Polyamide-6, tetra-n-butylammonium bromide (TBAB; ≥98%), and potassium antimony tartrate hydrate (≥99%) were purchased from Sigma-Aldrich whereas trifluoroacetic acid (TFA, 99%), acetone (Ace, ACS reagent grade, 99.5%), potassium dihydrogen phosphate (KH₂PO₄; 99.3%), sulfuric acid, and ascorbic acid (99.4%) were purchased from Fisher Scientific. Hematite (99.95%, average diameter of 3 nm APS powder) and ammonium molybdate tetrahydrate (99%) were purchased from Alfa Aesar. All materials were used without further treatment. Electrospinning solutions were prepared by dispersing α−Fe₂O₃ NPs in a solvent mixture of acetone (Ace) and trifluoroacetic acid (TFA) 60:40 mol%. Once α−Fe₂O₃ NPs were well-dispersed using sonication, PA6 pellets and TBAB (various concentrations: 0, 0.1, 1 wt.%) were added to the solution, which was then sealed for magnetic stirring until homogeneous.

Three solution properties (i.e., viscosity, surface tension and electrical conductivity) were measured. The viscosity was measured using a CPA-40 spindle connected to a Brookfield DV2THB viscometer. The solution viscosity was determined to be independent of the shear rate. Thus, the viscosity was measured at 95% torque at each rotational speed ranging from 0.5 rpm to 200 rpm. An automatic surface tensiometer (Shanghai Fangrui Instrument, QBZY-1) with platinum-coated plate was used to measure the surface tension. Solution electrical conductivity was measured using an electrical conductivity probe from Apera Instruments (Al1311, K = 0.1) connected to EZO conductivity circuit (Atlas Scientific), on Tentacle T3 using Raspberry Pi (Whitebox T3, Mkll). All solution property measurements were taken at room temperature prior to electrospinning to correlate them closely to the resulting nanofiber properties.

Electrospinning was conducted by injecting the prepared solution through a 5-mL BD Luer-Lok syringe with a 20-gauge stainless steel needle using a syringe pump (New Era, NE-100). Negative voltage was applied to the needle tip while the drum collector, wrapped with aluminum foil and rotating around 400 rpm, was grounded to collect the sample. Electrospinning and environmental conditions including applied voltage, feed rate, temperature, and absolute humidity were fixed at 12 kV, 0.25 mL/h, 23°C ± 1°C, and 0.008 ± 0.001 kg H₂O/kg dry air, respectively.

An experimental approach that varies one factor at a time can overlook conditions critical to developing a full systematic understanding of experimental design and fails to examine the interaction between the factors (Yu et al., 2022; Yu and Myung, 2018). DOE allows one to systematically vary factors of different conditions at the same time and fully visualize the effect of each factor on the resulting system properties. From the DOE analysis, an equation can be obtained that allows one to predict the response or conditions to achieve a specific, target response.

The first DOE consists of varying the α−Fe₂O₃:PA6 ratio and TBAB concentration while keeping the concentration of PA6 at 6.5 wt.%. The conditions for the α−Fe₂O₃:PA6 ratio and TBAB concentration were based on previous work done by Wang et al. (2021) with PAN tri-composites, which had adsorption capacity of 8.76 mg P/g of nanofiber membrane. The first DOE was conducted to study how adding and increasing α−Fe₂O₃ NP and TBAB concentration to PA6, a more hydrophilic polymer compared to PAN, would affect solution and nanofiber properties. The lower and upper limit for the α−Fe₂O₃ NP:PA6 ratio were chosen to be 0 and 0.43 with middle point set at 0.21 for α−Fe₂O₃ NP:PA6 ratio. TBAB concentration was altered from 0 wt.% to 1 wt.%. The middle point for TBAB concentration was set at 0.1 wt.% rather than median to study the effect of TBAB more drastically within a narrow range. These samples are listed as Sample 9, 8, 2, 1, and 3 in Table 1. The code for each sample in Table 1 shows what factors (the α−Fe₂O₃ NP:PA6 ratio and TBAB concentration) and the range (the lower, middle, or upper point) have been altered to synthesize each sample.

The second DOE was designed and carried out based on the nanofiber properties and batch testing results of the first DOE. To target a bigger range of average fiber diameters, the concentration of PA6 was increased from 6.5 wt.% to 7.5 wt.%. From the first DOE, samples with higher ratio of α−Fe₂O₃ NP:PA6 ratio had higher phosphate removal efficiency. To test if increasing the α−Fe₂O₃ NP:PA6 ratio further increases the adsorption capacity, the range of α−Fe₂O₃ NP:PA6 ratio was increased from 0 to 0.43 (i.e., the range from the first DOE) to 0.21 to 0.65. For second DOE, the TBAB concentration was kept at 1 wt.%; while addition of TBAB helped to decrease the bead density, the first DOE revealed there was less than 2% difference in phosphate removal efficiency between 0.1 wt.% and 1.0 wt.% TBAB. The samples from the second DOE are listed as Sample 6, 7, 4, 5, and 10 in Table 1.

Morphology of the as-spun nanofiber was observed with a scanning electron microscope (Prisma E SEM, Thermo Fisher Scientific, USA). Prior to analysis, a thin layer of gold was sputtered using Electron Microscopy Sciences 575X over the samples at 20 mA for 30 s to minimize surface charging. Obtained SEM images were imported to ImageJ software to measure the average fiber diameter, which was obtained by measuring the diameter of 30 unique nanofibers. The bead density was calculated by dividing the total number of beads from a single SEM image by the total area of the image. Fiber fraction was determined by the proportion of nanofibers in the total product, which could include beads and clumps. Transmission electron microscopy (TEM) samples were collected by placing a carbon-coated copper grid directly in front of the drum collector for 1 min during electrospinning. TEM images were captured using 300 (S)TEM Ceta™.

The specific surface area and pore volume measurements of the composite nanofibers were obtained using a surface area and pore size analyzer (Quantachrome Nova 4200e) under nitrogen. Specific surface area (S_BET) was determined from multi-point BET, where the relative pressures (P/P₀) ranged from 0.05 ≤ P/P0 ≤ 0.30 (Hesterberg Butzlaff et al., 2023). The t-plot (V-t, where V is the volume of gas adsorbed and t is the statistical film thickness) method was applied to the adsorption isotherm to determine the micropore volume and surface area (Kruk et al., 1996; Galarneau et al., 2014; Daraghmeh et al., 2017).

Mechanical properties were examined using a Discovery hybrid rheometer (DHR-30, TA Instruments, USA) attached with RH Linear Tension Rectangular Fixture. Nanofiber samples were placed between the plates between the plates and the samples were pulled apart at a constant linear rate of 1.0 mm per second until 50 mm is reached at room temperature. All measurements were obtained directly from the manufacturer supplied computer software (TRIOS, TA Instruments).

To evaluate the P remediation efficiency of PA6/α-Fe₂O₃/TBAB nanofiber membranes, adsorption kinetics and isotherm studies were carried out using 40 mg of composite nanofiber membranes in 40 mL of phosphate solution with phosphate concentrations of 2, 5, 10, 15 and 20 mg/L. The experiment was performed in a 50 mL sealed polypropylene centrifugal tube at 23°C. The solution was collected using a 5-mL BD Luer-Lok syringe with a syringe filter attached (0.22 μm, PTFE Teflon filter) at different time intervals (0.25, 0.5, 1, 2, 4, 8 and 24 h). The phosphate concentration of collected samples was measured at 880 nm with a UV-Vis spectrophotometer (Agilent Cary 60) based on the ascorbic acid molybdate blue method (Wang et al., 2021).

The equilibrium adsorption capacity (q_e) was calculated using the following equation (Equation 1): q e = C o − C e ∗ V m (1) where q_e (mg/g) is the adsorption capacity at time t, C₀ and C_e are the initial and equilibrium phosphate concentrations (mg/L), respectively, V is the volume of the solution (L), and m is the mass of the composite nanofiber mat (g).

The adsorption kinetic was investigated using the pseudo first order and pseudo second order equations (Equations 2, 3 respectively). q t = q e ∗ 1 − ⁡ exp   k 1 ∗ t (2) t q t = 1 k 2 q e 2 + t q e (3) where rate constant of pseudo first order adsorption as k₁ (min⁻¹) and k₂ as rate constant of pseudo second order adsorption (g mg⁻¹ min⁻¹); q_e is the amount of phosphate adsorption at equilibrium (mg/g); q_t is the amount of phosphate adsorption at time t (min) in mg/g. Additionally, the adsorption isotherms were investigated using Langmuir and Freundlich adsorption equations (Equations 4, 5 respectively): q e = q max K L C e 1 + K L C e (4) q e = K F C e 1 / n (5) where q_max is the maximum adsorption capacity (mg/g); C_e is the equilibrium phosphate concentration (mg/L); n is the parameter of the Freundlich adsorption isotherm; and K_L (L/mg) and K_F ((mg/g)*(L/mg)^1/n) are the equilibrium constants related to the Langmuir and Freundlich adsorption isotherms, respectively.

Table 1 lists the solution properties and nanofiber morphologies from DOE #1 and #2. The analysis for the DOE #1 of viscosity, electrical conductivity, and surface tension as function of α−Fe₂O₃ NP:PA6 ratio and TBAB concentrations are shown in Supplementary Figure S1. The viscosity of the solution drastically increased from 17.13 cP to 57.29 cP as α−Fe₂O₃ was added and the ratio of α−Fe₂O₃ NP:PA6 was increased to 0.43 without adding TBAB to the solution. The addition of TBAB to solution resulted in a decrease in the viscosity from 17.13 cP to 14.52 cP. The decrease in viscosity is attributed to increase in free charge in polymer phenomenon aligns with observations reported in previous literatures (Mingzheng et al., 2012; Zheng et al., 2014). However, when 1 wt.% of TBAB was added to the electrospinning solution with α−Fe₂O₃ NP:PA6 ratio, the viscosity of the solution increased. Although both factors lead to increase in viscosity of the electrospinning solution, increase in α−Fe₂O₃ NP:PA6 ratio had more significant impact on the viscosity. Based on these results, the equation shown in Supplementary Figure S1A can be used to predict the solution viscosity when the α−Fe₂O₃ NP:PA6 ratio and TBAB concentration are varied.

The electrical conductivity of the solution reflects the charge density on a jet, and the elongation level of a jet by an electrical force (Wendorff et al., 2012; Alves et al., 2013; Ko and Wan, 2014). It was reported in previous literature that an increase in electrical conductivity can result in a thinner fiber diameter of electrospun polymer fibers (Xue et al., 2019; Sun et al., 2014). Low electrical conductivity leads to more beads, resulting from insufficient elongation of a jet by electrical force needed to produce uniform nanofibers (Cooper et al., 2018; Wang et al., 2014; Bhardwaj and Kundu, 2010). The electrical conductivity of the solution was mainly affected by TBAB concentration. Supplementary Figure S1B shows that increasing α−Fe₂O₃ NP:PA6 ratio from 0 to 0.43 decreased the electrical conductivity, which may be explained by increased amount of α−Fe₂O₃, which has poor conductivity. Adding TBAB to the solution sharply increased the electrical conductivity from 1,292 μS/cm to 1862 μS/cm, similar to the trend we previously observed for PAN/Fe₂O₃/TBAB composite solutions (Wang et al., 2021).

Surface tension plays a critical role in the electrospinning process and determines electrospinnability and determines the upper and lower boundaries of the electrospinning window (Haghi and Akbari, 2007). A Taylor cone is produced when a certain threshold voltage exceeds the value of surface tension (Xue et al., 2019). As shown in Supplementary Figure S1C, the surface tension of the as-prepared solution was generally maintained throughout at 21 dynes/cm. While previous literature showed that incorporation of TBAB to the solution decreases the surface tension for other polymers and/or solvents (as expected for a surfactant), notably that was not the case for this solution (Tanvir and Qiao, 2012).

Analysis for DOE #2, where the concentration of PA6 and α−Fe₂O₃ NP:PA6 ratio was varied, can be seen in Supplementary Figure S2. Changing these factors had a bigger impact on the viscosities of the solutions. As the PA6 concentration increased from 6.5 wt.% to 7.5 wt.%, the viscosity of the solution increased, as expected. Chisca et al., reported that an increase in the polymer concentration reduces polymeric chain mobility in solution due to chain entanglement that prevents chain re-ordering and raises the solution viscosity (Chisca et al., 2012). When the α−Fe₂O₃ NP:PA6 ratio was increased from 0.21 to 0.65, a similar trend that seen in DOE #1 was observed, and the viscosity of the solution increased from 65.14 cP to 159.91 cP. Although both factors had a direct effect on the viscosity, when the concentration of PA6 and the ratio of α−Fe₂O₃ NP:PA6 both increased, the combination of factors had the biggest impact on increasing the solution viscosity, which is shown as the steepest slope shown in Supplementary Figure S2A and through the equation describing this relationship (given above the graph) as well. The electrical conductivity of the solution decreased from 1717 μS/cm to 1,673 μS/cm when the PA6 concentration increased. Unlike the DOE #1, when the α−Fe₂O₃ NP:PA6 ratio increased, the electrical conductivity increased slightly.

Figure 1 shows the morphology of electrospun tri-composite nanofibers. As shown in Supplementary Figure S5, tri-composite nanofibers with different composition and fiber diameters ranging from 66 nm to 235 nm were fabricated. For Figures 1A–G, the nanofibers show texture from different amounts of α−Fe₂O₃ NP that have been added to the solution, while Figures 1H, I shows smooth electrospun nanofiber without any texture as the solution did not include α−Fe₂O₃ NP. The rough surfaces appearing in SEM images indicate that α−Fe₂O₃ NP were enriched on the surface of the nanofiber. This was attributed to the interaction between surfactant and α−Fe₂O₃ NP leading to improved dispersion of α−Fe₂O₃ NP.

Before conducting detailed adsorption kinetic and isotherm studies, samples were exposed to a fixed phosphate concentration of 10 mg/L (Supplementary Figure S6). Two additional samples, which are α−Fe₂O₃ NPs (Sample #11) and TBAB coated α−Fe₂O₃ NPs (Sample #12), have been added as control to determine maximum adsorption capacity of pure materials. As shown in the figure, free floating TBAB coated α−Fe₂O₃ NPs show the highest extent of phosphate removal (i.e., 90.35%), whereas pristine PA nanofiber (Sample #9) shows the lowest degree of phosphate removal (i.e., 4.68%). Adding 1 wt.% TBAB to PA6 nanofibers (Sample #8) showed slightly higher phosphate removal (i.e., 16.05%) than pristine PA6 nanofiber, which might be attributed to ion exchange characterized by electrostatic attraction between positively charged quaternary ammonium sites (QAS) and oxyanions (i.e., phosphate) (Nalbandian et al., 2022).Initially, α−Fe₂O₃ NPs: PA6 ratio was increased to test for increased phosphate uptake, however, Sample #12 showed that both TBAB and α−Fe₂O₃ NPs contributed in uptake of phosphate. Increased α−Fe₂O₃ NPs content did not necessarily lead to increased phosphate uptake, and Supplementary Figure S6 shows that there exists an optimal ratio of 8.5:1 between α−Fe₂O₃ NPs and TBAB that lead to maximum phosphate uptake of the tri-composite nanofiber mat.

Using a batch testing system for phosphate removal from the solution helps to establish the time until equilibrium and rate of phosphate removal. Figure 4A shows the phosphate removal batch testing with different initial phosphate concentrations (i.e., 2, 5, 10, 15 and 20 mg/L). The two most popular adsorption kinetic models, the pseudo first order and pseudo second order were used to fit data. Figure 4B, Supplementary Figures S7, S8 show pseudo first order and pseudo second order fittings of data of PA6/α-Fe₂O₃/TBAB nanofiber phosphate removal at different initial phosphate concentrations. Supplementary Table S2 shows summarized kinetic correlation coefficients from the data fitted by pseudo first order and pseudo second order kinetic models. Based on the resulting correlation coefficients (i.e., R² values), it is concluded that PA6/α-Fe₂O₃/TBAB nanofiber is best-described by a pseudo second-order kinetic model, which suggests chemisorption via ligand exchange of negatively charged phosphate ions with iron oxide surface (Suresh Kumar et al., 2017). Pseudo-second order fitting has been used to calculate the k₂, which shows the rate of phosphate adsorption by each of the materials with different contents of α-Fe₂O₃ and TBAB. Supplementary Figure S9 shows kinetic rate constant changes as a function of α-Fe₂O₃:PA6 ratio. It is observed that while keeping the TBAB concentration constant at 1 wt.%, increasing the α-Fe₂O₃ content resulted in a monotonic decrease in k₂. We interpret this behavior based upon pristine PA6 nanofiber having a higher k₂ value of 0.38 g/(mg min) than α-Fe₂O₃ NP, which only exhibits a value of 0.00386 g/(mg min) (see Supplementary Table S2). Therefore, by adding α-Fe₂O₃ into PA6, it is expected that the reaction rate constant will decrease as α-Fe₂O₃ content increases.

Meanwhile, increasing the TBAB concentration from 0 to 1 wt.% contributes to a higher adsorption rate constant of 0.00437 g/(mg min) rather than 0.00716 g/(mg min) while the α−Fe₂O₃ NP:PA6 ratio remained constant (i.e., 0.43) As discussed previously, incorporating a surfactant such as TBAB to α-Fe₂O₃ results in surface enrichment of α-Fe₂O₃ nanoparticle in the composite, where the phosphates will be adsorbed by the ligand-exchange reaction with the positively charged surface that may possibly increase the reaction rate (Wang et al., 2021; Jung et al., 2019).

To further understand the adsorption capacity of these composite nanofibers, nonlinear and linear Langmuir and Freundlich isotherm models were used to fit the experimental data. Figure 4C, Supplementary Figures S10, S11 show nonlinear fit to Langmuir and Freundlich model, linear fit to Langmuir model, and linear fit to Freundlich model, respectively. The Langmuir isotherm model describes homogenous, monolayer adsorption onto the finite adsorptive sites whereas the Freundlich isotherm is based on multilayer adsorption on heterogenous sites (Wang et al., 2021; Li et al., 2016; Katal et al., 2012). Supplementary Table S3 summarizes the obtained data, demonstrating that the synthesized composite nanofiber exhibits result comparable to previous finding presented in Supplementary Table S4. Monolayer adsorption onto the active sites was determined by the higher correlation coefficient (R²) obtained for Langmuir fitting compared to Freundlich fitting. While the experiment was carried out for all the samples, Sample #7 and Sample #10 failed to be fitted using both Langmuir and Freundlich isotherm equations, which may be because of their high α-Fe₂O₃: PA6 ratio. The free-floating particles (as mentioned above) of Sample #11 had q_max of 13.22 mg/g of nanoparticles. Figure 5 shows the 3D and contour plot of adsorption capacity as a function of TBAB and α-Fe₂O₃ content. The contour plot shows that the adsorption capacity increases as TBAB and α-Fe₂O₃ content increases. However, it can be concluded that adsorption capacity of composite nanofiber is more dependent on the α-Fe₂O₃ content rather than the TBAB content. As illustrated in Figure 5, an optimal ratio between α-Fe₂O₃ and TBAB content is observed, corresponding to optimized adsorption capacity.

One of the key requirements for a water filtration system is having excellent mechanical properties to withstand high pressure and flux during application (de-Bashan and Bashan, 2004). To understand the impact of adding chemically active ingredients to pristine PA6 nanofibers, composite materials were investigated with tensile strength testing. Figure 6 shows the stress versus strain curves of different samples. Various parameters such as Young’s modulus, yield strength, ultimate tensile strength, toughness, and strain to fracture extracted from the stress-strain plot and the results obtained are listed in Supplementary Table S5. Results showed that in the absence of α-Fe₂O₃ NPs, addition of TBAB reduced most of the mechanical properties of PA6 nanofibers. Similar results were observed when TBAB concentration increased from 0.1 wt% to 1.0 wt% while maintaining α-Fe₂O₃ content, where the Young’s modulus, yield strength, ultimate tensile strength and toughness decreased but the yield point increased. With an increase in TBAB and α-Fe₂O₃ content, the nanofiber became more brittle. Although this is affected by both factors, the slope of the contour plot in Supplementary Figure S12 shows that TBAB content has a predominant effect on the decrease in toughness of the material. Supplementary Figure S13 shows that when there are no α-Fe₂O₃ nanoparticles present, increasing the content of TBAB resulted in a decrease in Young’s modulus. Furthermore, when there is no TBAB present, increasing α-Fe₂O₃ content also led to a decrease in Young’s modulus; however, when both TBAB and α-Fe₂O₃ are added to produce tri-composite nanofibers, Young’s modulus reaches its maximum when α-Fe₂O₃ content is between 10%–20% and TBAB content is 2%–10% of the composite nanofibers. The decrease in the mechanical properties with increased amount of α-Fe₂O₃ may be due to the inhomogeneous dispersion and aggregation of Fe₂O₃ nanoparticles (Yadav, 2018; Liu et al., 2016). The slope of the contour plot of Supplementary Figure S13 shows that while the Young’s modulus is dependent on both TBAB and α-Fe₂O₃, it is more dependent on the content of α-Fe₂O₃.

Various results were shown in previous literatures in regards of addition of TBAB, such as increasing the tensile strength until 0.66 w/v % but decreasing the tensile strength when TBAB content increases further (Fan et al., 2020). Arora et al. reported that up to 2 wt% TBAB modified clay showed an increase in tensile modulus and tensile strength, but these values decreased when TBAB content was further increased (Arora et al., 2011). In addition, changing the polymeric host results in different mechanical properties. PAN/Fe₂O₃/TBAB tri-composite nanofibers have been synthesized by Wang et al. in our previous work (Wang et al., 2021). For their sample with the highest adsorption capacity for phosphate, Young’s modulus, yield strength, ultimate tensile strength, toughness, and strain to fracture was determined to be 6.13 × 10⁶ Pa, 2.30 × 10⁵ Pa, 2.96 × 10⁵ Pa, 1.46 × 10⁴ J×m⁻³, and 7.10 × 10⁻² ε, respectively. Compared to this work utilizing polyamide 6 instead of PAN, the respective mechanical properties are measured to be 20.51, 16.76, 17.33, 32.33, and 0.27 times higher with PA6, as depicted in Supplementary Table S6. Indeed, it has been reported that the Young’s modulus and tensile strength of nanofibers can depend on several variables such as the chemical structure of polymer, molecular orientation, fiber diameter, fiber fraction, as well as alignment of nanofibers (Lasenko et al., 2023; Pelipenko et al., 2013).