Optimised Operating Conditions and Performance Landscape of Metal-Doped Carbon Dots for Dye Decolourisation in Water Treatment Systems

Abstract

Synthetic dyes frequently persist through conventional wastewater treatment, motivating the use of advanced oxidation processes capable of breaking down these stable molecules. Metal-doped carbon dots (CDs) offer a tuneable platform for catalytic dye degradation in water, although their performance varies strongly with operating conditions. The aim of this work was to determine how temperature, H₂O₂ dosage, and pH influence the catalytic behaviour of Fe-, Cu-, Zn-, and Mg-doped CDs during the degradation of methylene blue (MB) and rhodamine B (RB), optimised using a Taguchi L27 orthogonal array design. Temperature and oxidant loading were the dominant factors: higher temperatures accelerated reactions through Arrhenius-type kinetics, while increasing H₂O₂ availability improved removal until excessive levels began to suppress •OH generation. Across all condition sets, apparent rate constants spanned 7.0 × 10⁻⁴–2.65 × 10⁻² min⁻¹, with t₅₀ values of 26–217 min and t₉₀ extending from ~86 min to >700 min; final decolourisation ranged from ~17% to nearly 100%. pH played a secondary role, mainly affecting dye speciation and surface adsorption. Dopant identity shifted the optimum operating region for each catalyst: Fe- and Cu-CDs achieved complete or near-complete removal of both dyes at pH 7 and 50 °C with relatively low H₂O₂ dosage (0.5–1.0 mL); Zn-CDs reached equivalent performance at pH 7 and 25 °C but required higher oxidant loading (1.5 mL of H₂O₂), reflecting their photo-induced rather than thermally driven activation mechanism; Mg-CDs performed comparably under the same conditions as Fe- and Cu-CDs. The resulting condition–catalyst map highlights the operating regimes that maximise efficiency while minimising chemical input, providing a practical framework for selecting carbon-dot-based catalysts for water treatment applications.

1. Introduction

Synthetic dyes are persistent chromophore pollutants that challenge conventional treatment due to their aromaticity and functionalisation [1]. Advanced oxidation processes (AOPs) offer rapid colour removal by generating highly reactive oxygen species (ROS) [2,3]. However, the efficacy and economy of AOPs hinge on operating conditions as much as on the catalyst: among these, reaction temperature and oxidant (H₂O₂) dosage typically dominate performance by governing both intrinsic reaction rates and the net ROS flux available for chromophore breakdown [4,5]. Carbon dots (CDs) combine low-cost synthesis, rich surface functionality, and tuneable electronic structure with accessible metal doping that introduces redox-active sites [6,7]. Metal-doped CDs can efficiently activate oxidants and concurrently pre-concentrate dyes via surface interactions [8,9,10,11]. However, comparative reports often appear inconsistent because catalytic outcomes emerge from the coupling of two levers: (i) activation kinetics, largely dictated by temperature and oxidant availability at doped active sites, and (ii) adsorptive pre-conditioning, shaped by surface charge and functionalities [12,13,14]. Distinguishing which lever dominates under practical operating windows is essential for rational deployment [15]. In practice, temperature and H₂O₂ dosage are the primary drivers of decolourisation [16]: temperature accelerates elementary steps (Arrhenius behaviour) and can mitigate mass-transfer limitations, while oxidant dosage sets the instantaneous ROS generation rate [17]. At low dosage, under-supply of ROS limits conversion [18]; at high dosage, self-scavenging and non-productive recombination diminish catalytic efficiency, creating a finite optimal range rather than a monotonic benefit [19]. By contrast, pH predominantly acts as a secondary modulator that shifts these optimal windows by altering dye speciation [20,21], surface charge [22], and solution equilibria [23]. Once temperature and H₂O₂ are appropriately tuned, pH rarely overrides their control but can improve robustness and reagent economy within each CD–dye pair [24]. Despite many positive demonstrations, practitioners lack an integrated, condition-aware performance landscape that prescribes which metal-doped CD is optimal under which combination of temperature and oxidant dosage [25] and how pH should be adjusted to widen the operating window or reduce chemical input [26]. Moreover, connecting observed performance to surface chemistry [27] and surface-charge descriptors (e.g., FTIR-inferred functionalities and zeta-potential trends) can supply mechanistic handles for purposeful tuning beyond empirical trial and error [28,29,30].

Building on our previous work, which established synthesis protocols and surface-characterisation profiles for Fe-, Cu-, Zn-, and Mg-doped CDs [31], the present study determines how temperature, oxidant dosage, and pH govern their catalytic performance during dye degradation. A Taguchi L27 orthogonal array design was adopted to systematically cover the three-factor parameter space while keeping the experiment count manageable, enabling both main-effect ranking and identification of optimal operating conditions for each catalyst–dye combination. Methylene blue (MB) and rhodamine B (RB) were chosen as model dyes because they differ meaningfully in molecular architecture and oxidation behaviour: MB is a cationic phenothiazine dye with a relatively simple planar structure that is readily oxidised [32], whereas RB is a xanthene dye whose bulkier aromatic framework and lower oxidation potential present a more demanding degradation challenge [32]. Both dyes are widely used as model substrates in AOP studies due to their relevance to coloured industrial effluents [33,34]. Decolourisation is quantified using rate-based metrics (apparent pseudo-first-order rate constants) [35] and time-based metrics (t₅₀, t₉₀) [36], and outcomes are linked to FTIR and zeta-potential descriptors that capture differences in active-site chemistry and interfacial interactions [37]. The four dopants were selected because prior results indicated that Fe-CDs excel in redox-driven activation, Cu-CDs favour ligand-mediated charge transfer, Zn-CDs offer structural stability under neutral conditions, and Mg-CDs promote hydrogen-bonding interactions [37]—suggesting fundamentally different kinetic regimes that have not previously been compared across a systematic operating window. The novelty of this work lies in constructing a condition–catalyst performance map that integrates these datasets to reveal how each dopant reaches its optimal activity through different reaction pathways, advancing from qualitative mechanistic insights to a decision-oriented framework for selecting metal-doped CDs in practical decolourisation systems [38,39].

2. Materials and Methods

2.1. Reagents and Solutions

Citric acid (C₆H₈O₇, ≥99.5%, Sigma-Aldrich, St. Louis, MO, USA), urea (99.0–100.5%, Sigma-Aldrich, St. Louis, MO, USA), ferric citrate (FeC₆H₅O₇, ≥98.5%, Sigma-Aldrich, St. Louis, MO, USA), copper(II) acetate monohydrate (Cu(CH₃COO)₂·H₂O, 98%, Sigma-Aldrich, St. Louis, MO, USA), zinc acetate dihydrate (Zn(CH₃COO)₂·2H₂O, 99.99%, Sigma-Aldrich, St. Louis, MO, USA), and magnesium chloride hexahydrate (MgCl₂·6H₂O, ≥98%, Sigma-Aldrich, St. Louis, MO, USA) were used as received. Hydrogen peroxide (H₂O₂, 30% w/w, Sigma-Aldrich, St. Louis, MO, USA), methylene blue (MB, dye content ≥82%, Sigma-Aldrich, St. Louis, MO, USA), and rhodamine B (RB, ≥95%, Sigma-Aldrich, St. Louis, MO, USA) were used as received. For the reactive oxygen species quenching experiments, isopropanol (IPA, ≥99.5%, Sigma-Aldrich, St. Louis, MO, USA), tert-butanol (TBA, 99+%, Thermo Fisher Scientific, Scoresby, VIC, Australia), dimethyl sulfoxide (DMSO, ≥99.9%, Sigma-Aldrich, St. Louis, MO, USA), sodium azide (SA, ≥99%, Sigma-Aldrich, St. Louis, MO, USA), and ethylenediaminetetraacetic acid (EDTA, ≥99%, Sigma-Aldrich, St. Louis, MO, USA) were used as received. Ultrapure water (≥18.2 MΩ cm) was used for all preparations.

2.2. Synthesis of Metal-Doped CDs

Fe-, Cu-, Zn-, and Mg-doped CDs were prepared by a one-step hydrothermal carbonisation. Typically, citric acid (1.0 g) and urea (0.5 g) were dissolved in 30 mL of deionised water under stirring; then 1.0 g of the metal precursor (FeC₆H₅O₇, Cu(CH₃COO)₂, Zn(CH₃COO)₂, or MgCl₂) was added [40,41,42,43]. Equal masses rather than equal molar quantities of precursor were used, consistent with the referenced hydrothermal protocols [40,41,42,43]. In a one-step carbonisation at 260 °C, actual dopant incorporation is governed by precursor decomposition kinetics and metal–ligand coordination affinity rather than by initial molar input; standardising by mass therefore provides a reproducible synthetic input and ensures direct comparability with prior studies [31]. The homogeneous solution was transferred to a 50 mL Teflon-lined autoclave and heated at 260 °C for 4 h [44]. After cooling, the dispersion was centrifuged (10,000 rpm, 15 min), filtered (0.22 µm), and dialysed (MWCO 3.0 kDa, 24 h) against water to remove small molecules and free ions. Purified CDs were stored as aqueous dispersions at 4 °C or freeze-dried for later use [44,45]. Full structural and surface characterisation of these CD systems, including TEM imaging and XPS analysis, is reported in our prior work [31].

2.3. Fourier-Transform Infrared Spectroscopy (FTIR)

FTIR spectra were recorded on a PerkinElmer Frontier spectrometer. Each spectrum was acquired using 32 scans at a resolution of 4 cm⁻¹, with background subtraction performed using pure water. Data were collected across the mid-infrared (400–4000 cm⁻¹) and far-infrared (40–700 cm⁻¹) regions to comprehensively capture vibrational modes associated with functional groups and metal–ligand interactions.

2.4. Zeta Potential

Zeta potential was measured at 25 °C (equilibrated for 120 s to ensure thermal stability) using a Zetasizer Nano ZS with a folded capillary cell (DTS1070) [46]. CD dispersions (0.1 mg mL⁻¹) were prepared in deionised water, maintaining pH between 6.8 and 7.0; each condition was measured in triplicate with up to 100 runs per replicate [47]. Specific refractive indices were applied for accurate sizing (Mg-CD: 1.69; Cu-CD: 2.10; Fe-CD: 1.55; Zn-CD: 2.00). The refractive index (RI) of the CD dispersions was independently measured using an Abbe refractometer (Atago DR-A1, Tokyo, Japan) at 589 nm to support optical characterisation/DLS settings; RI values are unitless and used in Zeta calculations where samples were homogenised by vortexing (5 min at 500 rpm), filtered (0.45 μm) to remove aggregates, and applied to the prism, equilibrated at 20 °C, and measured at 589 nm with a precision of ±0.0001 in refractive index. Triplicates were averaged (standard deviation ≤0.0002) following calibration with distilled water [48,49].

2.5. Decolourisation Experiments

Catalytic decolourisation of MB (C₁₆H₁₈N₃SCl) and RB (C₂₈H₃₁ClN₂O₃) was assessed under dark, isothermal conditions with H₂O₂ as oxidant [50,51]. Different pHs, temperatures, and dosages of H₂O₂ were used to degrade 0.25 mg mL⁻¹ of cationic (MB) and zwitterionic (RB) dye, respectively. Before oxidation, 4 mL of CD dispersion was mixed with dye and shaken in the dark for 60 min to reach adsorption equilibrium. The reaction was then initiated by combining 1 mL of equilibrated mixture and 1.5 mL of dye solution with 1.5 mL of 3% H₂O₂ and pH buffer in a cuvette (total 4 mL) and immediately placing it into a dark incubator. Absorbance was first recorded 60 s after initiation and then every 10 min for 300 min (monitoring MB at 664 nm and RB at 554 nm) [51,52]. Decolourisation was calculated as (A_i − A_t)/A_i, where A_i is the initial absorbance, and A_t is the absorbance at time t [53].

2.6. Reactive Oxygen Species (ROS) Assays

Reactive species contributing to the degradation of MB were examined through quenching experiments. The approach relied on the assumption that inhibiting an active oxidative pathway would lead to a measurable decrease in dye degradation. Isopropanol (IPA), dimethyl sulfoxide (DMSO), and tert-butanol (TBA) were added to suppress hydroxyl radicals (•OH). Singlet oxygen (¹O₂) was inhibited using sodium azide (SA), while thylenediaminetetraacetic acid (EDTA) was included to capture positive holes (h⁺).

For each test, MB or RB solutions with an initial concentration of 20 mg L⁻¹ were prepared. IPA, DMSO, and IPA were added at 5% (v/v), whereas SA and EDTA were dosed at 50 mM and 4 mM, respectively. The catalytic reaction was initiated using the optimised reaction conditions which included Zn-CDs (0.50 mg mL⁻¹) together with 3% H₂O₂ (1 mL) at 25 °C. Changes in dye concentration were monitored for 30 min by UV–visible spectroscopy. Identical experiments performed without any scavenger served as controls.

2.7. Kinetic Modelling

A pseudo-first-order kinetic model was applied, as H₂O₂ was present in excess relative to the dye concentration throughout the reaction, maintaining an approximately constant oxidant level and reducing the rate expression to a first-order dependence on dye concentration [54,55]. The apparent pseudo-first-order rate constant (k) for dye decolourisation was obtained by monitoring changes in dye absorbance by UV–visible spectroscopy over the course of the catalytic reaction, with dye concentrations inferred from absorbance measurements:

$$\ln \left({\displaystyle \frac{C_0}{C_t}}\right)=k t$$

where $C_0$ is the initial concentration, $C_t$ is the concentration at time $t$, and $k$ is the apparent rate constant; $\mathrm{l}\mathrm{n}(C_0/C_t)$ was plotted versus $t$ to obtain the slope. Concentrations were inferred from absorbance at 664 and 554 nm as described above [54,55].

2.8. Experimental Design and Analysis (Taguchi Approach)

To systematically evaluate how temperature (T), initial H₂O₂ dosage ([H₂O₂]₀), and pH govern performance while maintaining a compact experiment count, Taguchi’s orthogonal-array method was adopted:

• Design. An L27 (3¹³) orthogonal array was used to allocate three-level factors across 27 runs. In the main array, T, [H₂O₂]₀, and pH were set at three practical levels each (e.g., representative low/medium/high settings); other controllable variables (e.g., CD dose, sampling horizon) were fixed to isolate their effects [56,57,58,59]. Dye identity (MB vs. RB) and CD identity (Fe-, Cu-, Zn-, Mg-doped) were evaluated via blocked/stratified execution of the same L27 across dye/CD sets so that the three-level factors remained balanced; pooled ANOVA across strata was used to compare trends while avoiding mixed-level confounding.
• Response metrics. Primary responses were apparent pseudo-first-order rate constant (k_app) (larger the better) and/or t₉₀ (smaller the better). Secondary responses (endpoint removal at fixed time; oxidant economy) were recorded for trade-off analysis [60].
• Signal-to-noise (S/N) ratios. For each trial i in a condition set of size n, a larger k_app or % removal is better. The corresponding S/N ratio was calculated using the larger-the-better criterion:

$$\mathrm{S}/\mathrm{N}=-10{\mathrm{l}\mathrm{o}\mathrm{g}}_{10} ({\displaystyle \frac{1}{n}}\sum _{i=1}^n {\displaystyle \frac{1}{y_i^2}})$$

A smaller t₉₀/t₉₉ is better. The S/N ratio was therefore calculated using the smaller-the-better criterion:

$$\mathrm{S}/\mathrm{N}=-10{\mathrm{l}\mathrm{o}\mathrm{g}}_{10} \left({\displaystyle \frac{1}{n}}\sum _{i=1}^n y_i^2\right)$$

• Analysis of effects. Main-effect plots and Δ(S/N) contrasts were used to rank factor influence. ANOVA on S/N and on the raw response (k_app or t₉₀) were used to quantify contributions and significance; normality/variance were checked, and non-parametric substitutes were applied if assumptions were violated [61].
• Optimum prediction and confirmation. The predicted optimum was obtained by superposing best levels from main-effect means/S/N [62]. Confirmation runs at the predicted settings were conducted to verify gains and to compute absolute/relative error against predictions; where applicable, minimal-dosage operating points (lowest [H₂O₂]₀ achieving the target removal within the time constraint) were identified to support oxidant-economy claims [63].
• Table S1 in Supplementary Files shows the experimental design matrix for Taguchi optimisation, with a total of 54 runs for each CD, including replicates. “a” series represent assessment for RB. For instance, runs 1–6 (number in black colour) represent Mg-CD-catalysed MB, and runs 1a–6a (number in red colour) represent Mg-CD-catalysed RB. Three sets of control groups were used: 1, dyes + pH buffer; 2, dyes + hydrogen peroxide; 3, dyes + CDs + pH buffer. Absorbance was measured by UV–Vis spectroscopy at specific wavelengths (around 664 nm for MB and 554 nm for RB) over a duration of 300 min of degradation time [64].

2.9. Quality Control, Statistics, and Safety

All catalytic runs were performed in triplicate and reported as mean ± SD; significance was set at p < 0.05. Experiments were performed under dark conditions to exclude photo effects. Handling of H₂O₂, dyes, and metal-bearing wastes followed institutional hazardous waste protocols.

3. Result and Discussion

3.1. Decolourisation of MB and RB Under Varying Conditions

To support the kinetic analysis, a set of control tests was run in parallel to distinguish true catalytic oxidation from background degradation. As expected, the dyes-only systems showed negligible colour loss over 300 min (MB < 3%; RB < 1%), confirming that neither dye undergoes spontaneous decay under the reaction conditions (

Table 1. Summary of the Taguchi-optimised best catalytic runs for metal-doped carbon dots (M-CDs) in the degradation of methylene blue (MB) and rhodamine B (RB).

Metal-CDs	Dye	Final Decol. (%)	k (min−1)	t90 (min)	Experimental Conditions (pH/T/H2O2 mL)	Proposed Mechanism (Description)
Mg-CDs	MB	100	0.0229	112	pH 7/50 °C/1.0 mL	Surface-mediated photo-Fenton-like process; •OH and •O2− generation assisted by surface adsorption and polar sites
Mg-CDs	RB	99	0.0169	142	pH 7/50 °C/1.0 mL	Photo-Fenton synergy; adsorption-controlled oxidation with weaker electrostatic interaction for RB
Cu-CDs	MB	100	0.0213	94	pH 7/50 °C/1.5 mL	Cu2+/Cu+ redox cycling activating H2O2 to produce •OH; efficient electron transfer at neutral pH
Cu-CDs	RB	100	0.0204	65	pH 7/50 °C/1.5 mL	Cu-mediated Fenton-like reaction with π–π and electrostatic adsorption synergy
Zn-CDs	MB	100	0.0156	162	pH 7/25 °C/1.5 mL	Lewis-acid Zn sites and photo-induced electron transfer forming •OH/•O2− radicals
Zn-CDs	RB	97	0.0170	155	pH 7/25 °C/1.5 mL	Similar photo-oxidation mechanism; lower rate due to zwitterionic structure of RB
Fe-CDs	MB	100	0.0228	112	pH 7/50 °C/0.5 mL	Classical Fe2+/Fe3+ Fenton cycle stabilised on CD surface; photo-induced redox enhancement
Fe-CDs	RB	98	0.0222	123	pH 7/50 °C/0.5 mL	Surface-bound Fe sites driving ROS generation with slower adsorption of RB molecules

). Adding H₂O₂ alone increased removal to 18–40% for MB and 10–35% for RB, indicating that peroxide-driven oxidation is possible but remains limited without a catalyst. In contrast, the dye + M-CDs systems (no H₂O₂) produced considerably higher removal—40–57% for MB and 30–48% for RB—demonstrating that the doped CDs themselves contribute to dye breakdown, likely through surface-mediated ROS formation or adsorption-enhanced pathways [59,65].

These trends were consistent across pH 3, 7, and 11. The acidic condition produced slightly higher removal in the M-CD-only tests, which is consistent with proton-assisted activation pathways reported for carbon materials [34,66]. However, neither low pH nor elevated H₂O₂ dosage alone accounted for the high efficiencies observed in the full catalytic experiments. Collectively, the controls confirm that the strong degradation seen in the Taguchi-optimised runs originates from the combined action of M-CDs and H₂O₂ rather than from adsorption or weak peroxide oxidation alone [67,68,69,70]. The catalytic behaviour of undoped CDs under comparable conditions was characterised in our previous study [31], where N-doped CDs showed markedly lower dye degradation efficiency; this observation motivated the use of metal-doped systems in the present work. A comparison of the catalytic performance of the present M-CD systems with the representative literature catalysts, including operational conditions and rate constants, is provided in our previous publication [31].

All decolourisation curves for MB and RB are provided in the Supplementary Information for clarity and to avoid repetition in the main text.

Each entry lists the experimental run achieving the highest decolourisation efficiency within the L27 design. The final percentage, apparent pseudo-first-order rate constant (k), and time for 90% decolourisation (t₉₀) were determined directly from absorbance–time data by linear regression of ln(C/C₀) versus time. Runs without the suffix “a” correspond to MB, whereas “a” denotes the RB tests performed under identical conditions. Mechanistic assignments are proposed based on catalyst-specific FTIR and zeta-potential data (Section 3.2 and Section 3.3) and, for Zn-CDs, direct ROS quenching experiments (Section 3.4). For Fe- and Cu-CDs, the proposed redox mechanisms are further supported by the established literature [71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88].

The catalytic behaviour of the best-performing runs—Mg-CDs (5/5a), Cu-CDs (8/8a), Zn-CDs (17/17a), and Fe-CDs (23/23a)—shows how the dopants influence both the kinetics and the underlying reaction pathways in this system. Notably, all optimised conditions occurred at pH 7. This is important because conventional Fenton processes typically require strongly acidic conditions (pH ≈ 2–3), which can reduce H₂O₂ stability, promote sludge formation, and limit practical application. In contrast, all four doped CDs operate efficiently under neutral conditions, indicating that the catalysts can sustain radical generation, maintain colloidal stability, and interact effectively with both MB and RB without the drawbacks associated with acidic environments [59,89]. Each optimal run reflects the best combination of pH, temperature, and H₂O₂ dosage within the Taguchi matrix and results in near-complete decolourisation for both dyes under visible-light irradiation.

For Mg-CDs, the most effective conditions were obtained in Run 5 and Run 5a, both conducted at pH 7, 50 °C, and 1.0 mL of H₂O₂. Under these mild settings, the catalysts achieved complete MB and 99% RB removal, with apparent rate constants of 0.0229 min⁻¹ and 0.0169 min⁻¹ and corresponding t₉₀ values of 112 min and 142 min, respectively. Although Mg is not a redox-active metal, its incorporation into the carbon matrix enhances the electronic conductivity and introduces polar surface sites that promote photo-induced H₂O₂ activation and the formation of •OH and •O₂⁻ radicals [39,90,91]. The elevated surface polarity enhances the adsorption of cationic MB molecules, leading to rapid initial degradation. In contrast, the zwitterionic RB exhibits weaker interaction with the surface, resulting in a slower yet more consistent reaction trajectory. The pseudo-first-order linearity observed in both reactions indicates that Mg-CDs can maintain a continuous radical flux under neutral conditions, eliminating the necessity for the strongly acidic environments characteristic of traditional Fenton systems.

Cu-CDs exhibited superior early-stage kinetics under the same neutral-pH conditions (pH 7, 50 °C, 1.5 mL of H₂O₂). Run 8 for MB produced a rate constant of 0.0213 min⁻¹ with t₉₀ = 94 min, whereas Run 8a for RB achieved complete decolourisation in 65 min at k = 0.0204 min⁻¹; Run 8 represents decolourisation to MB, and Run 8a represents decolourisation to RB. This rapid degradation is attributed to the Cu²⁺/Cu⁺ redox couple, which perpetually activates H₂O₂ to produce hydroxyl radicals via a heterogeneous Fenton-like mechanism. The graphitic domains in Cu-CDs enhance charge transfer between copper centres and the carbon lattice, thereby reducing recombination and maintaining the production of ROS. The aromatic structures of both MB and RB facilitate robust π–π interactions with the Cu-CD surface, thereby enhancing degradation rates [92]. However, despite the fast kinetics of Cu-CDs, they remain more sensitive to oxidant dosage and pH fluctuations than Zn- or Mg-based systems.

Zn-CDs exhibited distinct catalytic behaviour. Although their apparent rate constants were slightly lower than those of Fe and Cu systems, they achieved exceptionally stable and linear kinetics under neutral conditions, indicating a well-controlled photo-oxidation mechanism. Both Run 17 (MB) and Run 17a (RB) were conducted at pH 7, 25 °C, and 1.5 mL of H₂O₂, yielding 100% dye degradation with k = 0.0156 min⁻¹ (t₉₀ = 162 min) for MB and k = 0.0170 min⁻¹ (t₉₀ = 155 min) for RB. Unlike Fe or Cu systems, Zn does not rely on variable-valence redox cycling. Instead, Zn²⁺ functions as a Lewis-acidic site that enhances electron transfer and polarises H₂O₂ molecules, facilitating their decomposition into •OH and •O₂⁻ radicals. The reaction proceeds through photo-induced charge separation within the carbon core, providing uniform radical generation and minimal self-quenching. The outstanding linearity (R² ≈ 0.999) and reproducible kinetics indicate that Zn-CDs maintain an ideal balance between ROS production and stability. Their neutral-pH operation and consistent degradation of both cationic (MB) and zwitterionic (RB) dyes collectively establish Zn-CDs as the most sustainable and predictable catalyst among the studied systems.

Fe-CDs, by contrast, represent the classical Fenton mechanism. The optimal runs—Run 23(MB) and Run 23a (RB)—were carried out at pH 7, 50 °C, and 0.5 mL of H₂O₂, achieving complete removal of MB and RB with rate constants of 0.0228 min⁻¹ and 0.0222 min⁻¹ and t₉₀ values of 112 min and 123 min, respectively. The Fe²⁺/Fe³⁺ redox cycle efficiently activates H₂O₂ into •OH radicals, and the CD surface prevents Fe precipitation by stabilising Fe³⁺ species through coordination with oxygenated functional groups. Elevated temperature enhances both radical generation and dye diffusion, enabling high catalytic turnover even at reduced oxidant input.

A cross-comparison of all optimum decolourisation profiles for each metal CD reveals two distinct mechanistic families: Fe- and Cu-doped CDs act as redox-driven Fenton-like catalysts with rapid but condition-sensitive activity, whereas Mg- and Zn-doped CDs operate via photo-induced surface activation governed by charge separation and electronic modulation of the carbon framework. Among these, Zn-CDs demonstrate the most balanced performance, showing high efficiency, reproducible kinetics, and environmental stability. Although Fe- and Cu-CDs offer slightly faster initial reaction rates, their reliance on redox cycling and sensitivity to pH reduce long-term sustainability. Mg-CDs provide reliable performance at neutral pH but at slower rates. By contrast, Zn-CDs achieve full degradation of both MB and RB at mild temperature and neutral pH through a purely photo-driven mechanism, avoiding acidification, secondary sludge formation, or metal leaching. This combination of kinetic consistency, structural stability, and eco-compatibility confirms that Zn-CDs are the most optimal catalyst among all metal-doped systems, providing a sustainable pathway for efficient Fenton-like degradation under real-world conditions.

3.2. FTIR Analysis of M-CDs Before and After Catalytical Reactions

3.2.1. Surface Functional Group (O–H, C=O, C-N, C–O–C)

Fe-, Cu-, Zn-, and Mg-CDs exhibit distinct FTIR peak shifts and intensity changes for surface O–H, C=O, COO⁻/C-N, and C–O–C groups after treatment with MB and RB. Figure 1a–d shows the surface function groups of Mg-, Cu-, Zn-, and Fe-CDs, respectively.

Fe-, Cu-, Zn-, and Mg-CDs were characterised by FTIR before and after treatment with MB and RB (Figure 1a–d). With the exception of the COO⁻/C–N region, all principal surface functional group bands—including O–H (~3270 cm⁻¹), C=O (~1635 cm⁻¹), and C–O–C (~1075 cm⁻¹)—remained at their pre-reaction positions across all four CD systems after both MB and RB treatment, with any movement falling within the instrument’s ±8 cm⁻¹ resolution limit. These groups therefore remain largely intact under the reaction conditions and do not play a significant role in the surface chemistry changes during catalytic degradation. Full peak positions for all functional groups are provided in Table S1 (Supplementary Information).

The only region showing reproducible and meaningful shifts was the COO⁻/C–N band. Before reaction, all samples exhibited this feature at approximately 1395–1397 cm⁻¹. After MB treatment, the band shifted well beyond the resolution threshold: to +77 cm⁻¹ for Fe-CDs (to 1474 cm⁻¹), +43 cm⁻¹ for Mg-CDs (to 1440 cm⁻¹), and +23 cm⁻¹ for both Cu- and Zn-CDs (to 1420 cm⁻¹). After RB treatment, the same trend was observed but with smaller magnitudes—Fe-CDs shifted to ~1415 cm⁻¹ (+18 cm⁻¹) and Mg-CDs to ~1417 cm⁻¹ (+20 cm⁻¹), while Cu- and Zn-CDs moved by only ~3 cm⁻¹, which falls within measurement variability. These shifts are summarised in

Table 2. COO⁻/C–N band positions (cm⁻¹) for metal-doped CDs before and after MB and RB treatment. Bold values indicate shifts exceeding the ±8 cm⁻¹ significance threshold.

CD	Before (cm−1)	After MB (cm−1)	Shift	After RB (cm−1)	Shift
Fe-CDs	1397	1474	+77	1415	+18
Mg-CDs	1397	1440	+43	1417	+20
Cu-CDs	1397	1420	+23	1400	+3
Zn-CDs	1397	1420	+23	1400	+3

.

These blue shifts in the COO⁻/C–N region reflect a redistribution of electron density at carboxylate sites during the catalytic reaction. As dye molecules adsorb onto the CD surface, carboxylate groups act as primary binding sites through electrostatic interaction or coordination with surface metal centres, weakening the symmetric C–O bond and shifting the vibrational frequency to higher wavenumbers [93,94]. MB’s cationic character drives stronger electrostatic attraction to the negatively charged CD surface than RB, accounting for the larger COO⁻ perturbation seen after MB treatment. Across the four dopants, the shift magnitude follows the degree of metal–ligand involvement during degradation: Fe sites, which cycle actively between Fe²⁺ and Fe³⁺, sustain ongoing disruption of their local coordination geometry, while Zn²⁺—redox-inactive under these conditions—leaves neighbouring carboxylate groups relatively undisturbed [93,95].

3.2.2. Metal–Oxygen Coordination (M–O Bond)

The metal–oxygen (M–O) stretching bands in the FTIR provide insight into the state of the dopant metals within the CDs. The complete FTIR spectra for all samples are provided in the Supplementary Information (Figures S5–S8). Peak positions and shifts are summarised in Table S2. Before decolourisation, each doped sample exhibits an M–O vibration in the 500–600 cm⁻¹ region, confirming successful incorporation of the metal into the carbon dot structure. Specifically, Mg–CDs show an Mg–O stretch around 549 cm⁻¹, Cu–CDs a Cu–O band near 590–600 cm⁻¹, Zn–CDs a Zn–O band at ~567 cm⁻¹, and Fe–CDs an Fe–O band around 580–585 cm⁻¹ (Table S2). These values align with known M–O vibrations in oxide environments [87,96].

After the dye degradation reactions, the M–O bands undergo red shifts (lower wavenumbers) indicating alterations in the metal centers’ bonding and coordination. In general, MB treatment produces a larger shift (greater change in metal coordination) than RB treatment, mirroring the trend seen for carboxylate bands. For Fe-CDs, the Fe–O band was initially ~582 cm⁻¹, corresponding to Fe–O–C linkages or small Fe–oxide clusters on the CD.

After MB treatment, the Fe–O stretching band shifts markedly from ~582 to ~525 cm⁻¹ (Table S2), corresponding to a red shift of −57 cm⁻¹. Such a large downshift is characteristic of Fe–O vibrations in iron(III) oxide or hydroxide species (e.g., Fe₃O₄, Fe₂O₃, FeO(OH)) and indicates a substantial reorganisation of the Fe coordination environment. This suggests that part of the Fe initially bound as Fe–O–C linkages on the CD surface migrated into a more extended Fe–O–Fe network during the MB/H₂O₂ reaction. Under Fenton-like conditions, Fe²⁺/Fe³⁺ redox cycling in the presence of H₂O₂ is known to promote the formation of amorphous or poorly crystalline Fe(III) oxide/hydroxide phases, which exhibit lower-frequency Fe–O vibrations. In contrast, after RB treatment, the Fe–O band shifts only to ~546 cm⁻¹, corresponding to a smaller red shift of ~36 cm⁻¹. This indicates partial modification of the Fe coordination environment, likely involving limited formation of Fe–O–Fe bridges, while a fraction of Fe remains in its original coordination state. Overall, MB induces a more extensive transformation of Fe centres toward oxide/hydroxide phases, whereas RB causes a detectable but less complete structural rearrangement. The disappearance of the ~580 cm⁻¹ Fe–O–C mode together with the emergence of a new low-frequency Fe–O band is consistent with the literature reports describing Fe migration from carbon-bound sites into oxide-like domains during Fenton reactions.

Pristine Cu–CDs exhibit a Cu–O stretching band at ~592 cm⁻¹ (Table S2), consistent with Cu–O bonding in a CuO-like environment. After MB treatment, this band shifts to ~577 cm⁻¹, corresponding to a moderate red shift of ~15 cm⁻¹. Although smaller than that observed for Fe, this shift is significant and indicates a change in the local Cu–O coordination environment. The band position remains within the typical CuO vibrational range, suggesting that Cu largely retains its Cu²⁺ oxidation state while experiencing subtle structural rearrangements such as increased coordination number or the formation of Cu–O–Cu linkages (e.g., growth of small CuO or Cu(OH)₂ domains). Such behaviour is consistent with the oxidative nature of the reaction conditions and with reports showing that Cu-doped carbon materials maintain Cu²⁺-based coordination under Fenton-like environments [95,97]. The persistence and intensity of the Cu–O band after MB treatment further indicate that Cu remains predominantly in the solid phase, with modification of its local bonding environment rather than dissolution. After RB treatment, the Cu–O band shifts only slightly to ~584 cm⁻¹ (a ~8 cm⁻¹ red shift), which lies within normal experimental variation. This suggests that RB degradation induces minimal structural change in the Cu–O coordination. Overall, MB causes minor but detectable restructuring of Cu sites, whereas RB has little impact on the Cu–O bonding environment.

Zn–CDs display a Zn–O stretching band at ~567 cm⁻¹ (Table S2), typical of Zn–O bonding in a ZnO-like environment [42]. Notably, neither MB nor RB treatment results in a meaningful shift in this band, which remains at ~565–567 cm⁻¹ in all cases. Both the position and intensity of the Zn–O vibration are essentially unchanged after reaction, indicating that the Zn coordination environment is highly stable under the applied H₂O₂/thermal conditions. This behaviour is consistent with Zn²⁺ being redox-inactive in Fenton-like systems and therefore not directly involved in catalytic redox cycling. The results imply that Zn remains bound in its original coordination environment (e.g., Zn–O–C linkages or small ZnO domains) throughout both MB and RB degradation. In contrast to Fe, Cu, and Mg, the Zn dopants act largely as spectator species, with no detectable structural transformation of the Zn–O framework.

The most pronounced M–O structural change is observed for Mg–CDs. Prior to reaction, Mg–CDs exhibit an Mg–O stretch at ~549 cm⁻¹, characteristic of Mg–O bonding in an MgO-like environment [43,98]. After MB treatment, this band disappears or becomes extremely weak, while a new band emerges at ~426 cm⁻¹ (Table S2). Such a low-frequency vibration is characteristic of Mg–O stretching in magnesium hydroxide (Mg(OH)₂) or related highly coordinated, hydrated Mg–O environments. The shift of more than 120 cm⁻¹ reflects a major softening of the Mg–O bond and indicates conversion of Mg from an oxide-like state to a hydroxide or strongly hydrated coordination environment. This transformation is consistent with the strong affinity of Mg²⁺ for oxygen ligands and its tendency to undergo hydration under hot aqueous peroxide conditions. Rather than redox chemistry, the Mg dopants appear to undergo ligand re-coordination and hydration, forming amorphous Mg(OH)₂-like species with longer and weaker Mg–O bonds. After RB treatment, a similar transformation is observed, with the appearance of a new band at ~432 cm⁻¹ and the loss of the original 549 cm⁻¹ Mg–O band. The slightly higher wavenumber compared to MB-treated Mg–CDs suggests a less complete or less developed hydroxide-like environment. Overall, both dyes induce conversion of MgO-like domains into hydrated Mg–O species, with MB driving this structural transformation more strongly than RB.

Across all systems, MB consistently drives more noticeable structural changes than RB.

Redox-active dopants, such as Fe and Cu, exhibit measurable red shifts in their M–O stretching bands, reflecting progressive modification of their coordination environments and the development of oxide- or hydroxide-like domains, with the effect being particularly pronounced for Fe during MB degradation. By contrast, Mg, despite being redox-inactive, undergoes an even more substantial spectral change, indicating a major reorganisation of its bonding environment toward a hydrated or hydroxide form. In sharp contrast, the Zn–O vibration remains essentially unchanged after both MB and RB treatments, highlighting the exceptional structural stability of Zn sites under the applied reaction conditions. The systematically larger spectral shifts observed in MB-treated samples suggest that MB degradation creates a more disruptive interfacial environment, likely involving higher radical fluxes or stronger perturbation of metal–ligand coordination. The stronger structural disruption seen under MB conditions is tied to the more intense radical environment generated during its degradation. In Fe-CDs, accumulation of Fe(III) oxide/hydroxide phases progressively consumes surface Fe²⁺ sites; with fewer active centres available for H₂O₂ activation, a gradual decline in the catalytic rate under extended reaction times is anticipated; CDs behave differently—the small Cu–O shift points to efficient Cu²⁺ regeneration within the Cu²⁺/Cu⁺ cycle, meaning active sites are replenished rather than consumed, which is reflected in the more consistent degradation kinetics seen experimentally. Zn-CDs show no M–O shift at all, confirming that H₂O₂ activation through Lewis-acid polarisation proceeds without any restructuring of the Zn coordination environment—a key reason why Zn-CDs maintain stable pseudo-first-order kinetics across all tested conditions [31,66,99]. Under such conditions, metal centres are more prone to restructuring, giving rise to features consistent with Fe(III) oxide or hydroxide formation, local reorganisation of Cu–O domains, and conversion of Mg–O units into hydroxide-like species. RB degradation induces the same types of transformations but to a noticeably lesser extent. Importantly, these changes in the M–O region occur in parallel with the blue shifts observed for the COO⁻ bands. Taken together, the opposing trends in the carboxylate and metal–oxygen vibrational modes indicate that dye molecules—most notably MB—interact directly with metal–carboxylate sites, altering local electron density and coordination geometry rather than causing loss of surface functional groups or metal species. The M–O bands therefore act as sensitive reporters of the dopant chemical state, and their evolution complements the functional group analysis discussed earlier. Overall, the FTIR results provide a coherent, molecular-level picture of how CD surface chemistry and dopant structure evolve during catalytic dye decolourisation, with Zn–CDs remaining largely unaffected while other metal centres respond to varying degrees. At the mechanistic level, Fe²⁺ initiates H₂O₂ cleavage via Fe²⁺ + H₂O₂ → Fe³⁺ + ·OH + OH⁻, with Fe³⁺ subsequently reduced back to close the redox cycle; the accumulation of Fe(III) oxide/hydroxide phases in the FTIR indicates that a fraction of Fe³⁺ escapes this cycle over time [83,100,101]. At Cu sites, Cu²⁺ + H₂O₂ → Cu⁺ + ·OOH + H⁺ initiates radical generation, and the mild Cu–O shift confirms that Cu²⁺ is largely regenerated rather than accumulating as inactive phases. For Zn²⁺, Lewis-acid polarisation of the H₂O₂ O–O bond lowers its dissociation energy without requiring any change in metal oxidation state, and the invariant Zn–O framework confirms this activation carries no structural cost to the catalyst [31,66].

3.3. Zeta-Potential Analysis

Zeta-potential measurements obtained before and after dye treatment (Figure 2) reflect not only dye adsorption but also surface changes induced by the Fenton-like reaction itself. All doped CDs exhibited strongly negative ζ-potentials at pH 7 (−21 to −31 mV), indicating abundant deprotonated oxygen-containing groups that stabilise the colloids. Following reaction with MB or RB, all samples showed less negative ζ values. This shift arises from a combined effect of surface charge neutralisation by adsorbed dye molecules and reaction-driven modifications of surface chemistry, including changes in metal–ligand coordination and oxygenated functional groups during oxidative degradation. The magnitude of the ζ-potential change therefore depends on both the dopant chemistry and the specific reaction pathways associated with each dye.

3.3.1. MB Interaction Trends

For MB, the measured ζ-potential changes were:

• Fe-CDs: −25.4 → −7.8 mV;
• Cu-CDs: −21.4 → −6.7 mV;
• Mg-CDs: −30.7 → −15.3 mV;
• Zn-CDs: −28.6 → −26.6 mV.

The pronounced neutralisation observed for Fe- and Cu-CDs reflects not only strong initial adsorption of MB but also reaction-induced restructuring at the metal–ligand interface. Electrostatic attraction between the cationic dye and the highly anionic CD surface promotes rapid adsorption, while concurrent Fe³⁺/Fe²⁺ or Cu²⁺/Cu⁺ redox cycling during Fenton-like degradation modifies the local coordination environment and surface charge [83,102,103,104]. In addition, π–π stacking between the aromatic MB molecules and the carbon core facilitates close interfacial contact, enhancing electron transfer during the reaction. As degradation proceeds, partial consumption of surface-accessible dye and reorganisation of metal-coordinated oxygen species collectively contribute to the substantial ζ-potential shifts and are consistent with the rapid MB degradation kinetics observed for these systems [88].

Mg-CDs display an intermediate ζ-potential shift. While MB adsorption on Mg-CDs is governed mainly by hydrogen bonding and dipole interactions on hydroxyl-rich surfaces [105,106,107,108,109], the Fenton-like reaction induces further surface hydration and ligand rearrangement around Mg²⁺, moderating the overall extent of charge neutralisation. By contrast, Zn-CDs show only a minor change in ζ-potential. Although MB can π-stack with the carbon framework, it lacks functional groups capable of strong Lewis-acid coordination to Zn²⁺; moreover, Zn sites remain largely inert during the reaction. Consequently, reaction-driven changes in surface charge are minimal, consistent with the small ζ shift observed [110].

3.3.2. RB Interaction Trends

The corresponding ζ-potential changes for RB were:

• Fe-CDs: −25.4 → −19.6 mV;
• Cu-CDs: −21.4 → −15.1 mV;
• Zn-CDs: −28.6 → −14.2 mV;
• Mg-CDs: −30.7 → −9.1 mV.

In contrast to MB, RB exhibits a zwitterionic structure, and its interaction with CDs is less dominated by simple electrostatics and more strongly influenced by coordination and hydrogen bonding [77,79]. For Zn-CDs, the substantial ζ-potential shift is consistent with coordination between the RB carboxylate group and Zn²⁺ combined with π–π stacking of the aromatic xanthene ring onto the carbon surface [110,111,112,113]. Importantly, this coordination persists during reaction, so that even as RB undergoes oxidative degradation, the metal–ligand environment around Zn is perturbed, producing a marked change in surface charge.

Mg-CDs show the largest ζ-potential change after RB treatment. This reflects the coupled effects of hydrogen-bond-mediated adsorption and reaction-induced surface hydration, as RB degradation proceeds on the highly hydroxylated and dynamically hydrated Mg-CD surface [107,114]. In contrast, Fe- and Cu-CDs exhibit comparatively small ζ-potential shifts. Although initial RB adsorption occurs, the anionic component of RB experiences electrostatic repulsion on their acidic surfaces, and steric constraints limit dense surface coverage. As degradation progresses, these factors restrict both adsorption-driven and reaction-driven charge compensation, resulting in more modest changes in ζ potential [102,103,104,115].

3.3.3. Comparison of the Interactions Between Methylene Blue and Rhodamine B with CDs

The zeta-potential results (Figure 2) illustrate the differences in the interactions of MB and RB with CDs, emphasising the significant influence of metal dopants on surface chemistry. In the case of MB, adsorption is primarily influenced by electrostatic attraction and π–π stacking interactions [39,116]. Fe³⁺ and Cu²⁺ dopants improve these processes by augmenting the density of carboxylate groups and introducing metal sites that facilitate redox coordination [102,105]. As a result, Fe- and Cu-CDs demonstrate the most significant ζ shifts and the most notable surface charge compensation after MB adsorption, which directly correlates with their enhanced kinetic rates in MB degradation. The larger zwitterionic structure of RB necessitates more intricate binding mechanisms [117]. The carboxylate group can coordinate with metal centres, while the cationic amines and aromatic moieties engage in hydrogen bonding and π–π stacking interactions [118]. Zn²⁺ and Mg²⁺ dopants, characterised by polar, Lewis-acidic, and hydroxyl-rich surfaces, demonstrate greater efficacy in anchoring RB. This elucidates the more pronounced ζ shifts noted for RB on Zn- and Mg-CDs in comparison to Fe- and Cu-CDs [102]. The balanced surface charge distribution of Zn- and Mg-CDs prevents over-neutralization, thereby maintaining colloidal stability and facilitating efficient interaction with both cationic and anionic regions of RB [88,103].

Among all systems, Zn-CDs represent the most balanced and versatile catalyst. The evolution of their ζ-potential suggests that Zn-CDs experience only limited electrostatic interaction with MB, consistent with the small change in surface charge. This indicates that adsorption is dominated by π–π stacking and hydrogen bonding rather than strong Coulombic attraction as well as by accessible Zn²⁺ centres that effectively coordinate with the carboxylate group of RB [113,119]. The simultaneous presence of electrostatic interactions for MB and coordinative/hydrogen-bonding interactions for RB endows Zn-CDs with a dual affinity that is not independently attained by Fe-, Cu-, or Mg-doped systems. Zn²⁺ modifies the electronic environment of the CDs, enhancing visible-light absorption and promoting photo-induced charge separation, thereby aiding in radical generation under neutral conditions [113]. As a result, Zn-CDs exhibit stable adsorption–desorption equilibria, enhanced surface reactivity, and a notable resistance to metal leaching [34,120,121].

In conclusion, Fe- and Cu-CDs exhibit effective electrostatic interactions with cationic dyes, whereas Mg-CDs reveal considerable adsorption driven by hydrogen bonding. Zn-CDs proficiently amalgamate these advantages. Their performance remains reliable for both MB and RB, guaranteeing surface stability across a range of pH conditions, and they operate efficiently without inducing acidification or sludge accumulation. Zn-CDs exhibit the ability to engage with dyes through a range of complementary mechanisms such as electrostatic attraction, Lewis-acid coordination, and π–π/hydrogen-bonding interactions. This establishes them as the most efficient and sustainable catalyst within the realm of doped systems for the degradation of dyes via Fenton-like processes in neutral-pH conditions.

3.4. Reactive Oxygen Species Identification

Radical quenching experiments were conducted to probe the identity of the reactive species driving Zn-CD-catalysed dye decolourisation. Five scavengers were selected to target distinct oxidative pathways, and decolourisation efficiencies measured after 30 min at 25 °C with 1 mL of H₂O₂ are presented in Figure 3. All values are referenced against a scavenger-free control, which achieved 68% MB and 54% RB removal under identical conditions.

Among the scavengers tested, DMSO produced the most severe suppression, bringing decolourisation down to 4% for MB and 3% for RB—losses of roughly 64% and 51% from the control. IPA caused less pronounced inhibition, reducing decolourisation to 17% (MB) and 7% (RB). Both reagents act as competitive ·OH scavengers that intercept hydroxyl radicals in the bulk solution before they can reach the dye chromophore [122]. The greater suppression of DMSO compared with IPA can be readily explained by its higher second-order rate constant for ·OH (~7 × 10⁹ M⁻¹ s⁻¹ versus ~1.9 × 10⁹ M⁻¹ s⁻¹ for IPA) [122]. TBA reduced decolourisation to 18% (MB) and 11% (RB), which is also reasonable considering that its rate constant for ·OH scavenging (~6 × 10⁸ M⁻¹ s⁻¹) is roughly an order of magnitude below that of DMSO [31,122]. Taken together, the graded response observed in three chemically unrelated ·OH scavengers (DMSO, IPA, and TBA) are closely related to their known rate constants, providing strong and internally consistent evidence that ·OH is the principal oxidant in the Zn-CD/H₂O₂ system.

The addition of SA, a well-known quencher of singlet oxygen (¹O₂) [31,122], reduced the decolourisation efficiencies to 24% for MB and 14% for RB. This drop in decolourisation efficiency suggests that some ¹O₂ may be generated in the system and contribute to dye degradation. However, the substantially stronger suppression produced by the ·OH scavengers indicates that ¹O₂ plays only a secondary role. It should also be noted that the concentration of SA (50 mM) was considerably lower than those of the 5% (v/v) hydroxyl radical scavengers (IPA ≈ 830 mM; TBA ≈ 530 mM). Under these conditions, SA must compete with dye molecules and other matrix constituents for available ¹O₂, leaving part of the ¹O₂ pathway unquenched. These dosing constraints mean the inhibition data likely capture only part of the ¹O₂ contribution, and a genuine involvement of this species in the oxidation process cannot be ruled out.

EDTA suppressed decolourisation to 11% (MB) and 8% (RB). Although these values are numerically close to those observed with IPA, the underlying mechanisms are distinctly different. EDTA is widely recognised as an efficient hole scavenger in heterogeneous catalytic systems [53]. In such systems, surface holes may contribute to oxidation by directly attacking adsorbed dye molecules or by promoting the formation of hydroxyl radicals from water or H₂O₂. In addition, EDTA tends to form complexes with metal-associated surface sites, blocking the associated coordination sites and disrupting local charge transfer at the catalyst interface, thereby attenuating Lewis-acid-driven oxidation routes [53]. In classical homogeneous Fenton systems, Fe²⁺ chelation by EDTA significantly reduces the catalytic activity, as it disrupts the Fe²⁺/Fe³⁺ redox cycle responsible for sustaining ·OH production [83]. In contrast, the comparatively moderate inhibition observed in Figure 3 reflects the fact that Zn²⁺, possessing a filled d¹⁰ electronic configuration, does not readily participate in redox cycling [31,66]. Instead, Zn²⁺ sites mainly function as Lewis-acid centres that polarise H₂O₂ at the CD surface, facilitating the generation of ROS. Once these Zn-associated sites are complexed by EDTA, Lewis-acid activation of H₂O₂ is partially disrupted, and radical production decreases. However, the carbon framework of the CDs can still sustain a residual oxidative flux through surface-mediated reactions, preventing complete loss of catalytic activity. The inhibition caused by EDTA therefore indicates that metal-associated surface sites contribute to H₂O₂ activation and that hole-mediated oxidation pathways may play a secondary role in the overall process, while hydroxyl radicals remain the dominant oxidising species. These findings agree well with the FTIR data presented earlier, where the Zn–O stretching band held its position after reaction. This confirms that Zn²⁺ functions as a structurally stable Lewis-acid activator rather than a centre that undergoes redox cycling.

Overall, the scavenger data indicate that ·OH is the primary oxidant responsible for MB and RB degradation, generated mainly through Lewis-acid activation of H₂O₂ at Zn²⁺ surface sites with an additional contribution from the carbon framework. The SA results suggest that singlet oxygen (¹O₂) may also occur in the system, although its contribution appears secondary relative to the dominant ·OH pathway. The inhibition observed in the presence of EDTA further indicates a modest role of metal-associated surface sites and possible hole-mediated pathways, neither of which dominates the overall oxidative outcome. The inhibition sequence (DMSO > EDTA ≈ IPA > TBA > SA) is consistent with this mechanistic interpretation and distinguishes the Zn-CD system from both classical Fenton chemistry and the redox-driven mechanisms of Fe- and Cu-doped CDs examined in this work.

3.5. Integrated Mechanistic Analysis of Zn-CDs: Zeta Potential, FTIR, and ROS

This integrative analysis focuses on Zn-CDs, as the ROS quenching experiments were conducted exclusively under Zn-CD optimised conditions, and Zn-CDs demonstrated the most consistent kinetics and structural stability among the four systems studied. A closer look at the FTIR spectra, zeta-potential data, and ROS quenching results shows a set of trends that correspond well to the interactions sketched in Figure 4. For MB, the dye mainly lies flat against the graphitic surface of the Zn-CDs, so the π–π contact highlighted in the figure is the dominant contribution. This interpretation fits the measurements: the COO⁻ band moves only slightly (1397 → ~1420 cm⁻¹), and the Zn–O vibration does not shift, which confirms that MB does not form a stable coordination bond with Zn²⁺. The small change in ζ potential (−28.6 to −26.6 mV) is also consistent with this picture, since such adsorption disturbs the surface charge only mildly [123].

RB shows a noticeably different pattern, matching the right-hand pathway in the schematic. Its carboxylate group can form hydrogen bonds with surface –OH or –COOH groups, and the dye is able to approach Zn²⁺ centres more closely than MB. This behaviour aligns with the experiments: the COO⁻ band moves to ~1400 cm⁻¹, and the ζ potential drops to −14.2 mV, both of which suggest a stronger alteration of the interfacial environment. The Zn–O band, however, remains unchanged, in agreement with the diagram indicating only limited Zn²⁺ involvement rather than a well-defined coordination interaction [124].

The lower part of Figure 4, which shows the formation of short-chain products and eventual mineralisation, also reflects what is seen in the FTIR results. The major surface groups (O–H, C=O, C–O–C) remain essentially intact after reaction, implying that Zn-CDs mainly provide a surface for activating H₂O₂ in the dark rather than undergoing significant structural modification. This supports the central pathway in the figure: generation of •OH on Zn-CDs, followed by stepwise breakdown of MB and RB into smaller acids and finally CO₂ and water. The intermediates depicted in Figure 4—including a demethylated MB derivative (Azure B) above the MB oxidation step, and the ring-opening products phenol and benzoic acid from the MB and RB backbones, respectively—are consistent with degradation pathways reported in the literature for these dyes under ·OH-driven conditions [37,66]. Their separate positions in the figure reflect the structurally distinct routes by which the phenothiazine (MB) and xanthene (RB) cores are cleaved.

Taken together, the spectral signatures and ζ-potential changes follow the mechanistic routes shown in Figure 4: MB interacts predominantly through π–π stacking; RB binds through hydrogen bonding together with weak contact with Zn²⁺; and throughout the process, the Zn–O framework remains stable enough to sustain dark H₂O₂ activation without major structural loss [123,124].

4. Conclusions

This study systematically established the performance landscape of metal-doped CDs as Fenton-like catalysts for the decolourisation of MB and RB under varied temperature, oxidant dosage, and pH conditions. Through a Taguchi L27 orthogonal design, the relative contributions of these parameters were quantitatively disentangled, revealing that temperature and H₂O₂ dosage dominate the kinetic response, whereas pH serves mainly as a secondary fine-tuning factor. The dopant identity critically modulates the optimal operating window through its influence on redox chemistry, surface charge, and electron-transfer efficiency, leading to distinctive catalyst–condition matchups across the parameter space.

Among the results, Zn CDs emerged as the most balanced catalyst, combining high catalytic activity, robust dispersion stability, and structural integrity after reaction. Zn CDs maintained consistent M–O bonding and minimal band shifts in FTIR spectra, indicating structural resilience and reversible adsorption behaviour. In contrast, Fe and Cu CDs exhibited the fastest kinetics owing to active redox cycling (Fe²⁺/Fe³⁺ and Cu⁺/Cu²⁺) but also underwent pronounced M–O band red shifts, signifying partial transformation to oxide or hydroxide phases during reaction. Mg CDs, while redox-inert, displayed moderate catalytic activity and excellent pH tolerance, with a large Mg–O shift from 549 to 426 cm⁻¹ revealing structural reordering into hydrated forms. The collective FTIR and zeta-potential analyses confirmed that carboxylate blue shifts (≈+50–80 cm⁻¹) correspond to weakened metal–ligand coordination and re-exposure of surface –COO⁻ groups, whereas M–O red shifts trace the evolution of dopant environments under oxidative stress. These spectroscopic transformations correlate strongly with catalytic performance and highlight the dynamic restructuring of surface-active sites during oxidation cycles. Mechanistically, the action of electrostatic attraction, π–π stacking, hydrogen bonding, and dopant-facilitated H₂O₂ activation governs dye degradation. Fe and Cu contribute through redox-driven radical generation, Zn primarily stabilises charge distribution and promotes electron transfer, and Mg tunes surface polarity to enhance adsorption. The resulting hierarchy—Fe > Cu ≈ Zn > Mg in intrinsic kinetics, yet Zn > Fe > Cu > Mg in structural robustness—demonstrates the trade-off between catalytic reactivity and material durability. Also, this work delivers an actionable framework for condition-aware optimisation of Fenton-like catalysts. The identified minimal-dosage operating points—where high decolourisation is achieved with reduced oxidant input—offer pathways towards greener, cost-efficient dye treatment. The integration of kinetic modelling, FTIR band changes, and zeta-potential trends helps link catalytic behaviour to surface chemistry. During degradation, H₂O₂ generates •OH and HO₂• radicals that attack MB and RB, leading to chromophore cleavage and subsequent mineralisation. Despite these advances, several directions merit further investigation. A dedicated ion leaching study measuring post-reaction metal concentrations in solution would confirm the long-term stability of the catalysts under repeated use, and recycling experiments would establish their practical reusability. Mass spectrometric identification of degradation intermediates at successive time points would provide a more complete picture of the mineralisation pathway. Extension of the present framework to real wastewater matrices containing competing ions and natural organic matter would also be valuable for assessing performance under environmentally relevant conditions. These insights help clarify how metal-doped carbon nanostructures operate in broader advanced oxidation systems. Such insights extend beyond dye degradation to broader advanced oxidation processes involving metal-doped carbon nanostructures.