1. Introduction
The motion of dispersed particles in shear flow can have very different trajectories. A well-known result is Jeffery orbits, which refers to the closed orbits of rotating ellipsoids in linear shear flows. When inertia of both the fluid and the particle are negligible, the Jeffery orbits can be described by the ellipsoid’s aspect ratio
$\lambda$
, ambient shear rate
$\dot {\gamma }$
, polar angle
$\theta$
and azimuthal angle
$\phi$
(Jeffery Reference Jeffery1922; Leal Reference Leal1980; Voth & Soldati Reference Voth and Soldati2017)
\begin{equation} \tan \theta =\frac {C\lambda }{\left (\sin ^2\phi +\lambda ^2\cos ^2\phi \right )^{1/2}},\quad \tan \phi =\lambda \tan \left (\frac {\dot {\gamma }t}{\lambda +\lambda ^{-1}}\right )\!, \end{equation}
where
$C$
is the orbit parameter and
$t$
is the time, and
$\theta$
and
$\phi$
are defined according to the particle orientation
$\boldsymbol {d}$
in the coordinate system (figure 1
$a$
). In principle, the rotation of ellipsoids can consist of infinite stable orbits, enveloped between a tumbling orbit in the plane perpendicular to the vorticity (
$C = \infty$
) and a spinning orbit parallels to the vorticity (
$C = 0$
). However, completely inertia-free systems are rare in nature. Jeffery (Reference Jeffery1922) hypothesised that a weak inertial effect would make the particles eventually rotate around the local vorticity from the initial position. The hypothesis was soon verified by the experiments of Taylor (Reference Taylor1923). The particles will gradually rotate from their initial orbit to a stable orbit, a process known as orbit drift. Inertia-induced orbit drift has also been verified and investigated by theoretical analysis (Subramanian & Koch Reference Subramanian and Koch2005, Reference Subramanian and Koch2006; Guazzelli & Pouliquen Reference Guazzelli and Pouliquen2018), experiments (Di Giusto et al. Reference Di Giusto, Bergougnoux, Marchioli and Guazzelli2024) and simulations (Feng & Joseph Reference Feng and Joseph1995; Huang et al. Reference Huang, Yang, Krafczyk and Lu2012). In particular, prolate particles drift towards a tumbling orbit over a few Jeffery rotations, and the increase of the orbital parameter
$C$
with time is accompanied by oscillations (Mao & Alexeev Reference Mao and Alexeev2014).
The orbit drift of particles in shear flow is closely related to the aspect ratio of their geometry. For a spheroid, the aspect ratio
$\lambda$
is normally defined as the ratio between the half-length of the symmetry axis and the radius of the spheroid equator. The effect of
$\lambda$
on the orbit drift of spheroids has been investigated by theoretical analysis under the assumption of weak inertia (Mao & Alexeev Reference Mao and Alexeev2014; Einarsson et al. Reference Einarsson, Candelier, Lundell, Angilella and Mehlig2015; Dabade, Marath & Subramanian Reference Dabade, Marath and Subramanian2016; Marath & Subramanian Reference Marath and Subramanian2017). Dabade et al. (Reference Dabade, Marath and Subramanian2016) reported that prolate spheroids (
$\lambda \gt 1$
) would drift towards the tumbling orbit (
$C = \infty$
), while oblate spheroids display a bifurcation in orbit drift at a critical aspect ratio
$\lambda ^*$
(
${\approx} 0.14$
). Specifically, they found that oblate spheroids with
$\lambda ^*\lt \lambda \lt 1$
do not have a stable tumbling orbit but a stable spinning orbit, and for those with
$\lambda \lt \lambda ^*$
, the eventual orbit depends on the initial polar angle. For other axisymmetric particles, Karnis, Goldsmith & Mason (Reference Karnis, Goldsmith and Mason1963) found experimentally that fibres (
$\lambda \gg 1$
) and disks (
$\lambda \ll 1$
) suspended in circular tubes would eventually rotate in the tumbling and spinning orbits, respectively. To quantitatively describe the orbit dynamics of axisymmetric particles in general, the solution of Jeffery orbit was extended by introducing an effective aspect ratio
$\lambda _e$
(Trevelyan & Mason Reference Trevelyan and Mason1951), e.g.
$\lambda _e=1.24\lambda /\sqrt {\textrm {log}}\lambda$
for slender cylinders (Cox Reference Cox1971),
$\lambda _e=1.14\lambda ^{0.844}$
for rod-like particles (Harris & Pittman Reference Harris and Pittman1975) and
$\lambda _e=1.12\lambda ^{0.75}$
for cylindrical disks (Singh et al. Reference Singh, Koch, Subramanian and Stroock2014). Furthermore, rigid particles can have chaotic orbits when they are non-axisymmetric (Yarin, Gottlieb & Roisman Reference Yarin, Gottlieb and Roisman1997; Thorp & Lister Reference Thorp and Lister2019). However, deformable particles may also rotate in shear flow, such as flexible fibres (Slowicka, Stone & Ekiel-Jezewska Reference Slowicka, Stone and Ekiel-Jezewska2020; Bonacci et al. Reference Bonacci, Chakrabarti, Saintillan, du Roure and Lindner2023) and red blood cells (Mauer et al. Reference Mauer, Mendez, Lanotte, Nicoud, Abkarian, Gompper and Fedosov2018). Zhang & Graham (Reference Zhang and Graham2020) numerically observed that deformable prolate capsules in shear flow might have multiple stable orbits, especially for the capsules with large bending stiffness.
Oriented compound drops such as Janus drops can exhibit complex dynamic behaviours in shear flow. Due to their unique structure and inherent anisotropy, Janus drops display dual behaviour: they rotate like rigid ellipsoids while simultaneously deforming like conventional drops. This combination, further complicated by their fore–aft asymmetry, leads to highly complicated dynamics. As a result, Janus drops have significant potential for diverse applications such as colour displayers (Goodling et al. Reference Goodling, Nagelberg, Kaehr, Meredith, Cheon, Saunders, Kolle and Zarzar2019), drug delivery (Song et al. Reference Song, Chao, Zhang and Shum2021) and biomedicine (Hao et al. Reference Hao, Du, He, Yang, Wang, Liu and Wang2022). For Janus drops with ‘ideal’ shape (spherical with a flat, inner interface separating the two drops of equal size), Díaz-Maldonado & Córdova-Figueroa (Reference Díaz-Maldonado and Córdova-Figueroa2015) theoretically demonstrated that the drops could behave as pure drops or as solid particles with broken fore and aft symmetry. Experiments have also been conducted to study dynamical differences between Janus drops and pure drops, including drop generation via microfluidic technology (Guo et al. Reference Guo, Zeng, Wang and Zhang2017; Zhu et al. Reference Zhu, Chen, Zhu, Huang, Mu, Si and Xu2023), drop breakup in bifurcating microchannels (Wang et al. Reference Wang, Wang, Mu, Han and Cheng2024) and drop impact on substrates (Yun, Kim & Kim Reference Yun, Kim and Kim2022; Hwang & Yun Reference Hwang and Yun2023; Jung & Yun Reference Jung and Yun2023 ). In addition, Torza & Mason (Reference Torza and Mason1970) experimentally observed the rotation and deformation of Janus drops in shear flow. In these experiments, Janus drops are asymmetrical due to differences in surface tension, viscosity or volume between their two component drops. By contrast, numerical investigation of Janus drops in shear mainly focused on the periodic rotation and deformation of symmetric two-dimensional (2-D) drops (Liu & Park Reference Liu and Park2022; Zhang et al. Reference Zhang, Chen, Qian and Ding2023) and engulfing behaviour of three-dimensional (3-D) drops (Liu & Park Reference Liu and Park2023). Among them, Zhang et al. (Reference Zhang, Chen, Qian and Ding2023) proposed a model based on Jeffery’s theory to predict the rotation periods, by introducing a corrected shear rate and an effective aspect ratio for drop deformation. However, it remains unknown whether 3-D oriented compound drops encounter orbit drift similar to 3-D rigid ellipsoids, and if so, how the drop deformation and the initial polar angle would affect their orbit dynamics.
In this study, we investigate the orbit dynamics of 3-D symmetric Janus drops in simple shear flow using an improved ternary-fluid phase field model. To simplify the flow problem, we restrict our study to the Janus drop consisting of two component drops of equal size, matched density and viscosity. We focus on two equilibrium configurations: near-spherical (similar to the ’ideal’ shape and characterised by significantly lower surface tension coefficient at the inner interface compared with the outer interfaces) and dumbbell-shaped (where surface tension coefficients are comparable), to evaluate the geometric effect of Janus drops on the orbit dynamics. With the help of numerical simulations, we identify the flow regimes in the range of flow parameters considered, measure the instantaneous deformation and orientation of the Janus drops in shear flow, investigate the effect of the initial polar angle and deformation of the Janus drops on their orbit dynamics, and analyse the relevant mechanisms. In particular, we propose an effective aspect ratio model and calculate the orbit parameter for dumbbell-shaped Janus drops, to quantitatively evaluate the effect of drop deformation on the orbit drift. Finally, we make predictions of the orbit dynamics of Janus drops, following the spirit of Jeffery’s theory for ellipsoidal particles.
(
$a$
) A sketch of a 3-D Janus drop in simple shear flow. The vector
$\boldsymbol{d}$
denotes the orientation of the Janus drop, which can be represented by the polar angle
$\theta$
and azimuthal angle
$\phi$
in a spherical coordinate. (
$b$
) Geometry measurement of drop deformation, including major semi-axis
$L$
, and two radii of the inner interface
$R_a$
and
$R_b$
.

2. Problem statement and methodology
2.1. Problem statement
We consider 3-D numerical simulations of a neutrally buoyant Janus drop in shear flow. A sketch of the Janus drop considered and its geometry parameters are shown in figure 1(
$a$
,
$b$
), in which the Janus drop consists of two immiscible component drops
$1$
and
$2$
of equal volume
$2\pi R^3/3$
, where
$R$
is the effective radius of the component drop. The component drops (of fluids
$1$
and
$2$
) and the surrounding fluid
$3$
have matched density (
$\rho$
) and viscosity (
$\mu$
). The surface tension coefficient at the interface between fluids
$i$
and
$j$
is represented by
$\sigma _{ij}$
. To avoid the lateral migration of Janus drops in the shear flow owing to geometric asymmetry, we consider symmetric Janus drops, particularly with
$\sigma _{23}=\sigma _{13}$
. Consequently, the shape of the Janus drop at equilibrium is determined by the ratio
$\sigma _{12}/\sigma _{13}$
, and details can be found in Appendix A. Two kinds of Janus drops at equilibrium are considered in the present study, namely dumbbell-shaped (with
$\sigma _{12}\sim \sigma _{13}$
; specifically,
$0.347\leqslant \sigma _{12}/\sigma _{13}\leqslant 1.532$
) and near-spherical (with
$\sigma _{12}\ll \sigma _{13}$
).
Initially, the Janus drop is placed at the centre of a domain of
$(-3R, 3R)^3$
, with an undisturbed flow field
$\boldsymbol {U}_0=(\dot {\gamma }y, 0, 0)$
. To maintain the shear flow, two moving plates are placed at the upper and lower boundaries at a speed of
$3\dot {\gamma }R$
but in the opposite direction. Periodic boundary conditions are used in the other two directions. The dynamics of Janus drops is determined by capillary number
$ \textit{Ca} =\mu \dot {\gamma }R/\sigma _{13}$
and Reynolds number
$Re=\rho \dot {\gamma }R^2/\mu$
. The orientation of the Janus drop is described by a vector
$\boldsymbol {d}$
, which points from the centre of mass of component drop
$1$
to that of component drop
$2$
. The angle between
$\boldsymbol {d}$
and the
$z$
-axis is defined as the polar angle
$\theta$
, and the azimuthal angle
$\phi$
is defined as the angle between the
$y$
-axis and the projection of
$\boldsymbol {d}$
on the
$x$
–
$y$
plane. In the present study, we consider relatively small Janus drops, and specifically,
$0.2\leqslant Re\leqslant 1.5$
and
$0.02\leqslant Ca\leqslant 0.2$
. In this range of parameters, the flow have weak inertia and Janus drops do not encounter breakup. The initial orientation of Janus drops, denoted by
$\theta _0$
, ranges from
$0$
to
$\pi /2$
.
2.2. Numerical methods
We employ an improved phase field model for ternary fluids to track the interface evolution (Olsson, Kreiss & Zahedi Reference Olsson, Kreiss and Zahedi2007; Abadi, Rahimian & Fakhari Reference Abadi, Rahimian and Fakhari2018; Chen et al. Reference Chen, Zhang, Liu and Ding2025). The interfaces are modelled by diffuse interfaces with finite thickness and represented by volume fractions of the fluids,
$\boldsymbol {A}$
$=(\alpha _1, \alpha _2, \alpha _3)$
, where
$\alpha _i$
denotes the volume fraction of the
$i$
th fluid, with the constraint
$\alpha _1+\alpha _2+\alpha _3=1$
. Time evolution of
$\alpha _i$
is governed by the following equation:
\begin{equation} \frac {\partial \alpha _i} {\partial t}+\boldsymbol{\nabla }\boldsymbol{\cdot }(\alpha _i \boldsymbol {u})=\frac {1}{\textit{Pe}}\boldsymbol{\nabla }\boldsymbol{\cdot }\left (\boldsymbol{\nabla }\alpha _i -\frac {\alpha _i(1-\alpha _i)}{\sqrt {2}Cn}\frac {\boldsymbol{\nabla }\alpha _i}{|\boldsymbol{\nabla }\alpha _i|}+\frac {1}{3}\sum _{j=1}^3\frac {\alpha _j(1-\alpha _j)}{\sqrt {2}Cn}\frac {\boldsymbol{\nabla }\alpha _j}{|\boldsymbol{\nabla }\alpha _j|}\right )\!, \end{equation}
where
$\boldsymbol {u}$
is the flow velocity. The Cahn number
$Cn$
represents a dimensionless measure of the thickness of the interfaces (Ding, Spelt & Shu Reference Ding, Spelt and Shu2007). In the present study, the Cahn number is set to
$Cn=0.5h/R$
, where
$h$
is the mesh size. It has been shown that the Péclet number
$ \textit{Pe} $
should obey
$ \textit{Pe} \sim Cn^{-2}$
to ensure the sharp interface limit of the phase field model (Chen et al. Reference Chen, Zhang, Liu and Ding2025). Furthermore, to maintain the interface profile and avoid unphysical changes of interfaces, the Péclet number
$ \textit{Pe} $
is set to
$ \textit{Pe} = 0.007Cn^{-2}$
following the recommendation of Chen et al. (Reference Chen, Zhang, Liu and Ding2025). The motion of ternary fluid flows is governed by the dimensionless continuity equation and Navier–Stokes equations,
where
$p$
is the pressure. The interfacial tension
${\boldsymbol {f}}_s$
is calculated by
\begin{equation} {\boldsymbol {f}}_s= - \sum _{j=1}^{3}\gamma _j \boldsymbol{\nabla }\boldsymbol{\cdot }\left (\frac {\boldsymbol{\nabla }\alpha _j}{|\boldsymbol{\nabla }\alpha _j|}\right ) \frac {\boldsymbol{\nabla }\alpha _j}{|\boldsymbol{\nabla }\alpha _j|}\delta _j, \end{equation}
where
$\delta _j = 6\alpha _j (1-\alpha _j)|\boldsymbol{\nabla }\alpha _j|$
is a smoothed delta function (Chen et al. Reference Chen, Zhang, Liu and Ding2025);
$\gamma _j$
represents the phase-specific surface tensions and relates to the surface tension coefficients by (Zhang et al. Reference Zhang, Ding, Gao and Wu2016)
\begin{equation} \begin{cases} \gamma _1=(\sigma _{12}-\sigma _{23}+\sigma _{13})/(2\sigma _{13}),\\[2pt] \gamma _2=(\sigma _{12}+\sigma _{23}-\sigma _{13})/(2\sigma _{13}),\\[2pt] \gamma _3=(-\sigma _{12}+\sigma _{23}+\sigma _{13})/(2\sigma _{13}). \end{cases} \end{equation}
Details about numerical discretisation and solution procedure are available from Chen et al. (Reference Chen, Zhang, Liu and Ding2025).
2.3. Validation and convergence study
We validate our simulations by comparing against experimental data from Torza & Mason (Reference Torza and Mason1970), where a Janus drop consisting of component droplets
$1$
(distilled water and 10 vol % Tween 20) and
$2$
(castor oil) is suspended in silicone oil (ambient fluid;
$\rho _3=0.979\,{\rm kg\,m}^{- 3}$
,
$\mu _3=6.4$
Pa s) under shear (
$\dot {\gamma }=0.106$
s
$^{-1}$
). Detailed flow conditions are listed in table 1. Due to inconsistencies between the reported volume ratio (1.125) and the experimental equilibrium drop image of Torza & Mason (Reference Torza and Mason1970), we set the component 1-to-2 volume ratio to 0.728 in our simulations. Surface tension coefficients were also slightly adjusted (relative error < 3.6 %) to match the experimental equilibrium shape (figure 2
$a$
). Although the flow in the experiment is in the Stokes regime (
$Re\approx 0$
), we used
$Re=0.1$
for computational efficiency, which is expected to have negligible effect on the results. In the simulation, the grid size and time step are set to
$h = R/40$
and
$\varDelta t=0.0002{\dot {\gamma }}^{-1}$
, respectively. Figures 2(
$b$
) and 2(
$c$
) show the comparison between numerical results and experimental data, in terms of drop snapshots at elongation and squeeze, respectively. We can see that our numerical simulation can accurately capture the deformation and rotation of Janus drops in experiments. Furthermore, the rotation period of Janus drop is
$14.47{\dot {\gamma }}^{-1}$
in the simulation, rather close to the experimental result of
$14.20{\dot {\gamma }}^{-1}$
.
Set-up of experiments (Torza & Mason Reference Torza and Mason1970) and simulation for validation.

Snapshots of experimental observations of Janus drop in shear flow (Torza & Mason Reference Torza and Mason1970) superimposed by numerical results in the
$z=0$
plane (dashed lines): (
$a$
) drop at equilibrium; (
$b$
) elongation and (
$c$
) squeeze.

To assess the convergence of the numerical results, we compared the drop shape at the moment of most elongation at
$Re=0.5$
,
$ \textit{Ca} =0.2$
,
$\sigma _{12}=\sigma _{13}$
and
$\theta _0=\pi /2$
on progressively refined meshes, with mesh sizes of
$h = R/32$
,
$R/40$
and
$R/64$
. Note that
$ \textit{Ca} = 0.2$
corresponds to the highest capillary number in this study and, thus, is expected to cause significant drop deformation. Figure 3(
$a$
) presents the interfaces of the Janus drop in the
$x$
–
$y$
plane. It is clear that the profiles of the interfaces converge with the mesh refinement. In the simulations hereafter,
$h = R/40$
and the time step
$\varDelta t=0.0002{\dot {\gamma }}^{-1}$
are adopted, and satisfy the computational stability when solving the equation for interface evolution, i.e.
$\dot {\gamma }\Delta t\lt (h/R)^2$
. Moreover, to assess the influence of domain boundaries, we perform two simulations at
$\theta _0 = \pi /2$
,
$Re = 0.5$
and
$ \textit{Ca} = 0.2$
, with the wall being positioned at
$y=\pm 4.5R$
and
$\pm 6R$
, respectively. Figure 3(
$b$
) shows that the drop shapes for three different wall positions are virtually overlapped; this demonstrates that the wall effect can be neglected when the distance from the wall to the droplet’s centre of mass exceeds
$3R$
.
(
$a$
) Convergence study of the drop shape at the moment of most elongation at
$\theta _0=\pi /2$
,
$Re$
= 0.5 and
$ \textit{Ca} $
= 0.2:
$h/R = 1/32$
(dash-dotted), 1/40 (dashed) and 1/64 (solid). (
$b$
) Comparison of drop shape at the moment of most elongation at
$\theta _0=\pi /2$
,
$Re$
= 0.5 and
$ \textit{Ca} $
= 0.2, with the wall placed at
$y=\pm 3R$
(dashed),
$\pm 4.5R$
(dash-dotted) and
$\pm 6R$
(solid).

3. Results and discussion
Unless stated otherwise, dumbbell-shaped Janus drop refers to the case of
$\sigma _{12}=\sigma _{13}$
and near-spherical to
$\sigma _{12}=0.0175\sigma _{13}$
.
3.1. Flow regimes: tumbling versus spinning
Janus drops in shear flow exhibit two distinct flow regimes: tumbling and spinning. In the tumbling regime, the Janus drop continuously rotates around the direction of the vorticity of the shear flow, accompanied by periodic deformation of interfaces. By contrast, the Janus drop in the spinning regime experiences stable deformation, surrounded by steady flows and with the inner interface in the plane perpendicular to the vorticity (more precisely, the plane at
$z=0$
). Figure 4 illustrates the flow regimes for the dumbbell-shaped Janus drops at
$ \textit{Ca} =0.2$
,
$Re=0.5$
and different initial orientations
$\theta _0$
, with respect to interface shapes and velocity vectors. For
$\theta _0=\pi /2$
, the drop falls into the tumbling regime, with a rotation period
$t_R\approx 23.0{\dot {\gamma }}^{-1}$
. Specifically, during the rotation, the outer interfaces (separating the drop from fluid 3) of the drop experience pronounced periodic deformation, while the inner interface (between the two component droplets) remains more or less flat. For
$\theta _0=0$
, the drop quickly evolves into the spinning regime (and maintains the state until the termination of computation at
$\dot {\gamma }t=200$
); the flat inner interface is deformed and located in the symmetry plane, with a vortex enclosed in it.
Dumbbell-shaped Janus drops in shear flow at
$ \textit{Ca} =0.2$
and
$Re=0.5$
in terms of drop snapshots and velocity vectors: (
$a$
) tumbling regime with
$\theta _0=\pi /2$
, at times
$\dot {\gamma }t=$
3.2, 10.8, 14.7, 22.3 from left to right and from top to bottom; (
$b$
) spinning regime with
$\theta _0=0$
at time
$\dot {\gamma }t=$
15.0, where the velocity vectors in the
$x{-}y$
plane at
$z = 0$
are shown in the top panel and streamlines in fluid 2 in the bottom panel.

Figure 5 depicts the flow around and the deformation of near-spherical Janus drops at
$ \textit{Ca} =0.02$
,
$Re=0.5$
and different
$\theta _0$
. Similar to the dumbbell-shaped Janus drops, the near-spherical drop is tumbling at
$\theta _0=\pi /2$
, while spinning at
$\theta _0=0$
. However, the deformation behaviours of the drops are very different. For example, in both regimes, the deformation of the outer interfaces resembles that of pure drops in shear flow: nearly ellipsoidal deformation under relatively weak shear (
$ \textit{Ca} =0.02$
) and surrounded by steady flows. During tumbling (figure 5
$a$
), the inner interface rotates periodically owing to the presence of internal vortices in the Janus drop. Because for near-spherical Janus drops the surface tension coefficient of the inner interface is much lower than the outer ones, (here
$\sigma _{12}/\sigma _{13}=0.0175$
), significant deformation (or even breakup) of the inner interface may occur at low
$ \textit{Ca} $
. In contrast, the spinning regime maintains a flat inner interface, which is perpendicular to the vorticity direction. Therefore, the assumption of a flat inner interface in the theoretical analysis of dynamics of spherical Janus drops (Díaz-Maldonado & Córdova-Figueroa Reference Díaz-Maldonado and Córdova-Figueroa2015) is valid for the spinning regime, but questionable for the tumbling regime.
Dynamics of (a) tumbling and (b) spinning of near-spherical Janus drops in shear flow at
$ \textit{Ca} =0.02$
and
$Re=0.5$
, in terms of the snapshots of interface evolution and the projected area of velocity vectors and interface shapes. In panel (a),
$\theta _0=\pi /2$
and the snapshots from left to right and from top to bottom correspond to
$\dot {\gamma }t=$
3.3, 6.4, 9.8, 12.9, respectively. In panel (b),
$\theta _0=0$
and the snapshot corresponds to the drop shape at time
$\dot {\gamma }t=$
15.0, and the velocity vectors in the
$x{-}y$
plane at
$z = 0$
are shown in the top panel, while streamlines in fluid 2 in the bottom panel. In the top panel of panel (b), the interface represents the triple contact line where all three fluids meet.

Dynamics of a 3-D tumbling Janus drop in shear flow at
$\theta _0=\pi /6$
,
$ \textit{Ca} =0.2$
,
$Re=0.5$
and
$\sigma _{12}=\sigma _{13}$
, in terms of the snapshots of (
$a$
) interface evolution, (
$b$
) temporal evolution of
$\theta$
and
$\phi$
, and (
$c$
) the major semi-axis
$L$
and
$\sqrt {R_aR_b}$
. The symbols correspond to
$\dot {\gamma }t =$
4.3 (
), 8.0 (
), 56.0 (
), 59.6 (
), 67.5(
), 71.1 (
).

3.2. Effect of initial orientation
As shown in the previous section for
$\theta _0=0$
and
$\pi /2$
, the initial orientation of Janus drops significantly influences the resulting flow regimes. In this section, we examine how the initial orientation
$\theta _0$
affects the trajectory of Janus drops for other angles. Figure 6 presents the dynamics of a dumbbell-shaped Janus drop at
$\theta _0=\pi /6$
,
$ \textit{Ca} =0.2$
and
$Re=0.5$
, in terms of drop shapes and time variations of
$\theta$
,
$\phi$
, and the geometric parameters of the drop. In particular, figure 6(
$a$
) shows the snapshots of the Janus drop evolving from its initial state to the tumbling regime. Figure 6(
$b$
) quantifies the process through the temporal evolution of the orientation angles
$\theta$
and
$\phi$
. It is clear that
$\theta$
increases with time in an oscillatory manner (
$0\leqslant \dot {\gamma }t\lt 50$
) until the drop enters the tumbling regime (
$\dot {\gamma }t\gt 50$
), i.e. the orbit of
$\theta =\pi /2$
in figure 4(
$a$
). The oscillation of
$\theta$
is related to the drop rotation, of which the period may vary with time and can be defined by the change of
$\phi$
by
$2\pi$
. Accordingly, the period of the
$k\mathrm{th}$
rotation represents the duration of
$2(k-1)\pi \leqslant \phi \leqslant 2k\pi$
. Concurrently with the
$\theta$
-increase, we observe enhanced drop deformation and an expanding projected area in the
$x{-}y$
plane, as evidenced by the variation of
$L$
and
$\sqrt {R_aR_b}$
in figure 6(
$c$
). Figure 6(
$c$
) also indicates that the maxima in
$L$
coincide with the minima in
$\sqrt {R_aR_b}$
, and vice versa.
(
$a$
) Dynamics of near-spherical Janus drops in shear flow at
$ \textit{Ca} =0.05$
,
$Re=1$
and
$\theta _0=$
$\pi /6$
. The snapshots from left to right correspond to the times
$\dot {\gamma }t=$
52.5, 78.5, 104.5, 130.0, 156.0, 182.0 and 208.0. The dashed line indicates the asymptotic approach of
$\theta$
to zero by data fitting of
$\theta = 0.5 \exp (-0.0072{\dot {\gamma }}t)$
. (
$b$
) Time evolution of the deformation parameter
$D_T =(\varGamma _a-\varGamma _b)/(\varGamma _a+\varGamma _b)$
.

Compared with dumbbell-shaped Janus drops, the initial orientation exerts a fundamentally different influence on the near-spherical Janus drops. Figure 7 shows the orbit evolution of a near-spherical Janus drop in shear flow at
$\theta _0=\pi /6$
,
$ \textit{Ca} =0.05$
and
$Re=1$
with respect to the temporal variation of
$\theta$
. We note that the global deformation of drop stabilises after the startup stage (approximately at
$\dot {\gamma }t= 12.0$
) (figure 7
$b$
), with the outer interfaces conforming to the streamlines in the surrounding shear flow. The temporal evolution of
$\theta$
exhibits an oscillatory decay, gradually decreasing over time with diminishing oscillation amplitude. Although
$\theta$
does not reach zero within the extended computation period, its trajectory strongly suggests an asymptotic approach to zero (
$\theta = 0.5\exp (-0.0072{\dot {\gamma }}t)$
by data fitting, see figure 7
$a$
). This implies that the drop will ultimately enter the spinning regime. More dynamic evolution of orbit drift are provided in the supplementary movies available at https://doi.org/10.1017/jfm.2025.10988.
3.3. Orbit drift: mechanisms and dynamics
The bifurcation in orbit drift originates from fundamental differences in deformation dynamics and shape-dependent hydrodynamics between the dumbbell-shaped and near-spherical Janus drops. Specifically, dumbbell-shaped Janus drops exhibit orbit drift dominated by the coupling between the outer interface deformation and the surrounding flow field, steering them towards
$\theta =\pi /2$
. In contrast, near-spherical drops show orbit drift governed by the interplay between the inner interface deformation and the vortex enclosed within the drop, directing them towards
$\theta =0$
. For the dumbbell-shaped cases, the rotation dynamics of drops resembles that of ellipsoids with various aspect ratios in shear flow. The shear-induced global deformation effectively increases the drop aspect ratio (see more details in § 3.4), driving it to higher orbit angles (
$\theta$
). This elevation further intensifies shear and deformation – a feedback mechanism causing orbit drift towards the tumbling regime. It is noteworthy that in the present study, the tumbling regime occurs for
$\theta _0\ne 0$
(specifically,
$\pi /36 \leqslant \theta _0\leqslant \pi /2$
), suggesting that the spinning regime at
$\theta _0=0$
is unstable. For the near-spherical cases, both the surrounding flow field and the shape of the outer interface enter a steady state during the orbit drift, and the rotation of the inner interface is primarily driven by the hydrodynamic torque from the vortex enclosed within the drop (see figure 5
$a$
). Moreover, the asymmetry of the inner interface about the
$z=0$
plane leads to an unbalanced torque, thereby inducing a slow orbit drift towards
$\theta =0$
(where the hydrodynamic torque vanishes). Based on the above-mentioned analysis, the tumbling regime at
$\theta _0=\pi /2$
is essentially unstable for near-spherical Janus drops, since any perturbation on the inner interface would trigger the orbit drift towards the spinning regime. Figure 8 shows the orbit drift of Janus drops in terms of
$\theta$
versus time at
$Re=1$
,
$ \textit{Ca} =0.05$
,
$\theta _0=\pi /6$
and different values of
$\sigma _{12}/\sigma _{13}$
. We can see that
$\theta$
increases very slowly with time for
$\sigma _{12}/\sigma _{13}=0.174$
, while decreases slowly for
$\sigma _{12}/\sigma _{13}=0.140$
. Compared with the cases with extreme surface tension ratios such as
$\sigma _{12}/\sigma _{13}=1$
(dominated by outer interface deformation) and
$\sigma _{12}/\sigma _{13}=0.0175$
(dominated by inner interface deformation), the combined effect of the two mechanisms in these intermediate cases yields significantly slower orbit drift.
Orbit drift of Janus drops in terms of
$\theta$
versus time at
$Re=1$
,
$ \textit{Ca} =0.05$
,
$\theta _0=\pi /6$
and different values of
$\sigma _{12}/\sigma _{13}$
. The drop snapshots in the middle of the figure correspond to the drop shapes at equilibrium, with decreasing value of
$\sigma _{12}/\sigma _{13}$
from left to right.

Orbit drift of Janus drops in the
$x$
–
$y$
plane at
$Re=0.5$
and
$\theta _0=\pi /6$
. Dumbbell-shaped Janus drops with (
$a$
)
$ \textit{Ca} =0.05$
and (
$b$
)
$ \textit{Ca} =0.15$
. (
$c$
) Near-spherical Janus drop with
$ \textit{Ca} =0.05$
and
$Re=1$
. Filled circles indicate the initial state of the drop.

The orbit of a Janus drop can be quantitatively described by the projection of the unit vector of
$\boldsymbol {d}$
onto the
$x$
–
$y$
plane, i.e. (
$\sin \theta \sin \phi , \sin \theta \cos \phi$
). Figure 9 shows the orbit evolution of Janus drops at
$\theta _0=\pi /6$
. Given initially
$\theta =\pi /6$
and
$\phi =0$
, the starting point of the orbit is located at
$(0, 0.5)$
. For the dumbbell-shaped Janus drops with
$Re=0.5$
, the orbit then gradually migrates from the lower polar angle to the higher one, ultimately aligning with that of
$\theta =\pi /2$
when reaching the tumbling regime. Furthermore, the drop at relatively high
$ \textit{Ca} $
(
$=0.15$
) requires less rotation periods to reach the tumbling regime than at relatively low
$ \textit{Ca} $
(
$=0.05$
). The orbit evolution of the near-spherical Janus drop with
$ \textit{Ca} =0.05$
and
$Re=1$
is in sharp contrast to that of dumbbell-shaped ones: it gradually drifts towards (0, 0). Therefore, it is reasonable to expect that the deformation history of the Janus drop can have a significant effect on the orbit drift.
Temporal evolution of
$\theta$
for dumbbell-shaped Janus drops: (
$a$
,
$b$
)
$Re=0.5$
and
$ \textit{Ca} =0.05$
, (
$c$
,
$d$
)
$Re=0.5$
and
$ \textit{Ca} =0.15$
, (
$e$
,
$f$
)
$Re=1.5$
and
$ \textit{Ca} =0.15$
. Note that panels (
$b$
), (
$d$
) and (
$f$
) correspond to the results with certain time offsets of panels (
$a$
), (
$c$
) and (
$e$
), respectively. Specifically, for
$\theta _0=\pi /12$
,
$\pi /4$
and
$\pi /3$
, the time offsets are
$t_0=0$
,
$76.2$
,
$106.8$
and
$129.7$
in panel (b),
$t_0=0$
,
$64.4$
,
$80.8$
and
$80.8$
and
$89.2$
in panel (
$d$
), and
$t_0=0$
,
$38.5$
,
$69.3$
and
$92.4$
in panel (
$f$
), respectively.

Figure 10 shows the temporal evolution of the polar angle
$\theta$
with various
$ \textit{Ca} $
,
$Re$
and
$\theta _0$
. Several observations can be made. First, the variation of
$\theta$
with time is not a monotonic process, but characterised by periodic oscillations, which is very similar to the orbit drift of a spheroid in shear flow (Mao & Alexeev Reference Mao and Alexeev2014). Second, Janus drops reach the tumbling regime more rapidly with the increasing of
$\theta _0$
. Furthermore, if the curves of various
$\theta _0$
are shifted to align with that of
$\theta _0=\pi /12$
according to the time reaching the tumbling regime, all the shifted curves collapse onto a single one (see figures 10
$b$
, 10
$d$
and 10
$f$
). In other words, the initial polar angle essentially has no effect on the orbit dynamics, while the start-up effect of Janus drops in shear flow at relatively large
$ \textit{Ca} $
might slightly affect the alignment (figure 10
$d$
). However, this is not surprising, considering the flows are essentially Stokes flow. Third, the comparison between figures 10(
$b$
) and 10(
$d$
) shows that drop deformation of Janus drops speeds up the progression of the polar angle towards
$\pi /2$
. At last, the orbit drift speeds up with increasing inertia (from the comparison between figures 10
$c$
and 10
$e$
), similar to ellipsoids in shear flow (Di Giusto et al. Reference Di Giusto, Bergougnoux, Marchioli and Guazzelli2024).
3.4. Effective aspect ratio of dumbbell-shaped Janus drops
To evaluate the deformation of dumbbell-shaped Janus drops in the tumbling regime, we define an aspect ratio
$\lambda =L/\sqrt {R_aR_b}$
, where
$L$
represents the semi-major axis of the Janus drop, and
$R_a$
and
$R_b$
denote the long and short radii of the inner interface of component drops, respectively (see the definition in figure 1
$b$
). For the convenience of discussion, we fix
$\theta _0=\pi /6$
in the following studies. The aspect ratio of a Janus drop at rest,
$\lambda _0$
, is determined by
$\sigma _{12}$
and
$\sigma _{13}$
in the present set-up, and can be expressed as (see Appendix A)
\begin{equation} \lambda _0=\sqrt {\frac {2\sigma _{13}+\sigma _{12}}{2\sigma _{13}-\sigma _{12}}}. \end{equation}
It is straightforward to obtain
$\lambda _0=\sqrt {3}$
for
$\sigma _{12}=\sigma _{13}$
. Figure 11(
$a$
) shows the typical evolution of
$\lambda$
in the tumbling regime at
$Re=0.2$
,
$\theta _0=\pi /6$
and various
$ \textit{Ca} $
. The variation of
$\lambda$
is periodic with time, and characterised by its maximum and minimum values, denoted by
$\lambda _{\textit{max}}$
and
$\lambda _{\textit{min}}$
, respectively. Moreover, the drop experiences linear deformation at small
$ \textit{Ca} $
(
$=0.02$
) such that
$\lambda _0\approx (\lambda _{\textit{max}} + \lambda _{\textit{min}})/2$
, while encounters nonlinear deformation at relatively large
$ \textit{Ca} $
(
$=0.15$
), where
$\lambda _{\textit{max}}-\lambda _0\gg \lambda _0-\lambda _{\textit{min}}$
. This implies that the drop deformation is dominated by elongation and thus can be represented by
$\lambda _{\textit{max}}$
. The dimensionless rotating period,
$\dot {\gamma }t_R$
, is related to the drop shape at equilibrium and the drop deformation, as depicted in figures 11(
$b$
) and 11(
$c$
). More specifically,
$\dot {\gamma }t_R$
increases with both
$\lambda _0$
and
$\lambda _{\textit{max}}$
, and is not sensitive to
$Re$
at low
$ \textit{Ca} $
(e.g.
$ \textit{Ca} =0.02$
and
$0.05$
).
(
$a$
) Temporal evolution of
$\lambda$
at
$Re=0.2$
and various
$ \textit{Ca} $
. The solid straight line denotes
$\lambda _0$
. (
$b$
)
$t_R$
versus
$\lambda _0$
at
$Re=0.2$
and various
$ \textit{Ca} $
. (
$c$
)
$t_R$
versus
$\lambda _{\textit{max}}$
at various
$ \textit{Ca} $
and
$Re$
. In panels (
$a$
) and (
$c$
),
$\lambda _0=\sqrt {3}$
.

Considering the dynamic similarity between Janus drops and spheroids in shear flow, we tentatively use Jeffery’s theory (Jeffery Reference Jeffery1922) to predict
$t_R$
by introducing an effective aspect ratio
$\lambda _e$
, such that
Similarly, the rotational angular velocity
$\omega$
(
$=\textrm {d}\phi /\textrm {d} \textrm {t}$
) can be estimated by
\begin{equation} \omega =\dot {\gamma }\left (\frac {\sin ^2\phi +\lambda _{e}^2\cos ^2\phi }{1+\lambda _{e}^2}\right )\!. \end{equation}
Note that
$\lambda _e$
should take account of the effect of the drop shape at equilibrium
$\lambda _0$
and drop deformation (characterised by
$\lambda _{\textit{max}}$
), and that its value should lie between
$\lambda _0$
and
$\lambda _{\textit{max}}$
. Therefore, we tentatively propose
where
$0\leqslant \beta \leqslant 1$
is an unknown constant. It is noteworthy that
$\lambda _e$
in (3.4) differs from its counterpart for 2-D Janus drops,
$\lambda ^{\text{2-D}}_e=\lambda _{max }/\lambda _0$
(Zhang et al. Reference Zhang, Chen, Qian and Ding2023). In the 2-D case, it is drop deformation that determines the change in aspect ratio and consequently the rotation period, while the equilibrium shape of the drop has insignificant effect (Zhang et al. Reference Zhang, Chen, Qian and Ding2023).
(
$a$
) Comparison of
$t_R$
between numerical results and model prediction of (3.2) (denoted by the solid line). The cases with different
$\lambda _0$
are the same as in figure 11(
$b$
), including
$ \textit{Ca} =0.02$
(
$\otimes$
) and
$0.05$
(
$\oplus$
) at
$Re=0.2$
. Other cases correspond to Janus drops with
$\sigma _{12}=\sigma _{13}$
, covering various
$Re$
(
$=0.2$
(
), 0.3 (
), 0.5 (
) and 1 (
)), and various
$ \textit{Ca} $
(indicated by colours). For each case, solid symbols regime is as in figure 11(
$c$
), while hollow symbols correspond to
$t_R$
in the orbital drift. (b)–(
$e$
)
$\omega$
versus
$\phi$
at
$Re=0.5$
and various
$ \textit{Ca} $
: (b)
$0.02$
; (
$c$
)
$0.05$
; (
$d$
)
$0.1$
and (
$e$
)
$0.2$
. Symbols denote the numerical results and solid curves denote the prediction of (3.3). The effective aspect ratios from panels (b) to (
$e$
) are
$\lambda _e=$
$1.81$
,
$1.93$
,
$2.21$
and
$3.21$
, respectively.

To determine the value of
$\beta$
and validate the model for
$\lambda _e$
, we compare
$t_R$
obtained from numerical simulations and model predictions based on (3.2), and the results are shown in figure 12(
$a$
). We find that the model with
$\beta =0.6$
accurately predicts
$t_R$
for tumbling Janus drops in the range of parameters considered, which are the same as in figure 11(
$b$
,
$c$
). Furthermore, figure 12(
$b$
–
$e$
) presents the comparison of
$\omega$
versus
$\phi$
between the numerical results at
$Re=0.5$
and the analytical solution of (3.3) using
$\lambda _e$
with
$\beta =0.6$
. At small
$ \textit{Ca} $
(
$=0.02$
and 0.05), the effect of drop deformation on
$\lambda _e$
is insignificant (
$\lambda _e/\lambda _0=1.04$
and
$1.11$
, respectively), and the angular velocity
$\omega$
virtually overlaps with the model prediction. With further increase of
$ \textit{Ca} $
(e.g.
$ \textit{Ca} =0.2$
and
$\lambda _e/\lambda _0=1.85$
), the numerical results deviate from the model prediction, primarily because the drop deformation becomes progressively nonlinear with increasing
$ \textit{Ca} $
(Zhang et al. Reference Zhang, Chen, Qian and Ding2023). It is interesting to note that a higher effective aspect ratio results in a lower maximum value of
$\omega$
, which is opposite to the cases of prolate ellipsoids (Daghooghi & Borazjani Reference Daghooghi and Borazjani2015).
3.5. Effect of deformation on orbit drift of dumbbell-shaped drop
We define a parameter
$D$
to quantitatively measure the drop deformation
Specifically,
$D=0$
corresponds to the Janus drop at equilibrium and
$D_{\textit{max}}=(\lambda _{\textit{max}}-\lambda _0)/(\lambda _{\textit{max}}+\lambda _0)$
. It is easy to obtain from (3.4) and the definition of
$D_{\textit{max}}$
,
Figure 13(
$a$
) indicates
$D_{\textit{max}}\propto Ca$
at relatively small
$ \textit{Ca} $
(
$\leqslant 0.1$
) for Janus drops in the tumbling orbit, consistent with linear deformation of 2-D Janus drops (Zhang et al. Reference Zhang, Chen, Qian and Ding2023) and 3-D pure drops in shear flow (Liu et al. Reference Liu, Ng, Chong, Lohse and Verzicco2021). Figure 13(
$b$
,
$c$
) presents the time evolution of
$\theta$
and
$D$
at
$Re=0.5$
and various
$ \textit{Ca} $
(
$=0.02$
and
$0.2$
, respectively). For the drop with linear deformation (
$ \textit{Ca} =0.02$
), the orbit drift takes place slowly, and the variations of
$\theta$
and
$D$
are apparently coupled in an anti-phase manner: the peak of
$D$
coincides with the trough of
$\theta$
. By contrast, for the drop with nonlinear deformation (
$ \textit{Ca} =0.2$
),
$D_{\textit{max}}$
increases more quickly with time and converges to a constant value within 2–3 rotation periods (or entering the tumbling regime). Owing to the significant deformation, the vector
$\boldsymbol {d}$
is not perpendicular to the inner interface (see figure 6), and
$\theta$
and
$D$
do not vary in an anti-phase manner anymore.
(
$a$
)
$D_{\textit{max}}$
versus
$ \textit{Ca} $
in the tumbling regime. (
$b$
,
$c$
) Orbit drift at
$Re=0.5$
and various
$ \textit{Ca} $
: (
$b$
)
$ \textit{Ca} =0.02$
and (
$c$
)
$ \textit{Ca} =0.2$
, in terms of evolution of
$\theta$
and
$D$
.

Evolution of orbit parameter
$C$
, polar angle
$\theta$
,
$\lambda _e$
and
$t_N$
of the dumbbell-shaped Janus drops at
$Re=0.2$
and various
$ \textit{Ca} =$
(
$a, b$
) 0.02 and (
$c, d$
)
$0.15$
. In panel (
$a$
), the variation of
$C$
with time follows
$C\sim {\textrm {exp}}(0.0075\dot {\gamma }t)$
. In panel (
$b$
), the variation of
$C$
has two trends:
$C\sim {\textrm {exp}}(0.078\dot {\gamma }t)$
and
$C\sim {\textrm {exp}}(0.025\dot {\gamma }t)$
. In panels (
$b$
) and (
$d$
),
$t^*_N=\pi (\lambda _e+\lambda _e^{-1})/{\dot {\gamma }}$
.

To quantitatively assess the effect of drop deformation on the orbit drift, we calculate
$\lambda _e$
in each half-period of rotation in the orbit drift, based on (3.4) and
$\lambda _{\textit{max}}$
in the corresponding half-period. Here, the
$N\mathrm{th}$
half-period of rotation is denoted by
$t_N$
, and defined as
$(N-1)\pi \leqslant \phi \leqslant N\pi$
; an example
$t_{N=5}$
is displayed in figure 14(
$a$
). Note
$t_R=t_{N=2k-1}+t_{N=2k}$
for the Janus drop in the
$k\mathrm{th}$
rotation period. Then, we calculate the orbit parameter
$C=\lambda _e^{-1}\tan \theta (\sin ^2\phi +\lambda _e^2\cos ^2\phi )^{1/2}$
of rotating Janus drops, by integrating
$\lambda _e$
,
$\theta$
and
$\phi$
obtained from numerical simulations into (1.1). In turn, we can provide a prediction of the polar angle,
$\theta ^*=\arctan [C_N\lambda _e/ (\sin ^2\phi ^*+\lambda _e^2\cos ^2\phi ^*)^{1/2} ]$
, in each half-period of rotation, where
$\phi ^*=\arctan [\lambda _e\tan (\dot {\gamma }(t-t_0)/(\lambda _e+\lambda _e^{-1}) ) ]$
,
$t_0$
and
$C_N$
denote the starting time and the characteristic orbit parameter of the half-period.
Figure 14 shows the process of orbit drift at
$Re=0.2$
and two
$ \textit{Ca} $
(
$=0.02$
and
$0.15$
), in terms of time evolution of
$C$
,
$\theta$
,
$\lambda _e$
and
$t_N$
. At small
$ \textit{Ca} $
(
$=0.02$
), the Janus drop drifts towards the tumbling orbit, accompanied by the increase of
$C$
in an oscillatory way. This remarkably resembles the orbit drift of prolate spheroids in shear flow (Subramanian & Koch Reference Subramanian and Koch2006). Moreover, we find that the time variation of
$C$
generally follows an exponential function,
${\textrm {exp}}(\eta \dot {\gamma }t)$
(figure 14
$a$
), where
$k$
(
$=0.0075$
) is a constant. Because of insignificant drop deformation (
$1\leqslant \lambda _e/\lambda _0\leqslant 1.03$
),
$t_N$
is more or less the same and can be well predicted by Jeffery’s theory (3.2) with
$\lambda _e=1.03\lambda _0$
(figure 14
$b$
). At all times, the prediction of
$\theta ^*$
is in excellent agreement with the numerical results of
$\theta$
(figure 14
$a$
). At relatively large
$ \textit{Ca} $
(
$=0.15$
), the time variation of
$C$
appears to transit from one exponential function of relatively low
$\eta$
,
${\textrm {exp}}(0.078\dot {\gamma }t$
) for
$N\leqslant 2$
, to another one of relatively high
$\eta$
,
${\textrm {exp}}(0.025\dot {\gamma }t$
) for
$N\geqslant 8$
(figure 14
$c$
). Clearly, a higher value of
$k$
results in a faster orbit drift. Such transition is closely related to the drop deformation: the values of
$\lambda _e$
are very close for
$N\leqslant 2$
and apparently saturated from
$N=8$
. Interestingly, the half-period of rotation during the orbit drift also agrees with the relation in (3.2) (figure 14
$d$
), which is initially obtained from tumbling Janus drops.
3.6. Deformation of near-spherical Janus drops
Near-spherical Janus drops in shear flow exhibit deformation behaviour remarkably similar to pure drops. Figures 15(
$a$
) and 15(
$b$
) show the time variation of the azimuthal angle
$\phi$
for a surface tracer particle (in the
$x{-}y$
plane at
$z=0$
) over one period for tumbling and spinning near-spherical Janus drops, respectively. Simulations of a pure drop in shear flow with identical parameters (volume, density, viscosity and surface tension
$\sigma _{13}$
) confirm that the shapes of Janus and pure drops overlap, independent of the flow regime (tumbling or spinning). However, the tracer particle completes one revolution in a shorter time on the tumbling Janus drop (13.06) compared with the pure drop (14.65), while the spinning Janus and pure drops share roughly the same period. This accelerated motion in the tumbling regime arises from the hydrodynamic torque on the inner interface due to the enclosed vortex within the Janus drop. To quantify the global deformation of the near-spherical Janus drop, we compute the deformation parameter
$D_T = (\varGamma _a - \varGamma _b)/(\varGamma _a + \varGamma _b)$
in the same way for pure drops, where
$\varGamma _a$
and
$\varGamma _b$
are semi-major and semi-minor axes of the deformed drop. Figure 15(
$c$
) shows the numerical results of deformation of Janus drops in shear flow at
$Re=0.5$
with respect to
$D_T$
versus
$ \textit{Ca} $
, along with those for 3-D pure drops: theoretical prediction (
$D_T=35Ca/32$
) (Taylor Reference Taylor1932) and numerical solutions (Liu et al. Reference Liu, Ng, Chong, Lohse and Verzicco2021). Clearly, a good agreement has been achieved, confirming the equivalent deformation dynamics.
Temporal evolution of
$\phi$
for near-spherical Janus drops and pure drops at
$Re=0.5$
,
$ \textit{Ca} $
=
$0.02$
and different initial orientations: (
$a$
)
$\theta _0=\pi /2$
and (
$b$
)
$\theta _0=0$
. The 2-D snapshots of Janus drop in the
$x{-}y$
plane at
$z=0$
are superimposed by that of the pure drop. (
$c$
) Taylor deformation parameter
$D_T$
as a function of
$ \textit{Ca} $
. Solid symbols denote the present results, hollow ones represent the results of Liu et al. (Reference Liu, Ng, Chong, Lohse and Verzicco2021) and the dashed line corresponds to
$D_T=(35/32)Ca$
.

4. Conclusion
The orbit dynamics of 3-D Janus drops in shear flow has been numerically investigated for dumbbell-shaped and near-spherical drop configurations. We find that dumbbell-shaped drops eventually evolve into the tumbling regime, while near-spherical drops enter the spinning regime. We attribute this bifurcation in orbit drift to contrasting deformation dynamics and shape-dependent hydrodynamics of the two configurations. For the dumbbell-shaped drops, the shear-induced global deformation effectively increases the drop aspect ratio, driving it to the higher orbit angle. This elevation further intensifies shear and deformation – a feedback mechanism causing orbit drift towards the tumbling regime. For the near-spherical drops, the rotation of the inner interface is driven by the hydrodynamic torque from the vortex enclosed within the drop. The asymmetry of the inner interface about the
$z=0$
plane leads to an unbalanced torque, thereby inducing a slow orbit drift towards the spinning regime.
For the dumbbell-shaped Janus drops, the tumbling dynamics is found to resemble ellipsoids with different aspect ratios. Moreover, the drop trajectories during orbit drift collapse onto a universal curve, independent of the initial orientations; significant deformation and inertia accelerate the orbit transition. To describe the effect of drop deformation on the orbit drift, we examine the deformation behaviour of dumbbell-shaped Janus drops at various capillary number and observe that the magnitude of drop deformation effectively increases with the orbit, as well as the effective aspect ratio. Accordingly, we propose a model for effective aspect ratio,
$\lambda _e = \lambda _0 + \beta (\lambda _{\textit{max}}-\lambda _0)$
, in which the drop shapes at equilibrium (
$\lambda _0$
) and at the maximum elongation (
$\lambda _{\textit{max}}$
) are taken into account, and
$\beta =0.6$
is a fitting parameter. Given the similarity in the orbit drift between the particles and the Janus drops, we incorporate the effective aspect ratio into Jeffery’s theory for solid particles to predict the orbit dynamics of dumbbell-shaped Janus drops. By comparison with the numerical results, we show that the resulting model can accurately predict the rotation period and angular velocity of Janus drops in the tumbling regime and during the orbit drift, especially for the drops with linear deformation. Finally, we find that the orbit parameter
$C$
varies exponentially with time for the drops with linear deformation, while the time variation of
$C$
transits from one exponential function to another for drops with nonlinear deformation.
In the present study, we only consider the Janus drops with two symmetrical component drops. Any difference in density, viscosity, surface tension coefficient and volume between the two component drops would introduce new asymmetries to the flow, and give rise to more complex dynamics of the Janus drops, which will be our future research topics.
Supplementary movies
Supplementary movies are available at https://doi.org/10.1017/jfm.2025.10988.
Funding
We are grateful for the support of the National Natural Science Foundation of China (grant numbers: 12388101, 12241204, 12302342, 11932019).
Declaration of interests
The authors report no conflict of interest.
Author contributions
J.-Q. C. and C.-Y. Z. have contributed equally to the article.
Appendix A. Calculation of the initial aspect ratio
$\boldsymbol{\lambda}_{\textbf{0}}$
and radius
$\boldsymbol{R}_{\textbf{0}}$
for 3-D symmetric Janus drops at equilibrium
The aspect ratio
$\lambda _0$
and radius
$R_0$
of the 3-D symmetric Janus drops at equilibrium can be calculated by the surface tension coefficient
$\sigma _{ij}$
between the
$i$
th and
$j$
th fluid. A sketch of the interface shape of a 3-D symmetric Janus drop at equilibrium is shown in figure 16. The interfacial angles of fluid
$i$
(
$=1$
, 2 or 3) are represented by
$\varphi _i$
. Clearly,
$\varphi _1+\varphi _2+\varphi _3=2\pi$
. A balance of surface tensions at the triple-phase line can be written as (Zhang et al. Reference Zhang, Gao, Li and Ding2021)
Given
$\sigma _{13}=\sigma _{23}$
for 3-D symmetric Janus drops, we have
$\varphi _1=\varphi _2=\arccos (-\sigma _{12}/(2\sigma _{13}))$
and
$\varphi _3=2\pi -2\varphi _1$
.
At equilibrium, the surfaces of the component drops are spherical, with radius
$R_0$
and size
$2\pi R^3/3$
, and the inner interface is a plane circle. Geometrically, the initial semi-major axis of the Janus drop
$L_0$
and the radius of the inner interface
$R_{a,0}$
(
$=R_{b,0}$
) yield
The aspect ratio of a Janus drop is defined as
$\lambda _0=L_0/(R_{a,0}R_{b,0})^{1/2}$
. Substituting
$L_0$
,
$R_{a,0}$
,
$R_{b,0}$
and
$\varphi _1$
into the definition of
$\lambda _0$
, it can be expressed by
\begin{equation} \lambda _0=\frac {1-\cos \varphi _1}{\sin \varphi _1}=\sqrt {\frac {2\sigma _{13}+\sigma _{12}}{2\sigma _{13}-\sigma _{12}}}. \end{equation}
Given that the volume of the component drop yields
$\pi (3R_0-L_0)L_0^2/3=2\pi R^3/3$
, the initial radius of the component drop can be calculated by
Sketch of a 3-D Janus drop at equilibrium.































































































































































































































