Intramolecular Spin State Locking in Iron(II) 2,6-Di(pyrazol-3-yl)pyridine Complexes by Phenyl Groups: An Experimental Study

: Here we report a series of 1-phenyl-5-substituted 2,6-di(pyrazol-3-yl)pyridine complexes with iron(II) ion found in a high spin state in solids (according to magnetochemistry) and in solution (according to NMR spectroscopy), providing experimental evidence for it being an intramolecular effect induced by the phenyl groups. According to X-ray diffraction, the high spin locking of the metal ion is a result of its highly distorted coordination environment (with a very low ‘twist’ angle atypical of 2,6-di(pyrazol-3-yl)pyridine complexes), which remains this way in complexes with different substituents and counterions, in a diamagnetic zinc(II) analogue and in their solutions. Three possible reasons behind it, including additional coordination with the phenyl group, energy penalty incurred by its rotation or intramolecular stacking interactions, are addressed experimentally.


Introduction
The ability of some transition metal complexes to exist in two spin states and to switch between them under an applied stimulus (temperature, pressure or light irradiation) [1,2] is behind many concepts of high-density data storage, switching, sensing and other molecular devices and materials [3][4][5]. Of them, iron(II) complexes with six coordinating nitrogen atoms are often used [1], as the transition occurs between differently colored diamagnetic low-spin (LS, S = 0) and paramagnetic high-spin (HS, S = 2) states and is accompanied by large structural changes that promote its cooperativity essential for the above applications [6] Isomeric 2,6-di(pyrazol-x-yl)pyridines (x = 1 or 3, 1-bpp or 3-bpp) [7,8] are among the most established ligands for this purpose, owing to their synthetic versatility [7,8] and possibility [9] to control the spin state of the metal ion by a judicious choice of substituents.
To get an experimental insight into this HS locking (which may be due to the HS state being thermodynamically more stable than the LS state, as suggested in a computational study for an iron(II) complex 1,3-bpp [17]), we reproduced the ligand L1 (the resulting complex Fe(L1)2OTf2 was obtained as another crystallosolvate [23] but with no significant changes in [Fe(L1)2] 2+ structure) and synthesized a new one (L2) with acetoxy groups instead of OH groups (Scheme 2). The two groups have different electron donating/withdrawal ability (as gauged by their Hammett constants [9] and calculated charges [24] of coordinating nitrogen atoms in the free ligands L1 and L2) and therefore exert different electronic effects on the HS/LS stabilization. For iron(II) complexes of 1-bpp, a more electron-withdrawing group at this position of the ligand favors the HS state (mostly through inductive effects [9]). For complexes of 3-bpp, it seems the opposite [12], although not as clear [15] mostly due to them being not as extensively studied as those of 1-bpp [12]. In addition, the bulkier acetoxy group in L2 has no H-donor ability of the hydroxy group, although it can still act as an acceptor of hydrogen bonds, which together with stacking interactions are important for spin state behaviour of transition metal complexes [1], especially with 3-bpp ligands [15]. The new iron(II) complex [Fe(L2)2] 2+ has been synthesized as two salts with triflate and tetrafluoroborate anions (Fe(L2)2OTf2 and Fe(L2)2(BF4)2), introducing different perturbations into its crystal environment (similar to different lattice solvents in Fe(L1)2OTf2) that should only slightly affect the HS locking and molecular features of [Fe(L2)2] 2+ if they are not induced by intermolecular interactions. Complexes [Fe (3-bpp) 2 ] 2+ , which are much less understood [12] than [Fe(1-bpp) 2 ] 2+ owing to fewer functionalized 3-bpp ligands reported to date [12], are less prone to distortion [22] (typical θ values > 70 • ) [14] but many of them are still HS only [16,22]. Although uncommon for 3-bpp, a recent iron(II) complex Fe(L1) 2 OTf 2 (L1 = 2,6-di(1-phenyl-5-hydroxy-1H-pyrazol-3-yl)pyridine, Scheme 1), a side-product to Fe(L1)OTf 2 for reductive activation of small molecules [23], has a very low value θ of 64.6-66.5 • at 100 K according to X-ray diffraction and thus adopts only the HS state, as spin transitions in bpp complexes generally occur at much higher temperatures [7,8].
To get an experimental insight into this HS locking (which may be due to the HS state being thermodynamically more stable than the LS state, as suggested in a computational study for an iron(II) complex 1,3-bpp [17]), we reproduced the ligand L1 (the resulting complex Fe(L1) 2 OTf 2 was obtained as another crystallosolvate [23] but with no significant changes in [Fe(L1) 2 ] 2+ structure) and synthesized a new one (L2) with acetoxy groups instead of OH groups (Scheme 2). The two groups have different electron donating/withdrawal ability (as gauged by their Hammett constants [9] and calculated charges [24] of coordinating nitrogen atoms in the free ligands L1 and L2) and therefore exert different electronic effects on the HS/LS stabilization. For iron(II) complexes of 1-bpp, a more electron-withdrawing group at this position of the ligand favors the HS state (mostly through inductive effects [9]). For complexes of 3-bpp, it seems the opposite [12], although not as clear [15] mostly due to them being not as extensively studied as those of 1-bpp [12]. In addition, the bulkier acetoxy group in L2 has no H-donor ability of the hydroxy group, although it can still act as an acceptor of hydrogen bonds, which together with stacking interactions are important for spin state behaviour of transition metal complexes [1], especially with 3-bpp ligands [15]. The new iron(II) complex [Fe(L2) 2 ] 2+ has been synthesized as two salts with triflate and tetrafluoroborate anions (Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 ), introducing different perturbations into its crystal environment (similar to different lattice solvents in Fe(L1) 2  The best experimental approach to decouple the spin state from intermolecular effects (and thus to allow for a truly molecular design [15] of spin transition compounds) is to probe it in a solution. Variable-temperature NMR spectroscopy is an easy access for solution studies similar to UV-vis and vibrational spectroscopies, but unlike them, it also allows measuring magnetic susceptibility via the Evans technique [25,26]. As its recent alternative (among others [20,27,28]), an approach [29] to the analysis of chemical shifts in NMR spectra, which is assisted by quantum chemistry calculations of spin density and molecular geometry in different spin states (see Materials and Methods), has been proposed for following spin transitions in solutions of paramagnetic compounds with unknown concentrations or admixtures and even obtained in situ (all are well-known limitations of the Evans method) and successfully tested on a series of cobalt(II) complexes of terpy and its derivatives [29]. For iron(II) complexes, such as Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2, it measures the population of the paramagnetic HS state over temperature and provides insight into its overall molecular geometry.
In addition, we synthesized a zinc(II) complex Zn(L1)2(ClO4)2 to check how an introduction of a non-spin transition metal ion (that shows no Jahn-Teller effect) would affect the distortion of the complex [M(L1)2] 2+ and to identify if it is caused by the ligand design. Diamagnetic analogues of spin transition compounds are not usually structurally analysed [30,31] (e.g., one zinc(II) complex out of a hundred of Fe(II) complexes with 1-or 3-bpp ligands available in CSD, as of 2018), even though they are often used for metal dilution in magnetic measurements and are known to influence the spin state behaviour [32]. They can also be useful as a diamagnetic analogues of the paramagnetic compounds, such as cobalt(II) or HS iron(II) complexes, in an analysis of the NMR data by the recently proposed approach [29].
Here we report how all these modifications to the ligand (L1 or L2), counterion (OTf, ClO4 or BF4), metal (Fe or Zn) and intermolecular interactions influence the HS state of the complexes [FeL2]X2 and a distorted molecular geometry of [ML2] 2+ associated with it and discuss possible reasons behind it by magnetochemistry, X-ray diffraction and multi-temperature NMR spectroscopy.

Spin State from Magnetometry
Spin state of the iron(II) ion in the complexes Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2 has been assessed by variable-temperature dc magnetic susceptibility measurements of their fine-crystalline samples at 3-300 K (Figure 1). For all of them, χT values are constant at 3.7 cm 3 mol −1 K between 50 and 300 K, suggesting them to be in the HS state over this entire temperature range. A lowtemperature decrease in χT and the values χT themselves differing from the theoretical pure HS value for S = 2 (3.00 cm 3 mol −1 K) result from the zero-field splitting. Modelling these data with PHI software [33] afforded the isotropic value of g-tensor giso and the zero-field splitting D equal to 2.18-2.20 and 10.9-13.6 cm −1 ; the latter being not too low indicates a minor contribution from a 'rotational' distortion, as suggested for the iron(II) complexes of 1-bpp [34]. Scheme 2. Ligands in this study: L1, 2,6-di(1-phenyl-5-hydroxy-1H-pyrazol-3-yl)pyridine and L2, 2,6-di(1-phenyl-5-acetoxy-1H-pyrazol-3-yl)pyridine.
The best experimental approach to decouple the spin state from intermolecular effects (and thus to allow for a truly molecular design [15] of spin transition compounds) is to probe it in a solution. Variable-temperature NMR spectroscopy is an easy access for solution studies similar to UV-vis and vibrational spectroscopies, but unlike them, it also allows measuring magnetic susceptibility via the Evans technique [25,26]. As its recent alternative (among others [20,27,28]), an approach [29] to the analysis of chemical shifts in NMR spectra, which is assisted by quantum chemistry calculations of spin density and molecular geometry in different spin states (see Materials and Methods), has been proposed for following spin transitions in solutions of paramagnetic compounds with unknown concentrations or admixtures and even obtained in situ (all are well-known limitations of the Evans method) and successfully tested on a series of cobalt(II) complexes of terpy and its derivatives [29]. For iron(II) complexes, such as Fe(L1) 2 OTf 2 , Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 , it measures the population of the paramagnetic HS state over temperature and provides insight into its overall molecular geometry.
In addition, we synthesized a zinc(II) complex Zn(L1) 2 (ClO 4 ) 2 to check how an introduction of a non-spin transition metal ion (that shows no Jahn-Teller effect) would affect the distortion of the complex [M(L1) 2 ] 2+ and to identify if it is caused by the ligand design. Diamagnetic analogues of spin transition compounds are not usually structurally analysed [30,31] (e.g., one zinc(II) complex out of a hundred of Fe(II) complexes with 1-or 3-bpp ligands available in CSD, as of 2018), even though they are often used for metal dilution in magnetic measurements and are known to influence the spin state behaviour [32]. They can also be useful as a diamagnetic analogues of the paramagnetic compounds, such as cobalt(II) or HS iron(II) complexes, in an analysis of the NMR data by the recently proposed approach [29].
Here we report how all these modifications to the ligand (L1 or L2), counterion (OTf, ClO 4 or BF 4 ), metal (Fe or Zn) and intermolecular interactions influence the HS state of the complexes [FeL 2 ]X 2 and a distorted molecular geometry of [ML 2 ] 2+ associated with it and discuss possible reasons behind it by magnetochemistry, X-ray diffraction and multi-temperature NMR spectroscopy.

Spin State from Magnetometry
Spin state of the iron(II) ion in the complexes Fe(L1) 2 OTf 2 , Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 has been assessed by variable-temperature dc magnetic susceptibility measurements of their fine-crystalline samples at 3-300 K ( Figure 1). For all of them, χT values are constant at 3.7 cm 3 mol −1 K between 50 and 300 K, suggesting them to be in the HS state over this entire temperature range. A low-temperature decrease in χT and the values χT themselves differing from the theoretical pure HS value for S = 2 (3.00 cm 3 mol −1 K) result from the zero-field splitting. Modelling these data with PHI software [33] afforded the isotropic value of g-tensor g iso and the zero-field splitting D equal to 2.18-2.20 and 10.9-13.6 cm −1 ; the latter being not too low indicates a minor contribution from a 'rotational' distortion, as suggested for the iron(II) complexes of 1-bpp [34].  Figure S1. The red line represents the fit.

Spin State from NMR
To exclude solid-state effects behind the observed HS locking, variable-temperature NMR experiments have also been performed for Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2. The 1 H NMR spectra collected from their acetonitrile solutions at 240-330 K show sets of paramagnetically shifted signals that confirm the composition of the obtained iron(II) complexes and their HS state. In addition, these chemical shifts closely follow the linear dependence on temperature as δ ~ T −1 ( Figure  2) expected at high T (>200 K) for systems with a g value close to 2 and a moderate zero-field splitting [35]. As there are no deviations stemming from a spin transition [29] (their other source is large magnetic anisotropy [36][37][38]), it does not occur between 240 and 330 K. This is quantified by the Evans method [25,26], which measures magnetic susceptibility based on a difference in chemical shifts of an inert substance, e.g., TMS, in a solution of a paramagnetic compound and in a pure solvent recorded simultaneously. Within the experimental error, which is fairly large due to the limitations of the method [29], the resulting χT value is constant with temperature at 3.5-4.0 cm 3 mol −1 K ( Figure  1), thereby corroborating the HS locking of the complexes Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2.   Figure S1. The red line represents the fit.

Spin State from NMR
To exclude solid-state effects behind the observed HS locking, variable-temperature NMR experiments have also been performed for Fe(L1) 2 OTf 2 , Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 . The 1 H NMR spectra collected from their acetonitrile solutions at 240-330 K show sets of paramagnetically shifted signals that confirm the composition of the obtained iron(II) complexes and their HS state. In addition, these chemical shifts closely follow the linear dependence on temperature as δ~T −1 (Figure 2) expected at high T (>200 K) for systems with a g value close to 2 and a moderate zero-field splitting [35]. As there are no deviations stemming from a spin transition [29] (their other source is large magnetic anisotropy [36][37][38]), it does not occur between 240 and 330 K. This is quantified by the Evans method [25,26], which measures magnetic susceptibility based on a difference in chemical shifts of an inert substance, e.g., TMS, in a solution of a paramagnetic compound and in a pure solvent recorded simultaneously. Within the experimental error, which is fairly large due to the limitations of the method [29], the resulting χT value is constant with temperature at 3.5-4.0 cm 3 mol −1 K (Figure 1

Spin State from NMR
To exclude solid-state effects behind the observed HS locking, variable-temperature NMR experiments have also been performed for Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2. The 1 H NMR spectra collected from their acetonitrile solutions at 240-330 K show sets of paramagnetically shifted signals that confirm the composition of the obtained iron(II) complexes and their HS state. In addition, these chemical shifts closely follow the linear dependence on temperature as δ ~ T −1 ( Figure  2) expected at high T (>200 K) for systems with a g value close to 2 and a moderate zero-field splitting [35]. As there are no deviations stemming from a spin transition [29] (their other source is large magnetic anisotropy [36][37][38]), it does not occur between 240 and 330 K. This is quantified by the Evans method [25,26], which measures magnetic susceptibility based on a difference in chemical shifts of an inert substance, e.g., TMS, in a solution of a paramagnetic compound and in a pure solvent recorded simultaneously. Within the experimental error, which is fairly large due to the limitations of the method [29], the resulting χT value is constant with temperature at 3.5-4.0 cm 3 mol −1 K ( Figure  1), thereby corroborating the HS locking of the complexes Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2.  Analysis of chemical shifts in the NMR spectra of paramagnetic compounds is in the core of another approach [29] that evaluates their spin state population without isolation or purification required for the Evans method. For iron(II) complexes in the paramagnetic HS state, chemical shifts in the NMR spectra at the desired temperature (such as down to 240 K for acetonitrile) are treated as a sum of diamagnetic (δ dia ), contact (δ c ) and pseudocontact (δ pc ) contributions. The values δ dia are directly measured by NMR spectroscopy of a diamagnetic analogue (such as a zinc(II) complex Zn(L1) 2 (ClO 4 ) 2 or a free ligand L2) at the same temperature, the contact contributions δ c are accessed by simple DFT calculations of the paramagnetic complex and the pseudocontact contributions δ pc are obtained in a fit of the total chemical shifts δ to those measured experimentally by varying the value of magnetic susceptibility tensor anisotropy ∆χ ax (see Materials and Methods). If combined with NMR spectroscopy on heating/cooling, it provides the population of the HS state over temperature [29]. A good agreement ( Figure 3) between the chemical shifts in the 1 H NMR spectra of Fe(L1) 2 OTf 2 , Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 and those estimated by this approach with molecular geometries of [ML 2 ] 2+ optimized at PBE0/def2-TZVP level of theory or taken from X-ray diffraction data (see below), together with acceptable ∆χ ax values [39] resulting from the fit, further confirms their HS state in a solution down to 240 K. All these NMR-based techniques combined show unequivocally that the HS locking in Fe(L1) 2 OTf 2 , Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 is an intramolecular effect rather than a consequence of crystal packing [12,13]. Analysis of chemical shifts in the NMR spectra of paramagnetic compounds is in the core of another approach [29] that evaluates their spin state population without isolation or purification required for the Evans method. For iron(II) complexes in the paramagnetic HS state, chemical shifts in the NMR spectra at the desired temperature (such as down to 240 K for acetonitrile) are treated as a sum of diamagnetic (δdia), contact (δc) and pseudocontact (δpc) contributions. The values δdia are directly measured by NMR spectroscopy of a diamagnetic analogue (such as a zinc(II) complex Zn(L1)2(ClO4)2 or a free ligand L2) at the same temperature, the contact contributions δc are accessed by simple DFT calculations of the paramagnetic complex and the pseudocontact contributions δpc are obtained in a fit of the total chemical shifts δ to those measured experimentally by varying the value of magnetic susceptibility tensor anisotropy Δχax (see Materials and Methods). If combined with NMR spectroscopy on heating/cooling, it provides the population of the HS state over temperature [29]. A good agreement ( Figure 3) between the chemical shifts in the 1 H NMR spectra of Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2 and those estimated by this approach with molecular geometries of [ML2] 2+ optimized at PBE0/def2-TZVP level of theory or taken from X-ray diffraction data (see below), together with acceptable Δχ values [39] resulting from the fit, further confirms their HS state in a solution down to 240 K. All these NMR-based techniques combined show unequivocally that the HS locking in Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2 is an intramolecular effect rather than a consequence of crystal packing [12,13].

Spin State and Molecular Geometry from X-Ray Diffraction
It also follows from X-ray diffraction data collected for all the three iron(II) complexes at 120 K (upon going to this temperature, crystals retained their orange color typical for HS iron(II) complexes [1] and bond lengths Fe-N typical for HS iron(II) complexes of bpp [12]) and of their diamagnetic analogue Zn(L1)2(ClO4)2. Single crystals of a sufficient quality have been obtained with one symmetry-independent [ML2] 2+ species (Figure 4

Spin State and Molecular Geometry from X-Ray Diffraction
It also follows from X-ray diffraction data collected for all the three iron(II) complexes at 120 K (upon going to this temperature, crystals retained their orange color typical for HS iron(II) complexes [1] and bond lengths Fe-N typical for HS iron(II) complexes of bpp [12]) and of their diamagnetic analogue Zn(L1) 2 (ClO 4 ) 2 . Single crystals of a sufficient quality have been obtained with one symmetry-independent [ML 2 ] 2+ species (Figure 4), two counterions and different lattice solvents: diethyl ether in Fe(L1) 2 OTf 2 (instead of acetonitrile in its earlier reported solvatomorph [23]), water and dichloromethane in Fe(L2) 2   Despite different crystal packing effects on [ML2] 2+ , nature of the ligands and even of the metal ions, all four complexes have a similarly distorted coordination geometry (Table 1). This distortion can be described by easily accessible angles ϕ and θ, the measures of 'rotation' (ϕ < 180°) and 'twist' (θ < 90°) component [10] that are independent from each other. For iron(II) complexes of 1-bpp, they are rather good indicators of a spin transition to the LS state [8], which in general [13] does not occur if ϕ < 172° and/or θ < 76°. In Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2, the θ values are all below 70° (66.7-69.8° with the value for the diamagnetic zinc(II) complex in between). Given that this is very low for the iron(II) complexes with 3-bpp ligands [14], they should almost [13] certainly be locked in Despite different crystal packing effects on [ML 2 ] 2+ , nature of the ligands and even of the metal ions, all four complexes have a similarly distorted coordination geometry (Table 1). This distortion can be described by easily accessible angles ϕ and θ, the measures of 'rotation' (ϕ < 180 • ) and 'twist' (θ < 90 • ) component [10] that are independent from each other. For iron(II) complexes of 1-bpp, they are rather good indicators of a spin transition to the LS state [8], which in general [13] does not occur if ϕ < 172 • and/or θ < 76 • . In Fe(L1) 2 OTf 2 , Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 , the θ values are all below 70 • (66.7-69.8 • with the value for the diamagnetic zinc(II) complex in between). Given that this is very low for the iron(II) complexes with 3-bpp ligands [14], they should almost [13] certainly be locked in the HS state, as evidenced by the variable-temperature magnetochemistry and NMR measurements.
Their ϕ values are, on the other hand, close to 180 • , so a major contribution to the distortion comes from the 'twist' component, which is consistent with the zero-field splitting parameters [34].

Molecular Geometry from NMR
Such a distorted molecular geometry even remains in solutions, as supported by the above NMR-based approach [29] (Figure 3 and Figures S3 and S4) is a good indication that these iron(II) complexes are similarly distorted in solution.

Distortion of Molecular Geometry from Symmetry Measures
A better descriptor of this distortion (and another good indicator of the spin state [42]) is continuous symmetry measures (Table 1) [10]. They measure how close the shape of the coordination polyhedron of a metal ion fits the ideal polyhedron, e.g., it should be zero for an ideal octahedron (OC-6) or an ideal trigonal prism (TRP-6). For iron(II) complexes with a N6 donor set, the transition from LS to HS state upon raising the temperature is associated with a trigonal twist distortion (the Bailar twist) from octahedral geometry [42]. The symmetry measures for Fe(L1) 2 OTf 2 , Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 are significantly removed from this pathway (Figure 5a), as typical of iron(II) complexes with tridentate ligands spanning mer positions (such as bpp or terpy) owing to the constrains imposed by these rigid ligands [14]. The corresponding S(OC-6) and S(TRP-6) values, which nicely overlap for NMR (PBE0-optimized) and X-ray (experimental) geometries ( Figure 5), fall into the range of HS only for iron(II) complexes with both 1-bpp [12] and 3-bpp [14] ligands. Surprisingly, Zn(L1) 2 (ClO 4 ) 2 is not an exception, although its coordination geometry is slightly shifted towards an octahedron (similar to the isolated [Fe(L1) 2 ] 2+ and [Fe(L2) 2 ] 2+ in the HS state). This behaviour indicates a D 2 d double axial distortion common for the iron(II) complexes of terpy [43] and both families of bpp ligands [12], with a geometry of the HS state deformed toward an edge-bicapped tetrahedron (ebcT-6) [10]. Such a distortion of the zinc(II) complex, however, cannot be explained by the Zn-N bond lengths being close to those of the HS iron(II) ion, as the zinc(II) ion has a more octahedral-like coordination environment in other complexes of bpp available in Cambridge structural database, CSD (see, e.g., refcodes JIZCUS [44], YODNOW [32], YODNUC [32], XEGFEU [30] and XIMSIV [31]). which nicely overlap for NMR (PBE0-optimized) and X-ray (experimental) geometries ( Figure 5), fall into the range of HS only for iron(II) complexes with both 1-bpp [12] and 3-bpp [14] ligands. Surprisingly, Zn(L1)2(ClO4)2 is not an exception, although its coordination geometry is slightly shifted towards an octahedron (similar to the isolated [Fe(L1)2] 2+ and [Fe(L2)2] 2+ in the HS state). This behaviour indicates a D2d double axial distortion common for the iron(II) complexes of terpy [43] and both families of bpp ligands [12], with a geometry of the HS state deformed toward an edge-bicapped tetrahedron (ebcT-6) [10]. Such a distortion of the zinc(II) complex, however, cannot be explained by the Zn-N bond lengths being close to those of the HS iron(II) ion, as the zinc(II) ion has a more octahedral-like coordination environment in other complexes of bpp available in Cambridge structural database, CSD (see, e.g., refcodes JIZCUS [44], YODNOW [32], YODNUC [32], XEGFEU [30] and XIMSIV [31]).   (Table 1) and NMR-based approach, respectively.
If referenced to OC-6 and ebcT-6 ( Figure 5b), the four complexes feature a rather small deviation from this, more accurate pathway [10] towards a vacant cape trigonal prism (as found for the HS complexes of [Fe(terpy)2] 2+ with phenyl, p-tolyl or pyridyl groups at α-positions [10]) or towards a divacant bicapped trigonal prism (as found for some bpp complexes [10]). In all those cases, the deviations were attributed to weak additional coordination of the metal centre either by the above phenyl (p-tolyl or pyridyl) substituent [10] or by outer-sphere anions [12].

Distortion of Molecular Geometry by Secondary Interactions
There is, however, very little indication that the metal ion in Fe(L1)2OTf2, Fe(L2)2OTf2, Fe(L2)2(BF4)2 or Zn(L1)2(ClO4)2 is involved in any of these secondary interactions. The outer-sphere species (such as anions or solvate molecules) in their crystals are too far away from the metal ion (>4.9 Å), and it would disagree with the distortion being an intramolecular effect as follows from the above NMR data. Coordination by the phenyl groups may be suspected only if the shortest M…C distances (3.73, 3.81 and 3.67 Å for Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2 vs. 3.81 Å for Zn(L1)2(ClO4)2) are compared to the sum of the van-der-Waals radii of Alvarez (4.21 and 4.16 Å for Fe…C and Zn…C contacts) [18]; relative to more conventional radii of Bondi [45] (2.09 Å for Zn…C) or Batsanov [46] (3.75 and 3.8 Å for Fe…C and Zn…C, respectively), they are too long for such an interaction to occur. In addition, these distances do not correlate with the average Fe-N distances, as in some of [Fe(1bpp)2] 2+ complexes [47], nor there is an elongation of the opposite Fe-N bonds [10], as caused by secondary coordination with a phenyl or similar group in [Fe(terpy)2] 2+ . Therefore, the observed   (Table 1) and NMR-based approach, respectively. If referenced to OC-6 and ebcT-6 (Figure 5b), the four complexes feature a rather small deviation from this, more accurate pathway [10] towards a vacant cape trigonal prism (as found for the HS complexes of [Fe(terpy) 2 ] 2+ with phenyl, p-tolyl or pyridyl groups at α-positions [10]) or towards a divacant bicapped trigonal prism (as found for some bpp complexes [10]). In all those cases, the deviations were attributed to weak additional coordination of the metal centre either by the above phenyl (p-tolyl or pyridyl) substituent [10] or by outer-sphere anions [12].

Distortion of Molecular Geometry by Secondary Interactions
There is, however, very little indication that the metal ion in Fe(L1) 2 OTf 2 , Fe(L2) 2 OTf 2 , Fe(L2) 2 (BF 4 ) 2 or Zn(L1) 2 (ClO 4 ) 2 is involved in any of these secondary interactions. The outer-sphere species (such as anions or solvate molecules) in their crystals are too far away from the metal ion (>4.9 Å), and it would disagree with the distortion being an intramolecular effect as follows from the above NMR data. Coordination by the phenyl groups may be suspected only if the shortest M . . . C distances (3.73, 3.81 and 3.67 Å for Fe(L1) 2 OTf 2, Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 vs. 3.81 Å for Zn(L1) 2 (ClO 4 ) 2 ) are compared to the sum of the van-der-Waals radii of Alvarez (4.21 and 4.16 Å for Fe . . . C and Zn . . . C contacts) [18]; relative to more conventional radii of Bondi [45] (2.09 Å for Zn . . . C) or Batsanov [46] (3.75 and 3.8 Å for Fe . . . C and Zn . . . C, respectively), they are too long for such an interaction to occur. In addition, these distances do not correlate with the average Fe-N distances, as in some of [Fe(1-bpp) 2 ] 2+ complexes [47], nor there is an elongation of the opposite Fe-N bonds [10], as caused by secondary coordination with a phenyl or similar group in [Fe(terpy) 2 ] 2+ . Therefore, the observed distortion of coordination geometry in Fe(L1) 2 OTf 2, Fe(L2) 2 OTf 2 , Fe(L2) 2 (BF 4 ) 2 and Zn(L1) 2 (ClO 4 ) 2 ) and the related HS locking of the iron(II) complexes are not associated with the existence of such weak coordinating interactions; the energy penalty to rotation of the phenyl groups [17] is a more likely reason for it.

Rotation of Phenyl Groups
According to X-ray diffraction, the phenyl groups in all the four complexes are rotated relative to the pyrazol-3-yl planes ( Figure 6) by 42.7 • in Fe(L1) 2 OTf 2 (slightly higher than in its early solvatomorph [23]) to 66.4 • in Zn(L1) 2 (ClO 4 ) 2 , the average angle between the two planes being 49.9, 53.5 and 56.5 • in Fe(L1) 2 OTf 2 , Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 and 59.7 • in Zn(L1) 2 (ClO 4 ) 2 . A similar rotation of the phenyl group but relative to the pyrazol-1-yl moiety in the iron(II) complex of 1,3-bpp (by 45.3-49.2 • ) [17] has previously been proposed to result mainly from avoiding steric hindrances with the second ligand (although interligand stacking interactions were also found to contribute slightly to the stability of a likewise distorted HS state) [17]. It was computationally shown to cause an energetic penalty associated with the loss of planarity in the phenyl-pyrazol-1-yl moiety, preventing the complexes from adopting an undistorted LS state in which the phenyl groups are nearly perpendicular to the pyrazolyl plane. However, the derivatives of 3-bpp may feature a different behaviour when compared to 1-bpp (as hinted by an observed opposite trend in the influence of electron donating/withdrawing substituents on the spin state of iron(II) complexes) [12,15].
Magnetochemistry 2018, 4, x FOR PEER REVIEW 9 of 17 distortion of coordination geometry in Fe(L1)2OTf2, Fe(L2)2OTf2, Fe(L2)2(BF4)2 and Zn(L1)2(ClO4)2) and the related HS locking of the iron(II) complexes are not associated with the existence of such weak coordinating interactions; the energy penalty to rotation of the phenyl groups [17] is a more likely reason for it.

Rotation of Phenyl Groups
According to X-ray diffraction, the phenyl groups in all the four complexes are rotated relative to the pyrazol-3-yl planes ( Figure 6) by 42.7° in Fe(L1)2OTf2 (slightly higher than in its early solvatomorph [23]) to 66.4° in Zn(L1)2(ClO4)2, the average angle between the two planes being 49.9, 53.5 and 56.5° in Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2 and 59.7° in Zn(L1)2(ClO4)2. A similar rotation of the phenyl group but relative to the pyrazol-1-yl moiety in the iron(II) complex of 1,3-bpp (by 45.3-49.2°) [17] has previously been proposed to result mainly from avoiding steric hindrances with the second ligand (although interligand stacking interactions were also found to contribute slightly to the stability of a likewise distorted HS state) [17]. It was computationally shown to cause an energetic penalty associated with the loss of planarity in the phenyl-pyrazol-1-yl moiety, preventing the complexes from adopting an undistorted LS state in which the phenyl groups are nearly perpendicular to the pyrazolyl plane. However, the derivatives of 3-bpp may feature a different behaviour when compared to 1-bpp (as hinted by an observed opposite trend in the influence of electron donating/withdrawing substituents on the spin state of iron(II) complexes) [12,15]. Indeed, the phenyl groups in free 3-bpp ligands available in CSD (refcodes ILESUP and GETSUL) [23] are already rotated by 28-43°, and in a rare case of M(L)X2 complexes (refcodes GECSEP and GESCIP) [23], by up to ~35°. In the 1-bpp ligand with the phenyl groups at the same position (2,6-bis(5-(4-Methoxyphenyl)-3-phenyl-1H-pyrazol-1-yl)pyridine, SUPZIN), the corresponding angles are ~14 and 28°. Moreover, an analysis of CSD for all the ordered organic compounds that contain a phenyl-pyrazol-1-yl or a phenyl-pyrazol-3-yl fragment ( Figure S7) shows that the angle between the two planes in the case of the latter adopts a wider variety of values with a maximum at ~40°, while the former are clustered around 0°. Their behaviour mirrors calculated energy variation upon the rotation of the phenyl group in the corresponding phenylpyrazole molecules [17] and suggests that the 3-bpp ligands have a lower penalty to their rotation (Figure 7). Indeed, the phenyl groups in free 3-bpp ligands available in CSD (refcodes ILESUP and GETSUL) [23] are already rotated by 28-43 • , and in a rare case of M(L)X 2 complexes (refcodes GECSEP and GESCIP) [23], by up to~35 • . In the 1-bpp ligand with the phenyl groups at the same position (2,6-bis(5-(4-Methoxyphenyl)-3-phenyl-1H-pyrazol-1-yl)pyridine, SUPZIN), the corresponding angles are~14 and 28 • . Moreover, an analysis of CSD for all the ordered organic compounds that contain a phenyl-pyrazol-1-yl or a phenyl-pyrazol-3-yl fragment ( Figure S7) shows that the angle between the two planes in the case of the latter adopts a wider variety of values with a maximum at~40 • , while the former are clustered around 0 • . Their behaviour mirrors calculated energy variation upon the rotation of the phenyl group in the corresponding phenylpyrazole molecules [17] and suggests that the 3-bpp ligands have a lower penalty to their rotation (Figure 7).
In Fe(L1) 2 OTf 2 , Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 , however, the phenyl groups rotate to higher angles than in an isomeric [Fe(Ph 2 -1-bpp) 2 [17], in which they follow the degree of the 'twist' distortion as gauged by the θ values (~61 and 70 • , respectively; Figure S8); no statistically significant correlation between these two parameters was observed. Given the initially higher rotation angles in the HS state (and a generally lower energy variation upon the rotation; Figure 7), it should be easier for the iron(II) complexes with the phenyl-substituted 3-bpp ligands to go to an undistorted LS state. Indeed, according to our PBE0/def2-TZVP calculation for a model complex [Fe(Ph 2 -3-bpp) 2  In Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2, however, the phenyl groups rotate to higher angles than in an isomeric [Fe(Ph2-1-bpp)2] 2+ (32.4-36.5°) [34] or the above [Fe(Ph2-1,3-bpp)2] 2+ (45.3-49.2°) [17], in which they follow the degree of the 'twist' distortion as gauged by the θ values (~61 and 70°, respectively; Figure S8); no statistically significant correlation between these two parameters was observed. Given the initially higher rotation angles in the HS state (and a generally lower energy variation upon the rotation; Figure 7), it should be easier for the iron(II) complexes with the phenylsubstituted 3-bpp ligands to go to an undistorted LS state. Indeed, according to our PBE0/def2-TZVP calculation for a model complex

Stacking Interactions
Additional contribution to the HS locking in Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2 by the phenyl groups may come from stabilization of a distorted molecular geometry by intramolecular stacking interactions, which all the four complexes have in common. Thus, the phenyl groups are significantly rotated to the plane of the pyrazol-3-yl moieties, but are nearly parallel to the pyridine ring of the second ligand, as in an earlier solvatomorph of Fe(L1)2OTf2 [23]. The angles between the planes are 4.4-16.1° (Table 1) with those in Zn(L1)2(ClO4)2 being on a smaller side (7.4° on average vs. 13.2, 11.8 and 10.2° in Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2). The zinc(II) complex, however, does not stand out if the shortest C…C distances are considered (3.413-3.667 Å); their average being 3.477, 3.553 and 3.499 Å in Fe(L1)2OTf2, Fe(L2)2OTf2 and Fe(L2)2(BF4)2 and 3.566 Å in Zn(L1)2(ClO4)2. These stacking interactions bind the phenyl and pyridine groups in a so-called 'parallel-displaced' orientation [48] (Figure 6), which is significantly preferred (over 'face-to-face' when the two groups are parallel, or 'T-shaped' when they are perpendicular to each other) for stacking interactions between benzene and pyridine molecules [49]; the binding energy may be as high as 3 kcal/mol at similar intermolecular separations.
The complexes keep the same orientation (with the phenyl groups slightly displaced towards the pyrazol-3-yl moiety) in solution, as hinted by a good correlation between the calculated chemical shifts and those measured experimentally for the HS iron(II) complexes, and is supported by chemical shifts for Zn(L1) 2 (ClO 4 ) 2 (Figures S9 and S10) and, more suited for this purpose, ROESY NMR data ( Figure  S11). Signals of the protons of the phenyl and pyridine groups in its 1 H NMR spectra (Figures S9 and S10) are shifted to higher field as compared to the free ligand L1, which is consistent with the shielding expected if these groups are involved in a 'parallel-displaced' stacking interaction. In the ROESY spectrum of Zn(L1) 2 (ClO 4 ) 2 ( Figure S11), the cross-peaks appear between the nuclei of the phenyl and pyridine groups, which are separated by more than 7 Å in the ligand but closely approach each other in the complex. The observed cross-peaks agree well with the molecular geometry as obtained by X-ray diffraction for Zn(L1) 2 (ClO 4 ) 2 .
A final piece of experimental evidence for an important role of intramolecular stacking interactions in the stabilization of a distorted molecular geometry of [ML 2 ] 2+ is based on a tendency of the 3-bpp ligands to form heteroleptic iron(II) complexes with terpy in solution [55]. Mixing equimolar quantities of L1, FeCl 2 and terpy in different solvents (methanol, DMSO and acetonitrile) did not result in the formation of a heteroleptic complex; only a mixture of [Fe(L1) 2 ]Cl 2 (with a metal ion in the HS state as gauged by the paramagnetic shifts) and diamagnetic [Fe(terpy) 2 ]Cl 2 was always observed ( Figure S12). Efforts to combine the two (L1 favoring the HS state and terpy favoring the LS state [55] of iron(II) ion) in one complex to 'unlock' the spin transition are underway in our group.

Materials and Methods
Synthesis. All synthetic manipulations were carried out in air. Solvents were purchased from commercial sources, purified using Innovative Technology SPS-400 PureSolv solvent system or by distilling from conventional drying agents and degassed by the freeze-pump-thaw method twice prior to use; THF was stored over activated 4 Å molecular sieves or sodium metal pieces. The compounds L1 and Fe(L1) 2 OTf 2 were synthesized as previously reported [23].
Ligand L2. Compound L1 (395 mg, 0.1 mmol) was dispersed in 10 ml of THF, and 1 mL of Et 3 N was added. The mixture was cooled down to −15 • C using an ice-salt bath, and 151 µL (0.21 mmol) of AcCl was added dropwise using a micro syringe. After warming up to r.t., the reaction mixture was stirred for additional 12 h. Reaction mixture was then filtered from ammonium salt, and volatiles were removed under vacuum to give light-yellow fine-crystalline solid. Yield 405 mg (85%). 1

Zn(L1) 2 (ClO 4 ) 2 .
To the suspension of 30 mg (0.076 mmol) of the compound L1 in 2 mL of ethanol, 12 mg (0.036 mmol) of Zn(ClO 4 ) 2 ·4H 2 O was added in one portion. Reaction mixture was stirred for 2 h at r.t. and then refluxed for 1 h, cooled to r.t., and evaporated to dryness. The solid product was washed with diethyl ether (3 × 7 mL), dichloromethane (2 × 7 mL) and extracted with acetonitrile (2 × 5 mL). The extract was filtered and evaporated to dryness. Yield: 30 mg (79%). Anal. Calcd for C 46  Magnetic measurements. Magnetic susceptibility for the iron(II) complexes was measured in the temperature range 3-300 K with a Quantum Design PPMS-9 (Quantum Design Inc., San Diego, CA, USA) device under the dc magnetic field of 1 kOe. Finely ground microcrystalline powders were immobilized in mineral oil matrix inside a polyethylene capsule. The data were corrected for the sample holder, the mineral oil and the diamagnetic contribution; the latter by using the Pascal constants [56]. Smoothing was performed by the Savitzky-Golay method with five points of window and polynomial order of two. The observed variable-temperature dc magnetic susceptibility data were fitted in PHI software [33] using the spin Hamiltonian ( Figure 1 and Figure S1): X-ray crystallography. X-ray diffraction experiments of single crystals, which were grown by recrystallization of powder samples from dichloromethane (for Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 ) or diffusion of diethyl ether to the solution in acetonitrile (for Fe(L1) 2 OTf 2 and Zn(L1) 2 (ClO 4 ) 2 ) on air, were carried out with a Bruker APEX2 DUO CCD diffractometer (for Fe(L2) 2 (BF 4 ) 2 ) and with a Bruker APEX2 CCD diffractometer (for all others), using the graphite monochromated Mo-Kα radiation (λ = 0.71073 Å, ω-scans) at 120 K. The structures were solved by direct method and refined by the full-matrix least-squares against F 2 in anisotropic approximation for non-hydrogen atoms. Hydrogen atoms of OH groups (in Fe(L1) 2 OTf 2 and Zn(L1) 2 (ClO 4 ) 2 ) and those of solvate water molecules (in Fe(L2) 2 OTf 2 and Zn(L1) 2 (ClO 4 ) 2 ), presumably absorbed from air, were located in difference Fourier synthesis. Positions of other hydrogen atoms were calculated, and they all were refined in the isotropic approximation in the riding model. Crystal data and structure refinement parameters for these complexes are given in Table S1. All calculations were performed using the SHELXTL software [57]. CCDC 1843948-1843951 contain the supplementary crystallographic data, which can be obtained free of charge via http://www.ccdc.cam.ac.uk/conts/retrieving.html. NMR spectroscopy. 1 H 1D and 2D NMR spectra were recorded from CD 3 CN solutions with a Bruker Avance 600 FT-spectrometer (600.22 and 150.96 MHz, respectively). The measurements were done using the residual signals of CD 3 CN ( 1 H 1.94 ppm, 13 C 118.26 ppm). No change in chemical shifts was observed when using another solvent, such as methanol.
Evans method. Temperature dependence of magnetic susceptibility of the iron(II) complexes in acetonitrile solutions was evaluated by the Evans method [25,26] in the temperature range 240-330 K (Figure 1 and Figure S1), using a Wilmad NMR tube with a coaxial insert. The inner (reference) tube was filled with acetonitrile-d 3 with approximately 1% of Me 4 Si, and the outer tube contained the solution of the complex (~5-10 mg/cm 3 ) in acetonitrile-d 3 with the same concentration of Me 4 Si. Molar magnetic susceptibility was calculated from the difference between the chemical shift of Me 4 Si in pure acetonitrile-d 3 and its shift in a solution of the complex (∆δ in Hz) in acetonitrile-d 3 using the following equation: (M-molar weight of the iron(II) complex; υ 0 -frequency of the spectrometer; S f -shape factor of the magnet (4π/3); c-concentration of the complex; χ M dia -molar diamagnetic contribution to the paramagnetic susceptibility, calculated using the Pascal's constant [56]). The concentration c was recalculated for each temperature in accordance with the density change of the solvent ρ:c T = m s ρ/m sol , where m s is the mass of the complex and m sol is the mass of the solution. Analysis of paramagnetic chemical shifts. Paramagnetic shifts in 1 H NMR spectra for acetonitrile solutions of the iron(II) complexes were measured (as hundreds of ppm) as a difference between the observed chemical shifts and those for their diamagnetic analogue, the zinc(II) complex Zn(L1) 2 (ClO 4 ) 2 (for Fe(L1) 2 OTf 2 ) or a free ligand L2 (for Fe(L2) 2 OTf 2 and Fe(L2) 2 (BF 4 ) 2 ). The theoretical chemical shifts for the iron(II) complexes were obtained as follows: δ obs = δ dia + δ CS + δ PCS (δ dia is a diamagnetic contribution to the observed chemical shift (in ppm), which is taken as a chemical shift of the corresponding nucleus in the NMR spectra of a suitable diamagnetic analogue). Isotropic paramagnetic (contact) contribution δ CS , which arises from spin polarization conveyed through molecular orbitals, was evaluated through the following equation: (S-electron spin; g N -nuclear g-factor; µ B -Borh magneton; µ N -nuclear magneton). The value of g-tensor and isotropic values of hyperfine interaction tensors A iso are taken from quantum chemical calculations [58], using optimized and X-ray geometries (with the positions of hydrogen atoms optimized as described below) of the iron(II) complexes. The pseudocontact contribution δ PCS , which arises from dipolar coupling between magnetic moments of a nucleus and of an unpaired electron [35,59,60], was estimated by fitting the observed chemical shifts with the following equation: (θ i and r i -polar coordinates of the nuclei in the coordinate frame of the magnetic susceptibility tensor χ; ∆χ ax -axial anisotropy of χ). Quantum chemistry. All quantum chemical calculations were done using the ORCA package, v. 4.0 [61]. Geometry optimization of the ligands was performed with a PBE0 functional [62] and a basis set def2-TZVP [63]. For the metal complexes, X-ray diffraction geometries were used as a starting point for geometry optimization with the hybrid PBE0 functional (providing simultaneously good results for the energy difference between HS and LS states in octahedral Fe(II) complexes [64] and for hydrogen spin densities [65] used in calculating chemical shifts in NMR spectra of paramagnetic compounds [29]), the scalar relativistic zero-order regular approximation (ZORA) [66], Grimme's DFT-D3 dispersion correction [67] and the scalar relativistically recontracted (SARC) [68] version of the def2-TZVP basis set [63]. Extra tight thresholds for forces and displacements were used. The solvation effects were included using the Conductor-like Polarizable Continuum Model, as implemented in ORCA 4.0, with acetonitrile as a solvent. The resulting geometries of [Fe(L1) 2 ] 2+ and [Fe(L2) 2 ] 2+ in the HS state were used to compute g-tensor and isotropic values of hyperfine interaction tensors A iso [58]. A model complex [Fe(Ph 2 -3-bpp) 2 ] 2+ was calculated in a similar way in its LS and HS states.

Conclusions
A combined study of four complexes with the 1-phenylsubstituted 3-bpp ligands that have different other substituents (with distinct steric and electronic properties and H-bonding ability), counterions, lattice solvents and even the metal ions (bistable iron(II) and diamagnetic zinc(II)) by magnetochemistry, X-ray diffraction and multi-temperature NMR spectroscopy provided an experimental evidence for the intramolecular origin of their HS locking (the HS state being thermodynamically more stable than the LS state) and the distorted molecular geometry, both found in iron(II) complexes of similar 1-bpp [34], 1,3-bpp [17] and terpy [51,52] ligands. It was attributed to the pendant phenyl groups that act through an energetically unfavourable rotation to avoid steric clashes with the second ligand [17] and, apparently, intramolecular 'parallel displaced' stacking interactions; the latter are shared by complexes with different tri-dentate ligands and aromatic substituents, such as phenyl [34], p-tolyl [51], metoxyphenyl [53] and pyridyl [54]. They are therefore guaranteed to yield the HS iron(II) complexes, which in the case of 1-unsubstituted 3-bpp is hindered by a dramatic influence of the environment on the spin state [69] owing to its NH groups (so that there is no clear correlation with the electromeric characteristics of other substituents) [15].
Even if found in one spin state, a thorough characterization of iron(II) complexes with new derivatives of 3-bpp and the like (including heteroleptic complexes [55]), especially in solutions, is a key to more accurate magneto-structural correlations. Together with a possibility for their high-throughput screening by variable-temperature NMR spectroscopy to identify a spin transition by paramagnetic shifts and even quantify it by the proposed approach (applicable to compounds obtained in situ without their isolation or purification, in contrast to the popular Evans method) [29], it paves the way to a truly 'molecular' design [15] of spin transition compounds for many future applications [3].