Polymer Diffusive Instability (PDI)

• Polymer Diffusive Instability (PDI) is a diffusion-enabled instability in polymer flows characterized by transient stress overshoots and wall-localized shear bands.
• It is analyzed using linear stability methods within models like Rolie–Poly, Oldroyd-B, and FENE-P, revealing non-monotonic transient constitutive responses.
• Diffusive terms both regularize the governing equations and enable instability, influencing length-scale selection and numerical simulation fidelity.

Searching arXiv for recent and foundational papers on polymer diffusive instability and related usages. Polymer Diffusive Instability (PDI) denotes a class of diffusion-mediated instabilities in polymeric media, but the term is not used uniformly across subfields. In the rheology of entangled and dilute viscoelastic flows, it refers most specifically to an instability of the polymer conformation or stress field that appears only when a finite diffusion term is present in the constitutive dynamics, producing transient or wall-localized banded states even when the underlying steady constitutive response is homogeneous (Adams et al., 2010). In other polymer contexts, closely related usages of “diffusive instability” describe shear-induced demixing in blends, evaporation-driven pluming in drying solutions, diffusion-mediated freezing-interface destabilization, and surface-diffusion-driven stress instabilities in elastic films. The common structural element is coupling between polymer transport or conformation gradients and a driving field—shear, evaporation, solidification, or internal stress—so that diffusion acts not merely as a stabilizer but as a selector of length scales and, in some cases, an enabler of instability (Beneitez et al., 2022).

1. Terminological scope and principal meanings

Within contemporary viscoelastic-flow literature, PDI most often denotes a linear instability of wall-bounded shear flows that requires finite polymer diffusion in the conformation tensor equation. In this usage, the instability is wall-localized, exists in planar geometries, and disappears when polymer diffusion is removed from Oldroyd-B or FENE-P dynamics (Beneitez et al., 2022). A closely related but earlier manifestation appears in entangled-polymer start-up and step-strain transients, where the diffusive Rolie–Poly framework predicts early-time spatial instability of the viscoelastic stress or conformation field during strongly nonlinear transients, even when the steady constitutive curve is monotonic (Adams et al., 2010).

The same label has also been mapped onto several other polymer-specific instability classes. In bidisperse and polydisperse blends, PDI is associated with shear-induced demixing, where stress–diffusion coupling amplifies concentration fluctuations under shear (2002.04556). In drying polymer solutions, it denotes an evaporation-driven Rayleigh–Taylor-type instability caused by formation of a dense polymer-rich surface layer (Mossige et al., 2020). In directional freezing, the phrase does not appear in the source paper itself, but the observations have been summarized as a polymer-specific diffusion-mediated planar-interface instability with a global mode distinct from the local Mullins–Sekerka behavior of ionic solutions (Zhang et al., 2020). In elastic polymer films, the term maps onto the Asaro–Tiller–Grinfeld mechanism, in which residual in-plane stress drives surface undulations by surface diffusion (Closa et al., 2010).

This suggests that “PDI” is best understood as a family resemblance term rather than a single canonical instability. Its most technically developed and internally consistent meaning, however, is the diffusion-enabled instability of polymer conformation or stress in viscoelastic shear flows.

2. Transient PDI in entangled polymers

In the diffusive Rolie–Poly framework, PDI refers to the early-time, spatial instability of the viscoelastic stress or conformation field in an entangled polymer solution when driven far from equilibrium by a large step strain or by start-up to a fixed shear rate (Adams et al., 2010). The governing description couples creeping-flow force balance to a conformation-tensor evolution equation containing reptation, chain stretch, convected constraint release, and a small conformation diffusion term. The total stress is written as

$$\boldsymbol{\sigma} = -p\,\mathbf{I} + 2\eta_s\,\mathbf{D}^{s} + G\,\mathbf{W},$$

with polymer stress

$$\boldsymbol{\sigma}_p = G\,\mathbf{W} \equiv G\,(\mathbf{A}-\mathbf{I}),$$

and in simple shear

$\sigma_{xy} = G\,W_{xy} + \eta_s\,\dot{\gamma}.$

The diffusive Rolie–Poly evolution is

$$(\partial_t + \mathbf{v}\cdot\nabla)\,\mathbf{W} -(\nabla\mathbf{v})^{T}\cdot\mathbf{W} -\mathbf{W}\cdot(\nabla\mathbf{v}) +\frac{1}{\tau_d}\,\mathbf{W} = 2\,\mathbf{D}^{s} - \frac{2}{\tau_R}(1-a)\left[\mathbf{I}+\mathbf{W}+\beta\,a^{-2\delta}\,\mathbf{W}\right] + \mathcal{D}\,\nabla^{2}\mathbf{W},$$

with

$$a(\mathbf{W}) = \left(1+\frac{\mathrm{tr}\,\mathbf{W}}{3}\right)^{-1/2}.$$

The defining distinction from steady shear banding is that the steady constitutive curve remains monotonic, so the final steady state is homogeneous. Instability instead arises during the transient, when the instantaneous constitutive response $\sigma_{xy}(\dot{\gamma},t)$ can become non-monotonic shortly after the stress overshoot. The key criterion is the negative instantaneous slope,

$$\frac{\partial \sigma_{xy}}{\partial \dot{\gamma}}(\dot{\gamma}, t) < 0,$$

which correlates with linear instability of the homogeneous base state and with the appearance of transient shear bands (Adams et al., 2010).

A representative parameter set used for the monotonic steady curve is $(\beta,\epsilon,Z)=(0.71,10^{-5},265)$, with $\epsilon=\eta_s/(G\tau_d)$ and $Z=\tfrac13\,\tau_d/\tau_R$. In this regime the stress overshoot occurs at early times, the most unstable eigenvalue becomes positive near the overshoot, and transient bands appear and later decay over a few $$\boldsymbol{\sigma}_p = G\,\mathbf{W} \equiv G\,(\mathbf{A}-\mathbf{I}),$$0, returning the system to uniform shear.

3. Linear stability, diffusion regularization, and band selection

The stability analysis of transient PDI proceeds by perturbing a homogeneous base state $$\boldsymbol{\sigma}_p = G\,\mathbf{W} \equiv G\,(\mathbf{A}-\mathbf{I}),$$1 with Fourier modes in the gradient direction,

$$\boldsymbol{\sigma}_p = G\,\mathbf{W} \equiv G\,(\mathbf{A}-\mathbf{I}),$$2

The linearized dynamics satisfy

$$\boldsymbol{\sigma}_p = G\,\mathbf{W} \equiv G\,(\mathbf{A}-\mathbf{I}),$$3

where the time-dependent stability matrix contains stabilizing $$\boldsymbol{\sigma}_p = G\,\mathbf{W} \equiv G\,(\mathbf{A}-\mathbf{I}),$$4 terms from diffusion. A representative diagonal contribution is

$$\boldsymbol{\sigma}_p = G\,\mathbf{W} \equiv G\,(\mathbf{A}-\mathbf{I}),$$5

If $$\boldsymbol{\sigma}_p = G\,\mathbf{W} \equiv G\,(\mathbf{A}-\mathbf{I}),$$6 are eigenvalues of $$\boldsymbol{\sigma}_p = G\,\mathbf{W} \equiv G\,(\mathbf{A}-\mathbf{I}),$$7, instability occurs when

$$\boldsymbol{\sigma}_p = G\,\mathbf{W} \equiv G\,(\mathbf{A}-\mathbf{I}),$$8

Diffusion plays a dual role. It regularizes the problem by imposing a finite interfacial width and suppressing short wavelengths through the $$\boldsymbol{\sigma}_p = G\,\mathbf{W} \equiv G\,(\mathbf{A}-\mathbf{I}),$$9 shift, but it also permits spatial communication of fluctuations and hence transient amplification into banded states (Adams et al., 2010). In the cylindrical Couette simulations, a dimensionless diffusion coefficient $\sigma_{xy} = G\,W_{xy} + \eta_s\,\dot{\gamma}.$0 was used, with typical value $\sigma_{xy} = G\,W_{xy} + \eta_s\,\dot{\gamma}.$1, giving an effective diffusion length $\sigma_{xy} = G\,W_{xy} + \eta_s\,\dot{\gamma}.$2.

The temporal strength of the instability is quantified by the integrated growth weight

$\sigma_{xy} = G\,W_{xy} + \eta_s\,\dot{\gamma}.$3

When $\sigma_{xy} = G\,W_{xy} + \eta_s\,\dot{\gamma}.$4 is sufficiently large, initially small perturbations are amplified into observable transient bands before decaying as the long-time steady state reasserts stability (Adams et al., 2010). The strongest instability occurs near the steepest negative slope of the instantaneous constitutive curve and just after the stress overshoot.

4. Wall-bounded PDI in Oldroyd–B and FENE-P flows

A later line of work redefined PDI more narrowly as a linear instability of wall-bounded viscoelastic shear flows that requires finite polymer diffusion in the conformation tensor equation. In inertialess plane Couette flow, the laminar state is linearly stable without diffusion, but becomes unstable when a small but nonzero diffusivity is added to the polymeric degrees of freedom (Beneitez et al., 2022). In this setting the instability is a wall mode localized near the moving wall.

For the FENE-P formulation in channel or Taylor–Couette settings, the nondimensional equations are

$\sigma_{xy} = G\,W_{xy} + \eta_s\,\dot{\gamma}.$5

$\sigma_{xy} = G\,W_{xy} + \eta_s\,\dot{\gamma}.$6

$\sigma_{xy} = G\,W_{xy} + \eta_s\,\dot{\gamma}.$7

with

$\sigma_{xy} = G\,W_{xy} + \eta_s\,\dot{\gamma}.$8

For Oldroyd-B with diffusion, the dimensional conformation equation is

$\sigma_{xy} = G\,W_{xy} + \eta_s\,\dot{\gamma}.$9

The defining asymptotics are singular. In straight-channel flow, PDI is a wall mode with boundary-layer thickness $$(\partial_t + \mathbf{v}\cdot\nabla)\,\mathbf{W} -(\nabla\mathbf{v})^{T}\cdot\mathbf{W} -\mathbf{W}\cdot(\nabla\mathbf{v}) +\frac{1}{\tau_d}\,\mathbf{W} = 2\,\mathbf{D}^{s} - \frac{2}{\tau_R}(1-a)\left[\mathbf{I}+\mathbf{W}+\beta\,a^{-2\delta}\,\mathbf{W}\right] + \mathcal{D}\,\nabla^{2}\mathbf{W},$$0 and preferred streamwise wavenumber $$(\partial_t + \mathbf{v}\cdot\nabla)\,\mathbf{W} -(\nabla\mathbf{v})^{T}\cdot\mathbf{W} -\mathbf{W}\cdot(\nabla\mathbf{v}) +\frac{1}{\tau_d}\,\mathbf{W} = 2\,\mathbf{D}^{s} - \frac{2}{\tau_R}(1-a)\left[\mathbf{I}+\mathbf{W}+\beta\,a^{-2\delta}\,\mathbf{W}\right] + \mathcal{D}\,\nabla^{2}\mathbf{W},$$1, while its growth rate remains finite as $$(\partial_t + \mathbf{v}\cdot\nabla)\,\mathbf{W} -(\nabla\mathbf{v})^{T}\cdot\mathbf{W} -\mathbf{W}\cdot(\nabla\mathbf{v}) +\frac{1}{\tau_d}\,\mathbf{W} = 2\,\mathbf{D}^{s} - \frac{2}{\tau_R}(1-a)\left[\mathbf{I}+\mathbf{W}+\beta\,a^{-2\delta}\,\mathbf{W}\right] + \mathcal{D}\,\nabla^{2}\mathbf{W},$$2 (Beneitez et al., 2024). In planar Taylor–Couette flow restricted to the $$(\partial_t + \mathbf{v}\cdot\nabla)\,\mathbf{W} -(\nabla\mathbf{v})^{T}\cdot\mathbf{W} -\mathbf{W}\cdot(\nabla\mathbf{v}) +\frac{1}{\tau_d}\,\mathbf{W} = 2\,\mathbf{D}^{s} - \frac{2}{\tau_R}(1-a)\left[\mathbf{I}+\mathbf{W}+\beta\,a^{-2\delta}\,\mathbf{W}\right] + \mathcal{D}\,\nabla^{2}\mathbf{W},$$3–$$(\partial_t + \mathbf{v}\cdot\nabla)\,\mathbf{W} -(\nabla\mathbf{v})^{T}\cdot\mathbf{W} -\mathbf{W}\cdot(\nabla\mathbf{v}) +\frac{1}{\tau_d}\,\mathbf{W} = 2\,\mathbf{D}^{s} - \frac{2}{\tau_R}(1-a)\left[\mathbf{I}+\mathbf{W}+\beta\,a^{-2\delta}\,\mathbf{W}\right] + \mathcal{D}\,\nabla^{2}\mathbf{W},$$4 plane, no linear instability exists in the diffusionless equations used, but explicit polymer diffusion produces unstable eigenmodes and chaotic states resembling previously reported elastic-turbulence-like dynamics (Beneitez et al., 13 May 2025).

These studies also show that diffusion need not be explicit to be dynamically relevant. Finite-volume and finite-difference discretizations can introduce effective diffusive errors near walls, sufficient to trigger PDI in Oldroyd-B and FENE-P computations even when no polymer stress diffusion is intentionally prescribed (Beneitez et al., 13 May 2025). This has made PDI central to the interpretation of numerical transition scenarios in wall-bounded viscoelastic flow.

5. Inertia, concentration, and boundary-condition sensitivity

Finite inertia strengthens wall-bounded PDI. In plane Couette and plane Poiseuille flow, increasing $$(\partial_t + \mathbf{v}\cdot\nabla)\,\mathbf{W} -(\nabla\mathbf{v})^{T}\cdot\mathbf{W} -\mathbf{W}\cdot(\nabla\mathbf{v}) +\frac{1}{\tau_d}\,\mathbf{W} = 2\,\mathbf{D}^{s} - \frac{2}{\tau_R}(1-a)\left[\mathbf{I}+\mathbf{W}+\beta\,a^{-2\delta}\,\mathbf{W}\right] + \mathcal{D}\,\nabla^{2}\mathbf{W},$$5 enlarges the unstable region toward lower $$(\partial_t + \mathbf{v}\cdot\nabla)\,\mathbf{W} -(\nabla\mathbf{v})^{T}\cdot\mathbf{W} -\mathbf{W}\cdot(\nabla\mathbf{v}) +\frac{1}{\tau_d}\,\mathbf{W} = 2\,\mathbf{D}^{s} - \frac{2}{\tau_R}(1-a)\left[\mathbf{I}+\mathbf{W}+\beta\,a^{-2\delta}\,\mathbf{W}\right] + \mathcal{D}\,\nabla^{2}\mathbf{W},$$6 and lower $$(\partial_t + \mathbf{v}\cdot\nabla)\,\mathbf{W} -(\nabla\mathbf{v})^{T}\cdot\mathbf{W} -\mathbf{W}\cdot(\nabla\mathbf{v}) +\frac{1}{\tau_d}\,\mathbf{W} = 2\,\mathbf{D}^{s} - \frac{2}{\tau_R}(1-a)\left[\mathbf{I}+\mathbf{W}+\beta\,a^{-2\delta}\,\mathbf{W}\right] + \mathcal{D}\,\nabla^{2}\mathbf{W},$$7, and the associated growth rates increase linearly with $$(\partial_t + \mathbf{v}\cdot\nabla)\,\mathbf{W} -(\nabla\mathbf{v})^{T}\cdot\mathbf{W} -\mathbf{W}\cdot(\nabla\mathbf{v}) +\frac{1}{\tau_d}\,\mathbf{W} = 2\,\mathbf{D}^{s} - \frac{2}{\tau_R}(1-a)\left[\mathbf{I}+\mathbf{W}+\beta\,a^{-2\delta}\,\mathbf{W}\right] + \mathcal{D}\,\nabla^{2}\mathbf{W},$$8 when other parameters are fixed (Couchman et al., 2023). At finite $$(\partial_t + \mathbf{v}\cdot\nabla)\,\mathbf{W} -(\nabla\mathbf{v})^{T}\cdot\mathbf{W} -\mathbf{W}\cdot(\nabla\mathbf{v}) +\frac{1}{\tau_d}\,\mathbf{W} = 2\,\mathbf{D}^{s} - \frac{2}{\tau_R}(1-a)\left[\mathbf{I}+\mathbf{W}+\beta\,a^{-2\delta}\,\mathbf{W}\right] + \mathcal{D}\,\nabla^{2}\mathbf{W},$$9, the Schmidt number

$$a(\mathbf{W}) = \left(1+\frac{\mathrm{tr}\,\mathbf{W}}{3}\right)^{-1/2}.$$0

collapses neutral stability curves across different $$a(\mathbf{W}) = \left(1+\frac{\mathrm{tr}\,\mathbf{W}}{3}\right)^{-1/2}.$$1 and $$a(\mathbf{W}) = \left(1+\frac{\mathrm{tr}\,\mathbf{W}}{3}\right)^{-1/2}.$$2, making $$a(\mathbf{W}) = \left(1+\frac{\mathrm{tr}\,\mathbf{W}}{3}\right)^{-1/2}.$$3 the organizing parameter for inertially enhanced PDI. In highly concentrated polymeric fluids, corresponding to $$a(\mathbf{W}) = \left(1+\frac{\mathrm{tr}\,\mathbf{W}}{3}\right)^{-1/2}.$$4, PDI becomes a short-wavelength instability localized in thin boundary layers adjacent to surfaces, with inertia reducing the smallest Weissenberg number for instability from $$a(\mathbf{W}) = \left(1+\frac{\mathrm{tr}\,\mathbf{W}}{3}\right)^{-1/2}.$$5 in inertialess flows to $$a(\mathbf{W}) = \left(1+\frac{\mathrm{tr}\,\mathbf{W}}{3}\right)^{-1/2}.$$6 when inertia is significant (Lewy et al., 2023).

The instability is also highly sensitive to how diffusion is implemented at walls. A recent asymptotic analysis argued that the unstable PDI branch is primarily induced by the conformation boundary conditions introduced by artificial conformation diffusion. From a stable near-wall asymptotic solution, a new set of conformation boundary conditions was derived: Neumann for $$a(\mathbf{W}) = \left(1+\frac{\mathrm{tr}\,\mathbf{W}}{3}\right)^{-1/2}.$$7, Robin-type for $$a(\mathbf{W}) = \left(1+\frac{\mathrm{tr}\,\mathbf{W}}{3}\right)^{-1/2}.$$8 in DNS form, and Dirichlet for the remaining conformation components. In the reported tests these conditions removed the PDI branch in Oldroyd-B and FENE-P while leaving other physical instabilities intact (Dong et al., 13 Aug 2025).

This has sharpened an active controversy. One line of work treats polymer diffusion as a physically meaningful small regularization, potentially capable of initiating elastic or elasto-inertial turbulence through a genuine wall instability (Beneitez et al., 2024). Another interprets much observed PDI as a consequence of artificial conformation diffusion and its associated boundary conditions, implying that reliable high-$$a(\mathbf{W}) = \left(1+\frac{\mathrm{tr}\,\mathbf{W}}{3}\right)^{-1/2}.$$9 simulation requires suppressing that mode rather than exploiting it (Dong et al., 13 Aug 2025). The literature does not resolve this question definitively.

6. Relation to transition, banding, demixing, and other polymer instabilities

PDI provides a concrete route from laminar viscoelastic flow to more complex spatiotemporal states. In two-dimensional channel simulations of FENE-P fluids, a small but nonzero polymer stress diffusion coefficient destabilizes the laminar base flow through a short-streamwise-wavelength, wall-localized mode. That primary PDI saturates into a finite-amplitude travelling wave, which then becomes unstable to large-scale secondary modes and ultimately breaks down through a Ruelle–Takens route to either elastic turbulence at $\sigma_{xy}(\dot{\gamma},t)$0 or elasto-inertial turbulence at finite $\sigma_{xy}(\dot{\gamma},t)$1 (Beneitez et al., 2024). In this interpretation, PDI is the linchpin that connects smooth laminar profiles to chaotic states in parameter regimes where previously conjectured centre or wall modes of the laminar base flow were absent.

The relation to transient shear banding is complementary rather than identical. In the entangled-polymer Rolie–Poly setting, PDI is tied to the time-dependent constitutive curve and to the elastic window around the stress overshoot, producing transient bands even though the steady constitutive curve is monotonic (Adams et al., 2010). In wall-bounded Oldroyd-B and FENE-P flows, by contrast, PDI is a wall-layer eigenmode of the diffusive constitutive operator, not an instability of an instantaneous constitutive curve.

Shear-induced demixing in blends has been described as a polymer diffusive instability in a broader two-fluid or multi-fluid sense, because diffusive concentration modes become unstable when stress gradients drive relative motion between species under shear (2002.04556). This suggests a structural analogy: in each case, a diffusive degree of freedom that would ordinarily smooth heterogeneity instead participates in a feedback loop with viscoelastic stresses and flow, producing growth over a finite band of scales.

Other uses of the term are more distant but conceptually related. In drying polymer solutions, the instability arises when evaporation builds a dense surface layer that eventually undergoes a Rayleigh–Taylor-type overturning, with onset times governed by competition among buoyancy, molecular diffusion, effective interfacial tension, and viscosity (Mossige et al., 2020). In polymer films, stress-driven surface diffusion produces an Asaro–Tiller–Grinfeld instability of the free surface (Closa et al., 2010). In freezing polymer solutions, the summarized observations indicate a global, polymer-specific instability mode not captured by a diffusion-controlled Warren–Langer treatment that succeeds for ionic solutions (Zhang et al., 2020).

7. Conceptual synthesis and unresolved issues

Across the viscoelastic-flow literature, the most robust definition of PDI is a diffusion-regularized instability of polymer conformation or stress that is absent in the strictly diffusionless constitutive equations. In entangled polymers it appears transiently around the stress overshoot and produces temporary shear bands despite a monotonic steady constitutive curve (Adams et al., 2010). In wall-bounded Oldroyd-B and FENE-P flows it appears as a thin-layer wall mode with $\sigma_{xy}(\dot{\gamma},t)$2, $\sigma_{xy}(\dot{\gamma},t)$3, finite growth rate as $\sigma_{xy}(\dot{\gamma},t)$4, and substantial sensitivity to inertia, concentration, and boundary treatment (Couchman et al., 2023).

Three issues dominate current interpretation. The first is physicality: whether the relevant diffusion term represents an experimentally meaningful centre-of-mass or stress diffusion, or whether the instability is primarily an artifact of the constitutive closure at very small scales. The second is numerical realism: because finite-volume and finite-difference schemes can generate effective polymer diffusion near walls, PDI can contaminate simulations even when diffusion is not explicitly included (Beneitez et al., 13 May 2025). The third is boundary-condition dependence: if the unstable branch can be removed by asymptotically motivated conformation conditions without altering other instabilities, then part of what has been called PDI may reflect the mathematical completion of the diffusive model rather than an unavoidable physical mode (Dong et al., 13 Aug 2025).

A plausible implication is that PDI occupies an intermediate status between regularization and mechanism. Diffusion is introduced to regularize constitutive equations and to set finite interfacial or boundary-layer scales, yet once present it can qualitatively reorganize the linear spectrum and the transition pathway. For that reason, PDI has become important not only as a candidate physical instability, but also as a diagnostic concept for distinguishing genuine viscoelastic dynamics from diffusion-enabled or discretization-enabled modes in polymer-flow theory and computation.