Conservation Laws, Scaling, and the Virial Theorem

For a bound gravitating system, the virial relation connects kinetic energy to the homogeneity of the potential: in a steady or long-time-averaged state, $2\langle T\rangle+\langle U\rangle=0$ for inverse-square gravity. It fixes characteristic speeds and times but does not prove equilibrium, boundedness, or stability. The averaging hypotheses are therefore as important as the familiar formula.

Required background. Calculus and Linear Algebra provides multivariable differentiation, linear maps, inner products, and basis changes. Helpful background. Ordinary Differential Equations provides initial-value problems, linearization, and asymptotic solution methods.

Mechanical conservation laws

For particles with positions $\mathbf r_a$, momenta $\mathbf p_a$, and conservative potential $U$, time-translation, spatial-translation, and rotational symmetry give

$$E=T+U, \qquad \mathbf P=\sum_a\mathbf p_a, \qquad \mathbf L=\sum_a\mathbf r_a\times\mathbf p_a.$$

Energy conservation requires $U$ to be time independent. Momentum conservation requires invariance under a common translation, and angular momentum requires rotational invariance. For an isolated Newtonian $N$-body system,

$$U=-\sum_{a\lt b}\frac{Gm_am_b}{r_{ab}}\lt0.$$

The negative sign permits bound states. These symmetry conditions and conserved quantities are reviewed in Tong 2005, §§1.2–1.3, pp. 2–9, PDF.

Derivation of the virial identity

Define the scalar moment of inertia

$$I=\sum_a m_a r_a^2.$$

Two derivatives and Newton’s equations give

$$\frac12\ddot I =2T+\sum_a\mathbf r_a\cdot\mathbf F_a.$$

If $U(\lambda\mathbf r_a)=\lambda^kU(\mathbf r_a)$, Euler’s theorem gives $\sum_a\mathbf r_a\cdot\nabla_aU=kU$. Since $\mathbf F_a=-\nabla_aU$,

$$\frac12\ddot I=2T-kU.$$

Gravity has degree $k=-1$, hence

$$\frac12\ddot I=2T+U.$$

Average from $0$ to $\mathcal T$:

$$2\langle T\rangle_{\mathcal T}+\langle U\rangle_{\mathcal T} =\frac{\dot I(\mathcal T)-\dot I(0)}{2\mathcal T}.$$

Only if the boundary term vanishes as $\mathcal T\to\infty$—for example, for a bounded periodic system—does one obtain

$$\boxed{2\langle T\rangle+\langle U\rangle=0}.$$

Then $\langle E\rangle=\langle U\rangle/2=-\langle T\rangle\lt0$. This direct differentiation provides the independent sign check required for gravitational potential energy.

Golwala 2007, §1.3.2, pp. 47–51, PDF gives a complementary derivation that makes the boundedness and long-time-average assumptions explicit and applies the inverse-square result to astrophysical mass estimates.

Scales inferred from virial balance

For total mass $M$ distributed over size $R$, write schematically $T\sim Mv^2$ and $|U|\sim GM^2/R$. Virial balance gives

$$v^2\sim\frac{GM}{R}, \qquad t_{\rm dyn}\sim\frac{R}{v} \sim\sqrt{\frac{R^3}{GM}}.$$

The same time follows from the scaling symmetry of $\ddot{\mathbf r}=-GM\mathbf r/r^3$: $\mathbf r\mapsto\lambda\mathbf r$ requires $t\mapsto\lambda^{3/2}t$. This is the dimensional core of Kepler’s third law. The numerical coefficient in $U$ depends on the density profile, so the estimates determine scaling rather than an exact structure model.

For a circular Kepler orbit, $T=GMm/(2r)$ and $U=-GMm/r$, so $2T+U=0$ instantaneously. For an eccentric bound orbit it generally oscillates, while its orbital average vanishes. A collapsing or expanding system may have a large boundary term and need not be virialized even when $2T+U$ happens to cross zero at one instant.

The sign relation also explains a characteristic feature of self-gravitating systems. If a slowly evolving virialized system radiates a small amount of energy, its energy change is $\delta E\lt0$. At fixed mass and within the same structural family, $E=U/2=-T$ then implies $\delta T=-\delta E\gt0$: the remaining bound system becomes kinetically hotter as its total energy decreases and it contracts. This “negative heat-capacity” reasoning concerns a sequence of re-virialized states; the radiating subsystem is not energetically isolated during the transition. It does not automatically describe a gravitating system in an ordinary canonical heat bath. Rapid mass ejection, tidal work, or collapse invalidates the quasi-static inference.

Virial mass estimates invert the scaling $v^2\sim GM/R$, but their accuracy inherits every structural assumption: whether the tracers are bound, whether velocity components are isotropic, what density profile fixes the coefficient, and whether surface terms are negligible. The theorem supplies the dimensional backbone of such estimates; it does not turn a measured speed and projected radius into a model-independent mass.

The boundary term can be assessed directly rather than assumed away. Since $\dot I=2\sum_a m_a\mathbf r_a\mathbin{\cdot}\mathbf v_a$, the time-averaged theorem is trustworthy only when $\dot I/\mathcal T\to0$. Periodic motion satisfies this condition exactly over an integer number of periods. A statistically stationary bound system may satisfy it in a long-time sense. Homologous expansion with $R(t)$ growing indefinitely generally does not: then $I\sim MR^2$ and the endpoint contribution can remain comparable to $T$ and $U$.

For a continuum in a finite control volume, the same derivation acquires surface pressure and momentum-flux terms. The pressure contribution follows directly from integration by parts:

$$-\int_V \mathbf x\mathbin{\cdot}\nabla P\,dV =3\int_V P\,dV-\oint_{\partial V}P\,\mathbf x\mathbin{\cdot}\mathbf n\,dA.$$

For a fixed boundary with uniform pressure $P_{\rm s}$, the last integral is $3P_{\rm s}V$. If $K$ is bulk kinetic energy and $T=K+\tfrac32\int_VP\,dV$ includes the thermal energy of an isotropic monatomic ideal gas, then—assuming no mass or ram-pressure flux, no magnetic stress, and no external body force—the scalar identity becomes

$$\frac12\ddot I=2T+U-3P_{\rm s}V.$$

A stationary configuration with $\ddot I=0$, or a long-time average for which the averaged $\ddot I$ vanishes, therefore obeys $2T+U=3P_{\rm s}V$ (with averages understood in the latter case). Krumholz 2015, §6.1, pp. 91–95, PDF derives the more general Eulerian balance and retains thermal, magnetic, surface-stress, and momentum-flux terms; the compact coefficient above follows from the displayed pressure integral rather than being imported without its hypotheses. The relation encodes a particular boundary model, not an optional correction to the isolated-particle theorem. Likewise, an external tidal potential contributes its own virial $\int \rho\,\mathbf x\mathbin{\cdot}\mathbf g_{\rm ext}\,d^3x$. State the system boundary before diagnosing “virial equilibrium.”

Each reduction has a visible failure mode. If the surface pressure varies, the full surface integral must remain; a single $P_{\rm s}$ is then not defined. If the equation of state is not that of a nonrelativistic monatomic ideal gas, $3\int P\,dV/2$ need not equal the thermal energy, so the pressure integral should remain explicit. For an evolving boundary, the moment-of-inertia and flux terms must be derived for that moving volume. Finally, a time average applies to $P_{\rm s}V$ as well as to $T$ and $U$ unless the boundary data are stationary. These qualifications prevent a measured pressure from being inserted into the compact formula independently of the model that produced it.

Common pitfalls

Mistaking the theorem for a stability test. The relation constrains an average of an assumed bounded evolution. It neither guarantees that small perturbations remain small nor establishes thermodynamic equilibrium.

Omitting surface or external terms. A pressure-confined continuum, imposed tidal field, mass flux, or non-isolated system adds boundary or external-force contributions to the scalar virial equation.

Exercise: homogeneous potentials

For a bounded system with a potential homogeneous of degree $k$, derive the averaged relation and specialize it to a harmonic potential.

Solution

The identity above gives $\ddot I/2=2T-kU$. Vanishing of the averaged boundary term yields $2\langle T\rangle=k\langle U\rangle$. For $U\propto r^2$, $k=2$, so $\langle T\rangle=\langle U\rangle$. This differs in sign and proportion from gravity because the harmonic potential is positive and grows with distance.

Orbits and Celestial Dynamics applies these invariants and scales to exact conics, encounters, and perturbed motion.

References

• Golwala, Sunil. Lecture Notes on Classical Mechanics for Physics 106ab. California Institute of Technology, revision of January 15, 2007. §1.3.2, pp. 47–51, PDF.
• Krumholz, Mark R. Notes on Star Formation. Open Astrophysics Bookshelf, 2015, arXiv:1511.03457. §6.1, pp. 91–95, Open PDF.
• Tong, David. Classical Dynamics. University of Cambridge lecture notes, 2004/2005. §§1.2–1.3, pp. 2–9, PDF.