Adsorptive removal of phenanthrene and naphthalene from wastewater using coal derived carbon nanoparticles with kinetic and thermodynamic evaluation

Abstract

This study presents the synthesis, comprehensive characterization, and application of coal-derived carbon nanoparticles (CNPs) for the highly efficient removal of polycyclic aromatic hydrocarbons (PAHs), specifically phenanthrene and naphthalene, from aqueous solutions. Subbituminous coals sourced from the Gombe and Kogi regions of Nigeria were transformed into CNPs via a single-step CO₂-assisted solid-state activation process conducted at two distinct temperatures: 550 °C and 650 °C. The synthesized materials underwent rigorous characterization using proximate/ultimate analysis, Scanning Electron Microscopy (SEM), Energy Dispersive X-ray Spectroscopy (EDX), Fourier Transform Infrared Spectroscopy (FTIR), Thermogravimetric Analysis Differential Thermal Analysis (TGA–DTA), and Brunauer–Emmett–Teller (BET) surface area analysis. Characterization results revealed predominantly spherical nanoparticles (10–100 nm) with high carbon purity (> 90%) and tunable oxygen functionalities. FTIR confirmed the presence of conjugated C = C domains, while TGA–DTA demonstrated superior thermal stability for 650 °C-activated samples, exhibiting less than 10% mass loss up to 800 °C, indicative of extensive carbonization. Batch adsorption experiments were systematically performed to optimize parameters such as contact time, initial concentration, adsorbent dosage, pH, and temperature for phenanthrene and naphthalene removal. The 650 °C-activated Gombe sample (GomCNp650) exhibited the highest monolayer adsorption capacities (Qmax) of 1.26 mg g⁻¹ for phenanthrene and 1.45 mg g⁻¹ for naphthalene, achieving equilibrium within 150 min and 120 min, respectively. Kinetic data fitted pseudo-second-order models (R² > 0.99), and equilibrium data were best described by the Langmuir isotherm (R² > 0.97), indicating monolayer chemisorption on uniform sites. Thermodynamic analysis revealed that phenanthrene adsorption onto 650 °C samples was spontaneous and exothermic (ΔG < 0; ΔH ≈ -8 to -10 kJ mol⁻¹), whereas naphthalene uptake required higher temperatures to become favorable due to entropy contributions. The study highlights that activation temperature critically tunes the balance between hydrophobic π–π interactions and surface functionality, influencing adsorption mechanisms. While the observed adsorption capacities are relatively low and reusability showed a significant decline over three cycles, these findings offer valuable insights into the fundamental interactions governing PAH adsorption on coal-derived carbon nanoparticles, informing future material design for enhanced performance.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-27116-4.

Introduction

Polycyclic Aromatic Hydrocarbons (PAHs) are the kind of compounds that are composed of several aromatic rings; and compounds with up to four rings are said to be of the “light” PAHs group, while those with five and more rings are referred to as “heavy” PAHs, being more thermally stable, more toxic, and more resistant to biodegradability (Lawal, 2017)¹. The property of the physical characteristics of the PAHs including a high melting point, a high boiling point, a low vapor pressure, and minimal water solubility gets more significant with the number of rings, while lipophilicity gives rise to bioaccumulation in lipid-rich tissues². The predominant source of PAH exposure is generally considered to be manufacturing processes, including smoking, drying, and cooking of food. The production of compounds such as PAH in food is influenced by several factors, including duration, fuel type, proximity to the heat source, fat drainage, and cooking methods (e.g., grilling, frying, roasting). These compounds exhibit significant toxicity and are linked to various adverse health outcomes, such as carcinogenicity, respiratory issues, metabolic disorders, mutations, and cancer affecting multiple organs^3,4. Water contamination may lead to the intake of PAHs through drinking water and cooked foods. The concentrations in drinking water are typically below 0.1 ng/L. The amount may increase when the asphalt or coal tar coating of storage tanks and water delivery pipes is utilized³. On the other hand, the oxidation and the reduction resistances found in heavy PAHs increase with molecular weight, making them particularly impersistent to change⁵. The pollution risk caused to the environment and human health because of PAHs is very high, there are PAHs (many) that exhibit the properties of being mutagenic, teratogenic, carcinogenic, and immunotoxic^6,7. Long-term exposures have been linked to long-term high blood pressure and a higher rate or incidence of hypertension in a few studies. Their production may make fruits and vegetables unsafe to eat if they are in soil or air that is polluted with PAH. A link was also found between background exposure to PAH and changes in heart rate in the general population³.

PAHs constitute a class of ubiquitous environmental contaminants formed primarily through the incomplete combustion of fossil fuels, biomass, and waste materials. Sources such as automobile exhaust, petroleum refining operations, and accidental oil spills release PAHs into the atmosphere, while agricultural residue incineration adds to their environmental burden⁸. Natural events (forest fires and volcanic activity) contribute minor PAH loads relative to anthropogenic emissions, which dominate global PAH inventories⁹. Once released, PAHs’ fate in soils and aquifers hinges on their strong hydrophobicity and affinity for organic matter, factors that prolong persistence and facilitate groundwater migration¹⁰.

Given these hazards, numerous remediation strategies have been explored. Conventional physicochemical treatments (coagulation/flocculation, activated sludge, and bioremediation) offer varying efficacy depending on matrix and conditions¹¹. Membrane technologies can achieve high removal efficiencies but at elevated cost and fouling risk¹². Advanced oxidation processes, including photocatalytic degradation and electrochemical catalysis, generate reactive species capable of PAH mineralization but often require energy-intensive conditions^13,14. Integrated approaches combining biological and chemical treatments have shown promise but remain complex to implement at larger scale⁵.

Adsorption, in comparison to other methods, is effortless, affordable, and versatile. In solid-liquid adsorption systems, the interfacial layer model is applied whereby PAH molecules are believed to adsorb onto solid surfaces through π–π stacking, hydrophobic bonding, as well as hydrogen bonds in certain situations¹⁵. Traditional adsorbents like activated carbon and biochar produced from agricultural waste possess a strong affinity for PAHs. For instance, the carbon obtained from the leaf litter of Vitis vinifera could remove 98% of phenanthrene at 1 ppm under optimized conditions¹⁶. Other, less expensive options (like rice husk activated carbon) could also achieve more than 90% removal of mixed PAHs at moderate concentrations¹⁷. The focus of recent developments has been on carbon and nanosized carbon materials. Bituminous coal activated by CO₂ results in a greater pore diameter which increases specific surface area to as high as 1200 m²/g and the ability to hold phenanthrene over 5 mg/g¹⁸. Naphthalene-absorbing carbons generated from microwave-assisted activation of coal tar pitches have a Qₘₐₓ of about 20 mg/g but require intricate processing¹⁹. Wheat-straw biochar produced by co-combustion shows rapid kinetics and equilibrium adsorption capacities of 0.6 mg/g for phenanthrene, underlining the role of surface oxygenates in adsorption mechanisms²⁰.

The existing studies on the adsorption efficiency of carbon-based materials for the removal of polycyclic aromatic hydrocarbons (PAHs) highlight the potential of these materials in environmental remediation efforts. Numerous investigations have revealed that activated carbon, graphene, and carbon nanotubes possess significant surface areas and porous configurations, thereby enhancing their ability to adsorb PAHs²¹. Carbon-based nanomaterials (such as graphene, carbon nanotubes, and fullerenes) bring distinctive advantages: high external surface areas, tunable functionalization, and enhanced π-electron systems. Graphene derivatives functionalized with oxygen or nitrogen moieties exhibit phenanthrene Qₘₐₓ values of 4–6 mg/g and rapid equilibrium times (< 60 min), driven by strong π–π stacking^22,23. However, production costs and aggregation tendencies remain challenges. In contrast, coal‐derived carbon nanoparticles synthesized via solid‐state methods strike a balance between performance and scalability. (Do et al., (2024)²⁴ reported nanoparticle frameworks with surface areas of 50–100 m²/g capable of removing > 90% of naphthalene under mld conditions, while retaining thermal stability for reuse. Thermodynamic and kinetic considerations further inform adsorbent design. Phenanthrene sorption typically follows pseudo‐second‐order kinetics (K₂ ≈ 0.3–0.5 g mg⁻¹ min⁻¹) and Langmuir isotherms indicative of monolayer coverage²⁵ Thermodynamic parameters often reveal exothermic, spontaneous adsorption with ΔH in the range − 5 to − 15 kJ mol⁻¹ and negative ΔG, supporting physisorption predominance²⁶. Naphthalene adsorption may be endothermic on certain biochar’s, requiring elevated temperatures to achieve spontaneity, highlighting the role of entropy gains from water desorption^27,28.The combination of the adverse health effects and environmental persistence of polycyclic aromatic hydrocarbons (PAHs) necessitates the development of effective and scalable removal technologies. Among various approaches, adsorption onto carbon-based materials, including activated carbon, biochar, and advanced nanostructures, remains the most versatile and widely adopted technique due to its cost effectiveness and operational simplicity. This study investigates coal derived carbon nanoparticles synthesized via a chemical solid-state activation process. It evaluates their performance in adsorbing phenanthrene and naphthalene through comprehensive isotherm, kinetic, and thermodynamic analyses fitting both linear and nonlinear models. By integrating contemporary insights from activation chemistry, surface functionalization, and adsorption modelling, this work advances current understanding of PAH remediation using ultrafine carbon particles. The approach reflects modern strategies in nanoparticle-mediated PAH removal under simulated realistic environmental conditions, offering a promising pathway toward efficient and sustainable remediation technologies.

Materials and methods

Materials

Coal samples were collected from Okaba Coal Depot in Kogi state and Mai Ganga coal deport in Gombe State Nigeria. Ferrocene 98% was purchased from TCI Chemicals, Ethanol95%, and Methanol was purchased from HMBG chemicals, Sigma chemicals Phenanthrene, TCI Naphthalene, Carbon dioxide gas, Chemiz Hydrochloric Acid and Sodium Hydroxide from Merck.

Method

The flowchart for the synthesis of carbon nanoparticles from coal is shown in Fig. 1. The coal was crushed to a particle size of 200 nm, and 8.5 g of ferrocene was measured and dissolved in 250 mL of ethanol for 24 h. Crushed coal (100 g) was measured into a beaker, and 250 mL of ethanol ferrocene was added to the sample and stirred for 30 min on a stirrer. The sample was then covered with foil paper, and small holes were made. The sample was then transferred to the oven at 60 °C until the ethanol evaporated. coal and ferrocene (ratio of 14:1) samples were then fed into the vertical furnace and were heated under the flow of CO₂ gas for various temperature and time as presented in Table 1. The resulting mixture was then brought out to cool at room temperature and the sample was eventually stored meticulously in a sample vial.

Fig. 1.

Flowchart for the synthesis of carbon nanoparticles from coal.

Table 1.

Parameters for the synthesis of different nanoparticles.

Samples	Temperature (°C)	Time (min)
KogCNp650 (°C)	650	50
KogCNp550 (°C)	550	60
GomCNp650 (°C)	650	50
GomCNp550 (°C)	550	60

Although ferrocene and ethanol proved efficient in this investigation, they may lack cost-effectiveness and scalability. Iron salts, like FeCl₃, have demonstrated effective catalysis of graphitization at a reduced cost²⁹, whereas glycerol has surfaced as a sustainable, economical, and environmentally friendly solvent alternative to ethanol in the production of nanomaterials³⁰.

Characterization of synthesized coal nanoparticles

Proximate analysis of the coal samples was carried out following ASTM D3172 using gravimetric methods. Moisture content and volatile matter were determined using a drying oven, while ash content and fixed carbon were measured using a muffle furnace. Ultimate analysis was performed using a CHNS elemental analyzer (PerkinElmer 2400 Series II) in accordance with ASTM D3176 to determine the elemental composition, including carbon, hydrogen, nitrogen, and Sulphur contents. Oxygen content was calculated by difference.

The synthesized carbon nanostructures were characterized for morphology and elemental composition using an ultra-high resolution scanning electron microscope equipped with energy dispersive X ray spectroscopy (SEM EDX, HITACHI REGULUS 8220). Functional groups were identified using potassium bromide-based Fourier transform infrared spectroscopy (FT IR). Surface area, pore size, pore volume, and nitrogen adsorption desorption isotherms were determined using Brunauer Emmett Teller analysis (BET, MICROMERITICS ASAP 2020). Thermal stability was assessed via thermogravimetric analysis (TGA). Crystalline structure was examined using X ray diffraction (XRD, BRUKER D8 ADVANCE), and optical properties were analyzed using ultraviolet visible spectroscopy (UV VIS, PerkinElmer Lambda 35).

Batch adsorption experiment of phenanthrene and naphthalene

A 100-ppm stock solution of phenanthrene (PHE) was prepared by dissolving 25 mg of PHE in 250 mL of a methanol and distilled water mixture (1:9 v/v). The stock was serially diluted to obtain concentrations of 1, 3, 5, 7, 9, and 11 ppm. A calibration curve was generated using the 1–11 ppm range, and concentrations between 1 and 5 ppm, which obeyed Beer Lambert’s law, were used for optimization studies. For each experiment, 50 mL of PHE solution (1, 3, or 5 ppm, pH 4) was placed in a conical flask, and 200 mg of GomCNP650°C adsorbent was added. The mixture was agitated in a water bath shaker at 30 °C and 120 rpm for time intervals ranging from 0 to 150 min. Absorbance readings were taken at each time point using a UV VIS spectrophotometer at wavelengths of 250 and 220 nm, respectively. The percentage removal (%R) and equilibrium adsorption capacity (Qe) were calculated using Eqs. (1) and (2), respectively.

The adsorption of phenanthrene (PHE) and naphthalene (NAP) onto GomCNP650°C was investigated by optimizing key parameters, including adsorbent dosage, contact time, solution pH, and temperature. All experiments were conducted in a water bath shaker at 30 °C and 120 rpm. To examine the effect of adsorbent dosage, 50 mL of 5 ppm PHE solution (pH 4) was treated with varying amounts of GomCNP650°C ranging from 200 to 1000 mg. The effect of contact time was assessed under similar conditions by monitoring adsorption over a time range of 0 to 150 min. The influence of pH was evaluated by adjusting the PHE solution to pH values between 2 and 10 using 0.1 M NaOH and 0.1 M HCl. A fixed adsorbent dosage of 400 mg was used with 50 mL of 5 ppm PHE solution at 30 °C. For temperature studies, 400 mg of GomCNP650°C was added to 50 mL of 5 ppm PHE solution at pH 7, and the mixture was subjected to adsorption at 30 °C, 50 °C, and 70 °C. All samples were analyzed spectrophotometrically, and the corresponding adsorption parameters were calculated using Eqs. (1) and (2).

1

2

where C_i is the initial concentration, C_e is the final concentration at equilibrium, V is the volume of adsorbate used, and M is the mass of adsorbent used.

Adsorption kinetics

The adsorption kinetics and interaction processes between the absorbent and adsorbate were analyzed utilizing adsorption kinetic models. The optimized kinetic model correlates with the adsorbent’s suitable equilibrium adsorption behavior³¹. In this study, the pseudo-first, Pseudo-Second order, and Elovich models were fitted linearly and non-linearly using the data collected from the UV-Vis machine at equilibrium conditions for all four adsorbents synthesized. In Pseudo First order, the model denotes that the concentration of the adsorbate in the solution solely dictates the rate of adsorption. The adsorption process is fundamentally governed by the diffusion of the adsorbate to the surface of the adsorbent³². Whereas for pseudo-second order, the kinetic model describes the adsorption rate, in which the step that is the rate-limiting phase is a chemical reaction that most likely involves chemisorption^31,33. The Elovich model further elucidates the pseudo-second-order kinetics based on the premise that the sorbent surface exhibits energy heterogeneity. The Elovich model has been employed for the chemisorption of liquids and gases onto the heterogeneous surfaces of adsorbents³⁴. Equations (3), (4), and (5) represent the linearized forms of the pseudo first order, pseudo second order, and Elovich kinetic models, while the non-linear forms are donated by Eqs. (6), (7), and (8), respectively.

3

4

5

6

7

8

In this context, Qt and Qe signify the adsorption capacity (mg g⁻¹) at contact time(t) and at equilibrium, respectively, whereas K₁ (min⁻¹) and K₂ (g/mg/min) indicate the first and second order equilibrium rate constants. The values of K₁ and Qe for first-order kinetics were derived from the graph of ln (Qe - Qt) over time (t). The values of K₂ and Qe for the second order were determined from the plot of t versus Qt. For the Elovich equation, α (mg/g/min) denotes the initial rate constant of adsorption and β (g/mg) reflects the adsorption capacity^31,34.

Adsorption isotherm

In this study, the Langmuir, Freundlich, and Temkin isotherm models were linearly fitted to the equilibrium adsorption data obtained for concentrations ranging from 1 ppm to 11 ppm across all four synthesized adsorbents. The linear expressions of the models are presented in Eqs. (9), (10), and (11), respectively. Model fitting was performed using regression analysis to determine the isotherm constants and assess the best-fitting model for each adsorbent based on the correlation coefficients (R² values)^34,35.

9

10

11

where q_m represents the maximal monolayer adsorption capacity (mol g-1), K_L denotes the Langmuir constant (L mol-1), and qe and Ce signify the adsorption capacity (mol g^{− 1}) and equilibrium concentration (mol L^{− 1}), respectively. The Langmuir linear plot was obtained by plotting Ce/qe against Ce. The constants related to the Freundlich isotherm model are sorption capacity (K_F) and sorption intensity (1/n). Furthermore, the exponent (1/n) indicates the favorability and capacity of the adsorbent/adsorbate system which can be achieved by Plotting ln qe against ln Ce. In the Temkin isotherm, K_T represents the equilibrium binding constant (L mol⁻¹) associated with the maximal binding energy, b pertains to the adsorption heat, R denotes the universal gas constant (8.314 J K⁻¹ mol⁻¹), and T signifies the temperature (K). and the Temkin model can be gotten by Plotting qe against ln (Ce)^33,35,36.

Adsorption thermodynamics

The enthalpy change (ΔH◦), entropy change (ΔS◦), and Gibbs free energy change (ΔG◦) were assessed to provide insights into the spontaneity of an adsorption process³¹. The thermodynamics of adsorption were established utilizing the thermodynamic equilibrium coefficients acquired at various temperatures and concentrations to validate potential adsorption mechanisms³⁵. In this study, the thermodynamics for the removal of phenanthrene and naphthalene were done at different temperatures ranging from 30˚C − 70˚C. Equations 12 and 13 show the ΔG◦, ΔH◦, and ΔS◦, respectively, and they are typically derived from the van Hoff Eq.

12

13

where; 

14

R (8.314 J K⁻¹ mol⁻¹) and T (K) represent the molar gas constant at room temperature, respectively. The distribution coefficient, Kd (L g⁻¹), indicates the capacity of an analyte to retain a solute and its mobility within the solution^31,37.

Adsorbent reusability

The reusability of the synthesized carbon nanoparticles was tested through the experiment of adsorption and desorption over 3 rounds. Following the completion of the centrifugation process, the absolute is discarded, and the adsorbent is cleaned with deionized water and absolute ethanol before being centrifuged for thirty minutes individually. Following the removal of the ethanol, the nanoparticle is next dried in an oven at 60˚C for 20 h to ensure that it is entirely dry.

Results and discussion

Coal characterization

The proximate and ultimate analyses in Table 2 show that both Kogi (Kograwcoal) and Gombe (Gomrawcoal) coals are classified within the subbituminous rank due to their moderate moisture and volatile matter contents. Kogi coal has 5% moisture with volatile matter of 26.3%, while Gombe coal has higher moisture (10%) and volatile matter (38.8%). The ash contents of Kogi (15%) and Gombe (10%) coal, along with fixed carbon values of 46.3% Kogi and 41.2% Gombe further justify their classification as subbituminous since these values abundantly exist for these coals³⁸. From ultimate analysis, the carbon contents are 64.64% for Kogi and 58.84% for Gombe, corresponding to hydrogen contents of 5.58% and 4.39%, respectively, along with very low sulfur (< 0.4%) levels. The comparatively higher volatile matter and oxygen content in Gombe coal (28.08% for Kogi, 35.73% for Gombe) implies a greater concentration of labile organic materials, suggesting the possibility of forming specific functional groups during nanoparticle fabrication. The subbituminous nature of these coals indicates that there is a relatively optimized ratio of carbon to volatile content, which is beneficial for carbon nanoparticle production where sufficient nanostructured graphitic domains are required alongside surface functionalization for PAH adsorption.

Table 2.

Proximate and ultimate analysis of coal.

Coal	MC (%)	VM (%)	AC (%)	FC (%)	C (%)	H (%)	N (%)	S (%)	O (%)
Kograwcoal	5	26.30	15	46.30	64.64	5.58	1.38	0.32	28.08
Gomrawcoal	10	38.80	10	41.20	58.84	4.39	0.95	0.08	35.73

where MC is moisture content, AC is ash content, VM is volatile matter. FC is fixed carbon, C is carbon, H is hydrogen, N is nitrogen, S is sulphur, and O is oxygen.

FTIR analysis

The FTIR spectra in Fig. 2a and b show distinctive changes between the raw coal and the synthesized carbon nanoparticles due to thermal treatment. These changes reflect modifications in the material’s chemical structure resulting from the conversion process. In the raw coal spectrum, a broad O–H stretching band occurs around 3400 cm^{− 1} due to moisture and hydroxyl groups, which undergoes significant diminishment in the nanoparticles indicating dehydration and loss of weak structural hydroxyls at high. The range of 2800 cm^{− 1} containing symmetric CH₃ and CH₂ sp³ vibrations is suggestive of residual aliphatic and the corresponding peaks also remain present. Most importantly, there is a strong peak at ~ 1600 cm^{− 1} in the nanoparticles which is due to structurally conjugated C = C stretching and carbonyl (C = O) bond which is in line with partial graphitization³⁹. The 650 °C sample is more graphitic in nature with lower polarity groups leading to a more hydrophobic surface which is more favorable for π–π stacking interactions with hydrophobic PAHs. The sample synthesized at 550 °C is less graphitic in nature and exhibits increased oxygenated and aliphatic functional groups. This rise in overall surface polarity is likely to influence adsorption kinetics and enhance selectivity towards lighter polycyclic aromatic hydrocarbons (PAHs)⁴⁰.

Fig. 2.

FTIR Spectra of synthesized carbon nanoparticles (a) Gomrawcoal and GomCNPs (b) Kograwcoal and KogCNPs.

The synthesis temperature is essential in influencing the ultimate chemical and physical characteristics of carbon nanoparticles. The sample synthesized at 650 °C demonstrates a stronger graphitic character, marked by an increased level of aromaticity and a reduced presence of polar groups⁴¹. The enhanced graphitization results in a more hydrophobic surface, which is advantageous for π–π stacking interactions with hydrophobic PAHs, and the improved hydrophobicity and expanded π-electron systems create optimal locations for the adsorption of PAHs via non-covalent interactions, rendering these materials efficient adsorbents for these chemicals⁴².

The carbon nanoparticles produced at 550 °C have reduced graphitic characteristics. The diminished degree of graphitization leads to a heightened abundance of oxygenated and aliphatic functional groups on the surface⁴³. The increased concentration of these polar functional groups enhances the overall surface polarity. The augmented surface polarity is expected to markedly affect adsorption kinetics and improve selectivity for lighter PAHs, and the polar interactions between the oxygenated groups and the PAHs, especially those with increased polarity or reduced molecular size, can enhance adsorption, providing an alternative adsorption mechanism relative to the more graphitic, hydrophobic surfaces⁴⁴.

TGA analysis

As shown in Fig. 3, Gombe and Kogi coals exhibit multi-stage decomposition patterns in the TGA–DTA thermograms, with remarkably similar thermal behavior. The mass loss observed below ~ 150 °C is attributed to moisture evaporation, which is more pronounced in Gombe coal due to its higher initial moisture content. The main devolatilization stage, occurring between approximately 250 °C and 500 °C, corresponds to a sharp decline in sample weight because of the release of volatile organic compounds, as shown in Fig. 3a and b. The DTA traces show an initial endothermic response at lower temperatures, turning exothermic around 450 °C, marking the onset of char combustion. Kogi coal shows a lower and sharper peak than Gombe coal, indicating fewer volatile components and greater coal maturity. Above 600 °C, both coals undergo further changes through residual carbon combustion and mineral ash transformation. These behaviors confirm their subbituminous nature and identify temperature zones crucial for efficient carbon nanoparticle synthesis, where volatiles are removed with minimal carbon loss. Figure 3c and d clearly demonstrate that carbon nanoparticles synthesized at 650 °C possess markedly enhanced thermal stability relative to raw coal, exhibiting clean combustion with minimal mass loss (less than 10%) up to 800 °C. The TGA curve is almost completely horizontal after the initial dehydration stage of 100 °C, which suggests that most easily decomposable oxygen-containing groups and some of the volatiles have been removed during the synthesis. Similarly, the DTA trace shows only slight fluctuations between endothermic and exothermic responses, indicating the absence of significant phase transitions or decomposition events. These observations confirm that treatment at 650 °C promotes effective carbonization and partial graphitization, resulting in a thermally stable and structurally secure carbon framework. One key advantage of this thermal resilience is its suitability for adsorption applications, which ensures that the structural integrity and surface area of the carbon nanoparticles are maintained across a wide range of operating temperatures. Figure 3e and f show that the nanoparticles synthesized at 550 °C undergo two distinct mass loss events. The first is a minor loss below 150 °C, attributed to residual moisture. The second is a more significant weight reduction between 200 °C and 450 °C, corresponding to the release of volatile components. The derivative thermogravimetry (DTA) curve reveals peaks near 300 °C and 400 °C, which correspond to the decomposition of oxygenated functional groups such as carboxyl and ether moieties. This suggests that the lower synthesis temperature of 550 °C retains a higher concentration of polar surface functionalities. These groups can enhance initial adsorption kinetics through hydrogen bonding or electrostatic interactions with PAH molecules. However, the reduced graphitic content and lower thermal stability compared to the 650 °C sample may compromise the long-term structural integrity of the material. This observation highlights a fundamental trade-off in synthesis temperature: while 550 °C favors the preservation of surface polarity, 650 °C offers improved thermal and chemical stability through enhanced graphitization^45–47.

Fig. 3.

Thermograms for raw coal and synthesized carbon nanoparticles.

XRD analysis

X-ray diffraction in Fig. 4 further clarifies the structural ordering of the nanoparticles. All samples display a broad diffraction peak centered at approximately 2θ = 23°, corresponding to the (002) plane of turbostratic carbon, and a weaker, broad (100) reflection near 43°. The breadth and low intensity of these peaks indicate a predominantly amorphous carbon structure with short-range graphitic domains. Such turbostratic disorder is characteristic of low-temperature activation products and correlates with the thermal stability trends observed in the TGAs²⁴.

Fig. 4.

XRD patterns for the carbon nanoparticles synthesized.

This peak signifies a graphitic character, yet its low intensity and broad profile further validate the mostly amorphous carbon structure with short-range graphitic ordering. The overall width and low intensity of both the (002) and (100) peaks are direct markers of this structural attribute⁴⁸. Turbostratic carbon is characterized by its graphitic layers arranged in a disorderly manner, devoid of the specific ABCABC stacking sequence present in highly crystalline graphite. This intrinsic disorder produces broad and less intense XRD peaks in contrast to the crisp, well-defined peaks of highly crystalline graphite. This structural configuration frequently arises from low-temperature activation mechanisms during material production⁴⁹.

Morphology and elemental composition

High-resolution SEM images presented in Fig. 5a and b reveal that carbon nanoparticles synthesized at 650 °C from Gombe and Kogi coals exhibit a highly uniform morphology, characterized by well-dispersed, nearly spherical particles with diameters predominantly in the 10 to 100 nm range. These particles exhibit a distinctly rough and porous surface texture, indicating extensive devolatilization and the formation of micropores and mesopores during carbonization. The elevated synthesis temperature increases thermal energy availability, which facilitates rapid nucleation and the formation of smaller domains with partial graphitic ordering. This resulting morphology, marked by high surface roughness and well-defined porosity, correlates with a greater density of active adsorption sites and enhances π-π stacking interactions, thereby improving the material’s affinity for hydrophobic PAH molecules. The absence of large agglomerates further suggests that CO₂-assisted activation effectively prevents particle coalescence, preserving the high specific surface area critical for adsorption applications⁵⁰. Compared to many carbonaceous adsorbents reported in the literature, such as biochar or activated carbons that typically exhibit broader particle size distributions, these nanoparticles offer a narrower size range combined with pronounced surface irregularities. This unique combination enhances both diffusion pathways and surface accessibility, thereby improving adsorption performance for phenanthrene and naphthalene molecules⁵¹. At 550 °C, SEM micrographs of GomCNp550 and KogCNp550 (Fig. 5c and d) reveal a markedly broader particle size distribution and irregular morphology, reflecting reduced structural uniformity relative to the samples synthesized at 650 °C. The particle diameters range from tens to several hundred nanometers, and the surfaces appear relatively smoother, suggesting incomplete volatilization and limited pore development. The lower thermal energy at 550 °C restricts the extent of carbonization and graphitic domain growth, resulting in residual aliphatic structures and a less defined microporous network⁵². These morphological features indicate a reduced number of high-energy adsorption sites and potentially slower adsorption kinetics, particularly for larger PAH molecules. However, the retention of oxygen containing functional groups, as inferred from the smoother and more heterogeneous surface textures, may enhance initial uptake through hydrogen bonding or electrostatic interactions, especially for phenanthrene, which exhibits moderate polarity compared to heavier PAHs. Therefore, nanoparticles synthesized at 550 °C reflect a tradeoff between surface functionality and pore driven adsorption capacity, consistent with previous reports where lower activation temperatures favor higher oxygen content but result in lower specific surface areas⁵³.

Fig. 5.

Micrographs for synthesized carbon nanoparticles.

EDX presented in Fig. 6a and b confirms the predominantly carbonaceous nature of the nanoparticles synthesized at 650 °C, with carbon atomic percentages of approximately 93% for GomCNp650 and 91% for KogCNp650. Oxygen is the second most abundant element, contributing about 5 to 6%, which suggests the presence of residual surface functionalities such as carboxyl, hydroxyl, and carbonyl groups that were not completely removed during the carbonization process. Silicon (≈ 1%) and aluminum (< 1%) appear as trace elements in the EDX spectra, likely originating from mineral residues not fully removed during the final acid purification step. The slightly higher oxygen content observed in GomCNp650 suggests a greater presence of oxygen containing functional groups, which increases surface polarity. These functionalities can facilitate interactions with naphthalene molecules, such as hydrogen bonding, thereby enhancing the adsorption capacity. On the other hand, the slightly more carbon purity of KogCNp650 allows larger aromatic graphene on the surface to which the PAH molecules with the same system of electrons can still stick by the so-called π–π stacking effect, according to the explanations of adsorption-mechanism theories that mostly deal with the issue on surface-conjugation in PAH capture. The EDX spectra of GomCNp550 and KogCNp550, presented in Fig. 6c and d, confirm carbon as the predominant element, accompanied by detectable levels of oxygen and trace amounts of iron. The presence of these elements, even after thermal treatment at 550 °C, indicates that the lower synthesis temperature facilitates the retention of oxygen containing functional groups and metal residues⁵⁴. While this may slightly reduce the extent of exposed carbon surfaces, these residual species can contribute positively by enhancing specific interactions with PAH molecules through Lewis’s acid base mechanisms and potentially acting as catalytic sites for the partial oxidation of adsorbed organics. The similarity in impurity profiles between the 550 °C and 650 °C samples suggests that the primary effect of temperature is on carbon structuring rather than impurity incorporation. This finding aligns with prior reports that thermal activation temperature primarily governs pore evolution and graphitic ordering, whereas elemental composition remains largely dictated by the precursor feedstock and catalyst loading^55,56.

Fig. 6.

EDX spectra for carbon nanoparticles synthesized (a) GomCNP650˚C (b) KogCNP650˚C (c) GomCNP550˚C (d) KogCNP550˚C.

BET analyses for synthesized carbon nanoparticles

Figure 7 presents Brunauer Emmett Teller (BET) measurements which notably reveal low apparent surface areas for all four nanoparticle samples. GomCNp650 exhibits the highest specific surface area at 4.56 m²/g, while KogCNp650 shows the lowest at 1.29 m²/g with Langmuir Surface, which typically yield higher estimates due to their monolayer coverage assumption, indicated specific surface areas of 13.3–27.8 m²/g (Table 3). The limited surface areas are attributed to the mild carbon dioxide activation conditions and the predominance of macropores or large mesopores, which contribute minimally to nitrogen uptake per unit mass. Despite low BET areas, all samples share a remarkably similar total pore volume (~ 40 cm³/g), suggesting that the pore architectures are dominated by a few large voids. The adsorption desorption hysteresis loops and corresponding pore size distributions, with adsorption average pore diameters ranging from 5.996 to 9.419 nm and desorption values between 10.667 and 16.021 nm, classify these materials as mesoporous to microporous. This classification is consistent with the findings of Do et al. (2024), who reported hierarchical pore structures in coal derived carbons synthesized under similar conditions²⁴. The relatively larger average pore sizes observed in the 550 °C samples, such as KogCNp550 with 94.19 Å during adsorption and 160.22 Å during desorption, reflect the influence of activation temperature on pore development. Lower synthesis temperatures tend to promote the formation of wider pores, likely due to reduced structural densification and incomplete volatilization during carbonization⁴⁵. At 550 °C, weaker gasification reactions yield fewer but larger pores, while stronger CO₂ etching at 650 °C progressively narrows and multiplies mesopores. Despite the low surface areas, adsorption performance may still be enhanced through favorable surface chemistry and accessible pore structure, rather than relying on extensive microporosity to drive capacity per unit mass. The lack of sharp graphitic peaks confirms that complete graphitization was not achieved, preserving a higher density of edge sites and defects that can enhance interactions with aromatic pollutants⁴⁶. Despite their limited surface areas, the materials demonstrated excellent PAH adsorption efficiencies, achieving removals approaching 90%. This challenges the conventional paradigm that adsorption capacity scales linearly with surface area. Instead, adsorption performance appears to be governed by surface chemistry and accessible pore geometry rather than total BET area. The absence of sharp graphitic peaks indicates incomplete graphitization, which preserves abundant edge defects and oxygenated functionalities. These moieties act as high-affinity sites for PAH binding via π–π interactions, hydrogen bonding, and electrostatic forces⁵⁷. Additionally, the dominance of mesopores facilitates diffusion and minimizes steric hindrance for large aromatic molecules, a limitation often encountered in ultramicroporous carbons⁵⁸. Beyond surface adsorption, partitioning into amorphous domains and absorption within the bulk carbon matrix further contribute to pollutant sequestration⁵⁹.

Fig. 7.

BET analyses for synthesized carbon nanoparticles.

Table 3.

Parameters for BET analyses of synthesized carbon nanoparticles.

Nanoparticle (oC)	BET(m²/g)	Pore Volume (cm³/g)	Langmuir Surface Area (m²/g)	Pore Size Adsorption (nm)	Pore Size Desorption (nm)
GomCNp650	4.5636	40.0091	27.7859	7.0625	11.2865
GomCNp550	2.0967	40.0045	13.3262	8.5165	12.4503
KogCNp650	1.2862	40.0019	23.9198	5.9963	10.6667
GomCNp550	1.6541	40.0039	19.5202	9.4185	16.0215

Collectively, the BET and Langmuir analyses confirm that while these carbons lack extensive microporosity, their defect-rich surfaces, functional group chemistry, and mesoporous architectures provide the primary drivers of PAH uptake. These findings are consistent with emerging literature emphasizing that the adsorptive efficiency of nanostructured carbons is dictated more by surface chemistry and pore accessibility than by total surface area⁶⁰.

The characterization data collectively validate the effective production of carbon nanoparticles from coal. SEM micrograms demonstrate a nanoscale spherical and irregular morphology that differs from bulk coal, while EDX analysis indicates a carbon-dominant composition with minimal oxygen-containing functions⁵³. The XRD pattern displays a large peak at around 23° (2θ), indicative of amorphous carbon nanoparticles²⁴. FTIR spectra reveal surface groups including C = O and C = C, aligning with conventional nanoparticle surface chemistry³⁹. The increased BET surface area compared to raw coal further corroborates the production of nanosized carbon⁴⁵. Collectively, these observations confirm that our material is consistent with research on coal derived CNPs.

Kinetics of adsorption of phenanthrene and naphthalene

The adsorption kinetics were rigorously analyzed to elucidate the rate determining mechanisms governing phenanthrene uptake onto the synthesized coal derived nanoparticles. Parameters derived from the linear fitting of pseudo first order, pseudo second order (PSO), and Elovich models were employed to interpret the dynamic behavior of the system as shown in Supplementary material (SM) 1,2,3 and 4.

The nonlinear fitting of PHE adsorption kinetics to the PFO, PSO, and Elovich models, as shown in Fig. 8, demonstrated that the PSO model consistently yielded the most precise fit across all carbon nanoparticles synthesized. PSO fits produced somewhat superior coefficients of determination (R²) compared to PFO and lower reduced chi-square c values. Additionally, the computed equilibrium capacities qₑcal derived by PSO closely aligned with the experimental values, but PFO underestimated the uptake, as shown in Table 4. In contrast, Elovich fits yielded lower R² values and significantly elevated x²_red, signifying inadequate descriptive efficacy.

Fig. 8.

Non-linear fitting for the kinetic of phenanthrene adsorption.

Table 4.

Parameters for the non-linear fitting of kinetic of phenanthrene adsorption.

Kinetic Model	Parameters	GomCNp650˚C	GomCNp550˚C	KogCNp650˚C	KogCNp550˚C
	QeExp (mg/g)	0.5959	0.5792	0.5872	0.5656
Pseudo First Order	Qecal (mg/g)	0.5855	0.5519	0.5772	0.5387
	K1(g mg min − 1)	0.1094	0.0985	0.1387	0.1387
	R2	0.9501	0.9389	0.9490	0.9366
	X2	0.0052	0.0058	0.0052	0.0057
Pseudo Second order	Qecal (mg/g)	0.6328	0.6085	0.6089	0.6008
	K2 (g mg min − 1)	0.3336	0.2706	0.5549	0.2307
	R2	0.9506	0.9440	0.9503	0.9411
	X2	0.0052	0.0053	0.0052	0.0053
Elovich	β(g/mg)	1	1	1	1
	α (mg g min 1)	0.0412	0.0129	0.0142	0.0124
	R2	0.5237	0.5757	0.4685	0.5953
	X2	0.0499	0.0401	0.0555	0.0364

Both the PFO and PSO models exhibited a strong correlation with the experimental data for phenanthrene adsorption, indicated by elevated R² values and Both the PFO and PSO models exhibited a strong correlation with the experimental data for phenanthrene adsorption, indicated by elevated X² values. For GombCNp650˚C, the PFO model produced a q_ecal of 0.5855 mg/g, a rate constant (k1) of 0.1094 min⁻¹, a R² of 0.9501, and χ² of 0.00523. The PSO model for the identical material exhibited a q_ecal of 0.6328 mg/g, k2 of 0.3336 g/mg·min, a R² of 0.95061, and a χ² of 0.00517. The q_eexp for GombCNp650˚C was 0.5959 mg/g. Conversely, the Elovich model regularly demonstrated a suboptimal fit (R² < 0.60), suggesting that heterogeneous chemisorption was not the principal rate-controlling mechanism. This observation corresponds with prior research indicating that the Elovich model frequently exhibits suboptimal performance in adsorption processes primarily governed by physisorption³³. The strong correlation between the q_ecal and q_eexp values for both PFO and PSO models further validates their applicability. The PSO dominance indicates that chemisorption is the rate-limiting process, involving valence forces such as π–π electron donor–acceptor interactions between the aromatic rings of PHE and the graphitic sp² domains, as well as surface functional groups on the carbon nanoparticles. This aligns with multiple publications indicating that the PSO model most accurately characterizes phenol kinetics on carbonaceous sorbents, such as agricultural biochars⁶¹, magnetic multi-walled carbon nanotubes⁶², and waste-derived activated carbons⁶³. Tian et al. (2022) achieved comparable results for PHE uptake in soils (R² = 0.999), corroborating the predominance of chemisorption. These data confirm that the larger, more polarizable three-ring PHE molecule establishes stronger, more specific contacts with carbon surfaces, hence extending the surface-reaction phase¹⁶. Conversely, the adsorption kinetics of NAP, as shown in Fig. 9, were optimally characterized by the PFO model across all carbon nanoparticle versions, exhibiting remarkably high R² values and substantially lower x² compared to the PSO model. The qₑ, Cal results from PFO are strongly aligned with qₑ exp, surpassing PSO in predictive precision as shown in Table 5. This indicates that film diffusion or intraparticle diffusion is the rate-limiting phase for NAP adsorption, aligning with its lower molecular size and lesser interaction energy with carbon surfaces compared to PHE. Both the PFO and PSO models yielded exceptional fits, with R² values continuously exceeding 0.99. For GombCNp650˚C, the non-linear PFO model produced a qe cal of 0.5984 mg/g, k1 of 0.06317 min⁻¹, R² of 0.9991, and χ² of 0.0000049. The PSO model exhibited a q_eexp of 0.6769 mg/g, k2 of 0.1332 g/mg·min, R² of 0.9932, and χ² of 0.000037. The q_eexp for GombCNp650˚C was 0.6059 mg/g. The Elovich model for NAP adsorption repeatedly demonstrated inadequate fits (R² < 0.41). The close alignment between the estimated and q_eexp values from the PSO model further substantiates this claim. The PFO model, although offering a satisfactory match, especially for NAP, is frequently linked to physical interactions like hydrophobic partitioning and π–π stacking⁶⁴. The marginal statistical advantage of the PFO model in certain instances indicates that physical adsorption mechanisms play a substantial role in the absorption of the PAHs.

Fig. 9.

Non-linear fitting for the kinetic of phenanthrene adsorption.

Table 5.

Parameters for the non-linear fitting of kinetic of naphthalene adsorption.

Kinetic Model	Parameters	GomCNp650˚C	GomCNp550˚C	KogCNp650˚C	KogCNp550˚C
	QeExp (mg/g)	0.6043	0.6004	0.5994	0.6004
Pseudo First Order	Qecal (mg/g)	0.5984	0.5814	0.5913	0.5827
	K1(g mg min − 1)	0.0632	0.0636	0.0605	0.0543
	R2	0.9991	0.9976	0.9979	0.9981
	X2	4.89 × 10− 5	1.24 × 10− 5	1.09 × 10− 5	9.80 × 10− 5
Pseudo Second order	Qecal (mg/g)	0.6769	0.6579	0.6733	0.6728
	K2 (g mg min − 1)	0.1332	0.1380	0.1056	0.1254
	R2	0.9932	0.9972	0.9966	0.9965
	X2	3.74 × 10− 5	1.46 × 10− 5	1.75 × 10− 5	1.83 × 10− 5
Elovich	β(g/mg)	1	1	1	1
	α (mg g min 1)	0.0108	0.0104	0.0106	0.0102
	R2	0.2882	0.2938	0.3288	0.4014
	X2	0.0389	0.0362	0.0355	0.0305

The non-linear fitting results of PHE and NAP adsorption on carbon nanoparticles exhibit substantial concordance with existing concepts and discoveries in the literature on adsorption kinetics. The persistent inadequate fit of the Elovich model (R² < 0.60 for PHE and R² < 0.41 for NAP) corroborates the notion that this model is typically ill-suited for adsorption processes predominantly governed by physisorption⁶⁵. Wang and Guo (2020) underscore the necessity of accurate interpretation of kinetic models, and our results corroborate the assertion that the Elovich model, although beneficial for specific chemisorption processes, fails to sufficiently characterize the kinetics observed in this study, indicating that heterogeneous chemisorption is not the rate-limiting step.

The robust efficacy of both the PFO and PSO models in elucidating the adsorption kinetics of PHE and NAP aligns with the prevailing literature. Both models demonstrated a strong fit for PHE (R² > 0.94), while both exhibited an exceptional fit for NAP (R² > 0.99). The marginal statistical advantage of the PFO model in certain cases, especially for NAP, indicates a substantial influence of physical adsorption mechanisms, including hydrophobic partitioning and π–π stacking. Rezvani et al. (2025) underlines the significance of these interactions in the improved adsorption capabilities of graphene oxide sponges⁶⁴. The finding that both PFO and PSO models yield adequate fits, indicating heterogeneous adsorption mechanisms. The nonlinear regression explains the kinetics of PHE and NAP adsorption compared to the linear regression.

The adsorption kinetics were rigorously evaluated using pseudo-first-order, pseudo-second order, and Elovich models, with both linear and non-linear regression applied to the experimental data. The pseudo-second-order model consistently provided the best description of the kinetic data, evidenced by high correlation coefficients (R² > 0.99). This strong fit suggests that the rate-limiting step in the adsorption process is predominantly chemisorption, involving electron sharing or transfer between the adsorbent and adsorbate. Although non-linear regression is generally favored for kinetic modeling to avoid potential biases introduced by data linearization. The reported R² values indicate that the pseudo-second-order model, regardless of the specific fitting method, accurately captured the time-dependent adsorption behavior of phenanthrene and naphthalene onto the synthesized carbon nanoparticles.

The apparent discrepancy between linear and non-linear kinetic fittings observed in this study is not uncommon in adsorption research. Linearized forms of the pseudo-first-order (PFO) and pseudo-second-order (PSO) models often yield artificially high correlation coefficients due to mathematical transformation bias, which can overemphasize the fit of the PSO model and misleadingly suggest chemisorption as the dominant mechanism⁶⁶.Recent analyses have shown that non-linear regression is more reliable as it preserves the original error structure and typically provides a more realistic description of the adsorption kinetics Therefore, the moderate R² values obtained for both PFO and PSO in the non-linear analysis of phenanthrene adsorption suggest that the process cannot be fully explained by a single kinetic model, but may involve contributions from surface reactions and diffusion simultaneously⁶⁷.In many studies, implausibly small ΔG° values have been reported due to such methodological oversights, even though typical ranges for physisorption (–20 to 0 kJ mol⁻¹) and chemisorption (> − 40 kJ mol⁻¹) are well established³³. Recent reviews strongly recommend re-evaluating ΔG°, ΔH°, and ΔS° based on rigorously defined, dimensionless equilibrium constants to ensure consistent and meaningful interpretation of spontaneity and adsorption affinity³³. By aligning our interpretation with these established considerations, we acknowledge the limitations of model selection and parameter derivation, while reinforcing the reliability of the observed adsorption behavior.

Isotherm analysis

To understand the mechanisms of adsorption, the equilibrium adsorption data for phenanthrene and naphthalene on the four coal-derived carbon nanoparticles were fitted using both linear and nonlinear isotherm models. The Langmuir, Freundlich, and Temkin models are shown in SM 5,6,7 and 8. The adsorption data for both PHE and NAP were analyzed using non-linear regression to fit Langmuir, Freundlich, and Temkin isotherm models as shown in Figs. 10 and 11. The R values and chi-squared values denote the adequacy of fit for each model. A higher R value and a lower X² value typically indicate a superior match. For PHE, the Langmuir model often exhibits the greatest R values among all four adsorbents. and R² values and X² values are likewise comparatively modest as shown in Table 6. This indicates that the Langmuir model effectively describes PHE adsorption, signifying monolayer adsorption on a uniform surface. The Temkin model demonstrates favorable R² values and minimal X values indicating that the heat of adsorption diminishes linearly with increasing coverage because of adsorbent-adsorbate interactions. The Freundlich model, although yielding acceptable R² values, typically exhibits elevated X² values relative to the Langmuir and Temkin models, indicating a suboptimal match for heterogeneous surface adsorption. The fit for NAP has greater variability across the models and adsorbents. For all adsorbent the Langmuir model demonstrates elevated R² values and minimal X² values, signifying an excellent fit as shown in Table 7. The Freundlich model demonstrates excellent fits, indicating a more heterogeneous adsorption process for NAP on some adsorbents. The Temkin model yields satisfactory fits for NAP, typically exceeding that of PHE.NAP typically demonstrates superior maximum adsorption capacities (Qmax) relative to PHE among the evaluated adsorbents. This may be ascribed to variations in molecule size, polarity, and solubility. Naphthalene, as a smaller molecule with two fused benzene rings, may exhibit superior accessibility to the adsorbent’s pores and active sites in comparison to phenanthrene, which consists of three fused benzene rings and is bigger. Although the Langmuir model often offers a satisfactory fit for both PHE and NAP, the Freundlich model seems to be a more robust alternative for NAP adsorption on certain adsorbents, indicating a more heterogeneous adsorption mechanism for NAP. The Langmuir model consistently demonstrates a robust match for PHE, suggesting a mostly monolayer adsorption. Wang et al. (2020) reported the Langmuir model is a suitable fit, corroborating the monolayer adsorption evident in adsorption of PHE biochar⁶¹. In another study by Cao et al. (2023) on PHE adsorption on multi-walled carbon nanotubes also reported the Langmuir and Freundlich models as the best fits, corroborating our finding that both models may be applicable depending on the particular adsorbent⁶⁸. Funari et al. (2023) reported on naphthalene adsorption on high-density materials examines isotherms, aligning with our observation of differing fits for NAP across various adsorbents⁶⁹. Similarly in a study by Gosh et al. (2025) on naphthalene removal utilizing synthetic silica doped PVA hydrogel highlights the Langmuir isotherm model as the best fit however, our results indicates that the Freundlich model may also provide a robust fit for naphthalene, underscoring the significance of adsorbent heterogeneity. Another study on the Langmuir isotherm for naphthalene adsorption onto nano-FeO carbon composite reinforces the applicability of the Langmuir model for NAP, while also referencing alternative models such as Freundlich and Temkin. This is consistent with our observations that various models can characterize NAP adsorption based on the adsorbent used⁷⁰. Generally, the non- linear regression for the isotherm provides a better result and best fit for both PHE and NAP adsorption compared to the results of the linear fits.

Fig. 10.

Parameters for non-linear fitting for the isotherm of phenanthrene adsorption.

Fig. 11.

Parameters for non-linear fitting for the isotherm of napthalene adsorption.

Table 6.

Parameters for non-linear fitting for isotherm of phenanthrene adsorption.

Isotherm Models	Parameters	GomCNp650˚C	GomCNp550˚C	KogCNp650˚C	KogCNp550˚C
Langmuir	Qmax(mg/g)	1.1917	1.1208	1.1749	1.1115
	KL (L/mg)	2.9728	2.6760	3.2640	2.3474
	R2	0.9961	0.9405	0.9653	0.9081
	X2	0.0057	0.0091	0.0058	0.0139
Freundlich	KF	0.8093	0.7252	0.8187	0.6950
	1/n	0.3953	0.3984	0.3893	0.4243
	R2	0.9383	0.9271	0.9405	0.93822
	X2	0.0103	0.0111	0.0100	0.0093
Temkin	B	0.2487	0.2368	0.2428	0.2237
	A	32.0866	27.6636	36.1035	29.7439
	R2	0.9767	0.9552	0.9777	0.9309
	X2	0.0039	0.0068	0.0038	0.0105

Table 7.

Parameters for non-linear fitting for isotherm of naphthalene adsorption.

Isotherm Models	Parameters	GomCNp650˚C	GomCNp550˚C	KogCNp650˚C	KogCNp550˚C
Langmuir	Qmax(mg/g)	1.3911	1.4644	1.7359	1.5608
	KL	4.9789	3.2612	2.2615	3.0348
	R2	0.9092	0.9661	0.9079	0.9640
	X2	0.0195	0.0072	0.0224	0.0079
Freundlich	KF	1.1386	1.1462	0.4993	0.9686
	1/n	0.3652	0.4753	0.4653	0.9465
	R2	0.8139	0.9673	0.8927	0.9684
	X2	0.0399	0.0007	0.0055	0.0069
Temkin	B	0.2923	0.3025	0.3688	0.2951
	A	52.4831	39.1894	24.2768	46.4448
	R2	0.8931	0.9769	0.8921	0.9446
	X2	0.0229	0.0049	0.0263	0.0123

For phenanthrene (PHE), adsorption was spontaneous at room temperature for nanoparticles activated at higher temperatures, as shown in Fig. 12a and confirmed by the negative ΔG° values in Table 8.The moderately exothermic ΔH∘ values (less than 20 kJ mol⁻¹) and negative ΔS∘ indicate a physisorption-dominated process involving weak π–π stacking and van der Waals forces, consistent with findings by Gupta and Gupta (2016)⁷¹.The negative entropy change further suggests reduced molecular randomness at the solid–liquid interface, reflecting the ordered arrangement of PHE molecules during adsorption, similar to observations for biochar reported by Stavrinou et al. (2022). In contrast, samples activated at lower temperatures exhibited non-spontaneous adsorption despite being exothermic, likely due to the presence of polar surface functionalities that disrupt hydrophobic interactions²⁷.

Fig. 12.

Thermodynamics for phenanthrene and Naphthalene adsorption.

Table 8.

Thermodynamics of phenanthrene adsorption.

Nanoparticles	R2	R	DH j/mol	DH Kj/mol	DS	DG j/mol	DG Kj/Mol
GomCNp650˚C	0.9276	8.314	−2696	−2.696	−85.291	−1061.54	−1.0615
GomCNp550˚C	0.9024	8.314	−12,883	−12.884	−43.970	220.76	0.2208
KogCNp650˚C	0.9192	8.314	−15,964	−15.964	−51.507	−330.80	−0.3308
KogCNp550˚C	0.9465	8.314	−12,883	−12.884	−26.785	995.11	0.9951

For naphthalene (NAP), the thermodynamic parameters show a different trend (Table 9). Adsorption was generally non-spontaneous at ambient temperature, with positive ΔG∘\Delta ΔG∘ values (Fig. 12b). Low or slightly positive ΔH∘ values suggest weak van der Waals interactions, consistent with Ge et al. (2015), who reported similar behavior for coal-derived adsorbents¹⁹. In some cases, adsorption was entropy-driven and endothermic, where positive ΔS∘ indicates a gain in disorder, likely due to desorption of structured water molecules at the adsorbent interface⁷². This entropy-driven mechanism has also been noted for naphthalene adsorption on wheat-residue biochar, where spontaneity occurred only at elevated temperatures²⁷.

Table 9.

Thermodynamics of naphthalene adsorption.

Nanoparticles	R2	R	DH j/mol	DH Kj/mol	DS	DG j/mol	DG Kj/Mol
GomCNp650 ˚C	0.95455	8.314	−11798.2	−11.7982	−32.0861	1994.42	1.9944
GomCNp550 ˚C	0.97221	8.314	15672.77	15.6728	45.0669	1522.24	1.5222
KogCNp650 ˚C	0.83997	8.314	1385.594	1.3856	49.3699	1425.51	1.4255
kogCNp550 ˚C	0.82495	8.314	12994.99	12.9949	47.71388	2219.10	2.2191

Overall, PHE adsorption was more favorable than NAP across all conditions, particularly for high temperature activated nanoparticles, where extended aromatic domains and hydrophobic sites enhanced molecular affinity. In contrast, NAP adsorption required additional thermal input and was primarily governed by entropy effects. These findings are in agreement with Gupta and Singh (2018), who also attributed negative ΔS∘ values in PHE adsorption to restricted molecular mobility and organized π–π interactions at the adsorbent surface²⁵.

Adsorbent reusability

The reusability of PHE was assessed during three successive runs, with the findings detailed in Fig. 13a. The percentage removal (%R) demonstrated a significant decline with each successive run, signifying a decrease in the material’s efficacy upon repetition. In Run 1, PHE attained a removal effectiveness of 95.34%. In Run 2, the efficiency declined to 83.68%, indicating a reduction of approximately 11.66% points. In Run 3, the removal efficiency decreased to 80.18%, representing a reduction of approximately 3.50% points from Run 2. After 3 cycles, the removal effectiveness of PHE diminished by roughly 15.16% from its original performance.

Fig. 13.

Reusability of adsorbent for phenanthrene and Naphthalene adsorption.

The reusability of NAP was evaluated during three successive runs, with the findings detailed in Fig. 13b. Like PHE, NAP exhibited a decrease in efficiency across repeated runs; however, it retained a somewhat larger clearance % in subsequent runs.

In Run 1, NAP had a removal efficiency of 96.05%. During Run 2, the efficiency declined to 91.11%, representing a reduction of approximately 4.94% points. At Run 3, the removal efficiency was 86.71%, reflecting a decline of approximately 4.40% points from Run 2. Throughout the 3 trials, the overall reduction in removal efficiency for NAP was around 9.34%.

The disparity aligns with the molecular properties of the two PAHs: PHE, possessing a larger aromatic framework, exhibits stronger interactions with the adsorbent surface, resulting in possible pore obstruction and incomplete desorption during regeneration, whereas the smaller NAP molecule desorbs more easily, enabling the adsorbent to preserve a higher proportion of its active sites. Comparable performance trends have been recorded in the literature, hence corroborating the data provided above. Li et al. (2016) indicated that Fe@SiO₂@PDA magnetic nanocomposites exhibited sustained high removal efficiency for polycyclic aromatic hydrocarbons, including phenanthrene, over several cycles; however, they noted a gradual decrease in performance, which they ascribed to the accumulation of strongly bound PAH residues on active sites⁷³. This observation corresponds with the decrease in PHE reusability seen in the current study. Similarly, Stavrinouet al. (2023) demonstrated that phenanthrene adsorption on carbon generated from coffee waste diminished after four cycles, necessitating cold plasma regeneration to restore efficacy. The results indicate that phenanthrene, owing to its greater hydrophobicity and π–π stacking propensity, typically induces more significant efficiency losses with repeated utilization compared to lighter PAHs like naphthalene⁷⁴.The increased retention of NAP elimination noted here corresponds with previously documented findings on smaller PAHs. In another study it was discovered that activated carbon maintained over 85% of its adsorption performance for naphthalene after three cycles, with only slight reductions due to partial fouling of surface micropores. This confirms the current observation that NAP may be efficiently eliminated with consistent performance, even after multiple cycles⁷⁵.These comparisons substantiate the current findings by illustrating that efficiency loss is a predictable consequence of repeated adsorption–desorption cycles, the rate of efficiency loss is contingent upon the molecular characteristics of the adsorbate, and NAP typically demonstrates superior reusability performance compared to PHE across various adsorbents. The strong alignment with published studies not only enhances the credibility of the current findings but also underscores the overarching issue of formulating regeneration systems that may completely restore adsorption efficiency, particularly for bigger polycyclic aromatic hydrocarbons like phenanthrene.

According to Aungthitipan et al. (2025), after several regeneration cycles, the adsorption performance gradually declined. This reduction can be attributed to surface fouling, pore blockage, and partial structural degradation, which limit the availability of active sites during successive reuse. Such instability in reusability emphasizes that, for effective PAH removal, robust desorption and regeneration strategies are essential to maintain long-term adsorption efficiency. These findings suggest that while reusability is achievable, optimization of regeneration protocols is necessary to ensure the economic viability and sustainability of the adsorption process⁷⁶.

Conclusion

This study successfully demonstrated the synthesis and comprehensive characterization of carbon nanoparticles (CNPs) derived from subbituminous coals (Gombe and Kogi regions) using a CO₂-assisted solid-state activation method at 550 °C and 650 °C. Characterization confirmed the formation of spherical nanoparticles (10–100 nm) with high carbon purity and surface functionalities that were critically influenced by the activation temperature. Specifically, 650 °C activation yielded CNPs with enhanced graphitic character, superior thermal stability (< 10% mass loss up to 800 °C), and a more hydrophobic surface, which promoted π–π interactions. In contrast, 550 °C activation resulted in CNPs with a higher abundance of oxygenated and aliphatic functional groups, favoring polar interactions.

Batch adsorption experiments showed that the 650 °C-activated Gombe sample (GomCNp650) achieved monolayer adsorption capacities (Qmax) of 1.26 mg g⁻¹ for phenanthrene and 1.45 mg g⁻¹ for naphthalene. While these capacities are modest when compared to some advanced adsorbents reported in the literature, the study provides valuable insights into the underlying adsorption mechanisms. The adsorption kinetics were well-described by the pseudo-second-order model, indicating a chemisorption-dominated process, and the Langmuir isotherm model provided the best fit for the equilibrium data, suggesting monolayer adsorption on homogeneous sites. Thermodynamic analysis further elucidated the adsorption mechanisms, revealing that phenanthrene adsorption was spontaneous and exothermic, while naphthalene adsorption became favorable at higher temperatures due to entropy gains.

Furthermore, reusability tests indicated a significant decline in adsorption performance over just three cycles (e.g., phenanthrene removal dropped from 95% to 80%). This highlights a limitation in the long-term practical application of these specific CNPs without further optimization. Despite the modest adsorption capacities and reusability challenges, these findings underscore the critical role of activation temperature in tailoring the physicochemical properties of coal-derived CNPs and influencing their interaction mechanisms with PAHs. The study contributes to a fundamental understanding of PAH adsorption on these materials, informing future research directions for enhancing performance, improving reusability, and exploring their potential in specific niche applications for wastewater treatment.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

D.U.N, A.A.A and A.S.G conceived the research idea. D.U.N carried out the experiment and wrote the original draft. A.A.A, A.S.G, NHHA and Z.U.Z supervised and validated the work. H.A.R and A.A.B revised and edited the manuscript. All authors have read and agreed to the published version of the manuscript.

Data availability

Data will be made available upon request. Please contact Ummulkhairi Nasiru Danmallam.

Declarations

Competing interests

The authors declare no competing interests.