Ultrathin organosiloxane membrane for precision organic solvent nanofiltration

Membrane-based separations can drastically reduce energy use compared to thermal processes, particularly for organic solvent systems prevalent in chemical, petrochemical, and pharmaceutical industries. However, polymeric OSN membranes typically exhibit limited selectivity for small solutes due to chain mobility and swelling, with most demonstrating MWCO around 300 g mol−1 and struggling in the critical 150–300 g mol−1 window. This range is highly relevant to active pharmaceutical ingredients (APIs), such as Acyclovir (225.20 g mol−1) and its prodrugs (e.g., Valacyclovir, 324.34 g mol−1), where high-purity separations are essential. Inorganic membranes face fabrication challenges for thin, defect-free films, while polymer membranes made by interfacial polymerization can lack precise control over thickness/roughness and are prone to fouling. Plasma-enhanced CVD improves stability but can leave residual functionalities that limit precision sieving. The study aims to develop a chemically robust, ultrathin, highly selective membrane capable of precise solute–solute separations within 150–300 g mol−1 under harsh organic solvent conditions, using an initiated CVD (iCVD) approach to form a highly interconnected organosiloxane selective layer.

Prior OSN materials include rigid polymers such as PEEK, PBI, and PIMs, which showed improved solvent stability but typically targeted solute rejections above ~300 g mol−1 due to polymer swellability and limited rigidity. Inorganic systems have difficulty forming ultrathin, defect-free layers. Interfacial polymerization can yield <100 nm polyamide films but with crumpled morphologies and limited control over thickness and roughness, potentially increasing fouling. PECVD has been applied to improve OSN performance but often produces films with partially retained functional groups and complex chemistry, limiting precision separations. Recent advances using conjugated polymers and aligned macrocycles improved selectivity, yet high-precision solute–solute separation in the 150–300 g mol−1 regime remains rare due to polymer relaxation in solvents. These gaps motivate solvent-free iCVD synthesis of a highly interconnected organosiloxane network to achieve uniform nanoscale sieving with chemical robustness.

Support membranes: Mesoporous PAN supports were fabricated via nonsolvent-induced phase separation from a PAN:1,3-dioxolane:DMSO (10:45:45 w/w) dope, cast on nonwoven support (speed 2.5 m/min, thickness 200 μm), coagulated in water, hot-water washed (80 °C, 3 h), rinsed, IPA-exchanged, and dried. Cross-linking was performed in 20% v/v hydrazine monohydrate at 85 °C for 6 h (XP6) or 12 h (XP12), followed by rinsing and drying. Cross-linking reduced pore size and porosity (e.g., PAN mean pore 17.7 nm to 14.8 nm for XP6), forming asymmetric supports with a denser skin.

iCVD selective layer: V4D4 monomer and TBPO initiator were vaporized (flows ~1.61 and 1.62 sccm) and delivered to a custom iCVD reactor (process pressure 200 mTorr, filament 140 °C, substrate 42 °C). Deposition rate was 1.41 nm min−1 with Pm/Psat = 0.093 to favor surface polymerization and prevent deep pore infiltration. Ultrathin poly(1,3,5,7-tetravinyl-1,3,5,7-tetramethylcyclotetrasiloxane) (pV4D4) layers were deposited to target thicknesses ~16–170 nm on XP6 or XP12 supports. Post-deposition DMF activation was performed by permeating DMF for ~24 h to remove unreacted monomers/oligomers from the percolated region within the support, reducing transport resistance.

Characterization: Morphology via SEM (top-view, cross-section) and EDS mapping; porometry for pore size; ToF-SIMS depth profiling and 3D tomography tracking SiCH3+ (pV4D4) and C3H4N+ (XP) to define zones (pV4D4 layer, percolated region, XP bulk) and quantify thickness/etch rates; XPS and FT-IR to confirm organosiloxane composition and polymerization (Si–O–Si, Si–CH3, reduced C=C vs monomer); AFM for roughness (RMS reduced after coating); density and refractive index (XRR and ellipsometry), indicating high cross-link density (n≈1.49–1.50, ρ≈1.70 g cm−3); GIWAXS to probe amorphous halos and intrinsic d-spacings (~4.1 Å and ~7.6 Å) consistent with siloxane ring and alkyl-chain organization.

Performance testing: Gas permeation for integrity (He/N2 ideal selectivity) across thicknesses/supports; organic solvent permeation in cross-flow cells (area 14.2 cm², room temp, typically 30 bar) across MeOH, acetone, DMF, etc., including sequential solvent exposure and DMF activation; pressure-dependent flux tests (10–30 bar). Solute rejection assessed using polystyrene (PS) oligomers (MW 162–370 g mol−1) in MeOH/acetone/DMF; HPLC-UV (254 nm) quantified feed/permeate; MWCO defined at 90% rejection. API mixture (Acyclovir and Valacyclovir in MeOH) analyzed by HPLC under specified mobile phase to determine rejections and solute–solute selectivity. Permeance P = V/(A Δt Δp); transport correlations with Hansen solubility parameters evaluated, notably δp η−1 dm².

• Fabrication: Conformal, ultrathin pV4D4 selective layers (down to 16 nm) formed by iCVD on cross-linked PAN supports. DMF activation effectively cleared percolated monomer/oligomer from support pores, slightly thinning the layer (e.g., 29 to 28 nm) while maintaining adhesion and integrity.
• Structure/composition: ToF-SIMS identified three zones and showed significant reduction of the V4D4-percolated region after DMF activation. XPS/FT-IR confirmed organosiloxane network formation; AFM showed smooth surfaces (RMS reduced). High refractive index (n≈1.49–1.50) and density (≈1.70 g cm−3) indicate a highly interconnected network. GIWAXS revealed amorphous halos with intrinsic transport channels (d-spacings ~4.1 Å and ~7.6 Å).
• Gas transport: He/N2 ideal selectivity increased with thickness, indicating elimination of pinholes (e.g., 7 nm/XP6: 2.69; 16 nm/XP6: 9.10; 55 nm/XP6: 17.88). For 35 nm/XP12, DMF activation decreased He/N2 from 12.35 to 6.38 concomitant with increased permeance, consistent with reduced percolated resistance.
• Solvent permeation and stability: For 29 nm/XP12, MeOH permeance increased from 0.042 to 0.070 L m−2 h−1 bar−1 through MeOH→acetone exposure; after DMF activation, MeOH permeance rose to ~0.42 L m−2 h−1 bar−1. Solvent permeance of 29 nm/XP12-D correlated linearly with δp η−1 dm². MeOH flux was ~11.1 L m−2 h−1 with linear pressure dependence (10–30 bar). The membrane maintained stable DMF flux over multiple days, indicating chemical robustness.
• Sieving performance (PS markers): Support pore size critically impacted selectivity. 30 nm/XP6-D exhibited broad rejection with MWCO > 1000 g mol−1 (non-precision sieving). In contrast, 29 nm/XP12-D showed sharp PS rejection profiles with MWCOs of ~252 g mol−1 (MeOH), ~255 g mol−1 (acetone), and ~369 g mol−1 (DMF). Increasing pV4D4 thickness to 35 nm further enhanced rejections. Rejection profiles were stable over ~24 h.
• Solute–solute selectivity and upper bound: 29 nm/XP12-D achieved record-high solute–solute selectivity of 39.88 for different-sized solutes (PS markers). Trade-off analysis against MeOH permeance set new upper bounds for selectivity in MW ranges 150–250 and 250–350 g mol−1, outperforming conjugated polymers, aligned macrocycles, polyarylate, PIMs, polyamides, graphene oxide, and COF membranes.
• API case study: In MeOH, Acyclovir (225.20 g mol−1) and Valacyclovir (324.34 g mol−1) exhibited rejections of 88.95 ± 3.26% and 99.00 ± 0.53%, respectively, yielding solute–solute selectivity of 11.04 despite <100 g mol−1 MW difference, demonstrating practical precision separation.

The work addresses the longstanding challenge of precise solute–solute separations in the 150–300 g mol−1 range for organic solvent systems, which is highly relevant to API purification. By employing solvent-free iCVD to form an ultrathin, highly interconnected organosiloxane (pV4D4) network on carefully tuned cross-linked PAN supports, the membrane achieves uniform nanoscale sieving with chemical and mechanical robustness. DMF activation is pivotal to remove percolated monomer/oligomer within the support, decreasing transport resistance and revealing intrinsic membrane performance without compromising integrity. The membranes exhibit stability under harsh solvents (including DMF), linear flux–pressure relationships, and sustained performance over time. Structural analyses (ToF-SIMS, XPS/FT-IR, AFM, GIWAXS) confirm a dense, interconnected network with intrinsic channels consistent with observed selectivity. Compared to state-of-the-art OSN materials, the membranes surpass current permeability–selectivity trade-offs in the targeted MW windows and demonstrate practical API separations (Acyclovir/Valacyclovir) with high selectivity, validating the approach for energy-efficient, high-purity separations in pharmaceutical processing.

An ultrathin, highly interconnected organosiloxane selective layer (pV4D4) fabricated by iCVD on optimized cross-linked PAN supports enables precision organic solvent nanofiltration in the previously challenging 150–300 g mol−1 regime. The approach combines precise thickness control, support pore-size optimization, and DMF activation to deliver high permeance with sharp sieving, setting new upper bounds for solute–solute selectivity at practical flux. The membranes demonstrate chemical/mechanical stability across harsh solvents and pressures, and achieve challenging API separations (Acyclovir/Valacyclovir) with high selectivity. Future work could explore copolymerization within the iCVD framework, leveraging diverse monomer libraries to tailor pore size distributions and functionalities, and scale-up strategies for large-area, defect-free production for industrial OSN applications.

• The membrane’s performance relies on a DMF activation step (~24 h) to remove percolated monomers/oligomers; without activation, transport resistance is elevated and gas selectivity differs.
• Slight thickness reduction after activation (e.g., 29 to 28 nm) and decreased He/N2 ideal selectivity for some samples indicate structural changes accompanying activation, though overall separation performance improved.
• PS rejection in DMF was somewhat lower than in MeOH/acetone for MW < ~370 g mol−1, reflecting solvent-dependent selectivity.
• Support morphology is critical; larger surface pores (XP6) led to deep percolation and poor precision sieving (MWCO > 1000 g mol−1), narrowing operational windows for suitable supports.
• Instrumental limitations prevented direct porometry characterization for cross-linking times >12 h; inferences were made from permeation data.
• The study focuses on selected solvents and solutes; broader validation across diverse chemistries and long-term fouling studies remain for future work.