We synthesized the yellow PIM-1 powder by spirobisindane-tetrafluoroterephthalonitrile polycondensation [17]. As a type of microporous polymer, the ladder-like and rigid chains give the polymer numerous interconnected nanopores [17–19]. Nitrogen adsorption–desorption isotherms and pore width distribution indicated the high porosity of PIM-1 with a specific surface area of 790 m² g⁻¹ calculated by the Brunauer–Emmett–Teller (BET) theory and a uniform nanopore diameter of 1.7 nm from the slit-pore model-based nonlocal density functional theory (Fig. 1C and D and Fig. S1). Water vapor isotherm (Fig. 1C), with a low uptake even close to saturation pressure, was a typical curve for the porous materials containing superhydrophobic nanopores [20]. Such pore properties were conducive to gas permeability and plasma impermeability of membranes. PIM-1 membranes are typically prepared by solvent evaporation phase conversion and have a thickness of tens to hundreds of micrometers, which restricts their permeance [18,21]. Asymmetric membranes have thin skin layers [22,23]. Here, NIPS with a chloroform solvent and a methanol nonsolvent was developed to construct the AsNp membranes. Through casting the PIM-1 chloroform ink on a polytetrafluoroethylene substrate (PTFE) (Figs. S2 to S6), the AsNp composite membrane was formed by simply standing at atmosphere and quenching in methanol. As illuminated in scanning electron microscopy (SEM) images (Fig. 1E to G), the AsNp membrane consisted of a skin layer and a porous sublayer. Solvent chloroform was evaporated during atmosphere standing to form dense shin layer. When the casted layer was immersed in a coagulant bath, the solvent exchange between nonsolvent and solvent, i.e., methanol and chloroform, migrated into and out of the PIM-1 matrix, respectively, and induced phase separation and solidification. With the shortening of evaporation time, the skin layers became thinner (Figs. S2 to S5); e.g., the AsNp membrane with a standing time of 1 min had a thickness of ~4.0 μm (Fig. 1E), which was an order of magnitude smaller than that of conventional PIM-1 membranes with an entirely dense structure [17,20]. An atomic force microscopy (AFM) image illustrated a smooth surface with a root mean square roughness of 2.1 nm and an arithmetic average roughness of 1.6 nm, from a probe area of 1 × 1 μm² (Fig. 1H). A smoother surface usually meant less protein adsorption and mechanical friction to blood cells.

Transport behaviors of CO₂, O₂, and N₂ through the AsNp membranes are measured and presented in Fig. 2A and B. Gas permeation in polymer membranes is related to molecular sorption/solution and diffusion [24]. Sorption is determined by gas condensability and its affinity to membrane, and diffusion is governed by gas kinetic diameter and membrane pore size. Benefitting from the preferential adsorption of PIM-1 toward CO₂ [25], the CO₂/O₂ and CO₂/N₂ selectivities of the AsNp membrane with a standing time of 1 min were 6.7 and 11.1, respectively, which were larger than those of PMP membranes [7,11]. Relative to the entirely dense PIM membrane with high CO₂ permeability but moderate permeance [21], the AsNp membrane exhibited a CO₂ permeance of 1,137 gas permeation units (1 GPU = 3.35 × 10⁻¹⁰ mol m² s⁻¹ Pa⁻¹) due to the thin skin layer. Meanwhile, the O₂ permeance was achieved at 203 GPU. Considering the thickness of the selective layer, the permeability of AsNp was poorer than the intrinsic one, implying the transfer resistance from the sublayer. As the thickness decreased, the permeance and CO₂/N₂ selectivity were enhanced and deteriorated, respectively. For example, the AsNp membrane with a standing time of 0.5 min had CO₂ and O₂ permeances of 3,545 and 1,169 GPU, respectively. We prepared the classic PDMS and PSF membranes and measured their permeation property for comparison (Fig. 2C and D and Figs. S7 to S9). The permeance and selectivity of the PDMS membrane were good but still obviously inferior to those of AsNp. We studied the aging behavior of the AsNp membrane (Fig. S10). After aging for 270 days, the CO₂ permeance (600 GPU) of AsNp was higher than that of PDMS as well. Apart from gas permeation, water inaccessibility is critical, especially for oxygenation with pressure drop at the blood side. Unlike the porous PSF membrane having SEM-detectable pores and being permeable to water (Figs. S8 and S11), the AsNp membrane with superhydrophobic nanopores was inaccessible to water like the PDMS membrane, even at a transmembrane pressure of 0.4 MPa and after filtration for several hours (Fig. S11). The above features revealed that the AsNp membrane was promising for eliminating plasma leakage and performing fast gas exchange.

Some typical experiments, protein adsorption, platelet adhesion and activation, hemolysis, and clotting time, were performed to investigate hemocompatibility. Protein adsorption on foreign materials initiates coagulation cascade and thrombosis. Bovine serum albumin, a representative protein, adsorbed on AsNp with an amount of 10.5 μg cm⁻², which was lower than that of the widely biologically used PSF and PDMS and other biomedical materials (Fig. 3A and Fig. S12) [3,11,13]. Platelet adhesion, activation, and clotting time were tested by using sheep blood. In contrast to PSF and PDMS showing few fluorescent speckles from platelet aggregation, almost no speckle could be observed on the AsNp membrane surface (Fig. S12D to F). The irregular platelets on PSF and PDMS indicated platelet activation, whereas the platelet adhered to AsNp kept the morphology without pseudopodia or deformation (Fig. S12D to F, insets). Hemolysis usually refers to erythrocyte rupture and hemoglobin (HGB) release and will cause a reduction in exchange efficiency and will have a toxic effect. After incubation in erythrocyte solution for 1 h, the hemolysis ratio (HR) of AsNp analyzed by HGB was only 0.29% (Fig. 3B). This value could meet the clinical requirement (2.0%) according to American Society for Testing and Materials standard and was much less than that of the PSF and PDMS membranes. Blood clotting index (BCI) denotes that blood components still disperse in plasma rather than adhere to membranes after clotting. Clearly, the AsNp membrane had a higher BCI than PDMS and PSF as time prolonged (Fig. 3C). Activated partial thromboplastin time (APTT), prothrombin time (PT), and thrombin time (TT) reflect intrinsic, extrinsic, and common coagulation pathways, respectively. The APTT, PT, and TT of AsNp (30.22 s, 17.00 s, and 11.85 s) were greater than those of PDMS and PSF (Fig. 3D to F). The fibrinogen (FIB) content (138.8 mg dl⁻¹) of AsNp was lower than that of PDMS and PSF (Fig. 3G), which could reduce fibrin formation and inhibit the occurrence of blood coagulation. All the above experimental results suggested the great anticoagulation and antithrombotic properties of the AsNp membrane.

Hemocompatibility correlates with various surface physicochemical properties of membranes. Both hydrophilic and hydrophobic modifications are applied to improve hemocompatibility, which drives the formation of a hydration layer or a low-energy interface and then relieves protein and cell attachments [10–14]. Compared with the other 2 membranes, the AsNp membrane showed moderate water (102.8^o) and blood (101.1^o) contact angles (Fig. 3H and Fig. S13A and B). In view of the superhydrophobicity of inner nanopores, the AsNp membrane surface was relatively hydrophilic. This discrepancy was probably attributed to the fact that the nanopores were generated after nonpolar chloroform evaporation, while the surface was solidified in methanol and had more exposed polar groups. A good hydrophobic–hydrophilic balance and a good distribution of membrane surface, derived from homogeneously nonpolar phenyl and methyl groups and polar cyanides and oxygen heterocycles, might be the reasons for the weakened adhesiveness and coagulation of AsNp, like amphiphilic polymers [26,27]. Besides the hydrophobic–hydrophilic nature, the repelling of the negatively charged surface, as identified by Zeta potential data (Fig. 3H), toward blood components with the same charged property contributed to hemocompatibility as well. Additionally, the smoother surface of AsNp than that of PSF and PDMS and other porous membranes, which could minimize contact area and shear force to blood components, was another important factor for ameliorating antithrombosis and antihemolysis (Fig. 1G and Fig. S13C and D) [3,26]. Thrombosis and hemolysis are complex and difficult to fully understand. Nevertheless, the data presented here confirmed the hemocompatibility of the AsNp membrane.

The gas exchange of the membranes was evaluated by oxygenation using water and porcine blood (Fig. S14). With a water flow acceleration from 10 to 400 ml min⁻¹, the O₂ exchange rate of the AsNp membrane averaged from 3 samples changed from 3 to 90 ml m⁻² min⁻¹ and was greater than that of PDMS (Fig. 4A). Notably, the leakage of the PSF membrane occurred due to wetting after an operation lasting tens of minutes (Fig. 4A, inset). Originating from the better gas solubility in blood, the O₂ exchange rate of the AsNp membrane for porcine blood with an oxygen saturation (SaO₂) of about 90% and a flow rate of 1 to 50 ml min⁻¹ was larger than that for deionized water with the same flow rates and reached 2 to 36 ml m⁻² min⁻¹ (Fig. 4B). Correspondingly, the CO₂ in blood was removed with a transport rate of 20 to 735 ml m⁻² min⁻¹ owing to preferential solubility and permselectivity (Fig. 4B). It was clear that blood, after making contact with the membrane, displayed elevated O₂ and decreased CO₂ concentrations (Fig. S15). Since a slower flow rate resulted in a longer dwell time, the increment and decrement of O₂ and CO₂ concentrations became more obvious as flow rate decreased. At 1 ml min⁻¹, the SaO₂ for the AsNp membranes increased to 99%, while the total CO₂ concentration (TCO₂) dropped from 21.8 to 18.0 mmol l⁻¹.

Further, the blood circulation was carried out with a SaO₂ of 60% ± 5% to simulate oxygenation for venous blood (Fig. 4C and D and Figs. S16 and S17). As circulation time prolonged, the blood was oxygenated and the SaO₂ increased to over 95%, which was similar to that of arterial blood. Comparatively, the AsNp membrane required a shorter oxygenation period of 25 min and exhibited a faster O₂ exchange rate of 20 to 60 ml m⁻² min⁻¹ than PDMS with values of 35 min and 10 to 30 ml m⁻² min⁻¹. With a blood circulation of 30 and 40 min for AsNp and PDMS with an initial TCO₂ (pCO₂) of 18.3 mmol l⁻¹ (70.1 mmHg) and 16.4 mmol l⁻¹ (60.7 mmHg), the TCO₂ (pCO₂) substantially reduced to 11.6 mmol l⁻¹ (26.8 mmHg) and 11.3 mmol l⁻¹ (26.4 mmHg), accompanied by a CO₂ removal rate of 100 to 350 and 60 to 160 ml m⁻² min⁻¹, respectively. Rapid CO₂ transport might be useful for mitigating CO₂ toxic effects. A downward trend of gas exchange rates over time was ascribed to the gradually weakened driving force for exchange due to the CO₂ removal and O₂ accumulation. Distinct from the CO₂/O₂ selectivity of AsNp and the permeance difference (4 to 6 times) between AsNp and PDMS, the related exchange rates differed by 5 and 2 times, respectively, which were attributed to partial pressure difference and transfer resistance from blood boundary [28,29]. In general, the HGB concentration of blood is positively correlated to exchange rate. By replacing the blood with an HGB concentration of 58 g l⁻¹ with another one (65 g l⁻¹), at a flow rate of 50 ml min⁻¹, the O₂ and CO₂ exchange rates of AsNp were enhanced to 170 and 740 ml m⁻² min⁻¹, respectively (Fig. S18). It was worth emphasizing that the parameters of membrane modules and operations affected gas transports immensely. Under similar conditions, the commercial PMP membranes had an O₂ exchange rate of 10 ml m⁻² min⁻¹ as reported in the literature [14]. In light of more HGB in the human blood and the rational parameters of the oxygenator, the AsNp membrane is expected to have a higher efficiency during practical application. Furthermore, the clean surface after blood circulation verified the hemocompatibility of the AsNp membrane (Fig. S19).

Anhydrous potassium carbonate (K₂CO₃), methanol, dichloromethane, toluene, acetone, chloroform, N-methyl-2-pyrrolidone (NMP), polysulfone (PSF), N,N-dimethylacetamide (DMAC), polydimethylsiloxane (PDMS), tetraethyl orthosilicate, dibutyltin dilaurate, cyclohexane calcium chloride (CaCl₂), sodium chloride (NaCl), and polytetrafluoroethylene (PTFE) substrate were purchased from Kutai Chemical Reagent Co., China. K₂CO₃ was dried at 393 K overnight before use. 2,3,5,6-Tetrafluoroterephthalonitrile (TFTPN) and 5,5’,6,6’-tetrahydroxy-3,3,3’,3’-tetramethyl-1,1’-spirobisindane (TTSBI) were purchased from Sigma-Aldrich. Spirobisindane was purified by crystallization in methanol and dichloromethane. Bovine serum albumin (BSA) was purchased from Biofroxx Biological Co., Germany. Sheep whole blood was purchased from Bestbio Biological Co., China. Porcine whole blood was purchased from Yuanye Biological Co., China. Calcein AM was purchased from UElandy Biological Co., China. APTT, PT, TT, and FIB reagents were purchased from Shanghai Sun Biotech Co., China.

Nitrogen adsorption–desorption isotherms were collected by using a physisorption analyzer (Autosorb iQ Station 1, Quantachrome Co.) at 77 K held using a liquid nitrogen bath. The multipoint BET method and the slit-pore model-based nonlocal density functional theory method were applied to calculate the specific surface area and the pore width distribution, respectively. Water vapor isotherms were collected using the physisorption analyzer at 298 K. A field-emission scanning electron microscope (SEM, Ultra-55, Zeiss Co.) with an accelerating voltage of 5 kV was used to observe the morphology of samples. An ultrathin gold layer was coated on the prepared sample by using an ion sputter coater to minimize the recharging effect. An atomic force microscope (AFM, Bioscope Catalyst Nanoscope-V, Bruker, USA) was employed to measure the surface roughness of membranes. An ultraviolet absorption experiment was carried out by using an ultraviolet-visible (UV-vis) spectrophotometer (UV-1780, SHIMADZU). An optical contact angle and interface tension meter (ST200KB, USA KINO Industry Co.) was used to investigate the water and blood contact angles of samples. The zeta potential of membranes was measured by using a streaming potential analyzer (SurPASS, Anton Paar, Austria) with the background electrolyte of 1 mmol l⁻¹ KCl solution.

TFTPN (0.01 mol), TTSBI (0.01 mol), and K₂CO₃ (0.03 mol) were added into a mixture of toluene (20.0 ml) and NMP (40.0 ml) in a 3-necked flask by stirring under nitrogen atmosphere. For polycondensation of PIM-1, the temperature of the flask was increased to 433 K using an oil bath and kept for 40 min. Then, massive methanol was added to precipitate the yellow PIM-1 product and stirring continued for 3 h. After synthesis, the yellow product was isolated by centrifugation and dried at 333 K. For purification, the PIM-1 product was dissolved in chloroform, re-precipitated and stirred in methanol overnight, separated by centrifugation, and washed 3 times with deionized water and acetone. Ultimately, the prepared PIM-1 powder was dried at 373 K for 2 days.

For fabrication of the AsNp membrane, the PIM-1 powder was dissolved in chloroform and stirred overnight at room temperature to obtain the polymer solution with a concentration of 5.0 wt%. To remove bubbles, the solution was stood for 12 h. A clean scraper was used to evenly cast the polymer ink on a PTFE substrate. Through standing at room temperature for a certain time (0.5, 1, 2, 5, or 10 min), the solvent was volatilized to form a skin layer. Then, the PIM-1/PTFE membrane was transferred and quenched in a nonsolvent coagulant bath at room temperature. Because of the methanol-chloroform mutual solubility and the insolubility of PIM-1 in methanol, the solvent exchange between the polymer matrix and coagulant bath induced the phase separation and PIM-1 solidification to form the AsNp membrane. After NIPS, the AsNp membrane was dried at room temperature for use.

For fabrication of the PSF membrane, the PSF after vacuum drying overnight was dissolved in DMAC and stirred overnight at room temperature to obtain the polymer ink with a concentration of 18 wt%. To remove bubbles, the PSF solution was stood for 12 h. A clean scraper was used to cast the polymer ink on a glass plate. After standing at room temperature for 5 s, the casted PSF layer and glass plate were immersed in deionized water for NIPS. Then, the PSF membrane was peeled off from the glass plate, washed several times with deionized water, and stored in deionized water for use.

For fabrication of the PDMS/PTFE membrane, the PDMS solution with a concentration of 5 wt% was prepared with cyclohexane solvent. For crosslinking, tetraethyl orthosilicate and dibutyltin dilaurate were used as a crosslinker and catalyst, respectively. PDMS, the crosslinking agent, and the catalyst with a weight ratio of 2:2:1 were mixed in cyclohexane. Then, the PTFE substrate was soaked in the solution for 5 min and the excess solution was removed. After dip-coating, the PDMS/PTFE membrane was thermally treated at 80 °C for 24 h to crosslink completely. For the PDMS/PSF membrane, the preparation procedures were similar to those for the PDMS/PTFE membrane, except that the PTFE substrate was replaced by the PSF membrane. It should be noted that the PDMS membrane in the manuscript usually represented the PDMS/PTFE membrane, unless otherwise specified.

The gas permeation property of the prepared membranes was assessed at room temperature. A membrane with an effective area of 3.14 cm² was sealed in a permeation cell by an O-ring. CO₂, O₂, or N₂ was injected into the feed side with a transmembrane pressure of 0.1 MPa. After running steady, the permeate gas was collected and recorded to evaluate the gas transport.

Permeance (P, GPU, 1 GPU = 3.35 × 10⁻¹⁰ mol m² s⁻¹ Pa⁻¹) was calculated using Eq. 1:where p_u and p_d are the upstream and downstream pressures, respectively. A is the effective membrane area. ΔV and Δt are the permeate volume and time of gas through the membrane, respectively. R and T are the gas constant value and temperature, respectively.

Permeability (P_er, Barrer, 1 Barrer = 3.35 × 10⁻¹⁶ m mol m⁻² s⁻¹ Pa⁻¹) was calculated using Eq. 2:where l is the membrane thickness.

Selectivity (S) was calculated using Eq. 3:where P_i and P_j are permeances of gas i and j, respectively.

The water permeation property of the prepared membranes was measured at room temperature by using a cross-flow filtration system. A membrane with an effective area of 7.0 cm² was sealed in a permeation cell by an O-ring. Water was injected into the feed side with a transmembrane pressure of 0.1, 0.2, 0.3, or 0.4 MPa and a cross-flow rate of 30.0 l h⁻¹. After running steady, the permeate water was collected and recorded to calculate flux.

Water flux (F, L m⁻² h⁻¹) was calculated using Eq. 4:where ΔV and Δt are the permeate volume and time of water through the membrane, respectively. A is the effective membrane area.

Before measurement, the membrane sample with an area of 1 × 1 cm² was immersed in phosphate buffer solution (PBS) solution (pH 7.4) for 24 h to remove possible impurities. For protein adsorption, the membrane was incubated in BSA PBS solution with a concentration of 1.0 mg ml⁻¹ at 310 K for 2 h. After soaking, the membrane sample was taken out and washed with PBS solution to remove non-adsorbed protein. Subsequently, the membrane sample was immersed in sodium dodecyl sulfate (SDS) solution with a concentration of 2 wt% and subjected to ultrasonication for 1 h to dissolve the adsorbed protein on the membrane surface. More than 95% of the adsorbed protein could be eluted into SDS solution. Ultimately, the amount of BSA was measured by using a UV-vis spectrophotometer to study the adsorption.

Platelet adhesion and activation were investigated by using a fluorescence microscope and SEM. Sheep whole blood (10 ml) was centrifuged at 1,000 rpm for 15 min and the supernatant was separated to obtain platelet-rich plasma (PRP). For platelet adhesion test, the membrane sample with an area of 1×1 cm² was incubated in PRP (500 μl) and was shaken at 100 rpm with a temperature of 310 K for 2 h. After incubation, the membrane sample was taken out and slightly rinsed with PBS to remove the non-adherent platelets. For fluorescence staining, the membrane sample was incubated in Calcein-AM PBS (2 μM) solution at 310 K for 30 min and rinsed with PBS. A fluorescence microscope (490/515 nm) was used to observe the platelet aggregation on the membrane. For platelet activation, the membrane sample with an area of 1×1 cm² was immersed in PBS solution at 310 K for 1 h and incubated in PRP (500 μl) at 310 K for 2 h. After incubation, the membrane sample was taken out and slightly rinsed by PBS solution to remove the non-adherent platelets. The membrane sample was immersed in glutaraldehyde solution with a concentration of 2.5 wt% at 277 K for 24 h to fix the adhered platelets. After immersion, the membrane sample was rinsed with PBS solution and dehydrated through a series of gradient concentrations of alcohol/PBS solutions (25%, 50%, 75%, 95%, and 100%) for 15 min each. After dehydration, the platelet adhesion and activation on membrane was observed using SEM.

A membrane sample with an area of 1 × 1 cm² was rinsed 3 times with deionized water and normal saline, and then was immersed in normal saline (10 ml) at 310 K for 30 min. Sheep whole blood (8.0 ml) was diluted by normal saline (10 ml), and the diluted whole blood (200 μl) was dropped into the above normal saline containing the soaked membrane sample. After incubation of the membrane sample at 310 K for 60 min, the normal saline containing diluted whole blood was treated by centrifugation at 2,500 rpm for 5 min. Then, the supernatant was separated and its absorbance was measured by using a UV-vis spectrophotometer at 545 nm. Meanwhile, the diluted whole blood (200 μl) was directly added into deionized water (10 ml) or normal saline (10 ml), and the absorbance of the supernatant from centrifugation was measured as a positive or negative control, respectively.

HR was calculated using Eq. 5:

HS represents the absorbance of sample. HN and HP represent the absorbance of negative control and positive control, respectively.

A membrane sample with an area of 1 × 1 cm² was placed at the bottom of a beaker and kept at 310 K for 5 min. Sheep whole blood (100 μl) was dropped on the membrane surface, and the CaCl₂ solution (10 μl, 0.2 mol·l⁻¹) was added to the membrane surface as well. After incubation at 310 K for a certain time (5, 10, 15, or 20 min), the deionized water (25 ml) was added in the beaker and was shaken at 30 rpm and 310 K for 10 min. A UV-vis spectrophotometer was used to measure the absorbance of the solution at 545 nm. Meanwhile, sheep whole blood (100 μl) was directly added into the deionized water (25 ml) and its absorbance was measured using a UV-vis spectrophotometer as a control value.

BCI was calculated using Eq. 6:

I_s represents sample absorbance. I_w represents control absorbance.

Activated partial thromboplastin time (APTT), prothrombin time (PT), thrombin time (TT), and fibrinogen (FIB) were measured by a semi-automatic blood coagulation analyzer (MTN-II, matenu) to evaluate the anticoagulant properties of the membrane. Sheep whole blood (10 ml) was centrifuged at 3,000 rpm for 15 min and the supernatant was separated to obtain platelet-poor plasma (PPP). The membrane sample with an area of 0.5 × 0.5 cm² was incubated in PPP (100 μl) with a temperature of 310 K for 30 min. After incubation, the incubated PPP (50 μl) and APTT agent (50 μl) were added into a test cup. After incubation at 310 K for 3 min, the CaCl₂ solution (50 μl, 0.025 mol·l⁻¹, incubated at 310 K for 10 min before using) was added, and then the APTT was measured. For PT test, the incubated PPP (50 μl) was added into a test cup at 310 K for 3 min and the PT agent (100 μl, incubated at 310 K for 10 min before using) was added, and then the PT was measured. For the TT test, the incubated PPP (100 μl) was added into a test cup at 310 K for 3 min and a TT agent (100 μl) was added, and then the TT was measured. For the FIB test, the incubated PPP (50 μl) was diluted by the FIB buffer solution (450 μl) and the diluted PPP (50 μl) was added into a test cup. After incubation at 310 K for 3 min, the FIB thrombin solution (50 μl) was added, and then the FIB content of PPP was calculated based on a standard curve.

The water oxygenation performance of the prepared membranes was evaluated by using a self-made apparatus shown in Fig. S7. A membrane with an effective area of 3.5 × 10 cm² was sealed into a rectangular membrane module. Deionized water was deoxygenated by N₂ with a flow rate of 160 ml min⁻¹ to reduce O₂ concentration below 1.0 mg l⁻¹, and was injected into one side of the membrane module with a flow rate of 10, 50, 100, 200, or 400 ml min⁻¹. For oxygenation, O₂ was input into the other side of the membrane module with a flow rate of 200 ml min⁻¹. Correspondingly, the flow rate ratios of feed gas to liquid were 20:1, 4:1, 2:1, 1:1, and 0.5:1. Oxygenation system temperature was maintained at 310 K. Water O₂ concentration was measured by using a dissolved oxygen meter (JPB-607A, Leici). The oxygen exchange rate of the membranes was calculated by the O₂ concentration of the water before and after oxygenation.

The oxygen exchange rate for water (R_W, ml m⁻² min⁻¹) was calculated using Eq. 7:where Q_W (l min⁻¹) is the flow rate of water. c_b (ml l⁻¹) and c_a (ml l⁻¹) are the O₂ concentrations of solutions before and after oxygenation, respectively. A (m²) is the effective membrane area.

Similar to water oxygenation, blood oxygenation of the prepared membranes was investigated by using the self-made apparatus. Animal porcine whole blood (500 ml) was firstly deoxygenated by the N₂ and CO₂ mixture with a ratio of 9:1 and a flow rate of 50 ml min⁻¹ to obtain the blood with an oxygen saturation (SaO₂) of about 90%. To reduce the hemolysis from flow shear force, porcine blood was injected into one side of the membrane module at a flow rate of 50, 10, 5, or 1 ml min⁻¹. Meanwhile, O₂ was injected into the other side of the membrane module with a flow rate of 200 ml min⁻¹. Correspondingly, the flow rate ratios of feed gas to liquid were 4:1, 20:1, 40:1, and 200:1. The oxygenation system temperature was maintained at 310 K and the membrane with an effective area of 3.5 × 10 cm² was sealed in the membrane module. Blood sample was taken at the inlet and outlet of the membrane module. A blood gas analyzer (i-STAT G300, Abbott) was employed to analyze HGB concentration, O₂ partial pressure, oxygen saturation, CO₂ partial pressure, HCO₃⁻, total CO₂ concentration, etc. of blood before and after oxygenation.

The oxygen concentration (cO₂) of blood was calculated using Eq. 8:where cO₂ (ml l⁻¹) is the O₂ concentration of blood. c_Hb (g l⁻¹) is HGB concentration. SaO₂ is the oxygen saturation of blood. pO₂ (mmHg) is the partial pressure of O₂. 1.34 is the amount in milliliters of O₂ that could be carried by 1.0 g of HGB. 0.03 is the amount in milliliters of O₂ that could be dissolved for each 1 mmHg per liter of blood.

Oxygen exchange rate (R_O2,1, ml m⁻² min⁻¹) was calculated using Eq. 9:where Q_b (l min⁻¹) is the blood flow rate. cO_2,b (ml l⁻¹) and cO_2,a (ml l⁻¹) are the O₂ concentrations of blood before and after oxygenation at the inlet and outlet of the membrane module, respectively. A (m²) is the effective membrane area.

Carbon dioxide exchange rate (R_CO2,1, ml m⁻² min⁻¹) was calculated using Eq. 10:where Q_b (l min⁻¹) is the flow rate of blood. TCO_2,b (mmol l⁻¹) and TCO_2,a (mmol l⁻¹) are the total CO₂ concentrations of blood before and after oxygenation at the inlet and outlet of the membrane module, respectively. A (m²) is the effective membrane area.

Oxygenation circulation was carried out by using the self-made apparatus. Animal porcine whole blood (100 ml) was firstly deoxygenated by the N₂ and CO₂ mixture with a ratio of 9:1 and a flow rate of 50 ml min⁻¹ to obtain the blood with an oxygen saturation (SaO₂) of 60% ± 5%. Porcine blood with a flow rate of 50 or 400 ml min⁻¹ was circulated from a reservoir tank to a membrane module with an effective membrane area of 3.5 × 10 cm² and then placed back to the tank. Meanwhile, O₂ was injected into the membrane module with a flow rate of 200 ml min⁻¹. Correspondingly, the flow rate ratios of feed gas to liquid were 4:1 and 0.5:1. Oxygenation system temperature was held at 310 K. A blood gas analyzer (i-STAT G300, Abbott) was employed to monitor the HGB concentration, O₂ partial pressure, oxygen saturation, CO₂ partial pressure, HCO₃⁻, total CO₂ concentration, etc. of the blood in the reservoir.

Oxygen exchange rate (R_O2,2, ml m⁻² min⁻¹) was calculated using Eq. 11:where cO₂ (ml l⁻¹) is the O₂ concentration of the blood in the reservoir. V (l) is the blood volume. t (min) is the oxygenation time. A (m²) is the effective membrane area.

Carbon dioxide exchange rate (R_CO2,2, ml m⁻² min⁻¹) was calculated using Eq. 12:where TCO₂ (mmol l⁻¹) is the total CO₂ concentration of the blood. V (l) is the blood volume. t (min) is the oxygenation time. A (m²) is the effective membrane area.