Apparent Superluminal Motion: The Jet That Fakes Faster Than Light

The Mechanism: A Blob Chasing Its Own Photons

Begin with something that has nothing to do with relativity: light takes time to travel. A knot of plasma is launched from the accretion flow of a supermassive black hole at speed v = βc, along a straight line that makes an angle θ with our line of sight. It emits a flash at event A, and a time Δt later — measured in our frame, the frame of Earth and the host galaxy — a second flash at event B.

Between A and B the knot covers a path length vΔt. Split it into two components:

• Radial: vΔt·cosθ, straight toward us. This is the piece that fools us.
• Transverse: vΔt·sinθ, sideways across the plane of the sky. This is the piece we actually watch move.

Because the knot has closed the gap by vΔt·cosθ, the second photon has a shorter journey than the first by exactly that distance, and so arrives earlier than it "ought" to by vΔt·cosθ/c. The gap between the two arrivals is therefore not Δt but

Δt_obs = Δt − (vΔt·cosθ)/c = Δt·(1 − β·cosθ).

Now do what any observer naively does: divide the sideways displacement you saw by the time you waited to see it.

β_app = (β·Δt·sinθ) / [Δt·(1 − β·cosθ)] = β·sinθ / (1 − β·cosθ).

Nothing in that derivation used a Lorentz transformation, time dilation, or length contraction — only the finiteness of c and Pythagoras. The knot never exceeds c; its worldline stays strictly inside the light cone. What has been compressed is not the knot's motion but our clock for it — the denominator (1 − β·cosθ), which collapses toward zero as β → 1 and θ → 0.

Working the Numbers: Ten Years of Flight, Delivered in 73 Days

The formula has three sharp consequences worth deriving rather than quoting.

1. There is a hard speed threshold. Maximise β_app over θ: setting d/dθ = 0 gives cosθ = β, and substituting back yields β_app^max = β/√(1 − β²) = βγ. Demanding βγ > 1 gives β² > 1 − β², i.e. β > 1/√2 ≈ 0.7071 (γ > √2). Below 70.7% of light speed no viewing angle can fake superluminal motion. This is the only place special relativity enters the phenomenon — as a threshold, not as a cause.

2. The best viewing angle is θ ≈ 1/γ. For β → 1 and small θ, 1 − β·cosθ ≈ 1/(2γ²) + θ²/2, so β_app ≈ 2γ²θ/(1 + γ²θ²), peaking at γθ = 1 with β_app = γ. Fast jets have a narrow "sweet cone" of orientations, which is exactly the opening angle of their own beamed emission.

3. A worked case. Take β = 0.99, so γ = 1/√(1 − 0.9801) = 7.09. The optimum is cosθ = 0.99, θ = 8.11°, sinθ = 0.1411. Let the knot fly for Δt = 10 years of real, honest coordinate time. It covers 9.9 light-years of path, of which 1.40 ly is transverse. But the arrivals are separated by Δt_obs = 10 × (1 − 0.9801) = 0.199 yr ≈ 73 days. Divide: 1.40 ly in 0.199 yr = 7.0c apparent — matching βγ = 0.99 × 7.09 = 7.02, as it must.

Reading it backwards. Since β_app ≤ βγ = √(γ² − 1), a measured apparent speed sets a floor on the true one: γ ≥ √(1 + β_app²) and β ≥ β_app/√(1 + β_app²). A blazar clocked at 50c cannot have a bulk Lorentz factor below ~50. This is how an illusion becomes a measuring instrument.

How It Is Actually Measured: Milliarcseconds Per Year

Nobody sees a jet move in real time. What is measured is a proper motion μ — an angular drift of a compact feature, in milliarcseconds per year — recovered by imaging the same source at many epochs and tracking identifiable knots. Converting μ into a speed requires the distance and, for cosmological sources, the redshift:

β_app = μ · D_A · (1 + z) / c = μ · D_L / [c(1 + z)],

where D_A and D_L are the angular-diameter and luminosity distances. The (1 + z) factor undoes cosmological time dilation, which stretches the observed epoch spacing.

The instrument of choice is very long baseline interferometry (VLBI). The Very Long Baseline Array — ten 25 m dishes spanning 8,611 km from Mauna Kea to St. Croix — reaches roughly 0.5 milliarcsecond resolution at 15 GHz, enough to resolve structures a few parsecs across in a quasar billions of light-years away. The MOJAVE programme (and its 2 cm Survey predecessor, running since 1994) has monitored several hundred bright jets this way, producing the largest catalogue of apparent speeds ever assembled; its distribution peaks near ~10c and extends to ~50c. The Event Horizon Telescope pushed the same technique to ~20 microarcseconds, imaging the base of the 3C 279 jet in 2017.

Scale matters enormously. At 3C 273 (z = 0.158) a proper motion of 1 mas/yr already corresponds to about 10c. At M87, only 16.8 Mpc away, 1 mas/yr is a mere 0.27c — which is why the Hubble Space Telescope, with far coarser angular resolution than VLBI but a nearby target, could measure knot HST-1 sliding at roughly 23 mas/yr and report 6.0c. Optical, radio and X-ray instruments all measure the same geometry; they simply trade resolution against proximity.

The Real Catalogue: From a 1966 Prediction to a Neutron-Star Merger

The prediction came first. In 1966 Martin Rees published "Appearance of Relativistically Expanding Radio Sources" (Nature 211, 468), showing that a source expanding at near-c would appear to expand faster than light, and using that to explain why quasars varied far too quickly for their apparent sizes. The geometry was on paper five years before the observation that would demand it.

The detection came in 1971. Early VLBI experiments found changing structure in 3C 279 and 3C 273 implying transverse speeds near 10c — Whitney and collaborators in Science, Cohen and collaborators in the Astrophysical Journal. The results were unwelcome enough that alternatives were seriously entertained — that quasar redshifts were not cosmological, or that a "Christmas-tree" of independently flaring components was masquerading as motion. Rees's geometry won.

Then the effect spread across the astrophysical zoo.

• M87 — Biretta, Sparks & Macchetto (1999) used HST to track knots in the optical jet at up to 6.0c, in a galaxy whose black hole the EHT would later image directly.
• GRS 1915+105 — Mirabel & Rodríguez (Nature, 1994) found the first Galactic case: a stellar-mass black hole ejecting twin blobs, the approaching one at 1.25c apparent. Because both jets were seen, the two apparent speeds could be combined — β·cosθ = (β_a − β_r)/(β_a + β_r) = 0.32. That ratio is distance-independent, and by itself caps the angle at θ ≲ 72°; adopting the 12.5 kpc distance then fixes β ≈ 0.92 and θ ≈ 70°. The term "microquasar" had been coined two years earlier by the same group, for 1E 1740.7−2942.
• GW170817 — Mooley et al. (Nature, 2018) tracked the radio afterglow of a neutron-star merger with global VLBI between 75 and 230 days and measured 4.1 ± 0.5 c apparent. That single number showed the merger had launched a genuine relativistic, collimated jet seen ~20° off-axis, tying gravitational-wave astronomy to short gamma-ray bursts.

Its Constant Companion: Doppler Beaming — And Why It Is a Different Effect

The same denominator that fakes the speed also controls the brightness, through the Doppler factor

δ = 1 / [γ(1 − β·cosθ)].

For our β = 0.99, θ = 8.11° example, δ = 1/(7.09 × 0.0199) = 7.09 — numerically equal to γ, which is always true at the optimal angle cosθ = β. Beaming does three things at once: it blueshifts frequencies by δ, compresses variability timescales by δ, and amplifies the observed flux density as S_ν ∝ δ^3+α for a discrete moving blob (or δ^2+α for a steady flow), where α is the spectral index in the convention S_ν ∝ ν^−α. With a typical α ≈ 0.7, δ = 7.09 brightens the jet by δ^3.7 ≈ 1,400×.

The receding counter-jet suffers the inverse. Its de-boosting gives a jet-to-counterjet ratio of [(1 + β·cosθ)/(1 − β·cosθ)]^3+α, which for these numbers is ~2 × 10⁷. That is why M87, 3C 273 and nearly every blazar look like one-sided jets even though the launching mechanism is symmetric.

Beaming also rescued jets from the inverse-Compton catastrophe: a synchrotron source cannot sustain a rest-frame brightness temperature above ~10¹² K without its own electrons scattering their own photons and radiating themselves to death (Kellermann & Pauliny-Toth, 1969). Yet observed brightness temperatures reach 10¹³–10¹⁴ K, with space-VLBI missions such as RadioAstron pushing higher still. Since the observed T_b scales roughly with δ, beaming reconciles the numbers.

But it is not the same effect. Superluminal motion is a projection and timing phenomenon: it moves a position. Beaming is radiative transfer: it changes how bright and how blue a source looks without moving it anywhere. They share a denominator and a cause, and are routinely conflated. One consequence of their pairing: any flux-limited survey preferentially selects small θ, so catalogues of superluminal sources are heavily biased toward the very geometries that produce the illusion.

Where the Simple Picture Breaks, and What It Is Confused With

Pattern speed is not bulk speed. The formula assumes we are watching a lump of matter. What VLBI actually tracks is a brightness feature — plausibly a shock front, a recollimation node, or a kink instability — and a pattern can move faster or slower than the plasma carrying it. MOJAVE finds jets whose components accelerate over tens of parsecs, and sources such as M87 showing slow, near-stationary features alongside fast ones in the same jet. A single (β, θ) pair is a cartoon of a sheared, accelerating flow.

The "Doppler crisis." In TeV-emitting BL Lac objects like Mrk 421 and Mrk 501, minute-timescale gamma-ray variability and gamma-gamma transparency arguments demand Doppler factors of order 50, yet their VLBI knots crawl along at sub-luminal to, at most, mildly superluminal apparent speeds. Proposed fixes — spine-sheath velocity structure, decelerating flows, reconnection-driven "mini-jets" moving relativistically within a slower bulk flow — remain an active research front.

Look-alikes that are not this effect.

• Genuine FTL travel. Never observed. The knot's speed is always < c; the illusion is entirely undone by correcting for light-travel time.
• Cherenkov radiation. A real particle outrunning light in a medium — a physical effect with a physical shock cone, not a projection.
• Light echoes. The expanding dust ring around V838 Monocerotis (2002) appeared to spread superluminally, and nothing moved at all — pure light-travel-time geometry, the same bookkeeping with a zero-velocity source.
• Sweeping spots and phase velocities. A laser spot swept across the Moon, or a wave's phase velocity in a plasma or waveguide, can exceed c because no energy or information rides along.
• Cosmological recession. Galaxies beyond the Hubble radius recede faster than c, but that is the expansion of space, not motion through it.

Open questions sit upstream of the illusion: whether jets are launched by the Blandford–Znajek mechanism extracting black-hole spin or by magnetocentrifugal disc winds; where the flow reaches Γ ~ 10–50 (EHT and GMVA imaging place M87's acceleration and collimation zone out to ~10⁵ gravitational radii); and whether the plasma is electron-positron or electron-proton. Apparent superluminal motion answers none of these — but every floor γ ≥ √(1 + β_app²) it delivers is one those models must clear.

Measured apparent superluminal speeds, and the minimum bulk Lorentz factor each one forces (γ ≥ √(1 + β_app²)).

Object	Apparent transverse speed β_app	Implied minimum γ	Instrument and landmark measurement
3C 279 (quasar, z = 0.536)	~10c in 1971; up to ~20c since	≥ 10 (≥ 20 for the fastest knots)	VLBI — Whitney et al., Science (1971); later VLBA/EHT
3C 273 (quasar, z = 0.158)	~5–10c (≈1 mas/yr knots)	≥ 5–10	VLBI — Cohen et al., ApJ (1971); VLBA monitoring
M87 knot HST-1 (16.8 Mpc)	6.0c (≈23 mas/yr)	≥ 6.1	Hubble Space Telescope — Biretta, Sparks & Macchetto (1999)
GRS 1915+105 (Galactic microquasar)	1.25c approaching, 0.65c receding	≥ 1.6	VLA — Mirabel & Rodríguez, Nature (1994)
GW170817 radio jet (40 Mpc)	4.1 ± 0.5 c	≥ 4.2	VLBI (VLBA+VLA+EVN) — Mooley et al., Nature (2018)
MOJAVE blazar sample (fastest)	~50c	≥ 50	VLBA 15 GHz, monitoring since 1994 — Lister et al.