) in solution during the adsorption step. The amount of H₂ generated (using 2.0 wt% NaBH₄, 2.5 wt% NaOH, and 50 mg catalyst), in this case, was insignificant or lower than the detectable limit of the employed experimental apparatus. In the case of Fe(III)-sulfate, there is no evidence to date that it has any catalytic activity for the process studied in this article. Therefore, in light of the above observations, the authors think that the observed catalytic activity is associated with the presence of the spinel crystals, also the catalytically active phase in the present context, and could be even enhanced if the spinel phase (CoFe₂O₄) could be produced in a state of higher purity.

Results obtained during catalyst recycling tests (Figure 10) for a first run (fresh catalyst) at 25 °C were compared with two other successive experiments conducted with the catalyst physically recovered from reaction media. It can be seen that catalytic activity progressively decreases during cycles. Figure 11 shows the XRD catalyst sample signal before and after being exposed three times to the same reaction conditions (50 mg catalyst, 2.5 wt% NaOH, and 2.0 wt% NaBH₄).

A preliminary Rietveld analysis was performed in order to achieve a first description of the possible crystalline new phases formed and also to quantify the amount of cobalt ferrite after reuse (Figure A7). Results confirmed the total absence of hematite after reuse and also suggested the presence of a sodium cobalt salt (CoNa₄O₃), together with B and Co bearing phases (BCo₃ and BCo₃O₅).

These findings strongly suggest that a chemical interaction of cobalt ferrite nanocrystals with boron aqueous complexes present in solution took place, a phenomenon that can be stimulated by the presence of Fe²⁺ or Fe³⁺ cations generated previously through hematite leaching under the imposed reaction conditions. This fact, together with the reduction of catalyst mass from one reaction cycle to the other, could explain the significant loss of catalytic activity during reuse (Figure 10). Reduction of catalyst mass can be explained by the strong magnetic character of the sample (Figure A4), mainly associated with CoFe₂O₄ particles. As a magnetic stirrer was employed, a residual catalyst amount stayed adhered to its surface during each reaction batch, also explaining why the catalyst mass was not constant between consecutive reaction cycles. Therefore, through the use of a mechanical stirrer instead of the magnetic one, catalytic activity could be even further enhanced. The strong magnetic character of the catalyst sample can be clearly depicted in Figure A4, which illustrates the significant attraction of Co-ferrite-containing particles to the externally applied magnetic field.

3.4. Kinetic Modeling

Based on kinetic data obtained at 25 °C for 50 mg catalyst, and concentrations of NaOH and NaBH₄ equal to 2.5 wt% and 2.0 wt%, respectively (Figure 8), a preliminary kinetic model for the global process was constructed. According to the stoichiometry of the global process involved (Equation (1)), the volume of H₂ at each time was used to calculate the amount of NaBH₄ present, and, after a second-order polynomial fit, the NaBH₄ consumption rate was computed, whose variation with time was correlated based on the linear model defined by Equation (3). Both the polynomial fit and subsequent linear regression resulted in a quantitative description of the experimental data (Figure 12). The reaction’s global order was equal to 0.584, and the kinetic constant was equal to 0.437, with a very good R² value of 0.999. The non-integer nature associated with the estimated reaction order should be a direct consequence of the chemical complexity associated with the hydrolysis mechanism [30].

Regarding the L–H model, both the reaction constant (k) and the adsorption constant (K) were simultaneously estimated against experimental data, according to Equation (4), and, according to Figure 13, a very good agreement with experimental data was also observed (R² = 0.997). In Figure 13, “vkt” denotes the product of the reaction constant against time ($kt$).

Employing the estimated kinetic parameters for computing NaBH₄ concentration as a function of time, with both models for the time interval considered during parameter fitting (Figure 14). Clearly, regarding the empirical model, there are some measurable differences between experimental and calculated values, with a mean relative deviation (absolute deviation of calculated and experimental values divided by the experimental value) of 19.9% for a time interval between 0 and 15 min, indicating that the proposed empirical model is not able to describe the complexity of the reaction system under study, even considering data in the region covered during the estimation procedure. On the other hand, for the L–H model, the same is not true, and a mean relative deviation of 2.1% was observed.

According to literature [30,43], indeed, the reaction mechanism should involve, besides participation of different boron aqueous complexes, for example, BH₄⁻ or B(OH)₄⁻, multiple reaction steps, some of them including direct participation of water molecules. Moreover, the reaction system can also be affected by mass transfer from species in solution to the catalytically active sites. In this context, it should be noted that the L–H model [39] assumes that both adsorption and chemical reaction contribute to the observed reaction rate and was the model that best fit the experimental data used for k and K estimation.

Regarding H₂ production conducted with 50 mg catalyst, 2.0 wt% NaBH₄, and 2.5 wt% NaOH, temperature was varied, and kinetic data for 10 °C and 55 °C were obtained and were treated in the same way as data at 25 °C. Through an Arrhenius (Figure 15), global activation energies have been determined: 42.2 ± 5.8 kJ mol⁻¹ (L–H), 38.4 ± 5.3 kJ mol⁻¹ (empirical model). In both cases, R² values very close to unity were observed, being equal to 0.998 and 0.882 for L–H and the empirical model, respectively.

It should be noted that the underestimated precision of these values is consistent with each other and also with earlier data for other cobalt catalysts already reported in the literature (see

Table 3. Global activation energy for H₂ extraction from NaBH₄ hydrolysis for different cobalt-containing catalysts.

Catalyst	Activation Energy (kJ mol −1)	Reference
CoFe2O4/oxidized graphene	31.4	[34]
CoFe2O4/Pd	63.1	[48]
CoBFe	74.0	[22]
CH3COOH-kaolin-CoB	49.4	[30]
AC (Ni1/Co3/AC)	50.0	[31]
Co3O4 hollow microspheres	42.5	[50]
CNSs@Pt0.1Co0.9	38.0	[52]
CoO/CaO	16.7	[53]
Plasma-treated Co-B-P	49.1	[28]
Ni-Co-B	62.0	[54]
CoFe2O4/Fe2O3	38.4 ± 5.3	This work (empirical model)
CoFe2O4/Fe2O3	42.2 ± 5.8	This work (L–H model)

). Values for estimated parameters and R₂ coefficients can be found in

Table A2. Estimated parameters for the empirical model at 10 °C, 25 °C, 40 °C, and 55 °C.

Temperature [°C]	Estimated Parameters for Empirical Model
10	n	0.63
	k (g·L−1 min−1)	0.0925
	R2	0.997
25	n	0.584
	k (g·L−1 min−1)	0.437
	R2	0.999
40	n	0.704
	k (g·L−1 min−1)	0.71
	R2	0.996
55	n	0.597
	k (g·L−1 min−1)	0.908
	R2	0.998

and

Table A3. Estimated parameters for the L–H model at 10 °C, 25 °C, 40 °C, and 55 °C.

Temperature [°C]	Estimated Parameters for L–H Model
10	K (L·g−1)	0.0852
	k (g L−1 min−1)	0.899
	R2	0.999
25	K (L·g−1)	1.267
	k (g·L−1 min−1)	1.771
	R2
40	K (L·g−1)	0.237
	k (g L−1 min−1)	4.553
	R2	0.999
55	K (L·g−1)	0.071
	k (g L−1 min−1)	8.917
	R2	0.999

. The low $E_a$ values calculated with the present data for both models could suggest that either diffusion should have a strong influence on kinetics, considering the experimental setup employed [54], or that the reaction path achievable through CoFe₂O₄ crystal presence is really effective in reducing the activation energy needs for H₂ molecule formation over the catalyst surface, topics that could be handled in future publications.

Therefore, our results have demonstrated a technically feasible alternative for H₂ production by catalysis in an alkali solution. Future studies should evaluate economic feasibility and scalability towards greener production of H₂ for electricity generation and to meet sustainable development goals [56].

4. Conclusions

This study synthesized and characterized a Co-ferrite catalyst for H₂ production from NaBH₄ aqueous solutions using a novel nanocellulose (TCNF) templated chemical route. Rietveld analysis of the oxide material revealed a Co-spinel phase (CoFe₂O₄-74.26 wt%), hematite (Fe₂O₃-24.98 wt%), and residual Fe-sulfate (Fe₂(SO₄)₃-0.76 wt%). TEM showed agglomerates of spheroidal particles (mean diameter: 19.6 nm), and EDS confirmed the uniform spatial distribution of Co, Fe, and O within the same phase. The synthesized sample exhibited notable magnetic behavior, with a saturation magnetization of 37.2 emu g⁻¹ and appreciable magnetic hysteresis, despite the small particle size, likely due to particle agglomeration.

The catalyst demonstrated significant H₂ generation potential from NaBH₄ in NaOH aqueous solutions. A maximum H₂ volume of 1785 mL was achieved (close to the stoichiometric maximum of 2086 mL) within 15 min, using 4.0 wt% NaBH₄, 2.5 wt% NaOH, and 50 mg catalyst. At 40 °C, a specific H₂ generation rate of 2.84 L min⁻¹ g_cat⁻¹ (with 2.5 wt% NaOH and 2.0 wt% NaBH₄) was observed, comparable to other Co-based catalysts. Activation energies were determined using empirical (38.4 ± 5.3 kJ mol⁻¹) and L–H (42.2 ± 5.8 kJ mol⁻¹) models, aligning with reported values for similar reactions and cobalt oxide catalysts.

Reuse tests conducted at 25 °C for NaOH and NaBH₄ initial concentrations of 2.5 wt% and 2.0 wt% indicated an appreciable reduction in catalytic activity of 84%, considering the first and third reaction cycles. This could be attributed to possible chemical interactions of CoFe₂O₄ crystals with borohydride ions available in solution, as suggested by Rietveld analysis of catalyst XRD data after reuse, but also to an appreciable reduction of catalyst mass from the first run (50 mg) to the third (27 mg) due to losses resulting from catalyst particles adhering to the magnetic stirrer employed. Our study fits the sustainable development goals for green H₂ production.