Introduction

An important step in the reprocessing of spent nuclear fuel is the separation of the various components, which then are subjected to different further usages. Uranium and plutonium, used to fabricate new mixed oxide (MOX) fuels, [1] can be effectively separated by the PUREX process [2]. The radiotoxic minor actinides (Np, Am, Cm) possessing long decay should be transmuted to short-lived radionuclides by neutron bombardment [3]. The efficiency of this process, however, is hindered by the lanthanide components, which have large neutron absorption cross sections and therefore have to be removed selectively.

The trivalent lanthanide ions have very similar chemical properties to the trivalent americium and curium ions. Their separation is based on a liquid–liquid extraction process using complexing ligands with soft donor atoms, which are likely to bind stronger to the actinides than to the lanthanides owing to the somewhat stronger donor–acceptor abilities of the actinide 5f and 6d atomic orbitals with the molecular orbitals of the soft donor ligands than the 4f orbitals of the lanthanides. The key element in these ligands is the soft donor nitrogen atom in an aromatic environment. From the large number of heterocyclic N-donor extractants tested [4], 2,6-bis([1,2,4]-triazin-3-yl)pyridines (BTPs) proved to be a most effective and convenient option [5]. The main factor for the efficient binding of trivalent actinides likely lies in the smaller steric demands of the [1,2,4]-triazin-3-yl ring with respect to the 2-pyridinyl ring, allowing the nitrogens to approach more closely to the metal. Another advantage of BTPs as extraction ligands is that they work also in highly acidic aqueous solutions, generally used in the reprocessing of spent fuel.

The focus in this research field has been to elucidate the differences between actinides and lanthanides [5–11] upon binding to BTP ligands, while less attention was paid to the characteristics within the Ln series. It can be expected that if the ‘lanthanide contraction’ [12, 13] is significant in these complexes, the metal–ligand interactions should increase along the Ln series. Indeed, the Ln elements have been found to show a trend in selectivity as a function of their ionic radii [14]. Hence, the two goals of our present study were (1) to elucidate the contraction effects in [Ln(BTP)3] complexes and (2) to determine the character and trend of the interactions. This has been achieved by a joint application of experimental (FTIR, Raman and X-ray) and theoretical (DFT calculations) methods.

Experimental

General preparation procedure of the complexes [Ln(BTP)3](CF3SO3)3

Approximately 45 mg (~0.13 mmol) of [Ln(H2O)9](CF3SO3)3 was dissolved in 1 ml EtOH, and then three equivalents of 2,6-bis(5,6-dipropyl-1,2,4-triazin-3-yl)pyridine (~65 mg) in 1 ml EtOH (dissolved at elevated temperature) were added. On slow cooling to room temperature, the complexes form crystals of the constitution [Ln(BTP)3](CF3SO3)3*EtOH in yields of 50–60 %. By further evaporation of the solvent, more complex precipitated resulting in an overall yield above 95 %. The MeCN adduct of Pr was formed by re-crystallisation of the obtained precipitate in MeCN.

IR spectroscopy

IR spectroscopic measurements were taken on a Perkin Elmer Spectrum 1000 FT spectrometer on KBr pellets for the MIR range (4000–400 cm−1) and polyethylene pellets for the FIR range (400–120 cm−1) with a resolution of 2 cm−1.

Raman spectroscopy

The Raman spectra of [Ln(BTP)3](CF3SO3)3 (Ln = La, Sm, Eu and Gd) were measured at room temperature using a Bruker Senterra dispersive Raman microscope at 785 nm, 25 mW, at a spectral resolution of 4 cm−1. For each compound, eight spectra were averaged and background was corrected using the OPUS software from Bruker.

Crystallographic data

The XRD measurements were taken on a Siemens SMART CCD 1000 diffractometer [15] with monochromated MoKα radiation collecting a full sphere of data (La, Eu) or a Bruker Apex II Quazar diffractometer [16] collecting four full spheres of data (Pr, Nd). Frames were collected with an irradiation time of 20 or 10 s per frame and ω-scan technique with Δω = 0.45° on the SMART diffractometer, whereas on the Apex II diffractometer a mixed ω- and φ-scan technique was employed for data collection with Δω = Δφ = 0.5° with irradiation times of 12 s per frame appropriate to size and diffracting abilities of the crystals. Data were integrated with SAINT, corrected to Lorentz and polarisation effects, and an empirical adsorption correction with SADABS was applied. The structures were solved by direct methods and refined to an optimum R1 value with SHELX-2013 [17]. Visualisation for evaluation was performed with xpma [18], and the figures were created with Winray-32 [19]. For more detailed explanations, see Table 1.

Table 1 Crystallographic details of the measured [Ln(BTP)3](CF3SO3)3 ([Ln]) complexes
Full size table

The structures have been deposited at the Cambridge Crystallographic Data Centre with the reference CCDC numbers 1405122-1405125, and they contain the supplementary crystallographic data for this paper. These data can be obtained free of charge from the CCDC via www.ccdc.cam.ac.uk/data_request/cif.

Computational details

The large size of the BTP ligand necessitated to a truncation in the model structures. The n-propyl chains are expected to exert a very small influence on the electronic interactions between Ln and BTP, and therefore, in the calculation of the Ln series, they were replaced by hydrogens. The truncated ligand is denoted in the following as BTP′. The applicability of BTP′ was confirmed by a test calculations on the La complexes: the La–N distances comparing La(BTP)3 and La(BTP′)3 were within 0.01 Å. (The Cartesian coordinates of the computed [La(BTP′)3] and [La(BTP)3](CF3SO3)3 structures are given as Supplementary Material.)

In contrast, the aliphatic groups are needed for proper modelling of the experimental IR spectra. As the calculation of vibrational frequencies failed for the complex with the n-propyl groups, they were replaced by ethyl groups. This truncation caused changes by a few thousandths of angstrom in the La–N distances.

The computations were performed with the Gaussian 09 suit of programs [20] using the M062X [21] exchange–correlation functional in conjunction with the Stuttgart–Cologne 4f-in-core quasi-relativistic pseudopotentials for the lanthanides [22, 23]. The valence basis set treating the 5s5p5d6s orbitals had the contraction scheme of [7s6p5d]/[5s4p3d] [22]. For N, C and H, the standard 6-31G** basis set was used. Test calculations using the more expensive small-core 4f-in-valence quasi-relativistic pseudopotentials [24] were performed on the Ce and Tm complexes. The valence basis set treating the 4s4p4d4f5s5p5d6s orbitals had the contraction scheme of [14s13p10d8f6g]/[10s8p5d4f3g] [25]. These results agree with the main features of ‘lanthanide contraction’ predicted by the 4f-in-core calculations. The atomic charges and charge transfer properties were assessed using the natural bond orbital (NBO) model [26] by means of the NBO 5.9 code [27].

Results and discussion

The Ln(III) and An(III) ions form 1:3 complexes with charge–neutral BTP-type ligands, in which BTP acts as a tridentate ligand [8, 28] (Fig. 1). The metal is bonded to the pyridine (N1) and two azine nitrogens (N2, N2′). In the complex molecule, the three BTP ligands surround the metal ion in a spherical fashion with dihedral angles of around 120°/240° between the planes of the ligands. As BTP is neutral, the [Ln(BTP)3] moieties have a charge of +3 from the Ln(III) ion. Trifluoromethanesulfonate, \({\text{CF}}_{3} {\text{SO}}_{3}^{ - }\), was used as counter ion both in the experiments and modelling of the IR spectra. This structure from the computations is presented in Fig. 2.

Fig. 1
figure 1

Coordination of lanthanides and actinides to BTP

Full size image
Fig. 2
figure 2

Structure of the computed [La(BTP)3](CF3SO3)3 complex (hydrogens are not depicted)

Full size image

The main structural features of the complex molecules in the crystal agree with those of the isolated molecules from the computations. The coordination sphere around the lanthanide ion is ninefold (Fig. 3). Three N atoms of each BTP ligand are embedded in coordinative bonds. The nine donor N atoms form a distorted tri-capped prismatic arrangement around the metal ion with more or less equal Ln–N distances in the range of 2.45–2.65 Å (see Table 2) exhibiting no real differences between the Ln–N bond distances to a pyridine or a triazine unit. With these findings, the described structural features of the [Ln(BTP)3]3+ complex cations are in good agreement with literature reports on BTP [28], BTP derivative complexes [7, 29, 30] or complexes with comparable structural motifs arising from ligands which are related to BTP in terms of symmetry and coordination behaviour [5, 31–42].

Fig. 3
figure 3

Ninefold coordination sphere around the lanthanide ion. The numbering of atoms corresponds to those in the cif (X-ray structure) files

Full size image
Table 2 Selected bond distances (Å) from X-ray diffraction on [Ln(BTP)3](CF3SO3)3 and computations on [Ln(BTP′)3] complexes
Full size table

The experimentally determined structures of the crystalline [La(BTP)3](CF3SO3)3*EtOH, [Nd(BTP)3](CF3SO3)3*EtOH, [Eu(BTP)3](CF3SO3)3*EtOH compounds are virtually identical due to isostructural crystal formation (cf. Table 1). The Pr complex, however, does not crystallize well from the solvent of choice (EtOH). Using MeCN as solvent, crystals with the composition of [Pr(BTP)3](CF3SO3)3*2.4 MeCN were obtained. Differences in crystal packing between the two families of compounds with ethanol or acetonitrile adducts influenced very slightly the Ln–O distances (vide infra).

Selected geometrical parameters from our (La, Pr, Nd, Eu) and literature (Sm, Tm, Yb) [28] X-ray studies and from our computations are compared in Table 2. The experimental and calculated bond distances are in good agreement (taking into account the approximations used in our modelling, the intermolecular effects in the solid state, and the different physical meaning of the distance parameter from the two methods [43]). Most interesting are the Ln–N1 and Ln–N2 distances representing the bonding between the lanthanide and BTP(BTP′) ligands. These distances are between 2.4 and 2.7 Å, and around the sums of the ionic and covalent radii of Ln and N (e.g. the sum of covalent radii of La and N is 2.78 Å [44]; the sum of ionic radii is 2.62 Å [45]). This is in agreement with the mixed ionic/covalent character of the bonding in such complexes. The Ln–N3 distance is around 3.5 Å indicating very weak interaction in the van der Waals range. We note that while the computations resulted in a symmetric Ln–BTP′ arrangement, in the crystal the packing introduces minor deviations: within an Ln(BTP)3 moiety, both the Ln–N1 and Ln–N2 distances can differ up to 0.03 Å. For an easier comparison, in Table 2 we presented the average values.

The main question in the discussion of the geometry is the ‘lanthanide contraction’ [12, 13]. This can be well recognised in both the experimental and computed Ln–N distances. The decreasing trend (ca. 0.2 Å from La to Lu) is nearly parallel to the Ln3+ ionic radii (cf. Figure 4).

Fig. 4
figure 4

Computed Ln–N bond distances and the Ln3+ ionic radii. N1 and N2 correspond to the pyridine and azine nitrogens, respectively (cf. Figure 1)

Full size image

The bonds within the BTP′ ligand are considerably less affected. As the computed data in Table 2 show, the N1–C bond shows a slight sensitivity on the atomic number of the lanthanide (0.003 Å from La to Lu), while the effect in N2–N3 and N2–C is negligible. As these changes are within experimental error, they cannot be observed in the X-ray data. Comparing some characteristic geometrical parameters of BTP′ in the complexes with those of the free ligand (cf. Table 2), we can observe a lengthening of N1–C, while shortenings of N2–N3 and N2–C in the complex. These changes are associated with the change in the electron density distribution in the ligand upon complex formation.

Very interesting is the relative magnitude of the La–N1 and La–N2 distance demonstrated by the data in Table 2 and Fig. 4. At the beginning of the lanthanide row, La–N1 > La–N2 (Δ = 0.025 Å), which turns over from the middle of the row to Lu–N1 < Lu–N2 (Δr = 0.015 Å). The trend can also be recognised in the available experimental data. We note that the slightly deviating experimental Pr–N2 distance can be explained by the somewhat different cell structure of this compound compared with the other measured complexes (vide supra). The trend points to a significant change in the Ln…BTP′ interactions as the atomic number of Ln increases.

The origin of the presented feature lies in the ‘lanthanide contraction’. As shown in Fig. 4, the Ln–N1 distance changes parallel with the Ln3+ ionic radii. It indicates that the approach of the pyridine nitrogen by the Ln ion is controlled by the Ln3+ radius. At the same time, the less flexible azine nitrogens cannot accommodate in the same degree to the decreasing Ln–N1 distance; hence, the decrease in the Ln–N2 distance is somewhat less pronounced. In this way, the different behaviour of the two Ln–N distances can qualitatively be explained by steric reasons.

In order to uncover the consequences of this effect on the bonding interactions in the complexes, we investigated the atomic charges and charge transfer interactions by means of the NBO method [26].

The computed bonding data of the La and Lu complexes are compared in Table 3. The natural charges are around 2 e, indicating a strong ionic character of the Ln–BTP′ interaction. From the two binding nitrogens, N1 has a larger negative character in the complexes and hence a stronger ionic interaction with Ln.

Table 3 Comparison of charge transfer propertiesa in [La(BTP′)3] and [Lu(BTP′)3] from NBO analysis
Full size table

The main covalent contributions consist of donations from the lone pairs (LP) of N1 and N2 to the empty valence orbitals of Ln and from the backdonation from Ln to the π* antibonding orbitals of the N1–C, N2–N3 and N2–C bonds (cf. Table 3). In agreement with the stronger electrostatic interaction of Ln with N1, the donation from the LP of this nitrogen is smaller than the donation from the LP of N2. In contrast, the back donation to the pyridine and triazine rings is similar in terms of energy.

Along the lanthanide row, the following trends can be recognised in Table 3: (1) The atomic charge of the lanthanide decreases from La to Lu, while no change can be observed in the natural charges of N1 and N2 (the compensating charge increase is dissipated in the inner part of the ligand). These observations point to somewhat weakening ionic interactions from La to Lu. (2) All charge transfer energies are significantly increased in the Lu complex as compared to [La(BTP′)3] meaning the increase in covalent interactions. (3) The charge transfer interactions between Ln and N1 are considerably more evident in [Lu(BTP′)3] than the charge transfer interactions between Ln and N2. This is in agreement with the more pronounced shortening of the Ln–N1 bond distances from La to Lu than that of the Ln–N2 bond lengths, resulting in a turnover in the relative magnitudes (vide supra).

The experimental FTIR spectra of the [Ln(BTP)3](CF3SO3)3 series show slight differences with the change of the lanthanide from La to Lu. (The list of the band positions and intensities is available as Table S1 in the Supplementary Material.) In order to interpret these changes properly, we had to assign the absorption bands in the spectra first, which was done with the help of the calculations. The calculated IR spectrum of the La complex is compared with the experimental one in Fig. 5. In the figure, the calculated frequencies above 1100 cm−1 are scaled by a factor of 0.940, suggested for the M06-2X/6-31+G** level (differing only by a set of diffuse functions from our basis set) [46].

Fig. 5
figure 5

Comparison of the experimental and computed (scaled) IR spectra of [La(BTP)3](CF3SO3)3

Full size image

The two spectra in Fig. 5 are in good agreement, confirming that our calculations can provide a proper interpretation of the experimental IR spectrum. This has been performed by qualitative visual observation of the fundamentals by means of the GaussView program [47], facilitating the recognition of their main vibrational components. The bands of the \({\text{CF}}_{3} {\text{SO}}_{3}^{ - }\) counter ions were supported by literature IR data on rare earth trifluoromethanesulfonate crystals [48]. The characterisation of selected important peaks of [La(BTP)3](CF3SO3)3 is given in Table 4.

Table 4 Assignment of the characteristic bands in the IR spectra of [La(BTP)3](CF3SO3)3 on the basis of M06-2X/6-31G** calculations
Full size table

In Fig. 6a, we show the range of 400–600 cm−1 of the experimental IR spectra, which demonstrates nicely the modifications induced by exchanging of the lanthanide. The bands at 423 and 463 cm−1 in the spectrum of the La complex show a significant gradual shift up to 433 and 470 cm−1, respectively, in the spectrum of the Lu complex. A qualitative agreement with the trend of Ln3+ ionic radii and calculated Ln–N bond distances, shown in Fig. 4, can be well recognised. Hence, the experimental IR data confirm the appearance of ‘lanthanide contraction’ in [Ln(BTP)3](CF3SO3)3 complexes.

Fig. 6
figure 6

a The 400–600 cm−1 section of the FTIR spectra of [Ln(BTP)3](CF3SO3)3 complexes demonstrating the effect of ‘lanthanide contraction’ on the vibrational frequencies. b Selected experimental fundamental frequencies

Full size image

The fundamentals affected by the ‘lanthanide contraction’ are complex, but all contain some movement of the Ln atom. In the high-wavenumber fundamentals, these correspond to Ln–N stretching vibrations, while in the low-wavenumber region bending, mainly out-of-plane bending (with respect to the plane of the aromatic ring) vibrations.

The experimental wavenumbers of the most characteristic blueshifting bands are compiled in Fig. 6b involving all the studied lanthanides. The magnitude of shifts from La to Lu amounts to 7–10 cm−1. There are two somewhat different behaviours in the depicted trends: for ν 1–ν 3 and ν 6 the changes in the band positions are monotonic from La to Lu, while for ν 4 and ν 5 the majority of the shift appears in the first lanthanides (La–Nd), and the change in the second part of the Ln row is considerably smaller. This can probably be attributed to a change in the composition of these latter normal modes, probably to a decreasing Ln–N contribution. However, the change in this contribution is so small that it cannot be recognised by a qualitative visualisation.

In contrast to IR, the room temperature Raman spectra of solid [Ln(BTP)3](CF3SO3)3 (Ln = La, Sm, Eu, Gd) complexes given in the Supplementary Material as Figure S1 and Table S2 show no obvious shifts of any characteristic band.

Conclusions

Lanthanide complexes with 2,6-bis(5,6-dipropyl-1,2,4-triazin-3-yl)pyridine show clearly the effects of ‘lanthanide contraction’. Evidences of the Ln–N bond contraction appear both in the X-ray data and in the shift of some bands in the FTIR spectra.

We performed DFT calculations on model structures in order to uncover the bonding basis of the contraction effects in these complexes and to assist the interpretation of the experimental IR spectra. The computed geometries revealed that upon contraction the covalent interactions of Ln with the pyridine ring strengthen somewhat more than with the triazine ring. This is due to the more pronounced increasing charge transfer interactions between Ln and N1 compared to those with N2. Hence, the specialities of the ‘lanthanide contraction’ are related to these different covalent interactions with the pyridine and triazine rings.

Analysis of the fundamental vibrations revealed that the blueshifting bands are attributed to modes including significant metal–ligand vibrations, particularly in which the Ln-N distance changes or Ln moves out of the ligand plane.

Supplementary Material

The online version of this article contains supplementary material (list of experimental IR and Raman data, Cartesian coordinates of the computed [La(BTP′)3] and [La(BTP)3](CF3SO3)3 structures are given), which is available to authorised users.