Boosting the Thermoelectric Properties of PEDOT:PSS via Low‐Impact Deposition of Tin Oxide Nanoparticles

Poly(3,4‐ethylenedioxy thiophene):poly(styrenesulfonate) (PEDOT:PSS) exhibits valuable characteristics concerning stability, green‐processing, flexibility, high electrical conductivity, and ease of property modulation, qualifying it as one of the most promising p‐type organic conductors for thermoelectric (TE) applications. While blending with inorganic counterparts is considered a good strategy to further improve polymeric TE properties, only a few attempts succeed so far due to inhomogeneous embedding and the non‐ideal organic‐inorganic contact. Here a new strategy to include nanoparticles (NPs) without any ligand termination inside PEDOT:PSS thin films is proposed. Spark discharge‐generated tin oxide NPs (SnOx‐NPs) are “gently” and homogenously deposited through low‐energy diffusion mode. Strong interaction between naked SnOx‐NPs and PSS chains occurs in the topmost layer, causing a structural reorganization towards an improved PEDOT chains crystalline packing at the bottom, providing a positive contribution to the electrical conductivity. Meanwhile, dedoping and energy filtering effect introduced by the SnOx‐NPs cause dramatic Seebeck coefficient enhancement. The optimized power factor of 116 μWm−1 K−2 achieved is more than six times higher than the value found for the film without NPs. This easy and efficient strategy promises well for future mass production of flexible TE devices and the mechanism revealed may inspire future research on TEs and flexible electronics.


Introduction
Polymeric thermoelectric (TE) materials have recently attracted increasing interest from the scientific community. [1,2] Their flexibility and low weight make them highly suitable for a broad range of applications such as portable electronic devices powered by human body temperature. [3] However, polymeric TE

Results and Discussion
The morphology of the NPs produced by spark discharge was studied at first. To this aim, we deposited them on Cu grids and analyzed them by transmission electron microscopy (TEM), as shown in Figure 1c. As observed, the NPs were widely and homogeneously dispersed on the Cu grid, which is very important for the preparation of homogeneous hybrid films with PEDOT:PSS. Their shape is spherical with sizes ranging between 10 and 20 nm. This value is further confirmed by scanning electron microscopy (SEM) and grazing incidence small-angle X-ray scattering (GISAXS) as shown in Figure S1, Supporting Information. In particular, an average diameter of 10.3 nm was derived from fitting the GISAXS data (see Figure  S1b, Supporting Information). Before analyzing the hybrid film, a thicker layer of NPs was deposited to perform grazing incidence wide-angle X-ray scattering (GIWAXS) as shown in Figure S2, Supporting Information. The GIWAXS pattern exhibits a sharp peak located at q = 1.85 Å −1 suggesting a d-spacing of 3.4 Å, which can be attributed to the (110) plane of SnO 2 . [18] Oxidation of the NPs is indeed expected in our case due to the exposure to air. The exact composition of the NPs will be further discussed below based on X-ray photoelectron spectroscopy (XPS) results. We will thus call the NPs as SnO x -NPs in the following.
Next, the SnO x -NPs were deposited on PEDOT:PSS thin films and the surface morphology of the hybrid was investigated by atomic force microscopy (AFM). Figure 1d shows phase mode AFM images of the PEDOT:PSS:DMSO thin film before and after deposition of SnO x -NPs. Comparing the two images, it is obvious that the SnO x -NPs (the brighter objects in the phase images) were successfully deposited onto the PEDOT:PSS:DMSO thin films. In line with what was observed by TEM for the deposition on Cu grids, SnO x -NPs appear well dispersed onto the polymer film, with minimal aggregation (Figure 1d). After NP deposition, the roughness as calculated from AFM increased only slightly, from 1.8 nm for the pristine PEDOT:PSS:DMSO film to 2.0 nm for the PEDOT:PSS:DMSO:SnO x -NPs one. This observation indicates that during the deposition process the SnO x -NPs penetrate into the polymer matrix, in line with a previous report identifying the reduction of the Gibbs surface free energy as the driving force for embedding NPs inside polymer films. [19] Moreover, a specific interaction between the charged PSSand the SnO x -NPs can act as an additional driving force for the embedding in our case, as will be discussed below based on XPS data.
The TE properties (namely σ, S, and PF) against the NP deposition time are summarized in Figure 2. The effect of the SnO x -NP incorporation was studied both on neat PEDOT:PSS and PEDOT:PSS:DMSO thin films. As shown in Figure 2a,b, the electrical conductivity varies little upon NP deposition. An opposite trend is observed here: while after continuous deposition of SnO x -NPs for 1 h the conductivity of the neat PEDOT:PSS system increased from 3 ± 0.5 to 4 ± 0.5 S cm −1 , for PEDOT:PSS:DMSO thin films a decrease of the conductivity from 600 ± 50 to 575 ± 25 S cm −1 was observed. An additional PEDOT:PSS:SnO x -NPs sample was prepared using 2 h deposition time and it exhibited a slight decrease in conductivity to ≈3.3 S cm −1 , suggesting that prolonging the deposi-tion time well behind 1 h is not necessary. In both cases, the observed conductivity changes are minor, implying that the diffusion mode deposition adopted here is a mild method to add NPs to the polymer thin film. Interestingly, a marked positive effect was observed for the Seebeck coefficient (Figure 2c  organic-inorganic hybrid systems so far. [20,21] Moreover, the preparation method reported here is extremely simple, without any complex chemical synthesis, post-treatment or washing steps, holding great potential in industrialized large-scale production. To shed more light onto the possible mechanism for the observed TE property increase, we first verified the (surface) composition of the deposited SnO x -NPs. As already mentioned, Sn is very sensitive to oxygen and can be easily oxidized in air, forming SnO and SnO 2 at the surface. Oxygen vacancies at the NP surface would render the SnO x -NPs a p-type semiconductor. [17] To quantify the composition of the SnO x -NPs, XPS spectra of the Sn3d core level region were collected. The Sn3d 5/2 component is shown in Figure 3d. The slightly asymmetric structure of the line suggests the presence of Sn with different oxidation states or in different chemical environment. The fitting suggests that there is predominantly Sn 4+ from SnO 2 , giving rise to the component at a BE of 487.0 eV (black dashed line in Figure 3d). This is consistent with reports for SnO x films of thickness below 60 nm, where the same high oxidation state was observed. [17] The peak at higher binding energy (BE) (red dashed line in Figure 3d), is attributed to Sn 4+ coordinated by oxygen from PSS − . [22,23] This is clear evidence for the interaction between SnO x -NPs and PSS.
We then employed GIWAXS to learn more about the poly mer blend structure. GIWAXS is particularly suited to probe the crystal structure of thin films on molecular length scales and has been proven highly successful in characterizing PEDOT:PSS. [24] Figure 3a  hybrid film, respectively; the out-of-plane linecut profiles are presented in Figure 3c. The PEDOT:PSS thin film exhibits the typical structure with an intense low angle peak located at q z = 0.26 Å −1 together with a weaker peak located at 0.51 Å −1 , both strongly focused along the out-of-plane q z direction. It is generally accepted that in PEDOT:PSS the second peak at q z = 0.51 Å −1 is the second order 200 peak. [25] At higher scattering angles, two more peaks are visible: the free PSS peak located at q = 1.26 Å −1 and the π-π stacking PEDOT peak located at q = 1.78 Å −1 . Upon NPs deposition, a clear difference in the ratio of the intensities of the 100 and 200 diffraction peaks of PEDOT:PSS is observed; for PEDOT:PSS:SnO x -NPs, the 200 peak shows a relative intensity increase of 80%. Similar effect is found for the PEDOT:PSS:DMSO:SnO x -NPs films (see Figure S3, Supporting Information). This observation suggests a modification in the crystalline structure of the PEDOT crystals. As reported by Bießmann et al. [26] and also as recently observed by us, [24] the first order of diffraction of the so-called type-II packing mode of PEDOT (100 type-II ) overlaps in position with the 200 peak for the type-I packing (200 type-I ). Due to the relatively low doping efficiency of type-II, this process can be considered as a dedoping effect and can partially explain the increase in the Seebeck coefficient reported in Figure 2c,2b. [26] The driving force for the dedoping could be the strong interaction between SnO x -NPs and the PSS − , which is normally combined with positively charged PEDOT + by Coulombic interactions. [27,28] The appearance of type-II PEDOT packing is paired by a shift in the peak position of the free PSS peak from 1.26 to 1.30 Å −1 after NPs incorporation, suggesting a concurrent change of PSS packing.
XPS spectra confirm the changes in the polymer blend observed by GIWAXS. In Figure 3e, the S2p core-level regions of the PEDOT:PSS:DMSO thin films without and with the SnO x -NPs show clear differences. As commonly reported, [29] the S2p signal of the PSS chains, is peaked at higher BE (≈168.4 eV; Peak 1), than the one of PEDOT (≈164.4 eV; Peak 2). Upon SnO x -NPs incorporation, the ratio between the intensities of the PEDOT and the PSS S2p signals decreases from 1:2.1 to 1:3.2. This suggests that the amount of PSS in the topmost part of the thin film has increased upon NP deposition-taking into account the surface sensitivity of XPS (about 10 nm information depth for PEDOT:PSS for S2p and Al K α excitation as estimated from ref. [30,31]). This surface accumulation of PSS is the result of the interaction between PSS and the SnO x -NPs, in line with the observation from GIWAXS. Fitting of the S2p core level spectra (shown in Figure 3d), discussed in Table S1, Supporting Information, reveals a third component (≈166.1 eV; Peak 3) that can be attributed to S from highly doped PEDOT unit (labeled as PEDOT-S + ). [32,33] This indicates that the doping level of PEDOT in the hybrid system is slightly lower (2 ± 1% PEDOT-S + ) compared to the doping level in PEDOT:PSS (8 ± 1% PEDOT-S + ) and PEDOT:PSS:DMSO (4 ± 1% PEDOT-S + ). This is a confirmation of the dedoping process induced by the NPs and discussed above on the basis of the GIWAXS results.
To further verify the PSS migration towards the film/air interface in the thin films containing the NPs, GIWAXS patterns were acquired using two different incident angles (α i ), that is, two different penetration depths (see Figure S5, Supporting Information). The ratio between the free PSS and the PEDOT π-π stacking peaks in the GISAXS linecuts along the q z direction is higher when a α i = 0.13° is used (minimal penetration inside the film) with respect to α i = 0.23° (full film penetration), confirming the enrichment of PSS at the film/ air interface as concluded from the XPS data. X-ray reflectivity (XRR) is then used to probe the vertical structure of the hybrid system as shown in Figure S6, Supporting Information. It can be clearly observed that there is a layer with higher electron density on the top region, the composition of which should be an enriched PSS phase (as verified by XPS) containing the SnO x NPs. Compared with PEDOT:PSS:DMSO film, the hybrid system also shows a different electron density distribution at the film/air interface region, which indicates a restructuring of the film surface (even smoother) in line with our AFM observation. [34] Besides, the lower electron density at the bottom part suggests that PSS migration on the top region of the film induces formation of a bottom region with more densely packed PEDOT crystallites, resulting in a more coherent carrier charge transport path. This positive effect on the electrical conductivity will compensate for the lowering in the doping level of the PEDOT chains, [35] effectively limiting the drop in σ observed for PEDOT:PSS:DMSO:SnO x -NPs ( Figure 2b) and even enhancing slightly the conductivity in PEDOT:PSS:SnO x -NPs ( Figure 2a).
Another important aspect is that the incorporation of the SnO x -NPs in the PSS enriched top layer implies the presence of phase boundaries acting as scattering centers, which allow only the high-energy charge carriers to go through. [36] In addition, the PEDOT to PSS ratio decrease suggests a longer hopping length among the doped PEDOT crystallite, which can be considered as the increase of scattering center amount. According to the energy filtering theory, charge carriers with low energy contribute negatively to the Seebeck coefficient, and if the amount of low energy charge carriers is reduced, the Seebeck coefficient can be improved very effectively. [36][37][38] This constitutes an additional important cause for performance improvements shown in Figure 2a,b. Our observations are in line with different works from the Ouyang group on energy filtering strategies applied to PEDOT:PSS based films, including blending with polyelectrolyte, ionic liquids, and even n-type MXene. [39][40][41] Schemes of the morphology of the hybrid film and of the conduction contributing to the TE property enhancement are also included in Figure 3f.

Conclusion
In summary, we successfully created homogenous PEDOT:PSS/ SnO x -NPs hybrid thin films. SnO x NPs generated using spark discharge, were deposited with low energy in diffusion mode onto the PEDOT:PSS surface. The NPs compose a p-type semiconductor and interact with PSS − , as shown by GIWAXS and XPS. The NP incorporation induces PSS migration toward the top surface of the films with subsequent structural modifications strongly impacting the TE properties of the thin films. The phase boundaries introduced by NPs and the prolonged hopping length act as scattering centers and thus cause the energy filtering of charge carriers. Together with the observed PEDOT dedoping, energy filtering effect well explains the dramatically improved Seebeck coefficient (43 uV K −1 ) for the PEDOT:PSS:DMSO:SnO x -NPs thin film. Moreover, because of the induced phase separation, more coherent carrier charge transport paths are formed in the PEDOT-rich regions, which compensate for the dedoping effect and counteract the electrical conductivity drop. As a consequence of this structural reorganization mechanism, one of the highest PF (116 µW m −1 K −2 ) for hybrid PEDOT:PSS-based thin films is achieved using this nonexpensive and easy-to-implement method. Our results highlight the importance of the interactions between the polymeric components and the added NPs. The method proposed here avoids any extra steps such as chemical synthesis, post-treatment, or washing, which is very important for possible, future industrial large-scale production. In other words, the easy and not expensive deposition method presented here shows great potential in the production of high-performance flexible p-type TE materials. We are confident that this new strategy will inspire future research towards using simple physical NP production methods to improve properties of polymer for soft electronics and TE applications.
Film preparation: The borosilicate glass substrates (10.0 mm × 10.0 mm × 0.7 mm) were sequentially washed using detergent, acetone, and isopropanol with sonication. Then, the substrates were dried using a nitrogen gun and treated by UV-Ozone for 10 min. The thin films were prepared by spin-coating PH1000 or PH1000-5% DMSO solution on the borosilicate glass at 2000 rpm. These films were subsequently annealed on a hot plate for 10 min at 130 °C and let cool down to room temperature. The film thickness (d) was 60 ± 2 nm as determined by AFM. For NPs fabrication, a spark discharging generator (VSP-G1) from VSPARTICLE B.V (Delft, Netherlands) was used. The generating current and voltage were kept as I = 8.1 mA; V = 1.3 kV, pure tin (provided by VSParticle, Delft) was used as the electrode and an inert atmosphere was achieved with a constant argon flow set at 10 L min −1 . As discussed in the main text, diffusion mode, where the substrates are placed parallel to the gas flow, was used to get homogenously distributed NPs. Before starting the deposition, the system was allowed to run 30 min to let it stabilize and to remove the possible tin oxides on the surface of the electrodes. The deposition process over the substrate was homogeneous, as the whole deposition chamber is filled by a cloud of NP, as verified by COMSOL calculations performed by VSParticle. Thus any in-plane gradient of NPs on the surface of the sample can be excluded and was never noticed by SEM or AFM investigations.
Characterization of the Thin Films: Four-point-probe measurements were performed in an N 2 -controlled environment. The electrical conductivity was calculated with the equation: where L is the channel length (1 mm), w is the channel width (4.5 mm), and d is the thickness of the active layer. [42] The final electrical conductivity was obtained by averaging 4 devices. For the Seebeck coefficient measurements, the setup was kept the same as reported before. [43] The surface morphology was investigated by tapping mode AFM performed on a Bruker AFM multimode MMAFM-2 equipped with an RTESPA-300 probe (resonant frequency 300 kHz, spring constant 40 N m −1 , Burker). The height images and phase images were captured at a scan rate of 0.8 Hz and 640 points per line.
The data were analyzed with the Nanoscope Analysis 1.5 program (provided by Bruker). To determine the thickness, small scratches were made in the film using a very fine needle. The scan direction was set perpendicular to the scratch direction, to allow for the determination of the height of the scratch (and therefore the film thickness). SEM images were recorded in vacuum on an FEI NovaNano SEM 650 with an acceleration voltage of 5 kV. GIWAXS measurements were performed using a MINA X-ray scattering instrument built on a Cu rotating anode source (λ = 1.5413 Å). [44] 2D patterns were collected using a Vantec500 detector (1024 × 1024 pixel array with pixel size 136 × 136 microns) located 102 mm away from the sample. The beam center was estimated using the known position of diffracted rings from standard Silver Behenate and α-Al 2 O 3 powders. The scattering vector q was defined with respect to the center of the incident beam and has a magnitude of q = (4π/λ)sin(θ), where 2θ is the scattering angle and λ is the wavelength of the X-ray beam. Herein the authors opted to present the wedge-shaped corrected images, where q xy and q z are the in-plane and near out-of-plane scattering vectors, respectively. The scattering vectors are defined as follows: s in where α f is the exit angle in the vertical direction and 2θ f is the in-plane scattering angle, in agreement with standard GIWAXS notation. An incident angle α i = 0.23° was used for all the samples except for the probe of different penetration length discussed in Figure S6, Supporting Information, where a small incident angle α i = 0.13° was applied. GISAXS measurement was performed with a Vantec2000 detector (pixel size 68 × 68 microns) and sample-to-detector distance equaled to 3000 mm. Fit of the GISAXS horizontal intensity cut has been performed using the equation for a polydisperse ensemble of spherical objects. In order to successfully fit the curve and take into account for aggregation and size dispersion properly, two different spherical populations have been considered. A single population did not allow to fit the data properly. Log-norm distributions have been assumed. The scattered intensity, in this case, is given by the sum of the intensities scattered by the two populations of NPs according to the equations where N 1 (R) and N 2 (R) are the log-norm distribution functions for the first and second populations of particles, characterized by the location and width parameters μ and σ: (4) and P(q, R) is the well-known form factor for a spherical object:  (5) Δρ is the contrast term, in this case, given by the difference of the electron density of the SnO x NPs and the surrounding media (air). The constant A, contains several terms including instrumental factors as well as the Fresnel transmission and reflection coefficients, that are constant at a fix q z . [45] The fit was achieved using MATLAB. In the data, evidence for significant spatial correlation between the NPs was not seen. This is plausible for the low concentration of the particles and the disordered aggregated that are formed (as visible in TEM). So, any structure factor here was not assumed.
XRR measurements were performed using a PANalytikal X'Pert thinfilm diffractometer (lab source with λ = 1.541 Å) in the 2θ range from 0.15° to 4.00° and a step size of 0.01°. Analysis of the XRR curves was achieved using the GenX software. [46] A model composed of bulk Si, SiO2 layer, PEDOT:PSS bottom interfacial layer, PEDOT:PSS mid bulk layer and PEDOT:PSS (or PEDOT:PSS:SnOx-NPs) top interface layer was successfully used to fit the experimental data. A model composed of only two polymeric layers was not able to describe the XRR curves.
For TEM measurements, the NPs were directly deposited onto the Cu grid. The morphologies were observed under a Philips CM120 Microscope coupled to a 4k CCD camera using an acceleration voltage of 120 kV.
XPS data were collected with a Surface Science SSX-100 ESCA instrument with a monochromatic Al K α X-ray source (h ν = 1486.6 eV). The measurement was done at a pressure below 5 × 10 −9 mbar. The spot size was 1000 µm. For each sample, at least two different spots were measured; fitting results are averaged over these spots. The energy resolution was set to 1.26 eV and the electron take-off angle with respect to the surface normal was 37°. The spectra were analyzed using the least-squares curve-fitting program Winspec, developed at the LISE laboratory, University of Namur, Belgium. A Shirley background was used. BEs are reported with a precision of ±0.1 eV and referenced to the C1s peak at 284.6 eV. [47,48] The S 2p core-level spectra shown in Figure  S4, Supporting Information, were fitted with three Voigt doublets, each consisting of two Voigt lines with the same width, separated by 1.16 eV and with intensity ratio (0.511) due to spin-orbit splitting.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.