3D printing of inherently nanoporous polymers via polymerization-induced phase separation

3D printing offers enormous flexibility in fabrication of polymer objects with complex geometries. However, it is not suitable for fabricating large polymer structures with geometrical features at the sub-micrometer scale. Porous structure at the sub-micrometer scale can render macroscopic objects with unique properties, including similarities with biological interfaces, permeability and extremely large surface area, imperative inter alia for adsorption, separation, sensing or biomedical applications. Here, we introduce a method combining advantages of 3D printing via digital light processing and polymerization-induced phase separation, which enables formation of 3D polymer structures of digitally defined macroscopic geometry with controllable inherent porosity at the sub-micrometer scale. We demonstrate the possibility to create 3D polymer structures of highly complex geometries and spatially controlled pore sizes from 10 nm to 1000 µm. Produced hierarchical polymers combining nanoporosity with micrometer-sized pores demonstrate improved adsorption performance due to better pore accessibility and favored cell adhesion and growth for 3D cell culture due to surface porosity. This method extends the scope of applications of 3D printing to hierarchical inherently porous 3D objects combining structural features ranging from 10 nm up to cm, making them available for a wide variety of applications.


Supplementary Note 1. Ink compositions and printing settings used in DLP 3D printing
The compositions of the inks used in this study are listed as follows: Printing settings: In DLP 3D printing, the layer thickness and the layer cure time are two important parameters that directly influence the printing speed and quality. In principle, the choice of layer thickness is only limited by the translation precision of the platform. However, there is always a trade-off between printing speed and resolution when adjusting the layer thickness. The layer cure time, on the other hand, needs to be judiciously selected: too short cure time will lead to layer delamination or missing parts, while too long cure time will result in overgrowth and lamination issues between the specimen and vat lining.
Based on these considerations, the layer thickness for all 3D printing experiments in this study was selected as 50 μm to ensure good printing resolution with acceptable printing speed. For determining the cure time, the working curve (cure depth as a function of cure time) for each ink was first measured (Supplementary Fig. 6b and Supplementary Fig. 10). The working curve provides a reference to determine the cure time. Then the cure time was empirically determined to provide objects with visibly good printability, without overgrowth, missing parts, or lamination issues with vat lining. The cure time was set as 12.5 s for ink Mix-1 to Mix-6, and 25 s for ink Mix-7 and Mix-8 due to the lower curing rate of DMAEA.

Supplementary Note 2. UV-Vis spectrum of the inks prepared with different photoinitiators
Supplementary Figure 1. UV-Vis spectra of the inks (30 wt% HEMA, 20 wt% EDMA, 40 wt% cyclohexanol and 10 wt% 1-decanol) without photoinitiators and with different photoinitiators (4 wt% to monomers), compared with the emission spectrum of the DLP printer. To realize the 3D printing of inherently nanoporous polymers, the ink should undergo phase separation upon photopolymerization. Therefore, we screened inks with a range of monomer and porogen ratios and identified the compositions that could undergo phase separation upon photopolymerization. Specifically, we irradiated the ink with the UV light source of the 3D printer for 300 s to prepare a 75 μm thick film and measured its transmittance at 500 nm ( Supplementary   Fig. 2a). At a fixed porogen composition, when the porogen content in the inks exceeded a certain value, the light transmittance of the dry film experienced a dramatic decrease ( Supplementary   Fig. 2b), indicating phase separation in the photopolymerization process. This method was used to construct the ternary diagram shown in Fig. 1c.
Another criterion in selecting the ink composition is the mechanical strength of the 3D printed structures. We found that the compressive strength of the 3D printed structures decreased with increasing the porogen content in the ink due to a decrease in their density ( Supplementary Fig. 3) When the porogen content was higher than 80 wt%, the mechanical strength of the 3D printed structures was too weak leading to a spontaneous degradation of the structures during 3D printing. For investigation of the printing resolution, we used a DLP printer with a higher resolution (Miicraft prime 110, pixel size 40 μm, https://miicraft.com/). An array of pillars with different diameters was designed and printed ( Supplementary Fig. 4a). It can be seen from the SEM images that pillars with diameter larger than 100 μm can be printed, and the surface of the pillars remained highly porous ( Supplementary Fig. 4b). The variance between the designed and printed feature was calculated by measuring the diameter of the printed pillars shown in the SEM images. All printed pillars showed -8~13% variance compared to the designed value ( Supplementary Fig. 4c). This is attributed to the shrinkage (~13%) of the printed porous polymers during supercritical drying, calculated from the length of a 3D printed cube (5×5×5 mm 3 ) before and after drying using optical microscopy. We note that there is a slight difference between the geometry of the printed pillars and the CAD files: the printed pillars are slightly wider at the bottom. This might be attributed to the higher crosslinking density (potentially due to the reflection of the substrate or the effect of the surface on the composition of the porogens) at the bottom and thus a lower shrinkage in the drying process.

Supplementary Note 7. Theoretical simulation of polymerization-induced phase separation using the phase-field method
A phase-field model within the framework of the Cahn-Hilliard approach is adopted to simulate the polymerization-induced phase separation. In our model, we use 1 , 2 , 3 to depict the concentrations of monomer, polymer, and solvent, respectively. The time evolution of the respective concentration follows 1 : ). (1) In Supplementary Equation 1, is the mobility representing the kinetic parameter of the system and assigned in accordance with the Onsager relationship as ].
Here, 0 stands for the inter-diffusivity of the polymer solution and is formulated as  Table 1). As 1-decanol is a poorer solvent than cyclohexanol, increasing 1-decanol concentration leads to a poorer solvent system, giving rise to the larger repulsive interactions between solvent and polymer. The last parameter 123 is assigned as -3.0 to simulate the polymerization-induced phase separation.  b, Compressive strain-curve for 3D polymer cubes (5×5×5 mm 3 ) printed using the five inks.

Representative of three independent experiments (N=3).
Supplementary Note 9. Pore size distribution of the 3D printed object with tri-disperse pore size Supplementary Figure 8. Pore size distribution of the 3D printed object with tri-disperse porosity by switching the inks during printing (Fig. 3f). a, Cross-sectional SEM images; b, Local thickness mapping; c, Pore size distribution.
The cross-sectional SEM images were taken at the upper-part, middle-part and lower-part of the 3D printed object (Fig. 3f), which were printed by using ink Mix-1, Mix-3 and Mix-5, respectively ( Supplementary Fig. 8a). The pore sizes were measured from the cross-sectional SEM images using the 'Local Thickness' plug-in for ImageJ. The SEM images were first converted to binary images using greyscale thresholding. Then the binary images were analyzed using the 'Local Thickness' plug-in, which measures the diameter of the largest sphere that fits into the dark region (pores), giving back a colored map of local thickness (Supplementary Fig. 8b). Here we assume that the pores preserve a spherical shape and the measured local thickness represents the pore diameter. The histograms for pore size distribution were calculated from the pixel counts for a given thickness ( Supplementary Fig. 8c). We note that the extraction of pore size from 2D cross-sectional instead of 3D z-stack images may lead to a systematic error, which we do not correct as here we care more about the general trends rather than very exact numbers for the pore diameter.    Fig. 19). The value of the slope a and the y-intercept b of the line were used to calculate the monolayer adsorbed gas quantity vm:

Supplementary
The specific surface are SBET can be then calculated from the following equation: Where N is the Avogadro number (6.02 × 10 23 ), s is the effective cross-sectional area of N2 ( 1.62 × 10 −19 2 ), and V is the molar volume of N2 at standard temperature and pressure (0.224 3 /mol).