Introduction to spin wave computing

Cite as: J. Appl. Phys. 128, 161101 (2020); https://doi.org/10.1063/5.0019328

Submitted: 22 June 2020 • Accepted: 28 September 2020 • Published Online: 22 October 2020

Abdulqader Mahmoud, Florin Ciubotaru, Frederic Vanderveken, et al.

ARTICLES YOU MAY BE INTERESTED IN

Introduction to antiferromagnetic magnons

Journal of Applied Physics 126, 151101 (2019); https://doi.org/10.1063/1.5109132

Spin–orbit-torque magnonics
Journal of Applied Physics 127, 170901 (2020); https://doi.org/10.1063/5.0007095

Magnetization dynamics of nanoscale magnetic materials: A perspective

Journal of Applied Physics 128, 170901 (2020); https://doi.org/10.1063/5.0023993

J. Appl. Phys. 128, 161101 (2020); https://doi.org/10.1063/5.0019328 128, 161101

© 2020 Author(s).
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

Introduction to spin wave computing

Cite as: J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328
Submitted: 22 June 2020 · Accepted: 28 September 2020 · View Online Export Citation CrossMark
Published Online: 22 October 2020

Abdulqader Mahmoud,1 Florin Ciubotaru,2 Frederic Vanderveken,2,3 Andrii V. Chumak,4

1 2,a)
Said Hamdioui, Christoph Adelmann, and Sorin Cotofana1,b)

AFFILIATIONS
1
Department of Quantum and Computer Engineering, Delft University of Technology, 2628 CD Delft, The Netherlands
2
Imec, 3001 Leuven, Belgium
3
Department of Materials, SIEM, KU Leuven, 3001 Leuven, Belgium
4
Faculty of Physics, University of Vienna, 1090 Wien, Austria

a)
Electronic mail: [email protected]
b)
Author to whom correspondence should be addressed: [email protected]

ABSTRACT
This paper provides a tutorial overview over recent vigorous efforts to develop computing systems based on spin waves instead of charges
and voltages. Spin-wave computing can be considered a subfield of spintronics, which uses magnetic excitations for computation and
memory applications. The Tutorial combines backgrounds in spin-wave and device physics as well as circuit engineering to create synergies
between the physics and electrical engineering communities to advance the field toward practical spin-wave circuits. After an introduction
to magnetic interactions and spin-wave physics, the basic aspects of spin-wave computing and individual spin-wave devices are reviewed.
The focus is on spin-wave majority gates as they are the most prominently pursued device concept. Subsequently, we discuss the current
status and the challenges to combine spin-wave gates and obtain circuits and ultimately computing systems, considering essential aspects
such as gate interconnection, logic level restoration, input–output consistency, and fan-out achievement. We argue that spin-wave circuits
need to be embedded in conventional complementary metal–oxide–semiconductor (CMOS) circuits to obtain complete functional hybrid
computing systems. The state of the art of benchmarking such hybrid spin-wave–CMOS systems is reviewed, and the current challenges to
realize such systems are discussed. The benchmark indicates that hybrid spin-wave–CMOS systems promise ultralow-power operation and
may ultimately outperform conventional CMOS circuits in terms of the power-delay-area product. Current challenges to achieve this goal
include low-power signal restoration in spin-wave circuits as well as efficient spin-wave transducers.

Published under license by AIP Publishing. https://doi.org/10.1063/5.0019328

I. INTRODUCTION technology have been steadily improved by geometric Dennard

Current computing systems rely on paradigms, in which infor- scaling,6 following the famed Moore’s law.7 This progress has been
mation is represented by electric charge or voltage, and computa- orchestrated first in the USA by the national technology roadmap
tion is performed by charge movements. The fundamental circuit for semiconductors and, after 1998, worldwide by the international
element in this framework is the transistor, which can serve both as technology roadmap for semiconductors (ITRS).8 This has allowed
a switch and an amplifier. Today’s large-scale integrated circuits are CMOS technology to simultaneously drive and respond to an
based on complementary metal–oxide–semiconductor (CMOS) exploding information technology market. Today, CMOS has
field-effect transistors because of their high density, low power con- clearly consolidated its leading position in the digital domain.
sumption, and low fabrication cost.1–3 Using CMOS transistors, In the last two decades, CMOS scaling has increasingly required
logic gates can be built that represent a full set of Boolean algebraic the introduction of disruptive changes in the CMOS transistor and
operations. Such basic Boolean operations are fundamental for the circuit architecture beyond Dennard scaling to sustain Moore’s
design of mainstream logic circuits and, together with charge-based law,9,10 e.g., Cu interconnects,11 high-κ dielectrics,12 or the FINFET
memory devices, of computing systems.4,5 architecture.13 In the future, CMOS scaling is expected to deceler-
In the first decades after its introduction into the mainstream ate14 mainly due to unsustainable power densities, high source–drain
in 1974, the device density and the performance of the CMOS and gate leakage currents,15,16 reduced reliability,17 and economical

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-1

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

inefficiency.15,17 Yet, despite the slowdown, Moore’s law and CMOS

scaling are not expected to end in the next decade and even beyond.
The roadmap for future developments is summarized in the
International Roadmap for Devices and Systems (IRDS).18
For many years, Moore’s law (especially the threat of its end)
has been accompanied by research on alternative computing para-
digms beyond the CMOS horizon to further improve computation
platforms.18–28 Recently, this has accelerated due to a surge of inter-
est in non-Boolean computing approaches for machine learning
applications.29–31 Such computing paradigms can be based on
devices with transistor functionality (e.g., tunnel FETs)32 or alterna-
tives (e.g., memristors).33,34 Among all beyond-CMOS approaches,
spintronics, which uses magnetic degrees of freedom instead of
electron charge for information coding,35–40 has been identified as
particularly promising due to the low intrinsic energies of magnetic
excitations as well as their collective nature.25–27,41,42 Numerous
implementations of spintronic Boolean logic devices have been inves-
tigated based on magnetic semiconductors,43 individual atomic
spins,44 spin currents,45–47 nanomagnets,48–52 domain walls,53–55 sky-
rmions,56,57 or spin waves.58–60 While some approaches try to
provide transistor-like functionality,43,45–47,61 others aim at replacing
logic gates rather than individual transistors.58–60,62,63 Among the
latter group of spintronic logic gates, majority gates have received
particular attention due to the expected simplification of logic cir-
cuits.27,59,64,65 While majority gates have been researched for FIG. 1. Overview of different envisaged applications of spin waves. This Tutorial
decades,66 their CMOS implementation is inefficient and therefore focuses on applications in digital logic based on hybrid spin-wave–CMOS com-
puting systems. Other application fields are reviewed briefly in Sec. VIII.
has not been widely used in circuit design. However, the advent of
compact (spintronic) majority gates has recently led to a revival of
majority-based circuit synthesis.64,67,68
A group of disruptive spintronic logic device concepts have successfully compete with CMOS in practice. Such an effort is inher-
been based on spin waves as information carriers.60,61,69–76 Spin ently multidisciplinary and needs to involve both spin-wave and
waves are oscillatory collective excitations of the magnetic device physics as well as circuit and systems engineering. This paper
moments in ferromagnetic or antiferromagnetic media77–79 and are provides a tutorial introduction to spin-wave computing technology
introduced in more detail in Sec. II. As their quanta are termed and its potentials from a circuit and computation viewpoint. The
magnons, the field is also often referred to as magnonics. The fre- focus is on the achievements but also on the gaps in the current
quency of spin waves in ferromagnets is typically in the GHz range, understanding that still inhibit the realization of practical competitive
their intrinsic energy is low (
μeV for individual magnons), and spin-wave circuits. The main goal of the Tutorial is to provide simul-
their propagation velocity can reach values up to several km/s taneous insight into the underlying physics and the engineering chal-
(μm/ns). At low amplitude, spin waves are noninteracting, enabling lenges to facilitate mutual synergistic interactions between the fields.
multiplexing and parallelism in logic devices and interconnec- The paper starts with an introduction to the physics of spin waves
tions.80 By contrast, spin waves can exhibit nonlinear behavior at (Sec. II). Subsequently, the computation paradigm based on spin
high amplitudes (Sec. II C), which can be exploited in spintronic waves is introduced and the fundamental requirements for the realiza-
devices and circuits (Secs. V and VI). As shown in Sec. IV D, spin tion of spin-wave circuits are discussed (Sec. III). Next, we provide an
waves are especially suitable for the implementation of compact overview of different spin-wave transducers (Sec. IV) and devices
majority gates due to their wave-like nature. Their short wave- (Sec. V). This is followed by a discussion of the current understanding
lengths down to the nm range at microwave (GHz) frequencies of spin-wave circuits (Sec. VI) and computing platforms (Sec. VII).
allow for the miniaturization of the devices while keeping operating Beyond digital computation, spin waves have also the potential to be
frequencies accessible. used in a number of additional application fields in electronics
In the last two decades, magnetic devices have been successfully (Fig. 1). This is briefly reviewed in Sec. VIII. Finally, Sec. IX concludes
commercialized for nonvolatile memory applications (magnetic the paper with an overview of the state of the art of spin-wave technol-
random-access memory, MRAM)81–85 and as magnetic sensors.86–88 ogy and identifies the challenges ahead toward the design and realiza-
Yet, despite tremendous progress in the theory and numerous tion of competitive spin-wave-based computing systems.
proof-of-concept realizations of spintronic and magnonic logic
devices, no competitive spintronic or magnonic logic circuits have
been demonstrated to date. It is clear that the step from individual II. PHYSICS OF SPIN WAVES
basic spintronic device concepts to operational circuits and systems This section provides an introduction to spin waves and their
is large and an additional complementary effort is still required to characteristics. We first start by explaining the relevant basic

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-2

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

magnetic interactions, followed by a discussion of the resulting exist, depending on the crystal structure.128 As an example, the
magnetization dynamics. energy density for uniaxial magnetocrystalline anisotropy can be
expressed by
A. Magnetization and magnetic interactions
Eani ¼ K1 (u  ζ)2  K2 (u  ζ)4 , (4)
Magnetic materials contain atoms with a net magnetic dipole
moment μ. Therefore, they can be considered a lattice of magnetic with u being the easy axis, ζ ¼ M=Ms being the magnetization
dipoles with specific amplitude and orientation at every lattice site. direction, and K1 and K2 being the first and second order anisot-
At dimensions much larger than the interatomic distances, it is ropy constants, respectively.
more convenient to work with a continuous vector field than with It is often convenient to describe magnetic interactions by cor-
discrete localized magnetic dipoles, i.e., with the so-called semi- responding effective magnetic fields. The general relation between a
classical approximation. The continuous vector field is called the magnetic energy density and its corresponding effective field is
magnetization and is defined as the magnetic dipole moment per given by
unit volume,126
P 1 dE(M)
μi Heff ¼  : (5)
M¼ i
: (1) μ0 dM
δV
At temperatures far below the Curie temperature, the magnetiza- For the magnetocrystalline interaction, this becomes
tion norm is constant throughout the material and is called the
saturation magnetization Ms . On the other hand, the magnetization 2K1 4K4
Hani ¼ (u  ζ)u þ (u  ζ)3 u: (6)
orientation can be position dependent and is determined by μ0 M s μ0 Ms
various magnetic interactions. In the following, the most important
magnetic interactions are briefly explained. In the case of polycrystalline materials, every grain may possess a
The Zeeman interaction describes the influence of an external different easy axis orientation. Therefore, the average magnetocrys-
magnetic field Hext on the magnetization. The Zeeman energy talline anisotropy in macroscopic polycrystalline materials is zero
density (energy per unit volume) is given by and can be neglected, as it can be for amorphous materials.
Another important magnetic interaction is the exchange inter-
EZ ¼ μ0 M  Hext , (2) action. It describes the coupling between neighboring magnetic
dipoles and has a quantum-mechanical origin. In continuum
with μ0 being the vacuum permeability. Hence, the energy is theory, the exchange energy density is given by
minimal when the magnetization is parallel to the external mag-
netic field. Aex  
Apart from external magnetic fields, the magnetization itself Eex ¼ 2
(∇Mx )2 þ (∇My )2 þ (∇Mz )2 , (7)
Ms
also generates a magnetic field, termed the dipolar magnetic field.
For a given magnetization state, it is found by solving Maxwell’s with Aex being the exchange stiffness constant. In ferromagnetic
equations.77 The dipolar magnetic field inside the magnetic mate- materials, the exchange stiffness constant is positive, which means
rial is called the demagnetization field, whereas the field outside is that the exchange energy is minimum when the magnetization is
called the stray field. The energy density of the self-interaction of uniform. In antiferromagnetic materials, the exchange stiffness
the magnetization with its own demagnetization field is given by constant is negative, and the exchange energy is minimum when
μ0 neighboring atomic dipoles are antiparallel. The corresponding
Ed ¼  M  Hd , (3) exchange field is given by
2
with Hd being the demagnetization field. The demagnetization field Hex ¼
2Aex
ΔM ¼ lex
2
ΔM ; λex ΔM, (8)
itself strongly depends on the shape of the magnetic element.126,127 μ0 Ms2
The demagnetization energy is minimal when the magnetization is
oriented along the longest dimension of the magnetic object. This with Δ being the Laplace operator, λex is the exchange constant,
magnetization anisotropy is therefore often called shape anisotropy. and lex the exchange length. This length is typically a few nm
The crystal structure of the magnetic material can also intro- (Table I) and characterizes the competition between the exchange
duce an anisotropic behavior of the magnetization. This is called and dipolar interaction. At length scales below lex , the exchange
magnetocrystalline anisotropy and originates from the spin–orbit interaction is dominant, and the magnetization is uniform. At larger
interaction, which couple the magnetic dipoles to the crystal orien- length scales, the dipolar interaction dominates and domains with
tation.128 As a result, the magnetization may have preferred orien- different magnetization orientations can be formed.
tations with respect to the crystal structure. Magnetization In addition to the previously described interactions, various
directions that correspond to minimum energy are called easy axes, other interactions exist, such as the Dzyaloshinskii–Moriya inter-
whereas magnetization orientations with maximum energy are action or the magnetoelastic interaction. Detailed discussions of
called hard axes. Different types of magnetocrystalline anisotropy the physics of these different interactions can be found in

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-3

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

TABLE I. Material properties of representative ferromagnetic materials, as well as propagation properties (group velocity, lifetime, and propagation distance) of surface spin
waves with a wavelength of λ = 1 μm in a 500 nm wide and 20 nm thick waveguide (external magnetic bias field μ0H = 100 mT).

Ms Gilbert damping Exchange length Group velocity Lifetime Propagation

Material (MA/m) α (×10−3) lex (nm) (μm/ns) (ns) distance (μm) Reference
Fe 1.7 60 3.4 5.8 0.08 0.5 89–93
Co 1.4 5 4.8 4.6 1.2 5.5 94–98
Ni 0.5 45 7.4 1.1 0.3 0.3 89 and 99–102
YIG (Y3Fe5O12, μm films) 0.14 0.05 17 42 600 25 000 70 and 103–107
YIG (Y3Fe5O12, nm films) 0.14 0.2 17 0.3 150 44 108–114
Permalloy (Ni80Fe20) 0.8 7 6.3 2.2 1.4 3.2 115–118
CoFeB 1.3 4 3.9 3.9 1.7 6.6 119–121
Co2(MnxFe1−x)Si 1.0 3 4.5 2.8 2.7 7.9 122–125

Refs. 126–128. Basic notions of the magnetoelastic interaction are with M0 being the static magnetization component, ω being the
also discussed in Sec. IV C. angular frequency, and k being the wavenumber. The effective
magnetic field is then given by
B. Magnetization dynamics and spin waves
The dynamics of the magnetization in the presence of one or ~ i(ωtþkr) ,
Heff (r, t) ¼ H0 þ h ¼ H0 þ he (11)
several of effective magnetic fields are described in the Landau—
Lifshitz—Gilbert (LLG) equation129,130
 
dM α dM
¼ γμ0 (M  Heff ) þ M , (9)
dt Ms dt

where γ is the absolute value of the gyromagnetic ratio, μ0 is the

vacuum permeability, α is the Gilbert damping constant, and Heff
is the effective magnetic field. This effective field is the sum of all
effective fields due to magnetic interactions and the external mag-
netic field. Hence, every magnetic interaction contributes to the
magnetization dynamics via the cross product of the magnetization
with its corresponding effective field.
In equilibrium, the magnetization is parallel to the effective
field. However, when the magnetization is not parallel to the effective
field, it precesses around this field, as described by the first term in
the LLG equation. The second term describes the attenuation of the
precession and represents the energy loss of the magnetic excitations
into the lattice (phonons) and the electronic system (electrons, eddy
currents). All these effects are subsumed in the phenomenological
Gilbert damping constant α. The combined effect of both terms in
the LLG equation results in a spiral motion of the magnetization
around the effective magnetic field toward the equilibrium state, as
graphically depicted in Fig. 2(a).
The LLG equation indicates that small oscillations of the effec-
tive magnetic field in time result in a precession of the magnetiza-
tion. The precession can be either uniform or nonuniform over the
magnetic volume. The case of uniform precession with a spatially
constant phase is called ferromagnetic resonance. For nonuniform
precession, the phase of the precession is position dependent and
FIG. 2. Schematic of the magnetization dynamics described in the LLG equa-
wave-like excitations of the magnetization exist, called spin waves
tion. (a) The trajectory of the magnetization is determined by the combination of
[see Fig. 2(b)]. Spin waves can thus be considered stable wave-like two torques [Eq. (9)]: (i) the precessional motion stems from M  Heff , whereas
solutions of the LLG equation. The ansatz for the magnetization (ii) the damping term M  dM dt ¼ M  (M  Heff ) drives the magnetization
dynamics of a spin wave in a bulk ferromagnet can be written as toward the direction of Heff . (b) Schematic representation of a spin wave in a
two-dimensional lattice of magnetic moments: top view of the first lattice row
~ i(ωtþkr) ,
M(r, t) ¼ M0 þ m ¼ M0 þ me (10) (top) and side view of the two-dimensional lattice (bottom).

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-4

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

with H0 and h being the static and dynamic components of the For short wavelengths (for large k), the exchange interaction is
effective magnetic field, respectively. As discussed above, this effec- dominant. In this limit, the dispersion relation shows a quadratic
tive magnetic field is the sum of the different effective fields due to behavior ωn,ex ¼ ωM λex k2tot , independent of the magnetization ori-
the relevant magnetic interactions. entation. By contrast, for long wavelengths (for small k), the
For weak excitations, i.e., kmk  kM0 k  Ms , the LLG equa- dipolar interaction becomes dominant. Then, the dispersion rela-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
tion can be linearized by neglecting terms quadratic in m. After a tion is given by ωn,dip ¼ ω0 (ω0 þ ωM F). The factor F strongly
temporal Fourier transform, we obtain depends on the magnetization orientation, indicating that the
dipolar interaction leads to anisotropic spin-wave properties. In the
iωα limit of infinite wavelengths, the frequency approaches the ferro-
iωm ¼ γμ0 (M0  h þ m  H0 ) þ (M0  m): (12)
Ms magnetic resonance frequency, which can be considered a spin
wave with k ¼ 0.
For specific values of k and ω, this linearized LLG equation has Figure 3 represents the spin-wave dispersion relations for dif-
nontrivial solutions, which represent stable collective magnetization ferent geometries in a 500 nm wide CoFeB waveguide (see Table I
excitations of the form me~ i(ω(k)tþkr) , i.e., spin waves. The function for material parameters) for an external magnetic field of
ω ¼ f (k) that relates the spin-wave oscillation frequency to the μ0 H ¼ 100 mT. In general, the dispersion relation of long-
wavevector is called the dispersion relation. The group velocity of a wavelength dipolar spin waves depends on the direction of the
(spin) wave is defined by the gradient of the dispersion relation, wavevector (the propagation direction) and the static magnetiza-
vg ¼ ∇k ω and represents the direction and the speed of the wave tion, as described in Eq. (13). It is, however, instructive to discuss
energy flow. By contrast, the phase speed, vp ¼ kω=kkk2 , describes three limiting cases of dipolar spin waves that are often called
the direction and speed of the wave phase front propagation. surface spin waves, forward volume waves, and backward
As discussed in detail in Sec. III, waveguide structures are of volume waves.
crucial importance for spin-wave devices and circuits. Therefore, in
the following, we briefly discuss the behavior of spin waves in wave-
guides with dimensions comparable or smaller to the wavelength. In
such waveguides, the behavior and specifically the dispersion relation
of spin waves are strongly affected by waveguide boundaries and
lateral confinement effects. Considering a waveguide with a thickness
d that is much smaller than its width w and with a rectangular cross
section, the spin-wave dispersion relation is given by131
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ωn ¼ (ω0 þ ωM λex k2tot )(ω0 þ ωM λex k2tot þ ωM F), (13)

with ω0 ¼ γμ0 H0 , ωM ¼ γμ0 M0 , and the abbreviations


F ¼ P þ sin f  1  P(1 þ cos2 (θk  θM ))
2


ωM P(1  P) sin2 (θk  θM )
þ (14)
ω0 þ ωM λex k2tot

and

1  edktot
P ¼1 : (15)
dktot

Here, k2tot ¼ k2 þ k2n with kn ¼ nπ=w is the quantized wavenumber,

n is the mode number, k is the wavenumber in the propagation
direction, θk ¼ arctan (kn =k), f is the angle between the magnetiza-
tion and the normal to the waveguide, and θM is the angle between
the magnetization and the longitudinal waveguide axis. Note that FIG. 3. Dispersion relation of backward volume spin waves (BVSWs), surface
this equation is only valid if the waveguide is sufficiently thin, i.e., spin waves (SSWs), and forward volume spin waves (FVSWs) in a 500 nm
kd  1, and the dynamic magnetization is uniform over the wave- wide and 30 nm thick CoFeB waveguide. For BVSW and SSW, the dispersion
guide thickness. We also remark that, depending on the magnetiza- relations of the first two laterally confined width modes (n1 and n2 ) are shown.
tion distribution and the demagnetization field at the waveguide The material parameters are listed in Table I, and the external magnetic field
was μ0 H ¼ 100 mT. The top panel depicts the mode profiles (top view) for con-
edges, it may be necessary to use an effective width instead of the
fined width modes with mode numbers as indicated.
physical width to accurately describe the dispersion relations.132,133

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-5

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

The first case corresponds to the geometry, in which both the

static magnetization and the propagation direction (the wavevector)
lie in the plane of the waveguide and are perpendicular to each
other, i.e., f ¼ π2 and θ M ¼ π2. Such spin waves are called surface
spin waves (SSWs) since they decay exponentially away from the
surface.134 Despite their name, the magnetization can still be con-
sidered uniform across the film for sufficiently thin films with
kd  1. The dispersion relations of the first two SSW width modes
(n1 and n2 ) in a 500 nm wide CoFeB waveguide are depicted in
Fig. 3 for an external field of μ0 H ¼ 100 mT. The curves indicate
that the group and phase velocities are parallel and point in the
same direction.
In the second geometry, the static magnetization is perpendic-
ular to both the propagation direction and the waveguide plane,
i.e., θ M ¼ π2 and f ¼ 0. The spin waves in this geometry have
dynamic magnetization components in the plane of the waveguide
and a group velocity parallel to the phase velocity. Such spin waves
are called forward volume spin waves (FVSWs) and their disper-
sion relation is also represented in Fig. 3.
In the third geometry, the static magnetization is parallel to
the propagation direction, both lying in the plane along the wave-
guide, i.e., f ¼ π2 and θM ¼ 0. In this case, dipolar spin waves have
a negative group velocity, which is antiparallel to the positive phase
velocity, i.e., group and phase velocities point in the opposite direc-
tions. Therefore, such waves are referred to as backward volume
spin waves (BVSWs). Their dispersion relation is also depicted in
Fig. 3 for the first two width modes (n1 and n2 ).
When the external driving magnetic fields are removed, the
spin-wave amplitude decreases exponentially with a characteristic
lifetime given by77

 1
@ωn
τ¼ αωn : (16)
@ω0

The spin-wave attenuation length represents the distance that a

spin wave can travel until its amplitude has been reduced by 1=e.
It is given by the product of the lifetime and the group velocity
δ ¼ τ  vg . As shown in Table I, spin-wave lifetimes are on the
order of ns in metallic ferromagnets, such as CoFeB or Ni, whereas
they can reach values close to the μs range in low-damping insula-
tors, such as Y3 Fe5 O12 (yttrium iron garnet, YIG). Since spin-wave
group velocities are typically a few μm/ns (km/s), attenuation
lengths are on the order of μm for metallic ferromagnets to mm
for YIG.
The spin-wave group velocity, lifetime, and attenuation length
(normalized to the wavelength) for the three cases of SSW, FVSW,
and BVSW are plotted in Fig. 4 as a function of the wavenumber
for a CoFeB waveguide and an external magnetic field of FIG. 4. Propagation characteristics of backward volume spin waves (BVSWs),
μ0 H ¼ 100 mT. Note that, when the static magnetization orienta- surface spin waves (SSWs), and forward volume spin waves (FVSWs) in a
tion is intermediate between the three limiting cases, the spin-wave 500 nm wide and 30 nm thick CoFeB waveguide, derived from the dispersion
properties also show intermediate characteristics. As a final remark, relations in Fig. 3. (a) Group velocity, (b) lifetime, and (c) attenuation length of
the BVSW and FWSV geometries both lead to volume waves, the spin waves normalized by the wavelength as a function of their wavevector.
which means that increasing the waveguide thickness may lead to For BVSW and SSW, data are shown for the first laterally confined width mode
(n1 ). In all cases, the material parameters were those of CoFeB (see Table I)
the formation of quantized spin-wave modes along the thickness of and the external magnetic field was μ0 H ¼ 100 mT.
the film at higher frequencies.

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-6

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

For SSW and BVSW, the group velocity reaches a maximum order,127,146 three-magnon confluence (i.e., the combination of
at small wavenumbers, which stems from the dipolar interaction. two magnons forming a single one), and four-magnon scattering
For BVSW, the group velocity becomes zero at a finite wavenumber (i.e., the inelastic scattering of two magnons) that is fundamental
(frequency) beyond the maximum due to the competition between for some spin-wave transistor concepts in Sec. V C.61
the dynamic dipolar and exchange fields. In the exchange regime, In all nonlinear scattering processes, the total energy and
the group velocities of SSW and BVSW become equal and further momentum are conserved. The magnon spectra in macroscopic
increase with the wavenumber. For logic applications, it is desirable structures always consist of a practically infinite number of modes
to use spin waves with large group velocities that ensure fast signal with different wavevector directions. Hence, an initial pair of
propagation and thus reduced logic gate delays. Moreover, large magnons, which participates e.g., in a four-magnon scattering
attenuation lengths reduce losses during spin wave propagation and process, can always find a pair of secondary magnons.127,140
are therefore also favorable for spin-wave devices. This will be However, in magnetic nanostructures,133,147 the magnon density of
further discussed below. states (scaling with the structure size) also decreases, which makes
Group velocities depend in general on the properties of the the “search” for secondary magnon pairs more complex.148,149
ferromagnetic medium, as shown in Table I. The group velocity Thus, the downscaling of magnonic nanostructures leads to a
decreases typically strongly with decreasing film (or waveguide) strong modification of nonlinear spin-wave physics, which offers
thickness. This can be compensated by using magnetic materials the possibility to control (in the simplest case, switch on or off )
with larger saturation magnetization Ms . The spin-wave lifetime in nonlinear processes by the selection of the operating frequency and
Eq. (16) depends on the Gilbert damping α. As the attenuation the external magnetic field.
length is given by the product of the group velocity and the lifetime, By contrast, processes (ii), which describe nonlinear frequency
the largest values are obtained for low-damping magnetic materials shifts of the spin-wave dispersion with increasing spin-wave ampli-
with large Ms . In practice, the two parameters α and Ms may need tude, are typically more pronounced at the nanoscale.133 These
to be traded off against each other, as indicated in Table I. phenomena do not require any specific adjustment of the operating
Additional material properties for ideal magnetic materials for logic point and can thus be useful for spin-wave devices. In particular,
computing applications are the possibility for co-integration along the nonlinear shift of the spin-wave dispersion relation allows
CMOS as well as a high Curie temperature to ensure temperature for the realization of nonlinear directional couplers, as discussed
insensitivity. This renders the complexity of the materials selection in Sec. V E.143
process and currently no clearly preferred materials has emerged yet.
Future material research in this field is thus of great interest to opti- III. FUNDAMENTALS OF SPIN-WAVE COMPUTING
mize conventional materials or to establish novel magnetic materials
In this section, we discuss the fundamental principles of dif-
for spin-wave applications.
ferent disruptive computation paradigms based on spin waves to
establish a framework for the architecture of a spin-wave-based
C. Nonlinear spin-wave physics computer. We start by introducing the basic components of a com-
puting system, their implementations using spin waves, and the
Section II B has discussed spin-wave physics using the linear-
limitations of an all-spin-wave system.
ized LLG equation (12). Such an approach is valid for small ampli-
tudes and describes noninteracting spin waves. However, the full
A. Basic computer architectures
LLG equation (9) is nonlinear, and thus nonlinear effects can arise
for large spin-wave amplitudes. Since nonlinear effects are central Despite many advances in computer architecture, the majority
for several spin-wave device concepts, this section provides a brief of today’s computing systems can still be considered to be concep-
overview over the topic. More details can be found in Refs. 72, 77, tually related to the Von Neumann architecture that was developed
127, 135, and 136. originally in the 1940s.150 Such a system consists of three essential
The theoretical model for nonlinear spin-wave interactions parts: (i) a central processing unit that processes the instructions of
was originally developed by Suhl, and thus nonlinear spin-wave the computer program and controls the data flow, (ii) a memory to
processes are often referred to as Suhl instabilities of the first and store data and instructions, and (iii) a data bus as interconnection
second order.127,137,138 Later, a generalized quantum-mechanical that links the various parts within the processor and the memory
description of nonlinear magnons (quantized spin waves), termed and provides communication with the outside world. A schematic
S-theory, was developed by Zakharov et al.139,140 Today, these of such a system is shown in Fig. 5. Hence, to design a computer
models are primarily used to describe a variety of different nonlin- system that operates entirely with spin waves, spin-wave processors,
ear and parametric spin-wave phenomena.70,141–145 spin-wave memory, as well as spin-wave interconnects need to be
In general, the diverse nonlinear effects can be categorized developed. Moreover, interfaces between the spin-wave processor
into two groups: (i) multimagnon scattering127,140 and (ii) the and the outside periphery—presumably charge-based—are required,
reduction of saturation magnetization at large precession including a power supply.
angles.141,143 However, (ii) can also be described by four-magnon The performance of a computing system is generally limited
scattering, so the separation into groups is not strict. Multimagnon by the weakest component. Its computing throughput is restricted
scattering effects (i) primarily include three-magnon splitting by the slowest part, and the power consumption is determined by
(i.e., the decay of a single magnon into two), which can be used for the most power-hungry subsystem. As detailed below, there is
the amplification of spin waves as a parametric process of the first currently no comprehensive concept for a full spin-wave computer.

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-7

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

FIG. 5. Schematic of a Von Neumann computer consisting of a central process-

ing unit and a memory, interconnected by a data bus.

In the following, we discuss requirements, basic approaches, and

potential spin-wave-based implementations of the main compo-
nents of a computer and finally suggest how a spin-wave-based
computing system may resemble.
Recently, there has been growing interest in alternative com-
puting paradigms beyond Von Neumann architectures, especially
in the field of machine learning.29–31 Whereas the implementation
of such architectures by spin waves is an intriguing prospect,
research on this topic is still in its infancy.151–157 A detailed discus-
sion of such systems is beyond the scope of the Tutorial.
Nonetheless, it is clear that many of the arguments below remain
relevant. Further information can be found in Sec. VIII A.

B. Information encoding
Before discussing spin-wave computing concepts, we need
to define how information can be encoded in a spin wave. Waves
are characterized by amplitude (intensity), phase, wavelength, and
frequency, which can all be used for information encoding. It is
clear that the encoding scheme determines the interactions that
can be employed for information processing and computation.
Presently, device proposals typically rely on information encoded FIG. 6. Different schemes to encode information in (spin) waves: (a) binary
amplitude encoding, (b) binary phase encoding, and (c) quaternary (nonbinary)
in spin-wave amplitude and/or phase (see Fig. 6). Moreover, mixed amplitude and phase encoding.
the usage of different frequency channels has been proposed to
enable parallel data processing based on frequency-division
multiplexing.80,158
In amplitude-based information encoding, two main schemes certain threshold and logic 0 otherwise (or vice versa). Multiple
can be pursued: (i) amplitude level encoding and (ii) amplitude thresholds can be defined to represent nonbinary information and
threshold encoding. In amplitude level encoding, the presence of a enable multivalued logic and computing. For example, if fX, Y g
spin wave in a waveguide is referred to as logic 1 and no spin wave with X , Y are defined as a set of thresholds, a spin-wave ampli-
as logic 0 [Fig. 6(a)]. By contrast, in amplitude threshold encoding, tude greater than Y can represent a 1, an amplitude between X and
logic 1 is represented by a spin wave with an amplitude above a Y a 0, and an amplitude below X a 1.

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-8

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

Alternatively, information can be encoded in the (relative) wavelength can also be (efficiently) utilized to evaluate logic func-
spin-wave phase such that e.g., a relative phase of 0 (i.e., a spin tions is still an open research question with the potential for addi-
wave in phase with a reference) refers to a logic 1, while a relative tional avenues toward novel computation paradigms.
phase of π refers to logic 0 [Fig. 6(b)]. Furthermore, additional For in-phase waves with equal frequency, constructive interfer-
phases can be utilized for multivalued  logic, e.g., f1, 0, 1g can ence leads to a peak-to-peak amplitude of the generated wave that
be represented by the set of phases 0, π2 , π . Such ternary com- is equal to the sum of the peak-to-peak amplitudes of the input
puting schemes can have advantages over binary ones and the waves. By contrast, destructive interference leads to a subtraction of
implementation of ternary logic circuits using (spin) waves may be the peak-to-peak amplitudes of input waves when their phase dif-
an interesting future research topic, e.g., for computer arithmetics ference is π. For spin waves, the corresponding magnetization
or neural networks. dynamics are depicted in Fig. 7. In narrow waveguides, the spin-
Combinations of amplitude and phase encoding schemes are wave modes [see Fig. 7(a) for the mode pattern of the first width
also possible and open further pathways toward effective nonbinary mode] may deviate from plane waves due to lateral confinement
data processing [Fig. 6(c)]. For example, the dataset f0, 1, 2, 3g can and the effect of the demagnetizing field, as discussed in Sec. II B.
be encoded using two amplitude levels fA, 2Ag and two phases Nonetheless, micromagnetic simulations, which rely on solving
{0, π} by 0 :¼ {A, 0}, 1 :¼ {A, π}, 2 :¼ {2A, 0}, and 3 :¼ {2A, π}. the LLG equation numerically,161,162 for a CoFeB waveguide
Such schemes can be easily generalized to larger sets of nonbinary [Figs. 7(b) and 7(c)] indicate that confined spin waves still show
information. the expected interference. By placing two spin-wave sources on the
The different encoding schemes have specific advantages and same waveguide, destructive [Fig. 7(b)] or constructive [Fig. 7(c)]
drawbacks when implemented in spin waves. Spin waves have interference is obtained for a relative phase of π or 0, respectively.
typical propagation distances of μm to mm, depending on the host The observation of incomplete destructive interference in Fig. 7(b)
material. For amplitude coding, the maximum size of a spin-wave can be linked to spin-wave attenuation, which leads to slightly dif-
circuit needs to be much smaller than the spin-wave attenuation ferent amplitudes of the two waves at both sides of the spin-wave
length, since the logic level may otherwise change during propaga- sources.
tion. By contrast, the phase of a wave is not affected by attenuation. Wave interference can be exploited to compute basic Boolean
While computing schemes may still require well-defined ampli- operations using the different encoding schemes. For example,
tudes, as further outlined below, the logic value encoded in the spin using amplitude level encoding, it is easy to see that the construc-
wave is nonetheless stable during propagation. Moreover, the phase tive interference of two waves generates output of an OR opera-
coherence times of spin waves are long and phase noise can be kept tion, whereas their destructive interference (with a phase shift of
under control even for nanofabricated waveguides with, e.g., con- π between the waves) produces the output of an XOR operation.
siderable linewidth roughness,159 rendering phase encoding rather Many proposals and experimental studies have focused on phase
stable. However, the largest differences between encoding schemes encoding and the calculation of the majority function,
lie in the different interactions and processes required for computa- MAJ.59,64,163–169 This stems from the fact that the phase of the
tion, which is the topic of Sec. III C. output wave, ensuing from the interference of three input waves,
We finally note that spin waves are noninteracting in the is simply the majority of the phases of the input waves when logic
small signal approximation, i.e., for small amplitudes. Therefore, 1 is encoded in phase 0 and logic 0 in phase π (or vice versa).
parallel data processing is possible using, e.g., frequency-division or Together with recent advances in MAJ-based circuit design,67,68,170,171
wavelength-division multiplexing. An information encoding this has led to a strong interest in spintronics42,50,52,55,64 and, in partic-
scheme can then be defined at each frequency or wavelength and ular, spin-wave majority gates.59,64,172,173 As an example, the carry out
computation can occur in parallel in the same processor. bit in a full adder (a fundamental building block in processor design)
Multiplexing in spin-wave systems is discussed further in Sec. VII. is directly computed by a three-input majority function [cf. Eq. (17)].
In addition, many error detection and correction schemes rely on
n-input majority logic.174,175
C. How to compute with (spin) waves? For novel computation paradigms, including (spin) wave com-
When logic levels are encoded in spin-wave amplitude or puting, a main requirement is the possibility to implement any
phase, performing a logic operation requires the combination of arbitrary logic function that can be defined within its basic formal-
different input waves and the generation of an output wave with an ism by means of a universal gate set. For example, within Boolean
amplitude or phase corresponding to the desired logic output state. algebra, any logic function can be expressed as a sum of products
In principle, the superposition of waves can lead to the addition of or as a product of sums. Using double complements and De
either their intensity or their amplitude, depending on whether the Morgan’s laws, it can be demonstrated that any logic function can
waves are incoherent or coherent.160 Since practical spin-wave be implemented by either NAND or NOR gates only. Therefore,
signals typically have a large degree of phase coherence, further dis- NAND or NOR constitutes each a universal gate with efficient
cussion can be limited to coherent superposition. In the absence of CMOS implementations. As mentioned above, (spin) wave interfer-
nonlinear effects, the interaction of coherent waves is described by ence provides a natural support to implement majority gates, MAJ,
interference, i.e., the addition of their respective amplitudes at each which form a universal gate set in combination with inverters,
point in space and time. We also limit the discussion to the super- INV. In phase encoding, an inverter can be realized by a passive
position of waves with identical frequency and wavelength. delay line of length (n  12 )  λ (with λ being the spin-wave wave-
Whether the interference of waves with a different frequency or length and n ¼ 1, 2, 3, . . . an integer) that leads to a phase shift of

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-9

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

FIG. 7. Out-of-plane component of magnetization (Mz ) in a 50 nm wide and 5 nm thick CoFeB waveguide obtained by micromagnetic simulations: Snapshot images of the
spin waves emitted by a single port (a) and two in-phase (b) or anti-phase (c) ports at a frequency of 15 GHz. The corresponding amplitudes along the magnetic wave-
guide are shown in panels (d) and (e), respectively. The material parameters considered in simulations were taken from Table I. The magnetic waveguide was initially mag-
netized longitudinally, whereas the simulations of spin-wave propagation were carried out in zero magnetic bias field. Spin waves were excited by a uniform out-of-plane
magnetic field at positions P1 and P2 in the waveguide center.

π during propagation. In amplitude encoding, inverters are more waves/phonons.180,181 A discussion of the advantages and disadvan-
complex and typically require active components. In this case, an tages of different physical implementations of wave computing is
inverter can be realized by interference with a reference wave with a beyond the scope of this Tutorial, but it is clear that many of the dis-
phase of π. As an example, XOR, XNOR, and a full adder (sum Σ cussions concerning devices, circuits, and hybrid systems are general
and carry out Cout ) can then be implemented with majority gates and remain valid for other wave-based computing approaches.
and inverters as follows:

A  B ¼ MAJðMAJðA, B,  B, 0Þ, 1Þ,

 0Þ, MAJðA, D. Spin-wave interconnects
 B,
A  B ¼ MAJðMAJðA,  0Þ, MAJðA, B, 0Þ, 1Þ, In Sec. III C, the basic principles of spin-wave interference
have been discussed and it has been shown that they can be used
 in Þ, Cin ,
Σ ¼ MAJ MAJðA, B, Cin Þ, MAJðA, B, C for logic operations. However, in a computing system, data need to
Cout ¼ MAJðA, B, Cin Þ: (17) be transmitted to the inputs of the logic circuit, exchanged between
gates, and finally output data need to be transmitted to, e.g., a
It should be mentioned that wave-based computing is not memory. This is the task of the interconnect, which may also trans-
limited to the usage of spin waves. Similar concepts have been mit clock signals as well as power. In conventional digital inte-
proposed for surface plasmon polaritons176–179 or acoustic grated circuits, the logic states 0 and 1 are encoded in voltages,

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-10

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

which allows for data transmission by metal wires. While intercon-

nect performance is today often limiting the overall performance of
integrated circuits, solutions are mature and well understood from
the point of view of their capabilities and associated overhead.
A natural approach to connect spin-wave logic gates is by
means of waveguides, in which spin waves propagate from, e.g., a
gate output to an input of a subsequent gate. Besides cascading
issues for specific implementations discussed in more detail in
Sec. VI, the rather slow and lossy spin-wave propagation leads to
fundamental limitations for spin-wave interconnects.182–184 Since
the spin-wave group velocity is much lower than that of electro-
magnetic waves in (nonmagnetic) metallic wires, interconnection
by spin waves propagating in waveguides adds a considerable delay
overhead, which depends on waveguide length and material. Some
representative numbers for the spin-wave group velocity are listed
in Table I. Typical delays are about 1 ns/μm (μs/mm), which means
that spin waves propagating in waveguides cannot be efficiently uti-
lized for long-range data transmission. Even for short range data
communication, the delay introduced by spin-wave propagation
may not be negligible. As an example, for a spin-wave circuit with
a waveguide length of a few μm, the propagation delay may already
FIG. 8. Schematic of a clocked spin-wave interconnect. Reproduced with per-
exceed the duration of a typical clock cycle of a high performance mission from S. Dutta, S.-C. Chang, N. Kani, D. E. Nikonov, S. Manipatruni,
CMOS logic processor of about 300 ps (
3 GHz clock frequency). I. A. Young, and A. Naeemi, Sci. Rep. 5, 9861 (2015). Copyright 2015 Nature.
It is worth noting that the overall delay is determined by the
longest propagation path in the circuit. Hence, propagation delays
may limit the computing throughput of a spin-wave circuit.
Additional boundaries for the throughput of spin-wave circuits and phase-sensitive switching of nanomagnets with perpendicular mag-
systems are discussed in Sec. VI. netic anisotropy in the repeater stages. It therefore also offers some
Moreover, the spin-wave amplitude decays during propagation memory functionality. A 2D-mesh configuration of such structures
due to intrinsic magnetic damping. Such propagation losses remain has also been proposed.186,187
limited when spin-wave circuits are much smaller than the attenua- An alternative approach is the use of conventional charge-based
tion length, which strongly depends on the waveguide material (see memories after signal conversion in the hybrid spin-wave–CMOS
Table I for indicative numbers). This can impose severe limits on systems discussed in Sec. III F. An introduction to charge-based
the size (and therefore the complexity) of spin-wave circuits. Losses memory devices is beyond the scope of this Tutorial and can be
can in principle also be compensated for by spin-wave amplifiers found, e.g., in Refs. 2, 188, and 189.
or repeaters. As an example, a clocked interconnect concept based
on spin-wave repeaters has been reported in Ref. 185 (see Fig. 8).
While such approaches can mitigate limitations of signal propaga- F. Hybrid spin-wave–CMOS computing systems
tion by spin waves, they add a significant overhead to the circuit Earlier, we have argued that spin-wave propagation in mag-
and need to be carefully considered when the energy consumption netic waveguides may add considerable delay and is therefore not
and delay of a spin-wave computing system is assessed. Spin-wave competitive over distances of more than a few 100 nm to 1 μm. To
repeaters and amplifiers are discussed in more detail in Sec. V F. address this issue, metallic or optic interconnects can be used for
long range data transmission after spin-wave signals have been con-
verted to electric or optical signals. Voltages and light travel very
E. Spin-wave memory fast through metal wires and optical fibers, respectively, with propa-
To date, rather little work has been devoted to specific spin- gation velocities given by the speed of light in the host materials.
wave memory elements that are required for computing systems Such solutions lead naturally to hybrid system concepts, in which
based on spin waves only. Spin waves are volatile dynamic excita- spin-wave circuits coexist with conventional CMOS or mixed-signal
tions, which decay at time scales of ns to μs (see Table I). There are integrated circuits, including memory. Such solutions rely on
two different basic approaches to memories for spin waves. The (frequent) forth-and-back conversion between spin-wave and
natural spintronic memory element is a nanomagnet, in which the charge domains using transducers, which may themselves add sub-
information is encoded in the direction of its magnetization. In such stantial delay and energy consumption overhead. To minimize the
a memory element, an incoming spin wave deterministically sets overhead, the number of necessary transducers should remain
(switches) the orientation of the magnetization of the nanomagnet. limited. The acceptable conversion granularity depends on the rela-
When phase encoding is used, the interaction between the spin wave tion between delay and energy consumption of spin-wave circuits,
and the nanomagnet needs to be phase dependent. The clocked transducers, and CMOS/mixed-signal circuits. In practice, it is of
interconnect concept185 depicted in Fig. 8 employs the deterministic course technology dependent.

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-11

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

Today, design guidelines for such hybrid circuits are only with mðxÞeiωt being the distribution of the dynamic magnetiza-
emerging. Their development and the benchmarking of the tion. Note that finite spin-wave lifetimes due to magnetic damping
ensuing hybrid circuits constitute a crucial step toward real-world broaden the δ-function. For inductive antennas transversal to the
applications for spin-wave computing. Since hybrid systems require direction of the waveguide, the magnetic field points essentially in
efficient and scalable transducers, the approaches to generate and the x-direction along the waveguide. For thin films, the magnetiza-
detect coherent spin waves are discussed in Sec. IV. Such trans- tion is uniform over the film thickness and the dynamic magnetiza-
ducers form critical elements of the spin-wave devices and circuits ~
tion of a plane wave can be written as mðxÞ ¼ m(y)e ikx ~
, with m(y)
that are reviewed in Sec. V. describing the transverse mode pattern. Equation (18) then
becomes115
IV. SPIN-WAVE TRANSDUCERS ð ð
As argued earlier, spin-wave computing systems require Γn / hx (x)eikx dx ~ x (y)dy  δ ðω  ωh Þ:
hx (y)m (19)
transducers to convert spin-wave-encoded signals to/from voltage
signals. The scalability and the energy efficiency of the transducers
can be expected to be crucial for the overall performance of a The first integral indicates that the wavelength dependence of the
hybrid system. This section introduces different concepts of spin- spin-wave excitation efficiency is determined by the Fourier spec-
wave transducers. As discussed in Sec. II, spin waves are a response trum of the driving Oersted field along the waveguide. The second
of a magnetic material to oscillatory external (effective) magnetic term leads to a dependence on the symmetry of the spin-wave
fields. In the linear regime, i.e., for weak excitation, excited spin mode. For symmetric dynamic magnetic field distributions (as in
waves have the same frequency as the applied oscillatory field, a the case of an inductive antenna), only spin-wave modes with sym-
well-defined phase, which depends on the specific interaction, and metric transverse mode patterns can be excited and the excitation
an amplitude proportional to the magnitude of the excitation. In efficiency is zero for antisymmetric modes. For an inductive
principle, any oscillatory effective field can launch spin waves in a antenna with width w, this leads to
waveguide. From a practical point of view, the need to generate
oscillatory effective magnetic fields at GHz frequencies has led to w kw
n sinc 2  δ ðω  ωh Þ for odd n,
Γn / (20)
several preferred approaches. It should be mentioned that the scal- 0 for even n:
ability and the energy efficiency of such transducers at the nano-
scale have not been definitively assessed and are currently actively Here, the dependence on the sinc function stems from the
researched. As argued in Sec. IX, the demonstration of a nanoscale Fourier transform of the uniform magnetic field underneath the
spin-wave transducer with high energy efficiency is one of the key antenna, whereas the explicit dependence on the mode number n
prerequisites for the ultimate goal of hybrid spin-wave–CMOS is caused by the transverse integral over the mode pattern. This
computing systems. discussion shows that the shape and the dimensions of inductive
antennas have strong impact on the spin-wave excitation band-
A. External magnetic fields: Inductive antennas width. Reducing the dimensions (width, gap) of an antenna
increases its bandwidth and the peak magnetic field strength
The “reference” method to excite spin waves is by means of underneath.
external magnetic fields generated by an AC current in a microwave While inductive antennas can be rather efficient at “macro-
antenna. The AC current generates an alternating Oersted field via scopic” scales 10 μm, scaling their dimensions into the μm and
Ampère’s law, which in turn exerts a torque on the magnetization sub-μm range strongly reduces the antenna quality factor, i.e., the
in an adjacent ferromagnetic medium. At excitation frequencies ratio between inductance and resistance, Q ¼ L=R, and the spin-
above the ferromagnetic resonance, the Oersted field can then wave excitation efficiency. In general, since the Oersted field is pro-
excite spin waves in the medium, as outlined in Sec. II and portional to the current via Ampère’s law, antennas do not scale
described in the LLG equation (9). favorably, with strongly increasing current densities (and thus
Different antenna designs have been used in spin-wave experi- degraded reliability) at smaller dimensions. More details on the
ments, such as microstrip antennas, coplanar waveguide antennas, relation between antenna design, spin-wave excitation efficiency,
or loop antennas. An overview can be found in Ref. 190. The spe- and bandwidth can be found in Ref. 190. It has also been shown
cific antenna design has strong repercussions on the spin-wave that a magnetic near field resonator in the vicinity of the antenna
spectrum that can be excited. It is intuitive that an oscillating mag- can enhance the spin-wave excitation efficiency.191,192
netic field that is uniform over a distance L along the waveguide Inductive antennas can also detect spin waves. The dynamic
cannot efficiently excite spin waves with wavelengths λ  L. More dipolar field generated by the spin waves induces a current in an
quantitatively, the excitation efficiency Γn of a spin wave propagat- adjacent antenna via Faraday’s law. Thus, inductive antennas can
ing along the x-direction with mode number n, wavenumber k, and be used as both input and output ports in all-electrical spin-wave
angular frequency ω by a dynamic magnetic field distribution transmission experiments.70,193–199 A schematic of such an experi-
h(x)eiωh t is proportional to the overlap integral over the magnetic ment is shown in Fig. 9. A first inductive antenna launches spin
volume V, waves in a ferromagnetic waveguide, which are subsequently
detected by a second antenna. The microwave power transmitted
Γn / jV hðxÞ  mðxÞdxj  δ ðω  ωh Þ, (18) by the spin waves can be measured with phase sensitivity using a

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-12

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

Mfix is given by200,211

 
dM jgj μB 1 J
¼ P½M  ðM  Mfix Þ , (21)
dt STT 2 Ms d e

where J is the current density and d is the thickness of the free

layer. P represents the current polarization, μB is the Bohr magne-
ton, g is the Landé factor, and e denotes the elementary charge.
The effect of a spin-polarized current on the magnetization
dynamics can be calculated by introducing an STT term in the
LLG equation (9). Using the notations

M Mfix H
ζ¼ , ζ fix ¼ , η¼ , τ ¼ γ 0 Ms t (22)
Ms Ms,fix Ms

and

 1 1J
h
χ¼ P, (23)
2 μ0 Ms2 d e

the LLG equation including the STT term can be written in a

dimensionless form as211
 
FIG. 9. (a) Sketch of a typical experimental setup for spin-wave transmission dζ dζ
¼  (ζ  η)  χ ½ζ  ðζ  ζ fix Þ þ α ζ  : (24)
based on a waveguide and two inductive antennas. Spin waves are excited by dτ |fflfflffl{zfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflffl} dτ
the Oersted field created by a microwave current in one of the antennas and precession spin transfer torque |fflfflfflfflfflfflffl
ffl {zfflfflfflfflfflfflffl
ffl }
detected inductively by the second. The power transmitted by the spin waves is damping
measured using a vector network analyzer and extracted from S-parameters.
The arrows inside the waveguide symbolize spin precession during propagation. Depending on the direction of current flow, the torque
(b) Scanning electron micrograph of a 500 nm wide CoFeB waveguide and two exerted by the spin-polarized current can enhance or compensate
125 nm wide inductive antennas. for the intrinsic damping. When the damping is exactly compen-
sated for, the STT enables a steady precession of the magnetization.
Even larger polarized current densities lead to a negative damping
vector network analyzer. Both the fraction of transmitted (S21 , S12 ) torque and the magnetization precession is strongly amplified. The
and reflected (S11 , S22 ) microwave power can be used to analyze the critical current required to excite the magnetization in the free layer
measurements. (from an initially parallel orientation of both magnetic layers) is
given by211,212
 
2e α Ms
B. Spin-transfer and spin–orbit torques Icrit ¼ Vμ Ms H þ Hk þ , (25)
h P 0 2
In Sec. IV A, it was discussed that spin waves can be excited
by the oscillatory Oersted field created by a microwave AC current where V represents the volume of the magnetic free layer, and H
in an inductive antenna. In addition, DC currents can also generate and Hk denote the external and the anisotropy magnetic fields,
spin waves or switch nanomagnets as long as they are spin respectively.
polarized.200–210 When an electric current passes through a uni- To limit the critical currents necessary for stable magnetiza-
formly magnetized layer, the electron spins align themselves with tion precession, the volume of the magnetic layers V is typically
the magnetization direction, generating a spin-polarized current reduced by patterning pillars with sub-μm diameters. These devices
[Fig. 10(a)].200,201 When such a polarized current flows through a have been termed spin-torque nano-oscillators (STNOs) and can
second magnetic layer, the spins reorient again if the direction of also emit spin waves if the free layer is coupled to a waveguide.
the magnetization is not aligned with the spin polarization. This It has been demonstrated that spin waves emitted by STNO can
leads to the transfer of angular momentum to the magnetization of travel for several μm and that their propagation direction can be con-
the second layer, which can change its orientation if the layer is trolled by a magnetic bias field [see Figs. 10(d) and 10(e)].208,209
thin enough (a few nm). The transfer mechanism of angular Another mechanism to generate spin currents is based on the
momentum by spin-polarized currents to the magnetization is spin Hall effect (SHE). This effect originates from the spin-
known as spin-transfer torque (STT). The spin-transfer torque that dependent electron scattering in a charge current flowing through a
acts on the magnetization M of a “free layer” due to a spin- nonmagnetic metal or a semiconductor with (large) spin–orbit
polarized current from a “fixed” reference layer with magnetization interaction.213,214 The resulting spin current is perpendicular to the

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-13

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

FIG. 10. Schematic illustrations of (a) spin-transfer torque (STT) and (b) spin–orbit torque (SOT) processes. (c) Magnetization dynamics in an effective field including pre-
cession, damping, and both STT and SOT. (d) Device layout used for the excitation of spin waves by STT. Reproduced with permission from M. Madami, S. Bonetti,
G. Consolo, S. Tacchi, G. Carlotti, G. Gubbiotti, F. B. Mancoff, M. A. Yar, and J. Åkerman, Nat. Nanotechnol. 6, 635 (2011). Copyright 2020 Springer Nature. (e)
Attenuation of the excited spin waves during propagation. Reproduced with permission from M. Madami, S. Bonetti, G. Consolo, S. Tacchi, G. Carlotti, G. Gubbiotti, F. B.
Mancoff, M. A. Yar, and J. Åkerman, Nat. Nanotechnol. 6, 635 (2011). Copyright 2020 Springer Nature. (f ) Scanning electron micrograph and (g) schematic layout of a
device based on an SOT emitter and an inductive antenna detector. (h) and (i) Intensity of spin waves generated by a SOT antenna [(h), magnified 20] and an inductive
antenna (i). Reproduced with permission from G. Talmelli, F. Ciubotaru, K. Garello, X. Sun, M. Heyns, I. P. Radu, C. Adelmann, and T. Devolder, Phys. Rev. Appl. 10,
044060 (2018). Copyright 2018 American Physical Society. The plots show the magnetic-field derivative of the forward-transmission S-parameter, dS21 =dH
(emitter-to-detector distance 4 μm, applied magnetic field μ0 H ¼ 80 mT). The insets show field–frequency signal maps corresponding to spin waves emitted by the two
types of antennas (magnetic fields μ0 H ¼ 52–145 mT, frequencies 8–15 GHz).

charge current and can therefore be transferred to an adjacent fer- (SOT), whereas the factors εk and ε? account for the efficiency of
romagnetic material even if the charge current is only flowing in the spin-transfer process. h is the reduced Planck constant, p repre-
the nonmagnetic metal. The spin current exerts a torque on the sents the spin-polarization orientation of the injected spin current,
magnetization of the ferromagnet, as illustrated in Fig. 10(b). In tFM is the thickness of the ferromagnetic layer, and Js represents the
addition, the spin–orbit interaction of the conduction electrons in a spin current density.
two-dimensional system can also generate an effective magnetic The first experimental observation of SOT effects on
field—the so-called Rashba effect.215,216 The torque on the magne- spin waves was a damping reduction due to a spin current
tization due to spin–orbit effects can be expressed by217,218 generated via the SHE in permalloy/Pt bilayers.219 The excitation
  of spin waves by SOT has been demonstrated in YIG/Pt
dζ heterostructures,220–222 whereas device nanopatterning allowed
¼ γβ k ½ζ  ðp  ζ Þ þ γβ ? ðp  ζ Þ, (26) for the demonstration of spin Hall nano-oscillators
dt SOT
(SHNOs),219,223–225 their synchronization to external microwave
with signals,226 and the mutual synchronization of SHNOs by pure
spin currents.227 Recently, it has also been shown that SOT anten-
h Js h Js nas can excite spin waves when driven by microwave currents. It
β k ¼ εk , β ? ¼ ε? : (27) was estimated that the generated antidamping spin-Hall and
2e tFM 2e Ms tFM
Oersted fields contributed approximately equally to the total
Here, β k and β ? are the coefficients for the antidamping (in-plane) effective field, providing an improvement over conventional
and field-like (out-of-plane) components of the spin–orbit torque inductive antennas.228

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-14

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

C. Magnetoelectric transducers resonance, is well understood,255–257 few studies have addressed the
Magnetoelectric transducers are a more recent addition to coupling to acoustic waves at GHz frequencies (hypersound).
the approaches to excite and detect spin waves. They are based When the transducer launches propagating acoustic waves, spin-
on magnetoelectric compounds, which consist of piezoelectric wave excitation is generally nonlocal and occurs in the waveguide
after acoustic wave propagation.258 For mechanical resonators with
and magnetostrictive bi- or multilayers. In such transducers,
high quality factors, the emission of elastic waves is however weak,
effective magnetoelastic fields are generated in the magnetostric-
tive ferromagnetic layer(s) via application of stress/strain due to and thus spin waves are generated locally at the transducer. As for
the inverse magnetostriction (Villari) effect. The stress/strain antennas, the spin-wave excitation efficiency is proportional to the
itself can be generated by an electric field applied across the pie- overlap integral of the spatial distribution of the dynamic excitation
field due to the standing waves in the transducer and the dynamic
zoelectric layer(s). Magnetoelectric transducers thus couple vol-
magnetization of the spin-wave mode, as described in Eq. (18). For
tages with magnetic fields indirectly via mechanical degrees of
freedom. Reviews of the magnetoelectric effect can be found in Hmel ¼ hmel ðx, yÞeiωmel t , the excitation efficiency of a spin-wave
mode with dynamic magnetization m(x, ~ y)eiωsw t in a thin waveguide
Refs. 229–237.
In a magnetostrictive material, the application of a strain with can be written as
tensor ε generates an effective magnetoelastic field. The magnetoe-
ðð
lastic field is given by
Γ/ ~ y) dx dy  δ ðωmel  ωsw Þ:
h(x, y)  m(x, (30)
1 dEmel (M)
Hmel ¼  : (28)
μ0 dM Here, the integral is carried out over the waveguide volume that is
mechanically excited by the transducer. In contrast to the Oersted
as outlined above in Eq. (5). In general, the magnetoelastic field field generated by an inductive antenna, the magnetoelastic field is
depends on the crystal symmetry of the magnetic material. An not necessarily uniform along the transverse y direction, so modes
explicit formula can be derived for cubic crystal symmetry. In this with both odds and even mode numbers can in principle be
case, the magnetoelastic field is given by238 excited. For small spin-wave amplitudes, the magnetization ζ in
Eq. (29) is equal to the static magnetization and does not change
0  1
with time. By contrast, large spin-wave amplitudes can lead to
B1 εxx ζ x þ B2 εxy ζ y þ εxz ζ z
2 B C considerable nonlinearities when ζ precesses in time. For linear
B B ε ζ þB ε ζ þε ζ C
Hmel ¼ B 1 yy y 2 xy x yz z C: (29) elastic systems, the integral in Eq. (30) can be evaluated for each
μ0 Ms @  A strain tensor component individually. As an example, the excita-
B1 εzz ζ z þ B2 εxz ζ x þ εyz ζ y
tion efficiency of spin wave propagating along the x-direction
with mode number n, wavenumber k, and angular frequency ωsw
Here, B1 and B2 are the magnetoelastic coupling constants of the by an oscillating shear strain εxy (x, y)eiωmel t in a waveguide uni-
waveguide material and ζ ¼ M=Ms . Equation (29) also describes formly magnetized along the transverse y-direction is
the case of an isotropic material. In this case, the magnetoelastic
coupling constants are equal, i.e., B1 ¼ B2 . This indicates that the ð
magnetoelastic field depends on both the magnetization orientation Γεxy / B2 εxy (x)eikx dx  δ ðωmel  ωsw Þ: (31)
and the strain tensor geometry. For uniform magnetization,
Eq. (29) indicates that normal strain parallel or perpendicular to
the magnetization does not exert a torque T ¼ Hmel  M on the Thus, the excitation efficiency is in this case given by the Fourier
magnetization since the magnetoelastic field is either parallel to the transform of the mechanical (strain) mode of the transducer in
magnetization or zero. By contrast, torques on the magnetization the waveguide direction and can thus feature resonances that are
are exerted by oblique normal strain (with respect to the magneti- linked to the mechanical response of the transducers. However,
zation) or shear strain.239 the mechanical behavior of realistic devices is expected to be
So far, experimental studies have focused mainly on spin-wave rather complex and the understanding is currently only
excitation by propagating surface acoustic waves.240–252 However, emerging.239,258–260
the interdigitated transducers used to excite surface acoustic waves In magnetoelectric compounds based on linear piezoelectric
are difficult to scale to small dimensions and resonance frequencies materials, the strain is proportional to an applied voltage—and
are typically well below ferromagnetic resonance even in low-Ms therefore also the magnetoelastic field. Schematics of magnetoelec-
ferrites. In all cases, the excitation of spin waves requires strain tric transducers are depicted in Fig. 11. Because typical charging
fields oscillating at GHz frequencies. Thus, the strain tensor is not energies of scaled magnetoelectric capacitors can be orders of mag-
static but determined by the dynamic oscillating strain field gener- nitude lower than Ohmic losses in inductive antennas or STT
ated by the transducer, which is typically characterized by a series devices, magnetoelectric transducers are potential candidates to
of electromechanical resonances (standing waves) in the transducer enable low-power and high-efficiency transduction. Moreover,
itself and propagating elastic (acoustic) waves in the magnetic since the mechanism depends on electric fields, it shows favorable
waveguide.59,253,254 Whereas the magnetoelastic coupling at low- scaling properties with larger magnetoelectric voltage coupling for
frequency electromechanical resonances, below ferromagnetic thinner piezoelectric films.

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-15

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

FIG. 11. Schematics of magnetoelec-

tric transducers consisting of a piezo-
electric element and a magnetic
spin-wave waveguide formed by (a) a
ferromagnetic and magnetostrictive
bilayer system and (b) by a simultane-
ously ferromagnetic and magnetostric-
tive single layer. (c) Schematic of a
spin-wave transmission experiment
including on a magnetic waveguide for
spin-wave propagation and two magne-
toelectric transducers. Similar to the
case of two antennas in Fig. 9, the
power transmitted by spin waves can
be measured by a vector network ana-
lyzer connected to the ground (G) and
signal (S) microwave electrodes of the
devices.

Beyond the generation of spin waves by the magnetoelectric D. Voltage control of magnetic anisotropy (VCMA)
effect, also an inverse magnetoelectric effect exists, which can be
A different type of magnetoelectric effects relies on the voltage
used to detect spin waves. A spin wave in a magnetostrictive mate- control of magnetic anisotropy (VCMA).265–268 VCMA describes
rial creates a dynamic displacement field and thus an elastic wave.
the modulation of the perpendicular magnetic anisotropy (PMA)
This inverse effect therefore acts as an energy conversion mecha-
of ultrathin magnetic films in a magnetic tunnel junction by an
nism from the magnetic to the elastic domain. The effect can
electric field. In many cases, PMA can be induced in ultrathin films
cause additional losses of propagating spin waves by emission of
and multilayers of 3d ferromagnets (e.g., Fe, Co, Ni, or their alloys)
elastic waves. These magnetoelastic losses can be limited by reduc-
by forming interfaces with nonmagnetic metals (e.g., Pt, Pd, W,
ing the “inverse” overlap integral between the dynamic magneti-
zation of the spin wave and the displacement field of the elastic and Au)269 or metal oxides (e.g., Al2 O3 , MgO, Ta2 O5 , and
wave as well as the overlap with elastic resonances. However, this HfO2 ).269–275 As an example, the interfacial PMA in CoFeB/MgO
inverse coupling can also be applied to design spin wave detec- heterostructures originates from the strong bonding of the 3d orbit-
tors. When the displacement field of the elastic wave induces als of Fe with the 2p orbitals of O. The electric field induced by
strain in an adjacent piezoelectric capacitor, it creates an oscilla- applying a voltage across the interface between the MgO and
tory charge separation and an oscillating electric polarization in CoFeB layers changes the electron density in the 3d orbitals of Fe
the piezoelectric material. The polarization can then be read out and implicitly their coupling strength with the 2p orbitals, impact-
as a microwave voltage. ing thus the interfacial PMA.276
The mutual interactions between spin waves and elastic waves Recent studies have demonstrated that the dynamic VCMA
(action and back action) in magnetostrictive media can lead to the effect by microwave (GHz) electric fields can excite ferromagnetic
formation of strongly coupled magnetoelastic waves when the resonance (FMR) in μm-scale277 to nm-scale278 magnets with a
respective dispersion relations cross. The physics of magnetoacous- power consumption of at least two orders of magnitude less than
tic waves is well understood in bulk materials,238,261–263 although the direct current induced STT excitation.277 Furthermore, it was
their behavior in thin films and waveguides has only recently been demonstrated that VCMA-based transducers can emit propagating
studied.264 When magnetoelectric transducers are employed, the spin waves.279–281 A disadvantage of VCMA-based transducers is,
excitation of magnetoacoustic waves may allow for the maximiza- however, that no spin waves or FMR-like magnetic excitations
tion of the transduction efficiency although concrete device propos- can be excited in magnets that are uniformly magnetized either
als based on magnetoacoustic waves are still lacking. in-plane (θ ¼ π2) or out-of-plane (θ ¼ 0).282 However, spin waves

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-16

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

can still be generated in these configurations by means of nonlinear of spin waves.289–292 Furthermore, it was demonstrated that the
parallel parametric pumping, in which the VCMA transducer is properties of the emitted spin waves, such as their wavelength and
driven at twice the frequency of the excited spin-wave modes energy flow direction, can be steered by shaping the laser spot or
(cf. Sec. II C).282,283 tuning the sequence of the excitation pulses.286,290,293
While VCMA-based transducers are established for the gener- The study of the magnetization dynamics induced by (sub)-ps
ation and amplification of spin waves with promise for scalability laser pulses relies typically on pump-and-probe techniques. The
and low power consumption, the detection of spin waves by first (pump) pulse triggers the magnetization oscillation, whereas
VCMA-like effects is an emerging topic.284 In addition, homodyne the probe pulse interacts with the sample after a delay [Fig. 12(a)].
detection schemes may be used, in which the microwave signal The magnetization orientation can be measured by the change in
from a spin wave is rectified and generates a DC voltage.277,278 A the polarization of a reflected probe pulse due to the magneto-optic
drawback for such detection schemes is, however, the low output Kerr effect (MOKE). Alternatively, the Faraday effect can be used
voltage, typically a few μV, which needs to be amplified to be read in a transmission geometry.291,294–296 The time resolution of the
by conventional CMOS circuits. Furthermore, the phase informa- measurement is provided by the delay between pump and probe
tion is lost since the output is converted to a DC signal. pulses and can easily reach ps time scales. High spatial resolution
can be obtained by focusing the pulses on the sample. The resolu-
tion is limited by diffraction effects and the numerical aperture of
E. Optical excitation and detection of spin waves the used microscope. The time-resolved MOKE can also be used to
While the integration of optical transducers into hybrid detect spin waves emitted by electric transducers. In this case, only
spin-wave systems is not practical, optical spin-wave excitation and the probe beam is in operation.
measurement schemes are widely used in many magnonic experi- In addition, Brillouin light scattering (BLS) spectroscopy
ments. Moreover, optical methods are capable of accessing the (Fig. 12) is a powerful technique to investigate magnetization
magnetization dynamics at ultrashort time scales of ps down to fs, dynamics because of its very high sensitivity to small spin-wave
which are difficult or impossible to assess by microwave electronics. amplitudes (e.g., thermal spin waves)297,298 and high versatility.299
Therefore, in this section, different optical methods to excite and BLS allows to study magnetization dynamics with spatial,300,301
detect spin waves are briefly reviewed. temporal,302 and phase resolution,303 as well as with wavevector
It is well known that ultrashort optical pulses with durations selectivity.304
of ps down to fs can generate spin waves in magnetic samples by The physical mechanism of BLS is based on the interaction of
different mechanisms. For example, the inverse Faraday effect can monochromatic light with a material whose optical density varies
be exploited to generate an effective magnetic field in a transparent with time and changes the light energy (frequency) and path. The
ferromagnet, generated from a circularly polarized light pulse. optical density may vary due to the presence of acoustic excitations
The effective magnetic field is parallel to the direction of the laser (phonons), magnetic excitations (spin waves), or thermal gradients
beam and can exert a torque on the magnetization. Hence, it can in the medium. The presence of spin waves in the material creates
cause the emission of spin waves.285,286 In addition, laser-induced a phase grating in the dielectric permittivity, which propagates with
thermal effects can either decrease the magnetic anisotropy287,288 the spin-wave phase velocity. The incident light is Bragg reflected
or lead to an ultrafast demagnetization process with the generation by the phase grating and its frequency undergoes a Doppler shift

FIG. 12. Simplified scheme of a time-resolved magneto-optic Kerr effect setup (a) and of a Brillouin light scattering setup (b).

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-17

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

corresponding to the spin-wave frequency. The change in the direc- waves in orthogonal planar waveguides are generally different. The
tion of the scattered light is related to the periodicity of the phase anisotropy also affects spin-wave propagation around corners and
grating. Thus, Brillouin scattered light contains information about in curved waveguides, in addition to the effects of inhomogeneous
magnetization dynamics in solids and can be used to probe the magnetization and demagnetizing field in such structures.
characteristics of magnetic excitations. The frequency analysis of Although spin waves can be guided along curved waveguides, this
the scattered light can be realized by a tandem Fabry–Pérot inter- typically results in additional losses.310–313 Although special wave-
ferometer [Fig. 12(b)].305,306 The frequency range of the interfer- guide designs alleviate the issue to some extent,314–317 the routing
ometer is typically several hundred GHz, whereas the frequency capabilities of spin waves at the nanoscale are limited, with reper-
resolution depends on the frequency range and can reach a few cussions on the spin-wave devices layout and scalability.
10 MHz at frequencies of a few GHz. The minimum detectable In planar conduits, these issues can be avoided when the
spin-wave wavelength is given by half the wavelength of the used magnetization is perpendicular to the plane since the in-plane
laser light (e.g., λSW,min ¼ 266 nm for a green laser with spin-wave properties are in this case isotropic.318 While the use of
λ ¼ 532 nm). BLS microscopy integrates a microscope objective forward volume spin waves in such a configuration is clearly advan-
with a high numerical aperture to focus the light onto the sample. tageous with more flexible device design options,164 the implemen-
Scanning the focus position can then be used to image the spin- tation is hampered by the lack of magnetic materials with
wave intensity with a spatial resolution of about 250 nm.117 simultaneous strong perpendicular anisotropy and low damping. In
thin waveguides, the demagnetization field (see Sec. II B) leads to a
V. SPIN-WAVE DEVICES strong magnetic anisotropy with an in-plane easy axis. To rotate
the easy axis out of plane, the in-plane shape anisotropy must be
After introducing basic concepts of spin-wave computing and
overcompensated by a perpendicular anisotropy. While this can
the transducers at the input and output ports of spin-wave devices,
be achieved using, e.g., magnetocrystalline319–321 or interfacial
we now discuss practical implementations of logic elements
anisotropies,322–324 the integration of such materials with low
and gates that can be used to design spin-wave logic circuits. While
damping in real devices is still challenging.
nonlinear devices such as spin-wave transistors and directional
Beyond patterned waveguides, spin waves can also be routed
couplers are also reviewed, the section focuses on passive linear
in ferromagnetic domain walls.325,326 While this may allow in
logic gates based on spin-wave interference. Linear passive gates
principle for high-density conduits structures, the fabrication of
take the most advantage of the wave computing paradigm and bear
stable domain-wall networks connecting logic gates is challenging.
the highest promise for ultralow-power electronics. The repercus-
Concepts for routing spin-wave information in three-dimensional
sions of such approaches for circuit design are then discussed
networks including multiple layers connected by vias are emerging
in Sec. VI.
only very recently.327 Multilevel spin-wave interconnects allow for
more flexible routing and potentially smaller spin-wave devices and
A. Spin-wave conduits circuits, although this is not a sine qua non requirement for spin-
The most fundamental element for information processing wave circuit design. Such approaches are, however, again strongly
and transfer by spin waves is a waveguide: the spin-wave conduit. affected by the anisotropic spin-wave dispersion relation.
In the conduit, information encoded in the spin-wave amplitude or Similar to the noisy voltage signal propagation in metallic
phase propagates at the spin-wave group velocity, which depends wires,328 the spin-wave propagation in ferromagnetic waveguides is
on material, frequency, and the effective static magnetic bias field affected by thermal noise.329 At nonzero temperature, spin waves
in the waveguide. When the spin wave wavelength is comparable to are thermally excited according to the Bose–Einstein distribution
the conduit length, the phase of the spin wave oscillates along the since the quanta of spin waves, i.e., magnons, are bosons.330
conduit. An ideal conduit material combines low Gilbert damping Thermally excited spin waves are incoherent and produce a
and high Curie temperature. Large saturation magnetization Ms background superimposed to coherent spin-wave signals used for
maximizes the spin wave power transmission and increases the computation. Moreover, adjacent waveguides may also suffer from
output signal by inductive antennas but also reduces the magnetoe- crosstalk. The dipolar magnetic fields generated by propagating
lastic coupling [cf. Eq. (29)]. Typical materials include YIG with spin waves extend beyond the waveguide and can excite spin waves
very low Gilbert damping in single-crystal form or more in adjacent waveguides. This leads to signal crosstalk between wave-
CMOS-compatible polycrystalline or amorphous metallic ferromag- guides as well as to additional propagation losses. Ultimately, this
nets such as CoFeB or permalloy (Ni80 Fe20 ), with Heusler alloys effect may limit the density of spin-wave conduits and devices in a
such as Co2 (Mnx Fe1x )Si emerging.125,307–309 Basic magnetic prop- circuit. More details on noise, crosstalk, and mitigation techniques
erties of these materials are listed in Table I. Spin-wave conduits can be found in Refs. 331 and 329.
show excellent scalability at the nanoscale and propagation of back-
ward volume spin waves in YIG waveguides as narrow as 50 nm
B. Magnonic crystals
has been demonstrated (Fig. 13), albeit with reduced attenuation
length.147 The spin-wave propagation can be further manipulated by engi-
The routing of spin waves in conduits is, however, complicated neering locally the magnetic properties or the shape of the wave-
by the anisotropic dispersion relation in the dipolar regime (see guide. Periodic manipulations lead to magnonic crystals. Magnonic
Sec. II B). For example, for a given frequency and an in-plane mag- cystals are magnetic media whose magnetic properties change peri-
netization direction, the wavelength and group velocity of spin odically in one,332,333 two,334,335 or three dimensions.336–338 They can

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-18

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

FIG. 13. (a) Schematic of the BLS

experimental configuration and scan-
ning electron micrographs of a 50 nm
wide YIG conduit. (b) Spin-wave dis-
persion relations for YIG waveguides
with different widths (w ¼ 1000, 300,
50 nm) in the backward volume geom-
etry. (c) Experimental spin-wave attenu-
ation (decay) length vs structure width.
Reproduced with permission from
B. Heinz, T. Brächer, M. Schneider,
Q. Wang, B. Lägel, A. M. Friedel,
D. Breitbach, S. Steinert, T. Meyer,
M. Kewenig, C. Dubs, P. Pirro, and
A. V. Chumak, Nano Lett. 20, 4220
(2020). Copyright 2020 American
Chemical Society.

be considered the magnonic equivalents to optical Bragg mirrors. in the magnetic properties of the structure [Fig. 14(a)]. Thus, the
The transmission spin-wave spectra through a magnonic crystal spectral positions of the bandgaps are determined by the spatial
show rejection bandgaps, i.e., frequency intervals, in which spin modulation periodicity of the crystal [see Fig. 14(b)]. The Bragg
waves are forbidden to propagate.339,340 condition for the forbidden spin-wave modes can be written as
The formation of such bandgaps can be attributed to Bragg
reflections of the spin waves by the artificial spatial grating created 2Λ ¼ nλ, n [ N, (32)

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-19

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

C. Spin-wave transistors
The basic building block of CMOS circuits is a transistor.
Given success of CMOS, one may find it thus natural to mimic the
transistor functionality using spin waves. A conventional transistor
can act both as a switch and as an amplifier and shows nonlinear
characteristics. Spin-wave transistors thus typically employ nonlin-
ear effects (see Sec. II C) beyond the linear small-signal approxima-
tion in Sec. II B.135,136,348,349
A proposal of a nonlinear spin-wave transistors has been pub-
lished in Refs. 61 and 347. They are based on nonlinear interactions
of spin waves propagating in a waveguide from “source” to “drain”
with spin waves that are injected in a “gate” section of the wave-
guide (see Fig. 15). The presence of spin waves in the gate modulates
the spin-wave transmission along the “channel” via four-magnon
scattering. To optimize the modulation and to confine the spin
waves in the gate, the central section of the transistor consists of a
magnonic crystal, as discussed in Sec. V B.
Recently, a “linear” transistor that does not require nonlinear
interactions between spin waves has been demonstrated.350 In this
device, spin waves propagate in a waveguide from source to drain
and interfere constructively or destructively with spin waves with
variable phase from the gate. In this way, the spin-wave flow from
source to drain can be modulated by the gate spin waves.
The modulation of spin-wave transmission between source
and drain by spin-wave injection into the gate allows for the opera-
tion of such a device as a switch. By contrast, the proposed spin-
wave transistors show no (or at best weak) gain and thus cannot be
operated as amplifiers, which complicates their usage in spin-wave
circuits (cf. Sec. VI). Together with the rather weak modulation of
the spin current (well below the typical on–off current ratios of 106
in CMOS transistors), this entails that spin-wave transistors are no
direct alternative to CMOS transistors. Nevertheless, the spin-wave
transistor prototype61 opened a new research avenue for all-magnon
data processing. In this concept, the spin-wave nonlinearity is used
FIG. 14. (a) Schematic representation of Bragg reflection of spin waves (wave- to process as much information as possible in the magnetic system
length λ) from a one-dimensional magnonic crystal with periodicity Λ. (b) instead of conversion of spin-wave energy in electric signals after
Dispersion relation of a SSW mode in a magnonic crystal. The frequency each gate. This approach was used for the realization of a directional
bandgaps corresponding to wavenumbers kΛn ¼ nπ=Λ with n ¼ 1, 2, 3 are
indicated.

where Λ is the periodicity of the magnonic crystal modulation and

λ is the wavelength of the spin wave. The depth and the width of
the bandgaps are controlled mainly by the amplitude modulation
of the magnetic or geometric parameters. The spin-wave transmis-
sion and the spectral position of the bandgaps have been investi-
gated for various types of magnonic crystals based on numerous
magnetic materials,332,341 different shapes of the waveguide,342,343
local modulation of the saturation magnetization,344,345 or local
variations of the bias field.346
Magnonic crystals can be potentially used in a number of appli-
cations, such as spectral filters, delay lines, or phase shifters (inverters,
see below). They also form a central part of some spin-wave transistor
approaches, as discussed in Sec. V C, and can be integrated in basic FIG. 15. Schematic of a spin-wave transistor. Reproduced with permission from
V. Chumak, A. A. Serga, and B. Hillebrands, Nat. Commun. 5, 4700 (2014).
computation elements as presented in Sec. V D. More details on
Copyright 2014 Nature.
magnonic crystals can be found, e.g., in Refs. 73 and 347.

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-20

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

coupler based on spin waves,351 and a first integrated magnonic shift with respect to an unperturbed waveguide. Alternatively,
circuit in a form of a half-adder.143 These concepts will be discussed external magnetic bias fields can also be used, including effective
in Sec. V E. fields generated by magnetoelectric effects (cf. Sec. IV C) or VCMA
(cf. Sec. IV D), which promise to be more energy efficient than
D. Spin-wave logic gates Oersted fields generated by a current. An advantage of such con-
cepts is that they can be reconfigurable, e.g., when a VCMA capaci-
Conventional logic CMOS circuits are not designed directly tor is used to generate the effective magnetic field. Magnonic
on a transistor level but rather constructed based on a set of certain crystals can also be used to generate phase shifts and invert a
universal building blocks (standard cells), e.g. NAND or NOR logic phase-coded signal. A disadvantage is the more complex device
gates or SRAM cells. Therefore, it is interesting to develop an structure as well as potentially the required additional power, e.g.,
equivalent set of spin-wave-based logic gates. As argued above, con- when an electromagnet is used. A highly beneficial property of
structing logic gates from spin-wave transistors does currently not such inverters is that they do not need to be separate logic gates
appear promising. A better approach is the design of logic gates but can be integrated in the design of, e.g., the spin-wave majority
using the interference-based paradigm discussed in Sec. III C. gates discussed below. Extending the length of an input or output
Different concepts for the implementation of spin-wave logic gates waveguide by λ2 renders the input or output inverting. In general,
have been proposed, using the different encoding schemes intro- this can be expected to reduce the size of spin-wave circuits
duced in Sec. III B. A main advantage is that these gates are linear considerably.
passive devices and do not require any energy beyond the energy in In the case of amplitude level encoding, inverters can be
the spin waves themselves, which renders such approaches promis- obtained by interference with a reference wave of phase π. For a
ing for ultralow-power computing applications assuming that the suitably chosen geometry [Fig. 16(b)], the reference wave interferes
involved spin waves can be efficiently excited. destructively with a potential signal wave. If a wave is present, its
amplitude is reduced to zero, i.e., an output of 0 is obtained for an
1. Inverters and phase shifters input of 1. For an input of 0, the reference wave reaches the output,
Before discussing more complicated logic gates, it is instructive leading to logic 1. Such inverters are not passive, unlike the above
to review inverter concepts for different encoding schemes. The delay lines, and therefore require additional power to generate the
simplest inverter is obtained by using phase encoding since in this reference wave.
case, logic inversion corresponds simply to a phase shift of π. Such
a phase shift can be achieved by propagation in a waveguide with a 2. Amplitude level encoding: Logic gates based on
length of L ¼ n  12  λ, with λ being the spin-wave wavelength interferometers
and n ¼ 1, 2, 3, . . . an integer. The advantage of such inverters is
that they are passive and do not require additional external power. Initial work on spin-wave logic gates has mainly focused on
A schematic of such an inverter is graphically depicted in Fig. 16(a). amplitude level coding in combination with a device design based
In addition, phase shifting concepts can be based on the local on an analog of a Mach—Zehnder interferometer.58,62,281,355,356
modification of the spin-wave dispersion relation. Such inverters In such a spin-wave interferometer, an incoming spin wave is split
can potentially be even more compact than delay lines.352–354 Local into two waves in the interferometer arms. A current flowing
changes in saturation magnetization or waveguide width can lead through a wire perpendicular to the plane of the interferometer
to a local change in wavelength, leading to an additional phase generates an Oersted field, which leads to a relative phase shift of
the spin waves in the two interferometer arms. Subsequently, the
waves are recombined and interfere. The relative phase shift, and
therefore the amplitude of the output wave, depends thus in an
oscillatory way on the current in the wire.
This approach can be used to design different logic gates, such
as XNOR, NOR, or NAND. Basic gate structures and their opera-
tion principles are depicted in Fig. 17. It should be mentioned that
such logic gates are inherently hybrid devices since input signals
are encoded in currents whereas output signals employ spin waves
for information encoding. For logic gate operation, the parameters
are chosen so that an input current leads to destructive spin-wave
interference in the interferometer (logic 0), whereas no current
leads to constructive interference (logic 1). Additional interference
between spin waves emanating from different interferometers can
in principle be used for more complex logic gates or circuits.
Alternative proposals use voltages rather than currents, e.g., via
FIG. 16. Implementation of spin-wave inverters. (a) Phase encoding: inversion VCMA or magnetoelectric effects, to modulate the spin-wave phase
occurs by propagation along a “delay line” with a length of n  12  λ, with λ during propagation.281,356
being the spin-wave wavelength and n ¼ 1, 2, 3, . . . an integer. (b) Amplitude
encoding: inversion occurs by interference with a reference wave with phase π.
Several logic gates—e.g., NOT, NAND, or XNOR—have been
demonstrated experimentally, as illustrated in Fig. 17(a) for

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-21

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

FIG. 17. Implementation of spin-wave logic gates based on Mach–Zehnder interferometers. (a) XNOR gate consisting of two yttrium iron garnet (YIG) waveguides. The
currents I1 and I2 represent the logical inputs, whereas the logical output is given by the spin-wave interference signal. Reproduced with permission from T. Schneider,
A. A. Serga, B. Leven, B. Hillebrands, R. L. Stamps, and M. P. Kostylev, Appl. Phys. Lett. 92, 022505 (2008). Copyright 2008 AIP Publishing LLC. (b) NOR gate consisting
of the two Mach–Zehnder interferometers in a serial configuration. Reproduced with permission from S. Lee and S.-K. Kim, J. Appl. Phys. 104, 053909 (2008). Copyright
2008 AIP Publishing LLC. (c) Voltage-controlled universal NAND gate consists of two parallel waveguides. Reproduced with permission from B. Rana and Y. Otani,
Phys. Rev. Appl. 9, 014033 (2018). Copyright 2018 American Physical Society.

XNOR.58,355 Device sizes were a few mm. Since the device opera- to conventional Boolean-based circuit design. It is rather natural to
tion is based on Oersted fields generated by currents, scaling the employ phase encoding for spin-wave majority gates since the inter-
devices leads to a strongly increasing current density in the wires ference of three (or any other larger odd number) input waves with
and to reliability (e.g., electromigration) issues. If the distance phases 0 or π generates an output wave with the phase that corre-
between the wire and the waveguide is also scaled, a part of the sponds to the majority of the input waves.
increase in current density can be avoided. Nonetheless, such Spin-wave majority gates consist in general of transducers and
current-based devices scale significantly worse than devices operat- input waveguides that provide input spin waves to the logic gate, a
ing with voltages or current densities. In addition, the hybrid char- region where the spin waves can interfere, and an output port
acter of the logic gates leads to cascading issues since the output of where the phase of the output wave is detected or transferred to an
a logic gate (spin-wave amplitude/intensity) cannot be used as an input waveguide of a subsequent gate. The input spin waves must
input for a subsequent gate, which requires encoding in a current. have the same wavelength λ and amplitude in the interference
Therefore, practical spin-wave circuits entail additional electric cir- region. When the amplitudes of the three spin waves decay differ-
cuits for signal conversion. Such issues are discussed in more detail ently during propagation, it may be necessary to compensate for
in Sec. VI. the unequal decay at the input level. For correct operation, the spin
waves representing the same logic level need to be in phase at the
output. This is best realized in logic gates, in which the path
3. Phase encoding: Spin-wave majority gates lengths of the three spin waves between their respective inputs and
Beyond the initial hybrid devices, recent work has focused the output, Di (i ¼ 1, 2, 3), differ only by integer multiples of λ,
on spin-wave logic gates that encode both input and output i.e., Di  Dj ¼ n  λ with n ¼ 0, 1, 2, . . .. Such “resonant” condi-
signals in spin waves. Conventional AND and OR logic gates have tions are preferred since they allow for the utilization of the same
been demonstrated using colinear182,357 or cross junction358 input phases for all three waves. When such conditions are not
geometries. Multivalued logic gates have also been proposed by met, the spin waves accumulate different phases during propagation
combining phase and amplitude coding.359,360 The most studied to the output port, which need to be compensated for at the trans-
device is, however, the spin-wave majority gate, originally proposed ducer or external signal level.
by Khitun and Wang.59 Majority gates have recently elicited much Alternatively, an inverting input Ii can be obtained when the
interest due to potential reductions of circuit complexity with respect path length of the corresponding spin wave, Di , is extended or

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-22

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

shortened so that the spin wave accumulates an additional phase of selection to avoid strong spin-wave attenuation at the bends of the
π with respect to the others, i.e., Di  Dj ¼ n  12  λ with trident.163,168 As discussed above, using forward volume spin waves in
n ¼ 1, 2, 3, . . .. Moreover, shifting the output port by the same devices with perpendicular magnetization can alleviate these
distance leads to an inverted output signal MAJ, i.e., to an inverted constraints.164,168,169,361
logic majority (or “minority”) function. This indicates that invert- The operation of a trident-shaped spin-wave majority gate has
ers do not have to be distinct logic gates as in the case of CMOS been demonstrated experimentally at the mm scale using YIG
but can be integrated into the majority gate design in a straightfor- waveguides.361,362 Figures 18(b) and 18(c) show photographs of
ward way. the devices. The device in Fig. 18(b) used in-plane magnetized
The initial proposals of spin-wave majority gates were based YIG and backward volume spin waves,362 whereas the device in
on a trident-shaped (also referred to as Ψ-shaped) device layout Fig. 18(c) operated with forward volume spin waves in out-of-plane
[Fig. 18(a)].59,163–165 In this layout, three parallel input wave- magnetized YIG.361 The phase of the output wave was extracted
guides are combined into a single output waveguide in a region from time-domain measurements and used to assemble the full
where the spin waves interfere. It should be kept in mind that the truth table of the majority function. These proof-of-concept dem-
three waveguides are generally not equivalent, and thus the onstrations clearly indicate the feasibility of the approach. However,
lengths of the trident prongs must be adapted to the spin-wave to become competitive with CMOS, these gates need to be minia-
wavelength and the relative phase shifts that are accumulated turized to the nanoscale and their throughput needs to be
during propagation.163,164,168 Reducing the dimensions of such a improved, e.g., by selecting different spin-wave configurations with
structure to the nanoscale requires careful design and parameter high group velocity.

FIG. 18. Overview over interference-based spin-wave majority gates. (a) Schematic of a trident-shaped spin-wave majority gate. I1 to I3 indicate the three input ports,
whereas O indicates the output port. (b) and (c) Photographs of experimental trident-shaped spin-wave majority gates using YIG. Reproduced with permission from
T. Fischer, M. Kewenig, D. A. Bozhko, A. A. Serga, I. I. Syvorotka, F. Ciubotaru, C. Adelmann, B. Hillebrands, and A. V. Chumak, Appl. Phys. Lett. 110, 152401 (2017).
Copyright 2017 AIP Publishing LLC and Kanazawa, T. Goto, K. Sekiguchi, A. B. Granovsky, C. A. Ross, H. Takagi, Y. Nakamura, H. Uchida, and M. Inoue, Sci. Rep. 7,
7898 (2017). Copyright 2017 Nature. (d) Schematic of an inline spin-wave majority gate. Since the gate is reconfigurable, every port can serve as input (In ) or output (O).
(e) Scanning electron micrograph of an 850 nm wide inline spin-wave majority gate (by courtesy of G. Talmelli). (f ) Micromagnetic simulations of the operation of an
850 nm wide spin-wave majority gate. (g) Schematic of a fan-out-enabled spin-wave majority gate and (h) demonstration of the majority functions by micromagnetic simula-
tions: (i) input (0,0,0); (ii) input (0,0,π); and (iii) (0,π,0) on ports (I3 /I4 , I2 , I1 ). Reproduced with permission from A. Mahmoud, F. Vanderveken, C. Adelmann, F. Ciubotaru,
S. Hamdioui, and S. Cotofana, AIP Adv. 10, 035119 (2020). Copyright 2020 AIP Publishing LLC.

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-23

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

To tackle the scaling challenge, colinear (inline) designs of been used to design and realize directional couplers for spin
majority gates [Fig. 18(d)] have been proposed, which are more waves.143,351 For spin-wave computing, directional couplers can
compact, more scalable, and easier to fabricate than the trident- provide multiple functionalities. In the linear regime, directional cou-
shaped gates.63,167,182,357,363 In inline majority gates, spin-wave plers can act as power splitters, frequency dividers, or signal multi-
transducers are placed along a single straight waveguide.364 When plexers. In the nonlinear regime, the coupling depends on the
the transducer distance dt is equal to an integer multiple of the spin-wave amplitude and directional couplers can be used for ampli-
spin-wave wavelength λ, i.e., dt ¼ n  λ with n ¼ 1, 2, 3, . . ., tude normalization368 and the realization of logic gates.143
in-phase electrical signals at the transducers generate in-phase spin Figure 19(a) depicts a scanning electron micrograph of a
waves throughout the device, which is ideal for spin-wave interfer- nanoscale (350 nm wide waveguides, 320 nm wide gap) directional
ence. Snapshots of micromagnetic simulations of the steady-state coupler fabricated from an 85 nm thick YIG film. Spin waves were
magnetization dynamics in an 850 nm wide CoFeB waveguide are excited by inductive antennas, and their intensity distribution in
depicted in Fig. 18(f) and indicate that strong and weak majority the device was mapped by BLS microscopy [Fig. 19(b)].143 Due to
can be clearly distinguished (red representing logic 0, blue repre- the presence of the second waveguide nearby, the spin-wave disper-
senting logic 1) despite rather complex spin-wave modes and wave sion in the first waveguide splits into antisymmetric (as) and sym-
patterns. Based on the position of the output port, both a majority metric (s) modes due to the dipolar interaction between the
gate and, after additional propagation over λ2, an inverted majority waveguides. This results in an oscillation of the spin-wave energy
(minority) gate can be obtained. The output port can also be posi- between the two coupled waveguides. This means that after the
tioned between the input ports, which renders the design reconfigur- propagation for a “coupling length,” the energy of spin waves in
able.363,365 The operation of an inline majority gate has been recently one waveguide is completely transferred to the adjacent other. The
demonstrated experimentally using CoFeB as the waveguide material coupling length is defined by the wavenumber of the spin-wave
and surface spin waves with high group velocity.363,365 This approach mode and thus strongly depends on the spin-wave dispersion. The
has also allowed for the scaling of the waveguide width down into ratio of the waveguide and coupling lengths determines the power
the sub-μm range [see Fig. 18(e)].365 transmission ratio and decides, into which output waveguide the
An additional advantage of inline spin-wave majority gates is spin wave is guided.143,351 Controlling the spin-wave dispersion,
the possibility of a fan-out of 2 since spin waves can travel in both e.g., by an external magnetic bias field, can lead to multifunctional-
directions in the waveguide.365 The importance of fan-out for the ity and reconfigurability of the device.
realization of spin-wave circuits is discussed in more detail in Sec. The general transfer characteristics of directional couplers are
VI. To improve the fan-out of the majority gates, a modified design nonlinear and therefore complimentary to the linear logic gates
has been recently proposed using forward volume spin waves in based on interference that were introduced above. As discussed in
perpendicularly magnetized waveguides.169 A schematic of such a Sec. VI, logic circuits require nonlinear elements. In CMOS circuits,
gate is depicted in Fig. 18(g). Again, adding distances of λ2 can be the nonlinearity is provided by the current–voltage characteristics
used for logic inversion in specific sections of the device with the of the transistors themselves. Analogously, directional couplers may
possibility to design, e.g., inverting inputs or outputs. provide the necessary nonlinearity in spin-wave circuits. In the
Micromagnetic simulations of the operation of such majority gates nonlinear regime, an increase in spin-wave amplitude results in a
are shown in Fig. 18(h) for an excitation frequency of 9 GHz downward shift of the spin-wave dispersion relation and, conse-
(λ ¼ 2πk ¼ 60 nm) and CoFeB material parameters (cf. Table I).169 quently, in the change of the coupling length. Figure 19(b) shows
The snapshots of the resulting magnetization dynamics (blue repre- that the output spin-wave intensity strongly depends on the input
senting logic 0, red representing logic 1) represent different sets of microwave power: at lower excitation power (here 2 dBm), the spin-
input phases that demonstrate that the entire majority function can wave energy is transferred to the second waveguide, whereas a
be obtained. The snapshots also clearly demonstrate that shifting the higher excitation power (10 dBm) leads to a transfer of the energy
output position by λ2 leads to the inverted majority (minority) func- back to the first waveguide.143
tion. The advantage of such a gate is that it has two distinct output The behavior of the directional couplers can be exploited to
ports with equal spin-wave signals. Since forward volume waves can design a spin-wave half adder as an example of a simple spin-wave
be guided around bends in the waveguide, such a design can be logic circuit that consists of two directional couplers, the first
used to generate circuits of connected majority and minority gates. working in the linear regime and the second in the nonlinear
They can thus be used as “standard cells” for spin-wave circuits, regime [Fig. 19(c)]. The functionality of the half adder has been
which is the starting point of Sec. VI. It should also be noted that verified by micromagnetic simulations351 as well as by experi-
the device design concept can be extended to different output geom- ments.143 In such devices, the data are encoded in the spin-wave
etries and a fan-out .2.169 amplitude. The first linear directional coupler is designed so that it
divides the incoming spin-wave energy equally into two parts when
spin waves are present in only one of the waveguides [top two
E. Directional spin-wave couplers panels in Fig. 19(c)]. In this case, the second directional coupler
Directional couplers are passive devices commonly used in remains in the linear regime and transfers the energy to output S.
radio technology or photonics. They couple a defined amount of However, when spin waves propagate in both input waveguides,
the electromagnetic power in a transmission line into a port, which constructive interference leads to a 4 stronger spin-wave intensity
allows for the use of the signal in another circuit. In magnonics, the that is transferred entirely to the upper waveguide. In this case, the
dipolar coupling between two adjacent spin-wave conduits366,367 has second directional coupler enters the nonlinear regime and

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-24

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

FIG. 19. (a) Scanning electron micrograph of a directional coupler (shaded in blue). A small external magnetic field is applied along the YIG waveguide in the x-direction
to saturate the directional coupler in a backward volume geometry. Reproduced with permission from Q. Wang, M. Kewenig, M. Schneider, R. Verba, B. Heinz, M. Geilen,
M. Mohseni, B. Lägel, F. Ciubotaru, C. Adelmann, C. Dubs, S. D. Cotofana, T. Brächer, P. Pirro, and A. V. Chumak, Nat. Electron. (to be published 2020). Copyright 2020
Author(s), licensed under a Creative Commons Attribution (CC BY) license. (b) Nonlinear transfer characteristics of a nanoscale directional coupler. Reproduced with per-
mission from Q. Wang, M. Kewenig, M. Schneider, R. Verba, B. Heinz, M. Geilen, M. Mohseni, B. Lägel, F. Ciubotaru, C. Adelmann, C. Dubs, S. D. Cotofana, T. Brächer,
P. Pirro, and A. V. Chumak, Nat. Electron. (to be published 2020). Copyright 2020 Author(s), licensed under a Creative Commons Attribution (CC BY) license. The color
maps represent the two-dimensional spin-wave intensity distributions measured by BLS microscopy for input powers of 2 dBm (top) and 10 dBm (bottom). (c) Operating
principle of a magnonic half-adder: two-dimensions spin-wave intensity maps from micromagnetic simulations for different input combinations. Normalized spin-wave
spectra at the output ports S and C are shown on the right-hand side.

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-25

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

transfers the energy to output C, leading to the full half adder truth G. Spin-wave multiplexers
table. Further details of the operation mode of directional spin- A multiplexer is a device that selects from several analog or
wave couplers can be found in Refs. 143 and 351. digital input signals and forwards the chosen one to a single output
line. Multiplexers are mainly used to increase the amount of data
that can be sent over a network with a fixed bandwidth. Conversely,
F. Spin-wave amplifiers and repeaters a demultiplexer is a device that disentangles a single input signal into
several output signals. Parallel data transmission can, e.g., be enabled
In addition to logic devices, spin-wave circuits may also require
using different (spin-wave) frequencies in frequency-division multi-
“auxiliary” elements, such as repeaters or amplifiers. As discussed
plexing. Several approaches have been reported for the realization of
above, spin waves have a lifetimes of ns to μs and thus lose energy
a spin-wave (de-)multiplexer. A number operates by guiding spin
during computation or information transfer. Spin-wave amplifiers
waves into one arm of Y- or T-shaped structures by controlling the
are thus crucial to compensate for such losses. Similarly, propagation
magnetization using magnetic fields,376,377 including current-induced
losses can be compensated for by repeaters, which are devices that
local magnetic field control.378 A drawback of these approaches is
receive signals and retransmit them. Amplifiers and active repeaters
that they increase the power consumption.
can also provide gain in otherwise passive linear interference-based
By contrast, passive devices, which do not require electric
logic circuits.
currents, may offer much lower energy consumption. Two pro-
The amplification of spin-wave signals can be realized by
posals for such passive (de-)multiplexers have been published to
different mechanisms. In principle, the transducer concepts dis-
cussed in Sec. IV can also be used for amplification. The spin- date. The first one is based on the directional spin-wave cou-
wave signal can be enhanced by decreasing the magnetic damping plers143,351 discussed in Sec. V E. The second one is based on
in a waveguide using STT or SOT210 generated by a DC current the utilization of caustic spin-wave beams.379,380 Such caustic
(see Sec. IV B). Alternatively, spin waves can be amplified para- beams are nondiffractive spin-wave beams with stable subwave-
metrically through a temporally periodic variation of a system length transverse aperture381 and are a consequence of the
parameter. For spin waves, two cases of parametric amplification strong anisotropy of the spin-wave dispersion relation in
can be distinguished: (i) parallel and (ii) perpendicular pumping. in-plane magnetized films (cf. Sec. II B). In an anisotropic
Perpendicular parametric pumping is often described in terms of medium, the direction of the group velocity does not generally
multi-magnon (three- or four-magnon) scattering processes that coincide with the direction of the phase velocity and the wave-
are discussed in Sec. II C. This process requires the generation of vector. For sufficiently strong anisotropy, the direction of the
large-amplitude spin waves to reach the nonlinear regime and is group velocity can become independent of the wavevector in a
therefore potentially not energetically efficient for logic applica- certain part of the spectrum. In such a case, wave packets
tions. In the case of parallel pumping, the spin-wave signal can be excited with a broad (angular) spectrum of wavevectors in the
amplified by generating an alternating magnetic field with twice specific part of the dispersion relation are channeled along the
the spin-wave frequency parallel to the longitudinal component of direction of the group velocity.379–381 These caustic beams are
the magnetization. This can, e.g., be realized using inductive linear and do not interact with each other, allowing in principle
antennas,145,166,369–372 but also STT,373 VCMA,281,356 or magneto- for the realization of complex two-dimensional spin-wave net-
electric effects,374,375 which intrinsically support the coupling to works in unpatterned magnetic films.
the longitudinal component of the magnetization. The similarity These effects have been used to route spin waves in unpat-
between transducers and amplifiers has the advantage that these terned thin magnetic films. The direction of such beams depends
components do not require very different integration schemes to on the spin-wave frequency and can be controlled by an external
be embedded in the same circuit and chip. magnetic field. Thus, caustics can selectively transfer information
Spin-wave repeaters are an alternative to amplifiers and can encoded in spin waves. The frequency dependence of the phe-
provide additional memory or clocking functionality. A schematic nomenon was successfully used to realize multiplexer and demul-
of a proposed repeater based on magnetoelectric transducers tiplexer functionalities first by micromagnetic simulations379 and
in combination with out-of-plane nanomagnets is depicted in recently experimentally.380 The device concept and the operating
Fig. 8.185 As an alternative, the use of nanomagnets with canted principle are illustrated in Fig. 20. The device consists of a 30 nm
magnetic anisotropy has been proposed.59,63 For suitably designed thick narrow CoFeB waveguide as input and two output wave-
devices, spin waves propagating in a waveguide can switch a nano- guides. In the unpatterned central part of the device, caustic
magnet in a magnetoelectric element when synchronized electric beams are propagating under different angles for different spin-
signals are applied to the latter. Based on the orientation of wave frequencies. As a result, the spin-wave intensity is transferred
the magnetization of the nanomagnet, spin waves can then be to different output waveguides, depending on the frequency. This
re-emitted into the waveguide by a second clock cycle. In this way, behavior can be used to separate information encoded in spin waves
a spin-wave signal can be transferred from one stage to the next at different frequencies in frequency-division multiplexing schemes
within a clock cycle. Micromagnetic simulations have indicated to enhance the computational throughput. It provides an “all-
that the relative phase of the incoming and outgoing spin wave can magnonic” alternative to demultiplexing in the electric domain after
be controlled. Such repeaters can compensate for losses or even detection of the complex multifrequency signal by the output trans-
provide gain, as well as regenerate and normalize spin-wave signals. ducer, leading to reduced bandwidth requirements at individual
This functionality is discussed in Sec. VI. output ports.

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-26

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

as well as the main hurdles on the road to spin-wave circuits, with

a focus on gate interconnection, fan-out achievement, and input–
output consistency. The goal of the section is to provide insight
into the requirements for spin-wave devices from the viewpoint of
circuit design.
Fundamental devices, such as transistors or logic gates, have
to fulfill the following criteria so that they can be used to design
logic circuits:24,387

• Cascadability, i.e., the possibility to use the output signal of a

logic gate as input signal for a subsequent gate.
• Fan-out, i.e., the capability to drive several gates with an output
signal of a single gate.
• Logic-level restoration and robust logic levels, i.e., the logic
signals should not degrade during data transfer between individ-
ual cascaded stages in the circuit; in particular, the separation
between 0 and 1 logic levels should remain large.
• Input/output isolation, i.e., the input logic signals should only
physically affect the output logic signal but not vice versa.
FIG. 20. Device structure and experimental demonstration of spin-wave demulti-
plexing using caustic beams. Reproduced with permission from Heussner, The combination of the above criteria is currently still a major chal-
G. Talmelli, M. Geilen, B. Heinz, T. Brächer, T. Meyer, F. Ciubotaru, lenge for the practical realization of spin-wave circuits. The output
C. Adelmann, K. Yamamoto, A. A. Serga, B. Hillebrands, and P. Pirro, of a spin-wave logic gate must be capable to drive several inputs of
Phys. Status Solidi RRL 14, 1900695 (2020). Copyright 2020 Wiley. The images
subsequent logic gates in the circuit. In CMOS, this is achieved by
show the distribution of the spin-wave intensity mapped by BLS microscopy for
two different frequencies. (a) The spin-wave intensity is guided into output 1 at representing logic values of 0 and 1 by voltages of 0 and VDD ,
11.2 GHz and (b) into output 2 at 13.8 GHz. respectively, at both the logic gate inputs and outputs. Thus, an
output signal can directly drive the input of a cascaded logic gate.
Since transistors provide gain, a single transistor (or logic gate)
output can drive several other inputs of transistors or logic gates,
providing fan-out. Moreover, in digital integrated CMOS circuits,
VI. THE ROAD FROM LOGIC GATES TO SPIN-WAVE
solutions exist for communication and data exchange between
CIRCUITS
gates, for power distribution, and for local and/or global synchroni-
In Sec. V, numerous spin-wave devices have been introduced zation via a clock signal. These functions are currently provided by
that can be used as building blocks for spin-wave circuits. In spin- the interconnect system using metal wires, with optical/photonic or
wave circuits, spin-wave logic gates are combined to calculate more plasmonic interconnects being actively researched. These intercon-
complex logic functions. An example of such a more complex nection as well as power and clock distribution solutions are
circuit is an arithmetic logic unit that can perform different opera- mature and well understood from the point of view of their capabil-
tions on binary integer numbers, such as addition, subtraction, ities and the associated overhead.
multiplication, or bit shift operations. Unfortunately, this is not the case for spin-wave logic gates.
For CMOS, the circuit design methodology has been developed Straightforward cascading can be based on signal conversion
for decades and highly sophisticated design and routing software between spin-wave and electronic domains at the gate level
tools (electronic device automation, EDA) are available to enable the [Fig. 21(a)]. This means that the spin-wave signal at the output of
very large-scale integration (VLSI, also ultra-large-scale integration, a logic gate is read out by a transducer (see Sec. IV), treated if
ULSI) of billions of transistors on a chip.382,383 Such EDA tools typi- needed, and converted to a spin-wave input signal of a subsequent
cally use standard cells to design (and layout) specific circuits based logic gate by a second transducer. Such an approach appears manda-
on their logic representations. Standard cells can provide logic (e.g., tory for mixed-signal devices, specifically for the amplitude-level
NAND and NOR) or memory functions (e.g., a flip-flop). This hier- encoded gates discussed in Sec. V D 2. The advantage of this
archical design approach has been developed in the late 1970s by approach is that it fulfills all criteria. Gain can be provided after
Mead and Conway384 and has allowed to separate technology and transduction in the CMOS domain, so this scheme also allows for
system development. fan-out.
By contrast, few attempts to design spin-wave circuits have In such an approach, the overhead due to signal conversion
been made,59,165,172,182,357,385,386 and a methodology for spin-wave and CMOS data treatment needs to be considered carefully. A key
circuit design has not yet been established. While circuit design parameter that determines the overhead is the signal level generated
based on MAJ and INV is well understood67,68 and can be auto- by the transducer. On the one hand, the signal level determines the
mated, the implementation of complex circuits by spin-wave logic complexity of the CMOS circuit required to detect it. Signal levels
gates and interconnects is still challenging and has not yet been of a few 100 mV may be large enough to directly drive a transistor
demonstrated. In this section, we discuss the current understanding for amplification. Lower voltages require the usage of, e.g., sense

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-27

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

amplifiers. Phase-sensitive detection entails even more complex cir- provide phase-sensitive amplitude normalization and spin-wave
cuits.388 These CMOS circuits consume power and occupy area and signal restoration. In this approach, an incoming spin wave switches
therefore contribute significantly to the overall circuit performance. the orientation of a nanomagnet depending on its phase, as demon-
While a complete benchmark of hybrid interconnection schemes strated by micromagnetic simulations.59,63,185 In the next clock cycle,
has not yet been carried out, it is questionable whether such an an electric (pulse) signal relaunches a spin wave from the repeater
approach can operate at sufficiently low energy to outperform the that is in phase with the initial spin wave. Such schemes require,
direct implementation of the desired circuit in (low-power) CMOS. however, low gate granularity and complex clocking schemes and the
Moreover, the signal level may limit the conversion through- operation of the entire circuit may thus last multiple clock cycles,
put. As an example, the Johnson–Nyquist voltage noise in the resis- determined by the longest path in the circuit. This means that
tive component R of a transducer (e.g., in an inductive antenna) is during every clock cycle, only one gate result can be evaluated, while,
given by328,389 e.g., current CMOS logic processors employing instruction-level
parallelism can execute several full operations per clock cycle.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Enhancing the throughput of spin-wave circuits can be achieved by,
vrms ¼ 4kB TRΔf , (33) e.g., frequency-division multiplexing or pipelining.59,390 Yet, the
energy and delay overhead of such cascading schemes may still be
with vrms being the root mean square noise of the voltage, kB being significant. To date, no circuit simulation of such a scheme has been
the Boltzmann constant, T being the temperature, and Δf being the reported and future work is thus required to assess its competi-
bandwidth of the measurement. For resistances R of a few kΩ and tiveness with respect to CMOS. In addition, the switching of a
a readout bandwidth of 10 GHz, the noise is about 1 mV. The nanomagnet by a spin wave has not yet been experimentally dem-
signal thus should be at least (several) 10 mV to enable fast read onstrated, in particular, not with phase sensitivity.
out even with sensitive circuits. Similar arguments apply for capaci- Recently, an alternative method of signal normalization has
tive (e.g., magnetoelectric) transducers. Hence, hybrid interconnec- been proposed using directional spin-wave couplers (see Sec. V E).368
tion schemes may add also a significant delay to the circuit. Directional couplers operate based on nonlinear spin-wave interac-
It is therefore strongly preferred to cascade and interconnect tions and can be designed to couple a spin wave with a certain
logic gates in the spin-wave domain without conversion to electronic amplitude (i.e., a normalized amplitude min ) into an adjacent wave-
signals. However, additional issues arise for spin-wave logic gates guide, independent of the amplitude of the propagating spin wave.
using phase-encoded information. While the interference of spin As demonstrated by micromagnetic simulations,368 this allows for
waves with phases 0 or π and amplitude min in a majority gate gen- “passive” spin-wave amplitude normalization without the need to
erates the correct output phase, the amplitude of the resulting spin switch nanomagnets and for clocked signal repetition.
wave mout is different in the cases of strong (fully constructive inter- Yet, approaches to connect spin-wave gates by means of wave-
ference) and weak (partially destructive interference) majority. guides, repeaters, or directional couplers may still add substantial
Concretely, if two input phases of a spin-wave majority gate are iden- overhead to the circuit since spin waves propagate rather slowly
tical and the third is different (weak majority), the amplitude of the through waveguides. While the actual gate interconnection delay is
generated spin wave is mout ¼ min , whereas it is mout ¼ 3  min in circuit dependent, it is in any case much longer than that of metallic
the case of strong majority, i.e., when all three input phases are iden- or optical interconnects. Indicative numbers for spin-wave group
tical. Consequently, if two majority gates are directly cascaded, velocities can be found in Table I. The propagation delay is typically
amplitude differences at the output of the driving gate can lead to a few 100 ps/μm, which can add significant delays for large circuits
wrong results at the driven gate, which has been designed to operate and impedes the utilization of waveguides and repeaters for long
with equal spin-wave input amplitudes min . For example, if a driving range interconnects. Moreover, when a spin wave propagates along a
gate produces a strong 0 output, whereas the other two input signals waveguide, its amplitude is attenuated due to Gilbert damping,
of the driven gate are weak 1 signal, the output of the driven gate is which may affect the next logic gate if the amplitude is much lower
0 and not 1 as expected. Therefore, a certain mechanism to restore than the expected value of min . This may require the utilization of
or normalize the spin-wave amplitude is required between gates to spin-wave amplifiers or repeaters (Sec. V F) to compensate for losses
guarantee proper circuit behavior. Note that since the amplitude nor- with added energy and delay overhead. Hence, all these schemes rely
malization is a nonlinear operation, it cannot be implemented using on the availability of a variety of energy-efficient and fast spin-wave
linear devices, e.g., based on spin-wave interference. devices beyond the logic gates themselves. However, the granularity
Two main approaches have been proposed to normalize the of the signal conversion, amplification, or repetition is still crucial
amplitude of a spin wave. In spintronics, an obvious nonlinear oper- for the performance of the spin-wave circuit. Cascading in the
ation is the switching of a nanomagnet, which provides a threshold CMOS domain or by switching nanomagnets entails a granularity
function. Moreover, the information storage in nanomagnets is non- at the logic gate level. By contrast, directional couplers may
volatile, which provides a route toward nonvolatile logic circuits. This increase the granularity to dimensions comparable to the spin-
points to the usage of spin-wave repeaters (see Sec. V F) between wave attenuation length with much less associated overhead, e.g.,
logic gates [Figs. 21(b) and 22] that can both normalize and restore from clocking circuits.
spin-wave signals. Repeaters can also compensate for propagation Apart from cascadability, circuits require gate fan-out since
losses and provide gain as well as fan-out. Different repeater concepts one gate output signal is often used as input signal for more than
have been proposed based on canted nanomagnets59,63,64 or magne- one gate, as illustrated in Fig. 23. In CMOS, fan-out achievement is
toelectric elements with perpendicular anisotropy,185 which both can straightforward due to the inherent gain of CMOS transistors.

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-28

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

FIG. 21. Spin-wave gate interconnection schemes. (a) Hybrid interconnection with signal conversion from the spin wave to the electronic or photonic/plasmonic domain.
The signal is then regenerated, transmitted to the next gate inputs by electronic or photonic/plasmonic interconnects, and converted again to the spin-wave domain. (b)
Clocked interconnection is possible by phase-sensitive switching of a nanomagnet (NM) by a spin wave. In a next clock cycle, a secondary spin wave is launched again
from the nanomagnet with a defined relative phase. (c) All spin-wave interconnections require a nonlinear device that normalizes the amplitude of the output spin wave.
Directional spin-wave couplers can provide such functionality.

Thus, the output voltage of a logic gate can be directly fed into Beyond spin-wave circuits designed by majority-gate and
several inputs by metallic wires. By contrast, achieving fan-out in inverter synthesis, the computation with waves opens other pos-
spin-wave circuits is less straightforward as it requires replication of sibilities for circuit design, in particular, for network-like cir-
the spin-wave signal. Signal division can be achieved using cuits, such as reconfigurable meshes,392 cellular nonlinear
Y-shaped waveguides (as in Fig. 23) or directional couplers. networks,393,394 or systolic arrays.395 These approaches can
However, since the spin-wave energy (intensity) is conserved, split- enable parallel computing using specific algorithms and bridge
ting a spin wave reduces the amplitude of the two resulting spin the gap to neuromorphic computing schemes. Different
waves by p1ffiffi2 even without additional losses. This needs to be com- spin-wave-based implementations have been pro-
pensated for by spin-wave amplifiers (Fig. 23) with additional posed.59,151,165,385,396,397 Such circuits can be represented by a
energy and possibly delay overhead. In contrast, fan-out enabled set of nodes, e.g., spin-wave repeaters, connected by a network of
majority gates [Fig. 18(g)] provide two equivalent outputs without waveguides, as represented in Fig. 22. A discussion of such com-
the need to split the spin wave after computation.169,391 An alterna- puting architectures is beyond the scope of this Tutorial. More
tive is the replication of the logic gate or the subcircuit itself to details can be found, e.g., in Refs. 398–400. To date, none of
provide two (or more) identical outputs for the realization of these computing architectures has been experimentally realized.
fan-out. However, for large circuits, this leads to considerable area A major obstacle is the rather strong spin-wave attenuation in
and energy overhead. As an example, if the output of a 32-bit many magnetic materials that limits the maximum size of such
adder is required at the input of two or more gates, the entire networks, especially since spin waves may have to propagate
32-bit adder needs to be replicated twice or more. along complex pathways.

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-29

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

FIG. 22. Schematic view of a spin-wave circuit with

nanomagnet-based cascading. Spin waves propagating
between nodes of the circuit switch the magnetization of
bistable nanomagnets. Clock electrodes then provide
trigger signals to launch spin waves from one node to the
next in the following clock cycle. Reproduced with permis-
sion from A. Khitun and K. L. Wang, J. Appl. Phys. 110,
034306 (2011). Copyright 2011 AIP Publishing LLC.

VII. HYBRID SPIN-WAVE–CMOS SYSTEMS concepts for spin-wave memory elements. These limitations can be
Section VI has outlined potential solutions to design spin- overcome by embedding spin-wave circuits in a CMOS and/or
wave circuits based on a set of basic devices, namely, waveguides, mixed signal environment, resulting in hybrid spin-wave–CMOS
systems. The performance of such a system is determined by the
majority gates, inverters, amplitude normalizers, amplifiers, and
individual performances of the spin-wave circuit, the CMOS envi-
transducers. The extension of such circuits to complete competitive
spin-wave-based computing systems is however limited, e.g., by the ronment, and last but not least the interdomain transducers.
lack of high-performance long-distance interconnection or To date, little attention has been devoted to hybrid systems and
experiments have been typically carried out using vector network
analyzers or optical detection techniques like BLS. Whereas such
techniques are useful for fundamental research and proof-of-concept
demonstrations, they cannot be employed in real-world applications
and need ultimately to be replaced by CMOS-based (mixed-signal)
periphery circuits that provide input signals and analyze the output
of the spin-wave circuit. It is clear that the benchmarking of spin-
wave computing technology must ultimately be accomplished on
complete systems including periphery, not only on the spin-wave
circuit or at the device level. Although no hybrid circuit has been
realized experimentally to date, a benchmark of hybrid spin-wave–
CMOS arithmetic circuits has been recently performed and reported,
based on the design and simulation of specific logic circuits.65,172,386
The benchmark suite included adders (BKA264, HCA464, CSA464),
multipliers (DTM32, WTM32, DTM64, GFMUL), a multiply-and-
accumulate (MAC) module, a divider (DIV32), and a cyclic-
redundancy-check (CRC32) module. These logic circuits have been
implemented using majority-based design approaches and layouted
using majority-gate and inverter primitives (see Refs. 172 and 386
for specific designs and device footprints). The input signals of the
spin-wave circuits were synthesized using a 10-nm-CMOS-based
circuit. The output signals were detected using sense amplifiers, also
implemented in 10 nm CMOS. The signal conversion at the bound-
aries between spin-wave and CMOS domains was realized by magne-
toelectric transducers. A schematic of such a system is represented in
Fig. 24(a). Due to the CMOS-based periphery at the inputs and
outputs of the spin-wave circuit, the system can operate with logic
levels encoded in voltages and thus interact with conventional
charge-based electronic systems, including memory.
For comparison, the same circuits were implemented in 10 nm
FIG. 23. Schematic of cascaded spin-wave majority gates with a fan-out of 2. CMOS using conventional EDA software. The performance of both
Amplitude normalizers and amplifiers are required at the inputs of the secondary
types of circuits was then simulated by commercial software tools,
majority gates.
using spin-wave gate delays and energies obtained from

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-30

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

FIG. 24. Benchmark of hybrid spin-wave–CMOS systems. (a) Schematic of the hybrid system. Reproduced with permission from Zografos, P. Raghavan, L. Amarù,
B. Sorée, R. Lauwereins, I. Radu, D. Verkest, and A. Thean, in 2014 IEEE/ACM International Symposium on Nanoscale Architectures (NANOARCH) (2014) pp. 25–30.
Copyright 2014 IEEE. (b) Area-delay-power (ADP) product of several arithmetic circuits (see the text) implemented in hybrid spin-wave–CMOS technology as well as
10 nm CMOS as a reference. Reproduced with permission from O. Zografos, B. Sorée, A. Vaysset, S. Cosemans, L. Amarù, P.-E. Gaillardon, G. D. Micheli,
R. Lauwereins, S. Sayan, P. Raghavan, I. P. Radu, and A. Thean, in 2015 IEEE 15th International Conference on Nanotechnology (IEEE-NANO) (2015) pp. 686–689.
Copyright 2015 IEEE.

micromagnetic simulations. The performances have then be com- (including the spin-wave wavelength) is 48 nm; (ii) the spin-wave
pared in terms of power, area, and delay. The results in terms of excitation and detection is performed by means of magnetoelectric
the area-delay-power product (ADPP) are depicted in Fig. 24(b).386 transducers (delay 0.42 ns, energy consumption 14.4 aJ); (iii) the
Currently, no complete methodology to assess the properties and delay and energy loss due to spin-wave propagation within the
performance of spin-wave circuits and transducers is available, and waveguide are negligible with respect to the overhead due to spin-
thus several assumptions were made:386 (i) the critical dimension wave generation and detection; and (iv) the signals provided by the

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-31

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

magnetoelectric transducers at the output ports of the spin-wave (well beyond the individual gate level) and the number of trans-
circuit (
100 mV) are read out using a CMOS sense amplifier ducers and sense amplifiers should be minimal. On the other hand,
(delay of 0.03 ns, energy consumption 2.7 fJ). Under such assump- large spin-wave circuits require frequent signal amplification and
tions, the results in Fig. 24(b) indicate that the ADPP of hybrid restoration to compensate for losses due to magnetic damping and
spin-wave–CMOS arithmetic circuits can be significantly lower possibly dephasing. Moreover, the layout of large-scale spin-wave cir-
than that of their 10 nm CMOS counterparts. cuits is complicated by losses due to bent waveguides as well as the
Although several assumptions in the benchmark are certainly current lack of multilevel interconnects and spin-wave vias. Large cir-
not fully realistic and actual area-delay-power product of hybrid cuits may thus have to be partitioned into spin-wave islands embed-
spin-wave–CMOS can be expected to be larger, several interesting ded in a CMOS periphery. Inside these islands, data processing is
conclusions can be drawn from this exercise. As an example, performed by cascaded spin-wave gates, whereas the islands them-
the area-delay-power product for a very complex circuit such as selves are interconnected using electric (charge, voltage) signals after
DIV32 implemented in hybrid spin-wave–CMOS is roughly about conversion by (magnetoelectric) transducers. These conversion
800 lower than its CMOS implementation; individually, the blocks can also restore the signal, reducing the need for signal resto-
power consumption is about 1800 lower, the area is about ration and amplification in the spin-wave domain. A possible length
3:5 smaller, whereas the delay is about 8 longer. The results scale for the spin-wave islands could be the spin-wave attenuation
indicate that (i) hybrid spin-wave–CMOS circuits are promising as length, which suggests the usage of low-damping magnetic materials.
ultralow power circuits although at the expense of latency (delay, Nonetheless, the conversion blocks contribute to the overall circuit
throughput). Nonetheless, under the above assumptions, the delay and the overall energy consumption,386 and therefore the
power-delay product may still be lower than that of 10 nm CMOS. optimum spin-wave island granularity depends on the properties of
(ii) The power dissipation in the CMOS periphery is typically both the spin-wave system and the conversion block, consisting of
much larger than in the spin-wave circuit itself. This means that transducers and CMOS periphery.
the performance advantage of hybrid spin-wave–CMOS circuits Finally, practical circuits require clocking schemes—a neces-
typically increases with their size since the CMOS periphery over- sary evil that most computation platforms cannot properly function
head becomes relatively smaller. As an example, Fig. 24(b) indicates without. Clocking spin-wave circuits and systems can also be an
that hybrid spin-wave–CMOS implementations of large multipliers important contributor to the circuit complexity and performance.
(DTM64) or dividers (DIV32) outperform CMOS in this bench- For example, if the information is converted from spin wave to
mark, whereas smaller adders (BKA264, HCA464) show little to no charge and back at the individual logic gate level, a complex clock-
advantage. (iii) The area of hybrid spin-wave–CMOS circuits can ing circuit is required to control the gate-output sampling process.
be competitive with CMOS circuits despite the comparatively large A similarly complex clocking system is required for nanomagnet-
critical dimension of 48 nm, which is within the limits of single- based spin-wave repeaters, which require clock control of each
exposure immersion lithography. This is due to the efficiency of nanomagnet node, potentially with large overheads. By contrast, if
the majority gate design. Again, advantages increase with the size cascading can be achieved by “passive” spin-wave amplitude nor-
of the spin-wave circuit since the area overhead occupied by the malizers, signals need to be converted only at the island outputs, in
CMOS periphery becomes relatively smaller. the same way as pipeline stage outputs are sampled in a pipelined
While this first benchmark clearly indicates promise of hybrid processor.390 This substantially diminishes the clock distribution
spin-wave–CMOS circuits, the assumptions appear not yet fully network complexity and allows for lower clock frequency and sig-
realistic. Future improved benchmarking studies should include, nificantly reduced energy consumption.
e.g., the material-dependent propagation delay of spin waves, as well Another essential aspect for the energy consumption of spin-
as the overheads due to gate cascading, signal renormalization, and wave circuits is the operation mode. When spin waves are excited
fan-out achievement. The computing throughput can be enhanced by continuous-wave microwave signals at the input transducers,
in principle by frequency-division multiplexing, although this also the overall energy consumption is determined by the input power
increases the overhead due to the associated multifrequency CMOS and the delay by the critical path in the island (i.e., the longest
periphery and the system-level advantages are not yet clear. The spin-wave propagation distance) and/or the bandwidth of the trans-
availability of compact models for spin-wave devices and for trans- ducers as well as the readout circuitry. Therefore, materials with
ducers is essential for the accurate behavior and performance evalua- fast spin-wave propagation velocity are clearly favored for wave-
tion of hybrid spin-wave–CMOS circuits with a SPICE-based guides and logic gates. Alternatively, spin waves can be excited by
simulation framework.331,401 However, despite its limitations, the microwave pulses to reduce the energy consumption per operation.
benchmark clearly indicates that hybrid spin-wave–CMOS systems This may also allow for pipelined computation schemes using a
bear promise for ultralow-power applications. Moreover, it demon- spin-wave pulse train propagating in the circuit. However, the exci-
strates that future spin-wave-based technologies need to be assessed tation of propagating stable wave packets by microwave pulses is
at the systems level—and not on the device or (small) circuit level. not trivial due to the nonlinear spin-wave dispersion relation that
An open question relates to the spin-wave processing island can lead to pulse distortion. In addition, long spin-wave lifetimes
granularity, i.e., the maximum complexity of a practical spin-wave are detrimental to the formation of short wave packets since the
circuit that can be implemented without requiring forth and back magnetization dynamics at the excitation site decay only slowly.
conversion to the charge (CMOS) domain. To minimize energy Spin-wave solitons136,402,403 due to nonlinear “bunching” or pulse
consumption of the full system, signal conversion between spin- compression effects may offer a potential solution but require spe-
wave and charge domains must be sufficiently coarse-grained cific conditions as well as high excitation power. In such cases,

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-32

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

when short wave packets or solitons are used, the control of the into the third dimension has intensified,76 and several
propagation and the interference in the structure is challenging and proof-of-concept experiments have been demonstrated.416,417 The
requires complex clocking schemes. Although frequent spin-wave fabrication of such three-dimensional structures was enabled by
repeaters may alleviate the issue (while adding considerable overhead), the recent advances in focused electron beam induced deposition
the feasibility of hybrid systems built on throughput-optimized spin- (FEBID).418 FEBID is a promising three-dimensional direct-write
wave islands realized with waveguide interconnection without repeat- nanofabrication technique,418,419 which opens prospects to build-
ers is still and open issue. ing magnonic three-dimensional nanoarchitectures with complex
interconnectivity and the development of novel types of human
VIII. SPIN-WAVE APPLICATIONS BEYOND LOGIC GATES brain-inspired neuromorphic networks using spin waves. In addi-
tion, the ease of area-selective tuning of the magnetization in
A. Unconventional and analog computing approaches spin-wave conduits via their postgrowth irradiation with ions420
Beyond digital spin-logic circuits and wave computing systems, or electrons,421 or the proximity to superconductors422 opens
spin-wave-based “unconventional” and analog circuits have also been pathway to the fabrication of spin-wave circuits with graded
proposed. A brief discussion has already been presented at the end refractive index for the steering of spin waves in curved wave-
of Sec. VI. While less universal than digital systems, these concepts guides or into the third dimension.
take particular advantage of the wave nature of spin waves and can
be very efficient for specific tasks such as signal and data process- C. Toward quantum magnonics
ing,153,404,405 prime factorization,406,407 or Fourier transforms.408
Pioneering work on wave-based computing in the 1970s and One of the prominent advantages of magnonics is the
1980s has used photons to develop optical computers.19,20,24 While possibility to exploit complex data processing concepts at room
optical data communication is today ubiquitous, optical computing temperature. Nevertheless, in recent years, increasing attention has
has not become competitive with CMOS. The challenges of optical been devoted to the behavior of spin waves at cryogenic tempera-
computing overlap with those of spin-wave computing and the tures for two reasons. First, the physics of hybrid superconductor-
realization of competitive optical computers has been hindered by ferromagnet structures provides access to fascinating new physics
difficulties to confine photons at ultrasmall length scales and the that may potentially be exploited for data processing or quantum
power efficiency at the transducer level.24,409 Nonetheless, both computing. Second, decreasing the temperature below 100 mK
digital and analog computing concepts have been developed and the leads to the freeze-out of thermal magnons, which enables experi-
work on optical computing has inspired spin-wave computing.408 ments with single magnons. Thus, such conditions give access to
An example for a nonbinary computing architecture is the quantum magnonics.
magnonic holographic memory. It consists of a two-dimensional The combination of ferromagnetism and superconductivity in
network of crossing waveguides with transducers for spin-wave hybrid ferromagnet/superconductor (F/S) systems leads to emerg-
excitation and detection at the edges.153,186,187,410 After spin waves ing physical phenomena. For instance, in proximity-coupled S/F/S
have been excited, they propagate through the structure, interfere three layers, a substantial reduction of the ferromagnetic resonance
with each other, and generate an interference pattern in the field is attributed to the generation of unconventional spin-triplet
network. In such a structure, all inputs directly affect all outputs, superconductivity.423 It has been demonstrated that coupling of
which can be used for parallel data processing.151,153,404,405,408 spin waves in F with S results in an enhanced phase velocity of the
Cellular nonlinear networks are structurally similar to magnonic spin waves due to the Meissner screening of AC magnetostatic stray
holographic memories and consist also of an array of magnetic fields by S.424 Several novel effects emerge for proximity-decoupled
waveguides.151 By contrast, active transducers at every waveguide S/F hybrids in out-of-plane magnetic fields.425 When the S layer is
crosspoint can be used to locally manipulate the magnetization. in the mixed state, an external magnetic field can penetrate in the
Wave superposition and interference can again be used for parallel form of a lattice of Abrikosov vortices (fluxons). The stray fields
data or image processing.404,406,411 emanating from the vortex cores produce a periodic modulation of
Spin waves can also be employed for the design of reversible the magnetic order in F such that the S/F bilayer can be viewed as
logic gates.412 Here, both reversibility of the logic operation and the a fluxon-induced magnonic crystal. It has been shown that the
physical processes are used to perform ultralow energy operations. Bragg scattering of spin waves on a flux lattice moving under the
Moreover, several spin-wave-based concepts for neuromorphic action of a transport current in the S layer is accompanied by
computing have been proposed.151–154,413–415 Finally, the asymmet- Doppler shifts.425 An additional promising research direction is
ric propagation and nonlinear behavior of spin waves renders them related to the experimental examination of a Cherenkov-like radia-
promising candidates for reservoir computing.155–157 tion of spin waves by fast-moving fluxons when the vortex velocity
exceeds a threshold value.426 To prevent instability and the collapse
of vortices at the velocity of required 5–15 km/s, one can use, e.g.,
B. Three-dimensional magnonics
superconductors with fast relaxation of disequilibrium.427
The spin-wave devices described in this Tutorial are based Hybrid systems based on superconducting circuits allow also
on films and multilayers that are prepared by thin film deposition for the engineering of quantum sensors that exploit different
techniques and lithographically patterned into the desired struc- degrees of freedom. Quantum magnonics,428–433 which aims to
tures. Hence, the resulting structures are all planar and two- control and read out single magnons, provides opportunities for
dimensional. Recently, research to extended the planar structures advances in both the study of spin-wave physics and the

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-33

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

development of quantum technologies. The detection of a single electric (microwave) and magnetic domains. Reducing the amount
magnon in a millimeter-sized YIG crystal with a quantum effi- of magnetic material in scaled devices degrades the power conver-
ciency of up to 0.71 was reported recently.428 The detection was sion efficiency and lead to similar issues that need to be overcome
based on the entanglement between a magnetostatic mode and the for nanoscale logic circuits. Therefore, advances in spin-wave trans-
qubit, followed by a single-shot measurement of the qubit state. ducer technology may additionally enable nanoscale analog micro-
The strong coupling of magnons and cavity microwave photons is wave applications with interesting prospects for telecommunication.
one of the routes toward quantum magnonics, which is intensively More recently, increasing interest has been devoted to magne-
explored nowadays.429–435 toelectric antennas. Conventional dipolar antennas are difficult to
In addition to single-magnon operations expected to be real- scale due to the large wavelength of electromagnetic waves in
ized at mK temperatures, macroscopic quantum states such as air455,456 and often suffer from losses due to near-field interactions
magnon Bose–Einstein Condensates (BECs) at room temperature with the environment.457,458 Lately, an alternative antenna type
have also been considered potential data carriers. The fundamental based on magnetoelectric composites has been proposed,459,460
phenomenon of Bose–Einstein condensation has been observed in which consists of a piezoelectric–magnetostrictive bilayer. Applying
different systems of both real particles and quasiparticles. The con- a microwave signal to such an antenna produces an oscillating
densation of real particles is achieved through temperature reduc- magnetic dipolar field, which acts as a source of electromagnetic
tion while for quasiparticles like magnons, a mechanism of external radiation.461–463 The response can be enhanced by acoustic and
boson injection by irradiation is required,436,437 or, as demonstrated magnetic resonances. Due to the much shorter wavelengths of
recently, a rapid-cooling mechanism can be exploited.438 Moreover, acoustic and magnetic waves at microwave frequencies, magneto-
a supercurrent in a room-temperature Bose–Einstein magnon con- electric antennas can be more compact that conventional dipolar
densate was demonstrated experimentally.436 The observation of a antennas and may require less power.459,464,465
supercurrent confirms the phase coherence of the observed
magnon condensate and may be potentially used in future mag-
nonic devices for low-loss information transfer and processing. F. Antiferromagnetic magnonics and terahertz
applications
D. Spin-wave sensors In recent years, antiferromagnetic spintronics have received
increasing attention as an extension of established spintronic
The on-chip integrability and miniaturization of spin-wave
approaches based on ferromagnets or ferrimagnets.466–468 The spin-
devices can also be employed for magnetic field sensing applica-
wave frequencies in antiferromagets are in the THz range469–472 and
tions. CMOS compatible magnetic sensors play a crucial role in a
therefore antiferromagnetic magnonics are of interest for THz appli-
variety of industries, including the automotive industry, biomedical
cations.473,474 In principle, antiferromagnetic media may conceptu-
applications, navigation, and robotics. Specifically, magnetoresistive
ally enable spin-wave logic at THz frequencies with prospects of
sensors,40,439,440 based on anisotropic magnetoresistance, giant
better scalability and higher operating speed.60 However, methods of
magnetoresistance, or tunnel magnetoresistance, have found wide-
controlling and detecting magnetic excitations in antiferromagnets
spread commercial application due to their high sensitivity as well
are only emerging.475–478 To date, logic devices utilizing antiferro-
as low noise and power consumption.40,439,440 Recently, several
magnetic spin waves have not yet been demonstrated. In particular,
pioneer investigations have been performed to explore the possibil-
the controlled excitation and the detection of phase-coherent THz
ity to use spin waves for magnetic sensors.441–447 In particular,
spin waves in antiferromagnetic waveguides is still lacking, as are
magnonic crystals, periodic magnetic structures, have been pro-
concepts to efficiently generate THz logic signals by CMOS circuits.
posed as sensors with very high sensitivity.441,442,445,446 Magnonic
Yet, if fundamental research on antiferromagnetic spintronics
crystals have also been used for the sensing of magnetic nanoparti-
continues at a fast pace, spin-wave logic at THz frequency may
cles.443 Finally, magnon polaritons in PT-symmetric cavities have
become an interesting alternative to the GHz approaches based on
been proposed for sensors with very high sensitivity.444 Such mini-
ferromagnetic media.
ature sensor applications share many properties of the logic circuits
discussed in this Tutorial and may also strongly benefit from opti-
mized spin-wave transducers and readout circuitry. IX. CONCLUSIONS, STATE OF THE ART, AND FUTURE
CHALLENGES
E. Microwave signal processing
The state-of-the-art of spin-wave computing has experienced
To date, commercial applications of ferromagnetic resonance tremendous advances in the last decade with several proof-of-
and spin waves mainly include macroscopic tunable microwave concept realizations of key devices, such as the spin-wave majority
filters, power limiters, circulators, or gyrators based on ferrite mate- gate.361,362,365 Much progress has been made, in particular, in the
rials, especially low-damping YIG.448,449 Much research has been understanding of the properties of spin waves in nanostructures.
devoted to such devices between the 1960s and 1980s.450–454 The overview of spin-wave devices in Secs. IV and V clearly indi-
Several devices are today commercially available, although typically cates that methods to manipulate spin waves at the nanoscale are
for niche applications. These devices employ typically magnetic ele- ever improving. Scaled individual spin-wave logic gates and many
ments in the mm size range. For such large quantities of magnetic features of wave-based computing have been demonstrated.365
material, the microwave absorption by ferromagnetic resonance or Hence, the field of magnonics is rapidly evolving. Moreover, bench-
spin waves is large, leading to efficient power conversion between marking studies have suggested that hybrid spin-wave–CMOS

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-34

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

computing systems can potentially operate at much lower power Efficient spin-wave detection is also still challenging. As for
than conventional CMOS circuits. generation, the microwave power induced in an antenna decreases
Yet, several obstacles still exist on the road toward the realiza- with the magnetic volume underneath. To efficiently convert the
tion of competitive hybrid spin-wave–CMOS computing systems. result of a spin-wave computation to a CMOS-compatible signal,
In Secs. IX A–IX F, we present our view on the most critical the transducer should ideally generate output signals of about
hurdles. For a number of these obstacles, potential solutions have 100 mV. Such large signals have been typically an issue for many
been proposed but need to be demonstrated and properly assessed spintronic logic technologies. Magnetoelectric transducers may
in terms of energy and delay overhead, while others have been less provide a potential solution but the detailed coupling of spin waves
addressed in the research literature so far. to strain and acoustic oscillations in realistic devices has not yet
been studied in detail. The demonstration of 1 mV output
signals in magnetoelectric transducers would certainly be a major
A. Cascading, fan-out, and signal restoration in breakthrough for spin-wave-based computing as well as for other
spin-wave circuits potential applications.
As discussed in Sec. VI, the step from individual spin-wave
devices to circuits requires the possibility to cascade logic gates. C. Device scaling
Signal normalization, restoration, and fan-out achievement are crit-
As mentioned above, the scaling of the magnetic volume in a
ical requirements that need to be fulfilled for a practical circuit.
spin-wave device reduces the efficiency of transducers, for both
Cascading using conventional charge-based interconnects is a possi-
generation and detection. Scaling device dimensions also has reper-
bility but the frequent transduction between spin-wave and charge
cussions on the properties of the spin waves themselves. Narrow
domains almost certainly consumes much energy, which may and
waveguides exhibit strong internal dipolar magnetic fields due to
render such approaches uncompetitive. Phase sensitive switching of
shape anisotropy. The magnetization is thus preferentially aligned
nanomagnets by spin waves remains to be demonstrated experimen-
along the waveguide, which means that scaled devices typically
tally, and the energy efficiency of real devices needs to be established.
operate with backward-volume spin waves. A distinct advantage of
The development of compact models for spin-wave repeaters and
this geometry can be the “self-biasing” due to the strong anisotropy
clocked interconnects that are calibrated to experimental devices can
field, which does not require external magnetic bias fields. By con-
then quantify energy and delay overheads. Similar arguments apply
trast, the excitation of surface waves requires large external fields to
to cascading approaches in the spin-wave domain using directional
rotate the magnetization transverse to the waveguide, which may
couplers. Experimental demonstrations together with calibrated
not be practical.
models can allow the assessment of the energy efficiency and
Device scaling also has strong repercussions on the spin-wave
throughput of spin-wave circuits. A first breakthrough would be the
group velocity. Reducing the waveguide thickness diminishes the
experimental demonstration of an operational spin-wave circuit
group velocity. Smaller devices also require the utilization of back-
based on a flexible scheme for circuit design.
ward volume spin waves with shorter wavelengths, with complex
effects on the group velocity. Reaching the exchange regime can be
B. Transducer efficiency advantageous since it reduces the anisotropy of the spin-wave disper-
sion and increases the group velocity. However, the high frequencies
A major limitation for all applications of spin waves at the
of exchange spin waves in large-Ms ferromagnetic materials may
nanoscale is the energy efficiency of spin-wave generation and
impose severe conditions on mixed-signal periphery circuits.
detection. While large mm-scale antennas and magnetic wave-
The benchmarking of hybrid spin-wave–CMOS systems has
guides can be efficient to transfer electrical energy into ferromag-
indicated that the possibility to design compact majority gates can
netic resonance and the spin-wave system, the radiated power and
lead to significant area gains with respect to CMOS circuits. In
the efficiency decrease with the magnetic excitation volume. Hence,
practice, the benchmark suggests that competitive areas can already
energy-efficient nanoscale spin-wave transducers are still lacking.
be achieved for characteristic dimensions (i.e., waveguide width) of
From a systems point of view, the relevant energy is the external
the spin-wave circuit of about 50 nm. Such dimensions have been
electric energy needed to excite spin waves and not the energy of
reached experimentally recently.147 This indicates that scaling the
the spin waves themselves. Hence, the transducer efficiency is a key
spin-wave wavelength and the device dimensions should not be a
property for ultralow-power applications of spin-wave computing
major roadblock. However, the scalability of spin-wave devices may
systems. Magnetoelectric transducers currently appear to be most
be ultimately limited by other effects, such as the dipolar crosstalk
promising. However, energy-efficient spin-wave excitation by mag-
or transducer efficiency.331
netoelectric transducers has not been demonstrated experimentally.
Moreover, research of magnetoelectric devices at the nanoscale and
at GHz frequencies is only starting. The physics of the magneto- D. High-throughput computation
electric coupling in nanoscale spin-wave transducers is not yet fully To date, experimental spin-wave logic gates have been oper-
established and is expected to be complicated by the complex ated in the frequency domain using vector network analyzers.
acoustic response of “real” nonideal devices.259 Here, a major In real applications, however, the devices have to be operated in the
breakthrough would be the demonstration of a scaled (or scalable) time domain. For cascading by nanomagnets, clocking schemes
efficient spin-wave transducer based on a magnetoelectric com- enable time-domain operation but still remain to be developed and
pound material. benchmarked. Moreover, input–output isolation may be a challenge

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-35

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

for such schemes. All-spin-wave cascading schemes may require close collaboration between researchers in spin-wave physics as
the use of spin-wave wave packets or solitons. While the time- well as device and circuit design. Physics-based compact models of
domain response of spin-wave transmission can be studied via the spin-waves devices and transducers401 may enable circuit simulation,
Fourier transform of the spectral response, excitation, interference, periphery design, and ultimately the refinement of the benchmarking
dephasing, and detection of wave packets are not fully understood procedure to embolden the promises of spin-wave technology. We
and remain to be studied experimentally. Electric crosstalk between hope that this Tutorial can be a keystone in establishing this collabo-
transducers is a major issue for nanoscale spin-wave devices due to ration and contribute to the realization of the exciting prospect of a
the low efficiency of spin-wave generation and detection. More effi- competitive hybrid spin-wave–CMOS computing technology.
cient transducers may facilitate such experiments. A major break-
through would be a time-resolved spin-wave transmission ACKNOWLEDGMENTS
experiment with phase sensitivity. Note that high-throughput
applications require single pulse operation. This work has been funded by the European Union’s Horizon
2020 research and innovation program within the FET-OPEN
E. CMOS periphery circuits Project CHIRON under Grant Agreement No. 801055. It has also
been partially supported by imec’s industrial affiliate program on
In hybrid spin-wave–CMOS systems, spin-wave circuits are beyond-CMOS logic. F.V. acknowledges financial support from the
embedded in mixed-signal CMOS-based periphery circuits that Research Foundation—Flanders (FWO) through Grant No. 1S05719N.
provide a link with cache/memory and input/output devices. A.V.C. acknowledges financial support from the European Research
However, only very few studies have been reported on concrete Council within the Starting Grant No. 678309, MagnonCircuits. The
periphery circuits.65,172,386,388 The design of periphery circuits is authors would like to thank Odysseas Zografos (imec), Davide Tierno
currently hindered by the lack of equivalent circuit models for spin- (imec), Giacomo Talmelli (KU Leuven, imec), Kevin Garello (imec),
wave devices and transducers. The development of calibrated Hasnain Ahmad (imec), Diogo Costa (imec), Xiao Sun (imec), Iuliana
compact models401 for a complete set of spin-wave devices and Radu (imec), Inge Asselberghs (imec), Bart Sorée (imec), Ian Young
transducers is thus a key first step toward the development of low- (Intel), Dmitri Nikonov (Intel), Philipp Pirro (TU Kaiserslautern),
power periphery circuits and complete hybrid systems. This is an Burkard Hillebrands (TU Kaiserslautern), Thibaut Devolder (U Paris
important conditio sine qua non for an accurate benchmark of the Sud), Silvia Matzen (U Paris Sud), Umesh Bhaskar (U Paris Sud),
performance of hybrid spin-wave–CMOS systems and ultimately Madjid Anane (CNRS), Paolo Bortolotti (Thales), Matthijn Dekkers
for a final assessment of their potential in commercial applications. (Solmates), Thomas Aukes (Solmates), George Konstantinidis
(FORTH Heraklion), Alexandru Muller (IMT Bucharest), Oleksandr
F. New materials Dobrovolskiy (U Vienna), and Azad Naeemi (Georgia Tech) for many
Spin-wave computing is also an interesting field for material valuable discussions.
scientists. Many spin-wave experiments have been performed using
single-crystal YIG. Epitaxy of high quality YIG on Si (100) has not DATA AVAILABILITY
been demonstrated, and thus YIG is incompatible with integration The data that support the findings of this study are available
alongside CMOS. Ferromagnetic metals, such as CoFeB or permal- within the article.
loy, are routinely integrated in MRAM memory cells and are com-
patible with Si technology. Nonetheless, insulating ferrites remain
REFERENCES
an interesting alternative since they typically show lower losses at
1
microwave frequencies. However, thin ferrite films with low R. L. Boylestad and L. Nashelsky, Electronic Devices and Circuit Theory, 11th
damping that can be cointegrated with Si-based CMOS still have to ed. (Pearson, Harlow, 2014).
2
be demonstrated. S. M. Sze and K. K. Ng, Physics of Semiconductor Devices, 3rd ed. (Wiley,
Hoboken, 2006).
Magnetoelectric compound materials are also a fascinating 3
B. Lojek, History of Semiconductor Engineering (Springer, Berlin, 2007).
research field in material science. Challenges include the combina- 4
E.-H. Jiang and W.-B. Jiang, Circuits Syst. Signal Process. 38, 5817 (2019).
tion of Pb-free high-performance piezoelectrics and ferromagnets 5
J. Uyemura, CMOS Logic Circuit Design (Kluwer, New York, 2007).
with large magnetostriction coefficients and low damping. In par- 6
R. H. Dennard, F. H. Gaensslen, V. L. Rideout, E. Bassous, and A. R. LeBlanc,
ticular, the piezoelectric response at GHz frequencies is often IEEE J. Solid-State Circuits 9, 256 (1974).
limited due to dielectric and ferroelectric relaxation, although some 7
G. E. Moore, Proc. IEEE 86, 82 (1998).
progress has recently been reported.479 8
“International technology roadmap for semiconductors,” see http://www.itrs2.
The above discussion indicates that many obstacles still exist net (last accessed 31 May 2020).
before spin-wave technology can lead to competitive computing
9
M. Bohr, Proc. SPIE 7974, 797402 (2011).
systems. Initial benchmarking has, however, clearly established
10
K. J. Kuhn, IEEE Trans. Electron Devices 59, 1813–1828 (2012).
11
S. P. Murarka and S. W. Hymes, Crit. Rev. Solid State Mater. Sci. 20, 87
the promise of such a technology for ultralow-power electronics.
(1995).
The large-scale effort in magnonic research will certainly advance 12
M. Bohr, R. Chau, T. Ghani, and K. Mistry, IEEE Spectr. 44, 29 (2007).
the state of the art further in the near future. Hence, one can 13
C. Auth, C. Allen, A. Blattner, D. Bergstrom, M. Brazier, M. Bost, M. Buehler,
anticipate that spin-wave circuits will become a reality in the next V. Chikarmane, T. Ghani, T. Glassman, R. Grover, W. Han, D. Hanken,
decade. The remaining obstacles relate to their embedding into M. Hattendorf, P. Hentges, R. Heussner, J. Hicks, D. Ingerly, P. Jain, S. Jaloviar,
the CMOS periphery, including transduction. This field requires R. James, D. Jones, J. Jopling, S. Joshi, C. Kenyon, H. Liu, R. McFadden,

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-36

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

B. McIntyre, J. Neirynck, C. Parker, L. Pipes, I. Post, S. Pradhan, M. Prince, 46

K. C. Hall and M. E. Flatté, Appl. Phys. Lett. 88, 162503 (2006).
S. Ramey, T. Reynolds, J. Roesler, J. Sandford, J. Seiple, P. Smith, C. Thomas, 47
H. Dery, P. Dalal, Ł. Cywiński, and L. J. Sham, Nature 447, 573 (2007).
D. Towner, T. Troeger, C. Weber, P. Yashar, K. Zawadzki, and K. Mistry, in 2012 48
R. P. Cowburn and M. E. Welland, Science 287, 1466 (2000).
IEEE Symposium on VLSI Technology (IEEE, 2012), pp. 131–132. 49
A. Ney, C. Pampuch, R. Koch, and K. H. Ploog, Nature 425, 485 (2003).
14
M. M. Waldrop, Nat. News 530, 144 (2016). 50
A. Imre, G. Csaba, L. Ji, A. Orlov, G. H. Bernstein, and W. Porod, Science 311,
15
D. Mamaluy and X. Gao, Appl. Phys. Lett. 106, 193503 (2015). 205 (2006).
16
B. Hoefflinger, Chips 2020: A Guide to the Future of Nanoelectronics (Springer, 51
B. Behin-Aein, D. Datta, S. Salahuddin, and S. Datta, Nat. Nanotechnol. 5, 266
Berlin, 2012). (2010).
17 52
N. Z. Haron and S. Hamdioui, in 2008 3rd International Design Test S. Manipatruni, D. E. Nikonov, C.-C. Lin, T. A. Gosavi, H. Liu, B. Prasad,
Workshop (IEEE, 2008), pp. 98–103. Y.-L. Huang, E. Bonturim, R. Ramesh, and I. A. Young, Nature 565, 35 (2019).
18
“International roadmap for devices and systems,” see https://irds.ieee.org (last 53
D. A. Allwood, G. Xiong, C. C. Faulkner, D. Atkinson, D. Petit, and
accessed 25 May 2020). R. P. Cowburn, Science 309, 1688 (2005).
19
D. G. Feitelson, Optical Computing (MIT Press, Cambridge, 1988). 54
P. Xu, K. Xia, C. Gu, L. Tang, H. Yang, and J. Li, Nat. Nanotechnol. 3, 97
20
N. Streibl, K.-H. Brenner, A. Huang, J. Jahns, J. Jewell, A. Lohmann, D. Miller, (2008).
M. Murdocca, M. Prise, and T. Sizer, Proc. IEEE 77, 1954 (1989). 55
D. E. Nikonov, G. I. Bourianoff, and T. Ghani, IEEE Electron Device Lett. 32,
21
J. A. Hutchby, G. I. Bourianoff, V. V. Zhirnov, and J. E. Brewer, IEEE Circuits 1128 (2011).
Devices Mag. 18, 28 (2002). 56
X. Zhang, M. Ezawa, and Y. Zhou, Sci. Rep. 5, 9400 (2015).
22
G. Bourianoff, Computer 36, 44 (2003). 57
K. Koumpouras, D. Yudin, C. Adelmann, A. Bergman, O. Eriksson, and
23
G. I. Bourianoff, P. A. Gargini, and D. E. Nikonov, Solid-State Electron. 51, M. Pereiro, J. Phys.: Condens. Matter 30, 375801 (2018).
58
1426 (2007). M. P. Kostylev, A. A. Serga, T. Schneider, B. Leven, and B. Hillebrands, Appl.
24
D. A. B. Miller, Nat. Photonics 4, 3 (2010). Phys. Lett. 87, 153501 (2005).
25
K. Bernstein, R. K. Cavin, W. Porod, A. Seabaugh, and J. Welser, Proc. IEEE 59
A. Khitun and K. L. Wang, J. Appl. Phys. 110, 034306 (2011).
98, 2169 (2010). 60
A. V. Chumak, V. I. Vasyuchka, A. A. Serga, and B. Hillebrands, Nat. Phys.
26
D. E. Nikonov and I. A. Young, Proc. IEEE 101, 2498 (2013). 11, 453 (2015).
27
D. E. Nikonov and I. A. Young, J. Mater. Res. 29, 2109 (2014). 61
A. V. Chumak, A. A. Serga, and B. Hillebrands, Nat. Commun. 5, 4700
28
D. E. Nikonov and I. A. Young, IEEE J. Explor. Solid-State Computat. 1, 3 (2014).
(2015). 62
K.-S. Lee and S.-K. Kim, J. Appl. Phys. 104, 053909 (2008).
29 63
T. M. Mitchell, Machine Learning (McGraw-Hill, Boston, 1997). O. Zografos, M. Manfrini, A. Vaysset, B. Sorée, F. Ciubotaru, C. Adelmann,
30
K. P. Murphy, Machine Learning: A Probabilistic Perspective (MIT Press, R. Lauwereins, P. Raghavan, and I. P. Radu, Sci. Rep. 7, 12154 (2017).
64
Cambridge, 2012). I. P. Radu, O. Zografos, A. Vaysset, F. Ciubotaru, J. Yan, J. Swerts, D. Radisic,
31
E. Alpaydin, Introduction to Machine Learning, 4th ed. (MIT Press, B. Briggs, B. Soré, M. Manfrini, M. Ercken, C. Wilson, P. Raghavan, S. Sayan,
Cambridge, 2020). C. Adelmann, A. Thean, L. Amarù, P.-E. Gaillardon, G. D. Micheli,
32
K. Boucart and A. M. Ionescu, IEEE Trans. Electron Devices 54, 1725 (2007). D. E. Nikonov, S. Manipatruni, and I. A. Young, in 2015 IEEE International
33
D. S. Jeong, K. M. Kim, S. Kim, B. J. Choi, and C. S. Hwang, Adv. Electron. Electron Devices Meeting (IEDM) (IEEE, 2015), p. 32.5.
Mater. 2, 1600090 (2016). 65
O. Zografos, A. Vaysset, B. Sorée, and P. Raghavan, in Beyond-CMOS
34
Y. Li, Z. Wang, R. Midya, Q. Xia, and J. J. Yang, J. Phys. D Appl. Phys. 51, Technologies for Next Generation Computer Design, edited by R. O. Topaloglu
503002 (2018). and H.-S. P. Wong (Springer, Cham, 2019), Chap. 7, pp. 231–262.
35
S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J. M. Daughton, S. von Molnár, 66
D. Hampel and R. O. Winder, IEEE Spectr. 8, 32 (1971).
M. L. Roukes, A. Y. Chtchelkanova, and D. M. Treger, Science 294, 1488 (2001). 67
L. Amarù, P.-E. Gaillardon, and G. De Micheli, Proceedings of 51st Design
36
C. Felser and G. H. Fecher, Spintronics: From Materials to Devices (Springer, Automation Conference (DAC) (IEEE, 2014), p. 1.
Dordrecht, Heidelberg, 2013). 68
L. Amarù, P.-E. Gaillardon, S. Mitra, and G. De Micheli, Proc. IEEE 103, 2168
37
S. Bandyopadhyay and M. Cahay, Introduction to Spintronics, 2nd ed. (CRC (2015).
Press, Boca Raton, 2015). 69
V. V. Kruglyak, S. O. Demokritov, and D. Grundler, J. Phys. D Appl. Phys. 43,
38
Handbook of Spintronics, edited by Y. Xu, D. D. Awschalom, and J. Nitta 264001 (2010).
(Springer, Dordrecht, 2016). 70
A. A. Serga, A. V. Chumak, and B. Hillebrands, J. Phys. D Appl. Phys. 43,
39
D. Sander, S. O. Valenzuela, D. Makarov, C. H. Marrows, E. E. Fullerton, 264002 (2010).
P. Fischer, J. McCord, P. Vavassori, S. Mangin, P. Pirro, B. Hillebrands, 71
B. Lenk, H. Ulrichs, F. Garbs, and M. Münzenberg, Phys. Rep. 507, 107
A. D. Kent, T. Jungwirth, O. Gutfleisch, C. G. Kim, and A. Berger, J. Phys. D (2011).
Appl. Phys. 50, 363001 (2017). 72
Magnonics: From Fundamentals to Applications, edited by S. O. Demokritov
40
B. Dieny, I. L. Prejbeanu, K. Garello, P. Gambardella, P. Freitas, R. Lehndorff, and A. N. Slavin (Springer, Berlin, 2013).
W. Raberg, U. Ebels, S. O. Demokritov, J. Akerman, A. Deac, P. Pirro, 73
M. Krawczyk and D. Grundler, J. Phys.: Condens. Matter 26, 123202 (2014).
74
C. Adelmann, A. Anane, A. V. Chumak, A. Hiroata, S. Mangin, A. D. Karenowska, A. V. Chumak, A. A. Serga, and B. Hillebrands, in
S. O. Valenzuela, M. C. Onbasli, M. d’Aquino, G. Prenat, G. Finocchio, Handbook of Spintronics, edited by Y. Xu, D. D. Awschalom, and J. Nitta
L. L. Diaz, R. Chantrell, O. Chubykalo-Fesenko, and P. Bortolotti, Nat. Electron. (Springer, Dordrecht, 2016), Chap. 37, p. 1505.
3, 446 (2020). 75
G. Csaba, Á. Papp, and W. Porod, Phys. Lett. A 381, 1471 (2017).
41 76
Nanomagnetic and Spintronic Devices for Energy-Efficient Memory and Three-Dimensional Magnonics: Layered, Micro- and Nanostructures, edited by
Computing, edited by J. Atulasimha and S. Bandyopadhyay (Wiley, Chichester, G. Gubbiotti (Jenny Stanford Publishing, Singapore, 2019).
77
2016). D. D. Stancil and A. Prabhakar, Spin Waves (Springer, New York, 2009).
42
S. Manipatruni, D. E. Nikonov, and I. A. Young, Nat. Phys. 14, 338 (2018). 78
F. Bloch, Z. Phys. 61, 206 (1930).
43
I. Žutić, J. Fabian, and S. Das Sarma, Rev. Mod. Phys. 76, 323 (2004). 79
C. Herring and C. Kittel, Phys. Rev. 81, 869 (1951).
44
A. A. Khajetoorians, J. Wiebe, B. Chilian, and R. Wiesendanger, Science 332, 80
A. Khitun, J. Appl. Phys. 111, 054307 (2012).
1062 (2011). 81
W. J. Gallagher and S. S. P. Parkin, IBM J. Res. Devel. 50, 5 (2006).
45
S. Datta and B. Das, Appl. Phys. Lett. 56, 665 (1990). 82
C. Chappert, A. Fert, and F. Nguyen Van Dau, Nat. Mater. 6, 813 (2007).

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-37

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

83
A. D. Kent and D. C. Worledge, Nat. Nanotechnol. 10, 187 (2015). 120
X. Liu, W. Zhang, M. J. Carter, and G. Xiao, J. Appl. Phys. 110, 033910
84
B. Dieny, R. B. Goldfarb, and K.-J. Lee, Introduction to Magnetic (2011).
121
Random-Access Memory (Wiley, Piscataway, 2016). A. Conca, J. Greser, T. Sebastian, S. Klingler, B. Obry, B. Leven, and
85
S. Bhatti, R. Sbiaa, A. Hirohata, H. Ohno, S. Fukami, and B. Hillebrands, J. Appl. Phys. 113, 213909 (2013).
S. N. Piramanayagam, Mater. Today 20, 530 (2017). 122
C. Liu, C. Mewes, M. Chshiev, T. Mewes, and W. Butler, Appl. Phys. Lett. 95,
86
J. Lenz, Proc. IEEE 78, 973 (1990). 022509 (2009).
87
S. Parkin, X. Jiang, C. Kaiser, A. Panchula, K. Roche, and M. Samant, Proc. 123
S. Trudel, O. Gaier, J. Hamrle, and B. Hillebrands, J. Phys. D Appl. Phys. 43,
IEEE 91, 661 (2003). 193001 (2010).
88
P. Ripka and M. Janosek, IEEE Sens. J. 10, 1108 (2010). 124
M. Oogane, T. Kubota, Y. Kota, S. Mizukami, H. Naganuma, A. Sakuma, and
89
H. Danan, A. Herr, and A. J. P. Meyer, J. Appl. Phys. 39, 669 (1968). Y. Ando, Appl. Phys. Lett. 96, 252501 (2010).
90
Y. K. Kim and M. Oliveria, J. Appl. Phys. 74, 1233 (1993). 125
T. Sebastian, Y. Ohdaira, T. Kubota, P. Pirro, T. Brächer, K. Vogt, A. Serga,
91
M. Kin, H. Kura, M. Tanaka, Y. Hayashi, J. Hasaegawa, and T. Ogawa, H. Naganuma, M. Oogane, Y. Ando, and B. Hillebrands, Appl. Phys. Lett. 100,
J. Appl. Phys. 117, 17E714 (2015). 112402 (2012).
92 126
P. Vavassori, D. Bisero, F. Carace, A. di Bona, G. C. Gazzadi, M. Liberati, and F. Brailsford, Physical Principles of Magnetism (Van Nostrand, London,
S. Valeri, Phys. Rev. B 72, 054405 (2015). New York, 1966).
93
J. Bishop, IEEE Trans. Magn. 13, 1638 (1977). 127
A. G. Gurevich and G. A. Melkov, Magnetization Oscillations and Waves
94
M. J. Aus, C. Cheung, B. Szpunar, U. Erb, and J. Szpunar, J. Mater. Sci. 17, (CRC Press, Boca Raton, 1996).
128
1949 (1998). S. Chikazumi, Physics of Ferromagnetism (Oxford University Press, Oxford,
95
S. Vernon, S. Lindsay, and M. B. Stearns, Phys. Rev. B 29, 4439 (1984). New York, 1997).
96
A. Michels, J. Weissmüller, A. Wiedenmann, J. S. Pedersen, and J. Barker, 129
L. Landau and E. Lifshitz, Phys. Z. Sowjetunion 8, 101 (1935).
Philos. Mag. 80, 785 (2000). 130
T. L. Gilbert, IEEE Trans. Magn. 40, 3443 (2004).
97
M. A. W. Schoen, D. Thonig, M. L. Schneider, T. J. Silva, H. T. Nembach, 131
B. A. Kalinikos and A. N. Slavin, J. Phys. C Solid State Phys. 19, 7013 (1986).
O. Eriksson, O. Karis, and J. M. Shaw, Nat. Phys. 12, 839 (2016). 132
K. Y. Guslienko and A. N. Slavin, Phys. Rev. B 72, 014463 (2005).
98 133
B. Heinrich, J. Cochran, M. Kowalewski, J. Kirschner, Z. Celinski, A. Arrott, Q. Wang, B. Heinz, R. Verba, M. Kewenig, P. Pirro, M. Schneider, T. Meyer,
and K. Myrtle, Phys. Rev. B 44, 9348 (1991). B. Lägel, C. Dubs, T. Brächer, and A. Chumak, Phys. Rev. Lett. 122, 247202 (2019).
99
L. Sun, Y. Hao, C.-L. Chien, and P. C. Searson, IBM J. Res. Devel. 49, 79 134
J. R. Eshbach and R. W. Damon, Phys. Rev. 118, 1208 (1960).
135
(2005). M. G. Cottam, Linear and Nonlinear Spin Waves in Magnetic Films and
100
P. Talagala, P. S. Fodor, D. Haddad, R. Naik, L. E. Wenger, P. P. Vaishnava, Superlattices (World Scientific, Singapore, 1994).
and V. M. Naik, Phys. Rev. B 66, 144426 (2002). 136
P. E. Wigen, Nonlinear Phenomena and Chaos in Magnetic Materials (World
101
A. Michels, J. Appl. Phys. 87, 5953 (2000). Scientific, Singapore, 1994).
102
J. Walowski, M. D. Kaufmann, B. Lenk, C. Hamann, J. McCord, and 137
H. Suhl, Proc. IRE 44, 1270 (1956).
M. Münzenberg, J. Phys. D Appl. Phys. 41, 16 (2008). 138
H. Suhl, J. Phys. Chem. Solids 1, 209 (1957).
103
V. Cherepanov, I. Kolokolov, and V. L’vov, Phys. Rep. 229, 81 (1993). 139
V. E. Zakharov, V. S. L’vov, and S. S. Starobinets, Sov. Phys.-Usp. 17, 896
104
H. L. Glass, Proc. IEEE 76, 151 (1988). (1975).
105
S. Geller and M. A. Gilleo, Acta Crystallogr. 10, 239 (1957). 140
V. S. L’vov, Wave Turbulence Under Parametric Excitation (Springer, Berlin,
106
S. Klingler, A. Chumak, T. Mewes, B. Khodadadi, C. Mewes, C. Dubs, 1994).
O. Surzhenko, B. Hillebrands, and A. Conca, J. Phys. D Appl. Phys. 48, 015001 141
P. Krivosik and C. E. Patton, Phys. Rev. B 82, 184428 (2010).
142
(2015). A. A. Serga, A. V. Chumak, A. André, G. A. Melkov, A. N. Slavin,
107
C. R. Serrao, J. R. Sahu, K. Ramesha, and C. N. R. Rao, J. Appl. Phys. 104, S. O. Demokritov, and B. Hillebrands, Phys. Rev. Lett. 99, 227202 (2007).
143
016102 (2008). Q. Wang, M. Kewenig, M. Schneider, R. Verba, F. Kohl, B. Heinz, M. Geilen,
108
P. Pirro, T. Brächer, A. V. Chumak, B. Lägel, C. Dubs, O. Surzhenko, M. Mohseni, B. Lägel, F. Ciubotaru, C. Adelmann, C. Dubs, S. D. Cotofana,
P. Görnert, B. Leven, and B. Hillebrands, Appl. Phys. Lett. 104, 012402 (2014). O. V. Dobrovolskiy, T. Brächer, P. Pirro, and A. V. Chumak, “A magnonic direc-
109
C. Hahn, V. V. Naletov, G. de Loubens, O. Klein, O. d’Allivy Kelly, tional coupler for integrated magnonic half-adders,” Nat. Electron. (2020) (in press).
144
A. Anane, R. Bernard, E. Jacquet, P. Bortolotti, V. Cros, J. L. Prieto, and P. Pirro, T. Sebastian, T. Brächer, A. Serga, T. Kubota, H. Naganuma,
M. Muñoz, Appl. Phys. Lett. 104, 152410 (2014). M. Oogane, Y. Ando, and B. Hillebrands, Phys. Rev. Lett. 113, 227601 (2014).
110
C. Dubs, O. Surzhenko, R. Linke, A. Danilewsky, U. Brückner, and J. Dellith, 145
T. Brächer, P. Pirro, and B. Hillebrands, Phys. Rep. 699, 1 (2017).
J. Phys. D Appl. Phys. 50, 204005 (2017). 146
C. L. Ordóñez-Romero, B. A. Kalinikos, P. Krivosik, W. Tong, P. Kabos, and
111
Y. Sun, Y.-Y. Song, H. Chang, M. Kabatek, M. Jantz, W. Schneider, M. Wu, C. E. Patton, Phys. Rev. B 79, 144428 (2009).
H. Schultheiss, and A. Hoffmann, Appl. Phys. Lett. 101, 152405 (2012). 147
B. Heinz, T. Brächer, M. Schneider, Q. Wang, B. Lägel, A. M. Friedel,
112
H. Yu, O. d’Allivy Kelly, V. Cros, R. Bernard, P. Bortolotti, A. Anane, D. Breitbach, S. Steinert, T. Meyer, M. Kewenig, C. Dubs, P. Pirro, and
F. Brandl, R. Huber, I. Stasinopoulos, and D. Grundler, Sci. Rep. 4, 6848 (2014). A. V. Chumak, Nano Lett. 20, 4220 (2020).
113 148
M. C. Onbasli, A. Kehlberger, D. H. Kim, G. Jakob, M. Kläui, A. V. Chumak, M. B. Jungfleisch, A. V. Chumak, A. Kehlberger, V. Lauer, D. H. Kim,
B. Hillebrands, and C. A. Ross, APL Mater. 2, 106102 (2014). M. C. Onbasli, C. A. Ross, M. Kläui, and B. Hillebrands, Phys. Rev. B 91, 134407
114
T. Liu, H. Chang, V. Vlaminck, Y. Sun, M. Kabatek, A. Hoffmann, L. Deng, (2015).
and M. Wu, J. Appl. Phys. 115, 17A501 (2014). 149
M. Mohseni, Q. Wang, B. Heinz, M. Kewenig, M. Schneider, F. Kohl,
115
V. E. Demidov and S. O. Demokritov, IEEE Trans. Magn. 51, 0800215 B. Lägel, C. Dubs, A. V. Chumak, and P. Pirro, arXiv:2006.03400 (2020).
150
(2015). D. A. Patterson and J. L. Hennessy, Computer Organization and Design: The
116
S. S. Kalarickal, P. Krivosik, M. Wu, C. E. Patton, M. L. Schneider, P. Kabos, Hardware/Software Interface, 4th ed. (Morgan Kaufmann, Waltham, 2011).
T. J. Silva, and J. Nibarger, J. Appl. Phys. 99, 093909 (2006). 151
A. Khitun, M. Bao, and K. L. Wang, Superlatt. Microstruct. 47, 464 (2010).
117
T. Sebastian, K. Schultheiss, B. Obry, B. Hillebrands, and H. Schultheiss, 152
F. Macià, A. D. Kent, and F. C. Hoppensteadt, Nanotechnology 22, 095301
Front. Phys. 3, 35 (2015). (2011).
118
C. E. Patton, J. Appl. Phys. 39, 3060 (1968). 153
A. Kozhevnikov, F. Gertz, G. Dudko, Y. Filimonov, and A. Khitun, Appl.
119
A. Brunsch, J. Appl. Phys. 50, 7603 (1979). Phys. Lett. 106, 142409 (2015).

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-38

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

154
H. Arai and H. Imamura, Phys. Rev. Appl. 10, 024040 (2018). 186
A. Khitun, J. Appl. Phys. 113, 164503 (2013).
155
R. Nakane, G. Tanaka, and A. Hirose, IEEE Access 6, 4462 (2018). 187
F. Gertz, A. V. Kozhevnikov, Y. A. Filimonov, D. E. Nikonov, and A. Khitun,
156
G. Tanaka, T. Yamane, J. B. Héroux, R. Nakane, N. Kanazawa, S. Takeda, IEEE J. Explor. Solid-State Computat. 1, 67 (2015).
H. Numata, D. Nakano, and A. Hirose, Neural Netw. 115, 100 (2019). 188
A. K. Sharma, Advanced Semiconductor Memories: Architectures, Designs,
157
S. Watt and M. Kostylev, Phys. Rev. Appl. 13, 034057 (2020). and Applications (Wiley–IEEE Press, Piscataway, Hoboken, 2002).
158
P. Shabadi, S. N. Rajapandian, S. Khasanvis, and C. A. Moritz, SPIN 2, 189
Emerging Non-Volatile Memories, edited by S. Hong, O. Auciello, and
1240006 (2012). D. Wouters (Springer, New York, 2014).
159 190
M. Collet, O. Gladii, M. Evelt, V. Bessonov, L. Soumah, P. Bortolotti, Y. Zhang, T. Yu, J. Lei Chen, Y. Guang Zhang, J. Feng, S. Tu, and H. Yu,
S. O. Demokritov, Y. Henry, V. Cros, M. Bailleul, V. E. Demidov, and A. Anane, J. Magn. Magn. Mater. 450, 24 (2018).
Appl. Phys. Lett. 110, 092408 (2017). 191
Y. Au, E. Ahmad, O. Dmytriiev, M. Dvornik, T. Davison, and V. V. Kruglyak,
160
M. Born and E. Wolf, Principles of Optics, 7th ed. (Cambridge University Appl. Phys. Lett. 100, 182404 (2012).
Press, Cambridge, 1999). 192
Y. Au, M. Dvornik, O. Dmytriiev, and V. V. Kruglyak, Appl. Phys. Lett. 100,
161
M. J. Donahue and D. G. Porter, “OOMMF user’s guide, version 1.0,” 172408 (2012).
Technical Report 6376, National Institute of Standards and Technology, 193
M. Bailleul, D. Olligs, and C. Fermon, Appl. Phys. Lett. 83, 972 (2003).
Gaithersburg, 1999. 194
V. Vlaminck and M. Bailleul, Phys. Rev. B 81, 014425 (2010).
162 195
A. Vansteenkiste, J. Leliaert, M. Dvornik, M. Helsen, F. Garcia-Sanchez, and J. H. Kwon, S. S. Mukherjee, M. Jamali, M. Hayashi, and H. Yang, Appl.
B. Van Waeyenberge, AIP Adv. 4, 107133 (2014). Phys. Lett. 99, 132505 (2011).
163 196
S. Klingler, P. Pirro, T. Brächer, B. Leven, B. Hillebrands, and A. V. Chumak, R. Huber, M. Krawczyk, T. Schwarze, H. Yu, G. Duerr, S. Albert, and
Appl. Phys. Lett. 105, 152410 (2014). D. Grundler, Appl. Phys. Lett. 102, 012403 (2013).
164
S. Klingler, P. Pirro, T. Brächer, B. Leven, B. Hillebrands, and A. V. Chumak, 197
N. Sato, N. Ishida, T. Kawakami, and K. Sekiguchi, Appl. Phys. Lett. 104,
Appl. Phys. Lett. 106, 212406 (2015). 032411 (2014).
165
A. Khitun, M. Bao, and K. L. Wang, IEEE Trans. Magn. 44, 2141 (2008). 198
F. Ciubotaru, T. Devolder, M. Manfrini, C. Adelmann, and I. P. Radu, Appl.
166
T. Brächer, F. Heussner, P. Pirro, T. Meyer, T. Fischer, M. Geilen, B. Heinz, Phys. Lett. 109, 012403 (2016).
B. Lägel, A. A. Serga, and B. Hillebrands, Sci. Rep. 6, 38235 (2016). 199
U. K. Bhaskar, G. Talmelli, F. Ciubotaru, C. Adelmann, and T. Devolder,
167
K. Ganzhorn, S. Klingler, T. Wimmer, S. Geprägs, R. Gross, H. Huebl, and J. Appl. Phys. 127, 033902 (2020).
S. T. B. Goennenwein, Appl. Phys. Lett. 109, 022405 (2016). 200
J. Slonczewski, J. Magn. Magn. Mater. 159, L1 (1996).
168
O. Zografos, S. Dutta, M. Manfrini, A. Vaysset, B. Sorée, A. Naeemi, 201
L. Berger, Phys. Rev. B 54, 9353 (1996).
P. Raghavan, R. Lauwereins, and I. P. Radu, AIP Adv. 7, 056020 (2017). 202
M. Tsoi, A. G. M. Jansen, J. Bass, W.-C. Chiang, M. Seck, V. Tsoi, and
169
A. Mahmoud, F. Vanderveken, C. Adelmann, F. Ciubotaru, S. Hamdioui, and P. Wyder, Phys. Rev. Lett. 80, 4281 (1998).
S. Cotofana, AIP Adv. 10, 035119 (2020). 203
J. A. Katine, F. J. Albert, R. A. Buhrman, E. B. Myers, and D. C. Ralph, Phys.
170
L. Amarù, P.-E. Gaillardon, and G. De Micheli, in 2015 52nd ACM/EDAC/ Rev. Lett. 84, 3149 (2000).
IEEE Design Automation Conference (DAC) (IEEE, 2015), pp. 1–6. 204
S. Urazhdin, N. O. Birge, W. P. Pratt, and J. Bass, Phys. Rev. Lett. 91, 146803
171
L. Amarù, P.-E. Gaillardon, and G. De Micheli, IEEE Trans. Comput. Aided (2003).
Des. Integr. Circuits Syst. 35, 806 (2016). 205
K.-J. Lee, A. Deac, O. Redon, J.-P. Nozières, and B. Dieny, Nat. Mater. 3, 877
172
O. Zografos, P. Raghavan, L. Amarù, B. Sorée, R. Lauwereins, I. Radu, (2004).
206
D. Verkest, and A. Thean, in 2014 IEEE/ACM International Symposium on D. Houssameddine, U. Ebels, B. Delaët, B. Rodmacq, I. Firastrau,
Nanoscale Architectures (NANOARCH) (IEEE, 2014), pp. 25–30. F. Ponthenier, M. Brunet, C. Thirion, J.-P. Michel, L. Prejbeanu-Buda,
173
I. P. Radu, O. Zografos, A. Vaysset, F. Ciubotaru, M. Manfrini, P. Raghavan, M.-C. Cyrille, O. Redon, and B. Dieny, Nat. Mater. 6, 447 (2007).
207
S. Sayan, C. Adelmann, Z. Tőkei, and A. Thean, in 2016 IEEE International A. Ruotolo, V. Cros, B. Georges, A. Dussaux, J. Grollier, C. Deranlot,
Interconnect Technology Conference/Advanced Metallization Conference (IITC/ R. Guillemet, K. Bouzehouane, S. Fusil, and A. Fert, Nat. Nanotechnol. 4, 528
AMC) (IEEE, 2016), pp. 51–52. (2009).
174
R. Lucas, M. P. C. Fossorier, Y. Kou, and S. Lin, IEEE Trans. Commun. 48, 208
V. E. Demidov, S. Urazhdin, and S. O. Demokritov, Nat. Mater. 9, 984
931 (2000). (2010).
175
R. Palanki, M. P. C. Fossorier, and J. S. Yedidia, IEEE Trans. Commun. 55, 209
M. Madami, S. Bonetti, G. Consolo, S. Tacchi, G. Carlotti, G. Gubbiotti,
1099 (2007). F. B. Mancoff, M. A. Yar, and J. Åkerman, Nat. Nanotechnol. 6, 635
176
H. Wei, Z. Wang, X. Tian, M. Käll, and H. Xu, Nat. Commun. 2, 387 (2011). (2011).
177
Y. Fu, X. Hu, C. Lu, S. Yue, H. Yang, and Q. Gong, Nano Lett. 12, 5784 210
B. Divinskiy, V. E. Demidov, S. Urazhdin, R. Freeman, A. B. Rinkevich, and
(2012). S. O. Demokritov, Adv. Mater. 30, 1802837 (2018).
178 211
S. Lal, J. H. Hafner, N. J. Halas, S. Link, and P. Nordlander, Acc. Chem. Res. Spin Dynamics in Confined Magnetic Structures III, edited by B. Hillebrands
45, 1887 (2012). and A. Thiaville, Topics in Applied Physics (Springer, Berlin, 2006), Vol. 101.
179 212
S. Dutta, O. Zografos, S. Gurunarayanan, I. Radu, B. Soree, F. Catthoor, and J. Grollier, V. Cros, H. Jaffrès, A. Hamzic, J. M. George, G. Faini, J. Ben
A. Naeemi, Sci. Rep. 7, 17866 (2017). Youssef, H. Le Gall, and A. Fert, Phys. Rev. B 67, 174402 (2003).
180
M. Maldovan, Nature 503, 209 (2013). 213
M. Dyakonov and V. Perel, Phys. Lett. A 35, 459 (1971).
181
S. R. Sklan, AIP Adv. 5, 053302 (2015). 214
J. E. Hirsch, Phys. Rev. Lett. 83, 1834 (1999).
182
A. Khitun, D. E. Nikonov, M. Bao, K. Galatsis, and K. L. Wang, 215
Y. A. Bychkov and E. I. Rashba, J. Exp. Theor. Phys. Lett. 39, 78 (1984).
Nanotechnology 18, 465202 (2007). 216
I. Mihai Miron, G. Gaudin, S. Auffret, B. Rodmacq, A. Schuhl, S. Pizzini,
183
A. Khitun, D. E. Nikonov, M. Bao, K. Galatsis, and K. L. Wang, IEEE Trans. J. Vogel, and P. Gambardella, Nat. Mater. 9, 230 (2010).
Electron Devices 54, 3418 (2007). 217
A. A. Kovalev, G. E. W. Bauer, and A. Brataas, Phys. Rev. B 75, 014430
184
S. Rakheja, A. Ceyhan, and A. Naeemi, in CMOS and Beyond: Logic Switches (2007).
218
for Terascale Integrated Circuits, edited by T.-J. K. Liu and K. Kuhn (Cambridge G. Allen, S. Manipatruni, D. E. Nikonov, M. Doczy, and I. A. Young, Phys.
University Press, Cambridge, 2015), Chap. 15, p. 381. Rev. B 91, 144412 (2015).
185 219
S. Dutta, S.-C. Chang, N. Kani, D. E. Nikonov, S. Manipatruni, I. A. Young, K. Ando, S. Takahashi, K. Harii, K. Sasage, J. Ieda, S. Maekawa, and
and A. Naeemi, Sci. Rep. 5, 9861 (2015). E. Saitoh, Phys. Rev. Lett. 101, 036601 (2008).

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-39

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

220 253
V. E. Demidov, S. Urazhdin, R. Liu, B. Divinskiy, A. Telegin, and M. Balinskiy, A. C. Chavez, A. Barra, H. Chiang, G. P. Carman, and
S. O. Demokritov, Nat. Commun. 7, 10446 (2016). A. Khitun, Sci. Rep. 8, 10867 (2018).
221
M. Evelt, V. E. Demidov, V. Bessonov, S. O. Demokritov, J. L. Prieto, 254
M. I. Bichurin, V. M. Petrov, S. V. Averkin, and E. Liverts, J. Appl. Phys. 107,
M. Muñoz, J. Ben Youssef, V. V. Naletov, G. de Loubens, O. Klein, M. Collet, 053905 (2010).
255
K. Garcia-Hernandez, P. Bortolotti, V. Cros, and A. Anane, Appl. Phys. Lett. M. I. Bichurin, V. M. Petrov, O. V. Ryabkov, S. V. Averkin, and G. Srinivasan,
108, 172406 (2016). Phys. Rev. B 72, 060408 (2005).
222 256
Y. Kajiwara, K. Harii, S. Takahashi, J. Ohe, K. Uchida, M. Mizuguchi, V. M. Laletin, N. Paddubnaya, G. Srinivasan, C. P. D. Vreugd, M. I. Bichurin,
H. Umezawa, H. Kawai, K. Ando, K. Takanashi, S. Maekawa, and E. Saitoh, V. M. Petrov, and D. A. Filippov, Appl. Phys. Lett. 87, 222507 (2005).
Nature 464, 262 (2010). 257
D. A. Filippov, U. Laletsin, and G. Srinivasan, J. Appl. Phys. 102, 093901
223
V. E. Demidov, S. Urazhdin, H. Ulrichs, V. Tiberkevich, A. Slavin, D. Baither, (2007).
G. Schmitz, and S. O. Demokritov, Nat. Mater. 11, 1028 (2012). 258
A. Barra, A. Mal, G. Carman, and A. Sepulveda, Appl. Phys. Lett. 110,
224
L. Liu, C.-F. Pai, D. C. Ralph, and R. A. Buhrman, Phys. Rev. Lett. 109, 072401 (2017).
259
186602 (2012). D. Tierno, F. Ciubotaru, R. Duflou, M. Heyns, I. P. Radu, and C. Adelmann,
225
P. Dürrenfeld, A. A. Awad, A. Houshang, R. K. Dumas, and J. Åkerman, Microelectron. Eng. 187–188, 144 (2018).
Nanoscale 9, 1285 (2017). 260
F. Vanderveken, H. Ahmad, M. Heyns, B. Sorée, C. Adelmann, and
226
V. E. Demidov, H. Ulrichs, S. V. Gurevich, S. O. Demokritov, F. Ciubotaru, J. Phys. D: Appl. Phys. 53, 495006 (2020).
V. S. Tiberkevich, A. N. Slavin, A. Zholud, and S. Urazhdin, Nat. Commun. 5, 261
A. Akhiezer, V. Bar’iakhtar, and S. Peletminskii, Sov. Phys. JETP 8, 157
3179 (2014). (1959), available at http://www.jetp.ac.ru/cgi-bin/dn/e_008_01_0157.pdf.
227
S. Urazhdin, V. E. Demidov, R. Cao, B. Divinskiy, V. Tyberkevych, A. Slavin, 262
P. A. Fedders, Phys. Rev. B 9, 3835 (1974).
A. B. Rinkevich, and S. O. Demokritov, Appl. Phys. Lett. 109, 162402 (2016). 263
V. W. Tucker and J. W. Rampton, Microwave Ultrasonics in Solid State
228
G. Talmelli, F. Ciubotaru, K. Garello, X. Sun, M. Heyns, I. P. Radu, Physics (North Holland, Amsterdam, 1972).
C. Adelmann, and T. Devolder, Phys. Rev. Appl. 10, 044060 (2018). 264
F. Vanderveken, F. Ciubotaru, and C. Adelmann, arXiv:2003.12099
229
G. Srinivasan, S. Priya, and N. Sun, Composite Magnetoelectrics: Materials, (2020).
265
Structures, and Applications (Woodhead Publishing, Boston, 2015). M. Weisheit, S. Fähler, A. Marty, Y. Souche, C. Poinsignon, and D. Givord,
230
M. Fiebig, J. Phys. D Appl. Phys. 38, R123 (2005). Science 315, 349 (2007).
231
G. Srinivasan, Ann. Rev. Mater. Res. 40, 153 (2010). 266
C.-G. Duan, J. P. Velev, R. F. Sabirianov, Z. Zhu, J. Chu, S. S. Jaswal, and
232
J. Ma, J. Hu, Z. Li, and C.-W. Nan, Adv. Mater. 23, 1062 (2011). E. Y. Tsymbal, Phys. Rev. Lett. 101, 137201 (2008).
233
L. Martin and R. Ramesh, Acta Mater. 60, 2449 (2012). 267
T. Maruyama, Y. Shiota, T. Nozaki, K. Ohta, N. Toda, M. Mizuguchi,
234
C. A. Fernandes Vaz and U. Staub, J. Mater. Chem. C 1, 6731 (2013). A. A. Tulapurkar, T. Shinjo, M. Shiraishi, S. Mizukami, Y. Ando, and Y. Suzuki,
235
S. Fusil, V. Garcia, A. Barthélémy, and M. Bibes, Ann. Rev. Mater. Res. 44, 91 Nat. Nanotechnol. 4, 158 (2009).
268
(2014). P. K. Amiri, J. G. Alzate, X. Q. Cai, F. Ebrahimi, Q. Hu, K. Wong, C. Grèzes,
236
Z. Chu, M. J. PourhosseiniAsl, and S. Dong, J. Phys. D Appl. Phys. 51, H. Lee, G. Yu, X. Li, M. Akyol, Q. Shao, J. A. Katine, J. Langer, B. Ocker, and
243001 (2018). K. L. Wang, IEEE Trans. Magn. 51, 3401507 (2015).
237
N. A. Spaldin and R. Ramesh, Nat. Mater. 18, 203 (2019). 269
S. Pal, B. Rana, O. Hellwig, T. Thomson, and A. Barman, Appl. Phys. Lett.
238
C. Kittel, Phys. Rev. 110, 836 (1958). 98, 082501 (2011).
239
R. Duflou, F. Ciubotaru, A. Vaysset, M. Heyns, B. Sore, I. P. Radu, and 270
T. Miyazaki and T. Tezuka, J. Magn. Magn. Mater. 139, L231 (1995).
C. Adelmann, Appl. Phys. Lett. 111, 192411 (2017). 271
J. S. Moodera, L. R. Kinder, T. M. Wong, and R. Meservey, Phys. Rev. Lett.
240
H. Zhou, A. Talbi, N. Tiercelin, and O. B. Matar, Appl. Phys. Lett. 104, 74, 3273 (1995).
114101 (2014). 272
S. Yuasa, Y. Suzuki, T. Katayama, and K. Ando, Appl. Phys. Lett. 87, 242503
241
S. Cherepov, P. Amiri, J. G. Alzate, K. Wong, M. Lewis, P. Upadhyaya, (2005).
273
J. Nath, M. Bao, A. Bur, T. Wu, G. P. Carman, A. Khitun, and K. L. Wang, Appl. M. K. Niranjan, C.-G. Duan, S. S. Jaswal, and E. Y. Tsymbal, Appl. Phys.
Phys. Lett. 104, 082403 (2014). Lett. 96, 222504 (2010).
242 274
M. Foerster, F. Macià, N. Statuto, S. Finizio, A. Hernández-Mínguez, H. X. Yang, M. Chshiev, B. Dieny, J. H. Lee, A. Manchon, and K. H. Shin,
S. Lendínez, P. V. Santos, J. Fontcuberta, J. M. Hernàndez, M. Kläui, and Phys. Rev. B 84, 054401 (2011).
L. Aballe, Nat. Commun. 8, 407 (2017). 275
B. F. Vermeulen, F. Ciubotaru, M. I. Popovici, J. Swerts, S. Couet, I. P. Radu,
243
M. Weiler, L. Dreher, C. Heeg, H. Huebl, R. Gross, M. S. Brandt, and A. Stancu, K. Temst, G. Groeseneken, C. Adelmann, and K. M. Martens, ACS
S. T. B. Goennenwein, Phys. Rev. Lett. 106, 117601 (2011). Appl. Mater. Interf. 11, 34385 (2019).
244 276
L. Dreher, M. Weiler, M. Pernpeintner, H. Huebl, R. Gross, M. S. Brandt, T. Kawabe, K. Yoshikawa, M. Tsujikawa, T. Tsukahara, K. Nawaoka,
and S. T. B. Goennenwein, Phys. Rev. B 86, 134415 (2012). Y. Kotani, K. Toyoki, M. Goto, M. Suzuki, T. Nakamura, M. Shirai, Y. Suzuki,
245
P. G. Gowtham, T. Moriyama, D. C. Ralph, and R. A. Buhrman, J. Appl. and S. Miwa, Phys. Rev. B 96, 220412 (2017).
Phys. 118, 233910 (2015). 277
T. Nozaki, Y. Shiota, S. Miwa, S. Murakami, F. Bonell, S. Ishibashi,
246
D. Labanowski, A. Jung, and S. Salahuddin, Appl. Phys. Lett. 108, 022905 H. Kubota, K. Yakushiji, T. Saruya, A. Fukushima, S. Yuasa, T. Shinjo, and
(2016). Y. Suzuki, Nat. Phys. 8, 491 (2012).
247 278
L. Thevenard, C. Gourdon, J. Y. Prieur, H. J. von Bardeleben, S. Vincent, J. Zhu, J. A. Katine, G. E. Rowlands, Y.-J. Chen, Z. Duan, J. G. Alzate,
L. Becerra, L. Largeau, and J.-Y. Duquesne, Phys. Rev. B 90, 094401 (2014). P. Upadhyaya, J. Langer, P. K. Amiri, K. L. Wang, and I. N. Krivorotov, Phys.
248
P. G. Gowtham, D. Labanowski, and S. Salahuddin, Phys. Rev. B 94, 014436 Rev. Lett. 108, 197203 (2012).
279
(2016). R. Verba, M. Carpentieri, G. Finocchio, V. Tiberkevich, and A. Slavin, Sci.
249
S. Bhuktare, A. Bose, H. Singh, and A. A. Tulapurkar, Sci. Rep. 7, 840 (2017). Rep. 6, 25018 (2016).
250
X. Li, D. Labanowski, S. Salahuddin, and C. S. Lynch, J. Appl. Phys. 122, 280
B. Rana, Y. Fukuma, K. Miura, H. Takahashi, and Y. Otani, Appl. Phys. Lett.
043904 (2017). 111, 052404 (2017).
251
R. Verba, I. Lisenkov, I. Krivorotov, V. Tiberkevich, and A. Slavin, Phys. Rev. 281
B. Rana and Y. Otani, Commun. Phys. 2, 90 (2019).
Appl. 9, 064014 (2018). 282
R. Verba, V. Tiberkevich, I. Krivorotov, and A. Slavin, Phys. Rev. Appl. 1,
252
R. Verba, V. Tiberkevich, and A. Slavin, Phys. Rev. Appl. 12, 054061 (2019). 044006 (2014).

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-40

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

283 314
Y.-J. Chen, H. K. Lee, R. Verba, J. A. Katine, I. Barsukov, V. Tiberkevich, C. S. Davies, A. Francis, A. V. Sadovnikov, S. V. Chertopalov, M. T. Bryan,
J. Q. Xiao, A. N. Slavin, and I. N. Krivorotov, Nano Lett. 17, 572 (2017). S. V. Grishin, D. A. Allwood, Y. P. Sharaevskii, S. A. Nikitov, and
284
A. S. Shukla, A. Chouhan, R. Pandey, M. Raghupathi, T. Yamamoto, V. V. Kruglyak, Phys. Rev. B 92, 020408 (2015).
H. Kubota, A. Fukushima, S. Yuasa, T. Nozaki, and A. A. Tulapurkar, Sci. Adv. 315
A. Haldar, D. Kumar, and A. O. Adeyeye, Nat. Nanotechnol. 11, 437 (2016).
6, eabc2618 (2020). 316
M. Vogel, R. Aßmann, P. Pirro, A. V. Chumak, B. Hillebrands, and G. von
285
T. Satoh, Y. Terui, R. Moriya, B. A. Ivanov, K. Ando, E. Saitoh, T. Shimura, Freymann, Sci. Rep. 8, 11099 (2018).
and K. Kuroda, Nat. Photonics 6, 662 (2012). 317
E. Albisetti, D. Petti, G. Sala, R. Silvani, S. Tacchi, S. Finizio, S. Wintz,
286
I. V. Savochkin, M. Jäckl, V. I. Belotelov, I. A. Akimov, M. A. Kozhaev, A. Calò, X. Zheng, J. Raabe, E. Riedo, and R. Bertacco, Commun. Phys. 1, 56
D. A. Sylgacheva, A. I. Chernov, A. N. Shaposhnikov, A. R. Prokopov, (2018).
V. N. Berzhansky, D. R. Yakovlev, A. K. Zvezdin, and M. Bayer, Sci. Rep. 7, 5668 318
A. Haldar, C. Tian, and A. O. Adeyeye, Sci. Adv. 3, e1700638 (2017).
319
(2017). R. C. O’Handley, in Modern Magnetic Materials, 1st ed. (Wiley, New York,
287
N. Ogawa, W. Koshibae, A. J. Beekman, N. Nagaosa, M. Kubota, 1999), Chap. 6, p. 179.
M. Kawasaki, and Y. Tokura, Proc. Natl. Acad. Sci. U.S.A. 112, 8977 (2015). 320
L. Soumah, N. Beaulieu, L. Qassym, C. Carrétéro, E. Jacquet, R. Lebourgeois,
288
N. E. Khokhlov, P. I. Gerevenkov, L. A. Shelukhin, A. V. Azovtsev, J. Ben Youssef, P. Bortolotti, V. Cros, and A. Anane, Nat. Commun. 9, 1 (2018).
321
N. A. Pertsev, M. Wang, A. W. Rushforth, A. V. Scherbakov, and J. J. Bauer, E. R. Rosenberg, S. Kundu, K. A. Mkhoyan, P. Quarterman,
A. M. Kalashnikova, Phys. Rev. Appl. 12, 044044 (2019). A. J. Grutter, B. J. Kirby, J. A. Borchers, and C. A. Ross, Adv. Electron. Mater. 6,
289
M. van Kampen, C. Jozsa, J. T. Kohlhepp, P. LeClair, L. Lagae, W. J. M. de 1900820 (2020).
Jonge, and B. Koopmans, Phys. Rev. Lett. 88, 227201 (2002). 322
M. J. Pechan and I. K. Schuller, Phys. Rev. Lett. 59, 132 (1987).
290 323
Y. Au, M. Dvornik, T. Davison, E. Ahmad, P. S. Keatley, A. Vansteenkiste, M. T. Johnson, P. J. H. Bloemen, F. J. A. den Broeder, and J. J. de Vries, Rep.
B. Van Waeyenberge, and V. V. Kruglyak, Phys. Rev. Lett. 110, 097201 (2013). Prog. Phys. 59, 1409 (1996).
291
A. Kirilyuk, A. V. Kimel, and T. Rasing, Rep. Prog. Phys. 76, 026501 (2013). 324
B. Dieny and M. Chshiev, Rev. Mod. Phys. 89, 025008 (2017).
292 325
S. Iihama, Y. Sasaki, A. Sugihara, A. Kamimaki, Y. Ando, and S. Mizukami, F. Garcia-Sanchez, P. Borys, R. Soucaille, J.-P. Adam, R. L. Stamps, and
Phys. Rev. B 94, 020401 (2016). J.-V. Kim, Phys. Rev. Lett. 114, 247206 (2015).
293 326
M. Jäckl, V. Belotelov, I. Akimov, I. Savochkin, D. Yakovlev, A. Zvezdin, and K. Wagner, A. Kákay, K. Schultheiss, A. Henschke, T. Sebastian, and
M. Bayer, Phys. Rev. X 7, 021009 (2017). H. Schultheiss, Nat. Nanotechnol. 11, 432 (2016).
294
W. K. Hiebert, A. Stankiewicz, and M. R. Freeman, Phys. Rev. Lett. 79, 1134 327
E. N. Beginin, A. V. Sadovnikov, A. Y. Sharaevskaya, A. I. Stognij, and
(1997). S. A. Nikitov, Appl. Phys. Lett. 112, 122404 (2018).
295
M. R. Freeman and B. C. Cjoi, Science 294, 1484 (2001). 328
J. B. Johnson, Phys. Rev. 32, 97 (1928).
296 329
M. Bauer, R. Lopusnik, H. Dötsch, B. Kalinikos, C. Patton, J. Fassbender, and S. Dutta, D. E. Nikonov, S. Manipatruni, I. A. Young, and A. Naeemi, Sci.
B. Hillebrands, J. Magn. Magn. Mater. 226–230, 507 (2001). Rep. 7, 1915 (2017).
297
J. Sandercock and W. Wettling, IEEE Trans. Magn. 14, 442 (1978). 330
J. Van Kranendonk and J. H. Van Vleck, Rev. Mod. Phys. 30, 1 (1958).
298
B. Hillebrands, P. Baumgart, and G. Güntherodt, Phys. Rev. B 36, 2450 331
S. Dutta, D. E. Nikonov, S. Manipatruni, I. A. Young, and A. Naeemi, IEEE
(1987). Trans. Electron Devices 62, 3863 (2015).
299
V. E. Demidov, M. P. Kostylev, K. Rott, P. Krzysteczko, G. Reiss, and 332
D. S. Deng, X. F. Jin, and R. Tao, Phys. Rev. B 66, 104435 (2002).
S. O. Demokritov, Phys. Rev. B 83, 054408 (2011). 333
M. P. Kostylev, A. A. Stashkevich, and N. A. Sergeeva, Phys. Rev. B 69,
300
V. E. Demidov, S. O. Demokritov, B. Hillebrands, M. Laufenberg, and 064408 (2004).
P. P. Freitas, Appl. Phys. Lett. 85, 2866 (2004). 334
G. Gubbiotti, S. Tacchi, M. Madami, G. Carlotti, S. Jain, A. O. Adeyeye, and
301
A. Banholzer, R. Narkowicz, C. Hassel, R. Meckenstock, S. Stienen, O. Posth, M. P. Kostylev, Appl. Phys. Lett. 100, 162407 (2012).
D. Suter, M. Farle, and J. Lindner, Nanotechnology 22, 295713 (2011). 335
D. Kumar, J. W. Kłos, M. Krawczyk, and A. Barman, J. Appl. Phys. 115,
302
O. Büttner, M. Bauer, S. O. Demokritov, B. Hillebrands, Y. S. Kivshar, 043917 (2014).
V. Grimalsky, Y. Rapoport, and A. N. Slavin, Phys. Rev. B 61, 11576 336
M. Krawczyk and H. Puszkarski, Phys. Rev. B 77, 054437 (2008).
337
(2000). J. Romero Vivas, S. Mamica, M. Krawczyk, and V. V. Kruglyak, Phys. Rev. B
303
A. A. Serga, T. Schneider, B. Hillebrands, S. O. Demokritov, and 86, 144417 (2012).
M. P. Kostylev, Appl. Phys. Lett. 89, 063506 (2006). 338
G. Gubbiotti, A. Sadovnikov, E. Beginin, S. Nikitov, D. Wan, A. Gupta,
304
C. W. Sandweg, M. B. Jungfleisch, V. I. Vasyuchka, A. A. Serga, P. Clausen, S. Kundu, G. Talmelli, R. Carpenter, I. Asselberghs, I. P. Radu, C. Adelmann,
H. Schultheiss, B. Hillebrands, A. Kreisel, and P. Kopietz, Rev. Sci. Instrum. 81, and F. Ciubotaru, arXiv:2007.13707 (2020).
073902 (2010). 339
Y. Gulyaev and A. Nikitov, Dokl. Phys. 46, 687 (2011).
305
R. Mock, B. Hillebrands, and R. Sandercock, J. Phys. E Sci. Instrum. 20, 656 340
S. L. Vysotskii, S. A. Nikitov, and Y. A. Filimonov, J. Exp. Theor. Phys. 101,
(1987). 547 (2005).
306
B. Hillebrands, Rev. Sci. Instrum. 70, 1589 (1999). 341
V. V. Kruglyak and R. Hicken, J. Magn. Magn. Mater. 306, 191
307
C. Felser and A. Hirohata, Heusler Alloys: Properties, Growth, Applications (2006).
342
(Springer, 2015). A. V. Chumak, A. A. Serga, B. Hillebrands, and M. P. Kostylev, Appl. Phys.
308
C. J. Palmstrøm, Prog. Crystal Growth Charact. Mater. 62, 371 (2016). Lett. 93, 022508 (2008).
309 343
L. Wollmann, A. K. Nayak, S. S. Parkin, and C. Felser, Ann. Rev. Mater. Res. A. V. Chumak, P. Pirro, A. A. Serga, M. P. Kostylev, R. L. Stamps,
47, 247 (2017). H. Schultheiss, K. Vogt, S. J. Hermsdoerfer, B. Laegel, P. A. Beck, and
310
K. Vogt, H. Schultheiss, S. Jain, J. E. Pearson, A. Hoffmann, S. D. Bader, and B. Hillebrands, Appl. Phys. Lett. 95, 262508 (2009).
B. Hillebrands, Appl. Phys. Lett. 101, 042410 (2012). 344
B. Obry, P. Pirro, T. Brächer, A. V. Chumak, J. Osten, F. Ciubotaru,
311
V. S. Tkachenko, A. N. Kuchko, M. Dvornik, and V. V. Kruglyak, Appl. Phys. A. A. Serga, J. Fassbender, and B. Hillebrands, Appl. Phys. Lett. 102, 202403
Lett. 101, 152402 (2012). (2013).
312
X. Xing, Y. Yu, S. Li, and X. Huang, Sci. Rep. 3, 2958 (2013). 345
F. Ciubotaru, A. V. Chumak, B. Obry, and A. Serga, Phys. Rev. B 88, 134406
313
A. V. Sadovnikov, C. S. Davies, V. V. Kruglyak, D. V. Romanenko, (2013).
S. V. Grishin, E. N. Beginin, Y. P. Sharaevskii, and S. A. Nikitov, Phys. Rev. B 96, 346
A. V. Chumak, T. Neumann, A. A. Serga, B. Hillebrands, and M. P. Kostylev,
060401 (2017). J. Phys. D Appl. Phys. 42, 205005 (2009).

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-41

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

347
A. V. Chumak, A. A. Serga, and B. Hillebrands, J. Phys. D Appl. Phys. 50, 378
K. Vogt, F. Y. Fradin, J. E. Pearson, T. Sebastian, S. D. Bader, B. Hillebrands,
244001 (2017). A. P. Hoffmann, and H. Schultheiss, Nat. Commun. 5, 3727 (2014).
348 379
H. G. Bauer, P. Majchrak, T. Kachel, C. H. Back, and G. Woltersdorf, Nat. F. Heussner, M. Nabinger, T. Fischer, T. Brächer, A. A. Serga, B. Hillebrands,
Commun. 6, 8274 (2015). and P. Pirro, Phys. Status Solidi RRL 12, 1800409 (2018).
349
B. A. Kalinikos and A. B. Ustinov, Solid State Phys. 64, 193 (2013). 380
F. Heussner, G. Talmelli, M. Geilen, B. Heinz, T. Brächer, T. Meyer,
350
M. Balinskiy, H. Chiang, and A. Khitun, AIP Adv. 8, 056628 (2018). F. Ciubotaru, C. Adelmann, K. Yamamoto, A. A. Serga, B. Hillebrands, and
351
Q. Wang, P. Pirro, R. Verba, A. Slavin, B. Hillebrands, and A. V. Chumak, P. Pirro, Phys. Status Solidi RRL 14, 1900695 (2020).
Sci. Adv. 4, e1701517 (2018). 381
T. Schneider, A. A. Serga, A. V. Chumak, C. W. Sandweg, S. Trudel, S. Wolff,
352
S. V. Vasiliev, V. V. Kruglyak, M. L. Sokolovskii, and A. N. Kuchko, J. Appl. M. P. Kostylev, V. S. Tiberkevich, A. N. Slavin, and B. Hillebrands, Phys. Rev.
Phys. 101, 113919 (2007). Lett. 104, 197203 (2010).
353
V. E. Demidov, S. Urazhdin, and S. O. Demokritov, Appl. Phys. Lett. 95, 382
N. H. E. Weste and D. M. Harris, CMOS VLSI Design: A Circuits and Systems
262509 (2009). Perspective, 4th ed. (Addison Wesley, Boston, 2011).
354
T. Liu and G. Vignale, Phys. Rev. Lett. 106, 247203 (2011). 383
Electronic Design Automation for IC Implementation, Circuit Design, and
355
T. Schneider, A. A. Serga, B. Leven, B. Hillebrands, R. L. Stamps, and Process Technology, edited by L. Lavagno, I. L. Markov, G. Martin, and L. K.
M. P. Kostylev, Appl. Phys. Lett. 92, 022505 (2008). Scheffer, 2nd ed. (CRC Press, Boca Raton, 2016).
356
B. Rana and Y. Otani, Phys. Rev. Appl. 9, 014033 (2018). 384
C. Mead and L. Conway, Introduction to VLSI Systems (Addison-Wesley,
357
A. Khitun and K. L. Wang, Superlatt. Microstruct. 38, 184 (2005). Reading, 1979).
358
K. Nanayakkara, A. Anferov, A. P. Jacob, S. J. Allen, and A. Kozhanov, IEEE 385
A. Khitun, M. Bao, and K. L. Wang, J. Phys. D Appl. Phys. 43, 264005
Trans. Magn. 50, 3402204 (2014). (2010).
359 386
M. Balynsky, A. Kozhevnikov, Y. Khivintsev, T. Bhowmick, D. Gutierrez, O. Zografos, B. Sorée, A. Vaysset, S. Cosemans, L. Amarù, P.-E. Gaillardon,
H. Chiang, G. Dudko, Y. Filimonov, G. Liu, C. Jiang, A. A. Balandin, R. Lake, G. D. Micheli, R. Lauwereins, S. Sayan, P. Raghavan, I. P. Radu, and A. Thean, in
and A. Khitun, J. Appl. Phys. 121, 024504 (2017). 2015 IEEE 15th International Conference on Nanotechnology (IEEE-NANO)
360
S. Khasanvis, M. Rahman, S. N. Rajapandian, and C. A. Moritz, in 2014 (IEEE, 2015), pp. 686–689.
IEEE/ACM International Symposium on Nanoscale Architectures (NANOARCH) 387
R. W. Keyes, Science 230, 138 (1985).
(IEEE, 2014), pp. 171–176. 388
E. Egel, C. Meier, G. Csaba, and S. Breitkreutz-von Gamm, AIP Adv. 7,
361
N. Kanazawa, T. Goto, K. Sekiguchi, A. B. Granovsky, C. A. Ross, H. Takagi, 056016 (2017).
Y. Nakamura, H. Uchida, and M. Inoue, Sci. Rep. 7, 7898 (2017). 389
H. Nyquist, Phys. Rev. 32, 110 (1928).
362 390
T. Fischer, M. Kewenig, D. A. Bozhko, A. A. Serga, I. I. Syvorotka, O. Zografos, A. D. Meester, E. Testa, M. Soeken, P.-E. Gaillardon,
F. Ciubotaru, C. Adelmann, B. Hillebrands, and A. V. Chumak, Appl. Phys. Lett. G. D. Micheli, L. Amarù, P. Raghavan, F. Catthoor, and R. Lauwereins, in
110, 152401 (2017). Design, Automation Test in Europe Conference Exhibition (DATE), 2017 (IEEE,
363
F. Ciubotaru, G. Talmelli, T. Devolder, O. Zografos, M. Heyns, C. Adelmann, 2017), pp. 1306–1311.
391
and I. P. Radu, in 2018 IEEE International Electron Devices Meeting (IEDM) A. Mahmoud, F. Vanderveken, C. Adelmann, F. Ciubotaru, S. Cotofana, and
(IEEE, 2018), pp. 36.1.1–36.1.4. S. Hamdioui, in 2020 IEEE Computer Society Annual Symposium on VLSI
364
N. Sato, K. Sekiguchi, and Y. Nozaki, Appl. Phys. Express 6, 063001 (2013). (ISVLSI) (IEEE, 2020), pp. 60–65.
365 392
G. Talmelli, T. Devolder, N. Träger, J. Förster, S. Wintz, M. Weigand, R. Miller, V. Prasanna-Kumar, D. Reisis, and Q. Stout, IEEE Trans. Comput.
H. Stoll, M. Heyns, G. Schütz, I. Radu, J. Gräfe, F. Ciubotaru, and C. Adelmann, 42, 678 (1993).
arXiv:1908.02546 (2019). 393
L. Chua and L. Yang, IEEE Trans. Circuits Syst. 35, 1257 (1988).
366
A. V. Sadovnikov, E. N. Beginin, S. E. Sheshukova, D. V. Romanenko, 394
L. Chua and T. Roska, IEEE Trans. Circuits Syst. I. Fundam. Theor. Appl. 40,
Y. P. Sharaevskii, and S. A. Nikitov, Appl. Phys. Lett. 107, 202405 (2015). 147 (1993).
367 395
A. V. Sadovnikov, E. N. Beginin, M. A. Morozova, Y. P. Sharaevskii, S. V. Grishin, N. Petkov, Systolic Parallel Processing (Elsevier, New York, 1992).
S. E. Sheshukova, and S. A. Nikitov, Appl. Phys. Lett. 109, 042407 (2016). 396
J. G. Alzate, P. Upadhyaya, M. Lewis, J. Nath, Y. T. Lin, K. Wong,
368
A. Mahmoud, F. Vanderveken, C. Adelmann, F. Ciubotaru, S. Cotofana, and S. Cherepov, P. K. Amiri, K. L. Wang, J. Hockel, A. Bur, G. P. Carman,
S. Hamdioui, “Spin wave normalization towards all magnonic circuits,” IEEE S. Bender, Y. Tserkovnyak, J. Zhu, Y. Chen, I. N. Krivorotov, J. Katine, J. Langer,
Trans. Circuits Syst. I Reg. Papers (to be published 2020). P. Shabadi, S. Khasanvis, S. Narayanan, C. A. Moritz, and A. Khitun, in 2012
369
Z. Haghshenasfard and M. G. Cottam, IEEE Magn. Lett. 7, 1 (2016). IEEE/ACM International Symposium on Nanoscale Architectures (NANOARCH)
370
T. Brächer, F. Heussner, P. Pirro, T. Fischer, M. Geilen, B. Heinz, B. Lägel, (IEEE, 2012), pp. 196–202.
A. A. Serga, and B. Hillebrands, Appl. Phys. Lett. 105, 232409 (2014). 397
M. M. Eshaghian-Wilner, A. Friesz, A. Khitun, S. Navab, A. C. Parker,
371
R. Verba, M. Carpentieri, G. Finocchio, V. Tiberkevich, and A. Slavin, Appl. K. L. Wang, and C. Zhou, J. Phys. Conf. Ser. 61, 288 (2007).
Phys. Lett. 112, 042402 (2018). 398
J. A. Sharp, Data Flow Computing: Theory and Practice (Ablex Publications,
372
F. Ciubotaru, A. A. Serga, B. Leven, B. Hillebrands, and L. Lopez-Diaz, Phys. Norwood, NJ, 1992).
Rev. B 84, 144424 (2011). 399
R. Dogaru, Systematic Design for Emergence in Cellular Nonlinear Networks
373
T. Meyer, T. Brächer, F. Heussner, A. A. Serga, H. Naganuma, with Applications in Natural Computing and Signal Processing (Springer, Berlin,
K. Mukaiyama, M. Oogane, Y. Ando, B. Hillebrands, and P. Pirro, IEEE Magn. 2008).
Lett. 9, 1 (2018). 400
C. Bobda and R. Hartenstein, Introduction to Reconfigurable
374
A. Khitun, D. E. Nikonov, and K. L. Wang, J. Appl. Phys. 106, 123909 Computing: Architectures, Algorithms, and Applications (Springer, Dordrecht,
(2009). 2010).
375
P. Chowdhury, P. Dhagat, and A. Jander, IEEE Trans. Magn. 51, 1300904 401
S. Dutta, D. E. Nikonov, S. Manipatruni, I. A. Young, and A. Naeemi, IEEE
(2015). Trans. Magn. 50, 1300411 (2014).
376
A. V. Sadovnikov, C. S. Davies, S. V. Grishin, V. V. Kruglyak, 402
B. A. Kalinikos, N. G. Kovshikov, and A. N. Slavin, Sov. Phys. JETP Lett. 38,
D. V. Romanenko, Y. P. Sharaevskii, and S. A. Nikitov, Appl. Phys. Lett. 106, 413 (1983).
403
192406 (2015). A. A. Serga, S. O. Demokritov, B. Hillebrands, and A. N. Slavin, Phys. Rev.
377
C. S. Davies, A. V. Sadovnikov, S. V. Grishin, Y. P. Sharaevsky, S. A. Nikitov, Lett. 92, 117203 (2004).
and V. V. Kruglyak, IEEE Trans. Magn. 51, 1 (2015). 404
A. Khitun, J. Appl. Phys. 118, 243905 (2015).

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-42

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

405 434
F. Gertz, A. Kozhevnikov, Y. Khivintsev, G. Dudko, M. Ranjbar, S. Kosen, A. F. van Loo, D. A. Bozhko, L. Mihalceanu, and
D. Gutierrez, H. Chiang, Y. Filimonov, and A. Khitun, IEEE Trans. Magn. 52, A. D. Karenowska, APL Mater. 7, 101120 (2019).
435
3401304 (2016). Y. Li, W. Zhang, V. Tyberkevych, W.-K. Kwok, A. Hoffmann, and
406
M. Balynsky, D. Gutierrez, H. Chiang, A. Khitun, A. Kozhevnikov, V. Novosad, arXiv:2006.16158 (2020).
436
Y. Khivintsev, G. Dudko, and Y. Filimonov, in 2016 IEEE International D. A. Bozhko, A. A. Serga, P. Clausen, V. I. Vasyuchka, F. Heussner,
Conference on Rebooting Computing (ICRC) (IEEE, 2016), pp. 1–4. G. A. Melkov, A. Pomyalov, V. S. L’vov, and B. Hillebrands, Nat. Phys. 12, 1057
407
Y. Khivintsev, M. Ranjbar, D. Gutierrez, H. Chiang, A. Kozhevnikov, (2016).
Y. Filimonov, and A. Khitun, J. Appl. Phys. 120, 123901 (2016). 437
A. A. Serga, V. S. Tiberkevich, C. W. Sandweg, V. I. Vasyuchka,
408
G. Csaba, A. Papp, and W. Porod, J. Appl. Phys. 115, 17C741 (2014). D. A. Bozhko, A. V. Chumak, T. Neumann, B. Obry, G. A. Melkov, A. N. Slavin,
409
D. Miller, Proc. IEEE 97, 1166 (2009). and B. Hillebrands, Nat. Commun. 5, 3452 (2014).
410 438
F. Gertz, A. Kozhevnikov, Y. Filimonov, and A. Khitun, IEEE Trans. Magn. M. Schneider, T. Brächer, D. Breitbach, V. Lauer, P. Pirro, D. A. Bozhko,
51, 4002905 (2015). H. Y. Musiienko-Shmarova, B. Heinz, Q. Wang, T. Meyer, F. Heussner, S. Keller,
411
D. Gutierrez, H. Chiang, T. Bhowmick, A. Volodchenkov, M. Ranjbar, E. T. Papaioannou, B. Lägel, T. Löber, C. Dubs, A. N. Slavin, V. S. Tiberkevich,
G. Liu, C. Jiang, C. Warren, Y. Khivintsev, Y. Filimonov, J. Garay, R. Lake, A. A. Serga, B. Hillebrands, and A. V. Chumak, Nat. Nanotechnol. 15, 457
A. Balandin, and A. Khitun, J. Magn. Magn. Mater. 428, 348 (2017). (2020).
412 439
M. Balynskiy, H. Chiang, D. Gutierrez, A. Kozhevnikov, Y. Filimonov, and D. Suess, A. Bachleitner-Hofmann, A. Satz, H. Weitensfelder, C. Vogler,
A. Khitun, J. Appl. Phys. 123, 144501 (2018). F. Bruckner, C. Abert, K. Prügl, J. Zimmer, C. Huber, S. Luber, W. Raberg,
413
M. Rahman, S. Khasanvis, J. Shi, and C. A. Moritz, IEEE Trans. T. Schrefl, and H. Brückl, Nat. Electron. 1, 362 (2018).
Nanotechnol. 14, 742 (2015). 440
C. Zheng, K. Zhu, S. Cardoso de Freitas, J.-Y. Chang, J. E. Davies, P. Eames,
414
C. Pan and A. Naeemi, IEEE J. Explor. Solid-State Comput. 3, 101 P. P. Freitas, O. Kazakova, C. Kim, C.-W. Leung, S.-H. Liou, A. Ognev,
(2017). S. N. Piramanayagam, P. Ripka, A. Samardak, K.-H. Shin, S.-Y. Tong,
415
N. Locatelli, V. Cros, and J. Grollier, Nat. Mater. 13, 11 (2014). M.-J. Tung, S. X. Wang, S. Xue, X. Yin, and P. W. T. Pong, IEEE Trans. Magn.
416
S. Sangiao, C. Magén, D. Mofakhami, G. d. Loubens, and J. M. D. Teresa, 55, 0800130 (2019).
Beilstein J. Nanotechnol. 8, 2106 (2017). 441
M. Inoue, A. Baryshev, H. Takagi, P. B. Lim, K. Hatafuku, J. Noda, and
417
O. V. Dobrovolskiy, R. Sachser, S. A. Bunyaev, D. Navas, V. M. Bevz, K. Togo, Appl. Phys. Lett. 98, 132511 (2011).
M. Zelent, W. Śmigaj, J. Rychły, M. Krawczyk, R. V. Vovk, M. Huth, and 442
P. Talbot, A. Fessant, and J. Gieraltowski, Procedia Eng. 120, 1241
G. N. Kakazei, ACS Appl. Mater. Interf. 11, 17654 (2019). (2015).
418
M. Huth, F. Porrati, and O. V. Dobrovolskiy, Microelectron. Eng. 185–186, 9 443
P. J. Metaxas, M. Sushruth, R. A. Begley, J. Ding, R. C. Woodward,
(2018). I. S. Maksymov, M. Albert, W. Wang, H. Fangohr, A. O. Adeyeye, and
419
P. Fischer, D. Sanz-Hernández, R. Streubel, and A. Fernández-Pacheco, APL M. Kostylev, Appl. Phys. Lett. 106, 232406 (2015).
Mater. 8, 010701 (2020). 444
Y. Cao and P. Yan, Phys. Rev. B 99, 214415 (2019).
420 445
L. Flajšman, K. Wagner, M. Vaňatka, J. Gloss, V. Křižáková, M. Schmid, S. Atalay, A. O. Kaya, V. S. Kolat, H. Gencer, and T. Izgi, J. Supercond. Nov.
H. Schultheiss, and M. Urbánek, Phys. Rev. B 101, 014436 (2020). Magn. 28, 2071 (2015).
421
O. V. Dobrovolskiy, M. Kompaniiets, R. Sachser, F. Porrati, C. Gspan, 446
R. G. Kryshtal and A. V. Medved, Appl. Phys. Lett. 100, 192410 (2012).
H. Plank, and M. Huth, Beilstein J. Nanotechnol. 6, 1082 (2015). 447
M. Balynsky, D. Gutierrez, H. Chiang, A. Kozhevnikov, G. Dudko,
422
M. Kompaniiets, O. V. Dobrovolskiy, C. Neetzel, F. Porrati, J. Brötz, Y. Filimonov, A. A. Balandin, and A. Khitun, Sci. Rep. 7, 11539 (2017).
W. Ensinger, and M. Huth, Appl. Phys. Lett. 104, 052603 (2014). 448
D. M. Pozar, Microwave Engineering, 4th ed. (Wiley, Hoboken, 2012).
423
K.-R. Jeon, C. Ciccarelli, H. Kurebayashi, L. F. Cohen, X. Montiel, 449
V. G. Harris, IEEE Trans. Magn. 48, 1075 (2012).
M. Eschrig, T. Wagner, S. Komori, A. Srivastava, J. W. Robinson, and 450
J. M. Owens, J. H. Collins, and R. L. Carter, Circuits Syst. Signal Process. 4,
M. G. Blamire, Phys. Rev. Appl. 11, 014061 (2019). 317 (1985).
424 451
I. A. Golovchanskiy, N. N. Abramov, V. S. Stolyarov, V. V. Bolginov, J. Helszajn, YIG Resonators and Filters (Wiley, Chichester, 1985).
V. V. Ryazanov, A. A. Golubov, and A. V. Ustinov, Adv. Funct. Mater. 28, 452
W. Ishak, Proc. IEEE 76, 171 (1988).
1802375 (2018). 453
J. D. Adam, Proc. IEEE 76, 159 (1988).
425
O. V. Dobrovolskiy, R. Sachser, T. Brächer, T. Böttcher, V. V. Kruglyak, 454
H. Tanbakuchi, D. Nicholson, B. Kunz, and W. Ishak, IEEE Trans. Magn. 25,
R. V. Vovk, V. A. Shklovskij, M. Huth, B. Hillebrands, and A. V. Chumak, Nat. 3248 (1989).
Phys. 15, 477 (2019). 455
J. S. McLean, IEEE Trans. Antennas Propag. 44, 672 (1996).
426
A. Shekhter, L. N. Bulaevskii, and C. D. Batista, Phys. Rev. Lett. 106, 037001 456
H. A. Wheeler, Proc. IEEE 35, 1479 (1947).
457
(2011). J. C.-E. Sten, A. Hujanen, and P. K. Koivisto, IEEE Trans. Antennas Propag.
427
O. V. Dobrovolskiy, D. Y. Vodolazov, F. Porrati, R. Sachser, V. M. Bevz, 49, 829 (2001).
M. Y. Mikhailov, A. V. Chumak, and M. Huth, Nat. Commun. 11, 3291 (2020). 458
D. M. Pozar, Proc. IEEE 80, 79 (1992).
428 459
D. Lachance-Quirion, S. P. Wolski, Y. Tabuchi, S. Kono, K. Usami, and Z. Yao, Y. Wang, S. Keller, and G. P. Carman, IEEE Trans. Antennas Propag.
Y. Nakamura, Science 367, 425 (2020). 63, 3335–3344 (2015).
429
H. Huebl, C. W. Zollitsch, J. Lotze, F. Hocke, M. Greifenstein, A. Marx, 460
J. P. Domann and G. P. Carman, J. Appl. Phys. 121, 044905 (2017).
R. Gross, and S. T. B. Goennenwein, Phys. Rev. Lett. 111, 127003 461
J. Xu, C. M. Leung, X. Zhuang, J. Li, S. Bhardwaj, J. Volakis, and
(2013). D. Viehland, Sensors 19, E853 (2019).
430 462
Y. Tabuchi, S. Ishino, T. Ishikawa, R. Yamazaki, K. Usami, and Y. Nakamura, R. L. Kubena, X. Pang, K. G. Lee, Y. K. Yong, and W. S. Wall, J. Phys. Conf.
Phys. Rev. Lett. 113, 083603 (2014). Ser. 1407, 012026 (2019).
431
X. Zhang, C.-L. Zou, L. Jiang, and H. X. Tang, Phys. Rev. Lett. 113, 156401 463
T. Nan, H. Lin, Y. Gao, A. Matyushov, G. Yu, H. Chen, N. Sun, S. Wei,
(2014). Z. Wang, M. Li, X. Wang, A. Belkessam, R. Guo, B. Chen, J. Zhou, Z. Qian,
432
R. G. E. Morris, A. F. van Loo, S. Kosen, and A. D. Karenowska, Sci. Rep. 7, Y. Hui, M. Rinaldi, M. E. McConney, B. M. Howe, Z. Hu, J. G. Jones,
11511 (2017). G. J. Brown, and N. X. Sun, Nat. Commun. 8, 296 (2017).
433 464
M. Pfirrmann, I. Boventer, A. Schneider, T. Wolz, M. Kläui, A. V. Ustinov, R. V. Petrov, A. S. Tatarenko, S. Pandey, G. Srinivasan, J. V. Mantese, and
and M. Weides, Phys. Rev. Res. 1, 032023 (2019). R. Azadegan, Electron. Lett. 44, 506 (2008).

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-43

Published under license by AIP Publishing.
 Journal of TUTORIAL scitation.org/journal/jap
Applied Physics

465
M. Manteghi and A. A. Y. Ibraheem, IEEE Trans. Antennas Propag. 62, 6491 J. Sibik, T. M. Korter, B. Ellison, S. Rea, P. Goldsmith, K. B. Cooper, R. Appleby,
(2014). D. Pardo, P. G. Huggard, V. Krozer, H. Shams, M. Fice, C. Renaud, A. Seeds,
466
T. Seifert, S. Jaiswal, U. Martens, J. Hannegan, L. Braun, P. Maldonado, A. Stöhr, M. Naftaly, N. Ridler, R. Clarke, J. E. Cunningham, and
F. Freimuth, A. Kronenberg, J. Henrizi, I. Radu et al., Nat. Photonics 10, 483 (2016). M. B. Johnston, J. Phys. D Appl. Phys. 50, 043001 (2017).
467
T. Jungwirth, X. Marti, P. Wadley, and J. Wunderlich, Nat. Nanotechnol. 11, 474
K. Zakeri, Physica C 549, 164 (2018).
475
231 (2016). A. V. Kimel, A. Kirilyuk, A. Tsvetkov, R. V. Pisarev, and T. Rasing, Nature
468
V. Baltz, A. Manchon, M. Tsoi, T. Moriyama, T. Ono, and Y. Tserkovnyak, 429, 850 (2004).
Rev. Mod. Phys. 90, 015005 (2018). 476
B. G. Park, J. Wunderlich, X. Martí, V. Holý, Y. Kurosaki, M. Yamada,
469
C. Kittel, Phys. Rev. 82, 565 (1951). H. Yamamoto, A. Nishide, J. Hayakawa, H. Takahashi, A. B. Shick, and
470
T. Kampfrath, A. Sell, G. Klatt, A. Pashkin, S. Mährlein, T. Dekorsy, M. Wolf, T. Jungwirth, Nat. Mater. 10, 347 (2011).
M. Fiebig, A. Leitenstorfer, and R. Huber, Nat. Photonics 5, 31 (2011). 477
P. Wadley, B. Howells, J. Železný, C. Andrews, V. Hills, R. P. Campion,
471
D. Bossini, S. Dal Conte, Y. Hashimoto, A. Secchi, R. V. Pisarev, T. Rasing, V. Novák, K. Olejník, F. Maccherozzi, S. S. Dhesi, S. Y. Martin, T. Wagner,
G. Cerullo, and A. V. Kimel, Nat. Commun. 7, 10645 (2016). J. Wunderlich, F. Freimuth, Y. Mokrousov, J. Kuneš, J. S. Chauhan,
472
K. Grishunin, T. Huisman, G. Li, E. Mishina, T. Rasing, A. V. Kimel, M. J. Grzybowski, A. W. Rushforth, K. W. Edmonds, B. L. Gallagher, and
K. Zhang, Z. Jin, S. Cao, W. Ren, G.-H. Ma, and R. V. Mikhaylovskiy, ACS T. Jungwirth, Science 351, 587 (2016).
Photon. 5, 1375 (2018). 478
S. Y. Bodnar, L. Šmejkal, I. Turek, T. Jungwirth, O. Gomonay, J. Sinova,
473
S. S. Dhillon, M. S. Vitiello, E. H. Linfield, A. G. Davies, M. C. Hoffmann, A. A. Sapozhnik, H.-J. Elmers, M. Kläui, and M. Jourdan, Nat. Commun. 9, 348
J. Booske, C. Paoloni, M. Gensch, P. Weightman, G. P. Williams, (2018).
479
E. Castro-Camus, D. R. S. Cumming, F. Simoens, I. Escorcia-Carranza, J. Grant, D. Tierno, M. Dekkers, P. Wittendorp, X. Sun, S. C. Bayer, S. T. King, S. Van
S. Lucyszyn, M. Kuwata-Gonokami, K. Konishi, M. Koch, C. A. Schmuttenmaer, Elshocht, M. Heyns, I. P. Radu, and C. Adelmann, IEEE Trans. Ultrason.
T. L. Cocker, R. Huber, A. G. Markelz, Z. D. Taylor, V. P. Wallace, J. A. Zeitler, Ferroelectr. Freq. Control 65, 881 (2018).

J. Appl. Phys. 128, 161101 (2020); doi: 10.1063/5.0019328 128, 161101-44

Published under license by AIP Publishing.