Ultraselective Macrocycle Membranes for Pharmaceutical Ingredients Separation in Organic Solvents

MaterialsPolyacrylonitrile porous flat-sheet membranes cast on polypropylene nonwoven support were purchased from GMT Membrantechnik GmbH, Germany. Two commercial OSN membranes, GMT-oNF-2 and DuraMem® 150, were respectively supplied by GMT Membrantechnik GmbH and Evonik. GMT-oNF-2 is a TFC membrane with an apolar rubbery polydimethylsiloxane-based active layer and a reported MWCO of 350 g mol–1. DuraMem®150 is a crosslinked polyimide with a lower reported MWCO of 150 g mol–1. 4,4’-(Ethane-1,2-diylbis(oxy))dianiline (ethoxyaniline) (99%) and 4,4’-(diamino)-1,4:3,6-dianhydro-2,5-di-O-phenyl-D-sorbitol (isosorbide) (>98%) were purchased from Luminescence Technology Corporation. Hydrazine monohydrate (98%), trimesoyl chloride (TMC) (>98%), m-phenylene diamine (MPD) (99%), and sodium metabisulfite (97%) were all purchased from Sigma Aldrich. n-Hexane (97%), N,N-dimethylacetamide (DMAc) (>99.9%), 2,2,2-trifluoroethanol (TFE) (98.8%), and HPLC-grade acetonitrile were purchased from VWR. Deionized (DI) water used in all experiments was filtered through a Millipore Milli-Q water purification system.Synthesis of bis(aminobenzo)−18-crown-6 (18C6)The monomer, bis(aminobenzo)−18-crown-6 (18C6), was synthesized according to the following procedure (Supplementary Fig. 1). Dibenzo-18-crown-6 (I) (20 g, 1 eq, 55 mmol) was dissolved in chloroform (104 mL). Acetic acid (78 mL) was added to the solution over 10 min, which was then stirred at room temperature for an additional 5 min. A solution of nitric acid (II) (6.3 g, 4.5 mL, 1.8 eq, 0.10 mol) in acetic acid (10 mL) was added dropwise over 15–20 min. The solution was stirred at room temperature for 1 h and then heated to reflux overnight, whereupon a precipitate formed. The solution was allowed to cool to room temperature. After 48 h, the precipitation was filtered, and the compound was isolated as a pale yellow solid. Residual acetic acid was removed by dissolving the sample in dimethylformamide (DMF), followed by the addition of water to precipitate pure bis(nitrobenzo)−18-crown-6 (III) (2.7 g, 6.0 mmol, 11%). Supplementary Fig. 2 shows the 1H NMR and 13C NMR for bis(nitrobenzo)−18-crown-6 in deuterated dimethyl sulfoxide (DMSO-d). 1H NMR (400 MHz, DMSO-d) δ 3.85 (m, 8H), 4.21 (m, 8H), 7.15 (d, 2H, J = 9 Hz), 7.72 (d, 2H, J = 2.7 Hz), 7.89 (dd, 2H, J = 9, 2.7 Hz). 13C NMR (75 MHz, DMSO-d) δ 153.8, 147.7, 140.6, 117.6, 111.3, 106.6, 68.4, 68.0.Subsequently, the intermediate product (III) (20 g, 1 eq, 44 mmol) was suspended in ethanol (800 mL), and hydrazine (IV) (14 g, 14 mL, 10 eq, 0.44 mol) was added. After 10 minutes of stirring, the reaction mixture was heated to reflux. After 30 minutes, the reaction mixture was filtered while hot, then cooled, and the product crystallized as a white compound. The product was purified by recrystallization from ethanol. Supplementary Fig. 3 shows the 1H NMR and 13C NMR for 18C6. 1H NMR (400 MHz, DMSO) δ 6.63 (d, J = 8.5 Hz, 1H), 6.27–6.22 (m, 1H), 6.06 (dd, J = 8.5, 2.4 Hz, 1H), 3.95 (dtd, J = 17.4, 5.7, 3.0 Hz, 4H), 3.79 (dq, J = 18.1, 5.0 Hz, 5H). 13C NMR (101 MHz, DMSO) δ 149.64, 149.53, 143.94, 143.87, 139.61, 139.53, 116.05, 115.86, 105.72, 101.17, 101.09, 69.87, 69.71, 69.65, 69.52, 68.13, 68.04.Crosslinking of the polyacrylonitrile porous supportThe polyacrylonitrile support was crosslinked according to a previously published protocol48. To enhance its chemical stability, the support was crosslinked at 85 °C for 6 h in 20% (v/v) hydrazine hydrate in DI water. The crosslinked membranes were washed and stored in DI water to remove the excess crosslinker and maintain their wettability.Solubility of 18C6 and analogous ethyleneoxy-based monomersThe ethyleneoxy-based monomers, shown in Table 2, have poor solubility in water at neutral pH. Therefore, the monomers were dissolved in a co-solvent system containing different volume ratios of an organic solvent to water. Initial solubility studies revealed that all three monomers are soluble in polar aprotic solvents such as dimethyl sulfoxide (DMSO), N,N-dimethylacetamide (DMAc), and N,N-dimethylformamide (DMF). 18C6 and ethoxyaniline are also soluble in fluorinated alcohols such as 2,2,2-trifluoroethanol (TFE). The volume ratio of the organic solvent to water was determined at a fixed concentration of 20.5 mM for each monomer. The monomers were first dissolved in the organic solvents and sonicated for 2 min. Then, water was slowly added and sonicated for another 10 min. These aqueous solutions were then filtered through a 0.20 µm syringe filter and used to prepare thin-film composite membranes on top of crosslinked polyacrylonitrile supports.Table 2 Chemical properties and solubility of 18C6 and analogous ethyleneoxy-based monomersInterfacial polymerizationThe interfacial polymerization method was used to make polyamide composite membranes (Fig. 1). In this method, an amine monomer and an acyl chloride monomer are dissolved in immiscible solvents, and at their interface, the polymerization takes place, forming a thin film on top of a microporous support. The wet, crosslinked polyacrylonitrile support was allowed to dry at room temperature for 30 min. The pristine support was mounted on a Teflon plate and frame set-up and impregnated with 10 mL of 20.5 mM 18C6 aqueous solution for 1 min (Table 2). After removing the excess solution with a custom-made air knife, the membrane’s surface was exposed to 10 mL of 2.5 mM (0.1 wt.%) TMC in hexane for 1 min. The membrane was afterward washed with 10 mL of hexane, heat treated at 80 °C for 5 min, washed with DI water, and then stored in a 1000 ppm sodium aqueous metabisulfite solution at 4 °C. The support impregnation and interfacial polymerization reaction were carried out at room temperature.Control membranes were fabricated following the same method, but the aqueous phase monomer was replaced by either 20.5 mM ethoxyaniline or 20.5 mM isosorbide (Table 2). Classical polyamide membranes were also prepared from 188.7 mM (2 wt.%) MPD in water as the aqueous phase solution.Fabrication of freestanding filmsFreestanding films were collected from the aqueous-organic interface, as shown in Supplementary Fig. 4. 10 mL of aqueous and organic phase solutions were added to a glass Petri dish and allowed to react for 1 min. The films were afterward collected from the interface on a substrate, rinsed with hexane to remove residual acyl chloride, floated on the water, and picked up on either a silicon wafer, a gold-coated silicon wafer, or an aluminum oxide anodisc filter (pore size 0.02 µm). A 24 h reaction time was carried out for samples used for Fourier Transform Infrared (FTIR) measurements. The films were collected from the interface, washed with hexane and water, and then dried overnight at 80 °C.Characterization methodsThe liquid phase nuclear magnetic resonance (NMR) measurements were recorded on a Bruker Avance III operating at a 400 MHz resonance frequency. Fourier Transform Infrared (FTIR) spectra were acquired using a Nicolet iS10 spectrometer (Thermo Fisher Scientific) for wavenumbers between 4000–400 cm−1. The FTIR spectra for the TFC membranes were obtained in attenuated total reflectance (ATR) mode with 16 scans and a resolution of 4 cm–1. The FTIR spectra for the powder samples were acquired in transmission mode on potassium bromide pellets (0.7 wt.%) with 16 scans and a resolution of 4 cm–1. X-ray photoelectron spectroscopy (XPS) studies were carried out in a Kratos Axis Supra DLD spectrometer equipped with a monochromatic Al Kα X-ray source (hν = 1486.6 eV) operating at 150 W, a multi-channel plate, and a delay line detector under a vacuum of ~10−9 mbar. All spectra were recorded using an aperture slot of 300 μm × 700 μm. Survey spectra were collected using a pass energy of 160 eV and a step size of 1 eV. A pass energy of 20 eV and a step size of 0.1 eV were used for the high-resolution spectra. Scanning electron microscopy (SEM) images were acquired on a Magellan field-emission scanning electron microscope with an accelerating voltage of 3 kV and a working distance of 4 mm. The membranes were cut and mounted on an aluminum stub using conductive aluminum tape. The cross-section samples were fractured in liquid nitrogen. To avoid surface charging, a 6 nm thick coating of iridium was sputter-coated using Quorum Technologies Q150T under an argon atmosphere. Atomic force microscopy (AFM) images were obtained on a Dimension ICON scanning probe microscope under tapping mode in air to analyze the 3D morphologies and roughness of the membranes. The root mean square roughness (RMS) was calculated for the height profile of each 5 μm × 5 μm sample. To study the surface-wetting nature of the membranes, water contact angles were evaluated via the sessile drop method on the FM40 Easy Drop instrument (KRÜSS) at room temperature. A 2-μL water droplet was carefully placed on the membrane using a microsyringe. The reported contact angle values are the averages of three measurements.Membrane performanceThe organic solvent nanofiltration experiments were carried out using a 4-way dead-end stainless-steel cell (Textop Ltd., Hungary) at room temperature and a constant stirring speed of 150 rpm. The feed chamber was pressurized using nitrogen gas. Two membrane samples with an effective area of 4.9 cm2 were loaded into the cell and tested in parallel using a feed volume of 400 mL. The membranes were initially compacted at 25 bar or 35 bar for one hour. The transmembrane pressure was afterward reduced to 20 or 30 bar, and the steady-state flux of pure solvents was determined by measuring the volume (V) per unit area (A) per unit time (t) according to Eq. (1).$${\rm{J}}=\frac{\Delta V}{A\cdot \Delta t}\left[\frac{{\rm{L}}}{{{\rm{m}}}^{2}{\rm{h}}}\right]$$
(1)
The permeance was calculated according to Eq. (2).$${\rm{P}}=\frac{\Delta V}{A\cdot \Delta t\cdot p}\left[\frac{{\rm{L}}}{{{\rm{m}}}^{2}\cdot {\rm{h}}\cdot {\rm{bar}}}\right]$$
(2)
where p is the transmembrane pressure.The permeance of different solvents was measured in the following sequence: methanol, ethanol, acetonitrile, acetone, tetrahydrofuran, heptane, and toluene. The properties of the solvents are summarized in Supplementary Table 1. The solute rejection was evaluated using a mixture of solutes with the properties shown in Supplementary Table 2. The mixture was prepared in acetonitrile at a fixed concentration of 0.25 mM of each solute. A 400 mL of the mixture was used as feed; the first 10 mL was discarded, and three permeate samples of 1.5 mL were collected. The samples were analyzed using a high-performance liquid chromatography (HPLC) system (Ultimate 3000, Thermo Scientific) equipped with a UV detector (Diode Array Detector, Thermo Scientific). The column was a Hypersil GOLD, 100 × 2.1 mm, 1.9 µm (Thermo Scientific), using 0.1 wt.% ammonium acetate in LCMS-grade water (A) and HPLC-grade acetonitrile (B) as the mobile phase. The total run time for each injection was 42 min with a 3-µL injection volume. The selected wavelength included the highest absorption wavelength for each solute, namely 220, 254, 272, 320, 215, 232, and 262 nm. The column oven temperature was maintained at 45 °C, the vaporizer temperature of the ISQEM was set to 200 °C, and the ion transfer tube temperature was 300 °C. A hybrid isocratic gradient elution with a flowrate of 0.5 mL min–1 was used for the measurements. The data was manually integrated into Chromeleon 7. At least two samples of each membrane type were tested to ensure the reproducibility of the results.The concentrations of the feed (CF), permeate (CP), and retentate (CR) were estimated based on the area under the curve (AUC) for each chemical and used to determine the rejection (R) values as follows:$${\rm{R}}=1-\frac{{{\rm{C}}}_{{\rm{p}}}}{{{\rm{C}}}_{{\rm{R}}}}=1-\frac{{{\rm{AUC}}}_{{\rm{p}}}}{{{\rm{AUC}}}_{{\rm{R}}}}$$
(3)
The separation factor for compound A over compound B was calculated using the following equation:$${\beta }_{{\rm{A}}/{\rm{B}}}=\frac{{{\rm{C}}}_{{\rm{P}}}^{{\rm{A}}}}{{{\rm{C}}}_{{\rm{F}}}^{{\rm{A}}}}\cdot \frac{{{\rm{C}}}_{{\rm{F}}}^{{\rm{B}}}}{{{\rm{C}}}_{{\rm{P}}}^{{\rm{B}}}}=\frac{1-{R}_{{\rm{A}}}}{1-{R}_{{\rm{B}}}}$$
(4)
To determine the separation factors with respect to all the solutes, a matrix was formed such that the rows and columns represent the thirteen markers arranged in order of their molecular weight. The separation factors were normalized by the ratio of the molecular weights of the solute pairs to account for the size difference as shown in Eq. (5). This normalized parameter was proposed by Marchetti et al. 5. and referred to as the selectivity figure of merit (\({\rm{SFM}}\left({\beta }_{{\rm{A}}/{\rm{B}}}\right)\)).$${\rm{SFM}}\left({\beta }_{{\rm{A}}/{\rm{B}}}\right)=\frac{{\beta }_{{\rm{A}}/{\rm{B}}}}{{{\rm{MW}}}_{{\rm{B}}}/{{\rm{MW}}}_{{\rm{A}}}},\, {{\rm{MW}}}_{{\rm{B}}}/{{\rm{MW}}}_{{\rm{A}}}\ge 1$$
(5)
The average selectivity figure of merit (\(\overline{{SFM}}\)) was calculated according to Eq. (6). Any solute that reported a rejection value of 1.0 was excluded from the arithmetic average.$$\overline{{\rm{SFM}}}=\frac{1}{n}\mathop{\sum }\limits_{i=0}^{n}{{\rm{SFM}}}_{{\rm{A}}/{\rm{B}}}\left(\beta \right)$$
(6)
Solubility parametersThe solubility parameters are used to describe the affinity between the polymers, the solutes, and the solvent. The Hansen solubility parameters were calculated using the HSiP software and Eq. (7), where \({{\rm{\delta }}}_{{\rm{D}}}\), \({{\rm{\delta }}}_{{\rm{P}}}\), and \({{\rm{\delta }}}_{{\rm{H}}}\) represent the dispersion forces, polar interactions, and hydrogen bonding, respectively49. For the polymers, the parameters were calculated for the repeating unit, assuming all reactive groups were consumed during the reaction. Supplementary Table 3 summarizes the Hansen solubility parameters of acetonitrile, solutes49, and polymers47,49 used in membrane preparation.$${\rm{\delta }}=\sqrt{{\left({{\rm{\delta }}}_{{\rm{D}}}\right)}^{2}+{\left({{\rm{\delta }}}_{{\rm{P}}}\right)}^{2}+{\left({{\rm{\delta }}}_{{\rm{H}}}\right)}^{2}}$$
(7)
The distance between compounds A and B in the Hansen space (Ra) was calculated using Eq. (8). If two compounds show good affinity or solubility, the Ra value is lower than the radius of interaction (Ro). The relative energy difference can be predicted using the ratio of Ra/Ro, the RED number.$${\rm{R}}{{\rm{a}}}^{2}=4{\left({{\rm{\delta }}}_{{\rm{D}}.{\rm{A}}}-{{\rm{\delta }}}_{{\rm{D}}.{\rm{B}}}\right)}^{2}+{\left({{\rm{\delta }}}_{{\rm{P}}.{\rm{A}}}-{{\rm{\delta }}}_{{\rm{P}}.{\rm{B}}}\right)}^{2}+{\left({{\rm{\delta }}}_{{\rm{H}}.{\rm{A}}}-{{\rm{\delta }}}_{{\rm{H}}.{\rm{B}}}\right)}^{2}$$
(8)
Partitioning coefficient calculationsThe partitioning coefficients (LogP) were taken from the PubChem database of the National Library of Medicine50. If measured LogP values were not available, the XLogP3 values from the same source were used instead.Pore size distribution calculationsThe pore size distribution calculations were performed according to the literature51. First, the critical volumes of the solutes were calculated using the Joback group contribution method52.$${V}_{c}=17.5+\sum \Delta {V}_{c}^{i}$$
(9)
Where \(\sum \Delta {V}_{c}^{i}\) is the sum of group contributions for critical volume. From the critical volume, we can calculate the molar volume at the boiling point using the following empirical equation:$${V}_{{\rm{m}},{\rm{i}}}=0.285\times {V}_{{\rm{c}}}^{1.048}$$
(10)
Using the boiling point molar volume of the solute, the diffusivity coefficient can be calculated from the Wilke-Chang equation:$${{\rm{D}}}_{{\rm{s}},{\rm{j}}}=7.4\times 10-4\frac{T\sqrt{\Phi M\,}}{{\mu }_{p,i}{V}_{m,i}^{0.6}}$$
(11)
Where \(T\) is the temperature, \(\Phi\) is the affinity coefficient of the solvent, \(M\) is the molar weight of the solvent. \({\mu }_{p,i}\) is the kinematic viscosity, and \(V\) is the molar volume at the boiling point. From the diffusivity coefficient (\({{\rm{D}}}_{{\rm{s}},{\rm{j}}}\)), we can calculate the equivalent sphere diameter that will have the same diffusivity according to the Stokes-Einstein equation:$${{\rm{D}}}_{s,{ij}}=\frac{{KT}}{6\pi {r}_{s,j}{\mu }_{p,i}}$$
(12)
Where \(K\) is the Boltzmann constant, \(r\) is the radius, and \({\mu }_{p,i}\) is the matrix (solvent) dynamic viscosity. The resulting radius values can be plotted against the rejection. Then, a generalized logistic function was fitted on the data points, adjusting five parameters according to the equation:$$g(x)=A+\frac{K-A}{{\left(C+Q\cdot {e}^{-{Bx}}\right)}^{1/\nu }}$$
(13)
The distribution function can be calculated by calculating the derivative of the \(g(x)\):$${g}^{{\prime} }\left(x\right)=\frac{\left(K-A\right)\cdot \left(-B\right)\cdot Q\cdot {e}^{-{Bx}}}{\nu \cdot {\left(C+Q\cdot {e}^{-{Bx}}\right)}^{\frac{1}{\nu }+1}}$$
(14)
Where A, K, B, Q and \(\nu\) are internal fitting parameters. The normalized mean expected value of \({g}^{{\prime} }\left(x\right)\) is calculated using the numerical integration method on \({g}^{{\prime} }\left(x\right)\) after normalization using the scipy’s quad function.