Eccentric Roche Lobe Overflow

• Eccentric Roche Lobe Overflow is mass transfer in eccentric binaries that occurs mainly near periastron, with a phase-dependent Roche geometry.
• Analytic, hydrodynamic, and stellar evolution models reveal that transfer episodes can be impulsive or smooth, leading to disk formation, self-accretion, or common-envelope evolution.
• The phenomenon impacts systems from main-sequence binaries to X-ray transients by affecting orbital evolution, tidal interactions, and high-energy transient events.

Eccentric Roche-lobe overflow is mass transfer in a binary for which the donor fills its Roche lobe on an eccentric orbit, typically near periastron rather than throughout the orbit. In this regime the Roche geometry, the mass-transfer rate, and the orbital response are explicitly phase dependent, so the standard circular, synchronized RLOF approximation is generally inadequate. Across recent analytic, stellar-evolution, ballistic, and hydrodynamic studies, eccentric RLOF appears in systems ranging from main-sequence binaries and giant-star binaries to supergiant X-ray transients, planet-star systems, and star-black-hole binaries, with outcomes that include transient disks, self-accretion, eccentricity damping, eccentricity pumping, common-envelope onset, and residual post-detachment eccentricity (Hamers et al., 2018, Rocha et al., 2024, Bareli et al., 10 Apr 2026).

1. Orbital definition and Roche geometry

The defining feature of eccentric RLOF is that the donor’s effective Roche volume varies around the orbit because the instantaneous separation varies as

$r(\theta)=\frac{a(1-e^2)}{1+e\cos\theta},$

with semi-major axis $a$, eccentricity $e$, and true anomaly $\theta$. A widely used approximation is to evaluate the usual Eggleton Roche-lobe radius at this instantaneous separation, so that $R_{\rm L}(\theta)=R_{\rm L}^c(q)\,r(\theta)/a$, where $q=M_d/M_a$ and $R_{\rm L}^c(q)$ is the circular-orbit Eggleton radius (Hamers et al., 2018). In that approximation, a donor can underfill its lobe over most of the orbit and still overflow at periastron, where $r=a(1-e)$.

More detailed treatments include asynchronous rotation and the explicitly time-dependent Roche potential. Davis et al. introduced an asynchronism parameter,

$$\mathcal A(f)=\Bigl(\frac{\Omega_1}{\omega_{\rm peri}}\Bigr)^2 \frac{(1+e)^4}{[1+e\cos f]^3},$$

and computed the volume-equivalent Roche-lobe radius from the corotating equipotential rather than from the instantaneous Eggleton scaling alone (Davis et al., 2013). This matters because donor spin modifies both the size of the effective lobe and the duration of the overflow episode.

The same geometric logic is used in compact-object applications. In IGR J08408-4503, Ducci et al. evaluated the donor’s Roche-lobe radius at periastron as $R_L({\rm peri})\approx14.4\,R_\odot$, while the donor radius is $a$0–$a$1, implying periastron overflow (Ducci et al., 2019). In massive pre-common-envelope binaries, Vick et al. similarly approximated the onset of RLOF by the Eggleton radius evaluated at $a$2 (Vick et al., 2020). The common structure of these calculations is that eccentric RLOF is set by the periastron race between orbital separation and donor expansion.

2. Transfer prescriptions and stream morphology

A central modeling issue is whether the transfer episode is treated as impulsive or smoothly phase distributed. Sepinsky et al. formulated the high-eccentricity limit as a periastron delta-function mass-loss event,

$a$3

with $a$4 the true anomaly, and derived secular orbit-averaged expressions for $a$5 and $a$6 from that prescription (Sepinsky et al., 2010). The same formalism was later implemented in MESA for compact-object binaries, where it is coupled to tides, winds, spin-orbit coupling, gravitational radiation, and Eddington-limited accretion (Rocha et al., 2024).

Hamers and Dosopoulou constructed a smooth phase-dependent alternative valid for arbitrary eccentricity. In their conservative model with polytropic index $a$7, the orbit-dependent transfer rate is written as

$a$8

so that the circular limit is recovered continuously and the high-$a$9 limit reduces to the impulsive-periapsis picture (Hamers et al., 2018). This removes the unphysical behavior that delta-function prescriptions can exhibit as $e$0.

In full stellar-evolution calculations, the orbital modulation of $e$1 can be resolved directly. For the $e$2, $e$3, $e$4 d main-sequence binary studied with BINSTAR, the mass-transfer profile is Gaussian-like with a maximum at periastron. In the non-distorted model, $e$5 and the episode lasts $e$6 of the orbit; tidal and rotational deformation inflate the donor by $e$7 and raise the peak rate by about a factor of ten to $e$8 (Davis et al., 2013). In that same reference model, Davis et al. found a lower spin threshold $e$9 below which no transfer occurs and an upper threshold $\theta$0 above which transfer persists through the entire orbit.

Once material leaves $\theta$1, ballistic calculations identify three generic outcomes: direct impact onto the accretor, self-accretion onto the donor, or a quasi-periodic orbit around the accretor that can lead to disk formation (Sepinsky et al., 2010). In planet-star systems, Dosopoulou, Naoz, and Kalogera found that for prograde systems with $\theta$2, low eccentricity favors disk formation over much of parameter space, whereas for $\theta$3, retrograde spin, or $\theta$4, self-accretion dominates (Dosopoulou et al., 2017). Eccentric RLOF is therefore not synonymous with steady accretion onto the companion; the stream topology is part of the problem.

3. Orbital evolution, tides, and the failure of instant circularization

A persistent misconception is that binaries are effectively circular and synchronized when RLOF begins. Several studies now show that this assumption is not generically justified. In the ballistic framework of Sepinsky et al., direct impact can either increase or decrease the orbital semi-major axis and the eccentricity, whereas self-accretion always decreases both $\theta$5 and $\theta$6 (Sepinsky et al., 2010). In some circularizing configurations the eccentricity-damping timescale is only a few percent of the mass-transfer timescale, but in other direct-impact configurations mass overflow can increase $\theta$7, so the long-term behavior is controlled by competition between transfer and tides rather than by an automatic march toward $\theta$8.

The consequences of explicit eccentric mass transfer were quantified in self-consistent MESA calculations of star-compact-object binaries. Using the Sepinsky secular rates, the 2024 study of eccentric mass transfer found that a large fraction of systems remain eccentric after RLOF when one does not impose ad hoc instant circularization: in the full-physics sample, $\theta$9 remain eccentric post-RLOF and $R_{\rm L}(\theta)=R_{\rm L}^c(q)\,r(\theta)/a$0 circularize naturally (Rocha et al., 2024). Even binaries that do circularize differ substantially from instant-circularization predictions: final $R_{\rm L}(\theta)=R_{\rm L}^c(q)\,r(\theta)/a$1 are larger by $R_{\rm L}(\theta)=R_{\rm L}^c(q)\,r(\theta)/a$2–$R_{\rm L}(\theta)=R_{\rm L}^c(q)\,r(\theta)/a$3, and final $R_{\rm L}(\theta)=R_{\rm L}^c(q)\,r(\theta)/a$4 by $R_{\rm L}(\theta)=R_{\rm L}^c(q)\,r(\theta)/a$5–$R_{\rm L}(\theta)=R_{\rm L}^c(q)\,r(\theta)/a$6. The same work also showed that about $R_{\rm L}(\theta)=R_{\rm L}^c(q)\,r(\theta)/a$7 of systems differ qualitatively between the two treatments, and $R_{\rm L}(\theta)=R_{\rm L}^c(q)\,r(\theta)/a$8 of those are stable in the explicit eccentric-transfer calculation but unstable in the instant-circularized treatment. This directly affects predictions for X-ray binaries, gravitational-wave progenitors, and high-energy transients.

Pre-contact tides do not necessarily rescue the classical approximation. Vick et al. coupled post-main-sequence stellar evolution to the Vick-Lai turbulent-viscosity tidal formalism and found that, for a $R_{\rm L}(\theta)=R_{\rm L}^c(q)\,r(\theta)/a$9 primary with a $q=M_d/M_a$0 companion, systems with initial pericenter $q=M_d/M_a$1 AU retain essentially their initial eccentricity at the Roche radius; for a $q=M_d/M_a$2 primary the threshold is $q=M_d/M_a$3 AU (Vick et al., 2020). In systems that do not circularize, the donor spin is strongly sub-synchronous at RLOF. Even in systems that do circularize, the donor typically reaches only $q=M_d/M_a$4–$q=M_d/M_a$5, because stellar-expansion spin-down competes with tidal spin-up. The onset of eccentric RLOF can therefore be both eccentric and non-synchronous.

4. Residual eccentricity at detachment and stellar-population channels

In some binaries the orbit is nearly circular during the transfer phase, but a small eccentricity is imprinted at detachment. This is the mechanism analyzed for the formation of millisecond pulsars with intermediate-mass progenitors and CO white-dwarf companions. In the stable Case A/B channel, with progenitor masses $q=M_d/M_a$6, tides maintain an almost perfectly circular orbit during RLOF, but convection in the donor’s envelope drives stochastic fluctuations in the gravitational quadrupole moment. At detachment, when tidal damping weakens catastrophically as the donor contracts, the previously excited epicyclic energy is frozen in as a residual eccentricity (Bareli et al., 10 Apr 2026).

Following Phinney’s fluctuation-dissipation argument, the detachment eccentricity is obtained by equating epicyclic energy and convective kinetic energy. The resulting scaling is

$q=M_d/M_a$7

with normalization

$q=M_d/M_a$8

for typical $q=M_d/M_a$9–$R_{\rm L}^c(q)$0 (Bareli et al., 10 Apr 2026). Intermediate-mass donors detach at core-helium ignition, not when the hydrogen-burning shell can no longer be supported, so the envelope mass at detachment is larger than in low-mass degenerate-core donors. However, because the dependence is only $R_{\rm L}^c(q)$1, the increase in $R_{\rm L}^c(q)$2 is modest rather than order unity. This explains why intermediate-mass systems with $R_{\rm L}^c(q)$3 can have eccentricities similar to those of lower-mass He-white-dwarf systems. The same paper also derived an analytic $R_{\rm L}^c(q)$4 relation using $R_{\rm L}^c(q)$5 with $R_{\rm L}^c(q)$6, $R_{\rm L}^c(q)$7–$R_{\rm L}^c(q)$8, and $R_{\rm L}^c(q)$9, fitted to MESA tracks for initial masses $r=a(1-e)$0–$r=a(1-e)$1.

A different stellar-population application arises in Ba-star progenitors, where wind Roche-lobe overflow, circumbinary disks, and eccentricity pumping are coupled. In BINSTAR calculations, WRLOF is activated when $r=a(1-e)$2, or equivalently when $r=a(1-e)$3 (Krynski et al., 15 Apr 2025). WRLOF shrinks the orbit during the AGB phase, and the combination of WRLOF, circumbinary-disk resonant pumping, and tidally enhanced winds can drive RLOF on eccentric orbits down to $r=a(1-e)$4 d. Non-conservative RLOF can reduce the period before circularization to $r=a(1-e)$5 d, provided at least $r=a(1-e)$6 percent of the transferred mass leaves the system. By contrast, the same models do not reproduce the observed eccentricity distribution of Ba stars at $r=a(1-e)$7 d, where common-envelope evolution appears unavoidable.

5. Compact-object transients, disks, and disruption-limited regimes

Eccentric RLOF has direct observational consequences in high-energy transients. Ducci et al. argued that the supergiant fast X-ray transient IGR J08408-4503 undergoes RLOF around periastron because its high eccentricity and large donor radius imply $r=a(1-e)$8 (Ducci et al., 2019). Using the orbital solution $r=a(1-e)$9 d and $$\mathcal A(f)=\Bigl(\frac{\Omega_1}{\omega_{\rm peri}}\Bigr)^2 \frac{(1+e)^4}{[1+e\cos f]^3},$$0, with $$\mathcal A(f)=\Bigl(\frac{\Omega_1}{\omega_{\rm peri}}\Bigr)^2 \frac{(1+e)^4}{[1+e\cos f]^3},$$1 and $$\mathcal A(f)=\Bigl(\frac{\Omega_1}{\omega_{\rm peri}}\Bigr)^2 \frac{(1+e)^4}{[1+e\cos f]^3},$$2, they obtained $$\mathcal A(f)=\Bigl(\frac{\Omega_1}{\omega_{\rm peri}}\Bigr)^2 \frac{(1+e)^4}{[1+e\cos f]^3},$$3 and $$\mathcal A(f)=\Bigl(\frac{\Omega_1}{\omega_{\rm peri}}\Bigr)^2 \frac{(1+e)^4}{[1+e\cos f]^3},$$4, compared with a donor radius of $$\mathcal A(f)=\Bigl(\frac{\Omega_1}{\omega_{\rm peri}}\Bigr)^2 \frac{(1+e)^4}{[1+e\cos f]^3},$$5–$$\mathcal A(f)=\Bigl(\frac{\Omega_1}{\omega_{\rm peri}}\Bigr)^2 \frac{(1+e)^4}{[1+e\cos f]^3},$$6. In the Regős et al. picture for highly eccentric binaries, such periastron overflow can launch both $$\mathcal A(f)=\Bigl(\frac{\Omega_1}{\omega_{\rm peri}}\Bigr)^2 \frac{(1+e)^4}{[1+e\cos f]^3},$$7 and $$\mathcal A(f)=\Bigl(\frac{\Omega_1}{\omega_{\rm peri}}\Bigr)^2 \frac{(1+e)^4}{[1+e\cos f]^3},$$8 streams. Most of the $$\mathcal A(f)=\Bigl(\frac{\Omega_1}{\omega_{\rm peri}}\Bigr)^2 \frac{(1+e)^4}{[1+e\cos f]^3},$$9 stream binds to the neutron star and can form an accretion disk, while $R_L({\rm peri})\approx14.4\,R_\odot$0 mass loss can contribute to a clumpy circumbinary environment.

For IGR J08408-4503, the inferred circularization radius is of order $R_L({\rm peri})\approx14.4\,R_\odot$1–$R_L({\rm peri})\approx14.4\,R_\odot$2 cm, and the viscous timescale is of order days for $R_L({\rm peri})\approx14.4\,R_\odot$3 and $R_L({\rm peri})\approx14.4\,R_\odot$4, consistent with day-long outbursts (Ducci et al., 2019). In a magnetically gated interpretation, episodic periastron RLOF supplies the transient mass-transfer surge, and transitions occur when the inner-disk radius approaches the corotation radius. The observed $R_L({\rm peri})\approx14.4\,R_\odot$5 s flares with $R_L({\rm peri})\approx14.4\,R_\odot$6–$R_L({\rm peri})\approx14.4\,R_\odot$7 are then linked to periodic RLOF supply and barrier crossing. For $R_L({\rm peri})\approx14.4\,R_\odot$8–$R_L({\rm peri})\approx14.4\,R_\odot$9 and $a$00, the inferred magnetic moment is $a$01–$a$02, corresponding to $a$03–$a$04 G and $a$05–$a$06 s.

At the most extreme end, eccentric mass transfer can bridge into tidal-disruption-like behavior. Chen and Lai followed a Sun-like star orbiting a $a$07 stellar-mass black hole with SPH simulations and found two sharply separated pathways: a stable branch for $a$08 and a runaway branch for $a$09, where $a$10 is the pericenter distance in units of the tidal radius (Chen et al., 3 Jun 2026). In the stable branch, the system settles into long-lived eccentric mass transfer lasting up to 150 orbits in the simulations, with repeated pericenter flares and negative feedback from orbital widening. In the runaway branch, adiabatic stellar expansion outpaces the Roche-lobe response, leading to unstable RLOF and eventual disruption; the stripped debris forms a thick torus with $a$11–$a$12, and the accretion rate through the sink reaches $a$13. This establishes a hydrodynamic continuity between self-regulated eccentric transfer and micro-TDE-like runaway destruction.

6. Common-envelope onset and unresolved boundaries

Hydrodynamic calculations of giant-star binaries indicate that once eccentric RLOF becomes strong, it can be intrinsically short-lived. In simulations of a $a$14 AGB star of radius $a$15 au interacting with a companion, Staff et al. found that periastron RLOF requires approximately $a$16 for $a$17, and that systems with strong RLOF enter a common envelope within $a$18–$a$19 periastron passages (Staff et al., 2015). In their representative $a$20, $a$21 case, the first periastron passage strips $a$22 into a disk, with $a$23 unbound, and the orbital period shrinks from $a$24 yr to $a$25 yr after the first strong interaction. Final separations are $a$26 au, subject to numerical resolution limits.

The pre-common-envelope accretion structures in those AGB calculations are themselves astrophysically consequential. Disk masses are a few $a$27, $a$28, and the accretion flow is formally super-Eddington by up to a factor $a$29 for a $a$30 companion (Staff et al., 2015). With a magneto-centrifugal wind efficiency $a$31, the resulting jet estimates are $a$32–$a$33, $a$34–$a$35, and $a$36–$a$37 erg over a 10 yr episode. The same simulations, however, do not support a long-lived series of periastron RLOF eruptions recurring over centuries in a pure binary.

Taken together, these studies imply that eccentric Roche-lobe overflow is best viewed as a family of related regimes rather than a single phenomenology: impulsive or smooth periastron transfer, disk-forming or self-accreting stream trajectories, secular circularization or eccentricity pumping, residual post-detachment eccentricity, and hydrodynamic transition to common-envelope evolution or disruption. Several boundaries remain open. Massive CO and ONe white-dwarf companions with $a$38 have measured eccentricities that the intermediate-mass detachment model does not explain (Bareli et al., 10 Apr 2026). Ba-star calculations still leave the $a$39 d population outside the modeled eccentric-RLOF channel (Krynski et al., 15 Apr 2025). Vick et al. explicitly concluded that the effects of pre-common-envelope eccentricity on the resulting compact binary merit further study (Vick et al., 2020).