Encoding and multiplexing information signals in

magnetic multilayers with fractional skyrmion tubes

Runze Chen1, Yu Li1, William Griggs1, Yuzhe Zang1, Vasilis F. Pavlidis2, and Christoforos

Moutafis1

1
Nano Engineering and Spintronic Technologies (NEST) research group, Department of

Computer Science, The University of Manchester, Manchester M13 9PL, United Kingdom

2
Advanced Processor Technologies (APT) research group, Department of Computer Science,

The University of Manchester, Manchester M13 9PL, United Kingdom

1
ABSTRACT

Tailored magnetic multilayers (MMLs) provide skyrmions with enhanced thermal stability,

leading to the possibility of skyrmion-based devices for room temperature applications. At the

same time, the search for additional stable topological spin textures has been under intense research

focus. Besides their fundamental importance, such textures may expand the information encoding

capability of spintronic devices. However, fractional spin texture states within MMLs in the

vertical dimension have yet to be investigated. In this work, we demonstrate numerically fractional

skyrmion tubes (FSTs) in a tailored MML system. We subsequently propose to encode sequences

of information signals with fractional skyrmion tubes (FSTs) as information bits in a tailored MML

device. Micromagnetic simulations and theoretical calculations are used to verify the feasibility of

hosting distinct FST states within a single device, and their thermal stability is investigated. A

multilayer multiplexing device is proposed, where multiple sequences of the information signals

can be encoded and transmitted based on the nucleation and propagation of packets of FSTs.

Finally, pipelined information transmission and automatic demultiplexing is demonstrated by

exploiting the skyrmion Hall effect and introducing voltage-controlled synchronizers and width-

based track selectors. The findings indicate that FSTs can be potential candidates as information

carriers for future spintronic applications.

2
INTRODUCTION

Magnetic skyrmions are particle-like topological spin configurations [1,2]. They are

promising candidates as information carriers in future information technologies, owing to their

topologically protected morphological stability and dynamical properties [2,3]. This stability

comes from the competition between magnetic interactions which favor colinear spin

configurations, such as Heisenberg exchange coupling and perpendicular magnetic anisotropy

(PMA), and interactions favoring orthogonal configurations such as dipolar coupling and

Dzyaloshinskii-Moriya interaction (DMI) [4,5]. In both bulk magnetic and magnetic multilayer

(MML) systems with broken symmetry, skyrmion spin states become one of the system energy

minima. The magnetic skyrmion can be described by an integer topological index, called the

skyrmion number N, which counts how many times the magnetisation wraps around a unit sphere.

The skyrmion number is defined as [6]:

1 𝜕𝒎 𝜕𝒎
𝑁 = 4𝜋 ∫ 𝒎 ∙ ( 𝜕𝑥 × )𝑑𝑥𝑑𝑦 (1)
𝜕𝑦

where m is the normalized magnetisation and N = ±1 corresponds to the case of magnetic

skyrmions, with the sign reflecting the polarity. Skyrmions can be competitive candidates as

information carriers in low power and highly efficient computational devices because of their non-

volatility, nanoscale size, and ease of manipulation [3]. These advantages have inspired proposals

for their implementation in skyrmionic transistors [7], logic gates [8–10], racetrack memory [11–

13], nano-oscillators [14], resonant diodes [15], neuromorphic computing [16,17], and reservoir

computing [18]. Notably, Ref. [17] is an experimental demonstration of a skyrmionic synapse

device, while the other references contain numerical results.

3
 For each of these applications, room temperature operation is a critical requirement for

realistic device integration. Recently, tailored MMLs have been explored as a means to host

skyrmions at room temperature [19,20], where the stacking of a repeated layer structure leads to

enhanced thermal stability. This stability in MMLs can be attributed to the increased DMI from

the asymmetric interfaces [Heavy metal (HM1)|Ferromagnetic (FM)] and [FM|HM2] and the

increased magnetic volume of skyrmions.

At the same time, there has been much recent effort to find skyrmion-like quasiparticles

with topological charge other than ±1. Such skyrmionic quasiparticles have been proposed as

information carriers to enhance device functionality [21,22]. For example, skyrmionium, which

comprises a skyrmion-antiskyrmion pair, is a type of magnetic quasiparticle with a vanishing

topological charge N = 0 which has been proposed for use in racetrack memory applications [21].

Similarly, anti-skyrmionite (i.e., a double-antiskyrmion-skyrmion pair with N = ∓1 [23]) has the

potential to be used as an additional information carrier in skyrmionic devices. More recently,

several studies have shown that nanomagnets can indeed host a plethora of topological and non-

topological quasiparticles, including both theoretical calculations [22] and experimental

demonstrations of skyrmion bags in liquid crystals [24] and skyrmion bundles in chiral magnets

[25]. However, these proposals are explorations of skyrmionic particles limited in two-dimensional

systems. A natural next step is to explore skyrmion-like quasiparticles which are distinguishable

by their three-dimensional profile, including variations in the film-perpendicular direction. In fact,

partially stabilised skyrmion quasiparticles have recently been theoretically predicted in bulk chiral

magnets, including chiral bobbers [26] and dipole strings [27], enriching the diversity of the

skyrmionic family.

4
 Previous studies in MML systems have mainly focused on skyrmion ‘tube’ states, in which

skyrmions are stabilised throughout the MML stack. In contrast, here we propose to encode

information signals via fractional skyrmion tubes (FSTs) in MMLs, allowing distinct vertical

skyrmion states to be explored. We will show that such fractional skyrmion tubes can exhibit

effective tunability in their thermal stability. Based on their topological properties and propagation

by electric currents, we propose a multilayer multiplexing device which relies on the nucleation,

propagation, and automatic selection of multiple distinct FSTs in a single device. Our proposal in

this work highlights the potential of encoding information via distinct fractional skyrmion states

in MMLs. Specifically, we use simulations to demonstrate pipelined information transmission and

automatic demultiplexing of information signals via use of voltage-controlled synchronizers and

track selectors, respectively. Most results of this work are performed via the micromagnetic

simulation package Mumax3 [28]. Mathematical calculations are performed via the Python

mathematical module SciPy [29].

RESULTS

1. Fractional skyrmion tubes in MMLs as multi-bit information carriers

1.1. Potential multi-bit information carriers

As aforementioned, tailored MMLs have been proposed to stabilize room temperature

skyrmions, typically as skyrmion tubes which extend vertically throughout the depth of the MML

[30,31]. In contrast, here we investigate the possibility of obtaining fractions of full skyrmion

tubes, wherein skyrmions are stabilised in select layers of the system. Fig. 1(a) shows an example

of a 4MML nanotrack hosting four distinct skyrmion states, each comprising a stack of skyrmions

which extend over one, two, three, or all four layers of the system. Such skyrmion states are

5
hereafter referred to as fractional skyrmion tubes (FST). Thus, Fig. 1(a) exhibits four FSTs, namely

(from left to right) a 4MML FST, a 1MML FST, a 3MML FST, and a 2MML FST. We will show

that each of these can be nucleated and manipulated individually, leading to a fourfold

enhancement of the capacity of the 4MML device to store and process information.

To demonstrate the effect of the micromagnetic parameters on the stability of the four

distinct FST states in our example 4MML system, we conducted micromagnetic simulations which

explore the phase diagram of FSTs by modifying i) the external out-of-plane magnetic field, ii) the

DMI constant, and iii) the ferromagnetic interlayer exchange coupling. The phase diagrams of

distinct FST states in a 4MML film with JInterlayer = 0 and JInterlayer = 0.04 mJ m-2 is shown in Fig.

1(b). The interlayer exchange coupling can be effectively modified by tuning the thickness or

changing the spacer layer material [32–34], where either ferromagnetic or antiferromagnetic

interlayer coupling can be achieved.

In the simulations, the external magnetic field was scanned in the range of 10 to 100 mT

with a 10 mT step applied out-of-plane, and the DMI constants were varied from 1.5 to 2.6 mJ m-
2
with a step of 0.1 mJ m-2, while keeping the interlayer exchange coupling constant at 0 and 0.04

mJ m-2. For each data point in Fig. 1(b), we configured the system with an initial ansatz that

contained a 1MML FST, a 2MML FST, a 3MML FST, and a 4MML FST, respectively. We then

allowed the system to equilibrate with the corresponding initial states and magnetic parameters.

Each cell of Fig. 1(b) shows the number of distinct FST states which are stabilised in the

equilibrated system. For instance, a “0” signifies that we obtain no FST states (i.e. we have the

ferromagnetic (FM) state), while a “4” denotes that all of the 1MML FST, 2MML FST, 3MML

FST, and 4MML FST states can be stabilised and distinctly identified. Note that we only record

6
the total number of distinct states stabilised in each case; we do not distinguish between different

combinations of FSTs.

Figure 1. Fractional skyrmion tubes in MMLs. (a) Schematic illustration of the 4MML, 1MML,

3MML, and 2MML FSTs, respectively, stabilised in a 4MML system. (b) Phase diagram of the

number of distinct FST states with varied external out-of-plane magnetic field and DMI constant

in a 4MML film without and with the interlayer coupling. The integers 0-4 represent the number

of distinct states. (c) Energy change of the 2MML, 3MML, 4MML, and 6MML FSTs as a function

of the x displacement of the top layer skyrmion with the interlayer exchange constant JInterlayer =

0.02 mJ m-2. (d) Energy change of the 4MML FST as a function of the x displacement of the top

layer skyrmion with the interlayer exchange coupling constant ranging from 0 to 0.20 mJ m-2.

Magnetic parameters in simulations of (c) and (d): external magnetic field 90 mT and DMI

constant 1.9 mJ m-2 which are selected from (b) in order to obtain stable FSTs states in the system.

7
 According to the results shown in Fig. 1(b), there is a noticeable transition from the no FST

state at a low out-of-plane magnetic field and high DMI constant to four distinct FST states at a

high field and low DMI constant. There is a broad parameter window for stabilizing four distinct

FST states in the 4MML systems without interlayer exchange coupling. However, with increasing

the interlayer exchange coupling, the parameter window for four FST states shrinks, while the

parameter windows for three FST states and one/two FST states expand towards the initial 4 FST

phase and 0 FST phase, respectively (see also the phase diagram at higher interlayer exchange

coupling constant JInterlayer = 0.08 to 0.12 mJ m-2 in Fig. S1 of the Supporting Information).

Therefore, by modifying the magnetic parameters of the material system, we can effectively tune

the distinct states phase diagram to achieve the maximum number of distinct FST states.

Apart from the results exhibited in Fig. 1(b), we also performed simulations on an 8MML

film to demonstrate the extension of the FSTs in even higher-level cascaded MMLs. As shown in

Fig. S2 of Supporting Information, we can obtain eight distinct FST states with a large parameter

window in the 8MML film. A similar transition to that seen in Fig. 1(b) is also observed. Similarly,

by tunning the interlayer ferromagnetic exchange coupling JInterlayer from 0 to 0.08 mJ m-2, we

observe an apparent shrinking of region corresponding to the larger number of distinct FST states.

In summary, the data in Fig. 1(b) shows that to maximize the number of distinct FST states, a

relatively high external magnetic field (70 to 100 mT), a moderate DMI constant (1.8 to 2.2 mJ m-
2
), and a relatively weak interlayer ferromagnetic exchange coupling (lower than 0.08 mJ m-2) are

required.

1.2. Thermal stability of fractional skyrmion tubes

8
 To reveal the reliability and feasibility of using multiple FSTs in a single MML device, it

is important to assess their thermal stability, which is related to their binding energy. We thus

calculated the binding energy of different FSTs as a function of the in-plane separation between

two nearest-neighbour skyrmions in the two topmost layers of the FST stack, respectively. The

binding energy of FSTs was calculated according to Ref. [35]; the results are shown in Fig. 1(c).

First, we relaxed the skyrmions in a 2MML FST, 3MML FST, 4MML FST, and 6MML FST,

respectively. Then we artificially shifted the top-layer skyrmion from -50 nm to 50 nm in the x-

direction while fixing the position of skyrmions in the other layers. The total micromagnetic energy

of the whole system at every x-shifted position is calculated without relaxing and equilibrating the

system. As shown in Fig. 1(c), we normalised the results by subtracting the energy of initial FST

states from the energy after shifting the top layer skyrmion. Therefore, the binding energy as a

function of x-shifted position in Fig. 1(c) is greater or equal to 0. The relative value represents the

increased total energy induced from shifting the top layer skyrmion.

The binding energy in Fig. 1(c) shows that by shifting the top layer skyrmion away from

the centre, the total energy first rises gradually and then remain approximately constant after an x-

displacement of ~35 nm. The difference between the highest energy state and the initial state is the

binding energy of the FST, which quantifies the energy barrier separating the FST and decoupled

skyrmion states. The interlayer exchange coupling JInterlayer = 0.02 mJ m-2, external magnetic field

of 90 mT, and DMI constant of 1.9 mJ m-2 were used when simulating FSTs in Fig. 1(c). The

binding energy Eb of the 2MML FST, 3MML FST, 4MML FST and 6MML FST are calculated as

1.26 eV, 2.46 eV, 4.60 eV, 12.81 eV, respectively. The increased binding energy from 2MML FST

to 6MML FST can be attributed to enhanced dipolar coupling. The micromagnetic total energy of

a 1MML FST is calculated as 0.6 eV. Therefore under these parameters, the FSTs have sufficient

9
binding energy Eb to prevent skyrmions from decoupling, demonstrating the thermal stability of

FSTs proposed in this work.

To demonstrate the tunability of the FST thermal stability, we next modified the interlayer

exchange coupling constant JInterlayer from 0 to 0.08 mJ m-2 for each FST. The binding energy of

FSTs is calculated and displayed in Fig. S3 of the Supporting Information. It is observed that the

binding energy of each FST follows a linear relationship with the interlayer exchange coupling of

the MMLs. To explore this phenomenon, we calculated the binding energy of a 4MML FST with

JInterlayer ranging from 0 to 0.2 mJ m-2. The results, shown in Fig. 1(d), verify that the binding energy

increases with increasing interlayer exchange coupling. Similar results can also be observed for

2MML, 3MML, and 6MML FSTs, as shown in Fig. S4 of the Supporting Information.

The larger calculated binding energy results in an improved thermal stability of FSTs in

the MML system. The thermal stability can be quantified using the Arrhenius-Néel law to estimate

the lifetime of a metastable state [36,37],

𝐸
𝜏(𝐸b ) = 𝜏0 exp (𝑘 b𝑇), (2)
B

where 𝑓 = 𝜏0−1 is attempt frequency, kB is the Boltzmann constant, and T is the temperature under

consideration. Here, we assume T = 300 K in order to estimate the FST lifetime at room

temperature. An attempt frequency of 109 - 1012 Hz is typically used [36–38]. However, there is

some debate about the correct value of the attempt frequency to use [39], which can be as large as

1021 Hz. Precisely estimating the lifetime of quasiparticles is beyond the scope of this paper. We

therefore choose a large value of 1021 Hz for the attempt frequency here and give a conservative

estimation of the lifetime of FSTs. The binding energy of FSTs extracted from Fig. 1(c) with

JInterlayer = 0.02 mJ m-2 lead to estimated lifetimes of 20.1 ns, 1.47 s, 2.12×1020 s, 1.89×1056 s, and

10
1.58×10194 s for the 1MML FST, 2MML FST, 3MML FST, 4MML FST, and 6MML FST,

respectively. Note that the estimated results merely reflect the thermal stability and annihilation

probability, rather than the precise lifetime of FSTs in realistic devices. As we are using multiple

FSTs in a single device, the least stable one defines the thermal stability of the whole device.

However, we can enhance the lifetime of each FST by 5 orders of magnitude by tuning the

magnetic parameters and interlayer exchange coupling. Our results here demonstrate the tunability

of the thermal stability of FSTs, a key consideration for experimental and commercial device

design.

2. Magnetic and topological properties of fractional skyrmion tubes

To utilize FSTs in realistic devices and applications, it is necessary to evaluate their

magnetic and topological properties, especially the propagation behaviour under the applied

electric current. Fig. 1 demonstrates that the various MML FSTs have different dimensions in the

x-y plane. We thus stabilised a series of FSTs in an 8MML film to demonstrate the MML-

dependent diameter of skyrmions.

In the simulations, we artificially equilibrated skyrmions in the select MMLs in order to

obtain a 1MML FST, 2MML FST, 3MML FST, 4MML FST, 5MML FST, 6MML FST, 7MML

FST, and 8MML FST. The layers are marked as L1 to L8 for the 8MML film in Fig. 2(a), where

the bottom layer is L1, and the top layer is L8, such that the stabilised 1MML FST contains only

one skyrmion in L1 while L2 to L8 remain in the saturated FM state. Similarly, the 4MML FST

has skyrmions within L1 to L4, and 8MML FST has skyrmions within all layers as a full tube. The

relative size and skyrmion diameter of these FSTs can be visualized by superposing 1MML to

8MML FSTs together, as shown in Fig. 2(a). The skyrmion diameter of FSTs grows monotonically

11
with the number of MMLs. This trend can be attributed to the increased dipolar field as the number

of cascaded skyrmions increases [16]. The measured skyrmion diameter of FSTs is shown by the

black lines in Figs. 2(c) and 2(d), where an approximately linear relation with the number of MMLs

can be observed.

We then investigated the transport behaviour of different MML FSTs under electric

current. The spin-orbit torque (SOT) induced by a current perpendicular to the plane (CPP)

geometry and the spin-transfer torque (STT) from a current in plane (CIP) geometry are

considered. For the CIP case, the charge current is applied in the FM layer in the x-direction with

a density of 15 MA cm-2 within each layer of the MML. In the case of CPP, a charge current with

density 50 MA cm-2 is applied in the bottom HM layer only; the thickness of the HM layers in

each MML is smaller than the electron diffusion length [40] so that no spin Hall effect will be

induced [41]. A spin current with spin polarisation in +y direction is created and injected into the

first MML starting from the bottom, namely L1 mentioned above. As a result, FSTs move along a

trajectory at an angle to the direction of the applied current, which is the well-known skyrmion

Hall effect (SkHE) [42]. The angle between the FST trajectory and the direction of applied current

(+x in this example) is defined as the skyrmion Hall angle θSkHE. Under the SOT, FSTs propagate

along the +y transverse direction, while under the STT, FSTs move towards the opposite transverse

direction. As shown in Fig. 2(b), the 1MML FST has the largest θSkHE, while the 4MML FST

exhibits a smallest θSkHE for both CPP and CIP.

To better understand the simulated spin dynamics, it is instructive to employ an analytical

model for the motion of non-collinear spin textures. Therefore, we utilize the Thiele equation by

imposing the stationary limit that the FSTs move with a constant velocity and that the texture does

12
not deform. We consider both CIP and CPP geometries using the Zhang-Li [43] and Slonczewski

[44] torques in the LLG equation (see Methods), respectively.

Figure 2. Magnetic and topological properties of fractional skyrmion tubes. (a) Cross-sectional

view of the 1MML to 8MML FSTs stabilised in the same 8MML system. White outlines of the

FSTs indicate the varying skyrmion diameter from 1MML to 8MML. (b) Simulated trajectories of

FSTs driven by SOT with the current density of 15 MA cm-2 and a spin Hall ratio of 0.6, and STT

with the current density of 50 MA cm-2. (c) and (d) show simulation results and theoretical

predictions from the Thiele equation of the skyrmion diameter, the velocity of movement, and

skyrmion Hall angle for 1MML to 8MML FSTs under (c) SOT with current perpendicular to the

plane and (d) STT with the current in the plane. Note that 7MML and 8MML FSTs transformed

to the stripe domain wall states after applying the electric current. The magnetic parameters of

13
simulations are interlayer exchange coupling constant JInterlayer = 0.02 mJ m-2, external magnetic

field 90 mT, and DMI constant 1.9 mJ m-2.

In the CPP geometry, assuming a periodical boundary condition in the x-y plane, the

translational motion of spin textures driven by the spin Hall effect can be described by a modified

Thiele’s equation [45,46]

𝑮 × 𝒗 − 𝛼𝓓 ∙ 𝒗 − 𝒯𝑆𝑂𝑇 𝓘 ∙ 𝐦p = 0, (3)

where 𝑮 = (0, 0, −4𝜋𝑁) is the Gyroscopic vector with the topological charge 𝑁 defined in Eq.

1, and 𝒗 = (𝑣x , 𝑣y ) is the skyrmion drift velocity along the 𝑥 and 𝑦 axes, respectively. The first

term 𝑮 × 𝒗 in Eq. 3 is the topological Magnus force that results in the transverse motion of

skyrmions as a function of the driving current, which directly results in the SkHE [42]. 𝛼 is the

magnetic damping parameter, and 𝓓 is the dissipative tensor which is calculated by 𝒟ij =

1 𝜕𝑴 𝜕𝑴 𝒟𝑥𝑥 𝒟xy
∬ 𝜕𝑥 ∙ 𝜕𝑥 𝑑𝑥𝑑𝑦 = [ 𝒟 ]. The term 𝒯𝑆𝑂𝑇 𝓘 ∙ 𝐦p quantifies the effect of the SOT over the
𝑀𝑠2 i j yx 𝒟yy

𝛾𝑒 ℏ𝑗𝑒 𝜃𝑆𝐻
magnetic quasiparticle, where 𝒯𝑆𝑂𝑇 = is the amplitude of SOT over the quasiparticle, 𝛾𝑒 =
2𝑒𝑀𝑠 𝑡

𝛾
= 1.76 × 1011 𝑇 −1 𝑠 −1 is the gyromagnetic ratio of an electron, ℏ is the reduced Planck
𝜇0

constant, 𝑗𝑒 is the current density, 𝜃𝑆𝐻 is the spin Hall ratio, 𝑒 is the electron charge, 𝑀𝑠 is the

saturation magnetisation, and 𝑡 is the thickness of the FM layer. 𝓘 is the driving torque tensor

1 𝜕𝑴 ℐ𝑥𝑥 ℐxy
which is calculated by ℐij = 𝑀2 ∬ (𝜕𝑥 × 𝑴) 𝑑𝑥𝑑𝑦 = [ ], and 𝐦p is the polarisation
𝑠 i j ℐyx ℐyy

direction of the spin current. By solving Eq. 3, we can obtain the skyrmion Hall angle of the FSTs

as

14
 𝑣y 4𝜋𝑁
𝜃SkHE(CPP) = arctan ( 𝑣 ) = − 𝛼𝒟 . (4)
x xx

In CIP geometry, the Thiele equation has the form [11,46]

𝑮 × (𝒋 − 𝒗) + 𝓓(𝛽𝒋 − 𝛼𝒗) = 0, (5)

where 𝑮, 𝓓, 𝛼, and 𝒗 are as defined above. 𝛽 is the non-adiabaticity of spin-transfer-torque, for

which a small value of 0.02 is used, and 𝒋 is the current vector describing the direction and

amplitude of the applied current. From Eq. 5, the skyrmion Hall angle of FSTs under CIP is

therefore

𝑣y 4𝜋(𝛼−𝛽)𝒟xx 𝑁
𝜃SkHE(CIP) = arctan ( 𝑣 ) = 𝛼𝛽𝒟2 2𝑁2
. (6)
x 𝑥𝑥 +16𝜋

We calculated 𝜃SkHE for both CPP and CIP by Eq. 5 and Eq. 6, respectively. Comparison

of the theoretical prediction and simulation results of 𝜃SkHE for 1MML to 8MML FSTs are

exhibited in Figs. 2(c) and 2(d), where Fig. 2(c) is for CPP geometry and Fig. 2(d) is for CIP. The

simulated results of 𝜃SkHE are shown by the discrete data points, and the calculated results from

Thiele’s equation are exhibited as solid lines. There is a decrease of 𝜃SkHE both in simulations and

theoretical calculations when cascading MMLs for FSTs in both CPP and CIP geometry. However,

there is a slight disparity between the theoretical predictions and simulations for 𝜃SkHE , which

becomes more significant as the number of MML repeats increases. The overestimation of

skyrmion Hall angle by the Thiele equation can be explained by additional dissipation mechanisms

related to dynamic variation of the skyrmion shape, which cannot be captured by the rigid shape

approximation used to derive the Thiele equation.

We also extracted the velocity of FST propagation in CPP and CIP from simulations. The

results are shown in Figs. 2(c) and 2(d). FSTs generally moves faster in CPP than in CIP, even

15
with a smaller amplitude of the applied electric current. On the other hand, a marked drop of FST

velocity from 7.5 ms-1 for 1MML FST to 3 ms-1 for 6MML FST under the same driving current

can be observed in the CPP geometry, while the velocity of FSTs remains virtually constant as the

number of MML repeats increases for the case of CIP geometry. The different current-dependent

characteristics of skyrmion velocity could result from the way that torque is injected into the

magnetic textures via the CPP and CIP geometry. In the CIP geometry, the torques act on

individual skyrmions in each layer, while in the CPP geometry, the torque is merely injected from

the top layer skyrmion so that velocity would decrease as the number of MML repeats increases.

Overall, these results indicate that the proposed FSTs have tuneable thermal stability and distinct

topological and magnetic properties, which may be used as signatures to identify them from each

other. In the next chapter, we demonstrate one of the potential uses for FSTs: an MML

multiplexing device.

3. Magnetic multilayer multiplexing device with fractional skyrmion tubes

3.1. Demonstration of the proposed multiplexing device

In this section, a multilayer multiplexing device is proposed using multiple FSTs as

information carriers. Such a device can perform signal multiplexing, signal transmission, and

automatic signal demultiplexing. We have shown that as many as eight distinct FST states,

possibly more, can be achieved through careful selection of the system and magnetic parameters.

In this section, an FST-based device built from a 4MML nanotrack is illustrated for simplicity.

The schematic of the proposed 4MML multiplexing device is illustrated in Fig. 3. The multiplexing

device has three parts: a 4MML nanotrack for FST transmission, terrace-like MML stages for FST

16
nucleation via magnetic tunnel junctions (MTJs) fabricated on the top, and a four-branched track

for FST automatic selecting.

The proposed device shown in Fig. 3 can multiplex and transmit four distinct sequences of

information signals simultaneously, where each information signal is encoded by one type of FST,

and the presence/absence of the FST encodes information “1”/“0”. As shown in Fig. 3, each type

of FST acts as a distinct information carrier. The workflow of the proposed multilayer multiplexing

device contains four procedures: 1) FSTs are nucleated at the terrace-like MML stages via STT by

injecting electric current from MTJs; 2) Multiple FSTs are chambered into the 4MML nanotrack

in preparation for the transmission of signals; 3) Information transmission via FSTs propagation

along the 4MML nanotrack; 4) Automatic demultiplexing of the information signals (FSTs) via

the SkHE and the four-branched track selector. These four procedures are explored in detail in the

following sections.

Figure 3. Schematic of the proposed use of FSTs in a multilayer nanotrack as a multiplexing

device. The workflow of the proposed device is: nucleation of FSTs at the terrace-like MML

17
stages, chambering the 4MML nanotrack with multiple FSTs, FSTs propagation along the 4MML

nanotrack, and automatic demultiplexing of FSTs via the four-branched track selectors. A group

of FSTs (i.e. 1MML FST, 2MML FST, 3MML FST, and 4MML FST) serve as information carries.

3.2. Nucleation of the fractional skyrmion tubes

The FSTs can be nucleated in the terrace-like MML stage structure shown in Fig. 4(a) in

the proposed multilayer multiplexing device. There are individual MTJs on the surface of each

step of the four-level stage, each of which is used to inject electric current providing STT for

nucleating skyrmions in the MMLs beneath it. The main staircase structure and MTJs shown in

Fig. 3 and Fig. 4(a) can be fabricated by additive or subtractive lithography processes. Single-layer

precision may be achieved either through careful calibration of the material growth/removal rate

or by dose-modulated electron-beam lithography [47]. In order to contact the top electrode of the

MTJs for tunnelling in the perpendicular direction, each of them must be sheathed by an insulating

material such as SiO2.

18
Figure 4. Nucleation of the fractional skyrmion tubes at a multilayer terrace-like stage in the

proposed multiplexing device at room temperature. (a) Schematic drawing of the terrace-like

nucleation site. An MTJ is placed on the surface of each terrace as an electric writing head. With

this design, four FSTs can be nucleated individually. (b) The nucleation process of 1MML, 2MML,

3MML, and 4MML FSTs by a 1.5 ns width electric current pulse (full width at half maximum).

The screenshots of each FM layer within FSTs at 0 ns, 0.5 ns, 1.0 ns, 1.5 ns, and 2.0 ns are

presented in a timeline. (c) Probability phase diagram of successfully nucleating four FSTs at room

temperature, with varied pulse width (0.1 ns to 12.8 ns) and the amplitude of current density (100

MA cm-2 to 800 MA cm-2). Each data point illustrates the probability of successful nucleation of

19
FSTs out of 20 distinct attempts. (d) Evolution of topological charge density for 1MML, 2MML,

3MML, and 4MML FSTs during the nucleation process in (b).

The simulated system has a four-level terrace geometry with cell size 2 nm x 2 nm x 1 nm.

From left to right, the geometry of each terrace is 600 nm x 600 nm x 2nm, 600 nm x 600 nm x

4nm, 600 nm x 600 nm x 6nm, and 600 nm x 600 nm x 8nm. The MML has a small interlayer

exchange coupling of JInterlayer = 0.02 mJ m-2. The simulations were performed at room temperature

by including an extra thermal fluctuation field [28] and with room temperature magnetic

parameters. The magnetic parameters were retrieved from the literature of experimental

measurements (see Methods). Electric current is injected into the MMLs from the MTJ using a

CPP geometry, and the FSTs are nucleated through the STT. The pulse width of the injected current

is 1.5 ns full-width at half maximum (FWHM) with a peak current density of 250 MA cm -2.

Although the pulse length is very short, the requisite current density is very large and may damage

to the ultrathin tunnel barrier of the MTJ through Joule heating [48]. Therefore, future work is

required to minimise the effects of Joule heating, either through optimisation of the spin injection

process or by including a heatsink.

Fig. 4(b) presents each FM layer of the 1MML FST, 2MML FST, 3MML FST, and 4MML

FST during the nucleation process. Each simulation is initialised in the uniform FM state at the

beginning. After receiving the STT from the electric current pulse, the magnetisations in each case

first experience a fluctuation period of around 0.5 ns, before individual skyrmions form in each

MML. After a time frame of 1.5 ns, we can see from Fig. 4(b) that skyrmions are nucleated and

stabilised in all four FSTs. To better understand the nucleation process of FSTs, the topological

charge density for the four regions of each FST from left to right in Fig. 4(a) is extracted and

20
displayed in Fig. 4(d). The topological charge density remains unchanged until 0.5 ns. From 0.5

ns to 1.5 ns, a few fluctuations in the topological charge density for FSTs can be seen before they

reach the final skyrmion state. The final values of the topological charge density are approximately

40×1012 m-2, 80×1012 m-2, 120×1012 m-2, and 160×1012 m-2 for 1MML FST, 2MML FST, 3MML

FST, and 4MML FST, respectively. The calculated topological charge density displayed in Fig.

4(d) verifies the results in Fig. 4(b).

Since a stochastic thermal field was included in the simulations of FST nucleation, the

successful nucleation of FSTs occurs probabilistically, rather than being deterministic. We

therefore determined the probability phase diagram of the FSTs by scanning the electric pulse

width from 0.1 ns to 12.8 ns in a geometric sequence (i.e. 0.1, 0.2, 0.4, 0.8, 1.6, 3.2, 6.4, 12.8 ns)

and the current density from 100 MA cm-2 to 800 MA cm-2 in 50 MA cm-2 increments. We

simulated the nucleation process for each FST with each given pulse width and current density 20

times, then calculated the probability of successful nucleation. The probability phase diagrams of

successful nucleation of 1MML to 4MML FSTs are shown in Fig. S5 in Supporting Information.

There are 15×8=120 data points displayed in the figures, where each data point illustrates the

probability of successful nucleation out of 20 distinct attempts. The colour coding describes the

probability value, where white represents 0% of the probability and black represents 100%. A

transition from low nucleation probability for 1MML FST to high nucleation probability for

4MML FST can be seen in Fig. S5. However, for the multiplexing device proposed in this work,

four different FSTs need to be nucleated at the four-stage terraces individually at the same time.

Therefore, we also calculated the probability for successfully nucleating four FSTs simultaneously,

with results summarised in the phase diagram displayed in Fig. 4(c). The results suggest that FSTs

are more likely to be nucleated when we apply a large amplitude of current density (200 MA cm -

21
2
to 300 MA cm-2) with a relatively large pulse duration (greater than 1 ns). This information is

critical for experimental realisation of the multiplexer device.

We neglected the spin memory loss (SML) [43] when injecting STT to magnetisations in

cascaded MMLs to simplify the nucleation procedure. Careful consideration of SML in the device-

level simulations may be the focus of follow-up work. The results in this section demonstrate the

nucleation of FSTs in the proposed multiplexing device, including the structure of the four-stage

terrace nucleation site, the detailed nucleation process of FSTs, and the probability phase diagram

of successful nucleation. In the following section, the second important stage of the device

workflow will be addressed, namely the FST initialisation process.

3.3. Chambering the multiplexing device with FSTs

After nucleating FSTs in the four-stage terraces, the proposed device needs to be initialised

before all of the FSTs propagate in the main 4MML nanotrack. The FSTs must propagate upwards

from the nucleation sites into the main nanotrack during the initialisation process. We refer to the

initialisation process as “chambering” in this work. FSTs must overcome an energy barrier when

moving into the 4MML nanotrack from a film with a different number of MMLs [35], which can

be attributed to the dipolar field in the two systems. As shown in Fig. 5(a), the 4MML multiplexing

device needs to be chambered with all four FSTs from the four-stage terraces. We used CIP

geometry for the chambering process because STT can act on every skyrmion throughout the FSTs,

which decreases the risk of FSTs decoupling.

Fig. 5(a) illustrates the chambering process enabled by an STT resulting from the CIP in

the y-direction with the current density of 100 MA cm-2 and pulse width of 4 ns. FSTs first

propagate towards the boundary of different MMLs systems in the y-direction, i.e. along the

22
direction of the applied current, before moving along the boundary in the x-direction for a short

distance. They subsequently cross the boundary and successfully move into the main track. The

1MML FST exhibits the largest displacement along the x direction; this displacement gradually

reduces as the number of MML repeats increases. We then varied the interlayer exchange coupling

constant JInterlayer and the amplitude of the current density to generate a phase diagram of the

initialisation process. By changing JInterlayer and current density, the initial state leads to four phases

which we label “chamber”, “stuck”, “nucleate” and “decouple”, as shown in the left panel of Fig.

5(b). If the interface traps the FST, we mark it as a “stuck”; if all four FSTs propagate into the

main track, we mark it as a “chamber”; if any of FSTs annihilate, we mark it as a “decouple”; if

the skyrmions in different layers separates we also mark this as a “decouple”; if a new skyrmion

is nucleated within the FM layers during the process, we mark it as a “nucleate”.

23
Figure 5. Chambering the multiplexing device with FSTs (the initialisation process). (a)

Micromagnetic simulation results of chambering the 4MML nanotrack with a 1MML FST, a

2MML FST, a 3MML FST, and a 4MML FST. The trajectories of each FST are marked as C1,

C2, and C3, respectively. The magnetic parameters used were: interlayer exchange coupling

constant JInterlayer = 0.02 mJ m-2, external magnetic field 90 mT, and DMI constant 1.9 mJ m-2. (b)

Phase diagram of chambering four FSTs into the main track. Energy paths for the chambering

process C1 and C3 are shown in (c) and (d), respectively. The energy path for process C2 is shown

in Fig. S6 in Supporting Information.

The results of the initialisation process after injecting electric current are summarised in

the phase diagram shown in Fig. 5(b). The four possible phases “chamber”, “stuck”, “nucleate”,

and “decouple” are marked with a blue rectangle, a red triangular, a green circle, and a yellow star

respectively. As the target phase, the “chamber” phase has a decently large parameter window,

which occupies the upper left 1/4 of the phase diagram. In the multilayers with large interlayer

exchange coupling constants (JInterlayer ≥ 0.06 mJ m-2), 1MML FSTs are stuck by the boundary

when applying insufficient current densities (Jdc < 50 MA cm-2) and will annihilate when the

current density is large (Jdc ≥ 50 MA cm-2). 3MML FSTs will nucleate a new skyrmion in the top

FM layer when crossing the boundary with high interlayer exchange coupling constants (JInterlayer

> 0.03 mJ m-2) and applied current densities (Jdc > 37.5 MA cm-2). 2MML FSTs behave more

robustly and produces the “chamber” phase in most situations except for at extremely high

interlayer exchange coupling constants (JInterlayer > 0.08 mJ m-2) and current densities (Jdc > 125

MA cm-2), wherein annihilation is observed. 4MML FSTs always produce the “chamber” phase,

resulting from the fact that there is no energy barrier during the propagation process in this case.

24
The superposed markings in the phase diagram indicate different behaviours from different FST

types. For instance, all cases of the yellow stars superposing green circles correspond to the

situation that the 1MML FST annihilates and the 3MML FST nucleates a new skyrmion; the red

triangle with the green circle indicates that the 1MML FST is stuck and the 3MML FST nucleates

a new skyrmion.

In order to better explain the phases shown in Fig. 5(b), we extracted the total

micromagnetic energy of the regions containing 1MML FST, 2MML FST, 3MML FST, and

4MML FST individually. The trajectories when chambering 1MML FST, 2MML FST, and 3MML

FST are marked as C1, C2, and C3, respectively, in Fig. 5(a). The energy evolution of C1, C2, and

C3 can be seen from Fig. 5(c), Fig. S6 of Supporting Information, and Fig. 5(d). Here we varied

the interlayer exchange coupling constant from 0 to 0.08 mJ m-2 while fixing the current density

Jdc = 125 MA cm-2. With a higher interlayer exchange coupling constant, the “decouple” phase is

observed in C1, while the “nucleate” phase could be seen in C3. The energy evolution is further

demonstrated by the change of topological charge of the system shown in the insets of Figs. 5(c)

and 5(d). The energy barrier for chambering 1MML FST into the central track increases as a linear

relationship with the amplitude of JInterlayer. When chambering 1MML FST, the “chamber” phase

is obtained when JInterlayer ≤ 0.06 mJ m-2; therefore, no net change of topological charge is

observed. Annihilation of the skyrmion happens at JInterlayer = 0.08 mJ m-2, so the topological charge

of the system changes to zero. Similarly, when chambering the 3MML FST, a net increase of the

topological charge by around 1/3 occurs, suggesting the nucleation of a new skyrmion in the

system at higher interlayer exchange coupling constants (JInterlayer ≥ 0.04 mJ m-2). Note that the

initialisation process of 2MML FST, marked as C2 in Fig. 5(a), merely exhibits “chamber” and

“stuck” phases in Fig. 5(b), and therefore no noticeable variation can be observed in the topological

25
charge. The energy evolution when chambering the 2MML FST is presented in Fig. S6 of the

Supporting Information.

3.4. Voltage-controlled synchronizers for pipelining

After the initialisation process, we now have the multiplexing device chambered with FSTs

in the main nanotrack. FSTs can then be transmitted along the track towards the demultiplexing

region. Considering the long transmission distances likely required in real-world applications, it is

better to divide the transmission track into several regions, as shown in Fig. 6(a). In such a design,

we can achieve pipelined transmission for information carriers to enhance device throughput. The

results in Fig. 2 have demonstrated that FSTs with different MMLs propagate with various

velocities, where the 4MML FST moves the most slowly, and the 1MML FST the most quickly.

Therefore, synchronizers are required to maintain the correct order of information sequences and

avoid sticky packets during data transmission [23]. Here we can utilise voltage gates as

synchronizers based on the voltage-controlled-magnetic-anisotropy (VCMA) effect. The VCMA

effect was first reported in a 3D transition ferromagnetic layer in 2–4 nm thick FePt (Pd) films [49].

Surprisingly, it has been reported that a small electric field of 100 mV nm-1 is sufficient to change

the PMA by 40%, corresponding to a VCMA efficiency of 210 fJ V-1 m-1 at room temperature [50].

In this work, the simulation of the VCMA effect is based on a linear relationship [12]:

𝐾𝑢𝑣 = 𝐾𝑢0 + 𝜗𝑉𝑏 , (7)

where 𝜗 is the VCMA coefficient, 𝑉𝑏 is the bias voltage on the VCMA gate, and 𝐾𝑢0 is the

background anisotropy constant. Regions in which the PMA is elevated provide an increased

energy barrier, whereas regions of reduced PMA lead to a potential well. When simulating the

pipelined transmission of FSTs schematised in Fig. 6(a), we periodically set VCMA gates on and

26
off by changing the amplitude of the PMA constant by 10%. An electric driving current with the

amplitude of 3 MA cm-2 is applied in the HM layer in the x-direction. Compared to the chambering

process of FSTs where STT is utilized, we apply SOT for the transmission of FSTs instead to

achieve higher driving efficiency and thus lower energy consumption [13].

Figure 6. Pipelined transmission and automatic demultiplexing of FSTs in the proposed device.

(a) Schematic showing pipelined transmission of four FST packets. Each information packet

consists of a combination of multiple FSTs. The red shaded areas denote the VCMA synchronizers.

(b) Schematic illustration for tuning the branch width of the track selector. SOT drives the FSTs

with the current density of 3 MA cm-2. (c) Simulation results of automatically demultiplexing FSTs

via a four-branch track selector. (d) Track selecting phase diagram for 1MML FST, 2MML FST,

3 MML FST, and 4MML FST, respectively, while varying the width of the branch and fixing the

current density at 3 MA cm-2.

The pipelined transmission of four FSTs in the 4MML main track is illustrated in Fig. 6(a).

We deployed three VCMA gates in the track, which are marked as red shaded rectangles. The track

27
is divided into four regions as registers, each of which can store a packet of information consisting

of four bits. The presence (absence) of a 1MML FST/2MML FST/3MML FST/4MML FST

encodes a single binary “1” (“0”) for the first/second/third/fourth bit respectively. Initially, four

packets of information signals “1111”, “1110”, “0011”, and “0001” are positioned in consecutive

registers of the device via propagation mediated by the VCMA gates. They are then transmitted

along the x-direction and synchronised by adjacent VCMA gates. It should be noted that the FSTs

are nucleated and propagated from left to right in order of the number of MML repeats, such that

within an information packet, the order is [1MML FST | 2MML FST | 3MML FST | 4 MML FST].

This design can retain the order of FSTs and prevent mismatch between information packets during

transmission. Fig. 6(a) indicates that the information integrity and sequence order are well retained,

allowing the transmission efficiency to be quadrupled. Furthermore, this pipelined design is not

limited to four-bit packets of information. The VCMA gates can support as many information

carries in the nanotrack as can be stabilised in the MML system. Here we use four carries in each

packet for illustration and consistency with the other results in this paper.

3.5. Automatic demultiplexing via track selectors

The final stage of the proposed MML multiplexing device workflow is to filter the FSTs

out of the information packets to decode the information signals. Here we propose to use a four-

branch track selector to filter the FSTs into individual tracks for detection. We can utilise a track

selector to filter different FSTs, as illustrated in Fig. 6(b). As has been discussed above, there are

two choices to filter the FSTs: the first one is via the skyrmion Hall angle 𝜃SkHE and the second is

the size difference of the skyrmion diameter. Changing the angle or the width of the branch can

filter and demultiplex different FSTs. As for the skyrmion Hall angle 𝜃SkHE , our results suggest

that there is a difference in 𝜃SkHE for FSTs with different MMLs. Namely, 𝜃SkHE for 1MML FST

28
to 4MML FST under SOT is -47.5°, -44.0°, -40.7°, and -35.0°, respectively. At the same time,

there is also a difference in the skyrmion diameter among FSTs, where the skyrmion diameter for

1MML FST to 4MML FST is 21.5 nm, 26.5 nm, 32.5 nm, and 38.0 nm, respectively. The angle-

based selector would require a much larger device footprint to filter the FSTs for their relatively

small angle difference. Therefore, in this work, we propose to demultiplex the information signals

by a width-based track selector.

Simulations of the demultiplexing process were performed to obtain a phase diagram for

the four-branch track selector. Here, we modified the width of the branch from 50 nm to 130 nm

while the angle of the branch was fixed at 70°, and the electric current density was 3 MA cm-2 (see

Fig. 6(b)). The four FSTs individually propagated through the selector along the track, driven by

the SOT. If the FST propagated into the branch, we marked the result as “go upwards”. Otherwise,

it was marked as “go straight”. The track selecting phase diagram for 1MML to 4MML FSTs is

shown in Fig. 6(d), where the sky-blue coloured region represents FSTs propagating into the

branch and the cream-yellow coloured region denotes FSTs moving straight, passing by the

branch. The results indicate enough space to design a track selector for filtering four FSTs out of

one packet because each FST has an exclusive window with 20 nm width. Guided by the track

selecting phase diagram for FSTs, we configured a four-branch track selector as shown in Fig.

6(c). The width of these four branches is 60 nm, 80 nm, 100 nm, and 120 nm for demultiplexing

1MML FSTs, 2MML FSTs, 3MML FSTs, and 4MML FSTs respectively. Successful

demultiplexing of the FSTs shown in Fig. 6(c) verifies the feasibility and validity of such a width-

based track selector.

In summary, these results demonstrate that the proposed fractional skyrmion tubes in the

MML system can be potential candidates as information carriers, based on the fact that multiple

29
FSTs can be nucleated, transmitted, and filtered in a single MML device. The results highlight the

potential for three-dimensional skyrmionics quasiparticles to significantly expand the information

storage and processing capability of tailored magnetic multilayer systems.

CONCLUSIONS

This study was designed to assess the hypothesis that in MML systems, skyrmions can

exist within part of a multilayer, as fractional skyrmion tubes. We confirmed this with magnetic

energy analysis and micromagnetic simulations. The findings suggest that distinct FST states may

coexist in a single MML system which are tuneable in their thermal stability and magnetic

properties. Their topological properties and current-driven behaviour were analysed both with

theoretical calculations and micromagnetic simulations. This work also proposes to use such FSTs

to encode information in an MML multiplexing device, where multiple FSTs can be nucleated,

transmitted, and filtered. This study provides the first comprehensive assessment of encoding

information by multiple FSTs in a single MML device, highlighting the potential utility of distinct

skyrmion states in magnetic multilayer systems. Further investigations are required to explore

device settings for FSTs and establish effective nucleation and detection methods.

METHODS

Micromagnetic simulations:

The micromagnetic simulations were performed using the GPU-accelerated micromagnetic

programme Mumax3 [28]. The time-dependent magnetisation dynamics are conducted by the

Landau-Lifshitz-Gilbert (LLG) equation:

𝑑𝐦 𝑑𝐦
= −|𝛾LL |𝐦 × 𝐡eff + 𝛼𝐦 × + 𝒯𝑆𝑂𝑇 𝐦 × (𝐦p × 𝐦) (8)
𝑑𝑡 𝑑𝑡

30
where 𝐦 = 𝐌/𝑀s is the reduced magnetisation, 𝑀s is the saturation magnetisation, 𝛾LL is the

gyromagnetic ratio, 𝐡eff = 𝐇eff /𝑀s is the reduced effective field, α is the damping parameter,

𝛾e ℏ𝑗e 𝜃SH 𝛾
𝒯SOT = is the SOT efficiency with 𝛾e = 𝜇 = 1.76 × 1011 T-1 s-1 being the gyromagnetic
2𝑒𝑀s 𝑡 0

ratio of an electron, ℏ is the reduced Planck constant, 𝑗e is the current density, 𝜃SH is the spin Hall

angle, 𝑒 is the electron charge, 𝑡 is the thickness of the FM layer, and 𝐦p is the polarization

direction of the spin current. The energy density 𝐸 is a function of 𝐦, which contains the exchange

energy term, the anisotropy energy term, the Zeeman energy term, the magnetostatic energy term,

and the DMI energy term. The material parameters to perform the simulations are chosen

according to previous reported room temperature experimental results [20]: damping parameter

α = 0.1, DMI constant 𝐷int = 1.5 mJ m-2 to 2.6 mJ m-2, the value for Gilbert gyromagnetic ratio

γ = −2.211 × 105 mA-1 s-1, saturation magnetization 𝑀s = 956 kA m-1, the spin Hall

polarisation ΘSH = 0.6 to enhance the spin Hall effect, the uniaxial out-of-plane magnetic

anisotropy 𝐾u = 717 kJ m-3, the polarisation of the spin current is in the +𝑦 direction, and the

exchange constant is assumed to be A = 10 pJ m-1. To ensure the accuracy of calculation, the mesh

size is set to 2 nm × 2 nm × 1 nm, which is smaller than the exchange length 𝑙EX =
2𝐴
2√𝐴/(𝜇0 𝑀s2 ) = 6.0 nm and DMI length 𝑙DMI = 𝐷 = 11.8 nm. An external magnetic field of 10
int

mT to 100 mT in the out-of-plane direction is applied for the simulations. In the simulated MMLs,

the intermediate HM1 and HM2 layers are thinner than the electron spin diffusion length. In this

case, the torques would be efficient only in the external layers. In the simulation of FSTs

propagation along the nanotrack, the SOT created via a CPP is applied only in the bottom layer,

and the injected spin polarization is uniform in the layer. The injected current is then modelled as

a fully polarized vertical spin current. In the simulation of the chambering process, electric current

31
is applied in the +y direction via the CIP geometry, where skyrmions in all FM layers within FSTs

are driven by STT.

SUPPORTING INFORMATION

Figures S1-S6 of the supporting information PDF show:

(S1) Phase diagrams for FST stability with respect to DMI and external magnetic field in a four-

layer system for different values of the interlayer exchange constant;

(S2) as in Fig. S1, but for an eight-layer system;

(S3) the binding energy of different FST states for four values of the interlayer exchange constant;

(S4) the binding energy over a range of interlayer exchange couplings for 2MML, 3MML, 4MML,

and 6MML FSTs;

(S5) a phase diagram of the nucleation probability for different FST states as a function of applied

current density and pulse width, and;

(S6) the evolution of the total micromagnetic energy during the chambering process for a 4MML

multiplexing device.

AUTHOR INFORMATION

Corresponding Author

[Link]@[Link]

32
Author Contributions

R.C. and C.M. conceived the project., and R.C., C.M., Y.L., and Y.Z. contributed to the project

design. R.C. and Y.L. performed the micromagnetic simulations and theoretical calculations. R.C.,

C.M, and W.G. prepared the manuscript. All authors discussed and commented on the manuscript.

All authors have approved the final version of the manuscript.

ACKNOWLEDGMENT

This work is supported by the Engineering and Physical Sciences Research Council (EPSRC)

under the grant ‘Skyrmionics for Neuromorphic Technologies’, EP/V028189/1. The authors would

also like to acknowledge the assistance provided by Research IT and the use of the Computational

Shared Facility at the University of Manchester. R.C. and Y.L. wish to acknowledge the

Department of Computer Science Kilburn Scholarship for the funding support. R.C. would like to

thank the University of Manchester President’s Scholarship.

REFERENCES

[1] S. Muhlbauer, B. Binz, F. Jonietz, C. Pfleiderer, A. Rosch, A. Neubauer, R. Georgii, and

P. Boni, Skyrmion Lattice in a Chiral Magnet, Science (1979) 323, 915 (2009).

[2] A. Fert, N. Reyren, and V. Cros, Magnetic Skyrmions: Advances in Physics and Potential

Applications, Nat. Rev. Mater. 2, 17031 (2017).

[3] X. Zhang, Y. Zhou, K. M. Song, T. E. Park, J. Xia, M. Ezawa, X. Liu, W. Zhao, G. Zhao,

and S. Woo, Skyrmion-Electronics: Writing, Deleting, Reading and Processing Magnetic

Skyrmions toward Spintronic Applications, J. Phys. Condens. Matter 32, 143001 (2020).

33
[4] A. Bernand-Mantel, C. B. Muratov, and T. M. Simon, Unraveling the Role of Dipolar

versus Dzyaloshinskii-Moriya Interactions in Stabilizing Compact Magnetic Skyrmions, Phys.

Rev. B 101, 045416 (2020).

[5] A. Fert and P. M. Levy, Role of Anisotropic Exchange Interactions in Determining the

Properties of Spin-Glasses, Phys. Rev. Lett. 44, 1538 (1980).

[6] C. Moutafis, S. Komineas, and J. A. C. Bland, Dynamics and Switching Processes for

Magnetic Bubbles in Nanoelements, Phys. Rev. B 79, 224429 (2009).

[7] X. Zhang, Y. Zhou, M. Ezawa, G. P. Zhao, and W. Zhao, Magnetic Skyrmion Transistor:

Skyrmion Motion in a Voltage-Gated Nanotrack, Sci. Rep. 5, 11369 (2015).

[8] M. Chauwin, X. Hu, F. Garcia-Sanchez, N. Betrabet, A. Paler, C. Moutafis, and J. S.

Friedman, Skyrmion Logic System for Large-Scale Reversible Computation, Phys. Rev. Appl. 12,

064053 (2019).

[9] X. Zhang, M. Ezawa, and Y. Zhou, Magnetic Skyrmion Logic Gates: Conversion,

Duplication and Merging of Skyrmions, Sci. Rep. 5, 9400 (2015).

[10] B. W. Walker, C. Cui, F. Garcia-Sanchez, J. A. C. Incorvia, X. Hu, and J. S. Friedman,

Skyrmion Logic Clocked via Voltage-Controlled Magnetic Anisotropy, Appl. Phys. Lett. 118,

192404 (2021).

[11] X. Zhang, G. P. Zhao, H. Fangohr, J. P. Liu, W. X. Xia, J. Xia, and F. J. Morvan,

Skyrmion-Skyrmion and Skyrmion-Edge Repulsions in Skyrmion-Based Racetrack Memory, Sci.

Rep. 5, 7643 (2015).

34
[12] W. Kang, Y. Huang, C. Zheng, W. Lv, N. Lei, Y. Zhang, X. Zhang, Y. Zhou, and W.

Zhao, Voltage Controlled Magnetic Skyrmion Motion for Racetrack Memory, Sci. Rep. 6, 23164

(2016).

[13] J. Sampaio, V. Cros, S. Rohart, A. Thiaville, and A. Fert, Nucleation, Stability and

Current-Induced Motion of Isolated Magnetic Skyrmions in Nanostructures, Nat. Nanotechnol.

8, 839 (2013).

[14] L. Shen, J. Xia, G. Zhao, X. Zhang, M. Ezawa, O. A. Tretiakov, X. Liu, and Y. Zhou,

Spin Torque Nano-Oscillators Based on Antiferromagnetic Skyrmions, Appl. Phys. Lett. 114, 1

(2019).

[15] L. Zhao, X. Liang, J. Xia, G. Zhao, and Y. Zhou, A Ferromagnetic Skyrmion-Based

Diode with a Voltage-Controlled Potential Barrier, Nanoscale 12, 9507 (2020).

[16] R. Chen, C. Li, Y. Li, J. J. Miles, G. Indiveri, S. Furber, V. F. Pavlidis, and C. Moutafis,

Nanoscale Room-Temperature Multilayer Skyrmionic Synapse for Deep Spiking Neural

Networks, Phys. Rev. Appl. 14, (2020).

[17] K. M. Song et al., Skyrmion-Based Artificial Synapses for Neuromorphic Computing,

Nat. Electron. 3, 148 (2020).

[18] D. Pinna, G. Bourianoff, and K. Everschor-Sitte, Reservoir Computing with Random

Skyrmion Textures, Phys. Rev. Appl. 14, 054020 (2020).

[19] O. Boulle et al., Room-Temperature Chiral Magnetic Skyrmions in Ultrathin Magnetic

Nanostructures, Nat. Nanotechnol. 11, 449 (2016).

35
[20] C. Moreau-Luchaire et al., Additive Interfacial Chiral Interaction in Multilayers for

Stabilization of Small Individual Skyrmions at Room Temperature, Nat. Nanotechnol. 11, 444

(2016).

[21] B. Göbel, A. F. Schäffer, J. Berakdar, I. Mertig, and S. S. P. Parkin, Electrical Writing,

Deleting, Reading, and Moving of Magnetic Skyrmioniums in a Racetrack Device, Sci. Rep. 9,

12119 (2019).

[22] F. N. Rybakov and N. S. Kiselev, Chiral Magnetic Skyrmions with Arbitrary Topological

Charge, Phys. Rev. B 99, 064437 (2019).

[23] R. Chen, Y. Li, V. F. Pavlidis, and C. Moutafis, Skyrmionic Interconnect Device, Phys.

Rev. Res. 2, 1 (2020).

[24] D. Foster, C. Kind, P. J. Ackerman, J. S. B. Tai, M. R. Dennis, and I. I. Smalyukh, Two-

Dimensional Skyrmion Bags in Liquid Crystals and Ferromagnets, Nat. Phys. 15, 655 (2019).

[25] J. Tang, Y. Wu, W. Wang, L. Kong, B. Lv, W. Wei, J. Zang, M. Tian, and H. Du,

Magnetic Skyrmion Bundles and Their Current-Driven Dynamics, Nat. Nanotechnol. (2021).

[26] F. N. Rybakov, A. B. Borisov, S. Blügel, and N. S. Kiselev, New Type of Stable

Particlelike States in Chiral Magnets, Phys. Rev. Lett. 115, 117201 (2015).

[27] G. P. Müller, F. N. Rybakov, H. Jónsson, S. Blügel, and N. S. Kiselev, Coupled

Quasimonopoles in Chiral Magnets, Phys. Rev. B 101, 184405 (2020).

[28] A. Vansteenkiste, J. Leliaert, M. Dvornik, M. Helsen, F. Garcia-Sanchez, and B. Van

Waeyenberge, The Design and Verification of MuMax3, AIP Adv. 4, 107133 (2014).

36
[29] P. Virtanen et al., SciPy 1.0: Fundamental Algorithms for Scientific Computing in

Python, Nat. Methods 17, 261 (2020).

[30] J. Xia, X. Zhang, O. A. Tretiakov, H. T. Diep, J. Yang, G. Zhao, M. Ezawa, Y. Zhou, and

X. Liu, Bifurcation of a Topological Skyrmion String, Phys. Rev. B 105, 214402 (2022).

[31] S. Zhang, D. M. Burn, N. Jaouen, J.-Y. Chauleau, A. A. Haghighirad, Y. Liu, W. Wang,

G. van der Laan, and T. Hesjedal, Robust Perpendicular Skyrmions and Their Surface

Confinement, Nano. Lett. 20, 1428 (2020).

[32] R. Lo Conte et al., Tuning the Properties of Zero-Field Room Temperature

Ferromagnetic Skyrmions by Interlayer Exchange Coupling, Nano. Lett. 20, 4739 (2020).

[33] H. Jia, B. Zimmermann, M. Hoffmann, M. Sallermann, G. Bihlmayer, and S. Blügel,

Material Systems for FM-/AFM-Coupled Skyrmions in Co/Pt-Based Multilayers, Phys. Rev.

Mater. 4, 094407 (2020).

[34] E. Liu et al., Synthetic-Ferromagnet Pinning Layers Enabling Top-Pinned Magnetic

Tunnel Junctions for High-Density Embedded Magnetic Random-Access Memory, Phys. Rev.

Appl. 10, 054054 (2018).

[35] L. Liu, W. Chen, and Y. Zheng, Skyrmion Transport Modified by Surface Terraces in

Magnetic Multilayers, Phys. Rev. Appl. 16, 014050 (2021).

[36] D. Cortés-Ortuño, W. Wang, M. Beg, R. A. Pepper, M. A. Bisotti, R. Carey, M.

Vousden, T. Kluyver, O. Hovorka, and H. Fangohr, Thermal Stability and Topological

Protection of Skyrmions in Nanotracks, Sci. Rep. 7, 4060 (2017).

37
[37] M. A. Bisotti, D. Cortés-Ortuño, R. Pepper, W. Wang, M. Beg, T. Kluyver, and H.

Fangohr, Fidimag – A Finite Difference Atomistic and Micromagnetic Simulation Package, J

Open Res. Softw. 6, 1 (2018).

[38] T. Schrefl, H. Forster, D. Suess, W. Scholz, V. Tsiantos, and J. Fidler, Micromagnetic

Simulation of Switching Events, in Advances in Solid State Physics (Springer Berlin Heidelberg,

Berlin, Heidelberg, 2007), pp. 623–635.

[39] L. Desplat and J. Von Kim, Entropy-Reduced Retention Times in Magnetic Memory

Elements: A Case of the Meyer-Neldel Compensation Rule, Phys. Rev. Lett. 125, 107201 (2020).

[40] M. H. Nguyen, D. C. Ralph, and R. A. Buhrman, Spin Torque Study of the Spin Hall

Conductivity and Spin Diffusion Length in Platinum Thin Films with Varying Resistivity, Phys.

Rev. Lett. 116, 126601 (2016).

[41] W. Legrand, J.-Y. Chauleau, D. Maccariello, N. Reyren, S. Collin, K. Bouzehouane, N.

Jaouen, V. Cros, and A. Fert, Hybrid Chiral Domain Walls and Skyrmions in Magnetic

Multilayers, Sci. Adv. 4, eaat0415 (2018).

[42] W. Jiang et al., Direct Observation of the Skyrmion Hall Effect, Nat. Phys. 13, 162

(2017).

[43] S. Zhang and Z. Li, Roles of Nonequilibrium Conduction Electrons on the Magnetization

Dynamics of Ferromagnets, Phys. Rev. Lett. 93, 127204 (2004).

[44] J. C. Slonczewski, Current-Driven Excitation of Magnetic Multilayers, J. Magn. Magn.

Mater. 159, L1 (1996).

38
[45] A. A. Thiele, Steady-State Motion of Magnetic Domains, Phys. Rev. Lett. 30, 230 (1973).

[46] R. Tomasello, E. Martinez, R. Zivieri, L. Torres, M. Carpentieri, and G. Finocchio, A

Strategy for the Design of Skyrmion Racetrack Memories, Sci. Rep. 4, 6784 (2015).

[47] A. Schleunitz and H. Schift, Fabrication of 3D Nanoimprint Stamps with Continuous

Reliefs Using Dose-Modulated Electron Beam Lithography and Thermal Reflow, J. Micromech.

Microeng. 20, 095002 (2010).

[48] D. H. Lee and S. H. Lim, Increase of Temperature Due to Joule Heating during Current-

Induced Magnetization Switching of an MgO-Based Magnetic Tunnel Junction, Appl. Phys. Lett.

92, 233502 (2008).

[49] M. Weisheit, S. Fahler, A. Marty, Y. Souche, C. Poinsignon, and D. Givord, Electric

Field-Induced Modification of Magnetism in Thin-Film Ferromagnets, Science (1979) 315, 349

(2007).

[50] T. Maruyama et al., Large Voltage-Induced Magnetic Anisotropy Change in a Few

Atomic Layers of Iron, Nat. Nanotechnol. 4, 158 (2009).

Supporting information for

Encoding and multiplexing information signals in

Runze Chen1, Yu Li1, William Griggs1, Yuzhe Zang1, Vasilis F. Pavlidis2, and Christoforos

Moutafis1

39
1
Nano Engineering and Spintronic Technologies (NEST) research group, Department of

Computer Science, The University of Manchester, Manchester M13 9PL, United Kingdom

2
Advanced Processor Technologies (APT) research group, Department of Computer Science,

The University of Manchester, Manchester M13 9PL, United Kingdom

40
Figure S1. Distinct FST states phase diagram in 4 MMLs system with different interlayer
exchange coupling. Phase diagram of the number of distinct stable FST states with various
external out-of-plane magnetic fields and DMI constant in a four magnetic layers system with the
interlayer coupling of (a) 0 mJ m-2, (b) 0.04 mJ m-2, (c) 0.08 mJ m-2, and (d) 0.12 mJ m-2,
respectively. The colour code is the same as in Fig. 1(b) in the main text, representing the number
of supported distinct states.

41
Figure S2. Distinct FST states phase diagram in 8 MMLs system with different interlayer
exchange coupling. Phase diagram of the number of distinct stable FST states as a function of
out-of-plane magnetic field and DMI constant in an 8MML system with the interlayer coupling of
(a) 0 mJ m-2, (b) 0.04 mJ m-2, and (c) 0.08 mJ m-2, respectively. The colour code represents the
number of supported distinct states.

42
Figure S3. Comparison of the binding energy of 2MML, 3MML, 4MML, and 6MML FSTs
with various amplitude of interlayer exchange coupling. The calculated total micromagnetic
energy as a function of in-plane displacement of the skyrmion in the top layer of the FSTs with the
amplitude of interlayer exchange coupling of (a) 0 mJ m-2, (b) 0.02 mJ m-2, (c) 0.04 mJ m-2, and
(d) 0.08 mJ m-2, respectively. The schematic drawing of 2MML to 6MML FSTs are illustrated in
the left panel of the figure. The 6 MML FST without interlayer exchange coupling exhibits
considerable binding energy, as shown in the inset of (a), which is attributed to the labyrinth
domain phase under these parameters.

43
Figure S4. Comparison of the binding energy with various amplitude of interlayer exchange
coupling as for (a) 2MML, (b) 3MML, (c) 4MML, and (d) 6MML FSTs, respectively. The
solid lines represent the energy change while shifting the top layer skyrmion and the colour code
stands for the amplitude of interlayer exchange coupling varying from 0 mJ m-2 to 0.2 mJ m-2.
Insets (a) to (d) depict the schematic illustration of shifting the top layer skyrmion in FSTs.

44
Figure S5. Phase diagram for nucleation probability of (a) 1MML FST, (b) 2MML FST, (c)
3MML FST, and (d) 4MML FST by changing the pulse width of the electric current and the
amplitude of the current density under room temperature. The schematic drawing of 1MML
to 4MML FSTs are illustrated in the left panel of the figure. The colour code represents the
probability of successful nucleation from 0% (white) to 50% (purple) to 100% (black). Each data
point illustrates the probability of successful nucleation out of 20 distinct attempts.

45
Figure S6. Evolution of the total micromagnetic energy during the chamber process in a
4MML skyrmionic multiplexing device. (a) Illustration of the chamber process in the device.
The energy change when (b) a 1MML FST, (c) a 2MML FST, and (d) a 3MML FST is chambered
into the 4MML transmission nanotrack. “Decouple” is observed when chambering 1MML FST in
(a), and “Nucleation” is observed in chambering a 3MML FST into the main track in (d), while
only the “Chamber” phase is observed when chambering the 2MML FST.

46