Pith. sign in

REVIEW 2 major objections 5 minor 66 references

Longer bacterial chains swim faster because faster cells preferentially join them, a self-sorting process the authors call swimming-limited aggregation.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-11 07:33 UTC pith:X674HAR6

load-bearing objection Clear experimental observation of faster longer chains, explained by relative-velocity self-sorting; quantitative match is a one-parameter consistency check under a dry model, not an independent prediction. the 2 major comments →

arxiv 2607.05239 v1 pith:X674HAR6 submitted 2026-07-06 cond-mat.soft physics.bio-ph

Swimming-limited aggregation of bacteria in liquid crystals

Guillaume Sint\`es , Martyna Goral , Teresa L\'opez-Le\'on , Anke Lindner , Maria T\u{a}tulea-Codrean This is my paper
classification cond-mat.soft physics.bio-ph
keywords swimming-limited aggregationbacterial chainsnematic liquid crystalsself-sortingspeed varianceE. colione-dimensional aggregationactive matter
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Escherichia coli swimming in one-dimensional lanes inside a nematic liquid crystal spontaneously form long-lived chains. Force balance alone would make every chain swim at the average speed of its members, yet experiments show that longer chains are systematically faster. The authors resolve the paradox by showing that chains are not random assemblies: cell-to-cell speed differences make faster bacteria catch slower ones first, so the fastest cells are preferentially incorporated into longer chains. A minimal three-body encounter model and agent-based simulations of thousands of particles with log-normally distributed speeds reproduce the measured length–speed curve and its dependence on speed variance. The same framework predicts that the correlation weakens as the system ages. The result identifies a distinct aggregation regime whose kinetics and statistics are controlled by the intrinsic speed distribution of the swimmers rather than by diffusion or elastic attraction.

Core claim

Longer chains of E. coli swimming along the director of a nematic liquid crystal move systematically faster than shorter ones because of a dynamic self-sorting process: relative velocities cause faster bacteria to meet and join chains earlier, so longer chains become enriched in high-speed cells. This swimming-limited aggregation is constrained by the variance of the individual speed distribution and becomes weaker with time.

What carries the argument

Swimming-limited aggregation: a one-dimensional process in which irreversible chain formation is driven solely by relative swimming speeds (sampled from a lognormal distribution) under a dry force-balance rule that sets every new chain’s speed to the arithmetic mean of its constituents.

Load-bearing premise

The model assumes every bacterium has the same drag coefficient and that flagella rearrange instantly after collisions, so chain speed is always exactly the average of the members’ intrinsic speeds.

What would settle it

Measure the length–speed curve of chains formed from a genetically or environmentally prepared population whose swimming-speed variance is much narrower than the wild-type distribution used here; if the positive correlation vanishes or becomes non-monotonic while mean speed remains unchanged, the self-sorting claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The manuscript reports that motile E. coli in nematic liquid crystals form long-lived 1D chains whose average swimming speed increases with chain length n. A single merging event is consistent with a dry force-balance model in which chain speed equals the arithmetic mean of precursor speeds (Fig. 1d; Appendix E, Eqs. 3–5). The authors attribute the positive length–speed correlation to a dynamic self-sorting process driven by cell-to-cell speed variability: faster bacteria encounter neighbours sooner and are preferentially incorporated into longer chains. This is supported by a minimal three-body encounter model (Eqs. 1–2, Fig. 2) and by agent-based simulations of irreversible 1D aggregation that, after estimating mean speed V from the experimental weighted average and fitting one free parameter σ_V at the simulation time t* where ⟨n⟩ matches experiment (≈2.2), reproduce both the mean speed–length curve and the shape of the speed distributions (Fig. 1e–f). The authors term the process “swimming-limited aggregation” and contrast its time scale with passive LC-mediated aggregation.

Significance. If the self-sorting interpretation holds, the work identifies a clean, experimentally accessible regime of active aggregation whose kinetics and emergent length–speed correlation are controlled by the intrinsic speed distribution of the agents rather than by reaction rates or elastic attractions. The combination of large trajectory statistics, a transparent dry kinematic rule, a minimal analytic encounter model, and publicly released simulation code is a genuine strength and makes the mechanism falsifiable in related confined active systems. The result is of clear interest to soft-matter and active-matter communities studying motility-induced organization in anisotropic media.

major comments (2)
  1. §IV and Fig. 1e–f: The quantitative match is obtained by estimating V from the already-aggregated experimental dataset (conserved under the dry model) and fitting the single free parameter σ_V so that the simulated speed–length curve at the dynamical age t* (defined by matching experimental ⟨n⟩≈2.2) reproduces experiment. The paper itself notes that the solitary-bacterium histogram in Fig. 1f is residual after faster cells have joined chains and that experimental t=0 is ill-defined. The agreement is therefore a consistency check under the dry-model assumptions rather than an independent prediction of the sorting mechanism. An independent measurement (or at least a robust bound) of the pre-aggregation speed distribution, or an explicit demonstration that the qualitative positive correlation survives without fitting σ_V to the same curve, is needed to secure the central claim.
  2. Appendix E, Eqs. (3)–(5) and §III.B: The dry mechanical coupling model forces chain speed to equal the arithmetic mean of constituents by assuming identical drag coefficients, instantaneous flagellar rearrangement after head-on collisions, and complete neglect of hydrodynamics and LC elasticity once a chain has formed. These assumptions are used both to interpret the single merging event (Fig. 1d) and to drive the simulations that are then fitted to experiment. The manuscript acknowledges that hydrodynamics would likely increase the slope of the speed–length curve (and therefore shift the matching time t*). Because any systematic n-dependent enhancement of chain speed would produce a positive correlation even without self-sorting, the paper should either (i) provide a quantitative estimate of the hydrodynamic/elastic correction or (ii) show that the observed correlation cannot be explain
minor comments (5)
  1. Fig. 1e inset and §II: The observation count drops rapidly with n; error bars on the longest chains are large. A brief statement of the minimum sample size used for the mean-speed points would help the reader assess robustness.
  2. §III.A, Eq. (1): The notation for the cumulative distribution F_V and the absolute-value arguments is dense; a short sentence defining the two bracketed factors would improve readability.
  3. Fig. 3(d) inset: The non-monotonic speed–length relation reported for σ_V/V=1/3 is interesting but only briefly discussed; a sentence on whether this feature is expected to appear in experiment would be useful.
  4. Appendix B: The passive-aggregation estimate relies on an anchoring energy taken from B. subtilis; a short caveat that W has not been measured for E. coli would be appropriate.
  5. Code availability is excellent (GitHub links for both image analysis and simulations). Adding a short README note on the exact random-seed protocol used for the 50-run averages would further aid reproducibility.

Circularity Check

2 steps flagged

The quantitative speed–length match is a one-parameter fit of σ_V at a matched dynamical age, not an independent prediction; the qualitative self-sorting claim remains non-circular.

specific steps
  1. fitted input called prediction [§IV (Comparison of Experimental and Theoretical Results); Fig. 1(e) and caption]
    "The standard deviation σ_V is the only remaining free parameter, which we determine through a one-parameter fit of the theoretical model evaluated at time t* to the experimental data in Fig. 1(e). The best fit is obtained for σ_V ≈ 7.26 µm s−1 ... The excellent agreement between the average chain speed predicted by the numerical simulations and the one observed experimentally, shown in Fig. 1(e), is a direct consequence of our approach to finding σ_V."

    σ_V is adjusted so that the simulated average speed-versus-length curve at the age-matched time t* (defined by matching experimental ⟨n⟩≈2.2) reproduces Fig. 1(e). The reported ‘agreement’ on that curve is therefore forced by construction of the fit, not an independent prediction of the sorting mechanism. V itself is also taken from the post-aggregation experimental weighted average under the dry-model conservation assumption, so the initial speed distribution is not measured independently before aggregation.

  2. fitted input called prediction [§IV; Fig. 1(f) and abstract claim of consistency]
    "Although it uses only a single fitting parameter to match the average speed of bacterial chains, the theoretical model captures the shape of the full speed distributions remarkably well, as shown in Fig. 1(f). ... Consistent with experimental observations, our agent-based simulations reveal a positive correlation between the length and speed of dynamically self-assembled chains"

    After σ_V is fixed to the mean speed–length curve, the same simulations are presented as capturing the full n=1,3,5 speed distributions and as ‘consistent with’ the positive length–speed correlation. The distributions are a weaker, partially independent check (same parameter, related observables), but the abstract/figure framing still treats the overall theory–experiment match as confirmatory of swimming-limited aggregation when the primary quantitative curve was the fitting target. The paper also notes that the experimental solitary histogram is the residual after faster cells have already chained, so it is not an independent pre-aggregation input.

full rationale

The paper’s central qualitative claim—that longer chains are faster because relative-velocity self-sorting preferentially places fast bacteria in long chains—is not circular. Under the dry model (Appendix E), chain speed is exactly the arithmetic mean of constituents, so any positive length–speed correlation must be compositional; the minimal nearest-neighbour model and agent-based simulations then show that faster agents encounter neighbours sooner and are enriched in longer chains. That logic is independent of the experimental fit. However, the main quantitative comparison in Fig. 1(e) is not an independent prediction: V is taken from the already-aggregated experimental weighted average (justified by conservation under the dry model), the observation time t* is chosen so that simulated ⟨n⟩ matches experiment (≈2.2), and the single free parameter σ_V is fitted so that the simulated speed–length curve at that t* matches Fig. 1(e). The paper itself states that the excellent agreement on averages is a direct consequence of this fitting procedure. The residual solitary-bacterium histogram is acknowledged not to be the pre-aggregation distribution. The full speed distributions (Fig. 1f) provide a partial out-of-sample check with the same fitted σ_V, so the circularity is partial rather than total. Score 5 reflects one clear fitted-input-called-prediction step on the load-bearing quantitative curve, with the qualitative mechanism and distribution shapes retaining independent content.

Axiom & Free-Parameter Ledger

2 free parameters · 5 axioms · 1 invented entities

The central claim rests on one fitted width of the speed distribution, the dry force-balance idealization that forces chain speed to equal the arithmetic mean of constituents, the assumption of irreversible aggregation without fragmentation or speed fluctuations, and the identification of dynamical age via mean chain length. No new physical entities are postulated; the mechanism is built from standard active-matter and liquid-crystal ingredients plus the sorting dynamics that follow from relative velocities.

free parameters (2)
  • σ_V (standard deviation of individual bacterial speeds) = ≈7.26 µm s⁻¹
    Single free parameter fitted to the experimental average speed-versus-length curve at the simulation time t* defined by matching experimental mean chain length; value ≈7.26 µm s⁻¹.
  • V (mean individual speed) = ≈9.18 µm s⁻¹
    Estimated directly from the experimental weighted average across all tracked objects under the conservation property of the dry model; used as the scale for the lognormal initialization.
axioms (5)
  • ad hoc to paper Chain swimming speed equals the arithmetic mean of the intrinsic speeds of its constituent bacteria (dry mechanical coupling, identical drag coefficients).
    Appendix E, Eqs. (3)–(5); used both to interpret the single merging event and as the kinematic rule of the agent-based simulations.
  • domain assumption Aggregation is irreversible; no fragmentation occurs on experimental time scales.
    Stated throughout §III; elastic attraction is assumed to stabilize chains permanently once formed.
  • domain assumption Bacterial speeds are drawn once from a lognormal distribution and remain constant until a merging event.
    §III.B; fluctuations and run-and-tumble reversals are neglected.
  • ad hoc to paper Hydrodynamic interactions and liquid-crystal elasticity can be neglected for the kinematics of already-formed chains.
    Explicitly stated as the “dry” approximation in §III and Appendix E.
  • ad hoc to paper Dynamical age of the experimental system can be identified with the simulation time at which mean chain length equals the experimental value ⟨n⟩≈2.2.
    §IV; used to select the comparison time t*.
invented entities (1)
  • swimming-limited aggregation regime no independent evidence
    purpose: Name the distinct aggregation process whose rate and composition are set by the intrinsic speed distribution of active agents rather than by diffusion or reaction kinetics.
    Introduced in the abstract and §I; the entity is a descriptive label for the observed dynamics rather than a new physical particle or force.

pith-pipeline@v1.1.0-grok45 · 28082 in / 3071 out tokens · 21488 ms · 2026-07-11T07:33:10.197052+00:00 · methodology

0 comments
read the original abstract

Aggregation and fragmentation processes are widespread in engineering and the natural world. Here, we investigate a distinct colloidal aggregation mechanism in an active system of motile bacteria in highly anisotropic environments. Specifically, we examine \textit{Escherichia coli} bacteria swimming in one-dimensional confinement within nematic liquid crystals and observe long-lived chains of bacteria swimming along the nematic director. Crucially, we find that longer chains swim faster, in apparent contradiction to fundamental force-balance models that predict the swimming speed to be independent of chain length, as chains should swim at the average speed of their individual components. The seeming discrepancy is resolved by recognizing that chains do not form randomly but self-organize due to the relative velocities between bacteria. To elucidate the physical mechanism behind this active aggregation process, we combine our experimental findings with a minimal model of nearest-neighbour aggregation and agent-based simulations of active particles aggregating in one dimension. Consistent with experimental observations, our agent-based simulations reveal a positive correlation between the length and speed of dynamically self-assembled chains of active particles, with the correlation depending on the variance of the individual speed distribution and diminishing over time. Together, our experiments and theoretical models indicate a distinct regime of swimming-limited aggregation whose evolution is constrained by the intrinsic speed distribution of active agents, providing new insight into bacterial self-organization.

Figures

Figures reproduced from arXiv: 2607.05239 by Anke Lindner, Guillaume Sint\`es, Maria T\u{a}tulea-Codrean, Martyna Goral, Teresa L\'opez-Le\'on.

Figure 1
Figure 1. Figure 1: FIG. 1. (Colour online) [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. (Colour online) [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. (Colour online) [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. (continued) [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

66 extracted references

  1. [1]

    L. R. Middleton and K. I. Winey, Nanoscale aggregation in acid- and ion-containing polymers, Annu. Rev. Chem. Biomol. Eng.8, 499 (2017)

    2017

  2. [2]

    P. J. Lu and D. A. Weitz, Colloidal particles: Crystals, glasses, and gels, Annu. Rev. Cond. Matt. Phys.4, 217 (2013)

    2013

  3. [3]

    Bourouiba, Fluid dynamics of respiratory infectious diseases, Annu

    L. Bourouiba, Fluid dynamics of respiratory infectious diseases, Annu. Rev. Biomed. Eng.23, 547 (2021)

    2021

  4. [4]

    Villermaux, Fragmentation, Annu

    E. Villermaux, Fragmentation, Annu. Rev. Fluid Mech.39, 419 (2007)

    2007

  5. [5]

    A. B. Burd and G. A. Jackson, Particle aggregation, Annu. Rev. Mar. Sci.1, 65 (2009)

    2009

  6. [6]

    Pumir and M

    A. Pumir and M. Wilkinson, Collisional aggregation due to turbulence, Annu. Rev. Cond. Matt. Phys.7, 141 (2016)

    2016

  7. [7]

    D. A. Weitz, J. S. Huang, M. Y. Lin, and J. Sung, Dynamics of diffusion-limited kinetic aggregation, Phys. Rev. Lett.53, 1657 (1984)

    1984

  8. [8]

    Family, P

    F. Family, P. Meakin, and T. Vicsek, Cluster size distribution in chemically controlled cluster–cluster aggregation, J. Chem. Phys.83, 4144 (1985)

    1985

  9. [9]

    Meakin and F

    P. Meakin and F. Family, Structure and dynamics of reaction-limited aggregation, Phys. Rev. A 36, 5498 (1987). 16

    1987

  10. [10]

    Meakin and F

    P. Meakin and F. Family, Structure and kinetics of reaction-limited aggregation, Phys. Rev. A38, 2110 (1988)

    1988

  11. [11]

    L. Gey, P. Le Gal, and G. Verhille, Formation and fragmentation of nylon fibre aggregates, Proc. R. Soc. A481, 20250246 (2025)

    2025

  12. [12]

    Borgnino, J

    M. Borgnino, J. Arrieta, G. Boffetta, F. De Lillo, and I. Tuval, Turbulence induces clustering and segregation of non-motile, buoyancy-regulating phytoplankton, J. R. Soc. Interface16, 20190324 (2019)

    2019

  13. [13]

    Y. Z. Sinzato, R. Uittenbogaard, P. M. Visser, J. Huisman, and M. Jalaal, Fragmentation and aggregation of cyanobacterial colonies, eLife14, RP103503 (2025)

    2025

  14. [14]

    Muˇ seviˇ c and M.ˇSkarabot, Self-assembly of nematic colloids, Soft Matter4, 195 (2008)

    I. Muˇ seviˇ c and M.ˇSkarabot, Self-assembly of nematic colloids, Soft Matter4, 195 (2008)

    2008

  15. [15]

    Muˇ seviˇ c, Nematic liquid-crystal colloids, Materials11, 24 (2018)

    I. Muˇ seviˇ c, Nematic liquid-crystal colloids, Materials11, 24 (2018)

    2018

  16. [16]

    Grober, I

    D. Grober, I. Palaia, M. C. U¸ car, E. Hannezo, A. ˇSari´ c, and J. Palacci, Unconventional colloidal aggregation in chiral bacterial baths, Nat. Phys.19, 1680 (2023)

    2023

  17. [17]

    Maier, How physical interactions shape bacterial biofilms, Annu

    B. Maier, How physical interactions shape bacterial biofilms, Annu. Rev. Biophys.50, 401 (2021)

    2021

  18. [18]

    Savorana, T

    G. Savorana, T. Redaelli, D. Truzzolillo, L. Cipelletti, and E. Secchi, Stress-hardening behaviour of biofilm streamers, Nat. Commun.16, 9497 (2025)

    2025

  19. [19]

    David and S

    J. David and S. Thutupalli, Explosive dispersal of non-motile microbes through metabolic buoyancy (2025), arXiv:2512.16288 [cond-mat.soft]

    arXiv 2025

  20. [20]

    M. E. Black, C. Fei, R. Alert, N. S. Wingreen, and J. W. Shaevitz, Capillary interactions drive the self-organization of bacterial colonies, Nat. Phys.21, 1444–1450 (2025)

    2025

  21. [21]

    Figueroa-Morales, L

    N. Figueroa-Morales, L. Dominguez-Rubio, T. L. Ott, and I. S. Aranson, Mechanical shear controls bacterial penetration in mucus, Sci. Rep.9, 9713 (2019)

    2019

  22. [22]

    Repula, E

    A. Repula, E. Abraham, V. Cherpak, and I. I. Smalyukh, Biotropic liquid crystal phase transforma- tions in cellulose-producing bacterial communities, Proc. Natl. Acad. Sci. U.S.A.119, e2200930119 (2022)

    2022

  23. [23]

    S. Zhou, A. Sokolov, O. D. Lavrentovich, and I. S. Aranson, Living liquid crystals, Proc. Natl. Acad. Sci. U.S.A.111, 1265 (2014)

    2014

  24. [24]

    H. Baza, F. Chen, T. Turiv, S. V. Shiyanovskii, and O. D. Lavrentovich, Bend instabilities and topological turbulence in shear-aligned living liquid crystal, Soft Matter22, 1297 (2026)

    2026

  25. [25]

    P. C. Mushenheim, R. R. Trivedi, H. H. Tuson, D. B. Weibel, and N. L. Abbott, Dynamic self- assembly of motile bacteria in liquid crystals, Soft Matter10, 88 (2014)

    2014

  26. [26]

    Goral, E

    M. Goral, E. Clement, T. Darnige, T. Lopez-Leon, and A. Lindner, Frustrated ‘run and tumble’ of swimmingEscherichia colibacteria in nematic liquid crystals, Interface Focus12, 20220039 (2022)

    2022

  27. [27]

    A. G. Prabhune, A. S. Garc´ ıa-Gordillo, I. S. Aranson, T. R. Powers, and N. Figueroa-Morales, Bacteria navigate anisotropic media using a flagellar tug-of-oars, PRX Life2, 033004 (2024)

    2024

  28. [28]

    Cheon, J

    J. Cheon, J. Son, S. Lim, Y. Jeong, J.-H. Park, R. J. Mitchell, J. U. Kim, and J. Jeong, Motile bacteria crossing liquid–liquid interfaces of an aqueous isotropic–nematic coexistence phase, Soft Matter20, 7313 (2024)

    2024

  29. [29]

    Z. Lin, Y. Wang, Y. Quan, Z. Yu, T. Gao, and S. Chen, Microorganisms swimming in lyotropic liquid crystal polymers near a wall, Phys. Rev. Fluids10, 083302 (2025)

    2025

  30. [30]

    Y. Quan, Y. Xia, Z. Ouyang, Z. Yu, J. Lin, and Z. Lin, Microorganisms swimming in nematic liquid crystal at finite Reynolds numbers, Phys. Fluids38, 021901 (2026)

    2026

  31. [31]

    S. G. La Corte, T. G. J. Chandler, S. E. Spagnolie, N. S. Wingreen, and S. S. Datta, Morphogenesis of bacterial colonies in liquid crystalline environments (2025), arXiv:2512.12406 [cond-mat.soft]

    arXiv 2025

  32. [32]

    Berg and D

    H. Berg and D. Brown, Chemotaxis inEscherichia colianalysed by three-dimensional tracking, Nature239, 500 (1972)

    1972

  33. [33]

    Volovik and O

    G. Volovik and O. Lavrentovich, Topological dynamics of defects: Boojums in nematic drops, J. Exp. Theor. Phys.85, 1997 (1983)

    1997

  34. [34]

    E. M. Terentjev, Disclination loops, standing alone and around solid particles, in nematic liquid crystals, Phys. Rev. E51, 1330 (1995)

    1995

  35. [35]

    Masuda and T

    N. Masuda and T. Hiraoka, Waiting-time paradox in 1922, Northeast J. Complex Syst.2, 1 (2020)

    1922

  36. [36]

    E. M. Purcell, Life at low Reynolds number, Am. J. Phys.45, 3 (1977)

    1977

  37. [37]

    G. F. Carnevale, Y. Pomeau, and W. R. Young, Statistics of ballistic agglomeration, Phys. Rev. Lett.64, 2913 (1990)

    1990

  38. [38]

    Piasecki, Universal distribution of masses in a one-dimensional sticky gas, Physica A190, 95 (1992)

    J. Piasecki, Universal distribution of masses in a one-dimensional sticky gas, Physica A190, 95 (1992)

    1992

  39. [39]

    Frachebourg, Exact solution of the one-dimensional ballistic aggregation, Phys

    L. Frachebourg, Exact solution of the one-dimensional ballistic aggregation, Phys. Rev. Lett.82, 1502 (1999)

    1999

  40. [40]

    Parrenin, S

    A. Parrenin, S. Kooij, C. J. M. van Rijn, and D. Bonn, Effect of coalescence on the propagation of 17 water droplets from a jet, Phys. Fluids36, 103323 (2024)

    2024

  41. [41]

    Vani and D

    N. Vani and D. Bonn (unpublished)

  42. [42]

    Tailleur and M

    J. Tailleur and M. E. Cates, Statistical mechanics of interacting run-and-tumble bacteria, Phys. Rev. Lett.100, 218103 (2008)

    2008

  43. [43]

    M. E. Cates and J. Tailleur, Motility-induced phase separation, Annu. Rev. Cond. Matt. Phys.6, 219 (2015)

    2015

  44. [44]

    J. S. Parkinson, Complementation analysis and deletion mapping ofEscherichia colimutants de- fective in chemotaxis, J. Bacteriol.135, 45 (1978)

    1978

  45. [45]

    P. Y. Yap and D. Trau, DirectE. colicell count at OD600,Application Note. Tip Biosystems Pte Ltd, Singapore(2019)

    2019

  46. [46]

    Junot, T

    G. Junot, T. Darnige, A. Lindner, V. A. Martinez, J. Arlt, A. Dawson, W. C. K. Poon, H. Auradou, and E. Cl´ ement, Run-to-tumble variability controls the surface residence times ofE. colibacteria, Phys. Rev. Lett.128, 248101 (2022)

    2022

  47. [47]

    Suh and D

    A. Suh and D. K. Yoon, Nanoscratching technique for highly oriented liquid crystal materials, Sci. Rep.8, 9460 (2018)

    2018

  48. [48]

    Darnige, N

    T. Darnige, N. Figueroa-Morales, P. Bohec, A. Lindner, and E. Cl´ ement, Lagrangian 3D tracking of fluorescent microscopic objects in motion, Rev. Sci. Instrum.88, 055106 (2017)

    2017

  49. [49]

    Darnige, D

    T. Darnige, D. Midtvedt, R. Baillou, B. P. Estay, C. Wu, A. L. Guen, G. Volpe, and E. Clement, Deep learning-enhanced Lagrangian 3D Tracking of motile microorganisms (2026), arXiv:2603.20764 [physics.bio-ph]

    arXiv 2026

  50. [50]

    Sint` es,Dynamic self-assembly of bacteria in liquid crystals, PhD thesis, Universit´ e Paris Sciences et Lettres (2024)

    G. Sint` es,Dynamic self-assembly of bacteria in liquid crystals, PhD thesis, Universit´ e Paris Sciences et Lettres (2024)

    2024

  51. [51]

    Habibi, C

    A. Habibi, C. Blanc, N. B. Mbarek, and T. Soltani, Passive and active microrheology of a lyotropic chromonic nematic liquid crystal disodium cromoglycate, J. Mol. Liq.288, 111027 (2019)

    2019

  52. [52]

    S. Zhou, K. Neupane, Y. A. Nastishin, A. R. Baldwin, S. V. Shiyanovskii, O. D. Lavrentovich, and S. Sprunt, Elasticity, viscosity, and orientational fluctuations of a lyotropic chromonic nematic liquid crystal disodium cromoglycate, Soft Matter10, 6571 (2014)

    2014

  53. [53]

    E. E. Riley, D. Das, and E. Lauga, Swimming of peritrichous bacteria is enabled by an elastohydro- dynamic instability, Sci. Rep.8, 10728 (2018). 18 Appendix A: Materials and Methods

    2018

  54. [54]

    super-bundle

    Bacterial Strains and Culture a. E. coli RP437 For the majority of tracking experiments, we used theEscherichia coli(E. coli) strain RP437, a wild-type strain with run-and-tumble motility derived from the K12 strain [44]. The RP437 strain carries a plasmid expressing yellow fluorescent protein (YFP) that makes the bacterial cell body fluorescent with an e...

  55. [55]

    To preserve bacterial motility, the DSCG molecules were dispersed in MB instead of water

    Liquid Crystal Sample Preparation We used the biocompatible liquid crystal Cromolyn sodium salt (DSCG), purchased from Sigma-Aldrich. To preserve bacterial motility, the DSCG molecules were dispersed in MB instead of water. The solution was placed on the vortex shaker for around 15 min until fully dissolved, then transferred to an ultrasonic bath for 30 m...

  56. [56]

    For experiments using the long working distance 40×objective, the chamber was created by sandwiching two scratched glass slides

    Aligning Chamber Preparation The chambers were constructed in a sandwich configuration using either two glass slides or a glass slide and a cover slip, separated by a Parafilm spacer. For experiments using the long working distance 40×objective, the chamber was created by sandwiching two scratched glass slides. When the 63×objective was required for flage...

  57. [57]

    Images were captured with a Hamamatsu Orca Flash 4.0 camera, using a 40×long working distance objective

    Fluorescence Microscopy The experiments were conducted using a Zeiss inverted microscope equipped with a control- lable motorized stage and a fluorescent lamp fitted with a cyan colour filter at a wavelength of 511 nm. Images were captured with a Hamamatsu Orca Flash 4.0 camera, using a 40×long working distance objective. For fixed swimming experiments, i...

  58. [58]

    super-bundle

    3D Tracking We used a 3D Lagrangian tracking device developed in the PMMH laboratory by Darnige et al.[48] to visualize the “super-bundle” configuration of flagella over long periods (Fig. 1(g) and Fig. S2) and to capture rare merging events such as the one depicted in Fig. 1(d). The device integrates a 3-axis motorized microscope stage, a Hamamatsu Orca ...

  59. [59]

    Further details can be found in the doctoral dissertation of Sint` es [50]

    Image Analysis We provide a brief description of the image analysis pipeline for detecting and tracking bacterial chains. Further details can be found in the doctoral dissertation of Sint` es [50]. a. Chain Detection The background signal was calculated as the minimum intensity value of each pixel over the full duration of the experimental video and was s...

  60. [60]

    lane” in the nematic LC (see Supp. Fig. S1). By partitioning a 1 mm square into 1000 “lanes

    Thus, when the particles start closer than ten radii apart,d 0 <10R, the encounter occurs within approximately 6 min. However, the encounter takes longer than 1 h whend 0 >14R. This suggests that only neighbouring bacteria could be passively aggregated by the LC on experimentally accessible time scales of less than an hour. To estimate the typical distanc...

  61. [61]

    Relative Velocity Distribution The setup of the minimal model of aggregation between three isolated bacteria is depicted in Fig. 2(a). The central bacterium swims forward at a given speed,v c, while the signsS r,f indicate the swimming directions (+1 forward, -1 backward) of its rear and front neighbours, andV r,f indicate their absolute speeds. The appar...

  62. [62]

    The central bacterium will encounter its rear or front neighbours only if ˆVr <0 or ˆVf >0, respectively

    First Encounter Time Distribution Having described the probability of observing a certain range of relative velocities, we now consider the expected encounter time between the central bacterium and its nearest neighbours. The central bacterium will encounter its rear or front neighbours only if ˆVr <0 or ˆVf >0, respectively. The apparent time of these en...

  63. [63]

    Infinite Encounter Time Probability Naturally, it is also possible that the central bacterium never meets either its front or rear neighbours if ˆVr >0 and ˆVf <0. The probability of infinite encounter time is given by p(Te = +∞) =p ˆVr ≥0 p ˆVf ≤0 ,(D12) = 1 2 [1−sgn(−v c)FV (vc)]× 1 2 [1 + sgn(−vc)FV (vc)],(D13) = 1 4 1−F V (vc)2 .(D14) In particular, t...

  64. [64]

    Speed Selection for Individual Bacteria The velocity of a freely swimming bacterium emerges from the balance of thrust and drag on its body and flagella [36]. Let us assume that an isolated bacterium generates a total thrustF by rotating its flagella and experiences a net viscous drag−ζVfrom both its body and flagella as it moves forward through a viscous...

  65. [65]

    Speed Selection for Bacterial Chains A swimming bacterial chain is itself a force-free swimmer since no external force is applied to the chain. Therefore, the total thrust generated by all flagella must balance the total viscous drag on the bodies and flagella of all bacteria in the chain, nX k=1 F k − nX k=1 ζkV k =0.(E1) The dry mechanical coupling mode...

  66. [66]

    Selection of Swimming Direction First, we note the distinction between rear-end and head-on collisions depicted in Fig. 3(a). In rear-end collisions, the precursor agents swim in the same direction with sgn( eV1) = sgn(eV2), and Eqs. (E3)-(E4) trivially reduce to Eqs. (3)-(4). In head-on collisions, we postulate that the new swimming direction is determin...