Dinaphthotetrathienoacenes: Synthesis, Characterization, and Applications in Organic Field‐Effect Transistors

Abstract The charge transport of crystalline organic semiconductors is limited by dynamic disorder that tends to localize charges. It is the main hurdle to overcome in order to significantly increase charge carrier mobility. An innovative design that combines a chemical structure based on sulfur‐rich thienoacene with a solid‐state herringbone (HB) packing is proposed and the synthesis, physicochemical characterization, and charge transport properties of two new thienoacenes bearing a central tetrathienyl core fused with two external naphthyl rings: naphtho[2,3‐b]thieno‐[2′′′,3′′′:4′′,5′′]thieno[2″,3″:4′,5′]thieno[3′,2′‐b]naphtho[2,3‐b]thiophene (DN4T) and naphtho[1,2‐b]thieno‐[2′′′,3′′′:4′′,5′′]thieno[2′′,3′′:4′,5′]thieno[3′,2′‐b]naphtho[1,2‐b]thiophene are presented. Both compounds crystallize with a HB pattern structure and present transfer integrals ranging from 33 to 99 meV (for the former) within the HB plane of charge transport. Molecular dynamics simulations point toward an efficient resilience of the transfer integrals to the intermolecular sliding motion commonly responsible for strong variations of the electronic coupling in the crystal. Best device performances are reached with DN4T with hole mobility up to μ = 2.1 cm2 V−1s−1 in polycrystalline organic field effect transistors, showing the effectiveness of the electronic coupling enabled by the new aromatic core. These promising results pave the way to the design of high‐performing materials based on this new thienoacene, notably through the introduction of alkyl side‐chains.


Experimental procedures
General All reagents were purchased Sigma-Aldrich (now Merck), VWR, Acros, Alfa Aesar, TCI and Fluorochem and were used without further purification. Technical grade solvents were purchased from Chem-Lab and used as supplied. Anhydrous solvents as chloroform, dichloromethane, N,Ndimethylformamide and tetrahydrofurane were distilled using common methods. Air-and/or moisture-sensitive liquids and solutions were transferred via a syringe or a Teflon cannula. Analytical thin-layer chromatography (TLC) was performed on aluminum plates with 10-12 μm silica gel containing a fluorescent indicator (Merck silica gel 60 F254). TLC plates were visualized by exposure to ultraviolet light (254 nm and 365 nm). Flash column chromatography was performed on Grace Davisil LC60A (70-200μm) silica. All NMR spectra were recorded on Jeol 400 MHz spectrometer. Chemical shifts are reported in parts per million (ppm, δ scale) from tetramethylsilane for 1 H NMR (δ 0 ppm in chloroform and 1,1,2,2-tetrachoroethane) and from the solvent carbon for 13 C NMR (e.g., δ 77.16 ppm for chloroform). The data are presented in the following format: chemical shift, multiplicity (s = singlet, d = doublet, t = triplet, m = multiplet), coupling constant in hertz (Hz), signal area integration in natural numbers, assignment (italic). The Mass Spectrometry analyses have been performed in the Organic Synthesis & Mass Spectrometry Laboratory at the University of Mons (Prof Pascal Gerbaux) using MALDI-MS on a Q-TOF Premier mass spectrometer (Waters, Manchester, UK) in the positive ion mode.

2-bromo-3-methylthionaphthalene 2
To a solution of 2,3-dibromonaphthalene 1 (1.00 g, 3.50 mmol) in DMF (35 mL) was added sodium thiomethoxide (355 mg, 1.45 eq) and the mixture was heated at 110°C overnight. 0.65 eq of sodium thiomethoxide were added to reach completion of the reaction and the mixture was heated at the same temperature for 24h more. Then the solvent was evaporated under reduced pressure, the residue was suspended in water and the mixture was extracted 2 times with EtOAc. The combined organic layers were dried with MgSO4 and evaporated under reduced pressure to give 1.229 g of crude compound. Column chromatography (pure heptane) afforded 764 g of pure 2 as a white solid with 87% yield. 1

1-methylthio-2-trifluoromethanesulfonylnaphthalene 8
To a solution of 1-methylthio-2-naphthol 7 (3.53 g, 18.6 mmol) in DCM (90 mL) at 0°C were successively added pyridine (4.73 mL, 58.4 mmol, 3.15 eq) and trifluoromethanesulfonic anhydride (3.60 mL, 21.3 mmol, 1.15 eq) under argon atmosphere and the mixture was stirred at room temperature overnight. The reaction was quenched by the addition of an aqueous solution of HCl (90 mL, 1M), the phases were separated, and the aqueous layer was extracted with DCM. The combined organic layers were dried with MgSO4 and evaporated under reduced pressure to afford 6.51 g of crude product that was purified by column chromatography (heptane/DCM 9:1) to afford 5.64 g of pure 8 as a white solid with 95% yield. 1

Experimental UV-vis spectra
The optical absorption spectra, shown in Figure S4, were recorded with a Perkin Elmer's Lambda 950UV/Vis/NIR spectrophotometer. A background correction was performed, prior to measurement, for 100% transmittance or 0 absorbance over the wavelength range. The spectra of DN4T and isoDN4T were both taken both in solution (1,2,4 trichlorobenzene) and in thin-film (quartz glass substrate). Thin films of 10 nm nominal thickness were grown by thermal evaporation in ultra-high vacuum (UHV 10 -8 mbar). Maximum absorption wavelengths of solutions of DN4T and isoDN4T in TCB are 421 and 394 nm respectively, and 448 and 348 nm for thin films, respectively. In the case of isoDN4T in solution, strong absorption from aggregates is observed at a concentration of ~1 mg/mL concentration, as can be seen from the lower energy shoulder of the peak at 3.15 eV in Figure  S4 (c). Upon dilution we observe that the aggregate absorption is reducing until its essentially gone around ~0.02 mg/mL.

Single-crystal X-ray diffraction (SCXRD)
The diffraction data of DN4T and isoDN4T were collected on a Rigaku Oxford Diffraction XtaLAB Synergy diffractometer equipped with a HyPix-6000HE area detector using Cu Kα (λ = 1.54184 Å) from PhotonJet micro-focus X-ray Source from PhotonJet. The diffraction images were processed and scaled using the CrysAlis Pro software suite (Rigaku Oxford Diffraction. CrysAlis Pro Software system, version 1.171.40.25a, Rigaku Corporation: Oxford, UK, (2018)).The structure was solved using the charge-flipping algorithm, as implemented in the program SUPERFLIP [2] and refined by full-matrix least-squares techniques against F 2 using the SHELXL program [3] All non-H atoms in the structure were refined with anisotropic thermal parameters, and hydrogen atoms were placed in calculated positions. The structures were examined using the Addsym subroutine of PLATON [4] to ensure that no additional symmetry could be applied to the models. Crystallographic data collection and refinement parameters are collated in Table S1 and full crystallographic details in CIF format have been deposited in the CCDC, reference numbers 2103738 and 2103739 for DN4T and isoDN4T, respectively. Measurements on L-DBTTA were made with a Rigaku Synergy-i instrument using CrysAlis Pro software for data collection and reduction. The solution was found by direct methods using SHELXT and the structure was refined to convergence against F 2 using all unique reflections and the program SHELXL-2018 as implemented within WINGX. [3,5] All H atom positions were observed by difference synthesis but the H atoms were placed in idealized positions and refined in riding modes. C-H bond length was 0.95 Å and Uiso(H) = 1.2Ueq(C) of parent atom. Selected crystallographic details and refinement parameters are given in Table S1 and full crystallographic details in CIF format have been deposited with the CCDC, reference number 2092479.
Despite the good reliability of the crystal structure of DN4T solved from SCXRD results, the possibility for a centrosymmetric molecule to pack in a polar space group is actually still a debate, [6] and as a matter of fact one suggestion resulting from the PLATON analysis of this crystal structure was to add an inversion center to it, yielding in the space group P 21/m instead of the current P 21. However, the DN4T molecule is not planar according to the XRD results (notably showing two bend angles of 5.6° and 8.4° between the tetrathienyl core and the two naphthalene rings), and in addition the molecule is not perpendicular to the b axis of the crystal lattice. Moreover, switching from P 21 to P 21/m led to only 80% overlap of atoms with respect to the raw data according to PLATON, and finally the initial crystal structure was already ordered and occupied reasonable displacement parameters. We thus kept the P 21 the space group in DN4T's crystal structure for being the one that effectively fits the best the collected data from SCRXD on DN4T crystals.
Nevertheless, the crystal structure of DN4T also initially included a commonly used DFIX restraint on the length of the C-C bonds of the naphthyl rings, that could be inappropriate for this particular crystal structure. We thus tried to remove the DFIX restraint on the C-C bonds, which slightly improved the R-factor of DN4T's crystal structure (see Table S1, second column). However, as all the calculations coming from DN4T's crystal structure in this work (i.e., the determination of the transfer integrals, the molecular dynamics, and the kinetic Monte Carlo simulations) have been based on the initial crystal structure (see section 5 of this document), we performed another transfer integrals calculation from the DFIX restraint-free one as a comparison. As a matter of fact, the values calculated using the DFIX restaint vs DFIX restraint-free crystal structures are 93 meV vs 99 meV, 36 meV vs 33 meV and 59 meV vs 61 meV, giving evidence of negligible differences in the electronic coupling values between the two crystal structures. We thus considered all the results coming from the initial crystal structure of DN4T presented in this work to be satisfyingly accurate, and most importantly not affecting the comparison of its properties with the other aromatic cores also discussed in this article. Note that consequently, for the sake of consistency, the corresponding CIF file related to DN4T available with this article is the one from which all the calculations have been done, i.e., still including the DFIX restraint on the length of the C-C bonds of the naphthyl rings.
172 (2) 172 (2) 100 (2) 100 (2) Crystal  Table S1. Crystal data collection and structure refinement for DN4T, isoDN4T and L-DBTTA. The first column presents the data corresponding to the crystal structure of DN4T solved with a DFIX restraint on the length of the C-C bonds of the naphthyl rings and is the one that has been used in all the calculations regarding this molecule in this work. The second column (*) present the data corresponding to the crystal structure of DN4T solved without this DFIX restraint.

Hirshfeld Surface Analysis
The Hirshfeld surfaces [7,8] were generated through CrystalExplorer17 software. [9] In CrystalExplorer, X-H bond lengths were normalized to be the same as those obtained from neutron diffraction experiments (e.g. C-H=1.083 Å). [10] The normalized contact distance (dnorm) based on de (the distance from the point on the surface to the nearest nucleus external to the surface) and di (the distance from the point on the surface to the nearest nucleus internal to the surface) and van der Waals radii of the atom is calculated by : where and are the van der Waals radii of the atoms. The parameter dnorm is negative or positive, which is illustrated by a surface with a red-white-blue colored graph. On the surface, red spots represent the lengths shorter than the van der Waals distance, whilst white and blue regions show lengths around and longer than the van der Waals distance, respectively. 2D fingerprint plot derived from a Hirshfeld surface [7,11] reveals visually the frequency of each combination of de and di over the molecular surface. The color on the plot with a range from blue (relatively few points) through green (moderate fraction) to red (highest fraction) reflects the contribution from different interatomic contacts. Curvedness surface, [8] a measure of "how much" shape is useful to measure curvature, offers further chemical insight into molecular packing. Low values of curvedness designate essentially a flat region of the surface and may be a sign of π-π stacking in the crystal. High curvedness is highlighted as dark-blue edges and tends to divide the surface into patches. Curvedness, C, is given by : Where, k1 and k2 are principal curvatures.

Variable Temperature Powder X-Ray Diffraction
Powder X-ray diffraction patterns in different temperature points were obtained using a Panalytical Empyrean X-ray diffractometer [Cu Kα (λ = 1.54184 Å) radiation] with a TTK 600 variable temperature stage attached. For both DN4T and isoDN4T, the samples were ground using a mortar and pestle and mounted on the sample holder over a zero-background film. The program on the diffractometer was set to obtain the patterns in vacuum using a gonio scan with 2θ range from 5 ° to 40 °, scan speed of 0.11 °/s and steps of 0.13 °, in different temperatures, varying from 93 K to 303 K in steps of 30 K.
The unit cell parameters of DN4T were obtained using XCell program within Reflex Powder Indexing tool, which is part of Materials Studio software [12] and space group P 21. Pawley refinement [13,14] with ultra-fine quality with Pseudo-Voigt function was used for the refinement of the FWHM, profile and line-shift parameters, as well as unit cell parameters. For isoDN4T, patterns at variable temperature were indexed using EXPO2014 software. [15] Unit cell parameters and crystal structure obtained from single-crystal x-ray diffraction were used as input. The calculated powder patterns were refined against the experimental pattern using Rietveld refinement, [16,17] with 2θ range between 8° to 30°, automatic refinement of the profile, non-structural parameters with the Le Bail fitting [18] and line shifts. Calculated diffractograms were obtained from the crystal structures presented in this work using the Powder Pattern tool on Mercury software. [19,20] X-Ray diffraction patterns are illustrated in Figure  S8, and details of the unit cell parameters and fitting of the experimental powder patterns can be observed in Table S2. As it can be observed in Figure S9, there are no great variations in the unit cell parameters from low temperature (93 K) to room temperature (303 K) with data for each axis or angle varying less than 1.5 % in comparison with the value obtained by single-crystal XRD.

Thermogravimetric analysis (TGA)
Thermogravimetric analysis (TGA) measurements were conducted on a Pyris 6 TGA instrument with Pyris software. About 7 mg of DN4T and isoDN4T were placed in an open ceramic crucible and were scanned at a rate of 5 °C/min under a nitrogen flow at 20 mL/min. TGA curves reveal that sublimation of both compounds starts around 400°C. First derivatives clearly show that no thermal event takes place before 400 °C for both compounds, suggesting that DN4T and isoDN4T possess high thermal stability. Figure S10. TGA curves of DN4T (red solid line) and isoDN4T (blue solid line); first derivatives of DN4T (red dotted line) and isoDN4T (blue dotted line). The data were recorded with a heating rate of 5°C/min.

Differential Scanning Calorimetry (DSC)
DSC analysis was performed on a Netzsch Polyma 214 instrument with Proteus software. Approximately 5 mg of DN4T and isoDN4T were placed in a covered aluminum crucible with pierced lids. The heating and cooling rates of 10°C/min were adopted. The dry nitrogen acted as both purge and protect gas at 20 mL/min. No phase transitions are observed up to 300°C.

Methods
The Gaussian 16-A03 package was used to optimize the geometry, to compute the electronic structure of the neutral isomers and to estimate their reorganization energies at the Density Functional Theory (DFT) level using the B3LYP functional and a 6-31G(d,p) basis set. [21] The reported experimental crystal structures at 100K/173K for DN4T/Iso-DN4T were used as input to calculate transfer integrals between close neighbors in the frozen crystal structure. The ADF package was used to estimate these parameters at the DFT B3LYP/DZ level of theory within a fragment orbital approach. [22,23] In practice, we created a 3×3×3 supercell for each molecule and computed electronic couplings among all pairs of close neighbors involving molecules in the unit cell at the center of the system. A close neighbor to a given molecule A is defined as any molecule B for which at least one atom is within a 5 Å range of any atom of molecule A. The reported experimental X-ray structures were also used as inputs for the Molecular Dynamics simulations on the crystals, aimed at analyzing the fluctuations in the transfer integrals induced by thermal disorder. To do so, we exploited the oplsaa force field, as assigned by the LigParGen server [24] and implemented in the LAMMPS package; [25] this is motivated by the fact that this force field was recently applied with success to organic semiconductors made of fused aromatic rings to depict lattice vibrations and their role in heat transport. [26] The force field was modified by quadrupling the coefficients of the improper terms to avoid excessive bending of the molecules. 4×4×4 supercells were generated, optimized at 0K and heated up first at 100K, then at 298K in the NVT ensemble for a 1ns recording run. 500 snapshots of the dimers of interest were selected in the final 3ns of the NVT run at room temperature. Mobility anisotropy plots were computed in the scope of a hopping regime, with a Kinetic Monte Carlo algorithm based on the first reaction method. [27] In this method, a single charge is set at a localized site (i.e., molecular unit) at t=0 and hopping rates are computed as a function of the transfer integrals and reorganization energies previous calculated at the quantum-chemical level. Waiting times are generated randomly for each hopping process, and the one with the shortest waiting time is selected. The simulation clock is then advanced by the chosen waiting time, the distance d travelled by the charge is incremented by the distance between the two centers of mass of the molecules involved in the charge transfer. This cycle repeats until simulation clock reaches a time limit set to 10 -9 s. The mobility is computed according to the equation: µ = . (3) Where E is the amplitude of the electric field (set to 1000 V/cm). Mobilities were computed for different directions of the electric with regard to the unit cell vector a, from θ=0° to θ=350°.

Calculation of the reorganization energy
The contribution of each intramolecular vibrational mode to the nuclear relaxation energy occurring when going from the neutral to charged geometry has been estimated within a displaced harmonic oscillator model, while neglecting Duchinsky rotation effects [28] : with the index i running over all intramolecular vibrational modes of energy ħi. g(i) is the local electron-phonon coupling constant associated to the normal coordinate Qi that has been computed numerically and Si the corresponding Huang-Rhys factor.
with Mi the effective mass of mode i. Figure S12 depicts the corresponding distribution for DN4T and isoDN4T. Figure S12. Huang-Rhys factor associated with each vibrational mode and corresponding energetic contribution to the total relaxation energy in DN4T and isoDN4T. In both cases, the nuclear relaxation is dominated by a low-energy, acoustic-like, mode around 200 cm -1 and by a high-energy, optical, mode at 1500 cm -1 . These vibrations involve in-phase breathing of all the atoms in the molecule (at 200 cm -1 ) and out-of-phase local displacements of the atoms (the ubiquitous carbon-carbon aromatic vibration at 1500 cm -1 ). The deformation amplitude along both modes is strongly damped in the more aromatic DN4T isomer, explaining the corresponding lower relaxation energy.

Electronic couplings between dimers
To further emphasize the importance of electronic density over S atoms, we have evaluated the spatial overlap between the HOMOs in the dimers of DN4T and isoDN4T coupled in red in Figure 4 of the main text. The overlap maps displayed in Figure S13 and S14 refer to the default 95% probability density convention to represent molecular orbitals, and can be compared based on extent of bonding interactions in the overlap region. [29] From the overlap pictures, it is clear that the marked decrease in the magnitude of the transfer integral going from DN4T to isoDN4T can be explained by the reduced wavefunction overlap over the cores in the latter case, associated with the lack of electron density on the S atoms in the HOMO of isoDN4T (Figure 4 of the main text).

DN4T
isoDN4T Another interesting feature of DN4T is the asymmetry in the herringbone transfer integrals ( Figure  4 of the main text), which is consistent with the absence of an inversion center in the system. While the overlap maps of two different dimers along a diagonal direction are relatively similar ( Figure  S14), there is an overall greater degree of overlap in the 99 meV coupled dimer, particularly in the internal thiophene moieties, close to the sulfur atoms.

Molecular dynamics simulations
Since previous works on rod-like semiconductors demonstrated that the relative displacements of molecules over their long axis (mostly aligned along the Z axis) is detrimental for charge transport due to a strong modulation of the transfer integral values, we have further computed the evolution of the coupling for the two dimers of DN4T displayed in Figure S14 as a function of their relative translation (Z). [30,31] Figure S15, left panel shows that the transfer integrals initially at 99 meV (blue) and 33 meV (green) follow opposite variations as a function of the shift along the Z axis. This can be explained on the basis of short-distance sulfur-sulfur contacts (below 4 Å). At ΔZ=0Å (i.e., the crystal geometry) we count 4 short contacts in the 99 meV pair but only 2 in the 33 meV pair. At ΔZ = -2Å ( Figure S15, 2), the relative number of short S-S contacts between the two pairs has swapped, and so does the relative magnitude of the transfer integrals. However, Molecular Dynamics simulations revealed that the displacement of the molecules along the Z axis does not exceed 0.5 Å. Figure S15, right panel shows the ΔZ distribution of the geometrical centers of mass of the dimers located in the herring bone layer along an NVT run at 300K, respecting the color code set in Figure  4 of the main text. It appears that the maximum Z shift DN4T molecules reach is on the order of 0.22 Å for the green and blue dimer. From Figure S15 left panel, this means the system never reaches a state where the electronic couplings are inverted (above a 1.5 Å shift in absolute value).

Kinetic Monte Carlo simulation for hole mobility assessment on a pure hopping regime
In a first attempt to assess how the different connectivity in DN4T vs isoDN4T affects charge transport, we have performed kinetic Monte Carlo (kMC) simulations of hole transport assuming a purely hopping regime in the dynamic limit. We consider the two following approximations: (i) dynamic and static energetic disorder prompts the formation of charge carriers that are localized on single molecules, (ii) charge migration is slow enough so that each molecule experiences effective electronic interactions with its neighbors that equate to their time averaged values. On the basis of a qualitative analysis of the comparative amplitudes of the electronic transfer integrals versus reorganization energy, the first approximation very likely holds for isoDN4T. DN4T on the other hand might be at the limit of the validity of this hypothesis since the electronic couplings are much closer to the reorganization energies. Additional calculations are definitively required to address this question. The second approximation results in an upper limit of the mobility in the hopping regime, such results are said to correspond to the dynamic limit. The transfer rates injected in the KMC simulations are obtained in the framework of the Marcus-Levich-Jortner formalism [32] as: A single high-frequency, effective, vibrational mode with an energy ħeff estimated to 0.2 eV (related to the ~1500 cm -1 from Figure S12), accounts for quantum-mechanical nuclear tunneling (using a Huang-Rhys factor of 0.76 for DN4T and 1.15 for isoDN4T) while the low energy modes are treated classically through the external reorganization energy s set equal to 0.1 eV. Part of it arises from the 200 cm -1 mode which results in a contribution of 0.05 eV while the contribution from the environment is roughly estimated to be also 0.05 eV. [21] Total simulation time was set to 10 ns. The mobility of the equilibrium structure was obtained by injecting the mean electronic couplings of the corresponding dimers in the transfer rate equation. These mean values were obtained by retrieving the geometry of each dimer amongst 500 snapshots during an NVT run at 300K and computing the transfer integrals. Each snapshot is separated by 2 ps. The maximum mobility the charge can reach in this framework is defined by the dynamic limit: when the hopping rate is much lower than the frequency of the intermolecular modes, the crystal has the time explore many configurational states and the charge will most likely hop when a high transfer integral dimer configuration will appear in the thermal noise. This equates to injecting <t²> in the master equation. Results are displayed in Figure 5 of the main text.

Determination of the ionization energy by UPS
Ultraviolet photoelectron spectroscopy measurements were carried out with a UHV system (10 -10 mbar) using the monochromatized He Ia line (21.22 eV) that was produced with a Helium discharge lamp. All spectra were recorded at room temperature using a hemispherical SPECS Phoibos 100 analyzer. Prior to the measurement, a polycrystalline gold foil was used to calibrate the detector and also to determine the energy resolution. For the valence region spectra, the pass energy was set at 5 eV that gives an energy resolution of 120 meV. The work function (Φ) values were determined from the secondary electron cutoff (SECO) spectra, measured with −10 V sample bias and a pass energy 2 eV that allows an energy resolution of 78 meV. Before the UPS measurements, the molecules were grown on Au (111) and PEDOT:PSS by thermal evaporation under ultra-high vacuum (UHV 10 -8 mbar). The Au (111) substrates were cleaned prior to deposition to remove any carbon contamination on the surface. For this purpose, the substrates were introduced to the UHV preparation chamber (10 -8 -10 -9 mbar) and two cycles of 30 min sputtering with Argon ions and annealing at 450 °C were carried out. Films of ~1-2 Å nominal thickness were then grown for both molecules which corresponds approximately to the thickness of a flat-lying DN4T (isoDN4T) monolayer. The nominal thickness was monitored via a quartz crystal microbalance. The UPS spectra were recorded at 45° angle between the sample and the detector as the intensity of the HOMO was maximized, due to the higher photoionization cross section for molecules lying flat on the surface, which is also the case of many other planar molecules. [33][34][35] The PEDOT:PSS samples were prepared on solvent cleaned ITO (indium tin oxide) substrates. ITO substrates were exposed to UV/ozone for 10 mins and afterwards PEDOT: PSS (Ossila Al 4083) was spin-coated at 2000 rpm from an aqueous solution. The samples were then annealed in ambient conditions for 5 mins at 180°C and then introduced in UHV where a film of ~100 Å was grown. In this case the UPS spectra were taken with normal emission angle at which the intensity of the HOMO was taking the maximum value. The ionization energy (IE) can be extracted from the UPS spectra using equation (7) : Where Φ is the work function obtained from the SECO onset as mentioned above. HIB stands for the hole injection barrier which is commonly determined from the onset of the HOMO peak. The extracted IP values from the UPS spectra showed in Figure S16 are reported in  show the x-ray photoelectron spectroscopy (XPS) survey spectra obtained using a standard hemispherical analyzer (SPECS Phoibos 100 analyzer) and a Mg X-ray source (Mg-Ka: 1254 eV). Thin films of DN4T and isoDN4T were deposited on clean Au (111) substrate by thermal evaporation in an ultra-high vacuum chamber (UHV 10 -8 mbar). The instrumental resolution was determined by measuring the crystalline Au (111) and was 0.4 eV. The survey spectra showed contributions from Au, C and S without any presence from extrinsic contamination ultimately. A more precise analysis on the higher resolution spectra of C and S helped us to determine the stoichiometry by fitting the peaks, as shown in Figure S18Error! Reference source not found. (a)-(d). For the fitting of the photoelectron peaks we used Voigt functions, which are convolution of Gaussian and Lorentzian peaks to account for both the instrumental broadening but also the natural linewidth of the energy levels. A Shirley background has been subtracted to consider the background associated with the inelastically scattered electrons. [36] The peak fitting results are listed in Table S3 and Table  S4 for the case of DN4T and isoDN4T, respectively. The stoichiometry of the DN4T and isoDN4T compounds (C26H12S4) is given by the ratio of carbon to sulfur atoms and is therefore: The stoichiometry determined by the XPS measurements relates to the intensity of the peaks corresponding to the carbon and sulfur orbitals that have interacted. The XPS intensities were corrected using the sensitivity factors , calculated from the method reported by Seah. et.al. [37] Therefore, the resulting stoichiometry is given by: Using the results from Table S3 and Table S4 we determined that the stoichiometry of the two compounds is: DN4T: Seemingly, the real stoichiometric value of 6.5 lies within the interval of confidence of the experimentally determined values. However, the small discrepancy of the values can be caused due to the weak photoelectron signals associated with the S 2p energy levels, which make harder the accurate determination of the intensity of the peaks and therefore overestimate the stoichiometry as is the case of isoDN4T.

IP measurements by PYS
The ionization potential (IP) of DN4T, isoDN4T and L-DBTTA powders has been measured by Photoelectron Yield Spectroscopy (PYS) at ambient conditions. Since the photoemission yield curves were recorded from powder samples, the measured IP value averages over all the possible crystallographic orientations. Photoemission yield spectroscopy was performed with a Riken Keiki spectrophotometer (Japan) model AC-2 with an energy step of 0.05 eV and a UV spot intensity of 10 nW. Average IP values over 3 trials gave 5.27 ± 0.02 eV for DN4T powder, 5.23 ± 0.01 eV for isoDN4T powder and 5.16 ± 0.01 eV.

Device fabrication and performances
All the thin film transistors (TFT) were fabricated on highly doped silicon wafers with 30 nm of Al2O3 (Christian-Albrecht University of Kiel, Institute for Electrical Engineering and Information Technology). The Al2O3 substrates were exposed to oxygen plasma (Diener Electronic; oxygen flow rate 20 sccm, pressure 0.50 mbar, plasma power 100 W, duration 2 min) and then immersed overnight in 1.5 mM solution of n-tetradecylphosphonic acid (TDPA, Sigma-Aldrich) in 2-propanol (Acros Organics) to obtain a 1.5-nm-thick self-assembled monolayer (SAM). Subsequently, the substrates were rinsed first in 2-propanol then in deionized water and finally in 2-propanol again and dried on a hot plate at 100 °C for 10 min.  Transfer and output curves of best performing DN4T and isoDN4T TFTs with BC and TC configurations are reported in Figure S20, and charge carrier mobilities (μ) and threshold voltages (Vth) as a function of the substrate temperature during DN4T and isoDN4T evaporation are reported in Table S6. The values of μ were calculated with the conventional gradual channel model approximation using the equations (12) and (13)

Contact Resistance
Evaluation of the channel-width-normalized contact resistance (RcW), using the transmission line method (TLM) [38,39] at a substrate temperature of 100 °C for DN4T and of 120 °C for isoDN4T TFTs was carried out following equation (14) : where Rtot is the total resistance of the transistor and Rc is the contact resistance. The model assumes that the channel resistance, but not the contact resistance, depends on the channel length. Hence, plotting the total resistance of the transistor using different channel length, it is possible to calculate the RcW and μ0, through a linear fitting. Both for the BC and TC devices of DN4T and isoDN4T, TLM was applied ( Figure S22) and RcW and μ0 were extracted ( explained by their dependance on the charge carrier mobility. [40][41][42] Considering that the injection barrier is almost similar for the two isomers due to comparable HOMO and LUMO energies, μ plays a crucial role in determining the contact resistance. Therefore, a higher μ would turn out in a lower RcW, in both BC and TC devices.   Table S7. Intrinsic mobility and channel-width-normalized contact resistance for bottom and top contact structures of DN4T and isoDN4T TFTs.

Thin film morphologies
The thin films morphologies were evaluated through atomic force microscopy (AFM, Brucker Dimension Icon) in tapping mode with Olympus OMCL-AC240TS-R3. DN4T and isoDN4T thin films of ca. 25 nm were evaporated on Al2O3 substrates treated with TDPA, using the previously mentioned procedure. The substrates were held at 40, 100 and 140 °C. Figure S23 shows the effect of temperature on DN4T and isoDN4T thin-films morphology. A terrace pattern covered by needle-shaped crystals arises from the images at all substrate temperatures. Particularly for DN4T, the quantity of these three-dimensional needle-shaped crystals is higher compared to isoDN4T samples, covering the terrace structures and making difficult to perform the image analysis. In Table S8 are reported the estimated height of the terrace steps and the grain diameters are reported. Remarkably, the longest grain diameter is achieved at a substrate temperature of 100 °C for both DN4T and isoDN4T samples, which correspond to the highest mobility values for both molecules (Figure S24), highlighting that the field-effect mobility is enhanced with increasing the grain size. Moreover, the height of the terrace steps is close to the molecular length of DN4T and isoDN4T, suggesting that in the crystalline films the molecules are oriented approximately upright with respect to the substrate. This observation was furtherly confirmed by XRD patterns recorded on DN4T and isoDN4T thin films (Rigaku SmartLab, Cu Kα radiation). The same samples used for AFM measurements were employed for XRD