Effect of the amide bond diamine structure on the CO2, H 2S, and CH4 transport properties of a series of novel 6FDA-based polyamide-imides for natural gas purification

Article abstract of DOI:10.1021/ma301249x

A series of higher permeability polyamide-imides based on 2,2′-bis(3,4-dicarboxyphenyl) hexafluoropropane dianhydride with comparable plasticization resistance to Torlon were synthesized and formed into dense film membranes. Polymers possessing 2,4-diamino mesitylene (DAM) were stable up to 56 atm of pure CO₂, which is due to enhanced charge transfer complex formation compared to polymers containing 4,4′- (hexafluoroisopropylidene) dianiline (6FpDA) and 2,3,5,6-tetramethyl-1,4- phenylenediamine (TmPDA). The new polymers containing DAM and TmPDA showed ideal CO₂/CH₄ selectivities of near 50 with CO₂ and H₂S permeabilities over an order of magnitude higher than Torlon. CO₂ and CH₄ sorption in the DAM- and TmPDA-based materials was reduced, whereas H₂S sorption was enhanced relative to membranes containing fluorinated 6FpDA. Consequently, DAM- and TmPDA-based membranes showed increased stability toward high pressure CO₂ but lower plasticization resistance toward pure H₂S. These results highlight the differences between CO₂ and H₂S that challenge the rational design of materials targeting simultaneous separation of both contaminants.

Full text of DOI:10.1021/ma301249x

Article
pubs.acs.org/Macromolecules
Effect of the Amide Bond Diamine Structure on the CO2, H2S, and CH4
Transport Properties of a Series of Novel 6FDA-Based Polyamide−
Imides for Natural Gas Purification
J. Vaughn and W. J. Koros*
School of Chemical and Biomolecular Engineering, Georgia Institute of Technology, 311 Ferst Drive, Atlanta, Georgia 30308, United
States
ABSTRACT: A series of higher permeability polyamide−
imides based on 2,2′-bis(3,4-dicarboxyphenyl) hexafluoropro-
pane dianhydride with comparable plasticization resistance to
Torlon were synthesized and formed into dense film
membranes. Polymers possessing 2,4-diamino mesitylene
(DAM) were stable up to 56 atm of pure CO2, which is due
to enhanced charge transfer complex formation compared to
polymers containing 4,4′-(hexafluoroisopropylidene) dianiline
(6FpDA) and 2,3,5,6-tetramethyl-1,4-phenylenediamine
(TmPDA). The new polymers containing DAM and
TmPDA showed ideal CO2/CH4 selectivities of near 50 with
CO2 and H2S permeabilities over an order of magnitude higher
than Torlon. CO2 and CH4 sorption in the DAM- and TmPDA-based materials was reduced, whereas H2S sorption was
enhanced relative to membranes containing fluorinated 6FpDA. Consequently, DAM- and TmPDA-based membranes showed
increased stability toward high pressure CO2 but lower plasticization resistance toward pure H2S. These results highlight the
differences between CO2 and H2S that challenge the rational design of materials targeting simultaneous separation of both
contaminants.
N−C imide bond, suggesting high permeabilities could be
achieved in polyamide−imides possessing bulky groups that
increase steric interactions along the polymer backbone. No
information was given, however, regarding their performance
toward H2S or elevated CO2 feed pressures. Huang et al.
measured O2/N2 transport properties in 6FDA-based poly-
amide−imides possessing non fluorinated diamines.5 They
reported O2 permeabilities of 2.71 barrer and an O2/N2 ideal
selectivity of 5.27 for structures with 3,3′,5,5′-tetramethylbis[4-
(4-aminophenoxy)phenyl] sulfone as the diamine. However,
neither CO2 nor H2S transport was measured in their work. We
recently reported on the effects of thermal annealing on a
6FDA-based polyamide−imide (6FDA-3-aminobenzoic acid-
4,4′-hexafluoroisopropylidene dianiline, or 6F-PAI-1). The
CO2/CH4 separation performance was characterized using
feed streams containing toluene as well as H2S and revealed a
stable, mixed gas CO2/CH4 selectivity of 30 at 69 atm total feed
pressure.6 The amide−imide backbone functionality affords
good plasticization resistance due to a combination of
intermolecular hydrogen bonding and charge transfer complex
(CTC) formation. In this work, the diamine forming a tail−tail
connection (amide to amide) during polycondensation in 6F-
PAI-1, namely 4−4′-hexafluoroisopropylidene dianiline, was
1. INTRODUCTION
Polyamide−imides have gained attention for their use as gas
separation materials due to their good mechanical and thermal
properties. They also display high intrinsic efficiencies and
resistance toward aggressive feeds containing condensable
components due to their ability to form interchain hydrogen
bonding as well as charge transfer complexes (CTC). An
example of the robust nature of these materials was
demonstrated by Kosuri and Koros, who measured single gas
CO2 permeability in Torlon 4000T, a commercially available
polyamide−imide, and showed only ∼4% increase in
permeability up to almost 800 psia.1,2 Ideal CO2/CH4
permselectivities of 52 were also reported. Despite these
excellent properties, Torlon’s use in natural gas purification is
not practical because of very low fluxes resulting from a high
degree of chain packing. Several researchers have shown that
gas permeabilities can be increased in polyamide−imides by
incorporation of bulky groups along the main chains which
frustrate chain packing. The work by Nagel and Fritsch gives an
excellent overview of a number of different polyamide−imides
based on isomers of 2,2′-bis(3,4-dicarboxyphenyl)-
hexafluoropropane dianhydride (6FDA) and 4,4′-hexafluoroi-
sopropylidene dianiline as the diamine, along with a detailed
correlation of gas transport properties with free volume.3,4 They
showed CO2 permeabilities as high as 56 barrer along with a
CO2/CH4 ideal selectivity of 36 for structures containing both
fluorine moieties as well as pendent methyl groups ortho to the
Received: June 19, 2012
Revised: August 1, 2012
Published: August 21, 2012
© 2012 American Chemical Society
7036
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
Figure 1. Reaction sequence for 6FDA−polyamide−imide (6F-PAI-1,-2,-3) synthesis.
Table 1. Structure of Torlon 4000T and 6F-PAI-1,2,3 as Well as Diamines Used To Form the Amide Bond in the 6F-PAI Series
of Materials
replaced with nonfluorinated 2,4-diaminomesitylene (DAM)
and 2,3,5,6-tetramethyl-1,4-phenylenediamine (TmPDA) with
the goal to increase the CO2 and H2S permeability and
selectivity.
ity was not expected to decrease as a result of the anticipated
increase in diffusivities, as only a slight size difference between
H2S and CH4 (3.6 vs 3.8 Å) contributes to low diffusion
selectivities that range between 1 and 2. Additionally, an
expected increase in H2S solubility relative to CH4 due to lower
fluorine content was expected to offset any loss in size
discrimination. In this paper, the physical properties of three
different polyamide−imides based on diamines 6FpDA, DAM,
and TmPDA are discussed along with the CO2, H2S, and CH4
transport properties. Additionally, the stability of these
materials toward pure gas H2S and CO2 as well as mixtures
of the two gases is reported and discussed in terms polymer
chain packing, CTCs, and fluorine concentration in the
polymer backbone.
Work by Merkel et al. has shown interesting hydrocarbon
and fluorocarbon solubility trends in various perfluorinated
media.7,8 They showed that H2S solubility is significantly
reduced in both fluorinated glassy and rubbery polymers.9 In
the case of rubbery perfluoroelastomer TFE/PMVE49, which is
based on tetrafluoroethylene, CO2 solubility was higher than
H2S. This was also shown for Cytop, a high free volume, highly
fluorinated glassy polymer. They attributed this unusual
behavior to “unfavorable” interactions between H2S and the
polymer, however no attempt was made to describe the nature
of these interactions. In this work, nonfluorinated DAM and
TmPDA monomers were chosen to increase H2S/CH4
selectivity through an increase in H2S sorption resulting from
lower fluorine concentration in the polymer backbone.
Additionally, pendant methyl groups on DAM and TmPDA
located ortho to the rotatable N−C bond in polyimides have
been shown to induce steric repulsions with neighboring
carbonyl groups, causing the bond planes to configure in a
highly nonplanar conformation, which significantly hinders
chain packing. This is demonstrated in polyimides 6FDA-DAM
and 6FDA-TmPDA,10 which have CO2 permeabilities as high
as 350 and 450 barrer, respectively. These steric interactions
were targeted in the 6F-PAI series in order to increase H2S and
CO2 diffusion coefficients. The overall H2S/CH4 permselectiv-
2. MATERIALS/METHODS
2.1. Synthesis and Dense Film Formation. 6FDA was
purchased from Alfar Aesar and dried at 120 °C under 30 mmHg
vacuum prior to synthesis. All other chemicals were purchased from
Sigma-Aldrich. DAM and TmPDA were dried at 90 °C and 15 mmHg
reduced pressure before the synthesis.
All polyamide−imides were synthesized via two step reaction
sequences originally described by Fritsch and Avella.4 In the first
reaction, the intermediate diacid containing the diimide functionality
was formed by reaction between 6FDA and 2 equiv of 3-aminomethyl
benzoic acid (3-ABA). Imidization was accomplished through thermal
means, where temperatures between 160 and 200 °C were used to
induce ring closure. The water formed was distilled off through
7037
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
azeotropy with xylene. Imidization was confirmed using FTIR-ATR
spectroscopy and shown in Figure 2.
Harrick MVP 2 Series ATR utilizing a N2 purge at 30 cm3/min. An
average of 128 scans with a resolution of 4 cm−1 was taken. This was
done to confirm imide and amide bond formation in 6FDA-3ABA and
final polymer structures, respectively. Bulk powder was used for 6FDA-
3ABA, and dense films with thicknesses ranging from 25 to 50 μm
were used for structure identification of the polymer membranes.
Absorbance spectra are shown in Figure 2.
UV−Vis/Fluorescence Spectroscopy. Ultraviolet−visible spectros-
copy was performed on a Beckman Coulter DU Series 700 general
purpose spectrometer to compare differences in absorbance properties
arising from intramolecular interactions of monomers and dilute
polymer solutions. Concentrations of 1 × 10−6 mol/L in DMF were
used to avoid saturation of the detector.
The diacid containing diimide, 6FDA−3ABA, was subsequently
reacted with the various diamines through direct polycondensation as
reported by Yamazaki et al.11 With this method, a large batch of
6FDA−3ABA can be synthesized and reacted with different diamines
to form a series of polyamide−imides differing only by the diamine
forming the tail-to-tail connection. A schematic of the reaction is given
in Figure 1, and the structures of 6F-PAI-1, -2, and -3 are shown in
Table 1. The details of this reaction are given in our past work6 as well
as in the work by Fritsch et al.4,12 With this reaction, molecular weights
ranged from 50 to 60 kDa as measured by gel permeation
chromatography.
Dense films membranes were fabricated using a draw casting
technique. Polymer was dissolved in a given solvent at 18−20% wt.
solids and spread over a clean glass surface using a stainless steel
casting block with 254−500 μm clearance. 6F-PAI-1 was dissolved in
tetrahydrofuran (THF) within a THF-saturated glovebag. For 6F-PAI-
2 and -3, the casting procedure was modified because of different glass
surface wetting characteristics and solubility of the new polymers. For
6F-PAI-1, the surface was treated with Siliclad (Geleste Inc.) to
increase hydrophobicity, thereby easing the removal of the vitirfied
polymer film. 6F-PAI-2 and -3 exhibited a dewetting effect on this
surface, whereby the polymer solution arranged itself in discontinuous
“pools” in order to minimize surface contact. This is likely due to
increased hydrophilicity of these materials resulting from lower
fluorine content. Because of these issues, a nontreated, plain glass
surface was used to cast 6F-PAI-2 and -3. Additionally, these materials
were only soluble in high boiling, polar solvents such as N-methyl
pyrrolidone (NMP) or N,N-dimethylformamide (DMF). Immediately
after drawing the polymer solution, the glass plate was placed on a
heating element at approximately 120 °C for 2−3 h in order to slowly
evaporate enough solvent to vitrify the film. The solid films were then
peeled off of the plate and soaked in methanol for 12 h at room
temperature, followed by slow evaporation of methanol over 2 h. The
films were then annealed in a vacuum oven at 30 mmHg absolute
reduced pressure at 220 °C for 24 h, followed by slowly cooling over a
period of 4 h. Without solvent exchanging, a significant amount of
residual solvent remained in the membranes which greatly affected the
resulting transport properties.
Solid state fluorescence spectroscopy (Horiba Scientific) was used
to study differences in intermolecular charge transfer formation. A right
angle configuration was used, and both excitation and emission profiles
were collected. Excitation profiles were taken between 250 and 480 nm
and monitored at 485 nm. Emission profiles were excited at 350 and
440 nm. The resulting spectra are shown in section 3.2.
2.4. Gas Permeation and Sorption. Permeability. Pure gas CO2,
CH4 and H2S permeability measurements were taken as a measure of
the intrinsic separation capability of the dense polymer films. A
constant volume, variable pressure apparatus was used.14 The
permeability coefficient is defined as the flux of the gas normalized
by the transmembrane fugacity difference and film thickness (l) as
defined in eq 1. The flux of a penetrant gas molecule, neglecting bulk
flow, is described by diffusion jumps that depend on both the size of
localized free volume fluctuations of the polymer chains as well as the
frequency of these fluctuations. The steady state flux is given by eq 2.
JAl
PA =
ΔfA
(1)
∂CA
∂x
J = −D
A
A
(2)
For mixed gas measurements, the permeability was calculated by eq
3. JT is the total flux of all components permeating through the
membrane and yA is the mole fraction of component A in the
downstream. The compositions of the gases were determined using a
gas chromatograph (Varian 450). The average of 3 injections was
taken.
2.2. Density and Fractional Free Volume. The density of dense
film membranes was measured using a density gradient column
prepared from calcium nitrate. The average of at least three sections of
a dense polymer film was added to the column to obtain the density of
each sample. Fractional free volume was then calculated from the
following equation:
JTlyA
PA =
ΔfA
(3)
The ideal permselectivity given in eq 4 is described as the intrinsic
separation capability of a polymer material, and is calculated from the
ratios of the pure gas permeability coefficient of the fast gas relative to
the slow gas. For mixed gas feeds, a separation factor is used as shown
in eq 5. In the case of very low pressure in the downstream relative to
the upstream side of the membrane, as was the case in this study, the
ratio of the mixed gas permeability coefficients is equal to the
separation factor.
̂
̂
o
V − V
g
FFV =
̂
V
g
Vg is the specific volume of the glass at a certain temperature and is
determined from the density measurements. Vo is the specific occupied
volume and is calculated from group contribution methods according
to Bondi.13
2.2. Thermal Analysis. Residual solvent content in the films was
checked with thermal gravimetric analysis (TGA) (STA 409 PC Luxx)
using N2 at 30 cm3/min as the purge gas. Up to 5% residual solvent
remained in the films not treated with MeOH. After solvent
exchanging, films generally showed less than 0.5% residual solvent
content.
Differential scanning calorimetry (TA Instruments, Q200 Series)
was used to determine the glass transition temperature (Tg) of the
dense films. N2 at 50 cm3/min was used as the purge gas. The films
were heated at 10 °C/min from 25 to 370 °C, followed by cooling to
25 °C, then repeating the cycle. The second heating run was used to
calculate the tangent line at the step transition and was taken as the
intrinsic Tg. The glass transitions temperatures are reported in Table 2.
2.3. Spectroscopy. FT-IR ATR Spectroscopy. Fourier transform
infrared attenuated total reflectance spectroscopy (FT-IR ATR) was
performed using a Bruker-Tensor 27 spectrometer equipped with a
PA
α*A/B
=
P
(4)
B
[yA/yB]permeate
[yA/yB]feed
α
=
A/B
(5)
Gas transport through polymer membranes also depends on gas
solubility. As such, the permeability coefficient can be described as the
product of the diffusion and sorption coefficients via eq 6, where DA
and SA are the diffusion and sorption coefficients, respectfully. Finally,
from eq 4, the ideal permselectivity can be given as the product of the
diffusion and solubility selectivities as shown in eq 7.
PA = D × SA
(6)
A
7038
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
Figure 2. (a−d) FT-IR ATR spectra in the 1900−1400 cm−1 range.
⎡
⎤
⎡
⎤
CH′bp
1 + bp
D
SA
A
CH
=
α*A/B = ⎢ ⎥ × ⎢
⎥
⎦
(9)
(10)
D
S
⎣
⎦
⎣
(7)
B
B
CD = kDp
Single gas permeability measurements were done at 4.5 atm
upstream pressure while the downstream was maintained at less than
1% of 1 atm. In this situation, the transmembrane fugacity difference in
eq 1 can be approximated as the upstream pressure without serious
error. The gas was allowed to permeate through the membrane for a
time period of 8−10 times the diffusion time lag in order to reach
steady state, as discussed by Koros and Paul.15 In accordance with their
work, the single gas diffusion coefficients were calculated using eq 8,
which relates gas diffusivity to the steady state permeation time lag.
p is the equilibrium pressure of the gas in contact with the membrane.
CH′ is the Langmuir capacity constant and represents the maximum
amount of penetrant gas that can adsorb into these cavities. The
Langmuir affinity constant, b, is the affinity of the gas for this region.
The Henry's law coefficient, kD, characterizes the ease to which
polymer chains can open up to accommodate a penetrant gas
molecule, and is analogous to the sorption coefficient in rubbery
polymers. Together, these regions sum to give the total concentration
as shown in eq 11.
l2
6θ
CH′bp
1 + bp
D =
C = kDp +
A
(8)
(11)
The sorption coefficient in eq 6 is the pressure normalized
concentration of the penetrant gas molecule inside the polymer
membrane, and is given by eq 12.
For the permeability vs pressure isotherms, the permeability was
measured at a given pressure after 8−10 time-lags. The pressure was
then increased by 2−3.5 atm and the membrane was allowed to reach
steady state again before a measurement was taken. This process of
incremental pressure increases was performed up to the point where
the permeability coefficient began to increase, whereby the pressure
was held constant for approximately 12 h until a measurement was
taken. Because plasticization involves small and/or large scale chain
relaxations, permeability beyond the plasticization pressure can be time
dependent. This phenomena is discussed in detail in the work by Wind
et al.16 The 12 h permeation time was performed to provide a
meaningful comparison of the magnitude of permeability increase
between the different materials once membrane plasticization began.
Gas Sorption. Gas sorption in glassy polymers is envisioned to
occur in two different idealized environments.17 The Langmuir mode
consists of molecular scale packing defects or “holes” that occur as a
result of the polymer being below its glass transition temperature (Tg).
The concentration in this environment, CH, is described by eq 9. The
Henry’s Law environment consists of polymer chains in packing
equilibrium with one another and is analogous to gas dissolution in a
rubbery polymer. The concentration in this mode, CD, is described in
eq 10.
CH′b
1 + bp
S = kD +
(12)
In this work, the concentration vs pressure isotherms were
measured via a pressure decay technique, which is described in detail
elsewhere.18 The dual-mode parameters CH′, b, and kD were obtained
by fitting the concentration isotherms to eq 11 using a nonlinear least-
squares method. The sorption coefficients were then calculated from
eq 12. The data fit the equation well, and in most cases, R2 was above
0.98. Standard deviations of the fit are reported in Table 5. For H2S,
membrane swelling occurred above ∼5.5 atm for 6F-PAI-2 and -3,
which was signified by an upswing in the isotherm that was concave to
the y-axis. This caused difficulty in fitting the data due to nonlinearity
in the isotherm at elevated pressures. In this case, only qualitative
estimates can be made.
3. RESULTS AND DISCUSSION
3.1. IR Spectroscopy. ATR absorbance spectra for 6FDA-
3ABA and the various 6F-PAI analogues is shown in Figures 2
7039
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
Figure 3. FT-IR ATR spectra in the 3700−2600 cm−1 range showing the N−H stretching mode centered around 3350 cm−1.
and 3. Symmetric and asymmetric imide stretches are shown
between 1700 and 1800 cm−1, confirming imide bond
formation in 6FDA-3ABA and preservation of this bond in
the subsequent polymers. Additionally, the carbonyl stretch of
the carboxylic acid in 6FDA-3ABA occurs near 1690 cm−1,
which is depicted in Figure 2d. The carbonyl stretch in the
amide bond occurs at lower wavelengths, and is shown near
1660 cm−1 for the polyamide−imides. The amide bond also
results in a broad, medium strength N−H stretching frequency
centered around 3350 cm−1. This confirms the formation of the
amide bond as well as highlights the nature of interchain
hydrogen bonding. The broad peak results from varying
degrees of hydrogen bonding between the randomly packed
polymer chains. Thus, the shape and span of this peak
qualitatively suggests the N−H groups exist in a hydrogen
bonded state. It is difficult, however, to distinguish between
bonded and nonbonded protons, and so the authors do not
Figure 4. UV−vis spectra comparing (a) 6F-PAI-1, (b) Torlon, (c)
6FDA-3ABA, (d) 6F-PAI-2, and (e) 6F-PAI-3.
make any attempt to assess different degrees of hydrogen
bonding between the three different structures.
3.2. UV−Vis/Fluorescence Spectra. Solution state UV−
vis was performed in dilute conditions to wash out
intermolecular interactions and thus examine differences in
intramolecular charge transfer, such as π electron conjugation
across the polymer backbone. Figure 4 compares the precursor
diacid 6FDA-3ABA with polymers 6F-PAI-1, -2, and -3. Torlon
was also measured as a reference for which to compare the 6F-
PAI series.
The wavelengths of maximum absorption occur at 270 nm
for all four polymers as well as 6FDA-3ABA. This absorption
results from the conjugated π electrons of the aromatic rings,
which exist in all five structures. The absolute intensities are not
as significant, as small differences in concentration can affect
this. Small variances in the scale can occur when measuring very
low milligram quantities, and while attempts were made to keep
the concentrations the same, it is likely that small differences in
solid addition occurred, which may give rise to different
intensities. Therefore, only the qualitative aspects, such as peak
position and shape are analyzed. As can be seen in Figure 4, the
shape of the peaks as well as absorption cutoff wavelength are
approximately equal for 6FDA-3ABA and 6F-PAI-2. This shows
that very little difference in conjugation occurred by addition of
DAM as the amide linker, which is the only difference between
the two structures. The methyl group ortho to the N−C imide
bond in 6FDA-3ABA inhibits planarity between the 6FDA and
3ABA moiety, thus it is unlikely for the lone pair electrons of
the imide nitrogen to conjugate with the aromatic π electrons.
Similarly, methyl groups on DAM that are located ortho to the
amide nitrogen in 6F-PAI-2 inhibit planarity around the N−C
amide bond. Thus, intrinsically, no significant difference in
intramolecular charge transfer occurs across the backbones of
the diacid monomer 6FDA-3ABA and 6F-PAI-2. 6F-PAI-3,
however, shows an absorption cutoff around 280 nm, which is
lower than all other materials measured. Para configurations
7040
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
such as in TmPDA have greater rotational mobility compared
to meta.3 Thus, rotation around the amide N−C bond may be
less restricted in 6F-PAI-3 compared to DAM in 6F-PAI-2,
which further decreases planarity across this bond. This may
reduce electron conjugation across the N−C bond, resulting in
lower wavelengths of light absorption. This is also supported by
the lower extent of intermolecular CTC formation in 6F-PAI-3
as will be discussed in the next section.
6F-PAI-1 and Torlon both show broader absorption peaks
that are red-shifted toward longer absorption wavelength
cutoffs. 6F-PAI-1 contains six aromatic rings per repeat unit,
which is one higher than Torlon and 6F-PAI-2 and -3. The
peak intensity and broadening may be due to increased
concentration of aromatic π electrons, however, Torlon also
shows peak broadening that is similar in shape to 6F-PAI-1 and
higher in intensity than 6F-PAI-2 and -3. Therefore, the red
shift in peak width and longer absorption cutoff wavelengths is
due to increased electron conjugation across the polymer
backbone. Because Torlon has no bulky pendent groups on the
aromatic rings, it has a higher degree of backbone planarity than
6F-PAI-2 and -3. The lone pair electrons on the phenyl ether
oxygen in Torlon can also conjugate with the aromatic rings,
which results in a broader peak that is shifted toward longer
absorption wavelengths. For 6F-PAI-1, the increase in intra-
molecular conjugation occurs over the amide bond since it
contains the same imide functionality as 6F-PAI-2 and -3.
Although the bulky CF3 groups of the 6FpDA diamine inhibit
close intermolecular chain packing in the solid state, this effect is
not relevant in the dilute solution state, where interactions
between adjacent polymer chains are unlikely. 6FpDA is also
absent of pendent methyl groups which could induce rotation
around the N−C bond and reduce planarity. Thus, the UV−vis
spectra reveal the backbones of Torlon and 6F-PAI-1 have
intrinsically higher extents of π-electron conjugation, which is
believed to be due to a greater degree of planarity resulting
from the absence bulky pendent groups along the polymer
backbones.
Solid State Fluorescence Spectroscopy. Intermolecular
charge transfer has been shown to increase membrane stability
against swelling and plasticization by increasing dipole−dipole
types of physical interactions in the solid state between
electron-rich and -poor regions on adjacent chains.6,19 Red
shifts in wavelength of maximum absorption indicate greater
electron conjugation, which can be either intra or intermo-
lecular in nature. Because of the difficulty in obtaining films thin
enough to avoid saturation of the UV−vis detector,
fluorescence spectroscopy was used instead. Changes in
position and shape of the fluorescence excitation profiles
signify the formation of CTCs, thus it can be used to compare
changes to the polymer microstructure in the solid state. Figure
5 shows the fluorescence excitation profiles for 6F-PAI-1,2 and
3, Torlon, and polyimides 6FDA-6FpDA and 6FDA-DAM.
6FDA-TmpDA polyimide was not measured, however, the
authors expect a similar absorption spectra as 6FDA-DAM due
to similar steric interactions resulting from pendent methyl
groups ortho to the N−C imide bond.
Figure 5. Fluorescence excitation spectra comparing 6FDA polyimides
and Torlon (solid lines) and 6FDA polyamide−imides (dashed lines).
wavelengths of maximum absorption. Although 6F-PAI-1
showed increased intramolecular charge transfer in the UV−
vis spectra, in the solid state, the bulky −(CF3)- groups inhibit
close, parallel chain packing. This can significantly interfere
with parallel stacking of adjacent aromatic and imide groups
required for intermolecular CTC formation. For 6F-PAI-3, the
pendent methyl groups ortho to the amide bond induce
rotation around the N−C bond as a result of TmPDA having
greater rotational mobility. This can also hinder the close
overlap of adjacent polymer chains as do the CF3 groups in
6FpDA, thus limiting the formation of intermolecular CTCs.
The low UV−vis absorption cutoff wavelength in 6F-PAI-3
shown in Figure 4 supports the idea that its backbone is highly
nonplanar. A comparison with the analogue polyimides shows
that CTC formation decreased with meta connected DAM in
the polyimide series relative to para connected 6FpDA. This
suggests that DAM may inhibit planarity and polymer chain
packing when connected over an imide bond, however, the
opposite effect occurs when DAM is incorporated over an
amide bond as it is in 6F-PAI-2, where intermolecular CTC
formation is enhanced. This leads to large permeability
increases in the polyimide series whereas permeabilities are
reduced in the polyamides−imides. The permeability trends
will be further discussed in section 3.4.
The excitation spectra for Torlon shows two absorption
bands, the first near 370 nm and the second near 450 nm. This
longer wavelength absorption gives rise to the more deeply
colored (yellow) appearance of Torlon, and suggests charge
transfer is enhanced relative to the 6F-PAI-1 and -3, which are
both much lighter in color. 6F-PAI-2, which also absorbs near
460 nm, also displays a much darker color than 6F-PAI-1 and
-3. Because 6F-PAI-2 showed a lower UV−vis absorption cutoff
wavelength compared to Torlon, the second absorption band
near 460 nm for 6F-PAI-2 shown in the fluorescence excitation
spectra reveal that CTC formation in this material most likely
occurs intermolecularly, whereas in Torlon, the darker color
resulting from increased charge transfer is likely both inter as
well as intramolecular in nature.
As shown in Figure 5, 6F-PAI-2 formed additional CTCs
relative to 6F-PAI-1 and -3. Both the shape and bandwidth
changed along with a red-shift in the wavelength of maximum
absorption. Meta-connected DAM increases chain packing
relative to para connected TmPDA. This facilitates electronic
interactions between adjacent chains in the solid state. 6F-PAI-
1 and -3 show similar shaped absorption bands as well as
Fluorescence emission profiles are given in Figure 6. All
polymers show wavelengths of maximum intensity near 490 nm
when excited at 350 nm. Additionally, 6F-PAI-2 excited at both
350 and 440 nm emits at the same wavelength. This indicates
the formation of a ground state CTC, further suggesting the red
7041
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
A comparison of the calculated fractional free volumes (FFV)
reveals that the polyamide−imide FFV decreases with
incorporation of DAM and TmPDA monomers, while the
same substitutions in the polyimides results in an increase in
the FFV. This unexpected trend is likely due to differences
between amide and imide bond sterics, and correlates well with
the differences in fluorescence behavior as shown in Figure 5.
DAM and TmpDA are connected over an imide bond in the
polyimides, where two carbonyl groups contribute to steric
repulsions with pendent methyl groups. In contrast, only one
carbonyl exists over the amide bond. Furthuremore, the imide
nitrogen in the polyimides likely has less degrees of vibrational
freedom due to it being bonded to two carbonyl carbons. This
effectively secures the imide nitrogen in place, and allows only
for rotation around the N−C diamine bond to minimize the
steric repulsions between the methyl and carbonyl groups. On
the other hand, the amide nitrogen may have more vibrational
degrees of freedom to minimize the carbonyl methyl group
interactions. These factors are believed to contribute to much
lower planarity between the diimide and diamine moieties in
DAM and TmPDA polyimides compared to their correspond-
ing polyamide−imides, thus causing the opposite trends in
FFV. Within the polyamide−imide series, chain packing is
promoted by incorporation of DAM and TmPDA over the
amide bond. The decrease in FFV for 6F-PAI-2 and -3 relative
to 6F-PAI-1 is due to the packing inhibiting CF3 groups
attached to 6FpDA, as well as the more rigid, linear structures
of DAM and TmPDA. Previous researchers have show that
rigid, linear rod-like diamines increase chain packing in amide
bonds as compared to diamines containing a flexible kink such
as in the isopropyl carbon of 6FpDA in 6F-PAI-1, which can
increase chain mobility and disrupt chain packing along the
backbone.4 These interesting trends will be further discussed in
context with the transport data in section 3.4.
Figure 6. Fluorescence emission profiles for 6F-PAI-1, -2 and -3. 6F-
PAI-2 excited at both 350 and 440 nm is shown.
shifting of excitation wavelengths shown in Figure 5 is due to
intermolecular π−π stacking. The fluorescence emission shape
and bandwidth are similar for 6F-PAI-1 and -3, as it was for the
excitation profiles. This indicates similar intermolecular
interactions exist between polymer chains for both materials.
Excitation of 6F-PAI-2 at 440 nm, corresponding to the
maximum excitation wavelength in Figure 5, caused a second
band to appear near 515 nm, which may be due to an additional
CTC arising from the solid state packing promoting meta
connection. The effect of charge transfer on the permeation
properties and plasticization resistance will be further discussed
in section 3.4 and 3.5.
3.3. Physical and Thermal Properties. The physical
properties for the 6F-PAI series as well as the analogue
polyimides are given in Table 2. An increase in Tg is observed in
The X-ray diffraction patterns in Figure 7 confirm the
amorphous nature of all three structures. d-spacing, which was
Table 2. Physical Properties of 6F-PAI and Analogue
a
Polyimides
d-
spacing
(Å)
ρ
Tg
vf (cm3/
mol)
ε (dielectric
constant)
(g/cm3) (°C)
FFV
Polyamide−Imide
6F-PAI-1
6F-PAI-2
6F-PAI-3
Torlon
1.3867
1.3374
1.3281
1.3550
324
330
350
295
135
0.18
0.14
0.13
0.15
4.5
6.3
6.3
4.2
3.6
4.1
4.0
5.3
87
84
70
Analogue Polyimides
6FDA-
6FpDA
1.4660
1.3342
1.3247
320
395
420
80
79
80
0.16
0.19
0.19
5.9
6.3
6.3
3.3
3.4
3.4
6FDA-
DAM
6FDA-
TmPDA
a
Polyimide values for d-spacing, Tg and density were adapted from the
following: 6FDA-6FpDA, Coleman and Koros;286FDA-DAM, Kim et
al.;29 6FDA-TmPDA, Powell30 and Lui;31 Torlon, Teoh et al.32
Figure 7. XRD patterns depicting the amorphous nature of 6FDA-
based polyamide−imides. For 6F-PAI-1, the average of the two peaks
was used.
the structures bearing pendent methyl groups. This trend is
shown in both the polyamide−imides as well as the polyimides,
and results from steric interactions from the methyl groups
which inhibit localized, rotational mobilities. Consequently,
long-range, segmental scale rigidity is increased in 6F-PAI-2 and
-3 as well as their corresponding analogue polyimides.
calculated using the Bragg equation, increases in structures
based on DAM and TmPDA, which is shown across both series
of polymers. Higher d-spacing in the polyimide series is
consistent with higher FFV in 6FDA-DAM and 6FDA-
TmPDA compared to 6FDA-6FpDA, however, they are not
7042
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
a
Table 3. Pure Gas Permeabilities and Permselectivities at 35 °C and 4.5 atm
polyamide−imide
PCO
PCH
PH S
6.2 7 × 10−3
PCO /PCH (ideal selectivity)
PH S/PCH (ideal selectivity)
2
4
2
2
4
2
4
6F-PAI-1
6F-PAI-2
6F-PAI-3
Torlon
32.8 0.05
14.2 0.1
0.79 0.01
42 0.6
49 0.02
47 0.02
52.8 0.6
8.1 0.04
10.3 0.02
10.9 0.02
14.8 1.9
0.29 5 × 10−3
0.46 0.01
3
0.04
21.6 0.03
0.7 8 × 10−3
5.0 0.04
0.2 0.02
1.3 × 10−2 1.0 × 10−3
polyimides
PCO
PCH
PH S
PCO /PCH (ideal selectivity)
PH S/PCH (ideal selectivity)
2 4
2
4
2
2
4
6FDA-6FpDA
46
1.5
30
20
16
b
6FDA-DAM
350
456
17.6
28.4
nd
nd
b
6FDA-Durene
a
Permeability is given in units of barrer:
cm3(STP)·cm
1 barrer = 1 × 10−10
cm2·s·cmHg
b
Sridhar et al.10
consistent with decreases in FFV as shown in the polyamide−
imides. These anomalous trends in d-spacing were also
observed by Nagel et al., where materials with lower d-spacing
displayed higher gas permeabilities.3 They measured the
average hole size of the free volume elements using positron
annihilation spectroscopy (PALS) and found that the materials
with lower d-spacing had higher average hole sizes, which they
attributed to the reason for higher permeability. Such
phenomena would help to explain the trends in d-spacing,
fractional free volume, and permeability in the materials in this
work, however PALS was not available to the authors at the
time of this work. Nevertheless, reduced fractional free volume
as calculated in Table 2 is consistent with the permeability data
which will presented in section 3.4 for the polyamide−imides,
and supports the idea that DAM and TmPDA diamines
increase chain packing over amide bonds but decrease it when
incorporated over an imide bond.
The dieletric constant was also predicted from group
contribution methods and shown in Table 2 to illustrate the
effect of fluorine content in the polymer backbone. The
predicted increase with lower fluorine concentration in both the
polyamide−imide as well as the polyimide series can be
attributed to increased polarizability due to lower fluorine
content across the polymer backbone. The increased polar-
izability has significant implications on the sorption of H2S and
will be further discussed in section 3.4.
3.4. Permeation and Sorption Analysis. The overall goal
of this work is to characterize robust, plasticization resistant
membrane materials for aggressive natural gas separations.
Industrially attractive materials need to have both high
permeability and selectivity. An increase in selectivity from 15
to 30 can result in a 25% reduction in overall process cost,20
thus selectivity is extremely important when considering
materials for a given separation. Additionally, H2S is an
important contaminant that has not been widely studied due to
handling issues associated with its toxic nature. In our lab, state
of the art facilities with the proper safety precautions were
constructed to allow us to safely study pure H2S at elevated
pressures. All permeation cells were located within a sealed
plexiglass fume cabinet under lower total pressure relative to
atmospheric pressure outside of the cabinet. Before accessing
permeation cells that had been exposed to H2S, the gas was
purged from the systems into a NaOH scrubber followed by
rigorous purging of the cell with N2 and pulling vacuum for at
least 6−8 times the diffusion time lag in order to completely
evacuate residual H2S. Additionally, respirators and H2S
monitoring badges were donned while working in the lab.
The pure gas transport properties for the polyamide−imides
are shown in Table 3. The CO2 and CH4 permeabilities for the
polyimide series, which were gathered from the literature, are
shown for comparison. H2S was not measured in the
polyimides due to their high susceptibility to plasticization at
low feed pressures.
The selectivities of 6F-PAI-2 and -3 increased relative to 6F-
PAI-1, and approach those shown by Torlon; however the CO2
and H2S permeabilities are over an order of magnitude higher.
These results show that polyamide−imides based on the
packing inhibiting 6FDA and rigid, rod-like diamines containing
pendent methyl groups combine to form materials with good
gas permeability while still maintaining excellent selectivity.
A comparison of the permeabilities between the two types of
analogous polymer series in Table 3 clearly shows the effect
DAM and TmPDA being connected over different types of
bonds. 6F-PAI-2 and -3 show lower permeabilities for all gases
relative to 6F-PAI-1 in the polyamide−imides. In sharp contrast,
the permeabilities are significantly increased in the polyimide
series. As previously discussed with the fluorescence spectra
and FFV values, chain packing is enhanced in the DAM and
TmPDA connected polyamide−imides. Within the 6F-PAI
series, the 6FpDA connected material tends to increase
fractional free volume and chain mobilities, thus permeability
is lowered in the more rigid, linear DAM- and TmPDA-based
materials. This leads to higher selectivities for both the CO2/
CH4 and H2S/CH4 gas pairs for 6F-PAI-2 and -3.
The apparent diffusion coefficients were calculated from eq 8.
Sorption coefficients were predicted from dual-mode fits to the
experimentally determined concentration isotherms as de-
scribed in section 2.4. In general, reasonable fits to the
concentration isotherms were obtained as shown in Figures
8−10. Along with the difficulties in fitting H2S isotherms as
already discussed, for 6F-PAI-3, very low sorption at higher
pressures reduced the ability of the nonlinear least-squares
fitting program to obtain accurate fits to the data. In this case,
the dual-mode parameters were estimated manually, with the
resulting values bringing reasonable fits to the data. These
issues contributed to overall error in the predicted sorption
values, and as such are associated with higher experimental
uncertainty. Nevertheless, a couple of important trends can still
be noted from the data shown in Tables 4−6.
7043
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
volume available for gas sorption. Additionally, for 6F-PAI-2,
the diffusion coefficients for all gases decreased relative to 6F-
PAI-1 and -3, which is an interesting trend considering it shows
higher free volume, FFV, and increased sorption relative to 6F-
PAI-3. Increased charge transfer interactions in 6F-PAI-2 cause
higher cohesive forces between the polymer chains. This leads
to lower diffusivities as more energy is required for a transient
gap to open up for a diffusion step. Consequently, 6F-PAI-2
shows the highest CO2/CH4 diffusion selectivity.
The sorption and diffusion coefficients of all gases in Torlon
are lower than all three 6FDA-based polyamide−imides. The
absence of packing inhibiting groups and greater backbone
planarity, as supported by the UV−vis and fluorescence
excitation spectra, contribute to a tighter packed micro-
structure, which reduces both the free volume available for
gas sorption as well as the frequency of transient gap openings
available for diffusion jumps. On the other hand, para
connected TmPDA showed slightly increased diffusivities
Figure 8. CH4 concentration vs pressure at 35 °C. Solid lines
represent dual-mode model fits to the experimental data.
relative to 6F-PAI-2, while compared to 6F-PAI-1, only DCO
2
and DCH appear to decrease outside of the standard deviation.
4
This caused a reduction in the CO2/CH4 diffusion selectivity
relative to 6F-PAI-2, whereas the H2S/CH4 diffusion selectivity
was not significantly affected. It is interesting to note that the
H2S/CH4 diffusion selectivity is between 0.8 and 1.6 for all
materials, as diffusion should favor the molecule with lower
kinetic diameter, which in this case is H2S. Such low diffusion
selectivity may be due to more favorable sorption interactions
between H2S and various polar substituents along the polymer
backbone such as the amide N−H or carbonyl CO. This
would increase the energy barrier required to make a diffusion
jump. In summary, it is evident that for H2S, the permselectivity
is predominantly governed by sorption selectivity.
From the concentration isotherms shown in Figures 8−10,
the sorption of both CH4 and CO2 was reduced in 6F-PAI-2
and -3 relative to 6F-PAI-1, causing an increase in the CO2/
CH4 and H2S/CH4 sorption selectivity. These changes can be
further analyzed through the dual-mode framework. As shown
in Table 6, the Langmuir capacity constant, or CH′, is not
substantially affected by the diamine structures. This is also
visually evident in the concentration isotherms, where CH′ can
be found from the intersection of the back-extrapolated, high
pressure linear region of the isotherm with the concentration
axis. From this data, it appears that the diamine structure has a
much greater impact on the equilibrium packing region, as
sharp decreases in Henry’s law sorption occurred. This is
probably due to the more rigid and linear structure of DAM
and TmPDA, which reduced free volume and increased the
rigidness of the polymer matrix. The kink in the isopropyl
carbon of 6FpDA increases flexibility in this region, which
allows for better accommodation of gas molecules into the
densely packed polymer chains. These factors are expressed by
significantly reduced kD values for CO2 and CH4 in 6F-PAI-2
and -3.
For H2S, total concentration in the membrane increased in
the DAM and TmPDA analogues despite the fact that CO2 and
CH4 sorption was lowered. Reduced fluorine content in these
structures increases overall polarizability, which is expressed by
an increase in the dielectric constants as shown in Table 2. The
highly electronegative fluorine atoms, combined with a short
atomic radius, reduce the ability of the valence electrons to be
distorted by an external field. Dipole−induced dipole
interactions between polar H2S and the polymer backbone
Figure 9. CO2 concentration vs pressure at 35 °C. Solid lines
represent dual-mode model fits to the experimental data. The dual-
mode fit to 6F-PAI-3 was estimated manually.
Figure 10. H2S concentration vs pressure at 35 °C.
As can be seen in Tables 4 and 5 as well as Figures 8−10,
reductions in CO2 and CH4 permeability for 6F-PAI-2 and -3
relative to 6F-PAI-1 were due to both lower sorption and
diffusion coefficients. Reduced sorption in these materials is
particularly pronounced in the high pressure, Henry’s Law
region, where rigid, rod-like DAM and TmpDA promote chain
packing compared to kinked 6FpDA. This reduces the free
7044
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
a
Table 4. Apparent Diffusivities and Diffusion Selectivities at 35 °C and 4.5 atm
polyamide−imide
DCO (×108)
DCH (×108)
DH S (×108)
DCO /DCH
DH S/DCH
2
4
2
2
4
2
4
6F-PAI-1
6F-PAI-2
6F-PAI-3
Torlon
1.9 0.1
1.1 0.2
1.4 0.3
0.1
0.2 0.01
0.1 0.04
10 0.8
12 0.2
0.8 0.2
0.1 6 × 10−3
0.1 7 × 10−3
0.3 0.1
1
0.08
0.2 3 × 10−3
9 × 10−3 8 × 10−4
9
0.2
1.6 0.6
8 × 10−3 2 × 10−5
10 8 × 10−4
0.8 0.07
a
Units of the diffusion coefficients are cm2/s.
a
Table 5. Sorption Coefficients and Selectivities at 35 °C and 4.5 atm
polyamide−imide
SCO
SCH
SH S
SCO /SCH
SH S/SCH
2
4
2
2
4a
2
4
0.03
0.02
6F-PAI-1
6F-PAI-2
6F-PAI-3
Torlon
7.2 0.7
6.8 0.9
2.4 1.2
1.8 0.5
0.7 0.05
0.4
12.2 2.4
14.4 1.6
13 2.4
3.0 0.02
3.7 0.02
6.4 0.03
6
5
8
4.5
2
18 0.04
18
2.4
7.2 0.9
a
Sorption coefficients are given in units of (cm3(STP)/cm3 × atm). Values given are predicted from dual-mode parameters as estimated from the
concentration isotherms.
Table 6. Dual-Mode Sorption Parameters at 35 °C
a
CO2
CH4
H2S
CH′
CH′
CH′
polyamide− kD (cm3(STP)/cm3
(cm3(STP)/
cm3)
b
kD (cm3(STP)/cm3
× atm)
(cm3(STP)/
cm3)
b
kD (cm3(STP)/cm3
× atm)
(cm3(STP)/
cm3)
b
imide
× atm)
0.1
(atm−1
)
(atm−1
)
(atm−1
)
6F-PAI-1
6F-PAI-2
6F-PAI-3
Torlon
31.2
37.6
33.5
18
0.05
0.04
0.02
0.02
0.04
15
0.02
0.02
0.1
0.4
0.4
−
32
0.2
0.2
−
0.04
3 × 10−5
2 × 10−3
3 × 10−3
14.4
3.5
36.8
b
5 × 10−6
5 × 10−5
−
17
2.1
0.03
0.28
0.06
a
b
H2S values for 6F-PAI-3 could not be calculated because of membrane swelling. Values were estimated manually.
Figure 11. Pure gas CO2 permeability vs pressure at 35 °C. The Torlon permeability axis is depicted on the right. CO2 permeability vs pressure data
for Torlon was adapted from Kosuri.1
are enhanced by more polarizable DAM and TmPDA relative
to 6FpDA. This suggests H2S sorption is more strongly
dependent on intermolecular interactions than are the
quadrapolar CO2 and nonpolar CH4, where sorption is mostly
governed by condensability and free volume. This increased the
H2S/CH4 solubility selectivities for both 6F-PAI-2 and -3, thus
overall H2S/CH4 permselectivity was increased slightly.
3.5. Plasticization Analysis. Natural gas wells can reach
pressures above 70 atm, where elevated partial pressures of
CO2 and H2S can cause significant swelling and plasticiza-
tion,21,22 resulting in large selectivity losses. Cellulose acetate
and most un-cross-linked polyimides plasticize at CO2 partial
pressures near 14 atm,23,24 and with the combined effect of
highly condensable H2S, this pressure is expected to be much
lower. Thus, the need for intrinsically stable materials that can
extend high selectivities into these aggressive conditions
underscores material development for natural gas separation
materials. Figures 11 and 12 show pure gas permeability vs
pressure isotherms for CO2 and H2S, respectively.
7045
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
Figure 12. Pure gas H2S permeability vs pressure at 35 °C. The Torlon permeability axis is shown on the right.
Torlon, widely recognized for its outstanding plasticization
resistance, is shown again as a baseline for which to compare
the performance of the 6F-PAI series. From Figure 11, CO2
plasticization resistance increases in the order 6F-PAI-3 ∼ 6F-
PAI-1 < 6F-PAI-2 < Torlon. This is inferred from the
plasticization pressure, which is the pressure at which the
permeability coefficient begins to increase from its minimum
value. The plasticization pressure for 6F-PAI-3 of 24 atm is
slightly lower than 6F-PAI-1, which begins to increase around
28 atm. The overall extent of permeability increase, however, is
∼10% for both materials, suggesting similar resistance to chain
swelling once a threshold CO2 concentration in the membrane
was reached. This similar resistance occurs despite the fact that
kD for 6F-PAI-3 was orders of magnitude lower than kD in 6F-
PAI-1. The fluorescence excitation spectra in Figure 5 suggest
that similar intermolecular interactions exist between the
polymer chains in 6F-PAI-1 and -3, and that no additional
CTCs formed, which would certainly alter the excitation
spectra. Thus, the similar magnitudes in permeability increase
are likely the result of similar intermolecular interactions
throughout the entire polymer network in both materials, not
just chain mobility in the Henry’s environment as inferred from
kD and previously suggested by Wind et al.23 Instead, for
structures with similar charge transfer interactions, the absolute
plasticization pressure appears to depend on the pressure at
which the Langmuir sorption sites become saturated. As can be
seen in Figure 11, both 6F-PAI-1 and -3 show initial upswing in
permeability once the permeability stops decreasing with
pressure. In other words, the permeability continues to
decrease as long as the microcavities have not been saturated
by penetrant gas molecules. As sorption begins to occur
predominantly in the Henry’s environment with higher
pressure, the permeability begins to increase, which suggests
polymer chain swelling occurs when gas sorption occurs
primarily in the densely packed region. Therefore, the higher
plasticization pressure in 6F-PAI-1 may be due to higher
pressures being required to saturate the Langmuir sorption
sites. 6F-PAI-2 shows the highest plasticization pressure out of
the fluorinated polyamide−imides, which occurs near 30 atm.
Additionally, beyond the plasticization pressure, the CO2
permeability increases by less than 5%. Similar to 6F-PAI-1
and -3, the permeability in 6F-PAI-2 continues to decrease
slightly until the point at which the permeability begins to
increase, again suggesting that the polymer chains begin to
swell once additional sorption occurs primarily in the Henry’s
environment. For 6F-PAI-2, however, lower increases in
permeability correlates with significantly greater CTC for-
mation relative to 6F-PAI-1 and -3 as discussed in section 3.2.
Such charge transfer interactions strengthen the physical
interactions between neighboring chains in a manner similar
to covalent cross-linking, allowing 6F-PAI-2 to operate up to
almost 55 atm of pure CO2 with only minor increases in CO2
permeability. In general, the plasticization pressures of all three
materials represent substantial improvements from conven-
tional un-cross-linked materials such as cellulose acetate, and
even compared to cross-linked materials, they hold a potential
key advantage in the fact that this high plasticization resistance
is intrinsic to these materials, so an additional cross-linking step
is not required.
Despite improved stability against high pressure CO2, the
H2S plasticization resistance is lower in both 6F-PAI-2 and -3
relative to 6F-PAI-1. This is suggested by lower plasticization
pressure as well as a greater extent of permeability increase
beyond the initial permeability upswing. For 6F-PAI-2 and -3,
the plasticization pressure is around 5.8, whereas for 6F-PAI-1,
it occurs closer to 7 atm. The upswing in the concentration
isotherms in 6F-PAI-2 and -3 does not occur until
approximately 7 atm, while the permeability isotherms increase
near 5.8 atm. This means the initial upswing in permeability is
due to an increase in the H2S diffusion coefficient. Because CH4
is nonpolar, its sorption in this region is governed by its kinetic
diameter and the energy required for the densely packed
polymer chains to open up to sufficient size for accommodation
of a CH4 molecule. The fact that kD for CH4 was reduced from
0.04 (cm3(STP)/cm3·atm) in 6F-PAI-1 to 3 × 10−5
(cm3(STP)/cm3·atm) and 2 × 10−3 (cm3(STP)/cm3·atm) in
6F-PAI-2 and -3 respectively suggests that additional energy is
required to open up the chains in this region. Lower kD for CO2
is also visually evident in Figure 9. Although kD for H2S could
not be accurately determined in 6F-PAI-2 and -3, it appears that
the increased H2S sorption in 6F-PAI-2 and -3 relative to 6F-
PAI-1 is more pronounced in the Henry region, where
nonfluorinated DAM and TmPDA were shown to have the
greatest impact on gas sorption. Thus, this increase is due to
7046
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
increased thermodynamic affinity toward the polymer chains as
gas sorption transitions into a Henry's law dominated
mechanism. As a result of the higher sorption levels, chain
swelling in these materials occurs at lower H2S pressures,
causing increased permeability. Perhaps even more surprising is
the H2S plasticization pressure for Torlon, which is clearly
more stable against CO2 than 6F-PAI-1; however, as can be
seen in Figure 12, Torlon’s H2S plasticization pressure occurs
near 5.8 atm, which is slightly lower than 6F-PAI-1. Because
Torlon is nonfluorinated, H2S sorption may be even more
thermodynamically favorable. As such, Torlon is less stable
toward H2S despite its remarkable stability against CO2. These
results highlight the inherent difficulties in designing high
selectivity materials targeting simultaneous removal of CO2 and
H2S. As shown in Table 4, DH S/DCH is generally between 1
2
4
and 2. Smaller kinetic diameter differences between the gases
relative to CO2 and CH4 limit the size discrimination
capabilities for H2S relative to CH4 in pure polymer
membranes. On the other hand, differences in H2S and CH4
solubility, as inferred from critical temperatures, are higher for
H2S than CO2. It is therefore more advantageous to target
solubility selectivity for H2S/CH4. However, as the sorption
and permeability data in this work show, this can compromise
plasticization resistance even when CTC formation, which
generally stabilizes materials against plasticization, is enhanced
as it is in Torlon and 6F-PAI-2.
Figure 13. CO2/CH4 selectivity vs pressure at 35 °C. The feed gas for
6F-PAI-1 was composed of 10% H2S, 20% CO2, and 70% CH4. For
6F-PAI-2, the feed gas was composed of 20% H2S, 20% CO2, and 60%
CH4.
3.6. Mixed Gas Permeation Analysis. The previous
analysis on membrane stability toward CO2 and H2S was based
only on single gas feed streams. Although these results were
promising in that high ideal permselectivities and pure gas
plasticization resistances were shown, the true separation
performance of a material must be assessed using mixed gas
feeds. Factors such as competitive sorption and bulk flow
effects can reduce permselectivities relative to their correspond-
ing ideal, pure gas values.25,26 Also, increases in CO2 and H2S
partial pressure can cause localized increases in polymer chain
mobility as their concentration inside the membrane increases.
This can increase the diffusion coefficients of the larger, slower
gases while the fast gas is unaffected,27 thus by only using single
gas analysis, even the plasticization resistance can be over-
estimated. The mixed gas separation performance of 6F-PAI-1
was addressed in our previous paper. In this paper, we wish to
show the effects of the apparent greater sorption affinity of H2S
for 6F-PAI-2, which does not plasticize in mixtures containing
only CO2 and CH4, but does so in the presence of H2S at low
total feed pressures. Figures 13−15 compare 6F-PAI-2 using a
ternary feed stream containing CO2, H2S, and CH4. Higher
fluorinated 6F-PAI-1 is shown for comparison. It should be
noted that absolute values of the permselectivities cannot be
accurately compared due to different H2S feed concentrations
that were used for the two different materials. In this situation,
only a comparison of the plasticization resistance at similar H2S
partial pressures in the feed is meaningful.
Figure 14. H2S/CH4 permselectivity vs pressure isotherm at 35 °C.
The feed gas for 6F-PAI-1 was composed of 10% H2S, 20% CO2, and
70% CH4. For 6F-PAI-2, the feed gas was composed of 20% H2S, 20%
CO2, and 60% CH4.
As can be observed in Figures 13−15, H2S causes significant
performance reductions in lower fluorinated 6F-PAI-2. On the
other hand, more fluorinated 6F-PAI-1 does not show any
permeability increases up to 60 atm total feed pressure. 6F-PAI-
2 showed permeability upswings near 10 atm total feed
pressure, corresponding to approximately 1 atm of H2S partial
pressure. At 60 atm, H2S partial pressure is 6 atm. It is clear that
6F-PAI-1 does not plasticize even at higher H2S partial
pressures than 6F-PAI-2. This suggests that the material with
Figure 15. CO2 permeability vs pressure isotherm at 35 °C. The feed
gas for 6F-PAI-1 was composed of 10% H2S, 20% CO2, and 70% CH4.
For 6F-PAI-2, the feed gas was composed of 20% H2S, 20% CO2, and
60% CH4.
7047
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
Figure 16. CO2 and CH4 permeability vs pressure for 6F-PAI-2 at 35 °C. The feed gas was composed of 50% CO2 and 50% CH4.
higher fluorine content can support higher H2S concentrations
without plasticizing. The decrease in the H2S/CH4 selectivity in
6F-PAI-2 is not as high as the CO2/CH4 selectivity drop
because PH S also increases with feed pressure. Also, PCO for
2
2
6F-PAI-2 does not appear to increase until 40 atm total
pressure, while H2S permeability began increasing at much
lower feed pressure. This is believed to be due to the fact that
CO2 has a smaller kinetic diameter. Small scale increases in
local chain mobilities are believed to affect the diffusion
coefficient of the larger gases to a greater extent, and as such,
PCO does not appear to show any plasticization effects until 40
2
atm total feed pressure. This idea of different gases being
affected differently by changes in free volume is the basis of
Park and Paul’s modified method for calculating free volume
from group contributions.13 It also demonstrates the
importance of mixed gas permeation for the assessment of
plasticization resistance, as the permeability of the larger kinetic
diameter gases more accurately reveal swelling and plasti-
cization behavior with increases in acid gas concentration.
To test whether H2S was the cause of the low plasticization
resistance, 6F-PAI-2 was degassed for greater than 1 week
followed by subjecting it to a clean 50% CO2 50% CH4 feed
stream. Figures 16 and 17 show the PCO , PCH , and PCO /PCH
Figure 17. PCO /PCH vs pressure isotherm for 6F-PAI-2 at 35 °C. The
feed gas was composed of 50% CO2 and 50% CH4.
2
4
to H2S, as the separation performance was still quite promising
upon removal of the gas. Additionally, 6F-PAI-2 shows a CO2/
CH4 permselectivity near 40 above 55 atm of total feed
pressure. On the basis of these results, 6F-PAI-2 has strong
potential for CO2/CH4 separations, which is suggested by its
intrinsically high plasticization resistance and ability to maintain
good selectivities at high feed pressures.
2
4
2
4
isotherms up to 55 atm of total feed pressure. As can be
observed, the CO2 permeability is restored to its original, pure
gas value of ∼14 barrer. PCH is slightly higher than its pure gas
4
3.7. Conclusion. A new series of polyamide−imides based
on bulky 2,2-dicarboxyphenyl hexafluoropropane dianhydride
were synthesized and characterized toward high pressure CO2
and H2S for natural gas separations. 2,4-Diaminomesistylene
(DAM) and 2,3,5,6-tetramethyl phenylenediamine (TmPDA)
were used as the amide−amide bond diamine in attempts to
reduce chain packing in the solid state through steric
interactions. To our surprise, diffusivities and sorption
coefficients for both CO2 and CH4 decreased in the DAM-
and TmPDA-based polymers relative to structures containing
6FpDA (4,4′-(hexafluoroisopropylidene) dianiline). This de-
crease was not expected as 6FDA-based polyimides containing
DAM and TmPDA show significant permeability increases
relative to 6FpDA analogues. These results suggest that steric
interactions with ortho pendent groups around the amide bond
are not as strong as they are in imide bonds, and that
value of 0.29 barrer. This is due to swelling from the presence
of H2S in the original applied feed stream. Localized chain
dilations can become locked into the glassy chain conformation
as the membrane is degassed. This resulted in reduced
permselectivity at low pressure. With increasing total feed
pressures, however, PCH decreases, which causes the
4
permselectivity to increase with feed pressure. This increase is
believed to be the result of CO2 out competing CH4 for the
limited number of Langmuir sorption sites. These results
confirm that H2S has a significantly deleterious impact on the
separation performance of 6F-PAI-2. However, Figures 16 and
17 suggest that the permeabilities can be restored after removal
of H2S, and that large scale, glassy relaxations did not occur.
This suggests that 6F-PAI-2 may be able to tolerate disruptions
in process streams that may potentially expose the membrane
7048
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049

Macromolecules
Article
incorporating rigid, rod-like DAM and TmPDA diamines into
amide bonds results in lower free volume and polymer chain
mobility. Interestingly, the opposite trend was shown for H2S,
where H2S had higher sorption in DAM- and TmPDA-based
structures, which enhanced the H2S/CH4 ideal permselectivity.
These results suggest that the sorption of polar H2S is more
dependent on intermolecular interactions with the membrane
material than nonpolar CO2 and CH4.
REFERENCES
■
(1) Kosuri, M. R.; Koros, W. J. Ind. Eng. Chem. Res. 2009, 48, 10577−
10583.
(2) Kosuri, M. R.; Koros, W. J. J. Membr. Sci. 2008, 320, 65−72.
(3) Nagel, C.; Gunther-Schade, K.; Fritsch, D.; Strunskus, T.; Faupel,
̈
F. Macromolecules 2002, 35, 2071−2077.
(4) Fritsch, D.; Avella, N. Macromol. Chem. Phys. 1996, 197, 701−
714.
(5) Huang, S.; Hu, C.; Lee, K.; Liaw, D.; Lai, J. Eur. Polym. J. 2005,
42, 140−148.
The plasticization resistance of the polyamide−imides was
determined using both pure CO2 and H2S, as well as mixed gas
feed streams. The DAM-based analogue showed almost no
upswing through 55 atm of pure CO2. Furthermore, mixed gas
permeation up to 55 atm total feed pressure showed a CO2/
CH4 mixed gas permselectivity of ∼40, which is much higher
than typical un-cross-linked polymer materials at similar feed
condtions. On the other hand, pure gas H2S plasticization
resistance was actually reduced in DAM- and TmPDA-based
polyamide−imides, which is likely due to lower fluorine in
these materials. Mixed gas testing using a ternary feed stream
showed plasticization in 6F-PAI-2 at low total feed pressures,
whereas 6F-PAI-1 showed no permeability upswings up to 60
atm total feed pressure. Upon removal of the H2S containing
feed stream, permeabilities in 6F-PAI-2 showed only minor
hysteresis, and in fact the membrane maintained a CO2/CH4
binary feed stream selectivity of ∼40 at 55 a.m. total feed
pressure. On the basis of these results, it is unlikely that H2S
caused large scale, glassy relaxations. Rather, localized
disruptions to the polymer microstructure occurred. These
results suggest increased CTC formation, as was the case in 6F-
PAI-2, may not be sufficient in stabilizing polymers against H2S
plasticization and that intermolecular interactions should be
taken into account. In particular, fluorine content in the
polymer backbone may be used to reduce H2S affinity for the
polymer, which can stabilize the membrane toward elevated
(6) Vaughn, J. T.; Koros, W. J.; Johnson, J. R.; Karvan, O. J. Membr.
Sci. 2012, 401−402, 163−174.
(7) Merkel, T. C.; Bondar, V.; Nagai, K.; Freeman, B. D.
Macromolecules 1999, 32, 370−374.
(8) Merkel, T. C.; Bondar, V.; Nagai, K.; Freeman, B. D. Langmuir
2000, 273−296.
(9) Merkel, T. C.; Toy, L. G. Macromolecules 2006, 39, 7591−7600.
(10) Sridhar, S.; Smitha, B.; Aminabhavi, T. M. Sep. Purif. Rev. 2007,
36, 113−174.
(11) Yamazaki, N.; Matsumoto, M. J. Polym. Sci., Part A: Polym.
Chem. 1975, 13, 1373−1380.
(12) Fritsch, D.; Peinemann, K.-V. J. Membr. Sci. 1995, 99, 29−38.
(13) Park, J. Y.; Paul, D. R. J. Membr. Sci. 1997, 125, 23−39.
(14) Damle, S.; Koros, W. J. Ind. Eng. Chem. Res. 2003, 42, 6389−
6395.
(15) Paul, D. R.; Koros, W. J. J. Polym. Sci., Part B: Polym. Phys. 1976,
14, 675−685.
(16) Wind, J. D.; Sirard, S. M.; Paul, D. R.; Green, P. F.; Johnston, K.
P.; Koros, W. J. Macromolecules 2003, 36, 6442−6448.
(17) Kanehashi, S.; Nagai, K. J. Membr. Sci. 2005, 253, 117−138.
(18) Koros, W. J.; Paul, D. R. J. Polym. Sci., Part B: Polym. Phys. 1976,
14, 1903−1907.
(19) Zhou, F.; Koros, W. J. Polymer 2006, 47, 280−288.
(20) Baker, R. W.; Lokhandwala, K. Ind. Eng. Chem. Res. 2008, 47,
2109−2121.
(21) Wessling, M.; Strathmann, H.; Bos, A.; Pu, I. G. M. J. Membr.
Sci. 1999, 155, 67−78.
(22) Sanders, E. S.; Jordan, S. M.; Subramanian, R. J. Membr. Sci.
1992, 74, 29−36.
concentrations of this highly condensable gas. Finally, DH S
/
2
DCH was between 0.8 and 1.6 for all three polyamide−imides
4
(23) Wind, J. D.; Staudt-Bickel, C.; Paul, D. R.; Koros, W. J.
Macromolecules 2003, 36, 1882−1888.
as well as Torlon. Thus, overall H2S/CH4 permselectivity is
governed mostly by solubility, whereas diffusion selectivity
governs CO2/CH4 separations. This work highlights the
difficulty in designing materials that are highly selective for
both CO2 and H2S, a separation that would require materials
with high mobility as well as solubility selectivity. Introduction
of polar groups such as Cl or Br, which can increase the
cohesive energy density, is one potential strategy. This total
sour gas separation has not been actively targeted in the
literature but nevertheless remains a relevant challenge in
natural gas separations.
(24) Bos, a.; Punt, I. G. M.; Wessling, M.; Strathmann, H. Sep. Purif.
̈
Technol. 1998, 14, 27−39.
(25) Visser, T.; Koops, G.; Wessling, M. J. Membr. Sci. 2005, 252,
265−277.
(26) Kamaruddin, D. H.; Koros, W. J. J. Membr. Sci. 1997, 135, 147−
159.
(27) Chern, R. T.; Provan, C. N. Macromolecules 1991, 24, 2203−
2207.
(28) Coleman, M. R.; Koros, W. J. J. Polym. Sci., Part B: Polym. Phys.
1994, 32, 1915−1926.
(29) Kim, J. H.; Koros, W. J.; Paul, D. R. Polymer 2006, 47, 3094−
3103.
(30) Powell, C. E.; Duthie, X. J.; Kentish, S. E.; Qiao, G. G.; Stevens,
G. W. J. Membr. Sci. 2007, 291, 199−209.
(31) Liu, Y.; Wang, R.; Chung, T.-S. J. Membr. Sci. 2001, 189, 231−
239.
(32) Teoh, M. M.; Chung, T.-S.; Wang, K. Y.; Guiver, M. D. Sep.
Purif. Technol. 2008, 61, 404−413.
AUTHOR INFORMATION
Corresponding Author
*E-mail: wjk@chbe.gatech.edu. Telephone: (404) 751-6985.
Fax: (404) 385-2683.
Notes
■
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This publication is based on work supported by Award No.
KUS-I1-011-21, made by King Abdullah University of Science
and Technology (KAUST). The authors would also like to
thank Dr. JR Johnson and Dr. Oguz Karvan (Ga. Tech.) for
design and construction of the H2S systems as well as Megan
Lydon (Ga. Tech.) for assistance in the XRD measurements.
7049
dx.doi.org/10.1021/ma301249x | Macromolecules 2012, 45, 7036−7049