We report the all-optical observation of intrinsic spin dynamics and extraction of magnetic material parameters from arrays of sub-100 nm spin-transfer torque magnetic random access memory (STT-MRAM) devices with a CoFeB/MgO interface. To this end, the interference of surface acoustic waves with time-resolved magneto-optic signals via magneto-elastic coupling was suppressed using a dielectric coating. The efficacy of this method is demonstrated experimentally and via modeling on a nickel nanomagnet array. The magnetization dynamics for both coated nickel and STT-MRAM arrays shows a restored field-dependent Kittel mode from which the effective damping can be extracted. We observe an increased low-field damping due to extrinsic contributions from magnetic inhomogeneities and variations in the nanomagnet shape, while the intrinsic Gilbert damping remains unaffected by patterning. The data are in excellent agreement with a local resonance model and have direct implications for the design of STT-MRAM devices as well as other nanoscale spintronic technologies.

As current memory architectures approach scaling limits, there is a strong motivation for a technology that utilizes new physical principles to provide scalability. Spin-transfer-torque magnetic random access memory (STT-MRAM) has emerged as one of the most promising spintronic technologies, offering nanosecond read/write operation, non-volatile storage, and high areal bit densities.1 It is well known that the Gilbert damping parameter, α, determines the critical operation characteristics such as the threshold write current,2,3 switching time,4 and transition jitters.5 Harnessing interfacially induced spin-orbit interactions, highly anisotropic CoFeB layers with intrinsically low α have been identified as the best choice for STT-MRAM.6 There are a limited number of reports that have addressed what effect nanopatterning has on the damping in this class of devices7–11 and none that investigates them in a densely packed geometry, which is essential for commercial STT-MRAM.

Optical techniques, such as the frequency and time-resolved magneto-optic Kerr effect (FR and TR-MOKE, respectively), provide a high spatial resolution and have been used to probe damping in films, arrays, and single nanomagnets.12–16 In TR-MOKE measurements, the sample is irradiated by a femtosecond pulse to instigate dynamics. The measured response is a superposition of the ultrafast demagnetization, thermal decay, and a damped small-angle precession of the magnetization about the effective applied field. By omitting the initial thermalization and subtracting the subsequent exponentially decaying background, the time-evolution of the magnetic signal is isolated and fit with a damped sinusoid to obtain an effective damping parameter, αeff. This parameter is heavily influenced by extrinsic contributions that can only be quantified by evaluating the field dependence of αeff.16–18 However, it was recently shown that the rapid thermal expansion of periodically arranged nanomagnets generates surface acoustic waves (SAWs) that dynamically couple to the magnetization dynamics over a wide range of fields, resulting in a convoluted magnetoelastic (MEL) response19 that makes damping analysis impossible.

In this letter, we report the restoration of the natural magnetic response of a nanopatterned array by adding a smoothing dielectric film (Fig. 1(a)) to quench SAWs while preserving, or improving, the dynamic magneto-optic response of the sample. Applying this technique to two distinct nanomagnet arrays of polycrystalline nickel and a pre-commercial STT-MRAM prototype, we present measurements of damping in densely packed STT-MRAM devices. We observe a pronounced enhancement of αeff at low applied fields. These findings are explained by the same Gilbert damping parameter as in an unpatterned film and dynamic dephasing caused by magnetic inhomogeneities (MI) and inter-element shape variations.

FIG. 1.

(a) Schematic side-view of the Ni array with the SiN coating and the geometry of the applied field (b) (c) Top-down SEM image of the (b) Ni and MTJ (c) arrays without the SiN coating.

FIG. 1.

(a) Schematic side-view of the Ni array with the SiN coating and the geometry of the applied field (b) (c) Top-down SEM image of the (b) Ni and MTJ (c) arrays without the SiN coating.

Close modal

To demonstrate the technique, two identical nickel arrays composed of square nanomagnets with nominal dimensions of 125 × 125 × 30 nm3 and pitch (p) of 250 nm (Fig. 1(b)) were fabricated on an antireflection (AR) coated silicon (100) substrate using an established electron beam lithography and lift-off process.19–21 A 65 nm SiN film was then uniformly deposited onto the surface of one sample using plasma-enhanced chemical vapor deposition. The arrays were studied using a previously described two-color TR-MOKE setup,22 whose pump and probe pulses (pulse width - 165 fs, repetition rate - 76 MHz) were focused onto the sample with 1/e2 radii of approximately 5 and 2.5 μm, respectively. The nanomagnets are quasi-instantaneously thermalized upon absorption of the pump pulse (modulated at 1 kHz), instigating both magnetization and acoustic dynamics. The probe pulse is delayed with a mechanical delay line and experiences a polarization rotation upon reflection from the sample. Lock-in detection at the pump modulation frequency is then used to record the Kerr rotation by measuring the polarization direction in a balanced photodetector setup.

A series of measurements were taken with an externally applied field Happ kept at a fixed angle θH=30° from the surface normal (see Fig. 1(a)). At each field value, the time-dependent signal is converted into the frequency domain using a FFT algorithm. The FFT power spectra are taken in 250 Oe increments, and stitched together to create the field-dependent Fourier maps shown in Fig. 2. In Fig. 2(a), the normalized Fourier spectra of a Ni film evince only the well-defined Kittel mode in stark contrast with measurements of the uncoated Ni nanomagnet array (Fig. 2(b)), which contain several pronounced field independent frequencies. Inspection of the time-dependent reflectivity signal (nonmagnetic, not shown) reveals that the pinning occurs at well-defined SAW modes (marked by arrows), known to exist when periodically arranged elements are irradiated with an ultrafast pulse.23,24 As the spin waves approach the SAW frequencies, the two become magnetoelastically coupled and the magnetic response is pinned at the acoustic eigenmodes over a range of fields; enhancing the oscillation amplitude and lifetime thereby rendering the damping analysis impossible. This behavior can be modeled by simulating both elastic motion and the magnetization dynamics. The strain profile obtained from finite element simulations of the optically excited nickel elements creates a magneto-elastic contribution HMEL in the effective applied field entering the Landau-Lifshitz-Gilbert equation in Object-Oriented Micromagnetic Framework (OOMMF).25 In order to suppress this coupling, we coated the array with a dielectric silicon nitride (SiN) film (Fig. 2(c)). Qualitatively, the elements' physical motion is restrained which quenches the SAWs. The field-dependent intrinsic magnetization dynamics are restored as seen in Fig. 2(c).

FIG. 2.

Normalized, field-dependent TR-MOKE Fourier spectra (θH=30°) of (a) a Ni film displaying the characteristic Kittel mode; (b) an uncoated nanomagnet array showing distinct pinning of resonances to surface acoustic wave frequencies marked by dark arrows as well as a surface skimming longitudinal wave (SSLW) resonance marked by the white arrow, and (c) a nanomagnet array coated with 65 nm SiN coating which quenches the SAWs and restores the intrinsic response of the magnetization.

FIG. 2.

Normalized, field-dependent TR-MOKE Fourier spectra (θH=30°) of (a) a Ni film displaying the characteristic Kittel mode; (b) an uncoated nanomagnet array showing distinct pinning of resonances to surface acoustic wave frequencies marked by dark arrows as well as a surface skimming longitudinal wave (SSLW) resonance marked by the white arrow, and (c) a nanomagnet array coated with 65 nm SiN coating which quenches the SAWs and restores the intrinsic response of the magnetization.

Close modal

Numerical simulations of the dynamics of both arrays are shown in Fig. 3. Fig. 3(a) reveals that the oscillation amplitude of the HMEL field in the coated sample is reduced by more than a factor of 10 and decays in picoseconds compared to the MEL coupled response of the bare sample that persists for nanoseconds. Fig. 3(b) then shows that the coating eliminates the Fourier components at the elastic resonances, in agreement with the removal of the pinning of the magnetic response as seen in Fig. 2(c).

FIG. 3.

(a) Micromagnetic simulation of magnetoelastic field created by SAWs with and without a dielectric coating (traces are offset for clarity); (b) the discrete Fourier transform (Happ=5kOe|θH=30°). The coated response is offset for display; insets: FEM geometries used for modelling.

FIG. 3.

(a) Micromagnetic simulation of magnetoelastic field created by SAWs with and without a dielectric coating (traces are offset for clarity); (b) the discrete Fourier transform (Happ=5kOe|θH=30°). The coated response is offset for display; insets: FEM geometries used for modelling.

Close modal

We now turn our attention to the array of cylindrical STT-MRAM devices shown in Fig. 1(c) with an average diameter of 70 nm on a pitch p=250nm. The sample is a pre-commercial prototype with a CoFeB free layer (FL) embedded between two MgO tunneling barriers and grown on thermal oxide Si wafer with a metal seed layer. Fig. 4(a) shows the TR-MOKE spectra measured (θH=80°) of a film possessing perpendicular magnetic anisotropy (PMA) and low damping of the FL. However, measurements of the uncoated array at a few representative field values (Fig. 4(b)) yield a completely field-independent response due to MEL coupling and enhanced leakage of the nonmagnetic signal attributed to the highly reflective substrate (R80%). Again, we deposit a 65 nm thick layer of SiN uniformly onto the array surface to both suppress SAWs and maximize the magneto-optical response via the cavity enhancement effect described in Ref. 26. Once coated, the array response (Fig. 4(c)) is restored and a well-defined Kittel mode is observed throughout the field range. Although the layer sequence is similar, it is worth noting that the film has a FL which is approximately 7 Å thicker than in the array sample which produces a smaller anisotropy field, HK, and translates to a shift in the spin wave frequencies.

FIG. 4.

Normalized Fourier spectra (θH=80°) measured on (a) unpatterned CoFeB film (HK2kOe), (b) uncoated patterned array (HK3kOe) and (c) SiN coated array. The higher frequency resonance of the array is due to the larger anisotropy field.

FIG. 4.

Normalized Fourier spectra (θH=80°) measured on (a) unpatterned CoFeB film (HK2kOe), (b) uncoated patterned array (HK3kOe) and (c) SiN coated array. The higher frequency resonance of the array is due to the larger anisotropy field.

Close modal

Now, the time-domain response at each field can be fit with a damped sinusoid to obtain the oscillation lifetime (τeff) which is used to determine the effective damping using the relation αeff=1/2πfτeff.27,28 First comparing the Ni film and uncoated array in Fig. 5(a), we find the SAWs extend τeff and result in much smaller αeff values than the intrinsic α of 0.04 observed in the film. The most extreme instance was observed at Happ=5kOe, where the decay constant measured for the array (τeff5ns) is nearly 16 times larger than the purely magnetic response observed in the film (τeff=300ps) due to MEL coupling. In contrast, the SiN coated array (solid circles) shows an increased αeff throughout the entire field range that converges to the film value at large external fields. This increase can be predominantly attributed to variations between the individual magnetic elements that lead to a dynamic dephasing and, thus, more rapid damping of the averaged temporal response.29 Here, the significance of the data lies in the restoration of the magnetization dynamics by the dielectric coating which re-opens the door for extraction of pertinent material parameters.

FIG. 5.

(a) Effective damping of the Ni film and array with and without coating (θH = 30°).The dashed lines are guides to the eye. Without coating, the SAWs strongly couple to the precession and extend the precession lifetime. Suppression restores the natural relaxation revealing a small enhancement of αeff. (b) Effective damping in the CoFeB|MgO array compared to a film (θH=80°), solid lines are fits using a local resonance model to account for the distribution of PMA; (c) Micromagnetic modelling of the dephasing effect due to shape variations alone (Happ=5kOe|θH=80°).

FIG. 5.

(a) Effective damping of the Ni film and array with and without coating (θH = 30°).The dashed lines are guides to the eye. Without coating, the SAWs strongly couple to the precession and extend the precession lifetime. Suppression restores the natural relaxation revealing a small enhancement of αeff. (b) Effective damping in the CoFeB|MgO array compared to a film (θH=80°), solid lines are fits using a local resonance model to account for the distribution of PMA; (c) Micromagnetic modelling of the dephasing effect due to shape variations alone (Happ=5kOe|θH=80°).

Close modal

For a more quantitative analysis, we focus on the damping behavior of the technologically important CoFeB-based STT-MRAM devices. Fig. 5(b) compares uncoated film and coated nanomagnet array which exhibits an even larger increase of αeff for the patterned sample. This increase is primarily due to two extrinsic effects. The first is a contribution from magnetic inhomogeneities (MI), due to variations of the local effective anisotropy field, HK.27,28 The second one is a variation in nanomagnet shapes which result in a spread of precession frequencies and thus, through dynamic dephasing, larger damping of the time response. This mechanism applies only to the patterned array. When the applied field cannot reach large enough values to converge to the intrinsic Gilbert damping α, as is the case here, one can relate the effective lifetime, τeff, obtained with TR-MOKE to the Lorentzian resonance linewidth, Δωeff, using the relation Δωeff=2/τeff. The extrinsic contributions to the effective linewidth are given by27 

Δωeff=Δωint+ΔωextΔωint+ΔωMI+ΔωS,
(1)

where Δωint is given by the Smit-Suhl formula, ΔωMI represents the dispersion in the resonance frequencies due to MI and ΔωS the spread due to nanomagnet shape fluctuations. If the distribution of resonant frequencies is assumed to be primarily caused by a distribution of the local anisotropy field with a peak-to-peak width, ΔHK, then ΔωMI can be approximated as30 

ΔωMI|ω/HK|ΔHK+|ω/θH|ΔθH,
(2)

where ΔθH represents a spread due to orientation of grains and can be directly related to ΔHK in the vicinity of a resonance.31 However, it can be shown that ΔωMI vanishes when the external field is applied perpendicular to the sample plane.32 FMR measurements employing this configuration were performed on both samples (films) and the intrinsic α for the film, and array FLs were found to be 0.003 and 0.005, respectively. We reiterate that the two samples are not identical; the FL in the array is approximately 7 Å thinner than the film shown which is known to correspond to larger HK and α values.6 First fitting τeff for the film using only ΔHK as a free parameter we find an excellent agreement to the data (lines in Fig. 5(b)) using ΔHK,film=280Oe. In order to quantify ΔωS, we analyzed the SEM image (Fig. 1(c)) to extract the actual device shapes for 20 elements and used the resulting demagnetization field profiles as input for micromagnetic modelling of the magnetization dynamics. The simulations confirm a significant spread in the precession frequencies due to shape variations. For example, at Happ=4kOe, the individual nanomagnet resonances are distributed between 13.4–14.5 GHz. We verified that the effect of this spread on αeff is independent of the number of devices once 10 or more elements are considered. By superimposing the time evolution of all the elements, we then estimate the impact of dephasing due to shape variations. In Fig. 5(c), the calculated effective lifetime of the nanomagnet ensemble is considerably shorter than for a single device. Using this approach we computed the field dependence of the ensemble and fit the time-domain response at each field with a damped harmonic to obtain τeff. We then estimate Δωs using Equation (2) and obtain the spread of local anisotropies due to shape distortions ΔHK,shape150Oe. If the spread ΔHK,array in the array is a combination of the distribution present before nanopatterning and the additional inhomogeneity due to shape, ΔHK,array=ΔHK,film+ΔHK,shape, our modelling estimates a value of ΔHK,array=430Oe which is remarkably close to the value extracted from the fit in Fig. 5(b) using ΔHK=480Oe. The difference between the experimental and predicted values of Δωext may be explained by a slightly larger inherent ΔHK,film in the patterned sample due to the thinner CoFeB layer thickness, as well as unaccounted roughness at the edges of the device or oversimplification of the dynamics using a macrospin approximation.

We have demonstrated that addition of a dielectric smoothing layer can suppress the optically excited SAWs that interfere with the extraction of intrinsic magnetic material parameters. This technique was used to resolve the spin-dynamics of an array of 70 nm STT-MRAM devices. Our findings reveal a significant enhancement of the αeff due to nanopatterning which is fit remarkably well using a local resonance model. Using SEM imaging and micromagnetic modelling we confirm that local variations of the anisotropy field and nanomagnet shape variations are the primary additions to the effective linewidth, as opposed to a change in the intrinsic damping.

This work was supported by the National Science Foundation under Grant Nos. ECCS-1509020 and DMR-1506104. Work at the Molecular Foundry, Lawrence Berkeley National Laboratory was supported by the Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. We acknowledge T. Yuzvinsky and the W. M. Keck Center for Nanoscale Optofluidics at the University of California at Santa Cruz for SEM imaging.

1.
S.
Tehrani
,
J. M.
Slaughter
,
M.
Deherrera
,
B. N.
Engel
,
N. D.
Rizzo
,
J.
Salter
,
M.
Durlam
,
R. W.
Dave
,
J.
Janesky
,
B.
Butcher
,
K.
Smith
, and
G.
Grynkewich
,
Proc. IEEE
91
,
703
(
2003
).
2.
3.
P. B.
Visscher
and
D. M.
Apalkov
,
J. Appl. Phys.
97
,
10C704
(
2005
).
4.
X.
Wang
,
K.
Gao
,
H.
Zhou
,
A.
Itagi
,
M.
Seigler
, and
E.
Gage
,
IEEE Trans. Magn.
49
,
686
(
2013
).
5.
P.-W.
Huang
and
R. H.
Victora
,
J. Appl. Phys.
115
,
17B710
(
2014
).
6.
A. V.
Khvalkovskiy
,
D.
Apalkov
,
S.
Watts
,
R.
Chepulskii
,
R. S.
Beach
,
A.
Ong
,
X.
Tang
,
A.
Driskill-Smith
,
W. H.
Butler
,
P. B.
Visscher
,
D.
Lottis
,
E.
Chen
,
V.
Nikitin
, and
M.
Krounbil
,
J. Phys. D: Appl. Phys.
46
,
074001
(
2013
).
7.
S.
Ikeda
,
K.
Miura
,
H.
Yamamoto
,
K.
Mizunuma
,
H. D.
Gan
,
M.
Endo
,
S.
Kanai
,
J.
Hayakawa
,
F.
Matsukura
, and
H.
Ohno
,
Nat. Mater.
9
,
721
724
(
2010
).
8.
S.
Tamaru
,
H.
Kubota
,
K.
Yakushiji
,
M.
Konoto
,
T.
Nozaki
,
A.
Fukushima
,
H.
Imamura
,
T.
Taniguchi
,
H.
Arai
,
S.
Tsunegi
,
S.
Yuasa
, and
Y.
Suzuki
,
J. Appl. Phys.
115
,
17C740
(
2014
).
9.
T.
Devolder
,
K.
Garcia
,
G.
Agnus
,
M.
Manfrini
,
S.
Cornelissen
, and
T.
Min
,
Appl. Phys. Lett.
103
,
182402
(
2013
).
10.
E.
Hirayama
,
S.
Kanai
,
H.
Sato
,
F.
Matsukura
, and
H.
Ohno
,
J. Appl. Phys.
117
,
17B708
(
2015
).
11.
K.
Mizunuma
,
M.
Yamanouchi
,
H.
Sato
,
S.
Ikeda
,
S.
Kanai
,
F.
Matsukura
, and
H.
Ohno
,
Appl. Phys. Express
6
,
063002
(
2013
).
12.
J. M.
Shaw
,
T. J.
Silva
,
M. L.
Schneider
, and
R. D.
McMichael
,
Phys. Rev. B
79
,
184404
(
2009
).
13.
A.
Barman
,
S.
Wang
,
J.
Maas
,
A. R.
Hawkins
,
S.
Kwon
,
J.
Bokor
,
A.
Liddle
, and
H.
Schmidt
,
Appl. Phys. Lett.
90
,
202504
(
2007
).
14.
B.
Rana
,
D.
Kumar
,
S.
Barman
,
S.
Pal
,
Y.
Fukuma
,
Y.
Otani
, and
A.
Barman
,
ACS Nano
5
,
9559
(
2011
).
15.
S.
Mizukami
,
J. Magn. Soc. Jpn.
39
,
1
7
(
2015
).
16.
G.
Malinowski
,
K. C.
Kuiper
,
R.
Lavrijsen
,
H. J. M.
Swagten
, and
B.
Koopmans
,
Appl. Phys. Lett.
94
,
102501
(
2009
).
17.
S.
Mizukami
,
D.
Watanabe
,
T.
Kubota
,
X.
Zhang
,
H.
Naganuma
,
M.
Oogane
,
Y.
Ando
, and
T.
Miyazaki
,
Appl. Phys. Express
3
,
123001
(
2010
).
18.
H.-S.
Song
,
K.-D.
Lee
,
J.-W.
Sohn
,
S.-H.
Yang
,
S. S. P.
Parkin
,
C.-Y.
You
, and
S.-C.
Shin
,
Appl. Phys. Lett.
102
,
102401
(
2013
).
19.
Y.
Yahagi
,
B.
Harteneck
,
S.
Cabrini
, and
H.
Schmidt
,
Phys. Rev. B
90
,
140405
(
2014
).
20.
A.
Barman
,
S.
Wang
,
J. D.
Maas
,
A. R.
Hawkins
,
S.
Kwon
,
A.
Liddle
,
J.
Bokor
, and
H.
Schmidt
,
Nano Lett.
6
,
2939
(
2006
).
21.
Z.
Liu
,
R.
Brandt
,
Y.
Yahagi
,
B.
Hansen
,
B.
Harteneck
,
J.
Bokor
,
A. R.
Hawkins
, and
H.
Schmidt
,
Appl. Phys. Lett.
98
,
052502
(
2011
).
22.
A.
Barman
,
S.
Wang
,
O.
Hellwig
,
A.
Berger
,
E. E.
Fullerton
, and
H.
Schmidt
,
J. Appl. Phys.
101
,
09D102
(
2007
).
23.
J.-F.
Robillard
,
A.
Devos
, and
I.
Roch-Jeune
,
Phys. Rev. B
76
(
9
),
092301
(
2007
).
24.
C.
Giannetti
,
B.
Revaz
,
F.
Banfi
,
M.
Montagnese
,
G.
Ferrini
,
F.
Cilento
,
S.
Maccalli
,
P.
Vavassori
,
G.
Oliviero
,
E.
Bontempi
,
L. E.
Depero
,
V.
Metlushko
, and
F.
Parmigiani
,
Phys. Rev. B
76
(
12
),
125413
(
2007
).
25.
A. G.
Gurevich
and
G. A.
Melkov
,
Magnetization Oscillations and Waves
(
CRC Press
,
New York
,
1996
), pp.
10
20
.
26.
S.
Wang
,
A.
Barman
,
H.
Schmidt
,
J. D.
Maas
,
A. R.
Hawkins
,
S.
Kwon
,
B.
Harteneck
,
S.
Cabrini
, and
J.
Bokor
,
Appl. Phys. Lett.
90
,
252504
(
2007
).
27.
A.
Capua
,
S.
Yang
,
T.
Phung
, and
S. S. P.
Parkin
,
Phys. Rev. B
92
,
224402
(
2015
).
28.
S.
Iihama
,
S.
Mizukami
,
H.
Naganuma
,
M.
Oogane
,
Y.
Ando
, and
T.
Miyazaki
,
Phys. Rev. B
89
,
174416
(
2014
).
29.
H. T.
Nembach
,
J. M.
Shaw
,
T. J.
Silva
,
W. L.
Johnson
,
S. A.
Kim
,
R. D.
McMichael
, and
P.
Kabos
,
Phys. Rev. B
83
,
094427
(
2011
).
30.
J.
Linder
,
I.
Barsukov
,
C.
Raeder
,
C.
Hassel
,
O.
Posth
, and
R.
Meckenstock
,
Phys. Rev. B
80
,
224421
(
2009
).
31.
J.
Dubowik
,
K.
Zaleski
,
H.
Głowinski
, and
I.
Goscianska
,
Phys. Rev. B
84
,
184438
(
2011
).
32.
P.
Krivosik
,
S. S.
Kalarickal
,
N.
Mo
,
S.
Wu
, and
C. E.
Patton
,
Appl. Phys. Lett.
95
,
052509
(
2009
).