Ultrafast reduction of exchange splitting in ferromagnetic nickel

TL;DR

This paper introduces a new time-dependent density functional theory approach to simulate ultrafast exchange splitting reduction in ferromagnetic nickel, capturing experimental trends more accurately than previous models.

Contribution

The authors develop TDLDFT, integrating the Liouville equation into DFT to dynamically simulate band structure changes during ultrafast laser excitation.

Findings

Exchange splitting is strongly reduced during ultrafast laser excitation.

Majority and minority bands shift toward the Fermi level, with majority shifting more.

The model predicts trends consistent with experimental observations.

Abstract

A decade ago Rhie \et (Phys. Rev. Lett. {\bf 90}, 247201 (2003)) reported that when ferromagnetic nickel is subject to an intense ultrashort laser pulse, its exchange splitting is reduced quickly. But to simulate such reduction remains a big challenge. The popular rigid band approximation (RBA), where both the band structure and the exchange splitting are held fixed before and after laser excitation, is unsuitable for this purpose, while the time-dependent density functional theory could be time-consuming. To overcome these difficulties, we propose a time-dependent Liouville and density functional theory (TDLDFT) that integrates the time-dependent Liouville equation into the density functional theory. As a result, the excited charge density is reiterated back into the Kohn-Sham equation, and the band structure is allowed to change dynamically. Even with the ground-state density functional, a larger demagnetization than RBA is found; after we expand Ortenzi's spin scaling method into an excited-state (laser) density functional, we find that the exchange splitting is indeed strongly reduced, as seen in the experiment. Both the majority and minority bands are shifted toward the Fermi level, but the majority shifts a lot more. The ultrafast reduction in exchange splitting occurs concomitantly with demagnetization. While our current theory is still unable to yield the same percentage loss in the spin moment as observed in the experiment, it predicts a correct trend that agrees with the experiments. With a better functional, we believe that our results can be further improved.