Morphological Anisotropy and Proton Conduction in Multiblock

Article pubs.acs.org/Macromolecules

Morphological Anisotropy and Proton Conduction in Multiblock Copolyimide Electrolyte Membranes Christoph F. Kins,† Esha Sengupta,† Anke Kaltbeitzel,† Manfred Wagner,† Ingo Lieberwirth,† Hans Wolfgang Spiess,† and Michael Ryan Hansen*,†,‡ †

Max Planck Institute for Polymer Research, Ackermannweg 10, D-55128 Mainz, Germany Interdisciplinary Nanoscience Center (iNANO) and Department of Chemistry, Aarhus University, Gustav Wieds Vej 14, DK-8000 Aarhus C, Denmark

‡

ABSTRACT: Performance improvement of advanced polymer electrolyte membranes requires control over the morphology as it plays an important role for mechanical, thermal, and proton transport properties. The ionic domain orientation of films cast from sulfonated block copolymer solutions is generally anisotropic, and to maximize proton conductivity in fuel cell applications, hydrophilic channel alignment is desirable. In this work, a series of multiblock copolymers based on sulfonated copolyimides were synthesized and characterized by NMR, TEM and AFM microscopy, and impedance spectroscopy. For constant ion exchange capacity, the higher the block length, the higher the proton conductivity and water uptake for a given relative humidity. A random copolymer exhibited the lowest performance, in particular at low relative humidity, caused by a reduced phase separation as derived from AFM and TEM measurements. Orientational order probed by 2H NMR on absorbed D2O showed preferential alignment in the through-plane direction. However, proton pulsed-field-gradient (PFG) NMR along the two orthogonal membrane directions revealed water diffusion to be faster in-plane than in the through-plane direction. This difference in diffusion is attributed to a lamella-like structure composed of rather short, through-plane hydrophilic channels in our systems. For the block copolyimide with the highest block length, two distinct diffusion processes could be identified. This is ascribed to a superimposed morphology on the micrometer scale, leading to an opaque appearance of the membrane.

■

nated polyimides (SPIs).4,5 These aromatic polymers are in general inexpensive, easy to modify (e.g., sulfonation), and have favorable film forming abilities in addition to a high chemical resistance. The IEC is the most important factor influencing both membrane stability and proton conductivity. However, both of these features are competing material properties, where, for sulfonated aromatic structures, the higher the degree of sulfonation and thereby proton conductivity, the larger the swelling and hydrolysis susceptibility.6 To counterbalance the effect of high IECs for obtaining high performance polymers, it is essential to choose appropriate monomers. A considerable variety of monomers have been studied, including 1,4,5,8naphthalenetetracarboxylic dianhydride (NTDA),7 4,4′-binaphthyl-1,1′,8,8′-tetracarboxylic dianhydride (BTDA),8−10 sulfonated BTDA, 11 , 1 2 2,2′-benzidinedisulfonic acid (BDSA), 13 2,2′- or 3,3′-bis(4-sulfophenoxy)benzidine (BSPOB),14,15 2,2′-bis(4-aminophenoxy)biphenyl-5,5′-disulfonic acid (oBAPBDS),16 2,2′- or 3,3′-bis(3-sulfopropoxy)-

INTRODUCTION The development and study of new polymer electrolyte membranes (PEMs) has attracted considerable attention in the past decades due to their wide range of applications, especially in the field of clean energy sources.1 PEMs are important components in polymer electrolyte fuel cells (PEFC). Their performance depends on a variety of factors, such as water uptake, morphology, ion exchange capacity (IEC), and specific chemical structure. Besides chemical stability as well as favorable mechanical and thermal properties, high proton conductivity is a crucial prerequisite for efficient fuel cell operation. In this regard, perfluorosulfonic acid membranes (e.g., commercially available Nafion) are considered state-of-the-art PEMs. Although perfluorosulfonic acid membranes have been the benchmark for decades, their industrial application is limited due to several drawbacks. These include poor conductivity at low relative humidity ( 100 °C), high price, and fuel gas crossover.2,3 For this reason considerable effort has been devoted to the development of alternative PEMs, where a particular focus has been on sulfonated aromatic structures, including sulfonated poly(aryl ether sulfone)s, sulfonated poly(aryl ether)s, sulfonated poly(ether ketone)s, and sulfo© 2014 American Chemical Society

Received: February 6, 2014 Revised: April 2, 2014 Published: April 8, 2014 2645

dx.doi.org/10.1021/ma500253s | Macromolecules 2014, 47, 2645−2658

Macromolecules

Article

benzidine (BSPB),17 and 4,4′-oxidianiline-2,2′-disulfonic acid (ODADS).18 The aim of this work is to elucidate the morphology and its associated anisotropy for a given block copolyimide using a number of different experimental techniques, focusing on different length scales. We have chosen NTDA, 19−21 2,2-bis(4-aminophenyl)hexafluoropropane (BAHF),6,22−24 and ODADS9,12,18,25 as dianhydride, diamine, and sulfonated diamine, respectively (see Scheme 1). These chemical building blocks have proven to exhibit good performance and are readily available.

performance.36 Recent studies have shown that membrane swelling leads to different conductivity values in the throughplane and in-plane directions, pointing toward anisotropic structure formation.6,12,20,23 Notably, through-plane conductivity measurements are not straightforward, mainly due to impedance contributions from the large interfacial regions between the polymer membrane and the electrodes. For this reason care must be taken if conductivity values are compared with those measured through-plane.5,23,32,37 Here, we characterize the relationship between water diffusion and alignment of the proton conducting channels determined via 1H pulsed-field-gradient (PFG) NMR diffusometry38 and 2H NMR spectroscopy,39 respectively, as a function of adsorbed water, relative humidity, and block copolymer length. Pulse-field gradient NMR is in contrast to most other techniques a noninvasive technique and does not rely on any complex theoretical models or assumptions.40−44 To assess the anisotropic proton transport, we have used 1H PFG NMR experiments applied to different membrane orientations with respect to the static magnetic field. It is important to note that the proton conductivity and water diffusion coefficients do not necessarily correlate with each other despite they both depend on the mobility of water molecules. However, water mobility does depend not only on the local environment but also on the percolation between different ionic domains, which provides diffusion pathways through the membrane. Conductivity, on the other hand, is a macroscopic effect, whereas the length scale probed by 1H PFG NMR is in the range 100 nm−10 μm, depending on the experimental setup. The orientational anisotropy, describing the degree of alignment of the proton conduction channels, has previously been elucidated by 2H NMR spectroscopy in a variety of systems.45−50 We note that considerable work related to the combination of 1H PFG NMR and 2H NMR has been reported in recent years by Madsen and co-workers.51−53 Using

Scheme 1. Chemical Structures of the Monomers 1,4,5,8Naphthalenetetracarboxylic Dianhydride (NTDA), 4,4′Oxydianiline-2,2′-disulfonic Acid (ODADS), and 2,2-Bis(4aminophenyl)hexafluoropropane (BAHF) Used in This Work To Form Block Copolyimides

Hydrophilic−hydrophobic microphase-separated structures are known to enhance the proton conductivity of PEMs and have been extensively studied by numerous experimental techniques, including small-angle X-ray scattering (SAXS), 3 , 2 6 − 2 9 transmission electron microscopy (TEM),19,30−32 and atomic force microscopy (AFM).24,33−35 It is widely accepted that block copolymers show better fuel cell performance than their random counterparts, especially at low relative humidity.4 Thus, it should be possible to control the specific self-organization by varying the relative volumes and chain lengths of the hydrophilic and hydrophobic blocks. Furthermore, the procedure of film preparation is crucial and complex, providing an additional handle to impact conductivity

Scheme 2. Reaction Pathways for the Synthesis of Multiblock Cosulfonated Polyimides (Co-SPIs)

2646


Macromolecules

Article

%). The glass plates were dried in an oven at 120 °C and ambient pressure, followed by 24 h at 120 °C in vacuo to remove residual solvent. The obtained membranes of random co-SPI and multiblock co-SPIs with block lengths up to 20 were transparent. However, the co-SPI NTDA-ODADS/BHF (50/50) membrane was opaque although it formed a transparent solution in m-cresol before film casting. The as-cast membranes were soaked in methanol at 50 °C for 1 h, and then proton exchange was conducted by immersing the films in 1.0 N hydrochloric acid at room temperature overnight. The membranes in proton form were thoroughly washed with deionized water and dried in vacuo at 120 °C for 24 h. The thickness of the coSPI films was in the range of 30 μm. Membrane Characterization. Solution-state 1H NMR spectra were measured on a Bruker AVANCE-III 250 spectrometer. Thermogravimetric analyses (TGA) were conducted with a Mettler Toledo SDTA851 machine at a heating rate of 10 °C/min in air (30− 900 °C). Differential scanning calorimetry (DSC) was performed in air from 30 to 200 °C at a heating rate of 10 °C/min using a Mettler Toledo 822/400. The molecular weight of the protonated polymers and oligomers was measured via gel permeation chromatography (GPC). Before the addition of the hydrophobic oligomer reaction mixture to the flask containing the sulfonated oligomer, a sample of about 0.1 mL of each oligomer reaction mixture was taken and precipitated in acetone (hydrophilic block) and methanol (hydrophobic block), respectively. Prior to the GPC measurement, the hydrophilic oligomers were transferred to the protonated state as described above. It turned out that the hydrophobic oligomers as well as the multiblock copolyimide with the longest blocks (50 repeat units) were not soluble in DMF, resulting in turbid solutions. Thus, molecular weights are only given for hydrophilic oligomers and multiblock copolyimides with block lengths shorter than 50. A Waters machine equipped with three PSS GRAM columns and a Soma S-3702 UV detector (270 nm) were used with DMF containing 0.01 M LiBr as eluent (flow rate: 1.0 mL/min; temperature: 60 °C). The polymer solutions were filtered through a 0.45 μm PTFE filter prior to injection. Molecular weights were calculated against poly(styrene) standards. Proton conductivity of the membranes was measured by dielectric spectroscopy in a two-electrode in-plane geometry over a frequency range from 0.1 to 106 Hz, using a SI 1260 impedance/gain-phase analyzer and a Novocontrol broadband dielectric converter. The samples with a typical size of 10 × 10 mm were contacted by E-TEK electrodes and placed in a climatic chamber (Binder KBF 240). The membranes were exposed to the specified relative humidity (RH) and temperature until impedance data showed that the moisture content had equilibrated. Care was taken to avoid contributions from external water.57 Because of the small membrane thicknesses of 20−40 μm, this state was obtained within several hours. The specific conductivity was calculated from Bode plots. At least two samples of each membrane were measured and values averaged. Light scattering experiments were performed at room temperature using an ALV unit equipped with an ALV/CGS3 compact goniometer (ALV-Laser Vertriebsgessellschaft m.b.h., Langen, Germany), ALV/ LSE-5004 correlator, and a He/Ne laser (λ = 632.7 nm). Signals from the detector were processed by ALV5000 software. The samples were dissolved in H2O, DMF (1 g/L LiBr), and DMSO to give a concentration of 1 g/L and finally filtered with 0.45 μm PTFE filters prior to measurements. For TEM characterization the samples were sectioned at room temperature to a nominal thickness of 80 nm by microtomy using a diamond knife and transferred to a 400 mesh copper TEM grid. Microstructural characterization of these thin sections was done using a FEI Tecnai F20 transmission electron microscope equipped with a Gatan Tridiem 863 post column energy filter. All micrographs were taken at an acceleration voltage of 200 kV. The elemental distribution of sulfur was acquired by applying the three-window method under the assumption of a power law background in the electron energy loss spectrum.58 Membranes were stained with silver by immersion of samples in a 0.5 M AgNO3 aqueous solution for 24 h, rinsed with distilled water, and dried at room temperature.

2

H NMR experiments, we take advantage of the fact that the electric-field gradients of the 2H nucleus are not completely averaged to zero through the nonisotropic molecular reorientations of the deuterated water molecules (D2O). This results in a characteristic 2H quadrupole splitting that is related to the specific orientational anisotropy. For PEMs, it has been shown that the morphology induced partial ordering of absorbed D2O molecules allowed the identification of the symmetry axis of diffusion.54−56 This behavior is strongly related to the morphology, which can be obtained from TEM or AFM. Thus, the combination of multiple techniques, including NMR spectroscopy, dielectric spectroscopy, and scanning probe techniques, provides a direct connection between morphology, orientation, and transport, aiding the rational design and evaluation of the next-generation PEMs.

■

EXPERIMENTAL SECTION

Materials. 1,4,5,8-Naphthalenetetracarboxylic dianhydride (NTDA), 4,4′-diaminodiphenyl ether (ODA), m-cresol, and triethylamine (TEA) were purchased from Sigma-Aldrich and used as received. 2,2-Bis(4-aminophenyl)hexafluoropropane and benzoic acid were obtained from Acros. 4,4′-Oxydianiline-2,2′-disulfonic acid (ODADS) was prepared according to the method previously reported.18 All monomers were stored under vacuum to avoid moisture contamination. Synthesis of Random Sulfonated Copolyimide. To a 100 mL completely dried flask equipped with a condenser were successively added ODADS (540.5 mg, 1.5 mmol), m-cresol (15 mL), and TEA (0.5 mL, 3.6 mmol) under argon flow with stirring. After ODADS was completely dissolved, NTDA (804.5 mg, 3.0 mmol), BAHF (501.4 mg, 1.5 mmol), and benzoic acid (512.9 mg, 4.2 mmol) were added. The mixture was stirred at room temperature for a few minutes and then heated at 80 °C for 4 h and subsequently at 180 °C for 20 h. After cooling to 100 °C, an additional 15 mL of m-cresol was added to dilute the highly viscous solution. The solution was then poured into 300 mL of acetone. The precipitate was filtered off, washed with acetone, and dried in vacuo for 24 h at 80 °C. Synthesis of Multiblock Sulfonated Copolyimides. Multiblock co-SPIs were synthesized by a two-pot method as described below in Scheme 2. As an example, the synthesis of NTDA-ODADS/BHF (20/ 20), where the numbers in parentheses refer to the hydrophilic/ hydrophobic block lengths, is described in the following. The anhydride-end-capped hydrophilic oligomer was synthesized as follows: a completely dried 100 mL flask equipped with a condenser was charged with ODADS (540.5 mg, 1.5 mmol), m-cresol (6 mL), and TEA (0.5 mL, 3.6 mmol) under argon flow with stirring. After ODADS was completely dissolved, NTDA (418.2 mg, 1.5592 mmol) and benzoic acid (268.7 mg, 2.2 mmol) were added to the flask. The reaction solution was heated to 80 °C, left to stir for 4 h, and then kept at 180 °C for 20 h. The amine-end-capped hydrophobic oligomer was synthesized as follows: a completely dried 100 mL flask equipped with a condenser was charged with BAHF (496.3 mg, 1.4847 mmol) and mcresol (8 mL) under argon flow with stirring. After BAHF had dissolved, NTDA (382.2 mg, 1.4253 mmol) and benzoic acid (268.7 mg, 2.2 mmol) were added to the flask. The reaction solution was stirred at 120 °C for 24 h. The hydrophobic block oligomer solution was carefully added to the flask with the hydrophilic oligomer solution. The mixture was left to stir for 2 h at 120 °C, followed by 22 h at 180 °C. After cooling to 120 °C, additional m-cresol (15 mL) was added to dilute the viscous solution. The dark mixture was poured into 300 mL of acetone, and the precipitate was filtered, washed with acetone, and finally dried in vacuo for 24 h at 80 °C. Preparation of Sulfonated Copolyimide Membranes. The sulfonated copolyimide membranes were prepared using a solvent-cast method. Co-SPIs in TEA salt form with block lengths up to 20 were dissolved in DMSO (∼5 wt %) under heat, and the solution was filtered and cast onto glass plates. Because of solubility problems, coSPI NTDA-ODADS/BHF (50/50) was dissolved in m-cresol (∼5 wt 2647


Macromolecules

Article

into a handmade Teflon sample cell (see Figure 1a). This cell fits vertically into a regular 5 mm NMR tube. For the through-plane

The AFM topography mapping was performed in intermittent contact mode under ambient conditions. Measurements were done with a Bruker D3100. Tapping mode cantilevers were used (nominal resonance frequency 70 kHz and 2 N/m spring constant, OMLAC 240 TS, Al back side coated). Tips were changed at regular intervals to avoid artifacts resulting from tip changes during scanning. For domain size analysis, all points in an image above 65% of height were selected. The average domain size was calculated using the GWYIDDION software. Ion exchange capacities (IECs) were determined by titration and compared with the theoretical values calculated from the monomer feed ratios. The membranes in proton-form were immersed in a 1 M NaCl solution for 24 h to liberate the protons. The solution with released H+ ions was then titrated with aqueous 0.01 M NaOH solution using phenolphthalein as indicator. Additionally, membranes in proton form were dissolved in DMSO-d6 for 1 H NMR measurements. No residual TEA resonances were visible, pointing toward quantitative proton exchange. Water uptake experiments at room temperature were carried out by drying ∼30 mg per membrane sheet in vacuo at 100 °C overnight, followed by an immersion of the sheets in deionized water for 2 days. The films were then taken out, wiped clean with tissue paper, and quickly weighed on a microbalance. Water uptake at a defined RH and elevated temperature was measured by equilibrating membrane sheets in a climatic chamber for at least 3 h at 50 °C. Water uptake of the films in wt % was calculated as (Wwet − Wdry)/Wdry, where Wdry and Wwet are the weights of dry and water-swollen membrane, respectively. Dimensional change of the SPI membranes was measured by drying a round-shaped sample (2 cm diameter) in vacuo at 100 °C overnight, followed by an immersion of the sheets in deionized water for 2 days. Through-plane and in-plane dimensional changes as well as the anisotropic membrane-swelling ratio were calculated as

Δt =

t − t0 , t0

Δl =

l − l0 , l0

Δt / l =

Δt Δl

Figure 1. Schematic diagram of the NMR sample cells fabricated from polytetrafluoroethylene. Care was taken to make tight-fitting caps to prevent loss of water during the experiments. At least 10 membrane sheets were stacked to ensure an optimal filling factor, where, for (a) in-plane diffusion measurements, the membranes were cut into rectangular pieces of 3 × 4 mm in size. Round shaped sheets were punched out of membranes and stacked on top of each other in the cylindrical cavity for (b) through-plane diffusion measurements. A larger cell based on that shown in (a) was used for the 2H NMR measurements.

(1)

where t0 and l0 are the thickness and length of the dry membrane, respectively; t and l refer to the values measured after immersion in water. Solid-state 1H magic-angle spinning (MAS) NMR spectra of the dried membranes were acquired on a Bruker AVANCE 700 spectrometer. Experiments were carried out in a 2.5 mm double resonance MAS probe (Bruker) spinning at 25.0 kHz with a π/2 pulse length of 2.5 μs and a recycle delay of 1 s (well above T1). All 1H MAS NMR and double-quantum filtered 1H MAS NMR spectra were recorded using 32 transients and referenced to TTSS (0.27 ppm, 1 H).59 The sample temperature was corrected to include heating effects arising from high-speed MAS.60 The 13C{1H} cross-polarization (CP)/MAS NMR spectra were acquired on a Bruker ASX 500 spectrometer using a CP contact time of 1 ms by coadding 32 768 transients. These experiments were carried out using a 2.5 mm double resonance MAS probe (Bruker) spinning at 25.0 kHz, a π/2 pulse length of 2.5 μs, and a recycle delay of 2 s. Self-diffusion coefficients of water in the membranes were measured both in-plane (Dip) and through-plane (Dtp) using a narrow-bore Bruker AVANCE-III 700 magnet equipped with a single-axis diffusion probe. This probe had a maximum gradient of 1192 G/cm in the B0 direction. Pulsed-gradient stimulated-echo (PGSTE) experiments with δ = 1 ms (≤T2) and Δ = 10 ms were performed at 25 °C for membranes equilibrated to 60%, 45%, and 30% RH, whereas Δ was set to 20 ms for the membrane with the lowest water content (19% RH). Dip and Dtp proved not to vary with Δ in the range of 10−50 ms. T1 was in all cases ∼10 ms, and for this reason a recycle time of 0.43 s was chosen for all measurements. Sixteen scans were coadded for each gradient step. The temperature was calibrated using a standard solution of 4% CH3OH in CD3OD. The gradient constant was calibrated by measuring the diffusion coefficient (32 gradient steps) of 1% H2O in a solution of 0.1 mg/mL GdCl3 in D2O (“doped water”) to a literature value of 1.91 × 10−9 m2/s at 25 °C.61,62 For the in-plane diffusion measurements, a stack of membranes (∼10−15) was loaded

measurements, a second Teflon cell composed of a cylindrical cavity was used (see Figure 1b). In the first case illustrated in Figure 1a, the magnetic field direction (B0) is parallel to the membrane surface, whereas in the second case (Figure 1b) it is normal to the membrane surface. In both cases care was taken to place the stacked membranes at the center of the RF coil. Membrane water uptake was adjusted by putting the open (without the piston cap), loaded sample cell into a climatic chamber with a defined RH at 25 °C overnight. Fitting of the resulting data and calculation of diffusion coefficients was performed with the Bruker TopSpin 2.1 software. Static 2H NMR spectroscopy was performed at 25 °C on a widebore Bruker AVANCE-I 300 spectrometer operating at 2H Larmor frequency of 46.09 MHz. A single-channel static NMR probe equipped with an 8 mm inside diameter horizontal solenoid coil was used. The sample cell was rotated inside the solenoid coil using a goniometer to adjust the orientation of the sample cell to ±2° accuracy. The shape of the sample cell was identical to the cell used for 1H PFG NMR experiments for in-plane diffusion measurements (Figure 1a). However, due to the larger inside diameter of the solenoid coil, the slit for the membrane stack had a size of 6 × 6 mm. Experiments were carried out using single-pulse excitation with a π/2 pulse of 15 μs, a recycle delay of 0.5 s (well above T1), and typically 256 scans per spectrum. 2H quadrupole splittings were extracted by fitting each spectrum using the DMFit software.63 All membranes were soaked in D2O for at least 24 h, cut into 10 pieces of 6 × 6 mm in size, stacked together, and trimmed to match the rectangular shaped cavity of the cell. The wt % of water in unsaturated membranes was measured using relative NMR signal intensities; i.e., the signal intensity of fully watersaturated membranes was set as reference D2O was allowed to slowly evaporate from the membranes by removing the sample cell cap in a nitrogen atmosphere. 2648


Macromolecules

Article

Table 1. Molecular Weight and IEC of the Sulfonated Multiblock Polyimides sampleb

x:y ratioc

Mnd (kg/mol)

co-SPI-r co-SPI-5 co-SPI-10 co-SPI-20 co-SPI-50a co-SPI-f

5:5 10:10 20:20 50:50

11 23 37 81

Mne (kg/mol)

Mwe (kg/mol)

PDIe

IECf (mequiv/g)

93 94 90 67

269 305 274 321

2.9 3.2 3.0 4.8

45

128

2.8

1.64 1.66 1.67 1.70 1.71 3.21

a

This sample was opaque and not soluble in DMF. bThe samples are named according to the ratio between hydrophilic and hydrophobic parts (numbers), whereas r and f correspond to a random and fully sulfonated polyimide. cRatio between hydrophilic (x) and hydrophobic (y) parts. d Sulfonated oligomers. eSulfonated polymers. fDetermined via titration. A theoretical value of 1.73 mequiv/g is predicted except for sample f (3.37 mequiv/g).

■

RESULTS AND DISCUSSION Synthesis and Membrane Characterization. Multiblock cosulfonated polyimides (co-SPIs) derived from NTDA, ODADS, and BAHF (see Scheme 1) were synthesized using a two-step procedure as described in Scheme 2. Anhydride endcapped sulfonated oligomers (hydrophilic) and amine-endcapped hydrophobic oligomers were prepared separately in two flasks. Subsequently, the hydrophobic solution was carefully poured into the hydrophilic solution and both mixed thoroughly. Care was taken to conduct this transfer quantitatively to obtain multiblock copolymers with high molecular weights. The reaction mixture was kept at 180 °C for at least 22 h to ensure full conversion and imidization. A random copolyimide (co-SPI-r) and fully sulfonated polyimide (co-SPI-f), corresponding to polyimides derived from NTDA and ODADS only, were prepared in a similar fashion by a onestep procedure. A series of block-sulfonated copolyimides with different block lengths were synthesized. The length of the hydrophilic (x) and hydrophobic (y) segments was controlled by varying the ratio of NTDA to diamines (ODADS or BAHF). In this work we have kept x = y for all multiblock copolymers with x = y = 5, 10, 20, and 50. As nomenclature for these samples we have used co-SPI-z with z designating the ratio between hydrophilic and hydrophobic parts. To be able to assess the effect of varying block lengths on morphology and proton conductivity, the ion exchange capacity (IEC) was held constant for all copolymers. BAHF was chosen as nonsulfonated diamine due to the good solubility of the resulting hydrophobic oligomers in m-cresol. All multiblock co-SPI and random copolyimide membranes were transparent, flexible, tough, and ductile, except for the membrane based on co-SPI50. Attempts to form non-sulfonated oligomers via the reaction of NTDA and ODA in m-cresol failed; the reaction mixture was opaque, and subsequent addition of the hydrophilic oligomers resulted in multiblock copolyimides with low molecular weights. The obtained films are opaque and brittle. These observations are in accordance with previous reports, where hydrophobic blocks derived from NTDA and other diamines with aryl ether linkages did not form clear solutions either.22,23 However, the random copolyimide obtained from the polycondensation of NTDA, ODADS, and ODA did exhibit a high molecular weight and was soluble in m-cresol. The IEC and molecular weights of the synthesized copolyimides, random co-SPIs, and fully sulfonated SPI are summarized in Table 1. The calculated IEC for all polymers (except for co-SPI-f) was 1.73 mequiv/g. Experimental values obtained by conventional titration were only slightly lower, and additional 1H NMR data of proton-exchanged membranes showed no residual triethylamine (TEA) signal intensity,

pointing toward quantitative removal of TEA. All synthesized polymers showed relatively high molecular weights and were readily soluble in DMF or DMSO, except block for co-SPI-50, which had the highest block length. Most likely, the long hydrophobic segments in multiblock copolyimide hamper dissolution in polar aprotic solvents, as all hydrophobic oligomers were also insoluble in DMF irrespective of the chain length. Although no molecular weight data could be determined for co-SPI-50, the resulting film was as flexible and ductile as the other membranes, suggesting a sufficient high molecular weight. Absolute molecular weight values for the hydrophilic oligomers were 3 times higher than expected (e.g., Mn = 11.0 kg/mol for the sample with theoretically 5 repeat units; the calculated value is around 3.4 kg/mol). This might be due to the presence of a significant amount of aggregates in the GPC eluent despite the fact that the film dissolves in DMF, giving optically clear solutions.64 Indeed, DLS experiments of co-SPI-f in DMF (data not shown) revealed significant aggregates through the presence of polymer particles with hydrodynamic radii exceeding 100 nm. However, during the GPC experiment, these aggregates possibly broke up through shearing forces in the column. The GPC data are usually referenced relative to a standard and since the molecular weight of hydrophilic (sulfonated) oligomers was approximately proportional to the calculated number of repeat units, this proves good control over the block length. Water uptake, or level of hydration, is a crucial factor determining the properties of a PEM. Generally, proton conductivity increases with increasing water content and density of acidic protons.65 However, the mechanical strength decreases with increasing IEC, limiting the practicable amount of water uptake.6 Block copolymers are believed to be potentially superior to their random counterparts in maintaining membrane stability at a higher level for a given water uptake.4 Table 2 lists the water uptake for the synthesized block copolyimides and the corresponding random polyimide. The water uptake was lowest for the random copolyimide co-SPI-r, rising steadily with increasing segment length. Sample co-SPI-5 Table 2. Water Uptake and Anisotropic Membrane Swelling sample

water uptake (%)

λwatera

Δtb

Δlb

Δt/lb

co-SPI-r co-SPI-5 co-SPI-10 co-SPI-20 co-SPI-50

40 41 57 58 69

12.8 13.1 18.3 18.6 22.2

0.21 0.21 0.30 0.55 0.54

0.14 0.08 0.17 0.13 0.19

1.5 2.6 1.8 4.2 2.8

Number of water molecules absorbed per sulfonic acid group λ = [H2O]/[SO3H]. bCalculated according to eq 1.

a

2649


Macromolecules

Article

membranes it is only ∼5 nm, whereas for co-SPI-20 the average domain size doubles in size to ∼10 nm. The membrane based on co-SPI-50 exhibits the largest microphase separated structures with ∼30 nm in size. This increase in domain size is most likely related to the increasing block length, facilitating the clustering of larger hydrophilic and hydrophobic domains. More importantly, the TEM images were taken under high vacuum conditions and may not represent the actual morphology at ambient conditions accurately. For this reason, we have recorded AFM tapping mode images at ambient conditions. These are shown in Figure 2b,d,f for comparison with the TEM images. Basically, the AFM images confirm the information derived from the TEM images, even though the domain sizes appear to be slightly larger than in TEM. However, the trend observed in the TEM images that the microphase separation increases with increasing block length of the hydrophilic and hydrophobic parts is still visible. Most likely, the apparent increase in domain size in the AFM images is due to water-induced swelling of the hydrophilic domains; i.e., water is absorbed from the air and incorporated into the hydrophilic channels, increasing the apparent diameter of these channels. Figure 3a shows a conventional bright-field TEM image and its corresponding electron energy loss spectroscopy (EELS)

had the lowest swelling of all membranes with a water uptake that is marginally higher than co-SPI-r, whereas the water uptake rises dramatically for co-SPI-10. This suggests that a block length of more than 10 is necessary to induce significant microphase separation. All membranes in this study exhibited anisotropic membrane swelling with Δt/l > 1, i.e., larger through-plane than in-plane swelling. This finding has previously been attributed to the degree of polymer chain alignment in the in-plane direction of the membrane combined with the rigid structure of aromatic polyimides.12,23 Thermal stability of the membranes in their protonated form was investigated by TGA (data not shown). All polymers exhibited the typical three-step degradation process. The TGA data were similar for both random and block copolyimides, suggesting that the specific morphology does not have a large impact on thermal stability.11,12 The first weight loss up to 280 °C was due to the removal of absorbed water; the second weight loss up to 500 °C is ascribed to desulfonation of the polymer backbone. Above this temperature, the polymer backbone started to decompose (formation of CO2) and evaporates. No glass transition was detected up to 200 °C by DSC measurements. Membrane Morphology from Scanning Probe Techniques. Figure 2a,c,e displays the TEM bright-field images for membranes based co-SPI-r, co-SPI-20, and co-SPI-50. These images indicate that spherical clusters are uniformly dispersed throughout the membranes, pointing toward a channel-like morphology, rather than a layer-like structure. The domain size increases with increasing block length.66 For co-SPI-r

Figure 3. (a) TEM image of a membrane based on co-SPI-50 and (b) EELS mapping image of the same area showing the sulfur content. Bright yellow colors indicate high sulfur content corresponding to the ionic domains. The AFM images of membranes (c) co-SPI-50 and (d) co-SPI-r illustrate that the superimposed morphology is only present for co-SPI-50.

image in Figure 3b for the membrane based on co-SPI-50. Besides the observation of the characteristic phase separation on the nanometer scale (cf. Figure 2), Figure 3a,b also reveals a superimposed morphology on the micrometer scale, indicating a higher organization for extended block length. According to the EELS image in Figure 3b, showing the sulfur content of the membrane, the hydrophobic clusters are ∼1 μm in diameter and are embedded in the matrix with mainly ionic character. This explains the opaque appearance of the membrane based on co-SPI-50, since visible light is scattered on these density inhomogeneities, which sizes are on the order of the wavelength. No distinct morphology on the micrometer scale

Figure 2. AFM and TEM images of membranes based on (a, b) coSPI-r, (c, d) co-SPI-5, and (e, f) co-SPI-50. 2650


Macromolecules

Article

ratio for the water free polymer of 1:11 at both temperatures. If the membranes were not dried sufficiently, a very intense 1H resonance from water at ∼6.8 ppm is observed. Moreover, the 1 H chemical shift of the acidic SO3H protons shifts to higher values with increasing block length (see Figures 4c and 5). This

could be found for the other membranes. Accordingly, these membranes appear transparent to the eye. Hydrogen Bonding Monitored via 1H MAS NMR. To evaluate the structural impact of the enhanced water uptake on the hydrogen bonding with increasing block length of the multiblock co-SPIs, we have used 1H MAS NMR to determine if this results in an increase of the hydrophilic domains only or if the ionic character of hydrophilic domains in the membrane also grows. This evaluation relies on the fact that protons involved in hydrogen-bonded structures typically exhibit wellresolved 1H chemical shifts, allowing them to be identified via 1 H MAS NMR.67−69 A clear evidence of strong hydrogen bonding is a high-frequency shifted 1H resonance.70−73 Thus, it is reasonable to conclude that with increasing ionic character of hydrophilic domains for the multiblock co-SPIs the hydrogen bonding also becomes stronger. Figure 4 shows a series of

Figure 5. Plot of the 1H chemical shift as a function of temperature for the acidic SO3H protons in dried multiblock co-SPI membranes.

is a result of enhanced hydrogen bonding, pointing toward a better phase separation for longer block lengths, but most likely also a result of increased hydrophilicity differences as well as dimension and symmetry of the channels in the membranes. The lowest 1H chemical shift was found for the random copolyimide co-SPI-r (10.7 ppm at 320 K), whereas the highest shift was observed for fully sulfonated polyimide co-SPI-f (11.4 ppm). In fact, co-SPI-f can be considered as infinite, perfect ionic domain, since it does not contain any hydrophobic sequences. Hence, the 1H chemical shift of SO3H in co-SPI-f is expected to be the highest value that can be observed as the ionic interactions should be maximal. The co-SPI with the largest block length is co-SPI-50, and this also revealed the highest 1H chemical shift of 10.93 ppm (see Figure 4c), which is slightly lower than that found for co-SPI-f. This demonstrates that even for long block lengths, the phase separation between ionic and hydrophobic domains is far from being defined and sharp. Figure 5 shows the chemical shift of all co-SPIs as a function of temperature. At low temperatures, as in Figure 4a, no distinct intensity maxima can be observed for the acidic SO3H protons, and the 1H chemical shifts could for this reason only be determined by deconvolution of the experimental spectra. This leads to an estimated error of ±0.1 ppm for the random (co-SPI-r) and block copolyimides (co-SPI-5, co-SPI10, co-SPI-20, and co-SPI-50), whereas the error for fully sulfonated polyimide (co-SPI-f) is ±0.3 ppm. Figure 5 demonstrates that with increasing temperature the 1H chemical shift of SO3H decreases for all samples by 0.5−0.8 ppm over a temperature range of 80 K. The shift to lower frequencies is a direct result of weaker effective hydrogen bonding caused by thermal motion. Nevertheless, co-SPI-f maintains the highest 1 H chemical shift when dried at 400 K, followed by co-SPI-50, having the longest block length. Notably, DQ-filtered 1H MAS NMR spectra of dried membranes using one rotor period (50 μs) excitation/reconversion period did not include a SO3H proton resonance, while the conventional single-pulse 1H MAS NMR experiments did. The difference between the two experiments is ascribed to the fact that the DQ filter suppresses weak 1H−1H dipolar couplings that are averaged due to rapid

Figure 4. Deconvoluted 1H MAS NMR spectra of dried membranes based on co-SPI-20 at (a) 320 K and (b) 400 K. (c) Comparison of 1H MAS NMR spectra of fully sulfonated dried membranes of co-SPI-f, co-SPI-50, and co-SPI-r at 400 K. These spectra demonstrate that the 1 H chemical shift for the acidic SO3H protons depends on the type of membrane. The minor impurity signals in the low ppm region at ∼3 ppm most likely originate from side products formed during the polycondensation reaction.

solid-state 1H MAS NMR spectra for a number of co-SPI membranes dried at 100 °C in vacuo for 2 days. These reveal well-resolved, albeit broad 1H resonances from both acidic SO3H protons centered at ∼10−11 ppm and those of the aromatic groups of the polymer backbone at ∼7−8 ppm. From a comparison of Figure 4a,b, the ratio of the integrals for the acidic proton and polymer backbone is close to the theoretical 2651


Macromolecules

Article

irrespective of the degree of sulfonation or block length. However, co-SPI-20 and co-SPI-50 appear to have higher activation energies than co-SPI-5 and co-SPI-10 even though the phase separation (or ionic character) is better for co-SPIs with the higher block lengths (cf. Figure 3). This difference is further addressed below using 1H PFG NMR, demonstrating that the obtained activation energy for co-SPI-50 is just an apparent value as result of a superposition of two distinct 1H environments. We note that the activation energy range determined here for the co-SPIs is slightly lower than that found in dehydrated Diels−Alder material (∼18 kJ/mol),71 dried sulfonated poly(ether ketone)s (∼20 kJ/mol), and dried Nafion (∼16 kJ/mol).70 Moreover, recent PFG NMR experiments on Nafion further showed that the EA is strongly dependent on the specific hydration level and thereby the water:sulfonate ratio.75 It has to be stressed that the NMR activation energies determined via 1H line widths as performed in this work provide information about the local environment of the acidic protons only. Hence, from T2* data alone, no conclusions can be drawn about the overall proton conductivity, since this is a macroscopic property that strongly depends on morphological issues (e.g., vehicle transport).72 Proton Conductivity. Proton conductivity is extremely sensitive to the relative humidity (RH) of the environment and increases with increasing water content of the membrane. As presented in Figure 7, the multiblock co-SPI membranes

motional averaging. This effect is typical for sulfonated aromatic polymers, and it correlates well with the fact that the high mobility of the acidic protons is responsible for proton conduction.70,71 Proton Conduction. To learn more about proton mobility, the proton line width of the acidic SO3H resonance was studied as a function of temperature. For rapid molecular dynamics, the temperature dependence of the acidic proton spin−spin relaxation T2* is sensitive to the rate with which protons hop between different sites. If the exchange rate Ω is much larger than the frequency difference between the sites undergoing exchange, proton mobility on the nanoscale can be probed by determining proton hopping activation energies according to ⎛ −E ⎞ 1 = πT2* ∼ Ω = Ω 0 exp⎜ A ⎟ ⎝ RT ⎠ Δν

(2)

where Δν is the line width of the resonance and T*2 is the effective transverse relaxation time.72,74 Distinct from that, the aromatic proton resonance showed an almost invariable line width for all of the studied membranes, implying that the mobility of the polymer chains does not increase significantly until temperatures above 400 K. For sample co-SPI-f an initial narrowing of the acidic 1H resonance with increasing temperature is observed followed by a plateau value of ∼1100 Hz at 370 K. This value is the inherent 1 H line width that is no longer dominated by anisotropic NMR interactions, which can be averaged by increasing temperature. Hence, only data from the initial slope (320−370 K) were used to calculate the activation energies.72 As an example, Figure 6

Figure 7. Proton conductivity as a function of relative humidity at 50 °C of the synthesized co-SPI membranes and Nafion 117 determined from dielectric spectroscopy in a two-electrode in-plane geometry. Figure 6. Arrhenius plot of the temperature dependence for the relaxation time T*2 of the sulfonic acid resonance in a membrane based on co-SPI-20. Table 3 summarizes the activation energies for all other samples.

showed good proton conductivities approaching the values of Nafion 117 at high humidity. Above 60% RH, both random and block co-SPIs all exhibit approximately the same conductivity (>0.05 S/cm at 90% RH). At lower RH conditions, the performance strictly depends on the block length, where those with longer blocks show better conductivity values. For example, at 5% RH, the conductivity of co-SPI-50 is more than 12 times higher than that measured for the random copolyimide (4.6 × 10−6 S/cm vs 3.7 × 10−7 S/cm). This difference is most likely related to an enhanced microphase separation for membranes with longer blocks (see Figure 2).

shows the Arrhenius plot of −ln(T2*) for SO3H in co-SPI-20 as a function of temperature. Here, the best linear fit gives the activation energy directly as the slope; i.e., EA = 15.5 ± 1.0 kJ/ mol. Table 3 summarizes the activation energies for the other membranes. Interestingly, Table 3 reveals that the activation energies for all samples fall in the range 12−16 kJ/mol, Table 3. Activation Energies for the Proton Mobility in Co-SPIs sample EAa a

(kJ/mol)

co-SPI-r

co-SPI-5

co-SPI-10

co-SPI-20

co-SPI-50

co-SPI-f

15.4 ± 0.8

12.9 ± 1.5

12.8 ± 1.0

15.5 ± 1.0

14.0 ± 0.6

12.4 ± 1.2

The error is given as the standard deviation of the fit for the activation energy. 2652


Macromolecules

Article

where CQ is the quadrupole coupling constant (∼260 kHz for a rigid O−D bond)78 and S the oriental order parameter of the O−D bond with respect to the alignment axis of the material.79 Smatrix is the order parameter of the channel network matrix, whereas the interaction of the host matrix and the water molecules depends on several complex factors, such as channel size, nature of the network surface, etc. These contributions are summarized in the scaling factor ρ. The second-order Legendre polynomial P2(cos θ) = 1/2(3 cos2 θ − 1) describes the angular dependence of the 2H quadrupole splitting with respect to the magnetic field B0.39 Equation 3 can be used to fit the experimentally determined ΔνQ as a function of the angle θ, where the prefactor Δν0 is the maximum splitting. Figure 9 displays a series of solid-state 2H NMR spectra for a fully hydrated membrane of co-SPI-5 for three selected

Two main criteria can be defined how improved phase separation can lead to a better proton transport: (i) welldeveloped ionic percolation pathways within the hydrophobic matrix and (ii) better water retention at low relative humidity.76 To test these criteria, Figure 8 displays the water uptake for co-

Figure 8. Water uptake as a function of the relative humidity for membranes based on co-SPI-50 and co-SPI-r at 50 °C. The water uptake at a relative humidity 0% was determined after storing the membrane overnight at 100 °C in vacuo. All other relative humidities were adjusted in a climatic chamber.

SPI-r and co-SPI-50 determined at 50 °C. Here, we have defined 0% water uptake as the state of the membrane after being kept under vacuum at 100 °C overnight. At 5% RH, coSPI-50 adsorbs almost twice the amount of water compared to co-SPI-r. At higher relative humidities the water uptake for both membranes grows in step with each other, leading to a smaller relative difference. Hence, the relative difference in water uptake of membranes based co-SPI-50 compared to co-SPI-r is highest at low relative humidity; i.e., exactly where the strong dependence of conductivity on block length is observed and a characteristic feature for most sulfonated block copolyimides.19,24 Thus, the better performance in terms of proton conductivity for the multiblock co-SPIs is to some extent due to enhanced water absorption under dry conditions, although the contribution from percolation effects cannot be excluded. Hydrophilic Channel Alignment from 2H NMR. The degree of alignment, or equivalently the anisotropy, of the hydrophilic channels in the co-SPI membranes was assessed by 2 H NMR, examining the quadrupole splitting (ΔνQ) observed in 2H NMR spectroscopy of absorbed D2O. For an isotropic system, such as liquid water, a single 2H resonance for D2O is observed due to fast molecular reorientations of the D2O molecules. Nonisotropic morphology on the nanometer scale, e.g., aligned hydrophilic channels or stacked lamellar layers, leads to only partial averaging of the quadrupole coupling and results in a splitting of the 2H resonance. This quadrupole splitting, ΔνQ, depends on the anisotropy of the environment and the direction of the material alignment axis relative to the externally applied magnetic field. For a uniaxial system with ηQ = 0, the quadrupole splitting ΔνQ may be written as51,53,77 1 ΔνQ = CQ SP2(cos θ ) = CQ Smatrixρ (3 cos2 θ − 1) 2 1 2 = Δν0(3 cos θ − 1) (3) 2

Figure 9. Room temperature 2H NMR spectra of a fully hydrated membrane based on co-SPI-5. θ defines the angle between the material alignment axis, which is along the through-plane direction (red arrow), and the magnetic field B0.

orientations with respect to the magnetic field direction B0. When the membrane through-plane axis is oriented perpendicular to B0, corresponding to θ = 90°, a splitting of ΔνQ = 252 Hz is observed. This is half the size of ΔνQ when the throughplane axis is parallel to B0 with θ = 0° (ΔνQ = 505 Hz). At θ = 54.7°, corresponding to the angle where the second-order Legendre polynomial P2(cos θ) is zero (also known in solidstate NMR as the magic angle), the doublet collapses into a single resonance as expected if the hydrophilic channels are oriented parallel to the through-plane direction (see Figure 9). Least-squares optimization of eq 3 to the experimental data points (in red) of θ ranging from 0° to 90° shown in Figure 10 gives the solid black fitted line with Δν0 = 505 Hz. This yields a partial ordering S of D2O molecules of ∼10−3 in the fully hydrated membrane of co-SPI-5, illustrating that the length scale of the hydrophilic channels is on the order of a few nanometers.80 To estimate the length scale probed by the 2H NMR measurements, we can use the random walk expression ⟨r2⟩1/2 = (2D/ΔνQ)1/2, where D is the water diffusion coefficient (≈ 5 × 10−11 m2/s, see below) and ⟨r2⟩1/2 the 2653


Macromolecules

Article

water uptake as well as the absolute intensity of 2H NMR experiments on fully saturated membranes at room temperature is known (Table 2), the intensity of 2H NMR experiments on partially hydrated membranes can be employed to calculate the water uptake. The 2H quadrupole splitting dependence on the water uptake is summarized for all co-SPI membranes in Figure 11a, where Figure 11b illustrates the relative dependence. In

Figure 10. Plot of the 2H quadrupole splitting ΔνQ for D2O as a function of the angle θ for a fully hydrated membrane based on coSPI-5. The solid black line represents the fitted curve using eq 3.

diffusion length.81 This gives a diffusion length of ∼0.5 μm for ΔνQ = 500 Hz that is significantly larger than the channel length estimated from TEM images. The difference in apparent length scale is caused by the fact that the 2H quadrupole splitting does not probe the channel dimensions directly. Rather, it is a measure of the average bulk anisotropy, since each water molecule is sampling a large number of channels over the ∼0.5 μm diffusion length, reflecting the average channel ordering combined with the degree of coupling to the channels (via the scaling factor ρ, see eq 3). The observation of a finite 2H quadrupole splitting proves that the hydrophilic channels must be orientationally ordered (Smatrix is nonzero).80 The black curve shown in Figure 10, representing a fit to the second-order Legendre polynomial only, fits the experimental data points very well. This shows that the hydrophilic channels exhibit uniaxial orientation (or alignment) in the through-plane direction of the membrane as expected for solution-cast membranes. We note that it has been proposed that the alignment of proton conducting channels arises from solvent evaporating in the direction normal to the membrane plane during the casting process and possibly gets even further pronounced via additional processes during annealing.82 Further control over the direction and extent of orientational order of the hydrophilic channels can increase the proton conductivity. For example, it has been demonstrated that Nafion 112 and Nafion 117 (extruded) show axially oriented channels along the extrusion direction, whereas the channels in Nafion NRE212 (dispersion-cast) are oriented perpendicular to the membrane plane.83 However, also uniaxial elongation of cast membranes may lead to significant orientation and strongly anisotropic morphology.84 Water uptake and relative humidity are crucial parameters when determining the properties of a proton exchange membrane; e.g., the proton conductivity may vary over several orders of magnitude when changing the relative humidity as demonstrated in Figure 7. Thus, quantitative analysis of diffusion coefficients, as well as 2H quadrupole splittings, is debatable and nonreproducible if no control over water content is achieved. To this end both sealed and low-dead-volume sample cells are indispensable.51 The Teflon sample cell depicted in Figure 1a meets these requirements. The proportionality between the 2H NMR signal intensity for sufficient long recycle delays and the D2O content of the membrane was used to determine the level of hydration. As

Figure 11. Room temperature 2H quadrupole splittings plotted as (a) absolute values and (b) relative changes in percent as a function of water uptake.

general, the quadrupole splitting ΔνQ increases as the D2O content of the membrane decreases (Figure 11a), which can be attributed to an increased anisotropic interaction of D2O (e.g., caused by restricted rotations) with the channel walls. This effect originates from a shrinking of the average hydrophilic domain diameter upon loss of water and is consistent with the observation of water uptake-dependent ΔνQ for Nafion membranes51,84 and sulfonated poly(arylene ether sulfone) block copolymers.53 Specifically, it is the scaling factor ρ that increases with decreasing hydrophilic channel diameter as CQ and Smatrix should not vary much with water content (eq 3). Similarly, ΔνQ decreases by 0.5−1.0 Hz per kelvin with increasing temperature (data not shown). This is most likely due to thermal expansion of the hydrophilic channels and enhanced isotropic D2O rotation induced by higher kinetic energy of the water molecules.51 Figure 11b illustrates that the longer the block length of the co-SPI, the higher the relative increase in ΔνQ if the water uptake is lowered. Such a phenomenon has already been observed for sulfonated block copolymers and may be related to a faster, relative shrinking of hydrophilic channels for larger domains (see Figure 2).53 Even for the random copolymer co-SPI-r a small quadrupole splitting is observed, which may be attributed to finite blocks even in 2654


Macromolecules

Article

this polymer. A simple, intuitive explanation for the differences in absolute splitting is however not straightforward, since ΔνQ for different batches of membranes with formally the same block length varied considerably for identical water uptake. For example, the maximum ΔνQ for a saturated membrane with a formal block length of 10 was 470, 675, and 900 Hz for three different batches. Notably, conductivity measurements of these batches show virtually identical values. This suggests that very small differences in the membrane casting procedure and/or polymer characteristics (block length, IEC, polydispersity, etc.) may have a tremendous effect on the interactions at the molecular scale, even though macroscopic features such as the conductivity remain the same, keeping in mind that the length scale probed in the 2H NMR experiment is ∼0.5 μm (see above) and conductivity measurements depend on the whole sample (several micrometers). Water Diffusion Anisotropy from 1H PFG NMR. By applying a magnetic-field gradient with strength g for a period δ, molecules can be labeled depending on their position in a sample. After a diffusion time Δ, the new position of the spatially labeled molecule is decoded via a second gradient of the same strength and duration. The NMR signal attenuation caused by diffusion is described by the Stejskal−Tanner equation42 2 2 2

I = I0e−Dγ g

δ (Δ− δ /3)

= I0e−Db

(4)

,where I is the spin−echo signal intensity, I0 is the signal intensity with no gradient applied, γ is the gyromagnetic ratio of the probe nucleus, and D is the self-diffusion coefficient of the observed molecule. For reasons of simplification, γ2g2δ2(Δ − δ/ 3) is summarized to b. Practically, it is impossible to generate perfect rectangular gradient pulses, as extremely high currents are required, leading to irreproducible pulses. Hence, sineshaped pulses are applied. Furthermore, we have used the pulse-gradient stimulated echo (PGSTE) sequence, since this mainly depends on T1 instead of T2 (T1 > T2).85 Self-diffusion coefficients D were determined by evaluating ln(I/I0) as a function of b, i.e., incrementing the gradient g in linear steps with each experiment. In case of a monomodal intensity decay, where all water molecules have the same diffusion coefficient, eq 4 predicts a linear relationship with a slope D. Similar to the 2 H NMR results presented above, control over water content is a prerequisite for 1H PFG NMR experiments on proton exchange membranes. As described in the Experimental Section, tight-fitting caps with low dead volume were used to minimize water evaporation once the cell is put out of the climatic chamber and into the NMR tube. Figure 12a shows the plot of ln(I/I0) vs b for two membranes based on co-SPI-r and co-SPI-50 together with their respective optimized linear fits. Clearly, the expected straight line is obtained for co-SPI-r, but not for co-SPI-50. If the same experimental data for co-SPI-50 are plotted with the intensity as a function of the gradient strength g (Figure 12b), it becomes evident that the experimental data and monomodal fit deviate not only for large values of g but also for low values, corresponding to the initial part of the experiment. However, applying a superposition of two exponentials (bimodal) yields satisfying results for the diffusion data of co-SPI-50. Physically, this corresponds to two distinct diffusion processes whose NMR signals overlap, albeit with two different diffusion coefficients. As the only diffusing species is H2O, the difference in D must be a results of two different environments within the

Figure 12. (a) 1H PFG NMR data for membranes based on co-SPI-r and co-SPI-50 at a relative humidity of 60% and 25 °C. The data are plotted as the normalized signal amplitudes (ln(I/I0)) vs b (see eq 4). Solid lines represent best fits assuming a monoexponential decay. (b) Plot of the signal intensity as a function of gradient strength g for coSPI-50, illustrating the deviations from a monoexponential fit (yellow). The biexponential fit (black) has two components: a fast with D1 = 5.66 × 10−11 m2/s (blue) and a slow with D2 = 1.85 × 10−11 m2/s (green). The relative proportions are 73% and 27%, respectively.

membrane, where water diffuses fast in one domain (D1) and slow (D2) in the other; i.e., D1 > D2. Moreover, it can be seen that the fast process (D1) accounts for the steep intensity decay for small values of g, whereas the slow diffusion process (D2) explains the relatively slow decay of intensity for strong gradients. The observation of two different diffusion coefficients compares well with the inhomogeneities on the micrometer scale found for membrane co-SPI-50 using TEM and EELS (see Figure 3). Water in the sulfur-rich domains is expected to diffuse faster due to the more ionic character of the environment. In contrast, the less sulfur-rich environment offers water molecules a more hydrophobic diffusion pathway that hinders a fast passage, resulting in a lower diffusion coefficient. The length scale of a 1H PFG NMR experiment can be determined as ⟨r2⟩1/2 = (2DΔ)1/2.81 Here, Δ is the diffusion time (10 ms), and with D = 5 × 10−11 m2/s, we obtain a length scale ⟨r2⟩1/2 of ∼1 μm, which again is in good agreement with the length scale of the superimposed morphology found for coSPI-50 (see Figure 3). However, since co-SPI-50 is the only membrane with observable micrometer scale phase separation, all other membranes are expected to exhibit 1H PFG NMR curves that can be fitted with a single diffusion coefficient. Indeed, 1H PFG NMR data of co-SPI-50 are the only requiring a bimodal fit, irrespective of the relative humidity conditions (see below). 2655


Macromolecules

Article

Figure 13 summarizes the results of 1H PFG NMR experiments performed using the cell shown in Figure 1 for

strated in Figure 13a,b, where the diffusion coefficients rise with increasing block length in the order co-SPI-r, co-SPI-5, and coSPI-10, before a drop for co-SPI-20 is observed. However, the conductivity is higher for co-SPI-20 than for membranes with smaller block length (see Figure 5). We speculate that even though the local environment is unfavorable for proton diffusion in co-SPI-20, a very effective long-range percolation network might be present in this membrane in addition to a higher water uptake for a given relative humidity (see Figure 8), both compensating for the comparable slow small-range diffusion. This is supported by the data of Table 3. In analogy to the 1H PFG NMR data, the analysis of temperaturedependent 1H line widths is also a local probe of proton mobility and a comparable high activation energy EA for the hopping of acidic protons is found for co-SPI-20 also pointing toward an unfavorable local environment. As explained above, two diffusion processes could be identified for co-SPI-50: a fast one (D1) and a slow one (D2). D1 is higher than the diffusion coefficient of the other membranes, whereas D2 is lower over the whole relative humidity range. This is consistent with the 1 H MAS NMR data (Figure 4), which suggests that hydrophilic domains in co-SPI-50 have the highest ionic character, owing to a more pronounced phase separation. The calculated contribution of the fast component D1 to the overall signal intensity is about 2/3 for measured relative humidities. We note that the opaqueness of co-SPI-50 is not related to the formation of large crystallites in the membrane, as solid-state 13C MAS NMR spectra are virtually identical for all membranes. A significant narrowing of the 13C resonances would be expected for crystalline regions. A slight diffusion anisotropy, defined as the ratio of throughplane to in-plane diffusion coefficient, is observed for all membranes (cf. Figure 13a,b). Assuming a channel-like morphology as proposed by TEM and AFM images (Figure 2), and taking into account the results of 2H NMR, the hydrophilic channels are oriented perpendicular to the membrane plane. The diffusion coefficients were found to be highest for in-plane diffusion (Figure 13a). We attribute this to a lamella-like structure composed of rather short channels in our systems as similar diffusion anisotropies in PEMs with lamellar structures have been reported previously.51−53 Through-plane translational diffusion on the micrometer length scale requires water molecules to hop across several short, possibly nonaligned hydrophilic channels. Such a diffusion process is more likely to happen between adjacent channels aligned parallel to each other rather than to channels in the next layer above. Indeed, in related molecular and supramolecular systems, like smectic liquid crystals and columnar discotics, similar anisotropic diffusion has been observed.88,89 Thus, in the systems studied here, which exhibit limited phase separation, the usual conjecture that channel alignment in through-plane direction is desirable may not be justified.

Figure 13. Water diffusion coefficients for (a) in-plane and (b) through-plane as a function of the relative humidity for all synthesized co-SPI membranes. All experiments were performed at 25 °C. The annotations fast and slow for membrane co-SPI-50 denote that the deconvolution of the experimental data included two processes.

in- and through-plane diffusion measurements for all co-SPI membranes at relative humidities of 19%, 30%, 45%, and 60%. For all membranes, regardless whether in-plane or throughplane, the water diffusion coefficients can be seen to increase with increasing relative humidity. At 19% relative humidity, the diffusion coefficient is below 10−12 m2/s and rises to ∼5 × 10−11 m2/s at 60% relative humidity. This is due to an increasing water uptake that swells the hydrophilic channels in the membrane and facilitates proton transport. Diffusion coefficients of absorbed water in Nafion and sulfonated block poly(sulfones) were reported to be 1 order of magnitude higher,51,53 indicating that the local environment for water transport is more favorable in these materials compared to that of the multiblock co-SPIs investigated here. The self-diffusion coefficient of pure water is reported to be 2.30 × 10−9 m2/s at 25 °C.86 This value is 2 orders of magnitude higher than those measured here for the co-SPIs membranes, suggesting that diffusion of water in the hydrophilic channels is highly restricted by its surroundings.87 Nevertheless, it is not possible to draw conclusions about the proton transport mechanism responsible for the conductivity based on the values of diffusion coefficients alone. Even though diffusion processes should be related to conductivity due to the required motion of water in both cases, the length scale of both experiments is different making a direct correlation impossible. This fact is demon-

■

CONCLUSIONS A series of sulfonated block copolyimides were synthesized using a two-step procedure. The observed differences in performance between the membranes could solely be ascribed to differences in block length as the IEC was held constant and only the block length was varied from 5 to 50 repeat units. Proton conductivity increased with increasing block length, where, especially at low relative humidity conditions, the performance of a random copolyimide is inferior to its block counterparts. Water uptake measurements revealed that to 2656


Macromolecules

Article

(16) Guo, X.; Fang, J.; Tanaka, K.; Kita, H.; Okamoto, K.-I. J. Polym. Sci., Part A: Polym. Chem. 2004, 42, 1432. (17) Yin, Y.; Yamada, O.; Tanaka, K.; Okamoto, K.-I. Polym. J. 2006, 38, 197. (18) Fang, J.; Guo, X.; Harada, S.; Watari, T.; Tanaka, K.; Kita, H.; Okamoto, K. Macromolecules 2002, 35, 9022. (19) Asano, N.; Miyatake, K.; Watanabe, M. J. Polym. Sci., Part A: Polym. Chem. 2006, 44, 2744. (20) Yaguchi, K.; Chen, K.; Endo, N.; Higa, M.; Okamoto, K.-I. J. Power Sources 2010, 195, 4676. (21) Guo, X.; Fang, J.; Watari, T.; Tanaka, K.; Kita, H.; Okamoto, K. Macromolecules 2002, 35, 6707. (22) Chen, K.; Hu, Z.; Endo, N.; Higa, M.; Okamoto, K.-I. Polymer 2011, 52, 2255. (23) Hu, Z.; Yin, Y.; Yaguchi, K.; Endo, N.; Higa, M.; Okamoto, K.-i. Polymer 2009, 50, 2933. (24) Nakano, T.; Nagaoka, S.; Kawakami, H. Polym. Adv. Technol. 2005, 16, 753. (25) Yan, J.; Liu, C.; Wang, Z.; Xing, W.; Ding, M. Polymer 2007, 48, 6210. (26) Kreuer, K. D. J. Membr. Sci. 2001, 185, 29. (27) Jutemar, E. P.; Jannasch, P. J. Membr. Sci. 2010, 351, 87. (28) Elabd, Y. A.; Napadensky, E.; Walker, C. W.; Winey, K. I. Macromolecules 2006, 39, 399. (29) Zhao, C.; Li, X.; Wang, Z.; Dou, Z.; Zhong, S.; Na, H. J. Membr. Sci. 2006, 280, 643. (30) Bae, B.; Miyatake, K.; Watanabe, M. Macromolecules 2009, 42, 1873. (31) Chen, L.; Hallinan, D. T.; Elabd, Y. A.; Hillmyer, M. A. Macromolecules 2009, 42, 6075. (32) Soboleva, T.; Xie, Z.; Shi, Z.; Tsang, E.; Navessin, T.; Holdcroft, S. J. Electroanal. Chem. 2008, 622, 145. (33) Badami, A. S.; Lane, O.; Lee, H.-S.; Roy, A.; McGrath, J. E. J. Membr. Sci. 2009, 333, 1. (34) Matsumoto, K.; Higashihara, T.; Ueda, M. Macromolecules 2009, 42, 1161. (35) Roy, A.; Hickner, M. A.; Yu, X.; Li, Y.; Glass, T. E.; McGrath, J. E. J. Polym. Sci., Part B: Polym. Phys. 2006, 44, 2226. (36) Park, M. J.; Balsara, N. P. Macromolecules 2010, 43, 292. (37) Yamada, O.; Yin, Y.; Tanaka, K.; Kita, H.; Okamoto, K.-I. Electrochim. Acta 2005, 50, 2655. (38) Kimmich, R. NMR: Tomography, Diffusometry, Relaxometry; Springer: Berlin, 1997. (39) Schmidt-Rohr, K.; Spiess, H. W. Multidimensional Solid-State NMR and Polymers; Academic Press: New York, 1994. (40) Cohen, Y.; Avram, L.; Frish, L. Angew. Chem., Int. Ed. 2005, 44, 520. (41) Price, W. S. Mod. Magn. Reson. 2006, 1, 105. (42) Stejskal, E. O.; Tanner, J. E. J. Chem. Phys. 1965, 42, 288. (43) Ohkubo, T.; Kidena, K.; Ohira, A. Macromolecules 2008, 41, 8688. (44) Kidena, K. J. Membr. Sci. 2008, 323, 201. (45) Smith, R. L.; Oldfield, E. Science 1984, 225, 280. (46) Toriumi, H.; Deloche, B.; Herz, J.; Samulski, E. T. Macromolecules 1985, 18, 304. (47) Deloche, B.; Samulski, E. T. Macromolecules 1981, 14, 575. (48) Matsumura, K.; Hayamizu, K.; Nakane, T.; Yanagishita, H.; Yamamoto, O. J. Polym. Sci., Part B: Polym. Phys. 1987, 25, 2149. (49) Matsumura, K.; Hayamizu, K.; Nakane, T.; Yanagishita, H.; Yamamoto, O. J. Polym. Sci., Part B: Polym. Phys. 1988, 26, 2215. (50) Migchelsen, C.; Berendsen, H. J. C. J. Chem. Phys. 1973, 59, 296. (51) Li, J.; Wilmsmeyer, K. G.; Madsen, L. A. Macromolecules 2009, 42, 255. (52) Li, J.; Park, J. K.; Moore, R. B.; Madsen, L. A. Nat. Mater. 2011, 10, 507. (53) Hou, J.; Li, J.; Madsen, L. A. Macromolecules 2010, 43, 347. (54) Rankothge, M.; Haryadi; Moran, G.; Hook, J.; Van Gorkom, L. Solid State Ionics 1994, 67, 241. (55) Xu, G.; Pak, Y. S. Solid State Ionics 1992, 50, 339.

some extent this is due to the good water retaining properties of the block copolymers. The morphology was assessed by TEM and AFM measurements. Both techniques revealed a growing domain size with increasing block length, pointing toward a channel-like morphology. For the longest block copolyimides (co-SPI-50), the hydrophilic character of the phase-separated, ionic channels was examined by solid-state NMR. This demonstrated that the quality of the phase separation rises with growing domain size. Swelling experiments indicated an anisotropic morphology for all membranes. Combination of 1H PFG NMR and 2H NMR proved pronounced locally anisotropic structures and gave insight into domain alignment. 2H NMR further showed that the water molecules exhibit pronounced locally anisotropic motion along the axis perpendicular to the membrane plane, where translational diffusion of the water molecules, however, is found to be faster in the membrane plane.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected] (M.R.H.). Notes

The authors declare no competing financial interest. C.F.K.: Based on Part II of the doctoral thesis entitled “Synthesis and Investigation of Functional Polymer Materials” by Christoph Kins, available online http://ubm.opus.hbz-nrw. de/volltexte/2012/3139/.

■

ACKNOWLEDGMENTS We thank G. Brunklaus for discussions and the anonymous reviewers for helpful suggestions. M.R.H. acknowledges financial support from the Villum Foundation under the Young Investigator Programme (VKR023122).

■

REFERENCES

(1) Joerissen, L. In Hydrogen and Fuel Cells Fundamentals, Technologies and Applications; Edited by Stolten, D., Ed.; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim, 2011; Vol. 50. (2) Mauritz, K. A.; Moore, R. B. Chem. Rev. 2004, 104, 4535. (3) Schmidt-Rohr, K.; Chen, Q. Nat. Mater. 2008, 7, 75. (4) Peckham, T. J.; Holdcroft, S. Adv. Mater. 2010, 22, 4667. (5) Marestin, C.; Gebel, G.; Diat, O.; Mercier, R. Adv. Polym. Sci. 2008, 216, 185. (6) Yamazaki, K.; Tang, Y.; Kawakami, H. J. Membr. Sci. 2010, 362, 234. (7) Faure, S.; Cornet, N.; Gebel, G.; Mercier, R.; Pineri, M.; Sillion, B. New Mater. Fuel Cell Mod. Battery Syst. II, Proc. Int. Symp., 2nd 1997, 818. (8) Zhang, F.; Li, N.; Cui, Z.; Zhang, S.; Li, S. J. Membr. Sci. 2008, 314, 24. (9) Li, N.; Cui, Z.; Zhang, S.; Xing, W. Polymer 2007, 48, 7255. (10) Chen, X.; Chen, K.; Chen, P.; Higa, M.; Okamoto, K.-I.; Hirano, T. J. Polym. Sci., Part A: Polym. Chem. 2009, 48, 905. (11) Li, N.; Liu, J.; Cui, Z.; Zhang, S.; Xing, W. Polymer 2009, 50, 4505. (12) Li, N.; Cui, Z.; Zhang, S.; Li, S. J. Polym. Sci., Part A: Polym. Chem. 2008, 46, 2820. (13) Meyer, G.; Gebel, G.; Gonon, L.; Capron, P.; Marscaq, D.; Marestin, C.; Mercier, R. J. Power Sources 2006, 157, 293. (14) Endo, N.; Matsuda, K.; Yaguchi, K.; Hu, Z.; Chen, K.; Higa, M.; Okamoto, K. J. Electrochem. Soc. 2009, 156, B628. (15) Sutou, Y.; Yin, Y.; Hu, Z.; Chen, S.; Kita, H.; Okamoto, K.-I.; Wang, H.; Kawasato, H. J. Polym. Sci., Part A: Polym. Chem. 2009, 47, 1463. 2657


Macromolecules

Article

(56) Chen, R. S.; Jayakody, J. P.; Greenbaum, S. G.; Pak, Y. S.; Xu, G.; McLin, M. G.; Fontanella, J. J. J. Electrochem. Soc. 1993, 140, 889. (57) Kaltbeitzel, A.; Schauff, S.; Steininger, H.; Bingol, B.; Brunklaus, G.; Meyer, W. H.; Spiess, H. W. Solid State Ionics 2007, 178, 469. (58) Jeanguillaume, C.; Trebbia, P.; Colliex, C. Ultramicroscopy 1978, 3, 237. (59) Muntean, J.; Stock, L.; Botto, R. J. Magn. Reson. 1988, 76, 540. (60) Langer, B.; Schnell, I.; Spiess, H. W.; Grimmer, A.-R. d. J. Magn. Reson. 1999, 138, 182. (61) Holz, M.; Weingaertner, H. J. Magn. Reson. 1991, 92, 115. (62) Zick, K. Diffusion NMR User Manual (Version 003), Bruker BioSpin GmbH, 2009. (63) Massiot, D.; et al. Magn. Reson. Chem. 2002, 40, 70. (64) Lienkamp, K.; Kins, C. F.; Alfred, S. F.; Madkour, A. E.; Tew, G. N. J. Polym. Sci., Part A: Polym. Chem. 2009, 47, 1266. (65) Choi, P.; Jalani, N. H.; Datta, R. J. Electrochem. Soc. 2005, 152, E123. (66) Bates, F. S.; Fredrickson, G. H. Annu. Rev. Phys. Chem. 1990, 41, 525. (67) Chierotti, M. R.; Gobetto, R. Chem. Commun. 2008, 1621. (68) Bohle, A.; Brunklaus, G.; Hansen, M. R.; Schleuss, T. W.; Kilbinger, A. F. M.; Seltmann, J.; Spiess, H. W. Macromolecules 2010, 43, 4978. (69) Shu, J.; Dudenko, D.; Esmaeili, M.; Park, J. H.; Puniredd, S. R.; Chang, J. Y.; Breiby, D. W.; Pisula, W.; Hansen, M. R. J. Am. Chem. Soc. 2013, 135, 11075. (70) Ye, G.; Janzen, N.; Goward, G. R. Macromolecules 2006, 39, 3283. (71) Cherry, B. R.; Fujimoto, C. H.; Cornelius, C. J.; Alam, T. M. Macromolecules 2005, 38, 1201. (72) Goward, G. R.; Schuster, M. F. H.; Sebastiani, D.; Schnell, I.; Spiess, H. W. J. Phys. Chem. B 2002, 106, 9322. (73) Schmidt, J.; Hoffmann, A.; Spiess, H. W.; Sebastiani, D. J. Phys. Chem. B 2006, 110, 23204. (74) Slichter, C. P. Principles of Magnetic Resonance, 3rd ed.; SpringerVerlag: Berlin, 1990. (75) Lingwood, M. D.; Zhang, Z.; Kidd, B. E.; McCreary, K. B.; Hou, J.; Madsen, L. A. Chem. Commun. 2013, 49, 4283. (76) Wu, L.; Zhang, Z.; Ran, J.; Zhou, D.; Li, C.; Xu, T. Phys. Chem. Chem. Phys. 2013, 15, 4870. (77) Callaghan, P. T.; Samulski, E. T. Macromolecules 2003, 36, 724. (78) Hauch, A.; Bildsøe, H.; Jakobsen, H. J.; Skibsted, J. J. Magn. Reson. 2003, 165, 282. (79) Hansen, M. R.; Graf, R.; Spiess, H. W. Acc. Chem. Res. 2013, 46, 1996. (80) Terzis, A. F.; Snee, P. T.; Samulski, E. T. Chem. Phys. Lett. 1997, 264, 481. (81) Callaghan, P. T. Principles of Nuclear Magnetic Resonance Microscopy; Clarendon Press: Oxford, 1991. (82) Li, J.; Wilmsmeyer, K. G.; Madsen, L. A. Macromolecules 2008, 41, 4555. (83) Hou, J.; Li, J.; Wilmsmeyer, K. G.; Zhang, Z.; Madsen, L. A. In NMR Spectroscopy of Polymers: Innovative Strategies for Complex Macromolecules; American Chemical Society: Washington, DC, 2011; Vol. 1077, p 251. (84) Park, J. K.; Li, J.; Divoux, G. M.; Madsen, L. A.; Moore, R. B. Macromolecules 2011, 44, 5701. (85) Tanner, J. E. J. Chem. Phys. 1970, 52, 2523. (86) Mills, R. J. Phys. Chem. 1973, 77, 685. (87) Hou, J.; Li, J.; Mountz, D.; Hull, M.; Madsen, L. A. J. Membr. Sci. 2013, 448, 292. (88) Dong, R. Y.; Goldfarb, D.; Moseley, M. E.; Luz, Z. J. Phys. Chem. 1984, 88, 3148. (89) Noack, F. In Handbook of Liquid Crystals, 1st ed.; Demus, D., Goodby, J., Gray, G., Spiess, H. W., Vill, V., Eds.; Wiley-VCH Verlag GmbH: Weinheim, 1998; Vol. 1, Chapter VII.13.

2658


Morphological Anisotropy and Proton Conduction in Multiblock

Recommend Documents