Photoconductive Properties and Electronic Structure in 3,5-Disubstituted 2-(2′-Pyridyl)Pyrroles Coordinated to a Pd(II) Salicylideneiminate Synthon

The synthesis and the electrochemical, photophysical, structural, and photoconductive properties of three new heteroleptic Pd(II) complexes with various 3′,5′- disubstituted-2-(2′-pyridil) pyrroles H(N^N) as coordinated ligands are reported. The coordination of the metal center was completed by a functionalized Schiff base H(O^N) used as an ancillary ligand. The [(N^N)Pd(O^N)] complexes showed highly interesting photoconductive properties which have been correlated to their electronic and molecular structures. Theoretical density functional theory (DFT) and time-dependent DFT calculations were performed, and the results were confronted with the organization in crystalline phase, allowing to point out that the photoconductive properties are mainly a consequence of an efficient intramolecular ligand-to-metal charge transfer, combined to the proximity between the central metal and the donor moieties in the solid-state molecular stacks. The reported results confirm that these new Pd(II) complexes form a novel class of organometallic photoconductors with intrinsic characteristics suitable for molecular semiconductors applications.

. Relative energies of the 1, 2 and 3 cis (Npyrr…O) and trans (Npyrr…O) isomers (kJ/mol) in vacuum and different solvents  Table S2. Decomposition of the ligand-metal energy bond in 1, 2 and 3 cis (Npyrr…O) and trans (Npyrr…O) isomers Table S3. Decomposition of the interaction energy (energy term) associated to Step 3 only. Energies are kJ/mol   Table S4. Absorption data of 1-3 recorded in dichloromethane solution  Table S8. Computed Cartesian coordinates of the 1molecular cluster with ONIOM (singlet and triplet ground states) Table S9. Test on the reliability of PM6 computations: single molecule in vacuo Table S10. Test on the reliability of PM6 computations: 1 couple a-b Table S11. Cartesian coordinates of the T 1 excited states of 1-3 Figure S22. Views of the 1 and 3 infinite stacks in the crystals Table S12. Computed singlet and triplet excited states of 1-3. Table S13. Excited states of 1-3 computed with inclusion of spin-orbit effects.

Photoconductivity measurements details
Photoconductivity measurements were carried out by applying on the sample a DC voltage and measuring the current in the dark and under illumination, as detailed in the following. A DC voltage is applied in the dark and the current flowing through the sample is measured. Once the current becomes stable, the light is turned on and the increase of the current is observed. When the current is again stabilized, the light is turned off. The difference between the value of the current measured in the dark and the one measured with the light on is called photocurrent i ph . From i ph , applying Ohm's, it possible to calculate the photoconductivity  ph : where d is the sample thickness, A is the sample area and V is the applied voltage. The current was measured with a Keithley 6517A electrometer, while the light was provided by a He-Ne laser at 633 nm for (compound 2) or by a solid-state laser at 532 nm (for compound 3) or at 600 nm by using a lamp/monochromator system (for compound 1). The light was turned on/off by a Thorlabs Optical Shutter. Figure SI6 shows typical data of current versus time obtained in a sample of complex 2.
S-3 In Figure S10 is represented one of the most interesting synthon formed through C-H---F interactions in complex 3. Two adjacent molecules are connected via C H---F weak hydrogen bonds with the complementary association of the hydrogen atoms and CF 3 groups of the pyrrolic ring of the coordinated pyridilpyrrole ligands, forming a hydrogen bond ring R 2 2 (10) graph set.

On the relative stability of the 1, 2 and 3 cis (Npyrr … O) and trans (Npyrr … O)
The relative stability of the cis (Npyrr … O) and trans (Npyrr … O) isomers of 1, 2 and 3 was investigated with the aim to explain the different isomery observed in 1 and 3 crystals (trans (Npyrr … O) and cis (Npyrr … O) , respectively). Table S1 lists the relative energies of the two isomers, computed after full structure optimization in vacuum and full optimization in four solvents of different dielectric properties (chloroform, dichloromethane, ethanol and water). Solvation effects were included through the Self Consistent Reaction Field (SCRF) method described in the "Computational Methods" in the paper. These computations will be labeled as G09/SDD09/D95d when performed in vacuum and G09/M06/SDD09/D95d/SOLV in the case of SCRF computations (SOLV= chloroform, dichloromethane, ethanol and water, = 4.73, 8.92, 24.852, 78,3553, respectively). Figure S11 is a graphical view of the relative stabilities of the two isomers in the four studied solvents.  The first point to be discussed is the origin of the higher trans (Npyrr … O) stability observed in vacuum. It is particularly evident in 2, much less in 3, however, it is a constant along the series of the studied compounds (in vacuum). For this goal, the formation process of cis (Npyrr … O) and trans (Npyrr … O) complexes from Pd 2+ and the two anionic ligands was decomposed according to the following sequence of ideal steps, which is detailed below for and written for a generic complex: Step 1 Step 2 Step 3 Step 4 In the first and second steps, the (O^N) and (N^N) fragments, computed as anions, change their structure starting from the relaxed one (in vacuum) toward the structure assumed in the complex. These steps always imply a positive energy variation and are called "preparation energies" so that the deformed ligands are labeled as "prepared". The third step consists in the two prepared ligands approaching one another and reaching the relative position they undertake in the complex. Being the two fragments anionic and closed-shell, a positive energy change is expected also in this case. The fourth step consists in the formation of the metal-ligand bonds. This step includes both charge-transfer between ligands and central metal and also polarization of the electron densities induced by the interactions among fragments. It is normally associated to a negative energy term.  Table S2 collects the change in electronic energies of each step computed at the G09/M06/SDD09/D95d level of approximation.
Step 3 was evaluated, for comparison, also at the higher level of approximation M06/6-311+G(d) in the G09/M06/6-311+G(d) column, that is, using a triple-zeta quality basis set plus one polarization (on C, O, N and F atoms) and one diffused function as provided by the program. The same step 3 was computed using the ADF2019 (Amsterdam Density Functional, 2019.03) [1] software at the M06/ATZP level, that is with a Slater-type basis set of the same quality (the standard "AUG/ATZP" basis set provided by the program, triple-zeta quality with polarization and diffused functions on all the atoms). The numerical accuracy Trans more stable Cis more stable S-13 of the ADF computations was improved to the "good" level ("NUMERICALQUALITY good" keyword). Structures were not optimized in the G09/M06/6-311+G(d) and ADF/M06/ATZP, we used the same structures obtained at the lower G09/SDD09/D95d level. The additional computations for step 3 were performed to rule out computational reliability problems often associated to anions computations, which could be especially present in the case of two interacting anionic fragments as the case of step 3 products. Table S2 allows being confident that the low level computations are sufficiently reliable. A further reason for having performed ADF/ATZP computations is to check its similarity to the G09 computations before a further energy decomposition discussed below in this paragraph. From Table S2, the higher stability of trans (Npyrr … O) isomers is always mainly related to Step 3, that is, to a lower repulsive interactions between chelants in the trans (Npyrr … O) conformation. This contribution (in favor to trans (Npyrr … O) ) is only partially counterbalanced by the better metal-ligand interaction (Step 4) in the case of cis (Npyrr … O) . As often, better metal-ligand interactions imply a more positive (destabilizing) preparation energy, but no equally relevant differences in this contribution are observed between the two isomers in each compound. Table S2. Decomposition of the ligand-metal energy bond according to Steps 1-4 described above. Electronic energy changes are collected (kJ/mol) at the G09/M06/SDD09/D95d for all the steps where not otherwise specified. Step 3 energy change was computed also at the G09/M06/6-311+G(d) and ADF/ATZP levels. Small differences between the total metal-ligand bond energies reported here respect the same value Table S1 (for example, +11.5 kJ/mol versus +11.0 kJ/mol for 1) arises from the different way used for this computation in the two tables. In this table, the total energy is the sum of the Step 1-4 contributions, in Table S1 it is the energy difference between the two complexes. Numerical errors are different in the two computations.

S-14
For further understanding, step 3 was analyzed according to the decomposition scheme of the bonding energy implemented in the ADF program. [1,2] The interaction energy between the two anionic ligands (as they place one another in the complex) is decomposed in three contributions: electrostatic repulsion, Pauli repulsion (repulsion between closed-shell electron densities due to the antisymmetric character of the state function of a system of electrons, roughly associated to the two orbitals-four electrons interactions and commonly called steric repulsion) and polarization of the interacting fragments plus charge transfer between themselves. Regarding the last step, charge transfer between the two anionic ligands is almost inexistent, for obvious reasons, so this term can be considered only polarization of the two ligands due to mostly electrostatic interactions. According to Table S3, the larger repulsion between the two ligands is associated to a larger electrostatic repulsion. The changes of Pauli repulsion and polarization energies (passing from the trans (Npyrr … O) isomer to cis (Npyrr … O) isomer) sum to a relatively small value in all the cases. Our conclusion is that in the cis (Npyrr … O) isomer, the proximity between the formally negative oxygen and nitrogen atoms of the (O^N) and (N^N) ligands leads to a higher repulsion between the chelants. Figure S12 shows the SCF Coulomb potential computed for the two isomers of 1 and plotted on the surface corresponding to a 0.03 bohr -3 electron density. It is possible to observe that the pyrrole cycle bears more negative charge in comparison to the pyridine one. On the (O^N) ligand, the oxygen atom shows a particularly negative charge. As detailed in the paper, cis (Npyrr … O) isomers are favored in condensed phases by their higher dipole moment. Figure S12 allows a direct association between dipole moment and SCF charge distribution in the two isomers. isomers. Geometries were optimized at M06/ECP28MDF:cc-aug-VDZPP/cc-pVDZ/CLF. Chemical shifts were computed at this geometry but with M06, B3PW91 and mPW1PW91 (red, green and violet respectively) xc-functionals and DFT/ECP28MDF:cc-aug-VTZPP/cc-pVTZ/CLF level of theory. isomers. Geometries were optimized at M06/ECP28MDF:cc-aug-VDZPP/cc-pVDZ/CLF. Chemical shifts were computed at this geometry but with M06, B3PW91 and mPW1PW91 (red, green and violet respectively) xc-functionals and ECP28MDF:cc-aug-VTZPP/cc-pVTZ/CLF basis set.
S-17 Figure S14 shows that, in case of 1, the agreement between the computed and experimental chemical shifts are better for the trans isomer than in case of cis. The other way around is true in case of the compound 3 where the agreement ( Figure S15) is better for the cis than for the trans isomer. Both calculations agree with the SCXRD findings validating this approach also in case of compound 2 ( Figure S13) where SCXRD data aren't available as discussed in the paper. From Figure S13 it is evident as in case of 1 that the best agreement is for isomer trans than cis.     S-20 Figure S19. UV-Vis spectra in dichloromethane of 1 (black),2 (red) and 3 (green) and the computed transitions by TD-DFT at the mPW1PW91/SDD09/D95d/DCM level of theory (vertical arrows, same colour convention). (The number of computed states has been extended up to 32500, 36000, 42500 cm -1 for 1, 2 and 3 respectively).            Table S7. Computed Cartesian coordinates at the MPW1PW91/SD09/D95(d)/DCM level of approximation. Neutral and charged compounds were computed and reported here. These structures were used for the evaluation of the oxidation and reduction potentials ( Figure 4 in the paper) as energy difference between charged and neutral compounds (see "Computational Details" for more information).  Figure S20. Emission spectrum of 1 dissolved in a 5:5:2 mixture of cyclohexane, ethanol and 2-methylbutane, at 77 K (a); time-resolved luminescence intensity of 1 at 77 K (b).

Effect of packing on the internal torsion angles
Structure optimizations of 1 ground and T 1 states were repeated after placing the molecule in a cage of five nearest neighbour molecules and modelling the system with the ONIOM approach (see "Computational Details" for more information). Using Figure S22 as reference, the ONIOM computations were performed by modelling the molecule b in the stack at the MPW1PW91/ SDD09/D95d level, molecules a and b along the stack plus three additional molecules surrounding molecule b in the crystal were modelled at the semiempirical PM6 level. Figure S21 (top part) shows the cluster of molecules used in the ONIOM investigation and the way in which molecule b is surrounded.  S-37 Geometry optimization was performed on all the internal degree of freedom, both in the low (PM6) and in the highlevel layers (DFT) of ONIOM, apart from all the Pd-Pd distances, which were frozen to the experimental value for preventing drifts of the molecules which are impossible in solid phase. Figures S21a and S21b show the obtained ground state structure from two different viewpoints (the structure of molecule b, the one computed at high level, referred to as "ONIOM structure" in the following) which is overlapped to the experimental structure (labelled as "crystal" in the figure). The two structures are in turn overlapped to the one computed in vacuo. The ONIOM structure appears very similar to the experimental one, with higher planarity in comparison to the structure computed in vacuo. The average unsigned variation of the torsion angles (dihedral angles among four connected atoms) is 4 degrees, if we exclude from this average the lateral aliphatic chains on the O^N ligand. The larger distortion is confined to one of the O^N phenyl, which points outside the cage of molecules in the ONIOM computations and is consequently expected to be relatively free to distort in comparison to the real crystal.
In the case of the T 1 state, (Figure S21c and S21d), the difference between the in vacuo and crystal structures is more evident. Conversely, the T 1 ONIOM structure is comparable to the crystal one, similarly to what it is observed in the ground state. Also, in this case, the values of torsion angles remain substantially unchanged. We can then conclude that the crystal packing substantially hampers any significant modification of the values of molecular torsion angles both in the ground and in the excited states. The results allow to assume that even in the case of the other compounds packing constraints can be mimicked in a single molecule excited state calculation, freezing the torsional variable while allowing all other variables free to relax. Table S8. Cartesian coordinates (Å) obtained after geometry optimization with ONIOM. Compound 1 is reported. The high level layer was set to MPW1PW91/SDD09/D95(d), as all the DFT computations in this paper, and it was specified in the coordinates below with "H" at the end of the lines. The low level layer was treated at the PM6 levels ("L" label in the end of the line).

ONIOM computations: optimized structures
Singlet: the high level layer was computed as an unrestricted singlet state. The low level layer was computed as a restricted singlet state.
Table S10. The same structural parameters of Table S9 obtained after geometry optimization of couple a-b of 1 in the crystal (see Figure 8a in the paper). The experimental parameters are reported for the molecules in the couple (an inversion centre exchange the two molecules both in the crystal and in our computations).
With the aim to avoid the drifting and tilting of one molecule respect to the other (impossible in solid phase), some structural parameters were frozen during optimization: the Pd-Pd experimental distance was frozen, like all the angles of D1-  In all the cases (as discussed in the paper), all the dihedral angles were frozen during the structure optimization. In 1 and 3, structure optimization was performed starting from the experimental structure, in 2 the computed (in vacuo) structure was used as starting point. The Gaussian09 revision D01 was used within the TD-DFT approach. The obtained structure was used for the characterization of the T 1 excited states (Figure 11 and associated discussions in the "Photoconductivity Studies" paragraph).

S-47
The infinite stacks in 1 and 3 crystals Figure S22a shows the molecular infinite stack of the 1 crystal. The shown A-B couple of molecules highlights how the O^N ligand of one molecule is eclipsed to the N^N ligand of the second molecule. Furthermore, two donor atoms of one complex are close to the analogue ones of the second molecule. The C-D couple of molecules are symmetrically related to the A-B one, resulting from the application of an inversion operation. Molecules B and C face one another through their O^N ligands. The A-D sequence in Figure S22a is repeated and gives rise to an infinite stack, which could be suitable for charge conduction. No other infinite stacks can be found in the crystal organization of 1. Its finding is a good clue in favor of efficient photogeneration and photoconduction. Figure S22b shows the possibly conductive stack in the 3 crystal. A sequence of four molecules is replicated so to produce, in principle, an infinite pattern which could be suitable for photogeneration and/or conduction. All the couples involve a good "facing" of the O^N ligand. Furthermore, couples A-B and C-D show a proximity to the central metals and its surrounding donor atoms. Such a stack is substantially formed by couples of almost planar molecules in close parallel relative disposition, as observed in the case of the 1 crystal.  Table S12. Some computed properties of the lowest singlet and triplet excited states of 1-3. All the structures were optimized in the corresponding excited state and in DCM solutions as detailed in the "Computational Methods" in the paper. The main character of the excited state is labelled as metal-centred (MC), metal-ligand charge transfer (MLCT) or ligand-ligand charge transfer (LLCT). The emission wavelength of each state is reported in nm.
computational methods applied through the two programs (in particular Slater basis sets and ZORA in ADF 2019 in comparison to the used relativistic pseudopotential and contracted Gaussian basis functions in Gaussian 09). More important, no significant mixing is computed between triplet and singlet excited states up to the fourth excited state, in which a 5,9 % contribution (in 1) comes from singlet states. In all the studied complexes, the lowest three excited states result from spin-orbit mixing mostly among pure triplet states components only. This makes the oscillator strengths of these excited states very small. These findings allow to explain the not-emissive character of the lowest (we could say T1) excited state in all the complexes. Furthermore, the fact that no phosphorescent emission is observed also in solid solutions (see the discussions in the paper relative to Figure 7) cannot exclude the presence of relatively long-living triplet states, which could be not emissive because of their very low oscillator strength rather than fast non-radiative deactivation processes. A long lifetime of the lowest triplet state, combined to the unique nature of T1 (a MLCT excited state), allows a direct explanation of the high observed photogeneration in 1.
TABLE S13. Excited states of 1-3 computed with inclusion of spin-orbit effects. The lowest six excited states were computed on the T 1 optimized structure (optimized with Gaussian 09 as detailed in the paper about the computational study of emission in solid solution at 77 K, see also T2 62.0% T1 36.0% S3 1.7% T3 0.1% a Respect to Excitation 1, the structure is relaxed in T 1 without spin-orbit coupling and it is computed with Gaussian 09 (see the paper for details and discussions therein about Figure S12).