Virial Theorem in Stellar Systems

The virial theorem is the single most-used equilibrium relation in stellar dynamics, and it earns that status by trading something you cannot measure for two things you can. You cannot watch the individual orbits of a billion stars, nor weigh a galaxy cluster on a scale. But you can measure how fast its members move — a velocity dispersion $\sigma$ — and how big it is — a radius $R$ . In its scalar form $2\langle T\rangle+\langle W\rangle=0$ , the theorem links the total random kinetic energy $T$ of a bound, self-gravitating system to its total gravitational potential energy $W$ , with no reference to any orbit, and so converts kinematics directly into a mass, $M\sim\sigma^2 R/G$ . Physically it is a statement about a standoff. Gravity pulls the swarm inward; the internal motions of its members resist collapse; equilibrium sits precisely where twice the kinetic energy cancels the (negative) potential energy. The cleanest derivation starts from the moment of inertia $I=\sum_i m_i|\mathbf r_i|^2$ : two time derivatives and Newton's law give the exact identity $\tfrac12\ddot I=2T+W$ , and in a relaxed system the long-time average of $\ddot I$ vanishes, leaving the virial balance. A more general route runs through the tensor virial theorem from velocity moments of the collisionless Boltzmann equation. The celebrated consequence is a weighing. Zwicky's 1933 analysis of the Coma cluster used exactly this estimator and found a mass far larger than the summed starlight — the first dynamical evidence for what he named dark matter. That estimator, suitably refined, still underlies the cluster mass functions of modern cosmology, even as its strong assumptions — boundedness, isolation, relaxation, isotropy — mark where it must be handled with care.

Key takeaways

The virial theorem relates the time-averaged kinetic energy $T$ and gravitational potential energy $W$ of a bound, relaxed, self-gravitating system: $2\langle T\rangle+\langle W\rangle=0$ .
It is the trace and steady-state limit of the exact identity $\tfrac12\ddot I=2T+W$ , following from the moment of inertia and Newtonian gravity — or from the tensor virial theorem built on the collisionless Boltzmann equation.
Solved for mass it gives the workhorse estimator $M\sim\langle v^2\rangle R/(\alpha G)$ , with $\langle v^2\rangle=3\sigma_{\mathrm{los}}^2$ and $\alpha=3/5$ for a uniform sphere.
Applied to Coma by Zwicky (1933), the dynamical mass exceeded the luminous mass by roughly two orders of magnitude — the discovery of dark matter.
Because $E=T+W=-T$ , gravitating systems have negative heat capacity: lose energy and they contract and heat up.
Key idea: for a bound, relaxed self-gravitating system, twice the time-averaged kinetic energy exactly cancels the gravitational potential energy, $2\langle T\rangle+\langle W\rangle=0$ , so the total energy is $E=T+W=-T=\tfrac12 W<0$ .

Where this lives today

The virial theorem is not a museum piece; it runs inside modern cosmology. Cluster masses from galaxy velocity dispersions feed the cluster mass function, whose evolution constrains the matter density and fluctuation amplitude — one of the cleaner tests of dark energy. The definition of a dark-matter halo in simulations is itself virial: the "virial radius" $R_{200}$ encloses a mean density two hundred times the cosmic mean, inside which material has relaxed enough to satisfy $2T+W\approx0$ . Dwarf spheroidal galaxies are weighed the same way, and their dispersions imply mass-to-light ratios in the hundreds or thousands, making them the most dark-matter-dominated systems known and prime targets for dark-matter searches. Closer to home, the negative heat capacity in $E=-T$ drives the gravothermal core collapse seen in old globular clusters, and the same budget sets the Jeans mass that gates star formation. Where precision is needed, the estimate is cross-checked against gravitational lensing, X-ray hydrostatic masses, and full Jeans and Schwarzschild modelling — but it remains the fast first weighing every other method is measured against.

The physical picture

Picture a swarm of stars, each on its own orbit inside the gravitational well the swarm itself digs. Too slow, and gravity wins and the swarm falls inward; too fast, and it escapes and evaporates. Equilibrium is the knife-edge between these fates, and the virial theorem names its exact location: the kinetic energy is precisely half the depth of the well, $2\langle T\rangle=-\langle W\rangle$ . This is the same $\langle T\rangle=-\tfrac12\langle W\rangle$ a single planet obeys, time-averaged over its Kepler ellipse — promoted from one body to an entire population. The moment of inertia $I=\sum_i m_i r_i^2$ measures how spread out the swarm is, and its second time derivative is exactly $2T+W$ . A relaxed system breathes but neither systematically expands nor contracts, so averaging over many crossing times sends $\langle\ddot I\rangle\to0$ and the balance emerges. A collapsing cloud, by contrast, has $\ddot I<0$ and is not virialized — which is why the theorem is about time averages, never instants.

Figure

A bound swarm of stars sits in the gravitational well it digs: gravity pulls inward while random stellar motions σ resist collapse, and equilibrium locks in at 2⟨T⟩ = −⟨W⟩, so the total energy is E = ½⟨W⟩ = −⟨T⟩ < 0

Because the total energy is $E=-T$ , gravity hides a paradox: inject energy into the system and it expands and cools, its random speeds falling as $E$ rises. That is negative heat capacity, $C=\mathrm dE/\mathrm dT=-1$ , the hallmark of self-gravity — it drives the gravothermal catastrophe of globular clusters and is why a star, radiating energy away, contracts and gets hotter. The virial theorem is the compact statement of all of it. The second figure is the payoff: the theorem turned into a scale. Feed it a dispersion and a size, and it returns a mass — and for Coma that mass towers over the visible starlight.

Figure

Zwicky's Coma estimate: with σ_los ≈ 1000 km/s and R ≈ 1.5 Mpc the estimator M ≈ 5σ²R/G gives a dynamical mass of 1.7×10¹⁵ M⊙, about 70 times the stellar mass 2.5×10¹³ M⊙ — a mass-to-light ratio near 300 and the first evidence for dark matter

The core equations

The whole subject condenses into one scalar relation. For a bound, relaxed, isolated system the time-averaged kinetic energy and gravitational potential energy are locked in a fixed ratio, independent of the detailed orbits.

2\langle T\rangle + \langle W\rangle = 0 \qquad (1)

where: $\langle T\rangle$ — time-averaged total kinetic energy, $T=\tfrac12\sum_i m_i v_i^2\ge0$ ; $\langle W\rangle$ — time-averaged total gravitational potential (binding) energy, $W<0$ . Reading: the kinetic energy is exactly half the magnitude of the potential energy, so a colder system is more tightly bound and a hotter one more diffuse. Equation (1) is the time-averaged shadow of an exact mechanical identity: differentiate the trace moment of inertia twice, and Newton's second law hands you the virial of Clausius on the right.

\tfrac12\ddot I = 2T + W,\qquad I=\sum_i m_i|\mathbf r_i|^2 \qquad (2)

where: $I$ — trace moment of inertia about the centre of mass; $\ddot I$ — its second time derivative, which vanishes on time-average in a steady state. Reading: nothing is averaged yet — (2) holds every instant, and (1) is what survives once $\langle\ddot I\rangle=0$ . The potential energy is the pairwise sum over all Newtonian bonds, and for any system it packages into a mass, a size, and a dimensionless structure factor carrying the shape of the density profile.

W = -\frac12\sum_i\sum_{j\neq i}\frac{Gm_im_j}{|\mathbf r_i-\mathbf r_j|} = -\alpha\,\frac{GM^2}{R} \qquad (3)

where: $M=\sum_i m_i$ — total mass; $R$ — characteristic (e.g. half-mass) radius; $\alpha$ — order-unity structure factor, $3/5$ for a uniform sphere and larger for concentrated profiles. Reading: all the geometry is quarantined inside $\alpha$ , the one place a mass estimate stays model-dependent. Combining (1) and (3) with $T=\tfrac12 M\langle v^2\rangle$ solves the equilibrium for the one thing you cannot see — the mass — in terms of the two things you can measure.

M \simeq \frac{\langle v^2\rangle\,R}{\alpha\,G} \qquad (4)

where: $\langle v^2\rangle=2T/M$ — mean-square 3D speed of the tracers; $R$ — the structural radius entering $W$ ; $\alpha$ — the same structure factor as in (3). Reading: mass scales as velocity squared times size, since $\sigma$ measures kinetic energy and the theorem trades it for binding energy $\propto GM^2/R$ . Observations give not the full 3D speed but a single line-of-sight component. For an isotropic velocity ellipsoid the three Cartesian variances are equal, so the 3D mean-square speed is three times the 1D dispersion.

\langle v^2\rangle = 3\,\sigma_{\mathrm{los}}^2 \qquad (5)

where: $\sigma_{\mathrm{los}}$ — observed line-of-sight velocity dispersion. Reading: forgetting this factor of three misweighs a cluster threefold; anisotropy replaces it with a case-dependent factor. The scalar theorem is the trace of a richer tensor statement. Spatial moments of the collisionless Boltzmann equation yield a nine-component relation whose steady-state form constrains the shape of a system, not just its energy.

2K_{jk} + W_{jk} = 0 \qquad (6)

where: $K_{jk}=\tfrac12\int\rho\,\overline{v_jv_k}\,\mathrm d^3x$ — kinetic-energy tensor (ordered plus random motion); $W_{jk}=-\tfrac12\int\rho\,x_j\,\partial\Phi/\partial x_k\,\mathrm d^3x$ (symmetrized) — Chandrasekhar potential-energy tensor. Reading: its unequal diagonal components let the theorem explain why a slowly rotating elliptical galaxy can be flattened by anisotropy rather than rotation. Finally, the theorem fixes the total energy and, through it, the counter-intuitive thermodynamics of gravity. Since $T=-\tfrac12 W$ , the total energy is negative and equals minus the kinetic energy.

E = T + W = -T = \tfrac12 W < 0 \qquad (7)

where: $E$ — total mechanical energy, negative for a bound system. Reading: the effective heat capacity is $C=\mathrm dE/\mathrm dT=-1$ , so radiating energy away raises the kinetic energy — a self-gravitating system heats as it loses energy.

Derivation AThe scalar virial theorem from the moment of inertia

The goal is to show that $2\langle T\rangle+\langle W\rangle=0$ is not an assumption but a consequence forced on any bound system of point masses by Newton's laws and the $1/r$ shape of gravity. We differentiate the trace moment of inertia twice, replace accelerations by gravitational forces, use the pairwise-force symmetry to collapse the double sum into the potential energy, and argue that the second derivative time-averages to zero for a system that stays finite forever. The one subtle input is that gravity is a degree $-1$ homogeneous potential, which is exactly what fixes the factor of two.

Step 1. Define $I=\sum_i m_i\,\mathbf r_i\cdot\mathbf r_i$ about a fixed centre of mass and differentiate once: $\tfrac12\dot I=\sum_i m_i\,\mathbf r_i\cdot\dot{\mathbf r}_i$ . Why: $I$ is the natural scalar measure of how spread out the system is, and its rate of change tracks whether the swarm is expanding or contracting.

Step 2. Differentiate again: $\tfrac12\ddot I=\sum_i m_i\,\dot{\mathbf r}_i\cdot\dot{\mathbf r}_i+\sum_i m_i\,\mathbf r_i\cdot\ddot{\mathbf r}_i$ . The first sum is manifestly $2T$ , and Newton's second law replaces $m_i\ddot{\mathbf r}_i$ by the net force $\mathbf F_i$ , so $\tfrac12\ddot I=2T+\sum_i\mathbf r_i\cdot\mathbf F_i$ . Why: this splits the geometry's acceleration into a kinetic piece and a force piece — the second sum is precisely Clausius's "virial".

Step 3. Insert the gravitational force $\mathbf F_i=-\sum_{j\neq i}Gm_im_j(\mathbf r_i-\mathbf r_j)/|\mathbf r_i-\mathbf r_j|^3$ and symmetrize the double sum by pairing $(i,j)$ with $(j,i)$ ; one power of the separation cancels, leaving $\sum_i\mathbf r_i\cdot\mathbf F_i=-\tfrac12\sum_i\sum_{j\neq i}Gm_im_j/|\mathbf r_i-\mathbf r_j|=W$ . Why: the virial of a $1/r$ force collapses exactly onto the potential energy — Euler's theorem for a degree $-1$ potential — which is what makes the coefficient of $T$ exactly two.

Step 4. Assemble the exact identity $\tfrac12\ddot I=2T+W$ and time-average over $\tau$ : $\langle\ddot I\rangle=[\dot I(\tau)-\dot I(0)]/\tau$ . For a bound system $I$ stays finite and $\dot I$ bounded forever, so as $\tau\to\infty$ this tends to zero. Why: boundedness is the whole content of "relaxed" — a system that neither disperses to infinity nor collapses to a point cannot have a persistently growing $\dot I$ .

Step 5. Setting $\langle\ddot I\rangle=0$ leaves $2\langle T\rangle+\langle W\rangle=0$ , and substituting into $E=T+W$ gives $E=\tfrac12\langle W\rangle=-\langle T\rangle<0$ . Why: the same identity that yields the virial relation pins the total energy at half the potential energy — the source of gravity's negative heat capacity. (Verified against Binney and Tremaine eq. 4.253 and Carroll and Ostlie section 2.4.)

Go deeper · Derivation B · 1 minThe virial mass estimator and Zwicky's dark-matter argument

Derivation BThe virial mass estimator and Zwicky's dark-matter argument

The aim is to turn the abstract balance $2T=-W$ into a number a telescope can deliver, and then to run that machine on the Coma cluster to reproduce the discovery of dark matter. We combine the kinetic energy as a mean-square speed, the potential energy as a structure factor, and the projection linking the observable line-of-sight dispersion to the full 3D speed. The result reads mass off from a dispersion and a radius, and the punchline is that for Coma this dynamical mass overshoots the starlight by nearly two orders of magnitude.

Step 1. Write $T=\tfrac12 M\langle v^2\rangle$ and $W=-\alpha GM^2/R$ , then impose $2T=-W$ : this gives $M\langle v^2\rangle=\alpha GM^2/R$ . Why: both sides now speak the same language of $M$ , $R$ , and a speed, so equilibrium becomes an algebraic equation for the mass.

Step 2. Cancel one power of $M$ to get the estimator $M=\langle v^2\rangle R/(\alpha G)$ . Why: one factor of mass measures how much stuff moves, the other how deep the well is; the theorem ties them and leaves a single power of $M$ to solve for.

Step 3. Replace the unobservable 3D speed using $\langle v^2\rangle=3\sigma_{\mathrm{los}}^2$ and adopt the uniform-sphere $\alpha=3/5$ : then $M=3\sigma_{\mathrm{los}}^2R/[(3/5)G]=5\sigma_{\mathrm{los}}^2R/G$ . Why: the catalogue-ready form; the prefactor $5$ is the projection factor $3$ times the geometric $5/3$ , and it becomes order-unity for any other profile.

Step 4. Insert Coma values $\sigma_{\mathrm{los}}\approx10^6\ \mathrm{m\,s^{-1}}$ and $R\approx1.5\ \mathrm{Mpc}=4.6\times10^{22}\ \mathrm m$ : $M\approx5(10^6)^2(4.6\times10^{22})/(6.67\times10^{-11})\approx3.5\times10^{45}\ \mathrm{kg}\approx1.7\times10^{15}\,M_\odot$ . Why: the arithmetic makes the abstraction physical — one dispersion and one size deliver a numerical cluster mass.

Step 5. Compare with the light: Coma's luminosity implies a stellar mass near $2.5\times10^{13}\,M_\odot$ , so the dynamical mass exceeds it by a factor near $70$ , a mass-to-light ratio of a few hundred solar. Why: no plausible amount of faint stars closes this gap — precisely Zwicky's 1933 conclusion that most of the cluster is dark, unchanged in spirit by ninety years of refinement. (Estimator verified against Carroll and Ostlie section 23.2 and Ryden chapter 8.)

Go deeper · Derivation C · 1 minThe tensor virial theorem from the collisionless Boltzmann equation

Derivation CThe tensor virial theorem from the collisionless Boltzmann equation

The purpose here is to lift the scalar theorem to its full tensor form and expose its deepest root: it is a moment of the collisionless Boltzmann equation, the master equation of stellar dynamics. We take the momentum moment to obtain a Jeans equation, multiply by a coordinate and integrate over all space, discard surface terms for a finite system, and identify the surviving integrals as the kinetic- and potential-energy tensors. Taking the trace and steady-state limit returns the scalar theorem, showing the moment-of-inertia and phase-space derivations are two faces of one result.

Step 1. Start from the Jeans (momentum) equation $\partial_t(\rho\bar v_k)+\partial_{x_j}(\rho\,\overline{v_jv_k})=-\rho\,\partial_{x_k}\Phi$ , the first velocity moment of the collisionless Boltzmann equation, with $\rho$ the density and $\Phi$ the potential. Why: the theorem is not about individual particles but the smooth phase-space flow, so the natural start is the fluid-like moment equation.

Step 2. Multiply by $x_l$ and integrate over space: the potential term becomes $-\int\rho\,x_l\,\partial_{x_k}\Phi\,\mathrm d^3x$ , which symmetrized in $(l,k)$ is the Chandrasekhar potential-energy tensor $W_{lk}$ . Why: weighting by position converts an equation about momentum flux into one about the second moments of the mass distribution.

Step 3. Integrate the stress term by parts, $-\int x_l\,\partial_{x_j}(\rho\,\overline{v_jv_k})\,\mathrm d^3x=\int\delta_{lj}\rho\,\overline{v_jv_k}\,\mathrm d^3x=2K_{lk}$ , dropping the surface term since $\rho$ vanishes at large radius. Why: the finite-extent assumption does real work — an embedded subsystem in a hot medium keeps that surface term and picks up an external-pressure correction.

Step 4. Collect the time-derivative piece into $\tfrac12\ddot I_{jk}$ with $I_{jk}=\int\rho\,x_jx_k\,\mathrm d^3x$ , giving the tensor virial theorem $\tfrac12\ddot I_{jk}=2K_{jk}+W_{jk}$ . Why: this is the full nine-component identity; its diagonal controls energy balance while its off-diagonal and anisotropic parts control the equilibrium shape of the system.

Step 5. Impose steady state ( $\ddot I_{jk}=0$ ) to get $2K_{jk}+W_{jk}=0$ , then take the trace, with $\mathrm{Tr}\,K_{jk}=T$ and $\mathrm{Tr}\,W_{jk}=W$ , recovering $2T+W=0$ . Why: the scalar theorem is literally the trace of the tensor one, so the two derivations must — and do — agree. (Follows Binney and Tremaine section 4.8, eqs. 4.247–4.253.)

Limiting cases

The virial relation sharpens, generalizes, or breaks down in a handful of instructive regimes; each row is a place where the naive $2T+W=0$ must be adjusted or abandoned.

Case	Result	Meaning
Isolated bound equilibrium	$2\langle T\rangle+\langle W\rangle=0$	The clean scalar theorem: no surface pressure, no external tides, steady state over many crossing times, with total energy $E=\tfrac12 W<0$ .
Confined by external pressure	$2T+W=3\oint P\,\mathbf r\cdot\mathrm d\mathbf S$	A subsystem embedded in a hot medium keeps the surface term; ignoring it overestimates the internal mass — the virial theorem of the interstellar and intracluster medium.
Single Kepler orbit (time-averaged)	$\langle T\rangle=-\tfrac12\langle W\rangle,\ E=-GMm/2a$	For one bound two-body orbit the relation holds as an average over a period; the semimajor axis $a$ fixes the energy — the two-body seed of the whole theorem.
Uniform-sphere structure factor	$W=-\tfrac35\,GM^2/R$	The reference case $\alpha=3/5$ ; concentrated profiles have larger $\alpha$ , so the same $\sigma$ and $R$ imply a smaller inferred mass.
Collapsing or merging system	$\tfrac12\ddot I=2T+W\neq0$	During collapse or a major merger the balance fails instantaneously; the system must relax by violent relaxation and phase mixing before any virial mass applies.

Common trapThe virial theorem holds at every instant, so a single snapshot gives the mass

It does not hold at every instant. The exact statement is $\tfrac12\ddot I=2T+W$ , and only the time-averaged version $2\langle T\rangle+\langle W\rangle=0$ is the virial theorem — it requires $\langle\ddot I\rangle=0$ , meaning the system neither systematically expands nor contracts over the window. A collapsing cloud, an ongoing merger, or a cluster caught mid-relaxation has $\ddot I\neq0$ and is not virialized; the estimator then gives a meaningless number. Two further traps ride along. First, $W=-\alpha GM^2/R$ carries a structure-dependent $\alpha$ that is $3/5$ only for a uniform sphere, so a mass quoted without the assumed profile is false precision. Second, $\langle v^2\rangle=3\sigma_{\mathrm{los}}^2$ holds only for isotropic orbits; anisotropy and unbound interlopers both bias the mass — which is why the theorem is a calibrator alongside Jeans and distribution-function modelling, not a final word.

Worked examples

1 · Weighing the Coma cluster and finding dark matter (fully worked)

The Coma cluster has a line-of-sight velocity dispersion $\sigma_{\mathrm{los}}\approx1000\ \mathrm{km\,s^{-1}}$ and a characteristic radius $R\approx1.5\ \mathrm{Mpc}$ . Estimate its dynamical mass, and compare with the stellar mass implied by its light, $M_{\mathrm{stars}}\approx2.5\times10^{13}\,M_\odot$ .

Step 1. Adopt the uniform-sphere estimator from Derivation 2, $M=5\sigma_{\mathrm{los}}^2R/G$ , and convert to SI: $\sigma_{\mathrm{los}}=10^6\ \mathrm{m\,s^{-1}}$ and $R=1.5\times3.086\times10^{22}\ \mathrm m=4.63\times10^{22}\ \mathrm m$ .

Step 2. Numerator: $5\sigma^2 R=5\times(10^6)^2\times4.63\times10^{22}=5\times10^{12}\times4.63\times10^{22}=2.31\times10^{35}$ (SI). Divide by $G=6.674\times10^{-11}$ : $M=2.31\times10^{35}/6.674\times10^{-11}=\mathbf{3.5\times10^{45}\ kg}$ .

Step 3. Convert to solar masses with $M_\odot=1.989\times10^{30}\ \mathrm{kg}$ : $M=3.5\times10^{45}/1.989\times10^{30}\approx\mathbf{1.7\times10^{15}\,M_\odot}$ — a thousand-trillion suns inferred from just a dispersion and a size.

Step 4. Compare with the light: $M/M_{\mathrm{stars}}=1.7\times10^{15}/2.5\times10^{13}\approx\mathbf{70}$ , and the dynamical mass-to-light ratio is a few hundred in solar units. The stars account for barely more than one percent of the mass; the rest is dark. M ≈ 1.7×10¹⁵ M⊙, about 70× the stellar mass — Zwicky's dark-matter conclusion, reproduced.

2 · A globular cluster has no dark matter

A globular cluster shows $\sigma_{\mathrm{los}}\approx7\ \mathrm{km\,s^{-1}}$ within a half-mass radius $R\approx3\ \mathrm{pc}$ , and shines with $L\approx8.5\times10^4\,L_\odot$ . Estimate its mass and mass-to-light ratio, and contrast with Coma.

Step 1. Same estimator $M\approx5\sigma_{\mathrm{los}}^2R/G$ as an order-of-magnitude mass. Convert: $\sigma_{\mathrm{los}}=7\times10^3\ \mathrm{m\,s^{-1}}$ , $R=3\times3.086\times10^{16}\ \mathrm m=9.26\times10^{16}\ \mathrm m$ .

Step 2. Numerator: $5\times(7\times10^3)^2\times9.26\times10^{16}=5\times4.9\times10^{7}\times9.26\times10^{16}=2.27\times10^{25}$ . Divide by $G$ : $M=2.27\times10^{25}/6.674\times10^{-11}\approx3.4\times10^{35}\ \mathrm{kg}\approx\mathbf{1.7\times10^{5}\,M_\odot}$ .

Step 3. Mass-to-light ratio: $M/L=1.7\times10^5/8.5\times10^4\approx\mathbf{2}$ in solar units. A value near $2$ is exactly what a population of ordinary stars gives — no dark matter is needed or implied. M ≈ 1.7×10⁵ M⊙, M/L ≈ 2 — a purely stellar system, in stark contrast to Coma's M/L ≈ 300. The two examples share one formula and reach opposite conclusions: the same theorem that reveals dark matter in a cluster confirms its absence in a globular. It does not decide the answer — it faithfully reports whatever the kinematics and size demand.

Physics of the portrait

The virial theorem is a piece of nineteenth-century thermodynamics that became, almost by accident, the twentieth century's first clue to dark matter. Rudolf Clausius stated it for a gas in 1870 and coined "virial" from the Latin vis, force. But its most consequential use belongs to a combative Caltech astronomer who pointed it at a swarm of galaxies and found most of their mass invisible.

Origins · 1 minFritz Zwicky and how this was discovered

Fritz Zwicky

1898–1974 · the Coma cluster, dark matter, and dynamical mass

Fritz Zwicky was a Bulgarian-born Swiss-American astrophysicist whose instinct for the big, uncomfortable result was matched only by his talent for antagonizing colleagues. In 1933, applying the virial theorem to the Coma cluster, he measured a velocity dispersion so large that the cluster could stay bound only if it held far more mass than its visible galaxies supplied — by his estimate a couple of hundred times more. He called the missing component dunkle Materie, dark matter, and concluded it dominated the universe's mass. The claim ran decades ahead of its acceptance: not until the 1970s, with Vera Rubin and Kent Ford's flat rotation curves, did dark matter become mainstream. Zwicky's range was extraordinary — with Walter Baade in 1934 he coined the term "supernova," proposed that supernovae mark the birth of neutron stars, and predicted they power cosmic rays. That the virial theorem is now the front-line tool for weighing dark-matter halos is his most enduring monument: a thermodynamic identity turned into a scale for the invisible.

History

1870

— Clausius derives the virial theorem for a bounded system and coins the term virial

1933

— Zwicky applies the theorem to the Coma cluster and infers unseen dark matter

1937

— Zwicky's expanded English analysis sharpens the cluster missing-mass problem

1967

— Lynden-Bell explains violent relaxation, how collisionless systems reach virial equilibrium

1969

— Chandrasekhar systematizes the tensor virial theorem in Ellipsoidal Figures of Equilibrium

Going further

Texts. Binney and Tremaine, Galactic Dynamics (2nd ed.), section 4.8 (tensor and scalar virial theorems, eqs. 4.247–4.253) and its applications · Carroll and Ostlie, An Introduction to Modern Astrophysics (2nd ed.), section 2.4 (moment-of-inertia derivation) and section 23.2 (virial masses of clusters) · Mo, van den Bosch and White, Galaxy Formation and Evolution, chapters 5 and 6 (virial equilibrium and halo structure) · Ryden, Introduction to Cosmology (2nd ed.), chapter 8 (dark matter and virial masses) · Dodelson, Modern Cosmology, chapter 1 (cluster masses and the case for dark matter). The key results to keep in mind are: (1) the identity $\tfrac12\ddot I=2T+W$ follows from Euler's theorem for a degree $-1$ potential; (2) the projection $\langle v^2\rangle=3\sigma_{\mathrm{los}}^2$ holds for isotropic orbits; (3) the uniform-sphere result $W=-\tfrac35 GM^2/R$ yields the prefactor $5$ in $M=5\sigma^2R/G$ ; (4) the energy relations $E=\tfrac12 W=-T$ and heat capacity $C=-1$ ; (5) the numerical Coma mass $\approx1.7\times10^{15}\,M_\odot$ and globular mass $\approx1.7\times10^5\,M_\odot$ from the worked examples.

Prerequisites. Gravitational N-Body Problem · Collisionless Boltzmann Equation · Distribution Function and Phase Space · Gravitational Two-Body Problem. Leads to. Jeans Equations · Two-Body Relaxation Time · Galactic Potential Models.

Ask the tutor: ask me to walk through the Coma mass estimate line by line, or how the tensor virial theorem explains an elliptical galaxy's flattening.

Why does the virial theorem carry a factor of exactly two, and not some other number?

The factor two comes from the shape of gravity, not from any choice. Gravity's $1/r$ potential is a homogeneous function of degree $-1$ , and Euler's theorem then makes the position-weighted force sum — Clausius's virial — equal to $-1$ times the potential energy. That $-1$ , combined with the kinetic term entering as $2T$ from the second derivative of $I$ , produces $2T+W=0$ . If gravity instead scaled as $r^2$ (a harmonic potential, degree $+2$ ), the same argument would give $T=W$ with no factor of two. The number two is a fingerprint of the inverse-square force law.

What is the difference between the scalar and tensor virial theorems, and when do I need the tensor one?

The scalar theorem, $2T+W=0$ , is a single equation about the energy budget — it tells you how fast things must move on average, and it is all you need for a mass. The tensor theorem, $2K_{jk}+W_{jk}=0$ , is nine equations (six independent) about the second moments of the mass and velocity distributions, constraining shape and internal dynamics, not just energy. You reach for the tensor form when the question is geometric: why is an elliptical galaxy flattened — is it spinning, or is its dispersion anisotropic? The tensor theorem relates the flattening to the ratio of ordered rotation to random-motion anisotropy, which the scalar version cannot see.

Why is the observed velocity dispersion multiplied by three to get the mass?

Because a spectrograph measures only one component of each star's velocity — the piece along your line of sight, through the Doppler shift. The kinetic energy, though, involves the full 3D speed, $\langle v^2\rangle=\langle v_x^2\rangle+\langle v_y^2\rangle+\langle v_z^2\rangle$ . If the orbits are isotropic, all three components share the same variance, equal to $\sigma_{\mathrm{los}}^2$ , so $\langle v^2\rangle=3\sigma_{\mathrm{los}}^2$ . Skip this factor and you underestimate the kinetic energy — and hence the mass — threefold. When orbits are anisotropic the factor is no longer exactly three, one of the leading systematic uncertainties in dispersion-based masses.

How can a self-gravitating system have negative heat capacity — doesn't that violate thermodynamics?

It violates the intuition built from ordinary gases, not thermodynamics itself. For a virialized system $E=-T$ , so raising $E$ lowers $T$ (the random kinetic energy that plays the role of temperature): $C=\mathrm dE/\mathrm dT=-1<0$ . The resolution is that self-gravitating systems are not extensive — gravity's long-range, unshielded nature means you cannot treat pieces as independent, and the standard theorems guaranteeing positive heat capacity do not apply. This is genuine and important: it makes gravitating systems thermodynamically unstable and directly causes the gravothermal catastrophe that collapses globular-cluster cores.

If Zwicky found dark matter in 1933, why did it take forty years to be accepted?

Several reasons converged. Zwicky's masses depended on the assumed cluster radius and on the distance scale (the Hubble constant), both poorly known and biased at the time, so the discrepancy could be argued down. His abrasive personality did not help, and a single hard-to-check cluster is easy to set aside. Acceptance required independent evidence: Rubin and Ford's flat rotation curves in the 1970s showed the same missing mass in ordinary spirals, X-ray gas revealed extra binding mass, and gravitational lensing gave a third, geometry-based weighing. When three unrelated methods agreed, the case became overwhelming — but the virial theorem got there first.

Does the virial theorem apply to a single star or the Solar System, or only to swarms?

It applies to both, as a time average. A planet on a bound Kepler orbit obeys $\langle T\rangle=-\tfrac12\langle W\rangle$ averaged over one period, which is why its total energy is $E=-GMm/2a$ , fixed by the semimajor axis alone. A star in hydrostatic equilibrium obeys the continuous version, $2U+\Omega=0$ , relating internal thermal energy $U$ to gravitational binding energy $\Omega$ — the reason a contracting protostar heats up. The theorem is scale-free: it governs a two-body orbit, a star's interior, a globular cluster, and a cluster of galaxies with the same line of algebra, provided the system is bound and averaged over its dynamical time.