Autonomous life-like behavior emerging in active and flexible microstructures

Daniela J. Kraft; Mengshi Wei

REVIEW 3 major objections 6 minor 3 references

Autonomous life-like behavior emerging in active and flexible microstructures

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Flexible chains of active micro-units navigate, beat, and burrow without programming.

desk verdict A genuinely new microprinting platform that shows a rich catalog of active-flexible behaviors, but the 'embodied intelligence' headline outruns the evidence because the authors never separate shape-motion feedback from direct dielectrophoretic steering. read the letter →

arxiv 2506.15198 v1 pith:UQ4ZW7ZR submitted 2025-06-18 cond-mat.soft

Mengshi Wei , Daniela J. Kraft This is my paper · ORCID

classification cond-mat.soft

keywords activematterself-dielectrophoresismicroswimmersflexiblemicrostructuresembodiedintelligencetangentiallydrivenpolymers3Dmicroprintingshape-motionfeedback

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper seeks to establish that adding mechanical flexibility to shape-anisotropic active particles is enough to produce autonomous, adaptive behavior at the micrometer scale. The authors 3D-print chains of hinged half-cylinder units and actuate them with a uniform AC electric field; because each unit propels along its own orientation and the units are linked so that orientation follows the chain's contour, the chain's shape and its motion continually reshape each other. On this minimal feedback they demonstrate a repertoire of life-like behaviors: railway-like following, flagellar beating when clamped, undulatory swimming with a load, wall reorientation, navigation through obstacle arrays, collision avoidance, and burrowing through crowded environments. If the claim is right, it establishes a minimal physical route to embodied intelligence at the micrometer scale, where conventional sensing and control components are impractical.

What carries the argument

The central object is a chain of 3D-printed half-cylinder units joined by handle-and-beam hinges that restrict bending to about ±50 degrees. In a uniform AC electric field, each anisotropic unit polarizes and propels itself along its own orientation; the authors attribute the dominant contribution to self-dielectrophoresis, in which the particle moves within the non-uniform field it induces around itself. The units are arranged so that each unit's propulsion direction follows the local chain contour, making the chain an experimental realization of a tangentially driven active polymer. Dipolar repulsion between neighboring units adds effective bending rigidity, while the geometric hinge limit sets a maximum buckle angle; the combination turns shape deformation into reorientation of propulsion, and that reorientation back into further deformation—the feedback loop that carries the reported behaviors.

What would settle it

Print a chain whose units are deliberately linked so their self-propulsion points sideways instead of along the chain, and run it on the same wall and obstacle course. If it still reorients and navigates, the shape–motion feedback is not the mechanism; if it cannot turn, the feedback is necessary.

Watch

Extended reading notes

Core claim

The paper's central claim is that a flexible chain of shape-anisotropic, self-propelling units can behave as an autonomous physical system whose conformation and locomotion are coupled. Free chains straighten and exhibit railway motion, with each segment following the one ahead. Clamping the head, or attaching a load, causes active forces to buckle the chain, and the buckled shape reorients the units' propulsion directions, producing self-oscillations that trace a limit cycle; the beating frequency is proportional to the speed of a single unit. Raising the AC frequency reverses propulsion and converts oscillation into extension, allowing compression–extension cycles. In structured environments the same feedback lets chains reorient away from walls, turn 90 and 180 degrees to pass through pillar arrays, slip past one another, and push through dense crowds of spheres. The authors attribute all of these behaviors to shape–motion feedback, with dipolar repulsion between units providing effective elasticity.

Load-bearing premise

The load-bearing premise is that each unit pushes itself along its own orientation and the links keep those orientations aligned with the chain's local direction; if external field gradients or nearby walls instead set the movement direction, the shape–motion feedback would not be the cause of the reported autonomy.

Editorial extensions

If this is right

Autonomous wall reorientation, obstacle navigation, and collision avoidance emerge in micrometer-scale structures with no sensors, controllers, or pre-programming.
Clamping or loading a chain switches it from railway motion to self-oscillation, and raising the field frequency reverses propulsion, so a single chain can be externally toggled between oscillation and extension modes.
The beating frequency of a clamped chain grows linearly with single-unit speed, so the oscillation rate is set predictably by the applied voltage.
Because the design is modular, chains with different numbers of units, hinge ranges, loads, and propulsion directions can be printed, giving a broad design space for autonomous microstructures.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Beyond the paper: because the hinge range, unit shape, and propulsion direction are independent design parameters, chains with heterogeneous stiffness or propulsion strengths could be built to perform task-specific gaits without reprogramming.
Beyond the paper: an explicit model that treats the connector's hard maximum bend angle as the buckling limiter could connect the observed linear beating-frequency scaling to the geometric-hinge mechanism.
Beyond the paper: repeating the wall-reorientation and obstacle-navigation tests at MHz actuation would show whether the shape–motion feedback persists when the propulsion direction reverses.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Dear colleague,

You should know two things before spending an evening on this one. First, it is a real experimental advance: 3D-printed concatenated microchains that self-propel by shape-anisotropic self-dielectrophoresis and display a remarkable catalog of behaviors — railway motion, flagellar-like beating under clamping, undulatory load pushing, rotation, and what looks like navigation through obstacle arrays. Second, the most provocative half of the title, "autonomous life-like behavior" and "embodied intelligence," is not yet backed by the controls needed to pin the mechanism. The stress-test note lands.

What is actually new and good: the fabrication route to flexible hinges via two-photon polymerization, with thermal fluctuation data showing the hinge flexibility; the demonstration that these chains are experimental realizations of tangentially driven active polymers; and the careful oscillation analysis with limit cycles and geometric buckling. The propulsion mechanism is also reasonably supported: bulk levitation rules out electrode EHD flow, tracer experiments suggest no convective flow, and the COMSOL sDEP calculation uses literature parameters and predicts speeds in the observed range without fitting to the reported speeds. That is honest, reproducible work.

The main soft spot is causal attribution. The SI states that the ellipsoidal obstacles "impose a repulsive force at the equator of the sphere via DEP," so the obstacle interactions include direct dielectrophoretic forces on the chain that have nothing to do with its shape. Walls also perturb the field and can exert torques on the head. Without a rigid control — same geometry with hinges fused, or a straight rigid rod — you cannot tell whether the reported reorientation, turning, and collision avoidance are due to shape-motion feedback or simply the chain being pushed by field gradients. The wall-buckling observation is consistent with DEP steering with buckling as a passive consequence. This is a load-bearing gap for the sense-response claims, and it is acknowledged in the SI, even if not framed as a limitation.

Two smaller issues. The main text says speed is proportional to applied voltage, while Fig. S2A collapses the data as U versus E^2; the SI says U ∝ E^2. That inconsistency should be fixed. And several quantitative plots (Fig. 1E, Fig. 2H) lack error bars and replicate counts, which matters when the paper claims scaling laws.

The propulsion mechanism is not fully closed — the material parameters are literature PMMA values, not measured AC properties of the actual photoresist, and the COMSOL is 2D DC — but the qualitative evidence is strong enough that I would not block on that.

Who is this for? Soft matter, active matter, and micro-robotics readers will get real value from the fabrication platform and the behavioral repertoire. It deserves a serious referee. My recommendation: send to peer review with a request for rigid controls or a substantial softening of the embodied-intelligence language. As it stands, the paper is a solid proof-of-concept that currently overstates its mechanism.

Referee Report

3 major / 6 minor

Summary. The paper reports 3D-microprinted chain-like microstructures made of concatenated half-cylinder units with flexible hinges, actuated by an AC electric field. The authors propose that each unit self-propels via self-dielectrophoresis (sDEP), with its propulsion direction aligned with the chain contour, making the chains experimental realizations of tangentially driven active polymers. They demonstrate a rich set of behaviors—railway motion, clamped self-oscillation, load-induced undulation, rotation and tumbling, wall reorientation, obstacle-array navigation, collision avoidance, and burrowing through crowded dispersions—and interpret these as arising from feedback between the chain's shape and its motion, without sensors, software, or pre-programming. The paper also presents a COMSOL simulation of sDEP forces and a tracer-particle test to support the propulsion mechanism, and it offers a geometric explanation for the observed linear scaling of beating frequency with propulsion speed.

Significance. If the interpretation is correct, this is a significant advance in microscale active matter and soft robotics: it would show that simple synthetic structures with only shape-anisotropic propulsion and mechanical flexibility can exhibit adaptive, autonomous behaviors usually associated with living organisms. The experimental work is visually rich and includes several thoughtful controls: bulk levitation tests that argue against electrode-based electrohydrodynamic flow, tracer experiments that show attractive dielectrophoretic capture consistent with negative dielectrophoresis, and a COMSOL force calculation using literature-based material parameters rather than fitted parameters. The paper's central prediction—that geometric hinge constraints, rather than a competition between active force and bending rigidity, set the buckling conformation and yield f∝U—is specific and falsifiable. However, the causal claim that wall reorientation and obstacle navigation arise from the shape-motion feedback loop rather than from direct dielectrophoretic steering by field gradients is not yet isolated, and several quantitative claims lack the statistical detail needed to support them.

major comments (3)

[Section 'Sensing and smart adaptation' and Materials and Methods] The central claim that reorientation at walls, turns in obstacle arrays, and two-chain avoidance are caused by the feedback between chain shape and motion is not separated from direct dielectrophoretic steering by field gradients around boundaries and obstacles. The text itself attributes obstacle interactions to 'dipolar and steric repulsion from the obstacles' and states in Materials and Methods that the ellipsoidal obstacles 'impose a repulsive force at the equator of the sphere via DEP.' Since DEP forces and torques act on all chain units, the observed buckling and head realignment in Fig. 3A-B and the turns in Fig. 3C-E are equally consistent with a direct DEP torque on the head followed by passive buckling of the flexible body. A decisive control would be to compare the behavior of flexible chains with otherwise identical rigid (hinge-fused) chains under the same field, or to time-resolve whether head reorientation precedes or follows the buckling deformation. Without such a control, the 'sense-response' and 'embodied intelligence' claims remain a plausible but unproven interpretation.
[Fig. 1E, Fig. 2H, Fig. S2A, Supplementary Text 'AC field induced propulsion mechanism'] The quantitative support for the central scaling claims is incomplete. In Fig. S2A, the U∝E² collapse is obtained by normalizing each dataset with a free prefactor a that varies from 0.0146 to 0.0435; the paper does not state how a is determined or whether the collapse is robust when a is constrained. Fig. 1E and Fig. 2H show speed and frequency data without error bars or replicate counts, so the reported linear proportionality f∝U is not statistically established. The COMSOL sDEP calculation yields predicted speeds of 6.8 and 4.0 µm/s that are said to 'closely align' with experiments, but the experimental speeds are not reported with confidence intervals, and the sensitivity of the simulated force to the assumed particle permittivity and conductivity (taken for PMMA, not measured for the printed photoresist) is not assessed. Please add replicate numbers, error bars, and a propagation-of-uncertainty statement for the simulated force.
[Supplementary Text 'Self-oscillations and bending rigidity'] The derivation of the geometric scaling f∝U rests on the assertion that the active force is 'always much larger' than the thermal or dipolar bending resistance, so that the buckled conformation is set purely by the maximum geometric bending angle of about 50°. This force hierarchy is not directly measured. The observation that the buckled conformations are similar across voltages (Fig. S5A-F) is consistent with the geometric saturation hypothesis, but it does not exclude the possibility that the maximum angle is reached only in part of the cycle or that a voltage-dependent bending rigidity contributes to the scaling. A direct test, such as measuring the oscillation frequency as a function of solvent viscosity or calibrating the active force through drag measurements, would substantially strengthen the mechanistic claim.

minor comments (6)

[Abstract and Introduction] The phrase 'without the need for pre-programming or external control' is overstated because the AC field is an external energy source that sets the speed and, through its frequency, the propulsion direction. Please state explicitly that 'external control' means no real-time feedback or pre-programmed actuation sequence, not the absence of an external driving field.
[Materials and Methods, AC field induced propulsion mechanism] The Clausius-Mossotti expression in the supplementary text is corrupted ('𝜖H,F∗=𝜖?𝜖H,F−𝑖2%,#J'); please replace it with standard notation, for example ε* = ε₀ε_r − iσ/ω.
[Page 20, Supplementary Text] There is a typo in the sentence about the COMSOL speeds: 'which. closely align' should read 'which closely align'.
[Fig. 3 and main text, Page 5] The notations ⟨|Δθ+|⟩ and |θ#−θ⟂| in Fig. 3A-B are used before they are defined; please define them in the figure caption or immediately before the equations in the main text.
[Supplementary Text 'Self-oscillations and bending rigidity'] The claim that the maximum bending angle is 'about 50°' is used as an input to the geometric explanation; please report its uncertainty and explain how it was estimated from the probability density function in Fig. 1B.
[Fig. S2A] Please clarify in the caption how the normalization prefactor a is obtained for each dataset and whether a common value can be used; as written, the collapse to E² is not parameter-free.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the central quantities are measured or independently modeled, and the few self-citations are contextual, not load-bearing.

full rationale

The paper's central claim is an experimental demonstration rather than a derivation from a fitted model. The propulsion direction being aligned with the local chain contour is a stated design premise, not an output fitted to the reported behaviors; the railway, beating, undulation, wall-reorientation, obstacle-navigation, and collision-avoidance behaviors are directly measured. The COMSOL sDEP calculation uses literature-based permittivity and conductivity values for PMMA and DI water and predicts speeds of about 6.8 and 4.0 um/s without tuning those inputs to measured speeds, so no fitted variable is renamed as a prediction. The f proportional to U scaling is explained post hoc from the geometric maximum bending angle, but it is an interpretation of measured scaling, not a parameter recycled as a result. Refs. 19 and 36 are by the present authors, but they are cited only as contextual examples (photo-switchable active colloidal gels; floppy colloidal square lattices) and carry no load in the central argument. The SI statements that ellipsoidal obstacles impose a repulsive DEP force at their equator (Materials and Methods, Fabrication of crowded and structured environments) and that the MHz propulsion inversion requires further research (Materials and Methods, AC field induced propulsion mechanism) were examined: they identify an alternative direct-DEP mechanism and an open question, respectively, which are causal-identification or completeness concerns, not circular reductions. No equation in the paper is equivalent by construction to its input, and no uniqueness theorem or ansatz is imported through self-citation. Accordingly, there is no significant circularity.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central claims rest on the assumed alignment of unit propulsion with chain contour, the dominance of sDEP over other electrokinetic mechanisms, and the interpretation that navigation arises from internal shape-motion feedback rather than external DEP steering. The COMSOL material parameters and the per-experiment normalization prefactors are the only numeric inputs that are not measured for the actual system; they do not enter as fitted predictions of the main behaviors.

free parameters (2)

Normalization prefactor a for speed scaling collapse (Fig. S2A) = 0.0146, 0.0182, 0.0435, 0.0431, 0.0435, 0.0354 for different experiments
In Fig. S2A, the chain speed U is divided by a per-experiment prefactor a to collapse data onto a V_E^2 line. These prefactors are fitted per experimental condition (N_s, f_E) and are not predicted by the model; they are used only for plotting normalization and do not enter the central qualitative claim.
Material parameters for COMSOL sDEP simulation (epsilon_p, sigma_p, sigma_m, c_infinity) = epsilon_p=4, sigma_p=10^-11 S/m, sigma_m=4x10^-3 S/m, c_infinity=2.5x10^-6 M
These are estimated from literature values for PMMA and from the pH of Milli-Q water, not measured for the actual IP-Dip photoresist used. The predicted speeds (6.8 and 4.0 um/s) are claimed to align with experiments. These parameters are inputs, not fitted to the target data, but their uncertainty propagates into the mechanism claim.

assumptions (5)

domain assumption Each unit's propulsion direction is determined by its orientation and aligns with the local chain contour, realizing tangentially driven active polymers.
Main text: 'The arrangement of the units in the chain is designed such that the propulsion direction of each unit aligns with the contour of the chain.' This is the basis of the shape-motion feedback that underlies all subsequent claims; if propulsion direction were not contour-aligned, the chain dynamics would differ qualitatively.
domain assumption At the operating frequencies (1-150 kHz), self-dielectrophoresis (sDEP) is the dominant propulsion mechanism, with ICEP and EHD subdominant.
Established via tracer capture tests, bulk levitation experiments, and a 2D DC COMSOL simulation (Fig. S4). The tracer test shows no sustained flow but does not rule out weak ICEP; the COMSOL model uses DC fields and assumed material parameters. The dominance of sDEP is load-bearing for the claim that anisotropic shape alone drives propulsion.
ad hoc to paper The maximum bending angle set by hinge geometry (about 50 degrees) dictates the buckling conformation and yields the linear scaling f proportional to U.
Supplementary text: 'the maximum relative bending angle set by the connections dictates the buckling profile and the oscillation frequency thus scales simply with the velocity of the single unit.' This explanation is offered for the observed linear scaling, contrasting with the 3/4 exponent predicted by active polymer theory; it is not derived from a quantitative model.
ad hoc to paper The active force is always much larger than the thermal or dipolar bending resistance, so buckled conformations are set by geometry rather than by force balance.
Supplementary: 'the active forces are always much larger than the resistance to buckling, leading to the buckled conformation being determined by the maximally possible bending angle.' This justifies why the observed limit cycles are similar across voltages; it is asserted without a measured force comparison.
domain assumption The observed navigation behaviors (wall avoidance, obstacle turns, collision avoidance) arise from the embodied shape-motion feedback and not from passive mechanical redirection or pre-existing field gradients.
The paper interprets wall buckling, obstacle turns, and obstacle bypassing as autonomous sense-respond abilities. No control experiments with passive flexible chains or rigid active particles are provided to exclude purely mechanical explanations.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Autonomous life-like behavior emerging in active and flexible microstructures." pith.science (2026). https://pith.science/paper/UQ4ZW7ZR

@misc{pith2026250615198,
  author       = {Pith},
  title        = {Pith review of: Autonomous life-like behavior emerging in active and flexible microstructures},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UQ4ZW7ZR}},
  note         = {Machine review of arXiv:2506.15198}
}

read the original abstract

Many organisms leverage an interplay between shape and activity to generate motion and adapt to their environment. Embedding such feedback into synthetic microrobots could eliminate the need for sensors, software, and actuators, yet current realizations are either active but rigid, or flexible but passive. Here, we introduce micrometer-scale structures that integrate both activity and flexibility by 3D microprinting concatenated units and actuating them with an AC electric field. This minimal yet versatile design gives rise to a rich array of life-like modes of motion - including railway and undulatory locomotion, rotation, and beating - as well as emergent sense-response abilities, which enable autonomous reorientation, navigation, and collision avoidance. Our approach offers a versatile platform for designing biomimetic model systems and autonomously operating microrobots with embodied intelligence.

Figures

Figures reproduced from arXiv: 2506.15198 by the authors.

**Figure 1.** Active chains of concatenated micrometer-sized units. (A) Design and SEM micrograph of a 3D microprinted anisotropic unit. Scalebar is 2 µm. (B) Concatenated units can move within the geometrically allowed range. Schematic with unit orientation 𝜃& (top) and bright field microscopy image (middle) of two units with overlaid the center of mass motions with respect to the left unit, Scalebar 5 µm; bottom: probability de… view at source ↗

**Figure 2.** Autonomous switching between different emergent life-like modes of motion. (A) - (C) Free ‘railway’ motion: (A) Schematic, (B) Overlaid microscopy snapshots with trajectory of the leading unit (purple) and (C) all units showing ‘railway’ motion, confirmed by (D) the relative mean squared displacement. (E-I) Head-clamping induces self-oscillation: (E) Schematic and (F) Overlaid time-lapse snapshots of the beating cha… view at source ↗

**Figure 3.** Adaptive navigation in complex environments. (A, B) Upon encountering a wall (red), the chain buckles as measured by the mean local curvature ⟨|Δθ+|⟩ = # %(# ∑ |𝜃&'# %(# # − 𝜃&|, and reorients the head as captured by the orientation of the head with respect to the tangential direction of the wall |θ# − θ⟂| (Movie S7). (C) Active, flexible chains adapt their motion to ordered arrays of obstacles (spacing 25 µm) as vi… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

3 extracted references · 3 canonical work pages

[14]

Kumar, A

M. Kumar, A. Murali, A. G. Subramaniam, R. Singh, S. Thutupalli, Emergent dynamics due to chemo-hydrodynamic self-interactions in active polymers. Nature Communications 15, 4903 (2024). 15. J. Veenstra, C. Scheibner, M. Brandenbourger, J. Binysh, A. Souslov, V. Vitelli, C. Coulais, Adaptive locomotion of active solids. Nature 639, 935–941 (2025). 16. A. C...

work page 2024
[30]

R. E. Isele-Holder, J. Elgeti, G. Gompper, Self-propelled worm-like filaments: Spontaneous spiral formation, structure, and dynamics. Soft Matter 11, 7181–7190 (2015). 31. R. E. Isele-Holder, J. Jäger, G. Saggiorato, J. Elgeti, G. Gompper, Dynamics of self-propelled filaments pushing a load. Soft Matter 12, 8495–8505 (2016). 32. O. Chepizhko, F. Peruani, ...

work page 2015
[45]

(𝑡,𝜏)=⟨[𝑟#(𝜏)−𝑟&(𝜏+𝑡)]

T. M. Squires, M. Z. Bazant, Breaking symmetries in induced-charge electro-osmosis and electrophoresis. Journal of Fluid Mechanics 560, 65 (2006). 46. A. M. Brooks, S. Sabrina, K. J. M. Bishop, Shape-directed dynamics of active colloids powered by induced-charge electrophoresis. Proceedings of the National Academy of Sciences 115, E1090–E1099 (2018). 47. ...

work page 2006

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [14]

Kumar, A

M. Kumar, A. Murali, A. G. Subramaniam, R. Singh, S. Thutupalli, Emergent dynamics due to chemo-hydrodynamic self-interactions in active polymers. Nature Communications 15, 4903 (2024). 15. J. Veenstra, C. Scheibner, M. Brandenbourger, J. Binysh, A. Souslov, V. Vitelli, C. Coulais, Adaptive locomotion of active solids. Nature 639, 935–941 (2025). 16. A. C...

work page 2024

[2] [30]

R. E. Isele-Holder, J. Elgeti, G. Gompper, Self-propelled worm-like filaments: Spontaneous spiral formation, structure, and dynamics. Soft Matter 11, 7181–7190 (2015). 31. R. E. Isele-Holder, J. Jäger, G. Saggiorato, J. Elgeti, G. Gompper, Dynamics of self-propelled filaments pushing a load. Soft Matter 12, 8495–8505 (2016). 32. O. Chepizhko, F. Peruani, ...

work page 2015

[3] [45]

(𝑡,𝜏)=⟨[𝑟#(𝜏)−𝑟&(𝜏+𝑡)]

T. M. Squires, M. Z. Bazant, Breaking symmetries in induced-charge electro-osmosis and electrophoresis. Journal of Fluid Mechanics 560, 65 (2006). 46. A. M. Brooks, S. Sabrina, K. J. M. Bishop, Shape-directed dynamics of active colloids powered by induced-charge electrophoresis. Proceedings of the National Academy of Sciences 115, E1090–E1099 (2018). 47. ...

work page 2006