Dipolar magnetostirring protocol for three-well

atomtronic circuits
arXiv:2501.05301v1 [cond-mat.quant-gas] 9 Jan 2025

Hector Briongos-Merino1,2,† , Felipe Isaule3 , Montserrat

Guilleumas1,2 , Bruno Juliá-Dı́az1,2
1
Departament de Fı́sica Quàntica i Astrofı́sica, Facultat de Fı́sica, Universitat de
Barcelona, Martı́ i Franquès 1, E-08028 Barcelona, Spain
2
Institut de Ciències del Cosmos, Universitat de Barcelona, Martı́ i Franquès 1,
E-08028 Barcelona, Spain
3
Instituto de Fı́sica, Pontificia Universidad Católica de Chile, Avenida Vicuña
Mackenna 4860, Santiago, Chile.
E-mail: † [email protected]

Abstract. We propose a magnetostirring protocol to create persistent currents on an

annular system. Under this protocol, polar bosons confined in a three-well ring circuit
reach a state with high average circulation. We model the system with an extended
Bose-Hubbard Hamiltonian and show that the protocol can create circulation in an
atomtronic circuit for a range of tunable parameters. The performance and robustness
of this scheme are examined, in particular considering different interaction regimes. We
also present a method for predicting the optimal protocol parameters, which improves
protocol’s scalability and enables its application to systems with large numbers of
bosons. This overcomes computational limitations and paves the way for exploring
macroscopic quantum phenomena.

Keywords: Persistent currents, Dipolar interactions, Magnetostirring, Bose-Hubbard,

Atomtronics, Vortices in BECs.
Dipolar magnetostirring protocol for three-well atomtronic circuits 2

1. Introduction

Superfluidity, a hallmark of quantum fluids, is characterized by frictionless flow and

exotic phenomena such as quantized vortices and persistent currents [1]. In this line,
Bose-Einstein condensates (BECs) provide an ideal platform for exploring these effects
due to their macroscopic quantum coherence [2, 3]. In particular, persistent currents,
analogous to superconducting loops, can offer insights into the stability and quantum
phase coherence in ring-shaped condensates [4].
Recently, the atomtronics field has emerged as a promising platform to study the
rich quantum phenomena offered by ultracold atomic gases and to control them for
developing quantum devices [5, 6]. This includes persistent currents of neutral matter
waves, which convey a technological significance and are intensively studied within
atomtronics [7]. Indeed, Superconducting Quantum Interference Devices (SQUIDs)
have already been realized experimentally [8, 9]. In particular, atomic SQUIDs are
very promising tools in quantum sensing [10] as compact Sagnac interferometers [11] or
gyroscopes [12, 13], as well as candidates for the realization of the atomic analog of the
Mooij-Harmans qubit [14, 15, 16].
The creation of vortices in BECs with different geometries, such as in ring-
shaped configurations, can be realized by multiple techniques [17, 18]. These range
from potential stirring [19] to suitable Raman transitions [20] and phase imprinting
techniques. The latter employ commercially available devices for light sculpting such as
Spatial Light Modulators (SLM) or Digital Mirror Devices (DMD) [21, 22].
A potential additional method for controlling atomtronic circuits is the use
of ultracold dipolar particles, as these interact through a tunable dipole-dipole
interaction. Dipolar systems can be achieved with either magnetic atoms, such as
dysprosium [23, 24], erbium [25], chromium [26], and europium [27], or with polar
molecules [28, 29, 30]. The dipolar interaction is long-range and anisotropic, with
head-to-tail dipoles attracting each other, and side-by-side dipoles are repelled. These
properties give rise to very rich phenomena, even for the smallest systems [31, 32, 33].
In this work, we propose a protocol for creating persistent currents based on
the magnetostirring technique, which was developed for generating vortices in dipolar
gases [34]. This technique exploits the anisotropic long-range interactions to induce
asymmetry in the system, called magnetostriction [35], to later rotate the direction of
polarization, which induces a rotation in the condensate. This method was employed to
produce the first observed vortices in a dipolar BEC [34] and has been further used to
explore vortex lattices in the dominantly dipolar regime for dipolar BECs and supersolids
[36, 37]. Rapid magnetostirring is also known to be able to produce systems where
atoms remain stationary, but the dipole moments rotate rapidly [38]. This leads to
a time-averaged dipole-dipole interaction, known as an anti-dipolar interaction, where
head-to-tail anti-dipoles repel and side-by-side anti-dipoles attract.
We consider a fully connected triple well ring, the smallest system that supports
angular momentum with a superfluid phase [39, 40]. The ring is filled with dipolar
Dipolar magnetostirring protocol for three-well atomtronic circuits 3

Figure 1. Schematic representation of the triple well system (in the x-y plane, left
panel) and the movement of the dipole orientation µ during the protocol (right panel).

bosons to examine the creation of circulation with the magnetostirring-based protocol.

We model the system with an extended Bose-Hubbard Hamiltonian for three sites in
triangular geometry, to then solve the time evolution of the system and study the
creation of a persistent circulation. We find the optimal physical parameters for inducing
circulation and study the scheme’s robustness, as well as its feasibility in systems with
many bosons.
This work is organized as follows: Section 2 introduces the system and the extended
Bose-Hubbard Hamiltonian. Section 3 presents the proposed protocol to create circular
currents. Then in Section 4 we discuss the obtained results, analyzing the evolution
of the system and the performance of the protocol. Finally, Section 5 presents the
conclusions and outlook of this work.

2. Theoretical Framework

We consider N dipolar bosons confined in a ring formed by 3 sites, as depicted in Fig. 1.

We model the system with the following extended Bose-Hubbard Hamiltonian
3 3 3 X 3
UX Vjk
(â†j+1 âj + â†j âj+1 ) +
X X
Ĥ = −J n̂j (n̂j − 1) + n̂j n̂k , (1)
j=1 2 j=1 j=1 k̸=j 2

where âj (â†j ) are the bosonic annihilation (creation) operators for the j-th site, n̂j = â†j âj
is the particle number operator, J is the nearest neighbors tunneling strength, U is the
on-site interaction strength, and Vjk is the strength of the dipolar interaction between
sites j and k. The dipolar interaction term is anisotropic and depends on the relative
orientation of the dipoles [1]. In this work, we consider that all individual dipoles are
aligned in the same direction at any given time, which is achieved by polarizing them
Dipolar magnetostirring protocol for three-well atomtronic circuits 4

with an external electromagnetic field. Under this assumption, the dipolar interaction
has the form  
" #2
Ud  µ · (rj − rk ) 
Vjk = 1−3 , (2)
|rj − rk |3  |rj − rk | 

where Ud is the dipolar coupling constant, µ is the unitary dipole orientation, and rj −rk
is the vector that points from the k-th to the j-th site. Due to the geometry of the fully
connected three-well system, the term |rj − rk |3 is constant for all pairs of sites and
can be absorbed into Ud . Therefore, in the following, we will not explicitly write such
dependence, as it will be included in the constant Ud .
The Hamiltonian (1) commutes with the number operator N̂ = 3j=1 n̂j , i.e. the
P

system is number conserving, and thus we can subtract a factor (U/2)N̂ (N̂ − 1) to the
Hamiltonian when working in sectors of well defined N . This enables us to rewrite it in
the following convenient form
3 3 X
3 
Vjk −U

′
ĤI′ (â†j+1 âj â†j âj+1 )
X X
Ĥ = ĤJ + = −J + + n̂j n̂k . (3)
j=1 j=1 k̸=j 2
This transformation explicitly shows the competition between the inter-site (dipolar)
and on-site local interaction terms. This competition is highlighted when the dipole
orientation µ is aligned perpendicular to the plane in which the triple well is contained,
µ = µ ez , with ez the normalized vector along the z direction. In this configuration,
µ · (rj − rk ) = 0 ∀j, k, which simplifies the interaction term of Hamiltonian (3) to the
following expression:
3 X 3
1X
ĤI′ = (Ud − U ) n̂j n̂k . (4)
µ=µ ez 2 j=1 k̸=j
Tuning the on-site interaction strength U via Feshbach resonances [41] to match Ud
can nullify the interaction term in Eq. (4). This results in a non-interacting Hamiltonian,
and thus the system is described by free particles in a three-well ring potential.
To study the onset of persistent currents we examine the behavior of the current
operator. For a real tunneling parameter J, the total azimuthal current operator [7, 42],
also called the azimuthal circulation operator, is defined in a triple-well circuit as
3 
2π JmR2 X 
L̂z = i â†j+1 âj − â†j âj+1 ,
3 h̄ j=1
where m is the atomic mass of the bosons and R is the radius of the atomtronic circuit.
This operator commutes with the tunneling term of the Hamiltonian, sharing a common
eigenbasis. The eigenvectors and eigenvalues of the N -boson system can be constructed
as tensor products or sums, respectively, of the single-particle eigenstates and energies.
These single-particle eigenstates and energies are given by
!
2πk
λJ (k) = −2J cos ,
3
4π JmR2
!
2πk
λL (k) = sin ,
3 h̄ 3
Dipolar magnetostirring protocol for three-well atomtronic circuits 5
1  T
vk = √ 1, wk , w2k ,
3
where w = e2πi/3 and k = 0, 1, 2. The eigenvalues λJ (k) and λL (k) correspond
to the tunneling term of the hamiltonian, ĤJ and the circulation operator, L̂z ,
respectively. The eigenstates are Bloch waves written in the single-particle Fock basis
{|1, 0, 0⟩, |0, 1, 0⟩, |0, 0, 1⟩}; where the Fock vectors are noted as, |n1 , n2 , n3 ⟩.
At zero temperature and in the absence of interactions, all the bosons populate
the same single-particle state. However, interactions or excitations can remove particles
from such a single-particle ground state, promoting them into excited ones. This depletes
a fraction of the condensate, or may even cause its fragmentation for sufficiently strong
interactions. The condensed fraction fc of the system described by the many-body wave
function Ψ is given by the largest eigenvalue of the one-body density matrix operator
(OBDM) [43]
1
ρ̂j,k = ⟨Ψ|â†j âk |Ψ⟩, (5)
N
while its associated eigenvector is the so-called natural orbit of the system, the most
populated eigenstate. For a singly condensed system, the largest eigenvalue is close to
unity (fc ≃ 1), while the others are roughly zero. On the other hand, if the system is
fragmented, two or more eigenvalues will have a comparable magnitude [44].
To numerically study the time evolution of our extended Bose-Hubbard model,
we construct a Fock basis of N bosons over three sites, using exact diagonalization
techniques. The size of this basis grows exponentially with the number of bosons,
which limits the systems that are computationally affordable to a few tenths of particles
[45]. For the system with three sites, we are able to study up to 40 particles within
reasonable computing times, performing time evolution using the fourth-order Runge-
Kutta method.

3. Circulation creation protocol

The proposed protocol for the creation of circulation in the three-well system consists
of a dynamic change of the dipole alignment µ, which breaks the reflection symmetry
of the system. Initially, at t = 0, the dipoles are polarized in the y-axis direction [see
Fig. 1]. The initial state is chosen as the ground state of such configuration, in which
sites 2 and 3 concentrate most bosons [32, 33]. Then, the polarization direction changes
over time with a spherical spiral motion parametrized by
µ · ex = sin(ωxy t) cos (ωz t) ,
µ · ey = cos(ωxy t) cos (ωz t) ,
µ · ez = sin(ωz t) ,
where ωxy and ωz , correspond to the rotation frequencies that define the spherical
movement. This time dependence only affects the dipolar interaction term [equation
Dipolar magnetostirring protocol for three-well atomtronic circuits 6

(2)]. The polarization is then frozen at time tf = π/(2ωz ) when the dipoles are aligned
in the z-direction, that is, perpendicular to the circuit’s plane [see Fig. (1)].
The hopping, on-site interaction, and dipolar interaction strengths are constant
throughout the protocol. Also, the on-site interaction strength is chosen such that
U = Ud , thus at the end of the protocol the system evolves under a free-particle
Hamiltonian due to the vanishing of the interaction term in equation (4).
This protocol relies on two main ideas. Firstly, the circulation will be generated by
the exchange of bosons between the occupied and empty sites when the dipole orientation
starts to rotate. Secondly, the azimuthal circulation generated is preserved at the end
of the protocol due to the cancellation of the interactions.
It should be stated that the protocol parameters ωxy and ωz can be tuned to regulate
the system’s state at tf . In this work, we focus on the creation of the maximum
circulation in the most predictable way, and thus do not analyze other potentially
interesting final states.
Throughout this paper, the energy is provided in units of the hopping parameter
J and the time in units of h̄/J. The circulation is normalized to the maximum possible
2π JmR2
circulation of a free boson in a three-site ring, namely L0 = √ 3 h̄
.

4. Numerical results

4.1. Evolution of the system under the protocol

We begin by examining the evolution of some characteristic quantities over time to
understand how the system evolves under the protocol and how circulation is produced.
In Fig. 2 we study the evolution of the quantum state |Ψ(t)⟩ under the
magnetostirring protocol. Panel a) shows the projection of the instantaneous state
of the system onto the eigenstates of the time-dependent Hamiltonian as a function
of time. The spectrum of this many-body system is a set of values that change with
the dipole orientation, modifying gaps and creating crossings between the instantaneous
eigenstates. An important aspect to note is that at the end of the protocol, the energy
levels are equidistant, as it is the spectrum of a free Hamiltonian, discussed previously.
As required by the protocol, the initial state corresponds to the ground state of the
Hamiltonian at t = 0. Since the initial polarization lies in the x − y plane and it is
aligned along the y axis, the ground state presents an equal population in the sites 2
and 3 due to the attractive dipolar interaction, whereas site 1 is empty, as it is shown
in panel c).
As the Hamiltonian starts changing the initial gap closes, allowing the system’s
state to start overlapping with the excited states. This closing happens because the
dipole orientation approaches a configuration for which the ground state of this system
is degenerate. In fact, this happens for an in-plane polarization direction perpendicular
to one side of the triangle [32, 33]. When the state of the system shows a large overlap
with excitations, it populates many highly excited states while depopulating the low-
Dipolar magnetostirring protocol for three-well atomtronic circuits 7

Figure 2. Evolution of the state of a system with N = 8 bosons and interaction

strengths U/J = Ud /J = 5.0 under the magnetostirring protocol with ωxy = 1.07 h̄/J
and ωz = 1.32 h̄/J. a) Eigenenergies Ej and projection of the system’s state |Ψ(t)⟩ onto
the instantaneous eigenstates of the Hamiltonian, {|φj ⟩}|t , as a function of time. The
color map (upper bar) corresponds to the probability density |⟨Ψ|φj ⟩|2 in logarithmic
scale. The eigenenergies are rescaled on each step to improve visibility. b) Expected
value of the azimuthal circulation as a function of time. c) Expected occupation of
each of the three sites as a function of time.

lying ones. At the end of the protocol the state of the system is a combination of highly
excited Hamiltonian eigenstates, far from the ground state of the system. Furthermore,
the final state is a superposition of multiple circulation states.
The initial value of the azimuthal circulation is, as expected, zero. This value
starts increasing around half-time of the protocol when the system’s state is no longer
the instantaneous ground state. The growth stops before the end of the protocol
without oscillations, as the system’s Hamiltonian at the end of the protocol preserves the
circulation. This change in the expected value of the azimuthal circulation during the
protocol is shown in Fig. 2 b). Therefore, the protocol successfully creates a persistent
circulation for t > tf .
At the beginning of the protocol, due to the anisotropic character of the dipolar
Dipolar magnetostirring protocol for three-well atomtronic circuits 8

Figure 3. Expected value of the azimuthal circulation at the end of the protocol
of a system with N = 8 bosons and interactions U/J = Ud /J = 5.0 for different
combinations of ωxy and ωz .

interaction, only two sites are populated (sites 2 and 3). However, as the system evolves
with time, the occupation changes, decreasing the population in sites 2 and 3, while
increasing it in the initially unpopulated one (site 1). This population evolution is
shown in Fig. 2 c). It must be noted that at the end of the protocol, the populations
in the three sites are not equal, which is what one would expect for a pure circulation
state. Therefore, one can consider this imbalance as a signature of the superposition of
circulation eigenstates in the system.

4.2. Parameter optimization

It is necessary to choose appropriate values for the frequencies ωxy and ωz , the two
parameters of the time-dependent Hamiltonian, to maximize the circulation produced
after the end of the protocol, To explore the parameter landscape, we have computed
the expected final azimuthal circulation of the system for a fixed number of bosons and
interaction strengths as a function of ωxy and ωz . One representative exploration is
shown in Fig. 3.
Inside the parameter landscape, it is possible to identify different parameter
regimes. Firstly, it is expected that no circulation will be induced in systems with a very
fast quench, limiting the maximum value of ωz h̄/J parameter (minimum value of tf ).
This is evident from the figure, where values of ωz h̄/J > 10 fail to generate circulation.
On the other side, for low values of the azimuthal parameter (ωz h̄/J < 0.5), the protocol
time is long enough to let the system undergo multiple oscillations. In this region the grid
chosen cannot resolve the continuous evolution of the system, obtaining a noisy pattern.
Regarding the parameter ωxy , we observe that for values ωxy h̄/J < 0.1 no circulation is
produced. In this case as the movement is not fast enough to produce excitations in the
system (adiabatic) and the ground state has no average circulation. For large rotation
Dipolar magnetostirring protocol for three-well atomtronic circuits 9

Figure 4. Maximum circulation generated in the system after the protocol for different
interaction strengths (indicated in the labels) as a function of the number of bosons
N . The lines are a guide for the eye.

frequencies, values ωxy h̄/J > 11, no circulation is produced. The rapid variation of the
polarization direction cannot be followed by the bosons, and thus the evolution of the
system approaches to one under a static antidipolar interaction, which is just a time-
averaged interaction for dipolar gases with large dipole rotation frequencies [38]. Hence,
the useful regime for the protocol is that of intermediate frequencies at the center of the
figure, where a large lobe of positive circulation (red region) appears.
Therefore, we conclude that the protocol has a parameter range where circulation is
produced consistently. This lobe is not always located in the same region as it depends
on the number of particles and the strength of the interactions. As a result, the specific
optimal parameters of the system vary with the number of bosons and the strength of
the interaction. Interestingly, we show in Appendix A that the values of the optimal
parameters can be predicted for large systems.

4.3. Circulation’s dependence on the number of bosons

As previously seen in Figs 2 and 3, the maximum value of the circulation produced
for eight particles is not even half of the circulation such a system can have. However,
this changes as the number of particles in the system increases. The circulation that is
created in the system grows faster than the maximum circulation that the system can
2π 2
hold, which is N L0 = √ 3
N JmR
h̄
. Therefore, the protocol becomes more efficient in the
production of circulation as the number of bosons increases. In Fig. 4 we show the
growth of the maximum circulation generated as a function of the number of particles
for different interaction strengths with U = Ud .
We stress that the points shown in Fig. 4 are obtained for the optimal parameters
(ωz and ωxy ) for the chosen interaction strengths and the number of particles [the lobes
discussed in the previous subsection]. The series of points for each interaction strength
Dipolar magnetostirring protocol for three-well atomtronic circuits 10

Figure 5. Condensed fraction fc at the end of the protocol for different interaction
strengths (indicated in the labels) as a function of the number of bosons. The lines
are a guide for the eye.

starts from a different number of bosons because we do not show results for which the
optimal parameters are outside the lobes. Finding more circulation for a combination
of frequencies outside the lobe only happens for smaller numbers of particles, where
the protocol is not easy to optimize. This also allows us to argue that the protocol’s
usefulness increases with the number of bosons in the system.
We also note that the circulation produced for a small number of particles N < 10
is usually larger for stronger interactions. However, the increase of the maximum
circulation as a function of N is slower with larger U . This means that for a large
number of particles, there will be more circulation in systems in which the interaction
parameters are closer to the hopping strength.

4.4. Condensed fraction at the end of the protocol

After characterizing the creation of circulation with different parameters, we now study
its connection with the condensed fraction fc at the end of the protocol, which is
extracted from the OBDM [equation (5)]. Naively, one would expect a low condensed
fraction at the end of the protocol because of the excitation of the system. One also
would expect a decrease of fc for increasing values of the inter-particle interactions.
However, we find large values of the condensed fraction at the protocol. Furthermore, we
find that fc grows with the number of particles almost independently of the interactions.
This behavior happens due to the cancellation of the on-site and dipolar interactions
at the end of the protocol, therefore reducing the fragmentation. This is shown in Fig.
5, where the value of fc is depicted at the end of the protocol as a function of N for
different interaction strengths at the optimal protocol parameters used in Fig. 4.
For each interaction’s value, we also observe a local minimum of fc , after which it
starts to grow. This is in contrast with the expected average tendency of the condensed
Dipolar magnetostirring protocol for three-well atomtronic circuits 11

Figure 6. The final value of the circulation as a function of the on-site interaction
deviation ∆U/J for fixed values of Ud and multiple numbers of particles (indicated in
the labels), where the protocol parameters are the ones that optimize final circulation
at ∆U/J = 0.

fraction. The minimum of fc is inversely proportional to the value of the interactions,

thus recovering the idea that increasing interactions deplete a larger fraction of the
system. We include a more detailed analysis of the fragmentation of the system in
Appendix B, and a discussion about the validity of the representation under a condensed
mean-field treatment.

4.5. Robustness of the protocol

In the previous discussions, we have followed all the proposed requirements of the
protocol. In the following, we analyze the robustness of the protocol under slight
modifications of the on-site interactions and the number of particles.
First, we consider a variation of the on-site interaction parameters. In general, we
expect a decrease in the circulation generated in systems where the on-site interaction
U is larger than the dipolar Ud , as the stirring effect will be less relevant. On the other
hand, a weaker on-site interaction would seem preferable. This is true if the deviation
is relatively small, while for larger deviations fails. The imperfect cancellation of the
interactions at the end of the protocol also reduces the final circulation. This results in
the restriction for U and Ud to be similar to achieve the largest circulation. The final
circulation obtained considering a deviation ∆U ≡ U − Ud in the on-site interaction
strength is presented in Fig. 6. We employ the optimal protocol frequencies ωxy and
ωz obtained for U = Ud . The behavior is qualitatively similar for all the systems
analyzed, all of them showing the previously described features. For positive deviations
∆U/J > 0, the circulation produced in the system decreases as the deviation increases.
On the other hand, the final circulation grows while ∆U/J decreases, but for deviations
∆U/J < −0.2, it starts decreasing. .
Having examined the robustness of the protocol under a change in the interaction
Dipolar magnetostirring protocol for three-well atomtronic circuits 12

Figure 7. Final value of the circulation as a function of the number of bosons N .

In purple, the protocol frequencies are fixed for constant U/J, while, in green, the
protocol frequencies are fixed for constant (N − 1)U/J. The lines are a guide for the
eye.

strengths, we now examine its robustness under a change in the number of particles.
In Fig. 7 we show the final circulation as a function of N , having chosen the optimal
parameters for N0 = 15 particles and maintaining U/J (purple markers and lines) and
(N − 1)U/J (green markers and lines) constant.
For constant interactions U/J = Ud /J = U0 /J (purple markers), the circulation
shows a significant decrease after varying N . This means that, in general, the chosen
protocol frequencies do not remain optimal under a change in the number of particles.
On the other hand, by maintaining the value of (N − 1)U/J = (N − 1)Ud /J =
(N0 − 1)U0 /J constant (green markers), the circulation fraction increases with N . This
means that to maintain and even increase the circulation for a larger number of particles,
it is necessary to have a constant ratio between the interaction strengths U = Ud and
(N − 1)/J. More importantly, this enables one method to predict optimal protocol
frequencies for large N . Indeed, one can find the optimal frequencies for a small number
of particles, and then choose the interaction that maintains (N − 1)U/J constant for a
larger N .

5. Conclusions

We have proposed a functional magnetostirring protocol to create azimuthal circulation

on a dipolar system on a triple-well Bose-Hubbard ring. Our method consists on the
excitation of the system using a spherical spiral-like movement of the polarization
direction. We have shown that it can promote the bosonic system from the ground
state to an excited one where the average circulation is of the order of the largest one
that the system can sustain. This scheme extends the magnetostirring technique to
atomtronic rings, increasing the methods that can be used for the creation of persistent
Dipolar magnetostirring protocol for three-well atomtronic circuits 13

currents in such systems.

This protocol has only two free parameters, which are the two frequencies that
dictate the movement of the dipole orientation. We find that there is a set of parameters
where the final circulation is maximized. In addition, the circulation produced in the
system and the final condensed fraction increase with the number of bosons, highlighting
that this effect is more efficient in highly populated systems. We also demonstrate that
the protocol is robust under a small offset in the on-site interaction strengths, favoring
its experimental realization.
Finally, for systems with a large number of particles we have provided a feasible
way to obtain the optimal frequencies from few-particles studies. This prediction can be
done by finding the optimal protocol frequencies using a system with a small number of
bosons and the same value of (N − 1)U/J = (N − 1)Ud /J as the desired system, to later
use those parameters for a system with many more bosons. This property improves the
scalability of the protocol, as it enables to reach systems computationally intractable
to simulate using the Bose-Hubbard model, opening new avenues for the exploration of
macroscopic quantum phenomena.
These results could be the ground for future developments in the field of
atomtronics, and a useful tool for creating and manipulating persistent currents. Further
studies could explore larger rings and different ring configurations, focusing on their
utility as rotation or gravity sensors.

Acknowledgments

We acknowledge useful and constructive discussions with Abel Rojo-Francàs.

This work has been funded by Grant PID2023-147475NB-I00 funded by MI-
CIU/AEI/10.13039/501100011033 and FEDER, UE, by Grant No. 2021SGR01095 from
Generalitat de Catalunya, and by Project CEX2019-000918-M of ICCUB (Unidad de
Excelencia Marı́a de Maeztu). H. B.-M. is supported by FPI Grant PRE2022-104397
funded by MICIU/AEI/10.13039/501100011033 and by ESF+. F.I. acknowledges fund-
ing from ANID through FONDECYT Postdoctorado No. 3230023.

Appendix A. Parameter prediction

As mentioned in Section 4.5, it is possible to predict optimal protocol parameters for

computationally intractable systems by exploiting the behavior observed in Fig. 7. For
fixed values of the protocol frequencies, the circulation introduced in the system grows
with the number of particles if (N − 1)U/J = (N − 1)Ud /J is kept constant. This allows
one to reuse the optimal parameters found for a system with few particles N1 (cheap
to compute) and employ them to efficiently introduce circulation on a system with a
larger number of particles N2 (more expensive or impossible to compute) as long as the
Dipolar magnetostirring protocol for three-well atomtronic circuits 14

following relation holds:

N2 − 1 U1 /J1
=
N1 − 1 U2 /J2
where U1 (J1 ) and U2 (J2 ) are the on-site interaction (hopping) strengths for the first
and second systems, respectively.
In Fig. A1 we show the values for the optimal frequencies of the protocol for the
data presented in Figs. 4 and 5. The optimal frequencies show a decaying power law
for increasing (N − 1)U/J = (N − 1)Ud /J, increasing the final time of the protocol.
The data presented also includes a tentative fitting to facilitate the prediction for high
values of the parameter, which can be used as a starting point for the search for the
optimal parameters in future research and applications.

Figure A1. Optimal protocol frequencies ωxy (left panel) and ωz (right panel) as a
function of (N − 1)U/J for the interaction strengths indicated in the labels. The area
shaded in red is the confidence interval of the fitting.

Appendix B. Mean-field description

Appendix B.1. Coherent mean field model

When the system is fully condensed, the state of N bosons can be written as a product
state of N identical single-particle states. In our three-well system, the state depends
only on three complex coefficients that represent the projection of the single-particle
state onto each of the sites of the ring. Thus, such a state can be written as
N
|ψ⟩, |ψ⟩ = (χ1 , χ2 , χ3 )T ,
O
|Ψ⟩ =
j=1

which evolves under a coherent mean-field set of equations [46]:

∂χj
= −J (χj+1 + χj−1 ) + (N − 1) (Vjk − U )|χk |2 χj .
X
ih̄
∂t k̸=j

In this evolution, the condensed fraction remains one throughout the evolution.
We stress that these equations provide only an approximation for our model. This is
because time-dependent Hamiltonians and interactions usually deplete a fraction of the
Dipolar magnetostirring protocol for three-well atomtronic circuits 15

Figure B1. Expected value of the azimuthal circulation at the end of the protocol of a
system with N = 8 and U/J = Ud /J = 5, leading to (N −1)U/J = (N −1)Ud /J = 35.0,
for multiple combinations of ωxy and ωz . Computations are done using the mean-field
model.

condensate, as shown in the main text. However, previous studies [33] have shown that
dipolar interactions prevent the system’s fragmentation for contact interaction strengths
of the same order, and thus these mean-field equations can provide a good approximation
for certain parameters.
This mean-field model represents the state of the system with only three parameters,
and its computational cost is independent of the number of bosons N in the system.
Therefore, we can use this approximation to extract results for large number of particles.

Appendix B.2. Coherent mean-field results

First, to benchmark the mean-field model, in Fig. B1 we show the circulation at the
end of the protocol for the same parameters used in Fig. 3, which is obtained from
exact calculations. The obtained final circulation resembles that obtained in the main
text. The relevant range of the parameters is qualitatively the same, but the pattern in
which the circulation is produced has changed. Nevertheless, the obtained circulations
still present a large lobe of positive circulation on the same parameter region, which can
help predict the optimal protocol frequencies. Moreover, it is important to highlight that
in the region where the pattern is not resolved, the circulation values obtained with the
mean-field model are much larger than those obtained using the exact Bose-Hubbard
Hamiltonian.
This disagreement between the results obtained using the two models shows that the
assumption that the system is condensed is not always fulfilled. There is no guarantee
that the system will remain condensed under all the protocol frequencies.
Dipolar magnetostirring protocol for three-well atomtronic circuits 16

Figure B2. Condensed fraction fc obtained with exact Bose-Hubbard model at the
end of the protocol. The threshold of fc = 0.75 is highlighted with the black line.

Appendix B.3. Validity of the mean-field model

To further test and compare whether and in which region the mean-field model can be
used to calculate the results for the stirring protocol, in Fig. B2 we show the final
condensate fraction computed using the Bose-Hubbard model. We also considered
a qualitative limit of a condensate fraction of 0.75 to separate the condensed and
fragmented systems. In the figure, we observe that the protocol fragments the system
in the same parameter region where circulation is produced. The latter means that the
system is being excited and, in most cases, the state cannot be written as a product
of single-particle states, which is an already expected result. We also observe that for
sufficiently small ωxy < 0.1 J/h̄, the system remains condensed, which is a fingerprint
of the onset of a critical excitation velocity of superfluid condensates.
To recognize if the optimal realization of the protocol can be well represented in the
condensed model, we have represented in Fig. B3 the final circulation produced in the
system with a shadowed area that represents the instances in which the condensed
fraction of the system is less than 0.75. These two figures clearly show that the
main cases where circulation is produced cannot be studied under the mean-field
approximation. Despite that, the big lobe in which we are interested is not completely
covered, meaning that the approximation can still be useful. Moreover, the results of the
circulation production for those instances in which the system is condensed are almost
identical.
To extract useful information with the mean-field model, one has to restrict the
calculations to the region where the system is condensed. Large differences appear in
the region in which this model is not valid. For example, looking at the best overall
configurations, [yellow stars in Fig. B3], we can observe that there is a noticeable
difference between the two approaches. Nevertheless, the mean-field approach is useful
for exploring large systems under the mentioned constraints, as discussed.
Dipolar magnetostirring protocol for three-well atomtronic circuits 17

Figure B3. Comparison between the final circulation obtained at the end of the
protocol using exact Bose-Hubbard (left panel), and coherent mean-field model (right
panel). The shadowed region marks those protocol parameters for which the final
condensed fraction is less than 0.75. The yellow star shows the position of the optimal
parameters.

References

[1] Pitaevskii L and Stringari S 2016 Bose-Einstein Condensation and Superfluidity (Oxford
University Press) ISBN 978-0-19-875888-4 URL https://doi.org/10.1093/acprof:oso/
9780198758884.001.0001
[2] Rayfield G W and Reif F 1964 Physical Review 136 A1194–A1208
[3] Matthews M R, Anderson B P, Haljan P C, Hall D S, Wieman C E and Cornell E A 1999 Physical
Review Letters 83 2498–2501
[4] Yakimenko A, Vilchinskii S, Bidasyuk Y, Kuriatnikov Y, Isaieva K and Weyrauch M 2015
Romanian Reports in Physics 67 249–272
[5] Amico L, Boshier M, Birkl G, Minguzzi A, Miniatura C, Kwek L C, Aghamalyan D, Ahufinger
V, Anderson D, Andrei N, Arnold A S, Baker M, Bell T A, Bland T, Brantut J P, Cassettari
D, Chetcuti W J, Chevy F, Citro R, De Palo S, Dumke R, Edwards M, Folman R, Fortagh J,
Gardiner S A, Garraway B M, Gauthier G, Günther A, Haug T, Hufnagel C, Keil M, Ireland P,
Lebrat M, Li W, Longchambon L, Mompart J, Morsch O, Naldesi P, Neely T W, Olshanii M,
Orignac E, Pandey S, Pérez-Obiol A, Perrin H, Piroli L, Polo J, Pritchard A L, Proukakis N P,
Rylands C, Rubinsztein-Dunlop H, Scazza F, Stringari S, Tosto F, Trombettoni A, Victorin N,
von Klitzing W, Wilkowski D, Xhani K and Yakimenko A 2021 AVS Quantum Science 3 039201
ISSN 2639-0213
[6] Polo J, Chetcuti W J, Domanti E C, Kitson P, Osterloh A, Perciavalle F, Singh V P and Amico
L 2024 Quantum Science and Technology 9 030501 ISSN 2058-9565
[7] Polo J, Chetcuti W J, Haug T, Minguzzi A, Wright K and Amico L 2024 Persistent currents in
ultracold gases URL https://arxiv.org/abs/2410.17318v1
[8] Cominotti M, Rossini D, Rizzi M, Hekking F and Minguzzi A 2014 Physical Review Letters 113
025301
[9] Aghamalyan D, Cominotti M, Rizzi M, Rossini D, Hekking F, Minguzzi A, Kwek L C and Amico
L 2015 New Journal of Physics 17 045023 ISSN 1367-2630
[10] Degen C L, Reinhard F and Cappellaro P 2017 Reviews of Modern Physics 89 035002
[11] Helm J L, Cornish S L and Gardiner S A 2015 Physical Review Letters 114 134101
[12] Gustavson T L, Bouyer P and Kasevich M A 1997 Physical Review Letters 78 2046–2049
[13] Adeniji O, Henry C, Thomas S, Sapp R C, Goyal A, Clark C W and Edwards M 2024 Double-target
BEC atomtronic rotation sensor (Preprint 2411.06585)
[14] Mooij J E and Harmans C J P M 2005 New Journal of Physics 7 219 ISSN 1367-2630
Dipolar magnetostirring protocol for three-well atomtronic circuits 18

[15] Astafiev O V, Ioffe L B, Kafanov S, Pashkin Y A, Arutyunov K Y, Shahar D, Cohen O and Tsai
J S 2012 Nature 484 355–358
[16] Gallemı́ A, Muñoz Mateo A, Mayol R and Guilleumas M 2016 New Journal of Physics 18 015003
ISSN 1367-2630
[17] Goldman N, Juzeliūnas G, Öhberg P and Spielman I B 2014 Reports on Progress in Physics 77
126401 ISSN 0034-4885
[18] Dalibard J, Gerbier F, Juzeliūnas G and Öhberg P 2011 Reviews of Modern Physics 83 1523–1543
[19] Fetter A L 2009 Reviews of Modern Physics 81 647–691
[20] Andersen M F, Ryu C, Cladé P, Natarajan V, Vaziri A, Helmerson K and Phillips W D 2006
Physical Review Letters 97 170406
[21] Kumar A, Dubessy R, Badr T, De Rossi C, de Goër de Herve M, Longchambon L and Perrin H
2018 Physical Review A 97 043615
[22] Henderson K, Ryu C, MacCormick C and Boshier M G 2009 New Journal of Physics 11 043030
ISSN 1367-2630
[23] Lu M, Burdick N Q, Youn S H and Lev B L 2011 Physical Review Letters 107 190401
[24] Tang Y, Burdick N Q, Baumann K and Lev B L 2015 New Journal of Physics 17 045006 ISSN
1367-2630
[25] Aikawa K, Frisch A, Mark M, Baier S, Rietzler A, Grimm R and Ferlaino F 2012 Physical Review
Letters 108 210401
[26] Griesmaier A, Werner J, Hensler S, Stuhler J and Pfau T 2005 Physical Review Letters 94 160401
[27] Miyazawa Y, Inoue R, Matsui H, Nomura G and Kozuma M 2022 Physical Review Letters 129
223401
[28] Bohn J L, Rey A M and Ye J 2017 Science 357 1002–1010
[29] Langen T, Valtolina G, Wang D and Ye J 2024 Nature Physics 20 702–712 ISSN 1745-2481
[30] Bigagli N, Yuan W, Zhang S, Bulatovic B, Karman T, Stevenson I and Will S 2024 Nature 1–5
ISSN 1476-4687
[31] Lahaye T, Menotti C, Santos L, Lewenstein M and Pfau T 2009 Reports on Progress in Physics
72 126401 ISSN 0034-4885
[32] Gallemı́ A, Guilleumas M, Mayol R and Sanpera A 2013 Physical Review A 88 063645
[33] Rovirola M, Briongos-Merino H, Juliá-Dı́az B and Guilleumas M 2024 Physical Review A 109
063331
[34] Klaus L, Bland T, Poli E, Politi C, Lamporesi G, Casotti E, Bisset R N, Mark M J and Ferlaino
F 2022 Nature Physics 18 1453–1458 ISSN 1745-2481
[35] Baillie D and Blakie P B 2012 Physical Review A 86 023605
[36] Bland T, Lamporesi G, Mark M J and Ferlaino F 2023 Comptes Rendus. Physique 24 1–20 ISSN
1878-1535 (Preprint 2303.13263)
[37] Casotti E, Poli E, Klaus L, Litvinov A, Ulm C, Politi C, Mark M J, Bland T and Ferlaino F 2024
Nature 1–5 ISSN 1476-4687
[38] Giovanazzi S, Görlitz A and Pfau T 2002 Physical Review Letters 89 130401
[39] Arwas G, Vardi A and Cohen D 2014 Physical Review A 89 013601
[40] Gallemı́ A, Guilleumas M, Martorell J, Mayol R, Polls A and Juliá-Dı́az B 2015 New Journal of
Physics 17 073014 ISSN 1367-2630
[41] Chin C, Grimm R, Julienne P and Tiesinga E 2010 Reviews of Modern Physics 82 1225–1286
[42] Escrivà A, Mateo A M, Guilleumas M and Juliá-Dı́az B 2019 Physical Review A 100 063621
[43] Löwdin P O 1955 Physical Review 97 1474–1489
[44] Mueller E J, Ho T L, Ueda M and Baym G 2006 Physical Review A 74 033612
[45] Raventós D, Graß T, Lewenstein M and Juliá-Dı́az B 2017 Journal of Physics B: Atomic, Molecular
and Optical Physics 50 113001 ISSN 0953-4075
[46] Gallemı́ A, Guilleumas M, Martorell J, Mayol R, Polls A and Juliá-Dı́az B 2016 New Journal of
Physics 18 075005 ISSN 1367-2630