The Shapiro Delay

The idea: spacetime is a longer road near a mass

Fire a radar pulse from Earth, let it bounce off a planet on the far side of the Sun, and time the round trip. If gravity did nothing to light, the echo would return after exactly twice the straight-line distance divided by c. It doesn't. When the line of sight passes close to the Sun, the echo comes back a little late — by up to a couple of hundred microseconds. That extra lag is the Shapiro delay, and it is one of the cleanest demonstrations that mass bends not just the path of light, but the geometry of time itself.

The intuition people reach for first — "the Sun's gravity slows the light down" — is almost right and slightly wrong. A photon's local speed is always exactly c; no observer ever clocks a passing light ray going slower. What changes is the relationship between the photon and a clock-and-ruler set far away. Two things conspire. First, gravitational time dilation makes clocks tick slower deep in the Sun's potential well, so a distant observer's clock counts more seconds while the photon crosses that region. Second, the spatial geometry is curved, so the photon's path is geometrically a touch longer than the flat-space straight line. Add them up and, in the distant observer's coordinates, the round trip takes longer. It is less "light slowed down" and more "the road got longer in both space and time."

The governing equation

For a signal that travels from an emitter at distance r_e from the Sun, grazes the Sun at closest approach (impact parameter) b, and reaches a receiver/reflector at distance r_r, the one-way gravitational time delay in the parameterised-post-Newtonian (PPN) framework is

Δt = (1 + γ) · (GM/c³) · ln[ (r_e + r_r + R) / (r_e + r_r − R) ]

  where  R = the straight-line distance from emitter to receiver
         b = impact parameter (closest approach to the mass)
         γ = PPN curvature parameter  (γ = 1 in general relativity)

For the practical case where both endpoints are far from the Sun and the ray skims close to it (b ≪ r_e, r_r), the logarithm simplifies to the form people usually quote:

Δt ≈ (1 + γ) · (GM/c³) · ln( 4 r_e r_r / b² )

The leading constant is the key timescale. GM_☉/c³ is the Sun's "light-crossing time of its own gravitational radius":

GM_☉/c³ = (6.674×10⁻¹¹ × 1.989×10³⁰) / (2.998×10⁸)³
        = 1.327×10²⁰ / 2.694×10²⁵
        = 4.93×10⁻⁶ s  ≈ 4.93 microseconds

Everything else is a logarithm — slowly varying, of order 10–13 for solar-system geometries — multiplied by (1+γ) = 2 in GR. So the delay is a few microseconds times a logarithm of order ten, landing in the tens-of-microseconds range one way, hundreds round trip. The whole effect lives or dies on that 4.93-microsecond number and the factor of (1+γ).

Why the delay is logarithmic, not 1/b

A natural guess is that the delay should blow up as the ray gets closer to the mass — diverge like 1/b. It doesn't; it grows only as ln(1/b²) = −2 ln b, which is extraordinarily gentle. Halving the impact parameter adds only a fixed increment of 2(1+γ)(GM/c³)·ln 2 ≈ 14 µs round trip, no matter how close you already were. The reason is that the time delay accumulates along the whole path, and the integrand falls off as 1/r away from closest approach; integrating 1/r along a nearly straight line gives a logarithm. This is the same logarithmic structure that appears in the Coulomb scattering of charged particles and in the deflection integral for light bending. It means the Shapiro delay is a "long-range" effect: a signal passing the Sun at ten solar radii still picks up most of the delay a grazing ray would.

The key numbers

Here is the effect across the scales where it has actually been measured, from a radar ping grazing the Sun to a pulse threading a binary pulsar.

System	Deflecting mass	GM/c³	Geometry	Delay (one way)
Earth–Mars radar past Sun	1 M☉	4.93 µs	grazing limb, b ≈ R☉	~120 µs (≈240–250 µs round trip)
Cassini ranging (2002 conj.)	1 M☉	4.93 µs	b ≈ 1.6 R☉	~120 µs (γ to ~2×10⁻⁵)
Signal grazing the Earth	3×10⁻⁶ M☉	~15 ps	limb	~0.1 ns
Binary pulsar (NS companion)	1.0–1.4 M☉	5–7 µs	edge-on, near conjunction	up to ~20–100 µs amplitude
Light grazing Sgr A* horizon	4.3×10⁶ M☉	21 s	strong field	tens of seconds (no longer "small")

The relevant constants: G = 6.674×10⁻¹¹ m³ kg⁻¹ s⁻², c = 2.998×10⁸ m/s, M_☉ = 1.989×10³⁰ kg, R_☉ = 6.96×10⁸ m, 1 AU = 1.496×10¹¹ m. The Sun's gravitational radius GM_☉/c² is just 1.48 km, against a physical radius of 696,000 km — which is exactly why the solar Shapiro delay is a weak-field, microsecond-scale effect and not a dramatic one.

Worked example: a radar echo grazing the Sun

Take the classic superior-conjunction geometry: Earth at r_e = 1 AU on one side of the Sun, a target planet at r_r = 1.5 AU on the other side, and the radar beam just grazing the solar limb so the impact parameter is b = R_☉ = 6.96×10⁸ m. Use the simplified logarithmic form for one way:

Δt ≈ (1 + γ)(GM_☉/c³) · ln(4 r_e r_r / b²)

r_e = 1.496×10¹¹ m
r_r = 2.244×10¹¹ m
b²  = (6.96×10⁸)² = 4.84×10¹⁷ m²

4 r_e r_r / b² = 4 × (1.496×10¹¹)(2.244×10¹¹) / 4.84×10¹⁷
             = 4 × 3.357×10²² / 4.84×10¹⁷
             = 1.343×10²³ / 4.84×10¹⁷
             = 2.77×10⁵

ln(2.77×10⁵) = 12.53

Δt ≈ 2 × 4.93×10⁻⁶ s × 12.53
   ≈ 1.24×10⁻⁴ s
   ≈ 124 microseconds  (one way)

So the round-trip delay is about 247 microseconds. Converted to an apparent range error, c·Δt(round trip)/2 ≈ 37 km: the planet appears about 37 km farther away than straight-line geometry says, purely because of the Sun's curvature of spacetime. Against an Earth–Mars distance of hundreds of millions of kilometres, that is a fractional effect of ~10⁻⁷ — yet 1970s radar timing, good to a fraction of a microsecond, resolved it cleanly. Notice that turning off general relativity (setting γ = 0, the Newtonian "light is just a fast particle" picture) would halve the delay to ~124 µs round trip; the measured value matching the γ = 1 prediction is the test.

How it's measured — radar, spacecraft, pulsars

There are three observational arenas, in increasing precision:

• Planetary radar (1966–1971). Shapiro's group bounced radar off Mercury and Venus from the Haystack 37 m antenna and the Arecibo 305 m dish as the planets swung behind the Sun. The delay appeared as an anomalous bump in the timing residuals peaking at superior conjunction, confirming GR to ~20%, then ~5%.
• Spacecraft ranging (1976–2003). Active transponders beat passive reflection because they return a clean, strong signal. The Viking landers and orbiters at Mars (1976–77) reached ~0.1% on γ. The decisive measurement came from Cassini in June 2002, during its cruise to Saturn: a multi-frequency radio link (X-band and Ka-band) let the team subtract the dispersive delay of the solar corona — the dominant systematic — and isolate the non-dispersive gravitational delay. Bertotti, Iess & Tortora (2003) reported γ = 1 + (2.1 ± 2.3) × 10⁻⁵, still the tightest constraint on the PPN curvature parameter.
• Binary-pulsar timing (1980s–today). Here the delaying mass is the pulsar's companion, and the "clock" is the pulsar itself. Once per orbit, in nearly edge-on systems, the pulses graze the companion and pick up a periodic Shapiro delay. Because pulse arrival times are measured to nanoseconds, this is now the highest-precision arena. The double pulsar PSR J0737−3039 and the millisecond pulsar PSR J1614−2230 are textbook cases.

Weighing neutron stars with the Shapiro delay

This is the application that turned a relativistic curiosity into a workhorse. In a binary pulsar the Shapiro delay is parameterised by two numbers extracted from the timing fit:

range:  r = G M_c / c³        (sets the amplitude — depends on companion mass M_c)
shape:  s = sin i             (sets the sharpness near conjunction — depends on inclination i)

Δt_orbit = −2r · ln[ 1 − e·cos E − s(sin ω (cos E − e) + √(1−e²) cos ω sin E) ]

Because r and s are independent observables, fitting them gives the companion mass and the inclination directly — and combined with the Keplerian mass function, the pulsar mass too. No other relativistic effect (periastron advance, gravitational redshift, orbital decay) is needed, though they provide cross-checks. The landmark result was PSR J1614−2230 (Demorest et al., 2010): an exceptionally edge-on (i ≈ 89.17°) pulsar–white-dwarf binary whose sharp Shapiro signature yielded a neutron-star mass of 1.97 ± 0.04 M_☉. That single number ruled out a swath of "soft" equations of state that could not support so massive a neutron star, reshaping dense-matter physics. Later, PSR J0740+6620 was weighed at ~2.08 M_☉ by the same technique — currently among the most massive neutron stars known.

Historical discovery and key people

By 1960 general relativity had passed three classical tests: Mercury's perihelion precession (43 arcsec/century), the gravitational deflection of starlight (1.75 arcsec at the limb, Eddington's 1919 eclipse), and the gravitational redshift (Pound–Rebka, 1959). Irwin I. Shapiro, then at MIT Lincoln Laboratory, realised in 1964 that the newly available planetary radar could deliver a fourth, independent test — one probing the time component of the metric directly rather than the spatial deflection. His Physical Review Letters paper, "Fourth Test of General Relativity," predicted the delay and laid out the radar experiment.

The first detections came in 1966–67 with Mercury echoes. The Mariner 6, 7 and 9 spacecraft (1970–71) and then the Viking program (1976–77) progressively tightened γ. The corona — whose free electrons add a frequency-dependent plasma delay — was always the enemy; the breakthrough of the Cassini experiment was using two well-separated radio bands so the dispersive coronal delay (∝ 1/ν²) could be measured and removed, leaving the achromatic gravitational delay clean. Meanwhile, after the 1974 discovery of the Hulse–Taylor binary pulsar opened relativistic pulsar timing, the Shapiro delay migrated from the solar system into binary systems, where it remains a precision tool today.

Related and generalised forms

• Gravitational lensing time delay. In a strong lens producing multiple images of a background quasar, each image's light takes a different path and samples a different depth of the lens potential. The arrival-time differences (days to years) are part geometric, part Shapiro. Measuring them — "time-delay cosmography," e.g. the H0LiCOW / TDCOSMO programs — yields the Hubble constant H₀ independently of the distance ladder.
• Frame-dragging (gravitomagnetic) delay. A spinning mass adds a small spin-dependent term to the delay; signals co-rotating versus counter-rotating with the body's spin differ slightly. It is the time-delay cousin of Lense–Thirring precession, far below current solar-system sensitivity but in principle present.
• Shapiro delay in gravitational-wave detection. Pulsar timing arrays searching for nanohertz gravitational waves must model the Shapiro delay from the Sun and planets along each pulsar line of sight; it is a known signal that has to be subtracted to expose the stochastic background.
• Strong-field regime. Near a black hole the weak-field logarithm breaks down and the delay must be computed from the full Schwarzschild or Kerr metric. For light grazing Sgr A*, GM/c³ ≈ 21 s, so the "delay" is no longer a small correction — it dominates photon-ring timing relevant to the Event Horizon Telescope.

Common misconceptions and subtleties

• "Light physically slows down." No local observer ever measures light below c. The delay is the difference between coordinate travel time and the flat-space straight-line time, and it splits roughly equally between temporal (clock) and spatial (geometric) curvature. Quoting "light slowed by the Sun" is shorthand that obscures the time-dilation half of the effect.
• Confusing it with light bending. Bending changes the direction of the ray; the Shapiro delay changes its travel time. Both come from the same (1+γ) metric factor, so a theory that gets one right generally gets the other right — but they are logically and observationally distinct tests, which is precisely why Shapiro's was called a fourth test.
• Ignoring the corona. The solar corona's plasma adds a delay that, at radio frequencies, can swamp the gravitational one and varies with solar activity. It scales as 1/ν², so it is dispersive; the gravitational delay is achromatic. Single-frequency ranging cannot separate them — this is why Cassini's dual-band link was essential.
• Thinking the delay diverges at the limb. It grows only logarithmically, so even a ray grazing the photosphere has a finite, modest delay; there is no singular blow-up in the weak field.
• Assuming γ being measured means GR is "proven." γ = 1 to parts in 10⁵ tightly constrains a broad class of alternative (scalar–tensor) gravity theories, but those theories can still differ in the strong field. Binary-pulsar and black-hole tests probe regimes the Cassini measurement cannot reach.