Introduction
Eutrophication is a persistent and globally distributed water-quality problem driven by the accumulation of nitrogen species in surface waters [1]. In aquatic systems at near-neutral pH, ammonium (NH4+) is the dominant inorganic-nitrogen species and the entry point for the nitrogen cycle through nitrification [2]. Anthropogenic inputs—industrial discharge, agricultural runoff, livestock effluent, and municipal wastewater—routinely enrich natural waters with NH4+ above the assimilation capacity of receiving ecosystems [3,4]. The ensuing algal blooms and their decomposition draw dissolved oxygen below the thresholds required by aquatic biota, producing hypoxic and anoxic zones that degrade biodiversity and compromise downstream water uses [5,6,7]. Controlling the NH4+ load that reaches receiving waters is therefore a precondition for sustainable water-quality management, and is increasingly a regulatory target for decentralised and on-site wastewater handling.
Existing NH4+ removal technologies fall into three broad classes. Physical approaches include air stripping, ion exchange, membrane separations, capacitive deionisation, and membrane bioreactors [8]. Chemical approaches include struvite precipitation, electrochemical oxidation, and electrocoagulation [9]. Biological approaches encompass microalgal uptake, the anammox pathway, bioelectrochemical systems, and treatment with photosynthetic bacteria [3]. Each class carries well-documented limitations: sensitivity to pH and temperature fluctuations, slow kinetics relative to industrially acceptable residence times, secondary waste or brine streams, and specialised microbial consortia that are not robust to real-world feed variability [3]. Among these, adsorption remains attractive for decentralised and point-of-use treatment because the unit operation is simple to implement, tolerant to feed composition variability, and readily engineered into passive filtration geometries.
Natural zeolites, particularly those of the clinoptilolite type, have a long-established capacity for NH4+ uptake through cation exchange on the aluminosilicate framework [10,11,12], and their high cation-exchange capacity makes them competitive with synthetic adsorbents for a wide range of water-treatment and environmental applications [13,14,15]. Prior work from our group demonstrated that a desilicated natural zeolite from Lampung (Indonesia) reaches NH4+ adsorption capacities of 22 mg g−1 from aqueous solution [16], and independent column studies report 60 %–80 % removal at optimal operating conditions for zeolite beds packed with treated municipal effluent [17,18]. Loose-powder deployment, however, has three practical limitations in flow-through applications: particles require downstream separation, the adsorbent bed introduces pressure drop proportional to its depth, and fine fractions migrate into the treated effluent. Integrating the adsorbent into an engineered membrane format can address all three limitations simultaneously. Electrospun polyacrylonitrile (PAN) nanofiber mats are a versatile platform for this purpose: their high surface-area-to-volume ratio, tensile robustness, and compatibility with common adsorbent slurries make them a natural host for cation-exchange particles [19,20,21,22], and hybrid nanofiber architectures have been developed across a wide range of water-treatment targets [23,24]. Prior exemplars combining zeolite with PAN- or PVDF-based matrices span from recyclable catalysis [25] to proton-conducting composite membranes [26], silica-modified aerogel fibres for coloured wastewater [27], and more recent antibacterial and air-filter designs [28]. Our group has recently demonstrated a PAN/PVDF–zeolite composite platform for efficient dye removal from aqueous streams [29,30], but its application to nutrient-species capture—specifically NH4+—has not been explored.
Two conceptually different methods can be used to integrate particulate adsorbent into an electrospun polymer matrix. In the dispersion method, particles are deposited on the upstream surface of a preformed nanofiber substrate by vacuum-assisted coating. In the electrospinning-integration method, particles are suspended in the polymer dope prior to spinning and are encapsulated inside the fibres during jet elongation. The two routes are not interchangeable: dispersion preserves unrestricted access to the particle surface but relies on relatively weak mechanical attachment, while electrospinning-integration stabilises the particle within the fibre but may bury a large fraction of the adsorption-active surface inside the polymer. Which route is preferable depends sensitively on the adsorbate, the filtration geometry, and the mass-transport regime. For dye removal from aqueous streams both routes have been reported to work in our lab’s earlier studies [29] [30]. For NH4+ capture, where the active step is ion exchange at exposed aluminosilicate sites rather than diffusion of a bulky organic dye through the matrix, the methodology question has not been systematically resolved, and the outcome is not obvious from the existing dye-removal literature.
Here, we address that gap. We fabricate two parallel series of zeolite-PAN composite nanofiber membranes—Ze-Dx by surface dispersion of natural zeolite onto a preformed PAN/PVDF substrate, and Ze-ESx by electrospinning-integration of zeolite into a PAN dope, where x denotes the zeolite mass used in the batch. All membranes are characterised by Fourier-transform infrared spectroscopy to verify component incorporation, by scanning electron microscopy to image particle distribution across the fibre network, and by water contact angle to quantify the functional hydrophilicity of each surface. NH4+ removal performance is then evaluated in a dead-end vacuum filtration geometry, first as a screening across all formulations at fixed feed concentration and then as a concentration- and reusability-dependence study on the best-performing sample. We show that the two fabrication routes produce qualitatively different surface-availability regimes, that the dispersion route at the highest zeolite loading (Ze-D0.1) alone delivers sustained near-quantitative NH4+ removal at low feed concentrations with good reusability, and that the electrospinning-integration route fails outright due to progressive zeolite leaching during flow. These findings extend our group’s prior Ze-PAN/PVDF dye-removal platform to nutrient capture and resolve the method-selection question, for the specific case of NH4+ removal, in favour of surface dispersion.
Materials and Methods
Materials
Polyacrylonitrile (PAN,
Membrane fabrication
Two complementary routes for integrating zeolite into the PAN nanofiber matrix were compared (Fig. 1): (i) direct electrospinning-integration (ES), in which zeolite particles were blended into the PAN dope prior to spinning, and (ii) surface dispersion of zeolite particles onto a preformed PAN/PVDF substrate. The resulting membranes are denoted Ze-ESx and Ze-Dx, respectively, where x is the zeolite mass (in grams) used per batch.
PAN dope solution
A PAN dope was prepared by dissolving 0.6 g of PAN in 10 mL of DMF in a beaker inside a fume hood. The beaker was sealed with parafilm to prevent solvent evaporation and stirred on a magnetic stirrer at 500 rpm and 60 °C for 2 h until a homogeneous, viscous solution was obtained. For the electrospinning-integration method, zeolite (0.3 g or 0.5 g, yielding samples Ze-ES0.3 and Ze-ES0.5, respectively) was then added stepwise under continued stirring until the particulate phase was visually uniformly suspended.
Electrospinning
Nanofiber membranes were fabricated using a single-needle electrospinning setup. A grounded iron collector plate was covered with aluminium foil to facilitate peel-off of the deposited fibres. The dope solution was loaded into a syringe mounted on a syringe pump, and the needle tip-to-collector distance, applied voltage, and feed rate were set to fixed values across all experiments. An applied voltage of 9 kV and a feed rate of 0.5 μm s−1 were used throughout. Electrospinning continued until the syringe was fully discharged, after which the voltage was ramped down stepwise to 0 kV to avoid corona damage to the freshly deposited fibres. The collected nanofiber mat was peeled from the aluminium foil and stored in a sealed bag prior to further use.
Surface dispersion of zeolite on PAN/PVDF
For the Ze-Dx series, zeolite (0.01 g, 0.05 g, & 0.1 g, yielding Ze-D0.01, Ze-D0.05, and Ze-D0.1, respectively) was suspended in 20 mL of deionized water and stirred at 500 rpm for 10 min, then ultrasonicated for 30 min to break up agglomerates and improve dispersion stability. The resulting suspension was poured onto a commercial PAN/PVDF support membrane mounted in a vacuum filter holder; the liquid phase was drawn through the membrane under vacuum and discarded, leaving the zeolite particles deposited on the upstream membrane surface. The zeolite-loaded membrane was used directly in subsequent NH4+ removal tests without any drying or post-treatment.
Membrane characterization
The fabricated membranes and the bare natural zeolite reference were characterised by three complementary techniques spanning chemical composition, morphology, and surface wettability. Fourier-transform infrared spectroscopy (FTIR) was used to probe the chemical signature of each component over the range 400 cm–4000 cm, using the C≡N (nitrile) and CH stretches of PAN, the CF stretch of PVDF, and the SiOSi / AlOSi framework vibrations together with OH stretches of the aluminosilicate zeolite as diagnostic bands. Surface morphology was examined by scanning electron microscopy (SEM) at four magnifications (1000×, 3000×, 5000×, and 10 000×) to resolve features from the nanofiber-mat scale down to individual zeolite particles, allowing assessment of nanofiber uniformity, zeolite particle distribution on or within the matrix, and any aggregation or dead-zone formation that could limit mass transport. Surface wettability was quantified by static sessile-drop water contact angle (WCA) measurements: a microlitre-scale droplet of deionized water was dispensed onto the membrane surface and imaged from the side with the camera axis horizontal to the substrate; the contact angle was extracted in ImageJ using the Drop Snake plug-in, which fits a polynomial to the drop profile and reports left and right contact angles independently [31]. Five droplets were measured per sample at distinct positions on a homogeneous region, and the per-sample WCA is reported as the mean of the five droplet-averaged values with the spread expressed as one standard deviation.
Ammonium quantification and calibration
NH4+ concentration was measured by UV-Vis
spectrophotometry using the Nessler method. To 5 mL of sample,
0.1 mL of Nessler’s reagent was added; the mixture was shaken
briefly and incubated for 10 min at room temperature to allow the
yellow-orange
NH4+−Nessler
chromophore to develop. Absorbance was then measured at
Dead-end vacuum filtration tests
Ammonium removal performance was evaluated in a batch dead-end
vacuum filtration configuration. A disk of the test membrane with
effective filtration area
Two series of experiments were conducted. In the
screening series, all five membrane formulations
(Ze-D0.01, Ze-D0.05, Ze-D0.1, Ze-ES0.3, Ze-ES0.5) plus the
unmodified PAN/PVDF control were tested at a single feed
concentration of
Removal efficiency (Eq. (1)) and permeate flux (Eq. (2)) were calculated for each cycle:
where
Results and Discussion
Membrane characterization
The membranes were characterized by FTIR to verify successful incorporation of zeolite into the PAN matrix, by SEM to visualize how each fabrication route distributes the zeolite across the membrane, and by water contact angle to translate those morphological differences into a functional surface property. Together the three techniques build a consistent picture: the dispersion route delivers zeolite at the membrane surface, while the electrospinning-integration route leaves a substantial fraction of the zeolite buried within, or aggregated between, the PAN fibres.
Fourier-transform infrared spectroscopy
Figure 2 shows the FTIR spectra of the PAN/PVDF control, the dispersion-method sample Ze-D0.1, and the two electrospinning-integration samples Ze-ES0.3 and Ze-ES0.5, together with the natural zeolite reference. Each component has diagnostic bands used to probe its presence at the membrane surface: the nitrile C≡N stretch of PAN near 2240 cm, the aliphatic CH stretch near 2930 cm, the CF stretch of PVDF between 1200 cm and 1400 cm, and the framework SiOSi and AlOSi vibrations of the aluminosilicate zeolite between 950 cm and 1050 cm. A broad OH stretch near 3450 cm and its bending partner near 1650 cm indicate water adsorbed on the zeolite framework [29,30].
For the dispersion-method sample Ze-D0.1 (Fig. 2a, c), the zeolite framework band at 1050 cm is the dominant spectral feature, consistent with zeolite particles exposed directly at the upstream surface of the membrane and strongly attenuating probe-beam transmission at aluminosilicate frequencies. The characteristic PAN/PVDF polymer bands are correspondingly weaker: zeolite dominates the surface-sensitive FTIR signal even though the underlying polymer support is unchanged. The broad OH feature is also stronger than in the other samples, because water is easily adsorbed onto the hydrophilic zeolite surface during sample handling. These observations are a direct optical confirmation that the dispersion route deposits zeolite on—rather than within—the support membrane.
For the electrospinning-integration samples Ze-ES0.3 and Ze-ES0.5 (Fig. 2b), both the PAN C≡N band (intact) and the zeolite SiOSi band are observed. Higher zeolite loading (Ze-ES0.5) intensifies the zeolite feature only modestly, far less than the doubled loading would suggest. This is consistent with a substantial fraction of the zeolite being encapsulated within the PAN fibre matrix after spinning, where it is less accessible to the infrared beam than the surface-deposited zeolite of Ze-D0.1. The preservation of the PAN nitrile feature in both ES samples confirms that the electrospinning step does not chemically modify the polymer.
Scanning electron microscopy
SEM imaging (Fig. 3)
provides complementary morphological evidence for how each
fabrication route distributes zeolite within the membrane. Four
magnifications were used (
In the dispersion series (Fig. 3a), the coverage density of zeolite on the PAN/PVDF support increases monotonically with the added zeolite mass. At the lowest loading, Ze-D0.01, the surface is only partially covered and gaps expose bare nanofiber substrate, so the adsorption-active area is inherently limited. At the intermediate loading Ze-D0.05, coverage is more uniform but residual agglomeration clusters remain visible. At the highest tested loading, Ze-D0.1, zeolite is uniformly distributed across the support surface with minimal agglomeration, producing the densest and most morphologically homogeneous interface of the three.
In the electrospinning-integration series
(Fig. 3b), the PAN
nanofiber network itself is well-formed at both loadings, but the
zeolite is clearly displaced from the fibre interior: particles
decorate fibre surfaces and interstitial spaces rather than being
uniformly encapsulated within the fibres. For Ze-ES0.3,
aggregation is evident at intermediate magnifications
(
Water contact angle
Water contact angle (WCA) translates the surface chemistry and morphology inferred from FTIR and SEM into a single macroscopic wettability number per sample. Figure 4 collects the measurements across the five formulations.
The unmodified PAN/PVDF control is moderately hydrophobic, with WCA =(82.99±0.87)∘, consistent with the low-polarity CF groups of PVDF dominating the accessible surface. Pure electrospun PAN is less hydrophobic, (71.41±0.55)∘, reflecting the polar nitrile group in the PAN backbone. Both zeolite-functionalized formulations are substantially more hydrophilic: Ze-D0.1 shows WCA =(32.77±4.09)∘; the electrospinning-integration samples Ze-ES0.3 and Ze-ES0.5 show (35.52±6.64)∘ and (24.11±2.91)∘, respectively. The shift of more than 50∘ between the PAN/PVDF control and the zeolite-bearing samples is consistent with the high cation-exchange capacity of clinoptilolite, whose exposed aluminosilicate surface readily hydrogen-bonds with water [35,36]. The monotonic decrease of WCA as zeolite loading increases confirms that the surface trends inferred spectroscopically (Fig. 2) and morphologically (Fig. 3) translate to a functional wetting property.
Two secondary features of the WCA data are worth noting. First,
the standard deviation on Ze-ES0.3 (
Ammonium removal performance
Screening across membrane formulations
The six tested formulations (Ze-D0.01, Ze-D0.05, Ze-D0.1,
Ze-ES0.3, Ze-ES0.5, and the unmodified PAN/PVDF control) were
compared at a feed concentration of
The dispersion-method samples form a clear performance rank that tracks the zeolite loading. At the highest loading, Ze-D0.1 approaches quantitative removal in the first two cycles and retains 47 % removal by cycle 5, consistent with the densest zeolite coverage seen by SEM (Fig. 3a). At intermediate loading, Ze-D0.05 begins at 91 % and degrades steeply to 22 % by cycle 5: coverage is sufficient for first-cycle capture but saturates rapidly. The lowest-loading sample Ze-D0.01 shows a more gradual decline, from 66 % to 35 % across cycles 1–5, but never matches the first-cycle efficiency of the higher-loaded samples, confirming that sparse zeolite coverage limits even the initial removal. The monotonic Ze-Dx ordering by loading is in full agreement with the morphological picture from SEM.
The two electrospinning-integration samples behave qualitatively differently. Ze-ES0.3 starts at 12 % removal in cycle 1, falls to 10 % in cycle 2, 3 % in cycle 3, and reaches effectively zero removal at cycles 4–5. Ze-ES0.5 is even worse: 1.7 % in cycle 1 and zero thereafter. At the cycles reported as zero, the raw Nessler-reported NH4+ concentration in the filtrate in fact matches or exceeds the feed concentration; since a membrane cannot produce more NH4+ than it receives, the apparent super-feed reading is clipped to zero in Fig. 6 and interpreted as no net removal plus an additive background. That background is consistent with zeolite leaching from the PAN matrix during filtration: loosely-held particles detach under vacuum-driven flow, co-elute with NH4+, and release adsorbed cations, inflating the apparent residual concentration. The SEM images of the Ze-ES samples (Fig. 3b) support this mechanism directly: the ES formulation fails to stably embed zeolite in the PAN fibre matrix, so the weakly-held particles shed mechanically during filtration.
The PAN/PVDF control maintains removal efficiency within noise of zero (never exceeding 12 % in any cycle, with several cycles reading at or near zero), with no monotonic trend. This confirms the intended baseline: the polymer matrix itself does not adsorb NH4+ meaningfully, so any removal observed in the Ze-D or Ze-ES series is attributable to the zeolite component rather than to the polymer.
Permeate flux data (Fig. 5) provide an independent, mechanistically informative view of the same screening runs and reinforce the removal-efficiency picture described above. Three features are most diagnostic. First, across the dispersion series, cycle-1 flux decreases monotonically as zeolite loading increases: \sim 7400 L m2 h−1 for Ze-D0.01, \sim 110 L m2 h−1 for Ze-D0.05, and \sim 60 L m2 h−1 for Ze-D0.1, spanning more than two orders of magnitude. The inverse scaling of flux with loading directly confirms that the dispersed zeolite forms a resistive surface layer on the support rather than simply resting on it [9]. Second, the Ze-D0.1 flux is essentially constant across all five cycles (60.8 L m2 h−1 in cycle 1 to 54.2 L m2 h−1 in cycle 5, a decline of only \sim 11 %), indicating that the zeolite coating is mechanically stable under repeated vacuum-driven flow. Third, and most diagnostically, the Ze-ES0.5 flux increases from 3870 L m2 h−1 in cycle 1 to 11265 L m2 h−1 in cycle 5—an almost threefold rise over the same five cycles in which its removal efficiency falls to zero. A rising flux with a falling removal can be explained only by a progressive loss of active adsorbent from the flow path: as loosely-held zeolite detaches cycle by cycle, the hydraulic resistance of the membrane drops while the NH4+ capture surface is simultaneously lost. Flux and removal thus corroborate each other on the leaching mechanism already inferred from SEM (Fig. 3b), and the contrast with the mechanically stable Ze-D0.1 flux profile establishes dispersion as the route of choice for durable NH4+ capture. The PAN/PVDF control follows an orthogonal trajectory—a two-orders-of-magnitude flux collapse (251000 L m2 h−1 down to 2200 L m2 h−1)—driven by cumulative compaction and fouling of the bare polymer matrix rather than by any adsorbent dynamics.
Taken together, the screening identifies Ze-D0.1 as the only formulation that combines high first-cycle efficiency with sustained performance across the five-cycle test [29]. It was therefore selected for the concentration-dependence study described next.
Concentration-dependence and reusability of the selected formulation
Ze-D0.1 was further evaluated at four feed concentrations
(
In the low-concentration regime (
At intermediate and high concentrations the expected
saturation-plus-fouling pattern emerges. At
Three quantitative caveats are consolidated here. First,
nominal removal values above 100 % at low and moderate
From an operational standpoint, these results define an envelope for the Ze-D0.1 formulation: single-pass removal of NH4+ at feed concentrations at or below 2 mg L−1 is essentially quantitative and reusable over at least five cycles, while at concentrations approaching or exceeding 10 mg L−1 a single membrane saturates within 2–3 cycles and must be regenerated or replaced. Although the per-membrane NH4+ capacity of the Ze-D0.1 formulation is intrinsically lower than that of loose-powder natural zeolite on a per-gram basis (∼22 mg/g for desilicated Lampung zeolite) [16], the integrated-membrane format eliminates the downstream separation step and enables direct incorporation into flow-through treatment units. This positions Ze-D0.1 most naturally as a polishing step on already-low-concentration streams (secondary wastewater effluent, aquaculture recirculation, environmental monitoring sampling lines) rather than as a bulk-removal unit on concentrated feeds.
Conclusions
Two zeolite-PAN composite nanofiber membranes were fabricated via
contrasting routes—surface dispersion (Ze-D series) and
electrospinning-integration (Ze-ES series)—and compared for
NH4+ removal by dead-end vacuum filtration
across five consecutive cycles. FTIR, SEM, and water contact angle
measurements established that the dispersion route places zeolite at
the upstream membrane surface, producing a dense and strongly
hydrophilic interface (WCA (32.8±4.1)∘
for Ze-D0.1 vs. (83.0±0.9)∘
for the unmodified PAN/PVDF control), whereas the
electrospinning-integration route encapsulates a substantial fraction
of the adsorbent inside the fibre matrix with progressive aggregation
at higher loading. The best-performing formulation, Ze-D0.1, delivered
near-quantitative NH4+ removal in the first two
cycles at