Eccentric Accretion Disks

• Eccentric accretion disks are non-circular gas flows with elliptical orbits that support slow m=1 precessing modes, shaping global disk dynamics.
• They modify angular-momentum transport and shock formation via pressure, self-gravity, and magnetized turbulence across a range of astrophysical environments.
• Excitation mechanisms such as tidal resonances, stream-fed infall, and anisotropic mass-loading drive their behavior in binaries, tidal disruption events, and AGNs.

Eccentric accretion disks are accretion flows in which the gas follows elliptical streamlines, or equivalently disks that support global, slowly precessing $m=1$ disturbances described by a complex eccentricity $E(r,t)=e(r,t)e^{i\varpi(r,t)}$. They occur in compact binaries, circumbinary disks, tidal disruption events, active galactic nuclei, and protostellar disks, where eccentricity may be free, forced, resonantly driven, or maintained by anisotropic mass loading. Across these settings, eccentricity changes angular-momentum transport, apsidal precession, vertical structure, shock formation, and radiative output, while the survival of the eccentric state depends sensitively on outer boundary conditions, thermodynamics, and the competition between excitation and damping (Lee et al., 2019, Lee et al., 2018, Ohana et al., 2024, Lynch et al., 2020).

1. Dynamical definition and modal structure

A standard diagnostic is the eccentricity vector, written in compact-binary simulations as

$$\boldsymbol{e}=\frac{1}{G M_1}\,\boldsymbol{v}\times(\boldsymbol{r}\times\boldsymbol{v})-\hat{\boldsymbol r}, \qquad \langle \boldsymbol e\rangle=\frac{\int \rho\,\boldsymbol e\,dV}{\int \rho\,dV},$$

with analogous cell-wise constructions used in circumbinary and protostellar calculations. In thin, nearly Keplerian disks, the corresponding $m=1$ eccentric modes are slow: their pattern frequency $\omega$ is much smaller than the orbital frequency $\Omega$, so the apsidal line precesses on secular rather than dynamical timescales (Ohana et al., 2024, Siwek et al., 2022, Ahmad et al., 13 Jan 2026).

For pressure-only disks, the linear eccentricity equation can be written as

$$\omega E=\frac{1}{2\Omega r^3\Sigma}\left\{\frac{d}{dr}\left(\gamma r^3P\frac{dE}{dr}\right)+r^2\frac{dP}{dr}E\right\},$$

and, after the transformation $y=(r^3P)^{1/2}E$, as a Schrödinger-like wave equation with an effective potential $\omega_p(r)$. In disks with realistic outer density tapers, $\omega_p(r)$ develops an inverted-U shape, the eccentric mode is trapped between turning points, and the longest-lived zero-node mode is an aligned elliptical pattern that decays coherently on the viscous timescale (Lee et al., 2019).

When self-gravity is included, the second-order WKB dispersion relation contains both pressure and self-gravity terms. The resulting mode taxonomy separates pressure-dominated and gravity-dominated regimes: when self-gravity is weak, $E(r,t)=e(r,t)e^{i\varpi(r,t)}$0, the modes are pressure dominated, whereas for $E(r,t)=e(r,t)e^{i\varpi(r,t)}$1 two gravity-dominated families appear, namely aligned elliptical patterns and one-armed spirals (Lee et al., 2018). This establishes that eccentric accretion disks are not a single phenomenon but a family of slow global states whose morphology depends on the relative importance of pressure, self-gravity, and boundary trapping.

2. Excitation mechanisms across astrophysical settings

In compact binaries, the canonical driver is the 3:1 mean-motion resonance. For a Keplerian disk,

$E(r,t)=e(r,t)e^{i\varpi(r,t)}$2

and eccentricity growth requires the disk to extend to $E(r,t)=e(r,t)e^{i\varpi(r,t)}$3 while remaining inside the tidal truncation radius $E(r,t)=e(r,t)e^{i\varpi(r,t)}$4. For $E(r,t)=e(r,t)e^{i\varpi(r,t)}$5, this condition is typically satisfied; in the reported simulations $E(r,t)=e(r,t)e^{i\varpi(r,t)}$6 comfortably meets it, and the companion’s $E(r,t)=e(r,t)e^{i\varpi(r,t)}$7 tidal forcing nonlinearly couples to an eccentric mode through the Lubow mechanism, driving global eccentricity and prograde apsidal precession associated with positive superhumps (Ohana et al., 2024).

In circumbinary disks, eccentricity can instead be injected at the cavity wall. Finite circumbinary rings that viscously spread into disks develop an eccentric cavity instability excited by repeated impacts of rejected accretion streams on the cavity wall. Compact, cold rings lack the large outer angular-momentum reservoir available to extended disks, so eccentricity can grow to $E(r,t)=e(r,t)e^{i\varpi(r,t)}$8–$E(r,t)=e(r,t)e^{i\varpi(r,t)}$9, whereas infinite disks saturate at $$\boldsymbol{e}=\frac{1}{G M_1}\,\boldsymbol{v}\times(\boldsymbol{r}\times\boldsymbol{v})-\hat{\boldsymbol r}, \qquad \langle \boldsymbol e\rangle=\frac{\int \rho\,\boldsymbol e\,dV}{\int \rho\,dV},$$0 (Betancourt et al., 2 Jan 2026). In binaries with nonzero orbital eccentricity, the circumbinary response divides into three regimes—free precession, forced precession, and apsidal locking—set by the pair $$\boldsymbol{e}=\frac{1}{G M_1}\,\boldsymbol{v}\times(\boldsymbol{r}\times\boldsymbol{v})-\hat{\boldsymbol r}, \qquad \langle \boldsymbol e\rangle=\frac{\int \rho\,\boldsymbol e\,dV}{\int \rho\,dV},$$1, so the disk eccentricity may be either a freely precessing mode or a forced response to the binary quadrupole (Siwek et al., 2022).

A distinct mechanism operates in equal-mass binary minidisks. There the instability is not a tidal 3:1 resonance but a phase-locked sequence of regular impacts between anti-aligned eccentric minidisks. The characteristic cadence is the beat frequency

$$\boldsymbol{e}=\frac{1}{G M_1}\,\boldsymbol{v}\times(\boldsymbol{r}\times\boldsymbol{v})-\hat{\boldsymbol r}, \qquad \langle \boldsymbol e\rangle=\frac{\int \rho\,\boldsymbol e\,dV}{\int \rho\,dV},$$2

and the measured growth rate is $$\boldsymbol{e}=\frac{1}{G M_1}\,\boldsymbol{v}\times(\boldsymbol{r}\times\boldsymbol{v})-\hat{\boldsymbol r}, \qquad \langle \boldsymbol e\rangle=\frac{\int \rho\,\boldsymbol e\,dV}{\int \rho\,dV},$$3 (Westernacher-Schneider et al., 2023).

Eccentric disks can also be born eccentric. Tidal disruption events inject matter on highly eccentric orbits, while Class 0 protostellar disks receive anisotropic streamer-fed accretion with a systematic angular-momentum deficit relative to local circular motion. In the latter case, the inflow continuously generates and sustains significant disk eccentricity, typically $$\boldsymbol{e}=\frac{1}{G M_1}\,\boldsymbol{v}\times(\boldsymbol{r}\times\boldsymbol{v})-\hat{\boldsymbol r}, \qquad \langle \boldsymbol e\rangle=\frac{\int \rho\,\boldsymbol e\,dV}{\int \rho\,dV},$$4, without requiring planets, binaries, or strong self-gravity (Chan et al., 2022, Ahmad et al., 13 Jan 2026).

3. Angular-momentum transport, MRI turbulence, and nonlinear structure

Magnetohydrodynamic transport in eccentric disks is not a small perturbation of circular-disk behavior. Global ideal-MHD simulations of moderately eccentric disks show that MRI saturation levels remain comparable to those of circular disks when magnetic topology is held fixed, but the Maxwell stress can be negative in some disk sectors even though the orbit-integrated stress remains positive. The net effect is outward angular-momentum transport, accompanied by an increase of eccentricity in the inner disk and a decrease in the outer disk (Chan et al., 2022).

Compact-binary calculations make the same point in a different setting. When a magnetized disk actually reaches the 3:1 resonance, MRI turbulence does not impede tidally driven eccentricity growth in any meaningful way; the Lubow instability operates in full MHD much as it does in viscous hydrodynamics. The major differences appear in how the disk reaches the resonance and in how eccentricity is redistributed afterward (Ohana et al., 2024).

Run	Transport state	max $$\boldsymbol{e}=\frac{1}{G M_1}\,\boldsymbol{v}\times(\boldsymbol{r}\times\boldsymbol{v})-\hat{\boldsymbol r}, \qquad \langle \boldsymbol e\rangle=\frac{\int \rho\,\boldsymbol e\,dV}{\int \rho\,dV},$$5
Hydro-$$\boldsymbol{e}=\frac{1}{G M_1}\,\boldsymbol{v}\times(\boldsymbol{r}\times\boldsymbol{v})-\hat{\boldsymbol r}, \qquad \langle \boldsymbol e\rangle=\frac{\int \rho\,\boldsymbol e\,dV}{\int \rho\,dV},$$6	$$\boldsymbol{e}=\frac{1}{G M_1}\,\boldsymbol{v}\times(\boldsymbol{r}\times\boldsymbol{v})-\hat{\boldsymbol r}, \qquad \langle \boldsymbol e\rangle=\frac{\int \rho\,\boldsymbol e\,dV}{\int \rho\,dV},$$7, stream-fed	$$\boldsymbol{e}=\frac{1}{G M_1}\,\boldsymbol{v}\times(\boldsymbol{r}\times\boldsymbol{v})-\hat{\boldsymbol r}, \qquad \langle \boldsymbol e\rangle=\frac{\int \rho\,\boldsymbol e\,dV}{\int \rho\,dV},$$8
MHD-stream	strongest total $$\boldsymbol{e}=\frac{1}{G M_1}\,\boldsymbol{v}\times(\boldsymbol{r}\times\boldsymbol{v})-\hat{\boldsymbol r}, \qquad \langle \boldsymbol e\rangle=\frac{\int \rho\,\boldsymbol e\,dV}{\int \rho\,dV},$$9, truncated inside $m=1$0	$m=1$1
MHD-$m=1$2	$m=1$3, $m=1$4	$m=1$5
MHD-$m=1$6	$m=1$7, $m=1$8	$m=1$9
MHD-$\omega$0	$\omega$1, $\omega$2	$\omega$3

These runs show that even modest total $\omega$4–$\omega$5 is sufficient once the disk already occupies $\omega$6, while a stream-fed MHD disk can still fail despite large stress if overdense rings form and tides truncate the disk too early (Ohana et al., 2024). Nonlinearly, the eccentric MHD disks break into two misaligned eccentric regions separated by a circular annulus, develop standing eccentric waves in the inner disk, and form an eccentric central void when inner-disk pericenters reach the inflow boundary. By contrast, the hydrodynamic $\omega$7-disk shows smoother, aligned eccentricity growth without inner void formation (Ohana et al., 2024).

A useful transport diagnostic is

$\omega$8

In circular MRI turbulence, $\omega$9, preserving circularity; in eccentric flows, the mismatch is generically nonzero. The measured MHD behavior is that outer regions tend to circularize while the inner disk is pumped, so MRI transport moves angular momentum more effectively than energy in a way that favors inner-disk eccentricity growth (Ohana et al., 2024, Chan et al., 2022).

4. Vertical breathing, shocks, and thermodynamics

Unlike circular disks, eccentric disks experience a strongly phase-dependent vertical gravity,

$\Omega$0

which strengthens near pericenter and weakens near apocenter. One-dimensional vertical-column simulations show that this gravitational pumping generically drives supersonic compression toward pericenter; for $\Omega$1, supersonic motion is expected once $\Omega$2. In high-eccentricity, thick-disk cases, the resulting shocks reach Mach numbers $\Omega$3–$\Omega$4, heat gas to a sizable fraction of the virial temperature, and expel mass with $\Omega$5 per orbit at $\Omega$6 (Ryu et al., 2021).

After only a few orbits, these vertical solutions settle into a limit cycle with a low-entropy midplane and a much higher-entropy outer layer. This two-layer structure is a robust consequence of eccentric forcing rather than a peculiarity of any one initial condition (Ryu et al., 2021). In highly eccentric tidal-disruption flows, the same physics becomes a nonlinear “breathing” problem: the vertical scale height obeys a second-order ODE forced by the varying gravity and by orbital convergence, and the solutions include a nozzle-like structure near pericenter (Lynch et al., 2020).

Thermodynamics then becomes decisive. In the nonlinear eccentric-disk theory for TDEs, radiation-pressure–dominated solutions with stress proportional to total pressure are thermally unstable, whereas a stress proportional to gas pressure admits a nearly adiabatic stable branch at sufficiently large $\Omega$7 (Lynch et al., 2020). Thin, thermally stable eccentric disks radiate orbital energy over $\Omega$8 orbits, while thermally unstable or strongly thickened states can radiate on $\Omega$9 orbits (Lynch et al., 2020). This suggests that the eccentric state is controlled not only by orbital mechanics but also by whether the flow remains thin, cooling-limited, and weakly dissipative near pericenter.

5. Circumbinary, tidal-disruption, and AGN manifestations

In compact binaries, eccentricity manifests observationally as positive superhumps: the disk precesses apsidally with

$$\omega E=\frac{1}{2\Omega r^3\Sigma}\left\{\frac{d}{dr}\left(\gamma r^3P\frac{dE}{dr}\right)+r^2\frac{dP}{dr}E\right\},$$0

while tides drive prograde precession, pressure drives retrograde precession, and magnetic stresses can modulate both. The MHD simulations show precession slowing at saturation and even transient retrograde phases, implying time-variable superhump periods (Ohana et al., 2024).

Ring-fed circumbinary disks display a different timing signature. Their dominant power-spectral feature is

$$\omega E=\frac{1}{2\Omega r^3\Sigma}\left\{\frac{d}{dr}\left(\gamma r^3P\frac{dE}{dr}\right)+r^2\frac{dP}{dr}E\right\},$$1

half the canonical circumbinary lump frequency $$\omega E=\frac{1}{2\Omega r^3\Sigma}\left\{\frac{d}{dr}\left(\gamma r^3P\frac{dE}{dr}\right)+r^2\frac{dP}{dr}E\right\},$$2. Cold, compact rings can produce bulk gas eccentricities $$\omega E=\frac{1}{2\Omega r^3\Sigma}\left\{\frac{d}{dr}\left(\gamma r^3P\frac{dE}{dr}\right)+r^2\frac{dP}{dr}E\right\},$$3, a significant fraction reaching $$\omega E=\frac{1}{2\Omega r^3\Sigma}\left\{\frac{d}{dr}\left(\gamma r^3P\frac{dE}{dr}\right)+r^2\frac{dP}{dr}E\right\},$$4, while accretion at $$\omega E=\frac{1}{2\Omega r^3\Sigma}\left\{\frac{d}{dr}\left(\gamma r^3P\frac{dE}{dr}\right)+r^2\frac{dP}{dr}E\right\},$$5 is suppressed to $$\omega E=\frac{1}{2\Omega r^3\Sigma}\left\{\frac{d}{dr}\left(\gamma r^3P\frac{dE}{dr}\right)+r^2\frac{dP}{dr}E\right\},$$6 of the $$\omega E=\frac{1}{2\Omega r^3\Sigma}\left\{\frac{d}{dr}\left(\gamma r^3P\frac{dE}{dr}\right)+r^2\frac{dP}{dr}E\right\},$$7 case (Betancourt et al., 2 Jan 2026). In eccentric binaries, forced precession can modulate preferential accretion by up to two orders of magnitude, with a representative precession period of $$\omega E=\frac{1}{2\Omega r^3\Sigma}\left\{\frac{d}{dr}\left(\gamma r^3P\frac{dE}{dr}\right)+r^2\frac{dP}{dr}E\right\},$$8 for $$\omega E=\frac{1}{2\Omega r^3\Sigma}\left\{\frac{d}{dr}\left(\gamma r^3P\frac{dE}{dr}\right)+r^2\frac{dP}{dr}E\right\},$$9, $y=(r^3P)^{1/2}E$0 (Siwek et al., 2022). Equal-mass eccentric-binary simulations likewise show that one component can accrete $y=(r^3P)^{1/2}E$1–$y=(r^3P)^{1/2}E$2 times more rapidly than its companion for intervals of order $y=(r^3P)^{1/2}E$3, with the identity of the favored accretor alternating on $y=(r^3P)^{1/2}E$4 as the eccentric circumbinary disk slowly precesses (Muñoz et al., 2016).

In TDEs, eccentric-disk models predict low radiative efficiency and large emission radii. For nearly uniform disk eccentricity, soft X-rays generated at pericenter are trapped and advected, the photon-trapping radius is $y=(r^3P)^{1/2}E$5, the blackbody spectrum has a nearly single temperature typically about $y=(r^3P)^{1/2}E$6 K, and the radiative efficiency is $y=(r^3P)^{1/2}E$7 few $y=(r^3P)^{1/2}E$8, yielding sub-Eddington optical/UV luminosities with large photospheric radii (Liu et al., 2020). A related eccentric-accretion picture emphasizes that MHD stresses can remove angular momentum faster than orbital energy, allowing debris to plunge ballistically before circularization and reducing the efficiency to $y=(r^3P)^{1/2}E$9–$\omega_p(r)$0 of a standard accretion disk’s efficiency (Svirski et al., 2015).

For AGN, one recent synthesis proposes that moderately eccentric flows create an “eccentricity cascade” in the broad-line region, from $\omega_p(r)$1 inward to $\omega_p(r)$2 outward, with soft X-rays generated at periapsis and hard X-rays produced when general-relativistic apsidal precession compresses the inner flow. In that model, the high-frequency PSD turnover is tied to the orbital frequency near $\omega_p(r)$3, and precession of the eccentric flow drives optical and X-ray variability on timescales shorter than viscous inflow (Deng, 13 Jul 2025). The proposal is explicitly a dynamical alternative to the conventional BLR-plus-corona phenomenology rather than a settled consensus.

6. Open problems, misconceptions, and current boundaries

A persistent misconception is that MRI turbulence itself suppresses eccentricity growth. The compact-binary simulations indicate the opposite: if the disk reaches the 3:1 resonance, eccentricity grows in full MHD much as in hydrodynamic $\omega_p(r)$4-disks, and previous MHD failures are plausibly explained by insufficient outward spreading rather than by turbulence quenching the instability (Ohana et al., 2024). The unresolved problem is therefore not whether MRI permits eccentricity, but how real stream-fed MHD disks avoid overdense rings and spread to the resonant radius.

A second misconception is that eccentric TDE debris must circularize promptly. Several works instead show that dissipation near pericenter can be weak, that apocenter shocks may dominate early optical/UV output, and that the fate of the flow depends sensitively on thermodynamics, stress closure, and cooling time (Svirski et al., 2015, Lynch et al., 2020). This does not imply that circularization never occurs; rather, the timescale and location of circularization are model-dependent.

Several eccentric-disk phenomena remain numerically delicate. Minidisk eccentricity is suppressed by locally isothermal equations of state in CBD-fed runs and by softening lengths $\omega_p(r)$5 of the binary semi-major axis, which raises the possibility that some null results are artifacts of thermodynamics or vertical softening. The same study argues that a three-dimensional follow-up with well-resolved, geometrically thin minidisks, $\omega_p(r)$6, is needed to decide whether eccentric minidisks occur in real systems (Westernacher-Schneider et al., 2023).

The importance of vertical stratification, radiation, and realistic boundaries is equally clear. Many global calculations are 2D or vertically unstratified, locally isothermal, and non-self-gravitating; Class 0 simulations still omit Hall and Ohmic terms and treat only tightly coupled dust; GR circumbinary calculations prescribe the binary orbit and do not include MHD (Ohana et al., 2024, Betancourt et al., 2 Jan 2026, Ahmad et al., 13 Jan 2026, DeLaurentiis et al., 2024). At the same time, the mode theory shows that long-lived eccentricity itself is boundary-sensitive: realistic outer density tapers trap slow $\omega_p(r)$7 modes, whereas disks without such trapping can leak eccentricity away (Lee et al., 2019).

Taken together, these results define eccentric accretion disks as a broad dynamical class rather than a specialized anomaly. They encompass resonantly driven compact-binary disks, cavity-driven circumbinary disks, impact-coupled minidisks, thermodynamically sensitive TDE flows, streamer-fed protostellar disks, and proposed AGN broad-line/X-ray structures. The common thread is that noncircular motion reorganizes transport and dissipation at every level, from global precession and mode trapping to pericenter shocks, magnetic stresses, and observable timing structure.