Machine learning-driven discovery of hard magnetic materials using high-throughput computation and screening
Halder A., Paudyal D., Sanvito S., Takac M., Ucar H.
Article, Acta Materialia, 2025, DOI Link
View abstract ⏷
We present a machine-learning-driven framework for discovering high-performance rare-earth-free hard magnetic materials integrating machine learning, a universal graph deep-learning interatomic potential, and density functional theory validation. Key contributions include the identification of FeCo-based ternary alloys with remarkable magnetic properties, such as uniaxial anisotropy constant, K1, Curie temperature, TC, and saturation magnetization, MS. Notable examples include Fe6CoB2 and FeCo5B, which exhibit K1 values of 1.76 MJ/m3 and 1.00 MJ/m3, respectively, with MS above 1.3 T, and TC exceeding 600 K. These properties align with the needs of high-temperature and high-performance applications. The universal graph deep-learning interatomic potential M3GNet accelerates the structural relaxation process, enabling the efficient screening of 48,000 candidate structures, while density functional theory validates the top performers with energy product (BH)max reaching more than 600 kJ/m3. Our study highlights a scalable, efficient pipeline for advancing the discovery of permanent magnets, reducing reliance on rare-earth elements.
Collapse of the standard ferromagnetic domain structure in hybrid Co/Molecule bilayers
Benini M., Shumilin A., Kabanov V., Rakshit R.K., Sahoo A., Halder A., Droghetti A., Cugini F., Solzi M., Bisero D., Graziosi P., Riminucci A., Bergenti I., Singh M., Gnoli L., Sanna S., Cinchetti M., Mertelj T., Sanvito S., Dediu V.A.
Article, Nature Communications, 2025, DOI Link
View abstract ⏷
The interplay between Hund’s coupling, exchange interaction and magnetic anisotropy is responsible for a multitude of magnetic phases, ranging from conventional ferromagnetism to exotic spin textures. Yet, engineering and fine-tuning a magnetic state remains a major challenge in modern magnetism. We show that the chemisorption of organic molecules over Co thin films offers a tool to transform the films from ferromagnetic to a glassy-type state. This emerges when the correlation length of the random anisotropy field, induced by the π-d molecule/metal hybridization, is comparable to the characteristic exchange length. Such a state is characterized by the collapse of the standard domain structure and the emergence of blurred pseudo-domains intertwined by diffuse and irregular domain walls. The magnetization reversal then involves topological vortex-like structures, which are here predicted and successfully measured by magnetic-force microscopy. At the macroscopic level this new glassy-type state is defined by a giant magnetic hardening and the violation of the magnetization-reversal Rayleigh law. Our work thus shows that the electronic interaction of a standard thin-film magnet with readily available molecules can generate structures with remarkable new magnetic properties, and thus opens a new avenue for the design of tailored-on-demand magnetic composites.
Half-Metallic Transport and Spin-Polarized Tunneling through the van der Waals Ferromagnet Fe4GeTe2
Halder A., Nell D., Sihi A., Bajaj A., Sanvito S., Droghetti A.
Article, Nano Letters, 2024, DOI Link
View abstract ⏷
We examine the coherent spin-dependent transport properties of the van der Waals (vdW) ferromagnet Fe4GeTe2 using density functional theory combined with the nonequilibrium Green’s function method. Our findings reveal that the conductance perpendicular to the layers is half-metallic, meaning that it is almost entirely spin-polarized. This property persists from the bulk to a single layer, even under significant bias voltages and with spin-orbit coupling. Additionally, using dynamical mean field theory for quantum transport, we demonstrate that electron correlations are important for magnetic properties but minimally impact the conductance, preserving almost perfect spin-polarization. Motivated by these results, we then study the tunnel magnetoresistance (TMR) in a magnetic tunnel junction consisting of two Fe4GeTe2 layers with the vdW gap acting as an insulating barrier. We predict a TMR ratio of ∼500%, which can be further enhanced by increasing the number of Fe4GeTe2 layers in the junction.
Theoretical perspective on the modification of the magnetocrystalline anisotropy at molecule-cobalt interfaces
Halder A., Bhandary S., O'Regan D.D., Sanvito S., Droghetti A.
Article, Physical Review Materials, 2023, DOI Link
View abstract ⏷
We study the modification of the magnetocrystalline anisotropy (MCA) of Co slabs induced by several different conjugated molecular overlayers, i.e., benzene, cyclooctatetraene, naphthalene, pyrene, and coronene. We perform first-principles calculations based on density functional theory and the magnetic force theorem. Our results indicate that molecular adsorption tends to favor a perpendicular MCA at surfaces. A detailed analysis of various atom-resolved quantities, accompanied by an elementary model, demonstrates that the underlying physical mechanism is related to the metal-molecule interfacial hybridization and, in particular, to the chemical bonding between the molecular C pz and the out-of-plane Co dz2 orbitals. This effect can be estimated from the orbital magnetic moment of the surface Co atoms, a microscopic observable accessible to both theory and experiments. As such, we suggest a way to directly assess the MCA modifications at molecule-decorated surfaces, overcoming the limitations of experimental studies that rely on fits of magnetization hysteresis loops. Finally, we also study the interface between Co and both C60 and Alq3, two molecules that find widespread use in organic spintronics. We show that the modification of the surface Co MCA is similar on adsorption of these two molecules, thereby confirming the results of recent experiments.
DFT+ ς2 method for electron correlation effects at transition metal surfaces
Droghetti A., Radonjic M.M., Halder A., Rungger I., Chioncel L.
Article, Physical Review B, 2022, DOI Link
View abstract ⏷
We present a computational approach for electronically correlated metallic surfaces and interfaces, which combines density functional and dynamical mean-field theory using a multiorbital perturbative solver for the many-body problem. Our implementation is designed to describe ferromagnetic metallic thin films on a substrate. The performances are assessed in detail for a Fe monolayer on a W(110) substrate, a prototypical nanoscale magnetic system. Comparing our results to photoemission data, we find qualitative and quantitative improvements in the calculated spectral function with respect to the results of density functional theory within the local spin density approximation. In particular, the spin splitting of the d states is drastically reduced and, at the same time, their spectral width becomes narrower. The method is, therefore, able to account for the main correlation effects in the system.
Machine learning classification of binary semiconductor heterostructures
Rom S., Ghosh A., Halder A., Dasgupta T.S.
Article, Physical Review Materials, 2021, DOI Link
View abstract ⏷
Heterostructures of two semiconductors are at the heart of semiconductor devices with tremendous technological importance. The prediction and designing of semiconductor heterostructures of a specific type is a difficult materials science problem, posing a challenge to experimental and computational investigations. In this study, we first establish that the prediction of heterostructure type can be made with good accuracy from the knowledge of the band structure of constituent semiconductors. Following this, we apply machine learning, built on features characterizing constituent semiconductors, on a known dataset of binary semiconductor heterostructures extended by a synthetic minority oversampling technique. A significant feature of engineering made it possible to train a classifier model predicting the heterostructure type with an accuracy of 89%. Using the trained model, a large number (872 number) of unknown heterostructure semiconductor types involving elemental and binary semiconductors is theoretically predicted. Interestingly, the developed scheme is found to be extendable to heterojunctions of semiconductor quantum dots.
The Critical Role of Stereochemically Active Lone Pair in Introducing High Temperature Ferroelectricity
Ali Saha R., Halder A., Fu D., Itoh M., Saha-Dasgupta T., Ray S.
Article, Inorganic Chemistry, 2021, DOI Link
View abstract ⏷
In this paper, a comparative structural, dielectric, and magnetic study of two langasite compounds Ba3TeCo3P2O14 (absence of lone pair) and Pb3TeCo3P2O14 (Pb2+ 6s2 lone pair) have been carried out to precisely explore the development of room temperature spontaneous polarization in the presence of a stereochemically active lone pair. In the case of Pb3TeCo3P2O14, mixing of both Pb 6s with Pb 6p and O 2p helps the lone pair to be stereochemically active. This stereochemically active lone pair brings a large structural distortion within the unit cell and creates a polar geometry, while the Ba3TeCo3P2O14 compound remains in a nonpolar structure due to the absence of any such effect. Consequently, polarization measurement under varying electric fields confirms room temperature ferroelectricity for Pb3TeCo3P2O14, which was not the case for Ba3TeCo3P2O14. A detailed study was carried out to understand the microscopic mechanism of ferroelectricity, which revealed the exciting underlying activity of a polar TeO6 octahedral unit as well as Pb-hexagon.
Understanding complex multiple sublattice magnetism in double double perovskites
Halder A., Das S., Sanyal P., Saha-Dasgupta T.
Article, Scientific Reports, 2021, DOI Link
View abstract ⏷
Understanding magnetism in multiple magnetic sublattice system, driven by the interplay of varied nature of magnetic exchanges, is on one hand challenging and on other hand intriguing. Motivated by the recent synthesis of AA′BB′O6 double double perovskites with multiple magnetic ions both at A- and B-sites, we investigate the mechanism of magnetic behavior in these interesting class of compounds. We find that the magnetism in such multiple sublattice compounds is governed by the interplay and delicate balance between two distinct mechanisms, (a) kinetic energy-driven multiple sublattice double exchange mechanism and (b) the conventional super-exchange mechanism. The derived spin Hamiltonian based on first-principles calculations is solved by classical Monte Carlo technique which reproduces the observed magnetic properties. Finally, the influence of off-stoichiometry, as in experimental samples, is discussed. Some of these double double perovskite compounds are found to possess large total magnetic moment and also are found to be half-metallic with reasonably high transition temperature, which raises the hope of future applications of these large magnetic moment half-metallic oxides in spintronics and memory devices.
Covalency driven modulation of paramagnetism and development of lone pair ferroelectricity in multiferroic Pb3TeMn3 P2 O14
Saha R.A., Halder A., Saha-Dasgupta T., Fu D., Itoh M., Ray S.
Article, Physical Review B, 2020, DOI Link
View abstract ⏷
We have investigated the structural, magnetic, and dielectric properties of the Pb-based langasite compound Pb3TeMn3P2O14 both experimentally and theoretically in light of metal-oxygen covalency, and the consequent generation of multiferroicity. It is known that the large covalency between Pb 6p and O 2p plays an instrumental role behind the stereochemical lone pair activity of Pb. The same happens here, but a subtle structural phase transition above room temperature changes the degree of such lone pair activity and the system becomes ferroelectric below 310 K. Interestingly, this structural change also modulates the charge densities on different constituent atoms and consequently the overall magnetic response of the system while maintaining the global paramagnetism behavior of the compound. This single origin of modulation in polarity and paramagnetism inherently connects both functionalities and the system exhibits mutiferroicity at room temperature.
Prediction of the properties of the rare-earth magnets Ce2Fe17-xCoxCN: A combined machine-learning and ab initio study
Halder A., Rom S., Ghosh A., Saha-Dasgupta T.
Article, Physical Review Applied, 2020, DOI Link
View abstract ⏷
We employ a combination of machine learning and first-principles calculations to predict magnetic properties of rare-earth lean magnets. For this purpose, based on a training set constructed out of experimental data, the machine is trained to make predictions on magnetic transition temperature (Tc), largeness of saturation magnetization (μ0Ms), and the nature of the magnetocrystalline anisotropy (Ku). Subsequently, the quantitative values of μ0Ms and Ku of the yet-to-be synthesized compounds, screened by machine learning, are calculated by first-principles density-functional theory. The applicability of the proposed technique of combined machine learning and first-principles calculations is demonstrated on 2-17-X magnets, Ce2Fe17-xCoxCN. Further to this study, we explore the stability of the proposed compounds by calculating vacancy formation energy of small atom interstitials (N/C). Our study indicates a number of compounds in the proposed family and offers the possibility to become a solution for cheap and efficient permanent magnets.
Understanding the curious magnetic state of Sr3OsO6
Das S., Halder A., Chakraborty A., Dasgupta I., Saha-Dasgupta T.
Article, Physical Review B, 2020, DOI Link
View abstract ⏷
Motivated by the recent report on the high-Tc ferromagnetic insulating state of a single transition metal containing a double perovskite compound, Sr3OsO6 [Wakabayashi et al., Nat Commun. 10, 535 (2019)2041-172310.1038/s41467-019-08440-6], we study this curious behavior by employing first-principles calculations in conjunction with exact diagonalization of the full t2g multiplet problem of two Os sites. Our analysis highlights the fact that stabilization of Sr3OsO6 in the cubic phase in epitaxially grown thin film is the key to both ferromagnetic correlation and the high-temperature scale associated with it. It also provides a natural explanation for why the sister compound, Ca3OsO6, exhibits low-TN antiferromagnetism in its monoclinic structure. Furthermore, the insulating property is found to be driven by the opening of a Mott gap in the half-filled spin-orbit coupled j=3/2 manifold of d2 Os. We point out that Sr2CaOsO6, which naturally forms in the cubic phase, would be worthwhile to explore as a future candidate to exhibit a high-Tc ferromagnetic insulating state in bulk form.
Magnetism in cation-disordered 3d-4d/5d double perovskites
Halder A., Sanyal P., Saha-Dasgupta T.
Article, Physical Review B, 2019, DOI Link
View abstract ⏷
Employing an exact diagonalization Monte Carlo solution of the first-principles-derived model Hamiltonian of a number of A2BB′O6 double-perovskite compounds, containing a 3d transition metal at the B site and a 4d or 5d transition metal ion at the B′ site, we investigate the effect of B/B′ cation disorder on their magnetic properties. Our exhaustive study reveals that the influence of cation disorder on both the magnetic transition temperature and magnetization depends strongly on the underlying exchange mechanism with a distinct difference between the double exchange mechanism and a combined double exchange and superexchange mechanism. We further find that the nature of the disorder has a remarkable effect. While the uncorrelated or random disorder has a severely detrimental effect, especially for magnetism having a superexchange contribution, correlated disorder with a high degree of short-range order retains the magnetic properties of the fully cation-ordered compounds to a large extent. Our findings shed light on the puzzling report of magnetic order in fully cation-disordered CrRu oxides.
Machine-learning-assisted prediction of magnetic double perovskites
Halder A., Ghosh A., Dasgupta T.S.
Article, Physical Review Materials, 2019, DOI Link
View abstract ⏷
Magnetism is an important property of materials that plays a key role in many different applications. In the present paper, we use a combination of computational tools: a machine-learning technique for screening of stable candidates, an evolutionary algorithm for crystal structure determination, and first-principles calculations for characterization of electronic and magnetic properties to make predictions on magnetic double perovskites, which are yet to be synthesized. Out of 412 scanned candidates of A2BB′O6 composition with 3d and 4d or 5d transition metals at B and B′ sites, we found 33 compounds to form stable double-perovskite structures, 25 of which were further considered for characterization of their structure and properties. Our exercise predicted 21 double perovskites of varying magnetic and electronic properties, ranging from ferromagnetic half metals to ferri- and antiferromagnetic insulators to ferromagnetic metals and a rare example of antiferromagnetic metals. Our computational study is expected to help in discovering new magnetic double perovskites.
Computer predictions on Rh-based double perovskites with unusual electronic and magnetic properties /639/766/119 /639/301/1034 /639/766/119 /639/301/1034 article
Halder A., Nafday D., Sanyal P., Saha-Dasgupta T.
Article, npj Quantum Materials, 2018, DOI Link
View abstract ⏷
In search for new magnetic materials, we make computer prediction of structural, electronic and magnetic properties of yet-to-be synthesized Rh-based double perovskite compounds, Sr(Ca)2BRhO6 (B=Cr, Mn, Fe). We use combination of evolutionary algorithm, density functional theory, and statistical-mechanical tool for this purpose. We find that the unusual valence of Rh5+ may be stabilized in these compounds through formation of oxygen ligand hole. Interestingly, while the Cr-Rh and Mn-Rh compounds are predicted to be ferromagnetic half-metals, the Fe-Rh compounds are found to be rare examples of antiferromagnetic and metallic transition-metal oxide with three-dimensional electronic structure. The computed magnetic transition temperatures of the predicted compounds, obtained from finite temperature Monte Carlo study of the first principles-derived model Hamiltonian, are found to be reasonably high. The prediction of favorable growth condition of the compounds, reported in our study, obtained through extensive thermodynamic analysis should be useful for future synthesize of this interesting class of materials with intriguing properties.
Magnetism in Sr2CrMoO6: A combined ab initio and model study
Sanyal P., Halder A., Si L., Wallerberger M., Held K., Saha-Dasgupta T.
Article, Physical Review B, 2016, DOI Link
View abstract ⏷
Using a combination of first-principles density functional theory (DFT) calculations and exact diagonalization studies of a first-principles derived model, we carry out a microscopic analysis of the magnetic properties of the half-metallic double perovskite compound Sr2CrMoO6, a sister compound of the much discussed material Sr2FeMoO6. The electronic structure of Sr2CrMoO6, though appearing similar to Sr2FeMoO6 at first glance, shows nontrivial differences with that of Sr2FeMoO6 on closer examination. In this context, our study highlights the importance of charge transfer energy between the two transition metal sites. The change in charge transfer energy due to a shift of Cr d states in Sr2CrMoO6 compared to Fe d in Sr2FeMoO6 suppresses the hybridization between Cr t2g and Mo t2g. This strongly weakens the hybridization-driven mechanism of magnetism discussed for Sr2FeMoO6. Our study reveals that, nonetheless, the magnetic transition temperature of Sr2CrMoO6 remains high since an additional superexchange contribution to magnetism arises with a finite intrinsic moment developed at the Mo site. We further discuss the situation in comparison to another related double perovskite compound, Sr2CrWO6. We also examine the effect of correlation beyond DFT, using dynamical mean field theory.
