Quasi-Periodic Eruptions: A Star Punching Through a Black Hole's Disk

The signature: a nucleus that flares on a clock

The defining feature of a Quasi-Periodic Eruption is regularity. In the archetype GSN 069 — a low-mass Seyfert nucleus reported by Miniutti and collaborators in Nature in 2019 — the soft X-ray flux jumps by nearly two orders of magnitude, peaks over roughly an hour, and returns to a low quiescent level, only to do it all again about nine hours later. Between bursts the source sits quietly; during a burst the count rate can rise ~100-fold in the 0.4–2 keV band. Crucially, the eruptions are soft: the emission is well described by a blackbody whose temperature climbs from about 50 eV in quiescence to ~100–250 eV (a few million kelvin) at peak, then cools back down. The spectrum heats up as the burst brightens and softens as it fades — a fingerprint of an expanding, cooling gas cloud rather than a change in the central engine's brightness.

A handful of confirmed and candidate sources now anchor the class: RX J1301.9+2747 (recognised as a QPE by Giustini et al. 2020, with recurrences of a few hours), and eRO-QPE1 and eRO-QPE2 (Arcodia et al. 2021), discovered blind by the eROSITA all-sky survey with recurrence times of ~18.5 hours and ~2.4 hours respectively. A striking detail is that several sources show an alternating long–short–long–short spacing of eruptions rather than a single fixed period. Any successful model has to reproduce not just the repetition but this two-timescale beat.

The leading model: an orbiter punching through the disk

The picture that best fits the data is almost mechanical. Around the supermassive black hole sits a thin accretion disk. On a tight, mildly eccentric orbit — and, importantly, inclined to the disk plane — travels a small, dense companion: most likely a stripped stellar core, a low-mass star, or a white dwarf. Because the orbit is tilted, the companion pierces the disk not once but twice per orbit, on the way up through the midplane and again on the way back down. Each impact drives a shock into the disk gas, ejects hot debris above and below the plane, and that debris radiates the soft X-ray flare we see. Two crossings per orbit is exactly why QPEs come in pairs, and an eccentric, inclined geometry naturally makes the two crossings unequally spaced in time — the long–short recurrence pattern falls straight out of the geometry.

This is the orbiter–disk collision model, developed in detail by Linial & Metzger (2023), Franchini et al. (2023), Tagawa & Haiman (2023) and others. It reproduces the observed burst durations, luminosities, temperatures, and the alternating timing with remarkably few free parameters. Competing ideas exist — radiation-pressure or magnetically driven limit-cycle instabilities in the inner disk, or the tearing and precession of a warped disk — but these struggle to explain the clockwork stability and the paired timing as cleanly. The collision model has one more virtue: it makes concrete, testable predictions about how the eruptions should drift as the orbit slowly shrinks.

The physics of a single crossing

The numbers explain why the flares are fast, hot, and soft. To orbit a black hole of ~10⁵–10⁶ solar masses with a period of only hours, the companion must sit close in — at roughly a few tens to a couple hundred gravitational radii (r_g = GM/c²). At ~100 r_g the Keplerian orbital speed is about c/√100 ≈ 0.1c, or ~30,000 km/s. When a body moving that fast plows through disk gas, the collision shock-heats the swept-up material to well above 10⁷ K.

But that ferociously hot shock front is not what we observe directly. The shocked gas is unbound and blasts outward as a small cloud, expanding and cooling as it goes. Its photosphere — the surface where the expanding debris finally becomes transparent, a few times 10¹²–10¹³ cm across — radiates a near-blackbody spectrum at ~10⁶ K, i.e. a few tens to ~100+ eV, squarely in the soft X-ray band. The burst rises as the cloud grows and heats, then fades as it thins and cools, which is precisely the observed hard-then-soft evolution. Each eruption radiates on the order of ~10⁴⁵–10⁴⁶ erg — enormous by everyday standards, yet a small fraction of the energy budget available in the orbit and the disk.

Where the disk and the orbiter come from

A natural question is: why is there a disk to hit at all? Many QPE hosts are otherwise unremarkable, quiescent galaxies, not brilliant quasars. The compelling answer, in at least some cases, is a prior tidal disruption event. In 2024, Nicholl and collaborators reported QPEs emerging in AT2019qiz, a galaxy where a star had been tidally shredded a few years earlier — the debris of that disruption settled into a compact accretion disk, and only then did the eruptions switch on, with a recurrence of roughly two days. That is a direct, causal link: build a disk with a TDE, and a pre-existing orbiter starts drumming on it. GSN 069 itself brightened in the years before its QPEs, consistent with a TDE-fed disk that later faded — and, tellingly, GSN 069's eruptions weakened and vanished as the underlying disk dimmed, then partially recovered.

The identity of the orbiter is still debated. To survive so close to the black hole without being torn apart, it should be compact and dense — favouring a white dwarf or the stripped, degenerate core of a star whose envelope was already lost. A humble main-sequence star can also survive around the lightest black holes. Recent extreme sources such as Ansky (in the galaxy SDSS1335+0728), which produces the most energetic and longest-recurrence eruptions yet seen, are pushing these models and hinting that the QPE population is more diverse than the first handful suggested.

How we catch them, and how we know they are not something else

QPEs live almost entirely in the soft X-ray band, so they are found and followed with X-ray telescopes: XMM-Newton and Chandra for deep spectroscopy, NICER on the ISS for dense timing, and eROSITA (on the Spektr-RG spacecraft), whose repeated all-sky scans let it catch sources flaring in and out and enabled the first blind QPE discoveries. The observational recipe is demanding: you need repeated visits spaced by hours to reveal that a nucleus is erupting on a clock rather than flickering at random.

Distinguishing QPEs from their cousins comes down to timing and spectrum. A tidal disruption event is a one-off: a giant rise and a slow, non-repeating decay. A changing-look AGN shifts its whole character over months to years and does not reset. Ordinary X-ray quasi-periodic oscillations are smooth, small-amplitude modulations — a few to ~15 percent — whether on the millisecond-to-second periods of stellar-mass black holes or the roughly hour-long period of the AGN QPO in RE J1034+396; they are not the factor-of-100 flares QPEs show. Thermal-viscous limit-cycle outbursts can repeat, but with quasi-regular, not paired, timing and different spectral behaviour. The QPE signature — large-amplitude, soft, blackbody bursts that recur every few hours with a long–short beat and heat-then-cool spectra — is specific enough to stand on its own.

Gravitational waves, LISA, and the open questions

If the orbiter model holds, QPEs are the electromagnetic face of an extreme-mass-ratio inspiral (EMRI): a stellar-scale object spiralling around a supermassive black hole, with a mass ratio of ~10⁻⁴ to 10⁻⁶. Such a system radiates continuous gravitational waves as it orbits, at twice the orbital frequency — for a ~9-hour orbit that is ~0.05–0.1 millihertz, near the low-frequency end of the LISA (Laser Interferometer Space Antenna) band. LISA, an ESA-led mission targeting launch in the mid-2030s, is built precisely for millihertz gravitational waves. Because several QPE hosts are nearby (tens to ~150 Mpc), a few may be loud enough for LISA to hear directly, making them prime candidates for multi-messenger study: X-ray eruptions we can time now, gravitational waves we may detect later, from the very same orbit.

Plenty remains unsettled. What exactly is the orbiter — white dwarf, stripped core, or ordinary star? Is the disk always TDE debris, or sometimes a pre-existing AGN disk? Why do eruptions sometimes disappear and return? How fast is each orbit shrinking under gravitational-wave emission and disk drag, and how long does a QPE phase last? And are the slower, weeks-long “quasi-periodic outflows” seen in sources like Swift J0230 the same phenomenon scaled up, or a separate family? With eROSITA and dedicated X-ray monitoring turning up new sources, and LISA on the horizon, QPEs have gone from a curiosity to one of the sharpest probes we have of the crowded, violent space just outside a supermassive black hole.

Quasi-Periodic Eruptions versus their look-alikes: other ways a black hole can flicker, flare, or change its face.

Phenomenon	Timescale	Amplitude	Driving mechanism	Example
Quasi-Periodic Eruption (QPE)	Recurs every ~2 h–few days	~10–100× in soft X-rays	Compact orbiter crossing the disk twice per orbit (EMRI)	GSN 069, eRO-QPE2
Tidal Disruption Event (TDE)	One-off; months–years decay	~10³–10⁶× rise, then fades	A whole star torn apart and accreted	ASASSN-14li, AT2019qiz
Changing-look AGN	Months–years, non-repeating	Factor ~5–100 in continuum	Intrinsic change in accretion rate	Mrk 590, NGC 2617
X-ray quasi-periodic oscillation (QPO)	~ms–s (stellar); ~1 h (AGN)	Few-percent modulation	Oscillation of the innermost accretion flow	RE J1034+396, GRS 1915+105
Thermal-viscous / limit-cycle outburst	Weeks–years	~10–100×	Radiation-pressure or ionisation disk instability	GRS 1915+105 'heartbeat'