The conversion of abundant low-grade thermal energy into useful electrical energy is essential for sustainable society development. Thermoelectric (TE) materials can convert waste heat generated by industry, fossil fuels, sunlight and even the human body directly into electricity with no moving parts2,10,11. Organic thermoelectric materials exhibit excellent thermoelectric properties near room temperature. This makes them ideal for low-grade heat recovery and for providing power to the devices on the internet of things (IoT)12–14. But these devices often require a relatively high voltage (> 1.5 V) and power density15. So, improving the thermopower, also known as Seebeck coefficient (S), and figure of merit (ZT) of TE materials is crucial for such applications. Furthermore, compared to traditional electronic TE materials, ionic thermoelectric (i-TE) materials exhibit superior thermoelectric and mechanical properties16–18. i-TEs are typically based on polymers that can be fabricated into ion conductors19,20, especially ionic gels5,21.
The working principle of i-TE materials is based on the Soret effect, whereby a temperature gradient induces an inhomogeneous distribution of cations and anions, resulting in a voltage difference. Electrostatic induction between the ions and electrons in the electrode enables charge storage and release to a load in an external circuit. Thus, it is regarded as a thermionic capacitor3,22,23. Previous reports focused on the role of polymer matrix in obtaining large thermopower, also known as Seebeck coefficient7,24,25. For example, the Seebeck coefficient of NaOH aqueous solution is less than 1 mV K− 1, but it can reach up to 11 mV K− 1 when incorporated into polyethylene oxide (PEO)3. Furthermore, the Seebeck coefficient can vary from − 4 to + 14 mV K− 1 through changing the interaction strength between poly(vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP) and the 1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide (EMIM:TFSI) ionic liquid depending on PEO content6. Furthermore, modified cellulose containing charged nanochannels generates a Seebeck coefficient of 24 mV K− 1 7. Similar strategy has also been applied to quasi-solid ionic gels26,27. For example, through the dipole moment interaction between PVDF-HFP and an ionic liquid 1-ethyl-3-methylimidazolium dicyanamide (EMIM:DCA), the thermal voltage reaches 26.1 mV K− 1 8. At present, the record Seebeck coefficient for all reported i-TE materials is 34.5 mV K− 1, achieved through a combination of waterborne polyurethane and EMIM:DCA28. The examples provided clearly demonstrates the dependency of Seebeck coefficient on polymer type, and its surface properties which dictate its interaction with ions. However, whether the thermopower can be enhanced via manipulating the intrinsic physical properties of ions remains to be demonstrated.
In this work, the properties of cations and anions are investigated separately in the context of ionic thermoelectrics. A series of inorganic compounds including KOH, KNO3, KCl, KBr, KI, NaI or CsI, is incorporated into polyvinyl alcohol (PVA) hydrogels. For all the potassium compounds, the resultant Seebeck coefficient changes with the structure breaking strength of the anions. By comparing the three iodides, the cation with smaller electronegativity and smaller binding energy with the hydroxyl groups on PVA shows greater Seebeck coefficient. After optimization, the PVA/CsI hydrogel exhibits a record Seebeck coefficient of 52.9 mV K− 1. Furthermore, an extremely high thermoelectric figure of merit (ZT) of 5.09 is obtained in the PVA/NaI hydrogels, which is more than three times the present ZT record. In addition to the excellent TE performance, our ionic hydrogels also have the advantages of facile preparation, low cost and mechanical robustness. These features together suggest hydrogels containing properly selected ions have great potential for low-grade heat harvesting.
For potassium compounds incorporated into the PVA hydrogel, the Seebeck coefficient changes from 17.7 mV K− 1 in PVA/KOH hydrogel to 49 mV K− 1 in PVA/KI hydrogel (Fig. 1a), while the ionic conductivities of these hydrogels are comparable (Supplementary Fig. 1). By comparing the hydrogels containing the three iodides, we found the Seebeck coefficient exhibited a trend, i.e. NaI < KI < CsI. Notably, the Seebeck coefficient of PVA/CsI hydrogel at 0.1 M reaches as high as 52.9 mV K− 1 (Fig. 1a and Supplementary Fig. 2a). To the best of our knowledge, this is the highest Seebeck coefficient reported for ionic thermoelectric materials till date (Fig. 1b). The result is highly reproducible (see Note 1 in Supplementary Information).
Figures 1c-e and Supplementary Fig. 4 show the thermoelectric properties of the modified PVA hydrogel as a function of CsI, KI or NaI concentration. The maximum Seebeck coefficient in all cases occur at a concentration of about 0.1 M (Fig. 1c, and Supplementary Fig. 4a). This trend is analogous to that observed in electronic thermoelectric materials29. However, note that ionic conductivity exhibits a monotonous increase with increasing salt concentration in all three hydrogels (Fig. 1d, and Supplementary Fig. 4b).
However, PVA/NaI hydrogel at a concentration of 1 M exhibits the best PF of 9.47 mW m− 1 K− 2, due to its large ionic conductivity (51.8 mS cm− 1) and a Seebeck coefficient of 42.8 mV K− 1. The PVA/NaI hydrogel shows a low thermal conductivity of 0.543 W m− 1 K− 1 (Supplementary Fig. 5b), yielding a record ionic ZT value of 5.09. This ionic ZT value is more than three times the previous ZT record26 for ionic conductors in literature (Fig. 1f).
In order to understand the role of the ions, we have conducted a systematic investigation on our i-TE system. First, potentially influencing factors such as electrode type for Seebeck measurement, water evaporation effects, and changing valence states of mobile ions are excluded (see Note 2 in S Supplementary Information). Afterwards, density functional theory (DFT) calculations were employed to reveal the cations effect on Seebeck coefficient. The binding energy of the cations towards PVA, from DFT calculations show a similar trend with their electronegativity (Fig. 2a). Thus, Cs+ which is furthest from the nearest oxygen atom on PVA (Supplementary Fig. 11) and less bound by hydroxyl groups on PVA, can migrate most rapidly within the hydrogel and thus produces the highest Seebeck coefficient.
Figure 2b shows a digital picture of the PVA hydrogels in inverted vials incorporated with five different potassium compounds. There is a difference in the fluidity in these samples. The PVA/KI hydrogel is on the verge of collapse and it has the best fluidity, followed by PVA/KBr, PVA/KCl and PVA/NO3; while PVA/KOH has the poorest fluidity. The is also reflected by morphological changes at the micro scale. Scanning electron microscopic (SEM) images (Supplementary Fig. 12) indicate the pristine PVA hydrogel and PVA/KOH sample exhibit a rough surface, implying a strong inter-molecular interaction among polymer chains. In contrast, when potassium halide or KNO3 is added, the surface of the hydrogel becomes smooth. Atomic force microscopic (AFM) images further confirm this observation. The pristine PVA hydrogel shows a roughness of 61.9 nm while the roughness in PVA/KOH and PVA/KI samples is 2.8 nm and 1.0 nm (Supplementary Fig. 13), respectively. This indicates that the ions can reduce the inter-molecular interaction between polymer chains.
Such a change in interaction is quantified by measuring the viscosity of these hydrogels (Fig. 2c). The viscosity follows a trend of PVA/KI (188 mPa·s) < PVA/KBr (212 mPa·s) < PVA/KCl (231 mPa·s) < PVA/KNO3 (235 mPa·s) < PVA/KOH (246 mPa·s) < pristine PVA (308 mPa·s). This implies that interaction among PVA chains is weakened with the addition of the potassium compounds, and different anions exhibits various degrees of weakening effect.
Fourier transform infrared spectroscopy (FT-IR) also confirmed this observation. The absorption peak within 3200–3600 cm− 1 relates to hydroxyl stretching mode30. Figure 2d shows the peak intensity of this mode in PVA/KI hydrogel is higher than that of the pristine and PVA/KOH hydrogels, indicating that more hydroxyl groups can stretch in PVA/KI hydrogels than in other two samples. Moreover, the absorption peak of PVA/KI is blue-shifted compared with pristine and PVA/KOH hydrogels, indicating weaker hydrogen bonding31. This further proves ions can break the hydrogen bonds among PVA chains, with I− ions exhibiting stronger ability for hydrogen bond breaking than OH− ions.
A measurement of the water retention ability (see Note 3 in Supplementary Information) of the salt-incorporated hydrogels (Fig. 2e and Supplementary Fig. 14) reveals PVA/KI hydrogel retained the highest water content after 210 min. This suggests I− ions liberate more hydroxyl groups that can hold more water molecules, resulting in slower water evaporation.
According to Y. Marcus et al., a water molecule forms 1.55 hydrogen bonds on average with surrounding molecules at room temperature32. The standard molar Gibbs free energy of transfer ΔGHB characterizes changes in local water structure (hydrogen bonding network) when perturbed by a solute ion32–35. A more positive (negative) ΔGHB value signifies a greater degree of promotion (destruction) of hydrogen bonds. All five anions used in this study exhibit a negative ΔGHB (Fig. 2e), meaning they are all structure breakers33. More specifically, the absolute ΔGHB value of the five anions follows a trend of -OH− < NO3− ≈ Cl− < Br− < I−. This trend is consistent with the above-mentioned results on TE performance and structural analyses.
Based on these results, a physical model is proposed to explain the ion-polymer interaction on the thermoelectric performance of our i-TE materials. As mentioned, ionic thermal voltage stems from the Soret effect (Fig. 3a). The Seebeck coefficient obtainable by the thermal diffusion of ions with dissimilar mobilities can be expressed as4,22:
where the subscript + (-) represents a cation (anion), e is the elementary charge, D and Ŝ are the mass diffusion coefficient and the Eastman entropy of transfer, respectively. Eq. (1) indicates the greater the difference between the mass diffusion coefficient or Eastman entropy of transfer of the cations and anions, the larger Seebeck coefficient is obtainable.
The dissolved ions in water can influence the nature of hydrogen bonds in their vicinity, i.e. either increasing (kosmotropic effect) or destroying (chaotropic effect) the local order around these ions4,6,36. If an ion promotes (disrupts) the order of the surrounding bonding network, it is called a structure maker (breaker)33,37,38. The degree of disruption of hydrogen bonding network varies for ions with different ΔGHB values, leading to different hydration shell structures39 (as sketched in Fig. 3b and 3c).
The different hydration shell structures caused by structure breakers have a direct impact on the whole hydrogel system. The structure breaker can disrupt the hydrogen bonds formed by the hydroxyl groups between adjacent PVA chain segments, in addition to the hydrogen bonds between water and structure breaker. On one hand, hydroxyl groups from disrupted hydrogen bonds are exposed to form stronger interaction with ions or H2O40. On the other hand, the disruption of the hydrogen bonds among water molecules will thin the hydration shell of ions, making ions more exposed and interacting more strongly with the hydroxyl group (as shown in Fig. 3e). Conversely, if the structure maker ion promotes the formation of hydrogen bonds, the ion will promote the cross-linking between polymers, as sketched in Fig. 3d. These two effects will ultimately slow down the movement of anions with structure breaker characteristics.
Different from anions, the chaotropic effect of cations is not obvious41. The electronegativity of the cation determines the binding strength between the cation and the polar groups on the polymer chains, and thus affects the thermal diffusion rate of the cations. Besides the above-mentioned ion-polymer interactions, we do not exclude possible contribution from water evaporation that occurs for all hydrogels.
Because ions cannot flow through external circuits, ionic thermoelectric generators convert heat to electricity conversion in the form of capacitors3,23. Details regarding capacitance performance test can be found in the Supplementary Information. Figure 4a shows the four operation stages of an ionic thermoelectric capacitor3,8,17,42. First, the voltage increases during charging process under a temperature gradient (Stage I). Then, the charge accumulated on the electrode is released to the external circuit to do work when a load is connected (Stage II). Upon removal of the temperature difference and the external circuit is disconnected, the ions diffuse back to their initial state (Stage III). Finally, the external circuit is closed and the accumulated charges on the electrode flow through the load again (Stage IV).
The stage II time discharge curves and power density as a function of external load resistance are shown in Figs. 4b and 4c. The power density profile exhibits a parabolic relationship, with a maximum reaching1.31 J m− 2 when the external load is 68 kΩ. The large power density generated under such a small temperature difference verifies the superior thermoelectric performance of our hydrogel.
It should be noted that the thickness of the hydrogel has a huge effect on the charging time and discharge power density, but does not affect its Seebeck coefficient. For example, the discharge time for a 300 µm PVA/NaI hydrogel is 1600 s, while that of a thicker (1000 µm) sample is 10000 s (Supplementary Fig. 16). The increase in charging time brings about simultaneous increase in power density (from 0.34 to 1.25 J m− 2) (Supplementary Fig. 17) due to an increased amount of charges stored in the gel network. Thus, a longer charging time leads to an increase in power density.
The quasi-continuous operation mode of the ionic thermoelectric capacitor is evaluated. After charging for ~ 1800 s under a temperature difference of ~ 1.4 K, the device reaches a saturation voltage of ~ 60 mV (Fig. 4d). Afterwards, the capacitor is connected to an external circuit with a load of 1 kΩ and discharged to almost 0 V in 1 ~ 3 s. Then it is recharged to its original voltage within 3 ~ 5 s (Fig. 4e and Supplementary Video 1). This charging rate is faster than other reported ionic thermoelectric capacitors22,27. Our hydrogel can easily complete more than 100 charge/discharge cycles within ~ 600 s, showing good cycle stability (Fig. 4d).
Despite the good cyclability and fast charge/discharge rate, the hydrogels are transparent in the visible region. The transmittance is more than 94% in the wavelength range between 400 nm and 800 nm (Fig. 4f). Finally, we demonstrate that good adhesion of the PVA/NaI hydrogel makes it suitable for wearable electronics. The hydrogel on a flexible PET substrate retains its original shape and does not de-laminate even when subjected to repeated bending and distortion (Supplementary Fig. 18 and Supplementary Video 2).
In summary, we have developed a series of excellent ionic thermoelectric materials, which are based on PVA/alkali metal compound hydrogels. The mechanistic study shows that electronegativity of the cations and Gibbs free energy of transfer of the anions are two key factors in modulating the ion-polymer interaction in our material system. The Seebeck coefficient of the PVA/CsI hydrogel prepared in this work reaches a record of 52.9 mV K− 1. The PVA/NaI hydrogel exhibits both high Seebeck coefficient (42.8 mV K− 1) and good ionic conductivity (51.5 mS cm− 1), leading to a superior ZT value of 5.09. Transparent, flexible and robust ionic thermoelectric capacitors are demonstrated. We believe the high-performance ionic thermoelectric materials will play an important role in harvesting low-grade thermal energy.