🔬 condensed matter

Mixing induced by microswimmers as probed by mutual information

This study demonstrates that mutual information serves as an effective global metric for quantifying fluid mixing by microswimmers, revealing that mixing efficiency is maximized by positional and orientational disorder and an optimal balance between swimmer translation and dipolar flow generation, while being suppressed by aggregation.

Original authors: Yihong Shi, Yuto Hosaka, Andrej Vilfan, Ramin Golestanian

Published 2026-07-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Yihong Shi, Yuto Hosaka, Andrej Vilfan, Ramin Golestanian

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a crowded dance floor where tiny, self-propelled dancers (microswimmers) are trying to mix a giant bowl of soup. The soup represents a fluid, and the dancers are microscopic organisms or artificial swimmers. The goal is to get the ingredients in the soup to blend together perfectly.

In the world of the very small, water feels thick like honey, and these tiny dancers can't just rely on stirring with a spoon (which doesn't work well at this scale) or waiting for the ingredients to slowly drift apart on their own (which takes forever). They need to move in a specific way to create chaos and mix things up.

This paper investigates how well these tiny dancers mix the soup and introduces a clever new way to measure that mixing.

The New Ruler: "Mutual Information"

Usually, scientists measure mixing by looking at how much a specific pattern (like a drop of dye) has stretched out. But what if there is no pattern to begin with? What if the soup is already a bit mixed, and you just want to know how efficiently it's being stirred?

The authors use a concept from information theory called Mutual Information. Think of it like a game of "Where's Waldo?"

High Mutual Information: If you know where a particle started, you can still guess exactly where it is now. The soup hasn't mixed well; the particle is still in its original neighborhood.
Low Mutual Information: If you know where a particle started, you have absolutely no idea where it is now. It could be anywhere. The soup is perfectly mixed.

The paper uses this "guessing game" score to measure mixing efficiency. The faster the score drops to zero, the better the mixing.

The Dancers: The "Squirmer" Model

The researchers used a mathematical model called a "squirmer" to represent these swimmers. Imagine a tiny, round ball that swims by wiggling its surface. Depending on how it wiggles, it creates different flow patterns:

The Puller: Pulls fluid from the front and pushes it out the sides (like a swimmer doing a breaststroke).
The Pusher: Pushes fluid from the back and pulls it in from the sides (like a swimmer doing a backstroke).
The Stresslet: A dancer that pushes and pulls but doesn't actually move forward (like a stationary propeller).
The Source Dipole: A dancer that moves forward but doesn't create much swirling flow.

What They Discovered

1. Chaos is Good, Order is Bad
If you line up all the dancers in a perfect grid, all facing the same direction, they create a very organized, predictable flow. It's like a military parade; everyone moves in sync, but the soup doesn't get mixed well.

The Finding: When the dancers are disordered (facing random directions and standing in random spots), they create a chaotic, swirling mess. This chaos is exactly what's needed to mix the fluid efficiently.
The Aggregation Trap: If the dancers huddle together in a tight group in the center of the room, they only mix the center. The edges of the soup stay unmixed. To mix the whole bowl, the dancers need to be spread out.

2. The Goldilocks Zone of Movement
The researchers asked: "If all dancers burn the same amount of energy, which type mixes best?"

They found that the answer isn't the strongest flow or the fastest swimmer. It's a balance.
If a dancer creates a huge swirling flow but doesn't move (like a Stresslet), it mixes well locally but stays in one spot.
If a dancer swims fast but creates no swirl (like a Source Dipole), it moves everywhere but doesn't stir the soup.
The Winner: The most efficient mixer is a dancer that does both—it creates a good swirl and moves around the room to spread that swirl everywhere. It's a "Goldilocks" scenario: not too much flow, not too much movement, but just the right mix of both.

3. Pushers vs. Pullers
When the dancers interact with each other's water currents (hydrodynamic interactions):

Pushers (like E. coli) turned out to be better mixers than Pullers (like Chlamydomonas).
Why? Pullers tend to get attracted to each other and form clumps (which, as we saw, is bad for mixing). Pushers tend to push each other apart, keeping the group spread out and the mixing efficient.

4. The Three Stages of Mixing
The paper breaks the mixing process into three distinct acts:

Act 1 (The Slow Start): At first, mixing is slow and driven only by the natural, slow drifting of particles (diffusion).
Act 2 (The Swirl): The dancers start moving, creating currents that speed up the mixing significantly. This is the "advection" phase.
Act 3 (The Big Picture): Eventually, the particles hit the walls of the container. Now, the size of the room matters most. The mixing slows down as the particles try to fill the entire space evenly.

The Bottom Line

This paper shows that to mix a fluid at the microscopic level, you don't just need strong swimmers; you need disorder. You want your microscopic dancers to be spread out, facing random directions, and moving in a way that balances creating swirls with traveling across the room. By using this new "information" measurement, the researchers proved that this chaotic, balanced approach is the key to efficient mixing in the tiny world.

Technical Summary: Mixing Induced by Microswimmers as Probed by Mutual Information

Problem Statement
Fluid mixing at the microscale is a critical process in both biological systems (e.g., nutrient uptake, biogenic mixing in oceans) and artificial applications (e.g., microreactors, lab-on-a-chip). Unlike macroscopic mixing, microscale flows are governed by low Reynolds numbers, rendering them time-reversible; thus, stirring alone is insufficient, and mixing relies on the interplay between advection and molecular diffusion. While conventional metrics for mixing efficiency exist (e.g., $L_2$ norms, Lyapunov exponents), they often depend on specific initial patterns or local stretching. The authors aim to investigate mixing induced by microswimmers using mutual information, an information-theoretic measure previously validated only in simplified model systems (such as Couette flow), to determine its applicability to complex, active fluid flows generated by swimmers.

Methodology
The study employs a two-dimensional (2D) squirmer model to describe non-deformable, force-free circular particles swimming in Stokes flow.

Flow Model: The flow field is generated by a tangential slip velocity on the swimmer's surface, characterized by modes $B_n$ . The authors retain the first two modes: $B_1$ (source dipole, determining translational velocity) and $B_2$ (force dipole/stresslet, determining the swimmer type: pusher, puller, or stresslet). The squirmer parameter is defined as $\beta = B_2/B_1$ .
Confinement: Simulations are conducted in a 2D square domain with perfect-slip (stress-free) boundary conditions. The confined Green's function is constructed using the method of images and computed in Fourier space to ensure numerical stability, including spectral regularization to represent finite-sized swimmers.
Mixing Quantification: Mixing efficiency is quantified by the mutual information $I[x_0; x_t]$ between a particle's initial position $x_0$ and its final position $x_t$ at time $t$ . This is calculated as the difference between the Shannon entropy of the final position distribution and the conditional entropy given the initial position.
Transport Dynamics: The evolution of the probability density function (PDF) of tracer particles is solved numerically using the advection-diffusion (Fokker-Planck) equation. The conditional PDF is obtained via a pseudo-spectral method.
Parameters: The study compares different swimmer configurations (ordered vs. disordered positions/orientations), swimmer types (stresslets, pushers, pullers, source dipoles) under a fixed total energy dissipation constraint, and systems with and without hydrodynamic interactions (HIs).

Key Results

Impact of Spatial Organization:
- Disorder Enhances Mixing: Highly symmetric configurations (e.g., ordered positions with identical orientations) suppress large-scale transport by creating flow domains separated by stagnation lines. Introducing disorder in swimmer orientations and positions breaks these symmetries, generating more irregular velocity fields that enhance tracer stretching and dispersion.
- Aggregation Suppresses Mixing: Strong aggregation of swimmers in a localized region concentrates the induced flow, leaving outer regions largely unaffected. As the spatial distribution of swimmers broadens, mixing efficiency increases.
Optimal Swimmer Type and Dissipation:
- Non-monotonic Dependence: At a fixed energy dissipation rate, mixing efficiency depends non-monotonically on the squirmer parameter $|\beta|$ .
- Trade-off: There is a trade-off between generating strong, long-range shear flows (dominated by the stresslet component, $|\beta| \to \infty$ ) and achieving spatial dispersion via swimmer translation (dominated by the source dipole component, $\beta=0$ ).
- Optimal Finite Value: The optimal mixing efficiency occurs at a finite value of $|\beta|$ , balancing the long-range flow of stresslets with the translational motion of the swimmer. Pure stresslets ( $\beta \to \infty$ ) are non-motile in the absence of HIs and thus less effective at global mixing than swimmers with finite $|\beta|$ .
Hydrodynamic Interactions (HIs):
- When HIs are included, pushers ( $\beta < 0$ ) outperform pullers ( $\beta > 0$ ). This is attributed to the tendency of pullers to form aggregates or exhibit polar alignment, which reduces mixing efficiency.
- The inclusion of HIs smooths the dependence of efficiency on $|\beta|$ and shifts the optimal swimmer type toward larger values of $|\beta|$ .
Mixing Regimes and System Size:
- The decay of mutual information occurs in three distinct stages:
  1. Diffusion-dominated: Initial stage where mixing follows pure diffusion scaling.
  2. Advection-enhanced: Intermediate stage where advection accelerates mixing; the decay rate is largely independent of system size for a fixed swimmer density.
  3. Relaxation stage: Final stage controlled by system size, where tracers interact with boundaries.
- Larger systems retain more mutual information for longer periods, and the relaxation time scales with system size, though the dependence is weaker than the $L^2$ scaling expected for pure effective diffusion due to the extended range of advective flows.

Significance and Claims
The paper demonstrates that mutual information is a robust, universal, and assumption-free measure for quantifying mixing efficiency in complex active flows, extending its validity beyond simplified shear flows. The study reveals that microswimmer-induced mixing is governed by a fundamental trade-off: maximizing mixing requires balancing the generation of strong shear flows (via dipolar components) with the spatial dispersion provided by swimmer translation. Furthermore, the results highlight that disorder in swimmer organization and collective hydrodynamic interactions (favoring pushers over pullers) are critical factors in determining the efficiency of biogenic mixing. The framework provides a quantitative basis for understanding transport phenomena in active fluids, with potential relevance to ciliary flows, microbial suspensions, and intracellular transport.

The New Ruler: "Mutual Information"

The Dancers: The "Squirmer" Model

What They Discovered

The Bottom Line

More like this