Metal-coordinated sub-10 nm membranes for water purification

Ultrathin membranes with potentially high permeability are urgently demanded in water purification. However, their facile, controllable fabrication remains a grand challenge. Herein, we demonstrate a metal-coordinated approach towards defect-free and robust membranes with sub-10 nm thickness. Phytic acid, a natural strong electron donor, is assembled with metal ion-based electron acceptors to fabricate metal-organophosphate membranes (MOPMs) in aqueous solution. Metal ions with higher binding energy or ionization potential such as Fe3+ and Zr4+ can generate defect-free structure while MOPM-Fe3+ with superhydrophilicity is preferred. The membrane thickness is minimized to 8 nm by varying the ligand concentration and the pore structure of MOPM-Fe3+ is regulated by varying the Fe3+ content. The membrane with optimized MOPM-Fe3+ composition exhibits prominent water permeance (109.8 L m−2 h−1 bar−1) with dye rejections above 95% and superior stability. This strong-coordination assembly may enlighten the development of ultrathin high-performance membranes.

SEM. Field emission scanning electron microscopy (Nanosem 430, Japan) was utilized to capture the surface and cross-section morphology of membrane. The Au coating was sputtered on samples surface (Q150T turbo-pumped sputter coater) with current of 25 mA for 1 min at argon atmosphere (2×10 -2 mbar) to achieve a minimum conductivity and avoid sample charging.
TEM. Transmission electron microscopy (JEM-2100F, Japan) was employed to capture the cross-section image of membrane. The membrane sample was encapsulated by epoxy resin and cut into thin (~90-100 nm) slices by Ultratome (Leica EM UC6). Finally, the slice-shaped samples were deposited on micro grid copper.
AFM. Atomic force microscopy (Bruker Dimension Icon, USA) was employed to characterize the surface roughness and thickness of membranes. The images were captured using tapping mode. NanoscopeAnalysis data visualization and analysis software was used to process the AFM images. Surface roughness was presented by root-mean-square (R rms ). Surface morphology, roughness parameters and thickness were evaluated from AFM scans.
To measure the thickness, the MOPMs was assembled on quartz glass wafers. A scratch was made with a sharp AFM probe 1 with approximately 2 nN applied force. Membrane thickness was detected from the height difference between glass substrate and the membrane using a one dimensional statistical function. The surface strength of the membranes was expressed by Young's modulus (E), and calculated using (Eq. 1) where E was the Young's modulus, d and z were the cantilever deflection and piezo displacement, respectively. k was the spring constant of the cantilever (k = 39 N/m), ν was Poisson's ratio (ν = 0.5), and α is the opening angle of the cone (α = 35°). AFM results were analyzed and exported by bundled software (NanoScopeAnalysis Version1.9, Bruker Dimension Icon).

EDX.
Energy dispersive X-ray spectroscopy (Genesis XM2 APEX 60SEM, USA) was utilized to achieve the elemental distribution and composition of membranes. The EDS detector was externally affiliated to the SEM apparatus and the EDS mapping images and elemental ratio could be obtained under similar condition of SEM characterization.
ATR-FTIR spectra. Attenuated total reflectance fourier transform infrared spectroscopy (Nicolet 560, USA) was utilized to characterize the chemical structures of MOPMs.

CA.
where C p (ppm) and C f (ppm) were the solute concentration in permeate and feed solutions, respectively.
where r (m) was the Stokes radii of PEG and Mw was the average molecular weight (Da). Subsequently, we related obtained solute rejection with the Stokes radii and transformed it into a correlation function. Finally, the pore size distribution was described by the following probability density function namely Eq. 4, where μ p was defined as the geometric mean diameter of solute at 50% solute rejection, σ p was defined as the ratio of the solute radius when solute rejections were 84.13% and 50%, representing the geometric standard deviation of μ p .

Filtration performance measurement
The membrane was loaded in the filtration cell with effective area of 4.1 cm 2 and maximum volume of 10 mL.
Before measurement, the membrane was prepressed with DI water at 1.5 bar for 30 min to obtain steady flux. Then, the filtration performance of membrane was measured at 1.0 bar with 10 ppm, 50 ppm and 100 ppm of dye solution as feed (Methyl blue, Congo red, Alcian blue, Rose Bengal and Orange GII). Especially, the graphene oxide quantum dots (GQDs) were synthesized based on our previous reported work 3 and used as filtration feed (1wt%) to simulated nanoadsorbents. Besides, the salt permeation was also measured by rejecting 1000 ppm of salt solution feed (NaCl, NaSO 4 , MgSO 4 , MgCl 2 ), which was calculated by the conductivity of feed and permeate with a conductivity analyzer (Leichi, DDS-11A, China) using following equation (Eq. 5), where the C p and C f were the salt concentration in permeate and feed solutions. The dye adsorption was measured by fixing membrane into filtration cell (effective area=1.77 cm 2 ) with 10 mL of feed solution (100 ppm dye). After 3-day adsorption, the dye concentration of feed was determined by ultraviolet-visible spectrophotometer and the amount of adsorbed dyes (M dye , μg/cm 2 ) on the membrane was calculated by following equation (Eq. 6) where the C 0 , C 3 (μg/mL) were the dye concentration at the beginning and after 3 days, respectively, V S (mL) was the volume of solution, the A (cm 2 ) was the effective membrane area. The reported data were the mean values of triplicate samples for each membrane.

Stability performance measurement
For antifouling performance, the MOPM-Fe 3+ /PAN composite membrane was pre-compacted at 1. Firstly, the membrane was tested with DI water at 1.0 bar for 30 min to get initial permeance (J n−1 ). Secondly, filtration process was carried out for 30 min with feed solution containing 1000 ppm foulant to get filtration permeance (J (n−1)f ). Finally, after rinsing membrane with DI water by magnetic stirring at 200 rpm for 5 min, recovered permeance (J n ) of membrane was measured at 1.0 bar for 30 min. The as-described filtration-rinse-test process was repeated five times and the permeance was normalized for comparison. The n (n=1,2,3,4,5) represented filtration cycle. Permeance recovery ratio (P RR = J n /J n−1 ) and permeance decline ratio (P DR =1− J (n−1)f /J n−1 ) were calculated to evaluate the antifouling performance of membrane for each cycle.
The 5-day stability performance of membranes was carried out by filling DI water into buffer tank to supply enough feed. Owing to the nearly 100% Congo red rejection of MOPM-Fe 3+ /PAN composite membrane, it was reasonable to assume there was no dye loss in feed and maintain a constant feed concentration (100 ppm) during long-term performance measurement.
The BSA adsorption measurement was conducted with method we previous reported 4 where the C 0 , C 1 (μg/mL) were the BSA concentration at the beginning and adsorption equilibrium (1 day), respectively, V S (mL) was the volume of solution, the A (cm 2 ) was the membrane area. The reported data were the mean values of triplicate samples for each membrane. The BSA concentration was determined by UV-vis spectrophotometer at 278 nm.
The stability of MOPM-Fe 3+ /PAN membrane in salty condition was conducted by immersing membrane in Na 2 SO 4 solution (1000 ppm) or NaCl solution (1000 ppm) followed by performance measurement every two days.
This immersion test was carried out for 2 weeks.
where k is the solvent permeation speed (L m −2 h −1 ), A is the pre-exponential factor (L m −2 h −1 ), E a is the activation energy associated with the permeation process (kJ mol −1 ), R is the gas constant (kJ mol −1 K −1 ), and T is absolute temperature (K). By taking the log of both sides of Eq. (8) and using R (8.314 × 10 −3 kJ mol −1 K −1 ), the E a can be evaluated.

Supplementary Figures
Supplementary Figure 1

Supplementary Tables
Supplementary