A computational study has identified NiPt- and CoIr-based alloy families as the main source of its largest predicted anomalous Hall and Nernst responses. The FePt family did not enter the study’s high-response regime, defined as an absolute anomalous Hall conductivity of at least 1,500 S cm−1 or an absolute anomalous Nernst conductivity of at least 4 A m−1 K−1.
The study is asking a practical materials question: which electronic features make a parent compound more responsive to chemical substitution and more likely to produce derivatives with very large transverse electrical and thermoelectric signals? The authors propose exchange splitting, the separation between a material’s spin channels, as a descriptor for that job. They present it as a guide to chemical tunability rather than a simple ranking of unchanged, stoichiometric compounds.
The work is an arXiv preprint, version arXiv:2608.25584v1, dated 26 August 2026. Its findings come from modeled candidates, not laboratory measurements.
The strongest modeled responses
The largest listed anomalous Hall conductivity was 2,809.07 S cm−1 for (Co0.8Fe0.2)(Ir0.7Pt0.3). The largest listed anomalous Nernst conductivity was 7.72 A m−1 K−1 for (Ni0.8Co0.2)(Pt0.7Ir0.3). Both figures came from the computational screen.
The modeled parent-to-derivative comparisons also showed a sharp increase in the selected substituted cases. CoIr’s anomalous Hall conductivity was −340 S cm−1, compared with 2,809 S cm−1 for a selected CoIr-derived alloy. NiPt’s anomalous Nernst conductivity was 1.59 A m−1 K−1, compared with 7.72 A m−1 K−1 for a selected NiPt-derived alloy.
For the thermal response, the researchers calculated anomalous Nernst conductivity at 300 K from the energy-dependent anomalous Hall conductivity using a finite-temperature Mott relation. In plain terms, the calculation used how the Hall response changes with energy to estimate the thermoelectric response at a fixed temperature.
How the virtual search worked
The search began with parent compounds retained from OQMD. The researchers kept 67 of them after requiring an energy above the stability hull of no more than 0.3 eV and a finite magnetic moment larger than 0.1 µB per formula unit.
They then modeled chemical substitutions in tetragonal L10 structures. The virtual crystal approximation represented each substituted site with concentration-weighted interpolation, producing transport data for 2,251 chemically substituted candidates.
An automated in-house Python workflow linked QUANTUM ESPRESSO, WANNIER90 and WANNIERTOOLS. The electronic-structure calculations used PBE, fully relativistic ONCV pseudopotentials and spin-orbit coupling. The Hall conductivity was obtained by integrating Berry curvature, a calculated property of the electronic bands, while the Nernst conductivity was derived from the energy-dependent Hall response at 300 K.
To compare materials with different magnetic moments, the candidates were grouped into net-magnetization bins 0.5 µB per formula unit wide. Each bin was summarized using its mean response and its 95th percentile.
A lower magnetic range stood out
Across those groups, both absolute anomalous Hall and anomalous Nernst conductivity had their largest mean and 95th-percentile values in the 1 to 2 µB per formula unit magnetization range.
That pattern is part of the case for using exchange splitting as a selection tool. The authors interpret the quantity as a practical descriptor for judging how much a parent compound may be reshaped by substitution, instead of using it only to rank the transport response of the parent itself.
Why the candidates responded
The study also examined the calculated electronic bands of the selected CoIr-derived candidate. It found a nodal-line structure, an extended set of band crossings, that was absent in the CoIr parent. Once spin-orbit coupling opened gaps along that structure, it coincided with strong contributions from Berry curvature and to the Hall conductivity.
In the selected NiPt-derived case, near-R crossings between upper majority-spin and minority-spin bands were associated with substantial contributions to the anomalous Nernst conductivity. A more extended nodal line in a lower band made a weaker additional contribution. The analysis describes these features in the selected case, not as a finding about every composition in the screen.
A prediction that still needs testing
The evidence is entirely computational. The candidates were treated as tetragonal L10 compositions in first-principles calculations using the virtual crystal approximation, and the proposed CoIr- and NiPt-derived alloys are presented as candidates for experimental validation.
The virtual crystal treatment does not settle how chemical disorder would affect the predicted responses. The study identifies stability checks, phonon calculations and screening of magnetic critical temperatures as next steps rather than reported results.
No uncertainty estimate was reported for the headline values. The authors also note a possible trade-off: smaller exchange splitting may coincide with lower Curie temperatures, putting chemical tunability in tension with thermal robustness.
The result is a proposed screening principle whose usefulness depends on whether the predicted alloys can be synthesized and whether measurements reproduce the calculated transport responses.
The acknowledgements report partial support from JST-CREST, JSPS KAKENHI and MEXT programs. They also state that the calculations used the Numerical Materials Simulator at NIMS.
Paper data and sources
Original title: Exchange splitting as a descriptor for giant anomalous Hall and Nernst effects in ferromagnets
Authors: Ivan Kurniawan, Guangzong Xing, Yoshio Miura, Keisuke Masuda
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text