”Adiabatic” Elastic Constants in Hubbard-Corrected Density-Functional Theory DFT+U: case UO 2

Mahmoud Payami, Samira Sheykhi · 2025

Read ”Adiabatic” Elastic Constants in Hubbard-Corrected Density-Functional Theory DFT+U: case UO 2 by Mahmoud Payami, Samira Sheykhi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-11200.txt.

Introduction

Mahmoud Payami and Samira Sheykhi wrote this 2025 computational physics paper. It flags a subtle error in crystal stiffness calculations. The method at fault is DFT+U. Materials scientists use it where electrons interact strongly. The test material is uranium dioxide.

DFT+U often has several self-consistent electronic states. A strained crystal can fall into a different state than the unstrained one. That switch happens silently. The computed elastic constants then come out wrong. The authors describe an adiabatic procedure against it. Every strained calculation stays on the same electronic branch. The branch is the one of the equilibrium state. Two unit cells of uranium dioxide were tested. The constants agreed closely with each other. They also agreed with experiment.

Catalog record

CREATORMahmoud Payami, Samira Sheykhi
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE14 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-11200

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

”Adiabatic” Elastic Constants in Hubbard-Corrected Density-Functional Theory DFT+U: case UO 2 Mahmoud Payami, Samira Sheykhi Abstract Since in DFT+U there are multiple self-consistent electronic solutions, the so called metastable states, the elastic constants computed from stress-vs-strain will be incorrect if some of the strained configurations fall into a different local electronic minimum than the equilibrium non-strained state. So, it is crucial to carefully take steps to keep the same electronic Hubbard occupation branch when computing the stresses for small strained geometries. In this work, we have explained this ”adiabatic” method of calculation for elastic constants and applied for UO 2 crystal described within two different unit cells of cubic 12-atom and tetragonal 6-atom basis sets. The calculation results for the two different unit cells are the same within 0.1 GPa, and agreement with experiment is excellent. I Introduction In DFT+U dudarev1998 ; coco-degironc2005 ; payami-sheikhi23 ; payami24 , because of the added orbital-dependent terms to DFT functional hohenberg1964 ; kohn1965self , the energy landscape will be multi-valued with respect to occupation matrices as shown schematically in Fig. 1 . Figure 1 : Schematic representation of the multi-valued energy functional with respect to different Hubbard occupation branches. The lowest lying minimum corresponds to the ground state and other local minima correspond to metastable states. The ground state, which is the lowest energy state out of many local minima, is determined via occupation matrix control (OMC) method payami-smcomc23 . In DFT+U, any small changes in ionic positions induced by strain or other dynamics, can flip the system to a different local minimum with a different occupancy pattern. Elastic constants are derivatives of stress with respect to strain, and if the underlying electronic state jumps discontinuously under strain, the calculated derivative is not the derivative in the same branch and will be incorrect. Keeping the occupation matrix fixed or consistently seeded and constrained, ensures that the differentiating is done in the same electronic branch with no discontinuity. In section II , we explain mathematically the calculation method of adiabatic elastic constants. Section III is devoted to calculation details. In section IV the results of application of the method to elastic constants of UO 2 crystal described within two different unit cells of cubic 12-atom and tetragonal 6-atom basis sets are presented; and finally section V concludes the work. II Adiabatic Elastic Constants The adiabatic zero-temperature static elastic constants in general DFT (no Hubbard correction) are the second derivatives of the ground-state energy with respect to strain as: C_{ijkl}=\left.\frac{1}{\Omega_{0}}\frac{\partial^{2}E^{\text{DFT}}_{tot}}{\partial\varepsilon_{ij}\partial\varepsilon_{kl}}\right|_{\varepsilon=0} (1) where \Omega_{0} is the equilibrium cell volume and \varepsilon_{ij} is the symmetric strain tensor. Equivalently, via the Hellmann-Feynman theorem, C_{ijkl}=\left.\frac{\partial\sigma_{ij}}{\partial\varepsilon_{kl}}\right|_{\varepsilon=0}, (2) with \sigma_{ij} the stress tensor. Adding orbital-dependent Hubbard correction to DFT energy, the DFT+U total energy is given by E^{\text{DFT+U}}[n,\{n^{I}\}]=E^{\text{DFT}}[n]+E^{U}[\{n^{I}\}], (3) where, n(\bf r) is the total electron density, n^{I} are the on-site occupation matrices for correlated Hubbard orbitals on atom I , and E^{U} is the Hubbard correction (Dudarev type, Liechtenstein-type, etc.). For example, in Dudarev’s simplified form we have dudarev1998 ; coco-degironc2005 : E^{U}=\sum_{I,\sigma}\frac{U^{I}}{2}{\rm Tr[{\bf n}^{\it I\sigma}({\bf 1}-{\bf n}^{\it I\sigma}})]. (4) Thus the stress tensor acquires an additional Hubbard contribution: \sigma_{ij}^{\text{DFT+U}}=\sigma_{ij}^{\text{DFT}}+\sigma_{ij}^{U}, (5) where \sigma_{ij}^{U}=\frac{1}{\Omega}\frac{\partial E^{U}}{\partial\varepsilon_{ij}}. (6) The Hubbard occupation matrices \{n^{I}\} play important role in the derivatives. They are variational degrees of freedom in DFT+U such that for each geometry consisting of ”lattice+ions”, the self-consistent solution finds a stationary point on both n(\bf r) and \{n^{I}\} . Formally, the elastic constants should be computed from the second derivative of the minimum energy branch: C_{ijkl}=\left.\frac{1}{\Omega_{0}}\frac{\partial^{2}}{\partial\varepsilon_{ij}\partial\varepsilon_{kl}}\left[\min_{\{n^{I}\}}E^{\text{DFT+U}}[n,\{n^{I}\}]\right]\right|_{\varepsilon=0}. (7) In DFT+U, the minimization over \{n^{I}\} has multiple local minima corresponding to metastable occupation matrices. Then, derivatives are well defined only if one stays on the same branch of \{n^{I}\} . Mathematically, one must treat \{n^{I}\} as implicit functions of strain constrained to the chosen minimum branch: n^{I}(\varepsilon)\;\;\;\;\text{with}\;\;\;\left.\frac{\partial E^{\text{DFT+U}}}{\partial n^{I}}\right|_{n^{I}(\varepsilon)}=0. (8) Thus, C_{ijkl}=\frac{1}{\Omega_{0}}\left[\frac{\partial^{2}E^{\text{DFT+U}}}{\partial\varepsilon_{ij}\partial\varepsilon_{kl}}+\sum_{I}\frac{\partial^{2}E^{\text{DFT+U}}}{\partial\varepsilon_{ij}\partial n^{I}}\frac{\partial n^{I}}{\partial\varepsilon_{kl}}\right]_{\varepsilon=0,\;n^{I}=n^{I}_{0}}. (9) Under the condition of ”fixed-occupation branch”, at self-consistency, the first-order derivatives with respect to occupations vanish: \left.\frac{\partial E^{\text{DFT+U}}}{\partial n^{I}}\right|_{n^{I}_{0}}=0. (10) Therefore, by implicit function theorem, the term involving \partial n^{I}/\partial\varepsilon drops out if one stays on the same local minimum branch. So, in practice, one just needs to ensure the occupations do not ”jump”, and then the usual stress-strain or energy-strain formula is valid. To sum up: i ) The occupation matrices act as auxiliary variational parameters; ii ) If one stays on the same minimum branch, their implicit strain-dependence does not contribute to the second derivative (elastic constants); and iii ) If the system hops to another local minimum, the derivative becomes discontinuous, and the computed ”elastic constants” will be unphysical. So, the theoretical formulation for the ”correct workflow” is: C_{ijkl}=\frac{1}{\Omega_{0}}\left.\frac{\partial^{2}E^{\text{DFT+U}}}{\partial\varepsilon_{ij}\partial\varepsilon_{kl}}\right|_{\varepsilon=0,\;\{n^{I}\}=\{n^{I}_{0}\}}, (11) where \{n^{I}_{0}\} are the occupation matrices of the equilibrium ground state. III Calculation Details The uranium dioxide crystal can be described with good accuracy using the cubic Fm\bar{3}m space group (space group number 225) with a lattice constant a=5.47\AA , as shown in the left of Fig. 2 , or by a simple tetragonal unit cell with a 6-atom basis set shown in the right of Fig. 2 . The antiferromagnetic (AFM) structure of uranium atoms is modeled by the simple one-dimensional AFM in z -direction. Figure 2 : UO 2 crystal structure with cubic 12-atom unit cell (left) and with simple tetragonal 6-atom unit cell (right). Large grey and small red balls represent uranium and oxygen atoms, respectively. The sets of lattice vectors \{{\bf a_{1}},{\bf a_{2}},{\bf a_{3}}\} for cubic and \{{\bf a^{\prime}_{1}},{\bf a^{\prime}_{2}},{\bf a^{\prime}_{3}}\} for tetragonal unit cells are shown in Fig. 3 and are defined in Cartesian coordinates as: Figure 3 : Top view of cubic 12-atom and simple tetragonal 6-atom unit cells. Lattice vectors \{{\bf a_{1}},{\bf a_{2}},{\bf a_{3}}\} and \{{\bf a^{\prime}_{1}},{\bf a^{\prime}_{2}},{\bf a^{\prime}_{3}}\} describe the cubic and tetragonal cells, respectively. {\bf a_{1}}=a(1,0,0)\;\;\;{\bf a_{2}}=a(0,1,0)\;\;\;{\bf a_{3}}=a(0,0,1) (12) {\bf a^{\prime}_{1}}=\frac{a}{2}(1,1,0)\;\;\;{\bf a^{\prime}_{2}}=\frac{a}{2}(-1,1,0)\;\;\;{\bf a^{\prime}_{3}}=a(0,0,1) (13) In this work, using DFT+U, the elastic constants of UO 2 is calculated using both cubic and tetragonal unit cells. The results are in excellent agreement with each other and with experiment. DFT+U Calculations For the electron-ion interactions we have used scalar-relativistic ultra-soft pseudo-potentials (USPP) with PBEsol approximation for the XC, generated using ”atomic” code. The valence configurations U( 6s^{2},\,6p^{6},\,7s^{2},\,7p^{0},\,6d^{1},\,5f^{3} ) and O( 2s^{2},\,2p^{4} ) were used in the USPP generation. All DFT+U calculations were based on the solution of the KS equations using the Quantum-ESPRESSO (QE) code package qe-2009 ; qe-2020 . Kinetic energy cutoffs for the plane-wave expansions were chosen as 90 and 720 Ry for the wave-functions and densities, respectively. The smearing method of Marzari-Vanderbilt for the occupations with a width of 0.001 Ry were used. For the unstrained and strained cell Brillouin-zone integrations, 6\times 6\times 6 and 10\times 10\times 7 grids were used for cubic and tetragonal cells, respectively; All geometries were fully optimized for total residual pressures on unit cells to within 0.5 kbar, and residual forces on atoms to within 10 -3 mRy/a.u. For Hubbard orbitals, we have used the ”atomic” projection operators for the expansion of KS orbitals. The value of Hubbard parameter for the 5f orbital of uranium atoms is set to 2.45 eV payami23 which gives the lattice constant of optimized structure equal to the experimental value 5.47Å. Elastic Constants Calculations To calculate the elastic

Read the complete text