Roche-Lobe Overflow and Mass Transfer

A single star sets its own boundary: gravity holds its gas in, and its surface is a sphere. Put two stars in a close orbit and that private geometry dissolves. In the frame that co-rotates with the binary, a parcel of gas feels three forces at once — the pull of its own star, the pull of the companion, and the centrifugal push of the rotating frame. Their combined Roche potential has equipotential surfaces that are nearly spherical deep in each star's well but distort outward into teardrops, and the critical surface that just touches at the inner Lagrangian point $L_1$ divides space into two lobes, one per star. That lobe is the largest a star can grow before its outer layers no longer belong to it. Roche-lobe overflow (RLOF) is what happens when a star fills its lobe. It can be driven from the star's side — a normal star swells by orders of magnitude when it leaves the main sequence — or from the orbit's side, as gravitational radiation and magnetic braking bleed away angular momentum and shrink the lobe onto a star of fixed size. Either way, gas piling up at the $L_1$ saddle spills through the nozzle into the companion. Whether the stream is a gentle trickle or a catastrophic flood is decided by a race between two radii — the star's and its lobe's — whose sign is fixed by the mass ratio. The consequences are among the loudest phenomena in the sky. The transferred gas cannot fall straight in; it carries orbital angular momentum, swings into an accretion disc, and lights up as it spirals down. This one process builds cataclysmic variables and novae on white dwarfs, the brightest X-ray binaries on neutron stars and black holes, the recycled millisecond pulsars, and — near the Chandrasekhar mass — the Type Ia supernovae that map the Universe's expansion.

Key takeaways

In a close binary each star's gravitational territory is bounded by its Roche lobe — the largest closed equipotential of the co-rotating (Roche) potential enclosing only that star, meeting the companion's lobe at the inner Lagrangian point $L_1$ .
When a star swells off the main sequence, or the orbit shrinks through angular-momentum loss, the photosphere reaches the lobe and gas streams through $L_1$ — Roche-lobe overflow.
The Eggleton formula gives the volume-equivalent lobe radius $R_L/a$ as a function of the mass ratio $q=M_1/M_2$ to better than 1 % across all $q$ ; at $q=1$ it is $R_L/a\approx0.38$ .
Stability turns on how the donor's radius and its lobe respond to mass loss: conservative transfer from the lighter to the heavier star widens the orbit and self-limits, while the reverse shrinks it and runs away.
The infalling gas keeps its angular momentum, forms an accretion disc, and liberates $L_{\mathrm{acc}}=GM\dot M/R$ — capped at the Eddington luminosity $L_{\mathrm{Edd}}\approx1.3\times10^{38}(M/M_\odot)\ \mathrm{erg\,s^{-1}}$ .
RLOF powers cataclysmic variables, X-ray binaries, Algol systems, Type Ia progenitors, and the recycling of old pulsars to millisecond spins.
Key idea: the Roche lobe is the critical equipotential of the co-rotating binary potential; a star that grows to fill it transfers mass through the $L_1$ saddle, and the sign of the orbital response to that transfer — set by the mass ratio — decides whether mass transfer self-regulates or runs away.

The core insight

The Roche lobe is the critical equipotential of the co-rotating binary potential; a star that grows to fill it transfers mass through the $L_1$ saddle point, and the sign of the orbital response to that transfer — set by the mass ratio — decides whether mass transfer is stable or runs away.

Where this lives today

Roche-lobe overflow is the working backbone of high-energy astrophysics. NASA's NICER times X-ray pulsations from accreting neutron stars to map their radii and constrain the dense-matter equation of state, while Chandra and XMM-Newton survey the X-ray binary populations of galaxies. Ultraluminous X-ray sources — points shining hundreds of times above the naive Eddington limit — are now read mostly as neutron stars accreting through beamed, magnetically channelled super-Eddington flows, a live test of equation (6). Radio timing arrays (NANOGrav, the EPTA) use recycled millisecond pulsars, each the fossil of a completed RLOF episode, as galactic-scale clocks to hunt the nanohertz gravitational-wave background. Type Ia supernovae, the standard candles that revealed cosmic acceleration, trace back to white dwarfs pushed toward the Chandrasekhar mass by RLOF or mergers. And LIGO-Virgo-KAGRA's merging compact objects are the end products of binaries that survived mass transfer and common-envelope evolution, so calibrating stability criteria like $\zeta_{\mathrm{ad}}$ versus $\zeta_L$ feeds directly into predicting the gravitational-wave population.

The physical picture

Imagine the effective-potential landscape as a rubber sheet with two deep wells, one under each star, joined by a low mountain pass. As long as a star's surface sits below the pass, its atmosphere is trapped in its own well and the star is round and self-contained. Let the star grow — or lower the pass by shrinking the orbit — until its surface reaches $L_1$ , and gas spills over the saddle into the neighbouring well. That spillover is mass transfer, and because $L_1$ is a single point rather than a rim, the outflow is a narrow, directed stream, not an isotropic wind.

Figure

Equipotentials of the co-rotating Roche potential in the orbital plane: nested surfaces hug each star, and the critical figure-eight through the inner Lagrangian point L₁ pinches off the two teardrop-shaped Roche lobes, while the outer points L₂, L₃ and the triangular L₄, L₅ complete the five equilibria. At equal masses q = 1 the pass sits at the midpoint and R_L/a ≈ 0.38

Whether transfer is gentle or catastrophic is a race between two radii. Strip mass from the donor and its radius changes by an amount set by its envelope physics; simultaneously the Roche lobe changes by an amount set by orbital mechanics. If the lobe contracts faster than the star can shrink, the overfill deepens and transfer runs away on a dynamical or thermal timescale. If instead the star withdraws inside a widening lobe, the flow chokes off and self-regulates, transferring only as fast as some slow driver pushes it. The mass ratio decides which happens, and the Eggleton curve below shows how large the lobe is to begin with.

Figure

Eggleton's approximation R_L/a as a function of the mass ratio q = M₁/M₂, accurate to better than 1 percent across all q: the volume-equivalent Roche-lobe radius climbs from a small fraction of the separation when the star is light to more than half the separation when it dominates. Marked: q = 0.25 (a cataclysmic-variable donor) at R_L/a ≈ 0.27, equal masses q = 1 at 0.38, and q = 3 at 0.48

Finally, the geometry of the infall. Gas leaving $L_1$ shares the donor's orbital motion, so it arrives with a large sideways velocity and cannot strike the compact star head-on; it swings past, self-intersects, and settles into a ring at the circularisation radius. Viscous friction spreads the ring into a disc that carries angular momentum outward while matter spirals inward, converting gravitational energy into accretion luminosity and spin-up torque. The disc is the engine room of every RLOF system.

The core equations

The starting point is the potential a co-rotating parcel of gas feels. In the frame rotating at the binary's angular frequency $\Omega$ , the two gravitational wells add to a centrifugal term to give the Roche potential — whose geometry defines everything downstream.

\Phi(\mathbf{r}) = -\frac{GM_1}{|\mathbf{r}-\mathbf{r}_1|} - \frac{GM_2}{|\mathbf{r}-\mathbf{r}_2|} - \frac{1}{2}\Omega^2\left(x^2+y^2\right) \qquad (1)

where: $\Phi$ — effective potential per unit mass in the co-rotating frame; $M_1,M_2$ — stellar masses at positions $\mathbf{r}_1,\mathbf{r}_2$ ; $\Omega$ — orbital angular frequency; $x,y$ — coordinates in the orbital plane with origin at the centre of mass; $G$ — Newton's constant. The first two terms are ordinary gravity; the third is the outward centrifugal push that flattens the outer equipotentials into teardrops. The centrifugal term is not free — a circular orbit fixes $\Omega$ through Kepler's third law, tying the potential's shape to the separation $a$ .

\Omega^2 = \frac{G(M_1+M_2)}{a^3} \qquad (2)

where: $a$ — orbital separation (semi-major axis); $M_1+M_2$ — total mass. Shorter periods mean smaller $a$ , so a close binary has a compact, tightly pinched pair of lobes. The lobe is not spherical, so one quotes its volume-equivalent radius $R_L$ — the radius of a sphere of the same volume. Eggleton's fit gives it as a function of the mass ratio alone, good to better than 1 % for every $q$ .

\frac{R_L}{a} = \frac{0.49\,q^{2/3}}{0.6\,q^{2/3} + \ln\!\left(1+q^{1/3}\right)} \qquad (3)

where: $R_L$ — volume-equivalent Roche-lobe radius of star 1; $q=M_1/M_2$ — mass ratio of the star whose lobe is computed; $a$ — separation. At $q=1$ this returns $R_L/a\approx0.38$ ; it rises toward $0.5$ – $0.7$ for a dominant star and falls toward $\sim0.46\,(M_1/M_{\mathrm{tot}})^{1/3}$ for a light one. The orbit's angular momentum is the conserved quantity governing the response to transfer; for a circular binary it takes a compact form.

J = M_1 M_2 \sqrt{\frac{Ga}{M_1+M_2}} \qquad (4)

where: $J$ — orbital angular momentum; $M_1,M_2$ — component masses; $a$ — separation. At fixed $J$ and total mass, moving mass toward equal masses (maximising $M_1M_2$ ) forces $a$ to a minimum — the fact behind Algol mass-ratio reversal. When gas lands on the accretor it releases its gravitational binding energy as radiation; that is the accretion luminosity, whose efficiency scales with the compactness $M/R$ of the accretor.

L_{\mathrm{acc}} = \frac{GM_{\mathrm{acc}}\dot{M}}{R_{\mathrm{acc}}} \qquad (5)

where: $L_{\mathrm{acc}}$ — luminosity liberated by accretion; $M_{\mathrm{acc}},R_{\mathrm{acc}}$ — mass and radius of the accretor; $\dot M$ — accretion rate. A neutron star's tiny $R$ makes $L_{\mathrm{acc}}$ enormous — tens of percent of $\dot M c^2$ — whereas a white dwarf yields only $\sim10^{-4}\dot M c^2$ . That luminosity cannot grow without bound: its own radiation pressure halts the infall, capping steady accretion at the Eddington luminosity.

L_{\mathrm{Edd}} = \frac{4\pi G M_{\mathrm{acc}} m_p c}{\sigma_T} \approx 1.3\times10^{38}\,\frac{M_{\mathrm{acc}}}{M_\odot}\ \mathrm{erg\,s^{-1}} \qquad (6)

where: $m_p$ — proton mass; $c$ — speed of light; $\sigma_T$ — Thomson cross-section. The limit depends only on the accretor's mass, not its radius, because both gravity and radiation force fall as $r^{-2}$ . Finally, stability is a comparison of two logarithmic slopes — how the lobe radius and the donor's own radius each respond to mass loss.

\zeta_L \equiv \frac{\mathrm{d}\ln R_L}{\mathrm{d}\ln M_1}, \qquad \zeta_{\mathrm{ad}} \equiv \left(\frac{\mathrm{d}\ln R_1}{\mathrm{d}\ln M_1}\right)_{\mathrm{ad}} \qquad (7)

where: $\zeta_L$ — response of the lobe radius to donor mass loss; $\zeta_{\mathrm{ad}}$ — adiabatic response of the donor's radius; $R_1,M_1$ — donor radius and mass. If the donor shrinks faster than its lobe ( $\zeta_{\mathrm{ad}}>\zeta_L$ ) transfer is stable; if the lobe shrinks faster, it runs away.

Derivation ALocating the inner Lagrangian point and the birth of the Roche lobe

The goal is to show, from the Roche potential alone, that a single interior saddle $L_1$ sits between the two stars, that the equipotential through it pinches into a figure-eight, and that the two teardrops it encloses are the Roche lobes whose volume-equivalent radius Eggleton later fitted. We restrict the potential to the star-connecting axis, demand a stationary point, examine its transverse character, and identify the critical surface — turning the abstract landscape of equation (1) into the geometry that governs mass transfer.

Step 1. Restrict equation (1) to the line joining the stars ( $y=z=0$ ): $\Phi(x)=-GM_1/|x-x_1|-GM_2/|x-x_2|-\tfrac12\Omega^2x^2$ , centre of mass at the origin so $M_1x_1+M_2x_2=0$ . Why: the collinear equilibria live on this axis by symmetry, so the three-dimensional problem collapses to one variable.

Step 2. Demand $\mathrm{d}\Phi/\mathrm{d}x=0$ . Between the stars this gives $GM_1/(x-x_1)^2-GM_2/(x_2-x)^2-\Omega^2x=0$ , where the two gravitational pulls and the centrifugal term exactly cancel. Why: a stationary point of $\Phi$ is a point of zero net force in the rotating frame, which is what a Lagrangian point is.

Step 3. Inspect the interior root. Along the axis $\Phi$ has a local maximum (a parcel displaced along the line rolls away), but transverse to the axis it curves upward (a sideways displacement is pushed back). The point is a saddle — the lowest pass between the two wells — named $L_1$ . Why: the saddle character is why the stream is narrow and directed; gas crosses only by threading this one low point, not over the whole surface.

Step 4. Follow the equipotential $\Phi=\Phi(L_1)$ . It self-intersects at $L_1$ , pinching into a figure-eight whose two closed lobes wrap the stars. Each lobe is the largest closed equipotential containing only its own star; a star that touches it begins to leak gas through $L_1$ . This is the Roche lobe. Why: this is the operational definition of "largest self-contained size" — beyond it, gas can flow to the companion at no energy cost.

Step 5. Since the lobe is a teardrop, characterise it by the radius $R_L$ of a sphere of equal volume; fitting that across all mass ratios yields the Eggleton formula, equation (3). Why: a single scalar $R_L$ lets one compare the lobe against a stellar radius $R_1$ directly, all the stability analysis needs. *As a check: equation (3) gives $R_L/a=0.379$ at $q=1$ and $R_L/a\to0.46\,(M_1/M_{\mathrm{tot}})^{1/3}$ as $q\to0$ , both standard limiting values.*

Go deeper · Derivation B · 1 minThe orbital response to conservative mass transfer

Derivation BThe orbital response to conservative mass transfer

The goal is to derive how the separation $a$ changes when the donor loses mass, assuming all mass and angular momentum stay in the binary. The result is a single expression whose sign flips at $M_1=M_2$ — the master switch between stable, self-regulating transfer and a runaway, and the most important qualitative fact in the subject. We differentiate equation (4) logarithmically, impose conservation, and read off the criterion.

Step 1. Start from $J=M_1M_2\sqrt{Ga/M}$ with $M$ held constant (conservative transfer: whatever leaves the donor lands on the accretor, $\dot M_2=-\dot M_1$ ). Take the logarithm: $\ln J=\ln M_1+\ln M_2+\tfrac12\ln a+\text{const}$ . Why: logarithmic differentiation turns the products and powers of $J$ into a simple sum of fractional rates — the natural language for "how does each quantity respond".

Step 2. Differentiate in time: $\dot J/J=\dot M_1/M_1+\dot M_2/M_2+\tfrac12\,\dot a/a$ ; the total-mass term is absent since $M$ is fixed. Why: this isolates $\dot a/a$ in terms of the transfer rate and any angular-momentum loss.

Step 3. Impose $\dot J=0$ and $\dot M_2=-\dot M_1$ , giving $0=\dot M_1/M_1-\dot M_1/M_2+\tfrac12\,\dot a/a$ . Solve: $\dot a/a=-2\dot M_1\left(1/M_1-1/M_2\right)=-2(\dot M_1/M_1)\left(1-M_1/M_2\right)$ . Why: this closed form predicts the direction the orbit moves for any mass ratio, using nothing but conservation laws.

Step 4. Read off the criterion. The donor loses mass, so $\dot M_1<0$ . If the donor is heavier ( $M_1>M_2$ ), the bracket is negative and $\dot a<0$ : the orbit shrinks, the lobe tightens onto the still-overflowing donor, and transfer runs away. If the donor is lighter ( $M_1<M_2$ ), $\dot a>0$ : the orbit widens and transfer throttles itself. Why: this single sign organises the whole zoo of interacting binaries — heavy-donor systems plunge into common envelopes while light-donor systems settle into long-lived, gently accreting states. *As a check: $\dot a=0$ at $M_1=M_2$ for fixed $J,M$ , since $J\propto M_1M_2$ peaks there — the mass-ratio-reversal turning point.*

Go deeper · Derivation C · 1 minThe Eddington limit on accretion

Derivation CThe Eddington limit on accretion

The goal is to find the maximum luminosity — and hence the maximum steady accretion rate — a compact object can sustain before its own radiation pressure blows the incoming gas back out. The result depends only on the accretor's mass and sets the brightness ceiling of X-ray binaries. We balance the outward radiation force on an electron-proton pair against the inward gravity on the same pair, and note that the radius cancels.

Step 1. Ionised infalling gas is a plasma of electrons and protons. Radiation of luminosity $L$ pushes chiefly on the electrons through Thomson scattering, with force per electron $F_{\mathrm{rad}}=\sigma_T L/(4\pi r^2 c)$ . Why: momentum flux is $L/c$ spread over a sphere, and $\sigma_T$ is the area an electron presents; protons scatter $\sim(m_e/m_p)^2$ as weakly and are ignored.

Step 2. Gravity pulls chiefly on the protons, $F_{\mathrm{grav}}=GM_{\mathrm{acc}}m_p/r^2$ ; electrostatic coupling locks each electron to a proton, so the pair moves together. Why: the plasma cannot separate its charges, so the balance is on the pair — radiation grabbing the electron, gravity holding the proton.

Step 3. Set the two equal, $\sigma_T L_{\mathrm{Edd}}/(4\pi r^2 c)=GM_{\mathrm{acc}}m_p/r^2$ . The $r^{-2}$ cancels, leaving $L_{\mathrm{Edd}}=4\pi GM_{\mathrm{acc}}m_p c/\sigma_T\approx1.3\times10^{38}(M_{\mathrm{acc}}/M_\odot)\ \mathrm{erg\,s^{-1}}$ . Why: because both forces fall as $1/r^2$ , the balance is scale-free — the limit is a property of the accretor's mass alone.

Step 4. Convert to a maximum rate using equation (5): $L_{\mathrm{acc}}=L_{\mathrm{Edd}}$ gives $\dot M_{\mathrm{Edd}}=4\pi m_p c\,R_{\mathrm{acc}}/\sigma_T$ ; above it the source drives outflows and accretes inefficiently. Why: RLOF can deliver gas faster than the accretor can radiate it away, so persistent super-Eddington sources demand beaming or magnetic channelling. *As a check: inserting the constants gives $L_{\mathrm{Edd}}=1.26\times10^{38}(M/M_\odot)\ \mathrm{erg\,s^{-1}}$ ; for a $1.4\,M_\odot$ neutron star this yields $1.76\times10^{38}\ \mathrm{erg\,s^{-1}}$ .*

Limiting cases

The lobe geometry and the fate of the transfer are both controlled by two numbers — the mass ratio $q$ and the donor's structural response $\zeta_{\mathrm{ad}}$ :

Case	Result	Meaning
Equal masses ( $q=1$ )	$R_L/a \approx 0.38$	Symmetric lobes, $L_1$ at the midpoint; the minimum-separation point for conservative transfer.
Light donor ( $q\ll1$ )	$R_L/a \approx 0.46\,(M_1/M_{\mathrm{tot}})^{1/3}$	Lobe set by the donor's mass fraction; with Kepler this gives the period–mean-density relation.
Heavy donor ( $q\gg1$ )	$R_L/a \to 0.6$ – $0.7$	The dominant star claims most of the volume; transferring FROM it shrinks the orbit and is typically unstable.
Radiative vs convective donor	$\zeta_{\mathrm{ad}} \gtrless \zeta_L$	Radiative envelopes shrink on mass loss ( $\zeta_{\mathrm{ad}}>0$ ), favouring stability; deep convective envelopes expand ( $\zeta_{\mathrm{ad}}\lesssim-1/3$ ), favouring runaway.
Conservative transfer, light → heavy	$\dot a/a > 0$	Orbit widens, lobe grows: stable, self-limiting transfer (Algols after mass-ratio reversal).

Common trapThe Roche lobe is not a density surface, and

q

is easy to mislabel

Two traps ensnare newcomers. First, the Roche lobe is a purely gravitational-plus-centrifugal construct: it assumes synchronous rotation, a circular orbit, and centrally condensed (effectively point) masses. It is a surface of constant effective potential, not of constant density or pressure — a real photosphere only approximately coincides with it, and rotation or eccentricity distorts the picture. Second, whether transfer widens or shrinks the orbit hinges entirely on whether the donor is the more or less massive star, yet $q=M_1/M_2$ is defined differently in different textbooks (donor-over-accretor here, sometimes the reverse). Many spurious "the orbit widens" conclusions are just a flipped ratio — always anchor the physics to "is the donor heavier than the accretor?", not to the symbol.

Worked examples

1 · A cataclysmic variable: lobe size and the period–density link

A white dwarf $M_2=0.8\,M_\odot$ accretes from a lobe-filling companion $M_1=0.2\,M_\odot$ in a binary of period $P=100\ \mathrm{min}$ . Find the separation $a$ , the donor's lobe radius $R_L$ , and its mean density, and show the density is fixed by the period alone.

Step 1. Separation from Kepler, equation (2), in solar units ( $a^3_{\mathrm{AU}}=(M/M_\odot)(P/\mathrm{yr})^2$ ). With $M=1.0\,M_\odot$ and $P=100\ \mathrm{min}=1.90\times10^{-4}\ \mathrm{yr}$ , $a=3.31\times10^{-3}\ \mathrm{AU}=\mathbf{0.711\,R_\odot}$ — a genuinely close binary.

Step 2. Donor lobe radius from Eggleton, equation (3), with $q=M_1/M_2=0.25$ : $R_L/a=0.49(0.25)^{2/3}/[0.6(0.25)^{2/3}+\ln(1+0.25^{1/3})]=\mathbf{0.268}$ , so $R_L=0.268\times0.711\,R_\odot=\mathbf{0.190\,R_\odot}$ . A $0.2\,M_\odot$ M dwarf has almost exactly this radius, which is why it fills its lobe at this period.

Step 3. Mean density: $\bar\rho=M_1/(\tfrac43\pi R_L^3)$ , with $M_1=3.98\times10^{32}\ \mathrm{g}$ and $R_L=1.32\times10^{10}\ \mathrm{cm}$ , gives $\bar\rho=\mathbf{41\ g\,cm^{-3}}$ .

Step 4. This is no coincidence. For a light donor $R_L/a\approx0.46(M_1/M)^{1/3}$ ; cube it, insert $a^3$ from Kepler, and the donor mass cancels: $\bar\rho\approx3\pi/(0.0986\,GP^2)\approx110\,P_{\mathrm{hr}}^{-2}\ \mathrm{g\,cm^{-3}}$ . At $P=1.667\ \mathrm{hr}$ this gives $\mathbf{40\ g\,cm^{-3}}$ , matching Step 3. The mean density of a lobe-filling donor depends only on the orbital period — which is why cataclysmic variables cluster near $P\sim2\ \mathrm{hr}$ with $\sim0.2\,M_\odot$ M-dwarf donors.

2 · How bright can a neutron star shine, and how efficient is it?

Estimate the Eddington luminosity of a $1.4\,M_\odot$ , $10\ \mathrm{km}$ neutron star, its maximum accretion rate, and the efficiency $\eta=GM/(Rc^2)$ ; compare $\eta$ with a white dwarf.

Step 1. Eddington luminosity from equation (6): $L_{\mathrm{Edd}}=1.3\times10^{38}\times1.4=\mathbf{1.8\times10^{38}\ erg\,s^{-1}}$ , some $5\times10^4\,L_\odot$ from an object the size of a city — the reason low-mass X-ray binaries are the brightest persistent point sources in a galaxy.

Step 2. Maximum accretion rate $\dot M_{\mathrm{Edd}}=L_{\mathrm{Edd}}R/(GM)$ : with $R=10^6\ \mathrm{cm}$ and $M=1.4\,M_\odot$ , $\dot M_{\mathrm{Edd}}\approx9.5\times10^{17}\ \mathrm{g\,s^{-1}}=\mathbf{1.5\times10^{-8}\,M_\odot\,yr^{-1}}$ . RLOF can easily exceed this, so the excess is blown off or advected.

Step 3. Efficiency: $\eta=GM/(Rc^2)=(6.67\times10^{-8})(1.4)(1.99\times10^{33})/[(10^6)(3\times10^{10})^2]=\mathbf{0.21}$ . About one-fifth of the accreted rest-mass energy emerges as light — far above the $0.007$ of hydrogen fusion (GR gives $\sim0.1$ – $0.2$ depending on spin).

Step 4. For a white dwarf ( $M=0.6\,M_\odot$ , $R=7000\ \mathrm{km}$ ) the same formula gives $\eta_{\mathrm{WD}}\approx1.3\times10^{-4}$ . Neutron-star accretion is roughly $1600\times$ more efficient than white-dwarf accretion and beats nuclear burning — which is why swapping a white dwarf for a neutron star turns a faint cataclysmic variable into a luminous X-ray binary.

Physics of the portrait

The Roche lobe carries the name of a nineteenth-century French astronomer who never saw a binary transfer mass, but whose analysis of how one body deforms another's gravitational reach became the master key to interacting stars.

Origins · 1 minÉdouard Roche and how this was discovered

ÉR

Édouard Roche

1820–1883 · celestial mechanics and the equilibrium figures of tidally distorted bodies

Édouard Albert Roche spent his career at Montpellier on a deceptively narrow problem: the shape a fluid body takes in the gravitational field of a companion. In the 1840s and 1850s he found the critical surface beyond which a satellite's own gravity can no longer hold it together against the primary's tides — the Roche limit, and the reason Saturn's rings sit where they do. The same effective-potential picture, extended to two comparable masses in a rotating frame, produces the critical figure-eight equipotential whose two teardrops now bear his name. Roche himself worked on tides, cometary tails, and the nebular hypothesis, not mass transfer; the astrophysical afterlife of his geometry came only when twentieth-century astronomers realised that close-binary stars evolve to fill these lobes. The modern theory rests on the equilibrium points found by Euler and Lagrange, on the 1950s resolution of the Algol paradox by mass exchange, and above all on Bohdan Paczyński's 1971 synthesis of close-binary evolution and Peter Eggleton's 1983 fit for $R_L/a$ — the results that turned Roche's static geometry into a working theory of how stars trade mass, spin each other up, and detonate.

History

1772

— Lagrange identifies the five equilibrium points of the restricted three-body problem

1849

— Roche formulates the critical equipotential and tidal limit of a body under a companion's gravity

1955

— Crawford resolves the Algol paradox by proposing mass exchange between close binary components

1971

— Paczyński codifies close-binary evolution and the mass-ratio stability criterion

1983

— Eggleton publishes his 1-percent-accurate analytic fit for the Roche-lobe radius

Going further

Texts. Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., Ch. 18 (Close Binary Systems) · Frank, King & Raine, Accretion Power in Astrophysics, 3rd ed., Ch. 4–5 · Binney & Tremaine, Galactic Dynamics, 2nd ed. (restricted three-body problem and the Jacobi integral) · Hilditch, An Introduction to Close Binary Stars · Eggleton, P. P. 1983, ApJ 268, 368 (the $R_L/a$ approximation) · Paczyński, B. 1971, ARA&A 9, 183 (evolutionary processes in close binaries). Key results to keep in mind: Eggleton's (1983) fit is accurate to better than 1 % for all $q$ ; the conservative-transfer response is $\dot a/a=-2(\dot M_1/M_1)(1-M_1/M_2)$ , obtained by log-differentiating $J=M_1M_2\sqrt{Ga/M}$ at fixed $J,M$ ; the Eddington luminosity evaluates to $L_{\mathrm{Edd}}=1.26\times10^{38}(M/M_\odot)\ \mathrm{erg\,s^{-1}}$ ; and the CV worked example gives the chain $a=0.711\,R_\odot$ , $R_L=0.190\,R_\odot$ , $\bar\rho\approx40\ \mathrm{g\,cm^{-3}}$ , and $\bar\rho\approx110\,P_{\mathrm{hr}}^{-2}\ \mathrm{g\,cm^{-3}}$ .

Prerequisites. Newtonian Gravity and Orbits · Effective Potentials and Rotating Frames · Angular Momentum Conservation. Leads to. White Dwarfs · Chandrasekhar Mass Limit · Millisecond and Recycled Pulsars · Neutron Stars.

Ask the tutor: ask me to walk through why the orbit's response to mass transfer flips sign exactly at equal masses, or to work a stability estimate for any donor you like.

What is the difference between the Roche lobe and the Roche limit?

They share a name and an author but describe different things. The Roche limit is the distance within which a fluid satellite held together only by self-gravity is torn apart by a primary's tides — it governs planetary rings and disrupted comets. The Roche lobe is the critical equipotential of the co-rotating two-body potential, the largest closed surface enclosing a single star before gas can flow to the companion through $L_1$ . The limit is about cohesion against tidal shear; the lobe is about a star's maximum size before mass transfer. Both come from Roche's analysis of one body in another's gravitational field.

Why does a star fill its Roche lobe in the first place?

Two independent drivers. The first is stellar evolution: when a star exhausts core hydrogen it swells enormously — a Sun-like star reaches tens to hundreds of solar radii as a red giant — and in a close binary that expansion reaches the lobe. The second is orbital shrinkage at fixed stellar size: gravitational-wave radiation and magnetic braking drain angular momentum, so $a$ and hence $R_L$ close down onto the star. Cataclysmic variables and low-mass X-ray binaries are kept in contact mainly by the second mechanism; giant-donor systems by the first.

What actually decides whether mass transfer is stable or runs away?

Compare two logarithmic slopes as the donor loses mass: $\zeta_{\mathrm{ad}}$ , how fast the donor's own radius shrinks (its adiabatic mass-radius response), and $\zeta_L$ , how fast its Roche lobe shrinks. If the star pulls inside a lobe retreating more slowly ( $\zeta_{\mathrm{ad}}>\zeta_L$ ), the overfill decreases and transfer self-limits. If the lobe collapses faster ( $\zeta_{\mathrm{ad}}<\zeta_L$ ), the overfill deepens and transfer runs away on a dynamical or thermal timescale. Since $\zeta_L$ depends on the mass ratio and $\zeta_{\mathrm{ad}}$ on internal structure, both the masses and the donor's envelope type (radiative or convective) matter.

Where does the accretion energy come from, and why is a neutron star so much brighter?

It is gravitational potential energy released as gas falls into the accretor's well: $\dot M$ landing on mass $M$ , radius $R$ liberates $L_{\mathrm{acc}}=GM\dot M/R$ , with efficiency $\eta=GM/(Rc^2)$ — the well depth in units of $c^2$ . A white dwarf ( $R\sim7000\ \mathrm{km}$ ) gives $\eta\sim10^{-4}$ ; a neutron star ( $R\sim10\ \mathrm{km}$ ) gives $\eta\sim0.2$ ; a black hole up to $\sim0.4$ . So the same rate yields a thousand times more light on a neutron star — which is why compactness, not the transfer rate, separates a faint nova system from a luminous X-ray binary.

What is the Algol paradox and how does RLOF resolve it?

In Algol-type binaries the less massive star is the more evolved one — a subgiant — while the more massive star still sits on the main sequence. Single-star evolution says the heavier star should age first, so this looks backwards. RLOF resolves it: the originally more massive star evolved first, swelled, and filled its lobe. As the heavier donor it transferred mass fast, shrinking the orbit and stripping its envelope onto the companion, until the mass ratio reversed and transfer stabilised. What remains is a stripped, over-luminous, low-mass subgiant beside a rejuvenated, now-more-massive companion — exactly the Algol configuration.

Can accretion exceed the Eddington limit?

In steady, spherical, fully ionised accretion, no — above $L_{\mathrm{Edd}}$ radiation pressure reverses the inflow. But real systems break the assumptions. If the flow is confined to a disc or funnelled by a strong magnetic field into small polar caps, radiation escapes sideways while gas keeps falling in elsewhere, allowing local rates far above the naive limit. Ultraluminous X-ray sources shine tens to hundreds of times above the neutron-star $L_{\mathrm{Edd}}$ and are now read as magnetically channelled, beamed super-Eddington accretors. So $L_{\mathrm{Edd}}$ is a robust ceiling for isotropic emission, not for anisotropic geometries.