PDF hosted at the Radboud Repository of the Radboud University
Nijmegen
The following full text is a publisher's version.
For additional information about this publication click this link.
http://hdl.handle.net/2066/75623
Please be advised that this information was generated on 2022-05-30 and may be subject to
change.
PRL 103, 117201 (2009)
jy Selected for a Viewpoint in Physics
PHYSICAL REVIEW LETTERS
week ending
11 SEPTEMBER 2009
S ’
Ultrafast Path for Optical Magnetization Reversal via a Strongly Nonequilibrium State
K. Vahaplar,1* A.M . Kalashnikova,1,5 A. V. Kimel,1 D. Hinzke,2 U. Nowak,2 R. Chantrell,3 A. Tsukamoto,46 A. Itoh,4
A. Kirilyuk,1 and Th. Rasing1
1Institute for Molecules and Materials, Radboud University Nijmegen, P.O. Box 9010 6500 GL Nijmegen, The Netherlands
2Fachbereich Physik, Universität Konstanz, D-78457 Konstanz, Germany
3Department of Physics, University of York, York Y010 5DD, United Kingdom
4College of Science and Technology, Nihon University, 7-24-1 Funabashi, Chiba, Japan
5Ioffe Physical-Technical Institute of the Russian Academy of Sciences, 194021 St. Petersburg, Russia
6PREST0, Japan Science and Technology Agency, 4-1-8 Honcho Kawaguchi, Saitama, Japan
(Received 10 April 2009; revised manuscript received 28 July 2009; published 8 September 2009)
Using time-resolved single-shot pump-probe microscopy we unveil the mechanism and the time scale
of all-optical magnetization reversal by a single circularly polarized 100 fs laser pulse. We demonstrate
that the reversal has a linear character, i.e., does not involve precession but occurs via a strongly
nonequilibrium state. Calculations show that the reversal time which can be achieved via this mechanism
is within 10 ps for a 30 nm domain. Using two single subpicosecond laser pulses we demonstrate that for a
5 domain the magnetic information can be recorded and readout within 30 ps, which is the fastest
‘‘write-read’’ event demonstrated for magnetic recording so far.
DOI: 10.1103/PhysRevLett.103.117201 PACS numbers: 75.40.Gb, 75.60.Jk, 85.70.Li
The fundamental and practical limit of the speed of
magnetization reversal is a subject of vital importance for
magnetic recording and information processing technolo
gies as well as one of the most intriguing questions of
modern magnetism [1- 8]. The conventional way to re
verse the magnetization M is to apply a magnetic field
H antiparallel to M. In this collinear M-H geometry the
reversal occurs via precession accompanied by damping
that channels the associated angular momentum into the
lattice. Although this process is perfectly deterministic, it
is also unavoidably slow, typically of the order of nano
seconds, due to the required angular momentum transfer
[
8].
Alternatively, the driving field can be applied orthogonal
to M, so that the created torque [M X H] leads to a rapid
change of the angular momentum and a possible switching
of the magnetization [
1,3,4,9]. However, such precessional
switching requires a magnetic field pulse precisely tuned to
half of the precession period. The fastest precessional
reversal demonstrated so far using an external magnetic
field [1,5] or a spin-polarized current [6,7,10] is limited to
100 ps. Moreover, it has been shown that for field pulses
shorter than 2.3 ps such a switching becomes nondetermin-
istic [5,11].
Ultrafast laser-induced heating of a magnetic material is
known to stimulate the transfer of angular momentum from
spins to lattice on a femtosecond time scale [12,13]. It has
been recently demonstrated that a 40 fs, circularly polar
ized, laser pulse is able to reverse magnetization in a col
linear M-H geometry [14], as if it acts as an equally short
magnetic field pulse H eff ~ [E X E*] (where E is the
electric field of light) pointing along the direction of light
[15]. Although this experiment showed the intriguing pos
sibility of triggering magnetization reversal with a subpi
cosecond stimulus, the relevant time scales and mechanism
of such an optically induced magnetization reversal are
still unanswered questions, since a precessional switching
within 40 fs would require enormous effective magnetic
fields above 102 T and unrealistically strong damping. To
address these questions we used femtosecond single-shot
time-resolved optical imaging of magnetic structures and
multiscale modeling beyond the macro-spin approxima
tion. The combination of these advanced experimental
and theoretical methods unveiled an ultrafast linear path
way for magnetization reversal that does not involve pre
cession but occurs via a strongly nonequilibrium state.
In our experiments the amorphous ferrimagnetic 20 nm
GdxFe100_x_yCoy films with perpendicular anisotropy [
14]
were excited by a single circularly polarized laser pulse
(FWHM of about 100 fs, a central wavelength at A0 =
800 nm). A single linearly polarized probe pulse
(FWHM = 100 fs, A0 = 640 nm) delayed with respect
to the pump was used for ultrafast imaging of the magnetic
domain structure by means of the magneto-optical Faraday
effect. Magnetic domains with magnetization parallel
(‘‘up’’) or antiparallel ( ‘‘down’’) to the sample normal
are seen as white or black regions, respectively, in an image
on a CCD camera [16]. After each ‘‘write-read’’ event, the
initial magnetic state was restored by applying a magnetic
field pulse. Taking images of the magnetic structure for
different delays between the pump and probe pulses we
were able to visualize the ultrafast dynamics of the laser-
induced magnetic changes in the material.
Figure 1(a) shows images of magnetic domains in a
Gd24Fe66 5Co9 5 sample at different delays after excitation
by right- ( ^ +) or left-handed (^ _ ) circularly polarized
pulses. The images were obtained for both types of do
mains with initial magnetization up and down. In the first
0031-9007/ 09 /103(11)/117201(4)
117201-1
© 2009 The American Physical Society
PRL 103, 117201 (2009)
PHYSICAL REVIEW LETTERS
week ending
11 SEPTEMBER 2009
FIG. 1 (color). (a) The magnetization evolution in
Gd2 4 Fe6 6 5 Co9 .5 after the excitation with a + and a~ circularly
polarized pulses at room temperature. The domain is initially
magnetized up (white domain) and down (black domain). The
last column shows the final state of the domains after a few
seconds. The circles show areas actually affected by pump
pulses. (b) The averaged magnetization in the switched areas
(~5 ^m) after a + and a~ laser pulses, as extracted from the
images in (a) for the initial magnetization up.
few hundreds of femtoseconds, pump pulses of both he-
licities bring the originally magnetized medium into a
strongly nonequilibrium state with no measurable net mag
netization, seen as a gray area in the second column of
Fig. 1(a), the size of which is given by the laser beam
intensity profile. In the following few tens of picoseconds
either the medium relaxes back to the initial state or a small
(~ 5 ^m ) domain with a reversed magnetization is formed.
It is thus obvious that (i) the switching proceeds via a
strongly nonequilibrium demagnetized state, clearly not
following the conventional route of precessional motion,
and (ii) the final state is defined by the helicity of the 100 fs
pump pulse [last column of Fig. 1(a)].
As one can see from Fig. 1(a), the metastable state
corresponding to reversed magnetization is reached within
60 ps after a + ( a ~) excitation. This state is, however,
slightly different from the final state [the last column in
Fig. 1(a)], as clearly seen from Fig. 1(b). This happens due
to the laser-induced heating of the sample followed by slow
( » 1 ns) heat diffusion [17]. To take into account renor
malization of the two metastable states of magnetization at
the subnanosecond time scale we introduce two asymptotic
levels [see dashed lines in Fig. 1(b)]. The characteristic
time of switching r sw can be identified as the time required
to reconstruct 63% (1 — e _1) the difference between the
metastable states [Fig. 1(b)]. For example, in Fig. 1, r sw =
60 ps. After 1.5rsw the difference reaches 80% and, as also
can be seen from Fig.
1(a), this time can be reliably
assumed as the period required for a write-read event
(rw-r = 90 ps for the example in Fig. 1). The switching
time is in fact surprising, because in contrast to heat-
assisted magnetic recording [
18], the reversal time is
much longer than the effective light-induced magnetic field
pulse H eff. The duration of the latter Ateff is still an open
question but can be different from the FWHM of the
optical pulse. However, Ateff can be estimated from the
spectrum of THz radiation generated by an Fe film when
the latter is excited by a subpicosecond visible laser pulse.
According to Ref. [19], the intensity of the THz emission
depends on the polarization of the incoming light and has
to be explained in terms of difference-frequency genera
tion. Phenomenologically, this is very similar to the inverse
Faraday effect. Based on a half-period oscillation with the
lowest frequency in the THz spectrum [
19], the maximum
A teff is about 3 ps. The pulse amplitude Heff, for a typical
pump fluence of 2.5 J/m 2 and the magneto-optical con
stant of GdFeCo (~ 3 X 105 deg/cm), reaches 20 T.
To understand this route for magnetization reversal via
such a strongly nonequilibrium state we solved the
Landau-Lifshitz-Bloch (LLB) equation. This macrospin
approach encapsulates very well the response of a set of
coupled atomic spins subjected to rapidly varying tempera
ture changes, including the reduction of the magnitude of
M [20,21]. The temperature dependence of the anisotropy
constant K u is introduced in the LLB equation via the
temperature dependence of the transverse susceptibility
[22]. The temperature-dependent parameters for the LLB
equation, i.e., the longitudinal and transverse susceptibili
ties and the temperature variation of the magnetization, are
calculated atomistically using Langevin dynamics com
bined with a Landau-Lifshitz-Gilbert equation for each
spin [22]. It is well known that, due to the small heat
capacity of electrons, optical excitation by a subpicosec
ond laser pulse can cause heating of the electron system
well above 1000 K, whereafter the electrons equilibrate
with the lattice to a much lower temperature on a (sub)
picosecond time scale given by the electron-phonon inter
action [13]. This laser-induced increase of the kinetic
energy (temperature) of the electrons is simulated using a
two-temperature model [23], the parameters for which
were taken to be typical for a metal [24] (electron heat
capacity Ce = 1.8 X 106 J/m 3 K at room temperature and
electron-phonon coupling Gel-ph = 1.7 X 1018 J/K s). The
simulations show that in the first 100 fs the electron tem
perature Tel increases from 300 K up to T*j and relaxes with
a time constant of 0.5 ps down to the vicinity of TC.
Simultaneously the spins experience a short pulse of effec
tive magnetic field with amplitude Heff = 20 T and dura
tion A teff. The possibility of magnetization reversal under
these circumstances has been analyzed numerically for a
volume of 30 X 30 X 30 nm3. The results of the simula
117201-2
PRL 103, 117201 (2009)
PHYSICAL REVIEW LETTERS
week ending
11 SEPTEMBER 2009
tions are plotted in Fig. 2(a) as a phase diagram, defining
the combinations of T*j and Ateff for which switching
occurs for the given H eff. The assumed perpendicular
anisotropy value was Ku = 6.05 X 105 J/m 3 at 300 K.
As can be seen from the diagram, a field pulse as short as
A teff = 250 fs can reverse the magnetization. For better
insight into the reversal process we simulated the latter for
A teff = 250 fs and T*j = 1130 K. The result is plotted in
Fig. 2(b), showing that, already after 250 fs, the effective
fields of two different polarities bring the medium into two
different states, while the magnetization is nearly
quenched within less than 0.5 ps. This is followed by
relaxation either to the initial state or to the state with
reversed magnetization, achieved already within 10 ps.
The considered pulse duration Ateff of 250 fs is only
2.5 times larger than the FWHM of the optical pulse in
our experiments [
25] and well within the estimated lifetime
of a medium excitation responsible for Heff. Importantly, in
simulations A teff was found to be sensitive to the parame
ters of the two-temperature model. In particular, an in
crease of Gel-ph leads to a reduction of the minimum field
pulse duration. This shows that the suggested mechanism
may, in principle, explain the experimentally observed
laser-induced magnetization reversal. This magnetization
reversal does not involve precession; instead, it occurs via a
linear reversal mechanism, where the magnetization first
vanishes and then reappears in a direction of H eff, avoiding
any transverse magnetization components, just as seen in
Fig. 1(a). Exactly as in the experiments, the initial 250 fs
effective magnetic field pulse drives the reversal process,
that takes 1-2 orders of magnitude longer.
The state of magnetization after the pulse is critically
dependent on the peak temperature T*j and the pulse du
ration. For ultrafast linear reversal by a 250 fs field pulse it
2
1000
■s
no reversal
0.2
0.4
0.6
0.8
1.0
1.2
Field pulse duration At (ps)
100
80
60
40
20
Switchability (%)
FIG. 2 (color). (a) Phase diagram showing the magnetic state
of the (30 nm) 3 volume achieved within 10 ps after the action of
the optomagnetic pulse with parameters Heff = 20 T, Aieff, and
T*j. (b) The averaged z component of the magnetization versus
delay time as calculated for 250 fs magnetic field pulses Heff =
±20 T and T*j = 1130 K. (c) Switchability versus the pump
intensity for Gd2 2 Fe6 8 2 Co9 .8 at room temperature. We calculated
the peak electron temperature T*j using Ce. Note that in this
range of intensities the amplitude of the effective light-induced
magnetic field varies within 19.2-20.8 T.
is necessary that, within this time, Tel reaches the vicinity
of TC. If, however, this temperature is too high and persists
above TC for too long, the reversed magnetization is de
stroyed and the effect of the helicity is lost. This leads to a
phase diagram [Fig. 2(a)], showing that the magnetization
reversal may occur in a certain range of T*l. Such a
theoretically predicted reversal window of electron tem
perature can be easily verified in the experiment when one
changes the intensity of the laser pulse. Figure 2(c) shows
the switchability, i.e., the difference between the final
states of magnetization achieved in the experiment with
^ +- and ^ —-polarized pulses, as a function of T^, calcu
lated from the laser pulse intensity. It is seen that, indeed,
switching occurs within a fairly narrow laser intensity
range [26]. For intensities below this window no laser-
induced magnetization reversal occurs, while if the inten
sity exceeds a certain level both helicities result in magne
tization reversal, since the laser pulse destroys the
magnetic order completely, which is then reconstructed
by stray fields [27,28]. Such a good agreement between
experiment and theory supports the validity of the pro
posed reversal mechanism.
Despite this qualitative agreement between simulations
and experiments, the experimentally observed reversal
time is several times larger than the calculated 10 ps. The
latter, however, is calculated for a 30 nm domain, whereas
in our experiments the magnetization in a 5 ^ m spot is
manipulated. This size is defined by (i) the minimum size
of the stable domain in the material and (ii) by the area
within the laser spot, where the intensity favors the
helicity-dependent reversal. Inhomogeneities in the sample
and the intensity profile will lead to variations of T^ over
the laser spot. If due to these factors every 30 nm element
of the 5 ^ m spot is reversed with a probability between
50% and 100%, the actual time of magnetization reversal
of this large spot will depend on its size and the speed of
domain walls. Their mobility increases dramatically in
GdFeCo alloys in the vicinity of their angular momenum
compensation point (Tcomp); i.e., the temperature where the
angular momenta of the two sublattices cancel each other
[29- 31]. Therefore, one should expect a dramatic accel
eration of magnetization reversal near Tcomp. Note that this
would also perfectly explain the difference between the
times required for the formation of the switched domain
and the relaxation to the initial state [Fig. 1(b)]. Indeed, in
the former case the domain wall motion is additionally
accelerated by the demagnetizing field, while in the latter
case this field slows the motion down.
This hypothesis was verified experimentally by inves
tigating the reversal process as a function of temperature
in three alloys Gd22Fe682Co9 .8, Gd24Fe665Co9 .5, and
Gd26Fe64:7Co9:3 that are characterized by different com
pensation temperatures. The observed write-read time r w-r
is plotted in Fig. 3 as a function of the difference between
the sample temperature and the compensation point T —
Tcomp. The write-read time is the fastest and weakly de-
117201-3
PRL 103, 117201 (2009)
PHYSICAL REVIEW LETTERS
week ending
11 SEPTEMBER 2009
T-T (K)
comp ' '
FIG. 3 (color). The write-read time Tw-r versus the relative
temperature T — Tcomp for Gd2 2 Fe6 8 ,2 Co9 , 8 (Tcomp = 100 K),
Gd24Fe66.5Co9.5 (TComp = 2 8 0 K), and Gd26Fe64.7Co9.3 (Tcomp =
390 K). We achieved magnetization reversal within 30 ps for
Gd2 2 Fe6 8 2 Co9 8 at 10 K. The dashed line is guide to the eye.
pends on temperature below Tcomp. This agrees with the
hypothesis that the relaxation time to the metastable state is
defined by the domain wall speed averaged over the photo
excited area. If the laser pulse brings the central part of the
excited area from initial temperature T < Tcomp to the
vicinity of TC, somewhere within this area the material is
at TComp, where the domain wall mobility is the largest.
Then, it is the mobility at TComp which dominates the
averaged domain wall speed in the photoexcited area
and, thus, determines the write-read time. Above TComp,
all-optical magnetization reversal can still be realized, but
the write-read time increases exponentially with increasing
temperature. For example, while r w-r for Gd22Fe68 2Co9 8
(TComp = 100 K) at room temperature is found to be ex
tremely slow (16 ns), a huge decrease of r w-r of 2 orders of
magnitude is observed as T — Tcomp decreases. Finally, at
10 K we succeeded to achieve all-optical magnetization
reversal within just 30 ps, which is the fastest write-read
event demonstrated for magnetic recording so far.
In conclusion, by time-resolved single-shot microscopy,
we found a novel and ultrafast path for magnetization
reversal triggered by a subpicosecond circularly polarized
laser pulse. The reversal does not involve precession, but
instead has a linear character, proceeding via a strongly
nonequilibrium state. This all-optical reversal occurs only
in a narrow range of pulse energies. Using two single
subpicosecond laser pulses we demonstrated the feasibility
of both all-optical recording and reading on an ultrashort
time scale. The magnetic information was recorded by a
subpicosecond laser pulse in a 5 ^ m domain and readout
by a similarly short pulse within 30 ps, which is the fastest
write-read event demonstrated for magnetic recording so
far. Simulations for 30 nm domains demonstrate the feasi
bility of reversing magnetization within 10 ps. This time
can be even faster for media with a higher magnetic
anisotropy constant than the one used in our calculations.
We thank A. J. Toonen and A. F. van Etteger for techni
cal support and Dr. I. Radu for his help with sample char
acterization and stimulating discussions. This research has
received funding from NWO, FOM, NanoNed and EC FP7
[Grants No. NMP3-SL-2008-214469 (UltraMagnetron)
and No. 214810 (FANTOMAS)].
*K.Vahaplar@science.ru.nl
[1] C. H. Back et al., Science 285, 864 (1999).
[2] B. C. Choi et al., Phys. Rev. Lett. 8 6 , 728 (2001).
[3] Th. Gerrits et al., Nature (London) 418, 509 (2002).
[4] S. Kaka and S.E. Russek, Appl. Phys. Lett. 80, 2958
(2 0 0 2 ).
[5] I. Tudosa et al., Nature (London) 428, 831 (2004).
[6 ] T. Devolder et al., J. Appl. Phys. 98, 053904 (2005).
[7] Y. Acremann et al., Phys. Rev. Lett. 96, 217202 (2006).
[8 ] J. Stohr and H. C. Siegmann, Magnetism: From Funda
mentals to Nanoscale Dynamics (Springer-Verlag, Berlin,
2006).
[9] H. W. Schumacher et al., IEEE Trans. Magn. 38, 2480
(2 0 0 2 ).
[10] J. C. Slonczewski, J. Magn. Magn. Mater. 159, L1 (1996).
[11] A. Kashuba, Phys. Rev. Lett. 96, 047601 (2006).
[12] C. Stamm et al., Nature Mater. 6 , 740 (2007).
[13] E. Beaurepaire et al., Phys. Rev. Lett. 76, 4250 (1996).
[14] C. D. Stanciu et al., Phys. Rev. Lett. 99, 047601 (2007).
[15] A. V. Kimel et al., Nature (London) 435, 655 (2005).
[16] The angles of incidence for the pump and probe beams
were 2 0 ° and 0 °, respectively.
[17] Note that the heat load brought to the sample by the laser
pulse is several orders of magnitude smaller than the one
during current-induced magnetization reversal used in real
devices. Moreover, the heat diffusion can be accelerated
provided a proper nanostructure design.
[18] J. Hohlfeld et al., Phys. Rev. B 65, 012413 (2001).
[19] D. J. Hilton et al., Opt. Lett. 29, 1805 (2004).
[20] D. A. Garanin, Phys. Rev. B 55, 3050 (1997).
[21] U. Atxitia et al., Appl. Phys. Lett. 91, 232507 (2007).
[22] N. Kazantseva et al., Phys. Rev. B 77, 184428 (2008).
[23] N. Kazantseva et al., Europhys. Lett. 81, 27004 (2008).
[24] G. Zhang, W. Hiibner, E. Beaurepaire, and J.-Y. Bigot,
Spin Dynamics in Confined Magnetic Structures I, Topics
in Applied Physics Vol. 83 (Springer, New York, 2002).
[25] The switching was observed even for laser pulses with a
FWHM of ~3 ps, which corresponds to a larger Aieff.
[26] For T
< Tcomp the intensity required for switching was
increasing with temperature decrease. For example, for
sample Gd2 4 Fe6 6 5 Co9 5 the decrease of T by 300 K caused
an increase of the required intensity by ~ 8 %.
[27] T. Ogasawara et al., Appl. Phys. Lett. 94, 162507 (2009).
[28] The slower reversal in heat-assisted recording actually
corresponds to Tel above our reversal window, when the
whole system is brought above TC and the dynamics is
determined by cooling in an external magnetic field [
18].
[29] T. Miyama et al., IEEE Trans. Magn. 14, 728 (1978).
[30] R. S. Weng and M. H. A. Kryder, IEEE Trans. Magn. 29,
2177 (1993).
[31] V. Randoshkin et al., Phys. Solid State 45, 513 (2003).
117201-4
