Virial Theorem for Central Forces

In 1933 Fritz Zwicky pointed the Mount Wilson telescopes at the Coma cluster — a swarm of a thousand galaxies — and measured how fast its members were moving. Then he applied a theorem from 1870, proved by Rudolf Clausius for the kinetic theory of heat, that relates the average kinetic energy of any bound system to the average of its potential energy. The speeds were absurdly high: for the cluster to hang together, it had to contain vastly more mass than its starlight showed. Zwicky called the missing ingredient dunkle Materie — dark matter — and the theorem that exposed it, decades ahead of its acceptance, was the virial theorem. What makes the theorem remarkable is what it does not require. It never solves the orbit. It does not care whether the motion is periodic, or chaotic, or a thousand galaxies swarming at once. It demands only one thing — that the motion stay bounded — and from that alone it extracts an exact algebraic relation between time-averaged kinetic and potential energy. For a central power-law potential $V = kr^n$ the relation is simply $2\langle T\rangle = n\langle V\rangle$ : one number, $n$ , fixes the entire energy budget of every bound orbit in that potential. This article builds the theorem from its engine — a bookkeeping quantity $G = \mathbf{p}\cdot\mathbf{r}$ whose growth is forbidden by boundedness — then verifies it explicitly on the Kepler ellipse, scales it up to self-gravitating systems of many bodies, and puts numbers on its two greatest hits: why a satellite losing energy to drag speeds up, and how a velocity dispersion weighs a cluster of stars.

Key takeaways

The virial theorem relates time-averaged kinetic and potential energies for any bounded motion: $2\langle T\rangle = \langle\mathbf{r}\cdot\nabla V\rangle$ , with no need to solve the orbit.
For a power-law central potential $V = kr^n$ , Euler's homogeneity theorem collapses this to $2\langle T\rangle = n\langle V\rangle$ .
Kepler/Coulomb ( $n=-1$ ): $2\langle T\rangle = -\langle V\rangle$ , so $E = -\langle T\rangle = \tfrac12\langle V\rangle$ — the energy budget behind binding energies, the drag paradox, and negative heat capacity.
Isotropic harmonic oscillator ( $n=2$ ): $\langle T\rangle = \langle V\rangle$ — energy equipartitioned on average, the classical seed of equipartition in statistical mechanics.
The only input is boundedness: the generator $G = \mathbf{p}\cdot\mathbf{r}$ stays finite, so its average rate of change over long times vanishes. Unbound (hyperbolic) motion breaks the theorem.
The many-body gravitational version $2\langle T\rangle + \langle V\rangle = 0$ turns a measured velocity dispersion into a total mass — Zwicky's dark-matter argument and today's routine cluster-weighing tool.
Key idea: for bounded motion in a homogeneous potential $V = kr^n$ , the time-averaged kinetic and potential energies obey $2\langle T\rangle = n\langle V\rangle$ — a constraint that follows purely from boundedness, without solving the orbit.

Where this lives today

The virial theorem is the default scale in every branch of gravitational astrophysics. Star-formation surveys with ALMA commonly use the virial parameter $\alpha_{\rm vir}=5\sigma^2R/(GM)$ — the dimensionless form of the virial mass estimate — with $\alpha_{\rm vir}\lesssim2$ indicating that a cloud can be gravitationally bound in the simplest model; cosmological simulations define halo boundaries by the "virial radius"; and cluster masses from dispersion, X-ray temperature, and gravitational lensing are routinely cross-checked against each other, with the lensing maps confirming the presence of the unseen mass that Zwicky's 1933 virial argument had inferred. The Gaia mission's fourth data release, expected in December 2026, will sharpen the game further: with proper motions for individual stars in hundreds of clusters, both sides of $2\langle T\rangle + \langle V\rangle = 0$ become directly measurable, turning the theorem from an estimator into a consistency test of dynamical equilibrium itself. Away from gravity, the theorem quietly audits the microscopic world. Quantum chemistry codes report the ratio $-\langle V\rangle/\langle T\rangle$ for converged electronic wavefunctions — Fock's quantum virial theorem demands exactly 2 for Coulomb systems, and deviations flag an unconverged basis set. The same check runs inside molecular-dynamics engines and stellar-structure codes, where a drifting virial residual $2T + \sum\mathbf{r}\cdot\mathbf{F}$ is the standard smoke alarm for a broken integrator or an unphysical boundary condition. A theorem with no dynamics in it at all — just boundedness and a derivative — has become one of computation's most useful error bars.

The physical picture

Watch a comet on an eccentric orbit. Near perihelion it whips around the Sun at tremendous speed — kinetic energy dominates the ledger. Out near aphelion it crawls, almost all of its energy parked as potential. Instant by instant the split between $T$ and $V$ swings wildly. And yet the virial theorem says the time averages of these two wildly swinging quantities are locked in a rigid ratio: for gravity, $\langle T\rangle$ is always exactly half of $|\langle V\rangle|$ , for every bound orbit, at every eccentricity. The fast-and-close excursions and the slow-and-far excursions conspire, over each period, to a fixed exchange rate.

Figure

The energy see-saw — kinetic and potential energy trade places around the orbit, but their time averages are locked at the ratio two-to-one for gravity

The engine behind this rigidity is almost embarrassingly simple. Build the scalar $G = \mathbf{p}\cdot\mathbf{r}$ — a kind of phase-space radius that grows when the body moves outward and shrinks when it falls back. Along any orbit, $G$ wiggles. But if the motion is bounded — positions finite, speeds finite — then $G$ is trapped in a band forever. A quantity trapped in a band cannot have any long-run trend: average its rate of change over a long enough time and you get zero. That single statement, $\langle\mathrm{d}G/\mathrm{d}t\rangle = 0$ , is the virial theorem; everything else is unpacking the derivative.

Figure

Boundedness is the whole proof — G wiggles inside a band forever, so the average slope of its graph is forced to zero

Why does the derivative of $G$ know about energies? Because $\dot G$ splits into exactly two pieces: $\mathbf{p}\cdot\dot{\mathbf{r}} = mv^2 = 2T$ , and $\dot{\mathbf{p}}\cdot\mathbf{r} = \mathbf{F}\cdot\mathbf{r}$ — Clausius named the average of this second piece the virial, from the Latin vis, force. The factor of 2 in the theorem is nothing deeper than $\mathbf{p}\cdot\mathbf{v}$ being twice the kinetic energy. And the exponent $n$ enters through a lovely piece of pure mathematics: for any potential homogeneous of degree $n$ , Euler's theorem says $\mathbf{r}\cdot\nabla V = nV$ identically — the force term simply is $-n$ times the potential energy, no averaging subtleties required. Keep one boundary post in view from the start: the theorem is about bounded motion. A hyperbolic flyby has $G$ growing without limit — the chord in the picture above never flattens — and the theorem simply does not apply. Boundedness is not a technicality; it is the entire content of the hypothesis.

The machinery

The time average that the theorem controls is the honest infinite-time average. For a periodic orbit it reduces to the average over one period, but the definition asks for no periodicity at all:

\langle f\rangle \equiv \lim_{\tau\to\infty}\frac{1}{\tau}\int_0^{\tau} f(t)\,\mathrm{d}t \qquad (1)

where: $f(t)$ — any quantity evaluated along the trajectory; $\tau$ — the averaging window. For a Kepler ellipse of period $P$ , $\langle f\rangle$ equals the one-period average; for a non-closing rosette orbit or a swarm of stars it is the long-run mean, which is exactly what observations of a cluster sample. The generator whose boundedness powers everything is the Clausius scalar:

G = \mathbf{p}\cdot\mathbf{r} = m\,\mathbf{v}\cdot\mathbf{r} \qquad (2)

where: $\mathbf{r}$ — position from the force center; $\mathbf{p} = m\mathbf{v}$ — momentum. Geometrically $G = \tfrac{m}{2}\,\mathrm{d}(r^2)/\mathrm{d}t$ : it is positive while the body recedes, negative while it falls back, and for bounded motion it oscillates in a fixed band — the fact the whole theorem rests on. Differentiate $G$ once, using nothing but the product rule and Newton's second law, and the two energies appear:

\frac{\mathrm{d}G}{\mathrm{d}t} = \dot{\mathbf{p}}\cdot\mathbf{r} + \mathbf{p}\cdot\dot{\mathbf{r}} = \mathbf{F}\cdot\mathbf{r} + 2T \qquad (3)

where: $\mathbf{p}\cdot\dot{\mathbf{r}} = m\mathbf{v}\cdot\mathbf{v} = 2T$ — twice the kinetic energy (this is where the theorem's factor of 2 is born); $\mathbf{F}\cdot\mathbf{r}$ — the force dotted into position, whose time average Clausius named the virial. Average both sides and use boundedness to kill the left side, and:

2\langle T\rangle = -\langle\mathbf{F}\cdot\mathbf{r}\rangle = \langle\mathbf{r}\cdot\nabla V\rangle \qquad (4)

where: $\mathbf{F} = -\nabla V$ for a conservative force. Equation (4) is the general virial theorem: true for any conservative force and any bounded motion, central or not. It becomes an energy relation the moment the potential has a definite scaling behavior. That scaling input is Euler's theorem on homogeneous functions. If $V(\lambda\mathbf{r}) = \lambda^n V(\mathbf{r})$ — true for any power law $V = kr^n$ — then differentiating with respect to $\lambda$ at $\lambda = 1$ gives an identity, not an approximation:

\mathbf{r}\cdot\nabla V = n\,V \qquad (5)

where: $n$ — the degree of homogeneity: $n = -1$ for gravity and Coulomb, $n = 2$ for the isotropic harmonic oscillator, and the limit $n \to 0$ (with $kr^n/n \to k\ln r$ ) covers the logarithmic potential of flat galactic rotation curves, for which $\mathbf{r}\cdot\nabla V = k$ exactly. Feeding (5) into (4) delivers the headline:

2\langle T\rangle = n\,\langle V\rangle \qquad (6)

where: both averages are taken along the same bound trajectory. Since the total energy $E = T + V$ is conserved (so $\langle E\rangle = E$ ), the theorem immediately parcels $E$ between the two accounts:

E = \Big(1 + \tfrac{2}{n}\Big)\langle T\rangle = \Big(1 + \tfrac{n}{2}\Big)\langle V\rangle \qquad (7)

where: for gravity ( $n=-1$ ) this reads $E = -\langle T\rangle = \tfrac12\langle V\rangle$ : a bound orbit's total energy is minus its average kinetic energy. That innocuous sign carries a shocking consequence — remove energy from an orbiting body and its average speed increases — unpacked in the warning and worked examples below.

Which relation answers which question

you want	use	notes
any conservative force, bounded motion	$2\langle T\rangle = \langle\mathbf{r}\cdot\nabla V\rangle$ , eq. (4)	the general form — no scaling assumed
power-law potential $V = kr^n$	$2\langle T\rangle = n\langle V\rangle$ , eq. (6)	Euler's identity does the work
energy budget from one average	eq. (7)	gravity: $E = -\langle T\rangle = \tfrac12\langle V\rangle$
flat rotation curve $V = k\ln r$	$2\langle T\rangle = k$	the $n\to0$ limit; $\langle T\rangle$ independent of orbit size
self-gravitating N-body system	$2\langle T\rangle + \langle V\rangle = 0$ , eq. (8)	Zwicky's scale; $M \approx 5\sigma_{\mathrm{los}}^2R/G$

Derivation AThe virial theorem for power-law central forces

We are deriving $2\langle T\rangle = n\langle V\rangle$ for bounded motion in $V = kr^n$ , from nothing but Newton's second law, one time average, and one scaling identity. The strategy: differentiate the generator $G = \mathbf{p}\cdot\mathbf{r}$ , average the result over an infinite window, and let boundedness annihilate the left-hand side; Euler's homogeneity theorem then converts the surviving force term into a multiple of the potential energy. Watch where boundedness enters — it is used exactly once, and it is the entire hypothesis: every celebrated failure of the theorem (hyperbolic orbits, evaporating clusters) is a failure of that single step, never of the algebra.

Step 1. Differentiate the generator. With $G = \mathbf{p}\cdot\mathbf{r}$ , the product rule and Newton's law $\dot{\mathbf{p}} = \mathbf{F}$ give

\frac{\mathrm{d}G}{\mathrm{d}t} = \mathbf{F}\cdot\mathbf{r} + \mathbf{p}\cdot\dot{\mathbf{r}} = \mathbf{F}\cdot\mathbf{r} + 2T

Why $2T$ and not $T$ : $\mathbf{p}\cdot\dot{\mathbf{r}} = m|\mathbf{v}|^2$ , which is twice $\tfrac12 mv^2$ — the theorem's factor of 2 is born here, in the mismatch between momentum-times-velocity and kinetic energy.

Step 2. Time-average both sides over a window $\tau$ and take $\tau \to \infty$ :

\langle\dot G\rangle = \lim_{\tau\to\infty}\frac{G(\tau) - G(0)}{\tau} = 2\langle T\rangle + \langle\mathbf{F}\cdot\mathbf{r}\rangle

Step 3. Kill the left side with boundedness. For bounded motion, $|G| \le m\,v_{\max}\,r_{\max}$ — a fixed number — so the numerator $G(\tau) - G(0)$ is bounded while the denominator grows without limit:

\langle\dot G\rangle = 0

Why this is the crux: for an unbound orbit $G$ grows linearly at late times ( $\mathbf{v}$ and $\mathbf{r}$ align as the body recedes), the limit is finite and nonzero, and the theorem dies. Periodicity, note, was never needed — only confinement.

Step 4. Convert the force term with Euler's identity. For $V = kr^n$ , homogeneity of degree $n$ gives $\mathbf{r}\cdot\nabla V = nV$ pointwise along the orbit, so $\mathbf{F}\cdot\mathbf{r} = -\mathbf{r}\cdot\nabla V = -nV$ and

0 = 2\langle T\rangle - n\langle V\rangle \;\;\Rightarrow\;\; 2\langle T\rangle = n\langle V\rangle

Why no averaging subtlety survives: the identity $\mathbf{r}\cdot\nabla V = nV$ holds at every instant, not just on average, so the conversion commutes with the time average exactly.

Go deeper · Derivation B · 1 minChecking the theorem on an explicit Kepler ellipse

Derivation BChecking the theorem on an explicit Kepler ellipse

The virial theorem claims $\langle V\rangle = -k/a$ and $\langle T\rangle = k/2a$ for a bound Kepler orbit with $V = -k/r$ and semi-major axis $a$ — for every eccentricity. We verify this by brute force, and en route we prove the classic average $\langle 1/r\rangle_t = 1/a$ rather than quoting it. The strategy: convert the time average to a true-anomaly average using angular momentum conservation ( $\mathrm{d}t = mr^2\,\mathrm{d}\psi/L$ ), integrate over the conic $r = p/(1+e\cos\psi)$ , and clean up with Kepler's second law in the form $L = 2\pi abm/P$ . The satisfying point to watch: the factor $r^2$ from the time-conversion softens the $1/r$ being averaged into a plain $\int r\,\mathrm{d}\psi$ — the eccentric orbit's slow, distant leg gets exactly the extra time-weight needed to make the average come out independent of $e$ .

Step 1. Convert the time average of $1/r$ to an angle average. With $L = mr^2\dot\psi$ ,

\Big\langle\frac{1}{r}\Big\rangle = \frac{1}{P}\int_0^P \frac{\mathrm{d}t}{r} = \frac{m}{LP}\int_0^{2\pi} r\,\mathrm{d}\psi

Why the weight matters: equal time is not equal angle — the body dawdles at large $r$ , so time-averaging emphasizes aphelion. The $r^2$ Jacobian encodes precisely that emphasis.

Step 2. Integrate over the ellipse. With $r = p/(1+e\cos\psi)$ and the standard integral $\int_0^{2\pi}\mathrm{d}\psi/(1+e\cos\psi) = 2\pi/\sqrt{1-e^2}$ ,

\int_0^{2\pi} r\,\mathrm{d}\psi = \frac{2\pi p}{\sqrt{1-e^2}} = 2\pi a\sqrt{1-e^2} = 2\pi b

using $p = a(1-e^2)$ and the semi-minor axis $b = a\sqrt{1-e^2}$ .

Step 3. Eliminate $L$ with Kepler's second law. The areal rate is $L/2m$ , and one period sweeps the full ellipse area $\pi ab$ , so $L = 2\pi abm/P$ . Then

\Big\langle\frac{1}{r}\Big\rangle = \frac{m}{LP}\,2\pi b = \frac{m\cdot 2\pi b\cdot P}{2\pi abm\cdot P} = \frac{1}{a}

Why this is a small miracle: every trace of eccentricity has cancelled — the time-averaged inverse distance depends only on the orbit's energy scale $a$ , which is exactly what the virial theorem, blind to orbit shape, requires.

Step 4. Assemble the energy averages. Immediately $\langle V\rangle = -k\langle 1/r\rangle = -k/a$ ; and since the conserved total energy of a bound Kepler orbit is $E = -k/2a$ ,

\langle T\rangle = E - \langle V\rangle = -\frac{k}{2a} + \frac{k}{a} = \frac{k}{2a}

Step 5. Check the virial relation: $2\langle T\rangle = k/a = -\langle V\rangle$ — the $n = -1$ theorem, confirmed term by term, for all $e$ . As a counterpoint, the direct position average is $\langle r\rangle_t = a(1 + e^2/2) \ne a$ — different functions of $r$ average differently, and only $\langle 1/r\rangle$ is this clean.

Go deeper · Derivation C · 2 minWeighing a star cluster with the many-body virial theorem

Derivation CWeighing a star cluster with the many-body virial theorem

We are deriving the tool Zwicky used on Coma: for a self-gravitating system of $N$ bodies in a settled (bounded, statistically steady) state, $2\langle T\rangle + \langle V\rangle = 0$ , and from it the practical mass estimate $M \approx 5\sigma_{\mathrm{los}}^2 R/G$ . The strategy mirrors the single-particle proof with the generator promoted to a sum, $G = \sum_i\mathbf{p}_i\cdot\mathbf{r}_i$ ; the new content is that the pair-summed gravitational potential energy is itself homogeneous of degree $-1$ in all coordinates at once, so Euler's identity applies to the whole configuration. The step to watch is the last one — converting the theorem into an observable requires modeling assumptions (isotropic velocities, a mass profile), and those assumptions, not the theorem, set the error bars.

Step 1. Promote the generator and differentiate. With $G = \sum_i \mathbf{p}_i\cdot\mathbf{r}_i$ ,

\frac{\mathrm{d}G}{\mathrm{d}t} = 2T + \sum_i \mathbf{F}_i\cdot\mathbf{r}_i

where $T = \sum_i \tfrac12 m_iv_i^2$ and $\mathbf{F}_i$ is the total gravitational force on body $i$ from all others.

Step 2. Apply Euler's identity to the pair potential. The total potential energy $V = -\sum_{i<j} Gm_im_j/|\mathbf{r}_i - \mathbf{r}_j|$ is homogeneous of degree $-1$ under scaling all positions, so $\sum_i \mathbf{r}_i\cdot\nabla_i V = -V$ , i.e. $\sum_i\mathbf{F}_i\cdot\mathbf{r}_i = V$ . Why internal forces do not cancel here: they cancel in the total momentum, but $\mathbf{F}\cdot\mathbf{r}$ weights each force by position — what survives is precisely the system's binding energy.

Step 3. Average and use boundedness (the cluster neither collapses nor evaporates over the averaging time):

0 = 2\langle T\rangle + \langle V\rangle \qquad (8)

Step 4. Model the two averages with observables. Kinetic: if velocities are isotropic, the line-of-sight dispersion $\sigma_{\mathrm{los}}$ seen in Doppler widths satisfies $\langle v^2\rangle = 3\sigma_{\mathrm{los}}^2$ , so $\langle T\rangle = \tfrac32 M\sigma_{\mathrm{los}}^2$ . Potential: for a uniform sphere of mass $M$ and radius $R$ , $\langle V\rangle = -\tfrac35\,GM^2/R$ .

Step 5. Solve for the mass:

3M\sigma_{\mathrm{los}}^2 = \frac{3}{5}\frac{GM^2}{R} \;\;\Rightarrow\;\; M = \frac{5\,\sigma_{\mathrm{los}}^2\,R}{G} \qquad (9)

where: the prefactor 5 is the uniform-sphere value — realistic density profiles shift it by factors of order unity, which is why virial masses are quoted with structural caveats. Why the estimate is nonetheless robust: it needs no orbit solutions, no individual distances, only a dispersion and a size — two of the easiest quantities astronomy can measure. *Dimensionally $[\sigma^2R/G] = \mathrm{kg}$ ; applied to Coma ( $\sigma_{\mathrm{los}} \approx 1000$ km/s, $R \approx 1$ Mpc) it gives $\sim 10^{45}$ kg $\sim 5\times10^{14}\,M_\odot$ , far above the luminous mass — Zwicky's 1933 discrepancy, confirmed by modern lensing.*

Limiting cases

Case	Result	Meaning
Kepler / Coulomb ( $n=-1$ )	$2\langle T\rangle = -\langle V\rangle$ , $E = -\langle T\rangle$	binding energy equals average kinetic energy; negative heat capacity
isotropic oscillator ( $n=2$ )	$\langle T\rangle = \langle V\rangle = E/2$	equipartition on average — the seed of the statistical-mechanics theorem
logarithmic potential ( $n\to0$ )	$2\langle T\rangle = k$ , independent of orbit size	flat rotation curves: same $\langle v^2\rangle$ at every radius
steep power law ( $n=4$ )	$\langle V\rangle = \tfrac12\langle T\rangle$ , $E = \tfrac32\langle T\rangle$	stiffer wells park more energy in motion than in position
unbound motion ( $E \ge 0$ , gravity)	theorem fails	$G = \mathbf{p}\cdot\mathbf{r}$ grows without bound; $\langle\dot G\rangle \ne 0$
circular orbit, any $n$	$2T = nV$ holds instantaneously	no averaging needed — $T$ and $V$ are separately constant

Common trapThe virial theorem says 2T equals −V at every moment of the orbit

It says nothing of the kind — the theorem constrains time averages only, and confusing the two is the classic exam error. On an eccentric Kepler orbit the instantaneous ratio $2T/|V|$ swings from $(1+e)/1$ -ish at perihelion to far below 1 at aphelion; vis-viva shows $2T = -V$ holds instantaneously only at the special radius $r = a$ . (Circular orbits are the one exception: with $T$ and $V$ separately constant, the instantaneous and averaged statements coincide.) Two more traps sit close by. First, the theorem requires bounded motion: apply $2\langle T\rangle = -\langle V\rangle$ to a hyperbolic flyby and you get nonsense, because the boundary term $G(\tau)/\tau$ no longer vanishes. Second, different averages differ: $\langle 1/r\rangle_t = 1/a$ but $\langle r\rangle_t = a(1+e^2/2)$ — you may not compute $\langle V\rangle$ by evaluating $V$ at the average radius. Average the function, not the argument.

Worked examples

1 · A satellite ledger, checked to the joule (confidence builder)

The ISS orbits at altitude $420$ km, so $r = 6371 + 420 = 6791$ km, on a near-circular orbit. Using $GM_\oplus = 3.986\times10^{14}\ \mathrm{m^3/s^2}$ , verify the virial theorem explicitly per kilogram of station, and find the energy needed to send that kilogram to infinity.

Step 1. Orbital speed: $v = \sqrt{GM_\oplus/r} = \sqrt{3.986\times10^{14}/6.791\times10^6} = \sqrt{5.870\times10^7} = 7.66\times10^3\ \mathrm{m/s}$ .

Step 2. Kinetic energy per kilogram: $T = \tfrac12 v^2 = 2.93\times10^7\ \mathrm{J/kg}$ . Potential energy per kilogram: $V = -GM_\oplus/r = -5.87\times10^7\ \mathrm{J/kg}$ .

Step 3. Check the theorem: $2T = 5.87\times10^7 = -V$ . $\boxed{2T = -V\ \text{exactly}}$ — and for a circular orbit this holds at every instant, no averaging required.

Step 4. Total energy: $E = T + V = -2.93\times10^7\ \mathrm{J/kg} = -T$ , confirming $E = -\langle T\rangle$ . The unbinding energy is $|E| = T$ : to escape, each kilogram needs a further $\boxed{29\ \mathrm{MJ}}$ — exactly as much energy again as its current kinetic energy. (Check: escape speed $= \sqrt2\,v$ , so escape kinetic energy $= 2T$ , and the top-up needed is $2T - T = T$ .)

2 · Weighing a globular cluster from two numbers (exam level)

A globular cluster has line-of-sight velocity dispersion $\sigma_{\mathrm{los}} = 10\ \mathrm{km/s}$ and half-light radius $R \approx 10\ \mathrm{pc} = 3.09\times10^{17}\ \mathrm{m}$ . Estimate its virial mass in solar masses ( $M_\odot = 1.99\times10^{30}$ kg, $G = 6.674\times10^{-11}$ SI), and state what physics each factor of the formula carries.

Step 1. Apply the virial mass estimate (9): $M = 5\sigma_{\mathrm{los}}^2R/G$ .

Step 2. Numerator: $5 \times (10^4)^2 \times 3.09\times10^{17} = 5 \times 10^8 \times 3.09\times10^{17} = 1.54\times10^{26}$ .

Step 3. Divide by $G$ : $M = 1.54\times10^{26}/6.674\times10^{-11} = 2.31\times10^{36}\ \mathrm{kg}$ .

Step 4. In solar units: $M = 2.31\times10^{36}/1.99\times10^{30} = \boxed{1.2\times10^6\ M_\odot}$ — a typical globular cluster, weighed without tracking a single stellar orbit.

Step 5. Audit the formula's anatomy. The $\sigma_{\rm los}^2$ is the kinetic piece converted from one observed component to three dimensions by isotropy, $\langle v^2\rangle=3\sigma_{\rm los}^2$ . The factor multiplying $R/G$ is a structure coefficient: for a uniform sphere it gives the coefficient $5$ , while using a half-light radius and a realistic density profile changes this coefficient by a factor of order unity. If the measured luminosity implied ten times less mass, you would have rediscovered dark matter, which is precisely what happened at cluster scale in 1933.

Origins · 1 minRudolf Clausius and how this was discovered

Rudolf Clausius

1822–1888 · Zürich and Bonn

Clausius was the great bookkeeper of nineteenth-century physics. He gave thermodynamics its second law in the form still taught ("heat cannot of itself pass from a colder to a hotter body"), invented the concept and the very word entropy in 1865 — choosing Greek roots deliberately so the word would survive translation — and closed his famous paper with the pair of sentences: the energy of the universe is constant; the entropy of the universe tends to a maximum. The virial theorem came in 1870, in a short lecture titled "On a Mechanical Theorem Applicable to Heat": averaging $\mathbf{p}\cdot\mathbf{r}$ was his device for connecting the microscopic kinetic energy of gas molecules to the pressure confining them, and he coined virial from the Latin vis, force. He never knew his averaging trick would one day weigh galaxy clusters and expose most of the matter in the universe as invisible — the theorem outgrew its inventor's century entirely.

History

1870

— Rudolf Clausius states the virial theorem in a lecture at Bonn, as a tool linking molecular kinetic energy to gas pressure.

1911

— Henri Poincaré applies virial reasoning to self-gravitating systems in his lectures on cosmogony, setting up its astrophysical career.

1930

— Vladimir Fock proves the quantum-mechanical virial theorem: the same

2\langle T\rangle = n\langle V\rangle

holds for stationary states, hydrogen included.

1933

— Fritz Zwicky applies the theorem to the Coma cluster's velocity dispersion and infers dunkle Materie — dark matter — sixty years before its general acceptance.

2016

— Gaia begins delivering space-astrometry proper motions that turn virial mass estimates of star clusters into precision measurements.

Going further

Texts. Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., §3.4 (The Virial Theorem) · Landau & Lifshitz, Mechanics, 3rd ed., §10 · Morin, Introduction to Classical Mechanics, ch. 7 · Taylor, Classical Mechanics, §8 problems (time averages) · Binney & Tremaine, Galactic Dynamics, ch. 4 (the tensor virial theorem in astrophysics). The central results follow from differentiating $G=\mathbf{p}\cdot\mathbf{r}$ to obtain $\dot G=2T+\mathbf{F}\cdot\mathbf{r}$ , then applying Euler's homogeneity theorem to reach $2\langle T\rangle=n\langle V\rangle$ ; the Kepler averages used are $\langle 1/r\rangle=1/a$ and $E=-k/2a$ . Zwicky's Coma-cluster dark-matter inference (1933) is the historical application. Worked-example numbers on this page (ISS: $v = 7.66$ km/s, $T = 2.93\times10^7$ J/kg; cluster: $M = 1.2\times10^6\,M_\odot$ from $\sigma = 10$ km/s, $R = 10$ pc) are illustrative arithmetic computed from the boxed formulas.

Builds on: Central Force Fundamentals · Reduction to the Radial Equation · The Two-Body Problem and Reduced Mass · Classification of Orbits by Energy. Leads to: The Kepler Problem · Kepler's Third Law: Period-Semimajor Relation · The Effective Potential.

Can I apply the virial theorem to a comet on a hyperbolic flyby?

No — and the failure is instructive rather than technical. On an unbound trajectory the body eventually recedes with $\mathbf{v}$ nearly parallel to $\mathbf{r}$ , so $G = m\mathbf{v}\cdot\mathbf{r}$ grows roughly linearly in time forever. The boundary term $[G(\tau) - G(0)]/\tau$ then tends to a finite nonzero number, and the step that sets $\langle\dot G\rangle = 0$ — the only place boundedness is used — collapses. Concretely: an $E > 0$ Kepler orbit spends almost all its time in the asymptotic region where $T \to E$ and $V \to 0$ , so the long-time averages give $2\langle T\rangle = 2E \ne -\langle V\rangle = 0$ . The theorem is not approximately true for unbound motion; it is simply about confined systems, which is exactly why it certifies that a cluster whose members obey it is gravitationally bound.

Is 2T = −V ever true at a single instant on an eccentric orbit, or only on average?

It is true instantaneously at exactly one radius. Per unit mass, $2T = -V$ reads $v^2 = GM/r$ , while vis-viva gives $v^2 = GM(2/r - 1/a)$ ; equating them yields $1/r = 1/a$ , so the instantaneous virial balance holds precisely when the body crosses $r = a$ , the semi-major-axis distance — twice per orbit. Inside $r = a$ the orbit is kinetic-rich ( $2T > -V$ ), outside it potential-rich. Circular orbits sit permanently at $r = a$ , which is why they satisfy the relation at every moment. This also explains why the theorem is genuinely about averages: the fast inner excursions and slow outer ones must be weighted by the time spent in each, and the Kepler check in Derivation B shows that weighting is exactly what makes $\langle 1/r\rangle = 1/a$ regardless of eccentricity.

Why does a satellite speed up when drag takes energy away? That still feels like a perpetual-motion trick

No trick — the energy audit balances perfectly once the potential account is included. For gravity, $E = -\langle T\rangle$ : total energy and average kinetic energy point in opposite directions. Drag drains $E$ , making it more negative, which shrinks $a$ (since $E = -k/2a$ ) and therefore raises $\langle T\rangle = k/2a$ . Where does the extra speed come from? Gravity: as the orbit spirals inward, the gravitational force does positive work. Of the potential energy released, exactly half goes into increased kinetic energy and half is surrendered to drag heating — the virial theorem's 2:1 ratio is the accountant enforcing the split. Astronomers call the flip side of this "negative heat capacity": a star or cluster that loses energy gets hotter. It is why stars contract and heat as they radiate, and why hardening binaries heat globular-cluster cores.

The theorem needs a power law. What if my potential is a sum of terms, like a Coulomb part plus a confining r² trap?

Use the general form, equation (4), and apply Euler's identity term by term. If $V = V_1 + V_2$ with $V_1 = k_1r^{n_1}$ and $V_2 = k_2r^{n_2}$ , then $\mathbf{r}\cdot\nabla V = n_1V_1 + n_2V_2$ pointwise, so $2\langle T\rangle = n_1\langle V_1\rangle + n_2\langle V_2\rangle$ . You no longer get a single clean ratio between $\langle T\rangle$ and total $\langle V\rangle$ , but you still get one exact linear constraint among three averages — often enough to pin down what you need when combined with energy conservation. This mixed form is the working version in practice: ions in a Paul-trap-plus-Coulomb system, stars in a disk-plus-halo galaxy potential, or quarkonium models with a Coulomb-plus-linear potential all use exactly this term-by-term bookkeeping.

How did Zwicky get from Doppler shifts to missing mass — and why did people believe the theorem rather than suspect the cluster was flying apart?

Zwicky measured redshifts of individual Coma galaxies, found a line-of-sight scatter near $1000$ km/s, and fed $\sigma_{\mathrm{los}}$ and the cluster radius into $M \sim 5\sigma^2R/G$ : the answer exceeded the mass implied by starlight by a large factor (modern accounting puts the dark share near ninety percent, split between dark matter and hot gas). The escape hatch — "the cluster is not bound, it is dispersing" — fails on timescales: at $1000$ km/s, Coma would evaporate in a few billion years, and finding many rich clusters today would then be wildly improbable; moreover an unbound swarm would show systematic expansion, not the observed settled profile. Modern confirmation is threefold and independent: X-ray gas temperature (a thermal virial measurement), and gravitational lensing (no dynamics at all) both give masses agreeing with the virial estimate.

Does the virial theorem survive quantum mechanics, or is it a classical accident?

It survives intact — Fock proved it in 1930 for stationary states. The quantum statement replaces time averages with expectation values in an energy eigenstate: $2\langle\hat T\rangle = \langle\hat{\mathbf{r}}\cdot\nabla V\rangle$ , so for Coulomb systems $2\langle\hat T\rangle = -\langle\hat V\rangle$ exactly. Hydrogen's ground state wears the numbers openly: $E = -13.6$ eV, $\langle\hat T\rangle = +13.6$ eV, $\langle\hat V\rangle = -27.2$ eV. The proof even parallels the classical one — the commutator $[\hat H, \hat{\mathbf{r}}\cdot\hat{\mathbf{p}}]$ plays the role of $\dot G$ , and its expectation value vanishes in a stationary state just as $\langle\dot G\rangle$ vanishes for bounded motion. Quantum chemists use the ratio $-\langle V\rangle/\langle T\rangle = 2$ as a convergence diagnostic: a variational wavefunction that violates the virial ratio is provably not the true eigenstate.

Do I need the orbit to be periodic, or to know anything about its shape?

Neither — and this is the theorem's superpower. The proof used confinement (to kill the boundary term) and the scaling of the potential (to convert the force term); it never asked whether the orbit closes, precesses, or fills a torus chaotically. That is why the theorem applies verbatim to a precessing rosette in a galactic halo, to the non-periodic tangle of a triple-star system, and to the statistically steady swarm of a globular cluster, where no individual orbit is periodic at all. The practical caveat hides in the averaging window: $\langle\dot G\rangle \to 0$ needs times long compared to the system's internal orbital times. A cluster observed today gives a snapshot, and using it assumes the system has had time to settle ("virialize") — for systems still relaxing after a merger, virial mass estimates carry genuine systematic errors.