Revisiting tea-leaf paradox with a deformable spoon

Abstract

The tea-leaf paradox refers to the migration of tea leaves toward the center of a tea cup when the water is put in rotation by a spoon. Although much attention has been given to liquid and particle transport generated by a spoon, the back-action of the generated flow on a deformable spoon has not been considered. In this work, we therefore propose to revisit the tea-leaf paradox with a deformable stirrer consisting of a floating ball attached to a soft string. The later is set in rotation and the ball trajectory results from the balance between string tension and the back action of the surrounding flow, leading to surprising effects. At low rotation speeds, we show that the circular trajectory of the ball shrinks when we stir faster, which contradicts common intuition based on inertia. For higher rotation rates, the ball is either suddenly attracted toward the center, or is repulsed away from it, depending on the string length. A model explaining these observations is proposed, and validated using experimental measurements of the generated flow. Interestingly, the system exhibits strong hysteretic behavior, leading to a modified version of the tea-leaf paradox in which the spoon itself can be trapped at the center in the flow generated by its past motion.

In its simplest version, the tea leaf paradox arises when mixing a liquid with a spoon performing circular trajectories (Fig. 1a). The paradox arises when one notices that particles placed at the bottom tend to aggregate at the center of the cup rather than on the outside, as expected from centrifugal force. The transport arises due to three-dimensional flow generated along the vertical direction by the edges of the container Thomson and Thomson (1997). Such flows imposed by boundary conditions are of prime interest from geophysics to physical chemistry and microfluidics, for example to understand the preferential concentration of sediments concentration near a river bank’s meanders (Einstein, 1926; Friedkin, 1945; Bowker, 2007; Leopold and Wolman, 1960; Jackson, 1975), dissolution mediated by rotating paddles McCarthy et al. (2003), for controlling the aggregation of nanoparticles Zhang et al. (2023) or blood-plasma separation (Yeo et al., 2006; Arifin et al., 2006).

The back action of the flow generated by rotating free-to-move objects has also been investigated. For instance, magnetic disks placed in rotation at a liquid interface were shown to interact and self-assemble under the generated flow’s action (Grzybowski et al., 2000, 2001, 2002), and interact with their environment (Gorce et al., 2019, 2021). More generally, the interplay between deformable bodies and flow generation is at the heart of propulsion mechanisms for living organisms (Gazzola et al., 2014; Taylor, 1997; Wu, 2011; Ashraf et al., 2017) or robots (Oliveira Santos et al., 2023; Esposito et al., 2012; Zhu et al., 2019), an idealized case of such situation being the flow disturbance generated by a flexible plate (Wu, 1961; Alben, 2021; Toomey and Eldredge, 2008).

Figure 1: (a) Sketch of a rigid and a deformable spoon mixing a liquid. (b) Definition of geometrical parameters for the flexible stirrer consisting of a plastic ball of radius 7 mm pulled by an inextensible string attached to a rotation arm. (c) Experimental idealization of a rigid spoon (left) using a metalic stem, and its deformable counterpart made by replacing the stem by a wool strand (top right). The flow induced by the deformable stirrer transports particles (pepper grains) placed at the bottom near the center, showing that the tea leaf paradox still occurs in this case (bottom right).

Inspired by these works, we propose here to revisit the tea leaf paradox with a deformable spoon, and to focus on the motion of the later rather than on the flow (Fig. 1a). As an experimental model, we study the motion of a ball attached to one end of an inextensible string, the other end being attached to a rotating arm as in Fig. 1b-c. Imposing a circular trajectory results in the emergence of several unexpected phenomena that we describe in this work. First, the radius of the stirrer’s circular trajectory surprisingly decreases when the rotation rate increases. This holds until a critical rotation speed is reached, for which a strong bifurcation occurs and the ball is either suddenly attracted toward the center, or repelled outwards,depending on the string length. A model explaining these observations is developed, and we then show in particular that the lift force generated by the ball’s motion must be taken into account to satisfactory explain the result. A quantitative criteria for the appearance of the collapsed regime is derived, and we show that strong hysteresis is observed once the stirrer is trapped.

Figure 2: (a) Example of experimentally measured trajectories for different rotation speed $\Omega=0.2,1.4,2$ round/s and two different string length $L_{s}=4.6$ cm (top) and $L_{s}=5.3$ cm (bottom). (b) Experimental kinogrammes showing a circular trajectory (top), a collapsed state (middle) and a diverging state (bottom). (c) Picture from below showing the loss of tension in the string when a collapsed state is reached. (d) Radius of the trajectory as a function of the rotation speed for different string length $L_{s}$. For the shortest strings ($L_{s}\leq 4.6$ cm), the ball reaches a diverging state while longest strings ($L_{s}\geq 5$ cm) lead to collapsing states. (e) Motion of the ball in the laboratory or comoving frame and definition of the forces and angles. (f) Experimental measurement of $\tan(\phi)$ (with $\phi$ the rope angle $(-\vec{e}_{r},\vec{T})$) for different string length as a function of rotation speed $\Omega$ (left) and trajectory radius $R_{t}$ (right). All data collapse in a single line of slope $a_{exp}=0.56$ cm^-1 (dashed line) while theory predicts $a_{th}=0.14$ cm^-1 (dot line).

All experiments (except illustrations of Fig. 1) are conducted in a rectangular tank of 55$\times$55 cm² and filled with tap water with depth 6.5 cm, preventing the ball to get close from any edge. A model rigid spoon has been built using a 3D-printed plastic ball of radius $R_{s}=7$ mm and mass $m=1.2$ g at the end of a metallic stem (Fig 1c). A stepper motor is used to set it up in rotation with rate $\Omega~{}\sim~{}1$ round/s, and the rotation arm length is fixed to $W=4.4$ cm. This model spoon is made flexible by replacing the stem by a strand of wool of length $L_{s}$ and diameter $\leq~{}0.2$ mm as sketched in Fig. 1b. Such strand is, in our regimes, non-extensible, of negligible mass compared to the ball and free of static or plastic deformation after being constrained. As we aim to study here the ball’s motion at the fluid interface, the final version of the setup includes a small tube to guide the strand close from the water surface as in Fig. 1c. This allows to confine the deformable system in the horizontal plane while avoiding the introduction of constrains on the string. When the rotation is turned on, a short transient regime of duration less than 10s is observed during which the ball converges toward a circular trajectory or radius $R_{t}$. On a minute scale, articles placed at the bottom converge toward the center of the tank (see Fig. 1c and supplementary movies), showing that three dimensional flows and associated transport (’tea-leaf’ paradox) still occur with our model deformable spoon. In what follows, we will focus on the surprising properties of the stirrer’s motion.

A camera (Basler) placed below the tank records 10 images per revolution of the rotation arm (see Fig. 2c for a typical image). The ball’s position is retrieved on each image using a convolution algorithm detailed in the Supplementary Informations (SI). Examples of measured trajectories are shown in Fig. 2a for two string length $L_{s}=4.6$ cm, $L_{s}=5.3$ cm and different rotation speeds $\Omega=0.2,1.4$ and 2 round/s. The trajectories are circular in all cases but their radius $R_{t}$ strongly depends on the rotation speed and the string length. Several remarkable features can be observed. First, when the rotation speed increases from $0.2$ round/s to $1.4$ round/s, the radius of the trajectory decreases. This is opposite from what one occurs when the ball hangs in the air, for which the increasing centrifugal force pushes the ball more on the outside. Second, for the case $L_{s}=5.3$ cm, the ball reaches a trajectory with zero radius for $\Omega=2$ round/s, and the string goes from straight to curved shape (Fig. 2c). In this case, the ball simply spins at the center in what we will call a ’collapsed’ or ’self-trapped’ behavior. For a slightly shorter string $L_{s}=4.6$ cm, the same rotation speed $\Omega=2$ round/s gives rise to opposite behavior as the radius of the trajectory brutally increases, in what we call a ’diverging’ behavior. Kinogrammes obtained by stacking images from the side are shown in Fig. 2b and display the three regimes discussed above (see also supplementary movies). Experiment snapshots in Fig. 2c show that the string is straight in non-collapsed cases, but gains some curvature in the collapsed state.

More systematic experiments allow to plot the trajectory’s radius as a function of rotation rate for different string length varying from 2.7 cm to 5.9 cm. The results are shown in Fig. 2d. Aside from the smallest values of string length $L_{s}=2.7$ cm, two regimes can be identified. Starting at low rotation rate, increasing the later leads to diminishing the radius trajectory. This holds until a critical rotation rate, for which the system exhibits strong bifurcation and the radius brutally changes. Depending on the string length, the ball is either suddenly expelled to the outward or attracted in the inward of the circular trajectory, corresponding to diverging and collapsing behavior discussed above. The first case occurs for $L_{s}\leq 4.6$ cm, while the second case is observed for $L_{s}\geq 5$ cm.

We now propose to perform an analysis of the forces acting on the ball in the comoving frame. We will restrain ourselves to the case where the string is tensed and the immersed volume constant. From experimental observations, this corresponds to the cases where the trajectory radius diminishes with $\Omega$ without being collapsed (see SI for details on data selection). For now, we will consider the inertial force $\vec{F}_{in}$, the viscous force $\vec{F}_{v}$ and the string tension $\vec{T}$ sketched in Fig. 2e. We assume that the ball performs a circular motion of radius $R_{t}$ with rotation speed $\Omega$. The inertial force due to the acceleration writes $\vec{F}_{in}=3/2mR_{t}\Omega^{2}\vec{e}_{r}$ where the 3/2 factor accounts for the added mass of fluid that needs to be pushed by the sphere. As Reynolds number $Re$ typically ranges from $500$ to $5000$ in our experiments, we take a quadratic law for the viscous force $\vec{F}_{v}=1/2\rho C_{D}S_{im}(R_{t}\Omega)^{2}\vec{e}_{\theta}$ with $C_{D}\approx 0.4$ is the dimensionless drag parameter Clanet (2015) and $S_{im}\approx 0.8\pi R^{2}$ is the immersed cross-sectional surface of the object in the direction of the motion.

As the ball is steady in the comoving frame, one gets $\vec{T}=-\vec{F}_{in}-\vec{F}_{v}$. In particular, the angle $\phi=(-\vec{e}_{r},\vec{T})$ (see Fig.3a) can be expressed as

$$\tan(\phi)=\frac{F_{v}}{F_{in}}=\alpha_{th}R_{t}$$ (1)

where $\alpha_{th}\approx 0.24C_{D}/R_{s}$. Experimental measurements of $\tan(\phi_{d})$ as a function of $\Omega$ are shown in Fig. 2f. All measurements collapse on a single curve $\tan(\phi)=\alpha_{exp}R_{t}$ as shown in 2f, confirming the scaling law in Eq. 1. However, the measured experimental value $\alpha_{exp}=0.56$ cm^-1 is much larger than the predicted value $\alpha_{th}=0.14$ cm^-1.

Figure 3: (a) Stack of 15 consecutive images taken at 150 frame/s showing the motion of particles at the water surface in the laboratory (top) and comoving (botton) frame. Direction of lift force $\vec{F}_{L}$ and fluid motion is also indicated. (b) Experimental measurement of velocity field’s circulation using PIV for different contour sizes and three different rotation rate of the ball. Error bar show the standard deviation of the measurement over 40 consecutive images. (c) Circulation as a function of rotation speed and linear fit $\Gamma=S_{M}\Omega$ with $S_{M}=-5.4$ cm^-2.

Such discrepancy can be solved by considering the additional lift force produced by the fluid. The later is generated by the complete solid rotation of the sphere during its revolution, shown in Fig. 2e, combined with non-zero relative velocity of the ball with respect to the fluid Falkovich (2018); Rubinow and Keller (1961). A quantitative characterization of the resulting surface flow is performed using particle image velocimetry (PIV) measurement with PIVLab (Thielicke and Stamhuis, 2014; Thielicke and Sonntag, 2021). The liquid surface was seeded with small particles (pepper grains) while water was made opaque using a bit of powder milk to ensure good contrast. A camera placed above the tank recorded 150 frames/s during a full revolution of the ball, ensuring small displacement of the ball and the particles between two images. A stack of 15 consecutive images is shown in Fig. 4a. The same stack in the co-moving frame shows the streamline’s bending by the ball. From this picture, one can expect a lift force directed toward the inward of the trajectory. Its magnitude can be estimated starting from the lift force on a cylinder of height $D$ in a two dimensional flow Falkovich (2018); Dupeux et al. (2010)

$$\vec{F}_{L}=\rho DR_{t}\Omega\Gamma\vec{e}_{r}$$ (2)

where $\Gamma=\oint\vec{v}.\vec{d}l$ is the circulation taken around the object far from the boundary layer. The circulation at the water surface can be measured experimentally from PIV measurement. The contour we chose is a square of side $L$ centered on the ball as shown in Fig. 3b. The circulation $\Gamma$ for different contour size $L$ is shown in the same figure for three rotation rates $\Omega=0.2,0.6,1.0$ rad/s. The error bars are computed from the standard deviation on the measurement over 40 consecutive frames, showing low dispersion of measurement along time. As the size of the contour increases, the circulation increases and eventually reaches a plateau for the largest contours, which is the circulation value $\Gamma$ introduced in Eq. (2) Falkovich (2018). The later is shown as a function of the rotation rate in Fig. 3e and shows excellent agreement with a linear fit $\Gamma=S_{M}\Omega$ with $S_{M}=-5.4$ cm². The lift force can therefore be expressed as $\vec{F}_{M}=-MR_{t}\Omega^{2}\vec{e}_{r}$ with $M=\rho DS_{M}$ homogeneous to a mass and $D$ a distance let as a free parameter for now. The lift force scales exactly as the inertial force but with an effective negative mass. One must therefore modify Eq. 1 as

$$\tan(\phi)=\frac{a_{th}}{1-2M/3m}R_{t}$$ (3)

Figure 4: (a) Angle $\phi$ as a function of the trajectory radius $R_{t}$ for different string length $L_{s}$ according to Eq. (4) (color lines) and to force equilibrium of Eq. (1) (dashed line). Either zero or two equilibrium angles fulfill both constraints depending on $L_{s}$, and only the largest one is stable in the later case (red crosses). (b) Increasing string elongation $\Delta L_{s}$ measured experimentally as a function of $\Omega$ of different string length. (c) Elongation of the string in the horizontal plane due to increasing tension with $\Omega$. (d) Radius of the equilibrium trajectories when the string is elongated from $L_{s}$ to $L_{s}+\Delta L_{s}$. (e) Radius of the trajectory for the diverging (left, $L_{s}=4.6$ cm) and collapsing (right, $L_{s}=5.0$ cm) when the rotation speed is progressively increased (blue circles) or decreased (red squares) (f) Critical rotation speeds $\Omega_{u}$ and $\Omega_{d}$ as a function of string length $L_{s}$.

The linear relation $\Gamma\propto\Omega$ ensures that both equations (1) and (3) have similar dependency with $R_{t}$ but with an increasing prefactor in the second case, which is perfectly consistent with our results. Using the found value $\alpha_{exp}$ in Eq. (3) fixes the only free parameter of our model $D=0.3$ cm. For a purely two-dimensional flow, $D$ should exactly be the penetration length of the sphere in water, which is $1.0$ cm. The mismatch between the two can be attributed to the spherical shape of our object as well as to possible three-dimensional flows generation, in line with the particle’s motion discussed in Fig. 1b.

So far, we have shown that our experimental measurement were consistent with the force analysis expressed in Eq. (1). Nevertheless, this is not enough to predict the trajectories radius $R_{t}$ shown in Fig. 2d. The later can be fully determined if one assumes that the string is straight. In this case, the triangle formed by the ball, the hooking point and the rotation center (Fig. 4a) imposes an extra relationship between the geometrical parameters that reads

$$\phi=\arccos\left(\frac{L_{s}^{2}+R_{t}^{2}-W^{2}}{2L_{s}R_{t}}\right)$$ (4)

A plot of this geometrical constrain is shown in Fig.4a for $L_{s}=5,5.2$ and $5.4$ cm (plain colored lines), while force equilibrium from Eq. (1) is represented by the black dashed line. For smallest string length, two solutions $(\phi,R_{t})$ are found while no solution exists when $L_{s}$ is too large. Among the two solutions, one can show that only the largest one materialized by red crosses in Fig. 4a is stable (see SI). This finally fully determines the radius trajectory for a given string length.

The last step comes by noticing that the string length in the interface plane (also denoted $L_{s}$) increases with the rotation rate. This is shown by experimental measurement of the string elongation $\Delta L_{s}=L_{s}(\Omega)-L_{s}(\Omega_{0})$ in Fig. 4b, which is by definition zero for $\Omega_{0}=0.2$ round/s. We observed experimentally that the string’s edges tends to become more horizontal when $\Omega$ increases, as sketched in Fig. 4c (see also pictures in SI). The tension increase, which scales as $T\propto\Omega^{2}$, leads to three-dimensional reorganization of the inextensible string and to the apparent flexibility in the horizontal plane. It is this varying string length in the horizontal plane that should be taken into account for the geometrical constrain (4). The resulting equilibrium radius $R_{t}$ therefore depends on the elongation $\Delta L_{s}$ and is plotted in Fig. 4d for three initial string length and various elongations. The radius $R_{t}$ progressively decreases when the later increases, which is consistent with our experimental observations. The collapse of the system can also be interpreted as the loss of solutions discussed in Fig. 4a when the string becomes too long. In this case, the ball’s equilibrium and the string tension cannot be guaranteed simultaneously, which leads to a shape change of the rope. The later occurs for smaller values of elongation when the initial string length is larger. Last, the predicted value for the radius before collapse is approximately 2 cm and independent of $L_{s}$. All those features are remarkably consistent with our experimental results of Fig. 2d.

A last aspect of the ball’s motion is the stability of the collapsed state. Starting from a collapsed state obtained for $L_{s}=5$ cm and $\Omega_{u}=2.6$ round/s (blue circles in Fig. 4e), we progressively decrease the rotation speed. Interestingly, the ball remains trapped until $\Omega$ goes below $\Omega_{d}=1.2$ round/s (red squares), for which circular trajectories appear again. Such hysteresis was observed in all collapsing configurations, but not in the diverging case as shown in Fig. 4f. We plot in Fig. 4g the critical rotation speeds $\Omega_{u}$ and $\Omega_{d}$ for various string length $L_{s}$. Their difference decreases with $L_{s}$, showing that the hysteresis is stronger near the transition between diverging and collapsing behavior for a reason that is still to be understood. The hysteresis is due to the existence of a memory in the system, which can be qualitatively interpreted in two ways. First, the tension lost in the string needs to be recovered to expel the ball from the flow’s center. Second, the surrounding flow exhibits circular symmetry that needs to be broken to recover circular trajectories. Both features require a strong perturbation to be broken, leading to a robust hysteresis.

We have discussed in this work the circular motion of a ball pulled by a string and show the prime importance of the string’s deformation and flow’s action to explain the variety of observed trajectories. This opens several interesting avenues for the future. Our results first strongly suggests to perform the same experiments with several objects to introduce pairwise interactions mediated by the flow as done in other systems Grzybowski et al. (2000). It would also be interesting to replace water with a non-newtonian fluid, which were for instance shown to modify swimmers efficiency Espinosa-Garcia et al. (2013). Last, replacing the inextensible rope by a flexible one could also lead to new physical features that are still to explore.