
2
an atomistic spin model framework [29,30]. However, the
lack of a theoretical model describing the magnetization
dynamics questions the full validity of the atomistic spin
model.
In this work, we demonstrate that the M3TM is
able to quantitatively describe the ultrafast dynamics
in 3dferromagnets, including electron, lattice and spin
degrees of freedom. To do so, we replicate the dynamics
of the experimentally measured electron and lattice
temperatures as well as magnetization for a range of laser
fluences in all three 3dferromagnetic transition metals
iron, nickel and cobalt. Notably, we are able to reduce the
number of fit parameters to only the spin-flip probability.
We achieve so by using system parameters that have been
calculated from ab-initio methods and experimentally
validated. By contrast to previous works, we find that the
spin-flip probabilities agree with to those calculated via
ab-initio [14,17]. We conclude that our work evidences
electron-phonon spin-flips as intrinsic mechanism driving
ultrafast magnetization dynamics in 3dferromagnets.
Our model is based on an extension to the M3TM [12,
31]. The energy flow dynamics are described by the two-
temperature model (see Fig. 1),
Ce
dTe
dt =gep(Tp−Te) + S(t) + ˙
Qe−s(1)
Cp
dTp
dt =−gep(Tp−Te).(2)
When a metallic thin film is subjected to an optical
laser pulse, only the electrons are excited by the photon
electric field. Initially, the absorbed energy is barely
transferred to the lattice and consequently the electron
system heats up. On the timescale of 100 fs, the
electron temperature will rise far above the critical
temperature, Tc. The magnetic system responds to this
fast temperature rise by reducing its magnetic order on
similar time scales. The electron and lattice systems are
assumed to be thermalized so that their energy can be
described by a temperature, Tefor the electrons and Tp
for the phonons. In Eq. (1), the absorbed laser pulse
power is represented by a Gaussian function, S(t) =
S0∗G(τp), where τpis the pulse duration. The electron
heat capacity is Ce. The electron-phonon coupling allows
for temperature equilibration of hot electrons and the
lattice on the time scale determined by the ratio gep/Ce.
Since gep accounts only for spin-conserving scattering
events, in Eq. (1) we include a term that accounts for
the finite energy cost (gain) of a spin-flip and couple
it to the electron dynamics: ˙
Qe−s=J0m˙m/Vat, with
J0/3=[S/(S+ 1)]kBTcin the MFA, mis the reduced
magnetization, J0is the exchange energy, Vat the atomic
volume and effective spin S. In phase I (dm/dt < 0)
the energy cost of an electron-phonon mediated spin-flip
of probability asf in minority direction is deducted from
the electronic energy, while in phase II (dm/dt > 0)
the direction of energy flow is reverted (Fig. 1). In
FIG. 1. Energy absorbed by the spinless electrons
S(t)is distributed to the lattice (gep) and localized spins
(˙
Qe−s). Energy is transferred from electron to spin
during demagnetization (I), and spin to electrons during
remagnetization (II). The angular momentum from electron-
phonon mediated spin-flips, with probability asf , is implicitly
exchanged with lattice.
other works the energy flow between the spin and the
electron systems has been taken into account by adding
it to the electron specific heat, Ce→Ce+Cs, where Cs
is the equilibrium spin specific heat [27,29]. However,
recent works suggest that this energy flow needs to
be calculated through the spin Hamiltonian for non-
equilibrium spin configurations [29,30]. An important
aspect of the two-temperature model in Eqs. (1) and
(2) is the exact value of Ce,Cp, and gep. In this work,
we use the ab-initio calculated parameters, which were
already used before in Ref. [29,30]. The parameters
are temperature dependent and were already validated
by direct comparison to the ultrafast lattice dynamics
in iron, nickel and cobalt [29,30]. The electronic and
lattice heat capacities agree very well with experimental
data [19,32]. The magnetization dynamics is calculated
using the M3TM for finite spin values, for details of
the M3TM see Supplemental Material and Ref. [31].
Within this model, the rate parameter defining the
magnetization dynamics scales linearly with the so-called
spin-flip probability asf . Here, we use S= 1/2for nickel,
S= 2 for iron and S= 3/2for cobalt [33]. For these
values, within the MFA, the equilibrium magnetization
as function of temperature is well reproduced for all
three 3dferromagnets. Figure 2shows an example
of the comparison of our model to experimental data
for the magnetization [Fig. 2(a)], electron [Fig. 2(b)]
and lattice [Fig. 2(c)] temperature dynamics in nickel.
Vertical orange line denotes the transition from phase I
(dm/dt < 0) to phase II (dm/dt > 0). The electron
temperature is retrieved from tr-ARPES measurements
on a 400 nm nickel film, where the electronic energy
distribution around the Fermi edge is measured and fitted
to the Fermi-Dirac distribution, yielding the electronic
temperature [27]. The lattice dynamics are retrieved