Ligand Synthesis and Physicochemical Properties

We commenced our studies with the synthesis of the requisite ethyl 1,2,4-triazine-3-carboxylate building blocks. Ethyl thiooxamate 4 was first converted into the known amidrazone 5 using hydrazine hydrate following the literature procedure,^14, 17 and used immediately in the next step to avoid potential self-condensation of 5 during storage. Condensation reaction of 5 with 3,4-hexanedione 6 a in refluxing ethanol afforded 7 a in 77 % yield (Scheme 1). Cyclic diketones 6 b and 6 c were also utilized as 1,2,4-triazine ligands derived from these diketones are more resistant towards acid-catalyzed hydrolysis and radiolysis in a process than those derived from acyclic diketones bearing hydrogen atoms in the alpha-position.¹⁸ Diketones 6 b and 6 c proved somewhat less reactive than 6 a in the condensation reactions with 5. However, use of titanium(IV) isopropoxide as Lewis acid catalyst provided satisfactory results and 7 b and 7 c were obtained in reasonable yields of 61 % and 55 %, respectively (Scheme 1).

With the building blocks 7 a–7 c in hand, we next explored their reactions with different polyamine cores to synthesize multidentate triazine ligands,¹⁹ using 7 a and piperazine to optimize the amidation step. Our initial approach was to convert esters 7 into the corresponding free acids by hydrolysis and explore amide coupling reactions with the polyamines. However, hydrolysis of 7 a to the corresponding free acid resulted in partial decarboxylation, which precluded the use of amide coupling reagents (DCC, HATU, etc) to synthesize the ligands. We then explored direct amidation reactions of ester 7 a with piperazine instead. Reaction of 7 a with piperazine 8 in THF under microwave conditions generated the corresponding monoamide 10 a in only 10 % yield, while attempted amidation reactions of 7 a with 8 using lipase enzymes^14f, 15a resulted in no conversion to the product. However, we found that reaction of monoamide 10 a with 7 a in refluxing THF did generate the desired diamide 9 a, albeit in low yield. To circumvent the low overall yield of this two-step approach, we then returned to our original approach and converted 7 a into the lithium salt of the acid, which was stable and did not undergo decarboxylation. Amide coupling (HATU) of the lithium salt of 7 a with 8 generated the desired ligand 9 a in 24 % yield (Scheme 2). In contrast to the low-yielding reaction of 7 a with piperazine 8, the analogous reaction of 7 b with 8 in THF under microwave conditions gave monoamide 10 b in 83 % yield. Reaction of monoamide 10 b with 7 b in refluxing THF then gave ligand 9 b in 37 % yield (Scheme 2).

We next explored potentially tetradentate triazine ligands 12 containing two bidentate 1,2,4-triazine-3-carboxamide chelating groups derived from the pre-organized diamine core (±)-trans-1,2-diaminocyclohexane. Direct amidation reactions of esters 7 b and 7 c with (±)-trans-1,2-diaminocyclohexane 11 in refluxing 1,4-dioxane afforded the desired triazine ligands 12 b and 12 c in 54 % and 56 % yields, respectively (Scheme 3).

Tris(2-aminoethyl)amine (TREN) 13 is a widely used triamine core for the synthesis of hexadentate ligands containing bidentate catechol or hydroxypyridinone coordinating groups for use as sequestering agents for actinides^16a or as biostatic agents for chelation of Fe(III).²⁰ We therefore explored the synthesis of analogous potentially hexadentate 1,2,4-triazine ligands 14 containing three bidentate 1,2,4-triazine-3-carboxamide chelating groups from reactions of 13 with ethyl 1,2,4-triazine-3-carboxylate building blocks 7. Amidation reactions of TREN 13 with each of the building blocks 7 a–7 c in refluxing 1,4-dioxane proceeded smoothly to form the target ligands 14 a–14 c, albeit in low yields of ≤49 % (Scheme 4). A slightly lower yield of 41 % was obtained for ligand 14 a under microwave conditions (THF, 2 h), whilst the use of lipase enzymes gave <20 % conversion of 14 a, as judged by ¹H NMR spectroscopy.^14f, 15a

The triamine tris(aminomethyl)ethane 18 is a frequently used core for the synthesis of tripodal ligands for coordination and sensing of different metal ions.²¹ Accordingly, we decided to employ 18 as a core for the synthesis of an analogous hexadentate 1,2,4-triazine ligand 19 b containing three bidentate 1,2,4-triazine-3-carboxamide chelating groups. Triamine 18 was synthesized from commercially available triol 15 in three steps according to the literature procedures.²² Triol 15 was first converted into the trimesylate 16, which was then treated with sodium azide in DMF to generate triazide 17. Finally, Staudinger reduction²³ of 17 using triphenylphosphine in the presence of water afforded triamine 18 in 31 % overall yield from 15. Reaction of 18 with ethyl 1,2,4-triazine-3-carboxylate building block 7 b in refluxing 1,4-dioxane gave the desired ligand 19 b in 15 % yield (Scheme 5).

Finally, we sought to synthesize a hexadentate 1,2,4-triazine ligand with three bidentate 1,2,4-triazine-3-carboxamide chelating groups containing the 1,3,5-triethylbenzene core derived from triamine 22, which has been utilized as a scaffold for many different ligands and receptors, including tripodal diglycolamide ligands for selective extraction of Am(III).²⁴ Triamine 22 was synthesized from commercially available tribromide 20 in two steps and 52 % overall yield according to the literature procedures.²⁵ Reaction of 20 with sodium azide afforded the crude triazide 21 which was reduced to triamine 22 using a Staudinger reduction. Reaction of 22 with the 1,2,4-triazine building block 7 b in refluxing 1,4-dioxane gave the desired ligand 23 b in 82 % yield (Scheme 6).

The physicochemical properties of the novel ligands 9 a, 9 b, 12 b, 12 c, 14 a–14 c, 19 b and 23 b are presented in Table 1 and are compared to those of the benchmark 1,2,4-triazine ligands 1–3 (Figure 1). As shown, the calculated partition coefficients (cLogP values) of the ligands span a wide range of values regardless of which method is used to calculate them. Whilst ligands 9 a, 9 b, 12 b, 12 c, 14 a and 14 b are all predicted to be less hydrophobic than ligands 1–3, ligands 14 c, 19 b and 23 b are predicted to be more hydrophobic than 1–3. Ligand hydrophobicity has obvious implications for the solubility and extractability of metal complexes into the organic phase, and hence the distribution ratios of the metal ions that are obtained. However, as mentioned previously, making a ligand more hydrophobic (e. g.: by adding alkyl groups) is usually detrimental to the extraction kinetics. For the benchmark ligands 1–3, the fraction of saturated carbons (Fsp³ value) ranges from 0.47 for ligand 3 to 0.59 for ligand 1. For the novel ligands reported herein, all of them have higher Fsp³ values than ligands 1–3, ranging from 0.6 for ligands 9 a and 14 a to 0.73 for ligand 12 c. All the novel ligands except 9 a and 9 b contain hydrogen bond donor (HBD) amide NH groups which are absent in ligands 1–3, whilst the number of hydrogen bond acceptor (HBA) groups in the novel ligands is comparable to (ligands 9 a, 9 b, 12 b, 12 c), or greater than (ligands 14 a–14 c, 19 b, 23 b), those in ligands 1–3.

Solvent Extraction Experiments

The novel ligands were then screened as potential selective extractants for the minor actinides Am(III) and Cm(III), as well as Eu(III) using solvent extraction experiments. Nitric acid solutions spiked with ²⁴¹Am, ²⁴⁴Cm and ¹⁵²Eu radionuclides were mixed with organic solutions of the ligands in the appropriate diluent (1-octanol, cyclohexanone or nitrobenzene), and distribution ratios for each metal ion (D) were measured by α- or γ-spectroscopy. Initial screening experiments were performed on ligands 12 b, 12 c, 14 b and 14 c using two nitric acid concentrations and 1-octanol, cyclohexanone or nitrobenzene as diluent, and the results are shown in the Supporting Information (Table S1). With the exception of ligand 12 b in cyclohexanone, all of the measured D values for Am(III) and Eu(III) were below 1 indicating that none of the novel ligands showed the ability to extract either metal ion from nitric acid solution. For ligand 12 b in cyclohexanone, the D values for Am(III) and Eu(III) extraction from 1 M nitric acid were 1.03 and 1.07, respectively, indicating that 12 b can co-extract both elements from nitric acid, albeit with no selectivity between them (SF_Eu/Am=1.04).

More detailed extraction experiments were then performed on ligands 9 a, 14 a, 19 b and 23 b across a range of nitric acid concentrations (see Supporting Information, Tables S2–S9). All of the D values for Am(III), Cm(III) and Eu(III) were below 1 and close to or below the detection limit, indicating that none of the novel ligands were able to extract Am(III), Cm(III) or Eu(III) from nitric acid solutions into 1-octanol. The observed extraction results could be attributed to a number of reasons, as solvent extraction of metal ions into an organic phase is governed by several criteria, including the thermodynamic stability of the formed metal:ligand complex, the metal:ligand stoichiometry of the formed complex, its solubility/lipophilicity and the mechanism of extraction.^4a, 26

NMR Titrations

In order to shed more light on the solvent extraction results, we next performed ¹H NMR titrations of selected ligands with the diamagnetic lanthanides La(III) and Lu(III) as well as Y(III) in deuterated acetonitrile. In each case, 0.5 and 1 equivalents of metal nitrate solution were added to the free ligand to check for complex formation prior to performing the full titration. For ligands 9 b, 12 b and 12 c, addition of 0.5 or 1 equivalents of La(III) or Y(III) gave NMR spectra that showed no new resonances for complex species (see Supporting Information, Figures S1–S7). In the case of 12 b and 12 c, a broadening of the free ligand resonances was observed on addition of either La(III) or Y(III), accompanied by a slight downfield shift in the amide NH protons from 8.36 ppm to ca. 9 ppm. This suggests complex species were not formed with these ligands.

Similar results were observed for ligand 14 b which gave poorly resolved spectra in the titration with Y(III) (see Supporting Information, Figure S10). However, some minor new resonances did appear towards the end of the titration (singlets at 0.65, 1.17, 1.39 and 1.74 ppm). These could be due to a complex species but their relative amounts are very small, suggesting complex formation of 14 b with Y(III) is minimal at best. Complex formation of the ligands with Y(III) could also be hampered by the relative kinetic inertness of Y(III) toward ligand substitution,²⁷ which we observed previously in NMR studies of bis-1,2,4-triazine ligands.²⁸ Some evidence for complex formation of 14 b with La(III) was also observed. However, the new resonances were very broad and poorly resolved (see Supporting Information, Figure S9). New resonances appeared for the amide NH protons at 8.77 ppm and 9.44 ppm, and the region associated with the methylene protons adjacent to the nitrogen atoms between 2.5 ppm and 4 ppm was very complex. Although the results are inconclusive, this could be accounted for by the formation of a chiral complex in which the methylene protons of 14 b are diastereotopic.

For ligand 23 b, addition of 0.5 or 1 equivalents of La(III) to the free ligand gave NMR spectra that showed the disappearance of the free ligand resonances and the appearance of a new set of resonances associated with a complex species (see Supporting Information, Figures S12–S19). However, addition of either Lu(III) or Y(III) to solutions of 23 b gave NMR spectra that showed no new resonances for complex species (see Supporting Information, Figures S21–S23). A full NMR titration of 23 b with La(III) was therefore carried out. The free ligand resonances gradually disappeared and a new set of resonances appeared during the course of the NMR titration of 23 b with La(III), suggesting a single complex species was formed with an overall metal:ligand stoichiometry of 1 : 1. Part of the stack plot for the titration of 23 b with La(III) is presented in Figure 2, while the full stack plot and a species distribution curve are shown in the Supporting Information (Figures S24–S28).

During the NMR titration of 23 b with La(III), we observed the appearance of two double-doublets at 4.30 ppm and 4.85 ppm for the methylene protons adjacent to the amide NH protons, and a complex multiplet at 2.95 ppm for the methylene protons of the ethyl groups. This suggests these methylene groups are diastereotopic, indicating formation of a chiral complex between 23 b and La(III). This could be accounted for by the formation of a chiral complex with a screw axis in which all three 1,2,4-triazine-3-carboxamide groups of 23 b coordinate to the metal in a bidentate fashion. However, other resonances are also present in addition to those expected for such a complex; an apparent doublet at 3.64 ppm, a broad singlet at 4.73 ppm and a broad singlet at 8.10 ppm (for amide NH protons). We tentatively propose that this is due to formation of a chiral complex in which two of the 1,2,4-triazine-3-carboxamide groups of 23 b coordinate to the metal in a bidentate fashion, while the third either does not coordinate, or coordinates in a different mode than the other two (e. g.: in a monodentate fashion). The lack of complex formation observed with 23 b and the smaller Lu(III) ion suggests the size of the coordination cavity of 23 b is better able to accommodate larger lanthanides such as La(III) but not smaller ones such as Lu(III).

DFT Calculations

DFT calculations were then performed to gain an insight into the variation in complexation behaviour among the ligands. Specifically, calculation of the binding energies of ligands 9 b, 12 b, 14 b, and 23 b with La(III) in acetonitrile was carried out to provide a comparison with the NMR titration results, which showed variation in complexation behaviour between the novel ligands. The energetics of complexation of 1 were also determined as a benchmark, since BTP ligands are reported to bind La(III) in non-aqueous environments.²⁹

The most favourable complexes of 9 b, 12 b, 14 b, and 23 b were all calculated to be of the form La(κ⁴-L)(NO₃)₃ in contrast with 1, which was calculated to be La(κ³-L)(NO₃)₃(H₂O)₂. Images of the complexes and full geometries are given in the Supporting Information (Figures S33–S35). Total Gibbs free energies of binding, ΔG, are given in Table 2, showing a negative binding energy was calculated for 1, consistent with its ability to bind La(III) mentioned above. Ligands 14 b and 23 b show negative and low positive binding energies, respectively, consistent with the NMR titrations suggesting complexation with lanthanides does occur. Furthermore, the calculated complexes of 14 b and 23 b are of very low symmetry, which is consistent with the experimental indications of chirality. Interestingly, the most favourable complex of 14 b shows only one of the three 1,2,4-triazine-3-carboxamide units chelating to La(III) in a bidentate fashion, with the other two coordinating only via the O-atom of the carboxamide group (Figure 3). The most favourable complex of 23 b with La(III) shows that two of the three 1,2,4-triazine-3-carboxamide units are chelating to the metal, with the third remaining unbound on the opposite side of the central benzene core (Figure 3). This structure is broadly consistent with the observed NMR titration results for 23 b. Ligands 9 b and 12 b are calculated to have relatively large positive binding energies, again consistent with the NMR titrations, in which complexation was not observed.

Encouraged by the consistency of the calculations with the experimental NMR results, further analysis was carried out to gain insights into the contributing factors to the binding energies. The enthalpic and entropic contributions to the binding are also listed in Table 2 and show significant variation between the ligands. All ΔH values are positive, indicative of stronger interactions between the water ligands and La(III) than between the synthesised ligands and La(III), but all have negative −TΔS terms, consistent with the release of water molecules on complexation of the ligands. The values of ΔH and −TΔS for 1 are significantly smaller than all the other ligands, consistent with its lower denticity and the smaller number of water molecules displaced.

A comparison of the novel ligands shows that 9 b and 12 b have very similar calculated contributions to their binding energies. The main difference between these ligands and ligands 14 b and 23 b, which showed some evidence of metal binding experimentally, is the enthalpy of binding rather than the entropic contribution. A further breakdown was calculated by determining the reorganisation energies, E_reorg, of the ligands (the energy differences between the ligands in their free and bound conformations), and the interaction energies, E_int, of the ligands (the energy difference between each ligand complex and the isolated ligand and remaining atoms of the complex, all in their complexed geometries), which are listed in Table 2. These results show that the trends in enthalpy of binding are largely consistent with both the reorganisation energies and interaction energies. Ligands 14 b and 23 b are calculated to have lower E_reorg and E_int values than 9 b and 12 b, indicating that their lack of binding is likely to arise from a combination of these factors. These results provide an indication that such calculations may be able to assist in the design of further novel ligands in future.

General Procedures

All solvents and reagents were purchased from Sigma-Aldrich, Acros Organics, Fluorochem or Alfa-Aesar and used without further purification unless otherwise specified. Reactions were monitored by TLC using silica gel with UV₂₅₄ fluorescent indicator. NMR spectra were recorded on a JEOL ECS400FT Delta spectrometer (399.78 MHz for ¹H NMR, 100.53 MHz for ¹³C NMR). Chemical shifts are reported in parts per million (ppm) relative to tetramethylsilane as internal standard. Coupling constants (J) are measured in hertz. Multiplets are reported as follows: b=broad, s=singlet, d=doublet, dd=double doublet, dt=double triplet, t=triplet, q=quartet, qu=quintet, m=multiplet, app d=apparent doublet, app t=apparent triplet. High resolution mass spectra were obtained on a Finnigan MAT900XLT high-resolution double focussing MS spectrometer using nano-electrospray ionisation (NESI) at the EPSRC UK National Mass Spectrometry Service (University of Swansea). Column chromatography was conducted using 0.060–0.20 mm silica gel (70–230 mesh), and automated flash column chromatography was performed using a Biotage Isolera One ISO-1SV instrument. The calculated logP (cLogP) values of the ligands were calculated using ChemSketch software (available at https://www.acdlabs.com/) and SwissADME (available at http://www.swissadme.ch/). The fractions of saturated carbon atoms (Fsp³ values) of the ligands were calculated as the ratio of the number of saturated sp³ hybridized carbon atoms to the total number of carbon atoms for each ligand. Diketones 6 b⁸ and 6 c^6c were prepared as described previously.