Professional Documents
Culture Documents
net/publication/353096274
CITATIONS READS
0 48
4 authors, including:
Stéphane Régnier
Sorbonne Université
312 PUBLICATIONS 3,834 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Johan Edilberto Quispe Navarrete on 04 August 2021.
Abstract—Helical swimmers, actuated by low-strength uni- A large number of studies on the helical swimming effi-
form rotating magnetic fields, can develop swimming and ciency have been carried out experimentally [8, 9, 10, 11]
pumping in low Reynolds number environments. They could and through simulations. In that regard, the versatility of
play an important role for future in vitro and in vivo biomedical
microrobotic tasks. Studying how their morphology influences numerical approaches was a key during optimization of
swimming and pumping tasks is then of importance. In this geometrical aspects and different physical conditions.
paper, we focus on two geometrical aspects for optimizing Previous works based on boundary element method (BEM)
both tasks: their helical shape and their cross-section. As a have provided an important framework to find optimal
first contribution, we investigate the optimal performance of helical structures at swimming. Several aspects such as he-
swimming and pumping. As a second contribution, we elucidate
between optimal shapes at performing swimming and pumping lical shape variation [12, 13], cross-section orientation [14],
at the same time. This study based on numerical simulation degree of confinement [15] could be analysed. Other works
is intended to serve as a guiding reference for building with CFD have studied the effect on the swimming of differ-
optimal helical structures either for swimming, pumping, or ent helical shapes on propulsion speed was explored [16]. In
a combination of both tasks. this paper, we address the effect of the cross-section and the
Index Terms—Micro/Nano Robots, Automation at Micro- helical shape to analyse their combined effect on swimming.
Nano Scales. Moreover, we provide an interesting framework to study
helical swimming through two criteria concerning the type
of magnetic source.
I. INTRODUCTION Regarding pumping, different aspects were studied such as
the flux generation [15], thrust [17], viscoelastic media [18],
W IRELESS microrobots are an attractive solution for
microassembly [1], local remote sensing [2], and
future medical interventions [3]. Due to their compactness
and mixing capabilities [19]. On the other hand, efficiency at
pumping has not been totally covered. Besides, the optimal
and ease to actuate in very restrained environments, He- operation of swimming and pumping at the same time
lical Swimmers (HS) stand out for their common actuation has not been deeply covered in the literature. And, in the
method that requires low-strength uniform rotating magnetic envisaged medical operations, swimmers might be required
fields of few milliteslas. to perform both tasks in one mission e.g. to arrive at blood
The importance of HS structures (HSs) relies on the func- clots sites and apply a constant pumping action to reestablish
tionalities they can display. The first is the swimming ca- the blood flow.
pability: HSs would be capable of navigating through low Numerical methods such as RFT and SBT methods are based
Reynolds (Re) fluids in the human conduits to transport on the partition of large bodies in segments that contribute
payloads such as drugs [4] or sperm cells [5]. The second independently to the total propulsion of the structure. Even
functionality is pumping fluids in obstructed conduits, for if those techniques are computationally less expensive com-
mixing fluids [6] or dispensing microparticles [7]. There- pared to BEM and CFD, aspects such as the shape or cross-
fore, studying how their morphological changes affect their section geometry, i.a. , cannot be analysed. In this work, we
performance when realizing certain of such tasks is of great exploit CFD to model helical swimming and pumping in
interest. We intend with this work, to provide guidelines for Newtonian fluids. This enables us to compute the generated
optimizing HSs for pumping and swimming. flux at pumping as well as to investigate geometrical aspects
that RFT and SBT methods cannot deal with e.g. low aspect
Manuscript received: February, 24, 2021; Revised May, 22, 2021; Ac- ratios, small helical radius, and the cross-section elongation
cepted June, 16, 2021. and orientation.
This paper was recommended for publication by Editor Xinyu Liu upon The formulation of the optimization problem is treated in
evaluation of the Associate Editor and Reviewers’ comments. This work
is part of Multiflag project (ANR-16-CE33-0019) funded by the French section II. In section III and IV we analyse the different
National Research Agency (ANR) and supported by EIPHI Graduate School criteria for swimming and then for pumping. In section V
(Contract No. ANR-17-EURE-0002) we analyse the optimization of both tasks at the same
+ Corresponding author: quispe@isir.upmc.fr
1 ISIR, Sorbonne Univ., CNRS, Paris, France time. Finally, we provide a summary of the relevant results
2 FEMTO-ST, UBFC, CNRS, Besançon, France obtained in this work in section VI.
3 ICube, Univ. of Strasbourg, INSA Strasbourg, Strasbourg, France
Digital Object Identifier (DOI): see top of this page.
© 2021 EU Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in IEEE Robotics and Automation Letters. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LRA.2021.3095291
ϕ= 0°
represent the helical centerline of HS structure (Fig. 1):
λs (c) b
ψ(s) = R cos(s)x̂ + R sin(s)ŷ + ẑ (1) b
p
2π p
n
a
b ϕ
with ψ : R → R3 , s ∈ [0, 2πN], R the helical radius, λ the n
pitch value, and N the number of turns. L
The HSs are considered initially to have a circular cross-
section with radius r. To compare structures during the
λ t
optimization, we set a constant total volume, V = Lπr2 , (d) ∂Ωfs
with L the total flagellum length. By doing that, we can
Ωf
reduce the number of possible geometrical configurations
z Rc
in Eq. (1). For that purpose, we introduce a dimensionless θ
set of parameters {β , θ , Rr } with the aspect ratio β = 2r L
, y ∂Ωfc
the helical angle θ = arcsin(t · ẑ), and the normalized helical boundaries
© 2021 EU Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in IEEE Robotics and Automation Letters. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LRA.2021.3095291
QUISPE et al.: ENHANCING SWIMMING AND PUMPING PERFORMANCE OF HELICAL SWIMMERS AT LOW REYNOLDS NUMBERS 3
θ( )
(a) U>0 (b) U=0 propulsion swimming speed (cf. Fig. 2(a)), which in the case
(b) ω of pumping is zero (cf. Fig. 2(b)). Finally, ω = ωẑ is the
(b) ω
HS angular velocity. The final boundary condition, defined
for the limits of the container (∂ Ω f c ), is:
u = 0 in ∂ Ω f c × [0,t0 ] (10)
fluid
which indicates a zero fluid speed in the fluid-container
boundary. Then, the fluidic force ff on the HS is computed
u by integrating the total stress (Γ) over all the HS structure’s
surface. Z
ff = Γn f ds (11)
∂Ωfs
Figure 2: Different tasks of HSs. (a) Swimming. (b) pumping, in
this task a tethered structure is considered (U = 0) and a flow is being n f the normal vector to the surface of each finite
generated in the opposite direction of the thrust (T ). element on the robot surface, and ds the surface element.
Analogously, we can compute the fluidic torque τf as fol-
lows:
the single-domain is magnetized parallel to the external
Z
magnetic field, achieving shape-independent magnetic prop- τf = (r − r0 ) × Γn f ds (12)
∂Ωfs
erties. In this work, let us consider that all HSs have
the same quantity of magnetic material Vm as well as the Finally, as HSs are immersed in a low Re fluid, getting the
volumic density of the magnetic moment m. In that regard, steady state is quasi-immediate as the sum of external forces
the method used in [20] can apply for our subsequent and torques is null:
evaluation. We consider structures with same volume that fm + ff = 0 τm + τf = 0 (13)
in average can keep the same quantity of magnetic material,
if such particles are magnetized through the HS easy axis, in
average, the magnetization of such structures will be almost
the same. Other cases where this study could be applicable D. Criteria For Optimization
is when using magnetic heads [23], nonetheless, in such a
Authors in [25] provides three different coefficients for
case, the head drag have to be negligible to not alter the
evaluating the optimal swimming performance, namely,
helical contribution to the thrust. Under such conditions, the
energetic efficiency, swimming efficiency, and propulsion
differential factor for optimizing those structures relies on
efficiency. To compute those coefficients, they made use
the fluidic torque τf generated by them.
of the swimming speed U (Fig. 2), the motor torque τ
and drag coefficients given by RFT. A helical flagellum
C. Fluid and Helical Structure Dynamics was then considered with a head as a base structure. Then,
HSs develop a rotary motion for swimming and pumping using the head drag coefficient A0 , the energy efficiency was
under low Reynolds number Re conditions. The fluid flow computed as the ratio Aτω 0U
. On the other hand, the drag
generated by HS structure is then governed by Navier- effect was not considered for the definition of swimming
U
Stokes (N-S) equations in the creeping flow regime given efficiency, computed as the ratio τω . In [12] a dimensionless
2
as follows: expression of swimming efficiency µLτ U is proposed, with
∇ · Γ = 0 in Ω f × [0,t0 ] (4) τ the fluidic torque and L the flagellum length of the HS. In
Γ = −pI + µ(∇u + (∇u)T ) (5) this work, we consider HSs without head or the case where
the head drag can be negligible. Then, the latter expression
∇ · u = 0 in Ω f × [0,t0 ] (6) is the most adapted. Swimming efficiency coefficient makes
with u the fluid velocity field, p the pressure, ρ the fluid the optimization problem to be interpreted as to find the
density and µ its dynamic viscosity. Γ represents the stress optimal structure that generates a greater speed consuming
on an infinitesimal fluid volume, Ω f depicts the fluid domain the least possible torque. On the other hand, the propulsion
U
(Fig. 1(d)), and [0,t0 ] the temporal domain. The boundary efficiency, defined as ω , and can be regarded as a possible
condition in the fluid-structure domain ∂ Ω f s are given by: measurement of the linear distance covered per rotation.
2
Previous expressions, µLτ U and ω U
, stand for two different
u = us in ∂ Ω f s × [0,t0 ] (7) ways of focusing on the optimization problem considering
dr the nature of the power source. The former definition
us = in ∂ Ω f s × [0,t0 ] (8) indicates that we have a limited magnetic source, and in
dt
order to find an optimum HS structure, we have to search
us = U + ω × (r − r0 ) in ∂ Ω f s × [0,t0 ] (9)
for structures that outperforms the others with the least
where us is the HS structure velocity in one point r situated possible torque. The latter definition refers to the fact that
on its surface. r0 is the HS center of mass, U = U ẑ is the we have an infinite magnetic source, typically a few teslas,
© 2021 EU Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in IEEE Robotics and Automation Letters. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LRA.2021.3095291
(μm/s)
Rd
Rd
Rd
Figure 3: Color-map depicting the swimming speed (in µm/s) of structures with the same volume V in the configuration space defined
by {β , θ , Rr , }. To compute these results, HSs were set at constant angular speed ω = 2π Hz. Rd represents a distance equivalent
to 18r100 = 0.6318µm, where r100 = 3.51 × 10−2 µm is the flagellum radius r for the structure with aspect ratio β = 100. Then,
18r100 ≈ 14.3r50 ≈ 10.5r20 .
which always provides the necessary torque to rotate the HS, III. SWIMMING PERFORMANCE OPTIMIZATION
thus, demonstrating no step-out frequencies or frequencies at In this section, the fluid and structure dynamics is treated
which the magnetic torque τm is not strong enough to rotate with the finite element method through a commercial soft-
the HS structure synchronously with B. Therefore, in such a ware, COMSOL Multiphysics. For that purpose, we used the
case, evaluating the swimming speed per rotation indicates CFD-package with the Math package “arbitrary Lagrangian-
an optimum performance. Eulerian” method (ALE) for treating the structure mesh
To evaluate pumping, in [26, 17] the total thrust (T = ff · ẑ) displacement. Simulations were generated using a Intel(R)
generated by HSs is used as a study variable. In [15], authors Core(TM) processor with 28GB-RAM. The mesh resolu-
introduce the total flux I generated Rby HSs for studying tion was extremely fine (about 200000 elements on the
pumping. They compute the flux I = H u · nH ds in a plane HS surface). The computation time for each simulation in
H perpendicular to axis of the HS, with u the fluid velocity the transitory regime is between 5 and 25 min depending
field and nH the normal vector of plane H. In those cases, on the geometry of the HSs. The swimming of HSs is
the pumping action is evaluated under the criterion of an studied considering an angular speed ω of 2π Hz, which
infinite magnetic source. However, for future HS that must is equivalent to 1 rotation per second. The choice of this
accomplish pumping and swimming tasks for one mission, value is appropriate since either flux and thrust have linear
the evaluation under the limited magnetic source criterion is behaviours with respect to ω. Afterwards, the choice of
of high importance. the constant volume V was based on the flagellum bundle
The criteria used in our study consists in evaluating sepa- volume of the E. coli bacteria considered in [27, 15, 28],
rately the swimming and pumping, considering the two types the values of L and r are the ones considered in [15, 28].
of magnetic sources namely, an infinite and a limited one. Since L and R are 7.1 µm and 0.035 µm respectively, the
Afterward, we analyse the performance of both tasks at the computed volume is about 2.73 × 10−20 m3 , ensuring that
same time. the structures characteristic length will be in the microscale,
For swimming, we evaluate propulsion efficiency with the Re can be computed as ρLU/µ, being in this study about
situation of infinite magnetic source by proposing the fol- 7.1 × 10−6 , the structures speed U being about 1µm/s, and
lowing expressions as study variables: the swimming speed the fluid properties density ρ and viscosity µ being that of
ω
U and the dimensionless speed LUf with f = 2π . Then, to distilled water.
evaluate the case where we account with a limited magnetic
2
source, we use the dimensionless swimming efficiency µLτ U . A. Helical Shape Influence
For pumping, we treat the case of the infinite source through Here, circular cross-sections are considered ( ab = 1).
analysing directly the total thrust T generated by HS struc- Figure 3 depicts the swimming speed U of the different
I
ture and the normalized flux ωR 3 at the middle plane of the structures in the space of variables {θ , Rr } for aspect ra-
c
HS (Hm ) with Rc the container radius. To treat the case of the tios β = {20, 50, 100} corresponding to total lengths L =
finite source we propose for the first time the dimensionless {2.41, 4.43, 7.03}µm and flagellum radius {r20 = 6.01, r50 =
energetic efficiency at pumping expression, TτL . 4.43, r100 = 3.51}10−8 m, respectively. Optimum propulsion
Finally, to consider at the same time swimming and pump- efficiency is achieved for HSs with a helical angle θ ∈
ing, we build an objective function g that is the weighed sum [40◦ , 60◦ ] and a normalized helical radius Rr with the largest
of swimming and energetic efficiencies using a coefficient possible value. Knowing that the scale Rr is different in
α ∈ [0, 1] to indicate the balance between swimming and each color-map for each aspect ratio β depicted in Fig. 3, a
pumping priorities: g = α SS0 + (1 − α) PP0 . fixed radial distance Rd for each color-map will be situated
© 2021 EU Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in IEEE Robotics and Automation Letters. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LRA.2021.3095291
QUISPE et al.: ENHANCING SWIMMING AND PUMPING PERFORMANCE OF HELICAL SWIMMERS AT LOW REYNOLDS NUMBERS 5
(a) 1.8 β = 20
(b) 1.8 β = 20
1.4 1.4
U (μm/s)
U (μm/s)
1.0 1.0
0.6 0.6
0.2 0.2
6 10 14 18 20 40 60 80
R/r θ (°)
(c) 1.8 (d) R/r = 18
θ = 50° 1.4
1.4
U (μm/s)
U (μm/s)
1.2
1.0
0.8
0.6
0.
0.4
0.2 20 40 60 80
6 10 14 18 Figure 5: Optimization of HSs with aspect ratio β = 20. (a)
R/r θ (°) 2
Performance per rotation ( LUf ). (b) Performance per torque ( µLτ U ).
Figure 4: Dependence of the swimming propulsion speed U on the (c) Fluidic torque τ. (d) θ vs. LUf for different Rr values.
configuration space set {β , Rr , θ }. (a) Rr vs. U for θ ∈ [10, 80] given
β = 20. (b) θ vs. U for Rr ∈ [6, 18] given β = 20. (c) Rr vs. U for
β ∈ {20, 50, 100} given θ = 50◦ . (d) θ vs. U for β ∈ {20, 50, 100} then considered of quasi-zero filament width, and moreover,
given Rr = 18.
the force interaction with the neighboring points is ne-
glected. Additionally, we considered structures with a small
helical normalized radius R/r, which makes the contribution
differently. Here, we propose arbitrarily Rd = 18r100 (when
R from the neighboring points much stronger. Experimental
r = 18 in graph with β = 100) to observe the different results in [20] are obtained with superparamagnetic HSs of
positions that the same distance represents in all graphs. For
regular composition and the same aspect ratio β and helical
the other color-map graphs, that distance represents other
normalized radius R/r. It is demonstrated that for θ = 52.5◦ ,
value in the scale Rr e.g. Rr ≈ 14.3 for graph with β = 50
the swimmer structures achieve a better dimensionless speed
and Rr ≈ 10.5 for graph with β = 20 (Fig. 3 magenta line).
U/L f . Hence, our simulation results are in good agreement
Rd can serve to establish a criterion for bounding the helical
according to such experimental data. Regarding the torque,
radius of HSs. Under that condition, HSs with β = 100 can
its value increases as Rr increases (cf. Fig. 5(c)). On the
achieve a greater swimming speeds U compared to HSs with
other hand, regarding the normalized swimming efficiency,
lower aspect ratios. From these results, structures with larger
the maximum efficiency is found for θ = 55◦ and Rr = 2.3
aspect ratios and a helical angle θ ≈ 50◦ experiences a better
(cf. Fig. 5(b)). Keaveny et al. in [12] found the optimum
propulsion efficiency. Fig. 4 represents the swimming speed
swimmer under this criterion for θ = 57.2◦ and Rr = 2.2. The
U with respect the set of variables {β , Rr , θ }. Graphs (a)
slight differences between these two results can be explained
and (c) show a linear behavior of U with respect to the
by the fact that in [12] authors used BEM-based corrections
normalized helical radius Rr . In addition, from graphs (b) and
for computing the traction and resistances coefficients of
(d) a maximum value of U is observed for θ ≈ 50◦ . In order
an RFT model, which provides a relationship of Uτ where
to have a better approximation, Fig. 5 proposes a zoom on
parallel and normal drag coefficients are intervening. Here,
the region of interest for β = 20. In that figure the propulsion
we compute the kinematic and dynamic parameters of the
efficiency (a), the swimming efficiency (b), the fluidic torque
HS using CFD by integrating the traction through solving
(c), and the behaviour of LUf as function of θ (d) are depicted.
Navier-stokes equations with the boundary conditions using
Regarding the propulsion efficiency (cf. Fig. 5(a) and (d)),
finite elements, thus, dispensing with drag coefficients.
we found a maximum efficiency for θ = 52.5◦ for every
R
r . In previous works [26, 29, 30], analytical models were
implemented using RFT and SBT approximations. There B. Cross-Section Influence
exists a relation that links the propulsion speed with drag Here, we analyze the influence of cross-section parameters
coefficients and the geometrical parameters of the helix a
b and φ on the propulsion efficiency. To do so, we choose
(χ−1) sin(2θc )
given by U = ωR 2[1+(χ−1) sin2 (θc )]
, with χ ≈ 2 using the arbitrarily a HS structure with β = 20, θ = 50◦ and Rr = 3.
slender body approximation. Through this assumption, the Fig. 6(a) represents the influence of the variables { ab , φ }
maximum speed is attained for θc = 12 arccos( χ−1 ◦
χ+1 ) = 35.26 ,
on the propulsion efficiency ( LUf ). The best performance are
which is with our nomenclature equivalent to θ = 54.74 . ◦ found for ab = 4 (the largest value) with φ = 90◦ that has an
Nonetheless, we found a maximum angle for θ ≈ 52.5◦ . The Archimedian pump-like structure with efficiency LUf = 0.187.
difference can be due to the slender-body approximation, The least performant is found for ab = 4 but φ = 0, having
that is not needed with finite element method. Structures are a ribbon like structure with efficiency LUf = 0.078. For
© 2021 EU Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in IEEE Robotics and Automation Letters. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LRA.2021.3095291
a
and (f) depicts the HS morphology as a function of b and
φ.
having a binormal-type structure such as the ones presented structure. The maximum ωR I
3 is generated by helical angles
in [24] at the micrometer scale, or in [31] at the millimeter ◦ ◦
c
θ ranging from [45 , 50 ].
scale could experience an increment of more than 100% in
their propulsion speed if their cross-sections were twisted
in φ = π2 becoming normal-type swimmers. Therefore, the B. Cross-section influence
studied structure with fixed helical parameters (β = 20, For the analysis of the cross-section influence, we choose
θ = 50◦ and Rr = 3) with an elongated cross-section ab = 4 as we did in section III.B, a HS structure with parameters
and φ = 90◦ demonstrates the best swimming performance. β = 20, θ = 50◦ and Rr = 3. Fig. 8(a-b) represents the
Regarding results of HSs in a limited power source frame- influence of the thrust at pumping, the maximum value
work (Fig. 6(c) and (d)) the best performance is achieved obtained is for ab = 4 (the largest value) and φ = 90◦ with
2
for a HS with a/b ≈ 2.25 and φ ≈ 85◦ ( µLτ U = 0.59) while T = 30.9 × 10−15 N. The least performant one is achieved
the less performant as in the previous case is for a HS with
2
by the structure whose parameters are ba = 4 and φ = 0
the larger a/b = 4 and φ = 0◦ ( µLτ U = 0.32). Such novel with T = 12.1 × 10−15 N. The best structure outperforms
result would impact future designs of future HS. Fig. 6(e) in 156% the least performant. Moreover, a structure with
© 2021 EU Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in IEEE Robotics and Automation Letters. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LRA.2021.3095291
QUISPE et al.: ENHANCING SWIMMING AND PUMPING PERFORMANCE OF HELICAL SWIMMERS AT LOW REYNOLDS NUMBERS 7
© 2021 EU Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
This article has been accepted for publication in IEEE Robotics and Automation Letters. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/LRA.2021.3095291
Figure 9: Optimization of swimming and pumping at the same time. The solutions are depicted for α = 0 (just pumping), α = 0.50
(half relevance on swimming and half on pumping), and α = 1.00 (just swimming).
© 2021 EU Copyright. Personal use is permitted. For any other purposes, permission must be obtained from the IEEE by emailing pubs-permissions@ieee.org.
View publication stats