Continuous Supercritical CO2 Process using Nanofiltration by

Apr 28, 2009 - Green Nuclear Research Laboratory, EIRC, Kyung Hee University, Yongin, Kyungkido, 449-701, South Korea. Ind. Eng. Chem. Res. , 2009, 48...
2 downloads 0 Views 2MB Size
Download PDF
5406

Ind. Eng. Chem. Res. 2009, 48, 5406–5414

SEPARATIONS Continuous Supercritical CO2 Process using Nanofiltration by Inorganic Membrane Moonsung Koh,* Bruno Fournel, Stephane Sarrade, Luc Schrive, and Ivan Stoychev CEA Marcoule DEN/DTCD/SPDE/Laboratoire des Fluides Supercritiques et Membranes (LFSM), BP 17171-30207 Pierrelatte Cedex, France

Patrick Lacroix-Desmazes† and Tiphaine Ribaut†,‡ Institut Charles Gerhardt, UMR5253 CNRS/UM2/ENSCM/UM1, Inge´nierie et Architectures Macromole´culaires, Ecole Nationale Supe´rieure de Chimie de Montpellier, 8 rue de l’Ecole Normale, 34296 Montpellier Cedex 5, France, and Institut de Chimie Se´paratiVe de Marcoule, UMR 5257, ICSM Site de Marcoule, BP 17171, 30207 Bagnols Sur Ceze Cedex, France

Kwangheon Park Green Nuclear Research Laboratory, EIRC, Kyung Hee UniVersity, Yongin, Kyungkido, 449-701, South Korea

Nanofiltration using inorganic membranes was conducted to develop a continuous process in supercritical CO2. Inorganic membranes of various sizes (1, 5, 50, 300 kDa), materials (Al2O3/TiO2/ZrO2, Al2O3/ZrO2/ TiO2), and modules (one and three channels) were used for the experiments. The effects of fluid viscosity and density, membrane active layer thickness, number of channels, pore size, and the flux of a fluid were analyzed when supercritical CO2 (sc-CO2) passed through the membrane. In the case of PE6100 surfactant, it was confirmed that sc-CO2 passed through the membrane at a flow rate and transmembrane pressure (∆P) greater than 10 mL/min and 0.04 MPa, respectively, although the size of the surfactant was larger than that of the membrane pore. The phenomena were proven to have been caused by the folding effect. In the case of microemulsions using NP-4 and H2O, it was confirmed that water adsorbed on the membrane surface passed through at ∆P greater than 0.7 MPa. Based on these findings, experiments were conducted to separate microemulsions and dispersions. The experimental results indicated that microemulsions achieved 47% separation efficiency and 90% surfactant recovery efficiency due to the displacement of metal ions through the water adsorbed on the membrane surface. Conversely, dispersions yielded high separation efficiency. Also, the correlation between the contaminant size and the membrane pore size was confirmed through SEM images. Introduction As environmental regulations are increasingly reinforced to avoid the use of toxic organic solvents and the generation of secondary wastes, alternative solvents or techniques are earnestly needed. It is reasonable to assume that over the past 20 years supercritical carbon dioxide (sc-CO2) has gained considerable attention in heavy metal extraction,1-4 dry cleaning,5,6 nuclear decontamination,7-13 and wafer cleaning.14,15 However, one difficulty of using CO2 is that most of the solutes may involve metal or polar materials as well as organics. Therefore, CO2 would require the use of specific cosolvents, surfactants, and metal chelating ligands. However, almost all synthesized surfactants and ligands have several problems, such as being expensive and complicated to synthesize. These concerns make it necessary to recover and recycle the CO2-philic chemicals (ligands or surfactants) for the purpose of improving the economics of the commercial application process. Also, the use of CO2 as an environmentally responsible solvent in industrial processes will require a recycling process to prevent emissions * To whom correspondence should be addressed. Tel.: +82 31 201 2917. Fax: +82 31 202 2410. E-mail: [email protected]. † Institut Charles Gerhardt. ‡ Institut de Chimie Se´parative de Marcoule.

into the atmosphere. For these reasons, much attention has focused on the recovery and recycling process. Many authors have shown various regeneration methods which are capable of separating solute-sc-CO2 mixtures: pressure and temperature change, adsorption, absorption, entrainers, and membranes. Probably the simplest method is depressurization and heating. However, this process influences not only effectiveness but also economical aspects. According to Lack and Seidlitz’s data,16 recompression is the most cost-intensive method compared with adsorption (e.g., activated carbon) or absorption in a wash tower. In this regard, adsorption or absorption makes a solvent cycle feasible at nearly constant pressure. Considering energy consumption, these are the effective operating modes. However, it is only available to the downstream operating mode. In addition, a further treatment of adsorbent or absorption solvent should be required.17,18 Recently, membranes have been applied to the regeneration of CO2 since the association of a membrane with the supercritical fluid extraction process may avoid intense depressurization, as well as the secondary treatment. In the initial stage, an asymmetric Kapton membrane was applied to separate sc-CO2/ ethanol and sc-CO2/isooctane mixtures, by Semenova and Ohya.19 They showed that the permeabilities of gases (CO2,

10.1021/ie800452g CCC: $40.75  2009 American Chemical Society Published on Web 04/28/2009

Ind. Eng. Chem. Res., Vol. 48, No. 11, 2009

Figure 1. Schematic diagram of the continuous CO2 process using inorganic membrane: (1) high pressure reactor, (2) inorganic membrane, (3) high pressure circulation pump. Table 1. Properties of Inorganic Membranes membrane material membrane size (kDa) 1 1 5 5 50 300

composition

membrane geometry

MWCO pore size thickness (g/mol) (nm) module type (mm)

Al2O3/TiO2/ZrO2 1000 Al2O3/ZrO2/TiO2 1000 Al2O3/TiO2/ZrO2 5000 Al2O3/ZrO2/TiO2 5000 Al2O3/TiO2/ZrO2 50000 Al2O3/TiO2/ZrO2 300000

0.6 0.6 2.0 2.0 20.0 100.0

single channel three channel single channel three channel three channel three channel

2.0 1.0 2.0 1.0 1.0 1.0

N2, CH4, H2, Ar, O2, and He) depended on the diffusion factor in the range of high temperature. In 1993, a reverse osmosis membrane was proposed to separate the ethanol-content aqueous solution by sc-CO2, by Hsu and Tan.20 The ethanol rejection could be 70% higher than the best result obtained in sc-CO2. In 1995, an inorganic membrane, which has a high compressible strength, thermal stability, and solvent resistance, was utilized to separate the low volatile compounds (LVC) extracted into sc-CO2, by Biritgh.21 As a result of this study, 80-90% of LVC was recovered by a nanofiltration membrane without a significant drop in pressure. Also, regeneration with the inorganic membrane could be established as being ideal for matching with the inorganic membrane; that is, silica, which is one type of material used for membranes, has a low affinity for hydrophobic compounds. Therefore, CO2 does not absorb well on microporous silica, compared to silicalite-1 zeolite or microporous and activated carbons. Especially for high pressure applications it is advantageous that CO2 is weakly adsorbed in microporous silica membranes.22 Since then, inorganic membrane techniques in sc-CO2 including recovery, separation, and purification have been investigated.23-26 Sarrade and co-workers discussed the characterizations of coupling a sc-CO2 extraction process with nanofiltration. They then studied various applications including natural products,27 used oils,28 and electrolytic solutions.29 Chiu and Tan24 studied the separation of sc-CO2/ caffeine mixture with a membrane possessing nanoscale pore size. They showed that the highest caffeine rejection and CO2 permeation flux were caused by the presence of a large cluster size in the near-critical region. Therefore, pressure reduction was required for CO2 regeneration. Yonker and co-workers25 discussed membrane separation in two different solvent systems: one microemulsion in near-critical propane and the second in supercritical ethane. They showed that the surfactant and water core were passed through the membrane, while the macromolecule selectivity was based on size and molecular weight. Patil and co-workers26 investigated the permeation of sc-CO2 and SF6 across polymeric and inorganic membranes as a function of temperature and feed pressure. They found that the main

5407

mechanism for the mass transport was viscous flow, and there was almost no contribution from surface diffusion through the micropores. To make the best use of a supercritical fluid as a solvent, it is desirable to use a flowing, continuous system. Therefore, we will discuss the membrane separation in two different solvent systems: one is microemulsion and the other is dispersion in sc-CO2. In principle, a process for continuously carrying out the solutes, such as contaminants using nanofiltration in sc-CO2, comprises the following steps: (a) providing an apparatus including a high pressure reactor, ceramic membrane, and circulation pump as shown in Figure 1; (b) extracting or dissolving the solutes (like contaminants) from the substrate with microemulsions or dispersions in sc-CO2; (c) carrying the solutes to the inorganic membrane by means of flushing CO2; (d) separating the solutes from microemulsions or dispersions using the inorganic membrane without a pressure drop; (e) recycling the separated surfactant to the high pressure cell using the circulation pump; and (f) making it possible to have a continuous process without a depressurization step. To do this, we will investigate the effects of the operating pressure, the membrane pore sizes, the transmembrane pressure, and the flow rate to demonstrate the continuous CO2 process with an inorganic membrane. Finally, the present paper will provide the availability to design a continuous CO2 system with an inorganic membrane. This technology can be used to recover a wide range of polar solute molecules from supercritical fluids. Experimental Section Chemicals and Materials. CO2 (99.7% in mass) was supplied from the Messer Co. Two kinds of surfactants, NP-4 and PE6100, were used for microemulsions and dispersions in sc-CO2, respectively. NP-4 (nonylphenol-4-polyethoxylated, MW ) 396 g/mol) was obtained from Nicca Korea and Hannong Chemicals. Pluronics PE6100 (copolymer of ethylene oxide and propylene oxide, MW ) 1650 g/mol) was supplied by BASF Corp. in Germany. CeO2 used as a particulate simulant for contamination (diameter of particles ranged from 20 to 30 nm) was purchased from Aldrich. Co(NO3)2 · 6H2O for preparing the cobalt solution (200 ppm) was purchased from Aldrich. All other chemicals used were of analytical reagent grade. The experiments described in the following were conducted with various inorganic membranes defined in Table 1. Singleand three-channel membranes were made of a mixture of Al2O3/ TiO2/ZrO2 or Al2O3/ZrO2/TiO2, respectively. The membranes used were 155 mm long and have an external diameter of 10 mm. Experimental Setup. The setup for nanofiltration using inorganic membrane in sc-CO2 is represented in Figure 2. The high pressure reactor (11) having an electromagnetic stirrer (10), is used to form the microemulsions or the dispersions. Also, the reaction taking place inside could be observed clearly through two sapphire windows on both sides of the reactor. The inorganic membrane (L ) 155 mm, o.d. ) 100 mm, i.d. ) 70 mm) is placed inside the membrane housing (12). Effective surface areas of the single channel and three channels are 25 and 52 cm2, respectively. This system could be operated at a maximum transmembrane pressure of 1.0 MPa. Liquid CO2 from the CO2 tank (1) goes through a liquid pump (2) into the reservoir to preheat with the heat exchanger and prevent flow fluctuation. The temperature is controlled at 40 ( 1 °C by two methods, which are the oven (25) and the water circulation pump. The pressure of permeation and retention is minutely regulated by the actuator valves (16, 17). The total flow of

5408

Ind. Eng. Chem. Res., Vol. 48, No. 11, 2009

Figure 2. Experimental setup for nanofiltration using inorganic membrane in sc-CO2 process: (1) liquid CO2 tank, (2) gas booster, (4) preheater, (5) CO2 reservoir (2000 mL), (10) electromagnetic stirrer, (11) high pressure reactor (270 mL), (12) membrane housing, (20) collection vial, (3, 6, 7, 8, 9) metering valves, (13, 14, 15) pressure transducers, (16, 17) forward pressure regulators, (18, 19) separators, (21, 22) gas meters, (23, 24) pressure indicators, (25) oven.

permeation and retention is measured by gas meters (21, 22). All permeation experiments were done on fresh membranes with cleaning for each run. All flux measurements have been experimentally measured with an experimental error of about (5-15%. Membrane Cleaning. After each experiment, the membrane is cleaned to restore the peculiar flux. In the case of a permeation experiment of surfactants or microemulsions with H2O, it is cleaned with EtOH solution for 2 h, while in the case of microemulsions with cobalt solution it is cleaned with 1 M HNO3 solution for 3 h and then EtOH solution for 2 h, respectively. Lastly, the cleaned membrane was flushed with pure CO2 at 20 MPa, 40 °C, and a flow rate of 30 mL/min, until the initial flux was completely restored. Analysis. To measure the permeation rate with time, the surfactants permeated through the membrane were collected in a separator and measured by the microbalance. The performance of the NF (nanofiltration) membrane was determined as a function of the surfactant permeation, defined as permeation rate )

CP × 100% CF

(1)

where CP and CF are the surfactant concentrations in the permeation and in the feed solution, respectively. The surface of the inorganic membrane, which was used in the dispersions with surfactant and CeO2, was observed, and the thickness of the selective layer of the membrane was determined using a scanning electron microscope (SEM; Steroscan 440, Leica, Cambridge, England) and a field emission scanning electron microscope (FE-SEM; LEO SUPRA 55, Germany). To confirm the separation of metal ions from microemulsions, the amounts of cobalt ions absorbed into the membrane were analyzed by inductively coupled plasma atomic emission spectroscopy (ICP-AES; LEEMAN ABS, Inc., U.S.A.). IR spectra were recorded with a Perkin-Elmer 1000 spectrometer using silicon plates in the range 400-4000 cm-1. Results and Discussion Effect of Feed Pressure in Pure sc-CO2 Permeation. The high pressure reactor and membrane housing were pressurized by sc-CO2 preconditioned in the reservoir. At this time, valve 7 (see Figure 2) should be opened to prevent the damage or breakage of the membrane. When the desired pressure at the feed side was obtained, valves 6 and 7 were closed. To perform the permeation experiment, CO2 was continuously fed to the

Figure 3. Flux variations of pure CO2 as a function of transmembrane pressure for various feed pressures ((A) 10, (B) 15, (C) 20, and (D) 25 MPa) with 1 kDa of three-channel membrane at 40 °C.

Figure 4. Flux variations of pure CO2 as a function of transmembrane pressure for various membrane geometries at 40 °C and 25 MPa. The pore sizes are denoted by solid and open symbols, respectively. Membrane resistances at a transmembrane pressure of 1.0 MPa.

feed side of the membrane by maintaining the working pressure. Fluxes were measured with increasing transmembrane pressure steps for each working pressure. Therefore, the fluxes of pure CO2 through 10, 15, 20, and 25 MPa were linear functions of the transmembrane pressure at constant temperature (40 °C). From Figure 3 it follows that the flux decreased more or less linearly with the feed pressure. In the region of the critical pressure of 10 MPa, the viscosity and density of CO2 changed in such a way that a maximum flux was observed in the permeation. Above the critical point, both the density and viscosity change with pressure, and as a result a slowly decreasing CO2 flux was obtained. It should be noted that there is a relatively large change in the viscosity for the experiment with the feed pressures of 15 and 10 MPa. To verify the effect of viscosity, the membrane resistance (Rm, [m-1]) was calculated in different components based on the resistance in Darcy’s law: Rm )

∆P µCO2J

(2)

where ∆P is the transmembrane pressure [Pa], J is the flux through the membrane [L/h/m2], and µCO2 is the viscosity [Pa · s] of sc-CO2 at the working conditions. In principle, any fluid may interact with a porous medium by adsorption, absorption, and capillary condensation, which means that the fluid may liquefy inside a micropore. That is, it can be said that the permeation of a fluid depends on the pressure applied: As the pressure rises,

Ind. Eng. Chem. Res., Vol. 48, No. 11, 2009

5409

Figure 5. SEM cross-sectional images of ZrO2-coated active layers for 1 and 5 kDa membranes: (A) 1 kDa, single-channel membrane; (B) 1 kDa, threechannel membrane; (C) 5 kDa, single-channel membrane; (D) 5 kDa, three-channel membrane.

Figure 6. Flux variations for various solutions in sc-CO2 as a function of transmembrane pressure with 5 kDa of three-channel membrane at 25 MPa and 40 °C.

molecular layers gradually deposit on the wall and, at a sufficiently high pressure, the pore may be completely filled with liquid. However, the supercritical fluids would not be subjected to capillary condensation, but only to a locally higher density that would increase near the pore in materials with pore diameters on the order of 1 nm. That is, since the fluid has simultaneously regions of both gas and liquid local densities, a part of the local density with liquid generates the aggregation of solvent molecules close to the membrane.

Figure 7. Flux variations for various solutions in sc-CO2 as a function of transmembrane pressure with 1 kDa of three-channel membrane at 25 MPa and 40 °C.

Effect of Membrane Geometry in Pure sc-CO2 Permeation. CO2 fluxes for different geometries of one and three channels, and pore sizes of 1 and 5 kDa, respectively, were carried out and are presented in Figure 4. For three-channel membranes with different pore sizes, the CO2 flux and membrane resistance of the 5 kDa membrane were 2530 L/h/m2 and 16 × 1012 m-1, and those of the 1 kDa membrane were 1525 L/h/m2 and 27 × 1012 m-1 at a transmembrane pressure of 1.0 MPa. The difference of resistance by means of pore sizes was over 50% between 1 and 5 kDa of membranes. On the contrary, single-channel membranes with different pore sizes were less different than three-channel membranes. However, identical pore

5410

Ind. Eng. Chem. Res., Vol. 48, No. 11, 2009 Table 2. Nanofiltration of Microemulsions Using 1 kDa (Three-Channel) Membranea material balance analysis

subject

separator

reactor

membrane

total

balance ICP

NP-4 Co2+

92% (trapped) 40% (residue)

1% 7%

2% (residue) 47% (trapped)

95 ( 6% 94 ( 5%

a At 250 bar, 40 °C, stirring time (20 min), NP-4 (1.4 g), solution (1.26 mL, Co2+ 252 µg).

Figure 8. Effect of flow rate in membrane separation with 1 kDa of singlechannel membrane at 40 °C and 25 MPa.

Figure 9. Microemulsion permeation as increasing the transmembrane pressure with 1 kDa (three-channel) membrane at 40 °C and 25 MPa.

size membranes with different channels are considerably different. For 5 kDa membranes with different channel numbers, the CO2 flux and membrane resistance of a single channel were 1125 L/h/m2 and 37 × 1012 m-1. Compared with three channels, the difference in resistance by means of pore sizes exceeded 2 times. It can be said that the membrane resistance depends on pore size and geometry. Especially, the geometry of the membrane could be more greatly influenced in permeation, due to the thickness of the active layer of membrane. Coating the membrane for the active layer, the three-channel membrane was coated heterogeneously since the geometry was so complex. Therefore, some parts of the surface could be thinner than other parts. To prove this, the average thickness of the active layer was measured by a scanning electron microscope (SEM). Figure 5 shows SEM cross-sectional images of four kinds of membranes. The active layers were about 4 µm (single channel) and 1 µm (three channels) for 1 kDa membranes and 16 µm (single channel) and 8 µm (three channels) for 5 kDa membranes. It can be proved that the thickness of the active layer could have an effect on the flux. Also, this explained that single-channel membranes could not obtain enough points of CO2 flux under the effect of geometry. Effect of Pore Sizes in Surfactant and Microemulsion Permeation. For the permeation of various solutions in sc-CO2, microemulsions with NP-4 and H2O and surfactants with different molecular weights of 396 and 1650 g/mol for NP-4 and PE6100, respectively, were prepared. First of all, to dissolve

the surfactant or make the microemulsions, all desired chemicals were placed into the high pressure reactor (270 mL) and then sealed, which was heated in the oven up to 40 °C. Subsequently, the high pressure reactor and membrane housing were pressurized by sc-CO2 in the reservoir. When all chemicals were dissolved clearly (after mixing with 500 rpm for 20 min), the permeation mode was set by opening the valve between the reactor and the membrane housing. The fluxes for 1 and 5 kDa membranes were measured at various transmembrane pressures and for CO2 flow rate. CO2 flux was recorded after flowing for 3 min in each condition. As seen from Figure 6, all of the flux lines are almost the same as that of pure CO2 feed flux. Since the pore size of the membrane is significantly larger compared to the molecular size of all chemicals, the fluxes for all conditions through 5 kDa membranes with a pore diameter of 2 nm are not influenced by any rejectivity or selectivity. In the literature, a separated water phase may form capillary water bridges in the narrow pore, effectively leading to complete blocking of these pores for transport.30 However, no substantial differences were observed in 5 kDa of membrane. Consequently, it can be said that the pores of the membrane are wide enough to prevent spontaneous capillary condensation of the surfactant and even water. On the contrary, the flux through 1 kDa of membrane exhibited some remarkable features as shown in Figure 7. All the fluxes through the 1 kDa membrane are lower than those through the 5 kDa membrane. Moreover, in contrast to the 5 kDa membrane, the fluxes are varied under different conditions, which means that the fluxes decreased as the molecular size increased. This flux decline is expected when the surfactant or microemulsions form a cake layer on the membrane surface. As filtration progresses, the solvent mass transport resistance in the cake layer increases with time and reaches a maximum resistance. Theoretically, since the molecular weight of the surfactant was larger than the pore size of the membrane, PE6100 cannot pass through the membrane. As seen in the flux with PE6100, however, it passed through the membrane over entire ranges of transmembrane pressure. According to the previous study, although the surfactants have a molecular size larger than the pore size of the membrane, surfactants still could penetrate the smaller pores. Because nonionic surfactants consist of long chains that can be folded to a substantial degree, they can penetrate and adsorb in pores with a smaller molecular weight cutoff (MWCO) than their molecular size.31 Comparison with the data between NP-4 and microemulsions (NP-4 + H2O, W ) 20) shows that the flux of microemulsions was decreased slightly more than that of NP-4 due to water adsorption on the membrane. However, since the amount of water contained per microemulsion was quite small, the blocking of pores was not enough for capillary condensation. Also, there must be an effect of bicontinuous microemulsions, which may coexist predominantly in sc-CO2. Namely, demicroemulsions and remicroemulsions could be repeated continuously. Therefore, the separated surfactants and water could be permeated more easily. The membrane resistances for various solutions were calculated from the fluxes. Based on the resistance of pure

Ind. Eng. Chem. Res., Vol. 48, No. 11, 2009

5411

Figure 10. FE-SEM cross-section images of membrane surface before and after test: (A) 5 kDa membrane before test; (B) 5 kDa membrane after test; (C) 300 kDa membrane before test; (D) 300 kDa membrane after test. At 250 bar, 40 °C, stirring time (30 min), PE6100 (1 g), and CeO2 (20 mg).

Figure 11. Evolution of IR spectra as a result of successive transformations of the initial NP-4 (dotted line) to product collected from 1 M HNO3 microemulsion (dashed line) and from 3 M HNO3 microemulsion (solid line).

CO2, it can be recalculated that each condition has an effect on the membrane excluding the effect of pure CO2: PE6100 ) 9 × 1012 m-1, NP-4 ) 4 × 1012 m-1, and NP-4 + H2O ) 5 × 1012 m-1. Effect of Flow Rate in Surfactant Permeation. Figure 8 shows the kinetic behavior of the permeation effect as a function of flow rate. For this, a triblock copolymer from the Pluronics family PE6100 (MW ) 1650 g/mol), which has a molecular

weight larger than the MWCO (1 kDa ) 1000 g/mol) of the membrane, was used. After mixing with PE6100 and CO2 in the reactor for 20 min, 270 mL of CO2 was fed at a constant flow rate. In a fixed quantity (270 mL) interval, the permeated surfactants were collected in the separator and were measured by the microbalance. Throughout each experiment, 1080 mL of CO2 was flushed totally. As mentioned above, in general, PE6100 cannot pass through the membrane since PE6100 has a greater molecular weight (1600 g/mol) than the MWCO (1000 g/mol) of the membrane. However, as seen from Figure 8, the surfactant permeated clearly as the flow rate increased. At the low (2 mL/min) flow rate, the kinetic effect was not indicated due to the adsorption of the surfactant into the membrane. However, about 15% of the surfactant permeated through the membrane. This is due to the polymer surfactants rarely having the exact molecular weight; there is always a distribution around an average value. Moreover, the surfactant can be folded and it is thus possible that it can enter the pores as has already been observed in the case of surfactant nanofiltration in water.32 In the case of PE6100, this is a highly probable situation, since the surfactant molecule comprises two hydrophilic ends that absorb preferentially onto the membrane, whereas only the hydrophobic block located in the middle of the molecule extends inside the sc-CO2. It can be also observed that the permeation steeply increased for flow rate over 10 mL/min; for transmembrane pressure over 0.04 MPa, a rather constant value of permeation amount is reached. This suggested that equilibrium occurs at the membrane surface between adsorption forces and permeation mechanical forces. Consequently, such a result implies that the flow rate and transmembrane pressure are very important factors that govern the separation rate of surfactant material at the membrane surface. Effect of Transmembrane Pressure for Microemulsion Permeation. Water-in-CO2 microemulsions were permeated using the 1 kDa membrane. As mentioned above, microemulsions having contaminants should be recovered to reuse the

5412

Ind. Eng. Chem. Res., Vol. 48, No. 11, 2009

Figure 12. Reaction of nitric acid with the primary alcohol of NP-4 to form the nitrate (1) and the carboxylic acid (2) as byproducts.

surfactants as well as the CO2. However, the current techniques have many drawbacks. Therefore, we investigated nanofiltration using an inorganic membrane in sc-CO2. That is, the water core with dissolved solute (such as contaminants) could be separated from microemulsions using nanofiltration by means of the sieving mechanism. Free surfactants flowing through the membrane can be recycled. Nonylphenol-4-polyethoxylated (MW ) 396 g/mol) was used to form microemulsions. The flux variations of pure CO2, CO2/NP-4, and microemulsions (W ) 20, 30) are presented in Figure 9. At low transmembrane pressure, ∆P ) 0.1 MPa (Figure 9A), no considerable difference was observed. However, as seen in the range of the transmembrane pressure from 0.4 to 0.7 MPa (Figure 9B), flux decline in the case of CO2/NP-4 was observed. That is, as soon as the microemulsions arrive and adsorb into the membrane, the hydrophilic polyethoxylated part adsorbs onto the membrane and pore surfaces, thus increasing the total resistance of the membrane. The phenomenon increases in the case of microemulsions and with the water loaded. Addressing the microemulsions case, micelles reach and adsorb onto the membrane, and the surfactants then spread along the pore walls. Spreading carried the water into the pores and simultaneously the pressure gradient pushed the solution into the filter pores with the flux to the pore walls. In Figure 9B the process is dominated by the setup of an equilibrium between adsorption and hydrodynamic forces. Above 0.7 MPa of transmembrane pressure (Figure 9C), the value of flux was increased suddenly and became the same as NP-4 because of the intense transmembrane pressure and flow rate. In that case, hydrodynamic forces overcame the adsorption force. Ultimately, the water core permeated though the membrane directly without any adsorption. Experimental observations are qualitatively similar to the case of the permeation of a single surfactant, suggesting a rather complex mechanism involving adsorption behavior balanced by hydrodynamic forces at the membrane surface. These preliminary results also suggest that a much lower cutoff should be used for membranes to avoid surfactant penetration in pores. However, inorganic membranes having cutoffs far below 1 kDa is still a highly challenging issue. Nanofiltration of Contaminated Microemulsions for Continuous Process. In order to implement a continuous process under high pressure, an experiment was conducted to separate and recover contaminated microemulsions using membranes. Cobalt metal ions were used as the contaminant, and microemulsions were created using NP-4. Cobalt ions in the separator, reactor, and membrane were analyzed using ICP to obtain the contaminant separation efficiency using the membrane. The surfactant for recovery and reuse was confirmed using a microbalance. As indicated in Table 2, hardly any surfactant remained in the membrane after membrane separation, and the separator displayed a high recovery rate of 92%. Contrary to expectation, only 47% of cobalt metal ions were separated by the membrane and 40% passed through the membrane were collected at the separator. As distinct from the suggestion of Figure 1, the contaminant was not separated

clearly. One possible reason is that trapped Co2+ passed through the membrane with water. Therefore, a more detailed study to trap the contaminant dissolved in solution must be performed. Nanofiltration of Contaminated Dispersions for Continuous Process. Dispersion does not use water, and the surfactant directly forms a bond with the contaminant. CeO2 and PE6100 were used for the contaminant and the surfactant, respectively. 5 kDa and 300 kDa membranes were used for separation. After dispersions were created, the solution was passed through the membranes to separate PE6100 and CeO2. The surfaces of the membranes used for the experiment were compared and analyzed using SEM images before and after the experiment. With a 5 kDa membrane, CeO2 was not impregnated on the membrane surface and some surfactant was adsorbed, as shown in Figure 10. Conversely, it was confirmed that CeO2 was impregnated on the membrane surface after separation when a 300 kDa membrane was used. There was not a significant difference in terms of transmembrane pressure during separation in the 5 kDa membrane, but a pressure difference of 0.5 MPa or greater was indicated for the 300 kDa membrane. The pressure difference can be explained by the fact that CeO2, with its size of 20-30 nm, was not collected on the surface of the 5 kDa membrane due to the pore size of 2 nm, but CeO2 was collected on the 300 kDa membrane with a pore size of about 50 nm. Accordingly, CeO2 lumps were found inside the 5 kDa membrane after the experiment. The result confirmed that the processing size of the membrane must be smaller than the size of the contaminant. Spectroscopy Studies of Used Surfactants for Reuse. In the case of microemulsions, small amounts of acidic solutions should be used to dissolve the contaminating metal oxides. Surfactants recovered from microemulsions containing different concentrations of HNO3 were analyzed by IR spectroscopy and compared to the initial NP-4. The objective is to determine if NP-4 is damaged by aqueous nitric acid microemulsions. No modification of the IR spectra is observed for the products collected from the microemulsions formed with pure water, 0.1 M nitric acid solution, or 0.5 M nitric acid solution. In contrast, the spectra of the surfactants collected from 1 and 3 M nitric acid microemulsions show additional bands (Figure 11). The band at 1660 cm-1 is present in both spectra and is assigned to the stretching vibration frequency of a nitrate group (R-ONO2), ν(NO2). The band is more intense for the product collected from the 3 M microemulsion. The absorption band at about 1730 cm-1 in the IR spectra of the product collected from the 3 M microemulsion is assigned to the stretching band ν(CdO) of a carbonyl group. In this region, it can correspond either to an aldehyde, a ketone, an ester, or a carboxylic acid. However, 13 C NMR (in D2O) of the product collected from the 3 M microemulsion does not show any peak above 163.5 ppm. Since carbonyls of aldehyde and ketone absorb in the range 190-220 ppm, these two possibilities can be excluded. Furthermore, there is no additional band in the region 1100-1300 cm-1 where the asymmetric stretching vibration band of the C-O bond of an ester is expected. Therefore, it is deduced that the band at 1730

Ind. Eng. Chem. Res., Vol. 48, No. 11, 2009 -1

cm is due to a carboxylic acid. In addition, this value of 1730 cm-1 agrees well with the value of 1729 cm-1 observed for the stretching band ν(CdO) of glycolic acid where the chemical environment of the carboxylic acid group is similar. It is known that primary monohydric alcohols are prone to oxidation when treated with 30% v/v and higher concentrated HNO3 solutions,33 and it is generally assumed that the primary product of this reaction is a nitric acid ester.34,35 The formation of the oxidation product (carboxylic acid with the same number of carbon atoms) depends on the chemical stability of the nitrate. In addition, a large enhancement of the oxidation ability of dilute nitric acid was observed in reverse micelle systems.36 Thus NP-4 can react with nitric acid microemulsions at 40 °C to give a nitric acid ester 1 and a carboxylic acid 2 (Figure 12). However, we have observed that noticeable oxidation of NP-4 only occurred with the most concentrated 3 M microemulsion. Therefore, NP-4 is not significantly modified unless concentrated 3 M nitric acid microemulsion is used. According to a previous study, acid-CO2 microemulsions with 1 M HNO3 can efficiently remove the contaminants from the substrate.37 It can be concluded that NP-4 used in acid-CO2 microemulsions could be reused for decontamination. Conclusions This study was achieved successfully in developing a continuous process of sc-CO2 using an inorganic membrane. A new method of separation was deployed using an inorganic membrane to reuse sc-CO2 and additives. The permeation of sc-CO2 was measured and analyzed in terms of the inorganic membrane’s material, pore size, and geometry. Based on the analyzed fundamental data, studies were conducted regarding microemulsions and dispersions. In the case of microemulsions, metal ions were displaced and passed through the membrane as water adsorbed on the membrane surface, resulting in 47% metal ion separation efficiency and 90% surfactant recovery rate. Conversely, dispersions yielded a high separation efficiency, which was confirmed through SEM images. The results of this study confirmed the possibility of a continuous process of supercritical CO2 and are expected to provide fundamental data for process development. Acknowledgment This work was supported by the Korea Research Foundation Grant funded by the Korean Government (MOEHRD) (KRF-2006-DOO211). Literature Cited (1) Wai, C. M.; Wang, S. Supercritical Fluid Extraction: Metal as ComplexessReview. J. Chromatogr., A 1997, 785, 369. (2) Smart, N. G.; Carleson, T. E.; Elshani, S.; Wang, S.; Wai, C. M. Extraction of Toxic Heavy Metals Using Supercritical Fluid Carbon Dioxide Containing Organophosphorus Reagents. Ind. Eng. Chem. Res. 1997, 36, 1819. (3) Koh, M. S.; Park, K. H.; Yang, D. H.; Kim, H. K.; Kim, H. D. The Synergistic Effect of Organophosphorus and Dithocarbamate Ligands on Metal Extraction in Supercritical CO2. Bull. Korean Chem. Soc. 2005, 26, 423. (4) Li, J.; Beckman, E. J. Affinity Extraction into CO2. 2. Extraction of Heavy Metals into CO2 from Low-pH Aqueous Solutions. Ind. Eng. Chem. Res. 1998, 37, 4768. (5) Marsal, A.; Celma, P. J.; Cot, J.; Cequier, M. Supercritical CO2 Extraction as a Clean Degreasing Process in the Leather Industry. J. Supercrit. Fluids 2000, 16, 217. (6) Jones, C. A., III.; Zweber, A.; DeYoung, J. P.; McClain, J. B.; Carbonell, R.; DeSimone, J. M. Applications of “dry” Processing in

5413

Themicroelectronics Industry using Carbon Dioxide. Crit. ReV. Solid State Mater. Sci. 2004, 29, 97. (7) Wang, J. S.; Koh, M. S.; Wai, C. M. Nuclear Laundry using Supercritical Fluid Solutions. Ind. Eng. Chem. Res. 2004, 43, 1580. (8) Murzin, A. A.; Babain, V. A.; Yu, A.; Shadin, A.; Smirnov, I. V.; Lumpov, A. A.; Gorshkov, N. I.; Miroslavov, A. E.; Muradymov, M. Z. Supercritical Fluid Extraction of Actinide Complexes: II. SFE of Actinide β-Diketonates. Radiochemistry 2002, 44, 467. (9) Laintz, K. E.; Tachlkawa, E. Extraction of Lanthanides from Acidic Solution using Tributyl Phosphate Modified Supercritical Carbon Dioxide. Anal. Chem. 1994, 66, 2190. (10) Koh, M. S.; Yoo, J. Y.; Park, Y.; Bae, D. I.; Park, K. H.; Kim, H. K.; Kim, H. D. Supercritical CO2 Extraction of Uranium(VI) from HNO3 Solution using N,N,N′,N′-Tetrabutyl-3-oxapentanediamide. Ind. Eng. Chem. Res. 2006, 45, 5308. (11) Meguro, Y.; Iso, S.; Yoshida, Z.; Tomika, O.; Enokida, Y.; Yamamoto, I. Decontamination of Uranium Oxides from Solid Wastes by Supercritical CO2 Fluid Leaching Method using HNO3-TBP Complex as a Reactant. J. Supercrit. Fluids 2004, 31, 141. (12) Lin, Y.; Smart, N. G.; Wai, C. M. Supercritical Fluid Extraction of Uranium and Thorium from Nitric Acid Solutions with Organophosphorous Reagents. EnViron. Sci. Technol. 1995, 29, 2706. (13) Lin, Y.; Wai, C. M. Supercritical Fluid Extraction of Lanthanides with Fluorinated β-Diketones and Tributyl Phosphate. Anal. Chem. 1994, 66, 1971. (14) King, J. W.; Williams, L. L. Utilization of Critical Fluids in Processing Semiconductors and Their Related Materials. Curr. Opin. Solid State Mater. Sci. 2003, 7, 413. (15) Chen, J. L.; Wang, Y. S.; Kuo, H.; Shu, D. Y. Stripping of Photoresist on Silicon Wafer by CO2 Supercritical Fluid. Talanta 2006, 70, 414. (16) Lack, E.; Seidlitz, H.; King, M. B.; Bott, T. R. Extraction of Natural Products using Near-Critical SolVents; Blackie Academic & Professional: London, 1993; Chapter 5. (17) Subra, P.; Vega-Bancel, A.; Reverchon, E. Breakthrough Curves and Adsorption Isotherms of Terpene Mixtures in Supercritical Carbon Dioxide. J. Supercrit. Fluids 1998, 12, 43. (18) Brunner, G.; Johannsen, M. New Aspects on Adsorption from Supercritical Fluid Phases. J. Supercrit. Fluids 2006, 38, 181. (19) Semenova, S. I.; Ohya, H.; Higashijima, T.; Negishi, Y. Dependence of Permeability through Polyimide Membranes on state of Gas, Vapor, liquid and Supercritical Fluid at High Temperature. J. Membr. Sci. 1992, 67, 29. (20) Hsu, J. H.; Tan, C. S. Separation of Ethanol from Aqueous Solution by a Method Incorporating Supercritical CO2 with Reverse Osmosis. J. Membr. Sci. 1993, 81, 273. (21) Birtigh, A.; Brunner, G. Regeneration of Loaded Supercritical Fluids. J. Supercrit. Fluids 1996, 9, 227. (22) Verkerk, A. W.; Goetheer, E. L. V.; van den Broeke, L. J. P.; Keurentjes, J. T. F. Permeation of Carbon Dioxide through a Microporous Silica Membrane at Subcritical and Supercritical Conditions. Langmuir 2002, 18, 6807. (23) Sarrade, S.; Rios, G. M.; Carles, M. Nanofiltration Membrane behavior in a Supercritical Medium. J. Membr. Sci. 1996, 114, 81. (24) Chiu, Y. W.; Tan, C. S. Regeneration of Supercritical Carbon Dioxide by Membrane at Near Critical Conditions. J. Supercrit. Fluids 2001, 21, 81. (25) Yonker, C. R.; Fulton, J. L.; Phelps, M. R.; Bowman, L. E. Membrane Separations using Reverse Micelles in Nearcritical and Supercritical Fluid Solvents. J. Supercrit. Fluids 2003, 25, 225. (26) Patil, V. E.; Meeuwissen, J.; Van Den Broeke, L. J. P.; Keurentjes, J. T. F. Permeation of Supercritical Fluids across Polymeric and Inorganic Membranes. J. Supercrit. Fluids 2006, 37, 367. (27) Sarrade, S. J.; Rios, G. M.; Carles, M. Supercritical CO2 Extraction Coupled with Nanofiltration Separation Applications to Natural Products. Sep. Purif. Technol. 1998, 14, 19. (28) Gourgouillon, D.; Schrive, L.; Sarrade, S.; Rios, G. M. An Envrionmentally Friendly Process for the Regeneration of Used Oils. EnViron. Sci. Technol. 2000, 34, 3469. (29) Rios, G. M.; Joulie, R.; Sarrade, S. J.; Carles, M. Investigation of Ion Separation by Microporous Nanofiltration Membranes. AIChE J. 1996, 42, 2521. (30) Christenson, H. K.; Blorm, C. E. Solvation Forces and Phase Separation of Water in a Thin Film of Nonpolar Liquid between Mica Surfaces. J. Chem. Phys. 1987, 86, 419. (31) Jo¨nsson, A. S.; Jo¨nsson, B. The Influence of Nonionic and Ionic Surfactants on Hydrophobic and Hydrophilic Ultrafiltration Membranes. J. Membr. Sci. 1991, 56, 49. (32) Cornelis, G.; Boussu, K.; Van der Bruggen, B.; Devreese, I.; Vandecasteele, C. Nanofiltration of Nonionic Surfactants: Effect of the

5414

Ind. Eng. Chem. Res., Vol. 48, No. 11, 2009

molecular weight cut-off and contact angle of flux behavior. Ind. Eng. Chem. Res. 2005, 44, 7652. (33) Svetlakov, N. V.; Nikitin, V. G.; Nikolaeva, E. A. Oxidation with Nitric Acid of Aliphatic Alcohols and Diols to Carboxylic Acids. Russ. J. Org. Chem. 2007, 43, 773. (34) Desseigne, G. Preparation of Nitric Esters of Aliphatic Primary Alcohols. Mem. Poudres 1957, 39, 147. (35) Brunnberg, B. (Nitroglycerin Aktiebolaget). Nitric Acid Esters of Aliphatic Alcohols. U.S. Patent 3009944, 1961. (36) Hojo, M.; Ueda, T.; Daike, C.; Takezaki, F.; Furuya, Y.; Miyamoto, K.; Narutaki, A.; Kato, R. Great Enhancement in the Oxidation Ability of

Dilute Nitric Acid in Nanoscale Water-Droplets of Reverse Micelle Systems. Bull. Chem. Soc. Jpn. 2006, 79, 1215. (37) Koh, M. S.; Yoo, J. Y.; Ju, M. S.; Joo, B. K.; Park, K. H.; Kim, H. W.; Kim, H. D.; Fournel, B. Surface Decontamination of Radioactive Metal Wastes using Acid-in-Supercritical CO2 Emulsions. Ind. Eng. Chem. Res. 2008, 47, 278.

ReceiVed for reView March 19, 2008 ReVised manuscript receiVed March 2, 2009 Accepted March 26, 2009 IE800452G