Perihelion Precession of Mercury

A planet on a strict inverse-square leash traces a closed ellipse forever: the perihelion — the point of closest approach to the Sun — returns to the same place in the sky every orbit. That perfect closure is not luck. It is a hidden symmetry of the $1/r$ potential, the conservation of the Laplace–Runge–Lenz vector, and it is fragile: bend the force law even slightly away from inverse-square and the ellipse stops closing. Its long axis begins to pivot, orbit after orbit, and the perihelion creeps around the Sun. By the mid-nineteenth century astronomers had accounted, to exquisite precision, for the pivoting that the other planets induce on Mercury through their Newtonian tugs — and a stubborn residue was left over. Urbain Le Verrier announced that residue in 1859: after every known perturbation was subtracted, Mercury's perihelion still advanced by roughly $43$ arcseconds per century that Newton could not explain. For half a century the anomaly resisted every fix — a hypothetical inner planet named Vulcan, a tweaked force law, an oblate Sun. Then in November 1915, before he had even written down the final field equations, Albert Einstein computed the extra advance from his new theory of gravity and got $43''$ per century with no adjustable parameter — a result he said gave him heart palpitations. This article rebuilds that calculation from the Schwarzschild geodesic equation, shows exactly which term uncloses the orbit, and carries the number to Mercury and beyond — to the inner planets, binary pulsars, and the PPN parameters that make this one of the sharpest tests of general relativity we have.

Key takeaways

A test body orbiting a Schwarzschild mass moves in an effective potential that carries, beyond the Newtonian $-GM/r$ and centrifugal $\ell^2/2r^2$ terms, a purely relativistic attraction $-GM\ell^2/c^2 r^3$ that breaks the closure of Keplerian ellipses.
To first post-Newtonian order the orbit is a slowly rotating ellipse whose major axis advances by $\Delta\phi = 6\pi GM/c^2 a(1-e^2)$ radians per revolution.
For Mercury this is $5.0\times10^{-7}$ rad per orbit, which over about $415$ orbits accumulates to $42.98''$ per century — precisely the residual Le Verrier found.
It is a weak-field, slow-motion effect of order $GM_\odot/c^2 r \sim 10^{-8}$ , not a high-speed phenomenon; it is detectable only because it is secular, building up without bound.
In the PPN framework the advance carries a factor $(2-\beta+2\gamma)/3$ , so the measured precession of Mercury directly constrains the metric parameters $\beta$ and $\gamma$ , which equal $1$ in general relativity.
Key idea: The $-1/r^3$ term in the Schwarzschild effective potential detunes the radial oscillation frequency from the orbital frequency, so the orbit traces a slowly rotating ellipse advancing by $6\pi GM/c^2 a(1-e^2)$ per revolution.

Where this lives today

Mercury's precession has aged from a triumphant one-off into a workhorse of precision gravity. Modern Solar System ephemerides — JPL's DE series, the French INPOP, the Russian EPM — fit decades of radar ranging and spacecraft tracking (Mariner, MESSENGER, and now BepiColombo at Mercury) to the metre level, and the relativistic advance is a mandatory ingredient: omit it and Mercury's predicted position drifts by tens of thousands of kilometres over a century. Turned around, these fits measure the PPN coefficient in equation (6) and bound $|\beta-1|$ and $|\gamma-1|$ at the $10^{-4}$ – $10^{-5}$ level, ruling out broad classes of alternative gravity theories; the cleanest handle on $\gamma$ is Cassini's Shapiro-delay measurement. The one nuisance is the Sun's quadrupole moment $J_2$ , which adds a competing Newtonian precession that must be modelled out — BepiColombo is expected to pin down $J_2$ and the relativistic term simultaneously. At the strong-field extreme the same formula governs binary pulsars: the Hulse–Taylor system PSR B1913+16 has a periastron advancing $4.2^\circ$ per year — about $35{,}000$ times Mercury's rate — and measuring it, with the other post-Keplerian parameters, weighs both neutron stars and overdetermines the system, delivering some of the most stringent confirmations of general relativity ever made, including the indirect detection of gravitational-wave energy loss.

The physical picture

Bound motion in any central potential is a two-clock problem. One clock is the radial oscillation: the planet falls in to perihelion, swings out to aphelion, and returns — one full radial period. The other is the angular sweep through $2\pi$ of azimuth around the Sun. In a strict $1/r$ potential these clocks are locked in exact synchrony — the radial period is precisely $2\pi$ of azimuth — so the perihelion recurs at the same angle and the ellipse closes. That lock is the geometric face of the conserved Runge–Lenz vector, special to the inverse-square law (and, curiously, the harmonic $r^2$ law). Nothing else closes. The Schwarzschild correction is a whisper of extra attraction, $-GM\ell^2/c^2 r^3$ , that bites hardest near perihelion where $r$ is smallest. It makes the inward pull a touch stronger than inverse-square just where the planet moves fastest, so it reaches its inner turning point a hair early in angle. The radial clock now runs slightly slow relative to the angular one: a radial period takes a little more than $2\pi$ of azimuth, so the perihelion recurs not at the same place but drifted forward by a tiny angle $\Delta\phi$ . Repeat every orbit and the ellipse becomes a rosette, its axis pivoting steadily about the focus.

Figure

Mercury's orbit is a slowly rotating ellipse: each revolution the perihelion advances by Δφ = 5.0×10⁻⁷ rad, drawn here exaggerated to 35° so the pivot is visible. Over about 415 orbits per century the drift accumulates to 43 arcseconds — the anomaly general relativity explains

The size of the effect is set by how deep the planet plunges, not by how fast it goes. The advance carries a factor $1/[a(1-e^2)]$ , the inverse semi-latus rectum — essentially the closest-approach distance: a small, eccentric orbit dives near the Sun where the $-1/r^3$ term is strongest. Mercury — innermost, eccentricity $e=0.206$ — is the Solar System's champion. Yet the correction is tiny at each pass: the dimensionless strength of gravity at Mercury's orbit is $GM_\odot/c^2 a \approx 10^{-8}$ , and the per-orbit advance is a microradian-scale $5\times10^{-7}$ rad. What makes it observable is that it is secular — it does not average away but adds up, orbit upon orbit, into a measurable $43''$ per century. The second figure shows where the extra term lives: in the shape of the effective potential.

Figure

The Schwarzschild effective potential adds a relativistic −GMℓ²/c²r³ term to the Newtonian well: at large r the two curves agree, but at small r the extra pull opens a bottomless pit that detunes the radial oscillation. For Mercury the fractional correction GM/c²r is only of order 10⁻⁸, yet it is enough to unclose the orbit

The core equations

The whole calculation flows from one object: the effective potential that governs radial motion in the Schwarzschild geometry. It looks like the Newtonian potential of a planet with angular momentum $\ell$ , plus one extra term — and that extra term is the entire story.

V_{\mathrm{eff}}(r) = -\frac{GM}{r} + \frac{\ell^2}{2r^2} - \frac{GM\ell^2}{c^2 r^3} \qquad (1)

where: $G$ — Newton's gravitational constant; $M$ — mass of the central body (the Sun); $r$ — Schwarzschild radial coordinate; $\ell$ — conserved angular momentum per unit rest mass, $\ell = r^2\,\mathrm{d}\phi/\mathrm{d}\tau$ ; $c$ — speed of light. Reading it: the first two terms are exactly Newtonian, and the third, scaling as $1/c^2$ , is the relativistic correction that vanishes when gravity is switched off. It is cleaner to track the shape of the orbit than its motion in time, so we set $u = 1/r$ and use azimuth $\phi$ as the independent variable. The geodesic equations then collapse to a single ODE — Newton's linear equation plus one nonlinear term.

\frac{\mathrm{d}^2 u}{\mathrm{d}\phi^2} + u = \frac{GM}{\ell^2} + \frac{3GM}{c^2}u^2 \qquad (2)

where: $u$ — inverse radius, $u=1/r$ ; $\phi$ — azimuthal angle; $GM/\ell^2$ — the constant Newtonian source term (the inverse semi-latus rectum); $3GM u^2/c^2$ — the relativistic nonlinear term. Reading it: drop the last term and this is the equation of a closed ellipse; keep it and the orbit precesses. With the last term switched off, equation (2) is a driven harmonic oscillator in $\phi$ whose solution is the familiar conic section. This is the zeroth-order Kepler orbit we perturb around.

u_0(\phi) = \frac{GM}{\ell^2}\left(1 + e\cos\phi\right) \qquad (3)

where: $e$ — orbital eccentricity; $u_0$ — the Newtonian solution, a closed ellipse with $r$ minimal (perihelion) at $\phi=0$ . Reading it: $\cos\phi$ has period exactly $2\pi$ , so the orbit closes — perihelion always returns at $\phi = 2\pi, 4\pi, \dots$ . Feeding $u_0$ back into the small relativistic term and keeping the piece that resonates with the natural frequency yields the headline result — the angle by which the perihelion runs forward each revolution.

\Delta\phi = \frac{6\pi GM}{c^2\, a(1-e^2)}\quad\text{per revolution} \qquad (4)

where: $a$ — semi-major axis; $a(1-e^2)$ — the semi-latus rectum $p=\ell^2/GM$ ; $\Delta\phi$ — forward shift of the perihelion per complete orbit, in radians. Reading it: the advance is largest for small, eccentric orbits, since $a(1-e^2)$ is the smallest length in the problem. To compare with observation we want a rate — arcseconds per century — so we divide the per-orbit advance by the period and use Kepler's third law $n^2 a^3 = GM$ .

\dot\varpi = \frac{n}{2\pi}\,\Delta\phi = \frac{3\,n\,GM}{c^2\,a(1-e^2)} \qquad (5)

where: $n=2\pi/T$ — mean orbital angular frequency, $T$ the period; $\dot\varpi$ — secular precession rate of the perihelion, radians per unit time. Reading it: the rate grows as $n\,a^{-1}\propto a^{-5/2}$ , so hugging the Sun is doubly rewarded. Alternative metric theories change the coefficient but not the form. The parametrized post-Newtonian (PPN) generalization packages every such theory into two numbers, $\beta$ and $\gamma$ ; general relativity is the point $\beta=\gamma=1$ .

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

where: $\beta$ — PPN measure of the nonlinearity of gravity's superposition ( $\beta=1$ in GR); $\gamma$ — PPN measure of space curvature produced per unit mass ( $\gamma=1$ in GR). Reading it: set $\beta=\gamma=1$ and the prefactor is exactly $1$ , restoring the pure $6\pi$ ; measuring the prefactor tests the metric. Putting Mercury's orbital elements and the Sun's mass into equation (4) and accumulating over a century gives the number that made history.

\Delta\phi_{\mathrm{Merc}} = 42.98''\ \text{per century} \qquad (7)

where: $''$ — arcseconds; $42.98$ is the canonical GR value, matching the observed residual $\approx 43.1''\pm 0.5''$ . Reading it: this single number, with no free parameters, is the first quantitative victory of general relativity.

Derivation AFrom Schwarzschild geodesics to the orbit equation

The goal of this derivation is to show where the anomalous $-1/r^3$ term physically comes from, and to reduce the full geodesic motion of a planet in curved spacetime to the single ordinary differential equation (2) for the orbit shape $u(\phi)$ . We start from the two conserved quantities guaranteed by the symmetries of the Schwarzschild metric, impose the normalization of the four-velocity, and change variables — bookkeeping that reveals the extra pull hiding inside the geometry.

Step 1. The Schwarzschild metric is static and axisymmetric, so its timelike and azimuthal Killing vectors each give a constant of the motion: the energy $E=(1-2GM/c^2 r)\,c^2\,\mathrm{d}t/\mathrm{d}\tau$ and the angular momentum $\ell=r^2\,\mathrm{d}\phi/\mathrm{d}\tau$ , per unit mass. Why: two conserved quantities cut the problem down to the $(t,r,\phi)$ subspace, just as energy and angular momentum do in Newtonian central-force motion.

Step 2. Confine the orbit to the equatorial plane $\theta=\pi/2$ — legitimate by spherical symmetry — and impose the timelike normalization $g_{\mu\nu}\,\dot x^\mu \dot x^\nu = -c^2$ on the four-velocity. Why: the normalization is the relativistic replacement for "kinetic plus potential energy is fixed", the single scalar constraint tying $E$ and $\ell$ back to the radial motion.

Step 3. Substituting $E$ and $\ell$ into the normalization and clearing the metric factors yields the first integral $\tfrac12(\mathrm{d}r/\mathrm{d}\tau)^2 + V_{\mathrm{eff}}(r) = (E^2-c^4)/2c^2$ , whose $r$ -dependence is exactly $V_{\mathrm{eff}}$ of equation (1). The metric's $1-2GM/c^2 r$ factor multiplied against the centrifugal $\ell^2/r^2$ manufactures the new term $-GM\ell^2/c^2 r^3$ . Why: the relativistic term is not an add-on force but the algebraic residue of curved-space geometry acting on the angular-momentum barrier — which is why it disappears the instant $c\to\infty$ .

Step 4. Change the independent variable from proper time $\tau$ to azimuth $\phi$ using $\mathrm{d}/\mathrm{d}\tau = (\ell/r^2)\,\mathrm{d}/\mathrm{d}\phi$ and substitute $u=1/r$ ; differentiating once leaves $u'' + u = GM/\ell^2 + (3GM/c^2)u^2$ , equation (2). Why: a closed ellipse is a pure cosine in $u$ , so this casting turns "does the orbit close?" into the sharp question "does the solution have period exactly $2\pi$ ?"

Go deeper · Derivation B · 2 minSolving the orbit equation for the perihelion advance

Derivation BSolving the orbit equation for the perihelion advance

The goal here is to extract the closed-form advance $\Delta\phi = 6\pi GM/c^2 a(1-e^2)$ from equation (2) by treating the relativistic term as a small perturbation, since its size relative to the Newtonian term is $3GM u/c^2\sim 10^{-8}$ for Mercury. The subtlety is that a naive perturbation series appears to blow up; the resolution — recognizing a resonant, secular response and resumming it into a tiny frequency shift — is the mathematical heart of the whole phenomenon.

Step 1. Take the Newtonian solution $u_0 = (GM/\ell^2)(1+e\cos\phi)$ as the zeroth approximation and insert it into the small term, giving $(3GM/c^2)u_0^2 = (3(GM)^3/c^2\ell^4)(1 + 2e\cos\phi + e^2\cos^2\phi)$ . Why: perturbation theory evaluates the small correction on the known leading solution; the three pieces that appear behave very differently, and only one of them matters.

Step 2. Sort the three driving pieces. The constant term merely renormalizes $GM/\ell^2$ (a negligible size shift), and $\cos^2\phi = \tfrac12(1+\cos2\phi)$ drives the oscillator off-resonance at frequency $2$ , a bounded ripple. The dangerous piece is $2e\cos\phi$ , which drives the oscillator at its own natural frequency $1$ . Why: an oscillator forced at its natural frequency responds resonantly, and that resonance accumulates orbit after orbit into a secular drift — the other terms cannot.

Step 3. Solve $u_1'' + u_1 = (6(GM)^3 e/c^2\ell^4)\cos\phi$ for the resonant response. Resonant forcing of a unit-frequency oscillator gives the standard growing particular solution $u_1 = (3(GM)^3 e/c^2\ell^4)\,\phi\sin\phi$ , linear in $\phi$ . Why: the factor of $\phi$ multiplying $\sin\phi$ is the signature of secular growth — the correction is not periodic but marches steadily with each revolution, which is exactly the behaviour a precessing perihelion should show.

Step 4. Add $u_0+u_1$ and recognize $e\cos\phi + \alpha\,e\,\phi\sin\phi$ as the first-order expansion of $e\cos((1-\alpha)\phi)$ with $\alpha = 3(GM)^2/c^2\ell^2$ . So $u \approx (GM/\ell^2)[1 + e\cos((1-\alpha)\phi)]$ — still a cosine, but with its argument slowed by $(1-\alpha)$ . Why: resumming the runaway secular term into a shifted frequency is the key move — the "blow-up" was an artifact of expanding a slightly detuned cosine; the orbit is perfectly well-behaved, just precessing.

Step 5. The perihelion is where $u$ is maximal, i.e. $(1-\alpha)\phi = 2\pi$ , so it recurs at $\phi \approx 2\pi(1+\alpha)$ instead of $2\pi$ . The excess per orbit is $\Delta\phi = 2\pi\alpha = 6\pi(GM)^2/c^2\ell^2$ , and with $\ell^2/GM = a(1-e^2)$ this is $\Delta\phi = 6\pi GM/c^2 a(1-e^2)$ . Why: the entire effect is the detuning $\alpha$ — the radial oscillation takes slightly more than $2\pi$ of azimuth, and $2\pi\alpha$ is the leftover angle the ellipse has turned. This reproduces MTW Eq. (40.17) and Carroll Eq. (5.78).

Go deeper · Derivation C · 1 minFrom per-orbit advance to a secular rate and the PPN test

Derivation CFrom per-orbit advance to a secular rate and the PPN test

The goal of this derivation is to convert the geometric per-orbit advance into the observable astronomers report — a precession rate in arcseconds per century — and then to widen the result into the PPN form (6) that turns Mercury into a laboratory for testing gravity theories, connecting the textbook formula to both the historical $43''$ figure and the modern precision-gravity programme.

Step 1. The perihelion advances by $\Delta\phi$ every orbital period $T$ , so the average precession rate is $\dot\varpi = \Delta\phi/T = (n/2\pi)\Delta\phi$ with $n=2\pi/T$ the mean motion. Substituting equation (4) gives $\dot\varpi = 3nGM/c^2 a(1-e^2)$ . Why: observations span many orbits, so the natural quantity is a rate, not a per-orbit angle; dividing by $T$ is the bridge between the two.

Step 2. Use Kepler's third law $n^2 a^3 = GM$ to trade $a$ for $n$ , exposing the scaling $\dot\varpi \propto n^{5/3}(GM)^{2/3}/(1-e^2)$ . Why: written this way the rate depends only on the observable period, eccentricity, and central mass — how ephemeris fits parametrize it — and it shows why short-period, eccentric systems precess fastest.

Step 3. Replace the Schwarzschild potential by its PPN generalization, whose metric coefficients carry $\gamma$ (space curvature per unit mass) and $\beta$ (nonlinearity of superposition). Re-running the orbit calculation multiplies the advance by $(2-\beta+2\gamma)/3$ , giving equation (6). Why: every metric theory reduces, in the weak field, to this two-parameter family, so the measured coefficient is a clean discriminator — and general relativity's prediction is exactly $1$ .

Step 4. Decompose the GR coefficient: about two-thirds comes from the $\gamma$ (space-curvature) sector and the rest from $\beta$ and nonlinear time dilation, while the purely special-relativistic velocity correction alone would give only about one-sixth. Why: the precession is dominantly a statement about curved space, not time dilation or fast motion — a common misreading — which is why Mercury constrains $\gamma$ and $\beta$ in a specific, checkable combination.

Limiting cases

The advance $\Delta\phi = 6\pi GM/c^2 a(1-e^2)$ is compact enough to interrogate directly. Its behaviour in the important limits confirms it reduces to known physics and shows where the effect is amplified.

Case	Result	Meaning
Newtonian, $c\to\infty$	$\Delta\phi \to 0$	The $-1/r^3$ term scales as $1/c^2$ and vanishes; the orbit closes, recovering the Kepler ellipse with its conserved Runge–Lenz vector.
Circular orbit, $e\to 0$	$\Delta\phi = \dfrac{6\pi GM}{c^2 a}$	The advance is smallest for a given $a$ ; with no eccentricity there is no "perihelion" to track, but the frequency detuning $\alpha$ is still nonzero.
Highly eccentric, $e\to 1$	$\Delta\phi = \dfrac{6\pi GM}{c^2 a(1-e^2)}\to\infty$	The semi-latus rectum shrinks and the orbit skims the mass at perihelion, amplifying the effect — the regime of tight binary pulsars.
PPN theories	$\Delta\phi = \dfrac{2-\beta+2\gamma}{3}\dfrac{6\pi GM}{c^2 a(1-e^2)}$	GR ( $\beta=\gamma=1$ ) restores the factor $6\pi$ ; a measured deviation would falsify GR in favour of an alternative metric theory.

Common trapThe 43 arcseconds is the whole precession of Mercury

It is not. Mercury's perihelion advances by roughly $5600''$ per century in total as seen from Earth. The bulk of that — about $5025''$ — is not Mercury moving at all but the precession of Earth's own equinoxes, a coordinate effect. Of the dynamical remainder, about $532''$ comes from the Newtonian tugs of the other planets, chiefly Venus, Jupiter and Earth. Only after every classical contribution is computed and subtracted does a residue of about $43''$ per century survive — and that is what general relativity explains, with no adjustable parameter. Quoting the raw $5600''$ (or even the $532''$ planetary part) as the relativistic prediction is a classic error. A second trap: $\Delta\phi$ in equation (4) is per orbit, not per year — multiply by the orbits per century (about $415$ for Mercury) and convert with $1\ \text{rad}=206265''$ .

Worked examples

1 · Mercury's 43 arcseconds, from scratch (fully worked)

Using $GM_\odot = 1.327\times10^{20}\ \mathrm{m^3\,s^{-2}}$ , $a = 5.79\times10^{10}\ \mathrm{m}$ , $e = 0.2056$ , and orbital period $T = 0.2408$ yr, verify the canonical figure.

Step 1. Form the semi-latus rectum, the length that controls the advance: $p = a(1-e^2) = 5.79\times10^{10}\,(1-0.04227) = 5.545\times10^{10}\ \mathrm{m}$ . This is the closest-approach scale that the $-1/r^3$ term cares about.

Step 2. Evaluate the per-orbit advance from equation (4), all in SI with $c^2 = 8.988\times10^{16}\ \mathrm{m^2\,s^{-2}}$ : $\Delta\phi = \dfrac{6\pi GM_\odot}{c^2 p} = \dfrac{6\pi(1.327\times10^{20})}{(8.988\times10^{16})(5.545\times10^{10})}$ . The numerator is $2.501\times10^{21}$ and the denominator $4.984\times10^{27}$ , so $\Delta\phi = 5.02\times10^{-7}$ rad per orbit — half a microradian, utterly invisible in a single pass.

Step 3. Count the orbits per century: $N = 100\ \mathrm{yr}/0.2408\ \mathrm{yr} = 415.3$ . Mercury laps the Sun about $415$ times per Julian century.

Step 4. Accumulate and convert to arcseconds using $1\ \mathrm{rad} = 206265''$ : $\Delta\phi_{\mathrm{century}} = N\,\Delta\phi\,(206265) = 415.3\times 5.02\times10^{-7}\times 2.06265\times10^5$ . This gives $\Delta\phi_{\mathrm{century}} \approx 43.0''$ per century, matching the canonical $42.98''$ and the observed residual $43.1''\pm0.5''$ within uncertainty. The relativistic geometry, with no fitted parameter, lands squarely on Le Verrier's leftover.

2 · Why Earth precesses less: the scaling in action

General relativity predicts a perihelion advance for every planet. Estimate it for Earth ( $a = 1.496\times10^{11}\ \mathrm{m}$ , $e = 0.0167$ , $T = 1$ yr) and explain why it is smaller than Mercury's despite Earth completing more orbits per century.

Step 1. Semi-latus rectum: $p = a(1-e^2) = 1.496\times10^{11}(1-0.000279) = 1.496\times10^{11}\ \mathrm{m}$ — Earth's near-circular orbit means $p$ is essentially $a$ , and $a$ itself is about $2.6$ times Mercury's.

Step 2. Per-orbit advance: $\Delta\phi = 6\pi GM_\odot/c^2 p = 2.501\times10^{21}/(8.988\times10^{16}\times1.496\times10^{11}) = 1.86\times10^{-7}$ rad, roughly a third of Mercury's $5.02\times10^{-7}$ rad, because $p$ is larger.

Step 3. Orbits per century: $N = 100$ . Accumulating, $\Delta\phi_{\mathrm{century}} = 100\times1.86\times10^{-7}\times206265 \approx \mathbf{3.8''}$ per century, in agreement with the standard GR value for Earth of about $3.84''$ /century.

Step 4. Read off the lesson: Earth's per-orbit advance is smaller and it makes only a quarter as many orbits per century, so its secular rate is far below Mercury's. The rate scales as $\dot\varpi\propto n/a \propto a^{-5/2}$ , and low eccentricity offers no help — Mercury wins on both counts, being innermost and among the most eccentric planets. This is precisely why Mercury, not Earth, was the decisive test.

Physics of the portrait

The perihelion of Mercury is a story of two computers separated by half a century: one who found the anomaly by heroic hand-calculation and mistakenly chased a phantom planet to explain it, and one who dissolved it in three weeks of new geometry.

Origins · 1 minAlbert Einstein and how this was discovered

Albert Einstein

1879–1955 · the November 1915 relativistic calculation of Mercury's perihelion advance

By 1915 the anomaly was over fifty years old. Urbain Le Verrier — who had discovered Neptune in 1846 by the pure force of perturbation theory, predicting where an unseen planet must lie and being proved right within a degree — turned the same machinery on Mercury and announced in 1859 that its perihelion advanced some $38''$ (later refined by Simon Newcomb to $43''$ ) per century faster than Newton allowed. Trusting the method that had won him Neptune, Le Verrier proposed a second unseen planet, Vulcan, inside Mercury's orbit. Astronomers hunted it for decades and never found it. Others tried modifying the force law, or invoking an oblate Sun — each fix ad hoc. When Einstein reached the problem in November 1915 he was still finishing general relativity; the field equations took their final form only that same month. Using an early approximate solution (Schwarzschild's exact metric came weeks later), he computed the extra advance and found $43''$ per century, dead on, with not a single free parameter. He later said the discovery left him unable to work for days — heart palpitations. It was the moment the theory stopped being mathematics and became a description of the world. Unlike Vulcan, general relativity added nothing to the Solar System; it changed the geometry the planets already moved through.

History

1846

— Le Verrier predicts Neptune from Uranus's anomalies, and it is found within a degree

1859

— Le Verrier reports Mercury's unexplained perihelion advance and proposes the planet Vulcan

1882

— Newcomb refines the anomalous residual to about 43 arcseconds per century

1915

— Einstein computes 43 arcseconds per century from general relativity, with no free parameter

1916

— Schwarzschild publishes the exact vacuum solution underlying the effective potential

1974

— Hulse and Taylor discover the binary pulsar, whose periastron advances 35000 times faster

Going further

Texts. Misner, Thorne & Wheeler, Gravitation, §25.5 and §40.5 · Wald, General Relativity, §6.3 · Carroll, Spacetime and Geometry, §5.4 (Eq. 5.78) · Hartle, Gravity, Ch. 9 · Schutz, A First Course in General Relativity, §11.3 · Weinberg, Gravitation and Cosmology, §8.6 · Will, Theory and Experiment in Gravitational Physics, Ch. 7. The orbit equation $u''+u = GM/\ell^2 + 3GMu^2/c^2$ and its resonant solution giving $\Delta\phi = 6\pi GM/c^2 a(1-e^2)$ (see also MTW Eq. 40.17) are derived from the effective potential (1) and (2) above. The standard heliocentric numerical inputs used throughout are: $GM_\odot = 1.327\times10^{20}\ \mathrm{m^3\,s^{-2}}$ , $a = 5.79\times10^{10}\ \mathrm{m}$ , $e = 0.2056$ , $T = 0.2408$ yr, $c = 2.998\times10^8\ \mathrm{m/s}$ , $1\ \mathrm{rad}=206265''$ ; these yield $42.98''$ /century. The Earth cross-check ( $\approx 3.8''$ /century) uses $a = 1.496\times10^{11}\ \mathrm{m}$ , $e=0.0167$ , $T=1$ yr. The PPN coefficient $(2-\beta+2\gamma)/3$ appears in Will, Theory and Experiment in Gravitational Physics, Ch. 7.

Prerequisites. Schwarzschild Geodesics and the Effective Potential · Newtonian Limit of General Relativity · Weak-Field Limit and Linearized Gravity. Leads to. Parametrized Post-Newtonian (PPN) Formalism · PPN Parameters Beta and Gamma · Geodetic Precession (de Sitter Effect) · Frame Dragging and the Lense–Thirring Effect.

Ask the tutor: Ask me why only the resonant $\cos\phi$ term drives the precession, or to recompute the advance for Venus, Mars, or a binary pulsar you choose.

Is Mercury's precession a fast-motion, special-relativistic effect?

No, and this is the most common misconception. Mercury's orbital speed is only about $48\ \mathrm{km/s}$ , so $v/c\sim1.6\times10^{-4}$ — hardly relativistic. The advance is a weak-field, first post-Newtonian correction of order $GM_\odot/c^2 r\sim10^{-8}$ , driven by the curvature of space near the Sun, not by the planet's velocity. Keeping only the special-relativistic kinematic correction recovers about one-sixth of the full advance; the remaining five-sixths come from spatial curvature and the nonlinear metric. The effect is subtle, not fast.

Why does a strict inverse-square force give no precession at all?

Because the $1/r$ potential has a hidden symmetry beyond energy and angular momentum: it conserves the Laplace–Runge–Lenz vector, which points along the major axis and is fixed in space. A conserved axis means the ellipse cannot rotate — it must close. Equivalently, the radial period equals precisely $2\pi$ of azimuth, so perihelion recurs at the same angle each orbit. The $-1/r^3$ term breaks this symmetry, detuning the radial frequency from the orbital frequency by a factor $(1-\alpha)$ , so the axis pivots and the perihelion advances by $2\pi\alpha$ per orbit. Only the inverse-square (and harmonic) law close; almost any other precesses.

Where exactly does the extra −1/r³ term come from?

It is a geometric residue, not a new force. When you write the timelike normalization of a planet's four-velocity in the Schwarzschild metric and substitute the conserved $E$ and $\ell$ , the metric factor $1-2GM/c^2 r$ multiplies the centrifugal term $\ell^2/r^2$ , generating a cross term $-GM\ell^2/c^2 r^3$ on top of the Newtonian $-GM/r$ and $\ell^2/2r^2$ . It scales as $1/c^2$ , so it vanishes as $c\to\infty$ . Physically, in curved space the angular-momentum barrier is slightly weaker close in than Newton assumes, letting the planet reach perihelion a touch sooner in angle each orbit.

How big is the total observed precession compared with the 43 arcseconds?

Mercury's perihelion appears to advance about $5600''$ per century from Earth. But roughly $5025''$ of that is the precession of Earth's equinoxes — a change in our reference frame, not in Mercury. Of the dynamical part, about $532''$ comes from Newtonian perturbations by the other planets. Subtract all of these and a residual of about $43''$ per century remains, and that is what general relativity accounts for exactly. So the relativistic effect is under one percent of the raw observed number — the triumph lay in computing the classical parts precisely enough to expose it.

Does the same formula apply to binary pulsars, and why is their effect so large?

Yes — the periastron of a binary pulsar advances by the same $6\pi GM/c^2 a(1-e^2)$ per orbit, with $M$ the total system mass. The effect is enormous because the ingredients are pushed to extremes: near-solar masses in objects kilometres across, orbital periods of hours not months, and high eccentricity. For the Hulse–Taylor pulsar the periastron advances about $4.2^\circ$ per year, some $35{,}000$ times Mercury's rate. Measuring it, with other post-Keplerian parameters, weighs both neutron stars and tests general relativity in the strong field, far beyond the Solar System's gentle regime.

Did Einstein fit any parameter to get 43 arcseconds?

No, and that is the point. The number $42.98''$ per century falls straight out of the Schwarzschild geometry once you insert Mercury's known orbital elements and the Sun's mass — quantities measured independently by astronomers. There is nothing to adjust: no Vulcan to place, no force-law exponent to tune, no solar oblateness to invent. This distinguished general relativity from every earlier attempt, all of which required a free ingredient chosen to fit. A theory that predicts a measured number with zero free parameters is doing something a curve-fit cannot — which is why the result convinced Einstein, and over the following decade most physicists.