Gravitational Two-Body Problem

The gravitational two-body problem asks a deceptively small question — how do two point masses move under nothing but their mutual Newtonian pull? — and returns the founding exact result of celestial mechanics. Everything a first course later treats as given descends from it: that binary stars, planets, moons, and comets trace conic sections; that an orbit's period fixes the total mass; that a close stellar encounter is a hyperbola. That the problem is exactly solvable at all is a small miracle — add a third body of comparable mass and integrability is generically lost, the orbits turn chaotic, and only statistics survive. The whole solution turns on one change of variables. Two position vectors carry six degrees of freedom; split them into the center-of-mass position $\mathbf{R}$ , which drifts uniformly because no external force acts, and the relative separation $\mathbf{r}=\mathbf{r}_1-\mathbf{r}_2$ , which obeys a single one-body equation for a fictitious particle of reduced mass $\mu$ orbiting a fixed center of total mass $M$ . The genuinely two-body dance is repackaged as the one-body Kepler problem you already know how to solve. Because gravity is central, angular momentum is conserved and the orbit is planar and sweeps equal areas in equal times; energy conservation then delivers the conic orbit and the period-mass relation. All three of Kepler's empirical laws fall out of Newtonian gravity plus two conservation laws — and the same hyperbolic solutions, summed over many weak encounters, define the relaxation time that tells galactic dynamics when it may forget individual orbits and treat the system as a smooth potential.

Key takeaways

Two point masses under mutual gravity separate exactly into a freely drifting center of mass and a one-body problem for the relative vector $\mathbf{r}=\mathbf{r}_1-\mathbf{r}_2$ .
The relative motion obeys $\ddot{\mathbf{r}}=-GM\hat{\mathbf{r}}/r^2$ with the total mass $M=m_1+m_2$ ; the reduced mass $\mu=m_1m_2/M$ enters only the energy and angular momentum.
Angular-momentum conservation makes the orbit planar and equal-area (Kepler II); energy conservation makes it a conic section $r(\phi)=p/(1+e\cos\phi)$ (Kepler I).
The energy sign fixes the orbit: $E<0$ ellipse, $E=0$ parabola, $E>0$ hyperbola; and Kepler III follows exactly, $P^2=4\pi^2a^3/[G(m_1+m_2)]$ — the workhorse for weighing binaries and exoplanets.
This is the only nontrivial gravitational many-body case with a full closed-form solution; three-body and $N$ -body motion are non-integrable, which is why galactic dynamics turns statistical.
Key idea: transform to center-of-mass and relative coordinates — the CM drifts freely and the relative vector solves one Kepler problem with mass $M=m_1+m_2$ and reduced mass $\mu=m_1m_2/M$ , giving conic-section orbits and all three Kepler laws exactly.

The core insight

Transform to center-of-mass and relative coordinates: the CM drifts freely, and the relative vector solves a single Kepler problem with mass $M=m_1+m_2$ and reduced mass $\mu=m_1m_2/M$ , giving conic-section orbits and all three Kepler laws exactly.

Where this lives today

Three and a half centuries on, the two-body solution is still doing frontline work. ESA's Gaia measures stellar positions to microarcseconds, resolving the astrometric reflex wobble for thousands of stars and turning Kepler's third law into a mass census of the Galaxy and its exoplanets; the same reflex-velocity signal feeds radial-velocity surveys hunting Earth analogues. The Hulse–Taylor binary pulsar showed a two-body orbit decaying by gravitational-wave emission at exactly the general-relativistic rate — the first evidence for gravitational waves — and LIGO–Virgo now catch the final seconds of black-hole and neutron-star inspirals, where the Newtonian Kepler orbit is the leading-order skeleton onto which post-Newtonian and numerical-relativity corrections are hung. Every interplanetary trajectory is still planned as patched two-body arcs, with vis-viva sizing each burn. And in galactic dynamics the hyperbolic two-body solution, summed over countless weak stellar encounters, defines the relaxation time — the clock that tells us galaxies are effectively collisionless over the age of the Universe, licensing the smooth-potential methods the rest of stellar dynamics is built on.

The physical picture

Picture two stars joined by an invisible attractive tether. Sit at their balance point — the center of mass — and you never move: the whole system glides through space at constant velocity, because the internal gravitational forces are equal and opposite and cancel in the sum. Every bit of the interesting physics lives in one vector, the separation $\mathbf{r}$ pointing from one body to the other. Nature has quietly turned a symmetric two-body dance into an asymmetric one-body problem, a fictitious particle of reduced mass $\mu$ swinging around a fixed center whose "mass" is the sum $M=m_1+m_2$ . The two real bodies each trace a scaled copy of that relative orbit about the common center of mass, and the heavier body traces the smaller copy.

Because gravity points along the line joining the bodies, it exerts no torque about the center of mass, so the relative angular momentum is frozen: the orbit is locked in one plane and sweeps equal areas in equal times — the same conservation that makes a skater spin faster as she pulls her arms in. What remains, the radial motion, is a single one-dimensional problem: a particle of mass $\mu$ sliding in an effective potential — the gravitational well plus a centrifugal barrier. Whether the orbit closes into an ellipse or opens into an escaping hyperbola is then decided by one number, the total energy $E$ ; just enough to crawl to infinity with zero speed to spare is the razor's-edge parabola between capture and escape.

Figure

The effective potential U_eff(r) = L²/2μr² − GMμ/r governs the radial motion — a bound orbit with E < 0 oscillates between perihelion r_min and aphelion r_max (ellipse), E = 0 just reaches infinity (parabola), E > 0 escapes with speed to spare (hyperbola), and the minimum at r₀ = L²/GMμ² is the circular orbit with E = −GMμ/2a

The core equations

Newton's law of gravitation, applied to each body in turn, gives two coupled second-order vector equations — six scalar equations for the six position components. They are coupled because each body's acceleration depends on the other body's position through the separation $r$ — the coupling the change of variables will untie.

m_1\ddot{\mathbf{r}}_1 = -\frac{G m_1 m_2}{r^3}(\mathbf{r}_1-\mathbf{r}_2),\qquad m_2\ddot{\mathbf{r}}_2 = +\frac{G m_1 m_2}{r^3}(\mathbf{r}_1-\mathbf{r}_2) \qquad (1)

where: $\mathbf{r}_i$ — position vector of body $i$ ; $r=|\mathbf{r}_1-\mathbf{r}_2|$ — separation between the bodies; $G=6.674\times10^{-11}\ \mathrm{N\,m^2\,kg^{-2}}$ — Newton's gravitational constant; $m_i$ — mass of body $i$ . The forces are equal and opposite (Newton's third law) — the seed of every conservation law below. Adding the two equations of motion makes the internal forces cancel, so the mass-weighted mean position — the center of mass — feels no net force at all. Its acceleration vanishes identically.

\mathbf{R}=\frac{m_1\mathbf{r}_1+m_2\mathbf{r}_2}{m_1+m_2},\qquad \ddot{\mathbf{R}}=0 \qquad (2)

where: $\mathbf{R}$ — center-of-mass position; $M=m_1+m_2$ — total mass. The CM drifts at constant velocity, so we may sit in the frame where $\mathbf{R}=0$ and study only the relative orbit. Subtracting the two equations (after dividing each by its own mass) collapses the coupled pair into one equation for the relative vector $\mathbf{r}$ . The two contributions add, so the total mass appears — the single most misremembered point in the subject.

\ddot{\mathbf{r}} = -\frac{GM}{r^2}\,\hat{\mathbf{r}},\qquad M=m_1+m_2 \qquad (3)

where: $\mathbf{r}=\mathbf{r}_1-\mathbf{r}_2$ — relative separation vector; $\hat{\mathbf{r}}$ — unit vector along $\mathbf{r}$ ; $M$ — total mass, which sets the orbital dynamics. This is identical to a test particle orbiting a fixed point mass $M$ — the Kepler problem you already know. The reduced mass is the effective inertia of that fictitious particle. It does not appear in the acceleration (3), but it multiplies $\ddot{\mathbf{r}}$ to give the force and carries the energy and angular momentum.

\mu = \frac{m_1 m_2}{m_1+m_2} \qquad (4)

where: $\mu$ — reduced mass, always smaller than either body and tending to the lighter mass when $m_1\gg m_2$ . Note $\mathbf{F}=\mu\ddot{\mathbf{r}}=-GM\mu\hat{\mathbf{r}}/r^2$ , so $GM\mu=Gm_1m_2$ — the honest two-body force reappears. Because the force is central, the torque about the CM is zero and the relative angular momentum is a constant of the motion. Writing it in polar coordinates $(r,\phi)$ in the orbital plane gives Kepler's second law directly: the radius vector sweeps area at a fixed rate.

L=\mu r^2\dot\phi = \text{const},\qquad \frac{\mathrm{d}A}{\mathrm{d}t}=\frac{L}{2\mu}=\text{const} \qquad (5)

where: $L$ — magnitude of the relative orbital angular momentum; $\phi$ — azimuthal angle in the orbital plane; $\mathrm{d}A/\mathrm{d}t$ — areal velocity. Equal areas in equal times is not special to gravity — it holds for any central force. Total mechanical energy is the second constant. Using $L$ to eliminate $\dot\phi$ turns the two-dimensional problem into one-dimensional radial motion in an effective potential — the gravitational well plus a centrifugal barrier, the curve plotted above.

E=\tfrac12\mu\dot r^2 + \underbrace{\frac{L^2}{2\mu r^2} - \frac{GM\mu}{r}}_{U_{\rm eff}(r)} = \text{const} \qquad (6)

where: $E$ — total energy of the relative motion; $\tfrac12\mu\dot r^2$ — radial kinetic energy; $L^2/2\mu r^2$ — centrifugal term; $-GM\mu/r$ — gravitational potential energy. Turning points ( $\dot r=0$ ) occur where $E=U_{\rm eff}(r)$ , which is why the energy line's intersections with the curve are perihelion and aphelion. Integrating the radial equation (the derivations do it in full) yields the orbit as a conic section — Kepler's first law. The eccentricity packages $E$ and $L$ into the single number that says which conic you have.

r(\phi)=\frac{p}{1+e\cos\phi},\quad p=\frac{L^2}{GM\mu^2},\quad e=\sqrt{1+\frac{2EL^2}{G^2M^2\mu^3}},\quad a=-\frac{GM\mu}{2E} \qquad (7)

where: $p$ — semi-latus rectum; $e$ — eccentricity; $\phi$ — true anomaly from perihelion; $a$ — semi-major axis. For a bound orbit $a=p/(1-e^2)$ and the energy is fixed by size alone, $E=-GM\mu/2a$ — remarkably, independent of $L$ . Integrating the areal velocity over one orbit and inserting the ellipse geometry links period to size; every trace of $\mu$ and $L$ cancels, leaving only the total mass. This is Kepler's third law, the equation that weighs the Universe.

P^2=\frac{4\pi^2 a^3}{G(m_1+m_2)} \qquad (8)

where: $P$ — orbital period; $a$ — semi-major axis of the relative orbit; $m_1+m_2$ — total system mass. In solar units this is simply $(m_1+m_2)/M_\odot = a_{\rm AU}^3/P_{\rm yr}^2$ — the fastest mass estimate in astronomy. The energy relation gives the speed at any point of the orbit without ever solving for the time — the vis-viva ("living force") equation, the everyday tool of orbit mechanics.

v^2=GM\left(\frac{2}{r}-\frac{1}{a}\right) \qquad (9)

where: $v$ — relative speed; $r$ — instantaneous separation; $a$ — semi-major axis (positive for ellipses, negative for hyperbolae). Set $r=a$ for the mean circular speed $\sqrt{GM/a}$ ; take $a\to\infty$ for the escape speed $\sqrt{2GM/r}$ .

Derivation AReduction to an equivalent one-body problem

The goal is to show, with no approximation, that the coupled six-dimensional problem of equations (1) is exactly two independent problems — a free particle for the center of mass and a single Kepler problem for the relative vector — and to make explicit why the total mass $M=m_1+m_2$ , not the reduced mass, controls the relative acceleration. It is the structural move the whole solution rests on.

Step 1. Add the two equations of motion in (1). The right-hand sides are equal and opposite by Newton's third law, so they cancel and leave $m_1\ddot{\mathbf{r}}_1+m_2\ddot{\mathbf{r}}_2=0$ , i.e. $\ddot{\mathbf{R}}=0$ for the CM defined in (2). Why: the internal force cannot move the mean position, so the CM is a free particle and can be transformed away by choosing its rest frame.

Step 2. Now divide each equation in (1) by its own mass to isolate the accelerations: $\ddot{\mathbf{r}}_1=-Gm_2\,\mathbf{r}/r^3$ and $\ddot{\mathbf{r}}_2=+Gm_1\,\mathbf{r}/r^3$ , with $\mathbf{r}=\mathbf{r}_1-\mathbf{r}_2$ . Why: writing each body's acceleration in terms of the single separation vector is what lets the next subtraction collapse two vectors into one.

Step 3. Subtract the second from the first: $\ddot{\mathbf{r}}=\ddot{\mathbf{r}}_1-\ddot{\mathbf{r}}_2=-G(m_1+m_2)\,\mathbf{r}/r^3=-GM\hat{\mathbf{r}}/r^2$ , which is equation (3). Why: the two masses add because each body accelerates toward the other — the total mass, not $\mu$ , sets how fast the separation shrinks, and this is the step most people get backwards.

Step 4. For completeness, recover the individual motions from $\mathbf{r}$ and $\mathbf{R}$ : inverting the definitions gives $\mathbf{r}_1=\mathbf{R}+(m_2/M)\mathbf{r}$ and $\mathbf{r}_2=\mathbf{R}-(m_1/M)\mathbf{r}$ . Why: each real body traces a scaled copy of the relative orbit about the CM, with linear size $\propto$ the other body's mass — so the heavier body moves on the smaller ellipse, exactly as the first figure shows. *As a check: adding equations (1) gives $\ddot{\mathbf{R}}=0$ ; subtracting gives $\ddot{\mathbf{r}}=-GM\hat{\mathbf{r}}/r^2$ ; and the mass split $a_1/a_2=m_2/m_1$ reproduces the standard binary-star result.*

Go deeper · Derivation B · 1 minThe orbit equation and Kepler's first law

Derivation BThe orbit equation and Kepler's first law

The aim is to integrate the reduced equation (3) into the shape of the orbit, $r(\phi)$ , and prove it is always a conic section. The trick is to change both the dependent variable, to $u=1/r$ , and the independent variable, from time to angle $\phi$ , using angular-momentum conservation — a substitution that turns the nonlinear inverse-square dynamics into a linear oscillator equation whose solution is elementary.

Step 1. Conservation of $\mathbf{L}=\mu\,\mathbf{r}\times\dot{\mathbf{r}}$ forces the motion into the plane perpendicular to $\mathbf{L}$ ; use polar coordinates $(r,\phi)$ there, with $L=\mu r^2\dot\phi$ . Why: reducing three dimensions to two is free — a central force can exert no torque, so the orbital plane is fixed once and for all.

Step 2. Substitute $u=1/r$ and eliminate time with $\dot\phi=Lu^2/\mu$ ; the radial velocity becomes $\dot r=-(L/\mu)\,\mathrm{d}u/\mathrm{d}\phi$ and $\ddot r=-(L^2u^2/\mu^2)\,\mathrm{d}^2u/\mathrm{d}\phi^2$ . Why: with $u$ and $\phi$ the awkward $1/r^2$ force law becomes constant on the right-hand side, which is the whole point of the change of variables.

Step 3. Insert these into the radial component of (3), $\ddot r-r\dot\phi^2=-GM/r^2$ , and divide through; the equation reduces to the linear Binet equation $\mathrm{d}^2u/\mathrm{d}\phi^2+u=GM\mu^2/L^2$ . Why: a linear ODE with constant coefficients is the fingerprint of the inverse-square law — no other power law gives closed, non-precessing orbits.

Step 4. The general solution is $u=(GM\mu^2/L^2)(1+e\cos\phi)$ , so $r(\phi)=p/(1+e\cos\phi)$ with $p=L^2/GM\mu^2$ ; the constant $e$ is fixed by the energy through $e=\sqrt{1+2EL^2/G^2M^2\mu^3}$ . Why: this is the polar equation of a conic with focus at the center — ellipse, parabola, or hyperbola as $e<1$ , $=1$ , or $>1$ — so Kepler's elliptical orbits are a theorem, not a postulate. *As a check: substituting $e=0$ gives the circular orbit $r=p=L^2/GM\mu^2$ , which sits at the minimum of $U_{\rm eff}$ .*

Go deeper · Derivation C · 1 minKepler's third law from the swept area

Derivation CKepler's third law from the swept area

The target is the period-mass relation $P^2=4\pi^2a^3/GM$ , and the elegant route is not to integrate the motion in time but to use Kepler's second law: since area is swept at a constant rate, the period is just the total ellipse area divided by that rate. Watching $\mu$ and $L$ cancel at the end is the payoff — it is why a binary's period reveals its total mass and nothing about how it is split.

Step 1. From (5) the areal velocity is constant, $\mathrm{d}A/\mathrm{d}t=L/2\mu$ , so over one period $P$ the swept area is the whole ellipse: $A=(L/2\mu)P=\pi a b$ , with semi-minor axis $b=a\sqrt{1-e^2}$ . Why: constancy of $\mathrm{d}A/\mathrm{d}t$ is what makes "area over rate" give the period exactly, sidestepping the hard time integral.

Step 2. Use the ellipse identity $b^2=a p$ together with $p=L^2/GM\mu^2$ from the orbit equation, so $b^2=aL^2/GM\mu^2$ . Why: this is the bridge that injects the dynamical constant $L$ into the purely geometric area, so it can cancel.

Step 3. Square the area relation and substitute: $P^2=4\pi^2\mu^2a^2b^2/L^2=(4\pi^2\mu^2a^2/L^2)(aL^2/GM\mu^2)=4\pi^2a^3/GM$ . Why: $\mu$ and $L$ vanish identically, leaving only $a$ and the total mass $M=m_1+m_2$ — the exact statement of Kepler's harmonic law with the Newtonian correction that the "constant" depends on total mass. *As a check: dimensionally $[a^3/GM]=\mathrm{m^3/(m^3\,s^{-2})}=\mathrm{s^2}$ , and inserting $a=1\,\mathrm{AU}$ and $M=1\,M_\odot$ returns $P=1\,\mathrm{yr}$ .*

Limiting cases

The whole taxonomy of orbits is set by the sign of the total energy $E$ — equivalently by the eccentricity $e$ — with one further useful limit at the extreme of mass ratio.

Case	Result	Meaning
Bound (elliptical)	$E<0,\ 0\le e<1,\ a=-\dfrac{GM\mu}{2E}>0$	Closed periodic orbit between $r_{\min}$ and $r_{\max}$ ; binary stars, planets, and moons live here. The circle $e=0$ sits at the bottom of $U_{\rm eff}$ .
Marginal (parabolic)	$E=0,\ e=1,\ v=\sqrt{2GM/r}$	Escape reaching infinity with zero residual speed — the exact boundary between capture and escape. Long-period comets skim this limit.
Unbound (hyperbolic)	$E>0,\ e>1,\ v_\infty=\sqrt{2E/\mu}$	A single flyby; the body escapes with speed to spare. This is the two-body scattering that drives stellar encounters and gravity assists.
Test-particle limit	$m_2\ll m_1\Rightarrow \mu\to m_2,\ M\to m_1$	The elementary "planet about a fixed Sun" Kepler problem; the CM sinks into the massive body and the relative orbit becomes the light body's orbit.

Common trapThe reduced mass does not set the orbital acceleration

The single most common error is to write the relative equation of motion with the reduced mass $\mu$ where the total mass $M$ belongs. The correct relative acceleration is $\ddot{\mathbf{r}}=-GM\hat{\mathbf{r}}/r^2$ with $M=m_1+m_2$ — the two masses add because each body falls toward the other. The reduced mass appears only when you convert acceleration to force, $\mathbf{F}=\mu\ddot{\mathbf{r}}$ , and in the conserved energy $E=\tfrac12\mu\dot r^2+\dots$ and angular momentum $L=\mu r^2\dot\phi$ . Concretely, Kepler's third law carries $m_1+m_2$ , never $\mu$ : use $\mu$ there and you would badly underestimate the mass of any binary whose components are comparable — for an equal-mass binary $\mu=M/4$ , a factor-of-four error. A good sanity check: the force written two ways must agree, $GM\mu/r^2=Gm_1m_2/r^2$ , which is only true with $M=m_1+m_2$ .

Worked examples

1 · Reduced mass and the Sun's wobble (confidence builder)

Jupiter ( $m_2=1.898\times10^{27}\,\mathrm{kg}$ ) orbits the Sun ( $m_1=1.989\times10^{30}\,\mathrm{kg}$ ) at $a=5.20\,\mathrm{AU}$ . What is the reduced mass, and where is the Sun–Jupiter barycenter?

Step 1. Total mass $M=m_1+m_2=1.9909\times10^{30}\,\mathrm{kg}$ . The reduced mass is $\mu=m_1m_2/M$ . Since $m_2\ll m_1$ , expand: $\mu=m_2/(1+m_2/m_1)\approx m_2(1-9.5\times10^{-4})=\mathbf{0.99905\,m_2}$ . So $\mu$ is Jupiter's mass to within one part in a thousand — the reduced mass collapses onto the lighter body in the test-particle limit.

Step 2. The barycenter lies on the Sun–Jupiter line at distance $d_\odot=a\,(m_2/M)$ from the Sun's center. Numerically $m_2/M=1.898\times10^{27}/1.9909\times10^{30}=9.53\times10^{-4}$ , and $a=5.20\,\mathrm{AU}=7.78\times10^{11}\,\mathrm{m}$ , so $d_\odot=7.78\times10^{11}\times9.53\times10^{-4}=\mathbf{7.42\times10^{8}\ m}$ .

Step 3. Compare to the solar radius $R_\odot=6.96\times10^{8}\,\mathrm{m}$ : the barycenter sits at $d_\odot/R_\odot=\mathbf{1.07\,R_\odot}$ — just above the Sun's surface. The Sun therefore traces a small ellipse about a point outside itself, wobbling at $\approx12$ m/s over Jupiter's period. This reflex wobble, in radial velocity or astrometry, is exactly how Jupiter-mass exoplanets are found.

2 · Weighing the Sirius binary (Kepler's third law)

Sirius A and its white-dwarf companion Sirius B orbit with period $P=50.1\,\mathrm{yr}$ and relative semi-major axis $a=19.8\,\mathrm{AU}$ . Astrometry shows Sirius A traces the smaller apparent ellipse, with $a_A/a_B\approx0.49$ . Find the total mass and the two individual masses.

Step 1. Kepler's third law in solar units needs no constants: $(m_1+m_2)/M_\odot = a_{\rm AU}^3/P_{\rm yr}^2$ . Compute the cube and square: $a^3=19.8^3=7762$ and $P^2=50.1^2=2510$ , so $m_1+m_2=7762/2510=\mathbf{3.09\,M_\odot}$ .

Step 2. The mass ratio comes from the sizes of the two orbits about the CM: $m_A/m_B=a_B/a_A=1/0.49=2.03$ . The more massive star (Sirius A) traces the smaller ellipse — the same inverse relation as the first figure.

Step 3. Split the total mass with this ratio: $m_A=M\dfrac{a_B}{a_A+a_B}=3.09\times\dfrac{2.03}{3.03}=\mathbf{2.07\,M_\odot}$ and $m_B=3.09-2.07=\mathbf{1.02\,M_\odot}$ . A solar-mass white dwarf packed into an Earth-sized volume — one of the first stellar remnants ever weighed, and it agrees with the modern values $2.06\,M_\odot$ and $1.02\,M_\odot$ .

3 · Perihelion and aphelion speeds of the Earth (vis-viva)

Earth's orbit has $a=1.496\times10^{11}\,\mathrm{m}$ and eccentricity $e=0.0167$ , with $GM_\odot=1.327\times10^{20}\,\mathrm{m^3\,s^{-2}}$ . Find its mean, perihelion, and aphelion speeds.

Step 1. The mean (circular-equivalent) speed sets the scale: put $r=a$ in vis-viva (9), giving $v_{\rm circ}=\sqrt{GM/a}=\sqrt{1.327\times10^{20}/1.496\times10^{11}}=\sqrt{8.87\times10^{8}}=\mathbf{29.8\ km/s}$ — the familiar figure for Earth's orbital speed.

Step 2. Perihelion is at $r=a(1-e)$ . Vis-viva gives $v_{\rm peri}=v_{\rm circ}\sqrt{(1+e)/(1-e)}=29.8\times\sqrt{1.0167/0.9833}=29.8\times1.017=\mathbf{30.3\ km/s}$ . The planet moves fastest when closest, as Kepler's second law demands.

Step 3. Aphelion is at $r=a(1+e)$ , so $v_{\rm aph}=v_{\rm circ}\sqrt{(1-e)/(1+e)}=29.8\times0.983=\mathbf{29.3\ km/s}$ . The speed swings by about $\pm2\%$ around the mean — a 1 km/s gap between January perihelion and July aphelion, and the reason the seasons run slightly unequal in length.

Physics of the portrait

The two-body problem is where empirical astronomy became exact physics. Kepler distilled Tycho Brahe's naked-eye positions of Mars into three quantitative laws with no mechanism behind them; three generations later those laws became corollaries of a single inverse-square force. The portrait below belongs to the man whose ellipses Newton's mathematics finally explained.

Origins · 1 minJohannes Kepler and how this was discovered

Johannes Kepler

1571–1630 · the empirical laws of planetary motion that Newton's two-body solution later explained

Johannes Kepler inherited from Tycho Brahe the most accurate planetary observations ever made without a telescope, and spent years fighting the data for the orbit of Mars. Trained to expect circles, he found that no combination of circles and equants could fit Tycho's positions to better than eight arcminutes — and, refusing to discard those eight arcminutes, he abandoned two thousand years of circular dogma. In Astronomia nova (1609) he announced that Mars moves on an ellipse with the Sun at one focus (his first law) and that the line to the Sun sweeps equal areas in equal times (his second). A decade later, in Harmonices mundi (1619), searching for cosmic harmony, he found the relation $P^2\propto a^3$ linking every planet's period to its orbital size — the third law. Kepler had the what but not the why: he imagined a magnetic emanation from the Sun, not a central force. The explanation waited for Isaac Newton, who in the Principia (1687) proved that an inverse-square attraction produces exactly Kepler's conics and derived the harmonic law with its crucial amendment — the constant depends on the total mass, so a heavy companion shortens the period. What Kepler read off the sky, Newton read off a differential equation, and the two-body problem is where the two meet.

History

1609

— Kepler publishes the first two laws of planetary motion in Astronomia nova

1619

— Kepler states the harmonic third law in Harmonices mundi

1687

— Newton derives all three laws from an inverse-square force in the Principia

1799

— Laplace systematizes gravitational dynamics in Mécanique céleste

1890

— Poincaré proves the three-body problem is non-integrable, ending hopes of a general closed form

Going further

Texts. Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., Ch. 2 (Celestial Mechanics) and §7.2 (binary stars) · Binney & Tremaine, Galactic Dynamics, 2nd ed., Ch. 3 (The Orbits of Stars) and Ch. 1 (relaxation) · Murray & Dermott, Solar System Dynamics, Ch. 2 · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Ch. 3 (The Central-Force Problem) · Ryden, Introduction to Cosmology, 2nd ed. (dynamical masses). The key results to keep in mind: adding equations (1) gives $\ddot{\mathbf{R}}=0$ and subtracting gives $\ddot{\mathbf{r}}=-GM\hat{\mathbf{r}}/r^2$ ; the Binet equation is $u''+u=GM\mu^2/L^2$ with orbit $r=p/(1+e\cos\phi)$ ; Kepler's third law is $P^2=4\pi^2a^3/GM$ via the swept area. The worked examples give the Sun–Jupiter barycenter at $1.07\,R_\odot$ , Sirius total $3.09\,M_\odot$ splitting to $2.07$ and $1.02\,M_\odot$ , and Earth's speeds $29.8/30.3/29.3$ km/s from vis-viva. The constants used throughout are $G=6.674\times10^{-11}\ \mathrm{N\,m^2\,kg^{-2}}$ (CODATA) and $GM_\odot=1.327\times10^{20}\ \mathrm{m^3\,s^{-2}}$ .

Prerequisites. Newton's law of gravitation and Newton's second law · Conservation of energy and angular momentum · Conic sections in polar coordinates. Leads to. The Gravitational N-Body Problem · Two-Body Relaxation Time · Orbits in Spherical Potentials · The Virial Theorem in Stellar Systems.

Ask the tutor: Ask me to weigh a real binary from its period and separation, or to trace how a single hyperbolic encounter builds up into a cluster's relaxation time.

Why does the total mass, not the reduced mass, set the orbital period and acceleration?

Because each body accelerates toward the other, so when you subtract the two equations of motion the two pulls add: $\ddot{\mathbf{r}}=-G(m_1+m_2)\hat{\mathbf{r}}/r^2$ . The reduced mass $\mu$ is the inertia of the relative coordinate and appears only when you convert this acceleration into a force, $\mathbf{F}=\mu\ddot{\mathbf{r}}=-Gm_1m_2\hat{\mathbf{r}}/r^2$ , or into the energy and angular momentum. Kepler's third law inherits the same fact: $P^2=4\pi^2a^3/G(m_1+m_2)$ . The clean check is that a binary's period gives its total mass and says nothing, by itself, about how it is divided.

Is the reduced mass a real object, or just a bookkeeping trick?

It is a bookkeeping trick — but an exact and useful one. There is no physical particle of mass $\mu$ ; rather, the relative vector $\mathbf{r}=\mathbf{r}_1-\mathbf{r}_2$ obeys the same equation as a particle of mass $\mu$ in the field of a fixed mass $M$ . It works because the two-body Lagrangian, in CM and relative coordinates, splits cleanly into a free-CM piece and a one-body piece with kinetic energy $\tfrac12\mu\dot{\mathbf r}^2$ . The real bodies never move on the "relative orbit" — they trace scaled copies of it about the CM, with sizes $a_1=(m_2/M)a$ and $a_2=(m_1/M)a$ .

Why do the orbits close exactly, with no precession?

Closed orbits are special to just two force laws — the inverse-square ( $1/r^2$ ) and the harmonic ( $\propto r$ ) — a result known as Bertrand's theorem. For the inverse-square case the radial oscillation between perihelion and aphelion completes in exactly the same angle, $2\pi$ , that the body takes to go once around, so the ellipse retraces itself. Any deviation from a pure $1/r$ potential — relativistic corrections, an oblate central body, or a third mass — makes the two frequencies differ and the orbit precesses. Mercury's 43-arcsecond-per-century perihelion advance is exactly this: the leading breakdown of the Newtonian ellipse.

What quantities must be measured to get both individual masses of a binary?

Three things. First the period $P$ ; second the relative semi-major axis $a=a_1+a_2$ , which with Kepler's third law fixes the total mass; and third the ratio of the two bodies' excursions about the center of mass, $a_1/a_2=m_2/m_1$ , the mass ratio. Total mass plus ratio give each mass separately. For a visual binary the two apparent ellipses supply $a_1/a_2$ directly; for a double-lined spectroscopic binary the ratio of radial-velocity amplitudes does the same. A single-lined spectroscopic binary yields only the mass function, which constrains the companion but does not pin it down without the inclination.

Does the two-body problem still matter in general relativity?

Very much, as the backbone that relativity corrects. Two masses in general relativity do not trace closed Keplerian ellipses — the orbit precesses and slowly shrinks as it radiates gravitational waves, confirmed by the Hulse–Taylor pulsar. But at leading order the motion is still the Newtonian Kepler orbit; relativistic effects enter as a post-Newtonian series in powers of $v^2/c^2$ . When LIGO models a black-hole inspiral it starts from the two-body Kepler solution and adds these corrections until, in the final orbits, numerical relativity takes over. The Newtonian two-body problem is the zeroth-order skeleton of every relativistic binary.

How does a single close encounter between two stars relate to this problem?

A close pass between two field stars is a two-body problem with positive energy — a hyperbolic orbit. The stars approach, swing around once, and recede, each emerging with its velocity deflected by an angle set by the impact parameter and relative speed. One encounter barely changes anything, but a star in a cluster suffers many, and their random kicks accumulate. Summing the mean-square deflection over all encounters gives the two-body relaxation time — the timescale on which the cluster "forgets" its initial velocities. Because that time exceeds the age of the Universe for a galaxy, galaxies are treated as collisionless, which is the entire justification for smooth-potential stellar dynamics.