Lagrangian single-particle, multi-particle and topological analyses in turbulent Rayleigh-B'enard convection
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.
Kai: I'm Kai.
Mira: And I'm Mira, here with Lev.
Kai: We're looking at the paper: "Lagrangian single-particle, multi-particle and topological analyses in turbulent Rayleigh-B'enard convection".
Mira: Here's the gist of the paper. Lagrangian single-particle, multi-particle and topological analyses in turbulent Rayleigh-Bénard convection present three-dimensional direct numerical simulations of turbulent Rayleigh-Bénard convection (RBC) in the Lagrangian frame of reference for Rayleigh numbers 105 ⩽ Ra ⩽ 1010 and a Prandtl number Pr = 0.7 in a plane layer at an aspect ratio L: L: H = 4: 4: 1 with a horizontal length L and height H <ref:2605.22326/pg1,three-dimensional direct numerical simulations of turbulent Rayleigh-Bénard convection (RBC) in the Lagrangian frame of reference for Rayleigh numbers 105 ⩽ Ra ⩽ 1010 and a Prandtl number Pr = 0.7 in a plane layer at an aspect ratio L: L: H = 4: 4: 1 with a horizontal length L and height H <ref:2605.22326#pg2>.
Kai: Characterization of Lagrangian Heat Transfer Lagrangian heat transfer is extremely intermittent, as individual massless Lagrangian particles can carry convective heat fluxes up to 500 times the global Eulerian mean, although higher-order heat flux moments decrease toward Gaussian values with increasing Ra <ref:2605.22326#pg2>.
Mira: The analysis of velocity gradient invariants in the Q–R plane along trajectories identifies a distinct topological footprint of dust-devil-like convective vortices in the quadrant of Q > 0, R < 0, associated with vortex stretching, plume detachment, and intense localised heat transfer <ref:2605.22326#pg2>.
Kai: Global unconditioned pair dispersion exhibits neither extended Richardson nor Bolgiano–Obukhov scaling plateaus; rather, scale-dependent eddy viscosity and conditioned PCA of dense particle clouds reveal that buoyancy- and shear-driven dispersion are temporally organised, producing a short t5–like episode followed by sustained Richardson-like t3–scaling <ref:2605.22326#pg2>.
Paper summary: Mira: So to wrap up this paper, we’ve seen how Lagrangian single-particle, multi-particle and topological analyses in turbulent Rayleigh-Bénard convection use particle accelerations, heat transfer, Q–R topology, and PCA to characterize transport along material trajectories. The main claim is that these tools reveal the organization of convective transport by showing a distinct topological footprint of dust devils in the upper left quadrant of the Q–R plane.
Kai: And this means that even though we see turbulence as random noise, it has a structure—a specific shape to its coherent vortices that you can map out by tracking particles along their paths. It shows how these vortices stretch and detach plumes from the boundary layers near the walls.
Lev: From an engineering perspective, this implies that when designing systems that need to handle transport in turbulent environments, you need models that account for these specific vortex structures because they drive the most intense, intermittent events.
Kai: The authors found that particle pair dispersion shows a t to the power of three scaling after an initial t to the power of five episode driven by thermal plumes, and this is governed by shear forces across different scales. It’s about understanding those two different scaling regimes sequentially within the inertial subrange.
Mira: And they confirmed this with conditional sampling based on initial vertical velocity, which gave better results than just sampling the temperature field for modeling these transport mechanisms. It really points to how focusing your measurement can clarify what is actually driving the separation events.
Lev: For running real hardware, knowing that we have these distinct regimes—the fast plume detachment and the slower shear-driven dispersion—is crucial because it dictates which physical processes we need to model when predicting error propagation.
Kai: So, in short, this paper gives us a way to see the turbulence not just as a statistical average but as a series of organized events along material trajectories, characterized by specific topological features and sequential transport scaling.
Mira: It’s about moving beyond just knowing the bulk Rayleigh number dependence and understanding the underlying physical mechanisms that dictate how heat and momentum move through the fluid in these turbulent flows.
Lev: The limitation they point out is that these methods rely on three dimensional direct numerical simulations, which means we are seeing a specific idealized setup, and we don't know if those exact results hold up perfectly when scaled to different physical constraints.
Kai: Yeah, so the work is very powerful for understanding the organization of transport in this specific setup, but it’s important to remember that these findings are tied to the conditions they simulated at a Rayleigh number between one hundred five and one thousand ten and a Prandtl number of zero point seven <ref:2605.22326#pg1,1010 and a Prandtl number>.
Conclusion: Kai: So we’ve been looking at how these Lagrangian analyses—the single particle tracking, multi particle stuff—combined with topological views in turbulent Rayleigh-Bénard convection work together to map out what happens along the actual paths of fluid particles.
Mira: Exactly, and this paper is titled "Lagrangian single-particle, multi-particle and topological analyses in turbulent Rayleigh-Bénard convection". It sounds like a lot of math, but the core idea is tracking how heat actually moves in these turbulent flows.
Lev: Tracking the path of a particle means you’re looking at transport along specific trajectories rather than just where things are at one fixed point in space. That’s what makes this approach different from standard Eulerian measurements, which just take snapshots.
Kai: Right, and the paper shows how these Lagrangian methods reveal a distinct topological footprint of convective vortices—specifically these dust-devil-like structures that cause intense localized heat transfer when you look at the Q R plane.
Mira: That’s where it gets interesting for me because it suggests that the turbulence isn't just random noise; there’s a physical organization to those vortex events that dictates how heat gets moved around.
Lev: From an engineering standpoint, knowing these specific vortex structures exist along trajectories tells us exactly where we need to focus our error correction efforts if we were trying to build a simulator or a model for this convection.
Kai: So, the authors are essentially using all these different tools—particle acceleration statistics and Principal Component Analysis—to show that you can see the physics of these vortex events by looking at their shape in phase space.
Mira: And they connect that shape back to the physical mechanisms, showing how buoyancy and shear forces organize the dispersion of particle clouds over time.
Lev: The numbers they give on those scaling regimes, like t to the power of five transitioning to t to the power of three, those are crucial because they tell us which physical force—the thermal plume or the shear—is controlling the transport at different times.
Kai: It’s about understanding that sequence. So if you're looking at a turbulent fluid flow, it’s not just one process happening all at once; it’s a series of events with different scaling rules governing them.
Mira: And this paper provides the framework to see those rules by looking at the particles themselves, which is much more direct than trying to measure the temperature field alone.
Lev: It gives us a better blueprint for how we might design simulations or experiments that capture these specific, intermittent transport events rather than just averaging out and losing that crucial information.
Kai: So this work lays out a way to see the turbulence as a series of organized events along particle paths, characterized by specific topological features and sequential transport scaling.
Mira: It’s about moving beyond just knowing the bulk Rayleigh number dependence and understanding the underlying physical mechanisms that dictate how heat and momentum move through the fluid in these turbulent flows.
Lev: The next thing we need to figure out is how to translate this particle-level organization into a practical, efficient algorithm that can run on real hardware.
Institute of Thermodynamics and Fluid Mechanics, Technische Universiät Ilmenau · University of Arizona
physics.flu-dyn
Submitted: 2026-05-21
Updated: 2026-11-12
Comments: 52 pages, 16 figues, 6 tables
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
Importance score: 81/100
The gist: Lagrangian single-particle, multi-particle and topological analyses in turbulent Rayleigh-Bénard convection present three-dimensional direct numerical simulations of turbulent Rayleigh-Bénard…
Key concepts
- Lagrangian Heat Transfer
- This tracks the movement of individual particles carrying convective heat. It is highly intermittent, meaning some particles move with very high heat fluxes—up to 500 times the average—while others carry much less, showing extreme variability in how heat is transported.
- Dust-Devil-Like Vortices
- These are distinct topological structures found in the flow when analyzing velocity gradients. They appear as a specific pattern in particle trajectories (Q > 0, R < 0) and are linked to intense localized heat transfer events caused by vortex stretching and plume detachment.
- Sequential Scaling Regimes
- Particle dispersion exhibits different scaling behaviors depending on the time scale. Initially, thermal plumes cause violent one-dimensional separation (t⁵), which then transitions to a sustained superdiffusive regime governed by shear forces (t³), before eventually reaching normal diffusion (t).
- Principal Component Analysis (PCA)
- PCA is a statistical method used to simplify complex particle movement data into essential features. By calculating the gyration tensor, it reveals the instantaneous shape of particle clouds and helps identify flow characteristics, such as strong stretching in regions with high aspect ratios.
Terminology
Summary
Lagrangian single-particle, multi-particle and topological analyses in turbulent Rayleigh-Bénard convection present three-dimensional direct numerical simulations of turbulent Rayleigh-Bénard convection (RBC) in the Lagrangian frame of reference for Rayleigh numbers 105 ⩽ Ra ⩽ 1010 and a Prandtl number Pr = 0. p. 0 2Program in Applied Mathematics & Department of Mathematics, University of Arizona, Tucson, AZ
87521, USA. (Received xx; revised xx; accepted xx)
We present three-dimensional direct numerical simulations of turbulent Rayleigh-Bénard
convection (RBC) in the Lagrangian frame of reference for Rayleigh numbers 105 ⩽
𝑅𝑎 ⩽ 1010 and a Prandtl number 𝑃𝑟 = 0. (Received xx; revised xx; accepted xx)
We present three-dimensional direct numerical simulations of turbulent Rayleigh-Bénard
convection (RBC) in the Lagrangian frame of reference for Rayleigh numbers 105 ⩽
𝑅𝑎 ⩽ 1010 and a Prandtl number 𝑃𝑟 = 0. 7 in a plane layer at an aspect ratio 𝐿 : 𝐿 :
𝐻 = 4 : 4 : 1 with a horizontal length 𝐿 and height 𝐻. (Received xx; revised xx; accepted xx)
We present three-dimensional direct numerical simulations of turbulent Rayleigh-Bénard
convection (RBC) in the Lagrangian frame of reference for Rayleigh numbers 105 ⩽
𝑅𝑎 ⩽ 1010 and a Prandtl number 𝑃𝑟 = 0. 7 in a plane layer at an aspect ratio 𝐿 : 𝐿 :
𝐻 = 4 : 4 : 1 with a horizontal length 𝐿 and height 𝐻. 2Program in Applied Mathematics & Department of Mathematics, University of Arizona, Tucson, AZ
87521, USA. (Received xx; revised xx; accepted xx)
We present three-dimensional direct numerical simulations of turbulent Rayleigh-Bénard
convection (RBC) in the Lagrangian frame of reference for Rayleigh numbers 105 ⩽
𝑅𝑎 ⩽ 1010 and a Prandtl number 𝑃𝑟 = 0.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Characterization of Lagrangian Heat Transfer
Lagrangian heat transfer is extremely intermittent, as individual massless Lagrangian particles can carry convective heat fluxes up to 500 times the global Eulerian mean, although higher-order heat flux moments decrease toward Gaussian values with increasing Ra <ref:2605.22326#pg2>. The analysis of velocity gradient invariants in the Q–R plane along trajectories identifies a distinct topological footprint of dust-devil-like convective vortices in the quadrant of Q > 0, R < 0, associated with vortex stretching, plume detachment, and intense localised heat transfer <ref:2605. p. 0 Lagrangian heat transfer is extremely intermittent: individual massless Lagrangian
particles can carry convective heat fluxes up to 500 times the global Eulerian mean, although
higher-order heat flux moments decrease toward Gaussian values with increasing 𝑅𝑎. The
analysis of velocity gradient invariants in the 𝑄–𝑅 plane along trajectories identifies a distinct
topological footprint of dust-devil-like convective vortices in the quadrant of 𝑄 > 0, 𝑅 < 0,
associated with vortex stretching, plume detachment, and intense localised heat transfer. Global unconditioned pair dispersion exhibits neither extended Richardson nor Bolgiano–
Obukhov scaling plateaus. Rather, scale-dependent eddy viscosity and conditioned PCA
of dense particle clouds reveal that buoyancy- and shear-driven dispersion are temporally
organised: rapid plume-driven ejection produces a short 𝑡
5–like episode, followed by
sustained Richardson-like 𝑡
3–scaling.Where the paper says it
Where the paper says it
Analysis of Particle Acceleration Statistics
The data from series A were obtained at every integration time step and subsequently filtered using calibrated cubic smoothing splines to provide physically well-resolved acceleration statistics, as detailed in Appendix A <ref:2605.22326#pg9>. The probability density functions (PDFs) of the vertical (az) and horizontal (ah) acceleration components are computed, showing a distinct symmetrical shape with stretched exponential tails that get wider with increasing Ra <ref:2605. p. 7 The higher-order moments of all distributions underline a
distinct symmetrical shape with stretched exponential tails. The tails of the PDFs get wider with increasing
𝑅𝑎 (panel h), indicating increasing intermittency that is particularly strong within the thermal boundary
layers (panels e,f). Statistics are shown for the total domain (a,b), the bulk region 𝑧 ∈ [𝛿𝑇 , 1−𝛿𝑇 ] (c,d), and
the boundary layers 𝑧 ∈ [0, 𝛿𝑇 ] ∪ [1−𝛿𝑇 , 1] (e,f). The higher-order moments of all distributions underline a
distinct symmetrical shape with stretched exponential tails. The tails of the PDFs get wider with increasing
𝑅𝑎 (panel h), indicating increasing intermittency that is particularly strong within the thermal boundary
layers (panels e,f). Acceleration PDFs exhibit wider tails in the horizontal than in the vertical direction,
indicating that vertical vortices cause more extreme acceleration events than thermal plume detachment
(Schumacher 2008).Where the paper says it
Where the paper says it
Where the paper says it
Investigation of Particle Cloud Geometry
The Principal Component Analysis (PCA) is a statistical tool that transforms data into a new coordinate system, reducing it down to its essential features, the principal components (Greenacre et al. 2022) <ref:2605. p. 0 Abstract must not spill onto p. 2
arXiv:2605. Turbulent particle pair dispersion regimes
In homogeneous isotropic turbulence, Richardson (1926) empirically discovered that the
mean squared separation between two particles, ⟨|R|
2
(𝑡)⟩ with 𝑅 = |R| = |x2 − x1|, does
not grow diffusively (∼ 𝑡), but rather follows a 𝑡
3–scaling until their trajectories become
uncorrelated. The dimensional framework of Kolmogorov (1941) reveals that this inertialrange dispersion is solely governed by the turbulent kinetic energy dissipation rate 𝜖𝜈 and
scales according to
⟨𝑅
2
(𝑡)⟩ ∼ ⟨𝜖𝜈⟩𝑡
3
.Where the paper says it
Where the paper says it
Mechanism-Resolving Diagnostics for Transport
The Lagrangian viewpoint is essential because these mechanisms are organised along trajectories rather than at fixed spatial points <ref:2605.22326#pg15>. The analysis of particle pair dispersion reveals that the scaling exponents exhibit a pronounced overshoot of ∼ t3 before falling monotonically to approach the final diffusive limit (∼ t) at large times <ref:2605. p. 6 Turbulent particle pair dispersion regimes
In homogeneous isotropic turbulence, Richardson (1926) empirically discovered that the
mean squared separation between two particles, ⟨|R|
2
(𝑡)⟩ with 𝑅 = |R| = |x2 − x1|, does
not grow diffusively (∼ 𝑡), but rather follows a 𝑡
3–scaling until their trajectories become
uncorrelated. The dimensional framework of Kolmogorov (1941) reveals that this inertialrange dispersion is solely governed by the turbulent kinetic energy dissipation rate 𝜖𝜈 and
scales according to
⟨𝑅
2
(𝑡)⟩ ∼ ⟨𝜖𝜈⟩𝑡
3
. 11)
marks the transition point between the inertial and buoyancy-driven regimes. While at smaller
separations, 𝑅 < 𝐿𝐵, shear and inertial forces dictate the energy cascade and thus suggest a
𝑡
3–scaling, buoyant forces in connection to thermal plumes dominate the separation dynamics
for scales 𝑅 > 𝐿𝐵 and lead to 𝑡
5
-scaling.Where the paper says it
Where the paper says it
Topological Signatures of Convective Vortices
Conducting a Lagrangian Q–R analysis recovered the classical teardrop shape and Vieillefosse tail across all regimes, observing a pronounced skewness toward the upper-left quadrant of (Q > 0, R. This directly represents the topological footprint of so-called ‘dust-devils’ – vigorous convective vortices that originate in the thermal boundary layer <ref:2605. p. 0 Lagrangian heat transfer is extremely intermittent: individual massless Lagrangian
particles can carry convective heat fluxes up to 500 times the global Eulerian mean, although
higher-order heat flux moments decrease toward Gaussian values with increasing 𝑅𝑎. The
analysis of velocity gradient invariants in the 𝑄–𝑅 plane along trajectories identifies a distinct
topological footprint of dust-devil-like convective vortices in the quadrant of 𝑄 > 0, 𝑅 < 0,
associated with vortex stretching, plume detachment, and intense localised heat transfer.Where the paper says it
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
-
Improved analysis of turbulent Rayleigh-Bénard convection (RBC) by decoupling physical intermittency from numerical aliasing signatures using a cubic smoothing spline approach to determine Lagrangian particle accelerations. This allows for a
clean, unamplified view of Lagrangian intermittency at high Rayleigh numbers.
-
Enhanced understanding of convective transport mechanisms by employing Principal Component Analysis (PCA) on dense particle clouds to quantify the full spatiotemporal deformation of fluid volumes, distinguishing between plume-driven buoyant ejection and subsequent shear-driven turbulent spreading.
-
Improved characterization of local flow topology by analyzing the invariants of the velocity gradient tensor in a plane spanned by Q and R along Lagrangian particle trajectories, allowing for identification of a
distinct topological footprint
associated with dust-devil-like convective vortices in the quadrant of Q > 0, R < 0. -
Refined prediction of dispersion regimes by using conditioned sampling approaches on dense particle clouds to isolate the kinematic effect that thermal plumes have on dispersion, revealing a
sequential dispersion mode: thermal plumes drive initial, violent one-dimensional separation (t5), after which the cloud dispersion settles near the Richardson limit (t3).
-
Development of a scale-dependent Lagrangian eddy viscosity metric evaluated conditioned on spatial separation to probe turbulent dispersion regimes and recover a small scaling range close to the Richardson limit at high Rayleigh numbers.
Abstract
We present 3d Lagrangian DNS of turbulent RBC for Pr=0.7, spanning 10 5 at most Ra at most10 10 in a laterally periodic domain of Γ=4, complemented by dense multi-particle cloud simulations for 10 4 at most Ra at most 10 8 at Γ=16. Using single-particle accelerations, Lagrangian heat transfer, Q-R invariant topology, pair dispersion, scale-dependent Lagrangian eddy viscosity, and principal component analysis of particle clouds, we establish a hierarchical, scale-bridging framework of convective transport. Controlling spectral-element signatures via cubic smoothing splines yields highly intermittent distributions and Heisenberg-Yaglom scaling in the core, while a mixed viscous-thermal model reveals that thermal dissipation significantly drives vertical bulk and horizontal boundary-layer accelerations. Evaluating exact local temperature gradients along tracer trajectories reveals extreme intermittency, where particles carry convective heat fluxes up to 250 times the global Eulerian mean. Dust-devil-like convective vortices, isolated via an ANOVA-validated filtering algorithm at Ra=10 10, dominate the vortex-stretching quadrant of the Q-R invariant map and drive local heat transfer up to 30 times the global Eulerian mean. Our results do not support a Bolgiano-Obukhov (BO59) cascade in the Lagrangian frame. Complementing pair dispersion with scale-dependent Lagrangian eddy viscosity isolates initial-separation and viscous catch-up effects at the viscous-to-inertial crossover, revealing an extended Richardson shoulder and bulk behaviour increasingly reminiscent of HIT with rising Ra. Analysing particle clouds via logarithmic shape invariants, MSD, and the logarithmic geometric mean radius extends this picture to 3d Richardson-like volumetric expansion, showing that plume entrainment induces uniaxial stretching that reduces kinematic catch-up rather than sustaining a BO59 cascade.
Related papers
- Emergent Transfer of a Physics Foundation Model from Simulation to Laboratory Turbulence
- Early onset of secondary shear instability in Kelvin-Helmholtz braids at high Reynolds number
- Rotational covariance, continuous spectrum and non-modal growth of three-dimensional planar weak viscous shock layers
- Deep Learning-based Algebraic Reynolds Stress Closures for RANS Simulations of Turbulent Flows
- Neptuna: A Comprehensive Machine Learning Framework for Benchmarking Complex Multiphase Flows
- Size and shape of condensing steam jets and plumes