International Journal of Greenhouse Gas Control 10 (2012) 285–294 Contents lists available at SciVerse ScienceDirect International Journal of Greenhouse Gas Control journal homepage: www.elsevier.com/locate/ijggc CO2 sorption and desorption performance of thermally cycled hollow fiber sorbents Ryan P. Lively a,∗ , Ronald R. Chance a , Joshua A. Mysona a , Vinod P. Babu a , Harry W. Deckman b , Daniel P. Leta b , Hans Thomann b , William J. Koros a,1 a b School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, 778 Atlantic Dr. NW, Atlanta, GA 30332-0100, United States Corporate Strategic Research, ExxonMobil Research and Engineering, Annandale, NJ 08801, United States a r t i c l e i n f o Article history: Received 14 February 2012 Received in revised form 21 June 2012 Accepted 23 June 2012 Available online 24 July 2012 Keywords: Fiber sorbents CO2 capture Column chromatography Desorption Sorption a b s t r a c t We describe here the CO2 sorption and desorption properties of hollow fiber sorbents—a polymer/zeolite hybrid sorbent that possesses a coated bore that allows for exceedingly fast heat transfer between the heat transfer fluid and the sorbent. The fiber sorbents are intended for use in post-combustion CO2 . The fiber sorption properties are tested chromatographically with simulated flue gas in the absence of a heat transfer fluid and are found via in situ thermal measurements to be highly non-isothermal, which results in up to 40% losses in CO2 breakthrough capacities. The thermal front moving through the fiber wall was found to progress approximately 30% faster than the propagating speed of the CO2 sorption front. Upon the addition of a heat transfer fluid (water) in the bores of the fibers, breakthrough CO2 capacities were maintained at all flue gas superficial velocities studied (up to 50 cm/s). The propagation speed of the CO2 front was reduced by 38% by the addition of cooling water in the bores, and the in situ thermal measurements revealed that the fiber sorbents were nearly isothermal during the CO2 sorption step. One of the main conceptual advantages of a fiber sorbent CO2 capture platform is the ability to transfer the released sorption enthalpy to the bore-side cooling water, which can then be later used in a beneficial way. By varying the cooling water velocity, in situ thermal measurements showed that 22,000 J per mol of flowing CO2 could be transferred to the cooling water out of a possible 36,000 J per mol. Finally, plug flow-mode desorption experiments were performed, and a 40.5 mol% CO2 product was obtained. The small scales of the system prohibit sharper thermal fronts, which likely causes unwanted product CO2 and interstitial N2 mixing. © 2012 Elsevier Ltd. All rights reserved. 1. Introduction As increasing amounts of fossil fuels are consumed to supply the world’s explosive growth in energy demand, a corresponding rise in atmospheric CO2 concentration has been observed. This increase is invoked as a significant contributor to global climate change (HoltzEaken and Selden, 1995; Annual Energy Outlook, 2010). Prudence dictates that methods for controlling CO2 emissions be examined and developed. To curb this increase in atmospheric CO2 concentration, all CO2 emission-reducing strategies should be considered. Pacala et al. (2004) described this challenge in terms of seven CO2 emission “wedges,” where each wedge represents 50 gigatonnes of CO2 avoided over 25 years. The CO2 reduction wedges include strategies such as improving the fuel efficiency of vehicles in the fleet, constructing new nuclear power plants, increasing installed ∗ Corresponding author. Tel.: +1 404 385 4717; fax: +1 404 385 2683. E-mail address: ryan.lively@chbe.gatech.edu (R.P. Lively). 1 Principal Investigator. 1750-5836/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijggc.2012.06.019 renewable energy capacity and installing 90% CO2 capture on 90% of the coal-fired power plant infrastructure. In particular, the CO2 capture wedge has resulted in serious debate, as out of all of the possible CO2 emission reduction platforms, the capture of CO2 from power stations is the only reduction strategy that comes with an energy tax. Thus, to become a viable route for reducing atmospheric CO2 concentrations, CO2 capture techniques that minimize the amount of energy consumed must to be developed. 1.1. Fiber sorbent overview Lower energy costs for CO2 capture can potentially be achieved via a recently proposed method involving hollow fiber sorbents (Fig. 1) (Lively et al., 2009, 2010, 2011a). Fiber sorbents ideally can achieve the lower fundamental energy requirements of adsorption relative to absorption while mitigating many of the typical processing issues that prohibit adsorption processes from being widely adopted. This overarching objective is being pursued by creation of a porous polymer–sorbent composite in a hollow fiber morphology with an impermeable barrier on the inside (lumen side) of the fiber 286 R.P. Lively et al. / International Journal of Greenhouse Gas Control 10 (2012) 285–294 Fig. 1. Pictographic depiction of an ideal hollow fiber sorbent, demonstrating a sorbent supported within a highly porous matrix and a lumen-side barrier layer. (Fig. 1). The fiber sorbents can easily be modularized, similarly to hollow fiber membranes. The CO2 -laden power station flue gas is ideally flushed over the outside (shell side) of the fibers, whereupon the CO2 will sorb into the sorbents contained within the fiber walls. The shell-side gas pressure drop for flue gas flow will be significantly lower than a similarly sized packed bed. The thin fiber walls – coupled with an ideally impermeable lumen layer – allow for rapid heating and cooling cycles via hot and cold water flowing through the fiber bores. We have previously analyzed the energetic requirements for a fiber sorbent system (as well as other competing CO2 capture technologies) and shown the fiber sorbent system to be among the least energetically taxing systems (Lively et al., 2010). Other work has established the viability of spinning fiber sorbents with high sorbent loadings (75 wt% sorbent) (Lively et al., 2009), the post-treatment technique for creating an impermeable lumen side barrier layer (Lively et al., 2011b), and the reduction of external mass transfer resistances of fiber sorbents in a parallel flow module (Lively et al., 2011a). Fiber sorbents are spun using a wet phase inversion process, where a polymer solution is first mixed with zeolite 13X (a “proof of concept” sorbent filler we have used in these studies) and then extruded through an annular die along with a neutral bore-forming fluid (McKelvey et al., 1997). After spinning and solvent exchange, the fiber sorbents are finished by post-treatment to create a lumenside barrier layer using a polyvinylidene chloride (PVDC) latex. By washing PVDC latex through the bores of the fiber, a polymer layer is deposited on the lumen side of the fiber. This crucial step has been described and discussed in detail elsewhere (Lively et al., 2009, 2011b). The barrier layer is one of the enabling technologies of the fiber sorbent platform, and allows for heat transfer fluids to be moved through the bores of the fibers while flue gas is flushed over the outside of the fibers. The sorption enthalpy released by the CO2 –zeolite interaction can be captured by the bore-side flowing heat transfer fluid. If the heat transfer fluid is boiler feed water (in a coal-fired power station scenario), the sorption enthalpy transfer to the water can be used to reduce the amount of boiler water preheating that is required by the plant (Lively et al., 2010), thereby using much of the sorption enthalpy, that would otherwise add to the parasitic load on the power plant. This is a key advantage of the hollow fiber sorbent platform. Hot plant water is used bore-side for the CO2 desorption step. By carefully matching the heating front moving through the fibers with a shell-side N2 sweep, a near-pure CO2 product can be obtained (Fig. 2). The thin fiber walls allow for very rapid mass and heat equilibration times, thus enabling the possibility of quick thermal cycles. Fiber sorbents for post-combustion CO2 capture are intended for use in a rapid thermal swing adsorption (RTSA) cycle (Fig. 2), with the rapid cycles allowing minimum device volumes and more effective use of the sorbent. 1.2. Column chromatography using fiber sorbents Fiber sorbents for CO2 capture are based on an equilibrium separation of the shell-side flowing CO2 and N2 . The fibers are modularized such that the fiber bed resembles a column, and the CO2 /N2 separation occurs chromatographically. Here, the fiber sorbents are the stationary phase and the flue gas acts as the mobile phase. The difference in carbon dioxide’s and nitrogen’s Fig. 2. Overview of fiber sorbent RTSA operation. Four beds are assumed in this study, with each bed operating in phase with the others. During the sorption step (top left), CO2 -containing flue gas flows on the shell side of the fibers, while cooling water runs through the bores of the fibers carrying away the released heat of sorption. After the sorption step, the bed is heated using hot water while the bed gas feed is closed, which allows desorbing CO2 to create a local high pressure zone, thereby displacing downstream interstitial gas. Shortly after this, a N2 sweep is used to help push the desorbing CO2 out of the bed. In this way, a plug of high purity CO2 is captured. After the desorption step, the fiber sorbents are returned to the sorption temperature via cooling water while clean flue gas pushes interstitial CO2 out of the bed and blankets the bed in N2 . R.P. Lively et al. / International Journal of Greenhouse Gas Control 10 (2012) 285–294 287 Fig. 3. Representative plot of the response of an adsorption column to a step input. The spreading of the outlet curve is a result of dispersive forces as well as kinetic limitations within the adsorbent. The lag time in the outlet response is due to the adsorbates affinity for the adsorbent. affinity for the fibers results in a selective partitioning of the two gases, such that the CO2 moving through the bed is retained longer than the N2 . In lab-scale experiments, the effluent from the column is measured in real time, and typically a sweep gas is fed over the sample, followed by a tracer input. This tracer will move down the column at a speed that is dependent on the affinity of the absorbate for the adsorbent, and the output tracer signal will spread based on the rate of diffusion from the mobile gas phase to the stationary zeolite phase, as shown in Fig. 3. A pulse or a step tracer input may be used; however, for this work only a step input is considered. Ideally, the spreading of the curve can be used to analyze kinetic information about the fiber sorbents. However, non-idealities such as channeling and axial dispersion down the column often limit the accuracy of the approach unless great care is taken to minimize these factors. One way to account for the bed non-idealities is to use a mixed gas tracer, with one gas being strongly adsorbing and the other being non-adsorbing. The non-adsorbing tracer will progress through the bed at the mean retention time of the system, essentially capturing the “background” information in the column, which can then be easily factored out. By integrating the non-adsorbing tracer signal and the adsorbate signal with respect to time, the area bounded by the two resulting curves yields the equilibrium sorbed concentration (when the sorbent mass is accounted for), as demonstrated in Fig. 4. The equilibrium behavior of the sorbent not only affects the retention time of the adsorbate in the column, it also affects the shape of the adsorbate breakthrough front in the column. The velocity of the concentration front through the column is defined as, wc,i = v 1 + (1 − ε)/ε × (∂qi /∂ci )T (1) where v is the superficial gas velocity, ε is the bed void volume fraction available for gas flow, and dqi /dci is the slope of the isotherm where qi (mol sorbate/cm3 sorbent) is the concentration of the sorbate in the sorbent and ci (mol sorbate/cm3 total gas) is the sorbate gas concentration. As mentioned above, the sorption of CO2 into the zeolites is an exothermic process which releases a significant amount of heat. This heat will progress through the bed in a thermal front, similar to the concentration front discussed in the previous section. The thermal wave propagation velocity given by Farooq and Ruthven (1990), wT = v 1 + (1 − ε)/ε × (CP,s /CP,f ) − (1 − ε)/ε × (−Hs /CP,f ) × (∂qi /∂T )C (2) where CP,s is the heat capacity of the solids, CP,f is the heat capacity of the gas and H is the heat of sorption. For a Langmuir-type Fig. 4. Example plot of a typical adsorption run, where an equimolar mixture of non-adsorbing tracer and adsorbing component are flowed past a fiber column. The non-adsorbing tracer accounts for the bed “background,” or mean retention time, while the adsorbate signal is delayed due to affinity of the adsorbate for the sample in the bed. The area between the tracer signal and the adsorbate signal represents the total molar amount of adsorbate sorbed into the sample. isotherm, the q/T term (as an approximation) is often quite small, and can usually be neglected (Pan and Basmedjian, 1970). When comparing Eq. (2) to Eq. (1) it is clear that the thermal wave will lead the concentration wave when the heat capacity ratio is less than the concentration step ratio q/c (as an approximation), or the waves will align when the two ratios are similar, or the concentration wave will lead the thermal wave when the concentration step ratio q/c is greater than the heat capacity ratio. For zeolites, which sorb CO2 strongly at low pressures, and have heat capacities similar to that for CO2 /N2 (Chue et al., 1995; Felder and Rousseau, 1986), the thermal wave will almost always lead the concentration wave. As previously discussed, a key factor for using fiber sorbents efficiently is the ability to transfer the thermal wave effectively to the cooling water flowing in the bores of the fiber. The ability to match these two propagating waves will be discussed in this paper. This paper aims to be the culmination of the previous work by testing the sorption properties of the fiber sorbents in a simulated RTSA environment with sorption and desorption steps. The paper first reviews the equilibrium, kinetic, cyclic and chromatographic uptake of CO2 by bare (PVDC-free) fiber sorbents. These are expanded to chromatographic experiments on coated, actively cooled, fiber sorbents and finally a proof-of-concept heating cycle to desorb and purify CO2 . 2. Materials and methods 2.1. Materials Hollow fiber sorbents used in this work have been previously spun (Lively et al., 2009) using the wet-quench spinning process (McKelvey et al., 1997). The fiber sorbent is a matrix of cellulose acetate (CA) and zeolite 13X (1–3 micron particles, Sigma–Aldrich) with a dry zeolite loading of 75 wt%. The fibers are porous throughout, and are approximately 1100 microns in diameter, with a bore diameter of approximately 320 ␮m. Polyvinylidene chloride (PVDC) was chosen as the barrier layer polymer in this work. PVDC latex for application in the lumen layer was supplied by SolVin Chemicals (Northwich, UK), batch name XB 202. 288 R.P. Lively et al. / International Journal of Greenhouse Gas Control 10 (2012) 285–294 Fig. 5. Schematic of multi-component rapid thermal swing adsorption system. The system is partitioned into an upstream and downstream that are at different temperature (Tdownstream ≫ Tupstream ). The temperatures denoted are all measured continuously, as is the gas concentration. According to SolVin Chemicals, the PVDC latex is 55% by volume solids, the latex serum is anionic with a pH of 1.5. It has a glass transition temperature near −17 ◦ C (Ahareni, 1979). Finally, this batch of PVDC latex has copolymer components of polymethylmethacrylate (PMMA) to improve the temperature resistance of the polymer. The specifics of the fiber sorbent lumen layer creation have been published elsewhere (Lively et al., 2011b). Generally, the lumen layer allows the fibers to be used as “adsorbing heat exchangers,” and essentially prevents mass exchange while maintaining rapid heat exchange between the gas-side of the fiber and the water-side of the fiber. 2.2. Column chromatography sorption 2.2.1. Un-cooled fiber sorption A multicomponent competitive adsorption system was built to test the fiber sorbent modules in dry simulated flue gas feeds. A detailed schematic of this system is given in Fig. 5. Fiber sorbent modules were dried at 125 ◦ C for 2 days under 20.0 sccm of dry nitrogen and were then brought to the adsorption temperature (37 ◦ C). The fiber bed was then rapidly switched to a 10 mol% He/10 mol% CO2 /80 mol% N2 feed at varying flow rates between 20 sccm and 200 sccm. In this protocol, He acts as the inert tracer, N2 acts as the carrier gas, and CO2 is the adsorbate of interest. The fiber bed effluent composition was determined with a Pfeiffer Vacuum QMS 200 Omnistar Mass Spectrometer. After the sorption run was complete, the bed was heated at 125 ◦ C under 20.0 sccm of flowing N2 to prepare the sample for the next sorption run. Mass flow controllers were used to precisely hold the purge and feed gases at the flowrates required in this study (±0.1 sccm). The upstream of the fiber modules was held at a constant temperature of 37 ◦ C (chosen to enable large swing capacities of CO2 ) via heaters, while the fiber module was wrapped in heat tape and held at the desired experimental temperature. The immediate downstream of the module was held at 130 ◦ C to prevent any condensation of water on the walls of the tubing. This procedure was adopted, since in the absence of downstream heating, water vapor desorbing from zeolites tends to condense on the non-heated walls. All experiments were run in triplicate. 2.2.2. Cooled fiber sorbent experiments The multicomponent system described above was modified to allow hot or cold water to be pumped through the bores of the fibers. A 100 mL ISCO® pump attached to 1 m of 1/4′′ copper inch tubing wrapped in heat tape and controlled via heat controllers that supplied the cold or hot water to the fibers (the modification is shown in Fig. 5). Furthermore, the fiber modules were assembled using nylon PFA fittings from Swagelok® with a hypodermic needle thermocouple (Omega Instruments) nested in the middle of the fibers (Fig. 6), which minimized the thermal mass of the module and allowed for measurement of the thermal waves moving through the fiber bed, respectively. The sorption procedure described in the previous section was slightly modified by allowing the cooling water to begin running through the bores for 30 s before the sorption experiment was started. The effect of the water flow rate on the sorption capacity as well as well as the magnitude of the water and fiber thermal waves was studied. All experiments were run in triplicate. 2.2.3. Desorption experiments Desorption experiments were performed by first allowing the bed to come to CO2 saturation. The feed gas valve to the module was then closed as 95–105 ◦ C hot water was pushed through the fiber bores. The N2 sweep began as the observed thermal front passed R.P. Lively et al. / International Journal of Greenhouse Gas Control 10 (2012) 285–294 289 Fig. 6. Overview of module assembly for hollow fiber sorbents. Typically, six fibers are fit into approximately 18 cm of PFA tubing, and sealed into the two ends via epoxy. Three tees are attached to the tubing, two for gas inlet/outlet, and one for a hypo-needle thermocouple. This thermocouple is nested among the fibers, and epoxied into the tee. through the middle of the fiber module. The N2 flowrate was set to 100 sccm. This experiment was run in triplicate. 3. Results and discussion Pressure decay sorption experiments (Koros and Paul, 1976; Chandra, 2006) were previously performed (Lively et al., 2009) to elucidate the equilibrium uptake and kinetic response times of the fiber sorbents. Both equilibrium and diffusion experiments indicated that CO2 access to the 13X crystals is not hindered. The cyclic stability of the fiber sorbents was previously investigated by continuously cycling the fibers between 45 ◦ C and 100 ◦ C over the course of 48 h. No loss in fiber sorbent CO2 capacity was observed (Lively et al., 2009). Previous work investigated the kinetic resistance to CO2 sorption in modularized fiber sorbents in a flow system, and found that external mass transfer was the limiting resistance. These limitations were significantly reduced by increasing the fiber packing and increasing the superficial flue gas velocity flowing past the fibers (Lively et al., 2011a). to the small scales of the system studied here, much of that excursion was mitigated by the sensible heat of the system. As can be seen in Fig. 8, the overall intensity of the thermal wave decreases with decreasing flue gas flowrate, and the thermal wave itself becomes more diffuse. The sorption enthalpy is primarily carried away by both the gas flowing shell-side and conduction through the module walls. The rate at which the sorption enthalpy is delivered to the system is tied to the CO2 delivery rate. At lower superficial velocities, the fibers behave nearly isothermally due to adequate heat transfer to the module wall. At higher superficial velocities the fiber wall is unable to sufficiently accept the sorption enthalpy released, so that local heating dominates the CO2 sorption response due to the temperature-driven downward shift in the equilibrium sorption isotherm. 3.1.2. Effect of cooling water on sorption performance of hollow fiber sorbents One of the main motivations behind the use of hollow fiber sorbents is the ability to perform rapid thermal cycles which, in the absence of cooling, are quite non-isothermal due to the zeolite’s 3.1. Uptake of CO2 by bare hollow fiber sorbents in parallel flow module 3.1.1. CO2 sorption capacity in fiber sorbents, uncooled The CO2 uptakes of the “uncooled” fiber sorbent modules were measured 500 s after CO2 breakthrough. This time was chosen so that the bulk of the breakthrough curve could be observed (for inferring front velocities) while not allowing enough time for the bed to become saturated. This time frame was clearly system specific for our experiments, but can easily be defined for any system; industrial systems will define breakthrough at 1–10% CO2 leakage in the column eluent. These CO2 capacities decreased with increasing superficial flue gas velocity as seen in Fig. 7. Of course, if infinitely long times were used to analyze the CO2 uptake in the fiber sorbents, the equilibrium value would be obtained at all flue gas flow rates. In the absence of cooling water in the fiber bores, the sorption enthalpy will heat the zeolites within the fiber wall as seen in Fig. 8 which will in turn retard the CO2 adsorption kinetics until the heat has been dissipated. The data have been normalized by the highest temperature spike that was observed at 160 sccm; temperature excursions of 40 ◦ C are expected in actual operation, but due Fig. 7. Capacity of uncooled fiber sorbents (open squares) as a function of flue gas superficial velocity, and capacity of actively cooled fiber sorbents (closed diamonds) as a function of flue gas superficial velocity. Cooling water flow rate was set at 1500 mL/h. The dashed lines illustrate the likely capacities at increasing superficial velocities. 290 R.P. Lively et al. / International Journal of Greenhouse Gas Control 10 (2012) 285–294 Fig. 8. Thermal fronts measured at the middle of the six fiber module at three different flue gas flow rates. Bed length is 23.4 cm. The ordinate axis is the temperature signal normalized by the highest temperature observed at the highest flue gas flow rate. heat of sorption release. Without cooling water, the fiber sorbent bed will lose capacity at higher CO2 flowrates, as demonstrated above in Section 3.1.1. As established in previous work, a high CO2 velocity is desired not only for rapid cycles and bed size reduction (Lively et al., 2010) (due to the fact that each bed is processing more CO2 per minute), but also for minimization of the external boundary layer (Lively et al., 2011a). Cooling water running through the bores of the fibers allows for this advantage to be realized. A second, arguably greater, advantage of fiber sorbents is the ability to transfer the sorption enthalpy released to the cooling water flowing through the bores of the fibers for integrated heat management (Lively et al., 2010). In order for both of these goals to be achieved, an understanding of the effect of cooling water and cooling water flowrate on the sorption capacity must be developed as well as an understanding of the effect of the cooling water flowrate on the amount of sorption heat captured by the cooling water. To explore this effect, a sixfiber module with an impermeable PVDC layer (Lively et al., 2011b) was used and tested in a similar fashion to the experiments in the preceding section: deionized water at 37 ◦ C (the experimental temperature) was pumped through the bores of the fibers at 1500 mL per hour, the maximum flowrate that was attainable in the current RTSA system. In another set of experiments, the flowrate of the flue gas through the module was set at 200 sccm, the highest flue gas flowrate possible in the current setup, and the water flowrate in the bores was varied from 45 mL per hour to 1500 mL per hour. 3.1.3. Effect of cooling water on CO2 capacity With cooling water flowing through the bores, the thermal front data (Fig. 9) confirms that there is a drastic reduction in heat released when cooling water is used, lending credence to the heat release induced capacity loss in the absence of cooling described in Section 3.1.2. Fig. 7 illustrates the breakthrough capacity loss associated with an increasing rate of heat release (i.e., an increasing superficial velocity) for uncooled fiber sorbents, while the “expected” capacity is roughly maintained once cooling water is pumped through the bores of the fibers. Ideally, the difference in uncooled breakthrough capacities versus cooled breakthrough capacities should grow as the flue gas flow rate is increased, provided the cooling water flowrate is increased sufficiently to match the increased heat delivery rate. As Fig. 9 illustrates, the fibers are nearly isothermal relative to the un-cooled case, thus allowing for the CO2 capacity to be maintained at high flue gas velocities. Another possibility for the marked decrease in capacity with increased velocity is that the CO2 front velocity is fast enough Fig. 9. Thermal fronts in actively cooled fiber sorbent modules (cooling water flow rate = 1500 mL/h) as a function of flue gas flow rate shown in color. Gray thermal fronts correspond to uncooled fiber sorbents at varying flue gas flowrates. Thermal fronts in actively cooled fiber sorbents are denoted with arrows. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.) to cause internal mass transfer resistances to become an issue. However, this can easily discounted by the following estimate (Hines and Maddox, 1985): No external mass transfer resistance : t90 = 2 0.5 × lwall Deff 2 = 0.5 (400 × 10−4 cm) = 0.2 s 0.004 cm2 /s Severe external mass transfer resistance : t90 = 2 5.7 × lwall Deff (3) 2 = 5.7 (400 × 10−4 cm) = 2.3 s 0.004 cm2 /s These estimates show that even in the strongly internally mass transfer limited case, 90% sorption occurs in approximately 2 s whereas the sorption step is on the order of several minutes for these experiments. The diffusion coefficient was estimated as a serial combination of molecular diffusion and Knudsen diffusion (Hines and Maddox, 1985); the pore size was estimated from SEM microscopy. The above experiments show that cooling water in the bores is quite necessary to maintain sorption capacities at the high flue gas flowrates required for post-combustion CO2 capture. 3.1.4. Effect of cooling water on front velocity A key pair of parameters for designing a full scale RTSA system is the velocity of both the thermal front through the fiber bed and the concentration front through the fiber bed. These parameters can be estimated by dividing the bed length by the time required to reach the midpoint in either the CO2 front or the temperature front (the time for the temperature front must be multiplied by two, as the thermocouple is in the middle of the bed). Fig. 10 shows these front velocities in the fiber module as a function of the superficial flue gas velocity. In the case of fiber sorbents, the thermal wave should always propagate faster than the concentration front (see Section 1.2). Surprisingly, at low flue gas feed flowrates (low superficial velocities), the concentration front leads the thermal front, while at high feed flowrates, the thermal front propagates faster than concentration front. As discussed in Section 1.2, both the thermal and concentration front propagation velocities scale linearly with the feed gas superficial velocity. While the concentration front experimentally scales linearly with superficial velocity (as expected), the thermal front appears to have a discrete step in front velocity between 20 cm/s and 35 cm/s. The most likely cause of this is that the heat delivery rate (which, in theory, also scales linearly with the flue gas flowrate) is quite low for R.P. Lively et al. / International Journal of Greenhouse Gas Control 10 (2012) 285–294 291 Fig. 10. Concentration (open squares) and thermal (closed diamonds) front velocities as a function of flue gas superficial velocity. Fig. 12. Effect of cooling water velocity on the sorption heat capture by the cooling water. Parabolic fit is used as a guide for the eye. the slower superficial velocities. In the presence of significant heat leaks, the already weak thermal front at these low flowrates will be significantly diffused, causing the midpoint temperature rise to occur much later than it would in the absence of these heat leaks. At the higher feed flowrates, the thermal spike in the fibers is much more intense, and the midpoint is likely more accurately measured (though heat leaks are still significantly diffusing the thermal front). In the current configuration, the thermal front progress at a rate that is 2.6% of the superficial velocity, whereas the CO2 front progresses at a rate that is 2.0% of the superficial velocity (estimated from the high superficial velocity case). Even in these small, heat leak–limited systems, the thermal wave progresses 30% faster than the concentration front. Ideally, once the fibers are actively cooled with cooling water, the CO2 front velocity will be slowed due to the increase in capacity as a result of the fibers being more isothermal. Fig. 11 shows that the concentration fronts were indeed slowed by flowing cooling water through the bores; an average of a 38% reduction in front velocity was recorded. In the absence of severe thermal leaks, the reduction in front velocity via active cooling is expected to be even more substantial, as the uncooled fibers will become significantly hotter, thereby accelerating the front velocity. Clearly, relative to the uncooled case, the fibers are kept almost entirely isothermal by the cooling water flowing at 1500 mL per hour through the bores of the 6 fibers. The above results clearly show that cooling water is necessary to achieve one of the main advantages of fiber sorbents: retention of sorption capacity at high CO2 flowrates, thus allowing for rapid cycles and concomitant minimization of the CO2 capture system size. Fig. 11. Concentration front velocities for cooled (squares) fibers and uncooled fibers as a function of flue gas superficial velocity. Cooling water flow rate was set at 1500 mL/h. 3.1.5. Effect of cooling water velocity on heat of sorption capture The secondary motivation behind fiber sorbents is the ability to capture the heat of sorption released by the CO2 –zeolite interaction, as discussed in Section 1.1. To investigate this possibility, the flue gas flowrate was set at 200 sccm, as this was the most thermally intensive flowrate studied (and was the maximum flowrate attainable in the system) while the water flowrate through the bores was varied from 45 mL per hour to 1500 mL per hour. A needle thermocouple at the outlet of the module was used to measure the temperature rise in the cooling water, while a needle thermocouple placed at the inlet to the bores of the fibers measured the input water temperature. This allowed for observation of the temperature rise in the cooling water, and by integrating the outlet thermal wave with respect to time, the total amount of heat transferred to the cooling water could be estimated. There is an optimum water velocity for capturing the sorption heat via the water in the bores. As can be seen in Fig. 12, at the maximum heat captured, approximately 22,000 J of sorption heat per mole of CO2 captured has been transferred to the water, out of a maximum of 36,000 J per mole (the heat of sorption of CO2 on 13X). It is difficult to tell whether the remaining energy not captured is lost to the heat leaks present in the system, or is actually the upper limit that can be achieved due to heat transfer through the fiber sorbent walls. While it is quite difficult to quantify how the heat captured should scale with the cooling water velocity, the observations can be qualitatively explained as shown in the following. Fig. 13 shows the normalized outlet water temperature for three different water flowrates; the flowrates were chosen to represent high velocity/low heat capture (1500 mL/h), low velocity/low heat capture (100 mL/h), and high velocity/high heat capture (1100 mL/h). As can be seen in Fig. 13, the low water flowrate results in the largest temperature increase in the outlet water, and also the largest area of the three curves. As the water flowrate is increased, the temperature rise in the cooling water becomes less pronounced. However, as the cooling water flowrate is increased, the temperature front becomes increasingly sharper, which has important implications for efficient power-plant heat integration. Even though the slowest water flowrate has the largest temperature rise, only a small amount of water is heated, whereas at the higher flowrates, similar temperature rises (16% less and 34% less, respectively) are observed, but much larger quantities of water are 292 R.P. Lively et al. / International Journal of Greenhouse Gas Control 10 (2012) 285–294 Fig. 13. Normalized thermal fronts in bore-side cooling water measured at module outlet. Water flowrates and velocities are denoted in the figure. heated, hence the increase in sorption heat captured. One possible explanation for the above results centers around the fact that, at high water velocities, the thermal front propagation will lag significantly behind the water velocity (Fig. 14, Table 1). If rapid heat transfer is assumed (which is likely in the thin fiber walls), we hypothesize that the thermal front will transfer its heat to a plug of the rapidly moving cooling water. This plug will move well ahead of the thermal front in the fiber in the high water velocity scenario, creating a driving force for heat transfer from the water to the fiber wall (as the plug of hot water moves away from the thermal front in the fiber wall, the water plug is actually hotter than the fiber wall), causing the hot plug of water to become cool, as illustrated in Fig. 14 (top). Finally, once the thermal front nears the end of the fiber length, the hot plug of water will arrive at the ends of the fiber at the thermocouple, where the heat the water has absorbed can be detected. However, in the high water velocity scenario, there is less time for heat transfer between the length of the thermal front in the fiber wall and the flowing cooling water, resulting in lower water outlet temperatures. At the slowest water velocities, the water propagates through the bore of the fiber only marginally faster than the thermal propagation front in the fiber wall. This gives the plug of water much more contact time with the hot fiber wall, causing its temperature to rise considerably. However, since the water flowrate is quite slow, the total amount of sorption heat that is captured in the cooling water is reduced relative to the high flowrate scenarios. Fig. 14 (bottom) gives an overview of this hypothesis. These results indicate that an optimum exists for the cooling water flowrate; this optimum has been identified on the lab scale. While it is difficult to estimate exact values, this situation will likely extend to industrial scales as well: low water velocities will generate the hottest water, but also will generate less useful heat due to the small amounts of water used. On the other hand, extremely high water velocities will move much too fast to allow heating of the water as much as the low velocity case. To achieve the most useful energy from the heat of sorption release (a key factor disTable 1 Front velocities within hollow fiber sorbent module. Front Velocity (cm/s) Carrier gas Water, 100 mL/h Water, 1100 mL/h Water, 1500 mL/h CO2 Thermal, fiber 47.50 3.70 40.53 55.30 0.98 1.28 Fig. 14. Cartoon representing front propagation axially through the fiber in two extreme cases: rapid water velocity (top) and slow water velocity (bottom). cussed in Section 1), the water velocity must be carefully balanced between water temperature rise and total amount of water heated. 3.2. Desorption of CO2 from fiber sorbent modules using hot water 3.2.1. Challenges of recovering a pure product using TSA The previous section focused entirely on the sorption side of the RTSA cycle, with emphasis on investigating the optimum water flowrate to ensure isothermal operation and/or maximum heat of sorption capture. Of course, capturing the CO2 optimally does nothing for reducing CO2 emissions if the fibers are regenerated without producing a pure CO2 product. In this regard, there are serious difficulties associated with recovering a pure CO2 product via a TSA process. Typically, adsorption processes are split depending on whether trace or bulk contaminants need to be removed. For trace contaminants (<1 mol%), TSA is preferred; during regeneration (which occurs infrequently) the bed is simply heated with an inert purge, and the desorbing product is condensed and collected (Hines and Maddox, 1985). For bulk contaminants (5–25 mol%), PSA is preferred (Hines and Maddox, 1985; Yang, 1987). However, PSA has 293 R.P. Lively et al. / International Journal of Greenhouse Gas Control 10 (2012) 285–294 already been ruled out as a post-combustion CO2 capture technique due to the immense compression and/or vacuum requirements that go along with the enormous amount of flue gas (Lively et al., 2010). This mismatch in desorption mode—bulk contaminant (CO2 , 10 mol%) and thermal swing desorption—creates a challenging situation for recovering a pure CO2 product from fiber sorbents. The ideal case requires a sharp thermal step from 37 ◦ C to 100 ◦ C in the inlet water temperature, while the flue gas feed to the bed is shut off, and the vent to the bed is left open. If the sharp step is attained, the CO2 will rapidly desorb as the thermal front moves through the fiber bed; the desorbing CO2 will push the interstitial gas axially down the shell-side of the module. As the thermal wave passes the midpoint of the fibers, the upper half of the module will ideally be primarily CO2 . At this point, the module feed should ideally be opened to nitrogen (product gas from the sorption step) while hot water is still flowing through the bores. This sweep gas will serve to push a plug of CO2 out. After the plug has been collected, cooling water can be switched back on, cooling the fibers back down under nitrogen and preparing for the next sorption step. This desorption mode is labeled plug flow (PF) desorption. A much simpler desorption mode is continuous stirred tank (CST) desorption, where the bed is simply closed off and heated with hot water. In this mode, the desorbing CO2 mixes with the interstitial gas, thereby lowering product purity. 3.2.2. CO2 desorption from 6-fiber modules An experiment was performed to simulate the preferred PF desorption mode. In this experiment, the fiber bed was saturated with 10 mol% CO2 at 1 atm total pressure and 35 ◦ C. Next, the feed to the module was closed as the gas was switched to nitrogen, while simultaneously hot water at 100 ◦ C was pushed through the bores of the fiber. At the peak fiber temperature observed (measured at the axial and radial center of the bed), the feed valve was switched back on to flush nitrogen over the fibers and ideally push out a plug of CO2 . Fig. 15 (top) shows the results of the desorption experiment. Initially, before 100 s (in the “Equilibrium” period), the fiber bed is saturated with CO2 and no water is flowing. The inlet water temperature thermocouple is measuring stagnant water that is located between the 37 ◦ C module and the 100 ◦ C heat exchanger. After the equilibrium period, hot water begins to push through the bore (1500 mL per hour) during the heating period. Unfortunately, due to the small scale of the experiments, a sharp thermal front was not attained, despite many repeated attempts and system adjustments. While the inlet water certainly has a sharp thermal front (although it could be sharper—the front still spreads over 10 s), the fiber temperature front is severely dispersed. The main cause of this dispersion is that the fibers have contact both with the 100 ◦ C water in the bores and with the 38 ◦ C module wall. In essence, there is a large thermal heat sink in contact with every fiber in the module; the only way to avoid this is to construct modules that are much bigger in diameter. After a minute-long heating period, the fibers have risen to approximately 72 ◦ C. The outlet water temperature roughly matches the fiber temperature. Finally, the feed gas is turned back on to push the desorbed CO2 out of the bed in the sweep step resulting in an 18 mol% CO2 product. 3.2.3. CO2 desorption from 16-fiber modules To test the hypothesis that the small scales of the experiment were limiting the ability of the fibers to achieve a sharp thermal step – which in turn limits the ability of the system to form a near-pure CO2 product in PF mode – a 16 fiber module in 3/8′′ PFA housing was constructed and post-treated in the exact same fashion noted above. This module was subjected to the same desorption protocol as the 6-fiber module. If heat leaks through Fig. 15. Temperature and concentration profiles for two CO2 desorption experiments, (top) 1/4′′ , 6 fiber module, (bottom) 3/8′′ , 16 fiber module. Flue gas flow rate: 100 sccm. N2 flowrate: 100 sccm. the module wall are assumed to be responsible for the low CO2 purity in the earlier experiments, then it is fair to assume that decreasing the surface area for heat transfer through the module will result in a corresponding increase in CO2 purity. By this rationale, an increase in module ID from 1/4′′ to 3/8′′ should result in an increase in CO2 purity from 18 mol% to 40.5 mol%.2 As seen in Fig. 15 the observed CO2 purity (40.5 mol%) matches this hypothesis remarkably well. The same result was seen on a replicate experiment. These experiments support the PF-desorption mode as a route to achieve a near-pure CO2 product in a largescale RTSA operation where heat leaks through the module are minimal. 4. Conclusions This paper presented the sorption and desorption properties of the cellulose acetate/13X fiber sorbents in a simulated RTSA cycle. The fiber sorbents were assembled into parallel flow modules of two different sizes (1/4′′ and 3/8′′ ID), and installed into a custommade chromatographic system. The uncooled fibers breakthrough CO2 uptake was reduced at increasing superficial velocities. Thermal measurements revealed that the most likely cause of this loss of 2 If yCO2 ∝ As,wall then yCO2 ,large module = R2 i,large R2 i,small 0.405. × yCO2 ,small = (3/8 in)2 (1/4 in)2 × 0.18 = 294 R.P. Lively et al. / International Journal of Greenhouse Gas Control 10 (2012) 285–294 capacity was the increasing intensity of the thermal wave passing through the fiber module as superficial velocities were increased. The addition of flowing cooling water in the bores of the fibers allowed the fibers to maintain sorption capacities as expected, even at high gas velocities. The fibers were found to be almost completely isothermal upon addition of cooling water. Thermal front and concentration front velocities were measured and the thermal front was found to be 30% faster than the concentration front. In the absence of heat leaks, the thermal front is expected to lead by a larger margin. The cooling water velocity was varied to determine the optimum velocity required to capture the majority of the heat released during sorption. At the optimum water velocity for these experiments, 22,000 J per mol of CO2 flowing past the fibers were captured and transferred to the cooling water flowing through the bores; the maximum is 36,000 J per mol of CO2 . A PF-style desorption was performed using hot water and a CO2 product with a purity of 40.5 mol% was recovered. Currently, heat leaks through the fiber module wall are reducing the sharpness of the thermal front moving through the fibers, which likely corresponds to unwanted CO2 /N2 mixing. Considering the small scale of these experiments, the results are quite encouraging for the ability of the fiber sorbents to not only result in near-pure CO2 products, but also to capture a majority of the sorption enthalpy released during CO2 capture. Acknowledgement The authors thank ExxonMobil Corporation for funding this research. References Ahareni, S.M., 1979. On free volume of polymers above the glass transition. Journal of Applied Polymer Science 23 (1), 223–228. Annual Energy Outlook, 2010. Energy Information Administration. Report # DOE/EIA-0383, May 11, 2010. Chandra, P., 2006. Multi-component Transport of Gases and Vapors in Poly(Ethylene Terephthalate). Ph.D. Dissertation. Georgia Institute of Technology, GA. Chue, K.T., Kim, J.N., Yoo, Y.J., Cho, S.H., Yang, R.T., 1995. Comparison of activated carbon and Zeolite 13X for CO2 recovery from flue gas by pressure swing adsorption. Industrial & Engineering Chemistry Research 34 (2), 591–598. Farooq, S., Ruthven, D.M., 1990. Heat effects in adsorption column dynamics. Industrial & Engineering Chemistry Research 29, 1076–1084. Felder, R.M., Rousseau, R.W., 1986. Elementary Principles of Chemical Processes. John Wiley & Sons, Hoboken, NJ. Hines, A.L., Maddox, R.N., 1985. Mass Transfer: Fundamentals and Application. Prentice-Hall, New Jersey. Holtz-Eaken, D., Selden, T.M., 1995. Stoking the fires? CO2 emissions and economic growth. Journal of Public Economics 57 (1), 85–101. Koros, W.J., Paul, D.R., 1976. Design considerations for measurement of gas sorption in polymers by pressure decay. Journal of Polymer Science: Polymer Physics 14, 1903. Lively, R.P., Chance, R.R., Kelley, B.T., Deckman, H.W., Drese, J.H., Jones, C.W., Koros, W.J., 2009. Industrial & Engineering Chemistry Research 48 (15), 7314–7324. Lively, R.P., Chance, R.R., Koros, W.J., 2010. Enabling low-cost CO2 capture via heat integration. Industrial & Engineering Chemistry Research 49 (16), 7550–7562. Lively, R.P., Leta, D.P., DeRites, B.A., Chance, R.R., Koros, W.J., 2011a. Hollow fiber sorbents for CO2 capture: kinetic sorption performance. Chemical Engineering Journal 171 (3), 801–810. Lively, R.P., Mysona, J.A., Chance, R.R., Koros, W.J., 2011b. Formation of defect-free latex films on porous fiber supports. ACS Applied Materials & Interfaces 3 (9), 3568–3582. McKelvey, S.A., Clausi, D.T., Koros, W.J.A., 1997. guide to establishing hollow fiber macroscopic properties for membrane applications. Journal of Membrane Science 124 (2), 223–232. Pacala, Socolow, S., Stabilization Wedges, R., 2004. Solving the climate problem for the next 50 years with current technologies. Science 305, 968–972. Pan, C.Y., Basmedjian, D., 1970. An analysis of adiabatic sorption of single solutes in fixed beds: pure thermal wave formation & its practical implications. Chemical Engineering Science 25, 1653. Yang, R.T., 1987. Gas Separation by Adsorption Processes. Butterworth Publishers, Stoneham, MA.