betaFeedback

WorksPhysics and Astronomy

Microstructural Smoothed Particle Hydrodynamics Model and Simulations of Discontinuous Shear-Thickening Fluids

Abstract

Despite the recent interest in the discontinuous shear-thickening (DST) behaviour, few computational works tackle the rich hydrodynamics of these fluids. In this work, we present the first implementation of a microstructural DST model in Smoothed Particle Hydrodynamic (SPH) simulation. The scalar model was implemented in an SPH scheme and tested in two flow geometries. Three distinct ratios of local to non-local microstructural effects were probed: weak, moderate, and strong non-locality. Strong and moderate cases yielded excellent agreement with flow curves constructed via the Wyart–Cates (WC) model, with the moderate case exhibiting banding patterns. Weak non-locality produced stress-splitting instability, resulting in discontinuous stress fields and poor agreement with the WC model. The mechanism of the stress-splitting has been explored and contextualised by the interaction of local microstructure evolution and the stress-control scheme. Velocity profiles obtained in body force-driven channel flow were found to be in excellent agreement with the analytical solution, yielding an upward inflection corresponding to the typical S-curve. Simulations carried out at increasing driving forces exhibited a decrease in flow. We showed that even the simple scalar model can capture some of the key properties of DST materials, laying the foundation for further SPH study of instabilities and pattern formation.

Introduction

Suspensions of both colloidal [1] and non-colloidal [2, 3] particles are ubiquitous in industrially relevant flows [4], and their rich properties have been subject of extensive study. Some dense suspensions of fine non-Brownian particles can exhibit discontinuous shear thickening (DST) behaviour [5], which has been a subject of great interest in the last decade. With possible implications in industries spanning from manufacturing, environmental engineering, and military application, understanding the underlying mechanism responsible for the unique behaviour has become paramount. Multiple physical explanations have been proposed [6], with the most widely adopted being the mechanism of Seto et al. [7] or Wyart–Cates (WC) model [8], where the particles are held apart in a frictionless state by a characteristic repulsive force. Upon application of external stress, the interparticle repulsion is overcome, bringing about a transition to a frictional state where the developed contact network inhibits the ease of motion of the suspension. A defining feature of DST, as opposed to continuous shear-thickening (CST), is the characteristic ‘S’ shape of the flow curve. This unique non-monotonic shape facilitates unstable behaviours in rheometric [9, 10] and complex flows. For example, Texier et al. [11, 12] demonstrated that the negative gradients of flow curve (dγ˙dσ<0\frac{\mathrm{d}\dot{\gamma}}{\mathrm{d}\sigma}<0) present in DST materials cause instabilities in free surface flows of cornstarch suspension, leading to roll-waves.

WC-type models are inherently microstructural — the key parameter governing the state of the suspension at any point is the arrangement of the particle contact network. Typically, works in the context of WC theory propose a relationship between development of the microstructure and macro-hydrodynamic flow properties through a transient evolution equation. Nakanishi et al. [13] employed a scalar model with an assumed S-shape rheology curve in a range of geometries, including simple shear, Poiseuille flow, and free surface inclined plane. The results in simple shear exhibited periodic solutions, accompanied by formation of inhomogeneous fields of the microstructural parameter across the gap. Baumgarten and Kamrin [14] proposed a scalar model based on the balance of strain-dependant hardening and softening attributed to load buckling, shearing, and electrostatic repulsion. Their model predicted rheological behaviour in good agreement with the experimental evidence.

Another development comes from the work of Gillissen et al. [15], who proposed a full tensorial evolution equation. In addition to predicting DST rheology, they were able to predict negligible first normal stress difference, significant negative second normal stress difference, and flow behaviour in more complicated flow arrangements, including flow reversal and superposed transverse oscillations [16], although the values of normal stress differences predicted by their model were quantitatively too high (N1N_1) and too low (N2N_2) relative to the DEM values, along with the increases during thickening being underestimated. Their model also did not consider the role of contact friction in the development of microstructure and excluded tangential forces and torque balance.

So far, nearly all continuum microstructural models have been implemented in Eulerian frameworks. In recent times, there has been much interest in simulations of complex fluids with grid-less methods, primarily Smoothed Particle Hydrodynamics (SPH) [17, 18]. SPH was originally developed by Gingold and Monaghan [19] and Lucy [20] in 1977 in the study of astrophysical problems. It is a fully Lagrangian scheme, where the fluid is represented by a set of discrete ‘particles’ carried by flow. Hydrodynamic properties are computed at the position of each ‘particle’ by averaging via a smoothing kernel at each time step, and the trajectories of particles evolve over time. In comparison with Eulerian and mixed Lagrangian–Eulerian methods, SPH has a number of advantages and disadvantages. Generally, grid-based schemes benefit from superior convergence rates and easier implementation of boundary conditions [17]. On the other hand, SPH is inherently conservative and handles free surface boundaries and interfacial multiphase flows naturally [21] — which is of importance in a number of studies, including the behaviour of granular bed loads [22], mixing [23], and cavitating flows [24]. Moreover, SPH allows for direct access to the flow history — a crucial feature for microstructural models of complex fluids. This can be leveraged in heterogeneous multi-scale modeling (HMM) [25], with an example application in biologically relevant flows [26], or by directly solving integral constitutive equations [27].

Beyond Newtonian fluids, SPH has been used in a host of fields, including solid and complex fluid mechanics. In 2002, Ellero et al. [28] simulated viscoelastic flow by incorporating the corotational Maxwell model in SPH, followed by Oldroyd-B [29]. Since then, Fang et al. [30, 31] studied the behaviour of free surface flows of an Oldroyd-B fluid, with a focus on addressing tensile instability. Rafiee et al. [32] implemented both Maxwell and Oldroyd-B models in SPH free surface flows, including impacting drop and jet buckling. Xu et al. [33] simulated 3D injection molding of a Cross-model fluid. Implementation of Oldroyd-B was studied in a periodic array of cylinders by Vázquez-Quesada and Ellero [34] and Grilli et al. [35], and in extrudate swell by Xu and Deng [36].

Furthermore, a range of viscoplastic and elasto-viscoplastic models have been implemented in SPH. Rodriguez-Paz and Bonet [37] simulated debris flow down an incline using the Bingham model and the generalized viscoplastic model. Minniti and Paris [38] considered free surface granular flows, with a constitutive relation based on the works of Pouliquen et al. [39] and Jop et al. [40], and validated their model against granular column collapse experiments. Furthermore, Bingham-like viscoplastic models have been applied to sedimenting flows [41, 42], rheometric flows [43] and mixed fluid-structure cases [44, 45]. Recently, Rossi et al. [46] implemented a modified microstructural Papanastasiou model to study thixoviscoplastic flows around a cylinder. The first application of SPH to DST flow was carried out in the work of Vázquez-Quesada et al. [47], wherein they considered a non-microstructural inverse biviscous model in a simple planar flow.

In this work, we present in detail the first implementation of a scalar microstructural DST rheological model in an SPH scheme. Unlike previous models [47], we obtain typical S-shaped DST rheology in the simple shear and characteristically inflexed velocity profiles accompanied by flow reduction behaviour in channel flow. In both cases, stress-splitting is observed. § 1.1 lays out the details of the model used in this study. Section III presents the details of the numerical scheme and boundary condition treatment. Finally, results for simple shear and planar channel flow are presented in § 2 A and IV B, respectively.

Governing Equations

In this section, we outline the main equations used to model our DST fluid system. The continuum system can be written in the Lagrangian frame:

dρdt=−ρ∇⋅v,(1)\frac{d\rho}{dt}=-\rho\nabla\cdot\boldsymbol{v}, \tag*{(1)}
dvdt=1ρ∇⋅σ+F,(2)\frac{d\boldsymbol{v}}{dt}=\frac{1}{\rho}\nabla\cdot\boldsymbol{\sigma}+\boldsymbol{F}, \tag*{(2)}
σ=−pI+η(∇v+∇vT),(3)\boldsymbol{\sigma}=-p\boldsymbol{I}+\eta\left(\nabla\boldsymbol{v}+\nabla\boldsymbol{v}^{T}\right), \tag*{(3)}

where d/dt≡∂/∂t+v⋅∇d/dt\equiv\partial/\partial t+\boldsymbol{v}\cdot\nabla is the material derivative, ρ(x,t)\rho(\boldsymbol{x},t) is the density field, v(x,t)\boldsymbol{v}(\boldsymbol{x},t) is the velocity field, σ(x,t)\boldsymbol{\sigma}(\boldsymbol{x},t) is the stress field, pp is the pressure, and F\boldsymbol{F} is a body force acting on fluid. In this DST model, the local suspension viscosity is determined using the Maron–Pierce expression [48].

η(ϕ,ϕJ,t)=ηs(1−ϕϕJ)−2,(4)\eta(\phi,\phi^{J},t)=\eta_{s}\left(1-\frac{\phi}{\phi^{J}}\right)^{-2}, \tag*{(4)}

where η(ϕ,ϕJ,t)\eta(\phi,\phi^{J},t) and ηs\eta_{s} are, respectively, the suspension and constant solvent viscosities. ϕ\phi is the particle volume fraction, and ϕJ\phi^{J} is the jamming volume fraction. Assuming particle migration is negligible for the problems studied in the present work, the value of ϕ\phi is kept constant. The value of ϕJ\phi^{J} is inferred dynamically as

ϕJ(f(x,t))=ϕmf(x,t)+ϕ0{1−f(x,t)},(5)\phi^{J}(f(\boldsymbol{x},t))=\phi^{m}f(\boldsymbol{x},t)+\phi^{0}\{1-f(\boldsymbol{x},t)\}, \tag*{(5)}

where ϕm\phi^{m} and ϕ0\phi^{0} are frictional and frictionless divergence volume fractions, respectively.

The discontinuous shear-thickening phenomenon wherein the viscosity of the medium undergoes a sharp transition from a lower to a higher value is modeled with the help of a single microstructure scalar field called the fraction of frictional contacts f(x,t)f(\boldsymbol{x},t). The use of this scalar field model follows from the Wyart–Cates model [8], in which the value of ff is determined as exp⁡(−σ∗/σ)\exp(-\sigma^{*}/\sigma). Here, σ∗\sigma^{*} is the onset stress at which the suspension begins to shear-thicken. Figure 1 shows (a) typical flow curves produced by the model and (b) the corresponding viscosity dependence on shear stress. For sufficiently low volume fractions (ϕ<0.50\phi<0.50), continuous shear-thickening (CST) is achieved. Increasing volume fraction above 0.50 leads to a transition from CST to a DST regime, marked by the appearance of the characteristic S-shape, where stress is a multi-valued function of shear rate. Further increases in volume fraction sharpen the S-curve up to a point of divergence (ϕ=ϕm\phi=\phi^{m}).

WC model of the DST fluids

Figure 1. WC model of the DST fluids: (a) Flow curves in the stress-shear plane, (b) viscosity dependence on shear stress. Plots for σ∗=0.005\sigma^{*}=0.005, ϕm=0.562\phi_m=0.562, and ϕ0=0.693\phi_0=0.693.

The value of f(x,t)f(\boldsymbol{x},t) at a point in the flow field depends on the local value of stress, and the dependency follows a sigmoid function. Baumgarten and Kamrin [14] proposed an evolution equation for ff considering the influence of different physical mechanisms towards the formation and destruction of the frictional contacts between suspended particles. In the present study, we propose a modified evolution equation combining some features of WC and Baumgarten’s model. The evolution equation for ff is set as

df(x,t)dt=Kfγ˙(f^(x,t)−f(x,t))+α∇2f(x,t),(6)\frac{df(\boldsymbol{x},t)}{dt}=K_{f}\dot{\gamma}\left(\hat{f}(\boldsymbol{x},t)-f(\boldsymbol{x},t)\right)+\alpha\nabla^{2}f(\boldsymbol{x},t), \tag*{(6)}
f^(x,t)=exp⁡(−[σ∗σxy(x,t)]β),(7)\hat{f}(\boldsymbol{x},t)=\exp\left(-\left[\frac{\sigma^{*}}{\sigma_{xy}(\boldsymbol{x},t)}\right]^{\beta}\right), \tag*{(7)}

where KfK_{f} is the microstructural rate constant governing formation and destruction of structure, and γ˙(x,t)=γ˙:γ˙/2\dot{\gamma}(\boldsymbol{x},t)=\sqrt{\dot{\boldsymbol{\gamma}}:\dot{\boldsymbol{\gamma}}/2} is the local shear calculated as the second invariant of the strain rate tensor γ˙(x,t)=(∇v+∇vT)/2\dot{\boldsymbol{\gamma}}(\boldsymbol{x},t)=(\nabla\boldsymbol{v}+\nabla\boldsymbol{v}^{T})/2. The first term on the right-hand side of Eq. (6) accounts for the local evolution of microstructure, whereas the second term introduces non-local effects via microstructure diffusivity, where α\alpha is the diffusion coefficient.

SPH–DST model

SPH equations of motions

The Navier–Stokes equations are solved using the SPH methodology. A possible discretized form of the governing equations is given below [49]:

dρidt=∑jmjvij⋅∇iWij,(8)\frac{d\rho_{i}}{dt}=\sum_{j}m_{j}\boldsymbol{v}_{ij}\cdot\nabla_{i}W_{ij}, \tag*{(8)}
dvidt=−∑jmjpi+pjρiρj∇iWij+∑jmj(ηi+ηj)xij⋅∇iWijρiρjrij2vij+gi,(9)\frac{d\boldsymbol{v}_{i}}{dt}=-\sum_{j}m_{j}\frac{p_{i}+p_{j}}{\rho_{i}\rho_{j}}\nabla_{i}W_{ij}+\sum_{j}\frac{m_{j}(\eta_{i}+\eta_{j})\boldsymbol{x}_{ij}\cdot\nabla_{i}W_{ij}}{\rho_{i}\rho_{j}r_{ij}^{2}}\boldsymbol{v}_{ij}+\boldsymbol{g}_{i}, \tag*{(9)}

where mjm_{j} is the mass of a particle, Wij=W(∣xij∣)W_{ij}=W(|\boldsymbol{x}_{ij}|) is the smoothing kernel, vij=vi−vj\boldsymbol{v}_{ij}=\boldsymbol{v}_{i}-\boldsymbol{v}_{j}, xij=xi−xj\boldsymbol{x}_{ij}=\boldsymbol{x}_{i}-\boldsymbol{x}_{j}, and rij=xij⋅xijr_{ij}=\sqrt{\boldsymbol{x}_{ij}\cdot\boldsymbol{x}_{ij}}.

Following the weakly compressible SPH (WCSPH) formulation, a 1% variation in density is permitted. The fluid pressure is then calculated based on the Tait’s equation of state:

p=ρ0c2γ[(ρρ0)γ−1]+pb,(10)p=\frac{\rho_{0}c^{2}}{\gamma}\left[\left(\frac{\rho}{\rho_{0}}\right)^{\gamma}-1\right]+p_{b}, \tag*{(10)}

where ρ0\rho_{0} is the density of the fluid, pbp_{b} is the background pressure, and cc is the numerical speed of sound typically set as 10 times the maximum flow velocity vmax⁡v_{\max}.

SPH–DST microstructure model

So far, we have described details of SPH for general flows with Navier–Stokes equations. In this section, we discuss the closure SPH model for the definition of viscosity ηi(fi,t)\eta_i(f_i,t) in the SPH–DST model. The constitutive equations are written as follows, where the subscript ii indicates the particle index (that is f(xi,t)=fif(x_i,t)=f_i):

f˙i=Kfγ˙i(f^i−fi)+α∇2fi,(11)\dot{f}_i=K_f\dot{\gamma}_i(\hat{f}_i-f_i)+\alpha\nabla^2 f_i, \tag*{(11)}
ϕiJ=ϕmfi+ϕ0(1−fi),(12)\phi_i^J=\phi_m f_i+\phi_0(1-f_i), \tag*{(12)}
ηi=ηs(1−ϕiϕiJ)−2.(13)\eta_i=\eta_s\left(1-\frac{\phi_i}{\phi_i^J}\right)^{-2}. \tag*{(13)}

In (11), fif_i is calculated to evaluate ϕJ\phi^J in (12), which in turn is used to obtain viscosity via (13). The value of γ˙\dot{\gamma} in (11) is computed as the contraction of the strain rate tensor γ˙\dot{\boldsymbol{\gamma}} as follows:

γ˙=12γ˙:γ˙,γ˙αβ=12(∂uα∂xβ+∂uβ∂xα).(14)\dot{\gamma}=\frac{1}{2}\sqrt{\dot{\boldsymbol{\gamma}}:\dot{\boldsymbol{\gamma}}},\qquad\dot{\gamma}_{\alpha\beta}=\frac{1}{2}\left(\frac{\partial u_\alpha}{\partial x_\beta}+\frac{\partial u_\beta}{\partial x_\alpha}\right). \tag*{(14)}

The SPH approximation of the gradient of the α\alpha-component of the velocity vector in β\beta-direction is obtained as

(∂uα∂xβ)i=∑jmjρj(ujα−uiα)∇iβWij.(15)\left(\frac{\partial u^\alpha}{\partial x^\beta}\right)_i=\sum_j\frac{m_j}{\rho_j}(u_j^\alpha-u_i^\alpha)\nabla_i^\beta W_{ij}. \tag*{(15)}

The non-local term in equation (11) is discretised in a manner consistent with the Morris formulation of the moment equation [17],

(∇2f)i∈f=∑j∈f2mj(fi−fj)xij⋅∇iWijρj2rij2.(16)(\nabla^2 f)_{i\in f}=\sum_{j\in f}\frac{2m_j(f_i-f_j)\boldsymbol{x}_{ij}\cdot\nabla_i W_{ij}}{\rho_j^2 r_{ij}^2}. \tag*{(16)}

To simulate the ‘S’ - curve (see Figure 1) with the above definition of order parameter ff, the choice of ϕ\phi has to be sufficiently close to ϕm\phi_m which in the above case is ϕ>0.50\phi>0.50. However, the value of η/ηs\eta/\eta_s for ϕ=0.54\phi=0.54 at a high shear rate is approximately 3000, which drastically lowers the time step size required for stable integration of the governing equations. A simple idea used to address this was to modify the definition of ff so that we can anticipate the ‘S’ at a lower volume fraction. Therefore, the modified definition of ff is as follows.

f^i=exp⁡{−(σ∗/σi)β},(17)\hat{f}_i=\exp\{-(\sigma^*/\sigma_i)^\beta\}, \tag*{(17)}

where β\beta is the parameter setting the ‘sharpness’ of the transition between the frictional and frictionless plateaus, without changing the plateau values. We set β=2\beta=2 in this work. Increasing it allows us to introduce the previously discussed S-shaped rheology to an otherwise CST volume fraction, resulting in DST behaviour without increasing viscosity to computationally prohibitive values. The rest of the parameters are obtained by fitting the model to experimental data and other models. Once the value of fif_i and ϕJ(fi)\phi^J(f_i) are obtained, the dynamic viscosity of a SPH particle is updated using (13).

Solid wall boundary modeling

The dummy particle method of Adami et al. [50] is employed for imposing no-slip velocity boundary conditions and impermeability conditions. Pressure, density, and velocity are assigned to the dummy particles as

Comparison of WC model flow curves

Figure 2. Comparison of WC model flow curves, i.e., (a) rate dependence of viscosity and (b) stress dependence of viscosity for β=1\beta= 1 and β=2\beta= 2.

pi∈w=∑j∈fpjWij+(g−ai∈w)⋅∑j∈fρjxijWij∑j∈fWij(18)p_{i\in w}=\frac{\sum_{j\in f}p_jW_{ij}+(\mathbf{g}-\mathbf{a}_{i\in w})\cdot\sum_{j\in f}\rho_j\mathbf{x}_{ij}W_{ij}}{\sum_{j\in f}W_{ij}} \tag*{(18)}
ai∈w=−∇pi∈fρi∈f+g(19)\mathbf{a}_{i\in w}=-\frac{\nabla p_{i\in f}}{\rho_{i\in f}}+\mathbf{g} \tag*{(19)}
ρi∈w=ρ0(pi∈w−pbp0+1)1γ(20)\rho_{i\in w}=\rho_0\left(\frac{p_{i\in w}-p_b}{p_0}+1\right)^{\frac{1}{\gamma}} \tag*{(20)}
p0=ρ0c2γ(21)p_0=\frac{\rho_0c^2}{\gamma} \tag*{(21)}
vi∈w=2vwall−v~i∈w(22)\mathbf{v}_{i\in w}=2\mathbf{v}_{\mathrm{wall}}-\tilde{\mathbf{v}}_{i\in w} \tag*{(22)}

where subscript ww and ff denote the dummy wall and fluid particles, respectively, vi∈w\mathbf{v}_{i\in w} is the prescribed wall velocity, and

v~i∈w=∑j∈fvjWij∑j∈fWij.(23)\tilde{\mathbf{v}}_{i\in w}=\frac{\sum_{j\in f}\mathbf{v}_jW_{ij}}{\sum_{j\in f}W_{ij}}. \tag*{(23)}

Time integrators

A semi-implicit predictor-corrector type integration scheme is used for time marching as follows.

(1) Predictor step:

vin+12=vin+Δt2(dvidt)n(24)\mathbf{v}_i^{n+\frac{1}{2}}=\mathbf{v}_i^n+\frac{\Delta t}{2}\left(\frac{d\mathbf{v}_i}{dt}\right)^n \tag*{(24)}
ρin+12=ρin+Δt2(dρidt)n(25)\rho_i^{n+\frac{1}{2}}=\rho_i^n+\frac{\Delta t}{2}\left(\frac{d\rho_i}{dt}\right)^n \tag*{(25)}
xin+12=xin+Δt2(vi)n(26)\mathbf{x}_i^{n+\frac{1}{2}}=\mathbf{x}_i^n+\frac{\Delta t}{2}\left(\mathbf{v}_i\right)^n \tag*{(26)}
fin+12=fin+Δt2(dfidt)n(27)\mathbf{f}_i^{n+\frac{1}{2}}=\mathbf{f}_i^n+\frac{\Delta t}{2}\left(\frac{d\mathbf{f}_i}{dt}\right)^n \tag*{(27)}

(2) Corrector step:

vin+1=vin+Δt2(dvidt)n+12(28)\mathbf{v}_i^{n+1}=\mathbf{v}_i^n+\frac{\Delta t}{2}\left(\frac{d\mathbf{v}_i}{dt}\right)^{n+\frac{1}{2}} \tag*{(28)}
ρin+1=ρin+Δt2(dρidt)n+12(29)\rho_i^{n+1}=\rho_i^n+\frac{\Delta t}{2}\left(\frac{d\rho_i}{dt}\right)^{n+\frac{1}{2}} \tag*{(29)}
xin+1=xin+Δt2(vi)n+12(30)\mathbf{x}_i^{n+1}=\mathbf{x}_i^n+\frac{\Delta t}{2}\left(\mathbf{v}_i\right)^{n+\frac{1}{2}} \tag*{(30)}
fin+1=fin+Δt2(dfidt)n+12(31)\mathbf{f}_i^{n+1}=\mathbf{f}_i^n+\frac{\Delta t}{2}\left(\frac{d\mathbf{f}_i}{dt}\right)^{n+\frac{1}{2}} \tag*{(31)}

The flow variables at the subsequent time step are obtained as follows:

vin+1=2vin+12−vin(32)\mathbf{v}_i^{n+1}=2\mathbf{v}_i^{n+\frac{1}{2}}-\mathbf{v}_i^n \tag*{(32)}
ρin+1=2ρin+12−ρin(33)\rho_i^{n+1}=2\rho_i^{n+\frac{1}{2}}-\rho_i^n \tag*{(33)}
xin+1=2xin+12−xin(34)\mathbf{x}_i^{n+1}=2\mathbf{x}_i^{n+\frac{1}{2}}-\mathbf{x}_i^n \tag*{(34)}
fin+1=2fin+12−fin(35)\mathbf{f}_i^{n+1}=2\mathbf{f}_i^{n+\frac{1}{2}}-\mathbf{f}_i^n \tag*{(35)}

where nn represents the current time instant, n+12n+\frac{1}{2} represents the predicted time instant (variables at an intermediate time level), and n+1n+1 represents the corrected time instant. This scheme is second order accurate in time. Finally, the time-step size Δt\Delta t is determined based on the following condition

Δt=min⁡{0.25hslc,0.125hsl2ν,0.25(hsl∣g∣)1/2},(36)\Delta t=\min\left\{0.25\frac{h_{\mathrm{sl}}}{c},0.125\frac{h_{\mathrm{sl}}^{2}}{\nu},0.25\left(\frac{h_{\mathrm{sl}}}{\lvert\mathbf{g}\rvert}\right)^{1/2}\right\}, \tag*{(36)}

where hslh_{\mathrm{sl}} is the smoothing length, and ν\nu is the kinematic viscosity.

Numerical Simulations

Stress-Imposed Shear Flow

The behaviour of the previously introduced DST model is explored by simulating the rheology in a simple shear geometry. The fluid is bound between two solid walls vertically and two periodic boundaries horizontally at a distance l=0.01l=0.01 m. The domain is planar with no depth in the vorticity direction. Unlike previous shear-imposed simulations [51], input stress σin\sigma_{\mathrm{in}} is set here on the upper wall by assigning wall particles an appropriate velocity.

To impose specified stress on the top wall of the Couette geometry, the following approach is used. The wall particles are assigned with velocity (vi∈w)(v_{i\in w}) computed from the wall force (Fdiff)(F_{\mathrm{diff}}), which is in turn determined from a predictor–corrector method.

vi∈wn+1=vi∈wn+KpFdiffΔt.(37)v_{i\in w}^{n+1}=v_{i\in w}^{n}+K_{\mathrm{p}}F_{\mathrm{diff}}\Delta t. \tag*{(37)}

The net wall force (Fdiff)(F_{\mathrm{diff}}) is computed as the difference between the applied wall force (Fin)(F_{\mathrm{in}}) and the resistive shear force (Fr)(F_{\mathrm{r}}) exerted by the fluid on the top wall

Fdiff=Fin−Fr,(38)F_{\mathrm{diff}}=F_{\mathrm{in}}-F_{\mathrm{r}}, \tag*{(38)}
Fr=∑j∈wFj,(39)F_{\mathrm{r}}=\sum_{j\in w}F_j, \tag*{(39)}
Fi∈w=∑j∈fFijν,(40)F_{i\in w}=\sum_{j\in f}F_{ij}^{\nu}, \tag*{(40)}

where subscripts ww and ff denote wall and fluid particles, respectively. FijνF_{ij}^{\nu} is the inter-particle viscous force in eq. (9). All simulations use a quintic kernel, and smoothing length hsl=1.1Δxh_{\mathrm{sl}}=1.1\Delta x, where Δx=0.02l\Delta x=0.02l is the initial spatial resolution. Simulations are carried out for the same material parameters (see table I) with different imposed stress values. Three microstructure diffusion coefficients are tested to probe effects of the non-local term — strong non-locality (α=10−5 m2 s−1)(\alpha=10^{-5}\,\mathrm{m}^{2}\,\mathrm{s}^{-1}), moderate non-locality (α=10−8 m2 s−1)(\alpha=10^{-8}\,\mathrm{m}^{2}\,\mathrm{s}^{-1}), and weak non-locality (α=0 m2 s−1)(\alpha=0\,\mathrm{m}^{2}\,\mathrm{s}^{-1}). In addition, three volume

fractions were simulated for the case of strong non-locality: CST (ϕ=0.48)(\phi=0.48), moderate DST (ϕ=0.50)(\phi=0.50), and strong DST (ϕ=0.54)(\phi=0.54). Steady-state values are used to construct the flow curve (Figure 4). All rheometric values are obtained by averaging over the entire domain and across time, with an equilibration time of 100 s. Resolution dependence of the microstructure was probed, and 50 × 50 particle configuration was found sufficient.

Results of the SPH simulation compared against the WC model for strong, moderate, and weak non-locality

Figure 4. The results of the SPH simulation compared against the WC model (black) for strong ((a), (d), and (g)), moderate ((b), (e), and (h)), and weak ((c), (f), and (i)) non-locality. The top row (a)–(c) consists of domain-time averaged shear stress and shear rate signal (SPH in blue and red). A single simulation (red) in each diagram (ΣE=1.25\Sigma_{\mathrm{E}} = 1.25 was chosen, and values of all SPH particles were plotted over the WC flow curve (d)–(f). Fields of the microstructure parameter ff visualised in (g)–(i).

Far below onset stress, behaviour is essentially Newtonian — all three measured properties (shear stress, shear rate, and microstructural parameter ff) approach their respective steady-state values asymptotically and within relatively short timescales. Once the stress is increased, the time required to achieve a steady state is significantly increased, and over/under-shoots are present prior to achieving a steady state. This is likely due to the formation of microstructure being relatively slow compared to purely inertial transient effects — in the initial stages, the lack of structure allows shear rate to peak above the equilibrium value. This overshoot then accelerates formation of microstructure, increasing viscosity, which in turn adjusts the shear rate (or undershoots). To mitigate this behaviour, we choose to initialise all simulations from a pre-defined homogeneous field ff corresponding to the correct prediction of the WC at a given input stress, along with the appropriate shear rate via the upper wall velocity. For all values of α\alpha, simulation results (Figure 4(a)–(Figure 4c)) on stable branches are in excellent agreement with the theoretical model predictions.

In the case of strong non-locality, the simulation results for all three volume fractions (Figure 3(a)) match the WC model very well across the entire flow curve, including the region of negative gradient for the DST volume fractions. Both the stress and microstructure fields are entirely homogeneous Figure 4(d) and (g) due to the strong non-local term, with minor deviations in the computed shear rate field, likely due to the imprecision associated with numerical approximations and low resolution. In the purely local (α=0\alpha= 0) case, for all imposed stresses falling within the unstable region of negative flow curve gradient, seemingly spurious ‘jumps’ in microstructure, local stress and viscosity were observed, where individual particles attain either much higher or lower stress and microstructural parameter with respect to the imposed stress. This behaviour had no impact on the numerical stability of the simulations, and a solution could be computed even in this extreme case. These jumps occurred with no clear spatial correlation, yet both the domain averaged properties, and the properties measured at the wall remained in close agreement with the correct theoretically predicted values. By plotting all individual SPH particles for a given applied stress over the expected flow diagram (Figure 4(f)), it is apparent that these jumps are directed towards the valid stable solutions of the WC model.

Results for three volume fractions

Figure 3. Results for three volume fractions: 0.54 (circles), 0.50 (triangles), and 0.48 (diamonds). Strong non-locality (α=10−5)(\alpha=10^{-5}).

hh [m]ll [m]ρ\rho [kg m−3^{-3}]ηs\eta_s [Pa s]ϕ\phiϕm\phi^mϕ0\phi^0σ∗\sigma^* [Pa]
0.010.0110000.0010.48/0.50/0.540.5620.6930.005

Table 1. Simulation Parameters

In principle, a DST fluid could undergo vorticity banding under stress-controlled regime, which, in the simplest case, would manifest as any stress applied within the unstable branch (the branch joining frictional and frictionless branches) being split vertically into two bands organised along the vorticity direction each with either higher or lower stress, but the same shear rate. The width of the bands is then decided by a lever rule, such that the total average stress is equal to the applied stress. Such phenomena in DST materials have been experimentally observed in the work of Herle et al. [9], but Hermes et al. [52] excludes the possibility of steady-state vorticity bands in non-Brownian suspensions on the account of particle migration and normal stresses. For a purely local model and a constant volume fraction field, no effects counteract such splitting, giving rise to a new steady state configuration, where all individual SPH particles occupy stable branches of the flow curve (dfidt=0\frac{df_i}{dt} = 0), while being split such that the stress control equation is also satisfied (dvwalldt=0\frac{dv_{\mathrm{wall}}}{dt} = 0). The lack of spatial correlation in structure (Figure 4(i)) is caused by the action of the microstructure evolution equation on the numerical noise introduced in the SPH solution. Suppose all particles are initialised exactly at the steady state solution, with small perturbation in a random direction in the rate-stress plane. The particles that find themselves to the right of the flow curve will experience dfidt>0\frac{df_i}{dt} > 0, leading to an increase in local viscosity. This will, in turn, increase the local shear stress, increasing the value of f^i\hat{f}_i, ensuring that the particle will continue to move upwards until it reaches a stable upper branch where dfidt=0\frac{df_i}{dt} = 0. Identical argument applies to the particles to the left of the curve, with the resultant downward movement instead. In the case of our simulations, the small perturbations are introduced by the inherent numerical noise associated with meshless calculation the local shear rate γ˙i\dot{\gamma}_i via eq. (14). These small imprecisions do not have a spatial correlation, hence after stress-splitting amplification we arrive at effectively random spatial distribution of microstructural parameter. The left-right split of particles should approach an even split (ψ1=ψ2=0.5\psi^1 = \psi^2 = 0.5); however, the flow curve is not symmetric about the horizontal imposed stress line, and such split would result in an increase of measured shear stress. This means that as the stress-splitting takes place, particles are simultaneously rearranged between the lower and upper branches via the stress-control mechanism of eqs. (37)–(40) to achieve correct average stress.

To further explore both the microstructural transients and steady states, new dimensionless parameters are defined:

ψ1=number on the lower branchtotal number,(41)\psi^1 = \frac{\text{number on the lower branch}}{\text{total number}}, \tag*{(41)}
ψ2=number on the upper branchtotal number,(42)\psi^2 = \frac{\text{number on the upper branch}}{\text{total number}}, \tag*{(42)}
ψu=number on the unstable branchtotal number.(43)\psi^u = \frac{\text{number on the unstable branch}}{\text{total number}}. \tag*{(43)}

The number of particles on either the upper or lower branch is measured with reference to the two threshold stress values σα\sigma^\alpha and σβ\sigma^\beta, which occur at the intersection of the unstable branch with the upper and lower branches, respectively. Typical response consists of a 50 s period for banding process to begin and almost 150 s to 200 s to reach a steady state. Different initial configurations were tested including a) relaxed microstructure, b) correct steady-state microstructure, and c) excessive level of structure. Due to the local nature of this process, the splitting is independent of the resolution as there is no finite characteristic length scale associated with this process. All three configurations yielded the same steady-state banding ratios, i.e., the abovementioned steady-state is broadly independent of the path taken to achieve it. Furthermore, shear-controlled cases were tested both within and outside the multivariate region and yielded results with no apparent banding; a result consistent with the theory where only stress-controlled flow of DST fluids should experience vorticity banding.

To better understand the exact split between ψ1\psi^{1} and ψ2\psi^{2} at steady-state, a simple model of DST vorticity banding is considered. Suppose that stress is applied within the unstable region of the flow curve at a coordinate (γ˙u,σu)(\dot{\gamma}^{u}, \sigma^{u}). This unstable point will split directly in the vertical direction until it meets the lower and upper branches, resulting in a new steady-state consisting of two stable points (γ˙u,σ1)(\dot{\gamma}^{u}, \sigma^{1}) and (γ˙u,σ2)(\dot{\gamma}^{u}, \sigma^{2}), respectively. We also assume that the total average stress is preserved such that:

σu=ψ1σ1+ψ2σ2.(44)\sigma^{u} = \psi^{1}\sigma^{1} + \psi^{2}\sigma^{2}. \tag*{(44)}

By considering that the system is now in a steady state (ψ1+ψ2=1\psi^1+\psi^2=1 and ψu=0\psi^u=0), both fractions can be uniquely determined for any input stress. It is immediately apparent that the exact split will be governed by the exact shape of the flow curve. Finally, we define a dimensionless stress coordinate qq using critical stress values σβ\sigma^\beta and σα\sigma^\alpha (see Figure 4(f)):

q=σu−σβσα−σβ,(45)q=\frac{\sigma^u-\sigma^\beta}{\sigma^\alpha-\sigma^\beta}, \tag*{(45)}

such that, for q<0q<0, input stress is on the lower stable branch, 0≤q≤10\leq q\leq1 input stress is on the unstable branch, and for q>1q>1, input stress is on the upper stable branch. The results along with simulations are presented in Fig. 5. The simulation results exhibit a significantly lower number of particles on the lower branch and a higher number of particles on the upper branch relative to the theoretical predictions. This is due to how the split rearrangement is achieved. Since an even split would result in an increase in the measured shear stress, initial splitting causes the stress-control scheme to reduce shear rate. The shear rate reduction is set by the decrease in wall velocity, which affects all particles, causing a shift to the left in the (γ˙\dot{\gamma}–σ\sigma) phase diagram. Particles on the upper branch can only shift to the lower branch by crossing the critical stress σβ\sigma^\beta, where particles can no longer follow the linear relationship between shear rate and shear stress, and ’fall’ down to the stable branch, due to the combined effects of the microstructure evolution equation and the stress-control scheme. This results in the particles on the upper stable branch, occupying shear stress and shear rate values lower than anticipated in the naive vertical splitting scenario. Lower stress value of σ1\sigma^1 in eq. (44) necessitates higher value of ψ2\psi^2 (and lower value of ψ1\psi^1) at steady state. This mechanism also accounts for the poor agreement between the simulations and the WC model Figure 4(c) — average shear stress is close to the correct input value, but the splitting causes much lower shear rate values.

Finally, the case of moderate non-locality is considered. The averaged results show good agreement with the WC model Figure 4(b) along the entire flow curve, just as in the homogeneous case. Unlike the homogeneous case, the plot of individual SPH particles in Figure 4(e) shows a clear and significant spread in both shear stress and shear rate. However, this spread is different from the local case — particles occupy mainly the unstable region between the two stable branches. The bands in the microstructure parameter are also accompanied by bands in shear stress and shear rate. This steady state occurs as a result of a balance between the split-inducing local term and the smoothing non-local term in the microstructure evolution equation. Furthermore, the competition between the local and non-local effects yields an apparent spatial correlation in the microstructure field Figure 4(h) reminiscent of the bands observed in the work of Nakanishi et al. [13].

Channel Flow

Theoretical Analysis

In this section, we consider flow in a planar channel (Figure 6), with gap d=0.01d=0.01 m and periodic boundary conditions in the flow direction. Body force FF parallel to the flow direction is applied to all particles. In the following, a derivation of the theoretical solution is provided. We begin by assuming a steady state, well-developed unidirectional (u=(u,0,0)\boldsymbol{u}=(u,0,0)) axial flow driven by a body force FF. Eq. (2) simplifies to

Channel geometry sketch

Figure 6. Channel geometry sketch

∂σxy∂y=−F.(46)\frac{\partial\sigma_{xy}}{\partial y}=-F. \tag*{(46)}

The constitutive equation for the only non-zero component of the stress tensor is σxy=ηdudy\sigma_{xy}=\eta\frac{du}{dy} subject to no-slip and vanishing stress boundary conditions u=0u=0 at y=0y=0 and σxy=0\sigma_{xy}=0 at y=hy=h. Integrating eq. (46) with appropriate boundary conditions yields

σxy=F(h−y).(47)\sigma_{xy}=F(h-y). \tag*{(47)}

Using eqs. (4) and (5), the constitutive relation can be rewritten to yield an ODE:

dudy=Fηs(h−y)(1−ϕϕmexp⁡(−[σ∗F(h−y)]β)+ϕ0(1−exp⁡(−[σ∗F(h−y)]β))).(48)\frac{du}{dy}=F\eta_s(h-y)\left(1-\frac{\phi}{\phi_m\exp\left(-\left[\frac{\sigma^*}{F(h-y)}\right]^\beta\right)+\phi_0\left(1-\exp\left(-\left[\frac{\sigma^*}{F(h-y)}\right]^\beta\right)\right)}\right). \tag*{(48)}

Eq. (48) can be solved numerically to obtain shear rate, velocity and viscosity profiles. In this work, the solution was obtained via ODE45 MatLab algorithm with a relative tolerance of 10−810^{-8}.

Numerical Results

With the chosen parameters, we focus primarily on transitional DST regime where both frictional and frictionless states are meaningfully present within the flow and consider only the purely local case. At the steady state, the velocity shape of DST flow displays a sharper profile with an upward inflection near the surface (Figure 7(c)) and are in good agreement with the theoretical solutions of our model. The inflection corresponds to the threshold transition region between primarily frictional and frictionless regimes.

Velocity profiles and corresponding microstructural profiles

Figure 7. Velocity profiles (left) and corresponding microstructural profiles (right) for dimensionless driving forces Fh/σ∗Fh/\sigma^* of 0.5 (a)–(b), 10 (c)–(d), and 100 (e)–(f) for ϕ=0.54\phi=0.54.

It is worth noting that for these values, the ratio between frictionless and frictional viscosity, ηˉ\bar{\eta}, is approximately 30 — far lower than typical magnitudes observed in experimental model systems such as cornstarch, where the viscosity can span multiple orders of magnitude during shear thickening [53, 54]. This low value of ηˉ\bar{\eta} was chosen to facilitate feasible simulation run time by avoiding excessively low or high viscosity at any point during shear thickening.

The resulting velocity profiles are presented in Figure 7. The driving force (and thus the stress profile) was increased to probe the influence of shear thickening on the shape of the velocity profiles.

For sufficiently low driving forces, the velocity profile remains parabolic (a) due to lack of developed microstructure (b) — viscosity remains close to the lower branch viscosity η0\eta_0. The microstructural profile is in good agreement with the theoretical profile, with minor deviation near the centre line. This is due to the initialisation of the structure from a finite

value — particles near centre line experience near-zero shear rate, and thus the structure can not evolve to the appropriate steady state. This deviation is not significantly reflected in the velocity profile due to the weak scaling of viscosity with the microstructure at such low values, resulting in an excellent agreement in the velocity profile.

At intermediate driving forces, flow in the centre (y/D=0.5y/D=0.5) remains unconstrained; however, the microstructure builds up from the walls (d) towards the centre leading to a sharpened velocity profile (c) with an upward inflection. It is apparent that the stress-splitting is present here, with large deviations in the simulated microstructure from the theoretical values. Despite this discrepancy, the simulated velocity profile has very good agreement with the theory. The banding behaviour seems to preserve average velocity, stress, and microstructural profiles within the channel.

Further increase in the driving force leads to almost completely shear frictional flow (f) — the velocity becomes broadly parabolic (e) again with the velocity corresponding to the upper branch viscosity η1\eta_1, however, small inflection persists at the tip, where the transition takes place over a very thin region. Any banding behaviour is constrained the small transitional slice, leading to a deteriorated microstructural solution near the centre line.

The model was compared to the simplest DST model, i.e., inverse biviscous [47]. The biviscous model does not predict S-curves; dγ˙/dσ>0\mathrm{d}\dot{\gamma}/\mathrm{d}\sigma> 0 for all cases. The two models were compared by plotting the average velocity for a range of driving forces (Figure 8) for an equivalent viscosity ratio. The biviscous model predicts two branches corresponding to lower and upper viscosities, joined together by a plateau where increasing the driving force does not affect the average flow. The model presented in this work exhibits similar low and high viscosity branches; however, the conjoining region exhibits a flow reduction pattern unique to the S-shape rheology. It is noteworthy that the presence of a negative flow curve gradient is not always sufficient on its own — a certain minimum viscosity ratio is required to achieve flow reduction behaviour.

Comparison between the present model and biviscous model

Figure 8. Comparison between the present model (blue) and biviscous model [47] (red). Simulations (black) carried out for ϕ=0.54\phi= 0.54, ϕ0=0.64\phi_0 = 0.64, and ϕm=0.56\phi_m = 0.56, uˉ\bar{u} is the average velocity in the channel, and uˉ∗\bar{u}^{*} is an arbitrary average velocity chosen at F/Fmax⁡=1F/F_{\max} = 1.

Conclusions

In this work, we have presented the first implementation of a microstructural model of discontinuous shear thickening in SPH simulations. A simple non-local scalar microstructural evolution equation based on the work of Baumgarten and Kamrin [14] was proposed and implemented in simple shear and channel geometries.

Acknowledgments

Measured transients in shear rate were found to be in qualitative agreement with experimentally measured shear rates under constant imposed stress.

The simple shear results were used to construct flow curves and compare the macroscopic results with the microscopic WC model which we employ. Simulation results were in good agreement along lower and upper viscosity branches and predicted the characteristic ‘S-shape’ transition between them. In the absence of non-locality stress-splitting instability has been observed as a result of the action of the local term in the microstructure evolution equation upon small numerical perturbations in the shear rate field. The resulting steady state was found to be characterised by correct domain-averaged shear stress, despite all SPH particles occupying stable upper and lower branches, lack of spatial correlation, and a low shear rate relative to the WC model. Both low shear rate and the discrepancy between theoretical and measured split ratios were contextualised in terms of the particle rearrangement mechanism resulting from the stress-control scheme. Both moderate and strong non-locality yielded good agreement in average shear rate and shear stress, with banding spatial correlation being observed in the case of moderate non-locality.

Channel flow simulations showed excellent agreement in velocity profiles with the theoretical solution of the WC model across a range of driving forces. Once channel flow is forced into the intermediate DST regime, the velocity profile forms an upward inflection as a result of the S-shaped rheology. Stress-splitting was observed in the channel flow, with no discernable impact on the velocity profiles. Comparison in channel flow was made with the inverse biviscous mode [47]. The present S-shape microstructure model produced a flow reduction behaviour, in contrast to the flow plateau displayed by the biviscous model.

Further study is required to assess stability of the present model and implementation. Exhaustive assessment of the stress-splitting phenomena, band spatial correlation, and the associated characteristic length scale is required. Reproduction of known instability phenomena [10, 55] along with introduction of depth in the vorticity direction could provide further insight into the chaotic rheology of the DST materials.

ACKNOWLEDGMENTS

We thank Zhongqiang Xiong for discussions. PA and BS acknowledges funding from the Engineering and Physical Sciences Research council [EP/S034587/1]. This research is partially supported by the Basque Government through the BERC 2022–2025 program and by the Spanish State Research Agency through BCAM Severo Ochoa excellence accreditation CEX2021-0011 42-S/MICIN/AEI/10.13039/501100011033 and through the project PID2020-117080RB-C55 (“Microscopic foundations of softmatter experiments: computational nano-hydrodynamics” and acronym “Compu-Nano-Hydro”). R.S. acknowledges funding from the National Natural Science Foundation of China (12174390, 12150610463) and Wenzhou Institute, University of Chinese Academy of Sciences (WIUCASQD2020002).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

References

  1. [1]J. Mewis and N.J. Wagner. Colloidal Suspension Rheology. Cambridge Series in Chemical Engineering. Cambridge University Press, 2011. ISBN 9781107376137. URL https://books.google.co.uk/books?id=mNogAwAAQBAJ.
  2. [2]Roger I. Tanner. Review Article: Aspects of non-colloidal suspension rheology. Physics of Fluids, 30(10):101301, 10 2018. ISSN 1070-6631. doi:10.1063/1.5047535. URL https://doi.org/10.1063/1.5047535.
  3. [3]Morton M. Denn and Jeffrey F. Morris. Rheology of non-brownian suspensions. Annual Review of Chemical and Biomolecular Engineering, 5(1):203–228, 2014. doi: 10.1146/annurev-chembioeng-060713-040221. URL https://doi.org/10.1146/annurev-chembioeng-060713-040221. PMID: 24655134.
  4. [4]R.I. Tanner. Engineering Rheology. Oxford Engineering Science Series. OUP Oxford, 2000. ISBN 9780191590160. URL https://books.google.co.uk/books?id=f4vco23dl0UC.
  5. [5]Jeffrey F. Morris. Shear thickening of concentrated suspensions: Recent developments and relation to other phenomena. Annual Review of Fluid Mechanics, 52(1):121–144, 2020. doi:10.1146/annurev-fluid-010816-060128. URL https://doi.org/10.1146/annurev-fluid-010816-060128.
  6. [6]Eric Brown and Heinrich M Jaeger. Shear thickening in concentrated suspensions: phenomenology, mechanisms and relations to jamming. Reports on Progress in Physics, 77(4):046602, apr 2014.
  7. [7]Ryohei Seto, Romain Mari, Jeffrey F. Morris, and Morton M. Denn. Discontinuous shear thickening of frictional hard-sphere suspensions. Phys. Rev. Lett., 111:218301, Nov 2013.
  8. [8]M. Wyart and M. E. Cates. Discontinuous shear thickening without inertia in dense non-brownian suspensions. Phys. Rev. Lett., 112:098302, Mar 2014.
  9. [9]Vishwasra Herle, Peter Fischer, and Erich J. Windhab. Stress driven shear bands and the effect of confinement on their structuresa rheological, flow visualization, and rheo-sals study. Langmuir, 21(20):9051–9057, 2005.
  10. [10]James Richards, John Royer, B. Liecbhen, Ben Guy, and Wilson Poon. Competing timescales lead to oscillations in shear-thickening suspensions. Physical Review Letters, 123, 07 2019. doi:10.1103/PhysRevLett.123.038004.DOI
  11. [11]Baptiste Texier, Henri Lhuissier, Yoël Forterre, and Bloen Metzger. Surface-wave instability without inertia in shear- thickening suspensions. Communications Physics, 3:1234567890, 12 2020.
  12. [12]Baptiste Darbois Texier, Henri Lhuissier, Bloen Metzger, and Yoël Forterre. Shear-thickening suspensions down inclines: from kapitza to oobleck waves. Journal of Fluid Mechanics, 959:A27, 2023.
  13. [13]Hiizu Nakanishi, Shin-ichiro Nagahiro, and Namiko Mitarai. Fluid dynamics of dilatant fluids. Phys. Rev. E, 85:011401, Jan 2012. doi:10.1103/PhysRevE.85.011401. URL https://link.aps.org/doi/10.1103/PhysRevE.85.011401.DOI
  14. [14]Aaron S. Baumgarten and Ken Kamrin. A general constitutive model for dense, fine-particle suspensions validated in many geometries. Proceedings of the National Academy of Sciences, 116(42):20828–20836, 2019.DOI
  15. [15]J. J. J. Gillissen, C. Ness, J. D. Peterson, H. J. Wilson, and M. E. Cates. Constitutive model for time-dependent flows of shear-thickening suspensions. Phys. Rev. Lett., 123:214504, Nov 2019.DOI
  16. [16]J. J. J. Gillissen, C. Ness, J. D. Peterson, H. J. Wilson, and M. E. Cates. Constitutive model for shear-thickening suspensions: Predictions for steady shear with superposed transverse oscillations. Journal of Rheology, 64(2):353–365, 2020.DOI
  17. [17]Steven J. Lind, Benedict D. Rogers, and Peter K. Stansby. Review of smoothed particle hydrodynamics: towards converged lagrangian flow modelling. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 476(2241):20190801, 2020.DOI
  18. [18]Marco Ellero, Mar Serrano, and Pep Español. Incompressible smoothed particle hydrodynamics. Journal of Computational Physics, 226(2):1731–1752, 2007.
  19. [19]R. A. Gingold and J. J. Monaghan. Smoothed particle hydrodynamics: theory and application to non-spherical stars. Monthly Notices of the Royal Astronomical Society, 181(3):375–389, 12 1977.
  20. [20]L. B. Lucy. A numerical approach to the testing of the fission hypothesis. Astronomical Journal, 82:1013–1024, December 1977.
  21. [21]J J Monaghan. Smoothed particle hydrodynamics. Reports on Progress in Physics, 68(8):1703, jul 2005.
  22. [22]Ahmed M. Abdelazek, Ichiro Kimura, and Yasuyuki Shimizu. Simulation of three-dimensional rapid free-surface granular flow past different types of obstructions using the sph method. Journal of Glaciology, 62(232):335–347, 2016.DOI
  23. [23]Georgina Reece, Benedict D. Rogers, Steven Lind, and Georgios Fourtakas. New instability and mixing simulations using sph and a novel mixing measure. Journal of Hydrodynamics, 32(4):684–698, Aug 2020.DOI
  24. [24]Farhoud Kalateh and Ali Koosheh. Simulation of cavitating fluid–structure interaction using sph–fe method. Mathematics and Computers in Simulation, 173:51–70, 2020. ISSN 0378-4754.DOI
  25. [25]Nicolas Moreno and Marco Ellero. Generalized lagrangian heterogeneous multiscale modelling of complex fluids. Journal of Fluid Mechanics, 969:A2, 2023. doi: 10.1017/jfm.2023.540.DOI
  26. [26]Nicolas Moreno, Suzana Nunes, and Victor Calo. Morphological transitions of block copolymer micelles: implications for isoporous membrane ordering, 2023.arxiv.org/abs/2304.03900
  27. [27]Luca Santelli, Adolfo Vázquez-Quesada, and Marco Ellero. Sph simulations of integral multimode and fractional viscoelastic models. J. Non. Newt. Fluid Mech., 2023 (Under Review).
  28. [28]Marco Ellero, Martin Kröger, and Siegfried Hess. Viscoelastic flows studied by smoothed particle dynamics. Journal of Non-Newtonian Fluid Mechanics, 105(1):35–51, 2002. ISSN 0377-0257.
  29. [29]M. Ellero and R.I. Tanner. Sph simulations of transient viscoelastic flows at low reynolds number. Journal of Non-Newtonian Fluid Mechanics, 132(1):61–72, 2005. ISSN 0377-0257.DOI
  30. [30]30 Jiannong Fang, Robert G. Owens, Laurent Tacher, and Aurèle Parriaux. A numerical study of the sph method for simulating transient viscoelastic free surface flows. Journal of Non-newtonian Fluid Mechanics, 139:68–84, 2006.DOI
  31. [31]31 Jiannong Fang, Aurèle Parriaux, Martin Rentschler, and Christophe Ancey. Improved sph methods for simulating free surface flows of viscous fluids. Applied Numerical Mathematics, 59(2):251–271, 2009.DOI
  32. [32]32 A. Rafee, M.T. Manzari, and M. Hosseini. An incompressible sph method for simulation of unsteady viscoelastic free-surface flows. International Journal of Non-Linear Mechanics, 42(10):1210–1223, 2007. ISSN 0020-7462.DOI
  33. [33]33 Xiaoyang Xu, Jie Ouyang, Bin-Xin Yang, and Zhijun Liu. Sph simulations of three-dimensional non-newtonian free surface flows. Computer Methods in Applied Mechanics and Engineering, 256:101–116, 2013.DOI
  34. [34]34 A. Vázquez-Quesada and M. Ellero. Sph simulations of a viscoelastic flow around a periodic array of cylinders confined in a channel. Journal of Non-Newtonian Fluid Mechanics, 167–168:1–8, 2012. ISSN 0377-0257.
  35. [35]35 Muzio Grilli, Adolfo Vázquez-Quesada, and Marco Ellero. Transition to turbulence and mixing in a viscoelastic fluid flowing inside a channel with a periodic array of cylindrical obstacles. Phys. Rev. Lett., 110:174501, Apr 2013.
  36. [36]36 Xiaoyang Xu and Xiao-Long Deng. An improved weakly compressible sph method for simulating free surface flows of viscous and viscoelastic fluids. Computer Physics Communications, 201:43–62, 2016. ISSN 0010-4655.
  37. [37]37 M. X. Rodriguez-Paz and J. Bonet. A corrected smooth particle hydrodynamics method for the simulation of debris flows. Numerical Methods for Partial Differential Equations, 20(1):140–163, 2004.
  38. [38]38 Lorenzo Minatti and Enio Paris. A sph model for the simulation of free surface granular flows in a dense regime. Applied Mathematical Modelling, 39(1):363–382, 2015.DOI
  39. [39]39 O Pouliquen, C Cassar, P Jop, Y Forterre, and M Nicolas. Flow of dense granular material: towards simple constitutive laws. Journal of Statistical Mechanics: Theory and Experiment, 2006(07):P07020, Jul 2006.
  40. [40]40 Pierre Jop, Yoël Forterre, and Olivier Pouliquen. A constitutive law for dense granular flows. Nature, 441(7094):727–730, Jun 2006.
  41. [41]41 G. Fourtakas and B.D. Rogers. Modelling multi-phase liquid-sediment scour and resuspension induced by rapid flows using smoothed particle hydrodynamics (sph) accelerated with a graphics processing unit (gpu). Advances in Water Resources, 92:186–199, 2016.DOI
  42. [42]42 M. Khanpour, A.R. Zarrati, M. Kolahdoozan, A. Shakibaienia, and S.M. Amirshahi. Mesh-free sph modeling of sediment scouring and flushing. Computers & Fluids, 129:67–78, 2016.DOI
  43. [43]43 Huangjiang Zhu, Nicos S. Martys, Chiara Ferraris, and Daniel De Kee. A numerical study of the flow of bingham-like fluids in two-dimensional vane and cylinder rheometers using a smoothed particle hydrodynamics (sph) based method. Journal of Non-Newtonian Fluid Mechanics, 165 (7):362–375, 2010.DOI
  44. [44]44 Afonso Paiva, Fabiano Petronetto, Thomas Lewiner, and Geovan Tavares. Particle-based viscoplastic fluid/solid simulation. Computer-Aided Design, 41(4):306–314, 2009.DOI
  45. [45]45 Chongqiang Zhu, Yu Huang, and Liangzong Zhan. Sph-based simulation of flow process of a landslide at hongao landfill in china. Natural Hazards, 93(3):1113–1126, Sep 2018.DOI
  46. [46]46 E. Rossi, I. Garcia de Beristain, A. Vazquez-Quesada, J.E. López-Aguilar, and M. Ellero. Sph simulations of thixoviscoplastic fluid flow past a cylinder. Journal of Non-Newtonian Fluid Mechanics, 308:104891, 2022. ISSN 0377-0257.DOI
  47. [47]47 A. Vázquez-Quesada, N. J. Wagner, and M. Ellero. Planar channel flow of a discontinuous shear-thickening model fluid: Theory and simulation. Physics of Fluids, 29(10):103104, October 2017.
  48. [48]48 Samuel H. Maron and Percy E. Pierce. Application of re-eyring generalized flow theory to suspensions of spherical particles. Journal of Colloid Science, 11(1):80–95, 1956.
  49. [49]49 Joseph P. Morris, Patrick J. Fox, and Yi Zhu. Modeling low reynolds number incompressible flows using sph. Journal of Computational Physics, 136(1):214–226, 1997. ISSN 0021-9991. doi:https://doi.org/10.1006/jcph.1997.5776. URL https://www.sciencedirect.com/science/article/pii/S0021999197957764.
  50. [50]50 S. Adami, X.Y. Hu, and N.A. Adams. A generalized wall boundary condition for smoothed particle hydrodynamics. Journal of Computational Physics, 231(21):7057–7075, 2012. ISSN 0021-9991.
  51. [51]51 Adolfo Vázquez-Quesada, Pep Español, Roger I. Tanner, and Marco Ellero. Shear thickening of a non-colloidal suspension with a viscoelastic matrix. Journal of Fluid Mechanics, 880:1070–1094, 2019.
  52. [52]52 Michiel Hermes, Wilson C. K. Guy, Ben M.and Poon, Guilhem Poy, Michael E. Cates, and Matthieu Wyart. Unsteady flow and particle migration in dense, non-brownian suspensions. Journal of Rheology, 60(5):905–916, 2016.DOI
  53. [53]53 Ivo R Peters, Sayantan Majumdar, and Heinrich M Jaeger. Direct observation of dynamic shear jamming in dense suspensions. Nature, 532(7598):214–217, April 2016. ISSN 0028-0836.DOI
  54. [54]54 Deren Ozturk, Miles L. Morgan, and Bjørnar Sandnes. Flow-to-fracture transition and pattern formation in a discontinuous shear thickening fluid. Communications Physics, 3:1–9, 2020.DOI
  55. [55]55 Suzanne M. Fielding. Complex dynamics of shear banded flows. Soft Matter, 3:1262–1279, 2007.

Paper details

vlnc:W_2ydh2mjd
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