• magneto-optical Kerr effect measurement
• ultrafast demagnetization

In 1996 a paper by Beaurepaire et al. [PRL, 1996] reported on an ultrafast magneto-optic
investigation on Nickel. It showed an ultrafast reduction of magnetization within the first
picoseconds and a subsequent recovery on a slower time scale. This was the first evidence of
an all-optical magnetization reduction on a femtosecond time scale initiating great interest of
the scientific community. Since then, theoretical and experimental efforts have been made to
understand what lies at the basis of optical ultrafast demagnetization. The great complexity
arises from the required transfer of energy and angular momentum between non-equilibrium
electronic, spin and lattice degrees of freedom on a femtosecond time scale. The most successful
theory developed in the last decade [Koopmans et al., J. Magn. Magn. Mat., 2005] assumes
the lattice as an angular momentum reservoir through the Elliott-Yafet spin flip scattering.
After optical excitation of the electronic system, a finite spin flip probability asf is attributed
to each electron-phonon scattering event. However, literature on theoretical calculations of
asf disagrees, whether this effect alone is sufficient to explain the magnitude of experimentally
observed ultrafast demagnetization, or not. In 2010 Battiato et al. [PRL, 2010] proposed a
model based on superdiffusive spin transport as a mechanism for ultrafast demagnetization.
Optically excited, non equilibrium (NEQ) electrons have a spin dependent velocity and lifetime,
hence leading to significant spin-currents. Majority spins have a larger mobility than minority
spins, causing a depletion of majority spins in the (probed) magnetic layer, therefore reducing
the magnetization.
In addition to fundamental issues, there are strong interests in possible technology applica-
tions, since magnetic materials are employed in magnetic data storage. The quest of technology
progress is for a smaller bit area and for a faster writing process. For this reason it is of great
interest to understand the limiting time scales in which the magnetic order can be manipulated
and develop new techniques for this purpose.
The availability of laser sources with short pulses with a temporal width of tens or few
hundreds of femtoseconds allows studying sub-picosecond dynamics with the so called pump
and probe technique. A short pump pulse triggers the ultrafast dynamics, while a short probe
pulse allows studying the induced non-equilibrium effects. This kind of investigation can be
extended to magnetic materials taking advantage of the magneto-optical Kerr effect (MOKE).
MOKE refers to the change in the state of polarization when polarized light is reflected off a
magnetized material. Detecting this change allows measuring the magnetization of a sample.
My thesis work was done at the Technische Universitaet Berlin, where a time resolved MOKE
(TR-MOKE) experimental set up is available. I started my work by implementing important
improvements of the setup, among others they included testing and installation of a new pho-
todetector and optimization of the micro-focusing of the pump and probe pulses at the sample
surface. The TR-MOKE set up allows acquiring demagnetization curves from thin samples,
which are then analysed to infer the demagnetization and re-magnetization time constants.
Since, at the moment, there is no generally accepted model to describe demagnetization curves,
several different approaches are found in literature. It was thus necessary to discuss each of
them and check whether fitting results depend on the model used. This was actually the case
and one model was chosen to analyse all data of this thesis, in order to compare results from
different samples.
The samples investigated were specifically designed to test the role of spin superdiffusion.
Gold gratings with nanometre dimension were structured onto the surface of a magnetic Co/Pt
multilayer (ML), leading to nanometre spatially modulated excitation patterns. Light impinging
directly on the gold stripes is absorbed, while the magnetic material between gold stripes is
excited. There is thus a lateral temperature gradient which favours spin superdiffusion. Indeed
a faster demagnetization is found in samples featuring the gold nanostructures compared to the
flat surface. Due to the geometry of the structures and to interface effects, the excitation intensity
is locally enhanced and this effect is visible in the remagnetization time of the structured sample
compared to the film. Quantification of this enhancement confirms the theoretical expectations.
In addition, ultrafast demagnetization processes in thin magnetic Co/Pt ML grown on dif-
ferent substrates was analysed. Strikingly, we find no significant difference in the magnitude
of the demagnetization, for films grown on conducting or insulating substrates. This is a clear
indication that superdiffusive spin transport alone cannot explain the phenomenon of ultrafast
demagnetization, because all diffusion phenomena are inhibited for an insulating substrate.
In the present days, theoretical and experimental ultrafast magnetism face the challenge to
develop a model where the spin flip scattering and the spin superdiffusion coexist as microscopic
mechanisms for ultrafast demagnetization. This work is a part of this effort and experimental
evidences are analysed taking into account both of these microscopic models.