Structured Materials: Magnetic Structures
and Electronic Properties
Z. Q. Qui,
Electron beam stimulated spin reorientation
T. L. Monchesky,a)J. Unguris, and R. J. Celotta
Electron Physics Group, National Institute of Standards and Technology, Gaithersburg, Maryland 20899
?Presented on 15 November 2002?
Using scanning electron microscopy with polarization analysis, we observed the electron beam
induced switching of the magnetic state of epitaxial single-crystal Fe?110? films grown on
atomically flat cleaved GaAs?110?. For low film thickness the magnetization lies along the ??110?
in-plane direction, while above a thickness of 19 monolayers, the ground state magnetization
configuration switches to the ?001? in-plane direction. If Fe films are grown to a thickness greater
than the critical thickness of the reorientation, the magnetization is caught in a metastable state,
oriented along ??110?. We discovered that we can locally switch the metastable state to the stable
?001? direction by irradiating the metastable magnetic state with a suitable electron current density.
The reversal proceeds by the nucleation and growth of lancet-shaped domains that move in discrete
jumps between pinning sites. Our results show that there is a permanent reduction of the strength of
defect sites without a permanent change in the overall anisotropy. We demonstrate how an electron
beam can be used to locally control domain structure. ?DOI: 10.1063/1.1556250?
We have discovered that epitaxial ultrathin Fe films can
be prepared in a metastable magnetic state that can be locally
switched by the electron beam from a scanning electron mi-
croscope ?SEM?. The surface anisotropy of Fe/GaAs?110?
has an easy axis along the in-plane ??110? direction, perpen-
dicular to the ?001? bulk easy-axis direction. Below a critical
thickness, tcrit, the stable direction of the magnetization
points along ??110?. Above tcrit, ?001? becomes the stable
direction.1However, when we grow a film through tcrit, we
find the magnetization is frozen in a ??110? metastable state.
Subsequently, we discovered that we can switch the magne-
tization into the stable ?001? configuration by irradiating the
sample with an electron beam of appropriate strength.
The GaAs?110? surface was prepared by cleaving either
0.5- or 1.0-mm-thick wafers in situ. Scanning tunneling mi-
croscope measurements show that surfaces prepared in this
fashion are atomically flat over areas several micons wide.2
The two-dimensional reflection high-energy electron diffrac-
tion ?RHEED? pattern composed of sharp spots demonstrated
the high surface quality of the cleaved GaAs surface. An
electron beam evaporator deposited Fe wedges on room tem-
perature GaAs?110? with a shutter that varied the exposure of
the Fe flux along the length of the substrate. All wedges in
this study had their thickness gradient along ??110?. The
film thickness was calibrated by spatially resolved RHEED
intensity oscillations.3Scanning electron microscopy with
polarization analysis ?SEMPA?4was used to image the mag-
netic structure of our films at remanence. SEMPA simulta-
neously measures two in-plane components of the magneti-
zation and the topography of the sample surface. Beam
currents ranged between 1 nA and 10 nA and the beam volt-
age was fixed at 10 keV.
When we grew Fe wedges on GaAs?110? we expected to
see a reorientation transition from ??110? to ?100? near a
thickness of 24 monolayers ?ML?, assuming the surface mag-
netocrystalline anisotropy of Au/Fe?110? is not substantially
different than that of a vacuum/Fe?110?.5SEMPA measure-
ments made directly following growth showed this not to be
the case. The magnetic moments of nearly the entire sample
pointed along the ??110? direction. The dashed curve in Fig.
1 shows that the ??110? direction is the energy minimum of
the anisotropy energy for the thinnest films. The magnetiza-
tion is trapped in a metastable state when the thickness is
increased above tcritsince the global minimum for low thick-
nesses changes to a local minimum, as illustrated by the solid
curve in Fig. 1.
a?Electronic mail: email@example.com
FIG. 1. The magnetocrystalline anisotropy energy of Au/Fe/GaAs?110? as a
function of angle with respect to the ?100? crystallographic direction, calcu-
lated from the anisotropy measurements of Ref. 5. The solid line is the
energy of a 26 ML Fe film (t?tcrit) and the dashed line is that of a 16 ML
Fe film (t?tcrit). The plots are scaled for comparison.
JOURNAL OF APPLIED PHYSICSVOLUME 93, NUMBER 10 15 MAY 2003
We also observed a few regions with ?100? orientation.
These regions are typically found next to macroscopic de-
fects. We hypothesize that the defects nucleate ?100? oriented
domains either by changing the anisotropy over a large
enough region or by the magnetostatic field nucleating clo-
sure domains. These stable domains are typically seen at
higher thickness, where the anisotropy energy barriers are
The electron microscope beam was then used to enlarge
these stable domains at the expense of the metastable do-
mains. The electron beam stimulates the reorientation; the
domain wall moves to a position along the edges of the re-
gion the electron beam has scanned. As the domain wall
approaches the thinner end of the wedge, at some point,
which varies from sample-to-sample, it stops moving into the
electron beam scanned region. If true equilibrium were
achieved, ttrans?the minimum wedge thickness at the domain
wall position? would equal tcrit. A ttrans?19ML was the low-
est value observed in this system. Figures 2?a?, 2?b?, and 2?c?
shows the motion of a stable ?100? domain under the influ-
ence of the electron beam. The final distortion of the once
straight magnetic domain wall, shown in the larger area scan
of Fig. 2?d?, was achieved by scanning over an 83 ?m ? 63
?m region for 3 h with a 1 nA current.
The domains in motion in Fig. 2 exhibit lancet-shaped
supplementary domains because of magnetostatics. A wall at
a 45° angle to both the ??110? and ?100? directions is pre-
ferred because it keeps the normal component of the magne-
tization continuous across the wall. In place of one wall
along ?100?, two different wall orientations combined to cre-
ate a lancet: one wall is at an angle of roughly 15° from the
45° or zero magnetic charge orientation, and the other wall is
at an angle of roughly ?15°. Unlike the out-of-plane spin-
reorientation transition,6there is no collapse of the magnetic
domain size at ttrans.
The wall velocity depends on film thickness. In Figs.
2?a?, 2?b?, and 2?c? the wall velocity decreased from 4 to 2
nm/s while moving down the wedge from a film thickness of
36 ML to 26.5 ML, while under the influence of a 1 nA
beam. A velocity of 6.4 nm/s was observed at 55 ML. The
reduction in wall velocity with film thickness correlates with
the increase in the anisotropy energy barrier that separates
the metastable and stable configurations. This suggests that
the strength of the pinning sites is related to the effective
anisotropy that is determined by the bulk anisotropy plus the
surface anisotropy weighted by the film thickness. The mo-
tion is not smooth, i.e., the wall makes jumps between pin-
ning sites at a speed much faster than can be imaged with
SEMPA. The velocities reported here are the average dis-
tances the wall travels per unit time as it moves down the
The wall velocities at a given thickness were very sen-
sitive to the quality of the substrate and varied by up to a
factor of 20 from sample-to-sample. For a good cleave with
mirror like surfaces there are virtually no defects visible to
the electron microscope. In this case, the wall velocities are
over 100 nm/s for a 40 ML thick film and a 1 nA beam. This
is in contrast to the sample of Fig. 2 that has a substrate
surface with step bunches visible to the SEM and a wall
velocity in the 40 ML region of 4.6 nm/s. This large sample-
to-sample variation in wall velocity suggests that the domi-
nant pinning sites are related to the Fe–GaAs interface. The
switching is also sensitive to the electron beam parameters,
although a given set of electron beam parameters produces
different results depending on the substrate condition. In
general, increasing the current or current density will in-
crease the wall velocity. We have also observed wall motion
that continues at a slower rate independent of observation by
the electron beam; in Fig. 2 this drift is 0.8 nm/s for a 36 ML
To determine whether adsorbates play a role in the elec-
tron beam switching, we covered the Fe wedge with 0.6 nm
of Au. When Fe is grown on GaAs, As is known to segregate
to the Fe surface.7Furthermore, we find carbon and oxygen
on the surface of Fe after long exposure to the residual gases
in the vacuum chamber. The addition of the Au layer
dropped the wall velocity by an order of magnitude suggest-
ing that the Au altered domain wall pinning sites at the Fe
surface. Since the Au cap does not prevent the electron beam
from being able to move the domain wall, adsorbate diffu-
sion or desorption can be ruled out as a switching mecha-
The effect of an external field was investigated by first
applying a field along the ?100? direction of a 0ML to ML
wedge. The thickness of the wedge increased by 1 ML every
6.51?0.02 ?m along ??110?. A 110 kA/m ?1.4 kOe? field
pushed the ttransto 33 ML. A small raster scan, indicated by
the white box in Fig. 3?b?, further reduced ttransto 28.5 ML
FIG. 2. ?a?–?c? A sequence of 83 ?m ? 63 ?m SEMPA images of a domain
wall moving down an Fe wedge on GaAs?110?. The electron beam does not
uniformly illuminate the region of interest: a 1 nA electron beam is rastered
with discrete steps of 0.3 ?m with a spot size of roughly 0.02 ?m. The
region above the wall is stable and the region below the wall and above a
thickness of roughly 20 ML is metastable. The electron beam is rastered
from top to bottom and then right to left. The frames were collected con-
tinuously, each requiring 10.5 min. The time between the frames shown here
is 84 min. The arrows in ?a? indicate the direction of the magnetization. ?d?
A 175 ?m ? 133 ?m SEMPA image of the region investigated in ?a?–?c?,
indicated by the dashed rectangle.
8242J. Appl. Phys., Vol. 93, No. 10, Parts 2 & 3, 15 May 2003Monchesky, Unguris, and Celotta
over a 71 ?m length region. This shows that a 110 kA/m Download full-text
field is not sufficient to bring the sample into equilibrium and
indicates that the pinning fields of the 33 ML region are very
roughly of that strength.
We then investigated the reversibility of the electron
beam induced switching by attempting to switch a stable
state back into a metastable one. A 110 kA/m magnetic field8
was applied along the ??110? direction but it did not push
the ttransto higher film thickness. Surprisingly, the field re-
duced the metastable domain by moving ttransto 28.5 ML
over the entire length of the sample, as illustrated in Fig.
3?c?. Also, evidence of the previous 71 ?m ? 54 ?m scan is
present in the form of a reverse domain, shown in Fig. 3?c?,
further demonstrating that the electron beam permanently al-
tered the sample in this region.
The wedge was then capped with 0.6 nm Au in order to
compare the ttransto a previously published critical thickness.
A magnetic field applied along ?100? was used to help the
system approach equilibrium. SEMPA images show that
ttrans?25.4?0.3ML. The wall that separates the two do-
mains zigzags about this average with a peak-to-peak ampli-
tude of 1.4 ML. This transition agrees well with tcrit?24
?1.2ML for Au/Fe/GaAs?110? as determined by magneto-
optic Kerr effect measurements.5Furthermore, a 10 nA cur-
rent could not change ttrans, an additional indication that
ttrans?tcrit. This demonstrates that the electron beam does
not permanently affect the average anisotropy of the sample,
but instead modifies the anisotropy of a relatively small num-
ber of pinning sites. This finding is in contrast to electron
stimulated switching of the magnetization of Co films on
Pt?111?,9where it was suggested that a 10 nA beam was able
to modify the average anisotropy of the Co by as much as
There are two switching mechanisms that can be imme-
diately ruled out. The first is the magnetic field from the
electron beam. For a 1 nA beam, the maximum magnetic
field is roughly 0.2 ?T. This is several orders of magnitude
smaller than the coercive field even for the best quality Fe
crystals. The second mechanism that can be ruled out is elec-
tron beam heating of the crystal lattice, also ruled out in Ref.
9. Based on calculations presented in Ref. 10, the maximum
possible increase in temperature of the surface of the crystal
is of the order of 0.1 K for a 1 nA, 10 keV beam.
We speculate that there may be two processes involved
in the electron beam induced switching of the sample: one
process permanently reduces the pinning strength, and the
other process temporarily weakens the remaining anisotropy
energy barrier. One possible origin of the temporary change
in anisotropy is excitations in the Fe film created by the
electron beam. Magnetocrystalline anisotropy is known to be
very sensitive to the Fermi level filling.11It has been recently
shown that the electron excitations from a femtosecond laser
pulses can alter the anisotropy sufficiently to create a spin-
reorientation transition.12In a similar fashion, the electronic
excitation from an electron beam will populate states above
the Fermi level that can alter the anisotropy. This possibility
is consistent with our recent measurements of bcc Co grown
reorientation transition in contrast to previous reports.14
When the Co thickness reaches 8 ML there is a reorientation
from ??110? to ?100? without an in-grown metastable state:
no domain wall motion was observed. Based on calculations
and spin-polarized photoemission, it has been shown that the
band structure of bcc–Co is very similar to that of Fe with
one additional electron.15If the electronic excitations pro-
duced by the incident electron beam cause the Fe to tempo-
rarily behave similarly to Co, they will remove the meta-
stable state and allow the Fe to reach equilibrium.
In conclusion we have demonstrated that metastable
states can be created in Fe/GaAs?110? by growing epitaxial
films to thicknesses greater than tcrit. These states can be
switched with the help of an electron beam. The electron
beam permanently reduces the strength of the pinning sites
without permanently affecting the overall magnetocrystalline
anisotropy that creates the metastable and stable configura-
This work is supported in part by the Office of Naval
Research. T.L.M. would like to acknowledge financial sup-
port from the Natural Sciences and Engineering Research
Council of Canada.
1G. A. Prinz, G. T. Rado, and J. J. Krebs, J. Appl. Phys. 53, 2087 ?1982?.
2J. A. Stroscio ?private communication?.
3J. Unguris, R. J. Celotta, and D. T. Pierce, J. Appl. Phys. 75, 6437 ?1994?.
4D. T. Pierce, J. Unguris, and R. J. Celotta, Ultrathin Magnetic Structures
?Springer, New York, 1994?, pp. 117–147.
5R. Ho ¨llinger et al., J. Appl. Phys. 89, 7136 ?2001?.
6M. Speckmann et al., Phys. Rev. Lett. 75, 2035 ?1995?.
7T. L. Monchesky et al., Phys. Rev. B 60, 10242 ?1999?.
8This is an order of magnitude larger than the coercive field along ??110?.
9R. Allenspach et al., Appl. Phys. Lett. 73, 3598 ?1998?.
10L. Reimer, Scanning Electron Microscopy: Physics of Image Formation
and Microanalysis ?Springer, New York, 1998?.
11G. H. O. Daalderop et al., Phys. Rev. B 50, 9989 ?1994?.
12F. Kronast et al. ?unpublished?.
13T. L. Monchesky, J. Unguris, and R. J. Celotta ?unpublished?.
14G. A. Prinz, Phys. Rev. Lett. 54, 1051 ?1985?.
15G. A. Prinz et al., J. Appl. Phys. 57, 3024 ?1985?.
FIG. 3. The influence of a magnetic field on an Fe wedge/GaAs?110?. ?a?
304 ?m?271 ?m secondary electron intensity of the wedge. ?b? The corre-
sponding SEMPA image after the sample was placed in a 110 kA/m ?1.4
kOe? field along ?100?. The 71 ?m?54 ?m region previously irradiated by
the electron beam with 5.92?10?6C is indicated by the white dashed line.
?c? The SEMPA image of the same region in ?b? after the sample was placed
in a magnetic field of 110 kA/m, applied along ??110?.
8243 J. Appl. Phys., Vol. 93, No. 10, Parts 2 & 3, 15 May 2003Monchesky, Unguris, and Celotta