Field trial of hollow fiber modules of hybrid facilitated transport membranes for flue gas CO2 capture in cement industry

Alarming increase in global CO2 emissions warrants acceleration of CO2 capture technologies. In this work, testing of pre-pilot scale membrane modules containing hybrid facilitated transport membranes in hollow fiber configuration is reported. The tests were carried out using real flue gas from a slipstream in the Colacem cement plant located in Gubbio, Italy. With the fabricated modules, CO2 flux of up to 750 NL m 2h− 1 with a CO2 permeate purity ranging from 50 to 55 vol% was recorded. All pre-pilot membrane modules showed increased CO2 permeance in industrial testing (1.5 to 1.9x higher) compared to laboratory evaluation owing to the reliable water profile and high temperature of the flue gas from the chimney. Influence of operating parameters (e.g., pressures in the feed and permeate) were studied. Long-term testing showed no obvious reduction in permeation performance. Furthermore, the membranes with mobile carriers when exposed to the feed gas containing SOx and NOx exhibited good resistance to performance deterioration despite high concentrations of acidic impurities. Simulation studies based on validated experimental performance under industrial conditions reveal the high potential of the fabricated membranes as an efficient separation unit capable of achieving industrial capture rate and CO2 purity requirements using a relatively low membrane area.


Introduction
Global greenhouse gas emissions have been increasing since the industrial revolution, leading to the average atmospheric CO 2 concentration rising at an alarming rate of 3 ppm(v)/year, reaching 414 ppm(v) in 2020 [1]. Overt effects of climate change in the 21st century, along with the international consensus on its direct dependence on CO 2 emissions, expedite the need for urgent implementation of emission reduction technologies. Carbon capture and storage (CCS) has been proven to be a strategic approach to significantly cut-down CO 2 emissions, especially from large-point sources like power generation, cement industries and iron and steel production facilities [2].
Although studied over decades, CCS implementation is mainly limited by the availability of versatile, environmentally benign and economically favorable technology to capture CO 2 from flue gas [3]. Among the state-of-the-art technologies for CO 2 capture, membrane technology is particularly advantageous due to its simplicity, high flexibility and modularity, low energy requirement and carbon footprint [4]. Over the past years, numerous membrane materials have been developed with improved separation properties to increase the efficiency of the flue gas treatment process [5]. Typical membrane material development involves various strategies to engineer materials that have simultaneous high CO 2 permeability and CO 2 /N 2 selectivity [6]. Such membranes are usually fabricated in lab-scale as thick self-standing films. Interesting materials are then fabricated as thin-film composites (TFC) on porous supports and evaluated for separation properties using gas mixtures as feed gas [7]. In most cases, the membrane development process terminates at this juncture most likely due to the knowledge gap between lab-based research and industrial requirements or lack of a collaborative possibility to establish a constant feedback loop between technology developers and potential end-users [8]. These limitations result in only a small number of membrane materials being eventually scaled-up and tested in conditions pertinent to the targeted end-users [9]. The upscaled modules are expected to treat incoming flue gas and result in certain targets of CO 2 capture rate and CO 2 purity, depending on the requirements of the end-users. Typical targets for CO 2 capture rate are over 90%, while the purity requirements of the downstream CO 2 depend on the use of the purified stream, i.e., the requirement for transport and storage is usually > 98% and varies from case to case for utilization [10,11]. In most cases, a two-stage membrane process is used to reach both targets simultaneously since the selectivity of commercially available membranes is less than 50 for CO 2 /N 2 [12][13][14].
Recently, hybrid facilitated transport membranes (HFTMs) as hollow fibers with ultrathin selective layers (less than 200 nm) were developed, and superior separation performances were documented [15]. These HFTMs contain sterically hindered polyallylamine (SHPAA)/polyvinyl alcohol (PVA) blend as polymer matrix and size-optimized porous graphene oxide (pGO) as nanofiller at a very low loading of 0.2 wt% with respect to polymer. Additionally, CO 2 -reactive small molecules were used as mobile carriers to enhance reactive transport. Lab-scale tests revealed high CO 2 permeance of up to 825 GPU for these HFTMs with a CO 2 /N 2 separation factor of 31 measured at 35 • C using binary CO 2 /N 2 gas mixture (10 vol% CO 2 in N 2 ).
In this work, three different hybrid facilitated transport membranes (HFTMs) were scaled up in hollow fiber configuration and tested with real flue gas from a cement plant. The effects of operating conditions, such as feed gas water contents and upstream and downstream pressures, on gas permeation performances were studied in detail. Long-term permeation tests lasting an extended two-week period with continuous exposure to untreated flue gas was carried out to study the durability of the fabricated pre-pilot module. The membranes consisting of mobile carriers were subject to an independent impurity testing campaign in a controlled environment to study the resistance of the membranes to SO x and NO x in flue gas. Consequently, the experimental results of the field tests were used as a basis to perform a stage-cut analysis of the studied membranes in a simulated single-stage membrane separation process.

Materials
Poly(allylamine hydrochloride) (Mw = 120,000-200,000, Thermo Fisher Scientific, Sweden) was purified and converted to sterically hindered polyallylamine (SHPAA). The reaction involved the stoichiometric conversion of poly(allylamine) into poly-N-isobutyl allyl amine using potassium hydroxide (pellets, 99.9%, Sigma-Aldrich, Norway) in methanol at 50 • C under reflux conditions. [15] The precipitated SHPAA dissolved in DI water for 2 days to obtain a 6 wt% solution. Polyvinyl alcohol (Mw = 89,000-98,000, 89% hydrolyzed, Sigma-Aldrich, Norway) was dissolved in DI water at 80 • C for 4 h under reflux conditions to result in a 4 wt% solution. SHPAA and PVA were used in the ratio of 90:10 by weight.
GO used for pGO synthesis was supplied by Graphene-XT, Italy. Hydrogen peroxide (H 2 O 2 , 30% in water) used for pGO synthesis was supplied by Sigma Aldrich, Norway, after size optimization, as described in our previous study [15,16]. The concentration of pGO in the membrane is maintained at 0.2 wt% with respect to the total solid content in coating solution (Polymer + mobile carrier).
Poly(p-phenylene oxide) (PPO) hollow fibers (Parker A/S, Norway) with an inside diameter of 350 µm and outside diameter of 540 µm were used as support membranes.

Fabrication of pre-pilot modules
The HFTMs in this study were fabricated using a facile dip-coating method. A coating solution containing polymer, pGO nanofillers and mobile carriers (if any), with a total solid content of 0.15 wt% was used to coat individual fibers using the semi-automatic dip-coating procedure. Each fiber was coated twice in opposing directions to ensure defectfree coating of the selective layer. The thickness of the selective layer was about 200 nm. More details on the coating procedure can be found in our previous study [15]. Representative SEM images of surface and cross-section of HFTM can be found in Fig. S1 in Supplementary Information. Longer fibers were coated for pre-pilot modules when compared to lab-scale ones. Assembly into hollow fiber module was proven scalable with varying numbers of fibers, membrane area and packing density between the lab-scale and pre-pilot modules, as shown in Fig. 1 150 cm 2 to 200 cm 2 were assembled. The material configuration in the modules and the design aspects are summarised in Table 1.

In-field tests
The in-field pre-pilot testing was carried out at Colacem cement plant located in Gubbio, Italy. The production facility uses marl, limestone and small amounts of alternatives as raw materials with petroleum coke as fuel. The plant can operate at an effective clinker production capacity of 3000 tonnes day − 1 (34.7 kg s − 1 ). The conditions of clinker production during the in-field testing and the corresponding emitted flue gas properties are furnished in Table 2.
Flue gas without major pre-treatment was used in this in-field testing campaign. The process flow diagram of the pre-pilot membrane testing system is shown in Fig. 2. The flue gas feed to the membrane module was retrieved from the emission stack from a sampling point located at a height of 75 m. A ceramic dust filter (SP180-H -M&C Tech Group, 2 µm porosity) was employed to remove particulate matter before retrieval. The flow rate of wet flue gas was at around 10-12 L min − 1 (0.17-0.2 L s − 1 ). The membrane module was placed in a heated cabinet. The feed stream into the heated cabinet was at a temperature of 60-65 • C, while the heating cabinet was maintained at a constant temperature of 60 • C. A needle valve was used on the retentate side to control the feed pressure serving as a manual back pressure regulator. The permeate side of the membrane module was either connected to a vacuum pump or to the sweep gas depending on the tests. The compositions of permeate and retentate was measured using gas analysers -Horiba PG 350 SRM (Horiba, Japan) and Testo Model 350XL-350S (Testo, USA). The flows were measured using a flow meter, TSI 4143 (USA), and verified using a bubble flow meter. All gas lines were encased with electric heating covers with water knockouts at multiple points to prevent condensation of water and obtain desirable operating temperature.
The flue gas from the stack contained 9-11 vol% water at 115 • C. This water content corresponds to a relative humidity of 5.40 to 6.59 % RH at 1 bar and 115 • C. For the study on the water content effect on permeation properties, an additional humid stream (makeup stream) was mixed with the flue gas in an evaporator at 115 • C. The makeup stream was prepared by using a carrier gas (simulated flue gas, 12.6% CO 2 /14% O 2 /73.4% N 2 by volume) and DI water in a gas generator (HovaCAL® digital MF, IAS GmbH, Germany). The total water content was changed to 15 vol% and 20 vol% by using the external evaporator. The synthetic air (IP grade 20.93% O 2 and 79.07% N 2 ) was used in the study on the effect of feed pressures.
The gas permeation performance of the fabricated pre-pilot membrane modules was evaluated by CO 2 flux and CO 2 purity in the permeate after a single-stage operation. The CO 2 flux across the membrane was calculated by equation (1).
, v P is the permeate flow rate in cm 3 (STP) s − 1 , y CO2,P is the concentration of CO 2 in the permeate side and A eff is the effective membrane area in cm 2 . The calculated flux was verified with feed and retentate side mass balance wherever possible. Pressure ratio (φ) is defined as the ratio of total feed pressure to total permeate pressure represented by equation ( where P F and P P are the feed and permeate pressures in bar.

Lab-scale permeation tests
Lab-scale modules of the chosen HFTMs were also evaluated using an inhouse built humid mixed gas permeation rig at two conditions of RH. Synthetic feed gas mixture containing CO 2 /N 2 mixture (10/90 v/v) was

Table 2
Combustion parameters of the kiln and flue gas properties.

Cement kiln
used as feed and pure CH 4 was used as the sweep gas. The permeance of component 'i' was obtained using the following equation where the total permeate flow including sweep V P (ml s − 1 ) was measured at the exit at steady state conditions using a bubble flow meter. y H2O and y i denote the molar fraction of water and permeating species in the permeate flow, respectively. Partial pressures (in cm Hg − 1 ) are p i,F , p i,R and p i,P of the species 'i' in the feed, retentate and permeate, respectively. 〈p i,F , p i,R 〉 represents the average of feed and where y and x are the gas contents in the permeate and feed side, respectively.

Accelerated aging tests
In addition to the field tests carried out at the cement industry, the HFTMs containing mobile carriers were subject to an impurity exposure testing campaign at an independent facility (Pilot-scale Advanced CO 2 Capture Technology (PACT)) at the University of Sheffield (USFD). Membranes have been integrated into a slipstream of a gas turbine with and without injection of additional CO 2 and other trace gasses to investigate performance over a range of CO 2 concentrations (1.3-30 vol %) and gas compositions representative of power generation flue gases and industrial process gases. Turbec T100 series 3 micro Combined Heat and Power (CHP) gas turbine operating on mains natural gas was utilized for the generation of flue gas for testing. Fig. 3 illustrates the overall experimental setup used for the aging tests.
Two DX4000 Gasmet FTIR (Fourier transform infrared spectroscopy) gas analysers were utilised for gas monitoring in the experimental setup, both using Calmet v11.2 data analysis software. Gasmet FTIR gas analyser 1 was connected to the monitoring of flue gas at the inlet (feed) and outlet (retentate) of the membrane (sequentially); analyser 2 was permanently connected to slipstream gas outlet for permeate analysis. Sample lines were connected via a sample conditioning unit fitted with 10-µm heated filters set at 180 • C. Analyser 1 was equipped with a 5 L min − 1 gas sampling membrane pump. For analyser 2, the pump was disabled relaying on the sweep gas to carry the permeate gas to the sample conditioning unit and then the analyser. A 180 • C heated line interconnects the conditioning units and the FTIR analysers. The membranes were tested with nitrogen and 10% vol CO 2 feed compositions from cylinders. Initial filter and sampling pump were bypassed. Feed gas flow rates were tested from 5 to 10 L min − 1 with the pressure of 1.7 bar (absolute). Vacuum was used in the permeate side. NO x and SO x were introduced from calibration gas cylinders in nitrogen at the same point and in the same manner as for CO 2 .

Stage-cut analysis
For simulated membrane module performance analysis, the stage-cut and the membrane area requirement were calculated as: where V P is permeate and L F is the feed volumetric flows in Nm 3 h − 1 and A mem is the membrane area required for the process in m 2 .

Effect of feed pressure and vacuum grade
The membranes considered in this study were previously evaluated for lab-scale permeation performances. The average permeance of the lab-scale modules (effective membrane area: 10-16 cm 2 ) was around 800 GPU, while the CO 2 /N 2 separation factor was found to be around 30 for all modules measured using binary CO 2 /N 2 feed gas mixture at a feed pressure of 1.7 bar and at 35 • C and fully humid conditions [15]. With the pre-pilot modules, two flow configurations were studied, as shown in Fig. 4. Configuration A uses sweep gas flowing in the bore side while feed gas flows through the shell side of the module at high pressure in a counter-current direction. This configuration reflects lab-scale evaluation under low stage-cut values, assuming complete mixing in the feed and nearly constant driving force across the module. The use of sweep gas in CO 2 separation membranes has been reported beneficial in multistage processes [17][18][19]. As reported by MTR, USA, in these processes, air, used as sweep gas, is concentrated with CO 2 in the second or third stage, and the CO 2 -rich air is recirculated back to the combustion step [15]. The main advantage of using sweep gas in such cascades is the energy advantage for compression that is otherwise required to increase the driving force for permeation.
The use of sweep gas was studied in the HFTM module with the dry synthetic air at a constant flow rate of 3.2 LPM as the sweep gas. The CO 2 concentration in the permeate gas was ranged from 1.5 to 3 vol% and the feed pressure varied from 1.5 to 2.5 bar. Fig. 5A presents the effect of feed pressure on the CO 2 flux. As it can be seen, with increasing feed pressure from 1.5 to 1.7 bar, the CO 2 flux increased due to increasing partial pressure of CO 2 in the feed side. However, increasing the feed pressure further to up to 2.5 bar reflected no obvious changes in the flux but a trend to decrease the flux. Such behaviour is the characteristic of facilitated transport membranes due to the carrier saturation phenomenon [20]. The optimum upstream feed pressure was found to be 1.7 bar.
Given the limitations in increasing upstream feed pressure in order to create higher driving force, vacuum was used on the downstream side instead of the sweep gas to increase the transmembrane pressure difference. Unlike the use of sweep gas, using vacuum helps in concentrating CO 2 on the permeate side to high purities, making it particularly attractive for applications requiring storage or utilization of captured CO 2 . The use of vacuum (Fig. 4, configuration B) positively contributes to enhanced permeation driven by two factors: (i) increased driving force of CO 2 partial pressure between the upstream and downstream side, since the permeating gases are constantly removed by the pump and (ii) increased pressure ratio (ratio of total feed pressure to total permeate pressure, equation (2)). Pressure ratio, along with stage cut (θ) and selectivity, are the three main factors that determine the performance of a gas separation system using membranes [21]. Decreasing vacuum pressure on downstream side increases pressure ratio as seen in Fig. 5B, consequently increasing CO 2 flux across the membrane to up to 13.75 × 10 − 3 cm 3 (STP) cm − 2 s − 1 at the downstream pressure of 0.2 bar which is 1.8x the maximum achievable flux using sweep at the same feed pressure. This high sensitivity of CO 2 flux to the changes in pressure ratio is because the separation happens at the so-called "pressure-ratio-limited region", where the performance of the membrane is determined solely by the pressure ratio and is independent of membrane selectivity as described by Baker [21]. While the magnitude of pressure ratio can be a theoretical maximum of infinity with the use of a high-grade vacuum, the separation eventually departs from the "pressure-ratio-limited region" and enters a "membrane-selectivity-limited region", wherein the response to pressure ratio changes are levelled off at an already achieved maximum. In the current scenario, given the lab-scale membrane selectivity is around 30, a further increase in vacuum grade could result in higher fluxes across the membrane. However, it should be noted that the pressure ratio is also an important factor that adds up to the cost of the membrane separation system and hence a very high grade of vacuum would make the process more capital and energy intensive.

Effect of water activity and temperature on performance of HFTMs
Water plays an important role in facilitated transport membranes containing reactive CO 2 carriers (Fig. S2). The water content in flue gas depends on a variety of factors such as fuel type (mainly moisture content), combustion parameters and exhaust temperature [22]. In this study, the effect of water on transport properties of the pre-pilot modules was studied by using a make-up stream of simulated flue gas carrying additional water using an external evaporator (Fig. 1). The water content in the flue gas from the chimney averaged at ~10 vol% at 115 • C and 1.05 bar pressure. Two more water contents corresponding to 15 vol % and 20 vol% in the feed were studied using external evaporator operating at the same temperature and pressure conditions. These water contents corresponded to ~51% RH, 78% RH and 99% RH at 1.7 bar and 60 • C.
Beyond an activity of 0.99, condensation problems occurred in the gas lines due to proximity to saturation and fluctuating temperatures at non-insulated regions of the feed gas line. All modules showed an increasing CO 2 flux and purity with increasing water activity from 0.51 to 0.78 as seen in Fig. 6. The effect of mobile carriers was eminent in the scaled-up modules in accordance with the lab scale results. At lower water activity of 0.5, both 10% [Emim][OAc] and 20% ProK containing membranes exhibited a CO 2 flux about 25% and 35% more than that of the one without any mobile carrier. This effect is attributed to the additional CO 2 transport pathway aided by the formation of carbeneadduct in the case of [Emim][OAc] [23] and carbonate/bicarbonate species in the case of ProK in the presence of water [24]. Increasing water activity to 0.78 eventually increased CO 2 flux in all modules by 15-30% than the fluxes at 0.5. A notable increase was found in HFTM without mobile carriers as the transport is mainly steered by facilitated transport mechanism via fixed-site amino groups. Higher water content leads to better utilization of amine groups in the polymer matrix, as evidenced by several studies [4,25,26]. Further increase in water content to 0.99 only resulted in marginal drops in fluxes and purities, indicating local regions of condensation in the module as the conditions are close to saturation. Nevertheless, HFTM with 10% [Emim][OAc] was an exception, where the increased water activity may have caused higher and faster association/disassociation of CO 2 with IL, thus leading to enhanced performance close to the fully humid condition.
In order to compare the performance of the pre-pilot membrane modules with the lab-scale modules, the separation performance achieved at the water activity of 0.78 (~78% RH) was chosen. In bigger modules, the concentration of CO 2 and H 2 O in the feed rapidly decreases across the length of the module, and hence the driving force for permeation is no longer constant [8]. This limitation leads to the requirement of a realistic model for driving force to estimate the permeance of individual gases confidently.
To achieve the estimation of gas permeances in the pre-pilot modules, a single-stage membrane separation process mimicking the field test conditions was designed in Aspen HYSYS V9.0 integrated with an in-house built membrane model, ChemBrane. ChemBrane, with counter-current configuration, uses the 4th order Runge-Kutta method for calculating flux across the membrane module length followed by an iteration of permeance values to convergence [27]. The flow diagram described in Fig. 7 was developed in HYSYS, and the permeate and feed compositions obtained from the field tests at the water activity of 0.78 were input for simulation. Subsequently, the permeance values of gas components for each HFTM modules were calculated by manually adjusting and matching the relative ratios with ChemBrane, so that the simulated compositions of streams matched with the gas compositions measured in the exit permeate and retentate streams during the field tests for the same membrane area. The estimated permeance values for gas components in the HFTMs are summarized and compared with their respective lab-scale membrane modules in Table 3. Sizeable differences in permeation performances from the field test conditions and the lab conditions have been observed. Two humidity levels were chosen for lab-scale investigation, i.e., 100% RH (where the potential of facilitated transport is fully utilized) and 78% RH (close to the conditions of industrial testing). The permeances of all upscaled modules at field tests were higher than that of the lab-scale ones in both cases. This phenomenon can be attributed to three important factors of difference between the different test conditions: i) temperature of operation, ii) pressure ratio, and iii) water profile in the feed and the module.
Given that the operating temperature in the field tests was at 60 • C as opposed to 35 • C used in lab conditions, the permeability coefficients of the membrane materials change. In usual cases, with an increase in temperature, the solubility decreases, and the diffusivity increases. Hence, the total permeability of the gas component depends on the transport mechanism of membrane material if it is solution or diffusiondominated. Clearly, the permeances of all modules, especially the ones with mobile carriers, increased at a higher temperature. HFTM with 10% [Emim][OAc] recorded the highest increase of about 88%, followed by HFTM with 20% ProK at 76%. The HFTM without mobile carriers exhibited an increase of 45% with respect to CO 2 permeance. An increase in temperature increased the diffusivity of CO 2 , more profoundly  in membranes with mobile carriers, which confirms the potential of using mobile carriers as CO 2 reactive diffusion enhancers [15]. As expected, the apparent solubilities in these modules are reduced due to the transition towards a more diffusion-dominated mechanism as temperature increases. Evidently, membranes with [Emim][OAc] had the lowest apparent selectivity, which is in line with expectations due to the lower viscosity and reduced solubility of the IL in the membranes, as IL acts as a physical solvent at higher temperatures [28]. On the other hand, membranes with ProK suffered a smaller drop in selectivity due to relatively lower response in dynamic viscosity with respect to changing temperature, as seen in Table 4. Another important factor for increasing CO 2 permeance is the use of vacuum on the permeate side, which leads to larger fluxes at higher pressure ratio, as described earlier. Additionally, most facilitated transport membranes document higher performance with the use of real flue gas as opposed to simulated flue gas used in laboratory evaluation; it is believed to be primarily due to the constant water content in the large volumetric flow rates from the combustion units. The additional water in the feed is also attributed to cooling down the pre-humidified flue gas from 115 • C to 60 • C [30][31][32], which is the operating temperature for the pilot test in this work. Compared to the lab test (at 35 • C), the faster sorption and desorption of CO 2 with amines at higher temperature can also contribute to the enhanced permeation [20,33]. Moreover, the use of pre-pilot modules of higher packing density when compared to lab scale modules results in a more reliable water profile across the length of the module, ensuring more accessible water for membrane for CO 2 facilitated transport. Besides, the larger volumetric flows lead to higher gas velocities, which results in more turbulent flow and hence reduced concentration polarization inside the modules.

Long-term stability testing of HFTM
Long-term continuous operation of membranes is necessary for evaluation of membrane performance in the pilot scale for successful separation benchmarking [34,35]. Hereof, one case (the HFTM without mobile carriers) was chosen to study the performance durability for an extended period, while the HFTMs with mobile carriers were subject to impurity testing in an independent facility. Untreated flue gas without the addition of water through external evaporator was used for this purpose. The feed pressure was maintained at 1.7 bar throughout the study. Both CO 2 flux and purity on the permeate side were recorded during the test, which lasted for about two weeks, as seen in Fig. 8. The corresponding changes in flue gas compositions were also monitored to correlate instantaneous responses to the permeation properties. It can be seen that the CO 2 flux of the module dropped slightly from 13.01 × 10 − 3 cm 3 (STP) cm − 2 s − 1 and flattened out at ~ 10.5 × 10 − 3 cm 3 (STP) cm − 2 s − 1 after 10 days of permeation. On the other hand, the CO 2 purity remained constant at ~ 51 and increasing up to ~ 56 at the end of the test campaign. Clearly, the feed conditions affected the permeation properties differently. Since the facilitated transport membranes are more sensitive to the feed water content as discussed earlier, the highest water content in the feed on day 15 corresponded to the highest CO 2 purity in the module and recovery of the dropping CO 2 flux that was recorded until day 12. This point also corresponds to the lowest O 2 content in the feed gas stream. Although O 2 is considered as a contaminant in the flue gas for CO 2 separation, O 2 permeabilities are generally lower than CO 2 and higher than N 2 for all polymeric membranes [36]. While oxidative degradation is an important phenomenon in aminebased absorption systems [37], facilitated transport membranes have been reported to have shown no significant degradation with the presence of O 2 in the feed stream [31,32]. On the other hand, it should be noted that the feed flue gas was untreated except for upfront removal of particulate matter and the presence of SO x and NO x in trace levels did not significantly affect the membrane performance.

Effect of impurities on HFTMs with mobile carriers
Typical flue gas contains minor levels of impurities such as SO x , NO x , H 2 S, NH 3 , hydrocarbons and some heavy metals [36]. The SO x and NO x concentrations vary between 500 and 2000 ppm(v) and between 100 and 500 ppm(v), respectively, depending on fuel and combustion technology used [38]. The presence of SO x and NO x can cause irreversible changes in membranes, thereby causing a potential threat for the durability of membrane performance [39,40]. Mixtures of SO x /NO x and water generates sulfuric/nitric acid, which could eventually degrade the polymer chains. Conventional polymeric membrane systems typically employ flue gas desulphurisation unit and NO x control using catalytic reduction technology prior to the membrane separation unit.
For the study of the effect of acidic contaminants in the flue gas on the membrane performance, an accelerated aging test was carried out Table 4 Viscosity and CO 2 solubility changes of mobile carriers with temperature. Data obtained from [28,29].  with exposure of SO 2 and NO, at a composition of 630 ppm(v) and 380 ppm(v) in simulated flue gas, respectively. Only the HFTMs with mobile carriers were chosen for this test as the host polymer matrix in these membranes is similar to that of the HFTM and additionally contain small organic molecules (mobile carriers), which can be more prone to degradation. The compositions of SO 2 and NO were measured both at the feed and retentate side and the corresponding CO 2 purity in the permeate side of the membranes were monitored over time, as seen in Fig. 9. Both HFTMs containing [Emim][OAc] and ProK exhibited no changes in CO 2 purity on the permeate stream during these tests, indicating that SO 2 and NO had a negligible effect on the CO 2 transport. The effluent SO 2 and NO concentration achieved stable values in different time periods based on the composition of the membrane material. Both HFTMs concentrated SO 2 and NO on the permeate side, indicating high SO 2 and NO permeabilities in these membranes. Generally, SO 2 is expected to react with carriers and compete with CO 2 transport in facilitated transport membranes due to its high reactivity [41]. Additionally, due to their acidic nature, both SO 2 and NO change the pH of the system,  which in turn regulates permeation performance [42]. However, the results from the current study confirm the trend of facilitated transport membranes that documented stable performances after exposure to both SO x and NO x [31,43]. Nevertheless, it should be noted that the reactivity of both acidic gases with amines in the presence of water is highly dependent on temperature [44]. In the current study, the temperature of operation was maintained to 21 • C due to practical limitations, although the long-term studies document similar stability to SO x and NO x (Fig. 8), albeit at lower concentrations but extended operation time.

Stage-cut analysis
The separation capabilities of developed HFTMs when used in a single-stage membrane process were studied using a process simulation analysis with Aspen HYSYS V9.0. The similar flow diagram, as described in Fig. 5, was used for this purpose. The validated permeance and selectivity obtained through experimental data (Table 3) served as the basis for these simulations.
The effect of stage-cut on both CO 2 purity and recovery was studied by varying the membrane area in m 2 in the simulated process. As seen in Fig. 10, increasing stage-cut increases the recovery of the membrane while significantly affecting the purity of the processed stream. In this study, since the field testing which corresponded to the lowest stage-cut, the conditions of field test yielded the highest obtainable purity. The achievable CO 2 purity in the treated stream after a single-stage purification is in the range of membranes reported earlier [12,45,46]. Such high purity in a single-stage is beneficial to achieve an economically favourable CO 2 capture and purity. Nevertheless, the economic benefit of the fabricated membranes, especially with the ones containing mobile carriers, is clearly established in terms of the membrane area requirement for the same target recovery in Fig. 10B. It can be seen that the use of mobile carriers significantly reduced the membrane area requirement and hence can correlate to major savings in terms of membrane capital costs [47]. For instance, to achieve a CO 2 recovery target of 90%, the HFTMs containing mobile carriers required ~33% lower membrane area when compared to HFTMs. Similar benefits are obtained with purity CO 2 targets (Fig. 10B).
In order to benchmark the developed membranes with potential membranes for CO 2 separation, the performance of the current membranes as a single-stage separation unit has been compared with other hollow fiber CO 2 separation membranes with validation in pre-pilot scale or higher as seen in Table 5.

Conclusions
Hollow fiber modules with hybrid facilitated transport membranes were fabricated in the pre-pilot scale and tested in field conditions using real flue gas at a cement plant. Operation conditions such as feed pressure, temperature and pressure ratio were optimized to maximize separation using hybrid facilitated transport membrane. The HFTMs containing mobile carriers clearly showed improved performances than the neat HTFMs, indicating the advantages of CO 2 -philic additives in membranes for post-combustion capture. Higher water content in the feed stream led to better performances, but temperature fluctuations close to saturation might risk water condensation in the modules. Elevated temperature and reliable water profile in the real flue gas enhanced the permeation properties of the membrane modules when compared to the labscale tests. Long-term testing of HFTM in untreated flue gas revealed the good performance durability of the membrane for a period of two weeks. The HFTMs with mobile carriers were subjected to resistance testing for SO 2 and NO impurities in simulated flue gas, and no obvious changes in CO 2 permeate purity were obtained. Simulation studies reveal that, to achieve the first stage separation of a typical postcombustion capture process, use of the fabricated membranes leads to significant improvements in the separation performance compared to the currently available hollow fiber membranes tested at pre-pilot scale or higher, thus leading to decreased membrane area requirements. However, to further study the cost benefits of the HFTMs in a large scale process, extended exposure tests should be performed in order to estimate the membrane life time. Nevertheless, the developed highperformance membranes tested in the relevant environment strengthens the commercial potential of implementing membrane technology in CO 2 separation applications.

Declaration of Competing Interest
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.

Table 5
Summary of hollow fiber membrane module performances for flue gas treatment tested at pre-pilot scale or higher.