Low-Field Regime of Magnon Transport in Yttrium Iron Garnet

Hossein Taghinejad,1,2,* Kohtaro Yamakawa,1,3 Xiaoxi Huang,4 Yuanqi Lyu,1,3 Luke P. Cairns,1,3
Ramamoorthy Ramesh,1,3,4 James G. Analytis1,2,3,*

1. Department of Physics, University of California, Berkeley, CA, USA.

2. Kavli Energy NanoSciences Institute, University of California, Berkeley, CA, USA.
3. Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA.
4. Department of Materials Science & Engineering, University of California, Berkeley, CA, USA.

*Corresponding Authors:
[email protected]
[email protected]

ABSTRACT: Diffusive propagation of spin waves and their quanta, magnons, in the archetypal

magnetic insulator yttrium iron garnet (YIG) are the subject of a surge of research for low-power

and low-loss data communication. However, operation under external magnetic fields reduces

magnon diffusion length, attenuates the voltage amplitude at measurement terminals, and

complicates the architecture of magnonic devices. Here, we explore the low-field and field-free

regime of diffusive magnon transport in YIG films. We demonstrate that the field-induced

suppression of magnon diffusion length can be fully inhibited only at the zero-field limit. Even a

modest field of 10mT attenuates the non-local spin voltage by ~20% in a transport channel of ~

1µm long. Using Stoner-Wohlfarth macrospin simulations, we reveal that an often overlooked,

in-plane uniaxial anisotropy becomes the critical parameter governing the field-free operation of

magnonic devices. We further demonstrate a tenfold enhancement in the effective field

associated with the in-plane uniaxial anisotropy of YIG films at low temperatures—a key finding

for field-free operation of magnonic devices under cryogenic conditions.

1
 Spin waves, the collective excitations of magnetic materials, and their quanta—magnons—

are currently the focus of intense research as potential alternatives to electronic charge for data

communication. The excitement surrounding this paradigm shift stems from the wide range of

possibilities that spin waves could unlock. First, spin waves offer a fundamental solution to the

problem of heat generation in electronic devices, an issue that has become more critical since the

sunset of the Dennard's scaling law. [1] Communicating data through spin waves of magnetic

insulators eliminate the need for electronic charge movement during device operation, hence

efficiently reducing the Joule heating issue encountered in charge-based electronics. Second, the

wave nature of magnons opens up alternative computation paradigms, where the phase and

frequency can be exploited to encode information. This capability is gaining momentum for non-

Boolean computing, particularly in emerging machine-learning applications. Third, spin waves

support high-speed operation, with frequencies ranging from gigahertz in dipolar magnons to the

terahertz range in exchange magnons. [2-4] The recent report of coherent generation of THz

waves could further fuel this direction. [5] These promising features have driven extensive

research into the use of magnons for communication, processing, and storage of information,

giving rise to the field of magnonics.

Ideal material candidates for magnonic devices typically meet three key criteria: (i) they

should be electronically insulating to prevent parasitic effects from charge conduction, (ii) they

must exhibit low magnetic damping to allow magnons to propagate over long distances, and (iii)

they should support room-temperature operation. While the number of materials meeting these

requirements is steadily growing, [6-9] yttrium iron garnet (YIG, Y3Fe5O12) remains the gold

standard in magnonics. YIG possesses the lowest magnetic damping of any known material, [10,

11] enabling robust magnon propagation over millimeter-scale distances [3, 12] with a giant spin

2
conductivity. [13] Combined with its desired insulating electronic properties (bandgap ≈ 2.8eV)

[14] and high Curie temperature (TC ≈ 560 K), [15] YIG has become the archetypal platform for

magnonic research. Spin waves in YIG are traditionally excited inductively via the Orsted field

generated by current-carrying wires fabricated on YIG surface. [16-18] While this method is

straightforward for proof-of-concept studies, the extended footprint of the Oersted field hinders

high-density device integration. An alternative approach relies on the integration of heavy metals

like platinum (Pt) with YIG for localized excitation and detection of magnons by exploiting the

spin Hall effect (SHE) and interfacial spin accumulation at the metal-YIG interface. [19] This

method has gained considerable interest due to its scalability and potential for dense device

integration.

In the linear regime, the localized nature of magnon generation via SHE in heavy metals

leads to a diffusive propagation mode that is characterized by a characteristic magnon diffusion

length. [19] Diffusive transport experiments are often conducted under the application of an

external magnetic field, which poses several drawbacks. The application of the magnetic field

reduces the magnon diffusion length in YIG, which, by extension, attenuates the amplitude of the

voltage produced by travelling magnons at a non-local detector placed away from the injection

site. This is a major challenge, as the low signal-amplitude is already a critical issue in

spintronics, and applying magnetic fields worsens the problem. Cornelissen et al. studied this

effect in medium-to-high magnetic fields (from 10mT to 1T), showing a severe attenuation of the

detected voltage. [20] A similar effect is reported in other garnet films such as Tm3Fe5O12. [9] It

is also important to consider that high-quality YIG films are often grown on gadolinium gallium

garnet (GGG, Gd₃Ga₅O₁₂) substrates. At high magnetic fields, the GGG substrate plays a non-

trivial role in magnon transport within the YIG layer. Although GGG lacks exchange stiffness, at

3
high fields it can contribute paramagnons to the hybrid YIG/GGG transport channel. [21]

Additionally, at high fields, enhanced dipolar coupling between the GGG substrate and the YIG

film can induce parasitic effects in YIG, [22] particularly at low temperatures where quantum

magnonics is of interest. Application of large magnetic fields also complicate magnonic device

design by significantly altering magnon dispersion, affecting key parameters such as the

excitation energy gap, group velocity, and the equilibrium density of magnons. [23, 24] Finally,

relying on external magnetic fields for device operation complicates the architecture of magnonic

devices and increases power consumption. Therefore, the low-field, and ideally zero-field,

operation is essential for unlocking the full potential of magnonic platforms.

Here, we investigate the non-local transport of diffusive magnons in YIG thin films at the

ultra-low magnetic field regime, below 1 mT, and down to the zero-field limit. Our experiments

reveal that the suppression of non-local voltage due to the magnetic field persists down to the

zero-field limit, highlighting the critical importance of field-free operation of magnonic devices.

In such an extremely low-field regime, we demonstrate that the weak, often overlooked in-plane

uniaxial magnetic anisotropy of YIG becomes the key factor governing details of non-local

magnon transport. To elucidate the role of this anisotropy for zero-field operation, we develop a

Stoner-Wohlfarth (SW) macrospin model that not only quantifies YIG’s in-plane uniaxial

anisotropy but also allows conceiving device concepts based on the geometric twist between the

device orientation and YIG's in-plane easy magnetic axis. The success of SW model in

explaining our experimental data has important implications: in the low-field regime, incoherent

spins waves can serve as a probe for coherent magnetization switching in YIG thin films.

Combined with harmonic analysis of spin signals, this finding effectively positions non-local

4
magnon transport platforms as magnetometers seamlessly embedded within magnonic devices,

expanding their utility beyond conventional spintronic applications.

Figure 1(a) presents an optical image of a representative magnon transport device, consisting

of two Pt wires with dimensions of 50 µm × 400 nm × 7 nm patterned on an 80 nm-thick YIG

film. One Pt wire serves as the magnon injector, and the second wire acts as the detector,

separated by a distance ~1 µm. As established in previous studies, [13, 19, 25] the application of

a charge current to the Pt injector drives the propagation of magnons in the underlying YIG film

via two distinct mechanisms. The first mechanism is the magnon-Seebeck effect, a thermally

driven process. Here, the charge current induces Joule heating near the injector wire, creating a

temperature gradient that propels magnons towards the detector. The second mechanism is an

electronic excitation that leverages the spin Hall effect (SHE) within the Pt wire. The strong

spin-orbit coupling in Pt generates a transverse spin current when a charge current is applied.

This spin current flows towards the Pt – YIG interface, leading to spin accumulation. Through

interfacial exchange coupling, this spin accumulation interacts with the local moments in YIG

and serves as a reservoir of angular momentum for the excitation of magnons. As such, a spin-
ℏ ℏ
flip scattering (e.g., from + " to − ") in Pt transfers one unit of angular momentum (+ℏ) from the

spin accumulation to the YIG film, leading to the out-of-equilibrium excitation of a magnon

directly underneath the injector wire. Such a localized excitation creates a gradient in the

magnon density, driving their diffusive propagation towards the detector wire.

At the detector, the propagating magnons are absorbed by conduction electrons in the Pt

wire, resulting in spin accumulation at the Pt – YIG interface. This spin accumulation

subsequently induces a measurable open-circuit voltage along the Pt wire via the inverse SHE.

To differentiate between the two excitation mechanisms, we employ the lock-in detection

5
technique and perform the harmonic analysis of the open-circuit voltage. Accordingly, driving an

AC charge current at a frequency f0 = 7.77 Hz into the injector wire generates two voltage

components at the detector: a first harmonic component (V1f, at frequency f0) corresponding to

the out-of-equilibrium magnons electronically excited via the SHE, and a second harmonic

component (V2f, at frequency 2f0) representing magnons thermally excited via the spin-Seebeck

effect. [19, 25]

As depicted in Figure 1(a), our measurements are performed under an external magnetic

field, Bext, applied within the device plane. When Bext is sufficiently strong, YIG’s magnetic

moments (MYIG) align with the external field. Thus, rotating Bext by an angle φB rotates MYIG and

generates characteristic cos²(φB) and cos(φB) angular dependences for the first and second

harmonic voltages at the detector. [19] In Figure 1(b), we have demonstrated such angular

characteristics for several magnetic fields. A critical observation, however, is the scaling of the

V1f and V2f voltages with Bext. As shown in Figure 1(c), increasing the magnetic field suppresses

the amplitude of the non-local voltage, particularly the first harmonic component. Notably, the

V1f decreases by over 80% as the magnetic field approaches 0.5 T. This significant reduction is

attributed, phenomenologically, to the reduced diffusion length of magnons at high magnetic

fields. [20] In contrast, the suppression of the V2f signal is much weaker and onsets only at

relatively larger fields compared to V1f. This discrepancy stems from the direct heat diffusion

towards the detector wire, and the localized generation of spin-Seebeck voltage underneath the

detector. [20] Thus, the field-induced reduction of magnon diffusion length impacts the V2f

voltage much less than the V1f component. The suppression of non-local voltages becomes more

pronounced as the distance between the injector and detector wires increases (Figure S1,

Supporting Information (SI)). These trends emphasize the need for exploring the magnon

6
transport in YIG under lower magnetic fields. Otherwise, high field operation significantly

diminishes signal amplitude, which impedes the use of magnons for long-distance data

communication and efficient computing in next-generation electronic devices.

We next focus on magnon transport in the low-field regime (Bext < 5mT). We start with field-

rotation experiments and study the angular characteristics of the non-local voltages under small

fields. As shown in Figure 2(a, b), reducing Bext results in a progressive deviation of the V1f and

V2f components from the characteristic cos²(φB) and cos(φB) line-shapes, indicating that the

azimuthal angle of the YIG magnetization (φYIG) no longer aligns with that of the external

magnetic field (i.e., φYIG ≠ φB). This deviation becomes more pronounced at smaller magnetic

fields, particularly at Bext = 150 µT. At such low fields, the angular dependence of V1f and V2f

voltages exhibit two abrupt jumps separated by 180o. Additionally, these jumps show hysteretic

behaviors in the bidirectional scan of φB between 0o and 360o. We attribute these observations to

the competing effect of the in-plane magnetic anisotropy in the YIG film. When Bext becomes

comparable to the anisotropy field, rotating the magnetic field across the hard magnetic axis

triggers sudden magnetization switching from one quadrant to the next, which leads to abrupt

jumps in the V1f and V2f signals. The significance of such an in-plane magnetic anisotropy is

further confirmed from the different coercive fields required to saturate the non-local voltage

when Bext is applied in different in-plane directions. For instance, a coercive field of

approximately 150 µT is enough to saturate the V2f signal when Bext is applied at φB = 0o (Figure

2c), whereas a coercive field exceeding 500 µT is required when Bext is applied at φB = 90o

(Figure 2d). Such direction-dependent coercivity leads to the observed distortion of V1f and V2f

line-shapes, as shown in Figure 2(a, b).

7
 To further elucidate the role of in-plane uniaxial magnetic anisotropy of YIG, we conduct

macrospin simulations based on the SW model. Considering the coordinate system shown in

Figure 3(a, b), the system’s energy (E), normalized to the effective YIG volume (Veff), can be

described as:

(𝐸/𝑉#$$ ) = −𝑀%&' 𝐵#() cos(𝜑%&' − 𝜑* ) + 𝐾sin" (𝜑%&' − 𝜑+, ), (1)

where the first term represents the Zeeman energy, and the second term corresponds to the

anisotropy energy. In Equation (1), K and φEA are, respectively, anisotropy energy and the

direction of the magnetic easy axis relative to the [1126] crystal axis of the YIG film. Figure 3(a)

illustrates the crystal axes of the YIG film, highlighting [111] direction as normal to the film, and

[1126] and [1160] directions aligning with the edges of the YIG substrate. The Pt wires are aligned

parallel to the [1160] crystal axis. As depicted in Figure 3(b), all angles in our macrospin

simulations are measured relative to the [1126] direction.

Within the described simulation framework, we calculate the system’s energy as a function

of Bext, φB, K, φEA, and φYIG. By identifying local minima in the energy landscape, while

accounting for the history of the applied magnetic field, we determine the equilibrium position of

the YIG magnetization, φYIG. This process generates a set of φYIG values that we use to simulate

the experimental V1f and V2f voltages and, hence, calculate the K and φEA parameters. Figure 3(c)

illustrates this process, where green circles mark the equilibrium position of the magnetization as

Bext = 150 µT is rotated from φB = 0o to φB = 360o in 10o incremental steps. Figure 3(d, e) shows

the best fits of the SW macrospin model to the first and second harmonic voltages, respectively,

yielding K = (16 ± 2) J/m3 and φEA = 30o ± 5o.

8
 Our simulations accurately capture the key experimental features, namely the angular

dependences, abrupt voltage jumps, and the hysteretic behaviors of V1f and V2f signals observed

during bidirectional field rotation (see, Figure S2, SI). Additionally, the simulations successfully

reproduce the experimental hysteresis loops shown in Figure 2(c, d). These agreements validate

that our model is suitable for analyzing the role of in-plane anisotropy of YIG in non-local

magnon transport experiments within the low-field regime. Thus, we rely on the SW fitting of

the experimental data to quantize the scaling of the V1f voltage as Bext approaches zero, while

fully accounting for the role of the anisotropy field. As shown in the highlighted part of Figure

1(c), the V1f signal increases asymptotically, but noticeably, by more than 20% as the external

field is reduced from 10 mT to 150 µT. This key finding reveals that the field-dependence of the

magnon diffusion length persists even in the limit of vanishingly small magnetic fields, further

highlighting the importance of zero-field operation. Field-free operation not only inhibits the loss

of the signal amplitude but also significantly simplifies the architecture of magnonic devices by

eliminating the need for supplying external magnetic fields, hence reducing the overall system

complexity.

The in-plane magnetic anisotropy of YIG is critical for the field-free operation of magnonic

devices. To further elaborate this point, we analyze three device configurations where Pt wires

are geometrically twisted relative to the easy axis of the YIG film (Figure 4(a-c)). Using the

macrospin model biased with the empirical values of K = 16 J/m3 and φEA = 30o, we simulate the

hysteresis behavior of V1f and V2f voltages in response to an applied magnetic field. In all cases,

Bext is applied within the YIG plane and perpendicular to the Pt wires. Distinct characteristics are

observed across the three configurations. In the A-configuration (Figure 4(a)), where the Pt wires

are oriented perpendicular to the easy axis of YIG, the V2f signal exhibits a box-like hysteresis,

9
while the V1f signal remains constant at its maximum value for all values of Bext. In contrast, in

the C-configuration (Figure 4(c)), where the Pt wires are aligned parallel to the YIG’s easy axis,

V2f and V1f signals exhibit linear and parabolic field dependencies, respectively, without any

hysteresis. The intermediate case, the B-configuration (Figure 4(b)), with Pt wires twisted 45°

relative to the easy axis, shows a hybrid behavior between these two extreme cases. We note that

the device structure measured in Figure 2 closely resembles the B-type configuration (with a

twist angle of ~ 30o). The successful simulation of the experimentally measured hysteresis loops

for this device (solid lines in Figures 2(c, d)) further validates that the simulations presented in

Figure 4 are reliable predictions of device performance in alternative designs.

A key parameter for the field-free operation is the amplitude of the V1f voltage in the

remnant state (i.e., at Bext = 0). In the A-configuration, the remnant V1f signal remains at its

maximum value without any drop. In the B-configuration, the remnant V1f decreases by 50%,

while in the C-configuration, the non-local signal drops to zero in the absence of an external

field. This systematic change in the remnant V1f voltage with the device twist angle stems from

the interfacial coupling between conduction electrons in Pt and local magnetic moments of the

underlying YIG layer. As schematically depicted in Figure 4(d), once the Pt wires are fabricated,

the SHE dictates a fixed polarization (𝜎⃗) for the spin accumulation at the Pt – YIG interface.

Consequently, the efficiency of magnon injection and detection, 𝜂-./,1#) , depends on the

orientation of the YIG magnetization and can be phenomenologically described as:

::⃗YIG).
𝜂-./,1#) = g(𝜎⃗.𝑀 (2)

In the limit of Bext = 0, the YIG magnetization aligns with the easy magnetic axis, that is

:::::::::⃗
𝑀%&' ∥ EA :::::⃗. Therefore, the twist angle between 𝜎⃗ and the magnetic easy axis, ΔφEA, becomes the

10
sole parameter determining the 𝜂-./,1#) and, by extension, the amplitude of the non-local voltage.

As plotted in Figure 4(e), Equation (2) yields an analytical expression for the coupling efficiency

in the form of 𝜂-./,1#) ∝ cos(ΔφEA), leading to a cos2(ΔφEA) modulation for the remnant value of

the first harmonic signal (V1f ∝ 𝜂-./ ×𝜂1#) ). Figure 4(e) also includes the remnant V1f values

obtained from our macrospin simulation for the three device configurations studied in Figure

4(a-c), demonstrating agreement between our simulations and the analytical equation. Thus, we

conclude that the 𝜎⃗ ∥ :::::⃗

EA (𝜎⃗ ⊥ :::::⃗
EA) condition in the A-configuration (C-configuration) leads to

the full retention (loss) of the non-local voltage in the limit of Bext → 0. Further predications

regarding the angular dependence of the first and second harmonic non-local signals for devices

with the A-, B-, and C-configuration are shown in Figure S3, SI.

We note that the parameter g in Equation (2) is a pre-factor accounting for the details of

spin-charge interconversion. It includes contributions from the spin diffusion length and Hall

angle of Pt wires and the Pt|YIG interfacial spin-mixing conductance. [19, 26] In the low-field

regime, we treat g as a constant parameter, an assumption that is supported by the field

independence of the V2f voltage as previously discussed in Figure 1. Due to the localized nature

::⃗YIG). Up to Bext ≈ 50 mT, the

of the spin-Seebeck effect, we can assume that V2f ∝ 𝜂1#) = g(𝜎⃗.𝑀

V2f signal exhibits no significant field dependence, suggesting that the g-factor remains largely

unaffected by the external field. In contrast to the V2f signal, the V1f component shows a

persistent field dependence down to zero field, indicating that the scaling of V1f with Bext

primarily stems from the influence of field on parameters related to the magnon transport within

the YIG channel, rather than the g factor, which is relevant only at the Pt-YIG interface.

11
 We emphasize that the SW macrospin model describes coherent magnetization switching,

where all spins switch direction nearly simultaneously. On the other hand, the SHE leads to

incoherent excitation of out-of-equilibrium magnons in YIG films, essentially generating a broad

spectrum of magnons with varying frequencies, phases, and linear momenta. Therefore, the

success of SW model in explaining our magnon transport experiments has a significant

implication: incoherently excited spins waves and their quanta, magnons, can effectively serve as

proxies for monitoring coherent magnetization switching in YIG thin films. While both first and

second harmonic signals can, in principle, provide such information, the V2f signal is particularly

advantageous due to its localized nature and its robustness against magnon transport parameters.

Given the relationship V2f ∝ g(σ.MYIG), with fixed values of g and σ, V2f reflects YIG’s

magnetization. As such, the second harmonic signal mimics the role of the classic vibrating

sample magnetometry (VSM) technique, but over a smaller length-scale defined by the size and

the footprint of the device under test. This approach complements the data obtained from VSM,

which averages over the entire material volume, thereby masking fine details and heterogeneities

within the sample. We note that the YIG film can be effectively treated as a monodomain over

our device footprint, supported by extensive studies showing that magnetic domains in YIG films

extend over hundreds of micrometers. [27] Thus, the MYIG probed via the V2f signal, near the

detector wire, properly represents the magnetization state of the YIG across the entire magnonic

transport channel, including near the injector wire. In other words, we can leverage the second

harmonic component as an embedded magnetometer within the magnon transport channel to

probe YIG’s magnetic properties concurrently with, yet independent of, the transport phenomena

revealed by the V1f harmonic. In the following, we showcase this capability by characterizing the

12
temperature-dependence of YIG’s in-plane anisotropy via analyzing V2f voltage at different

temperatures.

Despite its relatively low value of 16 J/m3 at 200K, the in-plane uniaxial anisotropy of YIG

dictates key characteristics of magnon transport at low- and zero-field regime. This observation

motivates further investigations into how uniaxial anisotropy energy scales at lower

temperatures. To achieve this, we rely on the temperature dependence of the second harmonic

signal. Figure 5(a) shows the response of the V2f signal to the external magnetic field at several

temperatures ranging from 300K to 50K. Notable changes in the width (ΔB) and height (ΔV2f) of

the hysteresis loops are observed. From ΔB we can obtain the switching field Bs = ΔB/2, the

external magnetic field at which switching of YIG’s magnetization to the opposite direction

occurs. The Bs value directly reflects the strength of YIG’s uniaxial magnetic anisotropy because

magnetization switching occurs when the Zeeman energy surpasses the anisotropy energy in

Equation (1). The extracted values of Bs and ΔV2f at different temperatures are plotted in Figure

5(b). By fitting our SW macrospin simulations to the experimental data, we extract the
2
anisotropy energy, K, and the effective anisotropy field, BK = 3! 4"#$
. The temperature

dependence of the YIG’s saturation magnetization is accounted for (see, Figure S4, SI). As

depicted in Figures 5(c) and 5(d), both K and BK increase rapidly as temperature decreases,

reaching saturation values of ~42 J/m3 and ~220 µT, respectively, at temperatures below 100K.

These values reflect more than 15- and 10-fold enhancement in the anisotropy energy and

effective anisotropy field compared to the room temperature values of 2.7 J/m3 and 20 µT,

respectively. Such an order-of-magnitude increase in the anisotropy field underscores the

importance of often overlooked in-plane uniaxial anisotropy of YIG films for low-temperature

and low-field magnonic applications.

13
 Our observation of uniaxial anisotropy aligns with previous studies on YIG thin films grown

by pulsed-laser deposition (PLD) on (111)-GGG substrates. [27, 28] However, from a

crystallographic standpoint, this finding is surprising. Ideally, the cubic structure of YIG favors

threefold magnetocrystalline anisotropy. To the first order, the anisotropy energy of a cubic

crystal can be expressed as Ecubic ∝ K1(𝛼5" 𝛼"" + 𝛼5" 𝛼6" + 𝛼"" 𝛼6" ), where 𝛼7 s are directional cosine

functions. Defining the polar angle θ relative to the [111] axis and the azimuthal angle φ relative

to the [1126] axis yields the following expression for the cubic anisotropy energy:

5 5 √"
Ecubic ∝ K1(8 sin8 (𝜃) + 6 cos 8 (𝜃) + 6
sin6 (𝜃) cos(𝜃) cos (3𝜑)) (3)

For the threefold anisotropy term (third term in Equation (3)) to appear, the magnetization must

obtain an out-of-plane component, meaning that θ ≠ π/2. While this condition can be met in bulk

YIG, the strong demagnetization field and easy-plane anisotropy field in thin films generally

constrain magnetization to the YIG plane, thus effectively quenching the threefold in-plane

anisotropy.

We attribute the uniaxial in-plane anisotropy of YIG films to extrinsic growth-induced

effects. However, we believe the root cause of this effect is not due to sporadic factors such as

material heterogeneities. Instead, it likely arises from a systematic, underlying mechanism

intrinsic to the pulsed-laser deposition (PLD) of YIG films. Interestingly, the easy-axis

orientation, φEA = 30o ± 5o relative to the [1126], obtained from our non-local magnon transport

experiments matches those extracted from magneto-optical Kerr effect (MOKE) [28] and spin

magnetoresistance (SMR) [27] measurements. The unanimity of conclusions made via three

different techniques (MOKE, SMR, and magnon transport) on independently grown PLD films

reinforces our hypothesis of a systematic origin for the in-plane uniaxial anisotropy. The 30o

14
offset from the [1126] direction aligns with the octahedral Fe–O bonds in the YIG lattice. Such

octahedral Fe atoms are prone to re-sputtering during the PLD process, [29] creating vacancies

that have been associated with the rhombohedral distortion and residual strain in PLD-grown

YIG films on (111)-GGG substrates. [28, 30] Therefore, we speculate that this vacancy-induced

strain contributes to the emergence of the in-plane uniaxial anisotropy observed in PLD-grown

(111)-YIG films. Additionally, the temperature-dependence of the uniaxial anisotropy in Figure

5(c, d) closely mirrors the behavior of YIG’s magneto-elastic coefficient, [31, 32] both showing a

sharp increase as temperature drops from the room temperature and saturating below 100K. This

correlation further implies strain-related effects as a key factor in the observed anisotropy.

Nonetheless, a dedicated study is required to fully understand this mechanism.

In conclusion, we explored non-local propagation of diffusive magnons under the

influence of external magnetic fields, demonstrating the continued attenuation of spin signals at

measurement terminals down to the zero-field limit. This finding carries a significant

implication; optimal signal amplitude in magnonic devices is achieved at zero-field conditions.

This conclusion addresses a major challenge in spintronics, the signal amplification, which is a

barrier currently limiting the widespread adoption of spintronic technologies. Our experiments

further emphasized the essential role of in-plane uniaxial anisotropy in understanding the

unconventional behavior of non-local magnon transport in the low-field regime. Through semi-

empirical micromagnetic simulations, we demonstrated how this anisotropy can be leveraged to

effectively design the coupling efficiency between spin reservoirs (i.e., Pt wires) and the magnon

transport channel (i.e., YIG) thorough geometric twist in device orientation. This approach offers

a practical pathway to designing complex magnonic devices without requiring external magnetic

bias, thus advancing the development of efficient, field-free magnonic systems. We also

15
presented evidence suggesting an extrinsic origin, likely growth-induced strain, as the underlying

mechanism driving the observed uniaxial in-plane anisotropy in YIG thin films. While this

anisotropy is subtle at room temperature, it becomes pronounced at lower temperatures. Our

findings advocate for the strategic use of strain-engineering in garnet films to enable field-free

magnonic operations. Potential approaches include applying epitaxial strain through substrate

engineering and targeted elemental doping to tailor the magnetic properties effectively.

Methods Section:
Growth of Thin Films. YIG thin films were grown on (111)-oriented GGG substrates using

reflection high-energy electron diffraction (RHEED)-assisted pulsed laser deposition technique.

The deposition was carried out at a substrate temperature of 700 °C under an oxygen partial

pressure of 100 mTorr. A chemically stoichiometric ceramic YIG target was employed. A KrF

excimer laser (λ = 248 nm) with an energy density of 1.5 J/cm² and a repetition rate of 5 Hz was

used to ablate the target. After deposition, the films were cooled at a controlled rate of 10 °C/min

under atmospheric oxygen conditions.

Device Fabrication. We employed two electron-beam lithography (EBL) steps to fabricate

the devices on the YIG substrate. In the first EBL step, the Pt injector/detector wires (50 µm ×

400 nm) are defined. Following the pattering, a 7 nm-thick Pt layer was deposited via RF

sputtering at a base pressure better than 2×10-6 Torr. Then, the sample underwent lift-off by

soaking in acetone overnight. In the second EBL step, electrical leads to the Pt wires and contact

pads were patterned. Subsequently, a 40nm/5nm Au/Ti bilayer was deposited in an e-beam

evaporator at a base pressure better than 5×10-6 Torr. The Ti layer serves as the adhesion layer.

Prior to the Au/Ti deposition, a gentle Ar+ ion milling was performed to ensure a clean interface

between Au/Ti and the underlaying Pt layer. Poly(methyl methacrylate) (PMMA) was used as the

16
electron resist in both EBL steps. The fabricated devices were wire bonded to a measurement

puck for measurements.

Non-local Transport Measurements. Non-local magnon transport measurements were

conducted using the lock-in detection technique. A 0.6 mA AC charge current at a frequency of f0

= 7.77 Hz was applied to the Pt injector wire to excite magnons in the underlying YIG film. At

the detector wire, two lock-in amplifiers were used for the simultaneous measurement of the first

and second harmonic components at frequencies of f0 and 2f0, respectively. The harmonic signals

were found to be independent of the frequency used in our lock-in measurements (Figure S5, SI).

All measurements were carried out under vacuum within a Quantum Design PPMS system. To

study the angular dependence of the first and second harmonic signals, samples were rotated

within a fixed magnetic field.

Stoner – Wohlfarth Macrospin Simulation. To obtain the equilibrium direction of the

YIG magnetization from SW model, we calculate the first and second derivatives of Equation (1)

with respect to φYIG. For each field condition (Bext, φB), the equilibrium position is identified

when both derivatives are positive. In our calculations, we consider the temperature dependence

of YIG’s saturation magnetization. To account for the slight misalignment of the substrate

relative to the applied field direction, cause by sample loading on the measurement puck, we

apply a small correction to the apparent φB values.

Acknowledgements:
JGA and HT acknowledge financial support from the Bakar Institute through the Bakar
Prize. HT acknowledges the financial support of the Kavli Energy NanoScience Institute (ENSI)
through the Heising-Simons Postdoctoral Fellowship at the University of California, Berkeley.

17
 References:
(1) Dennard, R. H.; Gaensslen, F. H.; Yu, H.-N.; Rideout, V. L.; Bassous, E.; LeBlanc, A. Design of
Ion-Implanted MOSFET’S with Very Small Physical Dimensions. IEEE J Solid-State Circuits
1974, 9 (5), 256-268.
(2) Mahmoud, A.; Ciubotaru, F.; Vanderveken, F.; Chumak, A. V.; Hamdioui, S.; Adelmann, C.;
Cotofana, S. Introduction to Spin Wave Computing. J Appl Phys 2020, 128 (16), 161101.
(3) Kajiwara, Y.; Harii, K.; Takahashi, S.; Ohe, J.; Uchida, K.; Mizuguchi, M.; Umezawa, H.; Kawai,
H.; Ando, K.; Takanashi, K.; Maekawa, S.; Saitoh, E. Transmission of Electrical Signals by Spin-
Wave Interconversion in a Magnetic Insulator. Nature 2010, 464 (7286), 262–266.
(4) Serga, A. A.; Chumak, A. V.; Hillebrands, B. YIG Magnonics. J Phys D Appl Phys 2010, 43 (26),
264002.
(5) Taghinejad, M.; Xia, C.; Hrton, M.; Lee, K.-T.; Kim, A. S.; Li, Q.; Guzelturk, B.; Kalousek, R.;
Xu, F.; Cai, W.; Lindenberg, A. M.; Brongersma, M. L. Determining Hot-Carrier Transport
Dynamics from Terahertz Emission. Science 2023, 382 (6668), 299-305.
(6) Lebrun, R.; Ross, A.; Bender, S. A.; Qaiumzadeh, A.; Baldrati, L.; Cramer, J.; Brataas, A.; Duine,
R. A.; Kläui, M. Tunable Long-Distance Spin Transport in a Crystalline Antiferromagnetic Iron
Oxide. Nature 2018, 561 (7722), 222–225.
(7) Yuan, W.; Zhu, Q.; Su, T.; Yao, Y.; Xing, W.; Chen, Y.; Ma, Y.; Lin, X.; Shi, J.; Shindou, R.; Xie,
X. C.; Han, W. Experimental Signatures of Spin Superfluid Ground State in Canted
Antiferromagnet Cr2O3 via Nonlocal Spin Transport. Sci. Adv 2018, 4 (4),1098.
(8) Parsonnet, E.; Caretta, L.; Nagarajan, V.; Zhang, H.; Taghinejad, H.; Behera, P.; Huang, X.;
Kavle, P.; Fernandez, A.; Nikonov, D.; Li, H.; Young, I.; Analytis, J.; Ramesh, R. Nonvolatile
Electric Field Control of Thermal Magnons in the Absence of an Applied Magnetic Field. Phys
Rev Lett 2022, 129 (8), 087601.
(9) Gao, J.; Lambert, C. H.; Schlitz, R.; Fiebig, M.; Gambardella, P.; Vélez, S. Magnon Transport
and Thermoelectric Effects in Ultrathin Tm3Fe5O12/Pt Nonlocal Devices. Phys Rev Res 2022, 4
(4), 043214.
(10) Cherepanov, V.; Kolokolov, I.; L’vov, V. The Saga of YIG: Spectra, Thermodynamics, Interaction
and Relaxation of Magnons in a Complex Magnet. Phys Reports 1993, 229 (3), 81-144.
(11) Glass, H. L.; Elliott, M. T. Attainment of the Intrinsic FMR Linewidth in Yttrium Iron Garnet
Films Grown by Liquid Phase Epitaxy. J Crys Growth 1976, 34 (2), 285-288.
(12) Maendl, S.; Stasinopoulos, I.; Grundler, D. Spin Waves with Large Decay Length and Few 100
nm Wavelengths in Thin Yttrium Iron Garnet Grown at the Wafer Scale. Appl Phys Lett 2017,
111 (1), 012403.

18
(13) Wei, X. Y.; Santos, O. A.; Lusero, C. H. S.; Bauer, G. E. W.; Ben Youssef, J.; van Wees, B. J.
Giant Magnon Spin Conductivity in Ultrathin Yttrium Iron Garnet Films. Nat Mater 2022, 21
(12), 1352–1356.
(14) Thiery, N.; Naletov, V. V.; Vila, L.; Marty, A.; Brenac, A.; Jacquot, J. F.; De Loubens, G.; Viret,
M.; Anane, A.; Cros, V.; Ben Youssef, J.; Beaulieu, N.; Demidov, V. E.; Divinskiy, B.;
Demokritov, S. O.; Klein, O. Electrical Properties of Epitaxial Yttrium Iron Garnet Ultrathin
Films at High Temperatures. Phys Rev B 2018, 97 (6), 064422.
(15) Anderson, E. E. Molecular Field Model and the Magnetization of YIG. Phys Rev 1964, 134 (6),
A1581-A1585.
(16) Morrison, N.; Taghinejad, H.; Analytis, J.; Ma, E. Y. Coherent Spin Wave Excitation with Radio-
Frequency Spin-Orbit Torque. J Appl Phys 2024, 136 (11), 113901.
(17) Pirro, P.; Vasyuchka, V. I.; Serga, A. A.; Hillebrands, B. Advances in Coherent Magnonics.
Nature Reviews Materials. 2021, 6 (12), 1114–1135.
(18) Chumak, A. V.; Vasyuchka, V. I.; Serga, A. A.; Hillebrands, B. Magnon Spintronics. Nat Phys
2015, 11(6), 453–461.
(19) Cornelissen, L. J.; Liu, J.; Duine, R. A.; Youssef, J. Ben; Van Wees, B. J. Long-Distance
Transport of Magnon Spin Information in a Magnetic Insulator at Room Temperature. Nat Phys
2015, 11 (12), 1022–1026.
(20) Cornelissen, L. J.; Van Wees, B. J. Magnetic Field Dependence of the Magnon Spin Diffusion
Length in the Magnetic Insulator Yttrium Iron Garnet. Phys Rev B 2016, 93 (2), 020403.
(21) Oyanagi, K.; Takahashi, S.; Cornelissen, L. J.; Shan, J.; Daimon, S.; Kikkawa, T.; Bauer, G. E.
W.; van Wees, B. J.; Saitoh, E. Spin Transport in Insulators without Exchange Stiffness. Nat
Commun 2019, 10 (1).
(22) Serha, R. O.; Voronov, A. A.; Schmoll, D.; Verba, R.; Levchenko, K. O.; Koraltan, S.;
Davídková, K.; Budinská, B.; Wang, Q.; Dobrovolskiy, O. V.; Urbánek, M.; Lindner, M.;
Reimann, T.; Dubs, C.; Gonzalez-Ballestero, C.; Abert, C.; Suess, D.; Bozhko, D. A.; Knauer, S.;
Chumak, A. V. Magnetic Anisotropy and GGG Substrate Stray Field in YIG Films down to
Millikelvin Temperatures. npj Spintronics 2024, 2 (29).
(23) Jin, H.; Boona, S. R.; Yang, Z.; Myers, R. C.; Heremans, J. P. Effect of the Magnon Dispersion
on the Longitudinal Spin Seebeck Effect in Yttrium Iron Garnets. Phys Rev B Condens Matter
Mater Phys 2015, 92 (5), 054436.
(24) Ritzmann, U.; Hinzke, D.; Kehlberger, A.; Guo, E. J.; Kläui, M.; Nowak, U. Magnetic Field
Control of the Spin Seebeck Effect. Phys Rev B Condens Matter Mater Phys 2015, 92 (17),
174411.
(25) Liu, J.; Cornelissen, L. J.; Shan, J.; Van Wees, B. J.; Kuschel, T. Nonlocal Magnon Spin
Transport in Yttrium Iron Garnet with Tantalum and Platinum Spin Injection/Detection
Electrodes. J Phys D Appl Phys 2018, 51 (22), 224005.
19
(26) Qiu, Z.; Ando, K.; Uchida, K.; Kajiwara, Y.; Takahashi, R.; Nakayama, H.; An, T.; Fujikawa, Y.;
Saitoh, E. Spin Mixing Conductance at a Well-Controlled Platinum/Yttrium Iron Garnet
Interface. Appl Phys Lett 2013, 103 (9), 092404.
(27) Mendil, J.; Trassin, M.; Bu, Q.; Schaab, J.; Baumgartner, M.; Murer, C.; Dao, P. T.; Vijayakumar,
J.; Bracher, D.; Bouillet, C.; Vaz, C. A. F.; Fiebig, M.; Gambardella, P. Magnetic Properties and
Domain Structure of Ultrathin Yttrium Iron Garnet/Pt Bilayers. Phys Rev Mater 2019, 3 (3),
034403.
(28) Sokolov, N. S.; Fedorov, V. V.; Korovin, A. M.; Suturin, S. M.; Baranov, D. A.; Gastev, S. V.;
Krichevtsov, B. B.; Maksimova, K. Y.; Grunin, A. I.; Bursian, V. E.; Lutsev, L. V.; Tabuchi, M.
Thin Yttrium Iron Garnet Films Grown by Pulsed Laser Deposition: Crystal Structure, Static,
and Dynamic Magnetic Properties. J Appl Phys 2016, 119 (2), 023903.
(29) Manuilov, S. A.; Khartsev, S. I.; Grishin, A. M. Pulsed Laser Deposited Y3Fe5O12 Films: Nature
of Magnetic Anisotropy I. J Appl Phys 2009, 106 (12), 123917.
(30) Manuilov, S. A.; Fors, R.; Khartsev, S. I.; Grishin, A. M. Submicron Y3Fe5O12 Film
Magnetostatic Wave Band Pass Filters. J Appl Phys 2009, 105 (3), 033917.
(31) Callen, H. B.; Callens, E. The Present Status of the Temperature Dependence of
Magnetocrystalline Anisotropy, and the l(l+1)2 Power Law. J Phys Chem Solid 1966, 27 (8),
1271-1285.
(32) Callen K, R. A.; Callen, H. B. Magnetostriction in Cubic Neel Ferrimagnets, with Application to
YIG. Phys. Rev. 1963, 130 (5), 1735-1740.

20
Figure 1: Attenuation of Non-Local Magnon Signal With Magnetic Field. (a) Optical image
of a representative non-local magnon transport device. Two Pt wires spaced 1 µm apart are used
for injection (top wire) and detection (bottom wire) of magnons in the underlying YIG film (~80
nm). Measurements are taken under an in-plane magnetic field (Bext) applied at various angles
(φB). (b) First (top) and second (bottom) harmonic voltages measured in response to rotating
magnetic fields of varying intensities. Solid lines represent V1f×cos²(φB) and V2f×cos(φB) fits to
the respective harmonic signals. (c) Scaling of V1f and V2f signals with the magnetic field
intensity. V1f and V2f signals at different magnetic fields are normalized to their respective
maximum values within each dataset. Solid lines are guides to the eyes. All measurements are
performed at 200 K.

21
Figure 2: Magnon Transport Under Low Magnetic Fields. (a, b) Angular dependence of the
first and second harmonic voltages, respectively, measured under low magnetic fields. For
clarity, collected data at different magnetic fields are shifted vertically. Deviation from
characteristic cos²(φB) and cos(φB) angular dependences and pronounced hysteresis are observed
at low magnetic fields. (c, d) Hysteresis loops of the second harmonic voltage, V2f, with the
magnetic field applied normal (i.e., φB = 0o) and parallel (i.e., φB = 90o) to Pt wires, respectively.
Black lines represent the SW macrospin simulations, as explained in the text. Arrows indicate the
direction of the magnetic field sweep.

22
Figure 3: Stoner – Wohlfarth (SW) Macrospin Simulations. (a) Schematic illustration of the
device geometry relative to the in-plane magnetic easy axis (EA) and crystal axes of the YIG
film. (b) Modeling parameters within the [1126] – [1160] plane, including intensity and angle of
external magnetic field (Bext and φB), equilibrium direction of YIG magnetization (MYIG, φYIG),
and the strength and direction of the uniaxial in-plane magnetic anisotropy (K, φEA). (c) Energy
landscapes obtained from Equation (1) when Bext = 150 µT is applied along φB = 0o-360o (in 10o
steps). The equilibrium direction of the YIG’s magnetization is identified as the metastable
energy minima, marked by green circles. (d) Simulating the first and second harmonic voltages
using the SW model yields K = 16 ± 2 J/m3 and φEA = 30o ± 5o.

23
Figure 4: Utility of Magnetic Anisotropy For Field-Free Magnon Transport. (a-c) Macrospin
simulation of hysteresis loops of V1f (top) and V2f (bottom) signals for three device
configurations with different twist angles between Pt wires and the easy axis. In the inset, the red
lines represent Pt wires, and the arrow marks the direction of the magnetic easy axis. In all
device configurations, Bext is applied in-plane and perpendicular to the Pt wires. (d) Definition of
the twist angle ΔφEA, the angle between the polarization (σ) of the spin accumulation in Pt and
the magnetic easy axis. Twist angles of ΔφEA = 0o, 45o, and 90o are assumed for configurations
shown in panels (a)-(c), respectively. (e) Modulation of the interfacial coupling efficiency (right
axis) and, by extension, the remnant value of the V1f voltage (left axis) with the twist angle. The
coupling efficiency assumes an analytical form of cos(ΔφEA), which leads to the cos2(ΔφEA)
dependence for the first harmonic signal. Three pentagon symbols represent the zero-field values
of V1f obtained from the macrospin simulation of devices shown in panels (a)-(c). The circle
shows the experimental Data obtained from our device (ΔφEA = 30o). Adjust main text for the
experimental data point on Panel (e)

24
Figure 5: Temperature Dependence of Magnetic Anisotropy. (a) Temperature dependence of
the hysteresis loop in V2f voltage. The height (ΔV2f) and width (ΔB) of the hysteresis loops are
indicated on the top panel. Red curves represent macrospin simulation fitted to the experimental
data. Bext is applied perpendicular to the Pt wires at φB = 0o. (b) Evolution of ΔV2f (right axis) and
the switching field Bs = ΔB/2 (left axis) with temperature. (c, d) Anisotropy energy and the
effective anisotropy field at different temperatures. These values are obtained by fitting the SW
model to the experimental data, while accounting for the temperature dependence of the YIG’s
saturation magnetization. A device with the injector – detector distance of ~0.5µm is used here.

25