12
Physical aging in carbon molecular sieve membranes Liren Xu a , Meha Rungta a , John Hessler a , Wulin Qiu a , Mark Brayden b , Marcos Martinez c , Gregory Barbay b , William J. Koros a, * a School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA b The Dow Chemical Company, Plaquemine, LA 70765, USA c The Dow Chemical Company, Freeport, TX 77541, USA ARTICLE INFO Article history: Received 29 April 2014 Accepted 16 August 2014 Available online 23 August 2014 ABSTRACT This paper considers physical aging in carbon molecular sieve (CMS) membranes. More- over, the performance of stabilized membranes under practical operating conditions is discussed. Physical aging has been studied extensively in glassy polymers, but aging in CMS membranes has previously focused primarily on adsorption: either chemisorption of oxygen, or physical adsorption of water and organics in the pore structures. Experimen- tally, in this study, for the samples considered, all of the above adsorption-induced aging mechanisms were excluded as significant factors through thoughtful experimental design. Physical aging appears to be the primary cause for rapid changes of transport properties in early stages after membrane fabrication for samples derived from high fractional free volume precursors. The CMS pores are believed to age analogously to the ‘‘unrelaxed free volume’’ in glassy polymers. Over time, these pores tend to shrink in order to achieve ther- modynamically more stable states. Results of sorption tests in CMS also support the above hypothesis. The significance of physical aging phenomena on membrane testing protocols, structural tailoring, and performance evaluation are discussed. A long term permeation test demonstrated excellent stability of stabilized CMS membranes under realistic conditions. Ó 2014 Elsevier Ltd. All rights reserved. 1. Introduction Membrane technology has emerged as a promising alterna- tive to conventional separation technologies such as cryo- genic distillation and adsorption. Membranes have potential to be energy-efficient, environmentally friendly and have small process footprints. The current commercially available membrane materials, polymeric membranes, appear to have reached an upper bound tradeoff between ‘‘productivity’’ and ‘‘efficiency’’ [1–4]. Therefore, seeking novel materials to overcome the polymer performance upper bound is of great interest. Molecular sieve materials, such as inorganic zeolites, metal organic frameworks (MOF), and carbon molecular sieves (CMS), are able to exceed the polymer performance limitation [5,6]. Due to the close relation with their precursor materials, manageable processability and excellent perfor- mance, CMS membranes are especially promising as a mem- brane material option. Significant efforts have been devoted to the development of CMS membranes, which have http://dx.doi.org/10.1016/j.carbon.2014.08.051 0008-6223/Ó 2014 Elsevier Ltd. All rights reserved. * Corresponding author. E-mail address: [email protected] (W.J. Koros). CARBON 80 (2014) 155 166 Available at www.sciencedirect.com ScienceDirect journal homepage: www.elsevier.com/locate/carbon

Physical aging in carbon molecular sieve membranes

Embed Size (px)

Citation preview

C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6

.sc ienced i rec t .com

Avai lab le a t www

ScienceDirect

journal homepage: www.elsevier .com/ locate /carbon

Physical aging in carbon molecular sievemembranes

http://dx.doi.org/10.1016/j.carbon.2014.08.0510008-6223/� 2014 Elsevier Ltd. All rights reserved.

* Corresponding author.E-mail address: [email protected] (W.J. Koros).

Liren Xu a, Meha Rungta a, John Hessler a, Wulin Qiu a, Mark Brayden b,Marcos Martinez c, Gregory Barbay b, William J. Koros a,*

a School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USAb The Dow Chemical Company, Plaquemine, LA 70765, USAc The Dow Chemical Company, Freeport, TX 77541, USA

A R T I C L E I N F O

Article history:

Received 29 April 2014

Accepted 16 August 2014

Available online 23 August 2014

A B S T R A C T

This paper considers physical aging in carbon molecular sieve (CMS) membranes. More-

over, the performance of stabilized membranes under practical operating conditions is

discussed. Physical aging has been studied extensively in glassy polymers, but aging in

CMS membranes has previously focused primarily on adsorption: either chemisorption

of oxygen, or physical adsorption of water and organics in the pore structures. Experimen-

tally, in this study, for the samples considered, all of the above adsorption-induced aging

mechanisms were excluded as significant factors through thoughtful experimental design.

Physical aging appears to be the primary cause for rapid changes of transport properties in

early stages after membrane fabrication for samples derived from high fractional free

volume precursors. The CMS pores are believed to age analogously to the ‘‘unrelaxed free

volume’’ in glassy polymers. Over time, these pores tend to shrink in order to achieve ther-

modynamically more stable states. Results of sorption tests in CMS also support the above

hypothesis. The significance of physical aging phenomena on membrane testing protocols,

structural tailoring, and performance evaluation are discussed. A long term permeation

test demonstrated excellent stability of stabilized CMS membranes under realistic

conditions.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

Membrane technology has emerged as a promising alterna-

tive to conventional separation technologies such as cryo-

genic distillation and adsorption. Membranes have potential

to be energy-efficient, environmentally friendly and have

small process footprints. The current commercially available

membrane materials, polymeric membranes, appear to have

reached an upper bound tradeoff between ‘‘productivity’’

and ‘‘efficiency’’ [1–4]. Therefore, seeking novel materials to

overcome the polymer performance upper bound is of great

interest.

Molecular sieve materials, such as inorganic zeolites,

metal organic frameworks (MOF), and carbon molecular

sieves (CMS), are able to exceed the polymer performance

limitation [5,6]. Due to the close relation with their precursor

materials, manageable processability and excellent perfor-

mance, CMS membranes are especially promising as a mem-

brane material option. Significant efforts have been devoted

to the development of CMS membranes, which have

Fig. 1 – Schematic of the physical aging process of a glassy

polymer [42].

156 C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6

demonstrated high performance and excellent stability for

several potential applications [7–10].

Considerable research has also focused on tailoring the

membrane molecular sieve properties for higher permeability

and selectivity [7,11–21]; however, relatively little work has

focused on the stability of CMS materials [22–26]. This lack

of attention is partially due to the rigid nature of CMS mem-

branes and their enhanced chemical and thermal resistance

compared to polymeric materials. In order to successfully

implement CMS membrane technology, however, it is impor-

tant to truly understand the operational stability of CMS

membranes to intelligently engineer their processing and

handling during fabrication and operation. The performance

of membranes under active feed vs. under storage conditions

is also important, as has been observed for traditional glassy

polymers and will be considered here.

First, the previously reported aging phenomena in glassy

polymers and CMS membranes are reviewed systematically.

Then transport properties of CMS membranes observed dur-

ing membrane characterization are compared to previously

reported scenarios, and clear evidence of physical aging is

identified in CMS membranes. Finally, the impact of physical

aging in CMS membranes is discussed in detail, in terms of

membrane fundamental exploration, testing protocol, and

realistic applications.

2. Background

As noted above physical aging in polymeric membranes has

been studied extensively [27–36]. Generally, membrane aging

can be divided into two categories: physical aging and chem-

ical aging. Physical aging is common with glassy polymers,

and to the best of our knowledge has not been reported pre-

viously in CMS materials. In CMS membranes, only sorption

or chemisorption based aging has been suggested [22–26].

2.1. Physical aging in polymeric membranes

Clearly, properties of polymeric membranes may be affected

by contaminants or penetrant induced plasticization [37–40],

etc. This fact notwithstanding, ‘‘aging’’ in polymeric mem-

branes generally refers to physical aging in non-equilibrium

glassy polymers undergoing physical rearrangements to

approach an equilibrium state. Fig. 1 illustrates the non-equi-

librium nature of glassy polymers in terms of volume versus

temperature relationship [41,42]. Above the glass transition

temperature (Tg), the polymer exists in an equilibrium state.

When a polymer is cooled from the rubbery state through

the glass transition temperature, long-range polymer chain

segmental movements become drastically hindered, and the

polymer ‘‘vitrifies’’ to form segmental-scale packing microv-

oids. The sum of these microvoids is also known as ‘‘excess

free volume’’ or ‘‘unrelaxed volume’’. Below the glass transi-

tion temperature, the polymer densifies over time and

approaches thermodynamic equilibrium due to the removal

of the excess free volume by slow segmental rearrangements.

Such physical aging can result in significant time-dependent

membrane transport properties [27,29,30,33,34,43,44], with

decreased permeability and increased selectivity over time.

The state of a glassy polymer and its corresponding aging

behavior depend not only on the immediate environment of

the polymer but also on its previous history (i.e., thermal his-

tory, vapor exposure) [29]. Physical aging accelerates with

increasing temperature and decreasing membrane thickness

[30,43]. Thin films have been shown to age much faster than

thick films of the same materials [27,43]. Due to the thin

separation layer in asymmetric hollow fiber membranes,

physical aging is an especially important topic for polymeric

membrane research.

2.2. Chemical aging in CMS membranes

Although we are not aware of a prior report of physical aging

in CMS membranes, chemical and sorption-induced aging of

carbon membranes under different environments has been

noted previously as a cause of property changes versus time

[22–26]. Chemical and sorption-induced aging requires inter-

action of CMS membranes with external species, and much

of the previously reported stability issues of CMS membranes

appear to be caused by adsorption. Water adsorption has been

reported to reduce membrane performance [24,25,45]. Jones

and Koros also exposed CMS membranes to organic contam-

inants (hexane, vacuum pump oil, phenol and toluene, etc.),

and significant permeance loss was observed [26]; however,

most of the loss was reversible upon exposure to propylene

near unity activity, indicating sorption-based aging rather

than a true physical aging. Menedez and Fuertes investigated

the aging of carbonized phenolic resin films on porous alu-

mina tubes in different environments (air, nitrogen and pro-

pylene) [23]. By comparison of oxygen exposed and oxygen

free conditions, they concluded that exposure to oxygen

was the major cause for aging of the carbon membranes they

studied. Further effort was made by Lagorsse and Mendes

et al. for the impact of air and humidity exposure on cellulose

C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6 157

derived CMS membranes [22]. They found that water strongly

adsorbed on these CMS membranes at medium to high

humidity. Again, chemisorption of oxygen was noted to be

the cause of membrane aging in dry conditions.

The aging phenomena revealed previously can be illus-

trated using the illustrative representations in Fig. 2. Oxygen

chemisorption based aging is possibly similar to the previ-

ously reported higher temperature oxygen doping concept

[14], but more extreme and undesirable. We hypothesize that

in both cases, depending upon the nature of the precursor,

pyrolysis conditions, and post pyrolysis exposure conditions,

oxygen selectively chemisorbs on the ‘‘edges’’ of ultramicrop-

ores. The defects in carbon structure present reactive sites for

oxygen chemisorption [23]. When disordered carbon materi-

als with still reactive edges are exposed to air at room temper-

ature, chemisorption of oxygen may take place slowly [23,46].

This process is slower than oxygen doping at high pyrolysis

temperatures. Moreover, the carbon microvoids are generally

hydrophobic, so they tend to adsorb organics (hexane, tolu-

ene, etc.) [25]. Some oxygen-containing surface groups may

also occur on carbon surfaces and act as primary sites to

attract additional water [25]. At high activities, even relatively

hydrophobic micropores may eventually lead to formation of

water clusters. In any case, adsorption of water and organics

may reduce the pore volume and pore size, thereby decreas-

ing the permeability of gases through the membranes;

however, these reductions, as shown by Jones et al., are often

reversible [26].

To address the stability issue noted above, several

approaches have been explored. Besides drying membranes

at elevated temperatures [45], Jones and Koros showed that

forming a carbon composite membrane with a water resistant

Teflon coating offered a preventative approach to address the

problem under operating conditions [25]. For membranes

after organic exposure, pure propylene at close to unit activity

was shown to be an effective cleaning agent to remove most

organics [26]. For the aging caused by oxygen chemisorption,

heat-treatments in a reduced oxygen atmosphere at high

temperatures were used to remove oxygen surface groups

[22].

As these prior studies show, CMS ‘‘aging’’ must be consid-

ered carefully to identify the primary causative factor. Using

Fig. 2 – Cartoon representation of mechanisms of aging in

CMS membranes caused by adsorption. (A color version of

this figure can be viewed online.)

such an approach, the results shown below indicate that

physical aging issues related to CMS appear to share features

analogous to those indicated for glassy polymers Fig. 1.

3. Experimental

3.1. Preparation of CMS membranes

Three polyimides were used as precursor materials in

this study. They were commercially available Matrimid�

(Huntsman International LLC) as well as lab-synthesized

6FDA-DAM and 6FDA/BPDA-DAM. The structures of above

polyimides are illustrated in Fig. 3. 6FDA-DAM and 6FDA/

BPDA-DAM were synthesized by a two-step reaction in which

the first step produced a high molecular weight polyamic acid

at low temperature (�5 �C) followed by a second, imidization

step and a final drying step at 210 �C under vacuum to

produce the polyimides [40].

In this study, both dense film membranes and hollow fiber

membranes were considered to assess any differences in the

two forms. Details of membrane fabrication are provided in

previous publications [3,11–13]; however, general procedures

and a few key issues are pointed out, as follows for conve-

nience. The precursor dense film membranes were prepared

via a solution casting method. The precursor asymmetric hol-

low fiber membranes were prepared by a dry-jet/wet-quench

spinning process, and only defect-free precursors were used

for pyrolysis to provide consistency between precursor mate-

rials. The precursor membranes were pyrolyzed in a quartz

tube heated in a three-zone furnace. The pyrolysis protocols

have been discussed previously [3,11–13]. For dense film

membranes, a slotted quartz plate was used as the membrane

support; for hollow fibers, stainless steel wire mesh was used

as the support for hollow fiber membranes.

After pyrolysis and removal from the furnace, the resul-

tant CMS membranes were stored in the desired conditions,

or masked and mounted in a dense film membrane testing

cell or formed into a hollow fiber module for testing. Storage

conditions are described with results later.

3.2. Permeation tests

Both dense film and hollow fiber membranes were tested

using a constant volume permeation system [47]. To remove

physi-sorbed gases in the membranes, the permeation units

were first evacuated. After sufficient vacuum time and leak

test, the upstream was fed with gases under desired pres-

sures, and pressure rise in the downstream volume was

recorded using LabVIEW (National Instruments, Austin, TX).

The rate of pressure rise was then used to calculate the per-

meance or permeability of gases through the membranes.

The pure gas selectivity was calculated using the ratio of per-

meances or permeabilities (fast gas to slow gas). For mixed

gas permeation tests, similar protocols were followed except

that a gas chromatography was used to analyze the permeate

gas composition. The stage cut of permeation was usually

kept at 1% or lower. The selectivities in mixed gas cases were

calculated from gas chromatography analysis, and individual

Fig. 3 – Chemical structures of 3 precursor polyimides: (a) Matrimid�; (b) 6FDA/BPDA-DAM; (c) 6FDA-DAM.

158 C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6

gas permeance or permeability values were calculated based

on the downstream pressure rise and the selectivity.

3.3. Sorption tests

Gas sorption measurements were made using a pressure

decay method [48,49]. CMS dense films were used for the

sorption tests, and the flat sheet membranes were gently bro-

ken into small pieces between two pieces of weighing paper.

The CMS pieces were loaded into a stainless steel Swagelok

filter element and wrapped loosely by aluminum foil to

ensure that the samples remained in the filter. The sorption

apparatus was placed in a temperature controlled oil bath.

The system was evacuated for 24 h prior to testing, then feed

gas was introduced into the reservoir chamber and allowed to

equilibrate for 10–15 min. The thermally equilibrated gas was

then introduced into the sample cell and the pressures in the

sample cell and reservoir were monitored by pressure

transducers and recorded over time using LabVIEW. Sorption

isotherms were then obtained for the equilibrium sorbed gas

concentration at each final pressure.

3.4. Structural characterization

X-ray photoelectron spectroscopy (XPS) was performed on a

Thermo ScientificTM K-AlphaTM XPS system. The analysis pro-

vided elemental information of samples. Raman spectroscopy

for carbon samples was performed on a Thermo Scientific

NicoletTM AlmegaTM XR Micro and Macro Raman Analysis

System. For both techniques, the surface of CMS fiber sam-

ples was analyzed.

4. Results and discussion

4.1. Discovery of physical aging in CMS membranes

Prior to the current research, to our knowledge, there has

been no report of physical aging in CMS membranes. In this

research, changes in transport properties of CMS membranes

over time were observed. The aging phenomenon was negligi-

ble in CMS derived from lower free volume Matrimid� precur-

sor, but quite obvious and impacted several cases when our

work progressed to higher free volume 6FDA-polymers

derived CMS as noted below. 6FDA-DAM and 6FDA/BPDA-

DAM polymers were originally chosen as precursor materials

due to their high rigidity and intrinsically higher starting per-

meability relative to Matrimid�, and as expected, significant

improvement in CMS membrane performance was seen with

these higher free volume precursors.

While the permeation results for Matrimid� CMS were

quite consistent, some 6FDA CMS showed considerable time

dependence during tests. Fig. 4 shows permeation results

for a Matrimid� derived CMS hollow fiber membrane, which

was prepared using a typical 550 �C/2 h pyrolysis protocol

under a UHP argon purge. Under atmospheric storage condi-

tions, 6 days after membrane preparation, the ethylene per-

meance decreased from 5.5 GPU to 3.6 GPU, while the

ethylene/ethane selectivity increased from 3.1 to 4.2. During

the testing, some vacuum was applied due to the constant

Fig. 4 – Time dependence of separation performance of a

Matrimid� derived CMS hollow fiber membrane

(atmospheric storage, mixed gas test, 100 psia feed, 35 �C).

(A color version of this figure can be viewed online.)

Fig. 5 – Time dependence of separation performance of a

6FDA-DAM derived CMS hollow fiber membrane (pure gas

test, 100 psia, 35 �C). (A color version of this figure can be

viewed online.)

Fig. 6 – Regeneration of an aged 6FDA-DAM CMS hollow fiber

membrane by propylene cleaning (pure gas test, 100 psia,

35 �C).

C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6 159

volume testing method. Such change was much smaller than

the later discovered 6FDA CMS, and did not trigger the inves-

tigation on physical aging of CMS membranes. Time-depen-

dence of permeation behavior was not considered at that

time.

Fig. 5 shows permeation results for a 6FDA-DAM derived

CMS hollow fiber membrane. The membrane was prepared

using a 675 �C/2 h pyrolysis protocol under UHP argon purge.

The higher temperature pyrolysis protocol was used for the

more open 6FDA-based polymer, in order to achieve better

selectivity. Between the permeation runs, the membrane

was either kept under active vacuum in the permeation sys-

tem or capped in atmospheric air. These conditions are

marked in Fig. 5. As shown in Fig. 5, the ethylene permeance

decreased and the ethylene/ethane selectivity increased over

time. This trend is similar to Matrimid� derived CMS, but it is

much more significant and called our attention to this matter.

The transport properties changed quickly during the first few

days after production, and tended to stabilize quickly. Never-

theless, as with conventional glasses, the aging process is

likely to continue with less obvious impact on properties,

since physical aging is typically a self-retarding process [50].

As noted earlier, previous researchers, for different precur-

sor-derived CMS, proposed that aging in carbon membranes

were attributed to oxygen chemisorption or water or organic

adsorption. Therefore, the permeance loss in atmospheric

storage in the current work might be attributed to oxygen

chemisorption; however, this is not consistent with the

essentially identical response of the vacuum-stored mem-

brane. In the vacuum conditions, all these adsorption pro-

cesses should be absent and if this aging is truly adsorption

controlled, the membrane should preserve its original state.

This led to the hypothesis that true physical aging also exists

in CMS membranes as it does in more conventional glassy

polymers.

Based on the above hypothesis, causes for aging of CMS

membranes were expanded to the following possibilities:

water vapor adsorption, organic adsorption, oxygen chemi-

sorption and true physical aging. In the active vacuum condi-

tions, water vapor and oxygen were absent in the membrane

module and could be excluded as causes for the permeance

reduction. It was still possible that vacuum pump oil vapor

may back diffuse to the membrane, although it was unlikely

to happen due to the installation of an alumina foreline trap.

As demonstrated previously, some organic contaminants

could be removed by propylene regeneration [26], so propyl-

ene with near unity activity (about 148 psi at 20 �C) was used

to clean the membrane. The propylene stream was fed on the

bore side and permeated through the membrane to the shell

side. The cleaning was typically more than 12 h, and regener-

ation was performed for the module showed in Fig. 5 after

aging for 52 days. The results shown in Fig. 6 indicate that

160 C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6

ethylene permeance was only slightly increased from 2.2 GPU

to 2.5 GPU, which was still well below its original value of

16.1 GPU, and the selectivity was only slightly reduced from

6.1 to 5.3. This suggests that removable organic contaminants

were not the major cause of the permeance loss, or the regen-

eration should have been more effective.

Propylene regeneration was also performed on other

membrane modules with various aging histories, and in all

cases, only very small recoveries in permeance were seen to

values well below their initial permeances. Of course, the

argument at this point can be: (1) aging was not caused by

organic adsorption; or (2) the organics from vacuum pump

oil caused aging, but the regeneration method was not effec-

tive. To test the latter possibility, further tests were performed

under vacuum-free conditions to exclude the pump oil effect.

Specifically, CMS membrane modules were maintained

under inert gas storage conditions. Fresh 6FDA/BPDA-DAM

CMS hollow fiber membranes from a 675 �C/2 h pyrolysis pro-

tocol were stored under active 30 psi g feed conditions under

Fig. 7 – Time dependence of separation performance of a

6FDA/BPDA-DAM derived CMS hollow fiber membrane

under argon storage (mixed gas test, 100 psi, 35 �C).

UHP argon. Oxygen and moisture content in the gas were less

than 1 ppm according to Airgas. The membranes were re-

tested after 5 days and 145 days storage using mixed gas at

35 �C. Fig. 7(a) shows ethylene permeance decreased over

time, while Fig. 7(b) shows ethylene/ethane selectivity

increased over time. Since substantial aging in inert argon

occurred, adsorption-based aging can also be excluded. The

results, therefore, indicate that like conventional glassy poly-

mers, some CMS derived from high free volume precursors

can also undergo physical aging.

The above discussion indicates that, as is the case with

conventional glassy polymers, the composition of the glass,

the formation conditions, the characteristic dimensions of a

sample and even the testing protocols used to probe the sam-

ple properties can have a significant impact on measured

CMS transport properties. Besides the hollow fiber configura-

tion, the physical aging phenomenon was also observed in

the case of flat sheet membranes. The simplicity of the homo-

geneous morphology avoids complexities associated with

asymmetric structures, and provides additional clear evi-

dence for physical aging in the CMS materials. As shown in

Fig. 8, over a period of about 4 weeks, despite inconsistency

in the initial separation performance, CMS membranes

derived using the same fabrication conditions eventually con-

verged to almost the same performance. The main difference

between the films was the timing for the first test: Film 1 was

masked for permeation measurement immediately after

pyrolysis, while Film 2 was stored under vacuum for about

two days prior to masking and then tested. These results

demonstrated that 6FDA-DAM derived CMS dense film mem-

branes show a decrease in ethylene permeability over time,

but the performance eventually approaches a stable value.

As illustrated for the hollow fiber tests, the testing protocol

for dense films was well controlled although there were some

necessary differences. The dense films were either actively

exposed to testing gases or under active vacuum between

Fig. 8 – Time dependence of pure gas transport properties of

two 675 �C pyrolyzed 6FDA-DAM CMS films (pure gas test,

50 psi, 35 �C). (A color version of this figure can be viewed

online.)

C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6 161

runs, thus eliminating the possibility of exposure to oxygen or

moisture. Therefore, it appears that similar physical aging

takes place in CMS dense film membranes as in hollow fiber

membranes.

4.2. Additional structural considerations of physical agingin CMS membranes

The above insights were enabled by exclusion of other possi-

ble aging mechanisms; however, it is also useful to further

probe physical aging from the standpoint of the CMS struc-

ture. Several structural analyses below were used to achieve

this objective.

On the elemental level, for 6FDA-DAM and 6FDA/BPDA-

DAM, precursor polymers consist of carbon (C), hydrogen

(H), oxygen (O), nitrogen (N) and fluorine (F) elements. After

pyrolysis (500–800 �C, mostly 550 �C or 675 �C), fluorine was

essentially completely removed, as confirmed by XPS mea-

surement of the precursor and the resultant carbon fiber.

The XPS spectra are shown in Fig. 9. XPS also provides infor-

mation of some quantitative elemental compositions except

for hydrogen and helium. This 6FDA-DAM CMS membrane

prepared by a 675 �C protocol contained 85.8 ± 1.6% atomic

carbon, 11.1 ± 3.0% atomic oxygen, and 3.1 ± 1.8% atomic

nitrogen on a hydrogen-free basis. The above result indicates

there are significant amounts of elements other than carbon,

and the carbon content was a bit lower than reported in other

studies [7,9,21]. Since XPS is highly surface sensitive (the typ-

ical detection depth is �5 nm), works by other researchers

have shown that the surface carbon content might be

increased by using long argon ablation to remove surface-

sorbed oxygen to probe the ‘‘true’’ value within the bulk of

the material [9,51].

The second structural level (porous structure) is important

for CMS membrane transport properties. Previously, it was

noted that the pores in CMS materials are believed to be

Fig. 9 – Comparison of 6FDA-DAM precursor and CMS

(675 �C) XPS spectra. (A color version of this figure can be

viewed online.)

formed by packing imperfections of small graphene-like lay-

ers, while on the long range CMS materials are amorphous

[52,53]. Raman spectroscopy is useful to characterize disorder

in sp2 carbon materials [54–56]. The Raman spectrum of crys-

talline graphite is marked by the presence of two strong peaks

at 1580 cm�1 and 2700 cm�1, named the G and G 0 bands,

respectively. Samples with small crystallite size show an

additional dispersive peak centered at approximately

1350 cm�1. This peak has been named the D band (D for

defect or disorder). The D band is not present in single crys-

tals of graphite. The intensity of the D band and the ratio

between the intensities of the disorder-induced D band and

the graphite G band provides a parameter that can be used

as some measure of disorder within the structure. Raman

spectra of Matrimid�, 6FDA-DAM and 6FDA/BPDA-DAM CMS

membranes pyrolyzed at 550 �C are shown in Fig. 10. As

clearly illustrated by the spectra, the structure of CMS mem-

branes fall into the amorphous carbon form as described pre-

viously, since both G and D bands are present. Presumably,

the different transport rates in CMS variants reflect different

degrees of packing of the constituent graphene-like sheets.

Clearly, however, this graphene-like picture is simplified,

since the presence of the residual N and O (and H) atoms

may not only be on the defect edges. The absolute intensity

of the peaks in Fig. 10 does not necessarily correlate to the

properties of CMS samples investigated. The ratio of the D

and G peaks is more useful to characterize the carbon struc-

ture [57–59]. Details of Raman analysis will be performed in

our future work.

Based on the above discussion, all of the CMS samples

have a non-graphitic nature, as indicated by their Raman

spectra. Sorption tests were then used to probe microstruc-

ture evolution over time using 675 �C pyrolyzed 6FDA-DAM

CMS membranes. After unloading from the pyrolysis furnace

(Day 0), the sample was tested 3 times for ethylene sorption

on Day 6, 48 and 65, respectively, while the ethane tests were

Fig. 10 – Comparison of Raman spectra of Matrimid�, 6FDA-

DAM and 6FDA/BPDA-DAM CMS membranes pyrolyzed at

550 �C. (A color version of this figure can be viewed online.)

Fig. 11 – Time dependence of ethylene and ethane sorption

isotherms in 675 �C pyrolyzed 6FDA-DAM CMS membranes.

(A color version of this figure can be viewed online.)

Fig. 12 – Schematic cartoon representation of envisioned

physical aging in glassy polymers and CMS materials.

162 C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6

on Day 25, 55 and 69, respectively. The experimental data

were fitted to Langmuir isotherms shown in Fig. 11, and the

Langmuir model fitting parameters are listed in Table 1. The

Langmuir isotherm is written as shown in Eq. (1):

C ¼ C0Hb1þ bp

ð1Þ

where C is the concentration of penetrant (cc(STP)/cc CMS), p

is the penetrant pressure, C 0H is the Langmuir capacity

constant, and b is the Langmuir affinity constant. In glassy

Table 1 – Time dependence behavior of Langmuir isotherm para6FDA-DAM CMS membranes.

C2H4

Test 1 Test 2

C 0H cm3/cm3 136.8 109.2b 1/psia 0.048 0.080

polymers, the trapped excess free volume can decrease over

extended time due to annealing below the glass transition

of the polymer. The decrease in C 0H over time for the

6FDA-DAM-derived CMS samples supports the hypothesis

that the aging behavior in CMS is analogous to the physical

aging phenomenon in glassy polymers. The reduced Lang-

muir sorption capacity can be primarily attributed to reduc-

tions in C 0H decrease from the just-made ‘‘young’’ glass to

the ‘‘aged’’ glass. While the fit to the Langmuir model is

clearly not as good for the young glass, comparison between

the estimated values of the C 0H over time is still valuable. In

CMS membranes, the micropores that provide the source of

the Langmuir capacity are created by packing imperfections

or disorder of graphene-like layers during pyrolysis at high

temperatures. Immediately after formation, these structures

appear not to be in a thermodynamically stable state [60].

We envision such layers to achieve denser packing, especially

at the early stage post-fabrication. Based on this concept, the

pore volume in CMS materials can be viewed as analogous to

the ‘‘excess free volume’’ in conventional glassy polymers.

This fact notwithstanding, a significant difference exists

between CMS glasses and simple polymer glasses, due to

the lack of a well-defined glass transition, and the lack of

corresponding sample ‘‘thermal reversibility’’, which can be

achieved in simple glassy polymers [44,50]. The schematics

in Fig. 12 seek to compare and contrast the key features in

such envisioned physical aging in glassy polymers and CMS

materials.

4.3. Effect of precursor polymers and processingconditions

Based on the above discussion, as with glassy polymers, the

starting materials and the processing conditions used to

meters for ethylene and ethane sorption in 675 �C pyrolyzed

C2H6

Test 3 Test 1 Test 2 Test 3

93.1 105.8 93.9 84.60.145 0.136 0.221 0.239

Fig. 13 – Impact of aging on transport properties of

penetrant gases with different sizes for CMS produced from

6FDA/BPDA-DAM precursor at 550 �C and UHP pyrolysis

(35 �C, 100 psia).

C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6 163

create the ‘‘young’’ glass are expected to have significant

impacts on aging of CMS membranes. Glassy polymers with

more open structures have a higher driving force for aging

than less open ones [27,31,34,61,62]. CMS membranes pre-

pared from Matrimid�, 6FDA-DAM, and 6FDA/BPDA-DAM

550 �C/2 h and UHP argon pyrolysis were investigated for time

dependence. Vacuum and atmospheric conditions were

investigated for the storage of CMS membranes. In the vac-

uum case, the membrane module was installed in constant

volume permeation systems, and both upstream and down-

stream were evacuated except during the short testing time.

In the atmospheric storage, the membrane modules were

sealed using Swagelok fittings and atmospheric air was

trapped in the modules.

The high free volume 6FDA-DAM-derived CMS started with

the highest permeance (above �100 GPU for ethylene), while

low free volume Matrimid�-derived CMS had the lowest start-

ing permeance (less than �6 GPU for ethylene). The aging of

6FDA-DAM derived CMS was the fastest, while that from the

Matrimid� derived CMS was the most stable, and 6FDA/

BPDA-DAM (with intermediate free volume) derived CMS

was intermediate in behavior. The free volume of the three

polyimides were reported previously [63]. Comparing vacuum

and atmospheric storage, in all cases, at the early stage after

pyrolysis, membranes stored in vacuum conditions aged

more than the ones in atmospheric conditions. This is consis-

tent for all the 3 polymers investigated. For example, for

6FDA/BPDA-DAM, after 6 days’ storage, the permeance of

membranes under atmospheric storage is about 3.7 times

higher than the one under vacuum storage. What really mat-

ters, of course, for the realistic application is the stabilized

separation performance. Among these three polymers, it

appears still 6FDA/BPDA-DAM derived CMS had a reasonably

stable performance after aging, and the ethylene permeance

was stable in about one week and remained higher than

�10 GPU.

4.4. Impact of aging on transport properties of penetrantgases

Also as in the case of glassy polymers, responses of penetrant

gases vary with membrane aging [35,64]. Although the perme-

ances of all penetrant gases decrease, the selectivities for a

fast gas relative to a slow gas tends to increase with aging.

The extent of permeance loss at any point during the aging

process depends on respective gas molecule sizes, with small

sized gases showing less effect of aging, while the influence of

aging on permeances of large gases is more dramatic. Fig. 13

summarizes the key permeance trends for 7 pure component

gases in CMS membranes derived from 6FDA/BPDA-DAM after

aging for about 228 days after the initial test. The measured

permeance values are compared to the freshly made samples

that were tested 24 h after their formation. The smallest gas,

helium showed only �18.6% decrease in permeance, whereas

the largest gas, ethane showed a �75.4% decrease in perme-

ance, with the intermediate sized gases showing correspond-

ingly intermediate impacts of aging for the extended 228 day

aging study. While the detailed changes in micropore and

ultramicropore size distributions during aging process at this

point are still not clear, these data provide strong evidence

that shifts are occurring. Clearly, fast gas pairs such as that

in the CO2/CH4 separation or the O2/N2 case are less affected

than the case for olefin/paraffin pairs, especially C2H4/C2H6

separation or presumably hydrocarbons with higher carbon

numbers. This suggests a possible tuning mechanism for

selectivity.

4.5. Considerations of permeation testing protocols inlight of aging behavior

As demonstrated by the previous sections, transport proper-

ties of CMS membranes have strong history dependence dur-

ing the periods soon after their formation, and without care,

scattered results can lead to poorly characterized sample

properties. The post-pyrolysis processing conditions affect

the membrane performance significantly; however, without

realizing the history dependence feature noted here, CMS

membranes are typically assumed to be rigid and stable over

time. Indeed, many CMS membrane tests are performed in

constant volume downstream vacuum systems. To ensure

removal of pre-sorbed components, membranes are typically

evacuated at the feed and permeate sides. Based on the above

discussion, this evacuation can impact the rate of approach to

a more-or-less equilibrium set of sample properties with

characteristic steady state properties. This observation may

explain the divergence of some measured values caused by

changing testing protocols between different research groups.

For example, in our group, a 10–20 GPU ethylene permeance

was reported for 6FDA/BPDA-DAM CMS produced by 550 �C/

argon purge pyrolysis in our tests with the long term evacua-

tion protocol [11], while the intrinsic values right after pyroly-

sis, prior to dramatic aging under vacuum would be much

higher.

Fortunately, due to the thin selective layers ideally present

on CMS hollow fibers, relatively rapid approach to an ‘‘effec-

tive’’ steady state should be possible. On the other hand for

relatively thick (�60–70 micron) CMS dense membranes often

Fig. 14 – Long term stability test of a stabilized 6FDA/BPDA-

DAM CMS hollow fiber membrane. (A color version of this

figure can be viewed online.)

164 C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6

used for characterization, the situation is more complicated.

In most dense film cases, to ensure the complete removal of

sorbed gases from such relatively thick membranes, a long

term evacuation is standard before starting gas permeation

tests. Therefore, unavoidably, the membranes age quickly

during this period. In all cases, maintaining a consistent test-

ing protocol or stabilizing membranes effectively before test-

ing is clearly important for achieving consistent permeation

results. Of course, alternatively, without adequate evacuation

prior to tests, steady state permeation rates in constant vol-

ume permeation systems may still be obtained by monitoring

the pressure increase rates in the downstream, but it is chal-

lenging to judge the required period without knowing the

time lag. Such a required period can also be estimated for

comparison of mixed gas tests and pure gas tests; however,

prior to each pure gas test, the membrane has to experience

sufficient evacuation, and during this period the membrane

ages. Therefore, for an unstabilized membrane, separate pure

gas permeation tests are effectively performed on different

‘‘membranes’’ due to structural changes. This reality makes

it complicated to compare pure gas permeation results and

mixed gas permeation results, since in addition to a competi-

tion effect, aging effects during evacuation between pure

component tests are also at play. In any case, mixed gas per-

meations tests are much more reliable and reflect the actually

intended use conditions, and do not overestimate or underes-

timate the transport properties due to the artifact of history

dependence.

4.6. Realistic stability of CMS membranes

The current aging study shows that an unstable state persists

primarily within a very short period of time after membrane

fabrication. While meta-stable states may provide interesting

transport properties, and trapping the membrane structure in

such states could be an interesting topic for future investiga-

tions, the primary interest here is the properties of the stabi-

lized membranes.

After 5–12 months aging, 4 samples of 6FDA/BPDA-DAM

CMS membranes (550 �C/2 h, UHP argon pyrolysis) showed

very consistent ethylene permeance of about 8–9 GPU after

mixed gas permeation tests at 35 �C. It is significant that dif-

ferent apparent starting points imposed by different starting

points of characterization ultimately converge despite the dif-

ferent history for the 6FDA-derived CMS membranes. More-

over, the final permeance is still very attractive relative to

Matrimid� CMS membranes. Of course, as demonstrated

previously, the processing conditions can be used as a tool

to tailor the molecular sieving structure, so rapid aging can

be achieved via quick vacuum conditioning.

The long term membrane stability under realistic feed

conditions is another important topic. This topic has also

been explored for glassy polymers exposed to a perturbing

effect [65]. Most lab-scale characterization studies are per-

formed with downstream vacuum permeation conditions,

which are efficient for materials characterization but not

practical for most industrial applications. Therefore, in this

test, the downstream pressure was kept at 1 atm, and the eth-

ylene/ethane mixed gas feed pressure was kept at �115 psi,

thereby providing a driving force of about 100 psi. The active

testing was run for �150 h, and the results are shown in

Fig. 14. In the first 35 h, the ethylene permeance increased

from 8.8 GPU to 12.9 GPU, and the ethylene/ethane selectivity

decreases from 4.3 to about 3.9. After this period of time, the

performance was quite stable for the next 115 h. The standard

deviation of permeance was only 0.09 GPU, and the standard

deviation of the selectivity was only 0.02. For the initial equi-

librium step, two factors may be used to explain the observed

results. One reason can be the relatively slow establishment

of sorption equilibrium for the large penetrants, ethylene

and ethane, in the membrane. A second factor may be the

‘‘flexible’’ CMS structure as it transitions closer to a stable

lower free volume state. The intense hydrocarbon feed may

also participate in the transient rearrangement of the settling

pore structure, leading to the observed increase in permeance

and a subtle decrease in selectivity. Nevertheless, after

achieving equilibrium, the membrane is very stable in the

aggressive hydrocarbons feed conditions. A similar response

was, in fact, also observed for the previously mentioned

glassy polymer case.

5. Conclusions

In this paper, we report physical aging in carbon molecular

sieve membranes. We call attention to some similarities and

differences between the physical aging seen for CMS and con-

ventional glassy polymers. We also show that in addition to

aging in CMS membranes that has been revealed previously,

related to adsorption, physical aging is an important factor

to consider. Experimentally, in this study, all adsorption

induced aging was excluded by thoughtful experimental

design. The aging phenomenon was observed in both dense

film and hollow fiber membranes. It was shown that physical

aging appears to be the main cause for rapid changes of trans-

port properties at the early stage after the membrane fabrica-

tion for the samples studied here. Pores responsible for CMS

properties are believed to be formed by packing imperfections

of graphene-like layers, and age somewhat analogously to the

‘‘unrelaxed free volume’’ in glassy polymers. Over time, these

C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6 165

pores tend to be reduced in order to achieve thermodynami-

cally more stable states. Sorption tests indicate decreases of

Langmuir sorption capacity over time, which supports the

hypothesis of physical aging of CMS materials. The effect of

polymer precursor and processing conditions were discussed

as well and shown to be important factors in obtaining repro-

ducible results. The impact of the discovered aging phenom-

ena on membrane testing protocol, structural tailoring, and

performance evaluation was also discussed and shown to

be manageable with adequate care. Finally, history depen-

dence was identified as a possible tool to tailor the CMS mem-

brane transport properties; however, long term permeation

tests demonstrated ultimate excellent stability of CMS mem-

branes under realistic conditions.

Acknowledgments

This work was supported by The Dow Chemical Company.

The authors acknowledge the additional support provided

by King Abdullah University of Science and Technology

(KAUST).

R E F E R E N C E S

[1] Robeson LM. Correlation of separation factor versuspermeability for polymeric membranes. J Membr Sci1991;62(2):165–85.

[2] Burns RL, Koros WJ. Defining the challenges for C3H6/C3H8

separation using polymeric membranes. J Membr Sci2003;211(2):299–309.

[3] Rungta M, Zhang C, Koros WJ, Xu L. Membrane-basedethylene/ethane separation: the upper bound and beyond.AIChE J 2013;59(9):3475–89.

[4] Robeson LM. The upper bound revisited. J Membr Sci2008;320(1–2):390–400.

[5] Koros WJ, Lively RP. Water and beyond: expanding thespectrum of large-scale energy efficient separation processes.AIChE J 2012;58(9):2624–33.

[6] Baker RW. Future directions of membrane gas separationtechnology. Ind Eng Chem Res 2002;41(6):1393–411.

[7] Vu DQ, Koros WJ, Miller SJ. High pressure CO2/CH4 separationusing carbon molecular sieve hollow fiber membranes. IndEng Chem Res 2002;41(3):367–80.

[8] Steel KM, Koros WJ. An investigation of the effects ofpyrolysis parameters on gas separation properties of carbonmaterials. Carbon 2005;43(9):1843–56.

[9] Jones CW, Koros WJ. Carbon molecular sieve gas separationmembranes-I. Preparation and characterization based onpolyimide precursors. Carbon 1994;32(8):1419–25.

[10] Koresh JE, Sofer A. Molecular-sieve carbon permselectivemembrane. Part 1. Presentation of a new device for gas-mixture separation. Sep Sci Technol 1983;18(8):723–34.

[11] Xu L, Rungta M, Brayden MK, Martinez MV, Stears BA, BarbayGA, et al. Olefins-selective asymmetric carbon molecularsieve hollow fiber membranes for hybrid membrane-distillation processes for olefin/paraffin separations. J MembrSci 2012;423–424:314–23.

[12] Rungta M, Xu L, Koros WJ. Carbon molecular sieve dense filmmembranes derived from Matrimid� for ethylene/ethaneseparation. Carbon 2012;50(4):1488–502.

[13] Xu L, Rungta M, Koros WJ. Matrimid� derived carbonmolecular sieve hollow fiber membranes for ethylene/ethaneseparation. J Membr Sci 2011;380(1–2):138–47.

[14] Kiyono M, Williams PJ, Koros WJ. Effect of pyrolysisatmosphere on separation performance of carbon molecularsieve membranes. J Membr Sci 2010;359(1–2):2–10.

[15] Kiyono M, Williams PJ, Koros WJ. Effect of polymer precursorson carbon molecular sieve structure and separationperformance properties. Carbon 2010;48(15):4432–41.

[16] Kiyono M, Williams PJ, Koros WJ. Generalization of effect ofoxygen exposure on formation and performance of carbonmolecular sieve membranes. Carbon 2010;48(15):4442–9.

[17] Yoshino M, Nakamura S, Kita H, Okamoto K-i, Tanihara N,Kusuki Y. Olefin/paraffin separation performance ofcarbonized membranes derived from an asymmetric hollowfiber membrane of 6FDA/BPDA–DDBT copolyimide. J MembrSci 2003;215(1–2):169–83.

[18] Barsema JN, Balster J, Jordan V, van der Vegt NFA, Wessling M.Functionalized carbon molecular sieve membranescontaining Ag-nanoclusters. J Membr Sci 2003;219(1–2):47–57.

[19] Barsema JN, van der Vegt NFA, Koops GH, Wessling M. Carbonmolecular sieve membranes prepared from porous fiberprecursor. J Membr Sci 2002;205(1–2):239–46.

[20] Centeno TA, Fuertes AB. Carbon molecular sieve gasseparation membranes based on poly(vinylidene chloride-co-vinyl chloride). Carbon 2000;38(7):1067–73.

[21] Suda H, Haraya K. Gas permeation through micropores ofcarbon molecular sieve membranes derived from Kaptonpolyimide. J Phys Chem B 1997;101(20):3988–94.

[22] Lagorsse S, Magalhaes FD, Mendes A. Aging study of carbonmolecular sieve membranes. J Membr Sci 2008;310(1–2):494–502.

[23] Menendez I, Fuertes AB. Aging of carbon membranes underdifferent environments. Carbon 2001;39(5):733–40.

[24] Jones CW, Koros WJ. Characterization of ultrmicroporouscarbon membranes with humidified feeds. Ind Eng Chem Res1995;34(1):158–63.

[25] Jones CW, Koros WJ. Carbon composite membranes - asolution to adverse humidity effects. Ind Eng Chem Res1995;34(1):164–7.

[26] Jones CW, Koros WJ. Carbon molecular-sieve gas separationmembranes-II. Regeneration following organic exposure.Carbon 1994;32(8):1427–32.

[27] Cui LL, Qiu WL, Paul DR, Koros WJ. Physical aging of 6FDA-based polyimide membranes monitored by gas permeability.Polymer 2011;52(15):3374–80.

[28] Rowe BW, Pas SJ, Hill AJ, Suzuki R. Freeman BD, Paul DR.Ultrathin glassy polymer films for gas separation membranesand their physical aging behavior. Abstracts of papers of theAmerican Chemical Society 2010; 239.

[29] Rowe BW, Freeman BD, Paul DR. Influence of previous historyon physical aging in thin glassy polymer films as gasseparation membranes. Polymer 2010;51(16):3784–92.

[30] Kim JH, Koros WJ, Paul DR. Effects of CO2 exposure andphysical aging on the gas permeability of thin 6FDA-basedpolyimide membranes – Part 2 with crosslinking. J Membr Sci2006;282(1-2):32–43.

[31] Kim JH, Koros WJ, Paul DR. Physical aging of thin 6FDA-basedpolyimide membranes containing carboxyl acid groups. PartI. Transport properties. Polymer 2006;47(9):3094–103.

[32] Koros WJ, Madden WC. Effect of physical aging andaggressive feed components on hollow fiber gas separationmembranes. Abstracts of Papers of the American ChemicalSociety. 2005;230:U4111–U.

[33] Huang Y, Paul DR. Effect of temperature on physical aging ofthin glassy polymer films. Macromolecules2005;38(24):10148–54.

166 C A R B O N 8 0 ( 2 0 1 4 ) 1 5 5 – 1 6 6

[34] Huang Y, Paul DR. Physical aging of thin glassy polymer filmsmonitored by gas permeability. Polymer 2004;45(25):8377–93.

[35] Lin WH, Chung TS. Gas permeability, diffusivity, solubility,and aging characteristics of 6FDA-durene polyimidemembranes. J Membr Sci 2001;186(2):183–93.

[36] Wessling M, Huisman I, Vanderboomgaard T, Smolders CA.Time-dependent permeation of carbon-dioxide through apolyimide membrane above the plasticization pressure. JAppl Polym Sci 1995;58(11):1959–66.

[37] Omole IC, Bhandari DA, Miller SJ, Koros WJ. Toluene impurityeffects on CO2 separation using a hollow fiber membrane fornatural gas. J Membr Sci 2011;369(1–2):490–8.

[38] Ma C, Koros WJ. Effects of hydrocarbon and water impuritieson CO2/CH4 separation performance of ester-crosslinkedhollow fiber membranes. J Membr Sci 2014;451:1–9.

[39] Qiu W, Xu L, Chen C-C, Paul DR, Koros WJ. Gas separationperformance of 6FDA-based polyimides with differentchemical structures. Polymer 2013;54(22):6226–35.

[40] Qiu W, Chen C-C, Xu L, Cui L, Paul DR, Koros WJ. Sub-Tg cross-linking of a polyimide membrane for enhanced CO2

plasticization resistance for natural gas separation.Macromolecules 2011;44(15):6046–56.

[41] Huang Y, Paul DR. Effect of film thickness on the gas-permeation characteristics of glassy polymer membranes.Ind Eng Chem Res 2007;46(8):2342–7.

[42] Baker RW. Membrane technology and applications. 2nded. West Sussex, England: John Wiley & Sons Ltd; 2004.

[43] Cui LL, Qiu WL, Paul DR, Koros WJ. Responses of 6FDA-basedpolyimide thin membranes to CO2 exposure and physicalaging as monitored by gas permeability. Polymer2011;52(24):5528–37.

[44] Pfromm PH, Koros WJ. Accelerated physical ageing of thinglassy polymer films: evidence from gas transportmeasurements. Polymer 1995;36(12):2379–87.

[45] Kiyono M. Carbon molecular sieve membranes for naturalgas separations [Ph.D. thesis]. Georgia Institute ofTechnology; 2010.

[46] Boehm HP. Free radicals and graphite. Carbon2012;50(9):3154–7.

[47] Pye DG, Hoehn HH, Panar M. Measurement of gaspermeability of polymers. II. Apparatus for determination ofpermeabilities of mixed gases and vapors. J Appl Polym Sci1976;20(2):287–301.

[48] Costello LM, Koros WJ. Temperature dependence of gassorption and transport properties in polymers: measurementand applications. Ind Eng Chem Fundam 1992;31(12):2708–14.

[49] Koros WJ, Paul DR. Design considerations for measurementof gas sorption in polymers by pressure decay. J. Polym. Sci.Polym. Phys. Ed. 1976;14(10):1903–7.

[50] Struik LCE. Physical aging in amorphous polymers and othermaterials. Amsterdam: Elsevier; 1978.

[51] Singh A. Membrane materials with enhanced selectivity: anentropic interpretation [Ph.D. thesis]. The University of Texasat Austin; 1997.

[52] Pierson HO. Handbook of carbon, graphite, diamond, andfullerenes. Park Ridge, NJ: Noyes Publication; 1993.

[53] Jenkins GM. Polymeric carbons – Carbon fiber glass andchar. Cambridge University Press; 1976.

[54] Tuinstra F, Koenig JL. Raman spectrum of graphite. J ChemPhys 1970;53(3):1126–30.

[55] Dresselhaus MS, Jorio A, Souza Filho AG, Saito R. Defectcharacterization in graphene and carbon nanotubes usingRaman spectroscopy. Philos Transact A Math Phys Eng Sci1932;2010(368):5355–77.

[56] Reich S, Thomsen C. Raman spectroscopy of graphite. PhilosTrans A Math Phys Eng Sci 1824;2004(362):2271–88.

[57] Cancado LG, Jorio A, Ferreira EHM, Stavale F, Achete CA,Capaz RB, et al. Quantifying defects in graphene via Ramanspectroscopy at different excitation energies. Nano Lett2011;11(8):3190–6.

[58] Dresselhaus MS, Jorio A, Saito R. Characterizing graphene,graphite, and carbon nanotubes by Raman spectroscopy.Annu Rev Condens Matter Phys 2010;1(1):89–108.

[59] Ferrari AC, Robertson J. Interpretation of Raman spectra ofdisordered and amorphous carbon. Phys Rev B2000;61(20):14095–107.

[60] Oberlin A. Carbonization and graphitization. Carbon1984;22(6):521–41.

[61] Budd PM, McKeown NB, Fritsch D. Polymers of intrinsicmicroporosity (PIMs): high free volume polymers formembrane applications. Macromol Symp 2006;245–246(1):403–5.

[62] Morisato A, Shen HC, Sankar SS, Freeman BD, Pinnau I,Casillas CG. Polymer characterization and gas permeability ofpoly(1-trimethylsilyl-1-propyne) [PTMSP], poly(1-phenyl-1-propyne) [PPP], and PTMSP/PPP blends. J Polym Sci Part BPolym Phys 1996;34(13):2209–22.

[63] Rungta M. Carbon molecular sieve dense film membranes forethylene/ethane separations [Ph.D. thesis]. Georgia Instituteof Technology; 2012.

[64] Jordan SM. The effects of carbon-dioxide exposure onpermeation behavior in silicone rubber and glassypolycarbonates [Ph.D. thesis]. The University of Texas –Austin; 1989.

[65] Jordan SM, Koros WJ, Fleming GK. The effects of CO2

exposure on pure and mixed gas permeation behavior:comparison of glassy polycarbonate and silicone rubber. JMembr Sci 1987;30(2):191–212.