Unusual Lower Critical Solution Temperature Phase Behavior of Poly(benzyl methacrylate) in a Pyrrolidinium-Based Ionic Liquid

Polymer/ionic liquid systems are being increasingly explored, yet those exhibiting lower critical solution temperature (LCST) phase behavior remain poorly understood. Poly(benzyl methacrylate) in certain ionic liquids constitute unusual LCST systems, in that the second virial coefficient (A2) in dilute solutions has recently been shown to be positive, indicative of good solvent behavior, even above phase separation temperatures, where A2 < 0 is expected. In this work, we describe the LCST phase behavior of poly(benzyl methacrylate) in 1-butyl-1-methylpyrrolidinium bis(trifluoromethylsulfonyl)imide for three different molecular weights (32, 63, and 76 kg/mol) in concentrated solutions (5–40% by weight). Turbidimetry measurements reveal a strong concentration dependence to the phase boundaries, yet the molecular weight is shown to have no influence. The critical compositions of these systems are not accessed, and must therefore lie above 40 wt% polymer, far from the values (ca. 10%) anticipated by Flory-Huggins theory. The proximity of the experimental cloud point to the coexistence curve (binodal) and the thermo-reversibility of the phase transitions, are also confirmed at various heating and cooling rates.


Introduction
Ionic liquids (ILs) are an important class of "designer solvents", able to access numerous desired physical properties through modest modifications of their chemical structure. ILs offer thermal, chemical, and electrochemical stability, as well as high ion conductivity and negligible volatility, thereby showing great promise as greener alternatives to traditional solvents [1][2][3][4][5]. Incorporating a polymer matrix into these ILs imparts structural and mechanical integrity to the resulting material, while retaining the liquid-like diffusivity of the IL [6][7][8]. This has enabled their use for separation membranes, printable electronics, battery electrolytes, and self-healing materials [9][10][11][12][13][14][15]. However, this growing class of smart materials poses some interesting fundamental questions.
A system of particular interest is poly(benzyl methacrylate) (PBzMA) in ILs. Watanabe and co-workers first observed its LCST behavior and proposed that the formation of clathrates via cation-π interactions between imidazolium cations and benzyl groups leads to entropic stabilization during mixing [26][27][28][29]. Later work also observed intramolecular π-π stacking between adjacent benzyl groups, creating compact polymer aggregates rather than flexible random coils, as is typically observed within good solvents [17,30]. Recently, Kharel et al. determined the second virial coefficient to be consistently positive for PBzMA in dilute imidazolium-and pyrrolidinium-based ionic liquid solutions, indicating the ILs are reasonably good solvents, even above temperatures wherein phase separation occurs at higher concentrations. Thus, the theta temperature determined in very dilute solutions is well above the separation temperature of these systems, an unusual characteristic [31]. The critical concentration of several PBzMA/IL systems have also been observed to be significantly shifted toward polymer-rich compositions, contrary to the solvent-rich compositions anticipated by Flory-Huggins theory [16,31,33,34].
Few phase diagrams for polymers in ILs have been explored over wide composition ranges, and, to our knowledge, no polymer system involving pyrrolidinium-based ILs has been fully characterized. These non-aromatic ILs have been shown to be less toxic than their aromatic counterparts; thus, they can potentially be used as safer alternatives for applications once there is a better understanding of their behavior in polymer solutions [35]. In this work, we expand upon the previous studies by exploring the phase behavior of PBzMA in 1-butyl-1-methylpyrrolidinium bis(trifluoromethylsulfonyl)imide ([BMP][TFSI]) for a wide range of concentrations via turbidimetry as a function of polymer molecular weight ( Figure 1). Additionally, the thermal reversibility of the phase transition is examined and discussed.

Results
At low temperatures, PBzMA is completely miscible with [BMP][TFSI] and a clear solution is observed. As can be seen in Figure S3, upon heating, a sharp drop in transmittance occurs during phase separation as the solution becomes cloudy; the temperature at which this transpires is the cloud point (TCP). The subsequent increase in transmittance at even higher temperatures is due to the formation of distinct phases. When annealed at temperatures above TCP for several days, the formation of clear coexisting layers is achieved. The lower and upper phases consist of a [BMP][TFSI]-rich liquid and a PBzMArich viscoelastic liquid, respectively, as the density of [BMP][TFSI] (≈1.40 g/cm 3 ) is higher than that of PBzMA (≈1.18 g/cm 3 ). However, if these distinct layers have not fully developed before the solution is cooled to room temperature, the top layer remains opaque for

Results
At low temperatures, PBzMA is completely miscible with [BMP][TFSI] and a clear solution is observed. As can be seen in Figure S3, upon heating, a sharp drop in transmittance occurs during phase separation as the solution becomes cloudy; the temperature at which this transpires is the cloud point (T CP ). The subsequent increase in transmittance at even higher temperatures is due to the formation of distinct phases. When annealed at temperatures above T CP for several days, the formation of clear coexisting layers is achieved. The lower and upper phases consist of a [BMP][TFSI]-rich liquid and a PBzMArich viscoelastic liquid, respectively, as the density of [BMP][TFSI] (≈1.40 g/cm 3 ) is higher than that of PBzMA (≈1.18 g/cm 3 ). However, if these distinct layers have not fully developed before the solution is cooled to room temperature, the top layer remains opaque for months when stored in a vacuum desiccator. This suggests that the PBzMA-rich phase is metastable under these conditions, presumably due to kinetic limitations (e.g., reflecting chain entanglement and/or the glass transition).
The concentration dependence of T CP was explored for PBzMA-63 in [BMP][TFSI] and is illustrated in Figure 2, where the relative transmittance of each solution is plotted as a function of temperature on heating. The relative transmittance rapidly decreases (within seconds) upon crossing T CP , irrespective of concentration. Accordingly, T CP for the 10,15,20,25 Figure 2, where the relative transmittance of each solution is plotted as a function of temperature on heating. The relative transmittance rapidly decreases (within seconds) upon crossing TCP, irrespective of concentration. Accordingly, TCP for the 10,15,20,25  An experimentally observed cloud point generally lies between the binodal (coexistence) and spinodal (stability limit) curves. For polymer solutions, TCP determined via slow temperature variation often serves as a good approximation for the binodal curve (where distinct phases co-exist in equilibrium), as the rate of phase separation is dependent on nucleation of the phases, plus possibly the disassociation kinetics of the solvation structure. This assignment will be confirmed subsequently. The temperature-composition phase diagram in Figure 3 shows the cloud points, as determined by transmittance measurements exemplified in Figure 2, for three PBzMA molecular weights as a function of polymer concentration. The cloud point decreases monotonically with increasing concentration, with no suggestion of an approaching minimum. Flory-Huggins theory predicts that the critical composition, Фc, (the composition corresponding to the lowest TCP) should be shifted toward more dilute concentrations with increasing chain length (Фc ≈ N −1/2 , where N is the volumetric degree of polymerization with respect to a solvent molecule). However, it is observed that TCP is apparently independent of molecular weight, and that it is still decreasing as the concentration of polymer increases to 30/40 wt% PBzMA, indicating that Фc lies within the polymer-rich regime of the phase diagram, far above the expected values. An experimentally observed cloud point generally lies between the binodal (coexistence) and spinodal (stability limit) curves. For polymer solutions, T CP determined via slow temperature variation often serves as a good approximation for the binodal curve (where distinct phases co-exist in equilibrium), as the rate of phase separation is dependent on nucleation of the phases, plus possibly the disassociation kinetics of the solvation structure. This assignment will be confirmed subsequently. The temperature-composition phase diagram in Figure 3 shows the cloud points, as determined by transmittance measurements exemplified in Figure 2, for three PBzMA molecular weights as a function of polymer concentration. The cloud point decreases monotonically with increasing concentration, with no suggestion of an approaching minimum. Flory-Huggins theory predicts that the critical composition, Φ c , (the composition corresponding to the lowest T CP ) should be shifted toward more dilute concentrations with increasing chain length (Φ c ≈ N −1/2 , where N is the volumetric degree of polymerization with respect to a solvent molecule). However, it is observed that T CP is apparently independent of molecular weight, and that it is still decreasing as the concentration of polymer increases to 30/40 wt% PBzMA, indicating that Φ c lies within the polymer-rich regime of the phase diagram, far above the expected values. The reversibility of the phase transition indicated by the experimental cloud point was then explored by heating a 10 wt% PBzMA-63 in [BMP][TFSI] solution from 23 to 172 °C, and cooling from 172 to 23 °C, at a rate of ±1 °C/min for seven cycles. The solution was not allowed to anneal above TCP, thereby inhibiting the formation of individual layers. The temperature-dependent transmittance curve for the first cycle is shown in Figure 4a, where TCP upon heating and cooling are 167 and 161 °C, respectively, demonstrating that the mixture persists after phase separation until cooled a few degrees below TCP. This narrow hysteresis upon slow temperature variation (≈6 °C) confirms that the cloud point is a good approximation for the binodal temperature. It can also be seen that TCP increases modestly from 167 to 170 °C upon heating, and from 161 to 163 °C upon cooling, as the cycle number increases (Figure 4b); the hysteresis remains approximately constant.   The reversibility of the phase transition indicated by the experimental cloud point was then explored by heating a 10 wt% PBzMA-63 in [BMP][TFSI] solution from 23 to 172 • C, and cooling from 172 to 23 • C, at a rate of ±1 • C/min for seven cycles. The solution was not allowed to anneal above T CP , thereby inhibiting the formation of individual layers. The temperature-dependent transmittance curve for the first cycle is shown in Figure 4a, where T CP upon heating and cooling are 167 and 161 • C, respectively, demonstrating that the mixture persists after phase separation until cooled a few degrees below T CP . This narrow hysteresis upon slow temperature variation (≈6 • C) confirms that the cloud point is a good approximation for the binodal temperature. It can also be seen that T CP increases modestly from 167 to 170 • C upon heating, and from 161 to 163 • C upon cooling, as the cycle number increases (Figure 4b); the hysteresis remains approximately constant.  The reversibility of the phase transition indicated by the experimental cloud point was then explored by heating a 10 wt% PBzMA-63 in [BMP][TFSI] solution from 23 to 172 °C, and cooling from 172 to 23 °C, at a rate of ±1 °C/min for seven cycles. The solution was not allowed to anneal above TCP, thereby inhibiting the formation of individual layers. The temperature-dependent transmittance curve for the first cycle is shown in Figure 4a, where TCP upon heating and cooling are 167 and 161 °C, respectively, demonstrating that the mixture persists after phase separation until cooled a few degrees below TCP. This narrow hysteresis upon slow temperature variation (≈6 °C) confirms that the cloud point is a good approximation for the binodal temperature. It can also be seen that TCP increases modestly from 167 to 170 °C upon heating, and from 161 to 163 °C upon cooling, as the cycle number increases (Figure 4b); the hysteresis remains approximately constant.  To probe the heating/cooling rate sensitivity of the phase separation, the relative transmittance of 10 wt% PBzMA-32 in [BMP][TFSI] solutions was measured as a function of temperature for heating and cooling rates of 0.1, 0.5, 1.0, and 5.0 • C/min ( Figure 5). The tight grouping of the T CP values and the uniform sharp drop in transmittance upon reaching T CP for the heating rates under study further confirms that T CP upon heating lies close to the binodal curve; a factor of 50 increase in heating rate only changes T CP by 2 • C. On the other hand, the thermal hysteresis widens from 1 to 14 • C as the cooling ramp rate increases from 0.1 to 5.0 • C/min. This is quite reasonable, as re-mixing typically requires finite time. To probe the heating/cooling rate sensitivity of the phase separation, the relative transmittance of 10 wt% PBzMA-32 in [BMP][TFSI] solutions was measured as a function of temperature for heating and cooling rates of 0.1, 0.5, 1.0, and 5.0 °C/min ( Figure 5). The tight grouping of the TCP values and the uniform sharp drop in transmittance upon reaching TCP for the heating rates under study further confirms that TCP upon heating lies close to the binodal curve; a factor of 50 increase in heating rate only changes TCP by 2 °C. On the other hand, the thermal hysteresis widens from 1 to 14 °C as the cooling ramp rate increases from 0.1 to 5.0 °C/min. This is quite reasonable, as re-mixing typically requires finite time.

Discussion
Polymer solutions exhibiting strong and directional interactions, such as hydrogen bonding, can exhibit LCST phase behavior, and therefore do not conform to the assumptions of the basic Flory-Huggins theory (i.e., only weak, dispersive interactions). Many aqueous polymer systems have an LCST, due to the H-bonding network of water, and the tendency to form cage-like solvation structures. However, aqueous applications that exploit LCST behavior are limited to relatively low temperature windows [36,37]. Polymers in ILs are an emerging class of solutions that can exhibit this interesting phase behavior, presumably due to strong interactions that structure the solvent ions around the polymeric solute. In general, ILs show great promise due to their thermal stability, negligible vapor pressure, and tunable physico-chemical properties. As previously mentioned, small changes in the IL or polymer structure can cause large differences in the behavior of such systems.  [20,25,31]. We hypothesize that this dependence arises from interpolymer interactions between benzyl groups playing a pivotal role during phase separation. Previous work by Fujii et al. demonstrated that there are no significant interpolymer interactions between PBzMA chains in [EMIM][TFSI] at room temperature (below TCP) [30]. However, as the cation-π clathrate de-solvates during phase separation, benzyl groups from different chains begin to overlap as segments aggregate. By increasing the concentration of PBzMA, this stabilization becomes more prevalent and drives phase separation to lower temperatures by promoting the formation of solvent-rich and polymer-rich domains, favored by cation-π and π-π interactions, respectively.

Discussion
Polymer solutions exhibiting strong and directional interactions, such as hydrogen bonding, can exhibit LCST phase behavior, and therefore do not conform to the assumptions of the basic Flory-Huggins theory (i.e., only weak, dispersive interactions). Many aqueous polymer systems have an LCST, due to the H-bonding network of water, and the tendency to form cage-like solvation structures. However, aqueous applications that exploit LCST behavior are limited to relatively low temperature windows [36,37]. Polymers in ILs are an emerging class of solutions that can exhibit this interesting phase behavior, presumably due to strong interactions that structure the solvent ions around the polymeric solute. In general, ILs show great promise due to their thermal stability, negligible vapor pressure, and tunable physico-chemical properties. As previously mentioned, small changes in the IL or polymer structure can cause large differences in the behavior of such systems.  [20,25,31]. We hypothesize that this dependence arises from interpolymer interactions between benzyl groups playing a pivotal role during phase separation. Previous work by Fujii et al. demonstrated that there are no significant interpolymer interactions between PBzMA chains in [EMIM][TFSI] at room temperature (below T CP ) [30]. However, as the cation-π clathrate de-solvates during phase separation, benzyl groups from different chains begin to overlap as segments aggregate. By increasing the concentration of PBzMA, this stabilization becomes more prevalent and drives phase separation to lower temperatures by promoting the formation of solvent-rich and polymerrich domains, favored by cation-π and π-π interactions, respectively.
In contrast to Flory-Huggins theory, the critical composition of PBzMA in [BMP][TFSI] appears to be shifted to the polymer-rich regime of the phase diagram. A similar shift has also been observed in PEO/[EMIM][BF 4 ] solutions, with critical compositions varying from 50 to 80 wt% polymer [20,21]. The origin of this unusual phenomenon remains to be fully elucidated, but several mechanisms have been considered. White and Lipson hypothesize these shifts to be caused by a stronger cohesive energy density in the IL, and possible aggregation of the IL within polymer-rich mixtures [38]. However, these criteria should also be satisfied by poly(ethyl glycidyl ether) in [C n MIM][TFSI], whereas Watanabe and co-workers have revealed that the critical composition of this system is shifted toward dilute concentrations [18,19]. Another explanation, due to de Gennes, posits the formation of a dense polymer phase in equilibrium with the bulk solution due to inter-chain associations [39]. This could serve as a plausible explanation for PBzMA/IL solutions, as π-π stacking between benzyl groups would facilitate the formation of this phase, but recent light scattering studies have shown that PBzMA does not aggregate prior to phase separation in dilute solutions [31]. A third possibility presented by Schafer-Soenen and co-workers is that the interaction parameter is strongly dependent on the solution concentration [40]. This looks to be a promising path, as recent work from our group has already observed unusual behavior with PBzMA and ILs, in that in very dilute solutions, the second virial coefficient is positive, indicating good solvent behavior, at temperatures that exceed T cp at higher concentrations [31]. An increase in the interaction parameter with concentration could resolve this apparent contradiction, as this would result in a decrease in miscibility with higher concentrations. It would also imply that the second virial coefficient, which is inevitably measured in very dilute solutions, is not a reliable indicator of the phase boundary. Note that the independence of T CP on molecular weight for PBzMA in [BMP][TFSI] at high molecular weights (Figure 3) is not unusual for LCST systems. For example, the phase separation temperatures of PEO in [EMIM][BF 4 ] becomes independent of N above 20 kDa [21]. Within the context of Flory-Huggins theory, this result is a natural consequence of a negative enthalpy of mixing combined with a substantial unfavorable entropy of mixing [17].
In general, liquid-liquid phase separation in the metastable region of the phase diagram proceeds by nucleation and subsequent growth of droplets of the minor phase (in this case, the polymer-rich phase). The alternative is spinodal decomposition, but the near reversibility of the experimental cloud point, with significant hysteresis, suggests a nucleation and growth process. Cloud point values upon heating and cooling typically do not coincide, as phase separation and remixing are kinetic processes that occur from different initial conditions. However, the concentrations are quite high in this system, and therefore the possibility of becoming trapped in a very dilute, metastable one-phase state is correspondingly low. For PBzMA/[BMP][TFSI] solutions, the visible onset of phase separation upon heating is rapid, typically occurring within seconds upon crossing T CP . These results suggest rapid dissociation of the cation-π clathrate and aggregation of PBzMA segments, both of which are molecular scale processes that can occur on much shorter time scales than the accessible heating rates. On the other hand, re-solvation and re-mixing upon cooling are limited by the mutual diffusion of PBzMA and solvent. This becomes readily apparent when examining the reversibility of the phase transition. As samples are annealed above T CP , the formation of two phases is thermodynamically favored and PBzMA migrates preferentially toward the top of the vial ( Figure S3). If the system is cooled below T CP before distinct layers are formed, the polymer chains become re-solvated and begin to diffuse into the solvent-rich lower domains. The apparent increase in T CP as the cycle number increases (Figure 4b) suggests that the local polymer concentration within the path of the beam decreased. Without stirring, the change in local concentration after 10 min spent above T CP is not completely reversed after almost 5 h spent below T CP , indicating that remixing is mass-transport-limited. This is also exemplified by the widening of the cloud point hysteresis ( Figure 5) as the ramp rate increases. Thus, caution needs to be taken when handling solutions near T CP as their compositions are rapidly altered during phase separation and are not fully reversible over short time periods (<5 h).

Polymer Synthesis
PBzMA homopolymers were synthesized by atom transfer radical polymerization, adapting a procedure reported by Kharel et al. [31]. Copper (I) bromide, 1,1,4,7,10,10hexamethyltriethylenetetramine (HMTETA), ethyl 2-bromoisobutyrate (EBriB), and anisole were purchased from Sigma-Aldrich (St. Louis, MO, USA) and used as received. Copper (I) bromide was stored in an inert argon-atmosphere glove box. Benzyl methacrylate was purchased from Tokyo Chemical Industry (Tokyo, Japan) and purified via filtration through a neutral alumina column prior to use. A representative polymerization is described. A Schlenk flask outfitted with a magnetic stir bar was charged with benzyl methacrylate (33 g, 188 mmol), HMTETA (67 mg, 0.29 mmol), EBriB (58 mg, 0.30 mmol), and anisole (95 mL). The reaction vessel was sealed with a rubber septum and the reaction mixture was sparged with argon gas for 20 min. The vessel was then placed under positive pressure of argon. A needle was used to puncture the septum and relieve pressure within the vessel. Under a constant flow of argon, the septum was removed, and copper (I) bromide (0.25 mg, 0.17 mmol) was added to the reaction mixture. The Schlenk flask was sealed and the argon pressure within the vessel increased to 5 atm. The reaction was allowed to proceed at 80 • C under constant stirring. After 12 h, the reaction was quenched by opening the reaction mixture to atmosphere and placing the Schlenk flask in liquid nitrogen. The reaction solution was precipitated into cooled methanol, redissolved in dichloromethane (DCM), and filtered through a neutral alumina column to remove the copper complex. The residue was dissolved in DCM, precipitated into cooled methanol, and dried at 100 • C under dynamic vacuum (<200 mTorr) to afford PBzMA as a white powder. The degree of polymerization for three homopolymer samples was varied by changing the reaction time and ratio of monomer to initiator.
The weight-average molecular weight (M w ) and dispersity (Ð) of the synthesized homopolymers were obtained from size exclusion chromatography performed in tetrahydrofuran (THF) at room temperature on an Agilent 1260 Infinity System (Agilent Technologies, Santa Clara, CA, USA). A Wyatt Optilab T-rEX (Wyatt Technology, Santa Barbara, CA, USA) refractive index detector monitored the eluent ( Figure S1). The M w and Ð are summarized in Table 1 as determined from Zimm plots produced by a Wyatt Dawn Heleos II multiangle laser light scattering detector, using the value of 0.144 mL/g for the dn/dc of PBzMA in THF [31]. The synthesized polymers are designated PBzMA-x, where x indicates the M w of PBzMA in kg/mol. The polymer purity was assessed via 1 H NMR spectroscopy in CD 2 Cl 2 ( Figure S2) with a Bruker Avance III HD nanobay AX-400 spectrometer (Bruker, Billerica, MA, USA) with a shielded Ascend magnet and a SampleXpress autosampler.

Solution Preparation
The ionic liquid [BMP][TFSI] was purchased from IoLiTec GmbH (Heilbronn, Germany) and dried under a dynamic vacuum for 48 h. An 1 H NMR spectrum of the ionic liquid in CDCl 3 is shown in Figure S2. PBzMA and [BMP][TFSI] were combined gravimetrically and dissolved in THF as a co-solvent. The solutions were stirred until they achieved homogeneity and were then filtered using a 0.2 µm PTFE syringe filter. THF was evaporated under stirring during a nitrogen purge and the remaining solution was dried at 80 • C in an oven under a dynamic vacuum (<200 mTorr) for 12 h. Samples were stored in a vacuum desiccator prior to experiments.

Cloud Point Measurements
The PBzMA/[BMP][TFSI] solutions were flame-sealed in glass ampoules under reduced pressure (<130 mTorr) and were placed onto a temperature-controlled heating stage. Optical transmittance measurements were conducted by directing a HeNe laser (wavelength λ 0 = 632.8 nm) through the sample and focusing the beam on a photodiode detector while the stage temperature was heated or cooled at ±1.0 • C/min. For varying ramp rate experiments, four ampoules were charged with a 10 wt% PBzMA-32 solution, flame-sealed, and subjected to temperature ramps at ±0.1, ±0.5, ±1.0, and ±5.0 • C/min. Labview software (National Instruments, Austin, TX, USA) automated the temperature control and data collection processes, recording the laser intensity every 5 s. The raw intensity values were averaged at each temperature and normalized against the maximum intensity throughout the temperature ramp, yielding a relative transmittance at each temperature. The cloud point temperature (T CP ) is defined as the temperature when the transmittance drops to 80% of the value for the original homogenous solution.

Conclusions
The LCST-type phase behavior of PBzMA/[BMP][TFSI] concentrated solutions was explored for three molecular weights via optical transmittance. The phase separation temperatures demonstrate a strong dependence on concentration, decreasing as the incorporation of PBzMA increases, but are not influenced by molecular weight. The former trend is hypothesized to reflect a composition-dependent interaction parameter, while the latter is quite common in LCST systems. The critical compositions for these systems were shown to be shifted toward the polymer-rich regime of the phase diagram, in contrast to the solvent-rich regime anticipated by Flory-Huggins theory. These unusual characteristics add to the list of interesting solution phenomena exhibited by polymers in ILs. We also report the reversibility and thermal sensitivity of this phase transition to serve as a guide for polymer processing and thermo-responsive material design.