Shapiro Delay

The intuition — a radar echo that runs late

In 1964 Irwin Shapiro proposed a test of general relativity that nobody had thought of before. Bounce a radar pulse off another planet — Venus or Mercury — and time the round trip. When the planet is on the far side of the Sun (near superior conjunction), the radar beam must skim past the Sun's edge on the way out and back. Shapiro predicted the echo would come back late — not by milliseconds, but by tens to hundreds of microseconds — purely because the signal had passed through the Sun's curved spacetime.

Two things conspire to slow the pulse. First, space near a mass is stretched: the true distance the light covers is slightly longer than the flat-space straight line. Second, to a far-away clock the coordinate speed of light dips below c as it descends into the Sun's gravitational potential. Neither effect violates relativity — a local observer riding alongside the beam always clocks it at exactly c. But the cumulative bookkeeping, measured on Earth, is a real, repeatable lag.

This became the fourth classic test of general relativity, joining the perihelion precession of Mercury, the bending of starlight, and the gravitational redshift. Unlike the first three, it was a new prediction — a phenomenon Einstein himself never wrote down.

How the delay arises in the metric

Outside a spherical mass, spacetime is described by the Schwarzschild metric. In the weak-field limit (which is all we need for the Sun) the relevant line element, for light travelling in a plane, behaves like:

ds² = -(1 - 2GM/rc²)c²dt² + (1 + 2GM/rc²)dr² + r²dΩ²

For light, ds = 0. Solving for the coordinate speed of a radial light ray gives:

dr/dt ≈ c·(1 - 2GM/rc²)

The factor in parentheses is less than 1, so to a distant observer the light coordinate-slows. Integrate the time of flight along the path and subtract the flat-space value, and the extra one-way coordinate time for a signal between points at radii r₁ and r₂, with closest approach b to the mass, is:

Δt ≈ (2GM/c³)·ln[ (r₁ + r₂ + R) / (r₁ + r₂ − R) ]

where R is the straight-line distance between the endpoints. When both endpoints are far away and the ray grazes the body (b ≪ r₁, r₂), this reduces to the form most often quoted:

Δt ≈ (2GM/c³)·ln(4·r₁·r₂ / b²)

The logarithm is the signature of the Shapiro delay: the closer the grazing distance b, the larger the delay, but it grows only logarithmically, so it never diverges for a ray that misses the surface.

The 2GM/c³ prefactor — the whole scale of the effect

Every Shapiro delay is set by a single mass-dependent timescale, 2GM/c³. It is just twice the light-crossing time of the body's gravitational radius. The logarithm is dimensionless and usually of order 5–15, so the delay is a small multiple of this prefactor.

Body	Mass M	2GM/c³	Typical grazing delay (one-way)
Sun	1.99×10³⁰ kg	9.85 μs	~110–120 μs (× ln factor ≈ 11–12)
Jupiter	1.90×10²⁷ kg	9.4 ns	~0.1 μs near the limb
Earth	5.97×10²⁴ kg	30 ps	sub-nanosecond (matters for GPS/VLBI)
White dwarf (0.6 M☉)	1.2×10³⁰ kg	5.9 μs	μs-scale in binary pulsar systems
Neutron star (1.4 M☉)	2.8×10³⁰ kg	13.8 μs	tens of μs in tight binaries

Because the prefactor scales linearly with mass, the Sun dominates inside the Solar System, and compact companions dominate in binary-pulsar timing.

Worked example — radar echo off Venus past the Sun

Take a radar pulse sent from Earth, reflected off Venus near superior conjunction, and received back on Earth. Use the grazing form for each leg. With the Sun's prefactor 2GM☉/c³ = 9.85 μs, Earth–Sun distance r₁ ≈ 1.50×10¹¹ m, Venus–Sun distance r₂ ≈ 1.08×10¹¹ m, and a grazing impact parameter just outside the solar radius, b ≈ R☉ ≈ 6.96×10⁸ m:

ln(4·r₁·r₂ / b²)
  = ln(4 × 1.50e11 × 1.08e11 / (6.96e8)²)
  = ln(6.48e22 / 4.84e17)
  = ln(1.34e5) ≈ 11.8

One-way delay  Δt ≈ 9.85 μs × 11.8 ≈ 116 μs
Round trip     (out + back, two grazing legs) ≈ 2 × 116 μs
                ≈ 230 μs at closest grazing

In practice the maximum measured excess round-trip delay for Venus near conjunction is around 200 microseconds — corresponding to about 60 km of extra light-travel path. Shapiro's team at MIT's Haystack Observatory confirmed the prediction to ~3% by 1968, and later Viking-lander ranging on Mars pushed agreement with general relativity to ~0.1%.

The tests — Venus radar to Cassini to pulsars

Experiment	Year	Method	Result / precision
Haystack & Arecibo radar	1966–71	Radar echoes off Venus & Mercury near conjunction	First confirmation, ~3–5%
Mariner 6 & 7	1970	Active spacecraft transponder ranging	~3%
Viking landers (Mars)	1976–79	Lander transponder ranging through conjunction	γ = 1 to ~0.1%
Cassini	2002	Doppler tracking at solar conjunction	γ = 1 + (2.1 ± 2.3)×10⁻⁵ (best)
Hulse–Taylor pulsar B1913+16	1970s–	Binary pulse-arrival timing	Consistent with GR
Double pulsar J0737−3039	2004–	Shapiro delay of pulses past companion	Masses to <0.1%; γ confirmed

The Cassini measurement is the gold standard. Rather than timing a pulse, NASA tracked the frequency shift of a continuous radio link as the time-varying delay changed while the probe slid behind the Sun. That converted a microsecond timing problem into a frequency-stability problem, which radio science does far better — hence the 10⁻⁵ precision on the parameter γ.

Where the Shapiro delay shows up

• Spacecraft navigation. Deep-space ranging (Voyager, Cassini, the Deep Space Network) must include the solar Shapiro term — ignoring it would mis-locate a probe by tens of km near conjunction.
• Pulsar timing arrays. NANOGrav, EPTA, and PPTA model the Shapiro delay from the Sun (and planets) on every pulse, and the Shapiro delay from binary companions to weigh neutron stars.
• Weighing neutron stars. The shape and amplitude of the binary Shapiro delay yield the companion mass and orbital inclination — the most precise stellar-mass scale we have.
• Gravitational-wave detection prospects. A passing gravitational wave imprints a tiny correlated Shapiro-like delay across a pulsar timing array — the Hellings–Downs signature recently detected.
• VLBI and geodesy. Very-long-baseline interferometry corrects for the Earth's and Sun's gravitational delay (picosecond–nanosecond level) to align radio telescopes worldwide.
• Testing alternative gravity. Because the delay scales with the PPN parameter γ, every new measurement bounds scalar-tensor theories and screened modifications of gravity.

Shapiro delay vs light bending — same metric, different observable

Light bending and the Shapiro delay are siblings: both come from the same weak-field Schwarzschild metric and both carry the same factor of (1 + γ)/2 in the parameterized-post-Newtonian framework. But they are different measurements.

Property	Light bending	Shapiro delay
Observable	Angular deflection of the ray	Extra arrival time of the signal
Solar magnitude	1.75 arcsec at the limb	~200 μs round trip (Venus)
How measured	Star positions during eclipse; VLBI	Radar/Doppler timing through conjunction
Dominant cause	Spatial curvature (half) + potential (half)	Coordinate slowing + path lengthening
Distance scaling	Depends only on impact parameter b	Grows logarithmically with r₁·r₂/b²
Best precision	VLBI: γ to ~10⁻⁴	Cassini: γ to ~2×10⁻⁵
Discovered	Einstein 1915; Eddington 1919	Shapiro 1964; new prediction

Common misconceptions and edge cases

• "Light slows down, so c isn't constant." No. The coordinate speed slows for a distant bookkeeper, but every local measurement gives exactly c. The delay is geometry, not a change in the local speed of light.
• "The delay is the same as gravitational time dilation." They share metric roots but measure different things — clock rate at a place versus transit time along a path. The Shapiro delay also includes a path-lengthening piece pure time dilation lacks.
• "It diverges if the ray grazes the surface." No — the dependence on impact parameter is logarithmic, so even a near-grazing ray gives a finite delay. Only a path threading the exact center would blow up, which is unphysical.
• "Only the Sun matters." In binary pulsars the companion star produces the dominant delay; in precision geodesy even Earth's ~30 ps and Jupiter's ~9 ns terms must be modeled.
• "It needs a one-way trip." The classic test is a round trip (out and back), so the measured delay is twice the one-way value — and the signal grazes the mass on both legs.
• "It bends the light and delays it independently." They are two projections of one curved-spacetime trajectory: the same null geodesic that delays the signal also deflects it.