Metrics, Volume Forms, Hodge Star, and Laplace Operators

A smooth nondegenerate symmetric metric identifies vectors with covectors and induces pairings on differential forms. The metric alone also defines a canonical volume density. After an orientation is chosen, that density becomes an ordinary volume form, and the wedge product and metric pairing determine the Hodge star. Combining the Hodge star with the metric-independent exterior derivative then gives the codifferential and the Hodge–de Rham Laplacian:

$$\begin{aligned} g &\longrightarrow \bigl( \flat,\sharp, \langle\,\cdot,\cdot\,\rangle_g, \mathrm dV_g \bigr), \\ (g,\text{orientation}) &\longrightarrow \bigl(\operatorname{vol}_g,\star\bigr), \\ (\star,\mathrm d) &\longrightarrow \delta, \qquad (\mathrm d,\delta) \longrightarrow \Delta_{\mathrm H}. \end{aligned}$$

Signature matters at every stage after the metric enters. In four-dimensional Euclidean signature, $\star^2=+1$ on two-forms and $\Delta_{\mathrm H}$ is elliptic and nonnegative under suitable domain hypotheses. With the site’s Lorentzian convention $(+---)$, $\star^2=-1$ on two-forms and $\Delta_{\mathrm H}f=-\Box_gf$ on scalars. The Lorentzian operator is wave-type, not a positive elliptic Laplacian.

These constructions are the local geometric input for kinetic terms, dual field strengths, and Euclidean mode operators in QFT. Connection-based curvature formulas and the developed gauge theory and quantization of $p$-form fields are handed to their later canonical pages.

Required background. Smooth Manifolds, Tangent and Cotangent Bundles, and Tensor Fields supplies tensor fields, nondegeneracy, coordinate transformations, and tangent–cotangent duality; and Differential Forms, Integration, Orientation, and Stokes Theorem supplies wedge products, exterior differentiation, orientation, integration, and the boundary term used below.

A metric creates musical duality

Let $M$ be a smooth $n$-manifold. A pseudo-Riemannian metric is a smooth section

$$g\in\Gamma(T^*M\otimes T^*M)$$

such that each $g_p$ is symmetric and nondegenerate. Write its signature as $(p,q)$, where $q$ is the number of negative directions. The signature is locally constant, so it is constant on every connected component. A Riemannian metric has $q=0$; the site’s four-dimensional Lorentzian convention has signature $(1,3)$, written $(+---)$, and therefore $q=3$.

Nondegeneracy makes the bundle map

$$\begin{aligned} \flat:TM&\longrightarrow T^*M, & X^\flat&=g(X,\mathord\cdot) \end{aligned}$$

an isomorphism. Its inverse is denoted $\sharp$. Equivalently,

$$\alpha^\sharp = g^{-1}(\alpha,\mathord\cdot), \qquad g(\alpha^\sharp,Y)=\alpha(Y).$$

In coordinates these operations lower and raise indices:

$$X_\mu=g_{\mu\nu}X^\nu, \qquad \alpha^\mu=g^{\mu\nu}\alpha_\nu.$$

For the flat metric

$$\eta = \mathrm dt\otimes\mathrm dt -\mathrm dx\otimes\mathrm dx -\mathrm dy\otimes\mathrm dy -\mathrm dz\otimes\mathrm dz,$$

one obtains

$$\begin{aligned} (\partial_t)^\flat&=\mathrm dt, & (\partial_i)^\flat&=-\mathrm dx^i, & (\mathrm dx^i)^\sharp&=-\partial_i. \end{aligned}$$

The inverse metric extends the covector pairing to $k$-forms:

$$\langle\alpha,\beta\rangle_g = \frac{1}{k!} \alpha_{\mu_1\cdots\mu_k} \beta^{\mu_1\cdots\mu_k}.$$

In Riemannian signature this is a positive-definite pointwise inner product. In Lorentzian signature it is only a nondegenerate symmetric pairing. For example, the nonzero one-form $\mathrm dt+\mathrm dx$ is null:

$$\langle \mathrm dt+\mathrm dx, \mathrm dt+\mathrm dx \rangle_\eta =1-1=0.$$

Nondegeneracy is essential. On $\mathbb R^2$, the symmetric tensor $h=\mathrm dx\otimes\mathrm dx$ has rank one. It has no inverse, so it defines neither a $\sharp$ map nor the nondegenerate form pairing needed for an ordinary Hodge star. Lee 2018, Chapter 2 is the structural source for metrics, pseudo-Riemannian signatures, and the induced tensor constructions.

A metric gives a density; orientation gives a volume form

Let $g$ also denote the coordinate matrix $(g_{\mu\nu})$. In any chart the metric determines the positive density

$$\mathrm dV_g = \sqrt{\left|\det(g_{\mu\nu})\right|} \left| \mathrm dx^1\cdots\mathrm dx^n \right|.$$

The absolute-value notation records the density transformation law. An orientation-reversing coordinate change does not reverse the sign of $\mathrm dV_g$, so the density exists even when $M$ is not orientable.

If $M$ is oriented, then every positively oriented chart represents the metric volume form as

$$\operatorname{vol}_g = \sqrt{\left|\det(g_{\mu\nu})\right|} \, \mathrm dx^1\wedge\cdots\wedge\mathrm dx^n.$$

Thus the roles are distinct:

• the metric fixes local size and supplies $\mathrm dV_g$;
• the orientation fixes the sign needed to turn that density into $\operatorname{vol}_g$;
• neither structure by itself performs both jobs.

Reversing the orientation sends $\operatorname{vol}_g\mapsto-\operatorname{vol}_g$ while leaving the metric density unchanged. A Möbius strip, for example, admits Riemannian metrics and their globally defined densities, but it has no global ordinary metric volume form. One can instead let the Hodge operation take values in the orientation line bundle, but that twisted construction is outside the present scope.

The metric also does not choose a time orientation in Lorentzian geometry. Choosing which timelike directions are future-pointing is additional data, separate from both spacetime orientation and metric volume.

The Hodge star converts degree using metric and orientation

On an oriented pseudo-Riemannian $n$-manifold, the pointwise Hodge star is the unique bundle isomorphism

$$\star: \Lambda^kT^*M \longrightarrow \Lambda^{n-k}T^*M.$$

It induces the linear isomorphism on forms

$$\star: \Omega^k(M) \longrightarrow \Omega^{n-k}(M)$$

characterized by

$$\boxed{ \alpha\wedge\star\beta = \langle\alpha,\beta\rangle_g \operatorname{vol}_g } \qquad \text{for } \alpha,\beta\in\Omega^k(M).$$

This definition needs the metric and orientation, but no connection. In particular, it is pointwise and algebraic. Exterior differentiation $\mathrm d$, by contrast, needs neither the metric nor orientation.

Choose an oriented pseudo-orthonormal coframe. Applying $\star$ twice first permutes a $k$-fold basis wedge through its $(n-k)$-fold complement, producing $(-1)^{k(n-k)}$. Multiplying the norms of all negative coframe directions contributes $(-1)^q$. Hence

$$\boxed{ \star^2\big|_{\Omega^k(M)} = (-1)^{k(n-k)+q}\operatorname{id} }.$$

Two useful endpoint checks are

$$\star 1=\operatorname{vol}_g, \qquad \star\operatorname{vol}_g=(-1)^q.$$

Orientation reversal changes both $\operatorname{vol}_g$ and $\star$ by a minus sign. It does not change $\star^2$, and the two occurrences of $\star$ in the codifferential below make $\delta$ and $\Delta_{\mathrm H}$ orientation-independent.

Four-dimensional convention check

Take

$$\eta=\operatorname{diag}(+1,-1,-1,-1), \qquad \operatorname{vol}_\eta = \mathrm dt\wedge\mathrm dx\wedge\mathrm dy\wedge\mathrm dz.$$

This covariant volume form has $(\operatorname{vol}_\eta)_{0123}=+1$. It is not the object denoted by a naked epsilon in the site’s separate convention $\epsilon^{0123}=+1$, $\epsilon_{0123}=-1$. In these coordinates,

$$(\operatorname{vol}_\eta)_{0123} = -\epsilon_{0123}, \qquad (\operatorname{vol}_\eta)^{0123} = -\epsilon^{0123}.$$

An alternating symbol, a tensor or density, a volume multivector, and the covariant volume form obey different transformation and index-lowering rules. The calculation below therefore uses $\operatorname{vol}_\eta$ explicitly.

The invariant definition gives

$$\begin{aligned} \star\mathrm dt &= \mathrm dx\wedge\mathrm dy\wedge\mathrm dz, & \star\mathrm dx &= \mathrm dt\wedge\mathrm dy\wedge\mathrm dz, \\ \star(\mathrm dt\wedge\mathrm dx) &= -\mathrm dy\wedge\mathrm dz, & \star(\mathrm dy\wedge\mathrm dz) &= \mathrm dt\wedge\mathrm dx. \end{aligned}$$

Consequently,

$$\star^2(\mathrm dt\wedge\mathrm dx) = -\mathrm dt\wedge\mathrm dx.$$

Four-dimensional Euclidean–Lorentzian convention comparison

Metric setting	Negative directions q	Hodge star squared on two-forms	Codifferential on forms	Flat scalar Hodge operator
Euclidean	0	+1	δ = −⋆d⋆	−Σᵢ ∂ᵢ²
Lorentzian (+−−−)	3	−1	δ = +⋆d⋆	−□

Real Euclidean two-forms therefore split into the $\star$-eigenspaces with eigenvalues $+1$ and $-1$. In Lorentzian signature, a real two-form satisfying $\star F=\pm F$ must vanish: applying $\star$ once more would imply $-F=F$. After complexification, the possible eigenvalues are instead $+i$ and $-i$. This is a sign and signature check, not a development of self-dual field theory.

Frankel 2012, §§ 14.1a–14.1b and Nakahara 2003, §§ 7.9.1–7.9.3 give complementary physics-facing treatments of metric volume and the Hodge operation.

Translating the cited Lorentzian conventions

Frankel and Nakahara use the mostly-plus Lorentzian convention in the relevant examples, whereas this site uses $(+---)$. In four dimensions, with the same orientation,

$$g_{\text{site}}=-g_{\text{source}}, \qquad \star_{\text{site}}\big|_{\Omega^k} = (-1)^k \star_{\text{source}}\big|_{\Omega^k}.$$

The two-form Hodge star and its square are unchanged, but raising a one-form index changes sign. If the same algebraic definitions of $\delta$ and $\Delta_{\mathrm H}$ are compared on the two metrics, then

$$\begin{aligned} \delta_{\text{site}} &= -\delta_{\text{source}}, & \Delta_{\mathrm H,\text{site}} &= -\Delta_{\mathrm H,\text{source}}, \\ \Box_{\text{site}} &= -\Box_{\text{source}}. \end{aligned}$$

For that reason, the displayed $(+---)$ basis formulas above were recomputed from $\alpha\wedge\star\beta =\langle\alpha,\beta\rangle_g\operatorname{vol}_g$ rather than copied from either source.

The codifferential lowers degree

Sign conventions for the codifferential vary. This page defines it on $k$-forms by

$$\boxed{ \delta = (-1)^k\star^{-1}\mathrm d\star = (-1)^{n(k+1)+q+1} \star\mathrm d\star }.$$

Thus

$$\delta:\Omega^k(M)\longrightarrow\Omega^{k-1}(M), \qquad \delta^2=0.$$

The second equality follows from the stated formula for $\star^2$, and $\delta^2=0$ ultimately follows from $\mathrm d^2=0$. In four dimensions this convention reduces to $\delta=-\star\mathrm d\star$ in Euclidean signature and $\delta=+\star\mathrm d\star$ in $(+---)$ signature, as shown in the table.

The algebraic definition should be separated from the claim that $\delta$ is a formal adjoint. Let $\alpha\in\Omega^{k-1}(M)$ and $\beta\in\Omega^k(M)$. The graded Leibniz rule and the definition of $\delta$ give the exact boundary identity

$$\begin{aligned} &\int_M \mathrm d\alpha\wedge\star\beta - \int_M \alpha\wedge\star\delta\beta \\ &\qquad= \int_{\partial M} i_{\partial}^* \bigl(\alpha\wedge\star\beta\bigr). \end{aligned}$$

Therefore

$$\int_M \mathrm d\alpha\wedge\star\beta = \int_M \alpha\wedge\star\delta\beta$$

only when the boundary term vanishes. This holds, for example, on a compact manifold without boundary, when the forms have compact support in the interior, or under boundary conditions chosen to kill the pullback. The identity then makes $\delta$ the formal adjoint of $\mathrm d$ for the integrated pairing

$$(u,v)_g = \int_Mu\wedge\star v, \qquad u,v\in\Omega^r(M).$$

In Riemannian signature this pairing is positive on nonzero real forms with appropriate integrability. In Lorentzian signature it is indefinite, so it is not a Hilbert-space inner product and “formal adjoint” does not mean that a self-adjoint operator has already been defined. Operator domains and boundary conditions remain indispensable. Frankel 2012, § 14.1b and Nakahara 2003, § 7.9.4 support this codifferential and adjoint construction, subject to the convention translation stated above.

The same formula is elliptic or wave-like according to signature

The Hodge–de Rham Laplacian is

$$\boxed{ \Delta_{\mathrm H} = \mathrm d\delta+\delta\mathrm d }.$$

On a function, $\delta f=0$, so $\Delta_{\mathrm H}f=\delta\mathrm df$. In local coordinates this is

$$\Delta_{\mathrm H}f = -\frac{1}{ \sqrt{\left|\det(g_{\rho\sigma})\right|} } \partial_\mu \left( \sqrt{\left|\det(g_{\rho\sigma})\right|} \,g^{\mu\nu}\partial_\nu f \right) .$$

Riemannian signature

Suppose $g$ is Riemannian and the forms are compactly supported in the interior, or impose boundary and domain conditions that remove the boundary pairing. Then

$$\begin{aligned} (\omega,\Delta_{\mathrm H}\omega)_g &= (\mathrm d\omega,\mathrm d\omega)_g + (\delta\omega,\delta\omega)_g \\ &= \|\mathrm d\omega\|_g^2 + \|\delta\omega\|_g^2 \geq0. \end{aligned}$$

The principal symbol is $|\xi|_g^2\operatorname{id}$ for every nonzero covector $\xi$, so $\Delta_{\mathrm H}$ is elliptic. The displayed identity establishes nonnegativity and formal symmetry under the stated hypotheses; an actual self-adjoint realization still requires a specified operator domain.

Ellipticity and nonnegativity do not imply invertibility. On every connected closed Riemannian manifold,

$$\Delta_{\mathrm H}1=0,$$

so constants already supply a zero mode. Global Hodge decomposition and the classification of higher-degree zero modes belong to a later topology chapter, not to this definition page.

Lorentzian signature

For a Lorentzian metric define the scalar wave operator by

$$\Box_gf = \frac{1}{ \sqrt{\left|\det(g_{\rho\sigma})\right|} } \partial_\mu \left( \sqrt{\left|\det(g_{\rho\sigma})\right|} \,g^{\mu\nu}\partial_\nu f \right).$$

With this page’s codifferential convention,

$$\boxed{ \Delta_{\mathrm H}f=-\Box_gf }.$$

In flat four-dimensional spacetime with signature $(+---)$,

$$\Box = \partial_t^2-\boldsymbol\nabla^2, \qquad \Delta_{\mathrm H} = -\partial_t^2+\boldsymbol\nabla^2$$

on scalars. This is wave-type behavior. There is no Lorentzian analogue of the Riemannian positivity identity because the integrated form pairing is indefinite, and the operator is not elliptic.

No connection was needed to define $\star$, $\delta$, or $\Delta_{\mathrm H}$. Expressing the form Laplacian through covariant derivatives and comparing it with a connection or rough Laplacian introduces curvature. That comparison belongs to Levi–Civita Connections, Geodesics, and Riemann Curvature. Frankel, § 14.2a, and Nakahara, § 7.9.5, are the principal sources for the Laplace-operator claims in this section.

Controlled QFT bridge: kinetic terms and dual field strengths

On oriented Minkowski spacetime, let the gauge potential be a one-form and its field strength a two-form:

$$A\in\Omega^1(M), \qquad F=\mathrm dA\in\Omega^2(M).$$

Assume that the action integral below converges, and take all variations to have compact support away from any boundary. Equivalently, one may work on a compact spacetime region and impose the same interior-support condition.

The source-free Maxwell action can be written as

$$\begin{aligned} S[A] &= -\frac12\int_MF\wedge\star F \\ &= -\frac14 \int_M F_{\mu\nu}F^{\mu\nu} \operatorname{vol}_\eta. \end{aligned}$$

The factor $1/2$ in

$$F\wedge\star F = \frac12F_{\mu\nu}F^{\mu\nu} \operatorname{vol}_\eta$$

comes from the induced pairing on two-forms. The metric raises the indices, the orientation fixes the volume form, and $\star F$ packages the dual field strength. None of these operations changes the identity

$$\mathrm dF=\mathrm d^2A=0.$$

To see the separate field equation, vary $A_s=A+s\,a$ with $a\in\Omega^1(M)$ compactly supported. Symmetry of the form pairing and the graded Leibniz rule give

$$\begin{aligned} \left. \frac{\mathrm dS[A_s]}{\mathrm ds} \right|_{s=0} &= -\int_M\mathrm da\wedge\star F \\ &= -\int_Ma\wedge\mathrm d\star F. \end{aligned}$$

Hence stationary points satisfy

$$\mathrm d\star F=0.$$

Because $\delta=+\star\mathrm d\star$ on two-forms in four-dimensional $(+---)$ signature and $\star$ is invertible, this is equivalent to

$$\delta F=0.$$

The equation $\mathrm dF=0$ is a geometric identity following from $F=\mathrm dA$; the equation $\mathrm d\star F=0$ is the Euler–Lagrange equation. They do not have the same logical status. This is the bounded physical use developed here. Sources, gauge fixing, global flux sectors, higher-form reducibility, ghosts, and quantization belong to Differential-Form Fields and Reducible Gauge Systems.

Independently, on a compact oriented Riemannian background—or for a real $r$-form $F$ for which the displayed integral converges—define the Euclidean kinetic functional by

$$\begin{aligned} S_E[F] &= \frac12\int_MF\wedge\star F \\ &= \frac{1}{2r!} \int_M \sqrt{\det(g_{\mu\nu})}\, F_{\mu_1\cdots\mu_r} F^{\mu_1\cdots\mu_r} \,\mathrm d^n x. \end{aligned}$$

On a flat $n$-torus, a momentum $k$ in the dual lattice gives the genuine eigenfunction

$$\Delta_{\mathrm H}e^{ik\cdot x} = |k|^2e^{ik\cdot x}.$$

This nonnegative eigenvalue is a signature-dependent Euclidean statement; it must not be transferred unchanged to Lorentzian momentum modes. Frankel 2012, § 14.1c develops Maxwell theory in the language of forms, while Nakahara, § 7.9.5, connects the codifferential and Laplace operator to the same physics-facing framework.

Common pitfalls

Saying that a metric chooses an orientation. A metric gives a positive density. An orientation is separate data, and only their combination gives the ordinary metric volume form used in the Hodge definition.

Calling every integrated pairing a norm. The pointwise and integrated pairings are positive in Riemannian signature but indefinite in Lorentzian signature. The null form $\mathrm dt+\mathrm dx$ is an immediate counterexample.

Omitting the index from the Hodge-square formula. The factor $(-1)^q$ is what changes $\star^2$ on four-dimensional two-forms from $+1$ in Euclidean signature to $-1$ in $(+---)$ signature.

Importing a codifferential or wave-operator sign from another source. Always state the metric signature, orientation, definition of $\delta$, and definition of $\Box_g$ before comparing formulas.

Turning a formal identity into an operator theorem. Stokes’ boundary term must vanish before $\delta$ is a formal adjoint. Self-adjointness, invertibility, and spectral claims additionally require an operator domain and suitable boundary conditions.

Assuming the form Laplacian is componentwise on a curved manifold. Connection-based component formulas and curvature corrections require more structure. They are not consequences of the flat-coordinate expression.

Exercises

These checks review structure, a missing hypothesis, signature signs, and the QFT-facing kinetic term.

Structure check. Which structures are needed for $\mathrm d$, integration of an ordinary top form, the metric density, and the Hodge star?

Structure answer

The exterior derivative needs only the smooth structure. Integrating an ordinary top form requires an orientation and suitable support or convergence. A nondegenerate metric gives the metric density without an orientation. The ordinary-form Hodge star requires both the metric and an orientation, but no connection.

Hypothesis check. What survives if $h=\mathrm dx\otimes\mathrm dx$ is used on $\mathbb R^2$, and what survives if a genuine Riemannian metric is placed on a Möbius strip?

Hypothesis answer

The tensor $h$ is degenerate, so it has no inverse musical map and cannot induce the nondegenerate form pairing required by the Hodge star. A Riemannian metric on the Möbius strip is nondegenerate and gives musical maps and a global metric density, but nonorientability prevents a global ordinary volume form and an ordinary-form Hodge star.

Sign check. With $\operatorname{vol}_\eta =\mathrm dt\wedge\mathrm dx\wedge\mathrm dy\wedge\mathrm dz$, compute $\star^2(\mathrm dt\wedge\mathrm dx)$ in signature $(+---)$.

Sign answer

The defining wedge identity gives

$$\star(\mathrm dt\wedge\mathrm dx) = -\mathrm dy\wedge\mathrm dz, \qquad \star(\mathrm dy\wedge\mathrm dz) = \mathrm dt\wedge\mathrm dx.$$

Therefore

$$\star^2(\mathrm dt\wedge\mathrm dx) = -\mathrm dt\wedge\mathrm dx,$$

in agreement with $(-1)^{2(4-2)+3}=-1$.

QFT transfer check. Why does the flat-torus Euclidean mode $e^{ik\cdot x}$ have nonnegative Hodge-Laplacian eigenvalue, and why is that not a Lorentzian positivity statement?

Transfer answer

For the flat Euclidean metric on the torus, $\Delta_{\mathrm H}=-\sum_i\partial_i^2$, so

$$\Delta_{\mathrm H}e^{ik\cdot x} = |k|^2e^{ik\cdot x}.$$

The Euclidean metric makes $|k|^2\geq0$. In Lorentzian signature, $\Delta_{\mathrm H}=-\Box$ and the momentum contraction is indefinite, so the same positivity conclusion does not follow.

Synthesis and next steps

A nondegenerate metric supplies the musical isomorphisms and induced form pairings. It also supplies a canonical density; an orientation converts that density into the metric volume form. Their combination defines the Hodge star, whose square records both degree and metric index. The codifferential is built from $\star$ and $\mathrm d$, while its formal-adjoint interpretation depends on the Stokes boundary term. Finally, $\Delta_{\mathrm H}=\mathrm d\delta+\delta\mathrm d$ is elliptic and nonnegative in the Riemannian setting under suitable hypotheses, but its Lorentzian scalar specialization is $-\Box_g$.

Continue according to the structure needed:

• Levi–Civita Connections, Geodesics, and Riemann Curvature adds the unique torsion-free metric connection, geodesics, curvature, and connection-based Laplace formulas;
• Differential-Form Fields and Reducible Gauge Systems is the developed physical continuation for $p$-form gauge complexes, reducibility, harmonic and flux sectors, observables, and quantization.

References

• Theodore Frankel, The Geometry of Physics: An Introduction, third edition, Cambridge University Press, 2012, §§ 14.1a–14.1c and 14.2a. This is the physics-facing teaching source for the Hodge star, codifferential, Maxwell forms, and the Laplace operator.
• John M. Lee, Introduction to Riemannian Manifolds, second edition, Graduate Texts in Mathematics 176, Springer, 2018, Chapter 2. This chapter develops Riemannian and pseudo-Riemannian metrics and their induced constructions.
• Mikio Nakahara, Geometry, Topology and Physics, second edition, Institute of Physics Publishing, 2003, §§ 7.1.1 and 7.9.1–7.9.5. This gives an independent physics-facing treatment of metrics, volume, Hodge duality, adjoint exterior differentiation, and Laplace operators.