Publications

Du, M., Floyd, C., Jasrasaria, D., Vaikuntanathan, S., "Coherent advantage in the computational expressivity of excitionic networks," arXiv, 2026. [article] 

Floyd, C.*,  Manuel Lopez Rios, H.*, Dinner, A., Vaikuntanathan, S., "In-context learning emerges in chemical reaction networks without attention," arXiv, 2026. [article] 

Floyd, C.,  Dinner, A., Vaikuntanathan, S., "Local imperfect feedback control in nonequilibrium biophysical systems enabled by thermodynamic constraints," arXiv, 2025. [article] 

Lamtyugina, A., Behera, A., Nandy, A., Floyd, C., Vaikuntanathan, S., "Score-based generative diffusion with 'active' correlated noise sources," arXiv, 2024. [article] 

Publications

2026

[16] Floyd, C., Dinner, A., Vaikuntanathan, S., "Learning to guide non-equilibrium dynamics using local contrastive updates," Newton, 2026. [article] 

In this work we show that open-loop control policies for guiding dynamics across a range of non-equilibrium processes can be learned by leveraging "approximate" gradient updates. These updates are spatiotemporally local and hence do not require a global controller, but are imperfect in the sense that they only partially overlap with the optimal gradient direction. Nonetheless, this approach is sufficient to learn control policies in systems including active tissue models, suggesting that biologically feasible control strategies may be built from only locally accessible system measurements.

[15] Lannan, J., Floyd, C., Xu, L., Thompson, P., Yan, C., Marshall, W., Vaikuntanathan, S., Dinner, A., Bhamla, S., Elting, M., "A centrin-Sfi1 myoneme fishnet powers ultrafast calcium-triggered contraction in the giant ciliate Spirostomum ambiguum," PNAS, 2026. [article] 

[14] Lei, X.*,  Floyd, C.*, Cassas-Ferrer, L., Chakrabortty, T., Chandrasekaran, N., Dinner, A., Coyle, S., Honts, J., Bhamla, S., "Light-induced assembly and repeatable actuation in Ca2+-driven chemomechanical protein networks," Nature Communications, 2026. (Editor's pick) [article] 

This work, done in close collaboration with the Bhamla group, introduces a new experimental and modeling platform for light-controllable contractile protein networks based on the myonemal protein apparatus found in various protists. Using a new active elasticity theory that incorporates strong density inhomogeneities, we quantitatively capture the dynamics of these reconstituted driven materials, including a striking reversal of contractile direction upon repeated light stimulation, an unexpected behavior that offers new insight into how living materials can be programmed through external cues.

2025

[13] Floyd, C., Dinner, A., Murugan, A., Vaikuntanathan, S., "Limits on the computational expressivity of non-equilibrium biophysical processes," Nature Communications, 2025. [article]

This work asks a basic question: how powerful are biochemical networks as computational devices? Treating biochemical networks as Markov jump processes and training them to perform classification tasks, we uncover surprising and universal limitations on their computational expressivity, limitations rooted in a new non-equilibrium thermodynamic constraint we derive. Strikingly, these constraints can be overcome by a mechanism we call input multiplicity, a feature naturally realized when enzymes act on multiple targets. Much like how increasing the depth of a neural network expands its expressive power, tuning input multiplicity can yield an exponential increase in a cell's ability to classify and process high-dimensional chemical signals.  This work offers a concrete biophysical picture of how cells make sophisticated decisions.

[12] Sun, Z., Zimmberg, N., Kelly, P., Floyd, C., Papoian, G., Murrell, M., " Feedback between F-actin organization and active stress govern criticality and energy localization in the cell cytoskeleton," Nature Physics, 2025. [article]

This work, done in collaboration with the Murrell and Papoian groups, reveals that the cell's cytoskeleton, the dynamic protein scaffold that gives cells their shape and drives migration and division, operates near a state of self-organized criticality, much like the mechanics governing earthquakes and avalanches. Using a reconstituted actomyosin system as a minimal model, we show that this critical behavior emerges from feedback between F-actin architecture and myosin-generated stress: ordered, sparsely connected networks dissipate energy smoothly and quietly, while disordered, highly connected networks produce heavy-tailed, avalanche-like bursts of energy release. Whether this poised critical state is what allows cells to respond with remarkable sensitivity and flexibility to mechanical demands remains an exciting open question for future work.

[11] Floyd, C., Dinner, A., Vaikuntanathan, S., "Tailoring interactions between active nematic defects with reinforcement learning," Soft Matter, 2025. [article]

2024

[10] Floyd, C., Dinner, A., Vaikuntanathan, S., "Pattern formation in odd viscoleastic fluids," Physical Review Research, 2024. [article]

Pattern formation in soft active matter is typically thought to require chemical regulator fields, but this work shows that mechanics alone can do the job. Studying "odd" viscoelastic fluids, which break classical symmetry assumptions due to nonreciprocal, energy-consuming interactions, we use analytical theory and lattice Boltzmann simulations to identify a purely mechanical pattern-forming instability that spontaneously generates oscillating arrays of fluid vortices. The results open a new avenue for understanding spatiotemporal organization in living materials and engineered active systems.

[9] Redford, S., Colen, J., Shivers, J., Zemsky, S., Molaei, M., Floyd, C., Ruijgrok, P., Vitelli, V., Bryant, Z., Dinner, A., Gardel, M., "Motor crosslinking augments elasticity in active nematics," Soft Matter, 2024. [article]

2023

[8] Floyd, C., Molines, A., Lei, X., Honts, J., Chang, F., Elting, M., Vaikuntanathan, S., Dinner, A., Bhamla, S., "A unified model for the dynamics of ATP-independent ultrafast contraction," PNAS, 2023. (Journal cover) [article] 

Some of the fastest motions in biology are produced not by the familiar actomyosin machinery, but by myonemes, Ca2+-driven protein assemblies found in various single-celled protists. Despite their remarkable speed, the physics governing myonemal contraction has remained largely unexplored. This work derives a minimal continuum model that captures this ATP-independent ultrafast contraction, identifying three key timescales governing the dynamics: chemical driving, material stiffness, and viscous drag. The model quantitatively reproduces experimental measurements across multiple organisms and reveals distinct dynamical regimes, laying the groundwork for understanding and ultimately engineering control over this exotic biological actuator.

[7] Floyd, C., Dinner, A., Vaikuntanathan, S., "Simulating structured fluids with tensorial viscoelasticity," Journal of Chemial Physics, 2023. [article]

2022

[6] Floyd C., Ni, H., Gunaratne, R., Erban, R., Papoian, G., "On stretching, bending, shearing, and twisting of actin filaments I: Variational models," JCTC, 2022. [article]

2021

[5] Floyd, C., Levine, H., Jarzynski, C., Papoian, G., "Understanding cytoskeletal avalanches using mechanical stability analysis," PNAS, 2021. [article] 

Experiments have revealed that the cell cytoskeleton occasionally undergoes sudden, anomalously large rearrangements, so-called "cytoquakes," reminiscent of earthquakes. The physics underlying these events has been poorly understood. Using agent-based simulations of cytoskeletal self-organization, we show that mechanical energy accumulates gradually in the network and then releases in large avalanche-like bursts, and we use mechanical stability analysis to explain why. These results offer a physical picture of cytoskeletal remodeling rooted in the network's mechanical landscape, with implications for how cells sense and respond to their mechanical environment.

[4] Floyd, C., Chandrasekaran, A., Ni, H., Ni, Q., Papoian, G., "Segmental Lennard-Jones interactions for semi-flexible polymer networks," Molecular Physics, 2021. [article] 

2020

[3] Floyd, C., Papoian, G., Jarzynski, C., "Gibbs free energy change of a discrete chemical reaction event," Journal of Chemical Physics, 2020. (Editor's choice) [article] 

Stochastic simulations of biochemical networks inside cells treat molecular copy numbers as small, discrete quantities rather than continuous concentrations. This work reveals that standard textbook expressions for the Gibbs free energy change of a chemical reaction are inaccurate in this regime. We derive exact free energy expressions for the discrete case and show how the familiar continuum formulas emerge through a hierarchy of successive approximations. The result is a more rigorous thermodynamic foundation for stochastic simulation methods widely used to model cellular biochemistry.

2019

[2] Floyd, C., Papoian, G., Jarzynski, C., "Quantifying dissipation in actomyosin networks," Interface focus, 2019. (Journal cover) [article] 

2017

[1] Floyd, C., Jarzynski, C., Papoian, G., "Low-dimensional manifold of actin polymerization dynamics," New Jouranl of Physics, 2017. [article]