Direct Conversion of Phase‐Transition Entropy into Electrochemical Thermopower and the Peltier Effect

A thermocell generates thermopower from a temperature difference (ΔT) between two electrodes. The converse process of thermocells is an electrochemical Peltier effect, which creates a ΔT on the electrodes by applying an external current. The Seebeck coefficient (Se) of the electrochemical system is proportional to the entropy change of the redox reaction; therefore, a redox system having a significant entropy change is expected to increase the Se. In this study, a thermoresponsive polymer having a redox‐active moiety, poly(N‐isopropyl acrylamide‐co‐N‐(2‐acrylamide ethyl)‐N′‐n‐propylviologen) (PNV), is used as the redox species of a thermocell. PNV2+ dication undergoes the coil–globule phase transition upon the reduction to PNV+ cation radical, and a large entropy change is introduced because water molecules are freed from the polymer chains. The Se of PNV thermocell drastically increased to +2.1 mV K−1 at the lower critical solution temperature (LCST) of PNV. The entropy change calculated from the increment of Se agrees with the value evaluated by differential scanning calorimetry. Moreover, the electrochemical Peltier effect is observed when the device temperature is increased above the LCST. This study shows that the large entropy change associated with the coil–globule phase transition can be used in electrochemical thermal management and refrigeration technologies.


Introduction
Thermo-electrochemical conversion is a reversible energy transformation between heat and electrochemical potential, which is DOI: 10.1002/adma.202303341 gaining increasing attention in the field of waste-heat management and liquid cooling systems. [1][2][3][4][5][6][7][8][9][10][11][12][13] A thermocell generates a thermoelectric voltage (ΔV) by applying a temperature difference (ΔT) between two electrodes immersed in an electrolyte containing a redox pair. A converse system of thermocell is an electrochemical Peltier device, which creates a ΔT across the two electrodes by applying a current. [14,15] Seebeck coefficient (noted as S e with a unit of mV K −1 ) determines the values of ΔV of thermocells or the ΔT of Peltier devices. The S e of thermocells is an order of magnitude greater than a S e of typical solidstate thermoelectric materials, [16][17][18] which makes thermocells and electrochemical refrigerators a promising alternative to conventional thermoelectric devices. The S e of thermocells is proportional to the entropy change by a redox reaction (ΔS rc ): where n is a stoichiometric number of electrons exchanged in the redox reaction, and F is the Faraday constant. Equation (1) shows that a redox species with a large ΔS rc is advantageous to increase the S e . Conversely, a measurement of S e itself can be a convenient analytical method to estimate the ΔS rc , which otherwise requires a more cumbersome technique like a calorimetric analysis for the evaluation. [19,20] Previously, we reported that the addition of host molecules into thermocells can drastically increase the S e value because of the temperature-sensitive and guest-selective hostguest interaction. [21][22][23][24][25][26] The increment of S e caused by the host molecule can be explained by the concentration gradient of a guest species, [21,27] and also alternatively explained by using the entropy change induced by host-guest encapsulation. [1,2] These studies demonstrate a successful combination between supramolecular chemistry and thermo-electrochemical conversion.
Thermoresponsive polymers such as poly(N-isopropyl acrylamide) (PNIPAM) undergo a phase transition from hydrophilic coil phase to hydrophobic globule one at a lower critical solution temperature (LCST). [28][29][30] Addition of PNIPAM www.advancedsciencenews.com www.advmat.de and its copolymer in thermocells was recently discovered as an effective strategy to introduce a concentration gradient of triiodide or proton and increase the S e value according to the Nernst equation. [31][32][33] Pegylated -cyclodextrin shows LCST-type aggregation-dissolution behavior and its aggregation at the hot electrode increases the local concentration of triiodide. [25] Similarly, ethylated -cyclodextrin forms a complex with pentaiodide and causes the temperature-sensitive precipitation. [26] Temperature-sensitive precipitation of ferrocyanide (II/III) redox pair was also achieved by the addition of guanidinium cation, [4] or cationic surfactant. [9] These studies showed that the coilglobule or aggregation-dissolution phase transitions of the host molecules or host-guest complexes in the thermocell electrolyte can induce the concentration gradient of the guest redox species and increase the S e values.
Theoretical studies on the coil-globule phase transition of polymer materials showed that the phase transition is accompanied by a large entropy change (ΔS trs ). [34,35] The ΔS trs is expected to enhance the S e according to Equation (1), however, the previous studies were focused only on the concentration gradient of the guest species created by the phase-changing host materials, and the relationship between the ΔS trs and the S e manifested in Equation (1) remains elusive. To observe the direct conversion of ΔS trs into S e , the redox-active moiety must be chemically bonded to the polymer main chain so that the redox center can experience the same solvation entropy as the polymer chains do. We noticed that polymer materials showing both redox-activity and the coil-globule phase transition are suitable candidates for this purpose. [36][37][38][39][40] Herein, poly(NIPAM-co-N-(2-acrylamide ethyl)-N'-npropylviologen) (PNV) is studied as a model LCST-type redox polymer to evaluate the relationship between ΔS trs and S e of thermocell, and its electrochemical Peltier effect is measured for the first time (Figure 1a). At room temperature, the oxidized PNV 2+ is a hydrophilic cationic polymer and takes the coil state, while the reduced PNV + becomes more hydrophobic and takes the globule state ( Figure 1b). We expect that a reduction of PNV 2+ to PNV + above the LCST can produce a large entropy change and hence a considerable S e value because the water molecules trapped around the hydrophilic coils are released at the phase transition to from the hydrophobic globules. Experimental results show that the S e of PNV 2+/+ thermocell is increased from +0.09 to +2.1 mV K −1 when the hot-side temperature exceeds the LCST. The ΔS trs of PNV estimated from the increment of S e at the LCST is in good agreement with the ΔS trs value evaluated by a differential scanning calorimetry (DSC). Furthermore, an electrochemical Peltier effect of the PNV 2+/+ redox pair is observed above the LCST, which is the first demonstration of a reversible thermo-electrochemical conversion using the coil-globule phase transition. This study could be an important milestone showing that the phase transition of redox-active materials can dramatically improve the performance of thermocells and electrochemical Peltier devices.

Results and Discussion
The synthesis of PNV was carried out following the method reported by Okeyoshi and Yoshida, [40] with minor modifications (see Figures S1-S5, Supporting Information, for the full descrip- tions). The average molecular weight of the synthesized PNV was evaluated by gel permeation chromatography (GPC) as M w = 1.8 × 10 4 ( Figure S6, Supporting Information). 1 H-NMR measurement estimates the molar ratio of the viologen group incorporated in the PNV chain as 6 mol% ( Figure S7, Supporting Information). The as-synthesized PNV is in an oxidized form (PNV 2+ ), expressing a pale-yellow color when dissolved in water. The addition of sodium dithionate (Na 2 S 2 O 4 ) into the PNV 2+ solution causes a one-electron reduction of the viologen group to produce PNV + : [40][41][42] The solution turns into an intense-purple color after the reduction. UV-vis-NIR spectra of PNV + shows multiple broad absorption peaks between 200 and 1200 nm (Figure 2a). The coilglobule phase transition temperature of PNV 2+ and PNV + was evaluated by turbidimetry with an incident wavelength ( ) of 500 and 1200 nm for PNV 2+ and PNV + , respectively. The of 1200 nm was selected to mitigate the influence of the strong photoabsorption of PNV + . A sharp decrease in the transmittance along with an increasing temperature indicates the transformation from the coil state to the globule state, confirming an LCSTtype phase transition in both PNV 2+ and PNV + (Figure 2b,c). The cloud-point temperature (T cp ) was evaluated as the temperature at which the transmittance decreased to 50% of the initial value, and its dependency on the concentration of the polymer was plotted in Figure 2d. The T cp of PNV + is ≈20°C lower than that of PNV 2+ in the concentration range of 0.2-4 wt%. A redox reaction between PNV + and PNV 2+ is expected to induce the reversible coil-globule phase transition in the temperature range between the T cp of PNV + and PNV 2+ . The phase transition temperature was also evaluated by a DSC analysis (Figure 2e), and the onset temperature (T trs ) of the endothermic process agreed with the T cp measured in the turbidimetry. Phase transition enthalpy (ΔH trs ) per weight of PNV dissolved in the solution is shown in the inset of Figure 2e. The ΔS trs (= ΔH trs /T trs ) is calculated as 52 mJ K −1 g −1 for PNV + and 24 mJ K −1 g −1 for PNV 2+ . The larger ΔS trs of PNV + compared to PNV 2+ suggests a greater entropic gain of PNV + by forming hydrophobic globules (Figure 1b). In addition, the larger ΔH trs of PNV + possibly resulted from the bond between the viologen cation radicals. [43] Cyclic voltammetry (CV) of PNV was carried out at ambient temperatures (20-23°C) and 40°C on a platinum working electrode. Two reversible redox waves of the one-electron reaction, shown in Equation (2) were observed (Figure 3a). Interestingly, when the solution temperature was increased to 40°C, the peak current density of the oxidative wave significantly increased, and a sharp oxidative peak was observed. As a control experiment, CV of N,N'-di-n-propylviologen (DPV) was carried out at the same conditions. However, the peak current density of DPV shows only a slight increment when the temperature was increased to 40°C, and no sharp oxidative peak was observed. Therefore, the sharp oxidative peak is unique to PNV and possibly caused by cooperative oxidation among the viologen groups accumulating on the electrode surface. Similarly, Sagara and co-workers reported a one-electron transfer reaction of diphenyl viologen resulted in a sharp spike-like peak in a CV measurement because of the formation of the ordered -stacked cation radicals on a gold electrode surface. [44] Electrochemical quartz crystal microbalance (EQCM) was used to study the redox-induced aggregation of PNV on a gold electrode (Figure 3b). A constant reductive voltage  (−0.7 V vs Ag/AgCl) and an oxidative voltage (0 V) were applied in sequence to the electrode, and the relative electrode weight (ΔW) started to increase upon the reduction of PNV and spontaneously decreased to the initial ΔW upon its oxidation. The increment of ΔW was pronounced when the temperature was increased to 40°C. In contrast, no increment of ΔW was observed during the redox reaction of DPV. Therefore, we conclude that the increment of ΔW was caused by the reduction-induced LCSTtype aggregation of PNV on the electrode surface, as illustrated in Figure 3c. The spontaneous decrease of ΔW at the oxidation step explains that the spiky oxidative peak observed in the CV experiment (Figure 3a) is caused by a sharp phase transition of PNV globules accumulated on the electrode surface.
A thermocell was assembled with two platinum electrodes and an aqueous electrolyte of PNV 2+ (2 wt%) reduced by 2.5 mm Na 2 S 2 O 4 (Figure 4a , the cell design is shown in Figure S8, Supporting Information). UV-vis spectroscopic analysis revealed that a 2% PNV 2+ solution contains a 9 mm viologen unit ( Figure  S9, Supporting Information). An addition of 2.5 mm Na 2 S 2 O 4 can theoretically reduce 5 mm PNV 2+ , and therefore the molar ratio of the redox pair in a 2% PNV electrolyte is estimated to be PNV 2+ :PNV + ≈ 1:1. The cold-side temperature (T cold ) of the thermocell was maintained at 25°C, and the T hot was slowly increased from 25°C to 45°C, while the ΔV between the hot and cold electrodes was recorded (Figure 4b). A plot of ΔV vs ΔT (= T hot − T cold ) shows a drastic increase of the S e from +0.09 to +2.1 mV K −1 at ΔT = 10 K (T hot = 35°C, Figure 4c). The transition point of S e locates between the T cp of PNV + (22°C) and PNV 2+ (42°C). When the concentration of Na 2 S 2 O 4 is increased to 5 mm, the transition point of S e decreased to ΔT = 5 K (T hot = 30°C, Figure S10, Supporting Information). This result suggests that the transition point of S e corresponds to the T cp of the mixed solution of PNV + and PNV 2+ , which is determined by the molar ratio of the reductant and oxidant. As a control experiment, PNV was replaced with DPV molecules (9 mm reduced by 2.5 mm of Na 2 S 2 O 4 ); however, only a small and constant S e of −0.3 mV K −1 was observed. The negative S e value was observed in DPV, possibly because the dimerization between the two DPV + radical cations results in a negative entropy change at the reduction step. [43] Therefore, we confirmed that the drastic increment of S e in the T hot region between the two T cp values of PNV + and PNV 2+ is caused by the LCST-type phase transition.
The total ΔS calculated from Equation (1) using the increment of the S e at the transition point (+2.0 mV K −1 ) was 190 J K −1 per mole of viologen. In comparison, the ΔS trs of PNV + measured in the DSC analysis (Figure 2e) corresponds to 120 J K −1 per mole of viologen, which demonstrates the large ΔS trs of PNV + contributes to the drastic increase of S e value The long-term stability of the PNV thermocell was evaluated by monitoring the open-circuit voltage (V oc ) and the short-circuit current density (J sc ) while fixing the T hot at 45°C and the T cold at 35°C (ΔT = 10 K). A constant V oc of ≈18-20 mV was observed over 1000 s (Figure 4d). The slight decrease of V oc could be caused by the gradual precipitation of PNV at the hot electrode ( Figure  S18, Supporting Information). In contrast, the J sc dropped from 12 to 6 μA cm −2 in the first 100 s and then stabilized at 5-6 μA cm −2 for the next 500 s (Figure 4e). In comparison, the thermocell with molecular DPV showed a constant J sc of 4 μA cm −2 over 1000 s ( Figure S11, Supporting Information). This result suggests that a slower diffusion rate of the polymer PNV compared to the discrete DPV might cause the decreasing J sc observed in the PNV thermocell. The slow diffusion rate of a redox polymer may be overcome by using a flow-type thermocell, [6,45] and various types of redox flow batteries using redox-active polymers have been reported recently. [46][47][48][49] The maximum power density (P max ) of the PNV thermocell evaluated after a stabilization time of 120 s is an order of magnitude larger than that of the DPV thermocell, demonstrating the significance of attaching a redox molecule on a thermoresponsive polymer chain to improve the performance of thermocells (Figure 4f).
The converse process of a thermocell is the electrochemical Peltier effect, which creates a temperature difference between the electrodes by applying an external voltage. Electrochemical refrigeration has several advantages such as the integration into a liquid-cooling system to continuously carry away waste heat from the heat source. [14,15] Newman derived an equation balancing a heat flow and a heat generation in an electrochemical system: [50] −k∇T + k I ∇T I = i s + iΠ (3) k and k I are the thermal conductivity of the electrolyte and the electrode phase, respectively, ∇T is a temperature gradient in the vertical direction to the electrode, i is the current, s is the surface overpotential, and Π is the Peltier coefficient. Assuming the Dufour effect (i.e., the converse process of the Soret effect) is neg-ligibly small compared to the entropic one, the Peltier coefficient is proportional to the S e : [50] Π ≈ T nF The i s term in Equation (3) accounts for an irreversible heat generation associated with Joule heating, and the iΠ term accounts for reversible heat generation. Because the Joule heating (i s ) always takes a positive value independent of the direction of the current flow, we can disregard its influence by alternately applying a positive and negative current, which is an established method in the field of solid-state thermoelectrics. [51] Then, we can extract the reversible component of the heat generation (iΠ).
A positive and negative current (± 0.3 mA cm −2 ) was alternately applied to the PNV electrolyte for 4 s per step, and the ΔT of one of the electrodes was monitored (see Figure S12, Supporting Information for the cell design). Because the ΔT values were minimal (<1 mK), this measurement was repeated for 20 000 cycles and the ΔT values were averaged to improve the signal-to-noise ratio (Figure 5a; the raw data is shown in Figures  S13-S16, Supporting Information). The measurement was carried out at various temperatures, and the maximum ΔT of ≈−0.5 mK was observed in the reduction step at both 32 and 48°C, while no temperature change was observed at 25°C. The control measurement was carried out with the molecular DPV at 48°C; however, no temperature change was observed. These results suggest that the large ΔS trs of PNV is responsible for the observed electrochemical Peltier effect (Figure 5b). To the best of our knowledge, this is the first report of electrochemical refrigeration using the coil-globule phase transition of a redox polymer. Although the ΔT observed in this study is far too small for practical use, we believe the value could be improved by www.advancedsciencenews.com www.advmat.de increasing the concentration of a redox-active species and by applying a larger current.
We evaluated the S e at a higher PNV concentration (20 wt%) as +1.5 mV K −1 , and the transition point was observed at T hot = 30°C ( Figure S17a, Supporting Information). However, increasing viscosity of the electrolyte and severe precipitation of PNV + at such a high concentration decreased the output current and power ( Figure S17b, Supporting Information). Further materials optimization such as the formation of polymer nanoparticles [32] or the growth of polymer chains onto inorganic nanoparticles [29,37] may result in better particle dispersion and prevent their precipitation.

Conclusion
PNV was used as a model redox-active and thermoresponsive polymer to demonstrate that the entropy change associated with the coil-globule phase transition can increase the S e of the thermocell. A drastic increase of the S e from +0.09 to +2.1 mV K −1 was observed at the LCST of PNV 2+/+ mixed electrolyte. The corresponding entropy change (190 J K −1 mol −1 ) estimated from the increment of S e agrees with the entropy change evaluated from the DSC analysis (120 J K −1 mol −1 ). The electrochemical Peltier effect was observed when the cell temperature is above the LCST. As a control experiment, the molecular DPV was used as the redox species instead of the polymeric PNV; however, no increment of the S e and no Peltier effect was observed. This study demonstrates that the phase-transition entropy of redox-active polymers can make a significant contribution in the field of electrochemical thermoelectric conversion.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.