Thirty Years of Hide-and-Seek: Capturing Abundant but Elusive MIII@C3v(8)-C82 Isomer, and the Study of Magnetic Anisotropy Induced in Dy3+ Ion by the Fullerene π-Ligand

Our knowledge about endohedral metallofullerenes (EMFs) is restricted to the structures with sufficient kinetic stability to be extracted from the arc-discharge soot and processed by chromatographic and structural techniques. For the most abundant rare-earth monometallofullerene MIII@C82, experimental studies repeatedly demonstrated C2v(9) and Cs(6) carbon cage isomers, while computations predicted equal stability of the “missing” C3v(8) isomer. Here we report that this isomer is indeed formed but has not been recovered from soot using standard protocols. Using a combination of redox extraction and subsequent benzylation and trifluoromethylation with single-crystal XRD analysis of CF3 adduct, we prove that Dy@C3v(8)-C82 is one of the most abundantly produced metallofullerenes, which was not identified in earlier studies because of the low kinetic stability. Further, using the Dy@C3v(8)-C82(CF3) and Dy@C3v(8)-C82(CH2Ph) monoadducts for the case study, we analyzed the role of metal-fullerene bonding on the single-ion magnetic anisotropy of Dy in EMFs. The multitechnique approach, combining ab initio calculations, EPR spectroscopy, and SQUID magnetometry, demonstrated that coordination of the Dy ion to the fullerene cage induces moderate, nonaxial, and very fluid magnetic anisotropy, which strongly varies with small alterations in the Dy-fullerene coordination geometry. As a result, Dy@C3v(8)-C82(CH2Ph) is a weak field-induced single-molecule magnet (SMM), whose signatures of magnetic relaxation are detectable only below 3 K. Our results demonstrate that metal-cage interactions should have a detrimental effect on the SMM performance of EMFs. At the same time, the strong variability of the magnetic anisotropy with metal position suggests tunability and offers strategies for future progress.


Experimental and Computational Details
Synthesis.Dy-EMFs were synthesized by direct current arc discharge method.Specifically, a mixture of Dy2O3 and graphite powder (C) with a molar ratio of Dy:C = 1:7.5 was filled into the core-drilled graphite rods, which were then evaporated in 180 mbar He atmosphere with a direct current of 100 A. Endohedral metallofullerenes were extracted from the synthesized soot by N, N-Dimethylformamide (DMF).The DMF solution of EMF anions were reacted with benzyl bromide or Umemoto reagent II, as described previously. 1,2 he products were separated by high-performance liquid chromatography (HPLC) using toluene as the mobile phase.HPLC.HPLC separation was performed for toluene solutions of fullerene and with toluene as an eluent, employing analytical or semipreparative COSMOSIL Buckyprep chromatographic columns (Nacalai Tesque) and Agilent 1260 Infinity II LC System.Recycling HPLC separation was performed using the Sunflow 100 system (SunChrome).
Mass-spectrometry.Matrix assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectra were measured with a Bruker autoflex mass spectrometer.
UV-Vis-NIR.UV-vis-NIR absorption spectra were measured in carbon disulfide solution at room temperature with Shimadzu 3100 spectrophotometer.
Nuclear magnetic resonance (NMR) spectroscopy. 1H and 19 F NMR spectra were measured with 500 MHz Avance II spectrometer (Bruker) in CS2 solution.
X-ray diffraction.Single crystals were obtained by co-crystallization of Dy@C82(CF3) in toluene with nickel octaethylporphyrin (NiOEP) in benzene.X-ray diffraction data collection was carried out at the BL14.2 beamline (BESSY storage ring, Berlin-Adlershof, Germany). 3XDSAPP2.0suite was employed for data processing. 4,5 he structure was solved by direct methods and refined by SHELXL-2018. 6Hydrogen atoms were added geometrically and refined with a riding model.The crystal data are presented in Table S2.
Electron paramagnetic resonance (EPR) measurements were carried out on a pressed powder and oterphenyl solution of Dy@C3v(8)-C82(CH2Ph) at a fixed microwave frequency of ν = 9.56 GHz and at temperatures from 5 to 100 K using a commercial Bruker X-band spectrometer equipped with a 4 He gas flow cryostat from Oxford Instruments.The magnetic field was swept from 0 to 300 mT for most of the measurements.The EPR spectra were recorded as an absorption derivative.
Magnetization of powder samples was measured using a Quantum Design VSM MPMS3 magnetometer.In AC measurements, the oscillation amplitude was 2.5 Oe.Magnetic simulations were performed using PHI code 7 and employing powder-averaging to account for random orientation of molecules in experimentally-measured samples; single-ion ligand-field parameters were obtained in CASSCF calculations.
Computational studies.DFT calculations were first performed for Y analogs, molecular structures were optimized at the PBE 8 level using the fast DFT Priroda code 9, 10 with the implemented basis set of TZ2P quality with SBKJC-type effective core potential for Y atoms.The same code was used for transition state (TS) search and intrinsic reaction coordinate (IRC) calculations.DFT-level Born-Oppenheimer molecular dynamics were performed using in-house Python scripts and Atomic Simulation Environment libraries. 11Energies and gradients at each time step were evaluated at the PBE/TZ2P level with the Priroda code.The equation of motion was integrated for 25 ps with a step size of 1.0 fs, utilizing a canonical ensemble statistic and Nosé-Hoover thermostat for temperature control.
DFT calculations with Dy were performed using the Orca package. 12,13 or lanthanides with non-zero orbital momentum, all-electron DFT calculations can give ambiguous results because of the non-singledeterminant wavefunction.To avoid potential problems, we used 4f-in-core effective core potentials of ECPXXMWB by Dolg et al. with corresponding ECPXXMWB-II basis sets. 14,15 or C and F atoms, def2-TZVPP basis was used. 16The same code and basis sets were used for single-point calculations of energies with hybrid functionals PBE0 and B3LYP.
For consistency with earlier works, molecular structures for CASSCF calculations were optimized at the PBE/PAW level using the VASP code and recommended pseudopotentials with f-shell in-core treatment; 8,[17][18][19][20] the structures were very similar to those optimized with molecular codes.Ab initio energies and wave functions of Dy 3+ LF multiplets have been calculated at the CASSCF/SO-RASSI level of theory using the quantum chemistry package OpenMOLCAS 21 and SINGLE_ANISO module. 22The basis sets were ANO-RCC-VTZP for Dy and ANO-RCC-VDZP for other elements.To make axial term decomposition, we used ligand-field parameters calculated with the OPEN_SINGLE_ANISO module to construct a crystal-field Hamiltonian.Having constructed the Hamiltonian for different compositions, such as  ̂2  only or  ̂2  + ̂4  , etc., we estimated the weights of these compositions in terms of Euclidean norms. 22lecular structures, trajectories, and isosurfaces were visualized with VMD. 23olecular structures and relative energies of M III @C82 isomers Figure S1.DFT-optimized molecular structures of Dy@C82 isomers and their relative energies (PBE0//PBE level).Each structure is shown in two orientations.

Single-crystal X-ray diffraction
The asymmetric unit contains one NiOEP, one fullerene molecule, and two toluene molecules.Like the prevalent behaviour of fullerene-NiOEP co-crystals, the fullerene molecules were hold by NiOEP as shown in Figure S5.All molecule are fully ordered, the symmetry of fullerene cage is determined as C3v( 8)-C82.The encapsulated Dy is fully ordered, locating beneath one hexagon as shown in Figure S6.The CF3 functional group attaches to a hexagon-hexagon-pentagon (HHP) junction sitting on the mirror symmetry of fullerene cage, resulting in the Cs symmetry of C3v(8)-C82(CF3). .b opt PBE -optimized at the PBE/TZ2P level with SBKJC-type ECP for Y, Priroda code; c optimized at the PBE/def-TZVPP level with 4f-in-core ECP for Dy, Orca code; d Single-point calculations with PBE0 functional using geometry optimized at the PBE/def2-TZVPP level, basis set for Dy includes Dolg's 4f-in-core ECP, Orca code; e Single-point calculations with B3LYP functional using geometry optimized at the PBE/def2-TZVPP level, basis set for Dy includes Dolg's 4f-in-core ECP, Orca code; f Single-point calculations with B3LYP functional and D3BJ correction using geometry optimized at the PBE/def2-TZVPP level, basis set for Dy includes Dolg's 4f-in-core ECP, Orca code; Isomers of Dy@C3v(8)-C82(CF3) Figure S12.17 regioisomers of Dy@C3v(8)-C82(CF3) and their relative energies (in kJ mol −1 ) computed at the PBE0//PBE level.These relative energies were used in preparing the Figure 5a in the main text Calculations were first performed for Y. Metal atom was placed in the center of the cage and allowed to migrate to the preferable position during optimization.In most cases optimized were positions close to that in the non-functionalized Dy@C3v(8)-C82.In few cases, when different metal positions were obtained, re-optimization with metal shifted to the position as in bare Dy@C3v(8)-C82 showed lower energy.Thus, we can conclude one CF3 groups does not strongly affect he preferable position of the metal.The Y@C3v(8)-C82(CF3) structures were then re-optimized at the PBE/def-TZVPP level with Dy (including Dolg's 4f-in-core ECP for Dy), and then single-point energies were calculated with PBE0 functional.Relative energies obtained with different functionals are listed in Table S4.

Figure S6 .
Figure S6.Structure of Dy@C3v(8)-C82(CF3)•NiOEP•2Toluene showing the Dy coordinated fragment viewed from to perpendicular directions.The Dy-C bond lengths were compiled in the table.The distance between Dy and the coordinated hexagon is 1.9527(6) Å.The thermal ellipsoid probability was set at 30%.Color code: grey for C and green for Dy.

Figure S7 .Figure S9 .
Figure S7.Structure of Dy@C3v(8)-C82(CF3)•NiOEP•2Toluene.Solvent molecules and fullerene cage carbons are omitted for clarity to show the metal positions relative to the NiOEP.The thermal ellipsoid probability was set at 30%.Color code: grey for C, white for H, blue for N, green for Dy, and red for Ni.

CASSCF calculations of LF splittingFigure S13 .
Figure S13.Orientations of principal gz axes of four lowest Kramers doublets (KDs) in three conformers of Dy@C3v(8)-C82(CF3); principal gz-axes are shown in red (KD1), cyan (KD2), magenta (KD3), and violet (KD4), axis of the ground-state doublet KD1 is highlighted by a larger size, carbon atoms with Dy-C distance shorter 2.5 Å are shown in light green.A perpendicular orientation of molecules is chosen versus the one shown in Figure 6 of the main text.

Figure S14 .
Figure S14.Normalized X-band EPR spectra of Dy@C3v(8)-C82(CH2Ph) measured at different temperatures in glassy frozen solution in o-terphenyl.Broadening of the signal with temperature can be clearly seen in this representation.

Table S3 .
Relative energies (kJ mol −1 ) of conformers of M@C3v(8)-C82(CH2Ph) a for each metal position, rotation of benzyl group around the Cfullerene-C(H2) bond gives three additional rotamers, one with Ph group above the fullerene pentagon, and two with Ph group above fullerene hexagons; a