Homochiral Mn3+ Spin-Crossover Complexes: A Structural and Spectroscopic Study

Structural, magnetic, and spectroscopic data on a Mn3+ spin-crossover complex with Schiff base ligand 4-OMe-Sal2323, isolated in crystal lattices with five different counteranions, are reported. Complexes of [Mn(4-OMe-Sal2323)]X where X = ClO4– (1), BF4– (2), NO3– (3), Br– (4), and I– (5) crystallize isotypically in the chiral orthorhombic space group P21212 with a range of spin state preferences for the [Mn(4-OMe-Sal2323)]+ complex cation over the temperature range 5–300 K. Complexes 1 and 2 are high-spin, complex 4 undergoes a gradual and complete thermal spin crossover, while complexes 3 and 5 show stepped crossovers with different ratios of spin triplet and quintet forms in the intermediate temperature range. High-field electron paramagnetic resonance was used to measure the zero-field splitting parameters associated with the spin triplet and quintet states at temperatures below 10 K for complexes 4 and 2 with respective values: DS=1 = +23.38(1) cm–1, ES=1 = +2.79(1) cm–1, and DS=2 = +6.9(3) cm–1, with a distribution of E parameters for the S = 2 state. Solid-state circular dichroism (CD) spectra on high-spin complex 1 at room temperature reveal a 2:1 ratio of enantiomers in the chiral conglomerate, and solution CD measurements on the same sample in methanol show that it is stable toward racemization. Solid-state UV–vis absorption spectra on high-spin complex 1 and mixed S = 1/S = 2 sample 5 reveal different intensities at higher energies, in line with the different electronic composition. The statistical prevalence of homochiral crystallization of [Mn(4-OMe-Sal2323)]+ in five lattices with different achiral counterions suggests that the chirality may be directed by the 4-OMe-Sal2323 ligand.


■ INTRODUCTION
Manipulation of the internal electronic arrangement in spincrossover (SCO) complexes, 1−5 with the attendant changes in magnetic, 6−8 optical, 9−14 and electric properties, 7,15−22 constitutes one of the most versatile ways to build switchable molecular magnets. This versatility is underscored by the varied thermal evolution profiles which characterize spin-state switching. These can range from extremely sharp and hysteretic, 23,24 particularly suitable for memory applications, 25,26 to more gradual transitions which have good potential in neuromorphic or sensing roles. 27−29 The most studied Mn 3+ SCO complexes are the mononuclear examples with a hexadentate Schiff base ligand from the R-Sal 2 323 family prepared from condensation of 1,2-bis(3a m i n o p r o p y l -a m i n o ) e t h a n e w i t h a s u b s t i t u t e d 2-hydroxybenzaldehyde; for example, see Scheme 1. Metal complexes prepared with hexadentate chelates will be inherently chiral as the Schiff-base ligand has chirogenic amine nitrogen atoms. Crystallization of racemates of the Δ and Λ isomers in centrosymmetric space groups is typical, although recovery of mechanical mixtures of chiral conglomerates of the two forms is also possible but is less common. Here we report the serendipitous crystallization of the SCO complex cation [Mn(4-OMe-Sal 2 323)] + in conglomerate chiral form in five different lattices with achiral counterions. Such homochirality, without use of a chiral anion, has not previously been observed so systematically as is the case for [Mn(4-OMe-Sal 2 323)] + which suggests a ligand-directed effect. In total, [Mn(4-OMe-Sal 2 323)] + was isolated in conglomerate chiral form in lattices with ClO 4 − , BF 4 − , NO 3 − , Br − , and I − , all crystallizing in spacegroup P2 1 2 1 2 and all with a crystallographic C 2 axis bisecting the complex cation. A sixth example, that with a Cl − counterion, was also recovered in space group Pccn and the structure only of that complex is included (in the Supporting Information) for the sake of completeness. A check of the CCDC database reveals that only 6 of the 78 unique [Mn(R-Sal 2 323)] + complexes 30 published before 2021 crystallize adventitiously in a chiral space group, 31−36 with a seventh example targeted by introduction of a chiral anion. 37 The interplay between SCO and chirality 38−58 is increasingly recognized as an important route to switchable nonlinear optical (NLO) materials 51,59−64 and spin-state dependent changes in optical activity may also constitute an economic and low energy route to follow SCO in sensing applications. Here we use circular dichroism to confirm the spontaneous resolution in the case of the ClO 4 − complex (1) and to demonstrate that the complex is stable toward racemization in solution.
Another less studied aspect of the SCO phenomenon is the associated change in magnetic anisotropy which accompanies the spin pairing/unpairing process and, in particular, the change in magnitude of the zero-field splitting (ZFS) contribution. 65 The most commonly studied SCO complexes often include the Kramers ions Fe 3+ and Co 2+ which have a spin doublet ground state in the fully paired low-spin (LS) configuration and, hence, no ZFS at low temperature, or the non-Kramers Fe 2+ ion which has a spin singlet ground state when fully paired, i.e., again no ZFS in the cryogenic regime. In contrast, Mn 3+ , which is also a non-Kramers ion and for which thermal SCO is now wellestablished, 30,31,66−71 switches between the fully unpaired spin quintet and partially paired spin triplet forms; therefore, a considerable ZFS is expected to persist at low temperature. Spin triplet Mn 3+ is generally not common, with about 20 examples at room temperature, 32,72−88 and the ZFS interactions have been quantified via the D and E parameters in just two cases. 32,83 These studies have however demonstrated that spin triplet Mn 3+ has the highest ZFS parameters for any manganese ion, with D values in the range +15 to +20 cm −1 , 65 while the axially elongated spin quintet form shows small but negative values in the range of −4.5 to −1.2 cm −1 with only a small number of published spin quintet Mn 3+ examples with positive D values, i.e., axial compression of the Jahn−Teller ion. 65,89−104 In these examples, the sign and magnitude of the ZFS parameters have been examined in selected examples by Tregenna-Piggott, 89,90 Krzystek and Telser, 91,101−103 and Duboc and Neese, 92,93,99,100 who have built on the earlier studies of Gregson in probing the electronic structure. 98 Thermally induced switching to a different spin state in Mn 3+ can therefore be expected to profoundly affect the magnitude and possibly also the sign of D. Identification of the sign of D is particularly relevant to the current work because, although the majority of known (HS) Mn 3+ complexes display a pronounced axial elongation due to the Jahn−Teller effect, most Mn 3+ SCO complexes appear to have a marked compression in the spin quintet form. 19,31,33,37,69,88,105−112 An axially compressed 89−100 form may assist the transition to a spin triplet arrangement as the energetic order of orbitals should match that expected in the S = 1 form of the ion (Figure 1). Diffraction alone, however, is not sufficient to discern if this is a genuine compression, but it can be resolved by measuring the sign of D by EPR at high fields.
Here we use low-temperature multifrequency EPR spectroscopy to establish the magnitude and sign of the axial D parameter in the spin quintet and triplet forms of the [Mn(4-OMe-Sal 2 323] + complex cation 105 when it is crystallized in BF 4 − and Br − lattices, respectively complexes 2 and 4 in Scheme 1. In SCO complexes, the geometric structure is heavily dependent on the spin state due to population/depopulation of antibonding orbitals and associated bond length changes. Therefore, we present the temperature-dependent magnetic data at the outset as this guides the choice of temperature for the diffraction studies. The EPR investigation is new and knowledge of the spin state at the temperature of the measurement is essential for the study. Hence, we report EPR measurements and analysis in close alignment with the magnetic results, before discussing structural and optical properties. This ligand type is usually abbreviated as R-Sal 2 323 to indicate the 323 8-carbon alkyl connectivity in the starting tetraamine and the substitution (R) on the phenolate ring. Chirogenic amine nitrogen donors are indicated by an asterisk.   (6), which does not crystallize isomorphously and will therefore not be used further in the magnetic and spectroscopic investigation reported here on 1−5. Synthesis and structural details of 6 can be found in the Supporting Information (section S1).

■ RESULTS AND DISCUSSION
Magnetic Characterization of Complexes 1−5. The magnetic susceptibility, χ M , of the bulk samples of compounds 1−5 was measured using a SQUID magnetometer on cooling from 300 to 5.0 K under an applied direct current (dc) field of 1000 Oe, shown as temperature dependence in the form of the χ M T product ( Figure 2).
Complexes 1 and 2 are high-spin over the measured temperature range with values close to the expected spin only value of 3.0 cm 3 K/mol for a monomeric Mn 3+ complex with S = 2 and g = 2. The room temperature value of 2.9 cm 3 K/mol for complex 3 indicates that the full high-spin S = 2 state is not reached at room temperature and, on cooling, the χ M T value decreases steadily in a two-step transition, which becomes clearer in the derivative of the χ M T product (see Figure S1), where maxima at 142 and 37 K are apparent. The χ M T value displays a plateau at 2.3 cm 3 K/mol between these two temperature points, which is indicative of a 2:1 ratio of HS/LS sites. In order to confirm the second step at lower temperatures, the susceptibility was additionally measured in warming mode (see Figure S1). Bromide complex 4 is the only compound that exhibits a full thermal spin transition from S = 2 to S = 1 (expected spin-only value of 1.0 cm 3 K/mol using S = 1 and g = 2) between 300 and 70 K, following a gentle sigmoidal pathway and with T 1/2 of 136 K. The χ M T value of 2.0 cm 3 K/mol at room temperature for complex 5, with the larger iodide counterion, indicates a likely 1:1 mixture between the spin triplet and quintet sites within the crystal lattice. The χ M T product remains almost constant over the measured temperature range upon cooling, before a decrease below 35 K, suggesting a further adjustment to the spin-state ratio, as this temperature is too high for zero-field splitting effects, which are typically observed below 25 K.
High-Field Electron Paramagnetic Resonance Spectra of 2 and 4. High-field EPR (HFEPR) spectra were recorded at low temperatures (∼10 K) on polycrystalline powder samples of compounds 2 and 4 in order to characterize the ZFS parameters associated with the HS and LS species, respectively. We first present the results for compound 4, which undergoes a complete transition to the LS state below ∼50 K, resulting in very clean and simple HFEPR spectra, some of which are displayed in Figure 3.
Resonance peak positions determined from HFEPR spectra for 4 such as those in Figure 3 (and several others not shown) were then used to construct a 2D frequency versus field map, as shown in Figure 4, with colors denoting turning points due to the three principal (x, y, and z) components of the powderaveraged spectra. These data were then fit to the following spin Hamiltonian: 113 assuming a spin S = 1 ground state. The first and second terms in eq 1 respectively denote the axial and rhombic second-order ZFS interactions, with the associated D and E parameters. Meanwhile, the last term parametrizes the Zeeman interaction in terms of an anisotropic g '◊ -tensor. Ŝis the total spin operator with components Sî (i = x, y, and z), B ÷◊ ÷ is the magnetic induction, and μ B is the Bohr magneton. The data indicate more than one zerofield energy gap, 114 requiring a finite, albeit relatively small rhombicity factor E/D = 0.117 (the lowest frequency intercept  . Derivative mode (dI/dB 0 , where I is the microwave intensity transmitted through the sample and B 0 the applied magnetic field) HFEPR spectra of 4 recorded at 10 K and frequencies as indicated. The experimental spectra are shown in blue and simulations in red, generated using the parameters given in the main text. The resonances are labeled according to the associated components (x, y, and z) of the powder spectrum. The sharp feature at the g = 2.00 position, marked with an asterisk (*), is attributed to a Mn 2+ impurity; # denotes a weak signal possibly due to a small S = 2 Mn 3+ contaminant. corresponds exactly to 2E = 167 GHz = 5.58 cm −1 ). 115 The spectral simulations in Figure 3 were then generated with the following parameters: g x = 1.97 (2), g y = 2.13(5), g z = 2.00(1), D = +23.38(1) cm −1 , and E = +2.79(1) cm −1 . The relative intensities of the z and x spectral peaks confirm the positive sign of the axial ZFS (D) parameter. These parameters were used to simulate the low-temperature dc magnetic susceptibility of 4 ( Figure S2). We note that the obtained D and E values are slightly larger than those reported previously for a similar LS Mn 3+ compound, [Mn(napsal 2 323)]NTf 2 . 32 Finding one g-tensor component that is larger than 2.00 for a d 4 electronic configuration is initially surprising, as one normally anticipates values lower than 2.00 for a less than half-filled dshell. However, the non-Hund's rule LS configuration may be reduced to that of a greater than half-filled t 2g set for an octahedral coordination, i.e., 4 electrons (2 holes) occupying the three t 2g orbitals. 116 Distortions away from octahedral geometry may then give g-tensor components greater than 2.00.
Compound 2 was selected because magnetic measurements ( Figure 2) suggest that it remains in the S = 2 HS state down to the low temperatures necessary for achieving sufficient HFEPR sensitivity. The resultant powder spectra turned out to be surprisingly difficult to interpret. First, the increased spin multiplicity inevitably results in many more spectral features, especially when considering both parallel and perpendicular mode transitions and off-axis peaks. 117 The 2D frequency versus resonance field maps generated from multifrequency measurements collected at 5 K are displayed in Figures 5 and S3−S5. Attempts to simulate the results start by following the evolution of signals to zero field, thus gaining information on the ZFS without dependence on the g-tensor. In the present case, there are very obvious high-frequency intercepts at ∼560 and ∼675 GHz that fit well with the S = 2 model. However, there is another series of weaker peaks (red open circles) with an intercept at ∼600 GHz that is incompatible with an S = 2 state, which appears to better fit to the S = 1 parametrization in Figure  4 (with a slightly lower intercept value). These resonances can be fit assuming S = 1 and g x = 2.03(6), g y = 2.16(2), g z = 2.0(2), D = +23.3(8) cm −1 , and E = +3.1(8) cm −1 ; they are, thus, not very different from the ZFS parameters of 4 and also exhibit the same g y > 2 issue. The parameter errors are large due to the relatively small number of available data points and low signal quality. Therefore, we suspect that either some sites in crystals of 2 convert to a LS state or that the powder sample is contaminated with a small fraction of microcrystals that undergo a transition to a LS state, perhaps caused by stresses induced when grinding the sample. These findings illustrate the value in carrying out HFEPR in order to deduce ZFS parameters, as the mixtures of spin states would render any parameters deduced from a purely thermodynamic measurement highly unreliable.
After constraining the high-frequency intercepts, it is found that the best S = 2 simulation [based on eq 1] results in additional zero-field intercepts at lower frequencies that fall in a range (150−275 GHz) where many low-field peaks are observed. However, no single parametrization reproduces all peak positions. We therefore believe that the powder sample also contains multiple S = 2 species with a small spread in ZFS parameters; 118 to reflect this, we employ broader/lighter simulated curves in Figure 5. The existence of different species may be due to distinct sites within the lattice of an individual crystal, 119 or they could be due to crystal-to-crystal variations within the powder. A conservative analysis indicates the following axial ZFS parameter, D = +6.9(3) cm −1 , and a much more significant spread in E, with values from 0.17 to 0.63 cm −1 needed to reproduce all observed resonances (see Figures 5 and S3−S5). Again, the sign of D is constrained via the relative intensities of the different peaks. Indeed, a high-frequency spectral simulation that assumes D = +6.9(3) cm −1 , E = +0.63 cm −1 , g x = 1.97, g y = 1.98, and g z = 1.94 is in excellent agreement with the corresponding experimental spectrum, as  Finally, we comment on a possible relationship between the D parameters obtained for the HS and LS states. Here, we have assumed that the main contribution to the ZFS comes from spin−orbit coupling (SOC). It then follows that the Ligand-Field theory perturbative expression for D consists of three terms: (1) a numerator proportional to the SOC constant, ζ, and the sum of the squared matrix elements of the Lẑ operator between the ground and excited orbital states; (2) a denominator corresponding to the associated orbital excitation energies; and (3) a prefactor, 1/S 2 . 120 The first two terms may not vary significantly between the HS and LS states, as they primarily involve excitations between essentially the same orbital states in either (octahedral) case. However, the prefactor obviously differs by a factor of 4 between LS and HS states, potentially explaining the majority of the reduction in D between the two configurations (all other things being equal), i.e., D S=1 / D S=2 = 3.4. Interestingly, for a pure DSẑ 2 description of the ZFS Hamiltonian, the largest zero-field gap in the spectrum is proportional to (2S − 1) [= S 2 − (S − 1) 2 ] and should, thus, differ by a factor of 3 for the S = 1 and 2 states with identical D values. The fact that D S=1 /D S=2 ≈ 3 in the present case explains why the zero-field intercepts are close to 600 GHz for both states in Figures 4 and 5, i.e., the different D values and (2S − 1) factors more or less cancel, leading to similar aggregate magnetic anisotropies (as measured by the ZFS) for the two compounds.
Theoretical Calculations. Zero-field splitting is a result of spin−orbit coupling and ligand field splitting of energy levels of a paramagnetic atom possessing spin larger than 1/2. Dependencies between the ZFS and the ligand field energies for various electronic configurations are well-known, 121−125 but they are often difficult to apply as the ligand field bands are obscured by the charge-transfer bands. Calculations of the ZFS parameters for the Mn 3+ ions were thus attempted using the state-averaged complete active space self-consistent field (CASSCF) method as implemented in the ORCA 5.0.1 quantum mechanical software package. 126−129 Four electrons in five orbitals were used in the calculations; five quintets and the lowest ten triplet and ten singlet states were taken into account. 129−131 The initial quasirestricted orbitals (qro) were obtained from a DFT calculation employing the B3LYP/G functional and the diffuse def2-TZVPP function basis set for all atoms. 132 The Ahlrich's auxiliary basis sets were embedded into the ORCA software. 133,134 In the case of complex 2 (BF 4 − ), the calculation produced D = +3.84 cm −1 , compared to the experimental D = +6.9(3) cm −1 . The calculated E/D ratio of 0.11 compares reasonably with the experimental values that range from 0.025 to 0.09. The largest contribution to D (2.209 cm −1 ) comes from the lowest triplet state, derived from the free-ion term 3 H. In contrast, calculations of D in the S = 1 state of complex 4 (Br − ) were less satisfactory, and despite using the wave functions from an S = 1 DFT calculation, the CASSCF procedure converged to the S = 2 state. Using the structures optimized by the ORCA DFT calculations (Table S1) did not result in an improvement (see section S2 of the Supporting Information).
Structural Characterization of Compounds 1−5. All complexes 1−5 crystallize isostructurally in the orthorhombic space group P2 1 2 1 2, with Z = 2, where the asymmetric unit contains half of a [Mn(4-methoxy-sal 2 323)] + cation, as shown in Figure 7, and half of the respective anion, both located on a symmetry center. All complexes crystallized solvent-free.
Typically the average bond lengths in [Mn(R-sal 2 323)] + type complexes change upon spin transition, but only significantly for  . 110 The bond lengths of complexes 1−5 are summarized in Table 1 and clearly show the impact of the anion in the crystal lattice on the spin state of the Mn 3+ cation. At 100 K, compounds 1 and 2, with higher volume tetrahedral anions, have bond lengths typical for an S = 2 species, while those for complex 4 with the Br − counterion are in the correct range for S = 1, in line with the SQUID data. At 100 K, bond lengths for nitrate and iodide complexes 3 and 5 suggest a mixture of spin triplet and quintet states, again in line with the magnetic data. Upon warming, the bond lengths of complex 3 show the expected equatorial elongation (see Table 1) and indicate the transition to a state with a higher percentage of the S = 2 species. Structural data were also collected at higher temperature (190 K) for complex 4, [MnL 1 ]Br, revealing the bond lengths to be similar to those of complex 3, [MnL 1 ]NO 3 , at 100 K. This is in good agreement with the magnetic data for complex 4, [MnL 1 ]Br, where the χ M T value at 190 K is 2.48 cm 3 K/mol, i.e., a value similar to that for complex 3 at 100 K, which shows an almost identical value of 2.40 cm 3 K/mol, indicating a 2:1 ratio of quintet:triplet sites. Comparison of the Mn−nitrogen (amine or imine) bond lengths in the pure S = 1 state, for example, those of complex 4 at 100 K, with equivalent donors in the pure S = 2 state, for example, those of complex 1 at 100 K (Table 1), provide a useful scale of the bond length difference equivalent to 100% spin conversion. Therefore, distances within that range can be used to make a good estimate of the relative percentage of the two spin states at any temperature. For example, SQUID data for complexes 3 and 4 at 100 and 190 K, respectively, indicate a 2:1 ratio of quintet/triplet states (Figure 2). This is in good agreement with the closeness of the Mn−nitrogen bond lengths for these complexes at the indicated temperatures (Table 1). In the case of complex 3, the stepped profile of the SCO suggests an ordered phase at 100 K, whereas the gradual profile in complex 4 suggests a Boltzmann distribution of spin states at 190 K.
Since compounds 1−5 crystallize isotypically in the orthorhombic space group P2 1 2 1 2, the difference in spin state must be due to packing and intermolecular interactions. In all cases the [MnL 1 ] + cations arrange in a parallel way forming 1D chains along the b-axis (see Figure S7). These chains pack into a zigzag formation due to the crystallographic symmetry elements of the orthorhombic P2 1 2 1 2 space group. We suggest that the consistent recovery of enantiopure individual crystals of either the Λ or Δ forms of the associated manganese complex with a range of achiral counterions may be due to the unique position and size of the ligand substituent in the 4-OMe-Sal 2 323 ligand, i.e., a methoxy group para to the phenolate donor. We also suggest that steric effects of this substituent arrangement disfavor packing of both the Λ and Δ enantiomers of the complex cation in the presence of the medium sized counterions reported here: ClO 4 − , BF 4 − , NO 3 − , Br − , and I − . With smaller counterions, including Cl − (compound 6, structure reported in the Supporting Information), or larger counterions including CF 3 SO 3 − , PF 6 − , and BPh 4 − , 105 the packing pattern is altered, and the crystallization of both enantiomers is observed, as is typical for this type of chiral complex with achiral anions.
Compounds 1 and 2 with tetrahedral perchlorate and tetrafluoroborate anions have closely related intermolecular interactions, with H-bonding between one anion and four neighboring cations leading to a 3D network (see Figure S8). A similar behavior is observed in the room-temperature structure of complex 3, [MnL 1 ]NO 3 (see Figure S9). Upon cooling, the packing arrangement shows a weakening of the H-bonds, but each anion still exhibits short contacts to four neighboring cations. Complex 4, [MnL 1 ]Br, exhibits intermolecular H−Br bonds at 100 K, where one bromide anion shows close contacts to three neighboring cations (see Figure S10), which increases to four neighboring cations upon warming to 190 K. Complex 5, containing the slightly bigger iodide anion, exhibits short contacts at 100 K to four neighboring cations (see Figure S11), in a fashion similar to that observed for compounds 1 and 2 and the room-temperature structure of complex 3.
Spin-crossover Mn 3+ compounds exhibit a stronger distortion of the octahedral environment in the S = 2 state than in the almost perfect octahedron associated with the S = 1 state due to loss of the Jahn−Teller distortion upon cooling, as the antibonding orbital is depopulated. The degree of distortion can be analyzed by the distortion parameters Σ and Θ, as defined by McKee et al., 136 where Σ highlights the angular deviation from the 90°cis-octahedral angles, and Θ measures the trigonal distortion from a perfect octahedral environment toward trigonal prismatic geometry. In the case of a perfect octahedron, both values are zero. The reported literature values for typical spin-crossover Mn 3+ compounds are Σ = 28−45°for S = 1 (Σ = 48−80°for S = 2) and Θ = 79−125°for S = 1 (Θ = 135−230°f or S = 2) 19 and these values can be used to help assign the spin state.
Σ and Θ have been calculated for 1−5 using OctaDist 2.6.1, 137 and the observed parameters (   (Figure 7, right). While most of the Schiff base ligand overlaps almost perfectly, there are small discrepancies visible in the amine backbone as well as the peripheral methoxy substituent. CD and UV−Vis Spectroscopy. During the X-ray structure collection and analysis of compounds 1−5 at various temperatures, each single crystal was internally enantiopure, but both enantiomers have been observed ( Table 3). The appearance of both the dextro, Δ, and laevo, Λ, enantiomers highlights that the bulk material consists of a conglomerate of both enantiomers. Within the seven structures determined over different temperatures, we have observed a close-to 50:50 ratio distribution of the two enantiomers Λ/Δ, as highlighted in Figure 8 and Table  3.
Given the chiral nature of the complex and the crystallization in a chiral space group, circular dichroism (CD) spectroscopy was recorded at room temperature. Initial attempts to directly measure individual crystals of all compounds in the solid state showed optical rotation, although the resolution was poor. A solution-state study was then completed on high-spin compound 1 for which there was the highest yield of single crystals and which therefore offered the best chance of statistical analysis. The enantiomeric forms were clearly distinguishable on measurements of methanol solutions of individual single crystals of 1 (Figure 9), indicating they do not racemize in solution. Statistical studies in methanol solution on three batches each of 10 single crystals indicate an approximate ratio of 2:1 of the two enantiomers within each batch, with an overall distribution of 20:10 from 30 investigated crystals. All the solution spectra are included in Figures S12−S14.
Solid-state UV−vis spectra were collected for complexes 1 and 5 at room temperature, i.e., in a temperature regime where [MnL 1 ] + is fully in the S = 2 form (complex 1) and stabilized with a mixture of S = 1 and S = 2 forms (complex 5). The spectra for both have strong features in the 200−400 nm range ( Figure  S15). In addition, a broad but weak shoulder around 525 nm was observed for both compounds ( Figure S15). The intensity ratio of the peaks at 225, 255, and 305 nm is different for the pure S = 2 sample (1) and the mixed S = 1/S = 2 sample, (5), suggesting that the more intense feature at 225 nm is associated with the S = 1 electronic state ( Figure S15).
Solution UV−vis spectroscopy of the free ligand H 2 L 1 ( Figure  S16) in acetonitrile confirms the origin of the peaks at 225, 255, and 305 nm as being ligand-based and demonstrates that coordination to the metal ion in different spin states can result in different ligand-to-metal charge transfer (LMCT) bands in the solid state ( Figure S15). Although the solution spectra of complexes 1 and 5 were recorded in methanol ( Figure S17), a meaningful comparison with the spin states is not possible as the SQUID data were recorded only in the solid state. However, some differences were detected, notably a change in intensity of the UV absorption at 250 nm and a new absorption at around 770 nm for iodide complex 5 ( Figure S17).

■ CONCLUSIONS
Isolation of a cationic Mn 3+ complex in lattices with five different counterions resulted in stabilization of the ion in either the S = 2 state (complexes 1 and 2) or promoted thermal spin crossover behavior (complexes 3−5). HFEPR was used to estimate the magnitude and sign of the axial D parameter in both spin states by recording low-temperature variable-frequency spectra on complexes 2 and 4. This confirmed that the spin quintet form is axially compressed with a D value of +6.9(3) cm −1 which increased to D = +23.38(1) cm −1 in the spin triplet form. Both spin triplet and axially compressed spin quintet electronic states are less common for Mn 3+

■ EXPERIMENTAL SECTION
General Experimental Details. Physical Measurements. All measurements were recorded on powdered samples of the respective polycrystalline compound. Elemental analyses (C, H, and N) were carried out using a PerkinElmer Vario EL instrument and mass spectra were recorded on a Waters 2695 Separations Module Electrospray Spectrometer. Solution and solid-state UV−vis spectra were recorded on an Agilent Cary 60 UV−vis spectrometer. Solution measurements were carried out on ∼1.5 × 10 −5 M methanolic solutions of samples 1 and 5. Solid-state samples of complexes 1, 2 and 5 were prepared for UV−vis by grinding a small number of crystals with a few drops of silicone oil. The resulting paste was thinly spread on the edge of a quartz cuvette for collection.
CD spectra were recorded on a JASCO J-810 spectrometer. Solutions were prepared by systematically taking a single crystal of 1 and dissolving it in methanol (1 mL). The process was repeated 30 times to acquire a large distribution of samples. Solid-state samples were prepared by grinding a single crystal of 1 with 50 mg of KBr which was pressed into a disk under 12.5 tonnes of pressure for 90 s. This was then pasted onto the side of a cuvette using silicone oil, and the CD of the solid-state disc was recorded. It is important to note that a variety of single crystals with different size and shape were chosen. All CD spectra were recorded with a scan rate of 100 nm/min. Powder X-ray diffraction (PXRD) experiments were carried out for complex 3, [MnL 1 ]NO 3 , using a Bruker D2 Phaser with Cu Kα radiation (λ = 1.5418 Å). The Table 3. Enantiomer Determination from Single Crystal Structural analysis on Seven Different Crystals Inorganic Chemistry pubs.acs.org/IC Article powder diffraction pattern of complex 3 is shown in section S3 of the Supporting Information. Magnetic Measurements. The magnetic susceptibility measurements were recorded on a Quantum Design SQUID magnetometer (MPMS-XL) operating between 1.8 and 300 K. Direct current measurements were carried out on polycrystalline samples prepared in gelatin capsules in a field of 0.1 T. Diamagnetic corrections were applied to correct for a contribution from the sample holder, and the inherent diamagnetism of the sample was estimated with the use of Pascal's constants.
High-Field EPR Spectroscopy. HFEPR spectra were recorded at the National High Magnetic Field Laboratory (NHMFL, Tallahassee, FL) using the homodyne transmission spectrometer equipped with a 15/17 T superconducting magnet. 138 Measurements were carried out on powder samples of 2 and 4 (∼30 mg) mixed with eicosane wax. The samples were packed into a Teflon sample holder. Spectra were recorded at different temperatures and multiple frequencies from 203 to 634 GHz in the 0 to 14.5 T field range.
Materials and Synthesis Procedures. Starting Materials. All chemicals and solvents if not otherwise mentioned were purchased from chemical companies and were reagent-grade. They were used without further purification or drying. All reactions were carried out under ambient conditions. All measurements were carried out on powdered samples of the respective polycrystalline compound.
Synthesis and Characterization of Complexes 1−5. Complex [MnL 1 ]BF 4 (2). The ligand solution of H 2 L 1 was added to a solution of MnCl 2 ·4H 2 O (0.046 g, 0.25 mmol) dissolved in 1:1 ethanol/acetonitrile (10 mL) together with NH 4 BF 4 (0.030 g, 0.3 mmol). The solution turned dark red (almost black) and was stirred for 10 min at rt. Any precipitate was filtered off afterwards, and the reaction was left for slow evaporation. After a few days, small dark red−purple thin plates of crystals were isolated by filtration. Mass spectrometry (g/ mol): expected: 495. 18     Crystallography. Crystal Data Collection and Refinement. Suitable single crystals of complexes 1−5 were mounted on an Oxford Diffraction Supernova A diffractometer fitted with an Atlas detector; data sets were measured using monochromatic Cu Kα radiation or Mo Kα radiation and corrected for absorption. 139 The temperature (100 K) was controlled with an Oxford Cryosystem instrument. Structures were solved by direct methods (SHELXS) 140 and refined with full-matrix least-squared procedures based on F 2 , using SHELXL-2016. Nonhydrogen atoms were refined with independent anisotropic displacement parameters; organic H atoms, i.e., bound to C or N or OH groups from methanol, were placed in idealized positions. Not all hydrogen atoms of disordered solvents could be located. The hydrogen atoms of water molecules that could be located were first located in the difference Fourier map. In subsequent refinements the O−H bond lengths (0.84 Å) and H···H distances (1.33 Å) were restrained to their ideal values. After convergence, the water molecules were refined as rigid groups. Selected crystallographic data and structure refinements are summarized in Table S3, and crystallographic data for the structures reported in this paper have been deposited with the Cambridge Crystallographic Data Centre as supplementary publication numbers CCDC 2105399− 2105406. Copies of the data can be obtained free of charge from https://www.ccdc.cam.ac.uk/structures/.
The SI contains (i) synthesis details for compound 6, (ii) powder X-ray analysis on complex 3, [MnL 1 ]NO 3 , (iii) temperature dependence of the χ M T products at 1000 Oe for compound 3, [MnL 1 ]NO 3 , in cooling and warming mode, (iv) the first derivative of χ M T against the temperature, simulation of χT using the parameters observed from HFEPR for 5, (v) experimental resonance fields observed for 2 and assigned to the S = 2 state and to the S = 1 state, (vi) experimental HFEPR spectrum of 2 recorded at 5 K at a microwave frequency of 633 GHz, (vii) packing arrangement of compound 5, (viii) additional information about the DFT calculations, (ix) experimental and calculated donor bond lengths (Å) in complex 4, (x) individual contributions to the D '◊ -tensor, (xi) hydrogen bonding network found within complexes 1 and 3−5, (xii) CD spectra of all single crystals of complex 1, (xiii) solid-state (Nujol mull) UV−vis spectra of complexes 1 and 5 at room temperature, (xiv) solution UV−vis spectroscopy of the ligand H 2 L 1 , (xv) crystallographic details for compounds 1−6 (PDF)

Notes
The authors declare no competing financial interest.