Carbonised fibre-based materials for water purification and Industrial wastewater treatment: A review of textile-derived filtration systems

2.1. Polyacrylonitrile Fibres

PAN has been used as a carbon-fibre precursor since the 1960s, and the processing knowledge accumulated in the structural-composites sector applies largely unchanged to activated carbon fibre (ACF) manufacture for water treatment. Fibres are produced by wet or dry-jet wet spinning from acrylonitrile homo- or copolymers – small additions of itaconic acid or methyl acrylate moderate the exothermic cyclisation during stabilisation [24]. Heating in air at 200–300°C converts the nitrile groups into a thermally stable ladder structure of conjugated C–N and C–C bonds, after which the fibre can withstand the subsequent high-temperature step [25]. Carbonisation in nitrogen or argon at 800–1600°C removes most heteroatoms and leaves a turbostratic graphitic matrix with carbon content above 92 wt% [16]. Carbon yield is 50–55%, markedly higher than for most natural precursors, and partly compensates for the higher raw-material cost.

After activation, PAN-derived ACFs reach Brunauer–Emmett–Teller (BET) surface areas of 1000–2500 m² g⁻¹, with well-developed microporosity (pore width < 2 nm) that accounts for their high uptake of small organic molecules [14,29]. When the precursor is in nonwoven form, the bonded fibre architecture survives carbonisation, yielding an activated carbon nonwoven with a defined three-dimensional structure suitable for use in flow-through filter modules [30]. Basis weight, fibre diameter, and inter-fibre void volume are all adjustable in the precursor fabric, providing structural control over the finished filter that granular carbon beds cannot easily match. This tunability is the main practical justification for nonwoven PAN-ACF in cascade filter designs.

2.2. Cellulose Fibres and Their Carbonisation

Carbon fibres from rayon (regenerated cellulose) predate PAN-based fibres commercially, with early applications in aerospace dating to the 1950s [16]. Viscose, lyocell, and natural cotton have all been carbonised to produce porous carbon materials for environmental applications. The practical drawback versus PAN is low carbon yield: pyrolysis of cellulose fibres typically delivers only 20–30 wt% carbon because competing depolymerisation and dehydration reactions convert much of the mass to volatiles [30]. Pre-treating with phosphoric acid or zinc chloride, or using a hydrothermal step prior to carbonisation, promotes char formation and can raise yield to 35–45% [26].

Carbonised cellulose fibres retain meaningful amounts of surface oxygen functionality – carboxyl, lactone, and phenol groups – when carbonised at moderate temperatures (below about 900°C). This amphoteric character supports adsorption of cationic metal ions through ion exchange, complexation, and electrostatic attraction, the relative contributions of which shift with pH and the extent of surface oxidation [18]. It should be noted that aerogels cast from cellulose nanofibre or cellulose nanocrystal suspensions are not fibrous in the classical textile-precursor sense – they originate from nanocellulose suspensions rather than spun or natural cellulose fibres – but, after carbonisation, they give very-low-density carbon sponges with open macro- and mesopore networks [31], and are mentioned here only for functional comparison with fibre-based ACF media. This hierarchical porosity supports higher filtration velocities than predominantly microporous PAN-ACFs, which can be useful when high filtration velocity matters more than maximum surface area.

2.3. Chitosan-Based Fibres

Chitosan is obtained from crustacean shells by alkaline deacetylation of chitin. What distinguishes it from PAN fibres or cellulose fibres as a precursor is the high density of primary amine groups (−NH₂) along the backbone, which coordinate transition-metal ions through nitrogen lone-pair donation and electrostatic mechanisms. Upon pyrolysis, however, the −NH₂ groups are not retained as such; instead, nitrogen is incorporated into pyridinic-N, pyrrolic-N, and graphitic-N configurations. Pyridinic and pyrrolic sites at pore edges retain lone-pair donor character and are responsible for the selective affinity of carbonised chitosan for Cu²⁺, Pb²⁺, Hg²⁺, Cd²⁺, and Ni²⁺ [32]. The degree of deacetylation, typically 75–95% for commercial grades, controls amine density and hence metal-adsorption capacity [33]. Fibres can be wet-spun from dilute acetic acid solution; crosslinking with glutaraldehyde or epichlorohydrin before pyrolysis improves dimensional stability. Pyrolysis of crosslinked chitosan at 600–900°C under nitrogen yields nitrogen-doped porous carbons with 2–8 at % N, which may enhance electrostatic affinity for anionic species depending on the predominant nitrogen configuration (pyridinic vs. graphitic N) and solution pH, and can improve catalytic activity in oxidative processes [32].

2.4. Lignocellulosic and Waste-Derived Precursors

Flax, hemp, jute, coir, and kenaf have all been tested as ACF precursors. Their heterogeneous cell-wall structure – cellulose microfibrils in a hemicellulose-lignin matrix – tends to give broader and less predictable pore-size distributions than pure cellulose fibres or PAN [27]. Lignin deserves separate mention: available in large quantities from pulp and paper production, it has carbon yields of 40–50 wt% depending on lignin source (kraft, organosolv, or soda) and carbonisation temperature, which are closer to PAN than most polysaccharide precursors [28]. Post-consumer textile waste, mainly cotton-polyester blends, has also been examined as a feedstock. Pyrolysis of such mixed-fibre waste gives composite carbons in which the polyester-derived aromatic carbon structures provide a degree of structural reinforcement and the cellulose-fibre-derived fraction contributes surface functional groups [34]. Feedstock heterogeneity is a real problem for process control and reproducibility, but cost and end-of-life management considerations continue to motivate interest in waste-derived feedstocks. In practice, however, the cotton-to-polyester ratio in post-consumer streams varies widely between garment grades and collection sources, and few studies have evaluated whether carbonisation protocols optimised on model single-fibre textiles transfer reliably to unsorted blends. Reported BET surface areas and adsorption capacities for waste-derived carbons show substantially larger inter-study variability than for PAN or virgin cellulose precursors, and reproducibility within ±20% between batches has rarely been demonstrated [34]. Until robust pre-sorting and blending protocols are established at industrial scale, batch-to-batch consistency remains an unresolved obstacle for waste-textile carbonisation feedstocks.

2.5. Carbonisation and Activation Processes

Whatever the precursor, ACF production follows the same general sequence. An optional pre-treatment step – oxidative stabilisation for PAN, catalyst impregnation for cellulose fibres – prepares the fibre for carbonisation at 600–1600°C in an inert atmosphere, which removes heteroatoms and develops the turbostratic carbon structure. Activation then opens and expands the pore network. Physical activation uses steam or CO₂ at 800–1000°C; the Boudouard reaction (C + CO₂ → 2CO) preferentially burns off the most reactive carbon at defect sites and pore edges, giving predominantly microporous products with narrow pore-size distributions. Chemical activation with KOH – the most widely used agent – works by intercalation and subsequent volatilisation of potassium species, which generates additional mesopores alongside existing micropores; BET areas above 3000 m² g⁻¹ have been reported under optimised laboratory conditions, though at this level of porosity mechanical integrity tends to be compromised [35,36].

Temperature is the variable with the most pronounced effect on pore structure and surface chemistry. Below about 700°C, a significant share of heteroatom functionality is retained – oxygen and nitrogen groups useful for adsorbing polar contaminants but that limit graphitic order and electrical conductivity. Above 1200°C, graphitic ordering increases, which strengthens π–π stacking interactions with aromatic pollutants but reduces the density of ion-exchange-active surface groups [37]. For water-treatment purposes, 800–1000°C followed by mild activation is generally where the useful range lies: BET areas fall between 800 and 1500 m² g⁻¹, the pore structure is mixed micro-mesoporous, and enough residual oxygen functionality survives to maintain wettability and ion-exchange capacity. Heating rate and hold time at peak temperature have secondary but measurable effects: slow ramp rates (1–3°C min⁻¹) promote more uniform pore development, whereas rapid pyrolysis can create gaseous-product concentration gradients that cause cracking along the fibre cross-section [38].

For filter module applications, the dimensional changes that occur during carbonisation require attention. PAN nonwovens typically shrink 20–40% linearly depending on stabilisation conditions, applied tension, and precursor composition, and the shrinkage is anisotropic – more pronounced along the fibre length than across the cross-section – which can distort a module fabricated to defined dimensions [39]. Pre-tensioning the precursor fabric during the thermal treatment and optimising the heating profile can keep planar distortion below about 5%, sufficient to preserve inter-fibre pore geometry and ensure predictable pressure drop across the filter element.

PAN remains the most widely used starting material for high-performance ACFs, though bio-based and waste-derived alternatives have attracted growing interest as routes to more sustainable production [40]. The main precursor groups are compared in Table 1.

3.1. Physicochemical Characterisation of Carbonised Fibre Surfaces

Interpreting adsorption data and comparing results across studies requires consistent characterisation of both surface chemistry and pore structure. X-ray photoelectron spectroscopy (XPS) is the primary tool for quantifying surface elemental composition and for resolving the chemical state of heteroatoms: deconvolution of the N 1 s core-level spectrum distinguishes pyridinic-N (∼398.5 eV), pyrrolic-N (∼400.1 eV), graphitic-N (∼401.2 eV), and oxidised-N (∼402–405 eV), each of which has a distinct role in metal-ion coordination, electrostatic interactions, and Fenton-type catalysis [43,46]. Raman spectroscopy provides complementary information on graphitic order through the D/G band intensity ratio: the D band (∼1350 cm⁻¹) reflects defect-induced disorder at pore edges and grain boundaries, while the G band (∼1580 cm⁻¹) corresponds to in-plane vibrations of sp²-hybridised carbon [47,48]. A high D/G ratio indicates a defect-rich surface with greater density of adsorption-active edge sites; a low D/G ratio reflects more ordered graphitic domains with stronger π–π stacking capacity towards aromatic pollutants. [47,48] Fourier-transform infrared spectroscopy (FTIR-ATR) provides semi-quantitative identification of surface functional groups: carboxyl (∼1710–1730 cm⁻¹), conjugated carbonyl (∼1600–1650 cm⁻¹), hydroxyl (broad, 3200–3500 cm⁻¹), and amine N-H stretching and bending bands (3300–3500 and 1550–1650 cm⁻¹, respectively) [44]. Temperature-programmed desorption (TPD) under inert gas provides quantitative data on the density and thermal stability of individual oxygen-containing groups: carboxylic acids release CO₂ at 200–300°C, lactones and anhydrides at 350–600°C, and phenol and carbonyl groups at higher temperatures, though these ranges overlap in practice and peak assignments should be treated as indicative [44]. Pore structure is characterised by N₂ physisorption at 77 K, with the BET equation applied to the multilayer range and non-local density functional theory (NLDFT) used for micropore size distribution analysis; pore widths below 2 nm make a dominant contribution to adsorption capacity for small organic molecules and metal ions [46]. Mercury intrusion porosimetry extends the accessible range into the macropore region (0.1–10 µm) that governs hydraulic permeability and pressure drop in filter modules [49].

These characterisation methods and the parameters they provide are summarised in Table 2.

3.2. Functionalisation Methods

4.1. Heavy Metals

Cu²⁺ Pb²⁺, Cd²⁺, Hg²⁺, Ni²⁺, and Zn²⁺ enter textile effluents mainly from mordanting baths, metallic dye processes, and finishing chemicals. Their uptake on carbonised fibres involves overlapping contributions from physical adsorption into micropores by van der Waals forces, ion exchange at carboxylate and phenolate surface groups (which becomes increasingly important above pH_pzc), and surface complexation at nitrogen-donor sites on N-doped or chitosan-coated fibres [32]. These relative contributions depend on pH, ionic strength, and the coordination chemistry of each metal. Langmuir q _max values for Cu²⁺ on chitosan-coated PAN-ACF nonwovens are around 185 mg g⁻¹ at pH 5.5, roughly 2.5 times the uncoated ACF value under the same conditions, which is consistent with amine-group chelation as the primary uptake route [32]. For Pb²⁺and Hg²⁺, which form stable N- and S-donor complexes, functionalised ACF fibres have shown 36-fold higher batch capacities than granular activated carbon at sub-10 mg L⁻¹ concentrations [62,71].

The limitation of most published work is that single-solute batch tests dominate the dataset, and competitive effects from real effluent matrices are rarely examined. Ca²⁺ and Mg²⁺from water hardness compete with heavy metals for amine and carboxylate sites, and in synthetic hard water (250 mg L⁻¹ CaCO₃ equivalent) effective capacity can be measurably lower than in deionised water, though the magnitude of this competitive effect is highly system dependent, and reported values vary widely across studies. Introducing higher-affinity groups – thiol (–SH) or imidazole – partially restores selectivity since these form thermodynamically preferred complexes with soft-acid Hg²⁺ and Cd²⁺ over competing Ca²⁺ [71].

4.2. Textile Dyes and Aromatic Amines

Textile effluents carry dyes from several structural classes – reactive, acid, basic, vat, and disperse – which differ enough in charge and molecular size that no single adsorption condition handles all of them well. Cationic dyes such as methylene blue (MB) and crystal violet adsorb readily on 16 oxidised ACF at neutral pH via electrostatic attraction to the negative surface and π–π stacking between the aromatic dye rings and graphitic basal planes [15]. For MB on KOH-activated PAN-ACFs, reported q _max values exceed 300 mg g⁻¹, the combination of high microporosity and a net negative surface charge at pH 7 evidently places these materials among the better-performing carbon adsorbents for this dye class [15]. Anionic reactive dyes are more difficult: sulphonate groups give the molecule a net negative charge, so adsorption on an equally negative ACF surface is electrostatically unfavourable above pH 4. Either pH reduction below pH_pzc, or positive surface charge introduced by PEI grafting or chitosan coating (amino groups protonated below pKa ∼ 6.3), is needed for effective uptake [27,72,73].

Many aromatic amines – including 4-aminobiphenyl, benzidine, and 2-naphthylamine – are classified as confirmed or probable human carcinogens under REACH Annex XVII, making their removal from effluent a regulatory requirement. On N-doped ACFs at near-neutral pH, adsorption proceeds mainly through π–π electron donor–acceptor interactions between the amine-bearing aromatic rings and the graphitic surface [43,74]. First-order kinetics with half-lives in the 10–25 min range have been reported for aniline and substituted aromatic amines on 16 oxidised and nitrogen-treated ACFs [43,74]; residual concentrations below 0.1 mg L⁻¹ are achievable from 50 mg L⁻¹ initial solutions, which falls below the EU Water Framework Directive indicative thresholds for priority hazardous substances [7].

4.3. Pharmaceuticals and Personal Care Products

Two entry routes bring PPCPs into textile effluents: wash-off from fibres treated with biocidal or UV-protective finishes, and co-discharge with municipal wastewater at shared treatment sites. Tetracycline antibiotics, present in surface waters at 0.1–10 µg L⁻¹, adsorb on PAN-derived ACFs mainly through π–π stacking and hydrogen bonding at the fused ring system, with electrostatic contributions below the pKa values of the zwitterion (pH 3.3 and 7.7) [13]. Ciprofloxacin, a fluoroquinolone, shows higher q _max on a chitosan-modified activated carbon composite than on the unmodified carbon, attributed to complexation of the quinolone carbonyl and carboxylate groups with surface amine ligands [75]. For NSAIDs such as ibuprofen and diclofenac, the optimum lies around pH 5–6, where the undissociated form predominates and hydrophobic partitioning into ACF micropores is most effective [65].

4.4. Microplastics

Regulatory pressure on microplastics in surface water and drinking water is increasing, and domestic laundering of synthetic textiles is now recognised as a major diffuse source. An estimated 0.2–0.5 g of microfibres per kilogram of fabric is released per wash cycle, reaching municipal wastewater systems in large quantities [76], with textile-industry effluent treatment plants specifically identified as a disproportionate contributor relative to municipal systems [77]. Particle removal by carbonised nonwoven filters operates differently depending on size: larger particles are retained by size exclusion and depth filtration at inter-fibre pores, while smaller particles (below about 10 µm) are captured mainly through mechanical straining within the internal pore network and hydrophobic interaction between the non-polar polymer surface and the carbon matrix. Because oxidised ACF surfaces and PE/PP/PS microplastics both carry a net negative charge under near-neutral conditions – the same charge mismatch that limits anionic-dye uptake in Section 4.2 – direct electrostatic attraction is unlikely to be the dominant mechanism at unmodified surfaces. Positively charged PEI- or chitosan-modified ACF surfaces, by contrast, can capture microplastics electrostatically, analogous to their effect on anionic dyes [78].

Activated-carbon-based nonwoven filters in column/fixed-bed configurations have been reported to remove more than 95% of micron-scale polystyrene microplastics [66]. Multi-stage carbon-based filtration trains, combining coarse and fine media, have been reported to improve overall microplastic capture beyond what a single stage achieves [67]. The particle load is split between stages, which also slows the pressure-drop increase that signals the need for regeneration. The published evidence base for microplastic removal by carbonised fibre filters is, however, narrow. Most studies have used model suspensions of monodisperse polystyrene or polyethylene microspheres in clean water matrices, which do not reflect the heterogeneous size distribution, irregular morphology, and complex surface chemistry of microplastics encountered in real textile effluents. The effect of co-occurring dissolved organics and surfactants on electrostatic capture efficiency has received little attention, and long-term fouling behaviour under continuous operation has not been systematically characterised. Because the values compiled below come from studies using different feed concentrations, contact times, adsorbent doses, and temperatures, Table 4 should be read as an indicative comparison of orders of magnitude rather than as a rigorous ranking of materials. Representative removal efficiencies for microplastics and other priority pollutants discussed in Sections 4.1–4.4 are summarised in Table 4.

5.1. Cascade Filtration Design Principles

Good cascade design starts with feed characterisation: flow rate, suspended-solids load, chemical oxygen demand (COD) fractionation, and pollutant speciation together determine the number of stages, their order, and the surface chemistry at each layer. Working from coarse to fine is the natural sequence. An initial stage with inter-fibre pore size 100–500 µm removes macro-particles, textile fibres, and microplastics by depth filtration and inertial impaction, protecting downstream stages from rapid clogging [67]. A second stage uses oxidised or chitosan-modified ACF nonwoven with inter-fibre void dimensions of 5–50 µm to ensure adequate contact time; the actual removal of heavy metals and cationic dyes occurs by adsorption within the internal carbon micro- and mesoporosity of the fibres, not by mechanical interception at the structural pore level. A polishing stage of N-doped ACF with BET area above 1500 m² g⁻¹ targets trace micropollutants, aromatic amines, and pharmaceuticals [10]. As a proposed design approach rather than a validated protocol, stage mass could be sized from isotherm data at the design outlet concentration with a safety factor of 1.5–2.0 on breakthrough capacity to account for competitive adsorption in real multicomponent streams; this factor is an engineering estimate and has not itself been reported in the cited studies. It should be noted that Langmuir q _max values derived from batch isotherms at 25°C are not directly transferable to continuous-flow module performance; dynamic breakthrough capacity, empty-bed contact time, superficial velocity, and regeneration loss per cycle are the operationally relevant parameters and differ substantially from batch-derived figures. Figure 4 illustrates a conceptual treatment train, proposed here rather than reported as a tested integrated system, combining the three ACF stages with upstream oxidation and downstream membrane polishing; the individual component performance values shown are drawn from separate literature reports [10,15,67,79–82], but the train as a whole has not been validated experimentally.

Figure 4

Nonwoven architecture gives design freedom that granular carbon beds do not. Fibre diameter (5–20 µm), basis weight (100–500 g m⁻²), and sheet stacking density can each be specified independently, so pressure drop, residence time, and contact area need not move together as they do in packed beds. Multi-layer stacks with graded pore structure – coarser at the inlet, finer downstream – can be formed into compact cylindrical or flat-sheet modules that fit standard industrial pressure vessels [15]. Hydraulic resistance increases approximately linearly with basis weight and decreases with mean fibre diameter, which are straightforward parameters for scale-up calculations.

5.2. Integration with Advanced Oxidation Processes

Ozonation, UV/H₂O₂, Fenton chemistry, and photocatalysis all produce hydroxyl radicals (·OH, E° = 2.80 V) capable of mineralising persistent organics that resist biological treatment and adsorption [10]. This pairing works in both directions: AOP pre-treatment converts large, non-polar dye molecules into smaller, more polar fragments that adsorb more readily on ACF, while the ACF stage concentrates residual organics and prolongs their contact with the oxidative environment [79]. ACF pre-adsorption of hydrophobic micropollutants followed by in-situ photocatalytic regeneration with TiO₂-ACF under UV can regenerate capacity without added reagents [53,54,84]. For textile effluents, ozonation at 5–15 mg L⁻¹ with 5–10 min contact time typically gives more than 90% decolourisation and 40–60% COD reduction, cutting the adsorptive load on the downstream ACF stage and extending service life by a factor of 3–5 compared with direct filtration of raw effluent [10].

Fenton and photo-Fenton chemistry (Fe²⁺/H₂O₂, pH 3–4) can mineralise persistent azo-dye degradation products more completely than ozone, but require precise pH control and produce iron sludge as a secondary waste. Fe₃O₄-functionalised ACF nonwovens used as heterogeneous Fenton catalysts shift the working pH to 5.5–6.5, removing the need for acidification and neutralisation steps and cutting sludge production [55]. Residual H₂O₂ must be quenched or decomposed before the water reaches downstream RO membranes since oxidant carry-over causes irreversible membrane damage.

5.3. Integration with Membrane and Biological Processes

Where the goal is effluent of sufficient quality for reuse in dyeing or rinsing, membrane polishing – MF, UF, NF, or RO depending on the required separation – is the logical final step. Their main operational vulnerability is fouling: colloidal organics, surfactants, scaling salts, and biofilms raise transmembrane pressure, reduce permeate flux, and shorten service life [80,85]. ACF pre-filtration removes the colloidal fraction by size exclusion (in pores below 0.1 µm), reduces dissolved organic carbon by adsorption, and strips out the most surface-active species that initiate gel-layer fouling, allowing NF and RO to run at lower pressure and with longer intervals between cleaning events [15]. Studies on pre-treatment of secondary effluents before membrane filtration have reported substantially lower normalised flux decline rates and extended CIP intervals compared with direct membrane operation [81], though the magnitude of benefit varies with feed composition and the specific ACF configuration used.

In a membrane bioreactor (MBR), activated sludge or biofilm treatment is combined with in-line MF or UF in a single vessel. The secondary clarifier is eliminated, and the permeate is already clarified to a standard suitable for NF or RO polishing. Placing carbonised ACF nonwoven stages inside the MBR aeration tank – as structured carriers that double as biofilm support and adsorbent – has been reported to buffer peak organic loading on the membrane module and to protect the biomass from inhibitory compounds such as heavy metals and biocidal dyes that would otherwise suppress biological activity [82]. Fill-draw SBR operation suits ACF filter integration well. During the quiescent decant phase, suspended biomass settles and the ACF stage can be run under gravity-driven flow with little risk of biomass fouling.

6.1. Regeneration and Fouling

ACF nonwovens cost roughly three to five times more per unit mass than granular activated carbon [11], which makes regeneration an economic necessity rather than an optional step. Three main routes exist.

Thermal treatment at 700–900°C in steam or CO₂ pyrolyses the adsorbed organic layer and restores more than 90% of initial capacity, but each cycle burns off 5–10% of the carbon mass at pore edges, so cumulative micropore loss is significant over time [36]. It is important to distinguish these two figures: the >90% recovery refers to the specific (mass normalised) adsorption capacity of the carbon that remains after regeneration, not to the total capacity of the filter element. Because 5–10% of the carbon mass is lost at each thermal cycle, the total capacity of the element declines cumulatively – roughly by a factor of (0.90–0.95)ⁿ after n cycles – even though the per-gram performance of the surviving material is largely preserved, a distinction long recognised for granular activated carbon, where progressively more make-up carbon must be added over successive cycles even as per-unit-weight capacity recovery remains high [86]. Service-life projections in Sections 6.2 and 6.4 should therefore be based on cumulative mass loss of the element, not on the per-cycle capacity-recovery figure alone.

Acid-alkali washing – sequential rinse with HCl or H₂SO₄ at pH 2 to strip metal ions by proton competition, followed by NaOH at pH 12 to restore surface charge – suits metal-laden stages and avoids high-temperature structural damage, though it produces concentrated eluates requiring further treatment [29]. In-situ photocatalytic regeneration of TiO₂-bearing ACF nonwovens under UV irradiation sidesteps both thermal cycling and chemical reagent use, but works only for compounds that ·OH can oxidise and requires optically transparent housings that trade off against mechanical robustness [53,54].

A critical gap in the published record concerns long-term stability under industrially representative conditions. To the authors’ knowledge, identified regeneration studies rarely extend beyond 20–30 cycles, and none have sustained real textile effluent as feed throughout; stable adsorption performance of functionalised ACF nonwovens across more than 50 regeneration cycles under industrially representative conditions has not been documented in the reviewed literature. Whether surface coatings – PEI grafts, crosslinked chitosan layers, sol-gel nanoparticle deposits – retain structural and chemical integrity through the repeated pH swings (pH 2–12) and elevated temperatures (up to 60°C) of industrial acid–alkali cleaning sequences remains unanswered. This is not a minor technical detail: without credible service-life data covering at least 50 cycles, an operator cannot project consumable costs or size a replacement inventory, which in practice means the technology cannot clear a basic economic feasibility review.

Fouling – contaminant accumulation that backwashing does not fully reverse – is the other main cause of performance decline in continuous-flow operation. In practice, it does not happen all at once. Surface-active organics and humic-like substances are taken up rapidly at primary adsorption sites within the first few hours of operation. Over the following days, colloidal particles and macromolecules work their way into mesopores. Microbial colonisation develops on the timescale of weeks, gradually blocking surface area and reducing permeability [87]. Nano-Ag or nano-Cu functionalisation limits the biological stage, but the earlier two phases call for hydraulic or chemical intervention. Backwash protocols that combine upward-flow water jets with air-scour restore 80–95% of initial permeability if applied at the first measurable rise in pressure, but their effect diminishes once a compressible colloidal cake has formed [88].

6.2. Scale-Up Challenges

Batch laboratory data systematically overpredict performance in pilot-scale continuous-flow systems, and the discrepancy is not always discussed honestly in the literature. Axial dispersion and channelling in multi-layer nonwoven stacks reduce effective contact time below the design value. Non-uniform flow distribution in large-area modules (above 1 m²) arises from pressure gradients in feed manifolds. Elevated operating temperatures (40–60°C, typical of textile process streams) shift adsorption equilibria and can degrade polymer coatings [5]. Pilot trials on real textile effluents consistently report capacities 30–50% below laboratory isotherm predictions, attributed to competitive adsorption from background dissolved organic matter and to kinetic limitations at industrial flow velocities [43]. The membrane sector has workable pilot-testing and techno-economic frameworks; similar tools for ACF cascade systems do not yet exist, and their absence is a practical obstacle to industrial adoption.

6.3. Circular Economy Perspectives

These materials fit within circular-economy logic on two separate counts. Using sorted cotton-polyester waste, PAN-based acrylic scrap, or crustacean-processing residues as carbonisation feedstocks converts end-of-life waste into functional filter media, avoiding both landfilling and the energy-intensive production of virgin activated carbon from coal or coconut shell [28]. LCA comparisons generally show lower global warming potential and cumulative energy demand for waste-derived carbons than for coal-based GAC [89], though the margin narrows when feedstock sorting and pre-treatment are energy-intensive. The environmental benefit case, however, rests on the assumption of controlled feedstock composition. Post-consumer textile waste arrives as polydisperse blends of unknown fibre-type ratios, variable dyestuff loads, and residual finishing chemicals; each of these parameters shifts the optimal carbonisation temperature and the surface chemistry of the product. Without upstream sorting to textile-grade homogeneity – a step that is costly and not yet widely available at collection scale – the environmental and performance claims made for waste-derived ACF feedstocks should be treated with caution. On the water side, treating process effluent to reuse quality closes or partially closes the production water loop. A plant treating 300 m³ day⁻¹ with 80% water recovery, for instance, would avoid abstracting roughly 240 m³ day⁻¹ of freshwater and discharging the equivalent volume as polluted effluent – a benefit that grows as water tariffs and discharge fees rise.

A dedicated economic assessment of this pathway is largely absent from the literature and remains an important gap. Techno-economic feasibility of textile-waste-derived activated carbon is explicitly flagged as an open research need [90], and cost data specific to carbon fibre production from textile-waste feedstocks are only beginning to emerge: a recent techno-economic analysis of carbon fibre made from recycled acrylic textile waste reported approximately 20% lower production cost than virgin-precursor carbon fibre at 50% recycled content, alongside a roughly 40% reduction in cradle-to-gate greenhouse-gas emissions [91]. These figures concern fibre production rather than the full cascade filtration system discussed here, and do not include downstream costs such as activation, module fabrication, regeneration, or comparison against conventional treatment trains (coagulation, granular activated carbon, membrane processes). A dedicated techno-economic and comparative cost analysis of carbonised-fibre cascade filtration against these established alternatives is therefore identified as a priority for future work before the concept proposed here could be assessed for practical industrial adoption.

The spent adsorbent from chemical regeneration cycles carries concentrated heavy metals, dye fragments, and degradation products that themselves require management. Thermal regeneration recovers the carbon for reuse and concentrates inorganics in the ash residue, which may be worth processing by hydrometallurgical routes for copper, lead, or zinc recovery if concentrations are high enough [92]. Alkali eluates from chitosan-coated stages tend to be ammonium-rich; these can in principle be treated by nitrification-denitrification before discharge, integrating the regeneration circuit into the plant’s broader wastewater management.

6.4. Conclusions and Future Perspectives

The picture that emerges from this review is uneven: the materials science is well-developed, but the path to industrial deployment remains obstructed by gaps in long-term performance data and the absence of standardised evaluation tools.

• For continuous-flow ACF nonwovens, PAN continues to give the most predictable combination of surface area, microporosity, and mechanical strength [14,16,24]. Cellulose fibres, chitosan fibres, and lignocellulosic materials carry useful surface chemistries for specific contaminant classes and have real cost and environmental advantages, but carbon yield and structural reproducibility from these precursors remain below what PAN delivers [27,30,68].

• As-carbonised fibres without modification cover a narrow contaminant range. Oxidative treatment, polymer grafting, nanoparticle deposition, and nitrogen doping each extend performance into different areas of the pollutant spectrum [44,51,58]. Whether complex composite coatings hold up through many regeneration cycles is a practical question that the literature has not yet answered convincingly [29,36]. Identified cycle tests typically stop at 10–20 iterations using synthetic single-contaminant feeds, which is an insufficient basis for projecting operational service life in an industrial setting.

• Multi-stage ACF cascades can reach the effluent quality needed for process-water reuse, but only when combined with upstream oxidation and downstream membrane polishing [10,15,80]. No single stage achieves this on its own and claims to that effect in the literature should be read critically.

• Industrial uptake remains slow. ACF nonwovens cost 3–5 times more than granular carbon [11], real-effluent performance falls 30–50% short of laboratory isotherm projections [5,43], and there are no standardised testing or techno-economic tools that would let an end-user build a credible business case. Until those tools exist, the technology is unlikely to move beyond pilot demonstrations at scale.

• Rising virgin raw-material costs and increasing discharge charges are gradually improving the economics of waste-derived filter media and closed production water loops, without requiring any change in the regulatory framework [28,89].

Three gaps stand out as priorities for future work: continuous carbonisation and activation processes suitable for variable-composition waste-textile feedstocks at production scale [34]; systematic characterisation of competitive adsorption and fouling kinetics under realistic, dynamic multicomponent flow conditions, since most published data still rely on idealised single-contaminant batches [5,43]; and accelerated ageing protocols capable of supporting credible service-life predictions and regulatory acceptance [29,36].