Minimal Metric Coupling

Minimal coupling is the standard first attempt to place a special-relativistic matter theory on a curved spacetime: replace the Minkowski metric by the spacetime metric, use its invariant volume element, and replace ordinary derivatives by Levi-Civita covariant derivatives. This prescription produces generally covariant equations with the correct local flat-spacetime limit. It does not prove that the resulting theory is unique, because curvature couplings, derivative ordering, boundary terms, and the geometric variables needed for spin can carry information that flat spacetime cannot reveal.

Required background. Classical Fields and Their Actions provides action variation, Euler–Lagrange equations, and gauge symmetry; Flat-Background Hilbert Stress-Energy from Metric Variation provides the metric-response definition of stress–energy; Affine Connections and Covariant Derivatives provides Levi-Civita differentiation and curvature commutators.

Helpful background. Events, Intervals, and Minkowski Geometry reviews Lorentz invariance and causal vectors; Tensor Fields and Natural Operations reviews tensors, pullbacks, and coordinate-independent constructions; Continuum Kinematics and Balance Laws supplies the continuum meaning of local conservation; Presymplectic Potentials and Currents prepares boundary terms and variational ambiguities.

The covariantization prescription

For a real scalar field in Minkowski spacetime, start from

$$S_\eta[φ] =-\int d^4x \left[ \frac12\eta^{\mu\nu}\partial_\muφ\,\partial_\nuφ +V(φ) \right].$$

Minimal metric coupling gives

$$S_g[φ] =-\int d^4x\,\sqrt{-g} \left[ \frac12 g^{\mu\nu}\nabla_\muφ\,\nabla_\nuφ +V(φ) \right],$$

and varying φ yields $\Boxφ-V'(φ)=0$. At any event, local inertial coordinates make $g_{\mu\nu}=\eta_{\mu\nu}$ and $\Gamma^\rho{}_{\mu\nu}=0$, so the equation has its special-relativistic form at that event. Curvature and derivatives of the connection remain detectable over a finite neighborhood. Reall 2013, §5.1, pp. 49–50, PDF states this prescription and its equivalence-principle motivation explicitly.

For electromagnetism the geometrically primary object is the two-form $F=dA$. The minimally coupled vacuum action is

$$S_{\rm EM} =-\frac14\int d^4x\,\sqrt{-g}\, F_{\mu\nu}F^{\mu\nu},$$

with equations $dF=0$ and $\nabla_\mu F^{\mu\nu}=0$. Because antisymmetrization removes the symmetric Levi-Civita connection, $F_{\mu\nu}=2\nabla_{[\mu}A_{\nu]}=2\partial_{[\mu}A_{\nu]}$. Thus the first-order equations acquire no explicit curvature term, although the metric still enters the Hodge dual, index raising, volume element, and wave operator.

These examples illustrate what the prescription guarantees: tensorial equations, a definite metric response, and recovery of the chosen flat theory when $g_{\mu\nu}\to\eta_{\mu\nu}$. They do not establish that every allowed curved-spacetime term has been found.

Flat equivalence does not imply curved equivalence

In flat spacetime, partial derivatives commute. Two flat equations can therefore be identical before covariantization but inequivalent afterward. For a vector $V^\mu$,

$$\partial_\rho\partial_\nu V^\rho =\partial_\nu\partial_\rho V^\rho,$$

whereas the corresponding curved expressions differ by

$$\left(\nabla_\rho\nabla_\nu -\nabla_\nu\nabla_\rho\right)V^\rho =R_{\sigma\nu}V^\sigma.$$

Choosing an ordering in flat space therefore silently chooses a Ricci coupling after the comma-to-covariant-derivative replacement. Carroll uses precisely this kind of example to expose the nonuniqueness in Carroll 1997, Chapter 4, pp. 106–108, PDF.

There is a second ambiguity. Terms such as $\xi Rφ^2$, $R_{\mu\nu}V^\mu V^\nu$, or higher-curvature operators vanish on Minkowski spacetime but are compatible with diffeomorphism covariance. Flat data alone cannot determine their coefficients. In an effective field theory they are generally organized by symmetries, dimensions, and a separation of scales; “minimal” then means a chosen lowest-complexity baseline, not a theorem that the omitted coefficients vanish.

Even integrations by parts require care. Two flat actions that differ by a total derivative can define the same bulk equation under suitable boundary conditions, yet their curved extensions and boundary variations may differ if one covariantizes after discarding the boundary term. One must compare complete actions and variational problems, not just their bulk Euler–Lagrange equations.

Fields that require more than a metric substitution

Spinors do not carry ordinary tensor representations of the coordinate frame group. Coupling a Dirac field requires a tetrad or coframe, a spin structure, gamma matrices, and the spin connection. Varying the tetrad gives the gravitational source before it is translated into a symmetric spacetime tensor. If torsion is allowed, the spin connection is not fixed solely by the metric and spin can source torsion; that theory lies outside the Levi-Civita setting assumed here.

Gauge symmetry can remove some apparent ambiguities. For Maxwell theory, an arbitrary term $R_{\mu\nu}A^\mu A^\nu$ is not invariant under $A_\mu\mapsto A_\mu+\nabla_\mu\chi$, whereas curvature operators built from $F_{\mu\nu}$ can be gauge invariant. Symmetry narrows the possibilities but still does not normally select a unique action. Boundary conditions, field content, derivative order, and the regime of validity must also be stated.

The practical rule is therefore conditional: begin with minimal coupling when the equivalence-principle limit and symmetries make it appropriate, then list all curvature and boundary terms allowed at the accuracy being claimed. The resulting equations must be checked in the flat limit and against their gauge identities. The nonminimal-coupling page develops what changes when a curvature term is retained.

Exercise: detect an ordering ambiguity

Two flat-space expressions are related by commuting derivatives: $A_\nu=\partial_\rho\partial_\nu V^\rho$ and $B_\nu=\partial_\nu\partial_\rho V^\rho$. Covariantize each expression directly and find $A_\nu-B_\nu$. When do the two prescriptions agree?

Solution

Direct replacement gives

$$A_\nu=\nabla_\rho\nabla_\nu V^\rho, \qquad B_\nu=\nabla_\nu\nabla_\rho V^\rho.$$

The inherited curvature convention implies

$$A_\nu-B_\nu =[\nabla_\rho,\nabla_\nu]V^\rho =R^\rho{}_{\sigma\rho\nu}V^\sigma =R_{\sigma\nu}V^\sigma.$$

They agree on a Ricci-flat region, for a vector annihilated by $R_{\sigma\nu}$, or in the flat limit. General covariance alone does not prefer one ordering; the difference is a genuine curvature coupling.

References

• Carroll, Sean M. Lecture Notes on General Relativity. NSF-ITP/97-147, arXiv:gr-qc/9712019v1, 1997. arXiv. Open PDF.
• Reall, Harvey S. General Relativity. Cambridge Part III lecture notes, 2013. PDF.