Electronic engineering of a tetrathiafulvalene charge-transfer salt via reduced symmetry induced by combined substituents

A 1:1 metallic charge-transfer salt is obtained by cosublimation of ( Z,E )-(SMe) 2 Me 2 TTF and TCNQ. X-ray diffraction studies confirm that the formation of segregated stacks comprising donor and acceptor molecules in [( E )-(SMe) 2 Me 2 TTF](TCNQ). The crystal packing features lateral S ··· S interactions between TTF stacks, which is in sharp contrast to that in (TTF)(TCNQ). Structural analysis and theoretical studies afford a partial charge-transfer ( r » 0.52), leading to a system with the electronic structure close to quarter-filled. Resistivity measurements reveal that this material behaves as a metal down to 56 K and 22 K at 1 bar and 14.9 kbar, respectively. The thermopower is negative in the metallic regime, indicating the dominant role of the acceptor stacks for the observed conducting behavior. Analysis of a single-crystal EPR spectra shows the remaining spin susceptibility at 4.3 K, suggesting the importance of the Hubbard U correction. These results highlight the judicious engineering of electronic and geometrical effects on the TTF core; the combined use of methyl and thiomethy groups has decreased the TCNQ bandwidth while keeping the segregated stacks, converting the metal to insulator (M-I) transition more 4 k F like. In addition, the enhanced S ··· S contacts between the TTF stacks lead to more rapidly decreasing M-I transition temperature under pressures.


Introduction
The discovery of metallic behaviour in crystalline organic molecular materials has stimulated a plethora of different research areas covering chemistry and physics. 1 The most notable example is the charge transfer donor-acceptor compound, (TTF)(TCNQ) (TTF = tetrathiafulvalene; TCNQ = 7,7,8,8tetracyanoquinodimethane), which was the first organic conductor to present a metallic conductivity in a wide temperature range down to 59 K. 2 This compound features an anisotropic crystal structure where donor and acceptor molecules form segregated stacks which facilitate a good overlap of respectively HOMOs and LUMOs π-type frontier orbitals along the stack direction to render a 1D electronic structure. 3 The accumulated knowledge on such a system emphasizes the critical role of structural chemistry and crystal engineering. 4 In these molecular systems, a subtle change in chemical structure is known to drastically modify geometrical and electronic structures, which in turn enable to discover new exotic physical phenomena and to underpin physical understanding. 5 It is particularly interesting that the intriguing interplay of 2k F and 4k F instabilities have been extensively investigated in (TTF)(TCNQ). 4 Scheme 1. Tetrathiafulvalene derivatives with different substituents.
It has been known that tetramethyl-TTF (TMTTF) as well as hexamethylene-TTF (HMTTF) stabilize the metallic conduction, 12 whereas tetrathiomethyl-TTF (TTMTTF) makes insulating TCNQ complexes due to the mixed-stack structures in DDA and DDAA sequences. 13a This comes from the strong dimerization of TTMTTF associated with the steric hindrance of the out-of-plane thiomethyl groups. Methyl substitution increases the donor ability, but thiomethyl substitution reduces the donor ability ( Figure S5). Therefore, in order to finely control the electronic structure without destroying the segregated stacks, we have brought our attention to a TTF donor molecule with two thiomethyl-and two methyl-substituents (Z,E)-(SMe) 2 Me 2 TTF, (Z,E)-1 (Scheme 1).
At variance with classical TTF, TMTTF, TTMTTF and BEDT-TTF (bis(ethylenedithio)-tetrathiafulvalene) molecules with D 2h symmetry, such systems can exist into Z (C 2v ) and E (C 2h ) forms, which easily interconvert in solution upon oxidation, making often difficult their (electro)crystallization 6 despite their readily accessible synthesis. Such a difficulty is well demonstrated by the scarcity of the structurally characterized donor-acceptor two-chain charge transfer salts comprising such Z-or E-TTF isomers. 7 Our specific interest for 1 also comes from its first oxidation potential (E 1/2 = +0.41 V vs. SCE, Figure S1) close to that of TTF itself (+0.33 V), a consequence of the electron-rich methyl groups which compensate in part the electron withdrawing effect of the thiomethyl groups. We postulate that such a close proximity of electrochemical potentials likely triggers a non-zero charge transfer if this donor is associated with TCNQ, 8 as demonstrated below.

Co-crystallization and structural analyses
The neutral (Z,E)-1 was prepared according to the modified literature procedure, by successive deprotection of (Z,E)-(SCH 2 CH 2 CN) 2 (Me) 2 TTF with CsOH and further alkylation with MeI. 9 Cosublimation of (Z,E)-1 and TCNQ in a two-zones furnace 10 under vacuum afforded black, flake-shaped crystals of [(E)-1](TCNQ) in 3 weeks. In addition to [(E)-1](TCNQ), we have also observed the deposition of orange needle-shaped crystals of the donor molecules and yellow flake-shaped crystals of the acceptor molecules within the cosublimation tube (see Experimental section). NMR analysis on the orange crystals that remained after the sublimation indicated that it only contains E-isomer ( Figure S2). This observation strongly suggests that the Z-isomer converts to E-isomer during the vaporization in the presence of the TCNQ vapor, and therefore, exclusively affords [(E)-1](TCNQ) as a co-crystallized product. Please note that the sublimation of (Z,E)-1 without TCNQ afforded the mixture of Z-and E-isomers ( Figure S2). To the best of our knowledge, such a discrimination of isomers using co-sublimation has never been reported. It is worh to mention that the solution diffusion method for co-crystallization was not successful to obtain single crystals for further analysis, although it provided microcrystalline powders consisting of [(E)-1](TCNQ), as indicated by the powder X-ray diffraction (figure S3).  (b). Yellow, blue, gray, and pink spheres represent S, N, C, and H atoms, receptively.
The compound crystallizes in the triclinic system, P-1 space group, with both (E)-1 and TCNQ molecules on an inversion centre. The structure consists of segregated stacks of donor and acceptor molecules running both along the crystallographic a axis, and alternating along the c axis, as shown in Figure 1a.
The intermolecular overlaps along the stacks adopt the bond-over-ring motif commonly observed in highly conducting compounds, with plane-to-plane distances of 3.474 and 3.227 Å in the stacks of (E)-1 and TCNQ respectively. Note that these values are larger than those reported in the RT structure of (TTF)(TCNQ) (3.451 and 3.188 Å respectively) (Table S1), 11 an indication of possibly weaker intermolecular interactions along the stacks in [(E)-1](TCNQ). In the crystallographic ac plane ( Figure S4), all molecules are tilted in the same direction with respect to the (100) plane, with an angle of 22.23 and 31.65° for (E)-1 and TCNQ respectively, leading to an interplanar dihedral angle of 9.42°. A remarkable feature is the relative arrangement of the stacks which allows direct S···S inter-stack contacts along the b axis (Figures 1a and S4) while such lateral interactions are absent in (TTF)(TCNQ) (Figure 1b). Taken together, the arrangement and the packing of donor and acceptor molecules in [(E)-1](TCNQ) strongly differs from that of the unsubstituted (TTF)(TCNQ). 11 Analogous 1:1 salts were also reported with the D 2h -symmetry TTFs such as TMTTF and TTMTTF. The (TMTTF)(TCNQ) metallic salt crystallizes in the P2/c space group with stacks of TMTTF molecules interacting also sideways, but with large and opposite tilt angles vs. the stacking axis. 12 The (TTMTTF)(TCNQ) charge transfer complex is an insulating compound with mixed stacks, as explained in Introduction. 13 Although tetraethyl-TTF (TETTF) has a similar chemical potential located near TCNQ ( Figure S5), it also makes mixed stacks with TCNQ, (TETTF)(TCNQ) 2 , owing to the out-of-plane ethyl groups. 13b However, DEDMTTF forms segregated stacks in the TCNQ complex, 13c where the terminal methyl moieties are within the same plane as the TTF part but extend to side of the TTF. The resulting molecular packing of (DEDMTTF)(TCNQ) is very similar to the present complex (Figure 1a), and has direct contacts between the TTF stacks. These comparisons indicate that the balanced substitution between -SMe and -Me groups in the present donor 1 indeed allows for an original structure which enhances the structural dimensionality of the conducting sheets while avoiding DADA mixed stacks.
To estimate the degree of charge transfer, the bond distances within TCNQ were carefully examined using Kistenmacher's relationship (Table S2 and Figure S6). 14 The calculated charge of TCNQ is 0.52, implying a nearly quarter-filled band system. Note that this value is also coherent with r = 0.59 estimated from the C≡N stretch of the FT-IR spectrum 15 and the theoretical calculations (vide-infra). The variable-temperature electrical resistivity was measured under different pressures on single crystals along the a stacking axis (Figure 2). At 300 K, the conductivity is 118 S/cm under ambient pressure, which is comparable to the known organic charge transfer conducting compounds. 3 As temperature decreases, the resistivity decreases smoothly to reach a minimum value at 56 K before an abrupt increase, indicating the occurrence of a metal to insulator (M-I) transition. The activation energy in the insulating region is estimated at E a = 31 meV. Before the M-I transition, the resistivity drop shows a deviation around 130 K, which is confirmed by the maximum of the derivative ( Figure S7), likely related to the electronic instability in either the donor or the acceptor stacks (vide-infra).

Transport properties
Upon increasing pressure, the room-temperature conductivity increases and the M-I transition temperature significantly shifts to lower temperatures, to reach 470 S/cm and 22 K, respectively, at 14.9 kbar (Figures 2b and 2c). At this highest available pressure, the resistivity rise shows no saturation and E a decreases by half (E a = 16 meV). This is in stark contrast with (TTF)(TCNQ), where the conductivity increases less than twice and the M-I transition temperature hardly varies under similar pressures. 3a,16 The comparatively large pressure-sensitivity of the present compound is likely attributable to the enhanced structural dimensionality coming from the side-by-side S···S contacts. The applied pressure may indeed reinforce the inter-stack contacts which might ultimately lead to the suppression of the electronic Peierls instability expected for a purely 1D electronic system. As shown in Figure 3, the thermopower exhibits small negative values over the metallic region, indicating that the transport is dominated by the acceptor stacks. 17 However, in contrast to (TTF)(TCNQ) where it amounts to -40 µV/K at 300 K, the small absolute value for [(E)-1](TCNQ), -5 µV/K, indicates an increased contribution of the donor stacks to the conductivity. Upon lowering temperature, the thermopower shows a small peak around 150 K then a gradual decrease. Given the negative carriers become more dominant below 150 K, the anomaly at 150 K is probably to be associated with the setting of 4kF instability in the donor stack. Note that such a peak was observed also for (TMTTF)(TCNQ). 17 Below 70 K, the strong upturn of the thermopower is observed, which may be due to the setting of modulation in the TCNQ stacks that finally condense and open a gap around 50 K.  Figure S8). 18 Given that the two Lorentzians manifest at the same angles in all three samples investigated, this is likely an inherent property of the compound and the splitting stems from the existence of inversely tilted domains, which have borders in the ab plane ( Figure S9). In view of the extremely small deviation from the monoclinic cell (b = 90.292 o ), this seems to be inherently associated with the present crystal.  18,19 At low temperature, the spin susceptibility remains non-zero even at 4.3 K, indicating charge-spin decoupling in the compound. Note that (DEDMTTF)(TCNQ) seems to remain paramagnetic down to low temperatures as well. 13c Electronic structure calculation These bands are usually found to be wider than the experimental estimations in two-chain systems for which 4k F fluctuations associated with the TTF-based component have been observed. These observations mean that electronic repulsions in the TTF-based chains are important and that they are underestimated by the DFT approach. The simplest way to remedy for this drawback is by using an effective Hubbard correction term for S 3p orbitals which contracts them and leads to narrower HOMO bands and smaller charge transfers. 22 Since the present system contains TTF-based chains we have considered the U dependence of the electronic parameters of TTF-TCNQ, HMTTF-TCNQ and TMTTF-TCNQ (see for instance Figures S10 to S11) and found that a value U= 6 eV leads to reasonable results for band widths and charge transfers.

ESR analyses
For instance, using their room temperature crystal structure the calculated vs. experimental charge transfers (determined at 25 K) 4 for TTF-TCNQ, TMTTF-TCNQ and HMTTF-TCNQ are 0.57/0.59, 0.56/0.57 and 0.68/0.72. Thus, we have adopted this value in our study.
The calculated band structure and density of states are reported in Figure 6. The upper band at G is based on the (E)-1 HOMO and the lower band is based on the TCNQ LUMO. The two bands cross along the intrachain G-X line, at the Fermi energy. However, in these two-chain conductors the avoided crossing between the HOMO and LUMO bands is usually slightly overestimated by DFT so that we believe that the gap along G-X is smaller and probably irrelevant for most temperatures. The small but non nil dispersion along the interchain b*-direction (G-Y) of the donor band, which is associated with an S···S contact of 3.767 Å, slightly lowers the energy of the upper band and consequently overlaps with the top of the lower band, definitely leading to a band overlap inside the Brillouin zone and thus, a metallic-type behavior. The avoided band crossing occurs for a k F value of 0.13a* and consequently, the charge transfer (r = 4´k F /a*) is 0.52. Integration of the DOS curves (see Figure 6b where the total DOS as well as the separate contributions of the donor and acceptor bands are also shown) leads to r = 0.53. These values are very similar to the r = 0.52 one estimated above from the bond lengths of the acceptor thus confirming that the charge transfer must be very close to 0.5. This situation is clearly different from that of the archetypal (TTF)(TCNQ) for which the charge transfer (r = 0.59) is definitely far from the quarterfilled/empty case and thus a different physical behavior may be expected. Note that whatever the value of U used in the calculations, the difference with (TTF)(TCNQ) does not practically change ( Figure S10) so that the result is a robust one.

Comparison with (TTF)(TCNQ)
What is the origin of the charge transfer difference with (TTF)(TCNQ)? It may be argued that it is due to the decrease in the HOMO energy induced by the -SMe substituents, illustrated by the increase of first oxidation potential when compared with TTF. However, the charge transfer in these two-chain molecular conductors subtly depends on both electronic and structural factors. A decrease of the charge transfer can also occur if one or both of the HOMO and LUMO bands become narrower. The dispersion of the TCNQ band depends essentially of the C···C contacts along the intrachain a direction. As mentioned above, these contacts in [(E)-1](TCNQ) (Table S1) are clearly longer than even those for the RT structure of (TTF)(TCNQ). Hence we can conclude that the TCNQ band in the present salt must be narrower than in (TTF)(TCNQ).
The dispersion of the donor band depends essentially from the six intra-stack S···S contacts. In the RT structure of TTF-TCNQ, the same contacts are again found to be shorter (Table S1) In contrast, the decrease in the strength of the intrachain S···S interactions of the TTF core is more than compensated by the extra S SMe ···S TTF contacts, leading ultimately to an increase of the HOMO band width. However, we note that the variation for the TCNQ band is somewhat larger. In addition, the increase in the donor band width (and thus, of charge transfer) will mostly compensate the intrinsic lowering in the HOMO energy (and thus, the decrease of the charge transfer) due to the -SMe substituents. We thus suggest that the substantial decrease of the C···C contacts in the TCNQ stacks is the main reason behind the small but significant charge transfer decrease, now very close to the quarter-filled/empty value.

Low temperature behaviour scenario
What are the possible scenario for the low temperature behavior of this compound? An important factor to consider is the filling of the LUMO/HOMO bands which at low temperatures may become almost exactly quarter-filled/empty (r = 0.5) or can keep a slightly different value. In view of the transport measurements reported above and relying on analogies with other TTF-based two-chain salts 4,17 we suggest the following scenario. The small anomaly occurring around 150 K clearly seen in the thermopower measurements is probably to be associated with the setting of 4k F fluctuations in the donor stacks since the negative carriers become more dominant below 150 K. The strong upturn around 70 K may be due to the setting of a fluctuations regime in the TCNQ stacks which finally condense and open a gap at around 50 K. However, if a charge transfer very near to the quarter-filled/empty case remains at low temperatures the modulation on the TCNQ stacks could be a 2k F tetramerization and the regime change in the (E)-1 stacks could then be of the 4k F charge-ordering or Mott-dimer types, explaining that the magnetic susceptibility remains nonzero until very low temperatures. 23 X-ray diffuse scattering studies are needed in order to shed light on these issues.

Conclusions
In this study, we have successfully grown a 1:1 co-crystal of (E)-1 and TCNQ and have investigated its physical properties in detail. The obtained crystal of [(E)-1](TCNQ) features the segregated D and A stacks with the DA packing that is notably different when compared with the analogous charge transfer TCNQ salts comprising the classical TTF donors. In particular, the lateral S···S contacts between donor stacks give rise to the structural two-dimensionality in [(E)-1](TCNQ).
The band structure of the compound features close to a quarter-filled system with r » 0.52, leading to a metallic behaviour down to 56 K. Moreover, both the conductivity and the metal-to-insulator transition temperature are significantly influenced by the applied pressure, likely due to the enhanced structural dimensionality.
In particular, slight delocalization in the HOMO of the donor by means of the two thiomethyl-and two methyl-substituents leads to a situation where the donor and acceptor band widths are very similar and the charge transfer is very near the commensurate value 0.5. To the best of our knowledge the only other twochain salts of this family exhibiting both features are (DEDMTTF)(TCNQ) and (TMTTF)(DMTCNQ) (DMTCNQ = 2,7,8,. 4,13c These salts lie just at the borderline between those where the dominant structural instability changes from 2k F to 4k F . 4 Thus, it would be extremely interesting to study if this is also the case for the present salt. If it is indeed the case, small structural or electronic modifications could bring substantially different physical behaviors and provide invaluable information to settle the difficult issue of the competition between structural instabilities in twochain conductors. Work is underway to prove the ground state of [(E)-1](TCNQ) and to explore the promising role of (Z,E)-1 as electron donor to grow co-crystals with intriguing electronic structures.

General methods
Chemicals and materials from commercial sources were used without further purification. Solvents were dried using a commercial solvent purification system from Inert. The (Z,E)-(SCH 2 CH 2 CN) 2 (Me) 2 TTF was prepared according to the literature procedure. 9 Melting points were measured on a Kofler hot-stage apparatus. NMR spectra were recorded on a Bruker AV III 300 MHz NMR spectrometer at room temperature using CDCl 3 . Column chromatography was performed using Teledyne Isco CombiFlash® Rf400. Cyclic voltammetry measurements were carried out in a one-compartment cell under argon, equipped with a platinum wire, a glassy carbon, and a Saturated Calomel Electrode (SCE) as working, counter, and reference electrodes, respectively, using a Bio-Logic SP-50 potentionstat. Analyte solutions (10 −3 M solution of complex) were prepared in 0.1 M solutions of (NBu 4 )(PF 6 ) in CH 2 Cl 2 .

Synthesis of (Z,E)-1
To a solution of the (Z,E)-(SCH 2 CH 2 CN) 2 (Me) 2 TTF (300 mg, 0.75 mmol) in 75 mL of THF was added under argon a solution of CsOH·H 2 O (372 mg, 2.22 mmol) in 10 mL of MeOH dropwise at 0 °C. The resulting solution was stirred at 25 °C for 90 minutes to afford a dark mixture. MeI (0.3 mL,4.82 mmol) was then added to the flask, and the resulting mixture was stirred for 12 h to give a yellow solution. After quenching the reaction with 60 mL of distilled H 2 O, the organic solvent mixture was removed from the reaction under vacuum, and the resulting residue was extracted into CH 2 Cl 2 (4 x 60 mL). The organic extracts were dried over magnesium sulfate, filtered, and then concentrated under reduced pressure to give a red oil. The crude oil was purified by flash column chromatography on a silica gel to afford (Z,E)-1 as an orange solid (198 mg, 82 %). The product was recrystallized by slowly evaporating the saturated EtOAc solution. Column conditions: silica cartridge 12 g; solid deposit on Celite®; λ detection : (254 nm, 280 nm); CH 2 Cl 2 in light petroleum with a gradient 0 -30%. mp = 133 °C. 1 H NMR (300 MHz, Chloroform-d) δ = 2.31-2.32 (s 3H), 2. 12-2.13 (s, 3H).

Crystallization of [(E)-1](TCNQ)
Single crystals were prepared by the co-sublimation method. To opposite ends of a 30 cm thin-necked glass tube, (Z,E)-1 and TCNQ were placed. The glass tube was then sealed in vacuo with a pressure of 1 × 10 −5 mbar. In a two-zone furnace, the segments of the tube were heated to 130 and 250°C for the donor and the acceptor, respectively, to which the temperature was gradually raised from room temperature at a rate of 0.04 °C/min. The temperature of the central zone was approximately 180°C. After three weeks, black crystals grow near the center of the tube. Since a single central zone of crystals composed of a charge transfer salt was obtained, other complexes with (Z)-1 or different compositions were not obtained by this method. Note that our co-sublimation condition did not allow all vaporized donor and acceptor molecules flew to the central zone to react. Accordingly, there was also the deposition of each starting material along the tube, orange needle-shaped crystals of the donor molecules and yellow flake-shaped crystals of the acceptor molecules.

X-ray structure determination
Single crystals of (E-SMe 2 Me 2 TTF)(TCNQ) that are suitable for X-ray analysis were coated with Paratone-N oil and mounted on a MicroMounts TM rod. The crystallographic data were collected on a Bruker D8 venture diffractometer equipped with MoKα microsource and PHOTON100 detector at the Centre de Diffractométrie (CDIFX, Université de Rennes 1, France) at 150 K. Raw data were integrated and corrected for Lorentz and polarization effects using Bruker APEX3 v. 2015. 24 The structure was solved by direct methods using the SIR92 program, 25 and then refined with full-matrix least-square methods based on F 2 (SHELXL-2014) 26 with the aid of the WINGX program. 27 All non-hydrogen atoms were refined with anisotropic atomic displacement parameters.

Electrical resistivity and thermopower
Electrical resistivity was measured by the four-probe method down to 2.7 K. Four gold wires (15 µm diameter) were attached to a crystal using carbon paste. Low-frequency AC current was applied to the outer wires and the resistance was estimated from the voltage measured at the inside wires. Pressure was applied to a crystal by a conventional clamp cell with an oil Daphne #7373 as a pressure medium. 28 The absolute value of the pressure was estimated from the resistance change of a manganin wire at room temperature.
Results for 9.1 kbar to 14.9 kbar are taken using a different sample from that for 1 bar to 7.4 kbar. For thermopower measurement, a crystal was attached to two copper blocks via gold foils using carbon paste.
The copper blocks were alternately heated by glued tip resistors to generate DT less than 1K. ΔT was measured by an AuFe-chromel thermocouple. Thermopower was estimated from the slope of the generated voltage DV vs. the temperature difference DT. 29 All measurements were conducted along the a-axis.

ESR measurement
ESR measurements were carried out for a single crystal using X-band spectrometer (JEOL JES-TE100).
The g-values were calibrated by the 3 rd (g = 2.034) and 4 th (g = 1.981) signals of a Mn 2+ marker. The crystal was located so as to make the long axis parallel to the magnetic field, and rotated by 15 o steps up to 180 o for the angle dependence measurements ( Figure S5). Although the crystal was chosen carefully, the observed signal was composed of two Lorentzians when the crystal long axis was not parallel to the magnetic field. The signal was analyzed by a sum of two Lorentzians, where H was the magnetic field, I was the intensity of the signal, H 0 was a center of the signal, and DH pp was the linewidth of the signal. Temperature dependence measurements were carried out down to 4.3 K for the single Lorentzian and double Lorentzians.

Computational details
The density functional theory (DFT) calculations were performed using a numerical atomic orbitals DFT 30 approach developed for efficient calculations in large systems and implemented in the SIESTA code. 31 The generalized gradient approximation to DFT and, in particular, the functional of Perdew, Burke, and Ernzerhof (PBE) 32 has been used. Recently 22 we have found that inclusion of a Hubbard correction term 20 U for S 3p orbitals is needed for accurately describing the electronic structure of TTF-TCNQ and related two-chain molecular solids for which accurate experimental information on the bandwidth and charge transfer is available. In the present work, we have considered the dependence of the experimentally observed charge transfer for several sulfur-containing two-chain salts and found that a good agreement is found for U= 6 eV. Thus, except otherwise stated we have used this value in our calculations. The contribution of core electrons has been described by norm-conserving Troullier-Martins relativistic pseudo-potentials, 33 factorized in the Kleinman-Bylander form. 34 Valence electrons were treated explicitly using a split-valence basis set of double-ξ plus polarization functions obtained with an energy shift of 10 meV. 35 The energy cutoff for the real space integration mesh was set to 300 Ry and the Brillouin zone was sampled using a mesh with 9×25×3 k-points obtained using the method of Monkhorst and Pack. 36 All calculations were performed using the experimental crystal structure.