Changing-Look AGN: When a Supermassive Black Hole's Broad Lines Vanish in Months

What a changing-look AGN actually is

Active galactic nuclei are traditionally sorted into two optical classes. Type 1 AGN show broad permitted emission lines (velocity widths of thousands of km/s) on top of narrow forbidden lines; Type 2 AGN show only the narrow lines. The unified model of the 1980s explained this purely by orientation: a dusty, doughnut-shaped torus hides the fast-moving broad-line region (BLR) in Type 2s, while Type 1s let us look straight in.

A changing-look AGN breaks that tidy picture. The same object is seen to switch class, meaning its broad lines physically fade away or ignite while we watch. Because orientation cannot change in a human lifetime, the transition must be intrinsic — the ionizing engine itself is dimming or brightening. Intermediate classes (Type 1.8, 1.9) mark the in-between states where broad Hα survives but broad Hβ has already weakened.

• Broad lines = gas within light-days to light-weeks of the black hole, photoionized by the disk's UV continuum.
• When the continuum drops, there are no longer enough ionizing photons to sustain the broad lines.

The mechanism: an accretion-rate flip, not a moving cloud of dust

The favored driver is a real change in the mass accretion rate Ṁ onto the black hole. The disk's ultraviolet luminosity photoionizes the BLR; the broad-line flux tracks the ionizing continuum with a light-travel delay (this is the basis of reverberation mapping). If the disk luminosity plunges, the BLR is starved of ionizing photons and the broad lines fade within roughly the recombination time of the gas, which can be months.

Physically this looks like an accretion state transition, analogous to the hard-to-soft transitions seen in stellar-mass black-hole X-ray binaries. As the Eddington ratio λ = L/L_Edd drops below roughly 0.01, the inner disk is thought to switch from a radiatively efficient thin disk to a radiatively inefficient (ADAF-like) flow, collapsing the UV continuum.

The catch is timing. In a standard thin disk, luminosity responds on the viscous inflow time:

• t_visc ≈ (R²/ν) ≈ (α Ω)⁻¹ (H/R)⁻² — with α ~ 0.1 and H/R ~ 0.01 this gives ~10⁴–10⁵ years at the relevant radii.

Observed CL transitions take months to years — 100 to 100,000 times too fast. Proposed fixes include disk thermal/radiation-pressure instabilities, magnetically elevated disks, and heating/cooling fronts that propagate faster than the viscous inflow.

Characteristic numbers and a worked example

Take the classic case of Mrk 590 (redshift z ≈ 0.026, black hole mass M_BH ≈ 5×10⁷ M_☉). In the 1990s it was a luminous Type 1. Between about 2006 and 2013 its continuum luminosity fell by a factor of ~100, and the broad Hβ line essentially vanished, leaving a Type 1.9/2 spectrum. Sometime between 2013 and 2017 it began reappearing — a full round trip within roughly a decade.

We can sanity-check the light-crossing scale. For M_BH ≈ 5×10⁷ M_☉, the gravitational radius is R_g = GM/c² ≈ 7×10¹⁰ m ≈ 0.5 AU. The BLR sits at hundreds of light-days:

• Eddington luminosity: L_Edd ≈ 1.3×10³⁸ (M/M_☉) erg/s ≈ 6.5×10⁴⁵ erg/s.
• BLR radius from the R–L relation: R_BLR ~ light-days to ~light-months for L ~ 10⁴⁴ erg/s.
• Recombination time in the BLR: t_rec ≈ 1/(n_e α_B) ~ weeks–months for n_e ~ 10⁹–10¹⁰ cm⁻³.

So once the continuum drops, the broad lines can genuinely switch off within months — the observed behavior — even though rebuilding the disk luminosity that caused the drop is the part that strains theory.

How they are found and confirmed

Changing-look AGN are discovered by repeat spectroscopy combined with time-domain photometry. A candidate flags when an archival spectrum (often from SDSS) differs dramatically from a new one — a broad line present in one epoch and gone in another. Confirmation requires ruling out the boring alternative that a passing dust cloud simply obscured a still-active BLR.

• Continuum test: in true CL events the accretion continuum itself dims/brightens; obscuration would redden the light without killing the intrinsic engine.
• X-rays: CL transitions show low absorbing column (not Compton-thick), and the X-ray flux drops in step — evidence the central engine faded rather than being hidden.
• Mid-infrared echo: WISE monitoring shows the dusty torus re-radiates the changed UV with a delay, confirming an intrinsic luminosity change.

Time-domain surveys — SDSS, Pan-STARRS, the Zwicky Transient Facility, and now Rubin/LSST — have turned CL-AGN from a handful of curiosities into a population of hundreds, including 'turn-on' and 'turn-off' events found in systematic difference-spectrum searches.

How CL-AGN differ from their close cousins

Several phenomena can mimic a vanishing broad line, and separating them is the whole game:

• Obscuration (dust) variability: a torus cloud crossing our line of sight can hide the BLR without changing Ṁ. This is a real effect but produces reddening and no drop in intrinsic X-ray/UV power — the opposite of a true CL event.
• Tidal disruption events (TDEs): a star shredded by the black hole causes a bright flare with broad lines, but TDEs are one-off, decay on months, and occur in otherwise quiescent nuclei rather than persistently accreting AGN.
• Ordinary AGN 'flickering': all AGN vary at the ~10–30% level; CL-AGN are the extreme tail where variation is large enough to change the spectral classification.
• Supermassive-binary models: some authors propose a close binary black hole modulates the accretion, an alternative to the pure state-transition picture.

The consensus is that most confirmed CL-AGN are genuine intrinsic accretion-state changes, but the fraction attributable to each channel is still debated.

Significance, famous cases, and open questions

Changing-look AGN matter because they are a rare chance to watch a single black hole move between accretion states on human timescales — a natural laboratory for disk physics that otherwise evolves over millennia. They also stress-test the unified model: if type can change intrinsically, then a snapshot spectral class is not a permanent identity.

Landmark objects include:

• NGC 3516 and NGC 7603 — the earliest recognized cases (Andrillat 1968; Tohline & Osterbrock 1976).
• Fairall 9 — an early well-monitored transition (Kollatschny & Fricke 1985).
• Mrk 590 — the textbook 'faded and returned' Seyfert.
• NGC 2617 and NGC 1566 — dramatic 'turn-on' brightenings with multi-wavelength coverage.
• SDSS J1011+5442 — a genuine changing-look quasar, showing the effect scales to high luminosity.

Open questions: What exactly makes the disk change state so fast? Is it disk instability, magnetic elevation, or something else? Why do only some AGN do it, and can Rubin/LSST catch a transition in real time to pin down the trigger?

Ordinary AGN spectral types versus a changing-look AGN, and what distinguishes CL-AGN from other 'disappearing broad line' explanations

Property	Static Type 1 AGN	Static Type 2 AGN	Changing-Look AGN
Broad Balmer lines (Hβ, Hα)	Present, FWHM ~1000–10,000 km/s	Absent (only narrow lines)	Appear/vanish over months–years
Standard explanation	Face-on / unobscured	Dusty torus obscures BLR (orientation)	Intrinsic accretion-rate change
UV/optical continuum	Bright blue continuum	Weak / reddened	Rises or falls by factor 5–100
X-ray behavior	Steady, unabsorbed	Often Compton-thick absorbed	Correlated dimming/brightening, low column
Typical Eddington ratio	~0.01–1	Varies	Crosses ~0.01 during transition
Timescale to change	Stable for decades	Stable for decades	As short as a few months