betaFeedback

WorksPhysics and Astronomy

Capturing the ground state of uranium dioxide from first principles: crystal distortion, magnetic structure, and phonons

Abstract

Uranium dioxide (UO2_2) remains a formidable challenge for first-principles approaches, due to the complex interplay among spin-orbit coupling, Mott physics, magnetic ordering, and crystal distortions. Here we use DFT+UU to explore UO2_2 at zero temperature, incorporating all the aforementioned phenomena. The technical challenge is to navigate the many metastable electronic states produced by DFT+UU, which is acomplished using ff-orbital occupation matrix control to search for the ground state. We restrict our search to the high-symmetry ferromagnetic phase, including spin-orbit coupling, which produces a previously unreported occupation matrix. This newfound occupation matrix is then used as an initialization to explore the broken symmetry phases. We find the oxygen cage distortion of the 3k antiferromagnetic state to be in excellent agreement with experiments, and both the spin-orbit coupling and the Hubbard UU are critical ingredients. We demonstrate that only select phonon modes have a strong dependence on the Hubbard UU, whereas magnetic ordering has only a small influence overall. We perform measurements of the phonon dispersion curves using inelastic neutron scattering, and our calculations show good agreement when using reasonable values of UU. The quantitative success of DFT+UU warrants exploration of thermal transport and other observables within this level of theory.

Introduction

Uranium dioxide (UO2\mathrm{UO}_2) has attracted a great deal of research interest ever since the 1950s, given its use as a standard nuclear fuel. The partially filled ff shell in uranium sets the stage for rich physics, and extensive experiments have characterized the behavior of UO2\mathrm{UO}_2. At ambient temperature and pressure, UO2\mathrm{UO}_2 crystallizes in the flourite structure with a lattice parameter of 5.47 Å [1], and exhibits paramagnetism [2, 3]. Upon cooling, UO2\mathrm{UO}_2 undergoes an antiferromagnetic (AFM) transition at TN=30.8T_N = 30.8 K [4], and there is a concomitant oxygen cage distortion whereby the oxygen ions move 0.014 Å from their fluorite positions along the ⟨111⟩\langle111\rangle directions [5] (see Figure 1 of Ref [6] for a schematic, in addition to further details). More specifically, the antiferromagnetism is a noncollinear 3k ordering [7, 8, 9, 10], which is aligned with the oxygen cage distortion.

Extensive first-principles calculations were also performed on UO2\mathrm{UO}_2, and here we focus on ground state properties. It is well-known that UO2\mathrm{UO}_2 is incorrectly predicted to be metallic when using density functional theory (DFT) within conventional approximations to the exchange-correlation energy, including the local density approximation (LDA) and generalized gradient approximation (GGA), due to their failure to describe the strong correlation of the 5f5f electrons of uranium. Going beyond conventional DFT, various approaches have been developed to better capture the effects of strong local interactions; these approaches include hybrid functionals [11, 12, 13], self-interaction correction [14] [15], DFT+UU [16, 17], and DFT plus dynamic mean-field theory (DMFT) [18, 19]. Although all these approaches predict UO2\mathrm{UO}_2 to be an insulator, DFT+UU is the most widely used, thanks to its simplicity and relatively low computational cost. Furthermore, DFT+UU shares a degree of overlap with all the aforementioned techniques. DFT+UU is recovered from DFT+DMFT when solving the DMFT impurity problem within Hartree-Fock [19], DFT+UU can be viewed as a local hybrid functional [20], and the DFT+UU functional is general enough to account for self-interaction corrections. Given that DFT+UU relies on a Hartree-Fock-like approximation, it must break symmetry and strongly polarize the system in order to offer a reasonable description of energetics when dealing with strong interactions. Therefore, DFT+UU brings challenges to modeling UO2\mathrm{UO}_2, given the large number of possible spin and orbital orderings in the case of 5ff electrons. Indeed, many metastable states are found when performing DFT+UU on UO2\mathrm{UO}_2, and some calculated properties (e.g., the band gap) are highly dependent on the exact ordered state [21].

Multiple methods have been designed to search for the ground state in DFT+UU, including occupation matrix control (OMC) [21–24], the quasi-annealing scheme [25], and the UU-ramping method [26], all of which have been applied to UO2\mathrm{UO}_2. None of these schemes can guarantee that the ground state will be found, and different research efforts using these different methods have, in practice, generated different results. For example, both Thompson et al. [27] and Wang et al. [28] used the UU-ramping method and found 1k AFM ground states with different occupation matrices. Earlier work by Laskowski et al. [29] and Gryaznov et al. [30], which did not elaborate on how the ground state search was performed, found 3k AFM ground states with significantly larger oxygen cage distortions (0.16 and 0.09 Å, respectively). Dorado et al. [6] and Thompson et al. [28] employed OMC and the UU-ramping method, respectively, and both found a lowest-energy structure with 1k AFM ordering. The only study that found a 3k AFM ground state featuring oxygen cage distortion on the same scale as the experiments was Zhou et al. [24] (i.e., 0.024 Å), who used an OMC technique based on inputs from crystal field calculations, along with a different type DFT+UU functional.

Phonons may serve as a delicate probe of the electronic structure, and are a key ingredient in thermal transport. As such, a large number of experimental studies [31–34] and DFT-based computational studies [34–40] have been performed on UO2_2. While different first-principles methods have successfully captured various aspects of the phonon dispersion curves and elastic constants, large discrepancies still exist. For example, the predicted phonon frequency of the highest longitudinal optical mode, denoted LO2 [34], at the LL point (0.5, 0.5, 0.5) varies from 50 [38] to 70 meV [39]. More details on the discrepancies of previous DFT+UU calculations of phonons as compared with experiments are reviewed in Supplementary Material (SM)[41]. These discrepancies might be attributable to various differences in the calculations, such as the value of UU, the magnetic ordering, and whether or not spin-orbit coupling (SOC) was included. Furthermore, even for calculations performed under the exact same theory, there is no guarantee that the different researchers converged upon the same ground state, which potentially yields different phonons. Indeed, very few previous studies report details on their obtained UO2_2 ground state (e.g., the uranium ff-electron occupation matrices), making it extremely challenging to reproduce prior work and understand the origin of discrepancies.

In this DFT+UU study of UO2_2, we perform a ground state search using OMC within the ferromagnetic (FM) state, both with and without SOC, yielding a previously unreported occupation matrix. Our new ground state is then used as the starting point for exploring states with further broken symmetries, yielding unprecedented agreement with experiments. Given this new understanding, we then explore the elastic constants and phonon dispersion curves, comparing the phonons with new inelastic neutron scattering data and untangling the various roles of UU, magnetic ordering, and SOC.

Computational Details

Our DFT calculations were carried out using the projector augmented-wave (PAW) method [42, 43], as implemented in the Vienna ab initio Simulation Package (VASP) code [44, 45]. Two exchange correlation functionals were employed: LDA and GGA as formulated by Perdew, Burke, and Ernzerhof (PBE) [46]. A plane-wave cutoff energy of 550 eV was used, and the energy convergence criterion was 10−810^{-8} eV. The Hubbard correction was included in the LDA+UU or GGA+UU approximation to account for the strong local interactions of the uranium ff electrons. Specifically, we used the simplified rotationally invariant DFT+UU approach from Dudarev et al. [47], which only employs a single effective interaction. We customized the VASP code to monitor and impose the occupation matrices during the calculation [41]. The calculated states were controlled via two methods. First, specific occupation matrices were imposed during the first several electronic steps, and then the calculation was allowed to proceed self-consistently without constraint (i.e., OMC). Second, a specific charge density was used to initialize a calculation—a process we refer to as charge density initialization. Symmetry was turned off for all calculations.

Above TNT_N, the space group of UO2_2 is Fm3ˉmFm\bar{3}m and the point group symmetry of the uranium site is OhO_h. For FM calculations, the primitive cell of UO2_2 was used and a 13×13×1313 \times13 \times13 Monkhorst-Pack kk-point mesh [48] was applied; for AFM calculations, the conventional cubic cell was used, and a 7×7×77 \times7 \times7 Monkhorst-Pack kk-point mesh was applied. The phonons and elastic constants were calculated via the lone irreducible derivative (LID) approach [49]. LID uses central finite difference to individually compute each group theoretically irreducible derivative in the smallest possible supercell, minimizing the possibility of numerical inadequacies. Finite difference displacement amplitudes of 0.01, 0.02, 0.03, and 0.04 Å were used to construct quadratic error tails, ensuring that the discretization error is properly accounted for. The face-centered cubic lattice vectors are encoded in a 3×33 \times3 row stacked matrix a^=a2(J^−I^)\hat{\mathbf{a}} = \frac{a}{2}(\hat{\mathbf{J}}-\hat{\mathbf{I}}), where I^\hat{\mathbf{I}} is the identity matrix and J^\hat{\mathbf{J}} is a matrix in which each element is 1. The Brillouin zone is discretized using a real space supercell S^BZa^\hat{\mathbf{S}}_{BZ}\hat{\mathbf{a}}, where S^BZ\hat{\mathbf{S}}_{BZ} is an invertible matrix of integers which produces superlattice vectors that satisfy the point group [49]. In this study, all irreducible derivatives corresponding to S^BZ=2S^C\hat{\mathbf{S}}_{BZ}=2\hat{\mathbf{S}}_C were obtained, where S^C\hat{\mathbf{S}}_C generates the conventional cubic cell and is defined as S^C=(J^−2I^)\hat{\mathbf{S}}_C=(\hat{\mathbf{J}}-2\hat{\mathbf{I}}). The supercell 2S^C2\hat{\mathbf{S}}_C has a multiplicity of 32 and therefore contains 96 atoms; though the LID approach allows one to extract all of the irreducible derivatives from supercells that have multiplicity 4 (i.e. 12 atoms) or less. In the case of 1k and 3k AFM, the magnetism breaks the symmetry of the Fm3ˉmFm\bar{3}m space group, and therefore we use a primitive unit cell of S^Ca^\hat{\mathbf{S}}_C\hat{\mathbf{a}}, which is commensurate with both 1k and 3k. When computing phonons for 1k and 3k, we discretize the Brillouin zone using S^BZ=2I^\hat{\mathbf{S}}_{BZ}=2\hat{\mathbf{I}}, meaning that all irreducible derivatives consistent with the AFM supercell 2S^C2\hat{\mathbf{S}}_C will be extracted. In order to best compare with experiment, we also unfold the AFM phonon band structure back to the original primitive unit cell a^\hat{\mathbf{a}}, averaging any translational symmetry breaking. The Born effective charges (ZU∗=5.54Z_U^*=5.54 and ZO∗=−2.77Z_O^*=-2.77) of the U and O ions and the dielectric constant (ϵ=5.69\epsilon=5.69) were used to account for LO-TO splitting [50, 51].

Ground State Search for UO2_2

Determining the ground state of UO2_2 within DFT+UU is very complex, and exploring all of the phase space is not tractable. Our approach consists of three different stages, all executed using GGA. For stage 1, we used OMC for a given trial occupation matrix and performed computations in the FM state without SOC. For stage 2, the charge density from the stage 1 results was used to initialize a FM calculation with SOC. For stage 3, we used OMC with the resulting stage 2 occupation matrix, and performed an AFM calculation with SOC. Stage 1 was executed for a large number of trial occupation matrices (see below for details), while stage 2 was only executed for the low-energy subset of the stage 1 results. Stage 3 was only executed for the lowest-energy state of stage 2. All the calculations evaluated two types of structural relaxations: relaxing only the volume of the structural cubic crystal (denoted as the undistorted crystal), and full structural relaxations that may break the symmetry (denoted as the distorted crystal).

For stage 1, we followed Dorado et al.’s works [6] [21] and used randomly generated diagonal and non-diagonal occupation matrices (221 and 900 matrices, respectively) as trials. The nine lowest energies we found (denoted as S0S_0–S8S_8, with S0S_0 having the lowest energy) consisted of 25 distinct occupation matrices, some of which were related by symmetry. Details on this search process and the selected properties of the 25 states, including occupation matrices, can be found in SM[41]. Our lowest-energy class of states, S0S_0, contains an orbital ordering that induces structural symmetry breaking via a pure shear strain (±0.017\pm0.017 T2gT_{2g} mode), further lowering the energy by 10 meV relative to the relaxed cubic structure. The S1S_1 class is 13 meV higher in energy than the S0S_0 class, and one occupation matrix within S1S_1 is what Dorado et al. found to be the lowest-energy state in their search. Therefore, we have found a lower energy state within our search, though there is no guarantee that even lower-energy states do not exist. However, the shear strain distortion associated with the S0S_0 class of states is not observed in the experiments. This discrepancy is not particularly problematic, given that SOC will favor a different occupation matrix.

Ferromagnetism with SOC

Previous work reported that SOC may have important impacts on phonons [52]. However, the role of SOC in influencing the ground state occupation matrix of UO2_2 has not been carefully explored. Therefore, we now proceed to our stage 2, in which the low-energy subset of results from stage 1 are used to initialize FM calculations that include SOC. The results of stage 2 are given in Table I, comparing both the relative energy and the distortion. In regard to the energy, the nine lowest-energy classes found without SOC converge into four lowest-energy classes when using SOC, and the relative energies between the classes are significantly changed. For example, the calculation initialized by one state in S2S_2 ended up in S3S_3 with the highest energy, while the calculation initialized by higher-energy state S8S_8 ended up in S0S_0. Furthermore, calculations initialized from degenerate states within the same class sometimes converged to different classes when using SOC: the six degenerate states in S2S_2 converged to S0S_0, S1S_1, and S2S_2. It is clear that SOC significantly perturbs the system. SOC also changes the strain distortions. For example, a T2gT_{2g} strain exists in S0S_0, S3S_3, S6S_6, and S8S_8, but not in any of the reported SOC states. The same trend holds for the oxygen cage distortion (compare Table I to Table S1 in SM [41]).

StateDegeneracyInitialized from StatesEund−Emin⁡E_{\mathrm{und}} - E_{\min} (meV/UO2_2)Edis−Emin⁡E_{\mathrm{dis}} - E_{\min} (meV/UO2_2)Strain Distortion (ϵxx,ϵyy,ϵzz,ϵxy,ϵxz,ϵyz\epsilon_{xx}, \epsilon_{yy}, \epsilon_{zz}, \epsilon_{xy}, \epsilon_{xz}, \epsilon_{yz})×103\times10^{3}Oxygen Cage Distortion
S0S_01S0S_0, S1S_1, S2S_2, S4S_4, S5S_5, S6S_6, S8S_80.30(-1, -1, 2, 0, 0, 0)-
S1S_12S2S_2, S4S_4, S7S_768.967.7(-5, 2, 3, 0, 0, 0)a^{a}-
S2S_24S1S_1, S2S_272.370.8(-4, 5, 0, 0, 0, 0)a^{a}-
S3S_31S3S_372.971.5(-3, -3, 5, 0, 0, 0)-

Table I. States of FM UO2_2 with SOC, as found in stage 2 of our ground state search. The results are sorted by increasing relative energy, with respect to the fully relaxed crystal having occupation matrix S0S_0 and energy Emin⁡E_{\min}.

Stage 2 of our ground state search resulted in a lowest-energy state S0S_0 (see SM [41], Section I.D.1) that was substantially lower than all the others. Interestingly, the S0S_0 state is also found without using OMC (i.e. initializing the calculation in the default manner), which might be reasonable given that it is so much lower in energy than S1S_1. It should be noted that S0S_0 has no appreciable strain, thus there are no immediate inconsistencies with experiments. Although there is no oxygen cage distortion for S0S_0, we will demonstrate that magnetic ordering will generate such distortions.

AFM UO2_2 with SOC

References

  1. [5]J. W. L. Pang, W. J. L. Buyers, A. Chernatynskiy, M. D. Lumsden, B. C. Larson, and S. R. Phillpot, Physical Review Letters 110, 157401 (2013).
  2. [6]M. S. Bryan, J. W. L. Pang, B. C. Larson, A. Chernatynskiy, D. L. Abernathy, K. Gofryk, and M. E. Manley, Physical Review Materials 3, 065405 (2019).
  3. [7]G. Kaur, P. Panigrahi, and M. C. Valsakumar, Modelling and Simulation in Materials Science and Engineering 21, 065014 (2013).
  4. [8]B.-T. Wang, P. Zhang, R. Lizárraga, I. Di Marco, and O. Eriksson, Physical Review B 88, 104107 (2013).
  5. [9]M. Sanati, R. C. Albers, T. Lookman, and A. Saxena, Physical Review B 84, 041116 (2011).
  6. [10]D. L. Abernathy, M. B. Stone, M. J. Loguillo, M. S. Lucas, O. Delaire, X. Tang, J. Y. Y. Lin, and B. Fultz, Review of Scientific Instruments 83, 015114 (2012).
  7. [11]J. W. L. Pang, A. Chernatynskiy, B. C. Larson, W. J. L. Buyers, D. L. Abernathy, K. J. McClellan, and S. R. Phillpot, Physical Review B 89, 115132 (2014).
  8. [12]A. I. Liechtenstein, V. I. Anisimov, and J. Zaanen, Physical Review B 52, R5467 (1995).
  9. [13]S. L. Dudarev, G. A. Botton, S. Y. Savrasov, C. J. Humphreys, and A. P. Sutton, Physical Review B 57, 1505 (1998).
  10. [14]J. Faber, G. H. Lander, and B. R. Cooper, Physical Review Letters 35, 1770 (1975).

Paper details

vlnc:W_ekqkd33a
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents