Simulating ultrafast quantum magnetism with machine learning
Abstract of Johan Mentink
The explosive growth of digital data and its related energy consumption is pushing the need to develop fundamentally new physical principles for faster and more energy-efficient control of materials. Ultimately, the best compromise between speed and energy-efficiency is realized by employing coherent quantum dynamics. However, such dynamics is usually restricted to isolated systems temperatures far below room temperature which strongly limits application perspectives. Here we present our efforts towards finding alternative scenarios to control and magnetic materials in the quantum regime at room temperature. Instead of controlling magnetism with magnetic fields or spin-polarized currents, we focus on the ultrafast control of exchange interactions as we recently establised for Mott-Hubbard systems [1].
In particular, we show that short time-dependent electric fields can not only enhance and reduce the effective exchange interaction, but for strong electric fields even change the sign, as was recently confirmed in cold atom experiments [2]. Moreover, focusing on Heisenberg antiferromagnets, we show that ultrashort perturbations of exchange interactions triggers a time-dependent superposition of magnon pairs. We argue that this is a manifestation magnon entanglement and can be efficiently described by magnon-pair operators [3].
Furthermore, since the dynamics is governed by the energy of the exchange interactions it can be observed even at room temperature. Finally, we present our very first steps towards studying this magnon entanglement in the non-linear regime. To this end we apply the recently developed neural quantum states [4] to the dynamics of the 2D quantum Heisenberg model. We find excellent agreement for small systems where the dynamics is still accessible with exact diagonalization. Moreover, for large systems and for small perturbations, close correspondence with interacting spin-wave theory and rapid convergence with the number of neural network parameters is found [5]. This paves the way to explore the strongly nonequilibrium regime, with potentially enables the discovery of ultrafast reversal of magnetic order down to the quantum limit.
1. Mentink, J.H., Manipulating magnetism by ultrafast control of the exchange interaction. J. Phys.: Condens. Matter, 2017. 29 p. 453001
2. Görg, F., et al., Enhancement and sign change of magnetic correlations in a driven quantum many-body system. Nature, 2018. 553: p. 481.
3. Bossini, D., et al., Laser-driven quantum magnonics and THz dynamics of the order parameter in antiferromagnets. arXiv:1710.03143, 2017.
4. Carleo, G. and M. Troyer, Solving the quantum many-body problem with artificial neural networks. Science, 2017. 355(6325): p. 602.
5. G. Fabiani and J.H. Mentink, Investigating ultrafast quantum spin dynamics with machine learning. In preparation, 2018.
Contact
Johan Mentink, Radboud University, Nijmegen j.mentink@science.ru.nl