Pith. sign in

REVIEW 3 major objections 5 minor 18 references

Modelling the evolution of flow-induced anisotropy of concentrated suspensions

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Two relaxation rates outperform one in suspension flow model

desk verdict A clean, incremental fix to the Gillissen-Wilson fabric model with a genuine out-of-sample transient test, but the evidence is narrower than the claims and the untested closure is a real worry. read the letter →

arxiv 2506.05222 v1 pith:RUQ6IQWU submitted 2025-06-05 cond-mat.soft

classification cond-mat.soft
keywords fabrictensorconstitutivemodelnon-Browniansuspensionshearrotationtransientrheologydiscreteelementmethodflow-inducedanisotropydense
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper argues that the standard Gillissen–Wilson constitutive model for dense suspensions has a blind spot: it underestimates the isotropic part of the fabric tensor, which tracks how many particles are in near contact. The proposed fix is to give the trace of the fabric tensor its own relaxation coefficient, $A$, separate from the coefficient $\beta$ that controls the deviatoric (orientation) part. Using discrete-element simulations, the authors show that this one-parameter modification removes the steady-state error and, with parameters fixed from steady shear alone, significantly improves predictions of the fabric tensor's transient response when the shear direction is rotated. If the modification holds up, it gives a cheap, accurate way to model flow-history effects in suspensions, which matter for industrial processing and any application where the flow direction changes.

What carries the argument

The central object is the fabric tensor $\langle nn\rangle$, built from the dyadic products of the center-to-center unit vectors of near-contact particle pairs—in effect, the orientational average of contact directions. The modified evolution equation, Eq. (3), combines upper-convected transport with two relaxation terms: a deviatoric term proportional to $\beta$ acting on the elongation/compression split of the strain rate, and an isotropic term proportional to $A$ acting only on the trace. The fourth-rank tensor $\langle nnnn\rangle$, required by the equation, is supplied by the Hinch–Leal closure, Eq. (2). The separation of the strain rate into elongational and compressional parts, with the compressional trace feeding the $A$ term, is what lets the model distinguish the growth of the number of contacts from the reorientation of existing contacts.

What would settle it

Measure $\langle nnnn\rangle$ directly from the DEM simulations during a shear rotation and compare it with the Hinch–Leal closure; if the closure deviates significantly in the transient states where $\langle nn\rangle_{12}$ and $\langle nn\rangle_{23}$ evolve, then the modified model's advantage rests on a fragile approximation, and a shear-reversal test ($\theta = \pi$) at a higher volume fraction (e.g., $\phi = 0.5$) with the same fitted parameters should reveal whether the improvement persists.

Watch

Extended reading notes

Core claim

The central claim is that the evolution of the fabric tensor under shear is better described when the isotropic and deviatoric relaxation rates are decoupled. In the standard G-W model, one coefficient $\beta$ sets both rates; the modified model Eq. (3) replaces the trace-contribution term by an independent term weighted by $A$. Fitted to the steady startup shear, with $\beta \approx 9.4$ and $A \approx 42.4$, the modified model captures the diagonal components $\langle nn\rangle_{11}$ and $\langle nn\rangle_{22}$ that the standard model ($\beta \approx 10.9$) underestimates. The same parameters, without any transient fitting, yield substantially better predictions of the shear-rotation response of $\langle nn\rangle_{12}$ and $\langle nn\rangle_{23}$, although the modified model still overshoots $\langle nn\rangle_{12}$ for rotations near $\theta = \pi/2$. The paper concludes that the single-relaxation assumption, rather than the overall structure of the G-W model, is the key limitation.

Load-bearing premise

The modified model's transient predictions inherit the Hinch–Leal closure for the fourth-rank fabric tensor, and the paper does not test whether this 1976 closure remains accurate in the strongly anisotropic, off-diagonal microstructures that appear during shear rotation; if the closure degrades there, the claimed improvement may not carry over to other rotation angles, volume fractions, or flow histories.

Editorial extensions

If this is right

  • With parameters fitted only to steady shear, the modified model predicts the transient fabric evolution after shear rotation, making it a predictive tool for flows with changing shear direction.
  • Because the trace of the fabric tensor measures the average number of near contacts per particle, the new coefficient $A$ provides a separate, testable rate for contact-density relaxation.
  • The correct diagonal components of the steady-state fabric should feed directly into stress closures that depend on the full fabric tensor, improving predictions of normal stress differences.
  • The residual overshoot near $\theta = \pi/2$ marks a specific, localized deficiency that future refinements can target without reworking the whole model.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If $\beta$ and $A$ reflect distinct physical mechanisms, they should vary independently with volume fraction, friction coefficient, and particle polydispersity; refitting at other state points would test this and could map the model's domain of validity.
  • The Hinch–Leal closure is the most plausible source of the remaining overshoot; measuring $\langle nnnn\rangle$ directly in the simulations during shear rotation would show whether the closure or the relaxation form is the limiting approximation.
  • The same two-rate decoupling could be imported into other microstructure models that use a single timescale for the full fabric tensor, potentially improving their transient predictions at the cost of one extra parameter.
  • Because the stress in the G-W framework is a static function of the fabric and strain rate, the improved fabric predictions imply that stress predictions after shear rotation should also improve once the stress–fabric relation is calibrated.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes a modification of the Gillissen-Wilson constitutive model for the fabric tensor of non-Brownian suspensions. The standard model uses a single coefficient β for both the deviatoric and isotropic relaxation terms; the modified model, Eq. (3), introduces an independent coefficient A for the trace (isotropic) contribution. The authors derive the steady-state solution of the modified model in simple shear, fit β and A to DEM data at one volume fraction and friction coefficient, and then compare both models against DEM under shear rotations of the flow direction around the gradient axis. They report improved steady-state agreement and improved transient predictions for the off-diagonal fabric components, using parameters that are fixed before the transient comparison. The transient test is therefore a genuine out-of-sample check, although the steady-state comparison is partly circular because β and A are fitted to some of the components that are then shown to match.

Significance. If the reported improvement is quantitatively robust, the modification is valuable: it is minimal, physically interpretable (decoupling isotropic and deviatoric relaxation), and yields an analytic steady-state solution. The shear-rotation comparison is a real out-of-sample test because the model parameters are fitted only to steady-state startup data. However, the central claim currently rests on a visual comparison of color maps at a single state point, and the model inherits the Hinch-Leal closure for the fourth-rank fabric tensor without validation in the strongly anisotropic states generated by shear rotations. These issues need to be addressed before the claimed higher predictive power can be considered established.

major comments (3)
  1. [§4.2, Fig. 2] The central claim that Eq. (3) is 'significantly better' than Eq. (1) in transients is supported only by a visual comparison of color maps. No error bars, uncertainty estimates, or quantitative error metrics are provided, so the reader cannot judge whether the apparent improvement in panels (e)-(f) is statistically meaningful, especially because the modified model also shows an overshoot near θ≈π/2 that is absent in the simulations. Please report a quantitative comparison (for example, the L2 norm of (⟨nn⟩_ij^model − ⟨nn⟩_ij^DEM)/⟨nn⟩_12^ss over the (γ,θ) grid, for each component and each model) and an uncertainty estimate from repeated independent simulations or block averaging.
  2. [§2, Eq. (2)] The modified evolution equation inherits the Hinch-Leal closure for the fourth-rank fabric tensor, but the manuscript never validates this closure against the simulated ⟨nnnn⟩ in the strongly anisotropic states produced by shear rotations. Because β and A are fitted only to steady-state startup data, the transient comparison is a valid out-of-sample test only if the closure is accurate in those off-equilibrium states; if it is not, the improved match in Fig. 2(e)-(f) could be a fortuitous cancellation between the new coefficient A and closure error rather than evidence that Eq. (3) captures the relevant physics. Please test the closure on DEM-computed ⟨nnnn⟩ in the rotated states (for example, compare Eq. (2) to the simulated fourth-rank fabric as a function of γ and θ), or otherwise rule out such compensation.
  3. [§4.1, Fig. 1(b), Eq. (4)] The steady-state agreement for ⟨nn⟩11 and ⟨nn⟩12 is by construction, since β and A are fitted to those two components; the genuinely independent steady-state checks are ⟨nn⟩22 (which matches) and ⟨nn⟩33 (which is overestimated). The text should explicitly identify which steady-state components are fitted and which are predictions. As written, the conclusion that the model 'indeed shows quantitatively improved predictions in steady state' overstates the evidence from Fig. 1(b). This does not invalidate the transient test, but it changes the strength of the steady-state claim.
minor comments (5)
  1. [§3] The text states that the chosen contact stiffness keeps the system in the stiff-particle limit 'kn ≪ P', but the reported ratio kn/P ≈ 10^8 implies the opposite; this should read kn ≫ P (or kn/P ≫ 1).
  2. [§4.2 and Fig. 2] The second off-diagonal component is referred to as ⟨nn⟩13 in the main text and as ⟨nn⟩23 in the figure caption. Since the rotation protocol rotates the flow and vorticity directions around the gradient direction, the notation should be made consistent and the component definition explicitly stated.
  3. [Eq. (3)] The symbol ∂γ is used for the strain derivative; please define this notation at first use (for example, d/dγ) to avoid ambiguity with a partial derivative with respect to a coordinate.
  4. [Fig. 2] The color scale and normalization are not fully specified in the caption. Please state that the plotted quantity is normalized by ⟨nn⟩_12^ss and provide colorbars with numerical limits for each panel.
  5. [References] References [13] and [14] contain malformed bibliographic entries (incorrect years and DOIs); they should be corrected before publication.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the shear-rotation test is a genuine out-of-sample prediction with parameters fixed from steady-state fits.

full rationale

The paper is self-contained against DEM simulations and its only fitted parameters are explicitly fit. Section 4.1 states: 'We pick the β parameter in the G-W model, Eq. 1, to fit ⟨nn⟩12' and 'The modified G-W has two parameters, allowing to fit the steady-state value of ⟨nn⟩12 and ⟨nn⟩11'; the Fig. 1 caption likewise labels the curves as 'fitted'. Those components therefore match by construction, but the paper does not present that match as a prediction; it identifies it as fitting. The central out-of-sample claim is the shear-rotation transient: Sec. 4.2 says the model curves are produced 'with the model parameters we picked to fit the steady-state microstructure above', so the transient ⟨nn⟩12 and ⟨nn⟩23 evolution is computed with parameters fixed before seeing the rotation data. No equation in the paper reduces the transient output to the fitted inputs by construction. The Hinch-Leal closure, Eq. (2), is imported from an independent 1976 source and is not invoked via the authors' own prior work; whether that closure is accurate for the rotation states is a validation/correctness concern, not a circularity. The reduction Eq. (3) -> Eq. (1) for A=β is explicitly stated and is a controlled model nesting, not a renaming. No load-bearing self-citation or uniqueness-from-authors argument appears. Therefore no significant circularity is found.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the prior G-W model, the Hinch-Leal closure, and an ad hoc decoupling of isotropic and deviatoric relaxation introduced in this paper. The two model parameters are fitted to the same steady-state data used for the steady-state validation; only the shear-rotation transient is a true out-of-sample test. No new physical entities are introduced.

free parameters (3)
  • β (modified G-W model) = ≈9.4
    Controls the deviatoric relaxation rate of the fabric; fitted to steady-state ⟨nn⟩12 and ⟨nn⟩11 in the startup shear simulation (Sec. 4.1).
  • A (modified G-W model) = ≈42.4
    Controls the isotropic (trace) relaxation rate; introduced as a second parameter and fitted to the same steady-state data. Setting A=β recovers the standard model.
  • β (standard G-W model) = ≈10.9
    The single parameter of the original model, fitted to ⟨nn⟩12 only, used as the baseline for comparison (Sec. 4.1).
assumptions (4)
  • domain assumption Hinch-Leal closure (Eq. 2) expresses the fourth-rank fabric tensor ⟨nnnn⟩ as a function of the second-rank fabric tensor.
    This approximate closure from ref [16] is adopted without independent test for the strongly anisotropic transient states generated by shear rotation.
  • domain assumption The strain-rate tensor can be split into elongation and compression parts, Êe and Êc, from the signs of its eigenvalues.
    This decomposition is part of the original G-W model and is preserved in the modification; it assumes a local, rate-independent mechanism for contact formation and loss.
  • ad hoc to paper Isotropic and deviatoric fabric relaxations can be decoupled by giving the trace term its own coefficient A.
    This is the paper's central modification; it is motivated by the empirical mismatch of the standard model and is not derived from microstructural physics.
  • domain assumption The suspension is rate-independent, so fabric evolution is described as a function of strain rather than time.
    Standard for non-Brownian Stokesian suspensions; the paper relies on this to write ∂γ⟨nn⟩ instead of a time derivative.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Modelling the evolution of flow-induced anisotropy of concentrated suspensions." pith.science (2026). https://pith.science/paper/RUQ6IQWU

@misc{pith2026250605222,
  author       = {Pith},
  title        = {Pith review of: Modelling the evolution of flow-induced anisotropy of concentrated suspensions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RUQ6IQWU}},
  note         = {Machine review of arXiv:2506.05222}
}
read the original abstract

Suspensions, which exhibit complex behaviors such as shear thickening, thinning, and jamming, are prevalent in nature and industry. However, predicting the mechanical properties of concentrated suspensions, in both steady state and the transient regime, remains a significant challenge, impacting product quality and process efficiency. In this study, we focus on developing a robust theoretical framework to explain how flow history governs the anisotropy of mechanical responses in suspensions of hard particles under unsteady flow conditions. Our starting point is the Gillissen-Wilson constitutive model, which we confront to DEM simulation data of the micro-structure during steady shear, and shear rotations where the shear axis is rotated by a specific angle around the flow gradient direction. We introduce a simple modification to the Gillissen-Wilson model which leads to a model with higher predictive power in steady state and during shear rotations.

Figures

Figures reproduced from arXiv: 2506.05222 by the authors.

Figure 1
Figure 1. Fabric tensor evolution as a function of strain in sim￾ulations (symbols) and (a) the standard G-W model Eq. 1 (lines) fitted with β ≈ 10.9, (b) modified G-W model Eq. 3 (lines) fitted with β ≈ 9.4 and A ≈ 42.4. pair). Each near contact α is characterized by a center-to￾center unit vector nα. The fabric tensor is then defined as hnnii j = (1/N) PNp α=1 nα,inα, j , where N is the number of par￾ticles in the suspensio… view at source ↗
Figure 2
Figure 2. Polar plot of the fabric tensor hnni12/hnni ss 12 and hnni23/hnni ss 12 with color coding the corresponding values after shear rotation for (a-b) simulations, (c-d) standard G-W model, (e-f) modified G-W model. The radial coordinate is the post-rotation strain and the angular coordinate is the angle of rotation. 4 Results 4.1 Steady state predictions In our simulations, shear is applied along 12, where 1 is the flow… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

18 extracted references · 15 canonical work pages

  1. [1]

    C. Ness, R. Seto and R. Mari, The physics of dense suspensions. Annual Review of Condensed Matter Physics 13 (1), 97-117 (2022). https://doi.org/10.1146/ annurev-conmatphys-031620-105938

  2. [2]

    Gadala-Maria and A

    F. Gadala-Maria and A. Acrivos, Shear-Induced Struc- ture in a Concentrated Suspension of Solid Spheres. Journal of Rheology 24, 799–814 (1980). https://doi. org/10.1122/1.549584

  3. [3]

    M. E. Cates, J. P . Wittmer, J.-P . Bouchaud and P . Claudin, Jamming, Force Chains, and Fragile Matter. Physical Review Letters 81, 1841 (1998). https://doi. org/10.1103/PhysRevLett.81.1841

  4. [4]

    Blanc, E

    F. Blanc, E. Lemaire and F. Peters. Tunable fall ve- locity of a dense ball in oscillatory cross-sheared con- centrated suspensions. J. Fluid Mech. 746, R4 (2014). https://doi.org/10.1017/jfm.2014.160

  5. [5]

    N. Y . C. Lin, C. Ness, M. E. Cates, J. Sun, I. Cohen, Tunable shear thickening in suspensions. Proc. Natl. Acad. Sci. U.S.A. 113, 10774–10778 (2016). https:// doi.org/10.1073/pnas.1608348113

  6. [6]

    C. Ness, R. Mari and M.E. Cates, Shaken and stirred: Random organization reduces viscosity and dissipation in granular suspensions. Science Advances 4(3), eaar3296 (2018). https://doi.org/10.1126/sciadv. aar3296

  7. [7]

    R. Seto, A. Singh, B. Chakraborty, M. M. Denn and J. F. Morris. Shear jamming and fragility in dense suspen- sions. Granular Matter 21, 82 (2019). https://doi.org/10. 1007/s10035-019-0931-5

  8. [8]

    Blanc, F

    F. Blanc, F. Peters and J. JJ Gillissen, M. E. Cates, S. Bosio, C. Benarroche and R. Mari, Rheology of dense suspensions under shear rotation. Physical Re- view Letters 130, 118202 (2023). https://link.aps.org/ doi/10.1103/PhysRevLett.130.118202

Show all 18 references
  1. [9]

    Acharya and M

    P . Acharya and M. Trulsson, Tacking: Shear fragility and geometry reduces the dissipation for dense sus- pensions. Physical Review Research 6, 033327 (2024). https://link.aps.org/doi/10.1103/PhysRevResearch.6. 033327

  2. [10]

    N. K. Agrawal, Z. Ge, M. Trulsson, O. Tam- misola and Luca Brandt. Dense Suspensions in Ro- tary Shear. arXiv:2411.13463 (2024). https://doi.org/ 10.48550/arXiv.2411.13463

  3. [11]

    Rojas and K

    E. Rojas and K. Kamrin, Transient stress and fabric model for quasi-static granular flows in three dimen- sions. Soft Matter , (2025). http://dx.doi.org/10.1039/ D4SM01535E

  4. [12]

    Acharya and M

    P . Acharya and M. Trulsson, Shear jamming transi- tion in alternating shear rotation for frictional and fric- tionless suspensions. arXiv preprint , (2025). https:// doi.org/10.48550/arXiv.2503.12681

  5. [14]

    Ness, Simulating dense, rate-independent suspen- sion rheology using LAMMPS

    C. Ness, Simulating dense, rate-independent suspen- sion rheology using LAMMPS. JComputational Par- ticle Mechanics, 1–7,(2023). https://doi.org/10.1006/ jcph.1995.1039

  6. [15]

    R. Mari, R. Seto, J.F. Morris, M. M. Denn. Shear thickening, frictionless and frictional rheologies in non- Brownian suspensions. Journal of Rheology, 58 1693– 1724,(2014). https://doi.org/10.1122/1.4890747

  7. [16]

    E. J. Hinch and L. G. Leal. Constitutive equations in suspension mechanics. Part 2. Approximate forms for a suspension of rigid particles a ffected by Brownian ro- tations. Journal of Fluid Mechanics, 76 187-208,(1976). https://doi.org/10.1017/S0022112076003200

  8. [17]

    Gillissen, C

    J.J.J. Gillissen, C. Ness, J. D. Peterson, H. J. Wil- son, and M.E.Cates, . Constitutive model for shear- thickening suspensions: Predictions for steady shear with superposed transverse oscillations. Journal of Rheology,64 353–365,(2020). https://doi.org/10.1122/ 1.5129657

  9. [18]

    Gillissen and H

    J.J.J. Gillissen and H. J. Wilson. Modeling sphere suspension microstructure and stress. Physical Review E,98 033119,(2018). https://link.aps.org/doi/10.1103/ PhysRevE.98.033119

  10. [19]

    Transition from the viscous to inertial regime in dense suspensions

    Martin Trulsson Bruno Andreotti and Philippe Claudin. Transition from the viscous to inertial regime in dense suspensions . Physical Review Let- ters, 109 118305,(2012). http://dx.doi.org/10.1103/ PhysRevLett.109.118305

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.