Isolation of a californium(II) crown–ether complex

The actinides, from californium to nobelium (Z = 98–102), are known to have an accessible +2 oxidation state. Understanding the origin of this chemical behaviour requires characterizing CfII materials, but investigations are hampered by the fact that they have remained difficult to isolate. This partly arises from the intrinsic challenges of manipulating this unstable element, as well as a lack of suitable reductants that do not reduce CfIII to Cf°. Here we show that a CfII crown–ether complex, Cf(18-crown-6)I2, can be prepared using an Al/Hg amalgam as a reductant. Spectroscopic evidence shows that CfIII can be quantitatively reduced to CfII, and rapid radiolytic re-oxidation in solution yields co-crystallized mixtures of CfII and CfIII complexes without the Al/Hg amalgam. Quantum-chemical calculations show that the Cf‒ligand interactions are highly ionic and that 5f/6d mixing is absent, resulting in weak 5f→5f transitions and an absorption spectrum dominated by 5f→6d transitions. Californium is difficult to prepare in its divalent state. Now, crystals of a Cf(II) crown–ether complex have been synthesized by reduction of a Cf(III) precursor with an Al/Hg amalgam. They exhibit 5f→6d transitions in the visible region and near-infrared emission that are highly sensitive to changes in the coordination environment.

The actinides, from californium to nobelium (Z = 98-102), are known to have an accessible +2 oxidation state. Understanding the origin of this chemical behaviour requires characterizing Cf II materials, but investigations are hampered by the fact that they have remained difficult to isolate. This partly arises from the intrinsic challenges of manipulating this unstable element, as well as a lack of suitable reductants that do not reduce Cf III to Cf°. Here we show that a Cf II crown-ether complex, Cf(18-crown-6)I 2 , can be prepared using an Al/Hg amalgam as a reductant. Spectroscopic evidence shows that Cf III can be quantitatively reduced to Cf II , and rapid radiolytic re-oxidation in solution yields co-crystallized mixtures of Cf II and Cf III complexes without the Al/Hg amalgam. Quantum-chemical calculations show that the Cfligand interactions are highly ionic and that 5f/6d mixing is absent, resulting in weak 5f→5f transitions and an absorption spectrum dominated by 5f→6d transitions.
Recent advances in heavy element chemistry spurred by innovative molecular design, new spectroscopic techniques, and relativistic theory have evolved our understanding of the divergence between the chemistries of 5f elements and their 4f counterparts. Historically, the primary distinction between these two series has been that early actinides display facile redox chemistry compared to lanthanides, with the exception of those with special electronic configurations such as Eu II (4f 7 ) and Yb II (4f 14 ) 1 . This key attribute is largely lost in all trans-plutonium elements (excluding nobelium) where the most stable oxidation state is 3+, akin to lanthanides 2 . This has led to the notion that their chemistries are lanthanide-like. The combination of this bias with their scarcity, short half-lives, and the need for specialized research facilities has further stymied progress in understanding these elements. However, it has been known for several decades that actinides from californium to nobelium have accessible 2+ oxidation states 3 . In fact, electro-and thermochemical measurements have shown a progressive stabilization of the 2+ state, ultimately revealing that No II is more stable than No III by 1.45 V 4-6 .
Of late, substantial efforts have been dedicated to isolating complexes containing f elements in either new or rare oxidation states for the purpose of determining their electronic configurations, bonding trends, spectroscopic features, and reactivity. Using strong reductants such as KC 8 and low-temperature reaction conditions, the first examples of Th II , U II , Np II , and Pu II tris-cyclopentadienyl derivatives, such as [Pu(Cp′) 3 ] 1− (Cp′ = trimethylsilylcyclopentadienyl), have been prepared, characterized, and their reactivities explored [7][8][9][10][11] . These studies inspired the pursuit of other molecular scaffolds for stabilizing U II , such as the U II bis(amidate arene) [U(TDA) 2 ] 1− (TDA = N-(2,6-di-isopropylphenyl)pivalamido) that acts as U I synthon because of the non-innocence of the ligand, as well as the monoarene [U(( Ad,Me ArO) 3 mes)] that exhibits stabilization of U II via δ-backbonding 12,13 . These reductive methods can also yield compounds whose oxidation states are difficult to describe, such as [K(18-crown-6) (THF) 2 ] 2 [U(η 6 -C 14 H 10 )(η 4 -C 14 H 10 )(μ-OMe)] 2 ·4THF 14 . Similarly, Cf II complexes have been prepared in the gas phase, as exemplified by the reductive elimination of CH 3 SO 2 from Cf III (CH 3 SO 2 ) 4 − to yield Cf II (CH 3  Once this solution is placed over the Al/Hg amalgam and agitated, it begins to darken over the course of a few minutes to a peach-brown colour (Fig. 1c). We have evidence that [Cf III (18-crown-6)(H 2 O)(CH 3 CN)I]I 2 is first reduced to [Cf II (18-crown-6) (H 2 O) 2 (CH 3 CN)]I 2 , and that the water is lost to the amalgam to yield the desired Cf(18-crown-6)I 2 complex. It should be noted that if the mother liquor is removed from the amalgam at this stage, its colour will revert back to the original yellow-green colour of the starting solution.
Although crystals can be isolated from this solution, they contain a mixture of Cf II and Cf III in the form of [Cf II (18- This issue is circumvented by growing crystals over the amalgam itself via vapour diffusion of a 1:1 mixture of diethyl ether and pentane into the mother liquor at 0 °C. Red-brown crystals of Cf(18-crown-6)I 2 primarily grow (in surprisingly high yield) between the amalgam and the wall of the glass vial (Fig. 1d). Even with this crystal growth method, where a large excess of reductant is always present, these samples degrade at room temperature, 0 °C, and even at −173 °C in less than 16 h because of radiolytic effects.
It should be noted that KC 8 can be used in place of the Al/Hg amalgam, but, as we discussed above, over-reduction can occur, rendering these reactions unpredictable. When this takes place, the californium becomes intercalated into the graphite and is difficult to extract. Given the scarcity of 249 Cf, the herculean efforts required to recycle it, and the real external radiation hazards that it poses to researchers, even on small scales (~5 mg), it is ill-advised to utilize these overly strong reductants to prepare Cf II .

Structure
Single-crystal X-ray diffraction studies reveal that Cf(18-crown-6)I 2 (Fig. 2b) adopts a structure type first established with Ln II cations (Ln II = Eu, Sm, Yb, Dy, Tm) that has been the subject of recent investigations for single-molecule magnetism with Tm(18-crown-6)I 2 23,25,26 . The Cf(18-crown-6)I 2 molecule consists of a Cf II cation that is bound within the cavity of the crown ether by the six etheric oxygen atoms The issue with applying the above synthetic methods to the preparation of low-valent trans-plutonium complexes is that, unlike early actinides, the standard reduction potential for An III → An II can be quite similar to the potential for An III → An°. In fact, even with uranium, the reaction of UCl 4 with LiC 10 H 8 results in the formation of U° nanoparticles 16 . Thus, it would be expected that the reactions of alkali-metal-based reagents with Cf III should result in the formation of Cf° unless the supporting ligands prevent over-reduction, because the potential for Cf III → Cf II is −1.525 V, whereas the Cf III → Cf° potential is −2.01 V 17,18 . Moreover, for americium, the potential for Am III → Am II is −2.28 V versus Am III → Am° at −2.00 V, rendering the formation of Am° at lower energy than Am II19 . Reagents such as KC 8 are expected to yield ~2.93 V of reduction potential and thus are anticipated to be too strong for preparing most trans-plutonium An II complexes.
Finally, one cannot neglect that all trans-plutonium isotopes have high specific activities and that both their ionizing radiation and concomitant formation of radiolysis products will influence the longevity of certain oxidation states. For example, radiolysis products rapidly reduce Bk IV to Bk III , but instead oxidize Cf II I 2 to Cf III OI via α-particle induced abstraction of oxide from silica reaction vessels 20,21 . The latter provides an excellent illustration of the energetic differences between nuclear and chemical reactions. Taken together, there are two hurdles to overcome to successfully isolate a Cf II complex. First, a reductant must be identified that is strong enough to reduce Cf III to Cf II but does not over-reduce it to Cf 0 . Second, synthetic methodologies must be developed that allow for rapid complexation, reduction, and crystallization before radiolytic re-oxidation occurs. In this Article we show that Al/Hg amalgams can be used to achieve the proper reduction potential, and that crown-ether ligands are appropriate for rapid crystallization of Cf II complexes.

Synthesis
We have previously shown that when Sm II is coordinatively saturated by large crown-ether ligands such as dibenzo-30-crown-10, the resultant complex is stable in aqueous media for days 22 . This was somewhat surprising, because Sm II normally reduces water and forms {Sm III 2 O 2 } oxo-or hydroxo-dimers. From electrochemical measurements it has been confirmed that the stability of Sm II crown-ether complexes is kinetic in origin. Here, this chemistry has been used to guide our synthetic efforts for Cf II because of the similarities between the Sm III to Sm II standard reduction potential (−1.66 V) with that of californium 17 . Given the difficulties in preparing soluble and anhydrous trans-plutonium starting materials, this former observation suggests that a route that is tolerant to a small amount of water might be initially more flexible than strictly moisture-and air-free methods.
After optimizing the following reaction with both samarium and americium, as well as with dicyclohexano-18-crown-6 (Extended Data Fig. 4 23 . It is important to note that, although the CfI 3 ·nH 2 O (n ≤ 6) is obviously hydrated, this material has been dried in vacuo, and this process is known to convert hexahydrates to monohydrates 24 . CfI 3 ·nH 2 O (n ≤ 6) rapidly decomposes in the presence of O 2 or even excess water (as does Cf(18-crown-6)I 2 ), so these reactions are conducted in a glovebox under an argon atmosphere. No further removal of coordinated or lattice water is necessary, because the Al/ Hg amalgam also removes the water from the reaction mixture via its reduction and subsequent precipitation of Al(OH) 3  Article https://doi.org/10.1038/s41557-023-01170-9 and by two trans iodide anions. This yields a slightly distorted hexagonal bipyramidal environment where the Cf II cation is located on an inversion centre, so the three Cf II -O bonds and one of the Cf II -I bonds are generated by symmetry (as are the other M II analogues presented in Table 1). Cf-O bond distances range from 2.685(5) to 2.666(5) Å and are slightly shorter than those found with Eu(II), but longer than those of Dy(II), as summarized in Table 1. Computational calculations (Computational methods) suggest that the Cf-O and Cf-I bonds are assigned as being mostly ionic because the natural localized molecular orbitals (NLMOs) show a highly polarized bond with low Cf atomic participation (up to 8%; Fig. 4). Furthermore, the 5f-orbital participation is essentially negligible. The Cf-I bond distance is 3.1563(4) Å, which is notably shorter than that of Eu II -I, but statistically longer than that observed for Dy II -I 23 . The shortening of this bond is ascribed to the mixing of iodide 5p z orbitals with Cf II 6d z 2 orbitals (Fig. 3a). Seeing as this is the first condensed-matter Cf II molecule that has been isolated and characterized, these bond metrics help bracket the upper and lower limits of the ionic radius of Cf II , which is currently not established. As previously mentioned, crystals can be obtained in low yield from reaction mixtures where the mother liquor has been removed from the amalgam. These crystals, although complex in nature, shed some light on the reductive process. X-ray diffraction studies show that these crystals are composed of [Cf II (18- This justification for the trivalent parent complex is supported by results obtained from the reaction of putative AmI 3 ·nH 2 O (n ≤ 6) with the cis-syn-cis isomer of dicyclohexano-18-crown-6, where a similar nine-coordinate metal complex was obtained with an iodide ion in the apical position (Extended Data Fig. 4). In addition, isolation of nine-coordinate trivalent lanthanide complexes with 18-crown-6 have been reported exhibiting coordination of a halide ion at the apex, with solvent molecules (typically water) coordinating opposite the halide ion [27][28][29][30] . Previous work has demonstrated that this structure type is observed primarily with the intermediate lanthanides (Pr through Dy, excluding Pm), which are commonly used for structural comparisons with several of the trans-plutonium actinides 31 .
Unlike the lanthanide structures reported in the literature 27-30 and the Am III complex we report here, the key difference between the two Cf complexes observed in [Cf II (18- [Cf III (18-crown-6)(H 2 O)(CH 3 CN)I]I 2 exists primarily in the apical position, which can be occupied by either a H 2 O molecule or an iodide anion. The structure refinement shows that the occupancies of the superimposed atomic positions are 0.6865 (15) for the oxygen atom and 0.3135 (15) for the iodide anion. In short, it contains ~69% Cf II and 31% Cf III . This is important to consider, because the coordinated water   molecule that does not exhibit split occupancy has a bond length of 2.413(2) Å, which lies between the observed M-OH 2 bond distances for comparable trivalent Sm and Gd structures, and is shorter than the Am III -OH 2 we report here, consistent with the actinide contraction (Extended Data Fig. 4), and closely matches the distances observed for other reported Cf III -OH 2 bonds 30,32-34 . Contrarily, the Cf-OH 2 distance observed for the water molecule in the apical position deviates from this trend, displaying a distance of 2.705(11) Å and necessitating further consideration of the entire molecule before assuming that the metal centre is trivalent.  (3) 3.1383 (10) 3.1171 (3) 3.09211 (18) 3.1563(4) Reported in ref. 23  The presence of an oxygen atom in the apical position of [Cf(18-crown-6)(H 2 O) 2 (CH 3 CN)]I 2 does not necessarily mean that it is a water molecule. It could be a hydroxide anion instead, which would yield a Cf III complex. However, a computational model analysis (Supplementary Table 2) shows that the Cf III -OH bond distance would be 2.270 Å, resulting in a weighted bond distance of 2.547 Å for Cf III -L (L = 69% OH − , 31% I − ), which is much shorter than the observed distances of 2.705(11) and 2.742(2) Å for water and iodide, respectively. The same calculated weighted bond distance for Cf II -L (L = 69% H 2 O, 31% I − ) is 2.791 Å, which is in much better agreement with experimental data (see Computational methods for further analysis). This type of substitutional disorder is well precedented in general (for example, the origin of the bond shift isomerization misidentification) but is also known from numerous actinide halide compounds such as Pu 2 31,35 . The present system is somewhat different from the latter of these examples and better parallels that of the Mo=O/Mo-Cl averaging, where two different molecules are present in the same crystal structure on the same crystallographic sites 36 .
Given that the solution spectrum did not show the presence of Cf III and that O 2 has been excluded, two possible oxidants remain: water and radiolysis. In this study, the amount of water is low, with no more than six equivalents of water present in the initial reaction, and any remaining water after the CfI 3 ·nH 2 O is placed under vacuum is removed via reduction by the amalgam. Thus, we conclude from six replications of this experiment and numerous others that radiolytic oxidation occurs and is the primary cause of oxidation. This latter result is expected given the high specific activity of 249 Cf 3 .
Having established the isolation of two distinct Cf 2+ metal complexes between the compounds reported here, a comparison of their coordination environments is warranted. The structure of [Cf(18-crown-6)(H 2 O) 2 (CH 3 CN)]I 2 consists of a nine-coordinate Cf II ion within a hula-hoop geometry. Similar to the Cf(18-crown-6)I 2 structure, the 'hoop' is formed from six approximately co-planar etheric oxygen atoms from the 18-crown-6 ligand. The apex is occupied by the aforementioned water molecule, and the two positions below the plane are filled by an acetonitrile molecule and a second water molecule (Fig. 2), with two outer-sphere iodide anions rounding out the asymmetric unit. This structure type is well established from lanthanide crown-ether complexes and from this present study with Am III28-30 .

Spectroscopy
Room-temperature, solid-state absorption studies of single crystal of Cf(18-crown-6)I 2 yield broadband transitions in the UV-vis region, as shown in Fig. 4a. By comparison, typical solid-state absorption spectra of Cf 3+ -containing compounds are dominated by distinct 5f→5f transitions 30,31 . Based on reported absorption data for Cf 2+ binary halides, it is anticipated that compounds containing Cf 2+ would also exhibit characteristic 5f→5f transitions 21,37-39 . However, the broadband features observed in the experimental absorption spectra reported here are, in fact, 5f→6d transitions (Fig. 4), and there is a notable lack of expected 5f→5f transitions. For comparison with the experimental data presented here, as well as relevant literature data, ligand-field density functional theory (LFDFT) calculations were used to predict the absorption spectrum for this system (Fig. 5a). It is worth mentioning that the 5f→6d transitions reported in the literature for CfCl 2 and CfBr 2 in particular are higher in energy than the 5f→6d transitions exhibited by Cf(18-crown-6)I 2 .
The expected 5f→5f transitions in this region are not observed as a result of either (or both) the presence of a formal inversion centre in the molecule or masking of the Laporte forbidden transitions by the electric-dipole-allowed 5f→6d transitions.
The absorption spectra collected at room temperature and 93.15 K display three main peaks, indicative of 5f→6d transitions, appearing in the visible region at 16,303 cm −1 , 18,662 cm −1 and 20,368 cm −1 . These bands are produced by excitation from different 5f orbitals to the lowest-lying 6d orbitals, which in this case correspond to 6d yz and 6d xz (Extended Data Fig. 1). In addition, a shoulder resolves on the peak centred at 20,368 cm −1 . This shoulder is indicative of the oxidation of Cf 2+ . Under the N 2 atmosphere of the temperature-controlled stage used for solid-state measurements, this oxidation is slowed, such that reliable measurements can be obtained before complete oxidation of Cf 2+ occurs. At room temperature, the oxidation occurs more rapidly under ambient conditions, as indicated by the increased intensity of the observed shoulder at 20,829 cm −1 and the gradual decrease of the 5f→6d transitions in the visible region (Fig. 4b).
The observed Cf 2+ 5f→6d transitions are further supported by photoluminescence measurements of single crystals at 93.15 K. As shown in Fig. 4c, excitation of these crystals with 546-nm (18,315 cm −1 ) light yields broadband emission at 14,087 cm −1 (710 nm), which corresponds to emission of the first group of 5f→6d transitions (16,303 cm −1 ). Compared to the photoluminescence observed from the isomorphous Sm(18-crown-6)I 2 under the same conditions, the emission exhibited by Cf(18-crown-6)I 2 is far less intense 23 . The solution-phase absorption spectrum for this system corroborates the presence of Cf 2+ before crystallization. As shown in Fig. 4d, the absorption spectrum of [Cf (18-crown-6) x ] 2+ in acetonitrile exhibits several broad features between 32,500 and 15,000 cm −1 . Most of these features can be attributed to the transitions identified in the solid-state absorption spectra of Cf(18-crown-6)I 2 . However, there are notable differences in the spectra obtained from these two phases that are probably the result of competing coordination environments of the [Cf(18-crown-6) x ] 2+ ion. An example of this is an observed colour change, which is dependent on the concentration of [Cf (18-crown-6) x ] 2+ in solution. The more concentrated solution from which Cf(18-crown-6)I 2 crystallizes is a dark peach-brown colour, but the colour of this solution shifts dramatically when diluted with acetonitrile, resulting in a blue-green solution (Fig. 5b). In the presence of excess acetonitrile, it is probable that the more labile iodide ions are displaced by solvent molecules. Given the possibility of different coordination environments in solution, the allowance of 5f→5f transitions cannot be overlooked. However, the pseudo-axial symmetry imposed by the coordination of the 18-crown-6 molecule restricts the overall intensity of these transitions compared to the dipole-allowed 5f→6d transitions (Extended Data Fig. 2).
When the solution containing [Cf(18-crown-6) x ] 2+ is exposed to oxygen, a vivid colour change from blue-green to bright yellow is observed, resulting in the spectrum shown in Fig. 5b. Similar to studies with Sm 2+ , the introduction of oxygen to this system rapidly oxidizes the divalent species to [Cf(18-crown-6) x ] 3+ . The solution absorption spectrum obtained following this colour change shows that the low energy features exhibited by the reduced species in acetonitrile are notably absent, and a broad, intense feature is observed at 27,758 cm −1 (360 nm) instead. This latter feature is comparable to the high-energy transition displayed by [Cf II (18- Fig. 3). Overall, the ambiguity of the solution-phase absorption data obtained for the divalent and trivalent species highlights the necessity for reliable solid-state structural evidence of the presence of Cf 2+ in systems like the one reported here.

Conclusion
The isolation of Cf(18-crown-6)I 2 provides insight for stabilization of the 2+ oxidation state in the later actinides (Z ≥ 98) in condensed-matter, molecular systems. Although radiolytic re-oxidation of Cf 2+ is a substantial hurdle to overcome, we have shown that the use of a more suitable reductant in the form of an Al/Hg amalgam is sufficient to reduce and maintain stability of the divalent state as well as prevent potential oxidants like water from jeopardizing the integrity of Cf 2+ in solution. Moreover, these conditions are appropriate and conducive to the isolation of crystalline products, allowing for the solid-state characterization of Cf 2+ molecules. Thus, the methods and results presented here support further exploration of the chemical behaviour of californium and beyond.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41557-023-01170-9.

Experimental
Warning! Americium-243 (t 1/2 = 7,364 years; specific activity = 200 mCi g −1 ) is a strong α emitter (up to ∼5.35 MeV), with γ radiation also associated with it (max peak of ∼75 keV), presenting internal and external radiotoxic hazards. The daughter isotope neptunium-239 (t 1/2 = 2.356 days; specific activity = 232 kCi g −1 ) represents roughly half of the activity of the material once secular equilibrium is reached within about two weeks and emits powerful β particles, γ radiation and X-rays. Californium-249 (t 1/2 = 351 years; specific activity = 4.09 Ci g −1 ) is a potentially substantial internal and external radiologic hazard due to its 5.8 MeV α particles (82.2% branch) and high-energy γ-rays with 388 keV (66% branch) associated with it. The daughter of californium-249 is curium-245 (t 1/2 = 8,423 years; specific activity 173 mCi g −1 ) and is an internal radiologic hazard from 5.36 MeV (93.2% branch) α particles. Radioactive materials were handled with care at a Category II radiologic facility.
Materials. The 249 CfCl 3 used in this experiment was isolated as a solid residue following purification procedures. The 243 AmCl 3 used in this experiment was acquired from an Am stock solution in 2 M HCl. DyI 2 (Sigma-Aldrich, 99%) was used as received. 18-crown-6 (Sigma-Aldrich, 99%) and tetrabutylammonium tetraphenylborate (Sigma-Aldrich, 99%) were used as received. Dicyclohexano-18-crown-6 (98%) was acquired as a mixture of cis-isomers from Sigma-Aldrich. The cis-syn-cis and cis-anti-cis isomers were separated from one another using the procedure established by Izatt and colleagues 40 . Hydroiodic acid (contains no stabilizer, distilled, 57 wt% in H 2 O, 99.99% trace metals basis, Sigma-Aldrich) was stored in a freezer at less than 0 °C until use. Ammonium hydroxide (28% NH 3 in H 2 O, ≥99.99% trace metals basis, Sigma-Aldrich) was used as received. Diethyl ether (VWR, anhydrous ≥99.0% stabilized) was used as received for all manipulations on the benchtop. For manipulations in the glovebox, diethyl ether and acetonitrile (Sigma-Aldrich, anhydrous, 99.8%) were dried over sodium (Sigma-Aldrich, cubes, contains mineral oil, 99.9% trace metals basis) and benzophenone (Sigma-Aldrich, 99%) and CaH 2 (Sigma-Aldrich, reagent grade, 95% (gas-volumetric)), respectively, before being stored in a glovebox under an argon atmosphere over activated molecular sieves (3 Å, 4-8 mesh, Sigma-Aldrich). The Al/Hg amalgam was prepared as follows. Aluminium (Sigma-Aldrich, beads, 5-15 mm, 99.9% trace metals basis) was charged in a 100-ml Schlenk flask and submerged in concentrated HCl. Vigorous bubbling occurred as the oxide coating was stripped from the aluminium beads. The flask was stirred vigorously by hand for 5 min. The solution was then carefully decanted. The aluminium beads were washed three times with diethyl ether, then the entire flask was placed under vacuum overnight. The flask was then pumped into a glovebox under an argon atmosphere. In a 20-ml scintillation vial, enough Hg metal (Sigma-Aldrich, ≥99.99% trace metals basis) was added to just cover the bottom of the vial. Three aluminium beads were removed from the Schlenk flask and careful placed over the mercury. The vial was capped and placed on a hot plate and heated to 250 °C. The amalgam was kept heated overnight and stirred occasionally. Eventually, the aluminium began to dissolve in the mercury, forming a dull grey, liquid Al/Hg amalgam.

Synthesis of [Cf II (18-crown-6)(H 2 O) 2 (CH 3 CN)][Cf III (18-crown-6) (H 2 O)(CH 3 CN)I]I 2 .
In a fume hood, ~4.0 mg of CfCl 3 ·nH 2 O (0.016 mmol) was dissolved in 2 ml of 2 M HCl. This solution was transferred to a 15-ml centrifuge tube. An equivalent volume (2 ml) of concentrated ammonium hydroxide solution was added to the centrifuge tube and agitated, resulting in a light-green precipitate in the form of Cf(OH) 3 . The mother liquor was separated by centrifugation, and the solid was washed with 4 ml of deionized H 2 O. These washes were completed in triplicate. After removing the final H 2 O washing, the Cf(OH) 3 solid was dissolved in four drops of concentrated HI. This solution was transferred to a 20-ml scintillation vial. The solution was dried under a gentle stream of N 2 gas, resulting in a black residue. The residue was washed three times with 2-ml fractions of diethyl ether and agitated vigorously with a plastic spatula to remove excess iodine. The washings were each carefully pipetted away, and, after the third washing, the resulting yellow-green powder (putative CfI 3 ·nH 2 O) was immediately placed under vacuum in a glovebox antechamber for 40 min. After 40 min, the CfI 3 ·nH 2 O was brought into the glovebox and immediately dissolved in 0.5 ml of acetonitrile, resulting in a light-green solution. Separately, 18.1 mg (0.032 mmol) of tetrabutylammonium tetraphenylborate was dissolved dropwise in acetonitrile and added to the solution containing CfI 3 . Again, separately, 4.4 mg (0.017 mmol) of 18-crown-6 was dissolved in acetonitrile dropwise and added to the vial containing CfI 3 . The solution was stirred gently and remained a light-green colour. This solution was then transferred to a 20-ml scintillation vial containing an Al/Hg amalgam. The vial was capped and agitated vigorously. Over the course of 5 min, the solution slowly transitioned from a light-green colour to a light peach-brown colour. This solution was separated from the amalgam via pipette, and the solution was centrifuged to remove excess Al(OH) 3 from the amalgam. This solution was transferred to a 6-ml shell vial that was inserted in a 20-ml scintillation vial. The outer scintillation vial was filled to 75% capacity with a 1:1 pentane-diethyl ether solution and capped for vapour diffusion. Overnight, a yellowish solid precipitated. Amidst the yellow solid were small yellow-green rhombohedral crystals of [Cf II (18- No evident change in the colour of the solution was observed. This solution was then pipetted over an Al/Hg amalgam in a 20-ml scintillation vial. The vial was capped and then agitated vigorously. Over the course of 10 min, the solution shifted to a dark peach-brown colour, indicating reduction of the Cf 3+ to Cf 2+ . This solution was transferred to a 5-ml centrifuge tube and centrifuged to separate out solid Al(OH) 3 generated by the amalgam. Meanwhile, a second amalgam prepared in advance from 1 ml of Hg (measured with a volumetric pipette) and an excess of Al metal was stored inside a 6-ml glass shell vial that was inserted inside a 20-ml glass scintillation vial. After centrifugation, the solution was pipetted over the Al/Hg amalgam inside the 6-ml shell vial. The outer 20-ml vial was filled to 75% capacity with a 1:1 solution of diethyl ether and pentane, and capped for vapour diffusion. The solution was stored on a cold plate at 0 °C. Over the course of the next 4 h, reddish-brown crystals of Cf(18-crown-6)I 2 formed, first around the Al/Hg amalgam where it contacted the glass of the shell vial, and then along the sides of the glass vial as well. The crystalline yield was ~50%.
Preparation of Dy(18-crown-6)I 2 . In a glovebox under an argon atmosphere, a 20-ml glass scintillation vial was charged with 20.0 mg of DyI 2 . Then, 3 ml of cold (−30 °C) dimethoxyethane (DME) was added to the vial and swirled to encourage dissolution of the DyI 2 . The dissolution process was slow, and the solution began to turn a shade of dark green. The reaction vial was stored in a glovebox freezer at −30 °C to maintain the stability of Dy 2+ in solution. Every 10 min, the vial was removed from the freezer and a metal spatula was used to grind the remaining solid to encourage further dissolution of the DyI 2 . The solution was then gently swirled and replaced in the freezer for another 10-min period. This was repeated until all solids dissolved, resulting in a dark-green solution. Separately, 12.7 mg of 18-crown-6 was dissolved in 3 ml of cold (−30 °C) toluene to create a stock solution and then stored in a glovebox freezer at −30 °C to maintain the temperature of the solution. Once both solutions were prepared and chilled to −30 °C, they were removed from the glovebox freezer. A disposable glass pipette was used to transfer three drops of the toluene solution containing 18-crown-6 to the DME solution containing DyI 2 . Neither solution was stirred. Both solutions were then carefully (avoiding agitation of the solution) replaced in the glovebox freezer to maintain a solution temperature of −30 °C. Both solutions were allowed to sit in the freezer for 5 min untouched. Then, the solutions were carefully removed from the freezer to avoid agitation, and three drops of the toluene stock solution containing 18-crown-6 were again added to the solution containing DyI 2 . This process was repeated continually, avoiding agitation of the DyI 2 solution at all cost. Additions of the 18-crown-6 stock solution yielded a dark brownish-green precipitate almost immediately. After the slow, dropwise addition of almost half of the stock solution described previously, dark-green microcrystals began to form on the walls of the glass scintillation vial. Slow, dropwise addition of the ligand solution sometimes resulted in crystals of Dy(18-crown-6)I 2 that were large enough for single-crystal X-ray diffraction studies, but the total crystalline yield was low (<20%). Further characterization and interpretation of Dy(18-crown-6)I 2 will be reported elsewhere, but the crystal structure is crucial to the bond-length comparisons for Cf(18-crown-6)I 2 .
Crystallography. Crystals of all three compounds were isolated in a glovebox under an argon atmosphere and placed on a glass slide under immersion oil before being transferred to a microscope on the benchtop. Single crystals were then mounted on 75-μm-inner-diameter MiTeGen loops before being transferred to the goniometer of a Bruker D8 Quest single crystal X-ray diffractometer with an IμS X-ray source (Mo Kα; λ = 0.71073 Å). A stream of N 2 gas was flowed over the crystals to preserve crystallinity and regulate the temperature of the sample. All collections were performed at 100 K. Subsequent data integration was performed using APEXIII software, and space-group determination was performed using XPREP. The following structure refinement was completed with the SHELXTL suite using the OLEX2 GUI 41,42 .
Spectroscopy. For solid-state spectroscopy, slides containing crystals of each compound under immersion oil were prepared in a glovebox under an argon atmosphere before being transferred to the stage of a Craic Technologies 20/20 PVTM Dual microspectrophotometer with a 75-W xenon lamp used for absorption spectroscopy. A Linkham temperature-controlled stage was utilized to study the behaviour of the compounds at room and cold temperatures. Measurements were collected at ambient temperature and conditions before being transferred to the temperature-controlled stage under a blanket of N 2 gas. The stage was cooled at a rate of 5 °C min −1 to −180 °C (93.15 K), and was allowed to equilibrate for 5 min before measurement. For solution-phase spectroscopy, samples were isolated in a glovebox under an argon atmosphere and sealed in quartz cuvettes (Starna cells, 0.4-cm path length) and then transferred to an Agilent Technologies Cary Series UV-vis-NIR spectrophotometer (Cary 6000i) for measurements. All solution-phase experiments were conducted at ambient temperature.

Computational methods
Geometry optimization. Given that the [Cf II (18-crown-6) (H 2 O) 2 (CH 3 CN)][Cf III (18-crown-6)(H 2 O)(CH 3 CN)I]I 2 crystal structure presents uncertainty on the nature of the capping ligand, there are three different scenarios to be explored. If the apical position is occupied by an O atom (69% crystallographic occupancy; main text) the metal centre can be either in the trivalent state with coordination of a hydroxyl ligand or in the divalent state with a water molecule bound to Cf. The third possibility is with the apical position occupied by an iodide anion (31% occupancy), indicating that the Cf would be in its trivalent oxidation state. Two approaches were thus considered: constrained geometry optimizations of the three structures and scanning the potential energy surface along the Cf-apical ligand bond coordinate. We have previously shown that the presence of outer-sphere iodide counterions is crucial to reproduce properly the solid-state Cf-O crown bond lengths 43 , and 13 iodide ions surrounding the molecule in the crystal structure had to be incorporated (Extended Data Fig. 6). These iodide anions correspond to the total number of counterions directly interacting with the molecular complex, which were frozen, and the only constraint imposed in the geometry optimization steps. To show the importance of these constraints, Supplementary Table 2 shows a comparison between the experimental and theoretical bond lengths with and without the inclusion of the surrounding iodide ions. Given the rather small root-mean-square deviations of the constrained geometry-optimized and crystal structures observed (Supplementary Figs. 1-4), we consider this a valid model to give insight into the electronic structure and geometries of these complexes. Conversely, potential energy scans (PESs) were performed for the three complexes (Extended Data Fig. 7), with only the apical ligand allowed to be relaxed during the scan; that is, only hydrogens were relaxed at every step. Considering both approaches and the occupancy factor for O and I atoms in the crystal structure allows for calculating weighted averages that might be correlated with the crystal structure bond distances. With geometry optimizations on the complexes surrounded by iodide counterions, the weighted values are 2.547 Å and 2.792 Å for apical OH − and OH 2 , respectively. Conversely, PES scans predict these values to be 2.409 Å and 2.789 Å, respectively. Thus, both approaches suggest that the oxygen in the apical position corresponds to a bound water molecule, because a hydroxide anion should display a significantly shorter experimental bond length.
It is worth mentioning that solvation effects were not considered, given that accurate solvation parameters have not been reported beyond uranium due to the lack of experimental data to optimize them. We advise the reader to consider calculated distances for this structure carefully, as they might not necessarily be accurate estimations; we have included them in an attempt to provide further insight into the most likely scenario for the experimental crystal structure.
In the case of Cf(18-crown-6)I 2 , only H atoms are required to be optimized. However, a full geometry optimization was performed, and the Cf-I bond lengths were substantially overestimated, which is common when reproducing metal-iodide bond lengths (Supplementary Fig. 1). The highly polarizable character of I − and the lack of parameters for https://doi.org/10.1038/s41557-023-01170-9 Cf to include dispersion or solvation effects accurately leads to inaccurate geometry optimizations, regardless of the functional/basis set utilized, particularly when solid-state structures are to be reproduced.
These calculations were performed in the ADF engine of AMS2021 44 . The generalized gradient approximation (GGA) Perdew-Burke-Ernzerhof (PBE) functional was used for this task, along with the Slater-type orbital (STO)-DZ basis set for H and C atoms and STO-TZP for Cf, O, N and I atoms. Scalar relativistic effects were included using the ZORA approximation as implemented in ADF 45 . For those calculations involving surrounding counterions, the positions of iodide anions were constrained to simulate the crystal packing.
LFDFT. The LFDFT approach has been used widely to reproduce the electronic structure of f-element complexes of mostly lanthanides, but has also been applied to actinide complexes [46][47][48][49] . The inability of time-dependent (TD)-DFT to predict multiplet electronic structure due to the lack of static correlation makes the LFDFT a more appropriate model, which account for both static and to a certain extent dynamic correlation. LFDFT calculations were performed on the optimized structure of [Cf(18-crown-6)(H 2 O) 2 (CH 3 CN)] 2− excluding the 13 iodide anions.
The intensities associated with the excitation spectra were calculated using the oscillator strengths f a, b of an a→b electron transition, with a and b representing the initial final electronic states (in atomic units), f a, b = 2 3 ΔE a, b ⟨a|μ|b⟩ 2 , where the electric transition dipole moments were formulated in the dipole-length form (μ = −r). The non-zero oscillator strengths for 5f→5f transitions arise from the mixing between actinide 5f states and 6d or other ligand states in a non-centrosymmetric ligand-field regime. The ligand states for the actinide complexes are referred to the one-electron ligand-to-metal charge-transfer states that possess large parentage of ligands 2p orbitals. We note that we use a static model somewhat equivalent to the Judd-Ofelt model 50,51 , where non-zero oscillator strengths exist through parity mixing due to the coordination geometry of the actinide ion. Thus, dynamical effects (for example, vibrations) were not considered in our calculations.
Molecular orbital localization. The Kohn-Sham orbitals obtained from single-point calculations were used to obtain the NLMOs using the Natural Bond Orbital module included in ADF 52 . The single-point calculation was performed using the hybrid GGA PBE0 functional in conjunction with the STO-TZP basis set for all atoms except Cf and I, for which the STO-TZ2P basis set was used.
Complete active space self-consistent field (CASSCF) and NEVPT2 calculations. Spin-orbit CASSCF calculations were performed in ORCA 5.0.2 using starting Kohn-Sham orbitals obtained at the PBE0/ TZVP level of theory (except H atoms, for which we used the small SVP basis set) 53 . Scalar relativistic effects were included via the Douglas-Kroll-Hess (DKH) Hamiltonian and spin-orbit coupling via state interaction in a second step. The active space chosen included the 5f and 6d shell in a CAS (10,12) calculation. Dynamic correlation on all the calculated states was included via the second-order perturbation theory, NEVPT2. To be able to make the calculation computationally feasible, we only considered 100 septets, 200 quintets, 100 triplets and 100 singlets. Because the 6d shell is more involved in bonding interactions than the 5f orbitals, the excitation energies were not reproduced properly ( Supplementary Fig. 5) and larger (oftentimes unaffordable) active spaces are needed. However, a good molecular orbital analysis can be done in terms of the molecular orbitals involved in vertical electronic transitions.