Effect of N‑Methyl-2-pyrrolidone/Dichloromethane Solvent Ratio towards the Retention of Magnesium Sulfate and Ibuprofen by Polylactic Acid Membranes

Abstract

Polylactic acid (PLA) membranes are sustainable alternatives to petroleum-based polymers, yet inherent fragility often necessitates leachable additives for mechanical robustness that may contaminate the permeate. Cost-effective PLA 4032D pellets were used in this work to fabricate additive-free membranes via an evaporation-assisted non-solvent-induced phase separation (E-NIPS), employing N-methyl-2-pyrrolidone/dichloromethane (NMP/DCM) blends as a tuning parameter. E-NIPS implementation decoupled solvent evaporation from liquid–liquid demixing, facilitating a robust, high-crystallinity asymmetric structure. Optimized solvent ratio improved performance stability, yielding higher pure-water flux and enhanced rejection of ibuprofen and magnesium sulfate, chosen as representative pharmaceutical and salt contaminants. At an optimal 10:1 NMP/DCM ratio, membranes achieved a tensile strength of 9.65 ± 1.09 MPa and thermal stability onset near 345 °C. Differential scanning calorimetry confirmed that delayed demixing in NMP-rich formulations provided a kinetic window for ordered lamellae formation, significantly increasing crystallinity relative to DCM-rich amorphous structures. Additionally, 30 week soil burial tests revealed a controlled biodegradation profile regulated by the high-crystallinity matrix. These findings establish a sustainable framework for engineering high-performance, biodegradable membranes for pharmaceutical and salt removal.

1. Introduction

Membrane technology is increasingly pivotal across diverse applications such as water treatment, biotechnology, and food processing, driven by low energy consumption and operational simplicity. However, conventional membranes, typically fabricated from petroleum-based polymers like poly­(ether sulfone) (PES) and polyvinylidene fluoride (PVDF), pose a significant sustainability paradox. These materials are inherently non-biodegradable, resulting in substantial biohazardous waste that ends up in landfills or contributes to emissions through incineration, creating difficulties in universal waste management and a critical environmental challenge. This has catalyzed an urgent need to develop compostable, ecofriendly, and sustainable alternatives. Polylactic acid (PLA), a biodegradable aliphatic polymer derived from renewable resources, has emerged as a potential bio-derived polymer to address these concerns. PLA offers compatibility with established fabrication methods and reduced environmental footprint. To utilize PLA effectively in demanding processes such as nanofiltration (NF) and ultrafiltration (UF), new membrane structure must demonstrate mechanical and chemical robustness comparable to conventional petroleum-based polymers.

Despite the material advantages, the application of pristine PLA in membrane separation is severely limited by intrinsic material and structural trade-offs that must be overcome. PLA membranes suffer from intrinsic brittleness and limited permeability, which restrict their practical application. A pervasive technical challenge in designing high-flux polymeric membranes is the intrinsic trade-off between permeability and mechanical strength. High permeability requires high porosity, which often translates into fragile structures susceptible to failure under operational pressure differentials. For porous membranes, high-flux architectures often exhibit low yield stresses, typically ranging from 2 to 8 MPa, a range generally insufficient for high-pressure NF applications, which require substantial structural integrity.

To overcome the low intrinsic permeability and hydrophobicity of pristine PLA, researchers commonly employ chemical pore-forming agents (PFAs), such as polyethylene glycol (PEG) or polyvinylpyrrolidone (PVP), or incorporate inorganic fillers like ceramic nanomaterials. These additives are designed to increase porosity and enhance surface wettability, thereby improving flux and mitigating fouling. However, the reliance on chemical additives introduces long-term limitations. The major drawbacks are the risk of additives leaching out of the polymer matrix during operations. − This leaching phenomenon not only compromises the primary goal of water purification by potentially contaminating the filtrate but also leads to the inevitable collapse of the previously additive-stabilized pore structure, causing significant performance decay and structural instability. This structural and chemical instability drives a fundamental research gap with the need to develop entirely additive-free PLA membranes that achieve stable, tunable, and high-performance morphology purely through precise process control.

Recent attempts to address the limitations of additive-dependent membranes have explored additive-free PLA architectures using bio-based “green” solvents, such as Cyrene, DMI, or methyl lactate. However, these typically yielding modest pure water fluxes ranging from 10 to 390 L·m^–2·h^–1 without demonstrating significant rejections for low-molecular-weight solutes. Furthermore, binary solvent systems utilizing NMP/DMAc have been employed to fabricate asymmetric ultrafiltration membranes. However, the binary solvent systems are generally limited to low-flux operations while targeting high-molecular-weight solutes, such as bovine serum albumin (BSA) or organic substances. Consequently, a critical need remains for a fabrication strategy that achieves high permeability and high selectivity for low-molecular-weight solutes (ibuprofen and MgSO₄ in this work) using the PLA material.

This study strategically targets high-crystallinity PLA 4032D, a commercial grade notable for its low d-lactide content. This specific material selection is central to overcoming the inherent mechanical weakness of PLA structures. − PLA 4032D with high crystallinity exhibits low solubility in many common solvents used in non-solvent induced phase separation (NIPS). This processing bottleneck severely restricts the formulation window for conventional NIPS methods, hindering the reproducible fabrication of high-performance membranes from this mechanically advantageous material. To realize the full potential of PLA 4032D, a refined solvent strategy is mandatory to overcome this solubility limitation while simultaneously enabling precise morphological control without relying on leachable additives. The necessity of developing an additive-free process capable of handling PLA 4032D with low solubility mandates the use of a sophisticated co-solvent strategy via evaporation-assisted non-solvent-induced phase separation (E-NIPS). This study employs a binary solvent system consisting of N-methyl-2-pyrrolidone (NMP) and dichloromethane (DCM). This specific pairing is selected due to their highly contrasting physicochemical properties. Specifically, volatility and water miscibility allow the decoupled kinetic control necessary to engineer the asymmetric membrane structure.

The novelty of this study lies in the strategic implementation of an evaporation-assisted non-solvent-induced phase separation (E-NIPS) mechanism to overcome the structural instabilities inherent in conventional NIPS when chemical additives are omitted. In prior PLA studies, DCM has been used solely as a primary dissolution agent, a role in which its uncontrolled evaporation leads to premature vitrification and non-porous films. Current study redefines the volatile DCM as a kinetic crystallization modulator in which the low DCM fractions employed in E-NIPS (7.0 to 15.4 wt %) may quantitatively constrain pre-immersion enrichment of the surface polymer layer within the strictly maintained 15 s transfer window, rather than triggering bulk solidification as occurs when DCM dominates the solvent system. By leveraging the high vapor pressure of DCM, a controlled pre-immersion evaporation phase is established, which stabilizes the polymer-rich phase at the solution–air interface before contact with the coagulation bath. This approach effectively decouples the kinetics of solvent evaporation from the thermodynamics of liquid–liquid demixing, allowing for the precise tuning of an asymmetric structure with a finely porous selective layer.

NMP, characterized by high polarity and low volatility, delays phase separation due to its high boiling point, maintains sufficient polymer-chain mobility, and facilitates the partial alignment of PLA chains necessary for forming a uniform surface layer. The combined effects of delayed phase separation (by NMP) and regulated crystallization (by DCM) promote the formation of a sponge-like substructure and a thinner, more uniform selective skin layer. This combined solvent interaction potentially improves mechanical robustness, minimizes defect formation, and enhances water flux, thereby yielding a structurally stable and high-performance PLA membrane.

In this study, the effect of NMP/DCM solvent ratio on the fabrication of additive-free PLA 4032D membranes via E-NIPS is investigated. Membrane morphology, surface roughness, porosity, mechanical strength, and thermal stability are examined, along with separation performance for magnesium sulfate and ibuprofen (chosen as model solutes in water treatment). This research aims to systematically overcoming the traditional permeability–retention trade-off that dominates the additive-free polymeric membrane research. Through the optimization of the NMP/DCM solvent ratio, the potential for concurrent enhancements in both permeability and rejection efficiencies for low-molecular-weight solutes can be demonstrated. These advancements could represent a significant departure from previously reported PLA-based membranes, which have largely targeted contaminants with molecular weights exceeding 500 g·mol^–1. By establishing an additive-free fabrication pathway using commercial-grade PLA 4032D, this study provides a sustainable framework for high-performance membranes tailored for pharmaceutical and desalination pretreatment applications.

2. Methodology

The materials used in this study has been detailed in Supplementary Document S1.

2.1. Fabrication of PLA/NMP/DCM Membrane

The membrane fabrication process is illustrated in Figure . The temperature gradient between the heated dope solution (70 °C) and the ambient casting environment and coagulation bath (≈25 °C) drives a controlled (pre-immersion evaporation phase that defines the E-NIPS mechanism).

1.

Schematic representation of PLA membrane fabrication via the E-NIPS method.

A 23 wt % PLA dope was prepared by dissolving PLA pellets in a binary solvent mixture of NMP and DCM. The NMP/DCM compositions were adjusted according to Table . The 23 wt % PLA dope solution was selected after preliminary screening and used as the basis for subsequent membrane modification. The suspension was stirred at 350 rpm and 70 °C for 5 h to obtain a homogeneous solution in hermetically sealed vessels. After thermal equilibration at 70 °C, the solution was cast onto a clean glass plate at ambient room temperature (≈25°) using a 400 μm casting knife.

1. Weight Composition of Binary Solvent System (NMP/DCM) for PLA Membrane Fabrication with Corresponding Membrane Labelling.

membrane	NMP to DCM ratio	NMP (wt %)	DCM (wt %)	PLA (wt %)
M0	0:1	--	77.0	23
M1	4:1	61.6	15.4	23
M2	5:1	64.2	12.8	23
M3	6:1	65.9	11.1	23
M4	7:1	67.4	9.6	23
M5	8:1	68.4	8.6	23
M6	9:1	69.3	7.7	23
M7	10:1	70.0	7.0	23
M8	1:0	77.0	--	23

The nascent film was then immersed in a distilled-water coagulation bath maintained at room-temperature (25 ± 2 °C) and relative humidity of around 45%. The transfer time between casting and immersion was strictly maintained at 15 s in order to maintain consistency and minimize uncontrolled evaporation. The membrane was then gently removed, rinsed, and stored in distilled water until further analysis. The procedures were systematically employed for the fabrication of all membranes.

E-NIPS utilizes a strictly maintained 15 s transfer time to facilitate a controlled pre-immersion evaporation phase. This phase leverages the volatility of DCM in the binary system to produce a measurable but self-limiting enrichment (DCM constitutes only a minor fraction of the total solvent) of the polymer-rich layer at the solution–air interface. Once the surface DCM is evaporated during the 15 s window, the residual NMP-dominated matrix retains sufficient chain mobility for ordered lamellae formation upon immersion. The viscosity spike observed at M4 marks the upper composition boundary at which this balance is lost and densification occurred.

2.2. Membrane Characterization and Membrane Performance Evaluation

The details of the membrane characterization and membrane performance evaluation are provided in the Supplementary Document S1.

3. Results and Discussion

3.1. Membrane Morphologies and Wettability Studies

SEM micrographs M0-M8 show the PLA membranes prepared using NMP/DCM solvent ratios. (A) Provides a representative annotated SEM image identifying and labelling the structural regions.

The SEM cross sections in Figure demonstrate the impact of the binary solvent ratio on PLA membrane morphology, shifting from dense structures to asymmetric porous networks based on NMP/DCM concentrations. Increasing the NMP fraction shifts the membrane morphology from a dense and vitrified state (M0) through a transition of macrovoid-dominated morphologies toward an interconnected sponge-like substructure (M8).

2.

Cross-sectional SEM images of PLA membranes fabricated with varying NMP/DCM solvent ratios, illustrating structural evolution and internal morphology.

The kinetic pathways underlying these morphological transitions are summarized in Figure . The high volatility of DCM drives significant pre-immersion evaporation, which acts as the primary kinetic trigger to induce instantaneous surface vitrification in DCM-rich formulations like M0. This densification is driven by the low boiling point of DCM, causing significant pre-evaporation that seals the solution–air interface prior to the coagulation bath. In contrast, the higher boiling point of NMP favors delayed demixing, allowing for slower mass transfer that yields a robust sponge-like substructure beneath a controlled skin layer. Consequently, membrane cast with intermediate DCM fractions such as M3 develop finger-like macrovoid morphologies dominated by large irregular cavities embedded within a partly vitrified matrix, whereas the pure NMP membrane M8 exhibits a thin skin layer supported by a continuous and well-developed sponge-like substructure.

3.

Conceptual schematic of the kinetic pathways in E-NIPS as a function of the NMP/DCM ratio, illustrating the transition from instantaneous demixing (M0–M3) and vitrification-induced arrest (M4) to delayed demixing (M5–M8) for various membrane morphologies.

Unlike conventional NIPS, which frequently leads to structural collapse or dense vitrification in PLA 4032D, the E-NIPS mechanism promotes an asymmetric architecture with a sponge-like substructure. This controlled evaporation step enables a unique combination of high hydraulic permeability and superior solute rejection. The synergy between delayed demixing and regulated crystallization ensures mechanical robustness that standard NIPS pathways often fail to achieve in additive-free systems.

Notably, an intermediate solvent ratio of 7:1 (M4) triggers a critical morphological transition where pore channels collapse into a virtually sealed surface. This phenomenon is attributed to a vitrification-induced kinetic transition where polymer chains are immobilized by maximum local viscosity at the interface before pore formation can occur. While this is similar to the models reported for PVDF, the transition in the PLA 4032D system is distinctly governed by its aliphatic polyester chemistry and high crystallinity. The latter narrows the solubility window significantly, making the transition to a dense state (R _crit) more abrupt in the NMP/DCM blend compared to fluorinated polymers. ,, This assignment is confirmed by the viscosity maximum, the cold crystallization behaviour, and the retention data presented in the following section.

A clear correlations between the solvent composition and membrane surface roughness was established by analysing the topography of the PLA membranes fabricated from the NMP/DCM binary solvent system using AFM, as illustrated in Figure . The AFM images reveal that M1, which contains the lowest NMP fraction (highest relative DCM content) among the mixed-solvent formulations, exhibits the highest surface roughness (R _a ≈ 45.4 nm) with prominent peaks and valleys. This topology indicates a rapid, kinetically driven phase separation. Such accelerated demixing is primarily induced by the high volatility and low boiling point of DCM, which promotes rapid solvent exchange and the precipitation of a heterogeneous polymers structure, eventually resulting in a distinctly uneven surface topography. Conversely, as the NMP content increases (from M2 to M3), surface roughness progressively decreases. NMP, characterized by its lower volatility and higher boiling point, promotes a delayed demixing process. This slower solvent exchange rate is critical, as it maintains polymer chain mobility rearrangement and organization, enabling the formation of a more uniform surface layer and, consequently, a smoother topography.

4.

(a) Three-dimensional atomic force microscopy (AFM) images of PLA membranes fabricated with varying NMP/DCM solvent ratios with the respective surface roughness values (R _a) are shown for each sample. (b) Quantitative comparison of surface roughness (R _a) across all eight membranes, with values presented as mean ± standard deviation.

Further increases in NMP content yielded smoother surfaces, as observed in M4 (R _a = 14.4 nm), M5 (R _a = 11.4 nm), M6 (R _a = 6.8 nm), and M7 (R _a = 4.5 nm). Consequently, the PLA membrane fabricated using pure NMP exhibited the smoothest surface among all samples, with an R _a value of 2.4 nm. The mean roughness decreased significantly from M1 to M4, reflecting a transition from instantaneous to delayed demixing as the predominant formation mechanism. This aligns with literature reports that have documented similar surface topography changes in PVDF membranes fabricated using an acetone/DMF solvent system. Specifically, a transition from instantaneous to delayed de-mixing occurs as the solvent ratios shift, significantly impacting the final surface morphology.

The wettability of the fabricated membranes is primarily influenced by surface roughness, as described by two fundamental models, the Cassie–Baxter theory for enhanced hydrophobicity and Wenzel’s theory for roughness-induced wetting enhancement. In this study, the PLA membranes exhibited moderate hydrophobicity, with static water contact angles ranging from 57.8° to 82.3°. Notably, the experimental results revealed a trend where decreasing surface roughness correlated with an increasing contact angle, as shown in Figure . According to Wenzel, increased surface roughness enhances the inherent wetting behaviour of a material by expanding the contact area between the solid and the liquid. Since the intrinsic contact angle of flat PLA is typically below 90°, the Wenzel theory predicts that increased roughness should reduce the apparent contact angle, thereby enhancing hydrophilicity. The direction of the observed trend is therefore consistent with this prediction.

5.

Wettability of binary solvent-cast PLA membranes.

However, the magnitude of the contact angle increase observed for the smoothest membranes exceeds what roughness reduction alone can account for under the Wenzel framework, suggesting a transition toward a mixed or non-Wenzel wetting regime superimposed on this trend. This behavior is driven by solvent-ratio-induced changes in phase inversion kinetics and surface chemical orientation. Specifically, the delayed demixing process associated with higher NMP content evidenced by the transition from finger-like to sponge-like substructures, provides additional time for polymer rearrangement during membrane formation. This slower precipitation allows PLA chains to reorient at the interface, possibly exposing a slightly higher density of hydrophobic methyl groups to the water phase, thereby potentially increasing the intrinsic hydrophobicity of the membrane surface. Quantitatively, the inversed correlation between R _a and water contact angle indicates that the interfacial energy could also be caused by chemical composition rather than physical topography alone. This suggests that the delayed demixing kinetics in NMP-rich formulations (M5–M7) provide a sufficient kinetic window for the low-surface-energy methyl groups of the PLA chains to segregate toward the solution–air interface.

Concurrently, delayed phase inversion produces a denser and more uniform film while promoting the formation of isolated polymer nodules at the microscale. Although the overall roughness (R _a) based on AFM data decreases, these discrete surface features could trap air pockets beneath the water droplet, triggering a Cassie–Baxter wetting response. This localized air entrapment, combined with the chemically enhanced hydrophobic skin layer, produces a pronounced increase in contact angle despite the reduction in macroscopic roughness. The resulting wetting state is therefore governed by a synergistic interplay between enhanced intrinsic hydrophobicity and micro-structural features capable of supporting air-liquid interfaces. The observed increase in static contact angle on smoother topographies serves as a reliable macroscopic indicator of localized air entrapment in this study.

The pore architecture further reinforces this interpretation. M7 exhibits the highest contact angle, displays discrete micro-scale topographical features such as polymer nodules and shallow depressions that can support metastable air pockets at the liquid interface, hindering capillary-driven wetting. In contrast, M1 displays a continuous and densely packed surface with smaller asperities that promote direct liquid contact and suppress hydrophobic behaviour. The selective skin pores responsible for sieving are addressed separately in Section .

Previous studies provide the theoretical framework for understanding the wetting state transitions, confirming that the Wenzel and Cassie Baxter regimes result in fundamentally different interfacial properties. In this research, the shift from a Wenzel to a mixed Cassie–Baxter regime is demonstrated by the contrasting formation kinetics of M1 and M7. For M1, the high DCM fraction triggers rapid and kinetically-driven phase separation, yielding a rough surface (R _a = 45.4 nm) where the topography enhances the inherent wetting behavior. This aligns with the Wenzel model, resulting in a relatively low contact angle of 69.3°. Conversely, the high NMP fraction in M7 provides a wider kinetic window via delayed demixing. This delay allowed the chemical reorientation of PLA chains, while the formation of larger surface-adjacent openings facilitated localized air entrapment. These factors activate a Cassie–Baxter response, yielding a significantly higher contact angle of 82.3° despite a much smoother macroscopic topography (R _a = 4.5 nm). Thus, the binary solvent ratio significantly alters the thermodynamic stability of the liquid–solid interface, inducing a transition to a composite interface. Although M8 exhibits the smoothest macroscopic topography and the highest bulk porosity, its highly interconnected sponge-like substructure permits water ingress that progressively displaces air from the surface features, eliminating the Cassie–Baxter contribution and lowering the apparent contact angle. The discrete and partially isolated surface features of M7 are therefore more effective at stabilising the metastable composite interface than the continuous open network of M8.

3.2. Membrane Porosity

The concentrations of NMP and DCM in the solvent mixture significantly impact the porosity of the fabricated PLA membranes. As shown in Figure a, membrane porosity increases substantially with rising NMP fractions, from approximately 22% at a 4:1 NMP/DCM ratio to roughly 48% at 10:1. Observations from Figure b indicate that PLA membranes fabricated from pure DCM exhibit a translucent appearance whereas membranes fabricated from pure NMP appear opaque (Figure c). Literature reports have demonstrated that introducing as little as 10 wt % of NMP as a non-solvent during phase inversion was sufficient to induce a uniform pore structure in PLA membranes, highlighting the critical role of NMP concentration in governing membrane morphology. Higher NMP/DCM ratios yield membranes with more uniform pore structure, as the delayed demixing induced by NMP retards phase separation, allowing for more ordered polymer precipitation. Consequently, pores are distributed evenly throughout the matrix without the structural collapse typically associated with instantaneous demixing. These trends align with findings in other polymeric systems. For instance, polycarbonate membranes fabricated using NMP exhibit a highly porous and sponge-like substructure, while those prepared with DCM exhibit a non-porous architecture. In contrast to DCM-rich solutions, which frequently result in a dense skin layer and sparse pores, the data consistently indicate that NMP-rich solvent environments promote the development of larger and more numerous pores.

6.

(a) Porosity of PLA membranes under varying solvent compositions; (b) transparent PLA membrane fabricated using pure DCM; (c) white appearance PLA membrane fabricated using pure NMP.

Consistent with the morphological observations, the thermodynamic and kinetic variations in phase separation during E-NIPS account for these alterations in porosity. DCM tends to induce instantaneous demixing, causing localised phase separation that can suppress pore growth beneath the surface and result in a dense and polymer-rich skin layer. According to the National Institute of Standards and Technology (NIST) database, DCM is only sparingly soluble in water, with a solubility of approximately 17.5 g L^–1 at 25 °C, and it becomes increasingly immiscible at elevated temperatures, decreasing to 5.2 g L^–1 at 60 °C. In this E-NIPS framework, the rapid evaporation of DCM serves as the primary kinetic trigger for instantaneous surface demixing via an accelerated local polymer concentration gradient. This contrasts with conventional NIPS, where instantaneous demixing is typically governed by high solvent–non-solvent affinity.

The choice of co-solvent significantly impacts both the kinetics and thermodynamics of phase inversion. Consequently, DCM-rich solutions result in membranes with suppressed microvoids and lower overall porosity. In contrast, increasing the NMP proportion significantly increases the thermodynamic stability of the casting solution, a critical determinant of final porosity. Although the low diffusion rate of NMP out of the film induces delayed liquid–liquid demixing (a kinetic effect), leading to a symmetric sponge-like morphology, the enhanced thermodynamic stability exerts a dominant effect on the final void volume. Specifically, greater stability creates a stronger thermodynamic driving force for phase separation. This drive maximizes the formation of the polymer-poor phase, leading to an increase in overall pore volume and porosity. The resulting sponge structure is characterized by well-developed and interconnected pores, despite the absence of large macrovoids. Membrane M4 showed a dense skin layer formation due to vitrification, which avoided the formation of continuous pores even with high overall porosity. In this system, the solvent ratio acts as the primary modulator of surface enrichment. Unlike conventional systems where an air gap drives evaporation, the high vapor pressure of DCM ensures that the polymer concentration at the surface increases immediately upon casting. Thus, the variations observed between M1 and M8 effectively capture the kinetic transitions that would otherwise be induced by varying evaporation residence times.

3.3. Mechanical Strength

The E-NIPS kinetics are significantly altered by increasing the DCM content in the NMP/DCM solvent mixture, leading to variations in membrane mechanical strength as illustrated in Figure . The instantaneous demixing induced by DCM can trigger premature vitrification of the polymer-rich phase, wherein the polymer matrix solidifies before the pores can fully develop. A similar reduction in the mean pore size, permeance, and mechanical strength was documented in the literature for PVDF membranes due to the incorporation of water into the PVDF/N,N-dimethylacetamide (DMAc) dope solution, aligning with the trends observed in this study.

7.

Effect of solvent composition on tensile strength of PLA membranes.

As revealed by the SEM analysis, the formation of an open pore network is suppressed by the arrested phase separation, where these collapsed or incomplete pores serve as structural weak points, decreasing the mechanical strength of membranes cast with a high DCM content. This trend is evident in the figure, where the M1, M2, and M3, which were fabricated with higher DCM fractions, exhibited lower mechanical strengths of 6.02, 7.01, and 7.64 MPa, respectively. These “arrested” morphologies contain collapsed pores that act as localized stress concentrators, preventing effective load distribution. In summary, a high DCM fraction causes a pore-deficient, mechanically fragile structure by accelerating solidification via vitrification. The mechanical profile for the membranes, including the raw load-extension curve, stress–strain curve, and summarized tensile properties (eg. maximum tensile stress, maximum tensile strain, modulus), are provided in the Supplementary Document S2.

Consistent with literature, a reduction in the volatile-solvent fraction (DCM) leads to increased membrane porosity and pore development, characteristic of a delayed demixing process. As observed in the SEM analysis, membranes prepared with a high NMP concentration tend to produce smaller macrovoids at the sponge layers, thereby increasing the tensile strength of the PLA matrix. This is consistent with the increase in mechanical strength for M5, M6, and M7 as the NMP proportion rises. The increasing tensile strength for NMP-rich blends (M5–M7) is attributed to a wider kinetic window that facilitates crystalline development that highlights a fundamental difference between PLA and other semicrystalline systems like PVDF. Unlike the amorphous “frozen” states of DCM-rich membranes, the delayed demixing allows PLA 4032D chains to organize into ordered lamellae. This results in a higher baseline mechanical robustness than typically reported for PVDF-based ultrafiltration membranes of similar porosity. −

The vitrification-induced transition using the E-NIPS demixing could create a highly cohesive and dense structure by freezing polymer chains into a glassy matrix before the pores are fully connected. This mechanical transition is a direct consequence of the solvent-ratio-induced vitrification as discussed in morphology analysis. Thus, the elevated tensile strength of the M4 PLA membrane despite the entirely different underlying mechanism, can be attributed to such a vitrified morphology where immobilised chains within a compact and pore-deficient matrix provide mechanical cohesion in the absence of crystalline reinforcement. Similarly, literature reports indicate that the densification occurred when amorphous silica are partially vitrified, increasing mechanical strength while adversely affecting porosity due to the suppression of small pores. In brief, the tensile strengths of PLA membranes fabricated in this work are generally higher than those reported in the literature (which are only around 1.6 ± 0.2 to 4.4 ± 0.9 MPa) and are highly suitable for filtration processes. , Notably, membranes fabricated from pure solvents followed the conventional porosity–strength relationship. The dense but brittle DCM membrane (M0) exhibited relatively high strength due to its compact morphology, whereas the highly porous NMP membrane (M8) showed the lowest mechanical strength. Conversely, membranes fabricated with binary solvent mixtures have more uniform sponge-like structures formed in NMP-rich blends, allowing higher porosity to coexist with improved mechanical strength. Membrane M8 confirms that the absence of DCM’s kinetic modulation leads to hyper-porosity. The well-developed sponge-like substructure of M8 is morphologically continuous but mechanically weak, as the excessive void volume reduces the load-bearing cross-section and lowers the tensile strength to the minimum of the series.

3.4. Differential Scanning Calorimetry (DSC) Analysis

The degree of crystallinity (X _C) of the PLA membranes was analysed to elucidate the relationship between surface evaporation kinetics and bulk thermodynamic miscibility as illustrated in Figure . As the solvent system shifted from DCM rich to NMP rich composition, a progressive change in crystalline behaviour was observed, characterized by a progressive decrease in the cold crystallization peak temperature (T _CC) and a corresponding increase in the degree of crystallinity (X _C) as the NMP fraction rises. DCM-rich membranes, which vitrify prematurely, display large T _CC peaks indicative of high residual amorphous content, whereas NMP-rich membranes exhibit smaller T _CC peaks and larger melting endotherms, reflecting greater crystallinity.

8.

Degree of crystallinity (X _C) for PLA membranes fabricated with varying NMP/DCM ratios.

For the membranes fabricated with higher NMP content (M5 to M8), the DSC thermograms showed distinct melting endotherms, corresponding to higher crystallinity values and notably lower cold crystallization peaks where these data are provided in Supplementary Document S3. This enhancement in crystallinity is attributed to the delayed demixing mechanism. NMP has low volatility and high-water miscibility that led to the delayed polymer vitrification, providing a sufficient kinetic window for the PLA 4032D chains to organize into ordered lamellae during the solidification process. Crystallographic data corroborates the SEM observations where membranes with high crystallinity displayed a uniform and interconnected sponge like morphology. In the optimized sample M7, the high crystallinity led to the high tensile strength of 9.65 MPa, which is in good agreement with another work. However, membrane fabricated with pure NMP (M8) showed a drop in mechanical strength (5.8 MPa) despite its high crystallinity. This indicates that an excessive delay in the absence of DCM could lead to hyper porosity. This structurally compromises the membrane despite the crystalline nature.

Membranes cast with higher DCM ratios M1 to M3 exhibited a high vitrification level and significantly higher cold crystallization peaks. The high volatility of DCM drives the rapid surface evaporation, which acts as a kinetic trigger to increase the local polymer concentration instantly. This causes the polymer rich phase to solidify into a disordered state before the membrane pores can be fully developed, a process which is defined as premature vitrification. This phenomenon led to the formation of large and irregular finger like macrovoids with a dense and glassy skin layer. The absence of toughening effect of crystallinity led to the formation of amorphous and defect rich structures having lower tensile strength values.

The DSC thermogram for M4 revealed the highest and sharpest cold crystallization peak among all samples, confirming that this membrane was prepared in a predominantly amorphous and vitrified state. The large T _CC peak indicates that PLA chains were kinetically arrested into a disordered glassy matrix before crystalline order could develop. This is a direct consequence of vitrification-induced pore collapse rather than crystalline reinforcement. The high T _CC peak signifies low crystallinity, which distinguishes M4 fundamentally from the NMP-rich membranes (M5–M7) where low T _CC peaks reflect genuine semicrystalline character. At this specific intermediate ratio, the dope is expected to attain a maximum local viscosity. This severely hindered solvent nonsolvent exchange and caused the matrix to solidify before the pores could be fully formed. The membrane M4 retained a tensile strength of 9.26 MPa, comparable to the optimized M7 membrane, although the underlying reinforcing mechanism differs. In M4 the strength derives from physical immobilisation of polymer chains within a compact and pore-deficient matrix, whereas in M7 the strength derives from ordered lamellae formation enabled by the delayed demixing kinetic window. , The similar tensile values therefore mask fundamentally different microstructural origins.

In summary, the DSC analysis reveals two structurally distinct groups. The NMP-rich membranes (M5–M7) are semicrystalline with high X _C values enabled by the delayed demixing kinetic window. This crystallinity is the mechanistic origin of the simultaneous high permeability and high solute rejection. In contrast, M4 is amorphous-vitrified, which is mechanically dense and strong, but functionally non-permeable. Besides, M0–M3 are also amorphous but structurally defective.

3.5. Functional Group Analysis

The FTIR spectra acquired in this study are essentially aligned with the peer studies, with the observed peaks are tabulated in Supplementary Document S4 (Table S4.1) in conjunction with the other two literature for comparative purposes. Characteristic PLA absorption bands are present across all samples, and no additional functional groups were detected. Notably, the bending vibration associated with the terminal hydroxyl groups is absent in this study. This conforms to expectations for the high-molecular-weight PLA grade employed (i.e. Ingeo 4032D). This finding is supported by literature indicating that high-molecular-weight PLA possesses a relatively low content of terminal hydroxyl groups, making these fingerprints generally undetectable in the FTIR spectrum. Despite significant morphological and mechanical alterations, the choice of solvents system for PLA membrane fabrication did not affect the chemical structure or functional groups of the PLA, as evidenced by identical peak positions across all membranes. This is corroborated by previous research reporting that NMP as the non-solvent yields a porous film without shifting the FTIR spectra relative to pristine PLA, while pure DCM produces dense structures that likewise display only typical PLA peaks. This bulk spectroscopic signature reflects the chemical composition of the membrane and does not preclude localized differences in surface molecular orientation.

FTIR analysis enables the identification of residual solvent within the PLA membrane. NMP exhibits a weak N–H stretch at approximately 3400 cm^–1 and a distinct CO band near 1675 cm^–1, yet no membrane spectrum displays a discernible amide band between 1650 and 1675 cm^–1, as shown in Figure S4.1a (Supplementary Document S4), indicating that NMP was successfully removed during the phase inversion process. Furthermore, no detectable C–Cl absorption was observed near 600–700 cm^–1, suggesting that any residual DCM content is below the FTIR detection threshold. In summary, the FTIR results show no signs of residual NMP or DCM, confirming effective solvent removal from the fabricated PLA membranes.

3.6. Thermal Stability

The TGA follows the expected thermal profile of nearly pure PLA, wherein all membranes revealed a single-step thermal degradation under nitrogen. Significant weight loss commences above approximately 300 °C, and the peak rate of decomposition falls within the range of 320 °C–350 °C. By this point, about 90–100% of the PLA mass is lost near 500 °C, leaving only a trace amount of residual char. The single-step degradation beginning at nearly 320 °C matches pristine PLA, with negligible weight loss below 100 °C, confirming minimal residual solvent. This aligns with the FTIR analysis, which showed no distinct peaks corresponding to solvent residues for the membrane fabricated with pure NMP and DCM. The additive-free membranes is verified by single-step degradation profiles and residual masses below 7%, which is consistent with near-complete volatilization and negligible char characteristic of neat PLA. This result distinguishes the fabricated membranes from filled or stabilized systems, which typically exhibit significantly higher residual weight fractions. Comparative study indicates that polymeric membranes incorporating inorganic additives, such as basalt fibers, exhibit significantly higher residual values of approximately 30 wt % owing to the limited thermal decomposition.

TGA data indicate shifts in the thermal stability depending on the solvent system. Membranes fabricated with a higher proportion of DCM tend to decompose at slightly lower temperatures, whereas samples cast with a higher amount of NMP fractions exhibit higher degradation temperatures. This phenomenon implies that increased DCM concentrations could promote a more amorphous or defect-rich microstructure, thereby reducing thermal stability. A reduction in PLA thermal stability is typically observed with the addition of plasticizer or increased amorphous content. The total weight loss in all cases approaches 90–100%, and the final residue at 600 °C is less than 1% residual char.

The thermal stability trend stratifies the membranes into two regimes. The NMP-rich membranes M6, M7, and M8 exhibit elevated T _onset values of 324.89 to 348.30 °C, consistent with the dense lamellar packing established by the delayed demixing pathway. In contrast, the DCM-rich membranes M0 to M3 and the transitional membranes (M4 and M5) cluster within a lower range of 302.88–315.52 °C, reflecting the amorphous and defect-rich microstructure generated by premature vitrification. The slight depression of M5 relative to M6 indicates that the kinetic window for ordered lamellae formation widens progressively beyond the 8:1 ratio rather than activating sharply at M5.

In summary, the slight variations in T _max among samples are consistent with trends reported in the literature, with pure PLA having a T _max of around 350 °C and a shift of T _max towards 330 °C observed with blends or additives (e.g. poly­(butylene adipate-co-terephthalate, PBAT)). This is consistent with the behaviour of PLA membranes fabricated using a higher proportion of DCM in this study. Additionally, the very low residual masses across all samples confirm complete PLA degradation. , Recent PLA-based membranes reported in the literature exhibit earlier thermal degradation with T _onset ≈ 285–314 °C and DTG_max ≈ 326–346 °C. The observed peak degradation temperature (T _P ≈ 350 °C) is consistent with the upper range of reported values for high-crystallinity PLA grades. While numerical shifts in T _onset across the solvent series suggest a trend toward reduced stability at higher DCM fractions, these variations remain within the typical experimental margin for commercial PLA. Therefore, the performance of the M7 membrane reflects the successful preservation of the polymer’s intrinsic thermal robustness.

3.7. Dope Solution Viscosity

The viscosity of the M0 dope solution was measured within the temperature range of 25–60 °C to prevent rapid DCM evaporation, whereas formulations containing NMP (M1–M8) were analyzed from 65–95 °C to ensure the fluidity of these highly viscous solutions. Rheological measurements (Figure ) indicate that the dope viscosity generally decreases as the NMP fraction increases, with the notable exception where a pronounced viscosity peak observed at the M4 composition. This anomaly suggests stronger polymer–polymer interactions or a thermodynamic stabilization at that intermediate ratio. In practical terms, the high viscosity at M4 severely hinders solvent–nonsolvent exchange, contributing to the vitrification effect mentioned earlier.

9.

Viscosity of PLA dope solutions with (a) varying NMP/DCM ratios (M1–M7) and pure NMP solvent (M8) measured at 65–95 °C, and (b) pure DCM solvent (M0) measured at 25–60 °C.

NMP-rich dopes (which possesses lower viscosity) exhibit delayed demixing due to the strong affinity of NMP for the polymer, which stabilizes the polymer-rich phase and slows down water ingress. Conversely, DCM-rich dopes, which are generally more viscous at room temperature, undergo much faster demixing at the film–bath interface. This is attributed to DCM’s volatility and poor water miscibility, which promote rapid solvent removal. This underscores that the demixing rate is not governed by viscosity alone, but by a combination of rheology, polymer–solvent interactions, and solvent volatility.

The zero-flux behavior of M4 is elucidated by a vitrification-induced kinetic transition occurring at a critical thermodynamic point. At this intermediate composition, the dope reaches a maximum local viscosity where the synergistic solvency of the binary system is at a local minimum for the high-crystallinity PLA 4032D. This “viscosity spike” implies that the binary solvent mixture approaches a solvency limit, triggering thermal chain aggregation and enhanced entanglement that outpaces polymer–solvent interactions. Unlike PVDF systems where vitrification often follows rapid non-solvent ingress, the specific PLA/NMP interaction acts as a stronger thermodynamic stabilizer. In the M4 formulation, the severe imbalance between the rapid evaporation of high-vapor-pressure DCM and the low diffusion rate of NMP creates a localized “polymer-rich” environment. This effect is particularly pronounced in PLA 4032D due to its lower chain flexibility, leading to an immediate “freeze” into a glassy matrix before pores can nucleate.

This accelerated vitrification kinetics ensure the polymer-rich phase solidifies before appreciable pore growth or liquid–liquid demixing can occur, producing a dense, non-permeable skin. Beyond this vitrification point, increasing the NMP ratio (M5–M8) lowers the dope viscosity and shifts the system toward a delayed demixing pathway. While demixing remains slower than in DCM-rich systems, the increased NMP content restores the kinetic window necessary for polymer chains to organize into ordered lamellae. This transition promotes the development of more open and sponge-like substructures that restore hydraulic flux while maintaining high selectivity. In contrast, membranes prepared from DCM-rich dopes (M1–M3) undergo much faster demixing, favoring macrovoid formation at the sublayer where the balance between permeability and selectivity is governed by the continuity and thickness of the resulting top skin layer. ,

In summary, the superior performance of the NMP-rich membranes (M5–M7) and the unique “zero-flux” state of M4 can be explicitly articulated through the interplay of solvent volatility, polymer–solvent affinity, and demixing kinetics. The high affinity of NMP for PLA 4032D coupled with its low volatility shifts the phase inversion toward a delayed demixing pathway. This delay provides the necessary kinetic window for polymer chains to organize into ordered lamellae, creating a sponge-like substructure that simultaneously enables high hydraulic flux and high solute rejection. These dynamics demonstrate that the binary solvent ratio is not merely a composition variable, but the primary modulator of the membrane’s final anisotropic architecture.

3.8. Molecular Weight Cut Off (MWCO) and Pore Size Distribution

As shown in Figure , the retention profiles obtained with PEG tracers from 1000 to 10,000 Da stratify the fabricated membranes into a sequence that closely mirrors the morphological transitions and the kinetic pathways as discussed previously. , Membrane M7 attains the tightest cut off of the entire series at below 1000 Da, with retention rising from 95.4% at 1000 Da through 98.6% at 4000 Da and reaching 99.6% at 10,000 Da.

10.

Retention of PEG of varying molecular weight by PLA membranes fabricated with different NMP to DCM solvent ratios.

Applying the empirical PEG correlation, the M7 cut off of below 1000 Da corresponds to a representative upper bound pore diameter of approximately 1.57 nm. This value estimates the largest pore class that governs sieving in the selective skin, based on the Stokes radius of the PEG tracer at the 90% retention threshold. The hydrodynamic diameters of ibuprofen (approximately 0.7 nm) and MgSO₄ (approximately 0.8 nm) are both smaller than this upper bound, yet both are rejected at efficiencies exceeding 90% by M7. , This indicates that the rejection in M7 is not governed by pore size alone, since both solutes are smaller than the 1.57 nm cut off. The high rejection is more likely driven by the long and tortuous pore pathways within the dense and high crystallinity skin formed under delayed demixing, where extended diffusion lengths and frequent constrictions intercept solutes even when the local opening is wider than the solute itself.

The shape of each retention curve reflects the underlying phase inversion pathway. A sharp rise in retention over a narrow molecular weight span indicates a uniform pore population, while a gradual rise indicates a broad one. M7 retains 95.4 percent of PEG 1000 Da and 99.6 percent of PEG 10000 Da, confirming a tight and uniform selective skin formed under delayed demixing conditions. M6 and M5 follows the same trend at progressively looser cut offs of approximately 1400 and 3200 Da, while M3 reaches approximately 5800 Da, reflecting the systematic contraction of the kinetic window as the DCM fraction rises. M1 and M2 do not reach a cut off within the tested range, consistent with the defect populated skin formed by premature vitrification of DCM rich dopes. Membranes M0 and M4 return no measurable tracer retention because both compositions are in the vitrification arrested non permeable state, while M8 displays the lowest retention curve among all permeable membranes, rising only from 27.9% at 1000 Da to 59.4% at 10,000 Da, confirming that hyper porosity alone does not confer selectivity.

The physical sieving interpretation is further supported by the post filtration FTIR spectra in Supplementary Document S4 Figure S4.1b. The spectra were acquired after a light distilled water rinse, yet the characteristic ibuprofen signature at approximately 1710 cm^–1 is absent from all permeable membranes. This indicates that ibuprofen was removed along with a reversible cake layer rather than fully adsorbed into the PLA matrix, confirming that rejection is governed mainly by steric exclusion at the selective skin rather than chemical binding. , The reversibility of the cake layer is operationally advantageous, since flux can be recovered by low energy hydraulic backwash without aggressive chemical cleaning. For MgSO₄, steric exclusion alone cannot fully account for the 91.38% rejection at M7, because the hydrated ion pair at approximately 0.8 nm is smaller than the 1.57 nm cut off diameter. The residual retention of MgSO₄ beyond the steric contribution is attributed to additional rejection mechanisms acting in parallel with size exclusion. The high rejection of both SO₄ and Mg²⁺ at M7 is consistent with the known behaviour of polymeric membranes in which divalent ion rejection exceeds what steric hindrance alone can account for, and the electroneutrality constraint requires that retained anion and the co-rejected cation permeate together. These observations establish a dual-mode separation in M7, where size exclusion governs the rejection of the neutral organic probe while additional mechanisms contribute to the rejection of the divalent salt pair.

3.9. PLA Membrane Performance

The separation performance of the PLA membranes is significantly influenced by the NMP/DCM ratio. Figure illustrates the pure water flux (PWF) and rejection rates for MgSO₄ and ibuprofen, revealing that both rejection and PWF increase steadily with higher NMP proportion. The calibration curves and limit of detection values used to determine the permeate concentration MgSO₄ and ibuprofen are provided in Supplementary Document S5. The selected ibuprofen and MgSO₄ demonstrated the membrane’s ability to reject low molecular-weight species (approximately 120.4 to 206.3 g mol^–1), addressing a research gap where previous PLA-based membranes primarily targeted larger molecules. These two solutes were further selected to represent contrasting rejection regimes including a divalent electrolyte and a neutral organic acid which is consistent with the established practice of using representative model solutes to isolate a fabrication variable ahead of a broader contaminant panel. While environmental pharmaceutical concentrations are typically in the ng/L to μg/L range, a 2000 ppm feed was deliberately selected to represent worst-case industrial discharge scenarios such as pharmaceutical manufacturing effluents, where ibuprofen concentrations can reach tens to hundreds of mg/L.

11.

Effect of binary solvent composition on (a) salt and ibuprofen rejection, and (b) pure water flux of PLA membranes fabricated using varying NMP/DCM ratios.

Salt rejection increases from a modest range of 50.41–72.34% (for M1–M3) to 91.38% (for M7) at a higher NMP/DCM ratio. Simultaneously, ibuprofen rejection increases from 59.99 to 79.8% (for M1–M3) to 99.77% (for M7). These two solutes allow for a strategic comparison of separation mechanisms beyond molecular weight. For ibuprofen, high rejection is primarily attributed to steric exclusion provided by the dense and high-crystallinity selective skin layer. This mechanism is confirmed by post-filtration FTIR analysis (Figure b), which is identical to the pristine spectrum. The absence of characteristic ibuprofen peaks (such as the CO stretch at ∼1710 cm^–1) suggests that removal is a physical sieving process rather than adsorption. The molecular dimension of this bulky organic acid exceeds the effective pore size of the M7 matrix. Conversely, the rejection of divalent MgSO₄ ions is influenced by the other rejection mechanism.

Membrane M4 remained non-permeable as expected. At this composition of 7:1 NMP/DCM, the structure is “frozen” into a dense and non-permeable configuration, effectively sealing the surface. Beyond this point, as the NMP fraction increases further (M5–M7), membrane flux increases (from 34.8 to 114.9 L·m^–2·h^–1) while maintaining the high solute rejection. The recovery in flux is driven by a shift towards delayed liquid–liquid demixing. NMP with a low volatility provides a wider kinetic window, facilitating a highly porous, continuous sponge-like substructure that enables efficient water transport. Simultaneously, the slower exchange favors a more uniform and defect-free selective layer, which enhances the effectiveness of size-exclusion.

These trends can be correlated with the phase-inversion morphology. As indicated by the SEM analysis, M1 to M3 exhibit lower effective porosity in terms of continuous flow channels, leading to reduced rejection, whereas M5 to M7 develop a thin and continuous selective skin layer supported by a high-porosity sponge substructure with a narrower pore size distribution. The skin layer governs solute rejection, while the underlying high-porosity substructure provides low-resistance pathways for water transport. These observations align with membrane separation principles, which state that flux is governed by porosity, whereas the rejection is governed by pore structure. The vitrification of M4 prevents formation of continuous pore structure, resulting in low water flux. This vitrification-limited state is well documented in polymer membrane studies. Once the polymer-rich phase solidifies early, further coarsening stops, trapping a dense skin that is essentially non-permeable.

The simultaneous increase in both permeability and rejection in NMP-rich membranes arises from the anisotropic structure. A porous sponge layer provides low-resistance pathways for bulk water flow, while a thin, dense skin layer formed early in the demixing process supplies strong size-exclusion capability. Although DCM-rich membranes have lower contact angles, poorly connected pore networks dominate transport resistance and suppress flux.

These results demonstrate that permeability and selectivity increase concurrently rather than trading off across the solvent series. Pure water flux rises from 34.8 to 114.9 Lm^–2 h^–1 between M1 and M7, over the same range in which MgSO₄ rejection increases from 50.41 to 91.38% and ibuprofen rejection from 59.99 to 99.77%. Both properties therefore improve by approximately three-fold together, rather than one improving at the expense of the other. This behavior departs from the conventional permeability selectivity trade-off, in which higher flux is typically obtained at the cost of reduced rejection, and indicates that optimization of the binary solvent ratio simultaneously enhances water transport and size exclusion in the additive free PLA membrane. The fouling conditions of all the membranes are presented in the Supplementary Document S6 after exposure to BSA as a model foulant.

3.10. Biodegradability Test

The soil burial profiles presented in Figure confirm the inherent biodegradability of all additive-free PLA 4032D membranes under the ISO 17556 protocol, with a progressive mass loss recorded across the full test duration. The extent of biodegradation stratified into two clearly distinct regimes that correlate with the solvent-induced structural differences rather than with bulk porosity or thickness alone. The DCM-rich membranes M1, M2, and M3 exhibited the most pronounced mass loss of approximately 14.0, 12.5, and 10.8% by the end of the monitoring period, while the optimized NMP-rich membrane M7 retained the highest resistance to degradation with only 4.2% weight loss over the same interval. The pure-solvent end members M0 and M8 occupied intermediate positions at 9.7 and 6.2%, whereas the vitrified membrane M4 recorded a low degradation extent of 4.9% despite its amorphous nature. A short lag phase common to all samples during the initial weeks is consistent with the well-documented two-step degradation mechanism of PLA, in which abiotic hydrolysis of the ester backbone must proceed before microbial assimilation of the oligomeric fragments can drive measurable mass loss.

12.

Degree of biodegradation of PLA membranes (M0 to M8) fabricated with varying NMP/DCM solvent ratios as a function of soil burial duration.

The observation cross-validates the crystallographic and thermal findings in this work. The NMP-rich membranes M5 to M7, which attained the highest degree of crystallinity owing to the delayed demixing kinetic window that permitted ordered lamellae formation, consistently resisted soil degradation more effectively than the DCM-rich counterparts. This trend agrees with the established understanding that hydrolytic and enzymatic attack of PLA proceeds preferentially through the amorphous regions, as the densely packed crystalline lamellae sterically hinder water diffusion and restrict depolymerase accessibility. The accelerated mass loss of M1 to M3 therefore reflects the amorphous and defect-rich microstructure generated by the premature vitrification pathway, an interpretation further supported by the lower thermal decomposition onset of 304 to 315 °C reported in Table for these same samples. Membrane M4 represents a notable outlier because its vitrified and pore-deficient skin exposes very little internal surface to soil moisture and microbial colonization, producing a hydrolysis rate comparable to the highly crystalline membranes. This observation aligns with prior reports demonstrating that degradation kinetics in porous PLA are co-regulated by accessible surface area and pore architecture rather than by crystallinity alone. Consequently, the slowest degradation registered by M7 establishes the 10:1 NMP/DCM formulation as the optimum balance between operational longevity and controlled end-of-life composability, bridging the biodegradation outcome with the mechanical robustness of 9.65 MPa and the thermal stability onset near 345 °C already established for the optimized membrane in the preceding sections.

2. TGA Characteristics Temperatures (T _onset, T ₁₀, and T _max) and Residual Weight of Additive-Free PLA 4032D Membranes.

membranes	T onset (°C)	T 10 (°C)	T max (°C)	residual mass (%)
M0	313.36	292.54	332.06	0.94
M1	311.36	284.17	332.54	2.85
M2	315.52	297.89	332.71	3.17
M3	304.13	294.81	333.76	3.28
M4	308.12	298.96	320.02	5.83
M5	302.88	289.94	339.50	3.95
M6	324.89	310.30	346.03	6.67
M7	345.33	325.48	358.44	3.26
M8	348.30	333.09	360.01	2.14

3.11. Comparison to Peer Studies

Table contextualizes the performance of the present membranes against recently reported PLA-based systems. The critical differentiator of this work is the simultaneous achievement of high rejection for low-molecular weight solutes (MgSO₄ at ∼120 g mol^–1 and ibuprofen at ∼206 g mol^–1) and meaningful permeate flux (114.9 L·m^–2·h^–1), without any leachable additive or inorganic filler. Prior additive-free PLA membranes using green solvents achieved flux in the 10–390 L·m^–2·h^–1 range but demonstrated no qualified rejection for dissolved species below 500 g mol^–1. Systems that do report high rejection for small solutes rely on functional additives such as nanotubes, graphene oxide, or hydroxyapatite which introduce the leaching risks this work specifically targets to eliminate. The high flux in Table (greater than 1000 L·m^–2·h^–1) are optimized for macroscale oil-water separation or high-molecular-weight macromolecule rejection and are not directly comparable targets. The specific niche this work occupies is the additive-free, high solute rejection at meaningful flux which is unaddressed in the prior literature.

3. Performance Comparison of PLA-Based Membranes Fabricated by NIPS in Recent Peer Studies.

solvent(s)	additives	pure water flux (L m–2.h–1)	rejection (contaminant)–molecular weight	reference(s)
NMP	hydroxyapatite (HAp) filler and PDA coating	359 ± 76 to 788 ± 58 (pristine PLA to PLA + Hap/PDA)	87% (NOM) to 500–2000 g mol–1
DMAc	positively charged functionalized multi-walled carbon nanotube/graphene oxide (f-MWCNT/GO) nanohybrid	24.3 to 106.7 (0 to 6 wt % nanohybrid)	90.1% (NH4 +–N); 71.3% (PO4 3P) raw wastewater. ∼18 g mol–1 (NH4 +–N) and 95 g mol–1 (PO4 3P)
methyl lactate	poly(ethylene oxide), PEO (pore former)	>10,000	99.9% (dextran) to 70,000 g mol–1
Cyr, DMI, EL	none	10 to 390	N/A
NMP/DMAc	none	25 to 250	92% (BSA); 89% COD to 66,500 g mol–1 (BSA) and 180 g mol–1 (COD)
DMF/chloroform	halloysite nanotubes (HNTs)	1330.9 to 3566.5	94.9% (oil) to 600–900 g mol–1
DCM	green banana peel (BP) nanoparticles	105.31	95.2% (100 ppm emulsions)
			96.3% (300 ppm emulsions)
DCM	hesperidin (HSP) nanoparticles	220.5	99.15% (300 ppm oils)	,
NMP/DCM	none	34.8 to 114.9	91.38% (MgSO4); 99.77% (ibuprofen) ∼120.4 g mol–1 (MgSO4) and 206.3 g mol–1 (ibuprofen)	this study

Based on the comparative analysis in Table , the PLA membranes fabricated in this study demonstrated superior rejection capabilities for low-molecular-weight solutes. The proposed binary-solvent fabrication method is straightforward and additive-free. Previous studies that focused exclusively on selectivity alone (often at the expense of flux) further illustrate the typical trade-offs. For instance, literature has reported 20% PLA UF membranes that achieved ∼92% BSA and 89% COD removal, but only under low-flux conditions. Additionally, other investigations embedded 1.5% positively charged MWCNT/GO into PLA membranes and attained ∼90.1% NH₄ ⁺-N and 71.3% PO₄ ^3–-P removal, yet the highest water flux was only 24–107 L·m^–2·h^–1. A comparative analysis highlights the distinctive technical merits of the present research. While the modification of polylactic acid (PLA)/poly­(butylene adipate-co-terephthalate) (PBAT) blends with hesperidin or banana peel additives improves hydrophilicity and flux for oily wastewater treatment, these membranes are largely optimized for the rejection of macro-scale oil emulsions. ,

The key novelty of this work lies in achieving the separation results solely through the optimization of the NMP/DCM binary solvent ratio, eliminating the need for nanoparticles or leachable additives. This approach not only ensures a chemically stable filtration process but also results in superior mechanical robustness, with a tensile strength of 9.65 MPa, which exceeds the reported values for both banana peel-modified (8.4 MPa) and hesperidin-modified (7.5 MPa) variants. This tensile strength also exceeds the 2 to 6 MPa range typically reported for conventional PES and PVDF phase inversion membranes as shown in Table , while M7 membrane matches the PWF and salt rejection through a single step route. Unlike commercial polyamide membranes such as NF90 and NF270 that rely on a multilayer composite construction with a separate interfacial polymerization step, M7 attains comparable low molecular weight solute rejection without any composite selective layer, filler, or leachable additives.

4. Performance Comparison of the Optimized PLA Membrane with Commercial and Conventional NF and Loose NF Membranes.

membrane material	fabrication strategy	PWF (L·m–2 h–1 bar–1)	solute rejection (contaminant)	tensile strength (MPa)	references
polyamide TFC (NF90)	interfacial polymerization on polysulfone support	17 ± 1.3	98% (MgSO4)	N/A
			90–96% (NaCl)
polyamide TFC (NF270)	interfacial polymerization on polysulfone support	9.0 ± 1.1	98% (MgSO4)	N/A
			50% (NaCl)
PES (ultrafiltration to loose NF)	NIPS with PVP pore former	642	85–98% (BSA)	5.2
PES/SPES NF	NIPS with Ti3C2T x	18.5	>90% (metanil yellow and methylene blue); ≈5% (NaCl); ≈12.5% (Na2SO4); ≈16% (MgSO4)	3.5–7	68
PVDF (loose NF)	NIPS with plasma surface modification	>700	87% (alginate)	N/A	69

The selection of DCM at 7.0 to 15.4 wt % of the casting solution is justified primarily by its unique kinetic function rather than by solvent-class preference. No currently available bio-based or green solvent combines the vapor pressure and low boiling point necessary to generate the controlled pre-immersion surface enrichment that enables additive-free processing of high-crystallinity PLA 4032D within a practical 15 s window. The FTIR and TGA results confirm that DCM is effectively removed during phase inversion, yielding a solvent-free membrane product. From a process engineering perspective, DCM low boiling point makes closed-loop vapor recovery technically straightforward, and its recovery energy requirement is substantially lower, supporting the net process sustainability of the fabrication route. , Furthermore, its volatility acts as a kinetic trigger to stabilize high-crystallinity PLA without the need for toxic and leachable additives. , Analytical verification via FTIR and TGA confirms a solvent-free product, maintaining the membrane’s inherent high-performance capabilities. Similarly, the use of NMP is justified by its unique ability to induce high-performance membrane morphology that bio-based solvents currently cannot replicate at room temperature. Its high boiling point and low vapor pressure facilitate a closed-loop recovery system, minimizing environmental discharge through controlled distillation. By ensuring complete polymer dissolution and precise phase inversion kinetics, NMP eliminates the need for energy-intensive heating or auxiliary chemical stabilizers. This technical necessity establishes a performance benchmark for future bio-based PLA membranes, providing a foundational step toward reducing the carbon footprint of filtration technology while maintaining material integrity and recyclability.

4. Conclusions

This study systematically examined the influence of varying the binary solvent ratio (NMP/DCM) on the structural, mechanical, thermal, and filtration properties of additive-free PLA membranes prepared through E-NIPS. Adjusting the NMP/DCM ratio produced distinct changes in morphology, porosity, crystallinity, tensile strength, and contact angle, which directly influenced permeability and solute rejection. Notably, the utilization of E-NIPS proved superior to conventional NIPS by enabling a controlled pre-immersion evaporation phase. Morphological comparisons indicate that E-NIPS effectively suppresses the disordered macrovoid formation typically associated with standard immersion precipitation. An optimal ratio of 10:1 yielded membranes with tensile strength of 9.65 MPa, porosity of 47.77%, and thermal stability onset near 345°, while achieving pure water flux of 114.9 L·m^–2·h^–1 together with a rejection of 91.4% for magnesium sulfate and 99.7% for ibuprofen. It also possessed a surface roughness of 4.501 nm, while water contact angle of 82.3° was measured. This behaviour is notable because rejection and flux improved simultaneously, contrary to the conventional permeability–rejection trade-off that dominates membrane research. Biodegradability assessments revealed a controlled degradation profile over 30 weeks, where the high-crystallinity matrix regulated abiotic hydrolysis to maintain operational integrity before eventual mass loss. By using PLA 4032D pellets and avoiding additives or fillers, this approach offers both economic and environmental advantages. Future work is proposed to assess the extended filtration operation, fouling study, and evaluate a wider panel of pharmaceutical of contrasting charge and hydrophobicity, alongside a simulated or real wastewater matrix, building on the rejection mechanism established in this study. Such testing typically constitutes a dedicated investigation given the distinct solute interaction and fouling effects present in real matrices.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.6c07439.

• Materials, full characterization and performance testing protocols, covering membrane morphology, contact angle, functional group analysis, mechanical properties, porosity, differential scanning calorimetry, thermogravimetric analysis, rheological measurement, molecular weight cut off, pore size distribution, biodegradability, permeability, separation performance, antifouling evaluation; load extension and stress strain curves of the membranes; differential scanning calorimetry thermograms, calculated thermal properties and degree of crystallinity of the membranes; Fourier transform infrared spectra of cleaned and fouled membranes with peak assignments; calibration curves and limit of detection calculations for magnesium sulfate and ibuprofen; flux recovery ratio, total fouling ratio, and irreversible fouling ratio of the membranes after exposure to bovine serum albumin (PDF)

The authors declare no competing financial interest.