Constrained DFT-based magnetic machine-learning potentials for magnetic alloys: a case study of FeAl | Scientific … – Nature.com
Magnetic multi-component moment tensor potential (mMTP)
The concept of magnetic multi-component Moment Tensor Potential (mMTP) presented in the current research is based on the previously developed non-magnetic MTP for multi-component systems41,42 and magnetic MTP for single-component systems35.
The mMTP potential is local, i.e., the energy of the atomistic system is a sum of energies of individual atoms:
$$begin{aligned} E = sum _{i=1}^{N_a}E_i, end{aligned}$$
(1)
where i stands for the individual atoms in an (N_a)-atom system. We note that any configuration includes lattice vectors ({{varvec{L}}} = {{{varvec{l}}}_1,{{varvec{l}}}_2,{{varvec{l}}}_3}), atomic positions ({{varvec{R}}} = {{{varvec{r}}}_1, ldots , {{varvec{r}}}_{N_a}}), types (Z = {z_1,ldots ,z_{N_{a}}}) (we also denote (N_{rm types}) by the total number of atomic types in the system), and magnetic moments (M = {m_1,ldots ,m_{N_a}}). The energy of the atom (E_i), in turn, has the form:
$$begin{aligned} E_i = sum _{alpha =1}^{alpha _{rm max}} xi _{alpha }B_{alpha }({mathfrak n}_i), end{aligned}$$
(2)
where ({{varvec{xi }}} = {xi _{alpha } }) are the linear parameters to be optimized and (B_alpha) are the so-called basis functions, which are contractions of the descriptors25 of atomistic environment ({mathfrak n}_i), yielding a scalar. The (alpha _text {max}) parameter can be changed to provide potentials with different amount of parameters35.
The descriptors are composed of the radial part, i.e., the scalar function depending on the interatomic distances and atomic magnetic moments, and the angular part, which is a tensor of rank (nu):
$$begin{aligned} M_{mu ,nu }({mathfrak n}_i)=sum _{j} f_{mu }(| {{varvec{r}}}_{ij}|,z_i,z_j,m_i,m_j)underbrace{{{varvec{r}}}_{ij}otimes ...otimes {{varvec{r}}}_{ij}}_nu text { times }, end{aligned}$$
(3)
where ({mathfrak n}_i) stands for the atomic environment, including all the atoms within the (R_text {cut}) distance (or less) from the central atom i, (mu) is the number of the radial function, (nu) is the rank of the angular part tensor, (|{{varvec{r}}}_{ij}|) is the distance between the atoms i and j, (z_i) and (z_j) are the atomic types, (m_i) and (m_j) are the magnetic moments of the atoms.
The radial functions are expanded in a basis of Chebyshev polynomials:
$$begin{aligned} f_{mu }(|r_{ij}|,z_i,z_j,m_i,m_j) = sum _{zeta =1}^{N_{phi }} sum _{beta =1}^{N_{psi }}sum _{gamma =1}^{N_{psi }}c_{mu ,z_i,z_j}^{zeta ,beta ,gamma } phi _{zeta }(|{varvec{r}}_{ij}|) psi _{beta }(m_i)psi _{gamma }(m_j) (R_{rm cut} - |{varvec{r}}_{ij}|)^2. end{aligned}$$
(4)
Here ({{varvec{c}}} = {c_{mu ,z_i,z_j}^{zeta ,beta ,gamma }}) are the radial parameters to be optimized, each of the functions (phi _{zeta }(|{varvec{r}}_{ij}|)), (psi _{beta }(m_i)), (psi _{gamma }(m_i)) is a Chebyshev polynomial of order (zeta), (beta) and (gamma) correspondingly, taking values from (-1) to 1. The function (phi _{zeta }(|{varvec{r}}_{ij}|)) yields the dependency on the distance between the atoms i and j, while the functions (psi _{beta }(m_i)) and (psi _{gamma }(m_j)) yield the dependency on the magnetic moments of the atoms i and j, correspondingly. The arguments of the functions (phi _{zeta }(|{varvec{r}}_{ij}|)) are on the interval ((R_{rm min},R_{rm cut})), where (R_{rm min}) and (R_{rm cut}) are the minimum and maximum distance, correspondingly, between the interacting atoms. The functions (psi _{beta }(m_i)) and (psi _{gamma }(m_j)) are of the same structure, which we explain for the case of the former one. The argument of the function (psi _{beta }(m_i)) is the magnetic moment of the atom i, taking the values on the ((-M_{rm max}^{z_i},M_{rm max}^{z_i})) interval. The value (M_{rm max}^{z_i}) itself depends on the type of atom (z_i), and is determined as the maximal absolute value of the magnetic moment for atom type (z_i) in the training set. Similar to the conventional MTP, the term ((R_{rm cut} - |{varvec{r}}_{ij}|)^2) provides smooth fading to 0 when approaching the (R_{rm cut}) distance, in accordance with the locality principle (1).
We note that magnetic degrees of freedom (m_i) from (4) are collinear, i.e., they can take negative or positive values as projection onto the Z axis (though the choice of the axis is arbitrary). This way, in comparison to non-magnetic atomistic systems with N atoms, in which the amount of degrees of freedom equals 4N (namely 3N for coordinates and N for types), for the description of magnetic systems additional N degrees of freedom are introduced, standing for the magnetic moment (m_i) of each atom. The amount of parameters entering the radial functions (Eq. 4) also increases in mMTP compared to the conventional MTP41,42. Namely, in MTP this number equals (N_{mu } cdot N_{phi } cdot N_{rm types}^2), while in mMTP it is (N_{mu } cdot N_{phi } cdot N_{rm types}^2 cdot N_{psi }^2). Thus, if we take (N_{psi } = 2) (which is used in the current research), the amount of the parameters entering the radial functions would be four times more in mMTP then in MTP.
We denote all the mMTP parameters by ({varvec{theta }}= {{varvec{xi }}, {varvec{c}} }) and the total energy (1) of the atomic system by (E=E({{varvec{theta }}})=E({{varvec{theta }}};M)=E({{varvec{theta }}};{{varvec{L}}},{{varvec{R}}},Z,M)).
The tensor (Eq. (4)) includes collinear magnetic moments in its functional form. However, it is not invariant with respect to the inversion of magnetic moments, i.e., (E({{varvec{theta }}};M) ne E({{varvec{theta }}};-M)), while both original and spin-inverted configurations should yield the same energy due to the arbitrary orientation of the projection axis, which we further call the magnetic symmetry.
We use data augmentation followed by explicit symmetrization with respect to magnetic moments to train a symmetric mMTP as we discuss below. Assume we have K configurations in the training set with DFT energies (E_k^{rm DFT}), forces ({varvec{f}}^{rm DFT}_{i,k}), and stresses (sigma ^{rm DFT}_{ab,k}) ((a,b=1,2,3)) calculated. We find the optimal parameters (bar{{{varvec{theta }}}}) (fit mMTP) by minimizing the objective function:
$$begin{aligned} &sum _{k=1}^{K} Biggl [ w_{rm e} Biggl | frac{E_k ({varvec{theta }}; M) + E_{k}({varvec{theta }}; -M)}{2} - E_{k}^{rm DFT}Biggr |^2 \&quad + w_{rm f} sum _{i=1}^{N_a} Biggl | frac{{varvec{f}}_{i,k}({varvec{theta }};M) + {varvec{f}}_{i,k}({varvec{theta }};-M)}{2} - {varvec{f}}^{rm DFT}_{i,k}Biggr |^2 \&quad +w_{rm s} sum _{a,b=1}^{3} Biggl | frac{sigma _{ab,k}({varvec{theta }};M)+sigma _{ab,k}({varvec{theta }};-M)}{2} -sigma ^{rm DFT}_{ab,k}Biggr |^2 Biggr ], end{aligned}$$
(5)
where (w_{rm e}), (w_{rm f}), and (w_{rm s}) are non-negative weights. By minimizing (5) we find such optimal parameters (bar{{{varvec{theta }}}}) that yield (E_k (bar{{varvec{theta }}}; M) approx E_k (bar{{varvec{theta }}}; -M)), (k = 1, ldots , K) (the same fact takes place for the mMTP forces and stresses), i.e., we symmetrize the training set to make mMTP learn the required symmetry from the data itselfthis is called data augmentation.
Next, we modify mMTP to make the energy used for the simulations (e.g., relaxation of configurations) to satisfy the exact symmetry:
$$begin{aligned} E^{rm symm}(bar{{{varvec{theta }}}};M) = dfrac{E(bar{{varvec{theta }}};M)+E(bar{{varvec{theta }}};-M)}{2}. end{aligned}$$
(6)
That is, we substitute the mMTP energy (1) into (6) and get a functional form which satisfies the exact identity (E^{rm symm}(bar{{{varvec{theta }}}};M) = E^{rm symm}(bar{{{varvec{theta }}}};-M)) for any configuration. We also note that (E (bar{{varvec{theta }}}) approx E^{rm symm}(bar{{{varvec{theta }}}})).
We use the cDFT approach with hard constraints(i.e., Lagrange multiplier) as proposed by Gonze et al. in Ref.19. One way to formulate it is to first note that in a single-point DFT calculation we minimize the Kohn-Sham total energy functional (E[rho ; {{varvec{R}}}]) with respect to the electronic density (rho =rho (r)) (here (rho) combines the spin-up and spin-down electron densities), keeping the nuclei position ({{varvec{R}}}) fixed. In other words, we solve the following minimization problem:
$$begin{aligned} E_{rm DFT}({{varvec{R}}}) = min _rho E[rho ; {{varvec{R}}}], end{aligned}$$
and from the optimal (rho ^* = mathrm{arg,min} E[rho ; {{varvec{R}}}]) we can, e.g., find magnetization (m(r) = rho ^*_+ - rho ^*_-), where the subscripts denote the spin-up ((+)) and spin-down () densities. The magnetic moment of the ith atom can be found by integrating m(r) over some (depending on the partitioning scheme) region around the atom:
$$begin{aligned} m_i = int _{Omega _i} m(r) textrm{d}r. end{aligned}$$
(7)
Since the minimizer (rho ^*) depends on ({{varvec{R}}}), (m_i) are also the functions of ({{varvec{R}}}).
According to the cDFT approach19, we now formulate the problem of minimizing (E[rho ; {{varvec{R}}}]) in which not only ({{varvec{R}}}),but also (rho) is allowed to change only subject to constraints (7):
$$begin{aligned} begin{array}{rcl} E_{rm cDFT}(rho, {{varvec{R}}}, M) =&{} min _rho &{} E[rho ; {{varvec{R}}}] \ &{} text {subject to} &{} m_i = int _{Omega _i} big (rho _{+}(r)-rho _-(r)big ) textrm{d}r. end{array} end{aligned}$$
The algorithmic details of how this minimization problem is solved, and how the energy derivatives (forces, stresses, torques) are computed, are described in detail in Ref.19.
We used the ABINIT code43,44 for DFT (and cDFT recently developed and described in Ref.19) calculations with (6times 6times 6) k-point mesh and cutoff energy of 25 Hartree (about 680 eV). We utilized the PAW PBE method with the generalized gradient approximation. We applied constraints on magnetic moments of all atoms during cDFT calculations.
We fitted an ensemble of five mMTPs with 415 parameters in order to quantify the uncertainty of mMTPs predictions. For each mMTP we took (R_{rm min} = 2.1 ~) , (R_{rm cut} = 4.5 ~), (M_{rm max}^{rm Al} = 0.1 ~mu _B), and (M_{rm max}^{rm Fe} = 3.0 ~mu _B). The weights in the objective function (5) were (w_{rm e} = 1), (w_{rm f} = 0.01) (^2), and (w_{rm s} = 0.001).
See the article here:
Constrained DFT-based magnetic machine-learning potentials for magnetic alloys: a case study of FeAl | Scientific ... - Nature.com
- Color Sensor Array and Machine Learning Classify Tea and Measure Antioxidant Levels - Bioengineer.org - November 15th, 2026 [November 15th, 2026]
- Data Modality Determines Machine Learning and Deep Learning Success in Breast Cancer Prediction - Bioengineer.org - November 15th, 2026 [November 15th, 2026]
- Machine Learning Identifies Varied Signal Patterns Across Diverse Tremor Interventions - Bioengineer.org - November 15th, 2026 [November 15th, 2026]
- Machine Learning Models Classify Frailty in Adults With Mild Cognitive Impairment - Bioengineer.org - November 15th, 2026 [November 15th, 2026]
- Network Features Boost Machine Learning Predictions of Food Safety Risk Severity - Bioengineer.org - November 15th, 2026 [November 15th, 2026]
- Machine Learning Links Metabolic and Inflammatory Shifts to Age-Related Hearing Loss - Bioengineer.org - November 13th, 2026 [November 13th, 2026]
- Evaluation of machine learning-based prediction of pregnancy complications and adverse birth outcomes using routine data - Nature - November 13th, 2026 [November 13th, 2026]
- Machine Learning Identifies Palmitoylation Biomarkers and Candidate Repurposed Drug in Lung Cancer - Bioengineer.org - November 13th, 2026 [November 13th, 2026]
- Machine Learning Identifies Changing Predictors of Medication Experience in Schizophrenia - Bioengineer.org - November 13th, 2026 [November 13th, 2026]
- Machine Learning Predicts Nuclear Reactor Containment Seismic Demands With Explainable Models - Bioengineer.org - November 7th, 2026 [November 7th, 2026]
- Explainable Machine Learning Identifies Predictive Thresholds Governing Urban Heat Island Dynamics - Bioengineer.org - November 7th, 2026 [November 7th, 2026]
- Machine learning reshapes yeast enzyme to favor a stubborn aldehyde and survive heat - Bioengineer.org - November 7th, 2026 [November 7th, 2026]
- Machine learning and winter storm predictions to be the focus of Nov. 6 talk - The Pennsylvania State University - November 3rd, 2026 [November 3rd, 2026]
- Zelims ZOE Taps AI, Machine Learning to Enhance S&R Operations - Marine News Magazine - November 3rd, 2026 [November 3rd, 2026]
- From Meters to Machine Learning: Building AI-Ready Utilities in India - CSIS | Center for Strategic and International Studies - October 30th, 2026 [October 30th, 2026]
- SPINE: A machine-learning program is on the trail of ghost particles - Phys.org - October 30th, 2026 [October 30th, 2026]
- Identification of ferroptosis-related key genes influencing synovitis development in osteoarthritis based on machine learning - Nature - October 30th, 2026 [October 30th, 2026]
- Machine learning simulations reveal how plastic transforms into hard carbon - EurekAlert! - October 30th, 2026 [October 30th, 2026]
- Machine learning predicts wastewater aeration efficiency, but only when data stay close to home - Bioengineer.org - October 30th, 2026 [October 30th, 2026]
- Machine learning reveals when changing shot direction pays off in elite tennis - Bioengineer.org - October 30th, 2026 [October 30th, 2026]
- Machine learning and blockchain team up to cut supply chain costs and carbon - Bioengineer.org - October 30th, 2026 [October 30th, 2026]
- Machine learning simulations uncover how plastic transforms into hard carbon - Open Access Government - October 30th, 2026 [October 30th, 2026]
- Machine Learning Surrogates Predict Wing Aerodynamics and Stress in Seconds - bioengineer.org - October 30th, 2026 [October 30th, 2026]
- Kernel-Tuned Machine Learning Pushes Network Intrusion Detection Toward 99% Accuracy on Benchmark Data - Bioengineer.org - October 30th, 2026 [October 30th, 2026]
- Machine learning and water quality indices reveal monsoon pollution hotspots in Odishas Baitarani River - bioengineer.org - October 30th, 2026 [October 30th, 2026]
- Machine Learning Predicts ICU Deaths in PAD Patients, but Hemodynamics Add Little - Bioengineer.org - October 30th, 2026 [October 30th, 2026]
- Stacked machine learning pairs ultrasound and Raman spectra to flag thyroid cancer - Bioengineer.org - October 30th, 2026 [October 30th, 2026]
- Machine Learning and Entropy Indices Reveal Most Mahanadi Water Near Paradip Unfit to Drink - Bioengineer.org - October 30th, 2026 [October 30th, 2026]
- Machine learning simulations trace how plastic turns into hard carbon atom by atom - Bioengineer.org - October 30th, 2026 [October 30th, 2026]
- My Machine Learning Pipeline Cheated, and Cross-Validation Never Noticed - Towards Data Science - October 26th, 2026 [October 26th, 2026]
- Machine learning predicts next-year fitness assessment failure in university students - Bioengineer.org - October 26th, 2026 [October 26th, 2026]
- Serrated Airfoils Reshape Stall Behavior in CFD, Wind Tunnel and Machine Learning Study - Bioengineer.org - October 26th, 2026 [October 26th, 2026]
- Applications of artificial intelligence and machine learning in food processing, quality control, and food safety: a critical review - Nature - October 26th, 2026 [October 26th, 2026]
- Machine-learning simulations reveal how sodium batteries form their protective interface - Bioengineer.org - October 26th, 2026 [October 26th, 2026]
- Reimagining demand forecasting at Bayer with AI and machine learning - Supply Chain Management Review - September 28th, 2026 [September 28th, 2026]
- What Is Machine Learning Fairness? What You Need to Know - Coursera - September 28th, 2026 [September 28th, 2026]
- Howard University Hosts Machine Learning Symposium in Collaboration with Amazon Web Services - The Dig at Howard University - September 26th, 2026 [September 26th, 2026]
- Machine learning helps identify chemicals that repel honey bees from pesticides - University of California, Riverside - September 26th, 2026 [September 26th, 2026]
- Machine Learning Hunts Down Hidden Gene Switches Across the Fungal Tree of Life - Bioengineer.org - September 26th, 2026 [September 26th, 2026]
- Integrated Quantum Technologies Advances ICA Research for IEEE Secure and Trustworthy Machine Learning Conference Submission - TMX Newsfile - September 26th, 2026 [September 26th, 2026]
- I surprisingly don't hate the Logitech Blue Yeti 2's use of machine learning - Yahoo Tech - September 26th, 2026 [September 26th, 2026]
- LANDR expands AI fraud detection built with machine learning lab Mila and says artificial streams flagged by Spotify for music it distributes fell... - September 26th, 2026 [September 26th, 2026]
- University Of California Machine Learning Spots Scents Bees Avoid In Real Fields - Quantum Zeitgeist - September 26th, 2026 [September 26th, 2026]
- Inside the AI Investing Think Tank: How Machine Learning Drives Modern Financial Research - Rebellion Research - September 26th, 2026 [September 26th, 2026]
- New machine learning methods set to improve forecasts - University of Reading - September 24th, 2026 [September 24th, 2026]
- What this machine learning model boasting 65% accuracy says is coming next for the 10-year Treasury - MarketWatch - September 24th, 2026 [September 24th, 2026]
- Quantitative Investment Strategies: Institutional Architecture and Machine Learning Alpha in 2026 - Rebellion Research - September 24th, 2026 [September 24th, 2026]
- Machine learning forecasts suggest a concentration paradox in international student mobility to the United Kingdom - Nature - September 22nd, 2026 [September 22nd, 2026]
- Machine learning model distinguishes levels of psychological resilience in health care workers with 75% accuracy - Medical Xpress - September 22nd, 2026 [September 22nd, 2026]
- How Machine Learning Is Changing the Way We Inspect the Railway - Railway-News - September 22nd, 2026 [September 22nd, 2026]
- Microsoft file patent for a machine learning system that would "train" itself on your playtime to work out when you'll be least annoyed by... - September 22nd, 2026 [September 22nd, 2026]
- Quantitative Finance Conferences NYC 2026: The AI and Machine Learning Summit Guide - Rebellion Research - September 22nd, 2026 [September 22nd, 2026]
- Nanoparticle Electrode and Machine Learning Team Up to Catch Toxic Lead and Cadmium in Water - Bioengineer.org - September 22nd, 2026 [September 22nd, 2026]
- Cosmic tensions: can machine learning find a fairer story of the universe - Times of Malta - September 22nd, 2026 [September 22nd, 2026]
- Lessening the shock of defibrillation with machine learning - AIP.ORG - September 17th, 2026 [September 17th, 2026]
- Algorithmic Stock Picking: The Architecture of Machine Learning Alpha in 2026 - Rebellion Research - September 17th, 2026 [September 17th, 2026]
- Canada is pushing ahead with machine learning to improve medical diagnostics - Digital Journal - September 17th, 2026 [September 17th, 2026]
- How researchers use machine learning to re-create chirps and trills produced by forest insects when Dinos - The Times of India - September 15th, 2026 [September 15th, 2026]
- Machine Learning Reshapes Credit Scoring - Communications of the ACM - September 13th, 2026 [September 13th, 2026]
- Machine Learning With Threshold Optimization Could Help Reduce Unnecessary Appendectomies in Adults - bioengineer.org - September 13th, 2026 [September 13th, 2026]
- Information Theory Meets Machine Learning to Catch Industrial Cyberattacks - bioengineer.org - September 13th, 2026 [September 13th, 2026]
- Machine Learning Gets a Robustness Boost by Turning Labels into Preferences - bioengineer.org - September 13th, 2026 [September 13th, 2026]
- Machine Learning Meets X-Rays to Reveal the Hidden Architecture of Pea Seeds - bioengineer.org - September 13th, 2026 [September 13th, 2026]
- Machine Learning Predicts Which Women Will Face Early Ovarian Failure Within Three Years - bioengineer.org - September 13th, 2026 [September 13th, 2026]
- Machine Learning Cracks the Code of Nitinol Wear, a Metal That Remembers Its Shape - bioengineer.org - September 13th, 2026 [September 13th, 2026]
- Using machine learning to see how living brains learn - The University of Utah - September 8th, 2026 [September 8th, 2026]
- Applying causal machine learning to assess and improve cleantech policy design - Nature - September 8th, 2026 [September 8th, 2026]
- Frontier Tech Leaders Programme Celebrates First Machine Learning Bootcamp Graduation and AI for Sustainable Tourism Hackathon in Angola - United... - September 8th, 2026 [September 8th, 2026]
- From the Knowledge to machine learning: Wayve takes AI driving to London - IOT Insider - September 8th, 2026 [September 8th, 2026]
- Algorithm optimizes machine learning techniques that use linear, tunable resistor networks - AIP.ORG - September 2nd, 2026 [September 2nd, 2026]
- Math Modeling Seminar: Applications of Topological Data Analysis and Machine Learning Models in Predictive Biology and Drug Discovery | Events | RIT -... - September 2nd, 2026 [September 2nd, 2026]
- UC Berkeley Announces New Professional Graduate Degree in AI and Machine Learning - University of California, Berkeley - August 25th, 2026 [August 25th, 2026]
- DedeepyaYarraand the rise of Trustworthy AI: Where Machine Learning meets cybersecurity - India.com - August 25th, 2026 [August 25th, 2026]
- Machine learning smooths the road from idea to real-world climate impact - EurekAlert! - August 18th, 2026 [August 18th, 2026]
- Chris Latham Interviews Henry Zelikovsky, Founder & CEO of Softlab360: Successful Applications of AI/Machine Learning in Wealth Management -... - August 18th, 2026 [August 18th, 2026]
- Integrated data and machine learning transform lung cancer diagnosis and treatment - Bioengineer.org - August 18th, 2026 [August 18th, 2026]
- Machine learning accelerates climate solutions from ideas to real-world impact - Bioengineer.org - August 18th, 2026 [August 18th, 2026]
- Identification of weight loss predictors using machine learning approaches in adolescents with obesity - Nature - August 16th, 2026 [August 16th, 2026]
- Quantitative Hedge Fund Strategies: The Machine Learning Revolution of 2026 - rebellionresearch.com - August 16th, 2026 [August 16th, 2026]
- Healthcare Machine Learning Hits Production Scale as Governance Falls Behind, Black Book's Fourth Annual Report Finds - bhpioneer.com - August 16th, 2026 [August 16th, 2026]