Ultra-selective ligand-driven separation of strategic actinides

Metal ion separations are critical to numerous fields, including nuclear medicine, waste recycling, space exploration, and fundamental research. Nonetheless, operational conditions and performance are limited, imposing compromises between recovery, purity, and cost. Siderophore-inspired ligands show unprecedented charge-based selectivity and compatibility with harsh industry conditions, affording excellent separation efficiency, robustness and process control. Here, we successfully demonstrate a general separation strategy on three distinct systems, for Ac, Pu, and Bk purification. Separation factors (SF) obtained with model compound 3,4,3-LI(1,2-HOPO) are orders of magnitude higher than with any other ligand currently employed: 106 between Ac and relevant metal impurities, and over 108 for redox-free Pu purification against uranyl ions and trivalent actinides or fission products. Finally, a one-step separation method (SF > 3 × 106 and radiopurity > 99.999%) enables the isolation of Bk from adjacent actinides and fission products. The proposed approach offers a paradigm change for the production of strategic elements.

I sotope production has been the cornerstone of many research fields and applications throughout the last century 1 and relies largely on separation science. Contemporary examples illustrating the primary importance of separations include radionuclide purification for use in radiopharmaceuticals 2 or in radioactive thermoelectric generators that are vital to space exploration 3 . The development of efficient separation methods is also critical for forensics analysis, recycling of ageing weapon materials, fabrication of nuclear fuels 4 , production of radiotracers for research 5 , as well as manufacturing of high-purity actinide targets for the discovery of new elements 6 . Regardless of the application, product purity must be as high as possible, which requires highly efficient and cost-effective separation methods. Radionuclide production through either target irradiation ( 225 Ac, 177 Lu, 90/86 Y, 89 Zr, 47/44 Sc, 238 Pu, 248 Cm, 249 Bk, 249/252 Cf, etc.) or milking from long-lived sources ( 227 Ac → 227 Th → 223 Ra, 241 Pu → 241 Am, 233 U → 229 Th → 225 Ra → 225 Ac, 232 Th → 212 Pb, etc.) involves the handling of mixtures of metal ions where major impurities are often neighboring elements in the periodic table. In most cases, the ratio between the valuable element and impurities is extremely high (a few μg diluted in multi-g targets), rendering purification very challenging, albeit critical for the availability of the coveted isotope.
Recent reports have highlighted the creative use of solid matrices, such as crystalline selenites 7 or borates 8 , to quantitatively separate f-elements. However, most at-scale chemical purifications rely on chromatographic separations or liquid-liquid extraction methods or both, depending on the process scale and the desired specifications. These bi-phasic techniques are based on intrinsic interactions between metal ions and organic molecules dissolved in an organic diluent (liquid-liquid extraction) or grafted onto a solid matrix (chromatography). Ideally, these organic molecules (hereafter referred to as extractants) are amenable to transfer ions of interest from one phase to another in a selective manner relative to impurities. Extractant performance is typically not predictable and not always transposable from one separation system to another. In fact, predicting the efficiency and metal selectivity of a given process formulation (aqueous phase, diluent, and extractant) is a scientific challenge with numerous correlated variables 9,10 , such as metal speciation in the aqueous phase, metal-extractant compound speciation in the organic phase, free extractant speciation, influence of the diluent, loading capacity of the organic phase, etc. Most extractants currently used in hydrometallurgy can potentially co-extract multiple elements depending on chemical conditions (acidity, extractant concentration, phase ratio etc.) [11][12][13] . Separation selectivity is only achieved by finely tuning those chemical conditions and operational conditions for these processes are generally highly constrained, with many required steps to reach desired purities.
To overcome these challenges, a class of hydroxypyridinone (HOPO) chelators was investigated for its high metal-binding selectivity and applicability to separation needs. These molecules exhibit a unique combination of properties long sought in separation science: (i) water-solubility, (ii) structures consisting of solely H, C, N, and O atoms, (iii) ability to control metal oxidation states without additional redox-active species, (iv) extremely high charge-based selectivity, and (v) high metalligand complex stability even in strong acid (up to 10 M H + ). Using the model octadentate HOPO chelator, 3,4,3-LI(1,2-HOPO) (hereafter referred to as 343HOPO), and taking advantage of these unprecedented characteristics, highly efficient and robust chemical separation processes were developed for three strategic examples: the purification of 225 Ac, Pu-isotopes, and 249 Bk.

Results and Discussion
Choice of aqueous chelator as hold-back reagent. Numerous drug development studies [14][15][16] have focused on synthetic siderophore-inspired compounds because of their ability to form stable, and sometimes luminescent, complexes with metal ions of interest for medical imaging, radionuclide decontamination, and cancer treatments. While this class of ligands, encompassing HOPO and catecholamide (CAM) derivatives, has been known for decades, it had never been studied in details for separation applications. In fact, a few exceptions aside, most of these chelators have only been studied with a single cation, such as Gd 3+ , Th 4+ , or Pu 4+ , impairing broader evaluation of their metal-metal selectivity. The chemistry of some HOPO ligands was recently extended across the periodic table, highlighting their outstanding selectivity, and large superiority over polyaminocarboxylate chelators (IDA, EDTA, DTPA, etc.), typically encountered in separations (Fig. 1). A comparison of metal complex formation constants for the octadentate 343HOPO shows striking differences between tetravalent species and corresponding divalent or trivalent ones. The observed selectivity seems mainly chargebased and only slightly dependent on the ionic radius of the cations (Supplementary Fig. 1), although some selectivity is noted across the trivalent lanthanide and actinide series. 343HOPO outperforms any known chelator in terms of charge-specific selectivity and, in particular, for the binding of tetravalent ions. As such, the Ce 4+ /Ce 3+ selectivity of 343HOPO is about 15 orders of magnitude higher than what could be expected from a carboxylate ligand. Extreme selectivity is also observed even for ions of similar size, such as in the Th 4+ /Am 3+ , Th 4+ /Gd 3+ , Ce 4 + /Lu 3+ , or Th 4+ /Cd 2+ pairs ( Fig. 1 and Supplementary Fig. 2). As shown in Fig. 1, a handful of other molecules from the same family (octadentate 3,4,3-LI-CAM, Bis-TREN-Me-3,2-HOPO, and TAM-macrocycle; tetradentate 5LI-Me-3,2-HOPO, 5LIO-Me-3,2-HOPO; and bidentate PR-Me-3,2-HOPO) 17-20 exhibit known solution thermodynamic properties that match those of 343HOPO but none of the chelators typically used for separation methods is nearly as selective. Many additional HOPO and CAM ligands have been designed 15 and previously reported octadentate 1,2-HOPO 21 , mixed 3,4,3-LI(1,2-HOPO/Me-3,2-HOPO) 22 , and DFO/1,2-HOPO 23 structures would certainly make excellent candidates for charge-based selectivity; however, their chelation properties have so far only been investigated with either Eu 3+ , Zr 4+ , or Pu 4+ . Due to its current kg-scale availability and established solution thermodynamics, 343HOPO was used here as a model case. It was not initially designed for separation applications and its performance should not be considered as the upper limit for a separation strategy that could be extended to an entire molecular family.
The selectivity of 343HOPO for tetravalent ions is so high that it is expected to form complexes even under very acidic conditions (experimentally observed 24 in 3 M HCl for Sn 4+ ), whereas it should release trivalent 25 and divalent 24,26 ions completely below pH~2 (Fig. 1). To our knowledge, no other reported class of ligands exhibits such behavior. The clear metal discrimination afforded by 343HOPO (bound M 4+ versus free M 3+/2+/1+ ) can therefore be leveraged as a chemical switch to isolate charged ions, a needed tool when separating Ac 3+ /Th 4+ or Pu 4+ /Am 3+ mixtures. If such an ultra-selective complexant is present in the aqueous phase, overall process selectivity is expected to be decoupled from extractant selectivity, since a system containing 343HOPO in the aqueous phase and a completely non-selective extractant in the organic phase will still result in highly efficient separation. Such ligand-driven performance enables more flexible solvent formulation and operational process conditions. It will also spare the cumbersome development of new extractants and can be applied to a variety of M x+ /M y+ pairs, as demonstrated below.
Actinium purification. While there is great interest in using 225 Ac for targeted alpha therapy 27,28 , the development of 225 Acbased pharmaceuticals is still hindered by low isotope availability relative to potential market needs 1 . Furthermore, Ac chemistry is largely unexplored since (i) Ac 3+ being the biggest trivalent ion of the periodic table 28,29 , there is no adequate surrogate to study its chemistry, and (ii) its highly-radioactive longest-lived isotope ( 227 Ac, t 1/2 = 21.8 year) is available in only μg amounts. Both 227 Ac decay and 225 Ac production yield mixtures of Ac 3+ , Th 4+ , Ra 2+ , and trivalent lanthanides. The ratio between unwanted elements and Ac is typically very high and purification options are limited 30 . 343HOPO exhibits very low affinity toward Pu4+  Ac 3+ compared with other trivalent ions and has extremely high affinity for tetravalent ions, providing a straightforward tool to selectively isolate Ac. Figure 2 shows the Ac 3+ and Pu 4+ extraction yields by HDEHP (extractant widely used in hydrometallurgy, also known as D2EHPA) 31,32 in the presence of 343HOPO at different pH values. Full Ac 3+ extraction in the organic layer was achieved, while scavenging Pu 4+ (a surrogate for Th 4+ ) in the aqueous layer. Similar experiments in the presence of reference aqueous chelator DTPA, also used in the socalled TALSPEAK process 31 , showed partial extraction of both isotopes and no practical separation or recovery. After only a single step and despite a large initial Pu/Ac ratio (~10,000 mol/ mol), SF Ac/Pu values reached at least 1,000,000 in the presence of 343HOPO, combined with recovery yields of up to 99.70% for Ac in the organic phase and higher than 99.97% for Pu in the aqueous phase. In comparison, DTPA SF Ac/Pu values were between 1 and 100, the typical selectivity range observed in hydrometallurgical processes, with less than 50% Ac recovery. Separation of Ac from trivalent impurities was also investigated. both HDEHP/343HOPO-HNO 3 and HDEHP/DTPA-HNO 3 systems. Significant discrimination was observed for Ac 3+ against Am 3+ and Gd 3+ , with SF Ac/Am and SF Ac/Gd reaching 130,000 and 1,300,000, respectively. In the DTPA case, SF Am/Ac and SF Gd/Ac values were below 10 and 5300, respectively, and Ac recovery was low. A process flowsheet for 225 Ac purification using the 343HOPO-HDEHP combination is proposed in Supplementary Fig. 3.
Plutonium purification. Since the 1940's, worldwide Pu inventory has evolved from almost 0 to~2,500,000 kg due to anthropogenic activities 33,34 , and is estimated to increase by 70,000 kg/year 35 based on civilian nuclear power generation forecasts. Pu materials must be properly safeguarded throughout their lifespan, which necessitates advanced nuclear forensic controls and reprocessing activities. In this context, isolation of Pu from minor actinide and fission product impurities is essential. The standard PUREX (Plutonium Uranium Redox EXtraction) liquid-liquid extraction process 36 allows recovering and separating Pu and U from minor actinides and fission products. PUREX operates in concentrated HNO 3 media and includes two critical steps: (i) UO 2 2+ and Pu 4+ co-extraction into the organic phase (30 % TBP in diluent) while leaving trivalent ions in the aqueous phase, and (ii) reductive back-extraction of Pu as Pu 3+ while leaving UO 2 2+ in the organic phase. The latter is indispensable for U/Pu separation because of TBP's lack of selectivity between tetravalent and actinyl species. While reduction of Pu 4+ to Pu 3+ is achieved by addition of strong reducing agents into the liquid-liquid extraction batteries 37 , Pu 3+ is inherently unstable, and this redox component creates constraints in the reactive TBP-nitric medium (reduction of HNO 3 to HNO 2 and formation of potentially explosive compounds such as HN 3 ).
A simple but effective modification of PUREX was investigated for non-reductive separation of Pu and U (Supplementary Fig. 4). Figure 4 shows that Pu 4+ and UO 2 2+ extraction behaviors under typical PUREX conditions are expectedly very similar, with quantitative extraction of both elements and SF U/Pu values below 3. After co-extraction of U and Pu by TBP, the unprecedented chelation properties of 343HOPO at high acidity and its chargebased selectivity can be leveraged to selectively strip Pu 4+ from the organic phase. Efficient U/Pu separation was observed in the presence of 343HOPO (Fig. 4) by selective chelation of Pu 4+ , without significant interactions with UO 2 2+ over a broad range of acidity (up to 8 M HNO 3 ) and SF U/Pu values as high as 5800. The use of 343HOPO-type chelators could afford straightforward and efficient separation methods for the purification of U and Pu, and so, without using any redox-active chemical or non-volatile contaminant. This method also leads to very flexible process control, as demonstrated by the broad acidity range under which 343HOPO selectively scavenges Pu 4+ .
Another route, using TODGA as extractant, was explored for the purification of Pu not only from divalent ions but also from trivalent metals. TODGA is from the diglycolamide family used for resins separations 38,39 , and is currently investigated for nextgeneration nuclear waste treatment processes, such as EURO-GANEX 11,40,41 . TODGA is effective at extracting trivalent lanthanides and actinides from concentrated nitric media but its selectivity toward other ions, such as Pu 4+ , is very low. Very high yields were observed for the extraction of Gd 3+ , Lu 3+ , UO 2 2 + , Pu 4+ , and Cf 3+ from HNO 3 (>99.5% if HNO 3 > 1 M), emphasizing the lack of potential for practical separation (Fig. 4).
In the presence of 343HOPO, Pu 4+ is selectively held-back in the aqueous phase whereas the extraction behaviors of trivalent actinides, lanthanides, and uranyl are not impacted, offering a direct avenue for Pu recovery. SF values between the trivalent ions and Pu 4+ are above 450,000,000 (LoD reached) whereas SF U/Pu values are around 50,000. SF U/Pu values are limited by the relatively low distribution factors of uranyl (D U < 30) when using TODGA, as observed here and elsewhere 11 . Combining the high affinity of TODGA for trivalent lanthanides and actinides with the high affinity of TBP for uranyl and the selectivity of 343HOPO for Pu, a redox-free process was devised for the flashrecovery and purification of Pu from materials containing Pu, uranyl, minor actinides, and fission products ( Supplementary  Fig. 5).
Berkelium purification. Bk and Cf are of particular interest due to their use as targets for the production of super-heavy elements, such as elements 117 (tennessine, named after a 249 Bk target manufactured at Oak Ridge National Laboratory -ORNL, Tennessee) 42 and 118 (oganesson, produced from 249 Cf) 43 . 252 Cf is also a strategic isotope for oil and gas exploration as well as quality control of nuclear reactors 1 . Bk and Cf are produced via prolonged neutron irradiation of Am/Cm targets, yielding mixtures of actinides from Am to Fm and some fission products 5 . These transplutonium elements have traditionally displayed very similar chemistries as they exhibit the +III oxidation state in solution and have almost identical ionic radii 44 . To separate Cf 3+ from Bk 3+ , Bk 3+ can eventually be oxidized to Bk 4+ under harsh conditions (heating combined with excess NaBrO 3 in 8 M HNO 3 ) but it is unstable, which adds another level of complications to the eventual separation scheme. Campaigns conducted at ORNL for the purification of mg amounts of 249 Bk take several months 5,45 and result in relatively limited purification factors. The Bk isolation process 5 comprises about 25 steps, with a purification factor for the entire procedure (product of SF values from all steps) of~10 7 .
343HOPO was recently shown 46 to oxidize Bk 3+ and stabilize Bk 4+ in aqueous solution without addition of any redox-active species; a direct consequence of the ligand's thermodynamic preference for tetravalent cations (vide supra). The separation strategy detailed above can therefore be applied to the isolation of Bk from all trivalent ions. Preliminary tests with extractant HDEHP at high acidity (0.1-6 M HNO 3 ) show a drastic influence of 343HOPO on the Bk extraction by HDEHP ( Supplementary  Fig. 6). Comparisons with Pu 4+ , Gd 3+ , and Cf 3+ confirm that Bk exists as Bk 3+ in nitric media without chelator but forms a Bk 4+ complex with 343HOPO even under very acidic conditions ( Supplementary Fig. 7). Thus, isolation of Bk from its trivalent neighbors and lanthanides was studied in the HDEHP-HNO 3 −343HOPO system (Fig. 5). After a single step at room temperature, high extraction yields were observed for all tested trivalent ions, whereas Bk was selectively sequestered in the aqueous phase. SF Bk/Lu values as high as 320,000 were obtained, and SF values between 3000 and 10,000 were reached between Bk and adjacent actinides Am 3+ , Cf 3+ , and Es 3+ . Similar tests in the presence of NTA, EDTA, CDTA, and DTPA showed no separation at all between Bk and Cf, (similar to what is observed without chelator) since this type of ligands is not strong enough to bind metal ions under acidic conditions and is not selective enough to oxidize Bk 3+ to Bk 4+ . Importantly (Fig. 5 and Supplementary Fig. 8), the behavior of Bk in the presence of 343HOPO is completely decoupled from classic extraction parameters (extractant concentration, phase ratio), resulting in stable separation performance and robustness over a wide range of conditions. Given the system SF values, two consecutive steps would yield purification factors similar or higher than the current state-of-the-art process 5 .
Bk purification using TODGA as extractant was also investigated. TODGA is more relevant than HDEHP for the production of heavy actinides since it is effective at high acidity and is therefore compatible with the post-irradiation metallic target dissolution step. The extraction behavior of Bk in a TODGA-based system has never been reported 11 , but high extraction yields for Bk 3+ could be expected based on the behavior of Cf 3+ . In nitric solutions without chelator, quantitative co-extraction of 249 Bk 3+ and 249 Cf 3+ by TODGA was observed, leading to no practical separation between the metals (Fig. 5). In stark contrast, addition of 343HOPO drastically changes the Bk extraction profile: less than 1% Bk is extracted throughout the acidity range whereas Cf extraction remains undisturbed, due to the non-interaction between trivalent ions and the aqueous chelator at high acidity. The two adjacent actinides can be separated over a very broad acidity range (0.5-10 M HNO 3 ). Even if an unfavorable initial ratio Cf/Bk was used in those experiments (Cf/Bk~12,000), 249 Bk samples completely exempt of 249 Cf were obtained, notably so after only a single extraction. Experiments at various concentrations of extractant further confirmed the reliability of the TODGA-HNO 3 −343HOPO separation formulation ( Supplementary Fig. 9) with SF Cf/Bk > 1,000,000, regardless of the TODGA and HNO 3 concentrations. Additional separation experiments with Gd 3+ and Lu 3+ further evidenced the reliability of this system for purifying Bk (Supplementary Fig. 10). Finally, the separation method was tested at the mCi level. Legacy samples of 249 Bk/ 249 Cf used in previous studies 44,47 were gathered, evaporated, incinerated and reconstituted in 3 M HNO 3 solutions containing multiple mCi of 249 Bk with a Cf/Bk ratio of~3.4 mol/mol. The solution was split in five, 343HOPO was added, and an extraction step was performed using TODGA. As detailed in Table 1  the conditions used, very efficient separation was observed with the recovery of essentially pure 249 Bk in the aqueous phase (radiopurity > 99.999%, chemical purity > 99.8%) and high-purity 249 Cf in the organic phase (chemical purity > 99.5%). These results further confirm the proposed strategy could be used with high-activity samples, and scaled up to either produce high-purity Bk isotopes or to remove Bk traces during Cm, Cf, Es, or Fm production. Subsequent recovery of all isotopes would be facilitated by incineration and reconstitution in the desired matrix.
Noteworthy, the results presented above, although already providing better Cf/Bk separation than with any published method, do not represent optimum system performance since 343HOPO was not initially developed for separation applications and molecules with even higher selectivity could be designed. In addition, the three isotope purification examples examined in this work demonstrate the versatility of the proposed separation strategy, which could undoubtedly be transposed to other critical challenges 1 , such as those encountered with 89 Zr 4+ /Y 3+ , 177 Lu 3 + /Hf 4+ , 134 La 3+ / 134 Ce 4+ , or 117m Sn 4+ / 116 Cd 2+ purifications. We anticipate that the use of HOPO ligand derivatives as aqueous chelating hold-back reagents could pave the way to more reliable, flexible and efficient methodologies for metal cation separations.

Data availability
All data generated or analysed during this study are included in this published article (and its supplementary information files).