Aqueous thermogalvanic cells with a high Seebeck coefficient for low-grade heat harvest

Thermogalvanic cells offer a cheap, flexible and scalable route for directly converting heat into electricity. However, achieving a high output voltage and power performance simultaneously from low-grade thermal energy remains challenging. Here, we introduce strong chaotropic cations (guanidinium) and highly soluble amide derivatives (urea) into aqueous ferri/ferrocyanide ([Fe(CN)6]4−/[Fe(CN)6]3−) electrolytes to significantly boost their thermopowers. The corresponding Seebeck coefficient and temperature-insensitive power density simultaneously increase from 1.4 to 4.2 mV K−1 and from 0.4 to 1.1 mW K−2 m−2, respectively. The results reveal that guanidinium and urea synergistically enlarge the entropy difference of the redox couple and significantly increase the Seebeck effect. As a demonstration, we design a prototype module that generates a high open-circuit voltage of 3.4 V at a small temperature difference of 18 K. This thermogalvanic cell system, which features high Seebeck coefficient and low cost, holds promise for the efficient harvest of low-grade thermal energy.


Supplementary Figures
Supplementary Figure 1| Test of the Seebeck effect by a planar device. a, Photograph of the planar device. The device consisted of two graphite electrodes and a polyamide frame. The distance between the two electrodes was 15 mm. The hot side was heated by a metal ceramic heater (MCH). The cold side was cooled by a thermoelectric cooler contacting a water-cooling plate. The thermocouple wires were inserted into holes on the top of the graphite electrodes. b, The instantaneous voltage for four TGC systems at temperature differences of 0 to 5 K. At a T of 5 K, the temperatures at the cold side and the hot side were ca. 298 and 303 K.  Supplementary Figure 11| Thermal conductivity measurement. a-b, Schematic and photograph of the cell used for the thermal conductivity measurement. c, The infrared thermography and corresponding temperature gradient across the box were measured by an infrared imaging device (TiX520, Fluke, USA).

Supplementary Figure 12| The temperature differences in the different systems.
Dependence of the temperature differences across the TGC on heat flux. The temperature difference increased with increasing heat input. For the urea and GdmCl systems, the increase in the temperature difference slowed at high heat input due to extensive heat convection. However, the temperature difference for the UGdmCl systems linearly increased with the heat input, further indicating suppression of convection. The error bar is received by measuring temperature gradient at three different positions on the cross-section of cells by an infrared imaging device as shown in Supplementary Figure 11c. To illustrate the operation at cold temperature, the top side of the module was covered with an ice bag (~273 K), and its other side was contacted with a steel platform. At the steady state, a voltage of ~1.5 V was achieved at a temperature difference of ~7.9 K. The average S e value (S e =1.5/50/7.9) for the module was calculated to be 3.  The SPC/Fw model 4 , which is capable of taking into account water flexibility 2 , was adopted for water. Furthermore, the force field parameters for urea and GdmCl were based on the OPLS model 5,6 . The detailed simulation setup with the number of ions/molecules is summarized in Table 1 for all studied simulation systems. Periodic boundary conditions were used in all three directions of the MD system. A cut-off distance of 1.4 nm was employed for both electrostatic interactions and van der Waals terms, whereas the long-range electrostatic interactions were accounted for through the particle mesh Ewald (PME) method 7 . Temperature and pressure were controlled through the Berendsen 8 weak-coupling scheme with coupling constants of 0.1 and 1 ps, respectively. The leapfrog integration algorithm was used to solve the equations of motion with a time step of 0.5 fs. The trajectory was saved every 0.5 ps. Each simulation was equilibrated within 10 ns. Another 20-ns production run was subsequently performed for analysis.
The radial density profiles between the centres of mass of the ion and water molecules when adding different amounts of GdmCl, urea and UGdmCl are shown in Supplementary Figure 2a 6 ] 4− decrease to a certain degree. Although the size of the first solvation shell does not change greatly, some water molecules in the first solvation shell are squeezed out. To provide a quantitative description, the hydration energy of the first solvation energy is calculated 10 (Supplementary Figure 3). The hydration energy of the ferri species decreases from -1276.48 kJ mol −1 to -1007.25 kJ mol −1 , -1091.18 kJ mol −1 and -837.71 kJ mol −1 when adding GdmCl, urea and urea/GdmCl, respectively. For the ferro species, the hydration energy decreases from -1804.42 kJ mol −1 to -853.21 kJ mol −1 , -1568.97 kJ mol −1 and -671.86 kJ mol −1 , respectively.
A deep insight into the decrease in the hydration capacity of the [Fe(CN) 6 ] 3− and [Fe(CN) 6 ] 4− ions can be obtained by examining the interactions between these two anions and other species, as well as the pair correlation function between these anions and water (Supplementary Figure 2a-b). When GdmCl salt is added, due to its higher charge, Gdm + has a stronger Coulomb interaction with the [Fe(CN) 6 Supplementary Figure 10. The water molecules pack the first solvation shell and then compactly interact with urea molecules and Gdm + cations. However, there are some differences in the solvation shell between the ferri and ferro species. The value of the first peak of urea with the ferri species is slightly larger than with the ferro species. Nevertheless, Gdm + has a sharp peak with the ferro species that is 5.3 times higher than that for the ferri species.
To better describe the degree of destruction of the hydration shell, the coordination numbers for water molecules around the two anions are calculated according to the following Equation: (1) Here, min is the first valley of the RDF between the anion and water, and is the average number density of the water molecules. As shown in Supplementary Figure  9, the coordination number decreases as expected. The hydration shell is destroyed when GdmCl is added; that is, the coordination number of water decreases by 3.50. However, the hydration shell is destroyed more thoroughly for the ferrocyanide anion (its coordination number decreases from 23.58 to 12.03) because of the stronger Coulomb interaction between ferrocyanide and Gdm + . When urea is added, the ferri species' coordination number is reduced by 4.11, and the coordination number of the ferro species is reduced by 1.96. The damaged solvation shell induced a higher entropy, thus producing a larger Seebeck coefficient.   11 . Ions are usually categorized as chaotropes or kosmotropes based on their perceived influence on the water structure. The ions on the left side of the series, defined as kosmotropes, exhibit strong interactions with water molecules, whereas the ions on the right side of the series, defined as chaotropes, are weakly hydrated by water molecules. In general, the chaotropic (kosmotropic) cations and chaotropic (kosmotropic) anions tend to bond together based on the ion specificity effect, whereas the opposite phenomena are observed between the kosmotropic cation (anion) and the chaotropic anion (cation) 12 .