Single-walled carbon nanotubes (SWCNTs) used in this project were kindly offered by OCSiAl and had ˂2 nm diameter and >5 µm length. The nanotubes concentration is ≥ 80% and impurities account for ≤15%. Cyrene was kindly supplied by Circa Group. NMP was purchased from Acros Organics. Ultrason E3020 P Polyethersulfone (PES) of 55,000 Da and polyvinylpirrolidone Luvitek K-90 (PVP) Pulver of 1,500,000 Da were obtained from INGE.BASF, Germany. Both polymers were solvent exchanged in ethanol and ethyl acetate respectively and dried in a vacuum oven overnight before use. Strains, materials and reagents for the antimicrobial activity assessment (Escherichia coli XL10-Gold and Bacillus subtilis 168, DSM 402) were sourced from Agilent and Leibniz Institute DSMZ respectively. Whatman Filter Papers Grade 1 and phosphate-buffered saline (PBS) Tablets were sourced from Thermo Fisher Scientific. Miller Luria Broth and Agar were purchased from Melford and quaternary ammonium compound-based disinfectant (Q SHIELD Surface Care Spray was sourced from Q Biotechnologies.

Dispersions of 0.1% pristine SWCNTs in Cyrene and NMP (w/v) were prepared using a bath sonicator (GT SONIC-D3 with an ultrasonic power of 100 W) for 1 hour in water maintained at a temperature below 25°C (Supplementary Figure S1). 15 mL centrifuge tubes containing the desired solvent and SWCNTs were placed in a beaker, which in turn was placed into an ultrasonic bath and sonicated for an hour. After sonication, the samples were centrifuged for 10 min at 2,500 g, at room temperature using a Biofuge Primo centrifuge with an angle of 90° and 17 cm rotor. The upper 80% of the supernatant was then carefully decanted, separating the big bundles of SWCNT ropes and impurities from the commercial nanotubes. The supernatant was then collected and filtrated through 0.22 µm pore size nylon membranes with a diameter of 13 mm (pastel green P/N: FIL-S-NY-022-13-100 from Chromatography Direct). The nanotubes deposited onto the membrane were washed three times with 10 mL ethanol to remove any residual solvent and left to dry in the air for several days. To calculate the concentration of SWCNTs dispersed, the SWCNTs retained from the filtration membrane were recovered, dried, and weighed and the actual concentration registered.

The synthesis of buckypaper was accomplished through a straightforward filtration process. A dispersion of 1.25% w/v SWCNTs (plural form for single-walled carbon nanotubes) in Cyrene was achieved by employing the identical ultrasonic bath described in Section 2.2 for a duration of 1 hour. Following dispersion, the sample underwent centrifugation (10 min at 2,500 g, at room temperature, using a Biofuge Primo centrifuge with an angle of 90° and 17 cm rotor.). The resulting supernatant was then collected and filtrated through a 0.22 µm pore size nylon membrane with a diameter of 13 mm. The retained nanotubes on the filter were subsequently subjected to three ethanol washes (10 mL each) and left to air-dry at room temperature.

The surfactant-free Cyrene-based buckypaper, deposited on the nylon membrane, was employed to filter household wastewater and effect the separation of water and cooking oil from an emulsion. A sample of household wastewater and a 1:1 water:oil emulsion were subjected to filtration using a syringe equipped with the buckypaper, deposited atop a 0.45 μm nylon membrane.

The CNT-based flat-sheet PES membranes have been manufactured via non-solvent induced phase separation process (NIPS), using water as a coagulant. In this project, four membranes have been prepared, and their casting solution composition can be seen in Table 1.

The SWCNTs were dispersed first in the solvent by sonicating for 1 hour. This step was skipped for the pristine membranes. This solution was used to dissolve PES by dissolving 10–15 wt% of PES pellets at a temperature of 70°C for 4 h. Then concentrations of 3% PVP or 5% PVP were added under continuous stirring. The casting solution was degassed and then placed on a glass plate and a film of thickness of 150 µm was obtained using a manual casting knife. The casting film was submerged in a coagulation bath containing deionised water at RT. Membranes were then washed three times in distilled water for 10 min while under sonication in order to wash out the residual solvent. The fabricated membranes were then stored in deionised water until further use. To characterise the prepared membranes, they were dried in a vacuum oven at 80°C overnight.

The PES flat sheet membranes were dried and frozen in liquid nitrogen followed by Au/Pd coating. Cross-sectional SEM images of the coated membranes were recorded using a JEOL JSM-6490LV Scanning Electron Microscope. SEM images of the surface of one membrane and of the buckypapers were measured using a Thermoscientific Scios 2 scanning electron microscope, EDX analysis was performed on this instrument using a Bruker XFlash 6.60 EDX detector. All images shown here are secondary electron images.

200 mesh formvar/carbon copper grids were pre-treated by glow discharge in Polaron E6000 vacuum coating unit to make them hydrophilic. 5 μL of each sample were loaded onto these pre-treated grids and left to dry for 48 h in a fume hood before analysis. The grids were analysed on FEI Tecnai 12 BioTwin G2 Transmission Electron Microscope operating at 120 kV. The images were collected on a SIS CCD camera at magnifications of 6.8 k and 98 k.

Optical microscopy images were recorded using a Leica S6D Microscope with 6.3×-40× magnification and flat image field after placing a droplet of sample (wastewater or emulsion solution before and after filtration) onto a glass slide. The working distance is 110 mm, from microscope to specimen, providing space for manipulation.

For AFM, uncoated samples were used. The samples were scanning using an AFMWorkshop HR-AFM instrument in vibrating mode. Vibrating mode cantilevers with resonant frequency about 200 kHz were used. Images were collected in 6 × 6 µm areas of each sample, and all images shown here are height images.

Dead-end membrane filtration measurements were carried out using a Sterlitech HP4750 membrane filtration cell (max volume 300 mL) connected to a nitrogen gas supply. The cell can hold a membrane disk of ∼49 mm in diameter and the active membrane area 14.6 cm². The membrane holder and cell body are made of 316L stainless steel. The stirrer was removed for the pure water flux tests. The arrangement of the filtration and flux measurement equipment is seen in Figure 3.

The filtration cell was pressurized via nitrogen gas (oxygen free) out of a nitrogen cylinder, which was controlled by the reducing valve at the gauge of the cylinder. The applied pressure was monitored by an in-line pressure sensor with an overall measurement error of <1.5%. The cell was not stirred. In this case, rates of filtration J _w (L m⁻² h⁻¹ (LMH)) have been determined via recording the weight of water obtained in a given time interval and then using the Equation 1: J w = t + ∆ t − m t   ρ   A m   ∆ t (1)

where m is the mass at a given time, ρ is the density of water, A _m is the membrane area and t is the time (all in appropriate units to get LMH).

A UV–Vis spectrophotometer (Shimadzu UV-1800) was used to measure the spectra of wastewater and oil-water emulsion samples before and after filtration by Cyrene-based buckypaper, across a wavelength range of 200–800 nm.

The functional groups present in samples were investigated using PerkinElmer Spectrum 400 ATR-FTIR Spectrometer with transmittance peaks in 4,000–500 cm⁻¹ region, with rapid scanning (4 scans) and resolution 4 cm⁻¹ at room temperature. The obtained data was analysed using the OriginPro 2024 software.

As a measure of membranes’ hydrophilicity, the water contact angle was measured via the sessile drop method using a Theta Lite optical tensiometer at a room temperature of 20°C. A range of 1–2.5 µL droplet sizes of water were placed on the membrane surface and the images were recorded using the automated OneAttension software. The contact angles were measured at a minimum of three random locations and the mean values reported to minimize experimental error.

Membrane samples were prepared by cutting 1.3 × 2 cm (2.6 cm²) sections. Nylon filter bucky paper was used as is. Samples were autoclaved at 121°C for 20 min to sterilise. 10 mL overnight cultures of E coli XL10 Gold and B. subtilis 168 were set up in Luria Broth (LB) media and cultured at 37°C shaken at 180 rmm. They were then diluted to an OD of 0.1 (595 nm) in LB media. Using aseptic technique, the membranes were dipped into the diluted culture, and any residual drops were shaken off. The membranes were then placed individually on the inside of a separate sterile 50 ML falcon tube, and spun to remove excess liquid from the membrane, (500 rpm for 1 min). The samples were then incubated for 6 h at 37°C. Membranes were aseptically removed and added to 9.9 mL of sterile phosphate-buffered saline (PBS) and vortexed for 30 s. Serial dilutions were made and plated out on LB agar. The plates were incubated overnight at 37°C and colonies were counted. This was conducted in triplicate for each membrane type. A Whatman NO1 filter paper and nylon filter were included as negative controls and Whatman NO1 filter paper soaked in Quaternary ammonium compound-based disinfectant and then dried, served as a positive control.

In this study, we investigated the nanofluids of carbon nanotubes (CNTs) dispersed in Cyrene and NMP without the aid of surfactants, focusing on their debundling and stability characteristics. The authors have previously conducted a comprehensive study on the dispersion of single-walled carbon nanotubes (SWCNTs) in Cyrene and NMP, both with and without the inclusion of additives (R. A. Milescu, 2021). Following 1 hour of sonication in Cyrene, the nanotubes were observed across the entire grid surface, indicating effective debundling (unzipping) of aggregates into individual ropes (Figures 4A,B).

Due to the high boiling points of Cyrene (227°C) and NMP (202°C), combined with Cyrene’s significantly higher viscosity (14.5 cP compared to NMP’s 1.6 cP), the solvent tends to remain adhered to the nanotube surfaces. As a result, the walls of the nanotubes cannot be clearly observed. The TEM micrographs reveal the presence of only a few nanotube aggregates, confirming their successful exfoliation in the solvent. In contrast, the NMP-based nanofluid exhibited an uneven dispersion of carbon nanotubes, evident from the gaps on the grid surface (Figures 4C,D). Small bundles of nanotubes can be observed, suggesting an inefficient unzipping to individual ropes. Additionally, both examples showed the presence of small particles of impurities. The Cyrene- and NMP-based nanofluids containing SWCNTs were subjected to stability testing (Supplementary Figure S2). Generally, the stability of particles in a liquid is assessed by monitoring their sedimentation at the fluid’s base due to gravity (Phan and Haes, 2019). A longer settling time suggests higher stability for nanoparticles. In our study, it was noted that the nanoparticles settled within 1 hour in NMP, whereas the process took more time in Cyrene. Supplementary Figure S2A illustrates that, after a day, the Cyrene-based nanofluid exhibited significant SWCNTs agglomeration. The NMP-based nanofluid displayed minimal nanotube presence, evident from its colourless appearance, indicating low debundling and stability (Supplementary Figure S2B). Specifically, the dispersions of pristine SWCNTs in pure NMP proved unstable, with re-agglomeration observed within a short time frame. Surprisingly, even after 1 week, the agglomerates were still suspended in Cyrene, making it a better dispersant than NMP.

Following the stability test, the samples underwent filtration using a 0.22 μm filter, followed by three ethanol washes to eliminate the residual solvent. Subsequently, the filtered material was dried at room temperature overnight. The concentration of dry SWCNTs dispersed in NMP was approximately three times lower than that in Cyrene, measuring 0.013 mg mL⁻¹ compared to 0.038 mg mL⁻¹ (Supplementary Table S1).

SEM analysis of the buckypapers fabricated using NMP and Cyrene reveals a network of fiber bundles coating the nylon membrane surface (also shown in Figure 5A; Supplementary Figure S3). Cyrene appears to have produced a densely packed layer of SWCNTs (Figure 5B). In contrast, the coating formed using NMP is noticeably thinner, with the underlying membrane surface features still partially visible (Figure 5C).

Additional images in Supplementary Figure S4 also show regions of the NMP buckypaper where the Nylon membrane is clearly visible. These regions were not found on the Cyrene sample (Supplementary Figure S5). Both NMP and Cyrene buckypapers contained a number of impurities visible on their surface (Supplementary Figure S6). Energy-dispersive X-ray spectroscopy (EDX) suggested that these impurities are predominantly iron, in keeping with the data sheets provided by the supplier (Supplementary Figure S7). As such, for certain applications, a preprocessing chemical purification step would be required (Hou, Liu, and Cheng, 2008). The buckypaper exhibits well-defined nanotube ropes and interpores. As no surfactant was employed, no additional additive removal is required, rendering this novel buckypaper type more sustainable. The surface roughness of the buckpapers was compared with that of the nylon membrane (Figure 5D) using atomic force microscopy (AFM). The buckypaper surface was many times less rough than that of the original membrane (see z scales images Figures 5D–F). In keeping with the SEM data, the Cyrene-CNTs buckypaper was also smoother (Figure 5E) than that produced using NMP (Figure 5F). Infrared spectra of both Cyrene- and NMP-based buckypapers revealed residual solvent signals around 3,000 cm⁻¹, corresponding to O-H and N-H stretching frequencies (Figures 5G,H). The presence of these solvent residues raises concerns regarding the safety of using such filtration materials in applications involving water. This residual solvent could leach out, necessitating further investigation to ensure the safety and suitability of these materials for human-related applications.

Following the washing and drying process, the Cyrene- and NMP-based buckypapers deposited on a nylon membrane underwent thorough analysis and testing for their filtration capabilities. Two distinct and straightforward filtration applications were chosen for this evaluation: the treatment of waste-water and the separation of a cooking oil-water emulsion. These tests aimed to assess the effectiveness of the Cyrene-based buckypaper in removing contaminants and improving water quality. In the waste-water filtration test, the membrane’s ability to remove suspended solids, and potential pollutants was scrutinised. For the oil-water emulsion separation, the membranes’ performance was evaluated based on its capacity to selectively separate water from oil, thereby achieving a high degree of purity in the filtered water (Figure 6).

As depicted in Figure 6A, the permeate resulting from waste-water filtration revealed clear water samples following a single-step extrusion through a syringe containing the Cyrene- and NMP-based buckypapers. These results align with the UV-Vis spectra of the samples before and after filtration. Although the analysis of complex matrices such as waste-water may not be straightforward, the broad peak area between 220 and 300 nm is significantly reduced after filtration by both buckypapers. In the case of the oil-water emulsion (Figure 6B), a substantial peak reduction is observed after filtration using a Cyrene-based buckypaper. A less effective oil-water separation can be observed in case of NMP-buckypaper with a higher intensity peak observed in the UV-Vis. These results highlight the substantial potential of the Cyrene-based buckypaper, especially in critical scenarios such as accidental oil spillages in marine environments and the purification of other turbid solutions. Moreover, the versatility of this buckypaper extends beyond its application in oil-water separation, opening avenues for its utilisation in various filtration processes across diverse industries. The adaptability and efficacy demonstrated in this study suggest that the Cyrene-based buckypaper holds promise as a versatile solution for addressing environmental challenges and meeting filtration needs in different contexts.

Flat-sheet membranes produced with PES, PVP, and Cyrene have been previously reported in the literature by the team (Milescu et al., 2019). It was found that PES membranes produced with Cyrene exhibited higher total porosity than NMP-based membranes and the same pore diameter as typical NMP-based membranes produced with 5% pore-forming additive PVP, highlighting that Cyrene-based membranes do not require pore-forming additives in their composition. In this project, the authors incorporated a low concentration of SWCNTs to enhance the physico-chemical properties of the membranes for applications where high flux and anti-fouling properties are critical. Uniformity was maintained across all carbon nanotube-based membranes, which were produced using a consistent 0.1% mass ratio relative to the solvent. To prevent pore-blocking aggregations of nanotubes, which could negatively affect permeability, we first ensured the dispersion of nanotubes in the solvents before dissolving the polymer in these nanofluids. The flat-sheet membranes containing SWCNTs exhibit distinct visual characteristics when fabricated using Cyrene compared to NMP as solvents (Supplementary Figure S8). In the case of PES/C0 membrane, the surface appears as a wrinkled, dark grey layer. This texture suggests a more compact arrangement of SWCNTs and possibly a unique interaction between the nanotubes and the Cyrene solvent during the fabrication process. On the other hand, PES/CNT/N0 membrane displays a starkly different morphology. The surface features visible “ropes” of nanotubes that stand out through the white polymeric matrix. This transparency highlights a less uniform dispersion of SWCNTs within the polymer matrix. The nanotubes appear to cluster, potentially indicating weaker compatibility or interaction between the SWCNTs and the NMP solvent during processing. This uneven distribution might impact the mechanical, electrical, or permeability properties of the membranes, reflecting differences in the miscibility and solvation dynamics of the two solvents with the polymer and nanotubes.

Pristine PES membranes were produced with Cyrene and NMP for comparison and the cross-section figures of all membranes can be observed in Figure 7.

Commonly referred to as “Loeb-Sourirajan” membranes, this type is characterized by a thick, porous surface and an asymmetric distribution of pore sizes and porosity across the membrane’s thickness. Notably, smaller voids are present near one surface, while larger ones are evident on the opposite surface (Loeb and Sourirajan, 1962). The top layer of the membranes is supported by a sponge-like substructure, resulting from a slow demixing, and macrovoids formed due to instantaneous demixing during phase inversion. Finger-like structures can be observed near the top layer of the membrane, resulting from rapid demixing between the solvent (Cyrene or NMP) and non-solvent (water). The finger-like channels observed in membranes fabricated using Cyrene exhibited a more pronounced vertical orientation and a well-organised structure compared to those produced with NMP. Additionally, membranes prepared from Cyrene showcased a higher density of interconnected pores within the walls of the finger-like channels, consistent with prior findings (Ferreira et al., 2017). The porosity of PES/CNT/C0 (Figure 7A) is different from PES/CNT/N0 (Figure 7D) for the sponge structure; larger pores are observed in the Cyrene-based membrane. This phenomenon of pores forming in Cyrene without the aid of an additive was previously reported (Milescu et al., 2019). The introduction of PVP, commonly employed as a pore-forming additive, resulted in the formation of macrovoids in the Cyrene-based membranes (Figures 7B,C). The incorporation of PVP led to a reduction in sponge-like structures, with the extent of this reduction correlating with the amount of additive introduced into the casting solution. PES/CNT/C3 (Figure 7B) membrane reveals diminutive finger-like structures near the upper surface, a sponge-like configuration adjacent to the lower side (where the casting gel interfaces with the glass plate), and substantial macrovoids in the central area. This observation can be attributed to a rapid demixing process occurring from the top side, a stagnation of solvent-antisolvent interaction in the middle, and a gradual demixing process at the bottom side. In Figure 7C, the upper portion of the PES/CNT/C5 membrane reveals a thin layer of finger-like structures, resembling those seen in PES/CNT/C3 (Figure 7B).

Pristine membranes produced with Cyrene (PES/C in Figure 7E) and NMP (PES/N in Figure 7F) show larger macrovoids compared to the same membranes produced with added SWCNTs (Figures 7A,D). However, the Cyrene-based membrane (Figure 7E) exhibits interconnected pores within these macrovoids, similar to all Cyrene and SWCNTs-based membranes (Figures 7A–C). However, in case of NMP-based membranes, these interconnected pores can only be seen when SWCNTs were employed (Figure 7D). Additionally, large macrovoids traverse the PES/CNT/N membrane (Figure 7D), while a modest presence of sponge-like structures is observed at the membrane’s centre. This observation suggests a rapid demixing process occurring at the edges of the membrane, with a notable deceleration in the central region. This slower demixing in the center is attributed to the higher viscosity of the casting solution caused by the presence of carbon nanotubes, which hinders the diffusion and separation process. The rapid demixing from the bottom side indicates a potential separation of the emerging membrane from the glass plate, allowing water to infiltrate the space between the plate as demixing initiates or a fast water leaching through the displaced side of the membrane (Supplementary Figure S9). On the upper surface, demixing occurs swiftly, leading to the precipitation of the polymer and the formation of voids due to water displacement. Comparable voids become apparent when the membrane is separated from the glass plate, allowing water to ingress into the space or release the membranes porosity (Supplementary Figure S9B). The uniform thickness of all membranes, set at only 150 μm, coupled with the introduction of the hydrophilic pore-forming additive, contributes to an accelerated demixing rate. PVP serves as an antisolvent agent during the demixing step, owing to its high solubility in water. Moreover, it is essential to note that the water-Cyrene demixing process deviates from conventional solvent/non-solvent demixing. When water is introduced to Cyrene, the resulting mixture transforms into a complex blend of Cyrene, water, and geminal diol (De Bruyn et al., 2019). Furthermore, the viscosity of this new mixture varies depending on the amount of water added to Cyrene, ranging from a low viscosity, similar to that of water, to a more viscous solution when only 25% water is present in Cyrene (Milescu et al., 2023). Throughout the demixing process, if the mixture contains more Cyrene than water at any given time, the viscosity of the solution increases, impacting pore formation. In this case, a higher viscosity would create macrovoids. Therefore, the formation of large macrovoids in PES/CNT/C3 and PES/CNT/C5 membranes may result from two contributing factors: 1) the hydrophilic nature of PVP, which facilitates faster demixing by interacting with water and local viscosity increase and 2) the viscosity of the water/Cyrene/diol combination during demixing, leading to the development of extensive zones at the contact interface.

The water contact angle of the flat sheet membranes was found to be influenced by the dispersion of carbon nanotubes within the polymer matrix (Figure 8; Supplementary Table S2).

As shown in Figure 8, PES membranes containing SWCNTs (PES/CNT/C0 and PES/CNT/N0) exhibited highest water contact angles (up to 89.2°) compared to those without SWCNTs (PES/C0 and PES/N0) with contact angle up to 64.3°. This increase can be attributed to the inherent hydrophobicity of the pristine nanotubes. The non-uniform dispersion of SWCNTs within the polymer matrix produced with NMP led to the formation of aggregates, which, in turn, caused a reduction in the contact angle of the PES/CNT/N0 membrane. These agglomerates likely altered the surface characteristics, making it more hydrophilic, as water could interact more readily with the exposed areas of the membrane. The inclusion of polyvinylpyrrolidone (PVP), a hydrophilic additive, in the casting solution further enhanced membrane hydrophilicity. Increasing PVP concentration significantly reduced the water contact angle, from an average of 68° for PES/CNT/C3 to 65° for PES/CNT/C5. The difference in contact angle between the active and reverse sides of the membranes can be attributed to the non-uniform dispersion of nanotubes during membrane fabrication. A higher concentration of nanotubes on the reverse side results in increased hydrophobicity on that surface.

To explore the effects of membrane cross-section morphology on its permeability and the practical application of the prepared PES membranes with carbon nanotubes, their water permeation fluxes were investigated. The membranes were investigated using both sides, due to their different porosities: 1) active side (top) and reverse side (the side close to the glass plate). The average water permeability of the membranes can be seen in Figure 9.

The presence of macrovoids is considered a deviation from the desired membrane morphology, impacting filtration efficiency and potentially leading to superior filtration outcomes (Kosma and Beltsios, 2012). Conversely, the sponge structures present in the absence of PVP support a less effective filtration ability, which may be enhanced by increasing the applied pressure. Based on these criteria, the permeate flux seen in Figure 9 shows differences between the membranes depending on the solvent used, presence of SWCNTs and side of the membrane used. The passive side of the membrane generally exhibited higher permeability compared to the active side, aligning with the observed skin porosity in Supplementary Figure S10. All Cyrene-based membranes display pores on the passive side, which become larger when SWCNTs are added. In contrast, NMP-based membranes do not show any visible skin pores, when optical microscopy was employed. An exception was noted for PES/CNT/C3, where the active side displayed greater permeability than the passive side of the membrane. This could be attributed to either 1) a thicker sponge-like layer beneath the passive side, impeding filtration or 2) to a more hydrophilic active side (contact angle of 67.9°) compared to reverse side (69.1°). Additionally, finger layers in this membrane’s morphology feature large zones of end ends, adversely affecting the flux. All Cyrene-based membranes demonstrate significantly higher water permeance than their NMP-based counterparts. The PES/CNT/C0 membrane demonstrated a water permeability that was thirteen times higher than that of its NMP counterpart. This significant increase is attributed to the presence of large pores in the skin layer, as well as the large, interconnected macropores and the porous sponge layer of the Cyrene-based membrane. The addition of PVP resulted in even greater water permeability, attributed to the presence of large macrovoids. It is crucial to note, however, that these membranes are susceptible to collapse under high pressure, rendering them impractical for certain applications.

Adding carbon nanotubes to the casting solution increased the pore size on the membrane surface, enhancing water permeability and resulting in higher water permeability compared to pristine membranes. Cyrene-based membranes, which were operated at low pressures (less than 5 bar) as shown in Supplementary Table S3, are versatile and can be used in various filtration ranges: nanofiltration (10–100 LMH and 4–30 bar), ultrafiltration (50–200 LMH and 2–10 bar), and microfiltration (greater than 100 LMH and 0.1–2 bar). These membranes are suitable for applications involving the filtration of small molecules, proteins, viruses, and suspended solids in water treatment and food processing. In contrast, NMP-based membranes required higher pressures, up to 8 bar, to achieve efficient filtration.

To assess if the membranes formed with and without SWCNT in both NMP and Cyrene displayed any difference in terms of bacterial activity, a simple bacterial challenge was devised whereby the membranes and buckypapers were exposed to a Gram-negative (E. coli) and a Gram-positive (B.subtilis) strain of bacteria. No active filtration was conducted, bacteria were taken up into the porous network of the membrane through adsorption with any free solution removed by mild centrifugation. Results are shown in Supplementary Figures S11, S12. Data shows that all membranes have sufficient porosity to uptake bacteria, but that the presence of SWNT did not show any significant difference to the controls in terms of bacterial colony count. The negative control of filter paper showed a higher colony count compared to the PES membranes, but no significant differences were observed between membranes cast in Cyrene or NMP, with or without the presence of SWCNT. Similarly, the presence of SWCNT supported on the surface of a nylon membrane did not show any significant difference to the negative control of just the nylon filter. To ensure the test was functioning, positive control of filter paper treated with a quaternary ammonium salt was also trialed. This sample showed no bacterial growth for both Gram-negative or Gram-positive bacteria. This shows that the presence of SWCNT doesn’t impart any antibacterial properties when using a simple challenge test.

Cyrene, a bio-based solvent derived from cellulose, is generally considered safer and more environmentally friendly than traditional solvents like NMP (Milescu, 2021). Cyrene is non-mutagenic, biodegradable, and does not contain potentially harmful amide groups associated with reproductive toxicity, as observed with NMP. It also avoids the formation of harmful by-products during degradation, offering a safer profile compared to many petroleum-based solvents. Still, any residual Cyrene in materials, like filtration membranes and buckypapers, requires thorough investigation to ensure it does not leach into water or other fluids in concentrations that could compromise safety for human exposure. Carbon nanotubes, celebrated for their exceptional mechanical strength, electrical conductivity, and thermal stability, are remarkable nanomaterials with vast applications in fields like electronics, medicine, and environmental engineering. However, they are also linked to significant concerns, including toxicological risks, environmental persistence, workplace safety challenges, gaps in research and regulatory frameworks, and the ongoing need for mitigation strategies and safer alternatives. Integrating green solvents like Cyrene with CNT technology in filtration membranes and buckypapers, underpinned by robust green metrics and Life Cycle Assessment, offers a pathway to optimise performance while reducing environmental and health risks. Mass-based metrics like Process Mass Intensity (PMI) and E-factor are valuable tools for assessing the sustainability of production processes. PMI quantifies the total mass of inputs required to produce 1 kg of the desired product, while E-factor focuses on the mass of waste generated per kilogram of product (Martínez et al., 2022). These metrics are straightforward to calculate for optimised systems and provide clear benchmarks for resource efficiency and waste reduction. LCA considers environmental impacts across all stages of the product’s life, from raw material extraction to end-of-life disposal or recycling. That said, performing an LCA is most meaningful at the pilot or industrial scale, where processes and resource flows are sufficiently developed and stable to provide reliable data (Hong et al., 2023).