A solution-processable and ultra-permeable conjugated microporous thermoset for selective hydrogen separation

The synthesis of a polymer that combines the processability of plastics with the extreme rigidity of cross-linked organic networks is highly attractive for molecular sieving applications. However, cross-linked networks are typically insoluble or infusible, preventing them from being processed as plastics. Here, we report a solution-processable conjugated microporous thermoset with permanent pores of ~0.4 nm, prepared by a simple heating process. When employed as a two-dimensional molecular sieving membrane for hydrogen separation, the membrane exhibits ultrahigh permeability with good selectivity for H2 over CO2, O2, N2, CH4, C3H6 and C3H8. The combined processability, structural rigidity and easy feasibility make this polymeric membrane promising for large-scale hydrogen separations of commercial and environmental relevance.


Preparation of 3D, 2D and 1D conjugated microporous thermoset (CMT)
The thermosetting polymerization was conducted in a tube furnace with an Ar gas flow of 100 ml min -1 . In the tube furnace, precursors with various substrates were put in a jar equipped with a lid to minimize the disturbance of gas flow. Normally, samples were heated to 540 º C at a heating rate of 15 ºC min -1 and held at 540 ºC for 120 minutes, followed by cooling down to room temperature at ~10 ºC min -1 .

Synthesis of patterned 3D CMT
To produce the bulk CMT with a patterned surface, 50 mg of 3-TBTBP powder was loaded in a small glass holder with patterned silicon bottom plate. The precursor melted and polymerized in the holder during the heating. After cooling to room temperature, the patterned CMT was peeled off and used directly for testing.
Synthesis of 3D CMT/AAO membrane: the precursor 3-TBTBP was coated on an AAO disc by filtering 3-TBTBP/IPA (4 mg in 50 ml) dispersion through a porous AAO membrane (Anodisc, 47 mm in diameter, 0.02 μm pore size, Whatman). The 3-TBTBP coated AAO disc was covered with a flat Al foil to ensure no precursor evaporated out during the heating. After cooling down to room temperature, the Al foil was easily peeled off and the CMT/AAO film was directly used for tests.

Preparation of 2D CMT on Si Wafer
Prior to the growth of CMT films, O 2 plasma treatment was conducted to remove organic contamination on the Si/SiO 2 wafer. 10 mg of 3-TBTBP was used as precursor and Si wafer was placed around the precursor source with a certain distance inside the jar. A 10 cm distance results in a CMT film on the wafer with thickness ~ 5 nm.

Preparation of 1D CMT nanotubes
The substrates Cu nanowires were prepared as follows: into 19 ml of CuCl 2 2H 2 O aqueous solution (0.01 M) was added 0.12 ml oleylamine. After the mixture was sonicated for 1 min to form a white-blue emulsion, 1 ml L-ascorbic acid (0.2 M) was added and the solution was kept at 60 º C overnight. The precipitate was collected by filtration and washed 3 times with isopropanol (IPA). 10 mg of 3-TBTBP was used as the precursor source and 5 mg of Cu nanowire was put near the source (~2 cm in distance) inside the jar. Then the sample was heated to produce CMT coated Cu nanowires. Cu nanowires were etched away by HCl solution (37 wt% in water), then the CMT nanotubes were collected and washed with deionized water, followed by drying at ambient conditions.

Preparation of CMT ultrathin sheets
3-TBTBP and NaCl was mixed evenly and then heated in a jar. When the weight ratio of 3-TBTBP:NaCl = 1:6000, the resulting sheets have a thickness ~ 5.0 nm. After cooling down to room temperature, the CMT coated NaCl was immersed in a large amount of DI water for 3 hours to remove NaCl. The floating CMT sheets were filtered and washed with DI water. Finally, the obtained CMT sheets were freeze-dried.

2.1.Methods and instruments
Atomic Force Microscope (AFM) measurements were operated with a Bruker Dimension Fast Scan Atomic Force Microscope in tapping mode at ambient conditions. Scanning electronic microscopy (SEM) images were obtained using a Jeol JSM-6701F instrument. For X-ray photoelectron spectroscopy (XPS) investigations, samples were dispersed on Au-coated Si wafers. The XPS spectra were collected with the Phobios 100 electron analyzer equipped with 5 channeltrons, using an unmonochromated Al Ka X-ray source (1486.6 eV). Fourier-transform infrared spectroscopy (FTIR) measurements were performed at room temperature in a continuous vacuum environment. The FTIR samples were prepared by compressing samples with KBr into a disc. Solid 13 C NMR spectra were recorded with Bruker 400 MHz NMR. UV-VIS-NIR adsorption spectra were collected with Shimadzu UV-3600.
The samples were dispersed in DMF. Bulk CMT was ground into finepowders for Ar sorption measurements, which were performed on a Micromeritics ASAP 2020 with micropore option using liquid Ar bath. Thermal gravimetric analysis (TGA) was performed in the range of 150 to 900 ºC at a heating rate of 10 ºC /min in nitrogen atmosphere. Differential scanning calorimetry (DSC) was measured in the range of 100 to 600 ºC at a heating rate of 15 ºC min -1 in nitrogen atmosphere.
The free volume size, fraction and distribution of CMT polymer powders as a function of temperature were characterized by positron annihilation lifetime spectroscopy (PALS) using a conventional fast-fast coincidence spectroscope. Two Kapton® films were used to sandwich a radioactive source of 22 Na as the source of positrons. The samples were vacuumed for at least 12 h prior to testing and five million counts were collected for each spectrum. The data were resolved into three lifetimes using the PATFIT program, which assumes a Gaussian distribution of the logarithm of the lifetime for each component. The free volume distribution was analysed based on the MELT program.
Density of CMT was determined using a balance (Mettler Toledo, XS205) and a density determination kit according to the Archimedes' principle. The CMT density ( c ) was calculated based on the following equation: where W a is the CMT weight in air, W l is the CMT weight in the auxiliary liquid and ρ l is the density of the auxiliary liquid (n-hexane).

Crystallography data collection
The X-ray data were collected with a Bruker AXS D8 Venture Kappa four cycles X-ray diffractometer system equipped with a Photon 100 detector, using a Mo sealed microfocusing source, with the Bruker Apex 2 suite program. Data were integrated with the Bruker SAINT program using a narrow-frame algorithm. SADABS was used for absorption correction. Structural solution and refinement were carried out with the SHELXTL suite of programs. The structures were solved by direct methods, followed by difference maps and refined with full-matrix least-squares on F2. All non-hydrogen atoms were generally given anisotropic displacement parameters in the final model. All hydrogen atoms were placed at calculated positions. large face-to-face π-π interactions with an intermolecular distance of 3.434 Å. The stacked structure is stabilized by supramolecular multiple interactions such as H-bonds, face-to-face π-π interactions, and van der Waals forces.

Elemental analyses of 3-TBTBP and CMT
Supplementary  Supplementary Figure 6｜SEM image showing the inner structure of bulk CMT with a dense morphology.

AFM image of a CMT film grown on Si wafer
Supplementary Figure 7｜A CMT film grown on Si wafer with a thickness ~20 nm.
a, optical microscope image. b, AFM topography image and corresponding height profile.

Raman spectrum of CMT film on Si wafer
Supplementary Figure 8｜Raman spectrum of CMT film grown on Si wafer.

SEM image of CMT coated Cu nanowires
Supplementary Figure 9｜SEM images of CMT coated Cu nanowires.

Gas permeation measurements
Gas permeation properties of pure gases were conducted on a constant-volume variable-pressure gas permeation cell. The membrane was mounted onto the permeation cell and vacuumed for at least 12 h before tests. Pure gases including He, H 2 , O 2 , N 2 , CH 4 , C 3 H 6 , C 3 H 8 and CO 2 were tested. The measurement condition was held at 30 °C with a transmembrane pressure of 1 bar. The gas permeability through the membrane was calculated according to the steady state pressure increment (dp/dt) as given by the following equation: where P denotes the gas permeability in barrer (1 barrer = 1×10 −10 cm 3 (STP)· cm· cm -2 s -1 · cmHg -1 ), V is the volume of the downstream reservoir (cm 3 ), A is the effective membrane area (cm 2 ), l represents the membrane thickness (cm), T is the testing temperature (K) and P 2 is the upstream pressure of the system.
To measure the mixed gas separation performance, an equimolar mixture of H 2 and CO 2 was applied as the feed from 30°C to 150 °C at a transmembrane pressure of 1 bar.
The permeabilities of H 2 and CO 2 were obtained from equations (2) and (3): The molar fractions of gases in the feed and the permeate sides are denoted as x and y, respectively.  H 2 and  CO 2 refer to the fugacity coefficients of the respective upstream gases, which were determined by Thermosolver software according to the Peng-Robinson equation of state 1 .
The ideal or mixed gas selectivity (α) between two different gases across a membrane is the ratio of their single gas permeability as described in the following Equation: α ⁄ (4) where P A and P B refer to the permeabilities of gases A and B, respectively.  Adsorbed CO 2 decreases as a function of increasing temperature. It shows that the concentration of adsorbed CO 2 inside the membranes decreases as a function of increasing temperature owing to the larger thermal energy of CO 2 at higher temperatures. Due to the decrease of CO 2 adsorbed concentration, the solubility coefficient decreased (S=C/P, C refers to the concentration of adsorbed CO 2 , P refers to the applied pressure), thereby reducing the permeability of CO 2 . Therefore, the H 2 permeability is increased together with the increase of H 2 /CO 2 selectivity. 2

Comparison of gas separation performance of CMT with various reported membranes
Supplementary  In order to validate the reasonability of the built CMT model, the pore size distributions (Supplementary Figure 19) of CMT models were obtained by sampling the test particle radii proposed by Bhattacharya et al 29 . The difference between the simulated and experimental pore size distribution could be considered as favorable from the simulation method. 30 Furthermore, both mass density and accessible surface area were tested to compare with experimental data.

MD simulation of gas separation process
For H 2 and CO 2 , three-site models were adopted 31 . The 12-6 Lennard-Jones potential was used to calculate the van der Waals interactions between gas and CMT membrane.
Partial atomic charges and L-J parameters of H 2 (CO 2 ) and CMT are listed in Supplementary Table 4. Lorentz-Berthelot mixing rules were used to obtain the missing heteronuclear parameters. The cutoff distance for all the short-range vdW interactions was set as 12 Å. The long-range electrostatic interactions were computed by using the particle-particle particle-mesh (PPPM) algorithm. The simulation was conducted in a canonical ensemble (NVT) at 423 K controlled by the Nose-Hoover thermostat method 32 . The time step was set as 1 fs and the data were collected every 1 ps. The total simulation time of each model was 20 ns. The simulation shows that he fluxes are 88.9 and 22.4 molecules/ns for H 2 and CO 2 , respectively, giving a H 2 /CO 2 selectivity ~4 for the 5 nm-thick monolayer CMT sheet.