High-Temperature Cooperative Spin Crossover Transitions and Single-Crystal Reflection Spectra of [FeIII(qsal)2](CH3OSO3) and Related Compounds

: New Fe(III) compounds from qsal ligand, [Fe(qsal) 2 ](CH 3 OSO 3 ) ( 1 ) and [Fe(qsal) 2 ] (CH 3 SO 3 ) · CH 3 OH ( 3 ), along with known compound, [Fe(qsal) 2 ](CF 3 SO 3 ) ( 2 ), were obtained as large well-shaped crystals (Hqsal = N -(8-quinolyl)salicylaldimine). The compounds 1 and 2 were in the low-spin (LS) state at 300 K and exhibited a cooperative spin crossover (SCO) transition with a thermal hysteresis loop at higher temperatures, whereas 3 was in the high-spin (HS) state below 300 K. The optical conductivity spectra for 1 and 3 were calculated from the single-crystal reﬂection spectra, which were, to the best of our knowledge, the ﬁrst optical conductivity spectra of SCO compounds. The absorption bands for the LS and HS [Fe(qsal) 2 ] cations were assigned by time-dependent density functional theory calculations. The crystal structures of 1 and 2 consisted of a common one-dimensional (1D) array of the [Fe(qsal) 2 ] cation, whereas that of 3 had an unusual 1D arrangement by π -stacking interactions which has never been reported. The crystal structures in the high-temperature phases for 1 and 2 indicate that large structural changes were triggered by the motion of counter anions. The comparison of the crystal structures of the known [Fe(qsal) 2 ] compounds suggests the signiﬁcant role of a large non-spherical counter-anion or solvate molecule for the total lattice energy gain in the crystal of a charged complex. Anal. C


Introduction
Spin crossover (SCO) between a high-spin (HS) and low-spin (LS) state in a transition metal coordination compound is one of the molecular switching phenomena responsive to various external stimuli such as temperature, pressure, light, magnetic field, and chemicals. Significant attention has been attracted to SCO phenomena in the wide range of fields of chemical and physical sciences [1][2][3][4][5]. The SCO switches not only a spin-state but also color and coordination structure in a transition metal

Physical Measurements
Variable temperature direct current magnetic susceptibilities of polycrystalline samples were measured on a Quantum Design MPMS-XL magnetometer under a field of 0.5 T in the temperature range of 2−320 K. The oven option was used for the measurement in the temperature range of 300−450 K. The magnetic susceptibilities were corrected for diamagnetic contributions estimated by Pascal constants [45].
The Mössbauer spectra were recorded on a constant acceleration spectrometer with a source of 57 Co/Rh in the transmission mode. The measurements at low temperature were performed with a closed-cycle helium refrigerator (Iwatani Co., Ltd., Japan). Velocity was calibrated by using an α-Fe standard. The obtained Mössbauer spectra were fitted with asymmetric Lorentzian doublets by the least squares fitting program (MossWinn).
The polarized reflection spectrum was recorded using an infrared microscope Spectratech IR-Plan combined with an FT-IR spectrometer Thermo Nicolet NEXUS 870 for 5000−12000 cm −1 , and a multichannel visible spectrograph Atago Macs 320 for 11000−33000 cm −1 . The spectrum was obtained from the developed plane of the thin plate crystal. The crystal orientation was adjusted so that the infrared reflectivity was maximized for the plane polarized light. The optical conductivity spectrum was calculated from the reflection spectrum by the Kramers-Kronig analysis.

Synthesis of Compounds 1-3
All chemicals were purchased and used without further purification. [Fe(qsal) 2 ]Cl·1.5H 2 O was prepared according to the literature [7].

Physical Measurements
Variable temperature direct current magnetic susceptibilities of polycrystalline samples were measured on a Quantum Design MPMS-XL magnetometer under a field of 0.5 T in the temperature range of 2−320 K. The oven option was used for the measurement in the temperature range of 300−450 K. The magnetic susceptibilities were corrected for diamagnetic contributions estimated by Pascal constants [45].
The Mössbauer spectra were recorded on a constant acceleration spectrometer with a source of 57 Co/Rh in the transmission mode. The measurements at low temperature were performed with a closed-cycle helium refrigerator (Iwatani Co., Ltd., Japan). Velocity was calibrated by using an α-Fe standard. The obtained Mössbauer spectra were fitted with asymmetric Lorentzian doublets by the least squares fitting program (MossWinn).
The polarized reflection spectrum was recorded using an infrared microscope Spectratech IR-Plan combined with an FT-IR spectrometer Thermo Nicolet NEXUS 870 for 5000−12000 cm −1 , and a multichannel visible spectrograph Atago Macs 320 for 11000−33000 cm −1 . The spectrum was obtained from the developed plane of the thin plate crystal. The crystal orientation was adjusted so that the infrared reflectivity was maximized for the plane polarized light. The optical conductivity spectrum was calculated from the reflection spectrum by the Kramers-Kronig analysis.

Crystal Structure Determinations of 1-3
A crystal was mounted on a roll of 15 µm thick polyimide film by using the Araldite TM adhesive. A Nihon Thermal Engineering nitrogen gas flow temperature controller was used for the temperature variable measurements. All data were collected on a Bruker APEX II CCD area detector with monochromated Mo-Kα radiation generated by a Bruker Turbo X-ray Source coupled with Helios multilayer optics. All data collections were performed using the APEX2 crystallographic software package (Bruker AXS). The data were collected to a maximum 2θ value of 55.0 • . A total of 720 oscillation images were collected. The APEX3 crystallographic software package (Bruker AXS) was used to determine the unit cell parameters. Data were integrated by using SAINT. Numerical absorption correction was applied by using SADABS. The structures at all temperatures were solved by direct methods and refined by full-matrix least-squares methods based on F 2 by using the SHELXTL program. All non-hydrogen atoms were refined anisotropically. Hydrogen atoms were generated by calculation and refined using the riding model. CCDC 1891471-1891477 contains the supplementary crystallographic data for this paper. These data can be obtained free of charge via http://www.ccdc.cam.ac.uk/conts/retrieving.html (or from the CCDC, 12 Union Road, Cambridge CB2 1EZ, UK; Fax: +44 1223 336033; E-mail: deposit@ccdc.cam.ac.uk).

Density Functional Theory (DFT) Calculations
All theoretical calculations were performed using the Gaussian 09 program package [46]. The atomic coordinates for the LS and HS states of the [Fe(qsal) 2 ] cation were taken from the single crystal structural data for 1 and 3, respectively. All geometry optimization and frequency calculations of the compounds were carried out at the B3LYP functional [47,48]. The Wachters-Hay basis set [49,50] for Fe atoms and the 6-31+G(d) basis set [51] for H, C, O, and N atoms were used. No imaginary frequencies were found in the optimized structures.

Synthesis of Compounds 1-3
Compounds 1-3 were prepared by the metathesis reaction between [Fe(qsal) 2 ]Cl·1.5H 2 O and corresponding anion salts using the diffusion methods. Compound 1 was obtained as very large rhombic platelets (Figure 2a), whereas compounds 2 and 3 gave relatively large parallelogrammatic and rhombic platelets, respectively (Figure 2b,c). The compositions of 1 to 3 were confirmed by microanalyses and crystal analyses described below.

Crystal Structure Determinations of 1-3
A crystal was mounted on a roll of 15 μm thick polyimide film by using the Araldite TM adhesive. A Nihon Thermal Engineering nitrogen gas flow temperature controller was used for the temperature variable measurements. All data were collected on a Bruker APEX II CCD area detector with monochromated Mo-Kα radiation generated by a Bruker Turbo X-ray Source coupled with Helios multilayer optics. All data collections were performed using the APEX2 crystallographic software package (Bruker AXS). The data were collected to a maximum 2θ value of 55.0°. A total of 720 oscillation images were collected. The APEX3 crystallographic software package (Bruker AXS) was used to determine the unit cell parameters. Data were integrated by using SAINT. Numerical absorption correction was applied by using SADABS. The structures at all temperatures were solved by direct methods and refined by full-matrix least-squares methods based on F 2 by using the SHELXTL program. All non-hydrogen atoms were refined anisotropically. Hydrogen atoms were generated by calculation and refined using the riding model. CCDC 1891471-1891477 contains the supplementary crystallographic data for this paper. These data can be obtained free of charge via http://www.ccdc.cam.ac.uk/conts/retrieving.html (or from the CCDC, 12 Union Road, Cambridge CB2 1EZ, UK; Fax: +44 1223 336033; E-mail: deposit@ccdc.cam.ac.uk).

Density Functional Theory (DFT) Calculations
All theoretical calculations were performed using the Gaussian 09 program package [46]. The atomic coordinates for the LS and HS states of the [Fe(qsal)2] cation were taken from the single crystal structural data for 1 and 3, respectively. All geometry optimization and frequency calculations of the compounds were carried out at the B3LYP functional [47,48]. The Wachters-Hay basis set [49,50] for Fe atoms and the 6-31+G(d) basis set [51] for H, C, O, and N atoms were used.

Synthesis of Compounds 1-3
Compounds 1−3 were prepared by the metathesis reaction between [Fe(qsal)2]Cl·1.5H2O and corresponding anion salts using the diffusion methods. Compound 1 was obtained as very large rhombic platelets (Figure 2a), whereas compounds 2 and 3 gave relatively large parallelogrammatic and rhombic platelets, respectively (Figure 2b,c). The compositions of 1 to 3 were confirmed by microanalyses and crystal analyses described below.

Magnetic Susceptibility for 1-3
The temperature variations of magnetic susceptibility for compounds 1-3 are shown in Figure 3. The χ M T value for compound 1 at 300 K was 0.56 cm 3 K mol −1 , suggesting 1 was almost in the LS state. Below 300 K, the χ M T values were temperature-independent. The χ M T value was 0.43 cm 3 K mol −1 at 10 K. Meanwhile, on heating compound 1 above 300 K, the χ M T values smoothly increased up to 410 K and then an abrupt change in χ M T was observed (T 1/2 ↑ = 414 K). After the steep transition, the χ M T values still increased gradually again and reached 3.38 cm 3 K mol −1 at 450 K, suggesting compound 1 exhibited an incomplete SCO transition. On successive lowering temperatures, a steep decrease in χ M T was observed at 406 K (T 1/2 ↓ = 402 K), and then the χ M T curve traced that measured in the heating scan below 396 K. Thus, the compound 1 showed the cooperative SCO transition with a thermal hysteresis width of 12 K.
The temperature variations of magnetic susceptibility for compounds 1−3 are shown in Figure  3. The χMT value for compound 1 at 300 K was 0.56 cm 3 K mol −1 , suggesting 1 was almost in the LS state. Below 300 K, the χMT values were temperature-independent. The χMT value was 0.43 cm 3 K mol −1 at 10 K. Meanwhile, on heating compound 1 above 300 K, the χMT values smoothly increased up to 410 K and then an abrupt change in χMT was observed (T1/2↑ = 414 K). After the steep transition, the χMT values still increased gradually again and reached 3.38 cm 3 K mol −1 at 450 K, suggesting compound 1 exhibited an incomplete SCO transition. On successive lowering temperatures, a steep decrease in χMT was observed at 406 K (T1/2↓ = 402 K), and then the χMT curve traced that measured in the heating scan below 396 K. Thus, the compound 1 showed the cooperative SCO transition with a thermal hysteresis width of 12 K.
The χMT value for compound 2 at 300 K was 0.54 cm 3 K mol −1 , suggesting 2 was also almost in the LS state. On heating the sample, the χMT values gradually rose, and an abrupt increase in χMT was observed at 362 K (T1/2↑ = 365 K). The χMT value reached to 3.84 cm 3 K mol −1 at 450 K. On lowering temperatures, the χMT values traced those of the heating scan and then abruptly decreased at 364 K (T1/2↓ = 361 K), resulting in a cooperative transition with a thermal hysteresis width of 4 K.
The χMT value for compound 3 at 300 K was 4.15 cm 3 K mol −1 , suggesting 3 was almost in the HS state. On decreasing temperatures, the χMT values were almost constant in the temperature range of 2−300 K.

Mössbauer Spectroscopy for 1 and 3
To confirm the spin-states of the Fe ion for compound 1 and 3 at 293 K, the Mössbauer spectra for 1 and 3 were recorded ( Figure 4). The spectrum of 1 at 293 K consisted of only a sharp asymmetric doublet. The asymmetry of the spectrum may originate from the preferred orientation of the crystals. As compared with the Mössbauer parameters reported in the literature (Table 1), the spectrum of 1 can be ascribed to the LS spectrum, which is consistent with the χMT value of 1 at 293 K. On the other hand, the spectrum of 3 at 293 K consisted of a very broad asymmetric doublet. The isomer shift of 3 is similar to those of the simulated HS spectra reported in the literature. Note that the quadrupole splitting of 3 is the largest among those of the HS spectra in the related compounds and corresponds to that of the [Ni(nmt)2] compound having the significantly distorted coordination octahedron (mnt = maleonitriledithiolate) [33]. This suggests that the compound 3 might have a distorted coordination sphere due to some crystal packing effect. The χ M T value for compound 2 at 300 K was 0.54 cm 3 K mol −1 , suggesting 2 was also almost in the LS state. On heating the sample, the χ M T values gradually rose, and an abrupt increase in χ M T was observed at 362 K (T 1/2 ↑ = 365 K). The χ M T value reached to 3.84 cm 3 K mol −1 at 450 K. On lowering temperatures, the χ M T values traced those of the heating scan and then abruptly decreased at 364 K (T 1/2 ↓ = 361 K), resulting in a cooperative transition with a thermal hysteresis width of 4 K.
The χ M T value for compound 3 at 300 K was 4.15 cm 3 K mol −1 , suggesting 3 was almost in the HS state. On decreasing temperatures, the χ M T values were almost constant in the temperature range of 2-300 K.

Mössbauer Spectroscopy for 1 and 3
To confirm the spin-states of the Fe ion for compound 1 and 3 at 293 K, the Mössbauer spectra for 1 and 3 were recorded ( Figure 4). The spectrum of 1 at 293 K consisted of only a sharp asymmetric doublet. The asymmetry of the spectrum may originate from the preferred orientation of the crystals. As compared with the Mössbauer parameters reported in the literature (Table 1), the spectrum of 1 can be ascribed to the LS spectrum, which is consistent with the χ M T value of 1 at 293 K. On the other hand, the spectrum of 3 at 293 K consisted of a very broad asymmetric doublet. The isomer shift of 3 is similar to those of the simulated HS spectra reported in the literature. Note that the quadrupole splitting of 3 is the largest among those of the HS spectra in the related compounds and corresponds to that of the [Ni(nmt) 2 ] compound having the significantly distorted coordination octahedron (mnt = maleonitriledithiolate) [33]. This suggests that the compound 3 might have a distorted coordination sphere due to some crystal packing effect.

Reflection and Optical Conductivity Spectra of Single Crystals for 1 and 3
Since 1 and 3 have relatively large flat crystal surfaces, we tried to record reflection spectra at room temperature. The obtained reflection spectrum for 1 and 3 and their optical conductivity spectrum calculated by the Kramers-Kronig analysis are shown in Figure 5a and Figure 5b, respectively.
As is shown in the panel (a), the spectrum of the LS state in 1 displays the two clear absorption maxima: the strong visible absorption at 428 nm and the near-infrared medium band at 904 nm. On  Since 1 and 3 have relatively large flat crystal surfaces, we tried to record reflection spectra at room temperature. The obtained reflection spectrum for 1 and 3 and their optical conductivity spectrum calculated by the Kramers-Kronig analysis are shown in Figure 5a Some examples of the light-induced excited spin-state trapping (LIESST) effect on the [Fe(qsal) 2 ] compounds have ever been reported to date. The excitation wavelengths used for the LIESST effect were 808 [11] and 830 nm [22,23]. These wavelengths are in good agreement with the absorption band observed for 1, namely the LS [Fe(qsal) 2 ] compound. Therefore, the LIESST scheme from the LS to HS states in the [Fe(qsal) 2 ] compounds is evidenced.

Time-Dependent Density Functional Theory (TD-DFT) Calculations for the LS and HS [Fe(qsal)2] Cations
To provide an insight into the absorption bands for 1 and 3, the transition energies of all electron transitions in the LS [Fe(qsal)2] and HS [Fe(qsal)2] cations were calculated by using the time-dependent density functional theory (TD-DFT) method. The transition energies were strongly dependent on both the density functionals and [Fe(qsal)2] structures, and the most reasonable energies could be obtained by calculating the B3LYP-optimized structures at the CAM-B3LYP level. The transition wavelengths and assignments are summarized in Table 2.
The wavelengths calculated for the HS [Fe(qsal)2] cation were ascribed mainly to ligand-to-metal charge transfer (LMCT) transition and intra-ligand π-π* transition and were shorter than 656 nm, which was consistent with no remarkable absorption band observed in the near-infrared region for 3. The absorption bands observed at 467 and 366 nm for 3 were ascribed to π-π* and LMCT transitions, respectively. On the other hand, the wavelengths calculated for the LS [Fe(qsal)2] cation were ascribed mainly to d-d and LMCT transitions. Very weak absorption bands for the LS [Fe(qsal)2] cation were calculated in the near-infrared region. This is probably because these absorption bands would originate mainly from forbidden d-d transitions. The observed absorption bands at 904 and 428 nm for 1 can be ascribed to d-d and LMCT transitions, respectively.

Correlation between the Crystal Structures and Magnetic Behaviors for the [Fe(qsal)2] Compounds
The magnetic behaviors and short cation···cation distances for the [Fe(qsal)2] compounds with nonplanar counter-anions whose crystal structures were deposited to date are summarized in Table  9. Interestingly, the [Fe(qsal)2] cation arrangements except compound 3 are quite similar to each

Time-Dependent Density Functional Theory (TD-DFT) Calculations for the LS and HS [Fe(qsal) 2 ] Cations
To provide an insight into the absorption bands for 1 and 3, the transition energies of all electron transitions in the LS [Fe(qsal) 2 ] and HS [Fe(qsal) 2 ] cations were calculated by using the time-dependent density functional theory (TD-DFT) method. The transition energies were strongly dependent on both the density functionals and [Fe(qsal) 2 ] structures, and the most reasonable energies could be obtained by calculating the B3LYP-optimized structures at the CAM-B3LYP level. The transition wavelengths and assignments are summarized in Table 2. The wavelengths calculated for the HS [Fe(qsal) 2 ] cation were ascribed mainly to ligand-to-metal charge transfer (LMCT) transition and intra-ligand π-π* transition and were shorter than 656 nm, which was consistent with no remarkable absorption band observed in the near-infrared region for 3. The absorption bands observed at 467 and 366 nm for 3 were ascribed to π-π* and LMCT transitions, respectively. On the other hand, the wavelengths calculated for the LS [Fe(qsal) 2 ] cation were ascribed mainly to d-d and LMCT transitions. Very weak absorption bands for the LS [Fe(qsal) 2 ] cation were calculated in the near-infrared region. This is probably because these absorption bands would originate mainly from forbidden d-d transitions. The observed absorption bands at 904 and 428 nm for 1 can be ascribed to d-d and LMCT transitions, respectively.

Crystal Structures of 1-3
The variable temperature single crystal X-ray structural analyses for 1, 2, and 3 were performed using a Bruker AXS APEXII Ultra diffractometer. Crystallographic data are listed in Tables 3 and 4. The crystal structures for 1 and 3 at 293 K belonged to a triclinic system with P1, whereas the crystal structure of 2 at 293 K belonged to monoclinic P2 1 /n. Each asymmetrical unit consisted of one [Fe(qsal) 2 ] cation and one corresponding anion, and additionally one methanol molecule for 3. Fortunately, it was successful to determine the high-temperature phase structures of 1 and 2 at 400 and 425 K, respectively. The crystal space group for 2 maintained at 400 K, whereas that for 1 was changed into monoclinic P2/n. As associated with the crystal transformation, the crystallographically independent molecules in 1 at 425 K increased to three and a half of the [Fe(qsal) 2 ] cations and CH 3 OSO 3 anions. The independent [Fe(qsal) 2 ] cations in 1 at 425 K are hereafter designated as A, B, C, and D (a half independent cation). The atomic numbers of the [Fe(qsal) 2 ] cations A to D also add the notations of A to D to the corresponding atomic numbers.  The π-ligand qsal anion was coordinated to a central Fe atom as a tridentate chelate ligand and thus two coordinated ligand molecules were arranged in an almost perpendicular manner (Figure 6a-c). The coordination bond lengths and distortion parameters (Σ, Θ, Φ, see Figure 6d) for 1, 2, and 3 along with those of the HS and LS [Fe(qsal) 2 ] compounds confirmed by Mössbauer spectra are listed in Table 5. The distortion parameter Σ is the sum of the absolute differences in 12 coordination bite angles from 90 • . The distortion parameter Θ is the sum of the absolute differences in 24 angles from 60 • on 8 surface triangles of a coordination octahedron. Both Σ and Θ are zero if the coordination sphere is a regular octahedron. The distortion parameter Φ is the deviation of the angle between two coordination bonds from Fe to N atoms of two imine groups from 180 • . All the distortion parameters increase their values on distorting the coordination octahedron.
As compared with the coordination bond lengths and distortion parameters, the Fe-O and Fe-N bond lengths and distortion parameters for 1 and 2 at 293 K were quite similar to those of the LS NCSe compound, indicating that the [Fe(qsal) 2 ] cations in 1 and 2 were in the LS state at 293 K. On the other hand, the coordination bond lengths and distortion parameters for 3 at 293 K were much larger than those in 1 and 2, whereas they were similar to those in the HS [Ni(dmit) 2 ] compound (dmit = 4,5-dithiolato-1,3-dithiole-2-thione). Thus, the [Fe(qsal) 2 ] cation in 3 was in the HS state at 293 K. These observations are in good agreement of the spin-states of 1-3 from the magnetic susceptibilities and Mössbauer spectra at 293 K. On increasing temperatures, the coordination bond lengths of 1 were gradually lengthened, which was consistent with the gradual increase in the χ M T in 1. Note that the distortion parameters Σ in 1 decreased on increasing temperature to 400 K. This suggests that the distortion parameter Σ may not closely correlate with the spin-state of the [Fe(qsal) 2 ] cation.
At 425 K all the coordination bond lengths of three and a half independent [Fe(qsal) 2 ] cations except the Fe-O bond lengths in cation A were longer than those in 1 at 400 K. Although the distortion parameters Σ of the three and a half cations were varied, all the distortion parameters Θ and Φ were larger than those at 400 K. As compared with the LS NCSe and HS [Ni(dmit) 2 ] compounds, the cations A and D may contain certain LS fractions. Therefore, the SCO in 1 proved to be an incomplete cooperative transition accompanying a crystal structure phase transition. On the other hand, the coordination bond lengths and distortion parameters Θ and Φ in 2 at 400 K were a little smaller than those in the HS [Ni(dmit) 2 ] compound. This indicates that the SCO transition in 2 was also an incomplete one.
The  2 ] compound were deviated from the above bond length range. Since the degrees of deviations were small, the occurrence of SCO cannot be judged only from the coordination bond lengths. As mentioned above, the parameters Σ may not be related to the spin-states of the [Fe(qsal) 2 ] cation. On the other hand, the parameters Θ and Φ are probably reflected in their spin-states. In particular, the parameters Φ may be useful to judge the occurrence of SCO because the largest Φ values are observed in the HS [Ni(mnt) 2 ] compound and HS compound 3. Halcrow reported that the similar deviations of Φ are related to the occurrence of SCO in the Fe(II) complexes from bpp ligands (bpp = 2,6-di(pyrazol-1-yl)pyridine) [54]. These findings indicate that the central donor atoms of tridentate ligands can greatly impact the ligand field splitting energies in their homoleptic complexes, which is in good agreement with the role of the azo-functional group in new anionic SCO Fe(III) complexes from azobisphenolate ligands [55].  Table 5. The distortion parameter Σ is the sum of the absolute differences in 12 coordination bite angles from 90°. The distortion parameter Θ is the sum of the absolute differences in 24 angles from 60° on 8 surface triangles of a coordination octahedron. Both Σ and Θ are zero if the coordination sphere is a regular octahedron. The distortion parameter Φ is the deviation of the angle between two coordination bonds from Fe to N atoms of two imine groups from 180°. All the distortion parameters increase their values on distorting the coordination octahedron.    (11) 15.80 (7) 13.79 (9) 1.50 (14) 17.83 (8)

Molecular Arrangement of 1
The molecular arrangement of the [Fe(qsal) 2 ] cations and CH 3 OSO 3 anions in 1 along the b−c direction is shown in Figure 7a. Selected intermolecular distances in 1 are listed in Table 6. The phenyl ring (C1-C6) in the [Fe(qsal) 2 ] cation was stacked with the quinolyl ring (C8-C16,N2) in the nearest neighboring [Fe(qsal) 2 ] cation in a parallel-displaced manner, to form a π-stacking [Fe(qsal) 2 ] dimer (p in Figure 7b). The π-stacking dimers were arranged along the b−c direction through short C···C contacts between the imine and quinolyl moieties (q in Figure 7b) and thus afforded an alternate one-dimensional (1D) [Fe(qsal) 2 ] array. This type of 1D [Fe(qsal) 2 ] arrangement was found in [Fe(qsal) 2 ](NCX)·solv (X = S, Se) [9][10][11] and [Fe(qsal) 2 ](I 3 ) [13]. Moreover, a parallel-displaced π-stacking between the phenyl rings (C17-C22) (r in Figure 7c) and edge-shared π-stacking between the quinolyl rings (C8-C16,N2) (s in Figure 7d) were observed along the b+c and a directions, respectively. Therefore, the formation of a three-dimensional (3D) π-stacking interaction network is elucidated in 1. The CH 3 OSO 3 anions were located in the cavity of the 3D π-stacking network of the [Fe(qsal) 2 ] cation. All the edge oxygen atoms in the CH 3 OSO 3 anion were contacted with the hydrogen atoms of four neighboring [Fe(qsal) 2 ] cations. On increasing temperatures from 293 to 400 K, the intermolecular distances for 1 were gradually increased ( Table 6), but the overlapping modes were not changed. On the other hand, the molecular arrangement of the [Fe(qsal)2] cations and CH3OSO3 anions at 425 K were dramatically changed, which can be easily recognized by comparison between Figure 7e and Figure 7f. The periodic units of the 1D arrays of the [Fe(qsal)2] cation and CH3OSO3 anion were two and seven at 293 and 425 K, respectively. The incommensurate periodic units suggest that a large structural rearrangement would take place through the structural phase transition accompanying SCO. Within the 1D [Fe(qsal)2] array at 425 K, the π-overlaps similar to p shown in Figure 7b were found both between cations A and between cations B and D. Although the π-overlaps were deformed and twisted, several C···C contacts   (3) 3.384 (4) a The positions corresponding to letters are shown in Figure 7.
On increasing temperatures from 293 to 400 K, the intermolecular distances for 1 were gradually increased ( Table 6), but the overlapping modes were not changed. On the other hand, the molecular arrangement of the [Fe(qsal) 2 ] cations and CH 3 OSO 3 anions at 425 K were dramatically changed, which can be easily recognized by comparison between Figure 7e,f. The periodic units of the 1D arrays of the [Fe(qsal) 2 ] cation and CH 3 OSO 3 anion were two and seven at 293 and 425 K, respectively. The incommensurate periodic units suggest that a large structural rearrangement would take place through the structural phase transition accompanying SCO. Within the 1D [Fe(qsal) 2 ] array at 425 K, the π-overlaps similar to p shown in Figure 7b were found both between cations A and between cations B and D. Although the π-overlaps were deformed and twisted, several C···C contacts shorter than the sum of van der Waals (vdW) radii (C: 1.70 Å) [56] were observed between cations A and C, whereas there is only one short C···C contact between cations B and C. Thus, the 1D [Fe(qsal) 2 ] array seemed to consist of C···A···A···C tetramer and D··B···D trimer at 425 K. Between the 1D [Fe(qsal) 2 ] arrays, most π-overlaps similar to s shown in Figure 7d were lengthened in the range of 3.407-3.610 Å along the b axis, whereas one half of π-overlaps similar to r shown in Figure 7c disappeared and the remained π-overlaps were observed only between cations A and B along the a axis. Thus, the 3D π-stacking interaction network was rearranged and weakened by the deformation of the [Fe(qsal) 2 ] array. Since the orientations of the CH 3 OSO 3 anions were partly different from those below 400 K and their thermal ellipsoids were also much larger than those of the [Fe(qsal) 2 ] cation at 425 K, the crystal structure phase transition in 1 seems to arise from the thermal motion of the CH 3 OSO 3 anions.

Molecular Arrangement of 2
The molecular arrangement of the [Fe(qsal) 2 ] cations and CF 3 SO 3 anions in 2 along the a+c direction at 293 and 400 K are shown in Figure 8a,b. Selected intermolecular distances are listed in Table 7. Similar to the π-stacking [Fe(qsal) 2 ] dimer in 1, the qsal ligand of the [Fe(qsal) 2 ] cation was stacked with that of the neighboring [Fe(qsal) 2 ] cation in a head-to-tail manner (t in Figure 8c). On the other hand, the π-stacked [Fe(qsal) 2 ] cations were related each other by a symmetry operation of n glide plane and thus gave a uniform 1D molecular array along the a+c direction. This type of the uniform 1D [Fe(qsal) 2 ] arrangement was found in [Fe(qsal) 2 ](NCS) [11] and [Fe(qsal) 2 ](I) [15]. The π-overlaps between the 1D [Fe(qsal) 2 ] arrays were found between the phenyl moieties along the a−c direction (u in Figure 8a,c). Thus, the [Fe(qsal) 2 ] cations formed a two-dimensional (2D) regular network along the ac plane. Meanwhile, there was no remarkable short contact between the 2D [Fe(qsal) 2 ] networks.
The molecular arrangement of 2 at 400 K was very similar to that at 293 K. The difference in the 1D π-stacking arrays (Figure 8a,b) and 2D π-stacking network structure (Figure 8c,d) were hardly observed, but the 2D π-stacking layers glided alternately along the c axis (Figure 8e,f). Since the short O5···C23(imine) distance was shortened to be 3.039(6) Å at 400 K, this structure transition would be involved in the Coulomb interaction. The CF 3 SO 3 anions in a 2D π-stacking layer were located near the quinolyl rings in the neighboring 2D layer at 293 K, whereas the CF 3 SO 3 anions were shifted from the quinolyl rings in the neighboring 2D layer at 400 K. Moreover, the thermal ellipsoids of the CF 3 SO 3 anion and quinolyl moiety were larger than those of the other molecular components at 400 K. These observations indicate that this structure transition in 2 may result from the thermal motion of the CF 3 SO 3 anion, which is reminiscent of the structural transition in 1.    (3) 3.585 (7) a The positions corresponding to letters are shown in Figure 8. b (−0.5+x, 1.5−y, −0.5+z). c (0.5+x, 1.5−y, 0.5+z). d (0.5+x, 1.5−y, −0.5+z).

Correlation between the Crystal Structures and Magnetic Behaviors for the [Fe(qsal)2] Compounds
The magnetic behaviors and short cation···cation distances for the [Fe(qsal)2] compounds with nonplanar counter-anions whose crystal structures were deposited to date are summarized in Table  9. Interestingly, the [Fe(qsal)2] cation arrangements except compound 3 are quite similar to each other.  The CH 3 SO 3 anions and methanol molecules were located between the 2D [Fe(qsal) 2 ] networks. The S-O bond lengths imply that the negative charge in the CH 3 SO 3 anion may be localized mainly at O3 and O4 atoms. The short O4···C7 and O3···O6 distances were found to be 3.070(3) and 2.814(4) Å, respectively. The former suggests the existence of effective Coulomb interactions between the [Fe(qsal) 2 ] cation and CH 3 SO 3 anion, the latter indicates that of hydrogen bonding interactions between the CH 3 SO 3 anion and methanol molecule. We can assume that both strong Coulomb interaction and hydrogen bonding interaction may induce the distortion of a coordination octahedron, leading to the HS state in the whole temperature range.

Correlation between the Crystal Structures and Magnetic Behaviors for the [Fe(qsal) 2 ] Compounds
The magnetic behaviors and short cation···cation distances for the [Fe(qsal) 2 ] compounds with nonplanar counter-anions whose crystal structures were deposited to date are summarized in Table 9 [57,58], which seemed to be driven by Coulomb interactions. In general, the Coulomb interaction is much stronger than other intermolecular interactions. Thus, the structure transformation through desolvation in charged complexes may be driven by the Coulomb energy gain, namely shrinking the distances between the cations and anions. Several C···N and C···S distances shorter than the sum of vdW radii are found between the [Fe(qsal) 2 ] cation and NCS anion for [Fe(qsal) 2 ](NCS), indicating the Coulomb interactions may operate the crystal structure transformation for [Fe(qsal) 2 ](NCX)·solv.
One may ask why solvate molecules are included in a crystal lattice although they will reduce Coulomb energy gain. If the energy gain from other intermolecular interactions in a solvate compound is larger than the loss of Coulomb energy gain, the total energy gain can be larger than that in a non-solvate compound. The shorter intermolecular distances involved in π-stacking interactions in [Fe(qsal) 2 ](NCX)·solv suggest that the strong π-stacking interactions may result from the [Fe(qsal) 2 ] cation arrangement given by the inclusion of the solvate molecules. Therefore, we can assume that the solvate molecule in a solvate complex seems to play a significant role in the enhancement of the total lattice energy gain by various intermolecular interactions between its molecular components.
This idea is also applicable to 1, 2, and the I 3 compound. Although the crystal structure from a charged non-solvate complex should be a typical ionic crystal which has a larger coordination number and shorter cation···anion distances, the crystal structures for 1, 2, and the I 3 compound were not those of a typical ionic crystal but those similar to [Fe(qsal) 2 ](NCS)·CH 2 Cl 2 and [Fe(qsal) 2 ](NCSe)·CH 2 Cl 2 . This suggests that the total lattice energy gain from the strong π-stacking interactions found in 1, 2, and the I 3 compound may exceed the loss of Coulomb energy gain in the crystal structure of a typical ionic crystal. Consequently, we can discuss the role of large non-spherical counter-anions or solvate molecules for a crystal packing in a charged complex by means of the competition between Coulomb and intermolecular interactions. Further investigations on other charged complexes are needed for verification of the present finding.
Next, we will discuss the large thermal hysteresis loops of the magnetic susceptibility found in [Fe(qsal) 2 ](NCS) and [Fe(qsal) 2 ](NCSe). As shown in Table 9, more than two-step variations in the cooperative SCO transition are one of the characteristic points for 1, 2, [Fe(qsal) 2 ](NCS) and [Fe(qsal) 2 ](NCSe). Moreover, large thermal ellipsoids or disorder of counter-anions were observed in the crystal structures of their HS phases. This implies that the thermal motion of the counter-anion may play an important role in their HS crystal structures. The thermal variations in the crystal structures of 1 from 293 to 400 K (Table 6) revealed that strong intermolecular π-stacking interactions were retained despite the gradual SCO conversion. Recently, we found similar gradual SCO conversion in the charged Fe(II) compound having strong intermolecular interactions, in which the large difference in lattice enthalpy between the LS and HS states leads to a gradual SCO conversion despite the existence of strong intermolecular interactions [58]. Therefore, to undergo an abrupt SCO transition with a large hysteresis loop, it may be important to realize a small difference in lattice enthalpy between the LS and HS states by the choice of a counter-anion or solvate molecule.

Conclusions
New [Fe(qsal) 2 ] compounds, [Fe(qsal) 2 ](CH 3 OSO 3 ) 1 and [Fe(qsal) 2 ](CH 3 SO 3 )·CH 3 OH 3, along with the reported compound, [Fe(qsal) 2 ](CF 3 SO 3 ) 2, were prepared and characterized. The compounds 1 and 2 exhibited a cooperative SCO transition at higher temperatures. The optical conductivity spectrum from the single-crystal reflection spectrum of 1 revealed that the photo-excitation band for the LIESST effect on the LS [Fe(qsal) 2 ] compounds is attributed to d-d transition. The successful crystal structure determinations of 1 and 2 in the high-temperature phase reveal that large structural changes were triggered by the motion of counter anions. The structural comparison among the [Fe(qsal) 2 ] compounds determined to date suggests that the counter-anions and solvate molecules play a significant role in the total lattice energy gain in a charged complex. The present finding may lead to elucidation of the role of counter-anions and solvate molecules for controlling SCO transition behaviors. To do this end, the quantitative evaluation of lattice enthalpy for each spin-state in SCO compounds may be required. Since the [Fe(qsal) 2 ] derivatives are suitable candidates for this purpose, we are now investigating a family of the [Fe(qsal) 2 ] derivatives.