Dwarf Nova Outburst: An Accretion Disc That Flips to Hot

The system: a white dwarf, a stream, and a disc that keeps changing its mind

A dwarf nova is a cataclysmic variable: a white dwarf of roughly 0.6-1.0 solar masses in a very tight orbit with a low-mass, usually K or M dwarf companion. Orbital periods run from about 78 minutes (the observed CV period minimum) to a few hours, with a conspicuous scarcity between 2 and 3 hours known as the period gap: SS Cygni orbits in 6.6 hours, U Geminorum in 4.25, WZ Sagittae in 81.6 minutes.

Those periods imply cramped geometry. For SS Cygni, Kepler's third law gives a separation of about 1.4 x 10^11 cm, roughly two solar radii, and the disc reaches about 0.3 of that, some 4 x 10^10 cm. The white dwarf at the centre is Earth-sized, ~7 x 10^8 cm in radius. A disc half the width of the Sun is feeding a star the size of a planet.

The donor fills its Roche lobe and leaks gas through the inner Lagrange point at 10^15 to 10^17 g/s, or 10^-11 to 10^-9 solar masses per year. That stream carries too much angular momentum to fall straight in, so it splashes onto the disc rim and creates a bright spot which, in quiescence, can supply a third of the optical light. The key point is that the rate at which the disc is fed and the rate at which it drains do not have to match. In a dwarf nova, they never do.

The S-curve: why partially ionised hydrogen has no stable middle

Plot a ring of the disc in the plane of surface density (Sigma) against effective temperature and ask where it can sit in thermal equilibrium, with local heating from viscous dissipation exactly balancing radiative cooling. The answer is not a line but an S. That shape is the entire mechanism.

On the lower branch the gas is near 3,000-5,000 K and hydrogen is neutral. Free electrons are scarce, the H-minus opacity is small, and the disc radiates heat away efficiently. On the upper branch the gas is above ~10^4 K, hydrogen is fully ionised, opacity has settled again, and the ring is stable once more. Between them lies the trap. Around 6,500 K hydrogen part-ionises, and there the opacity climbs almost catastrophically with temperature, roughly as kappa proportional to rho^(1/2) T^9. Nudge such a ring hotter and it ionises a little more, its opacity soars, its radiation is trapped, and it heats further. There is no restoring force: the middle branch has negative slope and is thermally unstable.

So an annulus has only two allowed states, and the turning points define two critical surface densities. Using the standard fits (Lasota 2001), Sigma_max is about 11.4 R10^1.05 M1^-0.35 alpha_cold^-0.86 g/cm^2 and Sigma_min about 8.25 R10^1.05 M1^-0.35 alpha_hot^-0.80 g/cm^2, with R10 the radius in units of 10^10 cm. For a 0.8 solar mass white dwarf at 10^10 cm with alpha_cold = 0.01 and alpha_hot = 0.1, that is roughly 650 and 55 g/cm^2. The disc must build up about an order of magnitude in surface density before it can fire, and can only empty back down to the lower threshold.

The limit cycle, in numbers

The viscosity in Shakura and Sunyaev's 1973 prescription is nu = alpha c_s H, and fitting real outburst shapes requires alpha_hot / alpha_cold of about 4 to 10. Because the sound speed rises with the temperature jump as well, nu, and with it the inflow speed, climbs by one to two orders of magnitude when a ring flips: gas that took months to drift inward now takes days.

Follow one cycle. In quiescence the disc drains far more slowly than it is fed, so Sigma grows everywhere. Where Sigma first crosses Sigma_max a ring ionises; its neighbours are heated by radiation and radial diffusion and ionise in turn, so a heating front propagates at a speed of order alpha c_s, a few km/s, crossing the disc in hours to a day. Behind it the disc is hot, viscous and draining, and the accretion rate onto the white dwarf leaps from below ~10^15 g/s to ~10^18 g/s. That is the outburst.

The energy budget is easy to check. Each gram falling to the surface of a 0.8 solar mass, 7 x 10^8 cm white dwarf releases GM/R = 1.5 x 10^17 erg, half radiated by the disc and half by the boundary layer. Hydrogen fusion yields 6 x 10^18 erg/g, about forty times more per gram, and a nova burns an accreted shell far more massive than the gas one outburst drains, which is why a classical nova releases ~10^5 times more energy. Dumping the 10^23-10^24 g stored in a quiescent disc gives ~10^40 erg, and at 10^34-10^35 erg/s that takes days to a couple of weeks. The disc then falls below Sigma_min at its outer edge, a slower cooling front sweeps inward, and the cycle restarts.

One number decides whether a system does this at all. Above a critical feeding rate, roughly 10^-9 solar masses per year, even the outer edge stays hot and the disc sits permanently on the upper branch: those are the steady nova-likes such as UX UMa and RW Tri, which never outburst.

Fronts, delays and the boundary layer switch

The travelling-front picture makes sharp, testable predictions, and each has been confirmed.

• Two rise shapes. When Sigma first exceeds the threshold in the outer disc the front travels inward (outside-in) and the rise is fast, under a day; when it ignites near the centre it must climb outward against increasing Sigma, giving a slower, rounder rise (inside-out). Both are seen, often in the same star.
• The ultraviolet delay. If the optical comes mostly from the outer disc and the UV from the inner disc, an outside-in front should light the optical first. The International Ultraviolet Explorer found exactly that: the UV rise in SS Cygni and VW Hydri lags the optical by roughly half a day to a day and a half. Before the disc instability model this was a genuine puzzle.
• The X-ray switch. Half the accretion energy is dissipated in the boundary layer, where gas orbiting at nearly 4,000 km/s brakes onto a slowly spinning white dwarf. In quiescence that layer is optically thin and emits hard bremsstrahlung X-rays near 10 keV; in outburst it turns optically thick and radiates instead in the extreme ultraviolet at a few tens of eV. Coordinated RXTE and EUVE monitoring of SS Cygni (Wheatley, Mauche and Mattei 2003) caught the handover: a brief hard X-ray flare on the rise, suppression through the outburst as the EUV took over, then recovery on the decline.
• Linear decline. A cooling front moving at nearly constant speed shrinks the hot region steadily, so the decay is close to linear in magnitude. Bailey (1975) found the decline rate scales roughly as 0.4 P(hours)^0.84 days per magnitude: bigger discs take longer to empty.

How we know: 170 years of light curves, and one distance crisis

John Russell Hind found the first of these stars, U Geminorum, on 15 December 1855, mistaking it for a nova; Louisa D. Wells discovered SS Cygni at Harvard in 1896. Merle Walker's 1954 eclipse work on DQ Herculis and Robert Kraft's spectroscopy in the early 1960s established that these objects are close binaries. Yoji Osaki proposed the disc instability in 1974; Hoshi, Meyer and Meyer-Hofmeister, Smak and Cannizzo built it into the modern model, reviewed by Jean-Pierre Lasota in 2001.

Because outbursts are bright, slow and unpredictable, amateurs did most of the observing. The AAVSO light curve of SS Cygni runs continuously from 1896 and holds well over half a million measurements, the longest dense record of any variable star. Kukarkin and Parenago noticed in 1934 that amplitude and recurrence track each other, A ~ 0.7 + 1.9 log T(days); for SS Cygni's 49-day cycle that predicts 3.9 mag against ~4 observed, and for WZ Sagittae's ~10,000-day cycle 8.3 against ~8.5. Kepler later resolved the rise itself in V344 Lyrae and V1504 Cygni at one-minute cadence, while eclipse mapping of Z Chamaeleontis and OY Carinae (Horne's 1985 maximum-entropy method) recovered the radial temperature profile: close to the steady-state T proportional to R^-3/4 in outburst, much flatter in quiescence, exactly as a mass-hoarding disc requires.

The most instructive episode was a distance. In 1999 an HST Fine Guidance Sensor parallax placed SS Cygni at 166 +/- 12 pc, which made it too luminous for the model to work at all: at that distance its feeding rate exceeded the critical value and the disc should have been permanently hot. Radio detections of a transient jet (Kording et al. 2008) opened another route, and VLBA astrometry gave 114 +/- 2 pc (Miller-Jones et al., Science, 2013), later confirmed by Gaia. The model was fine; the parallax was wrong.

Superoutbursts, superhumps and the subtype zoo

U Gem stars show only normal outbursts. SU UMa stars, nearly all below the period gap, occasionally show a superoutburst: about 0.5-1 mag brighter and five to ten times longer, decorated with superhumps, photometric humps at a period a few percent longer than the orbit.

Whitehurst showed in 1988 that a disc reaching the 3:1 eccentric Lindblad resonance becomes eccentric and precesses slowly; the beat between orbital and precession periods is the superhump. The resonance sits at 3^(-2/3) times the separation, about 0.48a, which fits inside the tidal truncation radius only when the mass ratio q = M2/M1 is below about 0.25 - hence the confinement of superhumpers to short periods and low-mass donors. The fractional period excess therefore measures q, with Patterson et al. (2005) calibrating epsilon ~ 0.18q + 0.29q^2, making superhumps one of the few ways to weigh a non-eclipsing binary. Osaki's 1989 thermal-tidal instability model adds enhanced tidal torque in the eccentric state to explain the extra length.

The extremes are instructive. WZ Sagittae, 43 pc away with an 81.6-minute orbit and a very low feeding rate, erupts only every few decades - 1913, 1946, 1978, 2001 - by ~8.5 mag, and shows only superoutbursts; at the other end, ER UMa stars supercycle every 20-50 days. Z Camelopardalis stars, fed almost exactly at the critical rate, get stuck in standstills roughly 0.7 mag below maximum for weeks to years, neither erupting nor fading - a disc balancing on the knife edge of the S-curve. GK Persei ties the family together: a classical nova in 1901, it has since 1966 produced dwarf nova outbursts every ~3 years, because its unusually wide 2-day orbit gives it a disc big enough to be unstable.

Where the model strains, and what a dwarf nova is not

The model works, but its central parameter is still a fudge: nobody derives alpha_hot ~ 0.1 and alpha_cold ~ 0.02 from first principles. The magnetorotational instability (Balbus and Hawley, 1991) supplies the turbulence in the ionised state, and the cool state is too neutral for MRI to couple well, which motivates the hysteresis - but shearing-box simulations struggle to reach alpha ~ 0.1 without a net vertical field, and Hirose et al. (2014) found that convection in the cool state may supply the extra transport near the S-curve's upper knee. Whether superoutbursts come from Osaki's tidal instability or from a burst of enhanced mass transfer off the irradiated donor, as Smak argued, is still contested even with Kepler data.

Two persistent confusions are worth naming. First, a dwarf nova is not a faint classical nova. A classical nova is a thermonuclear runaway in the accreted hydrogen shell, ejecting ~10^-5-10^-4 solar masses at ~1,000 km/s and releasing ~10^45 erg once every 10^4-10^5 years. Dwarf novae eject essentially nothing, though outburst UV spectra do show wind lines such as C IV 1550 A with P Cygni profiles reaching several thousand km/s. The same binary does both: dwarf nova outbursts every few weeks while the shell accumulates, one classical nova when it ignites.

Second, the light is not the white dwarf flaring. Eclipse profiles in Z Cha and OY Car show the outburst source is extended and disc-shaped; the white dwarf itself only heats modestly, then cools for weeks afterwards, a signal HST has tracked in VW Hyi and U Gem. And the same instability, with irradiation of the outer disc added, explains black hole transients such as A0620-00 - which is why those outbursts last months and recur over decades.

Dwarf novae against the eruptions they are most often confused with. Only the first two rows are purely gravitational.

Event	Where the energy comes from	Amplitude and recurrence	Energy released
Dwarf nova outburst	Gravity: stored disc gas dumped onto the white dwarf	2-6 mag; days to decades, usually weeks to months	~10^39-10^41 erg
SU UMa superoutburst	Same, plus a 3:1 tidal resonance draining an eccentric disc	0.5-1 mag brighter than normal, 5-10x longer	~10^40-10^41 erg
Classical nova	Thermonuclear runaway in the accreted hydrogen shell	8-15 mag; ~10,000-100,000 yr	~10^45 erg
Recurrent nova	The same runaway on a massive, fast-fed white dwarf	7-9 mag; 10-100 yr (T CrB, RS Oph, U Sco)	~10^44-10^45 erg
Micronova	Localised, magnetically confined burning at one accretion pole	~1-2 mag; hours (found with TESS, 2022)	~10^38-10^39 erg
Type Ia supernova	Runaway carbon fusion that unbinds the whole white dwarf	~20 mag; once	~10^51 erg