Crystal nuclei templated nanostructured membranes prepared by solvent crystallization and polymer migration

Currently, production of porous polymeric membranes for filtration is predominated by the phase-separation process. However, this method has reached its technological limit, and there have been no significant breakthrough over the last decade. Here we show, using polyvinylidene fluoride as a sample polymer, a new concept of membrane manufacturing by combining oriented green solvent crystallization and polymer migration is able to obtain high performance membranes with pure water permeation flux substantially higher than those with similar pore size prepared by conventional phase-separation processes. The new manufacturing procedure is governed by fewer operating parameters and is, thus, easier to control with reproducible results. Apart from the high water permeation flux, the prepared membranes also show excellent stable flux after fouling and superior mechanical properties of high pressure load and better abrasion resistance. These findings demonstrate the promise of a new concept for green manufacturing nanostructured polymeric membranes with high performances.


Supplementary
. Typical pore size distribution of PVDF membranes measured by the gas-liquid displacement method. (a) CCD Glass/Glass 1 mm sample, which shows a pore size range from 286 to 389 nm and a mean flow pore size of 321 nm; (b) CCD Glass/Al 1 mm sample, which shows a pore size range from 103 to 240 nm and a mean flow pore size of 106 nm; (c) CCD Al/Al 1 mm sample, which shows a pore size range from 33 to 90 nm and a mean flow pore size of 41 nm; (d) NIPS DMSO 1 mm sample, which shows a pore size range from 19 to 80 nm and a mean flow pore size of 37 nm; (e) NIPS DMSO 0.3 mm sample, which shows a pore size range from 20 to 79 nm and a mean flow pore size of 57 nm; (f) NIPS NMP 0.3 mm sample, which shows a pore size range from 19 to 70 nm and a mean flow pore size of 41 nm.     * calculated based on a flat-sheet membrane of dimensions 1 × 2 m 2 (width × length), water flows along the length direction.

Supplementary Table 3. Comparison between the NIPS and CCD methods for flat-sheet PVDF membrane production NIPS CCD Preparation Method
The casting film of polymer solution together with the support is immersed in a coagulation bath composed of a non-solvent for the polymer. The polymer solution is then transformed from a liquid to a solid state due to the exchange of the solvent in the polymer solution with the non-solvent from the coagulation bath.
The casting film of polymer solution is unidirectionally cooled from one side to a certain temperature far below the freezing point of the solvent. As a result, the solvent starts nucleation and crystallization, and the polymer precipitates to form the final membrane structure. Then the solvent is leached out by iced water.

Typical Structure
Asymmetric structure with a dense skin layer supported by finger-like voids and sponge-like layer A thin separation layer of numerous tortuous pores supported by gradually changed, fully opened, interconnected and self-organized microchannels  The NMP Al/Al membrane shows a thick dense separation layer and a sponge-like structure in the supporting layer, and no micro-channels were formed. This thick and dense separation layer showed a very low pure water flux of 6.5 LMH bar -1 , and no pores larger than 18.6 nm were detected with gas-liquid displacement porosimetry.

Possible influencing factors during operation with a fixed composition of polymer solution
The DMAc Al/Al membrane also shows a homogeneous but porous supporting layer and a denser top layer. But this membrane broke apart when the crystallisation of DMAc is finished and only debris were obtained, which might due to the damaging shape of DMAc crystal grains that cuts the membrane. The SEM images show some deep cracks formed at the back side of the membrane. Permeation characteristics such as pore size and pure water flux were therefore not obtained for this membrane. Supplementary Fig. 3 shows typical pore size distributions of CCD PVDF membranes and NIPS PVDF membranes. All measurements used same pressure steps to ensure fair comparisons. Supplementary Fig. 3a shows a CCD Glass/Glass 1 mm sample, which has a sharp peak at 308 nm with a percent flow of 97.2%; Supplementary Fig. 3b shows a CCD Glass/Al 1 mm sample, which has a sharp peak at 103 nm with a percent flow of 98.3%; and Supplementary Fig. 3c shows a CCD Al/Al 1 mm sample that has a peak at 40 nm with a percent flow of 72.0%. On the other hand, the membranes prepared by the NIPS method showed much broaden pore size distributions. The NIPS DMSO 1mm sample showed a maximum percent flow of 18.9% at 38 nm ( Supplementary Fig. 3d), the NIPS DMSO 0.3 mm sample showed a maximum percent flow of 15.5% at 56 nm ( Supplementary Fig. 3e), and the NIPS NMP 0.3 mm sample showed a maximum percent flow of only 8.4% at 57 nm ( Supplementary Fig. 3f). The NIPS DMAc samples could not be measured, either due to the extremely low porosity, or because the pores are smaller than the testing limit of the equipment (18.6 nm).
In Supplementary Table 2, it is interesting to find that with a faster cooling rate used during the membrane fabrication process, the membrane shows better mechanical properties: higher fracture load, longer elongation and higher tensile stress. It is reasonable to attribute this trend to the microstructural change in the membrane due to the different cooling rates: with a faster cooling rate, the CCD membrane has smaller micro-channels, and the number of the micro-channel would be larger (this assumption agrees with the proposed membrane formation mechanism and is confirmed by SEM images). With smaller but more micro-channels, the stress would be better distributed in the membrane and the energy would be easier to be dissipated by deformation, and fatal damages would be less likely to happen.
The mercury intrusion results of the Al/Al and Glass/Al membranes show typical cumulative intrusion volume-pore size curves (Supplementary Fig. 13a) similar to those rigid pore structures such as in ceramic membranes, with an overall porosity of about 75-76 % and a broaden pore size distribution from around 20 μm to less than 0.1 μm. The gradually increased intrusion volume reflects the gradual change in the pore size from the backside to the top separation layer in the CCD membranes. As expected, the incremental intrusion data ( Supplementary Fig. 13b) of the Al/Al membrane reveals a smaller pore size (11 μm) than the Glass/Al membrane (17 μm) at which intrusion starts, which correspond to the openings of the micro-channels on the backside. The average pore size of the Al/Al membrane is also smaller than the Glass/Al membrane. Closer observation of the incremental intrusion results ( Supplementary Fig. 13c) shows that the Al/Al membrane has higher pore volume at the pore size range of less than 100 nm than the Glass/Al membrane. These results agree very well with SEM images and gas-liquid displacement porosimetry results, and also agree with the prediction of membrane structure based on the cooling rate.
With the experimental method employed for the abrasion tests, it is known that the most severe damages occur at the centre part of the membrane 5 , therefore all SEM images given here were taken from the centre of the membranes for fair comparison For the CCD Al/Al 0.3 mm membrane, it essentially kept the original pore structure in the separation layer and the whole membrane structure after the abrasion test. Although some extent of wearing can be found on the membrane surface, where debris were observed ( Supplementary Fig. 14d), the pore size on the surface and in the separation layer were not affected ( Supplementary Fig. 14c&e). And in the Supplementary Fig. 14c, it can be seen that the thickness of the separation layer basically didn't change compared with the untested same type of membrane shown in Fig. 2c, meaning that the wearing of the membrane under such accelerated test was very slight.
For the NIPS DMSO 0.3 mm membrane, the top separation layer was completely destroyed after the abrasion test. From the SEM images, it can be seen that there were only debris remained at the top layer and the separation layer was gone. And for the NIPS NMP 0.3 mm membrane, the extent of wearing is less than the DMSO sample and the top layer still remains. But the top layer has been largely deformed and the pore structure has been completely altered. Big holes appear on the membrane surface and the surface microstructure has become very rough with apparent worn parts. The NIPS DMAc 0.3 mm membrane was the least damaged sample among the NIPS samples after the abrasion test. The top layer remained almost unchanged after the test, but some debris can be seen on the surface. However, in high magnification SEM images, it is clear that intensive cracks start to appear on the membrane surface after the test, which would change the pore size and ruin the selectivity of the membrane.