Nonminimal Coupling and Effective Matter

A curvature–matter term can be written on either side of Einstein’s equation without changing the equation, provided every term is moved consistently and no singular division is performed. What changes is the interpretation: the effective gravitational coupling becomes field dependent, and derivatives of that field remain in the source. A conformal transformation can give an Einstein-frame description only where the transformation is smooth and invertible. It does not erase the coupling; it redistributes it among the metric, scalar, potential, and any additional matter.

Required background. Classical Fields and Their Actions provides field-action variation; Flat-Background Hilbert Stress-Energy from Metric Variation provides the metric-response definition and improvement issue; Affine Connections and Covariant Derivatives provides curvature and second covariant derivatives.

Helpful background. Events, Intervals, and Minkowski Geometry reviews field and observer normalization; Tensor Fields and Natural Operations reviews tensorial field redefinitions; Continuum Kinematics and Balance Laws supplies effective-source interpretation; Presymplectic Potentials and Currents prepares curvature-dependent variations and boundary terms.

A field-dependent curvature coefficient

Consider the four-dimensional action

$$S=\int d^4x\,\sqrt{-g} \left[ \frac{F(φ)}{16\pi G}R -\frac12(\nablaφ)^2 -V(φ) \right].$$

Here $F$ is dimensionless and $G$ is the fixed reference Newton constant. The bulk metric variation uses

$$\delta\!\left(\sqrt{-g}\,F R\right) =\sqrt{-g} \left[ FG_{\mu\nu} +(g_{\mu\nu}\Box-\nabla_\mu\nabla_\nu)F \right]\delta g^{\mu\nu}$$

up to a boundary variation. A complete Dirichlet variational problem requires the matching curvature boundary term; the displayed identity concerns the bulk equation. Combining it with the canonical scalar tensor

$$T^{(0)}_{\mu\nu} =\nabla_\muφ\nabla_\nuφ -g_{\mu\nu} \left[ \frac12(\nablaφ)^2+V \right]$$

gives

$$\boxed{ F G_{\mu\nu} =8\pi G T^{(0)}_{\mu\nu} +\nabla_\mu\nabla_\nu F -g_{\mu\nu}\Box F.}$$

The scalar variation is equally important:

$$\Boxφ-V'(φ) +\frac{F'(φ)}{16\pi G}R=0.$$

Thus the curvature coefficient changes both metric and scalar equations. Carroll introduces this scalar–tensor structure and its varying effective gravitational strength in Carroll 1997, Chapter 4, pp. 119–120, equation (4.77), PDF.

The ξ coupling and derivative terms

Set

$$F(φ)=1-8\pi G\xiφ^2.$$

The action then contains $-\tfrac12\xi Rφ^2$. The scalar equation becomes

$$\Boxφ-V'(φ)-\xi Rφ=0,$$

and the metric equation can be written

$$\begin{aligned} (1-8\pi G\xiφ^2)G_{\mu\nu} =8\pi G\Big[ T^{(0)}_{\mu\nu} +\xi(g_{\mu\nu}\Box-\nabla_\mu\nabla_\nu)φ^2 \Big]. \end{aligned}$$

Alternatively, treating the curvature coupling as part of the matter action gives the full Hilbert tensor

$$T^{(ξ)}_{\mu\nu} =T^{(0)}_{\mu\nu} +\xi \left( g_{\mu\nu}\Box-\nabla_\mu\nabla_\nu+G_{\mu\nu} \right)φ^2,$$

and the equation $G_{\mu\nu}=8\pi G T^{(ξ)}_{\mu\nu}$. Moving the $\xi G_{\mu\nu}φ^2$ term left recovers the previous form exactly. The second-derivative terms do not disappear. Kontou and Sanders give the same tensor and rearrangement, in conventions translated here to an explicit $G$, in Kontou–Sanders 2020, §2, pp. 6–7 and §2.3, pp. 14–16, equations (7) and (40)–(41), PDF.

Diffeomorphism invariance still implies conservation of the full coupled equation on shell. It does not generally imply that an arbitrarily selected subset of the effective right-hand side is conserved by itself. Reall’s general identity allows a matter Lagrangian containing $Rφ^2$ and makes the on-shell qualification explicit in Reall 2013, §10.3, pp. 120–121, PDF.

Effective coupling and singular surfaces

Where $F\ne0$, division gives an Einstein-like equation with

$$G_{\rm eff}(φ)=\frac{G}{F(φ)}.$$

Here $G_{\rm eff}$ denotes the coefficient multiplying the canonical stress tensor in the divided equation. It is not automatically the Newton constant measured by a force experiment: scalar exchange, the scalar mass, the background value of $F'$, and the matter coupling also enter that experiment. The rewriting is not a return to minimally coupled matter. The source still contains $\nabla_\mu\nabla_\nu F-g_{\mu\nu}\Box F$, and the scalar equation still contains $F'R$. At $F=0$ the divided equation is undefined and the coefficient of the metric’s Einstein operator degenerates. If $F<0$, the spin-2 kinetic term has the opposite sign relative to the chosen reference theory. Any claim that uses $G/F$ must therefore state a region on which $F$ is bounded away from zero and has the intended sign.

The derivative terms also change pointwise energy inequalities. Along a null vector $k^\mu$, the contribution $-\xi k^\mu k^\nu\nabla_\mu\nabla_\nuφ^2$ has no fixed sign. Nonminimally coupled classical scalars can violate the usual pointwise conditions even on shell and even in regimes that do not cross $F=0$; the examples and initial-value qualifications are discussed in Kontou–Sanders 2020, §2.3, pp. 14–16, PDF. Such a violation is not by itself a proof of instability or superluminal propagation.

Frame transformations and their domain

In $D>2$ dimensions, a common Einstein-frame metric is

$$\widetilde g_{\mu\nu} =F^{2/(D-2)}g_{\mu\nu}.$$

For real Lorentzian metrics in the usual branch one assumes $F>0$. In four dimensions, $\widetilde g_{\mu\nu}=Fg_{\mu\nu}$. After discarding the transformed boundary term, the scalar kinetic coefficient becomes

$$K(φ) =\frac1F +\frac{D-1}{8\pi G(D-2)} \left(\frac{F'}F\right)^2,$$

and the potential becomes $\widetilde V=V F^{-D/(D-2)}$. A canonical scalar variable exists only on intervals where the required field redefinition is real and invertible. The transformation fails at $F=0$ and can fail globally even when it works locally.

If other matter was minimally coupled to $g_{\mu\nu}$, it generally becomes φ-dependent after the conformal transformation. Lengths, clock rates, masses, and observer normalizations must be translated with the fields. Jordan and Einstein variables can describe the same local classical solutions where the map is regular, but a formula called “matter stress–energy” in one frame need not be the same tensorial split in the other. Frame language never licenses continuation through a singular transformation.

Exercise: constant field and the forbidden division

Let φ be constant on a region, so derivative terms vanish, and write $F(φ_0)=F_0$. Determine both field equations for $F_0>0$, $F_0<0$, and $F_0=0$.

Solution

The metric and scalar equations reduce to

$$F_0G_{\mu\nu}=-8\pi G V(φ_0)g_{\mu\nu}, \qquad -V'(φ_0)+\frac{F'(φ_0)}{16\pi G}R=0.$$

For $F_0>0$ one may divide the metric equation by $F_0$, with a positive effective coupling. For $F_0<0$ the algebraic division is possible, but the effective coupling has the opposite sign and the gravitational kinetic term is sign-reversed relative to the reference theory. In either case, the scalar condition must also hold; a constant scalar is not a solution merely because its derivatives vanish.

For $F_0=0$ division is impossible. In this canonical scalar model, the metric equation requires $V(φ_0)=0$ and no longer supplies the ordinary Einstein equation for the metric. The scalar equation remains:

$$F'(φ_0)R=16\pi G V'(φ_0).$$

If $F'(φ_0)\ne0$, it can still determine $R$; if $F'(φ_0)=0$, a constant solution requires $V'(φ_0)=0$ and this equation leaves $R$ undetermined. The singular case is therefore not obtained by taking a harmless limit of $G_{\rm eff}=G/F_0$.

References

• Carroll, Sean M. Lecture Notes on General Relativity. NSF-ITP/97-147, arXiv:gr-qc/9712019v1, 1997. arXiv. Open PDF.
• Kontou, Eleni-Alexandra, and Ko Sanders. “Energy Conditions in General Relativity and Quantum Field Theory.” Classical and Quantum Gravity 37 (2020): 193001. DOI. Open PDF.
• Reall, Harvey S. General Relativity. Cambridge Part III lecture notes, 2013. PDF.