betaFeedback

WorksPhysics and Astronomy

Nonlinear integral equations for stress-controlled rheology: Collocation methods for Volterra equations with functional kernels

Abstract

We present a class of numerical collocation methods to solve a typical inverse problem that arises in the stress-controlled rheology of complex fluids with pronounced history-dependent response, where the relation between stress and flow rate is given by an integral constitutive equation as the material law. We demonstrate the method by calculating the creep response of a nontrivial model of a shear-thinning viscoelastic fluid where a stress overshoot in rate-controlled startup flow leads to ``sudden yielding'' in stress-controlled creep. Our method is able to address the long-time behavior of the model, distinguishing the amorphous solid from the stress-induced yielded fluid, and from the in-between case of infinite deformation that never achieves a stationary shear rate (proper creep). To this end, we implement a method with flexible time-stepping, and prove uniqueness of the solution and convergence of the methods for a broad class of highly viscoelastic non-linear rheological laws in integral form.

Introduction

The description of the flow of non-Newtonian fluids rests on constitutive laws. They link the stress tensor that enters the coarse-grained macroscopic description based on the Navier-Stokes equations, with the microscopic material models. Recall the simplest example of a constitutive law of a fluid, the Newtonian fluid model: It links the velocity-gradient tensor γ˙=∇u⃗\dot{\boldsymbol{\gamma}} = \nabla\vec{u} of the fluid velocity field u⃗\vec{u}, linearly to the Euler stress tensor σ\boldsymbol{\sigma}. Respecting the expected symmetries of the stress tensor, one obtains for incompressible flow, ∇⋅u⃗=0\nabla\cdot\vec{u} = 0, the relation σ=η(γ˙+γ˙T)\boldsymbol{\sigma} = \eta(\dot{\boldsymbol{\gamma}} + \dot{\boldsymbol{\gamma}}^{T}), where η\eta is the Newtonian viscosity.1

Many materials of technological interest go far beyond the simple Newtonian fluid. The stress-response to a flow can be significantly delayed in time due to slow internal relaxation of the fluid’s microstructure, and it can be nonlinear especially in soft materials where the intrinsic forces are comparable to deformation forces. In principle, also spatial non-locality could be considered, but for simplicity we do not deal with this in the present paper.

A natural form to write a constitutive equation for such a material is thus an integral constitutive equation (ICE):

σ(t)=η(γ˙(t)+γ˙T(t))+∫t0tG(t,t′,[γ˙]):γ˙(t′) dt′(1)\sigma(t) = \eta\left(\dot{\gamma}(t) + \dot{\gamma}^{T}(t)\right) + \int_{t_0}^{t} G\left(t,t',[\dot{\gamma}]\right) : \dot{\gamma}(t')\,\mathrm{d}t' \tag*{(1)}

if the flow is started at some time t0<tt_0 < t in the past. The delayed and potentially non-linear response is encoded in a history integral whose kernel is a response function GG. It is a fourth-rank tensor field, obeying certain symmetries that we will not elaborate in detail here. In the presence of nonlinear response, it has a functional dependence on the velocity-gradient tensor γ˙\dot{\gamma}, denoted by square brackets. The first term on the r.h.s. of Eq. (1) is an instantaneous Newtonian contribution, representing degrees of freedom that can be assumed to relax instantaneously on the time scale of change of the considered flow. In principle it should be encoded in the integral kernel as a fast-decaying contribution followed by this time-scale separation argument; it is common practice to represent it as a singular t=t′t=t' contribution leading to the form of Eq. (1).

A common setup is shear flow, where the flow gradient can be considered to be only along one axis, so that γ˙\dot{\gamma} has only one non-trivial element, [γ˙(t)]ij=δi,xδj,yγ˙(t)[\dot{\gamma}(t)]_{ij} = \delta_{i,x}\delta_{j,y}\dot{\gamma}(t), where the scalar function γ˙(t)\dot{\gamma}(t) is called the shear-rate of the flow. In this case, if one restrcits to calculating just the corresponding shear-stress entry of the stress tensor, one can reduce Eq. (1) to a scalar equation,

σ(t)=ηγ˙(t)+∫t0tG(t,t′,[γ˙])γ˙(t′) dt′(2)\sigma(t) = \eta\dot{\gamma}(t) + \int_{t_0}^{t} G\left(t,t',[\dot{\gamma}]\right)\dot{\gamma}(t')\,\mathrm{d}t' \tag*{(2)}

The scalar function GG is then called the generalized non-linear shear modulus. We work with this scalar equation in the following for simplicity; we expect the generalization to tensorial quantities to be straight-forward.

Equations of this form in particular arise in a microscopic approach from statistical physics, the integration through transients (ITT) formalism [1, 2]. In general, the principles of perturbation theory in physics suggest that the stress-response σ(t)\sigma(t) to a shear perturbation γ˙(t)\dot{\gamma}(t) is given by a generalized convolution integral such as the one appearing in Eq. (2). The generalized shear modulus GG models the material’s response.

Most common rheological models for viscoelasticity can be seen as specializations of Eq. (2) with specific kernels; often such that the resulting integral equation can be converted into a differential equation for the stress or the strain. For example, the Maxwell model [3] is obtained by letting G(t,t′,[γ˙])=G∞exp⁡(−(t−t′)/τ)G(t,t',[\dot{\gamma}]) = G_{\infty}\exp(-(t-t')/\tau) (and η=0\eta= 0). We are however interested in models where the response is more complex, and the larger class of integral equations characterizes the response. This includes a number of “generalized” Maxwell-type models, but also models derived from the mode-coupling theory of the glass transition (MCT), where Eq. (2) is supplemented by another integral equation for the evolution of GG [4, 5].

The form of Eq. (2) as the result of microscopic modeling is dictated by the fact that a shear-flow perturbation is easily modeled as a driving term in the microscopic equations of motion – this is particularly true for colloidal suspensions [4, 5], but also for molecular fluids (save some thermostatting issues) [6, 7]. One thus imagines external forces to act such that a given shear-rate γ˙(t)\dot{\gamma}(t) is prescribed, and the material is characterized by the resulting shear stress σ(t)\sigma(t). This decouples the Navier-Stokes equation from Eq. (2), and the latter is solved by numerical quadrature. For example, linear viscoelasticity [8], constant shear flows [9], or startup-flow experiments involving so-called stress overshoots [10, 11] have been studied using various MCT-derived models, or simple generalizations of the Maxwell model [12].

We are concerned with numerical solutions of the inverse problem defined by Eq. (2): given a certain material stress σ(t)\sigma(t), what is the strain rate γ˙(t)\dot{\gamma}(t) that corresponds to that imposed stress? An especially important example of such stress-controlled rheology is creep, where a constant stress σ(t)=σ0Θ(t)\sigma(t)=\sigma_0\Theta(t) is fixed for all t>0t>0. The asymptotic behavior for γ˙(t→∞)\dot{\gamma}(t\to\infty) is then of interest, as it allows to distinguish flowing states (γ˙→γ˙0>0\dot{\gamma}\to\dot{\gamma}_0>0) from non-flowing ones (γ˙→0\dot{\gamma}\to0), where the latter can be further classified into solid-like states with finite deformation (γ=∫γ˙(t) dt<∞\gamma=\int\dot{\gamma}(t)\,\mathrm{d}t<\infty) from those showing creeping flow (γ→∞\gamma\to\infty even though γ˙→0\dot{\gamma}\to0). This asymptotic behavior is an important predictor of material failure. Also the more general setting of imposed oscillatory stress, σ(t)=σ0sin⁡(ωt)\sigma(t)=\sigma_0\sin(\omega t), is a commonly used and extremely powerful protocol to understand materials.

The solution of this inverse problem is numerically difficult. This is especially true if the integral kernel GG decays very slowly, so that γ˙(t)\dot{\gamma}(t) might change non-trivially over many decades in time, requiring adaptive-time-grid solvers. In principle, one would also need to take into account the Navier-Stokes equation in this case, but we will concentrate on slowly varying flows where the inertial term can be neglected. Recall that for a flow of average density ϱ\varrho and of lateral extension LL, one expects these inertial terms to decay on the timescale of hydrodynamic momentum diffusion, τu=ϱL2/η\tau_u=\varrho L^2/\eta.

We discuss in this paper a robust and versatile method for the numerical inversion of ICEs, using creep and a specific toy model of glass-forming colloidal suspensions as an illustrative example. The paper is organized as follows: We first (§ 2) provide a set of relevant examples of ICE that necessitate our approach. In § 3 we summarize some analytical background on the type of equations given by Eq. (1). In § 4 we demonstrate the use of collocation methods as a powerful numerical technique to solve the problem. Some numerical experiments to demonstrate convergence orders and the computational effort, together with a brief application example constitute § 5. We conclude the paper with the discussion in § 6. More rigorous proofs for existence and uniqueness of the solution, and convergence of the collocation method in some generic rheological cases are collected in the appendix.

Integral Models in Glassy Rheology

Let us give some examples with physical applications that fit the class of models we consider.

Example 1 (Kelvin-Voigt model). This is the trivial case where

GKV(t,t′,[γ˙])=G∞(3)G_{KV}(t,t',[\dot{\gamma}]) = G_{\infty} \tag*{(3)}

is constant. In the literature this model is usually given in the form of an equivalent differential equation for γ(t)=∫t0tγ˙(s) ds\gamma(t) = \int_{t_0}^{t}\dot{\gamma}(s)\,ds:

σ(t)=G∞γ(t)+ηγ˙(t);γ(t0)=0.(4)\sigma(t) = G_{\infty}\gamma(t) + \eta\dot{\gamma}(t); \qquad\gamma(t_0) = 0. \tag*{(4)}

The model describes a visco-elastic solid with shear modulus G∞G_{\infty}. Its creep response is an exponentially relaxing shear rate with time scale τV=η/G∞\tau_V = \eta/G_{\infty}.

γ˙creep,KV(t)=σ0ηe−t/τV.(5)\dot{\gamma}_{\mathrm{creep},KV}(t) = \frac{\sigma_0}{\eta}e^{-t/\tau_V}. \tag*{(5)}

Since the integrated strain γ<∞\gamma< \infty in this case, this is the response of a solid.

Example 2 (Jeffreys model). Another central model in the literature is the Jeffreys model for a visco-elastic fluid, given by

GJ(t,s,[γ˙])=G∞e−t−sτ(6)G_J(t,s,[\dot{\gamma}]) = G_{\infty}e^{-\frac{t-s}{\tau}} \tag*{(6)}

Also this model can be formulated in terms of a a differential equation,

σ˙(t)+1τσ(t)=(G∞+ητ)γ˙(t)+ηγ¨(t);γ˙(0)=σ(0)η.(7)\dot{\sigma}(t) + \frac{1}{\tau}\sigma(t) = \left(G_{\infty}+\frac{\eta}{\tau}\right)\dot{\gamma}(t) + \eta\ddot{\gamma}(t); \qquad\dot{\gamma}(0) = \frac{\sigma(0)}{\eta}. \tag*{(7)}

In the case η=0\eta=0 this is Maxwell’s model. The creep response of the Jeffreys model is

γ˙creep,J(t)=σ0ηM+σ0ηˉe−t/τJ(8)\dot{\gamma}_{\mathrm{creep},J}(t) = \frac{\sigma_0}{\eta_M} + \frac{\sigma_0}{\bar{\eta}}e^{-t/\tau_J} \tag*{(8)}

with τJ=ητ/ηM\tau_J = \eta\tau/\eta_M, ηM=G∞τ+η\eta_M = G_{\infty}\tau+\eta, and ηˉ=η(1+η/G∞τ)\bar{\eta}=\eta(1+\eta/G_{\infty}\tau). In the case η→0\eta\to0, the last term becomes a singular contribution around t=0t=0, and Maxwell’s result γ˙=σ0/G∞τ\dot{\gamma}=\sigma_0/G_{\infty}\tau is recovered. Since γ˙(t)\dot{\gamma}(t) approaches a constant, this is the response of a viscous fluid.

Many complex materials do not show the clear single-exponential relaxation that is predicted by the Jeffreys or Kelvin-Voigt models, but instead show broad relaxation-rate spectra that are well described by power laws. This motivates the introduction of fractional viscoelastic models, often formulated in terms of the Caputo fractional derivative (n∈N0n\in\mathbb{N}_0 and 0<α<10<\alpha<1),

∂n+α∂tn+αf(t):=1Γ(1−α)∫t0tf(n+1)(t′) dt′(t−t′)α.(9)\frac{\partial^{n+\alpha}}{\partial t^{n+\alpha}}f(t) := \frac{1}{\Gamma(1-\alpha)}\int_{t_0}^{t}\frac{f^{(n+1)}(t')\,dt'}{(t-t')^{\alpha}}. \tag*{(9)}

References

  1. [1]M. Fuchs and M. E. Cates, Phys. Rev. Lett. 89, 248304 (2002).
  2. [2]M. Fuchs and M. E. Cates, J. Rheol. 53, 957 (2009).
  3. [3]J. C. Maxwell, Philos. Trans. R. Soc. London 157, 49 (1867).
  4. [4]J. M. Brader, M. E. Cates, and M. Fuchs, Phys. Rev. Lett. 101, 138301 (2008).
  5. [5]J. M. Brader, T. Voigtmann, M. Fuchs, R. G. Larson, and M. E. Cates, Proc. Natl. Acad. Sci. (USA) 106, 15186 (2009).
  6. [6]S.-H. Chong, M. Otsuki, and H. Hayakawa, Phys. Rev. E 81, 041130 (2010).
  7. [7]K. Suzuki and H. Hayakawa, Phys. Rev. E 87, 012304 (2013).
  8. [8]M. Siebenbürger, M. Fuchs, H. Winter, and M. Ballauff, J. Rheol. 53, 707 (2009).
  9. [9]M. Fuchs and M. Ballauff, J. Chem. Phys. 122, 10.1063/1.1859285 (2005).
  10. [10]J. Zausch, J. Horbach, M. Laurati, S. U. Egelhaaf, J. M. Brader, T. Voigtmann, and M. Fuchs, J. Phys.: Cond. Matt. 20, 404210 (2008).
  11. [11]F. Frahsa, A. K. Bhattacharjee, J. Horbach, M. Fuchs, and T. Voigtmann, J. Chem. Phys. 138, 12A513 (2013).
  12. [12]M. Fuchs and M. E. Cates, Faraday Discuss. 123, 267 (2003).
  13. [13]A. Bonfanti, J. L. Kaplan, G. Charras, and A. Kabla, Soft Matter 16, 6002 (2020).
  14. [14]D. Bogue, Ind. Engin. Chem. Fundam. 5, 253 (1966).
  15. [15]M. Siebenbürger, M. Ballauff, and T. Voigtmann, Phys. Rev. Lett. 108, 10.1103/physrevlett.108.255701 (2012).
  16. [16]C. P. Amann, M. Siebenbürger, M. Krüger, F. Weyss­er, M. Ballauff, and M. Fuchs, J. Rheol. 57, 149 (2013).
  17. [17]A. Kaye, Non-Newtonian flow in incompressible fluids (College of Aeronautics, 1962).
  18. [18]T. Voigtmann, J. M. Brader, M. Fuchs, and M. E. Cates, Soft Matter 8, 4244 (2012).
  19. [19]W. Hackbusch, Integral equations: theory and numerical treatment (Birkhäuser, 1995).
  20. [20]A.-M. Wazwaz, Linear and nonlinear integral equations (Springer, 2011).
  21. [21]G. Gripenberg, S.-O. Londen, and O. Staffans, Volterra integral and functional equations (Cambridge University Press, 1990).
  22. [22]R. Zhang, H. Liang, and H. Brunner, SIAM J. Num. Analysis 54, 899 (2016).
  23. [23]H. Brunner and P. J. van der Houwen, The numerical solution of Volterra equations (Elsevier, Amsterdam, 1986).
  24. [24]P. Linz, Analytical and numerical methods for Volterra equations (SIAM, 1985).
  25. [25]J. Stoer, R. Bulirsch, R. Bartels, W. Gautschi, and C. Witzgall, Introduction to numerical analysis (Springer, 2002).
  26. [26]P. Wolkenfelt, BIT Numer. Math. 21, 232 (1981).
  27. [27]W. Götze and L. Sjögren, J. Math. Analysis and Appl. 195, 230 (1995).
  28. [28]H. Brunner, Collocation methods for Volterra integral and related functional differential equations (Cambridge university press, 2004).DOI
  29. [29]E. O. Tuck, Math. Comp. 21, 239 (1967).
  30. [30]M. Caraglio, L. Schrack, G. Jung, and T. Franosch, Commun. Comput. Phys. 29, 628 (2020).
  31. [32]S. Domann and T. Voigtmann, Collocation solvers for creep response of integral constitutive equations (2026), DOI:10.5281/zenodo.22978794.DOI
  32. [33]C. S. Kubrusly, Essentials of measure theory, 2nd ed. (Springer, 2026).
  33. [34]H. Amann and J. Escher, Analysis II (Birkhäuser, 2008).
  34. [35]S. Amat, Z. Li, J. Ruiz-Álvarez, C. Solano, and J. C. Trillo, Math. Comput. Sim. 210, 593 (2023).

Paper details

vlnc:W_g93h8fvh
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents