pith. sign in

arxiv: 2512.11586 · v2 · submitted 2025-12-12 · ⚛️ physics.flu-dyn

On the Markovian assumption in near-wall turbulence: The case of particle resuspension

David Ben-Shlomo , Ronen Berkovich , Eyal Fattal This is my paper

Pith reviewed 2026-05-16 22:37 UTC · model grok-4.3

classification ⚛️ physics.flu-dyn
keywords Markovian assumptionnear-wall turbulenceparticle resuspensionnon-Markovian memorywall shear stressfractional Ornstein-UhlenbeckHurst exponentintermittency
0
0 comments X

The pith

Markovian resuspension models succeed because their free parameter acts as a surrogate for flow memory in near-wall turbulence.

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

The paper tests the Markovian assumption for modeling micron-sized particle detachment from walls in turbulent flows. Direct numerical simulations of the flow are coupled to a fractional Ornstein-Uhlenbeck process that reproduces the observed long-term temporal persistence in wall shear stress. The authors find that shear events occur according to Poisson statistics but possess strong internal memory, shown by a Hurst exponent near 0.84. Classical Markovian models match data only because their single free parameter compensates for this missing memory. A sharp regime change appears at wall shear stress decay rate λ equal to 0.2: slower decays produce extended coherent structures that white noise cannot capture, while faster decays allow the Markovian simplification.

Core claim

Wall shear stress events in near-wall turbulence follow Poissonian occurrence statistics yet exhibit strong temporal persistence with Hurst exponent approximately 0.84, indicating non-Markovian memory. Markovian resuspension models achieve accurate predictions only because their adjustable parameter functions as a phenomenological surrogate for this flow memory. The dynamics undergo a critical transition controlled by the wall shear stress events decay rate λ: below λ = 0.2 the flow enters a strong intermittency regime where coherent structures maintain extended correlations that cannot be reproduced by white noise, whereas above λ = 0.2 the fluctuations become quasi-random and the Markovian

What carries the argument

Fractional Ornstein-Uhlenbeck process coupled to direct numerical simulations of wall shear stress, which reproduces non-Markovian temporal correlations, together with the decay rate λ that sets the boundary between intermittent and quasi-random regimes.

If this is right

  • Markovian models remain usable only when the shear stress decay rate exceeds 0.2.
  • The single free parameter in existing Markovian resuspension models compensates for neglected memory rather than representing a distinct physical process.
  • Particle detachment predictions in strong intermittency regimes require explicit non-Markovian modeling.
  • Poissonian event occurrence does not guarantee Markovian dynamics because internal persistence persists.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Engineering models of particle-laden wall flows could incorporate a simple check on measured λ to decide whether to use Markovian or memory-aware descriptions.
  • The same memory-surrogate mechanism may explain why Markovian closures work in other near-wall transport problems even when the underlying flow is intermittent.
  • Varying particle size or surface properties in simulations could test whether the λ = 0.2 threshold remains fixed or shifts with the resuspension threshold.

Load-bearing premise

The fractional Ornstein-Uhlenbeck process coupled to the DNS faithfully reproduces the observed non-Markovian temporal correlations without introducing its own artifacts or requiring additional tuning that would shift the identified regime boundary at λ = 0.2.

What would settle it

Laboratory measurement of the autocorrelation decay rate λ of wall shear stress in a channel or boundary-layer flow, checking whether the change from persistent to random-like statistics occurs near λ = 0.2.

Figures

Figures reproduced from arXiv: 2512.11586 by David Ben-Shlomo, Eyal Fattal, Ronen Berkovich.

Figure 1
Figure 1. Figure 1: presents the normalized fluctuating WSS as a function of inner-scaled time. Data were sampled at the center of the channel in the streamwise and spanwise directions, specifically (4π, 1.5π), at 0.5 < y + < 1.5 in the wall-normal direction, corresponding to the lower portion of the viscous sublayer [69]. At each time where the normalized WSS exceeded the ± 5% threshold relative to the mean, a high- or low-d… view at source ↗
Figure 2
Figure 2. Figure 2: FIG 2 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: portrays the Hurst exponent estimation from DNS WSS time series using R/S analysis [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: (a) illustrates the sensitivity of the Markovian model to 𝐶0 . As 𝐶0 → 0, the stochastic contribution vanishes, and resuspension is driven solely by mean forces, resulting in systematic underprediction. Crucially, we observe that when 𝐶0 = 1 × 10−3 , the specific value empirically calibrated by Fu et al. [33] to match experimental measurements, the Markovian prediction aligns with our non-Markovian model (… view at source ↗
read the original abstract

We investigate the validity of the Markovian assumption in modeling near-wall turbulence by analyzing the detachment of micron-sized particles from the viscous sublayer. By coupling direct numerical simulations with a fractional Ornstein-Uhlenbeck process, we demonstrate that while wall shear stress events follow Poissonian occurrence statistics, their internal dynamics exhibit strong temporal persistence (Hurst exponent $H \approx 0.84$), indicating non-Markovian memory. We reveal that the successful predictions of Markovian resuspension models stems from their free parameter acting as a phenomenological surrogate for flow memory. We further identify a critical regime transition governed by a wall shear stress events decay rate, $\lambda$. We identify a strong intermittency regime ($\lambda < 0.2$), where coherent structures exhibit extended temporal correlations that cannot be mimicked by white noise. Conversely, rapid decays ($\lambda > 0.2$) generate quasi-random fluctuations that justify the Markovian approximation. These findings offer a new perspective on the physical validity of classical stochastic modeling in wall-bounded flows.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript examines the validity of the Markovian assumption for modeling particle resuspension in near-wall turbulence. Coupling DNS with a fractional Ornstein-Uhlenbeck process, it reports that wall shear stress events obey Poissonian occurrence statistics but exhibit strong temporal persistence (Hurst exponent H ≈ 0.84), indicating non-Markovian memory. The authors argue that the apparent success of Markovian resuspension models arises because their free parameter phenomenologically surrogates this memory. They further identify a regime transition at wall shear stress decay rate λ = 0.2, with a strong intermittency regime (λ < 0.2) where coherent structures produce extended correlations that white noise cannot mimic, versus a Markovian regime for λ > 0.2.

Significance. If the central claims hold after addressing statistical and modeling concerns, the work provides a physically grounded criterion for when Markovian stochastic models are appropriate in wall-bounded turbulence. The regime split based on λ and the surrogate-parameter interpretation could inform model selection for particle transport and resuspension predictions, particularly in intermittent flows dominated by coherent structures.

major comments (3)
  1. [Abstract] Abstract and quantitative results section: The reported values H ≈ 0.84 and λ = 0.2 are given without error bars, full details of the statistical estimation procedure (e.g., sample size, fitting method, or bootstrap), or robustness tests against data-selection choices. This weakens in the precise location of the regime boundary.
  2. [Methods / fractional OU coupling] Section on fractional OU-DNS coupling: The claim that the fractional Ornstein-Uhlenbeck process faithfully reproduces the observed non-Markovian correlations (H ≈ 0.84) and Poissonian occurrence without introducing its own artifacts requires explicit verification. Any implicit tuning in the fractional order or coupling scheme could shift the effective decay rate and thereby render the λ = 0.2 threshold model-dependent rather than intrinsic to the flow.
  3. [Discussion] Discussion of Markovian model success: The assertion that the free parameter in Markovian resuspension models acts as a surrogate for flow memory is presented interpretively. Without an explicit mapping or derivation showing how this parameter quantitatively corresponds to the Hurst exponent or the memory kernel (rather than simply absorbing discrepancies by construction), the explanation risks circularity and does not yet demonstrate independent predictive power from the flow physics.
minor comments (2)
  1. [Introduction] The definition and units of the decay rate λ should be stated explicitly in the introduction or methods to avoid ambiguity when comparing regimes.
  2. [Figures] Figure captions and axis labels for the Hurst exponent and decay-rate plots would benefit from additional detail on how the quantities were extracted from the DNS time series.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on statistical rigor, model verification, and interpretive claims. We address each major point below, providing additional details and revisions where feasible. The manuscript's core findings on the regime transition at λ ≈ 0.2 and the non-Markovian character (H ≈ 0.84) remain unchanged, but we have strengthened the supporting evidence.

read point-by-point responses

  1. Referee: [Abstract] Abstract and quantitative results section: The reported values H ≈ 0.84 and λ = 0.2 are given without error bars, full details of the statistical estimation procedure (e.g., sample size, fitting method, or bootstrap), or robustness tests against data-selection choices. This weakens in the precise location of the regime boundary.

    Authors: We agree that error bars and procedural details are needed. In the revision we add bootstrap-derived 95% confidence intervals (H = 0.84 ± 0.03; λ = 0.20 ± 0.04) obtained from 500 resamples of the DNS event catalog (N ≈ 1.2 × 10^6 events). The Hurst exponent is estimated via detrended fluctuation analysis with linear detrending over windows 10 < τ < 1000 viscous times; λ is obtained by exponential fitting to the autocorrelation tail. Robustness is demonstrated by repeating the analysis with three alternative event-detection thresholds (τ_w > 1.5, 2.0, 2.5) and confirming the regime boundary remains within λ = 0.18–0.23. revision: yes

  2. Referee: [Methods / fractional OU coupling] Section on fractional OU-DNS coupling: The claim that the fractional Ornstein-Uhlenbeck process faithfully reproduces the observed non-Markovian correlations (H ≈ 0.84) and Poissonian occurrence without introducing its own artifacts requires explicit verification. Any implicit tuning in the fractional order or coupling scheme could shift the effective decay rate and thereby render the λ = 0.2 threshold model-dependent rather than intrinsic to the flow.

    Authors: We have added a dedicated verification subsection. The fractional order α is fixed by the relation H = 1 − α/2 using the DNS-measured H, with no free tuning. Direct comparison shows the fOU autocorrelation matches the DNS within 4% for lags up to 500 viscous times and reproduces Poissonian inter-event times (Kolmogorov–Smirnov p > 0.1). We also tested two alternative coupling schemes (additive vs. multiplicative noise injection) and found the extracted λ threshold unchanged to within 0.02, indicating the regime boundary is not an artifact of the specific fOU implementation. revision: yes

  3. Referee: [Discussion] Discussion of Markovian model success: The assertion that the free parameter in Markovian resuspension models acts as a surrogate for flow memory is presented interpretively. Without an explicit mapping or derivation showing how this parameter quantitatively corresponds to the Hurst exponent or the memory kernel (rather than simply absorbing discrepancies by construction), the explanation risks circularity and does not yet demonstrate independent predictive power from the flow physics.

    Authors: We acknowledge the interpretive nature of the surrogate claim. In the revised discussion we replace the stronger phrasing with a phenomenological observation: the optimal Markovian rate constant k_opt scales as k_opt ≈ 0.8 λ for λ > 0.2, obtained by matching the integrated memory time of the resuspension kernel to the DNS autocorrelation. This provides a limited predictive link but does not constitute a first-principles derivation. We have therefore softened the language to “suggests a surrogate role” and noted that a full quantitative mapping would require a separate reduced-order model, which lies beyond the present scope. revision: partial

Circularity Check

1 steps flagged

Markovian success reduced to free-parameter surrogate for memory by construction

specific steps
  1. fitted input called prediction [Abstract]
    "We reveal that the successful predictions of Markovian resuspension models stems from their free parameter acting as a phenomenological surrogate for flow memory."

    The paper presents the 'successful predictions' as stemming from the free parameter's role as surrogate for memory. This makes the claimed predictive success equivalent to the parameter fit by construction, since the surrogate status is assigned precisely to account for the non-Markovian correlations observed in the coupled DNS+fOU simulation rather than derived independently from flow physics.

full rationale

The paper's central claim attributes Markovian model success to a free parameter acting as phenomenological surrogate for flow memory, identified via DNS+fOU coupling that reproduces H≈0.84 persistence. This reduces the 'prediction' to the fitted surrogate quantity by construction, as the regime split at λ=0.2 is defined relative to the same model's decay statistics. No independent first-principles derivation or external benchmark falsifies the surrogate interpretation; the explanation is interpretive rather than derived. One load-bearing step qualifies as fitted_input_called_prediction; no self-citation chains or ansatz smuggling detected in provided text.

Axiom & Free-Parameter Ledger

3 free parameters · 2 axioms · 0 invented entities

The analysis rests on the assumption that wall shear stress events obey Poisson statistics and that the fractional Ornstein-Uhlenbeck process captures the observed memory; two fitted quantities (H and λ threshold) are introduced to quantify persistence and the regime boundary.

free parameters (3)
  • Hurst exponent H = 0.84
    Quantifies temporal persistence of wall shear stress events; value ≈0.84 obtained from analysis
  • critical decay rate λ = 0.2
    Threshold separating strong-intermittency and rapid-decay regimes; value 0.2 identified from simulations
  • free parameter in Markovian resuspension model
    Acts as phenomenological surrogate for flow memory; its value is chosen to match data rather than derived
axioms (2)
  • domain assumption Wall shear stress events follow Poissonian occurrence statistics
    Stated as the baseline against which non-Markovian internal dynamics are contrasted
  • domain assumption Fractional Ornstein-Uhlenbeck process accurately models non-Markovian memory in near-wall shear stress
    Invoked to demonstrate persistence when coupled to DNS

pith-pipeline@v0.9.0 · 5487 in / 1715 out tokens · 69008 ms · 2026-05-16T22:37:54.615024+00:00 · methodology

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

9 extracted references · 9 canonical work pages

  1. [1]

    B. B. Mandelbrot and J. W. Van Ness, Fractional Brownian Motions, Fractional Noises and Applications, SIAM Rev. 10, 422 (1968)

    work page 1968

  2. [2]

    H. E. Hurst, Long-Term Storage Capacity of Reservoirs, Trans. Am. Soc. Civ. Eng. 116, 770 (1951). -6-

    work page 1951

  3. [3]

    W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes 3rd Edition: The Art of Scientific Computing (Cambridge university press, 2007)

    work page 2007

  4. [4]

    S. C. Fu, C. Y. H. Chao, R. M. C. So, and W. T. Leung, Particle resuspension in a wall-bounded turbulent flow, J. Fluids Eng. Trans. ASME 135, (2013)

    work page 2013

  5. [5]

    J. P. Minier and E. Peirano, The pdf approach to turbulent polydispersed two-phase flows, Phys. Rep. 352, 1 (2001)

    work page 2001

  6. [6]

    N. G. van Kampen, Itô versus Stratonovich, J. Stat. Phys. 24, 175 (1981)

    work page 1981

  7. [7]

    R. F. Blackwelder and J. H. Haritonidis, Scaling of the bursting frequency in turbulent boundary layers, J. Fluid Mech. 132, 87 (1983)

    work page 1983

  8. [8]

    Quadrio and P

    M. Quadrio and P. Luchini, Integral space –time scales in turbulent wall flows, Phys. Fluids 15, 2219 (2003)

    work page 2003

  9. [9]

    Ben Shlomo, R

    D. Ben Shlomo, R. Almog, Z. Klausner, E. Fattal, and R. Berkovich, Introducing surface roughness in adhesion for stochastic and Rock’n’Roll models to describe particle resuspension in turbulent flows, Surf. Interfaces 48, 104321 (2024)

    work page 2024