Thermal energy havesting from low-temperature heat sources, such as human body, is a promissing method to feed energy to vast number of micro-devices and sensors in Internet of Things (IoT) technologies. Thermoelectric generator, used for this purpose, can directly convert heat into electric power through Seebeck effect. In recent years, organic semiconductor materials, such as π-conjugated polymers, have been attracting much attention as a potential candidate for light-weight, flexible, and low-toxic thermoelectric generators. However, thermoelectric performance of polymer-based devices are still low compared to that of inorganic materials, although various novel polymers exhibitting high mobilities have been synthesized so far. The possible reason for this is the absence of both the suitable doping method and the fundamental understanding of thermoelectric properties in these polymer materials.

In general, the power factor (P) of a thermoelectric device is described as a product of the electrical conductivity (σ) and Seebeck coefficient (S) as P=S²σ, whereas these two quantities tend to show opposite behavior upon carrier doping. Thus, it is indispensable to clarify the accurate relationship between σ and S in doped films in order to obtain a guideline to improve P. In semicrystalline polymers such as PBTTT (Fig. 1(a)), an empirical relation of S∝σ^-1/4 (P∝σ^1/2) has been reported in a wide range of σ by using chemical doping, although no physical background for this relation has been clarified [1]. However, different σ-S relation has also been pointed out recently based on the data analyses of various reports including the same data set in ref. 1 [2]. Such difference seems to arise from the different doping conditions in chemical doping. In this presentation, we report our recent challenge to determine the accurate σ-S relation in various high-mobility polymer materials by using an electrolyte-gated thin-film transistor (TFT) structure, which enables a continuous electrochemical doping on the same device up to high carrier densities.

Fig. 1(b) shows a schematic illustration of the TFT device and measurement setup [3]. We adopted an ionic liquid as the electrolyte, which was drop-casted on the polymer thin film fabricated on the glass substrate. A temperature gradient was given by a couple of Peltier devices and the induced voltage (ΔV) and temperature difference (ΔT) between source (S) and drain (D) electrodes were measured to obtain seebeck coefficient (S=ΔV/ΔT) under various gate voltages (V_g). The electrical conductivity was simultaneously measured at each V_g.

Fig. 1(c) shows an example of σ-S and σ-P relations upon electrochemical doping obtained for two PBTTT devices. We observe highly reproducible data for these devices. In the low conductivity region (σ<100 S/cm), the empirical relation of S∝σ^-1/4 is clearly confirmed, whereas the deviation from this relation becomes evident in the high conductivity region (σ>100 S/cm). The maximum of P appears around the boundary of low and high conductivity regions. In the high conductivity region, the σ-S relation seems to approach S∝σ^-1, typical to the Mott’s relation expected for metals. This behavior is consistent with the semiconductor-metal transition of PBTTT occuring at highly-doped regions, which is observed from the temperature dependence of the conductivity as well as the electron spin resonance (ESR) measurements.

Similar σ-S relation was commonly observed in a wide range of polymers such as donor-acceptor (DA) type copolymers. The maximum value of P can be enhanced by the surface treatment of the substrate by self-assembled monolayers or by aligning the polymer main chain, presumably due to the increase of the mobility, although the S∝σ^-1/4 relation is unchanged in the low conductivity regions. The above σ-S relation insensitive to the mobility is indicative of the contribution of domain boundaries in thermoelectric properties.

[1] A. M. Glaudell et al., Adv. Energy Mater. 5, 1401072 (2015).

[2] S. D. Kang et al., Nat. Mater. 16, 252 (2017).

[3] J. Pu et al., Phys. Rev. B. 94, 014312 (2016).