Carbon hollow fiber membranes for a molecular sieve with precise-cutoff ultramicropores for superior hydrogen separation

Carbon molecular sieve (CMS) membranes with rigid and uniform pore structures are ideal candidates for high temperature- and pressure-demanded separations, such as hydrogen purification from the steam methane reforming process. Here, we report a facile and scalable method for the fabrication of cellulose-based asymmetric carbon hollow fiber membranes (CHFMs) with ultramicropores of 3–4 Å for superior H2 separation. The membrane fabrication process does not require complex pretreatments to avoid pore collapse before the carbonization of cellulose precursors. A H2/CO2 selectivity of 83.9 at 130 °C (H2/N2 selectivity of >800, H2/CH4 selectivity of >5700) demonstrates that the membrane provides a precise cutoff to discriminate between small gas molecules (H2) and larger gas molecules. In addition, the membrane exhibits superior mixed gas separation performances combined with water vapor- and high pressure-resistant stability. The present approach for the fabrication of high-performance CMS membranes derived from cellulose precursors opens a new avenue for H2-related separations.


Supplementary Note 1. Tuning coagulation temperature
Temperatures of the dope solutions (Td) and coagulation bath (Tc) were investigated to determine the optimal membrane formation conditions. Different flat sheet membranes were cast under various Tc conditions in the range of 25 °C to 60 °C, while Td was maintained at 25 °C. A solvent exchange protocol was used to prevent pore morphology collapse. The water-wetted cellulose membranes were immersed into pure isopropanol for 2 h, followed by soaking in n-hexane for 2 h, and then all the membranes were allowed to dry under ambient conditions in air. It was found that the cellulose membranes had relatively dense morphology when the Tc was below 45 °C °C water bath presents a dense symmetric membrane morphology, as shown in Supplementary   Fig. 2. Thus, the coagulation bath temperature plays a crucial role for generating asymmetric cellulose membrane morphology. The critical temperature to generate asymmetric morphology is found to be ~45 °C, and asymmetric membranes could be obtained only when Tc ≥ 45 °C.
Moreover, in order to obtain a relatively thinner selective layer, a Tc of 60 °C was selected for spinning cellulose hollow fibers in the subsequent work.
Flat-sheet cellulose films with different thicknesses were cast and dried under the same conditions as those used for the spinning process, as shown in Supplementary Fig. 3. When a thin film with a thickness of smaller than 10 μm was made (Supplementary Fig. 3a), the bulk film presented entirely dense morphology, which is representative of the selective layer of the CHFMs. Thus, the thinnest cellulose films were carbonized, and the obtained carbon films were used for structural characterization.

Supplementary Note 2: preparation of asymmetric cellulose hollow fiber membranes
Supplementary Fig. 4 shows cross-sectional SEM images of a cellulose hollow fiber that was dried directly in air without solvent exchange treatment to prevent pore collapse. It presents a dense and symmetric structure. However, Supplementary Figs. 5a and 5b show the cross-sectional SEM images of the spun hollow fibers, and a clear asymmetric structure with a porous inner support layer and dense outer layer is evident. The asymmetric morphological structure was maintained by applying a solvent exchange protocol using first isopropanol, followed by n-hexane). The precursor fibers were dried in air before conducting the carbonization process. The dried cellulose hollow fibers were analyzed by FTIR and TGA, and the results are shown in Supplementary Fig.   5c and 5d, respectively.

Supplementary Note 3. Carbonization protocols and proposed carbonization mechanism for cellulose
The carbonization protocols selected are based on the TGA analysis of cellulose precursors in Supplementary Fig. 5d. A dwell-time of 2 h at 300 ℃ was employed to take into account the significant weight loss at this temperature due cellulose depolymerization. Three types of carbon membranes were obtained by carbonization protocols at different final temperatures of 550, 700 and 850 ℃, while all other carbonization parameters (e.g., heating rate, dwell time, etc.) were the same.
Supplementary Fig. 13 outlines the transformation mechanism from cellulose precursors to CMS membranes, based on the characterization results of HR-TEM, XPS, Raman spectra and TGA-MS.
When the final carbonization temperature is below 600 °C (in this work, 550 °C was used), disordered carbon "plates" were formed by intramolecular rearrangement, and a higher content of oxygen heteroatom existed in the carbon matrix by the formation of -OH, -COO, and -CH3 groups, which contributed to the more disordered structure (Supplementary Fig. 13c). As the final carbonization temperature was increased to over 600 °C, and especially over 800 °C, pendant groups, such as -OH and -CH3, were removed by forming H2O and CO2, which resulted in a more ordered carbon structure (Supplementary Fig. 13d). This was also supported by the higher sp 2 carbon content and lower oxygen content in the XPS spectra. Furthermore, according to HR-TEM and PSD, it can be proposed that the ultramicropores are from the inter-planar spacing, while the micropore contribution is from the imperfect packing of the carbon sheets.

Supplementary Note 4. Nanoindentation test
The hardness, the reduced elastic modulus, and Young's modulus of CHFMs were measured by nanoindentation tests using a Berkovich indenter. The CHFM samples were loaded to the maximum load (Pmax = 1 mN) in 5 s and then held for 2 s, followed by unloading in 5 s. The measured hardness and reduced elastic modulus are summarized in Supplementary Table 2. The Young's modulus (E, GPa) were estimated using the Oliver-Pharr method 1 as follows: Where ν and are E Poisson's ratio and Young's modulus of the CHFM samples, respectively. vtip and Etip are Poisson's ratio and Young's modulus of the indenter, respectively. vtip = 0.07 and Etip = 1140 GPa. Poisson's ratio of CHFMs is assumed to be the same and equal to 0.2. 2 Since Etip ≫ Er, the second term of the equation S1 is negligible. Hence, the Young's modulus of the samples is approximated to E = 0.96 Er.

Supplementary Note 5. XPS and Raman characterization
The CHFMs were characterized by XPS and Raman spectroscopy. The carbon content increases with the increase of carbonization temperature. Also, when the carbonization temperature is increased from 550 to 850 °C, the O contents was reduced from 9.26 Atomic % to 7.04 Atomic %.
The trace amount of N-element (ca. 0.6 Atomic % in CHFMs) presented in CHFMs is probably derived from residual EmimAc in the cellulose precursors. Such a low content may not have a significant effect on the micropore structure, as most of the reported N-doped porous carbon materials contain 3%-10%. Supplementary Fig. 12a shows each C 1s spectrum deconvoluted into three main peaks. The peak located at ~284.7 eV corresponds to sp 2 -hybridized carbon, while the peak at ~ 285.2 eV corresponds to sp 3 -hybridized carbon. The third peak at ~ 286.3 eV is assigned to C−N or C−O bonds 3,4 . The fourth peak with much lower intensity located at ~ 289.2 eV is attributed to the C=O bond 3,4 .
The Raman spectra of the CHFMs were obtained by deconvolution of the spectra into 5 peaks -D1, D2, D3, D4 and G (Supplementary Fig. 12b). All CHFMs exhibit two major peaks, namely the G peak (Graphite band), located at ~1600 cm -1 , which corresponds to the E2g-symmetry vibration mode of sp 2 hybridized carbon, and the D1 peak (Defect band) located at ~1346 cm -1 which is the A1g-symmetry vibration mode from the disordered graphite 5,6 . The D1 band is active when ring defects (ring breaks) are present within the graphite planes. The D2 band at ~ 1620 cm -3 is assigned to graphitic lattice vibrations mode with E2g symmetry (disordered graphitic lattice).
The D3 and D4 band are generally exhibited in highly disordered carbonaceous materials 5,6 . The D3 band, located at ~1525 cm -1 is usually ascribed to amorphous carbon, while the D4 band at ~1165 cm -1 is attributed to disordered graphitic lattice or sp 3 impurities 7 .

Supplementary Note 6. Apparent activation energies for the CHFMs
The apparent activation energies for the CHFMs were calculated by the Arrhenius relationship between gas permeance and testing temperature. Gas permeability (P, barrer) is the product of diffusivity (D) and sorption coefficient (S). Thus, gas permeance (P/l, GPU) can be described by Supplementary Equation (2) The apparent activation energies of for both H2 and CO2 permeate through different CHFMs are presented in Supplementary Fig. 15.

Supplementary Note 7. Possible issues for membrane fabrication scale-up
Scaling-up of CHFMs prepared from cellulose and ionic liquids may still face the following challenges, and the potential solutions are listed correspondingly.
1. Reducing the ratio of EmimAc/DMSO in dope solution and recovering EmimAc can be applied to bring down the relatively high cost of ionic liquids.
2. It is crucial to drain the tars and remove vapors during the carbonization if large amounts of fibers are carbonized in a furnace. By setting a small angle (e.g., 6°) between the quartz support and furnace can be used for draining tars. The prominent peaks of G and D1 correspond to the E2g-symmetry vibration mode of sp 2 hybridized carbon and the A1g-symmetry vibration mode from disordered graphite, respectively.
The D2 band at ~ 1620 cm -1 is assigned to graphitic lattice vibrations mode with E2g-symmetry (disordered graphitic lattice). The D3 band, located at ~1525 cm -1 is usually ascribed to amorphous carbon, while the D4 band at ~1165 cm -1 is attributed to disordered graphitic lattice or sp 3 impurities.