A Unified Treatment of the Relationship Between Ligand Substituents and Spin State in a Family of Iron(II) Complexes

Abstract The influence of ligands on the spin state of a metal ion is of central importance for bioinorganic chemistry, and the production of base‐metal catalysts for synthesis applications. Complexes derived from [Fe(bpp)2]2+ (bpp=2,6‐di{pyrazol‐1‐yl}pyridine) can be high‐spin, low‐spin, or spin‐crossover (SCO) active depending on the ligand substituents. Plots of the SCO midpoint temperature (T 1/2 ) in solution vs. the relevant Hammett parameter show that the low‐spin state of the complex is stabilized by electron‐withdrawing pyridyl (“X”) substituents, but also by electron‐donating pyrazolyl (“Y”) substituents. Moreover, when a subset of complexes with halogeno X or Y substituents is considered, the two sets of compounds instead show identical trends of a small reduction in T 1/2 for increasing substituent electronegativity. DFT calculations reproduce these disparate trends, which arise from competing influences of pyridyl and pyrazolyl ligand substituents on Fe‐L σ and π bonding.

(1) Solution phase magnetic susceptibility data and computational details Page Experimental details -solution magnetic measurements 3
9 Table S3 Computed energy difference between the high-spin and low-spin forms of the complexes.
11 Table S4 Calculated composition of the metal d-orbitals in [Fe(bpp H,H ) 2 ] 2+ . 12 Table S5 Computed d-orbital energies in the low-spin forms of the complexes. 13   2+ and [Fe(bpp H,Y ) 2 ] 2+ complexes in their high-spin and low-spin states.

17
(2) Synthetic, crystallographic and solid state magnetic data Page Scheme S1 Syntheses of the new bpp X,H ligands in this work. 42

Experimental details of the ligand syntheses 42
Experimental details of the synthesis of the complexes 44  64 Table S13 Selected bond lengths and angles in the crystal structures of the halogenated ligand complexes.

Figure S18
Packing diagrams of [Fe(bpp F,H ) 2 ] [BF 4 ] 2 in its high-spin and low-spin states. 66 Figure S19 Solid state magnetic susceptibility data for the new complexes in this work. 67

Experimental details -solution magnetic measurements
Magnetic susceptibility measurements in solution were obtained by Evans method using a Bruker DRX500 or Avance500 spectrometer operating at 500.13 MHz. [1] A diamagnetic correction for the sample, [2] and a correction for the variation of the density of the solvent with temperature, [3] were applied to these data. The spin-crossover midpoint temperatures from these data were derived by fitting the data to eq 1 and 2, where n HS (T) is the high-spin fraction of the sample at temperature T (Fig. S1).

Experimental details -Computational study
All DFT calculations employed the ORCA program system, version 3.0.1. [5] All complexes were fully optimized using the Becke Perdew (BP86) functional [6,7] and a def2-SVP basis set. [8] The resolution of identity approximation was also used with a def2-SVP/J auxiliary basis. Low spin systems were treated spin restricted and high spin systems spin unrestricted. For the latter, the spin expectation value <S 2 > was of the order of 6.06 compared to the ideal value of 6. Although the theoretical justification within DFT of trusting the value of <S 2 > is debateable, we judge the spin contamination to be negligible.
To account for condensed phase effects, the optimizations were carried out in a polarizable continuum solvent using the conductor like screening model (COSMO) [9] implemented in ORCA. The solvent used was acetone. Default convergence criteria for the SCF and geometry were applied. Frequency calculations were carried out for the unsubstituted complex [Fe(bpp H,H ) 2 ] 2+ which confirmed that both HS and LS forms were local minima. These structures were employed as the starting coordinates for substituted systems and no further frequency calculations were done. In common with our previous study of this type, [10] no corrections were made for zero point energies or dispersion and the analysis relies solely on the total electronic energies.
For substituents where multiple conformations are possible, the ones selected for this work are shown below (Fig. S5). Systematic conformational searches were not carried out. However, several conformations were calculated for CH 2 SCN and CH 2 OH, which showed minimal variation of the spin state and MO energy differences indicating significant error cancellations and that the particular choices of conformation should not significantly alter our conclusions.
After each structure minimization, a Loewden population analysis was performed in ORCA [5] using the following commands: !RKS BP RI def2-SVP def2-SVP/J TightSCF SlowConv Cosmo(Acetone) %Output Print [P_ReducedOrbPopMO_L] 1 End The five metal-based d-frontier orbitals were clearly identifiable in the resultant outputs, as containing ca. 65-70 % metal d character (Table S4). The atomic coordinates of all the minimized structures are listed in Table S6, while the d-orbital energies are summarized in Table S5. Plots correlating the spin state energies or orbital energies with substituent Hammett parameters were produced using SIGMAPLOT. [4] [5] F. Neese, U. Becker, D. Ganiouchine, S. Koßmann, T. Petrenko, C. Riplinger, F. Wennmohs, ORCA 3.0.1 edn., Max Planck Institute for Chemical Energy Conversion, Mülheim an der Ruhr, 2014.
[10] B. J. Houghton, R. J. Deeth, Eur. J. Inorg. Chem. 2014, 4573. Table S1. Solution phase spin-crossover data for [Fe(bpp X,H ) 2 ] 2+ (HS = fully high-spin over the liquid range of the solvent). Thermodynamic parameters are only quoted for equilibria that are fully resolved within the temperature range of the measurement.  [a] T ½ derived by extrapolation below the freezing point of the solvent. [b] Compound remains fully high-spin within the liquid range of the solvent. [c] Hammett constant value is for R = SSMe.
Some of these data are also included in Fig. 1 of the main article. No simulation of the data was undertaken for [Fe(bpp NH2,H [PF 6 ] 2 are of lower accuracy than the others, for similar reasons, and the errors on the quoted T ½ values for those compounds are correspondingly larger. The derived T ½ values for those compounds are in good agreement with predictions based on the  P and  P + Hammett parameters, however (Fig. 3, main article).

Fig. S2
. Plots of the measured T ½ values for [Fe(bpp X,H ) 2 ] 2+ vs. the basic pK a of the corresponding 4-substituted pyridine [20] (top), and of [Fe(bpp H,Y ) 2 ] 2+ vs. the basic pK a of the corresponding 4-substituted pyrazole [21] (bottom). The lines show the best fit correlations (R 2 = 0.78 [top] and 0.83 [bottom]), omitting the X = NH 2 and NMe 2 datapoints (white circles) whose T ½ values represent the upper limits for those measurements.
Both series of compounds show reasonable, but opposite, linear relationships between T ½ and the basicity of the corresponding substituted heterocyclic ligand donor. For [Fe(bpp X,H ) 2 ] 2+ , more basic pyridyl donors stabilize the high-spin state of the complex. In contrast, for [Fe(bpp H,Y ) 2 ] 2+ , more basic pyrazolyl donor groups favor the low-spin state.
Hence, the different relationships between T ½ and substituent Hammett parameter in [Fe(bpp X,H ) 2 ] 2+ and [Fe(bpp H,Y ) 2 ] 2+ complexes does not simply reflect the basicity of the bpp X,Y ligands, but must be a function of the metal-ligand interaction.
The correlation coefficient for this graph is R 2 = 0.79.

Table S5
Computed d-orbital energies in the low-spin forms of complexes in this work, and the average energies of the t 2g and e g subshells (all in kcal.mol -1 ). These data are plotted in Fig. 6 of the main article, and in Fig. S4.   (Table S4). The average orbital energies of the t 2g and e g subshells are also shown, along with their best fit correlations and slopes.
The slope of these correlations is almost exactly 2x greater for Y substituents than for X substituents, because there as twice as many Y substituents as X groups in a [Fe(bpp X,Y ) 2 ] 2+ molecule.

Ligand synthesis
The following ligands were prepared according to the literature procedure: bpp Cl,H ; [26] bpp Br,H ; [27] bpp I,H ; [22] bpp OH,H [25] and bpp CO2H,H . [28] The starting material 2,6-difluoro-4-dimethylaminopyridine was also prepared by the published method. [23] All other reagents were purchased commercially and used as supplied unless otherwise stated, although diglyme was always dried over sodium before use.
Synthesis of 4-fluoro-2,6-di(pyrazol-1-yl)pyridine (bpp F,H ). 48% Aqueous HBF 4 (10 cm 3 , 54.7 mmol) was syringed into a Schlenk tube containing bpp NH2,H (0.29 g, 1.28 mmol) under anaerobic conditions. The acidified contents were cooled to 0°C, and a degassed aqueous solution of NaNO 2 (0.20 g, 2.83 mmol) was gradually added with stirring causing the precipitation of a bright yellow solid. Addition of MeCN (15 cm 3 ) led to immediate evolution of N 2 , and the mixture was then heated at 80°C for 0.5 h until the evolution ceased. Once cool, the MeCN was removed in vacuo and the remaining yellow solution poured into H 2 O (30 cm 3 ) and neutralized with aqueous NaOH causing the formation of a pale yellow suspension. Extraction with CHCl 3 (3 x 50 cm 3 ), drying with MgSO 4 , filtration and removal of the volatiles yielded a crude pale yellow solid which was purified through silica gel column chromatography ( 19 F NMR spectrum (CDCl 3 ): δ −95.6 (t, 9.3 Hz).
Synthesis of 4-methoxy-2,6-di(pyrazol-1-yl)pyridine (bpp OMe,H ). MeI (0.19 g, 1.30 mmol) was added to a stirred suspension of K 2 CO 3 (0.18 g, 1.30 mmol) and bpp OH,H (0.18 g, 0.80 mmol) [25] in acetone (10 cm 3 ) under N 2 , and the resultant mixture was refluxed for 24 h. The solution was concentrated to ca. 25 % its original volume, and then diluted to 50 cm 3 with CHCl 3 . Aqueous NaOH (2 x 25 cm 3 ) was then used to wash the resultant suspension, and the volatiles were removed in vacuo. The resulting yellow solid was triturated in hexane (10 cm 3 ), collected on a glass frit and washed with a further few drops of hexane before drying in vacuo. Yield 0. 13
No disorder is present in any of these structures, and no restraints were applied to the refinements. All non-H atoms were refined anisotropically. For bpp NO2,H , H atoms were located in the Fourier map and allowed to refine freely, with U iso values constrained to 1.2x U eq of the associated C atom. Attempts to refine H atom positions for bpp H,tBu were unsuccessful, so H atoms in this structure were placed in calculated positions and refined using a riding model. 0.510 [a] 0.629 [a] 0.577 [a] 0.715 [a]   These crystals are isostructural, and their diffraction data were modelled in the same way. One nitromethane molecule in both structures was disordered over two equally occupied sites, which were refined using the fixed restraints C-N = 1.48(2), N-O = 1. 22(2), O...O = 2.14(2) and C...O = 2.32(2) Å. All non-H atoms were refined anisotropically, while C-bound H atoms were placed in calculated positions and refined using a riding model. The amino H atoms were located in the difference map and allowed to refine with a common U iso displacement parameter.  ≈ 0.25). Data from this pair of isostructural crystals were also refined using the same procedure. One of the two anions in each structure is disordered over two sites, with refined occupancies close to 0.75:0.25. Refined distance and angle restraints were applied to the minor anion disorder site. In addition to the cations and anions, a Fourier peak that was not bonded to any other residue was modelled as a partial water molecule, whose occupancy also refined to approximately 0.25. This partial water site, O(50) lies within hydrogen bonding distance of the disordered anion, and of its symmetry equivalent related by ½+x, y, ½−z. Since the anion disorder appears to correlate with the presence or absence of O(50), O(50) was refined with the same occupancy as the minor disorder residue in the final least squares cycles. Although the partial H atoms associated with this water site could not be located or refined, the water H content is included in the density and F000 calculations. All non-H atoms with occupancy >0.5 were refined anisotropically, and H atoms were placed in calculated positions and refined using a riding model. The water content in the refinement is supported by microanalysis, since bulk samples of both salts analyse reasonably for a hemihydrate formulation.

Structure refinements of [Fe(bpp F,H ) 2 ][BF 4 ] 2 .
Since cycling the crystal across the transition leads to rapid decay in diffraction quality, several attempts were required to obtain a good structure of the compound in its low-spin state. Hence, the two refinements reported here were obtained from different crystals, which have opposite handedness. One of the two unique anions was disordered at both temperatures, over three sites at 290 K and two sites at 150 K. This was modelled using refined B−F and F…F distance restraints. All non-H atoms except for the minor anion disorder sites were refined anisotropically, and H atoms were placed in calculated positions and refined using a riding model. There are ten residual peaks of +1.0-1.5 e.Å -3 in the low temperature refinement all of them <1 Å from Fe1 or another non-H atom in the complex dication.  (1) spanning the crystallographic C 2 axis at 0, y, 3 /4. In contrast, the complex molecule in [Fe(bpp OMe ) 2 ][PF 6 ] 2 lies on a general crystallographic site with no internal symmetry. No disorder was detected during refinement of any of these structures, and no restraints were applied. All non-H atoms were refined anisotropically, while H atoms were placed in calculated positions and refined using a riding model.

Other measurements
Elemental microanalyses were performed by the University of Leeds School of Chemistry microanalytical service. Infra-red spectra were obtained as nujol mulls pressed between NaCl windows, between 600-4,000 cm -1 , using a Nicolet Avatar 360 spectrophotometer. 1 H NMR spectra employed a Bruker DPX300 spectrometer operating at 300.2 MHz. Electrospray mass spectra (ESI MS) were obtained on a Waters ZQ4000 spectrometer, from MeCN feed solutions. All mass peaks have the correct isotopic distributions for the proposed assignments.
Solid state magnetic susceptibility measurements were performed on a Quantum Design SQUID or SQUID/VSM magnetometer, with an applied field of 1000 or 5000 G and a scan rate of 5 Kmin -1 . A diamagnetic correction for the sample was estimated from Pascal's constants; [2] a diamagnetic correction for the sample holder was also applied. Neighboring heterocyclic groups in each ligand structure have a transoid-coplanar orientation, which is normal for compounds of this type. [34] [34] C. A. Bessel, R.  Atomic displacement ellipsoids are at the 50 % probability level. Only one orientation of the disordered nitromethane molecule is shown. Solvent molecules that do not take part in hydrogen bonding, and C-bound H atoms, have been omitted. Symmetry codes: (ii) x, ½-y, ½+z; (iii) x, ½-y, -½+z.   [a] Symmetry code: (ii) x, 1 /2-y, 1 /2+z The view is parallel to the (100) plane, with the unit cell c-axis horizontal. Only one orientation of the disordered nitromethane molecule is shown. Solvent molecules that do not take part in hydrogen bonding, and C-bound H atoms, have been omitted.   29), showing the influence between the partial water site O(50) and the disordered anion. Thermal ellipsoids are at the 50% probability level except for the minor anion disorder site, which has arbitrary radii and paler coloration. All C-bound H atoms have been omitted for clarity. Symmetry code: (vii) ½+x, y, ½−z.  Color code: C, white; F/Cl, yellow; Fe, green; N, blue.

Table S13
Selected bond lengths and angles in the crystal structures of the halogenated ligands (Å, º). [a] See  Fig. S17 for the atom numbering schemes employed. [Fe(bpp F,H  Displacement ellipsoids are at the 50 % probability level except for the BF 4 − ions which are de-emphasized for clarity. Only one orientation of the disordered anion environment is shown. Both views are parallel to the [001] crystal vector.
These datasets were collected from different crystals, which were of opposite handedness. The lowtemperature diagram is shown in the opposite of its true handedness, to allow comparison. The sample of [Fe(bpp H,tBu ) 2 ][BF 4 ] 2 ·2H 2 O is a poorly crystalline powder, that contains a mixture of high-spin and SCO-active material according to these data. The compound is almost fully high-spin at room temperature, but more detailed interpretation of its solid state properties is impossible at this stage since the compound was not obtained in crystalline form.
The spin-crossover temperatures from these data are listed in Table S9.