WorksPhysics and Astronomy
Accurate calculation of light rare-earth magnetic anisotropy with density functional theory
Abstract
Density functional theory (DFT) has long struggled to treat light rare-earth magnetism. We show that this difficulty arises from an overestimate of the charge asphericity, and thus the magnetic anisotropy energy, due to the inadequacy of single Slater-determinant representations. We propose an effective solution by combining constrained DFT+U with crystal field theory and a systematic many-body correction to the charge asphericity. We confirm the validity of this combination on TbVSn and TbCo, and then show how the many-body correction adjusts the calculated magnetic anisotropy energy of SmCo to match experiment. Our method is an efficient DFT-based approach to address light-rare-earth magnetism.
The highly localized electrons of the rare earths (RE) are essential not only for fundamental science—from topological magnets [1, 2] to heavy fermions [3, 4, 5]—but also for technological applications, from high-performance permanent ferromagnets [6, 7, 8, 9, 10] to quantum transduction and memory [11]. These materials have large spin-orbit coupling (SOC) compared to the small crystal field (CF) scales, which allows for uniquely strong magnetocrystalline anisotropy (MA). The light REs (, –Sm) are particularly desirable in combination with transition metals (TM), as the antiferromagnetic - exchange takes place through the spin, allowing the total moment, to align with the moments, forming the strongest and most anisotropic permanent magnets, notably SmCo and NdFeB.
Despite the broad success of density functional theory (DFT) in many material systems, the open-shell materials remain a persistent challenge [12, 13, 14] due to the importance of strong interactions. Interactions may be better accounted for with other methods, but these are computationally expensive and struggle to capture the relatively small CF energy scales [12, 15, 16]. This challenge hinders the systematic, theory-driven design and discovery of RE-based materials. We have recently addressed one of the fundamental problems: the failure of DFT-based methods to reproduce the correct Hund’s rule ground state [12, 13], where we showed that using constrained DFT to enforce the correct Hund’s rule ground state allows DFT+U to capture the correct MA across a wide range of heavy pristine and doped RE-TM magnets [12, 17, 18, 19, 20, 21, 22]. Light RE materials provide an additional difficulty, as the SOC ground state cannot be represented by a single Slater determinant, while the DFT ground and metastable states will always be single Slater determinants.
In this paper, we address this long-overlooked problem and show that DFT’s single Slater determinant nature systematically overestimates the charge asphericity, leading to large errors in orbital-dependent quantities like MA. Fortunately, we show that this overestimation can be systematically countered with a many-body correction that depends only on the RE ion, not on specific material details, and argue that with this correction, DFT can, in fact, be successfully combined with CF theory to accurately treat magnetism throughout the REs. We propose a systematic method to probe the CF parameters (CFP) using DFT, and validate it in two distinct compounds before applying it to SmCo, where it allows DFT to reproduce the experimentally determined MA energy. We expect this approach to work for all integer-valent RE materials, and the obtained CFPs can be widely used, both for low-temperature properties and in the excited spin-orbit multiplets important for optical transitions in both heavy and light REs.
Multi--electron RE ions can be described in the Russell-Saunders (LS) coupling scheme, where the total and of a given configuration are first maximized and then aligned by SOC to give the total Hund’s rule ground state, where corresponds to . The LS coupling scheme is expected to work well for the REs, while intermediate or - coupling schemes will be more appropriate for the actinides. In the LS scheme, the states are written as the linear combination,
where are the Clebsch-Gordan coefficients. The many-body basis states, are themselves superpositions of single particle states with , but will be represented by a single Slater determinant for the maximally polarized state. can only be represented by a single Slater determinant for , and only when . Current DFT implementations can, at best, select only the largest term of this superposition, as they generally represent valence electrons by spherical harmonics . Here, we compare the angular charge densities obtained from a single (DFT) versus multiple (true) Slater determinant approximation for all the light ions; this anisotropy can be quantified by the three multipole moments, (), which are always overestimated by DFT. Here, are Legendre polynomials that are proportional to the spherical harmonics, but are not normalized [23]. The MA is directly influenced by the charge distribution, which changes its electrostatic interactions with the surrounding material as it is rotated. Therefore, an incorrect charge distribution translates into an incorrect MA energy. The consequences of this information loss manifest clearly in the angular part of the charge density, of the fully polarized state, as shown in Figure 1. For Ce, this approximation is not too bad, as and only 1/7 of the state is dropped. However, for Eu, , it misses out on 6/7 of the state and mistakenly predicts anisotropic magnetic behavior as it effectively admixes non-zero multiplets into the ground state.
Figure 1. DFT overestimates the charge asphericity of the states for light- ions. Polar plots show the true (blue) and DFT (red) -dependent charge density . The multipole moments () are shown for the true (solid bars) and DFT (empty bars) charge densities; these are projections of onto the (unnormalized) Legendre polynomials . The overestimation generally increases with both atomic number and multipole order .
Thus, DFT on its own cannot produce a fully accurate state for the light-REs—even with orbital constraints, the wave-function is distorted, and without the Hund’s rule constraints, it is entirely wrong for any open-shell ion. Nevertheless, we argue that the computed anisotropy energy from DFT can still be used to extract the true magnetic anisotropy by combining DFT with CF theory.
In CF theory, the spherically symmetric ground state determined by interactions and SOC experiences an anisotropic local environment captured by the effective Hamiltonian,
where are Stevens operators written in terms of , and are a set of often phenomenologically defined CF parameters that depend both on the ion and its environment. It is convenient to rewrite , where depends not only on the ion but also on the multiplet given by . The Stevens coefficients are commonly written as , , or for , respectively, and are tabulated for all the REs [6, 24, 25, 26, 27, 28, 29, 30].
CF theory assumes that the parameters are independent of both the ion and the configuration, and that a single set of parameters can capture the experimentally observed behavior over a wide range of temperatures and magnetic fields, during which the occupancies change substantially; the CF parameters are even thought to remain roughly constant throughout isostructural series of RE compounds. If the crystal potential is indeed independent of the configuration, we should be able to probe it with whichever non-ground-state charge distribution we have. While both CF theory and DFT are widely used, it has never before been tested whether the anisotropy of DFT calculations can truly be captured by CF theory, or where this assumption might break down.
For concreteness, we discuss the case of hexagonal symmetry, with point-group symmetry at the RE site, as relevant for . Here, there are generically four parameters: . It is useful to define the MA energy as the angle dependent energy of the fully polarized ferromagnetic ground state, , where is along the quantization axis, , rotated by from the crystallographic -axis.
where , depends on , and the coefficients and angular dependence are determined using the Wigner -matrices. All terms depend on the polar angle and contribute to the uniaxial MA, while only has -dependence (six-fold in-plane MA) [21].
Now we can introduce our DFT-based method to accurately calculate the CFP and MA throughout the REs. The process consists of three steps, detailed below:
Probe the crystal potential within DFT+ by calculating the total energy .
Check that CF theory is self-consistent within DFT for the system.
Obtain CFPs and apply a many-body correction to the DFT-derived anisotropies.
We calculate the total energy using constrained DFT+, ensuring that the occupancy remains in the designated Hund’s rule configuration as the spin is rotated. We then perform a least-squares fit of the calculated to Eq. (3) to extract the anisotropy parameters, and subsequently determine the CFPs, . Note that for DFT+ calculations, sizable values in the range of 6–12 eV are applied to the - shell to polarize the occupied and unoccupied states and remove them from the Fermi level. These large ’s ensure that the states are well-localized and integer valent, with minimal hybridization with the conduction electrons. Additional details of the DFT+ calculations, including the dependence, can be found in the Supplemental Material and in Ref. [12].
For the heavy REs, is a single Slater determinant and the CFPs calculated within constrained DFT can be used directly. For light REs, we assume (and justify below) that the DFT ground state probes the same CFPs, but the obtained ’s must be corrected for the overestimation of the charge asphericity. This many-body correction is done using the ratio of the three charge multipole moments, in the true many-body ground state, , and the DFT states. In a given configuration, the multipoles can be written as . The true many-body ground state is , while only the part of the DFT state is required; must be the appropriate Stevens coefficient, or , respectively [23, 31, 32]. The ratio of the two is
We have tabulated these ratios for the ground state multiplets of all light ions in Table I; these can also be calculated for excited multiplets if needed.
| , , | ||||
| Ce | ||||
| Pr | ||||
| Nd | ||||
| Pm | ||||
| Sm | ||||
| Eu |
Table I. Correction ratios of the true, many-body to DFT single-Slater determinant multipole moments for light ions.
Our procedure relies on the consistency of CF theory within DFT: the idea that the DFT single Slater determinant configuration and the true many-body configuration are experiencing the same CF potential, which is not
guaranteed. This validity can be checked by extracting and comparing the CFP for different meta-stable DFT configurations. While the strength of this consistency may depend on the material, we do expect it to be similar throughout series of RE compounds. Below, we check this for two well-characterized hexagonal compounds.
TbVSn has a crystal structure closely related to that of RCo magnets and exhibits giant easy-axis anisotropy at low temperatures [19, 21, 33]. It provides a clean test case because the transition-metal sublattice (the V kagome layers) is non-magnetic, so all anisotropy originates from the Tb ions. In DFT, Tb () is represented as seven electrons in the majority spin channel and one in the minority spin channel (). We calculate for each choice of the single minority-spin electron , with , thereby probing the CFPs using effectively four distinct charge distributions.
As shown in Figure 2(a), the DFT+ results for at various values are perfectly fit by the CF model using Eq. (3). Since states have identical , we average their calculated curves (see Supplemental Material). Note that the Hund’s-rule ground state, yields the expected giant easy-axis MA, consistent with experiment, while other configurations produce drastically different MA curves.

Figure 2. Rare-earth magnetocrystalline anisotropy and crystal-field parameters calculated using various Tb-4f configurations in TbV6Sn6 and TbCo5. (a) Magnetic energy variation, , in TbV6Sn6 as a function of spin-axis rotation calculated using DFT (dots) for all configurations of the single minority-spin electron () in Tb. Insets show polar plots of the corresponding wavefunctions, with phase denoted by color. Solid lines represent CF model fits using Eq. (3). The true ground state yields the correct giant easy-axis anisotropy. (b) Crystal structure of RV6Sn6. (c, d) Same as (a, b), but for TbCo5. (e, f) Crystal-field parameters (CFPs) extracted from the profiles of the states shown in (a, c), demonstrating consistency across different probes.
While each profile can be used to extract the CFPs, the response of different states is dominated by different CFPs due to their distinct . Most notably, , meaning that the MA for originates solely from higher-order () terms. Indeed, due to orbital-dependent self-interaction errors in DFT, this vanishing erroneously favors as the Tb ground state [18]. Similarly, as is relatively small, the weak in-plane MA in likely leads to errors in estimating , which can be better calculated from .
Now, we turn to RCo, where the Co atoms are also magnetic. As seen in Figure 2(c), TbCo exhibits a much smaller fourth-order term than TbVSn, with the uniaxial MA now dominated by the lowest-order term, . has the opposite sign compared to that in TbVSn, explaining the easy-axis MA of TbVSn [17, [19, 33] and the easy-plane MA of TbCo5 [34, 35]. This sign difference likely originates from structural differences: as shown in Figure 2(b) and (d), in RV6Sn6, a pair of “dumbbell” Sn (2e) atoms effectively replace an R site in RCo5 along the -axis and are located closer to the neighboring R. Overall, as we show later, the calculated CFPs can reproduce the experimental anisotropy [36] very well.
Crucially, despite the very different curves for each , the extracted CFPs are very similar across all cases for a given material. For TbV6Sn6, values extracted from are nearly identical, while those extracted from show slightly larger deviation (within across all cases). We note
that the deviations in TbCo5 are more pronounced than those in TbV6Sn6, which may be attributed to stronger hybridization between and wavefunctions.
This consistency provides a nontrivial confirmation of the validity of CFT within DFT: it demonstrates that the fitted CFPs are not very sensitive to the specific charge distribution used to probe them, with the caveat that cannot be used to probe , as . Even though crystal-field effects in real materials arise from both electrostatic and hybridization contributions, the excellent agreement between our DFT results and the CF model confirms the validity of extracting CFPs via self-consistent DFT total-energy calculations.
Now we use these corrections to calculate the MA energy (MAE) of SmCo5. Figure 3 compares the calculated for SmCo5 using DFT+ ( eV), with and without the many-body correction. Uncorrected DFT+ predicts a uniaxial MAE of meV per formula unit, almost double the experimental value of 13–16 meV/f.u. [36]. After applying our many-body correction, the MAE is significantly reduced, falling within the experimental range. The correction is also essential for understanding the lack of in-plane MA: DFT predicts a sizable in-plane MA, which vanishes completely after applying the many-body correction, as for . Experimentally, this anisotropy can be restored if there is mixing with the excited multiplet. As another example, in NdCo5, DFT is expected to overestimate the in-plane MA by roughly a factor of .

Figure 3. SmCo5 magnetic anisotropy energy, calculated using DFT+ both without (blue dots/line) and with (red line) many-body corrections. SmCo5 experimental MAE is in the range of 13–16 meV/f.u. (green bar) [36]. The insets show the DFT and corrected charge distributions.
We expect this approach to be broadly applicable to calculating RE crystal fields in both insulating and metallic and both magnetically ordered and paramagnetic materials; however, there are some limitations. Integer valence is required, as otherwise the effect of strong hybridization means DFT is no longer probing the material CFP. It is possible that DFT probes “bare” CFPs that can be renormalized by hybridization and interactions, but this requires further investigation. It is not clear how well this approach will work even for integer-valent Ce materials, where there can be substantial - hopping if the Ce-Ce distance is less than the Hill limit. While we have only tested this on ground-state multiplets, the CFPs should not be substantially different for excited multiplets with the same total and , which may allow for the calculation of optical transitions. Actinide materials will provide additional challenges, both due to the more extended wavefunctions leading to stronger hybridization and - hopping, and the different SO coupling schemes required, but it may well be possible to extend this method to some insulating actinide materials.
In summary, we have identified a fundamental issue in single-Slater-determinant DFT calculations for light REs: an exaggerated charge asphericity leads to an overestimation of the MAE. To resolve this issue, we presented a combined DFT and CF theory approach that uses constrained DFT+ to probe the CFPs, which we have shown to be largely independent of the electronic configuration. These DFT parameters are then corrected by to obtain the true CFPs and MAE. With this correction, our DFT+ calculations for SmCo now show close agreement with the experimental MAE. This many-body correction is especially important for heavier light-REs like Sm and Eu, and in cases where higher-order CF terms play a significant role. Higher-order CF terms are important in materials with substantial in-plane MA; those with large high-order contributions to the uniaxial MAE, as in RVSn; and in cubic materials, where the first anisotropy arises from . In cubic Sm materials, the many-body correction reduces the MA to approximately 15% of the DFT value. More broadly, the self-consistent constrained-DFT approach demonstrated here offers a practical and systematic method to extract CFPs and incorporate many-body effects in magnetism, paving the way for more accurate and efficient modeling of RE materials.
Acknowledgments— This work is supported by the U.S. Department of Energy, Office of Basic Energy Sciences, Materials Sciences and Engineering Division. The initial conceptual development by Ke was supported by the Office of Science Early Career Research Program through the Office of Basic Energy Sciences, Materials Sciences and Engineering Division. Ames Laboratory is operated for the U.S. Department of Energy by Iowa State University under Contract No. DE-AC02-07CH11358. We acknowledge Research Computing at the University of Virginia for providing computational resources.
References
References
- [1]J.-X. Yin, W. Ma, T. A. Cochran, X. Xu, S. S. Zhang, H.-J. Tien, N. Shumiya, G. Cheng, Q. Jiang, B. Lian, et al., Quantum-limit Chern topological magnetism in TbMn₆Sn₆, Nature 583, 533 (2020).DOI
- [2]W. Ma, X. Xu, J.-X. Yin, H. Yang, H. Zhou, Z.-J. Cheng, Y. Huang, Z. Qu, F. Wang, M. Z. Hasan, and S. Jia, Rare Earth Engineering in RMn₆Sn₆ (R = Gd–Tm, Lu) Topological Kagome Magnets, Phys. Rev. Lett. 126, 246602 (2021).
- [3]P. Pagliuso, N. Curro, N. Moreno, M. Hundley, J. Thompson, J. Sarrao, and Z. Fisk, Structurally tuned superconductivity in heavy-fermion CeMIn₅ (M=Co, Ir, Rh), Physica B: Condensed Matter 320, 370 (2002), proceedings of the Fifth Latin American Workshop on Magnetism, Magnetic Materials and their Applications.DOI
- [4]Q. Y. Chen, D. F. Xu, X. H. Niu, J. Jiang, R. Peng, H. C. Xu, C. H. P. Wen, Z. F. Ding, K. Huang, L. Shu, Y. J. Zhang, H. Lee, V. N. Strocov, M. Shi, F. Bisti, T. Schmitt, Y. B. Huang, P. Dudin, X. C. Lai, S. Kirchner, H. Q. Yuan, and D. L. Feng, Direct observation of how the heavy-fermion state develops in CeCoIn₅, Phys. Rev. B 96, 045107 (2017).DOI
- [5]S. Kirchner, S. Paschen, Q. Chen, S. Wirth, D. Feng, J. D. Thompson, and Q. Si, Colloquium: Heavy-electron quantum criticality and single-particle spectroscopy, Rev. Mod. Phys. 92, 011002 (2020).DOI
- [6]R. Skomski and J. M. D. Coey, Permanent magnetism (Institute of Physics Publishing, Bristol, 1999).
- [7]K. A. Gschneidner and L. Eyring, eds., Handbook on the Physics and Chemistry of Rare Earths (North Holland, Amsterdam, 1978).
- [8]K. Schüller, McCraig: Permanent magnets in theory and practice, Physikalische Blätter 35, 423 (1979).
- [9]L. H. Lewis and F. Jiménez-Villacorta, Perspectives on permanent magnetic materials for energy conversion and power generation, Metallurgical and Materials Transactions A 44, 20 (2013).DOI
- [10]R. McCallum, L. Lewis, R. Skomski, M. Kramer, and I. Anderson, Practical aspects of modern and future permanent magnets, Annual Review of Materials Research 44, 451 (2014).DOI
- [11]W. Tittel, M. Afzelius, A. Kinos, L. Rippe, and A. Walther, Quantum networks using rare-earth ions, Quantum Science and Technology 10, 033002 (2025).DOI
- [12]Y. Lee, Z. Ning, R. Flint, R. J. McQueeney, I. I. Mazin, and L. Ke, Importance of enforcing hund’s rules in density functional theory calculations of rare earth magnetocrystalline anisotropy, npj Computational Materials 11, 168 (2025).DOI
- [13]F. Zhou and V. Ozoliņš, Obtaining correct orbital ground states in f-electron systems using a nonspherical self-interaction-corrected LDA+U method, Phys. Rev. B 80, 125127 (2009).
- [14]C. E. Patrick, S. Kumar, G. Balakrishnan, R. S. Edwards, M. R. Lees, L. Petit, and J. B. Staunton, Calculating the magnetic anisotropy of rare-earth–transition-metal ferrimagnets, Phys. Rev. Lett. 120, 097202 (2018).DOI
- [15]I. L. M. Locht, Y. O. Kvashnin, D. C. M. Rodrigues, M. Pereiro, A. Bergman, L. Bergqvist, A. I. Lichtenstein, M. I. Katsnelson, A. Delin, A. B. Klautau, B. Johansson, I. Di Marco, and O. Eriksson, Standard model of the rare earths analyzed from the Hubbard I approximation, Phys. Rev. B 94, 085137 (2016).
- [16]P. Delange, S. Biermann, T. Miyake, and L. Pourovskii, Crystal-field splittings in rare-earth-based hard magnets: An ab initio approach, Phys. Rev. B 96, 155132 (2017).DOI
- [17]S. X. M. Riberolles, T. J. Slade, D. L. Abernathy, G. E. Granroth, B. Li, Y. Lee, P. C. Canfield, B. G. Ueland, L. Ke, and R. J. McQueeney, Low-Temperature Competing Magnetic Energy Scales in the Topological Ferrimagnet TbMn₆Sn₆, Phys. Rev. X 12, 021043 (2022).arxiv.org/abs/2110.14713
- [18]Y. Lee, R. Skomski, X. Wang, P. P. Orth, Y. Ren, B. Kang, A. K. Pathak, A. Kutevop, B. N. Harmon, R. J. McQueeney, I. I. Mazin, and L. Ke, Interplay between magnetism and band topology in the kagome magnets RMn₆Sn₆, Phys. Rev. B 108, 045132 (2023).
- [19]E. Rosenberg, J. M. DeStefano, Y. Guo, J. S. Oh, M. Hashimoto, D. Lu, R. J. Birgeneau, Y. Lee, L. Ke, M. Yi, and J.-H. Chu, Uniaxial ferromagnetism in the kagome metal TbV₆Sn₆, Phys. Rev. B 106, 115139 (2022).DOI
- [20]M. Xu, Y. Lee, X. Ke, M.-C. Kang, M. Boswell, S. L. Bud’ko, L. Zhou, L. Ke, M. Li, P. C. Canfield, and W. Xie, Giant Uniaxial Magnetocrystalline Anisotropy in SmCrGe₃, Journal of the American Chemical Society 146, 30294 (2024).
- [21]T. Han, R. D. McKenzie, J. Blawat, T. J. Slade, Y. Lee, D. M. Pajerowski, J. Singleton, B. Li, P. C. Canfield, L. Ke, R. McDonald, R. Flint, and R. J. McQueeney, Proximity to quantum criticality in the Ising ferromagnet TbV₆Sn₆ (2024), arXiv:2412.02010 [cond-mat.str-el].
- [22]M. E. Gazzah, P.-H. Chang, Y. Lee, H. Bhandari, R. Regmi, X. Zhou, J. F. Mitchell, L. Ke, I. I. Mazin, and N. J. Ghimire, Doping-induced spin reorientation in kagome magnet tmn₆sn₆ (2025), arXiv:2505.02936 [cond-mat.mtrl-sci].
- [23]A. Abragam and B. Bleaney, Electron Paramagnetic Resonance of Transition Ions, 1st ed., International Series of Monographs on Physics (Clarendon Press, 1970).
- [24]H. Bethe, Thermausfaltung in kristallen, Annalen der Physik 395, 133 (1929).
- [25]K. W. H. Stevens, Matrix elements and operator equivalents connected with the magnetic properties of rare earth ions, Proceedings of the Physical Society. Section A 65, 209 (1952).DOI
- [26]C. J. Ballhausen, Introduction to ligand field theory (McGraw-Hill, New York, 1962).DOI
- [27]M. T. Hutchings, Point-Charge Calculations of Energy Levels of Magnetic Ions in Crystalline Electric Fields (Academic Press, 1964) pp. 227–273.DOI
- [28]K. N. R. Taylor and M. I. Darby, Physics of rare earth solids (Chapman and Hall, London, 1972).
- [29]J. Sievers, Aspersicity of 4f-shells in their Hund’s rule ground states, Zeitschrift für Physik B Condensed Matter 45, 289 (1982).DOI
- [30]R. Skomski, Simple Models of Magnetism, Oxford Graduate Texts (OUP Oxford, 2008).DOI
- [31]P. Hoffmann, Generalization of stevens’ operator-equivalent method, Journal of Physics A: Mathematical and General 24, 35 (1991).DOI
- [32]O. Duros, A. Juhin, H. Elnaggar, G. S. Chiūbăian, and C. Brouder, General expressions for stevens and racah operator equivalents, Journal of Physics A: Mathematical and General 58, 025207 (2024).DOI
- [33]G. Pokharel, B. Ortiz, J. Chamorro, P. Sarte, L. Kautzsch, G. Wu, J. Ruff, and S. D. Wilson, Highly anisotropic magnetism in the vanadium-based kagome metal TbV₆Sn₆, Phys. Rev. Mater. 6, 104202 (2022).DOI
- [34]J. Greedan and V. Rao, An analysis of the rare earth contribution to the magnetic anisotropy in RCo₅ and R₂Co₁₇ compounds, Journal of Solid State Chemistry 6, 387 (1973).DOI
- [35]Z. Tie-song, J. Han-min, G. Guang-hua, H. Xiu-feng, and C. Hong, Magnetic properties of R ions in RCo₅ compounds (R=Pr, Nd, Sm, Gd, Tb, Dy, Ho, and Er), Phys. Rev. B 43, 8593 (1991).pubmed.ncbi.nlm.nih.gov/9996492
- [36]P. Larson, I. I. Mazin, and D. A. Papaconstantopoulos, Calculation of magnetic anisotropy energy in SmCo₅, Phys. Rev. B 67, 214405 (2003).DOI
