Реферат: Magnetic-material design is a demanding test for artificial intelligence (AI) because the measured response is rarely determined by composition alone. In systems such as molecular magnets, spin-crossover compounds, lowdimensional magnets, magnetocaloric phases, and permanent magnets, the magnetic structure is interconnected with atomic vibrations, the specifics of the crystal packing, defects, external thermodynamic stimuli. This selective review examines how digital methods address these relationships. We distinguish predictions of spin state, exchange and anisotropy from sample-level observables such as magnetic blocking, coercivity and reversible magnetocaloric response. Quantitative examples show both useful advances and important limits: a model may enrich a candidate list without accurately predicting the property magnitude, and agreement with calculated labels does not establish experimental performance. Curated datasets, quantum-chemical surrogates, graph models and adaptive calculations have extended the range of compounds that can be assessed, whereas prospective experiments and repeated experimental feedback remain less common in the studies considered. We argue that the appropriate data object for magnetic AI is a sample-context record linking chemical identity, structure, preparation, physical state, measurement protocol, raw signal, analysis and uncertainty. Such records support, but do not replace, independent validation. The priorities are chemically appropriate targets, comparisons with simple baselines, tests across chemical families and laboratories, and documented feedback between predictions and measurements. Human-supervised workflows offer a practical route to these goals while preserving the roles of synthesis expertise and physical interpretation.

Библиографическая ссылка: Tretyakov E.V. , Ananikov V.P.
Digital technologies and artificial intelligence in the design of molecular magnets and magnetic materials
Coordination Chemistry Reviews. 2027. V.572. 218726 :1-22. DOI: 10.1016/j.ccr.2026.218726

Даты:

Поступила в редакцию:	28 июл. 2026 г.
Принята к публикации:	25 окт. 2026 г.
Опубликована online:	3 нояб. 2026 г.

Идентификаторы БД: Нет идентификаторов

Альметрики: