Nd─Nd Bond in Ih and D5h Cage Isomers of Nd2@C80 Stabilized by Electrophilic CF3 Addition

Abstract Synthesis of molecular compounds with metal–metal bonds between 4f elements is recognized as one of the fascinating milestones in lanthanide metallochemistry. The main focus of such studies is on heavy lanthanides due to the interest in their magnetism, while bonding between light lanthanides remains unexplored. In this work, the Nd─Nd bonding in Nd‐dimetallofullerenes as a case study of metal–metal bonding between early lanthanides is demonstrated. Combined experimental and computational study proves that pristine Nd2@C80 has an open shell structure with a single electron occupying the Nd─Nd bonding orbital. Nd2@C80 is stabilized by a one‐electron reduction and further by the electrophilic CF3 addition to [Nd2@C80]−. Single‐crystal X‐ray diffraction reveals the formation of two Nd2@C80(CF3) isomers with D5h‐C80 and Ih‐C80 carbon cages, both featuring a single‐electron Nd─Nd bond with the length of 3.78–3.79 Å. The mutual influence of the exohedral CF3 group and endohedral metal dimer in determining the molecular structure of the adducts is analyzed. Unlike Tb or Dy analogs, which are strong single‐molecule magnets with high blocking temperature of magnetization, the slow relaxation of magnetization in Nd2@Ih‐C80(CF3) is detectable via out‐of‐phase magnetic susceptibility only below 3 K and in the presence of magnetic field.


Experimental details
HPLC: HPLC analysis and separation were performed for toluene solutions of fullerene and with toluene as an eluent, employing semipreparative COSMOSIL Buckyprep and Buckyprep-D chromatographic columns (Nacalai Tesque) and Agilent 1260 Infinity II LC System.Recycling HPLC separation was performed using Sunflow 100 system (SunChrome).
NMR spectrometry: 19 F and 13 C NMR spectra were measured with 500 MHz Avance II spectrometer (Bruker) in CS2 solution.
UV-Vis spectrometry: UV-vis-NIR absorption spectra were measured in CS2 solution at room temperature with Shimadzu 3100 spectrophotometer.
Vibrational spectroscopy.FT-IR spectra were recorded at room temperature on a Vertex 80 FT-IR spectrometer (Bruker) equipped with Hyperion microscope.The samples were drop-casted from toluene solution onto KBr substrate and measured in transmission mode.The same samples but cooled down to 78 K were used in Raman measurements performed with T64000 spectrometer (Horiba) and laser excitation at 532 nm (Torus laser by Laser Quantum).
X-ray diffraction.Single crystal X-ray diffraction data collection was carried out at 100 K at the BESSY storage ring (BL14.2,Berlin-Adlershof, Germany). 1 XDSAPP2.0suite was employed for data processing. 2,3 e structure was solved by direct methods and refined by SHELXL-2018. 4Hydrogen atoms were added geometrically and refined with a riding model.Magnetic measurements.DC and AC magnetic measurements of powder samples were performed using a Quantum Design VSM MPMS3 magnetometer.The sample was drop-casted from solution into standard propylene sample capsule.Magnetic simulations were performed using PHI code. 5In AC magnetometry studies, the in-phase and out-of-phase susceptibility, and , were measured as a function of oscillation frequency and then fitted with the generalized Debye model: where is the relaxation time.The linear term ( ) is included to correct for a linear background that was found in measurements at high frequency (the background has noticeable contribution to curves only above > 100 Hz).
DFT computations.DFT calculations were performed at the PBE level 6 using the Orca package. 7,8 or lanthanides with non-zero orbital momentum, all-electron DFT calculations can give ambiguous results because of the non-single-determinant wavefunction.To avoid potential problems, we used 4f-in-core effective core potentials of ECPXXMWB of Dolg et al. with corresponding ECPXXMWB-II basis sets. 9,10 or C and F atoms, def2-TZVPP basis was used. 11Molecular structures and isosurfaces were visualized with VMD. 12 DFT calculations of M2@C80 in different electronic states Table S1a.Relative energy of pT state (in eV) versus pS state in M2@Ih-C80 and M2@D5h-C80 series.

M
M2@Ih-C80 The structure in pS and pT state were optimized at the PBE level with def2-TZVPP basis for carbon and Dolg's 4f-incore MWB-II basis sets with core effective potentials for lanthanides, the relative energies obtained are denoted as "PBE"; singe-point calculations were then performed with the same basis using PBE0 functional, the energies are denoted as "PBE0//PBE".
Nd2@Ih-C80 and Ce2@D5h-C80 were chosen to benchmark density functionals.Based on experimental evidence, pT state should be lower in energy than pS for the former, whereas pS state should be more stable for the latter.Table S1b compares relative energies computed with various functionals, including hybrid (PBE0, B3LYP, O3LYP, X3LYP, B3PW), meta-hybrid (TPSSh, TPSS0, M06, and M06-2X), and two rangeseparated hybrid functionals (CAM-B3LYP and ω-B97X).As can be concluded from Table S1b, none of the functionals gives correct order of pS and pT energies for two molecules at once, while M06 and M06-2X also converged to pT with different spin distribution, which resulted in the strong deviation of the energies.
Other functional give close predictions, and we use PBE0 data in Fig. 2 in the main text.

Figure S1 (continued)
. HOMO and LUMO of selected pS-M2@Ih-C80 and pS-M2@D5h-C80 at the PBE0//PBE level.In the pT state of Nd2@C80, one SOMO is localized on the Nd2 dimer, and one on the fullerene cage.Accordingly, the spin density has an equal contribution on Nd2 dimer and on the fullerene cage.Upon single-electron reduction to [Nd2@C80] − , the fullerene SOMO is populated by the second electron and turns into doubly-occupied MO, hence the fullerene cage attains a closed-shell electronic structure, leaving Nd2 dimer with a single-electron Nd-Nd bond.The Nd-Nd bond is preserved in the neutral Nd2@C80(CF3) adduct.

Mass-spectra of EMF extracts
. MALDI-TOF mass-spectra of Nd-EMF and Pr-EMF extracts in CS2 in the range of M@C82 and M2@C80 species: (a) negative ion mode, (b) positive ion mode.Insets show isotopic distribution and simulated mass-spectra.Note that mass-spectra of Nd@Cn almost overlaps with Cn+12 (e.g., Nd@C82 with C94), but can be distinguished b the presence of two peaks with lower mass for Nd@C2n.Likewise, Nd2@Cn also overlaps with Cn+24, but Nd2@Cn has four peaks at lower mass which can be used to identify its presence.In mass-spectra of the CS2 extract, we can detect the presence of monometallofullerenes Nd@Cn but do not observe dimetallofullerenes Nd2@Cn.For Pr, both Pr@Cn and Pr2@Cn species can be detected.

Figure S4
. MALDI-TOF mass-spectra of Nd-EMF extracts in CS2 and DMF in the range of Nd@C82 and Nd2@C80 species: (a) negative ion mode, (b) positive ion mode.Insets show isotopic distribution and simulated mass-spectra.Whereas Nd2@C2n species are absent in mass-spectra of CS2 extract, they can be readily detected in DMF extract.Note that mass-spectra of Nd2@Cn species overlap with those of Nd@Cn+12, but di-EMFs have two additional peaks at lower masses, which allows to identify their presence in mixtures with mono-EMFs.

Synthesis and isolation of Nd2@C80(CF3) isomers
After arc-discharge evaporation, Nd-EMFs were extracted from the fullerene-containing soot by DMF under reflux.Umemoto reagent II was dissolved in DMF and added to the fullerene solution in DMF at room temperature.Optimal amount of URII for the synthesis of Nd2@C80(CF3) was determined by reacting aliquots of fullerene solution with different amount of URII (Fig. S5).After 5 minutes, DMF was removed under reduced pressure, the residue dissolved in toluene and processed by HPLC (Fig. S6).Note that reaction mixture also contained unidentified products of DMF decomposition, which elute at short retention time.

Figure S5
. Optimization of the URII amount for the synthesis of Nd2@Ih-C80(CF3).To identical aliquots of Nd-EMF extract dissolved in DMF (5 mL each), weighed amount of URII was added, after which DMF was removed by evaporation under reduced pressure, the residue dissolved in toluene, and toluene-soluble part analyzed with HPLC.The main component of the fraction eluting at 42-45 min is Nd2@Ih-C80(CF3).
Comparison of HPLC traces shows that addition of 0.5 mg of URII gives the best yield of the target compound.When 0.25 mg is added, the residual Nd@C82 can be well seen in the chromatogram, while amount of the target fraction is comparably small.Addition of 0.75 mg also gives smaller amount of the product because it reacts further and presumably forms multiaddducts.Therefore, the main trifluoromethylation reaction was performed using the reagent ratio corresponding to 0.5 mg URII per 5 mL of the Nd-EMF extract.

Regioisomers of Nd2@D5h-C80(CF3)
Figure S18.Isomers of Nd2@D5h-C80(CF3) obtained by addition of CF3 group to different carbons in the most stable conformer A of Nd2@D5h-C80.For the isomers with CF3 group added to the THJ carbon, the isomer number is underlined.Two more THJ isomers are not shown as we encountered sever SCF convergence problems for them and could not perform optimization.They are likely to be even less stable than other THJ isomer since convergence problems are caused by instabilities in their electronic structure.unique Nd pairs are designated as Nd1-Nd2, Nd3-Nd4, Nd5-Nd6, and 3 symmetry replicas are denoted by addition of "A", such as Nd1A-Nd2A.Ascribing metal dimers to particular cage orientations is not possible solely on the base of SC-XRD data.We therefore performed DFT calculations to aid the assignment.Optimized structures are shown with their relative energies.Nd1A-Nd2A and Nd5-Nd6 clearly correspond to the lowest-energy conformer of Nd2@D5h-C80(CF3).For Nd3-Nd4/Nd3A-Nd4A pairs, assignment is more ambiguous as neither gives the most stable conformer.Although Nd3A-Nd4A gives more stable structure after optimization, Nd3-Nd4 pair is more preferred as it is close to the metal positions of Nd1A-Nd2A and Nd5-Nd6 pairs.The disorder of Nd2 positions may be then caused by a largeamplitude motion of the Nd2 dimer near its minimum.

2283905 2283904 a
Figure S10.A, B, and C fullerene sites in 2Nd2@Ih-C80(CF3)•4NiOEP•1.63C7H8•0.37C6H6.The displacement parameters are shown at the 30% probability level.Upper row shows only major Nd2 sites, bottom row shows all Nd sites and a different orientation of the fullerene cage.Color code: grey for carbon, cyan for Nd, and yellow for F. The belt of hexagons around which metal dimer prefers to locate is highlighted yellow.The Nd disordered sites are labeled with site occupancies of 0.72, 0.20, and 0.07 for Nd1A/Nd2A, Nd3A/Nd4A, and Nd5A/Nd6A; 0.36, and 0.14 for Nd1B/Nd2B, and Nd3B/Nd4B; 0.36 and 0.14 for Nd1C/Nd2C, and Nd3C/Nd4C, respectively.

Figure S13 .
Figure S13.Relative position of endohedral Nd sites and NiOEP moieties in Nd2@D5h-C80(CF3)•2NiOEP•C6H6.C80(CF3) moieties and solvent molecules are omitted for clarity.The displacement parameters are shown at the 30% probability.Colour code: grey for carbon, blue for nitrogen, white for hydrogen, red for nickel, and cyan for Nd.

29 a
Figure S14.Conformers of Nd2@Ih-C80(CF3) obtained by varying a position of the Nd2 dimer inside C80(CF3) moiety.Two orientations of the fullerene cage are shown for each conformer along with its relative energy.

Figure S15 .
Figure S15.(a) Nd2@Ih-C80(CF3), site A from the SC-XRD structure; three Nd2 sites are colored by cyan (Nd1A, Nd2A, 0.72), red (Nd3A, Nd4A, 0.20) and purple (Nd5A, Nd6A, 0.07).(b) DFT-optimized molecule A with each of the Nd2 sites; also shown are symmetry replica of Nd atoms obtained by applying a mirror plane operation of the Cs-symmetric Ih-C80(CF3) moiety.(c) the same as (a), but also showing two nearest NiOEP and toluene molecules.For each Nd atom, the nearest Ni-N bond in the NiOEP molecule is highlighted with semitransparent triangles.(d) the same as (c) but showing only Nd, Ni, and N atoms.

Figure S16 .
Figure S16.(a) Nd2@Ih-C80(CF3), site C from the SC-XRD structure; one cage orientation is shown full-color, another one is semitransparent; two Nd2 sites and their symmetry replicas are colored by cyan (Nd1C, Nd2C, 0.36) and red (Nd3A, Nd4A, 0.14).The site Nd1C-Nd2C corresponds to the conformer Ih-1 (see Fig. S14), while the site Nd3C-Nd4C corresponds to the conformer Ih-2.(b) the same as (a), but also showing two nearest NiOEP; the second cage orientation is omitted.For each unique Nd atom, the nearest Ni-N bond in the NiOEP molecule is highlighted with semitransparent triangles.(c) the same as (b) but showing only Nd, Ni, and N atoms.

Figure S20 .
Figure S20.Electrostatic potential (ESP) maps inside the fullerene cage computed for [D5h-C80] 6− (upper row) and [D5h-C80(CF3)] 5− (middle row) compared to the molecular structure of the lowest-energy conformer of Nd2@D5h-C80(CF3) (bottom row).Each molecule is shown in three perpendicular orientations.While ESP in [D5h-C80] 6− is distributed uniformly, addition of CF3 group results in considerable inhomogeneity of ESP in [D5h-C80(CF3)] 5− .The minimum of ESP (colored red) is close to the position of one of the Nd atoms in Nd2@D5h-C80(CF3).Electrostatic stabilization of the metal position close to the ESP explains why this particular conformer is the most stable and hints to a possible hindered motion of metal atoms in Nd2@D5h-C80(CF3) in comparison to [Nd2@D5h-C80] − .However, this factor cannot explain the difference of metal dynamics between Ih and D5h cage isomers.

Figure S21 .
Figure S21.(a) SC-XRD structure of Nd2@D5h-C80(CF3) with two overlapping D5h-C80(CF3) moieties and 6 pairs of Nd sites (3 unique pairs and tree symmetry replica).(b) One of the cage orientations with a probable distribution of three Nd2 pairs in it.

Figure S22 .
Figure S22.DFT-optimization of Nd2@D5h-C80(CF3) structures, for which starting coordinates were based on one of the cage orientations from the SC-XRD structure combined with each of 6 pairs of Nd sites.3 unique Nd pairs are designated as Nd1-Nd2, Nd3-Nd4, Nd5-Nd6, and 3 symmetry replicas are denoted by addition of "A", such as Nd1A-Nd2A.Ascribing metal dimers to particular cage orientations is not possible solely on the base of SC-XRD data.We therefore performed DFT calculations to aid the assignment.Optimized structures are shown with their relative energies.Nd1A-Nd2A and Nd5-Nd6 clearly correspond to the lowest-energy conformer of Nd2@D5h-C80(CF3).For Nd3-Nd4/Nd3A-Nd4A pairs, assignment is more ambiguous as neither gives the most stable conformer.Although Nd3A-Nd4A gives more stable structure after optimization, Nd3-Nd4 pair is more preferred as it is close to the metal positions of Nd1A-Nd2A and Nd5-Nd6 pairs.The disorder of Nd2 positions may be then caused by a largeamplitude motion of the Nd2 dimer near its minimum.

Table S1b .
Relative All calculations are single-point energies for PBE-optimized structures
Analysis of experimental Nd-C distances in X-ray structure is complicated by the disorder, which makes the values not very reliable. a