Você está na página 1de 12

Polymer 55 (2014) 1028e1039

Contents lists available at ScienceDirect

Polymer
journal homepage: www.elsevier.com/locate/polymer

Methanol diffusion in polyimides: A molecular descriptionq


Michele Galizia a, Pietro La Manna a, b, Marianna Pannico b, Giuseppe Mensitieri a,
Pellegrino Musto b, *
a
Department of Chemical, Materials and Industrial Production Engineering, University of Naples Federico II, p.le Tecchio 80, 80125 Naples, Italy
b
Institute of Chemistry and Technology of Polymers, National Research Council of Italy, Via Campi Flegrei, 34, 80078 Pozzuoli, Italy

a r t i c l e i n f o a b s t r a c t

Article history: In this contribution the diffusion of methanol in two commercial polyimides (6FDA-ODA and UltemÒ
Received 16 October 2013 1000) is investigated in detail by gravimetric analysis and in-situ, time-resolved FTIR spectroscopy. Two
Received in revised form methods of spectral data analysis are employed, namely difference spectroscopy (DS) and least-squares
20 December 2013
curve fitting (LSCF). These approaches provided complementary information about the overall diffusivity,
Accepted 6 January 2014
Available online 17 January 2014
the interactions and the dynamics of the molecular species present in the investigated systems. Spec-
troscopic measurements on thin film samples (2e4 mm) allowed us to identify the interaction site(s) on
the polymer backbone and to propose likely structures for the H-bonding molecular aggregates.
Keywords:
Polyimides
By coupling gravimetric and spectroscopic data collected in the same experimental conditions it was
Methanol possible to evaluate the molar absorptivities of the different molecular species, and hence their pop-
FTIR spectroscopy ulations, in both polyimides. The results highlighted the relevance of the H-bonding interaction and of
the substrate molecular structure on the final transport properties of the systems.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction fact, their good performances in dehydration of alcohols [5e12] are


exploited in pervaporation and vapour permeation processes that
Polyimides displaying different molecular architectures have provide a feasible alternative to conventional distillation for
attracted, in the last decades, an increasing scientific and techno- alcoholewater mixtures, warranting a lower energy demand [13].
logical interest, in view of their widespread applications ranging The analysis of methanol sorption and transport in polyimides is of
from microelectronics to the process industry. In the latter field, the interest not only with reference to water/methanol separation but
membrane separations of liquid and gas mixtures are some of the also for other processes involving the removal of methanol from
most relevant applications of these materials [1]. Many routes have azeotropic mixtures with several organic compounds in all those
been developed to optimize and tune mechanical and transport cases where methanol is used as reactant or reaction product (e.g.
properties of polyimides. Among them, the partial or total fluori- esterification, transesterification).
nation of the polymer backbone and the introduction of crosslinks The optimal design of separation processes involving alcoholic
appear to be of particular effectiveness [2,3]. Such chemically compounds, requires a molecular level understanding of the in-
modified polyimides display improved mechanical properties, high teractions between the penetrant and the polymer matrix. Indeed,
gas permeability and selectivity, and strong resistance to organic it has been demonstrated that both water and alcohol are able to
solvent, that makes it possible to use these materials in aggressive form hydrogen bonds with polymers displaying proton acceptor
environments or in extreme conditions of temperature and groups on their backbone, and that such an occurrence strongly
pressure. affects the transport properties and the separation performances of
The interest of methanol/polyimides systems stems from the these systems. The occurrence of self-association of the penetrant
fact that polyimides have been identified as membrane materials to form larger molecular aggregates has been also documented for
with high separation factors for water/alcohol separations [4]. In the water/polyimide system [14]. The vast majority of literature
contributions report only experimental data on alcohol dehydra-
tion using polyimides, while there is a lack of information about the
q Presented in part to the Conference “Frontiers in Polymer Science 2013”, Sitges,
molecular mechanisms of these processes, and the role played by
Spain, 21e23rd May 2013.
the different types of H-bonding interactions.
* Corresponding author. Tel.: þ39 081 8675202. Modelling of alcohol sorption thermodynamics in glassy poly-
E-mail addresses: pellegrino.musto@ictp.cnr.it, musto@ictp.cnr.it (P. Musto). mers endowed with hydrogen bonding interactions is rather

0032-3861/$ e see front matter Ó 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.polymer.2014.01.009
M. Galizia et al. / Polymer 55 (2014) 1028e1039 1029

complex since one needs to account both for the out-of-equilibrium Table 1
state of the glassy system and for the occurrence of specific in- Properties of the investigated polyimides (in parentheses are indicated estimated
standard deviations).
teractions. Approaches based on the classical adsorption theories of
BET (BrunauereEmmetteTeller) or GAB (Guggenheim, Anderson, M w (g/mol)a Mn (g/mol)a Tg ( C)b Density (g/cm3)c
de Boer) or their modifications are mainly phenomenological and 6FDA-ODA 9.7  104 5.1  104 302 (0.5) 1.488 (0.001)
lack a physically sound description of the molecular mechanism UltemÒ 1000 3.0  104 1.2  104 210 (0.5) 1.260 (0.001)
since do not supply a realistic microscopic picture of the interaction a
Measured by GPC.
mechanism between polar molecules and the internal structure of b
Measured by differential scanning calorimetry: heating rate 20  C/min.
c
the polymeric sorbent [15]. Another simple way to interpret sorp- Measured by a helium pycnometer.
tion of low m.w. penetrants within glassy polymers is represented
by the so-called Dual Sorption model [16,17] that, properly modi-
fying the original theory, can be used to deal with specific poly- quantify their populations. Finally, spectroscopic measurements on
merepenetrant interactions. This approach is suitable for thin samples (2e4 mm) enabled us to identify the interaction site(s)
correlation purposes but it is not predictive. More recently, many on the polymer backbone.
efforts have been devoted to the development of a theoretical
framework grounded on rational non-equilibrium thermody-
namics aimed at extending the equilibrium mixture theories, 2. Experimental
suitable for the description of low m.w. sorption thermodynamics
in rubbery polymers endowed with hydrogen bonding interactions 2.1. Materials
to the case of non-equilibrium glassy polymer/penetrant mixtures
[18e20]. In the present contribution the thermodynamic modelling The structure of the two investigated polyimides is reported in
issue is not addressed. Work is in progress to model the experi- Scheme 1. Poly(hexafluoroisopropilydene-2,2-bis-phthalicanhy-
mental spectroscopic and gravimetric results presented herein, by dride-oxydianiline) (6FDA-ODA) was synthesized according to the
applying an extension of a thermodynamic model for rubbery procedure described in Ref. [14].
polymers to nonequilibrium glassy polymers. To this aim, the in- Poly([2,20 -bis(3,4-dicarboxyphenoxy)phenylpropane]-2-
formation gathered from the spectroscopic analysis is of primary phenylene-bisimide), commercially available under the trade-
relevance. name of UltemÒ 1000, was kindly supplied by General Electric.
In the present context, attention has been focused to the infor- Table 1 summarizes the main characteristics of the two polyimides.
mation at the molecular level that can be gathered from FTIR Methanol used for sorption experiments was purchased from
spectroscopy. This technique has recently emerged as a powerful SigmaeAldrich (Milano, I) with purity higher 99.6%; it has been
tool to investigate mass transport phenomena in polymers. For further purified and degassed through freezingethawing cycles.
example, in Ref. [14], by combining gravimetric and spectroscopic Free standing films of 6FDA-ODA, 30 mm thick, were obtained by
approaches, an investigation of water transport in polyimides was spreading the polyamic acid solution [13 wt% in N-methyl-2-pyr-
carried out and a reliable method was devised to quantify the rolidone (NMP)] on a clean glass support with a Gardner knife. The
different populations of water molecules absorbed in the polymer nascent films were then dried 1 h at room temperature and 1 h at
matrix. 80  C, allowing the solvent to evaporate. The castings were then
In this work, a gravimetric and spectroscopic analysis of meth- thermally treated in a stepwise manner at 100, 150, 200, 250 and
anol transport in two commercial polyimides differing for their 290  C for 1 h at each temperature. Finally, the films were detached
molecular structure has been performed. The spectral data have from the glass support by immersion in distilled water at 80  C.
been analysed by complementary techniques, namely difference Free standing films of UltemÒ 1000 were prepared by casting of
spectroscopy (DS) and least squares curve fitting (LSCF), which a 5 wt% solution of the polyimide in chloroform. The nascent film
provided information on the nature and the number of penetrant was first dried at atmospheric conditions for 2 days and then in a
species present in the systems. Coupling gravimetric and spectro- vacuum oven at 120  C for 24 h, to remove residual solvent traces.
scopic data collected in the same conditions, it was possible to The thickness of the film used in gravimetric and FTIR experiments
evaluate the molar absorptivity of the different species and to was 43  1 mm.

Scheme 1. Molecular structure of A) 6FDA-ODA and B) UltemÒ 1000.


1030 M. Galizia et al. / Polymer 55 (2014) 1028e1039

For both polyimides, thinner films (2e4 mm) were prepared by a the 3650e2750 cm1 range, corresponding to the v(OH) and v(CH)
two-step spin-coating process, using a Chemat KW-4A apparatus modes of methanol. To separate the individual components in the
(Northridge, CA). Spinning conditions were 12 s at 700 rpm for the case of unresolved bands, a curve fitting algorithm was applied,
first step and 20 s at 1500 rpm for the second step. After removal of based on the LevenbergeMarquardt method [23,24]. The peak
the films from their glass support in distilled water at room tem- functions used throughout were a log-normal and a mixed Gausse
perature, they were treated in the same conditions as for the Lorentz line shape, which are expressed respectively in the form
thicker samples. [24]:
" "  # #
2.2. Gravimetric sorption experiments ðx  x0 Þ r2  1
ln 2 2
f ðxÞ ¼ H exp ln þ1 (1)
 C,
ðln rÞ2 w
Methanol sorption experiments were performed at 30 using
a Quartz Spring Balance equipped with a couple of digital CCD " 2 #
  x  x0 H
cameras. Experimental sorption isotherms were obtained by f ðxÞ ¼ 1  lf H exp 4 ln 2 þ lf  (2)
w 2
increasing stepwise the penetrant relative pressure p/p0, p0 being xx0
4 w þ1
the penetrant vapour pressure at the experimental temperature.
Before each sorption test, the polymer sample was dried overnight
under vacuum at the experimental temperature, up to constant where x0 is the peak position, H the peak height, w the full-width at
weight. Full details about the experimental procedure are given in half height (FWHH), lf the fraction of Lorentz character and r the
Ref. [21]. asymmetry index (half width ratio). In order to keep the number of
adjustable parameters to a minimum, the baseline, the number and
2.3. FTIR sorption experiments the shape of the components were fixed, allowing the curve-fitting
algorithm to optimize the FWHH and the position of the peaks.
Time-resolved spectra were collected in the transmission mode
during sorption of methanol in polyimides. To this aim, a vacuum- 3. Results and discussion
tight FTIR sorption cell was used, in which a free standing polymer
film is exposed to methanol vapour at constant temperature (30  C) 3.1. Gravimetric measurements
and different penetrant relative pressures. The sorption cell,
equipped with a recirculation water bath to keep the temperature Sorption isotherms of methanol in 6FDA-ODA and UltemÒ 1000,
constant to within 0.1  C, was connected through service lines to a as obtained from gravimetric experiments, have been reported in
methanol reservoir, a turbo-molecular vacuum pump and an MKS Fig. 1 as a function of the penetrant relative pressure.
Baratron 121 (Andover, MA) pressure transducer (full scale Since the penetrant molecules are able to establish attractive
100 Torr, resolution 0.01 Torr and accuracy 0.5% of the reading). interactions with the carbonyl group on the polymer backbone in
Full details of the experimental setup are reported elsewhere [22]. the whole range of relative pressures inspected, the sorption iso-
Before each sorption measurement, the polymer film was placed in therms display in the low-medium p/p0 range a downward con-
the sorption cell and dried overnight under vacuum at the exper- cavity that, in the light of the Dual Sorption model [25,26], can be
imental temperature. ascribed to the ‘adsorption’ on polymer excess free volume
The sorption cell was coupled with a Spectrum GX spectrometer microvoids and specifically interacting sites. The slight upturn
from PerkineElmer (Norwalk, CT), equipped with a Ge/KBr beam observed in 6FDA-ODA at p/p0 higher than 0.5 can be ascribed to
splitter and a wide-band DTGS detector. The transmission spectra the onset of penetrant clustering, whereby molecular aggregates
were collected by setting the following instrumental parameters: larger than the dimer start to occur (more on this, later).
resolution ¼ 4 cm1; Optical Path Difference (OPD)
velocity ¼ 0.5 cm/s; spectral range 4000e600 cm1. Spectra were
acquired in the single-beam mode using a dedicated software
package for time-resolved spectroscopy (Timebase, PerkineElmer).
Differential sorption tests were performed by increasing step-
wise the relative pressures of methanol vapour within the range 0e
0.6.
Finally, for both polyimides, sorption experiments were also
performed on a much thinner film, to identify the functional
group(s) in the polymer backbone directly involved in H-bonding
interaction with the penetrant.

2.4. FTIR data analysis

Full absorbance spectra (i.e. polyimide plus absorbed methanol)


were obtained using a background collected on the cell with no
sample inside, at the test conditions. The spectra representative of
absorbed methanol were obtained by using as a background in the
single-beam spectrum of the cell containing the dry polymer film. It
is explicitly noted that this data processing approach is equivalent
to the more general difference spectroscopy method, provided that
no significant change in sample thickness takes place during the
sorption measurement, which was experimentally verified in the
present case. The above procedure allows us to eliminate the Fig. 1. Gravimetric sorption isotherms of methanol in 6FDA-ODA (open symbols) and
interference of the polymer spectrum in the regions of interest, i.e. UltemÒ 1000 (filled symbols). The continuous lines are a guide for the eye.
M. Galizia et al. / Polymer 55 (2014) 1028e1039 1031

Fig. 2. Absorbance spectra (polyimide þ methanol) for A) 6FDA-ODA and B) UltemÒ 1000.

3.2. Absorbance and difference FTIR spectra with the low frequency bands getting more and more pronounced
as the penetrant concentration grows. This effect is more marked for
In Fig. 2AeB are reported the absorbance spectra of dry 6FDA- UltemÒ 1000 than for 6FDA-ODA. A quantitative assessment of this
ODA and UltemÒ 1000, respectively, before (black trace) and after behaviour will be presented in a forthcoming paragraph. Below
(blue in web version trace) equilibration in methanol vapour at the 3000 cm1 is located the n(CH) pattern of sorbed methanol, showing
highest relative pressure inspected. a three component profile centred at 2945 cm1, plus a fully
The spectrum representative of sorbed methanol can be iso- resolved peak at 2830 cm1. This profile remains completely un-
lated, at each relative pressure, after suppression of the polymer perturbed in the case of 6FDA-ODA (coincident with the reference
interference by means of the subtraction analysis (see state, vide infra), while showing a typical first-derivative feature in
Experimental section). As reported in Fig. 3AeB, the difference UltemÒ 1000 at 2970 cm1. This effect is indicative of a (limited)
spectrum of methanol shows a quite complex band shape in the vOH perturbation in the v(CH) range of the UltemÒ 1000 spectrum.
region, indicating the occurrence of different species of methanol
involved in H-bonding interactions with the polymer matrix. 3.3. Sorption kinetics
In particular, in both polyimides a relatively sharp, partially
resolved component at around 3570 cm1 is superimposed onto a Sorption kinetics of methanol in 6FDA-ODA and UltemÒ 1000
much broader band approximately centred at 3440 cm1. The latter has been investigated by using gravimetric and spectroscopic ap-
feature is significantly more pronounced in 6FDA-ODA than in proaches. In view of the linear response of the spectroscopic ob-
UltemÒ 1000. It is anticipated that the intensity ratio between the servables with respect to concentration values (LamberteBeer
two components changes with the amount of sorbed methanol, behaviour, see Fig. 4AeB), the time evolution of the v(eCH) band

Fig. 3. Difference spectra collected at increasing relative pressures during sorption experiments of methanol in polyimides: A) 6FDA-ODA, and B) UltemÒ 1000.

Fig. 4. Absorbance of the v(OH), v(CH) and v(CeO) bands as a function of methanol concentration: A) 6FDA-ODA and B) UltemÒ 1000.
1032 M. Galizia et al. / Polymer 55 (2014) 1028e1039

intensities can be exploited to monitor the diffusion kinetics, which Table 2


have been analysed by reporting, in the Fick’s diagram, the Methanol diffusivity in 6FDA-ODA and UltemÒ 1000 as estimated from gravimetric
and FTIR experiments (in parentheses are indicated estimated standard deviations).
normalized absorbance A(t)/Ainf as a function of the ratio t0.5/L,
where t represents the sorption time and L the film thickness. Here Dgrav (cm2/s) DFTIR (cm2/s)
and in the following, A(t) is the integrated absorbance at time t and 6FDA-ODa 6.0  1010 (2  1011) 7.0  1010 (1  1011)
Ainf the final integrated absorbance, after reaching equilibrium UltemÒ 1000b 2.0  1010(2  1011) 2.5  1010(1  1011)
conditions. a
Measured at p/p0 ¼ 0.20.
The higher diffusivity of methanol in 6FDA-ODA as compared to b
Measured at p/p0 ¼ 0.04.
PEI is likely due the higher Fractional Free Volume (FFV) of the
former. In fact FFV of PEI has been reported to be equal to 0.112 as
opposed to the value of 0.165 for 6FDA-ODA [27]. This effect is their thickness has been estimated spectroscopically, according to
related mostly to the presence of CF3 groups in 6FDA-ODA that the relationship:
significantly increase the stiffness of the chain and reduce the
effective chain packing [28]. Athin
Lthin ¼ L (5)
The same analysis can be performed by considering the gravi- Athick thick
metric data and reporting, in the Fick’s diagram, m(t)/minf as a
function of t0.5/L. Fig. 5AeB demonstrates that the gravimetric and The analytical peak of the polymer spectrum (located at
spectroscopic data are essentially coincident and can be satisfac- 1015 cm1) was chosen so as to be fully resolved and of intensity
torily fitted by the Fick’s model, which is described, in the case of a lower than the saturation threshold in both thicker and thinner
plane sheet, by the following equations in terms of mass absorbed specimens. The thickness of the thin films resulted to be 2 mm for
(eq. (3)) or, equivalently, in terms of absorbance (eq. (4)) [29,30] 6FDA-ODA and 4 mm for UltemÒ 1000. The sorption test was per-
" # formed at four different relative pressures; in this case only equi-
mðtÞ 8 X
N
1 Dð2m þ 1Þ2 p2 t librium data were meaningful, owing to the very fast sorption
¼ 1 2 exp (3)
minf p m ¼ 0 ð2m þ 1Þ2 L2 kinetics.
In Fig. 6AeC the absorbance spectrum of dry 6FDA-ODA is
" # compared with the spectrum of the same polyimide after equili-
AðtÞ 8 X
N
1 Dð2m þ 1Þ2 p2 t bration in methanol vapour at p/p0 ¼ 0.6 in three different fre-
¼ 1 2 exp (4)
Ainf p m ¼ 0 ð2m þ 1Þ2 L2 quency ranges. The difference spectra (equilibrated  dry) are also
reported. In Fig. 6DeF is shown the same comparison for UltemÒ
In eqs. 3 and 4, D represents the mutual diffusivity and is the 1000 equilibrated at p/p0 ¼ 0.5.
sole parameter used to fit the experimental points. The Fickian The chosen frequency ranges are those where the most likely
behaviour of the systems at hand is confirmed by the linearity of proton-acceptor groups of the two polyimides display their major
the sorption curves as a function of t0.5/L, up to A(t)/Ainf and m(t)/ features. For 6FDA-ODA, we observe a well resolved carbonyl
minf values of 0.6. doublet at 1785e1727 cm1 [vsym(C]O) and vasym(C]O), respec-
The diffusivity values estimated from gravimetric and FTIR data tively] and a band at 1378 cm1 which, according to previous
are in excellent agreement one to another for both the systems literature reports [32] and to a recent normal coordinate analysis
investigated and compare very well with literature data [31] (see [33], has been assigned to a highly coupled normal mode
Table 2). comprising a significant contribution from the NeCeO in-plane
deformation. Finally, the prominent band at 1241 cm1 originates
3.4. Detecting the polymer active site(s) from the ether linkage of the ODA unit (the asymmetric stretching
vibration of the CeOeC bond [32]). Both carbonyl peaks display a
With the aim of identifying the interaction site(s) on the poly- red-shift which is clearly detected in the difference spectrum as a
mer backbone, a methanol sorption test has been performed on pronounced first-derivative feature with the negative lobe pre-
much thinner films of 6FDA-ODA and UltemÒ 1000. The samples ceding the positive (see Fig. 6A). Also for the 1378 cm1 band a first-
used for this kind of analysis have been obtained by a dual step derivative pattern is apparent in the difference spectrum, but this
spin-coating process, as described in the Experimental section: time the positive lobe occurs at higher frequencies than the

Fig. 5. Methanol sorption kinetics in A) 6FDA ODA and B) UltemÒ 1000. Red symbols refer to gravimetric data, blue symbols are relative to FTIR data. Continuous lines represent the
best fitting provided by 2nd Fick’s law (eqs. (3) and (4)). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
M. Galizia et al. / Polymer 55 (2014) 1028e1039 1033

Fig. 6. Spectra of the dry polyimide (black trace), of the polyimide after equilibration in methanol vapour at the higher p/p0 value (blue trace) and difference spectra (equilibrated
sample  dry sample, red traces). Spectra are reported in three different frequency regions. AeC): 6FDA-ODA; DeF) UltemÒ 1000. (For interpretation of the references to colour in
this figure legend, the reader is referred to the web version of this article.)

negative, i.e. the band is blue-shifted (see Fig. 6B). These observa- In summary, the above analysis demonstrates that, in the pre-
tions can be accounted for by assuming the involvement of the sent system the only proton acceptors are the imide carbonyls,
carbonyl groups in H-bonding interactions with methanol. In fact, while the ether oxygens are not appreciably affected.
the red shift is a direct consequence of the lowering of the C]O In view of the spectral features shown in Fig. 6DeF, exactly the
force constant caused by the interaction between the proton same conclusions can be drawn for UltemÒ 1000.
acceptor and the OeH groups. Conversely, it is known that H- The reason why the sp3 oxygen does not participate to the
bonding determines a stiffening of force constant of in-plane overall interaction equilibrium is not easily accounted for. Possible
deformation modes of the proton acceptor [33], which according steric effects can be invoked, originating from the well established
to our previous assignments, accounts for the blue-shift of the propensity of the polyimide chains to form charge-transfer in-
1378 cm1 band. The v(CeOeC) band is not significantly perturbed teractions among the aromatic rings of the backbone. These induce
in the presence of methanol, which rules out the involvement of the formation of mesophases in which the chains assume prefer-
ether linkages in H-bonding with the penetrant. ential conformations such that the carbonyls may be more
1034 M. Galizia et al. / Polymer 55 (2014) 1028e1039

Fig. 7. A): The absorbance spectra in the v(C]O) frequency range (1820e1640 cm1) collected on 6FDA-ODA (thin film) equilibrated at different vapour pressures of methanol; B):
Difference spectra (equilibrated sample  dry sample). The colour code is: black trace: p/p0 ¼ 0 (dry sample); blue trace: p/p0 ¼ 0.06; red trace: p/p0 ¼ 0.2; green trace: p/p0 ¼ 0.4;
cyan trace: p/p0 ¼ 0.6. In B) the black trace corresponds to the difference between the fully desorbed sample (eq. at p/p0 ¼ 0.6) and the initial, dry sample (reversibility test). (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

accessible than the ether groups. Similar effect has been already develop at p/p0 ¼ 0.20 and keeps increasing therefore. The spectral
reported for the system polyimide/water [14]. parameters of the first component are therefore experimentally
Figs. 7 and 8AeB display the carbonyl range of 6FDA-ODA and available: it is explicitly noted that the 3578 cm1 peak displays an
UltemÒ 1000 in the films equilibrated at different relative pressures intrinsic band asymmetry (that is, due to the vibrational relaxation
of methanol vapour. The extent of the red-shift, which is reflected mechanism rather then to an unresolved component) which has
in the peak-to-peak height of the difference spectra, is found to been satisfactorily simulated by use of a log-normal band-shape
increase gradually with increasing methanol concentration and is (see Fig. 9B). This model has been demonstrated to be well suited
fully reversible upon methanol removal, which further confirms for vibrational relaxation processes affected by multiple interfering
that the observed effects originate from the established in- mechanisms [34].
teractions, and are not due to spectroscopic artefacts. For the band at around 3440 cm1 the best fit over the whole
experimental profiles was achieved with a mixed LorentzeGauss
3.5. Quantitative assessment of the different methanol species line shape [23]. The results of the LSCF analysis are summarized in
Table 3. It explicitly noted that the n(OH) profile invariably displays
The quantitative analysis is based on the curve fitting of the a feature at around 3480 cm1 which resembles a partly resolved
spectral profiles in the v(OH) region; typical results are demon- component (see Figs. 3 and 9). We assign this feature to a
strated in Fig. 9A, relative to 6FDA-ODA, and in Fig. 9BeC relative to derivative-type profile appearing in the difference spectrum as a
UltemÒ 1000. For both polyimides two components are sufficient to consequence of the downward shift of the 3484 cm1 peak of the
satisfactorily reproduce the experimental profiles. However, the polyimides [a combination mode of the stretching fundamentals of
band shapes of the two components are significantly different. the imide carbonyls (niph þ nooph)]. The effect is similar to that
Luckily, for UltemÒ 1000 at the lowest activity only the sharp peak represented in Figs. 7 and 8 for the n(C]O) vibration but amplified
at 3578 cm1 is present: the broad band at lower frequency starts to twice (roughly) due to the non-fundamental nature of the vibration

Fig. 8. A) The absorbance spectra in the v(C]O) frequency range (1820e1640 cm1) collected on UltemÒ 1000 (thin film) equilibrated at different vapour pressures of methanol; B):
Difference spectra (equilibrated sample  dry sample). The colour code is: black trace: p/p0 ¼ 0 (dry sample); blue trace: p/p0 ¼ 0.05; red trace: p/p0 ¼ 0.2; green trace: p/p0 ¼ 0.3;
cyan trace: p/p0 ¼ 0.5. In B) the black trace corresponds to the difference between the fully desorbed sample (eq. at p/p0 ¼ 0.5) and the dry sample (reversibility test). (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
M. Galizia et al. / Polymer 55 (2014) 1028e1039 1035

Fig. 9. LSCF analysis of the vOH region (3800e3100 cm1): A) 6FDA-ODA/methanol system, p/p0 ¼ 0.07; B) UltemÒ 1000/methanol system, p/p0 ¼ 0.04; C) UltemÒ 1000/methanol
system, p/p0 ¼ 0.5.

involved. According to this interpretation, the feature was consid- first-shell species (second and higher layers in the BET nomencla-
ered as an artefact and neglected in the subsequent curve fitting ture). A schematic representation of the spectroscopically detected
analysis (see Fig. 9). species is reported in Scheme 2.
As demonstrated by Fig. 9AeC and by the data presented in Work is in progress to optimize the geometry of the above
Table 3, the simulation of the experimental data by the selected molecular aggregates by first principles computational approaches
curve-fitting model is satisfactory and consistent. so as to calculate theoretically the vibrational spectra and compare
Taking into account the whole of the spectroscopic results, we these simulations with the experimental results.
can assign the 3575 cm1 component to the OeH groups of It is useful to compare the spectrum representative of methanol
methanol directly bound to the imide carbonyls, which represent sorbed in the polyimides with that of the penetrant in the un-
the first shell layer of sorbed methanol in the frame of the multi- perturbed, reference state, which is represented by a dilute solu-
layer adsorption model of BrunauereEmmetteTeller (BET) [35]. tion in a low polarity, non-interacting solvent (CCl4). In Fig. 10 are
The band at 3440 cm1 is representative of self-associated meth- reported the spectra of three solutions of methanol in CCl4,
anol molecules, whereby the H-bond is established between the differing for their concentrations. At the highest dilution (0.012 M)
OeH group of this molecular species and the oxygen of the only the monomer is present, giving rise to a very sharp peak at
3643 cm1. Increasing the concentration, dimers start to appear,
producing a well resolved and symmetrical band at 3525 cm1.
Table 3
Results of the LSCF analysis. (Standard errors in parentheses). Note the considerable increase in FWHW when passing from the
monomer to higher aggregates, which is a direct consequence of
6FDA-ODA
the H-bonding interaction. Finally, at 0.1 M the equilibrium in-
p/p0 ¼ 0.07 p/p0 ¼ 0.20 p/p0 ¼ 0.40 p/p0 ¼ 0.60 volves also trimers and tetramers [36], which produce a further
Peak 1 Position (cm1) 3579 (0.1) 3577 (0.1) 3575 (0.2) 3573 (0.1) broad band centered at 3350 cm1. The spectrum of methanol
FWHH (cm1) 73.6 (0.2) 79.8 (0.3) 86.0 (0.2) 88.1 (0.1) sorbed in 6FDA-ODA displays a fully resolved component which,
Area (cm1) 9.4 (0.1) 16.3 (0.1) 23.9 (0.1) 28.4 (0.1) by comparison with the reference state, can be safely associated
r 0.64 (0.001) 0.66 (0.002) 0.62 (0.001) 0.60 (0.001)
Peak 2 Position (cm1) 3465 (0.1) 3445 (0.2) 3430 (0.1) 3423 (0.1)
with a monomeric species. However, the larger value of FWHH
FWHH (cm ) 1
198.8 (0.2) 231.5 (0.3) 236.7 (0.2) 249.5 (0.3) (from 22 to 90 cm1) and, especially, the considerable red-shift
Area (cm1) 13.1 (0.1) 30.8 (0.1) 52.9 (0.1) 82.1 (0.1) (68 cm1) demonstrates that in the polymer, the monomeric
lf 0 0 0 0 species is not free as in the CCl4 solution, but is involved in an H-
R2 0.993 0.996 0.982 0.990
bonding interaction of non-self-association type. Self-association
is connected with the broad band at 3440 cm1. Its rather sym-
UltemÒ 1000
metrical shape suggests the occurrence of a single molecular
p/p0 ¼ 0.04 p/p0 ¼ 0.20 p/p0 ¼ 0.35 p/p0 ¼ 0.5
aggregate (dimers) or, at least, a large predominance of this species
1
Peak 1 Position (cm ) 3577 (0.2) 3575 (0.2) 3575 (0.1) 3575 (0.2) over higher order aggregates. The available spectroscopic evidence
FWHH (cm1) 94.7 (0.5) 95.4 (0.3) 97.0 (0.4) 97.3 (0.2) allows us to rule out the occurrence “free” monomers or dimers,
Area (cm1) 6.8 (0.1) 18.9 (0.1) 23.5 (0.2) 25.6 (0.1)
i.e. of penetrant molecules or aggregates thereof not directly
r 0.60 (0.01) 0.64 (0.02) 0.64 (0.02) 0.63 (0.01)
Peak 2 Position (cm1) 3442 (0.5) 3441 (0.6) 3441 (0.5) anchored to the polymer backbone. In fact the highly characteristic
FWHH (cm1) 136.2 (0.9) 168.5 (0.8) 195.7 (1.0) signal of “free” OeH bonds, which would occur in the case of a
Area (cm1) 4.8 (0.1) 10.4 (0.2) 17.1 (0.1) non-interacting monomer and/or an open dimer (the energetically
lf 0.17 (0.01) 0.11 (0.01) 0.30 (0.01)
favoured conformation on the basis of ab-initio calculations [36]) is
R2 0.994 0.996 0.993 0.990
absent.
1036 M. Galizia et al. / Polymer 55 (2014) 1028e1039

Scheme 2. Schematic representation of the methanol species identified by the spectroscopic analysis.

The first step towards a quantitative assessment of the con- species, respectively (see Scheme 2). Substituting the Lamberte
centration of the different methanol species is the evaluation of Beer law in eq. (6), leads to:
their molar absorptivities. To this end, the mass balance for the total
absorbed methanol has been considered: Afs Ass
Ctot ¼ þ (7)
εfs L εss L
Ctot ¼ Cfs þ Css (6)
where Ai and εi are the integrated absorbance and the molar ab-
where Ctot is the total concentration as evaluated gravimetrically, sorptivity of the i-th species, while L represents the sample thick-
while Cfs and Css are the concentration of first shell and second shell ness. After some algebra, eq. (7) becomes:

Fig. 10. A): Spectra of methanol in CCl4 in the frequency interval 4000e2450 cm1. Concentrations as indicated. B) Comparison between the spectrum of a 0.10 M methanol solution
in CCl4 (upper trace) and the spectrum of methanol sorbed in 6FDA-ODA (bottom trace, p/p0 ¼ 0.07).

Fig. 11. Estimation of monomer and dimer molar absorptivity: A) 6FDA-ODA and B) UltemÒ 1000. Continuous line represents the best fit provided by eq. (8).
M. Galizia et al. / Polymer 55 (2014) 1028e1039 1037

 
Afs ε Ass Cfs, which may only occur when aggregates comprising more than
¼ εfs L  fs (8) two methanol molecules are formed. The plot of Fig. 12A allows us
Ctot εss Ctot
to clearly identify the onset of the clustering phenomenon as the
Thus, a linear correlation is expected between Afs/Ctot and Ass/ point of departure between the Css and Cfs curves.
Ctot, whose slope directly provides the εfs/εss ratio, while the For UltemÒ 1000 Cfs largely exceeds Css in the whole p/p0 range:
intercept gives an experimental estimation of the product εfsL. In in this case the predominant species is the monomer, and the self-
Fig. 11AeB, the experimental correlations between Afs/Ctot and Ass/ associated aggregate is likely to be the dimer.
Ctot are reported for the systems 6FDA-ODA/methanol and UltemÒ An approach has been proposed to quantify the population of
1000/methanol, respectively. Accordingly to eq. (8), in both cases a proton acceptor groups in polymer/penetrant systems where H-
very good linear trend is observed (R2 w 0.92); the molar absorp- bonding interactions among the penetrant and the substrate are
tivity evaluated therefore is εfs ¼ 88 km/mol and εss ¼ 183.3 km/mol formed. The method relies on the shift of the imide carbonyls,
for the 6FDA-ODA/methanol system; εfs ¼ 75.3 km/mol and brought about by the H-bonding interaction, which has been
εss ¼ 108 km/mol for UltemÒ 1000/methanol. The closeness of the described in detail in Refs. [33,14] and will be summarized here-
εfs and εss values for the two polyimides provides a clue to the after. The observed carbonyl red-shift is actually due to the pres-
consistency of our analysis. A value of εss higher that of εfs by a factor ence of two unresolved components originating, respectively, from
of about two appears to be consistent, as an order of magnitude, the interacting groups (at lower frequency) and the unperturbed
with previous results on molar absorptivity increase in similar H- ones (at higher frequency). The direct resolution of the components
bonding systems [14,37]. is not achieved because their Dn is comparable to their FWHH. The
As already mentioned, for UltemÒ 1000 the spectrum of sorbed unperturbed carbonyl peak is experimentally available (the peak of
methanol at the lowest investigated value of p/p0 displays only the the fully dry polyimide) and can be used as a reference for spectral
3575 cm1 component (see Fig. 9B), thus indicating that, in these subtraction. Thus, the interacting carbonyl component, Adiff, can be
experimental conditions, only the monomer is present in the sys- isolated as:
tem. Accordingly, in this case εfs ¼ Afs/LCtot the value calculated in
this way is 77.0 km/mol, in excellent agreement with the estimation Adiff ¼ As  K,Ar (9)
provided by eq. (8), which confirms the reliability of the more
general quantitative approach. where As and Ar refer, respectively, to the absorbance of the imide
A comparison of our εfs values with that of monomeric methanol peak in the sample spectrum (methanol saturated film) and in the
in CCl4 [36] indicates a fourfold increase of this parameter for reference spectrum (dry film). Owing to the invariance of the film
methanol sorbed in the polyimides with respect to the reference thickness upon methanol sorption, the subtraction factor K is
state (76 and 88 km/mol vs 21 km/mol). Once again, the compari- quantitatively related to the relative concentration of proton
son indicates that the present estimates are reasonable in terms of acceptor groups. The physical meaning of K, derived from the
orders of magnitude; the higher values we observe can be readily relevant absorbanceeconcentration relationships [33], is:
accounted for considering the well known increase of the absorp-
tivity value brought about by H-bonding formation. Cf C
With the absorptivity values at hand, the concentration of
K ¼ and b ¼ 1  K (10)
Ctot Ctot
methanol species is readily obtained from the relevant absor-
banceeconcentration relationships, i.e. Cfs ¼ A3575/εfsL and where the subscripts f, b and tot refer, respectively, to the non-
Css ¼ A3440/εssL. interacting (free) imide groups, the imide groups H-bonded to
In Fig. 12AeB are reported the concentrations of first-shell and methanol molecules, and to their total population. The criterion to
second-shell methanol in 6FDA-ODA and UltemÒ 1000, respectively, correctly choose the subtraction factor K is described in Ref. [33].
as a function of methanol relative pressure. In the light of the interaction model proposed in Scheme 2, the
In the case of 6FDA-ODA the two above concentrations can be concentration of carbonyl proton-acceptors is equal to the con-
considered coincident up to p/p0 values of 0.45, which infers that in centration of first-shell methanol. It is explicitly noted that in our
this relative-pressure range all absorbed methanol is present in the view, in terms of stoichiometry, the proton acceptor group is not
form of dimers. At higher p/p0 values (>0.5) Css significantly offsets represented by the single carbonyl, but rather by the whole imide

Fig. 12. Concentration of first-shell (red circles) and second-shell (blue circles) methanol in A) 6FDA-ODA and B) UltemÒ 1000. Continuous and dotted lines are a guide for the eye.
(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
1038 M. Galizia et al. / Polymer 55 (2014) 1028e1039

Fig. 13. Concentration of interacting carbonyls (red circles) and concentration of first-shell methanol (black triangles) as evaluated in the thin sample; concentration of first-shell
methanol (blue squares) as evaluated in the thick sample: A) 6FDA-ODA and B) UltemÒ 1000. The insets represent the analysis of C]O range in terms of difference spectroscopy,
with the main peak (black trace) representing the reference (dry sample). (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)

group. This because the formation of two H-bonds on the same by coupling gravimetric and in-situ FTIR analysis in the trans-
imide ring is unlikely both statistically and in terms of reactivity. mission mode. This approach allowed a detailed characterization of
Statistically, because of the large excess of non-interacting car- the investigated systems at a molecular level.
bonyls with respect to those involved in H-bonding; in terms of The analysis of gravimetric and spectroscopic kinetic data pro-
reactivity, because in the imide ring the electronic density is vided two independent estimations of the diffusion coefficients
delocalized over the whole condensed system. A withdrawing of which were in good agreement with each other.
electron density from one of the carbonyl by the OeH proton is also The spectral data were analysed by using two complementary
experienced by the second carbonyl in the imide ring, which, as a techniques, i.e. difference spectroscopy and least squared curve-
consequence, becomes an even weaker proton acceptor. From the fitting analysis: both methods indicated the occurrence of two
consideration enumerated so far, we can compare the concentra- distinct populations of penetrant, namely single methanol mole-
tion of imide acceptors as evaluated from the analysis of the cules directly interacting with the carbonyls of the imide group
carbonyl region, with the concentration of first shell methanol as (first-shell species), and self-associated methanol molecules form-
estimated in the n(OH) range. The two estimates are simultaneously ing second and higher shell layers. For 6FDA-ODA the dimer rep-
available for the thin film, where the n(OH) band, albeit weak, is still resents the largely predominant species up to p/p0 values of 0.5.
suitable for curve-fitting analysis. In the case of the thick sample Afterwards, higher aggregates (higher shell layers) are formed; for
only the Cfs value can be estimated since the carbonyl region is fully UltemÒ 1000 the isolated methanol molecules bound to the imide
saturated (see spectrum in Fig. 2). In Fig. 13-A is reported, for 6FDA- carbonyls represent the predominant species over the whole p/p0
ODA, the comparison between the Cfs and the Cb curves as a func- range investigated. Two independent methods to quantify the
tion of p/p0. The figure inset displays the analysis of the carbonyl population of the methanol species have been proposed and
region in terms of the difference spectroscopy method discussed compared. The first relies on the evaluation of the respective molar
earlier. It is found that the data compare very satisfactorily: the Cb absorptivities, the second employs the difference spectroscopy
curve is essentially coincident with the Cfs curve relative to the technique in the carbonyl range. The results obtained with the two
same sample (thin film); also the Cfs curve for the thick sample is approaches compare very satisfactorily to each other, thus con-
very close (within experimental uncertainty up to p/p0 ¼ 0.4); the firming the robustness of the quantitative analysis and the reli-
slight difference found at higher p/p0 values could be related to an ability of the proposed molecular models.
influence of the film thickness on the sorption behaviour. Similar
effects have been documented in the literature on polyimide/water Acknowledgements
and polyimideeoxygen systems [26,33]. The results discussed so
far confirm the robustness of the analytical methodology and verify Thanks are due to Mr. G. Orefice for his assistance in performing
the reliability of the proposed molecular models. the time-resolved FTIR experiments. Financial support from the
For UltemÒ 1000 similar conclusions can be drawn (see Fig. 13- National Research Council of Italy (CNR) in the frame of the Project
B): in this case the CbeCfs curves for the thin sample are slightly “Ricerca a Tema Libero” is acknowledged.
more separated. This is a consequence of the lower intensity of the
n(OH) band for this series of samples, due to the lower methanol
solubility compared to 6FDA-ODA. This makes the Cfs estimation on References
the thin sample by curve-fitting analysis more uncertain. The Cbe
[1] Robeson LM. Polymer membranes for gas separation. Curr Opin Solid State
Cfs curves relative to the films of different thickness are even closer Mater Sci 1999;4(6):549e52.
than in the previous case. [2] Tanaka K, Kita H, Okano M, Okamoto K. Permeability and permselectivity of
gases in fluorinated and non-fluorinated polyimides. Polymer 1992;33(3):
585e92.
4. Conclusions [3] Wind JD, Staudt-Bickel C, Paul DR, Koros WJ. The effects of crosslinking
chemistry on CO2 plasticization of polyimide gas separation membranes. Ind
Eng Chem Res 2002;41(24):6139e48.
In this work the transient and equilibrium sorption of methanol [4] Qiu Wulin, Kosuri Madhava, Zhou Fangbin, Koros William J. J Membr Sci
in 6FDA-ODA and UltemÒ 1000 polyimides has been investigated 2009;327:96e103.
M. Galizia et al. / Polymer 55 (2014) 1028e1039 1039

[5] Huang RYM. Pervaporation membrane separation processes. Amsterdam: from supercritical carbon dioxide solutions. Ind Eng Chem Res 2005;44:
Elsevier; 1990. 1795e803.
[6] Semenova IV SI. Separation properties of polyimides. In: Ohya H, [22] Cotugno S, Larobina D, Mensitieri G, Musto P, Ragosta G. A novel spectroscopic
Kudryavsev VV, Semenova SI, editors. Polyimide membranes e applications, approach to investigate transport processes in polymers: the case of water-
fabrications, and properties. Kodansha; 1996. pp. 103e78. epoxy system. Polymer 2001;42:6431e8.
[7] Feng X, Huang RYM. Pervaporation and performance of asymmetric poly- [23] Marquardt DW. Finite difference algorithm for curve fitting. J Soc Ind Appl
etherimide membranes for isopropanol dehydration by pervaporation. Math 1963;11:431e41.
J Membr Sci 1996;109:165e72. [24] Meier RJ. On art and science in curve fitting vibrational spectra. J Vib Spectrosc
[8] Yanagishita H, Maejima C, Kitamoto D, Nakane T. Preparation of asymmetric 2005;39:266e9.
polyimide membrane for water/ethanol separation in pervaporation by the [25] Mensitieri G, Del Nobile MA, Apicella A, Nicolais L. Moisture-matrix in-
phase inversion process. J Membr Sci 1994;86:231e40. teractions in polymer based composite materials. Rev l’Institut Francais Pet
[9] Huang RYM, Xianshe Feng. Dehydration of isopropanol by pervaporation us- 1995;50(4):551e71.
ing aromatic polyetherimide membranes. Sep Sci Technol 1993;28:2035e48. [26] Mensitieri G, Del Nobile MA, Monetta T, Nicodemo L, Bellucci F. The effect of
[10] Yanagishita H, Kitamoto D, Nakane T. Separation of alcohol aqueous solution film thickness on oxygen sorption and transport in dry and water-saturated
by pervaporation using asymmetric polyimide membrane. High Perform KaptonÒ polyimide. J Membr Sci 1994;89:131e41.
Polym 1995;7:275e81. [27] Fried JR. Materials science of membranes for gas and vapor separation. In:
[11] Okamoto K, Tanihara N, Watanabe H, Tanaka K, Kita H, Nakamura A, et al. Yampolskii Y, Pinnau I, Freeman BD, editors. Chap. 3. Molecular simulation of
Vapor permeation and pervaporation separation of water-ethanol through gas and vapor transport in highly permeable polymers. The Atrium, Southern
polyimide membranes. J Membr Sci 1992;68(1e2):53e63. Gate, Chichester, England: John Wiley & Sons Ltd; 2006.
[12] Chen WJ, Martin CR. Highly methanol selective membranes for the perva- [28] Powell CE, Qiao GG. Polymeric CO2/N2 gas separation membranes for the
poration separation of methyl-t-butyl ether/methanol mixtures. J Membr Sci capture of carbon dioxide from power plant flue gases. J Membr Sci 2006;279:
1995;104(1e2):101e8. 1e49.
[13] Nakagawa K, Asakura Y, Nakanishi S, Hoshino H, Kouda H, Kusuki Y. Sepa- [29] Crank J. The mathematics of diffusion. 2nd ed. Oxford: Oxford University
ration of vapor mixtures of water and alcohol by aromatic polyimide hollow Press; 1975.
fibers. Kobunshi Ronbunshu 1989;46:405e11. [30] Musto P, Galizia M, Scherillo G, Mensitieri G. Water sorption thermodynamics
[14] Musto P, Mensitieri G, Lavorgna M, Scarinzi G, Scherillo G. Combining gravi- and mass transport in poly(ε-caprolactone): interactional issues emerging
metric and vibrational spectroscopy measurements to quantify first and from vibrational spectroscopy. Macrom Chem Phys 2013;214:1921e30.
second shell hydration layers in polyimides with different molecular archi- [31] Kamaruddin HD, Koros WJ. Sorption of methanol/MTBE and diffusion of
tectures. J Phys Chem B 2012;116:1209e20. methanol in 6FDA-ODA polyimide. J Polym Sci B Polym Phys 2000;38(17):
[15] Pydà M, Lopez-Garzon FJ. Theory of sorption of gases on heterogeneous 2254e67.
solids-polymeric sorbents. Langmuir 1993;9:2676e81. [32] Ishida H, Wellinghoff TS, Baer E, Koenig J. Spectroscopic studies of poly-[N, N0 -
[16] Barrer RM, Barrie JA, Slater J. Sorption and diffusion in ethyl cellulose. Part III. bis(phenoxyphenyl)pyromellitimide]. 1. Structures of the polyimide and three
Comparison between ethyl cellulose and rubber. J Polym Sci 1958;27:177e97. model compounds. Macromolecules 1980;13:826e34.
[17] Michaels AS, Vieth WR, Barrie JA. Solution of gases in polyethylene tere- [33] Musto P, Ragosta G, Mensitieri G, Lavorgna M. On the molecular mechanism of
phthalate. J Appl Phys 1963;34(1):1e12. H2O diffusion into polyimides: a vibrational spectroscopy investigation.
[18] De Angelis MG, Doghieri F, Sarti GC, Freeman BD. Modeling gas sorption in Macromolecules 2007;40:9614e27.
amorphous teflon through the non equilibrium thermodynamics for glassy [34] Gaffney KJ, Piletic IR, Fayer MD. Hydrogen bond breaking and reformation in
polymers (NETGP) approach. Desalination 2006;193:82e9. alcohol oligomers following vibrational relaxation of a non hydrogen bond
[19] Pantoula M, von Schnitzler J, Eggers R, Panayiotou C. Sorption and swelling in donating hydroxyl stretch. J Phys Chem A 2002;106:9428e35.
glassy polymer/carbon dioxide systems part II e swelling. J Supercrit Fluids [35] Brunauer S, Emmett PH, Teller E. Adsorption of gases in multimolecular layers.
2007;39:426e34. J Am Chem Soc 1938;60:309e19.
[20] Scherillo G, Sanguigno L, Galizia M, Lavorgna M, Musto P, Mensitieri G. Non- [36] Dixon JR, George WO, Hossain Md Fokhray, Lewis R, Price JM. Hydrogen
equilibrium compressible lattice theories accounting for hydrogen bonding bonded forms of methanol IR spectra and ab-initio calculations. J Chem Soc
interactions: modelling water sorption thermodynamics in fluorinated poly- Faraday Trans 1997;93(20):3611e8.
imides. Fluid Phase Equilibria 2012;334:166e88. [37] Pimentel GC, McClellan L. The hydrogen bond. San Francisco, CA: Freeman and
[21] Cotugno S, Di Maio E, Mensitieri G, Iannace S, Roberts GW, Carbonell RG, et al. Co.; 1960.
Characterization of microcellular biodegradable polymeric foams produced

Você também pode gostar