TB2J: a python package for computing magnetic interaction parameters

arXiv:2009.01910 · cond-mat.mtrl-sci · Submitted 2020-09-03 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.

Kai: We're looking at the paper: "TB2J: a python package for computing magnetic interaction parameters".

Mira: Here's the gist of the paper. TB2J presents a Python package for automatically computing magnetic interaction parameters like exchange and Dzyaloshinskii-Moriya between atoms in magnetic crystals from density functional theory results It is crucial for fitting effective Hamiltonians to DFT data to enable simulations of larger systems and dynamics The core idea involves perturbing a localized spin in both the Heisenberg spin model and the DFT electronic model The Green’s function method is employed by treating local rigid spin rotation as a perturbation within the single-particle Green’s function method This approach allows determining Heisenberg magnetic interaction parameters from the ground-state solution regardless of whether it is FM or AFM The method also enables access to a bandby-band decomposition of different magnetic interactions Relativistic effects, including spin-orbit coupling, are taken into account to calculate the Dzyaloshinskii-Moriya interaction (DMI) and magnetic anisotropy Heisenberg Model Formulation The Heisenberg Hamiltonian is expressed as a sum of four parts: single-ion anisotropy, isotropic exchange, symmetric anisotropic exchange, and DMI terms The variation of the total energy with respect to spin vectors yields expressions for the interaction parameters The isotropic exchange parameter is given by J iso i j = Im(A 0 0 i j −A xx i j −A yy i j −A zz i j) Implementation and Workflow The TB2J package is structured with three main modules: the tight-binding (TB) module, the exchange module, and the output module The TB module interfaces with external DFT and Wannier function codes to calculate the Green’s function For LCAO codes like Siesta or GPAW, the Hamiltonian is read directly, and Green’s functions are calculated using eigenvalues and eigenvectors rather than inverting the matrix for every energy Application and Validation TB2J is applied to several crystals including BCC Fe, HCP Co bulk, SrMnO3, BiFeO3, and La2CuO4 to demonstrate its capabilities For BCC Fe, the isotropic exchange parameters are compared with results from the TB-LMTO method and KKR methods In HCP Co, the DMI parameters are compared with results from the KKR method The calculation of DMI and anisotropic exchange components requires an xyz average strategy to obtain the full set of parameters Advantages and Limitations The Green’s function method is more robust than total-energy mapping as it only requires a DFT calculation for a single spin configuration, avoiding convergence issues with unstable magnetic structures A limitation is that the calculation of DMI and anisotropic exchange parameters through the Wannier basis set might still be problematic for very small quantities due to noise introduced during the Wannierization process The Heisenberg model might not be valid when spins are itinerant or when non-bilinear magnetic interactions are nonnegligible TB2J provides a minimal input requirement, allowing easy integration into high-throughput workflows for automatic calculations of magnetic interactions The package is designed to be used either with Wannier functions or LCAO codes, making it suitable for automated high-throughput calculations Future versions plan to include single ion anisotropy and higher order parameters like three-spin and four-spin interactions The ability to calculate exchange magnetic interaction parameters at any distance in real space is a key feature User Instructions

Paper summary: Kai: To recap, we've looked at TB2J, which is this Python package presenting a way to automatically compute magnetic interaction parameters like exchange and Dzyaloshinskii-Moriya between atoms from density functional theory results.

Mira: The core thesis is that they use the Green’s function method by treating local rigid spin rotation as a perturbation within the single-particle Green’s function method, which lets them determine Heisenberg magnetic interaction parameters from the ground state solution regardless of whether it is ferromagnetic or antiferromagnetic.

Lev: They also highlight that this specific approach allows access to a band-by-band decomposition of these different magnetic interactions, which is useful because it separates what's happening at different energy levels.

Kai: And they didn't just stop at exchange; they incorporated relativistic effects like spin-orbit coupling to calculate the Dzyaloshinskii-Moriya interaction and magnetic anisotropy.

Mira: The input for the package comes from either Wannier90 codes that interface with many density functional theory packages, or codes built on localized orbitals.

Lev: What matters for us is how they handle the Hamiltonian formulation; they show a conversion where the exchange parameter J iso i j is given by taking the imaginary part of a certain combination of terms involving A zero zero i j and A xx i j and A yy i j and A zz i j.

Kai: That specific expression gives you the isotropic exchange parameter directly from those DFT results, which is pretty neat because it’s accessible from constrained density functional theory calculations too.

Mira: They also point out that the single-ion anisotropy parameter, which is part of the total energy mapping approach, isn't as easy to get from Wannier90 because you need to separately calculate the spin-orbit coupling part of the Hamiltonian for that.

Lev: So they are showing a way around those specific problems by using this perturbation scheme, avoiding convergence issues with unstable magnetic structures and relying on the ground state solution instead of trying to map total energy across an entire phase space.

Conclusion: Kai: So, TB2J is this Python package that lets you compute exchange and Dzyaloshinskii-Moriya parameters automatically from DFT results using the Green’s function method with local spin rotation as a perturbation.

Mira: It's important because it gives us a minimal input requirement, allowing for easy integration into high throughput workflows, whether you use Wannier functions or LCAO codes.

Lev: The implication is that we can get these parameters at any distance in real space, which is a key feature when trying to build effective Hamiltonians for large systems and dynamics simulations.

Kai: The authors applied it to several crystals like BCC Iron and HCP Cobalt bulk, comparing their isotropic exchange parameters against results from TB-LMTO and KKR methods for that BCC Iron system.

Mira: And they also showed how the DMI parameters can be calculated by taking an xyz average strategy to get the full set of components needed for anisotropic exchange.

Lev: The main thing here is that it provides a way to bridge the gap between large-scale DFT calculations and the effective Hamiltonians we need for simulating real magnetic materials.

Xu Hea, Nicole Helbig, Matthieu J. Verstraeteb, Eric Bousqueta

Physique Theorique des Mat é ris, Q-MAT, CESAM, University of Liège · Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC · Nanomat, Q-Mat, CESAM, and European Theoretical Spectroscopy Facility

cond-mat.mtrl-sci

Submitted: 2020-09-03

Updated: 2020-09-07

Journal ref: Computer Physics Communications 264 107938 (2021)

DOI: 10.1016/j.cpc.2021.107938

Code: https://github.com/mailhexu/TB2J

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 56/100

The gist: TB2J presents a Python package for automatically computing magnetic interaction parameters, including exchange and Dzyaloshinskii-Moriya interactions…

Key concepts

Green’s function method
This technique treats a localized spin's rotation as a small perturbation within the single-particle Green’s function framework. It allows researchers to determine Heisenberg magnetic interaction parameters from the ground-state solution, regardless of whether the material is ferromagnetic or antiferromagnetic. This approach is robust for finding interaction details.
Heisenberg Hamiltonian
This mathematical model describes a crystal's magnetic behavior using terms like single-ion anisotropy, isotropic exchange, symmetric anisotropic exchange, and Dzyaloshinskii-Moriya (DMI) terms. By calculating how the total energy changes with respect to spin vectors, the package extracts specific interaction parameters like J iso i j.
Tight-binding (TB) module
This component of TB2J interfaces with external DFT and Wannier function codes to calculate Green’s functions. For LCAO codes, it reads the Hamiltonian directly and uses eigenvalues and eigenvectors to find Green's functions, which is more efficient than inverting the matrix for every energy level.

Terminology

Summary

TB2J presents a Python package for automatically computing magnetic interaction parameters, including exchange and Dzyaloshinskii-Moriya interactions, from density functional theory calculations The program is based on the Green’s function method with local rigid spin rotation treated as a perturbation

Methodology Overview

The core idea involves perturbing a localized spin in both the Heisenberg spin model and the DFT electronic model For the DFT model, this is done via the rigid spin-rotation perturbation within the single-particle Green’s function method The Heisenberg parameters are then mapped to electron expressions using Eq. (1) The Green’s function in reciprocal space is defined by Eq. p. 2 (5) Real-space Green’s functions are obtained via Eq. p. 2 (6)

Hamiltonian Representation

The Heisenberg Hamiltonian consists of four parts: single-ion anisotropy, isotropic exchange, symmetric anisotropic exchange, and DMI The variation of the total energy with respect to the spin vectors yields Eq. (2) The conversion of Ji j to other conventions is detailed in Table 1

Implementation Details

The TB2J package is structured into three modules: the tight-binding (TB) module, the exchange module, and the output module The TB module interfaces with external DFT and Wannier function codes to calculate Green’s functions The exchange module then uses these Green’s functions to compute the magnetic interaction parameters using Eqs. (15)-(17)

User Workflow

The user workflow begins with a converged DFT calculation, either collinear or non-collinear, to obtain the electronic structure The next step involves constructing the tight-binding Hamiltonian using Wannier functions or LCAO results The package offers two main executables, wann2J.py and siesta2J. p. 6py, for calculating J from Wannier90 and Siesta output respectively

Results and Applications

The package calculates isotropic exchange, anisotropic exchanges, and DMI from the output of Wannier90 or LCAO Hamiltonian results The method allows for the calculation of magnon band structures via the Fourier transform of J to ~q space Examples include BCC Fe, HCP Co, SrMnO3, BiFeO3, and La2CuO4 crystals

Advantages and Limitations

The Green’s function method is robust because it only requires a DFT calculation for a single spin configuration, avoiding convergence issues with high-energy or unstable magnetic structures However, the calculation of DMI and anisotropic exchange using Wannier basis sets can be problematic due to noise of the order of a few µeV Furthermore, the Heisenberg model might not be valid when spins are itinerant or non-bilinear magnetic interactions are significant

Code Availability

The code is freely available under the BSD 2 clause license and can be found at https://github.com/mailhexu/TB2J The interface to OpenMX is in a separate package, TB2J OpenMX, under the GPLv3 license The documentation is provided at https://tb2j. p. 5readthedocs. p. 5io/en/latest/

Conclusion

TB2J provides a systematic way to calculate magnetic interaction parameters from Wannier and LCAO Hamiltonians using a Green’s function method It requires minimal user input, making it suitable for high-throughput workflows Future versions plan to include higher order parameters like three-spin and four-spin interactions The package can be easily integrated into spin dynamics and Monte Carlo codes

Acknowledgements

The authors thank Yajun Zhang, Alireza Sasani, Jorge Pilo Gonzlez, and Zachary Romestan for the testing of the code This work has been funded by the Communaute Franc¸aise de Belgique (ARC AIMED G.A. 15/19-09) XH thanks the support by the EU H2020-NMBP-TO-IND-2018 project ”INTERSECT” (Grant No. p. 7 814487) EB thanks the FRS-FNRS for support, as does MJV for an “out” sabbatical grant to ICN2 Barcelona in 2018-2019 The authors acknowledge the CECI supercomputer facilities funded by the F. p. 0R.S-FNRS (Grant No. 2.5020.1) and the Tier-1 supercomputer of the Wallonie-Bruxelles funded by the Walloon Region (Grant No. 1117545) Computing time was also provided by PRACE-3IP DECI grants 2DSpin and Pylight on Beskow (G.A. 653838 of H2020)

References

[1] Hirohata, A. et al., Journal of Magnetism and Magnetic Materials 509 (2020) 166711 [2] Baltz, V. p. 0 et al., Rev. Mod. Phys. 90 (2018) 015005 [3] Bhatti, S. et al., Materials Today 20 (2017) 530 [4] Joshi, V. p. 0 K., Engineering Science and Technology, an International Journal 19 (2016) 1503 [5] Lu, J. W., Chen, E., Kabir, M., Stan, M. R., and Wolf, S. A., International Materials Reviews 61 (2016) 456 [6] Zuti ˇ c, I. p. 0, Fabian, J., and Das Sarma, S., Rev. Mod. Phys. ´ 76 (2004) 323 [7] Fazekas, P., Lecture notes on electron correlation and magnetism, Series in modern condensed matter physics 5, World Scientific, 1999 [8] Chumak, A. V., Vasyuchka, V. I., Serga, A. A., and Hillebrands, B., Nat. Phys. 11 (2015) 453 [9] Kiselev, N. S., Bogdanov, A. N., Schfer, R., and Rßler, U. K., Journal of Physics D: Applied Physics 44 (2011) 392001 [10] Fert, A., Cros, V., and Sampaio, J., Nat. Nanotechnol. 8 (2013) 152 [14] Savrasov, S. p. 1 Y., Phys. Rev. Lett. 81 (1998) 2570 [15] Romero, A. H. et al., The Journal of Chemical Physics 152 (2020) 124102 [16] Cao, K. p. 0, Lambert, H., Radaelli, P. G., and Giustino, F., Phys. Rev. B 97 (2018) 024420 [17] Ricci, F., Prokhorenko, S., Torrent, M., Verstraete, M. J., and Bousquet, E., Phys. Rev. B 99 (2019) 184404 [18] Phillips, J. J. and Peralta, J. E., The Journal of Chemical Physics 138 (2013) 174115 [19] Liechtenstein, A. p. 1 I., Katsnelson, M. p. 1 p. 1, Antropov, V. p. 1, and Gubanov, V. p. 1, Journal of Magnetism and Magnetic Materials 67 (1987) 65 [20] Steenbock, T. p. 0 and Herrmann, C., Journal of Computational Chemistry 39 (2018) 81 [21] Katsnelson, M. p. 1 I. and Lichtenstein, A. I., Phys. Rev. B 61 (2000) 8906 [24] Antropov, V. p. 1, Katsnelson, M., and Liechtenstein, A. p. 1, Physica B: Condensed Matter 237 (1997) 336 [25] Mazurenko, V. p. 3 V. and Anisimov, V. I., Phys. Rev. B 71 (2005) 184434 [26] Mankovsky, S. and Ebert, H., Phys. Rev. B 96 (2017) 104416 [27] Secchi, A., Lichtenstein, A. I., and Katsnelson, M. p. 1 p. 1 I., Annals of Physics 360 (2015) 61 [28] Secchi, A., Lichtenstein, A. I., and Katsnelson, M. I., Journal of Magnetism and Magnetic Materials 400 (2016) 112 [30] Szilva, A. p. 0 et al., Phys. Rev. Lett. 111 (2013) 127204 [34] Korotin, D. p. 1 M., Mazurenko, V. V., Anisimov, V. I., and Streltsov, S. V., Phys. Rev. B 91 (2015) 224405 [37] Gonze, X. et al., Comput. Phys. Commun. 180 (2009) 2582 [38] Gonze, X. p. 0 et al., Comput. Phys. Commun. 248 (2020) 107042 [39] Giannozzi, P. p. 0 et al., Journal of Physics: Condensed Matter 29 (2017) 465901 [40] Soler, J. p. 3 M. et al., Journal of Physics: Condensed Matter 14 (2002) 2745 [41] Kresse, G. p. 3 and Furthmuller, J., Phys. Rev. B ¨ 54 (1996) 11169 [42] Kresse, G. and Joubert, D., Phys. Rev. B 59 (1999) 1758 [43] Blaha, P., Schwarz, K., Madsen, G. K., Kvasnicka, D., and Luitz, J., An augmented plane wave+ local orbitals program for calculating crystal properties (2001) [44] Blugel, S. p. 0 and Bihlmayer, G., Forschungszentrum J ¨ ulich GmbH ¨ (2006) 85 [45] Andrade, X. et al., Physical Chemistry Chemical Physics 17 (2015) 31371 [46] Tancogne-Dejean, N. p. 4 et al., The Journal of Chemical Physics 152 (2020) 124119 [47] Ozaki, T. p. 0 and Kino, H., Physical Review B 72 (2005) 045121 [48] Enkovaara, J. p. 4 et al., Journal of physics: Condensed matter 22 (2010) 253202 [49] The Elk Code, http://elk. p. 1sourceforge.net/ [50] Oroszlany, L., Ferrer, J., De ´ ak, A., Udvardi, L., and Szunyogh, ´ L., Phys. Rev. B 99 (2019) 224412 [51] Yoon, H., Kim, T. J., Sim, J.-H., and Han, M. J., Comput. Phys. Commun. 247 (2020) 106927 [52] Papior, N. p. 0 R., B., J. L., Frederiksen, T., and Wuhl, S. S., zerothi/sisl: v0. p. 19.8, 2020 [53] Solovyev, I. p. 0, Dederichs, P., and Mertig, I., Physical Review B 52 (1995) 13419 [54] Weingart, C. p. 4, Spaldin, N., and Bousquet, E., Phys. Rev. B 86 (2012) 094413 [56] Marzari, N. and Vanderbilt, D., Phys. Rev. B 56 (1997) 12847 [57] Wang, R., Lazar, E. A., Park, H., Millis, A. J., and Marianetti, C. A., Phys. Rev. B 90 (2014) 165125 [59] Evans, R. F. et al., Journal of Physics: Condensed Matter 26 (2014) 103202 [60] Di Gennaro, M. p. 3, Miranda, A. L., Ostler, T. p. 9 A., Romero, A. H., and Verstraete, M. J., Phys. Rev. B 97 (2018) 214417 [61] Catalan, G. and Scott, J. F., Adv. Matter. 21 (2009) 2463 [62] Pajda, M. p. 0, Kudrnovsky, J., Turek, I., Drchal, V., and Bruno, P., Phys. Rev. B 64 (2001) 174402 [63] Perdew, J. P., Burke, K., and Ernzerhof, M., Phys. Rev. Lett. 77 (1996) 3865 [64] Van Setten, M. et al., Comput. Phys. Commun. 226 (2018) 39 [65] Garc´ıa, A., Verstraete, M. J., Pouillon, Y., and Junquera, J., Comput. Phys. Commun. 227 (2018) 51 [67] Jollet, F., Torrent, M., and Holzwarth, N., Comput. Phys. Commun. 185 (2014) 1246 [68] Perdew, J. P. et al., Phys. Rev. Lett. 100 (2008) 136406 [77] Pizzi, G., Cepellotti, A., Sabatini, R., Marzari, N., and Kozinsky, B., Computational Materials Science 111 (218) [78] Damle, A. p. 0, Lin, L., and Ying, L., Journal of chemical theory and computation 11 (2015) 1463 [79] Damle, A. p. 0 and Lin, L., Multiscale Modeling & Simulation 16 (2018) 1392 [80] Vitale, V. p. 1 et al., npj Computational Materials 6 (2020) 1 [81] Varignon, J. p. 0, Bibes, M., and Zunger, A., Nature Commun. 10 (2019) 1658 [82] Himmetoglu, B., Floris, A., de Gironcoli, S., and Cococcioni, M., International Journal of Quantum Chemistry 114 (2014) 14 [83] Bousquet, E. p. 1 and Spaldin, N., Phys. Rev. B 82 (2010) 220402 [84] Sakuma, R., Phys. Rev. B 87 (2013) 235109 [85] Fedorova, N. S., Ederer, C., Spaldin, N. A., and Scaramucci, A., Phys. Rev. B 91 (2015) 165122 [86] Ostler, T. p. 9 A., Barton, C., Thomson, T., and Hrkac, G., Phys. Rev. B 95 (2017) 064415

Pruning redundant reasoning tokens

The Green’s function method is more robust than total-energy mapping because it only requires a DFT calculation for a single spin configuration, typically the ground-state one This avoids the problem of converging DFT calculations for magnetic structures which are much higher in energy or even unstable Spin dynamics at moderate temperatures typically stay close to the ground state, and fitting to extreme configurations will produce worse dynamics The method is most accurate around the magnetic phase for which the DFT calculation has been performed

How the training signal is built

The exchange parameters are derived from Eq.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Automatic calculation of magnetic interaction parameters from density functional theory (DFT) results without manual case-by-case construction. This allows for easy integration into highthroughput workflows by only requiring a minimal user input, such as the paths of the Wannier or DFT related files and the species of magnetic atoms.

  2. Systematic calculation of Heisenberg Hamiltonian parameters, including isotropic exchange, anisotropic exchange, and Dzyaloshinskii-Moriya interaction (DMI). The package uses a Green’s function method with local rigid spin rotation treated as a perturbation to determine these parameters from the ground-state solution of the system.

  3. Computation of magnon band structures for magnetic materials. Once the real space J parameters are obtained, they can be transformed into q space using a Fourier transform to obtain dispersion curves, which are then calculated by running commands like TB2J magnon.

  4. Generation of spin dynamics and Monte Carlo simulation inputs directly from the calculated magnetic interaction parameters. TB2J provides templates of input files for this code, such as for the Multibinit code, allowing for easy automation of spin system simulations.

  5. Automated generation of necessary structural data for complex magnetic configurations through rotation and averaging. The package includes scripts like TB2J rotate to generate structures rotated along x, y, and z directions to obtain the full set of J ani D vector components via a weighted average over all three directions.

Related papers