- Email: [email protected]
Novel silica membranes for high temperature gas separations
Journal of Membrane Science 371 (2011) 254–262
Contents lists available at ScienceDirect
Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci
Novel silica membranes for high temperature gas separations Neha Bighane, William J. Koros ∗ School of Chemical and Biomolecular Engineering, Georgia Institute of Technology, 778 Atlantic Drive, Atlanta, GA 30332-0100, USA
a r t i c l e
i n f o
a b s t r a c t
Article history: Received 20 October 2010 Received in revised form 22 January 2011 Accepted 25 January 2011 Available online 1 February 2011 Keywords: Silica Membrane Oxidative thermolysis
This article describes fabrication of novel silica membranes derived via controlled oxidative thermolysis of polydimethylsiloxane and their gas separation performance. The optimized protocol for fabrication of the silica membranes is described and pure gas separation performance in the temperature range 35–80 ◦ C is presented. It is observed that the membranes exhibit activated transport for small gas penetrants such as He, H2 and CO2 . The membranes can withstand temperatures up to 350 ◦ C in air and may ultimately find use in H2 /CO2 separations to improve efficiency in the water-gas shift reactor process. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Membrane separation processes can provide low cost improvements in thermally driven processes such as production of hydrogen from steam reforming of natural gas. This dominant H2 production process comprises a highly endothermic steam methane reforming reaction (∼800 ◦ C), a pair of water gas shift reactors (high temperature shift (HTS) at 350 ◦ C and low temperature shift (LTS) at 200 ◦ C) and downstream by-product CO2 absorption units [1,2]. The water gas shift reaction (Eq. (1)) is thermodynamically limited and can be driven forward by selective removal of product hydrogen by membrane separation. CO + H2 O ↔ CO2 + H2
H = −41.2 kJ/mol
(1)
Installation of a hydrogen selective membrane unit between the two water gas shift reactors can provide increased efficiency of the overall production process in the following ways: 1. Production of a high purity hydrogen stream from the membrane unit. 2. Higher conversion in the LTS, due to lower concentration of products in the equilibrium mixture entering LTS. 3. Higher rate of reaction in LTS, implying a smaller volume of reactor. 4. Excess reactant steam, that is currently used to drive the reaction forward, can be minimized.
∗ Corresponding author. Tel.: +1 404 385 2845; fax: +1 404 385 2683. E-mail addresses: [email protected], [email protected], [email protected] (W.J. Koros). 0376-7388/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2011.01.045
5. Product stream from LTS has been calculated to become CO2 rich, as opposed to H2 rich stream in the absence of membrane separation, and downstream CO2 capture becomes more efficient. H2 and CO2 are the most abundant species in the equilibrium mixture exiting from the HTS, where the installation of the membrane unit is proposed. Typically, the product stream from a HTS operating at 350 ◦ C comprises 73.9% H2 , 17.7% CO2 , 7.4% CH4 and 1.0% CO (dry basis) [1]. In addition to the gases, steam is also a major component in the reaction mixture. Hence, it is necessary to seek the development of economical hydrothermally stable membranes that can provide high H2 /CO2 selectivity at temperatures similar to the exit temperature of the HTS (300–350 ◦ C). Characterization of membrane materials for gas separations includes testing them for permeability (productivity) and selectivity (efficiency) toward the gases. Inorganic membranes, in particular silica membranes, have been well-recognized for their gas separation abilities at high temperatures. The efficiency of such membranes is well-known and there are a number of publications on this subject [1–5]. However, the complexity of the conventional membrane fabrication processes for silica membranes led the authors to devise a more efficient and more economically scalable alternative fabrication approach. There are two well-known techniques of preparing dense silica membranes, namely, the sol–gel technique [4] and the Chemical Vapor Deposition (CVD) technique [6,7]. Both of these techniques require high-quality supports and CVD requires high cost deposition reactors. While the sol–gel technique involves coating of liquid precursors on a support and subsequent heat treatment, CVD involves deposition of gaseous phase species on a support. A third technique of melt extrusion and leaching to make hollow fiber silica membranes (PPG Industries) was patented by Ham-
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
255
Table 1 Major research found in literature, on permeation through silica membranes. S.no.
Author(s)
Temperature in fabrication (◦ C)
Testing temperature (◦ C)
1
Nijmeijer (Thesis, University of Twente) [2]
2 3
J.D.Way et al. [5] Renate de Vos et al. [4]
600 625 625 – 600
200 400 600 70 300
mel et al. [8] and their gas transport properties were studied by many groups [5]. A fourth potentially low cost technique of making tubular silicon based membranes from a two-step calcination of silicone rubber tubes was explored by Lee and Khang [9]; however, the resultant membranes could not yield selectivities above the Knudsen range. Table 1 presents a brief outline of some major research developments in this field for membranes prepared by the conventional methods, which do provide performance better than the Knudsen values. In state-of-the-art sol–gel microporous silica membranes, it has been observed that the permeability of all penetrants, except CO2 , increase with increase in temperature. CO2 permeability decreases due to a decrease in its sorption (P = D·S) in the membrane with rise in temperature. High selectivities of small penetrants like He and H2 are obtained and gas penetrants larger ˚ are completely sieved out i.e., cannot than CH4 , like SF6 (5.5 A) permeate through the membrane. In the present work, silica membranes have been fabricated via a new approach involving the controlled oxidative thermolysis of a precursor rubbery film of polydimethylsiloxane to create ceramic microporous flat film silica membranes that also show performance above Knudsen selectivity. 2. Membrane fabrication The concept of fabrication pursued here is based on careful optimization of the formation and thermal oxidation of a polysiloxane, viz.,
R1 Si
O
n
R2 to silica (SiO2 ), while retaining the mechanical integrity of the final film. Polydimethylsiloxane is a liquid resin at room temperature. This resin is subject to chemical crosslinking using tetraethoxysilane (Si(OC2 H5 )4 ). The crosslinked precursor is thermally oxidized. The thermal oxidation step eliminates the organic side groups and stitches the polymeric chains together via oxygen, resulting in silica. The thermo-oxidative stability of polysiloxanes [10] varies with the side groups as C6 H5 > ClC6 H4 > Cl3 C6 H2 > Cl2 C6 H3 > CH2 = CH > CH3 The precursor polymer was chosen as polydimethylsiloxane or PDMS (R1 = R2 = CH3 ) due to (i) its low cost, (ii) availability of literature since it is the most widely studied polysiloxane and (iii) because the methyl group is more easily oxidized as compared to other side groups. This technique enables fabrication and testing of unsupported silica membranes (flat films) because the precursor is a solid rubbery film. This fabrication technique also holds the potential to enable its transformation into hollow fibers, which is the preferred membrane structure in industrial applications. Moreover, this approach allows the possibility of crosslinking PDMS with transition metal alkoxides to obtain silica–metal oxide membranes. This latter capability is particularly attractive, since it is reported
H2 /CO2 selectivity 45 17 21 5.36 70
that silica–metal oxide materials are more stable in hydrothermal environments than pure silica matrices [17–20]. The current study does not focus on this important extension, but this is envisioned for subsequent studies after the basic technique has been demonstrated. As will be discussed, the fabrication parameters can be varied to tune the micromorphology of the silica membranes, and this is envisioned to provide considerable flexibility in creating tailored morphology for various applications. 2.1. Casting crosslinked precursor PDMS film A precursor film of PDMS is cast using ‘condensation cure’ chemistry. Silanol terminated polydimethylsiloxane was purchased from Gelest Inc. (DMS-S45, Mw = 80,000–130,000) and it is a viscous liquid at ambient conditions. A solution of the polymeric resin in n-heptane is thoroughly mixed with an appropriate crosslinking agent, tetraethoxysilane (Si(OC2 H5 )4 ) (TEOS). A 5:1 weight ratio of PDMS to Si(OC2 H5 )4 was found optimal, and 12–15 mg of tin (din-butylacetoxytin 95%) and titanium (titanium-2-ethylhexoxide) catalysts, per gram polymer, were used in the current study. The PDMS:TEOS ratio was found to be optimum based on careful studies of the ratio, with the preferred value showing the best mechanical properties in the final silica films after the oxidative thermolysis process. A 2:1 (PDMS:TEOS) ratio resulted in minute clusters of silica particles within the precursor films, due to self crosslinking of TEOS. This created inhomogeneity and caused the precursor to crumble upon oxidative thermolysis. Precursor films that deviate by more than 3% from the 5:1 value result in either substructure cracks or crumbling or breaking of the films upon oxidative thermolysis. Hence, the value 5:1 was chosen as optimal and precursors fabricated with this value yielded defect free silica membranes upon thermal oxidation. The prepared solution was poured into a circular teflon® dish (solution casting), maintained in a sealed glove bag filled with N2 and water vapor. Moisture accelerates the crosslinking rate of the condensation cure reaction. Hence, a controlled humidity level between 70 and 85% RH (Omega HH311 humidity meter) at 20–22 ◦ C was maintained in the glove bag. When the solution is exposed to this atmosphere, the silanol ends of the PDMS chains undergo condensation with the ethoxy groups of Si(OC2 H5 )4 (Fig. 1). The polymer begins to crosslink to form a film in 22–30 min. The system is allowed to equilibrate for 21 h to yield a transparent rubbery film. The film is then peeled off the dish and dried in a vacuum oven at 100 ◦ C for 20 h and 120 ◦ C for 11 h to ensure complete removal of solvent, by-product and volatile materials. The obtained PDMS films are typically 7.9–8.4 mil thick (1 mil = 1/1000 inch). While the solution can crosslink even without humidity, it takes a very long time (∼1.5–2 h at 9% RH). A long crosslinking time is not favorable when scale up is desired. Humidity increases the crosslink density of the precursor and the silica membrane obtained after oxidative thermolysis would have a denser pore structure. Silica membranes formed from a no-moisture crosslinked PDMS would have large pores and separation performance would be equal to Knudsen values. Hence, a humidity level of 70–85% RH was used that results in a crosslinking time of 22–30 min.
256
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
Crosslinking PDMS (Condensation cure with alkoxy crosslinker) CH 3
OC 2 H 5 C2H5O
Si
OC 2 H 5
+
HO
Si
n
CH 3
OC 2 H 5
OC 2 H 5
CH 3 O
Si
OH
+
C2H5 O
CH 3
Tetraethoxysilane ( TEOS)
Si
OC 2 H 5
OC 2 H 5
Sn, Ti catalysts Moisture
Reacon involves simultaneous hydrolysis of TEOS and condensaon with PDMS chains
O O
Si
O
CH 3 O
O
Si
O
CH 3
n
Si
O
+ by-product C 2 H 5 OH
O
(3-d crosslinked network) Fig. 1. Schematic representation of TEOS crosslinking PDMS resin to obtain a rubbery PDMS film.
2.2. Oxidative thermolysis From the PDMS film obtained in the previous step, discs were cut and their weight, thickness and area (Scion ImageTM ) measured. These films were placed on a steel wire mesh inside a long quartz tube in a tubular furnace. The tube was sealed with an oxygen purge (99% O2 ) at a desired flow rate (discussed later in Section 4.1). The system is heated from room temperature to 390 ◦ C according to an optimized protocol (Table 2). The rubbery film undergoes oxidative thermolysis to yield a flat transparent ceramic silica membrane. At the end of the heating stage, the oxygen flow is stopped and the system cooled at 1 ◦ C/min. The accurate continuous flow of oxygen during the heating phase is important to complete the oxidation process as well as to purge out decomposition products. An initial ramp rate 2 ◦ C/min to soak at 250 ◦ C is used as the precursor film does not undergo any modification until this temperature. While the thermal program given in Table 2 was found to be optimum, some flexibility in the protocol may be tolerated; however, these conditions are recommended to prevent curling, which complicates permeation testing. During the thermal oxidation stage, the methyl groups in the PDMS film undergo homolytic cleavage by gaseous phase O2 and the siloxane free radicals from different chains are bridged via oxygen to form a three-dimensional silica matrix. Internal rearrangement and shrinkage occurs. A schematic of the hypothesized mechanism [16] is presented in Fig. 2. Based on the mechanism shown in Fig. 2, the expected total weight loss in the thermal oxidation step can be calculated. Specifi-
Table 2 Optimal temperature-time protocol for oxidative thermolysis of crosslinked PDMS films to silica membranes. Process point Starting set point Set point 1 Time 1 Set point2 Time 2 Set point 3 Time 3 Set point 4 Time 4 Set point 5 Time 5
Process point value
Rate
cally, removal of two methyl groups and addition of two half oxygen atoms (O shared by two Si) per unit PDMS should show an expected weight loss equal to =
(−(2 × 15) + (2 × 0.5 × 16)) × 100 = −18.92% 74
The experimentally observed weight loss of >20 samples was found to be 19 ± 2%, which is in good agreement with the theoretical value. On an average, approximately 23% increase in the density of the material is observed upon oxidative thermolysis. Further support for the claim that the weight loss is due to the methyl groups is offered later via FTIR evidence. In an inert atmosphere such as He, a black residue of SiOC results, rather than the clear pure silica found in our studies. Since only CH3 groups are burnt out in this method, it creates only CO2 and H2 O, simplifying the control of the pores created during the oxidative burn out process. Table 3 summarizes the key parameters found to affect the quality of the final membranes. As noted above, optimal values of these parameters were obtained empirically to achieve defect free flat silica membrane films. O2 flow rate during the oxidative thermolysis stage has been observed to most markedly result in changes in the gas transport properties of the membranes and to allow tuning the micromorphology of the membranes. 3. Material characterization 3.1. Thermogravimetric analysis Thermogravimetric analysis (TGA) in an air purge was employed to visualize the weight loss pattern during the oxidative thermolysis stage of fabrication, as crosslinked PDMS oxidizes to silica. An open pan TGA was used to ensure exposure to the oxidizing atmosphere. Air was used instead of oxygen due to safety concerns for the test equipment. A crosslinked PDMS film was subjected to heat
◦
23 C 250 ◦ C 1 h 53 min 250 ◦ C 5 min 390 ◦ C 1 h 56 min 390 ◦ C 15 min 25 ◦ C 6 h 5 min
2 ◦ C/min 0 ◦ C/min (hold) ◦
1.2 C/min ◦
0 C/min (hold) 1 ◦ C/min
Table 3 Summary of key process parameters. Parameter
Optimal value
PDMS:TEOS weight ratio Humidity level (precursor casting) Ramp rate Soak temperature O2 flow rate (cm3 /min)
5:1 70–85% RH 1.2 ◦ C/min 390 ◦ C Variable – 30, 50
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
257
CH 3 ------------ O
+
Si ----------
O-O
CH 3
CH 2 OOH ------------ O
Si -----------CH 3
CO ------ O
Si -------
+
OH
+
H2 O
HCOOH
CH 2 O
CH 3
+
O2 CO 2
+
H
2
OH ------ O
Si -----CH 3
Two chains link via an O atom, after the removal of a methyl group from each. CH 3 ----- O
Si ------
CH 3 ----- O
OH
Si -------O
+
O2 H2 O
SiO 2
OH ----- O
Si ------
----- O
Si -------CH 3
CH 3
- Petar R. Dvornic, High Temperature Stability of Polysiloxanes, Silicon Compounds: Silanes and Silicones, Gelest, Inc. Fig. 2. Oxidative thermolysis of PDMS to silica.
up to 500 ◦ C in the air purge (Fig. 3a). The thermal protocol followed was the same as presented in Table 2, up to 390 ◦ C. The ramp rate from 390 ◦ C to 500 ◦ C was also 1.2 ◦ C/min. The weight loss pattern obtained was an indicator of the conversion. A two-step weight loss pattern was observed. The first weight loss step begins from ∼250 ◦ C and stabilizes at 390 ◦ C. Further heating to 425 ◦ C disintegrates the film into particles. From this observed weight loss pattern, it is clear that the conversion of PDMS to silica is complete by 390 ◦ C as the weight loss reached the
theoretical value of ∼19% and the film was intact until this temperature. The same was observed in the actual fabrication process as discussed in Section 2.2. Until 390 ◦ C, elimination of only methyl groups occurs and polymeric chains get connected via oxygen to form a silica film. Beyond 425 ◦ C, breakdown of the Si–O bonds and redistribution (i.e., splitting and reforming) occurs to form volatile oligomers and a residue of fine silica particles results [10]. TGA was also used to verify the thermal stability of the silica films. The films are thermally stable up to 390 ◦ C in air (dry oxidiz-
258
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
Fig. 3. (a) TGA of crosslinked PDMS film in air. (b) TGA of silica film in air showing stability at 390 ◦ C in a 2 h test.
ing condition) (Fig. 3b). Consequently, the membranes can easily withstand a temperature of 300–350 ◦ C, as desired for application in the water gas shift process. 3.2. Fourier transform infra-red spectroscopy Fourier transform infrared spectroscopy was employed to investigate the chemical structure of the precursor PDMS films and final silica membranes. The obtained absorbance spectra are shown in Fig. 4a and b. The large reduction in the methyl peak (2970 cm−1 ) and the evolution of the broad shoulder between 1000 and 1250 cm−1 that is a signature of amorphous silica, in Fig. 4b relative to Fig. 4a demonstrate the transformation of PDMS to silica via the thermal oxidation process. 4. Gas permeation measurements Gas permeabilities of the silica membranes have been measured over a temperature range of 35–80 ◦ C, with six penetrants. Work is in progress on measurements at higher temperatures up to 300 ◦ C and will be presented in a later publication, after further optimizing the process demonstrated here, to create high perfor-
mance size discriminating silica membranes from the oxidative thermolysis of PDMS. This article focuses only on the first proof-of principle of the oxidative thermolysis of PDMS to make molecularly selective membranes that exceed Knudsen performance levels. 4.1. Permeation system The gas permeation testing system is primarily engineered into three sections: • Sections 1 and 3 always remain at ambient temperature. Section 1 comprises a gas feed line, upstream transducer (Sensotec Model Z, absolute 0–1000 psia), feed valve and vent line. • Section 2 comprises an upstream volume and the permeation cell. Section 2 is maintained at the desired operating temperature via a heating mechanism. A heavy insulated heating tape, controlled by PID tuned benchtop controller (Omega® ), is employed for controlling the temperature of this section. Section 2 is connected to section 3. • Section 3 has a downstream volume with downstream transducer (Baratron, 0–10 Torr). The temperature of the downstream is measured and used for permeability calculations.
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
259
Fig. 4. (a) FTIR of crosslinked PDMS precursor thin film. (b) FTIR of silica membrane (KBr pellet technique).
The upstream and downstream volumes are connected to a vacuum pump (BOC Edwards model RV3) for evacuation. All fitting are made of stainless steel 316/316L to ensure safe, stable and leak proof operations. In order to conduct gas permeation tests, it is vital that the seal between the membrane and the test cell be leak proof. While the seals used in the present work have high integrity for measurements up to 80 ◦ C, difficulties in sealing above this temperature must be resolved and will be discussed in subsequent works for high temperature operation. 4.2. Principle of operation The membrane is installed into the permeation cell using an impermeable aluminum foil tape (Fasson® ) and epoxy (Duralco 4703, Cotronics, Inc.), using a permeation cell whose detailed description is available [21]. After masking the membrane into the permeation cell, the entire system is thoroughly evacuated for 24–48 h to degas the system. A leak rate is measured by monitoring the downstream pressure rise while the system remains evacuated. For inorganic membranes in high temperature operating conditions, the leak rate is usually a negligible fraction of the permeation rate of the slowest penetrant. After evacuation of the entire system, gas is fed into the upstream volume at ambient temperature (feed valve closed) at a desired pressure and the desired temperature of operation established in section 2. The permeation cell and downstream volume
are isolated from the upstream by the feed valve and are maintained under vacuum. Once the feed valve is opened, transport of gas through the membrane occurs at the desired temperature. The permeate starts accumulating in the downstream volume. The rise in pressure in the downstream volume as a function of time dp/dt is recorded and this data is used to calculate the permeability of the gas through the membrane film at the operating temperature and pressure. In each experiment, the rate of rise in downstream pressure in the range 0.2–10 Torr is recorded. From the obtained data, the steady state data is used to calculate the permeability of the penentrant. Steady-state is assumed to be attained after roughly 10 time lag periods [11]. For each experiment, a second run was performed by evacuating the downstream in steady state to verify that the rate of rise in pressure remains the same as in the previous run. No time lag is observed in the second run, since the system is already under steady state conditions. The permeation rate is corrected for leaks by subtracting the leak rate from the measured permeation rate. No time lag was observed for He and H2 while the time lag for larger penetrants decreased with increase in temperature, due to an increase in the diffusivities. Before a different penetrant was studied, the entire system was evacuated for atleast 12 h. The area available for permeation and the thickness of the membrane were measured. In this work, the typical silica membrane area open for permeation was 0.5–1 cm2 and the typical membrane thickness was 7.4–7.9 mils.
260
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
Steady state flux was used to calculate the permeability (in Barrer) similar to the procedure described earlier [21]. 4.3. Theory The intrinsic separation performance of a membrane material is measured in terms of its productivity (permeability) and efficiency (selectivity). The permeability, P, of a gas through a membrane is defined as the flux normalized by the pressure driving force per unit thickness [11]. Pi = Di Si =
(Flux)i pi /l
(2)
D is the diffusivity and S is the sorption coefficient of the gas penetrant ‘i’ in the membrane, pi is the partial pressure difference and l is the thickness of the membrane. The detailed diffusion mechanism in porous membranes is dependent upon the size of the pores relative to the size of the permeating gas and the mean free path of the penetrant. If the pores are large compared to the molecular size of the penetrants but small relative to the bulk gas mean free path under the conditions of measurement, Knudsen diffusion tends to be dominant. For Knudsen transport, the selectivity is low and equal to the inverse ratio of the square root of the molecular weight of the two permeating gases. For example, in the case of O2 and N2 , the smaller but heavier O2 permeates slower than N2 by the factor (28/32)1/2 = 0.94. For non-porous polymeric membranes, a solution-diffusion mechanism prevails. In microporous (pore diameter < 2 nm) inorganic membranes such as the silica membranes considered here, with pores of sizes of the order of the kinetic diameters of penetrants, interaction between the penetrant and the pore walls become significant, especially at low temperatures. Hence, in addition to diffusion through size selective pores, the resultant mechanism becomes sorption–diffusion. In this case, the smaller O2 permeates faster than N2 . The gas transport becomes activated and shows molecular sieve-like separation properties, with high selectivities of He and H2 over larger molecules [2–5,12–15]. In such a case, the process is said to be thermally activated and follows a sorption-diffusion mechanism with permeability given by Eq. (2). As shown in Eq. (2), permeability is fundamentally related to the product of the diffusivity (dependent on molecular size) and the sorption coefficient (dependent on condensability) of the penetrant gas molecule. The unit of permeability commonly used is Barrer, which can also be reported in SI units, viz., 1Barrer = 10−10
cm3 (STP) cm mol m = 3.348 × 10−16 2 cm2 sec cm Hg m s Pa
(3)
Fig. 5. Permeability of silica membranes at 55.2 psia, 35 ◦ C, for He, H2 , CO2 , O2 , N2 and CH4 , fabricated with two different O2 flow rate parameter values.
The activation energy for permeation is the sum of the activation energy for diffusion Ed and the heat of sorption Hs (generally exothermic i.e., Hs is a negative value). As temperature rises, the diffusivity of a penetrant in a membrane increases while its sorption coefficient decreases, causing variation in the permeability according to Eq. (2). As pressure rises, the permeability of a gas penetrant either decreases or remains constant, primarily depending upon the sorption isotherm of the particular penetrant for the membrane. For gases such as O2 and N2 that have similar sorption enthalpies, one can detect the onset of activated permeation when significant deviations from Knudsen selectivity become apparent. Thus, when the smaller but high molecular weight O2 permeates significantly faster than the slightly larger but lower molecular weight N2 , activated sorption–diffusion is indicated. The case of H2 /CO2 is more complex due to the high condensability of CO2 . CO2 can permeate faster than the smaller H2 but as temperature increases, the high diffusivity of H2 starts dominating. These effects are considered in the following section. 5. Gas separation performance 5.1. Gas permeation properties In the present work, it has been observed that variation in the O2 flow rate, used during the fabrication of the silica membranes, results in variations in the gas separation properties. Hence, two different values of O2 flow rate fabrication parameter were investigated.
The ratio of the permeabilites of pure gases (for negligible downstream pressure) used in the experiments here provide a good measure of selectivity or permselectivity (Eq. (4)). ˛i/j =
(yi /yj )Permeate (yi /yj )Feed
DS Pi ∼ = i i = Pj Dj Sj
(4)
In the case of thermally activated permeation, the variation of permeability with temperature for a given feed partial pressure follows an apparent Arrhenius relationship (Eq. (5)) P = P0 exp
−E p
RT
(5)
where P0 is a pre-exponential factor, Ep is the apparent activation energy for permeation, T is the temperature in Kelvin and R is the universal gas constant. Ep = Ed + Hs
(6)
Fig. 6. Permeabilites of He, H2 , CO2 , O2 , N2 and CH4 , for a silica membrane (fabricated at 30 ml/min O2 flow rate) at 35 ◦ C, 55.2 psia and 80 ◦ C, 70 psia.
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
Fig. 7. Permeabilites of He, H2 , CO2 , O2 , N2 and CH4 , for a silica membrane (fabricated at 50 ml/min O2 flow rate) at 35 ◦ C, 55.2 psia and 80 ◦ C, 70 psia.
The permeability data (Figs. 5–7) and obtained selectivity values (Table 4) indicate that the mechanism of gas permeation through the membranes primarily follows a sorption–diffusion mechanism. It is particularly significant that the H2 /CO2 selectivity has “switched” as the temperature rose by only 45 ◦ C. These data show the tendency to shift from a sorption dominated selectivity for the condensable CO2 at 35 ◦ C to a size dominated selectivity for H2 at 80 ◦ C. This trend in anticipated to become more extreme and strongly favor H2 at still higher temperatures. A new permeation test equipment is under construction to cater to gas permeation tests at higher temperatures up to 300 ◦ C and for a flammable gas like H2 . Consistent with results for samples created by sol–gel and other more difficult membrane preparation techniques, the silica membranes appear to possess an interconnected network of pores and the pore size distribution is such that the smaller penetrants can access a larger fraction of the total pores. The average pore size of the majority of pores lies between the kinetic diameters of H2 ˚ and O2 (3.46 A). ˚ The O2 /N2 selectivity is 3 and is reasonable (2.89 A) given that a value between 3 and 8 is considered good for this tough ˚ and N2 (3.64 A) ˚ are very similar in size. The separation as O2 (3.46 A) combination of H2 permeability of 1200 Barrer and H2 /N2 selectivity of 14 lies on its upper bound curve [11] for polymer membranes for this separation. A fairly high CO2 /CH4 selectivity of 13 is also observed. The permeabilities of He, H2 , O2 , N2 and CH4 increase with increase in temperature while that of CO2 decreases. This is indicative of activated nature of permeation with positive activation energies (Ep ) except for CO2 , which has negative activation energy due to its sorption effect. As observed, variation in the O2 flow rate (during fabrication) leads to changes in the gas transport properties of He, H2 and CO2 . It is hypothesized that smaller pores that are more selective to He and H2 are obtained when a higher O2 flow rate is used. Hence, this parameter can be used to tune
261
Fig. 8. Variation of permeability of He, H2 , CO2 and N2 , with pressure (50–100 psia) at 35 ◦ C, of silica membrane fabricated at 50 ml/min O2 .
micro-morphology in the membranes, along with thermal ramp rates in the critical thermolysis zone. The ideal separation factor for H2 /CO2 increases from 0.87 (i.e., CO2 selective) at 35 ◦ C, 55.2 psia to 1.25 (i.e., H2 selective) at 80 ◦ C, 70 psia. The activated nature of the transport through the silica membranes created at 50 ml/min O2 during thermolysis is also apparent in the case of larger gases (O2 , N2 and CH4 ) where the permeabilities in Fig. 7 rise with temperature from 35 ◦ C to 80 ◦ C and the O2 /N2 selectivity in Table 4 exceeds the Knudsen value of 0.94. All these permeation data were reproducible. Fig. 8 shows the variation of gas permeability, through the silica membranes, with pressure from 50 to 100 psia, at 35 ◦ C. While the permeability of CO2 decreased by 11%, the permeabilities of He, H2 and N2 remain relatively constant. These data illustrate the effects of isotherm curvature with pressure. CO2 permeability decreases due to saturation in the sorption capacity of the membrane for the penetrant. It also indicates that even at the low temperature of 35 ◦ C, the size based selectivity of the membranes for H2 can be observed. High temperatures and high pressures favor H2 permeation over CO2 . It has been observed in sol–gel silica membranes [4] that increase in pressure causes small decrease in H2 /CO2 when a mixed gas feed is used, due to competitive sorption effect. However, at high temperature conditions at which we propose to employ the silica membranes, sorption effect is negligible and it is expected that not very significant deviations would occur between the H2 /CO2 selectivities obtained from pure and mixed gas experiments. As observed in Figs. 5–7, the O2 flow rate is a key fabrication parameter and can be used to tune the micromorphology of the membranes. Higher O2 flow rate results in higher permeabilities of small penetrants He, H2 and CO2 . During the oxidative thermolysis stage, higher O2 flow rate provides higher rate of mass transfer into the precursor film. At high concentration of O2 , smaller decompo-
Table 4 Pure gas selectivites of the silica membranes fabricated at different O2 flow rates. Pure gas selectivity O2 flow rate during fabrication
50 ml/min ◦
30 ml/min ◦
Feed conditions
35 C, 55.2 psia
80 C, 70 psia
35 ◦ C, 55.2 psia
80 ◦ C, 70 psia
H2 /CO2 O2 /N2 H2 /N2 H2 /CH4 CO2 /CH4 CO2 /N2
0.91 3.3 14.5 12.1 13.3 16.0
1.15 2.6 11.9 9.1 8.0 10.4
0.87 3.2 14.0 11.6 13.3 16.0
1.25 2.8 14.1 11.9 9.6 11.3
262
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
sition products are formed (Fig. 2) and hence, the pores formed due to the passage of these small products are smaller in diameter. The smaller pores are more selective to small gas molecules of He and H2 than toward larger gases like O2 , N2 and CH4 . It is also postulated that a very low or very high O2 flow rate results in incomplete oxidation of the precursor. A very low O2 flow rate would result in lack of sufficient mass transfer for thermal oxidation. At very high O2 flow rate, the outer layers of the precursor would undergo oxidation to form a layer so dense that they would hinder transport of O2 to the inner layers. Additionally, parameters such as thermal ramp rate and final soak temperature may also be varied to alter the micromorphology of the silica membranes. These issues will be investigated and reported in detail in subsequent articles. The purpose of this article is to provide a proof of concept that an appropriately processed polydimethylsiloxane sample can be converted in to molecularly selective silica membrane. 6. Conclusion A novel technique of fabricating H2 selective silica membranes has been described. The fabrication is a two-step process. In the first step, a precursor rubbery film of crosslinked polydimethylsiloxane was cast. In the second stage, the precursor film was subject to thermal oxidation, according to an optimized protocol, to yield flat microporous membrane films of silica. TGA and FTIR were used to characterize the thermal properties and chemical structure of these membranes. The membranes can withstand temperatures up to 300–350 ◦ C (temperature near the exit of the HTS watergas shift reactor) in air and have been developed for application in the water-gas shift process. Two different values of O2 flow rate (cm3 /min, during the second step in fabrication) were investigated for gas separation performance of six different gases. The membranes exhibit activated transport and provide selectivities higher than the Knudsen range. The aim is to make the silica membranes H2 /CO2 selective, and high temperatures and high pressures favor H2 permeation over CO2 . Further investigations of gas separation performance at higher temperatures up to 300 ◦ C and higher pressures are in progress. Moreover, additional research of the oxidative thermolysis stage should enable improvements in the high temperature H2 /CO2 selectivity of the developed silica membranes. The reported data show the viability of the novel silica membrane formation procedure, which was the focus of this article. Acknowledgements The authors gratefully acknowledge support for this work by Air Liquide, Inc., the Georgia Research Alliance, the Roberto C. Goizueta
Foundation and award no. KUS-I1-011-21 made by the King Abdullah University of Science and Technology (KAUST). References [1] N.W. Ockwig, T.M. Nenoff, Membranes for hydrogen separation, Chem. Rev. 107 (10) (2007) 4078–4110. [2] Arian Nijmeijer, Hydrogen selective silica membranes for use in membrane steam reforming, Thesis, University of Twente, 1999. [3] P. Molyneux, Permeation of gases through microporous silica hollow-fiber membranes: application of the transition-site model, J. Membr. Sci. 320 (1–2) (2008) 42–56. [4] R.M. de Vos, H. Verweij, Improved performance of silica membranes for gas separation, J. Membr. Sci. 143 (1998) 37–51. [5] J.D. Way, D.L. Roberts, Hollow fiber inorganic membranes for gas separations, Sep. Sci. Technol. 27 (1) (1992) 29–41. [6] Y.F. Gu, P. Hacarlioglu, S.T. Oyama, Hydrothermally stable silica–alumina composite membranes for hydrogen separation, J. Membr. Sci. 310 (1–2) (2008) 28–37. [7] D. Lee, L. Zhang, S.T. Oyama, S. Niu, R.F. Saraf, Synthesis, characterization, and gas permeation properties of a hydrogen permeable silica membrane supported on porous alumina, J. Membr. Sci. 231 (1–2) (2004) 117–126. [8] J.J. Hammel, W.J. Robertson, W.P. Marshall, H.W. Barch, Balbhadra Daas, M.A.Smoot, R.P. Beaver, Process of gas enrichment with porous siliceouscontaining material, U.S. Patent 4842620. [9] K.-H. Lee, S.-J. Khang, Physical characteristics of a porous silica material formed by pyrolysis of silicone rubber, Ceram. Eng. Sci. Proc. 8 (1–2) (1987) 85–92. [10] K.A. Andrianov, Metalorganic Polymers, Interscience Publishers, 1965. [11] W.J. Koros, G.K. Fleming, Membrane-based gas separation, J. Membr. Sci. 83 (1) (1993) 1–80. [12] M.H. Hassan, J.D. Way, P.M. Thoen, A.C. Dillon, Single-component and mixedgas transport in a silica hollow-fiber membrane, J. Membr. Sci. 104 (1–2) (1995) 27–42. [13] V. Boffa, D.H.A. Blank, J.E. ten Elshof, Hydrothermal stability of microporous silica and niobia-silica membranes, J. Membr. Sci. 319 (1–2) (2008) 256– 263. [14] S. Battersby, T. Tasaki, S. Smart, B. Ladewig, S. Liu, M.C. Duke, V. Rudolph, J.C. Diniz da Costa, Performance of cobalt silica membranes in gas mixture separation, J. Membr. Sci. 329 (1–2) (2009) 91–98. [15] C. Barboiu, B. Sala, S. Bec, S. Pavan, E. Petit, Ph. Colomban, J. Sanchez, S. de Perthuis, D. Hittner, Structural and mechanical characterizations of microporous silica–boron membranes for gas separation, J. Membr. Sci. 326 (2) (2009) 514–525. [16] Petar R. Dvornic, High temperature stability of polysiloxanes, silicon compounds: silanes and silicones catalogue, Gelest Inc. [17] M. Kanezashi, M. Asaeda, Hydrogen permeation characteristics and stability of Ni-doped silica membranes in steam at high temperature, J. Membr. Sci. 271 (1–2) (2006) 86–93. [18] K. Yoshida, Y. Hirano, H. Fujii, T. Tsuru, M. Asaeda, Hydrothermal stability and performance of silica–zirconia membranes for hydrogen separation in hydrothermal conditions, J. Chem. Eng. Jpn. 34 (4) (2001) 523–530. [19] S. Battersby, T. Tasaki, S. Smart, B. Ladewig, S. Liu, M.C. Duke, V. Rudolph, J.C. Diniz da Costa, Hydrothermal stability of cobalt silica membranes in a water gas shift reactor, Sep. Purif. Technol. 66 (2009) 299–305. [20] D. Uhlmann, S. Liu, B.P. Ladewig, J.C. Diniz da Costa, Cobalt-doped silica membranes for gas separation, J. Membr. Sci. 326 (2) (2009) 316–321. [21] T.T. Moore, S. Damle, P.J. Williams, W.J. Koros, Characterization of low permeability gas separation membranes and barrier materials: design and operations considerations, J. Membr. Sci. 245 (2004) 227–231.
Contents lists available at ScienceDirect
Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci
Novel silica membranes for high temperature gas separations Neha Bighane, William J. Koros ∗ School of Chemical and Biomolecular Engineering, Georgia Institute of Technology, 778 Atlantic Drive, Atlanta, GA 30332-0100, USA
a r t i c l e
i n f o
a b s t r a c t
Article history: Received 20 October 2010 Received in revised form 22 January 2011 Accepted 25 January 2011 Available online 1 February 2011 Keywords: Silica Membrane Oxidative thermolysis
This article describes fabrication of novel silica membranes derived via controlled oxidative thermolysis of polydimethylsiloxane and their gas separation performance. The optimized protocol for fabrication of the silica membranes is described and pure gas separation performance in the temperature range 35–80 ◦ C is presented. It is observed that the membranes exhibit activated transport for small gas penetrants such as He, H2 and CO2 . The membranes can withstand temperatures up to 350 ◦ C in air and may ultimately find use in H2 /CO2 separations to improve efficiency in the water-gas shift reactor process. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Membrane separation processes can provide low cost improvements in thermally driven processes such as production of hydrogen from steam reforming of natural gas. This dominant H2 production process comprises a highly endothermic steam methane reforming reaction (∼800 ◦ C), a pair of water gas shift reactors (high temperature shift (HTS) at 350 ◦ C and low temperature shift (LTS) at 200 ◦ C) and downstream by-product CO2 absorption units [1,2]. The water gas shift reaction (Eq. (1)) is thermodynamically limited and can be driven forward by selective removal of product hydrogen by membrane separation. CO + H2 O ↔ CO2 + H2
H = −41.2 kJ/mol
(1)
Installation of a hydrogen selective membrane unit between the two water gas shift reactors can provide increased efficiency of the overall production process in the following ways: 1. Production of a high purity hydrogen stream from the membrane unit. 2. Higher conversion in the LTS, due to lower concentration of products in the equilibrium mixture entering LTS. 3. Higher rate of reaction in LTS, implying a smaller volume of reactor. 4. Excess reactant steam, that is currently used to drive the reaction forward, can be minimized.
∗ Corresponding author. Tel.: +1 404 385 2845; fax: +1 404 385 2683. E-mail addresses: [email protected], [email protected], [email protected] (W.J. Koros). 0376-7388/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2011.01.045
5. Product stream from LTS has been calculated to become CO2 rich, as opposed to H2 rich stream in the absence of membrane separation, and downstream CO2 capture becomes more efficient. H2 and CO2 are the most abundant species in the equilibrium mixture exiting from the HTS, where the installation of the membrane unit is proposed. Typically, the product stream from a HTS operating at 350 ◦ C comprises 73.9% H2 , 17.7% CO2 , 7.4% CH4 and 1.0% CO (dry basis) [1]. In addition to the gases, steam is also a major component in the reaction mixture. Hence, it is necessary to seek the development of economical hydrothermally stable membranes that can provide high H2 /CO2 selectivity at temperatures similar to the exit temperature of the HTS (300–350 ◦ C). Characterization of membrane materials for gas separations includes testing them for permeability (productivity) and selectivity (efficiency) toward the gases. Inorganic membranes, in particular silica membranes, have been well-recognized for their gas separation abilities at high temperatures. The efficiency of such membranes is well-known and there are a number of publications on this subject [1–5]. However, the complexity of the conventional membrane fabrication processes for silica membranes led the authors to devise a more efficient and more economically scalable alternative fabrication approach. There are two well-known techniques of preparing dense silica membranes, namely, the sol–gel technique [4] and the Chemical Vapor Deposition (CVD) technique [6,7]. Both of these techniques require high-quality supports and CVD requires high cost deposition reactors. While the sol–gel technique involves coating of liquid precursors on a support and subsequent heat treatment, CVD involves deposition of gaseous phase species on a support. A third technique of melt extrusion and leaching to make hollow fiber silica membranes (PPG Industries) was patented by Ham-
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
255
Table 1 Major research found in literature, on permeation through silica membranes. S.no.
Author(s)
Temperature in fabrication (◦ C)
Testing temperature (◦ C)
1
Nijmeijer (Thesis, University of Twente) [2]
2 3
J.D.Way et al. [5] Renate de Vos et al. [4]
600 625 625 – 600
200 400 600 70 300
mel et al. [8] and their gas transport properties were studied by many groups [5]. A fourth potentially low cost technique of making tubular silicon based membranes from a two-step calcination of silicone rubber tubes was explored by Lee and Khang [9]; however, the resultant membranes could not yield selectivities above the Knudsen range. Table 1 presents a brief outline of some major research developments in this field for membranes prepared by the conventional methods, which do provide performance better than the Knudsen values. In state-of-the-art sol–gel microporous silica membranes, it has been observed that the permeability of all penetrants, except CO2 , increase with increase in temperature. CO2 permeability decreases due to a decrease in its sorption (P = D·S) in the membrane with rise in temperature. High selectivities of small penetrants like He and H2 are obtained and gas penetrants larger ˚ are completely sieved out i.e., cannot than CH4 , like SF6 (5.5 A) permeate through the membrane. In the present work, silica membranes have been fabricated via a new approach involving the controlled oxidative thermolysis of a precursor rubbery film of polydimethylsiloxane to create ceramic microporous flat film silica membranes that also show performance above Knudsen selectivity. 2. Membrane fabrication The concept of fabrication pursued here is based on careful optimization of the formation and thermal oxidation of a polysiloxane, viz.,
R1 Si
O
n
R2 to silica (SiO2 ), while retaining the mechanical integrity of the final film. Polydimethylsiloxane is a liquid resin at room temperature. This resin is subject to chemical crosslinking using tetraethoxysilane (Si(OC2 H5 )4 ). The crosslinked precursor is thermally oxidized. The thermal oxidation step eliminates the organic side groups and stitches the polymeric chains together via oxygen, resulting in silica. The thermo-oxidative stability of polysiloxanes [10] varies with the side groups as C6 H5 > ClC6 H4 > Cl3 C6 H2 > Cl2 C6 H3 > CH2 = CH > CH3 The precursor polymer was chosen as polydimethylsiloxane or PDMS (R1 = R2 = CH3 ) due to (i) its low cost, (ii) availability of literature since it is the most widely studied polysiloxane and (iii) because the methyl group is more easily oxidized as compared to other side groups. This technique enables fabrication and testing of unsupported silica membranes (flat films) because the precursor is a solid rubbery film. This fabrication technique also holds the potential to enable its transformation into hollow fibers, which is the preferred membrane structure in industrial applications. Moreover, this approach allows the possibility of crosslinking PDMS with transition metal alkoxides to obtain silica–metal oxide membranes. This latter capability is particularly attractive, since it is reported
H2 /CO2 selectivity 45 17 21 5.36 70
that silica–metal oxide materials are more stable in hydrothermal environments than pure silica matrices [17–20]. The current study does not focus on this important extension, but this is envisioned for subsequent studies after the basic technique has been demonstrated. As will be discussed, the fabrication parameters can be varied to tune the micromorphology of the silica membranes, and this is envisioned to provide considerable flexibility in creating tailored morphology for various applications. 2.1. Casting crosslinked precursor PDMS film A precursor film of PDMS is cast using ‘condensation cure’ chemistry. Silanol terminated polydimethylsiloxane was purchased from Gelest Inc. (DMS-S45, Mw = 80,000–130,000) and it is a viscous liquid at ambient conditions. A solution of the polymeric resin in n-heptane is thoroughly mixed with an appropriate crosslinking agent, tetraethoxysilane (Si(OC2 H5 )4 ) (TEOS). A 5:1 weight ratio of PDMS to Si(OC2 H5 )4 was found optimal, and 12–15 mg of tin (din-butylacetoxytin 95%) and titanium (titanium-2-ethylhexoxide) catalysts, per gram polymer, were used in the current study. The PDMS:TEOS ratio was found to be optimum based on careful studies of the ratio, with the preferred value showing the best mechanical properties in the final silica films after the oxidative thermolysis process. A 2:1 (PDMS:TEOS) ratio resulted in minute clusters of silica particles within the precursor films, due to self crosslinking of TEOS. This created inhomogeneity and caused the precursor to crumble upon oxidative thermolysis. Precursor films that deviate by more than 3% from the 5:1 value result in either substructure cracks or crumbling or breaking of the films upon oxidative thermolysis. Hence, the value 5:1 was chosen as optimal and precursors fabricated with this value yielded defect free silica membranes upon thermal oxidation. The prepared solution was poured into a circular teflon® dish (solution casting), maintained in a sealed glove bag filled with N2 and water vapor. Moisture accelerates the crosslinking rate of the condensation cure reaction. Hence, a controlled humidity level between 70 and 85% RH (Omega HH311 humidity meter) at 20–22 ◦ C was maintained in the glove bag. When the solution is exposed to this atmosphere, the silanol ends of the PDMS chains undergo condensation with the ethoxy groups of Si(OC2 H5 )4 (Fig. 1). The polymer begins to crosslink to form a film in 22–30 min. The system is allowed to equilibrate for 21 h to yield a transparent rubbery film. The film is then peeled off the dish and dried in a vacuum oven at 100 ◦ C for 20 h and 120 ◦ C for 11 h to ensure complete removal of solvent, by-product and volatile materials. The obtained PDMS films are typically 7.9–8.4 mil thick (1 mil = 1/1000 inch). While the solution can crosslink even without humidity, it takes a very long time (∼1.5–2 h at 9% RH). A long crosslinking time is not favorable when scale up is desired. Humidity increases the crosslink density of the precursor and the silica membrane obtained after oxidative thermolysis would have a denser pore structure. Silica membranes formed from a no-moisture crosslinked PDMS would have large pores and separation performance would be equal to Knudsen values. Hence, a humidity level of 70–85% RH was used that results in a crosslinking time of 22–30 min.
256
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
Crosslinking PDMS (Condensation cure with alkoxy crosslinker) CH 3
OC 2 H 5 C2H5O
Si
OC 2 H 5
+
HO
Si
n
CH 3
OC 2 H 5
OC 2 H 5
CH 3 O
Si
OH
+
C2H5 O
CH 3
Tetraethoxysilane ( TEOS)
Si
OC 2 H 5
OC 2 H 5
Sn, Ti catalysts Moisture
Reacon involves simultaneous hydrolysis of TEOS and condensaon with PDMS chains
O O
Si
O
CH 3 O
O
Si
O
CH 3
n
Si
O
+ by-product C 2 H 5 OH
O
(3-d crosslinked network) Fig. 1. Schematic representation of TEOS crosslinking PDMS resin to obtain a rubbery PDMS film.
2.2. Oxidative thermolysis From the PDMS film obtained in the previous step, discs were cut and their weight, thickness and area (Scion ImageTM ) measured. These films were placed on a steel wire mesh inside a long quartz tube in a tubular furnace. The tube was sealed with an oxygen purge (99% O2 ) at a desired flow rate (discussed later in Section 4.1). The system is heated from room temperature to 390 ◦ C according to an optimized protocol (Table 2). The rubbery film undergoes oxidative thermolysis to yield a flat transparent ceramic silica membrane. At the end of the heating stage, the oxygen flow is stopped and the system cooled at 1 ◦ C/min. The accurate continuous flow of oxygen during the heating phase is important to complete the oxidation process as well as to purge out decomposition products. An initial ramp rate 2 ◦ C/min to soak at 250 ◦ C is used as the precursor film does not undergo any modification until this temperature. While the thermal program given in Table 2 was found to be optimum, some flexibility in the protocol may be tolerated; however, these conditions are recommended to prevent curling, which complicates permeation testing. During the thermal oxidation stage, the methyl groups in the PDMS film undergo homolytic cleavage by gaseous phase O2 and the siloxane free radicals from different chains are bridged via oxygen to form a three-dimensional silica matrix. Internal rearrangement and shrinkage occurs. A schematic of the hypothesized mechanism [16] is presented in Fig. 2. Based on the mechanism shown in Fig. 2, the expected total weight loss in the thermal oxidation step can be calculated. Specifi-
Table 2 Optimal temperature-time protocol for oxidative thermolysis of crosslinked PDMS films to silica membranes. Process point Starting set point Set point 1 Time 1 Set point2 Time 2 Set point 3 Time 3 Set point 4 Time 4 Set point 5 Time 5
Process point value
Rate
cally, removal of two methyl groups and addition of two half oxygen atoms (O shared by two Si) per unit PDMS should show an expected weight loss equal to =
(−(2 × 15) + (2 × 0.5 × 16)) × 100 = −18.92% 74
The experimentally observed weight loss of >20 samples was found to be 19 ± 2%, which is in good agreement with the theoretical value. On an average, approximately 23% increase in the density of the material is observed upon oxidative thermolysis. Further support for the claim that the weight loss is due to the methyl groups is offered later via FTIR evidence. In an inert atmosphere such as He, a black residue of SiOC results, rather than the clear pure silica found in our studies. Since only CH3 groups are burnt out in this method, it creates only CO2 and H2 O, simplifying the control of the pores created during the oxidative burn out process. Table 3 summarizes the key parameters found to affect the quality of the final membranes. As noted above, optimal values of these parameters were obtained empirically to achieve defect free flat silica membrane films. O2 flow rate during the oxidative thermolysis stage has been observed to most markedly result in changes in the gas transport properties of the membranes and to allow tuning the micromorphology of the membranes. 3. Material characterization 3.1. Thermogravimetric analysis Thermogravimetric analysis (TGA) in an air purge was employed to visualize the weight loss pattern during the oxidative thermolysis stage of fabrication, as crosslinked PDMS oxidizes to silica. An open pan TGA was used to ensure exposure to the oxidizing atmosphere. Air was used instead of oxygen due to safety concerns for the test equipment. A crosslinked PDMS film was subjected to heat
◦
23 C 250 ◦ C 1 h 53 min 250 ◦ C 5 min 390 ◦ C 1 h 56 min 390 ◦ C 15 min 25 ◦ C 6 h 5 min
2 ◦ C/min 0 ◦ C/min (hold) ◦
1.2 C/min ◦
0 C/min (hold) 1 ◦ C/min
Table 3 Summary of key process parameters. Parameter
Optimal value
PDMS:TEOS weight ratio Humidity level (precursor casting) Ramp rate Soak temperature O2 flow rate (cm3 /min)
5:1 70–85% RH 1.2 ◦ C/min 390 ◦ C Variable – 30, 50
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
257
CH 3 ------------ O
+
Si ----------
O-O
CH 3
CH 2 OOH ------------ O
Si -----------CH 3
CO ------ O
Si -------
+
OH
+
H2 O
HCOOH
CH 2 O
CH 3
+
O2 CO 2
+
H
2
OH ------ O
Si -----CH 3
Two chains link via an O atom, after the removal of a methyl group from each. CH 3 ----- O
Si ------
CH 3 ----- O
OH
Si -------O
+
O2 H2 O
SiO 2
OH ----- O
Si ------
----- O
Si -------CH 3
CH 3
- Petar R. Dvornic, High Temperature Stability of Polysiloxanes, Silicon Compounds: Silanes and Silicones, Gelest, Inc. Fig. 2. Oxidative thermolysis of PDMS to silica.
up to 500 ◦ C in the air purge (Fig. 3a). The thermal protocol followed was the same as presented in Table 2, up to 390 ◦ C. The ramp rate from 390 ◦ C to 500 ◦ C was also 1.2 ◦ C/min. The weight loss pattern obtained was an indicator of the conversion. A two-step weight loss pattern was observed. The first weight loss step begins from ∼250 ◦ C and stabilizes at 390 ◦ C. Further heating to 425 ◦ C disintegrates the film into particles. From this observed weight loss pattern, it is clear that the conversion of PDMS to silica is complete by 390 ◦ C as the weight loss reached the
theoretical value of ∼19% and the film was intact until this temperature. The same was observed in the actual fabrication process as discussed in Section 2.2. Until 390 ◦ C, elimination of only methyl groups occurs and polymeric chains get connected via oxygen to form a silica film. Beyond 425 ◦ C, breakdown of the Si–O bonds and redistribution (i.e., splitting and reforming) occurs to form volatile oligomers and a residue of fine silica particles results [10]. TGA was also used to verify the thermal stability of the silica films. The films are thermally stable up to 390 ◦ C in air (dry oxidiz-
258
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
Fig. 3. (a) TGA of crosslinked PDMS film in air. (b) TGA of silica film in air showing stability at 390 ◦ C in a 2 h test.
ing condition) (Fig. 3b). Consequently, the membranes can easily withstand a temperature of 300–350 ◦ C, as desired for application in the water gas shift process. 3.2. Fourier transform infra-red spectroscopy Fourier transform infrared spectroscopy was employed to investigate the chemical structure of the precursor PDMS films and final silica membranes. The obtained absorbance spectra are shown in Fig. 4a and b. The large reduction in the methyl peak (2970 cm−1 ) and the evolution of the broad shoulder between 1000 and 1250 cm−1 that is a signature of amorphous silica, in Fig. 4b relative to Fig. 4a demonstrate the transformation of PDMS to silica via the thermal oxidation process. 4. Gas permeation measurements Gas permeabilities of the silica membranes have been measured over a temperature range of 35–80 ◦ C, with six penetrants. Work is in progress on measurements at higher temperatures up to 300 ◦ C and will be presented in a later publication, after further optimizing the process demonstrated here, to create high perfor-
mance size discriminating silica membranes from the oxidative thermolysis of PDMS. This article focuses only on the first proof-of principle of the oxidative thermolysis of PDMS to make molecularly selective membranes that exceed Knudsen performance levels. 4.1. Permeation system The gas permeation testing system is primarily engineered into three sections: • Sections 1 and 3 always remain at ambient temperature. Section 1 comprises a gas feed line, upstream transducer (Sensotec Model Z, absolute 0–1000 psia), feed valve and vent line. • Section 2 comprises an upstream volume and the permeation cell. Section 2 is maintained at the desired operating temperature via a heating mechanism. A heavy insulated heating tape, controlled by PID tuned benchtop controller (Omega® ), is employed for controlling the temperature of this section. Section 2 is connected to section 3. • Section 3 has a downstream volume with downstream transducer (Baratron, 0–10 Torr). The temperature of the downstream is measured and used for permeability calculations.
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
259
Fig. 4. (a) FTIR of crosslinked PDMS precursor thin film. (b) FTIR of silica membrane (KBr pellet technique).
The upstream and downstream volumes are connected to a vacuum pump (BOC Edwards model RV3) for evacuation. All fitting are made of stainless steel 316/316L to ensure safe, stable and leak proof operations. In order to conduct gas permeation tests, it is vital that the seal between the membrane and the test cell be leak proof. While the seals used in the present work have high integrity for measurements up to 80 ◦ C, difficulties in sealing above this temperature must be resolved and will be discussed in subsequent works for high temperature operation. 4.2. Principle of operation The membrane is installed into the permeation cell using an impermeable aluminum foil tape (Fasson® ) and epoxy (Duralco 4703, Cotronics, Inc.), using a permeation cell whose detailed description is available [21]. After masking the membrane into the permeation cell, the entire system is thoroughly evacuated for 24–48 h to degas the system. A leak rate is measured by monitoring the downstream pressure rise while the system remains evacuated. For inorganic membranes in high temperature operating conditions, the leak rate is usually a negligible fraction of the permeation rate of the slowest penetrant. After evacuation of the entire system, gas is fed into the upstream volume at ambient temperature (feed valve closed) at a desired pressure and the desired temperature of operation established in section 2. The permeation cell and downstream volume
are isolated from the upstream by the feed valve and are maintained under vacuum. Once the feed valve is opened, transport of gas through the membrane occurs at the desired temperature. The permeate starts accumulating in the downstream volume. The rise in pressure in the downstream volume as a function of time dp/dt is recorded and this data is used to calculate the permeability of the gas through the membrane film at the operating temperature and pressure. In each experiment, the rate of rise in downstream pressure in the range 0.2–10 Torr is recorded. From the obtained data, the steady state data is used to calculate the permeability of the penentrant. Steady-state is assumed to be attained after roughly 10 time lag periods [11]. For each experiment, a second run was performed by evacuating the downstream in steady state to verify that the rate of rise in pressure remains the same as in the previous run. No time lag is observed in the second run, since the system is already under steady state conditions. The permeation rate is corrected for leaks by subtracting the leak rate from the measured permeation rate. No time lag was observed for He and H2 while the time lag for larger penetrants decreased with increase in temperature, due to an increase in the diffusivities. Before a different penetrant was studied, the entire system was evacuated for atleast 12 h. The area available for permeation and the thickness of the membrane were measured. In this work, the typical silica membrane area open for permeation was 0.5–1 cm2 and the typical membrane thickness was 7.4–7.9 mils.
260
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
Steady state flux was used to calculate the permeability (in Barrer) similar to the procedure described earlier [21]. 4.3. Theory The intrinsic separation performance of a membrane material is measured in terms of its productivity (permeability) and efficiency (selectivity). The permeability, P, of a gas through a membrane is defined as the flux normalized by the pressure driving force per unit thickness [11]. Pi = Di Si =
(Flux)i pi /l
(2)
D is the diffusivity and S is the sorption coefficient of the gas penetrant ‘i’ in the membrane, pi is the partial pressure difference and l is the thickness of the membrane. The detailed diffusion mechanism in porous membranes is dependent upon the size of the pores relative to the size of the permeating gas and the mean free path of the penetrant. If the pores are large compared to the molecular size of the penetrants but small relative to the bulk gas mean free path under the conditions of measurement, Knudsen diffusion tends to be dominant. For Knudsen transport, the selectivity is low and equal to the inverse ratio of the square root of the molecular weight of the two permeating gases. For example, in the case of O2 and N2 , the smaller but heavier O2 permeates slower than N2 by the factor (28/32)1/2 = 0.94. For non-porous polymeric membranes, a solution-diffusion mechanism prevails. In microporous (pore diameter < 2 nm) inorganic membranes such as the silica membranes considered here, with pores of sizes of the order of the kinetic diameters of penetrants, interaction between the penetrant and the pore walls become significant, especially at low temperatures. Hence, in addition to diffusion through size selective pores, the resultant mechanism becomes sorption–diffusion. In this case, the smaller O2 permeates faster than N2 . The gas transport becomes activated and shows molecular sieve-like separation properties, with high selectivities of He and H2 over larger molecules [2–5,12–15]. In such a case, the process is said to be thermally activated and follows a sorption-diffusion mechanism with permeability given by Eq. (2). As shown in Eq. (2), permeability is fundamentally related to the product of the diffusivity (dependent on molecular size) and the sorption coefficient (dependent on condensability) of the penetrant gas molecule. The unit of permeability commonly used is Barrer, which can also be reported in SI units, viz., 1Barrer = 10−10
cm3 (STP) cm mol m = 3.348 × 10−16 2 cm2 sec cm Hg m s Pa
(3)
Fig. 5. Permeability of silica membranes at 55.2 psia, 35 ◦ C, for He, H2 , CO2 , O2 , N2 and CH4 , fabricated with two different O2 flow rate parameter values.
The activation energy for permeation is the sum of the activation energy for diffusion Ed and the heat of sorption Hs (generally exothermic i.e., Hs is a negative value). As temperature rises, the diffusivity of a penetrant in a membrane increases while its sorption coefficient decreases, causing variation in the permeability according to Eq. (2). As pressure rises, the permeability of a gas penetrant either decreases or remains constant, primarily depending upon the sorption isotherm of the particular penetrant for the membrane. For gases such as O2 and N2 that have similar sorption enthalpies, one can detect the onset of activated permeation when significant deviations from Knudsen selectivity become apparent. Thus, when the smaller but high molecular weight O2 permeates significantly faster than the slightly larger but lower molecular weight N2 , activated sorption–diffusion is indicated. The case of H2 /CO2 is more complex due to the high condensability of CO2 . CO2 can permeate faster than the smaller H2 but as temperature increases, the high diffusivity of H2 starts dominating. These effects are considered in the following section. 5. Gas separation performance 5.1. Gas permeation properties In the present work, it has been observed that variation in the O2 flow rate, used during the fabrication of the silica membranes, results in variations in the gas separation properties. Hence, two different values of O2 flow rate fabrication parameter were investigated.
The ratio of the permeabilites of pure gases (for negligible downstream pressure) used in the experiments here provide a good measure of selectivity or permselectivity (Eq. (4)). ˛i/j =
(yi /yj )Permeate (yi /yj )Feed
DS Pi ∼ = i i = Pj Dj Sj
(4)
In the case of thermally activated permeation, the variation of permeability with temperature for a given feed partial pressure follows an apparent Arrhenius relationship (Eq. (5)) P = P0 exp
−E p
RT
(5)
where P0 is a pre-exponential factor, Ep is the apparent activation energy for permeation, T is the temperature in Kelvin and R is the universal gas constant. Ep = Ed + Hs
(6)
Fig. 6. Permeabilites of He, H2 , CO2 , O2 , N2 and CH4 , for a silica membrane (fabricated at 30 ml/min O2 flow rate) at 35 ◦ C, 55.2 psia and 80 ◦ C, 70 psia.
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
Fig. 7. Permeabilites of He, H2 , CO2 , O2 , N2 and CH4 , for a silica membrane (fabricated at 50 ml/min O2 flow rate) at 35 ◦ C, 55.2 psia and 80 ◦ C, 70 psia.
The permeability data (Figs. 5–7) and obtained selectivity values (Table 4) indicate that the mechanism of gas permeation through the membranes primarily follows a sorption–diffusion mechanism. It is particularly significant that the H2 /CO2 selectivity has “switched” as the temperature rose by only 45 ◦ C. These data show the tendency to shift from a sorption dominated selectivity for the condensable CO2 at 35 ◦ C to a size dominated selectivity for H2 at 80 ◦ C. This trend in anticipated to become more extreme and strongly favor H2 at still higher temperatures. A new permeation test equipment is under construction to cater to gas permeation tests at higher temperatures up to 300 ◦ C and for a flammable gas like H2 . Consistent with results for samples created by sol–gel and other more difficult membrane preparation techniques, the silica membranes appear to possess an interconnected network of pores and the pore size distribution is such that the smaller penetrants can access a larger fraction of the total pores. The average pore size of the majority of pores lies between the kinetic diameters of H2 ˚ and O2 (3.46 A). ˚ The O2 /N2 selectivity is 3 and is reasonable (2.89 A) given that a value between 3 and 8 is considered good for this tough ˚ and N2 (3.64 A) ˚ are very similar in size. The separation as O2 (3.46 A) combination of H2 permeability of 1200 Barrer and H2 /N2 selectivity of 14 lies on its upper bound curve [11] for polymer membranes for this separation. A fairly high CO2 /CH4 selectivity of 13 is also observed. The permeabilities of He, H2 , O2 , N2 and CH4 increase with increase in temperature while that of CO2 decreases. This is indicative of activated nature of permeation with positive activation energies (Ep ) except for CO2 , which has negative activation energy due to its sorption effect. As observed, variation in the O2 flow rate (during fabrication) leads to changes in the gas transport properties of He, H2 and CO2 . It is hypothesized that smaller pores that are more selective to He and H2 are obtained when a higher O2 flow rate is used. Hence, this parameter can be used to tune
261
Fig. 8. Variation of permeability of He, H2 , CO2 and N2 , with pressure (50–100 psia) at 35 ◦ C, of silica membrane fabricated at 50 ml/min O2 .
micro-morphology in the membranes, along with thermal ramp rates in the critical thermolysis zone. The ideal separation factor for H2 /CO2 increases from 0.87 (i.e., CO2 selective) at 35 ◦ C, 55.2 psia to 1.25 (i.e., H2 selective) at 80 ◦ C, 70 psia. The activated nature of the transport through the silica membranes created at 50 ml/min O2 during thermolysis is also apparent in the case of larger gases (O2 , N2 and CH4 ) where the permeabilities in Fig. 7 rise with temperature from 35 ◦ C to 80 ◦ C and the O2 /N2 selectivity in Table 4 exceeds the Knudsen value of 0.94. All these permeation data were reproducible. Fig. 8 shows the variation of gas permeability, through the silica membranes, with pressure from 50 to 100 psia, at 35 ◦ C. While the permeability of CO2 decreased by 11%, the permeabilities of He, H2 and N2 remain relatively constant. These data illustrate the effects of isotherm curvature with pressure. CO2 permeability decreases due to saturation in the sorption capacity of the membrane for the penetrant. It also indicates that even at the low temperature of 35 ◦ C, the size based selectivity of the membranes for H2 can be observed. High temperatures and high pressures favor H2 permeation over CO2 . It has been observed in sol–gel silica membranes [4] that increase in pressure causes small decrease in H2 /CO2 when a mixed gas feed is used, due to competitive sorption effect. However, at high temperature conditions at which we propose to employ the silica membranes, sorption effect is negligible and it is expected that not very significant deviations would occur between the H2 /CO2 selectivities obtained from pure and mixed gas experiments. As observed in Figs. 5–7, the O2 flow rate is a key fabrication parameter and can be used to tune the micromorphology of the membranes. Higher O2 flow rate results in higher permeabilities of small penetrants He, H2 and CO2 . During the oxidative thermolysis stage, higher O2 flow rate provides higher rate of mass transfer into the precursor film. At high concentration of O2 , smaller decompo-
Table 4 Pure gas selectivites of the silica membranes fabricated at different O2 flow rates. Pure gas selectivity O2 flow rate during fabrication
50 ml/min ◦
30 ml/min ◦
Feed conditions
35 C, 55.2 psia
80 C, 70 psia
35 ◦ C, 55.2 psia
80 ◦ C, 70 psia
H2 /CO2 O2 /N2 H2 /N2 H2 /CH4 CO2 /CH4 CO2 /N2
0.91 3.3 14.5 12.1 13.3 16.0
1.15 2.6 11.9 9.1 8.0 10.4
0.87 3.2 14.0 11.6 13.3 16.0
1.25 2.8 14.1 11.9 9.6 11.3
262
N. Bighane, W.J. Koros / Journal of Membrane Science 371 (2011) 254–262
sition products are formed (Fig. 2) and hence, the pores formed due to the passage of these small products are smaller in diameter. The smaller pores are more selective to small gas molecules of He and H2 than toward larger gases like O2 , N2 and CH4 . It is also postulated that a very low or very high O2 flow rate results in incomplete oxidation of the precursor. A very low O2 flow rate would result in lack of sufficient mass transfer for thermal oxidation. At very high O2 flow rate, the outer layers of the precursor would undergo oxidation to form a layer so dense that they would hinder transport of O2 to the inner layers. Additionally, parameters such as thermal ramp rate and final soak temperature may also be varied to alter the micromorphology of the silica membranes. These issues will be investigated and reported in detail in subsequent articles. The purpose of this article is to provide a proof of concept that an appropriately processed polydimethylsiloxane sample can be converted in to molecularly selective silica membrane. 6. Conclusion A novel technique of fabricating H2 selective silica membranes has been described. The fabrication is a two-step process. In the first step, a precursor rubbery film of crosslinked polydimethylsiloxane was cast. In the second stage, the precursor film was subject to thermal oxidation, according to an optimized protocol, to yield flat microporous membrane films of silica. TGA and FTIR were used to characterize the thermal properties and chemical structure of these membranes. The membranes can withstand temperatures up to 300–350 ◦ C (temperature near the exit of the HTS watergas shift reactor) in air and have been developed for application in the water-gas shift process. Two different values of O2 flow rate (cm3 /min, during the second step in fabrication) were investigated for gas separation performance of six different gases. The membranes exhibit activated transport and provide selectivities higher than the Knudsen range. The aim is to make the silica membranes H2 /CO2 selective, and high temperatures and high pressures favor H2 permeation over CO2 . Further investigations of gas separation performance at higher temperatures up to 300 ◦ C and higher pressures are in progress. Moreover, additional research of the oxidative thermolysis stage should enable improvements in the high temperature H2 /CO2 selectivity of the developed silica membranes. The reported data show the viability of the novel silica membrane formation procedure, which was the focus of this article. Acknowledgements The authors gratefully acknowledge support for this work by Air Liquide, Inc., the Georgia Research Alliance, the Roberto C. Goizueta
Foundation and award no. KUS-I1-011-21 made by the King Abdullah University of Science and Technology (KAUST). References [1] N.W. Ockwig, T.M. Nenoff, Membranes for hydrogen separation, Chem. Rev. 107 (10) (2007) 4078–4110. [2] Arian Nijmeijer, Hydrogen selective silica membranes for use in membrane steam reforming, Thesis, University of Twente, 1999. [3] P. Molyneux, Permeation of gases through microporous silica hollow-fiber membranes: application of the transition-site model, J. Membr. Sci. 320 (1–2) (2008) 42–56. [4] R.M. de Vos, H. Verweij, Improved performance of silica membranes for gas separation, J. Membr. Sci. 143 (1998) 37–51. [5] J.D. Way, D.L. Roberts, Hollow fiber inorganic membranes for gas separations, Sep. Sci. Technol. 27 (1) (1992) 29–41. [6] Y.F. Gu, P. Hacarlioglu, S.T. Oyama, Hydrothermally stable silica–alumina composite membranes for hydrogen separation, J. Membr. Sci. 310 (1–2) (2008) 28–37. [7] D. Lee, L. Zhang, S.T. Oyama, S. Niu, R.F. Saraf, Synthesis, characterization, and gas permeation properties of a hydrogen permeable silica membrane supported on porous alumina, J. Membr. Sci. 231 (1–2) (2004) 117–126. [8] J.J. Hammel, W.J. Robertson, W.P. Marshall, H.W. Barch, Balbhadra Daas, M.A.Smoot, R.P. Beaver, Process of gas enrichment with porous siliceouscontaining material, U.S. Patent 4842620. [9] K.-H. Lee, S.-J. Khang, Physical characteristics of a porous silica material formed by pyrolysis of silicone rubber, Ceram. Eng. Sci. Proc. 8 (1–2) (1987) 85–92. [10] K.A. Andrianov, Metalorganic Polymers, Interscience Publishers, 1965. [11] W.J. Koros, G.K. Fleming, Membrane-based gas separation, J. Membr. Sci. 83 (1) (1993) 1–80. [12] M.H. Hassan, J.D. Way, P.M. Thoen, A.C. Dillon, Single-component and mixedgas transport in a silica hollow-fiber membrane, J. Membr. Sci. 104 (1–2) (1995) 27–42. [13] V. Boffa, D.H.A. Blank, J.E. ten Elshof, Hydrothermal stability of microporous silica and niobia-silica membranes, J. Membr. Sci. 319 (1–2) (2008) 256– 263. [14] S. Battersby, T. Tasaki, S. Smart, B. Ladewig, S. Liu, M.C. Duke, V. Rudolph, J.C. Diniz da Costa, Performance of cobalt silica membranes in gas mixture separation, J. Membr. Sci. 329 (1–2) (2009) 91–98. [15] C. Barboiu, B. Sala, S. Bec, S. Pavan, E. Petit, Ph. Colomban, J. Sanchez, S. de Perthuis, D. Hittner, Structural and mechanical characterizations of microporous silica–boron membranes for gas separation, J. Membr. Sci. 326 (2) (2009) 514–525. [16] Petar R. Dvornic, High temperature stability of polysiloxanes, silicon compounds: silanes and silicones catalogue, Gelest Inc. [17] M. Kanezashi, M. Asaeda, Hydrogen permeation characteristics and stability of Ni-doped silica membranes in steam at high temperature, J. Membr. Sci. 271 (1–2) (2006) 86–93. [18] K. Yoshida, Y. Hirano, H. Fujii, T. Tsuru, M. Asaeda, Hydrothermal stability and performance of silica–zirconia membranes for hydrogen separation in hydrothermal conditions, J. Chem. Eng. Jpn. 34 (4) (2001) 523–530. [19] S. Battersby, T. Tasaki, S. Smart, B. Ladewig, S. Liu, M.C. Duke, V. Rudolph, J.C. Diniz da Costa, Hydrothermal stability of cobalt silica membranes in a water gas shift reactor, Sep. Purif. Technol. 66 (2009) 299–305. [20] D. Uhlmann, S. Liu, B.P. Ladewig, J.C. Diniz da Costa, Cobalt-doped silica membranes for gas separation, J. Membr. Sci. 326 (2) (2009) 316–321. [21] T.T. Moore, S. Damle, P.J. Williams, W.J. Koros, Characterization of low permeability gas separation membranes and barrier materials: design and operations considerations, J. Membr. Sci. 245 (2004) 227–231.