betaFeedback

WorksPhysics and Astronomy

Pulsatile poromechanics in layered soft media controls fluid flow and solute transport: from fundamentals to brain clearance

Abstract

Soft porous media often feature a heterogeneous structure. Notably, biological tissues - such as cartilage and the brain tissue - consist of two or more layers, with varying mechanical and fluid-flow properties. Despite the ubiquity of periodic loading in these systems, the physical implications of layering on nonlinear poromechanics and solute transport remain poorly understood. Uncovering these coupled mechanisms could clarify the fundamental physics behind pressing topics, such as brain metabolic clearance. Here, we address this gap using a bilayer model of a generic soft porous medium. To isolate the specific role of layering, we select combinations of material properties (porosity, permeability, and p-wave modulus) that maintain an identical poroelastic timescale, TPET_{\mathrm{PE}}, across four layered configurations and a reference homogeneous case. We demonstrate that while TPET_\mathrm{PE} is the key parameter governing the response in a homogeneous medium, the same TPET_\mathrm{PE} leads to non-trivial localization/propagation patterns for strain, fluid flow and solute transport in a bilayer medium. These non-intuitive results suggest that layered architectures may provide functional benefits for cellular homeostasis over homogeneous ones. Finally, we show that pathological alterations to the brain's layered structure significantly disrupt fluid-flow and metabolic waste clearance, offering a possible mechanical explanation for impaired transport in disease.

Introduction

Numerous soft porous media feature a layered structure. In biological tissues, layered architectures can arise from collagen fibre alignment, which provide increased strength and elasticity along preferred load-bearing directions [1], or from differences in material composition. The brain, for example, is divided into grey and white matter at the macro-scale, while, at the meso-scale, perivascular spaces — regions of increased porosity surrounding blood vessels — form high permeability pathways, facilitating cerebrospinal fluid flow and thus playing an essential role in brain clearance from metabolic waste [2, 3].

Such division in layers provides space-dependent mechanical and fluid-flow properties, such as solid matrix stiffness, porosity and permeability. This is well studied in several different tissues, from articular cartilage [4, 5] to the brain at the macro [6] and meso-scale [7, 8]. In cartilage, these layers directly influence solute transport [9], as depth-dependent properties enhance advective and diffusive fluxes at the local layer scale [13] [10, 11, 12]. Furthermore, layers in soft tissues may provide functional advantages at the whole tissue scale [14]: in tissue-engineering, heterogeneous scaffolds with layered or permeability-graded architectures promote cell growth compared to homogeneous designs [15, 16, 17]. Conversely, disrupting organisation can have deleterious consequences: in osteoarthritis, damage to superficial cartilage layers can trigger feedback mechanisms that exacerbate global tissue degeneration [18, 19]. Guided by the analogy with layer-dependent bio-poromechanics in cartilage, we investigate the controversial biophysics of brain clearance: we hypothesise that layers in the brain could play a role in enhancing fluid-flow and waste transport, and that pathological modifications may disrupt them. Abnormal accumulations of metabolic waste, such as amyloid-β\beta associated with Alzheimer’s disease and other neuropathologies, indeed lead to alterations of the grey matter properties [20, 21, 22, 23, 24], which may disrupt the poromechanical coupling with the adjacent layer (white matter at the macro-scale and/or perivascular spaces at the meso-scale). As in damaged cartilage, this may alter the overall deformation driven by blood vessel pulsations, well known as regulators of brain clearance [25, 26] and trigger further toxic waste accumulation.

Understanding these complex physiological systems requires extending the framework of nonlinear poromechanics to account for macroscopic heterogeneities (layers). Rather than aiming for a homogenized multilayer medium [27, 28, 29], our objective is to capture the specific effects of macroscopic heterogeneity itself. While the response of homogeneous media to static [30, 31, 32] and periodic loading [33, 34] — and the resulting solute transport [35] — is well-characterised, the fundamental role of macroscopic heterogeneities remains largely unexplored. Periodic loading is particularly relevant here, not only for its biological ubiquity but also for the unique space-dependent squishy dynamics it triggers in soft media [36]. Moreover, the spatial cyclic poromechanical response can be significantly altered by local degradation [37] or reinforcement [38] of material properties. Building on this, we propose that macroscopic heterogeneities, such as layering, fundamentally reshape the distribution of stress, strain, fluid flow, and the resultant solute transport under large-deformation oscillations. Elucidating these effects is critical to understanding the homeostatic equilibrium of layered biological tissues.

This study is organized as follows. We begin by isolating the structural influence of layering on the overall poromechanical response under periodic loading conditions. To this end, we adopt a simplified bilayer configuration of equal thickness. By fixing the poroelastic timescale across all setups, we systematically evaluate four distinct layered cases against a baseline homogeneous medium. Building upon these mechanical insights, we then evaluate how these architectural variations modulate solute transport dynamics. Finally, we demonstrate the practical utility of our findings by applying the developed framework to a realistic physiological system: the cerebral grey matter–white matter bilayer. Evaluating this specific system provides essential insights into the physical mechanisms governing brain clearance, offering a foundation for understanding how they are disrupted in pathological conditions.

Isolating the role of layering: definition of the case studies

In homogeneous media subject to pulsatile loads, for fixed loading amplitude AA and period TT, the material response is driven by the poroelastic timescale:

Tpe=μL2k0M,(1)T_{\mathrm{pe}} = \frac{\mu L^2}{k_0\mathcal{M}}, \tag*{(1)}

defined by μ\mu, the fluid viscosity, LL the domain length, and k0k_0 and M\mathcal{M}, the material resting permeability and pp-wave modulus, respectively. Importantly, the initial porosity Φ0\Phi_0 does not impact TpeT_{\mathrm{pe}}. Varying TpeT_{\mathrm{pe}} significantly modifies the spatial dependence of the material response, with two asymptotic regimes: a slow-loading regime (T≫TpeT \gg T_{\mathrm{pe}}), with no strain/flow localisation, and a fast-loading regime (T≫TpeT \gg T_{\mathrm{pe}}) with emergence of localisation [33]. To isolate the role of macroscopic heterogeneities, we filter out these TpeT_{\mathrm{pe}}-dependent effects, by varying the individual layer properties under two constraints: (i) each layer is assigned the same TpeT_{\mathrm{pe}} (through a constant product k0Mk_0\mathcal{M}) to prevent contrasting mechanical regimes within the material halves; and (ii) a constant effective macroscopic TpeT_{\mathrm{pe}} is maintained across the entire material, computed via the harmonic average of k0k_0 and M\mathcal{M} across the layers. We consider four scenarios (Figure 1): we either co-vary M\mathcal{M} and k0k_0 at constant Φ0=0.5\Phi_0 = 0.5, or vary Φ0\Phi_0 at fixed M=k0=1\mathcal{M} = k_0 = 1. Table I lists the resulting layer-specific parameters. We set M\mathcal{M}, k0k_0, Φ0\Phi_0 individually, as the interdependencies between these parameters are non-trivial given the complex microstructure of soft biological tissues [20, 39]. For instance, an increased Φ0\Phi_0 does not intrinsically imply a higher k0k_0, if the restrictive dimensions of the pore throats remain unchanged [40, 41]. Notably, while we treat the initial parameters as independent, the deformation-dependent permeability is modelled by a Kozeny-Carman relation (see equation (5) below), where the current permeability k(X,ϕf)k(X,\phi_f) evolves as a function of the current true fluid fraction ϕf\phi_f relative to Φ0\Phi_0, with k0≡k(X,ϕf,0)k_0 \equiv k(X,\phi_{f,0}).

Image not extracted

Figure 1. Schematic representation of a layered medium subject to periodic, displacement-driven load, through a permeable piston (left boundary), whereas the right boundary is fixed and impermeable. The concentration of solutes is C=1C = 1 in the domain (X∈[0,L]X \in[0,L]), and C=0C = 0 outside. Below, we highlight the four bilayer cases studied (properties not to scale).

CaseLayer 1 (M,k0,Φ0\mathcal{M}, k_0, \Phi_0)Layer 2 (M,k0,Φ0\mathcal{M}, k_0, \Phi_0)
Homogeneous1, 1, 0.51, 1, 0.5
(i.a)3, 0.6, 0.50.6, 3, 0.5
(i.b)0.6, 3, 0.53, 0.6, 0.5
(ii.a)1, 1, 0.251, 1, 0.75
(ii.b)1, 1, 0.751, 1, 0.25

Table 1. Summary of the dimensionless parameters explored in each case, as schematised in Figure 1.

The domain consists of a bilayer medium of total length LL, partitioned into two subregions of equal thickness: 0≤X≤L/20 \le X \le L/2 and L/2≤X≤LL/2 \le X \le L. Note that, while the length ratio undoubtedly influences the overall poromechanical behaviour during simultaneous parameter shifts, the current study is designed to isolate the specific effects of layered heterogeneities.

This simplified configuration serves as an optimal baseline for this objective.

The medium is subject to a periodic displacement of period TT at a permeable left boundary, while the right boundary is fixed and impermeable. The initial symmetry in the layers does not imply equal thickness during loading, as the spatial variations in poroelastic properties induces non-uniform deformations in the two layers. To avoid interfacial discontinuities, we smoothly transition each initial property f(X,0)f(X,0) (representing M\mathcal{M}, k0k_0, or Φ0\Phi_0) between its respective values f1f_1 and f2f_2 using a hyperbolic tangent function

f(X,0)=f1+f2−f12[1+tanh⁡(κ(X−L2))],f(X,0) = f_1 + \frac{f_2-f_1}{2}\left[1+\tanh\left(\kappa\left(X-\frac{L}{2}\right)\right)\right],

where the steepness coefficient is set to κ=100\kappa=100 to ensure a rapid yet differentiable transition across the interface at X=L/2X=L/2 (see Figure 1). Given such spatial variations, we reformulate the model from [30] in Lagrangian coordinates, following [37, 42].

Model

In this section, we summarise the equations defining the model for poromechanics and solute transport in Lagrangian coordinates.

Kinematics

The Lagrangian solid displacement field is Us(X,t)=x(X,t)−X\mathbf{U}_s(X,t)=\mathbf{x}(X,t)-X, with x(X,t)\mathbf{x}(X,t) the current position of the material point with reference position XX. Our reference configuration is the relaxed configuration, where x(X,0)=X\mathbf{x}(X,0)=X, so that Us(X,0)=0\mathbf{U}_s(X,0)=0. Given the Jacobian determinant J=1+∂Us/∂XJ=1+\partial U_s/\partial X, for incompressible constituents and uniform initial porosity Φf,0\Phi_{f,0}, the local change in volume relates to the change in porosity as

J(X,t)=1−Φf,0+Φf→∂Us∂X=Φf−Φf,0,(2)J(X,t)=1-\Phi_{f,0}+\Phi_f \quad\rightarrow\quad\frac{\partial U_s}{\partial X}=\Phi_f-\Phi_{f,0}, \tag*{(2)}

where the nominal porosity Φf=Jϕf\Phi_f=J\phi_f measures the current fluid volume per unit reference total volume and is expressed as function of the true porosity ϕ\phi.

Darcy's law

Mass conservation for the fluid phase reads:

∂Φf∂t+∂∂X(Wf)=0,(3)\frac{\partial\Phi_f}{\partial t}+\frac{\partial}{\partial X}(W_f)=0, \tag*{(3)}

where WfW_f is the nominal flux of fluid through the solid skeleton. By definition, WfW_f is given by Darcy’s law, which reads

Wf=−1Jk(X,ϕf)μ∂p∂X,(4)W_f=-\frac{1}{J}\frac{k(X,\phi_f)}{\mu}\frac{\partial p}{\partial X}, \tag*{(4)}

where pp is the pore pressure. For the permeability law, we choose a normalised Kozeny-Carman relation of the form

k(ϕf)=k0(1−ϕf,0)2ϕf,03ϕf3(1−ϕf)2,(5)k(\phi_f)=k_0\frac{(1-\phi_{f,0})^2}{\phi_{f,0}^3}\frac{\phi_f^3}{(1-\phi_f)^2}, \tag*{(5)}

where k0≡k(X,ϕf,0)k_0 \equiv k(X,\phi_{f,0}) is the permeability of the initial state, as mentioned above.

Mechanical equilibrium

In 1D, mechanical equilibrium reads:

∂s∂X=0,(6)\frac{\partial s}{\partial X}=0, \tag*{(6)}

where ss is the nominal total stress. The nominal effective stress s′s' is then:

s′=s+p.(7)s'=s+p. \tag*{(7)}

We take the elasticity law to be Hencky elasticity [43]:

s′=Mln⁡(J)J.(8)s'=M\frac{\ln(J)}{J}. \tag*{(8)}

Fluid flow

Combining mass conservation, Darcy’s law, and mechanical equilibrium, the final nonlinear flow equations read:

∂Φf∂t−∂∂X[1Jk(X,ϕf)μ∂p∂X]=0and∂s′∂X=∂p∂X.(9)\frac{\partial\Phi_f}{\partial t}-\frac{\partial}{\partial X}\left[\frac{1}{J}\frac{k(X,\phi_f)}{\mu}\frac{\partial p}{\partial X}\right]=0 \quad\text{and}\quad \frac{\partial s'}{\partial X}=\frac{\partial p}{\partial X}. \tag*{(9)}

Note that the fluid and solid velocities (vfv_f and vsv_s respectively) are the same in Eulerian and Lagrangian space. Since

Wf=ϕf(vf−vs),(10)W_f=\phi_f(v_f-v_s), \tag*{(10)}

it is easy to derive that

vs=−Wfandvf=−(1−ϕfϕf)vs.(11)v_s=-W_f \quad\text{and}\quad v_f=-\left(\frac{1-\phi_f}{\phi_f}\right)v_s. \tag*{(11)}

With appropriate initial and boundary conditions, the above formulations close the Lagrangian model.

Solute transport

The model for solute transport is analogous to the one presented in [35]. In a Lagrangian frame, conservation of mass for the solute reads:

∂(ΦfC)∂t+∂∂X[CWf−1JϕfD∂C∂X]=0,withD=Dm+Dh=Dm+α∣Wfϕf∣(12)\frac{\partial(\Phi_f C)}{\partial t}+\frac{\partial}{\partial X}\left[CW_f-\frac{1}{J}\phi_f\mathcal{D}\frac{\partial C}{\partial X}\right]=0, \qquad\text{with}\quad\mathcal{D}=\mathcal{D}_m+\mathcal{D}_h=\mathcal{D}_m+\alpha\left|\frac{W_f}{\phi_f}\right| \tag*{(12)}

where CC represents the intrinsic solute concentration and the diffusivity D\mathcal{D} is written in terms of the coefficients for molecular diffusion Dm\mathcal{D}_m and for hydrodynamic dispersion Dh\mathcal{D}_h using the dispersivity α\alpha. The total solute flux is

Wc=CWf−1JϕfD∂C∂X.(13)W_c=CW_f-\frac{1}{J}\phi_f\mathcal{D}\frac{\partial C}{\partial X}. \tag*{(13)}

Boundary conditions

We define the left and right boundaries of the solid to be at X=0X=0 and X=LX=L, respectively, with initial conditions for the mechanics Φf(X,0)=ϕf,0\Phi_f(X,0)=\phi_{f,0} and Us(X,0)=Wf(X,0)=0U_s(X,0)=W_f(X,0)=0. For the solute, in analogy with brain clearance, we initially impose C(X,0)=1C(X,0)=1 throughout the domain, meaning that solutes are initially present inside the medium and can escape through the left boundary into a solute-free fluid reservoir. At the left boundary, we apply a periodic displacement a(t)=A2[1−cos⁡(2πtT)]a(t)=\frac{A}{2}\left[1-\cos\left(\frac{2\pi t}{T}\right)\right], with boundary conditions:

vs(X=0,t)=a˙(t),p(X=0,t)=0(14)v_s(X=0,t)=\dot{a}(t),\qquad p(X=0,t)=0 \tag*{(14)}

for the mechanics, and

C(0,t)=0,(15)C(0,t)=0, \tag*{(15)}

for solute transport. The right boundary is impermeable and fixed in place, such that

Us(L,t)=Wf(L,t)=0,(16)U_s(L,t)=W_f(L,t)=0, \tag*{(16)}

and

∂C∂X∣X=L=0.(17)\left.\frac{\partial C}{\partial X}\right|_{X=L}=0. \tag*{(17)}

We fix the values of the loading amplitude and period to typical mean values for biological media, with A/L=0.05A/L=0.05 and T/Tpe=1T/T_{\mathrm{pe}}=1 (see [33]); analogously, we take a typical dimensional diffusion coefficient Dm=10−10 m2/s\mathcal{D}_m=10^{-10}\,\mathrm{m}^2/\mathrm{s} [35], and we vary the value of α\alpha.

Scaling

We report here the model in dimensionless form. The evolution equations are nondimensionalised according to:

X~=XL,U~s=UsL,t~=tTpe,s~′=s′M,p~=pM,k~=k(ϕ)k0,W~f=WfL/Tpe.(18)\begin{aligned} \tilde{X} &= \frac{X}{L}, \quad\tilde{U}_s = \frac{U_s}{L}, \quad\tilde{t} = \frac{t}{T_{pe}}, \quad\tilde{s}' = \frac{s'}{\mathcal{M}}, \quad\tilde{p} = \frac{p}{\mathcal{M}}, \\ \tilde{k} &= \frac{k(\phi)}{k_0}, \quad\tilde{W}_f = \frac{W_f}{L/T_{pe}}. \tag*{(18)} \end{aligned}

for the poromechanics part, and

C~=CC0,α~=αL,D~=DDm=1+α~Pe∣W~fϕf∣(19)\tilde{C} = \frac{C}{C_0}, \quad\tilde{\alpha} = \frac{\alpha}{L}, \quad\tilde{\mathcal{D}} = \frac{\mathcal{D}}{D_m} = 1 + \tilde{\alpha}\mathrm{Pe}\left|\frac{\tilde{W}_f}{\phi_f}\right| \tag*{(19)}

for the solute transport part, where Pe=(L2/Tpe)/Dm\mathrm{Pe} = (L^2/T_{pe})/D_m is the Péclet number. The full problem can then be rewritten in dimensionless form as

∂Φf∂t~−∂∂X~(W~f)=0(20)\frac{\partial\Phi_f}{\partial\tilde{t}}-\frac{\partial}{\partial\tilde{X}}\left(\tilde{W}_f\right)=0 \tag*{(20)}

and

∂(ΦfC~)∂t~+∂∂X~[C~W~f−1JϕfPe−1D~∂C~∂X~]=0,(21)\frac{\partial(\Phi_f\tilde{C})}{\partial\tilde{t}}+\frac{\partial}{\partial\tilde{X}}\left[\tilde{C}\tilde{W}_f-\frac{1}{J}\phi_f\mathrm{Pe}^{-1}\tilde{\mathcal{D}}\frac{\partial\tilde{C}}{\partial\tilde{X}}\right]=0, \tag*{(21)}

with initial conditions

a~(0)=0,Φf(X~,0)=ϕf,0,W~f(X~,0)=v~s(X~,0)=0,C~(X~,0)=1.(22)\begin{aligned} \tilde{a}(0) &= 0,\quad\Phi_f(\tilde{X},0)=\phi_{f,0},\quad\tilde{W}_f(\tilde{X},0)=\tilde{v}_s(\tilde{X},0)=0,\\ \tilde{C}(\tilde{X},0)&=1. \tag*{(22)} \end{aligned}

The scaled left boundary conditions read:

U~s(0,t~)=a~(t~)=A~2[1−cos⁡(2πt~T~)],W~f(0,t~)=−a~˙(t~)andC~(0,t~)=0,(23)\begin{aligned} \tilde{U}_s(0,\tilde{t})=\tilde{a}(\tilde{t})&=\frac{\tilde{A}}{2}\left[1-\cos\left(\frac{2\pi\tilde{t}}{\tilde{T}}\right)\right],\\ \tilde{W}_f(0,\tilde{t})&=-\dot{\tilde{a}}(\tilde{t})\quad\text{and}\quad\tilde{C}(0,\tilde{t})=0, \tag*{(23)} \end{aligned}

where A~=A/L\tilde{A}=A/L and T~=T/Tpe\tilde{T}=T/T_{pe}. The right boundary conditions are

U~s(1,t~)=W~f(1,t~)=0and∂C~∂x~∣x~=1=0.(24)\tilde{U}_s(1,\tilde{t})=\tilde{W}_f(1,\tilde{t})=0\quad\text{and}\quad\left.\frac{\partial\tilde{C}}{\partial\tilde{x}}\right|_{\tilde{x}=1}=0. \tag*{(24)}

In the text, we refer to dimensionless values dropping the tildas for convenience everywhere except for the example of brain GM and WM, where dimensional parameters are essential to define the problem.

We solve our model numerically in MATLAB, using compact finite differences in space, with implicit Runge-Kutta integration in time, which we achieve via MATLAB’s built-in solver ODE15s [44]. The full description of the numerical method and the converge analsysi can be found in [35].

Results

Impact of layers on poromechanics

We evaluate the impact of macroscopic heterogeneity on poromechanics and solute transport by comparing five configurations: a homogeneous control and the four bilayer cases described in table I. Figure 2 displays the corresponding displacement UsU_s, solid stress s′s', and local change in porosity Φf/Φ0\Phi_f/\Phi_0, and Figure 3 the fluid flux WfW_f, fluid velocity vfv_f (both normalised by v⋆=2A/Tv^\star= 2A/T), and solute flux WcW_c. The presence of layers fundamentally alters the spatial distribution of mechanical and transport quantities. Notably, the system response is not solely a function of its average properties, but is critically dependent on the spatial arrangement of the layers relative to the pulsatile boundary. In details, in case (i.a), the first layer’s high impedance (↑M,↓k0\uparrow\mathcal{M}, \downarrow k_0) resists local deformation, generating high solid stresses that effectively transmit strain into the second layer (Figure 2(a-Figure 2f-Figure 2k)). This propagation effect amplifies fluid and solute fluxes deep into the medium (Figure 3(a-Figure 3f-Figure 3k)). Conversely, in case (i.b), the high deformability of the superficial layer (↓M\downarrow\mathcal{M}, ↑k0\uparrow k_0) localises strains and fluid flow near the piston, significantly attenuating distal solute transport (Figure 2(b-Figure 2g-Figure 2l) and Figure 3(b-Figure 3g-Figure 3l)).

Image not extracted

Figure 2. Space distribution of the normalised displacement UsU_s, solid stress s′s', and local variation in porosity Φf/Φ0\Phi_f/\Phi_0 at ten equally spaced time-steps in a loading cycle (dark to light), during loading vs unloading (solid vs dashed lines), for the five cases of bilayer media summarised in table I.

Image not extracted

Figure 3. Space distribution of the normalised fluid flux WfW_f, fluid velocity vfv_f, and solute flux WcW_c at ten equally spaced time-steps in a loading cycle (dark to light), during loading vs unloading (solid vs dashed lines), for the five cases of bilayer media summarised in table I. WcW_c is computed at the second cycle, with α=0.05\alpha= 0.05.

Variations in Φ0\Phi_0 (series ii) do not impact displacement and fluid flux spatial propagation. However, a lower proximal porosity (ii.a) enhances localization (Figure 2(i-Figure 2n)), characterised by higher velocities and solid stresses in the proximal layer, but diminished total solute transport (Figure 3(i-Figure 3n)). In contrast, increasing proximal porosity (ii.b) mitigates these localizations compared to the homogeneous case, thereby enhancing overall solute clearance (Figure 2(j-Figure 2o) and Figure 3(j-Figure 3o)).

Impact of layers on solute transport

In Figure 4, we quantify the impact of heterogeneities on solute transport. Considering 20 loading cycles, we show the final concentration profile across the domain (Figure 4(a-Figure 4b)) and the total solute amount over time (Figure 4(c-Figure 4d)). For each case, we consider diffusion only, advection-diffusion (no dispersion, α=0\alpha= 0), and advection-diffusion-dispersion (α=0.05\alpha= 0.05). Cases (i.a) and (ii.b) show a lower final solute concentration compared to the homogeneous case, i.e. enhanced clearance. On the other hand, case (i.b) and (ii.a) display higher solute concentration, highlighting an overall reduced solute transport. In series (i), such transport

Image not extracted

Figure 4. Solute transport dynamics for the five cases of bilayer media summarised in table I. Final concentration profiles (a)-(b) and normalised total solute mass over time (c)-(d) for mechanical property variations (a)-(c) and porosity variations (b)-(d), compared against the homogeneous baseline. Dotted lines show pure diffusion (α=0\alpha= 0, A=0A = 0). Dashed and solid lines represent advection-diffusion at α=0\alpha= 0 and advection-diffusion-dispersion at α=0.05\alpha= 0.05, respectively.

dynamics are intuitive: proximal resistance to deformation reduces localisation in (i.a) vs (i.b), enhancing solute transport, with no impact in pure diffusion, as expected. Conversely, series (ii) reveals a more complex interplay of poromechanical and diffusive effects. Even in the diffusion-only limit, the higher porosity near the exit in (ii.b) accelerates clearance by increasing the concentration gradient. Under periodic loading, moreover, series (ii) exhibits counter-intuitive behaviour: proximal resistance (low Φ0\Phi_0 near the piston) increases localization in fluid volume changes, further hindering transport in (ii.a). Consequently, transport physics is dictated by the resulting spatial distribution of fluid volume changes rather than proximal deformability. Configurations (i.a) and (ii.b) are the most favourable for clearance; by minimizing localization and shifting deformation toward the distal end, these setups maximize fluid and solute fluxes deep into the tissue.

Biological bilayer system: brain grey and white matter

We next implement this bilayer poromechanical analysis in a case study of amyloid-β\beta transport within the grey matter (GM) and white matter (WM) system in the brain. In this part, TpeT_{\mathrm{pe}} is not constant anymore and depends on the tissue parameters. On average, the thickness of the GM is 2.5 mm2.5\,\mathrm{mm} [45], and its surface is subject to pulsations from pial arteries. Such pulsations have the largest amplitude at the frequency of vasomotion [25, 46], and are likely to propagate from the GM surface to the WM surface, since parenchymal volumetric strain linked to cardiac pulsations is documented in those regions [47, 48, 49]. Hence, we define our domain as a bilayer of total length 5 mm5\,\mathrm{mm}, equally partitioned between GM and WM. We take T=10 sT = 10\,\mathrm{s} (dimensional value, corresponding to the typical vasomotion frequency of 0.1 Hz0.1\,\mathrm{Hz}) and A=15 μmA = 15\,\mathrm{\mu m}, since vasomotion leads to strains in the order of 1515–20%20\% of the vessel diameter (with typical diameter values of 200200–250 μm250\,\mathrm{\mu m} [50]). Reference properties of healthy GM and WM are taken from literature (post-mortem experiments) and are summarised in table II. The ratio MGM\mathcal{M}_{GM} to MWM\mathcal{M}_{WM} is assumed to be the same as reported for the Young modulus [6]. For the pathological changes, we focus on the impact of amyloid-β\beta accumulations, characterising the GM only; hence, we do not consider any changes in the WM properties. We assume that pathological alterations of the GM lead to: (i) a decrease in tissue stiffness, due to neuronal death [22, 23, 24]; (ii) a decrease in permeability, due to the progressive clogging of the interstitial fluid space [51, 52], usually reported as increased tortuosity [20, 21]; and (iii) a progressive decrease in porosity, due to reductions in functional pore space connectivity. Assumption (iii) is related to the experimental observations that amyloid plaques create “dead-space” microdomains that trap extracellular water [20] and that inflamed astrocytes swell [53], reducing the effective GM fluid fraction actively

TissueM\mathcal{M} [kPa]k0k_0 [m2\mathrm{m}^2]Φ0\Phi_0TpeT_{\mathrm{pe}} [s] (single layer)
WM - Healthy250 [59]5×10−155 \times10^{-15}0.2 [60]5
GM - Healthy0.7×MWM0.7 \times\mathcal{M}_{WM} [6] = 1754×10−154 \times10^{-15} [61]0.22 [60]9
GM - P1131.252×10−152 \times10^{-15}0.1524
GM - P287.51×10−151 \times10^{-15}0.1170

caption: TABLE II. Values considered in the model for WM, GM, and the GM pathological alterations P1 and P2 due amyloid-β\beta plaques. Note that in the Pathological GM cases, alterations of the parameters are speculative as they are challenging to find in literature.

:::

participating in clearance. Together with the increased cortical free water observed via MRI in Alzheimer’s disease [54], these observations imply that while more extracellular water is present, a larger portion remains stagnant. In details, for the GM pathological states P1 and P2, we assume M\mathcal{M} and k0k_0 to be respectively half and one quarter of their healthy value; Φ0\Phi_0 is reduced of one third and half of the reference healthy value (as summarised in table II).

For the solute transport, we take the dimensional amyloid-β\beta diffusion coefficient to be Dm=6×10−11m2/sD_m = 6 \times10^{-11}\mathrm{m}^2/\mathrm{s} [55]. Values for the dispersivity α\alpha usually scale with the size of the sources of heterogeneity in the medium. In a homogeneous gel, for example, α\alpha is set up by the size of the gel pores [56]. At the meso-scale considered here, the brain tissue is highly heterogeneous due to the presence of several different structures (such as, blood vessels of different sizes, perivascular spaces, cells). Hence, we choose dimensional dispersivity values of α=50\alpha= 50 and 250 μm250\,\mu\mathrm{m}, ranging from the typical length-scale of a single capillary vessel (50 μm50\,\mu\mathrm{m}) to the characteristic spacing between large penetrating vessels and their perivascular space (100100 to 300 μm300\,\mu\mathrm{m}), to account for the macroscale hydrodynamic spreading [57, 58].

Figure 5(a-Figure 5b-Figure 5c) summarises the impact of the pathological changes on the mechano-fluid properties of the GM-WM system. The progressive decrease in stiffness and permeability in the GM layer leads to progressively stronger localisation of the fluid flux at the surface: in the healthy condition, WfW_f spreads through the WM, while in the pathological cases it is almost entirely suppressed in the WM and strongly reduced in the distal portions of the GM. Such changes in fluid-flow distribution impact on solute transport dynamics (Figure 5(d)), noticeably impaired in both pathological cases. Given the narrow physio-pathological range assumed for Φ0\Phi_0 (see table II), its impact is minor and only marginally observable in the absence of dispersion (α=0\alpha= 0). We consider this scenario to be physiologically improbable, however, as the profound microstructural heterogeneity of brain tissue [20, 39] is expected to induce significant hydrodynamic dispersion [35].

Image not extracted

Figure 5. Poromechanics and solute transport of the GM-WM system, with healthy vs pathological properties (see table II). Space distribution of WfW_f (a)-(b)-(c) (dimensional), at equally spaced time-steps in a loading cycle (dark to light), during loading vs unloading (solid vs dashed lines). Temporal evolution of the normalised solute mass (d) for varying α\alpha, with α=0\alpha= 0 (dotted lines), α=0.01\alpha= 0.01 (dashed lines) and α=0.05\alpha= 0.05 (solid lines).

Conclusions

In this work, we have explored the pulsatile poromechanics of layered soft porous media, and its implications on solute transport. Our results reveal a clear dependence on the spatial sequencing of the layers relative to the loading boundary. This asymmetry arises from the phase lag between tissue deformation and pore-pressure relaxation, suggesting that the order of tissue heterogeneities is as critical as the properties themselves. Such interactions among layers lead to complex distributions of fluid fractions, stresses, and fluid fluxes, which are not achievable through classic changes in mechanical regimes (governed by the poroelastic timescale [33]) in a homogeneous medium.

Furthermore, our findings demonstrate that solute transport dynamics is strongly influenced by bilayer architectures. This insight is of fundamental significance for modelling transport phenomena in biological media. From a biophysical aspect, these results suggest that the division in layers could hence be a strategy for each tissue to reach an efficient response to fixed oscillating boundary conditions, such as blood vessel pulsations. Moreover, these results could guide the design of “smart” tissue engineering scaffolds: tuning their local stress/flow while controlling local transport of nutrients and/or waste substances could dramatically improve cell survival and proliferation.

In the final section of this work, we extend the developed framework to a physiologically relevant case: the grey and white matter bilayer of cerebral tissue. In this case, the empirical parameters sourced from literature naturally deviate from the uniform poroelastic timescales assumed in our theoretical analysis, but the model presented here is still fully valid. Although simplified, our 1D approach offers a fundamentally new perspective on brain clearance by isolating the interfacial physics of bilayer poroelasticity. We show that altering the mechanical balance between layers — even in this idealized limit — is sufficient to reduce clearance efficiency. This suggests that the brain’s architecture is inherently sensitive to pathological shifts in material properties: a pathological GM is likely to further hinder clearance because of the altered poromechanical equilibrium, thus generating a self-reinforcing “vicious circle” for waste accumulation [62]. Importantly, while qualitatively based on published literature, the pathological changes assumed here are speculative and would require experimental verification to advance our understanding of brain poromechanics. Specifically, local in vivo measurements of permeability, stiffness, and porosity — across both healthy tissues and different stages of pathological progression — are essential to rigorously characterize the biophysical coupling between poromechanics and solute transport. In this regard, more precise assessments of the specific transport mechanisms happening in vivo, such as the dynamic interplay between deformation-dependent diffusion and dispersion, remain highly necessary to validate upscaled continuum models.

In future work, we plan to extend the bilayer parameter space exploration, and to focus on the upscaling of solute transport to retrieve effective transport estimations for each layered case. Furthermore, we believe it would be interesting to refine our model of the cerebral grey matter–white matter system by incorporating dual-porosity poroelasticity, a framework particularly suited for brain tissue [63].

Acknowledgements: This work was supported by a grant from the Leducq Foundation and the Leducq Foundation for Cardiovascular Research (23CVD03). We thank the members of the Leducq Network for the stimulating environment for exchanges and interdisciplinary discussions. We are grateful to C. W. MacMinn for the inspiring discussions and for the assistance with the formulation of the model in Lagrangian coordinates. We also thank S. Waters, D. Moulton, and Z. Godard for insightful discussions.

References

References

  1. [13]G. A. Ateshian and J. A. Weiss, Journal of Biomechanical Engineering 132, 1 (2010).
  2. [14]S. Hashemi, L. M. Amirabad, F. D. Nazhvani, P. Zarrintaj, H. Namazi, A. Saadatfar, and A. Golchin, Bilayer scaffolds for interface tissue engineering and regenerative medicine: A systematic reviews, in Cell Biology and Translational Medicine, Volume 14: Stem Cells in Lineage Specific Differentiation and Disease, edited by K. Turksen (Springer International Publishing, 2021) pp. 83–113.DOI
  3. [15]S. Camarero-Espinosa, B. Rothen-Rutishauser, C. Weder, and E. J. Foster, Biomaterials 74, 42 (2016).
  4. [16]T. Mesallati, C. T. Buckley, T. Nagel, and D. J. Kelly, Biomechanics and Modeling in Mechanobiology 12, 889 (2013).
  5. [17]K. Sano, Y. Ishida, and T. Aida, Angewandte Chemie - International Edition 57, 2532 (2018).
  6. [18]J. Hwang, W. C. Bae, W. Shieu, C. W. Lewis, W. D. Bugbee, and R. L. Sah, Arthritis & Rheumatism 58, 3831 (2008).
  7. [19]M. E. Stender, R. A. Regueiro, and V. L. Ferguson, Computer Methods in Biomechanics and Biomedical Engineering 20, 319 (2017).
  8. [20]S. Hrabětová and C. Nicholson, Neurochemistry International 45, 467 (2004).
  9. [21]E. Syková, I. Voříšek, T. Antonova, T. Mazel, M. Meyer-Luehmann, M. Jucker, M. Hájek, M. Or, and J. Bureš, Proceedings of the National Academy of Sciences 102, 479 (2005).
  10. [22]M. C. Murphy, J. Huston III, C. R. Jack Jr, K. J. Glaser, A. Manduca, J. P. Felmlee, and R. L. Ehman, Journal of Magnetic Resonance Imaging 34, 494 (2011).
  11. [23]M. C. Murphy, D. T. Jones, C. R. Jack, K. J. Glaser, M. L. Senjem, A. Manduca, J. P. Felmlee, R. E. Carter, R. L. Ehman, and J. Huston, NeuroImage: Clinical 10, 283 (2016).
  12. [24]L. V. Hiscox, C. L. Johnson, M. D. J. McGarry, H. Marshall, C. W. Ritchie, E. J. R. van Beek, N. Roberts, and J. M. Starr, Brain Communications 2, fcz049 (2019).
  13. [25]S. J. van Veluw, S. S. Hou, M. Calvo-Rodriguez, M. Arbel-Ornath, A. C. Snyder, M. P. Frosch, S. M. Greenberg, and B. J. Bacskai, Neuron 105, 549 (2020).
  14. [26]H. Mestre, J. Tithof, T. Du, W. Song, W. Peng, A. M. Sweeney, G. Olveda, J. H. Thomas, M. Nedergaard, and D. H. Kelley, Nature Communications 9 (2018).
  15. [27]R. K. N. D. Rajapakse and T. Senjuntichai, Earthquake Engineering & Structural Dynamics 24, 703 (1995).
  16. [28]A. M. Kudarova, K. N. van Dalen, and G. G. Drijkoningen, Geophysical Journal International 195, 1337 (2013).
  17. [29]T. Liu and C. Zhao, Soil Dynamics and Earthquake Engineering 48, 15 (2013).
  18. [30]C. W. MacMinn, E. R. Dufresne, and J. S. Wettlaufer, Physical Review Applied 5, 044020 (2016).
  19. [31]D. R. Hewitt, D. T. Paterson, N. J. Balmforth, and D. M. Martinez, Physics of Fluids 28, 063304 (2016).
  20. [32]D. R. Hewitt, J. S. Nijjer, M. G. Worster, and J. A. Neufeld, Physical Review E 93, 023116 (2016).
  21. [33]M. Fiori, S. Pramanik, and C. W. MacMinn, Journal of Fluid Mechanics 974, A2 (2023).
  22. [34]J. Cacheux, J. Ordonez-Miranda, A. Bancaud, L. Jalabert, D. Alcaide, M. Nomura, and Y. T. Matsunaga, Science Advances 9, eadf9775 (2023).
  23. [35]M. Fiori, S. Pramanik, and C. W. MacMinn, Journal of Fluid Mechanics 1009, A15 (2025).
  24. [36]M. G. Worster, Journal of Fluid Mechanics 980, F1 (2024).
  25. [37]Z. C. Godard, D. E. Moulton, and S. L. Waters, Soft Matter 21, 7487 (2025).
  26. [38]D. Alcaide, A. Bancaud, J. Cacheux, and Y. Matsunaga, bioRxiv (2026).
  27. [39]C. Nicholson, Journal of The Royal Society Interface 20, 20230223 (2023).
  28. [40]P. H. Nelson, The Log Analyst 35, 38 (1994).
  29. [41]A. J. Katz and A. H. Thompson, Phys. Rev. B 34, 8179 (1986).
  30. [42]M. Fiori, Impact of Large, Periodic Deformations on Flow and Transport in Soft Porous Media, Ph.D. thesis, University of Oxford (2023).DOI
  31. [43]H. Hencky, Rubber Chemistry and Technology 6, 217 (1933).
  32. [44]L. F. Shampine and M. W. Reichelt, SIAM Journal on Scientific Computing 18, 1 (1997).
  33. [45]B. Fischl and A. M. Dale, Proceedings of the National Academy of Sciences 97, 11050 (2000).
  34. [46]T. Broggini, J. Duckworth, X. Ji, R. Liu, X. Xia, P. Mächler, I. Shaked, L. P. Munting, S. Iyengar, M. Kotlikoff, S. J. van Veluw, M. Vergassola, G. Mishne, and D. Kleinfeld, Neuron 112, 2349 (2024).
  35. [47]A. L. Adams, H. J. Kuijf, M. A. Viergever, P. R. Luijten, and J. J. Zwanenburg, NMR in Biomedicine 32, e4050 (2019).
  36. [48]J. J. Sloots, G. J. Biessels, A. de Luca, and J. J. Zwanenburg, NeuroImage 236, 118078 (2021).
  37. [49]J.-J. Sloots, M. Froeling, G. J. Biessels, and J. J. M. Zwanenburg, Magnetic Resonance in Medicine 88, 266 (2022).
  38. [50]S. Bollmann, H. Mattern, M. Bernier, S. D. Robinson, D. Park, O. Speck, and J. R. Polimeni, eLife 11, e71186 (2022).
  39. [51]X. Yue, Y. Mei, Y. Zhang, Z. Tong, D. Cui, J. Yang, A. Wang, R. Wang, X. Fei, L. Ai, Y. Di, H. Luo, H. Li, W. Luo, Y. Lu, R. Li, C. Duan, G. Gao, H. Yang, B. Sun, R. He, W. Song, H. Han, and Z. Tong, Alzheimer’s & Dementia: Translational Research & Clinical Interventions 5, 671 (2019).
  40. [52]A. Iorio, S. Melchionna, P. Derreumaux, and F. Sterpone, PNAS Nexus 4, pgae548 (2024).
  41. [53]C. Nicholson and E. Sýková, Trends in Neurosciences 21, 207 (1998).
  42. [54]B. J. Hall, E. Aumont, S. A. Hosseini, J. F. Arias, A. Boré, G. Bezgin, L. Trudel, T. Chan, J. Therriault, A. C. Macedo, M. S. Woo, D. Oliva-Lopez, N. Rahmouni, Y. Zheng, S. Servaes, J. Stevenson, S. Gauthier, A. L. Benedet, M. Dumont, J.-C. Houde, G. Triana-Baltzer, H. C. Kolb, N. J. Ashton, H. Zetterberg, Y. I. Medina, P. Vitali, M. Descoteaux, J. M. Klostranec, T. Pascoal, and P. Rosa-Neto, Neurology 106, e214606 (2026).
  43. [55]J. Waters, PLOS ONE 5, 1 (2011).
  44. [56]D. H. Noh, A. H. Zadeh, H. Zhang, et al., Annals of Biomedical Engineering 52, 1693 (2024).
  45. [57]H. Duvernoy, S. Delon, and J. Vannson, Brain Research Bulletin 7, 519 (1981).
  46. [58]S. Lorthois, F. Cassot, and F. Lauwers, NeuroImage 54, 1031 (2011).
  47. [59]G. Franceschini, D. Bigoni, P. Regitnig, and G. Holzapfel, Journal of the Mechanics and Physics of Solids 54, 2592 (2006).
  48. [60]A. Lehmenkuhler, E. Syková, J. Svoboda, K. Zilles, and C. Nicholson, Neuroscience 55, 339 (1993).
  49. [61]A. Greiner, N. Reiter, J. Hinrichsen, M. P. Kainz, G. Sommer, G. A. Holzapfel, P. Steinmann, E. Comellas, and S. Budday, Interface Focus 14, 20240026 (2024).
  50. [62]S. M. Greenberg, B. J. Bacskai, M. Hernandez-Guillamon, J. Pruzin, R. Sperling, and S. J. van Veluw, Nat. Rev. Neurol. 16, 30 (2020).
  51. [63]G. Sciumé, Acta Mechanica 232, 1445 (2021).

Paper details

vlnc:W_pxptdjv2
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