Organic thermoelectric (OTE) materials have garnered increased attention because of their potential to enable flexible and wearable devices that can convert waste heat into electricity. [1] The efficacy of a TE material in this energy conversion is given by the dimensionless figure of merit ZT = (S 2 σ/κ)T, where S is the Seebeck coefficient, σ is the electrical conductivity, κ is the thermal conductivity, and T is the absolute temperature. In comparison to traditional inorganic TE materials, OTEs typically have advantageously low thermal conductivities (0.1 -1 W m -1 K -1 ), but less favorable electronic properties; therefore, research is focused on optimizing the power factor (PF), S 2 σ. The PF of organic semiconductors can be fine-tuned using a variety of synthetic techniques (doping, synthetic modification, processing, etc.), [2] but often S and σ are anticorrelated, making PF optimization nontrivial. Decoupling S and σ so that both parameters can be improved simultaneously remains a challenge for the field.
A powerful method to fine-tune the thermoelectrical properties in organic semiconductors is molecular doping, where a dopant molecule creates additional charge carriers by reducing or oxidizing these organic semiconductors. [3] Several studies have reported new conjugated organic backbones and their structure-property relationships, [2c,2f-2j] and doping with a wide variety of molecular dopants. [4] Unlike traditional inorganic semiconductors, where dopant atoms are covalently bonded to the surrounding matrix and are homogenously incorporated on the atomic scale, doped organic semiconductors consist of spatially heterogenous dopants that are coulombically bound to the organic matrix. [5] The addition of molecular dopants impacts the organic matrix morphology, and the interrelationship between molecular dopant, aggregation, and material properties in OTEs is still not fully understood.
For impactful OTE device architectures to be realized, both p-and n-type OTE materials with high PFs are needed. Efficient p-type materials and high PFs have been reported, [2a-2c,6] but progress on n-type materials still lags far behind because of their relatively low charge-carrier mobility values and, often, the lower stability of n-doped materials in various environments. [7] Although numerous n-type OTEs are being explored, including new conjugated polymers [2g,2i] and metal-organic structures, [8] additional investigations on n-type systems are essential for advancing OTEs.
The work described here investigates the effects of dopant selection on the morphological, thermoelectric properties of n-doped FBDPPV (see Figure 1) with dimeric dopant (RuCp*mes) 2 and hydride dopant N-DMBI-H (although this dopant has often been referred to as "N-DMBI", we use "N-DMBI-H" here to emphasize that a hydrogen atom as well as an electron must be lost to form the N-DMBI + cation; Figure 1). To decouple the effects of differing doping mechanism from the counterion size, we also developed a dimer ndopant, (N-DMBI) 2 , that behaves in a similar way to (RuCp*mes) 2 (i.e., which forms the same N-DMBI + cation as N-DMBI-H, but without involving a hydrogen atom or hydride transfer). Using spectroscopic, structural characterization, and thermoelectric property measurements as functions of dopant species and concentration we found that (N-DMBI) 2 dopes FBDPPV most efficiently and produces the optimum thermoelectric properties. Finally, we performed temperature-dependent thermoelectric measurements to elucidate the chargetransport mechanisms. We conclude that charge transport in doped FBDPPV is best described by thermally activated polaron hopping (Mott polaron model), and that the electronic structure is dependent on the dopant. This study shows the importance of dopant selection for optimized n-type OTE materials and provides insight into design guidelines for future n-type OTEs.
2-Substituted-1,3-dimethylbenzimidazole derivatives of several types -hydride derivatives, Y-DMBI-H, such as N-DMBI-H (Figure 1); halide salts of benzoimidazolium cations, Y-DMBI + , that release the corresponding radicals on sublimation; and dimers, (Y-DMBI) 2 , formed by such radicals -and dimeric organometallic compounds, such as (RuCp*mes) 2 , are some of the most efficient n-dopants (so far) in terms of achieving the highest electrical conductivities with both the small molecule C 60 and the polymer P(NDI2OD-T2) (Figure 2, S1, and see Figure 1 for a summary of n-dopant molecules used).
Use of Y-DMBI-H molecules, such as N-DMBI-H, inevitably involves hydride and/or hydrogen-transfer reactions, [9] and the fate of the hydrogen atom in the n-doped systems is in many cases unknown; in contrast, (Y-DMBI) 2 and (RuCp*mes) 2 dimers are known to react effectively and cleanly with electron acceptors to contribute two electrons and form two monomeric cations, Y-DMBI + or RuCp*mes + , respectively. [10] In addition to differing mechanisms, different cation geometries are available through Y-DMBI-H and dimer approaches, which can be important since counterions can impact aggregation and morphology in solid-state, and hence thermoelectric properties. The differences in the shapes of the cations used in this work are evident in both DFT-optimized gas-phase geometries (Figure 2c) and in the single-crystal X-ray structures of RuCp*mes + I -and N-DMBI + PF 6 -(Figure S3 and S4). The dopant cation sizes are fairly similar (molecular volumes based on the X-ray geometries are 2388 and 2176 bohr 3 , respectively); however, the organometallic cation is a bulky cylindrical shape, similar to that of other sandwich compounds, [4d] whereas N-DMBI + has a more planar structure but with a significant twist between the planes formed by the imidazolium and arene portions of the cation (51.9° in the optimized structure (Figure 2c); 52.6° in the crystal structure of its PF 6 -salt (Figure S4)). Although, as noted above, Y-DMBI dimers are known, those reported to date have all employed bulkier 2-substituents (Y = cyclohexyl, ferrocenyl, ruthenocenyl) rather than planar aryl groups. [4f,10b] Given that high conductivity values have been obtained in some systems using N-DMBI-H, and in others using (RuCp*mes) 2 , we reasoned that the hypothetical dimer (N-DMBI) 2 would help us to decouple the role of cation shape from the reaction pathway, and to understand how these variables affect thermoelectric properties.
We found that the new dimer (N-DMBI) 2 (Figure 1) could be synthesized through Na-Hg reduction of N-DMBI + PF 6 -in 85% yield. NMR characterization does not unambiguously confirm the structure of the reduction product, with the 1 H spectrum exhibiting broad features and the 13 C spectrum numerous resonances, likely due to restricted rotation. However, elemental analysis, mass spectrometry (showing, as is typical for dimeric reductants, the corresponding monomer cation), electrochemical data (Figure S5), and chemical reactivity as a reductant are consistent with the proposed dimeric structure. The effective redox potential of the dimer, E(D + /0.5D 2 ), is estimated to be ca. -2 V vs. FeCp 2 +/0 (see details in Supporting Information, Figure S5), similar to that of other (Y-DMBI) 2 species [10b] and of (RuCp*mes) 2 [4d] , but is certainly at least as reducing as -1.45 V, as evidenced by its reduction of TIPS-pentacene to the corresponding radical anion (Figure S6).
In the following sections, we compare N-DMBI-H, (RuCp*mes) 2 (synthesized in 88% yield by a modification of literature procedures that, as described in the supporting information, replaces hazardous liquid alkali metals by silica-gel-supported sodium-potassium alloy), and (N-DMBI) 2 as n-dopants for polymer FBDPPV in terms of their effects on spectra, thermoelectric properties, structure, and morphology.
The doping behavior of the three dopants was investigated using UV-vis-NIR spectroscopy of doped FBDPPV films. Figure 3 shows that pristine FBDPPV has absorption peaks at 490, 710, and 785 nm. Upon doping, an absorbance peak at 1000 nm and a broad mid-infrared (MIR) absorption emerge in all systems with a concomitant decrease in the absorption peaks of the undoped polymer. The increased MIR absorption intensity with increasing doping concentration is consistent with a greater extent of reduction by the dopants.
In comparing doping effects, it is important to note that the dimeric molecules, (RuCp*mes) 2 and (N-DMBI) 2 , are expected to each contribute two electrons and form two monomeric cations, [4c,4d,4f,10] while N-DMBI-H can only contribute a single electron per molecule. [9] Polymers doped with 25 mol% dimer (i.e. potentially corresponding to two electrons for every three polymer repeat units) show a lower absorbance ratio of neutral polymer (i.e., around 710 and 785 nm) to polaron (i.e., over 1000 nm) than polymer doped with 50 mol% N-DMBI-H (potentially a higher doping level of one electron per repeat unit), qualitatively indicating that the dimer dopants have a higher doping efficacy.
The spectral changes with increased doping levels are qualitatively similar for both types of dopants: a feature peaked at ca. 1000 nm (with a subsidiary peak discernable in some cases at ca. 1300 nm), attributed to polaronic absorption, was observed, along with a broad feature extending into the MIR. At the highest levels of reduction (i.e., 50 mol% loadings of the dimeric dopants (RuCp*mes) 2 and (N-DMBI) 2 , corresponding to two electrons per repeat unit), a distinct broad peak is seen around 1700 nm. This may be associated with a compressed polaron, bipolaron, or π-dimer species with similar energetics to the original polaron absorption at 1100 nm. [11] To understand how the dopant-induced electronic states impact thermoelectric properties, the electrical conductivity and the Seebeck coefficient were measured on doped films. As shown in Figure 4a, the highest electrical conductivity for FBDPPV doped with (RuCp*mes) 2 is 1.6 S cm -1 at 23 mol% dimer. However, FBDPPV reaches an even higher electrical conductivity of ca. 8 S cm -1 with N-DMBI-based dopants. The maximum electrical conductivity observed is with 10.7 mol% (N-DMBI) 2 or 43 mol% N-DMBI-H, the data for the latter agreeing well with our previous reports. [2g] To reach highest electrical conductivity, the amount of N-DMBI-H is considerably more than double that of (N-DMBI) 2 , suggesting that the dimer (N-DMBI) 2 dopes more efficiently than N-DMBI-H (even when its ability to contribute two rather than one electrons is taken into account), consistent with the optical data discussed above (Figure 3). The higher electrical conductivities in N-DMBI + systems relative to those in the RuCp*mes + system further affirms the importance of dopant selection for conductivity optimization, and, therefore, perhaps thermoelectric property optimization as discussed below.
As dopant species are introduced into the FBDPPV films, the Seebeck coefficient (S) changes. The Seebeck coefficient is less sensitive to morphology and more dependent on transport parameters and energy levels. [2a,12] Figure 4b shows that the Seebeck coefficients of all films are negative (n-type behavior), as expected; values are -113, -63, and -150 µV K -1 at ca. 12 mol% for the dimers (RuCp*mes) 2 and (N-DMBI) 2 , and at ca. 24 mol% for N-DMBI-H, respectively. For the dimer dopants, this roughly equates to one electron donated for every 3.7 repeat units. For N-DMBI-H, this roughly equates to one electron potentially donated for every 3.2 repeat units. A smaller (magnitude) Seebeck coefficient is indicative of a higher extent of doping; therefore, based on Seebeck coefficient measurements and the nearly comparable dopant electron to monomer ratio, (N-DMBI) 2 is the most efficient dopant in this study.
The dopant concentrations corresponding to optimal thermoelectric PF depend on the interplay of the trends in conductivity and in the Seebeck coefficient. Dopants initially increase the electrical conductivity by introducing mobile charge carriers, but at higher dopant concentrations the electrical conductivity subsequently decreases by deleteriously impacting morphology and increasing carrier scattering; whereas, as noted above, the magnitude of the Seebeck coefficient decreases. Figure 4c shows the PF for the FBDPPV systems; the best PF we obtained for a FBDPPV film is ca. 7 µW m -1 K -2 for the case of doping with 9.2 mol% (N-DMBI) 2 .
To further develop an understanding of dopant geometry-polymer morphologythermoelectric property relationships, grazing-incidence wide-angle X-ray scattering (GIWAXS), atomic force microscopy (AFM), and scanning Kelvin probe microscopy (SKPM) measurements were performed. The neat film of FBDPPV is crystalline and smooth (Figure 5a,e), and the out-of-plane multiple order scattering features (along the q z axis) of the lamellar packing (h00) and the signal for both face-on (along the q z axis) and edge-on (along the q xy axis) π-π stacking (010) are clearly observed (Figure 5a). For the doping concentrations that yielded the highest electrical conductivity for each dopant-polymer system, the lamellar distance increases from 29.6 Å (neat film) to over 31 Å in all cases (Table 1).
Notably, only (100) can be observed in the RuCp*mes + system while the two N-DMBI + systems still exhibit multiple order scattering features (Figure 5b-d). Additionally, the π-π stacking distance shows negligible change in the two N-DMBI + systems, but ordered π-π stacking is lost in the RuCp*mes + system (Figure 5b-d, Table 1). In over-doped systems, the lamellar distance in the RuCp*mes + system increases by 12% (for 46 mol% doping), whereas the lamellar spacing of N-DMBI-H (86 mol%) and (N-DMBI) 2 (43 mol%) systems are largely maintained and the π-π distances increases (Table 1, Figure S7a-c). These results may indicate that for the N-DMBI-H and (N-DMBI) 2 systems there is excess space for facile intercalation of the relatively small and more planar N-DMBI + in the lamellar alkyl side-chain region, so that it less significantly impacts the π-π stacking, while bulky RuCp*mes + resides not only enlarges the lamellar distance but also strongly disrupts the π-π interaction.
AFM and SKPM were used to further characterize the relationships between dopant, film topography, and the measured thermoelectric properties. AFM surface morphology characterization reveals that the highest conductivity films maintained the pristine fibril microstructure, independent of dopant (Figure 5e-h). Moreover, all surface roughness measurements in the optimally doped films were lower than 1 nm (Table 1). SKPM was used to probe for structural homogeneity based on spatial fluctuations in the work function (potential). [13] Homogenous surface potential mappings (Figure 5j
Since OTE materials are not perfectly crystalline, charge transport can be analyzed by using transport formalisms developed for disordered materials. [14] Additional transport parameters can be extracted through non-linear regression of both the temperature-dependent electrical conductivity and Seebeck coefficient measurements. [2c,8a,15] In this study, we performed temperature-dependent electrical conductivity and Seebeck coefficient measurements on each dopant system at the doping concentrations with the highest electrical conductivity (i.e., 23 mol% for (RuCp*mes) 2 , 43 mol% for N-DMBI-H, and 10.7 mol% for (N-DMBI) 2 ). We observed that as temperature increases, electrical conductivity increases in all systems (Figure 6a), but in contrast, the Seebeck coefficients show a less evident temperature-dependence (Figure 6b). Based on the aforementioned observations, (specifically, the UV-vis-NIR polaronic signatures (Figure 3), the charge transport dependence on the extent of doping (Figure 4), and the temperature-activated electrical conductivity (Figure 6)), we explored the possibility that these materials exhibit characteristics that are phenomenologically consistent with the Mott polaron model. The Mott Polaron model expresses electrical conductivity, σ, and Seebeck coefficient, S, as functions of material charge transport constants that can be isolated with temperature dependent measurements (Equation 1,2). Here,
where e is the electron charge. 𝜎 ! is the pre-exponential conductivity that heavily depends on film morphology and hopping distance, and represents a maximum electrical conductivity achievable. E is the (average) ionization energy of the donor states and W H is the energetic spread of states. Collectively, E and W H (i.e., E + W H ) represent an Arrhenius activation energy (𝐸 ! ) that is related to the energy barrier for charge transport. [17] In Equation 2, S 0 is a constant. [15b] In a similar study, Emin, Crispin, and coworkers attributed S 0 to be a nearly temperature-independent constant associated with the bipolaron carrier concentration in doped poly(thiophene). [15a,15c] The carrier concentration contribution to the Seebeck coefficient is often expressed as an entropy of mixing term whose functional form is
, where c
is the ratio of transport-active polarons to thermally accessible hopping sites, but the functional form (admittedly) can vary depending on the polaronic species interactions. Based on these previous studies, we express the Seebeck coefficient as: . Values for 𝜎 ! , 𝐸 ! , and S 0 were evaluated for statistical significance using a t-tests and a 95% confidence interval. It was found that 𝜎 ! , 𝐸 ! , and S 0 can be significantly extracted from the collected data with P-values orders of magnitude lower than the significance level (0.05); P-value for 𝜎 ! on the order of 10 -29 to 10 - 20 , for 𝐸 ! on the order of 10 -29 to 10 -25 , for S 0 on the order of 10 -5 to 10 -4 . Therefore, we can reject the null hypothesis that these transport constants cannot be significantly extracted from the (arguably sparse) data set. (85 µV K -1 ) to N-DMBI-H (58 µV K -1 ). One reason for this observed difference could be that the polaronic species created in each dopant system could have different inter-polaronic interactions [16] and therefore different ratios of transport-active polarons to thermally accessible hopping sites. This idea is further explored in Figure S9, comparing different polaronic interaction models. Another reason could be that different doping counterions could have different impacts on energy-(in)dependent scattering processes (as seen in the Kang-Snyder empirical model, Figure S10). [17] Although some uncertainties exist about the observed qualitative trends between S 0 and S, we observe that (N-DMBI) 2 is the best choice for the high PF in this study, and that additional temperature-dependent thermoelectric property measurements are needed (broadly) for the organic thermoelectrics community to better understand the underlying transport characteristics.
In
ToC figure
General: All commercially available chemicals were used without further purification unless otherwise noted. N-DMBI-H were purchased from Sigma-Aldrich. FBDPPV were synthesized following the previously reported procedures. [S1] All operations involved in dimer synthesis were performed under an atmosphere of nitrogen using standard Schlenk techniques or in a glove box. Toluene and THF were dried using a solvent purification system from MBraun, benzene and hexane were dried using sodium, and NEt 3 was stored over KOH, and distilled prior to use. Sodium amalgam (1 wt%) was prepared immediately prior to use by addition of small pieces of Na metal to vigorously stirred Hg (electronic grade, 99.99%). 1 H and 13 C{ 1 H} NMR spectra were recorded on a Bruker AVIIIHD 500 MHz spectrometer and were referenced to tetramethylsilane using the residual proton signal of the solvent and the carbon resonances of the deuterated solvent, respectively. Mass spectra were measured on an Applied Biosystems 4700 Proteomics Analyzer using ESI mode. Elemental analyses were carried out by Atlantic Microlabs using a LECO 932 CHNS elemental analyzer.
Electrochemical data were acquired using cyclic voltammetry in 0.1 M n Bu 4 NPF 6 in dry THF under nitrogen, using a CH Instruments 620D potentiostat, a glassy carbon working electrode, a platinum wire auxiliary electrode, and, as a pseudo-reference electrode, a silver wire anodized in 1 M aqueous potassium chloride solution. A scan rate of 50 mV s -1 was used and ferrocene was used as an internal reference.
(RuCp*mes) 2 : This compound was synthesized by a modification of previous procedures, which used hazardous liquid alkali-metal reductants, either Na-Hg (which, in our hands, gives a lower yield than the literature) [S2] or Na-K (which gives a high yield, [S3] but is highly pyrophoric). Herein we used a stage 1 silica-gel supported Na-K alloy, NaK 2 -SG(I), a commercially available solid that is stable to dry air. [S4] Specifically, a slurry of [RuCp*mes] + PF 6 - [S5] (2.00 g, 3.98 mmol) in anhydrous THF was added to NaK 2 -SG(I) (from Sigma-Aldrich, 4.96 g, 5 eq) under inert atmosphere. The reaction was stirred for 1 h at room temperature. The solution was then decanted from the reductant via cannula and evaporated under reduced pressure. The solid residue was dissolved in toluene and the resulting solution was filtered through Celite, evaporated under reduced pressure, and dried under vacuum to yield pure (RuCp*mes) 2 as a pale yellow solid (1.26 g, 88%), 1 H and 13 C{ 1 H} NMR spectra of which were consistent with previous reports. [S2,S6] [RuCp*mes] + I -: [RuCp*mes] + PF 6 - [S5] (0.40 g, 0.78 mmol), [N-DMBI] + PF 6 -: A mixture of N,N'-dimethyl-o-phenylenediamine [S7] (350 mg, 2.57 mmol) and 4-(dimethylamino)benzoylchloride (465 mg, 2.53 mmol) was heated to reflux in toluene (20 mL) under nitrogen for 1 h with vigorous stirring. After allowing to cool, the solids were collected by filtration, washed with copious hexane, and dried under vacuum. The solids were dissolved in water and the mixture was filtered to remove insoluble residues. An aqueous solution of NH , 153.2, 133.2, 132.8, 127.4, 113.7, 112.6, 106.6, 40.0, 33.4 were corrected for absorption using the SADABS program. [S14] The crystal structures were solved by direct methods and refined by a full-matrix least squares technique on F 2 with anisotropic displacement parameters for non-hydrogen atoms. All Hydrogen atoms were geometrically placed and refined using a riding model. Crystal and refinement parameters are summarized in Table S1. The crystallographic data may be obtained in CIF format from the Cambridge Crystallographic Data Centre (www.ccdc.cam.ac.uk); the deposition numbers for the two structures are CCDC 1886261 and 1886262, respectively.
Absorption Spectroscopy: Absorption spectra were recorded on PerkinElmer Lambda 750 UV-vis-NIR spectrometer. The samples were encapsulated to avoid exposure to ambient air during measurement.
The GIWAXS data were recorded at beamline BL14B1 of the Shanghai Synchrotron Radiation Facility (SSRF) at a wavelength of 1.2396 Å. BL14B1 is a beamline based on bending magnet and a Si (111) double crystal monochromator was employed to monochromatize the beam. The size of the focus spot is about 0.5 mm and the end station is equipped with a Huber 5021 diffractometer.
NaI scintillation detector was used for data collection.
and SKPM studies were performed with a Cypher atomic force microscope (Asylum Research, Oxford Instruments). The surface morphology and the potential mappings were also recorded with a scan rate of 2 -3 Hz at AC mode (noncontact mode). Data analysis was performed by Igor Pro (Wavemetrics Inc., OR) software.
The doping method used in this study was solution blending. FBDPPV were dissolved in anhydrous toluene (3 mg mL -1 ) at 70 °C overnight prior to using. FBDPPV solution was blended with dopant solutions as a function of doping ratio at room temperature. All devices for conductivity and Seebeck coefficient measurements were fabricated using glass substrates.
The 20 nm thickness of gold layer (with 5 nm Cr adhesion layer) as electrodes were prepatterned by photolithography on the glass with a channel length of 100 µm and a channel width of 500 µm as contact pad. Prior to use, the substrates were cleaned with acetone, cleaning agent, deionized water (three times) and isopropanol under ultrasonics, and then were dried with a nitrogen flow. Doped thin films were deposited on the cleaned substrates by spin-coating at 1500 rpm for 30 s, and then annealed at 120 °C for 8 h for the hydride dopant N-DMBI-H, and 80 °C for 15 min for the dimer dopants (RuCp*mes) 2 and (N-DMBI) 2 . The conductivity (resistance) was collected by four-probe measurement in a N 2 glove box with Keithley 4200 SCS semiconductor parameter analyzer. The film thickness (ca. 20 nm) was determined by AFM. The Seebeck coefficient measurements were done in vacuum. The doped films for Seebeck measurements were deposited by the same procedures as those in conductivity measurements. The doped films were patterned to isolate the heater from the semiconductor and avoid electrical crosstalk with the thermal voltage probes and reduce the gate leakage current. The Seebeck coefficient is calculated by S = ΔV therm / ΔT, where ΔV therm is the thermal voltage between the hot and the cold ends of the device under a temperature difference, ΔT. Data were collected from 270 to 330 K. The ΔV therm was monitored by Keithley 4200 SCS, and the temperature difference was introduced by Joule heat (heater) and a liquid-nitrogen cooling system. To accurately establish the temperature difference, ΔT, between the two contact pads, two temperature-sensing wires (5 nm Cr/20 nm Au bilayer)
were introduced on the hot and the cold ends and were aligned with the patterned polymer layer. The temperature coefficient of resistance (TCR) of the temperature sensing wires was calculated from the slope of the measured resistance versus temperature. The resistance of the metal wires is linear correlated with the temperature. TCR was found to be 0.307 Ω K -1 with R 2 = 0.9999. By monitoring the resistance evolution of the temperature sensing electrodes, the accurate temperature of the contact pads was obtained by
)/TCR. The temperature difference was then given by the difference in temperature between the hot and the cold ends ΔT = T h -T c . The device architecture for Seebeck coefficient measurements is shown as following:
Polaron Interaction Model: Please see Figure S10.
Comparison to Kang-Snyder Model: Please see Figure S11.
Ultraviolet Photoemission Spectroscopy (UPS) and X-ray Photoelectron Spectroscopy (XPS): XPS and UPS were conducted on a Kratos AXIS AXIS Supra/Ultra Photoelectron Spectrometer under an ultrahigh vacuum of about 3×10 -9 Torr with an unfiltered He I gas discharge lamp source (21.22 eV) and a monochromatic Al Kα source (1486.7 eV, θ = 90°) as the excitation source, respectively. Al Kα source operated at 14 kV and 15 mA. The instrumental energy resolution for UPS and XPS were 0.1 eV and 0.5 eV, respectively. Data analysis was performed by CasaXPS software. For sample preparation, all films were deposited on heavily doped n-type Si wafers in a N 2 glove box and transferred through a transport system without air exposure into the spectrometer analysis chamber. In common with other DMBI-based with different 2-substituents, [S8] (N-DMBI) 2 exhibits an irreversible oxidation wave in its cyclic voltammogram and the reduction of the corresponding monomer cation is seen following scanning of this oxidation. Similarly the oxidation of the dimer is seen in the voltammogram of DMBI + following reduction of the cation. The peak oxidation potential of -0.75 V vs. FeCp 2 +/0 falls within the range reported for other DMBI dimers. The effective redox potential of the dimer, E(D + /0.5D 2 ), will depend on both the monomer potential (-2.38 V) and the strength of the C-C bond in the dimer (E(D + /0.5D 2 ) = E(D + /D) -0.5ΔG diss (D 2 )); [S3,S9] if the latter is assumed to be the same as that previously determined for (Fc-DMBI) 2 , [S9] E(D + /0.5D 2 ) would be ca. -2.0 V. In any case, reduction of TIPS-pentacene to its radical anion indicate that E(D + /0.5D 2 ) for (N-DMBI) 2 is at least as reducing as -1.45 V (Figure S6). UV-vis-NIR absorbance spectra in toluene of TIPS-pentacene before (black) and after (blue) addition of (N-DMBI) 2 . The peak at ca. 750 nm is attributable to TIPS-pentacene •- . [S6] Reduction of TIPS-pentacene to its radical anion indicate that E(D + /0.5D 2 ) for (N-DMBI) 2 is at least as reducing as -1.45 V. Polaron Interaction Models: Using the Seebeck coefficients and known dopant concentrations, polymer-dopant systems were compared with Seebeck coefficient models that account for the impact of charge carrier interactions on the entropy of mixing. [S18] Figure S10a shows the full range for Seebeck coefficients and its ability to change from positive to negative as a function of doping. Figure S10b shows the absolute Seebeck coefficient of the studied systems assuming that each dopant dopes the polymer backbone, produces a charge carrying species that contributes to charge transport, and that dimers only contribute 1e -. Figure S10c shows the results if each dimer dopant contributes 2e -. Nearest Neighbor Interactions accounted for only second neighbors (b = 2). Using these models in future studies, with more robust methods of measuring dopant efficacy, could lead to optimized Seebeck coefficients based on polaronic interaction engineering. Comparing to the Kang-Snyder Model: Different s values could be understood as different transport pathways. s = 2 best fits the benzoimidazole systems based on linear regressions, but s = 3 may also be appropriate based on visual interpretations. s is a transport exponent and s = 3 is possible if ionized impurities are the dominant scattering mechanism for charge carriers. [S13a] Because s may equal 2 for benzoimdidazole systems (definitively less than 3), it could indicate that benzoimdidazole dopants scatter charge carriers differently. The absorbance shown here is extracted with peak deconvolution and fitting at 786 nm as that attributable to the 0-0 transition of the neutral FBDPPV. [S1] To better compare the quenched repeat unit/counter ion ratio, what shown here is the concentration of counterion rather than dopant concentration. The effective redox potentials for the two dimers, E(D + /0.5D 2 ), are required to demonstrate that at equilibrium no unreacted dopant remains in the UV-vis-NIR doping experiments. As discussed in the caption for Figure S5, E(D + /0.5D 2 ), will depend on both the monomer potential and the strength of the C-C bond in the dimer. [S3,S9] However, both readily reduce TIPS-pentacene in solution, which has a much lower electron affinity (3.0 eV from inverse photoelectron spectroscopy; estimated at ca. 3.4 eV from electrochemistry) than the FBDPPV (estimated at ca. 4.2 eV from electrochemistry). [S1,S6] Additionally, at room temperature the change in absorbance of our doped systems was complete in seconds; therefore, it is reasonable to assume that at low doping concentrations, the reactions between FBDPPV and (RuCp*mes) 2 or (N-DMBI) 2 is quantitative if the reaction is observed to reach equilibrium.
After extracting the absorbance with peak deconvolution and fitting, the evaluated delocalization lengths are 2.9 repeat units for FBDPPV-(RuCp*mes) 2 system and 5.2 repeat units for FBDPPV-(N-DMBI) 2 system. The peaks that emerge at around 402 eV are attributable to N-DMBI + cation (Figure S13a). [S10a] Although polymer FBDPPV was blended with a lower dopant ratio of (N-DMBI) 2 (10.7 mol%) than of N-DMBI-H (43 mol%), the (N-DMBI) 2 -doped FBDPPV system showed a visually higher ratio of cation (around 402 eV) to polymer and neutral dopant (around 400 eV) than FBDPPV doped with N-DMBI-H. As a result, it is reasonable to believe that (N-DMBI) 2 has a higher doping efficiency than N-DMBI-H. On the other hand, the absence of signal at around 281.3 eV that corresponds to the Ru 3d orbital in (RuCp*mes) 2 and the presence of signal at around 282.2 eV that corresponds to the lower-electron density Ru 3d orbital in RuCp*mes + reveal that (nearly) all (RuCp*mes) 2 became RuCp*mes + . b) The incident energy is 21.22 eV and the secondary Electron Cutoff energy was 16.8 eV (pristine), so all graphically obtain values for the pristine film were then shifted by 4.4 eV such that the cutoff energy is representative of the vacuum energy level for the film sample. When measuring, the Fermi level was aligned to 0, but after secondary electron cutoff shifting, we determine the pristine Fermi level is 4.4 eV lower (more stable) than the vacuum level. c) The HOMO edge value was measured to be at 2.0 eV below the Fermi level, where the signal intensity dramatically increased and indicative of the valence band. d) After the shift, we determine the HOMO level for the pristine sample to be 6.4 eV below vacuum level. Similar procedures were applied to subsequent systems.