Polymorphism in a π stacked Blatter radical: structures and magnetic properties of

3-(Phenyl)-1-(pyrid-2-yl)-1,4-dihydrobenzo[ e ][1,2,4]triazin-4-yl ( 2 ) demonstrates the first example of polymorphism in the family of Blatter radicals. Two polymorphs, 2 α and 2 β , have been identified and characterized by single crystal X-ray diffractometry and magnetic susceptibility measurements to investigate their magnetism – structure correlations. Both polymorphs form one-dimensional (1D) π stacks of evenly spaced radicals with distinctly different π – π overlap modes. Within the 1D π stacks, radicals are located at evenly interplanar distances, 3.461 Å for 2 α and 3.430 Å for 2 β . Magnetic susceptibility studies indicate that both polymorphs exhibit antiferromagnetic interactions inside their 1D π stacks. The magnetic susceptibility data are best interpreted in terms of a regular chain model of antiferromagnetically coupled quantum spins H ¼ − 2 J X i S i ! · S i þ 1 !! ! ! with exchange-interactions of J / k B = − 36.7(3) K ( − 25.5(2) cm − 1 ) for 2 α and J / k B = − 72(3) K ( − 50(2) cm − 1 ) for 2 β . For polymorph 2 β , a crossover on the magnetic susceptibility around 20 K suggests the presence of a phase transition, which might be related to dimerization of the radicals along the chain. DFT calculations support the experimental structure – magnetism results and the antiferromagnetic nature of the local interactions between radicals within the 1D π stacks. Two polymorphs, 2 α and 2 β , were identified, isolated and characterized by means of single crystal X-ray diffractometry and magnetic susceptibility measurements. Polymorph 2 α shows a unique mode of overlap along the 1D stacking direction where the N1-(pyrid-2-yl) substituent interacts face-to-face (eclipsed conformation) with a 1,2,4-triazinyl ring of a subsequent radical. Polymorph 2 β forms a slipped 1D π stack wherein the radicals overlap extensively over the benzotriazinyl rings.


EPR and cyclic voltammetry
The solid-state and solution EPR spectra (CH 2 Cl 2 , ca. 20°C) of radical 2 have been previously reported, and we summarize the data here. 13 The solution EPR spectrum of radical 2 is typical of benzotriazinyls with the largest 14 N hyperfine coupling constant (hfcc) located at N1 followed by N4 and N2 (a N1 ≫ a N4 > a N2 ). 4 The experimentally determined hyperfine coupling constants for radical 2 are a N1 (6.74 G), a N2 (4.88 G), a N4 (4.9. G) with g solution = 2.0040 and g solid = 2.0046 which compare well with the values deduced from the modelling of the magnetic susceptibility data (vide infra). Cyclic voltammetry (CV) measurements of radical 2 (1 mM in CH 2 Cl 2 containing n-Bu 4 NBF 4 (0.1 M) as electrolyte, Ag/AgCl as reference electrode, 50 mV s −1 scan rate, ca. 20°C, Fc/Fc + as internal reference), show two fully reversible oxidation E 1/2 (ox) = 0.24 V and reduction waves E 1/2 (red) = −0.82 V and E cell = 1.06 V.

Single crystal and powder X-ray diffractometry
Single crystal X-ray diffraction data were collected on an Agilent SuperNova diffractometer, equipped with an Atlas detector and Cu Kα radiation source (λ = 1.5418 Å). Suitable crystals were attached to MiTeGen micro-mounts with Fomblin ® Y oil and transferred to a goniostat where they were cooled for data collection. Unit cell dimensions were determined and refined by using 2674 reflections (4.75 ≤ θ ≤ 74.49°) for polymorph 2α and 2797 (4.50 ≤ θ ≤ 76.65°) for polymorph 2β. Data acquisitions, reductions and empirical absorption corrections were applied using CrysAlis PRO software. 69 The structures were solved by direct methods and refined on F 2 using full-matrix least squares using SHELXL. 70,71 The non-H atoms were treated anisotropically. The hydrogen atoms were placed in calculated, ideal positions and refined as riding on their respective carbon atoms. Crystallography figures were generated using Mercury. 72 Powder X-ray diffraction (PXRD) patterns for samples 2α and 2β were recorded on a Shimadzu 6000 Series X-ray diffractometer at room temperature (Cu Kα radiation, λ = 1.5418 Å).

Magnetic susceptibility measurements
Magnetic measurements were performed on a Quantum Design SQUID magnetometers MPMS-XL (Quantum Design, San Diego, CA, USA) and MPMS3-VSM at temperatures between 1.8 and 300 K, and dc magnetic fields ranging from −7 to +7 T. The measurements were carried out on polycrystalline samples (22.21, 21.13 and 21.6 mg for 2α, and 21.35, 15.6, 14.5 and 15.1 mg for 2β) introduced in a sealed polyethylene bag (3 × 0.5 × 0.02 cm; typically, 18-22 mg) or gelatin capsules. Prior to the main experiments, the field-dependent magnetization was measured at 100 K on each sample to detect the possible presence of any bulk ferromagnetic impurities. Paramagnetic materials should exhibit a perfect linear dependence of magnetization that extrapolates to zero at zero dc field and all samples appeared to be free of any bulk ferromagnetic impurities. The magnetic data were corrected for the sample holder and intrinsic diamagnetic contributions. 73

Computational methodology
Exchange coupling constants were calculated using broken symmetry density functional theory (BS-DFT) by mapping the energies of the calculated states to the diagonal elements of the Heisenberg-Dirac-van Vleck Hamiltonian H = −2J comp Ŝ 1 ·Ŝ 2 , where J comp is the calculated exchange-coupling constant and Ŝ 1 and Ŝ 2 are spin operators acting on two spin sites. 74 The energies of the broken symmetry singlet and triplet states were determined by single point calculations using geometries extracted from crystal structures. Seven different functionals, namely B3LYP, X3LYP, CAM-B3LYP, ωB97x, LC-ωHPBE, M06-2X and M15, were employed together with the large def2-TZVPP basis sets. 75 B3LYP is the classic three-parameter hybrid functional consisting of Becke's 88 exchange functional 76 and the correlation functional of Lee, Yang and Parr, 77 whereas X3LYP replaces Becke's 88 exchange with an improved functional developed to provide a better description of non-bonded interactions, spin states and thermochemical properties. 78 CAM-B3LYP is a hybrid functional which combines B3LYP with a long-range correction based on the Coulomb-attenuating method. 79 ωB97X 80 is a long-range corrected functional based on Becke's work, whereas LC-ωHPBE 81 is Henderson's version of the long-range-corrected LC-ωPBE functional of Vydrov. 82-84 M06-2X 85,86 is a global hybrid with 54% HF exchange and empirically parameterized only for nonmetals. MN15 87,88 is a newer version of M06 with 44% HF exchange and parameterized for multi-reference systems and noncovalent interactions. All calculations were performed with Gaussian16 89 using XSEDE 90 resources and services.

Crystal structures
Radical 2 crystalizes in two polymorphs 2α (CCDC 1955680) and 2β (CCDC 1955684). Single crystals suitable for X-ray diffraction studies were obtained by slow cooling of a dilute and concentrated n-hexane solution for 2α and 2β, respectively. Careful recrystallization is required to avoid crystallization of both polymorphs as mixtures. Polymorph 2β comes out rapidly from a hot super saturated solution as it cools down to room temperature and is, tentatively, the kinetic polymorph or a metastable kinetic polymorph. Polymorph 2α comes out of solution slowly once at room temperature and is, tentatively, the thermodynamic polymorph. Both crystal structures were collected at 100(2) K. Polymorph 2α crystallizes in the orthorhombic space group P2 1 2 1 2 1 and polymorph 2β in the monoclinic space group P2 1 /c. Both polymorphs contain one molecule in the asymmetric unit. The intramolecular bond angles and bond lengths are similar to that of other benzotriazinyls, [21][22][23][24][25][26][27] however, there is a significant difference in the geometry of the amidrazonyl moiety. In polymorph 2α, the 1,2,4amidrazonyl moiety adopts a shallow boat conformation with deviations of the N1 and N3 atoms from the mean plane of C2, C3, N2, C1 of 0.09 and 0.06 Å, respectively (Fig. 2, top). A similar amidrazonyl structure was observed in 1,3-diphenyl-7trifluoromethyl-1,4-dihydrobenzo[e] [1,2,4]triazin-4-yl. 10 The 1,2,4-amidrazonyl moiety of polymorph 2β is closer to planarity with deviation of the N1 and N3 atoms from the mean plane of C2, C3, N2, C1 of 0.06 and 0.03 Å, respectively (Fig. 2, bottom). The torsion angle (C3, N1, C14, N4) of the N1-(pyrid-2-yl) group with respect to the plane of benzotriazine is similar for both polymorphs [38.9(3) and 36.4(2)°for 2α for 2β, respectively] and significantly less than the average of 59 ± 13°reported thus far. 19 This torsion angle is the result of steric repulsion between the H4 and the lone pair of N4. The torsion angle (N3, C1, C8, C9) between the C3-phenyl and the amidrazonyl plane is 15.3(3)°and 0.8(2)°f or 2α and 2β, respectively. Despite subtle differences in the intramolecular geometrical parameters of the radicals in CrystEngComm Paper these two polymorphs, their solid-state packing presents some striking distinctions in the way these radicals associate and π stack. Most benzotriazinyls form 1D supramolecular arrangements wherein the radicals π stack to obtain efficient SOMO-SOMO overlap. [21][22][23][24][25][26][27][28][29][30][31][32][33] This is also the case for both polymorphs 2α and 2β, and is attributed to the presence of the spin density primarily on the amidrazonyl unit (ca. 70%) and to a lesser degree on the fused benzene ring and the N1-(pyrid-2-yl) substituent (vide infra).
Solid-state packing of polymorph 2α. Polymorph 2α π stacks along the a-axis and forms supramolecular chains of evenly spaced radicals (Fig. 3, left). A 2-fold screw axis in the [1, 0, 0] direction at x, 1/4, 0 and screw component [1/2, 0, 0] places the N1-(pyrid-2-yl) substituent directly on top of a triazine ring (eclipsed conformation) of a subsequent radical inside the π stack to form a "head-to-tails" dimer ( Fig. 3, left). This packing is unique as most benzotriazinyls overlap in either a centrosymmetric manner or via translation parallel to the stacking direction. The centroid distance between these two N1-(pyrid-2-yl) and 1,2,4-triazine rings is 3.48 Å. There are three pairs of close intermolecular contacts between radicals inside the π stack, C16⋯C2 [d  Solid-state packing of polymorph 2β. Radicals in polymorph 2β π stack along the b-axis to form supramolecular chains of evenly spaced radicals (Fig. 3, right). The molecules inside the stack are related via a glide plane perpendicular to [0, 1, 0] with glide component [0, 0, 1/2] packing in a "head-to-head" orientation. This results in a slipped π stack wherein the radicals are not eclipsed but overlap with slippage angles of 76.45°(longitudinal) and 70.75°(latitudinal). The interplanar distance along the supramolecular chains (defined as the distance between subsequent planes of benzotriazinyl rings) is 3.432 Å (Fig. 3, right). The shortest contact inside the π stack is between carbons of the N1-(pyrid-2-yl) substituent C14⋯C15
The experimental diffraction signatures of polycrystalline samples at 300 K for 2α and 2β match well the patterns calculated from single crystal X-ray structures at 100 K. Additionally, the comparison of powder X-ray diffraction patterns for polymorphs demonstrates their phase purity.
However, the presence of small amounts of amorphous paramagnetic impurities cannot be fully excluded.

Magnetic properties
The temperature dependence of the magnetic susceptibility (χ) was collected on polycrystalline samples of 2α and 2β in the 2-300 K temperature region. The representative data are shown as χ vs. T and χT vs. T plots in Fig. 6. The χT product at 300 K is 0.34 and 0.29 cm 3 K mol −1 for 2α and 2β, respectively. These values are significantly smaller than the expected Curie constant of 0.375 cm 3 mol −1 K expected for an S ¼ 1 2 radical species with a g factor of 2. This apparent discrepancy is induced by the presence of significant antiferromagnetic interactions between radical molecules as confirmed by the marked decrease of the χT product (or the broad maximum of the magnetic susceptibility at 47 and 100 K, respectively) when decreasing the temperature. At 2 K, if one considers that the χT product should be null when all the radical spins are fully antiferromagnetically coupled, the observed residual paramagnetism of 0.004 and 0.006 cm 3 K mol −1 , respectively, corresponds to about 1 and 2% of an Curie impurity. Based on the crystal structures shown above, the strongest antiferromagnetic interactions should be present along the regular chain of radicals in both polymorphs (Fig. 3). The magnetic susceptibility data were thus modeled using a regular chain of S ¼ 1 2 quantum spins with a single magnetic interaction, J, between radical centers While the regular chain model is perfect for the magnetic data of 2α, the theory/experiment agreement for 2β is obviously less performant. Hence, alternative spin chain models with two different magnetic interactions have also been considered without being able to significantly improve the agreement. It is thus suspected that interchain magnetic interactions are indeed effective in polymorph 2β. Below 20 K (Fig. 6, inset), a clear anomaly is observed on the susceptibility of polymorph 2β, indicating a possible phase Fig. 5 Powder X-ray diffraction patterns of samples used for magnetic measurements: (top) 2α collected at 300 K (red line) and calculated from the single crystal X-ray structure at 100 K (blue line), (middle) 2β collected at 300 K (red line) and calculated from the single crystal X-ray structure at 100 K (blue line) and (bottom) comparison of the experimental powder X-ray diffraction patterns at 300 K for the two polymorphs 2α (blue line) and 2β (red line).

CrystEngComm Paper
transition. Attempts to collect the crystal structure of polymorph 2β at 3 K led to diffraction degradation in one direction which appeared to be reversible. The magnetic and X-ray diffraction experimental data suggested a phase transition below 17 K that most likely involves a dimerization of the radicals along the chain as observed in many related 1D spin systems. 94-99

DFT and ab initio calculations
Magnetism-structure correlations can be further supported by means of quantum chemical methods. 100 This is typically done either at post-HF levels or using BS-DFT of which the latter is a computationally efficient but theoretically lessrigorous approach that is typically employed for systems containing several tens, even hundreds, of atoms. Table 1 lists the exchange-coupling constants calculated for polymorphs 2α and 2β using different density functionals. The data can be compared to the experimentally determined radical⋯radical interactions, −25.5 and −50 cm −1 for polymorph 2α and 2β, respectively. In general, all functionals predict the coupling in 2α to be antiferromagnetic but the coupling strength varies greatly between different functionals. Most notably, functionals with improved long-range corrections, namely CAM-B3LYP, ωB97x and LC-ωHPBE, predict significantly weaker antiferromagnetic coupling compared to others, which, surprisingly, show only small variation and are in good agreement with the experimentally derived exchange coupling. The results for 2β are, however, more varied with some of the theoretically more just longrange corrected functionals even predicting the coupling to be ferromagnetic and, thus, at variance with experimental observations. It is also notable that even the best-performing functionals give J comp for 2β that is almost half (in magnitude) of the experimentally derived intrachain exchange coupling J. The reason for this discrepancy is unclear, though it can be related to problems in treating the singlet state via BS-DFT and it also parallels the problems observed in modelling the magnetic data of 2β with the regular chain model. It should also be noted that by simple examination of the SOMO and the associated spin density of radical 2 (Fig. 7), it is evident that the benzo triazinyl ring contains most of the spin density and overlap through this region, like that in polymorph 2β, should lead to strong antiferromagnetic exchange interaction unless prevented by appropriate stack slippage. However, if the radicals pack so that the interactions involve the N1-(pyrid-2-yl) substituents, such as in polymorph 2α, the antiferromagnetic interaction is expected to be weaker even in case of perfect stacking due to less efficient overlap.

CrystEngComm Paper
Two polymorphs, 2α and 2β, were identified, isolated and characterized by means of single crystal X-ray diffractometry and magnetic susceptibility measurements. Polymorph 2α shows a unique mode of overlap along the 1D stacking direction where the N1-(pyrid-2-yl) substituent interacts faceto-face (eclipsed conformation) with a 1,2,4-triazinyl ring of a subsequent radical. Polymorph 2β forms a slipped 1D π stack wherein the radicals overlap extensively over the benzotriazinyl rings.
Magnetic susceptibility studies reveal that in both cases the radical⋯radical interactions are antiferromagnetic. The intra-chain magnetic exchange interaction of polymorph 2β (−50 cm −1 ) was found to be double that of polymorph 2α (−25.5 cm −1 ), possibly due to more effective SOMO-SOMO overlap. The two low dimensional polymorphs of radical 2 demonstrate weak antiferromagnetic interactions often observed in organic open-shell molecules. 101,102 While recent advances in the chemistry of organic radicals led to some materials exhibiting substantial magnetic hysteresis (primarily in thiazyl radicals), 103-107 this class of Blatter radicals have yet to demonstrate their efficiency to generate large magnetic couplings and thus magnetic order at high temperature as observed often in purely inorganic systems and in a few metal-organic materials. 108 Polymorphism in Blatter-type radicals could be more prevalent than heretofore recognized and requires careful examination of the harvested crystals. We are currently working on other examples of a Blatter-type radicals that demonstrate polymorphism.

Conflicts of interest
The authors declare no conflicts of interest.