Coordination networks assembled from Co(NCS) 2 and 4 ′ -[4-(naphthalen-1-yl)phenyl]-3,2 ′ :6 ′ ,3 ′′ -terpyridine: Role of lattice solvents

The preparation and characterization of 4 ′ -[4-(naphthalen-1-yl)phenyl]-3,2 ′ :6 ′ ,3 ′′ -terpyridine ( 1 ) are described. Reactions of 1 with Co(NCS) 2 under conditions of crystal growth by layering using different solvent combinations produced crystals of [Co( 1 ) 2 (NCS) 2 ] n ⋅ 2 n CHCl 3 and [Co( 1 ) 2 (NCS) 2 ] n ⋅ 2 n C 6 H 5 Me, each of which comprised a (4,4) net. The orientations of 1 with respect to the planar network defined by the Co atoms are significantly different in [Co( 1 ) 2 (NCS) 2 ] n ⋅ 2 n C 6 H 5 Me compared to [Co( 1 ) 2 (NCS) 2 ] n ⋅ 2 n CHCl 3 , and the toluene molecules in [Co ( 1 ) 2 (NCS) 2 ] n ⋅ 2 n C 6 H 5 Me are involved in π -stacking interactions. The solvent-accessible void-space in the latter consists of a series of interlinked cavities in contrast to the open channels in [Co( 1 ) 2 (NCS) 2 ] n ⋅ 2 n CHCl 3 . Thermogravimetric analysis was used to investigate solvent loss and uptake in the two coordination networks. After solvent loss from [Co( 1 ) 2 (NCS) 2 ] n ⋅ 2 n CHCl 3 , CHCl 3 , CDCl 3 or CH 2 Cl 2 could be taken up by the lattice. In contrast, removal of toluene from [Co( 1 ) 2 (NCS) 2 ] n ⋅ 2 n C 6 H 5 Me was found to be irreversible.

We have previously carried out systematic investigations of the effects of different 4 ′ -substituents R on the assembly of [Co(NCS) 2 (4 ′ -R-3,2 ′ :6 ′ ,3 ′′ -tpy) 2 ] n 2-dimensional nets [10,15,16], and have demonstrated a change between conformations A and B (Scheme 1) as a consequence of changing the 4 ′ -substituent. Motivated by these results and the observation that many of the 2D structures contain lattice solvent molecules, we decided to focus on a single 3,2 ′ :6 ′ ,3 ′′ -tpy ligand (1, Scheme 2) and explore the effects of using different pairs of solvents in crystal growth experiments. Here we report the coordination networks assembled when MeOH or MeCN solutions of Co(NCS) 2 were layered over toluene or CHCl 3 solutions of 1. Ligand 1 was selected because of the potential for π-stacking interactions involving the extended aromatic system [24,25] and the possibility of using the supramolecular effects to further control and refine the structure.

General
1 H, 13 C{ 1 H} and 2D NMR spectra were recorded on a Bruker Avance III-500 spectrometer equipped with a BBFO probehead at 298 K. The 1 H and 13 C NMR chemical shifts were referenced with respect to residual solvent peaks (δ TMS = 0). A Shimadzu LCMS-2020 instrument was used to record electrospray ionization (ESI) mass spectra, and FT-infrared (IR) spectra were recorded on a PerkinElmer UATR Two instrument. A Shimadzu UV2600 spectrophotometer was used to record solution absorption spectra.
Thermogravimetric analysis (TGA) was performed on a TGA5500 instrument (TA Instruments) coupled to a Discovery II MS, Cirrus 3, Mass Spectrometer, DMS. The analysis was carried under nitrogen, using a Barchart scanning method in the mass range 10-125. In all the experiments, the temperature of the TGA instrument was initially stabilized at 30 • C. The samples were then heated to the appropriate temperature, depending upon the solvent included in the lattice and this temperature was maintained for 10-30 min. During this time, it was possible to detect the solvent being released from the coordination network (see Section 3.4 for details) and solvents were identified through mass spectrometry. Afterwards the samples were cooled to ambient temperature and put in contact with vapors of the same or a different solvent for 24-72 h, After this, the TGA was repeated using the same procedure. A set of measurements was performed on each coordination network.
All crystal growth experiments were carried out under ambient conditions using identical glass crystallization tubes (i.d. = 13.6 mm, 24 mL).

[Co(1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me
A solution of Co(NCS) 2 (5.3 mg, 0.030 mmol) in MeOH (6 mL) was layered over a toluene solution (6 mL) of 1 (13.1 mg, 0.030 mmol). Pink plate-like crystals grew after 7 days. A single crystal was selected for Xray diffraction and the remaining crystals were washed with MeOH and toluene, dried under vacuum and analysed by PXRD and FT-IR spectroscopy.

Crystallography
Single crystal data were collected on a Bruker APEX-II diffractometer (CuKα radiation) with data reduction, solution and refinement using the programs APEX [26], ShelXT [27], Olex2 [28] and ShelXL v. 2014/7 [29]. All H atoms were included at geometrically calculated positions and refined using a riding model with U iso = 1.2 of the parent atom. Structure analysis and structural diagrams used CSD Mercury 2020.1 [30]. In [Co(1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me, each toluene solvent molecule was disordered over a symmetry element, with site occupancies of 50%; the toluene molecules were refined isotropically. In [Co (1) 2 (NCS) 2 ] n ⋅2nCHCl 3 , the solvent area was treated using a solvent mask, and the electron density removed corresponded to 2 molecules of CHCl 3 per Co atom; this was added to the formulae and appropriate numbers.

Scheme 2.
Structure of compound 1 and numbering for NMR spectroscopic assignments.
using a Stoe Stadi P diffractometer with Cu Kα1 radiation (Ge(111) monochromator) and a DECTRIS MYTHEN 1 K detector. Whole-pattern decomposition (profile matching) analysis [31][32][33] of the diffraction patterns was performed with the package FULLPROF SUITE [33,34] (v. September 2020) using a previously determined instrument resolution function based on a NIST640d standard. The structural models were taken from the single crystal X-Ray diffraction refinements. Refined parameters in Rietveld were: scale factor, zero shift, lattice parameters, Co and S atomic positions, background points and peaks shapes as a Thompson-Cox-Hastings pseudo-Voigt function. Preferred orientations were included in the analysis as a March-Dollase multi-axial phenomenological model. (1) (1)

Ligand synthesis and characterization
Compound 1 was prepared by the one-pot strategy of Wang and Hanan [35] by reaction of two equivalents of 3-acetylpyridine with 4-(naphthalen-1-yl)benzaldehyde under basic conditions and addition of aqueous NH 3 . The ligand was isolated in 42.0% yield as a colorless solid, and the base peak in the ESI mass spectrum corresponded to the [M+H] + ion (m/z 436.17, Fig. S1). The 1 H and 13 C{ 1 H} NMR spectra (Figs. S2 and S3, respectively) were assigned using NOESY, COSY, HMQC (Fig. S4) and HMBC (Fig. S5) spectra, and were in accord with the structure shown in Scheme 2. Overlap of the signals for the naphthyl protons made unambiguous assignment of these signals difficult. The IR spectrum is shown in Fig. S6, with diagnostic, strong bands in the fingerprint region at 796, 773 and 695 cm − 1 . Fig. 1 displays the solution absorption spectrum of 1, and the absorption bands are assigned to π*←π transitions. The spectrum closely resembles that of 4 ′ -[4-(naphthalen-1-yl) phenyl]-4,2 ′ :6 ′ ,4 ′′ -terpyridine [36] which is an isomer of 1.

Coordination networks [Co(1) 2 (NCS) 2 ] n ⋅2nCHCl 3 and [Co (1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me
Single crystals of [Co(1) 2 (NCS) 2 ] n ⋅2nCHCl 3 and [Co (1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me were grown under ambient conditions by layering either a MeCN solution of Co(NCS) 2 over a CHCl 3 solution of 1, or a MeOH solution of Co(NCS) 2 over a toluene solution of 1, respectively. X-ray quality crystals were selected and the remaining crystals were used for PXRD analysis (see Section 3.3). The two compounds crystallize in the monoclinic space groups P2 1 /n and P2 1 /c, respectively, and each possesses a 2D-network with the Co(II) centers acting as 4-connecting nodes. The structures of the asymmetric units with symmetrygenerated atoms are shown in Fig. 2. The Co atom in each structure is six-coordinate and lies on an inversion center, being bonded to four different, but crystallographically equivalent, ligands 1. Selected bond lengths and angles in each cobalt(II) coordination sphere are given in Table 1 and are unexceptional.
In [Co(1) 2 (NCS) 2 ] n ⋅2nCHCl 3 , 1 adopts conformation A in Scheme 1, whereas in [Co(1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me, conformation B (Scheme 1) is observed. This results in distinct differences between the 2D-assemblies in [Co(1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me and [Co(1) 2 (NCS) 2 ] n ⋅2nCHCl 3 . This is quantified by a comparison of the angles between the planes of adjacent aromatic rings in the ligand in each structure presented in Table 2. When 1 adopts conformation A (Scheme 1), the vectorial properties of each N1 and N3 in each ligand are equivalent. In contrast, the different vectorial properties of the N-donor lone pairs when ligand 1 is in conformation B lead to three possible coordination modes for a trans-arrangement of ligands. In a previous publication, we defined the labels in and out to describe the orientation of each N lone pair with respect to the central pyridine ring (Scheme 3) [13]. Since the Co atom in [Co (1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me lies on a center of symmetry, only two of the three coordination modes shown in Scheme 3 are permitted. Since each Co in [Co(1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me is crystallographically equivalent, both trans out/out and in/in arrangements are present at each metal center (Fig. 3a). This contrasts with the arrangement in [Co (1) 2 (NCS) 2 ] n ⋅2nCHCl 3 in which all motifs are out/out (Fig. 3b), leading to a significant difference between the {Co(1) 4 } units in the networks (Fig. 3a versus 3b).
Part of the (4,4) network in [Co(1) 2 (NCS) 2 ] n ⋅2nCHCl 3 is shown in Fig. 4a. The Co atoms are coplanar, and the 4-(naphthalen-1-yl)phenyl substituents protrude above and below the plane defined by the Co atoms. Around each rhombus in the net, the 4-(naphthalen-1-yl)phenyl groups are arranged in an up/up/down/down configuration (Fig. S7), and adjacent 2D-sheets are locked together by face-to-face π-stacking interactions between naphthalenyl rings in one layer with the central pyridine ring of the 3,2 ′ :6 ′ ,3 ′′ -tpy units in the neighboring sheet ( Fig. 4c). Metrics for the π-π interaction are centroid…centroid = 3.60 Å, and angle between the ring planes = 3.1 • . The up/up/down/down configuration of the arene substituents leads to extended π-stacking interactions which interconnect all adjacent sheets in the lattice. Viewing the structure of [Co(1) 2 (NCS) 2 ] n ⋅2nCHCl 3 down the a-axis without solvent molecules present reveals approximately linear channels (Fig. 4d) and the solvent accessible void space in the lattice is ca. 25% of the total volume.
In [Co(1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me, each toluene molecule is disordered over a symmetry element, with 50% site occupancies of head-to-tail positions. For the discussion below, one position for each toluene molecule is considered. Two views of the (4,4) net in [Co(1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me are depicted in Fig. 5a and 5b. As in [Co(1) 2 (NCS) 2 ] n ⋅2nCHCl 3 (Fig. 4), the Co atoms in each sheet are coplanar, and the terpyridines adopt an up/up/ down/down arrangement around each rhombus (Fig. 5c). Inspection of Fig. 5a and 5c reveals that ligands 1 lie over the rhombuses in the net in  (Figs. 4a and S7). This impacts on the solvent accessible voids in the lattice (Fig. 5d and S8) which comprise a series of interlinked cavities (ca. 24% of the total lattice volume) in contrast to the open channels in [Co(1) 2 (NCS) 2 ] n ⋅2nCHCl 3 . The toluene molecules occupying the cavities are involved in a face-to-face π-interaction with the pyridine ring containing N2, and C-H py …π toluene contacts involving the pyridine rings with N2 and N3 (Fig. S10). For the face-to-face contact, the distance between the ring centroids is 3.8 Å and the angle between the ring planes is 9.8 • . For the C-H py …π toluene contacts, the H…centroid distances are 3.2 and 3.4 Å, and the C-H…centroid angles are 151 and 168 • , respectively.
Ellipsoids are plotted at a 40% probability level, and H atoms and solvent molecules are omitted.

Analysis of bulk samples of the single crystals
In order to verify that the crystals selected for single-crystal X-ray diffraction were representative of the bulk crystalline materials, we carried out PXRD on the remaining crystals in each crystallization tube (see Sections 2.3 and 2.4). Fig. 6 shows comparisons of the experimental PXRD patterns (in red in Fig. 6) and the patterns predicted from the single crystal structures (in black in Fig. 6). For each compound, there was a match for every peak in the predicted pattern with those in the experimental PXRD pattern. No additional peaks were observed in the PXRD of the bulk materials.

Thermogravimetric analysis: Solvent removal and re-entry
Crystals of [Co(1) 2 (NCS) 2 ] n ⋅2nCHCl 3 were heated to 80 • C in the TGA instrument, and this temperature was maintained for 10 min. Loss of CHCl 3 was detected (Fig. 7) with mass peaks at m/z 83.0, 85.0 and 87.0 corresponding to CHCl 2 + as the dominant fragment [37]. After cooling, the sample retained crystallinity. The weight loss of 17.1% corresponded to the loss of approximately 2 CHCl 3 molecules per formula unit (calculated 18.6% from the molecular formula). In order to investigate whether the coordination network maintained its integrity, the same sample was placed in contact with CHCl 3 vapor for 48 h and was again analysed using TGA (Fig. S12). Loss of CHCl 3 was again detected confirming that, after the initial removal of solvent, the coordination assembly was able to re-absorb CHCl 3 . A third cycle was performed with the same crystals exposed to CDCl 3 vapor for 24 h and then analysed by TGA (Fig. S13). Loss of CDCl 3 was confirmed by the presence of mass peaks at m/z 84.0, 86.0 and 88.0 corresponding to the CDCl 2 + ion. In a fourth cycle, the crystalline material was recovered from the preceding TGA experiment and exposed to CH 2 Cl 2 vapor for 24 h. The TGA trace of the resultant crystal showed loss of CH 2 Cl 2 (m/z 49.0, 51.0, 84.0 and 86.0 arising from CH 2 Cl + and CH 2 Cl 2 + ) at around 40 • C (Fig. S14). Finally, an analogous experiment was attempted using CCl 4 but the TGA of the crystals after exposure to CCl 4 vapor for 24 h revealed no change in mass over a period of 10 min heating at 80 • C. This is consistent with no uptake of CCl 4 into the crystal lattice.
[Co(1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me was also analysed by TGA. It was first heated to 150 • C and this temperature was maintained for 30 min. The loss of toluene was detected with peaks in the mass spectrum of the TGA-MS at m/z 92.0, 91.0 (highest intensity peak), 65.0 and 39.0 (Fig. 8). The weight loss of 14.4% (Fig. 8) corresponded to complete loss of two molecules of toluene from the lattice (calculated 15.0%). The same sample was then placed in contact with toluene vapor for 24 h, and the subsequent TGA analysis showed no appreciable mass loss, consistent with the fact that once toluene had been removed from [Co (1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me, it was unable to be re-absorbed. This is consistent with the presence of the aromatic solvent in [Co (1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me playing a structural role in contrast to the CHCl 3 in [Co(1) 2 (NCS) 2 ] n ⋅2nCHCl 3 occupying channels in a nonordered manner. The π-stacking interactions are key to the former, and we suggest that these may template the assembly of the network in [Co(1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me and that there is a partial loss in crystallinity (confirmed with PXRD) once the solvent is removed.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests that could have appeared to influence the work reported in this paper. Fig. 6. PXRD (CuKα1 radiation) patterns (red circles) of the bulk crystalline materials of (a) [Co(1) 2 (NCS) 2 ] n ⋅2nCHCl 3 , and (b) [Co(1) 2 (NCS) 2 ] n ⋅2nC 6 H 5 Me, with fitting to the predicted patterns from the single-crystal structures. The black lines are the best fits from the Rietveld refinements, and green lines show the Bragg peak positions. Each blue plot gives the difference between calculated and experimental points, and the differences in intensities arise from differences in the preferred orientations of the crystallites in the bulk samples.