This is a preprint version of the paper published in

This is a preprint version of the paper published in
Journal of applied physics. The full-text final article may be downloaded from
J. Appl. Phys. 115, 243901 (2014) © 2014 AIP Publishing LLC
Investigations of stacking fault density in perpendicular
recording media
S.N. Piramanayagam, Binni Varghese, Yi Yang, Wee Kiat Lee and Hang Khume Tan
Data Storage Institute, A*STAR (Agency for Science, Technology and Research), 5,
Engineering Drive 1, Singapore – 117608.
In magnetic recording media, the grains or clusters reverse their magnetization over a range
of reversal field, resulting in a switching field distribution. In order to achieve high areal
densities, it is desirable to understand and minimize such a distribution. Clusters of grains
which contain a stacking fault have lower anisotropy, an order lower than those without
stacking faults. It is believed that such low anisotropy regions reverse their magnetization at a
much lower reversal field than the rest of the material with a larger anisotropy. Such
clusters/grains cause recording performance deterioration such as adjacent track erasure and
dc noise. Therefore, the observation of clusters that reverse at very low reversal fields
(nucleation sites) could give information on the noise and the adjacent track erasure.
Potentially, the observed clusters could also provide information on the stacking faults (SF).
In this paper, we study the reversal of nucleation sites in granular perpendicular media based
on a magnetic force microscope (MFM) methodology and validate the observations with high
resolution cross-section transmission electron microscopy (HRTEM) measurements.
Samples, wherein a high anisotropy CoPt layer was introduced to control the nucleation sites
(NS) or stacking faults (SF) in a systematic way, were evaluated by MFM, TEM and
magnetometry. The magnetic properties indicated that the thickness of the CoPt layer results
in an increase of nucleation sites. TEM measurements indicated a correlation between the
thickness of CoPt layer and the stacking fault density. A clear correlation was also observed
between the MFM results, TEM observations and the coercivity and nucleation field of the
samples, validating the effectiveness of the proposed method in evaluating the nucleation
sites which potentially arise from stacking faults.
Corresponding author: [email protected]
Granular perpendicular recording media based on stacked layers of CoCrPt:Oxide are
currently used in hard disk drives.1,2 The layers are designed in such a way that the bottom
most layer has the highest anisotropy constant, which helps to achieve the desired thermal
stability for the whole stack of recording layer grains.3,4 The layers on top have gradually
reduced values of anisotropy constant. The magnetic layers with different anisotropy are
separated by some non-magnetic layers to fine tune the exchange coupling between them.
Such a design helps in switching of the whole stack through an incoherent reversal process,
which enables a writing process at a lower field than that is required in the case of coherent
reversal mechanism.6,7 The bottom layer with the highest anisotropy constant has a large
amount of Pt and plays a key role in the thermal stability.8 Although thermal stability can be
increased by increasing the Pt concentration to certain extent, it is commonly known that the
addition of Pt in excess of 20 at% causes stacking faults.9,10 Since the variation in Pt
concentration from grain to grain in the deposited film cannot be controlled precisely, the
stacking fault density will increase if Pt concentration is increased beyond 20 at%. As the
grains with the stacking faults have a much lower anisotropy, they widen the switching field
distribution and cause dc erase noise and adjacent track erasure issues.11 Therefore, the Pt
concentration in recording media is typically kept at a value less than 20 at% (besides cost
consideration).12 In order to achieve higher anisotropy in CoPt systems with Pt concentrations
higher than 20 at%, it is essential to establish methods which can give an estimation of
stacking fault density in a straightforward manner.
In longitudinal recording media, the stacking faults lie perpendicular to the disk surface and
hence a plane view transmission electron microscopy (TEM) could provide visual
information of the density of stacking faults in a given area. 13 However, in perpendicular
recording media, the stacking faults lie along the film-plane and hence cross-section TEM
provides information about stacking faults.14,15 However, the information from cross-section
TEM are statistically insufficient. Therefore, it is essential to identify suitable methods which
can provide an estimation of stacking fault densities as was possible during the era of
longitudinal recording media. In this study, we have made a special set of samples with
various amount of stacking faults and have characterized them using a new magnetic force
microscope (MFM) methodology.
Experimental Details
The samples for this study were prepared on glass disk substrates by dc magnetron sputtering
using Intevac Lean 200 production type sputtering machine. Typical seedlayers, which are
used for achieving hcp(00.2) texture in Co-alloy layer, were deposited prior to the deposition
of the CoCrPt:oxide recording layer.16 The recording layer was covered with a capping layer
and a carbon overcoat.17 The thickness of the granular CoCrPt:oxide layer was kept at 10 nm,
slightly thinner than the typical total thickness of recording media. The samples were
characterized using polar Kerr magnetometry (MOKE), transmission electron microscopy
(TEM), X-ray diffraction (XRD) and MFM. A CoPt layer with a near-equiatomic Pt
concentration (called stacking fault inducing layer, SFI) which increases the stacking faults
with an increase of thickness due to the high Pt concentration, was introduced below the
CoCrPt:Oxide recording layer. Figure 1 shows a schematic of the recording media layer
structure. Figure 1(b) shows the XRD patterns of samples with different SFI layer
thicknesses. It can be noticed from figure that the samples show peaks corresponding to the
Ru seedlayer and the Co alloy layer at around 42.4 and 43.2 degrees, respectively. An
fcc(111) peak induced by the SFI layer, corresponding to the stacking faults, emerges at
around 41.4 degrees in samples with thicker SFI layer, indicating that the SFI layer carries
out its proposed function of producing grains/clusters with lower anisotropy, probably
through the formation of stacking faults.
Figure1: (a) A sketch of the media layer structure with approximate thicknesses. (b) XRD
patterns of samples with different SFI layer thicknesses.
In granular perpendicular recording media, the stacking faults have been characterized using
synchrotron XRD techniques or cross-section TEM methods.18 In cross-section TEM, the
number of grains that can be seen in order to spot the stacking faults is very few. Therefore,
cross-section TEM technique is not statistically sufficient. Synchrotron XRD has been used
to measure the stacking faults.19 Quite recently, laboratory based XRD also has been used to
measure the stacking fault density.20 Although XRD method is suitable for CoPt alloys with
Pt concentration less than 30 at%, a specific problem to XRD arises for near-equiatomic
CoPt, particularly when the peaks of CoPt and Ru overlap with each other. Even though the
“c” lattice parameter of Co is lower than that of Ru, addition of Pt leads to an increase in the
lattice parameter, resulting in an overlap of hcp(00.2) peaks. Considering the drawbacks of
these existing techniques, we have attempted investigating if stacking faults could be
visualized using a simple but effective technique based on MFM. As MFM scan does not
involve any special sample preparation and does not take more than an hour, this method has
potential advantages when many samples need to be investigated. In this study, we have
hypothesized that the grains or clusters with stacking faults will reverse at much lower
reversal fields due to their lower anisotropy than those without stacking faults. Such a
reversal could happen even at remanence, in samples with a higher saturation magnetization
than that of our samples due to higher magnetostatic fields. Therefore, observation of clusters
that reverse at lower fields or even at remanence (nucleation sites) could potentially give
information on the stacking faults, although care needs to be taken in interpreting this
information. Therefore, we have also carried out cross-section TEM to validate the
MFM has been traditionally used to measure recorded bit-patterns of a recording medium.21 It
has also been used to measure the cluster sizes in an AC demagnetized (ACD) state.22 ACD
has also been used to measure the magnetic roughness. 23 Useful correlation has been
identified between the noise of the recording media and cluster sizes or magnetic roughness.
However, MFM has not been used much to image nucleation sites arising from stacking
faults or other sources. In our MFM methodology, we saturated the samples first and then
applied a small reversal field (typically less than the nucleation field of the sample ~ about
1000 Oe in the case of CoCrPt-based perpendicular media). Subsequently, the field was
reduced to zero and samples were moved to the MFM stage and MFM was carried out at
remanence. Figure 2 is an illustration of the methodology, which also shows an example of
MFM image of a sample in this study with reversed nucleation sites (which are potentially
due to stacking faults) as highlighted by dark circle.
Figure 2: (a) Schematic illustration of the magnetic states at which MFM measurements were
carried out. State A represents measurement at remanence after the application of a reversal
field Hr1, which is a fraction of Hn. State B represents measurement at remanence after
saturation (Inset) Typical MFM image of a media sample with reversed captured at the State
A. The circled regions and similar (uncircled) dots are the reversed regions most probably
due to stacking faults.
Results and Discussion
Figure 3 shows the hysteresis loops and other magnetic properties as obtained from MOKE
magnetometry. It can be noticed from figure 3(a) that the reference sample has an Hc and Hn
of about 4500 and 1500 Oe respectively. When the SFI layer with a thickness of 3 nm was
introduced, the Hn and Hc improved considerably to 4900 and 2300 Oe respectively. From
our previous studies on CoPt layer material using TEM and magnetic characterization
techniques, it is known that the CoPt layer does not cause significant stacking faults at this
thickness.24 Moreover, it also does not cause any increase of exchange coupling between the
grains. As a result, the increase in anisotropy energy (KuV) leads to an increase of Hc and Hn
induced by the SFI layer at this thickness. It must be mentioned that the increase in
anisotropy might come from increased Ku of the SFI layer, increase in V of the grain due to
the increased thickness or both. However, the sample with a 6 nm SFI layer shows a lower Hc
(4000 Oe) and Hn (1700 Oe), despite the increased volume. Although the low Hn observed in
this sample (as compared to the sample with 3 nm SFI layer) is believed to be due to the
reduction in Ku arising from stacking faults, the increase in exchange coupling between the
grains of SFI layer is probably another source of reduction in Hn and Hc. Even though the SFI
layer was sputtered at high pressures, it did not have segregants and hence it is expected to
increase exchange coupling in thicker layers. It should, however, be noted that any minor
increase in exchange coupling (as by a cap layer in perpendicular media) would only increase
the nucleation field. Therefore, the decrease of Hn in sample with 6 nm SFI layer is most
likely due to the increased number of stacking faults at this thickness.
Figure 3: (a) MOKE hysteresis loop of the samples with different SFI layer thickness. (b)
Estimated values of anisotropy field (Ho) and the thermal stability factor (KuV/kBT) (from
dynamic coercivity), as a function of SFI layer thickness. Variation of thermal stability with
SFI layer thickness for a constant Ku value of 7x106 erg/cc is also shown.
Time-dependent MOKE measurements (dynamic remanent coercivity) were carried out in
order to gain further understanding on the magnetic properties of these samples. The scanning
time was varied from 10s to 100s in 5 steps. The data were fitted to Sharrock’s equation and
the R2 of the fitting was about 0.99. R2 close to 1 indicates that the data from MOKE were
good enough to obtain information on the anisotropy field (Ho) and the thermal stability
factor (KuV/kBT) (shown in Figure 3(b)).25 The ideal KuV/kBT, (assuming a constant Ku of
7x106 erg/cc for the scenario that the SFI layer did not have any stacking faults), is also
plotted together. It can be noticed that the KuV/kBT increases with the introduction of SFI
layer. However, the increase is not seen to be along the expected lines, except for sample
with 3 nm thick SFI layer. For samples with thicker SFI layer, the increase of KuV/kBT
happens at a much slower rate than the ideal value. In unison, Ho shows the highest value for
SFI layer thickness of 3 nm. For further increase of SFI layer, Ho decreases. These results
indicate that the introduction of SFI layer beyond 3 nm causes stacking faults, which leads to
a reduction in the anisotropy constant of those grains with stacking faults. Since Ho and
KuV/kBT are measured from the time-dependent coercivity (from the magnetization reversal
of about 50% grains), they do not provide an accurate picture of the stacking faults. However,
as the number of the grains with stacking faults is expected to be only a small fraction of the
whole material, they do provide a signature effect of grains with stacking faults. The results
from figure 3 are an indication that the stacking faults, although occur at 3 nm thickness,
increase significantly with SFI thickness beyond 3 nm.
Figure 4(a) shows the MFM images and the number of clusters measured as a function of SFI
layer thickness. The reversed clusters, in dark brown color, can be seen from figure 4(a). The
diameter of the clusters measured from the MFM images in samples with thinner SFI is about
50 nm. It must be mentioned that the size of reversed clusters in MFM images is typically
larger than the actual magnetic dots (or clusters). In patterned media with 30 nm dot sizes, the
reversed dots in MFM were observed to have a size of about 70 nm.26 Therefore, these
clusters may have an actual size of 20-30 nm. These results confirm our hypothesis for this
work that, under the application of a small reversal field, the clusters with a low thermal
stability (due to stacking faults or for compositions that result in a very low anisotropy)
reverse their magnetization and appear in MFM and that the reversal at a small reversal field
(around 1000 Oe) is not due to the small grains.
Figure 4: MFM images of the samples with different SFI layer thicknesses, measured at a
reverse field (Hr) of 1000 Oe and 0 Oe.
To illustrate this point further, we provide a simple calculation based on an isolated CoPt
cluster with a diameter of 20 nm and a thickness of 10 nm. If all the grains in this cluster of
CoPt are in hcp phase, the cluster would have a thermal stability factor of about 380 and it
could not be reversed by a small reversal field (Since Hk of such a cluster is more than 10
kOe). On the other hand, if this cluster has a stacking fault, its anisotropy would be reduced
by a factor, resulting in a thermal stability factor of only about 38 (Clusters with diameters in
the range of 20-25 nm, would have a thermal stability factor between 40 and 60). Such
clusters would flip their magnetization at a small reversal field due to the thermal effects. An
hcp grain of 6 nm diameter will also have similar thermal stability factor value, but such a
grain may not be observable within the resolution limit of MFM (about 20 nm) when its
magnetization is flipped. In essence, the clusters that show up in MFM for small reversal
fields are more likely arising from grains with a lower anisotropy constant (due to stacking
faults, for example) than entirely due to changes in volume of grains. Figure 4(b) shows the
same set of samples, which were measured at state B, which is at remanence after a saturating
field was applied and removed. They do not show any reversed clusters, indicating that the
nucleation sites due to SF or low anisotropy grains must be measured at a suitable reversal
field, depending on the nature of the sample. It can also be noticed from figure 4(a) that the
number of clusters increases as a function of SFI thickness.
Figure 5 shows the number of reversed clusters for different samples, plotted as a function of
SFI thickness. The Hc and Hn values of the samples are also shown in figure 5, which show
an opposite trend, as compared to the number of reversed clusters. In conjunction with the
magnetometry results, such as KuV/kBT, Ho, Hc and Hn, the MFM results indicate that the
nucleation sites due to stacking faults increase as the SFI layer thickness increases from 3 to 6
nm. From figure 4, it was noticed that the size and contrast of the clusters also increase as a
function of thickness. This is due to the increase in exchange coupling, as the thickness of
SFI layer was increased. Although the SFI layer was sputtered at higher pressures to achieve
a segregated structure, this layer did not have the usual oxide segregants. As a result, the
decoupling between the grains deteriorated as thickness increased. This is further confirmed
by the increase of dM/dH with thickness of SFI layer in figure 5 (as measured from the slope
near coercivity of MOKE hysteresis loops).
Figure 5: Effect of SFI layer thickness on the coercivity, nucleation field, dM/dH and number
of clusters observed in MFM.
Although the trends discussed in figure 5 more or less establish the fact that the stacking
faults could be estimated from MFM, it is useful to validate this observation further by
carrying out additional measurements. Although TEM measurement is not as fast as the
MFM measurement, it provides clearer evidence of the stacking faults. Detailed HRTEM
measurements were carried out to find out the microstructure properties of CoCrPt and CoPt.
Figure 6 (a) and (b) show the TEM images of samples with 3 and 6 nm of SFI layer
thickness, respectively, focussed at SFI and CoCrPt layers. In the case of sample with 3 nm
SFI layer, no significant stacking fault could be found in the SFI layer, and the CoCrPt layer
growing on top of it shows stacking faults (indicated by blue arrows). On the other hand, in
the case of sample with 6 nm SFI layer, much more significant number of stacking faults can
be observed in CoPrPt layer as indicated by blue arrows. In both samples, a region of fcc
phase can be found at the CoCrPt/CoPt interface, and phase transformation from hcp to fcc
and vice versa happens. The fcc phase has an order lower anisotropy in comparison to that in
hcp phase. The thickness of this fcc phase increases with the increase of CoPt layer thickness,
which is in consistence with increased number of clusters observed in MFM analysis. the top
portion of the layer exhibits a phase transformation from hcp to fcc structure. The CoCrPt
layer that grows on top of such a grain exhibits stacking faults.
Figure 6. Typical cross-section TEM images of SFI-CoPt(t nm)/CoCrPt:Oxide (10 nm)
samples with (a) 3 nm SFI layer and (b) 6 nm SFI layer.
Figure 7 shows a plot of the number of MFM clusters in the samples as a function of the
stacking faults observed in the same set of samples. Although there is a correlation between
the two, the trend is not linear. This can be understood from the fact that the MFM measures
the number of clusters in a given area (L2) and the cross-section TEM measures the number
of clusters in a scale of length (L). In addition, it was reported that the Ku for pure Co film
can be reduced by more than a factor of 2 due to increase of 10% of the hcc phase,27 which
indicates the fcc phase or the stacking fault can largely alter the Ku value. Therefore, the
quadratic increase of MFM clusters (as compared to the TEM cross-section clusters) is
expected. It must also be mentioned that in order to make our observation representative, the
several MFM and TEM measurements were carried out and the average number of clusters
(or stacking faults) was considered in both cases. These results indicate the correlation
between the MFM clusters and the stacking faults, as observed by cross-section TEM.
Figure 7. Correlation between the MFM cluster density and the number of stacking faults
measured using cross-section TEM
As a conclusion, we have proposed the use of measuring clusters that reversed their
magnetization at small reversal fields as a method to study stacking faults. Our studies on
samples with stacking faults indicates a clear correlation between the number of reversed
clusters in MFM and the coercivity and nucleation field of the samples indicating the
effectiveness of the method. Validation of the MFM results was achieved using TEM
observation of stacking faults. Although MFM is not a direct method for visualizing stacking
faults, the proposed method is useful for the recording media applications in estimating the
density of grains with stacking faults.
S. N. Piramanayagam, J. Appl. Phys., 102, 011301 (2007).
H. J. Richter, J. Magn. Magn. Mater., 321 467 (2009).
D. Suess, Appl. Phys. Lett., 89, 113105 (2006).
S.N. Piramanayagam, J.Z. Shi, H.B. Zhao, C.S. Mah and J. Zhang, IEEE Trans. Magn.,
41(10) 3190 (2005).
J.G. Zhu and Y.M. Wang, IEEE Trans. Magn., 47(10) 4066 (2011).
J. -P. Wang, W. K. Shen, J. M. Bai, R. H. Victora, J. H. Judy, and W. L. Song, Appl. Phys.
Lett., 86, 142504 (2005).
N. Gaur, K. K. M. Pandey, S. L. Maurer, S. N. Piramanayagam, R. W. Nunes, H. Yang, and
C. S. Bhatia, J. Appl. Phys., 110, 083917 (2011).
H. Yuan and D. E. Laughlin, J. Appl. Phys., 105, 07A712 (2009).
T. Kubo, Y. Kuboki, M. Ohsawa, R. Tanuma, A. Saito, T. Oikawa, H. Uwazumi, and T.
Shimatsu, J. Appl. Phys., 97, 10R510 (2005).
T. Shimatsu, H. Sato, T. Oikawa, Y. Inaba, O. Kitakami, S. Okamoto, H. Aoi, H. Muraoka,
and Y. Nakamura, IEEE Trans. Magn., 41, 566 (2005).
J. –G. Zhu, X. Zhu, Y. Tang, IEEE Trans. Magn., 44, 125 (2008).
N. Nozawa, S. Saito, S. Hinata, M. Takahashi, J. Phys. D: Appl. Phys., 46 172001 (2013)
D.E. Laughlin, B. Lu, Y-N. Hsu, J. Zou, and D.N. Lambeth, IEEE Trans. Magn., 36(1) 48
M. Zheng, G. Choe, A. Chekanov, B. G. Demczyk, B. R. Acharya, and K. E. Johnson,
IEEE Trans. Magn., 39(4) 1919 (2003)
B. Lu, T. Klemmer, K. Wierman, G. P. Ju, D. Weller, A. G. Roy, D.E. Laughlin, C. H.
Chang, and R. Ranjan, J. Appl. Phys., 91, 8025 (2002).
S. N. Piramanayagam, K. Srinivasan, J. Mag. Mag. Mater., 321, 485 (2009).
G. Choe, M. Zheng, B.R. Acharya, E.N. Abarra, J.N. Zhou, IEEE Trans. Magn., 41(10)
3172 (2005).
Y. Takahashi, K. Tanahashi, and Y. Hosoe, J. Appl. Phys., 91, 8022 (2002).
H.S. Jung et al., IEEE Trans. Magn., 43(2), 615 (2007).
S. Saito, A. Hashimoto, D. Hasegawa, and M. Takahashi, J. Phys. D: Appl. Phys., 42,
145007, (2009).
M. Futamoto et al., IEEE Trans. Magn., 49(6), 2748, (2013)
Murayama A., Hyomi K., Ohshima K., Miyamura M., Maekawa M., Kondoh S., J. Appl.
Phys., 81 (8), 3925 (1997).
Glijer P., Sivertsen J. M., Judy J. H., IEEE Trans. Magnetics, 31, 2842 (1995).
24 B. Varghese, S.N. Piramanayagam, Y. Yang, S.K. Wong, H.K. Tan, W.K. Lee and I.
Okamoto, J. Appl. Phys., 115, 17B707 (2014)
25 M. P. Sharrock, J. Appl. Phys., 76, 6413 (1994).
M. Ranjbar, A. Tavakkoli, S.N. Piramanayagam, K.P. Tan, R. Sbiaa, S.K. Wong and T.C.
Chong, J. Phys. D: Appl. Phys., 44 265005 (2011)
S. Hinata, R. Yanagisawa, S. Saito and M. Takahashi, J. Appl. Phys., 105, 07B718 (2009)