A computer-based search has identified 12 new predictions for a type of magnetic material whose electronic bands split by spin while it remains metallic. The study found 16 metallic d-wave altermagnets in all, including four compounds already known from experiments, and selected six of the new predictions for closer attention.
The result is a computational prediction, not a demonstration that the compounds can be made or will behave exactly as calculated. The supplied record identifies the work as a preprint, and its material, magnetic and transport findings are reported as calculations rather than laboratory measurements.
A broad search, narrowed by stability
The researchers examined 280 TiNiSi-type MTA systems to ask whether this structural family could host metallic d-wave altermagnetism with useful spin-splitter and anomalous Hall responses. The work used high-throughput first-principles screening across the specified combinations of M, T and A elements.
To narrow the list, the team compared the calculated formation energies of TiNiSi- and ZrAlNi-type structures with ternary convex-hull calculations. In plain terms, that test asked how close each proposed composition was to the lowest-energy combinations of the same elements. The researchers then carried out magnetic-state calculations on 183 TiNiSi-type phases whose convex-hull distance was no more than 0.2 electronvolts per atom.
For those phases, several collinear magnetic arrangements were compared to identify the lowest-energy calculated state. The search then focused on compounds that were both metallic and showed the characteristic d-wave altermagnetic band splitting. Here, that label refers to a calculated magnetic state in which spin-dependent electronic bands separate near the Fermi level.
Six compounds move to the front of the queue
The six leading new candidates were ScMnP, TaMnAs, ScMnAs, MoMnAs, MoMnSi and WMnSi. Their calculated energy separations from the next-lowest magnetic state were 5, 45, 15, 22, 29 and 30 millielectronvolts per formula unit, respectively. Those values helped prioritize the compounds for follow-up; they are calculated energy differences, not measurements from laboratory samples.
Across the analyzed altermagnets, the electronic calculations showed metallic behavior and roughly 100 millielectronvolts of altermagnetic splitting near the Fermi level. Put simply, the calculation placed electronic bands associated with different spins at different energies in a part of the spectrum relevant to the transport analysis.
The researchers also tested whether a more complicated, canted magnetic arrangement might undercut the proposed state. In six compounds, they examined three prescribed paths in which the magnetic moments were tilted away from the collinear arrangement. Along all of those tested paths, the collinear altermagnetic state remained the lowest in energy, and no robust lower-energy intermediate state appeared.
Promising transport signals remain theoretical
The calculated spin-splitter angle was about 0.2 for ScMnP and TaMnAs. It was slightly below 0.1 for ScMnAs, MoMnSi and WMnSi, while MoMnAs showed a negligible angle. These are calculated transport quantities near the Fermi level under the study's stated approximation, not direct measurements of spin currents.
All six leading candidates also had a finite calculated anomalous Hall conductivity, a sideways electrical response in the calculation. One reported Hall-conductivity component reached several hundred siemens per centimetre in ScMnP, TaMnAs, MoMnAs and WMnSi, while the other component had comparable magnitudes in all six compounds.
For the spin and charge conductivities, the team used a Boltzmann transport calculation with a common, constant relaxation time. The intrinsic anomalous Hall conductivity was obtained by integrating the Berry curvature of occupied electronic states with WannierTools. The resulting conductivities are therefore outputs of a model rather than fixed performance ratings for a finished material.
What still has to be tested
The calculations do not establish that the 12 newly predicted compounds can be synthesized in the TiNiSi structure. The convex-hull distance indicates how likely that structure may be for a composition, but it does not definitively predict synthesizability. The stability comparison included TiNiSi and ZrAlNi candidate structures, leaving open the possibility that other polymorphs could be lower in energy.
The magnetic checks have a similar boundary. Noncollinear behavior was examined directly in six representative compounds and along three selected canting paths, not across every predicted material or every possible magnetic configuration. The collinear result therefore applies to the paths tested, rather than automatically to every candidate.
The magnetic label also has to be read within the design of the screen. The researchers compared several collinear configurations after applying the stability cutoff, which identifies the lowest-energy state among the configurations tested. A separate noncollinear check extended that examination, but only to six compounds and three paths.
That leaves synthesis and measurement as the practical follow-up. Researchers will need to determine whether the predicted compounds adopt the intended structure and whether their calculated spin-splitter angles and anomalous Hall conductivities appear in real samples. For now, the six compounds are prioritized computational targets, not experimentally demonstrated materials.
The work was supported by the U.S. Department of Energy's EPSCoR program, with computing carried out at Iowa State University and the University of Nebraska.
Paper data and sources
Original title: Mapping metallic d-wave altermagnetism across the TiNiSi structural family
Authors: Zhen Zhang, Subhadip Pradhan, Kirill D. Belashchenko, Vladimir Antropov
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-24
DOI: Not available
Original paper · Full text