Spin-Crossover Temperature Predictable from DFT Calculation for Iron(II) Complexes with 4-Substituted Pybox and Related Heteroaromatic Ligands.

Spin-crossover (SCO) is a reversible transition between low and high spin states by external stimuli such as heat. The SCO behavior and transition temperature (T 1/2) of a series of [FeII(X-pybox)2](ClO4)2 were studied to establish a methodology for ligand-field engineering, where X-pybox stands for 2,6-bis(oxazolin-2-yl)pyridine substituted with X at the 4-position of the pyridine ring. We utilized X = MeO, Me, 3-thienyl, Ph, H, MeS, 2-thienyl, N3, Cl, Br, 3-pyridyl, and 4-pyridyl. The solution susceptometry on five new derivatives with X = Me, 2-thienyl, N3, Br, and 3-pyridyl was performed in acetone, giving the SCO temperatures of 220, 260, 215, 280, and 270 K, respectively. The density-functional-theory molecular orbital (MO) calculation was performed on the ligands with geometry optimization. The atomic charge on the pyridine nitrogen atom [ρ(Npy)] was extracted from the natural orbital population analysis. Positive correlation appeared in the T 1/2 versus ρ(Npy) plot with R 2 = 0.734, being consistent with the analysis using the Hammett substituent constants (σp and σp +). This finding well agrees with the mechanism proposed: the rich electron density lifts the t2g energy level through the dπ-pπ interaction, resulting in a narrow t2g-eg energy gap and favoring the high-spin state and low T 1/2. The MO method was successfully applied to the known SCO-active iron(II) compounds involving 4-substituted 2,6-bis(pyrazol-1-yl)pyridines. A distinct positive correlation appeared in the T 1/2 versus ρ(Npy) plot. The comparison of correlation coefficients indicates that ρ(Npy) is a more reliable parameter than σp or σp + to predict a shift of T 1/2. Furthermore, this method can be more generalized by application to another known SCO family having 3-azinyl-4-p-tolyl-5-phenyl-1,2,4-triazole ligand series, where azinyl stands for a 2-azaaromatic ring. A good linear correlation was found in the T 1/2 versus ρ(NA) plot (NA is the ligating nitrogen atom in the azaaromatic ring). Finally, we will state a reason why the present treatment is competent to predict the SCO equilibrium position only by consideration on the electronic perturbation.

Introduction

Spin-crossover (SCO) is a reversible transition between low-spin (LS) and high-spin (HS) states associated with external stimuli such as heat, light, and pressure.[1−4] Iron(II) (3d6) coordination compounds are important SCO systems because SCO occurs between S = 0 dia- and S = 2 paramagnetic states, where a drastic magnetic change is expected. The six-nitrogen donor structures (i.e., FeIIN6) have been studied most extensively,[5−10] and in particular, meridional triimine-type ligands are frequently applied for this purpose. The analogy between the well-established SCO ligands 2,6-bis(pyrazol-1-yl)pyridine (1-bpp;[11] for the structural formula, see Scheme ) and 2,6-bis(pyrazol-3-yl)pyridine (3-bpp)[12] gives us a clue to the development of novel robust SCO ligands and furthermore a guide to the ligand-field engineering. We have developed the triimine ligand 2,6-bis(oxazolin-2-yl)pyridine (pybox) and its derivatives after the pioneering works by the Gao[13,14] and Halcrow groups.[15]

Scheme 1

Structural Formulas of 1-bpp, 3-bpp, Pybox, and Lazinyl (* = CH or N)

The physical properties should be predicted and controlled in a nonserendipitous manner. The SCO behavior and transition temperature of a series of [FeII(X-pybox)2](ClO4)2 were studied in connection to the substituent effect, where pybox is substituted with X at the 4-position of the pyridine ring. The introduction of substituents will bring about drastic SCO tuning through the covalent bonds, in comparison with the effects through intermolecular interaction from counter ions and/or solvent molecules incorporated. A part of the experimental data has been communicated and clarified to undergo near-room-temperature SCO with X = H, Cl, Ph, CH3O, CH3S,[16] 3-thienyl (3Th), and 4-pyridyl (4Py).[17] The magnetic susceptibility measurements were performed in solution to purge intermolecular interactions and rigid crystal lattice effects, and discussion can be focused on the substituent effect. A plot of T1/2 against the Hammett substituent constant σp[18,19] exhibited a strong correlation with a positive slope, indicating that electron-donating (-releasing) groups suppress T1/2 in this series. Here, the SCO temperature T1/2 is defined as the temperature at which equimolar fractions of the HS and LS species are present.

Five compounds with X = Me, 2-thienyl (2Th), N3, Br, and 3-pyridyl (3Py) are newly synthesized, and the SCO properties are investigated in the present work. Discussion will cover a wide range of the substituent effect in a T1/2 versus σp plot, after the known data[16,17] were superposed on the same plot (12 data points in total). As a main subject in this work, we planned to predict T1/2 by an aid of molecular orbital (MO) calculation. The atomic charge from population analysis at the pyridine nitrogen atom is assumed to be the most important for the crystal field. In this line, the MO calculation method seems to be versatile because it requires the chemical formula only. The experimental and calculation results are combined and discussed in connection with a plausible mechanism.

Finally, the MO calculation method will be justified by applying to the known SCO system, [FeII(X-1-bpp)2](BF4)2 with 4-X-substituted 1-bpp.[11] In addition, this method can be more generalized by applying to another known SCO family [FeII(Lazinyl)2(NCBH3)2] having 3-azinyl-5-phenyl-4-p-tolyl-1,2,4-triazole ligand series,[20] where azinyl stands for a 2-azaaromatic ring. We will make a comment on what method and metrics are recommended to predict T1/2 and a reason why the present treatment is competent to predict the SCO equilibrium position only by consideration on the electronic perturbation from the ligating atom.

Results

Preparation and Crystal Structure Analysis

Applying a wide variety of the substituents is an important ingredient for the purpose of proposing a magnetostructure relationship. A simple alkyl group Me was chosen as an electron-donating group and a bromine atom as an electron-withdrawing (-accepting) group. An azide group is selected as an example of a heteroatom substituent, and 2-thienyl and 3-pyridyl groups are selected as aromatic substituents. Known ligands Me-pybox[21,22] and Br-pybox[23] were prepared according to the literature method. New ligands N3-, 2Th-, and 3Py-pybox were prepared according to the established procedures[23−27] using a common intermediate Cl-pybox[26] via aromatic nucleophilic substitution reactions. The Suzuki coupling[28] was applied to heteroaromatic coupling in the preparation of 2Th- and 3Py-pybox (Scheme a).

Scheme 2

(a) Ligand Synthesis Route via the Aromatic Nucleophilic Substitution and the Suzuki Coupling Reaction; (b) Synthesis of [Fe(X-pybox)2](ClO4)2

X-pybox stands for 4-X-2,6-bis(oxazolin-2-yl)pyridine with X = MeO, Me, 3-thienyl (3Th), Ph, H, MeS, 2-thienyl (2Th), N3, Cl, Br, 3-pyridyl (3Py), 4-pyridyl (4Py).

According to the known method,[13−17] the complex formation from X-pybox and Fe(ClO4)2·6H2O was conducted in methanol (Scheme b). New complexes [Fe(X-pybox)2](ClO4)2 (X = Me, 2Th, N3, Br, 3Py) were synthesized and purified to give polycrystalline products. The elemental and spectroscopic analyses satisfied the target formula with solvation. As for the X = Me and 2Th derivatives, the X-ray diffraction study clarified the cell parameters (see the Experimental Section).

The crystal structures were successfully determined for [Fe(X-pybox)2](ClO4)2 (X = N3, Br, 3Py) at 100 K (Figure ). Selected crystallographic parameters are listed in Table . The crystals involve solvated methanol molecules. Compound [Fe(N3-pybox)2](ClO4)2·0.75MeOH crystallizes in a space group triclinic P1̅, and there are two crystallographically independent molecules in a unit cell with Z = 4. They are quite similar to each other except for the bending directions of the terminal N3 groups (Figure a,b). Regardless of the peripheral substituents, the double meridional motif is formed, affording an FeN6 coordination sphere. The Fe–N bond lengths (1.97 and 1.95 Å) on the average indicate that both molecules are assigned to an LS state at 100 K (Table ).

Figure 1

X-ray crystal structures of (a) the Fe1 molecule and (b) Fe2 molecule in [Fe(N3-pybox)2](ClO4)2·0.75MeOH, (c) [Fe(Br-pybox)2](ClO4)2·MeOH, and (d) [Fe(3Py-pybox)2](ClO4)2·2MeOH, determined at 100 K. The thermal ellipsoids are drawn at the 50% probability level. Hydrogen atoms, counter anions, and solvent molecules are omitted for the sake of clarity.

Table 1

Selected Crystallographic Parameters of [Fe(N3-pybox)2](ClO4)2·0.75MeOH, [Fe(Br-pybox)2](ClO4)2·MeOH, and [Fe(3Py-pybox)2](ClO4)2·2MeOH, Measured at 100 K

X	N3	Br	3Py
	C22.75H23Cl2FeN12O12.75	C23H24Br2Cl2FeN6O13	C34H36Cl2FeN8O14
formula weight	792.28	879.04	907.46
crystal system	triclinic	triclinic	orthorhombic
space group	P1̅	P1̅	P212121
a/Å	11.085(3)	9.678(3)	14.284(2)
b/Å	16.394(5)	9.790(3)	15.268(3)
c/Å	17.644(5)	16.669(5)	17.484(3)
α/deg	71.995(13)	85.581(15)	90
β/deg	83.901(12)	80.683(16)	90
γ/deg	83.563(13)	81.201(15)	90
V/Å3	3021.6(16)	1537.8(8)	3812.8(12)
Z	4	2	4
dcalcd/g·cm–3	1.695	1.898	1.581
μ (Mo Kα)/mm–1	0.759	3.345	0.615
no. of unique reflections	10 924	6976	8717
R(F) (I > 2σ(I))a	0.0731	0.0438	0.0659
wR(F2) (all reflections)b	0.1887	0.1038	0.0495
goodness-of-fit parameter	1.034	1.104	1.044
Flack parameter			–0.009(6)

R = ∑[|Fo| – |Fc|]/∑|Fo|.

wR = [∑w(Fo2 – Fc2)/∑wFo4]1/2.

Table 2

Fe–N Bond Lengths (d) in Å for [Fe(N3-pybox)2](ClO4)2·0.75MeOH, [Fe(Br-pybox)2](ClO4)2·MeOH, and [Fe(3Py-pybox)2](ClO4)2·2MeOH

X	N3 (Fe1)a	N3 (Fe2)a	Br	3Py
d/Å	1.989(5) (Fe1–N1)	1.967(5) (Fe2–N13)	1.974(3) (Fe1–N1)	1.965(4)(Fe1–N1)
	1.917(4) (Fe1–N2)	1.897(4) (Fe2–N14)	1.905(3) (Fe1–N2)	1.896(3) (Fe1–N2)
	1.987(5) (Fe1–N3)	1.970(5) (Fe2–N15)	1.958(3) (Fe1–N3)	1.984(4) (Fe1–N3)
	1.994(5) (Fe1–N7)	1.973(5) (Fe2–N19)	1.988(3) (Fe1–N4)	1.964(4)(Fe1–N5)
	1.922(4) (Fe1–N8)	1.905(5) (Fe2–N20)	1.900(3) (Fe1–N5)	1.891(3) (Fe1–N6)
	1.981(5) (Fe1–N9)	1.968(5) (Fe2–N21)	1.976(3) (Fe1–N6)	1.977(4) (Fe1–N7)
davg/Å	1.97	1.95	1.95	1.95

There are two crystallographically independent molecules.

The space group of the crystal of [Fe(Br-pybox)2](ClO4)2·MeOH is triclinic P1̅ with Z = 2, and one whole molecule is crystallographically independent. One methanol molecule is involved in the crystal. The Fe–N bond length (1.95 Å) indicates the LS state at 100 K.a There is a pseudo-polymorph, [Fe(Br-pybox)2](ClO4)2·3MeOH (see the Experimental Section). A crystal was manually picked up under a microscope, but selective crystallization was unsuccessful at present. This morph did not undergo SCO, so that the magnetic data must be analyzed with this finding taken into account. On the other hand, the solution experiments are not disturbed by the presence of the polymorph.

Compound [Fe(3Py-pybox)2](ClO4)2·2MeOH crystallizes in an orthorhombic P212121 space group. The Flack parameter was well reduced. There is a unique crystallographically independent molecule in a unit cell and Z = 4. The Fe–N bond lengths varying in ca. 1.90–1.98 Å suggest the LS state at 100 K. The pyridine nitrogen position does not show any disorder because the presence of hydrogen bonds between the pyridine nitrogen atoms and the hydroxyl groups from methanol molecules solvated (the Npyridyl···OMeOH distances are 2.820(6) Å for N4pyridyl···O13MeOH and 2.884(5) Å for N8pyridyl···O14MeOH).b

Magnetic Properties in the Solid State

The present compounds have a potential of SCO near room temperature, which has been confirmed by means of the SQUID magnetic susceptometry. Because the three derivatives (X = N3, Br, and 3Py) have structurally been characterized, their polycrystalline magnetic properties are reported here. As Figure a shows, the high-temperature limits of the χmT value were 3.0–4.5 cm3 K mol–1, typical of the S = 2 HS state of the iron(II) ion. The LS state is characterized with 0 cm3 K mol–1. For all the compounds investigated here, the χmT profiles on heating were not reproduced by those of the cooling process. This finding is rationalized by the presence of a crystal solvent molecule, which is assumed to escape during the heating measurements up to 400 K.

Figure 2

(a) Temperature dependence of χmT for solid specimens [Fe(N3-pybox)2](ClO4)2·0.75MeOH, [Fe(Br-pybox)2](ClO4)2·MeOH, and [Fe(3Py-pybox)2](ClO4)2·2MeOH. (b) Temperature dependence of the HS molar fraction, γHS, in acetone solution of [Fe(X-pybox)2](ClO4)2 (X = Me, 2Th, N3, Br, 3Py).

The χmT value of the N3 derivative is practically null below 150 K (Figure a), but above 320 K, it reached the HS level, indicating that this specimen underwent SCO. The heating process showed a two-step profile, which is in good agreement with the presence of two crystallographically independent molecules. The SCO temperatures were ca. 220 K for the first gradual SCO and 330 K for the second abrupt SCO. On cooling from 400 K, the SCO behavior took place in a single step and broad manner. The high-temperature χmT value of the Br derivative is practically saturated to reach a half level of the theoretical HS value. A possible reason for this behavior is the coexistence of polymorphic crystals of LS, as clarified by the crystallographic analysis. Owing to the crystal breakdown, details are unknown in the high-temperature phase. A relatively abrupt χmT jump was recorded at 310 K on heating. The measurements on cooling almost maintained the room-temperature χmT value. The 3Py derivative exhibited an abrupt χmT jump at 340 K on heating and an incomplete decrease on cooling.

The SCO temperatures in the heating process has been recorded as T1/2(N3) < T1/2(Br) < T1/2(3Py). The electron-withdrawing character has the same order in the σp scale:[18,19] σp(N3) < σp(Br) < σp(3Py). However, the analysis on the solid-state specimens involves approximation because the various space groups, molecular arrangements, solvation and desolvation effects, and coordination structure distortions were found, but such factors are neglected here. The structural distortion is known to regulate T1/2 to a considerable extent.[16,30−32] The solution experiments will afford a better venue for the discussion on the substituent effect.

Magnetic Properties in Solution

The solution magnetic susceptibilities in acetone were acquired on a SQUID susceptometer. Details of the experimental method have been described elsewhere.[16] Although the SCO profiles were all gradual with a similar shape, an appreciable difference in T1/2 was recorded (Figure b). The T1/2 values are determined as 220, 260, 215, 280, and 270 K for X = Me, 2Th, N3, Br, and 3Py, respectively. The trend observed seems to be approximately related to the electron-accepting/-donating properties. Namely, electron-donating substituents (X = Me) stabilized the HS state, whereas electron-withdrawing substituents (X = Br, 3Py) stabilized the LS state, which is qualitatively the same as that of the solid-state experiments.

The results on 12 derivatives in total[16,17] are accumulated in Figure a, and the relation between T1/2 and σp is surveyed to establish a structure–SCO property relationship. Actually, the plot displays a strong positive correlation. A plot using the alternative substituent constant σp+ suitable for substituent effects to a cationic center exhibits a positive correlation as well (Figure b). The R2 correlation coefficients are comparable; 0.780 for σp and 0.768 for σp+. The solution data have no structural information, but we suppose that the structural distortion factor would be negligible because of the vast conformational freedom. Accordingly, the solution experiments seem to be more appropriate than the solid-state experiments to assess substituent effects. As previously pointed out,[16,17] electron-donating substituents suppress T1/2 and electron-withdrawing substituents raise T1/2.

Figure 3

SCO temperature T1/2 for [Fe(X-pybox)2](ClO4)2 in acetone solution as a function of the Hammett substituent constants (a) σp and (b) σp+. Dashed lines represent the best linear fit.

The substituent dependence of T1/2 in [Fe(X-pybox)2](ClO4)2 in acetone solution can be verified with the trend known for T1/2 in [Fe(X-1-bpp)2](BF4)2 in acetone or nitromethane solution.[30]Figure displays the plot using the difference of T1/2 of an X-substituted derivative from that of the mother X = H compound as ΔT1/2. Because the 4-X-1-bpp and 4-X-pybox series contain the double meridional chelation in common, the steric effects are similar to each other. A relationship is formulated as eq and superposed in Figure . The slope close to unity implies that essentially the same substituent effects are operative in both. Moreover, the comparison study indicates that the counter anion dependence or solvent dependence could be negligible in the solution experiments.

Figure 4

Plot of ΔT1/2 for [Fe(X-pybox)2](ClO4)2 vs ΔT1/2 for [Fe(X-1-bpp)2](BF4)2.[30] A dashed line represents the best linear fit.

Density-Functional-Theory (DFT) MO Calculation

We investigated the theoretical MO calculation on the series of X-pybox ligands and performed DFT calculation in the Gaussian 03 package.[33] The B3LYP Hamiltonian[34] with the 6-311+G(2d,p) basis set was selected. To reproduce the solution data, the geometry optimization is required, and only the ligand portion (X-pybox) was subjected to calculation. This method would be welcomed for ab initio prediction of T1/2 when only the chemical formula of X-pybox is given. There have been several known DFT functionals, B3LYP* and TPSSh, for example,[35,36] suitable for the analysis of the SCO behavior on the whole coordination compounds. In the present work, the metal-free organic ligand is subjected to the calculation, so that the choice of B3LYP/6-311+G(2d,p) would be reasonable.

Because the crystal field effect is sensitive to the charge or electron population of the ligating atom, the Mulliken atomic charge at the pyridine nitrogen seems to be important. However, it has been pointed out that the natural population analysis gives better results than the Mulliken population analysis when precise basis sets are chosen.[37,38] Therefore, the natural orbital population analysis was performed with the natural bond orbital (NBO) option.[39]

Attention was paid to a configuration problem around the biaryl-type Csp–Csp single bonds. The pybox ligands and related chelatable ligands have a syn conformation in the coordination compounds, though the most stable form often is nearly anti[20] (Scheme ). To reproduce the conformation after complexation, the geometry optimization started from the syn form, eventually giving a configuration with a local potential minimum close to that of an ideal syn form.

Scheme 3

Conformational Isomers in Di- and Triimines

Syn and anti configurations are classified with respect to the ligating nitrogen atoms

The natural charge at the pyridine nitrogen atom, ρ(Npy), and the energy levels of the pyridine C=N π orbital, E(πpy), and the pyridine lone-pair n orbital, E(npy), were obtained and listed in Table . These values are varied depending on the substituent, and the T1/2 versus ρ(Npy) plot (Figure a) displays a positive correlation. The molecular electrostatic potential surfaces are demonstrated in Figure b,c for MeO- and 4Py-pybox, respectively, under the same isosurface conditions. The electronegative character can be seen around heteroatoms, and in particular, the pyridine N atom in the former is more negative than that of the latter. To argue the usefulness of ρ(Npy), correlations were checked by drawing the plots of σp and σp+ versus ρ(Npy) (Figure ). Positive correlations were found with R2 = 0.410 and 0.540, respectively, and it is notable that these coefficients are smaller than that of the T1/2 versus ρ(Npy) plot (0.734 in Figure a).

Table 3

Calculated ρ(Npy), E(πpy), E(npy), Experimental SCO Temperatures in Acetone (T1/2), and the Hammett σp Constant

	ρ(Npy)	E(πpy)/eV	E(npy)/eV	T1/2 (soln.)/K	σp
MeO	–0.37848	–0.30214	–0.33584	170[16]	–0.27
Me	–0.35957	–0.30197	–0.33839	220	–0.17
3Th	–0.35748	–0.30387	–0.34099	240[17]	–0.02
Ph	–0.35540	–0.30355	–0.34059	240[16,17]	–0.01
H	–0.35251	–0.30621	–0.34120	260[16]	0
MeS	–0.36901	–0.30408	–0.33930	210[16]	0.00
2Th	–0.35816	–0.30411	–0.34140	260	0.05
N3	–0.36376	–0.31199	–0.34715	215	0.08
Cl	–0.36322	–0.31510	–0.34947	270[16]	0.23
Br	–0.35422	–0.31533	–0.34953	280	0.23
3Py	–0.35275	–0.30873	–0.34538	270	0.25
4Py	–0.34978	–0.31135	–0.34800	310[17]	0.44

Figure 5

(a) Plot of T1/2 for [Fe(X-pybox)2](ClO4)2 in acetone solution vs ρ(Npy) from the DFT natural population analysis. A dashed line represents the best linear fit. (b,c) Molecular electrostatic potential from −0.10 (red) to +0.14 (blue) mapped on the total electron density surface at 0.02 e au–3 for X = MeO (b) and 4Py (c).

Figure 6

Plots of (a) σp vs ρ(Npy) and (b) σp+ vs ρ(Npy) for X-pybox. Dashed lines represent the best linear fit.

The matching of the energy levels has also been taken into consideration when the ligand MOs interact with iron(II) orbitals. Figure illustrates that T1/2 had a negative correlation with E(πpy) and E(npy), though the data points were relatively scattered. It is not surprising that the three calculated values, ρ(Npy), E(πpy), and E(npy), are related to each other because the rich electron population or negatively large ρ(Npy) lifts E(πpy) and E(npy).

Figure 7

Plots of T1/2 for solution [Fe(X-pybox)2](ClO4)2 vs E(πpy) (a) and vs E(npy) (b) from the DFT calculation. Dashed lines represent the best linear fit.

Discussion

Substituent Effect from an Empirical Approach

To establish a prediction method of SCO T1/2, the solution data were analyzed in connection with the substituent effect. We plotted the solution T1/2 against σp and σp+ (Figure ), where the data on the known derivatives[16,17] are included. We can find a considerably good positive correlation, which indicates that electron-donating substituents suppress T1/2, whereas electron-withdrawing substituents raise T1/2. An empirical relationship equation is expressed as eq (a dashed line superposed in Figure a). This method is satisfactory for the prediction of T1/2 when the substituent constant is available.

Substituent Effect from the DFT Calculation Approach

Very recently, Brooker and co-workers reported the relation between the NMR chemical shift of the ligating nitrogen atom and SCO behavior.[20] The chemical shifts were obtained by means of spectroscopic and DFT computational methods. In that research, the 15N NMR spectroscopy was applied, but such methods are not easily performed. Novikov et al. reported another way using a paramagnetic shift of 1H NMR on the ligand protons.[40] The pseudocontact shift depends on the calculated mutual geometry between the 1H nucleus and unpaired electrons, and the contact distributions are given from calculation. Both experimental and computational techniques are needed.

On the other hand, the MO calculation on the ligand is an easier way to approach SCO analysis. The electron density is caused by the substituents, as usually formulated with the σp and finally proven with the MO calculation. This method has an intervening parameter (σp) in the discussion of the relation between the SCO temperature and the substituent effect. However, it seems to be more convenient if the SCO temperature is directly predicted from the calculation. Furthermore, as the calculation results on [Fe(X-pybox)2](ClO4)2 (Figure a) show, the R2 correlation coefficient between T1/2 and ρ(Npy) is better than those of σp versus ρ(Npy) and σp+ versus ρ(Npy). A direct correlation using ρ(Npy) seems to be a more reliable and convenient way to predict T1/2. In short, the substituent constants are not needed any more. A correlation equation is empirically described with eq (a dashed line superposed in Figure ).

Application to the 1-bpp Compounds

The present work provides a prediction method for T1/2 only from a chemical formula, enabling us to apply the DFT method to well-known SCO-active [FeII(X-1-bpp)2](BF4)2 involving 4-substituted 1-bpp,[30] for example. The calculation protocol was the same as that of the 4-X-pybox series, and the result is displayed in Figure . An astonishingly good correlation (R2 = 0.980) appeared, as shown with a dashed line. The best fit line is formulated with eq .c

Figure 8

Plot of T1/2 for [Fe(X-1-bpp)2](BF4)2 in acetone or nitromethane solution[30] vs ρ(Npy) from the DFT natural population analysis. A dashed line represents the best linear fit.

Figure shows the correlations between σp versus ρ(Npy) and σp+ versus ρ(Npy). These plots still have a strong correlation but relatively scattered data points, as indicated with R2 = 0.831 and 0.932, respectively. As Halcrow et al. pointed out, the σp+ constants are slightly better than the σp ones to describe the results of the 1-bpp complexes.[30] However, similar to the case of the X-pybox compounds, the best way in the substituent effect analysis is to utilize the ρ(Npy) parameter. Furthermore, the present method is inexpensive in comparison with the usual theoretical treatments including the calculation on the whole coordination compounds.[20,30,35,36,41] The SCO temperature will be discussed in a generalized manner according to the present method, regardless of the different ligand platforms, such as pybox and 1-bpp.

Figure 9

Plots of (a) σp vs ρ(Npy) and (b) σp+ vs ρ(Npy) for X-1-bpp. Dashed lines represent the best linear fit.

Application to the Azine-Substituted Triazole Compounds

To pursue further generalization of the present method, we expanded the calculation work to another system, the azine (heteroaromatic ring) dependence on T1/2 of [FeII(Lazinyl)2(NCBH3)2] (azinyl = 2-pyridyl, 3-pyridazyl, 4-pyrimidinyl, 2-pyrimidinyl, pyrazinyl).[20] The electronic population on the ligating nitrogen atom (NA; see Scheme ) varies with respect to the heteroaromatic skeleton itself and not as a substituent on the ring. The natural charge on the ligating azinyl nitrogen atom, ρ(NA), was calculated at the same level as those of the calculations on the 4-X-pybox and 4-X-1-bpp families. In this calculation, the starting structure is restricted to a syn isomer (Scheme ), like the two previous cases. Figure shows a calculation result and clarifies a distinct positive correlation between T1/2 and ρ(NA). The best fit line is expressed as eq with R2 = 0.963.

Figure 10

Plot of T1/2 for [FeII(Lazinyl)2(NCBH3)2] in CDCl3 solution[20] vs ρ(NA) from the DFT natural population analysis. A dashed line represents the best linear fit.

Apparently, a similar result was obtained here, when compared with the plot of T1/2 versus δ(NA) (15N NMR chemical shift) reported by Brooker et al.,[20] but there seems to be a different meaning. Both the crystal field effect and NMR chemical shift originate in the atomic charge in common. However, the 15N chemical shift in the heteroaromatic ring involves an additional deshielding effect from an aromatic ring current.[42] After the deshielding effect is cancelled by accident in the comparison work, a linear relation in the T1/2 versus δ(NA) plot would appear. Therefore, the T1/2 versus ρ(NA) plot is concluded to demonstrate a straightforward causal relationship. In short, the NMR chemical shift is not needed in the discussion on T1/2.

The charge at the ligating nitrogen atom is concluded to be essential in discussion on SCO. The present DFT MO method would be applicable to other SCO-active FeII complexes carrying substituted pyridine-based N-donor ligands,[43] and even for different ligand systems involving azaaromatic rings.[44] The electron population analysis at the ligating nitrogen atom in a metal-free ligand by means of the DFT MO method provides one of the most convenient and powerful ways to explain and predict the SCO temperature.

Mechanism

The substituent effect described here is explained in terms of a perturbation from the substituent. When an electron-donating group is introduced to the pyridine ring, the t2g energy level is raised, and the eg–t2g energy gap (Δoct) becomes narrower, favoring the HS state and low T1/2 (Figure ). With an electron-withdrawing group, it is vice versa. The σp and σp+ constants contain two major contributions: one is the inductive effect along the σ electron system and another is the mesomeric or resonance effect through the π electron system. In general, the latter becomes obvious when the 2- or 4-position in the six-membered aromatic ring is substituted. Therefore, the σp and σp+ constants are adequate empirical representatives to comprehend the substituent effect in the SCO compounds involving 4-X-substituted pyridine-based ligands.

Figure 11

Model explaining the substituent effect on T1/2 from the d−π interaction. A π-electron-rich heteroaromatic ring stabilizes the iron(II) HS state.

The DFT calculation supports the proposed mechanism. Figures a, 8, and 10 highlight the role of the electron-withdrawing or -donating nature of the substituent group or heteroaromatic skeleton, and the charge of the ligating nitrogen atom is tuned. In the case of the Lazinyl series, the substituent effect belongs to the heteroaromatic ring itself. The logic never changes; the electron population on the ligating nitrogen atom increases or decreases, depending on the electronic structure of the heteroaromatic rings.

The dσ/dπ symmetry consideration tells us that the t2g energy level is perturbed through the dπ–pπ interaction from the ligand π electron system (Figure ). In addition to the orbital overlap, the matching of the energy level of each orbital has also been taken into account. The electron density at each atom is responsible for the shift of the energy levels of relevant MOs. Although correlations appear for both E(πpy) and E(npy) to T1/2, E(πpy) has a higher energy level than E(npy), and there seems to be an advantage in the interaction between the πpy and iron(II) t2g orbitals over the interaction between the npy and iron(II) eg orbitals. In other words, the dπ–pπ interaction serves a major path in tuning T1/2.d

Finally, we have to make a comment on the validity that a purely enthalpy term defined with the Coulombic interaction regulates T1/2 while an entropy term is set aside. The transition temperature and behavior depend on entropy as well as enthalpy; namely, ΔG = ΔH – T1/2ΔS and the equilibrium condition with ΔG = 0. According to the treatment of the linear free-energy relationships,[45] a constant ΔS regardless of the substituents would guarantee that the change of ΔG is connected to the change of ΔH. The entropy difference originates in the spin state and vibrational freedom.[46] In solution SCO, organization of solvation molecules may also contribute to the entropy term. The SCO in a series of substituent variations with the unique mechanism tends to have a constant ΔS. Even if ΔS is variable, linearly correlated ΔH and ΔS would be an alternative condition.[45] Kepp and co-workers[35] suggested the enthalpy–entropy compensation[47] in many SCO materials with various structures from the survey research and theoretical consideration. Therefore, the discussion on ΔG in the SCO can approximately be replaced with the argument of ΔH.

Conclusions

We investigated the SCO behavior on 12 derivatives [FeII(X-pybox)2](ClO4)2, and the substituent effect was analyzed with the DFT MO calculation performed on the metal-free ligands. The geometry was optimized and the natural population analysis gave the ligating pyridine nitrogen charge ρ(Npy). The T1/2 versus ρ(Npy) plot has a better correlation than the T1/2 versus σp or σp+ plot. When only the information of chemical formula of the ligand is available or a substituent constant is unavailable, the DFT calculation method is a promising option. Actually, the present method was applied to another SCO-active family, [FeII(X-1-bpp)2](BF4)2, which possesses an analogous meridional coordination structure from 4-substituted pyridine-based ligands. An excellent correlation appeared in the T1/2 versus ρ(Npy) plot.

This calculation method was diverted to the triazole ligand system substituted with a heteroaromatic ring, [FeII(Lazinyl)2(NCBH3)2], to generalize this method. The heteroaromatic groups were varied in pyridine, pyridazine, pyrimidine, and pyrazine rings. The charge at the ligating nitrogen atom (ρ(NA)) is again concluded to be an essential parameter in the discussion on the T1/2 shift.

The electron population analysis at the ligating nitrogen atom in a metal-free ligand by means of the DFT MO method provides one of the most convenient and powerful ways to explain and predict the SCO temperature.

Experimental Section

Materials

Caution! The perchlorate salts should be handled with care, though we have not yet encountered any hazard.

Me-,[21,22] Br-,[23] and Cl-pybox[26] were prepared according to the literature methods.

N3-pybox: a mixture of Cl-pybox (0.746 g; 2.96 mmol), NaN3 (1.932 g; 29.7 mmol), and N,N-dimethylformamide (10 mL) was heated at 65 °C for 4 h. Anhydrous conditions are required to avoid hydrolysis of the oxazoline rings. After the addition of aq NaHCO3, the organic layer was extracted with dichloromethane and hexane. The organic layer was washed with brine, dried over anhydrous MgSO4, filtered, and concentrated under reduced pressure. The resultant colorless powder was collected on a filter and washed with tetrahydrofuran. The yield was 0.575 g (2.27 mmol; 77%). mp 178–179 °C. MS (ESI+) m/z: 281.1 (M + Na+). 1H NMR (500 MHz, CDCl3): δ 7.80 (2H, s), 4.55 (4H, t, J = 9.8 Hz), 4.13 (4H, t, J = 9.8 Hz). 13C NMR (126 MHz, CDCl3): δ 163.09, 150.61, 148.43, 115.72, 68.62, 55.13. IR (neat, ATR): 564, 701, 868, 950, 1141, 1233, 1400, 1595, 1633, 2109, 2934 cm–1.

2Th-pybox: a mixture of Cl-pybox (0.394 g; 1.57 mmol), 2-thienyl boronic acid (0.198 g; 1.55 mmol), Pd(PPh4)3 (0.0232 g; 0.0201 mmol), and K2CO3 (0.890; 6.44 mmol) in dry N,N-dimethylformamide (4 mL) was heated at 130 °C for 3 days. After the addition of water, the organic layer was separated with ethyl acetate. The organic layer was washed with aq. NaHCO3 and brine, dried over anhydrous MgSO4, filtered, and concentrated under reduced pressure. Recrystallization from dichloromethane–hexane gave 2Th-pybox as a colorless solid. The yield was 0.210 g (46%). mp 185 °C (decomp.). MS (ESI+) m/z: 320.0 (M + Na+). 1H NMR (500 MHz CDCl3): δ 8.36 (2H, s), 7.67 (1H, d, J = 1.2, 4.0 Hz), 7.48 (1H, d, J = 1.2, 5.2 Hz), 7.16 (1H, t, J = 4.0, 5.2 Hz), 4.57 (4H, t, J = 9.8 Hz), 4.15 (4H, t, J = 9.8 Hz). 13C NMR (126 MHz, CDCl3): δ 163.63, 147.49, 143.28, 139.87, 128.76, 128.46, 126.68, 121.42, 68.51, 55.19. IR (neat, ATR): 539, 721, 940, 1104, 1191, 1384, 1595, 1634, 2936 cm–1.

3Py-pybox: a mixture of Cl-pybox (0.762 g; 3.03 mmol), 3-pyridineboronic acid (0.409 g; 3.32 mmol), Pd(PPh4)3 (0.0448 g; 0.0369 mmol), and K2CO3 (1.629; 11.8 mmol) in dry N,N-dimethylformamide (8 mL) was heated at 130 °C for 2 days. After the addition of water, the organic layer was separated with dichloromethane. The organic layer was washed with aq. NaHCO3 and brine, dried over anhydrous MgSO4, filtered, and concentrated under reduced pressure. Recrystallization from dichloromethane–hexane gave 3Py-pybox as a pale yellow solid. The yield was 0.376 g (42%). mp 220–222 °C. MS (ESI+) m/z: 317.1 (M + Na+), 295.1 (M + H+). 1H NMR (500 MHz CDCl3): δ 9.00 (1H, d, J = 1.4 Hz), 8.73 (1H, dd, J = 1.4, 5.0 Hz), 8.72 (2H, s), 8.05 (1H, td, J = 1.4, 7.7 Hz), 7.45 (1H, dd, J = 5.0, 7.7 Hz), 4.59 (4H, t, J = 9.7 Hz), 4.18 (4H, t, J = 9.7 Hz). 13C NMR (126 MHz, CDCl3): δ 163.50, 150.86, 148.26, 147.62, 147.10, 134.70, 132.54, 123.99, 123.28, 68.62, 55.17. IR (neat, ATR): 593, 652, 709, 769, 808, 916, 1108, 1155, 1246, 1309, 1401, 1598, 1636, 2873 cm–1.

Complexes [Fe(X-pybox)2](ClO4)2·X(solv.) (X = Me, 2Th, N3, Br, 3Py) were synthesized in a manner similar to those of [Fe(X-pybox)2](ClO4)2.[13−17] The elemental analysis was performed by a usual combustion method (PerkinElmer Series II CHNS/O 2400) on evacuated specimens, and accordingly, the solvation content was often different from those of the X-ray crystallographic analysis. The purity of the target complexes was confirmed with the elemental, spectroscopic, and magnetic analyses.

[Fe(Me-pybox)2](ClO4)2: yield 22%. mp 197 °C (decomp.). IR (neat, ATR): 524, 620, 914, 1071, 1268, 1374, 1589 cm–1. Anal. Calcd for C24H26Cl2FeN6O12: C, 40.19; H, 3.65; N, 11.72%. Found: C, 39.87; H, 3.67; N, 11.53%.

[Fe(2Th-pybox)2](ClO4)2·H2O: yield 86%. mp 231 °C (decomp.). IR (neat, ATR): 620, 725, 914, 1065, 1270, 1376, 1482, 1585 cm–1. Anal. Calcd for C32H28Cl2FeN6O12S2·1.5H2O: C, 40.92; H, 3.32; N, 9.54; S, 7.28%. Found: C, 40.80; H, 3.24; N, 9.64; S, 7.02%.

[Fe(N3-pybox)2](ClO4)2: yield 41%. mp 179 °C (decomp.). Anal. Calcd for C22H20Cl2FeN12O12: C, 34.26; H, 2.61; N, 21.79%. Found: C, 34.06; H, 2.41; N, 21.31%. No IR spectrum was measured because of avoiding hazard.

[Fe(Br-pybox)2](ClO4)2·H2O as a mixture of two pseudo-polymorphs: yield 18%. mp 220 °C (decomp.). IR (neat, ATR): 511, 619, 761, 916, 1067, 1272, 1386, 1487, 1574 cm–1. Anal. Calcd for C23H24Br2Cl2FeN6O12·MeOH: C, 40.92; H, 3.32; N, 9.54; S, 7.28%. Found: C, 40.80; H, 3.24; N, 9.64; S, 7.02%.

[Fe(3Py-pybox)2](ClO4)2·H2O: yield 45%. mp 223 °C (decomp.). IR (neat, ATR) 619, 708, 916, 1069, 1268, 1372, 1485, 1586 cm–1. Anal. Calcd for C32H28Cl2FeN8O12·H2O: C, 44.62; H, 3.51; N, 13.01%. Found: C, 44.30; H, 3.47; N, 12.92%.

Crystallographic Study

X-ray diffraction data of [Fe(X-pybox)2](ClO4)2·nMeOH (X = N3, Br, 3Py) were collected on a Saturn70 CCD diffractometer with graphite monochromated Mo Kα radiation (λ = 0.71073 Å). The structures were directly solved by a heavy-atom method and expanded using Fourier techniques in the crystalstructure program package.[48] The parameters were refined in the SHELXL program.[49] Numerical absorption correction was used. Hydrogen atoms were placed at calculated positions, and their parameters were refined as “riding.” The thermal displacement factors of nonhydrogen atoms were refined anisotropically. Selected crystallographic data are given in Table , and important bond lengths are listed in Table . A pseudo-polymorph of the X = Br derivative was found. Selected crystallographic parameters are as follows: [Fe(Br-pybox)2](ClO4)2·3MeOH, monoclinic P21/c, a = 9.627(2), b = 17.371(4), c = 20.100(5) Å, b = 96.333(11)°, V = 3340.8(13) Å3, Z = 4, d = 1.875 g·cm–3, μ(Mo Kα) = 3.090 mm–1, R(F) (I > 2σ(I)) = 0.0661, wR(F2) (all data) = 0.1667, G.O.F. = 1.031, T = 100 K for 7649 reflections. A crystallographic study on the X = Me and 2Th derivatives clarified the cell parameters as follows: [Fe(Me-pybox)2](ClO4)2·nMeOH, tetragonal, a = 20.433(7), b = 20.037(5) Å, V = 8366(4) Å3 and Z = 8 at 100 K; [Fe(2Th-pybox)2](ClO4)2·0.75MeOH, triclinic, P1̅, a = 17.084(4), b = 17.905(4), c = 24.247(7) Å, α = 90.225(10), β = 99.720(11), γ = 95.326(9)°, V = 7278(3) Å3, and Z = 8 at 100 K.

CCDC numbers 1836313, 1836314, and 1836315 for [Fe(X-pybox)2](ClO4)2·nMeOH (X = N3, Br, 3Py, respectively) include the experimental details and geometrical parameter tables.

Magnetic Study

Magnetic susceptibilities of polycrystalline and solution forms of [Fe(X-pybox)2](ClO4)2·X(solv.) (X = Me, 2Th, N3, Br, 3Py) were measured on a Quantum Design MPMS-XL7 SQUID magnetometer with a static field of 0.5 T. Solution specimens were mounted in a sealed NMR sample tube. Details have been described elsewhere.[16]

DFT Calculation Study

DFT MO calculation was performed with the GAUSSIAN 03 package[33] running on a Windows PC. The geometry was optimized, and the self-consistent-field (SCF) energy was converged on the B3LYP Hamiltonian[50] with the 6-311+G(2d,p) basis set. The criterion of the SCF energy convergence was set to 10–5 au. The NBO analysis was performed with the NBO option executing the Gaussian NBO version 3.1.[39] The natural charge at the pyridine nitrogen atom, ρ(Npy), and the MO energy levels, E(πpy) and E(npy), could be read out from the text output.