The Invariant Volume Element

Integration is the one operation that seems to resist the whole machinery of tensor calculus. A vector transforms cleanly, a tensor transforms cleanly, but the humble coordinate product $\mathrm{d}^n x = \mathrm{d}x^0\cdots\mathrm{d}x^{n-1}$ does not: under a change of chart it is rescaled by the Jacobian determinant of the transformation, exactly the way the change-of-variables theorem from ordinary calculus demands. So the innocent-looking integral $\int f\,\mathrm{d}^n x$ of a genuine scalar field $f$ gives a different number in polar coordinates than in Cartesian — a physically meaningless answer that depends on the grid you happened to draw. To integrate on a manifold at all, you need a measure that carries no memory of the chart. The metric supplies exactly the missing factor, through a quantity almost always misread as a scalar: the determinant $g = \det(g_{\mu\nu})$ . It is not a scalar. Because the metric transforms with two copies of the Jacobian matrix, its determinant transforms with the square of the Jacobian determinant — a scalar density of weight $-2$ . Take its square root and you obtain a density of weight $-1$ , the precise mirror image of $\mathrm{d}^n x$ , so the product $\sqrt{|g|}\,\mathrm{d}^n x$ has weight zero: it is invariant. Geometrically that combination is the Levi-Civita volume form, the unique metric-compatible top-degree form assigning unit volume to any orthonormal frame. This one factor $\sqrt{|g|}$ then reappears wherever serious physics is done on a curved background — in the Einstein-Hilbert action, in conserved charges, in the covariant continuity equation, and in the path-integral measure of curved-space field theory.

Key takeaways

The bare coordinate volume $\mathrm{d}^n x$ is not invariant: a change of chart multiplies it by the Jacobian determinant, so $\int f\,\mathrm{d}^n x$ depends on the coordinates and is physically meaningless on its own.
The metric determinant $g$ is a scalar density of weight $-2$ ( $g' = J^2 g$ ); its square root $\sqrt{|g|}$ has weight $-1$ and exactly cancels the weight $+1$ of $\mathrm{d}^n x$ .
The combination $\sqrt{|g|}\,\mathrm{d}^n x$ is the invariant volume element — a true $n$ -form, the Levi-Civita volume form $\boldsymbol{\epsilon}$ — and it is what makes integration on a manifold well defined.
The same factor turns the covariant divergence into an ordinary one: $\nabla_\mu V^\mu = \tfrac{1}{\sqrt{|g|}}\partial_\mu(\sqrt{|g|}\,V^\mu)$ , which is the engine behind conservation laws and Gauss's theorem in curved space.
In Lorentzian signature $g<0$ , so one writes $\sqrt{-g}$ ; in spherical coordinates $\sqrt{g}=r^2\sin\theta$ , recovering the familiar $\mathrm{d}V=r^2\sin\theta\,\mathrm{d}r\,\mathrm{d}\theta\,\mathrm{d}\phi$ .
Key idea: $\sqrt{|g|}$ is a scalar density of weight $-1$ that exactly cancels the Jacobian weight $+1$ of $\mathrm{d}^n x$ , so $\sqrt{|g|}\,\mathrm{d}^n x$ is the unique metric-built invariant measure for integrating scalars over a manifold.

Where this lives today

The invariant volume element is the silent partner in essentially every quantitative result in modern gravitational physics. In numerical relativity — the simulations behind the waveform templates LIGO and Virgo match against merging black holes — the factor $\sqrt{-g}$ (or its spatial cousin $\sqrt{{}^{(3)}g}$ in the 3+1 ADM split) weights every grid cell, and evolving it correctly is what keeps a simulated spacetime's total mass and angular momentum conserved to the precision a usable template demands. In cosmology the same measure defines the comoving volume that turns a galaxy survey's raw counts into a physical number density, and hence into the structure-growth statistics that constrain dark energy. And in quantum field theory on curved spacetime — the framework behind Hawking radiation and the primordial spectrum imprinted on the microwave background — the path-integral measure carries an implicit $\sqrt{-g}$ so that the partition function is diffeomorphism invariant. From a laptop lattice calculation to a gravitational-wave detector the rule is identical: if you want a number every observer agrees on, you integrate against $\sqrt{|g|}\,\mathrm{d}^n x$ .

The physical picture

Picture the smallest possible cell of your coordinate grid: a little box with edges $\mathrm{d}x^\mu$ . In flat space with Cartesian axes that box has volume equal to the plain product of its edge lengths, and $\mathrm{d}V = \mathrm{d}x\,\mathrm{d}y\,\mathrm{d}z$ needs no correction. But the moment the coordinates are curvilinear — polar, spherical, or the warped coordinates of a black hole — the same coordinate box is stretched and skewed in real space. A step $\mathrm{d}\phi$ in longitude covers a huge arc near the equator and almost nothing near a pole. The bare $\mathrm{d}^n x$ counts coordinate boxes, which is an arbitrary bookkeeping choice; the metric knows how much physical volume each box actually occupies. The factor $\sqrt{|g|}$ is precisely the conversion between the two, and in spherical coordinates it reproduces the schoolbook $r^2\sin\theta$ without any special pleading.

Figure

The metric converts a coordinate box dr dθ dφ into a physical box with edges dr, r dθ, r sinθ dφ. Its volume is √g dr dθ dφ = r²sinθ dr dθ dφ, and at r = 2, θ = 90° the conversion factor is √g = 4

Now watch two observers use different grids over the same region. Where one draws large boxes, $\mathrm{d}^n x$ is big but $\sqrt{|g|}$ is correspondingly small, because each big box is metrically thin; where the other draws small boxes, $\mathrm{d}^n x$ is small but $\sqrt{|g|}$ is large. The product $\sqrt{|g|}\,\mathrm{d}^n x$ is the same physical sliver of volume either way. This is not a lucky coincidence but a forced one: the coordinate volume carries Jacobian weight $+1$ , so a chart-independent measure needs a partner of weight $-1$ , and the metric determinant is the only natural scalar of that weight the geometry offers. The volume form is unique up to orientation and scale.

Figure

The weight ledger for a coordinate change. Under a uniform doubling of the coordinates the coordinate volume d³x grows by a factor of 8 while √g shrinks by the same factor of 8, so the product √g d³x is unchanged: 8 × ⅛ = 1

The core equations

Everything begins with how the plain coordinate volume responds to a smooth change of chart. This is nothing but the multivariable change-of-variables theorem: the new differentials relate to the old through the absolute Jacobian of the map, so $\mathrm{d}^n x$ behaves as a density rather than a scalar.

\mathrm{d}^n x'=\left|\det\!\frac{\partial x'^{\mu'}}{\partial x^{\mu}}\right|\,\mathrm{d}^n x \equiv |J|^{-1}\,\mathrm{d}^n x \qquad (1)

where: $\mathrm{d}^n x$ — coordinate volume product $\mathrm{d}x^0\cdots\mathrm{d}x^{n-1}$ ; $J=\det(\partial x^\mu/\partial x'^{\mu'})$ — Jacobian determinant of the inverse map; $|J|^{-1}$ — the factor that marks $\mathrm{d}^n x$ as a scalar density of weight $+1$ . Reading it physically: the coordinate box is a chart artifact, not a geometric quantity, so it cannot stand alone in an invariant integral. The metric, being a $(0,2)$ tensor, transforms with two factors of the Jacobian matrix. This double occurrence is the whole story: it is what promotes the determinant from weight $-1$ (which a single vector index would give) to weight $-2$ .

g_{\mu'\nu'}=\frac{\partial x^{\mu}}{\partial x'^{\mu'}}\frac{\partial x^{\nu}}{\partial x'^{\nu'}}\,g_{\mu\nu} \qquad (2)

where: $g_{\mu\nu}$ — metric components in the original chart; $\partial x^\mu/\partial x'^{\mu'}$ — Jacobian matrix $M$ of the transformation. In matrix form this is a congruence, $g' = M^{\mathsf T} g\, M$ , and congruence is what squares the determinant. Taking the determinant of that congruence, and using both $\det(M^{\mathsf T})=\det M$ and the multiplicativity of the determinant, yields the central density law of this article.

g'=\det(g_{\mu'\nu'})=J^{2}\,g,\qquad g\equiv\det(g_{\mu\nu}) \qquad (3)

where: $g$ — metric determinant in the old chart; $g'$ — in the new chart; $J^2$ — square of the Jacobian, one power for each metric index. The two powers of $J$ are the visible signature of weight $-2$ ; a single square root will halve that to the weight $-1$ we need. Because the determinant carries $J^2$ , its absolute square root carries a single $|J|$ — and the absolute value, taken before the root, is what keeps the measure real and positive in Lorentzian signature where $g<0$ .

\sqrt{|g'|}=|J|\,\sqrt{|g|} \qquad (4)

where: $\sqrt{|g|}$ — square root of the absolute value of the metric determinant; $|J|$ — absolute Jacobian. This is a scalar density of weight $-1$ : the exact partner that $\mathrm{d}^n x$ was missing. Multiplying (4) by (1), the $|J|$ and $|J|^{-1}$ annihilate identically. What remains is a quantity every observer computes to the same value — the invariant volume element.

\sqrt{|g'|}\,\mathrm{d}^n x' = |J|\sqrt{|g|}\cdot |J|^{-1}\,\mathrm{d}^n x = \sqrt{|g|}\,\mathrm{d}^n x \qquad (5)

where: $\sqrt{|g|}\,\mathrm{d}^n x$ — the invariant volume element (measure). For any scalar field $f$ , the integral $\int f\sqrt{|g|}\,\mathrm{d}^n x$ is a single coordinate-independent number. The same object has a second life as the totally antisymmetric Levi-Civita tensor, which must be sharply distinguished from the Levi-Civita symbol $\tilde\epsilon$ that is $\pm1,0$ in every chart by fiat. The symbol is a density; multiplying by $\sqrt{|g|}$ (or dividing, with a sign, for the raised version) produces a genuine tensor.

\epsilon_{\mu_1\cdots\mu_n}=\sqrt{|g|}\,\tilde\epsilon_{\mu_1\cdots\mu_n},\qquad \epsilon^{\mu_1\cdots\mu_n}=\frac{\mathrm{sgn}(g)}{\sqrt{|g|}}\,\tilde\epsilon^{\mu_1\cdots\mu_n} \qquad (6)

where: $\tilde\epsilon_{\mu_1\cdots\mu_n}$ — Levi-Civita symbol, a weight $+1$ density; $\epsilon_{\mu_1\cdots\mu_n}$ — the true $(0,n)$ tensor; $\mathrm{sgn}(g)$ — $+1$ for Riemannian, $-1$ for Lorentzian signature, from raising all $n$ indices with $g^{\mu\nu}$ . Packaged as a differential form, the same components define the Levi-Civita volume $n$ -form, the object on which Stokes' theorem and the Hodge star are built.

\boldsymbol{\epsilon}=\sqrt{|g|}\,\mathrm{d}x^0\wedge\mathrm{d}x^1\wedge\cdots\wedge\mathrm{d}x^{n-1} \qquad (7)

where: $\boldsymbol{\epsilon}$ — Levi-Civita volume form; $\wedge$ — exterior product, encoding antisymmetry and orientation. It is covariantly constant, $\nabla_\alpha\epsilon_{\mu_1\cdots\mu_n}=0$ , because the connection is metric-compatible. From covariant constancy of $\boldsymbol{\epsilon}$ (equivalently, from a determinant identity) follows the identity that makes the whole apparatus useful for physics: the covariant divergence of a vector is an ordinary coordinate divergence of $\sqrt{|g|}\,V^\mu$ .

\nabla_\mu V^\mu=\frac{1}{\sqrt{|g|}}\,\partial_\mu\!\left(\sqrt{|g|}\,V^\mu\right) \qquad (8)

where: $\nabla_\mu$ — Levi-Civita (metric-compatible) covariant derivative; $V^\mu$ — contravariant vector field. This is the curved-space form of $\nabla\cdot\mathbf V$ ; it collapses a conservation law into a plain divergence and, once integrated against $\sqrt{|g|}\,\mathrm{d}^n x$ , delivers Gauss's theorem. Finally, the reason the whole subject matters for dynamics: any action that is to be a coordinate-independent number must integrate a scalar Lagrangian against the invariant measure, which fixes the appearance of $\sqrt{-g}$ in the Einstein-Hilbert action.

S=\int \mathcal{L}\,\sqrt{-g}\,\mathrm{d}^4 x,\qquad S_{\mathrm{EH}}=\frac{1}{16\pi G}\int R\,\sqrt{-g}\,\mathrm{d}^4 x \qquad (9)

where: $\mathcal{L}$ — scalar Lagrangian; $R$ — Ricci scalar; $G$ — Newton's constant. Varying $\sqrt{-g}\,R$ with respect to $g^{\mu\nu}$ produces Einstein's field equations, and the variation of the measure itself, $\delta\sqrt{-g}$ , is a piece of that calculation (Derivation 3).

Derivation AInvariance of the metric volume element

The goal is to prove from the ground up that $\sqrt{|g|}\,\mathrm{d}^n x$ is genuinely the same in every chart, by tracking how each of its two factors responds to a smooth coordinate change and showing that their Jacobian weights are equal and opposite. This is the theorem that licenses every integral in general relativity, so it deserves to be seen not as a slogan about weights cancelling but as an explicit computation from the transformation laws of the coordinate volume and the metric determinant, ending in a term-by-term cancellation of the Jacobian factor. We fix a change $x^\mu\to x'^{\mu'}$ with matrix $M^\mu{}_{\mu'}=\partial x^\mu/\partial x'^{\mu'}$ and $J=\det M$ .

Step 1. The coordinate volume picks up $|J|^{-1}$ . The change-of-variables theorem gives $\mathrm{d}^n x' = |\det(\partial x'/\partial x)|\,\mathrm{d}^n x = |J|^{-1}\mathrm{d}^n x$ , since the forward Jacobian is the reciprocal of $J$ as defined through the inverse map. Why: this is where the chart-dependence of naive integration is quantified — the coordinate box is a density of weight $+1$ , nothing more.

Step 2. The metric determinant picks up $J^2$ . The tensor law $g_{\mu'\nu'}=M^\mu{}_{\mu'}M^\nu{}_{\nu'}g_{\mu\nu}$ is the matrix congruence $g'=M^{\mathsf T}gM$ ; taking determinants and using $\det(M^{\mathsf T})=\det M$ gives $g' = (\det M)^2 g = J^2 g$ , hence $|g'|=J^2|g|$ . Why: the two lowered indices of the metric each contribute one Jacobian factor, and only because they multiply do we get a square rather than a single power.

Step 3. The square root picks up $|J|$ . Taking the positive root of $|g'|=J^2|g|$ gives $\sqrt{|g'|}=|J|\sqrt{|g|}$ , where the absolute value is essential: it keeps the measure positive regardless of whether the transformation preserves or flips orientation, and it is well defined even in Lorentzian signature where $g$ itself is negative. Why: halving the weight from $-2$ to $-1$ is the entire purpose of the square root — it is what makes the metric determinant a usable partner for $\mathrm{d}^n x$ .

Step 4. Multiply the two factors. Combining Steps 1 and 3, $\sqrt{|g'|}\,\mathrm{d}^n x' = (|J|\sqrt{|g|})(|J|^{-1}\mathrm{d}^n x) = \sqrt{|g|}\,\mathrm{d}^n x$ . The Jacobian cancels identically, with no leftover. Why: the equality of the two numbers in any pair of charts is precisely the statement that $\int f\sqrt{|g|}\,\mathrm{d}^n x$ is coordinate-independent for a scalar $f$ — the manifold now has a well-defined notion of volume.

Go deeper · Derivation B · 1 minThe covariant divergence formula

Derivation BThe covariant divergence formula

The aim is to derive $\nabla_\mu V^\mu = \tfrac{1}{\sqrt{|g|}}\partial_\mu(\sqrt{|g|}V^\mu)$ , the workhorse identity that converts a curved-space divergence into an ordinary coordinate divergence and thereby makes conservation laws tractable. The route runs through the contracted Christoffel symbol and Jacobi's determinant identity; the payoff is that the messy connection terms collapse into a single logarithmic derivative of $\sqrt{|g|}$ , exposing the measure factor hiding inside the covariant derivative. We take $V^\mu$ a smooth vector field and $\nabla$ the Levi-Civita connection.

Step 1. Expand the divergence. By definition $\nabla_\mu V^\mu = \partial_\mu V^\mu + \Gamma^\mu{}_{\mu\nu}V^\nu$ , the single connection term coming from the one upper index, contracted against the derivative index. Why: the whole difference between $\partial_\mu V^\mu$ and the true divergence lives in that contracted Christoffel symbol, so if we can evaluate it in closed form we are done.

Step 2. Reduce the contracted Christoffel symbol. Starting from $\Gamma^\mu{}_{\mu\nu}=\tfrac12 g^{\mu\lambda}(\partial_\mu g_{\lambda\nu}+\partial_\nu g_{\mu\lambda}-\partial_\lambda g_{\mu\nu})$ , the first and third terms cancel against each other under the symmetric contraction with $g^{\mu\lambda}$ , leaving $\Gamma^\mu{}_{\mu\nu}=\tfrac12 g^{\mu\lambda}\partial_\nu g_{\mu\lambda}$ . Why: this isolates a trace $g^{\mu\lambda}\partial_\nu g_{\mu\lambda}$ that is exactly the shape Jacobi's identity resolves.

Step 3. Apply Jacobi's determinant identity. For any invertible matrix, $\partial_\nu\ln\det = \mathrm{tr}(g^{-1}\partial_\nu g)$ ; for the symmetric metric this reads $\partial_\nu\ln|g| = g^{\mu\lambda}\partial_\nu g_{\mu\lambda}$ . Substituting, $\Gamma^\mu{}_{\mu\nu}=\tfrac12\partial_\nu\ln|g|=\tfrac1{\sqrt{|g|}}\partial_\nu\sqrt{|g|}$ . Why: the identity is what turns an implicit trace over metric derivatives into an explicit derivative of the measure $\sqrt{|g|}$ — the measure was inside the connection all along.

Step 4. Assemble the formula. Substitute back: $\nabla_\mu V^\mu = \partial_\mu V^\mu + \tfrac{V^\mu}{\sqrt{|g|}}\partial_\mu\sqrt{|g|} = \tfrac1{\sqrt{|g|}}\partial_\mu(\sqrt{|g|}V^\mu)$ , recognizing the product rule in reverse. Why: integrating both sides against $\sqrt{|g|}\,\mathrm{d}^n x$ now turns $\int \nabla_\mu V^\mu\sqrt{|g|}\,\mathrm{d}^n x$ into $\int\partial_\mu(\sqrt{|g|}V^\mu)\mathrm{d}^n x$ , a pure boundary term — the curved-space Gauss theorem.

Go deeper · Derivation C · 1 minVarying the measure and the origin of the field equations

Derivation CVarying the measure and the origin of the field equations

The purpose here is to compute $\delta\sqrt{-g}$ under a variation of the inverse metric $\delta g^{\mu\nu}$ — the identity without which one cannot vary the Einstein-Hilbert action — and to show the measure itself contributes a term proportional to $g_{\mu\nu}$ in the field equations. This is where the volume element stops being a bookkeeping device and becomes dynamical: because the action carries $\sqrt{-g}$ , gravity feels the variation of the measure, and that variation supplies exactly the metric term sitting opposite the curvature in Einstein's equations. We use Jacobi's identity once more, in variational form.

Step 1. Vary the determinant. Jacobi's identity in variational form gives $\delta g = g\,g^{\mu\nu}\delta g_{\mu\nu}$ , where $g=\det g_{\mu\nu}$ . Equivalently, since $\delta(g_{\mu\nu}g^{\nu\lambda})=0$ implies $g^{\mu\nu}\delta g_{\mu\nu}=-g_{\mu\nu}\delta g^{\mu\nu}$ , we may write $\delta g = -g\,g_{\mu\nu}\delta g^{\mu\nu}$ . Why: expressing the variation through the inverse metric $\delta g^{\mu\nu}$ is what matches the variable one actually varies in the gravitational action.

Step 2. Take the square root. With $g<0$ in Lorentzian signature, $\delta\sqrt{-g} = \tfrac{-\delta g}{2\sqrt{-g}}$ . Substituting $-\delta g = g\,g_{\mu\nu}\delta g^{\mu\nu}$ and using $-g=(\sqrt{-g})^2$ gives $\delta\sqrt{-g} = -\tfrac12\sqrt{-g}\,g_{\mu\nu}\delta g^{\mu\nu}$ . Why: this compact result is the entire contribution of the measure to any metric variation; every $\sqrt{-g}$ in an action produces this factor when varied.

Step 3. Read off the consequence. Varying $S_{\mathrm{EH}}=\tfrac1{16\pi G}\int R\sqrt{-g}\,\mathrm{d}^4x$ , the term $R\,\delta\sqrt{-g} = -\tfrac12 R\,g_{\mu\nu}\sqrt{-g}\,\delta g^{\mu\nu}$ combines with $\sqrt{-g}\,\delta R$ to give the vacuum field equations $R_{\mu\nu}-\tfrac12 R\,g_{\mu\nu}=0$ . Why: the $-\tfrac12 R g_{\mu\nu}$ that completes the Einstein tensor comes directly from the variation of the volume element — the measure is not passive, it is half of the field equations. *The key result of this derivation is $\delta\sqrt{-g}=-\tfrac12\sqrt{-g}\,g_{\mu\nu}\delta g^{\mu\nu}$ .*

Limiting cases

How $\sqrt{|g|}$ specializes across the coordinate systems and signatures one meets in practice:

Case	Result	Meaning
Flat Euclidean, Cartesian	$\sqrt{g}=1,\quad \mathrm{d}V=\mathrm{d}x\,\mathrm{d}y\,\mathrm{d}z$	$g_{ij}=\delta_{ij}$ so $g=1$ ; the naive measure is already correct.
Spherical coordinates $(r,\theta,\phi)$	$\sqrt{g}=r^2\sin\theta$	From $g_{rr}=1,\ g_{\theta\theta}=r^2,\ g_{\phi\phi}=r^2\sin^2\theta$ , so $g=r^4\sin^2\theta$ .
Minkowski, inertial coordinates	$g=\det\eta_{\mu\nu}=-1,\quad \sqrt{-g}=1$	Lorentzian signature forces $g<0$ ; one uses $\sqrt{-g}$ throughout, equal to $1$ here.
Schwarzschild coordinates	$\sqrt{-g}=r^2\sin\theta$	The $g_{tt}=-(1-2GM/r)$ and $g_{rr}=(1-2GM/r)^{-1}$ factors cancel in the determinant, leaving the angular part.
Weak field $g_{\mu\nu}=\eta_{\mu\nu}+h_{\mu\nu}$	$\sqrt{-g}\approx 1+\tfrac12 h,\ h=\eta^{\mu\nu}h_{\mu\nu}$	To first order the measure tracks the trace of the perturbation — the volume response to a passing wave.

Common trapThe Levi-Civita symbol and the Levi-Civita tensor are the same thing.

They are not. The symbol $\tilde\epsilon_{\mu_1\cdots\mu_n}$ is defined to be $+1,-1,0$ in every coordinate system — it never transforms, which is exactly why it is a density of weight $+1$ and not a tensor. The tensor $\epsilon_{\mu_1\cdots\mu_n}=\sqrt{|g|}\,\tilde\epsilon_{\mu_1\cdots\mu_n}$ does transform correctly and is the object you may legitimately contract with other tensors. Conflating them is the single most common error in index gymnastics, and it silently drops factors of $\sqrt{|g|}$ . A companion trap: in Lorentzian signature $g<0$ , so writing $\sqrt{g}$ instead of $\sqrt{-g}$ produces an imaginary volume — always take the absolute value before the root.

Worked examples

1 · The volume of a ball, done the invariant way (confidence builder)

Compute the volume of a Euclidean ball of radius $R=2$ using the invariant measure in spherical coordinates, and confirm it reproduces $\tfrac43\pi R^3$ .

Step 1. Build $\sqrt{g}$ . The flat metric in spherical coordinates has $g_{rr}=1$ , $g_{\theta\theta}=r^2$ , $g_{\phi\phi}=r^2\sin^2\theta$ , and is diagonal, so $g=\det = r^2\cdot r^2\sin^2\theta = r^4\sin^2\theta$ and $\sqrt{g}=r^2\sin\theta$ . The invariant volume element is $\mathrm{d}V=r^2\sin\theta\,\mathrm{d}r\,\mathrm{d}\theta\,\mathrm{d}\phi$ — the $r^2\sin\theta$ everyone memorizes is nothing but $\sqrt{g}$ .

Step 2. Integrate over the ball. The three integrals separate: $V=\int_0^{2}r^2\,\mathrm{d}r\int_0^{\pi}\sin\theta\,\mathrm{d}\theta\int_0^{2\pi}\mathrm{d}\phi = \big(\tfrac{R^3}{3}\big)(2)(2\pi)$ . With $R=2$ the radial factor is $8/3$ , and $V=(8/3)(2)(2\pi)=\tfrac{32\pi}{3}$ .

Step 3. Evaluate and check. Numerically $V=\tfrac{32\pi}{3}=\mathbf{33.51}$ , and the closed form $\tfrac43\pi R^3=\tfrac43\pi(8)=\tfrac{32\pi}{3}$ agrees exactly. Had we integrated the bare $\mathrm{d}^3x=\mathrm{d}r\,\mathrm{d}\theta\,\mathrm{d}\phi$ instead, we would have obtained the meaningless coordinate volume $2\cdot\pi\cdot2\pi=4\pi^2\approx39.5$ — a number with no geometric content. The invariant measure gives $V=32\pi/3\approx33.51$ ; the factor $\sqrt{g}=r^2\sin\theta$ is what makes the answer real.

2 · Proper volume of a shell just outside a mass

A static observer measures the proper 3-volume of a thin spherical shell of coordinate thickness $\Delta r$ at Schwarzschild radial coordinate $r$ , where $r_s/r=0.2$ with $r_s=2GM$ the Schwarzschild radius. By how much does it exceed the naive Euclidean shell volume?

Step 1. Identify the spatial measure. On a surface of constant $t$ the induced 3-metric has $g_{rr}=(1-r_s/r)^{-1}$ , $g_{\theta\theta}=r^2$ , $g_{\phi\phi}=r^2\sin^2\theta$ , so the spatial determinant is $(1-r_s/r)^{-1}r^4\sin^2\theta$ and $\sqrt{{}^{(3)}g}=(1-r_s/r)^{-1/2}\,r^2\sin\theta$ . The radial stretch factor $(1-r_s/r)^{-1/2}$ is the whole difference from flat space.

Step 2. Integrate over the shell. The angular integral of $r^2\sin\theta$ over the full sphere gives $4\pi r^2$ , so the proper shell volume is $V_{\text{proper}}=(1-r_s/r)^{-1/2}\,4\pi r^2\,\Delta r$ , against the Euclidean $V_{\text{flat}}=4\pi r^2\,\Delta r$ .

Step 3. Put in the number. With $r_s/r=0.2$ , the stretch factor is $(1-0.2)^{-1/2}=(0.8)^{-1/2}=\mathbf{1.118}$ . The proper volume therefore exceeds the Euclidean estimate by 11.8 %: the same coordinate shell holds more real volume because radial distances are stretched near the mass. At a milder $r_s/r=0.1$ the excess drops to $(0.9)^{-1/2}-1=5.4\%$ , and it diverges as $r\to r_s$ — the geometric signature of the horizon. The metric determinant turns a coordinate shell into 11.8 % more proper volume at $r_s/r=0.2$ .

3 · Recovering the spherical divergence from equation (8)

Use $\nabla_\mu V^\mu=\tfrac1{\sqrt{g}}\partial_\mu(\sqrt{g}V^\mu)$ to reproduce the familiar vector-calculus divergence in spherical coordinates, without ever computing a Christoffel symbol.

Step 1. Insert the measure. With $\sqrt{g}=r^2\sin\theta$ independent of $\phi$ , equation (8) reads $\nabla_\mu V^\mu = \tfrac{1}{r^2\sin\theta}\big[\partial_r(r^2\sin\theta\,V^r)+\partial_\theta(r^2\sin\theta\,V^\theta)+\partial_\phi(r^2\sin\theta\,V^\phi)\big]$ .

Step 2. Pull constants through each derivative. The $\sin\theta$ is constant under $\partial_r$ and the $r^2$ under $\partial_\theta$ , and both are constant under $\partial_\phi$ , giving $\nabla_\mu V^\mu = \tfrac{1}{r^2}\partial_r(r^2 V^r)+\tfrac{1}{\sin\theta}\partial_\theta(\sin\theta\,V^\theta)+\partial_\phi V^\phi$ .

Step 3. Recognize the result. This is exactly the standard spherical divergence (in coordinate-basis components), with the characteristic $1/r^2$ and $1/\sin\theta$ weightings emerging purely from differentiating $\sqrt{g}$ . The measure factor $\sqrt{g}=r^2\sin\theta$ alone reproduces the entire spherical divergence — the connection coefficients are encoded in the derivative of the volume element.

Physics of the portrait

The invariant volume element is a monument to a nineteenth-century idea that had to wait for twentieth-century physics to find its purpose: that geometry itself could be made calculational, index by index, on any manifold.

Origins · 1 minTullio Levi-Civita and how this was discovered

Tullio Levi-Civita

1873–1941 · Italian mathematician, co-founder of tensor calculus

Tullio Levi-Civita, a student of Gregorio Ricci-Curbastro at Padua, is the mathematician whose name the volume form, the tensor, and the metric connection all carry. With Ricci he wrote the 1900 memoir Méthodes de calcul différentiel absolu et leurs applications, the document that turned Riemann's and Christoffel's scattered results into a systematic absolute differential calculus — the tensor calculus in which densities, weights, and invariant integration first became precise tools rather than ad hoc tricks. It was this calculus that Marcel Grossmann pressed into Einstein's hands in Zurich and that Einstein leaned on through the long struggle toward general relativity; the two later corresponded on the field equations, with Einstein voicing open admiration for Levi-Civita's command of the formalism. In 1917 Levi-Civita supplied the geometric heart the calculus had lacked: parallel transport, which gave the metric-compatible connection an intrinsic meaning and made covariant constancy a picture rather than an equation. The invariant measure $\sqrt{|g|}\,\mathrm{d}^n x$ sits squarely inside his legacy — it is the density of weight zero his calculus was built to recognize, and the Levi-Civita tensor $\epsilon_{\mu_1\cdots\mu_n}=\sqrt{|g|}\,\tilde\epsilon_{\mu_1\cdots\mu_n}$ that expresses it bears his name to this day. Dismissed from his Rome chair in 1938 under the Fascist racial laws, he died in 1941, but the apparatus he and Ricci built remains the working language of every calculation here.

History

1854

— Riemann's habilitation lecture founds the geometry of curved manifolds and the line element

1869

— Christoffel introduces the symbols governing covariant differentiation

1900

— Ricci-Curbastro and Levi-Civita codify the absolute differential calculus of tensors and densities

1915

— Hilbert derives the field equations by varying the action integral of R over √(−g) d⁴x

1917

— Levi-Civita defines parallel transport, grounding the metric-compatible connection geometrically

Going further

References. Misner, Thorne & Wheeler, Gravitation (1973), section 3.3, section 8.3 and Box 5.4 · Wald, General Relativity (1984), Appendix B (section B.2) and section 2.4 · Carroll, Spacetime and Geometry (2004), sections 2.8–2.10 and 4.3 · Schutz, A First Course in General Relativity (2009, 2nd ed.), sections 6.2–6.4 · Hartle, Gravity (2003), section 20.4. The central identities of this article are: (1) $g'=J^2 g$ from the metric congruence $g'=M^{\mathsf T}gM$ and $\det(M^{\mathsf T})=\det M$ ; (2) $\sqrt{|g'|}\,\mathrm{d}^n x'=\sqrt{|g|}\,\mathrm{d}^n x$ by cancellation of $|J|$ ; (3) $\Gamma^\mu{}_{\mu\nu}=\partial_\nu\ln\sqrt{|g|}$ via Jacobi's identity; (4) $\delta\sqrt{-g}=-\tfrac12\sqrt{-g}\,g_{\mu\nu}\delta g^{\mu\nu}$ ; (5) the tabulated values $\sqrt{g}=r^2\sin\theta$ (spherical), $\sqrt{-g}=1$ (Minkowski), $\sqrt{-g}=r^2\sin\theta$ (Schwarzschild); and (6) the worked numbers $V=32\pi/3\approx33.51$ for the $R=2$ ball and the $11.8\%$ proper-volume excess at $r_s/r=0.2$ .

Prerequisites. The Metric Tensor and the Line Element · The Tensor Transformation Law · Tensor Densities and Weights · Metric Signature and Lorentzian Geometry · Tensor Algebra: Contraction, Symmetrization and Antisymmetrization. Leads to. Vector Fields and Integral Curves · Tensor Densities and Weights.

Ask the tutor: Ask me to show why the same $\sqrt{-g}$ that fixes integration also generates half of Einstein's field equations when you vary the action.

Why isn't the plain coordinate volume dⁿx good enough for integration?

Because it is not a geometric object — it is a property of the grid you drew, not of the space itself. Under a change of coordinates it is multiplied by the Jacobian determinant, so $\int f\,\mathrm{d}^n x$ of a scalar $f$ gives different numbers in Cartesian and polar coordinates over the same region. That cannot be a physical volume, since physics does not care which chart you chose. The metric fixes this by supplying $\sqrt{|g|}$ , whose coordinate dependence exactly cancels that of $\mathrm{d}^n x$ , leaving a number every observer agrees on — without it there is no coordinate-independent notion of how much of anything is in a region.

Why does the metric determinant transform with the square of the Jacobian?

Because the metric has two lower indices, and each index contributes one factor of the Jacobian matrix under a coordinate change: $g_{\mu'\nu'}=(\partial x^\mu/\partial x'^{\mu'})(\partial x^\nu/\partial x'^{\nu'})g_{\mu\nu}$ . In matrix language that is a congruence $g'=M^{\mathsf T}gM$ . Taking the determinant and using $\det(M^{\mathsf T})=\det M$ together with multiplicativity gives $\det g' = (\det M)^2\det g = J^2 g$ . The two lower indices are the whole reason it is a square rather than a single power — a one-index density would carry just $J$ . This is exactly why the square root, carrying a single $|J|$ , is the right partner for the coordinate volume.

What is the difference between the Levi-Civita symbol and the Levi-Civita tensor?

The symbol $\tilde\epsilon_{\mu_1\cdots\mu_n}$ is defined by hand to be $+1,-1,0$ in every coordinate system, so it has the same components everywhere and does not transform like a tensor at all — it is a density of weight $+1$ . The tensor $\epsilon_{\mu_1\cdots\mu_n}=\sqrt{|g|}\,\tilde\epsilon_{\mu_1\cdots\mu_n}$ is the symbol dressed with the measure factor, and it does transform correctly, so it is the object you may contract with other tensors to build invariants. Physically the tensor measures oriented volume; the symbol is just its numerical shell. Confusing the two silently drops factors of $\sqrt{|g|}$ and is a leading source of sign and normalization errors.

Where does the √(−g) in the Einstein-Hilbert action actually come from?

It is forced by the demand that the action be a single number independent of coordinates. The action integrates the scalar $R$ over spacetime, and the only invariant measure available is $\sqrt{-g}\,\mathrm{d}^4x$ ; using the bare $\mathrm{d}^4x$ would make $S$ chart-dependent and the resulting field equations meaningless. The Lorentzian sign convention makes $g<0$ , so the absolute value shows up as $\sqrt{-g}$ . Remarkably, the measure is not just along for the ride: when you vary the action, $\delta\sqrt{-g}=-\tfrac12\sqrt{-g}\,g_{\mu\nu}\delta g^{\mu\nu}$ supplies precisely the $-\tfrac12 R\,g_{\mu\nu}$ term that completes the Einstein tensor. The volume element is half the field equations.

How does √g manage to encode the Christoffel symbols in the divergence formula?

Through the contracted Christoffel symbol $\Gamma^\mu{}_{\mu\nu}$ , which by Jacobi's identity equals $\partial_\nu\ln\sqrt{|g|}$ — a pure derivative of the measure. So when you write out $\nabla_\mu V^\mu=\partial_\mu V^\mu+\Gamma^\mu{}_{\mu\nu}V^\nu$ , the connection piece is exactly $\tfrac{V^\mu}{\sqrt{|g|}}\partial_\mu\sqrt{|g|}$ , and the product rule folds everything into $\tfrac1{\sqrt{|g|}}\partial_\mu(\sqrt{|g|}V^\mu)$ . Worked example 3 shows this concretely: differentiating $\sqrt{g}=r^2\sin\theta$ alone reproduces the entire spherical divergence, $1/r^2$ and $1/\sin\theta$ factors included, with no Christoffel symbol ever written down.

Does √g affect whether a quantity is conserved, or only how we compute it?

Both, and they are the same statement. A current is conserved when $\nabla_\mu J^\mu=0$ , which by the divergence formula is identical to $\partial_\mu(\sqrt{|g|}J^\mu)=0$ — an ordinary coordinate divergence of the densitized current $\sqrt{|g|}J^\mu$ . Integrating $J^0$ against $\sqrt{|g|}\,\mathrm{d}^3x$ then gives a charge $Q$ that is both coordinate-independent and constant in time. Drop the $\sqrt{|g|}$ and the divergence theorem fails in curvilinear settings, and total charge becomes observer-dependent nonsense. The measure is not a convenience layered on top of conservation — it is what makes conservation a coordinate-independent fact at all.