Ultrafast Laser-Induced Spin-Transfer Torque

Ultrafast Laser-Induced
Spin-Transfer Torque
0
1
Acknowledgements
 Eindhoven University of Technology





100 fs
Sjors Schellekens
Koen Kuiper
Wouter Verhoeven, Ruud de Wit, Taco Vader
Francesco Dalla Longa, Gregory Malinowski
Bastiaan Bergman, Jeroen Rietjens, Carlos Bosco,
Csaba Jozsa, Maarten Van Kampen, Harm Kicken
 Technische Universität Kaiserslautern
 Tobias Roth
 Mirco Cinchetti
 Martin Aeschlimann
experimentalists
of the week
 MPI Metalforschung Stuttgart
 Daniel Steiauf
 Manfred Fähnle
Bert Koopmans
Local dynamics vs. spin transport
Outline
2
3
 Introduction
 Local magnetization dynamics
 Spin transfer (super-diffusive)
FM
M
 Laser-induced spin transfer torque
 Why efficient
 Our recent experiments
 A first interpretation:
Super-diffusive spin currents or Spin-Dependent Seebeck?
 Its importance for local magnetization dynamics
Beaurepaire et al., Phys. Rev. Lett. 1996
What happens after fs laser excitation?
Sub-ps loss of magnetization
5
4
S
 Beaurepaire et al., PRL 1996
 Launching spin waves
N
S
 Van Kampen et al., PRL 2002
MO contrast
 Quenching magnetic moment
N
 AF  F phase transition
S
 Ju et al., PRL 2004;
Thiele et al. APL 2004
N
1.0
Ni thin film
 Switching by circularly polarized light
 Stanciu et al., PRL 2007
+
S
N
N
0.5
S
0
 “Toggle switching” ferrimagnets
 Radu et al., Nature 2011
linear
S
N
N
50 fs
pump/probe
S
Beaurepaire et al., Phys. Rev. Lett. 1996
Sub-ps loss of magnetization
5
10
15
∆t (ps)
Conservation of Angular Momentum
6
7
1,0
Temperature (K)
MO contrast
Angular Momentum Transfer
M
500
E =
magnetization
0,9
400
E
50 fs
pump/probe
0
th
1
Delay (ps)
0.5 ps
th <
0.1 ps
electrons
0,8
lattice
300
electr.
electr.
2
lattice
Transfer from spin to orbit
M <
0.2 ps
Photons in or out
spins
From spin to the lattice
J
spins
lattice
Are photons and hot (highly excited) electrons
relevant?
Boeglin et et al., Nature 2010
Distinguishing orbital and spin moments
8
Ni 10 nm film
low laser fluence
1,0
0,0
most demag.
after
thermalization
M / M0
-0,5
M / M (%)
9
high laser fluence
-1,0
-1,5
0,8
0,6
-2,0
0,0
0,5
1,0
0,4
1,5
delay (ps)
0
1
2
4
delay (ps)
data: TU/e & U. Kaiserslautern
The answer (?)
Koopmans et al., PRL 2005
Nat. Mater. 2010
Minimalistic model
10
Like de Haas & Einstein
Photons in or out
Transfer from spin to orbit
11
 Electrons: Constant DOS
 Phonons: Einstein model (+ Debye)
 Spins: mean-field S = ½ Weiss model
e
From spin to the lattice
e
e
 spin flip
Figure of
Assumption:
τth merit
~0
e
e
 e-p
Highly-excited electrons / laser field
e
U
 e-e
p
asf
TC
a
sf
Leading to:
μτat~ 0.4 ps
E
e-σ
eσ
DF
ED, Dp
TC, μat
p
Finite chance for spin-flip
upon momentum scattering
(Elliott-Yafet)
What it can reproduce
Open questions
But which
phonons, and how
do they carry
angular
momentum?
12
 Slow dynamics (Gd) vs. fast dynamics
Do we treat magnetic
excitations adequately?
Stoner vs. Magnons
 Koopmans et al., Nature Mater. 2010
 Temperature- and laser fluence dependence
 Roth et al., Phys. Rev. X 2012
 Toggle switching of ferrimagnets (two spin sub-lattices)
 Schellekens et al., Phys. Rev. B Rapid 2012
Is it really possible
to treat this highly
nonequilibrium
system thermodynamically?
Three-particle
interaction?
(e-p + spin-flip)
 Combined with experiment: asf ~ 0.1
 Roth et al. Phys. Rev. X 2012
 Carva et al. Phys. Rev. Lett. 2011
13
Isn’t our picture of the
rare earth dynamics too
crude?
 Note: very much like atomistic LLG and LLB
 Kazantseva et al. PRB 2008, etc.,
 Mentink et al., PRL 2012
Or is it something
completely else?
Or: Claims that spin transfer may explain all
Fs spin-transfer
Malinowski et al. Nature Physics 2008
15
Pt
Co
NiO
Pt
Co
Ru
Normalized  (arb.u.)
14
NiO
Ru
0,0
0,5
1
Delay (ps)
Majority spins travel further
2
3
And even more exciting / surprising…
The approach and the surprize
16
Can this highenergy state live
that long?
?
17
Fe
Ni
Fe
Ni
Outline
All previous results: collinear systems
18
 Introduction
 Local magnetization dynamics
 Spin transfer (super-diffusive)
19
 If spin transfer assists thermodynamically stable final
state: OK
 Dynamics just speeds up (Malinowski)
 But if spin transfer creates strong non-equilibrium?
 Laser-induced spin transfer torque
 Why efficient
 Our recent experiments
 A first interpretation:
Super-diffusive spin currents or Spin-Dependent Seebeck?
 Its importance for local magnetization dynamics
 Either local dissipation of angular momentum (100 fs)
 Or compensating spin transport
(also just femtoseconds) (?)
What if non-collinear spin transfer?
Motivation
20
21
 Ideal method for quantifying laser-induced spin
transport
 Addressing role of super-diffusive spin transport to
ultrafast demagnetization
 Device options? All-optical switching?
 Measuring thermal STT without lithography?
Quick dissipation of spin momentum
(mixing conductance or precession)




Efficient absorption of angular momentum
Causing laser-induced STT
Final state just rotation of quantization axis
So final state in thermal equilibrium
Recent experiments @ TU/e
Sample properties
22
Polar MOKE
measuring Mz
Probe pulse
after delay ∆t
Longitudinal MOKE
Sample stack
Field @ 45o with respect to sample plane
B
in-plane
in-plane
Co
HK
Cu
Pt 4 nm
HK
[Co/Ni]n
PMA
Pt 1 nm
Co 3 nm
Cu 0 - 20
[Co / Ni ]x4
0.2/0.6 nm
PMA
Four remnant configurations magnetic bilayer!
Schellekens et al. (TU/e), In preparation
Indeed two precessions!
Cu vs Pt spacer layer
24
Two oscillations
- with proper symmetry B
Can be assigned to
top and bottom
 Oscillation top in-plane layer suppressed:  STT!
 Precession bottom layer different origin  “∆K”
Schellekens et al. (TU/e), In preparation
Suppression of STT on top layer
Phase of the STT-induced precessions
angular momentum transfer (%)
26
27
3
2
0.06 deg
Cu
z
1
y
y
Pt
0
0
z
2
4
0.12o x 0.6o
6
t (nm)
8
10
12
All consistent with fs Spin-Transfer Torque pulse
M


 
H eff    M
depending on

x
maj. / min.
sin(t   )
 = 0 or 
 = ±/2
The ∆K artefact
Overview phases
28
29
sin(t   )
M
STT short
HK
Heff
HK
M
z
y
STT long
∆K short
∆K long
Maj.
Min.
Maj.
Min.
dec.
inc.
dec.
inc.
IP
π
π
π
0

π
π
π
OOP
π
π
0
π

π
π
π
Heff
majority
minority
decrease
M
Ha
Ha
experiment:
x
Analysis
Analysis
Field dependence
Angle dependence
 = -/2,  = -/2,
STT
∆K
STT
∆K
STT
STT
∆K
Top layer
Frequency
Amplitude
Phase
Top layer
Bottom layer
STT
K




Frequency

Phase

Amplitude
STT
K






Frequency
Amplitude
Phase
STT
Anisotropy






Precessions top layer
consistent with USTT!
∆K inefficient due to large Tc
Precessions bottom layer consistent
with ∆K,
USTT not visible
(poor sensitivity bottom layer +
overwhelmed by ∆K)
Possible origin of laser-induced STT
Calculated transient temperature profile
32
33
7 K / nm
Optical absorption
Laser-induced torques:
Transient T profiles
t = 0.1 ps
t = 0.5 ps

   sd   T   sc   K
t = 500 ps
t = -0.5 ps
 Super-diffusive spin currents (+ screening currents)
 Spin-Dependent Seebeck effect due to large T gradients
7 K / nm
Calculated spin currents due to SDS
The alternative: Super-diffusive
Transient T profiles
M
Continuity equations:
up/down
spins
absorbing
up/down
spins within
1 nm
Spin dependent current with T gradient:
Drift diffusion
34
35
 Experiment (top IP layer):
 Amplitude of precession:
Spin transfer corresponds to 2% of demagnetization
OOP layer
 Phase of precession:
Majority spin flow from bottom OOP layer
 Both consistent with super-diffusive spin current
( + screening spin polarized charge current)
M
Canting IP layer = 3 mdeg
experiment: 60 mdeg

Similar results for Co / spacer / Pt/Co/Pt
Outline
36
But more difficult
to interpret…
(combination
of
∆K and STT)
37
 Introduction
 Local magnetization dynamics
 Spin transfer (super-diffusive)
 Laser-induced spin transfer torque
 Why efficient
 Our recent experiments
 A first interpretation:
Super-diffusive spin currents or Spin-Dependent Seebeck?
 Also positive results and rich data for Co/Pt/Au/Co
Co
Au
Co
 Its importance for local magnetization dynamics
Pt
A decisive experiment?
Schellekens, Verhoeven et al. APL 2013
… nothing …
38
Schellekens, Verhoeven et al. APL 2013
39
Front pump
Back pump
Conclusions Questions
Acknowledgements
40
41
 Thermo-dynamical model using Elliott-Yafet process
seems to explain all local magnetization dynamics
 Did we prove minor rol of spin transport?
And what
about using
the effect?
 Or are we too naïve?
 What about all the questions I posed?
 Laser-induced spin transfer can play a role (?)
 And can exert a STT (?)
 Likely of super-diffusive (and not SDS) origin (?)





Are there pitfalls in our interpretation?
Is our optical / thermal / SDS correct?
How to model super-diffusive STT more accurately?
What about back flow / screening charge?
Etc. etc.
The ∆K artefact
Is it magnetism at all in Ni?
42
Koopmans et al. PRL 2000
Regensburger et al. PRB 2000
Guidoni et al. PRL 2002
MO rotation MO ellipticity
Ha
M
Haff
z
y
x
Ha
HK
demag. field
shape anisotropy
M
HK
ind. MO response (%)
~
 =  + i 
0
-5
0
1
2
delay (ps)
43