1
Supplementary Information
Fast current-induced domain wall motion controlled by the Rashba effect
Ioan Mihai Miron 1,2*, Thomas Moore 1,3, Helga Szambolics 1, Liliana Daniela Buda-
Prejbeanu1, Stéphane Auffret 1, Bernard Rodmacq 1, Stefania Pizzini 3, Jan Vogel 3, Marlio
Bonfim4, Alain Schuhl 1,3 and Gilles Gaudin 1
1 SPINTEC, UMR-8191, CEA/CNRS/UJF/GINP, INAC, F-38054 Grenoble, France 2 Catalan Institute of Nanotechnology (ICN-CSIC), UAB Campus, E-08193 Barcelona, Spain 3 Institut Néel, CNRS/UJF, B.P. 166, F-38042 Grenoble, France 4Departamento de Engenharia Elétrica, Universidade Federal do Paraná, Curitiba, Paraná, Brazil
DW dynamics observed by micromagnetic simulations
The goal of these micromagnetic simulations is to detail the understanding of the DW
dynamics observed experimentally. Three questions will be addressed:
i) Can the terminal DW velocity be measured using ultra-short current pulses?
ii) What is the influence of HR on the DW stability and velocity?
iii) How does a single DW chirality reversal driven by HR influence the DW displacement?
The extended Landau Lifshitz Gilbert (LLG) equation
In order to model the effect of current on the magnetization1, the Landau Lifshitz
Gilbert (LLG) equation will include besides the effective Rashba field2 (HR), the spin transfer
torque (STT)3. The Extended LLG equation reads:
( )[ ] ( ) ( )[ ]MuMMuMMMHHMReff ∇⋅×+∇⋅−
∂∂×+×+=
∂∂ βαγ
SS MtMt1
0 , (1)
Here M is the local magnetization, 0γ the gyromagnetic ratio, Heff the magnetic field
including contributions from the external, anisotropic, and magnetostatic fields, HR is the
Fast current-induced domain-wall motion controlled by the Rashba effect
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nmat3020
natuRe mateRials | www.nature.com/naturematerials 1
© 2011 Macmillan Publishers Limited. All rights reserved.
2
effective Rashba field, α the Gilbert damping paramater, β is the ratio of the nonadiabatic to
adiabatic spin torque, and 2
Be
s
gµ PeM
=u j where je is the current density. Note that HR is
included on equal footing with the other effective fields occurring in the ferromagnet.
DW motion opposed to electron flow, in the context of STT
DW motion opposed to the carrier flow4 was also observed in (Ga,Mn)As5, where it
was attributed to negative current polarization generated by the antiferromagnetic exchange
coupling between the itinerant and localized spins, and also in the PtCoPt trilayers6, where it
was assumed to be caused by a negative current polarization. Since it is not the purpose of this
work to understand the origin of this phenomenon, we will analyze the DW dynamics within
all scenarios that could explain this observation.
In the framework of the CIDM theory1 the inverse motion could have three possible
origins:
i) Both spin-torque components are negative. Physically, in the context of the present
current induced domain wall motion (CIDM) theory this would correspond to a negative
current polarization (P<0). In this case, the DW dynamics are expected to be identical to the
ones obtained for a positive current polarization, only that the displacements occur in the
opposite direction.
ii) The non-adiabatic spin-torque is negative while the adiabatic one is positive. This
corresponds to positive current polarization (P>0), but negative β. In this case the direction of
motion is reversed only for motion occurring below the Walker breakdown (WB).
iii) The non-adiabatic component is positive while the adiabatic one is negative. In this
case both the current polarization and β are negative (P<0, β<0). With respect to the previous
scenario, the sign change of the current polarization reverses the behavior: the DW moves
against the electron flow only above the WB.
Since it is impossible to a priori distinguish between them, both physical possibilities
(negative P and negative β) explaining the direction of DW motion are analyzed by
micromagnetic simulations.
2 natuRe mateRials | www.nature.com/naturematerials
SUPPLEMENTARY INFORMATION doi: 10.1038/nmat3020
© 2011 Macmillan Publishers Limited. All rights reserved.
3
Values of the physical parameters used in simulations
Most of the required values, such as the saturation magnetization magnetocristalline
anisotropy non-adiabaticity of STT and Rashba field were determined and reported
previously. The value of the Rashba field was previously measured using 100 ns current
pulses2. Since in the present work we use ultra-short pulses, we repeat the experiment to
determine the value of the effective HR for different pulse durations. Our observations are
summarized in Figure S1. As we apply shorter pulses, the current density necessary for
domain nucleations increases. This enhancement correlates with the decrease of the effective
value of HR. We believe that, this variation could be attributed to something as common as
thermal effects. The existing theoretical models predict that HR varies inversely proportional
to MS. As MS depends on the wire temperature, which in turn depends on the pulse duration
and intensity, thermal effects could account for the observed dependence of HR on pulse
duration. Note that this dependence is not central for the claims of our work. The DW motion
experiments prove that HR is sufficiently strong to prevent WB, and simulations show that its
presence doesn’t affect the velocity value of the steady motion regime. For the micromagnetic
study of the DW dynamics we use the smallest value of HR determined experimentally.
The value of the damping parameter was determined from the slope of the constant
mobility regime in the field induced DW motion. The slope can be fitted in principle with two
Figure S1. Dependence of the Rashba field intensity on pulse duration; in red the current density required for nucleation in half of the wires is plotted as a function of pulse length. In black we plot the variation of the Rashba field efficiency normalized to a current density of 1012A/m2. The increase of the current density required for nucleation correlates well with the loss of efficiency of the Rashba field.
natuRe mateRials | www.nature.com/naturematerials 3
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nmat3020
© 2011 Macmillan Publishers Limited. All rights reserved.
4
laws, one corresponding to the steady flow regime (α
γ Δ= 0Hv , where Δ is the wall width),
and a second corresponding to the turbulent DW motion ( 10
−+Δ
=αα
γHv ). In the previous
work by Metaxas et al.7 the DW mobility was in a range that could correspond to both these
scenarios, so that the damping value could not be uniquely determined. In our case, the
velocity can only be fitted with the law corresponding to the steady flow regime, and the
damping is uniquely determined to be approximately α ≈ 0.5 for a DW width of 4.2 nm8.
The physical parameters considered for the micromagnetic simulations are the
following: μ0Ms= 1T; AEX = 1011 J/m; KANIS = 1.19 106 J/m3 P = ±1 ; β = ±1; HR = 0.25
T/1012 A/m2; α = 0.5;
A domain wall was created in the middle of the wire by calculating the equilibrium
configuration starting from a sinusoidal configuration of the magnetization. Equation 1 was
implemented to simulate the effect of current pulses on the magnetization. The pulse shape is
similar to the measured pulses (Figure S2).
a) Measurement of the terminal DW velocity using current pulses.
The use of ultra-short current pulses, far from quasi-static conditions, requires
considering the transient regimes of DW motion.
In Figure S3 we plot an example of DW displacements observed for a 0.9 ns pulse of
6×108A/m2. Note that the DW does not actually have time to reach its terminal velocity
during the pulse (Figure S3). Moreover, its motion continues after the pulse has ended. This is
due to its structural transformation, measured by the x component of the DW’s magnetization.
Figure S2: Shape of the injected current pulses. a, used in the experiment b, used in the micromagnetic simulations
4 natuRe mateRials | www.nature.com/naturematerials
SUPPLEMENTARY INFORMATION doi: 10.1038/nmat3020
© 2011 Macmillan Publishers Limited. All rights reserved.
5
However, the velocity determination that we use does not seem to be affected: the gradient of
displacement vs. pulse length is the same as the terminal velocity of the steady flow regime
induced by dc current. This happens because the DW distortion is “elastic” and the delay
accumulated during the pulse is compensated by DW motion occurring after the pulse.
In the inset of Figure S3 we plot the x component of the magnetization, which
measures the DW deformation. It increases during the pulse, and as soon as the pulse ends, it
starts decreasing. Together with this variation, the DW motion continues after the pulse and
only ends when the DW deformation ceases. The effect of HR is to limit the deformation of
the wall structure. Consequently the DW moves more uniformly during the pulse and less
after the end of the pulse.
b) Effect of HR on the DW stability
Figures S4 shows the DW displacement as a function of the pulse length for various
values of current densities and with and without HR. In the absence of HR, the variation of the
displacement with pulse length seems to have, besides a monotonic trend, a “random”
component. The apparent noise of this dependence is related to the DW transformations above
the WB. While in the case of steady motion the DW deformation is reversible and therefore
the two transient regimes (the beginning and the end of the pulse) compensate, above the WB
this is no longer the case. The periodic transformations of the DW structure are irreversible
and the transient regime at the end of the pulse, determined by the final DW deformation, is
dictated only by the pulse duration and amplitude. Therefore, this apparently random
Figure S3: The magnetization variation during a 0.9 ns pulse. j = 6×108A/m2. The black curve is the DW position calculated in the scenario of P < 0. In red, we plot the same variation, for the scenario that includes HR. The effect of HR is to produce a steadier motion during the pulse and to reduce the residual displacement (occurring after the end of the pulse). HR regulates the displacement by limiting the DW distortion, measured here by the MX component (along the wire).
natuRe mateRials | www.nature.com/naturematerials 5
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nmat3020
© 2011 Macmillan Publishers Limited. All rights reserved.
6
component is deterministic and does not cancel by averaging. The fact that we do not observe
such variations experimentally further supports the conclusion that DW motion is steady, not
turbulent.
When HR is included, the displacements lose their “noisy” character as the DW
transformations no longer occur. Once again we find a “clean” linear dependence and the DW
mobility obtained from the linear regression of the displacement vs. pulse length corresponds
to the terminal velocity values expected from the steady motion regime (Figure S4).
Figure S4: DW displacement vs. pulse duration. a) and b) the effect of HR is not included. c) and d) once HR is included the dependence becomes perfectly linear and the difference between the two scenarios disappears. The slope of the gray dotted line on the four graphs is the theoretically8 predicted (v =β/α·u) DW velocity for a current density of 250·108 A/m2. Note that the inclination of the gray line is the same as the slope of the DW displacement vs. pulse length (for 250·108 A/m2) only in the presence of HR. Otherwise the DW displacement is much smaller.
6 natuRe mateRials | www.nature.com/naturematerials
SUPPLEMENTARY INFORMATION doi: 10.1038/nmat3020
© 2011 Macmillan Publishers Limited. All rights reserved.
7
c) DW chirality reversal by HR
Figure 4c of the main text shows the DW displacement obtained for the two possible
orientations of the DW chirality. First, it is set along the Rashba field, and after opposed. The
experimental observation of the decrease in the displacement produced by the chirality
reversal is reproduced by simulations. Once again, we find not only good qualitative
agreement, but also good quantitative similarity. Note that the two scenarios of negative β and
negative P give identical results.
Figure S6: Distribution of the measured DW displacements for a current density of 2×1012A/m2; a, 0.98 ns pulse; b, 1.18 ns; and c, 1.3 ns; Note that changes of the average displacement don’t influence much the dispersion. This indicates that the distribution is associated to the rise and fall time of the pulse, which is the same for all pulse lengths.
Figure S5: Delay time dependence on the injected current density. The delay time is defined as the intercept of the linear interpolation of displacement vs. pulse length with the abscissa (Figure 4a). As the current density is increased, the delay time decreases, and saturates for values close to the rise and fall time of the pulse (0.3 ns).
natuRe mateRials | www.nature.com/naturematerials 7
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nmat3020
© 2011 Macmillan Publishers Limited. All rights reserved.
8
Supplementary References:
1 Thiaville, A. Nakatani, Y. Miltat, J. & Suzuki, Y. Micromagnetic understanding of current-
driven domain wall motion in patterned nanowires. Europhys. Lett. 69, 990 (2005).
2 Miron, I.M. Gaudin, G. Auffret, S. Rodmacq, B. Schuhl, A. Pizzini, S. Vogel, J. &
Gambardella, P. Current-driven spin torque induced by the Rashba effect in a ferromagnetic
metal layer. Nat. Mater. 9, 230 - 234 (2010).
3 Szambolics H., Toussaint J.C., Marty A., Miron I.M., & Buda-Prejbeanu L.D. Domain wall
motion in ferromagnetic systems with perpendicular magnetization JMMM 321, 1912-1918
(2009)
4 Moore, T.A. et al. High domain wall velocities induced by current in ultrathin Pt/Co/AlOx
wires with perpendicular magnetic anisotropy. Appl. Phys. Lett. 93, 262504 (2008); ibid. 95,
179902 (2009).
5 Yamanouchi, M. Chiba, D. Matsukura, F. Dietl, T. & Ohno, H. Velocity of domain-wall
motion induced by electrical current in the ferromagnetic semiconductor (Ga,Mn)As. Phys.
Rev. Lett. 96, 096601 (2006).
6 Lee, J.C. Kim, K.J. Ryu, J. Moon, K.W. Yun, S.J. Gim, G.H. Lee, K.S. Shin, K.H. Lee,
H.W. Choe, S.B. Roles of adiabatic and nonadiabatic spin transfer torques on magnetic
domain wall motion arXiv:1006.1216v1
7 Metaxas, P.J. et al. Creep and flow regimes of magnetic domain-wall motion in ultrathin
Pt/Co/Pt films with perpendicular anisotropy. Phys. Rev. Lett. 99, 217208 (2007).
8 Miron, I.M. Zermatten, P.-J. Gaudin, G. Auffret, S. Rodmacq, B. & Schuhl, A. Domain wall
spin torquemeter. Phys. Rev. Lett. 102, 137202 (2009).
8 natuRe mateRials | www.nature.com/naturematerials
SUPPLEMENTARY INFORMATION doi: 10.1038/nmat3020
© 2011 Macmillan Publishers Limited. All rights reserved.