Introduction

A very small proportion of water on Earth is fresh water [1], with only 0.5–0.75% of the total amount of water constituting fresh groundwater and soil moisture whilst the vast majority corresponds to oceans or salty groundwater [2]. Unfortunately, this small amount of freshwater is further depleted owing to pollution by various pollutants [3], considering that about 80% of all wastewater worldwide is discharged into the environment without adequate treatment. This poses a serious threat to ecosystems and can have detrimental effects on human health as well as the quality of ambient freshwater resources for polluted water contains harmful industrial/agricultural wastes, heavy metals, poisonous textile dyes, pesticides, or oil [4,5,6]. The majority of water pollution, including heavy metal (HM) ions, is caused by human activities, driven by industrialisation, population growth, urbanisation, climate change, natural disasters and agricultural operations [7, 8]. Due to their toxicity and longevity in the biosphere, heavy metals are particularly problematic when present in fresh water sources [9] as high concentrations of HM ions can cause severe health problems for both animals and humans [10].

Numerous sectors, including mining, paper, petroleum, and metal industries discharge high volumes of effluents containing Cu(II) ions into the environment [11], making copper one of the most frequently found heavy metal pollutants in the environment, particularly in water [12, 13]. Currently, ion-exchange resins and polymer or ceramic filters and membranes are commonly used as effective tools and technologies for removing HM ions from water [14]. Although these approaches exhibit indeed promising HM adsorption capacities, each of these methodologies has disadvantages such as high energy requirements for manufacture, their synthetic material content and/or the generation of toxic sludge [15]. Thus, an efficient, economically, and environmentally sustainable technology needs to be developed to address these important issues with water treatment approaches [6].

Adsorption based on natural and renewable materials is one of the most cutting-edge approaches for water purification, particularly for removing heavy metal ions, due to its convenience, selective uptake [16], and cost efficiency, especially for low-level heavy metal treatment in water [17]. Amongst renewable materials, cellulose nanofibers (CNF) were shown to be potent absorbent materials for their high specific surface area [18] and excellent chemical and mechanical properties [19]. Moreover, CNF can be extracted from a broad variety of biomass such as trees, plants, and weeds but also agricultural side streams [20]. Thanks to its high number of surface hydroxyl groups, CNF can be modified for selective and enhanced interactions using a variety of chemical reactions, such as carboxylation utilising 2,2,6,6-tetramethylpiperidin-1-oxyl (TEMPO) [21]. TEMPO mediated oxidation is applied as an effective method to selectively convert hydroxyl groups on the surface of cellulose fibres located at the C6 position of cellulose into carboxylate groups [22]. The negatively charged functional carboxylic (COO−) groups exhibit significant attraction to a number of HM ions, including copper ions [23]. Cellulose nanofibrils generated therefrom are commonly denominated as TEMPO-CNF [24, 25] and have potential applications as an environmentally friendly and bio-based nanomaterial in a variety of fields including water treatment [26]. By utilising coating technologies, i.e. foam and cast coating, TEMPO-CNF can be attached to the surface of commercial filters with various coat weights in laboratory scale [14, 24], facilitating continuous adsorption during filtration processes, which would allow for an efficient process at relevant scale.

The aim of this study was to investigate foam and cast coating processes in pilot scale for depositing TEMPO-CNF on renewable filter substrates with varying coat weights for the removal of HM ions from water. The influence of the different coating processes and coat weights on various filter characteristics (mechanical properties, thickness and shrinkage) was investigated and the performance of these filters was evaluated in terms of their pure water permeance and copper ion adsorption ability. Furthermore, the capability to regenerate copper loaded filters was demonstrated and the re-adsorption capacity of the regenerated filters was investigated.

Results and discussion

Foam-coated TEMPO-CNF filters

TEMPO-CNF-coated viscose filters for the removal of heavy metal ions from water were fabricated by pilot-scale foam and cast coating. TEMPO-CNF was the active adsorption agent on the viscose filter surface, exhibiting COOH groups which facilitate capture of (heavy) metal ions from water. To maintain both high flux and reasonable adsorption capacity, the amount of TEMPO-CNF coating needed to be controlled. Moreover, to generate filters in a cost-efficient manner, filters with low coat weights were fabricated. Ideally, the substrate should be coated in a way that individual viscose fibres are coated with a thin layer of TEMPO-CNF. This assembly allows for a maximum effective surface area available for metal ion capturing. This required the starting TEMPO-CNF suspension to have a low solid concentration and yet a large enough volume to evenly cover the viscose filter during coating. However, direct dilution of a commercial TEMPO-CNF is not ideal, as the increased amount of water requires increased energy consumption for drying. As such, a foamed TEMPO-CNF suspension, stabilised by an anionic surfactant, was formulated and used as starting material for the coating. The type of the substrate and surfactant as well as surfactant concentration for foaming TEMPO-CNF have been optimised in a previous study [14]. The dry coat weight of TEMPO-CNF on viscose filters was set between 1.2 and 5.1 g m−2. During roll-to-roll foam coating and the subsequent drying process, the viscose filters were under tension in machine direction, retaining the dimensions of the filters. However, in the transverse direction, the filters were not affixed, and thus under tension resulting in shrinkage upon drying. Thus, the real coat weight was between 1.3 and 6.8 g m−2 (Table 1). The shrinkage increased with the set coat weight and was approximately 20% for most samples. Higher coat weight is connected to higher water content during drying and evaporating high amounts of water causes stress in the transverse fibre direction, thus resulting in significant shrinkage. Hence, a trend to higher shrinkage at higher coat weight was observed. However, for all coat weights, but in particular for low coat weights, a reasonable agreement between set and real coat weights was achieved.

Table 1 Set and real coat weights as well as shrinkage of foam coated filters
Full size table

Samples F 1–4 with low and moderate coat weights exhibited a surface morphology similar to the uncoated substrate. This was explained by the coat weights being too low to yield a “thick”, continuous surface layer of TEMPO-CNF. On the other hand, samples F 5–6 exhibited a dense but non-continuous layer, caused by their higher coat weights compared to F 1–4. Nevertheless, the dense coating layer did not fully cover the top surface of the viscose substrate, leaving many pores (Fig. 1), which are anticipated to be beneficial for rapid water passage. Micrographs of all foam-coated filters are shown in the supporting information (Fig. S1).

Fig. 1
figure 1

SEM images of the surface of foam coated TEMPO-CNF filters at 50 × magnification: a Sample F 1, b Sample F 6

Full size image

Pure water permeance of filters

The water permeance of uncoated and foam-coated filters was measured using a dead-end cell at various pressures (0.2, 0.5, and 1 bar). The relationship between the pure water permeance and real coat weight of foam coated filters is shown in Fig. 2. All filters permitted for high flux as required in water filtration application, even though the permeance slightly decreased with increasing coat weight. The insignificant decrease of the permeance with increasing coat weight was because also higher coat weights did not yield a complete continuous coating layer on the surface of the filters (also for sample F 6) [6, 14], allowing for water bypass through the filters at a high flux. Furthermore, the water permeance was reduced with increased pressure. This was explained by the compaction of the filters at higher pressure, resulting in a denser filter, whose water permeance was reduced. The stable water permeance indeed outperformed pure cellulose nanopapers and NC coatings [6, 14, 27,28,29]. It is noted that no deterioration of neither the substrate nor the coating was observed and no mass loss or loss of fibre (fragments) could be detected in the permeate at the pressures applied.

Fig. 2
figure 2

Water permeance vs. real coat weight of foam coated filters

Full size image

Adsorption capacity for Cu(II) ions

2.5 mM CuCl2 solutions, corresponding to strongly copper-contaminated water, were applied as feed solution to investigate the adsorption capacity of the foam-coated filters. The Cu(II) concentrations in the feed and filtered solutions were determined. After filtration, the foam-coated TEMPO-CNF filters were coloured greenish, indicating the adsorption of Cu(II) ions onto the filter surface. The adsorption per unit area represents the areic adsorption of the filters. For higher coat weights (6.8 g m−2), the adsorption per unit area was increased from 25 mg m−2 (for 1.3 g m−2) to 128 mg m−2 (Fig. 3). This was explained by the overall higher number of carboxyl groups (COO−) attached to the filter as the coat weight increased [13, 25].

Fig. 3
figure 3

Areic adsorption and adsorption capacity of copper ions vs. real coat weight of foam coated filters

Full size image

The adsorption capacity represents the adsorption per unit mass of TEMPO-CNF coating, allowing for consideration of the efficiency of the coating. The adsorption capacity increased from 19 to 52 mg g−1 with decreasing coat weight (from 6.8 to 1.3 g m−2), indicating that the adsorption efficiency was improved at lower coat weight (Fig. 3). This was explained by establishing thicker TEMPO-CNF coating layers with increasing coat weight. Considering that adsorption primarily occurs at the surface of TEMPO-CNF layers [11, 30], increasing the coat weight, i.e. coating layer thickness, does not provide a higher number of COO− sites on or near the surface but rather buries them in the bulk of the coating. Thus, the adsorption capacity and coat weight were indirectly proportional. Taking into account an atomic mass of copper of 63.55 mg mmol−1, the molar adsorption capacity is 0.82 mmol g−1. Thus, in this case, the ratio between carboxyl-groups and copper is 1.34. This ratio was found for the thinnest coatings for which apparently a majority of the carboxyl groups were available for adsorption and was accordingly higher for lower adsorption capacities found at higher coating grammages. Adsorption capacities were in the range of pure anionic nanocellulose [13, 31] papers [11] but with the advantage of higher permeance, almost achieving adsorption capacities as in static batch-wise experiments [32].

The adsorption capacity of TEMPO-CNF filters was also investigated for Ca(II) ions, commonly present in fresh water sources, from aqueous solution by filtration experiments using a test solution of 2.5 mM CaCl2. Similar to copper ions, the highest amount of Ca(II) ions was adsorbed (63.6 mg m−2) by sample F 6 with the highest coat weight (6.8 g m−2). However, the filters showed much higher areic adsorption for Cu(II) compared to Ca(II). Thus, little negative impact by competitive adsorption of Ca(II), which is commonly present in natural water sources, is expected for Cu(II) adsorption. Still, TEMPO-CNF filters exhibited relevant adsorption capacities for Ca(II) ions, enabling the filters to reduce water hardness, which is an important parameter of the quality of water.

Regeneration and reuse of foam-coated filters

After adsorption, Cu(II) ions were desorbed from the filters by soaking in HCl solution, which is known to be an efficient method to regenerate filters saturated with Cu(II) ions. More than 90% of the Cu(II) ions were recovered from the filters using optimised regeneration conditions, which was soaking the filters in HCl at pH = 2 for 2 h. A second washing cycle did not further desorb Cu(II) ions, indicating that copper ions were majorly desorbed from the foam-coated TEMPO-CNF filters already after one washing cycle. After desorption of Cu(II) ions, neutralised TEMPO-CNF-coated filters were subjected to a second filtration test. Areic re-adsorptions were 94 and 54 mg m−2, corresponding to 73 and 72% of the original areic adsorption for regenerated filter samples F 6 and F 3, respectively.

Mechanical properties of foam-coated filters

The mechanical properties of foam-coated TEMPO-CNF filters were analysed in wet and dry conditions by tensile tests (Fig. S4). The Young’s moduli and tensile strengths of foam-coated TEMPO-CNF filters in dry condition increased from 40 MPa for the pure substrate to about 500 MPa for coated samples and in wet conditions from 7 to more than 12 MPa, respectively (Fig. 4). The reinforcement effect of the dry TEMPO-CNF on the viscose filters was probably attributed to the TEMPO-CNF coating bonding the viscose fibres, thus increasing their tensile properties. With the increase of the coat weight, however, the numbers of binding sites between viscose fibres may not necessarily increase, leading to a plateau region of the stiffness of the coated filters. The increase of the stiffness was at the expense of reduced breaking strain (Fig. S4), which corresponded to the cleavage of binding sites between viscose fibres.

Fig. 4
figure 4

Young’s modulus and tensile strength in dry (left) and wet (right) condition vs. the coat weight of foam coated filters

Full size image

The Young’s moduli and tensile strengths of the wet substrate and coated viscose filters (20 and 4 MPa, respectively) were identical within measurement error (Fig. 4). TEMPO-CNF greatly absorbs water and swells, thus becoming softer whereby its reinforcement effect onto the viscose filters vanishes and the stiffness of the wet filters was predominately controlled by the viscose substrate. In wet condition, the uncoated and coated filters were elongated more than twice than in dry condition (Fig. S4), which was related to the plasticising effect of water in hydrophilic materials [33].

Cast-coated TEMPO-CNF filters

The adsorption capacity of foam-coated TEMPO-CNF filters was higher for lower coat weights, but their areic adsorption was higher at higher coat weight. However, for foam coating, the maximum coat weight is inherently limited by the volume of foam that can be applied and collapsed on the substrate. Thus, to increase the coat weight and hence the areic adsorption, TEMPO-CNF suspensions were also directly cast-coated on the viscose substrate. By adjusting the thickness of the coating, the coat weight was controlled between 2.3 and 15.8 g m−2. The upper limit of coat weight that could be realised was set by the amount of water that could be removed during drying (Table 2). Because of the high water content in the suspension, drying of the TEMPO-CNF coating could not be realised in continuous roll-to-roll mode and the coated fabrics had to be kept static in the drying section for approx. 30 min during which time they were affixed in transverse direction to reduce shrinkage.

Table 2 Set and real coat weights as well as shrinkage of cast coated filters
Full size table

Similar to foam-coated filters, a linear relationship between the coat weight and thickness of the filters was established. However, other than for foam coating, the slope was much smaller and an almost constant thickness of 200 to 250 µm (Fig. 5) with a tendency to higher thicknesses at higher coat weight was established. This could be explained by TEMPO-CNF suspensions penetrating deeper into the viscose substrates compared to foams, filling the pores rather than just being applied to the surface of the filters, due to the higher density of the suspensions (ca. 1000 kg m−3) as compared to the foams (ca. 200 kg m−3).

Fig. 5
figure 5

Relationship between thickness and coat weight of cast and foam coated filters

Full size image

Furthermore, cast-coated filters exhibited a more compact, continuous, and homogeneous surface structure compared to foam-coated filters (Fig. 6). In particular, sample C 7 exhibited a denser and smoother surface as compared to the other samples, especially C 1–4, caused by a higher coat weight (15.8 g m−2) (Fig. 6b). All cast coated TEMPO-CNF filters (samples C 1–7) are illustrated in the supporting information (Fig. S2). TEMPO-CNF were present as a very thin layer between fibres or coated on individual fibres. However, occasional holes can still be observed in the coating layers, especially for samples with low coat weight (Fig. 6c, d).

Fig. 6
figure 6

SEM images of the surface of cast coated TEMPO-CNF filters at 50 × (top) and 1000 × (bottom) magnification: a & c Sample C 1 (2.3 g m−2), b & d Sample C 7 (15.8 g m.−2)

Full size image

Pure water permeance of cast-coated filters

Cast-coated filters exhibited substantially lower permeance than foam coated-filters and the permeance significantly decreased with increased coat weight. This can be explained by overall higher coat weights realised with cast coating and the resulting denser layers on top of the substrate, which function as a barrier resisting the water throughput. Furthermore, the permeance of the filters generally decreased as the backpressure increased, which can be explained by filter compaction caused by the backpressure (Fig. 7). Still, permeances of more than 100 000 dm3 m−2 h−1 MPa−1 are considered sufficient for water filtration applications and are much higher than, for instance, reported for pure cellulose nanopapers [27, 28].

Fig. 7
figure 7

Water permeance vs. real coat weight of cast coated filters

Full size image

Adsorption capacity for Cu(II) ions

Cu(II) adsorption on cast-coated filters was analysed using the same method applied for foam-coated filters. It was noted that the feed solutions passed through the cast-coated filters much slower compared to foam-coated filters because of the higher coat weight and the presence of a dense continuous layer on top of the filters. Compared to the substrate and foam-coated filters, the cast-coated filters exhibited higher areic adsorption of up to 280 mg m−2, explained by the higher overall amount of TEMPO-CNF present, i.e. higher coat weight (Fig. 8 left). This was also observed by higher quantities of water that could be passed through the filters before saturation was reached (Fig. S3). Higher adsorption at the same coat weights for cast-coated filters compared to foam-coated filters could be ascribed to longer residence and contact times of Cu(II) ions enabled by the lower permeance.

Fig. 8
figure 8

Areic adsorption (left) and adsorption capacity (right) vs. real coat weight of foam and cast coated filters

Full size image

Increasing the number of free carboxyl groups on/in the filters improves the retention of divalent ions from aqueous solution [34]. On the other hand, the adsorption capacities decreased with increasing coat weight (Fig. 8 right), indicating that the adsorption preferentially took place at the surface of the filters. This phenomenon was also reported before for (dense, anionic) cellulose nanopapers and filters [27, 28, 35]. This favours the application of very thin deposits of TEMPO-CNF on filter substrates offering a high number of available adsorption sites; however, large areas of filter surface are required to facilitate adsorption of large absolute amounts of HM ions.

Regeneration and reuse of cast-coated filters

The optimised regeneration process was also applied on saturated cast-coated filters, which were soaked in HCl solution at pH = 2 for 2 h. More than 90% of Cu(II) were recovered from the filters during the first washing cycle. The amount desorbed by a second washing cycle was almost zero, further confirming that the copper ions were already mostly removed from the cast-coated TEMPO-CNF filters during the optimised regeneration procedure. A high amount of Cu(II) ions could be re-adsorbed after neutralisation (186 mg m−2 and 202 mg m−2), corresponding to 76% and 74% of the original amount adsorbed for the regenerated cast-coated filters samples C 5 and C 7, respectively, values which correspond well to foam-coated filters.

Mechanical properties of cast-coated filters

Representative stress-strain curves of cast-coated filters (Fig. S4) show a linear region within a strain range from 0 to 1% in dry and 0 to 10% in wet conditions, respectively, corresponding to the elastic deformation region of the TEMPO-CNF filters. Beyond that region, the curves flatten, indicating local plastic deformation, such as the pull-out or breaking of cellulose fibres in filters. The uncoated sample showed much higher yield points in both dry and wet conditions (15 and 24% strain, respectively) compared to cast-coated TEMPO-CNF filters. This could be explained by the mechanically stiff TEMPO-CNF coating, hindering the elongation of the viscose fibre network. Plastic deformation continued until the stress reached a maximum, indicating ultimate tensile strength.

In the dry state, all coated filters exhibited much higher Young’s moduli and tensile strengths compared to the uncoated substrate. The Young’s moduli of cast-coated TEMPO-CNF filters in dry condition ranged from 40 to 1000 MPa (Fig. 9 left), increasing with increasing coat weight. This indicated that the means of reinforcement of TEMPO-CNF in cast-coated filters differed to those in foam-coated ones. Cast coating and drying of TEMPO-CNF formed a continuous phase to reinforce the viscose filters. Generally, increased coat weight increased the continuous phase volume and, therefore, provides further reinforcement to the filters. However, the ultimate tensile strength was at almost the same level. Furthermore, compared to foam-coated filters and the pure substrate, cast-coated filters displayed significantly higher stiffness and strength. This can be explained by the high amount of coating deposited during cast coating that has a positive impact on the reinforcement of filters.

Fig. 9
figure 9

Young’s modulus and tensile strength in dry (left) and wet (right) condition vs. the coat weight of cast coated filters

Full size image

The Young’s moduli and tensile strengths of wet uncoated and coated viscose filters were identical within error (Fig. 9 right). Furthermore, all foam- and cast- coated samples had tensile strengths of more than 4 MPa in wet conditions, which was basically equivalent to the uncoated sample (4.4 MPa). This could be explained by the TEMPO-CNF layer absorbing water, swelling, and thus softening. Hence, there was hardly any reinforcing effect by the coating in wet conditions, similar to foam coating.

Conclusion

In this work, TEMPO-CNF were foam and cast coated on commercial viscose substrates in pilot scale to realise water filters. The coatings were shown as thin layers on the surface of single viscose fibres or between fibres. It was established that the efficiency of TEMPO-CNF filters was related to their coat weight. Cast-coated filters showed a higher areic adsorption of up to 280 mg m−2 for Cu(II) compared to foam-coated filters (up to 128 mg m−2) due to their higher coat weight and the number of carboxylic groups present which are responsible for adsorption. However, foam-coated filters had a higher adsorption capacity of up to 52 mg g−1 than cast-coated filters (up to 35 mg g−1), indicating higher efficiencies of the adsorption process and preferable adsorption at the surface of the CNF layer opposed to the bulk. The permeance of foam- and cast-coated TEMPO-CNF filters at 0.2 bar was about 330,000 to 430,000 dm3 m−2 h−1 MPa−1, similar to the uncoated substrate, and 100,000 to 380,000 dm3 m−2 h−1 MPa−1, respectively. Furthermore, more than 70% re-adsorption capacities were found for the regenerated foam- and cast-coated TEMPO-CNF filters. Thus it was demonstrated that foam- and cast-coated TEMPO-CNF filters have high potential for utilisation in adsorptive filtration also in pre-industrial scale, tackling water pollutants including heavy metal ions as demonstrated for Cu(II).

Experimental

TEMPO-CNF suspension (1.1 mmol g−1 carboxyl groups, consistency 1.7%) was produced and kindly supplied by Swiss Federal Laboratories for Materials Science and Technology (EMPA). Hostapur SAS 30 (secondary alkyl sulfonate CH3 – (CH2)m – CH-(SO3Na) – (CH2)n – CH3) was purchased from Clariant. Viscose fabric was supplied by Acondaqua (Spain) and had a thickness of about 170 µm and an areic mass of ~ 50 g m−2, the diameter of the viscose fibres was approx. 10 µm. CuCl2 (97%), ammonium chloride (97.5%), Murexide, HCl (37%), and NaOH were purchased from Sigma-Aldrich. Ethylenediaminetetraacetic acid (EDTA) was purchased from W. Neuber’s Enkel (Vienna, Austria). All chemicals were used as received without further purification and ultra-pure water (0.05 µS cm−1, Elga Pure lab, High Wycombe, UK) was used for all analytical procedures.

Preparation of foam- and cast-coated TEMPO-CNF filters

All filters were prepared and produced by foam and cast coating on the Surface Treatment Concept (SUTCO) pilot coating machine located at VTT (Espoo, Finland) [14, 24, 36].

Foam coating process

For foam coating, the TEMPO-CNF suspension with 1.7 wt.-% solid content was mixed with 1.5 wt.-% (based on the dry weight of TEMPO-CNF) SAS 30 as foaming agent. The mixing was done gently to avoid any air entrapping before foaming. The homogeneous dispersion was then fed to a foam generator (Hansa Top-Mix, Hansa Industry Mixer GmbH). Compressed air was injected into the foam generator and the suspension frothed under shear created by a rotor–stator assembly. The airflow rate was automatically controlled based on the flow rate of the liquid suspension and the predefined foam density. The produced foam was then pumped to the coating station of the pilot machine where it was applied on the moving web through a slot-die. The die had a width of 250 mm and the slot-web gap was fixed at 1 mm. After foam deposition, the web was run over a vacuum box (under-pressure of 4–5 kPa) to enhance the penetration of the foam into the bulk of the substrate. The coated wet web subsequently entered the drying section, which was composed of three separate infrared drying units (12–14 kW each) and five convection drying units (140 kW each). The water in the foam was evaporated; hence, the foam collapsed and dry TEMPO-CNF was deposited on the substrate. The line speed was varied between 2.5 and 4 m min−1 for the foam coating process to control the coat weight.

Cast coating process

Cast coating was used to generate higher coat weights of CNF on the substrate as compared to foam coating. The CNF suspension containing 1.7% solid content was directly fed into the coating feed and the bottom of the reservoir was opened toward the machine direction. The opening size, denoted as the casting gap, was measured as the vertical distance between the bottom surface of the casting bar and the web. The coat weight during casting depends on both the line speed and the casting gap. The coating width was kept constant at 300 mm and the line speed was fixed at 1 m min−1. Hence, the theoretical coat weight only depended on the casting gap and the solids content of the CNF suspension. Cast coating was followed directly by drying the web, the vacuum box was not used as the high amount of coating was observed to cause penetration of the CNF suspension to the backside of the web and exiting it, leading to low retention on the web. The cast-coated substrates were dried in batches. Due to the high coat weight and thus high water content, the web was kept in the IR drying section for 30 min at 2 kW power to gently evaporate all the water. During drying, the coated fabrics were fixed in the transverse direction using clips to reduce the web shrinkage.

Coat weight, shrinkage, and thickness of the filters

The grammage of TEMPO-CNF was determined by the ratio of the mass of the coating (= actual weight of the filter minus the mass of the substrate, calculated for the area of the specimen from the areal weight) and cross-sectional area of 0.24 m diameter filters. The foam-and cast-coated TEMPO-CNF filters shrunk in transverse direction during the drying process and the shrinkage was calculated by the ratio of set width (0.25 m) and the width of the coating on the filters. With the calculated shrinkage, the real coat weight (Grcw) of TEMPO-CNF filters was calculated (Eq. 1).

$$G_{{\text{rcw}}} \left[ {\frac{{\text{g}}}{{{\text{m}}^2 }}} \right] = \frac{{\text{set coat weight}}}{{{1} - {\text{shrinkage}}}}$$
(1)

The thickness of each specimen of foam and cast coated TEMPO-CNF was measured at three different spots using a digital micrometre (electronic outside micrometre 0–25 mm, accuracy 0.001 mm).

Micro structure of filters by scanning electron microscopy

The microstructure of uncoated as well as foam and cast coated TEMPO-CNF viscose filters was assessed by scanning electron microscopy (SEM, JCM-6000, JEOL, Germany). Prior to imaging, the samples were fixed onto SEM stubs using carbon tabs and gold coated for 1 min at 30 mA using a sputter coater (JEOL Fine coater JF-1200, Germany).

Pure water permeance of filters

The permeance of pure water through the filters was measured at room temperature (21 °C) using a dead-end cell (Sterlitech HP4750 stirred Cell) at a backpressure of 0.2, 0.5, and 1 bar nitrogen, respectively. Filters with a diameter of 49 mm were fixed on a porous stainless steel support, which was then placed on top of the bottom part of the dead-end cell. 200 cm3 of deionised water were filled into the cell body, which was sealed by the cell lid connected to N2. At the desired backpressure, the water was forced through the filter. The pure water permeance P [dm3 m−2 h−1 MPa−1] for the active filtration area (1460 mm2) was determined by measuring the volume permeated per unit area (A), time (t), and pressure (p) (Eq. 2).

$$P\left[ {\frac{{{\text{dm}}^3 }}{{{\text{m}}^2 \cdot {\text{h}} \cdot {\text{MPa}}}}} \right] = \frac{v}{A \cdot t \cdot p}$$
(2)

Adsorption of metal ions on filters

Two discs of the filters with a diameter of 0.24 m were placed in a Büchner-Funnel. The funnel was fixed on a filter flask, which was connected to a water-jet pump. A solution containing Cu(II) ions at an initial concentration of 2.5 mM was filtered through the TEMPO-CNF-coated filters under reduced pressure to reach saturation of the filter and the permeate fractions were collected. Cu(II) concentrations of the feed and filtrate solutions were determined by titration. 50 cm3 of the permeate were pipetted into a 0.25 dm3 Erlenmeyer flask and mixed with 50 cm3 pure water. The pH was set to 10 by adding 5 mg of NH4Cl and 2 cm3 of 30% NH3 solution. Murexide was dissolved in the solution as colour indicator. A yellowish/greenish solution was obtained and titrated with 0.11 M EDTA until the transition point (colour change from yellow/orange to pink). To calculate the amount of Cu(II) adsorbed [mg] from a certain filtration fraction, Eq. (3) was applied.

$$m_{{\text{CU}}} = \left( {C_{{\text{feed}}} - C_{{\text{permeate}}} } \right) \cdot 63.546 \cdot V_{{\text{permeate}}}$$
(3)

Cfeed is the concentration of the feed solution and Cpermeate is the concentration of the permeate solution [\({\mathrm{mmol\, dm}}^{-3}\)]. 63.546 [\(\mathrm{mg }\,{{\text{mmol}}}^{-1}\)] is the atomic weight of copper. Vpermeate is the volume of solution that was filtered through TEMPO-CNF membrane filters. The mass of adsorbed copper ions [mg] per unit filter area [m2], the areic adsorption capacity \({q}_{{\text{A}}}\) [mg m−2], was obtained using Eq. (4).

$$q_{\text{A}} = \frac{{\text{mass of the adsorbed copper ions}}}{{\text{area of the filter}}} \cdot \left[ {{\text{mg}} \cdot {\text{m}}^{ - 2} } \right]$$
(4)

The adsorption capacity \({q}_{{\text{e}}}\)[mg g−1], the mass of adsorbed copper ions \([{\text{mg}}]\) per total mass of active adsorption agent \([{\text{g}}],\) was then calculated by dividing \({{\text{q}}}_{{\text{A}}}\) by the coat weight (Eq. 5).

$$q_{\text{e}} = \frac{{q_{\text{A}} }}{{\text{coat weight}}} \cdot \left[ {{\text{mg}} \cdot {\text{g}}^{ - 1} } \right]$$
(5)

Regeneration of filters

After the adsorption test, two discs of the filters with a diameter of 0.24 m with adsorbed Cu(II) were soaked in 0.2 dm3 of 0.01 M HCl for 2 h. During this time, the pH value of the soaking solution was checked with pH indicator paper every 30 min. The concentration of Cu(II) ions in the acidic solution was then analysed by titration. The regenerated TEMPO-CNF filters were rinsed several times with deionised water to reach pH 7 and dried at room temperature. Afterwards, adsorption tests were conducted as described above to evaluate the regenerated TEMPO-CNF filters.

Mechanical properties of filters

Tensile tests were performed using a universal test frame (Model 5969 Dual Column Universal Testing System, Instron, Darmstadt, Germany) equipped with a 1 kN load cell. Dogbone specimens with shapes according to ISO 527 type 5 were cut from each filter longitudinal to the fibre/machine direction. To determine the mechanical properties of wet filters, prior to the tests, specimens were soaked for 30 min in deionised water. The dry and wet specimens were clamped between two clamps with a distance of 25 mm. The specimens were stretched at a speed of 1 mm min−1. The force during the tensile tests was recorded by the machine. The gauge length was set to 20 mm and the deformation of the specimens during the tests was recorded from dotted marks by a video extensometer (Blackfly GigE-PoE). From the stress-strain curves, the mechanical properties (ultimate tensile strength and elongation at break) of the filters were obtained. The Young’s modulus was calculated from the slope of the linear region of the stress-strain curves as secant between stress values separated by 0.2% strain.