Transient Covalency in Molten Uranium(III) Chloride

Uranium is arguably the most essential element in the actinide series, serving as a crucial component of nuclear fuels. While U is recognized for engaging the 5f orbitals in chemical bonds under normal conditions, little is known about its coordination chemistry and the nature of bonding interactions at extreme conditions of high temperature. Here we report experimental and computational evidence for the shrinkage of the average U–ligand distance in UCl3 upon the solid-to-molten phase transition, leading to the formation of a significant fraction of short, transient U–Cl bonds with the enhanced involvement of U 5f valence orbitals. These findings reveal that extreme temperatures create an unusual heterogeneous bonding environment around U(III) with distinct inner- and outer-coordination subshells.


General considerations.
Attn: the natural uranium(III) chloride sample is radioactive and was shipped to APS and SNS facilities through corresponding radiation safety departments.Another potential hazard with the sample is related to the high temperature used during the neutron scattering measurement, which could result in severe burns if handled improperly.The chloride salts are relatively low hazard at room temperature.However, when the salts are melted at high temperature, they can interact with atmospheric moisture if exposed to room air to produce small amounts of HCl.This should only happen in the event of a cell rupture.

Synthesis of anhydrous uranium(III) chloride.
There are several approaches for synthesizing uranium chloride salts 1 .In this work, we used the reaction between uranium trioxide and hexachloropropene to form uranium tetrachloride, which was subsequently reduced to uranium(III) chloride employing zinc metal as a reducing agent.Uranyl nitrate salt (containing natural uranium Unat: 99.28 wt% U 238 , 0.71 wt% U 235 , 0.01 wt% U 234 ) was taken as the initial compound in our adopted synthetic scheme.In the first stage, uranium peroxide was precipitated from uranium nitrate solution by the reaction with hydrogen peroxide (Sigma-Aldrich, 30 wt.% in water): Obtained sediment was filtered, thoroughly washed, and then thermally decomposed in the air at 873-923 K to solid uranium trioxide: The obtained uranium trioxide was chlorinated by hexachloropropene (Sigma-Aldrich, ≥ 90% purity) at 433 K in a reactor with reverse coolant for 6 hours: The sediment of uranium tetrachloride was carefully washed from the hexachloropropene excess and other reaction products using carbon tetrachloride and dried under vacuum.
The uranium trichloride was prepared using the metallothermic reduction method by zinc (Sigma-Aldrich, 97.5 wt.% purity) excess at 873 K in a quartz cell with a neck: The cell was evacuated, filled with nitrogen, and placed in a preheated furnace.The unreacted zinc, uranium tetrachloride, and zinc chloride sublimed and condensed in the top cold part of the cell.The UCl3 collected from the bottom part was carefully transferred and stored in a glovebox under an inert argon atmosphere.

Uranium trichloride sample purity analyses.
Sample handling: the UCl3 salt was stored and handled in a MBraun Unilab Pro glovebox under high purity argon (Linde, > 99.999%), with total O2 + H2O concentrations kept below 5 ppm.The samples were weighed using an Ohaus PA84C scale (±0.1mg) inside the glovebox.
Composition analysis: three UCl3 samples were digested and analyzed using inductively coupled plasma optical emission spectrometry (ICP-OES, Perkin-Elmer Avio 200) to measure impurity concentrations.A solution of 5% v/v of HNO3 from the High-Purity Standards (HPS) Company was used as a blank for the ICP-OES as well as a solvent.The three samples which averaged 9.47 mg were each dissolved in 8 ml of 5% v/v of HNO3, and digestion performed in a Teflon® (PTFE) vessel in a programable microwave digester (PerkinElmer's Titan MPS Microwave).The oven was ramped to 448 K and 20 bar in 30 min, and the samples held for 10 minutes and then cooled down to 323 K for removal from the digester.The trace element measurements utilized three different concentrations + blank (0, 0.5, 1, and 2 ppm) for calibration.The UCl3 is considered 99.5% pure with respect to other cations (Table S1).
Structural analysis: crystallographic purity was confirmed by measuring XRD spectra in transmission mode.The samples were loaded into quartz capillaries with 0.01 mm thin walls and a diameter of 0.5 mm (Charles Supper) inside the glovebox and sealed with clay plugs to avoid exposure to air or moisture before removal to the diffractometer.Measurements were performed with a Rigaku 9kW Rotating Anode SmartLab X-ray Diffractometer with a Mo source (Kα1 = 0.70926 Å and Kα2 = 0.71359 Å radiation, zirconium filter).The XRD patterns were collected at room temperature over 5-50° 2θ with a step size of 0.01°, a counting time of 0.133 s, using a rotation rate of 3Hz.Rietveld refinement was performed using the Full Prof-Win Plot program 2,3 A representative XRD pattern is provided in Figure S1.No additional peaks due to contaminants, such as those of hydrates or oxides, were observed, the sample being pure to the limit of the technique's sensitivity (2-3 wt.%).The obtained c/a ratio of 0.5809(1) is in excellent agreement with the values of Schleid et al. 4 (c/a = 0.5809(1)), and Vogel et al. 5 (c/a = 0.5800(1)).
Thermal analysis: heat-flux DSC measurements were carried out using a Netzsch-Gerätebau GmbH 404 F3 Pegasus ® instrument in flowing 70 mL .min -1 of Ar (Linde, > 99.999%).Both heating and cooling thermal effects were analyzed to establish the melting and crystallization behavior of the salt.A sample of 19.1 mg was placed inside Ni liner and hermetically sealed in a high-pressure 100 µl chrome-nickel stainless-steel crucible as previously described. 6,7 he instrument was calibrated using the a zero heating rate method of the IUPAC 8 , which consists of heating the sample at different four heating rates varying overall by a factor of at least 10, which were 10, 5, 3, 1 K .min -1 .The extrapolated peak onset temperatures are determined as a function heating rate and extrapolated linearly to a rate of 0 K .min -1 .The accuracy for both heating and cooling was estimated to be within 3 K, and the error for the enthalpy of fusion on heating was less than 4%.
Overall, it was estimated the purity was 98 ± 1 mol% due to the DSC small aberration at 10 K .min - 1 run, with no additional phases identified by XRD, and ICP-OES only revealing trace concentrations of Ca, Fe, Mg, and Ni.

X-ray absorption spectroscopy experiments at the Advanced Photon Source (APS).
Uranium L3-edge X-ray absorption spectroscopy (XAS) measurements were acquired at 12-BM of the Advance Photon Source.The synthesized UCl3 powder was physically mixed with BN matrix and then sealed within a Kapton capillary (2.67 mm inner diameter).For transport to the beamline, the sample was placed into a quartz capillary, which was then flame-sealed, and doubly sealed within two epoxy-sealed Kapton capillaries all within an Ar glove box.XAS measurements were acquired at room temperature.The U spectrum was energy-aligned with respect to a Zr metal foil.XAS was acquired in florescence orientation using a seven-element Vortex detector.
Ejected photoelectrons are defined by their wavenumber (k) in relation to the absorption edge energy (E0) through the equation: The experimental EXAFS oscillations of each sample, χ(k), are extracted from the normalized XAS data using subtraction of a spline and a cutoff distance (RBKG) of 1.1 Å.For analysis of the EXAFS region, we use the EXAFS relationship given by: where the index, i, is considered the path index and the χ(k) is calculated as the summation over all paths.For fitting of the EXAFS, FEFF6 12 was utilized with the experimental χ(k) data weighted by k 3 for all fits.In eq.6, Fi(k), δi(k), and λ(k) represent the effective scattering amplitude, total phase shift, and mean-free-path of the photoelectron and each are derived from FEFF6.The degeneracy of the scattering path (Ni) was fixed to 9, based on the single-crystal XRD data.Therefore, the parameters still to be fit include, S0 2 , the many-body amplitude-reduction factor, Ri, the half-path length, σ 2 i, the Debye-Waller factor and C3,i, the asymmetry of the distribution.Variation of the C3 parameter was found to provide negligible changes to resulting EXAFS fits and was not varied in finally EXAFS fits.Additionally, a single nonstructural parameter for all paths, ΔE0, is varied to align the k = 0 point of the experimental data and theory.Conventional EXAFS fitting focused exclusively on the first shell U-Cl scattering path, and the number of variables (4) stayed below the number of independent data points (6.9) available in the UCl3 data with kmax = 12.0 Å -1 .

Neutron scattering experiments at the Spallation Neutron Source (SNS).
Sample preparation: the obtained uranium trichloride was crushed to fine powder in an argon glove box and added to a thin-walled NMR quartz tube (4.9635 ± 0.0065 mm O.D., 4.2065 ± 0.0065 mm I.D., wall thickness 0.38 mm; supplier: Wilmad-LabGlass).The quartz tube with UCl3 was then flame sealed near the top under high vacuum (Figure S4).Before shipment to the beamline, the sample was tested for heating/cooling to the temperatures of the experiment (1173 K) to ensure there are no unwanted reactions or breaches.
Neutron total scattering experiments were performed using NOMAD beamline 13 at the Spallation Neutron Source (SNS), Oak Ridge National Laboratory.Flame-sealed quartz tube with UCl3 was loaded into a vanadium sample can (Figure S4), and placed in the beamline furnace chamber, which was subsequently evacuated.A standard vacuum furnace was used to control the sample temperature at 1173 K.The total scattering structure factor, S(Q), was obtained from the neutron time-of-flight data following standard data reduction protocols, including absolute normalization using a vanadium rod for reference and background subtraction.Scattering patterns obtained at room and temperatures above the melting point (1146 K) were collected for a total integrated proton charge of 4 C and 12 C, respectively.
The experimental structure function is given by: where Icoh is coherent scattering intensity,   and   are the molar fraction and coherent neutron scattering length 14 of species i, respectively, and Q denotes the momentum transfer.
Real space pair distribution function (PDF), G(r), was obtained by the Fourier transform of S(Q): where  0 is the average number density of the system (number of atoms per unit volume).One can also define the partial subcomponents of G(r) to provide information as to which pair interactions contribute to the overall PDF at specific distances: where   is neutron weighting factor for an ionic pair factor defined as: where   is one for i=j and zero for i≠j.

Ab Initio molecular dynamics (AIMD) simulations.
We first performed classical polarizable ion model (PIM)-based 15,16 nanosecond equilibrium molecular dynamics (MD) simulations for UCl3 at 1173 K and 1 bar in the isothermal-isobaric ensembles (NPT).A snapshot with the experimental density 17 of UCl3 from the equilibrated portion of the trajectory was chosen to generate the initial configuration for AIMD simulations (cubic box with length of 22.86979 Å, 96 U and 288 Cl, number density 0.032103 atoms .Å -3 , which corresponds to the UCl3 experimental density of 4.589 g .cm -3 at 1173 K).These were performed using the PBE exchange-correlation functional [18][19][20][21] and Grimme's D3 dispersion correction 22 implemented through the Quickstep 23 module of the CP2K 6.1 package. 24,25 he MOLOPT basis set 26 of double zeta valence plus polarization (DZVPMOLOPT) in conjunction with Goedecker-Teter-Hutter (GTH) pseudopotentials 27 were applied for all ions.Uranium was treated with a medium-core pseudopotential containing 78 electrons in core (Xe 4 14 5 10 ) and the remaining 14 electrons (6 2 6 6 7 2 5 3 6 1 ) in valence.The DFT+U method was employed for uranium ions with a partially filled 5f orbitals, using the Hubbard parameter Ueff = 2.00 eV recommended for CP2K. 24,25 he DFT+U method for f elements is known to converge to multiple electronic states.With three alpha electrons in seven 5f orbitals there are 35 different orbital occupations.To guarantee that the lowest electronic state is found, we enforced a specific orbital occupation using the &ENFORCE_OCCUPATION section in CP2K.The occupation constrains were released after the first 20 SCF cycles.The lowest energy electronic state identified from all possible orbital occupations was used in the subsequent AIMD simulations.The orbital transformation method was employed with a FULL_ALL preconditioner and a conjugate gradient minimizer to achieve and accelerate the SCF convergence.A Nose-Hoover chain thermostat 28 with a velocity rescaling time constant of 1.0 picoseconds (ps) was used for the temperature coupling.A trajectory of 100 ps in length at 1173 K was generated using a 1.0 fs timestep.The last 60 ps were used for structural analysis.The total neutron S(Q) and G(r) were calculated from the AIMD simulation trajectory using the software package debyer. 29For the Raman spectrum, we used the 40-70 ps trajectory interval for our simulations.

Reverse Monte-Carlo (RMC) modeling.
The neutron total scattering pattern of molten UCl3 (1173 K) was fitted through the reverse Monte-Carlo (RMC) approach with the AIMD cell expanded to a 4 × 4 × 4 supercell, targeting an optimal match with experimental data in a metropolis manner.The RMCProfile package 30,31 with such an algorithm implemented was employed to fit the neutron total scattering data in both real and reciprocal space.A total of 24576 ions (6144 U and 18432 Cl) in a cubic box with length of 91.479160 Å was used for the RMC fitting, with ∼500 moves per ion generated and ∼60 moves per ions accepted, overall.The minimum distance for each ionic pair was used as constraint to allow the system to fully relax.To guarantee the generality and good statistics of the obtained RMC results, the fit was performed using 15 different configurations and an average was taken over all RMC runs to obtain partial pair distribution functions.Hence, the resulting U-Cl distance in the molten state was reported considering a standard deviation: 2.78 ± 0.01 Å.

Chemical bonding analyses.
We used CP2K to produce the electron localization function (ELF) 32,33 cube file from the AIMD trajectory snapshot.Ground-state single-point density functional theory (DFT) calculations (PBE0-D3 level of theory) 34 were performed using the Gaussian 16, Revision A.03 program package 35 on the representative UCl9 6-and UCl8 5-clusters based on the experimental geometry (solid state) or AIMD snapshot structure (molten state).This enlisted unrestricted Kohn-Sham methods, with the aug-cc-pVDZ basis set 36 for the chloride atoms.The Stuttgart small-core (SSC) potential to account for relativistic effects and the associated contracted basis set 37 was used for U, and the cluster was treated as a negatively charged quartet with three unpaired f-electrons.The bonding in UCl3 and Wiberg bond indices (WBIs) were examined by using the natural bond orbital (NBO) methodology 38 , as implemented in the NBO7 program. 39,40 e further analyzed the electron densities at bond critical points (BCPs) for each U-Cl bond in the representative clusters at molten and solid states using the quantum theory of atoms in molecules (QTAIM) approach 41 , which is based on topological features of the electron density.The QTAIM analysis was performed using the Multiwfn program. 42The electron densities at BCPs for each U-Cl bond are summarized in Table S4.Molecular orbital diagrams were drawn with an isovalue of 0.02 a.u.Model representations in the figures were prepared using the UCSF Chimera software. 43For the density of states (DOS) analysis, we performed spin-polarized (ferromagnetic state) single point calculations using the VASP 6 software 44,45 on the snapshot configuration from UCl3 AIMD trajectory.The calculations were done using PBE+U DFT functional. 46Hubbard U value of 4 eV was used for U 5f electrons.A planewave basis set of 400 eV and standard PAW pseudopotentials were employed.Due to the large unit cell the Brillouin zone was sampled using the Gamma point approximation.

U-Cl coordination number analysis.
Considering   as the distance between the i th Cl -and a U(III) ion and  † as the location of the boundary of the first chloride coordination shell, the coordination number () is defined using a smooth function, f: † is obtained from the first minimum of the U-Cl g(r), which appears after the first peak. Cl is the total number of chloride ions.The function, f, allows smooth transitions of Cl -ions across the boundary.The powers, (12, 24), ensure that both the smoothness and correctness of CN are maintained as discussed in our earlier studies. 47We compute the free energy profile of CN from its probability distribution, P(CN), i.e., W(CN) = -kBT ln[P(CN)], where T is the temperature and kB in the Boltzmann constant.

Survival probability correlation function.
The survival probability correlation function, C(t), is defined as 48 : C(t)=⟨Pi(t,t + δt,)⟩i,t/⟨P(t,t)⟩i,t. (12)   where, the survival probability, P, is assigned a value of 1 when a Cl -is found in the first solvation shell of U(III) at both times t and t + δt.Otherwise, P is set to zero.⟨…⟩i,t indicates averaging over U-Cl pairs and time.We carried out the C(t) calculations for different cutoff distances representing the boundary of the coordination shell (from short U-Cl bond distance to the actual cutoff distance obtained from the U-Cl g(r)), and Fourier-transformed C(t) to resolve the vibrational signatures of the coordination shell.The expression for the Fourier transformed spectra, FT-C(ω), is: max 0 (13)   tmax is the maximum correlation time and ω is the frequency of U-Cl symmetric bond stretching vibration., amplitude-reduction factor; r, interatomic distance; σ 2 , Debye-Waller factor; ΔE0, a single non-structural parameter for all paths, was varied to align the k=0 point of the experimental data and theory.*Fixed parameters.k -window: 2.7 -12.0 Å -1 ; r -window: 2.0 -3.2 Å were used for the EXAFS fit.

Figure S1 .
Figure S1.Room temperature refined powder transmission XRD pattern for UCl3.

Figure S2 .
Figure S2.DSC curves for the UCl3 sample on heating at the four rates of 1, 2.5, 5 and 10 K .min - 1 .Inset is an expanded view of the peaks obtained at 5 K .min -1 and 10 K .min -1 .

Figure S3 .
Figure S3.DSC scan at 5 K .min -1 used to obtain the enthalpy of fusion for UCl3 sample measured.

Figure S4 .
Figure S4.L3-edge EXAFS spectrum of the synthesized UCl3 salt at room temperature where k is the energy of the photoelectron in wavenumbers and k 3 χ(k) is the k 3 -weighted EXAFS function.Data between 2.7 and 12.0 Å -1 were Fourier transformed using a Hanning window to obtain realspace information presented in Figure 1A (main text).

Figure S5 .
Figure S5.UCl3 sample in a sealed quartz tube and vanadium containment used for neutron scattering experiments at NOMAD beamline.

Figure S6 .Figure S7 .
Figure S6.(A) Neutron structure factors, S(Q)s, obtained from neutron scattering measurements of UCl3 at room temperature (blue) and 1173 K (red).(B) Comparison of S(Q)s for molten UCl3 at 1173 K, obtained experimentally (solid red line) and using AIMD simulations (dashed black line).

Figure S10 .
Figure S10.Total DOS and projected DOS of UCl3 molten salt using PBE+U level of theory (top panel).Projected DOS of the most reduced U center and surrounding Cl using PBE+U level of theory (bottom panel).

Table S2 .
12mparsion of UCl3 crystallographic data (coordination number (CN) and average U-Cl bond length from the available crystal structures identified by their respective Cambridge Crystallographic Data Centre (CCDC) number) and our EXAFS fitting parameters derived from the UCl3 sample at room temperature.1 errors in U-Cl bond distance are computed from the covariance matrix of the non-linear minimization of the EXAFS fit.12 2