PPN Parameters Beta and Gamma

Two numbers stand between Newton and Einstein. In the slow-motion, weak-field world of the solar system, every metric theory of gravity can be written as Newtonian gravity plus a series of small corrections, each carrying a dimensionless coefficient. The parametrized post-Newtonian (PPN) framework turns "which theory of gravity is right?" into "what are the values of these coefficients?" — and among the ten standard ones, two do all the heavy lifting for a static, spherically symmetric mass like the Sun. Following Eddington, Robertson and Schiff they are written $\gamma$ and $\beta$ . The parameter $\gamma$ lives in the spatial part of the metric and answers "how much does a unit of mass curve space?" — it is what samples light, so it governs the deflection of starlight and the Shapiro radar delay. The parameter $\beta$ lives at second order in the time-time component $g_{00}$ and answers "does gravity gravitate, and by how much?" — it measures the nonlinearity of the field's self-interaction and controls the relativistic part of Mercury's perihelion advance. General relativity predicts $\gamma=\beta=1$ exactly, not by convention but because a single geometric field with no extra scalars fixes $\gamma$ , and the Bianchi-identity nonlinearity of Einstein's equations fixes $\beta$ . Alternative theories drift away: Brans–Dicke, with a scalar field diluting the spatial curvature, gives $\gamma<1$ . This article shows where $\gamma$ and $\beta$ come from, why exactly these two suffice, and how a handful of precision experiments have pinned both to unity to parts in ten thousand or better.

Key takeaways

The Eddington–Robertson–Schiff parameters $\gamma$ and $\beta$ measure the two physically distinct leading departures from Newtonian gravity in a static, spherically symmetric field.
$\gamma$ sets the spatial curvature produced per unit mass — it controls light deflection, the Shapiro delay, and the $O(U)$ spatial metric; light, moving at $c$ , feels the full $(1+\gamma)$ effect.
$\beta$ sets the nonlinearity of gravity's self-coupling — it enters $g_{00}$ at $O(U^2)$ and governs perihelion advance and, through the Nordtvedt effect, strong-equivalence violation.
General relativity gives $\gamma=\beta=1$ exactly; the classical tests then read as formulas linear in $\gamma$ and $\beta$ , converting each precision measurement into a direct bound.
Experiment confirms both: Cassini gives $\gamma-1=(2.1\pm2.3)\times10^{-5}$ , and lunar laser ranging plus perihelion data bound $\beta-1$ at the $10^{-4}$ level.
Key idea: write the static, spherically symmetric metric in isotropic coordinates as $g_{00}=-(1-2U+2\beta U^2)$ , $g_{ij}=(1+2\gamma U)\delta_{ij}$ with $U=GM/(rc^2)$ . Then $\gamma$ sets spatial curvature per unit mass (light bending, Shapiro delay) and $\beta$ sets gravitational nonlinearity (perihelion shift); GR predicts $\gamma=\beta=1$ .

The core insight

Write the static, spherically symmetric metric in isotropic coordinates as $g_{00}=-(1-2U+2\beta U^2)$ , $g_{ij}=(1+2\gamma U)\delta_{ij}$ , where $U=GM/(rc^2)$ is the Newtonian potential. Then $\gamma$ sets the spatial curvature per unit mass (light bending, Shapiro delay) and $\beta$ sets the strength of gravitational nonlinearity (perihelion shift). General relativity predicts $\gamma=\beta=1$ .

Where this lives today

The two Eddington–Robertson–Schiff parameters remain the front line of experimental gravitation, and every improvement tightens the noose around alternatives to general relativity. The tightest bound on $\gamma$ still comes from the 2002 Cassini solar conjunction, where the Doppler shift of the spacecraft's radio link measured the changing Shapiro delay and returned $\gamma-1=(2.1\pm2.3)\times10^{-5}$ — a result that alone forces Brans–Dicke coupling above $4\times10^4$ and constrains a broad class of scalar–tensor and $f(R)$ dark-energy models. Lunar laser ranging, now reaching millimetre precision on the Earth–Moon distance via retroreflectors left by Apollo and Lunokhod, tests the Nordtvedt effect and bounds $\eta_{\text{N}}=4\beta-\gamma-3$ at the $10^{-4}$ level, giving the cleanest handle on $\beta$ . Radio interferometry (VLBI) routinely measures light deflection across the whole sky, not just at the limb, confirming $\gamma=1$ to $10^{-4}$ . Looking ahead, missions and campaigns aim to push $\gamma-1$ below $10^{-6}$ , binary-pulsar and pulsar-timing-array observations extend the tests into the strong field where the PPN expansion itself breaks down, and gravitational-wave observatories probe the fully nonlinear regime that $\beta$ only hints at. Whenever a new theory of gravity is proposed to explain dark energy or dark matter, its first exam is still: what are your $\gamma$ and $\beta$ ?

The physical picture

Newton's gravity is entirely a story about time: the whole force follows from a single potential sitting in $g_{00}$ , and space stays flat. Einstein's gravity curves space too, and $\gamma$ is exactly the dial that says how much — with $\gamma=1$ a unit of mass warps space just as strongly as it warps the flow of time. That extra spatial curvature is invisible to a slow planet, which crawls through space and barely samples it, but it is fully visible to a photon, which sweeps through space at the same rate it advances in time. The consequence is the most famous factor in the subject: a light ray grazing the Sun is deflected by $1.75''$ , exactly twice what a "Newtonian photon" argument gives, and the doubling is entirely the $\gamma$ contribution from curved space. The first figure separates that deflection into its two equal halves.

Figure

Starlight grazing the Sun is deflected by 4GM/(c²R) = 1.75″ at the limb when γ = 1 — half (0.875″) comes from the time curvature of g₀₀, the Newtonian-photon piece, and the other half (0.875″) comes from the spatial curvature set by γ, which is exactly why the full deflection is twice the naive value

The parameter $\beta$ tells a different story — not about how mass curves the world, but about whether gravity feels its own weight. The gravitational binding energy of a body is a form of energy, and energy gravitates, so a heavy self-gravitating body should generate a field slightly different from the sum of its parts. $\beta$ measures that nonlinearity; whether a massive body falls like a test particle — the strong equivalence principle — is fixed not by $\beta$ alone but by the Nordtvedt combination $4\beta-\gamma-3$ , which is why Brans–Dicke violates the SEP despite sitting exactly at $\beta=1$ . Because it enters $g_{00}$ only at $O(U^2)$ it hides from light and shows up in slow-body orbits, where it combines with $\gamma$ to advance Mercury's perihelion. The second figure is what the experiments actually deliver: a zoom into the $(\gamma,\beta)$ plane, where every metric theory is a point and general relativity sits at $(1,1)$ , hemmed in on all sides.

Figure

The observational verdict in the γ–β plane, zoomed on the general-relativity point at (1, 1): the Cassini Shapiro-delay measurement pins γ−1 = (2.1 ± 2.3)×10⁻⁵ (narrow vertical band) while lunar laser ranging and Mercury's perihelion bound β−1 at the ×10⁻⁴ level (horizontal band); their intersection is a tiny box that contains GR and excludes Brans–Dicke, which lies on the β = 1 line at γ < 1

The core equations

The whole framework starts from one line: the metric of a static, spherically symmetric mass, written in isotropic coordinates and expanded to the order at which solar-system tests operate. The Newtonian potential $U$ appears linearly in both blocks; the two PPN parameters are the coefficients of the specific higher pieces that distinguish theories.

g_{00}=-\left(1-2U+2\beta U^2\right),\qquad g_{0i}=0,\qquad g_{ij}=\left(1+2\gamma U\right)\delta_{ij} \qquad (1)

where: $U$ — Newtonian potential $U=GM/(rc^2)$ , dimensionless and positive; $\gamma$ — spatial-curvature-per-unit-mass parameter; $\beta$ — nonlinearity-of-superposition parameter; $\delta_{ij}$ — the flat spatial (Kronecker) metric. Reading: the $-2U$ term reproduces Newton, $\gamma$ multiplies the leading spatial curvature, and $\beta$ multiplies the second-order self-coupling term in $g_{00}$ ; setting $\gamma=\beta=1$ recovers general relativity. Written as a line element in isotropic spherical coordinates, the same metric shows that space is conformally flat — the spatial part is an overall factor times ordinary flat space, which is what makes $\gamma$ so cleanly interpretable as "curvature per unit mass".

\mathrm{d}s^2=-\left(1-2U+2\beta U^2\right)c^2\mathrm{d}t^2+\left(1+2\gamma U\right)\left(\mathrm{d}r^2+r^2\mathrm{d}\Omega^2\right) \qquad (2)

where: $\mathrm{d}\Omega^2=\mathrm{d}\theta^2+\sin^2\theta\,\mathrm{d}\varphi^2$ is the unit two-sphere metric; $r$ — isotropic radial coordinate, distinct from the areal (Schwarzschild) radius. Reading: the coefficient $1+2\gamma U$ multiplying the entire flat spatial line element is the physical content of $\gamma$ ; a photon's path length is stretched by it, and that stretch is half of the light deflection. Because a photon travels at $c$ , it samples the spatial metric at the same order as $g_{00}$ , so the deflection of a light ray by a mass depends on $\gamma$ but not $\beta$ . The result is the single most-cited PPN formula, and the $(1+\gamma)/2$ prefactor multiplies the general-relativistic value $4GM/(c^2b)$ .

\Delta\phi_{\text{light}}=\frac{1+\gamma}{2}\,\frac{4GM}{c^2 b} \qquad (3)

where: $b$ — impact parameter (distance of closest approach); $\Delta\phi_{\text{light}}$ — total deflection angle. Reading: with $\gamma=1$ this is $4GM/(c^2b)$ , giving $1.75''$ at the solar limb; with the naive $\gamma=0$ "Newtonian photon" it would be half that, so a measured deflection directly weighs $\gamma$ . The same $(1+\gamma)/2$ factor governs the Shapiro delay — the extra round-trip light-travel time when a radar signal passes near a mass — because it too is a pure null-geodesic effect controlled by the spatial metric.

\Delta t_{\text{Shapiro}}=\frac{1+\gamma}{2}\,\frac{4GM}{c^3}\ln\!\left(\frac{4 r_e r_p}{b^2}\right) \qquad (4)

where: $r_e$ — Sun-to-emitter distance; $r_p$ — Sun-to-reflector distance; $b$ — impact parameter of the signal path. Reading: this is the round-trip delay; the logarithm grows as the ray grazes closer to the Sun ( $b\to R_\odot$ ), which is exactly the geometry Cassini exploited near solar conjunction to bound $\gamma$ . For a slow massive body both parameters enter, because the orbit mixes the $O(U^2)$ piece of $g_{00}$ (carrying $\beta$ ) with the $O(U)$ spatial piece (carrying $\gamma$ ). The per-orbit perihelion advance is proportional to the combination $(2+2\gamma-\beta)/3$ .

\Delta\varpi=\frac{2+2\gamma-\beta}{3}\,\frac{6\pi GM}{c^2 a(1-e^2)} \qquad (5)

where: $\Delta\varpi$ — advance of pericenter per revolution; $a$ — semi-major axis; $e$ — eccentricity. Reading: for $\gamma=\beta=1$ the prefactor is unity and one recovers the textbook $6\pi GM/[c^2a(1-e^2)]$ , which for Mercury sums to $42.98''$ per century. Finally, whether a self-gravitating body obeys the strong equivalence principle is measured by the Nordtvedt parameter, a specific combination of $\beta$ and $\gamma$ that vanishes in general relativity and is generically nonzero once $\beta\neq1$ or $\gamma\neq1$ , although particular pairs of deviations can cancel inside it.

\eta_{\text{N}}=4\beta-\gamma-3 \qquad (6)

where: $\eta_{\text{N}}$ — Nordtvedt parameter; a nonzero value means bodies of different gravitational binding energy fall at different rates. Reading: GR gives $4(1)-1-3=0$ , so Earth and Moon fall toward the Sun identically; lunar laser ranging measures no such difference, bounding $\eta_{\text{N}}$ and hence $\beta$ .

Derivation ASchwarzschild in isotropic coordinates gives gamma and beta equal to one

The goal is to prove that general relativity's unique static, spherically symmetric vacuum solution sits exactly at $\gamma=\beta=1$ , the benchmark every alternative is measured against. We take the Schwarzschild metric in its usual areal-radius form, apply the coordinate change that renders the spatial part conformally flat, expand the resulting metric coefficients as power series in the small potential $U=GM/(c^2r)$ , and read off the coefficients of $U^2$ in $g_{00}$ and of $U$ in $g_{ij}$ by matching term-by-term against the PPN template of equation (1). No approximation beyond the weak-field expansion is used, so the result is exact for GR.

Step 1. Start from Schwarzschild in areal coordinate $\rho$ : $\mathrm{d}s^2=-(1-2GM/c^2\rho)c^2\mathrm{d}t^2+(1-2GM/c^2\rho)^{-1}\mathrm{d}\rho^2+\rho^2\mathrm{d}\Omega^2$ . Why: this is the exact vacuum solution, but its spatial part is not conformally flat, so the PPN parameters cannot be read off directly — we must change coordinates first.

Step 2. Substitute $\rho=r(1+GM/2c^2r)^2$ , the transformation that makes the spatial metric a scalar factor times flat space. With $m\equiv GM/c^2$ and $U=m/r$ , the time-time component becomes the compact ratio $g_{00}=-\left[(1-m/2r)/(1+m/2r)\right]^2$ . Why: isotropic coordinates are the gauge in which the canonical PPN definitions are stated; using any other gauge would misattribute the coefficients.

Step 3. Expand $g_{00}$ in powers of $U$ . The ratio gives $(1-U/2)/(1+U/2)=1-U+U^2/2-\cdots$ , and squaring yields $g_{00}=-(1-2U+2U^2-\cdots)$ . Matching to $-(1-2U+2\beta U^2)$ forces $\beta=1$ . Why: the $U^2$ coefficient is where the theory's nonlinearity lives, and GR delivers exactly the coefficient $2$ , i.e. $\beta=1$ , with no free choice.

Step 4. Expand the spatial factor $g_{ij}=(1+m/2r)^4\delta_{ij}=(1+2U+\tfrac32U^2+\cdots)\delta_{ij}$ . The leading correction is $2U$ , so matching to $(1+2\gamma U)\delta_{ij}$ forces $\gamma=1$ . Why: the coefficient of the linear spatial term is the spatial-curvature-per-unit-mass, and GR sets it equal to the time coefficient — this equality is the origin of the doubled light deflection.

Step 5. Conclude that GR's unique spherically symmetric vacuum solution is $\gamma=\beta=1$ exactly, providing the fixed point against which Brans–Dicke ( $\gamma<1$ ) and other theories are compared. Why: because these values come from the structure of Einstein's equations and not from data, any experimental deviation would falsify GR outright. *The symbolic expansion gives $g_{00}=-(1-2U+2U^2+\cdots)$ and $g_{ij}=(1+2U+\cdots)\delta_{ij}$ , confirming $\gamma=\beta=1$ by direct coefficient matching.*

Go deeper · Derivation B · 1 minThe light-deflection factor one plus gamma over two

Derivation BThe light-deflection factor one plus gamma over two

The goal is to show that the deflection of a light ray in the PPN metric is $(1+\gamma)/2$ times the general-relativistic value $4GM/(c^2b)$ , and thereby to expose how the total splits into equal halves — one from the time curvature that even a Newtonian photon would feel, one from the spatial curvature that only $\gamma$ supplies. We treat the ray in the weak-field, small-angle limit, recast the null condition as an effective index of refraction, integrate the transverse gradient of that index along the undeflected straight-line path, and evaluate the resulting standard integral for a point mass.

Step 1. Impose the null condition on the metric restricted to the plane of the ray: $0=-(1-2U)c^2\mathrm{d}t^2+(1+2\gamma U)\mathrm{d}\ell^2$ , where $\mathrm{d}\ell$ is spatial path length. Why: for a photon $\mathrm{d}s^2=0$ , and the $\beta U^2$ term is second order in $U$ so it drops out entirely — this is why light never sees $\beta$ .

Step 2. Solve for the effective refractive index $n=c\,\mathrm{d}t/\mathrm{d}\ell=\sqrt{(1+2\gamma U)/(1-2U)}\approx1+(1+\gamma)U$ . Why: gravity acts on light exactly like a medium with a position-dependent index, and the whole $\gamma$ -dependence is packaged into the single coefficient $(1+\gamma)$ .

Step 3. Compute the deflection as the integral of the transverse gradient of $n$ along the unperturbed path: $\Delta\phi=\int_{-\infty}^{\infty}(\partial n/\partial y)\,\mathrm{d}x=(1+\gamma)\int_{-\infty}^{\infty}(\partial U/\partial y)\,\mathrm{d}x$ , with $y$ perpendicular to the ray. Why: to leading order the bending is small enough that evaluating on the straight line is consistent, and the constant $(1+\gamma)$ factors out of the integral.

Step 4. Insert $U=GM/(c^2\sqrt{x^2+b^2})$ and evaluate the standard integral, $\int_{-\infty}^{\infty}(\partial U/\partial y)\,\mathrm{d}x=2GM/(c^2b)$ , giving $\Delta\phi=(1+\gamma)\cdot 2GM/(c^2b)=\tfrac{1+\gamma}{2}\cdot 4GM/(c^2b)$ . Why: this is the punchline — one factor $GM/(c^2b)$ from time curvature and one $\gamma\,GM/(c^2b)$ from space curvature, equal when $\gamma=1$ , hence the famous doubling. *At the solar limb with $\gamma=1$ , $4GM_\odot/(c^2R_\odot)=8.49\times10^{-6}\,\text{rad}=1.75''$ .*

Go deeper · Derivation C · 1 minThe perihelion factor two plus two gamma minus beta over three

Derivation CThe perihelion factor two plus two gamma minus beta over three

The goal is to trace how both PPN parameters enter a bound orbit and combine into the single coefficient $(2+2\gamma-\beta)/3$ that multiplies the general-relativistic perihelion advance. We linearize the geodesic equation for a slow planet in the metric of equation (1), keep the leading post-Newtonian corrections, reduce the radial motion to the equation for $u=1/r$ as a function of azimuth, identify the term that shifts the angular frequency of the radial oscillation away from $2\pi$ per orbit, and collect the $\gamma$ and $\beta$ contributions into one prefactor.

Step 1. Write the orbit equation $u''+u=GM/L^2+(\text{PN corrections})$ , where $L$ is the specific angular momentum and primes denote $\mathrm{d}/\mathrm{d}\varphi$ . Why: at Newtonian order this gives a closed ellipse with the perihelion fixed; any extra term proportional to $u^2$ is what makes the ellipse slowly rotate.

Step 2. Collect the post-Newtonian source of the $u^2$ term. The $O(U^2)$ piece of $g_{00}$ contributes with weight tied to $\beta$ , while the $O(U)$ spatial metric contributes with weight tied to $\gamma$ through the velocity-dependent part of the geodesic equation. Why: the slow planet feels $g_{00}$ to second order and $g_{ij}$ to first order, so both parameters appear, unlike the null case.

Step 3. Combine the coefficients. The net $u^2$ term shifts the radial oscillation frequency so that the perihelion advances per orbit by $\Delta\varpi=\tfrac{2+2\gamma-\beta}{3}\cdot 6\pi GM/[c^2a(1-e^2)]$ . Why: the specific linear combination $2+2\gamma-\beta$ is dictated by how the two metric pieces enter the effective potential; for $\gamma=\beta=1$ it equals $3$ , cancelling the denominator.

Step 4. Specialize to GR ( $\gamma=\beta=1$ ): the prefactor is unity and Mercury's advance is $42.98''$ per century after Newtonian planetary perturbations are subtracted. Why: this single number, unexplained for half a century, was Einstein's first triumph, and it constrains $\beta$ once $\gamma$ is fixed independently by Cassini. *With Mercury's $a=5.79\times10^{10}\,\text{m}$ and $e=0.206$ , evaluating $6\pi GM_\odot/[c^2a(1-e^2)]$ gives $0.1035''$ per orbit and $42.98''$ per century.*

Limiting cases

How $\gamma$ and $\beta$ specialize across theories and regimes:

Case	Result	Meaning
General relativity	$\gamma=1,\ \beta=1$	The unique values from Einstein's equations with no extra fields; every classical-test formula reduces to its textbook GR value.
Newtonian gravity	drop the $O(U)$ spatial and $O(U^2)$ time terms	Ignoring the $\gamma,\beta$ pieces entirely: orbits close, and light is unbent (or bent by half if one wrongly applies a Newtonian-photon argument).
Brans–Dicke scalar–tensor	$\gamma=\dfrac{1+\omega}{2+\omega},\ \beta=1$	A scalar field dilutes the spatial curvature; $\gamma\to1$ as the coupling $\omega\to\infty$ , so Cassini's bound forces $\omega\gtrsim4\times10^4$ .
Fully conservative metric theory	$\xi=\alpha_1=\alpha_2=\alpha_3=\zeta_i=0$	No preferred-frame or non-conservation effects; the eight remaining PPN parameters vanish and only $\gamma,\beta$ survive as meaningful deviations.
Strong-field regime	expansion in $U$ breaks down	Near neutron stars and black holes $U\sim1$ , the post-Newtonian series fails, and $\gamma,\beta$ no longer capture the physics — pulsar timing and merger waveforms take over.

Common trapReading gamma off a metric in the wrong coordinates

The numerical values of $\gamma$ and $\beta$ are coordinate-dependent, and the canonical definitions are tied to isotropic coordinates with the standard PPN normalization of the potential $U=GM/(rc^2)$ . If you take the Schwarzschild metric in areal (Schwarzschild) coordinates — where $g_{rr}=(1-2GM/c^2\rho)^{-1}$ — and try to read $\gamma$ from it directly, you get the wrong answer, because the areal radius is not the isotropic radius. The very factor that doubles the GR light deflection relative to the naive Newtonian value is the $\gamma$ contribution from the spatial metric, and it is precisely this term that a coordinate-careless treatment mangles. Always transform to isotropic coordinates and expand in $U$ before matching coefficients.

Worked examples

1 · Deflection of starlight at the solar limb (confidence builder)

A light ray from a distant star just grazes the edge of the Sun. Using $GM_\odot/c^2=1.477\ \text{km}$ and $R_\odot=6.96\times10^5\ \text{km}$ , find the deflection angle in general relativity, and state what a measured value of $1.61''$ would imply for $\gamma$ .

Step 1. For a grazing ray the impact parameter is the solar radius, $b=R_\odot$ . The GR deflection ( $\gamma=1$ ) is $\Delta\phi=4GM_\odot/(c^2R_\odot)$ . Substitute: $\Delta\phi=4\times1.477/6.96\times10^5=8.49\times10^{-6}$ radians.

Step 2. Convert to arcseconds with $1\ \text{rad}=206265''$ : $\Delta\phi=8.49\times10^{-6}\times206265=\mathbf{1.75''}$ . This is Einstein's 1915 prediction, confirmed by Eddington's 1919 eclipse expeditions.

Step 3. Now invert equation (3) for a hypothetical measurement. Writing $\Delta\phi_{\text{meas}}=\tfrac{1+\gamma}{2}(1.75'')$ , a value of $1.61''$ gives $(1+\gamma)/2=1.61/1.75=0.920$ , so $\gamma=0.84$ . A 8% low deflection would mean γ = 0.84, a 16% deviation from GR — which is why sub-arcsecond radio interferometry, reaching $\gamma-1\sim10^{-4}$ , is so much sharper than optical eclipse work.

2 · Perihelion advance of Mercury (fully worked)

Mercury has semi-major axis $a=5.791\times10^{10}\ \text{m}$ and eccentricity $e=0.2056$ ; it completes an orbit every $87.97$ days. With $GM_\odot=1.327\times10^{20}\ \text{m}^3/\text{s}^2$ and $c=2.998\times10^8\ \text{m/s}$ , compute the general-relativistic perihelion advance per century.

Step 1. In GR the prefactor $(2+2\gamma-\beta)/3=1$ , so the advance per orbit is $\Delta\varpi=6\pi GM_\odot/[c^2a(1-e^2)]$ . Compute the denominator: $c^2=8.988\times10^{16}$ , $1-e^2=0.9577$ , so $c^2a(1-e^2)=8.988\times10^{16}\times5.791\times10^{10}\times0.9577=4.984\times10^{27}$ .

Step 2. Numerator: $6\pi GM_\odot=6\times3.1416\times1.327\times10^{20}=2.501\times10^{21}$ . Per-orbit advance: $\Delta\varpi=2.501\times10^{21}/4.984\times10^{27}=5.02\times10^{-7}$ radians $=\mathbf{0.1035''}$ per orbit.

Step 3. Orbits per century: $100\times365.25\ \text{days}/87.97\ \text{days}=415.2$ orbits. Multiply: $0.1035''\times415.2=\mathbf{42.98''}$ per century. This matches the observed anomalous advance to within the measurement error — and combined with the Cassini value $\gamma=1$ , it pins $\beta-1$ to a few $\times10^{-4}$ .

3 · Turning a Cassini bound into a Brans–Dicke limit

The Cassini experiment gives $\gamma-1=(2.1\pm2.3)\times10^{-5}$ . In Brans–Dicke theory $\gamma=(1+\omega)/(2+\omega)$ . What lower bound on the coupling $\omega$ does this imply?

Step 1. Rearrange the Brans–Dicke relation: $\gamma-1=\dfrac{1+\omega}{2+\omega}-1=\dfrac{-1}{2+\omega}$ , so $|\gamma-1|=1/(2+\omega)$ .

Step 2. Use the measurement uncertainty as the ceiling, $|\gamma-1|\lesssim2.3\times10^{-5}$ . Then $2+\omega\gtrsim1/2.3\times10^{-5}=4.3\times10^4$ , giving $\mathbf{\omega\gtrsim4\times10^4}$ . A theory that looked like a serious rival to GR in the 1960s is now squeezed to within one part in $10^4$ of the GR limit $\omega\to\infty$ — the scalar field, if it exists, is almost perfectly decoupled.

Physics of the portrait

The parameter $\gamma$ is often called the Eddington parameter, and with reason: it was Arthur Eddington who first wrote the metric of the Sun as an expansion with adjustable coefficients, and who led the expedition that first weighed one of them.

Origins · 1 minArthur Eddington and how this was discovered

Arthur Eddington

1882–1944 · the parametrized expansion of the solar metric

Arthur Stanley Eddington was the astrophysicist who made general relativity a measured science rather than a mathematical one. In November 1919 he announced the results of two eclipse expeditions — his own to the island of Príncipe and a companion party to Sobral in Brazil — that measured the deflection of starlight grazing the Sun and found it consistent with Einstein's full $1.75''$ rather than the $0.87''$ of a Newtonian corpuscle. The result made Einstein world-famous overnight. But Eddington's deeper contribution to this article came three years later. In his 1923 treatise The Mathematical Theory of Relativity he wrote the Schwarzschild metric as a series in the potential with free numerical coefficients — the first "parametrized" post-Newtonian metric — precisely so that the observations could be phrased as measurements of those coefficients rather than as an all-or-nothing test of one theory. Howard Percy Robertson sharpened this in the early 1960s and Leonard Schiff independently, giving the modern $\gamma,\beta$ notation; Kenneth Nordtvedt and Clifford Will then generalized it to the full ten-parameter PPN formalism in the early 1970s and turned it into the standard tool for confronting every metric theory with solar-system data. The line from Eddington's 1919 photographic plates to the Cassini radio link runs straight through the two numbers he first isolated.

History

1915

— Einstein's field equations predict 43″ per century for Mercury and 1.75″ light deflection

1919

— Eddington's eclipse expeditions confirm the full deflection at the solar limb

1923

— Eddington writes the solar metric as a parametrized expansion in the potential

1968

— Nordtvedt derives the strong-equivalence combination 4β − γ − 3

2003

— Cassini pins γ − 1 to a few parts in 100000

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §39–40 (PPN formalism and classical tests) · Will, Theory and Experiment in Gravitational Physics, 2nd ed. (2018), Ch. 4–8 · Wald, General Relativity (1984), §6.3 · Carroll, Spacetime and Geometry (2004), §8.6 · Hartle, Gravity (2003), Ch. 10 · Schutz, A First Course in General Relativity, 2nd ed. (2009), Ch. 10. Key numerical anchors: the solar-limb deflection $4GM_\odot/(c^2R_\odot)=1.75''$ ; Mercury's anomalous advance $42.98''$ per century; and the Brans–Dicke inversion $\omega\gtrsim4\times10^4$ from $\gamma-1=(2.1\pm2.3)\times10^{-5}$ . The Cassini figure is from Bertotti, Iess & Tortora, Nature 425, 374 (2003).

Prerequisites. Parametrized Post-Newtonian (PPN) Formalism · Weak-Field Limit and Linearized Gravity · Newtonian Limit of General Relativity · Schwarzschild Geodesics and Effective Potential. Leads to. Deflection of Light by the Sun · Shapiro Time Delay · Perihelion Precession of Mercury · Geodetic Precession (de Sitter Effect).

Ask the tutor: Ask me to walk through why a photon feels the full $(1+\gamma)$ deflection while a slow planet's perihelion mixes $\gamma$ and $\beta$ , or to invert any solar-system measurement into a bound on the two parameters.

Why are only two of the ten PPN parameters nonzero in general relativity?

The ten PPN parameters are designed to catch every conceivable way a metric theory could differ from Newtonian gravity: preferred-frame effects ( $\alpha_1,\alpha_2,\alpha_3$ ), failures of momentum conservation ( $\zeta_i$ ), nonlinear superposition ( $\beta$ ), and spatial curvature ( $\gamma$ ). General relativity is a fully conservative theory with no preferred frame and no extra fields, so all the exotic parameters vanish identically, leaving only $\gamma$ and $\beta$ — and even those are fixed to unity. The eight zeros are as much a prediction of GR as the two ones, and experiments bound them all; the point of the formalism is that a single set of solar-system data constrains the whole slate at once.

What is the difference between gamma and beta in one sentence each?

$\gamma$ answers "how much space curvature does a unit of mass produce?" — it lives in the spatial metric, is felt fully by fast-moving light, and controls deflection and the Shapiro delay. $\beta$ answers "how strongly does gravity gravitate?" — it lives at second order in the time-time metric, is felt only by slow massive bodies, and controls the relativistic part of perihelion advance and strong-equivalence violation. The clean division of labour — $\gamma$ for null tests, $\beta$ for the nonlinear timelike tests — is what makes the pair so useful.

Why does light see gamma but not beta?

A photon moves at $c$ , so along its path the spatial displacement and the time displacement are of the same order; it therefore samples the spatial metric (where $\gamma$ lives) at the same order as $g_{00}$ . The parameter $\beta$ , by contrast, sits in the $U^2$ term of $g_{00}$ , which is second order in the potential and drops out of the null condition to the order that matters for deflection. A slow planet is the opposite case: its velocity is small ( $v^2\sim U$ ), so it barely feels the spatial curvature but does feel $g_{00}$ to second order, which is why its orbit depends on both parameters.

Why must you use isotropic coordinates to read off the parameters?

Metric components are not coordinate-invariant, so the numerical coefficients in $g_{00}$ and $g_{ij}$ depend on which radial coordinate you use. The canonical PPN definitions are stated in isotropic coordinates, where the spatial metric is a single conformal factor times flat space and the potential is normalized as $U=GM/(rc^2)$ . In Schwarzschild (areal) coordinates the spatial metric looks completely different, and reading $\gamma$ from it naively gives a wrong number. The fix is always the same: transform to isotropic coordinates, expand in $U$ , then match coefficients — as Derivation 1 does.

How does Brans–Dicke theory shift gamma, and why is it now disfavoured?

Brans–Dicke theory adds a scalar field alongside the metric; the scalar carries part of the gravitational interaction and effectively dilutes the spatial curvature, giving $\gamma=(1+\omega)/(2+\omega)<1$ while leaving $\beta=1$ . The single coupling constant $\omega$ tunes how strongly the scalar participates: small $\omega$ means a large deviation, and $\omega\to\infty$ recovers general relativity. The Cassini bound $\gamma-1=(2.1\pm2.3)\times10^{-5}$ forces $\omega\gtrsim4\times10^4$ , meaning any Brans–Dicke scalar must be so weakly coupled as to be observationally indistinguishable from GR in the solar system — which is why the theory, once a leading rival, is now heavily constrained.

Is the Nordtvedt effect a test of beta or of gamma?

Both, in a fixed combination. The Nordtvedt parameter is $\eta_{\text{N}}=4\beta-\gamma-3$ , which measures whether a self-gravitating body's gravitational binding energy contributes equally to its inertial and passive gravitational mass — the strong equivalence principle. Lunar laser ranging searches for a polarization of the Moon's orbit toward the Sun that a nonzero $\eta_{\text{N}}$ would cause, and finds none at the $10^{-4}$ level. Because $\gamma$ is already pinned near unity by Cassini, the null Nordtvedt result translates almost directly into a bound on $\beta-1$ , making it one of the two primary constraints on gravitational nonlinearity alongside Mercury's perihelion.