najoNanosystems: Physics, Chemistry, MathematicsНаносистемы: физика, химия, математика2220-80542305-7971Университет ИТМО10.17586/2220-8054-2020-11-6-642-650najo-362Research ArticlePHYSICSФИЗИКАMachine learning method for computation of optimal transitions in magnetic nanosystemsBushuevK. R.

Kronverkskiy, 49, Saint Petersburg, 197101

LobanovI. S.

Kronverkskiy, 49, Saint Petersburg, 197101

Saint Petersburg, 198504

lobanov.igor@gmail.comITMO UniversityRussian FederationITMO University; Saint Petersburg State UniversityRussian Federation202029072025116642–650Copyright © Bushuev K.R., Lobanov I.S., 20252025Bushuev K.R., Lobanov I.S.Bushuev K.R., Lobanov I.S.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://nanojournal.ifmo.ru/jour/article/view/362

Minimum energy path (MEP) is an important tool for computation of activation barriers and transition rates for magnetic systems. Recently, new methods for numeric computation of MEP were proposed based on conjugate gradient and L-BFGS methods [1] significantly improved convergence rate compared to nudged elastic band (NEB) method. Due to lack of strict mathematical theory for MEP optimization other more effective methods are expected to exist. In this article, we propose a machine learning based approach to search for MEP computation methods. We reformulate the NEB method as a differentiable transformation in the space of all paths parametrized by a family of metaparameters. Using rate of convergence as the loss function, we train NEB optimizer to find optimal metaparameters. This meta learning technique can be the basis for deriving new optimization methods for computing MEP and other non-classical optimization problems.

Transition stateminimum energy pathmachine learningmeta learningThe study of classical optimization methods for transition state computation in the sections 1 and 2 was funded by Government of the Russian Federation (Grant 08-08). The development of meta learning algorithm for nudged elastic band method in the section 3 was funded by Russian Science Foundation (Grant 19-42-06302).ReferencesIvanov A.V., Dagbartsson D., Tranchida J., Uzdin V.M., Jnsson H. Efficient optimization method for finding minimum energy paths of magnetic transitions. Journal of Physics: Condensed Matter, 2020, 32(34).Ivanov A.V., Dagbartsson D., Tranchida J., Uzdin V.M., Jnsson H. Efficient optimization method for finding minimum energy paths of magnetic transitions. Journal of Physics: Condensed Matter, 2020, 32(34).Everschor-Sitte K., Masell J.,Reeve R.M., Klaui M. Perspective: Magnetic skyrmions - Overview of recent progress in an active research field. J. Appl. Phys., 2018, 124, P. 240901.Everschor-Sitte K., Masell J.,Reeve R.M., Klaui M. Perspective: Magnetic skyrmions - Overview of recent progress in an active research field. J. Appl. Phys., 2018, 124, P. 240901.Mittal S. A survey of techniques for architecting processor components using domain-wall memory. ACM Journal on Emerging Technologies in Computing Systems (JETC), 2016, 13(2), P. 1–25.Mittal S. A survey of techniques for architecting processor components using domain-wall memory. ACM Journal on Emerging Technologies in Computing Systems (JETC), 2016, 13(2), P. 1–25.Parkin S.S.P., Hayashi M., Thomas L. Magnetic domain-wall racetrack memory. Science, 2008, 320(5873), P. 190–194.Parkin S.S.P., Hayashi M., Thomas L. Magnetic domain-wall racetrack memory. Science, 2008, 320(5873), P. 190–194.Fert A., Cros V., Sampaio J. Skyrmions on the track. Nature Nanotech, 2013, 8, P. 152–156.Fert A., Cros V., Sampaio J. Skyrmions on the track. Nature Nanotech, 2013, 8, P. 152–156.Schratzberger J., Lee J., Fuger M., Fidler J., Fiedler G., Schref T., Suess D. Validation of the transition state theory with Langevin-dynamics simulations. Journal of Applied Physics, 2010, 108, P. 033915.Schratzberger J., Lee J., Fuger M., Fidler J., Fiedler G., Schref T., Suess D. Validation of the transition state theory with Langevin-dynamics simulations. Journal of Applied Physics, 2010, 108, P. 033915.Desplat L., Suess D., Kim J-V., Stamps R.L. Thermal stability of metastable magnetic skyrmions: Entropic narrowing and significance of internal eigenmodes. Phys. Rev. B, 2018, 98, P. 134407.Desplat L., Suess D., Kim J-V., Stamps R.L. Thermal stability of metastable magnetic skyrmions: Entropic narrowing and significance of internal eigenmodes. Phys. Rev. B, 2018, 98, P. 134407.Bessarab P.F., Mller G.P., Lobanov I.S. et. al. Lifetime of racetrack skyrmions. Sci. Rep., 2018, 8, P. 3433.Bessarab P.F., Mller G.P., Lobanov I.S. et. al. Lifetime of racetrack skyrmions. Sci. Rep., 2018, 8, P. 3433.Potkina M.N., Lobanov I.S., Jnsson H., Uzdin V.M. Skyrmions in antiferromagnets: Thermal stability and the effect of external field and impurities. Journal of Applied Physics, 2020, 127, P. 213906.Potkina M.N., Lobanov I.S., Jnsson H., Uzdin V.M. Skyrmions in antiferromagnets: Thermal stability and the effect of external field and impurities. Journal of Applied Physics, 2020, 127, P. 213906.Uzdin V.M., Potkina M.N., Lobanov I.S., Bessarab P.F., Jnsson H. The effect of confinement and defects on the thermal stability of skyrmions. Physica B: Condensed Matter, 2018, 549, P. 6–9.Uzdin V.M., Potkina M.N., Lobanov I.S., Bessarab P.F., Jnsson H. The effect of confinement and defects on the thermal stability of skyrmions. Physica B: Condensed Matter, 2018, 549, P. 6–9.Lobanov I.S., Jnsson H., Uzdin V.M. Mechanism and activation energy of magnetic skyrmion annihilation obtained from minimum energy path calculations. Phys. Rev. B., 2016, 94, P. 174418-2016.Lobanov I.S., Jnsson H., Uzdin V.M. Mechanism and activation energy of magnetic skyrmion annihilation obtained from minimum energy path calculations. Phys. Rev. B., 2016, 94, P. 174418-2016.Bessarab P.F., Uzdin V.M., Jnsson H. Method for finding mechanism and activation energy of magnetic transitions, applied to skyrmion and antivortex annihilation. Comp. Phys. Commun., 2015, 196, P. 335–347.Bessarab P.F., Uzdin V.M., Jnsson H. Method for finding mechanism and activation energy of magnetic transitions, applied to skyrmion and antivortex annihilation. Comp. Phys. Commun., 2015, 196, P. 335–347.Bessarab P.F., Uzdin V.M., and Jnsson H. Harmonic transition-state theory of thermal spin transitions. Phys. Rev. B, 2012, 85, P. 184409.Bessarab P.F., Uzdin V.M., and Jnsson H. Harmonic transition-state theory of thermal spin transitions. Phys. Rev. B, 2012, 85, P. 184409.Langer J.S. Statistical theory of the decay of metastable states. Annals of Physics, 1969, 54(2), P. 258–275.Langer J.S. Statistical theory of the decay of metastable states. Annals of Physics, 1969, 54(2), P. 258–275.Liashko S.Y., Lobanov I.S., Uzdin V.M., Jnsson H. Thermal stability of magnetic states in submicron magnetic islands. Nanosystems: Physics, Chemistry, Mathematics, 2017, 8(5), P. 572–578.Liashko S.Y., Lobanov I.S., Uzdin V.M., Jnsson H. Thermal stability of magnetic states in submicron magnetic islands. Nanosystems: Physics, Chemistry, Mathematics, 2017, 8(5), P. 572–578.Lobanov I.S., Uzdin V.M. The lifetime of big size topological chiral magnetic states. Estimation of the pre-exponential factor in the Arrhenius law. 2020. arXiv:2008.06754Lobanov I.S., Uzdin V.M. The lifetime of big size topological chiral magnetic states. Estimation of the pre-exponential factor in the Arrhenius law. 2020. arXiv:2008.06754Henkelman G., Jnsson H. A dimer method for finding saddle points on high dimensional potential surfaces using only first derivatives. J. Chem. Phys., 1999, 111(15), P. 7010–7022.Henkelman G., Jnsson H. A dimer method for finding saddle points on high dimensional potential surfaces using only first derivatives. J. Chem. Phys., 1999, 111(15), P. 7010–7022.Olsen R.A., Kroes G.J., Henkelman G., Arnaldsson A., Jnsson H. Comparison of methods for finding saddle points without knowledge of the final states. J. Chem. Phys., 2004, 121(20), P. 9776–9792.Olsen R.A., Kroes G.J., Henkelman G., Arnaldsson A., Jnsson H. Comparison of methods for finding saddle points without knowledge of the final states. J. Chem. Phys., 2004, 121(20), P. 9776–9792.Gutierrez M.P., Argaez C., Jnsson H. Improved minimum mode following method for finding first order saddle points. J. Chem. Theo. Comput., 2017, 13(1), P. 125–134.Gutierrez M.P., Argaez C., Jnsson H. Improved minimum mode following method for finding first order saddle points. J. Chem. Theo. Comput., 2017, 13(1), P. 125–134.Henkelman G., Johannesson G., Jnsson H. Methods for finding saddle points and minimum energy paths. Theoretical Methods in Condensed Phase Chemistry, 2002, 5, P. 269–302.Henkelman G., Johannesson G., Jnsson H. Methods for finding saddle points and minimum energy paths. Theoretical Methods in Condensed Phase Chemistry, 2002, 5, P. 269–302.Henkelman G., Uberuaga B.P., Jnsson H. A climbing image nudged elastic band method for finding saddle points and minimum energy paths. J. Chem. Phys., 2000, 113(22), P. 9901–9904.Henkelman G., Uberuaga B.P., Jnsson H. A climbing image nudged elastic band method for finding saddle points and minimum energy paths. J. Chem. Phys., 2000, 113(22), P. 9901–9904.Maragakis P., Andreev S.A., Brumer Y., Reichman D.R., Kaxiras E. Adaptive nudged elastic band approach for transition state calculation. J. Chem. Phys., 2002, 117, P. 4651–4658.Maragakis P., Andreev S.A., Brumer Y., Reichman D.R., Kaxiras E. Adaptive nudged elastic band approach for transition state calculation. J. Chem. Phys., 2002, 117, P. 4651–4658.Hoffmann M.,Mller G.P., Blgel S. Atomistic Perspective of Long Lifetimes of Small Skyrmions at Room Temperature. Phys. Rev. Lett., 2020, 124, P. 247201.Hoffmann M.,Mller G.P., Blgel S. Atomistic Perspective of Long Lifetimes of Small Skyrmions at Room Temperature. Phys. Rev. Lett., 2020, 124, P. 247201.Henkelman G., Jnsson H. Improved tangent estimate in the nudged elastic band method for finding minimum energy paths and saddle points. J. Chem. Phys., 2000, 113(22), P. 9978–9985.Henkelman G., Jnsson H. Improved tangent estimate in the nudged elastic band method for finding minimum energy paths and saddle points. J. Chem. Phys., 2000, 113(22), P. 9978–9985.Weinan E., Weiqing R., Vanden-Eijnden E. String method for the study of rare events. Phys. Rev. B., 2002, 66, P. 052301(4).Weinan E., Weiqing R., Vanden-Eijnden E. String method for the study of rare events. Phys. Rev. B., 2002, 66, P. 052301(4).Sheppard D., Terrell R., Henkelman G. Optimization methods for finding minimum energy paths. J. Chem. Phys., 2008, 128, P. 134106(10).Sheppard D., Terrell R., Henkelman G. Optimization methods for finding minimum energy paths. J. Chem. Phys., 2008, 128, P. 134106(10).Heistracher P., Abert C., Bruckner F., Vogler C., Suess D., GPU-Accelerated Atomistic Energy Barrier Calculations of Skyrmion Annihilations. IEEE Transactions on Magnetics, 2018, 54(11), P. 1-5.Heistracher P., Abert C., Bruckner F., Vogler C., Suess D., GPU-Accelerated Atomistic Energy Barrier Calculations of Skyrmion Annihilations. IEEE Transactions on Magnetics, 2018, 54(11), P. 1-5.Schmidhuber J. Evolutionary Principles in Self-Referential Learning. On Learning how to Learn: The Meta-Meta-Meta...-Hook. PhD thesis, Institut f. Informatik, Tech. Univ. Munich, 1987.Schmidhuber J. Evolutionary Principles in Self-Referential Learning. On Learning how to Learn: The Meta-Meta-Meta...-Hook. PhD thesis, Institut f. Informatik, Tech. Univ. Munich, 1987.Naik D.K., Mammone R.J. Meta-neural networks that learn by learning. In International Joint Conference on Neural Networks, IEEE, 1992, 1, P. 437–442.Naik D.K., Mammone R.J. Meta-neural networks that learn by learning. In International Joint Conference on Neural Networks, IEEE, 1992, 1, P. 437–442.Bengio S., Bengio Y., Cloutier J. On the search for new learning rules for ANNs. Neural Processing Letters, 1995, 2(4), P. 26–30.Bengio S., Bengio Y., Cloutier J. On the search for new learning rules for ANNs. Neural Processing Letters, 1995, 2(4), P. 26–30.Schmidhuber J., Zhao J., Wiering M. Shifting Inductive Bias with Success-Story Algorithm, Adaptive Levin Search, and Incremental SelfImprovement. Machine Learning, 1997, 28, P. 105–130.Schmidhuber J., Zhao J., Wiering M. Shifting Inductive Bias with Success-Story Algorithm, Adaptive Levin Search, and Incremental SelfImprovement. Machine Learning, 1997, 28, P. 105–130.Hochreiter S., Schmidhuber J. Long Short-Term Memory. Neural Computation, 1997.Hochreiter S., Schmidhuber J. Long Short-Term Memory. Neural Computation, 1997.Thrun S., Pratt L. Learning to learn. Springer Science and Business Media, 1998.Thrun S., Pratt L. Learning to learn. Springer Science and Business Media, 1998.Younger A.S., Conwell P.R., Cotter N.E. Fixed-weight on-line learning. Transactions on Neural Networks, 1999, 10(2), P. 272–283.Younger A.S., Conwell P.R., Cotter N.E. Fixed-weight on-line learning. Transactions on Neural Networks, 1999, 10(2), P. 272–283.Runarsson T.P., Jonsson M.T. Evolution and design of distributed learning rules. In IEEE Symposium on Combinations of Evolutionary Computation and Neural Networks, 2000, P. 59–63.Runarsson T.P., Jonsson M.T. Evolution and design of distributed learning rules. In IEEE Symposium on Combinations of Evolutionary Computation and Neural Networks, 2000, P. 59–63.Diederik P. Kingma, Jimmy Ba. Adam: A Method for Stochastic Optimization. 3rd International Conference for Learning Representations, San Diego, 2015.Diederik P. Kingma, Jimmy Ba. Adam: A Method for Stochastic Optimization. 3rd International Conference for Learning Representations, San Diego, 2015.Andrychowicz M., Denil M., Gomez S., Hoffman M.W., Pfau D., Schaul T., Shillingford B., de Freitas N. Learning to learn by gradient descent by gradient descent. 2016, arXiv:1606.04474.Andrychowicz M., Denil M., Gomez S., Hoffman M.W., Pfau D., Schaul T., Shillingford B., de Freitas N. Learning to learn by gradient descent by gradient descent. 2016, arXiv:1606.04474.Chen Y., Hoffman M.W., Gomez S.C., Denil M., Lillicrap T.P., Botvinick M., de Freitas N. Learning to Learn without Gradient Descent by Gradient Descent. 2016, arXiv:1611.03824.Chen Y., Hoffman M.W., Gomez S.C., Denil M., Lillicrap T.P., Botvinick M., de Freitas N. Learning to Learn without Gradient Descent by Gradient Descent. 2016, arXiv:1611.03824.Wichrowska O., Maheswaranathan N., Hoffman M.W., Gomez S.C., Denil M., de Freitas N., Sohl-Dickstein J. Learned Optimizers that Scale and Generalize. 2017, arXiv:1703.04813.Wichrowska O., Maheswaranathan N., Hoffman M.W., Gomez S.C., Denil M., de Freitas N., Sohl-Dickstein J. Learned Optimizers that Scale and Generalize. 2017, arXiv:1703.04813.Hospedales T., Antoniou A., Micaelli P., Storkey A. Meta-Learning in Neural Networks: A Survey. 2020, arXiv:2004.05439.Hospedales T., Antoniou A., Micaelli P., Storkey A. Meta-Learning in Neural Networks: A Survey. 2020, arXiv:2004.05439.Carlon A., Espath L., Lopez R., Tempone R. Multi-iteration stochastic optimizers. 2020, arXiv:2011.01718Carlon A., Espath L., Lopez R., Tempone R. Multi-iteration stochastic optimizers. 2020, arXiv:2011.01718Khromova K. Optimization of the neural network training method. Master dissertation. ITMO University, 2020.Khromova K. Optimization of the neural network training method. Master dissertation. ITMO University, 2020.Rybakov F.N., Borisov A.B., Blgel S., Kiselev N.S. New Type of Stable Particlelike States in Chiral Magnets. Phys. Rev. Lett., 2015, 115, P. 117201.Rybakov F.N., Borisov A.B., Blgel S., Kiselev N.S. New Type of Stable Particlelike States in Chiral Magnets. Phys. Rev. Lett., 2015, 115, P. 117201.Mentink J.H., Tretyakov M.V., Fasolino A., Katsnelson M.I., Rasing Th. Stable and fast semi-implicit integration of the stochastic LandauLifshitz equation. Journal of Physics: Condensed Matter, 2010, 22(17), P. 176001.Mentink J.H., Tretyakov M.V., Fasolino A., Katsnelson M.I., Rasing Th. Stable and fast semi-implicit integration of the stochastic LandauLifshitz equation. Journal of Physics: Condensed Matter, 2010, 22(17), P. 176001.Desplat L., Vogler C., Kim J.-V., Stamps R. L., Suess D. Path sampling for lifetimes of metastable magnetic skyrmions and direct comparison with Kramers’ method. Phys. Rev. B., 2020, 101, P. 060403(R).Desplat L., Vogler C., Kim J.-V., Stamps R. L., Suess D. Path sampling for lifetimes of metastable magnetic skyrmions and direct comparison with Kramers’ method. Phys. Rev. B., 2020, 101, P. 060403(R).Moskalenko M.A., Lobanov I.S., Uzdin V.M. Demagnetizing fields in chiral magnetic structures. Nanosystems: Physics, Chemistry, Mathematics, 2020, 11(4), P. 401–407.Moskalenko M.A., Lobanov I.S., Uzdin V.M. Demagnetizing fields in chiral magnetic structures. Nanosystems: Physics, Chemistry, Mathematics, 2020, 11(4), P. 401–407.Jonsson H., Mills G., Jacobsen K.W., Berne B.J., Ciccotti G., Coker D.F. Nudged elastic band method for finding minimum energy paths of transitions, in Classical and Quantum Dynamics in Condensed Phase Simulations. World Scientific, Singapore, 1998, P. 385–404.Jonsson H., Mills G., Jacobsen K.W., Berne B.J., Ciccotti G., Coker D.F. Nudged elastic band method for finding minimum energy paths of transitions, in Classical and Quantum Dynamics in Condensed Phase Simulations. World Scientific, Singapore, 1998, P. 385–404.Henkelman G., Arnaldsson A., Jonsson H. Theoretical calculations of CH4 and H2 associative desorption from Ni(111): Could subsurface hydrogen play an important role? J. Chem. Phys., 2006, 124, P. 044706(9).Henkelman G., Arnaldsson A., Jonsson H. Theoretical calculations of CH4 and H2 associative desorption from Ni(111): Could subsurface hydrogen play an important role? J. Chem. Phys., 2006, 124, P. 044706(9).Einarsdottir D.M., Arnaldsson A., Oskarsson F., Jonsson H. Path optimization with application to tunneling. Lecture Notes in Computer Science, 2012, 7134, P. 45–55.Einarsdottir D.M., Arnaldsson A., Oskarsson F., Jonsson H. Path optimization with application to tunneling. Lecture Notes in Computer Science, 2012, 7134, P. 45–55.Malottki S.V., Dupe B., Bessarab P.F., Delin A., Heinze S. Enhanced skyrmion stability due to exchange frustration. Sci. Rep., 2017, 7(10), P. 12299.Malottki S.V., Dupe B., Bessarab P.F., Delin A., Heinze S. Enhanced skyrmion stability due to exchange frustration. Sci. Rep., 2017, 7(10), P. 12299.Lobanov I.S., Potkina M.N., Jnsson H., Uzdin V.M. Truncated minimum energy path method for finding first order saddle points. Nanosystems: Physics, Chemistry, Mathematics, 2017, 8(5), P. 586–595Lobanov I.S., Potkina M.N., Jnsson H., Uzdin V.M. Truncated minimum energy path method for finding first order saddle points. Nanosystems: Physics, Chemistry, Mathematics, 2017, 8(5), P. 586–595Bradbury J., Frostig R., Hawkins P., Johnson M.J., Leary C., Maclaurin D., Necula G., Paszke A., Vanderlas J., Wanderman-Milne S., Zhang Q. JAX: composable transformations of Python+NumPy programs. (http://github.com/google/jax), 2018.Bradbury J., Frostig R., Hawkins P., Johnson M.J., Leary C., Maclaurin D., Necula G., Paszke A., Vanderlas J., Wanderman-Milne S., Zhang Q. JAX: composable transformations of Python+NumPy programs. (http://github.com/google/jax), 2018.Mills G., Jonsson H., Schenter G.K. Reversible work based transition state theory: Application to H2 dissociative adsorption. Surf. Sci., 1995, 324, P. 305–337.Mills G., Jonsson H., Schenter G.K. Reversible work based transition state theory: Application to H2 dissociative adsorption. Surf. Sci., 1995, 324, P. 305–337.Zhu T., Li J., Samanta A., Kim H.G., Suresh S. Interfacial plasticity governs strain rate sensitivity and ductility in nanostructured metals. PNAS, 2007, 104(9), P. 3031–3036.Zhu T., Li J., Samanta A., Kim H.G., Suresh S. Interfacial plasticity governs strain rate sensitivity and ductility in nanostructured metals. PNAS, 2007, 104(9), P. 3031–3036.Fakoor M., Kosari A., Jafarzadeh M. Revision on fuzzy artificial potential field for humanoid robot path planning in unknown environment. International Journal of Advanced Mechatronic Systems (IJAMECHS), 2015, 6(4), P. 174–183Fakoor M., Kosari A., Jafarzadeh M. Revision on fuzzy artificial potential field for humanoid robot path planning in unknown environment. International Journal of Advanced Mechatronic Systems (IJAMECHS), 2015, 6(4), P. 174–183

The authors declare that there are no conflicts of interest present.