The Hodge Star Operator

The exterior algebra of differential forms is the part of geometry that survives with no ruler in the room. You can wedge forms, differentiate them with $\mathrm{d}$ , pull them back, and integrate them over surfaces — all without ever knowing a single length or angle. De Rham cohomology, $F=\mathrm{d}A$ , Stokes' theorem: pure topology and smooth structure, blind to the metric. But physics eventually needs the metric back. Maxwell's action weighs the field against itself; the Laplacian needs a notion of "size" of a form; a variational principle needs a genuine scalar to extremize. The Hodge star $\star$ is the single device that smuggles the metric into this metric-free world, and it does so with surgical economy. Concretely $\star$ is a linear isomorphism $\Omega^p(M)\to\Omega^{n-p}(M)$ , sending a $p$ -form on an $n$ -manifold to a form of complementary degree — legal because $\binom{n}{p}=\binom{n}{n-p}$ , so the two spaces have identical dimension. It uses the metric exactly twice: once to raise the indices of the form it acts on, and once through the volume element $\sqrt{|g|}$ that promotes the numerical Levi-Civita symbol into an honest tensor. The upstream articles own those two ingredients — the metric that raises indices from aq-gr-raising-lowering-indices, and the density $\sqrt{|g|}$ from aq-gr-tensor-densities-weights — and the star fuses them into one operator whose defining property is that $\alpha\wedge\star\beta$ is the inner product of $\alpha$ and $\beta$ times the volume form. From that one line flow the codifferential $\delta$ adjoint to $\mathrm{d}$ , the Laplace–de Rham operator, the whole of Hodge theory, and the cleanest statement of classical electromagnetism ever written: $\mathrm{d}F=0$ , $\mathrm{d}\star F=\star J$ .

Key takeaways

The Hodge star $\star:\Omega^p(M)\to\Omega^{n-p}(M)$ is a linear isomorphism on an $n$ -manifold carrying a metric and an orientation; it depends on both and is undefined without either.
On components it raises the indices of a $p$ -form with $g^{ab}$ and contracts them against the Levi-Civita tensor $\epsilon_{a_1\cdots a_n}=\sqrt{|g|}\,\tilde\epsilon_{a_1\cdots a_n}$ , landing on the complementary degree $n-p$ .
Applied twice it is the identity up to a sign: $\star\star\omega=s\,(-1)^{p(n-p)}\,\omega$ , where $s=\mathrm{sgn}(\det g)$ is $+1$ for Riemannian and $-1$ for Lorentzian metrics.
Its defining property $\alpha\wedge\star\beta=\langle\alpha,\beta\rangle_g\,\mathrm{vol}$ manufactures the $L^2$ inner product on forms, hence the codifferential $\delta$ adjoint to $\mathrm{d}$ and the Laplace–de Rham operator $\Delta=\mathrm{d}\delta+\delta\mathrm{d}$ .
On 2-forms in 4D Lorentzian spacetime $\star\star=-1$ , an almost-complex structure that encodes electromagnetic duality $F\leftrightarrow\star F$ and the self-dual / anti-self-dual splitting of the Weyl tensor.
Key idea: the Hodge star is the metric- and orientation-dependent isomorphism $\star:\Omega^p\to\Omega^{n-p}$ fixed by $\alpha\wedge\star\beta=\langle\alpha,\beta\rangle_g\,\mathrm{vol}$ ; it turns wedge complementarity into an inner product and thereby builds the adjoint $\mathrm{d}^\dagger=\delta$ .

The core insight

The Hodge star is the metric- and orientation-dependent isomorphism $\star:\Omega^p\to\Omega^{n-p}$ characterized by $\alpha\wedge\star\beta=\langle\alpha,\beta\rangle_g\,\mathrm{vol}$ ; it converts wedge complementarity into an inner product and thereby manufactures the adjoint $\mathrm{d}^\dagger=\delta$ .

Where this lives today

The star operator has quietly become a computational technology. Discrete exterior calculus and finite element exterior calculus rebuild $\mathrm{d}$ , $\wedge$ , and a discrete Hodge star on a simplicial mesh, and the resulting mimetic schemes are now standard in computational electromagnetics, fluid solvers, and numerical relativity, where preserving $\mathrm{d}^2=0$ and $\star$ at the discrete level is what keeps constraints like $\nabla\cdot\mathbf{B}=0$ satisfied to machine precision. On the theoretical side, the self-duality equation $\star F=\pm F$ — a first-order condition that automatically extremizes the Yang–Mills action — is the seed of instanton physics, and the moduli spaces of its solutions gave Donaldson his invariants distinguishing exotic four-manifolds, later streamlined by the Seiberg–Witten equations. In gravitation the same star splits the Weyl tensor into self-dual and anti-self-dual parts; the resulting Weyl scalars of the Newman–Penrose formalism are exactly the quantities LIGO and Virgo extract from a binary black-hole merger, with the outgoing radiation encoded in $\Psi_4$ . Twistor theory pushes the idea furthest, trading the Lorentzian $\star\star=-1$ for a complex structure and reformulating field equations as holomorphic geometry.

The physical picture

The one-sentence intuition is: $\star$ means "point in the orthogonal complement, sized by the metric." Feed it a direction — a 1-form that singles out one axis — and it hands back the entire hyperplane perpendicular to that axis, with its area scaled so the total is the volume element. In three-dimensional Euclidean space this is a familiar face wearing a mask: $\star\,\mathrm{d}x=\mathrm{d}y\wedge\mathrm{d}z$ is precisely the "area normal to $x$ ", which is why the cross product and the curl of ordinary vector calculus are secretly Hodge stars in disguise. The complement is the whole content of the operator: a $p$ -form and its $(n-p)$ -form dual carry the same geometric information, read from opposite sides.

Figure

⋆ sends a direction to its orthogonal complement, sized by the metric. In 3D Euclidean space the 1-form dx maps to the perpendicular area element dy ∧ dz, and the constant 1 maps to the full volume, with ⋆⋆ = +1 on every degree

Two features of the star are the whole story of this article, and both are visible the moment you write it down. The first is that the star is a reflection of the degree ladder: it swaps $\Omega^p$ with $\Omega^{n-p}$ , folding the ladder about its middle. The second is that composing it with itself is nearly the identity — $\star\star=\pm1$ — so the star behaves either like a mirror (eigenvalues $\pm1$ ) or like the complex unit $i$ (eigenvalues $\pm i$ ), depending on the dimension, the degree, and the signature. In four-dimensional spacetime the middle rung $\Omega^2$ is mapped to itself, and there $\star\star=-1$ : the star acts like $i$ , splitting the six-dimensional space of 2-forms into self-dual and anti-self-dual halves exactly as a complex structure splits a real vector space. This single fact powers electromagnetic duality, the Weyl-tensor decomposition, and the conformal invariance of light.

Figure

In 4D spacetime ⋆ reflects the degree ladder about Ω², whose 6 two-forms split into self-dual and anti-self-dual halves. Because ⋆⋆ = s(−1)^p(n−p) = −1 on Lorentzian 2-forms, the eigenvalues are ±i, so Ω² = Λ⁺ ⊕ Λ⁻ with 3 dimensions each

The core equations

Everything begins with a genuine tensor built from the numerical Levi-Civita symbol. The symbol $\tilde\epsilon$ is just $\pm1$ or $0$ and transforms as a density, not a tensor; multiplying by $\sqrt{|g|}$ produces the Levi-Civita tensor, the object with which the star is built. This is the one place the volume element enters.

\epsilon_{a_1\cdots a_n}=\sqrt{|g|}\;\tilde\epsilon_{a_1\cdots a_n} \qquad (1)

where: $\epsilon_{a_1\cdots a_n}$ — the Levi-Civita tensor, totally antisymmetric and transforming as a genuine tensor; $\tilde\epsilon_{a_1\cdots a_n}$ — the Levi-Civita symbol, $\pm1$ or $0$ , a density of weight $+1$ ; $g=\det(g_{ab})$ in the chart, $|g|$ its absolute value; $\sqrt{|g|}$ — the factor that converts the symbol into a tensor. Physical reading: the metric determinant is exactly the volume factor a change of variables demands, so $\epsilon$ measures oriented volume covariantly. The star itself raises the indices of the input form and contracts them into this tensor, leaving $n-p$ free lower slots. The combinatorial $1/p!$ corrects for the fact that the antisymmetric sum over $p$ indices counts each independent component $p!$ times.

(\star\omega)_{a_1\cdots a_{n-p}}=\frac{1}{p!}\,\epsilon_{a_1\cdots a_{n-p}}{}^{b_1\cdots b_p}\,\omega_{b_1\cdots b_p} \qquad (2)

where: $\omega_{b_1\cdots b_p}$ — the totally antisymmetric $p$ -form being acted on; $\epsilon_{a\cdots}{}^{b\cdots}$ — the Levi-Civita tensor with $p$ indices raised by $g^{ab}$ (the first appearance of the metric); $\tfrac{1}{p!}$ — combinatorial factor removing the $p!$ -fold overcount. Physical reading: raise-then-contract is the algebraic form of "hand back the orthogonal complement, metric-weighted." The defining, coordinate-free property is the reason the star matters at all. Wedging a form with the star of another produces the metric inner product on $p$ -forms, packaged as a multiple of the volume form. Removing the star returns you to pure topology; inserting it is the act of measuring.

\alpha\wedge\star\beta=\langle\alpha,\beta\rangle\,\mathrm{vol},\qquad \langle\alpha,\beta\rangle=\frac{1}{p!}\,\alpha_{a_1\cdots a_p}\beta^{a_1\cdots a_p} \qquad (3)

where: $\alpha,\beta$ — two $p$ -forms; $\langle\cdot,\cdot\rangle$ — the pointwise metric inner product on $p$ -forms, symmetric in $\alpha$ and $\beta$ ; $\mathrm{vol}=\tfrac{1}{n!}\epsilon_{a_1\cdots a_n}\mathrm{d}x^{a_1}\wedge\cdots\wedge\mathrm{d}x^{a_n}$ — the Riemannian volume form. Physical reading: the star is the unique operator making the wedge pairing coincide with the metric pairing. Composing the star with itself is nearly the identity — the deviation is a pure sign fixed by three integers. This is the single most error-prone formula in the subject, because the signature factor $s$ is easy to drop.

\star\star\,\omega = s\,(-1)^{p(n-p)}\,\omega,\qquad s=\mathrm{sgn}\,(\det g) \qquad (4)

where: $s$ — signature sign, $+1$ for Riemannian, $-1$ for Lorentzian metrics with one timelike direction; $p$ — degree of $\omega$ ; $n$ — dimension of the manifold. Physical reading: on 4D Lorentzian 2-forms $s(-1)^{p(n-p)}=(-1)(+1)=-1$ , so $\star$ squares to $-1$ and acts as a complex structure. At the two ends of the ladder the star reduces to the volume form. It sends the constant function $1$ to $\mathrm{vol}$ and sends $\mathrm{vol}$ back to $s$ , the cleanest instance of the double-dual formula.

\mathrm{vol}=\sqrt{|g|}\;\mathrm{d}x^1\wedge\cdots\wedge\mathrm{d}x^n=\star 1,\qquad \star\,\mathrm{vol}=s \qquad (5)

where: $\mathrm{d}x^i$ — coordinate basis 1-forms; $\star 1$ — the Hodge dual of the 0-form $1$ . Physical reading: the volume element is literally $\star 1$ , so the star is what "knows" how big a region is. The star's payoff for analysis is the codifferential, the formal adjoint of $\mathrm{d}$ with respect to the $L^2$ inner product. Combining $\mathrm{d}$ (which raises degree) with $\delta$ (which lowers it) gives the Laplace–de Rham operator that acts on forms of every degree.

\delta = (-1)^{n(p+1)+1}\,s\;\star\,\mathrm{d}\,\star,\qquad \Delta=\mathrm{d}\delta+\delta\mathrm{d} \qquad (6)

where: $\delta:\Omega^p\to\Omega^{p-1}$ — the codifferential, formal adjoint of $\mathrm{d}$ ; the prefactor carries signature- and degree-dependent signs; $\Delta$ — the Hodge Laplacian, the unique second-order operator preserving degree and self-adjoint under $\langle\cdot,\cdot\rangle_{L^2}$ . Physical reading: $\delta=\mathrm{d}^\dagger$ only exists because the star supplies an inner product to be adjoint with respect to. Cast in this language, classical electromagnetism collapses to two lines. The homogeneous pair is pure topology (no metric, no star); the inhomogeneous pair is where the star — and hence the spacetime geometry — enters.

\mathrm{d}F=0,\qquad \mathrm{d}\star F=\star J \qquad (7)

where: $F=\tfrac12 F_{ab}\,\mathrm{d}x^a\wedge\mathrm{d}x^b$ — the electromagnetic field 2-form; $\star F$ — its Hodge dual, swapping the electric and magnetic roles; $J$ — the 4-current 1-form, with $\mathrm{d}\star J=0$ encoding charge conservation. Physical reading: only the second equation carries the metric, which is why gravity couples to light exclusively through $\star$ .

Derivation AThe Hodge dual of a basis two-form in Minkowski spacetime

The goal is to compute $\star(\mathrm{d}t\wedge\mathrm{d}x)$ explicitly in flat spacetime so that every sign in the abstract definition becomes a concrete arithmetic step, and in particular so that the crucial minus sign contributed by the single timelike direction is exposed rather than asserted. We fix coordinates $(t,x,y,z)$ , the metric $\eta=\mathrm{diag}(-1,1,1,1)$ , and the orientation $\tilde\epsilon_{txyz}=+1$ , so that $\sqrt{|g|}=1$ and the Levi-Civita tensor equals the symbol. The target is the standard result $\star(\mathrm{d}t\wedge\mathrm{d}x)=-\,\mathrm{d}y\wedge\mathrm{d}z$ , and getting the minus honestly is the whole point.

Step 1. Read off the components of the input. Writing $\omega=\mathrm{d}t\wedge\mathrm{d}x$ , the only independent nonzero component is $\omega_{tx}=-\omega_{xt}=1$ ; all others vanish. Why: the wedge of two basis 1-forms is by definition the antisymmetrized tensor product, so it has a single independent slot.

Step 2. Raise both indices with the inverse metric. Since $\eta$ is diagonal, $\omega^{tx}=\eta^{tt}\eta^{xx}\omega_{tx}=(-1)(+1)(1)=-1$ . Why: the lone timelike factor $\eta^{tt}=-1$ is exactly where the Lorentzian signature injects the sign that Euclidean space would not have — this is the arithmetic origin of $s=-1$ .

Step 3. Apply the component definition, keeping the free indices on the complementary pair $\{y,z\}$ . Then $(\star\omega)_{yz}=\tfrac{1}{2!}\,\epsilon_{yz}{}^{ab}\,\omega_{ab}=\tfrac12\,\epsilon_{yzab}\,\omega^{ab}$ , and only $a,b\in\{t,x\}$ survive. Why: the star contracts the raised indices of $\omega$ against $\epsilon$ , and $\epsilon_{yzab}$ forces the summed indices into the two slots $t,x$ not already used by $y,z$ .

Step 4. Evaluate the sum. The two orderings $(a,b)=(t,x)$ and $(x,t)$ contribute equally and cancel the $\tfrac12$ : $(\star\omega)_{yz}=\epsilon_{yztx}\,\omega^{tx}=(+1)(-1)=-1$ , using $\tilde\epsilon_{yztx}=\tilde\epsilon_{txyz}=+1$ because $(t,x,y,z)\to(y,z,t,x)$ is an even permutation. Why: the permutation parity of the Levi-Civita symbol and the raised-index sign multiply, and here they compound to a single minus.

Step 5. Reassemble the 2-form. The component $(\star\omega)_{yz}=-1$ means $\star(\mathrm{d}t\wedge\mathrm{d}x)=-\,\mathrm{d}y\wedge\mathrm{d}z$ . Dualizing once more returns $-\mathrm{d}t\wedge\mathrm{d}x$ , consistent with $\star\star=-1$ on 4D Lorentzian 2-forms. Why: the round trip is the double-dual formula in action, and its $-1$ is the same signature sign that appeared in Step 2.

Go deeper · Derivation B · 1 minThe double-dual sign from the epsilon-epsilon contraction

Derivation BThe double-dual sign from the epsilon-epsilon contraction

The goal is to prove the master formula $\star\star\omega=s\,(-1)^{p(n-p)}\omega$ for an arbitrary $p$ -form, isolating where each factor comes from: the signature sign $s$ from raising a block of contracted Levi-Civita indices through the metric, and the reordering sign $(-1)^{p(n-p)}$ from sliding one block of indices past another. This is the identity every downstream sign convention rests on, so deriving it once — rather than quoting a textbook whose signature or orientation may differ — is what lets you trust your own computations across MTW, Wald, and Nakahara.

Step 1. Compose the component definition with itself. Applying $\star$ twice gives $(\star\star\omega)_{c_1\cdots c_p}=\tfrac{1}{p!\,(n-p)!}\,\epsilon_{c_1\cdots c_p}{}^{a_1\cdots a_{n-p}}\,\epsilon_{a_1\cdots a_{n-p}}{}^{b_1\cdots b_p}\,\omega_{b_1\cdots b_p}$ . Why: the first star produces an $(n-p)$ -form with free indices $a$ , and the second star contracts those against a fresh $\epsilon$ , raising $n-p$ then $p$ indices in turn.

Step 2. Contract the two Levi-Civita tensors. The identity $\epsilon^{a_1\cdots a_{n-p}}{}_{c_1\cdots c_p}\,\epsilon_{a_1\cdots a_{n-p}}{}^{b_1\cdots b_p}=s\,(n-p)!\,p!\,\delta^{[b_1}_{c_1}\cdots\delta^{b_p]}_{c_p}$ holds, where the antisymmetrization is over the upper indices. Why: contracting $n-p$ index pairs leaves a generalized Kronecker delta on the remaining $p$ slots, and the factor $s=\mathrm{sgn}(\det g)$ appears because raising the contracted indices pulls down one factor of $\det g^{ab}$ whose sign is $s$ .

Step 3. Restore canonical index order. The contracted $\epsilon$ carries its $n-p$ summed indices before the block of $c$ 's in Step 1 but after them in the standard identity of Step 2; sliding the block of $p$ lower indices past the block of $n-p$ summed indices costs $(-1)^{p(n-p)}$ . Why: each of the $p$ indices must hop over each of the $n-p$ indices exactly once, and every transposition of the totally antisymmetric symbol contributes a minus.

Step 4. Collect the factors. The $p!\,(n-p)!$ from the contraction cancels the $1/(p!\,(n-p)!)$ prefactor, the antisymmetrized deltas reproduce $\omega_{c_1\cdots c_p}$ , and the surviving signs multiply to $s\,(-1)^{p(n-p)}$ . Why: everything combinatorial cancels, leaving only the two genuine signs — signature and reordering — which is exactly the claimed result.

Go deeper · Derivation C · 1 minThe codifferential as the metric adjoint of the exterior derivative

Derivation CThe codifferential as the metric adjoint of the exterior derivative

The goal is to show that the star manufactures an operator $\delta$ that is the formal adjoint of $\mathrm{d}$ under the $L^2$ inner product $\langle\alpha,\beta\rangle_{L^2}=\int_M\alpha\wedge\star\beta$ , so that $\langle\mathrm{d}\alpha,\beta\rangle_{L^2}=\langle\alpha,\delta\beta\rangle_{L^2}$ for $\alpha\in\Omega^{p-1}$ and $\beta\in\Omega^p$ on a closed manifold. This is the analytic heart of Hodge theory: it is what makes $\Delta=\mathrm{d}\delta+\delta\mathrm{d}$ self-adjoint and non-negative, and hence what guarantees a unique harmonic representative in every cohomology class.

Step 1. Differentiate the natural pairing. Leibniz for $\mathrm{d}$ on the product of a $(p-1)$ -form and the $(n-p)$ -form $\star\beta$ gives $\mathrm{d}(\alpha\wedge\star\beta)=\mathrm{d}\alpha\wedge\star\beta+(-1)^{p-1}\alpha\wedge\mathrm{d}\star\beta$ . Why: this is the one identity that puts $\mathrm{d}\alpha$ and $\mathrm{d}(\star\beta)$ on the same footing, which is what an integration by parts needs.

Step 2. Integrate and drop the boundary. On a closed (compact, boundaryless) manifold Stokes' theorem gives $\int_M\mathrm{d}(\alpha\wedge\star\beta)=0$ , so $\int_M\mathrm{d}\alpha\wedge\star\beta=-(-1)^{p-1}\int_M\alpha\wedge\mathrm{d}\star\beta$ . Why: the vanishing boundary term is exactly the condition under which $\mathrm{d}$ and its adjoint trade places cleanly; with boundary, extra surface terms survive.

Step 3. Reinsert a star to expose an inner product. Using $\star\star=s(-1)^{p(n-p)}$ one writes $\alpha\wedge\mathrm{d}\star\beta=\pm\,\alpha\wedge\star(\star\,\mathrm{d}\star\beta)=\pm\,\langle\alpha,\star\mathrm{d}\star\beta\rangle\,\mathrm{vol}$ , converting the right-hand wedge into the $L^2$ pairing of $\alpha$ with $\star\mathrm{d}\star\beta$ . Why: the whole point of the star is that $\gamma\wedge\star\eta=\langle\gamma,\eta\rangle\,\mathrm{vol}$ , so a single reinserted star turns a wedge into a measurable overlap.

Step 4. Read off the adjoint. Matching $\langle\mathrm{d}\alpha,\beta\rangle_{L^2}=\langle\alpha,\delta\beta\rangle_{L^2}$ forces $\delta=(-1)^{n(p+1)+1}\,s\,\star\mathrm{d}\star$ , and on a Riemannian manifold ( $s=1$ ) the assembled signs give the standard codifferential. Why: the definition of $\delta$ is nothing more than "whatever operator makes $\mathrm{d}$ self-adjoint," and the star is what builds it.

Limiting cases

The star's behaviour is dictated entirely by the three integers $n$ , $p$ , and the signature sign $s=\mathrm{sgn}(\det g)$ . The following cases exhaust the qualitatively distinct regimes.

Case	Result	Meaning
Riemannian, $p(n-p)$ even	$\star\star=+1$	$s=+1$ and even exponent make $\star$ a genuine involution; its eigenspaces are the real $\pm1$ subspaces.
Lorentzian 2-forms in 4D	$\star\star=-1$	$s=-1$ , $n=4$ , $p=2$ , $(-1)^{2\cdot2}=+1$ , so $\star\star=-1$ : an almost-complex structure enabling complex self-duality.
Riemannian 2-forms in 4D	$\star\star=+1$	$s=+1$ gives a real splitting $\Lambda^2=\Lambda^+\oplus\Lambda^-$ into self-dual and anti-self-dual 2-forms, central to instanton physics.
3D Euclidean, 1-forms	$\star\,\mathrm{d}x=\mathrm{d}y\wedge\mathrm{d}z$	the star reproduces the cross product and curl of vector calculus; $\star\star=+1$ on every degree.
Top and bottom degree	$\star 1=\mathrm{vol},\ \star\,\mathrm{vol}=s$	the star and its inverse move between scalars and volumes; the round trip picks up the signature sign.

Common trapThe double dual is always +1, so the Hodge star is its own inverse

This is false for exactly the case general relativity cares about most. The correct statement is $\star\star=s\,(-1)^{p(n-p)}$ , and the signature factor $s=\mathrm{sgn}(\det g)$ equals $-1$ for any Lorentzian metric. On 2-forms in four-dimensional spacetime the exponent gives $(-1)^{2\cdot2}=+1$ , so $\star\star=-1$ , not $+1$ — the star squares to minus the identity and behaves like the complex unit $i$ , which is precisely what makes electromagnetic duality and the self-dual / anti-self-dual decomposition possible. Assuming the Euclidean answer $+1$ silently flips the sign of every duality relation. A second, closely related trap is the codifferential prefactor $(-1)^{n(p+1)+1}s$ : it too depends on $n$ , $p$ , and $s$ , and textbooks (MTW, Wald, Nakahara, Frankel) fix orientation and signature differently, so mixing their formulas produces sign errors that are invisible until a physical result comes out backwards. The cure is mechanical: fix $n$ , $p$ , and $s$ explicitly before writing a single star.

Worked examples

1 · Electromagnetic duality in flat spacetime (fully numeric)

In Minkowski space with $\eta=\mathrm{diag}(-1,1,1,1)$ and $\tilde\epsilon_{txyz}=+1$ , take a field 2-form with electric field of magnitude $E=3$ along $x$ and magnetic field of magnitude $B=2$ along $z$ , so that $F=3\,\mathrm{d}t\wedge\mathrm{d}x+2\,\mathrm{d}x\wedge\mathrm{d}y$ . Compute $\star F$ , verify $\star\star F=-F$ , and evaluate the Lagrangian density from $F\wedge\star F$ .

Step 1. Dualize each basis 2-form. From the first derivation, $\star(\mathrm{d}t\wedge\mathrm{d}x)=-\mathrm{d}y\wedge\mathrm{d}z$ ; an identical component calculation gives $\star(\mathrm{d}x\wedge\mathrm{d}y)=+\mathrm{d}t\wedge\mathrm{d}z$ (raise the two spacelike indices, no timelike sign, and $\tilde\epsilon_{tzxy}=+1$ ). Linearity then gives $\star F=3(-\mathrm{d}y\wedge\mathrm{d}z)+2(\mathrm{d}t\wedge\mathrm{d}z)=\mathbf{-3\,\mathrm{d}y\wedge\mathrm{d}z+2\,\mathrm{d}t\wedge\mathrm{d}z}$ .

Step 2. Interpret the dual physically. The new "electric" (time-space) part is $2\,\mathrm{d}t\wedge\mathrm{d}z$ , an electric field of magnitude $2$ along $z$ ; the new "magnetic" (space-space) part $-3\,\mathrm{d}y\wedge\mathrm{d}z$ is a magnetic field of magnitude $-3$ along $x$ . So $(\mathbf{E},\mathbf{B})=((3,0,0),(0,0,2))$ maps to $(\mathbf{E}',\mathbf{B}')=((0,0,2),(-3,0,0))$ — the duality rotation $\mathbf{E}\to\mathbf{B},\ \mathbf{B}\to-\mathbf{E}$ .

Step 3. Verify the double dual. Dualizing again with $\star(\mathrm{d}y\wedge\mathrm{d}z)=+\mathrm{d}t\wedge\mathrm{d}x$ and $\star(\mathrm{d}t\wedge\mathrm{d}z)=-\mathrm{d}x\wedge\mathrm{d}y$ gives $\star\star F=-3(\mathrm{d}t\wedge\mathrm{d}x)-2(\mathrm{d}x\wedge\mathrm{d}y)=-F$ . This confirms $\star\star=-1$ on 4D Lorentzian 2-forms with explicit numbers, no formula assumed.

Step 4. Evaluate the action density. Only $\mathrm{d}t\wedge\mathrm{d}x\wedge\mathrm{d}y\wedge\mathrm{d}z=\mathrm{vol}$ survives the wedge, giving $F\wedge\star F=(B^2-E^2)\,\mathrm{vol}=(4-9)\,\mathrm{vol}=-5\,\mathrm{vol}$ . Hence the Maxwell Lagrangian density is $-\tfrac12 F\wedge\star F=\tfrac12(E^2-B^2)\,\mathrm{vol}=\mathbf{+2.5\,\mathrm{vol}}$ , matching $-\tfrac14 F_{ab}F^{ab}=\tfrac12(E^2-B^2)$ exactly.

2 · The star in curved coordinates: the sphere's area from $\star\,\mathrm{d}r$ (fully numeric)

Work in flat three-dimensional space in spherical coordinates $(r,\theta,\phi)$ , where $g=\mathrm{diag}(1,r^2,r^2\sin^2\theta)$ so that $\sqrt{|g|}=r^2\sin\theta$ . Compute $\star 1$ and $\star\,\mathrm{d}r$ , and evaluate the latter on the sphere $r=2$ .

Step 1. Compute the volume form. By equation (5), $\star 1=\mathrm{vol}=\sqrt{|g|}\,\mathrm{d}r\wedge\mathrm{d}\theta\wedge\mathrm{d}\phi=\mathbf{r^2\sin\theta\,\mathrm{d}r\wedge\mathrm{d}\theta\wedge\mathrm{d}\phi}$ — the familiar spherical volume element, produced automatically by the star acting on the constant $1$ .

Step 2. Dualize the radial 1-form. For a 1-form the star raises the single index and contracts: $\star\,\mathrm{d}r=\sqrt{|g|}\,g^{rr}\,\mathrm{d}\theta\wedge\mathrm{d}\phi$ . With $g^{rr}=1$ and $\sqrt{|g|}=r^2\sin\theta$ , this is $\star\,\mathrm{d}r=\mathbf{r^2\sin\theta\,\mathrm{d}\theta\wedge\mathrm{d}\phi}$ .

Step 3. Recognize the result. $r^2\sin\theta\,\mathrm{d}\theta\wedge\mathrm{d}\phi$ is exactly the area 2-form on the sphere of radius $r$ : the Hodge dual of "the radial direction" is "the sphere of constant $r$ , sized correctly." This is the orthogonal-complement picture made quantitative — and the metric factor $\sqrt{|g|}=r^2\sin\theta$ is doing all the work.

Step 4. Integrate on $r=2$ . Then $\star\,\mathrm{d}r=4\sin\theta\,\mathrm{d}\theta\wedge\mathrm{d}\phi$ , and $\int_0^{2\pi}\!\!\int_0^{\pi}4\sin\theta\,\mathrm{d}\theta\,\mathrm{d}\phi=4\cdot 2\pi\cdot 2=\mathbf{16\pi}$ , the correct surface area $4\pi r^2=4\pi(2)^2=16\pi$ . The star turned a coordinate 1-form into a physical area element, and the numbers check.

Physics of the portrait

The star operator is named for the geometer who made it the centerpiece of a theory, but its lineage runs from Grassmann's exterior algebra through Cartan's differential forms to the harmonic integrals that fuse analysis, geometry, and topology.

Origins · 1 minW. V. D. Hodge and how this was discovered

W. V. D. Hodge

1903–1975 · Scottish geometer, founder of the theory of harmonic integrals

William Vallance Douglas Hodge was born in Edinburgh and spent most of his career at Cambridge, where he held the Lowndean Chair of Astronomy and Geometry and served as Master of Pembroke College. His great work, The Theory and Applications of Harmonic Integrals (1941), took a scattered set of ideas — de Rham's cohomology of differential forms, the Laplace operator, Riemann's period relations — and welded them into a single machine. The pivot of that machine is the star operator: by supplying a metric inner product on forms, $\star$ lets one define the codifferential $\delta$ , the Laplacian $\Delta=\mathrm{d}\delta+\delta\mathrm{d}$ , and the notion of a harmonic form $\Delta\omega=0$ . Hodge's decomposition theorem then states that on a compact Riemannian manifold every cohomology class contains exactly one harmonic representative, so the purely analytic kernel of $\Delta$ computes the topological Betti numbers — a bridge between the local and the global that still underlies index theory and much of modern geometry. Hodge's early proofs contained gaps that Hermann Weyl and Kunihiko Kodaira later closed rigorously, but the vision was his. His deepest conjecture, on which classes of cohomology on a projective variety come from algebraic cycles, remains one of the Clay Millennium Problems. Hodge received the Copley Medal of the Royal Society in 1974 and collaborated in his later years with Michael Atiyah, seeding the geometry that would grow into the Atiyah–Singer index theorem.

History

1844

— Grassmann's Ausdehnungslehre founds the exterior algebra of antisymmetric quantities

1899

— Élie Cartan builds the systematic calculus of differential forms and the exterior derivative

1931

— Georges de Rham proves that closed-modulo-exact forms compute the manifold's topology

1941

— Hodge's Theory and Applications of Harmonic Integrals introduces the star operator and harmonic forms

1949

— Kodaira and Weyl complete the rigorous proof of the Hodge decomposition theorem

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §3.5 and §4.3–4.5 (forms and duality) · Wald, General Relativity (1984), Appendix B and §C.2 · Carroll, Spacetime and Geometry (2004), §2.9–2.10 · Nakahara, Geometry, Topology and Physics (2003), §7.9 (Hodge theory) · Frankel, The Geometry of Physics (2011), Ch. 14.

Provenance. Conventions used throughout: signature $(-,+,+,+)$ , orientation $\tilde\epsilon_{txyz}=+1$ , and $\sqrt{|g|}$ a density of weight one. The metric-raising of indices is recalled by name from aq-gr-raising-lowering-indices and the promotion $\epsilon=\sqrt{|g|}\,\tilde\epsilon$ from aq-gr-tensor-densities-weights; the wedge and $\mathrm{d}$ are owned by aq-gr-differential-forms-exterior-algebra and aq-gr-exterior-derivative and are not re-derived here. A verifier should recompute: (1) $\star(\mathrm{d}t\wedge\mathrm{d}x)=-\mathrm{d}y\wedge\mathrm{d}z$ from the component definition (Carroll §2.9, MTW §3.5); (2) the double-dual sign $s(-1)^{p(n-p)}$ via the $\epsilon\epsilon$ contraction identity (Nakahara eq. 7.181, Wald App. B); (3) $\star\star F=-F$ and $F\wedge\star F=(B^2-E^2)\,\mathrm{vol}$ for the numeric field of Example 1; (4) $\star\,\mathrm{d}r=r^2\sin\theta\,\mathrm{d}\theta\wedge\mathrm{d}\phi$ integrating to $16\pi$ on $r=2$ ; (5) $\dim\Omega^p=\binom{n}{p}=\binom{n}{n-p}$ . All signs are convention-dependent but internally consistent and cross-checked against the cited texts.

Prerequisites. The Metric Tensor and the Line Element · Metric Signature and Lorentzian Geometry · Raising and Lowering Indices · Tensor Densities and Weights · Tensor Algebra: Contraction, Symmetrization and Antisymmetrization. Leads to. Tensor Densities and Weights · Metric Signature and Lorentzian Geometry.

Ask the tutor: Ask me to dualize any specific form — a basis 2-form, a Maxwell field, or a 1-form in curved coordinates — and I will walk the raise-contract-check steps with every sign shown.

Why does the Hodge star need an orientation as well as a metric?

The metric fixes all the magnitudes in the construction — it raises indices and supplies $\sqrt{|g|}$ — but it cannot fix an overall sign. The volume form $\mathrm{vol}=\sqrt{|g|}\,\mathrm{d}x^1\wedge\cdots\wedge\mathrm{d}x^n$ and the Levi-Civita tensor $\epsilon$ both carry a sign that only a choice of orientation can pin down. Reverse the orientation and $\epsilon\to-\epsilon$ , so $\star\to-\star$ : every Hodge dual flips sign. This is why a non-orientable manifold has no globally defined star, and why physical results that depend on $\star$ — the sign of a duality rotation, the direction of a flux — must always be quoted alongside an orientation convention.

How is the Hodge star related to the cross product and curl?

They are the same operation in disguise. In three-dimensional Euclidean space the star sends a 1-form to a 2-form, $\star\,\mathrm{d}x=\mathrm{d}y\wedge\mathrm{d}z$ , and using the metric to trade 1-forms for vectors, this is exactly the rule that turns two edge vectors into the area vector perpendicular to them — the cross product. Likewise the curl of a vector field is $\star\,\mathrm{d}$ applied to its associated 1-form: $\mathrm{d}$ builds the 2-form of circulation, and $\star$ turns it back into a vector. The reason these familiar operations exist only in three dimensions is that only there does $\star$ map 1-forms to 1-forms after the metric identification, so the "curl of a vector is a vector."

In what sense does the star make Maxwell's equations "geometric"?

Written as $\mathrm{d}F=0$ and $\mathrm{d}\star F=\star J$ , Maxwell's equations separate cleanly into a topological half and a metric half. The homogeneous pair $\mathrm{d}F=0$ uses no metric at all — it says $F$ is a closed 2-form, equivalent to the absence of magnetic monopoles and Faraday's law, and it holds identically once $F=\mathrm{d}A$ . The inhomogeneous pair carries every appearance of the metric inside the single star on $F$ and on $J$ . This is why gravity couples to electromagnetism exclusively through $\star$ : curving spacetime changes $\star F$ , hence the field's dynamics, but never touches $\mathrm{d}F=0$ . It also makes the source-free symmetry $F\to\star F$ manifest.

Why is source-free Maxwell theory conformally invariant, and how does the star show it?

Under a conformal rescaling $g\to\Omega^2 g$ the Hodge star on $p$ -forms scales as $\star\to\Omega^{\,n-2p}\star$ , because raising each of the $p$ indices brings a factor $\Omega^{-2}$ while $\sqrt{|g|}$ brings $\Omega^{\,n}$ . In four dimensions acting on a 2-form the exponent $n-2p=4-4=0$ , so $\star F$ is left completely unchanged by any conformal rescaling of the metric. Consequently $\mathrm{d}\star F=0$ is preserved: light does not "see" the conformal factor of spacetime, which is the precise sense in which massless electromagnetism is conformally invariant. This special cancellation happens only for middle-degree forms in the critical dimension $n=2p$ .

What is the difference between the exterior derivative d and the codifferential δ?

The exterior derivative $\mathrm{d}$ raises a form's degree by one, is built from partial derivatives alone, and needs no metric — it is the same on any smooth manifold. The codifferential $\delta$ lowers degree by one and is genuinely metric-dependent: $\delta=(-1)^{n(p+1)+1}s\,\star\mathrm{d}\star$ , so it cannot even be written without the star. Their defining relationship is adjointness — $\langle\mathrm{d}\alpha,\beta\rangle_{L^2}=\langle\alpha,\delta\beta\rangle_{L^2}$ on a closed manifold — which makes $\delta$ the "divergence" to $\mathrm{d}$ 's "gradient/curl." Both square to zero ( $\mathrm{d}^2=\delta^2=0$ ), and their symmetric combination $\Delta=\mathrm{d}\delta+\delta\mathrm{d}$ is the Laplacian whose kernel is the harmonic forms.

Does the Hodge star change the number of components of a form?

No — that is exactly why it can be an isomorphism. A $p$ -form on an $n$ -manifold has $\binom{n}{p}$ independent components, and its dual is an $(n-p)$ -form with $\binom{n}{n-p}$ components. The binomial identity $\binom{n}{p}=\binom{n}{n-p}$ guarantees these two numbers are equal, so $\star$ maps one space bijectively onto another of the same dimension. In four-dimensional spacetime, for instance, a 2-form and its dual both have $\binom{4}{2}=6$ components, which is why $\star$ maps the six-dimensional space of 2-forms to itself and can act there like a complex structure. The star never adds or discards information; it repackages it into complementary degree.

Why do textbooks disagree about the signs in the codifferential?

Because three independent conventions all feed into the sign: the metric signature (whether time is $-$ or $+$ ), the orientation (the sign of $\tilde\epsilon_{txyz}$ ), and the ordering choice in the definition of $\star$ itself. MTW, Wald, Carroll, Nakahara, and Frankel do not all make the same three choices, so their formulas for $\delta$ and for $\star\star$ differ by signs that are genuine, not typos. None is "wrong" — each is internally consistent. The practical rule is to never mix formulas across books: fix your signature, orientation, and star convention once, derive $\star\star=s(-1)^{p(n-p)}$ and $\delta$ from them yourself, and use only those. Every worked example here uses $(-,+,+,+)$ with $\tilde\epsilon_{txyz}=+1$ .

The Hodge Star Operator

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · Electromagnetic duality in flat spacetime (fully numeric)

2 · The star in curved coordinates: the sphere's area from $\star\,\mathrm{d}r$ (fully numeric)

Physics of the portrait

Going further

Before this

Where this leads

The Hodge Star Operator

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · Electromagnetic duality in flat spacetime (fully numeric)

2 · The star in curved coordinates: the sphere's area from ⋆ dr\star\,\mathrm{d}r⋆dr (fully numeric)

Physics of the portrait

Going further

Before this

Where this leads

2 · The star in curved coordinates: the sphere's area from $\star\,\mathrm{d}r$ (fully numeric)