Trace-Free Anisotropic Viscosity Model

• Trace-free anisotropic viscosity is defined by using the deviatoric part of the strain tensor to capture directional momentum transport in systems like magnetized plasmas and two-phase flows.
• It employs constitutive constructs—such as tensors of the form (b⊗b - I/3)—to precisely separate shear effects from isotropic compression, ensuring a traceless stress response.
• Numerical and analytical studies reveal that this formulation alters flow stability, heating rates, and the evolution of instabilities like Kelvin–Helmholtz modes compared to isotropic viscosity models.

A trace-free anisotropic viscosity model is a constitutive formulation in which viscous momentum transport is tied to a preferred direction and built from a deviatoric, hence traceless, part of the strain. In the literature, this structure appears most explicitly in Braginskii-type magnetized-plasma closures through tensors such as $\hat b\hat b-\frac13{\sf I}$, in filtered incompressible two-phase flow through interface-normal-dependent anisotropic stresses that remain traceless, and more indirectly in holographic anisotropic phases where the shear-viscosity tensor splits into inequivalent components under broken rotational symmetry (ZuHone et al., 2014, Magnaudet et al., 11 Apr 2025, Jain et al., 2015). The phrase therefore does not designate a single universal model; rather, it denotes a class of anisotropic closures in which shear is separated from isotropic compression and the viscous response becomes direction-dependent.

1. Constitutive meaning of “trace-free” in anisotropic viscosity

In the strict constitutive sense, “trace-free” means that the viscous stress is built from the shear, or deviatoric, part of the rate of strain rather than from an isotropic compressional term. This appears explicitly in the Braginskii form used for the intracluster medium, where the pressure anisotropy is written as

$$p_\perp-p_\parallel = 3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v},$$

and the viscous stress tensor is

$${\sf \Pi}_{\rm aniso} = -3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right)\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v}.$$

The operator $\hat{b}\hat{b}-\frac13{\sf I}$ is explicitly traceless, so the stress is constructed from the traceless part of the strain projected along the magnetic field direction (ZuHone et al., 2014).

An equivalent trace handling appears in the switching Braginskii-like solar-coronal model, which uses the rate-of-strain tensor

$$\mathbf{W} = \nabla\mathbf{u} + (\nabla\mathbf{u})^T - \frac{2}{3}(\nabla\cdot\mathbf{u})\mathbf{I}.$$

Its anisotropic contribution is proportional to

$$\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right),$$

whose trace vanishes because

$$\mathrm{tr}\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right) = \mathbf{b}\cdot\mathbf{b} - 1 = 0.$$

The model is therefore explicitly deviatoric in its field-aligned part (Quinn et al., 2021).

The same distinction between shear and isotropic compression is standard in hydrodynamic tensor language. In the anisotropic unitary Fermi gas, the dissipative shear sector is written with the trace-free condition

$\eta^{ijkl}\delta_{kl}=0,$

and the isotropic shear tensor takes the projector form

$$\eta_{ijkl}^{\rm iso} = \eta\,\frac12\left( \delta_{ik}\delta_{jl}+\delta_{il}\delta_{jk} -\frac23\delta_{ij}\delta_{kl} \right).$$

The anisotropic corrections are then added through extra traceless tensor structures built from the preferred direction $\hat E=\nabla\phi/|\nabla\phi|$ (Samanta et al., 2016).

Not every anisotropic viscosity model is trace-free in this constitutive sense. The stochastic and deterministic Navier–Stokes models with one-directional or horizontal-only dissipation replace the full Laplacian by $$p_\perp-p_\parallel = 3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v},$$0 or $$p_\perp-p_\parallel = 3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v},$$1; they are anisotropic and degenerate, but the papers do not formulate them as trace-free viscosity tensors (Liang et al., 2018, Bessaih et al., 2017). This distinction is central: anisotropy alone does not imply a trace-free constitutive law.

2. Field-aligned trace-free closures in magnetized plasma

The best-known trace-free anisotropic viscosity model in plasma physics is the Braginskii closure for weakly collisional, magnetized media. In the cold-front simulations of the intracluster medium, momentum transport is strongly directional because the ion mean free path is vastly larger than the ion gyroradius. Viscous dissipation therefore acts only on velocity gradients parallel to the magnetic field, while gradients perpendicular to the field are essentially not damped (ZuHone et al., 2014).

The stress-tensor interpretation is explicit. The paper writes the anisotropic pressure tensor as

$$p_\perp-p_\parallel = 3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v},$$2

with

$$p_\perp-p_\parallel = 3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v},$$3

This is not an isotropic Laplacian-like damping law. It is a field-aligned closure in which the magnetic geometry determines which parts of the velocity gradient are dissipated (ZuHone et al., 2014).

In sloshing cold fronts, this geometric dependence is decisive. The magnetic field is often stretched or draped so that it lies approximately parallel to the front surface. The dominant shear flow across the interface then has velocity gradients mostly perpendicular to the local $$p_\perp-p_\parallel = 3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v},$$4, which makes Braginskii viscosity much less effective than an isotropic viscosity in suppressing Kelvin–Helmholtz growth. The simulations show that Braginskii viscosity only partially suppresses Kelvin–Helmholtz modes, whereas full isotropic Spitzer viscosity essentially eliminates them; a reduced isotropic Spitzer viscosity with $$p_\perp-p_\parallel = 3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v},$$5 produces qualitatively similar smoothing to the Braginskii case (ZuHone et al., 2014).

The same paper emphasizes that the observational interpretation is not straightforward. Even weak magnetic fields can significantly alter cold-front morphology, Braginskii viscosity and reduced isotropic viscosity can produce similar observational signatures, and the Braginskii-MHD equations do not include finite-Larmor-radius physics. When

$$p_\perp-p_\parallel = 3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v},$$6

firehose and mirror instabilities should arise and regulate the anisotropy, but these kinetic effects are not resolved (ZuHone et al., 2014).

3. Regularized and numerical formulations

A direct Braginskii tensor becomes problematic near magnetic nulls because the preferred direction $$p_\perp-p_\parallel = 3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v},$$7 is singular as $$p_\perp-p_\parallel = 3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v},$$8. The switching model used for a dynamically twisted magnetic null point resolves this by interpolating between an isotropic Newtonian stress near nulls and a trace-free field-aligned stress in strong-field regions: $$p_\perp-p_\parallel = 3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v},$$9 The switching function is

$${\sf \Pi}_{\rm aniso} = -3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right)\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v}.$$0

This construction preserves a trace-free anisotropic part while avoiding singular behavior in weak-field regions (Quinn et al., 2021).

Its dynamical consequences differ substantially from isotropic viscosity. In the null-point problem, anisotropic viscosity allows faster Kelvin–Helmholtz growth, permits the collapse of the null point significantly sooner, and yields generally smaller viscous heating. At the same time, the stronger instability enhances current-sheet formation and therefore Ohmic heating, so the total heating rate can be greater when anisotropic viscosity is employed (Quinn et al., 2021).

High-order numerical realization of trace-free anisotropic viscosity has also been developed in mesh-free MHD. In the GIZMO MFM/MFV formulation, the diffusion equation is written as

$${\sf \Pi}_{\rm aniso} = -3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right)\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v}.$$1

with the mesh-free update

$${\sf \Pi}_{\rm aniso} = -3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right)\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v}.$$2

For Braginskii viscosity, the trace-free tensor is

$${\sf \Pi}_{\rm aniso} = -3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right)\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v}.$$3

so

$${\sf \Pi}_{\rm aniso} = -3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right)\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v}.$$4

The method combines a second-order accurate least-squares gradient estimator, face-based conservative fluxes, and a nonlinear flux limiter with a direct-flux sign check. The Braginskii viscosity test confirms the correct field-aligned, trace-free viscosity behavior and agreement with high-resolution ATHENA results (Hopkins, 2016).

4. Filtered interfacial anisotropy in incompressible two-phase flow

A different trace-free anisotropic viscosity model arises in fixed-mesh one-fluid formulations of incompressible two-phase flow. The central claim is that the equations actually solved by VOF, Level Set, Front Tracking, and IBM methods are spatially filtered near the interface, so the effective viscous law in mixed cells is not isotropic. Normal and shear stresses obey different averaging rules and require two distinct viscosity coefficients (Magnaudet et al., 11 Apr 2025).

In the reference two-dimensional configuration, the filtered normal or diagonal viscous stress is governed by the arithmetic mean

$${\sf \Pi}_{\rm aniso} = -3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right)\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v}.$$5

whereas the filtered shear stress is governed by the harmonic-type mean

$${\sf \Pi}_{\rm aniso} = -3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right)\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v}.$$6

This difference follows from continuity of tangential traction across the interface for shear stresses and from the distinct behavior of normal viscous stresses when $${\sf \Pi}_{\rm aniso} = -3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right)\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v}.$$7 (Magnaudet et al., 11 Apr 2025).

The general constitutive law is then written in intrinsic tensor form using the filtered interface normal $${\sf \Pi}_{\rm aniso} = -3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right)\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v}.$$8: $${\sf \Pi}_{\rm aniso} = -3\mu\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right)\left(\hat{b}\hat{b}-\frac{1}{3}{\sf I}\right):\nabla{\bf v}.$$9 Because the flow is incompressible, the viscous stresses of the pure fluids are traceless, and the filtered viscous stress is therefore also traceless. The paper identifies this with the tensor $\hat{b}\hat{b}-\frac13{\sf I}$0 in the Ericksen-type representation and fixes the coefficients by matching to the two-dimensional derivation (Magnaudet et al., 11 Apr 2025).

This model is validated against a 3D viscous buoyancy-driven exchange flow in a closed vertical pipe. The benchmark uses a $\hat{b}\hat{b}-\frac13{\sf I}$1 glass pipe of $\hat{b}\hat{b}-\frac13{\sf I}$2 inner diameter, fluids with $\hat{b}\hat{b}-\frac13{\sf I}$3, $\hat{b}\hat{b}-\frac13{\sf I}$4, and $\hat{b}\hat{b}-\frac13{\sf I}$5, and negligible interfacial tension and molecular diffusivity. Using different levels of grid refinement, the paper reports that the anisotropic model is the only one capable of predicting correctly the evolution of the front of the ascending and descending fingers at a reasonable computational cost (Magnaudet et al., 11 Apr 2025). The harmonic model is too fast for the ascending front and poor in early descending dynamics, whereas the arithmetic model over-dissipates shear and is too slow for the descending front (Magnaudet et al., 11 Apr 2025).

5. Holographic anisotropic shear and the horizon formula

In holographic anisotropic phases, the relevant object is not usually a single trace-free constitutive tensor in the plasma-physics sense. Instead, rotational symmetry is broken while translation invariance is preserved, so the shear viscosity becomes a tensor $\hat{b}\hat{b}-\frac13{\sf I}$6 with distinct components. The canonical result is that shear polarized partly along the anisotropic direction behaves differently from shear entirely within the residual isotropic plane (Jain et al., 2015).

For anisotropic black branes generated by translationally invariant forcing, the key formula is

$\hat{b}\hat{b}-\frac13{\sf I}$7

or, equivalently in the notation of the same paper,

$\hat{b}\hat{b}-\frac13{\sf I}$8

This relation is derived from the spin-1 metric perturbation $\hat{b}\hat{b}-\frac13{\sf I}$9, the Kubo formula for the retarded stress-tensor correlator, and radial conservation of the conjugate momentum in the $$\mathbf{W} = \nabla\mathbf{u} + (\nabla\mathbf{u})^T - \frac{2}{3}(\nabla\cdot\mathbf{u})\mathbf{I}.$$0 limit. In the isotropic case $$\mathbf{W} = \nabla\mathbf{u} + (\nabla\mathbf{u})^T - \frac{2}{3}(\nabla\cdot\mathbf{u})\mathbf{I}.$$1, the KSS value $$\mathbf{W} = \nabla\mathbf{u} + (\nabla\mathbf{u})^T - \frac{2}{3}(\nabla\cdot\mathbf{u})\mathbf{I}.$$2 is recovered; in anisotropic geometries, the ratio can be much smaller than one (Jain et al., 2015).

The paper verifies this formula in several examples. In the one-dilaton case at low temperature,

$$\mathbf{W} = \nabla\mathbf{u} + (\nabla\mathbf{u})^T - \frac{2}{3}(\nabla\cdot\mathbf{u})\mathbf{I}.$$3

so the viscosity is parametrically suppressed when $$\mathbf{W} = \nabla\mathbf{u} + (\nabla\mathbf{u})^T - \frac{2}{3}(\nabla\cdot\mathbf{u})\mathbf{I}.$$4. In the dilaton–axion system,

$$\mathbf{W} = \nabla\mathbf{u} + (\nabla\mathbf{u})^T - \frac{2}{3}(\nabla\cdot\mathbf{u})\mathbf{I}.$$5

and for $$\mathbf{W} = \nabla\mathbf{u} + (\nabla\mathbf{u})^T - \frac{2}{3}(\nabla\cdot\mathbf{u})\mathbf{I}.$$6,

$$\mathbf{W} = \nabla\mathbf{u} + (\nabla\mathbf{u})^T - \frac{2}{3}(\nabla\cdot\mathbf{u})\mathbf{I}.$$7

The magnetic-field example yields

$$\mathbf{W} = \nabla\mathbf{u} + (\nabla\mathbf{u})^T - \frac{2}{3}(\nabla\cdot\mathbf{u})\mathbf{I}.$$8

The paper’s summary is that spin-1 components can become parametrically small compared with the entropy density, while components that remain effectively isotropic recover the usual KSS value (Jain et al., 2015).

This component-wise structure is reinforced by other holographic studies. In anisotropic $$\mathbf{W} = \nabla\mathbf{u} + (\nabla\mathbf{u})^T - \frac{2}{3}(\nabla\cdot\mathbf{u})\mathbf{I}.$$9 SYM with $$\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right),$$0, the longitudinal shear viscosity is

$$\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right),$$1

whereas the purely transverse viscosity remains

$$\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right),$$2

Since $$\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right),$$3 for $$\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right),$$4, the longitudinal component satisfies $$\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right),$$5 (Rebhan et al., 2011). In the anisotropic Reissner–Nordström black brane with finite chemical potential, the quadratic correction to the spin-one viscosity is negative, so $$\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right),$$6 for nonzero anisotropy, especially at larger $$\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right),$$7 and smaller $$\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right),$$8 (Chakraborty et al., 2017). By contrast, in Horava–Lifshitz gravity the transverse $$\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right),$$9 channel still yields

$$\mathrm{tr}\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right) = \mathbf{b}\cdot\mathbf{b} - 1 = 0.$$0

so the KSS value is preserved in that sector (Sadeghi, 2019).

A common misconception is that anisotropy suppresses all shear viscosities uniformly. The holographic results show the opposite: the suppression is tied specifically to the spin-1 sector, while the spin-2 or purely transverse channel often remains universal (Rebhan et al., 2011, Ge, 2015).

6. Constraints, applications, and conceptual boundaries

The main constraint analysis in anisotropic Einstein gravity finds no stringent universal bound that restores isotropic behavior. The relevant criterion is whether the fluctuation equation can be written exactly as that of a minimally coupled massless scalar field. If it can, then $$\mathrm{tr}\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right) = \mathbf{b}\cdot\mathbf{b} - 1 = 0.$$1 survives; if it cannot, then the shear-viscosity bound can be violated. The spin-2 perturbation $$\mathrm{tr}\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right) = \mathbf{b}\cdot\mathbf{b} - 1 = 0.$$2 obeys the minimally coupled scalar equation, whereas the spin-1 perturbation $$\mathrm{tr}\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right) = \mathbf{b}\cdot\mathbf{b} - 1 = 0.$$3 does not. The same study concludes that causality and thermodynamic stability do not provide a sharp upper bound on the anisotropy parameter strong enough to prevent viscosity violation, and that the diffusion bound is also not saturated in the anisotropic regime (Ge, 2015).

The broader physical relevance of trace-free anisotropic viscosity depends strongly on context. In the intracluster medium, Braginskii viscosity smooths cold fronts but is far less effective than full isotropic Spitzer viscosity because field lines are oriented mostly along the front surface and therefore perpendicular to the dominant shear-gradient direction (ZuHone et al., 2014). In the solar corona, a switching Braginskii-like closure captures parallel viscosity in strong-field regions and isotropic behavior near magnetic nulls, but neglects perpendicular and drift terms from the full Braginskii tensor; the authors state that these neglected components could matter in strong velocity shears, although they are expected to affect results quantitatively rather than qualitatively (Quinn et al., 2021).

Cold-atom theory provides an intermediate perspective between constitutive tensor models and holographic transport splitting. In the anisotropic unitary Fermi gas, the shear-viscosity tensor acquires extra anisotropic pieces organized into five irreducible projectors $$\mathrm{tr}\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right) = \mathbf{b}\cdot\mathbf{b} - 1 = 0.$$4 built from $$\mathrm{tr}\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right) = \mathbf{b}\cdot\mathbf{b} - 1 = 0.$$5. The perturbative Boltzmann calculation gives

$$\mathrm{tr}\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right) = \mathbf{b}\cdot\mathbf{b} - 1 = 0.$$6

with $$\mathrm{tr}\left(\mathbf{b}\otimes\mathbf{b} - \frac{1}{3}\mathbf{I}\right) = \mathbf{b}\cdot\mathbf{b} - 1 = 0.$$7. The anisotropic corrections are therefore negative, and the paper interprets the model as a trace-free anisotropic shear viscosity rather than a bulk-viscosity-like modification (Samanta et al., 2016).

The term should therefore be used with care. In plasma closures and filtered incompressible two-phase flow, it refers directly to a constitutive law built from traceless tensor structures. In holographic transport theory, it more often refers to a component-wise anisotropic viscosity tensor whose distinguished channels may be much smaller than the isotropic KSS value. A plausible implication is that the unifying principle is not a single formula but a shared structural feature: viscous response is separated from isotropic compression and projected onto preferred directions set by magnetic fields, interfaces, or anisotropic forcing.