Thermodynamics of Counterion Release Is Critical for Anion

Jun 14, 2018 - Hickner, M. A.; Herring, A. M.; Coughlin, E. B. Anion Exchange Membranes: Current ...... Wehlin, Troian-Gautier, Sampaio, Marcélis, an...
1 downloads 0 Views 1MB Size
Article Cite This: J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

pubs.acs.org/JACS

Thermodynamics of Counterion Release Is Critical for Anion Exchange Membrane Conductivity Michael T. Kwasny,† Liang Zhu,‡ Michael A. Hickner,‡ and Gregory N. Tew*,† †

Downloaded via KAOHSIUNG MEDICAL UNIV on June 14, 2018 at 20:01:24 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Department of Polymer Science and Engineering, University of Massachusetts Amherst, Amherst, Massachusetts 01003, United States ‡ Department of Materials Science and Engineering, The Pennsylvania State University, University Park, Pennsylvania 16802, United States S Supporting Information *

ABSTRACT: As the field of anion exchange membranes (AEMs) employs an increasing variety of cations, a critical understanding of cation properties must be obtained, especially as they relate to membrane ion conductivity. Here, to elucidate such properties, metal cation-based AEMs, featuring bis(norbornene) nickel, ruthenium, or cobalt complexes, were synthesized and characterized. In addition, isothermal titration calorimetry (ITC) was used to probe counterion exchange thermodynamics in order to understand previously reported differences in conductivity. The ion conductivity data reported here further demonstrated that nickel-complex cations had higher conductivity as compared to their ruthenium and cobalt counterparts. Surprisingly, bulk hydration number, ion concentration, ion exchange capacity, and activation energy were not sufficient to explain differences in conductivity, so the thermodynamics of metal cation−counterion association were explored using ITC. Specifically, for the nickel cation as compared to the other two metal-based cations, a larger thermodynamic driving force for chloride counterion release was observed, shown through a smaller ΔHtot for counterion exchange, which indicated weaker cation−counterion association. The use of ITC to study cation−counterion association was further exemplified by characterizing more traditional AEM cations, such as quaternary ammoniums and an imidazolium cation, which demonstrated small variances in their enthalpic response, but an overall ΔHtot similar to that of the nickel-based cation. The cation hydration, rather than its hydration shell or the bulk hydration of the membrane, likely played the key role in determining the strength of the initial cation−counterion pair. This report identifies for the first time how ITC can be used to experimentally determine thermodynamic quantities that are key parameters for understanding and predicting conductivity in AEMs.



INTRODUCTION Producing highly efficient and chemically stable anion exchange membranes (AEMs), as cost-effective replacements for proton exchange membranes (PEMs), is a significant challenge due to an incomplete understanding of the fundamental properties and parameters that lead to enhanced performance in AEMs.1 Optimizing AEM performance is critical for promoting fuel cells’ widespread adoption as current fuel cell technologies, which utilize PEMs, require costly precious metal catalysts and exhibit poor performance across a range of operating conditions.2−10 Though AEMs currently offer several advantages over their PEM counterparts, including cheaper fabrication, improved oxygen reduction kinetics, and a wider range of available fuels, most AEMs still suffer from poor cation chemical stability and a generally low ion conductivity, although some recent reports have shown improvements in one or both of these areas.1,11−17 Although both limitations impact AEM performance, cation chemical instability, which arises from the degradation of the © XXXX American Chemical Society

cation under alkaline conditions, has garnered more attention, resulting in the synthesis of a number of different cationic species, the most common of which are summarized in Figure 1.18,19 The main categories of cations demonstrated in AEMs include ammonium-, 20−25 phosphonium-, 26−28 imidazolium-,21,24,29−31 sulfonium-,32,33 and metal-complex species.34−39 Some cations, namely the quaternary ammonium, imidazolium, and phosphonium examples, have also been functionalized with a variety of different moieties to further improve their alkaline stability. However, changing the functional groups around the cation also influences how the cations interact with their counterions and the surrounding water, resulting in significant modulation of ion exchange capacity (IEC), water uptake, and ion conductivity.22,24,26−28 Given the effects of the cation functional group on membrane properties and the massive library of potential cations, Received: April 13, 2018

A

DOI: 10.1021/jacs.8b03979 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

Figure 1. Chemical structures of various cations studied for use in AEM applications, where structures in black are amine-based cations, red are sulfur-based cations, green are phosphorus-based cations, and blue are metal-complex cations.

Table 1. Summary Comparing Reports of Different Cations Highlighting the Difficulty in Comparing between Reports Due to Changes in Membrane Properties or Experimental Conditions Polymer

Cation

Conductivtiy (mS/cm)

IEC (mmol/g)

Polyethylene

Ammonium Phosphonium Ammonium Sulfonium

40 22 55 15.6

1.29 0.67 1.2 1.37

Polysulfone

Condition 20 22 80 80

°C, liquid water °C, liquid water °C, liquid water °C,100% RH

Reference 46 26 47 33

developing highly conductive AEMs.1,48 One such parameter, the conducting ion concentration, or the number of free ions within the membrane, is important for ion conduction, but is also difficult to study experimentally.1,49 Therefore, comparing AEMs with different cations, while maintaining similar AEM properties and conductivity conditions, becomes critical for isolating the impact of changing the cation identity on conductivity. Furthermore, understanding the cation’s interactions with the surrounding aqueous environment becomes necessary so that parameters such as the percentage of free ions in an AEM can be determined. Toward that end, our group has previously developed a metal cation-based AEM platform that has enabled thorough structure−property relationship studies linking changes in the identity of the metal-cation to overall AEM performance, without also altering other AEM properties as described in the previously published work.34,36,37 In our materials platform, ring-opening metathesis polymerization (ROMP)-based, polymeric membranes containing bis(norbornene) nickel-, ruthenium-, and cobalt-terpyridine cations were synthesized and characterized, showing that AEMs with nickel complexes outperformed their ruthenium and cobalt counterparts.37 Here, we expand upon that work in an effort to elucidate why nickel-complex cations featured the highest chloride ion conductivity compared to the other cations within the same materials platform. To accomplish this goal, a series of AEMs were synthesized containing a poly(ethylene oxide) (PEO) cross-linker at different metal cation mole fractions, to compare between nickel-, ruthenium-, and cobalt-complex cations. Ion conductivity was measured in liquid water, confirming that nickel-complex cations facilitated higher chloride ion conductivity than ruthenium- or cobalt-complex materials. The mechanism by which some cations facilitate improved ion conductivity over others was elucidated through isothermal titration calorimetry (ITC) that probed the thermodynamics of

quantifying the effects that different cations have on AEM performance becomes critical, but has proven difficult to accomplish.32,40 The difficulty in constructing cation-conductivity relationships in these materials arises due to a lack of consensus in the field about how to directly measure the cation’s impact on AEM conductivity.24,41−45 This lack of consensus arises because there is no standard approach toward synthesizing and characterizing AEMs, resulting in a variety of polymer backbones, topologies and cations demonstrating a range of properties. As a result, few reports directly compare different types of cations while holding other membrane properties constant. AEM properties, such as IEC and morphology, and experimental conditions, such as the bulk hydration of the membrane and temperature, vary from one report to another, so it remains difficult to benchmark a material’s overall performance. For example, when reports of functional polyethylenes containing either quaternary ammonium- or aminophosphonium-based cations are compared (Table 1), the quaternary ammonium AEM appeared to have better conductivity than its aminophosphonium AEM counterpart, but also had a larger IEC.26,46 Likewise, when comparing polysulfone membranes functionalized with quaternary ammonium- or tertiary sulfonium-based cations, the ammonium-functionalized material again seemed to outperform the alternative, except in this case the quaternary ammonium AEM was tested in liquid water whereas the sulfonium AEM was characterized under 100% relative humidity (RH) (Table 1).33,47 Because conductivity is dependent on both the IEC and the bulk hydration of the membrane, it is difficult to compare AEMs that have differences in those properties and in the conditions of the conductivity experiment. The overall mechanism of conductivity in AEMs is not completely understood, and identifying parameters that impact conductivity, especially those related to the cation, is crucial for B

DOI: 10.1021/jacs.8b03979 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society Scheme 1. Synthetic Scheme for the PEO Series of AEMsa

a

The mole fraction, f, of metal cation and DCPD, X and Y respectively, was varied to change the overall metal content without affecting the crosslinking density. Nomenclature and components are shown.

cation−counterion pair strength for both metal and nitrogenbased cations. This report, for the first time, details specifically how changing the cation affected AEM conductivity by experimentally exploring cation−counterion association thermodynamics.

Table 2. Membrane Properties for all AEMs in the PEO Series Sample 0.36

RuPEO 0.55 RuPEO 0.36 NiPEO 0.55 NiPEO 0.36 CoPEO 0.55 CoPEO



RESULTS AND DISCUSSION AEM Synthesis and Characterization. To better understand why nickel cations showed improved ion conductivity, bis(norbornene) metal-terpyridine complexes, containing varying amounts of either nickel, ruthenium, or cobalt, were synthesized and copolymerized via ROMP with norbornene, dicyclopentadiene (DCPD), and a long, hydrophilic crosslinker poly(ethylene oxide) (PEO), to maintain the same crosslink density (Scheme 1).34,36,37 The dinorbornene functionalized PEO cross-linker (4 kg/mol, Figure S1), synthesized following previously reported procedures, was used to improve the water uptake and mechanical properties of the membranes for handling purposes.50−52 Six AEMs were synthesized: three at a cationic monomer mole fraction ( f) of 0.36 and three with f = 0.55 within the network. The IEC for these samples could not be determined experimentally due to the presence of the ester bonds in the PEO that could degrade in the IEC backtitration procedure. Therefore, the theoretical IEC values, calculated using the feed ratios of the monomers and ranging from 0.89 to 1.26 mmol/g, were used for these materials, although it should be noted that these theoretical values are known to be an overestimation, as demonstrated from analogous, non-PEO containing, cross-linked membranes through back-titration (Scheme S1 and Table S1). Full nomenclature for these samples is reported in Scheme 1, where the superscript before the element indicates the metal cation monomer mole fraction (f) in the network and the “PEO” subscript denotes the presence of the PEO cross-linker. In order to maintain the same cross-link density between both sets of membranes, the lower metal content samples had DCPD at f = 0.19, and the higher metal content samples had no DCPD, with all six membranes having norbornene and the PEO cross-linker at f = 0.36 and 0.09, respectively. All samples were characterized for both water uptake (Table 2) and mechanical properties (Figure S2 and Table S2). Ionic Conductivity of Metal Cations. The PEOcontaining samples were characterized for their chloride ion conductivity in liquid water as a function of temperature. Conductivity measurements were taken at 20, 30, 40, 60, and 80 °C, as shown in Figure 2. The conductivity of sample 0.55 CoPEO could not be studied as it was mechanically too weak and fractured when handled as it was inserted into the

Water Uptake (%)a

IEC (mmol/g)b

Ea (kJ/mol)

± ± ± ± ± ±

0.89 1.12 0.99 1.26 0.99 1.26

15.6 16.2 16.3 16.5 15.1

169 152 162 193 154 152

53 25 27 46 12 29

a

Liquid water uptake values, average of three trials, for AEMs in the Cl− form at room temperature, where water uptake = [(mwet − mdry)/ mdry]. bTheoretical IEC values for the AEMs calculated from the monomer feed ratio.

Figure 2. Chloride ion conductivity for the PEO series AEMs in liquid water as a function of temperature for each AEM. Sample 0.55CoPEO was not tested as it was too weak and broke during handling.

conductivity cell. As expected, all samples demonstrated an increase in conductivity with temperature. In addition, as the metal complex content was increased from f = 0.36 to 0.55, the conductivity increased along with the increase in IEC of the materials from 0.89−0.99 mmol/g to 1.12−1.26 mmol/g, respectively (Figure 2 and Table 2). All three membranes with a lower mole fraction had similar conductivity, contrary to a previous report that showed that nickel-based AEMs exhibited better conductivity than cobalt- and ruthenium-based AEMs. Importantly, that data was collected at 95% RH as opposed to liquid water, which may be significant.37 It is possible that the combination of being tested in liquid water and the higher water content of these samples promoted increased mobility of the chloride ions, which diminished the conductivity differences C

DOI: 10.1021/jacs.8b03979 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society between metal cations.53 As the metal content increased, however, 0.55NiPEO showed a larger increase in conductivity than 0.55RuPEO, consistent with the previously observed trend.37 To better understand this conductivity data, the conductivity values at 80 °C for each sample were plotted as a function of bulk hydration number (λ), as shown in Figure 3. The

c=

ρ × IEC 1 + WU

(3)

Figure 4. Chloride ion conductivity at 80 °C plotted versus ion concentration for each AEM. Sample 0.55CoPEO was not tested as it was too weak and broke during handling. Figure 3. Chloride ion conductivity at 80 °C plotted versus bulk hydration number, λ, for each AEM. Sample 0.55CoPEO was not tested as it was too weak and broke during handling.

where c is ion concentration of the chloride ions in the swollen membrane, in mmol Cl−/mL of membrane and water combined, and ρ is the density of the dried sample measured by determining its dried volume and mass.34,54 AEM 0.55CoPEO, again not shown, had an ion concentration of 0.58 mmol/mL. AEM 0.55NiPEO, similarly to 0.55CoPEO, had the largest ion concentration at 0.57 mmol/mL, whereas all other samples had ion concentrations between 0.28 and 0.52 mmol/mL (Figure 4). Though 0.55NiPEO had both the highest ion concentration and conductivity, ion concentration could not explain all of the conductivity data. For example, 0.55RuPEO had the second lowest ion concentration, 0.32 mmol/mL, but also the second highest conductivity, 12.3 mS/cm at 80 °C. Though the relationship between λ and ion concentration can be further explored through eqs S2 and S3, Figures 3 and 4 ultimately show that these three parameters (λ, ion concentration, and IEC) were not sufficient to account for the nickel cation-based AEM’s superior conductivity. Therefore, the activation energy of ion conduction, Ea, which relates to kinetics through the Arrhenius equation, was calculated for each membrane by plotting the natural logarithm of its chloride ion conductivity versus 1/T. Ea was then determined from the slope of the best fit linear regression, using the following form of the Arrhenius equation, eq 4:

hydration number, or number of water molecules in the membrane per counterion, for the samples was calculated using eqs 1 and 2, and is derived from the bulk hydration of the membrane: λ=

1000 × WU M H2O × IEC

WU =

(1)

mhyd − mdry mdry

(2)

where WU is the water uptake, MH2O is the molecular mass of water, 18.015 g/mol, IEC is the theoretical IEC of the membrane, mhyd is the hydrated mass of the AEM, and mdry is the dry mass of the AEM.34,54 AEM 0.55CoPEO, not shown, had a λ of 67. When the AEM conductivity data was plotted against λ in Figure 3, it became clear that increasing the metal complex content, from f = 0.36 to 0.55, decreased the amount of water molecules per counterion (λ). Because a positive correlation between increasing AEM cation content and an increased hydration number is often seen in literature for other AEMs and PEMs, it was surprising that here the hydration number appeared to decrease with increased IEC for each sample.15,54 In this case, the nickel complex showed the smallest decrease in λ, with a decrease of 7%, whereas ruthenium- and cobalt complex-based samples showed declines of 29% and 22%, respectively. All six samples had hydration numbers ranging from 75 to 105, and while the conductivity and λ of samples with f = 0.55 appeared to have a positive correlation, that correlation broke down when comparing AEMs with f = 0.36 and between different f. For example, for the samples with f = 0.36, 0.36CoPEO had the highest conductivity, followed by 0.36 NiPEO and then 0.36RuPEO, but the λ values showed the opposite trend. The conductivity at 80 °C was then plotted as a function of ion concentration, calculated using eq 3 (Figure 4):

ln(σCl−) = ln(σ0) −

Ea RT

(4)

where σCl is the ion conductivity, σ0 is the pre-exponential factor, R is the gas constant, and T is the absolute temperature (Table 2, Figure 5).34,55 All AEMs had similar activation energies, between 15.1 and 16.5 kJ/mol, indicating neither metal content nor metal center had an effect on the mobility of the chloride ions within the AEM for chloride conductivity (Table 2).56 Thermodynamic Characterization of Metal-Based Cations. Because the Ea of ion conduction is indicative of the kinetics of ion mobility and was similar regardless of the metal center, the thermodynamics of counterion conduction was characterized for the different cations.56 Counterion release for conduction depends on the strength of the cation− counterion association, with a weaker association increasing the −

D

DOI: 10.1021/jacs.8b03979 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

bicarbonate solution from zero to two molar equivalents of bicarbonate to chloride counterion. As the solutions were titrated, the heat of counterion exchange was monitored (Figures 6A and S4−S7) and the change in enthalpy for each injection (calculated as the area under the heat absorption spike) was plotted against the molar ratio of NaHCO3 to initial chloride counterion as a binding curve (Figure 6B). All titrations were compared to a control monitoring the heat of dilution of the NaHCO3 into pure RO water, where the curves in Figure 6 are corrected for this heat of dilution. The peaks in Figure 6A became larger after injections 6 and 15 because a larger volume of NaHCO3 was injected after those points, but the overall ΔH for the reaction continued to decrease as shown in Figure 6B. Traditionally, the binding curve obtained would be fit to determine a binding constant, however, the nickel complex did not produce a curve that could be fit, as it was too noisy, most likely due to the low amount of heat exchanged.59,60 Therefore, the total change in enthalpy, ΔHtot, for all three titrations was calculated by summing each data point in the plot (Table 3), as each data point represented

Figure 5. Chloride conductivity for PEO-containing AEMs as a function of temperature, where the Ea was calculated from the slope of the linear regression.

free ion concentration and ion conductivity.1,49 Therefore, the association strength between each metal cation and its corresponding chloride counterion was examined with a counterion exchange experiment using isothermal titration calorimetry (ITC) to determine if the thermodynamics of counterion exchange could account for the trends in conductivity. ITC is used for measuring the heat exchange upon binding between two species, making it the method of choice for this study.57,58 However, because ITC required the analyte to be soluble in an aqueous medium, neither the AEMs nor the metal complex monomers could be characterized. Therefore, three model, analogous bis(terpyridine) metal complexes were synthesized from nonfunctional terpyridine, bis(terpy)Ni, bis(terpy)Ru, and bis(terpy)Co, as shown in Scheme S2, to isolate the metal center’s effect on the strength of the cation−counterion pair. The association strength was quantified by measuring the enthalpic response of breaking the initial ion pair between the metal center and its chloride counterions by adding bicarbonate counterions to form new ion pairs. To probe the thermodynamics of cation−counterion association, the metal complexes in the chloride form were dissolved in RO water then titrated with an aqueous sodium

Table 3. ITC Data for Each Metal Complex Showing the Total Change in Enthalpy for the Overall Exchange Reaction Sample

ΔHtot (cal/mol HCO3−)

Bis(terpy)Ni Bis(terpy)Ru Bis(terpy)Co

586 ± 41 1,003 ± 48 1,120 ± 42

the ΔH for a single injection. Interestingly, the spontaneous exchange (negative ΔG) for all metal complexes was endothermic, with positive ΔHtot values, indicating that a positive change in entropy, ΔS, drove the reaction (eq 5). ΔG = ΔH − T ΔS

(5)

The change in entropy should be similar for all three reactions since the same number of principle molecules were involved in each exchange and the same exchange occurred for each titration. In addition, all experiments were run at the same temperature, 25 °C. Therefore, differences in the measured ΔHtot could be correlated to differences in the overall change in energy, ΔG, of the reaction, using eq 5. For the exchange, Bis(terpy)Ni had a ΔHtot of 586 cal/mol HCO3−, about half as

Figure 6. Counterion binding curves for bis(terpyridine)-metal complexes. (A) Raw ITC data for each metal complex stacked on the y-axis for clarity, average of two trials, which has been corrected for the heat of dilution of the NaHCO3 solution into water and (B) integration data plotted as a binding curve showing the decrease in the change in enthalpy for each injection as the molar ratio of NaHCO3 increased. Complexes contained either nickel (green), ruthenium (black), or cobalt (magenta). Samples were run in Milli-Q water at 25 °C exchanging chloride counterions for bicarbonate. E

DOI: 10.1021/jacs.8b03979 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

Figure 7. Nitrogen-based cations used for ITC characterization arranged in order of increasing ΔHtot from left to right.

surrounding aqueous environment (Figure 7 and Table S3). The titration of these cations demonstrated that increasing the alkyl chain length of the cation’s substituents appeared to have little effect on counterion binding. This was surprising since increasing the hydrophobicity of the cation was expected to increase the strength of the ion pair as it becomes less watersoluble, but that was not the case in this instance as ΔHtot was small for all three cations and similar for BEtA and BMeA.40,63 Interestingly, the ΔHtot values for the quaternary ammonium cations, either BMeA or TMA, and the imidazolium cation, EMI, correlate well with their respective AEM conductivity values from literature. These cations have been characterized extensively in AEMs, and in almost every case when they were compared in the same report the quaternary ammonium outperforms the imidazolium with higher conductivity values.34,41,43,64,65 Similarly, the quaternary ammonium cations were also able to release their chloride counterion more easily than the imidazolium, as demonstrated by lower ΔHtot values. This data further indicates that the differences in the ΔHtot values obtained from the exchange titrations in ITC can be related to the cation’s ultimate ability to conduct its counterion in an AEM, both for these more traditional, nitrogen-based cations as well as for the metal cations. Interpreting ITC Data in terms of Hydration. The differences, or lack of differences, among cations in the thermodynamics of cation−counterion association captured by ITC were most likely related to the cation hydration on the molecular level, as opposed to the bulk hydration number obtained from the swollen membrane. It is important to note that the ion hydration relates to the number of water molecules directly interacting with the cation, which is a separate parameter from the hydration shell, which corresponds to the number of water molecules surrounding the cation, but not necessarily directly interacting with it.66 In general, less hydrated cations have increased association strength with their counterions, which manifests as an increase in the magnitude of the enthalpic response during counterion exchange as measured by ITC.67 Furthermore, larger cations tend to have less hydration than smaller ions, as water molecules surrounding the cation interact more with each other and less with the cation, due to lower electrostatic forces and more delocalized charge associated with the larger ion.34,66,68,69 For example, a previous report studied the binding of various metal chloride salts to polyoxometalates using ITC.70 As the metal was changed between Na+, K+, Rb+, and Cs+, the heat produced by binding increased as the size of the cation increased. Given how increased ion size also results in reduced hydration, the increased enthalpic response was related to the decreased hydration of the ions creating stronger ion pairs. The relationship between ion hydration, ion pair strength, and the magnitude of the enthalpic response measured by ITC, regardless of whether it was exothermic or endothermic, can

large as bis(terpy)Ru and bis(terpy)Co, at 1003 and 1120 cal/ mol HCO3−, respectively. Because the nickel complex had a smaller ΔHtot, but likely a similar TΔS term, the ΔG of counterion exchange for the nickel complex was more negative than for the other metals, suggesting that there was more of a thermodynamic driving force for chloride ion release, a parameter that has previously been difficult to measure for AEMs and PEMs. Additionally, the lower ΔHtot indicated that less energy needed to be absorbed by the nickel system to facilitate counterion exchange, strongly suggesting that nickel had weaker initial binding to its chloride counterion.61 Weaker initial counterion binding and a stronger thermodynamic driving force for counterion release should increase the number of free ions and increase the AEM’s conductivity.1,49 Given that ion release is critical to any conductivity mechanism, and that the percentage of free ions is directly correlated to conductivity, this thermodynamic driving force for ion release appears to be the prevailing factor for the nickel-based AEM’s superior ion conductivity.1,62 Thermodynamic Characterization of Nitrogen-Based Cations. Characterizing the thermodynamic driving force for counterion release in metal cations led to a quantitative understanding of conductivity trends in the corresponding AEMs. Therefore, the same technique was used to study more traditional nitrogen-based cations in order to investigate the general applicability of ITC toward a wider range of cations. Consequently, five nitrogen-based cations, tetramethylammonium (TMA), benzyltrimethylammonium (BMeA), benzyltriethylammonium (BEtA), benzyltributylammonium (BBuA), and 1-ethyl-3-methylimidazolium (EMI), were studied using ITC (Figure 7). All five nitrogen-based cations were characterized following the same procedure used for the metal cations. The heat of counterion exchange was monitored (Figures S8A and S9A), with the change in enthalpy for each injection plotted against the molar ratio in Figures S8B and S9B. Interestingly, the titration of the nitrogen-based cations produced curves more similar to the bis(terpy)Ni titration curve, suggesting that the nickel cation had counterion exchange behavior more similar to the nitrogen-based cations than to the other metal cations. The ΔHtot was again calculated, and the results in Table S3 show that the nitrogen-based cations were also overall endothermic reactions, yet still spontaneous, meaning that entropy dominated their counterion exchange as well. All nitrogen-based cations demonstrated low ΔHtot values, where BEtA and BMeA had similar ΔHtot of 207 and 212, respectively, which were lower than BBuA’s ΔHtot of 287 cal/mol HCO3−, TMA’s ΔHtot of 350 cal/mol HCO3−, and EMI’s ΔHtot of 481 cal/mol HCO3− (Figure 7). Counterion exchange experiments with BMeA, BEtA, and BBuA probed the effect of cation hydrophobicity and steric crowding on counterion association, as those parameters could affect how the cation interacts with the counterion and the F

DOI: 10.1021/jacs.8b03979 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

Figure 8. Illustration relating the ion hydration to the ion pair association strength and the enthalpic response to breaking and forming ion pairs as measured by ITC, where green indicates the cation and red indicates the anion.

behaved more similarly to the nitrogen-based cations than to the other metal cations as its hydration level was more similar to the nitrogen-based cations.

be understood through Figure 8, which most likely hydration phenomenon most likely accounts for the changes in ITC measured here. As the hydration of the cation decreases, the strength of the ion pair association between the cation and its initial counterion increases, making it more difficult to perform the counterion exchange, as expressed by a larger, more endothermic ΔHtot. Because BMeA, BEtA, and BBuA all had low ΔHtot and it is known that tetramethylammonium and tetrabutylammonium analogues have similar levels of cation hydration, these three cations further establish the connection between enthalpic response measured by ITC and cation hydration.71 Despite the increased hydrophobicity among the cations, a similar enthalpic response indicated that cation hydration had a greater impact on the enthalpic response than the cation’s hydrophobicity. Furthermore, when comparing TMA, BMeA, and EMI, BMeA had a slightly lower ΔHtot than TMA, which was slightly lower than EMI. These slight differences could also be explained by small differences in cation hydration. The stabilizing effect of the benzyl group on the cation for BMeA would allow the counterion to dissociate more, increasing the hydration of the cation as compared to TMA. Furthermore, TMA should have greater hydration than EMI due to the increased size and delocalization of the charge on the imidazolium ring. As illustrated in Figure 8, this decreased hydration from BMeA to TMA to EMI would result in an increase of the initial ion pair strength with the chloride counterion, causing the larger enthalpic penalty (larger endothermic ΔHtot) measured by ITC for EMI over TMA, and TMA over BMeA (Figure 7). Interestingly, cation hydration could also explain the differences in ΔHtot observed for the metal cations. Metalcomplex cations are influenced by their metal-ligand field bonding and spin state, either low spin or high spin, where the spin state relates to the overall energy profile of the metal complex.72 Comparing the high- and low-spin states of an octahedral complex, like the metal complexes used here, shows that the high-spin state results in an increased charge localization within the center of the cationic complex.73 Specifically, for the complexes studied here, the diamagnetic bis(terpy)Ru is in the low-spin state, whereas the paramagnetic bis(terpy)Ni complex is in the high-spin state, meaning that its charge is increasingly localized in the center of the complex.37 Furthermore, while octahedral Co(II) complexes can either be in the high- or low-spin state, it has been shown that bis(terpy)Co complexes are primarily low-spin with chloride counterions.74 When in an aqueous medium, the increased charge localization for bis(terpy)Ni could cause more water molecules to directly interact with the complex, increasing the hydration of the metal complex cation. This would also explain why it



CONCLUSIONS Commonly studied parameters such as IEC, λ, ion concentration, and Ea were not sufficient to fundamentally explain the observed conductivity data, so the three metal cation complexes were evaluated for their counterion association thermodynamics. Using ITC, the nickel cation complex’s release of their chloride counterions was found to be more energetically favorable than in the case of the ruthenium and cobalt cation complexes, due to cation hydration. Nitrogen-based cations also had ΔHtot values more similar to that of the bis(terpy)Ni cation, indicating that the nickel cation behaved more similarly to the nitrogen-based cations than the other metal cations. Taken as a whole, ITC revealed that the thermodynamic driving force for counterion release appears to be a universal parameter for understanding cation behavior in ion exchange membranes. Moreover, these studies further implied that the cation hydration, as opposed to the standardly measured bulk membrane hydration, has a strong influence on cation− counterion association thermodynamics, which accounts for the significant differences in the measured ΔHtot for counterion exchange. In the future, more comprehensive studies about the thermodynamics of cation−counterion associations, including ΔG and ΔS, will likely produce more insight into the underlying parameters critical for ion conducting membranes.



EXPERIMENTAL SECTION

Monomer Synthesis and Membrane Preparation. The homoleptic bis(terpyridine) metal complexes and the norbornene functionalized telechelic PEO were synthesized according to previous reports.34,36,37,50,51 In order to keep the cross-link density consistent, the cationic metal complexes were polymerized in the presence of norbornene, dicyclopentadiene (DCPD), and the hydrophilic PEO. The metal-containing monomer, dicyclopentadiene, norbornene, and telechelic PEO were dissolved in a methanol/chloroform mixture. A solution of Grubb’s second generation catalyst (G2) in chloroform was added and the solution was stirred vigorously for up to 1 min. The solution was transferred to a preheated (40 °C) aluminum pan (diameter of ∼7 cm and depth of ∼1.5 cm) on a hot plate set to 40 °C. The pan was then covered with a glass jar (diameter of ∼7.5 cm and depth of ∼9 cm) to slow evaporation of the solvent. After 1 h at 40 °C, the cover was removed and the temperature remained at 40 °C for another hour, after which the temperature was raised to 70 °C for one more hour. The membrane was then cooled and transferred to a glass jar where it was swollen in 100% methanol, followed by 70% aqueous methanol, and then 30% aqueous methanol for at least 6 h for each solution. Finally, the membrane was swelled in 100% RO water for at least 12 h and the resulting membrane was stored in fresh RO water until characterization unless otherwise stated. For further information G

DOI: 10.1021/jacs.8b03979 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society



regarding monomer and membrane synthesis, refer to the Supporting Information. Nonfunctional terpyridine-metal complexes were synthesized by dissolving 2,2′:6′,2″-terpyridine and the desired metal chloride salt in methanol. The solution was stirred overnight after which the solvent was removed, and the resulting powder was dried under vacuum producing quantitative yields, unless otherwise stated. For further information on metal complex synthesis, refer to the Supporting Information. Characterization. Liquid water uptake was measured for all AEMs in the chloride form at room temperature. The fully hydrated AEM was removed from liquid water and the surface was blotted to remove excess surface water not absorbed into the membrane. The mass of the AEM was then recorded immediately and the membrane was placed back into water for 5−10 min. This process of weighing the hydrated membrane was repeated 3−5 times until consistent masses were obtained. They were then dried for 24 h in a vacuum at 50 °C, and the dried AEM was weighed for its dehydrated mass. Isothermal titration calorimetry (ITC) was performed using a Malvern MicroCal Auto-iTC200. The calorimetric titrations were performed at atmospheric pressure and 25 °C. An aqueous solution of NaHCO3 (∼10 mM) was titrated into the aqueous cation solution (all with a chloride counterion concentration of 1 mM) over the course of 28 injections with a stirring speed of 750 rpm. Injection volume was 0.4 μL for the first injection, 0.5 μL for the next five injections, 1 μL for the next nine injections, and 2 μL for the remaining 13 injections. The experiment was run using the single site binding analysis method and the first injection was removed due to probable leakage from the syringe.75 The heat of dilution of NaHCO3 into pure RO water measured by ITC and subtracted from each cation titration to ensure that NaHCO3 dilution did not influence the results. The membrane impedance was obtained using an impedance/gain phase analyzer (Solartron 1260A, Solartron Analytical, Farnborough Hampshire, ONR, UK). The membrane resistance was obtained from the real value of the impedance where the imaginary response was zero. The sample was immersed in liquid water and heated to 20 °C for 1 h. The membrane impedance was measured over a frequency range from 1 MHz to 100 Hz by two-point probe alternating current (AC) impedance spectroscopy three times during the hour. This was then repeated at 30, 40, 60, 80, 60, 40, 30, and 20 °C, in that order. The in-plane conductivity, σ in mS/cm, of each membrane was calculated from σ = L/RA, where L is the distance between reference electrodes, R is the resistance of the sample, and A is the crosssectional area of the sample. Any additional characterization information pertaining to this paper can be found in the Supporting Information.



ACKNOWLEDGMENTS This work was funded by the NSF NRT program DGE1545399 and the GAANN Fellowship DoED P200A150276. ITC data was obtained at the biophysical characterization core facility within the Institute for Applied Life Sciences (IALS) at The University of Massachusetts Amherst. The authors thank Ms. Kelly McLeod for providing the norbornene-functionalized PEO. The authors also thank Mr. Nicholas Posey for his assistance in preparing the manuscript. M.A.H. acknowledges the Corning Foundation for fellowship support and the NSF DMREF program under award CHE-1534326.



REFERENCES

(1) Amel, A.; Gavish, N.; Zhu, L.; Dekel, D. R.; Hickner, M. A.; EinEli, Y. J. Membr. Sci. 2016, 514, 125−134. (2) Robertson, N. J.; Kostalik, H. A.; Clark, T. J.; Mutolo, P. F.; Abruña, H. D.; Coates, G. W. J. Am. Chem. Soc. 2010, 132 (10), 3400− 3404. (3) Pan, J.; Zhu, L.; Han, J.; Hickner, M. A. Chem. Mater. 2015, 27 (19), 6689−6698. (4) Liu, D.; Kyriakides, S.; Case, S. W.; Lesko, J. J.; Li, Y.; McGrath, J. E. J. Polym. Sci., Part B: Polym. Phys. 2006, 44 (10), 1453−1465. (5) Lee, C. H.; Park, H. B.; Lee, Y. M.; Lee, R. D. Ind. Eng. Chem. Res. 2005, 44 (20), 7617−7626. (6) Lehmani, A.; Durand-Vidal, S.; Turq, P. J. Appl. Polym. Sci. 1998, 68 (3), 503−508. (7) Ono, H.; Miyake, J.; Miyatake, K. J. Polym. Sci., Part A: Polym. Chem. 2017, 55 (8), 1442−1450. (8) Cheng, J.; He, G.; Zhang, F. Int. J. Hydrogen Energy 2015, 40 (23), 7348−7360. (9) Zhang, H.-W.; Chen, D.-Z.; Xianze, Y.; Yin, S.-B. Fuel Cells 2015, 15 (6), 761−780. (10) Clark, T. J.; Robertson, N. J.; Kostalik, H. A., IV; Lobkovsky, E. B.; Mutolo, P. F.; Abruña, H. D.; Coates, G. W. J. Am. Chem. Soc. 2009, 131 (36), 12888−12889. (11) Merle, G.; Wessling, M.; Nijmeijer, K. J. Membr. Sci. 2011, 377 (1−2), 1−35. (12) Wu, B.; Ge, L.; Yu, D.; Hou, L.; Li, Q.; Yang, Z.; Xu, T. J. Mater. Chem. A 2016, 4 (38), 14545−14549. (13) Hickner, M. A.; Herring, A. M.; Coughlin, E. B. J. Polym. Sci., Part B: Polym. Phys. 2013, 51 (24), 1727−1735. (14) Couture, G.; Alaaeddine, A.; Boschet, F.; Ameduri, B. Prog. Polym. Sci. 2011, 36 (11), 1521−1557. (15) Ran, J.; Wu, L.; Wei, B.; Chen, Y.; Xu, T. Sci. Rep. 2015, 4 (1), 6486. (16) He, G.; Li, Z.; Zhao, J.; Wang, S.; Wu, H.; Guiver, M. D.; Jiang, Z. Adv. Mater. 2015, 27 (36), 5280−5295. (17) Ertem, S. P.; Tsai, T.-H.; Donahue, M. M.; Zhang, W.; Sarode, H.; Liu, Y.; Seifert, S.; Herring, A. M.; Coughlin, E. B. Macromolecules 2016, 49 (1), 153−161. (18) Sun, Z.; Lin, B.; Yan, F. ChemSusChem 2018, 11 (1), 58−70. (19) Edson, J. B.; Macomber, C. S.; Pivovar, B. S.; Boncella, J. M. J. Membr. Sci. 2012, 399−400, 49−59. (20) Liu, L.; He, S.; Zhang, S.; Zhang, M.; Guiver, M. D.; Li, N. ACS Appl. Mater. Interfaces 2016, 8 (7), 4651−4660. (21) Liu, Y.; Wang, J.; Yang, Y.; Brenner, T. M.; Seifert, S.; Yan, Y.; Liberatore, M. W.; Herring, A. M. J. Phys. Chem. C 2014, 118 (28), 15136−15145. (22) Gu, F.; Dong, H.; Li, Y.; Sun, Z.; Yan, F. Macromolecules 2014, 47 (19), 6740−6747. (23) Wang, J.; Chen, Y.; Wei, Q.; Yang, S.; Fang, H.; Wei, H.; Ding, Y. Polymer 2017, 112, 288−296. (24) Hugar, K. M.; Kostalik, H. A.; Coates, G. W. J. Am. Chem. Soc. 2015, 137 (27), 8730−8737. (25) Wang, J.; Li, S.; Zhang, S. Macromolecules 2010, 43 (8), 3890− 3896.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.8b03979. AEM and monomer synthesis and characterization description, synthetic schemes, IEC, 1H NMR, DMA, raw ITC, ITC analysis, water uptake (PDF)



Article

AUTHOR INFORMATION

Corresponding Author

*[email protected] ORCID

Liang Zhu: 0000-0003-4551-0963 Gregory N. Tew: 0000-0003-3277-7925 Notes

The authors declare no competing financial interest. H

DOI: 10.1021/jacs.8b03979 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

(56) Kornyshev, A. A.; Kuznetsov, A. M.; Spohr, E.; Ulstrup, J. J. Phys. Chem. B 2003, 107 (15), 3351−3366. (57) Maity, P.; Saha, B.; Kumar, G. S.; Karmakar, S. RSC Adv. 2016, 6 (87), 83916−83925. (58) Lounis, F. M.; Chamieh, J.; Leclercq, L.; Gonzalez, P.; Geneste, A.; Prelot, B.; Cottet, H. J. Phys. Chem. B 2017, 121 (12), 2684−2694. (59) Olesen, N. E.; Westh, P.; Holm, R. J. Colloid Interface Sci. 2015, 453, 79−89. (60) Tonga, G. Y.; Mizuhara, T.; Saha, K.; Jiang, Z.; Hou, S.; Das, R.; Rotello, V. M. Tetrahedron Lett. 2015, 56 (23), 3653−3657. (61) Cho, H.; Henkensmeier, D.; Brela, M.; Michalak, A.; Jang, J. H.; Lee, K. Y. J. Polym. Sci., Part B: Polym. Phys. 2017, 55 (3), 256−265. (62) Grew, K. N.; Chiu, W. K. S. J. Electrochem. Soc. 2010, 157 (3), B327−B337. (63) Marcus, Y. J. Solution Chem. 2008, 37 (8), 1071−1098. (64) Qiu, B.; Lin, B.; Qiu, L.; Yan, F. J. Mater. Chem. 2012, 22 (3), 1040−1045. (65) Chen, D.; Hickner, M. A. ACS Appl. Mater. Interfaces 2012, 4 (11), 5775−5781. (66) Hribar, B.; Southall, N. T.; Vlachy, V.; Dill, K. A. J. Am. Chem. Soc. 2002, 124 (41), 12302−12311. (67) Wang, H.; Wang, J.; Zhang, S.; Xuan, X. J. Phys. Chem. B 2008, 112 (51), 16682−16689. (68) Bernal, J. D.; Fowler, R. H. J. Chem. Phys. 1933, 1 (8), 515−548. (69) Samoilov, O. Y. A. Discuss. Faraday Soc. 1957, 24 (0), 141−146. (70) Pigga, J. M.; Teprovich, J. A.; Flowers, R. A.; Antonio, M. R.; Liu, T. Langmuir 2010, 26 (12), 9449−9456. (71) Bhowmik, D.; Malikova, N.; Mériguet, G.; Bernard, O.; Teixeira, J.; Turq, P. Phys. Chem. Chem. Phys. 2014, 16 (26), 13447−13457. (72) Housecroft, C. E.; Sharpe, A. G. Inorganic Chemistry, 3rd ed.; Pearson Education Limited, Prentice Hall: London, 2008; pp 637− 681. (73) Ioannidis, E. I.; Kulik, H. J. J. Chem. Phys. 2015, 143 (3), 034104. (74) Kremer, S.; Henke, W.; Reinen, D. Inorg. Chem. 1982, 21 (8), 3013−3022. (75) Francisco, V.; Basilio, N.; García-Río, L. J. Phys. Chem. B 2012, 116 (17), 5308−5315.

(26) Noonan, K. J. T.; Hugar, K. M.; Kostalik, H. A.; Lobkovsky, E. B.; Abruña, H. D.; Coates, G. W. J. Am. Chem. Soc. 2012, 134 (44), 18161−18164. (27) Zhang, W.; Liu, Y.; Jackson, A. C.; Savage, A. M.; Ertem, S. P.; Tsai, T.-H.; Seifert, S.; Beyer, F. L.; Liberatore, M. W.; Herring, A. M.; et al. Macromolecules 2016, 49 (13), 4714−4722. (28) Womble, C. T.; Coates, G. W.; Matyjaszewski, K.; Noonan, K. J. T. ACS Macro Lett. 2016, 5 (2), 253−257. (29) Yoshimura, K.; Zhao, Y.; Hasegawa, S.; Hiroki, A.; Kishiyama, Y.; Shishitani, H.; Yamaguchi, S.; Tanaka, H.; Koizumi, S.; Appavou, M.-S.; et al. Soft Matter 2017, 13 (45), 8463−8473. (30) Jheng, L.-C.; Tai, C.-K.; Hsu, S. L.-C.; Lin, B.-Y.; Chen, L.; Wang, B.-C.; Chiang, L.-K.; Ko, W.-C. Int. J. Hydrogen Energy 2017, 42 (8), 5315−5326. (31) Gong, X.; Yan, X.; Li, T.; Wu, X.; Chen, W.; Huang, S.; Wu, Y.; Zhen, D.; He, G. J. Membr. Sci. 2017, 523, 216−224. (32) Zhang, B.; Gu, S.; Wang, J.; Liu, Y.; Herring, A. M.; Yan, Y. RSC Adv. 2012, 2 (33), 12683−12685. (33) Hossain, M. A.; Jang, H.; Sutradhar, S. C.; Ha, J.; Yoo, J.; Lee, C.; Lee, S.; Kim, W. Int. J. Hydrogen Energy 2016, 41 (24), 10458− 10465. (34) Disabb-Miller, M. L.; Zha, Y.; DeCarlo, A. J.; Pawar, M.; Tew, G. N.; Hickner, M. A. Macromolecules 2013, 46 (23), 9279−9287. (35) Gu, S.; Wang, J.; Kaspar, R. B.; Fang, Q.; Zhang, B.; Coughlin, B. E.; Yan, Y. Sci. Rep. 2015, 5 (1), 1−11. (36) Zha, Y.; Disabb-Miller, M. L.; Johnson, Z. D.; Hickner, M. A.; Tew, G. N. J. Am. Chem. Soc. 2012, 134 (10), 4493−4496. (37) Kwasny, M. T.; Tew, G. N. J. Mater. Chem. A 2017, 5 (4), 1400−1405. (38) Chen, N.; Zhu, H.; Chu, Y.; Li, R.; Liu, Y.; Wang, F. Polym. Chem. 2017, 8 (8), 1381−1392. (39) Zhu, T.; Xu, S.; Rahman, A.; Dogdibegovic, E.; Yang, P.; Pageni, P.; Kabir, M. P.; Zhou, X.; Tang, C. Angew. Chem., Int. Ed. 2018, 57 (9), 2388−2392. (40) Li, N.; Leng, Y.; Hickner, M. A.; Wang, C.-Y. J. Am. Chem. Soc. 2013, 135 (27), 10124−10133. (41) Wang, B.; Sun, W.; Bu, F.; Li, X.; Na, H.; Zhao, C. Int. J. Hydrogen Energy 2016, 41 (4), 3102−3112. (42) Pan, J.; Han, J.; Zhu, L.; Hickner, M. A. Chem. Mater. 2017, 29 (12), 5321−5330. (43) Meek, K. M.; Nykaza, J. R.; Elabd, Y. A. Macromolecules 2016, 49 (9), 3382−3394. (44) Meek, K. M.; Elabd, Y. A. Macromolecules 2015, 48 (19), 7071− 7084. (45) Ran, J.; Wu, L.; Ru, Y.; Hu, M.; Din, L.; Xu, T. Polym. Chem. 2015, 6 (32), 5809−5826. (46) Kostalik, H. A.; Clark, T. J.; Robertson, N. J.; Mutolo, P. F.; Longo, J. M.; Abruña, H. D.; Coates, G. W. Macromolecules 2010, 43 (17), 7147−7150. (47) Pan, J.; Li, Y.; Han, J.; Li, G.; Tan, L.; Chen, C.; Lu, J.; Zhuang, L. Energy Environ. Sci. 2013, 6 (10), 2912−2915. (48) Sanoja, G. E.; Schauser, N. S.; Bartels, J. M.; Evans, C. M.; Helgeson, M. E.; Seshadri, R.; Segalman, R. A. Macromolecules 2018, 51 (5), 2017−2026. (49) Kiss, A. M.; Myles, T. D.; Grew, K. N.; Peracchio, A. A.; Nelson, G. J.; Chiu, W. K. S. J. Electrochem. Soc. 2013, 160 (9), F994−F999. (50) Walker, C. N.; Bryson, K. C.; Hayward, R. C.; Tew, G. N. ACS Nano 2014, 8 (12), 12376−12385. (51) Walker, C. N.; Versek, C.; Touminen, M.; Tew, G. N. ACS Macro Lett. 2012, 1 (6), 737−741. (52) Zhu, L.; Zimudzi, T. J.; Wang, Y.; Yu, X.; Pan, J.; Han, J.; Kushner, D. I.; Zhuang, L.; Hickner, M. A. Macromolecules 2017, 50 (6), 2329−2337. (53) Disabb-Miller, M. L.; Johnson, Z. D.; Hickner, M. A. Macromolecules 2013, 46 (3), 949−956. (54) Han, J.; Zhu, L.; Pan, J.; Zimudzi, T. J.; Wang, Y.; Peng, Y.; Hickner, M. A.; Zhuang, L. Macromolecules 2017, 50 (8), 3323−3332. (55) Chen, D.; Hickner, M. A. Macromolecules 2013, 46 (23), 9270− 9278. I

DOI: 10.1021/jacs.8b03979 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX