Butterfly Effects in Turbulence: A Unified View through the Finite Size Lyapunov Exponent
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Nonlinear Sciences. The day in nonlinear science: chaos, patterns, networks and adaptation.
Kai: We're looking at the paper: "Butterfly Effects in Turbulence".
Mira: Here's the gist of the paper. The paper synthesizes classical concepts of chaotic sensitivity with modern ideas of Eulerian spontaneous stochasticity using the Finite Size Lyapunov Exponent (FSLE) to characterize multiscale physics in turbulent flows.
Kai: The FSLE provides a unified framework bridging traditional predictability metrics with the complex phenomena observed at high Reynolds numbers and vanishing initial perturbations Unified Framework for Butterfly Effects The work aims to bring together different ideas about the "butterfly effect" into a single framework using the FSLE, which describes perturbation growth as a function of scale The FSLE is used to bridge classical approaches to predictability in turbulence with recent discussions on Eulerian spontaneous stochasticity This synthesis clarifies the physical mechanisms that dictate the fundamental boundaries of forecasting in high-Reynolds-number flows The FSLE allows for a comprehensive characterization of the multiscale physics of turbulent flows Modeling Turbulence and Noise The study employs shell models, such as the Sabra shell model, to investigate predictability in turbulence by mimicking the turbulent energy cascade and its multiscale character The deterministic Sabra shell model is used to establish classical arguments for predictability in chaotic multiscale systems A stochastic variant of the Sabra shell model is introduced to account for fluctuations induced by thermal noise, which becomes relevant below the Kolmogorov dissipative scale The presence of thermal noise modifies the deep dissipative range, inducing an equilibrium spectrum at such small scales Predictability Across Scales The paper reexamines classical arguments for predictability in turbulence by reframing them in terms of the finite-size Lyapunov exponent The FSLE, denoted as λ(δ), quantifies the growth rate of a perturbation of size δ, where it recovers the usual Lyapunov exponent in the limit δ → 0 For larger δ, it describes the nonlinear growth of perturbations which is generally non-universal and depends on the system The FSLE allows for a more realistic treatment of physically relevant uncertainties by using the relation Tp(∆, δ0) ∼ Z ∆ / δ0 dδ λ(δ)δ Spontaneous Stochasticity and Lagrangian Dynamics The FSLE is used to characterize Eulerian spontaneous stochasticity, which is fundamentally tied to the non-uniqueness of solutions of the Euler equations This phenomenon was first discovered in the Lagrangian context through Richardson dispersion, where particle separation grows as hr(2/3) (t)i ∝ ǫt cubed The connection between non-uniqueness of solutions and Richardson dispersion is highlighted by the fact that the limits lim Re→∞ lim κ→0 do not commute, which signifies the essence of Lagrangian spontaneous stochasticity Literal Butterfly Effect Analysis The "literal butterfly" scenario involves localized, sub-dissipative perturbations, where a small perturbation initially localized at the dissipative scales can undergo an initial decrease due to viscous dissipation before growing exponentially The evolution of such a perturbation is analyzed using finite-time Lyapunov exponents (FTLE) to determine the typical evolution ˜δ(t) = δ0 exp(hγtit) The paper concludes that the initial transient decay phase due to viscous forces can reduce the amplitude of the perturbation to a point where it is indistinguishable from thermal noise The FSLE reveals that at high Reynolds numbers, nature is fundamentally and spontaneously stochastic, moving beyond the standard dynamical systems perspective The FSLE attains a precise limiting functional form characterizing the intrinsic unpredictability of turbulent flows in both Eulerian and Lagrangian contexts This modern paradigm bridges Edward Lorenz’s intuitions and the rigorous mathematics of modern statistical physics The "real" butterfly effect is not just about sensitivity to initial conditions but the realization that at high Reynolds numbers, nature is fundamentally and spontaneously stochastic Appendix Details on FSLE Computation The appendix provides details on computing the Finite Size Lyapunov Exponent and considers the effects of intermittency using the multifractal formalism The FSLE is defined as λ(δ0, ∆) ≡ 1/T (δ0, ∆) ln ∆ / δ0 averaged along a very long reference trajectory This procedure is used to compute the FSLE for the shell model and particle dispersion in the Kraichnan flow The results show that when performing the limit Re → ∞ and vanishing noise as discussed above, the predictability time at scale l or, equivalently, for tolerance on the velocity field of order ∆ ≈ (ǫl) 1/3 is finite and basically given by the eddy turnover time τl = T0(l/L0) 2/3 This is the manifestation of the fact that the solution of the system for zero viscosity is not unique and thus of the (Eulerian) spontaneous stochasticity References
Title and authors: Kai: So we've seen that the paper, "Butterfly Effects in Turbulence: A Unified View through the Finite Size Lyapunov Exponent," is trying to unify chaotic sensitivity with Eulerian spontaneous stochasticity using the FSLE. It’s about bringing these different concepts together into one coherent way of looking at turbulence.
Mira: And what they do with it is they use this FSLE to clarify the physical mechanisms that actually determine where forecasting breaks down in high Reynolds number flows.
Lev: What I find interesting is how they use shell models, like the Sabra shell model, to test this theory. They are using these models to mimic the energy cascade and its multiscale nature.
Kai: They actually introduce a stochastic version of that Sabra model too, which accounts for thermal noise. This allows them to see how that thermal noise modifies the deep dissipative range at very small scales.
Mira: That's key because the paper suggests that this thermal noise induces an equilibrium spectrum at those small scales, which is a specific physical consequence they want to explore.
Lev: And they reframe classical predictability arguments using the FSLE, which quantifies how fast a perturbation of size delta grows, and it recovers the usual Lyapunov exponent when delta goes to zero.
Kai: But for larger deltas, it describes that nonlinear growth which isn't universal and depends on the system itself. This is important because it lets them treat physically relevant uncertainties in a more realistic way by using this relation involving Tp(Delta, delta zero) and lambda of delta.
The paper's summary: Kai: So, to summarize what the paper is saying, they are taking the standard idea that sensitive dependence on initial conditions is captured by the Lyapunov exponent and extending it to cover spontaneous stochasticity in turbulence.
Mira: They show how this FSLE lets them characterize both the deterministic chaotic sensitivity and this new Eulerian spontaneous stochasticity that comes from non-unique solutions of equations like the Euler equations.
Lev: The Lagrangian context is really important here because they connect the non-uniqueness of solutions to Richardson dispersion, where particle separation grows in a way related to t to the power of three.
Kai: They highlight this connection by showing that even though you take limits as the Reynolds number goes to infinity and noise goes to zero, those limits don't commute, which signals that Lagrangian spontaneous stochasticity is real.
Mira: Then they look at a very specific scenario called the "literal butterfly effect," where a small perturbation initially localized at the dissipative scales might actually decrease due to viscous dissipation before it grows exponentially.
Lev: They analyze this literal butterfly effect using finite-time Lyapunov exponents, FTLE, to track how that perturbation evolves over time, looking at something like delta t power of gamma t.
Kai: The paper concludes that this initial transient decay phase caused by viscosity can actually shrink the perturbation down so much that it becomes indistinguishable from thermal noise. This is a specific finding about how small perturbations behave near the dissipation scale.
The paper's improvements: Mira: Now, regarding the improvements they suggest, they are really pushing beyond just using standard Lyapunov exponents by using this FSLE as a more powerful tool for predictability assessment.
Kai: They want to use it to characterize what they call spontaneous stochasticity, which is that profound phenomenon where a flow stays intrinsically unpredictable even when you turn off the background noise.
Lev: From an engineering standpoint, we need to see how this translates to real hardware. If we have a system where we measure perturbations up to a certain threshold Delta, the predictability time Tp scales in a specific way based on that FSLE.
Mira: They offer this idea of an adaptive forecasting system that uses the scale-dependent growth rate lambda of delta, which they approximate as epsilon delta minus two, and they mention it's independent of the Reynolds number in the inertial range.
Kai: That independence from the Reynolds number in that specific regime is a big thing because it suggests a more universal scaling for predictability there.
Lev: What I need to know is how you actually measure this scale-dependent growth rate lambda of delta, since that's what drives the whole FSLE characterization.
Mira: They also provide technical details in the appendix on computing the FSLE, specifically mentioning using multifractal formalism to consider intermittency effects when calculating it for shell models and particle dispersion in flows like Kraichnan flow.
Conclusion: Kai: So to wrap up, this paper, "Butterfly Effects in Turbulence: A Unified View through the Finite Size Lyapunov Exponent," shows that at high Reynolds numbers, nature is fundamentally spontaneous stochastic. It moves beyond the standard dynamical systems view where you just look for a positive Lyapunov exponent.
Mira: The FSLE gives them a precise limiting functional form that describes the intrinsic unpredictability of turbulent flows, both in Eulerian and Lagrangian contexts.
Lev: For us on hardware, this suggests that if we can get close to these limits, the predictability time at scale delta is basically set by the eddy turnover time tau epsilon l over two to the power of three.
Kai: So it’s not just about how fast things diverge from a starting point; it’s realizing that at high Reynolds numbers, the flow itself has an intrinsic tendency to be unpredictable.
Mira: It bridges Edward Lorenz’s intuition with the rigorous math of statistical physics by offering this unified view through the FSLE.
Lev: I just want to say that for real hardware, we need to worry about those small initial perturbations hitting that dissipative range where the dynamics change dramatically due to viscosity before they can show their chaotic growth.
Kai: That's a fair point, Lev. It shows that even with all the complexity of turbulence, there are still physical boundaries dictated by things like thermal noise and viscosity.
Department of Physics and INFN, University of Rome "Tor Vergata" · Department of Computational Science and Technology, AlbaNova University Center · Dipartimento di Fisica and INFN - Università degli Studi di Torino · Istituto dei Sistemi Complessi, CNR Via dei Taurini 19 00185 Rome · INFN Sezione Roma "Tor Vergata" · Dipartimento di Fisica, Universitâa di Roma "La Sapienza"
nlin.CD, physics.flu-dyn
Submitted: 2026-07-27
Updated: 2026-11-04
Comments: 15 pages, 5 figures
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 79/100
The gist: The paper presents a systematic synthesis of different concepts related to the "butterfly effect" within fully developed turbulence using the Finite Size Lyapunov Exponent (FSLE) as a unifying…
Key concepts
- Finite Size Lyapunov Exponent (FSLE)
- The FSLE measures how a small perturbation grows as its size changes. It bridges classical predictability metrics with modern ideas of spontaneous noise, providing a unified way to characterize multiscale physics in turbulent flows by showing growth rates across different scales.
- Eulerian Spontaneous Stochasticity
- This refers to the intrinsic unpredictability of fluid flow solutions arising from the non-uniqueness of equations like the Euler equations. It means that even if you know the initial conditions perfectly, the system's evolution is fundamentally random due to this mathematical property.
- Literal Butterfly Effect Analysis
- This examines perturbations starting at small scales where viscous forces cause an initial decay before exponential growth. Using Finite-Time Lyapunov Exponents (FTLE), the paper shows that these localized disturbances can be reduced to noise, clarifying the physical mechanism behind sensitivity.
Terminology
Summary
The paper presents a systematic synthesis of different concepts related to the butterfly effect
within fully developed turbulence using the Finite Size Lyapunov Exponent (FSLE) as a unifying framework to characterize multiscale physics and spontaneous stochasticity. The FSLE bridges classical chaotic metrics with modern Eulerian spontaneous stochasticity by describing perturbation growth across scales
Unified Framework and Core Concepts
The Finite Size Lyapunov Exponent (FSLE) is introduced as a tool to describe the growth rate of perturbations as a function of their scale, enabling a comprehensive characterization of the multiscale physics of turbulent flows It describes how disturbances cascade dynamically across a vast hierarchy of scales rather than following a simple exponential amplification The FSLE is used to bridge classical approaches to predictability in turbulence to the recent discussions on Eulerian spontaneous stochasticity This synthesis clarifies the physical mechanisms that dictate the fundamental boundaries of forecasting in high-Reynolds-number flows
Modeling Turbulence and Perturbation Growth
The study employs shell models of turbulence, such as the Sabra shell model, to investigate predictability and the butterfly effect The Sabra shell model mimics the turbulent energy cascade and preserves its multiscale character Thermal noise is incorporated into a stochastic variant of this model to account for fluctuations below the Kolmogorov dissipative scale Perturbations are analyzed using the FSLE, which quantifies the growth rate of a perturbation of size delta, defined as ˙δ ≈ δλ(δ) The FSLE allows for the treatment of physically relevant uncertainties by writing Tp(∆, δ0) ∼ Z ∆ / δ0 dδ λ(δ) (16)
Characterizing Spontaneous Stochasticity
The FSLE is used to characterize Eulerian spontaneous stochasticity, which is fundamentally tied to the non-uniqueness of solutions of the Euler equations This phenomenon was first discovered in the Lagrangian context through Richardson dispersion The FSLE for particle separation shows that in the limit κ → 0, only the Richardson regime survives, demonstrating a link to Eulerian spontaneous stochasticity The time required for thermal fluctuations to reach large scales is bounded by the eddy turnover time of that scale
The Literal Butterfly Effect and Transient Dynamics
The literal butterfly
scenario, involving localized, sub-dissipative perturbations, is examined using shell models These perturbations initially undergo a transient decay due to viscous dissipation before potentially growing exponentially aligned with the maximal expanding direction The FSLE analysis shows that for scales smaller than dissipative ones, the perturbation can decrease until it reaches a minimum value, which is bounded from below by thermal noise This observation suggests that the real
butterfly effect is theoretically possible when thermal fluctuations are included
Conclusion on Predictability Boundaries
The FSLE reveals a scale-dependent growth rate λ(δ) ≈ ǫδ−2 for perturbations in the inertial range, which is independent of the Reynolds number This result shows that the predictability time at scale l is finite and proportional to the eddy turnover time of that scale The FSLE provides a way to characterize subtle aspects emerging when considering limits of fully developed turbulence, vanishing initial perturbations and noise strength This demonstrates that nature is fundamentally and spontaneously stochastic at high Reynolds numbers
Appendix Details on Intermittency
The appendix details the computation of the Finite Size Lyapunov Exponent and considers intermittency using multifractal formalism The analysis shows that the δ−2 behavior is insensitive to intermittency correction, as derived from Eq. (A6) where β = -2 This result can be physically understood by observing that in (17) λ(δ) depends on the energy dissipation rate ε to the power one, which is not affected by intermittency corrections
Summary of Key Findings
The FSLE unifies disparate paradigms into a single framework that clarifies why traditional chaotic metrics fail at macroscopic scales It shows that the real
butterfly effect is the realization that at high Reynolds numbers, nature is fundamentally and spontaneously stochastic The FSLE attains a precise limiting functional form characterizing the intrinsic unpredictability of turbulent flows It demonstrates that both Eulerian and Lagrangian spontaneous stochasticity are characterized by the FSLE The paper shows that the uncertainty induced by thermal fluctuations is enough to propagate to larger scales in a finite time bounded by the eddy turnover time of the scale of interest This is equivalent to the result (32) derived from combining classical arguments with FSLE
Final Synthesis
By organizing these interconnected ideas under the FSLE, the work bridges the gap between Edward Lorenz’s intuitions and modern statistical physics The real
butterfly effect is not just about sensitivity to initial conditions—it is the realization that at high Reynolds numbers, nature is fundamentally and spontaneously stochastic The FSLE can also be used to characterize the evolution of replicas with different realizations of the noise, providing a way to characterize subtle aspects emerging when considering the limits of fully developed turbulence The paper shows that in both Eulerian (with the shell model) and Lagrangian (with the Kraichnan model) spontaneous stochasticity, the FSLE attains a precise limiting functional form characterizing the intrinsic unpredictability of turbulent flows This is a fresh, comprehensive perspective on the fundamental boundaries of forecasting in high-Reynolds-number flows
References
E. N. Lorenz, “Predictability: does the flap of a butterfly’s wings in brazil set off a tornado in texas? p. 0” in The Essence of chaos (Routledge, 1995) p. p. 1 181, appendix 1: 3pg260724715
R. C. Hilborn, “Sea gulls, butterflies, and grasshoppers: A brief history of the butterfly effect in nonlinear dynamics,” Am. p. 3 J. Phys. 72, 425–427 (2004) pg3pgN
E. Ghys, “The butterfly effect,” in ´ The Proceedings of the 12th International Congress on Mathematical Education: Intellectual and attitudinal challenges (Springer, 2015) pp. p. 1 19–39 pg3pgN
T. N. Palmer, A. D¨oring, and G. Seregin, “The real butterfly effect,” Nonlinearity 27, R123–R141 (2014) pg3pgN
A. A. Mailybaev, “Spontaneous stochasticity of velocity in turbulence models,” Mult. p. 1 p. 1 Model. & Simul. 14, 96–112 (2016) pg3pgN
S. Thalabard, J. Bec, and A. A. Mailybaev, “From the butterfly effect to spontaneous stochasticity in singular shear flows,” Commu. p. 0 p. 0 Phys. 3, 122 (2020) pg3pgN
D. Bandak, A. A. Mailybaev, G. L. Eyink, and N. Goldenfeld, “Spontaneous stochasticity amplifies even thermal noise to the largest scales of turbulence in a few eddy turnover times,” Phys. p. 3 Rev. Lett. 132, 104002 (2024) pg3pgN
D. Ruelle, “Microscopic fluctuations and turbulence,” Phys. p. 2 Lett. A 72, 81–82 (1979) pg3pgN
M. Cencini, F. p. 0 Cecconi, and A. Vulpiani, Chaos: from simple models to complex systems (World Scientific, 2010) pg3pgN
E. N. Lorenz, “The predictability of a flow which possesses many scales of motion,” Tellus 21, 289–307 (1969) pg3pgN
L. Biferale, “Shell models of energy cascade in turbulence,” Annu. p. 2 Rev. Fluid Mech. p. 0 p. 0 p. 0 35, 441–468 (2003) pg3pgN
E. Aurell, G. Boffetta, A. p. 0 Crisanti, G. Paladin, and A. Vulpiani, “Predictability in systems with many characteristic times: The case of turbulence,” Phys. p. 3 Rev. E 53, 2337 (1996) pg3pgN
M. Cencini and A. p. 0 Vulpiani, “Finite size lyapunov exponent: review on applications,” J. p. 3 Phys. A: Math. Theor. 46, 254019 (2013) pg3pgN
L. F. Richardson, “Atmospheric diffusion shown on a distanceneighbour graph,” Proc. p. 9 Royal Soc. Lond. Series A 110, 709–737 (1926) pg3pgN
D. Bernard, K. Gaw¸edzki, and A. Kupiainen, “Slow modes in passive advection,” J. p. 8 Stat. Phys. 90, 519–569 (1998) pg3pgN
G. Falkovich, K. Gaw¸edzki, and M. Vergassola, “Particles and fields in fluid turbulence,” Rev. p. 0 Mod. Phys. 73, 913 (2001) pg3pgN
V. S. L’vov, E. Podivilov, A. Pomyalov, I. Procaccia, and D. Vandembroucq, “Improved shell model of turbulence,” Phys. p. 2 Rev. E 58, 1811 (1998) pg3pgN
D. Bandak, N. Goldenfeld, A. A. Mailybaev, and G. Eyink, “Dissipation-range fluid turbulence and thermal noise,” Phys. p. 3 Rev. E 105, 065113 (2022) pg3pgN
R. H. Kraichnan, “Small-scale structure of a scalar field convected by turbulence,” Phys. p. 8 Fluids 11, 945–953 (1968) pg3pgN
R. Betchov, “On the fine structure of turbulent flows,” J. p. 0 Fluid Mech. 3, 205–216 (1957) pg3pgN
D. K. Lilly, “Numerical simulation studies of two-dimensional turbulence: Ii. p. 5 stability and predictability studies,” Geophys. p. 0 Fluid Dyn. 4, 1–28 (1972) pg3pgN
E. N. Lorenz, “Predictability: A problem partly solved,” in Proc. p. 4 Seminar on predictability, Vol. p. 0 1 (Reading, 1996) pp. 1–18 pg3pgN
B.-W. Shen, R. A. Pielke Sr, X. Zeng, J. Cui, S. Faghih-Naini, W. Paxson, and R. Atlas, “Three kinds of butterfly effects within lorenz models,” Encyclopedia 2, 1250–1259 (2022) pg3pgN
R. A. Pielke Sr, B.-W. Shen, and X. Zeng, “Butterfly effects,” Phys. p. 0 Today 77, 10–10 (2024) pg3pgN
J. B. Bell, A. Nonaka, A. L. Garcia, and G. Eyink, “Thermal fluctuations in the dissipation range of homogeneous isotropic turbulence,” J. p. 2 Fluid Mech. 939, A12 (2022) pg3pgN
T.-S Thalabard, J. Bec, and A. A. Mailybaev, “From the butterfly effect to spontaneous stochasticity in singular shear flows,” Commu. Phys. 3, 122 (2020) pg3pgN
A. Mailybaev, “Spontaneous stochasticity of velocity in turbulence models,” Mult. Model. & Simul. 14, 96–112 (2016) pg3pgN
E Aurell, G. Boffetta, A Crisanti, G Paladin, and A Vulpiani, “Growth of small localized perturbations in surface quasi-geostrophic turbulence,” arXiv preprint arXiv:2605. p. 608244 (2026) pg3pgN
B. Dubrulle, A. Barlet, A. Barral, A. Cheminet, G Costa, P. Dragoni, A Harikrishnan, A Lopez, K Mallick, and Q Pikeroen “Can we predict the weather? p. 0 new tools for an old problem” Phys. p. 8 Rev. Fluids 10, 083801 (2025) pg3pgN
F. Flandoli, “Topics on regularization by noise,” Lecture Notes, University of Pisa (2013) http://users. p. 8dma.unipi.it/flandoli/Berlino Lectures Flandoli.pdf pg3pgN
M. Rehmeier, “Remarks on regularization by noise, convex integration and spontaneous stochasticity,” Milan J. p. 0 Math. 92, 349–370 (2024) pg3pgN
J.-S. She and E. Leveque, “Universal scaling laws in fully developed turbulence,” Phys. p. 0 Rev. Lett. 72, 336 (1994) pg3pgN
A. Berera and R. D. Ho, “Chaotic properties of a turbulent isotropic fluid,” Phys. p. 1 Rev. Lett.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Develop a predictability assessment tool that goes beyond standard Lyapunov exponents by using the Finite Size Lyapunov Exponent (FSLE). This system will be able to characterize
spontaneous stochasticity,
which is defined asthe profound phenomenon where a flow remains intrinsically unpredictable even as background noise vanishes.
-
Implement a simulation framework using shell models that incorporates both deterministic and thermal noise variants of the Sabra model. This allows the AI to distinguish between standard chaotic sensitivity and the effects of thermal fluctuations on predictability, such as how
thermal fluctuations modify the deep dissipative range, inducing an equilibrium spectrum at such small scales.
-
Create a mechanism to analyze localized perturbations in turbulent flows using Finite-Time Lyapunov Exponents (FTLE). This system can track how a perturbation
initially localized in the dissipative range
evolves through an initial transient decay before eventually growing exponentially, clarifying theliteral butterfly
scenario. -
Design a Lagrangian particle tracking module that utilizes FSLE to map
Lagrangian spontaneous stochasticity.
This module will be able to predict how particle separation grows, showing that for finite separation in the inertial range, the predictability time isfinite and of the order of the large-scale eddy turnover time.
-
Develop an adaptive forecasting system that uses FSLE to estimate a scale-dependent growth rate lambda(delta). This system can provide a more nuanced prediction by using the formula lambda(delta) approx epsilon delta - 2, which is
independent of the Reynolds number
in the inertial range.
Sources
Related papers
- PRISM: Unifying multiphysics optothermal transport and identifiability-guided inference for nanoscale thermal metrology
- Quantifying The Complex Spatiotemporal Chaos of Cardiac Fibrillation in Ionic Models Across Parameter Regimes
- Modal interactions in strongly nonlinear coupled oscillators with closely spaced modes and nonclassical damping distribution
- An active Lorentz gas: walking droplets in periodic media
- Stochastic-layer width as an order parameter for the integrability-to-chaos transition in the bouncer model
- When more data become less informative: Finite-precision periodicization and collapse of forecast-error Lyapunov estimates