Ultra-High Performance Organic Thermoelectric Generators through Interfacial Doping Gradients

The interfacial energetics are known to play a crucial role in organic electronic devices. However, their effects in organic thermoelectrics remain to be elucidated. In this work we optimize the output power density of an organic thermoelectric generator (OTEG) at ambient atmosphere to record high values, by varying the work function of the metal contacts. We nd that the effect is linked to extended gradients of doping states, which are induced by humidity and reside inside the organic layer oriented perpendicular to the metal contacts. The thermovoltage, arising from this contact phenomenon alone, reaches a magnitude similar to that of the Seebeck voltage of the conducting polymer itself, thereby providing a major contribution to the resulting thermoelectric performance. With this work, we put a new emphasis on the importance of the metal-polymer interface in thermoelectrics. The overall output performance can be greatly improved by ne-tuning the interfacial energetics, which then provides an attractive complementing route for enhancing the performance of OTEGs.


Introduction
Thermoelectric devices convert heat ux into electricity, and vice versa. Over the last decade, organic (semi)conductors have attracted major attention from the thermoelectrics community thanks to a series of compelling and unique properties, such as simple processing and manufacturing, mechanical exibility, high abundance with respect to their atomic elements, as well as electric-to-thermal conductivity ratios (σ/κ) similar to those of conventional inorganic alloys operating at low temperatures (< 200 °C). Various strategies have emerged to increase the Seebeck coe cient (S), power factor (σ S 2 ), and gure of merit ZT (= σ S 2 T/κ) of thermoelectric polymers [1][2][3][4] . Today the best-performing organic thermoelectric material comprises positively charged poly (3,4-ethylenedioxythiophene) chains that are charge-compensated with anionic counterions X (PEDOT:X), yielding ZT values in the range 0.25-0.4 at room temperature 5,6 . Although the device geometry has been found to be crucial for measuring the Seebeck coe cient of the materials accurately 7 , the reported values without these errors still vary from 15 up to 45 μV/K for PEDOT thin lms 1,6,[8][9][10][11][12] . Those differences have a signi cant impact on the thermoelectric power factor and the conversion e ciency, as these quantities are proportional to the square of the Seebeck coe cient.
For an ideal generator, the maximum output power density is found for a load resistance that equals the device (internal) resistance (R dev ) of the organic thermoelectric generator (OTEG), and is expressed by P max = V oc 2 /(4 R dev A) (Eq. 1), where V oc = S ΔT is the open-circuit voltage (thermovoltage) and A is the device cross-section area. In a simplistic model, one leg of a thermoelectric generator is composed of two metal contacts and the thermoelectric material. In inorganic thermoelectric generators, it is well established that the contact resistance between the metal contact and the thermoelectric material limits the e ciency and power output performance 13,14 . Also, the energetics at the metal-organic semiconductor interface is recognized to play a major role in determining the performance of organic (opto)electronic devices, such as organic transistors and solar cells, 15 but a similar dependence has never been identi ed or even investigated for OTEGs.
Here, we report a systematic study on the impact of the metal-organic semiconductor interface energetics on the thermoelectric performance of PEDOT-based OTEGs. We observe that while the Seebeck coe cient of PEDOT lms does not vary with the work function of the metal contact when measured in inert and dry atmosphere, S systematically increases with increasing the metal work function in humid air. Infrared spectroscopy allows us to identify an interfacial redox reaction occurring between PEDOT and the metal contact 16 , which modi es the oxidation level of the polymer in a signi cant volume of the thin lm extending beyond the metal contact interface. Learning from these new insights, we optimize the electrical power output of OTEGs and achieve a record-high value in power density for PEDOT:Toluenesulfonate (PEDOT:Tos)-based OTEGs with Pt as the metal contacts (643.5 μW/cm 2 at room temperature for ΔT= 30°C). We anticipate that our ndings of interfacial energetics-thermopower dependence will have a similar impact on the eld of organic thermoelectrics as it has had in the context of organic eld-effect transistors 17,18 and organic solar cells 19 .

Results And Discussion
PEDOT:Poly(4-styrenesulfonate) (PEDOT:PSS) and PEDOT:Tos lms (~100 nm thick) were deposited on glass substrates pre-patterned with metal contacts (~100 nm thick). While exposed to ambient air (40 % Relative Humidity (RH), 20 °C), a temperature gradient Δ is applied between the two electrodes and the open circuit voltage V oc between those two electrodes is measured to estimate the Seebeck coe cient = V oc ⁄Δ ( Fig. S1.1). The linewidth and inter-distance separation of the metal electrodes were chosen such to minimize the error associated with the estimation of the Seebeck coe cient ( Fig. S1.2). 7 Five different metals were explored, Al, Ni, Ag, Au, and Pt, covering a range of work function values from 3.7 to 5.2 eV, as measured by ultraviolet photoelectron spectroscopy ( Fig. S2 Fig. 1c), we attribute the change in R dev to the variation of the contact resistance R c with the contact metal WF (Fig. 2c). In order to provide further evidence on that, we extracted R c values also for the PEDOT:PSS/Ag contact (6.7 ohm cm) and PEDOT:PSS/Au contact (2.4 ohm cm) and indeed observed a similar dependence for the total device characteristics. Hence, the WF of the metal contact in an OTEG impacts the resulting Seebeck coe cient, the device resistance, and thus its electrical power conversion performance. These ndings raise two questions: what is the origin of the interfacial effects and what is the true Seebeck coe cient of PEDOT?
Before designing experiments to answer these two fundamental questions, it is important to remind that the humidity affects the thermoelectric properties of PEDOT:PSS. The reason to this is that the thermodiffusion of ions generates an additional contribution to the Seebeck voltage that has timedependent and diminishing characteristics. 10,24 In our measurements, carried out at 40 %RH, the Seebeck coe cient is constant over an extended period of time ( Fig. S3.1), thus indicating the absence of an ionic Seebeck effect. Moreover, there is no additional static voltage as indicated by the fact that the intercept of the thermovoltage vs ΔT curve runs through zero (see Fig. S2.2). The voltage contribution from the ionic Seebeck effect in PEDOT:PSS is also known to be negligible at relative humidity levels below 40 % 25 , and has never been observed in PEDOT:Tos even at elevated humidity levels 9 . With this in mind, we then decided to focus on PEDOT:Tos exposed to 40 (Fig. 2a) 4 . We believe that this value is the true Seebeck coe cient of the PEDOT:Tos lms (i.e., without any contribution introduced by the metal contact). When the samples were exposed to ambient atmosphere (40 %RH at room temperature), the Seebeck coe cient increased from 21 μV/K to 50 μV/K for OTEGs with Pt contacts (Fig.   2b). Interestingly, when the samples were re-introduced into the glovebox, the Seebeck coe cient returned back to its original value of 21 μV/K. Next, we identi ed whether the relative humidity and/or the presence of oxygen were key-parameters to control the Seebeck coe cient. We conducted the thermoelectric measurements in three different atmospheres: dry nitrogen, dry oxygen-nitrogen mixture and humid nitrogen. After transferring the samples to a glovebox lled with dry air (O 2 +N 2 , -30 °C dew point), we observed a slight increase in the Seebeck coe cient value reaching 25±3 µV/K (Fig. 2c). One set of samples was then kept in dry nitrogen atmosphere while it was characterized over an extended period of time. These OTEGs displayed a constant Seebeck coe cient over the entire measurement period. Another set of samples was instead exposed to humid nitrogen and underwent an immediate and rapid increase of the Seebeck coe cient, starting at 25 μV/K and nally reaching 43 μV/K after 20 h of exposure. After 20 h, the exposure to humidity was terminated and dry nitrogen was fed into the chamber. We then observed that the Seebeck coe cient slowly reduced towards its original value of 21 μV/K. The power output recorded from this Au/PEDOT:Tos/Au OTEG leg also displays an evolution that strongly depends on the relative humidity level (Fig. S2.3). Consequently, we can then conclude that humidity is a keyparameter to observe the effect of WF on the measured Seebeck coe cient for PEDOT lms.
The way S and R dev depend on the WF of the metal contacts pinpoints that the phenomenon responsible under vacuum conditions, is typically localized over a few nm. 15 However, when a PEDOT lm is exposed to ambient air with 40 %RH, ions from the polymer lm can potentially diffuse within the polymer bulk as the interchain interactions are dominated by electrostatic and Van der Waals forces. In other words, PEDOT lms are mixed ion-electron conductors with an ionic conductivity that increases exponentially with the relative humidity of the environment 34 . The presence of mobile ions accompanied by the interface charge transfer, driven by electronic chemical potential equilibration between PEDOT and the metal, is expected to form an inhomogeneous oxidation level pro le in the PEDOT bulk starting from the interface. This spontaneous reorganization of the electrons and ions due to the Fermi level alignment at the PEDOT-metal interface has not been identi ed to date. However, it is known that under an external applied bias, generated by two electrodes connecting a PEDOT layer that is also coated with an electrolyte, a potential gradient is created along the PEDOT-electrolyte interface. This then generates an oxidation pro le that results in an electrochromic gradient, a phenomenon that can be observed for instance along the channel of an electrochemical transistor 35 .
Vibrational spectroscopy is known to be sensitive to the oxidation level of PEDOT [36][37][38] and can thus be used to probe the metal-PEDOT interface. To identify the intrinsic vibrational transitions of the polymer, we rst measured ATR-FTIR (attenuated total re ectance Fourier transform infrared spectroscopy) of PEDOT:PSS and PEDOT:Tos lms on a non-conducting IR-transparent CaF 2 substrate (Fig. 3b,c 15 ; while if the polymer WF is larger than the metal WF (e.g. aluminum), the polymer gets reduced at the interface compared to the pristine oxidation state of PEDOT. Since the asymmetric peak position of the Au sample is the same as with the pristine PEDOT:PSS, there is no apparent interfacial redox reaction. However, there is an interfacial redox reaction between PEDOT:Tos and Au, as the asymmetric peak of the pristine PEDOT:Tos is lower than that of PEDOT:Tos/Au, which is likely due to the lower WF of PEDOT:Tos compared to PEDOT:PSS.
In IRAS, the incident infrared light is re ected at grazing incidence by the metal surface so that the absorbed wavelengths correspond to transitions in the vibrational modes of PEDOT at the metal interface, as well as in the bulk of the thin lm. Tuning the thickness of the PEDOT lm provides us with a probe to study the volumetric extension of this redox phenomenon. When there is no interfacial redox reaction, as identi ed for the PEDOT:PSS/Au system, the peak positions remain constant while increasing the polymer thickness from 100 to 800 nm, although we do observe a slight trend in the peak ratios of the asymmetric C=C doublet. In contrast, the C=C asym peak of PEDOT:PSS on Al [Ag] varies with the thickness from 1525 cm -1 [1531 cm -1 ] for a 100 nm thick lm to 1531 cm -1 [1533 cm -1 ] for a 800 nm thick lm (Fig 3e-h, Fig. S4.5). This supports the formation of an oxidation level pro le within the PEDOT lm that extends from the metal-interface into the polymer bulk, reaching up to several hundreds of nanometers of thickness.
Electrochemical doping/dedoping of PEDOT thin lms is known to alter its crystalline structure 43 . We studied changes in morphology, induced by the oxidation level pro le triggered by the redox interfacial chemistry, by using grazing incidence wide angle X-ray scattering (GIWAXS). As the metals themselves have a strong scattering contribution, we deposited a very thin layer (10 nm) of Au and Al on Si (with native oxide) followed by the deposition of the PEDOT lms on top. 2D GIWAXS scattering patterns (Fig.  S5.1) were recorded and a background (Si+metal) subtraction was carried out. This allowed us to decouple the pure PEDOT contribution (Fig. 4 inset), which was also integrated to obtain the 1D scattering patterns. The scattering features observed for both PEDOT:Tos and PEDOT:PSS agree well with those previously reported in literature 43,44 . Interestingly, a higher degree of order (i.e. a higher integrated (100) peak intensity) is observed for PEDOT in contact with Au as compared to Al, which is consistent with the higher oxidation level of PEDOT:Tos on Au 43 .
The interfacial redox chemistry found at the interface between the metal contact and the PEDOT lm in OTEG legs is promoted by the WF differences of the metal and the polymer, along with the kinetics of ions that are boosted by the presence of humidity. In order to better understand the full impact and opportunity of the WFs in OTEGs, we propose that the interfacial energetics should be taken into the account while characterizing and optimizing OTEGs in the future. We believe that the description of the electric potential pro le along the metal-polymer-metal OTEG leg, submitted to a temperature gradient, should include the drop in Seebeck potential within the PEDOT (S m =dV/dT) lm, as well as the two dissimilar interfacial potential drops located at the hot (V int (T H )) and cold sides (V int (T c )). With such a description ( Fig. S6.1 Finally, we highlight the importance of the interfacial energetics on the OTEG performance, by comparing the remarkably high value of P max obtained with Pt contacts with values found in the literature (Fig.5).
Since our devices are based on a single element, as a proof of concept for the phenomenon, we expect that even higher power output could be achieved through proper device fabrication (i.e., by optimizing the number and dimensions of the thermoelectric elements). It is evident from Fig.5 that interfacial energetics can severely enhance the output power of OTEGs, we manage to signi cantly enhance the thermoelectric performance and reach a record-high OTEG output power. Our study thus opens up a new pathway towards the optimization of organic thermoelectric technology.
Moreover, our discovery of a doping gradient within the polymer bulk, extending away from the various metal contacts, is of direct relevance for the eld of organic electronics in general, as metal-conducting polymer interface is a fundamental elementin many devices. A better understanding of these polymermetal interfaces will impact the performance and e ciencies of several devices, from solar cells to supercapacitors and batteries, to organic electrochemical transistors and more.

Methods
Details on all processes for material and device fabrication are provided in the Supplementary  Information (Section S9). For all metals the deposition rates were 1 Å per second. Pt was electrodeposited on Au electrodes following the work of Strakosas et al 46