Christian D.
Buch
a,
Steen H.
Hansen
a,
Dmitri
Mitcov
a,
Camilla M.
Tram
a,
Gary S.
Nichol
b,
Euan K.
Brechin
b and
Stergios
Piligkos
*a
aDepartment of Chemistry, University of Copenhagen, Universitetsparken 5, DK-2100, Copenhagen, Denmark. E-mail: piligkos@chem.ku.dk
bEaStCHEM School of Chemistry, University of Edinburgh, Edinburgh, UK
First published on 15th April 2021
Heterolanthanide complexes are difficult to synthesize owing to the similar chemistry of the lanthanide ions. Consequently, very few purely heterolanthanide complexes have been synthesized. This is despite the fact that such complexes hold interesting optical and magnetic properties. To fine-tune these properties, it is important that one can choose complexes with any given combination of lanthanides. Herein we report a synthetic procedure which yields pure heterodinuclear lanthanide cryptates LnLn*LX3 (X = NO3− or OTf−) based on the cryptand H3L = N[(CH2)2NCH–R–CH
N–(CH2)2]3N (R = m-C6H2OH-2-Me-5). In the synthesis the choice of counter ion and solvent proves crucial in controlling the Ln–Ln* composition. Choosing the optimal solvent and counter ion afford pure heterodinuclear complexes with any given combination of Gd(III)–Lu(III) including Y(III). To demonstrate the versatility of the synthesis all dinuclear combinations of Y(III), Gd(III), Yb(III) and Lu(III) were synthesized resulting in 10 novel complexes of the form LnLn*L(OTf)3 with LnLn* = YbGd 1, YbY 2, YbLu 3, YbYb 4, LuGd 5, LuY 6, LuLu 7, YGd 8, YY 9 and GdGd 10. Through the use of 1H, 13C NMR and mass spectrometry the heterodinuclear nature of YbGd, YbY, YbLu, LuGd, LuY and YGd was confirmed. Crystal structures of LnLn*L(NO3)3 reveal short Ln–Ln distances of ∼3.5 Å. Using SQUID magnetometry the exchange coupling between the lanthanide ions was found to be anti-ferromagnetic for GdGd and YbYb while ferromagnetic for YbGd.
With respect to QIP, Ln complexes form a very exciting but rather unexplored class of molecular materials.27,42–45 We have recently demonstrated that Yb(trensal),46 a member of the Ln(trensal) series,47–54 is a very promising candidate for the realisation of a molecule-based electronic quantum-bit (qubit).45 Yb(trensal) is one of the few known Ln-based molecular qubits to date. Furthermore, the potential multilevel nature of Ln molecular materials, resulting from the hyperfine interaction of electronic and nuclear angular momenta or from week ligand field splittings of the Stark sublevels in cases where the orbital angular momentum is quenched, can be exploited as an additional resource for the development of quantum logic algorithms.55–59 In recent ground-breaking work, Grover's quantum algorithm, relevant to identifying an element within an unsorted database, was implemented on the basis of the four nuclear spin wavefunctions of a bis(phthalocyanine) Tb(III) molecular magnet.60 Nuclear spins were also proposed as physical supports for the implementation of qubits,28,60 since the contraction of nuclear wave functions shields them from the environment, minimizing decoherence effects. This results in nuclear spins having much longer coherence times than their electronic counterparts. However, this isolation also leads to long manipulation times,61 which can be circumvented via the use of molecular systems in which hyperfine interactions are significant, resulting in faster manipulation time-scales.62 We very recently demonstrated that Yb(trensal)46 is a prototypical coupled electronic qubit–nuclear qudit63 (where a qudit is a quantum system comprising more than two levels). The nuclear qudit of the 173Yb isotope nucleus (nuclear spin I = 5/2) displayed coherence times of the order of 10 to 102 microseconds, similar to the best performing, state-of-the-art, nuclear-spin-based qudits.64 Most importantly, the electronuclear multilevel structure of the qudit allowed intrinsic implementation of quantum error corrections concerning encoding of both amplitude and phase shift error corrections.63 QIP algorithms can be performed as a sequence of single qubit rotations and gate operations performed on two entangled qubits. Thus, nuclear spins hosted in molecular magnetic materials are excellent candidates for the implementation of single qubit gates. However, their lack of interaction with the environment creates difficulties for the realization of coupled qubit gates. Hyperfine interactions mix the nuclear and electronic angular momentum wave-functions. Thus, construction of nuclear spin quantum gates appears feasible by simultaneously exploiting the long range coupling of electronic spins via magnetic exchange interactions and the hyperfine interaction-induced mixing of the electronic and nuclear angular momenta, at the single-ion level.65 In order to be individually addressable, the two angular momenta located on the two lanthanide centres composing the quantum gate have to be different. Thus, the need for the controlled synthesis of pure heterometallic complexes arises.
Synthesis of heterometallic Ln complexes is difficult since the 4f elements all have very similar chemistries. This is further compounded by the bonding interaction with ligands being largely ionic in character,35 resulting in many synthetic protocols producing impurities from homonuclear complexes (i.e. scrambling).66–69 Thus, examples of pure heterometallic Ln complexes remain rare. These are almost solely limited to phthalocyanine70–73 and DOTA complexes.74–76 An elegant strategy relying on size selection of dissymmetric ligands has been employed to synthesise heteronuclear Ln complexes without scrambling.31,33 However, this is only efficient when the included Ln(III) ions are of very different size, greatly limiting the choice of possible combinations.67
We present herein a synthetic strategy for the realization of pure heterodinuclear Ln complexes of composition not limited by size selectivity. Thus, for the second half of the lanthanide series (from Gd to Lu, and including Y), our synthetic strategy results in heterodinuclear complexes of arbitrary composition. To illustrate this point, we present the synthesis and characterisation of heterodinuclear complexes LnLn*L containing two large Ln centres (Y, Gd), two small Ln centres (Yb, Lu) and one small (Yb or Lu) and one large (Y or Gd) centre, corresponding to the following 10 complexes: YbGd 1, YbY 2, YbLu 3, YbYb 4, LuGd 5, LuY 6, LuLu 7, YGd 8, YY 9 and GdGd 10. We also demonstrate the solid-state and solution-state stability of 1–10, and present the static magnetic properties of the paramagnetic members of the family, namely, 1, 2, 3, 4, 5, 8 and 10.
As can be seen from Scheme 1, the heptacoordinated Ln ion in Ln(trensal) occupies all the available coordination positions within the complex. There are therefore no uncoordinated functional sites left to be used as expansion points of the structural motif to accommodate coordination of a second Ln* centre. Thus, use of a ligand bearing additional chemical functions is necessary. In the first instance we used tris(2-aminoethyl)amine (tren) and 2,6-diformyl-p-cresol (dfmp), a functionalised derivative of salicylaldehyde. Use of dfmp allows for the synthesis of complexes of type LnL1 (Scheme 1) that can be regarded as functionalized derivatives of Ln(trensal). Subsequently, the template effect of the Ln centre in LnL1 was exploited for the condensation of a second tren to create the cryptate complexes LnL, which possess a preformed vacant coordination site that ultimately acts as host for the second Ln centre. In the final step, the second Ln* centre is inserted in the cryptate resulting in the targeted heterodinuclear complexes [LnLn*L]X3, with X a monoanion. Cryptand complexes are known to be thermodynamically and kinetically stable,77 with Ln(III) cryptates being considered suitable candidates for MRI contrast agents.78 Such stability is a prerequisite for surface deposition protocols and, ultimately, inclusion in spintronic devices. To the best of our knowledge only homodinuclear Ln(III) cryptates exist.78–82
![]() | ||
Scheme 1 Strategy for the synthesis of pure heterodinuclear LnLn*L complexes based on the Ln(trensal) motif. |
We have previously presented the synthesis, magnetic and spectroscopic properties of LnL1 and LnL.83,84 Both these families of complexes have merits on their own and should not simply be regarded as mere intermediate compounds in the synthesis of our targeted heterodinuclear complexes. LnL1 contains pendant, reactive carbonyl functions that can be used as anchoring points for surface deposition protocols. These same functionalities can also be exploited in post-synthetic reaction schemes making them truly versatile modules for deposition on various surfaces.84 LnL have previously been considered as targets for the development of novel MRI contrast agents.83
![]() | ||
Fig. 1 MALDI positive mode mass spectrum of 1. Colour code: signal (black), predicted isotope patterns for [Gd2L(OTf)2]+ (red), [YbGdL(OTf)2]+ (green), and [Yb2L(OTf)2]+ (orange). |
![]() | ||
Fig. 2 1H NMR of 6 in CD3CN. The insert shows an enlargement of the four signals between 8.5 and 7.0 ppm. |
![]() | ||
Fig. 4 ΔΔ and ΛΛ enantiomers of the cations of 1N and 3N, respectively. The complexes are viewed along the pseudo-trigonal axis. |
PXRD measurements confirm that the isolated powders are isostructural, and have the same structure as the single crystal X-ray structure (Fig. S63 and S64†). PXRD also confirms that the complexes are obtained phase pure. Mass spectrometry reveals 1N–10N remain pure heterodinuclear complexes and that no exchange of Ln ions occurs (Fig. S11–S20†). This is also confirmed by NMR spectroscopy where 6N shows the same splitting in the 1H NMR as 6 (Fig. S60†). Additionally, the NMR spectra of 6N, 7N and 9N confirm that the complexes remain dinuclear after the recrystallization (Fig. S60–S62†).
To have a truly versatile heterodinuclear Ln complex, the molecule needs to be as robust as possible. This is especially true if the complex is to be included in spintronic devices where surface deposition techniques require stable components. The cryptate complexes presented herein are all stable in solution for several days. This was confirmed using NMR, where the 1H NMR of 6 dissolved in CD3CN did not change after seven days (Fig. 2 and S55†). Similar stability was found when dissolved in CD3OD (Fig. S57 and S58†). To further test the robustness of the complexes 100 equivalents of a third Ln(III) ion were added to a MeOH solution of the complexes. The complexes retain their original LnLn* composition. This was studied by dissolving 4 and 10 in methanol and subjecting them to 100 equivalents of either Y(OTf)3·xH2O or Lu(OTf)3·xH2O, respectively. After one week no signals from new species involving either Lu(III) or Y(III) could be observed by mass spectrometry (Fig. S24 and S25†), confirming the robustness of these complexes.
The d.c. susceptibility data of 1N–5N, 8N and 10N, are plotted in Fig. 5 as χT products, where χ = M/B with χ the molar magnetic susceptibility and M the magnetization. At 270 K, the χT products of all measured complexes approach their Curie constants (10.44, 2.57, 2.57, 5.14, 7.87, 7.87 and 15.74 cm3 K mol−1) with values of 9.91, 2.06, 1.99, 4.30, 7.58, 7.89 and 15.74 cm3 K mol−1, for 1N, 2N, 3N, 4N, 5N, 8N and 10N, respectively. Upon decreasing the temperature, the χT product decreases for 2N–5N, 8N and 10N reaching its lowest value at 2 K measuring 1.25, 1.22, 2.39, 7.72, 7.93 and 5.06 cm3 K mol−1, respectively. This decrease is due to thermal depopulation of excited mJ states in complexes containing one paramagnetic centre (2N, 3N, 5N and 8N). For 4N and 10N that both contain two paramagnetic centres, the decrease can be attributed to a combination of effects arising from thermal depopulation of excited mJ states and anti-ferromagnetic interactions between the coupled Ln centres. For 4N, thermal depopulation of excited mJ states will prevail at higher temperatures, and the weak exchange interactions will occur at lower temperatures. For 10N, given the magnetically isotropic nature of Gd(III), this picture is inversed. The χT product of 1N also decreases upon lowering the temperature down to 12 K. Below this temperature the χT product of 1N rapidly increases, reaching a maximum of 9.44 cm3 K mol−1 at 2 K. This behavior indicates ferromagnetic coupling between the Yb(III) and Gd(III) centres.
The quantitative analysis of the static magnetic properties of the studied complexes was performed by simultaneous numerical fitting, by use of the simplex algorithm,86 of both the χT and VTVB data to Hamiltonian (1)
![]() | (1) |
We performed the quantitative analysis of the static magnetic properties in two steps. Initially the CF parameters of Gd(III) or Yb(III) sites in complexes containing only a paramagnetic centre (2N, 3N, 5N and 8N) were determined. The number of CF parameters required depends on the symmetry of the coordination sphere of the Ln ion. For C1 symmetry, as is the case for 2N, 3N, 5N and 8N, twenty seven CF parameters are needed. Determining this number of parameters from the magnetic susceptibility and VTVB measurements alone would result in overparameterization. To remediate this problem, we assume that the local symmetry of the Ln sites is C3, neglecting the effect of the coordinated nitrate anions. By assuming C3 symmetry, the maximum number of CF parameters reduces to nine. However, nine CF parameters is still a large number of parameters to be determined solely from thermodynamic magnetization data. Thus, additional information can be incorporated in the fitting model by exploiting emission and Electron Paramagnetic Resonance (EPR) spectroscopies, as has been previously shown for the Ln(trensal) series.45–54 Once the CF parameters of Gd(III) and Yb(III) sites determined, quantitative analysis of the magnetic properties of complexes containing two paramagnetic centres was performed by fixing the CF parameters of Gd(III) and/or Yb(III) and fitting only the parameters relevant to the exchange term of Hamiltonian (1).
In the case of complexes containing a paramagnetic Gd(III) ion and a diamagnetic Lu(III) or Y(III) ion (5N and 8N, respectively), the 4f7 electronic configuration of Gd(III) results in an 8S7/2 ground term where the orbital angular momentum is quenched. This means that there are no first order contributions to anisotropy and that the associated CF splitting is relatively small with respect to the experimental conditions of magnetization measurements. Hence, very little information on the CF can be extracted solely from magnetization measurements. However, the small CF in Gd(III) means that all transitions within the ground multiplet can be observed using EPR spectroscopy and thus information on the CF parameters can be extracted based on these observations. Using the EPR simulation software Sim,89 the EPR spectra of 5N and 8N were simulated using Hamiltonian (1) resulting in the best-fit CF parameters (in Stevens formalism90): B20/hc = −1.84 × 10−2 cm−1, B40/hc = 1.95 × 10−4 cm−1 and B60/hc = −6.67 × 10−6 cm−1 with g = 2 and h and c the Planck and speed of light constants (Fig. S84 and S85†). Because of the very broad polycrystalline powder EPR spectra no off-diagonal CF parameters could be determined. The obtained CF parameters were then used to simulate the χT product and the VTVB data (Fig. 5, S78 and S79†). The associated eigen-spectrum is shown in Fig. S91† where it can be seen that the total CF splitting of 5N and 8N is of the order of 1 cm−1. We note that both complexes display very similar EPR spectra, albeit very broad (Fig. S84 and S85†), suggesting that the CF of Gd(III) is not dramatically influenced by the size of the second Ln ion.
For the complexes containing a paramagnetic Yb(III) ion and a diamagnetic Y(III) or Lu(III) ion (2N and 3N, respectively), the 4f13 electronic configuration of Yb(III) results in an 2F7/2 ground term where the orbital angular momentum of Yb(III) is unquenched. Thus, anisotropy to first order is expected for the CF splitting pattern of Yb(III). This means that in the case of Yb(III), emission spectroscopy can be exploited, in addition to EPR, as an additional source of information on the CF. The solid state, room temperature emission spectra of 2N and 3N, diluted at 3% in the diamagnetic host 9N, reveal that the total CF splitting of the ground term of Yb(III) in these complexes is of the order of 560 cm−1 (Fig. S88†). This is approximately 2/3 of the value observed in the parent Ln(trensal) complex,46 evidencing the influence of the electron density of one Ln centre as an additional contribution to the ligand field of the other one. However, the polycrystalline powder EPR spectra of these dilute samples of 2N and 3N, are too broad (Fig. S86 and S87†) to offer reliable information to be used in the fitting procedure. Furthermore, the χT products and VTVB measurements of 2N and 3N are very similar (Fig. 5, S70 and S72†). This suggests that the CF of Yb(III) in these two complexes is very similar, and similar to that observed for Gd(III) in 5N and 8N. Therefore, CF parameters were extracted only from measurements on 2N. Thus, the CF parameters of 2N were obtained by simultaneously fitting to Hamiltonian (1) its χT product, VTVB data and total multiplet splitting, determined by emission spectroscopy. As initial parameters for the modelling, the CF parameters of YbL were used.83 Using this model and allowing all nine CF parameters to vary, good agreement between the predicted and measured VTVB and susceptibility curves was obtained (Fig. 5 and S70†). The nine best-fit CF parameters are given in Tables S1 and S2† in Stevens and Wybourne notations, respectively. The associated eigenspectrum is shown in Fig. S91† where it can be seen that the total CF splitting of 2N and 3N is of the order of 560 cm−1. This set of parameters also reproduces the susceptibility and VTVB measurements of 3N, demonstrating that the CF of Yb(III) in these complexes does not dramatically depend on the size of the second Ln ion (Fig. S71 and S72†).
For the complexes containing two paramagnetic Ln ions (10N, 1N and 4N) inclusion of exchange interaction terms in Hamiltonian (1) is necessary since statistically significant deviations are observed between the experimental χT product of these as compared to the sum of their uncorrelated constitutive centres (Fig. S81–S83†).
In the case of 10N, the magnetic exchange interaction between the two Gd(III) centers can be treated by including in Hamiltonian (1) only isotropic exchange terms (JGd–Gd). Thus, by fixing the CF parameters of Gd(III) to the ones determined for 5N and 8N, JGd–Gd/hc was determined to be −0.1370 cm−1 ± 0.0005 cm−1 (in the −2J convention), by a simultaneous fit of the χT product and the VTVB data to Hamiltonian (1). The model reproduced the measurements well, with an anti-ferromagnetic interaction between the two Gd(III) ions (Fig. 5 and S80†). The associated eigenspectrum of 10N is shown in Fig. S91† where it can be seen that the antiferromagnetic interactions in 10N result in a group of sixty four closely packed states, spread over a total range of ∼9 cm−1. The antiferromagnetic ground state is a singlet separated from the first excited state by just 0.05 cm−1. The determined best-fit value is lower than that previously reported (Jiso/hc = −0.194 cm−1) for this complex using a spin-only expression,80 where only magnetic susceptibility data (not VTVB data) were analyzed.
For 1N and 4N containing the orbitally degenerate Yb(III) ion, anisotropic exchange, in addition to isotropic exchange, should be included. However, such a model (Table S3†) produces the same results, on a qualitative level, irrespective of whether the model contains only isotropic exchange (Fig. 5), only anisotropic exchange, or both anisotropic and isotropic exchange (Fig. S65–S68 and S73–S76†). This is most likely due to the fact that the polycrystalline powder data used in the modelling are insufficient to probe the orientation sensitivity of the anisotropic exchange terms, as previously discussed.88 Small crystal size precludes orientation dependent single crystal magnetization and EPR measurements. Performing the modelling without including exchange terms results in poorer agreement with experiment, clearly demonstrating that magnetic exchange interactions are operating in 1N and 4N (Fig. S65–S69† for 1N and Fig. S73–S77† for 4N). This is also observed in the temperature dependence of the χT product of 1N, which increases upon cooling from 12 to 2 K. Using only an isotropic exchange term in 1N and 4N yielded a ferromagnetic interaction in 1N (Jiso/hc = 0.01780 cm−1 ± 0.00005 cm−1) and an anti-ferromagnetic interaction in 4N (Jiso/hc = −0.0072 cm−1 ± 0.0007 cm−1). The associated eigenspectra of 1N and 4N are shown in Fig. S91.† In the case of 1N the sixty four levels are grouped into four sets of sixteen, corresponding to the interaction of each of the four Kramers doublets of Yb(III) with the four Kramers doublets of Gd(III), spread over a range of ∼700 cm−1. In the case of 4N the sixty four levels are grouped in ten sets spread over a range of ∼1100 cm−1.
The Ln–Ln distance, bridged by three phenoxide O-atoms, was found to be ∼3.5 Å with GdGd having the largest distance at 3.5035(7) Å and LuLu possessing the smallest distance at 3.4396(7) Å. This results in relatively strong magnetic exchange coupling between the metals with respect to QIP protocols. For both 4N and 10N the lanthanide ions are coupled antiferromagnetically. On the contrary, 1N shows ferromagnetic coupling between the two lanthanide ions. In future studies on single crystals, the exchange coupling between the isotropic Gd(III) ion and oblate or prolate Ln(III) ions, as well the coupling between two anisotropic ions, will be investigated in detail.
Footnote |
† Electronic supplementary information (ESI) available: Experimental details and characterisation and magnetisation data of 1–10. Crystallographic data (including structure factors) for 1N–10N. CCDC 2059654 (10N), 2059655 (5N), 2059656 (7N), 2059657 (3N), 2059658 (6N), 2059659 (1N), 2059660 (4N), 2059661 (2N), 2059662 (8N), 2059663 (9N). For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/d1sc00987g |
This journal is © The Royal Society of Chemistry 2021 |