Accretion
Soft-to-Hard State Transitions in Black-Hole X-ray Binaries
Over a few weeks, a stellar-mass black hole a dozen kilometers across can dim by a factor of a hundred and completely rearrange the geometry of the gas swirling into it — swapping a million-degree thermal glow for a hard, crackling X-ray hiss that reaches beyond 100 keV. This is the soft-to-hard state transition, the closing act of an outburst in a black-hole X-ray binary (BHXRB), when the source falls from its disk-dominated soft state back to the corona-dominated hard state.
It is one half of a hysteresis loop that every transient black hole traces on its way up and down in brightness. The transition marks a wholesale reorganization of the accretion flow — from a thin, luminous disk reaching almost to the innermost stable circular orbit, back to a geometrically thick, radiatively inefficient hot flow that launches a compact radio jet. Critically, the soft-to-hard switch happens at a much lower luminosity than the hard-to-soft switch that opened the outburst, a memory effect that remains one of the central puzzles of accretion physics.
- TypeAccretion-state spectral transition
- RegimeOccurs at ~0.3–3% of L_Edd (typically ~2%)
- First mappedRXTE era, 1996–2012; hysteresis clarified ~2004 (GX 339-4)
- TimescaleDays to a few weeks (slower than hard-to-soft rise)
- Key signaturePhoton index softens Γ~2.5 → hardens to Γ~1.7; disk fades
- Observed inGX 339-4, Cygnus X-1, MAXI J1348-630, Swift J1727.8-1613
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
What the transition is: two states and a memory effect
A black-hole X-ray binary is a stellar-mass black hole (typically 5–15 M_sun) pulling gas off a companion star. Most are transients: they spend years in quiescence, then brighten by orders of magnitude in an outburst lasting months. During an outburst the source cycles between two spectral extremes:
- Hard state — X-rays dominated by a hard power law (photon index Γ ≈ 1.7) from a hot electron corona, with a high-energy cutoff near 100–200 keV. A steady compact radio jet is present.
- Soft state — X-rays dominated by a ~1 keV thermal blackbody from an optically thick accretion disk reaching close to the innermost stable circular orbit (ISCO). The jet is quenched.
The soft-to-hard transition is the return leg: as the outburst fades, the disk-dominated soft state collapses back into the hard state. The defining feature is hysteresis — the source drops to the hard state at a far lower luminosity than it rose out of it, so the outburst traces a loop rather than a line.
The mechanism: disk truncation, a cooling corona, and a reborn jet
The leading physical picture is the truncated-disk / hot-inner-flow model. In the soft state, the thin, radiatively efficient disk (the Shakura–Sunyaev α-disk) extends inward to the ISCO at roughly 6 R_g (where R_g = GM/c²). As the mass accretion rate ṁ = Ṁ/Ṁ_Edd falls during the decay, the inner disk can no longer sustain itself: it evaporates or transitions into a geometrically thick, optically thin, two-temperature advection-dominated accretion flow (ADAF).
- The disk truncates and recedes outward to tens of R_g.
- Seed photons feeding the corona drop, so inverse-Compton cooling weakens and the electron temperature rises — the spectrum hardens (Γ steepens back from ~2.5 toward ~1.7) and the cutoff reappears near ~150 keV.
- The hot, thick flow re-establishes the poloidal magnetic geometry that relaunches the compact radio jet, whose flux scales roughly as L_radio ∝ L_X^0.7.
The unsolved part is why the switch-down luminosity differs from the switch-up. Hysteresis likely reflects that the trigger depends not just on ṁ but on the disk's history — its temperature, viscosity, and possibly the strength of large-scale magnetic flux threading the inner flow.
Characteristic numbers and a worked example
The hallmark of the transition is where it happens in Eddington units. The Eddington luminosity for a 10 M_sun black hole is
L_Edd ≈ 1.26 × 10^38 (M/M_sun) erg/s ≈ 1.3 × 10^39 erg/s.
- Hard→soft (rise): occurs high, anywhere from a few percent up to ~30% L_Edd — highly variable.
- Soft→hard (decay): occurs low and much more consistently, near 0.3–3% L_Edd (commonly quoted as ~2%). This narrow, reproducible fraction is the tell-tale of hysteresis.
Worked example (GX 339-4): take M ≈ 8 M_sun, d ≈ 8 kpc. Then L_Edd ≈ 1.0 × 10^39 erg/s. A soft-to-hard transition at ~2% L_Edd occurs at L_X ≈ 2 × 10^37 erg/s. At d = 8 kpc that corresponds to an observed 2–10 keV flux of order F = L/(4πd²) ≈ 2.6 × 10^-9 erg/cm²/s. As the source crosses down, the disk temperature falls below ~0.5 keV, the power-law tail hardens from Γ ≈ 2.5 to Γ ≈ 1.7, and radio emission from the recovering jet switches back on within days.
How it is observed: the q-shaped hardness-intensity diagram
The transition is read directly off the hardness–intensity diagram (HID), also called the 'q-diagram'. One plots X-ray hardness (the ratio of counts in a hard band to a soft band, e.g. 6–10 keV / 3–6 keV) against total intensity (count rate, a luminosity proxy) on log axes.
- Almost every transient traces the same counter-clockwise 'q': rise up the hard right branch, jump left across the top into the soft state, decay down the soft left branch, then jump back right to the hard state at low intensity.
- The horizontal gap between the top (hard→soft) and bottom (soft→hard) crossings is the hysteresis.
Instruments that mapped this include NASA's Rossi X-ray Timing Explorer (RXTE, 1995–2012), MAXI, Swift, NICER, NuSTAR, and China's Insight-HXMT. Complementary diagnostics include timing: fractional rms variability rises sharply, and characteristic low-frequency quasi-periodic oscillations (QPOs, ~0.1–10 Hz) reappear and drift down in frequency as the source re-enters the hard state. Simultaneous radio monitoring (e.g. ATCA, VLA) catches the jet switching back on.
How it differs from its cousins
The soft-to-hard transition is easy to confuse with related accretion phenomena; the distinctions matter:
- vs. the hard-to-soft transition: same two states, opposite direction. The upward transition is at high, variable luminosity and features bright, discrete ballistic jet ejections at the peak; the downward transition is at low, reproducible luminosity and re-establishes a steady compact jet. Together they form the hysteresis loop.
- vs. neutron-star XRB transitions: neutron stars show analogous but weaker hysteresis, plus a solid surface (thermonuclear bursts, boundary-layer emission) that black holes lack — a key way to tell the accretor apart.
- vs. AGN 'changing-look' events: the same disk–corona physics operates in supermassive black holes, but their viscous timescales are millions of times longer, so state changes that take weeks in a BHXRB would take millennia — making BHXRBs the fast-motion laboratory for accretion.
- vs. dwarf-nova outbursts: both are driven by the thermal–viscous disk instability, but white-dwarf accretion never reaches the relativistic, ISCO-scale regime that defines these states.
Significance, famous cases, and open questions
State transitions are the cleanest natural laboratory for how matter accretes onto black holes across all scales, from stellar-mass binaries to quasars. Because the soft state exposes the disk down to the ISCO, it is the regime used to measure black-hole spin via the continuum-fitting method; the hard state, with its truncated disk and relativistically broadened iron Kα line, probes coronal geometry.
- GX 339-4 is the textbook source, with multiple outbursts (2002–2011) that made hysteresis undeniable (Belloni, Homan, Miyamoto, Fender, Gallo, Maccarone, and collaborators mapped the q-track).
- Cygnus X-1 is a persistent source hovering near the hard-to-soft boundary, showing that transitions need not require a full outburst.
- Recent bright transients MAXI J1348-630 (2019) and Swift J1727.8-1613 (2023) gave NICER, NuSTAR and HXMT exquisite coverage of the decay.
Open questions: what sets the ~2% L_Edd trigger and the width of the hysteresis loop? Is disk truncation real or an artifact of complex reflection? What role does large-scale magnetic flux (the 'MAD' state) play in launching the jet? GRMHD simulations are only now reproducing the downward transition self-consistently.
| Property | Hard state (low/hard) | Soft state (high/soft) | Intermediate state |
|---|---|---|---|
| Dominant emission | Comptonized power law | Multicolor disk blackbody | Both, comparable |
| Photon index Γ | ~1.4–1.8 (hard) | ~2.2–2.8 (steep) | ~2.0–2.4 |
| High-energy cutoff | ~100–200 keV present | Absent / very high | Weak, rising |
| Inner disk radius | Truncated, tens of R_g | Near ISCO (~6 R_g) | Receding inward/outward |
| Inner disk kT_in | Cool/undetected (<0.5 keV) | ~0.5–1 keV | ~0.5–0.8 keV |
| Radio jet | Steady compact jet ON | Jet quenched (OFF) | Discrete ejections at peak |
Frequently asked questions
What triggers the soft-to-hard transition?
As the outburst decays, the mass accretion rate drops below a critical value near a few percent of the Eddington rate. The inner thin disk can no longer sustain radiative cooling and evaporates into a hot, geometrically thick, optically thin flow (an ADAF). This starves the corona of soft seed photons, so it heats up and the spectrum hardens, while the disk truncates and recedes. Exactly what sets the critical rate — and why it differs from the upward transition — is still debated.
Why does the transition show hysteresis?
Hysteresis means the source drops to the hard state at a much lower luminosity (~0.3–3% L_Edd) than the luminosity at which it rose into the soft state (up to ~30% L_Edd). This makes the outburst trace a loop, not a line, on the hardness-intensity diagram. The physical origin is uncertain but is thought to reflect a dependence on the disk's history and thermal/viscous state — possibly the buildup of magnetic flux — not just the instantaneous accretion rate.
What is the q-diagram or hardness-intensity diagram?
It plots X-ray hardness (ratio of hard-band to soft-band counts) against total intensity (count rate) on logarithmic axes. Black-hole transients trace a counter-clockwise 'q' shape: up the hard branch, across to the soft state, down the soft branch, then back to hard at low luminosity. The horizontal offset between the two horizontal crossings visually encodes the hysteresis of the state transitions.
What happens to the jet during the soft-to-hard transition?
In the soft state the compact radio jet is quenched (radio-quiet). As the source crosses back to the hard state, the steady compact jet switches back on, typically within days, and its radio flux correlates with the X-ray luminosity roughly as L_radio ∝ L_X^0.7. This differs from the hard-to-soft transition at outburst peak, which instead produces discrete, bright ballistic ejections.
How is the soft-to-hard transition different from the hard-to-soft one?
They are the two legs of the same hysteresis loop but in opposite directions. Hard-to-soft happens on the way up, at high and highly variable luminosity, with dramatic jet ejections. Soft-to-hard happens on the way down, at a low and remarkably consistent luminosity (~2% L_Edd), and re-establishes a steady compact jet. The reproducibility of the downward transition is one of its most striking features.
Which sources best show this transition?
GX 339-4 is the archetype, having undergone many well-sampled outbursts that established hysteresis in the early 2000s. Cygnus X-1 shows transitions as a persistent source near the state boundary. Recent bright transients MAXI J1348-630 (2019) and Swift J1727.8-1613 (2023) provided detailed modern coverage with NICER, NuSTAR, and Insight-HXMT.