Reassessing the role of carbon shells in magnetite nanoparticles: a comparative adsorption study of ionic dyes

Abstract

Carbon-coated magnetite nanoparticles have been widely explored for water treatment applications; however, the fundamental role, necessity, and true performance advantages of the carbon shell relative to bare magnetite remain insufficiently understood. We present a systematic comparative study of ionic dye adsorption on bare Fe₃O₄ (b-Fe₃O₄) and carbon-coated Fe₃O₄ (c-Fe₃O₄) nanoparticles. Kraft Lignin was the carbon precursor for the carbon shell developed herein. Transmission electron microscopy (TEM) revealed that b-Fe₃O₄ NPs possessed an average diameter of 74 ± 16 nm, while the carbon shell thickness in c-Fe₃O₄ NPs measured 7.2 ± 3.0 nm. BET analysis showed a substantial decrease in specific surface area upon coating application, from 71.30 ± 0.21 m² g⁻¹ for b-Fe₃O₄ to 12.01 ± 0.06 m² g⁻¹ for c-Fe₃O₄. FTIR spectra confirmed the incorporation of aromatic carbon structures, and XRD patterns indicated identical crystallographic peaks for both samples, demonstrating preservation of the magnetite core. Adsorption experiments with anionic Congo red (CRed) and cationic methylene blue (MBlue) were conducted across a pH range of 2–10. Interestingly, b-Fe₃O₄ NPs exhibited high adsorption of cationic CRed (more than 90%) throughout the entire tested pH range, while displaying negligible adsorption of anionic MBlue. In contrast, c-Fe₃O₄ showed minimal interaction with CRed except at pH 2, but achieved strong MBlue removal, reaching 85% at pH 10. Kinetic analyses revealed rapid CRed uptake by b-Fe₃O₄, reaching equilibrium within ∼10 minutes at pH 4, whereas c-Fe₃O₄ required more than 24 hours to reach equilibrium for both dyes. Isotherm studies further confirmed distinct adsorption behaviors associated with each material system. These findings highlight that bare magnetite offers superior performance for anionic CRed, whereas carbon coating selectively improves uptake of cationic MBlue, highlighting that the benefits of carbon shells are system-dependent rather than universal.

---

Introduction

Over the past decades, substantial progress has been made in the development of surface-functionalized magnetite nanoparticles (NPs), particularly those encapsulated with carbon-based shells. These nanocomposites typically consist of an iron oxide (Fe₃O₄) core encapsulated with a carbon shell, derived from a variety of precursors, including glucose, sucrose, kraft lignin, oleic acid, and polymer-based resins among others.^1–9 Carbon-coated magnetite (c-Fe₃O₄) NPs integrate the advantageous properties of both components: they retain the inherent magnetic properties of the magnetite core, while displaying enhanced chemical stability, improved biocompatibility, and resistance to oxidation compared to bare magnetite (b-Fe₃O₄) NPs.^2–4,10 In addition, the presence of a carbon shell mitigates the strong tendency of b-Fe₃O₄ NPs to agglomerate, thereby improving their colloidal stability and effective surface area.¹¹ Owing to these properties, c-Fe₃O₄ NPs have been extensively explored as the adsorbents for a wide range of environmental contaminants, including per- and polyfluoroalkyl substances, phenolic compounds, and synthetic dyes.^12–15

While being widely favoured among researchers, the fabrication of c-Fe₃O₄ NPs requires additional synthesis steps, increased processing time, and higher material complexity. Consequently, a fundamental understanding of the role, the necessity of the carbon shell, and the extent to which it enhances adsorption performance relative to b-Fe₃O₄ NPs is critically important. Most existing studies place emphasis on the design and optimization of core–shell structures and provide the in-depth analyses of their adsorption capabilities, while the b-Fe₃O₄ NPs remain overlooked or implicitly assumed ineffective as standalone adsorbents. Yet, emerging evidence suggests that b-Fe₃O₄ NPs can play a significant role in the adsorption processes and can even enhance the performance of well-established carbon-based adsorbents.^16–18 For instance, Xu et al. demonstrated that the incorporation of b-Fe₃O₄ NPs into the granular activated carbon significantly enhanced its adsorption capacity toward perfluorooctanoic acid, highlighting the active role of b-Fe₃O₄ NPs in the overall adsorption mechanism.¹⁶ Noteworthy, the underlying mechanism governing adsorption processes by b-Fe₃O₄ NPs has yet to be thoroughly elucidated.

Comparative adsorption studies of dyes on b- and c-Fe₃O₄ magnetite NPs have been reported by several groups. However, most of these investigations considered only a single class of dye – either anionic or cationic.^19–24 In studies where both types of dye were considered, the adsorption experiments were conducted only with c-Fe₃O₄ NPs.^15,25 To the best of our knowledge, only one comprehensive comparative study has investigated both types of magnetite NPs in combination with both cationic and anionic dyes.²⁶ Nevertheless, the experiments in that work were conducted at a single pH value, thereby neglecting the influence of NPs surface charge and dye ionization. Consequently, the authors concluded that c-Fe₃O₄ do not adsorb cationic dyes. In contrast, our results indicate that the adsorption process of cationic dye on c-Fe₃O₄ is possible under specific conditions, such as at pH 2. Therefore, a systematic comprehensive study examining the adsorption kinetics and equilibrium of both b- and c-Fe₃O₄ NPs encompassing cationic and anionic dyes across a broad pH range has yet to be reported.

Addressing this knowledge gap, the present work systematically evaluates and compares the adsorption behaviour of b-Fe₃O₄ and c-Fe₃O₄ NPs across a pH range of 2–10 in the presence of two ionic dyes: anionic Congo red (CRed) and cationic Methylene blue (MBlue). The opposite charges of the dyes allow us to explore the effect of the NPs overall surface charge on the adsorption process. Furthermore, the ionic nature of dyes, together with the aromatic structures, allow simultaneous evaluation of multiple adsorption mechanisms, including electrostatic attractions and π–π interactions. Kraft lignin was selected as the carbon precursor for the shell owing to its abundant functional groups, aromatic structure, and established relevance in wastewater treatment applications.^27–29 The findings of this study also contribute to a deeper understanding of the adsorption mechanism of b-Fe₃O₄ NPs. By critically assessing the strengths and limitations associated with the carbon shell incorporation, especially with respect to dye charge and structure, this work identifies the conditions under which the application of carbon shell is advantageous and offers new insights into the rational design of magnetite-based nanomaterials for environmental remediation.

Materials and methods

Chemicals and materials

The following materials were used without further purification: iron(III) hexahydrate trichloride (FeCl₃·6H₂O, Thermosphere Scientific, 97.0–102.0%), polyethylene glycol 4000 or PEG (C_2nH_4nO_n, Thermo Fisher Scientific, 99%), sodium acetate or NaAc (CH₃COONa, Thermo Fisher Scientific, 99%), ethylene glycol (C₂H₆O₂, Thermo Fisher Scientific, 99%), kraft lignin (Sigma Aldrich), Congo red (C₃₂H₂₂N₆O₆S₂·2Na, Sigma Aldrich, High Purity), and methylene blue (C₁₆H₁₈CN₃S·xH₂O, Sigma Aldrich, ≥82%).

Synthesis of b-Fe₃O₄ NPs

b-Fe₃O₄ NPs were synthesized using solvothermal method, previously reported by Deng et al.³⁰ In short, 1.35 g of FeCl₃·6H₂O was added to 40 mL of ethylene glycol and stirred until an orange clear solution was formed. Next, 3.6 g of NaAc and 1 g of PEG were introduced, accompanied by vigorous stirring. The entire mixture was then transferred to a 50 mL Teflon-lined autoclave reactor and baked at 200 °C for 8 hours. The resulting black solution was washed twice with DI water–ethanol solution (1 : 1 ratio) and centrifuged at 4000 rpm for 20 minutes. Probe sonication at 30% amplitude was applied for 3 minutes to disperse the nanoparticles in-between centrifugations. The final wash was done with DI water only and separated via magnetic decantation. The product was freeze-dried and grinded gently to form a thin black powder.

Synthesis of c-Fe₃O₄ NPs

The surface of b-Fe₃O₄ NPs was modified by adding a kraft lignin shell using hydrothermal method. In short, 0.250 g of kraft lignin and 0.100 g of previously synthesized b-Fe₃O₄ NPs were added to 10 ml of DI water. The solution was then transferred to a 50 mL Teflon-lined autoclave reactor and baked at 180 °C for 12 hours. The resulting solution was purified with DI water repeatedly to remove unreacted carbon using magnetic decantation. Purification was repeated 8 times. Finally, the product was freeze-dried and lightly milled to form a thin brown powder.

Characterization of c-Fe₃O₄ and b-Fe₃O₄ NPs

Transmission Electron Microscope (TEM) images and elemental mapping were obtained using Talos F200X transmission electron microscope. X-ray Diffraction (XRD) spectra of powders (vacuum-dried at 64 °C) were collected using a Rigaku SmartLab X-ray diffractometer. Fourier Transform Infrared Spectroscopy (FTIR) spectra were measured on the Bruker Invenio-X FTIR spectrometer using a Ge ATR accessory and a room temperature detector. Adsorption spectra were collected using a Genesys 50 UV-visible spectrophotometer. Zeta potential was measured using Malvern zeta sizer. Multi-point BET specific surface area was determined using the Micromeritics Tristar II 3020 gas physisorption analyser (static volumetric method, with nitrogen gas). The samples were degassed under vacuum at 40 °C for 1 hour prior to analysis. Magnetic behaviour of b-Fe₃O₄ and c-Fe₃O₄ was analysed by a Quantum Design MPMS-3 SQUID magnetometer.

Adsorption experiments

For pH-dependence adsorption experiments, 20 mL of MBlue or CRed (20 mg L⁻¹) solution was prepared using 0.01 M NaCl solution adjusted to the desired pH values (2–10). Subsequently, 20 mg of the NPs were added to each solution, followed by probe sonication at 60% amplitude for 1 minute to ensure uniform dispersion. The samples were then placed on the orbit shaker at 150 rpm for 24 hours to reach equilibrium.

Kinetics adsorption experiments were conducted at the pH values corresponding to the highest removal efficiency of MBlue and CRed. b-Fe₃O₄ (20 mg) or c-Fe₃O₄ (50 mg) NPs were added to 100 mL of corresponding dye solutions (20 mg L⁻¹). At the predetermined time intervals, 1.5 mL samples were withdrawn for the UV-vis absorption measurements.

In adsorption isotherm studies, 20 mg of b-Fe₃O₄ or c-Fe₃O₄ NPs were added to 20 mL of MBlue or CRed solutions, with initial concentrations ranging from 10 to 250 mg L⁻¹. The solutions were adjusted to the corresponding optimal pH values determined from the kinetic experiments and shaken for the equilibrium times established therein. Upon reaching equilibrium, nanoparticles were removed from the solution using external magnet. The remaining dye concentration was determined by measuring the absorbance of supernatants at the wavelengths of 611 nm and 339 nm for MBlue and CRed, respectively. Concentrations were computed using calibration curves prepared beforehand at identical conditions.

Results and discussion

Characterization of b-Fe₃O₄ and c-Fe₃O₄ NPs

The size and morphologies of b-Fe₃O₄ and c-Fe₃O₄ NPs are illustrated in Fig. 1 and Fig. S1, S2. b-Fe₃O₄ NPs exhibit crystalline structure with an average diameter of approximately 74 ± 16 nm (Fig. S1). Following hydrothermal treatment with kraft lignin, an amorphous carbon layer forms around the NPs. Shell thickness was estimated to be 7.2 ± 3.0 nm. The estimate was made based on 63 independent measurements taken from several TEM images in which the boundary between magnetite core and carbon layer was clearly distinguishable. The presence of the carbon coating is also supported by the elemental mapping obtained using Energy-Dispersive X-Ray spectroscopy (EDS). The EDS images reveal that the cores of the c-Fe₃O₄ NPs are primarily composed of iron and oxygen, consistent with magnetite structure, while carbon becomes increasingly predominant toward the particle edges (Fig. 2). Collectively, these observations confirm the successful deposition of a carbon shell and demonstrate that the surfaces of b-Fe₃O₄ and c-Fe₃O₄ NPs are both morphologically and elementally distinct.

Fig. 1 TEM images of b-Fe₃O₄ (left) and c-Fe₃O₄ (right) NPs.

Fig. 2 STEM and EDS elemental mapping of b-Fe₃O₄ (left) and c-Fe₃O₄ (right) NPs.

Multi-point BET specific surface area was estimated to be 71.30 ± 0.21 m² g⁻¹ and 12.01 ± 0.06 m² g⁻¹ for b-Fe₃O₄ and c-Fe₃O₄ NPs, respectively. We hypothesize that the reduction in surface area is most likely due to the encapsulation of multiple b-Fe₃O₄ NPs within the same shell, as the value is comparable to the reported value of lignin coated nanoparticles of 27.74 m² g⁻¹ for a size of 260 nm.³¹

The magnetic properties of both types of NPs were investigated. Fig. S3 shows the magnetization curves measured at 300 K. The magnetic saturation values are 81.4 and 42.8 emu g⁻¹ for b-Fe₃O₄ and c-Fe₃O₄ NPs, respectively. The reduction in magnetization is attributed to the presence of carbon shell and is consistent with other literature works.^12,26

To elucidate the surface chemical composition, FTIR spectra were collected for kraft lignin, b-Fe₃O₄, and c-Fe₃O₄ NPs (Fig. 3). For kraft lignin, a wide absorption band at ∼3382 cm⁻¹ was attributed to aromatic and aliphatic OH groups. The absorption peaks at 2938 cm⁻¹ and 2842 cm⁻¹ were assigned to C–H stretching in methyl, methylene and methoxy groups. The absorption peak at ∼1683 cm⁻¹ was assigned to carbonyl stretching. The absorption peaks at 1589 cm⁻¹ and 1510 cm⁻¹ were assigned to aromatic ring vibrations. The adsorption peaks at 1210 cm⁻¹ and 1026 cm⁻¹ were attributed to C–O stretching. The assignment of peaks for kraft lignin was done in accordance with previously published data.³² Compared to b-Fe₃O₄, the FTIR spectrum of c-Fe₃O₄ NPs clearly reveals absorption peaks in the 1600–1500 cm⁻¹ region, which are absent in b-Fe₃O₄ NPs and correspond to C–C stretching vibrations of aromatic rings. These features closely resemble those observed in the FTIR spectrum of kraft lignin, indicating that some of the lignin-derived functional groups are retained after the shell formation. The persistence of aromatic C–C vibrations supports the successful formation of a lignin-based carbon shell and is particularly significant, as they are expected to facilitate adsorption of aromatic organic contaminants via π–π interactions. The relative intensity of these aromatic bands is reduced compared to pristine kraft lignin, consistent with partial carbonization.

Fig. 3 FTIR spectra of kraft lignin, b-Fe₃O₄ and c-Fe₃O₄ NPs.

To determine crystal structure, XRD patterns were identified for both types of NPs. The patterned peaks were identified at 18.3°, 30.1°, 35.4°, 43.1°, 53.4°, 56.9°, and 62.5°, which correspond to magnetite Fe₃O₄ (Fig. S4). The crystallographic patterns of c-Fe₃O₄ resemble closely those of b-Fe₃O₄, suggesting that the magnetite crystal structure is not affected during the formation of carbon shell.

Adsorption studies of c-Fe₃O₄ and b-Fe₃O₄ NPs

The effects of pH. The removal efficiency of both types of magnetite NPs was studied as a function of pH in the range 2 to 10. The pH media plays an important role in the removal efficiency of ionic molecules because of the ionization effects at the surface of nanoparticles.

b-Fe₃O₄ NPs exhibited large positive (>30 mV) zeta potential at low pH and negative zeta potential at high pH values (Fig. 4). This observation is typically attributed to the protonation and de-protonation of surface Fe-OH groups in response to varying pH. Magnetite NPs dispersed in aqueous media are believed to be heavily passivated with hydroxyl (–OH) groups.³³ At low pH, due to the abundance of hydrogen ions in water, the hydroxyl groups on the surface of Fe₃O₄ NPs tend to protonate, resulting in positively charged –OH₂⁺ on the surface and a positive zeta potential.^7,34 An opposite effect, called de-protonation, takes place when nanoparticles are placed in media with high pH. Hydroxyl groups in water compete with hydroxyl groups on the surface of the Fe₃O₄, with the former pulling hydrogen away and leaving a negatively charged –O⁻ layer on the surface. Additionally, c-Fe₃O₄ NPs displayed highly negative zeta potential (>−20 mV) within most of the tested pH range, except for pH 2, where the surface charge is slightly positive (Fig. 4). These observations are consistent with other published results.¹⁴ Thus, depending on the charge of the surface layer, bare and carbon-coated magnetite nanoparticles are typically expected to adsorb polar molecules via electrostatic interactions.^35,36 Surprisingly, we find that when interacting with dyes, b-Fe₃O₄ NPs displayed an outstanding adsorption even under conditions when the electrostatic repulsion was anticipated to occur.

Fig. 4 Removal efficiency of cationic and anionic dyes on b-Fe₃O₄ (left) and c-Fe₃O₄ (right) NPs and zeta potential as a function of pH.

Congo Red (CRed), a known anionic dye, is negatively charged due to the sulfonate (SO₃⁻) groups present in the molecule. As expected, there was practically zero to no adsorption of CRed on c-Fe₃O₄ NPs along the entire pH range, except for pH 2 (Fig. 4). This can be attributed to the strong electrostatic repulsion between the overall negative zeta potential on the surface of c-Fe₃O₄ NPs and the dye's sulfonate groups. The adsorption occurs at pH 2 given that the net surface charge of c-Fe₃O₄ is slightly positive, consistent with the interactions driven by electrostatic forces.

More complex findings were observed in case of b-Fe₃O₄ NPs. The latter demonstrated an exceptionally high adsorption toward anionic CRed dye within the entire 2–10 pH range, regardless of their surface charge (Fig. 4). While the adsorption of CRed at low pH can be attributed to strong electrostatic attractions between the positively charged NPs’ surface layer and the negatively charged dye's sulfonate groups, this concept fails to explain an exceptionally high removal of CRed at higher pH values, where electrostatic repulsion is anticipated to occur among alike charges. To the best of the authors’ knowledge, these findings have not been previously reported. Prior work has reported a decrease in % removal of anionic CRed with increasing pH, which is consistent with electrostatic repulsion of anionic molecules at high pH.³⁷ In our experiments, the adsorption occurs slower, but the % removal remains high across the entire pH range. The possibility of natural degradation of CRed across the tested pH range has been ruled out experimentally (Fig. S5). It is evident that another dominant mechanistic pathway could be responsible for b-Fe₃O₄ NPs to strongly adsorb the anionic dye. These findings are further supported by the interactions with cationic dye.

Methylene blue (MBlue) is a cationic dye with a positively charged dimethylamino group. c-Fe₃O₄ NPs displayed evident adsorption toward MBlue across the pH range of 2 to 10, reaching removal efficiencies of 85% at pH 10. Adsorption of MBlue on c-Fe₃O₄ have been previously attributed to π–π interactions between the aromatic rings on the surface of the c-Fe₃O₄ NPs and the MBlue molecule.^15,29 Note that the estimate for the removal efficiency at pH 2 is approximated, taking into consideration the natural tendency of MBlue to undergo degradation at pH 2 (Fig. S6). Additionally, as reflected in the data, c-Fe₃O₄ NPs exhibit higher variability in adsorption capacity, which we attribute to batch-to-batch differences in shell formation. Interestingly, b-Fe₃O₄ NPs exhibited significantly weaker adsorption of cationic MBlue, even under favorable conditions. Specifically, at high pH values, the surface of b-Fe₃O₄ NPs is notoriously negatively charged, meaning that the adsorption of MBlue is expected to occur driven by electrostatic forces among opposite charges. Yet, a very weak adsorption is observed, supporting the presence of an alternative primary mechanism governing the interactions of b-Fe₃O₄ NPs with the ionic dyes.

Adsorption kinetics

To determine the rate-limiting step and further elaborate the adsorption mechanism, kinetics experiments were performed at the pH values with optimal performances for both types of NPs. The adsorption kinetics were fitted to the most widely used pseudo-first order (PFO) and pseudo-second (PSO) order kinetic models.^38,39 PFO model suggests that the adsorption is predominantly physical in nature and occurs mainly through the diffusion. In contrast, PSO model generally indicates that the adsorption is governed by chemisorption.

PFO:

ln(q_e − q_t) = ln(q_e) − k₁t

PSO:

Here, q_e (mg g⁻¹) and q_t (mg g⁻¹) are the adsorption capacities of the nanoparticles at equilibrium and at time t (min), respectively. The parameters k₁ (min⁻¹) and k₂ (g mg⁻¹ min⁻¹) are the PFO and PSO rate constants. The experimental data was fitted using non-linear least-squares regression implemented via the curve fit function in SciPy. Kinetic models were compared using the combination of correlation coefficients (R²), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC), calculated from the non-linear least-squares residuals.⁴⁰ The models’ parameters along with corresponding R², AIC, and BIC are summarized in Table 1.

Table 1 Adsorption kinetic parameters and conditions of b-Fe₃O₄ and c-Fe₃O₄

Adsorbent	Dye	pH	Dosage (mg)	PFO				PSO				Experiment
				R^2	qe	AIC	BIC	R^2	qe	AIC	BIC	qe
b-Fe3O4	CRed	4	20	0.957	57.68	17.25	17.65	0.973	62.33	17.49	17.88	54.24
	CRed	10	20	0.955	21.68	12.85	13.24	0.977	24.85	13.36	13.24	20.65
c-Fe3O4	CRed	2	50	0.976	74.90	23.10	23.50	0.986	90.60	18.46	18.85	77.88
	MBlue	10	50	0.822	39.11	15.15	15.31	0.930	40.57	7.70	7.86	40.88

---

Given that b-Fe₃O₄ NPs demonstrated high removal efficiency with CRed at highly acidic and highly alkaline conditions, the kinetics was studied at both conditions – pH 4 and pH 10 (Fig. 5). Note that pH 4 was selected over pH 2 given almost instantaneous adsorption at pH 2, which prevented reliable data collection. Our findings demonstrate that the adsorption of anionic CRed on b-Fe₃O₄ NPs at both pH 4 and pH 10 could be well-fitted by both models, however the differences in AIC and BIC indicate that the models are indistinguishable (ΔAIC/BIC < 2).^41,42 Notably, pH and consequently zeta potential had a significant effect on the adsorption rates. The adsorption of CRed slowed down considerably with increasing pH. At pH 4, equilibrium was reached within approximately 10 min, whereas adsorption at pH 10 required more than 5 h to attain equilibrium (Fig. 5). Furthermore, the equilibrium adsorption capacity (q_e) is also lower for highly alkaline conditions (Table 1). The pronounced differences in rates suggest that the origin of the adsorption process is either unique for each pH or it is the same and is hindered with increasing pH.

Fig. 5 Left: Adsorption kinetics of CRed on b-Fe₃O₄ NPs at pH 4 and pH 10. Inset shows a close-up of the pH 4 kinetics. Right: Adsorption kinetics of CRed at pH 2 and MBlue pH 10 on c-Fe₃O₄ NPs.

In case of c-Fe₃O₄ NPs, the adsorption of both anionic CRed (pH 2) and cationic MBlue (pH 10) is decisively fitted best by the PSO model (Table 1). The application of the shell clearly slowed down the adsorption process even further, requiring more than 24 hours to reach equilibrium in both cases (Fig. 5). Note, that the number of NPs was more than doubled for the experiments with coated nanoparticles to combat their slow kinetics. This extended equilibration time may represent a limitation for the practical application of c-Fe₃O₄ NPs in real-world water treatment processes, whereas b-Fe₃O₄ NPs, with their comparatively faster kinetics, may offer a more feasible alternative.

Additionally, the experimental adsorption capacities are generally in good agreement with the adsorption capacities predicted by the kinetic models (Table 1). The only notable exception is the case of CRed adsorption on c-Fe₃O₄ NPs at pH 2, where experimental q_e aligns more closely with the value predicted by the PFO, whereas the PSO model was identified as the best-fitting model. This discrepancy may indicate that either the true equilibrium was not reached during the experiment or that the difference in goodness-of-fit between the PFO and PSO models is marginal, making it difficult to unambiguously identify the most appropriate kinetic model.

Equilibrium isotherm studies

Equilibrium isotherms were obtained under the same conditions as the kinetic experiments. The experimental data was fitted using the Langmuir and Freundlich isotherm models to describe the adsorption behaviour.^43,44 The Langmuir isotherm,
assumes monolayer adsorption on a homogeneous surface with a finite number of identical adsorption sites, whereas the Freundlich isotherm,

q_e = K_FC^1/n_e,

describes adsorption on heterogeneous surfaces and allows for multilayer formation. q_max (mg g⁻¹) and q_e (mg g⁻¹) are the maximum and equilibrium adsorption capacities. K_L, K_F and 1/n are the constants for the respective models.

At pH 4, the rapid uptake of anionic CRed by b-Fe₃O₄ NPs is best described by the Langmuir model, indicating monolayer surface coverage (Fig. 6 and Table S1). This behaviour suggests that the adsorption proceeds until the available surface sites are saturated, after which equilibrium is reached. In contrast, the adsorption data of the same process at pH 10 follows the Freundlich model, consistent with the multilayer adsorption. This observation implies that once CRed molecules initially anchor to the b-Fe₃O₄ surface, further adsorption proceeds via intermolecular interactions and dye self-assembly process. Interestingly, the adsorption of either cationic MBlue (ph 10) or anionic CRed (pH 2) is also best fitted by Freundlich model. We hypothesize that the adsorption process here begins with the π–π interactions between aromatic regions of the carbon shell and dye molecules and is then facilitated by the self-assembly of the dye molecules.

Fig. 6 Left: Equilibrium isotherms with Langmuir and Freundlich adsorption isotherm models for CRed adsorption on b-Fe₃O₄ NPs at pH 4 and pH 10. Right: Equilibrium isotherm with Langmuir and Freundlich adsorption isotherm models for CRed adsorption on c-Fe₃O₄ NPs at pH 2 and pH 10.

Adsorption mechanism

Iron oxide magnetite has a complex inverse spinel structure where Fe³⁺ ions occupy tetrahedral sites, while both Fe²⁺ and Fe³⁺ ions occupy octahedral sites. The octahedral sites are often exposed to the surface making them crucial for surface reactions. While both divalent and trivalent ions are available at the surface, it has been previously reported that Fe²⁺ tends to oxidize at the surface of magnetite to Fe³⁺.⁴⁵ We postulate that the b-Fe₃O₄ NPs adsorb ionic dyes predominantly via complexation with localized uncoordinated iron Fe³⁺ cites at the magnetite surface (Table 2). Our hypothesis is strongly supported by an exceptionally high adsorption of anionic CRed on b-Fe₃O₄, regardless of the nanoparticles’ net surface charge, excluding electrostatic interactions as the primary mechanism. Though the adsorption is dominated by the surface complexation, the electrostatic interactions have a profound effect on the adsorption rate. As an example, at higher pH the access of dye molecules to Fe³⁺ sites is hindered by the electrostatic repulsion due the negatively charged –O⁻ on the surface, thus resulting in a considerably slower adsorption rate and lower adsorption capacity.

Table 2 Summary of the proposed adsorption mechanisms

System	Kinetics and isotherm	Plausible dominant mechanism
b-Fe3O4 CRed (pH 4)	Fast, monolayer	Fe^3+-surface complexation, enhanced by electrostatic attraction
b-Fe3O4 CRed (pH 10)	Slower, multilayer	Fe^3+-surface complexation, accompanied by π–π interactions between dye molecules
c-Fe3O4 CRed (pH 2)	Very slow, multilayer	π–π interactions between shell and dye molecules
c-Fe3O4 MBlue (pH 10)	Very slow, multilayer	π–π interactions between shell and dye molecules

---

Additionally, low adsorption of MBlue on b-Fe₃O₄ NPs indicates that adsorption is not driven by nonspecific surface interactions but instead depends on the chemical affinity of CRed to surface Fe³⁺ sites. Collectively, these observations support a mechanism in which CRed binds directly to reactive iron sites on the magnetite surface.

The mechanism of complexation with iron sites is further supported by the drastic reduction in CRed uptake upon application of a carbon shell, which effectively masks or passivates the reactive Fe³⁺ sites and results in a slower uptake of CRed, which is governed most likely via π–π interactions. Noteworthy, the adsorption becomes increasingly unfavourable as the solution pH rises and the zeta potential becomes highly negative. Increasing pH enhances electrostatic repulsion as the dye approaches the negatively charged shell surface, thereby inhibiting close contact and quenching π–π stacking interactions. Collectively, these results indicate that shell-coated systems rely on weaker and slower, pH-sensitive interactions for dye adsorption, whereas bare magnetite adsorbs primarily via surface complexation with uncoordinated iron sites. Authors would like to acknowledge that the proposed adsorption mechanisms are hypothetical and cannot be decisively established without additional investigation, including adsorption–desorption studies, post-adsorption spectroscopic analysis (e.g., FTIR, XPS), and thermodynamic evaluation.⁴⁶

Conclusions

In this work, a comparative analysis of the adsorption of cationic and anionic dyes on b-Fe₃O₄ and c-Fe₃O₄ NPs was performed. Our findings reveal that standalone magnetite NPs significantly outperform carbon-coated structures, in terms of both adsorption kinetics and capacity, with anionic CRed dye. The adsorption experiments demonstrate that b-Fe₃O₄ NPs display strong affinity toward anionic CRed across a wide pH range (2–10). Notably, high adsorption occurs at pH 10, even though the net surface charge of magnetite is negative, and electrostatic repulsion of the anionic dye is expected to occur. However, at pH 10 the adsorption rate is more than an order of magnitude slower, and the adsorption capacity decreases by approximately a factor of three compared to pH 4. The slower kinetics and reduced capacity at higher pH are attributed to electrostatic repulsion between negatively charged dye molecules and the magnetite surface. In contrast, b-Fe₃O₄ NPs demonstrate poor adsorption capacity toward cationic MBlue across the studied pH range. Taken together, these observations suggest that the primary adsorption mechanism on b-Fe₃O₄ is most likely surface complexation with uncoordinated iron Fe³⁺ sites. The adsorption of cationic MBlue is substantially improved with the introduction of carbon shell. For c-Fe₃O₄ NPs, the adsorption improves with increasing pH and reaches its maximum at pH 10, which is attributed to the increasingly negative surface charge which favors electrostatic attraction of the cationic dye. In contrast, the adsorption of anionic CRed on c-Fe₃O₄ NPs is negligible, except at pH 2. Although the adsorption capacity of c-Fe₃O₄ is slightly higher than that of b-Fe₃O₄, significantly longer times are required to reach equilibrium. While the carbon shell enables adsorption of both cationic and anionic dyes on the c-Fe₃O₄ surface, the process requires more than 24 hours to reach equilibrium. We hypothesize that the application of carbon shell masks the uncoordinated Fe³⁺ sites on the magnetite surface, thereby altering the adsorption mechanism from surface complexation to π–π interactions. This transition leads to much slower pH-sensitive adsorption of ionic molecules.

Author contributions

Kajibanga conducted experiments, analysed results and wrote the manuscript. Ngah and Walden assisted in conducting and analysing experiments. Morrell supervised experiments and wrote the manuscript.

Conflicts of interest

There are no conflicts to declare.

Data availability

TEM images, FTIR spectra, BET surface area, magnetic properties, python codes for all figures are available: https://gitlab.com/MariaMorrell/rsc-nanoscale-reassessing-the-role-of-carbon-shells-in-magnetite-nanoparticles-a-comparative-adsorption-study-of-ionic-dyes.

Supplementary information (SI) is available. See DOI: https://doi.org/10.1039/d6nr00230g.

Acknowledgements

This work was partially supported by the North Dakota Department of Agriculture (grant number 25-432) and University of North Dakota seed grant. The authors would like to thank Sergey Shilov (Bruker Scientific, LLC) for the acquisition of the FTIR spectra, Scott Payne (North Dakota State University) for the acquisition of TEM images, Neil Dilley for the acquisition of magnetic properties (Birck Nanotechnology Center at Purdue University), and Rawan Makhlouf (Particle Technology Labs) for BET specific surface area measurements. The authors would also like to acknowledge the University of North Dakota's Nanofoundry for providing access to their research facilities.

References

1. H. Zhao, H. J. Cui and M. L. Fu, J. Colloid Interface Sci., 2016, 461, 20.
2. S. Xuan, L. Hao, W. Jiang, X. Gong, Y. Hu and Z. Chen, Nanotechnology, 2007, 18, 035602.
3. L. Kong, X. Lu, X. Bian, W. Zhang and C. Wang, ACS Appl. Mater. Interfaces, 2011, 3, 35.
4. Z. Wang, H. Guo, Y. Yu and N. He, J. Magn. Magn. Mater., 2006, 302, 397.
5. F. A. Petrie, J. M. Gorham, R. T. Busch, S. O. Leontsev, E. E. Ureña-Benavides and E. S. Vasquez, Int. J. Biol. Macromol., 2021, 181, 313.
6. X. B. Zhang, H. W. Tong, S. M. Liu, G. P. Yong and Y. F. Guan, J. Mater. Chem. A Mater., 2013, 1, 7488.
7. X. Liu, M. D. Kaminski, Y. Guan, H. Chen, H. Liu and A. J. Rosengart, J. Magn. Magn. Mater., 2006, 306, 248.
8. G. Magnacca, S. Allera, E. Montoneri, L. Celi and E. Prevot, ACS Sustainable Chem. Eng., 2014, 2, 1518.
9. G. Bayramoglu, A. Akbulut, G. Liman and M. Y. Arica, Chem. Eng. Res. Des., 2017, 124, 85–97.
10. X. W. Wei, G. Zhu, C. J. Xia, Y. Ye, J. Wu and P. Du, Nanotechnology, 2006, 17, 4307.
11. K. Petcharoen and A. Sirivat, Mater. Sci. Eng., B, 2012, 177, 421.
12. Y. Gong, L. Wang, J. Liu, J. Tang and D. Zhao, Sci. Total Environ., 2016, 562, 191.
13. X. Zhao, Y. Shi, Y. Cai and S. Mou, Environ. Sci. Technol., 2008, 42, 1201.
14. X. Zhao, Y. Cai, F. Wu, Y. Pan, H. Liao and B. Xu, Microchem. J., 2011, 98, 207.
15. X. Li, Y. He, H. Sui and L. He, Nanomaterials, 2018, 8, 162.
16. J. Xu, Z. Liu, D. Zhao, N. Gao and X. Fu, Sci. Total Environ., 2020, 723.
17. Z. Duan, W. Zhang, M. Lu, Z. Shao, W. Huang, J. Li, Y. Li, J. Mo, Y. Li and C. Chen, Carbon, 2020, 167, 351–363.
18. B. Kakavandi, A. J. Jafari, R. R. Kalantary, S. Nasseri, A. Ameri and A. Esrafily, Iran. J. Environ. Health Sci. Eng., 2013, 10, 19.
19. A. Jangra, J. Singh, R. Khanna, P. Kumar, S. Dua and R. Kumar, Asian J. Chem., 2021, 33(12), 3031–3038.
20. J. Fresnais, M. Yan, J. Courtois, T. Bostelmann, A. Bée and J. F. Berret, J. Colloid Interface Sci., 2013, 395, 24–30.
21. Y. S. Ho, C. T. Huang and H. W. Huang, Colloids Surf., A, 2013, 434, 256–262.
22. H. Li, H. Jin, R. Li, J. Hua, Z. Zhang and R. Li, Sci. Rep., 2024, 14, 1217.
23. V. S. Munagapati, H. Y. Wen, A. R. K. Gollakota, J. C. Wen, C. M. Shu, K. Y. A. Lin, Z. Tian, J. H. Wen, G. M. Reddy and G. V. Zyryanov, J. Mol. Liq., 2022, 345, 118255.
24. S. Chatterjee, N. Guha, S. Krishnan, A. K. Singh, P. Mathur and D. K. Rai, Sci. Rep., 2020, 10, 111.
25. Z. Zhang and J. Kong, J. Hazard. Mater., 2011, 193, 325–329.
26. O. S. Ivanova, I. S. Edelman, C. R. Lin, E. S. Svetlitsky, A. E. Sokolov, K. A. Lukyanenko, A. L. Sukhachev, N. P. Shestakov, Y. Z. Chen and A. A. Spivakov, Materials, 2023, 16, 23.
27. B. Wang, D. Sun, T. Q. Yuan, G. Song and R. C. Sun, ACS Symp. Ser., 2021, 1377, 143.
28. L. Fang, H. Wu, Y. Shi, Y. Tao and Q. Yong, Front. Bioeng. Biotechnol., 2021, 9, 691528.
29. Y. Li, M. Wu, B. Wang, Y. Wu, M. Ma and X. Zhang, ACS Sustainable Chem. Eng., 2016, 4, 5523.
30. H. Deng, X. Li, Q. Peng, X. Wang, J. Chen and Y. Li, Angew. Chem., Int. Ed., 2005, 44, 2782.
31. S. Li, X. Li, S. Li, P. Xu, Z. Liu and S. Yu, Int. J. Biol. Macromol., 2024, 259, 128971.
32. M. N. M. Ibrahim, N. Zakaria, C. S. Sipaut, O. Sulaiman and R. Hashim, Carbohydr. Polym., 2011, 86, 112–119.
33. F. Kraushofer, F. Mirabella, J. Xu, J. Pavelec, J. Balajka, M. Müllner, N. Resch, Z. Jakub, J. Hulva, M. Meier, M. Schmid, U. Diebold and G. S. Parkinson, J. Chem. Phys., 2019, 151, 154702.
34. I. J. Bruce, J. Taylor, M. Todd, M. J. Davies, E. Borioni, C. Sangregorio and T. Sen, J. Magn. Magn. Mater., 2004, 284, 145.
35. B. Gu, J. Schmitt, Z. Li, L. Chen and J. J. McCarthy, Environ. Sci. Technol., 1994, 28, 38.
36. A. M. Vindedahl, J. H. Strehlau, W. A. Arnold and R. L. Penn, Environ. Sci.: Nano, 2016, 3, 494.
37. N. T. Nguyen, V. A. Nguyen and T. H. Nguyen, J. Dispersion Sci. Technol., 2023, 44(11), 2161–2168.
38. S. Lagergren and K. Sven, Vetensk. Handl., 1898, 24(4), 1–39.
39. G. Blanchard, M. Maunaye and G. Martin, Water Res., 1984, 18(12), 1501–1507.
40. J. P. Vareda, J. Mol. Liq., 2023, 376, 121416.
41. K. P. Burnham and D. R. Anderson, Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach, Springer, 2nd edn, 2002.
42. R. E. Kass and A. E. Raftery, J. Am. Stat. Assoc., 1995, 90, 773–795.
43. I. Langmuir, J. Am. Chem. Soc., 1918, 40(9), 1361–1403.
44. H. Freundlich, Z. Phys. Chem., 1906, 57, 385–471.
45. R. L. Rebodos and P. J. Vikesland, Langmuir, 2010, 26, 16745.
46. H. N. Tran, S. J. You, A. Hosseini-Bandegharaei and H. P. Chao, Water Res., 2017, 120, 88–116.

---