r-Modes, Viscosity, and Saturation

An r-mode is an axial-led oscillation whose restoring force is primarily the Coriolis force. Its current multipole can be driven by gravitational radiation, whereas viscosity, superfluid mutual friction, a crustal boundary layer, and nonlinear couplings oppose or redirect that growth. The linear sign is robust; the temperature–spin instability window and the amplitude at which growth stops are model-dependent frontier questions. Evidence and interpretations on this page are current through 2026-08-18.

Helpful background. Stress–Energy and Conservation supplies conserved currents; the Einstein equation supplies the rotating background; Relativistic Matter and Compact Objects orients stellar equilibria; Black-Hole Perturbation Theory supplies mode conventions; Gravitational Radiation and Waveforms supplies multipolar fluxes; and Computational Relativity supplies convergence tests for mode calculations.

The axial r-mode pattern

Adopt perturbations proportional to $e^{-i\omega t+im\phi}$ in a uniformly rotating star of angular velocity $\Omega$. In a Newtonian, slowly rotating perfect-fluid star, a purely axial leading-order r-mode has

$$\omega_{\rm rot}=-\frac{2m\Omega}{\ell(\ell+1)} +O\!\left(\frac{\Omega^3}{\Omega_{\rm dyn}^2}\right), \qquad \omega=\omega_{\rm rot}+m\Omega.$$

Here $\Omega_{\rm dyn}=\sqrt{GM/R^3}$ and $\Omega/\Omega_{\rm dyn}\ll1$. The familiar $\ell=m\ge2$ family is present in a barotropic Newtonian star; the broader pure-axial $\ell\ge m$ classification requires stratification and must not be assumed for every barotrope. General-relativistic barotropic inertial modes are generically axial–polar hybrids. Compactness corrections can change the coefficient of $\Omega$ itself, so slow rotation alone does not make the Newtonian value exact. The Newtonian frequency and frame translation are given in Haskell 2015, §2, Eqs. (1)–(4); the relativistic hybrid distinction is established in Lockitch, Andersson, and Friedman 2000, §§I and IV.

Thus the $\ell=m=2$ mode is retrograde in the fluid frame, $\omega_{\rm rot}=-2\Omega/3$, but prograde at infinity, $\omega=4\Omega/3$. It satisfies $\omega(\omega-m\Omega)<0$ and therefore has the Chandrasekhar–Friedman–Schutz sign needed for gravitational radiation to drive it. Andersson established this relativistic instability for rotating stars in Andersson 1998, §4, Eq. (15).

These frequencies are leading-order Newtonian statements, not precision templates. Rapid rotation mixes axial and polar harmonics, relativistic frame dragging shifts the spectrum, and stratification, elasticity, superfluid degrees of freedom, or a magnetic field can introduce avoided crossings. The sign check survives small continuous corrections, but an inferred frequency must use a specified equilibrium and frame convention. With $c=1$, $\Omega$ and $\omega$ have dimension $L^{-1}$; a dimensionless frequency is $GM\omega$, not $M\omega$ if $M$ is quoted as a physical mass.

A signed timescale balance

For a linear mode amplitude $\alpha$, collect independent weak driving and damping channels into

$$\frac{d\ln\alpha}{dt}=-\frac{1}{\tau}, \qquad \frac{1}{\tau}= \frac{1}{\tau_{\rm GR}}+ \frac{1}{\tau_{\rm shear}}+ \frac{1}{\tau_{\rm bulk}}+ \frac{1}{\tau_{\rm MF}}+ \frac{1}{\tau_{\rm Ek}}+\cdots .$$

The convention is $\tau_{\mathrm{GR}}<0$ for radiation driving and positive timescales for damping. Linear growth occurs when $1/\tau<0$; the neutral curve is $1/\tau=0$. The gravitational term rises steeply with spin because the leading radiation is a current multipole. Shear viscosity is usually more important on the cold side of a minimal-fluid window, while bulk viscosity becomes stronger on the hot side. Lindblom, Owen, and Morsink give the standard leading-order scalings and a concrete stellar model in Lindblom, Owen, and Morsink 1998, Eqs. (16)–(22), pp. 4844–4845.

For example, their minimal normal-fluid $\ell=m=2$ model has the scalings

$$\frac1{\tau_{\rm GR}}=-A\Omega^6,\qquad \frac1{\tau_{\rm shear}}=B T^{-2},\qquad \frac1{\tau_{\rm bulk}}=D\Omega^2T^6,$$

with positive, dimensionful constants $A,B,D$ fixed by the stellar model and viscosity prescription. The last scaling assumes modified-Urca bulk viscosity in the regime used in that calculation. Balancing gravitational driving with shear damping gives $\Omega_{\rm crit}\propto T^{-1/3}$; balancing it with bulk damping gives $\Omega_{\rm crit}\propto T^{3/2}$. These opposite slopes explain the familiar window between a cold and a hot stable region. They are not universal laws through superfluid transitions or into a different weak-reaction regime.

The sum is useful only when each term refers to the same mode normalization and background. A damping time computed for a barotropic fluid cannot simply be combined with a superfluid eigenfunction whose counter-moving component has a different norm. Likewise, a rigid-crust Ekman estimate and an elastic-crust calculation answer different boundary-condition problems. An independent check is to set every dissipative coefficient to zero: the formula must recover pure radiation growth. Setting the radiation flux to zero must instead leave monotonic damping.

Why instability windows differ

The familiar spin–temperature curve is not a universal exclusion boundary. It is the zero set of a selected microphysical model. A minimal npe fluid gives one position; nucleon superfluidity adds entrainment and mutual friction; hyperons or deconfined quarks change reaction rates and bulk viscosity; a solid crust introduces slippage and a viscous boundary layer. Pairing gaps, composition gradients, magnetic geometry, and differential rotation enlarge the parameter regime further. Haskell reviews how these formulation axes alter both theoretical windows and comparisons with measured spins and temperatures in Haskell 2015, §§4–5.

There is credible evidence for strong damping in some model classes, including boundary-layer and mutual-friction scenarios, but contrary conclusions follow when crustal slippage, pairing gaps, or interaction strengths are changed. Observed stars plotted outside a minimal-fluid window therefore challenge that model, not the CFS sign theorem. Temperature estimates can also be envelope- and distance-dependent, and an accreting star need not sit in thermal or torque equilibrium. The unresolved question is which matter model, boundary condition, and magnetic parameter regime describes a given source well enough to turn the curve into a quantitative constraint.

Saturation is a nonlinear energy budget

Linear theory predicts neither an infinite amplitude nor a unique saturation value $\alpha_{\rm sat}$. At finite amplitude, three-mode couplings can transfer energy from the unstable parent into damped daughters; differential rotation and magnetic winding can reshape the background; shocks or wave breaking may appear in suitable regimes; and viscous heating feeds back on temperature. A schematic amplitude system is

$$\dot\alpha=-\frac{\alpha}{\tau}+\mathcal N_\alpha, \qquad C_V\dot T=H_{\rm visc}(\alpha,T)-L_\nu(T)-L_\gamma(T),$$

where $\mathcal N_\alpha$ represents nonlinear transfer and the thermal equation shows why a fixed-temperature saturation calculation need not describe a long-lived star. A reported amplitude is interpretable only with its mode normalization, daughter network or evolution equations, resolution and convergence criteria, and duration relative to the secular timescales.

Competing saturation estimates are not necessarily inconsistent: a three-mode truncation, a hydrodynamic evolution, and a magnetized superfluid calculation have different initial data and extension classes. What remains not settled is whether any one mechanism controls realistic old accretors, young remnants, or both. A result that could change the interpretation is a converged general-relativistic multifluid evolution that reports its microphysical rates and demonstrates the same saturation amplitude under resolution, domain, and initial-perturbation changes; a revised reaction-rate or pairing-gap calibration would also require revisiting the window.

Check a toy instability window

At one point in a stellar model, take $\tau_{\rm GR}=-40\,\mathrm{s}$, $\tau_{\rm shear}=200\,\mathrm{s}$, $\tau_{\rm bulk}=100\,\mathrm{s}$, and neglect other channels. Does the mode grow, and what is its e-folding time?

Solution

The signed inverse time is

$$\frac{1}{\tau}=-\frac{1}{40}+\frac{1}{200}+\frac{1}{100} =-0.010\,\mathrm{s}^{-1}.$$

Because $d\ln\alpha/dt=-1/\tau$, the logarithmic growth rate is $0.010\,\mathrm{s}^{-1}$ and the amplitude e-folding time is $100\,\mathrm{s}$. Adding a damping channel faster than $100\,\mathrm{s}$ would move this toy point to stability. The calculation diagnoses the linear sign only; it says nothing about nonlinear saturation.

References

• Andersson, Nils. “A New Class of Unstable Modes of Rotating Relativistic Stars.” Astrophysical Journal 502 (1998): 708–713. DOI.
• Haskell, Brynmor. “R-Modes in Neutron Stars: Theory and Observations.” International Journal of Modern Physics E 24 (2015): 1541007. arXiv.
• Lindblom, Lee, Benjamin J. Owen, and Sharon M. Morsink. “Gravitational Radiation Instability in Hot Young Neutron Stars.” Physical Review Letters 80 (1998): 4843–4846. DOI.
• Lockitch, Keith H., Nils Andersson, and John L. Friedman. “The Rotational Modes of Relativistic Stars: Analytic Results.” Physical Review D 63 (2000): 024019. DOI; Open PDF.