Effect of UV irradiation and physical aging on O2 and N2 transport properties of thin glassy poly(arylene ether ketone) copolymer films based on tetramethyl bisphenol A and 4,4′-difluorobenzophenone

Article abstract of DOI:10.1016/j.polymer.2016.01.075

Modification of membranes to improve gas separation properties has been of considerable interest. Crosslinking is one route to modify membranes, but such studies need to be performed on thin membranes to quantify the impact of such modifications at thicknesses relevant to industrial membranes. In this study, the influences of UV irradiation and physical aging on O₂ and N₂ gas permeation properties of thin (～150 nm) glassy, amorphous poly(arylene ether ketone) (PAEK) copolymer films at 35°C and 2 atm were investigated. Thin PAEK copolymer films, prepared from tetramethyl bisphenol A (TMBPA) and 4,4′-difluorobenzophenone (DFBP), were UV irradiated on both sides in air or N₂ at 254 nm or 365 nm, which induced crosslinking and, in some cases, photooxidation. Gas permeability decreased and O₂/N₂ selectivity increased as UV irradiation and aging time increased. At 254 nm, samples irradiated in air have lower permeability coefficients and higher selectivities than samples irradiated in N₂, which was ascribed to additional decreases in free volume due to photooxidation in samples irradiated in air. Additionally, samples irradiated in air at 254 nm exhibit less physical aging than uncrosslinked and samples irradiated in N₂ at 254 nm, possibly due to interactions among photooxidative polar products that may restrict polymer chain mobility, thereby lowering the aging rate. The influence of water vapor on physical aging of samples irradiated in air was examined. Finally, irradiation at 254 nm leads to more extensive crosslinking and/or photooxidation than irradiation at 365 nm, possibly due to greater UV absorption by the polymer and the higher probability of radical formation at the lower wavelength.

Full text of DOI:10.1016/j.polymer.2016.01.075

Accepted Manuscript
Effect of UV irradiation and physical aging on O and N transport properties of thin
2
2
glassy poly(arylene ether ketone) copolymer films based on tetramethyl bisphenol A
and 4,4'-difluorobenzophenone
Qiang Liu, Andrew T. Shaver, Yu Chen, Gregory Miller, Donald R. Paul, J.S. Riffle,
James E. McGrath, Benny D. Freeman
PII:
S0032-3861(16)30075-1
10.1016/j.polymer.2016.01.075
JPOL 18427
DOI:
Reference:
To appear in: Polymer
Received Date: 9 December 2015
Revised Date: 17 January 2016
Accepted Date: 27 January 2016
Please cite this article as: Liu Q, Shaver AT, Chen Y, Miller G, Paul DR, Riffle JS, McGrath JE,
Freeman BD, Effect of UV irradiation and physical aging on O and N transport properties of
2
2
thin glassy poly(arylene ether ketone) copolymer films based on tetramethyl bisphenol A and 4,4'-
difluorobenzophenone, Polymer (2016), doi: 10.1016/j.polymer.2016.01.075.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

ACCEPTED MANUSCRIPT
Effect of UV irradiation and physical aging on O2 and N2
transport properties of thin glassy poly(arylene ether ketone)
copolymer films based on tetramethyl bisphenol A and
4,4'-difluorobenzophenone
Qiang Liu1, Andrew T. Shaver2, Yu Chen2, Gregory Miller2, Donald R. Paul1,
J. S. Riffle2, James E. McGrath2 and Benny D. Freeman1*
1. Department of Chemical Engineering, Texas Materials Institute, Center for Energy and
Environmental Research, The University of Texas at Austin, 10100 Burnet Road, Bldg. 133,
Austin, TX 78758, USA
2. Department of Chemistry, Macromolecules and Interfaces Institute, Virginia Tech,
Blacksburg, VA 24061, USA
Submission to Polymer
*Corresponding Author: Email freeman@che.utexas.edu Tel: 1-512-232-2803

ACCEPTED MANUSCRIPT
Abstract
Modification of membranes to improve gas separation properties has been of considerable
interest. Crosslinking is one route to modify membranes, but such studies need to be performed
on thin membranes to quantify the impact of such modifications at thicknesses relevant to
industrial membranes. In this study, the influences of UV irradiation and physical aging on O2
and N2 gas permeation properties of thin (~ 150 nm) glassy, amorphous poly(arylene ether
ketone) (PAEK) copolymer films at 35 oC and 2 atm were investigated. Thin PAEK copolymer
films, prepared from tetramethyl bisphenol A (TMBPA) and 4,4'-difluorobenzophenone (DFBP),
were UV irradiated on both sides in air or N2 at 254 nm or 365 nm, which induced crosslinking
and, in some cases, photooxidation. Gas permeability decreased and O2/N2 selectivity increased
as UV irradiation and aging time increased. At 254 nm, samples irradiated in air have lower
permeability coefficients and higher selectivities than samples irradiated in N2, which was
ascribed to additional decreases in free volume due to photooxidation in samples irradiated in air.
Additionally, samples irradiated in air at 254 nm exhibit less physical aging than uncrosslinked
and samples irradiated in N2at 254 nm, possibly due to interactions among photooxidative polar
products that may restrict polymer chain mobility, thereby lowering the aging rate. The
influence of water vapor on physical aging of samples irradiated in air was examined. Finally,
irradiation at 254 nm leads to more extensive crosslinking and/or photooxidation than irradiation
at 365 nm, possibly due to greater UV absorption by the polymer and the higher probability of
radical formation at the lower wavelength.
Keywords: Physical aging, UV crosslinking, Gas separation, Poly(arylene ether ketone)

ACCEPTED MANUSCRIPT
1. Introduction
Polymers with high permeability, high selectivity, and long term stability are desired for
gas separation applications, and glassy polymers are typically used. However, the inherent non-
equilibrium state of glassy polymers causes physical properties, including gas transport
properties, to drift with time towards a seemingly unattainable equilibrium in a process known as
physical aging [1, 2]. As polymer chains undergo self-retarding reorganization in the glassy state,
polymer segmental mobility decreases due to free volume reduction [2]. Consequently, gas
permeability generally decreases and selectivity increases with aging [3-5]. Several studies have
demonstrated the influence of physical aging on gas transport properties of thick films (> 1 µm)
[6-9]. As film thickness decreases into the range (~ 0.1 µm [10-12]) relevant for commercial gas
separation membranes, gas permeability decreases more rapidly, suggesting accelerated aging in
thin films [3, 13, 14]. While this phenomenon is not completely understood, it is believed to be
connected to differences in polymer chain mobility at the surface versus the bulk of the films.
As thickness decreases, from the tens of microns often used in initial screening of gas transport
properties to 100 nm or less, surface to volume ratio increases enormously, so the influence of
surface properties of such thin films on other properties (e.g., gas permeability) becomes much
more important.
One route to modifying existing polymers to achieve better gas separation properties is
via crosslinking. Introducing crosslinks into some polymers can improve selectivity without
significant losses in permeability [15]. Crosslinking reactions may occur by either step-growth
polymerization, such as reaction of an amine with an epoxy at elevated temperature [16, 17], or
by chain-growth polymerization, which can be activated thermally or by irradiation, such as
exposure to UV light [18, 19]. For example, Kita et al. studied the effect of UV irradiation on
1 | P a g e

ACCEPTED MANUSCRIPT
gas transport properties of a benzophenone-containing polyimide prepared from 3,3′, 4,4′-
benzophenone tetracarboxylic dianhydride (BTDA) and 2,4,6-trimethyl-1,3-phenylene-diamine
(TMPD) [15]. After 30 minutes of irradiation in air, the H2/CH4 selectivity of BTDA-TMPD
increased by a factor of 50, whereas H2 permeability decreased by a factor of only 5. Kita et al.
concluded that the selectivity increase upon UV exposure was mainly due to changes in gas
diffusion coefficients [15]. Similar observations were reported in a study involving polyarylates
[20]. In these studies, polymer film thicknesses on the order of tens of microns were used.
However, for thick aromatic polymer films absorbing significantly in the UV range, the radiation
penetration depth may be small, thereby promoting crosslinking at and near the film surface but
not much at the center of a film, yielding non-uniform crosslinking throughout the film [20, 21].
McCaig et al. investigated the influence of UV irradiation in N2 and physical aging on gas
transport properties of thin polyarylate films (≤ 1 µm). Both crosslinking and aging decreased
gas permeability and increased selectivity, but the gain in selectivity from crosslinking was more
pronounced than that due to aging. In addition, crosslinked samples exhibited slower aging than
uncrosslinked samples, and this effect was attributed to the lower free volume of the crosslinked
samples [22]. From a fundamental perspective, it is of interest to study uniformly irradiated
samples, but the thicknesses required (<1 µm) can be in the range where physical aging effects
become significant, so careful experimental design is required to separate these effects.
The irradiation environment can also influence gas transport properties. Meier et al.
investigated the effect of UV irradiation in air or N2 on permeation properties of a an
uncrosslinkable polyimide [23, 24]. Pure-gas O2/N2 selectivity increased by 80% after the
sample was irradiated in air for 30 minutes. However, the O2/N2 selectivity did not change after
UV irradiation in N2. FTIR analysis revealed that photooxidation products, such as hydroxyl and
2 | P a g e

ACCEPTED MANUSCRIPT
carbonyl groups, formed in the sample irradiated in air, but not in the sample irradiated in N2.
Meier et al. concluded that photooxidation induced polymer densification and found polar groups
at the surface. These polar groups can interact via hydrogen bonding, lower polymer free
volume, and increase gas selectivity. Similar photooxidative effects were observed in other
studies involving both uncrosslinkable and crosslinkable polyimides [25-27].
This study investigated the influence of UV irradiation and physical aging on gas
transport properties of ultra-thin glassy, amorphous poly(arylene ether ketone) (PAEK)
copolymer films (l ~ 150 nm) prepared from tetramethyl bisphenol A (TMBPA) and 4,4'-
difluorobenzophenone (DFBP). The chemical structure and bulk properties of TMBPA-BP are
presented in Table 1 [28]. This polymer was chosen because the tetramethyl substitution helps to
reduce the efficiency of polymer chain packing, thereby increasing the polymer free volume and
improving gas permeability and provides an abundant source of benzylic methyl groups for
crosslinking reactions [29].
In addition, the benzophenone moiety is a well-known
photosensitizer that can be excited upon irradiation at 250-370 nm to initiate crosslinking
between polymer chains [30, 31]. The uniformity of irradiation throughout the sample depth was
estimated using UV-Vis spectroscopy, and the progress of crosslinking and photooxidation was
monitored using Fourier transform infrared spectroscopy (FTIR). The effects of irradiation time,
irradiation environment (air vs. N2), and irradiation wavelength (254 nm vs. 365 nm) on gas
transport properties and aging behavior were explored.
3 | P a g e

ACCEPTED MANUSCRIPT
Table 1. Bulk material properties (film thickness ~ 31µm) of uncrosslinked TMBPA-BP.
Permeability
(Barrer)1
Tg Density
(oC) (g/cm3)
Selectivity
O2/N2
Polymer
FFV
1
O2
N2
TMBPA-BP
215
1.092
0.167
4.5
0.84
5.4
1Gas transport properties measured at 35 oC and 10 atm.
2. Experimental
2.1. TMBPA-BP Synthesis
Sulfuric acid, acetic acid and acetone were purchased from Spectrum Chemicals (New
Brunswick, NJ, USA) and used as received. N-Methyl-2-pyrrolidone (NMP) was also purchased
from Spectrum Chemicals and was dried over calcium hydride and vacuum distilled before use.
4,4′-Difluorobenzophenone (DFBP) was kindly donated by Solvay and recrystallized from 2-
propanol. Methylene chloride and potassium carbonate were purchased from Fisher Scientific
o
(Pittsburgh, PA, USA). The potassium carbonate was dried at 150 C under vacuum
immediately before use. Toluene and 2,6-dimethylphenol (2,6-xylenol) were purchased from
Sigma-Aldrich (St. Louis, MO, USA) and used as received.
The synthesis of tetramethyl bisphenol A (TMBPA) was adapted from the synthesis
procedure for tetramethyl bisphenol F [32]. Acetone (7.103 g, 122 mmol) and 2,6-xylenol
(18.92 g, 155 mmol) were added to a two-neck 250-mL round bottom flask fitted with a
mechanical stirrer and an addition funnel. The solution was stirred and cooled to 15 oC in an ice
bath. A solution of sulfuric acid (16.304 g, 0.163 mol) and acetic acid (17.392 g, 0.290 mol) was
4 | P a g e

ACCEPTED MANUSCRIPT
o
mixed, chilled to 15 C, and added dropwise to the reaction flask via the addition funnel. The
reaction mixture was maintained at room temperature overnight, diluted with 100 mL of water
and 150 mL of methylene chloride, and then transferred to a separatory funnel. The organic
layer was washed with deionized water three times before collection. The methylene chloride
layer was dried over magnesium sulfate overnight and then evaporated to yield the crude product.
The product was recrystallized from toluene. Yield was 70%. The melting point of the
recrystallized product was in good agreement with the reported value (162 oC) [33].
TMBPA (29.362 g, 103.2 mmol) and DFBP (22.527 g, 103.2 mmol) were added to a
three-neck 500 mL round bottom flask equipped with a mechanical stirrer, a condenser, a
nitrogen inlet and a Dean-Stark trap. NMP (300 mL) was added to the flask, and the mixture
was stirred to obtain a clear solution. Afterwards, K2CO3 (16.408 g, 118.7 mmol) was added,
o
followed by 150 mL of toluene. The reaction bath was heated at 155 C for 4 h to remove the
o
water and toluene, kept at 150 C for 12 h, and then cooled to room temperature. The reaction
mixture was filtered to remove any excess K2CO3 or by-product salts and precipitated in 1000
mL of deionized water. The resulting polymer was stirred in deionized water at 80 oC overnight
and then dried under vacuum at 110 oC for at least 48 h. The yield of TMBPA-BP was 81%. At
o
35 C, which is the temperature used for both aging and permeation experiments in this study,
TMBPA-BP is deep within its glassy state (about 180 oC below its Tg, as shown in Table 1).
2.2. Film preparation
In general, to avoid rapid evaporation of solvent during spin coating, solvents with high
o
boiling points (>100 C) and low vapor pressures are preferred [3, 4, 34]. For this reason,
toluene was used in this study. Film thickness was controlled by varying the solution
5 | P a g e

ACCEPTED MANUSCRIPT
concentration, typically between 2 and 3 wt%. About 15 mL of polymer solution was prepared
and filtered through 5 µm, 0.45 µm, and 0.1 µm PuradiskTM PTFE syringe filters (Whatman,
Pittsburg, PA, USA), in that order, to remove dust or particulates. Afterwards, the solution was
sonicated for 30 min to remove dissolved air bubbles, since air bubbles can introduce pinhole
defects in thin films [34].
Thin films were prepared by spin coating the polymer solution onto silicon wafers (Nova
Electronics Materials, Flower Mound, TX, USA) at 1000 rpm for 3 min. Next, the silicon wafer
o
was placed on a hot plate and heated at 130 C for 3 min to remove residual solvent. To avoid
dust and particulates in the environment, spin coating was performed inside a class 100 clean
room. Film thickness was measured using a variable angle spectroscopic ellipsometer (VASE)
(M2000D, J. A. Woollam Co., Lincoln, NE, USA). Psi (ψ) and delta (ꢀ) parameters were taken
at three different angles (65, 70, and 75 degrees) over a wavelength range of 450 – 1000 nm, and
the data was modeled using the Cauchy equation [4, 12, 14, 34].
A challenge in fabricating sub-micron films is to avoid the presence of pinhole defects,
which compromise gas selectivity and render the film useless for gas separation characterization
[14, 35]. To mitigate this problem, a second layer of highly permeable polymer, such as PDMS,
is often coated onto the selective layer to block the pinholes and, thus, the selectivity-destroying
convective flow [14, 36-38]. This practice of “caulking” the defects was initially employed in
industry to treat hollow fiber membranes [36-39], and recently was used to study the influence of
physical aging on gas transport properties of ultra-thin (< 400 nm) films [12, 14, 35, 40]. At the
o
aging temperature used for this study, PDMS is about 150 C above its Tg [14], and therefore,
will not undergo physical aging [13, 14]. In addition, the PDMS layer, with a thickness typically
in the range of 5-6 µm, can also provide mechanical support to the very thin selective layer.
6 | P a g e

ACCEPTED MANUSCRIPT
Rowe et al. [14] and Cui et al. [12] demonstrated that PDMS did not significantly change the gas
transport properties of the thin glassy polymers they considered, so we presume that it does not
influence the permeability of the glassy polymers considered in this study.
A PDMS solution was prepared by mixing Wacker Silicones Corporation Dehesive 944
and cyclohexane in a 2:3 ratio by mass and then adding a crosslinking agent (V24) and
proprietary catalyst (OL) as described elsewhere [34, 35]. After the glassy polymer thickness
was measured using an ellipsometer, the PDMS/cyclohexane solution was spun onto the glassy
o
polymer at 1000 rpm for 1 min. The wafer was then heated to 115 C for 15 min to fully
crosslink the PDMS. The PDMS layer thickness was measured using a Dektak 6M stylus
profilometer (Veeco, Plainview, NY, USA).
The bilayer film was lifted from the wafer using deionized water and transferred to a thin
rectangular loop of copper wire (Cat# 2781345, Radio Shack, Austin, TX, USA). The free-
o
standing thin film was then annealed at 233 oC (~ 15 C above Tg) for 5 min in an HP 5890 gas
chromatograph (GC) oven under an ultra-high purity nitrogen purge to erase thermal history and
any orientation effects [4]. The start of the aging process (t = 0) is defined as the time when the
thin film was removed from the GC oven and rapidly quenched to ambient conditions.
2.3 UV-Vis Spectroscopy
Ultraviolet-visible spectroscopy (UV-Vis) was performed using an Agilent Cary 5000
UV-Vis-NIR spectrophotometer (Santa Clara, CA, USA). Free-standing films were attached to a
sample holder and covered an aperture through which light at various wavelengths passed.
Absorbance was measured at wavelengths ranging from 200-800 nm at a scan rate of 600
nm/min.
7 | P a g e

ACCEPTED MANUSCRIPT
2.4 FTIR
FTIR was performed using a Thermo Nicolet 6700 spectrometer (Waltham, MA, USA)
equipped with a DTGS detector. Samples were tested in transmission mode with a resolution of
4 cm-1 and 256 scans per sample.
2.5 UV irradiation
Thin film samples were irradiated in air or N2 for various amounts of time on each side
using either a UV crosslinker (XL-1000, Spectronics, Westbury, NY, USA) equipped with 254
nm light filter bulbs or a 100W high pressure mercury arc UV lamp (Blak-Ray B-100, UVP,
Upland, CA, USA) equipped with a 365 nm light filter bulb. Samples were placed 12 cm from
the bulbs in the UV crosslinker and 4 cm from the bulb of the UV lamp. The UV intensity at the
film surface was monitored using a radiometer (Part# 97-0015-02, UVP, Upland, CA, USA) and
was found to be 6.55 mW/cm2 at 254 nm and 16.7 mW/cm2 at 365 nm. For irradiation in N2,
both the crosslinker and the UV lamp were placed inside a glove box (Cat# 50601-00, Labconco,
Kansas City, MO, USA). Prior to UV irradiation, the glove box was purged with N2 (UHP grade,
Airgas, Austin, TX, USA) until the O2 level was below 0.1% by volume, which was monitored
using an oxygen sensor (Part# 9452, Nuvair, Oxnard, CA, USA). At 254 nm, films were
irradiated up to 10 min (5 min per side) in air and 20 min (10 min per side) in N2. Samples
irradiated for longer times were not mechanically stable for subsequent characterization
experiments. At 365 nm, films were irradiated for 10 min (5 min per side) in air or N2.
8 | P a g e

ACCEPTED MANUSCRIPT
2.6 Gas permeation
The influence of physical aging (up to approximately 1000 hours) on O2 and N2
permeation properties of uncrosslinked and UV-irradiated thin films was investigated.
o
Permeability coefficients were measured at 35 C and an upstream pressure of 2 atm using a
constant-volume, variable-pressure permeation apparatus built in house [41]. Thin films were
supported on a 0.02 µm AnoPore membrane disc (Cat# 6809-5502, Whatman, Maidstone,
England) and masked on both sides with aluminum tape, as described previously [35]. The mass
transfer resistance from the PDMS layer can be accounted for using a series resistance model as
follows [14, 42]:
ꢀ
ꢀ
ꢎꢏꢐꢅꢅꢑ
ꢀ
ꢁꢂꢃꢄꢂꢅꢆꢇꢈ
ꢊꢋꢌꢍ
=
+
ꢉ
ꢎꢏꢐꢅꢅꢑ
(1)
ꢖꢀꢗꢘꢘꢙ , and ꢒꢚꢛꢜꢝꢛꢘꢞꢟꢠ are the thicknesses of the PDMS layer, the underlying
selective layer, and the overall composite ( ꢒꢉꢓꢔꢕ ꢖꢀꢗꢘꢘꢙ), respectively. Accordingly, ꢡꢉꢓꢔꢕ
ꢖꢀꢗꢘꢘꢙ, and are the permeability coefficients of PDMS, the glassy polymer, and the
ꢉ
ꢉ
ꢁꢂꢃꢄꢂꢅꢆꢇꢈ
ꢊꢋꢌꢍ
where ꢒꢉꢓꢔꢕ , ꢒ
+
ꢒ
,
ꢡ
ꢡ
ꢚꢛꢜꢝꢛꢘꢞꢟꢠ
composite structure, respectively. Permeability coefficients of gases in PDMS do not change
with time, and these permeability coefficients were obtained from the literature [14].
Permeability of gases in both the uncrosslinked and UV-irradiated TMBPA-BP thin films was
calculated using Equation (1). When polymer samples were not installed in the gas permeation
systems for permeation experiments, they were shielded from light and allowed to age inside a
dark gas chromatograph (GC) oven at 35 oC.
9 | P a g e

ACCEPTED MANUSCRIPT
3. Results and discussion
3.1 UV irradiation uniformity
One general question of interest in UV crosslinking is the crosslinking uniformity
through the sample. In the thin films considered in this study, we cannot directly measure the
extent of crosslinking as a function of depth into the sample. However, we can estimate the
uniformity of UV irradiation throughout the sample. Presumably, more uniform irradiation
throughout the sample will contribute to more uniform crosslinking. The UV intensity at any
point in a film, I, can be estimated using the Beer-Lambert law [20]:
ꢢ = ꢢꢣ × 10ꢤꢥꢦꢓ = ꢢꢣ × 10ꢤꢧ
(2)
where I0 is the incident UV intensity, E is the extinction coefficient, C is the concentration of
photoactive species in the film, D is the path length, and A is the absorbance. As UV irradiation
penetrates deeper into these aromatic polymer films, its intensity will decrease due to absorption
of light by the polymer. If there is a sharp gradient of UV intensity between the surface and the
center of a film, the resulting crosslinking may not be uniform. For such a non-uniform
crosslinked material, the measured gas permeability would not represent the intrinsic property of
a homogeneously crosslinked polymer and, therefore, could not be directly compared with the
permeability of uncrosslinked samples or other non-uniform crosslinked samples.
Fig. 1A presents the UV-VIS spectra of TMBPA-BP films of different thicknesses. A
strong band at ~200 nm represents the π→π* transition associated with the benzophenone
excitation, and a weaker band at ~280 nm represents the n→π* transition [30]. As film thickness
decreases, the absorbance at 254 nm (i.e., one of the crosslinking wavelengths considered)
decreases. Because the product EC in Equation (2) is a material property and is, therefore,
10 | P a g e

ACCEPTED MANUSCRIPT
independent of film thickness, absorbance should be linearly correlated with film thickness, as
shown in Fig. 1B. Using Equation (2), relative UV intensity at any depth in a film can be
calculated. Fig. 1C shows a calculated profile of relative UV intensity for TMBPA-BP films of
different thicknesses. Normalized film thickness values of 0 and 1 correspond to the two
surfaces of a film. Because each film was irradiated for the same amount of time on each side,
the calculated UV intensity profile has a parabolic shape. To increase the uniformity of an
irradiated sample, the film needs to be thin enough so that the UV intensity at the center of the
film is as similar as possible to that at the surface. As shown in Fig. 1C, the profile becomes
more uniform as film thickness decreases. For example, for a 259 nm TMBPA-BP film, there is
a 32% difference between the UV intensity at the surface and the center of the film. For a 145
nm film, there is only a 14% difference between the UV intensity at the surface and the center of
the film. Hypothetically, films thinner than 145 nm would have an even more uniform
irradiation, but to consistently make defect-free films at such thicknesses is very difficult. As a
result, films of around 145 nm were used for this study to balance the competing needs for
relatively uniform irradiation and the practicality of making very thin, defect-free films.
11 | P a g e

ACCEPTED MANUSCRIPT
Fig. 1. (A) Effect of film thickness on UV-Vis spectra of TMBPA-BP films. (B) Effect of film
thickness on UV absorbance at 254 nm. (C) Calculated UV intensity profile at 254 nm in
TMBPA-BP films of various thicknesses.
Because PDMS is coated onto the glassy polymer to prevent pinholes from affecting the
transport results, one question is whether the PDMS layer may significantly shield the UV light
12 | P a g e

ACCEPTED MANUSCRIPT
from the underlying glassy polymer, thereby affecting the irradiation uniformity in the glassy
polymer. Fig. 2A presents the UV-Vis spectrum of a TMBPA-BP film (l=145 nm) and a
crosslinked PDMS film (l = 6 µm). The absorbance for PDMS is very low at wavelengths above
250 nm. At an irradiation wavelength of 254 nm, the PDMS absorbance (0.069) is 85% lower
than that of the TMBPA-BP thin film (0.475). Fig. 2B presents the calculated UV intensity
profile at 254 nm for TMBPA-BP films with or without PDMS coating. Because the PDMS
coating is applied to only one side of the film (at a normalized film thickness = 1 in Fig. 2B), the
addition of PDMS skews the UV profile slightly away from the center. The maximum difference
in UV intensity between the film surface and the interior is 14% for the TMBPA-BP film without
a PDMS coating and 17% for the film with a PDMS coating. Therefore, the influence of the
PDMS layer on UV irradiation uniformity is believed to be small.
Fig. 2. (A) UV-Vis spectra for TMBPA-BP and PDMS. (B) Calculated UV intensity profile at
254 nm in TMBPA-BP films (l = 145 nm) with and without a PDMS coating.
13 | P a g e

ACCEPTED MANUSCRIPT
Additionally, UV treatment at 254 nm and 365 nm is not expected to change the physical
and gas transport properties of PDMS. PDMS is generally stable against UV irradiation at
wavelengths above 200 nm [43, 44]. Below 200 nm, UV irradiation in air can promote ozone
formation, which may induce polymer chain scission and oxidation of PDMS to SiO2 at the
polymer-air interface, thereby affecting PDMS physical properties [44-46]. However, at the
wavelengths considered in this study (i.e., 254 nm and 365 nm), UV absorbance in PDMS is
small (cf., Fig. 2A). Because light must be absorbed for photochemical reactions to occur (i.e.,
the Grotthuss-Draper law) [47], little photochemical activity, and thereby physical property
changes, are expected at 254 nm and 365 nm.
Because UV absorbance at 365 nm is much lower than that at 254 nm (cf., Fig. 1A),
irradiation at 365 nm is more uniform than at 254 nm. For example, the absorbance of a 145 nm
thick TMBPA-BP film at 365 nm (0.112) is 76% lower than that at 254 nm (0.475). As a result,
at 365 nm, the difference between the UV intensity at the surface and center of a 145 nm thick
film is less than 1%, resulting in a relatively uniformly irradiated sample.
3.2 Crosslinking and photooxidation
Crosslinking and photooxidation reactions via UV irradiation of benzophenone-
containing polymers have been reported [23, 24, 48]. Based upon previous studies, Fig. 3
presents several reaction pathways that may occur to various extents. Upon irradiation at 250-
370 nm, benzophenone (BP) can be excited and abstract a labile hydrogen to yield carbon-
centered radicals and benzylic radicals (reaction 1). If oxygen is present, benzylic radicals (D)
can react with oxygen to form peroxy radicals (E), which can then abstract another hydrogen to
14 | P a g e

ACCEPTED MANUSCRIPT
form peroxides (F). Further UV irradiation and subsequent hydrogen abstraction decompose the
peroxides to generate alkoxy radicals (G) and hydroxides (H) (reaction 2).
Fig. 3. Expected mechanism of UV-induced crosslinking and photooxidation [23, 24, 48].
15 | P a g e

ACCEPTED MANUSCRIPT
Radicals can couple according to either the crosslinking or photooxidation pathways, as
outlined in Fig. 3. For crosslinking reactions, benzylic radicals can couple with carbon-centered
radicals derived from the benzophenone (reaction 3) or benzylic radicals (reaction 4) to form
crosslinks between polymer chains. McCaig et al. speculated, based on CPK models, that steric
hindrance might limit the benzylic-carbon radical coupling (reaction 3) [22], although this has
not been proven.
Photooxidation reactions can occur via coupling of oxidized intermediates, according to
reactions 5-7. For example, benzylic radicals (D) can couple with either alkoxy (G) or peroxy (E)
radicals to form aliphatic ethers (reaction 5) or peroxides (reaction 6), respectively. However,
because peroxides are typically not stable, they tend to rapidly dissociate to yield more alkoxy
radicals (reaction 6). Finally, alkoxy radicals (G) can couple with other alkoxy radicals (G) to
form aldehydes, which can be subsequently oxidized to carboxylic acids (reaction 7).
Interactions among these polar groups, e.g., hydrogen bonding, etc., can potentially increase
interchain cohesion, thereby decreasing polymer free volume and increasing gas selectivity [24].
3.3 FTIR characterization
Figs 4A-4C present FTIR spectra of TMBPA-BP films before and after UV irradiation
under different conditions. It is noted that while these films were too thin to be able to
accurately measure gel fractions on the irradiated samples, none of the irradiated materials
dissolved in chloroform or appeared to swell significantly. The peak at around 1650 cm-1
corresponds to the carbonyl group in the benzophenone moiety, and the peak near 1600 cm-1 is
attributed to the stretching of substituted benzene moieties [48, 49]. The intensity of each
spectrum was normalized using the peak at 1600 cm-1 as an internal reference, which was
16 | P a g e

ACCEPTED MANUSCRIPT
assumed to remain unchanged with UV irradiation. As shown in Fig. 4A and 4B, the carbonyl
peak intensity at 1650 cm-1 decreases with increasing UV irradiation time. Concomitantly, a
broad band at around 3500 cm-1, attributed to the stretching of hydroxyl groups generated via
crosslinking and/or photooxidation [25, 49], forms and increases with increasing UV irradiation
time. In the spectra of samples irradiated in air at 254 nm (Fig. 4A and 4C), a broad absorption
at ~1730 cm-1, which is attributed to C=O stretching from aldehydes and carboxylic acids
generated from photooxidation, emerges and increases in intensity with increasing irradiation
time. Interestingly, this band was also observed in the spectra of certain samples irradiated in N2
(e.g., “20, N2”, Fig. 4B). As described in the Experimental section, the glove box was purged
with N2 until the O2 level was below 0.1% by volume. Thus, low levels of residual oxygen in the
predominantly N2 atmosphere might have caused low levels of photooxidation, especially for
samples irradiated for relatively long times. Interestingly, as shown in Fig. 4C, samples
irradiated in air and N2 at 365 nm have similar IR spectra to the uncrosslinked sample,
suggesting that the irradiation dose at 365 nm may not have been sufficient to cause significant
crosslinking or photooxidation for the irradiation times considered.
17 | P a g e

ACCEPTED MANUSCRIPT
Fig. 4. FTIR spectra of TMBPA-BP films before and after UV irradiation. Samples in (A) and
(B) were irradiated at 254 nm and are designated by the total irradiation time (in min) and the
irradiation environment. Samples in (C) were irradiated for 10 min (5 min/side) and are
designated by the irradiation environment and irradiation wavelength. All samples were 150-170
nm thick.
18 | P a g e

ACCEPTED MANUSCRIPT
The progress of crosslinking and/or photooxidation can be monitored by observing the
decrease in the benzophenone ketone absorption (1650 cm-1) in the FTIR spectra with irradiation.
Fig. 5 presents the relative BP absorption as a function of irradiation time. The relative amount
of absorption at 1650 cm-1 was calculated by dividing the BP absorption of irradiated samples by
that of an uncrosslinked sample. Because variations in sample thickness were small (< 10%),
they are not expected to significantly affect the FTIR absorption intensities. As irradiation time
and energy increase, the relative BP absorption decreases as BP undergoes excitation and forms
crosslinks or photooxidation products. Samples irradiated in air undergo photooxidation in
addition to crosslinking, so they experience a greater reduction in BP absorption than samples
irradiated in N2. For example, after irradiation at 254 nm for 10 min, the relative BP intensity of
a sample irradiated in air is 33% lower than that of a sample irradiated in N2.
Fig. 5. Relative FTIR intensities of the benzophenone peak at 1650 cm-1 as a function of total
irradiation time, irradiation wavelength, and irradiation environment.
19 | P a g e

ACCEPTED MANUSCRIPT
As shown in Fig. 5, for a fixed irradiation environment and exposure, irradiation at 254
nm results in lower BP intensities than at 365 nm, so UV irradiation at 254 nm may cause a
greater extent of crosslinking and/or photooxidation than irradiation at 365 nm. In general, the
amount of radicals generated via UV irradiation depends on the amount of UV energy absorbed
by the polymer and the efficiency of the absorbed energy to form radicals [18, 50]. Table 2
characterizes the different UV irradiation methods considered in this study. After undergoing
irradiation for 10 min, polymers irradiated at 254 nm absorbed more energy per unit area than
those irradiated at 365 nm. Moreover, because photons at 254 nm (472 kJ/mol) are more
energetic than those at 365 nm (328 kJ/mol), photons at 254 nm are more likely to overcome the
energy barrier (310 kJ/mol) required to promote the n→π* transition of benzophenone (cf., Fig.
1) [51, 52], which leads to more radicals available for photochemical reaction. Consequently,
due to greater UV energy absorption and a higher probability of radical formation, irradiation at
254 nm may lead to more crosslinking and/or photooxidation than irradiation at 365 nm.
20 | P a g e

ACCEPTED MANUSCRIPT
Table 2. Characteristics of UV irradiation methods at 254 nm and 365 nm. Films are typically
150-170 nm thick.
Irradiation at
254 nm
Irradiation at
365 nm
Absorbancea
0.475
0.112
UV intensity
(mW/cm2)
6.55
16.7
Emitted energy per area
after 10 min irradiation
(mJ/cm2)
Absorbed energy per area
after 10 min irradiation
(mJ/cm2)b
3900
2600
10,000
2300
Irradiation dose
After 10 min irradiation
1.6
×
105
1.4×
105
(kGy)c
a cf., Fig. 1
b determined using Equation (2)
cestimated using the film thickness and the bulk film density (cf., Fig. 1),
1kGy = 1 kJ absorbed/kg of polymer
3.4 Influence of irradiation time and environment on physical aging and gas transport properties
Fig. 6 presents O2 permeability (Fig. 6A) and relative O2 permeability (Fig. 6B) as a
function of aging time for uncrosslinked and irradiated TMBPA-BP films with thicknesses
ranging from 150 to 170 nm. The relative O2 permeability was calculated by dividing the O2
permeability coefficient (P) at a given aging time by its initial value (P0) at about 1 h of aging,
which was the time at which the first permeation measurement could be conducted. The films
were irradiated at 254 nm, and they are designated by the total irradiation time and environment.
For example, “20, N2” indicates a sample irradiated in N2 at 254 nm for 20 min (10 min/side).
The initial O2 permeability of the uncrosslinked thin film is higher than that of the bulk film (cf.,
Table 1), indicating that additional free volume may have been captured in the thin film due to
21 | P a g e

ACCEPTED MANUSCRIPT
rapid quenching from above Tg [3, 13, 14, 22]. As aging time increases, free volume of glassy
polymers typically decreases [4], leading to permeability decreases. For example, the O2
permeability of the uncrosslinked sample decreases by almost 50% after about 1000 h of aging.
Fig. 6. Influence of physical aging on: (A) oxygen permeability and (B) normalized oxygen
permeability of uncrosslinked and irradiated TMBPA-BP films. In (B), relative O2 permeability
was calculated by dividing the O2 permeability (P) at a given aging time by its initial value (P0)
at ~ 1 h of aging time. Samples were irradiated at 254 nm, and they are designated by the total
time (in min) and the environment of UV irradiation. Lines are drawn to guide the eye.
The oxygen permeability also decreases with UV irradiation in either N2 or air, which
may be due to a loss in polymer free volume with crosslinking and/or photooxidation [22, 25].
Interestingly, as shown in Fig. 6B, samples irradiated in N2 display similar rates of permeability
loss as that of the uncrosslinked sample. Materials having different permeability coefficients yet
similar aging rates have been reported in earlier studies [3, 5, 53]. For example, Kim et al.
observed that a chemically crosslinked thin polyimide film (281 nm thick) had a similar aging
rate as its uncrosslinked analog (312 nm thick), even though the initial O2 permeability of the
22 | P a g e

ACCEPTED MANUSCRIPT
crosslinked sample was 57% lower than that of the uncrosslinked film [53]. Similar observations
were also reported for thin films of polysulfone (annealed vs. non-annealed) [5] and poly(2,6-
dimethyl-1,4-phenylene oxide) (semicrystalline vs. amorphous) [3].
The aging rate, according to the self-retarding aging model developed by Struik, can be
expressed as the ratio of the driving force, i.e., the displacement of the specific volume at time t
away from that at equilibrium, to a characteristic relaxation time, ꢨ, which is a function of the
polymer’s current free volume state, glass transition temperature, Tg, and experimental
temperature [2]:
ꢬ ꢮ
ꢤ ꢪꢤꢪ
ꢭ
ꢩꢪ
ꢩꢫ
Aging rate ≡
=
(3)
ꢯ ꢰꢪ,ꢱ ꢤꢱꢳ
ꢲ
where v and ꢴꢵ are the specific volumes of the polymer at time t and at equilibrium, respectively.
One possible explanation for the similar aging rates of the uncrosslinked and samples irradiated
in N2 is that the decrease in polymer free volume due to crosslinking may not be significant
enough to considerably decrease the driving force, (i.e., lower ꢴ − ꢴꢵ), for aging. In addition,
the distance between crosslinks in samples irradiated in N2 may be large enough that local
polymer segmental mobility is not significantly restricted by crosslinking, leading to small
changes in relaxation time (i.e., higher ꢨ). The net effect of these factors (i.e., lower ꢴ − ꢴꢵ and
higher ꢨ) should reduce the aging rate (cf., Equation (3)). However, because samples irradiated
in N2 and the uncrosslinked polymer exhibit similar aging rates, the changes to the driving force
and the relaxation time are likely minor.
For a fixed UV irradiation time, samples irradiated in air have lower O2 permeability than
samples irradiated in N2. In addition to crosslinking, photooxidation can occur due to irradiation
23 | P a g e

ACCEPTED MANUSCRIPT
in air, which introduces polar groups into the polymer matrix (cf., Fig. 4B). Interactions among
these polar groups, such as hydrogen bonding, etc., can increase interchain cohesion, thereby
lowering free volume and gas permeability [23, 24]. Moreover, such interactions may also
restrict polymer chain mobility, increasing Tg, and according to Equation (3), increase the
relaxation time, thereby decreasing the aging rate. Consistent with this hypothesis, as shown in
Fig. 6B, samples irradiated in air have lower aging rates than uncrosslinked polymer and samples
irradiated in N2, and the aging rate of samples irradiated in air decreases as irradiation time
increases.
In addition to the interactions described above, polar groups introduced by
photooxidation can potentially increase polymer hydrophilicity [54]. Particularly in more
hydrophilic polymers, water can influence physical aging [55]. To explore the potential
influence of water on physical aging of polymers irradiated in air considered in this study, two
fresh thin film samples were prepared and irradiated in air for 10 min (5 min/side). One film was
kept in a permeation apparatus for up to 1000 hours. Except during permeation measurements,
o
which used dry O2 and N2, that film was aged in vacuum at 35 C and, consequently, had
minimal contact with water vapor. The other sample was aged in a dark GC, which had a
o
relative humidity of about 25%, at 35 C. The relative humidity was measured using a VWR
electronic digital humidity meter (P/N 35519-055, VWR, Radnor, PA, USA). After each
permeation measurement, the sample was immediately removed from the permeation apparatus
and continued aging in the ambient atmosphere. The aging behavior of these samples is
presented in Fig. 7. During the first 21 h, each sample was kept in its respective permeation
apparatus, so the samples were not exposed to water vapor, and the initial change in relative O2
permeability with time is essentially identical for both samples. After about 21 h, one sample
24 | P a g e

ACCEPTED MANUSCRIPT
was removed from its permeation system and aged in the ambient environment, as described
above. As shown in Fig. 7, the sample aged in the ambient environment, which contained water
vapor at about 25% relative humidity, underwent accelerated aging relative to the sample aged in
vacuum, which were not exposed to water.
Fig. 7. Influence of aging environment on aging rate of TMBPA-BP thin films irradiated in air.
Relative O2 permeability was calculated by dividing the O2 permeability (P) at a given aging
time by its initial value (P0) at ~ 1 h of aging time. The films were irradiated in air for 10 min (5
o
min/side) and were around 170 nm thick. One sample was aged in vacuum at 35 C, and the
other sample was aged in vacuum for 21 hours before being subsequently aged in the ambient
environment, which had a relative humidity of about 25%, at 35 oC.
This accelerated aging in the presence of water is similar to a phenomenon reported by
Horn et al., who investigated the influence of long term CO2 exposure on aging behavior of thin
Matrimid films (~200 nm) [35]. At a fixed feed pressure and temperature, CO2 permeability in
Matrimid initially increases with increasing aging time until a maximum is reached, then it
decreases for the remainder of the experiment. Initially, CO2 plasticizes the polymer and
25 | P a g e

ACCEPTED MANUSCRIPT
increases gas permeability. Then, once the polymer is plasticized, its free volume increases and
Tg decreases, leading to an increase in the driving force for physical aging and a decrease in the
relaxation time (cf., Equation (3)), both of which act to increase the aging rate. Consequently, at
later times, the influence of aging on permeability becomes more important than the initial
plasticization/swelling event, resulting in a decrease in CO2 permeability with time [35, 56]. A
similar interplay between plasticization and physical aging may also be at work in this study.
Because water (Tc = 647 K [57]) is much more condensable than CO2 (Tc = 304 K [57]), water
can potentially plasticize the more hydrophilic samples irradiated in air, leading to accelerated
aging, as shown in Fig. 7. Of course, other components in the ambient environment, such as
carbon dioxide or other species, may also have an impact on physical aging. A more detailed
and controlled study of aging environment influence on aging behavior deserves further
investigation, but it is beyond the scope of this study. In any case, the observed dependence of
aging rate on the aging environment highlights the importance of maintaining consistent aging
conditions and reporting how samples are stored between gas permeation measurements during
aging studies.
Fig. 8 presents the absolute (A) and relative (B) O2/N2 selectivity of uncrosslinked and
irradiated samples as a function of aging time. Relative selectivity was determined by dividing
selectivity at time t by the initial selectivity, which was measured after about 1 h of aging. The
uncertainty for the 10 min, air-irradiated sample is presented, calculated based on uncertainties in
permeability coefficients using propagation of errors [58]. Uncertainties in the other data points
are believed to be similar, but they are not shown for clarity in this figure. The initial O2/N2
selectivity of the uncrosslinked sample is similar to that of the bulk film (cf., Table 1), indicating
that the PDMS-coated thin film is essentially defect-free. Selectivity generally increases with
26 | P a g e