Resolving its shadow by correlating noise. Not one telescope — a tradespace of them.
TON 618 is the heaviest black hole we have ever put a number on. We see it as a quasar 10.8 billion light-years away — a portrait of the universe at a quarter of its present age.
That shadow — 2.4 millionths of an arcsecond — is the apparent size of a pea resting on the surface of the Moon, seen from your back garden. That is the thing this mission is built to measure.
A telescope's sharpness is set by its width in wavelengths. The Event Horizon Telescope turned the whole Earth into one dish to image M87* — a shadow 42 µas across.
TON 618's shadow is roughly 18× smaller still. To resolve it at radio wavelengths the array must stretch across about one million kilometres — nearly three times the distance to the Moon. No single planet is big enough. The dishes have to fly.
Classical interferometry records the radio wave itself — amplitude and phase — at every dish, then correlates them. Spread that across cislunar space and the bill comes due.
Hydrogen masers on every craft, locked to one part in 1015.
A phase-coherent radio tie between spacecraft, live, the whole observation.
Every baseline known to a fraction of a wavelength — millimetres — across a million km.
They correlated intensity — how the brightness flickers — at two separate detectors. Light from a thermal source arrives in faint bunches; how much two detectors share that bunching encodes the source's angular size.
Throw the phase away, and every phase-stability nightmare evaporates with it. The two dishes never have to talk to each other while they observe.
The price is a steep one: sensitivity scales as (source brightness / system noise)². This only works on a source that is both compact and genuinely bright. Almost nothing is. TON 618 is.
I pulled the radio catalogs for this sight line. TON 618 is a ~330 mJy flat-spectrum compact source — a self-absorbed core, not diffuse fluff.
NVSS 333 mJy and FIRST 324 mJy (inside 1.2″) at 1.4 GHz; GB6 345 mJy at 4.85 GHz. A spectral index near zero is the signature of a compact core.
At z = 2.219 the universe stretches rest-frame 138 GHz down to 43 GHz observed — exactly where 15-metre dishes and cooled HEMT amplifiers already exist. The cosmos down-converts the EHT band for us.
Picture only two spacecraft, each with a modest ~300 m² of collecting area — a rough placeholder; the real aperture isn't fixed yet — drifting apart on heliocentric orbits until ten million miles (0.11 AU) separate them. At 43 GHz that lone pair resolves 0.09 microarcseconds — below half a gravitational radius, well inside the photon ring. Resolution is lavish; the fight is sensitivity and coverage, because two small elements give one baseline at a time and little area to spare. A sparse long-baseline pair, or a cloud of small dishes close in — both are just points in the same space.
Records how its own signal flickers — power vs. time. That's all.
Spacecraft never talk during the observation. Clocks align in post to ~30 ps.
Later multiplies every pair of streams to recover |V(u,v)|².
The two-aperture array you just saw is one dot in a large design space, chosen to be concrete — not because it is the answer. A cloud of small cislunar dishes is another. OBLIVION's purpose is to chart that whole space: which combinations of aperture, element count, baseline and frequency actually buy each science goal, and at what cost.
More area per element lifts SNR, but mass and stowed volume climb with it.
More elements fill the u–v plane faster and lift SNR as √(pairs).
How far the elements drift apart sets the sharpest resolution.
Higher frequency sharpens the image but demands better surfaces and colder receivers.
Longer integration buys SNR and turns variability into a spin clock.
Two free-flyers is one shape — a swarm or a space–ground tie are others.
Every one of these axes is a slider in the explorer below. The gold marker shows where your current design sits on each axis — it moves as you tune the controls.
This is the question OBLIVION exists to answer. The defaults are the two-aperture interplanetary array above; every slider moves you to a different architecture. The readouts use the Hanbury Brown sensitivity law and the curve is the squared visibility that array would sample — watch which science survives.
Two small elements on a vast baseline buy lavish resolution — 0.4 r_g — but one baseline and tight sensitivity, landing at tier C. Lift the collecting area or add elements to push the SNR into tier D; or trade resolution for the many baselines an image needs with a cislunar swarm. Mapping those trades is OBLIVION's whole job.
15 m at 0.15 mm rms. 10 m designed (Millimetron); 12 m near-term.
Cryo HEMTs at 12–30 K; KVN proves simultaneous tri-band optics.
ALMA already processes 16 GHz/antenna; rad-tolerant RFSoCs bank the rest.
DSAC ion clock flew at 3×10-15/day. Optical time transfer <50 ps.
GRACE-FO laser ranging: micron-class at 220 km. We need centimetres.
~128 Gb/s recorded per node. Needs ~10 Gb/s lasercom + burst-mode ops — HBT tolerates the gaps.
Locate the first |V|² null to pin the ring diameter to a few percent — extending the Kerr shadow-size test of M87* and Sgr A* four orders of magnitude up the mass ladder.
Run the same equation backwards: assume GR, and the ring angle returns the angular-diameter distance to redshift 2.2 — beyond BAO's reach and independent of the distance ladder.
At 22/43 GHz the SNR runs into the hundreds. Core size and brightness temperature vs. frequency map jet collimation and confront the 1012.5 K inverse-Compton limit.
Each dish measures g²(τ) for free, tracking variability at the 7.5-day light-crossing time. Multi-year monitoring turns quasi-periodicity into a spin constraint.
Detecting photon bunching at ~200σ from z = 2.219 is the most distant second-order-coherence measurement conceivable — a probe of cosmological decoherence and spacetime foam.
Intensity interferometry keeps amplitude, not phase, so the deliverable is fitted geometry — ring diameter, size, asymmetry — not a true image. Respectable company: the EHT's M87* diameter leaned hard on visibility-amplitude nulls too.
The single biggest unknown is the slider marked shadow-scale flux: how much of that 150 mJy core lives below 10 r_g is unmeasured. So the first step is trivially cheap — a few hours of VLBA + ALMA at 43/86 GHz on a source never imaged at high frequency settles the whole tier-C/D case before any metal is bent.