Skip to content
PaperFren

Can an algorithm find the equations behind active nematic flows?

Open paper intelligence

A physics-constrained regression algorithm, fed with movies of motor-driven microtubules, recovered simple flow equations in which active stress balances viscous friction with no elastic term, differing from standard models.

Source

Physically informed data-driven modeling of active nematics

Golden M, Grigoriev RO, Nambisan J, et al. · Science advances · 2023

doi.org/10.1126/sciadv.abq6120Read the full paper ↗14 citationscc by

Study at a glance

Design
Computational / modelling — Data-driven model discovery (SPIDER: weak-form sparse symbolic regression constrained by symmetry) applied to director and velocity fields extracted from fluorescence movies of a microtubule-kinesin active nematic at an oil-water interface.
N
No sample size; the analysis uses spatiotemporal fields from experimental movies, excluding low-density regions near topological defects.
Population
Quasi-2D microtubule bundle suspension driven by kinesin motors at an oil-water interface
Outcome
Identified partial differential equations, their coefficients and residuals; predicted characteristic length scale versus measured defect spacing

Structured fields used in claim comparison tables when every cited study has a complete layer.

What they did

The authors extracted the microtubule orientation (director) field and flow velocity from fluorescence images of a microtubule suspension spread at an oil-water interface. They built symmetry-sorted libraries of candidate terms from these fields and their derivatives, converted the equations to a noise-robust weak form, and used sparse regression to find the fewest terms that fit. Regions near topological defects, where microtubule density is low and data are unreliable, were masked out.

What they found

The algorithm found nine relations that all follow from three: an incompressibility condition, a director evolution equation matching the Leslie-Ericksen model with coefficients near ±1, and a local balance between active and highly anisotropic viscous stress. No elastic or free-energy terms were detected, and the incompressibility relation had a residual of about 4%. Building on this, a force-balance argument predicted a characteristic length of about 270 μm, close to the measured spacing of about 240 μm between same-charge defects.

The limits

What it doesn't show

The model only describes regions of high, uniform microtubule density and low curvature; the defect neighbourhoods that actually control the dynamics were excluded, so elasticity may still matter there. It is based on one experimental system and geometry, and the length-scale estimate uses single-microtubule material values because bundle values are unmeasured. The two-dimensional 'stresses' are projections of three-dimensional stresses in the surrounding fluid layers, not stresses in the usual sense.

Key terms

Active nematic
A system of elongated, head-tail symmetric units that consume energy to generate stresses and flows, such as motor-driven microtubule bundles.
Director field
The local average orientation of the elongated units, defined up to a sign (n and −n are equivalent).
Topological defect
A point where the director field cannot be defined smoothly, such as ±1/2 defects in 2D nematics.
Sparse regression
A fitting method that selects the smallest set of candidate terms that still describes the data accurately.
Weak form
A version of a differential equation integrated against smooth weight functions, which reduces sensitivity to noise in derivatives.

Flashcards

1 / 10

0 of 10 answers reviewed

Research intelligence for this paper

See its role on concept claims, tensions it is part of, placement history, and related discoveries.

Open paper intelligence

Quiz yourself

1 / 5

What kind of output does SPIDER produce?

Common questions

How is SPIDER different from training a neural network on the flows?

It outputs interpretable partial differential equations that can be compared directly with existing theories, rather than a black-box predictor.

What was surprising about the discovered model?

The flow is set by a local balance between active and viscous stresses with no elastic contribution, whereas standard models usually balance active and elastic stresses.

Why were regions near defects excluded?

Microtubule density drops near defects and the images cannot resolve director or velocity reliably there.

More on Active matter