You are on page 1of 12

Physica B: Condensed Matter 570 (2019) 194–205

Contents lists available at ScienceDirect

Physica B: Condensed Matter


journal homepage: www.elsevier.com/locate/physb

Vibrational properties of uranium fluorides☆ T


a,∗ a a c b
Andrew Miskowiec , Ashley E. Shields , J.L. Niedziela , Yongqiang Cheng , Paul Taylor ,
Guillermo DelCulb, Rodney Huntb, Barry Spencerb, John Langforda, Douglas Abernathyc
a
Nuclear Nonproliferation Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA
b
Isotope and Fuel Cycle Technology Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA
c
Neutron Scattering Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA

A R T I C LE I N FO A B S T R A C T

Keywords: Multiphase mixtures of the uranium fluoride compounds UFx with x = 3, 4, 4.5, 5, whose local U–F bonding
Uranium geometry is conserved, may result from UF6 reduction. One method for identifying multiphase mixtures is op-
Phonons tical vibrational spectroscopy, but experimental Raman and IR spectra for UFx compounds are difficult to obtain.
Density functional theory To supplement those experimental measurements, we use density functional perturbation theory (DFPT) with
Inelastic neutron scattering
the on-site Coulomb correction (Hubbard + U) to calculate the phonon normal modes, their IR intensity, and
Lattice dynamics
their Raman activity for UF3, UF4, U2F9, and UF5. In addition, we use neutron spectroscopy to measure the
phonon density of states of the most stable UF4. Our measurements on the Wide-Angular Range Chopper
Spectrometer at the Spallation Neutron Source indicate that UF4 has a broad phonon spectrum centered at about
30 meV, but modeling of this spectrum indicates significant multiphonon scattering is present. DFPT calculations
for U2F9, which shares structural features with UF4 and UF5, suggest a dynamic instability and we speculate that
previous crystallographic measurements of this compound were possible only in multiphase mixtures. Analysis
of the detailed atomic motions of phonons in UF5 indicate that a series of high energy modes (between 69 and
75 meV) are described by motion of single-coordinated F atoms (bonded to only one U atom). These single-
coordinated F atoms are unique to the UF5 structure and those modes could be used to distinguish UF5 in a
multiphase mixture using optical spectroscopy.

1. Introduction in all four compounds is close to a uranium-centered square antiprism


(Fig. 1) [2–5]. At least one piece of experimental evidence supports this
Uranium tetrafluoride (UF4) is an important intermediate material hypothesis as well: Thoma et al. reported that UF3 and UF4 form a solid
in the nuclear fuel cycle; however, significantly less work regarding solution below 865 °C at all UF3 to UF4 ratios [6].
characterization and understanding of UF4 exists compared to, for ex- Stoichiometric differences in these compounds are provided by
ample, UO2. The primary technological interest in UF4 as a material is differences in the ligand bond order. In UF4, all fluoride ligands are
its place as a necessary intermediate between UF6 and U metal. coordinated to two U atoms, whereas in UF5 only seven of every eight F
Incomplete reduction of UF6 to UF4 could result in uranium penta- ligands are coordinated with two U atoms. In other words, UF5 contains
fluoride (UF5) contamination, potentially disrupting the quality of the some “dangling”, single-bonded fluorines. Coordination in UF3 and
product material. Diuranium nonafluoride (U2F9) is also a plausible U2F9 is U metal centers coordinated by nine F in a tricapped trigonal
contaminant and this compound has been identified as having a cubic prism. It is easy to imagine the pathway between the U–F coordination
structure [1]. Considering also UF3, the plausibility of coexistence of in UF4/UF5 and U2F9, since the tricapped trigonal prism in U2F9 is es-
UF3, UF4, UF5, and U2F9 is provided in part by the structural similarities sentially the square antiprism in UF4/UF5 with an additional terminal F.
to the crystal structures. UF3, UF4, UF5, and U2F9 have spacegroups Two phases of UF5, α-UF5 and β-UF5, have been identified, but it is the
P63cm, C2/c, I 4̄2d , and I 4̄3m but the underlying uranium coordination latter that shares structural similarities with UF4 and it is the one


This manuscript has been authored by UT-Battelle, LLC, under contract DE-AC05-00OR22725 with the US Department of Energy (DOE). The US government
retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide
license to publish or reproduce the published form of this manuscript, or allow others to do so, for US government purposes. DOE will provide public access to these
results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan).

Corresponding author.
E-mail address: miskowiecaj@ornl.gov (A. Miskowiec).

https://doi.org/10.1016/j.physb.2019.06.049
Received 22 January 2019; Received in revised form 10 May 2019; Accepted 21 June 2019
Available online 24 June 2019
0921-4526/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
A. Miskowiec, et al. Physica B: Condensed Matter 570 (2019) 194–205

Fig. 1. Structure of (a) UF3 [2], (b) UF4 [4], (c) U2F9 [3], and (d) 1 × 1 × 2 supercell of β-UF5 [5]. Crystal structure visualizations are produced with VESTA [18].

investigated here [7]. Because of the structural similarities in the set of complement to optical spectroscopy for UF4 is motivated by the pene-
compounds UF3, UF4, U2F9, and β-UF5, we are presented an opportunity tration of the neutron, the lack of vibrational selection rules, and the
to study the impact of differences in local U–F bonding on lattice dy- reasonable neutron cross-section of UF4 (compared to the weak Raman
namics. Two other UFx compounds, α-UF5 and UF6, contain distinct intensity). With the software package developed at the Spallation
octahedral U–F coordination, their lattice dynamics are understood, Neutron Source (SNS), the DFT calculations, including multiphonon
and a comparison of these dynamics to the antiprismatically co- effects, can be directly compared to the full dynamic structure factor
ordinated compounds may not be possible [1,8,9]. [12].
Optical spectroscopy is a critical tool for chemical characterization The secondary impetus for this study is to provide additional dis-
on the micron length scale. However, the identification of UF4, UF5, and cussion of the applicability and accuracy of the DFT + U method as it
U2F9 by Raman and infrared (IR) spectroscopy is experimentally chal- pertains to lattice vibrations in actinide systems. Neutron spectroscopy
lenging. First, UF4 is evidently a weak Raman scatterer [10]. Most re- is the most appropriate tool for measuring phonon density of state
cent high-resolution measurements by Villa-Aleman and Wellons show (pdos); however, for actinide oxides, only the pdos of PuO2 and UO2
that the Raman spectra of UF4 is complex (a symmetry analysis in- have been investigated [20–22]. As UO2 is a primary chemical form of
dicates 42 Raman-active bands should exist in UF4), with many nearly nuclear fuel, thermal transport properties are clearly important. Pre-
overlapping, low-frequency modes. Major bands appear in the < 47 vious comparisons of the pdos calculated using DFT + U to experi-
meV regime, with some additional Raman intensity appearing in broad mental neutron spectroscopy data showed that the one-phonon, neu-
shoulders above 75 meV. The authors noted that some weak spectral tron-weighted dos calculated from DFT + U did not entirely reproduce
intensity is observed above 75 meV as well. In the IR, the strongest the experimental spectrum [21]. Whether this problem extends to ur-
spectral features in UF4 are between 25 and 50 meV, but the IR spec- anium fluorides is pertinent.
trum of UF4 has not been measured with modern infrared spectroscopy Nevertheless, it is clearly valuable to apply the DFT + U method for
instrumentation [11]. this investigation for two reasons. First, performing experiments with
On the other hand, UF5 is identified by a single strong Raman band actinides is difficult because of handling requirements, and neutron
near 79 meV, but the production of pure UF5 is notoriously difficult spectroscopy measurements typically require multi-gram sample
[13–17]. One possible explanation for intensity in Villa-Aleman and quantities. Supplementing experimental data with calculation is clearly
Wellons' spectra above 75 meV is the presence of small amounts of UF5. desirable in this context. Second, although the DFT + U method cannot
There are no reported optical vibrational spectra of UF3 or U2F9. reproduce all experimental observables (such as excited states), it is the
Our primary motivator here is to supplement the high-quality UF4 only method currently available for systems of this size and complexity
Raman spectra of Villa-Aleman and Wellons' with associated DFT + U that could feasibly be used to calculate important lattice dynamical
calculations and inelastic neutron scattering. We are motivated to properties such as phonon linewidths [23]. The UF4 system has a lower
provide fundamental knowledge of the optical phonon spectra to assist symmetry crystal structure with significantly more phonon modes than
in the separation of UF4 from UF5 and U2F9 in materials produced from UO2 and PuO2 and is a next step in terms of complexity.
UF6 reduction reactions. To do this, we calculate the phonon modes in
UF4, UF5, and U2F9, their spectral intensities in the IR, and their activity 2. Computational details
in Raman based on symmetry with the DFT + U method. Because of the
structural similarities in all four UFx compounds mentioned, we are We use the Vienna ab-initio simulation package [24,25] with the
also interested in exploring the relationship between U–F coordination projector-augmented wave approach [26,27]. Details of the calcula-
and lattice dynamics. tions performed on UF3, UF4, UF5, and U2F9 are presented in Table 1.
To supplement these calculations, we provide measurements of the All calculations were spin-polarized, performed with a cutoff energy of
inelastic neutron spectra of UF4. The use of neutron spectroscopy as a 500 eV, a 5 × 5 × 5, Γ-centered k-point grid generated via the

195
A. Miskowiec, et al. Physica B: Condensed Matter 570 (2019) 194–205

Table 1 symmetry considerations of the unit cell and the Wyckoff positions of
DFT calculation parameters. atoms. Their activities in IR and Raman are proportional to the di-
System Cell No. Atoms Spacegroup Functional Ueff/eV electric susceptibility tensor (χij) and its spatial derivative with respect
to the phonon mode atomic displacements, respectively. For IR modes,
UF3 Conventional 6 U, 18 F P63mc PBEsol [30] 0.9 the imaginary part of the dielectric constant is given by
UF4 Conventional 12 U, 48 F C2/c PBEsol 1.3
U2F9 Conventional 8 U, 36 F I 4̄3m PBEsol 1.8 4π 2 |f n |2
UF5 1 × 1 × 2 Supercell 16 U, 80 F I 4̄2d PBEsol 2.3
I [ε (ω)] =

∑ 2ωn
δ (ω − ωn ),
n (3)

where Ω is the unit cell volume and f are the oscillator strengths for
n

Monkhorst–Pack scheme, and with smearing applied by a second-order each mode, n,


Methfessel–Paxton scheme [28,29]. Ion geometries were relaxed by
forces to within 0.1 meV/Å and phonon frequencies and eigenvectors fin = ∑ ZI∗,ij uIn,k ,
were calculated with density functional perturbation theory (DFPT). Ii (4)
Valence configuration for U atoms was [Rn]7s26d25f2 and for F atoms ΔFIj
with the Born effective charges ZI∗, ij = − ΔE equal to the force on atom I
was [He]2s22p5. i
in direction j from an electric field component Ei. The Born effective
charges can also be calculated via perturbation theory [39]. Equation
2.1. LSDA + U method (3), given the Born effective charges, depends only on the eigen-
frequencies and the eigendisplacements through the oscillator
The on-site Coulomb repulsion, quantified in the so-called Hubbard strengths. Hence, solving Eq. (2) and calculating the Born effective
U parameter, was introduced as a response to the failure of LDA to charges is sufficient to calculate the IR activities.
appropriately describe d and f shell electrons. In particular, its effect is Raman activity of phonon modes is determined by the mode irre-
to encourage idempotency in the on-site occupation matrix, ρσ, where σ ducible representations. Modes that preserve product combinations of
is the spin index. We used the approach of Dudarev et al. which is to Cartesian coordinates for a given crystallographic pointgroup are
describe the Coulomb term and the exchange term with a single Raman active. Generation of the mode irreducible representation is
spherically averaged Ueff = Ū − J̄ , where the bar indicates spherical done with the phonoPy software package, which calculates the in-
averaging, and U and J are the on-site Coulomb and exchange inter- variant symmetry operations for each phonon eigenvector, and as-
action parameters, respectively [31]. The Hubbard term is added to the signed Raman activity by us [40].
exchange-correlation energy

Ueff ⎡ ⎤ 3. Experimental
Exc +U = Exc +
2
∑ ⎢∑ (ρmσ1,m1 ) − ∑ (ρmσ1,m2 ρmσ1,m2 )⎥,
σ ⎣ m1 m1, m2 ⎦ (1) The sample of UF4 is produced by the reduction of UF6 by NOx, as
−3
where m1 and m2 go from −3 to 3 for the f to f levels and σ = ± 1/
3 originally discussed by Geichman et al. [41,42]. The production pro-
2. The Hubbard Ueff parameter is not restricted to LDA functionals. ceeds in two steps. First, nitrosylium hexafluorouranate (NOUF6) is
Beridze et al. used the GGA-level PBEsol functional and the linear re- prepared by exposing frozen UF6 to liquid N2O4 in a two-step reaction:
sponse approach of Cococcioni and de Gironcoli to calculate the op- 2UF6(s) + N2O4(l) → 2NO2UF6(s), NO2UF6(s) + N2O4(l) →
timal value of Ueff for UF4 to be 1.3 eV and for UF5 to be 2.3 eV [32,33]. NOUF6(s) + N2O5(l), which must be performed below the boiling point
Their results agreed with experiment with regard to lattice parameters of N2O4 (21.7 °C). Next, UF4 is produced by reacting H2 with the in-
and unit cell volume to within 1% and predicted a band gap of 1.0 eV. termediate:
We use these values for Ueff with the PBEsol functional as our target 2NOUF6(s) + H2(g) → 2UF4(s) + 2HF(g) + 2NOF(g).
value. To our knowledge, no DFT calculations of any kind have been The solid NOUF6 was held at 350 °C and H2 was flowed over the
performed on U2F9, and as such no optimal value of Ueff is known. material for 120 min. Any remaining NOx compounds are volatilized
Hence, we use a value of 1.8 eV for Ueff for U2F9 (i.e., the average value and removed during this step. A variety of tests (thermogravimetric and
for UF4 and UF5). The PBEsol functional has been shown to improve pyrohydrolysis) showed only F anions, and Davies–Grey analysis in-
lattice constants and bond distances over traditional GGA, which is dicated 76.9 wt% U (theoretical 75.8 wt%) [43]. X-ray diffraction of the
known to overestimate bond distances, by restoring the gradient ex- powder confirmed the material was UF4 (see Supplemental Informa-
pansion for exchange [30,34]. tion).
To avoid metastable electronic solutions, a method of occupation Structural parameters were measured using CuKα radiation on a
matrix control developed by us was used to UF4 and UF5 [35,36]. No Proto AXRD benchtop powder x-ray diffractometer (Proto
metastable states appeared for the UF5 case. For UF3, the number of Manufacturing), calibrated with NIST SRM 640 (Si powder line position
possible states is immense (over 130,000); instead, we use the U- standard) prior to measurement. A temperature controlled stage
ramping method which has been shown to effectively avoid metastable (Anton–Parr) was used for temperature-dependent measurements. Data
states in metal oxides [37]. were collected every 10 K between 40 and 190 °C over the range
16 < 2θ < 115° for approximately 2 h per point. Results were ana-
lyzed with GSAS-II and the sequential refinement method [44].
2.2. Infrared activities
Neutron spectroscopy measurements were performed on the Wide
Angular-Range Chopper Spectrometer (ARCS) at the SNS [45]. The
The vibrational eigenfrequencies and eigenmodes are the solutions
ARCS instrument is designed to study phonons in detail due to its 120°
to the harmonic equation
equatorial angular range coverage and tunable incident neutron energy.
1 ∂2E (R) ARCS can operate with incident neutron energies between 20 and
det − ωn2 = 0, 1400 meV. Phonon scattering intensity is proportional to the product of
MI MJ ∂RI ∂RJ (2) 2 2
Q2 and the Debye–Waller factor, exp−Q ⟨u⟩ , where 〈u〉2 is the mean-
where I and J index over all atoms, i and j are Cartesian indexes, and ωn square atomic displacement. To model the neutron spectra, we use the
are the eigenfrequencies. The energy derivative in Eq. (2) is solved software package OClimax, developed at SNS for measurements on the
using perturbation theory [38]. VISION spectrometer, but adapted for use on ARCS [12]. We expect
Optical vibrational modes are IR or Raman active based on broad phonon scattering in UF4 because of its large density of modes

196
A. Miskowiec, et al. Physica B: Condensed Matter 570 (2019) 194–205

below 80 meV and the low energy resolution of a neutron spectrometer


compared to a light spectrometer. OClimax uses results from DFT cal-
culations as inputs; thus, we can identify features in the phonon spectra
and correlate them directly to the underlying atomic displacements.
Furthermore, no fitting of the predicted spectra from OClimax to the
neutron data is performed; the comparison is done with no free para-
meters. In addition to the relaxed (zero-force) crystal structure, phonon
eigenvectors and eigenfrequencies, as calculated using DFPT, are used
to calculate the dynamic structure factor according to Eq. (1) and Eq.
(2) in Cheng et al. by OClimax [12].
Uranium tetrafluoride has a relatively weak neutron cross-section
compared to hydrogen. The intercalation of H into UF4 samples is
possible via several routes. First, UF4 is not particularly soluble in
water, but the physical adsorption of water is possible. Second, reports
exist of UF4 hydrolysis at higher temperatures as an intermediate re-
action to complete oxidation in atmosphere [46]. A slower reaction
process involving water (nominally of the form 2UF4 + 4H2O →
UO2 + UO2F2 + 6HF + H2) at ambient temperatures has not been
definitively studied but cannot be excluded. Related, although UF4 is
relatively stable, the oxidation intermediate UO2F2 is highly hygro-
scopic and may form as a by-product of the UF4 production process or
as a degradation product [47]. Third, the generation of UF4 via the NOx
reduction route generates HF as a by-product, which may remain in the
crystal matrix. For these reasons, the sample was annealed under va-
cuum (unsealed) at ARCS at 352 °C for several hours prior to mea-
surement. Measurements of the total scattering intensity before and
after annealing indicated that only a small amount of H, by weight, was
initially present, but due to its large incoherent neutron cross-section,
this amount was sufficient to affect the measured spectra. Therefore, we
present data collected at 100 K after vacuum annealing.

4. Results and discussion


Fig. 2. Electronic density of states for (a) UF3, (b) UF4, (c) U2F9, and (d) UF5
decomposed by spin up (blue lines) and spin down (red lines). The bands
4.1. Electronic structure
nearest the Fermi level are occupied in all cases by f electrons, and their sum
corresponds to the expected number of unpaired f electrons for the appropriate
Although this work is primarily concerned with the vibrational valence state of U (+3, +4, +4.5, and +5 for UF3, UF4, U2F9, and UF5, re-
spectra, the electronic structure is acquired during the course of cal- spectively). In the case of U2F9, the top valence band is half-filled, implying
culating the former. Fig. 2 shows the electronic density of states (edos) metallicity, although U2F9 is known to be an insulator. (For interpretation of the
for each UFx compound in increasing order of valence. The relatively references to colour in this figure legend, the reader is referred to the Web
sharp features in the UF4 edos are a consequence of the lower symmetry version of this article.)
of the UF4 crystal, with nine unique atomic positions compared to four,
three, and four for UF3, U2F9, and UF5, respectively. The ground state of Table 2
all structures are high-spin configurations, in agreement with magnetic Orbital character of U bonding electrons.
susceptibility measurements [51–53].
System s p d f
The four structures in the edos share the following common fea- e− e− e− e−
tures: (1) a broad (about 4 eV) bonding region 5.9, 3.8, 2.6, and 1.8 eV
below the Fermi level in UF3, UF4, U2F9, and UF5, respectively; (2) the UF3 0.19 0.83 1.29 0.69
top of the valence band containing exactly three, two, or one electron(s) UF4 0.12 0.38 0.81 0.68
U2F9 0.08 0.43 0.51 0.48
per U atom for UF3, UF4 and U2F9, and UF5; and (3) a set of unoccupied UF5 0.05 0.18 0.51 0.43
conduction bands approximately 1 eV above the Fermi level. The pro-
jection of the edos into orbital and magnetic angular momentum
components (i.e., onto the spherical harmonics) for UF4 was described contribution to the edos of each of the angular momentum components
in more detail by Miskowiec, showing that while the bonding region integrated over the bonding region. Of particular note for this plot are
includes contributions almost exclusively of p-type character from the F the following features. First, the s-type orbital component is weak for all
atoms, it also includes s, p, d, and f-type contributions in nontrivial bonding regions, which is noteworthy because the assumed “vacuum
amounts from the U atoms, indicating significant hybridization of the phase” electronic structure of U is [Rn][5f3][6d1][7s2]. Second, the
bonding U orbitals [35]. The same pattern is observed for UF3, U2F9, general trend is for less f character with decreasing U valence which is
and UF5, in which the F atoms' contribution to the edos in the bonding understood in reference to Fig. 2, showing that those f electrons transfer
region is predominantly p-type, but with significant s, p, d, and f-type to the top of the valence band in lower oxidation state U structures.
contributions from the U atoms, as enumerated in Table 2. Third, between UF3 and UF4 and between U2F9 and UF5, the increase in
This hybridization of U orbitals in the bonding region is shown in f character was most apparent. We note that the coordination en-
Fig. 3. Each structure shows nontrivial contributions from s, p, d, and f- vironment of UF4 and UF5 (square antiprismatic) is more isotropic and
type orbitals. We speculate that the large number of angular mo- more compact than the UF3 and U2F9 monocapped square antiprismatic
mentum components required to describe the electron density for the U coordination environment. In particular, the unidirectional anisotoropy
atoms is a consequence of the relatively complex coordination en- imposed by a singular capping ligand in that geometry favors p bonding
vironment of these UFx compounds. In the inset, we plot the fractional

197
A. Miskowiec, et al. Physica B: Condensed Matter 570 (2019) 194–205

Fig. 3. Orbital-decomposed and summed over spin channels


edos for U atoms in (a) UF3, (b) UF4, (c) U2F9, and (d) UF5 in
the bonding region, as plotted in stacked areas. Significant
hybridization between s, p, d, and f characters describes the
electron density for all structures. In the inset, the fraction of
the bonding region represented by s, p, d, or f character is
shown for each structure. The orbital-decomposed densities
of state are calculated by reprojecting the ground-state
electron density onto the set of spherical harmonics.

over d and f character. 2U(V) is an interesting question. It generally believed that the former
The orbital character just below the Fermi level for all structures is configuration is more stable, but experimental evidence is not definitive
f-type exclusively and is commensurate with the assumed ground-state [54]. As an aside, Prodan et al. discussed the relative stability of a
electronic structure of U(III), U(IV), or U(V) as 5f3, 5f2, or 5f1, respec- mixed valence CmO2 (with electronic configuration 5f6) configuration
tively [48,49]. The unoccupied conduction bands are the remaining and found it to be energetically unfavorable compared to the ground
magnetic f states, and the band gap is similar between structures (0.82, state by 1.6 eV per formula unit [55].
0.94, and 0.83 eV for UF3, UF4 and UF5, respectively). Strictly, U2F9 is Because the reported crystal structures for UF3, U2F9, and UF5
metallic in our calculations due to the half-filling of the top valence contain only one symmetry position for U, we might also expect to
band. The calculation unit cell contains only one distinct U atom (ar- observe long-range ferromagnetic ordering of differing magnetic mo-
ranged into eight symmetrically related positions), so the half-occupied ment magnitudes, structural distortions on the same length scale due to
orbital is required. A supercell calculation containing 64 U atoms and the additional localized electron, or both. We do not predict delocali-
288 F atoms was attempted and deemed unfeasible. Real U2F9 is re- zation of the valence 5f electrons, despite the application of the
ported to be a black solid powder and is probably not metallic (al- Hubbard Ueff which discourages non-integer orbital occupation. The
though no experiments of its conductivity have been performed) [50]. energy gap between the bonding bands and the top of the valence bands
The probable electronic structure of U2F9 is a linear combination of U containing the unpaired f electrons increases with decreasing U valence
(IV) + U(V) states in a 1:1 ratio or U(IV) + U(VI) in a 3:1 ratio. Again, in a nearly linear way.
no experimental evidence presents itself to definitively rectify these
possibilities, but Eller et al., on the basis of similarities in the spectral
characterization of U2F9 and a mixture of UF4 and UF5, suggested that 4.2. Geometries and phonon modes
U2F9 is not likely described by an unusual oxidation state or metallicity
[i.e., it is probably some combination of U(IV) and U(V) or U(VI) The relaxed geometric parameters for UF3, UF4, U2F9, and UF5 are
centers] [1]. This configuration was also suggested by Zachariasen who presented in Table 3. The x-ray diffraction pattern at 40 °C for UF4 (data
studied the structure and noted the same fact much earlier: “Accord- appearing in Table 3) is shown in the Supplemental Information, as well
ingly, it must be concluded that tetravalent and penta- or hexavalent as the atomic positions via DFT and x-ray Rietveld refinement.
uranium atoms replace one another isomorphously, or that each ur- While GGA functionals tend to overestimate lattice constants in
anium atom resonates between the tetravalent state and higher valence DFT, one of the design goals of the PBEsol functional was to correct this
states.” [3] overestimation [30,34]. Examining the overall comparison to experi-
Likewise, UF5 also contains only one symmetrically distinct U atom, ment does not indicate systematic deviations. The crystal structures of
so the possibility of an electronic structure of the form 1U(IV) + 1U(VI) UF3, U2F9, and UF5 are reproduced well compared to experiment. The
could not be explored. The relative stability of 1U(IV) + 1U(VI) versus calculated cell volumes are within 4%, 2%, and 1% for UF3, U2F9, and
UF5, respectively [2,3,5]. For UF4, experiment indicates a peculiar

198
A. Miskowiec, et al. Physica B: Condensed Matter 570 (2019) 194–205

Table 3
Geometric parameters as calculated from DFT.
System a b c Vol. U–F bond distances
Å Å Å Å3 Å

UF3 7.054 7.054 7.2822 313.82 (γ ≡120 °C) 2.38–3.03


UF3 (Exp.) [2] 7.179 7.179 7.345 327.83 (γ ≡120°) 2.42–3.09
UF4 12.779 10.786 8.368 924.92 (β = 126.68) 2.23–2.33
UF4(CuKα, 313 K) 12.794(1) 10.785(1) 8.368(3) 931.31 (β = 126.24(2)°) 2.21–2.42
U2F9 8.512 8.512 8.512 616.63 2.25–2.37
U2F9 (Exp.) [3] 8.454(5) 8.454(5) 8.454(5) 604.32 2.15–2.49
UF5 11.519 11.519 5.203 690.35 2.04–2.30
UF5 (Exp.) [5] 11.469(0) 11.469(0) 5.215(0) 685.97 1.98–2.33

extreme anisotropy in the thermal expansion coefficients along a, b, and


c. With increasing temperature, the lattice contracts along a and c and
expands significantly along b (average thermal expansion coefficient
35.7 × 10−6/K along b, − 28.4 and −5.5 × 10−6/K along a and c).
The unit cell volume in this temperature range (313–463 K) is in-
creasing, in agreement with previous measurements (it is interesting to
note that the volume expansion coefficient is negative below 300 K in
UF4) [4]. Previous neutron diffraction measurements indicated an in-
crease in the lattice volume from 931.74 Å3 at 300 K to 931.82 Å3 at
420 K. The present measurements indicate a lattice volume of
930.69 Å3 at 313 K and 930.80 Å3 at 423 K. Although the measured
volume here is only 1 Å3 smaller at 313 K, the thermal expansion
coefficient is larger than previously measured (9.8168 × 10−4 Å3/K
here vs. 6.67 × 10−4 Å−3/K using neutron diffraction) [4].
Fig. 5 shows the distribution of U–F bond distances for each struc-
ture as calculated with DFT. The number of unique bond distances is
fewer for U2F9 and UF5 than UF4, which is representative of the F po-
sitions being in higher-symmetry Wyckoff positions than in UF4. From
UF3 to UF5, there is a contraction of the average bond distance. Phy-
sically, this pattern represents the conversion of the U-centered poly-
hedral pointgroup from C4ν (monocapped square antiprismatic, CN = 9
in UF3) to D4d (square antiprismatic, CN = 8 in UF4), back to C4ν for
U2F9 (with a commensurate bond length increase, and finally back to
D4d in UF5). (The pointgroup assignments are approximate as each
structure displays some distortion from the ideal symmetry). The
shorter bond distances in UF5 comes at the price of U–F coordination.
While in UF4 and U2F9 all fluoride ligands coordinate with two U atoms,

Fig. 4. Temperature dependence of the lattice constants [(a)] and structural


parameters (β, panel (b), and unit cell volume, panel (c)) of UF4, showing
significant anisotropy in the coefficient of thermal expansion.

temperature dependence of the lattice expansion/contraction [4].


Nevertheless, the lattice volume via DFT (924.92 Å3) is within 1% of
experiment (931.31 Å3 measured by us and 931.82 Å3 measured by
Kern et al. [Table 3]) [4].
Our own experimental results agree with those measured by Kern Fig. 5. Histogram of equilibrium (zero-force) U–F bond distances for UF3
et al., who used neutron diffraction, with regard to the lattice constants (blue), UF4 (red), U2F9 (green), and UF5 (purple) as calculated with the PBEsol
and atomic positions. Our analysis of the temperature dependence of functional. Dashed vertical lines represent the average bond distance for each
structural features in UF4 is shown in Fig. 4. Of particular note is the structure. (For interpretation of the references to colour in this figure legend,
the reader is referred to the Web version of this article.)

199
A. Miskowiec, et al. Physica B: Condensed Matter 570 (2019) 194–205

Table 4 reported IR features for crystalline UF4 are two broad absorption bands
Most intense IR-active modes in UFx. Experimental data is given in bands between 42.5 and 50 meV and 19 and 32 meV [58–60]. The most de-
(s = strong; ms = medium strong; w = weak; br = broad; sh = shoulder). finitive work regarding the IR spectrum of UF4 was performed by
System Frequency Representation Activity Goldstein et al., who showed that these absorption bands in fact showed
meV 10−3e−2eV−1 Å3 significant structure and were not broad and featureless, as previously
reported [11]. Our results indicate a large number of modes are IR
UF3 19.8 E1 450
active in the regions defined by the two strong bands, as well, although
25.1 A1 396
31.0 E1 337 the predicted frequencies appear to be underestimated compared to
18.8 A1 317 experiment (see Supplemental Information for more detail). GGA
38.0 A1 125 functionals are known to overestimate bond distances and as a result
U2F9 25.0 T2 1298
tend to underestimate vibrational frequencies [34].
45.4 T2 273
17.9 T2 229
In the Raman, two modes are predicted at 69.9 and 70.0 meV. The
41.4 T2 66 next highest frequency Raman-active mode is at 49.0 meV, and the
19.5 T2 49 remaining 39 Raman-active modes are rather equally distributed in
UF4 15.4 Bu 756 frequency space between 0 and 50 meV.
21.9 Au 360
U2F9—The highest-frequency phonon mode in U2F9 is calculated to
47.5 Au 298
9.2 Bu 277 be 45.4 meV, significantly below both UF4 and UF5. This difference can
14.5 Au 268 be understood in the context of the distinct U–F coordination in U2F9
UF4(Exp. [11]) 21.3–29.6 s br (CN = 9, v. CN = 8 in UF4/UF5). Only the triply degenerate T2 re-
42.5–49.9 s br
presentations are IR active, such that the 27 IR-active modes in U2F9
14.8–19.4 ms br
31.0–37.1 w
occur at only seven unique frequencies. Among those modes, only three
UF5 14.6 E 1167 have strong IR intensity (25.0, 45.4, and 17.9 meV, in order of the
14.0 B2 627 magnitude of their IR intensity). Most importantly, calculations of the
46.9 E 576 phonon frequencies for U2F9 show a significant imaginary eigensolution
17.0 B2 487
at − 21.8 meV, suggesting dynamic instability in U2F9.
15.2 E 403
UF5(Exp. [62]) 70.8 s UF5—Eleven optical modes in UF5 are above 47 meV. Only one of
63.5 s those modes (B2, 70.5 meV) has any significant IR intensity, a result
50.6 s sh consistent with previous experiments [61,62]. The strongest IR-active
77.9 s sh modes in order of their intensity in UF5 are located at 14.6, 14.0, 46.9,
17.0, and 15.2 meV. Jacob and Asprey, separately, measured the IR
spectrum of β-UF5 and reported strong IR intensity of powder samples
eight of 20 F atoms in UF5 coordinate with only one U atom. With re-
at 77.8, 70.9, 63.5, 48.4, and 50.6 meV [17,62]. The authors report
gard to lattice dynamics, we would predict higher frequency phonon
difficulty separating β and α phase UF5 and probably only the 70.1 and
modes in UF5 and lower frequency modes in U2F9 based on the bond
48.4 meV modes were exclusively β-UF5; the others coming from α-UF5.
distances alone (shorter bond distances corresponding to higher fre-
Our calculation predicts two strong IR modes near these frequencies, as
quencies; i.e., Badger's rule [57]).
well, but also predicts a large number of low-frequency, high intensity
IR modes below 25 meV (200 cm−1, which was the experimental limit
4.3. Infrared modes in the experiments).
In Fig. 6 the pdos for UF3, UF4, U2F9, and UF5 is shown. All struc-
For UF3, UF4, U2F9, and UF5, there are 16, 30, 22, and 24 atoms in tures have similar low-energy pdos owing to the similarity of their
the primitive cells respectively, resulting in 45, 87, 63, and 69 optical fundamental coordination units. Both UF4 and UF5 have four major
modes. Based on the site symmetries and point groups (6mm, 2/m, 4̄3m , regions in their dos: (1) a low energy region between 0 and 38 (40) meV
and 4̄2m ), 20, 42, 51, and 59 optical modes are Raman active and 20, ending in a narrow dark band, (2) a band of finite width between 38
45, 27, and 43 are IR active for UF3, UF4, U2F9, and UF5, respectively. (40) and 51 (49) meV, (3) a relatively dark band beginning at 51 (49)
As an additional check on our calculation, we remain agnostic about the meV and ending at 66 (62) meV, and (4) a series of high energy modes
IR activities of each mode; forbidden optical modes should come from above 66 (62) meV. Although the pdos is finite in region three for both
the calculation itself, and we calculate the optical activities for all IR UF4 and UF5, no phonons vibrate at frequencies within this region at the
active modes. The frequencies of the most intense modes for each zone-center; consequently, this region is dark in the optical vibrational
compound are listed in Table 4, along with their irreducible character spectra.
and IR activities. A full list of phonon modes with IR intensity and The highest energy phonon modes are slightly higher frequency in
Raman activity is given in the Supplemental Information section, as UF5 than in UF4. Noteworthy, however, is that the fundamental char-
well as plots of the phonon band structures. All frequencies discussed acter of the high-energy modes in UF5 is dissimilar from the modes in
below refer to the Γ-centered phonon modes, relevant for optical vi- UF4. In Fig. 7 the phonon displacement amplitudes (magnitude of the
brational spectroscopy. phonon eigenvectors) are averaged over two subsets of F atoms in UF5
UF3—The UF3 optical phonon spectrum is softer than the other UFx as a function of mode frequency. Uranium atoms in both UF4 and UF5
compounds, terminating at 56.3 meV. There are 20 IR active modes, are coordinated with F ligands in square antiprismatic polyhedral co-
with the strongest occurring at 19.8, 25.1, 31.0, and 18.8 meV. No ordination units, but in UF5 a subset (8 of 20 F atoms in the primitive
strong IR intensity occurs above 31 meV and below 18 meV. The 20 cell) are coordinated with only one U atom (singly coordinated), and
Raman-active modes are rather equally distributed in frequency be- the balance are coordinated with two U atoms (doubly coordinated). In
tween 8 and 55 meV, with no significant dark bands predicted. No ex- UF4, all F atoms are doubly coordinated. In Fig. 7, it is clear the highest
perimental vibrational spectra of UF3 are reported. energy modes in region four (above 62 meV) are primarily described by
UF4—Of the 87 calculated optical phonon modes for UF4, five have motion of singly coordinated F atoms. Analogous motion in UF4 does
frequencies above 50 meV. The other 82 modes demonstrate fre- not exist. Five modes in a particular band (singly degenerate modes at
quencies below 48.8 meV, leaving a gap in the fundamental optical 70.5, 73.2, 74.9 meV and a doubly degenerate mode at 72.5 meV) in
spectra. The strongest IR-active modes in order of their intensity are UF5 are described almost exclusively by motion of singly coordinated F
predicted to be at 15.4, 21.9, 47.5, 9.2, and 14.5 meV. The strongest atoms. The relative intensity of the Raman-active 73.2 meV mode

200
A. Miskowiec, et al. Physica B: Condensed Matter 570 (2019) 194–205

Fig. 6. Phonon density of states for UF3 (blue solid line) and
U2F9 (tan dashed line), both with monocapped square anti-
prismatic local U structure, in top panel (a). In panel (b),
phonon density of states for UF4 (blue solid line) and UF5
(dashed tan line), both square antiprismatic U coordination.
Densities of states are calculated with the phonoPy software
package using a 48 × 48 × 48 k-point mesh with a non-
analytical correction for UF3, UF4, UF5 (U2F9 contains an
inversion center) [40]. Both UF3 and U2F9 demonstrate no
high energy (> 60) phonon modes, whereas a significant
number of phonon states are present in UF4 and UF5 above
60 meV. Although the lower-energy (< 50 meV) regions are
similar in UF4 and UF5, owing to the similarity of the U–F
coordination units in each structure, UF5 uniquely shows a
significant number of states between 70 and 80 meV, which
are associated with the motion of “singly-bonded” F atoms.
(For interpretation of the references to colour in this figure
legend, the reader is referred to the Web version of this ar-
ticle.)

between similarly prepared samples could potentially be an indicator of


bond order (i.e., the fraction of singly coordinated fluoride ligands).
In UF4, there are only two Raman-active modes in the high energy
region, both nearly degenerate (69.9 and 70.0 meV). Of the experi-
mental reports of Raman spectra of UF4, all report a Raman band near
78 meV [10,11,58]. Krasser reported a major spectral feature at
78.5 meV. Unfortunately, Villa-Aleman and Wellons' reported spectra
were truncated at 44 meV, but they do report a weak spectral feature at
75.5 meV. It is probable that the observed phonon modes are the cal-
culated Bg and Ag modes at 69.9 and 70.0 meV, respectively, implying
the calculation underestimates the frequency of those modes sig-
nificantly. The Bg and Ag modes represent distortions of the funda-
mental UF 84 − polyhedra, whose distortion is also underestimated in the
calculation, so an error of 5 meV in their frequencies may be plausible.
Only one experimental Raman spectrum of UF5 has been reported,
by Armstrong [8]. The major spectral feature reported by Armstrong is
a single sharp band at 75.6 meV. The question is whether the ob-
servation of this single band, in a sample of unknown chemistry, is
sufficient to identify UF5 from co-located UF4. The calculations for UF5
present three Raman-active modes at 76.7 (doubly degenerate) and
77.6 meV (singly degenerate), strongly suggesting a positive assignment
to the observed 75.6 meV mode to UF5.
No optical spectra of UF3 or U2F9 have been reported, and the re-
sults in Fig. 6 represent the first prediction of their frequencies. The
Fig. 7. Mode displacement amplitudes for UF5 decomposed according to singly highest frequency mode in U2F9 is predicted to be 45.4 meV, lower than
and doubly coordinated F atoms. A set of the highest energy modes between 69 in both UF4 and UF5 and reflective of the longer bond distances in U2F9.
and 74 meV are primarily composed of the motion of singly coordinated F Likewise, the highest frequency mode in UF3 is predicted to be
atoms. Total mode displacement amplitude is normalized according to 56.3 meV. Both structures have a significant number of modes below
∑I ∑i uiI = 1, where u is the phonon eigenvector in Cartesian coordinates 50 meV.
(indexed by i) for each ion, I. Values for a given mode do not add to unity as the
number of U and singly and doubly coordinated F atoms are not identical.
4.4. Instability of U2F9

Although U2F9 is known to be chemically unstable, the degree of its

201
A. Miskowiec, et al. Physica B: Condensed Matter 570 (2019) 194–205

Q-dependence of the scattering, most prominent along 12 meV, are


caused by satisfaction of an inelastic Bragg condition Qp = G ± 2mE ,
where G is the magnitude of a reciprocal lattice vector, and m is the
mass of the neutron. These peaks are only visible at lower energy as
2mE becomes large and Qp spreads in momentum space. At higher Q
(panels [c] and [d] accessing 80 and 125 meV incident energy neutrons,
respectively), the Q-dependence of the measured intensity is smoothly
varying, as expected for a powder sample. In energy space, the scat-
tering is also broad, with no exceptionally strong peaks. A minor peak
at 114 meV was observed that we believe is associated with an instru-
mental effect caused by high-energy neutrons backscattering from the
detectors themselves. Because they are past the radial collimators, these
neutrons are not rejected initially.
Plotted in the bottom panels of Fig. 9 are mesh plots of the calcu-
lated S(Q, ω) as generated by the OClimax software package with 100 K
temperature factor, which calculates a mathematically equivalent
function to Eq. (5) from the phonon band dispersions as calculated with
DFT [19]. A 5 × 5 × 5, Γ-centered k-point mesh generated with pho-
noPy provided the phonon band dispersions, which are suitably aver-
aged over the Brillouin zone. A subtle, but relevant, point for modeling
Fig. 8. U2F9 structure with superimposed phonon displacements (red arrows) Eq. (6) is that the quantity R(t) is a quantum mechanical operator,
for the unstable eigenmode. The directions of fluoride ligand displacements are which has the property that it does not generally commute with other
approximately parallel with the vector joining them to the U atoms, corre- operators. Often, an “incoherent approximation” is made to rectify this
sponding to “breathing” of the UF9 polyhedra. (For interpretation of the re- complication whereby Q is replaced with Q on the argument that 2π is
ferences to colour in this figure legend, the reader is referred to the Web version Q
typically much smaller than the lattice constant and correlations be-
of this article.)
tween atoms due to translational symmetry of the sample are not de-
tectable. It is not necessary to make the incoherent approximation if an
instability is unknown. Zachariasen reported conversion of the black appropriate sampling of the Brillouin zone is performed. Including the
U2F9 crystals to green UF4 rapidly in the presence of air [3]. However, proper phase relationships provides some additional detail in the Q-
Eller et al. claim U2F9 was stable for several months in air [1]. Very few dependence of the predicted S(Q, ω) at smaller values of Q, as seen in
experimental investigations of U2F9 exist, and they are restricted to the Fig. 9(e), but it appears not especially important in this case. Expansion
investigation of its crystal structure [1,3,50,63]. To wit, there are no of the exponential in Eq. (6) yields multiphonon scattering, which is the
reports of a dynamic instability in U2F9. The equilibrium structure of excitation of n phonons from a single scattering event, and is included
U2F9 with the eigenvector of the unstable mode superimposed on the in the total calculation of S(Q, ω). Multiphonon scattering up to order
atomic positions is shown in Fig. 8. Qualitatively, this mode represents 10 is included in Fig. 9(e)–(h). We omit simulation of the elastic line
an expansion/contraction of the fundamental U–F coordination poly- near ω = 0.
hedra. This interpretation is evident due to the collinearity of the lar- On ARCS, the instrumental energy resolution depends on both final
gest atomic displacement vectors with the U–F bond vectors. and incident neutron energy and is well-modeled by a third-order
The density of U2F9 is significantly higher (7.1 g/cm3) than UF4 polynomial function of the final neutron energy for each incident en-
(6.7 g/cm3) or UF5 (6.4 g/cm3). Both observations may suggest a dy- ergy setting [45]. The OClimax software packaged was modified to
namic preference for the conversion of U2F9 to UF4 in addition to the accommodate this shape and theoretical spectra in Fig. 9 have been
chemical preference driven by local instability in the U–F coordination convolved with this model of the ARCS resolution function. Q resolu-
polyhedra. tion was modeled at 0.03 Å−1. Besides instrument parameters, the
Although all previous studies agree that U2F9 is chemically unstable, generation of Fig. 9 requires no experimental inputs and is thus model-
none investigated phase-pure materials. All production of U2F9 was free.
done in the presence of either UF4 or UF5 (either α or β phase). To the The dynamic structure factor can be related to the neutron-weighted
extent that an infinite crystal of U2F9 may be dynamically unstable, generalized dos [gdos, g(ω)], defined in terms of S(Q, ω) as [65]
finite regions could be stabilized by co-location with UF4 or UF5 or by
the presence of minority species of UOx or other contaminants. ℏ2Q 2 g (ω)
S (Q, ω) = ∑ exp(−2Wi ) i [nT (ω) + 1],
i
2 ω (7)
4.5. Inelastic neutron scattering where exp(−2Wi) is the Debye–Waller factor for atom i and nT is the
Bose–Einstein occupation factor. The gi(ω) represent “partial” terms for
In the top panels of Fig. 9 we plot the dynamic structure factor, S(Q, each atom. In Eq. (7), gi(ω) includes all multiphonon scattering. The n-
ω), for UF4 collected at 100 K for four different incident neutron en- phonon gdos, representing n-phonon scattering, is defined recursively
ergies (15, 30, 80, and 125 meV). The measured S(Q, ω) is given by Ref. as [66].
[64]
+∞
1 ∞
gi, n (ω) = ∫−∞ gi, n − 1 (ω) gi,1 (ω − ω′) dω′. (8)
S (Q , ω) =

∫−∞ I αβ (Q , t )exp(−iωt ) dt
(5)
Attempts to connect the measured gdos with the “true,” one-phonon
1 N N dos [gi,1(ω)] are nontrivial [67,68]. In particular, for systems with large
I αβ (Q , t ) = Nα Nβ
∑i α ∑ j β 〈σα σβ exp(iQ⋅Ri (t )) differences in mass or scattering length of the consistent elements, the
× exp(−iQ⋅Rj (0)) 〉, (6) difference between the true and measured gdos can be significant.
However, the pdos calculated in Fig. 6 is exactly the true one-
where α and β index over all isotopic species (of which there are Nα and phonon gdos (if neutron-weighted): appropriately averaged over the
Nβ with neutron cross-sections σα and σβ, respectively). For incident Brillouin zone and in the harmonic approximation by definition of the
neutron energies of 15 and 30 meV (panels [a] and [b]), peaks along the DFPT methodology. The true value of the DFPT calculation for

202
A. Miskowiec, et al. Physica B: Condensed Matter 570 (2019) 194–205

Fig. 9. Comparison of collected S(Q, ω) (panels [a]–[d]) with theoretical values calculated using OClimax (panels [e]–[h]) for 15, 30, 80, and 125 meV incident
energy neutrons for UF4. Data were collected at 100 K.

comparing to experimental neutron spectroscopy data is that the cal-


culation intrinsically represents the one-phonon response whereas ex-
perimental data is unavoidably a sum of multiphonon scattering. The
one-phonon partial gdos is defined as
σi
gi,1 (ω) =
Mi
∑ |ei (j, Q)|2 δ (ω − ω (j, Q)),
j, Q (9)

where e(j, Q) and ω(j, Q) are the phonon eigenvectors and frequencies
and σi and Mi are the cross-section and mass of atom i. To compare Eqs.
(7) and (9), we use a self-consistent procedure for extracting the
phonon density of states from the measured data in Mantid, extended
with the multiphonon package [69–71]. This approach knits together
regions of different energy resolution and allows direct comparison to
the one-phonon pdos as it is calculated from the DFT calculation.
In Fig. 10, we plot g(ω) as reduced from the measured S(Q, ω) (data
points) compared to g(ω) calculated from DFT. Because of the differ-
ence in mass weighted scattering power between U and F, Fig. 10 is
almost entirely scattering from F. The modeled gdos is paradoxically
softer than the measured spectrum in the 40–70 meV region given that
the equilibrium unit cell volume was predicted to be smaller than the
measured unit cell volume. It is possible that this is due to the UF 84 −
polyhedral distortion that is underestimated in the calculation, indu-
cing a softening of the calculated modes.
The measured neutron gdos is generally smooth, with a small main Fig. 10. One-phonon generalized (neutron weighted) density of states as gen-
peak at 30.2 meV. Several shoulders are observed extending up to erated from S(Q, ω) [blue data points, error bars smaller than points, derived
80 meV, much broader than the calculated gdos. This broadness could from a self-consistent extraction based on Eq. (7)] and as calculated from the
be a consequence of the more varied coordination environments in the DFT phonon eigenvectors (red line, calculated from 9) for UF4. Atom-decom-
real sample compared to the ideal calculation cell. posed (U: purple and F: green) generalized densities of states show that the
Previously, in UO2, Pang et al. studied the phonon linewidths using motion is almost exclusively F atoms. (For interpretation of the references to
neutron spectroscopy and DFT + U [20]. They calculated the phonon colour in this figure legend, the reader is referred to the Web version of this
article.)
linewidths using a finite displacement method, reasonably expecting
the additional spatial derivative to potentially cause disagreement with
the measured gdos. But, they noted that the phonon frequencies were, in insufficient to determine certain quantities of interest, such as thermal
fact, the primary culprit for disagreement. Given the quality of our UF4 conduction due to phonon anharmonicities.
sample, we should also conclude that disagreements between the In this case, interpretation of the measured S(Q, ω) is difficult due to
DFT + U method gdos calculations and the measured gdos (Fig. 10) lie the comparably poor energy resolution of the neutron spectrometer, the
primarily with the fundamental inaccuracies in the calculation of the large number of phonon modes, and the significant contribution of
phonon eigenfrequencies. Overall qualitative agreement may be multiphonon scattering to the spectra. Given the good agreement with

203
A. Miskowiec, et al. Physica B: Condensed Matter 570 (2019) 194–205

the model spectrum despite these difficulties, we gain confidence in the frequency modes in UF5 should be higher than in UF4 because com-
qualitative correctness of the underlying calculation. We are not aware mensurate modes do not exist in UF4.
that modeling of the dynamic structure factor itself using electronic Our inelastic neutron spectroscopy measurement of UF4 supports
structure theory has been performed for a system of this complexity the picture that the phonon spectrum of UF4 is centered around 30 meV
(i.e., a highly correlated system with multiple U lattice sites, low and relatively smoothly decreasing with energy, containing a large
symmetry, and significant electronic metastability problems [35]). number of fundamental modes at similar frequencies, and terminating
It would be interesting to understand which phonons are most above 75 meV with only multiphonon scattering contribution to the
strongly affected by the change in unit cell volume in UF4. Using neu- measured gdos above this value. We interpret the smooth phonon
tron diffraction, Kern et al. showed that the temperature dependence of spectrum to be a consequence of the rather low symmetry of the un-
the unit cell volume was due to an increase in order of the fundamental derlying crystal. Our DFT + U calculations, which we used to model the
U–F polyhedron (toward more perfect square antiprismatic orienta- inelastic neutron spectra, qualitatively reproduce the data. Quantitative
tion). Kern et al. also noted that isostructural ThF4, which has a closed agreement with experiment may be insufficient to accurately calculate
electron shell, also shows this behavior, so we expect the lattice dy- further derived quantities using DFT + U alone, emphasizing the need
namics to play a driving role in the temperature-dependence of the unit for improved approaches to calculating lattice dynamics in complex
cell volume for UF4 [4]. Because UF4 is a difficult sample to study with actinides.
Raman spectroscopy, a measurement of the temperature dependence of Direct calculation of the dynamic structure factor from electronic
the IR spectrum could be valuable to understanding this behavior. structure calculations for UF4 represents, to our knowledge, the first
Just as in Fig. 8, a certain subset of modes in UF4 represents the application of this method for a system of this complexity, and de-
displacement of F atoms preferentially along the direction of the vector monstrates the applicability of this technique for establishing a self-
joining them to their U neighbors, and we might expect those modes to consistent experimental/computational physical picture of the lattice
be more strongly affected by changes in the local U–F structure. A more dynamics.
comprehensive study of the lattice dynamics using DFT or ab initio
molecular dynamics may also shed light on this question. Future work Declaration of interest
could explore the possibility of lattice dynamics studied via inelastic x-
ray scattering. Whereas Fig. 10 is primarily reflective of F motion, the Declaration of interest: none.
high-Z U atoms would be more apparent under x-rays, providing
complementary information with distinct contrast. In UF4, several low- Acknowledgments
energy magnetic states have been identified [72]. X-ray spectroscopy
could allow a detailed study of the coupling between magnetic states A portion of this research used resources at the Spallation Neutron
and lattice dynamical properties. Finally, the uranium fluorides studied Source, a Department of Energy, Office of Science User Facility oper-
here represent a class of compounds whose local structure is, in some ated by the Oak Ridge National Laboratory. This research also used
sense, higher symmetry than its unit cell symmetry. If local interactions resources of the Compute and Data Environment for Science (CADES) at
between the U metal center and its ligands are responsible for the dy- Oak Ridge National Laboratory, which is supported by the U.S.
namical properties (instead of longer wavelength phonon interactions) Department of Energy, Office of Science under Contract No. DE-AC05-
of these structures, interesting lattice dynamics in other high-local- 00OR22725. A portion of this work was performed by a postdoctoral
symmetry structures could be observable. fellow (A. E. S.), who is supported by the Department of Homeland
Security. The authors would like to thank Drs. Heath Huckabay and
5. Conclusion Douglas Duckworth for a critical reading of the manuscript.

We have calculated the electronic, phonon, and neutron-weighted Appendix A. Supplementary data
generalized density of states for UF3, UF4, U2F9, and UF5. Although the
bulk crystal structures are not related among these compounds, the Supplementary data to this article can be found online at https://
local U–F coordination structure is similar. This similarity is reflected in doi.org/10.1016/j.physb.2019.06.049.
the edos, which shows features including a bonding region consisting of
F p character with significant contributions from U s, p, d, and f orbitals, References
and a valence region containing 1–3 f electrons (commensurate with
the expected stoichiometry). The low energy pdos (< 50 meV) is very [1] P.G. Eller, A.C. Larson, J.R. Peterson, D.D. Ensor, J.P. Young, Inorg. Chim. Acta 37
similar in all structures, further representative of the similar local co- (1979) 129.
[2] J. Laveissiere, Bull. Soc. Fr. Mineral. Cristallogr. 90 (1967) 304.
ordination. [3] W.H. Zachariasen, J. Chem. Phys. 16 (1948) 425.
We predict U2F9 to be dynamically unstable due to the instability of [4] S. Kern, J. Hayward, S. Roberts, J.W. Richardson, F.J. Rotella, L. Soderholm,
the UF9 polyhedron. Observation of U2F9 coexistence with UF5 could be B. Cort, M. Tinkle, M. West, D. Hoisington, G.H. Lander, J. Chem. Phys. 101 (1994)
9333.
understood by the stabilization of U2F9 in the crystal matrix with other [5] J.C. Taylor, A.B. Waugh, J. Solid State Chem. 35 (1980) 137.
species [50]. [6] R.E. Thoma, H. Insley, G.D. Brunton, J. Inorg. Nucl. Chem. 36 (1974) 1095.
No IR spectra of UF5 have been reported below 25 meV (200 cm−1). [7] W.H. Zachariasen, Acta Crystallogr. 2 (1949) 296.
[8] D. Armstrong, R.J. Jarabek, W.H. Fletcher, Appl. Spectrosc. 43 (1989) 461.
We predict that UF5 should have a set of strong IR-active modes near
[9] J.C. Taylor, P.W. Wilson, J. Solid State Chem. 14 (1975) 378.
14 meV. We also predict several IR-active modes in UF4 at similar fre- [10] E. Villa-Aleman, M.S. Wellons, J. Raman Spectrosc. 47 (2016) 865.
quencies, but their activities are weak (which has been corroborated by [11] M. Goldstein, R.J. Hughes, W.D. Unsworth, Spectrochim. Acta 31A (1975) 621.
[12] Y.Q. Cheng, L.L. Daemen, A.I. Kolesnikov, A.J. Ramirez-Cuesta, J. Chem. Theory
experiment [11]). This prediction could be one method for separating
Comput. 15 (2019) 1974.
the presence of UF5 and UF4 in the context of other evidence. [13] G.W. Halstead, P.G. Eller, L.B. Asprey, K.V. Salazar, Inorg. Chem. 17 (1978) 2967.
In the Raman, the high-energy optical modes are at higher fre- [14] G.W. Halstead, P.G. Eller, M.P. Eastman, Inorg. Chem. 18 (1979) 2867.
quency in UF5 than in UF4. Although we cannot calculate the Raman [15] T.A. O'Donnell, R. Rietz, S. Yeh, J. Fluorine Chem. 23 (1983) 97.
[16] R.T. Paine, R.R. Ryan, L.B. Asprey, Inorg. Chem. 14 (1974) 1113.
activity of these modes, and we expect some uncertainty regarding the [17] L.B. Asprey, R.T. Paine, J. Chem. Soc., Chem. Commun. (1973) 920.
phonon frequencies calculated via DFT + U due to the complexity of [18] K. Momma, F. Izumi, J. Appl. Crystallogr. 44 (2011) 1272.
the systems investigated, the atomic displacement decomposition of the [19] A.J. Ramirez-Cuesta, Comput. Phys. Commun. 157 (2004) 226.
[20] J.W.L. Pang, W.J.L. Buyers, A. Chernatynskiy, M.D. Lumsden, B.C. Larson,
phonon eigenmodes shows that the higher frequencies are related to S.R. Phillpot, Phys. Rev. Lett. 110 (2013) 157401.
motion of the singly coordinated F atoms in UF5. Therefore, the highest

204
A. Miskowiec, et al. Physica B: Condensed Matter 570 (2019) 194–205

[21] J.W.L. Pang, A. Chernatynskiy, B.C. Larson, W.J.L. Buyers, D.L. Abernathy, Sci. Technol. A 35 (2017) 03E108.
K.J. McClellan, S.R. Phillpot, Phys. Rev. B 89 (2014) 115132. [48] H.G. Hecht, J.G. Malm, W.T. Carnall, J. Less Common. Met. 115 (1986) 79.
[22] M.E. Manley, J.R. Jeffries, A.H. Said, C.A. Marianetti, H. Cynn, B.M. Leu, M.A. Wall, [49] E. Barnes, P. Murray, J. Am. Ceram. Soc. 41 (1958) 246.
Phys. Rev. B 85 (2012) 132301. [50] C.J. Howard, J.C. Taylor, A.B. Waugh, J. Solid State Chem. 45 (1982) 396.
[23] A. Togo, L. Chaput, I. Tanaka, Phys. Rev. B 91 (2015) 094306. [51] M. Berger, M.J. Sienko, Inorg. Chem. 6 (1967).
[24] G. Kresse, J. Furthmŭller, Phys. Rev. B 54 (1996) 11169. [52] N. Elliott, Phys. Rev. 76 (1949) 431.
[25] G. Kresse, J. Hafner, Phys. Rev. B 47 (1993) 558. [53] V.N. Ikorskii, V.K. Goncharuk, P.S. Nikitin, S.A. Polishchuk, S.P. Gabuda, Zh. Strukt.
[26] P.E. Blöchl, Phys. Rev. B 50 (1994) 17953. Khim. 27 (1986) 64.
[27] G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758. [54] L.D. Trowbridge, H.L. Richards, Surf. Interface Anal. 4 (1982) 89.
[28] H.J. Monkhorst, J.D. Pack, Phys. Rev. B 13 (1976) 5188. [55] I.D. Prodan, G.E. Scuseria, R.L. Martin, Phys. Rev. B 76 (2007) 033101.
[29] M. Methfessel, A.T. Paxton, Phys. Rev. B 40 (1989) 3616. [57] R.M. Badger, J. Chem. Phys. 2 (1934) 128.
[30] J.P. Perdew, A. Ruzsinszky, G.I. Csonka, O.A. Vydrov, G.E. Scuseria, [58] W. Krasser, H.W. Nürnberg, Spectrochim. Acta 26A (1970) 1059.
L.A. Constantin, X. Zhou, K. Burke, Phys. Rev. Lett. 100 (2008) 136406. [59] T. Soga, K. Ohwada, M. Iwasaki, Appl. Spectrosc. 26 (1972) 482.
[31] S.L. Dudarev, G.A. Botton, S.Y. Savrasov, C.J. Humphreys, A.P. Sutton, Phys. Rev. B [60] K. Ohwada, J. Inorg. Nucl. Chem. 34 (1972) 363.
57 (1998) 1505. [61] E.J. Benham, Actinide Chemistry: High Oxidation-State Uranium Fluoride and
[32] G. Beridze, P.M. Kowalski, J. Phys. Chem. A 118 (2014) 11797. Fluoride Halide Complexes, Thesis (1992).
[33] M. Cococcioni, S. de Gironcoli, Phys. Rev. B 71 (2005) 035105. [62] V.E. Jacob, Z. Anorg. Allg. Chem. 400 (1973) 45.
[34] Z. Wu, R.E. Cohen, Phys. Rev. B 73 (2006) 235116. [63] J.C. Taylor, Coord. Chem. Rev. 20 (1976) 197.
[35] A. Miskowiec, Phys. Chem. Chem. Phys. 20 (2018) 10384. [64] M. Bèe, Quasielastic Neutron Scattering, Adam Hilger, Bristol, England, 1988.
[36] J.P. Allen, G.W. Watson, Phys. Chem. Chem. Phys. 16 (2014) 21016–21031. [65] G.L. Squires, Introduction to the Theory of Thermal Neutron Scattering, Third Ed.,
[37] B. Meredig, A. Thompson, H.A. Hansen, C. Wolverton, A. van de Walle, Phys. Rev. B Cambridge University Press, Cambridge, 2009.
82 (2010) 195128. [66] B. Fultz, Prog. Mater. Sci. 55 (2010) 247.
[38] S. Baroni, S. de Gironcoli, A. Dal Corso, Rev. Mod. Phys. 73 (2001) 515. [67] S.N. Taraskin, S.R. Elliott, Phys. Rev. B 55 (1997) 117.
[39] P. Umari, A. Pasquarello, Diam. Relat. Mater. 14 (2005) 1255. [68] J. Dawidowski, F.J. Bermejo, J.R. Granada, Phys. Rev. B 58 (1998) 706.
[40] A. Togo, I. Tanaka, Scripta Mater. 108 (2015) 1. [69] O. Arnold, J.C. Bilheux, J.M. Borreguero, A. Buts, S.I. Campbell, L. Chapon,
[41] J.R. Geichman, E.A. Smith, S.S. Trond, P.R. Ogle, Inorg. Chem. 1 (1962) 661. M. Doucet, N. Draper, R. Ferraz Leal, M.A. Gigg, V.E. Lynch, A. Markvardsen,
[42] J.R. Geichman, E.A. Smith, P.R. Ogle, Inorg. Chem. 2 (1963) 1012. D.J. Mikkelson, R.L. Mikkelson, R. Miller, K. Palmen, P. Parker, G. Passos,
[43] W. Davies, W. Gray, Talanta 11 (1964) 1203. T.G. Perring, P.F. Peterson, S. Ren, M.A. Reuter, A.T. Savici, J.W. Taylor,
[44] B.H. Toby, R.B. Von Dreele, J. Appl. Crystallogr. 46 (2013) 544. R.J. Taylor, R. Tolchenov, W. Zhou, J. Zikovsky, Nucl. Instrum. Methods A 764
[45] D.L. Abernathy, M.B. Stone, M.J. Loguillo, M.S. Lucas, O. Delaire, X. Tang, J.Y. Lin, (2014) 156.
B. Fultz, Rev. Sci. Instrum. 83 (2012) 015114. [70] M. Kresch, O. Delaire, R. Stevens, J.Y.Y. Lin, B. Fultz, Phys. Rev. B 75 (2007)
[46] Z. Jing-Rong, S. Lang, Y. Chun-Rong, R. Yi-Ming, Acta Phys. - Chim. Sin. 8 (2015) 104301.
240027. [71] J.Y.Y. Lin, F. Islam, M. Kresch, J. Open Source Software 3 (2018) 440.
[47] J.G. Tobin, A.M. Duffin, S.W. Yu, R. Qiao, W.L. Yang, C.H. Booth, D.K. Shuh, J. Vac. [72] Z. Gajek, J. Mulak, J.C. Krupa, J. Solid State Chem. 107 (1993) 413.

205

You might also like