Submitted to: New J. Phys.

Diode Effect for Skyrmions Interacting with Linear

Protrusion Defects
arXiv:2508.12090v1 [cond-mat.mes-hall] 16 Aug 2025

J. C. Bellizotti Souza1,∗ , C. J. O. Reichhardt2 , C. Reichhardt2 , N. P.

Vizarim1 , P. A. Venegas 3
E-mail: [email protected]
1 “Gleb Wataghin” Institute of Physics, University of Campinas, 13083-859 Campinas, São

Paulo, Brazil
2 Theoretical Division and Center for Nonlinear Studies, Los Alamos National Laboratory,

Los Alamos, New Mexico 87545, USA

3 Department of Physics, São Paulo State University (UNESP), School of Sciences, Bauru

17033-360, SP, Brazil

∗ Corresponding author

Abstract. We simulate collectively interacting skyrmions in a channel with periodic

asymmetry, and find a strong diode effect for the skyrmion flow. There is also an asymmetry in
the skyrmion annihilation rate for currents applied along the hard or easy substrate asymmetry
direction, with a higher annihilation rate for hard direction currents. We map out the diode
efficiency as a function of magnetic field and substrate asymmetry angle. We also show that the
Magnus force impacts the diode motion and annihilation rate asymmetry by forcing skyrmions
into corners of the protrusion geometry.

Keywords: Diode effects, Skyrmions, Annihilation

Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 2

1. Introduction

Diode effects are characterized by flow that is easy in one direction and hindered in the
other. In electric diodes produced with a p-n junction [1], there is an easy flow of electrons
when the junction is positively polarized, and no flow under negative polarization. The use
of diodes is essential for modern hardware logic devices. Diode-like behavior has been
observed in systems ranging from superconductors [2, 3, 4, 5], colloidal particles [6], fluids
[7, 8], photonics [9, 10, 11], and thermodynamics [12, 13]. For overdamped systems, such as
superconducting vortices, diode effects can be produced using a combination of asymmetric
potentials, collective particle interactions, and dc driving [3, 4], and the diode effect intensity
can be tuned by varying the number of particles that are present.
More recently, skyrmionic diodes were proposed for ferromagnetic skyrmion systems
[14, 15, 16, 17, 18, 19, 20, 21]. Skyrmions are particle-like magnetic textures that form
triangular lattices in clean samples [22, 23, 24] and can be set into motion by an applied spin
current [25, 26]. When skyrmions interact with defects or interfaces, they can experience
pinning effects that cause the threshold current for motion to become finite and nonzero [27].
A key feature that distinguishes skyrmion dynamics from the dynamics of most other systems
is that skyrmion motion has a strong non-dissipative Magnus force component [24, 27, 28].
The Magnus force causes skyrmions to travel at a finite Hall angle with respect to an external
drive and also affects how the skyrmions interact with interfaces, pinning, and other skyrmions
[29, 30, 31]. The presence of the Magnus force removes the necessity of collective effects
usually required for diodes in overdamped systems. The Magnus force also increases the
velocity of skyrmions interacting with defects under external driving [19, 32, 33, 34], making
it possible to achieve rapid skyrmion motion in skyrmionic diode devices.
In recent studies of skyrmion dynamics, ratchet effects were observed under different
potential landscapes, currents and magnetic fields [35, 36, 37, 38, 39, 40, 41, 42, 43, 21, 44,
45]. Systems in which ratchet effects appear under ac driving are promising candidates for
producing diode effects when collective interactions are introduced. For example, after ratchet
effects were identified for individual skyrmions moving over a funnel shaped drive under ac
driving [42], it was shown that multiple interacting skyrmions in the same potential could
generate diode effects under dc driving. [19]. Ratchet effects of both superconducting vortices
[46] and magnetic skyrmions [43] interacting with a linear protrusion defect arrangement have
been demonstrated under ac driving. Here, we show that the same type of linear protrusion
defect array can give diode motion for multiple interacting skyrmions under dc driving.
We use atomistic based simulations to investigate skyrmion dynamics in the presence of
an array of linear protrusion defects under dc driving. Under positive currents, the skyrmions
flow in the +x direction with velocities of order 20 m s−1 , and the average velocity increases
linearly with applied current. For negative currents the skyrmions can move along the −x
direction at reduced velocities. This −x motion occurs only for a small current interval, and
the flow disappears for most negative current values. The magnetic skyrmion diode effect
we observe is characterized by rapid flow along the easy or +x direction and diminished or
vanishing flow along the hard or −x direction. The rate at which skyrmions annihilate also
depends on the direction of the skyrmion flow, and is less rapid for +x flow and more rapid
for −x flow, with complete skyrmion annihilation occurring for large negative currents. The
skyrmion trajectories indicate that interactions with the protrusion defects generate a strong
Magnus velocity boost for easy direction flow, but that during hard direction flow the moving
skyrmions instead interact with pinned skyrmions. The angle of the linear protrusions strongly
affects the velocity of the +x flow, since protrusion defects with small angles are more closely
aligned with the drag forces on the skyrmions and can more effectively enhance the skyrmion
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 3

50 H
y(nm)

φ
0
0 50 100 150 200 250
x(nm)
Figure 1. Schematic of the defect arrangement used in this work. Gray regions indicate high
anisotropy regions. The angle φ is defined as the angle between the x axis and the protrusion.
The magnetic field points in the positive z direction, H = H ẑ.

velocity under positive currents. The linear protrusion angle has little to no effect on the
velocity of the −x flow. Larger protrusion angles reduce the annihilation probability for both
flow directions. Reducing the magnetic field does not modify the +x flow, but the resulting
softer skyrmions can deform to a greater extent during −x flow, enhancing the motion. The
skyrmion annihilation rate is also reduced when the magnetic field is lowered, indicating that
although larger magnetic fields stabilize stiffer skyrmions, these stiff skyrmions are also brittle
and are more susceptible to annihilation.

2. Simulation

We use atomistic simulations, which can capture the dynamics of individual atomic magnetic
moments [47], to model a ferromagnetic ultrathin film capable of holding Néel skyrmions.
Our sample has dimensions of 272 nm × 68 nm with periodic boundary conditions along the
x direction. We apply a magnetic field perpendicular to the sample along the +z direction and
work at zero temperature, T = 0 K.
The Hamiltonian governing the atomistic dynamics is given by [48, 47, 26]:

H = − ∑ Ji j mi · m j − ∑ Di j · (mi × m j ) (1)
⟨i, j⟩ ⟨i, j⟩

− ∑ µH · mi − ∑ Ki (mi · ẑ)2 .
i i

The ultrathin film is modeled as a square arrangement of atoms with a lattice constant
a = 0.5 nm. The first term on the right hand side of Eq. (1) is the exchange interaction
with an exchange constant of Ji j = J between magnetic moments i and j. The second
term is the interfacial Dzyaloshinskii–Moriya interaction, where Di j = Dẑ × r̂i j is the
Dzyaloshinskii–Moriya vector between magnetic moments i and j and r̂i j is the unit distance
vector between sites i and j. Here, ⟨i, j⟩ indicates that the sum is over only the first neighbors
of the ith magnetic moment. The third term is the Zeeman interaction with an applied external
magnetic field H. Here µ = gµB is the magnitude of the magnetic moment, g = |ge | = 2.002
is the electron g-factor, and µB = 9.27 × 10−24 J T−1 is the Bohr magneton. The last term
represents the sample perpendicular magnetic anisotropy (PMA), where we use two PMA
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 4

constants: K = 0.02J for i ∈

/ P, and K = 5J for i ∈ P, with P being the set of defects composing
the linear protrusion arrangement shown in Fig. 1. In ultrathin films, long-range dipolar
interactions act as a PMA (see Supplemental of Wang et al. [49]), and therefore merely
effectively shift the PMA values.
The time evolution of atomic magnetic moments is obtained using the Landau-Lifshitz-
Gilbert (LLG) equation [50, 51]:
∂ mi ∂ mi pa3
= −γmi × Heff
i + αm i × + (j · ∇) mi . (2)
∂t ∂t 2e
1 ∂H
Here γ = 1.76 × 1011 T−1 s−1 is the electron gyromagnetic ratio, Heff i = − µ ∂ mi is
the effective magnetic field including all interactions from the Hamiltonian, α is the
phenomenological damping introduced by Gilbert, and the last term is the adiabatic spin-
transfer-torque (STT) caused by application of an in plane spin polarized current, where p
is the spin polarization, e the electron charge, and j = jŷ the applied current density. Use
of this STT expression implies that the conduction electron spins are always parallel to the
magnetic moments m [48, 52]. In this work we only consider adiabatic STT contributions,
which corresponds to dragging forces perpendicular to j [26, 53], so dragging forces are along
x̂ in our system. It is possible to include non-adiabatic STT contributions in Eq. 2; however,
non-adiabatic STT does not appreciably affect the dynamics of nanoscale skyrmions at small
driving forces [30], which is the regime we consider. Other types of torques, such as spin
orbit torques, can also be investigated, but are not explored in the present work.
The skyrmion velocity is computed using the emergent electromagnetic fields [50, 54]:

h̄ ∂m ∂m
 
Eiem= m· × (3)
e ∂i ∂t
h̄ m ∂m
 
em ∂
Bi = εi jk m · × , (4)
2e ∂j ∂k
where εi jk is the totally anti-symmetric tensor. The skyrmion drift velocity, vd , is then
calculated using Eem = −vd × Bem .
We fix the following values in our simulations: α = 0.3 and p = −1.0. The material
parameters are J = 1 meV and D = 0.2J. The numerical integration of Eq. 2 is performed
using a fourth order Runge-Kutta method. For each set of µH and φ , the system is initialized
with a triangular skyrmion lattice, and then relaxed for 10 ns to allow the skyrmions to adjust
to the presence of the linear protrusion defects shown in Fig. 1. To ensure a steady state for
measurement we evolve Eq. 2 over 200 ns.

3. Diode effect

We first consider a diode effect in a sample with a magnetic field of µH = 0.5D2 /J and a
protrusion angle of φ = 45◦ . In Fig. 2(a) we plot the average skyrmion velocity ⟨vx ⟩ versus
current amplitude j for positive and negative currents. Under a positive current with j > 0,
⟨vx ⟩ increases linearly with j, reaching ⟨vx ⟩ ≈ 19 m s−1 at j = 5 × 1010 A m−2 . For negative
currents with j < 0, |⟨vx ⟩| slowly increases until j = −1 × 1010 A m−2 , and then it decreases
to the value |⟨vx ⟩| = 0 m s−1 at and above | j| = 2 × 1010 A m−2 . The significant difference
in absolute velocity between positive and negative currents indicates that a diode effect is
present, where the flow is easier along the +x direction and harder along the −x direction.
The plot of the annihilation probability p0 versus current amplitude j in Fig. 2(b) again
indicates that positive and negative currents produce differing behavior. Under a positive
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 5

(a) (b)
j>0
j<0 0.8
15

0.6
hvxi(m s−1)

10

p0
0.4
5
0.2

0
0.0
0 1 2 3 4 5 0 1 2 3 4 5
| j|(1010A m−2) | j|(1010A m−2)

Figure 2. (a) Skyrmion average velocity along x, ⟨vx ⟩, and (b) annihilation probability p0 vs
the magnitude j of the applied current in a sample with φ = 45◦ and µH = 0.5D2 /J. Black:
positive currents j > 0; red: negative currents j < 0.

current, we find nonzero skyrmion annihilation above j = 1 × 1010 A m−2 , while the onset
of annihilation begins at a lower current amplitude of | j| = 0.5 × 1010 A m−2 for negative
currents. In each case, p0 increases rapidly at lower | j| and then approaches a saturation value
of p0 = 1 with increasing | j|. For negative currents, the saturation of p0 is complete at and
above | js | = 3.5 × 1010 A m−2 , but we do not reach complete saturation of p0 in the range
of positive currents considered here. For all values of | j|, the rate of skyrmion annihilation is
greater for negative than for positive currents.
To better describe the annihilation process, in Fig. 3 we plot the skyrmion trajectories at
selected values of j. There is no skyrmion annihilation at j = 1 × 1010 A m−2 in Fig. 3(a), and
the skyrmions interact strongly with the linear protrusion defects while flowing, resulting in
a velocity increase via the Magnus velocity boost effect [19, 32, 33, 34]. Three well defined
flow channels are present, with the first and second closely following the protrusion defect
profile, and the third interacting with skyrmions in stagnant areas of the channel. Near the
protrusions, we observe compression of the skyrmions similar to those reported by Souza et
al. [56], where a skyrmion size gradient is present. In Fig. 3(b) at j = −1 × 1010 A m−2 ,
a portion of the skyrmions are annihilated during the motion. The skyrmions flow primarily
in a single channel along the bottom of the sample. A steady state flow is established in
which one skyrmion becomes trapped in the acute corner of the bottom linear protrusion
defects. This skyrmion undergoes large periodic fluctuations in size as the other skyrmions
in the moving channel travel past. At the same time, a trio of skyrmions becomes pinned
in each acute corner of the upper linear protrusion defects as a consequence of the chirality
exhibited by the skyrmions when interacting with defects [57]. At j = 2.5 × 1010 A m−2
in Fig. 3(c), an even greater amount of skyrmion annihilation occurs and the formation of
more than one flowing lane becomes impossible. Due to the sign of the current, this lane
follows the contour of the upper linear protrusion defects. When the current is reversed to
j = −2.5 × 1010 A m−2 , Fig. 3(d) shows that nearly all of the skyrmions annihilate, and
the surviving skyrmions become pinned in the upper and lower acute angles of the linear
protrusion defects.
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 6

(a)
50
y(nm)

25

0
(b)
50
y(nm)

25

0
(c)
50
y(nm)

25

0
(d)
50
y(nm)

25

0
0 50 100 150 200 250
x(nm)
Figure 3. Skyrmion trajectories from the system in Fig. 2 with φ = 45◦ and µH = 0.5D2 /J
at (a) j = 1 × 1010 A m−2 , (b) j = −1 × 1010 A m−2 , (c) j = 2.5 × 1010 A m−2 , and (d)
j = −2.5 × 1010 A m−2 . Blue dots and trajectories: skyrmions that did not annihilate. Red
dots and trajectories: skyrmions that annihilated during the flow. Animations of the skyrmion
dynamics can be found in the supplemental material [55].

4. Varied protrusion angle φ

We next study the effect on ⟨vx ⟩ and p0 of modifying the protrusion angle φ . Figure 4(a)
shows ⟨vx ⟩ versus j for different values of φ . Under positive currents j > 0, the magnitude
of the velocity decreases with increasing φ . For example, at j = 5 × 1010 A m−2 , we find
⟨vx ⟩ ≈ 24 m s−1 when φ = 30◦ , ⟨vx ⟩ ≈ 19 m s−1 when φ = 45◦ , ⟨vx ⟩ ≈ 15 m s−1 when
φ = 60◦ , and ⟨vx ⟩ ≈ 3.5 m s−1 when φ = 90◦ . In contrast, for negative currents j < 0, the
velocity is only weakly sensitive to the value of φ , and in almost all cases ⟨vx ⟩ = 0 m s−1
when | j| ≥ 2 × 1010 A m−2 . The exception is that at φ = 90◦ , the + j and − j behaviors are
almost perfectly mirrored across the zero velocity line. The small differences that we find
in these two curves are expected to appear as a consequence of the skyrmion chirality when
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 7

(a) (b) 1.0

φ = 90◦
20 φ = 60◦
0.8
φ = 45◦
15 φ = 30◦
hvxi(m s−1)

0.6

p0
10
0.4
5

0.2
0

0.0
0 1 2 3 4 5 0 1 2 3 4 5
| j|(1010A m−2) | j|(1010A m−2)

Figure 4. (a) ⟨vx ⟩ and (b) p0 vs the magnitude of the applied current j in a sample with
µH = 0.5D2 /J at different protrusion angles φ = 90◦ (black), φ = 60◦ (red), φ = 45◦ (blue),
and φ = 30◦ (green). Solid lines with circles represent positive currents with j > 0, while
dashed lines with diamonds represent negative currents with j < 0.

(a)
50
y(nm)

25

0
(b)
50
y(nm)

25

0
(c)
50
y(nm)

25

0
(d)
50
y(nm)

25

0
(e)
50
y(nm)

25

0
(f)
50
y(nm)

25

0
0 50 100 150 200 250
x(nm)

Figure 5. Skyrmion trajectories from the system in Fig. 4 with µH = 0.5D2 /J at (a) j = 2.5 ×
1010 A m−2 and φ = 90◦ , (b) j = −2.5 × 1010 A m−2 and φ = 90◦ , (c) j = 2.5 × 1010 A m−2
and φ = 60◦ , (d) j = −2.5 × 1010 A m−2 and φ = 60◦ , (e) j = 2.5 × 1010 A m−2 and φ = 30◦ ,
and (f) j = −2.5 × 1010 A m−2 and φ = 30◦ . Blue dots and trajectories: skyrmions that did not
annihilate. Red dots and trajectories: skyrmions that annihilated during the flow. Animations
of the skyrmion dynamics can be found in the supplemental material [55].
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 8

interacting with defects [57].

In Fig. 4(b) we plot the annihilation probability p0 versus j for different protrusion angles
φ . The positive j > 0 current behavior for all φ values is very similar to the behavior of the
φ = 45◦ system in Fig. 2(b), where p0 increases rapidly with increasing j at low currents
before beginning to saturate toward the value p0 = 1 at larger j. For a given value of j, the
annihilation probability decreases as φ increases. The onset of skyrmion annihilation falls at
j = 0.5 × 1010 A m−2 for φ = 30◦ and 90◦ , but shifts to higher values of j = 1 × 1010 A m−2 .
for φ = 45◦ and 60◦ . Figure 2(b) shows that negative currents with j < 0 also give behavior
similar to that shown in Fig. 2(b), with a more rapid increase in p0 with increasing | j| at low
| j| compared to the positive current system, followed by a saturation to p0 = 1 at large | j|. We
again observe that, for a fixed value of j, increasing φ decreases the annihilation probability.
Skyrmion annihilation occurs for all negative current values when φ = 30◦ , while at φ = 45◦ ,
60◦ , and 90◦ , the onset of annihilation is at | j| = 0.5 × 1010 A m−2 . For all simulated values
of φ , the annihilation rate is greater for negative currents than for positive currents, with the
exception of φ = 90◦ , where the annihilation for j > 0 and j < 0 is approximately the same
over the entire range of j considered.
In order to better understand the results of Fig. 4, in Fig. 5 we plot representative
skyrmion trajectories for different φ values at a fixed current magnitude of | j| = 2.5 ×
1010 A m−2 . In Fig. 5(a), at φ = 90◦ and a positive j > 0 current, the majority of the
skyrmions annihilate, while the remaining skyrmions form a single flowing channel following
the upper contour of the linear protrusion defects. Some skyrmions remain pinned along the
lower region of the sample as a consequence of their interactions with the defects. For a
negative j < 0 current at φ = 90◦ , shown in Fig. 5(b), a similar behavior occurs with the
roles of the top and bottom walls of the sample reversed. For both φ = 90◦ systems, when
skyrmions interact with the corners at which the linear protrusions meet the confining upper
and lower walls, their velocities are increased as a result of a Magnus velocity boost. The boost
comes only from the corners and does not occur along the length of the linear protrusions.
In Fig. 5(c), for a positive j > 0 current in a system with φ = 60◦ , approximately half of
the skyrmions annihilate. The surviving skyrmions gradually assemble into a single flowing
channel following the upper contour of the protrusion defects. Here the Magnus velocity
boost becomes very important, since it is produced not only at the corners where the linear
protrusions emerge from the boundary walls, but also along the length of the linear protrusion
defects. For negative j < 0 currents at φ = 60◦ , Fig. 5(d) shows that the majority of the
skyrmions annihilate, while the surviving skyrmions become pinned inside the upper contour
of the protrusions and cease to move. At φ = 30◦ and positive j > 0 currents in Fig. 5(e),
approximately 60% of the skyrmions annihilate. We again find that the surviving skyrmions
travel along the upper contour of the protrusion defects, similar to the behavior in Figs. 3(a,
c) and Fig. 5(c). The skyrmion velocity is enhanced by the Magnus velocity boost, but the
skyrmions have a higher velocity than in the φ = 45◦ and φ = 60◦ samples, indicating that the
magnitude of the Magnus velocity boost is φ dependent. In Fig. 5(f) at φ = 30◦ and negative
j < 0 current, a majority of the skyrmions annihilate and the remaining skyrmions become
pinned by the protrusion defects.

5. Varied Magnetic field µH

Next we fix φ = 45◦ and vary µH for different values of j. Figure 6(a) shows ⟨vx ⟩ versus j
at applied fields of µH = 0.3D2 /J, 0.4D2 /J, 0.5D2 /J, and 0.6D2 /J. Under positive currents
j > 0, ⟨vx ⟩ is affected very little by changes to the magnetic field, unlike the sensitivity of
⟨vx ⟩ to the protrusion angle φ displayed in Fig. 4(a). A small amount of variation in ⟨vx ⟩
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 9

(a) (b) 1.0

µ H = 0.3D2/J
15 µ H = 0.4D2/J 0.8
µ H = 0.5D2/J
µ H = 0.6D2/J
hvxi(m s−1)

10 0.6

p0
0.4
5

0.2
0
0.0
0 1 2 3 4 5 0 1 2 3 4 5
| j|(1010A m−2) | j|(1010A m−2)

Figure 6. (a) ⟨vx ⟩ and (b) p0 vs the magnitude of the applied current j in a sample with φ = 45◦
at different magnetic fields, µH = 0.3D2 /J (black), µH = 0.4D2 /J (red), µH = 0.5D2 /J
(blue), and µH = 0.6D2 /J (green). Solid lines with circles represent positive currents with
j > 0, while dashed lines with diamonds represent negative currents with j < 0.

appears only at large values of j, where increasing µH modestly increases the velocity.
Much more significant changes in ⟨vx ⟩ appear under negative currents j < 0, where both
the maximum value of |⟨vx ⟩| and the range of currents over which the velocity is nonzero
increase with decreasing magnetic field. There is a small increase in the maximum velocity
from |⟨vx ⟩| ≈ 0.3 m s−1 at µH ≥ 0.4D2 /J to |⟨vx ⟩| ≈ 0.4 m s−1 at µH = 0.3D2 /J. The window
of nonzero velocity expands significantly as the magnetic field decreases. For µH = 0.3D2 /J,
0.4D2 /J, 0.5D2 /J, and 0.6D2 /J, the width ∆ j of the finite velocity window is, respectively,
∆ j = 3 × 1010 A m−2 , 2.5 × 1010 A m−2 , 2 × 1010 A m−2 , and 1.5 × 1010 A m−2 . When
the magnetic field is smaller, or if the anisotropy is decreased, the skyrmions become softer
[56, 58], permitting deformations to become important for enabling the flow of skyrmions
along the hard or negative x direction. Our results show that it is possible to tune the magnetic
field for the purpose of controlling the motion of the skyrmions along the hard direction, so
that the skyrmions are mobile under low fields but become pinned under higher fields.
Figure 6(b) shows the annihilation probability p0 versus the magnitude of the applied
current j for different values of µH. Positive currents with j > 0 exhibit the same behavior
described in Fig. 2(b) and Fig. 4(b), with a rapid increase of p0 with increasing current for
low j, followed by saturation towards p0 = 1 at large j. As the magnetic field increases, the
overall magnitude of p0 increases somewhat, indicating that more annihilation is occurring at
larger µH. The onset of skyrmion annihilation is weakly affected by changing the magnetic
field, falling at j = 0.5 × 1010 A m−2 for µH = 0.6D2 /J and at j > 1 × 1010 A m−2 for
µH < 0.6D2 /J. For negative currents with j < 0, we observe a rapid increase in p0 at lower
| j|, and saturation towards p0 = 1 at larger | j|, similar to the behavior in Fig. 2(b) and Fig. 4(b).
We find that, for a given current amplitude, the annihilation probability for j < 0 is larger than
that for j > 0.
In Fig. 7 we plot the skyrmion trajectories for | j| = 2.5 × 1010 A m−2 and φ = 45◦ under
different magnetic fields µH. At µH = 0.3D2 /J and j > 0 in Fig. 7(a), approximately 45%
of the skyrmions annihilate. The remaining skyrmions exhibit a Magnus velocity boost and
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 10
(a)
50

y(nm)
25

0
(b)
50

y(nm)
25

0
(c)
y(nm) 50

25

0
(d)
50
y(nm)

25

0
(e)
50
y(nm)

25

0
(f)
50
y(nm)

25

0
0 50 100 150 200 250
x(nm)

Figure 7. Skyrmion trajectories from the system in Fig. 6 with φ = 45◦ at (a) j = 2.5 ×
1010 A m−2 and µH = 0.3D2 /J, (b) j = −2.5 × 1010 A m−2 and µH = 0.3D2 /J, (c)
j = 2.5 × 1010 A m−2 and µH = 0.4D2 /J, (d) j = −2.5 × 1010 A m−2 and µH = 0.4D2 /J, (e)
j = 2.5 × 1010 A m−2 and µH = 0.6D2 /J, and (f) j = −2.5 × 1010 A m−2 and µH = 0.6D2 /J.
Blue dots and trajectories: skyrmions that did not annihilate. Red dots and trajectories:
skyrmions that annihilated during the flow. Animations of the skyrmion dynamics can be
found in the supplemental material [55].

form two channels of flow while interacting with the defects, with the first channel following
the upper contour of the protrusion defects, and the second channel following the contour of
the first channel. The two channels merge and mix each time the skyrmions reach the upper
edge of a protrusion defect, and then reform when the skyrmions arrive at the next protrusion
defect. For a negative j < 0 current at µH = 0.3D2 /J, Figure 7(b) shows that approximately
80% of the skyrmions annihilate. One portion of the remaining skyrmions becomes pinned
near the corners of the protrusion defects, and facilitates the flow of the rest of the skyrmions
along the hard axis by pushing the moving skyrmions over the ends of the lower protrusion
defects. Both the pinned and moving skyrmions deform strongly during this motion, and in
some cases the moving skyrmions adapt to the shape of the protrusion barrier in the course
of jumping over the barrier with assistance from the pinned skyrmions. In Fig. 7(c), for a
positive j > 0 current at µH = 0.4D2 /J, approximately 60% of the skyrmions annihilate.
The motion is similar to that found in Fig. 7(a), with the skyrmions experiencing a Magnus
velocity boost while flowing along the easy axis. There are again two channels of flow, but
now the two channels merge before the interaction between the skyrmions and protrusions
ceases, indicating that the existence of two [or three, Fig. 3(a)] channels of flow depends on the
rigidity of the skyrmions. Figure 7(d) shows a system with µH = 0.4D2 /J for a negative j < 0
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 11

current, where approximately 80% of the skyrmions annihilate. No persistent motion appears
and all of the surviving skyrmions become pinned. The larger value of the magnetic field
causes the skyrmions to become more rigid, with reduced deformations, which destroys the
hard direction skyrmion flow that was observed at µH = 0.3D2 /J in Fig. 7(b). In Fig. 7(e) at
µH = 0.6D2 /J under a positive j > 0 current, approximately 55% of the skyrmions annihilate.
As in Fig. 7(a, c), the skyrmions flow along the easy direction and experience a Magnus
velocity boost. The flow now follows a single channel, with skyrmions moving closer to the
upper contour of the protrusion defects and walls. The absence of a second channel of flow
enhances the interactions of the skyrmions with the protrusions, resulting in the appearance of
a stronger Magnus velocity boost. Under a negative j < 0 current at µH = 0.6D2 /J, Fig. 7(f)
shows that approximately 85% of the skyrmions annihilate. As was the case in Fig. 7(d), no
persistent motion occurs due to the increased rigidity of the skyrmions at the larger magnetic
field.
To summarize our findings, in Fig. 8(a) we plot a heat map of ⟨vx ⟩ as a function of φ
versus j for systems with µH = 0.5D2 /J. The velocity reaches its greatest magnitude for low
values of φ , but due to the absence of a hard flow direction for φ ≤ 10◦ , no diode effect is
present. A hard flow direction emerges once φ > 10◦ , and the diode effect becomes visible,
as indicated by the appearance of lower velocity magnitudes for negative currents than for
positive currents. In the small hatched region with φ ≤ 30◦ and j ≤ −4 × 1010 A m−2 , all
of the skyrmions annihilate; very large current amplitudes must be applied to achieve this
annihilation. The diode effect occurs across most of the φ and ± j pairs in the phase diagram,
but the intensity of the diode motion varies with φ . In particular, the diode effect weakens
as φ approaches φ = 90◦ , and is lost at φ = 90◦ where ⟨vx ⟩ becomes symmetric for positive
and negative currents. We expect that a mirrored version of the heat map would appear for
protrusion defect angles φ > 90◦ . The corresponding heat map of the annihilation rate p0 ,
plotted as a function of φ versus j in Fig. 8(b), shows that there is a higher annihilation rate
for negative currents, indicated by the darker areas in the j < 0 region.
In Fig. 8(c) we show a heat map of ⟨vx ⟩ as a function of µH versus j at fixed φ = 45◦ ,
and the corresponding heat map of p0 as a function of µH versus j appears in Fig. 8(d). The
diode effect is present for all values of µH, as indicated by Fig. 8(c), but the magnitude of
the diode effect increases with increasing µH, as shown by the diminishing magnitude of ⟨vx ⟩
with increasing µH in the j < 0 region of the plot. This reduction in the magnitude of the
velocity with increasing magnetic field is a consequence of the stiffening of the skyrmions and
their diminished ability to deform as the magnetic field becomes higher. Figure 8(d) shows
that the annihilation rate is greater for negative currents, as indicated by the darker areas in
the j < 0 portion of the plot. The annihilation rate also increases with increasing magnetic
field, and complete annihilation of all skyrmions occurs for large current amplitudes in the
high magnetic field regime. This behavior is associated with the brittleness of the skyrmions.
At higher magnetic fields, the skyrmions become smaller and more rigid; however, under a
sufficiently large dragging force produced by the applied current, they can become brittle and
annihilate upon interacting with the protrusion defects.

6. Summary

Using atomistic simulations, we investigated the dynamics of multiple skyrmions interacting

with anisotropy defects forming a linear protrusion arrangement under adiabatic spin-transfer-
torque currents. We observe a diode effect in which the skyrmion velocity |⟨vx ⟩| is large
for positive j > 0 currents applied along the easy or +x direction but reduced for negative
j < 0 currents applied along the hard or −x direction. The flow direction strongly affects
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 12

hvxi(m s−1) p0
−20 0 20 0.0 0.2 0.4 0.6 0.8 1.0

(a) (b)
80

60
φ (◦)

40

20

0
(c) (d)
0.7

0.6
µ H(D2/J)

0.5

0.4

0.3
−4 −2 0 2 4 −4 −2 0 2 4
j(1010A m−2) j(1010A m−2)

Figure 8. Heat maps of (a, c) the skyrmion average velocity along x, ⟨vx ⟩ and (b, d) the
annihilation probability p0 plotted as a function of (a, b) protrusion angle φ vs applied current
j or (c, d) magnetic field µH vs applied current j. In (a, b) we fix µH = 0.5D2 /J, and in (c, d)
we fix φ = 45◦ . The velocity color map ranges from dark blue through white to dark red, while
the annihilation probability color map ranges from white to dark red. In the hatched regions in
(a, c), there are no surviving skyrmions present.

the skyrmion annihilation rate p0 , with annihilation occurring more rapidly and beginning at
lower current magnitudes for negative currents than for positive currents. The annihilation rate
increases rapidly with increasing current magnitude for low | j|, and saturates toward p0 = 1
as | j| increases. Skyrmions flowing along the easy +x direction experience a Magnus velocity
boost during their entire motion, while skyrmions flowing along the hard −x direction do not.
The presence of skyrmions pinned in the corners of the protrusion defects is necessary to
facilitate the motion of the skyrmions in the hard −x direction, since collective interactions
between skyrmions of oscillating size are needed to push the moving skyrmions over the hard
direction barrier. If the annihilation rate is too large, no hard direction flow can occur because
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 13

not enough skyrmions are left in the system, and the remaining skyrmions are all pinned
by the defects. The value of the protrusion angle φ strongly affects ⟨vx ⟩ for easy +x flow
since a reduction in φ increases the skyrmion velocity due to the Magnus velocity boost. For
negative −x flow, ⟨vx ⟩ is largely insensitive to the φ . Larger φ values decrease the skyrmion
annihilation rate for both positive and negative currents, The exception is vertical φ = 90◦
protrusions, which have the greatest annihilation rate and also show no diode effect due to
the lack of asymmetry. When we decrease the magnetic field, for positive currents ⟨vx ⟩ is
slightly reduced, while for negative currents the window of finite ⟨vx ⟩ values is significantly
extended. Thus, applying a larger magnetic field enhances the diode effect by reducing the
range of j for which motion along the hard axis can occur. Larger magnetic fields stabilize
more rigid skyrmions, while smaller magnetic fields produce softer skyrmions that can deform
more easily and are better able to travel in the −x direction. The skyrmions also become
more brittle when the magnetic field is large, and the annihilation rate p0 is enhanced.
Under extremely large magnetic fields, there is complete annihilation of the skyrmions for
large current amplitudes. We expect our results to be useful for designing skyrmion diode
devices. The velocity dependence of φ together with the ability to control the range of j over
which hard direction motion occurs by changing µH can provide a device with fine tuning
mechanisms.

Acknowledgments

This work was supported by the US Department of Energy through the Los Alamos National
Laboratory. Los Alamos National Laboratory is operated by Triad National Security, LLC,
for the National Nuclear Security Administration of the U. S. Department of Energy (Contract
No. 892333218NCA000001). J.C.B.S and N.P.V. acknowledge funding from Fundação de
Amparo à Pesquisa do Estado de São Paulo - FAPESP (Grants J.C.B.S 2023/17545-1 and
2022/14053-8, N.P.V 2024/13248-5). We would like to thank FAPESP for providing the
computational resources used in this work (Grant: 2024/02941-1).

Data Availability Statement

Data available on request from the authors.

References
[1] Kitai A 2011 Principles of Solar Cells, LEDs and Diodes: The role of the PN junction (John Wiley & Sons)
[2] Lyu Y Y, Jiang J, Wang Y L, Xiao Z L, Dong S, Chen Q H, Milošević M V, Wang H, Divan R, Pearson J E, Wu
P, Peeters F M and Kwok W K 2021 Nat. Commun. 12 2703
[3] Olson Reichhardt C J and Reichhardt C 2013 J. Supercond. Nov. Magn. 26 2005–2008
[4] Reichhardt C and Olson Reichhardt C J 2010 Physica C 470 722–725
[5] Harrington S A, MacManus-Driscoll J L and Durrell J H 2009 Appl. Phys. Lett. 95 22518
[6] Reichhardt C J O and Reichhardt C 2018 J. Phys. Condens. Matter 30 244005
[7] Shou D and Fan J 2018 Adv. Funct. Mater. 28 1800269
[8] Mates J E, Schutzius T M, Qin J, Waldroup D E and Megaridis C M 2014 ACS Appl. Mater. Interfaces 6
12837–12843
[9] Tocci M D, Bloemer M J, Scalora M, Dowling J P and Bowden C M 1995 Appl. Phys. Lett. 66 2324–2326
[10] Scalora M, Dowling J P, Bowden C M and Bloemer M J 1994 J. Appl. Phys. 76 2023–2026
[11] Wang D W, Zhou H T, Guo M J, Zhang J X, Evers J and Zhu S Y 2013 Phys. Rev. Lett. 110 93901
[12] Li B, Wang L and Casati G 2004 Phys. Rev. Lett. 93 184301
[13] Martínez-Pérez M J, Fornieri A and Giazotto F 2015 Nature Nanotech. 10 303–307
[14] Feng Y, Zhang X, Zhao G and Xiang G 2022 IEEE Trans. Electron. Devices 69 1293
[15] Song L, Yang H, Liu B, Meng H, Cao Y and Yan P 2021 J. Magn. Magn. Mater. 532 167975
Diode Effect for Skyrmions Interacting with Linear Protrusion Defects 14

[16] Jung D H, Han H S, Kim N, Kim G, S J, Lee S, Kang M, Im M Y and Lee K S 2021 Phys. Rev. B 104 L060408
[17] Shu Y, Li Q, Xia J, Lai P, Hou Z, Zhao Y, Zhang D, Zhou Y, Liu X and Zhao G 2022 Appl. Phys. Lett. 121
42402
[18] Zhao L, Liang X, Xia J, Zhao G and Zhou Y 2020 Nanoscale 17 9507
[19] Souza J C B, Vizarim N P, Reichhardt C J O, Reichhardt C and Venegas P A 2022 New J. Phys. 24 103030
[20] Bellizotti Souza J C, Vizarim N P, Reichhardt C J O, Reichhardt C and Venegas P A 2023 J. Phys.: Condens.
Matter 35 015804
[21] Souza J C B, Vizarim N P, Reichhardt C J O, Reichhardt C and Venegas P A 2025 J. Phys. Condens. Matter 37
165801
[22] Mühlbauer S, Binz B, Jonietz F, Pfleiderer C, Rosch A, Neubauer A, Georgii R and Böni P 2009 Science 323
915–919
[23] Yu X Z, Onose Y, Kanazawa N, Park J H, Han J H, Matsui Y, Nagaosa N and Tokura Y 2010 Nature 465
901–904
[24] Nagaosa N and Tokura Y 2013 Nature Nanotechnol. 8 899–911
[25] Jonietz F, Mühlbauer S, Pfleiderer C, Neubauer A, Münzer W, Bauer A, Adams T, Georgii R, Böni P, Duine
R A, Everschor K, Garst M and Rosch A 2010 Science 330 1648–1651
[26] Iwasaki J, Mochizuki M and Nagaosa N 2013 Nature Nanotechnol. 8 742–747
[27] Reichhardt C, Reichhardt C J O and Milošević M 2022 Rev. Mod. Phys. 94 035005
[28] Everschor-Sitte K, Masell J, Reeve R M and Klaüi M 2018 J. Appl. Phys. 124 240901
[29] Jiang W, Zhang X, Yu G, Zhang W, Wang X, Jungfleisch M B, Pearson J E, Cheng X, Heinonen O, Wang K L,
Zhou Y, Hoffmann A and te Velthuis S G E 2017 Nature Phys. 13 162–169
[30] Litzius K, Lemesh I, Krüger B, Bassirian P, Caretta L, Richter K, Büttner F, Sato K, Tretiakov O A, Förster
J, Reeve R M, Weigand M, Bykova I, Stoll H, Schütz G, Beach G S D and Kläui M 2017 Nature Phys. 13
170–175
[31] Brearton R, Turnbull L A, Verezhak J A T, Balakrishnan G, Hatton P D, van der Laan G and Hesjedal T 2021
Nature Commun. 12 2723
[32] Iwasaki J, Koshibae W and Nagaosa N 2014 Nano Lett. 14 4432–4437
[33] Chen X, Kang W, Zhu D, Zhang X, Lei N, Zhang Y, Zhou Y and Zhao W 2017 Appl. Phys. Lett. 111 202406
[34] Castell-Queralt J, Gonzalez-Gomez L, Del-Valle N, Sanchez A and Navau C 2019 Nanoscale 11 12589–12594
[35] Yamaguchi R, Yamada K and Nakatani Y 2020 Jpn. J. Appl. Phys. 60 10904
[36] Reichhardt C, Ray D and Reichhardt C J O 2015 Phys. Rev. B 91 184502
[37] Chen W, Liu L and Zheng Y 2020 Phys. Rev. Applied 14 64014
[38] Chen W, Liu L, Ji Y and Zheng Y 2019 Phys. Rev. B 99 64431
[39] Reichhardt C, Ray D and Reichhardt C J O 2015 New J. Phys. 17 73034
[40] Ma X, Reichhardt C J O and Reichhardt C 2017 Phys. Rev. B 95 104401
[41] Göbel B and Mertig I 2021 Sci. Rep. 11 3020
[42] Souza J C B, Vizarim N P, Reichhardt C J O, Reichhardt C and Venegas P A 2021 Phys. Rev. B 104 54434
[43] Souza J C B, Vizarim N P, Reichhardt C J O, Reichhardt C and Venegas P A 2024 Phys. Rev. B 109 054407
[44] Migita K, Yamada K and Nakatani Y 2020 Appl. Phys. Express 13 73003
[45] Moon K W, Kim D H, Je S G, Chun B S, Kim W, Qiu Z Q, Choe S B and Hwang C 2016 Sci. Rep. 6 20360
[46] Wells F S University of Wollongong, 2018
[47] Evans R F L 2018 Atomistic Spin Dynamics Handbook of Materials Modeling: Applications: Current and
Emerging Materials ed Andreoni W and Yip S (Springer Nature Switzerland AG) pp 1–23
[48] Iwasaki J, Mochizuki M and Nagaosa N 2013 Nature Commun. 4 1463
[49] Wang X S, Yuan H Y and Wang X R 2018 Commun. Phys. 1 31
[50] Seki S and Mochizuki M 2016 Skyrmions in Magnetic Materials (Springer)
[51] Gilbert T L 2004 IEEE Trans. Mag. 40(6) 3443–3449
[52] Zang J, Mostovoy M, Han J H and Nagaosa N 2011 Phys. Rev. Lett. 107(13) 136804
[53] Feilhauer J, Saha S, Tobik J, Zelent M, Heyderman L J and Mruczkiewicz M 2020 Phys. Rev. B 102(18) 184425
[54] Schulz T, Ritz R, Bauer A, Halder M, Wagner M, Franz C, Pfleiderer C, Everschor K, Garst M and Rosch A
2012 Nature Phys. 8 301–304
[55] See supplemental material for animations showing the motion of the skyrmions.
[56] Souza J C B, Vizarim N P, Reichhardt C J O, Reichhardt C and Venegas P A 2023 New J. Phys. 25 053020
[57] Zhang X, Xia J, Tretiakov O A, Ezawa M, Zhao G, Zhou Y, Liu X and Mochizuki M 2023 Nano Lett. 23
11793–11801
[58] Liu L, Chen W and Zheng Y 2023 Phys. Rev. Lett. 131 246701