Calculations on CeCrB4 favor a ferromagnetic arrangement, in which the chromium moments point in the same direction, over an antiferromagnetic one, in which they point in opposite directions, but only by a narrow margin. In the model comparison, the ferromagnetic state was lower in energy by 0.15 meV per atom. The preprint converts that gap to 1.72 K, a temperature scale derived from the calculation rather than a measured transition point. Because the calculation describes a 0 K ground state, it does not establish how the material behaves at finite temperature.
A model of paired chromium atoms
CeCrB4 is the subject of an arXiv preprint, version v1, dated 28 Aug 2026. The work uses first-principles electronic-structure calculations, computer models of how electrons occupy a crystal, rather than measurements from a physical sample. Its main comparison is between ferromagnetic and antiferromagnetic CeCrB4 structures, both of which host triplet Cr–Cr dimers. The calculations use density-functional theory in FPLO18.00-52, with fully relativistic spin-orbit coupling, the PBE-GGA approximation and a GGA+U treatment of the cerium 4f orbitals using U of 6 eV.
The calculated site moments were close to zero for Ce, while each Cr carried about 0.40 µB in both magnetic configurations. The authors attribute the dimer moment mainly to localized Cr 3dz2 states. They identify four such localized states per dimer, two in each spin channel; for the studied occupancy, that corresponds to 1 µB per dimer, or 0.5 µB per Cr. This is the basis for a molecular-like picture in which the pair has bonding-like and antibonding-like states. The interpretation comes from the calculated orbital structure, not direct experimental proof of a molecular bond.
Cerium adds a second layer of uncertainty
Another part of the calculation concerns cerium’s valence, the effective charge state associated with its electrons. The results qualitatively support a mixed-valence picture, but the estimated number is lower than a spectroscopy value reported in the study. Mulliken analysis gives a Ce 4f occupation of 0.66 and an approximate nominal valence of 3.34+. The reported HERFD-XAS comparison gives a room-temperature valence of 3.54+ and a 4f occupation of 0.46. The two estimates point in the same broad direction while differing quantitatively.
Mulliken effective charges are +1.45 for Ce and +0.54 for Cr, with about two electrons transferred to the boron sublattice per formula unit. The authors interpret the calculated Ce–Cr layers as donor-like and the boron sublattice as acceptor-like. The inferred charges depend on the population-analysis scheme, so they are best read as a model description of charge distribution.
The chromium pair changes with distance
The paper then changes the separation of the two chromium atoms and follows the density of states, a map of where electron energies are concentrated. At about 3.48 Å, the Cr 3dz2 contribution approaches a single, atomic-like peak. At shorter distances it separates into two localized features, which broaden at the 2.40 Å equilibrium separation and blur below about 2.0 Å. The distance pattern is consistent with bonding-like and antibonding-like levels whose splitting depends on how close the atoms are. The scan is nonmagnetic and omits spin-orbit coupling, however, so it supports the molecular-like interpretation without settling the magnetic behavior.
Changing the metal shifts the calculated states
Composition provides a second variable in the models. In related CeTB4 compositions, where T is the transition-metal site, the localized T 3dz2 states shift downward in a rigid-band-like way as the atomic number of T rises. Extensions to heavier 4d and 5d transition metals suggest that molecular-like states can persist there as well. The result associates both chemical identity and dimer spacing with changes in the modeled electronic structure, but the chemical trend has not been experimentally validated in this paper.
What the calculation leaves open
The authors set important limits on the result. The DFT calculations describe a 0 K ground state; the PBE functional can affect electron localization and calculated magnetic moments; and GGA+U does not fully capture dynamic many-body correlations. Those cautions matter because the FM–AFM energy difference is only 0.15 meV per atom. The preprint therefore offers a computational interpretation of CeCrB4’s magnetism and dimer states, not a demonstration of finite-temperature behavior or a measured magnetic phase transition. Its proposed composition-based changes remain model predictions rather than experimentally tested effects.
Paper data and sources
Original title: Interplay between crystal structure and magnetism in CeCrB$_4$
Authors: Mirosław Werwiński, Andrzej Szajek, Andrzej Kowalczyk
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-28
DOI: Not available
Original paper · Full text