Sub-100 nm carbon molecular sieve membranes from a polymer of intrinsic microporosity precursor: Physical aging and near-equilibrium gas separation properties

Journal Pre-proof Sub-100 nm carbon molecular sieve membranes from a polymer of intrinsic microporosity precursor: Physical aging and near-equilibrium gas separation properties Wojciech Ogieglo, Tiara Puspasari, Xiaohua Ma, Ingo Pinnau PII:

S0376-7388(19)31816-2

DOI:

https://doi.org/10.1016/j.memsci.2019.117752

Reference:

MEMSCI 117752

To appear in:

Journal of Membrane Science

Received Date: 17 June 2019 Revised Date:

17 November 2019

Accepted Date: 13 December 2019

Please cite this article as: W. Ogieglo, T. Puspasari, X. Ma, I. Pinnau, Sub-100 nm carbon molecular sieve membranes from a polymer of intrinsic microporosity precursor: Physical aging and nearequilibrium gas separation properties, Journal of Membrane Science (2020), doi: https://doi.org/10.1016/ j.memsci.2019.117752. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier B.V.

Graphical abstract

Sub-100 nm Carbon Molecular Sieve Membranes From a Polymer of Intrinsic Microporosity Precursor: Physical Aging and Near-Equilibrium Gas Separation Properties Wojciech Ogieglo, Tiara Puspasari, Xiaohua Ma, and Ingo Pinnau* Functional Polymer Membranes Group, Advanced Membranes and Porous Materials Center, Division of Physical Sciences and Engineering, King Abdullah University of Science and Technology, 23955 Thuwal, Saudi Arabia

*Corresponding author: [email protected]

ABSTRACT Challenging molecular separation tasks continue to drive the need for advanced membranes, which could combine high productivity and selectivity with outstanding chemical resistance and mechanical robustness. Here, ultrathin, sub-100 nm carbon molecular sieve (CMS) membranes based on a polymer or intrinsic microporosity precursor were fabricated and characterized with respect to their separation performance for six gases, He, H2, N2, O2, CO2 and CH4. The rarely reported in thin CMS membranes physical aging was tracked until near-equilibrium was reached. The use of commercially available, small pore size (2-5 nm) γ-alumina-coated α-alumina ceramic supports allowed for an easy membrane precursor fabrication by a simple solution coating process. The subsequent application of a protective polydimethylsiloxane layer, as often done in research and membrane production, assured excellent defect control and produced CMS composite membranes with very high membrane permselectivities (e.g. CO2/CH4 = 84.5, H2/CH4 = 360). No significant deviations from bulk properties in terms of the chemical decomposition were discovered in the film thickness range of 60 – 300 nm. However, a surprisingly massive acceleration of the physical aging in comparison to the bulk was detected in all sub-100 nm membranes including the precursor polymer. All CMS membranes reached near-equilibrium gas transport properties within 1 – 1.5 months providing a rare opportunity to study aging-stabilized performance. For the aging-stabilized sub-100 nm membranes a decrease in gas permeabilities of up to 3 orders of magnitude was detected. Remarkably, sub-

1

100 nm CMS membranes showed significantly increased selectivities typical to thick isotropic CMS membranes pyrolyzed at 100 – 200 °C higher temperatures. A clear aging-related shift to more significant size sieving behavior occurred for the majority of the studied CMS membranes throughout the aging period. Overall, our study provides an important observation that excessive reduction of the selective layer thickness, especially in initially highly microporous materials (PIMs, CMS etc.), may not present the best strategy for the optimization of the long-term membrane performance.

Keywords ultra-thin composite membranes, carbon molecular sieves, gas separations, physical aging, nano-confinement

2

1. Introduction Membrane technology promises significant energy savings in the sector of industrial separations (constituting ~10-15% of the world’s total energy consumption) and is widely viewed as one of the key technologies in combating climate change [1,2]. Membranes are expected to play an important role in providing worldwide supply of natural gas which has been replacing other fossil fuels and will continue to rapidly expand in this area [3]. In addition, multiple reports suggest excellent performance of some membrane materials in very challenging olefin/paraffin or organic solvent separations [1,4–8]. However, to fully rise to its potential membrane technology needs to provide high separation performance, excellent membrane stability and robust gas separation devices [9]. Carbon molecular sieves (CMS) are promising membrane candidates due to their extraordinary separation performance and easy tunability (mainly by adjusting pyrolysis temperature) [10,11], as well as scalability potential which can build on the experience of the existing technologies used for polymeric membranes (e.g. hollow fiber spinning, solution coating and module production) [12–14]. CMS membranes are fabricated by controlled pyrolysis of an organic polymer precursor (e.g. high performance polyimide) under essentially inert atmosphere in the range of 400 – 1000 °C while avoiding full graphitization occurring above 2500 °C. The resulting amorphous, turbostratic and carbon element-rich structure comprises hierarchical micropores with a coexistence of larger micropores (7-20 Å) and ultramicropores (<7 Å). The larger micropores are predominantly responsible for the high permeability whereas the smaller micropores are thought to provide sharp size exclusion leading to high gas-pair selectivities. In this respect, CMS membranes structurally resemble polymers of intrinsic microporosity (PIMs), which display high backbone rigidity combined with a contorted chemical structure leading to frustrated packing, which traps large amounts of excess free volume. PIMs, however, have rarely been used as actual CMS precursors and only a handful of studies exist primarily with a focus on bulk olefin/paraffin separation properties [4,5,15]. To provide sufficient gas permeances (expressed as pressure-normalized gas flux, P/L, where P is permeability and L is selective skin thickness) both polymeric and CMS membranes are commonly fabricated in the form of a dense skin on top of a highly porous support. The porous support can either be of a different chemical nature (e.g.

3

inorganic alumina, porous carbon, porous metal) or of the same chemical composition but with much higher porosity (e.g. asymmetric hollow fibers produced by phase inversion [13]). A few previous studies reported on fabrication and characterization of thin-skinned CMS composite membranes with selective layers in the range of several microns. In some of the earlier studies, Fuertes and Centeno et al. [16–19] presented selective thin-skinned CMS composite membranes made by carbonization of commercial polyimides, phenolic resin, and poly(vinylidene chloride-co-vinyl chloride) with thicknesses ranging from 0.8 to about 3 microns on various inorganic supports. Some of those membranes achieved interesting separation properties with O2/N2, CO2/N2 and CO2/CH4 selectivities of up to 14, 45 and 160, respectively. Ma et al. [20] fabricated 1.6 µm CMS composite membranes supported by a porous γ-alumina membrane with O2/N2 selectivity of 5.1 and high mixed-gas C3H6/C3H8 selectivity of up to 36. Very few reports exist on CMS membranes with selective layers significantly thinner than 1 micron that are expected to provide high gas permeances and high gas-pair selectivities. The reason is drastically increased probability of pinhole defects, which are detrimental to the membrane selectivity in gas separation applications. In their study, Ma et al. [20] fabricated CMS membranes as thin as 386 nm at 500 - 550 °C with the aim to study their olefin/paraffin separation performance and noticed that a reduction in CMS thickness below 0.5 micron resulted in a drop in selectivity due to changes in the CMS micropore structure. In particular, a higher packing density was suggested as the origin of increased selectivity for the thinnest CMS membranes. Most recently, Hou et al. [21] used carbon nanotube networks as nano-scaffolds to support ultrathin CMS membranes with selective layers in the range 100 – 1000 nm and found the best overall performance between the gas permeance and selectivity for the intermediate 322 nm CMS thickness. No reports exist on the separation performance of even thinner (sub-100 nm) CMS membranes, which could potentially provide a desirable combination of high gas permeances and selectivities and can be fabricated in an easily up-scalable manner for commercialization. In this work, we investigated the fabrication and separation performance of sub-100 nm γ-alumina/α-alumina supported flat-sheet CMS membranes produced by pyrolysis of a high carbon content PIM polyimide (PIM-PI) precursor made by one-pot polycondensation reaction of a spirobifluorene dianhydride and 3,3′-

4

dimethylnaphthidine (SBF-DMN) Fig. 1 [22]. As such, these membranes present some of the thinnest CMS membranes reported to date. By using the intermediate 2-5 nm pore size, low roughness γ-alumina support high quality ultra-thin precursor films could be directly spin coated in a single step and excellent defect control was achieved by using a thin (~ 1 µm) PDMS protective coating. Because of the greatly accelerated physical aging of the extremely thin CMS layers a unique opportunity to observe near-equilibrium properties of the CMS membranes could be provided.

Fig. 1. Chemical structure of the PIM-PI used as a carbon precursor.

2. Experimental Part 2.1 Sample preparation The CMS precursor polymer was synthesized by polymerization of a spirobifluorene-based dianhydride with 3,3’dimethylnaphthidine (SBFDA-DMN) as previously described [22]. The polymer possessed molecular weight Mn = 6.5 x 104 g mol-1 with a polydispersity of 1.9 and internal surface area SBET = 686 m2 g-1. The decomposition temperature for the pristine bulk polymer observed by TGA in nitrogen was ~520 °C (onset). The particularly high (aromatic) carbon content of 84 wt.% combined with its intrinsically microporous character was anticipated to result in high quality ultra-thin CMS membranes. The γ-alumina-coated α-alumina ceramic disc supports (25 mm diameter and 2 mm thickness) were obtained from Corbra Technologies B.V. (Rijssen, The Netherlands). The bulk of the support consisting of α-alumina was produced by sintering ~150 nm alumina particles, leading to a porosity in the range of 30-35%, followed by mechanical polishing on one side. Subsequently a 2-3 micron γ-alumina layer with a porosity of >55% was

5

fabricated using sol-gel dip coating (usually twice) with boehmite solution followed by high temperature calcination. The resulting γ-alumina surface was very smooth with pores in the range of 2-5 nm and allowed for high quality spin coating of the sub-100 nm CMS precursor. The deposition of the CMS precursor was done by a single spin-coating step at 2000 rpm with a diluted and filtered polyimide solution (~1 wt %) in chloroform. Pyrolysis was conducted in a tube furnace under a constant flow of high purity nitrogen gas. Oxygen concentration was monitored and kept below 7 ppm. The pyrolysis protocol involved heating the samples from room temperature to a set point (500, 600, 700 or 800 °C) at 3 °C/min followed by passive cooling to room temperature within about 10 h. Prior to gas separation performance measurements, the samples were coated with a ~500 nm polydimethylsiloxane (PDMS) layer to cover the non-selective defects and protect the sample surface.

2.2 Spectroscopic ellipsometry measurements and modeling

30

Psi [°]

25

500 °C 600 °C

Bare support

20

Precursor

15 700 °C

10

800 °C

5 0 200

700

1200

1700

2200

Wavelength [nm] Fig. 2. Spectroscopic ellipsometry psi spectra of the pristine γ-alumina support, as well as of the precursor and carbon membrane films pyrolyzed at different temperatures.

Spectroscopic ellipsometry [23–25] (M-2000 DI from J. A. Woollam Co., Inc.) was used to determine polyimide precursor and carbon film thickness and refractive index on top of the γ-covered-α-alumina supports. The measurement consisted of recording the ellipsometry spectra, Fig. 2, (expressed in psi and delta) in a wavelength range of 192 – 1700 nm at three angles of incidence (65, 70 and 75°) on 5 random locations on the membrane

6

surface with focusing optics reducing the spot size to 300 µm (short axis). The optical model representing the membrane samples comprised bulk α-alumina covered with a γ-alumina layer, Fig. 3. In most cases, the γ-alumina layer itself was split into two separate layers reflecting the actual substrate fabrication process in which the γalumina coating was applied twice. The two separate layers differed slightly in porosity, which had to be accounted for in the optical modeling (see Supporting Information). All model constituents, i.e. the α-alumina substrate and γ-alumina layers, were modeled using effective medium approximation, EMA which can be used to extract porosity [24] by mixing the optical properties of a dense Al2O3 (data from literature [26]) with void (n = 1.000) while the optical properties of the precursor and carbon films pyrolyzed up to 600 °C were modeled using the well-known Cauchy formula in the range 600 – 1700 nm. The approach was similar to the previous study with ultra-thin POSS films deposited on analogous supports [27]. For carbon films made above 600 °C the Cauchy formula was used with an added Urbach absorption tail and exponent to account for light absorption. Given the substantial number of fit parameters a multiparameter global fit option was applied where the chosen parameters (thickness of the carbon or precursor film, thicknesses and porosities of the γ-alumina layers as well as the porosity of the bulk) were varied within a certain range while calculating the error of the fit. The lowest value of the fit error was considered a global minimum and resulted in excellent fits and well-resolved fit parameters (see Supporting Information).

Fig. 3. (a) Scanning electron microscopy image of the CMS membrane cross-section together with a scheme of the membrane cross-section (note that for the performance measurements an additional ~0.5 micron PDMS layer

7

is present on top) (b) Optical images showing the top view of the sample surfaces as a function of the pyrolysis temperature. For the precursor samples deposited on silicon wafers 500 nm silicon oxide wafers were used to benefit from the interference enhancement effect [28,29] to more accurately probe properties of sub-100 nm films. For the temperature scans up to 550 °C, an INSTEC heated stage was used under pure N2 flow at 10 °C/min heating or cooling rate. Changes in the optical properties of crystalline silicon, as well as of silicon oxide, were accounted for by using temperature-dependent optical constants libraries built in the CompleteEASE 6.43 software package.

2.3 Gas separation performance Gas separation properties were measured at 21 ± 1 °C by sealing the membranes in a commercial 25 mm highpressure filter cell (Millipore Corp.) with the active membrane area limited by a 17.8 mm internal diameter O-ring and using a bubble flow meter to measure the flow rate of the permeate. The gas testing sequence was O2, N2, He, H2, CH4 and CO2 in the pressure range of about 6 – 8 bar with a sufficient time for reaching steady state (at least 1-2 hours and up to 16 hours for slowly permeating N2 and CH4 in the case of samples aged for more than 5-10 days). For the aging period up to 5 days all measurements could be completed within ~ 8 hours. The first membrane performance measurement was done after about 16 hours since membrane fabrication (pyrolysis and PDMS coating).The relatively high pressure was necessary to assure sufficiently high flow for accurate measurements due to generally low permeances. The permeance was expressed in gas permeation units (1 GPU = 1 x 10-6 cm3(STP) cm-2 s-1 cmHg-1 = 3.35 x 10-10 mol m-2 s-1 Pa-1). The CMS layer permeability, Pcarbon, is expressed in Barrer (1 Barrer = 1 x 10-10 cm3(STP) cm cm-2 s-1 cmHg-1 = 3.35 x 10-16 mol m m-2 · s-1 · Pa-1). To extract the thin CMS layer permeability, Pcarbon, in the presence of a PDMS overcoat a resistances-in-series model was applied [30] (see Supporting Information). Gas separation performance was followed in time for up to 1.5 months while keeping the membrane sealed in the measurement cell at room temperature. Ideal selectivity for a gas pair, αi/j, was calculated as the ratio of the permeances of gas i over gas j. For each pyrolysis temperature two separate membrane samples were prepared and characterized.

8

2.4 Atomic force microscopy measurements Atomic force microscopy was recorded using a Dimension Icon apparatus with a TESPA type probe for the roughness and topology determination in a tapping mode and the PeakForce option with a TAP150A probe for the stiffness assessment. For the stiffness assessment (a measure of how much the sample surface resists deformation by the cantilever tip), the spring constant, optical sensitivity and tip radius of the probe were first calibrated and later accounted for in the analysis. Image processing was done with Gwyddion 2.49 software package.

3. Results and Discussion A scanning electron micrograph (SEM) of the cross-section of an ultra-thin carbon layer composite membrane used in this study is shown in Fig. 3a. The use of a highly porous, small pore size (2-5 nm) and low roughness intermediate γ-alumina layer allowed for the deposition of a high quality carbon precursor with thickness just below 100 nm, as determined by spectroscopic ellipsometry [23,24] and consistent with SEM. In all cases, the ellipsometry-determined porosity of the γ-alumina layer did not change from the prior determined porosity of an uncoated substrate indicating virtually no precursor intrusion and thus a very sharp interface between the thin CMS layer and the intermediate γ-alumina layer (see Supporting Information). The sharpness of the interface is further evident from the well-resolved oscillatory features seen in the raw ellipsometry data, shown in Fig. 2. If the coating polymer significantly intruded the γ-alumina layer during the membrane fabrication the amplitude of those oscillations would be expected to significantly reduce or even disappear altogether as a result of the low refractive index contrast between the respective layers. This is clearly not the case in Fig 2. The inorganic membrane substrate remained unchanged upon pyrolysis up to the maximum temperature of 800 °C employed in this study. Fig. 3b displays the optical images of the membrane surface as a function of the pyrolysis temperature where a clear change in light absorption is seen going from a yellowish precursor to dark shiny grey at 800 °C. The carbon precursor film showed a significant volume reduction in response to increasing pyrolysis temperature with a simultaneous sharp increase in the refractive index (Fig. 4a). These changes reflected a collapse of the microporous organic polymer structure into a highly conjugated, aromatic (hence high refractive index) carbon molecular sieve following the partial chemical decomposition and evolution of volatiles originating from

9

heteroatoms. Corresponding TGA data is provided in the Supporting Information for bulk films. Even though the optical absorption of the samples increased significantly, all of the thin films remained transparent enough in the near-IR spectral region for accurate thickness and refractive index determination due to their very small thickness. The residual transparency is clearly seen in Fig. 2 where even after pyrolysis at 800 °C spectral oscillations caused by the γ-alumina layers are clearly observed above about 700 nm.

Fig. 4. (a) Relative thickness (hcarbon/hprecursor) and refractive index versus pyrolysis temperatures. (b) Urbach absorption indicating the extent of polymer degradation as a function of temperature for thin films deposited on Si-wafers. (c) Relative thickness of the precursor films scanned upon cooling from 550 °C under inert atmosphere (N2) to determine the apparent glass transition temperature for different film thicknesses; data offset on the Y-axis for better visibility. It is evident that the largest change in the morphology occurred between 600 and 700 °C with the sample pyrolyzed at 600 °C being still largely polymeric (refractive index not too different from the precursor and the 500 °C sample), whereas the sample pyrolyzed at 700 °C was already transforming into a true carbon molecular sieve. For the particular precursor used in this study, pyrolysis at 500 °C induced comparatively little change and rather led to a densified, annealed polymer with only slight chemical degradation. This is in agreement with the outstanding thermal stability of the bulk polymer with the onset of thermal degradation at around 520 °C under inert atmosphere as determined by the TGA [31]. However, it is known that thin polymer films may deviate from the bulk properties with respect to several important physical characteristics, such as glass transition temperature, physical aging and stability[32–40]. Hence, high temperature Si-wafer ellipsometry experiments were performed on several films of different thicknesses. Fig. 4b shows the Urbach absorption tail as a function of temperature under inert atmosphere heating at a rate of 10 °C/min for various film thicknesses. The Urbach absorption tail is

10

indicative of light absorption toward the UV part of the spectrum which is often increased for partial polymer degradation [41]. Clearly, for all films polymer degradation up to 550 °C occurred to some extent with a magnitude following the order of film thickness with the thickest film degrading the least. However, the extent of the degradation was somewhat limited especially for the 300 nm and 60 nm CMS films, with the latter corresponding best with the membrane samples used for gas separation performance. Similar conclusions can be drawn by examining the development of the near-IR light absorption (Supporting Information) related to increasing electrical conductivity. Upon temperature scanning in cooling mode while monitoring film thickness a weak change of slope in the curve could be detected (Fig. 4c). In glassy polymers, this slope change corresponds to a change of the thermal expansion coefficient upon traversing the glass transition temperature, Tg. In this view, by finding an intersection point between the linear regions above and below the break it is possible to tentatively assign a Tg in the range 522 – 525 °C which does not seem to vary as a function of the film thicknesses between 25 – 300 nm. Data for the 9 nm CMS film were not of sufficiently high quality for Tg assignment due to experimental scatter. The high value of the Tg is in agreement with the PIM-PI nature of the polymeric precursor. The tentative character of the Tg assignment is related to potential partial polymer degradation. The challenge of an overlap between the Tg region and the chemical degradation is a known issue in PIMs and has only recently been overcome by using an extremely fast heating rate (~104 °C s-1) to detect the Tg in PIM-1 [42]. For our membranes, given the results shown in Fig. 4b and c, it may be concluded that the thin carbon precursor films did not seem to deviate significantly from bulk in terms of the Tg or the degree of chemical degradation at least down to 60 nm film thickness.

11

Fig. 5. (a) AFM-derived surface roughness of polymer precursor and CMS membranes calculated for 5 micron square areas using the height topology map. (b) Stiffness distribution for all analyzed membranes. (c) and (d) Atomic force microscopy topology and stiffness, 5 micron and 3 micron square areas, respectively.

Fig. 5 shows the AFM-derived roughness, topology and stiffness characterization of all thin film membranes prior to deposition of the thin PDMS sealing layer. All surfaces seem relatively smooth with a roughness in the range of several nm. A slight increase from about 3 nm for the precursor film is seen with increasing pyrolysis temperature, peaking at 600 °C with about 5.5 nm, Fig. 5a. A gradual slight surface morphology evolution was observed going from the precursor sample to the one pyrolyzed at 800 °C. Circular “hole-like” features with a diameter in the range of 50-70 nm (Supporting Information) were detected on the surfaces of all samples, however, more distinctly only for the polymer precursor and the 500 °C CMS sample. These features most

12

probably originated from the dewetting instabilities related to a highly extended surface of the sub-100 nm films that can potentially form pinholes detrimental to membrane selectivity. However, their low surface coverage fraction (Adefect/Asample) estimated at ~2% and an application of a PDMS overcoat resulted in a rather small effect on the selectivity, as discussed below.

Fig. 6. Permeances of the sub-100 nm PIM-PI precursor and carbon molecular sieve membranes tracked over time for six permanent gases; some scatter especially in the data for the slowest permeating species reflects experimental inaccuracies because of the proximity to the detection limits of the used experimental apparatus. The film surface stiffness, determined using the PeakForce protocol of the AFM device, was around 1 ± 0.25 GPa for all samples with a significant increase going from the precursor to the sample pyrolyzed at 500 °C and a subsequent stepwise decline for increasing pyrolysis temperatures, Fig. 4b (inset). A slight increase in stiffness

13

inhomogeneity was observed by increasing pyrolysis temperatures as reflected by broadening of the stiffness histograms, Fig. 5b, and visualized in the stiffness topology images, Fig. 5d. The permeances of six gases as a function of the pyrolysis temperature and physical aging time are shown in Fig. 6. The data for samples pyrolyzed at 800 °C could not be determined accurately due to extremely low gas permeances. In all cases, a very quick decline in permeances was observed followed by stabilization within about 30 days. This remarkably quick physical aging is consistent with the substantially faster aging dynamics in thin films observed in many prior studies for amorphous polymers [36,37,43,44] and suggests that the intermediatecarbon matrix undergoes an aging process similar to organic polymers. This is in agreement with the findings of Ma et al. [20] who noticed a strongly accelerated physical aging in thinner carbon molecular sieve films in the range 386 – 2640 nm. Interestingly, our study showed that in almost all cases CO2 started as the most permeable gas for fresh samples but with aging its permeance quickly decreased just below that of the smaller He and H2. This indicates a surprisingly quick densification and tightening of the polymeric precursor (Fig. 6a) or CMS matrix (Fig. 6b-6d) leading to the development of a predominantly size-sieving aged structure. The quick loss of CO2 permeance in our thin films is also faster than what is normally observed in bulk membranes and might be related with the reduced solubility in ultra-thin microporous amorphous materials [45]. As a result, ideal CO2/CH4 and CO2/N2 selectivities do not increase with time contrary to the general behavior in bulk membranes. Similar decline of about 1-2 orders of magnitude occurred in the case of the larger, less permeable O2, N2 and CH4. The aging process appeared to be the fastest (stable permeance within 10 – 15 days) for the carbon films pyrolyzed at 700 °C in which case also the lowest permeances of all samples were observed. This is consistent with the general behavior of carbon molecular sieves where the higher pyrolysis temperature usually leads to a tighter, less permeable structure. We note here that some researchers suggested oxygen chemisorption to be one of the possible causes for the gradual loss of permeance in CMS membranes in time. However, the work of Xu et al [46] suggested that the performance of CMS membranes kept under vacuum might be similar to those exposed to the atmosphere and thus the physical process of matrix densification is the dominant mechanism of aging in those materials. To

14

examine the influence of moisture or oxygen, two additional membranes (~ 97 nm thick) were pyrolyzed at 600 °C and immediately coated with PDMS, transported to membrane cells, and exposed only to CO2 and N2. The results (Supporting Information) suggest that at least in the case of our membranes, the aging process in CMS membranes seems unaffected by oxygen or moisture. In addition, Supporting Information contains CO2 and N2 permeances and ideal selectivities for freshly made membranes without PDMS coating to provide evidence of the necessity to seal defects in the ultrathin CMS layer. We also provide separation performance of the thin PDMS composite membranes. The uncoated CMS membranes possessed approximately five times lower CO2/N2 selectivities in comparison to the PDMS-coated counterparts. It is worth noting, that the largest magnitude of physical aging-related loss of permeance was found for the samples pyrolyzed at 600 °C, Fig. 6c. At this particular pyrolysis temperature, as evident from a separate BET analysis of thick isotropic films (Supporting Information), pore creation event takes place in the bulk. This information agrees well with the well-established fact of faster aging for glassy films with larger excess fractional free volumes. In fact, the behavior of all samples presented in Fig. 6 seems to strongly point to the dominating role of large free volume of our precursor PIM and derived carbons (N2 BET in excess of 500 m2/g) in the strong acceleration of the physical aging. The pyrolytic collapse has been shown to be much more pronounced in PIM vs. non-PIM precursors in our previous study [47]. We therefore speculate that materials with less initial microporosity (e.g. non-PIMs) would potentially not show such dramatic acceleration. However, a separate dedicated study would be needed to provide more evidence, also regarding the differences between the pure PIM class (e.g. PIM-1) and the PIM-PI precursor described in the present study. Ideal selectivities as a function of time for the most technologically important gas pairs are displayed in Fig. 7. Within the first 10 days of physical aging much fluctuation is seen in all cases, which could again reflect the rather complex structural reorganizations occurring within the first days following pyrolysis. Alternatively, the large fluctuations could be explained by inaccuracies related with the significant time needed to complete measurement of permeances of all six gases (up to 8 hours). This effect diminishes naturally for samples aged for longer periods (above 5 days) where indeed less fluctuation is seen. Similar to the permeances data shown in Fig. 6, the selectivities nearly stabilized within approximately 30 days and for the samples pyrolyzed at 500 and 600

15

°C reached relatively high values (e.g. CO2/CH4_500C = 84, CO2/CH4_600C = 56). These high selectivities are matching some of the best purely polymeric membrane materials, including from the PIMs class [48–51]. In agreement with the work of Ma et al. [20] physical aging resulted in a significant increase in the membrane selectivity for gas pairs with significant differences in molecular size (e.g. H2/N2) strengthening the suggestion of an accelerated rate of aging-related collapse of the larger micropores versus the smaller ultramicropores [13]. In the referenced work, also acceleration of aging with reduced thickness was described in thick films region (23 – 172 µm), which is conceptually consistent with our work, For example, 172 µm thick chemically similar PIM-PI (EAD-DMN) lost about 44% of CO2 permeability over a period of 430 days of aging while a 23 µm thick film of the same polymer lost a comparable 52% of CO2 permeability only over a period of 50 days. In a recent work of Tiwari et al. on ultra-thin PIM-1 films [45], very significant reductions of permeabilities, consistent with our findings (above one order of magnitude), were found in a film slightly thicker than those used in our work, ~220 nm PIM-1. Somewhat surprisingly, the carbon molecular sieves pyrolyzed at 700 °C did not display the highest selectivities as would have been anticipated based on the data obtained for thick isotropic bulk samples [31]. Possible reasons for this behavior could include the development of a significant volume/area fraction of micro-scale defects for samples pyrolyzed above 600 °C. This behavior occurred consistently upon several fabrication attempts. The defects were, however, not unambiguously detected by AFM (Fig. 5). In the work of Zhang et al. [14] the authors found that for very tight, high-temperature carbons the nitrogen over methane selectivity might be strongly enhanced as a result of sorptive-exclusion of the larger methane molecules. The tight CMS layer would effectively show almost barrier properties for methane but still allow some permeance of the smaller nitrogen. If the tight CMS layer is defective (pinholes, cracks etc.) then most of methane transport would be forced through the PDMS-covered defects. This could lead to enhancement of methane transport (PDMS is methane over nitrogen selective, CH4/N2 = 3.3) and lack of significant aging-related permeance loss. Methane permeance would stay more or less constant in time. This is precisely the behavior we observe in Fig. 6d. We note that 600 °C thus

16

would represent a limiting pyrolysis temperature for obtaining defect-free (or at least PDMS-sealable) sub-100 nm CMS membranes made on γ-alumina supports.

Figure 7. Ideal selectivities for some of the most relevant gas pairs of the sub-100 nm PIM-PI precursor and carbon molecular sieve membranes tracked over time.

Fig. 8a depicts the permeability values calculated from the resistances-in-series model for the equilibrated samples as a function of the gas kinetic diameter for all investigated membranes. A clear size sieving effect is seen for all samples, including the precursor. This indicates that while the bulk precursor properties displayed a reverse selective (solubility-driven) behavior with the CO2 permeability larger than that of both He and H2 (Table 1) upon transformation into a sub-100 nm films and sufficient physical aging the character changed towards size sieving (Table 2). In addition, an enormous decrease (up to three orders of magnitude) in permeability for all gases and pyrolysis temperatures was observed going from bulk to aging-stabilized sub-100 nm films. These

17

results are consistent with findings of Sanyal et al. [13] who postulated existence of a “hyperskin” – an outermost submicron skin layer in carbon hollow fibers with greatly reduced permeability but essentially unaffected selectivity. The existence of such a layer reduces the overall permeances of thin skin carbon hollow fiber membranes much below that which would be anticipated based on bulk properties. However, a closer examination of the results for the bulk and sub-100 nm CMS membranes, shown in Tables 1 and 2, indicates that while the permeabilities were greatly reduced the selectivities for the polymer precursor and CMS membranes pyrolyzed at 500 and 600 °C increased substantially. The slower gases (N2 and CH4) appear to be excluded much more strongly in the sub-100 nm membranes leading to quite attractive selectivities that are three to eight times higher than for the bulk. This trend discontinued for the sample pyrolyzed at 700 °C, which shows the least attractive combination of selectivities, as discussed before. However even in this case, the selectivities remain substantially better than bulk polyimide precursor for three out of four gas pairs (Table 1). The massive physical aging is put into context in Table 3 where the experimentally obtained permeances are confronted with theoretical permeances calculated from the membrane thickness (derived from ellipsometry) and bulk permeabilities (Table 1). A difference of 2-3 orders of magnitude, especially for aged ultra-thin films, is clear in most cases. To more clearly reflect the dramatic increase in size sieving ability of the sub-100 nm CMS membranes with aging Fig. 8b shows the fresh and aging-stabilized selectivities of the size-selectivity sensitive gas pairs, He/CO2 and N2/CH4 plotted for all analyzed membranes. A universal enhancement of the size sieving effect with aging and increasing pyrolysis temperature indicated by increasing selectivities is seen for all samples, including the polymer precursor. Particularly sharp increases were observed for the He/CO2 pair for which the dissimilarity of the molecular sizes is significantly larger than for N2/CH4 (∆dHe/CO2 = 0.7 Å and ∆dN2/CH4 = 0.16 Å). For the sample pyrolyzed at 700 °C, as discussed earlier, this effect is less conclusive.

18

Fig. 8 (a) Permeability as a function of gas kinetic diameter for the aged sub-100 nm PIM-PI precursor and carbon molecular sieve membranes. (b) Ideal selectivities for the prominent size discrimination-sensitive gas pairs (He/CO2 and N2/CH4, with kinetic diameter ratios of 2.6 Å/3.3 Å and 3.64 Å/3.8 Å, respectively); selectivity over 1 indicates size-sieving behavior while selectivity below 1 indicates reverse-selective behavior dominated by solubility effects (CO2 and CH4 are more soluble than He and N2).

19

Table 1. Bulk Permeabilities and Selectivities for the Carbon Molecular Sieves and the SBFDA-DMN Polyimide Precursor Determined for ~70 µm Films [31].

Bulk Membrane

Bulk Permeability (Barrer)

Selectivity (-)

He

H2

CO2

O2

N2

CH4

O2/N2

CO2/CH4

CO2/N2

H2/N2

H2/CH4

Precursor

1340

2966

4700

850

226

326

3.8

14.4

20.8

13

9.1

500 °C

350

800

1233

210

54

70

3.9

17.6

22.8

14.8

11.4

600 °C

1073

3402

2853

663

110

77.5

6.0

36.8

25.9

31

44

700 °C

903

2190

236

85

9

6

9.4

39.3

26.2

243

365

800 °C

418

974

105

35

3.2

0.9

10.8

113

32.5

302

1047

20

Table 2. Permeabilities and Selectivities for the Fully Aged Carbon Molecular Sieves and the SBFDA-DMN Polyimide Precursor for the Sub-100 nm Composite Membranes.

Sub-100 nm Membrane

Thin Film Permeability (Barrer)

Selectivity (-)

He

H2

CO2

O2

N2

CH4

Precursor 87 nm

2.27

3.10

1.91

0.24

0.07

0.06

3.7

500 °C 69 nm

5.22

5.94

2.36

0.37

0.06

0.03

600 °C 82 nm

1.00

1.39

0.22

0.07

0.01

700 °C 72 nm

0.52

0.49

0.09

0.02

0.004

21

O2/N2

CO2/CH4

CO2/N2

H2/N2

H2/CH4

32.3

28.9

47.0

52.3

6.5

84.5

42.1

105

213

0.004

7.4

56.5

24.2

156

360

0.006

6.6

22.5

15.3

134

82.3

Table 3. Experimental versus Theoretical (Based on Bulk Permeability) Permeances for the Fresh and Fully Aged Carbon Molecular Sieves and the SBFDA-DMN Polyimide Precursor for the Sub-100 nm Composite Membranes.

Sub-100 nm Membrane

Permeance (GPU)

He

H2

CO2

O2

N2

CH4

Precursor 87 nm (Exp. Fresh)

41.6

55.6

238.6

14.5

4.5

5.2

Precursor 87 nm (Exp. Aged)

26.6

36.6

23.1

2.9

0.8

0.7

Precursor 87 nm (Theor.)

15 400

34 100

54 000

9 800

2 600

3 750

500 °C 69 nm (Exp. Fresh)

121.4

196.3

294.5

52.6

12.0

9.9

500 °C 69 nm (Exp. Aged)

59.5

69.4

28.7

4.4

0.7

0.3

500 °C 69 nm (Theor.)

5 100

11 600

17 900

3 000

780

1 000

600 °C 82 nm (Exp. Fresh)

107.2

195.0

319.6

59.1

2.5

15.9

600 °C 82 nm (Exp. Aged)

12.5

17.5

2.8

0.8

0.1

0.05

600 °C 82 nm (Theor.)

13 100

41 500

34 800

8 100

1 300

950

700 °C 72 nm (Exp. Fresh)

15.7

20.4

8.0

2.7

0.4

0.1

22

700 °C 72 nm (Exp. Aged)

7.7

7.4

1.4

0.4

0.06

0.09

700 °C 72 nm (Theor.)

12 500

30 400

3 300

1 200

130

80

Fig. 9. Performance of the sub-100 nm PIM-PI precursor and carbon molecular sieve membranes relative to the Robeson 2008 polymer upper bounds for selected gas pairs; for each membrane type the values with higher permeability indicate fresh (1 day aged) samples while the values with lower permeability indicate aged (1 – 1.5 months) samples. Data for several other gas pairs is presented in the Supporting Information.

Gas separation performance of polymeric membranes is often conveniently benchmarked using the so-called tradeoff relationships first reported by Robeson in 1991 and later updated in 2008 [48,52]. Upon examination of the existing experimental gas permeation Robeson established that polymer membrane materials with a high permeability usually possess reduced selectivity, and vice versa. The theoretical basis for this observation was later developed by Freeman [53] and supported by molecular dynamics simulations [54] tracing back to an interplay of the backbone stiffness and interchain spacing. Fig. 9 places the membranes fabricated in this work on the Robeson’s trade-off diagrams for selected gas pairs (the remaining data are shown in the Supporting

23

Information). In each case, the point at a higher permeability value indicates the freshly made membrane whereas the point for the lower permeability represents the aging-equilibrated membrane. Thus, aging trajectories can be compared among the investigated samples. Clearly, due to the massive physical aging effect all sub-100 nm membranes place themselves much below the trade-off lines. This presents important insight and suggests that excessive reduction of membrane selective layer thickness, especially in initially high microporosity materials, may not be the best strategy to optimize long-term membrane performance. For some gas pairs, for example the O2/N2 and CO2/N2, the aging results in a predominantly horizontal move (reducing the permeability with a small effect on the selectivity) whereas for others, like H2/N2 and H2/CH4 usually a very substantial increase in the selectivity is observed. The former modest effect on the selectivity seems to be related with rather small differences in the molecular size (CO2 3.3 Å, O2 3.46 Å and N2 3.64 Å). For the gas pairs where the molecular size difference is larger (H2 2.89 Å, N2 3.64 Å and CH4 3.8 Å) the aging-induced structural tightening seems to produce a greatly enhanced selectivity concurrently with the reduced permeability.

4. Conclusions In this study, formation and gas separation performance of ultra-thin, sub-100 nm carbon molecular sieve membranes supported on an inorganic alumina flat disc supports were investigated. The chosen support morphology with a smooth, low pore size (2-5 nm) γ-alumina intermediate layer can potentially allow for easy scale-up using only a single coating step from a dilute polymer solution. Excellent defect control was achieved with a subsequent application of ~ 500 nm polydimethylsiloxane protective coating which essentially eliminated the influence of unwanted pinholes for membranes pyrolyzed up to 600 °C. Focused-beam spectroscopic ellipsometry proved useful in accurate determination of the pyrolysis-induced volume changes and thus allowed extracting gas permeabilities. Atomic force microscopy analysis indicated a low average surface fraction of pinhole defects (~2%) and revealed relatively small changes in both roughness and stiffness up to a maximum employed pyrolysis temperature of 800 °C. Because of the extremely high temperature stability of the precursor polymer (degradation onset of the bulk ~520 °C) a dilatometric signature of a glass transition temperature, Tg,

24

could be detected at approximately 522 °C. Neither the Tg or the thermal degradation differed significantly in a thickness range 60 – 300 nm while below 60 nm the thermal stability of the films seemed slightly affected. A massively enhanced physical aging effect was detected leading to a reduction of gas permeabilities up to 3 orders of magnitude in comparison to the bulk. A clear shift from a reverse-selective to size-sieving behavior within the aging period was observed. Because of the fast aging all analyzed samples reached near-equilibrium within 1 – 1.5 months thus providing a unique opportunity to study intrinsic, aging-stabilized separation performance. Significant (even up to 8 fold) increases in gas selectivity were detected for ultrathin CMS films (e.g. CO2/CH4 = 84.5, H2/CH4 = 360) versus the bulk film properties. The very high selectivities combined with acceptable permeances (resulting from the sub-100 nm skin thickness) may render the prepared membranes of potential interest for commercial scale-up.

Acknowledgments This work was supported by funding (FCC/1/1972-35-01) from the King Abdullah University of Science and Technology (KAUST). Help of Dr. Semen Shikin from the KAUST Solar Center with the AFM measurements and data interpretation is gratefully acknowledged. The generous access to the spectroscopic ellipsometry setup of the KAUST Solar Center is greatly appreciated.

Supporting Information Supporting information presents more details of spectroscopic ellipsometry modeling, atomic force microscopy analysis, as well as additional gas separation performance data.

References [1]

[2] [3] [4]

R.P. Lively, D.S. Sholl, From water to organics in membrane separations: Membrane materials provide economical means to achieve various separation processes - And their capabilities for processing organic fluids look set to expand significantly, Nat. Mater. 16 (2017) 276–279. D.S. Sholl, R.P. Lively, Seven chemical separations to change the world, Nature 532 (2016) 435–437. W.K. Coal, Coal and Natural Gas Calculations, in: Flash Smelt., Springer, 2013: pp. 280–284. O. Salinas, X. Ma, Y. Wang, Y. Han, I. Pinnau, Carbon molecular sieve membrane from a microporous spirobisindane-based polyimide precursor with enhanced ethylene/ethane mixed-gas selectivity, RSC Adv. 7 (2017) 3265–3272.

25

[5]

[6] [7] [8] [9] [10] [11] [12] [13]

[14] [15]

[16] [17] [18] [19] [20] [21]

[22]

[23] [24]

[25] [26] [27]

R.J. Swaidan, X. Ma, I. Pinnau, Spirobisindane-based polyimide as efficient precursor of thermallyrearranged and carbon molecular sieve membranes for enhanced propylene/propane separation, J. Membr. Sci. 520 (2016) 983–989. J.S. Adams, A.K. Itta, C. Zhang, G.B. Wenz, O. Sanyal, W.J. Koros, New insights into structural evolution in carbon molecular sieve membranes during pyrolysis, Carbon 141 (2019) 238–246. S.J. Kim, P.S. Lee, J.S. Chang, S.E. Nam, Y.I. Park, Preparation of carbon molecular sieve membranes on low-cost alumina hollow fibers for use in C3H6/C3H8 separation, Sep. Purif. Technol. 194 (2018) 443–450. K.S. Liao, S. Japip, J.Y. Lai, T.S. Chung, Boron-embedded hydrolyzed PIM-1 carbon membranes for synergistic ethylene/ethane purification, J. Membr. Sci. 534 (2017) 92–99. H.B. Park, J. Kamcev, L.M. Robeson, M. Elimelech, B.D. Freeman, Maximizing the right stuff: The tradeoff between membrane permeability and selectivity, Science 356 (2017) 1138–1148. R. Lively, Carbon molecular sieve membranes aim to cut energy use in hydrocarbon separations, Membr. Technol. 2017 (2017) 9–10. A. Singh-Ghosal, Air separation properties of flat sheet homogeneous pyrolytic carbon membranes, J. Membr. Sci. 174 (2002) 177–188. S.M. Saufi, A.F. Ismail, Fabrication of carbon membranes for gas separation - A review, Carbon 42 (2004) 241–259. O. Sanyal, S.T. Hicks, N. Bhuwania, S. Hays, M.G. Kamath, S. Karwa, R. Swaidan, W.J. Koros, Cause and effects of hyperskin features on carbon molecular sieve (CMS) membranes, J. Membr. Sci. 551 (2018) 113–122. C. Zhang, W.J. Koros, Ultraselective carbon molecular sieve membranes with tailored synergistic sorption selective properties, Adv. Mater. 29 (2017) 1701631. X. Ma, R. Swaidan, B. Teng, H. Tan, O. Salinas, E. Litwiller, Y. Han, I. Pinnau, Carbon molecular sieve gas separation membranes based on an intrinsically microporous polyimide precursor, Carbon 62 (2013) 88–96. A.B. Fuertes, D.M. Nevskaia, T.A. Centeno, Carbon composite membranes from Matrimid® and Kapton® polyimides for gas separation, Microporous Mesoporous Mater. 33 (1999) 115–125. T.A. Centeno, A.B. Fuertes, Supported carbon molecular sieve membranes based on a phenolic resin, J. Membr. Sci. 160 (1999) 201–211. T.A. Centeno, A.B. Fuertes, Carbon molecular sieve gas separation membranes based on poly(vinylidene chloride-co-vinyl chloride), Carbon 38 (2000) 1067–1073. T.A. Centeno, A.B. Fuertes, Carbon molecular sieve membranes derived from a phenolic resin supported on porous ceramic tubes, Sep. Purif. Technol. 25 (2001) 379–384. X. Ma, Y.S. Lin, X. Wei, J. Kniep, Ultrathin carbon molecular sieve membrane for propylene/propane separation, AIChE J. 62 (2016) 491–499. J. Hou, H. Zhang, Y. Hu, X. Li, X. Chen, S. Kim, Y. Wang, G.P. Simon, H. Wang, Carbon nanotube networks as nanoscaffolds for fabricating ultrathin carbon molecular sieve membranes, ACS Appl. Mater. Interfaces. 10 (2018) 20182–20188. X. Ma, B. Ghanem, O. Salines, E. Litwiller, I. Pinnau, Synthesis and effect of physical aging on gas transport properties of a microporous polyimide derived from a novel spirobifluorene-based dianhydride, ACS Macro Lett. 4 (2015) 231–235. W. Ogieglo, H. Wormeester, K.J. Eichhorn, M. Wessling, N.E. Benes, In situ ellipsometry studies on swelling of thin polymer films: A review, Prog. Polym. Sci. 42 (2015) 42–78. W. Ogieglo, H. Wormeester, M. Wessling, N.E. Benes, Spectroscopic ellipsometry analysis of a thin film composite membrane consisting of polysulfone on a porous α-alumina support, ACS Appl. Mater. Interfaces. 4 (2012) 935–943. W. Ogieglo, I. Pinnau, M. Wessling, In-situ non-invasive imaging of liquid-immersed thin film composite membranes, J. Membr. Sci. 546 (2018) 206–214. M. Ritala, M. Leskelä, Atomic layer deposition, College of Engineering and Applied Science, University of Rochester, 2002. M.J.T. Raaijmakers, W. Ogieglo, M. Wiese, M. Wessling, A. Nijmeijer, N.E. Benes, Sorption behavior of

26

[28]

[29] [30] [31]

[32] [33] [34] [35] [36] [37] [38] [39] [40] [41]

[42]

[43] [44] [45] [46] [47]

[48] [49]

[50]

compressed CO2 and CH4 on ultrathin hybrid poly(POSS-imide) layers, ACS Appl. Mater. Interfaces. 7 (2015) 26977–26988. W. Ogieglo, B. Ghanem, X. Ma, M. Wessling, I. Pinnau, High-pressure CO2 sorption in polymers of intrinsic microporosity under ultrathin film confinement, ACS Appl. Mater. Interfaces. 10 (2018) 11369– 11376. W. Ogieglo, A. Furchner, B. Ghanem, X. Ma, I. Pinnau, M. Wessling, Mixed-penetrant sorption in ultrathin films of polymer of intrinsic microporosity PIM-1, J. Phys. Chem. B. 121 (2017) 10190–10197. I. Pinnau, J.G. Wijmans, I. Blume, T. Kuroda, K.V. Peinemann, Gas permeation through composite membranes, J. Membr. Sci. 37 (1988) 81–88. K. Hazazi, X. Ma, Y. Wang, W. Ogieglo, A. Alhazmi, Y. Han, I. Pinnau, Ultra-selective carbon molecular sieve membranes for natural gas separations based on a carbon-rich intrinsically microporous polyimide precursor, J. Membr. Sci. 585 (2019) 1–9. L. Hamon, Y. Grohens, Y. Holl, Thickness dependence of the glass transition temperature in thin films of partially miscible polymer blends, Langmuir. 19 (2003) 10399–10402. J.A. Forrest, K. Dalnoki-Veress, The glass transition in thin polymer films, Adv. Colloid Interface Sci. 94 (2001) 167–195. J.Q. Pham, K.P. Johnston, P.F. Green, Retrograde vitrification in CO2/polystyrene thin films, J. Phys. Chem. B. 108 (2004) J.R. Dutcher, M.D. Ediger, Materials science: Glass surfaces not so glassy, Science 319 (2008) 577–578. Y. Huang, D.R. Paul, Physical aging of thin glassy polymer films monitored by optical properties, Macromolecules. 39 (2006) 1554–1559. Y. Huang, D.R. Paul, Physical aging of thin glassy polymer films monitored by gas permeability, Polymer 45 (2004) 8377–8393. M.D. Ediger, J.A. Forrest, Dynamics near free surfaces and the glass transition in thin polymer films: A view to the future, Macromolecules 47 (2014) 471–478. W. Ogieglo, K. Tempelman, S. Napolitano, N.E. Benes, Evidence of a transition layer between the free surface and the bulk, J. Phys. Chem. Lett. 9 (2018) 1195–1199. S. Napolitano, E. Glynos, N.B. Tito, Glass transition of polymers in bulk, confined geometries, and near interfaces, Reports Prog. Phys. 80 (2017) 36602. B.T. Koziara, E.J. Kappert, W. Ogieglo, K. Nijmeijer, M.A. Hempenius, N.E. Benes, Thermal stability of sulfonated poly(ether ether ketone) films: on the role of protodesulfonation, Macromol. Mater. Eng. 301 (2016) 71–80. H. Yin, Y.Z. Chua, B. Yang, C. Schick, W.J. Harrison, P.M. Budd, M. Böhning, A. Schönhals, First clearcut experimental evidence of a glass transition in a polymer with intrinsic microporosity: PIM-1, J. Phys. Chem. Lett. 9 (2018) 2003–2008. Y. Huang, D.R. Paul, Effect of temperature on physical aging of thin glassy polymer films, Macromolecules 38 (2005) 10148–10154. M.S. McCaig, D.R. Paul, Effect of film thickness on the changes in gas permeability of a glassy polyarylate due to physical aging: Part I. Experimental observations, Polymer 41 (2000) 629–637. R.R. Tiwari, J. Jin, B.D. Freeman, D.R. Paul, Physical aging, CO2 sorption and plasticization in thin films of polymer with intrinsic microporosity (PIM-1), J. Membr. Sci. 537 (2017) 362–371. L. Xu, M. Rungta, J. Hessler, W. Qiu, M. Brayden, M. Martinez, G. Barbay, W.J. Koros, Physical aging in carbon molecular sieve membranes, Carbon 80 (2014) 155–166. W. Ogieglo, A. Furchner, X. Ma, K. Hazazi, A.T. Alhazmi, I. Pinnau, Thin composite carbon molecular sieve membranes from a polymer of intrinsic microporosity precursor, ACS Appl. Mater. Interfaces. 11 (2019) 18770–18781. L.M. Robeson, The upper bound revisited, J. Membr. Sci. 320 (2008) 390–400. N. Alaslai, B. Ghanem, F. Alghunaimi, E. Litwiller, I. Pinnau, Pure- and mixed-gas permeation properties of highly selective and plasticization resistant hydroxyl-diamine-based 6FDA polyimides for CO2/CH4 separation, J. Membr. Sci. 505 (2016) 100–107. F. Alghunaimi, B. Ghanem, N. Alaslai, M. Mukaddam, I. Pinnau, Triptycene dimethyl-bridgehead

27

[51] [52] [53] [54]

dianhydride-based intrinsically microporous hydroxyl-functionalized polyimide for natural gas upgrading, J. Membr. Sci. 520 (2016) 240–246. Y. Wang, X. Ma, B.S. Ghanem, F. Alghunaimi, I. Pinnau, Y. Han, Polymers of intrinsic microporosity for energy-intensive membrane-based gas separations, Mater. Today Nano. 3 (2018) 69–95. L.M. Robeson, Correlation of separation factor versus permeability for polymeric membranes, J. Membr. Sci. 62 (1991) 165–185. B.D. Freeman, Basis of permeability/selectivity tradeoff relations in polymeric gas separation membranes, Macromolecules. 32 (1999) 375–380. K. Zhang, S.K. Kumar, Defining the optimal criterion for separating gases using polymeric membranes, Soft Matter. 14 (2018) 9847–9850.

28

Highlights •

Sub-100 nm carbon molecular sieves made from a PIM precursor were characterized

•

Near-equilibrium gas transport properties reached within ~1.5 months

•

Massive losses of permeance and substantial increases in selectivity upon aging

•

Ultrathin CMS thickness not preferred for optimum membrane performance

Author Statement Wojciech Ogieglo: Conceptualization, membrane preparation and characterization, original draft preparation; Tiara Puspasari: Permeation measurements; Xiaohua Ma: Monomer and polymer syntheses. Ingo Pinnau: Conceptualization, reviewing and editing.

Declaration of interests X The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: