PHYSICAL REVIEW B 89, 174421 (2014)

Reversal mechanism, switching field distribution, and dipolar frustrations in Co/Pt bit pattern
media based on auto-assembled anodic alumina hexagonal nanobump arrays
T. Hauet,1 L. Piraux,2 S. K. Srivastava,2,3 V. A. Antohe,2 D. Lacour,1 M. Hehn,1 F. Montaigne,1 J. Schwenk,4 M. A. Marioni,4
H. J. Hug,4,5 O. Hovorka,6 A. Berger,7 S. Mangin,1 and F. Abreu Araujo2
1
Institut Jean Lamour, Université de Lorraine & CNRS, Vandoeuvre lès Nancy, F-54506, France
2
Institute of Condensed Matter and Nanosciences, Université catholique de Louvain, Place Croix du Sud, B-1348, Louvain-la-Neuve, Belgium
3
Amity Institute of Nanotechnology, Amity University, Noida, India-201303
4
Empa, Swiss Federal Laboratories for Materials Science and Technology, CH-8600 Dübendorf, Switzerland
5
Institute of Physics, Universität Basel, CH-4056 Basel, Switzerland
6
Faculty of Engineering and the Environment, University of Southampton, Southampton, SO17 1BJ, UK
7
CIC nanoGUNE Consolider, Tolosa Hiribidea 76, E-20018 Donostia-San Sebastián, Spain
(Received 23 January 2014; revised manuscript received 1 April 2014; published 19 May 2014)

We fabricated a perpendicularly magnetized bit pattern media using a hexagonally close-packed auto-assembled
anodic alumina template with 100 nm and 50 nm periods by depositing a Co/Pt multilayer to form an ordered
array of ferromagnetic nanodots, so-called nanobumps. We used Hall resistance measurements and magnetic
force microscopy to characterize the dot-by-dot magnetization reversal mechanism under applied field. The role
of interdot exchange coupling and dipolar coupling are investigated. Then we focus on separating the various
origins of switching field distribution (SFD) in this system, namely dipolar interactions, intrinsic anisotropy
distribution, and template packing faults. Finally we discuss the influence of triangular dipolar frustrations on
the energy stability of demagnetized and half-switched states based on an Ising model, including local exchange
coupling. The impact of SFD and lattice defects lines between misoriented ordered domains on the magnetic
configurations is studied in detail.

DOI: 10.1103/PhysRevB.89.174421 PACS number(s): 75.70.−i, 75.75.−c, 75.10.−b, 75.50.Ss

I. INTRODUCTION order needs to be promoted by templates using lithography

[11].
Ordered arrays of isolated magnetic nanostructures are of
In 2012, we proposed a method for designing low-cost
considerable interest in many fields of fundamental physics
densely packed magnetic particle assemblies on two dimen-
as well as in nanotechnology implementations. For the for-
sional (2D) curved substrates with perpendicular magnetic
mer, artificially frustrated nanomagnets have recently opened
anisotropy (PMA) [12]. This process involves the growth of
new ways of investigating complex phenomena that usually magnetic nanocaps on the back of nanoporous anodic alumina
occur in bulk systems such as spin ice with monopolelike (AAO) templates, which are a typical self-ordered nanopore-
excitations [1,2]. Regarding the latter, the best example is array material formed by the electrochemical oxidation of Al in
probably the research on bit pattern media (BPM), which acidic solutions [13,14]. In our previous report [12], as a proof
may lead to improved storage density of hard disks in of concept, we studied a hexagonally close-packed array of
the near future [3,4]. In both of the above examples, each [Co/Pt] nanocaps deposited on bumps having about a 100 nm
artificially fabricated magnetic nanostructure is capable of lateral modulation period. We showed that the magnetic
storing one uniform macrospin. Fabrication schemes involve hysteresis loops from flat films and bumpy films are very
either lithographic patterning [5,6] or a combination of e- different. Moreover, we observed specific lines of uniformly
beam lithography methods with other techniques such as magnetized bumps after out-of-plane demagnetization that we
electrodeposition [7] and self-assembling block copolymers explained considering dipolar frustration between bumps. We
[8]. However, lithographic techniques are expensive and deduced that direct exchange coupling between bumps must
time-consuming. Therefore, it is of interest to develop a be drastically reduced, as described in Ref. [15].
nonlithographic patterning method based on self-assembly Here we present a study of the magnetization reversal
as an alternative route in order to develop a fast and cheap mechanisms of a thin Co/Pt multilayer deposited on a closely
process capable of producing dense arrays of nanostructures packed nanobump array with 100 nm and 50 nm periods.
over millimeter-wide areas with precise long-range order. Using extraordinary Hall effect (EHE) and superconducting
One well-known approach to fabricate BPM is based on quantum interference device-vibrating sample magnetometer
self-assembly of colloidal nanospheres [9]. However, several (SQUID-VSM) measurements, we provide a macroscopic
problems associated with the nanosphere technique were characterization of reversal mechanisms in terms of the
pointed out. The presence of dislocations and domains is switching fields and switching field distribution (SFD) as a
difficult to prevent. The cleanness and hydrophilic properties function of field angle and temperature. The SFD is quantified
of the substrate surface are critical in making well-ordered in two ways: direct hysteresis loop differentiation and the
colloidal crystal films [10]. In addition, attractive forces so-called H (M, M) method [16,17]. This method requires
between the spheres prevent the formation of a layer of spheres as an input the measurement of partial reversal curves and
with good order when considering diameters below 50 nm so allows the disentanglement of the intrinsic SFD resulting from
that only short-range order can be obtained or their long-range magnetic and structural inhomogeneity of the bumps from

1098-0121/2014/89(17)/174421(13) 174421-1 ©2014 American Physical Society

T. HAUET et al. PHYSICAL REVIEW B 89, 174421 (2014)

the effect of dipolar coupling in between the bumps. Then, array of magnetic caps and magnetization pointing up or
using magnetic force microscopy (MFM), we demonstrate the down perpendicular to the substrate surface [12]. We estimated
bump-by-bump magnetization reversal and hereby reveal the the mean interbump distance of 105 nm and a full width at
competition between interdot exchange and dipolar couplings. half-maximum of 15.5 nm, as described in Ref. [12]. Height
Easy and hard switchers are detected, and their influence on modulation profiles extracted from atomic force microscopy
magnetic states during field cycles or after demagnetization (AFM) images show a maximum bump to bump height
is discussed. Finally, we investigated features of AAO lattice variation of 5 nm (only 1 nm maximum in region with no
stacking faults and their consequences on nanobump arrays in lattice defects) and an average bump height distribution of
conjunction with their dipolar interactions and SFDs. about 2 nm over a few microns, but larger variations (up to
50 nm) can be observed from region to region for areas of tens
of microns.
II. EXPERIMENTAL DETAILS
A. Sample fabrication and structural features
B. Magnetic characterization methods
To obtain ordered arrays of ferromagnetic nanodots, we
deposited thin Co/Pt multilayers onto the bumpy barrier- The extraordinary Hall resistivity measurements were
layer surface of AAO templates [12]. The AAO templates performed by a standard four-probe method. The Hall effect in
were fabricated by a two-step anodization process in order magnetic materials is commonly described by the phenomeno-
to obtain highly ordered structures [13,14]. The templates logical equation, ρH = Ro H + Rs Mz , where ρH is the Hall re-
with 50 nm and 100 nm average pore spacing consist of sistivity, Ro is the ordinary Hall coefficient, Rs is the extraordi-
long channels (50 μm) closed at the bottom end by a nary Hall coefficient, and Mz the component of magnetization
round-shaped Al2 O3 barrier layer [Fig. 1(b)]. The barrier- perpendicular to the four contacts plane (here perpendicular
layer surface self-organizes in a hexagonal close-packed to the film plane). As in our experiment the ordinary part is
lattice of nanobumps that possesses the same spacing as much smaller than the extraordinary part, and the extraordinary
the porous layer [Figs. 1(a) and 1(c)]. It is used as a Hall voltage serves as a direct measurement of magnetization
prepatterned substrate to modulate the thickness of thin component perpendicular to film plane. Indeed, the hysteresis
Co/Pt multilayers deposited on it. The multilayers consist loop observed in the ρH (H ) curve was well correlated with
of a Ta(5 nm)/Pt(5 nm)/[Co(0.4 nm)/Pt(0.7 nm)]4 /Pt(3.5 nm) the magnetization curve obtained in the same condition of field
sequence grown by magnetron sputtering. Due to the bumpy and temperature using a commercial SQUID-VSM.
surface, the deposited film experiences a variable thickness Two types of MFM setups have been used for the experi-
along the surface, up to the total thickness of the magnetic ments described here. The first one is a nanoscope AFM/MFM
stack (10 nm) at the top and much thinner close to the that allows measuring topography and the remanent magnetic
bottom of the bumps [15]. In our previous work, it was shown configuration at room temperature. The second is a modified
that the magnetic coating of the bumpy surface maintains an nanoscan high-resolution hr-MFM, operated in a vacuum in
out-of-plane magnetic anisotropy, thus forming an ordered dynamic mode, whereby the cantilever resonance frequency
shifts arising from the tip/sample interaction are recorded and
which allows image magnetic configurations in an applied
magnetic field. We used a Team-Nanotech hr-MFM-ML1 tip,
keeping a tip-sample distance of approximately 8 ± 1 nm on
average. Fields were applied using the calibrated permanent
magnet driven by the perpendicular field option of the
instrument. Prior to the measurements, the tip was separately
magnetized in the direction and sense of the field applied
during the ramp of field. For each applied field, a stable scan
condition was established for the desired tip-sample distance,
and the scan was subsequently carried out. In scan series where
the sequence of applied magnetic fields included a change of
sign, we removed the sample from the field area at zero applied
field, flipped the tip magnetization, repositioned the sample,
and continued the field sequence. Thus we avoided tip magne-
tization flip events during the sample magnetization reversal.
The tip is scanned at constant average height corrected for
large scale topographical variations of the sample but not for
the local topography variations from single bumps. The MFM
contrast then arises from changes of the local magnetic stray
field, i.e., it reflects the local magnetic state of the sample
but also includes contributions of the van der Waals force,
which varies with topography-induced changes of the local tip-
FIG. 1. SEM images of nanobump arrays of a 100 nm period sample distance. The latter becomes apparent [Fig. 7(a)] in the
[(a) top and (b) tilted views, respectively] and a 50 nm period data recorded at positive saturation (+350 mT). The black dots
[(c) top view]. visible on a gray background reflect the center positions of the

174421-2
REVERSAL MECHANISM, SWITCHING FIELD . . . PHYSICAL REVIEW B 89, 174421 (2014)

bumps, i.e., a more attractive van der Waals force arising from
the smaller tip-sample distance. In order to remove as much
as possible of this nonmagnetic contribution to the measured
MFM contrast and to focus on the magnetic signal, we use the
following procedure. First, we align all images, then we crop
them to the area that they all have in common. Subsequently,
we subtract from them the image obtained in saturation, which
contains essentially the same topographical information as all
other images. Finally we center and scale the contrast to ±1
except in the saturated images.

III. RESULTS AND DISCUSSION

A. Macroscopic features
In Fig. 2(a), the magnetization vs field loops measured by
SQUID-VSM at 300 K on the [Co/Pt] stack deposited on flat
Si substrate when the field is applied in-plane or out-of plane
is presented. This multilayer film shows PMA due to the Co/Pt
interfaces [18]. At 300 K, magnetization at saturation is Ms =
760 kA/m, and the effective anisotropy field extracted from
the in-plane field loop equals Hkeff = 470 mT. These values
are in good agreement with previous measurements on similar
samples [18]. In Fig. 2(b), the variation of these two parameters
is shown as a function of temperature. Ms monotonically
increases up to 860 kA/m at 5 K. Hkeff also increases by about
10% at 5 K compared to its room temperature value. At 5 K, FIG. 3. (Color online) (a) EHE measurements for Co/Pt multi-
layers on a bumpy surface for p = 100 nm measured at 20, 100,
Hkeff equals 0.52 T, and the anisotropy constant is 550 kJ/m3 .
200, and 300 K and for p = 50 nm measured at 300 K (dotted line).
The hysteresis loops of a large area specimen made of
The inset shows the variation of the coercive field with temperature;
100 nm period nanobumps measured by EHE at temperatures
the experimental results (black curve) are compared with theoretical
ranging from 300 K down to 20 K are shown in Fig. 3(a). predictions (red curve). (b) Variation of coercivity as a function of
the angle of the applied field measured at 20 K and 300 K for 100 nm
lateral size bumps and at 300 K for 50 nm lateral size bumps with a
Co/Pt multilayer deposited on top. The black continuous line is the
prediction of switching field in the Stoner-Wohlfarth model.

At 300 K, the loop shows a fully saturated magnetization at

remanence but the reversal branches are quite sheared. The
coercive field is 80 mT. For comparison, the [Co/Pt] film
deposited on a flat Si substrate shows a much sharper reversal
and a much lower coercive field of only 12 mT (see Fig. 2).
Such a difference between the bumpy sample and the flat
sample reveals a change of magnetization reversal mechanism,
as already reported in Ref. [12]. In the flat film, nucleation of
one or some reversed domains occurs at low anisotropy spots
and magnetization reversal proceeds through propagation of
the domain walls over the whole sample. In the bumpy sample,
the reversal takes place dot by dot due to a partial exchange
decoupling between the bumps, as explained in Refs. [12]
and [15]. This dot-by-dot reversal will be demonstrated in the
following by the MFM measurements. The evolution of the
coercivity as a function of magnetic field angle with respect
to the surface normal is presented in Fig. 3(b). The coercivity
decreases when the field angle increases from 0° to 45° and
FIG. 2. (Color online) (a) Saturation magnetization vs applied then increases again from 45° up to 75°. Coercivities at 0° and
field measured for a [Co(0.4 nm)/Pt(0.7 nm)] × 4 film at 300 K with 75° have almost the same value, and the value at 45° is about
out-of-plane (black solid squares) and in-plane (red open circles) ¾ of the value at 0°. Hc is considered the switching field of the
field orientation. (b) Saturation magnetization (green solid squares) average dots, and the field angle dependence is to be compared
and anisotropy field (blue open triangles) vs temperature. to the two extreme models that characterize a macrospin type

174421-3
T. HAUET et al. PHYSICAL REVIEW B 89, 174421 (2014)

reversal, namely the Stoner-Wohlfarth model [19], and a nucle-

ation/propagation type reversal, namely the Kondorsky model
[20]. The experimental curve, close to the Stoner-Wohlfarth
model but with a shallower variation, is typical of a dot-by-dot
reversal but with a nucleation/propagation process for each
dot [21,22]. This argument is corroborated by the low value
of Hc (80 mT) as compared to the anisotropy field (470 mT)
extracted from the hard axis loop measured on the flat film with
SQUID-VSM. In addition to regular anisotropy dispersion, our
system shows many reasons for having non-uniform reversal
within each dot. First simulations in Ref. [23] show that the
curvature of the bump can modify the ideal Stoner-Wohlfarth
field angle dependence. Second, the close packing of the
bumps induces a strong dipolar interaction between bumps
with a gradient of dipolar field inside each bump. Third, the
CoPt phase between the bumps must play a strong role through
a remaining interdot exchange coupling, similarly to the ion
milled edges in Ref. [24] or the irradiated region in Ref. [25].
Changing the temperature from 300 K to 20 K does not
drastically affect the hysteresis loop shape or the field angle
dependence of the coercivity [Figs. 3(a) and 3(b)]. As the
temperature decreases from 300 K, the main change concerns
the value of the coercive field, which linearly increases up
to 160 mT at 20 K [inset Fig. 3(a)]. This can be understood
by the continuous reduction of thermal effects acting on the
nucleation process. Since the coercivity’s angular dependences
at 300 K and 20 K are very similar [Fig. 3(a)], no change
in the reversal process is expected in the low temperature FIG. 4. (Color online) (a) Derivative of EHE loops for Co/Pt
range. In the inset of Fig. 3(a), the coercive field changes as a multilayers on a 100 nm period bump surface measured at 20, 100,
function of temperature and is compared with the theoretical 200, and 300 K. The inset shows the double peak Gaussian fitting
variation predicted by the Sharrock equation [26] in the of the curve derived from the measurement made at 300 K for p =
case of the Stoner-Wohlfarth-like reversal of a nucleation 100 nm. (b) Variation of both peaks’ maximum and σ , the standard
volume V of anisotropy K(T ) and magnetization Ms (T ) that deviation, as a function of temperature for the first and second peaks
depend both on temperature T . We used similar parameters shown in (a).
as those in Ref. [24]: the applied field time τ = 10 s, the
attempt frequency f0 = 10 GHz, and a nucleation volume V =
15 × 15 × 4.4 nm3 . 15 nm corresponds to an upper value the hysteresis loop reversal branches and uses a Gaussian fit
for the exchange length. The experimental anisotropy K(T ), to extract the single peak maximum field and the standard
extracted from Fig. 2(b), needs to be divided by three since deviation σ of this single peak [18]. Figure 4(a) shows the
Hc is much lower than Hkeff . Such a factor is typical of real derivatives for the descending branches of loops shown in
samples where the reversal process starts at a place where the Fig. 3(a). For all temperatures, the derivative curve shows
anisotropy is weaker [24,26–29]. Although the fit is clearly not a double peak typical of a two step reversal. As will be
perfect (it cannot be since the reversal is not macrospinlike), confirmed later by the MFM images, the double peak feature
the amplitude of the Hc variation with temperature using the in the derivative reveals the competition between exchange
Sharrock equation is of the right order of magnitude. coupling and strong dipolar interactions. Similar competition
In Figs. 3(a) and 3(b), we also compare the magnetic affects PMA continuous films, like Co/Pt thin films. Tuning the
features of 100 nm and 50 nm period nanobump arrays at dipolar field energy to a value of the same order as the exchange
300 K. Their hysteresis loops are similar and coercive fields coupling energy leads to a double peak derivative whose first
are identical. The reversal process seems unaffected by the peak usually occurs before reaching zero field coming from the
difference in dot diameter at these sizes. The 50 nm period saturation field [30,31]. Note that the double peak derivative
array has just a slightly sharper reversal and slightly larger is observed for continuous films of Co(0.4 nm)/Pt(0.7 nm)
end reversal tail that may mark a slightly stronger exchange multilayer (i.e., the composition we are using here) having
between bumps. Such similar behaviors for the 100 nm around 20 repeats [30], whereas our sample only holds four
and 50 nm period arrays tend to confirm a reversal process repeats. Therefore, in our sample the dipolar interactions
beginning at a nucleation region whose diameter is smaller are more effective, which implies a strong reduction of the
than 50 nm [29]. In the following we will focus only on the interbump exchange coupling in our bump array. Recently, a
100 nm period media that is much easier to image with high shallow double peak shape of the hysteresis loop derivative,
resolution by MFM. centered on the coercive field, has also been reported in a
In order to characterize BPM switching and especially BPM with large interbit dipolar interactions and possibly
SFD of BPM systems, one usually takes the derivative of under-etching, i.e., a nonzero interdot exchange coupling [32].

174421-4
REVERSAL MECHANISM, SWITCHING FIELD . . . PHYSICAL REVIEW B 89, 174421 (2014)

or magnetization, presence, or absence of misaligned grains) or

from the increase of dipolar field induced SFD as Ms increases
when the temperature decreases. The SFD value is high as
compared with the coercivity value. The relative SFD, i.e.,
SFD/Hc , is about 46% at 300 K, which is much larger than
usual values of approximately 10% found in Co/Pt-based BPM
[33,34]. One can wonder if this high relative SFD is mainly the
result of a combination of low Hc (related to the low Hk and
maybe a specific nucleation process at the bump side), high
dipolar fields (related to large Ms), and the close packing of
the BPM array, or if the SFD originates from a large dot-to-dot
distribution of magnetic features.
In order to further distinguish and quantify the SFD origins,
we use the H (M, M) method, which allows for separating
FIG. 5. (Color online) EHE measurement of out-of-plane major the intrinsic SFD from interaction effects. Originally, the
(black curve) and one minor loop (red dashed line) for Co/Pt method had been developed based on the mean-field model
nanobump array. Open red disks show the minor loop (red dashed line) of perpendicular magnetic recording media [16,17] and BPM
upshifted. The large black and white dots are a scheme of magnetic [16,18,35] and was later generalized beyond the mean-field
configuration discussed in the text. Inset: derivative of the open red approximation to be reliable in a broader range of exchange
disks curve (upshifted minor loop) and of black curve (major loop). and dipolar coupling values [36–38].
The arrows point out the difference of “easy switcher” switching field The method requires as an input the measurement of a
distribution as a function of their magnetic environment. hysteresis loop and a set of recoil curves, such as that shown
in Fig. 6(a) for the 100 nm period nanobump array at room
temperature. Here, the recoil curves have been generated at
If the interbump exchange coupling is difficult to extract reversal fields corresponding to fractions of switched bumps
and quantify here, the strong influence of dipolar interactions η = 23%, 50%, and 70%. The initial step before the actual
acting as a long-range antiferromagnetic coupling is easy to analysis requires transforming the data set to the H (M,
demonstrate. In Fig. 5, the reversal fields of the first 50%
of magnetic bits are compared when they have two different
environments [18]. From positive saturation, the magnetization
reversal of the 50% easiest switchers occurs in negative fields.
These switches align the bumps in negative field direction and
antiparallel to the remaining bumps. A comparison with the
switching of the remaining 50% of the bumps can be made
by starting from negative saturation, going to the (positive)
fields that reverse half the bumps, and subsequently reverting
back to negative saturation (dashed line in Fig. 5). Note that
in this reversal process, 50% of the bumps align in negative
field direction and parallel to the direction of the remaining
bumps. A direct comparison between the two reversal branches
shows that the reversal from the half-reversed state requires
field amplitudes larger by 50 mT on average. Moreover the
derivative in the inset of Fig. 5 shows that this reversal is much
shallower than the 50% reversal that starts from the saturated
state, revealing the influence of the dipolar field on the SFD
and possible differences in the reversal mechanism.
Because of its double peak shape, we chose not to fit
the hysteresis loop derivative with one Gaussian but with
two Gaussian curves [see inset Fig. 4(a)]. We call Hpeak the
fields at the first (lower reversal field) and second peak (larger
reversal field) maximum, and we call σ the standard deviation
of the first and second Gaussian curve. These parameters
are displayed in Fig. 4(b) as a function of temperature. The
difference between the two peak Hpeak values remains constant FIG. 6. (Color online) (a) EHE measurement of out-of-plane
with temperature, around 60 mT. The σ is very similar for the major and minor loops for Co/Pt nanobump array with p = 100 nm.
two peaks. We chose to define SFD as the standard deviation The percentage values correspond to the ratio of switched nanobumps
sigma of a function equal to the sum of two Gaussians. At during each loop. (b) H (M, M) data (open circles) obtained from
300 K, SFD is about 37 mT and increases up to 42 mT at the partial reversal curves in (a) and compared with the fit (red solid
20 K. The increase of SFD with temperature can originate line) of the H (M, M) model. The quality of the fit is attested by
both from the intrinsic SFD (dot-to-dot variation of anisotropy the correlation coefficient R 2 = 0.9689.

174421-5
T. HAUET et al. PHYSICAL REVIEW B 89, 174421 (2014)

FIG. 7. (Color online) 1 μm × 1 μm MFM images obtained successively after saturation at +350 mT (a), under 0 mT to −300 mT (b)–(o).
Since the magnetization is saturated at −350 mT, (a) corresponds to an average topography signal that is used to correct all the other MFM
images. Yellow hexagons in (a) mark lattice stacking faults.

M) form by first inverting the plots, i.e., plotting field H MFM figures using the procedure described in Sec. II B. The
vs M/Ms and then subtracting the field values along a recoil yellow spots in Fig. 7(a) indicate AAO lattice defect positions.
curve from the field values along the adjacent (decreasing) Figures 7(b)–7(o) show MFM images of the same nanobump
hysteresis loop branch (note the equivalence M ≈ VEHE due array region taken under applied fields (i.e., not at remanence)
to the normalization, from which follows M= 1−η/100%). from +350 mT to −350 mT. Figure 7(b) shows data obtained
The results of the transformation are the H (M, M) data at 0 mT. Fifteen from a total of 131 bumps visible in the
shown in Fig. 6(b) as open circles. Next, the data are fitted image have switched because of their low intrinsic reversal
by the set of reference functions of the generalized H (M, field combined with the dipolar field from the neighboring
M) method for variable exchange and dipolar interactions bumps. Some images show also that the stray field coming from
assuming a lognormal intrinsic SFD. The best-fit reference the MFM tip influences the bumps magnetization reversal.
function gives the statistically most reliable estimate of the The measurements were carried out with a commercial tip,
intrinsic SFD. As shown in Fig. 6(b), the fit is of very good maintaining a tip-sample distance of approximately 8 ±
quality with R 2 = 0.9689, yielding the standard deviation of 1 nm on average. At this distance, the maximum stray fields
the intrinsic SFD equal to σintrinsic = 14 .1mT. So the H (M, observed for this class of CoCrPt tips is typically of the order
M) method demonstrates that the intrinsic SFD is only of 5 mT, which adds to the external applied field. As the field
about one-third of the overall SFD (37 mT) extracted from amplitude increases from 0 mT towards negative values, more
EHE loop derivative (Fig. 4). The large dipolar effect deduced and more black spots occur until all dots have been reversed
from the comparison between the macroscopic SFD and the [see Figs. 7(c)–7(o)]. From the MFM images, we can calculate
H method sigma is fully consistent with the large shift the normalized magnetization as the number of white bumps
highlighted in Fig. 5. minus the number of black bumps divided by the total number
of bumps. The calculated values are plotted as a function of
applied magnetic field in Fig. 8.
B. Bump-by-bump reversal mechanism imaged by MFM The calculated MFM reversal branch is slightly different
An improved understanding of the roles of the intrinsic from the EHE one. Especially, the two reversal steps are much
SFD and the dipolar interactions for the magnetization reversal more visible in the MFM data, and the separation between
mechanism of Co/Pt nanobump arrays is obtained from the two peaks of the reversal branch derivative is much more
a field-dependent MFM study. Figure 7 shows the results pronounced than in the EHE data. Half reversal is reached
obtained for increasing negative fields after saturation in at about 80 mT, which is close to the coercive field value
positive fields, and Fig. 7(a) shows the positive saturation measured by EHE. Nevertheless, the reversal starts slightly
image (+350 mT). Black dots visible on a gray background before zero field (−20 mT for the EHE measurement) and ends
reflect the center positions of the bumps due to van der at around −300 mT (−170 mT for the EHE measurement).
Waals forces arising from the smaller tip-sample distance. All these differences must originate from a slightly larger
This nonmagnetic contribution has been reduced in the other thickness of Co in the sample measured by MFM. Indeed at

174421-6
REVERSAL MECHANISM, SWITCHING FIELD . . . PHYSICAL REVIEW B 89, 174421 (2014)

a labyrinthine pattern is observed, similar to what can be

observed in Co/Pt continuous films. In that these patterns
arise from a competition between short-range order (exchange
coupling) and long-range order (dipolar coupling), they are
reminiscent of patterns formed in a large range of physical
systems, e.g., Langmuir films, diblock copolymers, chemical
mixtures, and magnetic thin films [39,40–42]. As in PMA
continuous thin films, the method of sweeping the field
from the saturated magnetic configuration, the presence of
thermal fluctuations and the intrinsic SFD prevent the system
from reaching the well-ordered stripes’ domain state [43–45].
Section III C is devoted to a deeper investigation of the stability
of zero-magnetization configurations.
After the system has reached the labyrinthine state at
FIG. 8. (Color online) Normalized moment calculated from the −80 mT [Fig. 7(e)], the magnetic susceptibility strongly
MFM images in Fig. 7 as a function of the applied field. Each solid decreases, and only 14 bumps switch between −100 mT and
circle corresponds to one MFM image. The dashed line corresponds −180 mT (for comparison, 15 bumps switch between 0 and
to a field range where no MFM images have been performed. The −20 mT or between −20 and −40 mT). At this stage the
arrow shows the field sweep direction. The red stars correspond to reversal process consists in magnetization reversal mostly for
the normalized magnetization extracted from MFM images measured the bumps having two or three unswitched neighbors (11
during fields applied after out-of-plane demagnetization of the bumps over 14 to reverse). Here dipolar fields undoubtedly
sample. The inset presents the derivative of the solid circle branch. counteract the interbump exchange coupling; otherwise, the
latter would have prevented the system from creating more
switched/unswitched transitions, and white lines would have
300 K we measured Ms = 840 kA/m (instead of 760 kA/m), shrunk from their ends. We believe that the increase of
and the effective anisotropy field extracted from the in-plane switched/unswitched walls is the main origin of the slowdown
field loop is Hkeff = 380 mT (instead of Hkeff = 470 mT). As a in switching. After the field has been ramped to −200 mT, the
consequence dipolar field interactions are stronger in the MFM magnetic susceptibility increases again. Almost all unswitched
sample, so a stronger shearing of the hysteresis loop branches bumps are now surrounded by switched bumps, and all have
is measured. The discrepancy may also be an artifact caused comparable dipolar and exchange coupling energy. So, the
by the fact that MFM images comprise only about 131 bumps, intrinsic SFD becomes the leading parameter selecting the
whereas EHE senses millions of bumps. switching order.
Let us now analyze in more detail the MFM images to get We earlier used the H (M, M) method to show that
a deeper understanding of the reversal mechanism. First, from about one-third of the total SFD (14.1 mT) is intrinsic. With
all images, we can deduce that there are only two possible the MFM images in Fig. 7, we can now try to get more
magnetic states for each bump, i.e., fully magnetized up or insight on the impact of the intrinsic SFD on the magnetic
fully magnetized down. A nonuniform magnetization state is configurations during field cycling. Easy and hard switchers
not observed. We can also conclude that the magnetization (i.e., bumps having the lowest and largest switching field,
reverses bump by bump. Nevertheless, both exchange coupling respectively) can be detected by the observation of successive
and dipolar fields affect the location of a bump reversal. From images during field sweeps. Three magnetic field cycles have
0 mT [Fig. 7(b)] to −20 mT [Fig. 7(c)], 15 additional bumps been used to define the easy switchers as the bumps that have
have switched over 116 unswitched (white) bumps. If we do at least a 66% chance to be already switched at zero field
not consider the bumps at the image edges (because we do starting from a saturated state (8 over about 150 bumps in the
not know the magnetic states of all their neighbours), we have whole image). We defined the hard switchers as the bumps
to consider only 85 unswitched bumps at 0 mT. Of these 85 that have at least a 66% chance of remaining unswitched at
unswitched bumps, 58 have at least one switched neighbor, and −260 mT (9 over about 150 bumps in the whole image). Easy
27 have none. Nevertheless, of the 12 reversals (3 reversals and hard switchers’ positions are highlighted in the MFM
occur at the image edge that we disregard), 8 occur next to image in Fig. 9(a), which has been taken around half reversal
a switched neighbor, and 4 happen in unswitched regions. (60 mT) after negative field saturation. At this stage of the
The switching event location seems to be uncorrelated to the reversal process, they belong to a labyrinthine domain pattern
presence or the absence of switched neighbors. This result composed largely of black and white straight lines. In this
indicates that although the long-range dipolar fields affect pattern, all the easy switchers are black, and all hard switchers
the bump reversal (Fig. 5), locally the interbump exchange are white. On the same image [Fig. 9(a)], we have also placed
coupling and the dipolar interaction more or less compensate the position of the self-organized AAO network stacking faults
each other. Furthermore, we can also conclude that the dipolar [as well as in Fig. 7(a)]. Easy and hard switchers’ locations are
field opposes the exchange coupling since the subsequent clearly uncorrelated with the stacking defects. The dispersion
switching events do not lead to an isotropic growth of a in bump diameter is in our back AAO template of the order
reversed domain but tend to form lines of switched bumps of ±9 nm (cf. AFM processing in Sec. III D). We were not
having a width of one bump [Figs. 7(c) and 7(d) ]. At able to correlate easy or hard switchers to a larger or smaller
−80 mT [Fig. 7(e)], where half of the bumps have switched, bump diameter nor to specific bump shape. Since the reversal

174421-7
T. HAUET et al. PHYSICAL REVIEW B 89, 174421 (2014)

The main difference between the nanobump array and the

usual continuous film lies in the triangular lattice of the AAO
template. The loss of isotropy has an important impact since the
triangular lattice generates a dipolar field frustration between
neighboring bumps’ magnetization. Frustration of magnetic
spins interacting through dipolar fields in triangular spin lattice
has been heavily studied since the 1950s [47,48] and remains
a subject of current investigation due to the development of
micro- and nanopatterning techniques [1,49,50]. Kireev et al.
[51] have recently reported a work that is particularly well
suited to help understanding the consequences of dipolar
FIG. 9. (Color online) 1 μm × 1 μm MFM images obtained close frustrations in our bump assembly. They demonstrate that
to half reversal (−60 mT) after negative field saturation (a) and at the ground state of an assembly of hexagonal array of per-
remanence after demagnetization (b). The solid red disks correspond pendicularly magnetized Ising spins interacting only through
to hard switchers, open blue disks to easy switchers, and yellow the antiferromagnetic dipolar interactions consists in a series
hexagons to topological defects. of uniformly magnetized stripes of spins, each stripe being
oppositely magnetized to its neighboring stripes. This conclu-
sion holds for pure dipolar interactions. If a ferromagnetic
of a bump does not occur by coherent reversal, we do not coupling between closest neighbors has to be considered,
expect the bump diameter to affect the field of nucleation. the fundamental state is modified and should evolve towards
As a consequence, we expect the intrinsic SFD to originate wider domains. To quantify this effect, the lowest energy
mostly from the presence, absence, number, and location of configuration has been determined for a small network of
Co/Pt [100] misoriented grains on the bumps that differ from 25 elements for various coupling coefficients between first
one bump to another [27–29]. The location of easy and hard neighbors. With the unit of coupling coefficients chosen to
switchers clearly constrains the shape and location of the black be the dipolar coupling between first neighbors, J1 = −1
and white domains in the labyrinth pattern of Fig. 9(a). During corresponds to the pure dipolar coupling between bumps and
a field cycle, the magnetic configurations at 50% reversal in leads to single bump wide stripes, as shown in the scheme in
the ascending and descending branches are not identical. Less Fig. 10(a). This structure is unchanged until the first-neighbors
than 10% of bumps change from one MFM image to another coupling reaches −0.2, where magnetic bands are no longer
one. So, the switching process is not fully deterministic, and one dot wide [Fig. 10(b)]. Indeed at that point, in a real system,
thermal fluctuations slightly affect the 50% reversal magnetic the ferromagnetic coupling between first neighbors would
configurations. almost compensate the first-neighbors dipolar interaction.
As the relative ferromagnetic coupling increases, passing
through exact compensation at J1 = 0 and until J1 = +0.5,
C. Influence of dipolar frustrations, exchange-coupling, and the domains get wider and wider [see Figs. 10(d) and 10(e)].
intrinsic SFD on zero-magnetization states At J1 = +1, the fundamental state is a fully saturated state.
In Fig. 9, we compare the MFM images measured under We can conclude from our calculation that the presence of
a perpendicular field of −60 mT after positive saturation a “small” direct ferromagnetic exchange coupling between
at +350 mT [Fig. 9(a)] and one measured at remanence dots does not affect the nature of the fundamental state, so that
after ac demagnetization under decreasing out-of-plane field single bump wide stripes are stabilized by the triangular lattice
[Fig. 9(b)]. Unlike the labyrinth pattern discussed above, out- of the bump assembly, which does not exist in a regular full
of-plane ac demagnetization leads to a mostly well-ordered film deposited on flat substrate.
succession of up and down magnetized stripes. Each stripe
consists of a 1D chain of bumps. The width of both serpentine
domains and stripe domains, mostly of the order of the array
period, corroborates our previous comparison between the
nanobump assembly and the [Co(0.4)/Pt(0.7)]20 multilayer
thin film (see Sec. III A) and the strong decrease of lateral
ferromagnetic exchange-coupling in our AAO system. Indeed
100 nm domain periods would require 20 to 40 repeats in a
strongly coupled thin film [30]. Nevertheless, it is interesting
to note that a well-ordered stripe configuration in strongly
coupled PMA thin films is only observed after in-plane field ac
demagnetization [30,46]. Here all demagnetization processes,
with a careful orientation of the external field to as close as
possible to the normal to the film have led to a stripe state.
This result proves that our bump assembly is not directly FIG. 10. Ground states for characteristic values of first-neighbors
comparable to regular Co/Pt thin films. Another parameter magnetic coupling J1 with the unit of coupling coefficient being the
has to be considered in the energy calculation to explain the dipolar coupling between first neighbors; e.g., (a) holds only dipolar
stripe state stability of Fig. 9(b). couplings.

174421-8
REVERSAL MECHANISM, SWITCHING FIELD . . . PHYSICAL REVIEW B 89, 174421 (2014)

reported in Fig. 8. As the field increases, the stripe state region

is almost unchanged up to −180 mT. Only 4 of 35 unswitched
(white) bumps have switched. Only one of these 4 bumps
has previously been identified as easy switcher. At this stage,
under −180 mT, this region holds 31 unswitched bumps. For
a fair comparison, the same region in Fig. 7(j) holds only
21 unswitched bumps. This difference is highlighted in the
calculated magnetization vs field curves in Fig. 8, where the
two curves are quite different. We can conclude that a larger
field amplitude is required to destroy the zero-magnetization
stripe state than the zero-magnetization labyrinth state. It is
consistent with the fact that the stripe state is the energy
ground state. As expected on the same field range from 0
to −180 mT, some bumps have switched outside the stripe
region. They are either easy switchers or, more interestingly,
bumps connecting the shifted stripes. In Fig. 11(b), the new
boundary between the shifted stripes is marked with a green
zigzag. Again, as described in Ref. [51], this zigzag composed
of switched bumps is stabilized by the external field, which
is competing with the dipolar antiferromagnetic interactions.
Further increase of the field leads to successive switching
FIG. 11. (Color online) 1 μm × 1 μm MFM images obtained
events of individual bumps inside the unswitched stripes until
at remanence after out-of-plane demagnetization (a) and then under
−180 mT (b), −210 mT (c), and −250 mT (d), respectively. The
only mostly hard switchers remain. This stage of the process
red solid disks correspond to hard switchers, blue open disks to easy is essentially similar to the one reported in Figs. 7(j)–7(o).
switchers; the green lines highlight specific switched regions, and the
yellow hexagons correspond to stacking defects, as discussed in the D. Influence of arrays ordering faults on the
text. The white dashed lines mark the region taken into account to
zero-magnetization configurations
calculate the relative magnetization in Fig. 8.
We have demonstrated above that the intrinsic SFD is
uncorrelated with the stacking defects of the nanobump array.
The well-ordered stripe configuration does not cover the However, in the self-assembly of BPM, bit location and
whole MFM image of Fig. 9(b). Stripes have finite length, as ordering faults are known to be important for achieving
later highlighted by the two dashed lines in Fig. 11(a). Higher tight magnetic SFD and reliable control of the magnetic
energy configurations are observed. In the disordered region, configuration [8,52]. Moreover, the number of spins, the finite
as in the ordered one, intrinsic SFD does not seem to constrain size, and possible distortion of the well-ordered triangular
the final stripe domains state since up and down magnetized array of Ising spins is found to affect the energy (and
easy and hard switchers can be found in Fig. 9(b). On the thus the stability) of the various frustrated magnetic states
contrary, it seems reasonable to think that part of the magnetic [51,53]. In Fig. 9(b), we have already noticed that the lower
disorder originates from the large density of bump packing zone of the MFM image, which contains a large number
faults (i.e., change of dipolar environment) present in the lower of stacking defects (yellow spots), has a much less ordered
part of the MFM image. The short- and long-range influence magnetic configuration, i.e., no well-ordered stripe domains,
of these structural defects will be described in Sec. III D. In as compared to the upper part of the image where there is no
addition, white zigzag stripes spread perpendicularly to the lattice defect.
stripes. This topological magnetic defect typically originates To gain more insight into the role of stacking defects in the
from a succession of Y-shaped domains that connect stripes array on the magnetic ground state, we demagnetized the first
laterally shifted from one bump. This can be clearly observed sample (used for EHE measurement) with perpendicular fields
around the bottom white dashed line in Fig. 9(b). Such a and imaged with AFM and MFM a region of the sample with a
zigzag domain can be modeled by accounting for dipolarly large density of ordering defects. MFM and AFM at the same
coupled Ising spins [51], but it is of course stabilized if direct sample place (2 μm × 2 μm) are shown in Figs. 12(a) and
ferromagnetic exchange-coupling exists between neighboring 12(b), respectively. The resulting MFM pattern [Fig. 12(a)]
bumps. is superimposed on the AFM image in Fig. 12(b) for further
Let us now try to quantify the field stability of the stripe processing of the images. Let us use the AFM image to locate
domains’ configuration as compared to the labyrinth one. In the magnetic bump centers and to characterize the lattice
Fig. 11, we show MFM images measured when an external ordering (Fig. 13). First a Delaunay triangulation is applied to
magnetic field is applied onto the ac demagnetized remanent the set of bump centers so that the whole surface is covered by
state. Figure 11(a) corresponds to the same image of the triangles. After the triangulation, one can define the structural
ac demagnetized state as Fig. 9(b). The two dashed lines bump-ordering quality, which is related to the triangle quality
in Figs. 11(a), 11(c), and 11(d) delimitate the zone of the by imposing criteria on the triangle edges in such a way as
well-ordered stripes state (70 bumps total) that has been taken to keep only triangles that are equilateral within a defined
into account for calculating the magnetization vs the field tolerance. The criteria are defined as follows [see Fig. 13(b) for

174421-9
T. HAUET et al. PHYSICAL REVIEW B 89, 174421 (2014)

FIG. 12. (Color online) (a) MFM image of a 2 μm × 2 μm array

of bumps sample surface. (b) Corresponding 2 μm × 2 μm AFM
image with the MFM pattern superimposed on it.

notation]:
Cr1 = max(abs([d1 ,d2 ,d3 ]))/mean([a,b,c])
where
d1 = a − b, d2 = a − c, d3 = b − c,
Cr2 = 1 − min([a,b,c])/ max([a,b,c]),
Cr3 = max (abs(π/3[α,β,γ ]))/(π/3).
An additional criterion over the surface is used to avoid the
appearance of aberrant triangles, meaning triangles satisfying
the three criteria above but having a surface much larger than
the mean triangle surface. Triangles are accepted when their
surface differs from the mean surface by less than two times the FIG. 14. (Color online) Postprocessing results after triangles
standard deviation, i.e., S  [S − 2σ S,S +2σ S]. Introducing base angle (α) analysis giving rise to three main base angle
the tolerance τ the final selection conditions become orientations according to the number of occurrences distribution of
the α angle. The three different base angle orientations correspond to
(Cr1 < τ ) ∩ (Cr2 < τ ) ∩ (Cr3 < τ ) ∩ (S ∈ [S − 2σS ,S + 2σS ]). different magnetic domains delimited by the structural defects (black
Applying a tolerance of 30% (τ = 0.3) in the above filled circles). Domains with orange triangles correspond to α ranging
equation, one obtains 71.6% (587/820) of triangles satisfying from −30° to −17.5°, from 25° to 30°, the ones with blue triangles
correspond to α ranging from −17.5° to 6.5°, and the ones with green
the criteria [see Fig. 13(a)]. Computing the statistics over the
triangles correspond to α ranging from 6.5° to 25°. (b) Number of
occurrences of a given base angle α ranging from −30° to 30° (α =
60°). In the inset, triangle base angle α definition showing the six
equivalent triangles (due to hexagonal symmetry) for a given α base
orientation relative to the x axis. The angle range for the triangle base
is then limited to 60° (360°/6) due to symmetry.

triangles’ edges one obtains a mean interbump distance of

103.8 nm with a standard deviation of 9.0 nm. As shown in
Fig. 13(a) by black filled circles, the structural defaults are
almost outside the surface covered by the equilateral triangles
satisfying the criteria detailed above (at level τ = 0.3). The
quality of the equilateral triangles informs on the structural
ordering of the array of bumps and allows pointing out the
lattice defects. Although some hole locations seem stochastic
over the whole sample, Fig. 13(a) clearly demonstrates that
FIG. 13. (Color online) (a) AFM image with additional black lines of holes form boundaries between well-ordered areas.
filled circles showing structural defaults. The triangles obtained by To go further in characterizing the features of the different
Delaunay triangulation applied to the bump center coordinates are ordered areas, a triangle base angle analysis is performed. Due
filtered using criteria from Eq. (1) at τ = 0, 3 level and superimposed to hexagonal symmetry in the bumps’ ordering, there are six
on the AFM image. (b) Sketch of an equilateral triangle showing the nominally equivalent triangles corresponding to a given α base
angles and edges notations. angle relative to the x axis, as shown in Fig. 13(b). Figure 14(b)

174421-10
REVERSAL MECHANISM, SWITCHING FIELD . . . PHYSICAL REVIEW B 89, 174421 (2014)

shows the number of occurrences of a given α base angle in fields. One has to consider an additional interbump exchange
the range −30° to 30° (α = 60°). From the latter result coupling [31]. Nevertheless, such a double peak structure
three main base angle orientations are distinguished using requires 20 repeats of the same Co(0.4 nm)/Pt(0.7 nm) bilayer
three colors: orange for α  [−30° −17.5°]  [25°, 30°], as a thin film [30], whereas our bump array is only covered
blue for α  [−17.5° 6.5°], and green for α  [6.5° 25°]. with four repeats of Co(0.4 nm)/Pt(0.7 nm). Therefore, we
Then, using the same color code to fill in the corresponding can conclude that the exchange coupling is strongly reduced
triangles, one obtains a few colored regions on the AFM image in our bumpy system as compared to a continuous film.
[Fig. 14(a)]. At this point we can distinguish different “hyper Intermixing between the bumps is the most probable origin
grains” (ordered regions of different colors) defined by their of this partial exchange decoupling, as explained in Ref. [15].
respective orientation (α). The morphology of our surface High-resolution MFM allowed imaging the magnetization
thus mimics the surface of (111) textured body-centered-cubic reversal over hundreds of bumps during sweeps of an external
(bcc) polycrystals. The MFM demagnetized state pattern is magnetic field of uniform value over the whole imaged
then superimposed on the colorized AFM image on Fig. 14(a). area. Bump-by-bump individual magnetization switching is
From the MFM-AFM image comparison, it is clear that the demonstrated. After the first switching events, competition
magnetic lines are oriented along the α angle (modulo 2π /3) between exchange coupling and dipolar interaction is found
characterizing the hyper grain which they belong to. Hyper to favor lines of switched bumps until 50% of the bumps have
grain boundaries formed by a line of unique defects seem not been switched. At this stage, a labyrinthine domain pattern is
to affect the magnetic stripe too much. When passing through observed that has a low field susceptibility, i.e., it is more stable
a single defect boundary, the magnetic stripe just deviates to than the previous and subsequent configuration. This loss of
a new direction. The interbump exchange coupling may help magnetic susceptibility at 50% reversal leads to the double
to stabilize the boundary crossing. Zones with packed defects peak feature of the macroscopic hysteresis loops. Using an out-
affect the magnetic configuration more due to a strong change of-plane demagnetization, we achieved a zero-magnetization
in the exchange coupling and dipolar field environment, as configuration consisting of successive black and white stripes
observed in both Figs. 9(b) and 14(a). The remaining question with only one bump width. This configuration is even more
is whether the magnetic Y-shaped domains and zigzag domain stable than the labyrinthine one. When considering each bump
shape located inside the hyper grains are due to the finite size magnetization as an Ising spin, the dipolar interactions in a
of the hyper domain [51], due to lattice stress [53], or due triangular lattice are sufficient to explain the stability of the
to another feature. More statistics on hyper domain size and stripe state. Nevertheless, this configuration stability is not
shape are required to conclude on this point. drastically affected by the nonvanishing interbump exchange
coupling as long as it remains less than a fourth of the dipolar
interactions amplitude.
Magnetometry and MFM measurements have been used to
IV. SUMMARY AND CONCLUSIONS
quantify the SFD, to locate easy and hard switchers, and to
In our recent paper [12], we presented a method to grow distinguish the different origins of SFD and their influence
a perpendicularly magnetized BPM system, i.e., an ordered on the magnetic configurations. The H (M, M) method
assembly of magnetic islands. It was achieved by depositing shows that the intrinsic SFD is only one-third of the total
a Co/Pt multilayer on top of the barrier layer of AAO SFD. We could not find any structural reason to explain
templates with a 100 nm period to form an ordered array easy and hard switchers’ locations, so we conclude that the
of ferromagnetic nanodots, so-called nanobumps. Our self- intrinsic SFD mostly comes from misorientated Co/Pt grains
assembly method would have a number of advantages over the inside the bumps [27,28]. The strong dipolar-induced SFD
lithographic-based methods, including low cost and extremely is due to the close packing of the AAO template and might
simple processing. The AAO nanotemplates would display be a showstopper for a BPM technology implementation. In
tunable geometrical parameters, mild preparation conditions, addition, AAO lattice stacking faults are found to form mostly
robustness, and resistance to high temperatures. In addition, boundaries between well-ordered bump arrays. Lines of single
the extremely small attainable interbump period constitutes defects seem not to drastically affect the magnetic features
substantial advantages. Indeed regular nanopore arrays in of the neighboring bumps, whereas larger defect density
AAO with ultrasmall pore diameter (even less than 10 nm) does.
and pore interdistance as small as 15 nm have already been Finally, magnetic features of our nanobump assembly are
demonstrated [54,55]. very similar to those observed in continuous PMA thin films.
In the present paper, we have studied in detail two magnetic Therefore, besides being interesting as support for a BPM
features of a [Co(0.4 nm)/Pt(0.7 nm)]4 multilayer deposited template, we think that the present AAO system could be
on top of 100 nm and 50 nm AAO nanotemplates. EHE and a model for getting a deeper understanding of the influence
SQUID-VSM reveal values of coercivity five times lower than of the different thin film parameters (exchange coupling,
the anisotropy field. The latter’s dependence on field angle and dipolar interaction, and, more importantly, intrinsic SFD or
as a function of temperature reveal an incoherent magnetiza- thermal fluctuations) on the reversal mechanisms in PMA films
tion reversal mechanism, most probably based on a nucleation- [30,38,43,44]. Two main advantages of the bump assembly
propagation process inside each bump, as usually observed in system would be the control of interbump exchange coupling
nano-objects of this size. Surprisingly, the derivative of the and the precise characterization of SFD, which is difficult in
hysteresis loops shows a double peak structure that cannot usually strongly coupled systems [43,56]. More generally, the
be explained by taking into account only interbump dipolar present nanobump array is also a rich parameter space for

174421-11
T. HAUET et al. PHYSICAL REVIEW B 89, 174421 (2014)

testing in more detail isotropic and anisotropic triangular Ising ACKNOWLEDGMENTS

models with short- and long-range interactions [49,57] that
The authors thank S. Petit for help with numerical sim-
rule a wide range of physical systems such as Langmuir films,
ulations. F.A.A. acknowledges the Research Science Foun-
diblock copolymers, chemical mixtures, and not only magnetic
dation of Belgium (FRS-FNRS) for financial support (FRIA
thin films [39–41].
grant).

[1] R. F. Wang, C. Nisoli, R. S. Freitas, J. Li, W. McConville, B. J. [20] E. Kondorsky, J. Phys. Moscow 2, 161 (1940).
Cooley, M. S. Lund, N. Samarth, C. Leighton, V. H. Crespi, and [21] R. Dittrich, G. Hu, T. Schrefl, T. Thomson, D. Suess, B. D.
P. Schiffer, Nature (London) 439, 303 (2006). Terris, and J. Fidler, J. Appl. Phys. 97, 10J705 (2005).
[2] S. Zhang, J. Li, I. Gilbert, J. Bartell, M. J. Erickson, Y. Pan, [22] J. W. Lau, X. Liu, R. C. Boling, and J. M. Shaw, Phys. Rev. B
P. E. Lammert, C. Nisoli, K. K. Kohli, R. Misra, V. H. Crespi, 84, 214427 (2011), and references therein.
N. Samarth, C. Leighton, and P. Schiffer, Phys. Rev. Lett. 109, [23] T. C. Ulbrich, D. Makarov, G. Hu, I. L. Guhr, D. Suess,
087201 (2012). T. Schrefl, and M. Albrecht, Phys. Rev. Lett. 96, 077202 (2006).
[3] T. R. Albrecht, D. Bedau, E. Dobisz, Gao He, M. Grobis, [24] J. M. Shaw, S. E. Russek, T. Thomson, M. J. Donahue, B. D.
O. Hellwig, D. Kercher, J. Lille, E. Marinero, K. Patel, R. Ruiz, Terris, O. Hellwig, E. Dobisz, and M. L. Schneider, Phys. Rev.
M. E. Schabes, Wan Lei, D. Weller, and Wu Tsai-Wei, IEEE B 78, 024414 (2008).
Trans. Magn. 49, 773 (2013). [25] T. Aign, P. Meyer, S. Lemerle, J. P. Jamet, J. Ferré, V. Mathet,
[4] O. Ozatay, P. G. Mather, J. U. Thiele, T. Hauet, and P. M. C. Chappert, J. Gierak, C. Vieu, F. Rousseaux, H. Launois, and
Braganca, Spin-Based Data Storage, in Nanofabrication and H. Bernas, Phys. Rev. Lett. 81, 5656 (1998).
Devices, Vol. 4, of Comprehensive Nanoscience and Nanotech- [26] M. P. Sharrock and J. T. McKinney, IEEE Trans. Magn 17, 3020
nology, D. Andrews, G. Scholes, and G. Wiederrecht, eds. (1981).
(Elsevier, London, 561, 2010). [27] J. W. Lau, R. D. McMichael, S. H. Chung, J. O. Rantschler,
[5] S. Y. Chou, Proc. IEEE 85, 652 (1997). V. Parekh, and D. Litvinov, Appl. Phys. Lett. 92, 012506
[6] F. Luo, L. J. Heyderman, H. H. Solak, T. Thomson, and M. E. (2008).
Best, Appl. Phys. Lett. 92, 102505 (2008). [28] B. Pfau, C. M. Gunther, E. Guehrs, T. Hauet, H. Yang, L. Vinh,
[7] C. A. Ross, H. I. Smith, T. A. Savas, M. Schattenberg, X. Xu,, D. Yaney, R. Rick, S. Eisebitt, and O. Hellwig, Appl.
M. Farhoud, M. Hwang, M. Walsh, M. C. Abraham, and Phys. Lett. 99, 062502 (2011).
R. J. Ram, J. Vac. Sci. Technol. B 17, 3168 (1999). [29] J. M. Shaw, M. Olsen, J. W. Lau, M. L. Schneider, T. J. Silva,
[8] O. Hellwig, J. K. Bosworth, E. Dobisz, D. Kercher, T. Hauet, O. Hellwig, E. Dobisz, and B. D. Terris, Phys. Rev. B 82, 144437
G. Zeltzer, J. D. Risner-Jamtgaard, D. Yaney, and R. Ruiz, Appl. (2010).
Phys. Lett. 96, 052511 (2010). [30] O. Hellwig, A. Berger, J. B. Kortright, and E. E. Fullerton,
[9] M. Albrecht, G. Hu, I. L. Guhr, T. C. Ulbrich, J. Boneberg, J. Magn. Magn. Mater. 319, 13 (2007), and references therein.
P. Leiderer, and G. Schatz, Nature Mater. 4, 203 (2005). [31] A. Berger and H. Hopster, J. Appl. Phys. 79, 5619 (1996).
[10] L. Yan, K. Wang, J. Wu, and L. Ye, Colloids and Surfaces A: [32] N. Eibagi, J. J. Kan, F. E. Spada, and E. E. Fullerton, IEEE
Physicochem. Eng. Aspects 296, 123 (2007). Magnetic Lett. 3, 4500204 (2012).
[11] P. Kappenberger, F. Luo, L. J. Heyderman, H. H. Solak, [33] O. Hellwig, T. Hauet, T. Thomson, E. Dobisz, J. D. Risner-
C. Padeste, C. Brombacher, D. Makarov, T. V. Ashworth, Jamtgaard, D. Yaney, B. D. Terris, and E. E. Fullerton, Appl.
L. Philippe, H. J. Hug, and M. Albrecht, Appl. Phys. Lett. 95, Phys. Lett. 95, 232505 (2009).
023116 (2009). [34] T. Hauet, E. Dobisz, S. Florez, J. Park, B. Lengsfield, B. D.
[12] L. Piraux, V. A. Antohe, F. Abreu Araujo, S. K. Srivastava, Terris, and O. Hellwig, Appl. Phys. Lett. 95, 262504 (2009).
M. Hehn, D. Lacour, S. Mangin, and T. Hauet, Appl. Phys. Lett. [35] I. Tudosa, M. V. Lubarda, K. T. Chan, M. A. Escobar,
101, 013110 (2012). V. Lomakin, and E. E. Fullerton, Appl. Phys. Lett. 100, 102401
[13] H. Masuda and K. Fukuda, Science 268, 1466 (1995). (2012).
[14] O. Rabin, P. R. Hertz, Y.-M. Lin, A. I. Akinwande, S. B. Cronin, [36] O. Hovorka, Y. Liu, K. A. Dahmen, and A. Berger, Appl. Phys.
and M. S. Dresselhaus, Adv. Funct. Mater. 13, 631 (2003). Lett. 95, 192504 (2009); O. Hovorka, R. F. L. Evans, R. W.
[15] T. C. Ulbrich, C. Bran, D. Makarov, O. Hellwig, J. D. Risner- Chantrell, and A. Berger, ibid. 97, 062504 (2010); O. Hovorka,
Jamtgaard, D. Yaney, H. Rohrmann, V. Neu, and M. Albrecht, J. Pressesky, G. Ju, A. Berger, and R. W. Chantrell, ibid. 101,
Phys. Rev. B 81, 054421 (2010). 182405 (2012).
[16] A. Berger, Y. H. Xu, B. Lengsfield, Y. Ikeda, and E. E. Fullerton, [37] O. Hovorka, R. F. L. Evans, R. W. Chantrell, Y. Liu, K. A.
IEEE Trans. Magn. 41, 3178 (2005). Dahmen, and A. Berger, J. Appl. Phys. 108, 123901 (2010).
[17] A. Berger, B. Lengsfield, and Y. Ikeda, J. Appl. Phys. 99, 08E705 [38] O. Hovorka, Y. Liu, K. A. Dahmen, and A. Berger, J. Magn.
(2006). Magn. Mater. 322, 459 (2010).
[18] O. Hellwig, A. Berger, T. Thomson, E. Dobisz, Z. Z. Bandic, [39] A. D. Stoycheva and S. J. Singer, Phys. Rev. Lett. 84, 4657
H. Yang, D. S. Kercher, and E. E. Fullerton, Appl. Phys. Lett. (2000).
90, 162516 (2007). [40] M. Seul and D. Andelman, Science 267, 476 (1995), and
[19] E. C. Stoner and E. P. Wohlfarth, Philos. Trans. R. Soc. London references therein.
A 240, 599 (1948). [41] G. Malescio and G. Pellicane, Nature 2, 97 (2003).

174421-12
REVERSAL MECHANISM, SWITCHING FIELD . . . PHYSICAL REVIEW B 89, 174421 (2014)

[42] A. Hubert and R. Schäfer, Magnetic Domains, Chapters 3 and 5 [50] N. Rougemaille, F. Montaigne, B. Canals, A. Duluard,
(Springer-Verlag, Berlin, 1998). D. Lacour, M. Hehn, R. Belkhou, O. Fruchart, S. El Moussaoui,
[43] M. S. Pierce, J. E. Davies, J. J. Turner, K. Chesnel, E. E. A. Bendounan, and F. Maccherozzi, Phys. Rev. Lett. 106,
Fullerton, J. Nam, R. Hailstone, S. D. Kevan, J. B. Kortright, 057209 (2011).
K. Liu, L. B. Sorensen, B. R. York, and O. Hellwig, Phys. Rev. [51] V. E. Kireev, R. S. Khymyn, B. A. Ivanov, and C. E. Zaspel,
B 87, 184428 (2013). arXiv:1201.1747 (2012).
[44] M. S. Pierce, C. R. Buechler, L.B. Sorensen, S. D. Kevan, E. A. [52] K. Nielsch, R. B. Wehrspohn, J. Barthel, J. Kirschner,
Jagla, J. M. Deutsch, T. Mai, O. Narayan, J. E. Davies, K. Liu, S. F. Fischer, H. Kronmuller, T. Schweinbock, D. Weiss,
G. T. Zimanyi, H. G. Katzberger, O. Hellwig, E. E. Fullerton, and U. Gosele, J. Magn. Magn. Mater. 249, 234
P. Fischer, and J. B. Kortright, Phys. Rev. B 75, 144406 (2007). (2002).
[45] O. Hellwig, G. P. Denbeaux, J. B. Kortright, and E. E. Fullerton, [53] Y. Shokef and T. C. Lubensky, Phys. Rev. Lett. 102, 048303
Physica B: Phys. Cond. Matter 336, 136 (2003). (2009).
[46] M. A. Marioni, N. Pilet, T. V. Ashworth, R. C. O’Handley, and [54] Y. Matsui, K. Nishio, and H. Masuda, Small 2, 522
H. J. Hug, Phys. Rev. Lett. 97, 027201 (2006). (2006).
[47] G. H. Wannier, Phys. Rev. 79, 357 (1950); ,Phys. Rev. B 7, 5017 [55] G. Q. Ding, R. Yang, J. N. Ding, N. Y. Yuan, and Y. Y. Zhu,
(1973). Nanoscale Res. Lett. 5, 1257 (2010).
[48] R. M. F. Houtappel, Physica (Amsterdam) 16, 391 (1950); ,16, [56] O. Ozatay, T. Hauet, S. H. Florez, J. A. Katine, A. Moser, J.-U.
425 (1950). Thiele, L. Folks, and B. D. Terris, Appl. Phys. Lett. 95, 172502
[49] E. Mengotti, L. J. Heyderman, A. Bisig, A. Fraile Rodrı́guez, (2009).
L. Le Guyader, F. Nolting, and H. B. Braun, J. Appl. Phys. 105, [57] Y. I. Dublenych, J. Phys.: Condens. Matter 25, 406003
113113 (2009). (2013).

174421-13