Wei
You
a,
Jacob M.
Ganley
b,
Brian G.
Ernst
a,
Cheyenne R.
Peltier
a,
Hsin-Yu
Ko
a,
Robert A.
DiStasio
Jr
*a,
Robert R.
Knowles
*b and
Geoffrey W.
Coates
*a
aDepartment of Chemistry and Chemical Biology, Baker Laboratory, Cornell University, Ithaca, NY 14853, USA
bDepartment of Chemistry, Princeton University, Princeton, NJ 08544, USA
First published on 1st February 2021
Alkaline anion exchange membranes (AAEMs) with high hydroxide conductivity and good alkaline stability are essential for the development of anion exchange membrane fuel cells to generate clean energy by converting renewable fuels to electricity. Polyethylene-based AAEMs with excellent properties can be prepared via sequential ring-opening metathesis polymerization (ROMP) and hydrogenation of cyclooctene derivatives. However, one of the major limitations of this approach is the complicated multi-step synthesis of functionalized cyclooctene monomers. Herein, we report that piperidinium-functionalized cyclooctene monomers can be easily prepared via the photocatalytic hydroamination of cyclooctadiene with piperidine in a one-pot, two-step process to produce high-performance AAEMs. Possible alkaline-degradation pathways of the resultant polymers were analyzed using spectroscopic analysis and dispersion-inclusive hybrid density functional theory (DFT) calculations. Quite interestingly, our theoretical calculations indicate that local backbone morphology—which can potentially change the Hofmann elimination reaction rate constant by more than four orders of magnitude—is another important consideration in the rational design of stable high-performance AAEMs.
Given that cation degradation under alkaline conditions is a major problem, many cationic moieties have been investigated for AAEM applications, including ammonium,12–19 phosphonium,20–23 imidazolium,24–29 and metal-based cations.30–33 Model compound studies can shed light on cation stabilities and their mechanisms of degradation.10,34–38 Notably, in 2015, Marino and Kreuer suggested incorporating piperidinium cations within AAEMs due to their outstanding alkaline stability and ease of synthesis.38 Since then, several groups have developed strategies to synthesize piperidinium-functionalized AAEMs.39–53 For example, Jannasch, Yan, and Zhuang independently developed polyphenylene-piperidinium systems, and all of which displayed good alkaline stabilities.39–43 In addition, Jannasch and coworkers prepared a multi-piperidinium-functionalized AAEM with a hydroxide conductivity of 221 mS cm−1 at 80 °C.44 These results suggest that piperidinium cations are promising structural motifs for high-performance AAEMs.
Every previously reported piperidinium-functionalized AAEM is based on a poly(aromatic) backbone: poly(phenylene oxide), polysulfone, polystyrene, polyphenylene, etc. These aromatic structures can make the polymers rigid, and may also be adsorbed by electrocatalysts, thereby reducing the efficiency of electrochemical reactions.54–56 Additionally, it has been proposed that the aryl-ether linkages in poly(phenylene oxide)s and polysulfones are unstable under alkaline conditions.56–58 More recently, Kim and coworkers demonstrated that poly(aromatic) backbones lacking heteroatoms can undergo oxidative degradation in an operating membrane electrode assembly (MEA).59 Therefore, it is advantageous to prepare analogous AAEMs free of aromatic substituents. Some aromatic-free AAEMs have been prepared with polyolefin backbones, including polyethylene (PE),12,20 polypropylene,60 polynorbornene,14,61–65 and polytetrafluoroethylene (PTFE).66,67 Among these materials, PE-based AAEMs have shown promising chemical stability and mechanical integrity. These materials are conveniently prepared from a ring-opening-metathesis polymerization (ROMP)/hydrogenation sequence of cyclooctene (COE) monomers functionalized with various cations, including ammonium,12,13 phosphonium,20 imidazolium,25,26 and cobaltocenium.30,33 Despite the outstanding properties of PE-based AAEMs, one of the major limitations of these materials is the laborious synthesis of the cationic monomers, which usually requires four to eight linear steps from commercially available 1,5-cyclooctadiene (COD) (Scheme 1).
Most synthetic routes to these COE monomers first entail an oxidation state or functional group manipulation of COD, followed by the installation of a nitrogen group. A more direct and conceptually simple synthesis of these compounds would be the redox neutral hydroamination of COD.68,69 While significant progress in the field has been made, intermolecular hydroaminations of unactivated alkenes with simple amines remain rare.70–73 Indeed, only a single example of COD hydroamination has been published to date. In 2017, we reported that the photocatalytic hydroamination of COD with piperidine could afford a functionalized COE in a single step.74 With the robust ROMP/hydrogenation strategy, we sought to convert these easily accessible monomers to potential AAEM materials. Herein, we use a modified protocol to prepare a series of piperidinium-functionalized COE monomers in a two-step, one-pot process that directly combines COD, piperidine, and an alkyl iodide. This photocatalytic method allows for the efficient synthesis of multi-gram quantities of monomers in a 100% atom-economical fashion. The quinuclidinium-functionalized analogue can also be synthesized in two steps from commercially available starting materials. To the best of our knowledge, this is the first report of aromatic-free piperidinium- and quinuclidinium-functionalized AAEMs.
Entry | [Piperidine] | COD equiv. | [Ir-cat] mol% | Thiol (mol%) | Yield |
---|---|---|---|---|---|
a Reactions were run at 0.5 mmol scale (piperidine). Yields were calculated based on piperidine for isolated materials after purification. See the ESI for more details. | |||||
1 | 0.05 M | 5.0 | 2.0 | TRIP thiol (50) | 76% |
2 | 0.05 M | 10 | 2.0 | TRIP thiol (50) | 82% |
3 | 0.2 M | 10 | 2.0 | TRIP thiol (50) | 72% |
4 | 0.2 M | 10 | 0.25 | TRIP thiol (50) | 86% |
5 | 0.2 M | 10 | 0.25 | TRIP thiol (15) | 80% |
6 | 0.2 M | 10 | 0.25 | Thiophenol (15) | 69% |
7 | 0.2 M | 10 | 0.25 | Thiophenol (50) | 64% |
We then evaluated the possibility of a telescoped sequence to directly isolate the quaternary ammonium monomers. After the optimized hydroamination reaction described above (Table 1, entry 6), the toluene, unreacted piperidine, COD, and thiophenol catalyst were removed under reduced pressure. The residue was refluxed with 2 equiv. of MeI in THF for 2 hours providing white solid 1 in 81% yield after trituration in Et2O. This one-pot, two-step approach can be used to prepare monomers expediently in at least 1.5 gram scale (see the ESI† for more details). Similar strategies were applied to synthesize other piperidinium-functionalized monomers, including ethyl (2), n-butyl (3), and n-octyl (4) derivatives (Scheme 2A). Quinuclidinium monomer 5 was prepared in 30% yield over two steps via hydroamination of COD with 2-(piperidin-4-yl)ethanol followed by an Appel reaction (Scheme 2B).75
Samples | Ionic monomer | mol% of 1–5 | IECb (mmol I− per g) | σ (OH−, 22 °C) (mS cm−1) | WUd | ΔLe |
---|---|---|---|---|---|---|
a See the ESI for more details. b Determined by 1H NMR analysis before hydrogenation. c Determined from the average of three trials ± standard deviation. d Water uptake at 22 °C = 100 × [masswet − massdry]/massdry%. e Dimensional change at 22 °C = 100 × [lengthwet − lengthdry]/lengthdry%. | ||||||
PEPM0.16 | 1 | 16% | 1.09 | 26 ± 1 | 22% | 4% |
PEPM0.20 | 1 | 20% | 1.31 | 35 ± 2 | 26% | 7% |
PEPM0.23 | 1 | 23% | 1.41 | 43 ± 2 | 54% | 8% |
PEPM0.28 | 1 | 28% | 1.61 | 53 ± 2 | 92% | 9% |
PEPE0.20 | 2 | 20% | 1.25 | 36 ± 2 | 28% | 6% |
PEPB0.18 | 3 | 18% | 1.12 | 16 ± 1 | 24% | 6% |
PEPO0.23 | 4 | 23% | 1.23 | 15 ± 5 | 40% | 8% |
PEQ0.20 | 5 | 20% | 1.25 | 35 ± 1 | 36% | 11% |
Fig. 1 (A) Membrane hydroxide conductivity and water uptake of PEPMs with different ionic contents. (B) Membrane hydroxide conductivities of PEPMs under different temperatures. |
The membranes were soaked as strips in 1 M KOHaq at 80 °C in polypropylene (PP) vials for 30 days, and their alkaline stabilities were evaluated by both changes in hydroxide conductivities (Fig. 2) and Fourier-transform infrared (FT-IR) spectroscopy (Fig. 3). The methyl-substituted piperidinium PEPM20 and quinuclidinium PEQ20 showed the best conductivity stability (80% and 71% retained conductivity, respectively). The hydroxide conductivities of PEPB18 and PEPO23 boosted during the first three days with the plausible reasons discussed above, while drastic conductivity decreases were observed afterwards (21% and 40% remaining, respectively). AAEMs with ethyl substituents (PEPE20) exhibited obvious conductivity loss from the beginning of the stability tests (42% remaining after 30 days). These overall stability trends agreed well with previous reports.38,39 One of the most stable AAEMs, PEPM20, was further evaluated for its mechanical stability before and after the alkaline treatment (Fig. S38†). It was found that the as-synthesized membrane PEPM20 (fully hydrated in the iodide form) was strong and ductile (22 MPa stress and 330% strain at break), then the membrane became relatively weaker (15 MPa) yet more ductile (357% strain) after soaking in 1 M KOHaq at 80 °C for 30 days. Their excellent mechanical properties were comparable to Nafion™ N115 (32 MPa stress and 310% strain at break).78 The elongation properties of these PE-based AAEMs significantly outperformed the rigid AAEMs derived from poly(aromatic)s.54,79
Fig. 3 FT-IR spectra of membranes before and after alkaline treatment to study degradation mechanisms. |
To interrogate the degradation pathways, we used FT-IR spectroscopy to study the structural changes in these membranes before and after the alkaline treatment (Fig. 3). Piperidine-functionalized PE (PEP20) was also prepared from the HCl adduct of the hydroamination intermediate (compound S1†) using an analogous ROMP copolymerization strategy followed by deprotonation to mimic some of the proposed degradation products (pathways b and c in Scheme 3, see ESI† for more details).25 Since the as-prepared PE-based AAEMs cannot be dissolved in common organic solvents, solid-state characterization methods such as FT-IR are advantageous. Originally, we proposed that the piperidinium AAEMs would give extremely clear FT-IR spectra, as there are only three types of bonds in these aromatic-free polymers (i.e., C–C, C–H, and C–N bonds). However, as shown in Fig. 3 and S39,† we noticed that the signals of C–N bonds were too weak to be confidently assigned in comparison to the strong C–H bond signals (1460 cm−1 and 2800–3000 cm−1). The small hump peaks at 1650 cm−1 and 3300 cm−1 were from water in membranes.80 The FT-IR spectra of PEP, PEPM, PEPE, PEPB, PEPO, and PEQ were almost the same (Fig. S39†) suggesting that FT-IR spectroscopy might not be sensitive enough to distinguish piperidine and piperidinium in these polymer samples. A new peak at 964 cm−1 possibly suggested the presence of alkene groups in the polymer after alkaline treatment (the red trace in Fig. 3), yet it is hard to clearly assign it as pathways d or e in Scheme 3. As an important note, we observed that when we performed the stability studies in soda lime glass vials, there were always new broad peaks around 1050 cm−1 in FT-IR spectra for all the samples after alkaline treatment (see the blue trace in Fig. 3 and S40†). These peaks were attributed to Si–O bonds from glass etching, as we measured the weight of the glass vials dropping by near 0.5 g after the stability studies (Table S5†). We also observed relatively slower conductivity decrease in glass vials (Fig. S41†). Although similar glass etching problems have been observed in alkaline stability studies,81,82 this is the first spectroscopic evidence to prove that the detached Si–O substances could be adsorbed by polymer samples to affect their ionic conductivities. Plastic containers made from PP and PTFE are recommended to perform aqueous alkaline stability tests.38
To gain more insight into the degradation mechanism, we decided to use solution-based 1H NMR analysis, as it can provide quantitative degradation kinetics and clearer degradation product assignments. Our group has previously conducted alkaline stability studies under accelerated degradation conditions by treating model compounds or polymers with KOH solutions in CD3OH using 3-(trimethylsilyl)-1-propanesulfonic acid sodium salt as the internal standard at 80 °C in sealed NMR tubes.10,25,26,36 As a side note, control experiments showed that borosilicate glass NMR tubes were more resistant to basic corrosion and there were almost no changes to their weights after the KOH/methanol stability studies (Table S5†).10 One requirement of this NMR stability protocol is that the studied model compounds or polymers must be soluble in methanol. Fortunately, we can easily tune their solubilities by changing the cation incorporation ratio and molecular weight during the ROMP copolymerization process. For example, to get the previously mentioned mechanically strong AAEMs, we used a mol ratio of [COE]:[piperidinium]:[Ru] to be 800:200:1. By changing the ratio to 200:150:1, we obtained cationic polymers that were soluble in n-propanol (5 wt%) at 50 °C, which are ideal candidates as AEI solutions (Fig. S15†).12 By further changing the feeding ratio to 25:50:1, we were able to prepare a series of oligomers that were soluble in methanol at 50 °C. Notably, the cation incorporation percentage was slightly lower than expected in these oligomers, and monomers 1–5 gave rise to oligomers PEPM52, PEPE54, PEPB58, PEPO64, and PEQ55, respectively (see ESI† for more details).
We then subjected five cationic oligomers and a small molecule model compound N-hexyl-N-methylpiperidinium bromide (6) to the accelerated alkaline degradation conditions: 2 M KOH in CD3OH at 80 °C for 30 days (Fig. 4). The use of CD3OH makes anions ([OH]− and [OCD3]−) more reactive due to a smaller hydration sphere, and also prevents potential H/D exchange reactions.10,36 We found that cationic model compound 6 had very good alkaline stability (94% remaining after 30 days), and detected small amounts of byproducts from both nucleophilic demethylation (3%) and Hofmann elimination (2%).83 However, we observed much faster cation degradation in oligomers with the general trend being PEQ ≈ PEPM > PEPE > PEPB > PEPO. Oligomers PEPM52, PEPB58, and PEPO64 decomposed mostly through backbone Hofmann elimination (e) under these accelerated KOH/methanol conditions, as large quantities of N-alkylpiperidines were detected as byproducts (Fig. 4). We observed 27%, 55%, and 66% N-methyl-, N-butyl-, and N-octyl-piperidine compared to 41%, 87%, and 93% degradation from the corresponding PEPM52, PEPB58, and PEPO64, respectively. In all these cases, we also detected a broad single peak around 5.4 ppm, which is consistent with internal alkene peaks. The trend of increasing length of alkyl chains facilitating Hofmann elimination elsewhere is consistent with what Jannasch and coworkers demonstrated for their piperidinium-functionalized AAEMs, although they observed ring-opening Hofmann elimination as the major degradation pathway due to rigid substituents on the piperidinium ring.39 Ring-opening Hofmann elimination and nucleophilic ring opening degradation may also occur under our conditions, but these minor degradation pathways are extremely difficult to be accurately quantified by 1H NMR analysis in the complicated reaction system. Interestingly, a volatile degradation byproduct, ethylene, was detected from the degradation of PEPE54 oligomer (31% remaining). The backbone elimination (22%, e) and ethyl elimination (26%, c) gave similar rates under the testing conditions. Quinuclidinium PEQ55 gave 30% nucleophilic ring-opening degradation out of its 35% cation degradation, probably owing to the ring strain in this bicyclic scaffold. It is noteworthy that for all these oligomer studies in sealed NMR tubes, we observed the formation of dark insoluble materials, which likely formed as the oligomers became electroneutral and less polar after degradation, diminishing their solubility in methanol. These findings highlight the importance of using an internal standard to track polymer alkaline stabilities through 1H NMR analysis, as key information about volatile and insoluble byproducts could be easily missed otherwise.
To provide additional insight into the degradation mechanisms of these piperidinium- and quinuclidinium-functionalized AAEMs, we used dispersion-inclusive hybrid density functional theory (DFT) in conjunction with an implicit solvent model to compute free-energy barriers (ΔG‡) for the degradation pathways in Scheme 3 under the experimental conditions (2 M KOH/CD3OH, 80 °C). To do so, we constructed models of the PEPM, PEPE, and PEQ systems (i.e., model-PEPM, model-PEPE, and model-PEQ) in which the PE backbone was truncated to include only seven carbon atoms (see Fig. 5B for a graphical depiction of model-PEPM; see Theoretical and computational methods for a detailed description of the computational protocol).
Fig. 5 (A) Free-energy barriers (ΔG‡ in kcal mol−1) corresponding to the degradation pathways in Scheme 3 for model-PEPM as a function of an applied force (F in nN) to the distance (D) between γ-carbons on the polymer backbone. All barrier heights were computed with CH3O− as the nucleophile and base (i.e., at the experimental conditions of 2 M KOH/CD3OH, 80 °C) using the EFEI approach in conjunction with dispersion-inclusive hybrid DFT and an implicit solvent model. (B) Top: Graphical depiction of model-PEPM. Bottom: Tabulated statistics describing the model-PEPM backbone as a function of F. Even with the largest forces considered in this work (F = ±0.25 nN), the average backbone C–C bond length (〈dCC〉) and C–C–C bond angle (〈θCCC〉) remained largely unchanged from their force-free (F = 0 nN) values. |
For model-PEPM, ΔG‡ values associated with Hofmann elimination (d, e) were lower than that of SN2 (a, b) by ca. 4–6 kcal mol−1, which suggests that these nucleophilic substitution channels were not the primary degradation modes in this system (see Table 3). When comparing amongst the Hofmann elimination pathways, DFT predicted that d was preferred over e by ΔΔG‡ = ΔG‡e − ΔG‡d = 1.2 kcal mol−1. This ΔΔG‡ value corresponds to a factor of ca. 5.5× in the relative degradation rate constants (i.e., kd/ke = exp[(19.8 kcal mol−1 − 21.0 kcal mol−1)/RT ] = 5.5 at T = 80 °C in accordance with transition state theory (TST), see Theoretical and computational methods). As described above, 1H NMR analysis indicated that byproducts from e comprised 66% of the PEPM degradation (i.e., 27% of the total 41% degradation; see Fig. 4), while the remaining 34% remained unassigned (potentially due to solubility issues, vide supra). According to the theoretical results in Table 3, this unassigned portion most likely resulted from d, as the SN2 pathways have substantially higher ΔG‡ values. Although DFT errors in the prediction of relative barrier heights can exceed 1.0 kcal mol−1,84 our calculations on model-PEPM suggested that d should be the dominant degradation pathway, while 1H NMR analysis showed a slight preference for e.
System | Pathway a | Pathway b | Pathway c | Pathway d | Pathway e |
---|---|---|---|---|---|
a All barrier heights were computed with CH3O− as the nucleophile and base (i.e., at the experimental conditions of 2 M KOH/CD3OH, 80 °C) using dispersion-inclusive hybrid DFT and an implicit solvent model. Values in bold indicate primary degradation pathways based on the relative ΔG‡ values for a given system. b For consistency with model-PEPM and model-PEPE, we treated the bridgehead –CH2CH2– group that is orientated perpendicular to the backbone as the substituent (i.e., the pathways involving this –CH2CH2– group are b and c). When computing ΔG‡ for model-PEQ, the SN2 (a and b) and Hofmann elimination (c and d) degradation pathways depend on this orientation and are therefore reported with different values. Effective ΔG‡ values for these combined pathways are 22.9 kcal mol−1 (a, b) and 25.6 kcal mol−1 (c, d). See Theoretical and computational methods for more details. | |||||
Model-PEPM | 25.8 | 24.9 | N/A | 19.8 | 21.0 |
Model-PEPE | 26.1 | 25.2 | 17.1 | 20.7 | 22.7 |
Model-PEQb | 23.0 | 24.6 | 26.4 | 25.8 | 22.1 |
To further explore this issue, we investigated one aspect of the experiment that was not accounted for in the theoretical ΔG‡ in Table 3, namely, how the local conformation of the polymer backbone affects degradation kinetics in piperidinium-based AAEMs. To do so, we sampled both compressed and extended backbone conformations in model-PEPM, and computed ΔG‡ for each degradation pathway as a function of backbone geometry (see Fig. 5). This was accomplished by applying a small tensile force (F) to the distance (D) between the γ-carbons (Cγ) in model-PEPMvia the external forces explicitly included (EFEI) approach.85 In doing so, we explored a wide range of backbone conformations without perturbing the backbone C–C bond lengths (and C–C–C bond angles) in model-PEPM, thereby mimicking the distribution of local polymer morphologies present in PEPM. From the plot in Fig. 5A, backbone geometry had little to no effect on ΔG‡ for the SN2 pathways (a, b) as well as the Hofmann elimination on the piperidinium ring (d). However, Hofmann elimination along the backbone (e) was strongly dependent on the backbone conformation and became increasingly facile for extended, and increasingly difficult for compressed model-PEPM backbones. According to our calculations, this effect was remarkably large (i.e., modifying ΔG‡ by ca. 7 kcal mol−1) and can potentially change the TST rate constant by up to four orders of magnitude!
This finding is consistent with the Hofmann elimination (E2) mechanism, which proceeds more favorably when Hβ–Cβ is anti-periplanar to Cα–N, i.e., θHβCβCαN > 150° or θHβCβCαN < −150°.38 For compressed backbone conformations (in which Cγ–Cβ tended to be anti-periplanar to Cα–N), the observed increase in ΔG‡ resulted from unfavorable torsional rearrangement in the transition state to access the required anti-periplanar Hβ–Cβ–Cα–N orientation. For extended backbone conformations, Hβ–Cβ was naturally anti-periplanar to Cα–N, hence e can proceed with relative ease. Since PEPM exists as an ensemble of backbone conformations, accounting for the effects of local backbone morphology on ΔG‡ (via an appropriately weighted Boltzmann average over the ΔG‡ values in Fig. 5) could yield a very different picture of the degradation kinetics than that provided by the ΔG‡ (F = 0) values alone. As such, one explanation that is consistent with both our experimental and theoretical data is that the distribution of backbone conformations tends to be more extended in PEPM; this would lead to a more favorable ΔG‡ for e, making this pathway competitive with (or even more favorable than) d.
To investigate how the substitution of R = Me (PEPM) for R = Et (PEPE) affected the stability of these piperidinium-based AAEMs, we also computed barriers for the analogous model-PEPE system. Again, we found that the SN2 pathways (a, b) had higher barriers than the Hofmann elimination pathways (c, d, e), thereby suggesting that nucleophilic substitutions were not the primary degradation modes in PEPE as well. 1H NMR analysis attributed ca. 38% and 32% of the total PEPE degradation to c and e (i.e., 26% and 22% of the total 69% degradation; see Fig. 4); with higher SN2 barriers, we again hypothesized that the remaining 30% (unassigned) likely resulted from d. Since local backbone morphology substantially influenced ΔG‡ for e in model-PEPM (see Fig. 5), we expect the same in model-PEPE; if extended backbone conformations were favored in PEPE, e could again be competitive with d, which was observed experimentally.
Since rotation about Cα–Cβ in the ethyl moiety of model-PEPE was largely unhindered, one would expect that any one of the three corresponding β-hydrogens would be in the anti-periplanar position preferred by Hofmann elimination. DFT calculations on model-PEPE strongly agreed with these expectations and predicted ΔΔG‡ values for c (with respect to the other pathways) that were tantamount to >99% of the degradation proceeding through this channel. As mentioned above, 1H NMR indicated that c was only responsible for 38% of the PEPE degradation, in stark contrast to the theoretical result. Since the model-PEPE backbone conformation was unlikely to affect ΔG‡ for c, theoretical arguments based on local polymer morphology alone are insufficient to explain why c seems to be severely hindered in PEPE. Notably, a similar discrepancy also occurs in (even simpler) tetraalkylammonium systems, wherein experimental characterization of the kinetics associated with ethyl-based Hofmann elimination also yields markedly different results from the theoretical ΔΔG‡ values (cf. Table 3 of ref. 38 with Fig. 6 of ref. 86). In this case, experimental Hofmann elimination rates, i.e., half-life (τ) data, obtained by Marino and Kreuer38 for ethyltrimethylammonium (ETM) and propyltrimethylammonium (PTM) correspond to a ΔΔG‡ = ΔG‡ETM − ΔG‡PTM = −RTln[τPTM/τETM] = −RTln[33.2 h/2.8 h] = −2.1 kcal mol−1 at T = 160 °C (assuming pseudo-first order kinetics in 6 M NaOH and no side reactions), while the theoretical calculations of Long et al.86 yield ΔΔG‡ = ΔG‡ETM − ΔG‡PTM = 17.5 kcal mol−1 − 22.9 kcal mol−1 = −5.4 kcal mol−1 under the same conditions. As such, the (relative) rate of ethyl-based Hofmann elimination in tetraalkylammonium salts also seems to be markedly attenuated with respect to the theoretically predicted ΔΔG‡ values. Since a deeper understanding of this degradation pathway can be leveraged in the rational design of robust next-generation AAEMs for alkaline fuel cell applications, the underlying reason(s) (e.g., the use of TST, approximate treatment of electronic structure and solvent effects, etc.) for this apparent discrepancy is currently under investigation in our group.
For model-PEQ, the Hβ–Cβ in the bicyclic (quinuclidinium) ring cannot easily access an anti-periplanar position with Cα–N, thereby making Hofmann eliminations via the ring (c, d) less viable degradation pathways (Table 3). In direct contrast to model-PEPM and model-PEPE, the ring-opening SN2 reactions (a, b) have much lower ΔG‡ values (most likely due to ring strain) and now become competitive pathways in model-PEQ. 1H NMR analysis indicated that these SN2 reactions (a, b; experimentally indistinguishable) and the remaining Hofmann elimination (e) were indeed the primary degradation pathways, with a corresponding assignment of 86% (a, b) and 14% (e) of the PEQ byproducts (i.e., 30% and 5% of the total 35% degradation, respectively; see Fig. 4). In this case, there is strong experimental (cf. relative degradation kinetics for e for all polymers in Fig. 4) and theoretical (cf. relative ΔG‡ values for model-PEQ in Table 3) evidence that degradation through e was attenuated in PEQ. Taking local backbone morphology into consideration, these findings might be indicative of a preference for compressed backbone conformations in PEQ (in contrast to PEPM and PEPE). As such, this morphological picture of PEQ shares some similarities with the stacked layer model proposed by Trigg et al. for precisely sulfonated polyethylene materials.87
As such, the following working expression was used when computing G for a given reactant (Greact) or TS species (GTS):
G(F) = Eelec + H − TS + VEFEI(F) − RTlng. | (1) |
In this expression, Eelec is the electronic (or DFT) energy computed at the ωB97X-V/cc-pVTZ[HCN]/aug-cc-pVTZ[O]/PCM//B3LYP+D3/6-31G*[HCN]/6-31+G*[O]/PCM level. H is the enthalpic contribution:
H = H(T) = Etrans(T) + Erot(T) + Evib(T) + pV, | (2) |
S = S(T) = Strans(T) + Srot(T) + Svib(T), | (3) |
Since the primary purpose of the theoretical calculations in this work is to assess the relative viability among a set of competing degradation pathways (with the same reactants), all free-energy barriers (ΔG‡ = GTS − Greact) were computed with respect to a single reference, i.e., the global minimum reactant complex on the corresponding potential energy surface (PES), instead of the reactant complex directly connected to each TS via the intrinsic reaction coordinate (IRC). Doing so explicitly accounts for the equilibrium (Boltzmann) population of the corresponding IRC-based reaction complex—which is potentially a local minimum reaction complex on the PES—in each ΔG‡ value provided herein, and therefore allows for a fairer comparison among competing degradation pathways.
Due to the potential errors that arise when computing free-energy barriers with DFT (which often exceed 1.0 kcal mol−1),95 we avoid discussing absolute reaction rates throughout this work, and instead compare and contrast ΔG‡ values corresponding to different degradation pathways (ΔG‡1 and ΔG‡2) in relative terms, i.e., ΔΔG‡ = ΔG‡1 − ΔG‡2. We also discuss the relative differences between ΔG‡i values in terms of the corresponding reaction rate constants (ki), which have been estimated using the Eyring equation of transition state theory (TST):
(4) |
(5) |
(6) |
For identical degradation pathways (e.g., with two chemically equivalent SN2 or E2 sites), ΔG‡1 = ΔG‡2 and this expression simplifies to:
ΔG‡eff = ΔG‡1 − RTln2, | (7) |
All calculations were performed using the following convergence criteria: 10−8 au for the DIIS error and 10−11 au for the integral thresholding value during self-consistent field (SCF) optimizations; 3.0 × 10−6 hartree per bohr for the maximum gradient component, 1.2 × 10−5 bohr for the maximum atomic displacement, and 1.0 × 10−8 hartree for the energy change between successive steps during geometry optimizations (with convergence reached after satisfying any two of these criteria). A detailed breakdown of the free-energy components (Table S6†) as well as the optimized geometries for each species are included in the ESI.†
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0sc05789d |
This journal is © The Royal Society of Chemistry 2021 |