Dendrimer Liquid Membranes: CO2 Separation from Gas Mixtures

John Chau , Gordana Obuskovic , Xingming Jie , Tripura Mulukutla , and Kamalesh K. Sirkar. Industrial & Engineering Chemistry Research 2013 52 (31), 1...
61 downloads 0 Views 108KB Size
Download PDF
2502

Ind. Eng. Chem. Res. 2001, 40, 2502-2511

Dendrimer Liquid Membranes: CO2 Separation from Gas Mixtures A. Sarma Kovvali† and K. K. Sirkar* Department of Chemical Engineering, Chemistry and Environmental Science, New Jersey Institute of Technology, Newark, New Jersey 07102

A generation 0 poly(amidoamine) (PAMAM) dendrimer in the immobilized liquid membrane (ILM) configuration was studied using flat films and hollow fibers for CO2-N2 separation. This dendrimer as a pure liquid functions as a CO2-selective molecular gate with highly humidified feed gas (Kovvali et al. J. Am. Chem. Soc. 2000, 122, 7594). The present work broadens the range of relative humidity of the feed gas stream (RHf) by adding a small amount of glycerol to the pure dendrimer liquid. A 75% dendrimer-25% glycerol ILM was found to increase the operating range of RHf substantially while maintaining the CO2 permeance and the selectivity (RCO2/N2) close to the levels observed with a pure dendrimer ILM. The performances of pure and 75% dendrimer ILMs were found to be superior or comparable to the highest reported RCO2/N2’s. This behavior is explained in terms of the charged environment in the dendrimer liquid membrane under humidified feed conditions and facilitated transport of CO2. 1. Introduction In a recent short paper,1 we had introduced the notion of a stable liquid membrane based on a pure liquid of generation 0 poly(amidoamine) (PAMAM) dendrimers for highly selective separation of CO2 from N2. We had also provided there a few preliminary results for the above system for the dendrimer liquid membrane immobilized in a microporous hydrophilized polymeric substrate. In this paper, we explore this topic in greater detail to develop a broader perspective on the utility of dendrimer liquid membranes in the immobilized liquid membrane (ILM) form for gas transport and separation. Liquid membranes for gas separation have been studied by a number of investigators over the years.2-8 Of the various forms of liquid membranes, namely, ILM, solvent-swollen polymeric membrane,9 contained liquid membrane (CLM),10 ion-exchange membrane,11-14 and polyelectrolyte membrane,15-18 ILMs are simple and attractive if they are stable. They have high species permeance and generally possess higher selectivities compared to conventional polymeric membranes. For example, for CO2-N2 or CO2-CH4 separation, commercialized polymeric membranes generally have selectivities in the range of 15-40; liquid membranes, on the other hand, can have much higher selectivities, as much as 4000.19 Liquid membranes, however, are generally subject to problems of stability. For aqueous solution-based membranes, the membrane is unstable unless both feed and sweep streams are humidified (except in CLM). Further, the facilitating agents incorporated in the liquid membrane are often volatile and therefore are lost with time. This is true even if the substrate is an ion-exchange membrane.20 Chen et al.21-23 have therefore developed ILMs for CO2-N2 separation using glycerol as the solvent and nonvolatile species, such as sodium carbonate and sodium glycinate, as carriers. Glycerol is essentially a nonvolatile liquid; as a result, the membranes were shown to possess remarkable stability in environments † Current address: Compact Membrane Systems, Wilmington, DE. * To whom correspondence should be addressed. Tel.: (973) 596-8447. Fax: (973) 642-4854. E-mail: [email protected].

Figure 1. Schematic of the structure of the EDA core PAMAM dendrimer of generation 0.

having considerable humidity variations. The hypothesis behind the recent paper by Kovvali et al.1 is as follows: is it possible to replace the solvent altogether and use a pure liquid as the liquid membrane medium as well as the facilitating agent for facilitated CO2 transport over N2? This was prompted by the fact that pure glycerol has an inherently low selectivity for CO2N2, namely, 1.5. Amine functional groups in a form where the compounds are not volatile could lead to high carbon dioxide permeances and selectivity with improved stability. One such class of carriers is dendrimers. Dendrimers are a novel class of polymers having some unusual properties. A major class of dendrimers which has been studied in a wide range of applications is the PAMAM dendrimers. PAMAM dendrimers of generation 0 having an ethylenediamine (EDA) core were therefore chosen as the liquid to be immobilized in the pores of hydrophilized microporous polymeric membranes for selective CO2 separation. This work was focused on a generation 0 dendrimer instead of higher generations because of its suitable characteristics for the application at hand. As shown in Figure 1, the generation 0 dendrimer of molecular weight 518 has four primary amine groups and two tertiary amine groups; the molar concentration of primary amines is 9.1 M, and that for tertiary amines is 4.6 M. It is supposedly nonvolatile and is available

10.1021/ie0010520 CCC: $20.00 © 2001 American Chemical Society Published on Web 04/26/2001

Ind. Eng. Chem. Res., Vol. 40, No. 11, 2001 2503

in a 20% methanol solution, from which methanol could be removed by vacuum to yield a pure dendrimer liquid. PAMAM dendrimers are considered nontoxic up to generation 5.24 The PAMAM dendrimer of generation 0 is hygroscopic but has a finite solubility in water. Similarly, glycerol is strongly hygroscopic but completely miscible with water. When dendrimer-based ILMs are exposed to humidified feed gas, they pick up a significant amount of water from the feed. Some of the water will be transported across the membrane because the sweep gas used in our experiments was always dry helium. The water content in the liquid membrane is likely to have a significant influence on the performance of the dendrimer-based ILMs. Kovvali et al.1 employed experimental conditions in which the feed gas had 100% relative humidity (RH), whereas the sweep gas stream was dry. These conditions were similar to those employed by Chen et al.21-23 It is important to know the effect of feed gas humidity variation on the dendrimer membrane performance because it has been pointed out1 that when the feed gas is completely dry, the pure dendrimer membrane has a CO2-N2 selectivity of only 5 and a very high N2 permeance. However, addition of a significant amount of glycerol to the dendrimer liquid may potentially decrease the N2 permeance under dry feed gas conditions. Because glycerol-based ILMs containing nonvolatile carriers such as glycine-Na and Na2CO3 were found to be quite effective even when the feed gas had rather low RHs,21-23 we have studied specifically ILMs containing small amounts of glycerol in the dendrimer generation 0 liquid. The objective of the addition of glycerol to the pure dendrimer liquid was to see if the operating RH range of the feed gas stream can be broadened substantially without compromising the high selectivities associated with pure dendrimer liquids. The sorption of moisture in the dendrimer liquid is obviously important for the CO2 facilitation reactions. However, the data on water uptake by pure dendrimers is not available in the literature. Chen et al.21 reported preliminary data on water uptake by glycerol at different ambient RHs. Therefore, the extent of sorption of moisture in the pure dendrimer liquid has also been determined at various known RHs and compared with that in glycerol. An important observation in Kovvali et al.1 was that the permeability of N2 through the PAMAM generation 0 dendrimer liquid membrane was extraordinarily low when the feed gas was humidified. This was ascribed to the very highly charged environment in the presence of moisture due to a very high concentration of reactive functional groups in the pure dendrimer liquid. Whether such a behavior is retained in the presence of small amounts of glycerol in the ILM is also of interest. We have also briefly studied the CO2O2 system to the same end. Because hollow fibers are often the substrate of choice for gas separation, we have explored here pure dendrimer liquid membranes in two types of hollow fiber substrates. 2. Reaction Mechanisms for Carbon Dioxide Facilitation Reactions with Amines In general, amines as carriers perform better than carbonate/bicarbonate solutions for carbon dioxide separation. Among various amines, sterically hindered ones have higher capacities for CO2 than unhindered primary and secondary amines.25

In systems where amines are used as carriers, the reactions can occur with primary, secondary, or tertiary amines. A primary amine reacts with carbon dioxide according to

CO2 + 2RNH2 S RHNCOO- + RNH3+

(1)

The overall reaction between CO2 and a secondary amine is

CO2 + 2R2NH S R2NCOO- + R2NH2+

(2)

Although the tertiary amine is not supposed to react directly with CO2 like a primary or a secondary amine because they lack the proton needed in the deprotonation step, tertiary amines show considerable reactivity toward CO2; further, water is essential for this reaction. The reaction of CO2 with tertiary amines can be described satisfactorily with the base-catalysis reaction mechanism26,27

CO2 + R3N + H2O S HCO3- + R3NH+

(3)

3. Experimental Details Dendrimer liquids used in the current work are hydrophilic. Hydrophilic or hydrophilized membrane substrates were therefore used in the present study to immobilize the liquid membranes. The dendrimer liquid does not wet hydrophobic substrates spontaneously, but it does wet hydrophilic substrates without any special effort. All of the ILMs formed completely filled the pores of the substrate membrane, resulting in the ILM thickness being equivalent to that of the substrate. Both flat and hollow fiber membranes were employed. The flat membranes used were hydrophilized poly(vinylidene fluoride) (PVDF) membranes (Millipore, Bedford, MA) and hydrophilized Celgard 2500 membranes (Celgard LLC, Charlotte, NC). The hydrophilized Celgard 2500 membranes prepared from hydrophobic polypropylene microporous substrates are not yet commercially available. The developmental hydrophilized Celgard membranes studied here were prepared by an approach quite different from the surfactant-based approach used for the commercially available hydrophilized Celgard 2500 films. Two types of hollow fiber membranes were also used: hydrophilized poly(acrylonitrile) (PAN; Sepracor, Marlborough, MA) and polysulfone (PS; Minntech, Minneapolis, MN). It should be pointed out that the PAN fibers used are of the ultrafiltration type whereas the PS fibers are of the microfiltration type. Hollow fiber modules were fabricated by placing the fibers in a transparent polyethylene shell casing and separating the tube side and shell side with appropriate potting. The details of the substrate membranes and membrane modules fabricated are provided in Table 1. Starburst PAMAM dendrimers of generation 0 in methanol were obtained from Sigma (St. Louis, MO). Starburst PAMAM dendrimers of generation 0 are henceforth identified as “dendrimers”. Methanol from the dendrimer solution was removed by subjecting the solution to vacuum for several hours to several days; the resulting pure dendrimer solutions were stored in a desiccator until used. Dendrimer solutions of 44% (1M) and 75% in glycerol were also prepared for particular experiments. All mixtures of CO2 and N2 containing between 0.5% and 25% CO2 were obtained from Matheson (East Rutherford, NJ). Higher CO2 concentrations

2504

Ind. Eng. Chem. Res., Vol. 40, No. 11, 2001

Table 1. Details of Substrates and Modules Used membrane supplier

material

porosity

tortuosity

Millipore Celgard

Flat Membrane Substrates Durapore (hydrophilized PVDF) 0.7a 2.58b hydrophilized Celgard 2500 0.45c 2.54c

Sepracor Minntech

poly(acrylonitrile) (PAN) polysulfone (PS)

thickness, µm

pore diameter, µm

100a 25c

Hollow Fiber Substrates N/Ad N/A 0.3-0.4 N/A

0.1a 0.08c

50e 40f

70000 MWCO 0.1

a From the manufacturer’s catalog (Millipore). b From ref 21. c From ref 28. d N/A: not available. e Fiber i.d. 200 µm and o.d. 300 µm; logarithmic total membrane area of 116.4 cm2. f Fiber i.d. 280 µm and o.d. 360 µm; logarithmic total membrane area of 35.96 cm2.

in the feed gas mixture were obtained by mixing pure CO2 and N2 gas streams using Matheson electronic mass flow controllers (model 8272-0412) in required ratios. 3.1. Preparation of ILMs. The ILMs were prepared by the immersion technique.21 ILMs prepared by this technique normally have thicknesses comparable to those of the substrates. After immersing in the dendrimer solution (either pure or in a solvent), the substrate was removed from the liquid, and the excess liquid from the surfaces was carefully wiped with kimwipes. In some instances, any moisture absorbed by the liquid was removed by applying a vacuum during the wetting process. The procedure was similar if a 44% or 75% solution of dendrimer in glycerol was used as the liquid. Pure dendrimer ILMs were immobilized in the hollow fibers by filling the entire void fraction of the fibers. The lumen side of the fibers was filled with the 20% solution of dendrimer in methanol for a minute with the other end closed. Nitrogen was passed at a low flow rate on the shell side to evaporate the methanol. The procedure was repeated from the other tube end. In the case of polysulfone fibers, because of their pore size and porosity, methanol was seen coming out on the shell side. No attempt was made to evaporate methanol from the shell side during immobilization because it could hinder the ILM preparation. After the ILM was immobilized in the pores, nitrogen at high flow rates (∼50 cm3/min) was passed from the tube side for a few minutes to clear the lumen side of any possible blockage. This was verified by making sure all fibers were open. To remove any remaining methanol from the pores, a house vacuum (∼45.7 cmHg) was applied from the shell side with a small flow of nitrogen from the tube side overnight. The immobilization procedure was repeated for both sets of fibers to produce leak-free hollow fiber modules. Immobilization using the pure dendrimer liquid is potentially much simpler. 3.2. Measurement of the Equilibrium Water Concentration in Dendrimer and Glycerol. Because of its hygroscopic nature, fresh dendrimer liquid absorbs water from ambient air when exposed to it. If the contact time between the dendrimer liquid and ambient air is sufficient, it can be assumed that an equilibrium distribution of water has been reached between the dendrimer and ambient air. The change of weight of the dendrimer + water under different RHs was recorded using an electronic balance (model PB 303, Mettler-Toledo). The humidity in the balance environment was monitored by a humidity probe (model HMP 32UT, Vaisala, Woburn, MA). The measurements showed that this technique can give reasonably reproducible results. For comparison, the water uptake by glycerol was also measured. These measurements were done independently of the data reported earlier21 and covered a wider range of RH conditions.

3.3. Experimental Setup and Procedure. The experimental setup and procedure were similar to those used in our earlier studies.1,21-23 The flat membrane of approximately 10 cm2 area was placed between halfcells and was supported by a porous stainless steel screen (Pall Trinity Micro, Cortland, NY) in the bottom cell well. The gas space on the top of the membrane was sealed by a Viton O-ring on the top half of the cell. The CO2/N2 permeation performance through the ILMs was measured by the sweep gas technique.4 The experimental setup for the hollow fiber modules was the same as the one used for flat membranes. The feed for all hollow fiber ILMs was introduced through the lumen side of the membrane with the sweep gas flowing on the shell side. Experiments with the hollow fiber ILMs were performed in both cocurrent and countercurrent modes. The humidities of the gas streams were measured by humidity probes. In all experiments, the sweep gas was always dry helium. All experiments were conducted at room temperature, 23 ( 2 °C. 3.4. Calculation Procedure for Gas Permeances and the Separation Factor. On the basis of the calibration in the gas chromatograph (GC) and the peak areas obtained, the mole fractions of the species in the sweep gas stream can be obtained. When these mole fractions are multiplied by the sweep side gas volumetric flow rate, one obtains the volumetric permeation rate of the species through the membrane. The effective permeance of a species through the membrane was calculated by

() Qi tm

)

eff

Vi At∆pi

(4)

where Vi is the volumetric permeation rate of species i, ∆pi is the partial pressure difference of species i across the membrane, and At is the total membrane area. The true permeance of the species (Qi/tm)true is obtained from

() () Qi tm

)

true

Qi tm

τm eff m

(5)

The permeability Qi of a particular gas species through the ILM is obtained by multiplying its true permeance through the ILM and the membrane thickness:

Qi )

() Qi tm

tm

(6)

true

The separation factor of species i with respect to species j, Ri/j, is defined by

Ri/j )

Qi (Qi/tm)eff (Qi/tm)true ) ) Qj (Qj/tm)eff (Qj/tm)true

(7)

Ind. Eng. Chem. Res., Vol. 40, No. 11, 2001 2505

In the case of hollow fiber modules, the partial pressure differences of the species and the membrane area for transport are based on logarithmic mean averages. The logarithmic values of the total membrane area and partial pressure difference are calculated according to

Alm ) ∆pi,lm )

Ain - Aout ln(Ain/Aout) ∆pi,in - ∆pi,out

ln(∆pi,in/∆pi,out)

(8)

(9)

Here Ain and Aout are the total membrane areas based on the inside and outside diameters of the fiber; ∆pi,in and ∆pi,out refer to the partial pressure difference of species i at the inlet and outlet of the module, respectively, for the calculation of the logarithmic partial pressure difference. The subsequent reported values of permeances are denoted with subscript lm to highlight this aspect of the calculation. Gas permeances are reported in units of cm3(STP)/ cm2‚s‚cmHg. Wherever convenient, gas permeances are reported in units of GPU, which is equivalent to 10-6 cm3(STP)/cm2‚s‚cmHg. Gas permeabilities are reported in units of barrer, which is 10-10 cm3(STP)‚cm/cm2‚s‚ cmHg. 4. Results and Discussion Dendrimer-based liquid membranes are unique in terms of their preparation and behavior. The generation 0 PAMAM dendrimer was often used in its pure form without any additional solvent as a facilitated transport liquid membrane. For convenience, whenever a generation 0 dendrimer is used in its pure form without any solvent as the ILM, the ILM is referred to as a pure dendrimer ILM. 4.1. Effect of Glycerol on the Performance of Dendrimer-Based Liquid Membranes. Kovvali et al.1 observed that pure dendrimer liquid membranes tend to lose their ability to function as CO2-selective molecular gate vis-a`-vis N2 when the feed side inlet RH was reduced. However, when a significant amount of glycerol was added to dendrimer, the liquid membrane tended to be stable and essentially leak-free even under dry feed gas conditions. They postulated that glycerol present in the intramolecular and intermolecular spaces prevented the leakage of N2 in the absence of water in the gas streams. However, the addition of glycerol has other implications on the ILM performance. They observed that glycerol in a 44% (1 M) dendrimer membrane did not have much contribution to CO2 permeation. However, the selectivity of the ILM for CO2 over N2 decreased drastically from over 18000 to less than 5000. This was expected. A nonselective, nonfacilitating liquid like glycerol will reduce the highly charged ionic environment that a pure dendrimer liquid provides, resulting in much higher N2 transport. Is it, therefore, possible that a minimum amount of glycerol added to the pure dendrimer liquid could retain the molecular gating behavior of pure dendrimers while increasing its effective operating range of feed RH? This idea was explored by adding 25% glycerol to the dendrimer liquid. Figure 2 illustrates the results for an ILM containing 75% generation 0 dendrimer in a glycerol solution in

Figure 2. Variation of permeance of carbon dioxide and separation factor with partial pressure difference of carbon dioxide in hydrophilic PVDF substrates for different dendrimer concentrations in the ILM.

the hydrophilized PVDF substrate exposed to a feed gas having 100% RH. For comparison, the behaviors of pure dendrimer and 44% dendrimer in glycerol ILMs1 are also plotted. The substrate membranes were soaked in the dendrimer solutions for 8 h, and a vacuum was applied for about 5 h to remove any absorbed moisture. The removal of absorbed moisture proved to be important for successful operation of the ILM under dry or partially humid feed conditions. The ILM formed with a 44% (1 M) dendrimer solution in glycerol without removing any absorbed moisture tended to perform poorly under dry feed gas conditions with a high N2 permeability of 1260 barrer. However, the N2 permeability dropped significantly even under dry feed gas conditions to about 8 barrer when the ILM was formed by vacuuming the solution for about 5 h. The highest CO2 permeability observed was about 5400 barrer for a 0.5% CO2 feed gas. The sorbed water by glycerol, when left unremoved, would be essentially a third component in the ILM. Under dry feed gas conditions, this water would presumably evaporate, providing channels for excessive N2 transport. Hence, removal of sorbed water is considered critical for preventing excess N2 transport. However, once feed-side humidification was started, N2 permeability dropped to about 1 barrer because of the salting out effect and creation of a highly ionic environment in the ILM. Because of humidification of the feed side, CO2 facilitation occurs and CO2 permeability increased drastically. The 75% dendrimer ILM achieved a separation performance much closer to the pure dendrimer ILM than 44% dendrimer ILM. The highest selectivity obtained for a 0.76% CO2-N2 gas mixture (∆pCO2 of 0.82 cmHg) was about 16 300 (compare 18000 for a pure dendrimer at a lower ∆pCO2 of 0.27 cmHg). The corresponding CO2 and N2 permeabilities were 3200 and 0.19 barrer,

2506

Ind. Eng. Chem. Res., Vol. 40, No. 11, 2001

Table 2. Performance of Dendrimer Membranesa at Low CO2 Concentrations (Feed CO2 Concentration: ∼0.5%) carrier

CO2 effective permeance, GPU

CO2 permeability, barrer

RCO2/N2

ref

75% dendrimer-balance glycerol PAMAM pure dendrimer, generation 0 44% dendrimer-balance glycerol 2.25 M glycine sodium salt-glycerol 1 M Na2CO3-glycerol plasma-grafted AA with EDA

8.58 9.8 10.2 27.5 8.4 100

3200 3600 3800 10100 3100 ∼4000

16300 ∼18000 3000 g4000 1400 4000

current work 1 1 22 21 14

a

Substrate: hydrophilized PVDF.

respectively. The increase in nitrogen permeability from an average value of 0.17 barrer for pure dendrimer ILM to 0.19 barrer is not very significant. This indicates that the addition of small amounts of glycerol did not affect the CO2-selective molecular-gating behavior of PAMAM dendrimers. It is conceivable that glycerol present in small amounts (25%) in the ILM was able to occupy intermolecular and intramolecular spaces of dendrimer. However, the amount of glycerol added (25%) was not in excess to disrupt the highly charged environment that was created by the pure dendrimer liquid. This essentially ensures that the N2-blocking ability of the pure dendrimer liquid is retained. Note that the 44% dendrimer liquid has a QN2 of 0.94 barrer (pure glycerol QN2 is 2.85 barrer, while pure dendrimer QN2 is 0.17 barrer). It should be mentioned that nitrogen was detected by the GC in only a few experiments for 75% dendrimer ILM even though three test cells were used in series for the experiments. 4.2. Comparison of the Performance of Dendrimer Membranes. It would be useful to compare the performances of the dendrimer-based ILMs in PVDF substrates with some of the promising membranes in the literature at both low and high carbon dioxide partial pressures. The carbon dioxide permeances for pure dendrimer ILMs were lower compared to those obtained by glycine sodium salt in glycerol ILMs22 in PVDF substrates. The selectivities, however, were about 4-5 times those obtained for glycine sodium saltglycerol ILM. This could be due to a lower mobility of dendrimer-carbon dioxide complexes. The molecular weight of dendrimer of generation 0 is 518, whereas the molecular weights of glycine salt and glycerol are about 100. Hence, it can be expected that the diffusivities of dendrimer and the dendrimer-CO2 complex could be much less than those in the glycine salt-glycerol system. Another major reason for reduced carbon dioxide permeances in pure dendrimer ILMs is due to reduced solubilities of all gas species due to the salting out effect. There are no direct data available for solubilities of carbon dioxide and other gases in pure dendrimer solutions of different generations. However, it can be inferred from aqueous amine solutions that increased amine concentrations lower carbon dioxide loading per mole of amine at the same carbon dioxide partial pressure. For example, at a carbon dioxide partial pressure of 1 kPa (0.75 cmHg), a 0.5 M triethanolamine (TEA; corresponding to 7.37 wt %) aqueous solution can be loaded up to about 0.8 mol of CO2/mol of TEA. As the TEA concentration increases, the CO2 loading decreases. At 5 M TEA concentration (corresponding to 67 wt %), the CO2 loading is reduced to about 0.1 mol of CO2/mol of TEA.29,30 Table 2 compares the performances of dendrimerbased membranes at a low CO2 feed gas concentration of around 0.5%. At such low carbon dioxide partial pressures, the pure dendrimer membranes are highly

selective for CO2 over N2. The pure dendrimer membranes offer about the same effective permeance of CO2 as 1 M sodium carbonate-glycerol ILM but are about 13 times more selective for CO2. Similarly, glycine sodium salt-glycerol membranes offer the tradeoff of higher effective permeance for CO2 and about one-fourth of the pure dendrimer membrane selectivity for carbon dioxide. The presence of a small amount of glycerol in the dendrimer (25%) liquid does not affect either the permeances or RCO2/N2 very much. However, a further increase in the glycerol content in the ILM reduces the selectivity of the ILM for carbon dioxide without offering any significant advantages in terms of carbon dioxide permeance. The plasma-grafted thin membrane soaked with EDA19 offers by far the highest effective permeance for carbon dioxide of 100 GPUs and a selectivity that is comparable to or somewhat less than that of glycine sodium salt-glycerol ILM. It should be stressed that these thin membranes were tested under humidified conditions on both the feed and sweep sides of the membranes. There was no study performed on the loss of EDA with time. EDA is a volatile and toxic amine, and it could lead to process stream contamination. However, dendrimer membranes have permeabilities for carbon dioxide similar to those of the above membrane at these CO2 concentrations. This confirms that the dendrimer membranes facilitate carbon dioxide on par with other facilitated transport membranes, whereas the high selectivities are mainly due to the high density of charged functional groups present under humidified feed conditions. At higher partial pressures of carbon dioxide, the selectivity of dendrimer membranes is comparable to the highest levels of selectivity reported in the literature by polyelectrolyte salt membranes.17 Table 3 summarizes the comparison of dendrimer-based ILMs with various membranes studied in the literature at higher feed CO2 concentrations. The selectivity and permeances offered by the polyelectrolyte membranes were achievable within a narrow range of the feed and sweep RHs. It should also be mentioned that making a pinhole-free membrane from these salts is difficult.18 A poly(ethylene oxide)-containing and crosslinked polymer film31 could provide from a dry gas a CO2 permeability as high as 250 barrer and a RCO2/N2 of 60 based, however, on pure gas permeabilities. Poly(ethyleneimine)/poly(vinyl alcohol) blend membranes20 provide about 5 times higher permeance than dendrimer membranes, but the selectivity of the blend membranes for CO2 was an order of magnitude lower than that of dendrimer membranes. This indicates that the dendrimer membranes are in the right direction for commercialization of membrane processes for CO2 removal from flue gases, because Haraya et al.32 indicated that a RCO2/N2 of more than 210 is required to recover CO2 efficiently by a single-stage permeator. On the other

Ind. Eng. Chem. Res., Vol. 40, No. 11, 2001 2507 Table 3. Performance of Dendrimer Membranesa at High Concentrations of CO2 in the Feed Gas carrier

∆pCO2

75% dendrimer-balance glycerol PAMAM dendrimer, generation 0 44% dendrimer-balance glycerol 2.25 M glycine sodium salt-glycerol 1 M Na2CO3-glycerol polyelectrolyte membrane PEO-containing polymer PEI/PVA blend membrane

26 30 32 30 59 36 c 30

CO2 effective permeance, GPU 0.8 0.4 0.64 1.1 2.5 9.0 2.0

CO2 permeability, barrer 300 150 240 400 90-180b 250 ∼440

RCO2/N2

ref

760 720 250 200 100 835 60 70

current work 1 1 22 21 17 31 20

a

Substrate: hydrophilized PVDF. b The range based on the range for thickness of 10-20 µm. Membrane tortuosity and porosity are not accounted for in the calculation. c Pure gas permeability.

Figure 3. Effect of the feed side average RH on dendrimer ILM performance. Feed CO2 concentration: 0.27-1.03 cmHg.

hand, the pure dendrimer membranes spanning the full thickness of the thick PVDF substrates offer significantly lower effective permeance for carbon dioxide. At high feed CO2 concentrations, both 75% dendrimer and 44% dendrimer solutions in glycerol behave in a manner similar to that for low CO2 feed concentrations. 4.3. Effect of the Feed Gas RH on the ILM Performance. The facilitation reactions of amines and carbonates with CO2 require the presence of water. Most studies in the literature on the carbon dioxide reaction with the amines were carried out in an aqueous medium. However, glycerol-based ILMs21-23 employed glycerol as the solvent for the carrier instead of water. It was found that an absence of water vapor in the feed stream resulted in a complete loss of facilitation and drastically reduced the permeabilities for both nitrogen and carbon dioxide. In the case of pure dendrimer membranes, water is also required for creating a charged environment for effective rejection of nitrogen for achieving high CO2-N2 selectivities in addition to implementing the CO2 facilitation reactions. Figure 3 shows the effect of a variation of the average feed side RH, RHf,avg (average of feed inlet and feed outlet values), on the separation behavior of dendrimerbased ILMs in a PVDF substrate on a semilog scale. The feed inlet RH was adjusted by mixing a completely humidified feed stream with a dry feed stream before it entered the cell. For a pure dendrimer ILM, a reduction in the RH of the feed stream increased nitrogen permeation tremendously, resulting in a loss of selectivity by the ILM for carbon dioxide over nitrogen. It appears that, below a critical RHf,avg value

of around 81%, the CO2-selective gating ability of the pure dendrimer ILM is reduced substantially. However, the selectivity was regained as the feed was completely humidified. It should be stressed that, in all of these experiments, dry helium was used as the sweep gas. To observe the effect of an absence of humidity in the feed stream on the ILM performance, the humidification of the feed gas stream was stopped in between. The nitrogen permeance of the ILM increased drastically, resulting in a CO2/N2 separation factor of around 5. However, as the feed gas humidification was started again, the performance of the ILM returned to its original values of high CO2-N2 selectivity and very low N2 permeability. For a completely humidified feed inlet stream, the feed exit RH is usually around 70%. However, it depends on the gas flow rate. As shown in Figure 3, the addition of a small amount of glycerol to the pure dendrimer liquid (75% dendrimer-25% glycerol) improves the ability of the liquid membrane to withstand lower feed inlet RH (and hence lower average feed side RH) before a significant increase in N2 transport occurs. Up to about a RHf,avg of 50%, the 75% dendrimer membrane blocks N2 transport very effectively. For example, for RHf,avg of ∼50-55%, no N2 could be measured in the GC. (However, the CO2 permeance values at these higher RHf,avg’s for these membranes are somewhat lower.) Glycerol sorbs more water than pure dendrimer at a given RH (as will be discussed later). Hence, addition of 25% glycerol to dendrimer liquid maintained a higher level of water content in the liquid membrane compared to that in the pure dendrimer liquid. This maintains the charged environment even at lower RHf,avg values. Thus, as far as CO2 selectivity is concerned, the 75% dendrimer membrane behaves essentially the same as the pure dendrimer membrane over a larger RHf,avg. Table 4 summarizes the performances of dendrimerbased ILMs under dry and completely humid feed gas conditions. Data obtained for pure glycerol ILM under similar conditions21 are also presented for comparison. It should be stressed that, in all of these experiments, the sweep gas was dry helium. Pure glycerol ILM has a CO2 permeability of 5.62 barrer under dry feed gas conditions with a RCO2/N2 of 1.5. When 100% RH feed gas was introduced, CO2 transport is facilitated 20 times, resulting in a RCO2/N2 of around 39, which is close to that of water. For pure dendrimer membranes, under completely humidified feed conditions, the CO2 permeability was 3600 barrer with a RCO2/N2 of 18000 for a ∆pCO2 of 0.27 cmHg. However, the significant feature of the pure dendrimer ILM was that its N2 permeability was about 0.17 barrer. This value was almost 15 times lower than that of pure glycerol ILM; the pure den-

2508

Ind. Eng. Chem. Res., Vol. 40, No. 11, 2001

Table 4. Performance of Dendrimer-Based ILMsa under Dry and 100% Humid Feed Gas Conditions dry feed gas ILM pure glycerolb 44% (1 M) dendrimer (generation 0) in glycerolc 75% dendrimer in glycerold pure dendrimer (generation 0)e

QCO2, barrer

QN2, barrer

5.62 17.0 5800 4700

3.79 7.54 860 930

100% RH feed gas RCO2/N2 1.48 2.25 6.7 5.05

QCO2, barrer

QN2, barrer

RCO2/N2

111 900 3200 3600

2.85 0.94 0.19 0.17

39 900 16300 ∼18000

a Substrate: hydrophilized PVDF. b Reference 21. c Feed 5.2% CO -balance N mixture at 7 psig of feed pressure. 2 2 balance N2 mixture at 12 psig of feed pressure. e Feed 0.5% CO2-balance N2 mixture at 3 psig of feed pressure.

drimer ILM functions as a CO2-selective molecular gate in blocking the nitrogen transport. Under completely dry feed gas conditions, the permeabilities of 75% dendrimer membrane for CO2 and N2 were 5800 and 860 barrer, respectively. These values were close to those obtained with pure dendrimer membranes under dry feed gas conditions. This was expected. It appears that the amount of glycerol added was not sufficient to penetrate into all of the intermolecular and intramolecular spaces of the dendrimer liquid; therefore, the CO2 as well as N2 permeances were very high under dry feed gas conditions. The 44% dendrimer-balance glycerol ILM under completely dry feed gas conditions has CO2 and N2 permeabilities of 17 and 7.54 barrer, respectively, with a very low selectivity for carbon dioxide over nitrogen. Addition of a significant amount of glycerol to the pure dendrimer has therefore the effect of preventing the membrane from having very high nitrogen permeability under dry feed conditions. In the presence of humidified feed gas, carbon dioxide transport was facilitated about 52 times, while the nitrogen permeation rates were reduced by 8 times for the 44% dendrimer membrane. It should be mentioned that the data presented in the table were obtained using a 5.2% CO2-balance N2, where the effects of carrier saturation were evident in the low carbon dioxide permeabilities. ILMs with thicknesses lower than that of the PVDF substrate would increase CO2 permeances at the expense of a simultaneous decrease in RCO2/N2 as in any facilitated transport membrane. However, because the present ILMs possess exceptionally high selectivities for CO2, a tradeoff between increased CO2 permeances with thinner ILMs and the resulting lower RCO2/N2 is expected to be a viable approach. Because of the lower permeances for carbon dioxide observed in pure dendrimer liquid membranes, there is a need to improve the CO2 permeances while retaining the high selectivities from pure dendrimer solutions. One way to achieve this objective is to use thinner substrate membranes. Hydrophilized Celgard 2500 may be a suitable substrate having a thickness of 25 µm compared to the 100 µm thick PVDF substrate. However, Celgard 2500 has a lower porosity, which reduces the effective permeances. 4.4. Hydrophilized Celgard 2500 Substrate. Figure 4 shows the variation of carbon dioxide effective permeance and separation factor for a hydrophilized Celgard 2500 membrane containing pure dendrimer liquid. The data were obtained using a single cell having a membrane area of 12.56 cm2. The feed gas was completely humidified; the sweep gas was dry helium. CO2 effective permeances observed as a function of ∆pCO2 are close to those observed in the PVDF substrate. For example, the CO2 effective permeance for a ∆pCO2 of 1.86 cmHg was 4.27 GPU compared to 2.29 GPU in the PVDF substrate at ∆pCO2 of 1.7 cmHg. Similarly,

d

Feed 0.76% CO2-

Figure 4. Pure dendrimer ILM performance in hydrophilized Celgard 2500 substrate.

CO2 effective permeances were higher at 1.68 GPU (at 4.95 cmHg) and 0.96 GPU (at 9.87 cmHg) with hydrophilized Celgard substrates compared to 0.97 GPU (at 4.48 cmHg) and 0.82 GPU (at 8.95 cmHg) with PVDF substrates. This is encouraging because the porosity of the Celgard substrate is only about 65% that of the PVDF substrate. However, the N2 permeabilities were 3-4 times larger; therefore, only one test cell was needed in these experiments to detect N2 in the sweep gas. The increased nitrogen permeabilities are due to inherent or residual hydrophobicity of the developmental Celgard polypropylene substrate. This was also observed by Chen et al.22 The highest separation factor obtained for a feed gas mixture of 0.5% CO2-balance N2 was close to 3900, and carbon dioxide permeability was 1000 barrer. 4.5. Equilibrium Water Uptake by the Dendrimer Liquid. It is necessary to compare the equilibrium water uptake by the dendrimer liquid and glycerol. Chen et al.21 measured the equilibrium water concentration in glycerol in the ambient RH range of 48-85%. The equilibrium water uptake by a generation 0 dendrimer liquid was measured here in the ambient RH range of 17-75%. The water uptake by glycerol liquid was also measured in this RH range; they were found to be consistent with the reported data.21 Figure 5 shows the variation of the water content in both dendrimer and glycerol as a function of the ambient RH. The measurements of the water uptake did not show any signs of hysteresis with the direction of RH change.

Ind. Eng. Chem. Res., Vol. 40, No. 11, 2001 2509

Figure 5. Equilibrium sorption of water by dendrimer and glycerol.

Figure 6. Performance of pure dendrimer ILMs in poly(acrylonitrile) PAN fiber substrate.

For all RH levels studied here, glycerol holds much more sorbed water than the pure dendrimer liquid. This phenomenon explains the nitrogen rejection behavior of the pure dendrimer liquids at high feed inlet RHs and poor selectivities observed at low feed side RHs. While pure dendrimer liquids under highly humidified feed gas conditions can presumably sorb enough water to create a highly charged environment, lower feed side humidity conditions will not be able to do so effectively. This also explains why the performance of a 75% dendrimer membrane decreases with a decrease in the feed RH much more slowly than that of a pure dendrimer membrane (Figure 3). 4.6. Performance of Dendrimer Membranes in Hollow Fiber Substrates. A generation 0 PAMAM dendrimer was immobilized in two types of hydrophilic hollow fibers potted in a module. Completely humidified feed gas was introduced in all experiments from the lumen side while the sweep gas (dry helium) was passed on the shell side. Figures 6 and 7 show the effective CO2 permeance and the RCO2/N2 for the PAN and polysulfone fiber modules, respectively. Logarithmic mean average values for the membrane areas and the partial pressure differences across the ILM were employed for calculating the permeances of both species. The experiments with a PAN fiber module were done in both cocurrent and countercurrent modes for transmembrane pressure differences up to 139 cmHg (1.8

Figure 7. Performance of pure dendrimer ILMs in polysulfone fibers.

atm) on the feed side. Higher permeances for both carbon dioxide and nitrogen and separation factors were observed in the countercurrent mode. However, higher water fluxes (not reported here) were observed in the countercurrent mode of operation, resulting in lower feed side exit RHs. The highest CO2 effective permeance obtained with PAN fibers was about 1.3 × 10-5 cm3/ cm2‚s‚cmHg, resulting in a separation factor of 1300 for a ∆pCO2 of 0.28 cmHg when the feed and sweep gases were flowing countercurrently. The experiments with polysulfone fiber modules were conducted only in the cocurrent mode. The effective permeances and water fluxes observed were higher compared to those observed in the PAN fiber module at comparable ∆pCO2 values. This is expected because the PAN fibers, used for ultrafiltration, have a lower porosity whereas polysulfone fibers, used for microfiltration, have a significantly higher porosity. The highest effective permeance for the polysulfone fibers was about 9.7 × 10-6 cm3/cm2‚s‚cmHg for a feed CO2 partial pressure of 0.45 cmHg. Even though the effective permeances for carbon dioxide in these hollow fiber modules were comparable to those obtained in flat membrane ILMs, the effective permeances for nitrogen are, in general, higher in the case of both hollow fiber modules. The effective permeance of nitrogen through the PAN module was between 6.8 and 9.3 × 10-9 cm3/cm2‚s‚cmHg. These permeances for pure dendrimer membranes were 2-3 times higher than those obtained with sodium carbonate-glycerol ILMs in the PAN fibers.23 Similarly, the pure dendrimer ILMs in polysulfone hollow fibers have a nitrogen effective permeance in the range of (3.9-9) × 10-9 cm3/cm2‚s‚cmHg. These values are comparable to those obtained in polysulfone fibers with sodium carbonate-glycerol ILMs.23 The discrepancy in N2 permeation rates between ILMs in hollow fibers and flat membranes is expected because of the differences in the procedure for forming the dendrimer ILMs in flat membranes and hollow fibers and the very nature of dendrimer liquids. The dendrimer liquid was introduced as a solution in methanol into the hollow fibers, and methanol was subsequently evaporated by applying a mild vacuum (12.7-38.1 cmHg) from the shell side, leaving pure dendrimer behind in the pores of the hollow fibers. The dendrimer in a methanol solution was applied at atmospheric pressure from the lumen side because application of higher pressures may result in a loss of

2510

Ind. Eng. Chem. Res., Vol. 40, No. 11, 2001 Table 5. Gas Transport of Dendrimer-Based ILMs for CO2-O2 Mixtures ILM pure generation 0 75% generation 0 44% generation 0

Figure 8. Effect of the transmembrane pressure difference on QCO2 in pure dendrimer ILMs.

the dendrimer to the shell side. It is possible that some of the fine pores of the hollow fibers may not be filled with dendrimer but will permeate some nitrogen, resulting in a loss of selectivity for carbon dioxide vis-a`vis N2. The reason could also be the inherent hydrophobic nature of the substrates. This is particularly true in the case of PAN fibers which were hydrophilized, because poly(acrylonitrile) is inherently hydrophobic. Better techniques of hydrophilization of hydrophobic fibers to ensure uniform and permanent hydrophilization could improve this situation. Better approaches for immobilization of the dendrimer in the pores should result in higher separation factors by hollow fiber membrane ILMs. An additional reason for higher N2 permeance is the somewhat lower exit gas RH on the feed side compared to that for flat membranes which will increase the N2 permeance. 4.7. Effect of the Transmembrane Pressure Difference. ILMs consist of liquids immobilized in the pores of the substrates. The maximum transmembrane pressure difference (TMP) for an ILM varies inversely with the pore size of the substrate and directly with the surface tension of the ILM liquid. For any practical application, the ILM should be able to withstand the transmembrane pressure difference. The PVDF film used here has a water bubble point pressure of 4.8 atm (360 cmHg).21 Figure 8 shows the variation of carbon dioxide permeability for different transmembrane pressures for both hydrophilic PVDF and hydrophilized Celgard 2500 based pure dendrimer ILMs. The PVDF membranes were not tested for TMPs higher than 0.68 atm (51.7 cmHg). The hydrophilized Celgard 2500 membranes were tested up to a TMP of 1.7 atm (120 cmHg). It should also be noted that the PVDF membranebased ILM was tested with a 0.5% CO2-N2 feed gas mixture. The steep behavior of the ILM performance with transmembrane pressure is due to the strong dependence of the carbon dioxide permeability on CO2 partial pressure differences at low CO2 concentrations in the feed gas mixtures. This trend can be seen in Figure 8. Data for Celgard 2500 membrane were obtained using a 5.2% CO2-N2 feed gas mixture where the effect of facilitation will not be very evident at the relatively higher ∆pCO2’s. 4.8. Dendrimer Liquid Membranes for Separation of CO2/O2 Mixtures. Table 5 presents a limited amount of data on the performance of the generation 0 dendrimer-based liquid membranes for separation of carbon dioxide/oxygen mixtures using relatively low

∆pCO2 (QCO2/tm)eff 0.74 0.56 0.99

5.34 11.8 7.59

(QO2/tm)eff

QCO2 QO2 RCO2/O2

2.35 × 2000 0.87 1.16 × 10-2 4400 4.28 -3 8.41 × 10 2800 3.10 10-3

2300 1000 920

∆pCO2’s of up to 0.99 cmHg. At a ∆pCO2 of 0.74 cmHg, the pure dendrimer ILM has a QCO2 of 2000 barrer and a RCO2/O2 of 2300. The carbon dioxide permeability was comparable to that obtained using a humidified CO2N2 gas mixture under similar CO2 partial pressures. The permeability of oxygen through the pure dendrimer ILM was about 0.87 barrer compared to the nitrogen permeability of 0.17 barrer. This could be partly because of a significantly higher solubility of oxygen in water compared to nitrogen. However, the fivefold difference in O2 and N2 permeabilities cannot be due to higher solubility of oxygen in water alone. Additional experiments are needed to clarify this aspect. Because the feed gas stream was completely humidified, water sorbed by the dendrimer liquid yielded higher transport rates for oxygen than for nitrogen. The addition of glycerol to a generation 0 dendrimer increases the permeabilities of both carbon dioxide and oxygen with a reduction in the selectivity of the ILM for carbon dioxide over oxygen to about 1000. This trend was seen with both 75% dendrimer and 44% dendrimer ILMs. The presence of glycerol in the liquid membrane increases the effective permeance of oxygen more than the increase in the effective permeance of carbon dioxide for similar carbon dioxide partial pressures. As a result, the CO2/O2 separation factors are lower for dendrimer-glycerol mixture ILMs than for pure dendrimer liquid membranes. 5. Conclusions From the above discussion, the following conclusions can be drawn: (1) The CO2-selective molecular gating ability of pure dendrimer liquid membranes was preserved over a wide range of feed gas RHs by adding 25% glycerol to the dendrimer liquid. The addition of small amounts of glycerol does not seem to affect the CO2 permeances for the CO2 partial pressure range studied. (2) The dendrimer-based liquid membrane provides the highest selectivities for CO2-N2 separation both at low and high ∆pCO2’s. There is a need to improve the CO2 permeances, particularly at high CO2 partial pressures. (3) Hydrophilized Celgard membranes having a thickness lower than that of PVDF substrates marginally improved the CO2 effective permeances. This is encouraging because the porosity of the developmental hydrophilized Celgard substrate is only about 65% of that of the PVDF substrate. (4) The dendrimer liquid membranes were able to withstand TMPs of up to 52 cmHg in a 0.1 µm PVDF substrate and 120 cmHg in a 0.08 µm hydrophilized Celgard substrate. Studies were not made at higher TMP values. (5) The water uptake of pure dendrimer liquids was experimentally determined and compared with pure glycerol liquid at different ambient RHs. Dendrimer liquid absorbs less water than glycerol at comparable ambient RH values.

Ind. Eng. Chem. Res., Vol. 40, No. 11, 2001 2511

(6) The use of hollow fiber substrates for containing the ILMs was explored with two types of hollow fibers. Their performances were consistent with the observations made with glycerol-based ILMs conducted separately. Acknowledgment We acknowledge the support from the Membrane Separations and Biotechnology Program at NJIT. We thank Celgard LLC, Charlotte, NC, for providing their developmental hydrophilized Celgard 2500 films. A.S.K. appreciates the support of Dr. Hua Chen on various aspects during the present work. Nomenclature A ) membrane area pi ) partial pressure of species i P ) pressure Qi ) permeability coefficient RH ) relative humidity RHf,avg ) average of feed inlet and outlet RHs tm ) membrane thickness Vi ) volumetric permeation rate of species i Subscripts/Superscripts/Greek Letters eff ) effective i, j ) species in ) inlet condition lm ) logarithmic mean out ) outlet condition t ) total true ) true Ri/j ) separation factor of species i over j m ) porosity of the substrate τm ) tortuosity of the substrate

Literature Cited (1) Kovvali, A. S.; Chen, H.; Sirkar, K. K. Dendrimer membranes: A CO2-selective molecular gate. J. Am. Chem. Soc. 2000, 122, 7594. (2) Ward, W. J.; Robb, W. L. Carbon dioxide-oxygen separation: Facilitated transport of carbon dioxide across a liquid film. Science 1967, 156, 1481. (3) Meldon, J. H.; Stroeve, P.; Gregoire, C. E. Facilitated transport of carbon dioxide: A review. Chem. Eng. Commun. 1982, 16, 263. (4) Bhave, R. R.; Sirkar, K. K. Gas permeation and separation by aqueous membranes immobilized across the whole thickness or in a thin section of hydrophobic microporous Celgard films. J. Membr. Sci. 1986, 27, 41. (5) Bartsch, R. A., Way, J. D., Eds. Chemical Separations with Liquid Membranes; ACS Symposium Series 642; American Chemical Society: Washington, DC, 1996. (6) Meldon, J. H.; Paboojian, A.; Rajangam, G. Selective CO2 permeation in immobilized liquid membranes. AIChE Symp. Ser. 1986, 248, 114. (7) Way, J. D.; Noble, R. D. Facilitated Transport. In Membrane Handbook; Ho, W. S. W., Sirkar, K. K., Eds.; Chapman and Hall: New York, 1992. (8) Noble, R. D.; Way, J. D. Description of facilitated transport and environmental applications. In Membrane processes in separation and purification; Crespo, J. G., Boddeker, K. W., Eds.; Kluwer Publishers: Dordrecht, The Netherlands, 1994; p 317. (9) Matson, S. L.; Lee, E. K. L.; Friesen, D. T.; Kelly, D. J. Acid gas scrubbing by composite solvent-swollen membranes. U.S. Patent 4,737,166, 1988. (10) Majumdar, S.; Guha, A. K.; Sirkar, K. K. A new liquid membrane technique for gas separation. AIChE J. 1988, 34, 1135. (11) LeBlanc, O. H.; Ward, W. J.; Matson, S. L.; Kimura, S. G. Facilitated transport of CO2 in ion-exchange membranes. J. Membr. Sci. 1980, 6, 339.

(12) Way, J. D.; Noble, R. D.; Reed, D. L.; Ginley, G. M.; Jarr, L. A. Facilitated transport of CO2 in ion exchange membranes. AIChE J. 1987, 33, 480. (13) Matsuyama, H.; Teramoto, M.; Sakakura, H.; Iwai, K. Facilitated transport of CO2 through various ion exchange membranes prepared by plasma graft polymerization. J. Membr. Sci. 1996, 117, 251. (14) Matsuyama, H.; Teramoto, M. Facilitated transport of carbon dioxide through functional membranes prepared by plasma graft polymerization using amines as carrier. In Chemical Separations with Liquid Membranes; Bartsch, R. A., Way, J. D., Eds.; ACS Symposium Series 642; American Chemical Society: Washington, DC, 1996. (15) Laciak, D. V.; Quinn, R.; Pez, G. P.; Applyby, J. B.; Puri, P. S. Selective permeation of ammonia and carbon dioxide by novel membranes. Sep. Sci. Technol. 1990, 25, 1295. (16) Quinn, R.; Appleby, J. B.; Pez, G. P. New facilitated transport membranes for the separation of carbon dioxide from hydrogen and methane. J. Membr. Sci. 1995, 104, 139. (17) Quinn, R.; Laciak, D. V. Polyelectrolyte membranes for acid gas separations. J. Membr. Sci. 1997, 131, 49. (18) Quinn, R.; Laciak, D. V.; Pez, G. P. Polyelectrolyte-blend membranes for acid-gas separations. J. Membr. Sci. 1997, 131, 61. (19) Teramoto, M.; Nakai, K.; Ohnishi, N.; Huang, Q.; Watari, T.; Matsuyama, H. Facilitated transport of carbon dioxide through supported liquid membranes of aqueous amine solutions. Ind. Eng. Chem. Res. 1996, 35, 538. (20) Matsuyama, H.; Terada, A.; Nakagawara, T.; Kitamura, Y.; Teramoto, M. Facilitated transport of CO2 through polyethyleneimine/poly(vinyl alcohol) blend membrane. J. Membr. Sci. 1999, 163, 221. (21) Chen, H.; Kovvali, A. S.; Majumdar, S.; Sirkar, K. K. Selective CO2 separation from CO2-N2 mixtures by immobilized carbonate-glycerol membranes. Ind. Eng. Chem. Res. 1999, 38, 3489. (22) Chen, H.; Kovvali, A. S.; Sirkar, K. K. Selective CO2 separation from CO2-N2 mixtures by immobilized glycine-Naglycerol membranes. Ind. Eng. Chem. Res. 2000, 39, 2447. (23) Chen, H.; Obuskovic, G.; Majumdar, S.; Sirkar, K. K. Immobilized glycerol-based liquid membranes in hollow fibers for selective CO2 separation from CO2-N2 mixtures. J. Membr. Sci. 2001, 183, 75. (24) Roberts, J. C.; Bhalgat, M. K.; Zera, R. T. Preliminary biological evaluation of polyamidoamine (PAMAM) Starburst dendrimers. J. Biomed. Mater. Res. 1996, 30, 53. (25) Bosch, H.; Versteeg, G. F.; van Swaaij, W. P. M. Gas-liquid mass transfer with parallel reversible reactions. I. Absorption of CO2 into solutions of sterically hindered amines. Chem. Eng. Sci. 1989, 44, 2723. (26) Little, R. J.; van Swaaij, W. P. M.; Versteeg, G. F. Kinetics of carbon dioxide with tertiary amines in aqueous solution. AIChE J. 1990, 36, 1633. (27) Versteeg, G. F.; van Swaaij, W. P. M. On the kinetics between CO2 and alkanolamines both in aqueous and nonaqueous solutions: Tertiary amines. Chem. Eng. Sci. 1988, 43, 587. (28) Prasad, R.; Kiani, A.; Bhave, R. R.; Sirkar, K. K. Further studies on solvent extraction with immobilized interfaces in a microporous hydrophobic membrane. J. Membr. Sci. 1986, 26, 79. (29) Li, Y.-G.; Mather, A. E. Correlation and prediction of the solubility of carbon dioxide in a mixed alkanolamine solution. Ind. Eng. Chem. Res. 1994, 33, 2006. (30) Li, Y.-G.; Mather, A. E. Correlation and prediction of the solubility of CO2 and H2S in aqueous solutions of triethanolamine. Ind. Eng. Chem. Res. 1996, 35, 4804. (31) Hirayama, Y.; Kase, Y.; Tanihara, N.; Sumiyama, Y.; Kusuki, Y.; Haraya, K. Permeation properties to CO2 and N2 of poly(ethylene oxide)-containing and cross-linked polymer films. J. Membr. Sci. 1999, 160, 87. (32) Haraya, K.; Nakajima, M.; Itoh, N.; Kamisawa, C. Feasibility study of the application of membrane separation in CO2 removal from flue gases. Kagaku Kogaku Ronbunshu 1993, 19, 714. Cited in ref 20.

Received for review December 5, 2000 Revised manuscript received March 14, 2001 Accepted March 15, 2001 IE0010520