Classical Nova: A White Dwarf's Stolen Hydrogen Ignites

The binary engine: hydrogen stolen through the L1 point

A classical nova is never a single star. It is a cataclysmic variable: a white dwarf of typically 0.6–1.3 M☉ orbiting a companion so closely that a typical short-period system would fit inside the Sun. Most donors are K or M dwarfs on orbits of 1.4 to 10 hours; a few are subgiants or red giants on orbits of days to years, which is why RS Ophiuchi and T Coronae Borealis behave so unlike the rest.

The donor has been squeezed until it fills its Roche lobe — the teardrop equipotential, in the frame co-rotating with the binary, beyond which gas no longer belongs to it. At its tip sits the inner Lagrange point L1, a saddle in the combined gravitational and centrifugal potential. Gas spills through L1 at roughly the sound speed of the donor's atmosphere, ~10 km s⁻¹, and accelerates as it falls, but it carries the donor's orbital angular momentum, so it cannot fall straight in: it swings past the dwarf, wraps around, collides with itself and settles into an accretion disc, lighting a bright spot where the stream strikes the rim. Turbulence from the magnetorotational instability then drags angular momentum outward and lets matter creep inward.

Nothing about the nova drives this transfer; the binary bleeds orbital angular momentum on its own — magnetic braking above the 2–3 hour period gap, gravitational-wave radiation below it — giving rates of only 10⁻¹¹ to 10⁻⁸ M☉ yr⁻¹, typically 10⁻⁹. The gas lands on a body about 7,000 km in radius whose surface gravity is 10⁵–10⁶ times Earth's, releasing ~10¹⁷ erg per gram simply by landing — enough to make the disc glow, and trivial next to what the hydrogen is about to do.

Why degeneracy turns a fire into a bomb

Gravity compacts the delivered hydrogen into a shell only tens of kilometres deep — a fraction of a percent of the star's radius — reaching a column density of 10⁹–10¹⁰ g cm⁻² at a density of 10³–10⁴ g cm⁻³. There the electrons become degenerate: the Pauli exclusion principle, not thermal motion, supplies nearly all the pressure, and P ≈ 10¹³ (ρ/μₑ)^(5/3) dyn cm⁻² depends on density alone.

That is the whole physics of a nova. In an ordinary star the ideal-gas law ties pressure to temperature, so a shell that burns too fast expands, cools and throttles itself. A degenerate shell has no such thermostat: add heat and the pressure barely responds, the layer does not lift, and the temperature simply climbs.

At ~2×10⁷ K the CNO cycle switches on, and its bottleneck reaction ¹⁴N(p,γ)¹⁵O makes the energy generation rate scale roughly as ε ∝ T¹⁸ there: a 10% rise in temperature multiplies the burning by about 5.6. More heat, faster burning, more heat — a thermonuclear runaway.

The trigger is a pressure, not a mass. The base must reach P_crit ≈ 10¹⁹–10²⁰ dyn cm⁻², and for a thin shell P ≈ G M_wd ΔM / 4πR_wd⁴, so the fuel needed is ΔM ≈ 4πR⁴P_crit / GM — a few times 10⁻⁵ M☉ on a 1 M☉ dwarf, up to 10⁻⁴ M☉ on a cool low-mass one, and as little as 10⁻⁷–10⁻⁶ M☉ on a 1.35 M☉ dwarf whose small radius collapses the R⁴ term. That single scaling explains the entire recurrence spectrum: 10⁴–10⁵ years to reload at 10⁻⁹ M☉ yr⁻¹, but under a year for a near-Chandrasekhar dwarf gorging at 10⁻⁷.

Two things end the runaway together. Near 10⁸ K the thermal energy kT catches the electron Fermi energy (~8 keV at these densities), degeneracy lifts, and the layer expands at last. Meanwhile the CNO cycle saturates: proton captures outrun the β⁺ decays of ¹⁴O (70.6 s), ¹⁵O (122 s), ¹³N (~10 min) and ¹⁷F (64 s), so burning becomes beta-limited and freezes at ε ≈ 6×10¹⁵ Z_CNO erg g⁻¹ s⁻¹, nearly independent of temperature. Convection, turning over in ~100 s, carries those radioactive nuclei to the surface so their decay energy lands where it can drive an ejection. Peak base temperatures reach 1–4×10⁸ K.

One ingredient still resists calculation: the fuel's composition. Ejecta are wildly enriched — carbon, nitrogen and oxygen can be 20–50% of the ejected mass, and oxygen–neon dwarfs above ~1.1 M☉ add conspicuous neon, magnesium and aluminium — yet the accreted gas was ordinary hydrogen. The metals must be dredged from the dwarf's own core by shear or by Kelvin–Helmholtz instabilities at the convective boundary; Starrfield, Truran, Sparks and Kutter showed in 1972 that without them the runaway is too feeble to eject anything.

From ignition to naked-eye star, and back again

The eruption liberates ~10⁴⁵–10⁴⁶ erg. Within a day or two the envelope, no longer degenerate, swells from a 7,000 km dwarf to red-giant dimensions of 10¹¹–10¹³ cm; the photosphere cools to roughly 8,000 K while the luminosity climbs to and often past the Eddington limit, L_Edd = 4πGMc/κ ≈ 1.5×10³⁸ erg s⁻¹ (~4×10⁴ L☉) for a solar-mass dwarf. The optical brightness rockets 8–15 magnitudes to an absolute visual magnitude near −7 to −9; V1500 Cygni climbed about 19 magnitudes in roughly two days in 1975.

Then ~10⁻⁵–10⁻⁴ M☉ is unbound at 500–5,000 km s⁻¹ — up to ~10,000 km s⁻¹ in U Scorpii — carrying 10⁴⁴–10⁴⁵ erg of kinetic energy. The ejection is staged: a slow, dense, roughly equatorial outflow leaves first and is overtaken by a fast radiation-driven wind, and where the two collide internal shocks form. Those shocks are why novae turned out to be gamma-ray sources. Fermi-LAT found GeV emission first in V407 Cygni (2010) and in more than a dozen novae since; in 2021 H.E.S.S. and MAGIC caught TeV photons from RS Ophiuchi, proving the shocks accelerate particles past 10¹² eV. In V906 Carinae (2018) the BRITE nanosatellite watched optical flares track the Fermi flares exactly — evidence that shocks, not just the fireball, power much of the light.

Roughly a third of novae then condense dust 30–100 days after maximum, cutting the optical light with a deep dip while the infrared rises; DQ Herculis (1934) is the archetype. Underneath, the bolometric luminosity holds near Eddington as leftover hydrogen keeps burning on the surface. As the ejecta thin and the photosphere recedes, that luminosity emerges from a smaller, hotter surface, so the spectrum hardens through the ultraviolet into soft X-rays and the system becomes a supersoft X-ray source at kT ≈ 20–100 eV — a phase Swift's X-ray Telescope has tracked in dozens of novae. Turn-off comes when the residual hydrogen runs out: two weeks for M31N 2008-12a, over a decade for V723 Cassiopeiae. Then the disc rebuilds and the clock restarts.

How novae are actually measured

The workhorse observable is the light curve. Novae are graded by t₂ and t₃ — the days taken to fade 2 and 3 magnitudes below maximum — and sorted into speed classes from very fast (t₂ under 10 days) to very slow (t₂ over 150 days). U Scorpii drops 2 magnitudes in barely a day; HR Delphini took nearly half a year. Strope, Schaefer and Henden catalogued 93 well-observed curves in 2010 and sorted them into seven recurring shapes, from smooth and plateau to dust dip and jitter. Because faster novae are also intrinsically brighter, the maximum magnitude–rate of decline (MMRD) relation has long been used to turn t₂ into a distance, alongside the sturdier rule that novae converge near absolute visual magnitude −5.5 about 15 days after peak.

Firmer numbers come from elsewhere. VLA radio interferometry follows the shell's thermal free-free emission and gives the most model-independent ejecta masses available. The Hubble Space Telescope has resolved shells directly — the clumpy ring of DQ Herculis, the thousands of knots around T Pyxidis — and comparing a shell's angular expansion with its Doppler velocity yields an expansion parallax distance, while Gaia supplies trigonometric parallaxes outright. Swift, XMM-Newton and Chandra cover the supersoft phase, NuSTAR the shock-heated hard X-rays, Fermi-LAT everything above 100 MeV. Spectra split novae into the slow, narrow-lined Fe II class and the fast, broad-lined He/N class, a distinction Robert Williams drew in 1992 that maps onto the two-stage ejection.

Finding them is its own problem. The Galactic rate is roughly 25–50 eruptions per year, yet only about ten are discovered annually, because dust in the plane hides the rest — which is why Palomar Gattini-IR surveys in the near-infrared J band, alongside ASAS-SN, ZTF, OGLE, Gaia Alerts and a large contingent of amateurs. In M31, which can be surveyed whole, about 65 novae erupt each year.

Named cases and a four-century paper trail

The word is a historical accident, taken from Tycho Brahe's 1573 pamphlet De Nova Stella about the “new star” of 1572 — which was in fact a supernova. Edwin Hubble's 1929 survey of novae in M31 implied that one of them, S Andromedae of 1885, had been absurdly overluminous; in 1934 Walter Baade and Fritz Zwicky resolved the paradox by coining “super-novae” for a physically distinct class. The binary nature of ordinary novae waited until 1954, when Merle Walker found that Nova Herculis 1934 (DQ Herculis) is an eclipsing binary with a 4 hour 39 minute period.

• V603 Aquilae (1918) reached V ≈ −1.1 — the brightest nova of the twentieth century.
• GK Persei (1901) produced the first astronomical light echo ever photographed, by George Ritchey; it is now a magnetic intermediate polar sitting inside a still-expanding shell.
• RS Ophiuchi carries a ~1.35 M☉ dwarf fed by a red giant's wind and erupted in 1898, 1933, 1958, 1967, 1985, 2006 and 2021 — roughly every 15 years, and the source of that first TeV detection.
• T Pyxidis went off in 1890, 1902, 1920, 1944, 1967 and 2011, its ejecta sculpted into thousands of discrete knots.
• T Coronae Borealis, the “Blaze Star” some 800–900 pc away, erupted in 1866 and 1946 and has been watched closely since a 2023 dip echoed the one preceding 1946; it should reach roughly second magnitude.
• M31N 2008-12a erupts nearly every year on a ~1.38 M☉ dwarf accreting at ~10⁻⁷ M☉ yr⁻¹, inside a nova “super-remnant” some 130 parsecs across built from centuries of ejecta.
• Nova Scorpii AD 1437, recorded by Korean court astronomers, was tied by Michael Shara and colleagues in 2017 to a system that behaves as a dwarf nova today — support for the idea that post-novae hibernate.

What a classical nova is not

The comparison that matters most is with the Type Ia supernova, and the two are not variations on a theme. A nova burns a thin, borrowed skin of hydrogen and releases ~10⁴⁵ erg; a Type Ia detonates the carbon and oxygen of the entire dwarf, unbinds it, releases ~10⁵¹ erg and forges ~0.6 M☉ of radioactive ⁵⁶Ni. A nova peaks near absolute magnitude −8, a Type Ia near −19.3, some 10⁵ times brighter. After a nova there is still a white dwarf, a disc and a companion; after a Type Ia, no star at all.

Nor is a classical nova a dwarf nova outburst, despite the shared word. Systems like SS Cygni and U Geminorum brighten 2–8 magnitudes every few weeks because their accretion discs flip between a cool neutral state and a hot ionised one — a thermal-viscous instability that never touches the dwarf's surface and involves no nuclear burning at all. The energy scale differs by ~10⁵, and dwarf novae eject essentially nothing.

Two further impostors are worth naming. A luminous red nova is not a nova but the merger of two ordinary stars, as V1309 Scorpii proved when pre-eruption photometry caught its orbit shrinking to zero. A micronova, identified in TESS data in 2022, is genuinely nuclear but geographically confined: on strongly magnetic dwarfs the accreted gas is funnelled onto a small polar cap that ignites locally, releasing perhaps a millionth of a nova's energy. The closest true relative of a classical nova is none of these but the Type I X-ray burst — the same degenerate flash in the accreted layer on a neutron star's surface, where far stronger gravity makes it recur in hours and radiate in X-rays.

Open questions

Novae are well understood in outline and stubbornly unresolved in detail.

• Too much ejecta. Radio and infrared measurements routinely return ejected masses a factor of several — sometimes ten — above what hydrodynamic models predict for the same system.
• How the metals get mixed in. Every successful model needs the accreted layer seeded with core material, but whether that comes from shear, convective entrainment or breaking gravity waves is still argued — and the answer sets the eruption's strength.
• Does the dwarf grow or shrink? If a nova expels more mass than it accreted, dredging up core material as it goes, the dwarf erodes. If it expels less, the dwarf creeps toward the Chandrasekhar limit of 1.4 M☉ and recurrent novae become plausible single-degenerate Type Ia progenitors. Systems like M31N 2008-12a, at ~1.38 M☉, are the crux of the debate.
• Lithium. The detection of ⁷Be in the ejecta of V339 Delphini in 2015, and of ⁷Li in V1369 Centauri, suggests novae may dominate the Galaxy's ⁷Li budget.
• Missing gamma-ray lines. Models predict 478 keV photons from ⁷Be decay and 1275 keV photons from ²²Na, yet INTEGRAL and its predecessors have only ever set upper limits.
• How reliable is the MMRD? Faint-but-fast novae found in M31 by Mansi Kasliwal and colleagues in 2011 fall well off the relation, casting doubt on its use as a distance indicator.
• Hibernation. Do post-novae fade over centuries into detached, quiescent systems before the next eruption? Nova shells around present-day dwarf novae hint that they do.

Classical novae set against the eruptions they are most often confused with

Event	What actually burns or brightens	Peak brightness	Fate of the compact star
Classical nova	Runaway CNO burning of an accreted hydrogen shell (~10⁴⁵ erg)	Rise of 8–15 mag; peak absolute magnitude −7 to −9	Survives; reloads over ~10⁴–10⁵ yr
Recurrent nova	Identical mechanism on a massive (>1.2 M☉) dwarf fed at a high rate	Rise of ~7–10 mag; peak absolute magnitude −6 to −8	Survives; repeats within 1–100 yr
Dwarf nova outburst	Thermal-viscous instability in the accretion disc — no nuclear burning at all	Rise of 2–8 mag; peak absolute magnitude +3 to +6	Untouched; repeats every ~10 days to years
Micronova	Burning confined to one magnetic accretion pole (~10⁻⁶ of a nova's energy)	A few tenths of a magnitude, lasting hours	Survives; found with TESS in TV Col and EI UMa
Type Ia supernova	Carbon–oxygen detonation through the entire white dwarf (~10⁵¹ erg)	Peak absolute magnitude ≈ −19.3; ~10⁹–10¹⁰ L☉	Completely unbound; no star is left behind
Type I X-ray burst	The same degenerate flash staged on a neutron star instead	~10³⁹ erg over 10–100 s, visible only in X-rays	Survives; repeats in hours