Improved Performance of Organic Thermoelectric Generators Through Interfacial Energetics

Abstract The interfacial energetics are known to play a crucial role in organic diodes, transistors, and sensors. Designing the metal‐organic interface has been a tool to optimize the performance of organic (opto)electronic devices, but this is not reported for organic thermoelectrics. In this work, it is demonstrated that the electrical power of organic thermoelectric generators (OTEGs) is also strongly dependent on the metal‐organic interfacial energetics. Without changing the thermoelectric figure of merit (ZT) of polythiophene‐based conducting polymers, the generated power of an OTEG can vary by three orders of magnitude simply by tuning the work function of the metal contact to reach above 1000 µW cm−2. The effective Seebeck coefficient (S eff) of a metal/polymer/metal single leg OTEG includes an interfacial contribution (V inter/ΔT) in addition to the intrinsic bulk Seebeck coefficient of the polythiophenes, such that S eff = S + V inter/ΔT varies from 22.7 µV K−1 [9.4 µV K−1] with Al to 50.5 µV K−1 [26.3 µV K−1] with Pt for poly(3,4‐ethylenedioxythiophene):p‐toluenesulfonate [poly(3,4‐ethylenedioxythiophene):poly(4‐styrenesulfonate)]. Spectroscopic techniques are used to reveal a redox interfacial reaction affecting locally the doping level of the polymer at the vicinity of the metal‐organic interface and conclude that the energetics at the metal‐polymer interface provides a new strategy to enhance the performance of OTEGs.

S1: Thermopower measurement schematic, measured raw data, state-of-the-art on organic thermoelectric generator power densities, thermal stability and 7-leg generator.
The shadow masks for this measurement were designed following the guidelines provided by van Reenen & Kemerink on the effect of geometry on the thermoelectric performance, so that the measurements are as accurate as possible from a geometrical perspective 1 . All devices had a cross section area of 100 nm by 1 cm. The temperature difference didn't exceed 3 °C and was measured by two T-type thermocouples. We cross-checked the validity of our setup by measuring Ni and Au foils, following the approach used in Petsagkourakis et al. 2 .  We put the sample on top of the hot (red) and cold (blue) peltier elements. The sample is on a glass slide (grey) that has the deposited metal (yellow) and polymer layers (transparent blue). The dimensions and positions of the gold electrodes follow the geometrical aspects of Reenen & Kemerink 1 for a trustworthy thermoelectric measurement, where the aspect ratios of the electrodes and the channel should be less than 1. The open circuit voltage was measured with a Keithley nanovoltmeter and the temperature was measured with thermocouples (black) that are attached on the sample with thermally conductive paste (black). We identified that the thermocouple readings were unaffected if the thermocouples were put in a slightly different place (as long as it followed the instructions provide by van Reenen & Kemerink 1 . In order to cross-check the validity of the setup, Au and Ni foils were measured, such as in Petsagkourakis et al. 2 Moreover, our reported values for PEDOT:PSS/Au, PEDOT:PSS/Ag and PEDOT:Tos/Au agree with other reported values in the literature [2][3][4][5] . To further address the trustworthiness of our devices, we compared two different device geometries (L C is the distance of the electrodes, L e and W e are the length and width of the electrodes): (i) L C = 10 mm, L e = 0.1 mm , W e = 2 mm , while for configuration (ii) L C = 20 mm, L e = 0.2 mm , W e = 4 mm . (b) Comparison of the measured Seebeck for the different setup architectures (our configuration, (i), and wide electrode, (ii)).    # in Figure 1 Referen ce  Fig. 1.
Among the thermoelectric single leg generators, there are two kinds: either lateral or vertical architecture (see Fig. S1.8). For each of those types, we extract the power density by using the power generated and calculating the correct cross-section area (see formula below). We report the cross-section area in the Table S1.1. Our single leg elements have a thickness of 100nm and width of 2mm , thus a cross-section area of 2x10 -6 cm 2 . Then we have also projected the power density for a temperature gradient of 30 o C in the last column of the table to be able to compare them all together.
Figure S1.8 : The various architectures for a thermoelectric generator reported in the literature. A lateral single leg, a lateral module, a vertical single leg and a vertical module.The cross-section area is defined as width times thickness for the lateral architectures and as long side times short side for the vertical architectures. Meanwhile, the fill factor of a multi-leg module is defined as the thermoelectric active cross-section area of the whole device.
Among the thermoelectric modules, there are two kinds: either lateral or vertical architecture. Again, we can estimate the cross-section area as well as the fill factor. The fill factor is the part of the area cross-section of the device that is active. For the modules, we extrapolate the value for a fill factor of 100%, which would mean the maximum use of the area. In that way, the power density (µW/cm 2 ) of a module extrapolated to a fill factor of 100% is a fair comparison to the power density of single leg generators which have by default a fill factor of 100% (FF=1). Hence, the last column of the

S2: UPS spectra and Kelvin probe measurements
The UPS spectra were acquired with an UHV surface analysis system, consisting of an entry chamber (base pressure ~1×10 -7 mbar), a preparation chamber (~1×10 -8 mbar), and an analysis chamber (~2×10 -10 mbar). Monochromatized HeI (h = 21.22 eV) spectra were recorded with a in-lab designed and built photoelectron spectrometer, and were calibrated by using the Fermi edge and Au 4f 7/2 peak position of an Ar + ion sputter-cleaned gold foil. The UPS was performed with an error of ±0.05 eV.   PEDOT:To s

S4: IRAS spectra and analysis
For the IRAS samples, the PEDOT films were transferred on top of glass slides covered with the various metals (~100 nm thick) with dimensions of 25x10 mm 2 . The IRAS spectra were acquired using a Bruker Equinox 55 FTIR spectrophotometer.  Table S1.  Table S1.

S5 : Oxidation level estimate in the interfacial region and effect on contact resistance
We can trace back the oxidation level of PEDOT via the FTIR spectra recorded in Fig. 3.
Indeed, the vibrational frequencies of the C=C asymmetric and C=C symmetric peaks are coupled to the oxidation level as pointed out in the paper by Khan  Hence, we were able to correlate our FTIR data with optical absorption. Combining these information together with the evolution of the normalized absorbance vs. oxidation levels for PEDOT:Tos provided by Bubnova et al. 6 , we can estimate of the oxidation levels in the interfacial region probed by FTIR at the metal-PEDOT interface. We provide the details of this correlation in the Table S5.
We estimate that the PEDOT:Tos/Al has an oxidation level of ~17%. Moreover, our wavenumber shifts (Fig.3) for PEDOT:Tos/Al are similar to PEDOT:Tos that has been exposed to a pH ~ 10 solution, thus having a conductivity of ~250 S/cm (according to Khan et al). 19 For our films, this would correspond to a resistance about ~ 400 Ω. The OTEG single leg resistance metal/PEDOT/metal is constituted by the sum of contact resistances (R con1; R con2 ) and the channel PEDOT resistance, R dev = R con1 +R channel +R con2 . The measured total resistance for Al/PEDOT:Tos/Al is R dev ~ 900 Ω. If we consider that R con1 =R con2 = 400 Ω and R channel ~100 Ω , then find back the excepted measured R dev = 900 Ω. This supports our trainof-thoughts about the oxidation level extraction that we conducted. Note however, the presence of very thin oxide layer (10-20 Å) on low work function metals (Ni, Ag, Al) could also contribute slightly to the contact resistance (max few Ohms), but the main resistance is expected to come from the oxidation level of PEDOT in the interfacial chemistry region. The presence of these thin natural oxide layers are detectable by impedance spectroscopy (Section S7, Fig. S.7.3).  The wedge-shaped corrected images are presented herein, with q r and q z being the in-plane and near out-of-plane scattering vectors, respectively, defined as follows: q x =(2π/λ)(cos(2θ f )cos(α f )-cos(α i )), q y =(2π/λ)(sin(2θ f )cos(α f )), q z = (2π/λ)(sin(α f )+sin(α i )), . α f is the exit angle in the vertical direction and 2θ f is the in-plane scattering angle 21 .
Therefore, the scattering vector is calculated as follows, . The 1D scattering patterns were obtained by performing a radial integration of the 2D images with respect to the incident beam.   Table S7.1-3) to acquire a full image on the stability of the four metals. The metals were used as the working electrode in a three electrode configuration (Ag/AgCl reference and Pt mesh counter electrodes) and they were characterized with a Gamry Potentiostat. In Fig. S7.1 and Fig. S7.2 are presented the results of the spectra and in Table S7.1-3 are the equivalent circuits and the fitted parameters. As it can be observed in Fig. S7.1 and Fig. S7.2 all four metals are relatively stable in acidic and neutral conditions, while in basic conditions Al and Ag are starting to corrode in lower frequencies, which is evident from the respective increase of the phase (towards 0 o ) in those frequencies. For all metals, the EDLC was in the order of magnitude of 10 -5 …10 -4 F/cm 2 . EDLC is a means to characterize and understand the formation of interfacial layers between the metal and the media (i.e. the ions with water in this case). In order to acquire a more deep understanding on how this interfacial property is linked to the pH conditions and the metal work function, we plotted EDLC vs the pH and vs the metal work function (Fig. S7.3). Although for acidic or basic pH, there doesn't appear any correlation between the parameters, for neutral pH the EDLC is increasing with the metal work function; this is an indication that an oxide layer is formed between the metal and the electrolyte that decreases the capacitance of the system. Upon removal of that layer , i.e. in basic pH, EDLC is much higher.

S8 : Inductively Coupled Plasma with Optical Emission Spectroscopy
The metal samples (1x1 cm 2 ) were immersed in the different pH solutions (~5ml) for 30mins, to emulate the EIS in Section S7. Then those solution were dissolved to 10ml with additional DIW and a total of 16 samples were delivered to Gränges R&I for analysis with ICP-OES.
The samples were in plastic tubes, approximately in volume of 10 ml of each. The analysis was done without any sample preparation, except addition of 0.5 ml of nitric acid on each.
The samples and internal Wilab numbers are found in Table S8     pristine is un exposed sample. while pristine is un exposed sample. while pristine is un exposed sample.

S12: Materials and Fabrication
3,4-ethylenedioxythiophene (EDOT), pyridine, dimethylsulfoxide (DMSO) , n-butanol, ethanol, acetone, isopropanol, diethyle ether, chlorobenzene, p-toluenesulfonic acid, sulfuric acid, and chloroplatinic acid were purchased from Sigma-Aldrich. Clevios CB54 and Clevios PH1000 were purchased from Heraus. Al, Ni, Cr, Ag, and Au were deposited with thermal evaporation on glass slides cleaned in an ultrasonic bath with acetone (5 min) and isopropanol (5 min). A ~5 nm layer of Cr was deposited prior to the deposition of Ag and Au for better adhesion to the glass slides. For all metals the deposition rates were 1 Å per second. Pt was electrodeposited on Au electrodes following the work of Strakosas et al. 22 . A solution containing 5 mM chloroplatinic acid and 50 mM sulfuric acid was used for the Pt electrodeposition in a three electrode planar configuration. Ag/AgCl and a platinum mesh were used as the reference and counter electrodes, respectively. A probe with a positioner was used to contact the Au electrode that was used as a working electrode for the electrodeposition.
A potentiostat was used to apply -0.2V to the working electrode vs. the reference electrode for 5 minutes. The carbon paste DuPont 7102 was blade-coated on the samples. P(g 4 2T-T) was synthesized with a procedure reported elsewhere 23 . 10mg/ml solutions of the copolymer in chlorobenzene were spincoated on the substrates (2000 rpm, 45 s, 1000 rpm/sec) and annealed at 120 o C for 30 mins. A solution of p-toluenesulfonic acid in diethyl ether (0.1M) was consequently spin coated on the samples to dope them. PEDOT:PSS dispersions were formulated by mixing Clevios PH1000 with 5 v-% DMSO 24 , followed by spin coating (1500 rpm, 30 s, 800 rpm/sec) and annealing at 100 °C to remove the solvents. A part of the metal contact was then cleaned from PEDOT with a laboratory scrub that was dipped in distilled water. The samples were further annealed at 100 °C for 5 min. PEDOT:Tos was polymerized with in-situ wet chemical oxidative polymerization with a procedure that is reported elsewhere 3 . The Clevios CB54 was diluted at 40 wt-% in iron tosylate with n-butanol and then pyridine and DMSO were both added at 3 v-% concentration. The oxidant solution with the additives was left stirring for 12 hours and then stored in a fridge overnight before use. EDOT was added in a ratio of 3.3 μl per 100 μl of oxidant/additive solution. The EDOT/oxidant/additive was stirred vigorously for 30 s followed by spincoating on the desired substrate (1500 rpm, 30 s, 800 rpm/s). Afterwards the films/substrate were annealed at 100 °C for 15 min to initiate the polymerization. The films were washed for 5 min each in a bathing sequence of n-butanol, n-butanol, ethanol in order to remove excess oxidant. Finally, the samples were dried with nitrogen. As this polymerization is aggressive towards metals such as Ag, the PEDOT:Tos samples were initially polymerized on top of a silicon wafer for all systems. During the washing treatment with n-butanol, the PEDOT films were delaminated with tweezers and transferred to the substrates with the metal contacts. Those samples were carefully dried with dry air. For the contact resistance measurements, two shadow masks were fabricated to extract the contact resistance with a 4-point probe approach and with the transmission line method. PEDOT:PSS was later deposited on those films similarly to the approach described earlier.

S13. Stability of metals at various pH
Now the stability of the metal oxide depends on the electrical potential and the pH of the environment as described by Pourbaix's diagram. We investigated the stability of our metal surface in electrolyte solutions of pH=1.8, 7.4 and 11 by Electrochemical Impedance Spectroscopy (EIS, see Fig S7.1-3) and Inductively Coupled Plasma with Optical Emission Spectroscopy (ICP-OES, see Fig. S8.1-4). The electrodes were dipped in the three different solutions and EIS was performed. All metal surfaces display a capacitive behaviour at low  Fig. S8.1, Fig. S8.4). In the Fig S7.1, only Ag in the three different pH displays some deviation with a slight resistive behavior at low frequencies, while Al clearly undergoes an electro-induced dissolution only at basic pH (phase angle is close to zero deg). The capacitance for Ag is highest for the acidic and basic pH; while for Al, it is highest for the basic pH. The high capacitance is attributed to the dissolution of the oxide layer, which is also confirmed by the presence of Al and Ag ions in those solutions after dipping the electrodes for ~30 mins as measured by ICP-OES analysis (Fig. S8.2-3). The observations are also in good agreement with the microstructure of the metal surface observed by Scanning Electron Microscopy (SEM, see Section S9) after the 30 min exposure in those electrolytes. It is only the microstructure of the Ag surface that is modified in basic and acidic medium, together with that of the Al surface in alkali medium.