- Email: [email protected]
Generalization of effect of oxygen exposure on formation and performance of carbon molecular sieve membranes
CARBON
4 8 ( 2 0 1 0 ) 4 4 4 2 –4 4 4 9
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/carbon
Generalization of effect of oxygen exposure on formation and performance of carbon molecular sieve membranes Mayumi Kiyono, Paul J. Williams, William J. Koros
*
Department of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0001, USA Shell Projects and Technology, 3333 Highway 6 South, Houston, TX 77082, USA
A R T I C L E I N F O
A B S T R A C T
Article history:
Previously, a pyrolysis method was developed to control separation performance of carbon
Received 1 May 2010
molecular sieve (CMS) membranes by doping with trace oxygen. This method involved oxy-
Accepted 3 August 2010
gen exposure during pyrolysis to tune the selective pore windows. During the development,
Available online 6 August 2010
it was observed that oxygen concentration in inert gas, rather than the total amount of oxygen exposed controls performance. In this study, we hypothesized that mass transfer of oxygen in CMS membranes during pyrolysis is governed by chemical reaction at critical pore opening. Effect of thermal soak time, inert flow rate, and precursor thickness were conducted to test this hypothesis with 6FDA/BPDA-DAM and Matrimid polymer precursors. Results of oxygen consumption from pyrolysis process and CO2/CH4 separation performance showed that the process is likely governed by reaction kinetics. This observation implies simplicity and easy scale-up for the oxygen ‘‘doping’’ method on CMS formation by tuning the oxygen concentration in the pyrolysis atmosphere. Published by Elsevier Ltd.
1.
Introduction
Carbon molecular sieve (CMS) membranes are known to have attractive gas separation performance, well exceeding the polymer upper bound curve for challenging gas pairs, such as O2/N2, CO2/CH4, C3H6/C3H8 [1,2]. CMS membranes are formed by thermal decomposition of polymer precursors. As a result of this thermal decomposition process, amorphous turbostratic carbon membranes are usually formed [3]. Pores are formed from packing imperfections between microcrystalline regions in the material [4,5], and the pore structure in CMS membranes is described as ‘‘slit-like’’ with a bimodal pore distribution in which large pores connected by smaller pores known as ‘‘ultramicropores’’ [6,7]. This combination of ultramicropores and micropores provides the molecular sieving function. While traditional vacuum pyrolysis produces CMS membranes with attractive separation performance, the approaches are difficult to scale up economically.
* Corresponding author: Fax: +1 4043852683. E-mail address: [email protected] (W.J. Koros). 0008-6223/$ - see front matter Published by Elsevier Ltd. doi:10.1016/j.carbon.2010.08.003
With practical impact in mind, we extended our previously developed inert pyrolysis method with trace oxygen exposure using synthesized polymer 6FDA/BPDA-DAM [8] to allow use of economical precursors. Specifically we demonstrated the ability to use a commercially available polymer Matrimid precursor, and further identified factors to improve the efficacy of the method based on intrinsic, oxygen free pyrolysis case, structure of CMS membranes [9]. The method essentially utilizes oxygen chemisorption at selective pore windows during the high temperature pyrolysis process. Numerous studies have investigated the oxygen–carbon reaction process for various kinds of carbon materials [10–15]. When the carbon–oxygen reaction is believed to produce C@O chemisorptions entities on the surface, it is difficult to identify and control their true nature due to extreme sensitivity to the limitations imposed by energy and mass transfer processes [13,16]. When the oxygen doping method was initially developed, it was assumed that the total amount of oxy-
CARBON
4443
4 8 ( 20 1 0 ) 4 4 4 2–44 4 9
gen presented at the sample controlled the amount of chemisorption at selective pore windows. Based on surprising results discussed here, we now present evidence that the ultimate structure are more likely controlled by oxygen concentration in the inert gas at the temperature of the pyrolysis, rather than a total accessible oxygen, per se. This suggests an equilibrium limited structure an opposed to a more conventionally anticipated kinetically-controlled structure. If true, such a structure would be advantageous for practical processing. In this work, we seek to identify the controlling factor of the oxygen ‘‘doping’’ process during formation of CMS membranes. Specifically this study explains a series of well controlled experiments to determine the controlling factor governing the oxygen–carbon reaction by means of oxygen consumption during pyrolysis and separation performance of resulting CMS membranes. This identification is essential for transitioning from laboratory to industrial scale for the mass production of CMS membranes.
2.
Theory and background
2.1.
Structure and transport of carbon molecular sieve
In this study, CMS membranes are formed by thermal decomposition of polymeric membranes. When polymers are pyrolyzed, either coke or char is formed [3,17]. While coke forms graphite at a temperature above 2200 C, char remains in an amorphous structure [3,17]. Most CMS membranes are char, having a complex turbostratic structure consisting of parallel layers of condensed hexagonal rings with no three-dimensional crystalline order, as shown in Fig. 1a and conveniently represented by Fig. 1b for engineering purposes [3]. Pores of CMS membranes are described as slit-like, and have a bimodal pore size distribution made of ultramicropores and micropores [6]. The CMS structures as well as resulting separation performance properties of CMS membranes strongly depend on precursor property and pyrolysis process [2,5,6,18]. Like polymeric membrane systems, the sorption–diffusion mechanism is used to explain gas transport through carbon molecular sieve (CMS) membranes. Gas molecules sorb at the upstream, diffuse into the membrane under the influence of a chemical potential gradient, and desorb from the downstream. There are two parameters commonly used to measure separation performance of membranes. Permeability is used to measure the intrinsic productivity of a membrane
material, and is equal to the pressure and thickness normalized flux, viz., Pi ¼
ð1Þ
In the equation above, ni represents the flux of component ‘‘i’’ gas molecules through the membrane of the membrane thickness, l, and Dpi is the transmembrane partial pressure difference that acts as the driving force for component i across the membrane. The most common unit for permeability is the Barrer which is defined as: 1 Barrer½¼1010
cm3 ðSTPÞ cm : cm2 s cmHg
ð2Þ
An additional popular unit, kmol m m2 s1 kPa1, can be obtained by multiplying Barrers by 2.99 · 1015. Another parameter, selectivity, is a measure of the membrane’s separation efficiency and equals the ratio of the component permeabilities for the case of a negligible downstream permeate pressure.
2.2.
Oxygen ‘‘doping’’ method
As an alternative to the traditional vacuum pyrolysis, the authors developed a tuning method with utilizes oxidation during inert pyrolysis process [8]. The method relies on the fact that oxygen may chemisorb and form a stable bond at the selective ultramicropore windows at high temperature [8]. The method was based on evidence that: (i) the selective pores have sp2 carbon edges that are more reactive than basal plane [11,13,19], (ii) oxygen chemisorption is the limiting step of doping process [10], (iii) the bond is strong and stable and it does not break unless it experiences the same heating treatment responsible for its original formation [13]. Therefore, once applied, this doping process makes the sieving structure less permeable and more selective as the amount of doping increases up to a point before actual large scale oxidation occurs. Synthesized polymer, 6FDA/BPDA-DAM, was used to test the method. Separation performance of CO2 and CH4 supported the hypothesis that permeability decreases with an increase in selectivity up to a point when excessive doping makes the sieve lose both permeability and selectivity. This method was also demonstrated with a commercially available polymer, Matrimid, and FT-IR showed evidence of chemisorbed oxygen as C@O bonds [9].
2.3.
Fig. 1 – (a) Turbostratic structures of CMS membranes, adapted Pierson [17] and (b) an idealized slit-like structure.
ni l : Dpi
Oxidation mechanism
The oxygen ‘‘doping’’ method allows one to enhance membrane performance by >100 times in permeability with doubled selectivity when CMS pyrolysis atmosphere is optimized, compared with separation performance of polymeric precursors [8,9]. Oxidation of carbon is complex due to the fact that the oxygen–carbon reaction involves chemical kinetics with consequent heat transfer and mass transfer processes at a number of levels. Gaseous oxygen molecules move from the surrounding bulk atmosphere to the carbon surface and are adsorbed to form surface intermediates, which may rearrange, desorb, and return to the gas phase [13,16,19]. Moreover, emitted decomposition by-products can be reacted
4444
CARBON
4 8 ( 2 0 1 0 ) 4 4 4 2 –4 4 4 9
in the external bulk gas phase. This study seeks to identify the limiting factor of the doping process during formation of CMS membranes. Three regimes are considered during the oxidation process: (i) oxygen fully penetrates the solid and all active sites are available for the reaction, (ii) oxidant penetration is partial and oxygen diffusion into the solid is insufficient to supply all reaction needs, and (iii) reaction only takes place at the outer surface [20,21] for the available contact time. Investigation of the influence of diffusional resistance relative to reaction resistance can be carried out using a fundamental analysis of the Thiele modulus, h, which is defined as: 1 d rP kb 2 ð3Þ h¼ 2 De rOX where d is the particle diameter in meters, rP is its density in kg/m3, k is the carbon reaction rate in kg/kg s, b is the stoichiometric mass ratio of oxygen to carbon, De is the effective diffusion coefficient in m2/s, and rOX is the bulk gas concentration of oxygen in kg/m3 [14]. The Thiele modulus is applied by means of an effectiveness factor, which varies between 0 and 1 [14,22]. The effectiveness factor represents the fraction of internal surface which can react when exposed to the surface concentration of reactant gas [22]. When the Thiele modulus is larger than 1, the reaction rate hinders the ability of diffusion to supply oxygen to the reactive surface and the effectiveness factor is much smaller than 1. When the Thiele modulus is smaller than unity, there is no resistance to pore diffusion. In terms of formation of CMS membranes, factors like pyrolysis temperature, film thickness, and pore structure strongly influence the process and determine the three regimes described earlier. A preliminary calculation can be made to identify one of three oxidation process regimes from the Thiele modules in Eq. (3). The bulk oxygen concentration, rOX, of our interests is in 1–100 ppm, which corresponds to 1.41 · 106–1.41 · 104 kg/m3. The density, rP, of CMS membranes is reported to be in the order of 103 kg/m3, depending on polymeric precursors [6]. At temperatures of 500–800 C, where typical CMS membranes are produced, carbon oxidation produces CO as the dominant product [23]. According to Stantmore et al., this results in a b value of 4/3 [16]. Effective diffusion coefficients of oxygen in CMS membranes at high temperatures are in the order of 108 m2/s, according to Singh [24]. According to de Soete, the reaction is in the first order when oxygen adsorption dominates, and overall reaction rate can be in the range of 0.5–0.9 s1 [10,25]. When either homogeneous dense films or asymmetric hollow fibers have a selective skin thickness of d = 0.1–30 lm, calculation of the Thiele modulus based on these literature values results in a value of 0.6–2500. Since this range clearly spans values less than and greater than one, it suggests the possibility of pore diffusion resistance control of the oxidation reaction. Nevertheless, there is considerable uncertainty in the above parameters, so this study probes this issue experimentally in detail. While the above theoretical calculation suggests possible pore diffusion resistance of oxygen during high temperature
1
Table 1 – Pyrolysis temperature protocol. Tinitial (C)
Tfinal (C)
25 250 535 550
Ramp rate (C/min)
250 535 550 550
13.3 3.85 0.25 2 h soak
pyrolysis in carbon membranes, some researchers have seen inconsistent separation performance of CMS membranes pyrolyzed under atmosphere in which oxygen exists for possible oxidation reaction. Variations have been reported in CMS separation performance with respect to polymer precursor thickness [5], and researchers have seen separation performance independent of inert flow rates [4,5,26]: observations and the theoretical calculation have not reached an agreement. Therefore we seek to draw a more definite conclusion by identifying the controlling factor in the oxygen doping process by a series of well controlled experiments, specifically, by means of oxygen consumption during pyrolysis and separation performance of the resulting CMS membranes.
3.
Experimental
3.1.
Material
Two polymer materials were chosen: commercially available polyimide Matrimid from Vantico, Inc. and synthesized polyimide 6FDA/BPDA(1:1)-DAM. Their chemical structures, properties, and details of synthesis process are described elsewhere [8,9]. Polymeric dense films were prepared by the solution casting method, and they were heated accordingly with a pyrolysis protocol shown in Table 1 using the pyrolysis system shown in Fig. 2. A mixture of argon and various amounts of oxygen in ppm levels were used as an inert gas. The system was purged for at least 10 times the volume of the whole system to reach equilibrium before each pyrolysis experiment. During the pyrolysis process, oxygen concentration levels were monitored using an oxygen analyzer (Cambridge Sensotec Ltd., Rapidox 2100 series, Cambridge, England with ±1% accuracy between 1020 ppm and 100%1). Accuracy of the analyzer was confirmed by calibrating with air before each experiment, and the consistent oxygen level was checked before and after the pyrolysis process. Once an oxygen concentration profile with respect to time was obtained, the total amount of oxygen available and the amount of oxygen consumed were calculated using Eqs. (4) and (5) Total amount of O2 ¼ Q_ purge ½ðppm O2 ÞFeed t Amount of O2 consumed ¼ Q_ purge ½ðppm O Þ
ð4Þ
2 Feed
ðppm O2 ÞMeasured t
ð5Þ
The parameter Q_ purge is the volumetric flow rate of the inert gas and was measured by a bubble flow meter. Values (ppm O2)Feed and (ppm O2)Measured were obtained by the oxygen analyzer
Rapidox 2100 oxygen analyzer instruction manual. Cambridge, England: Cambridge Sensotec Limited; 2004.
CARBON
4445
4 8 ( 20 1 0 ) 4 4 4 2–44 4 9
Micro needle valve
O2 Sensor
Inert Gas Mass Flow Controller
Vent
Quartz Tube Thermocouple
Data Acquisition Temperature Controller
Fig. 2 – Pyrolysis set-up for CMS membrane preparation.
4.
Results and discussion
4.1.
Effect of thermal soak time
Effect of thermal soak time on the oxygen doping process was investigated by comparing oxygen consumption during pyrolysis and the separation performance between CMS membranes prepared at 550 C with two different thermal soak times of two and eight hours. First, 6FDA/BPDA-DAM polymer precursor was pyrolyzed with an eight hour soak time under 200 cc(STP)/min inert flow of 7 ppm oxygen in argon gas. The result was compared with data of a two hour thermal soak time previously reported [8]. Table 2 shows amounts of oxygen available and consumed during the pyrolysis process.
25
3
After the pyrolysis, the CMS membranes were immediately loaded into permeation cells. The permeability was measured using a constant-volume variable-pressure method [27,28]. Both upstream and downstream of the permeation system were evacuated for at least 12 h, and a leak rate was confirmed to be always less than 1% of the permeate rate of the slowest gas. Once the whole system was evacuated, the upstream was pressurized up to 50 psi with a test gas either CO2 or CH4 while the downstream was maintained at vacuum. The pressure rise in a standard volume on the downstream was monitored with time at 35 C by LabView (National Instruments) and permeability was calculated using Eq. (1).
(cm (STP)/(g of polymer precursor))
Characterization method
O2, consumed
3.2.
While the total amount of oxygen available during the eight hour soak time is higher than during the two hour soak, the consumption amount stays almost the same. Moreover, Fig. 3 shows that the data falls within the range of correlation observed in the previous study between the total amount of oxygen and the amount of oxygen consumption. This result indicates that a longer thermal soak time led to only a slight increase in the oxygen consumption. In addition, the separation performance of 6FDA/BPDADAM CMS membranes from two thermal soak periods was evaluated. As shown in Fig. 4, CMS membranes prepared with an eight hour soak time have a slightly higher CO2 permeability and lower CO2/CH4 selectivity compared with that of a two hour soak time. Based on these facts, it was speculated that the oxygen concentration may play a more major role during the doping process with oxygen than the period of soaking per se.
20 50 ppmO2 /Ar 2 hrs 15
30 ppmO2 /Ar 2 hrs 7ppmO2 /Ar
10
8 hrs
8 ppmO2 /Ar
5
q
described above. The value (ppm O2)Feed was the measured oxygen entering the pyrolysis chamber prior to the initiation of the pyrolysis. A duration of 720 min, time from onset of heating to cooling, was used to make an initial attempt to observe an oxygen consumption during pyrolysis of polymer membranes.
2 hrs 4 ppmO2 /Ar 2 hrs 0
Table 2 – Normalized values of the total amount of oxygen at different soak time obtained with 6FDA/BPDA-DAM. Inert pyrolysis under 200 cc/min of 7 ppm O2/Ar was conducted by eight hours soak time while the other was pyrolyzed for two hours 8 ppm O2/Ar. Experiments were repeated and have a standard deviation of less than 10%. Thermal soak time (h) 2 8
Total O2 available (ccSTP/g) 11.8 12.5
Total O2 consumed (ccSTP/g) 5.24 6.75
0
10
20 q
30
40
50
60
3
O2, tot
(cm (STP)/(g of polymer precursor)
Fig. 3 – Correlation between the total amount of oxygen and the amount of oxygen consumption normalized by weight of polymer precursors. Circles (d) represent data of two hour thermal soak time from previous study [8], and the rectangular (h) represents data of eight hour thermal soak time. Inert compositions are listed along with data points. The data with longer thermal soak time falls within the trend seen with two hour soak time.
4446
CARBON
4 8 ( 2 0 1 0 ) 4 4 4 2 –4 4 4 9
1000 4 ppmO2/Ar 8 ppmO2/Ar 30 ppmO2/Ar 50 ppm O2/Ar 7 ppmO2/Ar 8hrs Robeson 08 CO /CH Selectivity
100
30 ppm 8 ppm O2 /Ar O2 /Ar
4
7 ppm O2 /Ar 8hrs 4 ppm O2 /Ar
2
50 ppm O /Ar 10
2
1 100
1000
10
4
10
5
CO2 Permeability (Barrer)
Fig. 4 – Separation performance of 6FDA/BPDA-DAM CMS films. A filled red circle (d) represents CMS pyrolyzed with eight hour thermal soak time, and open circles (s) represent CMS pyrolyzed with two hour soak time. Each was repeated at 35 C.
Table 3 – Normalized values of the total amount of oxygen at different soak time obtained with Matrimid. Inert pyrolysis under 200 cc/min of 30 ppm O2/Ar was conducted in both cases. Experiments were repeated and have a standard deviation of less than 10%. Thermal soak time (h) 2 8
Total O2 available (ccSTP/g)
Total O2 consumed (ccSTP/g)
154.2 216.1
50.0 90.2
The effect of thermal soak time on Matrimid as a polymer precursor was also investigated. A pyrolysis atmosphere of 200 cc(STP)/min inert flow with 30 ppm oxygen in argon was used. Oxygen consumption and separation performance results are shown in Table 3 and Fig. 5, respectively. The results show that samples with a shorter soak time consumed a smaller amount of oxygen compared with a longer one; however, Fig. 5 indicates that their separation performances are similar. During the Matrimid pyrolysis process, we observed large amounts of tan colored by-products adsorbed on the pyrolysis tube wall that we did not see during the pyrolysis of 6FDA/BPDA-DAM. Based on this observation, we speculate that the difference in oxygen consumption could be caused by the oxidation of the by-products, which has little effect on the membrane properties. A combination of the oxygen consumption and the separation performance properties indicates that the effect of duration of oxygen exposure to the separation performance is relatively small. This led us to hypothesize that the oxygen doping process during the pyrolysis to produce attractive
Fig. 5 – Separation performance of Matrimid CMS films pyrolyzed with two different thermal soak times of two hours (d) and eight hours (s).The thickness of the films was 80 lm. Inert carrier of 30 ppm O2/Ar was used with a flow rate of 200 cc(STP)/min. All was repeated at 35 C.
Fig. 6 – Schematic of experiments to investigate external transfer limitation by applying two different inert flow rates. Total of 0.02 g polymeric films, m1 + m2 = m3 + m4 were pyrolyzed at 550 C with two hour soak time. Experiments were repeated twice.
CMS membranes is governed by oxygen concentration rather than total amount of oxygen exposed. A series of well controlled experiments was conducted to identify the limiting factor on the oxygen ‘‘doping’’ process to support this hypothesis. Three limiting factors are considered regarding the oxygen doping effect in dense films: external transport, internal transport, and chemical reaction. They were studied with respect to oxygen consumption and CO2/CH4 separation performance.
4.2.
External transfer limitation
Researchers have shown that inert flow rate during pyrolysis does not affect separation performances [4,5]. These studies showed that the external transport is not the limiting factor in determining the resulting CMS membrane performance. In order to demonstrate this phenomenon in terms of the oxygen exposure, two sets of experiments were conducted as shown in Fig. 6. A gas mixture of 30 ppm of oxygen in argon was used as an ‘‘inert’’ with two different flow rates: 50 and 200 cc(STP)/min. Two polymeric films (total of 0.02 g) were
CARBON
4447
4 8 ( 20 1 0 ) 4 4 4 2–44 4 9
Table 4 – Normalized values of the total amount of oxygen at different inert flow rates of 50 and 200 cc(STP)/min with Matrimid. Inert gas of 30 ppm O2/Ar was used in both cases. Experiments were repeated and have a standard deviation of less than 10%. Inert flowrate (cc(STP)/min) 50 200
Total O2 available (ccSTP/g) 67.3 154.2
Total O2 consumed (ccSTP/g) 45.4 50.0
Fig. 8 – Schematic design of experiments testing the internal mass transfer limitation. About 0.02 g of polymer precursor, m1 + m2 = m3, was prepared. Gas mixture of argon and a slightly higher than 30 ppm of oxygen in argon gas was used as an inert with the flow of 200 cc(STP)/min. Experiments were repeated twice.
Table 5 – Normalized values of the total amount of oxygen with different polymer precursor thickness. Inert gas of 35 ppm O2/Ar was used in both cases. Experiments were repeated and have a standard deviation of less than 10%. Thickness of polymer precursor (lm) 60 120
Total O2 available (ccSTP/g)
Total O2 consumed (ccSTP/g)
180 182
47 52
limited while by-products, or ‘‘molecular debris,’’ externally follow a kinetically limited reaction mechanism.
4.3.
Fig. 7 – Separation performance of Matrimid CMS films pyrolyzed with two different inert flow rates of 200 (d) and 50 (s) cc(STP)/min. Inert carrier of 30 ppm O2/Ar was used. All was repeated at 35 C.
pyrolyzed with the thermal protocol of 550 C and a two hour thermal soak time for consistency. It was hypothesized that if external mass transfer dominates the mechanism, oxygen consumption and separation performance would be dependent on the inert flow rates. Table 4 shows the result of oxygen consumption on the two different flow rates. This further confirms that the flow rates do not affect the amount of oxygen consumption as noted above. In addition, Fig. 7 shows that the separation performances are almost the same for different flow rates, which also supports our hypothesis that rate effects due to external transport resistance are negligible factors in fixing membrane performance. Clearly when CMS membranes experience longer thermal soak times, both the total amount of oxygen available and consumed should increase; however, consumption stayed almost consistent despite a higher inert (and oxygen) flow rate. Since the separation performances for the two different thermal soak times are almost the same, this discrepancy can be explained by suggesting that CMS oxidation may be equilibrium
Internal transfer limitation
The above investigation shows that the transport mechanism is not likely dominated by the external transport. Next, we conducted another investigation to determine any role of internal mass transfer limitations. This investigation consisted of two experiments as shown in Fig. 8. The first experiment consisted of pyrolysis of two polymer films (m1 and m2) which each film had a thickness of 60 lm. The second experiment consisted of a film (m3) whose mass was essentially a sum of m1 and m2 with a thickness of 120 lm. It was hypothesized that if the internal mass transfer dominated the mechanism, the oxygen consumption and separation performance would depend strongly on the film thickness. Recall that about four times longer exposure to a given amount oxygen showed almost no effect on CMS oxygen uptake, so these experiments sought to further probe if the process of doping is an equilibrium limited, internal reaction process. The results are shown in Table 5 and Fig. 9. Results of oxygen consumption in Table 5 indicate that oxygen consumption is almost the same regardless of the film thickness as long as the sample masses are the same. In addition, the separation performance of CO2/CH4 in Fig. 9 is similar as well. This indicates that the oxygen doping process is unlikely to be internal mass transfer limited.
4.4.
Chemical reaction limitation consideration
The above experiments show that the CMS membrane separation performance and oxygen consumption during the high temperature pyrolysis are not likely influenced by the thick-
4448
CARBON
4 8 ( 2 0 1 0 ) 4 4 4 2 –4 4 4 9
pyrolysis process. Once most of the by-products are evolved, the dehydrogenation process begins, but consumption of oxygen by the by-products continues at a significantly decreased rate, and a larger amount of oxygen became available to the CMS membranes compared with previous two stages. As Section 4.2 described, the oxidation of the by-products is likely controlled kinetically while the oxidation of the active carbon edges for ‘‘doping’’ is likely equilibrium controlled. Therefore, the likelihood of oxygen molecules adsorbing on the surface and penetrating through CMS membranes to react with active carbon edges depend on oxygen concentration in the bulk flow. The oxidation ‘‘doping’’ process also depends on the temperature which is related to the energetics of the chemisorption [10,30]. One should also note that a distribution of reactivity of carbon edges is speculated to exist to form carbonyl groups among ultramicropores, and the overall ‘‘doping’’ process is complex.
5. Fig. 9 – Separation performance of Matrimid CMS films pyrolyzed with two different precursor thickness, 60 (d) and 120 (s) lm. Inert carrier of 35 ppm O2/Ar was used. All was repeated at 35 C.
ness of the polymer precursors, inert, or oxygen, flow rates, nor thermal soak time during pyrolysis. The summation of all these facts implies that a carbon–oxygen equilibrium reaction governs the oxygen doping by a chemisorption process. The oxygen chemisorption mostly likely takes place at the same time polymer precursors decompose, and evolved products diffuse out of the membranes. This indicates that the actual full mechanism can be very complex.
4.5.
Possible mechanism
Previously, we successfully developed an oxygen doping process to tune separation properties of CMS membranes by controlling the CMS structure [8,9]. The method was built based on well-known scientific facts that (i) oxygen reacts with active carbon edges at high temperature during pyrolysis [13,16,19,29] and that (ii) the adsorption step which involves surface oxides dominate the reaction process and is endothermic and reversible in the temperature range between 350 and 700 C [10,15]. As described above, the oxygen–carbon mechanism can be very complex. Here we seek to understand the pyrolysis process of polymer membranes with oxygen exposure by normalization of the literature and our findings. Three stages are involved during the polymer decomposition process: precarbonation, carbonation, and dehydrogenation [3]. In this study, oxygen was continuously supplied during the pyrolysis process. Precarbonation mostly involves removal of excess solvent and monomer [3], and consumption of oxygen does not start until temperature is close to the decomposition temperature [8]. During the carbonation stages, it is believed that a majority of the oxygen was consumed by by-products. Meanwhile, a transformation from polymeric to CMS membranes takes place [3], and the ‘‘intrinsic’’ CMS structure results a product of the high temperature
Conclusions
Previously, a pyrolysis method that enables us to control separation performance of CMS membranes was developed. This method utilized oxygen chemisorption onto selective pore windows at high temperature. Literature has shown that the carbon–oxygen reaction is extremely sensitive to the limitations imposed by energy and mass transfer processes, and the mechanism is still not understood in detail [13,16]. In this study, the limiting factor of oxygen mass transfer in pyrolysis process was investigated. Specifically, effects of (i) thermal soak time, (ii) oxygen flow rate, and (iii) precursor thickness were studied by means of their oxygen consumption during pyrolysis and CO2/CH4 separation performance. The result indicated that the oxygen ‘‘doping’’ process on the carbon active pore windows is most likely limited by chemical reaction equilibrium. This finding is significant for practical applications, since the oxygen concentration, rather than the amount of exposure, may to be monitored for a better control of the separation performance.
Acknowledgment The authors acknowledge the financial support of Shell Global Solutions (US) Inc. (Houston, TX).
R E F E R E N C E S
[1] Baker RW. Membrane technology and applications. New York: McGraw-Hill; 2000. [2] Ismail AF, David LIB. A review on the latest development of carbon membranes for gas separation. J Membr Sci 2001;193(1):1–18. [3] Jenkins GM, Kawamura K. Polymeric carbons–carbon fiber, glass and char. London: Cambridge University Press; 1976. [4] Geiszler VC, Koros WJ. Effects of polyimide pyrolysis conditions on carbon molecular sieve membrane properties. Ind Eng Chem Res 1996;35(9):2999–3003. [5] Williams PJ. Analysis of factors influencing the performance of CMS membrane for gas separation. PhD thesis. Atlanta, GA, USA: Georgia Institute of Technology; 2006.
CARBON
4 8 ( 20 1 0 ) 4 4 4 2–44 4 9
[6] Steel KM. Carbon membranes for challenging gas separations. PhD thesis. Austin, TX, USA: The University of Texas at Austin; 2000. [7] Campo MC, Magalhaes FD, Mendes A. Comparative study between a CMS membrane and a CMS adsorbent: Part I – morphology, adsorption equilibrium and kinetics. J Membr Sci 2010;346(1):15–25. [8] Kiyono M, Williams PJ, Koros WJ. Effect of pyrolysis atmosphere on separation performance of carbon molecular sieve membranes. J Membr Sci 2010;359(1–2):2–10. [9] Kiyono M, Williams PJ, Koros WJ. Effect of polymer precursor on carbon molecular sieve structure and separation performance properties. Carbon, in press, doi:10.1016/ j.carbon.2010.08.002. [10] de Soete G. Catalysis of soot combustion by metal oxides. Western States Section Meeting of the Combustion Institute, 1988 March 21–22, Salt Lake City, UT, USA; 1988. [11] Grisdale RO. The properties of carbon contacts. J Appl Phys 1953;24(10):1288–96. [12] Ishiguro T, Suzuki N, Fujitani Y, Morimoto H. Microstructural changes of diesel soot during oxidation. Combust Flame 1991;85(1–2):1–6. [13] Marsh H. Introduction to carbon science. Borough Green, England: Butterworth and Co. Ltd.; 1989. [14] Walker Jr PL, Rusinko Jr F, Austin LG. Gas reactions of carbon. In: Eley DD, Frankenburg WG, editors. Advances in catalysis and related subjects. New York: Academic Press Inc.; 1959. p. 133–76. [15] Ong Jr JN. On the kinetics of oxidation of graphite. Carbon 1964;2(3):281–97. [16] Stanmore BR, Brilhac JF, Gilot P. The oxidation of soot: a review of experiments, mechanisms and models. Carbon 2001;38(15):2247–68. [17] Pierson HO. Handbook of carbon, graphite, diamond, and fullerenes. New York: Noyes Publication; 1993. [18] Jones CW, Koros WJ. Carbon molecular sieve gas separation membranes-I. Preparation and characterizaton based on polyimide precursors. Carbon 1994;32(8):1419–25.
4449
[19] Kinoshita K. Carbon: electrochemical and physicochemical properties. Berkeley, CA, USA: Wiley-Interscience; 1988. [20] Smith BC. Infrared spectral interpretation: a systematic approach. 1st ed. New York: CRC Press; 1999. [21] Welty JR, Wicks CE, Wilson RE, Rorrer GL. Fundamentals of momentum, heat, and mass transfer. New York: John Wiley and Sons, Inc.; 2001. [22] Levenspiel O. Chemical reaction engineering. 3rd ed. Hoboken, NJ: John Wiley and Sons, Inc.; 1999. p. 376–417. [23] Lear AE, Brown TC, Haynes BS. Formation of metastable oxide comples during the oxidation of carbons at low temperatures. In: Twenty-third symposium (international) on combustion; 1990; at the University of Orle´ans, Orle´ans, France, 1990. p. 1191–7. [24] Singh A. Membrane materials with enhanced selectivity: an entropic interpretation. PhD thesis. Austin, TX, USA: The University of Texas at Austin; 1997. [25] Neeft J, Nijhuis T, Smakman E, Makkee M, Moulijn J. Kinetics of the oxidation of diesel soot. Fuel 1997;76(12):1129–36. [26] Park HB, Nam SY, Jang JG, Lee YM. Preparation, characterization, and gas permeation properties of carbon molecular sieve membranes derived from dense P84polyimide film. Kor Membr J 2002;4(1):25–35. [27] Pye DG, Hoehn HH, Panar M. Measurement of gas permeability of polymers. I. Permeabilities in constant volume/variable pressure apparatus. J Appl Polym Sci 1976;20(7):1921–31. [28] Pye DG, Hoehn HH, Panar M. Measurement of gas permeability of polymers. II. Apparatus for the determination of permeabilities of mixed gases and vapors. J Appl Polym Sci 1976;20(2):287–301. [29] Laine NR, Vastola FJ, Walker Jr PL. The importance of active surface area in the carbon–oxygen reaction. J Phys Chem 1963;67(10):2030–4. [30] Hill Jr CG. An introduction to chemical engineering kinetics and reactor design. New York: John Wiley and Sons, Inc.; 1977.
4 8 ( 2 0 1 0 ) 4 4 4 2 –4 4 4 9
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/carbon
Generalization of effect of oxygen exposure on formation and performance of carbon molecular sieve membranes Mayumi Kiyono, Paul J. Williams, William J. Koros
*
Department of Chemical and Biomolecular Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0001, USA Shell Projects and Technology, 3333 Highway 6 South, Houston, TX 77082, USA
A R T I C L E I N F O
A B S T R A C T
Article history:
Previously, a pyrolysis method was developed to control separation performance of carbon
Received 1 May 2010
molecular sieve (CMS) membranes by doping with trace oxygen. This method involved oxy-
Accepted 3 August 2010
gen exposure during pyrolysis to tune the selective pore windows. During the development,
Available online 6 August 2010
it was observed that oxygen concentration in inert gas, rather than the total amount of oxygen exposed controls performance. In this study, we hypothesized that mass transfer of oxygen in CMS membranes during pyrolysis is governed by chemical reaction at critical pore opening. Effect of thermal soak time, inert flow rate, and precursor thickness were conducted to test this hypothesis with 6FDA/BPDA-DAM and Matrimid polymer precursors. Results of oxygen consumption from pyrolysis process and CO2/CH4 separation performance showed that the process is likely governed by reaction kinetics. This observation implies simplicity and easy scale-up for the oxygen ‘‘doping’’ method on CMS formation by tuning the oxygen concentration in the pyrolysis atmosphere. Published by Elsevier Ltd.
1.
Introduction
Carbon molecular sieve (CMS) membranes are known to have attractive gas separation performance, well exceeding the polymer upper bound curve for challenging gas pairs, such as O2/N2, CO2/CH4, C3H6/C3H8 [1,2]. CMS membranes are formed by thermal decomposition of polymer precursors. As a result of this thermal decomposition process, amorphous turbostratic carbon membranes are usually formed [3]. Pores are formed from packing imperfections between microcrystalline regions in the material [4,5], and the pore structure in CMS membranes is described as ‘‘slit-like’’ with a bimodal pore distribution in which large pores connected by smaller pores known as ‘‘ultramicropores’’ [6,7]. This combination of ultramicropores and micropores provides the molecular sieving function. While traditional vacuum pyrolysis produces CMS membranes with attractive separation performance, the approaches are difficult to scale up economically.
* Corresponding author: Fax: +1 4043852683. E-mail address: [email protected] (W.J. Koros). 0008-6223/$ - see front matter Published by Elsevier Ltd. doi:10.1016/j.carbon.2010.08.003
With practical impact in mind, we extended our previously developed inert pyrolysis method with trace oxygen exposure using synthesized polymer 6FDA/BPDA-DAM [8] to allow use of economical precursors. Specifically we demonstrated the ability to use a commercially available polymer Matrimid precursor, and further identified factors to improve the efficacy of the method based on intrinsic, oxygen free pyrolysis case, structure of CMS membranes [9]. The method essentially utilizes oxygen chemisorption at selective pore windows during the high temperature pyrolysis process. Numerous studies have investigated the oxygen–carbon reaction process for various kinds of carbon materials [10–15]. When the carbon–oxygen reaction is believed to produce C@O chemisorptions entities on the surface, it is difficult to identify and control their true nature due to extreme sensitivity to the limitations imposed by energy and mass transfer processes [13,16]. When the oxygen doping method was initially developed, it was assumed that the total amount of oxy-
CARBON
4443
4 8 ( 20 1 0 ) 4 4 4 2–44 4 9
gen presented at the sample controlled the amount of chemisorption at selective pore windows. Based on surprising results discussed here, we now present evidence that the ultimate structure are more likely controlled by oxygen concentration in the inert gas at the temperature of the pyrolysis, rather than a total accessible oxygen, per se. This suggests an equilibrium limited structure an opposed to a more conventionally anticipated kinetically-controlled structure. If true, such a structure would be advantageous for practical processing. In this work, we seek to identify the controlling factor of the oxygen ‘‘doping’’ process during formation of CMS membranes. Specifically this study explains a series of well controlled experiments to determine the controlling factor governing the oxygen–carbon reaction by means of oxygen consumption during pyrolysis and separation performance of resulting CMS membranes. This identification is essential for transitioning from laboratory to industrial scale for the mass production of CMS membranes.
2.
Theory and background
2.1.
Structure and transport of carbon molecular sieve
In this study, CMS membranes are formed by thermal decomposition of polymeric membranes. When polymers are pyrolyzed, either coke or char is formed [3,17]. While coke forms graphite at a temperature above 2200 C, char remains in an amorphous structure [3,17]. Most CMS membranes are char, having a complex turbostratic structure consisting of parallel layers of condensed hexagonal rings with no three-dimensional crystalline order, as shown in Fig. 1a and conveniently represented by Fig. 1b for engineering purposes [3]. Pores of CMS membranes are described as slit-like, and have a bimodal pore size distribution made of ultramicropores and micropores [6]. The CMS structures as well as resulting separation performance properties of CMS membranes strongly depend on precursor property and pyrolysis process [2,5,6,18]. Like polymeric membrane systems, the sorption–diffusion mechanism is used to explain gas transport through carbon molecular sieve (CMS) membranes. Gas molecules sorb at the upstream, diffuse into the membrane under the influence of a chemical potential gradient, and desorb from the downstream. There are two parameters commonly used to measure separation performance of membranes. Permeability is used to measure the intrinsic productivity of a membrane
material, and is equal to the pressure and thickness normalized flux, viz., Pi ¼
ð1Þ
In the equation above, ni represents the flux of component ‘‘i’’ gas molecules through the membrane of the membrane thickness, l, and Dpi is the transmembrane partial pressure difference that acts as the driving force for component i across the membrane. The most common unit for permeability is the Barrer which is defined as: 1 Barrer½¼1010
cm3 ðSTPÞ cm : cm2 s cmHg
ð2Þ
An additional popular unit, kmol m m2 s1 kPa1, can be obtained by multiplying Barrers by 2.99 · 1015. Another parameter, selectivity, is a measure of the membrane’s separation efficiency and equals the ratio of the component permeabilities for the case of a negligible downstream permeate pressure.
2.2.
Oxygen ‘‘doping’’ method
As an alternative to the traditional vacuum pyrolysis, the authors developed a tuning method with utilizes oxidation during inert pyrolysis process [8]. The method relies on the fact that oxygen may chemisorb and form a stable bond at the selective ultramicropore windows at high temperature [8]. The method was based on evidence that: (i) the selective pores have sp2 carbon edges that are more reactive than basal plane [11,13,19], (ii) oxygen chemisorption is the limiting step of doping process [10], (iii) the bond is strong and stable and it does not break unless it experiences the same heating treatment responsible for its original formation [13]. Therefore, once applied, this doping process makes the sieving structure less permeable and more selective as the amount of doping increases up to a point before actual large scale oxidation occurs. Synthesized polymer, 6FDA/BPDA-DAM, was used to test the method. Separation performance of CO2 and CH4 supported the hypothesis that permeability decreases with an increase in selectivity up to a point when excessive doping makes the sieve lose both permeability and selectivity. This method was also demonstrated with a commercially available polymer, Matrimid, and FT-IR showed evidence of chemisorbed oxygen as C@O bonds [9].
2.3.
Fig. 1 – (a) Turbostratic structures of CMS membranes, adapted Pierson [17] and (b) an idealized slit-like structure.
ni l : Dpi
Oxidation mechanism
The oxygen ‘‘doping’’ method allows one to enhance membrane performance by >100 times in permeability with doubled selectivity when CMS pyrolysis atmosphere is optimized, compared with separation performance of polymeric precursors [8,9]. Oxidation of carbon is complex due to the fact that the oxygen–carbon reaction involves chemical kinetics with consequent heat transfer and mass transfer processes at a number of levels. Gaseous oxygen molecules move from the surrounding bulk atmosphere to the carbon surface and are adsorbed to form surface intermediates, which may rearrange, desorb, and return to the gas phase [13,16,19]. Moreover, emitted decomposition by-products can be reacted
4444
CARBON
4 8 ( 2 0 1 0 ) 4 4 4 2 –4 4 4 9
in the external bulk gas phase. This study seeks to identify the limiting factor of the doping process during formation of CMS membranes. Three regimes are considered during the oxidation process: (i) oxygen fully penetrates the solid and all active sites are available for the reaction, (ii) oxidant penetration is partial and oxygen diffusion into the solid is insufficient to supply all reaction needs, and (iii) reaction only takes place at the outer surface [20,21] for the available contact time. Investigation of the influence of diffusional resistance relative to reaction resistance can be carried out using a fundamental analysis of the Thiele modulus, h, which is defined as: 1 d rP kb 2 ð3Þ h¼ 2 De rOX where d is the particle diameter in meters, rP is its density in kg/m3, k is the carbon reaction rate in kg/kg s, b is the stoichiometric mass ratio of oxygen to carbon, De is the effective diffusion coefficient in m2/s, and rOX is the bulk gas concentration of oxygen in kg/m3 [14]. The Thiele modulus is applied by means of an effectiveness factor, which varies between 0 and 1 [14,22]. The effectiveness factor represents the fraction of internal surface which can react when exposed to the surface concentration of reactant gas [22]. When the Thiele modulus is larger than 1, the reaction rate hinders the ability of diffusion to supply oxygen to the reactive surface and the effectiveness factor is much smaller than 1. When the Thiele modulus is smaller than unity, there is no resistance to pore diffusion. In terms of formation of CMS membranes, factors like pyrolysis temperature, film thickness, and pore structure strongly influence the process and determine the three regimes described earlier. A preliminary calculation can be made to identify one of three oxidation process regimes from the Thiele modules in Eq. (3). The bulk oxygen concentration, rOX, of our interests is in 1–100 ppm, which corresponds to 1.41 · 106–1.41 · 104 kg/m3. The density, rP, of CMS membranes is reported to be in the order of 103 kg/m3, depending on polymeric precursors [6]. At temperatures of 500–800 C, where typical CMS membranes are produced, carbon oxidation produces CO as the dominant product [23]. According to Stantmore et al., this results in a b value of 4/3 [16]. Effective diffusion coefficients of oxygen in CMS membranes at high temperatures are in the order of 108 m2/s, according to Singh [24]. According to de Soete, the reaction is in the first order when oxygen adsorption dominates, and overall reaction rate can be in the range of 0.5–0.9 s1 [10,25]. When either homogeneous dense films or asymmetric hollow fibers have a selective skin thickness of d = 0.1–30 lm, calculation of the Thiele modulus based on these literature values results in a value of 0.6–2500. Since this range clearly spans values less than and greater than one, it suggests the possibility of pore diffusion resistance control of the oxidation reaction. Nevertheless, there is considerable uncertainty in the above parameters, so this study probes this issue experimentally in detail. While the above theoretical calculation suggests possible pore diffusion resistance of oxygen during high temperature
1
Table 1 – Pyrolysis temperature protocol. Tinitial (C)
Tfinal (C)
25 250 535 550
Ramp rate (C/min)
250 535 550 550
13.3 3.85 0.25 2 h soak
pyrolysis in carbon membranes, some researchers have seen inconsistent separation performance of CMS membranes pyrolyzed under atmosphere in which oxygen exists for possible oxidation reaction. Variations have been reported in CMS separation performance with respect to polymer precursor thickness [5], and researchers have seen separation performance independent of inert flow rates [4,5,26]: observations and the theoretical calculation have not reached an agreement. Therefore we seek to draw a more definite conclusion by identifying the controlling factor in the oxygen doping process by a series of well controlled experiments, specifically, by means of oxygen consumption during pyrolysis and separation performance of the resulting CMS membranes.
3.
Experimental
3.1.
Material
Two polymer materials were chosen: commercially available polyimide Matrimid from Vantico, Inc. and synthesized polyimide 6FDA/BPDA(1:1)-DAM. Their chemical structures, properties, and details of synthesis process are described elsewhere [8,9]. Polymeric dense films were prepared by the solution casting method, and they were heated accordingly with a pyrolysis protocol shown in Table 1 using the pyrolysis system shown in Fig. 2. A mixture of argon and various amounts of oxygen in ppm levels were used as an inert gas. The system was purged for at least 10 times the volume of the whole system to reach equilibrium before each pyrolysis experiment. During the pyrolysis process, oxygen concentration levels were monitored using an oxygen analyzer (Cambridge Sensotec Ltd., Rapidox 2100 series, Cambridge, England with ±1% accuracy between 1020 ppm and 100%1). Accuracy of the analyzer was confirmed by calibrating with air before each experiment, and the consistent oxygen level was checked before and after the pyrolysis process. Once an oxygen concentration profile with respect to time was obtained, the total amount of oxygen available and the amount of oxygen consumed were calculated using Eqs. (4) and (5) Total amount of O2 ¼ Q_ purge ½ðppm O2 ÞFeed t Amount of O2 consumed ¼ Q_ purge ½ðppm O Þ
ð4Þ
2 Feed
ðppm O2 ÞMeasured t
ð5Þ
The parameter Q_ purge is the volumetric flow rate of the inert gas and was measured by a bubble flow meter. Values (ppm O2)Feed and (ppm O2)Measured were obtained by the oxygen analyzer
Rapidox 2100 oxygen analyzer instruction manual. Cambridge, England: Cambridge Sensotec Limited; 2004.
CARBON
4445
4 8 ( 20 1 0 ) 4 4 4 2–44 4 9
Micro needle valve
O2 Sensor
Inert Gas Mass Flow Controller
Vent
Quartz Tube Thermocouple
Data Acquisition Temperature Controller
Fig. 2 – Pyrolysis set-up for CMS membrane preparation.
4.
Results and discussion
4.1.
Effect of thermal soak time
Effect of thermal soak time on the oxygen doping process was investigated by comparing oxygen consumption during pyrolysis and the separation performance between CMS membranes prepared at 550 C with two different thermal soak times of two and eight hours. First, 6FDA/BPDA-DAM polymer precursor was pyrolyzed with an eight hour soak time under 200 cc(STP)/min inert flow of 7 ppm oxygen in argon gas. The result was compared with data of a two hour thermal soak time previously reported [8]. Table 2 shows amounts of oxygen available and consumed during the pyrolysis process.
25
3
After the pyrolysis, the CMS membranes were immediately loaded into permeation cells. The permeability was measured using a constant-volume variable-pressure method [27,28]. Both upstream and downstream of the permeation system were evacuated for at least 12 h, and a leak rate was confirmed to be always less than 1% of the permeate rate of the slowest gas. Once the whole system was evacuated, the upstream was pressurized up to 50 psi with a test gas either CO2 or CH4 while the downstream was maintained at vacuum. The pressure rise in a standard volume on the downstream was monitored with time at 35 C by LabView (National Instruments) and permeability was calculated using Eq. (1).
(cm (STP)/(g of polymer precursor))
Characterization method
O2, consumed
3.2.
While the total amount of oxygen available during the eight hour soak time is higher than during the two hour soak, the consumption amount stays almost the same. Moreover, Fig. 3 shows that the data falls within the range of correlation observed in the previous study between the total amount of oxygen and the amount of oxygen consumption. This result indicates that a longer thermal soak time led to only a slight increase in the oxygen consumption. In addition, the separation performance of 6FDA/BPDADAM CMS membranes from two thermal soak periods was evaluated. As shown in Fig. 4, CMS membranes prepared with an eight hour soak time have a slightly higher CO2 permeability and lower CO2/CH4 selectivity compared with that of a two hour soak time. Based on these facts, it was speculated that the oxygen concentration may play a more major role during the doping process with oxygen than the period of soaking per se.
20 50 ppmO2 /Ar 2 hrs 15
30 ppmO2 /Ar 2 hrs 7ppmO2 /Ar
10
8 hrs
8 ppmO2 /Ar
5
q
described above. The value (ppm O2)Feed was the measured oxygen entering the pyrolysis chamber prior to the initiation of the pyrolysis. A duration of 720 min, time from onset of heating to cooling, was used to make an initial attempt to observe an oxygen consumption during pyrolysis of polymer membranes.
2 hrs 4 ppmO2 /Ar 2 hrs 0
Table 2 – Normalized values of the total amount of oxygen at different soak time obtained with 6FDA/BPDA-DAM. Inert pyrolysis under 200 cc/min of 7 ppm O2/Ar was conducted by eight hours soak time while the other was pyrolyzed for two hours 8 ppm O2/Ar. Experiments were repeated and have a standard deviation of less than 10%. Thermal soak time (h) 2 8
Total O2 available (ccSTP/g) 11.8 12.5
Total O2 consumed (ccSTP/g) 5.24 6.75
0
10
20 q
30
40
50
60
3
O2, tot
(cm (STP)/(g of polymer precursor)
Fig. 3 – Correlation between the total amount of oxygen and the amount of oxygen consumption normalized by weight of polymer precursors. Circles (d) represent data of two hour thermal soak time from previous study [8], and the rectangular (h) represents data of eight hour thermal soak time. Inert compositions are listed along with data points. The data with longer thermal soak time falls within the trend seen with two hour soak time.
4446
CARBON
4 8 ( 2 0 1 0 ) 4 4 4 2 –4 4 4 9
1000 4 ppmO2/Ar 8 ppmO2/Ar 30 ppmO2/Ar 50 ppm O2/Ar 7 ppmO2/Ar 8hrs Robeson 08 CO /CH Selectivity
100
30 ppm 8 ppm O2 /Ar O2 /Ar
4
7 ppm O2 /Ar 8hrs 4 ppm O2 /Ar
2
50 ppm O /Ar 10
2
1 100
1000
10
4
10
5
CO2 Permeability (Barrer)
Fig. 4 – Separation performance of 6FDA/BPDA-DAM CMS films. A filled red circle (d) represents CMS pyrolyzed with eight hour thermal soak time, and open circles (s) represent CMS pyrolyzed with two hour soak time. Each was repeated at 35 C.
Table 3 – Normalized values of the total amount of oxygen at different soak time obtained with Matrimid. Inert pyrolysis under 200 cc/min of 30 ppm O2/Ar was conducted in both cases. Experiments were repeated and have a standard deviation of less than 10%. Thermal soak time (h) 2 8
Total O2 available (ccSTP/g)
Total O2 consumed (ccSTP/g)
154.2 216.1
50.0 90.2
The effect of thermal soak time on Matrimid as a polymer precursor was also investigated. A pyrolysis atmosphere of 200 cc(STP)/min inert flow with 30 ppm oxygen in argon was used. Oxygen consumption and separation performance results are shown in Table 3 and Fig. 5, respectively. The results show that samples with a shorter soak time consumed a smaller amount of oxygen compared with a longer one; however, Fig. 5 indicates that their separation performances are similar. During the Matrimid pyrolysis process, we observed large amounts of tan colored by-products adsorbed on the pyrolysis tube wall that we did not see during the pyrolysis of 6FDA/BPDA-DAM. Based on this observation, we speculate that the difference in oxygen consumption could be caused by the oxidation of the by-products, which has little effect on the membrane properties. A combination of the oxygen consumption and the separation performance properties indicates that the effect of duration of oxygen exposure to the separation performance is relatively small. This led us to hypothesize that the oxygen doping process during the pyrolysis to produce attractive
Fig. 5 – Separation performance of Matrimid CMS films pyrolyzed with two different thermal soak times of two hours (d) and eight hours (s).The thickness of the films was 80 lm. Inert carrier of 30 ppm O2/Ar was used with a flow rate of 200 cc(STP)/min. All was repeated at 35 C.
Fig. 6 – Schematic of experiments to investigate external transfer limitation by applying two different inert flow rates. Total of 0.02 g polymeric films, m1 + m2 = m3 + m4 were pyrolyzed at 550 C with two hour soak time. Experiments were repeated twice.
CMS membranes is governed by oxygen concentration rather than total amount of oxygen exposed. A series of well controlled experiments was conducted to identify the limiting factor on the oxygen ‘‘doping’’ process to support this hypothesis. Three limiting factors are considered regarding the oxygen doping effect in dense films: external transport, internal transport, and chemical reaction. They were studied with respect to oxygen consumption and CO2/CH4 separation performance.
4.2.
External transfer limitation
Researchers have shown that inert flow rate during pyrolysis does not affect separation performances [4,5]. These studies showed that the external transport is not the limiting factor in determining the resulting CMS membrane performance. In order to demonstrate this phenomenon in terms of the oxygen exposure, two sets of experiments were conducted as shown in Fig. 6. A gas mixture of 30 ppm of oxygen in argon was used as an ‘‘inert’’ with two different flow rates: 50 and 200 cc(STP)/min. Two polymeric films (total of 0.02 g) were
CARBON
4447
4 8 ( 20 1 0 ) 4 4 4 2–44 4 9
Table 4 – Normalized values of the total amount of oxygen at different inert flow rates of 50 and 200 cc(STP)/min with Matrimid. Inert gas of 30 ppm O2/Ar was used in both cases. Experiments were repeated and have a standard deviation of less than 10%. Inert flowrate (cc(STP)/min) 50 200
Total O2 available (ccSTP/g) 67.3 154.2
Total O2 consumed (ccSTP/g) 45.4 50.0
Fig. 8 – Schematic design of experiments testing the internal mass transfer limitation. About 0.02 g of polymer precursor, m1 + m2 = m3, was prepared. Gas mixture of argon and a slightly higher than 30 ppm of oxygen in argon gas was used as an inert with the flow of 200 cc(STP)/min. Experiments were repeated twice.
Table 5 – Normalized values of the total amount of oxygen with different polymer precursor thickness. Inert gas of 35 ppm O2/Ar was used in both cases. Experiments were repeated and have a standard deviation of less than 10%. Thickness of polymer precursor (lm) 60 120
Total O2 available (ccSTP/g)
Total O2 consumed (ccSTP/g)
180 182
47 52
limited while by-products, or ‘‘molecular debris,’’ externally follow a kinetically limited reaction mechanism.
4.3.
Fig. 7 – Separation performance of Matrimid CMS films pyrolyzed with two different inert flow rates of 200 (d) and 50 (s) cc(STP)/min. Inert carrier of 30 ppm O2/Ar was used. All was repeated at 35 C.
pyrolyzed with the thermal protocol of 550 C and a two hour thermal soak time for consistency. It was hypothesized that if external mass transfer dominates the mechanism, oxygen consumption and separation performance would be dependent on the inert flow rates. Table 4 shows the result of oxygen consumption on the two different flow rates. This further confirms that the flow rates do not affect the amount of oxygen consumption as noted above. In addition, Fig. 7 shows that the separation performances are almost the same for different flow rates, which also supports our hypothesis that rate effects due to external transport resistance are negligible factors in fixing membrane performance. Clearly when CMS membranes experience longer thermal soak times, both the total amount of oxygen available and consumed should increase; however, consumption stayed almost consistent despite a higher inert (and oxygen) flow rate. Since the separation performances for the two different thermal soak times are almost the same, this discrepancy can be explained by suggesting that CMS oxidation may be equilibrium
Internal transfer limitation
The above investigation shows that the transport mechanism is not likely dominated by the external transport. Next, we conducted another investigation to determine any role of internal mass transfer limitations. This investigation consisted of two experiments as shown in Fig. 8. The first experiment consisted of pyrolysis of two polymer films (m1 and m2) which each film had a thickness of 60 lm. The second experiment consisted of a film (m3) whose mass was essentially a sum of m1 and m2 with a thickness of 120 lm. It was hypothesized that if the internal mass transfer dominated the mechanism, the oxygen consumption and separation performance would depend strongly on the film thickness. Recall that about four times longer exposure to a given amount oxygen showed almost no effect on CMS oxygen uptake, so these experiments sought to further probe if the process of doping is an equilibrium limited, internal reaction process. The results are shown in Table 5 and Fig. 9. Results of oxygen consumption in Table 5 indicate that oxygen consumption is almost the same regardless of the film thickness as long as the sample masses are the same. In addition, the separation performance of CO2/CH4 in Fig. 9 is similar as well. This indicates that the oxygen doping process is unlikely to be internal mass transfer limited.
4.4.
Chemical reaction limitation consideration
The above experiments show that the CMS membrane separation performance and oxygen consumption during the high temperature pyrolysis are not likely influenced by the thick-
4448
CARBON
4 8 ( 2 0 1 0 ) 4 4 4 2 –4 4 4 9
pyrolysis process. Once most of the by-products are evolved, the dehydrogenation process begins, but consumption of oxygen by the by-products continues at a significantly decreased rate, and a larger amount of oxygen became available to the CMS membranes compared with previous two stages. As Section 4.2 described, the oxidation of the by-products is likely controlled kinetically while the oxidation of the active carbon edges for ‘‘doping’’ is likely equilibrium controlled. Therefore, the likelihood of oxygen molecules adsorbing on the surface and penetrating through CMS membranes to react with active carbon edges depend on oxygen concentration in the bulk flow. The oxidation ‘‘doping’’ process also depends on the temperature which is related to the energetics of the chemisorption [10,30]. One should also note that a distribution of reactivity of carbon edges is speculated to exist to form carbonyl groups among ultramicropores, and the overall ‘‘doping’’ process is complex.
5. Fig. 9 – Separation performance of Matrimid CMS films pyrolyzed with two different precursor thickness, 60 (d) and 120 (s) lm. Inert carrier of 35 ppm O2/Ar was used. All was repeated at 35 C.
ness of the polymer precursors, inert, or oxygen, flow rates, nor thermal soak time during pyrolysis. The summation of all these facts implies that a carbon–oxygen equilibrium reaction governs the oxygen doping by a chemisorption process. The oxygen chemisorption mostly likely takes place at the same time polymer precursors decompose, and evolved products diffuse out of the membranes. This indicates that the actual full mechanism can be very complex.
4.5.
Possible mechanism
Previously, we successfully developed an oxygen doping process to tune separation properties of CMS membranes by controlling the CMS structure [8,9]. The method was built based on well-known scientific facts that (i) oxygen reacts with active carbon edges at high temperature during pyrolysis [13,16,19,29] and that (ii) the adsorption step which involves surface oxides dominate the reaction process and is endothermic and reversible in the temperature range between 350 and 700 C [10,15]. As described above, the oxygen–carbon mechanism can be very complex. Here we seek to understand the pyrolysis process of polymer membranes with oxygen exposure by normalization of the literature and our findings. Three stages are involved during the polymer decomposition process: precarbonation, carbonation, and dehydrogenation [3]. In this study, oxygen was continuously supplied during the pyrolysis process. Precarbonation mostly involves removal of excess solvent and monomer [3], and consumption of oxygen does not start until temperature is close to the decomposition temperature [8]. During the carbonation stages, it is believed that a majority of the oxygen was consumed by by-products. Meanwhile, a transformation from polymeric to CMS membranes takes place [3], and the ‘‘intrinsic’’ CMS structure results a product of the high temperature
Conclusions
Previously, a pyrolysis method that enables us to control separation performance of CMS membranes was developed. This method utilized oxygen chemisorption onto selective pore windows at high temperature. Literature has shown that the carbon–oxygen reaction is extremely sensitive to the limitations imposed by energy and mass transfer processes, and the mechanism is still not understood in detail [13,16]. In this study, the limiting factor of oxygen mass transfer in pyrolysis process was investigated. Specifically, effects of (i) thermal soak time, (ii) oxygen flow rate, and (iii) precursor thickness were studied by means of their oxygen consumption during pyrolysis and CO2/CH4 separation performance. The result indicated that the oxygen ‘‘doping’’ process on the carbon active pore windows is most likely limited by chemical reaction equilibrium. This finding is significant for practical applications, since the oxygen concentration, rather than the amount of exposure, may to be monitored for a better control of the separation performance.
Acknowledgment The authors acknowledge the financial support of Shell Global Solutions (US) Inc. (Houston, TX).
R E F E R E N C E S
[1] Baker RW. Membrane technology and applications. New York: McGraw-Hill; 2000. [2] Ismail AF, David LIB. A review on the latest development of carbon membranes for gas separation. J Membr Sci 2001;193(1):1–18. [3] Jenkins GM, Kawamura K. Polymeric carbons–carbon fiber, glass and char. London: Cambridge University Press; 1976. [4] Geiszler VC, Koros WJ. Effects of polyimide pyrolysis conditions on carbon molecular sieve membrane properties. Ind Eng Chem Res 1996;35(9):2999–3003. [5] Williams PJ. Analysis of factors influencing the performance of CMS membrane for gas separation. PhD thesis. Atlanta, GA, USA: Georgia Institute of Technology; 2006.
CARBON
4 8 ( 20 1 0 ) 4 4 4 2–44 4 9
[6] Steel KM. Carbon membranes for challenging gas separations. PhD thesis. Austin, TX, USA: The University of Texas at Austin; 2000. [7] Campo MC, Magalhaes FD, Mendes A. Comparative study between a CMS membrane and a CMS adsorbent: Part I – morphology, adsorption equilibrium and kinetics. J Membr Sci 2010;346(1):15–25. [8] Kiyono M, Williams PJ, Koros WJ. Effect of pyrolysis atmosphere on separation performance of carbon molecular sieve membranes. J Membr Sci 2010;359(1–2):2–10. [9] Kiyono M, Williams PJ, Koros WJ. Effect of polymer precursor on carbon molecular sieve structure and separation performance properties. Carbon, in press, doi:10.1016/ j.carbon.2010.08.002. [10] de Soete G. Catalysis of soot combustion by metal oxides. Western States Section Meeting of the Combustion Institute, 1988 March 21–22, Salt Lake City, UT, USA; 1988. [11] Grisdale RO. The properties of carbon contacts. J Appl Phys 1953;24(10):1288–96. [12] Ishiguro T, Suzuki N, Fujitani Y, Morimoto H. Microstructural changes of diesel soot during oxidation. Combust Flame 1991;85(1–2):1–6. [13] Marsh H. Introduction to carbon science. Borough Green, England: Butterworth and Co. Ltd.; 1989. [14] Walker Jr PL, Rusinko Jr F, Austin LG. Gas reactions of carbon. In: Eley DD, Frankenburg WG, editors. Advances in catalysis and related subjects. New York: Academic Press Inc.; 1959. p. 133–76. [15] Ong Jr JN. On the kinetics of oxidation of graphite. Carbon 1964;2(3):281–97. [16] Stanmore BR, Brilhac JF, Gilot P. The oxidation of soot: a review of experiments, mechanisms and models. Carbon 2001;38(15):2247–68. [17] Pierson HO. Handbook of carbon, graphite, diamond, and fullerenes. New York: Noyes Publication; 1993. [18] Jones CW, Koros WJ. Carbon molecular sieve gas separation membranes-I. Preparation and characterizaton based on polyimide precursors. Carbon 1994;32(8):1419–25.
4449
[19] Kinoshita K. Carbon: electrochemical and physicochemical properties. Berkeley, CA, USA: Wiley-Interscience; 1988. [20] Smith BC. Infrared spectral interpretation: a systematic approach. 1st ed. New York: CRC Press; 1999. [21] Welty JR, Wicks CE, Wilson RE, Rorrer GL. Fundamentals of momentum, heat, and mass transfer. New York: John Wiley and Sons, Inc.; 2001. [22] Levenspiel O. Chemical reaction engineering. 3rd ed. Hoboken, NJ: John Wiley and Sons, Inc.; 1999. p. 376–417. [23] Lear AE, Brown TC, Haynes BS. Formation of metastable oxide comples during the oxidation of carbons at low temperatures. In: Twenty-third symposium (international) on combustion; 1990; at the University of Orle´ans, Orle´ans, France, 1990. p. 1191–7. [24] Singh A. Membrane materials with enhanced selectivity: an entropic interpretation. PhD thesis. Austin, TX, USA: The University of Texas at Austin; 1997. [25] Neeft J, Nijhuis T, Smakman E, Makkee M, Moulijn J. Kinetics of the oxidation of diesel soot. Fuel 1997;76(12):1129–36. [26] Park HB, Nam SY, Jang JG, Lee YM. Preparation, characterization, and gas permeation properties of carbon molecular sieve membranes derived from dense P84polyimide film. Kor Membr J 2002;4(1):25–35. [27] Pye DG, Hoehn HH, Panar M. Measurement of gas permeability of polymers. I. Permeabilities in constant volume/variable pressure apparatus. J Appl Polym Sci 1976;20(7):1921–31. [28] Pye DG, Hoehn HH, Panar M. Measurement of gas permeability of polymers. II. Apparatus for the determination of permeabilities of mixed gases and vapors. J Appl Polym Sci 1976;20(2):287–301. [29] Laine NR, Vastola FJ, Walker Jr PL. The importance of active surface area in the carbon–oxygen reaction. J Phys Chem 1963;67(10):2030–4. [30] Hill Jr CG. An introduction to chemical engineering kinetics and reactor design. New York: John Wiley and Sons, Inc.; 1977.